'W ••

Digitized by the Internet Archive

in 2018 with funding from
Kahle/Austin Foundation

https://archive.org/details/mathematicalanalOOOOwebe
MATHEMATICAL
ANALYSIS
MATHEMATICAL
ANALYSIS
Business and
Economic Applications
THIRD EDITION

JEAN E. WEBER
University of Arizona

HARPER & ROW, PUBLISHERS

New York, Hagerstown, San Francisco, London
Sponsoring Editor : John Greenman
Project Editor: Eleanor Castellano
Designer: Emily Harste
Production Supervisor: Stefania J. Taflinska
Compositor: Santype International Limited
Printer and Binder: Halliday Lithograph Corporation
Art Studio: J 8- R Technical Services Inc.

MATHEMATICAL ANALYSIS: Business and Economic Applications, Third Edition

The first edition of this book was coauthored by Jean E. Weber, formerly Jean E.
Draper, and Jane S. Klingman. Subsequent editions were revised by Jean E. Weber.
Copyright © 1967, 1972 by Jean E. Draper and Jane S. Klingman.
Copyright © 1976 by Jean E. Weber.

All rights reserved. Printed in the United States of America. No part of this
book may be used or reproduced in any manner whatsoever without written permission
except in the case of brief quotations embodied in critical articles and reviews.
For information address Harper & Row, Publishers, Inc., 10 East 53rd Street,
New York, N.Y. 10022.

Library of Congress Cataloging in Publication Data

Weber, Jean E.
Mathematical analysis.
Bibliography: p.
Includes index.
1. Business mathematics. 2. Mathematical
analysis. 3. Economics, Mathematical.
I. Title.
HF5691.W38 1976 510'.2'433 76-2078
ISBN 0-06-041762-5
Contents

Preface to the Third Edition, xi

Preface, xii

INTRODUCTION, 1
Sets, 2
Variables, 6
Relations and Functions, 8
Inverse Functions, 13

1. GRAPHICAL REPRESENTATION, 16
1.1 Introduction, 16
1.2 Rectangular Coordinates, 16
Fundamental Principle, 17
1.3 Straight Lines, 17
Slope of a Straight Line, 19
Equations for Straight Lines, 21
Two-Point Form, 21 Intercept Form, 22 Point-Slope Form, 23
Slope-Intercept Form, 24 Vertical Lines, 24 Horizontal Lines, 25
Parallel, Perpendicular, and Intersecting Lines, 26
Intersection of Two Straight Lines, 27
Independent Equations, 27 Consistent Equations, 27
Families of Lines, 33
1.4 Applications of Straight Lines in Business and Economics, 37
Linear Demand and Supply Curves, 37
Linear Demand Curves, 38 Linear Supply Curves, 40
Market Equilibrium, 43
Break-Even Analysis, 44
The Consumption Function, 46
1.5 General Methods of Graphing Nonlinear Curves, 50
Intercepts, 51
Symmetry, 51
Extent, 56
Asymptotes, 57
Factorization, 60
Real Curve, Point Locus, or Imaginary Locus, 61
Summary: General Methods of Graphing Nonlinear Curves, 62
1.6 Quadratic Curves, 70
Identification of a Quadratic Equation, 72
vi / Contents

The Circle, 72
The Ellipse, 74
The Parabola, 77
The Hyperbola, 81
Special Cases of the Equilateral Hyperbola, 86
Families of Quadratic Curves, 90
1.7 Applications of Quadratic Curves in Business and
Economics, 92
Demand and Supply Curves, 92
Market Equilibrium, 93
Product Transformation Curves, 98
Pareto's Law of Distribution of Income, 107
1.8 Exponential and Logarithmic Curves, 111
Exponential Functions, 111
Logarithmic Functions, 114
1.9 Applications of Exponential and Logarithmic Curves in
Business and Economics, 117
Compound Interest, 118
Growth Functions, 119
Biological Growth Curves, 119 Gompertz Curves, 120 Learning Curves, 121
1.10 Trigonometric Curves, 123
Degree and Radian Measure, 124
Trigonometric Functions, 125
Trigonometric Identities, 129
Polar Coordinates, 129
1.11 Applications of Trigonometric Curves in Business and
Economics, 133
Technical Notes I, II, III, IV, and V 133,133,134,137, and 137

2. DIFFERENTIAL CALCULUS: FUNCTIONS OF ONE VARIABLE, 139

2.1 Introduction, 139
2.2 Limits, 141
Definition of a Limit, 141
Properties of Limits, 146
2.3 Continuity, 151
Types of Discontinuities, 151
Properties of Continuous Functions, 159
2.4 Definition of the First Derivative, 161
Interpretation of the First Derivative, 165
Velocity of a Moving Body, 165 Rate of Change of a Function, 167
2.5 Applications of the First Derivative in Business and
Economics, 168
Cost, 1 68
Revenue, 170
Elasticity, 171
National Income, Consumption, and Savings, 172
2.6 Rules for Differentiation, 174
Algebraic Functions, 174
Rule for Differentiation of Polynomial Functions, 177
Logarithmic Functions, 185
Exponential Functions, 187
Trigonometric Functions, 191
Inverse Functions, 194
Composite Functions, 196
Summary of Procedure for Differentiation, 199
Summary of Rules for Differentiation, 201
Algebraic Functions, 201 Logarithmic Functions, 202 Exponential
Functions, 202 Trigonometric Functions, 202 Inverse Functions, 202
Composite Functions, 202
2.7 Differentials, 203
Approximations, 204
2.8 Higher-Order Derivatives, 207
 Contents / vii

2.9 Implicit Differentiation, 213

Procedure for Implicit Differentiation, 213
2.10 Differentiability and Continuity, 217
Smooth Functions, 218
2.11 Applications of Derivatives, 219
The First Derivative, 219
Increasing and Decreasing Functions, 219 Relative (Local) Maxima and
Minima, 220
Summary of the Information That Can Be Obtained from the First
Derivative, 230
The Second Derivative, 231
Concavity, 231 Points of Inflection, 236
Summary of Information That Can Be Obtained from the Second
Derivative, 240
Summary of the Procedure for Sketching the Graph of y = f(x)
Using Information Provided by First and Second Derivatives, 240
2.12 Applications of Derivatives in Business and Economics, 245
Cost, Average Cost, and Marginal Cost, 245
Elasticity, 256
Constant Elasticity of Demand, 263
Revenue, Marginal Revenue, and Elasticity of Demand, 266
Revenue from Taxation, 271
Profit Under Monopoly, 275
Effect of Taxation on Monopoly, 280
Inventory Models, 285
2.13 Indeterminate Forms, 289
2.14 Sequences and Series, 298
Summary of Procedure for Testing Convergence or Divergence of an
Infinite Series, 300
Power Series, 304
Taylor's Theorem, 308
Technical Notes I, II, and III 317, 321, and 321

3. DIFFERENTIAL CALCULUS: FUNCTIONS

OF MORE THAN ONE VARIABLE, 323
3.1 Introduction, 323
Functions of More Than One Variable, 323
Continuity, 324
3.2 Partial Differentiation, 324
The Total Differential, 330
The Total Derivative, 332
Differentiation of Implicit Functions, 335
3.3 Applications of Partial Derivatives in Business and
Economics, 340
Marginal Cost, 340
Demand Surfaces, 341
Marginal Demand, 341 Partial Elasticities of Demand, 344
Production Functions, 346
Marginal Productivity, 346
Euler's Theorem, 348
Linear Homogeneous Production Functions, 349
Constant Product Curves, 349
Returns to Scale, 351
Utility Functions, 352
3.4 Maxima and Minima of Functions of Two Variables, 356
3.5 Maxima and Minima Subject to Constraints, 367
Lagrange Multipliers, 368
Kuhn-Tucker Conditions, 375
3.6 Applications of Constrained Maximization and Minimization
in Business and Economics, 380
Constrained Utility Maximization, 380
Technical Note I 385
viii / Contents

4. INTEGRAL CALCULUS, 387

4.1 Introduction, 387
4.2 Indefinite Integration, 388
4.3 Applications of Indefinite Integration in Business and
Economics, 394
Cost, 394
Revenue, 395
National Income, Consumption, and Savings, 396
Capital Formation, 397
4.4 Definite Integration, 399
Area as a Definite Integral, 404
Interpretation of Negative Areas, 407 Area Between Two Curves, 408
Probability as an Area, 414
4.5 Applications of Definite Integration in Business and
Economics, 418
Consumers' Surplus, 418
Producers' Surplus, 422
Revenue Versus Cost, 425
4.6 Special Methods of Integration, 431
Standard Forms for Integration, 431
Integration by Parts, 435
Integration by Partial Fractions, 438
Integration by a Rationalizing Substitution, 443
Integration by Miscellaneous Substitutions, 447
4.7 Numerical (Approximate) Methods of Integration, 449
The Trapezoidal Rule, 449
Notes Concerning the Accuracy of the Trapezoidal Rule, 450
Simpson's Rule, 451
Taylor's Expansion, 453
4.8 Multiple Integrals, 459
Probability as a Double Integral, 464

5. DIFFERENTIAL EQUATIONS, 469

5.1 Introduction, 469
5.2 Definition and Classification of Differential Equations, 470
5.3 Solutions of Ordinary Differential Equations, 471
5.4 Differential Equations of First Order and First Degree, 476
Separable Differential Equations, 477
Homogeneous Differential Equations, 480
Exact Differential Equations, 487
Linear Differential Equations, 491
Differential Equations Linear in a Function of y or in a Function of
x, 495
5.5 Differential Equations of Higher Order and/or Higher
Degree, 498
5.6 Applications of Differential Equations in Economic
Models, 502
Domar Macro Model, 502
Domar Debt Model, 503
A Second Domar Debt Model, 504
Evans Price Adjustment Model, 505
Income-Consumption-Investment Model, 506

6. DIFFERENCE EQUATIONS, 509

6.1 Introduction, 509
6.2 Definition and Classification of Difference Equations, 510
Linear Difference Equations, 511
6.3 Solutions of Difference Equations, 512
Analogies Between Difference Equations and Differential Equations, 514
 Contents / ix

6.4 Linear First-Order Difference Equations with Constant

Coefficients, 515
Behavior of the Solution Sequence, 518
Equilibrium and Stability, 522
6.5 Applications of Difference Equations in Economic Models, 523
The Harrod Model, 523
The General Cobweb Model, 524
Consumption Model, 524
Income-Consumption-Investment Model, 525
6.6 Linear Second-Order Difference Equations with Constant
Coefficients, 528
Behavior of the Solution Sequence, 531
Nonhomogeneous Second-Order Difference Equations, 533
Equilibrium and Stability, 535
6.7 Applications of Second-Order Difference Equations in
Economic Models, 537
Samuelson's Interaction Model, 537
Metzler's Inventory Model, 539

7. MATRIX ALGEBRA, 541

7.1 Introduction, 541
7.2 Definition of a Matrix, 541
Definition of a Vector, 542
7.3 Matrix Operations, 544
Addition and Subtraction of Matrices, 544
Multiplication of a Matrix by a Scalar, 545
Multiplication of Matrices, 546
7.4 Special Types of Matrices, 551
Diagonal Matrices, 551
Identity Matrices, 552
Null Matrices, 553
7.5 The Transpose of a Matrix, 554
Transpose of a Sum or Difference of Matrices, 556
Transpose of a Product of Matrices, 556
7.6 Partitioned Matrices, 558
7.7 Determinant of a Matrix, 564
Properties of Determinants, 567
7.8 The Inverse of a Matrix, 569
Inversion of 2 x 2 Matrices, 570
Inversion of Larger Matrices, 573
Inversion Using Row or Column Operations, 574
Steps in Converting a Square Matrix to the Identity Matrix (Gaussian
Elimination), 574
Inversion Using Adjoints and Determinants, 577
Inversion of Partitioned Matrices, 580
Properties of Inverses, 582
7.9 Simultaneous Linear Equations, 585
Linear Dependence and Rank, 585
Solution of Simultaneous Linear Equations, 589
7.10 Vector Differentiation, 593
Vector Differentiation of a Linear Function, 593
Vector Differentiation of a Vector of Functions, 595
Vector Differentiation of a Quadratic Form, 595
Vector Differentiation of a Bilinear Form, 597
Applications of Vector Differentiation in Maximization and
Minimization, 598

8. APPLICATIONS OF MATRIX ALGEBRA, 602

8.1 Introduction, 602
8.2 Maxima and Minima of Functions of n Variables, 602
x / Contents

Unconstrained Maxima and Minima, 602

Lagrange Multipliers, 607
Kuhn-Tucker Conditions, 611
8.3 Input-Output Analysis, 614
8.4 Linear Programming, 619
Geometric Solution, 621
The Simplex Method, 625
The Dual of a Linear Programming Problem, 631
8.5 Game Theory, 639
Number of Persons, 639 The Payoff, 640 Strategies, 640 The Game
Matrix, 640 Saddle Points, 641
Solution of Two-Person, Two-Strategy Games, 642
2x2 Games, 643 Solution of 2 X 2 Games by Matrix Algebra, 646
2 x n Games and m X 2 Games, 650 Dominance, 651 Graphical
Solution, 653
Solution of Larger Games, 656
8.6 First-Order Markov Processes, 661
Definition of a First-Order Markov Process, 661
Steady State or Equilibrium, 662

Algebra Review Problems, 669

Tables, 675
Selected References, 679
Index, 683
Preface to
the Third Edition

The general outline of the previous editions is retained in the third edition. Within
this general framework, there has been considerable revision. Throughout the book
the exposition has, hopefully, been clarified; verbal explanation has been added to
many of the examples.
Some new material has been added. Specifically, additional material concerning
trigonometric functions is included. Capital formation and continuous compounding
are discussed as applications of integration. Several statistical concepts are consid¬
ered; probability is obtained by integration as the area under a curve, and joint
probabilities are obtained by double integration under a surface. The least squares
estimates for bivariate linear regression are derived as an application of maximiza¬
tion and minimization using partial differentiation. Vector derivatives of linear
functions, quadratic forms, and bilinear forms are considered; the application of
vector differentiation in obtaining maxima and minima of functions of several
variables is illustrated by the derivation of the least squares estimates for multi¬
variate linear regression. Problems have also been added. A revised solutions
manual, including complete solutions to all problems, is available to instructors
adopting the text.
Most of the changes, particularly with respect to additional material, were made
in response to suggestions by users and reviewers. These suggestions were very
helpful and are greatly appreciated; even more of them would have been followed
if excessive length were not a consideration.
Every effort has been made, by author and publisher, to avoid errors in the text
and solutions manual. The author accepts responsibility for any errors which have
gone unnoticed.

j. e. w.
Preface

This text offers the student an understanding of quantitative techniques both in their
mathematical contexts and as they are applied to business and economic problems.
The material has been used in various stages of development in graduate and
undergraduate quantitative analysis classes at The University of Wisconsin. It
assumes reasonable facility with algebra; there is, however, an algebra review
section at the end of the book for students whose skills in this area need
sharpening. The text is designed for a one-year course, allowing perhaps for some
omissions at the option of the instructor.
There are eight chapters: Graphical Representation, Differential Calculus: Func¬
tions of One Variable, Differential Calculus: Functions of More Than One
Variable, Integral Calculus, Differential Equations, Difference Equations, Vectors
and Matrices, and Applications of Matrix Algebra. Each chapter includes mathe¬
matical presentation of its subject and applications to business and economics.
Each also includes many problems, for we believe that substantial practice in prob¬
lem solving is essential in the study of mathematics. An extensive solutions
manual providing complete calculations for all problems is available on request to
instructors who adopt the text.
We wish to express our gratitude to the students who have used this material
through its many drafts and whose comments and suggestions have been incor¬
porated in the revisions. We would also like particularly to thank Myrna Edmonds
for her patient and painstaking typing of the manuscript. Finally, we accept full
responsibility for any errors that may remain.

j. E. D.
j. s. K.
Introduction

From the beginning of recorded history, cultural and scientific advances have
depended on the use of symbols. The history of civilization can be viewed as the
history of man’s increasingly sophisticated use of symbols. As thinking in any area
develops, the symbols used become increasingly abstract.
When the concepts to which symbols refer are essentially nonquantitative in
nature, the symbols and their relationships can be studied in the general framework
of logic and there is no need for mathematics as such. It is when symbols represent
essentially quantitative concepts that mathematics is useful and, in fact, essential for
analyzing their relationships. Mathematics is a branch of logic, the branch of logic
that provides a systematic framework within which quantitative relationships can be
studied. In pure mathematics definitions or axioms and assumptions are precisely
stated symbolically, and the analysis proceeds by deduction to obtain conclusions.
Applied mathematics differs from pure mathematics in a very important respect—in
pure mathematics symbols represent abstract concepts whose properties are assigned
by definition, while in applied mathematics many symbols represent variables ob¬
served in the real world; the properties of these variables must be determined by
observation, not by abstract definition, and then stated mathematically. In addition,
the empirical accuracy of the deductions of applied mathematics can be determined.
Applied mathematical analysis thus is based on empirically determined definitions
and assumptions from which empirically verifiable conclusions are obtained by
deduction. Pure and applied mathematical analysis differ only with respect to the
empirical aspect of the definitions and assumptions and conclusions, not with respect
to the methods of deduction.
Because economics is concerned with concepts that are essentially quantitative in
nature—for example, price, cost, wage rates, investment, income, and profit—much
of economic analysis is inescapably mathematical in nature. Mathematics provides a
logical, systematic framework within which quantitative relations can be studied.
When economic variables are represented by symbols and their properties stated
mathematically, mathematics provides the techniques for analyzing relations among
the symbols and thus among the variables they represent. Much of economic analysis
is therefore applied mathematical analysis.
In economic analysis, as in applied mathematics in general, deductions obtained
by mathematical analysis are interpreted and evaluated empirically. It should be
1
2 / Introduction

noted in this regard that if the deductions following from a set of definitions and
assumptions are not correct with respect to empirical observation, mathematical
analysis (if correctly performed) is not responsible and the difficulty is to be found in
the definitions or assumptions. Mathematics enables the economist to be precise in
defining relevant variables, to state clearly the assumptions made, to be logical
in developing the analysis, and to consider a larger number of variables than might
be feasible verbally. However, it does not and cannot prevent the omission or the
empirically incorrect definition of relevant variables nor can it prevent empirically
inaccurate or incomplete statements of assumptions. Mathematical analysis takes
definitions and assumptions as given and obtains the conclusions that follow log¬
ically from them. Mathematical analysis is thus by nature logical, not empirical, and
can be held responsible for conclusions only with respect to their logical validity
given the definitions and assumptions on which they are based and not with respect
to their empirical accuracy.
Thus if mathematical analysis is correctly performed, but its conclusions are
empirically incorrect, the definitions and assumptions must be examined for accur¬
acy and completeness. By providing a systematic framework for the deduction of
empirically verifiable conclusions, mathematical analysis helps the economist deter¬
mine the accuracy of his definitions and assumptions—if the conclusions are unten¬
able, the definitions and assumptions must be examined and revised.
The purpose of this book is to help the student understand, appreciate, and
perform applied mathematical analysis. Mathematical proofs are minimized except
when they can be made heuristic. The book is organized so that a type of analysis is
discussed first with respect to its mathematical (logical) procedure and then with
respect to its applications in economics and business. The assumptions required for
each type of analysis are emphasized and its applications are discussed in terms of
these assumptions.
The subsequent sections of this chapter review some of the basic mathematical
concepts and definitions used in later chapters.

Sets

A set is a collection of distinguishable well-defined objects or entities. The objects or

entities belonging to a set are said to be elements of the set. A set is determined either
by a list of its elements or by specifying a rule that determines whether a given object
or entity belongs to the set. Such a rule is referred to as a defining relation. Braces are
used to denote a set; either the elements of the set or its defining relation is written
inside the braces.

EXAMPLE

A = {a, b, c} means that the set A consists of the elements a, b, and c.

B — {x : x is an odd integer} means that the set B consists of the odd integers.
C = {1, 2, 3, 4, 5, 6} means that the set C consists of the numbers 1, 2, 3, 4, 5, and 6.
D = {y : y is an integer} means that the set D consists of the integers.
 Sets / 3

The notation x e S means that the entity or object x is an element of the set S. The
notation x 4. S means that x is not an element of the set S.
A defining relation that is satisfied by no element is said to define the empty set,
denoted by 0.

EXAMPLE

Referring to the preceding example,

a e A
b B
4 £ B
4 e C
4 E D
1$C
7 e B
7 e D
d <£ A
d£D

EXAMPLE

S = {x : x is an odd number ending in 2} = 0

P = {y : y is an odd number that is the square of an even number} = 0.

If every element of a set S is also an element of a set T, then S is said to be a subset

of T. If there is at least one element in T that is not also in S, then S is a proper subset
of T. The notation S c: T means that S is a subset of T; the notation S T means
that S is not a subset of T. Note that the empty set is a subset of every set. If S c= T
and T c: S, then every element in S is also in T and vice versa and S and T are the
same set. This is denoted S = T.

EXAMPLE

Referring to the preceding examples,

A cfi B
B a I)
BjzC
CcD

The union of two sets S and T is the set consisting of the elements of S and the
elements of T. The union of S and T, denoted by S u T, is described by

S u T = {x : x e S and/or x e T}
4 / Introduction

Note that an element common to both sets is not counted twice in the union.
The intersection of two sets S and T is the set consisting of the elements that are
common to S and T. The intersection of S and T is denoted by S n T and is
described by
S n T = {x: x e S and x e T}

The set difference of two sets S and T is the set consisting of the elements of S that
are not elements of T. The set difference of S and T is denoted by S — T and is
described by
S — T = {x : x e S and x $ T}

A universal set or universe is a specified set that contains all the elements of interest
for a particular discussion. If two sets have the property that their union is the
universal set and their intersection is the null set, one set is said to be the complement
of the other with respect to the universe. If U denotes the universal set, the comple¬
ment of a given set S is denoted by S' and is described by

S' = u -s
A statement of complementation must be qualified unless the universal set is clearly
understood. For example, if the complement of S is taken with respect to U and also
with respect to V, the notation S'v and S'v is used.

EXAMPLE

Referring to the preceding example,

B u D = D
A u C = {a, b, c, 1, 2, 3, 4, 5, 6}
A n D = 0
B n D = B
B n C = jl, 3, 5}
C n D = C
C-D = 0
C-B = {2, 4, 6}
D - A = D
D — B = {x : x is an even integer}

If U = {x : x is an integer}, then

B'u = {x : x is an even integer}

D'v = 0
If V = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, then

Cy = {7, 8, 9, 10}

If W = {x : x is a lowercase letter in the English alphabet}, then

A'w = {d, e,f g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y, zj

Sets formed by unions, intersections, and complements are shown here diagram-
matically; the shaded area is the set indicated below each diagram. This type of
representation of sets is called a Venn diagram.
 Sets / 5

A U B

PROBLEMS
1. If A, B, and C are sets such that A c B c= C, what is the relationship between C — B and
C - A?
2. Show that, in general, (A n B)' = A' u B'.
3. Show that, in general, A n (B vj C) = (A n B) yj (A n C).
4. Show that (S n 7') u (5 n 7) = S n (S u 7) = S u (S n 7) = S.
5. Show that K u (L n M) = (K u L) n (K u M).
6. If/4 n B = 0 and /4 n C = 0, is it necessarily true that B n C = 0?
7. If /4 ^ B and B ^ C, is it necessarily true that /4 ^ C?
8. If /4 B and B cfz C, is it necessarily true that /4 ^ C?
9. If /4 c: C and B c D, is it necessarily true that /luBcCuD?
10. If d c C and B <= D, is it necessarily true that /4 n B c C n D?
11. If S u T = {1, 2, 3, 4}, S n 7 = {1, 3}, and S — T = {2}, find S and T.
12. If d n B / 0, d n C # 0, and B n C # 0, is it necessarily true that Bn C#0?
13. If U is a universal set, determine which of the following statements are incorrect and
correct them by changing the right-hand side of the equation.
a. B u 0 = B h. C u C = C
b. C n U = C i. (D') = U
c. A u A' = U j. (4 — C)uC = 4 — C
d. B u U = U k. Bn (B — D) = ByjD
e. D n 0 = 0 l. If A = B\ then B — A'
f. 4 n 4' = 4 m. (C - D) = C' - D
g. B n B = 0 n. (4 u D) — D = 4 — D
14. If /4 = {c, / and B = {e, /i}, find
a. /4 — B c. A n B
b. B — A d. 4 u B
15. If 7 = {w, x, y}, S = {u, v, w} and 7 = {w, v, w, x}, and the universal set is
U = {m, v, w, x, y, z}, find
a. R' n T' n S' e. (5 u 7) - T
b. (R — S) n T f. (7 - 7) - (S - 7)
c. (7' - 7') u 5 g. (S - 7) - [(7 - 7) u (7 S)]
d. (7' u S')' h. (7 - 7) u S
16. If/4 n B = 0 and /!' = C, is it necessarily true that B c?
6 / Introduction

Answers to Odd-Numbered Problems

1. C -Be C - A
7. No
9. Yes
11.S' == {1, 2, 3}
T = {1, 3,4}
13. f. j- (A- C) u C = T u C
c

0
II

g- B n B = B k. B n (B - D) = B - D
i. (D')' = D m. (c- D) = C' u D
15. a. {z} c. S e. T g- 0
b. M d- {w} f. M h. S

Variables
In mathematics there are two types of quantities, constants and variables. A constant
is a quantity whose value remains unchanged throughout a particular problem. An
absolute or numerical constant has the same value in all problems; an arbitrary or
parametric constant (or a parameter) has the same value throughout any particular
problem but may assume different values in different problems. The absolute or
numerical value of a constant a is denoted by \a\ a and means the magnitude of a
regardless of its algebraic sign. Thus | + a | = — a a\ = a.
A variable is a quantity that assumes various values in a particular problem; the
set of values which a variable assumes is its range. A variable may be either contin¬
uous or discrete. A continuous variable is one that may assume any value within a
specified interval of real numbers (possibly all real numbers); successive values of a
continuous variable thus may differ by infinitesimal amounts. A discrete variable is
one that may assume only values specified in a countable range. Note that the
number of points or values in any specified real interval, say between 5 and 10, is
uncountable and no rule can be given for obtaining all possible such values.
However, the number of integers or integer values in any specified interval is count¬
able. The number of points on the real line (or on any part of the real line) is said to
be a noncountable or nondenumerable infinity. The number of integers on the real
line is a countable or denumerable infinity and any set obtained from the integers by
a specified rule is also countable or denumerable.
Thus in any problem there may be

1. Absolute constants that always have the same value; such quantities are either
numbers or symbols denoting numbers.
2. Parametric constants that have the same value throughout any given problem but
may have different values for different problems; such quantities depend on the
particular situation represented in the problem.
3. Variables that assume all values meaningful for the problem; such quantities may
vary discretely or continuously and may be restricted, for example, to positive
values.

EXAMPLE

In the equation of a straight line,

1 is a numerical constant, a and b are parameters, and x and y are variables.

 Variables / 7

EXAMPLE

In the equation for the area of a circle,

A = nr2

n is a numerical constant (approximately equal to 3.1416) and A and r are variables.

In pure mathematics, usually the letters at the beginning of the alphabet are used
to represent parameters and the letters at the end of the alphabet are used to repre¬
sent variables. However, in applied mathematics there are many exceptions to this
convention, and a variable is frequently represented by the first letter of its name—
for example, p for price, q for quantity, c for cost, s for savings, and so forth.
The only values of either constants or variables considered subsequently are real
numbers. The real number system includes all the familiar types of numbers—rational
numbers (positive and negative integers, fractions and decimal fractions) and irra¬
tional numbers that are roots of rational numbers and other numbers that cannot be
expressed as fractions; it does not include even roots of negative numbers (which are
called imaginary numbers) or numbers that have both real and imaginary parts
(which are called complex numbers). The set of all real numbers is denoted by

EXAMPLE

The equation x2 + 6 = 0 has no real solution since y/^-6 is an imaginary number.

EXAMPLE

The equation x5 + 32 = 0 has the real solution x = —2 (it also has four complex solutions not
of interest for the present discussion).

The quotient a/b of two numbers a and b is a number x such that a = bx. Division
by zero is not permissible as a result of this definition. Note that if b = 0, a = 0 * x,
which is true only if a = 0, in which case x can be any number. Thus the expressions

a 0
x and -
0 0

are meaningless. To distinguish between a/0 and 0/0, a/0 is sometimes said to be
“undefined” and 0/0 is sometimes said to be “indeterminate.” Note that 0/b = 0
for M 0, since 0/b is the value of x for which b ■ x — 0 and must be zero for
b^0.
Relationships involving inequalities are expressed as follows:

The notation < indicates “less than”

The notation > indicates “greater than”
The notation < indicates “less than or equal to”
The notation > indicates “greater than or equal to”
8 / Introduction

Relations and Functions

Many problems in mathematics involve sets of ordered pairs of numbers. For exam¬
ple, representation of an object moving in a straight line involves pairs of numbers
specifying distance of the object from the origin and the corresponding time in
motion. Representation of demand for a given commodity involves pairs of numbers
specifying quantity demanded and the corresponding price.
A set of ordered pairs of real numbers is referred to as a binary relation. The set of
first elements of a binary relation is called the domain of the relation; the set of second
elements is called the range of the relation. For a given set {(x, j/)}, x and y are
referred to as variables. The set of values variable x takes on is the domain and x is
usually called the independent variable; the set of values variable y takes on is the
range and y is usually called the dependent variable. When the number of variables is
clear from the context, a binary relation may be referred to simply as a relation.

EXAMPLE

Sx = {(1, 2), (2, 8), (2, 3)} is a binary relation whose domain is {1, 2} and whose range is
{2, 3, 8}.

52 — {(x, y): x, y real numbers, x < y} is a binary relation some of whose members are
(2, 2), (3, 4), (5, 5), and (8, 20). Note that (2, 1), (3, 2), and (25, 20), for example, are
not members of S2 ■

53 = {(x, y): y = x2, x e $} is a binary relation. The domain of S3 is & and the range is the
set of all nonnegative real numbers.

54 — {(x, y): y = x2 if 0 < x < 2, y — 3 — x if 2 < x < 3, and y = 3 if x = 3} is a binary

relation whose domain is the set x: 0 < x < 3 and whose range is the set
{y : 0 < y < 4}.

If a relation is such that to each element of the domain there corresponds one and
only one element of the range, the relation is said to be a function. Functions consti¬
tute a subset of relations—all functions are relations, but not all relations are func¬
tions. Note that the relations S3 and S4 of the above example are functions, but the
relations Sx and S2 are not functions. A special notation is used for functions to
denote the element of the range corresponding to an element of the domain. If /
denotes a function {(x, y)}, then the number y associated with a given x is denoted by
/(x), read “/ of x.” With this notation, the set of pairs defining/may be written as
{(*>/(*))} where y=f(x).

EXAMPLE

Referring to the preceding example, S3 and S4 can be written, respectively, as

53 = {(x, /(x)): /(x) = x2, x e

54 = {(x,/(x)):/(x) = x2 if 0 < x < 2 ,/(x) = 3- xif2<x<3, and f(x) = 3

if x = 3}
 Relations and Functions / 9

Other letters, for example, g, F, G, 0, and T, are also frequently used to indicate a
function. An equation such as g(x) = x2 + (1/x) provides a rule for finding the
second member of a pair whose first member is x. Such an equation or formula is said
to define the function, although the function is not the formula but is the set of
ordered pairs {(x, #(x))} or {(x, y)}. When a value of x is substituted in the formula for
a function, the result is said to be the function value or the value of the function for that
value of x.
In the case of a function involving two variables, whenever the value of the
independent variable is specified the value of the dependent variable is determined. It
is the independent variable whose value is thought of as being assigned arbitrarily
(except for nonpermissible values), thus determining also the value of the dependent
variable. It is conventional in applied mathematics to represent the independent
variable by x and the dependent variable by y.
For most problems in analytic geometry and other branches of pure mathematics,
the choice of independent and dependent variables is a matter of convenience and the
conventional designation of x and y is concerned only with graphical representation,
as discussed below. For example, in considering the equation

x — 4 v2 + 2y + 6 = 0
it is clearly more convenient to find pairs of points by regarding y as the independent
variable and x as the dependent variable as follows:

x = 4 y2 — 2y — 6

When the variables are purely mathematical and do not represent quantities in a
particular context, there is no basis other than convenience for making the choice of
independent and dependent variables. However, when the variables do represent
quantities in the context of a particular subject matter, the logic of the situation
usually determines the choice of independent and dependent variables. For example,
quantity produced is thought of as a primary determinant of total cost rather than
vice versa. There are exceptions even in these types of problems; for example, price
may be thought of as determining quantity demanded, or quantity demanded may be
thought of as determining price.

EXAMPLE

If/(x) = x2 - x + 2, then

f(z) = z2 - z + 2

/(2) = 4 - 2 + 2 = 4

/(_3) = 9 + 3 + 2= 14
/(0) = 0 — 0 + 2 = 2
f(a) — a2 — a + 2
f(x + 2) = (x + 2)2 — (x + 2) + 2

= (x2 + 4x + 4) — (x + 2) + 2

= x2 + 3x + 4

/(x + h) -/(x) = (x + h)2 - (x + h) + 2 - (x2 - x + 2)

= x2 + 2xh + h2 — x — h + 2 — x2 + x - 2
= 2 hx + h2 — h
10 / Introduction

EXAMPLE

If y — /(x) = (x + l)/x, then

/(!)= 1 +i = 2

/(-i)= =o

/(0) = 0 (undefined)

a h \ ci 1
f{a + h) — f(a) =
a + h a

a2 + ah + a — a2 — ah — a — h
a(a + h)
h
a(a + h)

A functional form can be obtained by the substitution of one form into another. If
y = f(x) and u = g(y) and if u = g[f(x)] = h(x), then h is called the g composite of f.

EXAMPLE

If/(x) = x2 — x — 1 and g(x) = x — 1, then

f[g{x)] = (x - l)2 - (x - 1) - 1 = x2 - 3x + 1

and

g[f(x)] = (x2 - x - 1) - 1 = x2 - x - 2

The above example illustrates the fact that, in general, f[g{x)\

EXAMPLE

If g(x) = x2 + 2, then

g[g{x)\ = (*2 + 2)2 + 2 = x4 + 4x2 -f 6

EXAMPLE

If/(x) = 10* and </>(x) = log10 x, show that f[4>(x)] = f[f{x)] = x.

f[(f)(x)] = 10^u> - 10logl°* = x

<t>[f(x)] - log10 f(x) = log10 lOx = x

 Relations and Functions / 11

PROBLEMS
1. For each of the following relations, state the domain and range and indicate whether the
relation is a function.
a. S = {(1, 3), (2, 3), (2, 4), (3, 2), (4, 1), (5, 5)}
b. A = {(1, 3), (2, 3), (3, 3), (4, 3)}
c. T = {(x, v): y = 4x 4- 1 if 0 < x < 2, y = 10 — x2 if 2 < x < 3}
d. B = {(x, y): y2 = x, y is an integer and |y | <8}
2. For each of the following, determine if the set {(x, y)} of pairs of real numbers formed
according to the given rule is a function.

a. yz = x x2 + 4
i- y =
x — 2
b. y3 = x
1
c. y4 = x j- y =
x2 — 6
d. x2 + y = 1
k. x =
e. x + y2 = 1 y -y + 2
f. x2 + y2 = 1
1. = X
g. y = x2 + 4 y2 + 2

h. xy = 1

3. If/(x) = x3 — x2 + 6, find a. /(0); b. /( -2); c.f(a); and d. f(y2).

3x2 - 8
4. If/(x) = , find a./(3); b./(- 1); c./(x - 2); and d./(a - h).
x — 1

5- = , find a./(-1); b./(4); c./(u2); and d./(x + 2).

6. If/(j;) = 2y + y, find a./(0); b./(- 1); c./(5); and d./(y + b).

7. If/(x) = 3x — x2, find a./(I); b./( —2); c./(a); and d. f(l/h).

8. If g(x) = a. ^(0); b. g{3); c. ^(1/x); and d. gf(x + h).

9. If h(x) = 4x — x2, find a. h(2) — h(4); b. h(?) • /i(2); c. h(a + b) — h(c);

and d. -7 4- [h(a)]2.
h(a)

10. Give the domain and range of each of the following relations; determine whether each
relation is a function and, if not, explain why.

a. y = x2 + 6 25
s- y = ~2
x
b. y = lOx — 5
x2 + 16
h. y =
c. y = + yj'4 — 2x2 x + 4

d. y = —y/4 — 2x2 1. y = x

j. 2x + y = 3
e. y = v/4 — 2x2
k. 3x2 + 2y = 4
9
f. y =
lOx — 5 l. 4x2 4- 6y2 = 10

t2 + 4
11. If/(x) = y - x and g(t) = , find
31

a. /(7) - 6f(3) and b.

/(3)
flf(2) + 1 •
12 / Introduction

x2 - 1 7
12. If q(x) = p{x) + g(x) and p(x) = —-—, g(x) = , find q(2).
J X

13. If h(x) = x3/2 and ()(x) = (x2 + 1)_1, find Q[h(x)].

14. If h(y) = ey and (9(x) = x2 + 4, find Q[h(y)].

15. If h(x) = x3 + 3x 4- 6 and g(y) = —, find g[h(2)].

16. If/(x) = 83 , g(x) = x2 and Q(x) - x3 - 10, find Q[f(-2) + g(2)].

X

17. If g(t) = t2 + 3 and Q(t) = t~\ find Q[g(t)].

18. If/(f) = f3 + a and g(x) - x-3, find g[f(t)\.

19. If f(t) — et + x, h(t) = eb2f and = yb\ find ■

h(t)
20. If /j(x) = | In x and g(x) = e2x, find h[g( 10)].

Answers to Odd-Numbered Problems

1. a. domain: {1, 2, 3, 4, 5}
range: {1, 2, 3, 4, 5}
not a function
b. domain: {1, 2, 3, 4}
range: {3}
a function
c. domain: {x : 0 < x < 3}
range: {y : 1 < y < 9}
a function
d. domain: {0, 1, 4, 9, 16, 25, 36, 49, 64}
range: (0, +1, +2, +3, +4, +5, +6, +7, +8}
not a function
3. a. 6 c. a 3-a2+6
b. —6 d. y 6 - / + 6

5. a. T5 [ x2 -I- 4x

b. ^3
d- V TTT
c.
a

7. a. 2 3h - 1

b. -10

c. 3a — a2

9. a. 4 1 + (4a — a2)3

b. 7 4a — a2

c. 4(a + 6 — c) — a2 — 62 + c2 — 2ab
11. a. b. 0

13. (x3 + 1)~ 1

15.

17. (f2 + 3)_ 1

19. e«b2
 Relations and Functions / 13

INVERSE FUNCTIONS
The domain and range of any relation may be interchanged to form a new relation.
Each pair in the new relation is obtained by interchanging the elements of a
corresponding pair in the original relation. Two such sets of pairs are said to be
inverse relations; each relation is said to be the inverse of the other. If both relations
are functions, they are called inverse functions. The inverse relation is a function if and
only if the original function is such that to each element of its range there corre¬
sponds one and only one element of its domain. The inverse of a function/is denoted
by the symbol f~l. In this notation — 1 is not an exponent; it means only that/ -1 is
the inverse of/ If/-1 is the inverse function of/ then

f[f-\x)\ = x for all x in the domain of/-1

and

/ 1 [/(*)] = x f°r x *n the domain of/

Providing the algebra can be done, the relation /-1(x) can be found by solving
f[f l(x)\ — x as though /- *(x) were a variable in the equation. When a domain is
not specified, it is assumed to be the set of all real numbers.

EXAMPLE

Let g = {(x, y): y = 2x — 1}. Find the inverse relation and determine whether it is a function.

y = 2x — 1

x = i(y+ 1)

which is a function; thus the inverse function is

g~l ={(x, y):y = i(x + 1)}

Since the letters used to indicate the values of the domain and the range are arbitrary, x and y
are used in their customary order.

EXAMPLE

Find the inverse of/= {(x, y): y = x2, x > 0} and determine whether it is a function.

y = x2

* = Vy
which is a function; thus the inverse function is

/-1 ={(*» l):>; = v/x}

Note that if / had as its domain the set of all real numbers in the above example,
f~1 would not be a function since its domain would be the set of all nonnegative real
numbers and its range would be the set of all real numbers.
14 / Introduction

EXAMPLE

If y = /(x) = (1 — x)/(l + x), find x = g(y). Solving for x,

1 — x
y = r+x
y(l + x) = 1 — x

y + xy = 1 — x

y + xy + x — 1=0

x -I- xy = 1 — y

x( 1 + y) = I - y

x = \~j = 9(y)

Note that /(x) = g(x) and thus/(x) = f~ x(x).

In summary, y = /(x) is a function if y is uniquely determined for every x, that is, if

there is one and only one value of y associated with each value of x. If one and only
one value of x is associated with each value of y, then there is an inverse function
x = d(y)i y — f(x) and x = g(y) are inverse functions if and only if f[g(y)\ =\ y and
g[f(x)] = x. The inverse function of /(x) is frequently denoted /_1(x); then
nr 1 (x)] = / 1 [/(*)] ~ x. Note that the choice of letter to represent the indepen¬
dent variable in a function is arbitrary; for example,/(x) and f(y) denote the same
function. Because it is customary to denote the independent variable by x and the
dependent variable by y, if g(y) is the inverse function of/(x), it is frequently denoted
by f~ J(x).

PROBLEMS
1. Determine for each of the following whether the inverse relation is a function; if it is not,
alter the range of the given function so that its inverse is a function.
a. {(x, y): y = x2 + 1} c. {(w, z): z = y/l - w2}
b. {(x, y): y = 4 - x2} d. {(«, v): v = \ u |}
2. For each of the following functions, find the inverse function f~1(x) and show that
/[/~1(x)]=/-1[/(x)] = x.
a. /(x) = 3x + 2 c. /(x) = (x - 2)/(x + 2)
b. /(x) = x/(x - 4) d. /(x) = (x + 3)/x
3. Find f[g{x)] and g[f(x)] for each of the following.
a- /(x) = l/(x - 1) and ^(x) = x2/(x2 - 1)
b. /(x) = x/(4 - x) and g(x) = x/(x - 4)
c- /(x) = ^(x) = (x + l)/(x - 1)
d. f(x) = y/x- 1 and g(x) = l/(x -f 1)
4. If/(x) = (ax + 1 )/(bx — 1), find values of a and b such that/is its own inverse—that is,
/[/(x)] = x.

5. If g(h) = he1,h and W~j = —find

a- ^[T(0] and b. F[g(t)].

6. If/(x) = x(x + 1), show that f(x + h) - f(x) = h(2x + 1 + h).

7. If/(x) = -, show that f(x + h) - f(x) = .

x x + hx
 Relations and Functions / 15

8- If g(y) = , 3 - , show that i[g(y) + <?(->>)] = g(y2).

i-y
9. If F(z) = log z, show that F(xy) = F(x) + F(y).

10. If </>(£) = 2K, show that (p(R -f 1) = 2(j)(R).

h
11. If P(x) = ^/x, show that P(x + /?) — P(x) =
x + h -k/x

12. Iff(x) = x2 — 1 and g(x) = 2x + 1, show that f[g(x)] = 4x(x + 1).

x
13. Iff(x) , show that/(x) +/( —x) = 2/( — x2).
1 + x
.2
14. If g(y) = y2 L/.A _
and h(y) = >' „u~„,
, show L/..2X _
that h(y2) = 9(y) .
i - y i - g(y)
15. If g(x) = 2 In x and/(x) = x3/2, show that Q[f(x)] = jQ(x).

16. Iff(x) = x2, show that f(x — h)—f(x)= f(h) — 2hx.

17. If h(x) x13, g(x) = (x9 + x6)1/2, and Q(x) = x(x -I- 1)1/2, show that
g[Hx)] <2(4
18. If/M = and g(y) = show that/M - #M = 2/M).
1 - y i
1 r r/.A ,
19. If/M = and g(y) = , show that f(y) -I- g(y) +, 0(4 1

r+M \+y2’— ” ' yv7/ ' /M /M

x + 2
20. If/(x) = , g(x) = ^ X , and h(x) = ~ — , show that f[g(x)] = ^ 12
x — 2 1 + x [M4];
i
21. If/(x) = x — 1 and g(x) = - , show that f(x2)g(x) — f(x).
X “1“ 1

22. If/M = 2> and g(y) = - —, show that f(y)g(y) =f(~y2).

i +y l-y

Answers to Odd-Numbered Problems

1. a. inverse a function
b. inverse a function if y < 0 or y > 0
c. inverse a function
d. inverse a function
3. a. f[g(x)] = x2 - 1; g[f{x)] = l/x(2 - x)
b. f[g(x)] = x/(3x - 16); g[f(x)] = x/(5x - 16)
c- /[<?(*)] = g[f{x)] = x _
d. f[g{x)] = J -xj(x + f); g[f(x)\ = (Jx-\ l)/(x — 2)
1 +f 1
5. a. b. 2^2/r
1 + v 1 + re
Graphical
Representation

1.1 Introduction
The French philosopher, Rene Descartes (1596-1650), developed a form of graphic
analysis in which algebraic equations are plotted in terms of geometric curves. Thus
it is possible to see relationships among such variables as price and quantity; direct,
indirect, and total cost; savings, investment, and consumption. In addition, much of
the theory of calculus can be presented in geometrical terms.
In order to relate algebra and geometry, and thus make possible this dual
representation, use is made of a coordinate system which provides a means for
locating specific points in a plane or in space. This correspondence may be estab¬
lished in many ways, but the system most commonly used is rectangular coordinates.
In this chapter, graphical representation using rectangular coordinates is dis¬
cussed for straight lines and for quadratic, exponential, trigonometric, and logarith¬
mic curves. General methods of graphing and special methods of graphing particular
types of curves are considered. Applications of each type of curve in business and
economics are also discussed.

1.2 Rectangular Coordinates

For rectangular coordinates, two straight lines intersecting at right angles are used as
lines of reference and a point is located in the plane of these lines by giving its
perpendicular direction and distance from each of them. These two distances with
signs indicating their directions are the coordinates of the point. The lines from which
the distances are measured are coordinate axes or, briefly, axes. The point of intersec¬
tion of the axes is the origin of coordinates or, briefly, the origin. The coordinate axes
divide the plane into four areas or regions called quadrants, numbered counterclock¬
wise for reference as shown in Figure 1.1.
Generally the horizontal line is called the x-axis and the vertical line is called the
y-axis. The choice of positive directions is a matter of convenience and may be
changed to simplify particular problems. However, it is customary to consider dis¬
tances measured to the right of the y-axis as positive and those measured to the left as
16
 1.3 Straight Lines / 17

Figure 1.1 The rectangular coordinate system.

y
II(-,+) I(+,+)

•(-3,2) • 0,2)
(-1,1) • •(3,1)
x

(-4,-2)
• • (3,-2).
(-2,-2)
• (2,-3)

III ( —» “) IV(+, -)

negative; similarly, it is customary to consider distances measured upward from the

x-axis as positive and those measured downward as negative.
If x and y represent variables having the same physical or geometrical character¬
istics, or if they are given abstractly as numbers with no physical interpretation, as is
frequently the case in analytic geometry, the same unit of measure or scale is used on
both axes whenever convenient. If the variables are measured in different units (for
example, if x is quantity and y is total cost), the units of measure appropriate for the
particular problem are used.
The x-coordinate, or abscissa, of a point is that coordinate which indicates the
direction and distance of the point to the right or left of the y-axis; the y-coordinate,
or ordinate, of a point is that coordinate which indicates the direction and distance of
the point above or below the x-axis. The position of a point is indicated by writing its
coordinates in parentheses in x, y order: (abscissa, ordinate). Locating a point when
its coordinates are given is called plotting the point.

Fundamental Principle
There is a one-to-one correspondence between number pairs and points in the plane
in which they are represented—that is, to each pair of numbers (coordinates) there
corresponds one and only one point in the plane and, conversely, to each point in the
plane there corresponds one and only one pair of numbers. The terms point and
coordinates of a point are generally used interchangeably.

1.3 Straight Lines

In this section graphical representation of straight lines is considered. The slope of a
straight line is defined and four forms of the equation for a straight line are discussed.
The relationships among parallel, perpendicular, and intersecting lines are con¬
sidered and the interpretation of the intersection of two straight lines as the simultan¬
eous solution of the two corresponding equations is discussed. The concept of a
family of lines is defined and illustrated.
The rectangular coordinate system can be used to represent any geometric figure
18 / Graphical Representation

that corresponds to an equation in two variables (for example, an equation in x and

y). The geometric figure corresponding to an equation consists of all the points, and
only those points, whose coordinates satisfy the equation; in particular, for every
straight line in the coordinate plane there is a corresponding linear equation in two
variables, and conversely.
An equation of the form

Ax + By + C = 0 (1)

where A, B, and C are constants and at least one of A and B is nonzero, is said to be
linear in x and y and such an equation has a straight line as its geometrical represen¬
tation. Equation (1) is also referred to as the general equation of the first degree in two
variables.
The degree of a variable is the value of the positive integral power to which the
variable is raised. The degree of a term of an equation in x and y is the sum of its
degrees in x and y; that is, the sum of the powers to which x and y appear in the term.
The degree of an equation is the degree of its term of greatest degree. For example,
x3 + y4 + 6 4- y — 12 = 0 is a fourth-degree equation; 4x2 + y4 + 2x3y3 — 21 = 0 is
a sixth-degree equation.
There are two aspects of the problem of correspondence between a straight line
and a linear equation: (1) given a linear equation, to graph the corresponding line,
and (2) given (conditions determining) the line, to find the corresponding linear
equation.
The coordinates x and y of every point (x, y) on a given line satisfy the equation
corresponding to the line; conversely, the straight line that passes through all the
points, and only through those points, whose coordinates satisfy the equation is
called the graph or locus of the equation. However, since there are an infinite number
of points on any given line or curve, establishing correspondence between algebraic
equations and their geometrical representations point by point is clearly not feasible;
it is the purpose of analytic geometry to develop methods using a minimum number
of points for establishing this correspondence. In general, the more simple the alge¬
braic equation (and thus its corresponding graph), the fewer points necessary for
establishing meaningful correspondence.
One easy way to graph a straight line is to compute its intercepts. The intercepts of
a line are the points where the line crosses the axes. Thus the y-intercept is the point
determined by setting x = 0 in the equation of the line; similarly, the x-intercept is
the point determined by setting y = 0 in the equation of the line.

EXAMPLE

Graph the line 4x + 5y — 20 = 0.

y-intercept x-intercept

if x = 0 if y = 0,

5y -20 = 0 4x - 20 = 0

5 y = 20 4x = 20

y=4
Intercept (0, 4) Intercept (5, 0)
 1.3 Straight Lines / 19

Figure 1.2

The intercepts may then be plotted and the line drawn through them, as shown in
Figure 1.2. As a check for accuracy, a third point is sometimes plotted as well. For
example, let y — 2 in the example above. Then

4x + 5(2) - 20 = 0

4x - 10 = 0
Y _ 10 __ 5
X — 4 — 2

The third point is then (f, 2).

SLOPE OF A STRAIGHT LINE

The direction of a straight line is indicated by its slope, which is defined in terms of
the angle between the line and the x-axis. When a straight line intersects the x-axis,
its angle of inclination is the angle 9, shown in Figure 1.3, which is measured counter¬
clockwise from the positive direction of the x-axis to the line and thus is always
between 0 and 180°. If a line is parallel to the x-axis, its angle of inclination 9 is
defined to be 0°.
The slope of a line is the tangent of its angle of inclination and is usually denoted
by m. The definition of the slope of a line is equivalent to the definition of the tangent
of its angle of inclination. The tangent is one of the six trigonometric functions of an
angle, which are defined in terms of a right triangle formed by the angle and a
perpendicular to its adjacent side.
Suppose 0° < 9 < 180°, and let a line having angle of inclination 9 intersect the
x-axis at A. From any other point on the line drop a perpendicular cutting the x-axis
at C, as in Figure 1.3. Now the tangent of 9, denoted tan 9, is defined by tan 9 =
CB/AC, where CB and AC are directed distances.
Equivalently, if (xls y^ and (x2, y2) are any two distinct points on a straight line,
then the slope of the line is

yi ~ yi Ay
m = tan 9
X2 - Xj Ax

It is customary, especially in calculus, to use the notation Ay, read delta y, to repre¬
sent the difference y2 — yx. This symbol Ay denotes the change in y or the difference
between two values of y. Do not think of Ay as delta times y. Similarly, Ax denotes
x2 - Xj.
20 / Graphical Representation

Figure 1.3

y y

It can be shown, by an argument based on properties of similar triangles, that the

slope of a straight line is the same regardless of the two distinct points chosen for
calculating it; correspondingly, the tangent of an angle is the same regardless of the
point at which the perpendicular intersects the adjacent side, provided the point is
not the intersection of the line and the x-axis.
Since

y2 - yi = yi - y2
x2 - Xi xx - x2
the slope of a line may be interpreted as the ratio of the directed change in vertical
distance to the corresponding directed change in horizontal distance as a point
moves along the line in either direction. In some problems (particularly in physics
and engineering), Ay = y2 — y1 is called the rise and Ax = x2 — Xi is called the run;
then the slope is defined as the rise per unit of run, that is, slope = rise/run = Ay/Ax.
If y\ — y2 and xx x2, the line through (xls yt) and (x2, y2) is parallel to the
x-axis, its angle of inclination 0 is zero, and tan 6 — Ay/Ax = 0.
If xx = x2 and yl y2, the line through (x1? yj and (x2, y2) is parallel to the
y-axis, its angle of inclination 6 is 90°, and tan 6 = Ay/Ax is undefined (see Figure
1.4). (Remember that an expression which involves division by zero is said to be
undefined or indeterminate and no numerical value can meaningfully be assigned to
it.)
summary: slope
The slope of the x-axis is zero.
The slope of the y-axis is undefined.
If a line slopes upward to the right, Ax and Ay have the same sign and tan 6 =
Ay/Ax > 0; in this case 0 < 6 < 90°.
If a line slopes downward to the right, Ax and Ay have opposite signs and tan 6 =
Ay/Ax < 0; in this case 90° < 6 < 180°.

EQUATIONS FOR STRAIGHT LINES

A straight line, the simplest geometric curve, may be specified uniquely by either of
the following: two points which lie on the line or the slope of the line and one
point which lies on the line. There are several formulas for obtaining the equation
of a straight line; the conditions that are given for obtaining the line determine
which formula is most convenient for a particular problem. The two-point form and
the intercept form determine a straight line by two points that lie on it; the point-
slope form and the slope-intercept form determine a straight line by one point that
 1.3 Straight Lines / 21

Figure 1.4

y y

o
o

(a) zero slope (b) undefined slope

(c) positive slope (d) negative slope

lies on it and its slope. These formulas are all equivalent to the general equation
for a straight line,

Ax By 4- C — 0

and are generally easily transformed into it and into each other.

Two-Point Form
One of the fundamental properties of a straight line is its constant slope; the slope
can be determined using any two distinct points on a straight line. These two facts
can be used to develop a formula for obtaining the equation of a nonvertical straight
line when the coordinates of two points on the line are known.
If (xl5 yx) and (x2, y2) are two distinct points on a nonvertical straight line, then
the slope m of the line is given by

*2 “ *1

If (x, y) is any other point (that is, a general point) on the straight line, then it and the
point (xj, >q) can also be used to determine the slope m of the line

X — Xj

and, since the slope is constant,

y - yi = y2 -yi
X — Xi X2 - Xj
22 / Graphical Representation

or

y - yl = y-2——~ (X-Xi) (la)

X2

See Figure 1.5, noting that the labeling of the points is arbitrary and may be chosen
for convenience.
Equation (la), referred to as the two-point form for a straight line, is generally most
convenient for determining the equation of a straight line when two of its points are
given.

EXAMPLE

Find the equation of the line passing through the points (3, 4) and ( — 5, 2).

(*i, yi) = (3, 4) (x2, y2) = (-5, 2)

y - 4 = i(x - 3)
x — 4y + 13 = 0

Check: Insert (3, 4) in x — 4y + 13 = 0:

3 - 4(4) +13 = 0

0 = 0

Insert (- 5, 2) in x - 4y + 13=0:

-5 -4(2)+ 13 =0
0 = 0

Intercept Form
For the special case when the point (x1? y.t) is the y-intercept, denoted by (0, b),
where b =/= 0, and the point (x2, y2) is the x-intercept, denoted by (a, 0), where a f 0,
 1.3 Straight Lines / 23

then equation (la) can be written

y (x-0)
a

y x
- 1 =

b a

x
= 1
a

Equation (lb), referred to as the intercept form for a straight line, is generally most
convenient for determining the equation of a straight line when its two intercepts are
given.

EXAMPLE

Find the equation of the line having intercepts (0, —6) and (4, 0).

b = —6 a = 4

Multiplying by the lowest common denominator 12,

3x — 2 y = 12

3.x — 2y — 12 = 0

Check: Insert (0, —6) in 3x — 2y — 12 = 0:

3(0) — 2( — 6) — 12 = 0

0 = 0

Insert (4, 0) in 3x — 2y — 12 = 0:

3(4) - 2(0) - 12 = 0
0 = 0

Point-Slope Form
Since the slope of a nonvertical line is m = (y2 — yi)/(x2 — xi), Equation (la) can be
written

y-y i = m(x-xl) (ic)

Equation (lc), referred to as the point-slope form for a straight line, is generally
most convenient for determining the equation of a straight line when one of the
points on a line and its slope are given.
24 / Graphical Representation

EXAMPLE _======^^

Find the equation of the line passing through the point ( — 1, 2) and having slope —4.

(*i, yi) = (— 1, 2) m = -4

y - yx = m(x - xx)

y — 2 = — 4(x + 1)

4x 4- y + 2 = 0

Check: m = — 4 in 4x + y + 2 = 0. Insert ( — 1, 2) in 4x +- y + 2 = 0:

4( — l) + 2 + 2 = 0
0=0

Slope-Intercept Form
For the special case when the point (x1? yt) is the ^-intercept, denoted by (0, b),
equation (la) can be written
y = mx + b (Id)
Equation (Id), referred to as the slope-intercept form for a straight line, is generally
most convenient for determining the equation of a straight line when its y-intercept
and slope are given.

EXAMPLE

Find the equation of the line having y-intercept (0, 5) and slope 3.

b — 5 m — 3

y = mx + b

— 3x + 5

3x — y + 5 = 0

Check: m = 3 in 3x — y + 5 = 0. Insert (0, 5) in 3x — y 4- 5 = 0:

3(0) -5 + 5 = 0
0 = 0

Vertical Lines
Since the slope of a vertical line is undefined, the above formulas are not appropriate
for obtaining the equations of vertical lines. A vertical line passing through the point
(xx, yx) has the equation x = xv

EXAMPLE

Find the equation of the vertical line passing through the point (5, —4).

x = xl = 5
x = 5
 1.3 Straight Lines / 25

Horizontal Lines
Since a horizontal line has zero slope, its equation may be obtained from the two-
point, point-slope, or slope-intercept forms. A horizontal line passing through the
point (xl5 yx) has the equation y = yx.

EXAMPLE

Find the equation of the horizontal line passing through the point ( — 3, —2).

y = yi = -2
y = -2

summary: straight lines

General form: Ax + By + C = 0

Two-point form: y — yx = —-— (x — x A

x2 - Xj

X V
Intercept form: - + 7=1
a b

Point-slope form: y — yx = m(x — xx)

Slope-intercept form: y = mx + b

Vertical lines: x = x{

Horizontal lines: y = yx

PROBLEMS
1. a. Which of the following points lie on the line 3x + 4y — 10 = 0?
i. (1,2) iv. (-25,21)
ii. (-2, 4) v. (0,0)
iii. (10, -5) vi. (4r, \)
b. Graph the line indicating points above which lie on it.
2. For each of the following equations,
i. Graph using intercepts.
ii. Put into slope-intercept form.
iii. Put into intercept form—this form is inappropriate for one of the equations; which one
and why?
a. 4y — 3x = 12 c. 2y + 3x + 2 = 0
b. 5x — y = 10 d. x — 3y = 0
3. a. Which of the following points lie on the line x — 5y + 4 = 0?
i. (0, 0) iv. (3, 2)
ii. (4, 0) v. (0, 4)
iii. (1, 1) vi. (— 1, 5)
b. Graph the line, indicating the points above which lie on it.
4. For each of the following pairs of points,
i. Find the slope of the line through the two points.
ii. Find the equation of the line using the slope.
iii. Find the equation of the line without using the slope.
iv. Graph the line.
a. (0, 0) and (6, 3) d. (3, -2) and (3, 5)
b. (-X, 0) and (0, |) e. (- 1, -2) and (4, 1)
c. (-7, 4) and (8, 4) f. (-2, -3) and (-5, -6)
26 / Graphical Representation

5. Find the equation of the line through the point (3, —2) and perpendicular to the line
through the points ( — 1, —3) and (3, 7).
6. Find the equation of the line through the point (4, 3) and parallel to the line through the
points (0, —3) and (6, 1).
7. Find the equation of the line through the point (5, 15) and parallel to the line y = x + 25.
What relation (parallel, perpendicular, coincident, or intersecting) does this line have to
the line through the two points (6, 0) and ( — 2, 8)?
8. Find the equation of the line which has intercept (0, —3) and is perpendicular to the line
through the points ( — 2, —1) and (2, 5).
9. Find the equation of the line which is parallel to the line through the points (5, 6) and
(7, 8) and also passes through the intersection of the line having slope —2 through the
point (-4, -6) and the line having slope 3 through the point (2, 2).
10. Find the equation of the line passing through the point (-1, 1) and perpendicular to the
line having slope -J and passing through the point (5, 2).

Answers to Odd-Numbered Problems

1. ii, iii, vi
3. a. (1,1), (0,|)
5. 2x + 5y + 4 = 0
7. x — y + 10 = 0
perpendicular
9. x _ y - 8 = 0

PARALLEL PERPENDICULAR,
AND INTERSECTING LINES
Any two lines in a plane are either parallel or intersecting; lines that intersect at right
angles are perpendicular. Two lines that are parallel have equal angles of inclination
and therefore have equal slopes, and conversely. Two lines that are perpendicular
have slopes which are the negative reciprocals of each other, and conversely.

EXAMPLE

The line 2x + 6y — 4 = 0 has the indicated relationship to each of the following lines (note
that 2x + 6y — 4 = 0 can be written in the form y = — ix + §):

(a) 4x + 12y — 8 = 0

12>’ = — 4x + 8

y = +i coincident with y = — ix + 3

(b) — 3x + y — 4 = 0

y = 3x + 4 perpendicular to y = — 3X + |

(c) x + 3y — 9 = 0

3y = — x + 9

y = -ix + 3 parallel to y = —-jx + 3

(d) 2x + y — 4 = 0
y = - 2x + 4 intersecting y = —3X -1- 3
 1.3 Straight Lines / 27

INTERSECTION OF TWO STRAIGHT LINES

The coordinates of the point of intersection of two straight lines must satisfy the
equations of both lines. Therefore, the point of intersection of two nonparallel lines
can be found by solving their equations simultaneously. The geometric property that
two lines intersect corresponds to the algebraic condition that their equations are
independent and consistent, and therefore have a simultaneous solution.
Consider two linear equations written in the form

A±x + B±y + Cj = 0 A2X y T C2 ~ 0

where Al9 A2, Blf B2, Cl9 and C2 are positive or negative constants.

Independent Equations
These equations are independent if one cannot be obtained from the other by multi¬
plication by a nonzero constant—that is, they are independent if the equality
a1/a2 = bjb2 = Cj tc2 does not hold. If the equality does hold, the equations are
dependent.

note: Twoequations that are dependent are equivalent—that is, they represent
the same line and have the same graph, since any point that lies on one of the
equations lies on the other also.

EXAMPLE

The equations

3x + 5y + 10 = 0 6x + lOy + 20 = 0

are dependent or equivalent, since the second equation is obtained by multiplying the first
equation by 2; they have the same straight line as their graph.

Consistent Equations
Two equations are said to be consistent if they can hold simultaneously—that is, they
are consistent if

A\ _ ^1 _ ^1 or if — ± —
A2 B2 c2 °r 1 A2 B2

If Ax jA2 = Bl /B2 = Cx /C2, the lines coincide and the equations are both consistent
and dependent. If A1 /A2 ^ Bv /B2, the lines have different slopes and thus intersect
at just one point; such lines are consistent and independent. Two linear equations
that are inconsistent—that is, not consistent—have parallel lines as their graphs.

EXAMPLE

The equations

3x + 5y + 10 = 0 6x 4- lOy +15=0

are inconsistent (since | ^ ^ {3) and cannot both hold simultaneously; they have parallel
lines as their graphs.
28 / Graphical Representation

Note that the equations Ay x + By y + Cy =0 and A2 x + B2 y + C2 = 0 can be

written

A, Ct
B,
and

C2
y=
B2

Thus if Ax /A2 = By /B2, the lines have the same slope. If By /B2 = Cy jC2, the lines
have the same intercept. If Ay/A2 = By /B2 = Cy /C2, the lines have the same slope
and the same intercept and are identical.
A point is said to be a simultaneous solution of two equations if its coordinates
satisfy both equations; geometrically, two lines intersect (or coincide) at a point
which is their simultaneous solution. Two straight lines have a unique simultaneous
solution if they are intersecting (consistent and independent), no simultaneous solu¬
tion if they are parallel (inconsistent), and infinitely many simultaneous solutions if
they are coincident (consistent and dependent). These cases are illustrated in
Figure 1.6.

Figure 1.6

lines inconsistent and independent lines consistent and independent

lines consistent and dependent

 1.3 Straight Lines / 29

Usually two linear equations can quite easily be solved simultaneously by elimina¬
tion or substitution; in some cases the method of determinants (to be discussed in
detail later) provides the solution more readily. To solve the equations

A^x-\-B^y-H Ui = 0 T 2 x T B2y T C2 — 0

by elimination, multiply the first equation by A2 and the second equation by Au

subtract to eliminate x, solve for y, and obtain x by substitution in either equation.
(Alternatively, multiply the equations by B2 and Bu respectively, and eliminate y.)
To solve by substitution, obtain x in terms of y (or, alternatively, y in terms of x)
from one equation, substitute in the other equation, and solve.
By either of these methods or, more directly, by the method of determinants, the
solutions for x and y are

BlC2 — B2Cl A2C1—AlC2

A±B2 — A2B± y A±B2 — A2 B1

provided AXB2 — A2Bl ^ 0.

For example, following the procedure for solution by elimination outlined above,
multiply the.first equation by A2 and the second equation by Ax:

A^A2x-\~A2Biy-\-A2Ci=0 A ^ A2 x T A j B2 y A^C 2 = 0

subtract to eliminate x,

(A2Bi — A1 B2)y + A2C1 — A1 C2 = 0

solve for y,

V — ^2^1 ~ A1 Cl
A\ B2 A2 Bi

and obtain x by substitution in the first equation

' A2 Ci ~ axc2\
Ax x + Bx + cx =0
,AlB2 — A2bx )

A2Ci — AxC2\ c±
x =
A, A]_ B2 A2 Bi J
—

— A2 B1 C1 T A1B1 C2 — A1 B2 C1 -I- A2B± C±

A1(A1B2 -- A2B{)

Bi C2 — B2 C i
AxB2 — A2 Bi

In many cases it is easier to obtain the simultaneous solution for a particular pair of
linear equations directly by elimination or substitution rather than by these
formulas.
30 / Graphical Representation

FXAMPLE

Find the point of intersection of the lines represented by the equations

3x — 4y + 6 = 0 x - 2y - 3 = 0

(see Figure 1.7)

Eliminating the x-term by multiplying the second equation by — 3 and adding it to the first
equation,

3x — Ay + 6 = 0

— 3x + 6y + 9 = 0

2y + 15=0

2y = -15

y - —t
Substituting y = in the first equation,

3x - 4(-^) + 6 = 0

3x + 30 + 6 = 0

3x = —36

x = —12

Answer: ( — 12, —ir).

Check: Equation 1: 3( — 12) — 4( —x) + 6 = 0

-36 + 30 + 6 = 0

0 = 0

Equation 2: — 12 — 2( —V) — 3 = 0

-12 + 15 - 3 = 0

0 = 0

Figure 1.7

y
 1.3 Straight Lines / 31

Alternatively, substituting for x obtained from the second equation in the first equation:

x = 2y + 3

3(2y + 3) - 4y + 6 = 0

6y + 9 — 4y + 6 = 0
2y + 15 = 0

2y = -15

y - —r

Substituting for y obtained from the second equation in the first equation:

3x — 2x+ 6 + 6 = 0

x + 12 = 0

x = —12

Answer: (-12, x), as above.

Using the formulas

_ C2 — B2 Ci — A2 C1 — A i C2
A1B2 — A2B1 A1B2 — A2B1

At = 3 Bt=- 4 Ci=6 A2 = l B2=- 2 C2 = -3

( 4)( — 3 )-( -2X6)

(3)( —2) - (l)(-4)

12 + 12 _ 24
-6 + 4~^2

(1X6) - (3X-3)
(3X-2 )-( l)(-4)
6 + 9 15
-6 + 4~^2

Answer: ( — 12, — x)> above.

These equations are consistent and independent and have a unique simultaneous solution,
the point (- 12, —x)-

EXAMPLE

Find the point of intersection of the lines represented by the equations

2x — 3y + 3 = 0 4x — 6y + 12 = 0
32 / Graphical Representation

Figure 1.8
y

Multiplying the first equation by 2 and rewriting the second equation,

4x — 6y -f 6 = 0 4x — 6y + 12 = 0

(see Figure 1.8). These equations are inconsistent and do not have a simultaneous solution;
they are represented by parallel lines.

EXAMPLE

Find the point of intersection of the lines represented by the equations

2x — 3y + 1 = 0 4x — 6y + 2 = 0

Multiplying the first equation by 2 and rewriting the second equation,

4x - 6y + 2 = 0 4x - 6y + 2 = 0

(see Figure 1.9). These equations are dependent and therefore equivalent and have infinitely
many simultaneous solutions; they are represented by the same straight line, every point of
which is a simultaneous solution of the equations.

Figure 1.9

y
 1.3 Straight Lines / 33

FAMILIES OF LINES
An important concept in both theoretical and applied mathematics is that of a family
of straight lines or curves. The members of a family possess some geometric property
(such as passing through a given point) in common, but differ in at least one other
geometric property (such as having different slopes). Algebraically, this corresponds
to the fact that some of the parameters in a general equation are specified and at least
one is not. Remember that an arbitrary (that is, unspecified) constant is called a
parameter.
The general equation for a straight line

Ax + By + C = 0

has two parameters, since if any one of the parameters A, B, or C is not zero, it may
be used as a divisor to reduce the equation to a form involving only two parameters.
The slope-intercept and intercept are two such forms.
The two essential parameters in the equation of a straight line correspond to the
two geometric conditions that may be imposed on the line. For each of these geome¬
tric conditions there is an algebraic equation showing the corresponding relation
between the parameters. Since any given number of variables is determined by that
same number of independent linear equations involving the variables, it follows that
the number of independent geometric conditions that can be put on a line is the same
as the number of essential parameters in the equation of a line, that is, two.
The general equation Ax + By 4- C = 0 with two essential parameters represents
all lines in the plane. When one of the two constants is specified and the other is
unspecified (that is, remains a parameter) a family of lines—such as all lines of a
given slope, or all lines passing through a particular point—is represented by the
equation. When both constants are specified, one particular line is represented by the
equation.

EXAMPLE

Consider the point-slope form of the equation for a straight line

t

y - yt = m(x - xx)

The two essential parameters are expressed in this form of the equation as the slope m and
the point (x1? yx). Suppose the condition that the line is to have slope 2 is imposed. This
geometric condition corresponds to the algebraic equation

y - yi = 2(x - xx)
The family of lines represented by this equation consists of an infinite number of lines parallel
to each other and all having slope 2. (See Figure 1.10.)
One further condition can be imposed on the line. Geometrically, a point through which
the line is to pass can be specified (as a special case, this point could be either the x-intercept or
the y-intercept); algebraically, a pair of constant values can be assigned to the coordinates
(x1? }q). Suppose it is specified that the line is to pass through the point (4, 5), that is,
(xi, yi) = (4, 5). (See Figure 1.11.) Now a particular line is determined by the equation

y — 5 = 2(x — 4) or 2x — y — 3 = 0

which is obtained from the equations

m = 2 and (xl9 yi) = (4, 5)

corresponding to the imposed geometric conditions; this line has slope 2 and passes through
the point (4, 5).
34 / Graphical Representation

Figure 1.10

Figure 1.11

EXAMPLE

Consider the intercept form of the equation for a straight line,

x
- +
a

The two essential constants are expressed in this form of the equation as the x-intercept
(a, 0) and the y-intercept (0, b). Suppose the condition that the line is to have the x-intercept
(3, 0) is imposed. This geometric condition corresponds to the algebraic equation

The family of lines represented by this equation consists of an infinite number of members,
each having a different y-intercept (and thus a different slope) but all passing through the point
(3, 0). (See Figure 1.12.)
One further condition can be imposed on the line. Geometrically, the y-intercept can be
specified; algebraically, a constant value can be assigned to b.
 1.3 Straight Lines / 35

Figure 1.12

Suppose that it is specified that the ^-intercept of the line is to be (0, — 1), that is, b = — 1.
(See Figure 1.13.) Now a particular line is specified by the equation

Figure 1.13

or

x — 3y = 3

x — 3y — 3=0

which is obtained from the two equations

a = 3 b = -1

corresponding to the imposed geometric conditions; this line has the x-intercept 3 and the
y-intercept — 1.

The idea of a general equation representing a large (frequently infinite) class of

straight lines or curves and the specification of subclasses of this general class (that is,
36 / Graphical Representation

families) by imposing certain geometric (and therefore corresponding algebraic) con¬

ditions on the lines or curves is very useful in both theoretical and applied
mathematics.

PROBLEMS
1. What relation (parallel, perpendicular, coincident, or intersecting) does the line
y — 2x — 4 = 0 have to the following lines?
a. y — x — 2 = 0 d. y — 3x — 4 = 0
b. 4y — 8x — 16 = 0 e. 2y + x — 6 = 0
c. 5y — lOx — 4 = 0
2. What relation (parallel, perpendicular, coincident, or intersecting) does the line
2x — 5y + 6 = 0 have to each of the following lines?
a. 15x + 6y + 9 = 0 d. 4x — 8y + 3 = 0
b. lOx + 4y — 5=0 e. 6x — 15y + 8 = 0
c. 4x — 10y +12 = 0 f. lx — 5y + 2 = 0
3. What relation (parallel, perpendicular, coincident, or intersecting) does the line
3x + 4y — 2 = 0 have to each of the following lines?
a. 15x + 20y - 10 = 0 d. 3x + y - 4 = 0
b. 8x — 6y + 5 = 0 e. 12x — 9y + 2 = 0
c. 9x + 12y + 7 = 0 f. 2x 4- y — 6 = 0
4. Determine whether each of the following pairs of equations are a. independent or depen¬
dent and b. consistent or inconsistent.
i. 2x — 6y + 5 = 0 and 3x — 8y + 3 = 1 * 3 * 5
ii. x + 5y — 2 = 0 and x + 5y — 5 = 0
iii. 3x — 9y + 12 = 0 and x — 3y + 4 = 0
iv. 5x — 4y — 6 = 0 and 4x — 5y + 6 = i
5. Referring to the pairs of equations in Problem 4, a. determine which pairs of equations
have simultaneous solutions and obtain those solutions and b. graph the pairs of
equations.
6. If a pair of equations are consistent and independent, do they necessarily have a unique
simultaneous solution?
7. If a pair of equations are consistent and dependent, do they necessarily have a simultan¬
eous solution? If so, is it unique?
8. Can a pair of equations be inconsistent and dependent?
9. If a pair of equations represent perpendicular lines, are the equations consistent? Are they
independent?
10. Graph the family of straight lines parallel to the x-axis and give the equation correspond¬
ing to this specification. Graph the member of this family that passes through the point
(10, -6).
11. Graph the family of straight lines passing through the point (— 1, 6) and give the equation
corresponding to this specification. Graph the member of this family which is parallel to
the line y + 6x — 5 = 0 and write the equation for this line.
12. Graph the family of straight lines perpendicular to the line 2x — 5y — 10 = 0 and give the
equation corresponding to this specification. Graph the member of this family which
passes through the point (4, — 1) and write the equation for this line.

Answers to Odd-Numbered Problems

1. a. intersecting d. intersecting
b. coincident e. perpendicular
c. parallel
3. a. coincident d. intersecting
b. perpendicular e. perpendicular
c. parallel f. intersecting
5. a. i. (11, f)
iii. all points on the line x — 3y + 4 = 0
iv. (6, 6)
 1.4 Applications of Straight Lines in Business and Economics / 37

7. yes; no
9. yes; yes
11. y = mx + 6 + m; y = —6x

1.4 Applications of Straight Lines

in Business and Economics

In this section several applications of straight lines in business and economics are
discussed. These applications include linear supply and demand curves and the
corresponding market equilibria, break-even analysis, and the consumption function.

LINEAR DEMAND AND SUPPLY CURVES

In practice, some demand and supply equations are approximately linear for the
relevant range; others are not linear. Even in the latter cases, linear equations may
provide reasonably accurate representations of supply and demand for a limited
range. In this section linear demand and supply equations are used for simplicity and
clarity in illustrating certain types of analyses. Nonlinear equations are discussed in
subsequent sections.
Figure 1.14(a) shows a more general representation of supply and demand curves.
Figure 1.14(b) represents supply and demand as linear functions. It should be noted,
as indicated in Figure 1.14, that only the segments of the equations which fall in the
first quadrant are pertinent to economic analysis. This is because supply, price, and
quantity are, in general, either zero or positive. For example, in simpler forms of
economic analysis:
Negative supply implies that goods are not available in the market either because
they are not produced or because they are withheld until a satisfactory price is
offered.
Negative price implies that prices are paid to buyers for the removal of goods from
the market.
Negative demand implies that price is so high as to preclude market activity until
quantities are offered at a satisfactory price.
These cases can occur, but their incidence is infrequent and is considered only in
more advanced economic analysis.
It is very important to realize that the straight line, mathematically, is perfectly

Figure 1.14
y y

Jt

(a) (b)
38 / Graphical Representation

Figure 1.15

y y
D

Price
O
a-

-x
Quantity Quantity
demanded demandec

(b) slope of (c) slope of

demand zero demand undefined

general; the formulas for a straight line do not indicate the range of values of x and y
which are to be considered. When it is specified, as it is for the present purpose, that
only zero or positive values of x and y are of interest—and conversely, that negative
prices or quantities are not meaningful—the range of values for x and y is restricted.
These restrictions are based on the interpretation and meaning of the equation for a
particular application; they are not based on its inherent mathematical properties.
This must be kept in mind to avoid misinterpretations, particularly when more
complicated equations are considered.

Linear Demand Curves

In the usual case, the slope of a demand curve is negative—that is, as price increases,
quantity demanded decreases, and as price decreases, quantity demanded increases.
In certain cases the slope of a demand curve may be zero—constant price regardless
of demand. In other cases the slope of a demand curve may be undefined—constant
demand regardless of price. Figure 1.15 illustrates these three cases.
Depending on the information available, various formulas for straight lines may
be most convenient for obtaining the demand function.

EXAMPLE

Ten watches are sold when the price is $80; 20 watches are sold when the price is $60. What is
the demand equation?

y 2 - yi
y-y i =
(x - Xi)
X2 — *1

*i = 10 yj = 80

*2 = 20 y2 = 60

/ 60 - 80^
V - 80 =
20 ^To)(x-10)

= — 2(x — 10)

2x + y - 100 = 0

(See Figure 1.16.)

 1.4 Applications of Straight Lines in Business and Economics / 39

Figure 1.16

EXAMPLE

When the price is $100, no watches are sold; when watches are free, 50 are demanded. What is
the demand equation?

-+H
a b

a = 50 b = 100

*+j^ = i
50 100

2x 4- y = 100

2 x + y — 100 = 0

(See Figure 1.17.)

Figure 1.17

y
40 / Graphical Representation

EXAMPLE

Because they are considered necessary for national security, 50 heavy-duty generators are
bought every year, regardless of price. What is the demand equation?

x = x i = 50

(See Figure 1.18.)

Figure 1.18

y
<u
_o
U-
Ou 50

(50, 0)
►-x
Quantity
demanded

Linear Supply Curves

In the usual case, the slope of the supply curve is positive—that is, as the price
increases, quantity supplied increases and as price decreases, quantity supplied
decreases. In certain cases, the slope of a supply curve may be zero—constant price
regardless of supply. In other cases, the slope of a supply curve may be undefined—
constant supply regardless of price (see Figure 1.19).
As in the discussion of demand curves, y represents the price, in appropriate units,
and x represents the quantity supplied, in appropriate units. Only positive values of x
and y are of interest, as previously stated. Note that for the supply curve in case (a)
the ^-coordinate of the ^-intercept may be positive, negative, or zero.
The x-coordinate of the x-intercept may be negative and, therefore, outside the
range of values of interest. This is reasonable since producers usually cease to supply
a commodity before the price reaches zero.

Figure 1.19

y y y
<L>
O
'c
Cl

- X -x
Quantity Quantity
supplied supplied

(a) slope of (b) slope of (c) slope of

supply positive supply zero supply undefined
 1.4 Applications of Straight Lines in Business and Economics / 41

EXAMPLE

When the price is $50, 50 cameras of a fixed type are available for market; when the price is
$75, 100 of the cameras are available for market. What is the supply equation?

y-yi = t
X2
—

~
~(x-xt)
A!

Xj = 50 yi = 50

x2 = 100 y2 = 75

sn = 1 75 - 50 )(x — 50)
y — 50
100 - 50

= i(x - 50)
x — 2y + 50 = 0

(See Figure 1.20.)

Figure 1.20

EXAMPLE

When the price is $25, no cameras of a fixed type are available for market; for every $10
increase in price, 20 more cameras are available for market. What is the supply equation?
y = mx + b m —\ b = 25
y = \x + 25
x — 2y + 50 = 0

(See Figure 1.21.)

Figure 1.21
X
42 / Graphical Representation

EXAMPLE

According to the terms of the contract between Company A and the telephone company,
Company A pays the telephone company $500 per month for long-distance calls of unlimited
length of time. What is the supply equation?

>’ = = 500

(See Figure 1.22.)

Figure 1.22

y
Price

(0, 500)
L k

\
o
o
II

Quantity
supplied

Figure 1.23

y y

y
 1.4 Applications of Straight Lines in Business and Economics / 43

Market Equilibrium
Market equilibrium is said to occur at that point (price) at which the quantity of a
commodity demanded is equal to the quantity supplied, as shown in Figure 1.23.
Thus, provided the same units for x and y are used in both equations, the equilibrium
amount and the equilibrium price correspond to the coordinates of the point of
intersection of the demand and supply curves. Algebraically, the equilibrium quan¬
tity and price are found by solving the supply and demand equations simultaneously
(again, provided the same units for x and y are used in both equations).
In general, for an equilibrium to be meaningful, the values of both x and y must be
positive or zero—that is, the demand and supply curves must intersect in the first
quadrant. This occurs if and only if the y-coordinate of the y-intercept of the demand
curve is greater than or equal to the y-coordinate of the y-intercept of the supply
curve and the x-coordinate of the x-intercept of the demand curve is greater than or
equal to the x-coordinate of the x-intercept of the supply curve. This can be seen
geometrically in Figure 1.23; the proof is given in Technical Note I.

EXAMPLE

Find the point of equilibrium for the following demand and supply equations.

y = 10 — 2x

y = fx + 1

Solving the equations simultaneously by substitution,

10 — 2x = fx + 1

tx = 9

y = 10 - 2(^)
_
—
34
~T

Answer: (^, ^).

(See Figure 1.24.)

Figure 1.24

y
44 / Graphical Representation

EXAMPLE

Find the point of equilibrium for the following demand and supply equations.
y = 5 — 3x
y = 4x + 12
Solving the equations simultaneously by substitution,
5 — 3x = 4x + 12
lx = — 7
x — —1
y = 5-3(—1)
8
Answer: (-1, 8) is not a meaningful equilibrium—note that bD = 5 < bs = 12.
(See Figure 1.25.)

Figure 1.25 y

BREAK-EVEN ANALYSIS
Break-even graphs are frequently used in business and economics to analyze the
implications of various pricing and production decisions. A break-even graph, in
simplified form, is shown in Figure 1.26.

Figure 1.26

y
 1.4 Applications of Straight Lines in Business and Economics / 45

In this particular case, costs have been divided into two general categories: fixed
and variable. Fixed costs remain constant at all levels of output and commonly
include such items as rent, depreciation, interest, plant, and equipment; variable
costs are those which vary with output and include such items as labor, materials,
and promotional expense. Total cost at any level of output is the sum of fixed cost
and the variable cost at that level of output.
In Figure 1.26 the straight line FC represents fixed cost. It is the straight line with
the y-intercept equal to the constant fixed cost (that is, cost when output is zero) and
slope zero.
The line TC represents total cost; its y-intercept is equal to fixed cost and its slope
is equal to the increase in variable cost per unit increase in output. Note the assump¬
tion that variable cost is proportional to output over the relevant range.
The line TR represents total revenue to the firm for different quantities sold; its
intercept is at the origin and its slope is equal to the price per unit, assuming that this
price is constant for all quantities sold.
The break-even point E is that point at which the lines TR and TC intersect. It
represents the quantity at which the producer just breaks even—that is, the quantity
for which there is just sufficient revenue to cover costs. Break-even analysis is used
more frequently in practice to demonstrate probable effects of changes than to deter¬
mine what those changes should be.

EXAMPLE

In Figure 1.27, suppose the firm decides that it could sell the same quantity of product,
denoted by point Q, if the price per unit were increased so that total revenue would be raised
from TR to TR'. TC remains the same, and the break-even point changes from E to E.
At the original price, profit to the firm is represented by the area of triangle EAB, at the
higher price, it is represented by the area of triangle EAC.
If, on the other hand, the firm assumes that the increase in price resulting in TR' would
reduce the quantity sold to that denoted by point Q', the following would be indicated:

Total cost reduced from that denoted by A to A'

Total revenue reduced from that of point C to C
Total profit changed from area EAB to area EA'C

Figure 1.27
46 / Graphical Representation

Since the profit area E'A'C' is greater than the profit area EAB resulting from the larger
quantity sold at the original price, the company might decide to increase the price even though
sales would be expected to drop.
Break-even analysis may be used to study possible results of many combinations of factors
involved in a particular problem.

EXAMPLE

Suppose the fixed cost of production for a commodity is $5000; the variable cost is $7.50 per
unit and the commodity sells for $10 per unit. What is the break-even quantity?

5000 = (10 - 7.50)x

5000 = 2.5x

x = 2000

The break-even quantity is 2000 units.

THE CONSUMPTION FUNCTION

National income analysis provides an interesting example of the use of linear func¬
tions, since the consumption function is frequently assumed to be linear over rela¬
tively short ranges or “ in the short run.” The analysis, in one of its simplest forms, is
based on the following assumptions:

1. There is some absolute amount of consumption necessary to maintain life, even

though there may be no money income.
2. Consumption is related to disposable income, that is, c =f(yd).
3. When disposable income increases, consumption will also increase, but by a smal¬
ler amount. Stated mathematically, if

Ayd represents an increase in disposable income

and

Ac represents the resulting change in consumption

then

Ac Ac
-t— will be positive but less than one, that is, 0 <-< 1
Ay,
4. The proportion of an increment in disposable income which will be consumed is
constant. This proportion is referred to as the “marginal propensity to consume.”

These assumptions can be translated into the point-slope form of the equation for
a straight line

c = a + byd

where c represents consumption; a is fixed, basic consumption regardless of income;

b is the marginal propensity to consume; and yd is disposable income.
 1.4 Applications of Straight Lines in Business and Economics / 47

EXAMPLE

When national disposable income is 0, national consumption is 5 (in billions of dollars). For
the economy as a whole consumption is linearly related to national disposable income as
follows: At each level of disposable income, consumption equals 5 (in billions of dollars) plus
80 percent of disposable income.
What is the equation that expresses this relationship?

c = 5 + 0.8yd

What is aggregate consumption when disposable income is 40 (in billions of dollars)?

c = 5 + 0.8(40) = 37

Graph the line representing aggregate consumption as a function of national disposable

income. (See Figure 1.28.)

Figure 1.28

EXAMPLE

Suppose that the consumption function is given by

c = 10 + 0.55yd

where c is aggregate consumption and yd is disposable income (in billions of dollars). What is
aggregate consumption when disposable income is 10 billion dollars? When disposable
income is 5 billion dollars, what proportion of it is consumed? When disposable income is 5
billion dollars, what proportion of aggregate consumption is consumption of disposable
income?
if yd = 10,

c = 10 + (0.55)(10)

= 15.5

Thus aggregate consumption is 15.5 billion dollars when disposable income is 10 billion
dollars.
48 / Graphical Representation

The proportion of disposable income that is consumed is 0.55 for any level of disposable
income.
If yd = 5,

c = 10 + (0.55)(5)

= 12.75

Thus, if disposable income is 10 billion dollars, consumption of disposable income is 2.75

billion dollars, which is yx or approximately 0.22 of aggregate consumption.

PROBLEMS
1. Which of the following equations represent demand curves, which represent supply curves,
and which represent neither demand nor supply curves? (Assume y represents price and x
represents quantity.)
a. x — 2y = 0 f. 2x T 5y + 4 = 0
b. 3x + 4y — 10 = 0 g. 3x + 4y — 12 = 0
c. y- 4 = 0 h. 5x — y — 10 = 0
d. x — 3 = 0 i. 2y T 3x -t- 2 = 0
e. 2x — 3y +1=0 j. x - 3y = 0
2. The demand curve for a commodity is x = 10 - y/4. (Assume y represents price and x
represents quantity demanded.)
a. Find the quantity demanded if the price is i. 4, ii. 16, iii. 25.
b. Find the price if the quantity demanded is i. 9, ii. 7, iii. 2.
c. What is the highest price that would be paid for this commodity?
d. What quantity would be demanded if the commodity were free?
e. Graph the curve.
3. The supply curve of a commodity is x = l.ly — 0.1. (Assume y represents price and x
represents quantity supplied.)
a. Find the price if the quantity supplied is i. 1, ii. 0.8, iii. 0.5.
b. Find the quantity supplied if the price is i. 8, ii. 6, iii. 4.1.
c. What is the lowest price at which this commodity would be supplied?
d. Graph the curve.
4. The demand equation of a commodity is x = A — By, where A and B are positive con¬
stants, y represents price, and x represents quantity demanded.
a. Find the price if the quantity demanded is A/3.
b. Find the quantity demanded if the price is A/2B.
c. Find the quantity demanded if the commodity is free.
d. What is the highest price that would be paid for this commodity?
5. The supply equation of a commodity is x = ay — b, where a and b are positive constants, y
represents price, and x represents quantity supplied.
a. Find the price if the quantity supplied is i. 5a — b, ii. a + 2b.
b. Find the quantity supplied if the price is i. 3b/a, ii. 5b/a.
c. What is the lowest price at which this commodity would be supplied?
6. For each of the following pairs of straight lines,
i. Determine which is the demand and which is the supply curve.
ii. Graph the curves and estimate the market equilibrium price and quantity.
iii. Solve the equations algebraically and verify the estimates for market equilibrium price
and quantity.
 1.4 Applications of Straight Lines in Business and Economics / 49

a. y = 10 — 2x and y = fx + 1 c. x = 15 - 3y and x = 2y — 3
b. y = 6 and x = 3y - 3 d. 2y + 3x = 10 and x = 4y - 6
7. A manufacturer sells his product at $5 per unit.
a. What is the total revenue for sales of 5000 units of product? What is the equation for
this revenue function? Graph the function.
b. Fixed costs are constant at $3000 regardless of the number of units of product involved.
Superimpose the graph of this function on graph a. above.
c. Total cost is equal to the sum of fixed costs and variable costs. In this company,
variable costs are estimated at 40 percent of total revenue. What is the total cost when
5000 units of product are sold? Graph the function superimposed on graph a.
d. What is the break-even point? Indicate this point on the graph, and solve for the
corresponding amount sold. Indicate on the graph the quantity at which the manufac¬
turer will cover his fixed costs.
8. At a price of $5 per unit, a firm will supply 5000 plastic battery lanterns every month; at
$3.50 per unit it will supply 2000 units. Determine the equation of the supply function for
this product. Graph the equation.
9. For the economy as a whole, consumption is linearly related to national disposable
income as follows: at each level of disposable income, consumption equals 3.5 (in billions
of dollars) plus 75 percent of disposable income, a. What is the equation that expresses this
relationship? b. What is aggregate consumption when disposable income is 50 (in billions
of dollars)?
10. A firm has analyzed its sales and found that its customers will buy 20 percent more units of
its products with each $2 reduction in the unit price. When the price is $12, the firm sells
500 units. What is the equation of the demand function for this product? Graph the
equation.
11. a. Assume that water is in unlimited supply in a municipality. The consumer pays $5.00
per month for water regardless of the amount used. Graph the demand and supply
equations.
b. There is only one genuine painting by Rembrandt titled “The Night Watch.” Assign
arbitrary values to the painting and graph the demand and supply equations.
12. A bus company has learned that when the price of a short excursion trip is $5.00, 30 people
will buy tickets; when the price is $8.00, only 10 tickets will be sold. Obtain the point-slope
form of the equation for the demand function and graph the equation.
13. Identify which of the following equations represents a demand curve and which represents
a supply curve; determine the equilibrium point and graph the curves.
a. x + y = 5
b. 2x — y = 5.5
14. Change equation b. in Problem 13 to 2x — y = 6. Graph the equation and identify it as
supply or demand. Has the equilibrium quantity involved increased or decreased relative
to that in Problem 13?
15. Suppose the fixed cost of production for a commodity is $45,000. The variable cost is 60
percent of the selling price of $15 per unit. What is the break-even quantity?
16. If profit is $100 per unit and the fixed cost of production is $225,000, what is the break¬
even quantity?
17. Aggregate national consumption is given (in billions of dollars) by the equation

c = 4.5 + 0.9 yd

where yd is disposable income. If disposable income is 15 (in billions of dollars), a. What is

aggregate consumption? b. What proportion of aggregate consumption is consumption of
disposable income?
18. Aggregate national income (in billions of dollars) is 4.8 plus 80 percent of disposable
income, a. What is the equation of the consumption function? b. What proportion of
disposable income is consumed? c. When disposable income is 60, what proportion of
aggregate consumption is consumption of disposable income?
50 / Graphical Representation

Answers to Odd-Numbered Problems

1. a. supply f. neither
b. demand g. demand
c. supply or demand h. supply
d. supply or demand i. neither
e. supply j. supply
3. a. i. 1, ii. 0.818, iii. 0.545 c. 0.091 (cannot be attained)
b. i. 8.7, ii. 6.5, iii. 4.41

a + 3b
5. a. i. 5, ii. c. b/a (cannot be attained)
a
b. i. 2b, ii. 4b
7. a. $25,000 d. break-even point at x = 1000;
TR = 5x recovery of fixed costs at
c. $13,000 x — 600
9. a. c = 3.5 + 0.15yd b. 41 billion dollars
13. equation a. is demand,
equation b. is supply
(3.5, 1.5)
15. 5000
17. a. 18 billion dollars b. 0.75

1.5 General Methods of Graphing

Nonlinear Curves

Any equation whose graph is not a straight line is referred to as nonlinear. There are
several types of nonlinear curves, all of which can be classified as either algebraic or
transcendental.
A polynomial in x and y is the sum of a finite number of terms of the type kxrys,
where k is a constant and each of the exponents r and s is either a positive integer or
zero. The degree of a term is r + s and the degree of the polynomial is the degree of the
term or terms of highest degree. Equations of the form f(x, y) = 0, where/(x, y) is a
polynomial in x and y, are referred to as algebraic; the graphical representation of an
algebraic equation is an algebraic curve. Note that a straight line is an algebraic
curve.
Any equation in x and y which is not algebraic is referred to as transcendental; the
graphical representation of a transcendental equation is a transcendental curve.
Included among the transcendental curves are the graphs of trigonometric, logarith¬
mic, and exponential functions. Curves represented by algebraic equations of degree
greater than two in x and y or by transcendental equations are referred to as higher
plane curves.
The general methods of graphing discussed below can be used for graphing any
type of algebraic or transcendental curve. However, for graphing many types of
curves, special methods of graphing, which are more efficient than the general
methods, are available. The general methods are discussed first. Then methods of
identifying and graphing specific types of quadratic curves (circles, ellipses, parab¬
olas, and hyperbolas) and trigonometric, exponential, and logarithmic curves are
discussed.
Graphing a straight line from its equation is a relatively simple problem, since any
two distinct points uniquely determine a straight line. However, graphing more
 1.5 General Methods of Graphing Nonlinear Curves / 51

complicated equations is less straightforward. The graph representing any equation

can be drawn fairly accurately if a sufficient number of points are plotted, but
frequently so many points are required that this method is very laborious. Further¬
more, plotting points in itself provides little or no information regarding the impor¬
tant properties of a curve.
From examining an equation, certain properties of the corresponding curve can
be determined which will facilitate drawing the curve with a minimum of point
plotting. Some of these properties are discussed below and their usefulness in gra¬
phing is illustrated. The properties discussed include intercepts, symmetry, extent,
asymptotes, factorization, and real, point, and imaginary loci.

INTERCEPTS
The intercepts of a curve are the points at which it crosses the axes. The x-intercepts
are obtained by setting y = 0 in the equation of the curve and solving for x; the
y-intercepts are obtained by setting x = 0 in the equation of the curve and solving
for y.

SYMMETRY
Two points are symmetric with respect to a line if that line is the perpendicular
bisector of the line segment joining the two points. Two points are symmetric with
respect to a third point if that third point is the midpoint of the line segment joining
the first two points. From these definitions it follows that the point (x, y) is a
symmetric to the point

(x, — y) with respect to the x-axis

(— x, y) with respect to the y-axis

( — x, — y) with respect to the origin

as shown in Figure 1.29.

A curve may also be symmetric with respect to the x-axis, the y-axis or the origin,
as shown in Figure 1.30. Refer to parts (a), (b), and (c) of that figure.
(a) A curve is symmetric to the x-axis if for each point (x, y) on the curve the
symmetric point (x, — y) is also on the curve—that is, if substitution of — y for y does
not change the form of the equation of the curve. For example, a curve is symmetric

Figure 1.29 Symmetry with respect to (a) the x-axis, (b) the .y-axis, and (c) the origin.
52 / Graphical Representation

Figure 1.30 Symmetry with respect to (a) the x-axis, (b) the .y-axis, and (c) the origin.

to the x-axis if it is represented by an algebraic equation in which y occurs only to

even powers. Note, however, that there are algebraic curves symmetric to the x-axis
whose equations contain odd powers of y. See the example which follows.
(b) A curve is symmetric to the y-axis if for each point (x, y) on the curve the
symmetric point (— x, y) is also on the curve—that is, if substitution of — x for x does
not change the form of the equation of the curve. For example, a curve is symmetric
to the y-axis if it is represented by an algebraic equation in which x occurs only to
even powers.
(c) A curve is symmetric to the origin if for each point (x, y) on the curve, the
symmetric point ( —x, — y) is also on the curve—that is, if substitution of — x for x
and — y for y does not change the form of the equation of the curve. For example, a
curve is symmetric to the origin if it is represented by an algebraic equation each of
whose terms is of even degree or each of whose terms is of odd degree.
In summary, the graph of an equation/(x, y) = 0 is symmetric with respect to

the x-axis if/(x, y) =/(x, — y) = 0

the y-axis if/(x, y) = /( — x, y) = 0

the origin if/(x, y) =/( —x, — y) = 0

Note that symmetry with respect to both the x- and y-axes implies symmetry
about the origin, but symmetry about the origin does not imply symmetry about
either axis. More generally, symmetry with respect to any two of the three (x-axis,
y-axis, and origin) implies symmetry with respect to the third. Thus symmetry can
occur with respect to none, one, or three of the x-axis, y-axis, and origin but never
with respect to just two of them. This can be seen from the definitions above:

Substitution of —y for y
leaves the equation unchanged
(x-axis symmetry) Substitution of — x for x and
and — y for y leaves the equation
substitution of — x for x ^ unchanged (symmetry to origin)
leaves the equation unchanged
(y-axis symmetry)

and similarly for the other two cases. (The symbol => means “ implies,” the symbol
=/> means “does not imply,” the symbol <= means “is implied by,” the symbol
means “ is not implied by.”)
 1.5 General Methods of Graphing Nonlinear Curves 53

EXAMPLE

The curve represented by the equation

3 x2y + y + x3 = 0

is symmetric to the origin but not to either axis (see Figure 1.31).

Figure 1.31

f(x, —y)= —3x2y — y + x3, so f(x, — y) = 0 is not the same equation as f(x, y) = 0 and
f(x, y) = 0 is not symmetric to the x-axis.
/( — x, y) = 3x2y + y — x3, so /( — x, y) = 0 is not the same equation as/(x, y) = 0 and
/(x, y) = 0 is not symmetric to the y-axis.
/(— x, — y) = — 3x2y — y — x3, so /(— x, — y) = 0 is the same equation as/(x, y) = 0 and
/(x, y) = 0 is symmetric to the origin.

EXAMPLE

The curve represented by the equation

2x4 + x2y + y + 6 = 0

is symmetric to the y-axis but not to either the x-axis or the origin (see Figure 1.32).
/(x, — y) = 2x4 — x2y — y + 6, so/(x, — y) = 0 is not the same equation as/(x, y) = 0 and
/(x, y) = 0 is not symmetric to the x-axis.

Figure 1.32
54 / Graphical Representation

f(-x, y) = 2x4 + x2y + y + 6, so /(-x, y) = 0 is the same equation as f(x, y) = 0 and

/(x, 3;) = 0 is symmetric to the y-axis.
f(-x, -y) = 2x4 - x2y - y + 6, so/(-x - y) = 0 is not the same equation asf(x, y) = 0
and f(x, y) = 0 is not symmetric to the origin.

EXAMPLE __=

The curve represented by the equation

5x2 — xy + 6x3 = 0
is not symmetric to the x-axis, the y-axis, or the origin (see Figure 1.33).

Figure 1.33 y

f(x, — y) = 5x2 + xy + 6x3, so /(x, — y) = 0 is not the same equation as/(x, y) = 0 and
/(x, y) = 0 is not symmetric to the x-axis.
/( —x, y) = 5x2 + xy - 6x3, so /( —x, y) = 0 is not the same equation as/(x, y) = 0 and
/(x, y) = 0 is not symmetric to the y-axis.
/( — x, — y) = 5x2 — xy — 6x3, so /( —x, — y) = 0 is not the same equation as/(x, y) = 0
and/(x, y) = 0 is not symmetric to the origin.

EXAMPLE

The curve represented by the equation

x3y + 3xy = 0
is symmetric to the x-axis, the y-axis, and the origin (see Figure 1.34).

Figure 1.34 y

= 0

x3y + 3xy = 0
 1.5 General Methods of Graphing Nonlinear Curves / 55

f(x, — y) = — x3y — 3xy, so f(x, — y) = 0 is the same equation as f(x,y) = 0 and

/(x, y) = 0 is symmetric to the x-axis.
/( — x, y) = — x3y — 3xy, so /( — x, y) = 0 is the same equation as /(x, y) = 0 and
/(x, y) = 0 is symmetric to the y-axis.
/( — x, — y) = x3y + 3xy, so /( —x, — y) = 0 is the same equation as /(x, y) = 0 and
/(x, y) = 0 is symmetric to the origin.

In graphing an equation it is seldom practical to consider symmetry with respect

to other than the coordinate axes and the origin; exceptions to this are curves whose
equations are of a particular form from which lines or points of symmetry are readily
identified. Examples of such equations are given in later sections.
Occasionally it is desired to construct an equation whose curve is symmetric with
respect to certain lines or points. In these problems the following rules, of which the
above are special cases, are useful (see Figure 1.35): The graph of the equation
/(x, y) = 0 is symmetric with respect to

the line x = h iff (h + c, y) = f (h — c, y) = 0 for all c and y

the line y = k if/(x, k 4- c) =f{x, k — c) = 0 for all c and x

the point (h, k) iff{h + c, k + d) =f(h — c, k — d) = 0 for all c and d

note: Symmetry about the line y = mx + b must be defined in terms of transfor¬

mations. This procedure is discussed in later sections.

Figure 1.35 Symmetry with respect to (a) the line x = h, (b) the line y = k, and (c) the point (h, k).

(c> 0)
fa) (b) (c)

Symmetry with respect to the line y = x is a property of inverse functions. Geo¬

metrically, the problem of finding the inverse function off (x) is the problem of viewing
the graph of y =/(x) with the x-axis and y-axis interchanged, which is equivalent to
reflection about the line y = x. Thus the graph of the inverse of a function can be
obtained by drawing a perpendicular from each point on the graph of the original
function to the line y = x and extending the perpendicular an equal distance beyond
the line y = x. The points thus obtained are the graph of the inverse function
y =f~1(x). (See Figure 1.36.)
56 / Graphical Representation

Figure 1.36 Symmetry with respect to the line y — x.

EXTENT
The rectangular coordinate system denned represents only those points (x, y) whose
coordinates are real numbers; thus values of x for which y is imaginary and values of
y for which x is imaginary must be excluded. When even powers of a variable appear
in an equation, the solution for that variable may involve square roots (or other even
roots) and the extent of the curve may then be limited, since negative numbers do not
have real square roots (or other even roots). When possible, an equation should be
solved explicitly for each variable to see whether the range of any variable is limited
and the curve is correspondingly limited in extent. Note that any limitation on the
extent of one variable may implicitly limit the extent of the other variable.

EXAMPLE

Determine whether there is any limitation on the extent of the curve represented by the
equation

x2 + y2 = 9

Solving for x,

x = ± y/9~— y2

the quantity under the radical sign 9 — y2 is negative if | y | > 3; the extent of the curve in the
y-direction is therefore limited to the interval — 3 < y < 3.
Solving for y,

y= ± — x2

and the extent of the curve in the x-direction is therefore similarly limited to the interval
— 3 < x < 3.
 1.5 General Methods of Graphing Nonlinear Curves / 57

EXAMPLE

Determine whether there is any limitation on the extent of the curve represented by the
equation

x2 - y2 = 4

Solving for x,

x = ± y/4 + y2

the quantity under the radical sign 4 + y2 is always positive and the extent of the curve in the
y-direction is therefore not limited.
Solving for y,

y = ± v/x2 - 4

the quantity under the radical sign x2 — 4 is negative if | x | <2; the extent of the curve in the
x-direction is therefore limited to the values — 2 < x < 2.

PROBLEMS
For each of the following equations, determine a. the intercepts; b. whether the curve repre¬
sented is symmetric to the x-axis, the y-axis, or the origin; and c. whether there is any
limitation on extent.

1. x3 — y2 — 9 = 0 4. xy + 5x — 15 = 0
2. x2 + y2 - 18 = 0 5. x2 + y4 - 6 = 0
3. y2 — 2x + 5 = 0 6. x2y2 — 25 = 0

Answers to Odd-Numbered Problems

1. a. intercepts: (^/9, 0)
b. symmetry: about the x-axis
c. extent: x3 > 9, y not limited
3. a. intercepts: (f, 0)
b. symmetry: about the x-axis
c. extent: x > f, y not limited
5. a. intercepts: (0, ^/6), (^/6, 0)
b. symmetry: about the x-axis, y-axis, and origin
c. extent: x2 < 6, y4 < 6

ASYMPTOTES
A curve may have the property that a point can move along it so that the distance
from the origin to the point increases without limit. If, as the distance (in some
direction) of a point from the origin increases without limit, the distance between the
point and a fixed straight line approaches zero through values of the same sign, the
line is called an asymptote of the curve.
To show that a particular line is an asymptote of a particular curve it is necessary
to show that the curve approaches arbitrarily close to the line as the distance from
the origin increases without bound. This property involves the concept of limits,
which will be discussed in succeeding sections.
In general, then, the line y = mx -F b is an asymptote of the curve y = f(x) iff(x)
becomes arbitrarily close to mx + b as x and y increase without bound. This may be
written f(x) -> mx + b as x, y -> oo. (See Figure 1.37.)
58 / Graphical Representation

Figure 1.37 /(x) -> mx + b as x, y -> oo; (a) from above, (b) from below.

y y

Asymptotes that are parallel to or coincident with one of the coordinate axes are
frequently of particular interest. These vertical and horizontal asymptotes are defined
as follows (see Figure 1.38).

The line x = h is a vertical asymptote of the curve y =f(x) if x h as y -► oo.

The line y — k is a horizontal asymptote of the curve y =f(x) if y -» k as x —► oo.

For purposes of graphing, it is helpful to determine the behavior of a curve with

respect to each of its asymptotes, in addition to determining the existence of the
asymptotes. Thus the equation should be examined as x and y in turn increase
without bound (x —► + oo, y -> + oo) and as x and y in turn decrease without bound
(x —► — oo, y -*■ — oo); the value of the variable which is not increasing or decreasing
without bound should also be noted, in order to determine whether the curve is
approaching its asymptote from the left or the right (vertical asymptotes) or from
above or below (horizontal asymptotes). When investigating asymptotes, it is usually
helpful to solve an equation explicitly first for one variable and then for the other
variable, if possible. As noted above, this procedure is also useful in determining
possible limitations on extent; since such limitations prevent the variable involved
from increasing or decreasing without bound, they are important in considering
asymptotes.

Figure 1.38

(a) vertical asymptote: x = h (b) horizontal asymptote: y = k

 1.5 General Methods of Graphing Nonlinear Curves / 59

EXAMPLE

Determine whether the curve represented by the following equation has vertical or horizontal
asymptotes:

xy + x — 3y — 2 = 0

(see Figure 1.39).

Solving for x,

3y + 2
* “ y + i

Figure 1.39

3;

as y -*■ + oo, x -► 3, and x < 3; as y -* — oo, x -► 3, and x > 3; thus x = 3 is an asymptote of the
curve and is approached from left and right.
Solving for y,

x —2
y - ,—
3 —x

as x -» + oo, y -*■ — 1, and y < —1; as x -► — oo, y -> — 1, and y > — 1; thus y = — 1 is an
asymptote of the curve and is approached from above and below.

EXAMPLE

Determine whether the curve represented by the following equation has vertical or horizontal
asymptotes:

x2 - y - 1 = 0

(see Figure 1.40).

Solving for x,

^ = ±V//)M- 1

asy->+oo, x->±oo;y> — 1, so y cannot approach — oo; thus the curve does not have a
vertical asymptote.
Solving for y,

y = x2 — 1

as x-++oo, y->+oo; as x—► — oo, y-++oo (as above); thus the curve does not have a
horizontal asymptote.
60 / Graphical Representation

Figure 1.40

FACTORIZATION
If the left member of the equation / (x, y) = 0 can be written as the product of two or
more factors, for example, if /(x, y) = g(x, y) • h(x, y) = 0, then only those points
(x, y) whose coordinates satisfy either g(x, y) = 0 or h(x, y) = 0 lie on the graph of
f(x, y) = 0. Thus the graph of/(x, y) = 0 consists of the graphs of g(x, y) = 0 and
h(x, y) = 0. It is important to check for factorization, since a factorable equation may
be difficult to graph correctly if it is not factored.

EXAMPLE

Graph the equation

x2y — xy2 — x + y = 0

(see Figure 1.41).

Figure 1.41

Factoring,

xy(x — y) - x + y = 0

*>’(* - y) - (x - y) = 0

(xy - l)(x - y) = 0
 1.5 General Methods of Graphing Nonlinear Curves / 61

Thus the graph of x2y — xy2 — x + y = 0 consists of the graphs of the curve xy — 1 = 0 and
the straight line x — y = 0.

EXAMPLE

Graph the equation

2x2 - 3xy - 2y2 = 0

(see Figure 1.42).

Figure 1.42

Factoring,

(2x -I- y)(x — 2y) = 0

Thus the graph of 2x2 — 3xy — 2y2 = 0 consists of the graphs of the straight lines 2x + y — 0
and x — 2y = 0.

REAL CURVE, POINT LOCUS, OR IMAGINARY LOCUS

Some equations are satisfied by the coordinates of only one point or of a finite
number of points; the graphs of such equations are called point loci. Other equations
are satisfied by the coordinates of no real point; since such equations have no
graphical representation in a system of real axes, they are said to represent imaginary
loci.

EXAMPLE

The equation x2 + y2 — 0 is satisfied only by the point (0, 0), and its graph is thus a point
locus.
62 / Graphical Representation

EXAMPLE

The equation (x* 1 2 3 * 5 — 4)2 + (y2 — 4)2 = 0 is satisfied only by the points (2, 2), (2, —2), ( — 2, 2),
and ( — 2, —2), and its graph is thus a point locus.

EXAMPLE

The equation x2 4- y2 — — 5 is satisfied by no pair of real numbers x and y and its locus is thus
imaginary.

PROBLEMS
For each of the following equations, determine a. the asymptotes; b. whether the equation can
be factored; and c. whether the equation represents a real curve, point locus, or imaginary
curve.

1. 2xy — x + y + 5 = 0 4. 3x2 + 3xy — 2x — 2y = 0

2. 3x2 + 2xy — y2 = 0 5. 3x2 + 6y2 + x2y2 = 0
3. 2x2 + 3y2 + 6 = 0 6. 3x2 - 4y2 - 9 = 0

Answers to Odd-Numbered Problems

1. a. asymptotes; y-> oo, x -> —■J; x -> oo, y
b. cannot be factored
c. real locus
3. a. asymptotes: none
b. cannot be factored
c. imaginary locus
5. a. asymptotes: none
b. cannot be factored
c. point locus

SUMMARY: GENERAL METHODS OF GRAPHING

NONLINEAR CURVES
In the preceding sections properties of curves that are helpful in graphing are dis¬
cussed. These properties include:

1. Intercepts 4. Asymptotes
2. Symmetry 5. Factorization
3. Extent 6. Real curve, point locus, or imaginary locus

The following examples illustrate the use of these six properties in graphing a
given equation. Although factorization occurs near the end of the list (because in
practice factorable equations occur relatively infrequently), it should be checked first,
as it may make investigation of some of the other properties unnecessary. The
 1.5 General Methods of Graphing Nonlinear Curves / 63

emphasis here is on determining properties of the curves from their equations;

greater accuracy in graphing can be attained by plotting more points and, partic¬
ularly, by methods of differential calculus to be discussed subsequently.

EXAMPLE

Graph the equation

y = (x + 2)(x — 3)2

(see Figure 1.43).

Figure 1.43

Intercepts: If y = 0, x = — 2, 3; if x = 0, y = 18.
Symmetry: Not symmetric with respect to either axis or the origin.
Extent: No apparent limits.
Asymptotes: None (as x -> + oo, y -*■ + oo and as x -► — oo, y -* — oo).
Factorization: Not factorable. [Note that factorability refers to the equation /(x, y) =
(x + 2)(x — 3)2 — y = 0, which is not factorable, although y = /(x) = (x + 2)(x — 3)2 is, in
fact, factored.]
Real, point, or imaginary locus: Real curve.

EXAMPLE

Graph the equation

x2 — 4x -I- y — 12 = 0

(see Figure 1.44).

Intercepts: If y = 0, x = — 2, 6; if x = 0, y = 12.
Symmetry: Not symmetric with respect to either axis or the origin.
Extent: No apparent limits.
64 / Graphical Representation

Figure 1.44

Asymptotes: None (as x -► + oo, y -► —oo and x-* — oo, y -► —oo).

Factorization: Not factorable.
Real, point, or imaginary locus: Real curve.

EXAMPLE

Graph the equation

xy — y — x — 2 = 0

(see Figure 1.45).

Intercepts: If y = 0, x = — 2; if x = 0, y = -2.
Symmetry: Not symmetric with respect to either axis or the origin.
Extent: Solving for x,

So no limitation on the extent of y, except x is not defined for y = 1. Solving for y,

x + 2

So no limitation on the extent of x except y is not defined for x = 1.

Asymptotes: x = (y + 2 )/(y — 1). As y -► + oo, x -► 1 and x>l; as y — oo, x -> 1 and
x<l; so x = 1 is an asymptote, y = (x + 2)/(x — 1). As x-»+oo, y-> 1 and y > 1; as
x —► — oo, y -*■ 1 and y < 1; so y = 1 is an asymptote.
Factorization: Not factorable.
Real, point, or imaginary locus: Real curve.
 1.5 General Methods of Graphing Nonlinear Curves / 65

Figure 1.45

EXAMPLE

Graph the equation

xy + x2 = 0

(see Figure 1.46).

Figure 1.46

xy + x2 = x(y + x) = 0, so the graph consists of the two straight lines x = 0 and x 4- y = 0.

The line x = 0 is the y-axis. The line x + y = 0 has intercept (0, 0) and slope — 1.

EXAMPLE

Graph the equation

y3 + xy2 — xy — x2 = 0

(see Figure 1.47).

66 / Graphical Representation

Figure 1.47

y3 + xy2 - xy - x2 = y2(y + x) - x(y + x) = (y2 - x)(y + x) = 0, so the graph consists

of the straight line x + y = 0 and the curve y2 - x = 0. The line x + y = 0 has intercept (0, 0)
and slope — 1. Graphing the equation y2 — x = 0,
Intercepts: If y = 0, x = 0; if x = 0, y — 0.
Symmetry: Symmetric with respect to the x-axis.
Extent: Solving for x, x = y2, so there is no limitation on the extent of y; solving for y,
y — T^/x, so x is limited to x > 0.
Asymptotes: None (as x + oo, y -*■ ± oo).

EXAMPLE

Graph the equation

x4 + x2 — y2 — y = 0

(see Figure 1.48).

x4 + x2 - y2 - y = (x2 + y)(x2 — y) + (x2 - y)

= (x2 - y)(x2 + y + 1) = 0

so the graph consists of the curves x2 — y = 0 and x2 + y + 1 = 0. Graphing the equation

x2 - y = 0,
Intercepts: If y = 0, x = 0; if x = 0, y = 0.
Symmetry: Symmetric with respect to the y-axis.
Extent: Solving for x, x = ±v^V’ so y is limited to y > 0. Solving for y, y = x2, so there is no
limitation on the extent of x.
Asymptotes: None (as x + oo, y -► f oo; as x ->• — oo, y -»• -f oo).
Graphing the equation x2 -I- y + 1=0,
Intercepts: If y = 0, x = y/ — 1 (imaginary); if x = 0, y = — 1.
Symmetry: Symmetric with respect to the y-axis.
Extent: Solving for x, x = +^J — y —so y is limited to y < — 1. Solving for y, y = — 1 — x2,
so there is no limitation on the extent of x.
Asymptotes: None (as x-> + oo,y-> + oo;asx-» — oo, y -foo).
 1.5 General Methods of Graphing Nonlinear Curves / 67

Figure 1.48

EXAMPLE

Graph the equation

x2y — x2 — 4y = 0

(see Figure 1.49).

Intercepts: If y = 0, x = 0; if x = 0, y = 0.
Symmetry: Symmetric with respect to the y-axis.
Extent: Solving for x,

y/(y — 1) is negative if 0 < y < 1 and is undefined if y = 1, so y is limited to the intervals y < 0
and y > 1. Solving for y, y = x2/(x2 — 4), so there is no limitation on the extent of x except y is
not defined for x = ± 2.
Asymptotes: x = ± 2 yjy/(y — 1). As y + oo, x-*■ ±2 and x> 121; as y —► — oo, x->±2
68 / Graphical Representation

Figure 1.49

and x < 121; so x = ±2 are asymptotes, y = x2/(x2 — 4). As x -► + oo, y -*■ 1 and y > 1; as
x —> — oo, y -> 1 and y > 1, so y = 1 is an asymptote.
Factorization: Not factorable.
Real, point, or imaginary locus: Real curve.

EXAMPLE

Graph the equation

x2y — x2 + 4y = 0

(see Figure 1.50).

Intercepts: If x = 0, y = 0; if y = 0, x = 0.
Symmetry: Symmetric with respect to the y-axis.
Extent: Solving for x,

y/( 1 — y) is negative if y < 0 or y > 1 and is undefined if y = 1, so y is limited to the interval

0 < y < 1. Solving for y, y = x2/(x2 + 4), so there is no limitation on the extent of x.

Figure 1.50

V = 1
 1.5 General Methods of Graphing Nonlinear Curves / 69

Asymptotes: x = ± 2y/y/(l - y). y is limited in extent and cannot become large or small
without bound, y = x2/(x2 + 4). As x -> + oo, y -*■ 1 and y < 1; as x -> — oo, y -► 1 and y < 1,
so y = 1 is an asymptote.
Factorization: Not factorable.
Real, point, or imaginary locus: Real curve.

EXAMPLE

Graph the equation

y2 — 4xy2 — 1=0

(see Figure 1.51).

Intercepts: If y = 0, x is not defined; if x = 0, y = ± 1.
Symmetry: Symmetric with respect to the x-axis.
Extent: Solving for x,

so there is no limitation on the extent of y, except that x is not defined for y = 0. Solving for y.

so x is limited to x <
Asymptotes: x = (y2 — l)/4y2. As y -> ± oo, x J, so x = J is an asymptote.

Figure 1.51

y
70 / Graphical Representation

x is limited in extent and cannot become large without bound. As x -► — oo, y -*■ 0, so y = 0 is an
asymptote.
Factorization: Not factorable.
Real, point, or imaginary locus: Real curve.

PROBLEMS
Sketch the curves represented by the following equations; give intercepts, extent, symmetry,
and asymptotes, as relevant.

1. x2y = 10 8. y = x3(x — l)(x + 6)

2. xy2 = -10 9. 4y = x3
3. y = x(x — 3)(x + 4) 10. y = x2(x - 3)2
4. y = x2(x2 — 4x + 4) 11. y = Xy/9 - x2
5. y = x4 — x2 12. y = x V'9 - x2
6. y = (x2 — l)(x2 — 4) 13. y = (x — 3)(x2 — 4x — 5)
7. y — x3 — 4x 14. y = x2(x — 6)(x2 — x — 6)

Answers to Odd-Numbered Problems

1. intercepts: none 9. intercepts: (0, 0)
extent: y > 0, x not limited extent: not limited
symmetry: about y-axis symmetry: about origin
asymptotes: x = 0, y = 0 asymptotes: none
3. intercepts: (0, 0), (3, 0), ( — 4, 0) 11. intercepts: (0, 0), (±3, 0)
extent: not limited extent: — 3 < x < 3, y not limited
symmetry: none except by range of x
asymptotes: none symmetry: about origin
5. intercepts: (0, 0), (± 1, 0) asymptotes: none
extent: not limited 13. intercepts: (0, 15), (3, 0), (5, 0),
symmetry: about y-axis (-1. 0)
asymptotes: none extent: not limited
7. intercepts: (0, 0), (±2, 0) symmetry: none
extent: not limited asymptotes: none
symmetry: about origin
asymptotes: none

1.6 Quadratic Curves

A quadratic equation may represent a circle, an ellipse, a parabola, a hyperbola, or
various degenerate or imaginary cases of these curves. The identification of a quadrat¬
ic equation as represencting one of these types of curves is discussed in this section;
the standard form for each type of equation is then considered for purposes of
graphical representation. Families of quadratic curves are also discussed.
An equation of the form

Ax2 -F Bxy 4- Cy2 + Dx + Ey + F = 0 (2)

where A, B, C, D, E, and F are constants and at least one of A, B, and C is nonzero, is

said to be a second-degree or quadratic equation. An equation is said to be quadratic
in a variable if that variable occurs to the second and no higher power. Equation (2)
is the general equation of second degree in two variables.
 1.6 Quadratic Curves / 71

Figure 1.52 Conic sections.

Since it is possible to remove the xy-term from any second-degree equation,

making B zero, the general quadratic equation actually can be written

Ax2 + Cy2 + Dx + Ey + F = 0

The removal of the xy-term is accomplished by changing from one set of axes to a
new set, which can be obtained by rotating the original axes about the origin. This
procedure will not be discussed in detail, because it is seldom necessary to use it in
practice.
Essentially, absence of the xy-term in a second-degree equation indicates that the
corresponding curve is symmetric to a line or lines (axes of the curve) parallel to one
or both of the coordinate axes. A circle, by definition, always has this type of sym¬
metry and B is always zero for a circle.
The curves represented by Equation (2) can be obtained by cutting a (double)
cone with a plane and thus are known as conic sections. The four special forms of a
conic section are the circle, the ellipse, the parabola, and the hyperbola (see Figure
1.52). In addition to these curves, Equation (2) may represent various types of degen¬
erate conics (actually special cases of conic sections). Degenerate conics may be
represented by two intersecting straight lines, two parallel straight lines, two coinci¬
dent straight lines (that is, one straight line), a point, or no points at all.
Any conic section can be defined as the locus of a point that moves in such a way
that the ratio of its distance from a fixed point (called its focus) to its distance from a
fixed line (called its directrix) is a constant e (called its eccentricity). Conic sections
can be classified into the four basic types according to the value of their eccentricity.
(For a circle, e = 0; for an ellipse, 0 < e < 1; for a parabola, e = 1; for a hyperbola,
e > 1.)
There are many interesting scientific applications of conic sections. For example,
the planets move in elliptic orbits about the sun with the sun at a focus; whispering
galleries usually have elliptical ceilings arranged so that a person standing at one
focus can hear a slight noise made at the other focus while someone standing between
the two foci hears nothing; elliptic arches are frequently used in architecture. The
cable of a suspension bridge hangs in the form of an arc of a parabola; a reflecting
surface made by rotating a parabola about its axis will send the light out in parallel
rays if the source of light is at the focus; the larger reflector in a reflecting telescope is
usually parabolic; the path of a projectile, neglecting air resistance, is a parabola. The
small reflector of a reflecting telescope is often a hyperbolic mirror (used to reflect the
image to the eyepiece); the hyperbola can be used to locate an invisible source of
72 / Graphical Representation

sound, such as an enemy’s guns, by considering two listening posts as the foci of a
hyperbola and performing certain computations on the difference in time of arrival of
the sound of the gun at the listening posts.
In economics and business, certain types of parabolic and hyperbolic curves are
appropriate for representing demand and supply functions, production functions,
and many other relationships. Applications of circular and elliptic curves are less
common but occur, for example, in representing product transformation functions.
The type of conic section represented by a second-degree equation can be
identified from an examination of its coefficients, whether or not the equation in¬
cludes an xy-term. However, only if the equation has no xy-term can it readily be put
into an appropriate standard form from which properties useful for graphing and for
other purposes can be obtained and from which degenerate cases can be identified.

Identification of a Quadratic Equation

The general quadratic equation

Ax2 + Bxy + Cy2 -(- Dx + Ey + F = 0

with at least one of A and C nonzero can be identified as a circle, ellipse, parabola, or
hyperbola as follows:

If B = 0 and A = C, the conic is a circle

If B2 — 4AC < 0, the conic is an ellipse
If B2 — 4AC = 0, the conic is a parabola
If B2 - 4AC > 0, the conic is a hyperbola

Direct identification of degenerate and imaginary loci is discussed in Technical Note

II.
For the special case B = 0, with at least one of A and C nonzero, the above
identification procedure can be reduced to the following:
If A = C, the conic is a circle
If A ^ C, but A and C have the same sign, the conic is an ellipse
If A = 0 or C = 0 but not both, the conic is a parabola
If A and C have opposite signs, the conic is a hyperbola

The standard forms for each of the four types of conic sections and their geometri¬
cal representations and properties are discussed in the following sections. The alge¬
braic procedure known as “ completing the square,” which is used to put equations
into these standard forms, is discussed in detail in Technical Note III.

THE CIRCLE
Geometrically, a circle is the locus of points in a plane which are a fixed distance
from a given point called the center. The distance of the points from the center is the
radius of the circle. The general equation of a circle can be written

Ax2 -F Ay2 + Dx + Ey + F = 0

(since A = C and B = 0 in the general quadratic equation). As shown in Technical

Note III, any equation that represents a circle can be written in the standard form
(x — h)2 -f (y — k)2 = r2

where (h, k) is the center of the circle and r is its radius, as shown in Figure 1.53.
 1.6 Quadratic Curves / 73

Figure 1.53

When an attempt is made to put an equation in the standard form for a circle,
degenerate or imaginary loci can be detected from the indicated value of r2:
If r2 < 0, there is no real locus (imaginary radius)
If r2 = 0, the locus is a point (zero radius)
If r2 > 0, the locus is a circle

EXAMPLE

Put the following equations into the appropriate standard form for a circle, check for degener¬
ate or imaginary loci, and graph.

(a) x2 -I- y2 — 4x — 6y -I- 19 = 0.

(b) 2x2 + 2yz + 16x — 4y 4- 17 = 0.
(c) x2 + y2 — lOx -I- Ay + 29 = 0.
Standard form for a circle:
(x — h)2 + (y — k)2 = r2
(a) (x2 — 4x + 4) -I- (y2 — 6y + 9) = —19 + 4 + 9

(x — 2)2 + (y — 3)2 = -6
r2 < 0, so no real locus.

(b) 2(x2 + 8x + 16) + 2(y2 — 2y + 1) = —17 + 32 + 2

(x + 4)2 + (y- 1)2=¥
Circle: center ( — 4, 1), radius y?- See Figure 1.54.

Figure 1.54

2x2 -f- 2y2 -j- 16x — 4y + 17 = 0

74 / Graphical Representation

Figure 1.55

(5, -2)

x2-j-y2 — lOx + 4y -+■ 29 = 0

(c) (x2 - lOx + 25) + (y2 + 4y + 4) = -29 + 25 + 4

(x - 5)2 + (y + 2)2 = 0

Point: (5, —2). See Figure 1.55.

THE ELLIPSE
Geometrically, an ellipse is the locus of points in a plane the sum of whose distances
from two fixed points (called foci) is constant. An ellipse has two (perpendicular) axes
of symmetry; the longer of these is the major axis, the shorter is the minor axis; the
point at which the axes intersect is the center of the ellipse (see Figure 1.56).
The general equation of an ellipse can be written
Ax2 + Cy2 + Dx + Ey + F = 0

(where A f C and A and C are of the same sign). As shown in Technical Note III,
any equation that represents an ellipse can be written in the standard form
x-h)2 (y-k)
7 b2

where (h, k) is the center of the ellipse and the major axis is parallel to the x-axis if
a> b and to the y-axis if a < b. (See Figure 1.57.)
The length of the axis parallel to the x-axis is 2a and the length of the axis parallel

Figure 1.56

y
 1.6 Quadratic Curves / 75

Figure 1.57

y y

(a) ellipse with major axis (b) ellipse with major axis
parallel to x-axis (a > b) parallel to r-axis(a < b)

to the y-axis is 2b. a and b are referred to as semiaxes. When a = b, the ellipse
becomes a circle with radius r = a = b.
When an equation is in the standard form for an ellipse, the locus is an ellipse.
However,

= C, where C < 0, there is no real locus

a

If (x-^ + (r-fe£
= 0, the locus is the point (h, k)
a

EXAMPLE

Put the following equations into the appropriate standard form for an ellipse, check for
degenerate or imaginary loci, and graph.

(a) 6x2 + 4y2 — 36x + 16y + 70 = 0.

(b) 9x2 + 2y2 + 36x + 4y + 20 = 0.
(c) 2x2 + y2 - 16x - 12y + 80 = 0.

Standard form for an ellipse:

(x - hf (y - kf
= 1
a2 + b2

(a) 6(x2 — 6x + 9) + 4(y2 + 4y + 4) = — 70 + 54 + 16

6(x - 3)2 + 4(y + 2)2 = 0

(x-3)2 (y + 2)2
"2T'“ + W

Point: (3, —2). See Figure 1.58.

76 / Graphical Representation

Figure 1.58
y

---x

(3, -2)
6x2 H- 4y2 — 36x -j-16+ 4- 70 = 0

(b) 9x2 +• 2 y2 + 36x + 4y + 20 = 0

9(x2 + 4x + 4) + 2(_y2 + 2y + 1) = —20 + 36 + 2

9(x + 2)2 + 2(y + l)2 = 18

(x + 2)2 (y + l)2
= 1
2 + 9
(x + 2)2 + (j, + l)2
1
(y /2 r 3 '

Ellipse: Center ( — 2, — 1), a = y/2, 6 = 3.

*

Major axis 6.
Minor axis 2+2.
Major axis parallel to y-axis.
See Figure 1.59.

(c) 2x2 + y2 — 16x — 12y + 80 = 0

2(x2 - 8x + 16) + (y2 - 12y + 36) = -80 + 32 + 36

2(x - 4)2 + (y - 6)2 = -12

(x - 4)2 (y - 6)2
+ = -1
12

(x — h)2 (y-k)2
a1 bz

so there is no real locus.

Figure 1.59

V
 1.6 Quadratic Curves / 77

PROBLEMS
Determine whether each of the following equations represents a circle or an ellipse, put the
equation into the appropriate standard form, check for degenerate or imaginary loci, and
graph.

1. x2 + y2 + 2x - 4y + 1 = 0 4. x2 + y2 - 8x - 4y + 18 = 0
2. 9x2 + 4y2 - 24y = 0 5. x2 + y2 - lOx + 25 = 0
3. x2 -I- 4y2 — 6x + 16y + 45 = 0 6. x2 + y2 — 2x + 4y -f 11 = 0

Answers to Odd-Numbered Problems

1. circle: (x + l)2 + (y — 2)2 = 22 5. circle: (x — 5)2 + (y — 0)2 = 0
center ( — 1, 2) degenerate locus: point (5, 0)
radius 2

3. ellipse:
(* - 3)2 . (y + 2f -5
4 1
so there is no real locus

THE PARABOLA
Geometrically, a parabola is the locus of points in a plane which are equidistant from
a fixed point (called the focus) and a fixed straight line (called the directrix). A
parabola is symmetric about a line, called its axis; the point of intersection of the
parabola and its axis is the vertex of the parabola (see Figure 1.60).
The general equation of a parabola can be written

Ax2 + Dx + Ey + F = 0

if the axis is parallel to the y-axis and

Cy2 + Dx + Ey + F = 0

if the axis is parallel to the x-axis. As shown in Technical Note III, any equation that
represents a parabola can be written in the standard form

(x - h f = 4p(y - k)
78 / Graphical Representation

Figure 1.61
y y

(d) (y — k)2 = 4p(x — h),p> 0

where (h, k) is the vertex of the parabola and the axis is parallel to the y-axis, or in the
standard form

(>’ - k)2 = 4p(x - h)

where (h, k) is the vertex of the parabola and the axis is parallel to the x-axis. The
orientation and curvature of a parabola are determined by the sign and magnitude of
p as follows:
For a parabola with axis parallel to the y-axis:

If p < 0, the parabola opens downward

If p > 0, the parabola opens upward

For a parabola with axis parallel to the x-axis:

If p < 0, the parabola opens to the left

If p > 0, the parabola opens to the right

(see Figure 1.61). The distance between the point (focus) and the line (directrix)
defining a parabola is 2p (distance from a point to a line is measured along the
perpendicular); as shown in Figure 1.62, the larger the value of p, the faster the
parabola opens.
 1.6 Quadratic Curves / 79

Figure 1.62

y y

When an equation is in either standard form for a parabola, the locus is a para¬
bola. However,

If (x — h)2 = C, where C < 0, there is no real locus

If (x — h)2 = 0, the locus is two coincident straight lines
(that is, geometrically, one straight line)
If (x — h)(x — k) = 0, the locus is two parallel lines

(and similarly for a parabola with axis parallel to the x-axis).

EXAMPLE

Put the following equations into the appropriate standard form for a parabola, check for
degenerate or imaginary loci, and graph.

(a) y2 - 8y + 17 = 0.
(b) x2 — 4x + y + 14 = 0.
(c) x2 - lx + 12 = 0.
(d) y2 - lOy + 25 = 0.
(e) y2 — 2y — 6x + 19 = 0.

Standard form for a parabola:

(y — k)2 — 4 p(x — h) or (x — h)2 = 4 p(y — k)

(a) y2 - Sy + 17 = 0

(y2 - 8y + 16) = -17 + 16

(y — 4)2 = -1

(y — k)2 < 0, so there is no real locus. Note that the x-term is missing in this example.

(b) x2 — 4x + y + 14 = 0

x2 — 4x + 4 = —y —14 + 4

(x — 2)2 = -(y+ 10)

Parabola: Vertex (2, —10), axis parallel to the y-axis, parabola opens downward.

See Figure 1.63. (Note that 4p = —1.)

Note the distinction between example (a), where the locus is imaginary, and example (b),
where the locus is real. In (a), the square of a number is said to be a negative constant, which is
80 / Graphical Representation

Figure 1.63

not possible for a real number. In (b), the square of a number is said to be the negative of a
quantity involving a variable, which implies that the quantity (y + 10) must be negative or
zero; y + 10 is negative or zero if y is limited in extent to the interval y < —10, as shown in
Figure 1.63.

(c) x2 - lx + 12 = 0

x2-7x + 4? = —12 + 4?

(x-j)2=i

x—
V- 7
1
_ IA
— ± a/3

x — ? ~ ±1

x=j or x — 4 = 0

and

x = -§ or x — 3 = 0

The locus is two parallel straight lines, parallel to the y-axis (see Figure 1.64). Note that the
y-term is missing in this example and the equation could have been factored (x — 4) x
(x - 3) = 0 without completing the square.

(d) y2 - lOy + 25 = 0

(y - 5)2 = o

Figure 1.64
 1.6 Quadratic Curves / 81

Figure 1.65

o
II
1

The locus is two coincident straight lines (see Figure 1.65). Note again that the x-term is
missing.

(e) y2 - 2y - 6x + 19 = 0

(y2 - 2y + 1) = 6x - 19 + 1

(y - l)2 = 6x - 18

(y - l)2 = 6(x - 3)

Figure 1.66

Parabola: Vertex (3, 1), axis parallel to the x-axis, parabola opens to the right (see Figure 1.66).

THE HYPERBOLA
Geometrically, a hyperbola is the locus of points in a plane the difference of whose
distances from two fixed points (called foci) is a constant. A hyperbola has two
(perpendicular) axes of symmetry; the axis that intersects the hyperbola is its trans¬
verse axis; the point at which the axes intersect is the center of the hyperbola. A
hyperbola always has two branches.
82 / Graphical Representation

The general equation of a hyperbola can be written

Ax2 + Cy2 -f Dx + Ey + F = 0

(where A and C are opposite in sign.) As shown in Technical Note III, any equation
that represents a hyperbola can be written in the standard form

{x-hf (y-kf
a2 b2

where (h, k) is the center of the hyperbola and the transverse axis is parallel to the
x-axis, or in the standard form

O' - k)2 (x ~h)2

b2 a2

where (h, k) is the center of the hyperbola and the transverse axis is parallel to the
y-axis. (See Figure 1.67.)
Every hyperbola has a pair of intersecting straight lines as asymptotes; these
asymptotes are given by the equations

Note that the equations for these asymptotes can be obtained from the standard form
by replacing 1 by 0, transposing the negative term, and taking square roots.
If a = h, the asymptotes of the hyperbola are perpendicular and the hyperbola is a
rectangular or equilateral hyperbola. Equilateral hyperbolas are discussed in more
detail below.
When an equation is in the standard form for a hyperbola, the locus is a hyper¬
bola. However,

Figure 1.67

y y

(b) OLlAI2 _ (* ~h)2

b2 a2
 1.6 Quadratic Curves / 83

(x — h* 2) (y - kf
If = 0, the locus is two intersecting straight lines,
a' b2

x — h y-k
= +
a b

Note that these are also the equations for the asymptotes of the hyperbola.
There is no possibility of an imaginary locus in the case of a hyperbola. If, for
example,

x — h)2 (y — k)2
-j-— = C where C < 0
a bz

then

(V - kf (x — h)2
^2
=c where C > 0
b2~

EXAMPLE

Put the following equations into the appropriate standard form for a hyperbola, check for
degenerate loci, and graph.

(a) 6x2 - 12x - 4y2 - 16y -34 = 0.

(b) 2y2 — 12y — x2 -F 6x + 7 = 0.
(c) 5x2 + 20x - 3y2 - 24y - 28 = 0.

Standard form for a hyperbola:

(x-hf (y-k)2 O' - kf (x - h)2

-= 1 ' b2 a2
(a) 6x2 — I2x - Ay2 — 16y — 34 = 0

(6x2 - 12x) - (4y2 + I6y) = 34

6(x2 — 2x + 1) — A(y2 + Ay + 4) = 34 + 6 — 16

6(x - l)2 - A(y + 2)2 = 24

(x ~ l)2 _ O' + 2)2

4 6

(x - l)2 (y + 2):
= 1
22 (V6);
Hyperbola: Center (1, — 2), a = 2, b = ^/6, transverse axis parallel to the x-axis.

. x — 1 y + 2
Asymptotes: —-— = ± —j=-
2 v/6

y + 2=±^6(x-1)

Plotting the asymptotes:

y « 1.23(x — 1) — 2 y « — 1.23(x — 1) — 2

If x = 0, y « —3.23 If x = 0, y « —0.77

If y = 0, x « 2.63 If y = 0, x « —0.63

(see Figure 1.68).

84 / Graphical Representation

Figure 1.68

6x2 - 12.v - 4y2 - \6y - 34 = 0

note: The symbol « means “approximately equal to” and is used because of the approxi¬
mation Of y/6.

(b) 2y2 — 12y — x2 + 6x + 7 = 0

2(y2 — 6y + 9) — (x2 - 6x + 9) = -7+18 — 9

2(y - 3)2 - (x - 3)2 = 2

(y - 3)2 (x - 3):
= 1
1 2

(y - 3)2 (X - 3)2
= 1
l2 (7+
Hyperbola: Center (3, 3); a = b = 1, transverse axis parallel to the y-axis.

1
Asymptotes: y — 3 = ± —t= (x - 3)
v^
Plotting the asymptotes:

y » 0.7l(x - 3) + 3 y * —0.71(x - 3) + 3

If x = 0, y % 0.87 If x = 0, y * 5.13

If y = 0, x * - 1.23 If y = 0, x % 7.23

See Figure 1.69.

(c) 5x2 + 20x - 3y2 - 24y - 28 = 0

5(x2 + 4x + 4) - 3(y2 + 8y + 16) = 28 + 20 - 48

5(x + 2)2 + 3 (y + 4)2 = 0

(x + 2)2 _ (y + 4)2 =
5
 1.6 Quadratic Curves / 85

Figure 1.69

The locus is two straight lines,

y= ±v/|(x + 2)-4

y « 1.29(x + 2) — 4 y « — 1.29(x + 2) — 4

If x = 0, y * - 1.42 If x = 0, y « -6.58

If y = 0, x « 1.10 If y = 0, x « —5.10

See Figure 1.70.

Figure 1.70

y
86 / Graphical Representation

Special Cases of the Equilateral Hyperbola

In the preceding discussion of the hyperbola, an equilateral hyperbola was defined as
a hyperbola whose asymptotes are perprpendicular to each other. A hyperbola
whose equation can be written in the standard form

(x - h)2 (y - k)1 (y- k)2 (x - h)2

-^-—z— =1 or —-j— = 1
az bz bz a

is an equilateral hyperbola if and only if a = b.

The equation of an equilateral hyperbola whose asymptotes are parallel, respec¬
tively, to the coordinate axes can be written in the standard form

(x — h)(y — k) = c

where (h, k) is the center of the hyperbola, about which it is symmetric, and x = h
and y = k are the asymptotes. (Note that this is the special case A = C = 0, B =/= 0, of
the general second-degree equation Ax2 + Bxy -I- Cy2 + Dx + Ey -l- F = 0.) If the
asymptotes are considered as coordinate axes, the hyperbola has its two branches in
the first and third quadrants if c > 0 and in the second and fourth quadrants if c < 0,
as shown in Figure 1.71. The intercepts of the hyperbola are the points (h — (c/k), 0)
and (0, k — (c/h)), where h — (c/k) and k — (c/h) may be positive or negative accord¬
ing to the values of h, k, and c. Note that this type of hyperbola has no more than one
x-intercept and one y-intercept; for a hyperbola to have two x-intercepts, its equa¬
tion must include an x2-term, and for a hyperbola to have two y-intercepts, its
equation must include a y2-term. Similarly, for a hyperbola to have one y-intercept,
its equation must include an x-term and for a hyperbola to have one y-intercept, its
equation must include a y-term.
The equation of an equilateral hyperbola whose asymptotes are the coordinate
axes can be written in the standard form

xy = c

[Note that this is the special case h = k = 0 of the equation (x — h)(y — k) = c.]

Figure 1.71

y y

(a) c > 0 (b) c < 0

 1.6 Quadratic Curves / 87

Figure 1.72

y y

This type of hyperbola has its center at the origin, about which it is symmetric; its
branches occur in the first and third quadrants if c > 0 and in the second and fourth
quadrants if c < 0. Since it is asymptotic to both coordinate axes, the hyperbola has
no intercepts (see Figure 1.72).
The equation xy — c expresses an inversely proportional relationship between x
and y—that is, as one variable increases, the other variable decreases proportionally.
The variable y is said to be inversely proportional to the variable x, if there is a
constant c such that

c
y=— or xy = c
x

By a generalization of this definition, the variable y is inversely proportional to a

positive power n of a variable x, if there is a constant c, such that

or xny = c

The curve corresponding to such an equation or to the analogous equation xym = c

is a generalized equilateral hyperbola, or hyperbola of Fermat. Its center is at the
origin; its asymptotes are the coordinate axes.
If n or m is odd, the branches of the equilateral hyperbola are symmetric to the
origin and are in the first and third quadrants if c > 0; the branches are in the second
and fourth quadrants if c < 0. If xny = c and n is even, the branches of the equilateral
hyperbola are symmetric to the y-axis and are in the first and second quadrants if
c > 0 and are in the third and fourth quadrants if c < 0. If xym = c and m is even, the
branches of the equilateral hyperbola are symmetric to the x-axis and are in the first
and fourth quadrants if c > 0 and are in the second and third quadrants if c < 0.

EXAMPLE

Graph the equation

(x - 4)(y + 12) = 2

(see Figure 1.73).

88 / Graphical Representation

Figure 1.73

Equilateral hyperbola: Center: (4, —12).

Asymptotes: x = 4, y = —12.
Lies in the first and third quadrants of its asymptotes.

EXAMPLE

Graph the equation

(x - 2)y = -4

(see Figure 1.74).

Equilateral hyperbola: Center: (2, 0).
Asymptotes: x = 2, y — 0.
Lies in the second and fourth quadrants of its asymptotes.

Figure 1.74
 1.6 Quadratic Curves / 89

EXAMPLE

Graph the equation

x3y = 16

(see Figure 1.75).

Generalized equilateral hyperbola, asymptotic to coordinate axes, symmetric to origin.

Points to aid graphing: (2, 2), (1, 16), (*/l6, 1), (-2), (-1, -16),
(— y/l6, — 1) (note that ^/l6 « |).

Figure 1.75
y

EXAMPLE

Graph the equation

xy2 = 25 (See Figure 1.76.)

Figure 1.76
90 / Graphical Representation

Generalized equilateral hyperbola, asymptotic to coordinate axes, symmetric to x-axis.

Points to aid graphing: (1, 5), (25, 1), (i 10), (1, -5), (25, -1), (i -10).

PROBLEMS
Determine whether each of the following equations represents a parabola or a hyperbola, put
the equation into the appropriate standard form, check for degenerate or imaginary loci, and
graph.

1. y1 2 3 - 2y - 2x + 9 = 0 4. y2 - 8y + 24 = 0
2. x2 - 3y2 - 4x + 12y - 11 = 0 5. xy - 4x - 5y + 5 = 0
3. 3x2 — 2y2 — 6x — 4y + 1 = 0 6. xy + 5x — y — 5 = 0

Answers to Odd-Numbered Problems

1. parabola: (y - l)2 = 2(x - 4) 5. equilateral hyperbola: (x — 5)(y — 4) = 15
vertex (4, 1) center (5, 4)
axis parallel to x-axis, opens to right asymptotes x = 5, y = 4
3. degenerate hyperbola: two intersecting first and third axes quadrants
lines 3x — 2y — 4 = 0 and 3x + 2y — 2 = 0

FAMILIES OF QUADRATIC CURVES

The general equation of the second degree

Ax2 + Bxy + Cy2 + Dx + Ey + F = 0

has five arbitrary constants, since if any of the constants A, B, C, D, £, and F is not
zero, it may be used as a divisor to reduce the equation to a form involving only five
arbitrary constants; after the values of five constants are set arbitrarily, the value of
the other constant must be set so that the sum of the terms is zero. These five
arbitrary constants in the equation of a second-degree curve permit five geometric
conditions to be imposed on the curve.
As discussed above, various conditions on the constants A, B, and C determine the
general form of the second-degree curve; complete specification of the constants
determines a particular curve. Families of various types of curves are represented by
an equation in which one (or more) of the constants is specified and the others are
unspecified (that is, remain parameters). For example, a family of circles may have
the same center (that is, be concentric) but have different radii, or a family of parab¬
olas may have the same vertex but have different degrees of curvature. Such families
of curves will be discussed in more detail in later sections.

PROBLEMS
For each of the following equations, identify the curve represented, put the equation into the
appropriate standard form and identify the parameters and properties thus obtained, and
sketch the curve.

1. x2 + y2 - 6x - 2y - 6 = 0 4. y2 - lOy = 0
2. y2 — 6y + 9 = 0 5. xy — 4y = — 4
3. 3x2 + 3y2 - 6x + 4y = 1 6. x2 - y2 + 4x - 2y + 1 = 0
 1.6 Quadratic Curves / 91

7. 2x2 + y2 = 50 17. y2 - 3x2 = 27

8. x2 + y2 — 4x — 2y + 5 = 0 18. 2x = 5y — y2
9. 4x2 + 9y2 - I6x - 18y + 133 = 0 19. 5x2 + 4y = 12
10. xy + 3y = x + 6 20. x^ + 15y + 3x = 15
11. 3x2 — y2 — 12x — 6y = 0 21. j;2 — 2^ — 8x + 25 = 0
12. x2 — y2 — 16 = 0 22. x2 + y2 — 4x — 2y + 6 = 0
13. y = 3 + 2x — x2 23. y2 - 4x2 - 4y + 4 = 0
14. 9x2 + 25y2 + 18x + 150y + 9 = 0 24. y2 - 12y + 46 = 0
15. x2 + 9y2 - 8x + 7 = 0 25. 3y2 + 2x = 0
16. 16x2 + y2 — 32x — 6y + 25 = 0 26. xy — 6x + 2 = 0

Answers to Odd-Numbered Problems

1. circle: (x — 3)2 + (y — l)2 = 42
center (3, 1)
radius 4

3. circle: (x - l)2 + (y +1)2 = (|)2

center (1, -|)
radius f

5. equilateral hyperbola: y(x — 4) = —4

center (4, 0)
asymptotes x = 4, y = 0
second and fourth axes quadrants

(X - o)2
7. ellipse: + (y - Q)2
52 (5s/2)2
center (0, 0)
semiaxes 5^2,5
major axis parallel to y-axis

>-2)2 , (y-1)
9. ellipse with no real locus: + = -3

(y + 3)2
11. hyperbola: —
(73)2
center (2, —3)
asymptotes y = J3 x - 2^3 - 3 and y = -y/3 x + 2^/3 - 3
transverse axis parallel to x-axis

13. parabola: (x — l)2 = — (y — 4)

vertex (1, 4)
axis parallel to y-axis, opens downward

i c ir (x ~ 4)2 (y - 0)2
15. ellipse:~ = 1

center (4, 0)
semiaxes 3, 1
major axis parallel to x-axis

17. hyperbola:
. (y - o); (X - o)2
(3+3): 32

center (0, 0)
asymptotes y = ^/3x and y = — y/3x
transverse axis parallel to y-axis
92 / Graphical Representation

19. parabola: (x — 0)2 = — |(y — 3)

vertex (0, 3)
axis parallel to y-axis, opens downward

21. parabola: (y — l)2 = 8(x — 3)

vertex (3, 1)
axis parallel to x-axis, opens to right

23. degenerate hyperbola: two intersecting lines: y = 2x + 2 and

y — — 2x + 2

25. parabola: (y — 0)2 = — ^(x — 0)

vertex (0, 0)
axis parallel to x-axis, opens to left

1.7 Applications of Quadratic Curves

in Business and Economics
In this section several applications of quadratic curves in business and economics are
discussed. These applications include quadratic demand and supply curves and the
associated market equilibria, product transformation curves, and Pareto’s law of the
distribution of incomes.

DEMAND AND SUPPLY CURVES

The first quadrant parts of various types of parabolas frequently are appropriate for
representing demand and supply functions, as illustrated in Figures 1.77 and 1.78.
The first quadrant part of an equilateral hyperbola is frequently used to represent a
demand function, as illustrated in Figure 1.79.
Note that each curve in Figures 1.77, 1.78, and 1.79 is only one of a family of
curves appropriate for representing the functions discussed. For example, the vertex
of the parabola in Figure 1.77(a) can lie anywhere in the second quadrant or on the

Figure 1.77 Examples of parabolic demand functions.

y y
 1.7 Applications of Quadratic Curves in Business and Economics / 93

Figure 1.78 Examples of parabolic supply functions.

y y

(b)

positive y-axis, provided the parabola has positive x- and ^-intercepts. The cond¬
itions on the coefficients of the equation which specify the family of curves appro¬
priate for representing a particular type of economic function can be determined
using the corresponding standard form of the equation (see the problems).

Figure 1.79 Example of a hyperbolic demand function.

Market Equilibrium
Market equilibrium price and quantity are the price and quantity represented by the
coordinates of the point of intersection of the demand and supply curves. An approx¬
imate solution for these coordinates can be obtained geometrically for any supply
and demand curves. However, algebraic simultaneous solution, even for second-
degree demand and supply functions, may involve solution of a third- or fourth-
degree equation.
Since general methods for solution of equations of degree greater than two are not
discussed here, the examples and problems are limited to cases involving solution of
94 / Graphical Representation

a quadratic equation. For example, if one equation is linear and the other equation is
parabolic or hyperbolic, their simultaneous solution involves solution of a quadratic
equation; similarly, simultaneous solution of two parabolic equations which are
quadratic with respect to the same variable involves solution of a quadratic equation.
Unless a quadratic equation can be factored, its solution is most readily obtained
by using the quadratic formula: The root(s) of a quadratic equation
ax2 + bx + c = 0 are given by

— b± yjb2 — 4ac
x =
2a
This formula is obtained by the process of completing the square, as shown in
Technical Note IV. Note that the number of real roots an equation has is determined
by the value of b2 — 4ac:

If b2 — 4ac < 0, no real roots

If b2 — 4ac = 0, one real root
If b2 — 4ac > 0, two real roots

EXAMPLE

Find the equilibrium price and quantity for the following demand and supply equations
(where x represents quantity and y represents price):

2x + y — 10 = 0

y2 — 8x — 4 = 0

Solving the equations simultaneously,

2x = — y + 10 from the first equation

8x = y2 — 4 from the second equation

so

y2 - 4 = —4y 4- 40

y2 + 4y — 44 = 0

and, using the quadratic formula,

-4±yi6^4)F44)
y 2
_ —4± 8^3
“ 2

= ~2± V3

x=i(-y+ 10)

= 6 ± 273

73 as 1.732, so the approximate solutions are (2.5, 4.9) and (9.5, —8.9) and the approximate
equilibrium point is (2.5, 4.9).
Plotting the curves to verify the result geometrically (see Figure 1.80),

2x 4- y — 10 = 0 y2 — 8x — 4 = 0

If x = 0, y = 10 y2 = 8x + 4

If y = 0, x = 5 y2 = 8(x + i)
 1.7 Applications of Quadratic Curves in Business and Economics / 95

Figure 1.80

EXAMPLE

Find the equilibrium price and quantity for the following demand and supply equations
(where x represents quantity and y represents price):

x2 + 5x — y + 1 = 0

2x2 + y — 9 = 0

Solving the equations simultaneously,

y = x2 + 5x + 1

y = — lx2 -f 9

so

x2 + 5x + 1 = — 2x2 4- 9

3x2 + 5x — 8 = 0

(3x + 8)(x — 1) = 0

x = -f, 1

y = — 2x2 + 9

The solutions are (— -§, —V) and (1, 7) and the equilibrium point is (i. n
Plotting the curves to verify the result geometrically (see Figure 1.81),

x2 + 5x — y + 1=0 2x2 + >’ — 9 = 0

(x2 + 5x + ^) = y — 1 + tt 2x2 = — (y — 9)

(x + i)2=y + ^ x2=~Uy~ 9)
96 / Graphical Representation

Figure 1.81

EXAMPLE

Find the equilibrium price and quantity for the following demand and supply equations
(where x represents quantity and y represents price):

y* 2 3 + y + x-20 = 0

2 y2 — x — 3y — 4 = 0

Solving the equations simultaneously,

x = — y2 — y + 20

x = 2 y2 — 3y — 4

so
— y2 — y + 20 = 2 y2 — 3y — 4

3y2 - 2y - 24 = 0

and, using the quadratic formula,

2±V4-4(3)( —24)
y 6
_ 2 ± 2v/73
6

= 1 ±/73
3

x = -y2 - y + 20

1 ± 2n/73 + 73 1 ± v/73
+ 20

77 ± 5^73
+ 20
 1.7 Applications of Quadratic Curves in Business and Economics / 97

Figure 1.82

y/Ti « 8.544, so the approximate solutions are (6.7, 3.2) and (16.2, —2.5), and the approxi¬
mate equilibrium point is (6.7, 3.2).
Plotting the curves to verify the result geometrically (see Figure 1.82),

y2 + y + x-20 = 0 2y2 — x — 3y — 4 = 0

y2 + y + i — — x + 20 + i 2 (y2 — iy + te) = x + 4 + |

(y + W= -(*-¥) (y - l)2 = i(x + ¥)

EXAMPLE

Find the market equilibrium price and quantity for the following demand and supply equa¬
tions (where x represents quantity and y represents price):

(x + 12)(y + 6) = 169

x — y 4- 6 = 0

Solving the equations simultaneously,

y = x + 6 from the second equation

Substituting in the first equation,

(x + 12)(x + 12) = 169

x + 12 = ±13

x = 1, -25

y=x + 6
y = 7,-19

The solutions are (1, 7) and ( — 25, —19) and the equilibrium point is (i. n
Plotting the curves to verify the result geometrically (see Figure 1.83),

x — y + 6 = 0 (x± 12)(y ± 6) = 169

If x = 0, y = 6 asymptotes x = —12, y = —6

If y = 0, x = — 6 If x = 0, y = fi

lfy = 0, x=^
98 / Graphical Representation

Figure 1.83

PRODUCT TRANSFORMATION CURVES

Some production processes can yield more than one output. Sheep raising is a classic
example of such a process: two outputs, wool and mutton, can be produced in
varying proportions by a single production process. Many industrial production
processes can also yield more than one output, for example, commodities that are
similar but of different type or quality. A production or product transformation
curve expresses the relationship between the quantities of two different commodities
(joint products) produced by the same firm using common supplies of labor and raw
materials. Note that the case of joint products is distinguished on a technical, not an
organizational basis; the quantities of two or more joint products are technically or
technologically interdependent. Cases in which a firm produces two or more tech¬
nically independent products are inappropriate for analysis by product transforma¬
tion curves.
A product transformation curve is defined as the locus of output quantity combina¬
tions which can be obtained from a given input. If the two output quantities
produced are x and y, the product transformation curve relating them must be such
that as one quantity increases the other quantity decreases. In addition, in order to
fulfill certain reasonable economic assumptions, which need not be discussed here,
product transformation curves are usually concave from below, as in Figure 1.84.
A product transformation curve is usually one of a family of possible product
transformation curves, where the curves of the family correspond to various inputs.
For example, Figure 1.84 shows product transformation curves represented by the
first-quadrant parts of four of a family of concentric circles; the farther a curve lies
from the origin, the greater the input to which it corresponds. Suitably located and
oriented elliptic, parabolic, and hyperbolic curves are also appropriate in some cases
for representing product transformation curves, as illustrated in Figures 1.85, 1.86,
 1.7 Applications of Quadratic Curves in Business and Economics / 99

Figure 1.84 Members of a family of cir¬

cular product transformation curves.

Figure 1.85 Example of an elliptic product transformation curve.

Figure 1.86 Examples of parabolic product transformation curves.

y y

(a) (b)
100 / Graphical Representation

Figure 1.87 Example of a hyperbolic product trans¬

formations curve.
y

and 1.87. Note that for circular, elliptic, and parabolic product transformation curves,
the entire first-quadrant part of the curve is appropriate. However, for hyperbolic pro¬
duct transformation curves, the range is limited to the lower branch of the hyperbola;
this limitation is usually stated as a restriction on the range of x:

c
0 < x < h —-
k

where (h — (c//c), 0) is the x-intercept.

EXAMPLE

A company produces amounts x and y of two different grades of steel using the same produc¬
tion process. The product transformation curve for the input used is given by

y2 + x + 4y — 20 = 0

(see Figure 1.88).

(a) What are the largest amounts x and y that can be produced?
(b) What amounts x and y should be produced in order to have x = 4y?

Figure 1.88

y
 1.7 Applications of Quadratic Curves in Business and Economics / 101

(a) x is as large as possible if y = 0, so the largest amount x is 20. y is as large as possible if

x = 0, so the largest amount y is

— 4±yi6~-(4)( —20)
y 2

_ -4 ± 4v//6
~ 2

= -2 ± 2^6

x/6 ^ 2.449, so y « 2.9, —6.9 and the largest amount y is (approximately) 2.9.
(b) Substituting x = Ay in y2 -f x + 4y — 20 = 0,

y2 + 4y + 4y — 20 = 0

y2 + 8y — 20 = 0

(y + 10)(j, - 2) = 0

y= -10,2

So the amounts produced are x = 8, y = 2.

EXAMPLE

A company produces amounts x and y of two different kinds of candy using the same produc¬
tion process. The product transformation curve for the input used is given by

5x2 -I- 2y2 = 98

(see Figure 1.89).

(a) What are the largest amounts x and y that can be produced?
(b) What amounts x and y should be produced in order to have y = |x?

(a) If y = 0,
5x2 = 98

■ X2 = IT

x = = = i^x/lO

Figure 1.89

y
102 / Graphical Representation

^10* 3.162, SO X ~ ± 4.4 and the largest amount x is (approximately) 4.4. If x = 0,

2 y2 = 98

y2 = 49

y = ±7

So the largest amount y is 7.

(b) Substituting y = |x in 5x2 + 2y2 = 98,

5x2 + 2(-^)x2 = 98

^x2 = 98

x2 = 16

x — ±4, y = ±3

So the amounts produced are x = 4, y = 3.

EXAMPLE

A company produces amounts x and y of two petrochemicals using the same production
process. The product transformation curve for the input used is given by

(x - 24){y - 36) = 240 x <^

(see Figure 1.90).

(a) What are the largest amounts x and y that can be produced?
(b) What amounts x and y should be produced to have x = iv?

(a) If y = 0,

x = — ^ + 24 = ^

Figure 1.90

y
 1.7 Applications of Quadratic Curves in Business and Economics / 103

So the largest amount x is If x = 0,

y = —10 + 36 = 26

So the largest amount y is 26.

(b) Substituting x = %y in (x — 24)(y — 36) = 240,

^ — 24)0--36) = 240

{y - 36)(y - 36) = 360

y -36= ±^360

y = 36 ± 6v//l0

y « 36 ± 19.0

55, 17

So the amounts produced are (approximately) x = y = 17. (Note that the solution x =
y = 55 violates the restriction x < Af and is on the upper branch of the hyperbola.)

EXAMPLE

A company produces amounts x and y of two different textiles using the same production
process. The product transformation curve for the input used is given by

(see Figure 1.91).

(a) What are the largest amounts x and y that can be produced?
(b) What amounts x and y should be produced to have x = yl

(a) If y = 0,

x2 = 100
x = ± 10
So the largest amount x is 10. If x = 0, y = 20, so the largest amount y is 20.

Figure 1.91

y
104 / Graphical Representation

-5 ± 5/17
2

-5 ± 5(4.123)

y « 7.8, -12.8

So the amounts produced are (approximately) x = y = 7.8.

PROBLEMS
For each of the following pairs of equations, i. determine which equation represents a demand
curve and which equation represents a supply curve, ii. determine the market equilibrium
quantity and price algebraically, and iii. check geometrically the algebraically determined
equilibrium points.

1. a. x = 16 - 2y 10. a. x = 84 - y2
b. 4x = 4y + y2 b. x = y + 4y2

2. a. x = 130 - 4y 11. a. x = lOy + 5y2

b. x = 64 — 8y — 2y

12. a. x = lOy + 4y2

b. x = 96 — 8y — 2y2

13. a. (x + 16)(y -f 12) = 480

b. y = 2x + 4

4. a. y = 16 — x 14. a. x = 2y2 — 2y — 6

b. y = 4 + x b. x = —y2 — y + 18

5. a. x = 32 — 4y — y 15. a. y = 10 - 3x2

b. y = 4 4- x2 + 2x

16. a. xy = 30
6. a. y = 9x + 12
b. 3y - x = 9
b. y = 39 — 3x2
17. a. xy = 15
x2
7. a. y = 6 + — b. y = x + 2
4
18. a. (x + 10)(y + 20) = 300
b. x = ^36 — y
b. x = 2y — 8
8. a. y = (x + 2)2
19. a. (x + 6)(y + 12) = 144
b. y = 39 - 3x2
i ^ x
9. a. y = 48 — 3x2
b- y =2 + 2
b. y = x2 + 4x 4- 16
 1.7 Applications of Quadratic Curves in Business and Economics / 105

20. a. (x + 12)(y + 6) = 169 25. a. (x + 4)(y -f- 2) — 24

b. x — y + 6 = 0 , . x
b.y=l+-
21. a. (x + 5)(j; + 6) = 80
26. a. y = x2 -I- 5x + 1
, x
b. y = - + 3
b. y + 2x2 — 9 = 0

22. a. (x + l)y = 5 27. a. x = 3y2 — 3y — 2

b. x = 10 — y2 — y

28. a. (x + 10)0; + 5) = 225

23. a. x(y + 6) = 24 , . n
' 7 b. x - y + 5 = 0
b. y — 2x + 4 = 0

24. a. y(x 4- 3) = 18

b. y — 3x + 6 = 0

Each of the following equations represents a product transformation curve for the amounts
x and y, respectively, of two related commodities; find the largest amounts x and y that can be
produced.

29. x = 36 — 6y2
30. y = 65 — 12x — 5x2
31. y = 45 — 9x2
32. x = 16 - 4y- 2y2
33. The production manager has decided that the marketing department can sell 126 units of
product daily, and he wants to produce that much. If he assumes that all factors other than
the number of employees and the resultant output remain constant within the range of this
total production, the production function can be expressed by the equation

2x2 + 4x — y = 0

with x representing the number of employees and y representing the units of production.
The production manager claims that he will need 7 men to produce the 126 units, a.
Assuming that the equation is appropriate, is the production manager correct in his
statement concerning the number of men needed? b. What type of curve does the equation
represent? Graph the curve, c. Draw up a schedule of units of production output per man
employed in the range 1 through 7 employees. Indicate the change in the number of units
produced through this range as each employee is added.
34. A company’s operations research director believes that the short-run average cost of
production can be expressed by the equation

x2 — 16x — y + 68 = 0

with x representing the number of units produced and y representing the average cost per
unit. He says that the average cost will be lowest when eight units are produced, a. Is his
statement correct? b. What type of curve is represented? Graph the curve, c. Draw up a
schedule of values of y for the range x = 4 through x = 12 and indicate the amount of
change in y for each change in x.
35. In national income analysis, the demand for a stock of money to be held, or the “ liquidity
preference,” as Keynes calls it, is often considered to be dependent on three motives: the
transactions motive, the precautionary motive, and the speculative motive. Assume that, at
a given level of national income, the effects of the transactions and precautionary motives
are constant. The speculative motive is considered to be a function of the interest rate as
expressed by the equation (x — l)y = 4, where x is the rate of interest (%) and y is the
demand for money to hold expressed in billions of dollars, a. What type of curve does the
equation express? Graph the curve, b. Draw up a schedule of values for y, the amount of
money demanded to be held (in billions of dollars) for the values of x from 2% through
106 / Graphical Representation

7%. What is the value of y when x = 100 (in billions of dollars)? c. Note and describe the
segment of the curve which represents the “ liquidity trap ”—that segment in which the
interest rate seems to lose its force as an effective factor in influencing the demand for
money to hold.
36. According to convention, in the economic analysis discussed in Problem 35, the dependent
variable (the demand for money to be held) is often assigned to the x-axis rather than the
y-axis. An equation used to express Keynesian ideas concerning the relationship between
the rate of interest and the money demanded to be hid is then x(y — 1) = 4. a. What type of
curve does this equation represent? Graph the curve, b. Draw up a schedule of values for
the interest rate y for the values of x from 1 through 7 (in billions of dollars). What is the
value of y when x = 100 (in billions of dollars)? c. Note and describe the segment of the
curve that represents the “ liquidity trap.”
37. Consider the parabola y = ix2 — x + 4 and the equilateral hyperbola (x + 2)(y - 2) = 4.
a. Show that for x = 0 and x — 2 both equations have the same value of y, but for x = 4
and x — —2 they have different values of y. b. Show that the two equations have the same
value of y only for x = 0 and x = 2. c. Sketch the two curves on the same set of axes.
38. A company produces x and y amounts of steel of two different grades, using the same
resources; the product transformation curve is

^20-3<^ <*<3°>
a. Sketch the curve, b. Determine the largest amounts of x and y that can be produced, c. If
the demand for grade x steel is twice that for grade y steel, determine the amounts the
company should produce.
39. A company manufactures two grades of candy from the same resources. If x and y
represent the quantities produced, the product transformation curve is

(x - 24)(y - 36) = 240 (x < 24)

a. Sketch the curve, b. Determine the largest amounts of x and y that can be produced, c. If
the demand for grade x candy is two-thirds that for grade y, determine the amounts the
company should produce.
40. A company manufactures two grades of paper from the same resources; if x and y repre¬
sent the quantities produced, the product transformation curve is
(x - 30)(y - 15) = 150 (x < 30)

a. Sketch the curve, b. If the demand for grade x paper is three times that for grade y paper,
determine the amounts of paper the company should produce, c. If the demand for grade y
paper exceeds that for grade x paper by four units, determine the amounts the company
should produce.

Answers to Odd-Numbered Problems

1. a. demand 11. a. supply
b supply b. demand
(8, 4) (40, 2)
3. a. supply 13. a. demand
b. demand b. supply
(6.91, 5.77) (4, 12)
5. a. demand 15. a. demand
b. supply b. supply
(20, 2) (1, 7)
7. a. supply 17. a. demand
b. demand b. supply
(4.90, 12) (3, 5)
9. a. demand 19. a. demand
b. supply b. supply
(2.37, 31.15) (3.22, 3.61)
 1.7 Applications of Quadratic Curves in Business and Economics / 107

21. a. demand 27. a. supply

b. supply b. demand
(3, 4)' (4,2)
23. a. demand 2Q v =36
b. supply >’max =

(3, 2) 31. xmax

25. a. demand .Vmax 45

b. supply
(2, 2)
33. a. yes ,
b. parabola, vertex (— 1, — 2), axis parallel to y-axis

Number of Total
employees units produced
X y Ay

1 6
2 16 10
3 30 14
4 48 18
5 70 22
6 96 26
7 126 30

35. a. equilateral hyperbola, transverse axis parallel to y-axis

center (1, 0)
lies in first and third quadrants of its asymptotes
asymptotes: x = 1, y = 0
intercept: (0, -4)

Interest Amount of money

rate demanded to be held
X y
2 4
3 2
4
4 3
5 1
4
6 3
2
7 3
4
100 ^5

c. segment for which x > 2

39. b. xmax = 1 ?i ymax = 26
c. x= 11.36, y= 17.04

PARETO S LAW OF DISTRIBUTION OF INCOME

The economist Vilfredo Pareto proposed the following law of distribution of income:
The number of individuals N from a given population of size a whose income exceeds
x is
108 / Graphical Representation

Figure 1.92

where b is a population parameter, usually approximately 1.5. This equation repre¬

sents a generalized equilateral hyperbola and is appropriate only for the range
0 < N < a and 0 < x < maximum income in the population (see Figure 1.92).
Note that Pareto’s equation implies that both N and x are continuous; in fact,
although x might be considered approximately continuous, N is by definition
discrete. For convenience, relationships involving discrete variables are frequently
represented as continuous; the discrepancy between the discrete nature of a variable
and its continuous representation must be considered in interpeting analyses based
on such representations.
For above-subsistence incomes, data indicate that Pareto’s law is generally fairly
accurate. Pareto suggested that the value of b is approximately 1.5; data indicate that
this varies from population to population, but 1.5 is usually a good approximation.
note: The notation used in the following examples is reviewed in the next
section, which deals with exponential and logarithmic functions. Note, however, that
Pareto’s law is represented by a power function, not by an exponential function.

EXAMPLE

Pareto’s law for the distribution of incomes for a particular group is

216 x 1010
N= .3/2

(a) How many people are millionaires?

(b) How many people have incomes between $3600 and $10,000?
(c) What is the lowest income of the 80 people with the highest incomes?

216 x 1010
= 2160 millionaires
(106)3/2

(b) The number with incomes exceeding $3600 is

216 x 1010 _ 216 x 1010

(3600)3/2 ~ 63 x 103

The number with incomes exceeding $10,000 is

_ 216 x 1010
= 216 x 104
(104)3/2
 1.7 Applications of Quadratic Curves in Business and Economics / 109

so the number with incomes between $3600 and $10,000 is 107 — 216 x 104 = (1000 — 216) x
104 = 784 x 104 or 7,840,000.

216 x 1010
(c) 80 = -3/2

216 x 1010
x-3/2
1 =--= 27 x 109
80
x = (27 x 109)2/3 = 9 x 106

or $9,000,000 is the lowest income of the 80 people with the highest incomes.

EXAMPLE

Pareto’s law for the distribution of incomes for a given population is

16 x 1012
N= .5/3

(a) How many people have incomes below $8000?

(b) How many people have incomes over $125,000 but less than $1,000,000?
(c) What is the lowest income of the 50 people with the highest incomes?

, , 16 x 1012 16 x 1012 c
(a) N"(8x 103)5'3 “ 25 x 105 “ 5 x 10

so 5 x 106 people have incomes above $8000 and 16 x 1012 — (5 x 106) = (16 x 106 — 5)
x 106 or 15,999,995,000,000 have incomes below $8000.
(b) The number with incomes over $125,000 is

16 x 1012 _ 16 x 1012 _ 16 x 107

~ (53 x 103)5/3 “ 55 x 105 “ 55
= 16 x 25 x 102 = 512 x 102 or 51,200

The number with incomes over $1,000,000 is

16 x 1012
N= = 16 x 102 or 1600
(106)5/3

So the number with incomes over $125,000 but less than $1,000,000 is 51,200 - 1600 = 49,600.

16 x 1012
(c) 50 =
x5/3

=
16 x 101"
x5/3
= 32 x 10 10
50

x = (32 x 10lo)3/5 = 8 x 106

or $8,000,000 is the lowest income of the 50 people with the highest incomes.

PROBLEMS
1. Pareto’s law for the distribution of incomes for a particular group is

„ 8 x 108

a. How many have incomes exceeding $1600?

b. How many have incomes between $1600 and $3600?
c. What is the lowest income of the 800 who have the highest incomes?
110 / Graphical Representation

2. Pareto’s law for the distribution of incomes for a particular group is

„ 1.9 x 1012

a. How many have incomes exceeding $50,000?

b. How many have incomes between $25,000 and $50,000?
c. What is the lowest income of the million who have the highest incomes
3. Pareto’s law for the distribution of incomes for a particular group is

XT 100,000
N =-2—
x

a. How many have incomes exceeding $15?

b. How many have incomes between $50 and $75?
c. What is the lowest income of the 5 who have the highest incomes?
4. Pareto’s law for the distribution of incomes for a particular group is

„ 32 x 1010

a. How many have incomes between $125,000 and $1,000,000?

b. What is the lowest income of the 200 who have the highest incomes?
5. Pareto’s law for the distribution of incomes for a particular group is

xr 625 x 109
N ~ X3'2

a. How many have incomes between $2500 and $10,000?

b. What is the lowest income of the 5000 who have the highest incomes?
6. Pareto’s law for the distribution of incomes for a particular group is

a. How many have incomes exceeding $2500?

b. How many have incomes between $2500 and $10,000?
c. What is the lowest income of the 6 who have the highest incomes?
7. Pareto’s law for the distribution of incomes is

a. How many have incomes exceeding /c?

b. How many have incomes exceeding $100,000?
c. What is the lowest income of the 10 who have the highest incomes?
d. How many have incomes between s and f?
e. How many have incomes between $500,000 and $1,500,000?
8. If the demand function is

17.6

(where x is quantity demanded and y is price):

a. Find the price if the quantity demanded is 2.
b. Find the quantity demanded if the price is $5.
9. If the demand function is

12.03

(where x is quantity demanded and y is price):

Find the price if the quantity demanded is a. 6, b. 3.3.
Find the quantity demanded if the price is c. $100, d. $67.
 1.8 Exponential and Logarithmic Curves / 111

10. If the demand function is

(where x is quantity demanded and y is price):

Find the price if the quantity demanded is a. 5a, b. 1.
Find the quantity demanded if the price is c. $1, d. a.

Answers to Odd-Numbered Problems

1 lb
1. a. 12,500 a
C.

b. 8796
To,
a(tb - sb)
c. $10,000 d.
sbtb
3. a. 444
e. a(lb ~ 1)
b. 22 (15 x 105)b
c. $141 9. a. $27,455
5. a. 4,375,000 b. $473.21

b. $250,000 c. 4.574

a d. 4.975
7. a. —r
kb
a
b.
10 5b

1.8 Exponential and Logarithmic Curves

Exponential and logarithmic curves are closely related. Graphical representations of
exponential and logarithmic functions are discussed in this section.

EXPONENTIAL FUNCTIONS
A function having a variable base and a constant exponent is a power function. For
example, y = xa is a power function, where x is the base and a is the exponent. Many
examples of power functions have been discussed in previous sections. A function
having a constant base and a variable exponent is an exponential function. For
example, y = ax is an exponential function, where a is the base and x is the exponent.
The following rules of exponents are applicable to both power functions and
exponential functions.
If x, y, and z are positive real numbers,

Rule Example

XmXn= xm+n io2 X io3 = 105

2. (24)6 = 224

3. (xyz)m = xmymzm (2 x 4 x 5)6 = 26 x 46 x 5'

m
3\44 = 34
4.
( 8/ 84
112 / Graphical Representation

5. x~m = m
x
m 1
x Ym-n _
6.
X' x"“m

7. xl,n = j/x, where n is a positive integer, 7i/3 = y?

read the nth root of x

8. xm,n = ^/x” = (</x)m, where m and n are y95 = (V9)s

integers, n > 0, read the mth power of
the nth root of x.

note: By definition, if x fi 0, x° = 1; that is, any base to the zero power is 1.

One very convenient use of exponents is in the representation of numbers which

are very large or very small in absolute value, by expressing them as a constant
multiplied by an exponent of 10. For example, 80,000 can be written 8 x 104;
5,000,000 can be written 5 x 106. Similarly, the fraction 0.002 can be written
2 x 10“3; the fraction 0.00000000096 can be written 9.6 x 10“10. This representa¬
tion of numbers is referred to as scientific notation; the significant digits in the
number are conventionally expressed with one digit to the left of the decimal point.
This notation was used in the examples of Pareto’s law, but the convention concern¬
ing decimal points was violated for convenience of calculation.
The simplest exponential functions are of the form y = bx, b > 0. The curve re¬
presenting the function y = bx lies entirely in the first two quadrants; it is monoton-
ically decreasing if 0 < b < 1 and monotonically increasing if b > 1; in both cases the
curve is asymptotic to the x-axis and has y-intercept (0, 1). The parameter b deter¬
mines the curvature of the function (see Figure 1.93). Note that if b = 1, y = bx
becomes y = 1, a straight line parallel to the x-axis.
The most frequently used exponential function is y — ex, where e is the base of the
natural logarithms, approximately 2.718. Almost all the exponential functions used
in economic theory have the base e. The precise definition of this base and its
advantages are discussed in connection with logarithms in the following sections.

Figure 1.93

y y

(a) 0<6< i (b) b> 1

Figure 1.94 y y

(c) a > 0, k > 0, c < 0, |c| < a (d) a > 0, k < 0, c < 0, \c\ < a

(if \c\ > a, intercepts negative) (if \c\ > a, intercepts negative)

y y

(if c < |a|, intercepts negative) (if c <\a|, intercepts negative)

(g) a < 0, k > 0, c < 0 (h) a < 0, k < 0, c < 0

114 / Graphical Representation

A more general form for exponential functions is

y = aekx + c

The curve representing the function y = aekx + c is asymptotic to the line y = c; it

approaches its asymptote from above if a > 0 and from below if a < 0. The curve is
monotonically increasing if a and k are of the same sign and monotonically decreas¬
ing if a and k are of opposite signs. The curve has y-intercept (0, a + c) and, if c < 0,
x-intercept ((1/A:) In \c/a\,0); see Figure 1.94. (The x-intercept is obtained in Techni¬
cal Note V.)

LOGARITHMIC FUNCTIONS
In 1614, John Napier wrote the first treatise on logarithms. He based his discus¬
sion on the comparison of two sets of numbers—one in arithmetic progression and
one in geometric progression. For example, consider the following two progressions:

Arithmetic: 0 1 2 3 4 5 6...
Geometric: 1 3 9 27 81 243 729...

If the numbers of the arithmetic progression are considered as exponents (powers) of

3, the corresponding numbers of the geometric progression result from raising 3 to
the indicated power. Thus
3° = 1
31 = 3

32 = 9
33 = 27

and so forth. Furthermore, the two progressions indicate that in order to multiply, it
is necessary only to add exponents: 32 x 33 = 32 + 3 = 35. And to divide, it is neces¬
sary only to subtract exponents: 35 -f- 32 = 35-2 = 33. This idea is the basis for the
computational short cuts possible using logarithms.
By definition, the logarithm of a positive number y to a positive base b, other than
1, is the exponent x to which the base must be raised to obtain the number. Thus, if
b > 0, b ± 1, and b, x, and y are related by the equation

y = bx (3)
then x, the exponent of b, is the logarithm of y to the base b. This relationship may
also be expressed by the equation

* = loSb y (4)
Equations (3) and (4) express the same relationship between b, x, and y in expon¬
ential and logarithmic forms, respectively, and are inverse functions—that is, if y is an
exponential function of x, then x is a logarithmic function of y.
Although the base of a logarithm can be any positive number other than 1, in
practice the base is almost always either 10 (common or Briggsian logarithms) or
e ~ 2.718 (natural or Naperian logarithms). The base e is defined in terms of limits as
follows:
n

e = lim
n-> oo

The concept of limits is discussed in detail later; for present purposes it is sufficient to
 1.8 Exponential and Logarithmic Curves / 115

note that e can be approximated as accurately as desired by increasing n in the

expression (1 + 1 /n)n.
In some cases, common logarithms are more convenient for computational work;
natural logarithms are more convenient for theoretical work. By convention, log x
denotes the common logarithm of x and In x denotes the natural logarithm of x. If
any other base is meant, it is specified.
The use of logarithms can save considerable work in computations involving
numbers that are very large or very small in absolute value. The essential properties
of logarithms are summarized in the following rules:
If x and y are positive real numbers,

Rule Example

1. log* (xy) = log*x + log*y log4 10 = log4 5 + log4 2

2- log* j = log* x - log* y logs (I) = logs 5 - logs 6

3. log* x” = n log* x log532 = 2 log5 3

1
4. log* x — — log* x log9 ^2 = i log9 2
n

5. lo&x = lo&Xlog,,*) logio 14 = (log102)(log2 14)

1 1
log6x log2 14
log6 a log2 10,
The simplest logarithmic functions, mentioned above, are of the form

y = log* x, b > 0 and b ± 1

The curve representing the function y — log* x lies entirely in the first and fourth
quadrants; it is monotonically decreasing if 0 < b < 1 and monotonically increasing
if b > 1; in both cases the curve is asymptotic to the y-axis and has x-intercept (1, 0).
The parameter b determines the curvature of the function (see Figure 1.95).

Figure 1.95

(a) 0<b< 1 (b) b> 1

116 / Graphical Representation

Figure 1.96
y

Figure 1.97

y y

(c) a > o, b < o (d) A <0, B<0

 1.9 Applications of Exponential and Logarithmic Curves / 117

Since y = bx and y = logb x are inverse functions, the graph of either one of these
families of curves (or of one particular curve for some specified value of b) can be
obtained from the graph of the other geometrically by reflection about the line y — x.
(Recall that reflection of a curve about a line consists of obtaining the curve that is
symmetric to the given curve with respect to the line; see Figure 1.96.)
A more general form for logarithmic functions is

y = ,4 1n(l + x) + # x> —1

The curve representing the function y = A In (1 + x) -I- B lies to the right of and is
asymptotic to the line x = — 1; the curve is monotonically increasing if A > 0 and
monotonically decreasing if A < 0. It has x-intercept (e~B/A — 1, 0) and y-intercept
(0, B); see Figure 1.97. (The x-intercept is obtained in Technical Note V.)

PROBLEMS
1. For what ranges of x are the following curves defined?
a. y = In (x + 1) c. y = In (x3 + 27)
b. y — In (x — 4) d. y — In (x4 — 16)
2. Sketch on the same set of axes the curves represented by the equations
a. y = 2X c. y = 2x/1
b. y = 2~x d. y = 2lx
3. Sketch on the same set of axes the curves represented by y = ax for a = 2, e, 3, and 10.
4. Sketch on the same set of axes the curves represented by y = logaX for a — 2, e, 3, and 10.
5. Sketch on the same set of axes the curves represented by the equations
a. y = 3* b. y = log3x
6. For what ranges of x are the following curves defined?
a. v = i In (1 — x2) d. y = —3 — 6 In (x — 2)
b. y = In (36 — x2) e. y = In (x — 27)
c. y — In (x2 — 36)

Answers to Odd-Numbered Problems

1. a. x > — 1 c. x > — 3
b. x > 4 d. x < — 2 and x > 2

1.9 Applications of Exponential and Logarithmic

Curves in Business and Economics

As can be noted in Figures 1.94 and 1.97, several types of exponential and logarith¬
mic functions are appropriate for representing demand or supply curves. Exponential
cost functions, representing total cost as an exponential function of number of units
produced, are also frequently used. Many examples of these uses of exponential and
logarithmic curves appear in later sections. The applications discussed in this section
are compound interest and various types of growth functions, problems for which
exponential and logarithmic functions are uniquely appropriate. A table of exponen¬
tial functions and a table of logarithms are provided at the end of the book.
118 / Graphical Representation

COMPOUND INTEREST
If the interest rate is 100* percent per year payable (that is, compounded) k times per
year, an amount of money x (the principal) becomes after n years

y = x(l + i)

(when k is large, y « xein, where e « 2.718 is the base of the natural logarithms).
This formula can be developed as follows: If the interest at rate 100* percent is
payable yearly, then the amount (principal plus interest) at the end of the first year
is

yt = x + ix — x(l + *)

The amount y2 after 2 years is

y2 = [x(l + /)][ 1 + i] = x(l + i)2

The amount y3 after 3 years is

y3 = [*0 + 02][i + 0 = *(i + if

The amount yn after n years is

y„ = x(\ + if

If interest is payable k times per year, then the rate of interest in any one period is
100(i/k) percent and the number of periods is nk. Thus the amount after n years is

EXAMPLE

A man deposits $5000 at 4% interest. How much has he (principal plus interest) after 10 years
(a) if interest is payable yearly, and (b) if interest is payable quarterly?

(a) y = x(l + i)"

= 5000(1 + 0.04)10

log y = log 5000 + 10 log 1.04

= 3.6990 + (10)(0.0170)

= 3.8690

y = $7396.67
nk
y = x 1 +

40
0.04
= 5000 1 +

log y = log 5000 + 40 log 1.01

= 3.6990 + 40(0.0043)

= 3.8710

y = $7430.00
 1.9 Applications of Exponential and Logarithmic Curves / 119

GROWTH FUNCTIONS
There are many relationships in business and economics which are appropriately
represented by curves referred to as growth functions, for example, number of em¬
ployees as a function of annual sales of a company, amount of finished stock as a
function of days after beginning a production run, sales as a function of advertising
expenditure, maintenance cost as a function of number of hours a machine is run,
and sales as a function of length of time a product has been on the market can be
represented by growth functions.
The basic property of the various curves referred to as growth functions is that
they are monotonically increasing. Growth functions may or may not have an upper
asymptote (although usually for business and economic applications absence of an
upper asymptote is reasonable only when relatively short periods of time are con¬
sidered) and they may be of various shapes. Three particular types of growth func¬
tions are discussed below: (1) very simple functions originally developed to describe
certain types of biological growth and appropriate only for growth that does not
have an upper asymptote, (2) Gompertz functions used to describe growth that starts
rather slowly and approaches an upper asymptote, and (3) learning functions orig¬
inally used by psychologists to describe human learning and appropriate for growth
that begins rapidly, levels off, and approaches an asymptote.

Biological Growth Curves

Many laws of biological growth are represented by the equation

N = N0Rt

where N is the number of individuals in the population at time t, N0 is the initial

number of individuals in the population (at time zero), and R > 0 is the rate of
growth (see Figure 1.98). This equation is based on the model of a population each of
whose members produces R — 1 additional members in each unit of time and none
of whose members die. (This situation is approximated for certain laboratory cul¬
tures, at least for limited periods of time.) The equation can be developed as follows:

Figure 1.98

N
120 / Graphical Representation

Time Number in Population

0 N0
1 N0 + N0(R- 1) = N0R
2 N0R + N0R(R - 1) = N0R2
3 N0R2 + N0R2(R - 1) = N0R3

t N0R{~1 + N0R‘-l{R - l) = N0Rt

Organization theorists have claimed that the function N = N0R* is appropriate
for describing the early growth of a rapidly developing company; because it has no
upper asymptote, the function is inappropriate for describing biological or organiza¬
tional growth for an indefinite period of time. In addition, the behavior of the curve,
as illustrated in the example below, indicates that it should be used with extreme
caution in theoretical analyses.

EXAMPLE

The National Aerospace Research Organization is beginning its operations with a staff of 5
men. At the end of each year of its operation, each employee will hire 3 assistants. How many
employees will National Aerospace Research Organization have after 10 years of operation?

N = N0R{

= (5)(410)
= 5,242,880 employees after 10 years

Gompertz Curves
Gompertz curves, named for their originator, are represented by the equation

N = caR'

where N is the number of individuals in the population at time t,R(0<R < 1) is the
rate of growth, a is the proportion of initial growth, and c is the growth at maturity
(that is, the upper asymptote). Note that when t = 0, N = ca, which corresponds to
N0 of the biological growth function.
The two basic types of Gompertz curves are characterized as follows and il¬
lustrated in Figure 1.99:

Type I: 0 < a <-

e

Type II: -< a < 1

e

Curves of Type I are positively accelerated for small positive t and negatively ac¬
celerated for large positive t; curves of type II are negatively accelerated for all
positive t. The mathematical definitions of positive and negative acceleration are
discussed later; essentially, a positively accelerated curve is increasing at an increas¬
ing rate and a negatively accelerated curve is increasing at a decreasing rate.
Gompertz curves have been used extensively by psychologists to describe various
aspects of human growth and development, including some types of learning. Organ¬
ization theorists have found Gompertz curves appropriate for describing the growth
 1.9 Applications of Exponential and Logarithmic Curves / 121

Figure 1.99

of many types of organizations. These curves are also appropriate for many other
functions in business and economics, for example, total revenue and production
functions.

EXAMPLE

On the basis of expected sales and data for similar companies, the Personnel Director of
National Industries predicts that the number of employees can be described by the equation

N = 200(0.04f5'

where N is the number of employees after t years. Assuming he is correct, how many em¬
ployees will National Industries have after 3 years? How many employees did the company
employ initially? How many will it employ when it reaches its maximum size?
The company employs (200)(0.04) = 8 people initially and 200 at maximum size. After 3
years it employs

N = (200)(0.04)°53

log N = log 200 + 0.53 log 0.04

= 2.3010 + (0.125)(-1.3979)

= 2.1263

N = 133.75 or approximately 134 people

Learning Curves
Because of their extensive use by psychologists to describe learning, exponential
curves of the form

y = c — ae~kx

where c, a, and k are positive, are frequently referred to as learning curves. [Note that
this is case (f) of the general form y = aekx + c discussed previously.] In his reinforce¬
ment theory of learning, Clark Hull used the special case c — a of this function as one
of the basic equations for describing the relationship between strength of learning y
and number of reinforcements x.
122 / Graphical Representation

The function y = c — ae~kx rises rapidly at first, flattens out, and then approaches
its asymptote y — c (see Figure 1.100). These curves have been found to be appro¬
priate for representing various cost and production functions.

EXAMPLE

If the number of items y manufactured per day x days after the beginning of a production run
is given by

y = 200(1 -e-°Ax)

how many items are manufactured per day 10 days after beginning the run and what per cent
of the maximum is this?

y = 200(1 - e-1)

= 200(1 -0.368)

= 200(0.632)

= 126.4 or approximately 126 items, 63.2% of the 200 maximum.

EXAMPLE

If the yearly maintenance cost y of a computer is related to its average monthly use x (in
hundreds of hours) by the equation

y = 35,000 - 25,000e-oo2x

what is the yearly maintenance cost for 200 hours average monthly use?

y 35,000 - 25,000c-0'04

= 35,000 - 25,000(0.135)

35,000 - 3375

= $31,625 yearly maintenance

 1.10 Trigonometric Curves / 123

PROBLEMS
1. Chem Corporation has $10,000 to deposit and expects to leave the deposit for 20 years.
Two options are available: 5% interest payable semiannually and 4\% interest payable
quarterly. Which option should Chem Corporation choose?
2. In order to have $20,000 after 20 years at 6% interest payable yearly, how much must be
deposited?
3. The National Professional Society has been formed with 10 original members. The rules
state that each member may invite 2 people to join at the beginning of each year. If each
member always takes advantage of this rule, how many members will the society have after
15 years?
4. Monthly total revenue (in dollars) for a particular firm can be described by the equation

R = 1000(0.10)° 8p

where p is the amount spent for promotion and advertising. What is total revenue when
nothing is spent for promotion and advertising? What is the maximum attainable total
revenue? What is total revenue if $20 is spent for promotion and advertising?
5. Production costs (in hundreds of dollars) for a firm are described by the equation

C = 100 - 10e~O O2x

where x is the number of units of output. What are the firm’s fixed costs? When output is
100 units, what proportion of the production costs are fixed?
6. One job on a production line consists of screwing a small screw into a metal plate. For a
typical employee, the number of plates completed per hour is described by the equation

y = 50 — 40c~°‘30jc

where x is number of hours the employee has worked on the production line. a. How many
plates can an employee complete the first hour? b. How many the sixth hour?
7. The number of firms in a particular industry is described by the equation

N = 6(0.5)°75'

where t is the number of years since the industry began. How many firms were there in the
industry after 5 years? How many firms were in the industry originally? How many firms
will there be when the industry reaches maximum size?
8. In order to have $10,000 after $6,000 has been deposited for 25 years with interest payable
yearly, what interest rate must be paid?
9. For a 10-year deposit, what interest rate payable annually is equivalent to 5% interest
payable quarterly?
10. For a 25-year deposit, what interest payable semiannually is equivalent to 4% payable
quarterly?

Answers to Odd-Numbered Problems

1. 5% payable semiannually ($26,841) rather than 4j% payable quarterly ($24,480)
3. 143,489,070
5. $3000 fixed costs
33.14% of production costs are fixed for 100 units output
7. 3.45 (approximately 3) firms after 5 years
3 firms originally
6 firms at maximum size
9. 5.1%

1.10 Trigonometric Curves

In this section the measurement of angles in degrees and radians is discussed and the
six basic trigonometric functions are defined. Some of the basic trigonometric identi¬
ties are given. The representation of a point in polar coordinates and the transforma¬
tion of an equation into polar coordinates are also discussed.
124 / Graphical Representation

Figure 1.101

x
r

DEGREE AND RADIAN MEASURE

An angle may be measured in either degrees or radians. Although measurement of
angles in degrees is probably more familiar, measurement in radians is more conven¬
ient for stating formulas and solving problems in calculus.
An angle is measured in the counterclockwise direction. The measurement of an
angle in degrees is based on the arbitrary assumption that a circle contains 360°; thus
a straight line contains 180° and a right angle contains 90°.
The radian or circular measure of an angle whose vertex is at the center of a circle
is defined as the ratio of the length of the intercepted arc to the radius of the circle
(see Figure 1.101). The unit angle, referred to as a radian, is the angle which, when
placed with its vertex at the center of a circle, intercepts an arc equal in length to the
radius of the circle (see Figure 1.102). Since the circumference of a circle is 2nr, there
are 2n radians in a circle, n radians in a straight line, and n/2 radians in a right angle.
The relationship between radian measure and degree measure may be expressed
by

n radians = 180°

Thus

1 radian

1° radians % 0.01745 radians

Figure 1.102

6 = 1 radian
 1.10 Trigonometric Curves / 125

EXAMPLE

200tt 1071
200 ° —— radians = — radians

3571 In
-35° — ~~~ radians = — — radians
180 36
3007c 571
300° -radians = radians
180 3

TRIGONOMETRIC FUNCTIONS
The trigonometric functions of an angle 9 are the sine of 9 (sin 9), the cosine of 9
(cos 9), the tangent of 9 (tan 6), the cosecant of 6 (esc 0), the secant of 6 (sec 9\ and
the cotangent of 6 (cot 0). When the angle 6 is at the center of a circle of radius r and
is measured counterclockwise, as in Figure 1.103, the trigonometric functions of 6 are
defined by the equations
a b a
sin 9 = cos 9 = tan 9 =
r r b
r b
esc 9 = — sec 9 = cot 9 =
a b a
Note that

esc 9 = ——- sec 9 = - cot 9 =--

sin 9 cos 9 tan 9
^ sin 9 . cos 9
tan 9 = - cot 9 =
cos 9 sin 9
The signs of the trigonometric functions can be summarized as follows:

+ + —
+ —
+
— — —
+ + —

sin, CSC cos, sec tan, cot

Figure 1.103

y
126 / Graphical Representation

The values of the trigonometric functions can be obtained geometrically for some
commonly used angles. For example, if 9 = 45° = n/4 radians, b = a and, using the
Pythagorean theorem, r2 — a2 + b2 and r — a-sjl, as shown below.

Thus

sin 6 =
a = sft
cos 0 =
a
A2 tan 8 = - = 1
a^fl 2 aj2 a

Similarly, if 0 = 90° = 7t/2 radians, r = a and b = 0, as shown below.

Thus

a u a
sin 0 = - = 1 cos 6 = - = 0 tan 9 = - not defined
a a 0
Values of the sine, cosine and tangent are given for some commonly used angles in
the table below. These values can all be obtained geometrically.

Degrees 0 30 45 60 90 180 270 360

71 n n n 3n
Radians 0 n 2n
6 4 3 2 T

Sine 0
1 V~2 75 1 0 -l 0
2 2 2
73 ■J~2 1
Cosine 1 0 -1 0 1
2 2 2
V3 not not
Tangent 0 1 73 0 0
3 defined defined

The graphs of the sine, cosine, and tangent functions are given in Figure 1.104.
Note that the function sin x is defined and continuous for all values of x. Sin x is a
periodic function with the period 2n, since sin (x + 2n) = sin x for any x, that is,
when the value of x is increased by 2n, the value of y is unchanged. The function cos x
 1.10 Trigonometric Curves / 127

Figure 1.104

is also defined and continuous for all values of x and is periodic with period In. The
graph of y — cos x can be obtained from the graph of y = sin x by taking the line
x = n/2 as the y-axis; that is, the graph of y = cos x is the same as the graph of
y — sin x when that graph is shifted to the right by n/l radians. The function tan x is
discontinuous for all values of x such that x = (n + j)n, where n is any positive or
negative integer; tan x is periodic with period n.
128 / Graphical Representation

The symmetry and periodicity of the sine, cosine, and tangent functions make it
possible to obtain a trigonometric function of any angle, 0 < 6 < 2n, from the trigon¬
ometric function of an angle 6 such that 0 < 6 < 7i/2. Thus trigonometric functions
are tabled only for angles between 0 and jc/2. The formulas for reducing angles are
given below. Note, for example, that sin ( — 6)= —sin 6, cos ( — 6) — cos 6,
tan ( — 9) = —tan 6, and so forth.

Angle sin cos tan cot sec CSC

-0 — sin 9 cos 9 — tan 9 — cot 9 sec 9 — esc 9

cos 9 sin 9 cot 9 tan 9 esc 9 sec 9

2-*

cos 9 — sin 9 — cot 9 — tan 9 — esc 9 sec 9

n —9 sin 9 — cos 9 — tan 9 — cot 9 — sec 9 esc 9

n + 6 — sin 9 — cos 9 tan 9 cot 9 — sec 9 — esc 9

— cos 9 — sin 9 cot 9 tan 9 — esc 9 — sec 9

!-»

— cos 9 sin 9 — cot 9 — tan 9 esc 9 — sec 9

T +e

2 n-0 — sin 9 cos 9 — tan 9 — cot 9 sec 9 — esc 9

EXAMPLE

Find the value of each of the following trigonometric functions: cos 3n/4, tan 57i/6, sin 5n/3,
sec ln/6, esc 2n/3, cot ( — n/4).

2n In n\ n_ yj2
cos — = cos - + - = —sin
4 2 4

5n I n\ n V3
tan — = tan I n — -1 =
6~~ 3

5n . I n . n 0
sin — = sin I in — - - sin - = -
3 2

In / n' 71 1 _ 2/3
sec — = sec \ n + sec - =
6 \ 6 6 n
cos -
6

271 / n\ n _ 1 _ 2^3
CSC — = CSC I 71 — I = CSC „
3 3/ 3 n
sin

7T n
cot — cot 7 = -1
4 n
tan
 1.10 Trigonometric Curves / 129

TRIGONOMETRIC IDENTITIES
There are a number of trigonometric identities which are useful in simplifying results
involving trigonometric functions. Some of the more commonly used identities are
given below without proof; additional identities and proofs can be found in any
trigonometry book.

sin2 6 + cos2 6=1 sec2 6 — tan2 6=1 esc2 6 — cot2 6=1

sin (x + y) = sin x cos y + cos x sin y

sin (x — y) = sin x cos y — cos x sin y

cos (x + y) = cos x cos y — sin x sin y

cos (x — y) = cos x cos y + sin x sin y

tan x 4- tan y tan x — tan y

tan (x + y) tan (x — y) =
1 — tan x tan y 1 + tan x tan y

2 tan x
sin 2x = 2 sin x cos x cos 2x = cos2 x — sin2 x tan 2x =
1 — tan2 x

1 — cos x x / 1 + cos x
sin — = + cos - = +
2 V1 2 2 ~

X 1 1 — COS X
tan - = 4-
2 yj 1 + cos x

sin2 x = 21 ~ j cos 2x 2 ' 2

POLAR COORDINATES
It is frequently more convenient to describe a point by its polar coordinates rather
than by its rectangular coordinates. If a point is described by the rectangular coor¬
dinates (x, y), then it can also be described by the polar coordinates (r, 6) as follows.
Consider a line drawn between the point and a given origin; the angle 6 which this
line makes with the horizontal axis and the distance r between the point and the
origin determine the point, as shown in Figure 1.105.
Note that the polar coordinates of a point are not unique. For example, the polar
coordinates (r, 6), (r, 6 + 2n), and (r, 6 — 2n) all represent the same point. In prac¬
tice, the polar coordinates of a point are usually written using the angle between 0
and 2n.

Figure 1.105
130 / Graphical Representation

The rectangular coordinates (x, y) and the polar coordinates (r, 6) of a point are
related as follows:

x = r cos 6 y = r sin 6 r = yjx2 + y2

• a y wO
a
H —
x
-
/-s s- tan 6 — —
X
7x + y 7*2 + /

These relations are used to change from one system of coordinates to the other.

EXAMPLE

Find polar coordinates for the following points given in rectangular coordinates,

(a) (0, ^2)

(b) (1, 1)

(a) (x, y) = (0, 72)

r = 7*2 + y1 = 70 + 2 = 72

cos 9 = 0 and sin 9 = 1

so

n
9 =

K'
Polar coordinates (r, 0) = ,

(b) (x, y) = (1, 1)

r = Vx2 + y2 = V1 + 1 = sfi
_ j_ = 72
cos 9 = and sin 9 = J_ = v7
72 2 T2 2
so

n
9 =

n'
Polar coordinates (r, 9) = J2,

EXAMPLE

Find rectangular coordinates for the following points given in polar coordinates
 1.10 Trigonometric Curves / 131

(a) M)

r cos 9 (3)(0) = o

r sin 9 (3)(1) = 3

(x, y) (0, 3)

(b) M)

r cos 9
4 4
^>( ) -

r sin 9
4 -

(*> y) (44 - )

EXAMPLE

Find the equation in rectangular coordinates of the curve whose equation in polar coordinates
is r = 4/(1 + 2 cos 9).

y** 1 2 + y

y'x2 + y2 + 2x 4

4 - 2x

x2 + y2 16 — 16x -I- 4x2

3x2 — y2 — 16x + 16 0 (hyperbola)

EXAMPLE

Find the equation in polar coordinates of the curve whose equation in rectangular coordinates
is Ax + By + C = 0.

Ar cos 9 4- Br sin 9 + C = 0

r(A cos 9 + B sin 9) + C = 0

PROBLEMS
1. Find the number of radians in each of the following angles given in degrees.
a. 9 = -240° d.9 = -40°
b. 9 = 70° e. 9 = -100°
c. 9 = 120° f.9 = 315°

2. Find the value of the following trigonometric functions.

In , 47T
a. sin — b. esc —
6 3
132 / Graphical Representation

5n 5n
c. tan d. cos
4 T
3. Obtain polar coordinates for the following points given in rectangular coordinates.
a. (1, 73) c. (3, 3,/3)
b. (-1, -1) d. (-1,0)

4. Obtain rectangular coordinates for the following points given in polar coordinates.

b. ( — 2, n)

Obtain the equation in rectangular coordinates that corresponds to each of the following
equations given in polar coordinates.

5. r = 5 12. r = a sec 9
6. r = sin 9 + cos 9 13. r2 = a2 cot 9
14. r — tan 9 sec 9
7. r = —-- —-
2 — cos 9 15. r3 = a3 esc 9
8. r = 1 + cos 9 16. r = a tan2 9
9. r = a cos 9 17. r2 = 9
10. r = a sin 0
18. 9 = ^
11. r = 9 2

Obtain the equation in polar coordinates that corresponds to each of the following equa¬
tions given in rectangular coordinates.

19. y2 = 4ax 24. (x2 + y2)2 = a2(x2 — y2)

20. x2 + y2 = a2 25. 2xy = a2
21. x2 + y2 — x + y 26. y — x2
22. y = x 27. xy2 = a
23. y3 = ax2 28. x + y = a

Answers to Odd-Numbered Problems

4tc
1. a. s = —
T
In
b. s =
18

2n d. (1, n)
c. s —
5. x2 + y2 = 25
2n 7. 3x2 + 4 y2 — 4x — 4 = 0
d. s =
9 9. x2 + y2 — ax = 0
11. y = x tan (^/x2 + y2)
5n
—
13. y3 4- x2}; — a2x = 0
9
15. x2y -I- y3 = a3
In 17. y = x tan (x2 -f y2)
f. s —
4 19. r = 4a cot 9 esc 9
21. r = cos 9 + sin 9
3. a. 2, *) 23. r — a cot2 9 esc 9
V 25. r2 = \a2 sec 9 esc 9
27. r3 = a sec 9 esc2 9
 1.11 Applications of Trigonometric Curves in Business and Economics / 133

1.11 Applications of Trigonometric

Curves in Business and Economics
Many economic phenomena can be described in part by sine or cosine functions. In
the study of business cycles, seasonal or other cyclic variations are described by sine
or cosine functions. In spectral analysis, the method most frequently used currently
to analyze periodicity or cyclic variation in time series, the detrended series is
analyzed in terms of cosine functions. Spectral analysis, and even less complicated
types of analysis of cyclic variation, involve statistical problems beyond the scope of
the present discussion. However, these analyses are an application of trigonometric
curves, and the logic on which they are based is intuitively reasonable: A series is
described in terms of components representing trends and cyclic variations.

Technical Note 1 Conditions for Equilibrium

Two linear equations of the form

A^x-FB^y-yCi — 0 A2x B2y C2 = ^

have the simultaneous solution (x, y), where

_ C2 — B2 C1 _ A2Ci Al C 2
A1B2 ~ A2Bl A±B2 ^2^1
provided At B2 — A2 Bl ^ 0.
Thus if the nonvertical demand and supply curves

y = mDx + bD y = msx + bs

(where the subscripts D and S distinguish demand and supply curves) are written

mDx — y + bD = 0 msx — y + bs = 0

then the equilibrium quantity and price are

bD ~ bs bDms - bsmD
x =- y —-
ms -mD ms mD —

since ms — mD f 0. Since mD < 0, ms > 0, yD > 0, and ys is unrestricted, ms — mD> 0

and x > 0 if bD — bs > 0, that is, if bD>bs, which is the first condition stated for
equilibrium in the first quadrant; y > 0 if

bDms - bsmD > 0

or
bD ms > bs mD

bn b<? . , .
<— (since mD < 0)
mD ms

mD ms

which is the second condition stated for equilibrium in the first quadrant, since the
x-coordinate of the x-intercept of the line y = mx + b is —(b/m).

Technical Note II Identification of Degenerate and Imaginary Loci

The type of conic section represented by a second-degree equation, Ax2 + Bxy +
Cy2 + Dx + Ey + F = 0, can be determined from certain conditions on the
coefficients as summarized in Table 1.1.
134 / Graphical Representation

Table 1.1

Case Conditions on the coefficients Type of conic

Proper conic, C > 0; /, D opposite in sign, A = C, Circle

0 B = 0 (no xy-term)
C > 0; /, D opposite in sign Ellipse
C <0 Hyperbola
c=o Parabola
C > 0, /, D same in sign No real locus

Degenerate C< 0 Two intersecting

conic, D = 0 lines
C = 0, J < 0 Two parallel
lines
c = 0, J = 0 Two coincident
lines
C>0 A point
C = 0, J > 0 No real locus

In this table,

I= A + C

B2
C = AC-
4

E2 D2 B2
J = AC + CF + AF-
4 4 4

- ^ BED CD2 AE2 FB2

D = ACE H-
4 4 4 4

In many cases it can be assumed (owing to the nature of the application) that the
given equation represents a proper conic (that is, D ± 0) and a real locus (that is, /
and D opposite in sign if C > 0). In such cases it is necessary only to compute C and
note the coefficients A, C, and B in order to determine which type of conic section is
represented by the equation.

Technical Note III Completing the Square

In many problems it is convenient to have an equation written so that variables

which occur in terms of second degree occur only as terms of a perfect square; this is
accomplished by putting the terms of the perfect square(s) on one side and all other
terms on the other side of the equation and adding the appropriate constant terms to
both sides of the equation.
The process of completing the square is based on the formula (x + a)2 = x2 +
2ax + a2. Thus, given x2 + cx, take half the coefficient of x and square it to complete
the square: x2 + cx + (c/2)2. Any constant term added to one side of an equation
must, of course, be added also to the other side in order to preserve the equality.
Completing the square is the essential step in putting a second-degree equation into
the appropriate standard form.
 1.11 Applications of Trigonometric Curves in Business and Economics / 135

If ax2 + bx + cy + d = 0, then the terms to be included in the perfect square are

ax and bx; thus

a \ x + - x I = — cy — d
a

a | x 2 -I- —
^x T
= — cy — d +
a 4a2 4a

b^2 cl d b4
x + - y H-—
2a > a\ c 4ac,

EXAMPLE

Complete the square for the equation

3x2 + 4x + 2y — 4 — 0

3(xz + I* + %) = — 2y + 4 + 3

(X + |)2 = -2(3; _ 8)

If ax2 + by2 + cx + dy + e = 0, then the terms to be included in the perfect squares are
ax2, cx, by2, and dy; thus

a(x2 + CaX)+b(y2 + iy) = -e

2 c c2 \
. / 2 ^ c2 d2
a x + -x + —c + -—f —
\ a 4? +T +by + 4p 4a 4b

dL
x + y+
2a, 2b; c2 d2
+ = 1/ +
a ab\4a + 4b 4 a2b 4 ab2 ab

EXAMPLE

Complete the squares for the equation

3x2 + 4y2 - 6x + ly - 10 = 0

3(x2 - 2x) + 4{y2 + ly) = 10

3(x2 — 2x + 1) + 4(y2 + ly + ^f) = 10 + 3 + y§

(* — l)2 ,
4-1-^- = TZ(ld +
{y + i)2 1 /17 I 49 \
TB) -
257
T9I

The essential aspect of the process of completing the square is the addition of the appro¬
priate constant terms so that groups of terms form perfect squares; the most convenient form
of writing the constant coefficients depends on the particular problem.
A second-degree equation can be put into the appropriate standard form for the type of
conic it represents by the process of completing the square.
136 / Graphical Representation

The circle:

Ax2 4- Ay2 4- Dx + Ey + F = 0

D D2 E2 r
a(x2 4- ^x +
4A 2j + ^( y +Ay + 4 A2 4A + 4 A

D Ex2 D2 E2 F
+
X + 2A +\y + 2A, 4 A4 4A2 ~ A

which is of the form (x — h)2 + (y — k)2 = r2, where

D D4 E2 F
h = - k = — r2 = +
2A " 2A 4 A2 4A2 A

The ellipse: /lx2 + Cy2 + Dx + Ey 4- F = 0, A and C same sign

D D2 \ J , E E2 D2 E2 ^
A\x +—X + — ]+C\y +~y + 2 — T — — f
4 A4 C 40 4A 4C

D2 E2
Denoting — + — - F by k,

D
x + y + 2C\
2A,
+ = 1
k k
A C

which is of the form

(x - h)2 , (y - k)2
+ = 1
a
where

D
h = - k = -
2A 2C

The parabola: Ax2 + Dx + Ey -f F = 0

. Dx+
A4 x22 + —
4A 2) \y + E 4AEi

D D4 4 AF]
x +
2A. 4AE

which is of the form (x — h)2 = 4p(y — k), where

D D2 - 4AF
h = - k = 4P = ~
2A 4AE A

(and similarly for the parabola Cy2 + Dx + Ey + F = 0).

The hyperbola: ^x2 + Cy2 4- Dx + Ey + F = 0, A and C opposite in sign

., 2 D D2 \ / - E E2 D2 E2
— _(_ — —
Aix +AX + 4A2 + CV +Cy + 4C~2 4A 4C

^ D2 E2
Denoting — + — - F by k.

D
x + y + 2C,
2A
+ = 1
k k
A C
 1.11 Applications of Trigonometric Curves in Business and Economics / 137

which is of the form

(x - h)2 (y - k)2
a2 b2

or

(y - k)2 (x - h)2
b2 a2

since A and C are opposite in sign, where

E
2C

The various properties of particular second-degree curves can, of course, be obtained by

identifying the appropriate type of conic section and then substituting appropriate numerical
values in the corresponding formulas given above, rather than by putting each equation into
standard form.

Technical Note IV Formula for Roots of a Quadratic Equation

The formula for the roots of a quadratic equation is obtained by completing the
square as follows:

ax2 + bx + c = 0

bx b2
a\x -K — ~C + ~T~

a 4 a2 4a

b'2 b2 — 4ac
x +
2a, 4a2

b b2 — 4ac
= +
X + Ta 4 a2

— b± ^/b2 — 4ac
2a

Technical Note V x-intercepts of y = aekx + c and y = /tln(l+x)TB

To obtain the x-intercept for the curve y = aekx + c: Let y = 0; then
If Y

ae = —c

In order for In | — -j to be defined,

c < 0, a > 0 [cases (c) and (d)]

or
c > 0, a < 0 [cases (e) and (f)]
138 / Graphical Representation

Note in Figure 1.94 that only cases (c), (d), (e), and (f) have x-intercepts and that

1
In — > 0 if a > 0, c < 0, k < 0, a > I c [case (d)]
a,

or

a <0, c > 0, k > 0, c > I a [case (e)]

1
k
In j < 0 if a > 0, c<0, k > 0, a > | c [case (c)]

or

a <0, c > 0, /c < 0, c > \a\ [case (f)]

To obtain the x-intercept for the curve y = /lln(l + x) + B: Let y = 0; then

A In (1 + x) = —B

In (1 + x) = — —
A

e-WA)= 1 + x

x= e~{B,A) — 1

and

e-(B/A)_ 1 >0 if — < 0

A

e~WA)_ i=0 if — = 0 that is, if B = 0

A

e~(SIA) _ 1 < 0 if — > 0

A

Thus
e~(BIA) _ j >Q

if

A > 0, B > 0 [case (a)]

A < 0, B < 0 [case (d)]
and

e~(B,A) - 1 <0

if

A > 0, B < 0 [case (c)]

A < 0, B > 0 [case (b)]
Differential
Calculus:
Functions of
One Variable

2.1 Introduction
Calculus is concerned with the mathematical analysis of change or movement. Be¬
cause everything in the world changes, calculus has applications in virtually all areas
of scientific inquiry. It is nearly impossible to exaggerate the importance of calculus,
particularly differential calculus, as a basis for mathematical analysis.
As a distinctly new mathematical method, calculus was developed in the seven¬
teenth century by Sir Isaac Newton and Gottfried Leibnitz, working independently.
For Newton, calculus originated in attempts to solve certain problems connected
with his work in physics and astronomy: finding the velocity of a moving body, the
work done by force, the center of mass of a body. For Leibnitz, calculus originated in
attempts to solve certain problems in geometry: finding the tangent to a curve, the
length of a portion of a curve, the area bounded by one or more curves, the volume of
a solid.
The basic operations of calculus are differentiation and integration; these opera¬
tions are the inverses of each other, as are addition and subtraction, and multiplica¬
tion and division. Differentiation is concerned essentially with determining the rate
of change of a given function. Integration is concerned essentially with the inverse
problem of finding a function when its rate of change is given.
The analogy of a moving-picture film is frequently used in discussing the processes
of differentiation and integration. A moving-picture film is a series of (static) pictures,
each at least slightly different from the others;each frame depicts the subjects in given
positions at a particular instant of time. When the film is run through a projector at
proper speed, the pictures are summed up and the illusion of motion is created.
Similarly, differentiation essentially breaks up a function into many infinitesimally
small (static) pieces and thus analyzes it at a particular point of time or for a
particular value of the independent variable; integration, on the other hand, sums up
the infinitesimally small pieces to obtain the function.
When relations among variables are stated in equations, calculus can be used to
analyze these relations. Calculus has been used by physicists, astronomers, chemists,
and engineers almost since its development; more recently, calculus has also been
used by biologists, sociologists, psychologists, and economists.
Since analysis in business and economics is frequently concerned with change,
139
140 / Differential Calculus: Functions of One Variable

calculus is an extremely valuable tool for these problems. Marginal analysis is per¬
haps the most direct application of calculus in business and economics; marginal rate
of change or variation on the margin is expressed analytically as the first derivative of
the relevant function. Differential calculus is also the method by which maxima or
minima of functions are obtained. Thus problems of maximizing profit or minimizing
cost under various assumptions can be solved using calculus. Mathematical
programming, which is concerned with maximizing or minimizing a function subject
to constraints, is applied increasingly in business and economics; the methods used
in mathematical programming are based on differential calculus.
The idea of the rate of change of a function, which is the basis of differential
calculus, is discussed in the following paragraphs. The applications of this concept to
marginal analysis and to various problems of maximization and minimization are
discussed in some detail in following chapters.
The simplest type of functional relationship between two variables is represented
by a straight line and corresponds to a constant, or uniform, rate of change in the
dependent variable with respect to change in the independent variable. A variable
rate of change in the dependent variable with respect to change in the independent
variable is represented by a curvilinear (or nonlinear) function. The average variable
rate of change is the average value over an interval of the variable rate of change.
For many analyses the most important concept is that of instantaneous rate of
change, that is, the variable rate of change at a particular instant of time or for a
particular value of the independent variable. Instantaneous rate of change is ob¬
tained by differentiation and is the first derivative of the function evaluated at the
point of interest. The concept of instantaneous change is the basis of marginal
analysis in economics; marginal analysis considers the effect on the dependent var¬
iable of small changes in the independent variable, that is, variation on the margin.
The mathematical definition and derivation of instantaneous or marginal rate of
change are discussed in detail later; the concept can perhaps best be understood
intuitively in terms of an example of physical movement. If a car is driven from city A
to city B at a constant speed, then the rate of change in its distance from city A is
constant with respect to change in time since leaving city A; the car is traveling at a
constant or uniform rate. However, if the car occasionally slows down for curves,
towns, traffic, and so forth, it travels at a variable rate. Suppose the trip from city A to
city B takes 5 hours; the number of miles traveled each hour could be averaged to
obtain the average variable rate of travel per hour. There is clearly another rate of
travel of interest to both drivers and traffic policemen—the rate of travel at a particu¬
lar instant of time. It is this instantaneous rate that is obtained by differential
calculus.
Although rate of change is perhaps most readily understood in terms of physical
motion, it can be generalized to any type of functional relationship. For example,
total cost is a function of the quantity produced and usually changes at a variable
rate as the quantity produced changes. The rate of change of total cost as quantity
produced changes is the marginal cost and is the first derivative of total cost; mar¬
ginal cost is a function of the quantity produced and can be evaluated for any
quantity of interest.
The first derivative, that is, the rate of change of a function, can be used to
determine its maximum and minimum points, if any. A function increases (positive
rate of change) until it reaches a maximum and then decreases (negative rate of
change); similarly, a function decreases until it reaches a minimum and then
increases. This basic method for determining the maxima and minima of a function
by differentiation has been developed and generalized for use in problems of
varying complexity; as noted above, it is the basis of the methods of mathematical
programming.
 2.2 Limits / 141

Calculus is concerned with infinitesimally small changes in the independent and

dependent variables. Mathematically, such changes are defined using the concepts of
limits and continuity. Thus the first sections of this chapter concern the mathematical
concepts of limits and continuity, which provide the foundation for the theory of
calculus.
The first derivative of a function is then defined and some of its applications in
business and economics are discussed. Rules for differentiation of various types of
functions are given. Differentials and their use in approximating functions, higher-
order derivatives, implicit differentiation, and differentiability and continuity are also
discussed. The applications of first and second derivatives in graphical representation
are discussed and illustrated. A number of applications of derivatives in business and
economics are considered. Indeterminate forms are defined and discussed. Conver¬
gence of sequences and series is considered and the use of Taylor’s theorem for
approximating functions is illustrated.

2.2 Limits

The concept of a limit seems to be one of the most troublesome in mathematics.

Apparently, the idea of approaching a point or value arbitrarily closely and still
never reaching it is not intuitively appealing. Actually, limit-type concepts are
frequently used in nonmathematical thinking and conversation. For example, the
theoretical maximum production of a machine or of a factory is a limit, the ideal (or
limiting) performance which is never attained in practice but which can be ap¬
proached arbitrarily closely. This same idea applies to the performance of any
mechanical or electronic device for which engineers can calculate an ideal (or limit¬
ing) performance; it applies also, for example, to profits under ideal conditions, gas
mileage under ideal conditions and operation, and so forth. Similarly, there are lower
limits of cost, waste, spoilage, and so forth.
The mathematical concept of a limit is fundamental in understanding differential
calculus and is discussed in some detail below. Properties of limits are also con¬
sidered; these properties are used in later sections concerning series and indetermin¬
ate forms.

DEFINITION OF A LIMIT
Consider a function f(x) and let the independent variable x assume values near a
given constant a; then the function f(x) assumes a corresponding set of values.
Suppose that when x is close to u, the corresponding values of/(x) are close to some
constant A. Moreover, suppose that the values of/(x) can be made to differ arbi¬
trarily little from A by taking values of x that are sufficiently close to a, but not equal
to a and that this is true for all such values of x. Then /(x) is said to approach the
limit A as x approaches a. More concisely, the definitions of the limit of a variable
and the limit of a function are as follows.
A variable x is said to approach a constant a as a limit when x varies in such a way
that the absolute difference | x — a | becomes and remains less than any preassigned
positive number, however small this number is chosen. This is indicated by the
notation

lim x = a or x —> a
142 / Differential Calculus: Functions of One Variable

EXAMPLE

If x assumes the sequence of values

1 3 7 15 2" - 1
2’ 4’ 8’ 16’ 2" ’

then x -*■ 1. But if x assumes the sequence of values

1 3 7 15 2" - 1
2’ 4’ 8’ 16’ ■’ 1 ' 2"

then x does not approach a limit.

If the function /(x) approaches a constant A when x approaches a in whatever

manner without assuming the value a, A is said to be the limit of/(x) as x approaches
a. This is indicated by the notation

lim f(x) — A or /(x) -► A as x —► a

x~>a

EXAMPLE

If/(x) = 2x + 5, lim*-^ /(x) = 5, since, for example,

/(1) = 7 /(-1) = 3
fii) = 6 f(~i) = 4
/(i) = 5t /(—i) = 4i
49
/(tto) = 5yts /(—tot) = 4TU
1 1499
/(Wot) = 5-jot /( two) =
etc. etc.

The following two statements of the definition of the limit of a function are
equivalent to the definition given above.

A function f(x) is said to approach a limit A as x approaches a if the absolute

difference between/(x) and A is less than an arbitrarily small positive number for all
values of x that are sufficiently close to a and for which x =£ a.
A function /(x) approaches a limit A as x approaches a if, and only if, for each
e > 0 there exists a S such that whenever 0 < | x — a \ < S, \f(x) — A \ <8.

It is implied in the above discussion of limits that both x and/(x) approach finite
constants (a and A, respectively) as limits. It is also possible, however, for either one
or both of x and f(x) to become arbitrarily large or arbitrarily small. These types of
limiting behavior are defined as follows:

If the difference between a function/(x) and a constant A is less in absolute value

than an arbitrarily small positive number for all positive values of x that are
sufficiently large, then/(x) is said to approach A as a limit when x becomes positively
infinite, that is, increases without limit. This is indicated by the notation

lim f(x) = A or /(x) -> A as x oo

X~* 00
 2.2 Limits / 143

note: The notation limJc^+ ^ f(x) = A or /(x) -> A as x -> + oo is sometimes

used. However, oo is understood to mean + oo, as would be the case if it were a
number, and the plus sign is usually omitted for convenience.

EXAMPLE

If/(x) = 1 — (1/x), then lim*-,^ /(x) = 1, since, for example,

/(i) = o
/(5)=l
/(20) = %
/(100) = -nni

/(1000) = «
/(10,000) = «
etc.

Similarly, the limit off(x) may be defined when x becomes negatively infinite, that
is, decreases without limit. This is indicated by the notation

lim f(x) = A' or f(x) -*• A as x —► — oo

X~* — 00

EXAMPLE

If f(x) = 1 - (1/x), then

lim /(x) = 1
X~+ 00

If a function f(x) is greater than an arbitrarily large positive number for all values
of x that are sufficiently near a constant a and for which x a, then f(x) is said to
become positively infinite (that is, increases without limit) as x approaches a. This is
indicated by the notation

lim f(x) = 00 or /(x) —> oo as x -► a

x~>a

Similarly,/(x) becomes negatively infinite (that is, decreases without limit) when it
assumes numerically large negative values. This is indicated by the notation

lim /(x) = — oo or /(x) -> — oo as x -► a

x->a

EXAMPLE

If/(x) = l/(x - 2)2, then

lim f(x)— oo
x-2
144 / Differential Calculus: Functions of One Variable

If a function /(x) is greater than an arbitrarily large positive number for all
positive values of x that are sufficiently large, then f(x) is said to become positively
infinite (that is, increases without limit) as x becomes positively infinite (that is,
increases without limit). This is indicated by the notation

lim f(x) = oo or /(x) -» oo as x —► oo

X-* 00

The cases indicated by the following notation are defined similarly:

lim f(x) = — oo or /(*) —> — oo as x -> oo

X~* 00
lim /(x) = oo or /(*) —► oo as x~> — oo
-*■ — 00
lim f(x) = — oo or f(x) —► — oo as x —>> —
X~* — 00

EXAMPLE

Iff(x) = x4 — 4, then

lim f(x) = oo and lim f(x) = oo

EXAMPLE

If/(x) = x3 — 8, then

lim f(x)= oo and lim f(x) = — oo

In some cases a function may approach either of two different limits, depending on
whether the variable approaches its limit through values larger or smaller than that
limit; in such a case, the limit is not defined (does not exist) but the right-hand and
left-hand limits exist.
The right-hand limit of a function is the value the function approaches when the
variable approaches its limit through decreasing values (that is, from the right); this
type of limiting behavior is indicated by the notation

lim f(x) = A+ or /(x) -> A+ as x -> a +

x~*a +

The left-hand limit of a function is the value the function approaches when the
variable approaches its limit through increasing values (that is, from the left); this is
indicated by the notation

lim f(x)=^A~ or /(x) -> A~ as x a~

Thus the limit of a function exists if and only if its right-hand and left-hand limits
exist and are identical; in that case

lim f (x) = lim /(x) = lim/(x)

x-*a+ x~>a~ x~>a
 2.2 Limits / 145

EXAMPLE

If/(f) = [f] = largest integer in f, then

lim /(f) = 3 and lim /(f) = 2

f-3 + 1-3-

Thus limf_3 f(x) is not defined.

It might seem that the limit of this function /(f) as f->3 should be 3. However,
when f is arbitrarily close to 3, some of the values of [f] are 2 (when f < 3), others are 3 (when
f > 3). Thus the values of f are not close to any one value A when f is arbitrarily close to 3 and
lim,_3 [f] does not exist, although the right-hand and left-hand limits do exist. This is in¬
dicated in Figure 2.1 by the unbroken horizontal lines between successive integer values of f.
That is,/(f) approaches 2 as f approaches 3 from the left;/(f) approaches 3 as f approaches 3
from the right. Clearly there is nothing unique about the integer 3; in fact,/(f) does not have a
limit (although it has right-hand and left-hand limits) as f approaches any integer value;/(f)
does, however, have the limit [f] as f approaches any noninteger value.
Figure 2.1

fit)

8
7
6
5
4 l

3
fit) = [t]
2
1
t

EXAMPLE_=============

If/(f) = 1/f, then

lim /(f) = oo and lim /(f) = — oo
f-0+ f-0~

(see Figure 2.2).

Figure 2.2

m
146 / Differential Calculus: Functions of One Variable

Thus limt_0 f(t) is not defined; as t approaches 0 from above,/(r) becomes positively
infinite; as t approaches 0 from below,/(f) becomes negatively infinite.

It is also possible that even right-hand and left-hand limits do not exist; this is the
case, for example, for the sine and cosine functions which are oscillatory in nature.

PROPERTIES OF LIMITS
The following properties of limits are useful in evaluating the limit of a function.

If limx^a f(x) = b, lim,,^ g(x) = c, and K is a constant,

1. lim K = K.
x~*a

That is, the limit of a constant is equal to that constant. For purposes of interpreting
the following properties of limits, a constant may be thought of as a special type of
function—that is, a function that assumes only one value. Thus the following state¬
ments are valid (as special cases) if some or all of the functions are constants.

2. lim [f(x) ± g(x)] = lim f(x) ± lim g(x) ~ b ± c.

x~+a x~>a x~>a

That is, the limit of the sum (or difference) of two functions is equal to the sum (or
difference) of their limits.

note: This rule applies to the sum or difference of any finite number of func¬
tions, that is, if limx^a f{(x) = a{ for i = 1, 2, ..., n, then

lim ( £ Mx)\
x -* a \ i = 1 /
= 2 lim j\(x) =
( = 1 x~* a
£ a,
i= 1

3. lim [f(x) • g{x)] = lim f(x) lim g(x) = be.

x -*a x~*a

That is, the limit of the product of two functions is equal to the product of their
limits.

note: This rule applies to the product of any finite number of functions, that is,
=
if lim*^ f^x) = at for i 1, 2, ..., n, then

Um ( fl fi(x)\ = fl lim
x -* a \ i — 1 / i= 1 jc -► a
f\(x) = f] ^
i= 1

fix) lim/(x) b
4. lim ') { = ——- = if c ^ 0.
x-a g(x) lim g(x) c
x-+a

That is, the limit of the quotient of two functions is equal to the quotient of their
limits, provided the limit of the divisor is not zero.

5. lim [/(x)]" = lim f(x) = bn.

That is, the limit of the nth power of any function is equal to the nth power of the
limit of the function. (This follows from rule 3 applied to the product of n equal
factors.)

6. lim (<//(x)) = </lim f(x) = </f>.

x~*a x~>a
 2.2 Limits / 147

That is, the limit of the principal nth root of a positive function is equal to the
principal nth root of the limit of that function. (This follows from rule 4.)

EXAMPLE

lim (3x + 5) — lim 3x + lim 5 (rule 2)

x~> 2 x~>2 x->2

= 6 + 5 (rules 3, 1)

= 11

EXAMPLE

/ X2 | -i
lim I - (rule 4)
x-- 1 \x2 + a)1 lim |(x2 + a)

lim x
u->- 1
(rules 5, 2)
lim x) +1 lim a
\x~* — 1 / \x-+-l

1
(rule 1)
1 + (3

EXAMPLE

1/3

lim (10^/x2 + 2) = [ lim 10 lim (x2 + 2) (rules 3, 6)

x~* 5 \ x~* 5 x-5

= 10(27)1/3 (rules 1, 2, 5)

= 30

EXAMPLE

lim (x2 + a)
'x2 + a] x->0
lim (rule 4)
x-»0 >x2 + b, lim (x2 + b)
x-+0

a
(rules 2, 1, 5)
b

EXAMPLE

lim a
X-> 00
lim (rules 4, 2, 4)
X-* 00 lim 1
X~> 00
lim 1 +
x-* 00 lim X
X~* 00

a
1 + 0 (rule 1)

= a
148 / Differential Calculus: Functions of One Variable

EXAMPLE

lim 2
lim lim b - _^.- ~c0_ (rules 2, 4)
X-* - 00 3x + 10J -*-» lim (3x + 10)

= b - 0 (rules 1, 2, 3)
= b

In applying rule 4 concerning the limit of a quotient, it may happen that the
quotient of the limits takes the form 0/0. This is one of the indeterminate forms
discussed in some detail in a later section; however, it is frequently possible to obtain
a determinate limit for the original quotient if a very simple procedure is used. This
procedure consists of dividing both the numerator and the denominator by an ex¬
pression which in the limit is equal to zero, thereby obtaining an expression having a
limit.

EXAMPLE

Iff(x) = (x2 — 4)/(x - 2), then limx^2 /(*) is the indeterminate form 0/0. But if x £ 2, both
numerator and denominator can be divided by x — 2. Then /(x) = (x2 — 4)/(x — 2) = x + 2
and limx_+2 f(x) = 4.
Recall that in evaluating limx^2 /(*), it is not the value of/(x) at x = 2 but only the values
of/(x) for x close to 2 which are of interest. It is assumed that x becomes arbitrarily close to 2,
but does not equal 2. For x =/= 2,/(x) = x + 2 and limx^2 /(x) = limx_>2 (x + 2) = 4. Thus,
even though/(x) is undefined at x = 2, it has the limit 4 as x -> 2.

EXAMPLE

If F(h) = [(2 + h)2 — 4\/h, then lim^o F(h) is the indeterminate form 0/0. However, for h =/=■ 0,
F(h) = (4 -I- 4h + h2 — 4)/h — (4h -F h2)/h = 4 + h and thus lim^_0 F(h) = 4.

A second type of indeterminate form that can arise as a result of applying rule 4 is
oo/oo. This is another of the indeterminate forms discussed in later sections; in this
case it is frequently possible to obtain a determinate limit for the original quotient if
both the numerator and denominator are divided by the highest power of the var¬
iable which appears in the denominator.

EXAMPLE

/(x) = (2x3 + x2 - 3)/(x3 + x + 2), then lim*^ /(x) is the indeterminate form oo/oo.
If

However, if both the numerator and denominator are divided by x3, then

2x3 -F x2 — 3
/M x3 + x + 2 i 1 2
1 + 2 -3
X X

and thus lim^ ^ /(x) = 2, since as x -*■ oo, 0, o.2

X
0, and
 2.2 Limits / 149

EXAMPLE

If/(x) = [x + (l/x)]/(2x — (1/x)], then limX-+Xf(x) is the indeterminate form oo/oo. However, if
both the numerator and denominator are divided by x, then

1 1 1
x 4— 1 4—o
/(*) =
2x - 1 2 - *
X X

and thus lim*-^ /(x) = since 1/x2 ->0 as x-^ oo.

PROBLEMS
Evaluate the following limits.
1. lim (t2 -f 6t 4- 5) 19. lim e r
t-> 2 f-> ao

/V 4- e
2. lim (x2 + 6x 4- 3) 20. lim L-
x->0 t->o \ 2

21. lim
3!'" (^) x-o \x + yj
4. lim (y — 3)2 '2X - 2 — x\
y--2 22. lim
x-*o \2JC 4- 2
5. lim (x3 — 3x 4- 5)
x-2 x-yy
23. lim
x\ x -I- 2' ro \* + ^
6. lim
x-+2 \ X + 1 x3 — 2x 4- 5'
24. lim
2 2x3 - 7
7. lim
x~>oo x2 + 1 ^x2 4- a2'
25. lim
x-ao \x3 4- a3
8. lim 1 +
x-0 X 1 x3 — 5x 4- 6\
26. lim
,2 x-o \*2 - 2x + 3/
9. lim
x-> - i \X +X+1 1
27. lim
X -> 00
1 + 21/xj
10. lim 4-
x 4- 1 1
X~+ OO \
28. lim
x~* — oo 1 + 21/A
31 - 5'

11. lim
t-»o \ f 4~ 2
29- (rrW
3r2 - 5t + 4^
12. lim
t~* oo \ t2 4- 2 /
3a(nW,
13. lim ax + h
/i-0
31. lim x3
x 2
14. lim 2~h
/i — o 32. limx-3
x-* 2
15. lim x-4
x->2
„ lax 4- 10'
33. lim
x~>a X
16. lim x4
x--2

17. lim 2X 34(rr^,

X “► — 00

18. lim 2~x

X-* oo 35- (i+W,
150 / Differential Calculus: Functions of One Variable

(h + l)e~h
36. lim I 1 l7x 52. lim
1 + ellxl /i-0 h2 4- 1

(x 4- ex2 + 3 4- x2^ 53. lim e1/x

37. lim
x-0 e3 4- x
x2 - y2 4- 2'
54. lim
38. lim - 3,-0 \ * -y
x-o U + ^

55. lim (x4-6x,- 4

39. lim ( 3 " x —2 \ X + 1
/i-i + 4h + 5
r 4- 412 4- 10'
f2 4 4 56. lim
40. lim 00 512 4- 12r
t—2 (f + 2)(f + 3)
1 4-
4/f 4- 5)
41. lim (^1/f 57. lim
1 — 00 /i— oo \ £

42. lim
(el — e 2t — e3{2 n + <?1//r
58. lim -5-
r-0 10 h— - oo V ^ /

16
43. lim 59. lim
x—4 (x - 4)2 fc-2 \x 4- h
x2 — 16 ^2 - 6f 4- 8\
44. lim 60. lim
X- -4 (x^4p f-2 ^2 — 5r 4 6/

x2 - 16
45. lim 61. lim (£±*
x—4 (x 4 4)2 *-« \3/i + 2h6
-t
h~x + h;
46. lim I 62. lim , .,
x-0 ' *-o U + e 1 >

47. lim e~, + 2 (t2 — t — 12\

1-2
63. lim
t— — 3 \fz 4- 4f 4 3/
tQ h _ £,^11 '
48. lim ;x2 4- 2ax 4- a2
/i-0 64. lim
x —0 \ ^

49. lim (1 4- 31/x)

X— 00
7 2 — e*'
65. lim ,14
x-0+ \ e I

50. lim (X~^ + 3)

y-0 \X 4 y - 6/ ‘2 - ex'
66. lim 1/x
x—0” \ ^
'x3 — 3x2 4- 2x - 6'
51. lim
x —► 3 x + 4

Answers to Odd-Numbered Problems

1. 21 25. 0
10
3. 27. ii
5. 7 29. 0
7. 0 31. -8
9. 2

11. 5 a2 + 10
—7
33.
13. ax a
i
15. T5 35. 1
17. 0 37. 1
19. 0
21. -1 1
39.
23. 1 lOe
 2.3 Continuity / 151

41. 6 57. 0
43. oo
45. 0 59. *
47. 1 X
49. 2 61. 0
51. 0 63. i
53. 1 65. 0
55. 0

2.3 Continuity

in this section continuity of a function is defined. Essentially, a function is continuous

if its graph consists of an unbroken curve; the mathematical definition of continuity
involves properties of limits. Types of discontinuities are discussed first and then
properties of continuous functions are given.
In the definition of limx^a /(x) the value off(x) for x = a is not specified; that is,
this limit depends only on the values of f(x) in the neighborhood of (i.e., close to)
x = a but not on the value of f(x) at x = a. Thus limx_+a /(x) may or may not be
equal to f(a). If lim x_>fl/(x) exists and the value f(a) exists and is equal to
limx^a /(x), then f(x) is continuous at x = a. That is, a function f(x) is said to be
continuous at x = a if

1. f(a) is defined.
2. lim f(x) exists.

3. lim f(x) =f(a).

x~*a

note: When a limit is said to exist, this should be understood to mean that the
limit exists finitely. Possible confusion arises because, for example, the expression
limx^a A = oo is written; however, this does not mean that A approaches a number
designated by oo, but only that A becomes arbitrarily large as x approaches a. It
should be remembered that oo is not a number and should not be thought of as a
number, even though for convenience it is used as a number would be used in some
expressions.

A function f(x) is said to be continuous in (or on) an interval b < x < c (or
b < x < c) if it is continuous at every point of the interval.
A function that is not continuous at a point x = a is said to be discontinuous at
x = a.
From the definition of continuity, it follows that the graph of a function that is
continuous on an interval consists of an unbroken curve (that is, a curve that can be
drawn without raising the pen from the paper) over that interval. Continuity was
tacitly assumed for the curves discussed in the preceding sections concerning graphi¬
cal representation. In subsequent sections this assumption, which makes it possible
to sketch a curve by plotting relatively few points and drawing an unbroken curve
through them, will be justified for several (large) classes of curves.

TYPES OF DISCONTINUITIES
The three conditions that must be satisfied for a function to be continuous can be
violated in various ways, resulting in different types of discontinuities. In general,
three types of discontinuities occur.
152 / Differential Calculus: Functions of One Variable

1. A function /(x) is said to have an infinite discontinuity at x = a if /(x) becomes

infinite (positively or negatively) as x-+a. That is, f(a) is not defined and
limx^a f(x) does not exist.
2. A function/(x) is said to have a finite discontinuity at x = a if/(x) remains finite
but changes abruptly at x = a. That is,f(a) is defined but lim*.^ /(x) does not
exist [although, in general, the right-hand and left-hand limits exist and f(a) is
equal to one of them].
3. A function/(x) is said to have a missing-point discontinuity at x = a iff(a) is not
defined but limx^fl f(x) exists.

It should be noted that a discrete function is defined only for a finite number of
values of x in any interval and is thus discontinuous at all other (infinitely many)
values of x in the interval. Since x assumes only discrete values, the concept of a limit
is not appropriate for discrete functions.
Geometrically, the curves representing functions having different types of discon¬
tinuities are quite different in appearance: The curve representing a function having
an infinite discontinuity at x = a approaches x = a as an asymptote; the curve re¬
presenting a function having a finite discontinuity at x = a has an abrupt jump or
step at x = a; the curve representing a function having a missing-point discontinuity
at x — a appears to be continuous, but the single point x = a is missing; the curve
representing a discrete function consists of discrete points and thus is not a curve in
the usual sense.
A missing-point discontinuity at x = a occurs because/(a) is not defined; this type
of discontinuity may be removed by defining

f(a) = lim f(x)

x~*a

since all three conditions of the definition of a continuous function at x = a are then
satisfied. It is possible for a function to have more than one missing-point discontinu¬
ity; as long as the number of these discontinuities is finite, they may be removed by
the procedure described. Note that there is also the logical possibility (although a
practical example is difficult to find) that there is a discontinuity at x = a because
f(a), although defined, is not equal to limx^fl/(x). Such a discontinuity is not remov¬
able, since f(a) is already defined and cannot therefore be defined as equal to
lim*-.. /(4
Although missing-point discontinuities are removable, finite and infinite discon¬
tinuities are not removable. This is the case because the definition of a function is an
arbitrary set of rules; adding another rule to this set, provided it is compatible with
the others, is perfectly permissible, and this is all that is required to remove a
missing-point discontinuity. On the other hand, finite and infinite discontinuities
occur because the function in question does not approach a limit under the specified
conditions. A limit either exists or does not exist as a mathematical fact associated
with the nature of the function in question and cannot be defined into existence; thus
there is no way to remove a discontinuity resulting from the nonexistence of a limit.

EXAMPLE

The function f(x) = l/(x — 2)2 has an infinite discontinuity at x = 2, since f(x) oo as x -► 2
and/(2) is undefined. However, this function is continuous at all values of x other than x = 2
(see Figure 2.3).
 2.3 Continuity / 153

Figure 2.3
fix)

EXAMPLE

The function f(x) = 4x/(4 — x2) has infinite discontinuities at x = ±2, since f(x)-> oo as
x -*• 2 /(x) -> — oo as x->2 + ,/(x)->oo as x —► — 2~, and /(x) -► — oo as x —► — 2 + , and
/(+ 2) and/(— 2) are undefined. However, this function is continuous for all values of x other
than x = ±2 (see Figure 2.4).
154 / Differential Calculus: Functions of One Variable

EXAMPLE

The function /(x) = 1/(1 + 21/x) has a finite discontinuity (that is, a “jump”) at x = 0, since
lim^o /(x) is undefined. However, this function is continuous at all values of x other than
x = 0 (see Figure 2.5).

Figure 2.5

fix)

EXAMPLE

The function /(x) = (x3 — 2x2 — 3x + 6)/(x — 2) is undefined and is thus discontinuous for
x = 2. However, for x ^ 2,

x3 — 2x2 — 3x + 6 (x2 — 3)(x — 2)

fix) = = x2 — 3
x — 2 x — 2
and lim/(x) = 1. Thus if, by definition,
x-*2

/(2) - lim /(x) = 1

x-2

then f(x) is continuous for all x and its graph is the parabola

y = x2 - 3

(see Figure 2.6).

Figure 2.6
 2.3 Continuity / 155

EXAMPLE

The function /(x) = (x2 — 9)/(x + 3) is undefined and is thus discontinuous for x = — 3.
However, for x ^ — 3,

x2 — 9
■^> 7TT= = *-3

and lim /(x) = — 6. Thus if, by definition,

x~* — 3

/(— 3) = lim f(x) = -6

x-* — 3

then/(x) is continuous for all x and its graph is the straight line

y= x —3
(see Figure 2.7).

Figure 2.7

Many functions in business and economics are discrete or have finite discontinui¬
ties of the step-function type. For example, price and cost functions are frequently
discrete because of the nature of the commodity involved or have discontinuities
because cost and price per unit decrease (or increase) abruptly for particular quanti¬
ties. Supply and demand functions and many other economic functions are also
frequently discrete because of the nature of the commodity involved. It should be
noted that functions which are, in fact, discrete are frequently represented as contin¬
uous for convenience; this applies, for example, to the supply and demand functions
for commodities sold in units—such as refrigerators, eggs, light bulbs, chairs, lawn
mowers, cars, and so forth. Representation as continuous of a function which is by
nature discrete makes possible the use of many tools of analysis not otherwise
applicable. However, in interpreting the results of such analyses, the inherent
discreteness should not be forgotten—for example, it is inappropriate to discuss the
price of 1.632 refrigerators or the wages of 29.2 workers.
156 / Differential Calculus: Functions of One Variable

EXAMPLE

A wholesale grocer sells number 2 size cans of a certain vegetable in case lots according to the
following price schedule:

$2.50 per case for 20 cases or less

$2.00 per case for orders of more than 20 cases and not more than 50 cases
$1.75 per case for orders of more than 50 cases and not more than 100 cases
$1.50 per case for orders of more than 100 cases

If y is total price and x is quantity in cases, the price function can be represented algebraically
as follows.

2.50x 0 < x < 20

2.00x 20 < x < 50

1.75x 50 < x < 100

1.50x x > 100

and geometrically as shown in Figure 2.8.

Figure 2.8

Note that, for this price function, total price is not a monotonic function of
quantity ordered. This makes certain orders unlikely. For example, the price of 17
cases is $42.50, while the price of 21 cases is only $42.00; thus an order of 17 cases
would probably be unlikely.

EXAMPLE

A company sells printed business stationery only in boxes of 200 sheets at $2.25 per box. If y is
total price and x is number of boxes, the price function can be represented algebraically by the
equation

y = 2.25x for x = 1, 2, 3, ...

 2.3 Continuity / 157

Figure 2.9

y
C/5
I—

—
20.25 -
o
"O

15.75-
u
o
a 11.25 f—
cd
o
h* 6.75 h

2.25 - •

J L
12 3 4 5 6 7 8
Quantity (in boxes)

and geometrically by the graph in Figure 2.9. This function is discrete, since it is defined only
for integer values of x.

EXAMPLE

An oil refinery has 5 distillation towers and operates as many of them as needed to process the
raw materials available. The overhead cost of operating each distillation tower (operator,
maintenance, and so forth) is $100 per week; in addition, the cost of raw materials is $0.40 per

Figure 2.10

y
158 / Differential Calculus: Functions of One Variable

gallon of refined oil. Each distillation tower can process raw materials yielding 10,000 gallons
of refined oil per week. If y is the total cost of operation and x is the quantity of refined oil (in
gallons), the cost function can be represented algebraically by the equation

x
f(x) = 10o( + 1 + 0.4x
10,000

where [x/10,000] denotes the largest integer part of x/10,000. Note that 1 is added to [x/10,000]
because [x/10,000] distillation towers produce a maximum of 10,000x gallons. Thus for any
additional production, another distillation tower must be operated. The graph of the function
is shown in Figure 2.10. This function has finite discontinuities at x = 10,000, 20,000, 30,000,
and 40,000.

EXAMPLE

United States first-class postage is 10 cents per ounce or fraction of an ounce. If y is postage
required and x is letter weight (in ounces), the cost function can be represented by the equation

y = 0.10([x] + 1)

where [x] denotes the largest integer part of x. The graph of the function is shown in
Figure 2.11. This function has finite discontinuities at all integer values of x.

Figure 2.11

EXAMPLE

A stationery store sells Christmas cards for 10 cents each or $1.00 for a dozen. If y is price and
x is number of cards sold, the price function for a customer can be represented by the equation

X
y= + 0.10 for x = 1, 2, 3, ...
12

where [x/12] denotes the largest integer part of x/12. This function is discrete since it is defined
only for integer values of x; it is also nonmonotonic, as a result of discounting for larger
quantities purchased. (See Figure 2.12.)
 2.3 Continuity / 159

Figure 2.12

y
Price (in dollars)

3.10
3.00

i
2.10
2.00 +*

1.10
1.00 *;•*

0.20
0.10 L X
11 12 23 24 35 36
Number of cards

PROPERTIES OF CONTINUOUS FUNCTIONS

The following properties of continuous functions follow from the definition of a
continuous function and the properties of limits.
If/(x) and g(x) are two functions that are continuous at x = a, then
f(x) = f(x) + g(x)
G(x) =f(x) - g(x)
H(x) =f(x) ■ g(x)
are also continuous at x = a. If g(a) ^ 0, then

is also continuous at x = a. [Note that if g(a) = 0, then 1(a) is undefined.]

The following properties can also be proved (but by more advanced arguments):
If a function/(x) is continuous on a closed interval a < x < b, then
1. f(x) has a greatest value (maximum) and a least value (minimum) in the interval.
2. If k is a number between a and fr, then there is at least one point c between f(a) and
f(b) such that f(c) = k.
3. If f(a) and f(b) are opposite in sign, then there exists at least one point d in the
interval such that f(d) = 0.

PROBLEMS
For each of the following functions, determine the values of x for which the function is
discontinuous, identify the types of discontinuities, and give appropriate definitions for remov¬
ing any removable discontinuities.

3x + 5 x —2
i- /M = 2. /(x) =
x2 + 4x + 4 x 4- 1
160 / Differential Calculus: Functions of One Variable

x2 - 1 x2 + 2x — 8
3- f(x) 26. /(x)
x — 1 x 4 4 ~

x2 + 4 1
27. /(x) =
4- f(x) 3e3x - 3
x2 — x — 2

log (2x — 5) (x + 5)2(x 4- 3)

5. f(x) 28. /(x) =
(x + 5)(x2 — 4x 4- 8)
1
6. f(x) x2 + 5x + 6
2^^1 29. /(x) =
x + 2
1
!■ f(x) x2 4- 3x 4- 6
x(x - 2) 30. /(x) =
x2 — 1
lx 4 1
-

8. f(x) 10
V x - 1 31. /(x) =
e6x - 1

9. fix)
_(x ~ l)2_ x2 -I- x — 2
(x — l)(x2 — 4x + 5) 32. /(x) =
x2 + 27x -f 50

10. fix) x2 + lOx + 1

33. f(x) =
x2 — 9
x2 + 5x 4- 6
11. fix) x2 — 5x + 4
x 4- 2 34. /(x) =
x —4
x2 — 3
12. fix)
x2 -4 35. /(x) = In (x2 — 6)

x2 — 2x x2 — 5x 4- 6
13. fix) 36. /(x) =
x3 — x2 + X (x — 2)(x2 — 2x — 3)

x2 — 5x 4- 6 2x2 + 3x
14. fix) 37. /(x) =
x — 2 x3 — 9x
x2 4- 1 1
15. f(x) 38. /(x) =
x3 — 4x 3ex - 3
4x ex -f 4x
16. fix) 39. /(x) =
4 — x2 3e4x -1
_1_
17. fix) x2 + 5x + 4
4x2 — 16 40. /(x) =
(x + 4)(x2 — 6x + 10)
x — 2
18. fix) x2 - 1
(x — 2)(x2 4- 2x + 10) 41. f(x) = for x ^ — 1
x 4- 1
1
19. fix) = —2 for x = — 1
e4x - 1
x — 2 Is/(x) continuous at x = —1?
20. fix)
(x — 2)(x2 + 2x — 3)
x2 4 3x 10 - —

42. /(x) =--- for x ^ 2.

ex + 2x2
21. /(x)
2e3x — 2
What value should be assigned to
x2 — 5x + 6 /(2) to make/(x) continuous at
22. /(x)
(x — 2)(x2 — 3x + 5) x = 2?
1 43. /(x) = ^/x. Is/(x) continuous at x
23. /(x)
ex - 1
x2 - 7x + 12
x2 — 3 44. f(x) = for x=^ 3
24. fix) x — 3
x2 — 16
= 1 for x = 3
(x - 3)3
25. /(x)
(x — 3)(x2 — 2x -f 6) Is/(x) continuous at x = 3?
 2.4 Definition of the First Derivative / 161

Answers to Odd-Numbered Problems

1. x = —2 23. x =0
3. x= 1,/(1) = 2 25. x = 3,/(3) = 0
5. x < | 27. x =0
7. x = 0, x = 2 29. x = — 2,/( —2) = 1
9. x= 1,/(1) = 0 31. x = 0
11. x= — 2, /(— 2) = 1 33. x = +3
13. x = 0,/(0) = -2 35. x = ±>/6
15. x = 0, x = ±2 37. x = ±3, x = 0,/(0) = -j
17. x = ±2 39. x =0
19. x = 0 41. continuous
21. x = 0 43. continuous

2.4 Definition of the First Derivative

In this section the first derivative of a function is defined and several interpretations
of the first derivative are considered. The first derivative of a function at a point is the
slope of the function at that point. This concept is defined precisely as follows:

The slope m of a straight line is defined as the tangent of its angle of inclination or,
equivalently, as the ratio of the change in vertical distance (rise) to the change in
horizontal distance (run) as a point moves along the line in either direction (see
Figure 2.13).

y2 - yi Ay
m = tan 6
X2 — Xj Ax

The slope of any given straight line is a constant—that is, the rate of change of y as
x changes is constant over the length of the line. However, for other curves the slope
is not constant and must be determined for each particular point of interest.
Suppose that (xls yx) and (x2, y2) are any two points on the curve y = /(x). Then
the slope of the line (called a secant) joining (x1? yj and (x2, y2) is

y2 - yi Ay
m sec
x2 — Ax

Figure 2.13

y
162 / Differential Calculus: Functions of One Variable

Suppose now that the point (x1? y^ is held fixed while the point (x2, y2) is moved
along the curve y = f(x) toward the point (xl5 yt); as the point (x2, y2) is moved
along the curve f(x), the slope of the line joining (x1? yx) and (x2, y2) will, in general,
vary. However, it may happen, and does happen for most curves encountered in
practice, that as the point (x2, y2) moves closer and closer to the point (x1? yx), the
slope of the secant line varies by smaller and smaller amounts and, in fact,
approaches a constant limiting value. When this happens, the limiting value is said to
be the slope of the tangent to the curve at (xl9 yj or the slope of the curve at (xls yx)
[see Figure 2.14].
More concisely, if, as the point (x2, y2) approaches the point (xl9 yj along the
curve y = f(x), the slope of the secant line approaches a constant limiting value, then
this limiting value is said to be the slope of the tangent to the curve at (x1? yx) or,
briefly, the slope of the curve at (xl9 yj, or the slope of/(x) at (xl9 yx). That is,
Ay
lim msec lim = slope of/(x) at (xx, yx)
Ax-» 0 Ax->0 Ax
The slope of a function at a given point is the first derivative of the function at the
given point.
It is customary in defining the first derivative to use the notation (x, y) for the
stationary point (xls yx) and the notation (x + Ax, y + Ay) for the moving point
(x2, y2). Then

d-f = lim ^ = lim /-(* + -^

"■X Ax-»0 Ax->0 Ax

is the first derivative with respect to x of the function y = /(x). This limit may exist
for some values of x and fail to exist for other values of x. At each point (x, y) where
this limit does exist, the function y =/(x) is said to have a derivative or to be
dy
differentiable and — is said to be the first derivative or the derivative of y =/(x). The

process of obtaining the first derivative of a function is referred to as differentiation.

Various types of notation, in addition to are used to denote the first derivative of
dx
y =/(x) with respect to x. The most common of these are

/' M / Jx(y) Dxy Dx(y)

Figure 2.14

(x2, y2) or (x + Ax, y + Ay)

x
 2.4 Definition of the First Derivative / 163

It should be noted that the first derivative of a function with respect to x is, in
general, another function of x which must be evaluated for particular values of
interest. This corresponds to the fact stated above that, except for straight lines,
curves do not in general have the same slope at different points.
From the definition of a derivative and the properties of limits, rules for obtaining
the derivatives of various types of functions can be obtained. These are given in a
later section. For illustration, the derivatives of several relatively simple functions are
obtained directly rather than by using the formulas.
The steps for obtaining a derivative directly are as follows:
1. Consider the function y =f(x).
2. Give increments to x and y to obtain

k + Ay =/(x + Ax)

3. Subtract to obtain

Ay =f(x + Ax) -/(x)

4. Divide by Ax to obtain

Ay = f(x + Ax) -f(x)

Ax Ax

5. Take the limit to obtain

d-f = !im^= li
dx a*-»o Ax ajc-*o Ax

EXAMPLE

Find the first derivative of y = 4x + 1.

dy = Hm f(x + Ax) -f(x)

dx Ax-»o Ax

4(x + Ax) + 1 — 4x — 1
= lim
Ax->0 Ax

4 Ax
= lim —
Ax->0 Ax

= lim 4 = 4
Ax-*0

dy
Note that y = 4x + 1 represents a straight line and thus — is a constant.

EXAMPLE

Find the first derivative of y = x3 — 12x +13.

dy lim /(* + Ax) ~/W

dx Ax-*0 Ax

j. (x + Ax)3 — 12(x + Ax) + 13 — x3 + 12x — 13

Ax-»0 Ax

_ x3 + 3x(Ax)2 + 3x2(Ax) + (Ax)3 — 12(x + Ax) + 13 — x3 + 12x — 13

Ax-+0 Ax
164 / Differential Calculus: Functions of One Variable

[note: (a + b)3 = a3 + 3a2b -I- 3ab2 + b3]

3x2(Ax) + 3x(Ax)2 + (Ax)3 — 12(Ax)

Ax-*o Ax

= lim [3x2 + 3x(Ax) + (Ax)2 — 12]

Ajc-^O

= 3x2 - 12

dy
Note that is a function of x and can be evaluated for any value of x.
dx

EXAMPLE

Find the first derivative of y = x2 + (1/x) for x ^ 0.

dy = /(* + Ax) -/(x)

dx Ax^0 Ax
1 1
(x + Ax)2 + -A-*-
x + Ax x
= lim
Ax-*0 Ax
1 1
x2 4- 2x(Ax) + (Ax)2 + -A- “ *-
x + Ax x
= lim
Ax~*0 Ax
x — x — Ax
2x(Ax) + (Ax)2 +
x(x + Ax)
= lim
Ax-*-0 Ax
1
= lim 2x + Ax —
Ajc-aO x(x + Ax)

= 2x — for x ^ 0

dy
Note that again — is a function of x and can be evaluated for any value of x ^ 0.
dx

PROBLEMS
dy
For each of the following functions/(x), find the first derivative using the definition
dx
dy
lim f{x + AaX> ~f(x)
dx Ax-*o Ax
1
1. /(x) = x2 — X + 1 4. /(x) = x-
X

2- f(x) = K 5- f(x) = 6 - 2x3

A
6. f(x) = 5x4 — 2
3- f(x) = V2x

Answers to Odd-Numbered Problems

1. 2x - 1 5. — 6x2
1
3.
 2.4 Definition of the First Derivative / 165

INTERPRETATION OF THE FIRST DERIVATIVE

Although the geometrical interpretation of the first derivative has historical
significance and is perhaps the most intuitive, other interpretations are equally valid
and sometimes are more useful in applications.
Two of the classical interpretations of a derivative are in terms of the velocity of a
moving body and the rate of change of a function. Although both of these interpreta¬
tions originated in the study of various problems in physics and mathematics, they
have since been applied in many fields. For example, marginal analysis in economics
can be more readily understood in terms of the rate of change of a function.

Velocity of a Moving Body

Consider a body (or a particle) moving along a straight-line path. Let f denote the
time measured from some fixed instant and let s denote the distance of the particle
from some fixed origin on the line, where s is positive or negative according to the
direction of movement from the origin. Suppose that distance from the origin is given
in terms of time by a function s — f(t), called the taw of motion. [For example, think of
the body or particle as being a car that is driven along a straight road in such a way
that its distance from the starting point is given as a function of t by s — f(t).]
At a certain time tu let the particle be at a distance sx from the origin 0, and
suppose that during the following time interval At it moves a distance As farther from
the origin (see Figure 2.15).
As
If the ratio — is constant, so that equal distances are always traversed in equal
At
As
intervals of time, the motion is said to be uniform and the ratio — is called the

velocity at any instant. The term speed is used frequently to mean the magnitude—
that is, absolute value—of the velocity.
As
If, however, the motion is not uniform, the ratio — varies as At varies and is no
At
longer the velocity of the particle at any instant. It is instead the average velocity of
the particle during the particular time interval At:

average velocity during interval At = ^5

At
As
As the time interval At approaches zero, this average velocity — may approach a

limit. If so, this limit is said to be the instantaneous velocity at time tx:
i • • v As
instantaneous velocity at time f, = lim —
A!-»0 At

As
But, by definition, limA(^0 — is the first derivative of/(f) at the point t = tv Thus,
at an instant tu the velocity of a particle moving in a straight line according to the

Figure 2.15

s As

Distance
0

Time
0 t\ 11 -f- At
166 / Differential Calculus: Functions of One Variable

law of motion s = f(t), where s is the directed distance from a fixed origin and t is the
time, is given by the value of the derivative of 5 with respect to t for t = tl.

EXAMPLE

The distance of a train from its starting point when it is traveling along a straight track is given
by the equation

s = 16f2 -l- 2t

where s is distance in miles and t is time in hours. Find the distance traveled and the velocity
after 2 hours.

s = 16r2 + 21

So after 2 hours, s = 64 + 4 = 68 miles traveled.

g=lim /fr + y-W

at Ar->0 At

_ lim 16(f + At)2 + 2(t -f At) - I6t2 - It

1612 + 32r(AQ + 16(At)2 + 2t + 2At - 16r2 - 2r

Ar-0 At

= Hm 32t(Af) + 16(At)2 + 2(Af)

Af-> 0 Af

= lim [321 + 16(Af) + 2]

Af-0

= lim (32f + 2)
Af-0

and the velocity after 2 hours is

ds
= 64 + 2 = 66 miles per hour
dt t=2

ds
(The notation means the function — evaluated for t — 2.)
dt t=2 dt ’

EXAMPLE

One of the attractions at a carnival is a “prove your strength ” bell, which rings if a lever is hit
with sufficient strength to propel an iron ball up a vertical shaft to hit the bell. When the lever
is hit with a force of 50 lb, the distance of the ball from the bottom of the shaft is given by

s — 401 — 3212

where s is distance in feet and t is time in minutes since the lever was hit.
(a) If the bell is at the top of a 15-ft shaft, is a force of 50 lb sufficient to ring it?
(b) How far from the bottom of the shaft is the ball after 30 seconds, 45 seconds, 1 minute, and
75 seconds, respectively?
ds si
(a) Velocity — is zero when the ball is at the top of the shaft, so solving the equation — = 0
dt dt
provides the time the ball travels up the shaft before coming back down.
 2.4 Definition of the First Derivative / 167

fk = lim /(f + Ar)-/(Q

dt At-*o A?

_ 4Q(f + Ar) - 32(f + Af)2 - 40r + 32r2

A(-*0 Af

- lim 401 + 40(At) - 32r2 - 64t(At) - 32(Af)2 - 40t + 32(2

Al-* o Af

= lim 40(Af) - 64r(At) - 32(At)2

Af-*0 Af

= lim [40 - 64( - 32(Af)]

At-*0

= 40 - 64r

ds
If ~ = 0, t = = | minutes (or 37.5 seconds) and s = 40(f) — 32(ff) = 12.5 feet, so the force

is not sufficient to ring the bell 15 ft up the shaft. (See Figure 2.16.)
(b) If t = 30 seconds = \, s = 40(f) — 32(f) = 12 ft (on the way up)
If t = 45 seconds = f, s = 40(f) — 32(^) = 12 ft (on the way down)
If t = 1, s = 40 — 32 = 8 ft (on the way down)
If t — 75 seconds = f, s = 40(f) — 32(ff) = 0 (back down)

Figure 2.16

Rate of Change of a Function

Let p and q denote the measures of two related variables and regard q as a function of

p—that is,q=f (p). If the ratio —^ of corresponding changes in the two quantities has
Ap
the same value for all values of Ap, it is called the rate of change of q with respect to p

and q is said to change uniformly with respect to p. But if the ratio ^ is not constant
Ap

as Ap changes, q does not change uniformly, and — is then called the average rate of
Ap
change of q with respect to p over the interval Ap.

If the ratio ^ approaches a limit when Ap approaches zero, then this limit is said
Ap
to be the instantaneous rate of change of q wi th respect to p.
From the definition of the derivative, it follows that the instantaneous rate of
change of a variable quantity q with respect to a related variable quantity p is given
dq
by the derivative of q with respect to p, that is,
dp
168 / Differential Calculus: Functions of One Variable

It is customary to use the expression “ rate of change of a function ” as equivalent

to the derivative of the function.
If the variable q can be expressed as a function of the variable time r, then the

derivative — of q with respect to t gives the time rate of change of q. Thus the velocity
dt
of a particle is the time rate of change of its distance from the origin.
The interpretation of a derivative in terms of the rate of change of a function is
frequently applicable in economics, as discussed below.

2.5 Applications of the First Derivative

in Business and Economics
In this section several applications of the concept of the first derivative in business
and economics are discussed. These applications include marginal cost, marginal
revenue, elasticity, and marginal propensity to save and to consume.
It is customary in economics to describe the variation of one quantity y with
respect to another quantity x in terms of two concepts, average and marginal. The
average concept expresses the variation of y over a range of values of x, usually the
range from zero up to a certain selected value. The marginal concept, on the other
hand, concerns the variation of y “ on the margin ”—that is, for very small variations
of x from a given value. The marginal concept is thus precise only when it is con¬
sidered in the limiting sense, as the variation in x approaches zero.
The economic concepts of average and marginal variation correspond to the more
general concepts of the average rate of change of a function over an interval and the
instantaneous rate of change (that is, the derivative) of a function.
Average and marginal variations in quantities are essential considerations in the
development of both microeconomic and macroeconomic theory. Examples of the
application of the derivative in both microeconomic theory (cost, revenue, elasticity)
and macroeconomic theory (income, consumption, savings) are discussed below;
additional examples are given in later sections.

COST
Suppose the total cost y of producing and marketing x units of a commodity is given
by the function

y=m
Then the average cost per unit is

y = Ax)
X X

If output is increased by an amount Ax from a certain level x and if the corre¬

sponding increase in cost is Ay, then the average increase in cost per unit increase in
. Ay
output is —- and marginal cost is defined as
Ax

That is, marginal cost is the derivative F(x) of the total cost function y =f(x) with
respect to x and is the rate of increase in total cost with increase in output.
 2.5 Applications of the First Derivative in Business and Economics / 169

EXAMPLE

Consider the total cost function

y = 20 + 2x + 0.5x2

where y denotes total cost and x denotes quantity produced. Average cost is

y = - = — + 2 + 0.5x
X X

and marginal cost is

dy
~r~ — 2 + x
dx

Thus total cost increases as output increases; average cost per unit decreases and then in¬
creases as output increases; and marginal cost (rate of increase in total cost) increases as
output increases (see Figure 2.17).

Figure 2.17

note: Derivatives are given in these examples without any explanation of how
they were obtained. The method of the preceding section or the rules given in the
next section can be used to verify the derivations.
170 / Differential Calculus: Functions of One Variable

REVENUE
For any given demand function

y =/(*)
where y is the price per unit and x is the number of units, total revenue R is the
product of x and y—that is,

R = xy = x • /(x)

Marginal revenue with respect to demand is the derivative with respect to x of total
revenue,
dR
R'(x)
dx
and is thus the rate of change in revenue with change in demand.
Note that average revenue, or revenue per unit, also represents the price per unit,
y—that is, the average revenue curve and the demand curve are identical.
Since x and y are always nonnegative within our previously stated analytical
dR
framework, R is also always nonnegative. However, — may be positive or negative—
dx
that is, although total revenue is always nonnegative, it may increase or decrease as
demand increases.

EXAMPLE

Consider the demand function

3x + 4y = 10

where y is the price per unit and x is the number of units. Then

5 3
y=2~4X

Total revenue is %

„ 5 3 .
R = - x - - x2
2 4

Figure 2.18

R
 2.5 Applications of the First Derivative in Business and Economics / 171

clR y
Tx’y

and marginal revenue is

dR _ 5 3
dx 2 2X

(see Figure 2.18).

For the example above, note the following properties, which hold for all linear
demand functions. Total revenue increases and then decreases as quantity increases;
both average revenue and marginal revenue decrease (linearly) as quantity increases.
The average and marginal revenue curves have the same y-intercept and, apart from
this common point, the marginal revenue curve lies below the average revenue curve.
The marginal revenue curve intersects the x-axis at the output for which total
revenue is greatest and the average revenue curve intersects the x-axis at twice that
output. The slope of the average revenue curve is one-half the slope of the marginal
revenue curve.

ELASTICITY

The point elasticity of the function y = /(x) at the point x is the proportional rate of
change in y per unit change in x:

dy
Ey y x dy
Ex dx y dx
x

Note that the elasticity of a function is independent of the units in which the variables
are measured. This results from the definition of elasticity in terms of proportional
changes, which are necessarily independent of units of measurement.
The point elasticity of demand, supply, cost, productivity, and other functions is
an important concept in economic theory. Detailed examples of these and other
applications of derivatives are discussed in later sections, following the discussion of
the techniques of differentiation.
172 / Differential Calculus: Functions of One Variable

NATIONAL INCOME, CONSUMPTION, AND SAVINGS

The relationship between total national disposable income and total national con¬
sumption is often referred to as the consumption function. In simple forms of theoreti¬
cal analysis of the consumption function, it is assumed that as income increases (or
decreases) consumption increases (or decreases), but by less than the increase (or
decrease) in income; that is, the marginal propensity to consume is greater than zero
but less than 1, where marginal propensity to consume is the rate of change in
consumption as disposable income changes.
If the consumption function is given by
c = /(*)
where c is total national consumption and x is total national income (and c and x are
in the same units), then the marginal propensity to consume is

In elementary theoretical analyses of national income, the assumption is often

made that disposable income equals consumption plus savings. This is expressed as
X = c + s

The marginal propensity to consume is

dc
=/'M
dx

and the marginal propensity to save is

ds dc
dx dx

In national income analysis, investment is considered as capital formation and

represents an increase in real capital as typified by equipment, buildings, inventories,
and so forth. Investment and consumption are assumed to be related in such a way
that an initial investment expenditure may result in an increase in income of several
times that amount. A precise numerical expression for this relationship is given by
the multiplier. The multiplier is the ratio of the ultimate increase in income to the
increase in investment that gave rise to it.
The multiplier k is related to the marginal propensity to consume and is given by

dc ds
dx dx

Note that if — = 0, then k = 1; that is, if none of the additional income is spent, the
dx
dc
total increase in income is equal to the initial expenditure; if-—► 1, k -> oo; that is, if
dx
all the additional income is spent, the total increase in income becomes infinitely
large.

EXAMPLE

Suppose the consumption function is

c — 10 + 0.8x + 0.5v/x
 2.5 Applications of the First Derivative in Business and Economics / 173

Figure 2.19
c, s

dc ds
dx, dx

where c is total consumption and x is total disposable income. Then the amount saved is
s = x — c

= x — (10 + 0.8x + 0.5^/x)

= — 10 + 0.2x — 0.5^/x
Note that c is positive for all positive values of x; s is negative for small positive values of x and
positive for larger values of x; for every value of x, c + s = x. Note also that c is an increasing
174 / Differential Calculus: Functions of One Variable

function of x; 5 is at first a decreasing and then an increasing function of x. (See Figure 2.19.)
The marginal propensity to consume is

The marginal propensity to save is

and

-—I—— = 1. The multiplier is

dx ax

dc ds
Note that — and k are decreasing functions of x and — is an increasing function of x. (See
dx dx
Figure 2.19.)

note: For certain economic analyses, it is more appropriate to consider income

as a function of consumption rather than consumption as a function of income. In
this case, income is plotted on the y-axis and consumption is plotted on the x-axis.

2.6 Rules for Differentiation

The procedure for obtaining derivatives directly was illustrated for several simple
functions in previous sections. Although straightforward, this procedure is tedious
and likely to result in errors, especially for more complicated functions. In practice,
derivatives are obtained using rules or formulas for differentiating particular types of
functions. These rules can be derived using the general procedure, as shown in
Technical Note I.
In this section, rules for differentiating polynomial, algebraic, logarithmic, expon¬
ential, trigonometric, inverse, and composite functions are stated and illustrated.

ALGEBRAIC FUNCTIONS
The formulas in this section apply to the products, quotients, powers, and roots of
differentiable functions.

Rule 1. The derivative of a constant is zero: If y = c,

EXAMPLE

10 = 6, d/
dx
= 0.
 2.6 Rules for Differentiation / 175

EXAMPLE

if y o.

Rule 2. The derivative of the nth power of a variable is the product of n and the
(n — l)st power of the variable: If y = x",

dx

for any positive integer n.

EXAMPLE

EXAMPLE

_ vl/4 dy 3/4
If y = = ix
dx

EXAMPLE

EXAMPLE

5
if y 3^*

Rule 3. The derivative of the product of a constant and a differentiable function

is the product of the constant and the derivative of the function: If y = cu, where
u =f(x) is a differentiable function of x,

dy du
dx dx

EXAMPLE

If y - lOx, f = 10.
dx
176 / Differential Calculus: Functions of One Variable

EXAMPLE

Uy = 6x.

EXAMPLE

EXAMPLE

If y — — 3x 1, = — (— l)(3bc 2 = 3x 2 = -i .
dx x

Rule 4. The derivative of the sum of a finite number of differentiable functions is

the sum of their derivatives: If y = u + v, where u = f(x) and v = g(x) are differen¬
tiable functions of x,

dy du dv
dx dx dx

In general, if y — YX=i ui> where =f(x) are differentiable functions of x for i = 1,

2, ..., n, then

dy " dut
dx dx

EXAMPLE

dy
If y = 3x2 + 4x + 2, 6x + 4
dx

EXAMPLE

EXAMPLE

If y = 10 — 6x~ ',2, dJx = — (—4)(6)x”3/2 = ix~312

 2.6 Rules for Differentiation / 177

EXAMPLE

d'y
If y = 6x5 + x4 + 2x3/2 + 5, 30x4 + 4x3 + 3x1/2
dx

Rule for Differentiation of Polynomial Functions

These four rules for differentiation can be summarized and generalized as the rule for
differentiation of polynomial functions. A single term of the form cx", where c is a
constant and n is zero or a positive integer, is called a monomial in x. A function that
is the sum of a finite number of such monomial terms is called a polynomial or a
polynomial function in x.
The derivative of a monomial in x is

— (cx") = cnxn~ 1
ax

and the derivative of the sum of a finite number of such terms (that is, the derivative
of a polynomial) is the sum of the derivatives of the terms:

d
( Zc.oc”') = Xc. HjX"1 1
dx \,
\ i -= 1 / i= 1

PROBLEMS
Find the first derivative with respect to x for each of the following functions y = f(x).

1. y = 6x + 2 7. y = 5x3/5 + 6x
2. y = x3 + x 8. y = lx2 + 8x~ 1/2 + 2
3. y — 4x2 + 2x 9. y = 9x3 + 5x" 1/4
4. y = x1/2 + 4 10. y = 10x4 + 2x5/4
5. y = 3x2 — x1/3 + 2 11. y = 12x3/2 + 6x1/2 + 2
6. y = 5x“1/3 + 5 12. y = x5 + 3x~7/3 + 5

Answers to Odd-Numbered Problems

1. 6 7. 3x_2/5 4- 6
3. 8x + 2 9. 21 x2 -|x~5/4
5. 6x — ix~2/3 11. 18x1/2 + 3x_ 1/2

Rule 5. The derivative of the product of two differentiable functions is equal to

the first function times the derivative of the second function plus the second function
times the derivative of the first function. Similarly, the product of more than two
differentiable functions is the sum of the products of the derivative of each function
and the other functions. If y = uv, where u =f(x) and v = g(x) are differentiable
functions of x,
178 / Differential Calculus: Functions of One Variable

In general, if y = ^]"=1 ut, where ut =/(x) are differentiable functions of x for i = 1,

2, ..., n, then

fi(x) n uj
j= i
jfi

EXAMPLE

If y = (x3 + 4)(x + 3),

dy_ du
+ v
dx dx

and

du
3x2
dx

l(x‘-‘)= 1
dx

jx = (*3 + 4)(1) + (x + 3)(3x2)

= x3 + 4 -f 3x3 + 9x2

= 4x3 + 9x2 + 4

EXAMPLE

If y = (^x + 3)(x2 + 6),

dy . dv du

and

div
dx

dv
2x
dx

dy
(y/x 4- 3)(2x) + (x2 + 6)(^x_1/2)
dx

= 2x3/2 + 6x + ix3/2 + 3x 1/2

= fx3/2 + 6x + 3x~1/2

EXAMPLE

If .y = (3x + 7)(x 2 + 8),

A,3x+1) = 3

7-(x-2 + 8)= — 2x~3

dx
 2.6 Rules for Differentiation / 179

and

^-(3x+7)(-2x 3) + (x 2 + 8)(3)

= — 6x~2 - 14x~3 + 3x~2 + 24

= —3x-2 - 14x“3 + 24

EXAMPLE

If y = (x + 3)(2x + 3)(x2 + 1),

^ + 3> = ‘
^(2x + 3) = 2

4- {x2 + 1) = 2x
dx

and

^ = (l)(2x + 3)(x2 + 1) + (2)(x + 3)(x2 + 1) + (2x)(x + 3)(2x + 3)

= (4x + 9)(x2 + 1) + 2x(2x2 + 9x + 9)

= 4x3 + 9x2 + 4x + 9 + 4x3 4- 18x2 + 18x

= 8x3 + 27x2 -I- 22x + 9

Rule 6. The derivative of the quotient of two differentiable functions is the

quotient of the product of the denominator and the derivative of the numerator
minus the product of the numerator and the derivative of the denominator divided
by the square of the denominator: If y = u/v, where u =f(x) and v = g(x) are differ¬
entiable functions of x,

du dv
dy dx dx
dx v2

EXAMPLE

x2 - 4x + 1
lfy= ■x—6-,

u = x2 — 4x + 1 v = x —6

du dv
2x — 4 1
dx dx
180 / Differential Calculus: Functions of One Variable

and
dy I(x — 6)(2x — 4)i-Cx2 -4x+ 1)(1)
dx (*: — 6 r
2x2 — 16x + 24 — x2 + 4x — 1
(x - 6)2
x2 — 12x -I- 23
(x - 6)2

EXAMPLE

Uy =

u = 4 v = x
dv
t- = o = 6x;
ax dx
and
dy _ (x6)(0) - 4(6x5)
dx (x6)2
— 24x5

-24

note: This example could also be solved using Rule 2, as follows:

y = 4x-t>
dy
= — 24x - 7
dx
-24

EXAMPLE

x3 + 16
\fy = ..2 ’

— (x3 + 16) = 3x"

dx

Tx{x2) = 2x

and
dy (x2)(3x2) — (x3 -I- 16)(2x)
dx x4
3x4 - 2x4 - 32x

x4 — 32x

x3 - 32
 2.6 Rules for Differentiation / 181

note: This example could also be solved using Rule 2, as follows:

y = x 4 16x~ 2
-

dy
32x~3
dx

32

EXAMPLE

x3 + 3x 4- 15
Uy =
x2 4- 2x 4- 2

d_
(x3 4- 3x 4- 15) 3x2 4- 3
dx

d
(x2 + 2x 4- 2) 2x 4- 2
dx

and

dy _ (x2 4 2x 4 2)(3x2 4 3) — (x3 4 3x 4 15)(2x 4 2)

- - - - - -

dx (x2 + 2x 4 2)2
-

3x4 4- 6x3 + 6x2 + 3x2 + 6x -f 6 — 2x4 — 6x2 — 30x — 2x3 — 6x — 30

(x2 + 2x 4 2)2
-

x4 + 4x3 + 3x2 — 30x — 24

= (x^+ 2x + 2)2

Rule 7. The derivative of the nth power of a differentiable function is the product
of n, the (n — l)st power of the function, and the derivative of the function: If y = un,
where u =f(x) is a differentiable function of x and n is any real number (positive or
negative, integer or noninteger),

dy du
nun
dx dx

Special Case: If u =f(x) = x, then un = xn and

i
nxn

which is Rule 2.
182 / Differential Calculus: Functions of One Variable

EXAMPLE

If y = (x2 + 3)3,

d du
nun
dx dx

u — (x2 + 3)

n = 3
'O

du
2x
dx

and

^ = 3(x2 + 3)2(2x)
dx

= 6x(x2 + 3)2

EXAMPLE

If ^ = (x + 3)~1/3,

d , . du
— {u ) = nu 1 —-
dx dx

u = (x + 3)

n = -i

du
= 1
dx

and

dy
= (_i)(x + 3)-4/3(l)
dx

-1
3(x + 3)4/3

EXAMPLE

but note that u = (x + l)/(x — 1) is a fraction and thus Rule 6 must be used to obtain dx.

du I(x i)l-( x + 1)
dx |(x - 1 )2

-2
 2.6 Rules for Differentiation / 183

and

dy -2i
fx+ 1\|I -2 \
dx !L-1/1 (*-i)i2/
— 4(x + 1)
(* - l)3

EXAMPLE

If y = {x+ l)2(x2 + I)”3,

dy x dv du
Txiuv) = uTx + vTx

du
but note that u = (x + l)2 and v = (x2 + 1) 3 and thus Rule 7 must be used to obtain — and
dx
dv
dx

(x + l)2 — 2(x + 1)
dx

^ (x2 + 1) 3 = -3(x2 + 1) 4(2x)

= — 6x(x2 + 1)~4

“ = (x + 1)2( —6x)(x2 + 1) 4 + (x2 + 1) 3(2)(x + 1)

= (x + l)(x2 + 1)“ 4[ —6x(x + 1) + 2(x2 + 1)]

= (x + l)(x2 + 1)~4( —4x2 — 6x + 2)

— 2(x + l)(2x2 + 3x — 1)
= (x2 + l)4

EXAMPLE

Ify = (x2-x)-2,

~ = ~2(x2 — x)~3(2x — 1)
dx

2(2x - 1)
(x2 - x)3

_ 2(1 - 2x)
” x3(x - l)3

EXAMPLE

If y = x2(x + l)"1,
184 / Differential Calculus: Functions of One Variable

dj- = 2x(x 4- 1) 1 + x2( — l)(x 4- 1) 2

= (x 4- l)~2[2x(x 4- 1) — x2]

= (x 4- l)~2(2x2 4- 2x — x2)

= (x 4- l)~2(x2 4- 2x)

x(x 4- 2)
“ (x + l)2

EXAMPLE

If y = (x 4- x~ *)2,

— 2(x 4- x_ 1 )(1 - x-2)

= 2(x - x_1 4- x_1 - x-3)

= 2(x — x“3)

= 2x~3(x4 - 1)

2(x4 - 1)
— v3

EXAMPLE

If y = 2x(3x2 4- 1)~ \

^ = 2(3x2 4- 1)~ 1 4- 2x( — l)(3x2 4- l)_2(6x)

= 2(3x2 4- l)~2(3x2 4- 1 — 6x2)

= 2(3x2 4- 1)~ 2( 1 — 3x2)

_ 2(1 - 3x2)
” (3x2 4- l)2

PROBLEMS
Find the first derivative with respect to x for each of the following functions y = f(x).

1. y = 2x3 4- 4x2 — 5x + 8 7. y = (x3 - 3x)4

2. y — — 5 4- 3x — fx2 — 7x3 8. y = (x + x-1)2

3. y = (2x2 4- 4x - 5)6 9. y = {x- 1 )3(x 4- 2)4

4- y = W2 + W12 10. y = (x4 2)2(2 - x)3

5. y = (1 — x2)1/2 11. y = (x 4- l)2(x2 4- 1)~3

, 6 4 3 2x + 1
6. y = - 4- - -j 12.
X X X y = x^l
 2.6 Rules for Differentiation / 185

2x
13. y = 21. y
x2 4- 1 "(x5 4- 10)1/8

1x + r 22. y (x 4- 2)3(x2 4- l)"1

14. y =
vx — 1
lx2 4- 1Q\10
23. y
15. y (x2 - x)~2

16. y x2(x 4- 1)“1

24. y =
(x3 - 4)2/3
_1 _
17. y =
(x2 — 9)1/2 (x2 + 1)1/2
25. y =
(2x + 4)1/4
1
18. y =
(16 — x2)1/2 26. y = (x + 2) 3/2(3x2 -I- 1)
x 27. y = x3(x2 -f 3)~1
19. y =
(x 4- 1)1/2
28. ^ = x6 + x4/3 + 6x1/2
(x2 4- 2)1/2
20. y =
x

Answers to Odd-Numbered Problems

1. 6x2 + 8x — 5 17. — x(x2 — 9)~3/2
3. 24(x + l)(2x2 + 4x - 5)5
19 x+2
2(x + 1)3/2
5. — x( 1 - x2)- 1/2
21. (x5 + 10)“9/8(Jx5 + 20)
7. 12x3(x2 — l)(x2 — 3)3

9. (x - 1 )2(x + 2)3(7x + 2)

11. —2(x + l)(x2 -I- l)-4(2x2 + 3x — 1)

25. i(x2 + 1)“ 1/2(2x + 4)_5/4(3x2 + 8x -1)
1 - x2
13 _
(x2 4- l)2 27. x2(x2 4- 9)(x2 + 3)-2

15. — 2x~3(x — l)~3(2x — 1)

LOGARITHMIC FUNCTIONS
The notation “ In” denotes natural logarithms; the base of the natural logarithms is
defined by e = lim„^00 (1 + (1 /n))n, and is approximately 2.718. The notation “log”
denotes common logarithms (base 10); if any other base is meant, it is specified.

Rule 8. If y = loga u, where u =f(x) is a differentiable function of x,

dy loga e du
dx u dx

EXAMPLE

If y=log(^rr)’
d / \ = (x + 1) - x
(Rule 6)
dx \x + 1/ (x 4- l)2
186 / Differential Calculus: Functions of One Variable

and
= log e /_1 \
dx x \(x + l)2/
x 1

log e
x(x + 1)

EXAMPLE

If y = >g x,
dy 1 log e]
= ^ (log x)“1/2|
dx x

log e
2xv1og

EXAMPLE

If y = (log x2)3,

6(log x2)2 log e

x

Special Case: If y = In u, where u =f(x) is a differentiable function of x,

dy 1 du
dx u dx

EXAMPLE

If y = In y/‘x2 - 1,

=-H*2 - 1 )-1,J(2x)

= x(x2-l)-'«
and

x2 - 1
 2.6 Rules for Differentiation / 187

EXAMPLE

If> =-,
x

d(\n x) = —
x

and

. |-)x — In x
dy \xj
(Rule 6)
dx x2

1 — In x

EXAMPLE

If y = In (x + ^/x2 -I- 4),

d
(x + ^/x2 + 4) = 1 + ^(x2 + 4) 1/2(2x)
dx
1/2
= 1 + x(x2 + 4)
and

dy =_
1
[l + x(x2 + 4)-1/2]
dx x + y x2 + 4
x
1 +
X2 + 4
x + ./x2 + 4

x 4- x/x2 + 4
x2 + 4

x + V*2 + 4
1
X2 + 4

EXPONENTIAL FUNCTIONS
Rule 9. If y = au, where u =f(x) is a differentiable function of x,

dy du
au In a
dx dx

EXAMPLE

Uy = 2~x,

2~x In 2
188 / Differential Calculus: Functions of One Variable

EXAMPLE

If y = 10x2-x,

d-f = 10*’-*(ln 10)(2x - 1)

u

= (2x — l)10x2-xln 10

EXAMPLE

If y = axxa,

^ = ax(In a)xa + axaxa~1

dx

= axx°~1(a + x In a)

Special Case: eu, where u =f{x) is a differentiable function of x,

dy
dx

EXAMPLE

dy _ xex — ex
dx x2

_ ex(x — 1)

EXAMPLE

If y = 10ex2 + \

% = 10eX’+4(2x)

= 20xe*2 + 4

EXAMPLE

e —e
X — X

lfy =
ex + e

(ex — e x) = ex + e x
dx

d
(ex -f e x) = ex — e~ X

dx
 2.6 Rules for Differentiation / 189

and

dy I(ex + e x)I2 -1(ex - e x)2

dx (e* + e -x)\2

e2x 4- 2 + e 2x — e2x + 2 — e 2x _ 4
(ex + e“*)2 (<?* + e~x)2

Rule 10. If y = uv, where u =f(x) and v = g(x) are differentiable functions of x,
differentiate In y = v In u to obtain

dy du dv
vu v — — + uL In u —
dx dx dx

Note that Rule 9 is a special case of Rule 10, where u = a. Some algebraic functions
are also easier to differentiate if the logarithm of each side is differentiated and the
resulting equation is solved for the derivative of the function. (See the fourth example
below.)

EXAMPLE

If y = xx\
In y — x2 In x

1 dy 1
2x In x + x'
ydx ,x.

= x(2 In x + 1)

— = xx' + *(1 + 2 In x)
dx

EXAMPLE

If y =
In y — ex In x

1 dy IT
— = ex In x + ex -
ydx \X)

1
— ex\ In x +

dx \X ;

EXAMPLE

If y = (x4 4- 5y\
190 / Differential Calculus: Functions of One Variable

In y = x2 In (x4 + 5)

1 dy 4x3 '
2x In (x4 + 5) + x2|
y dx x4 + 5.

2x^
= 2x In (x4 + 5) +
x4'+ 5

dy
= 2x(x4 + 5)x2 ![(x4 + 5) In (x4 + 5) + 2x4]
dx

EXAMPLE

In y = -j In (x + 1) — -J In (x — 1)

1 dy__1_1_
y dx 3(x + l) 3(x — 1)

x — 1 — (x + 1)
3(x2 — 1)

-2
~ 3(x2 - 1)

-2 /x+n
31(X2 - 1) \X - 1/

Find the first derivative

■ 1/x
1. y = log(l - 21) 13. y = e
14. y = x'
1 > - - (rrs) x2+ 1
15. y = xe
3. R = log,, (a2 — x2)3
ifi <x + 1>6
2u y~ (x2 + 2x + 2f
4. t = 6 -

r2_>c2 + 3
5. y — ex In x 17. y = 2x2ex

6. v =
jx
—
+ 1 18. y = log5 (x3 + x2)6
3 /x + 2 19. y = x2*4 + x
7. t = £>ln *
20. y log, (b - X3)2

8- y = axex /x + 3\2
21. y
9. y = log (x3 3x),1/3 F

10. y = t2 In t 22. y c>ln x2

1 -J2 23. y = 16 c2 —2x

11. y = 2 In
t 24. y = xx
12. y = 253x3“6
 2.6 Rules for Differentiation / 191

35 y —— £?x 4x 3 31. y = ie31nx

26. y = eln{x + 3) 32. y = xeln (x2 + 5^

27. y = log8 (6 - x2)4 33. y = xV2+4x+2

34 ^ _ gin (x4 + 3x2+10)
28. y = x*3+2

29. _y = 2 In (x3 4- 4x2)1/4 35. y = (x + l)x

30. y = (x2 + 4)Vx2 + 1 36. 4= (x+ l)x3 + 1

Answers to Odd-Numbered Problems

— 2 log e 17. 4x(x2 4- l)ex2 + 3
1.
l-2t 19. x2x4+x[2x3 +1-1- (8x3 + 1) In x]
6x loga e
2 2~
— 8(x + 3)
x — a
(x-1)3

5. — (1 4- x In x) 23. 2(x - 1)16x2_2x In 16

x
25. (2x + 4)ex2+4x+3
7. 1
— 8x log8 e
27
9. -1)1 log e 6 — x2
xl+ --3)1
3x + 8
29.
a - -1 2(x2 + 4x)
r(r2-l)
31. x2
13. \e~llx 33. 2x(x + 1 )2ex2+4x+2
X
35. (x + l)x_1[x + (x + 1) In (x + 1)]
1, +

TRIGONOMETRIC FUNCTIONS
Rule 11. If y = sin u, where u =f(x) is a differentiable function of x,

dy du
— = cos u —
dx dx

EXAMPLE

If y = sin (x2 + 9),

-j- = 2x cos (x2 + 9)

dx

EXAMPLE

If y = sin2 3x,

= 2(sin 3x)(cos 3x)(3)

tl Tv

= 6 sin 3x cos 3x
192 / Differential Calculus: Functions of One Variable

Rule 12. If y = cos u, where u = f(x) is a differentiable function of x,

dy . du
— = —sin u —
dx dx

Rule 13. If y = tan w, where u = f(x) is a differentiable function of x,

dy 9 du
— = sec u —
dx dx

Similarly,

1
y = esc u — —-
sin u

1
y = sec u =-
cos u

1
y = cot u = ——
tan u

can be differentiated to obtain the derivatives of the cosecant, secant, and cotangent
functions.

d y du
— (esc u) = — CSC u cot u —
dx dx

d , du
— (sec u) = sec u tan u —
dx dx

d ,
— (cot u) = — CSC u-j-
dx dx

EXAMPLE

If y = cos x + sec x,

dy
— = —sin x + sec x tan x
ax

= sin x(sec2 x — 1)

EXAMPLE

If y = (1 + cot x)2,

^ = 2(1 + cot x)( — esc2 x)

= — 2 esc2 x( 1 -|- cot x)

 2.6 Rules for Differentiation / 193

EXAMPLE

If y = tan2 (x + 3) + sec2 (x + 3),

~ = 2 tan (x -I- 3) sec2 (x + 3) + 2 sec2 (x + 3) tan (x + 3)

= 4 tan (x + 3) sec2 (x + 3)

EXAMPLE

if y = (sin x + cos x)2,

dy
= 2(sin x + cos x)(cos x sin x)
dx

= 2(cos2 x — sin2 x)

EXAMPLE

cot (x2 + 2)
If y
sec (x2 + 2) ’

dy — 2x esc2 1(x + 2)1 sec |(x2 + 2)1 — 2x cot 1(x2 + 2)1 sec 1(x2 -f 2) tan |(x2 + 2)
dx sec2 |(x2 + 2)1

— 2x[csc2 (x + 2) + 1]
sec (x2 + 2)

EXAMPLE

If y — tan (x3 + 2x + 3) sin (x3 + 2x + 3),

^ = (3x2 + 2) sec2 (x3 + 2x + 3) sin (x3 + 2x + 3)

+ (3x2 + 2) tan (x3 + 2x + 3) cos (x3 + 2x + 3)

= (3x2 + 2) sin (x3 + 2x + 3)[sec2 (x3 + 2x + 3) + 1]

PROBLEMS
Find the first derivative of each of the following functions.

. CSC X 5. y = sin 3x tan 3x

*' ^ = —

6. y = cot (x2 + 1)
^ sin x — cos x
7. y = x + cot x
2-^ =-X-

0 x -F sin x
3. y = x2 cot x 8 - y =---

4. y = tan x esc x
194 / Differential Calculus: Functions of One Variable

9. y = cos2 2x 18. y = tan x + sec 2x

10. y = sin x cos x sin x

19. y =-
sec x
11. y sin2 x + cos2 x
20. y = cos x esc x
12. y x sin x
21. y = sec2 3x
sin x
13. y
22. y =
sec x
14. y x — cos x
1
tan x 23. y —
15. y 1 + cot X
W+W
cot X
16. y sin x2 24. y =
1 + X'*

17. y tan 3x

Answers to Odd-Numbered Problems

1 —esc x(l + x cot x) (1 -I- x) sec2 x — 2 tan x
15.
L X2 (1 + x)3

3. 2x cot x — x2 esc2 x 17. 3 sec2 3x

5. 3 sin 3x(l + sec2 3x) 19. cos2 x — sin2 x

7. 1 — esc2 x 21. 6 sec2 3x tan 3x

9. —4 cos 2x sin 2x 1
23.
(sin x + cos x)'
11. 0

x cos x — sin x
13.

INVERSE FUNCTIONS
Rule 14. The derivative of the inverse of a function is equal to the reciprocal of the
derivative of the function. If y=f(x) and x ~ g(y) are inverse differentiable
functions,
dy 1
dx dx
dy

or

df(x) = 1
dx dg(y)
dy

EXAMPLE

If x = y 4- ^y3 + $y5, find

dl
dx

dx
1 + y2 + >’4
dy
 2.6 Rules for Differentiation / 195

dy 1 1
dx dx 1 + y* 1 2 + yA
dy

EXAMPLE

If x = In (y* + 3y2),

dx 4 y3 + 6 y
dy y4 + 3 y2

_ 4y2 + 6
y3 + 3y
dy _ 1 _ y3 + 3y
dx dx 4y2 + 6
dy

EXAMPLE

Ifx = e'y2+4y+2,
/j y
-f- — ( — 2y + 4r)e~y2+*yJr2
dy
dy _ 1 _^2-4>--2
dx dx 4 — 2y
dy

EXAMPLE

If x = tan2 (y3 + a),

dx
6y2 tan (y3 + a) sec2 (y3 -I- a)
dy

dy 1
cot (y3 + a) cos2 (y3 + a)
dx d^
dy

PROBLEMS
Find the first derivative with respect to x of each of the following functions x = f(y).

1. x = y2 + 3y + 2 4. x = In (y5 — 6)
2. x = y3 — 2y3/2 + 6y 5. x = cos2 y + sin y
3. x = 5y4 — ey 6. x = y sec y

Find the first derivative with respect to y of each of the following functions y = /(x).

9. y = tan x sin x

8. y = In 10. y =
COS X
196 / Differential Calculus: Functions of One Variable

Answers to Odd-Numbered Problems

1 sec y
1. 5.
2y + 3 1 — 2 sin y

1 7. -e*
3.
20 y3 e*
CSC X
9. --2
1 + sec x

COMPOSITE FUNCTIONS
If y is a function of w and u is a function of x, then y is a function of a function or a
composite function; the derivative of y with respect to x is the product of the deriva¬
tive of y with respect to u and the derivative of u with respect to x.

Rule 15. If y = <&(u) and u = F(x), that is, y = <D[F(x)] =/(x),

dy dy du
dx du dx

(This is sometimes called the chain rule of differentiation.)

Note that Rules 7, 8, 9, 10, 11, 12, and 13 are special cases of Rule 15.

EXAMPLE

If y = w2/3 and u = x2 + 1, find

dy
iu-1'3
du

3(x2 + 1)1/3

du
2x
dx

dy dy du 4x
dx du dx 3(x2 + 1)1/3

Alternatively,

y = w2/3 = (x2 + 1)2/3

g=j(x2 + l)-I'3(2x)

4x
“ 3(x2 4- 1)1/3

EXAMPLE

If y = —- and z = -I- 1, find .

z2 + 1 v du
 2.6 Rules for Differentiation / 197

dy _ 2z(z2 4 1) — z2(2z)
dz (z2 4- l)2

2z
= (z2 + l)2

2^/2m 4 1
(2m 4- 2)2

\j2u + 1
2(m 4 l)2

g = i(2U+l)-(2)

1
y/2ll 4 1

dy _ dy dz 1
du dz du 2 (u 4- l)2

Alternatively,

z2 2m 4 1
^ z2 4 1 2m + 2

2(2u + 2)1 ~ 1(2» + 1)(2)

1(2.m 4 2)I2

2
(2m 4 2)2

1
2(m + l)2

EXAMPLE

If y = In (w3 — 1) and w = z2 — 1, find .

dz

dy 3w2 3(z2 — l)2

dw w3 — 1 (z2 — l)3 — 1

dw
Tz = 2z

dy dy dw 6z(z2 — l)2
dz dw dz (z2 — l)3 — 1

Alternatively,

y = In (w3 - 1) = In [(z2 — l)3 — 1]

dy 3(z2 — l)2(2z)
dz = (z2 — l)3 — 1

6z(z2 — l)2
= (z2 - l)3 -1
198 / Differential Calculus: Functions of One Variable

PROBLEMS

Determine for each of the following functions.

dx

1. x = yn- 2y 12. x =
In y

2. x = </4 — 9y y
13. x = y2c_>1
3. x = (2 - 3y2)3
14. x = In (y2cy)
/ fr\2
4. x = a —
yj 15. x = In (ycos 2y)

5. x = yja2 + y2 4
16. x — ,
sec y
a -y
6. x =
a + y i /1 + sin y\1/2
17. x = In -;—— I
\ 1 — sin yf
7. x = + fry
18. x = (sin 2y)1/2
fr
8. x = - wa — y
a 19. y = a6; a = 1 + 2yx

9. x = In (ay2 + fr) 20. y = a sin a; u — In x

e* - 21. y — u2 cos u; u = ax2

10. x =
+ e y
__ a — a fr — x
22. y = —:— ; u =
a + a fr + x
11. x = --

23. If y = x4 4- 5 and x = log z, find .

dz
dy
24. If y = e3“ and u = 2x2 — 3x, find .
dx
du
25. If u = In (y + 4) and y = x2, find
dx
4v 2 dx
26. If x = 1 -- and y = a3 + 10a, find — .
y + 6 du

27. If y = e* + 6 and t = In (x2 + 6x), find dl_

dx

28. If y = ——and t = x2 — 4, find ™.

t + 4 dx

Answers to Odd-Numbered Problems

1. — x 15. — cot 2y
1 17. cos y
3.
18>>(2 - 3y2)2 6(i + 2 yif
10
5
(a2 + y2)1/2 yc
a2 + 2y2
21. 2a2x3(2 cos ax 2 — ax2 sin ax
2(a + by)1'2 4(log z)3
7.
2a + 3fry 23. - log e
z
ay2 + fr
9. 2x
2 ay 25.
x2 + 4
11. -V
27. 2x 4- 6
 2.6 Rules for Differentiation / 199

Summary of Procedure for Differentiation

1. Determine whether the entire function to be differentiated is a sum or difference of
functions; a product or quotient of functions; a logarithmic, exponential, or trig¬
onometric function of a function; a power of a function; a composite function; or
some combination of these.
2. Differentiate using the appropriate rule for the entire function and the appropriate
rules for the different parts of the function.

EXAMPLE

Find the first derivative of y = x2 + In x3.

The entire function is the sum of two functions

y = u(x) + v(x) where u(x) = x2

y(x) — In x3
dy du dv
Thus (Rule 4).
dx dx dx

u = x2 is differentiated using Rule 2, and v = In x3 is differentiated using Rules 2 and 8.

du , dv j
, = 2x and — = -
dx dx x

Thus

dy 3
2x +
dx x

EXAMPLE

Find the first derivative of y — ex2 + 2(x 4- l)2.

The entire function is the product of two functions

y = u(x) • v(x) where u(x) = exl+2

v(x) = (x + l)2
dy dv du
Thus —— = u ——h v —— (Rule 5).
dx dx dx

u = ex2 + 1 is differentiated using Rules 1, 2, 4, and 9, and v = (x + l)2 is differentiated using

Rules 1, 2, 4, and 7.

du
= 2xcx2 + 2
dx
dv
- 2(x + 1)
dx

Thus

dy 2„x2 + 2
= 2ex +2(x + 1) -F 2x(x + \)2e
dx
= 2(x 4- l)e*2 + 2(l + x2 + x)

= 2(x + l)(x2 + x + l)e*2+2

200 / Differential Calculus: Functions of One Variable

EXAMPLE

Find the first derivative of y = xtanx.

The entire function is a function raised to a functional power

y = u(x)t,(JC) where u(x) = x

v(x) = tan x

dv , du ,, v „ dv
Thus — — vu 1 — + (In u)u — (Rule 10).
ax ax ax

u = x is differentiated using Rule 2 and v = tan x is differentiated using Rule 13.

du dv
and — = sec" x
dx dx

Thus

dy
= (tan x)x(tanx) 1 + (In x)xtanx(sec2 x)
dx

_ xtan x j *al1 X _)_ sec2 x Jn x

PROBLEMS
Obtain the first derivative of each of the following functions.

13. y = In3 x
1. y =
x + 1
14. y — log
e+4
2. s =
e
15. y = 10"*
3. y = In (x 4- y/\ + x2)
16. y = In (ax2 + b)
2
4- y = 17. 5 = e~‘ cos It
ex
18. y = i tan3 9 - tan 0 + 9
x3 + 1
5 .y = ir. a + bt + ct2
x
19. s = —
3
6. y =
x2 + 2 a + bx + cx2
20. z =
1
7. s =
1 - 21 21. z = a2y

22. y = exl
8. y =
4 — x2 23. y = x*
9. y = eax sin bx 24. y = xn(a + bx)‘

10. >’ = x 25. y — xsm x

x2 T 1
26. y = xvX
ax + b
11. y =
bx + d 27. , = (f

12. y = In I tan 28. y = (cos x)J

 2.6 Rules for Differentiation / 201

Answers to Odd-Numbered Problems

15. 10"xn In 10
i. - 2
(x + l)2
17. — e_f(cos 2f + 2 sin 2r)
1
19. -iar312 +ibr1/2 +ict112
\/l +
21. 2a2* In a
2x3 - 1
5.
23. x*(l + In x)

sin
7. 25. x sin x | + cos x In x I
(1 - 2r)2
9. - eax(a sin bx + b cos bx)

ad — be
11.
(cx + d)2

3 In2 x
13.

SUMMARY OF RULES FOR DIFFERENTIATION

Algebraic Functions

Rule 1. If y = c, , = 0.
dx

Rule 2. If y — xn, —1- = nxn~ l.

dx

dy du
Rule 3. If y = cu, where u =f(x) = c
dx dx

Rule 4. If y = u + v, where u =/(x) and v = g(x),

dy du dv
dx dx dx

Rule 5. If y = uv, where u =/(x) and v = g(x),

dy dv du
dx dx

Rule 6. If y = where u =f(x) and v = g(x),

du dv
dy dx dx
dx

Rule 7. If y = un, where u = /(x),

dy du
nu n- 1

dx dx

Special Case: If u =f(x) = x, then un = xn and ~ = nun 1 = nxn 1 (Rule 3).

202 / Differential Calculus: Functions of One Variable

Logarithmic Functions
Rule 8. If y = loga u, where u = /(x),

dy logae du
dx u dx

1 du
Special Case: If a = e, then y = In u and
u dx

Exponential Functions
Rule 9. If y = au, where u = f(x).

dy du
= au In a
dx dx

dy du
Special Case: If a = e, then y = eu and
dx dx

Rule 10. If y = uv, where u =f(x) and v = g(x),

1
du dv
+ uv In u
V-
vu
dx dx

Note that Rule 9 is a special case of Rule 10.

Trigonometric Functions
Rule 11. If y = sin u, where u = f(x),

dy du
= cos v —
dx dx

Rule 12. If y = cos u, where u = f (x),

dy . du
— — —sin u —
dx dx

Rule 13. If y — tan u, where u = /(x),

dy . du
— = sec u —
dx dx

Inverse Functions
Rule 14. If >’ = /(x) and x = g(y) are inverse functions,
dy 1
dx dx
dy

Composite Functions
Rule 15. If >’ = <h(w) and u = F(x),

dy dy du
dx du dx

Note that Rules 7, 8, 9, 10, 11, 12, and 13 are special cases of Rule 15.
 2.7 Differentials / 203

2.7 Differentials
I

In the preceding discussion of derivatives, is considered not as a fraction with

dx
Ay
numerator dy and denominator dx, but as a symbol denoting the limit of — as Ax
Ax
approaches zero as a limit. In some problems it is useful to interpret dy and dx
separately. In this context, dy is referred to as the differential of y and dx is referred to
as the differential of x. Differentials are useful, for example, in applications of integral
calculus and in approximating changes in the dependent variable associated with
small changes in the independent variable. Approximation of the value of a function
using differentials is discussed in this section.
If f'(x) is the derivative of y = /(x) for a particular value of x and Ax is an
arbitrarily chosen increment of x, then the differential of y, denoted by df(x) or dy, is

dy — df(x) = f(x) Ax = -J- Ax

and the differential of x, denoted by dx, is

dx — Ax

Thus the differential of a function is defined as its derivative multiplied by an

increment of the independent variable. Note that if the function/(x) is taken to be x,
then/'(x) = 1 and it follows that the differential dx of x is Ax.
Geometrically, consider the curve y = /(x) (see Figure 2.20) and let /'(x) be the
value of the derivative at P. Then Ax — PQ and

dy =f(x) = (tan 6)(PQ) = ^PQ = QT

Therefore, dy is the increment (equal to QT) of the ordinate of the tangent corre¬
sponding to dx.

Figure 2.20

y
204 / Differential Calculus: Functions of One Variable

This geometrical argument provides the following interpretation of the derivative

as a fraction. If an arbitrarily chosen increment of the independent variable x for a
point P(x, y) on the curve y —f{x) is denoted by dx, then in the derivative

^ = /'(x) = tan 6
dx

dy denotes the corresponding increment of the ordinate of the tangent at P.

Note that the differential dy and the increment Ay of the function corresponding
to the same value of dx = Ax are not in general equal. In Figure 2.20, dy = QT but
Ay - QF.

Approximations
It can be seen in Figure 2.20 that Ay = QP' and dy = QT are approximately equal
when Ax = PQ is small. In fact, the differential of a function can be made arbitrarily
close to the increment of the function for sufficiently small changes in the indepen¬
dent variable. The value of the corresponding differential thus can be used as an
approximation for the change in a function when the independent variable changes
by a small amount.

EXAMPLE

If y — x4 — 2x3 + 9x + 7, use differentials to find an approximate value of y when x = 1.997.

Consider 1.997 to be the result of applying an increment Ax = —0.003 to an original value
of x = 2. Then

dy = (4x3 — 6x2 + 9) Ax

= (32 — 24 + 9)(—0.003)

= -0.051

Since y = 25 when x = 2 and —0.051 is the approximate change in y when x changes from 2 to
1.997,

y + dy = 25 - 0.051

= 24.949
and y is approximately 24.949 when x = 1.997.
To obtain an estimate of the accuracy of this approximation, compute the error of the
approximation, which is given by Ay — dy.

Ay =/(x + Ax) -/(x)

= (x + Ax)4 — 2(x -I- Ax)3 + 9(x + Ax) + 7 — x4 + 2x3 — 9x — 7

= x4 — 2x3 -f 9x 4- 7 + 4x3(Ax) + 6x2(Ax)2

+ 4x(Ax)3 + (Ax)4 — 6x2(Ax) — 6x(Ax)2

- 2(Ax)3 + 9(Ax) - x4 + 2x3 - 9x - 7

= 4x3(Ax) + 6x2(Ax)2 + 4x(Ax)3 -I- (Ax)4 — 6x2(Ax)

— 6x(Ax)2 — 2(Ax)3 + 9(Ax)

and

dy = (4x3 — 6x2 + 9)(Ax)

= 4x3(Ax) — 6x2(Ax) + 9(Ax)

 2.7 Differentials / 205

Thus

dy — Ay = 6x2(Ax)2 + 4x(Ax)3 + (Ax)4 — 6x(Ax)2 — 2(Ax)3

And, for x = 2,

dy- Ay = (24 - 12)( —0.003)2 + (8 - 2)( —0.003)3 + (-0.003)4

= 0.000107838081

Thus y + Ay = y + dy + (Ay — dy) = 24.949107838081 and the approximation y = 24.949 is

accurate to the three decimal places computed.

EXAMPLE

Using differentials, find an approximate value of ^98.

The number 98 is close to the number 100, which is a perfect square. Thus the change in
y = y/x corresponding to a change in x from 100 to 98 can be added to y = v/rKX) = 10 to
obtain an approximate value for y/9S. When x changes to x + dx, y changes to y + dy,
approximately.
For x = 100 and Ax = —2,

-2
” 27100
= -0.1

and thus ^798 = >/l00 — 0.1 = 9.9 approximately.

EXAMPLE

Using differentials, find an approximate value for tan 46° given tan 45° = 1, sec 45° = yfl,
and 1° = 0.0175 radians.
If y = tan x, dy = sec2 x dx. When x changes to dx, y changes to y + dy, approximately.

71
x = 45° = - radians
4

Ax = 1° = 0.0175 radians

dy = sec2 x dx

= (V2)2(0.0175)

= 0.0350

Since y = tan 45° = 1, y + dy = 1.0350, and tan 46° = 1.0350 approximately.

Differentials can also be used to estimate small errors in calculation of a function

which arise from lack of precision in the measurement of the independent variable or
from other causes.
206 / Differential Calculus: Functions of One Variable

EXAMPLE

Suppose C = 5 + 0.6x -I- 0.2v/x, where C is total consumption (billions of dollars) and x is
total disposable income (billions of dollars). If x = 25 with a maximum error of 0.3, find the
approximate maximum error in consumption.

dC — 0.6 + —^= dx

,
x

0
= 0.6 +
725
y<°.3)
= 10.6+ 531 (0.3)

= (0.62)(0.3)

= 0.186

If du is the error in u, then

du .
is the relative error in u
u
and
du .
100 is the percentage error in u
u

For this example, the approximate maximum relative error is

dC 0.186
_ = _ = 0.00744

and the percentage error is 0.744.

It is frequently convenient to find the relative error by logarithmic differentiation,

since if y = In x, dy = dx/x.

EXAMPLE

For the demand function

16

where x is number of units demanded and y is price in dollars, assume y is 200 with a
maximum error of 10 and determine the approximate maximum relative error in x.

In x = In 16 + 4 In y

dx A dy
— = -4
x y

and, for y = 200 and dy = 10, the relative error in x is

dx 40 1
~x ~ 200 ~~ ~ 5

and the percentage error is —20. The negative sign of the error indicates that the errors in
estimating x and y are opposite in sign.
 2.8 Higher-Order Derivatives / 207

PROBLEMS
Use differentials to find the approximate value of each of the following.

1. 0010 3. 766

2. 05 4. 04

Find dy and Ay for each of the following.

5. y = x4 — x2/2 for x = 2, Ax = 0.1 7. y = (x_+ l)3 for x = —3, Ax = —0.003

6. y = 12.8/y for x = 10, Ax = 0.24 8. y — yjx for x = 4, Ax = 0.04

9. A container is made in the form of a 10-cm cube to hold 1 liter (1000 cm3). How accurately
must the inner edge be made so that the volume will be correct to within 3 cm3?
10. Using differentials, find the allowable percentage error in the diameter of a circle if the area
is to be correct to within 4 percent.
11. Show that the relative error in the nth power of a measurement is approximately n times
the relative error in the measurement.
12. Show that the relative error in the nth root of a measurement is approximately 1/n times
the relative error in the measurement.

Answers to Odd-Numbered Problems

1. 10.033 7. dy = -0.036; Ay = -0.036027
3. 8.125 9. error must not exceed 0.01 cm
5. dy = 3; Ay = 3.2431

2.8 Higher-Order Derivatives

In some problems it is useful to differentiate a function more than once. The result of
two or more successive differentiations of a function is a higher-order derivative.
The derivative of y =f(x) with respect to x is, in general, a function of x and may
be differentiated with respect to x. The derivative of the first derivative is the second
derivative; its derivative is the third derivative, and so on.
In general, the nth derivative of a function y =f(x) is obtained by differentiating n
times. The result of this differentiation—that is, the nth derivative—is denoted by
dny
dxn

EXAMPLE

If y = x3 — 3x2 + 2,
dy
— 3x2 — 6x
dx
d2y
= 6x — 6
dx2
d3y
= 6
dx3
d4y
= 0
dx4
and higher-order derivatives are also zero.
208 / Differential Calculus: Functions of One Variable

EXAMPLE

If x = Jt2 + 1,

^= l)-'/2(2f)

= r(r2 + 1)_1/2

g = (t2+l)-‘'2 + (-i)t(t2 + l)-3'2(20

= (f2 + l)"3/2[r2 + 1 -t2]

= (t2 + 1 )_3/2

g=-^2 + l)-(2()
= — 3f(t2 + 1)"5/2

g= -3 (t> +1)~5'2 -H-m2 + ir7,2(2<)

= (r2 + 1)~ 7/2( — 3f2 - 3 + 15r2)

= 3(4r2 - l)(f2 + 1)“7/2

and so forth for higher-order derivatives.

EXAMPLE

If y = sin x,
dy
~ = COS X
dx
d2y
= —sin x
dx
d3y
= —cos x
dx
dAy
d? = smx
and so forth for higher-order derivatives.

Just as the derivative of the function y =f{x) with respect to x represents the rate
of change in y as x changes, the second derivative of y =f (x) with respect to x
represents the rate of change in the first derivative y' = /'(x) as x changes.
In general, the nth derivative with respect to x of a function y — f(x) represents the
rate of change in the (n — l)st derivative of y = f(x) as x changes.
Higher-order derivatives are very important for certain theoretical problems in
mathematics and statistics; however, in most applied work, derivatives of order
higher than the second are not frequently encountered.
Second-order derivatives are useful in graphical representation of functions, as
discussed in later sections, and are also useful whenever the rate of change of the
quantity represented by the first derivative is of interest. For example, the rate of
change of velocity is acceleration and the rate of change of marginal cost indicates
 2.8 Higher-Order Derivatives / 209

whether marginal cost is increasing, decreasing, or constant under varying conditions

of production.

EXAMPLE

If the position of a moving body at time t is represented by

s = (2f + 3)2

find its velocity, v = ~, and its acceleration, a = = ^-4.

dt dt dt2

v = j( = 2(2t + 3X2)

= 8t + 12

a = , =
dv e8
dt

EXAMPLE

An object projected vertically upward with a speed of 160 feet/second reaches an elevation of

s = 160f — 16f2

at the end of t seconds.

(a) How high does it rise?

(b) How fast is it traveling when it reaches an elevation of 256 feet going up and again when it
reaches this elevation coming down?
(c) What is its acceleration?
ds
(a) v = jt = 160 — 32f and v = 0 when t = 5. Thus its maximum height is

s = 160(5) - 16(52)

= 800 - 400

= 400 ft

(b) If s = 256,

256 = 160f - 1612

t2 - lOf + 16 = 0

(t - 8)(f - 2) = 0

t = 8, 2

The object is thus at an elevation of 256 feet going up after 2 seconds and coming down after 8
seconds.

ds
If t = 2, v = — = 160 - 32t = 160 - 64 = 96 ft/sec
dt

If t = 8, v = ^ = 160 - 32r = 160 - 256 = -96 ft/sec

dt

(c) a = ~ = - 32 ft/sec2
210 / Differential Calculus: Functions of One Variable

EXAMPLE

Total cost is represented by

y = 50 + 60x — 12x2 + x3

(see Figure 2.21).

Figure 2.21

where y is total cost and x is output. Average cost is y/x = (50/x) -f 60 — 12x + x2 and

marginal cost is —1- = 60 — 24x + 3x2 (see Figure 2.22).

ax

Figure 2.22
 2.8 Higher-Order Derivatives / 211

— 24 + 6x

— 0 when x = 4

< 0 when x < 4

> 0 when x > 4

Thus marginal cost decreases for x < 4, increases for x > 4, and is constant for x = 4.

EXAMPLE

Total cost is represented by

y = 200 + lOOOx - 24x2 + 4x3 4- x4

(see Figure 2.23).

where y is total cost and x is output. Average cost is

^ = 45+ 1000 _ 24x + 4x2 + x3

X X

and marginal cost is

= 1000 - 48x + 12x2 + 4x3

ax

(see Figure 2.24).

d-4 = -48 + 24x + 12x2

ax

= 0 when x2 + 2x — 4 = 0

Figure 2.23

y
212 / Differential Calculus: Functions of One Variable

Figure 2.24

dy_ y
dx'y

Appropriate value is x = -1-1- y/5 (approximately x = 1.236).

d2y n , .at
- -2 = 0 when x = — 1+^/5
tf A

<0 when x < -i + y5

> 0 when x > — 1 + y/S

Thus marginal cost decreases for x < — 1 + yj 5, increases for x > — 1 + yj~5, and is constant
for x = — 1 + y/5.

PROBLEMS
Find the first and second derivatives of each of the following functions.

1. y = e3x 4. y = x2ex

2. y = x In x 5. y = e,n (xJ~3)

3' fx - 5'
3. y = log 6. y =
.X, x +1

Answers to Odd-Numbered Problems

 2.9 Implicit Differentiation / 213

2.9 Implicit Differentiation

Functions of the form y =f(x) express y explicitly in terms of x and can be differen¬
tiated according to the rules for the types of functions involved. However, as noted in
preceding sections, some equations involving x and y, of the form/(x, y) = 0, do not
give y explicitly in terms of x and cannot conveniently be written so that they do.
Such equations define y as a function of x in the sense that for each value of x there is
a corresponding value of y which satisfies the equation; the equation is said to
dy
determine y as an implicit function of x. It is possible to calculate — from such an
dx
equation by the method of implicit differentiation. In this method, y is treated as an
unknown but differentiable function of x and the rules for finding derivatives are
dy
applied. In general, this method gives an expression for — in terms of both x and y.

Higher-order derivatives can also be obtained by implicit differentiation. The

procedure for implicit differentiation is discussed and illustrated below.

Procedure for Implicit Differentiation

When y is defined as an implicit function of x by an equation /(x, y) = 0, the
derivative of y with respect to x is obtained by differentiating the equation f(x, y) = 0
term by term, regarding y as a function of x, and then solving the resulting equation
_ , , . . dy
for the derivative —.
dx

EXAMPLE

dy
Find — for the equation xy2 — x + y = 0. Differentiating with respect to x,
dx

2xydy
dx
+ y2-2x + dx
dy = 0

(2xy + 1)^ = 2x - y2

dy 2x — y2
dx 2xy -I- 1

note: In this example, xy2 is the product of two functions of x, x and y2, and is differen¬
tiated using the product rule

d . . dv du
Tx(uv) = uTx + vTx

, du dx t dv _ dy d , ^ dy , _ , .
where u = x, v = y , — = — =1, and — = 2y —ihus — (xy ) = 2xy — -f v . The deriva-
dx dx dx dx dx dx

tive of —x2 is — 2x, as in differentiation of explicit functions (note again that — = 11. The
dx

derivative of y is —, again as in differentiation of explicit functions.

dx
214 / Differential Calculus: Functions of One Variable

EXAMPLE

Find -V for the equation x3 + y3 — 3axy = 0. Differentiating with respect to x,

dx

2 dy 0 dy
3x2 + 3y2 —-3ax -3ay = 0
dx dx

/(f-ax)Tx
2 , dy = ay-x 2

dy ay — x2
dx y2 — ax

EXAMPLE

dy
Find for the equation sin x cos y tan y = 0. Differentiating with respect to x,
dx
dy 7 dy
cos x cos y-sinxsiny--sec v — = 0
dx dx

dy cos x cos y
dx sin x sin y + sec2 y

EXAMPLE

dy
Find — for the equation x — y = 1. Differentiating with respect to x,
dx

2x-2y‘^ = 0
dx

dy x
dx y

Differentiating ^ with respect to x,

dx

dy
y X
dx

dy
Substituting
dx
 2.9 Implicit Differentiation / 215

or, since y2 — x2 = — 1,

d2y 1
dx2 v3

EXAMPLE

dy
Find and for the equation ey xex = 0. Differentiating with respect to x,
dx

ey --ex — xex — 0
dx
dy ex(x + 1)
dx ey
or, since ey = xex,
dy x + 1
dx x
and
d2y x — (x + 1)
dx2 x2
1

EXAMPLE

x 1/2 + y112 = a112. Differentiating with respect to x,

+iy~ll2^ = 0

dy
dx x 1/2
dy . .
Differentiating — with respect to x,
dx

1/2 f?F\xi/2
d2y dxj
dx2 x

. dy y11
Substituting — = — -ji
dx x 1

d2y i + jx~ 1/2y1/2

dx2 x

^x-1 +ix~3l2y112

= 7X~3/2(x1/2 + yl/1)

or, since x1/2 + y112 — a112,

216 / Differential Calculus: Functions of One Variable

PROBLEMS
d2y
Find a. dL and b. for each of the following functions by the method of implicit
dx dx2
differentiation.
1. x2 + y2 = 1 10. x2y2 + xy = 1
2. x3 + y3 = 1 11. x3y3 + x2y2 = a
3. x2/3 + y2/3 = j 12. x + y + xy + y2 =
4. xy + y2 = 1 13. x3y2 = a
5. y2 = x3 14. x2 + xy = a2
6. xy — a 15. xy2 + y2 = a
7. x2y2 = b 16. xy + y3 = b
8. x2y3 = c 17. (x + y)2 + (x + yf
9. x + y — xy = 2 18. x2 + y2 = ab

dy
Find — for each of the following functions by the method of implicit differentiation.
dx

x — 1 1 1
19. y2 27. + -= 1
X + 1 y X

20. x3 -xyry3 = 1 28. y --= (x2 + 3)1/3x"1

_ x - y 29. cos2 x 4- tan y sin y = 0

21. x2
x + y
30. esc x — sec y + tan y + cot x = 0
22. yz= x(x2 -1- 1)_ 1/2
31. xyex + ey = 0
23. yz= x1/2 + x1/3 4- x1/4
32. ye■* = 10 4- yey
x2 — 1
24. y2 33. x5 + 4xy3 — 3y5 = 2
X + 1
34. x2 + xy 4- y2 = 1
25. (x + y)3 + (x -■ y)3 = x4 + y4

26. yz= (x + 5)4(x2 -2)3

Answers to Odd-Numbered Problems

1 * u 1
1
1. a. — b. —^ 19.
y y y(x + i)2

. a. -x-'/y/3 b. y~4/y1/3
21. y
3
(x 4- yf
_ 3x2 3x
5. a. b. —
2y 4y 23. jx-1/2 +ix~2/3 + ix~3/4

2x3 — 3x2 — 3y2

7. a. — - b. % 25.
X X 6xy — 2y:
2
9. a. '—1 b. 2*r 11 27.
1 — X (l-x); x2

2 sin x cos x
11. a. — - b. 2); 29.
X X sin y(l -F sec2 y)

3y 15y y(l+_x)
13. a. 31. -
2x 4x2 x(l - y)

-v 3y 5x4 + 4y3
15. a. b. 33.
2(x 4- 1) ■ 4(x + l)2 15/ - 12xy2
17. a. -1 b. 0
 2.10 Differentiability and Continuity / 217

2.10 Differentiability and Continuity

The relationships between differentiability and continuity are discussed in this sec¬
tion and smooth functions are defined and illustrated.
It can be shown that if the function y =f(x) has a finite derivative

f(c + Ax) -/(c)

f\c) lim
Ax-^0 Ax

at x = c, then /(x) is continuous at x = c.

Continuity at a point does not, however, imply the existence of a derivative at that
point. Thus

Differentiability ^ Continuity

EXAMPLE

/(x) = |x | is continuous at x = 0, but does not have a derivative at x = 0, since

/'(x) =1 for x > 0

= —1 for x < 0

(see Figure 2.25).

Figure 2.25

fix)

EXAMPLE

/(x) = x1/3 is continuous at x = 0, but does not have a derivative at x = 0, since/'(x) = \x 2/3
is not defined for x = 0 (see Figure 2.26).
218 / Differential Calculus: Functions of One Variable

Figure 2.26

fix)

Smooth Functions
A function f(x) is said to be a smooth function if both f(x) and its first derivative are
continuous. The graph of a smooth function is said to be a smooth curve; such a
curve is not only continuous and unbroken but also has a continuously turning
tangent. As shown in the following sections, a smooth function has some useful
properties not possessed by a continuous curve whose first derivative is discontinuous
at certain points.

EXAMPLE

dy
y = 1 — x2 is a smooth function, since y and its derivative — = — 2x are everywhere contin-
dx
uous (see Figure 2.27).

Figure 2.27

y
 2.11 Applications of Derivatives / 219

EXAMPLE

y = (x — l)2 3 + 1 is not a smooth function at x = 1, since its derivative

dy 2
dx 3^x- 1
is discontinuous (becomes infinite) at x = 1 (see Figure 2.28).

Figure 2.28

2.11 Applications of Derivatives

In this section the information concerning a function or curve which can be obtained
from its first and second derivatives is discussed. Such information is useful not only
for graphing but also for determining the appropriateness of equations for represent¬
ing various relationships in business, economics, and the behavioral sciences.

THE FIRST DERIVATIVE

The first derivative of a function can be used to determine where the function is
increasing and where it is decreasing. The first derivative can also be used to locate
maximum and minimum values of the function (stationary points), if there are any.

Increasing and Decreasing Functions

Consider the function y =f(x) at the point x = a. The first derivative of y with
respect to x, y' = f'{x) is the slope at the point x of the curve representing the
function y =f(x); in particular,/'(a) is the slope of the curve y =f{x) at the point
x — a.
If the first derivative f'{a) is positive, y =f(x) is an increasing function of x at
x = a; that is, y =f{x) increases as x increases past x = a; if the first derivativef'(a)
is negative, y =/(x) is a decreasing function of x at x = a; that is, y =/(x) decreases
as x increases past x = a. See Figure 2.29.
220 / Differential Calculus: Functions of One Variable

Figure 2.29

fix)

EXAMPLE

f(x) = 2x2 4- 10, f'(x) = 4x. Since f'(x) > 0 for x > 0 and f'(x) > 0 for x < 0,/(x) is an in¬
creasing function of x for positive x and a decreasing function of x for negative x. Note that
/'(x) = 0 for x = 0.

EXAMPLE

/(x) = 3x3, f'(x) = 9x2. Since f'(x) is nonnegative for all values of x, /(x) is an increasing
function of x for all values of x.

EXAMPLE

/(x) = x3 4- 6x2 + 15, /'(x) = 3x2 4- 12x = 3x(x 4- 4). Since /'(x) > 0 for x > 0 or x < —4
and f'(x) < 0 for — 4 < x < 0,/(x) is a decreasing function of x for the interval — 4 < x < 0
and an increasing function of x for all x < —4 and x > 0. Note that f'(x) = 0 for x = 0.

In the examples above, the values of x for which f'{x) = 0 are not discussed; such
values are of particular interest, as discussed below.

Relative (Local) Maxima and Minima

A function y = /(x) is said to have a relative maximum or local maximum at x = a if
/(a) is greater than any value of/(x) for x in an interval around a; a function y =f(x)
is said to have a relative minimum or local minimum at x = a iff(a) is smaller than any
value of/(x) for x in an interval around a. Note that a relative maximum or mini¬
mum of a function is its maximum or minimum for a given interval; the (absolute)
maximum or minimum of the function over a larger interval can occur at an end
point of the interval, rather than at any relative maximum or minimum. Also note
 2.11 Applications of Derivatives / 221

Figure 2.30

f(x)

that it is possible for a relative maximum value of a function to be less than a relative
minimum value of the function (see Figure 2.30).
If a function/(x) has a relative maximum or minimum at a value x = a for which
its first derivative is continuous, then f'{a) = 0. Values of x for which/'(x) is discon¬
tinuous must be considered separately, as discussed below.

CAUTIONS
1. A relative maximum or minimum at x = a impliesf\a) = 0 only if/(x) and/'(x)
are continuous at x = a.
2. f'(a) = 0 does not imply a relative maximum or minimum at x = a even if/(x) and
/'(x) are continuous at x = a; that is, if/(x) and/'(x) are continuous at x = a,
f'(a) = 0 is a necessary but not a sufficient condition for a relative maximum or
minimum at x = a.

EXAMPLE (CAUTION 1)

Figure 2.31

fix)
222 / Differential Calculus: Functions of One Variable

Iff(x) = (x — 1)2/3 + 1, then

/ W = Y(x - i)1/3

and f'(x) is (infinitely) discontinuous at x = 1. Thus, although the function has a relative
minimum at x = !,/'(!) =/= 0 (see Figure 2.31).

EXAMPLE (CAUTION 2)

Iff(x) = x3, then

/'(*) = 3x2

and /'(x) = 0 for x = 0. However, the function/(x) = x3 does not have a relative maximum or
minimum at x = 0 (see Figure 2.32).

Figure 2.32

fix)

Consider a function f(x) at a value x — a for which f(x) and /'(x) are continuous.
It is evident geometrically that iff(a) is a relative maximum of/(x), then the slope
J'(x) off(x) changes from positive to negative as x passes through the point x = a;
similarly, if f(a) is a relative minimum of/(x), then the slope /'(x) of/(x) changes
from negative to positive as x passes through the point x = a (see Figure 2.33). The
corresponding algebraic argument is based on the fact that an increasing function
has a positive slope and a decreasing function has a negative slope.
Thus, in order to determine the relative maximum and minimum values, if any, of
a function y =/(x),
1. Solve the equation /'(x) = 0 to obtain its roots (sometimes called critical values).
2. For each root a, determine whetherf'(x) changes sign as x increases through a:
 2.11 Applications of Derivatives / 223

Figure 2.33

f(x)

f'(x) changes from + to — at x = a => relative maximum at x = a

f'(x) changes from — to -I- at x = a => relative minimum at x = a
f'(x) does not change sign at x = a => no relative maximum or
minimum at x = a

note: The above procedure locates all relative maxima and minima that occur
at values of x for which f(x) and f'(x) are continuous. Relative maxima and minima
which occur at values of x for which/'(x) is discontinuous are discussed below.

EXAMPLE

Find the relative maxima and minima (if any) of the function y = 2x3 — 3x2 — I2x +13.

dy
= 6x2 — 6x — 12
dx

= 0 if x2 — x — 2 = 0

(x — 2)(x + 1) = 0

x = 2, - 1

If -1 < x < 2, ^<0

dx
so minimum at x = 2
lfx>2, ^ > 0
dx

If x < -1, d-r > 0

dx
so maximum at x = — 1
If - 1 < x < 2, d-l < 0
dx

See Figure 2.34.

224 / Differential Calculus: Functions of One Variable

Figure 2.34

EXAMPLE

Find the relative maxima and minima (if any) of the function y = 3x4 — 4x3.

~ = 12x3 - 12x2
ax

= 0 if x3 - x2 = 0

x2(x — 1) = 0

x = 0, 1 x

Figure 2.35

y
 2.11 Applications of Derivatives / 225

dy
If X <0, <0
dx
so no maximum or minimum at x = 0
dl
IfO < X < 1, <0
dx

dy
IfO < X < 1, <0
dx
so minimum at x = 1
dy
If X > 1, >0
dx

See Figure 2.35.

EXAMPLE

Find the relative maxima and minima (if any) of the function y = x/^/x2 + 1.

dy (x2 + l)1 2 — ^x(x2 + l)~1/2(2x)

dx (x2 + 1)

(x2 + 1)~ 1/2(x2 + 1 - x2)

(x2 + 1)

1
(x2 + 1)3/2

T4 0

for all values of x, so there is no maximum or minimum (see Figure 2.36).

Figure 2.36

In each of the examples above, the first derivative is continuous for all values of x.
Thus relative maxima and minima can occur only for values x = a for which
f'(a) = 0. As noted above, relative maxima or minima may also occur for values
x = a for which f'{a) is discontinuous. That is, if f(x) is continuous at x = a but its
226 / Differential Calculus: Functions of One Variable

first derivative is discontinuous (finitely or infinitely) at x = a, then f(a) may be a

relative maximum or minimum even though fr(a) ± 0. Again, a change in the sign of
the first derivative as x passes through a is necessary for the existence of a relative
maximum or minimum at x = a.
Thus, in order to determine relative maximum and minimum values (if any) of a
function y = /(x) at points where /(x) is continuous but f'(x) is discontinuous,
1. Determine the values of x for which f'(x) is finitely or infinitely discontinuous and
/(x) is continuous.
2. For each value a such that/(x) is continuous at x = a and f'(x) is discontinuous at
x = a, determine whetherf\x) changes sign as x increases through a:

f'(x) changes from 4- to — at x = a => relative maximum at x = a

f'(x) changes from - to + at x = a => relative minimum at x = a

f'(x) does not change sign at x = a => no relative maximum or

minimum at x = a

EXAMPLE

_ v-2/3
Find the relative maxima and minima (if any) of the function y = x

dy _ 2 -1/3
dx~lX

± 0 for all values of x

dy A_ , dy +
-f- -> —oo asx-»0 and -—» oo as x -► 0
dx dx

dy
so — has an infinite discontinuity at x = 0.
dx

If x < 0, % < 0
dx
so minimum at x = 0
If X > 0, ~ > 0
dx

See Figure 2.37.

Figure 2.37

y
 2.11 Applications of Derivatives / 227

EXAMPLE

Find the relative maxima and minima (if any) of the function y = (1 — x)2/3(2 + x)1/3.

^ = -|(1 - x)“1/3(2 + x)1/3 + (1 - x)2/3(i)(2 + x)~2/3

= i(l - x)_1/3(2 + x)_2/3[-2(2 + x) + (1 - x)]

-3 - 3x
3(1 - x)1/3(2 + x)2/3

_ ~(x + 1)_
(1 - x)1/3(2 + x)2/3
= 0 if x = — 1
dy dy
— oo as x -> 1 and "f - oo as x -► 1+
dx dx
dy
so — has an infinite discontinuity at x = 1.
dx

^ —oo as x -> —2 and oo as x —2

dx dx
dy
so — has an infinite discontinuity at x = —2.
dx

If -2 <x < -1, % > 0

dx
so maximum at x = — 1
If -1 < x < 1, df < 0
dx

If — 1 < x < 1, dl<0

dx
so minimum at x = 1
dy
If x > 1, -- > 0
dx

Ifx< -2, ^>0

dx
so no maximum or minimum at x = —2
If -2 <x < -1, ^>0
dx
See Figure 2.38.

Figure 2.38
y
228 / Differential Calculus: Functions of One Variable

EXAMPLE

Find the relative maxima and minima (if any) of the function y = — (4 — x)1/2 — 2.

1
“ 2(4 - x)1/2

dy dy dy
-► oo if x -► 4 , so — has an infinite discontinuity at x = 4. Note that y and —- are
dx dx dx
undefined if x > 4. Thus y = —(4 — x)1/2 — 2 has no relative maximum or minimum (see
Figure 2.39).

Figure 2.39

The least (or greatest) value of a function in an interval—that is, its absolute
minimum or maximum in an interval—may occur at an end point of the interval
rather than at a relative minimum or maximum value.

EXAMPLE

Find the maximum and minimum values of the function y = x2 — 2x in the interval 2 < x < 4.

dl — 2x — 2
dx
= 0 if x = 1
<0 if x < 1 . .
.. J so minimum at x = 1
>0 if x > lj
But x = 1 is outside the interval 2 < x < 4.
If x = 2, y = 0

If x = 4, y = 8
Thus in the interval 2 < x < 4, the least value of y occurs at the end point x = 2, and the
dy
greatest value occurs at the end point x = 4; at neither of these points is — equal to zero (see
dx
Figure 2.40).
 2.11 Applications of Derivatives / 229

Figure 2.40

EXAMPLE

The total cost curve of a commodity is y = 15x — 8x2 + 2x3, where y represents total cost and x
represents quantity. Suppose that market conditions indicate that between 3 and 10 units
should be produced (that is, 3 < x < 10). Determine the quantity in this interval for which
average cost is a minimum.

Figure 2.41
y
230 / Differential Calculus: Functions of One Variable

Average cost = y/x = y = 15 — 8x + 2x2.

-f - = — 8 + 4x
ax

= 0 if x = 2

<0 if x < 2)
C so minimum at x = 2
>0 if x > 2)

But x = 2 is outside the interval 3 < x < 10.

Ifx = 3,5? = 9

If x = 10, y = 135

Thus in the interval 3 < x < 10, the least value of y occurs at x = 3 and the greatest value
dy
occurs at x = 10; at neither of these points is — equal to zero. Thus, for between 3 and 10
dx
units, average cost is minimum for 3 units (see Figure 2.41).

SUMMARY OF THE INFORMATION THAT CAN

BE OBTAINED FROM THE FIRST DERIVATIVE
Consider the function y =f{x) at the point x = a

m > 0 =>/(x) is an increasing function of x at x = a

f'(a) < 0 =>f(x) is a decreasing function of x at x = a

f'(x) changes from + to — at

/(x) is continuous at x = a
x = a=> relative maximum at x = a
and either
f{x) changes from — to + at
If
f(a) = 0
then x = a => relative minimum at x = a
or
f'(x) does not change sign at
f'(x) is discontinuous at
x = a => no relative maximum
x= a
or minimum at x = a

Note that the least or greatest value of a function in an interval, that is, its absolute
minimum or maximum in an interval, may occur at an end point of the interval
rather than at a relative minimum or maximum value.

PROBLEMS
For each of the following functions, determine maxima and minima; sketch the curve re¬
presenting each function.

1. y = 12 — 12x + x2 2x
6. y =
x2 - 1
2- y =
X + 1
7. y = x2 — 4x 4- 3
X3 X2
1 y = T-y-6x 8. y = Xy/l — x2

4. y = x4 — 32x + 48 9. y = x3 — 3x2 + 2

5. y = 10. y =
x2 + 7 x2 + 4
 2.11 Applications of Derivatives / 231

8x
11. y 16. v =
:2 - 8 x2 + 4

12. y x4 - 4x3 + 12

3x2
13. y
\/x2 + 3 18. y = x5 + 6

1 19. y = (x — l)1/3(x + 1)2/3

14. y
16 — x2
20. y = ^(x3 — 6x2 + 9x + 6)
15. y |x3 — 4x2 + 6x + 2

Answers to Odd-Numbered Problems

1. max (— 2, 28) min (+ 4, 4^2)
min (2, -4) 13. no max
3. max ( —2, min (0, 0)
min (3, 15. max (1, -^)
5. no max min (3, 2)
no min 17. no max
7. no max min (-6, -\)
min (2, — 1) 19. max (- 1, 0)
9. max (0, 2)
min (2, —2)
11. no max

THE SECOND DERIVATIVE

The second derivative of a function can be used to determine where the function is
concave upward and where it is concave downward. The second derivative can also
be used to locate the points of inflection of a function, if there are any.

Concavity
Consider the function y = /(x) at the point x — a. The second derivative of y with
respect to x, y" = /"(x), is the slope at the point x of the curve representing the first
derivative y' = /'(x) of the function/(x); in particular,/"(a) is the slope of the curve
y' = f'(x) at the point x = a.
If the second derivative /"(a) of a function y =f(x) is positive, y' = f'(x) is an
increasing function of x at x = a; the curve representing y =f(x) is said to be concave
upward where/"(x) is positive. If the second derivative f"(a) of a function y =f(x) is
negative, y' = f'(x) is a decreasing function of x at x = a; the curve representing
y =f{x) is said to be concave downward where f"(x) is negative (see Figure 2.42).
Consider a function/(x) at a value x = a for which f(x) and f'(x) are continuous.
It is evident geometrically that iff'(a) = 0 and f(x) is concave downward at x = a,
then/(x) has a maximum at a; similarly, iff'(a) = 0 and/(x) is concave upward at
x = a, then/(x) has a minimum at a. The corresponding algebraic argument is based
on the fact that a positive second derivative indicates an increasing slope and a
negative second derivative indicates a decreasing slope.

caution : The second derivative test is only sufficient, not necessary, for a rela¬
tive maximum or minimum, since f(x) may be concave upward or downward at
x = a iff"(a) = 0. That is, iff'(a) = 0, then
232 / Differential Calculus: Functions of One Variable

Figure 2.42

f"(a) > 0 H concave upward

f"{a) < 0 S concave downward

From this follows a very convenient, although not always applicable, test for
relative maximum and minimum values:
f(x) and f'(x) are continuous > 0 => relative minimum
at x = a at x = a

If { then < f"(a) < 0 => relative maximum

arid
at x -- a

f'(a) = 0 f"(a) = 0 => test does not apply

EXAMPLE

Find the relative maxima and minima (if any) of the function y = 3X3 — lx1 + 3x + 1.

4x + 3

= 0 if x2 — 4x + 3 = 0

(x — 3)(x — 1) = 0

x = 3, 1
 2.11 Applications of Derivatives / 233

Figure 2.43

dy2
= 2x-4
dx“

d2y
If x = 1, -r-j < 0, so maximum at x = 1
dx

d2y
If x = 3, —2 > 0, so minimum at x = 3

See Figure 2.43.

EXAMPLE

Find the relative maxima and minima (if any) of the function y = x4.

= 0 if x = 0

Figure 2.44

y
234 / Differential Calculus: Functions of One Variable

d2y
= 12x:
dx2

d2y
If x = 0, —~ = 0, so there may or may not be a maximum or minimum at x = 0
dx

If x < 0, ^ < 0
dx
so minimum at x = 0

If x > 0, ~>0
dx

See Figure 2.44.

EXAMPLE

Find the relative maxima and minima (if any) of the function y = x3.

dy
— 3x"
dx

= 0 if x = 0

d2y
= 6x
dx2

d2y
If x = 0, = 0, so there may or may not be a maximum or minimum at x = 0
dxz

if x < o, 7 > o
dx
so no maximum or minimum at x = 0

If x > 0, ~>0
dx

See Figure 2.45.

Figure 2.45

y
 2.11 Applications of Derivatives / 235

In the examples above, when f"(a) — 0—that is, when the second derivative test
for a maximum or minimum does not apply—the first derivative is examined for
change of sign as x passes through a. There is, however, an alternative procedure,
involving derivatives of higher order, which is more convenient in some cases.
f\a) =f"(a) --y<"“1>(a) = 0 | f{n){a) < 0 => relative
maximum at x = a
for n even
f{n)(a) > 0 => relative
If i and then minimum at x = a

for n odd,/(a) is neither a

/<-» + 0 relative maximum or minimum

EXAMPLE

Find the relative maxima and minima (if any) of the function y = 3x4 — x3 + 2.

^ = 12x3 - 3x2
dx

= 0 if 12x3 — 3x2 = 0

3x2(4x — 1) = 0

x = 0, ^

d2y
= 36x2 — 6x
dx2

= 6x(6x — 1)

d2 y
If x = i, —> > 0, so minimum at x = \
dx

d2y
If x = 0, —r = 0, so there may or may not be a maximum or minimum at x = 0
dxz

d3y
t = 72x — 6
dx

= 6(12x - 1)

d3y . .
If x = 0, T =/= 0, so no maximum or minimum at x = 0
dx

Alternative procedure using change of sign of the first derivative:

dy
If X <0, <0
dx
so no maximum or minimum at x = 0
dy
IfO < X < h <0
dx

dy
IfO < X < i> <0
dx
so minimum at x = \
dy
If X > i. >0
dx

See Figure 2.46.

236 / Differential Calculus: Functions of One Variable

Figure 2.46
y

Points of Inflection
A function y —f(x) is said to have a point of inflection at a point where the concavity
of the curve changes. Geometrically, at a point of inflection the tangent crosses the
curve. Since the sign of the second derivative indicates the concavity of a curve, it
follows that a change in the sign of the second derivative implies a change in concav¬
ity (and thus a point of inflection).
If a function f(x) has a point of inflection at a value x = a for which its second
derivative is continuous, then f"{a) = 0. Values of x for which f"(x) is discontinuous
must be considered separately, as discussed below.

CAUTIONS
1. A point of inflection at x = a implies f"{a) = 0 only if /(x) and f"(x) are
continuous at x = a.
2. f”(a) = 0 does not imply a point of inflection at x = a even if/(x) and /"(x) are
continuous at x = a; that is, iff(x) andf"{x) are continuous at x = a,f"(a) = 0
is a necessary but not a sufficient condition for a point of inflection at x = a.

Consider a function/(x) at a value x = a for which/(x) and f"(x) are continuous.

If/"(x) changes sign as x passes through the point x = a, then the curve representing
f(x) has a point of inflection at x = a. In this case f"(a) = 0.
A point of inflection may also occur at a value x = a for which f(x) is continuous
but f"(x) is discontinuous. Again, change in the sign of the second derivative as x
passes through a is necessary for the existence of a point of inflection at x = a.
Thus, in order to determine the points of inflection (if any) of a function y =/(x),
1. Determine the values of x for which /"(x) is either zero or discontinuous and/(x)
is continuous.
2. For each value a such that/(x) is continuous at x = a and f"(x) is either zero or
discontinuous at x = a, determine whether f"(x) changes sign as x increases
through a:

/"(x) changes sign at x = a => point of inflection at x — a

f "{x) does not change sign at x = a => no point of inflection at x = a

 2.11 Applications of Derivatives / 237

note: If f(x) and /"(*) are continuous at x = a and f"(a) = 0, then /"'(a) 0
implies a point of inflection at x = a; this may be used as an alternative to checking
for change of sign of the second derivative at x = a.

EXAMPLE

Find the relative maxima and minima and the points of inflection (if any) of the function
y = x1/3.

dy _ i _ 2/3
= ix
dx

^ 0 for all x

dy
discontinuous at 0
dx

If X < 0, ~ > 0
dx
> so no maximum or minimum at x = 0

If x > 0, % > 0
dx

d2y
2 yX~5/3
-£y
dx

=/=■■ 0 for all x

d2y
discontinuous at 0
dx2

d2y
If x < 0, >0
dx2
> so point of inflection at x = 0
d2y
If x > 0, ~ < 0
dx

See Figure 2.47.

238 / Differential Calculus: Functions of One Variable

EXAMPLE

Find the relative maxima and minima and the points of inflection (if any) of the function
y = *A/l ~ *2-

dl = - jx(l - x2)-3,2( —2x) + (1 - x2)-1'2

= (1 — x2) 3/2(x2 -hi — x2)

= (1 — x2)_3/2

— oo as x -+ ± 1, so infinite discontinuities at x = ±1
dx

— > 0 for all x, — 1 < x < 1, so no maximum or minimum

dx
(Note that y = x/^/l — x2 is defined only for — 1 < x < 1.)

g=-!( i-x2rs/2(-2x)

= 3x(l - x2)-5/2

= 0 if x =0

d2 y
If — 1 < x < 0, 2 <0
dx
so point of inflection at x = 0
d2y
IfO<x< 1, t4>0
i
dx '

See Figure 2.48.

Figure 2.48

EXAMPLE

Find the relative maxima and minima and the points of inflection (if any) of the function
y = x -h (1/x).
 2.11 Applications of Derivatives / 239

dy = t
dx x2
x2 - i

= 0 if x = + 1

dy discontinuous at x = 0
dx
dy
If x < — 1, — > 0
dx
so maximum at x = — 1
If -1 < x < 0, ^ < 0
dx

If 0 < x < 1, ^<0

dx
so minimum at x = 1
Ifx> 1, $->0
dx

If -1 < x < 0, ^ < 0

dx
so no maximum or minimum at x = 0
If 0 < x < 1, ™<0
dx
d2y _ 2
dx2 ~3
0 for all x

... ...
—r discontinuous at x = 0 but — also discontinuous at x = 0, so no point of inflection
dxz dx
at x = 0
See Figure 2.49.

Figure 2.49

y
240 / Differential Calculus: Functions of One Variable

SUMMARY OF THE INFORMATION THAT CAN

BE OBTAINED FROM THE SECOND DERIVATIVE
Consider the function y =f(x) at the point x = a:

f "(a) > 0 => f(x) concave upward at x = a

f"(a) < 0 =>f{x) concave downward at x = a
f"(a) > 0 => relative minimum at
f(x) and /'(x) are continuous
x= a
at x = a
If then < f"{a) < 0 => relative maximum at
and
x= a
f'(a) = 0
f"(a) = 0=> test does not apply

f(x) is continuous at x = a f"(x) changes sign at x — a

and either => point of inflection at x = a
then l
If /"(a) = 0 f"(x) does not change sign at x = a
or => no point of inflection at x = a

/"(x) is discontinuous at x = a

SUMMARY OF THE PROCEDURE FOR SKETCHING

THE GRAPH OF y=f(x) USING INFORMATION PROVIDED
BY FIRST AND SECOND DERIVATIVES
dv d2v
1. Obtain ~ and ——~.
dx dx

dy
2. Determine the ranges of values of x for which — is positive and for which it is
dx
negative.
dyit y
3. Check points where — = 0 for possible maxima or minima.
dx
dy
4. Check points where — is discontinuous for possible maxima or minima.
dx
d2y .
5. Determine the ranges of values of x for which ^ 4 positive and for which it is
dx
negative.
d2y
6. Check points where , ? = 0 for possible points of inflection.
ax
d2y
7. Check points where —j is discontinuous for possible points of inflection.
dx
8. Determine the intercepts.
9. Note the nature of the curve for very small and very large values of x.

10. Sketch the curve as indicated by the signs of ~ (increasing or decreasing) and of
A-

d2y
(concave upward or downward), the intercepts, asymptotic properties, and
dx2
discontinuities
 2.11 Applications of Derivatives / 241

EXAMPLE_

Sketch the curve represented by the function y = 4 + 3x — x3.

dy
3 — 3x2
dx

if x = +1

If X < -1, < 0 (decreasing)

dx

If — 1 < x < 1, ~ > 0 (increasing)

dx

If x > 1, ^ < 0 (decreasing)

dx

so minimum at (— 1, 2) and maximum at (1, 6). (As a check, if x 6, and if x 1,

if x = 0

> 0 if x < 0 (concave upward)

< 0 if x > 0 (concave downward)

so point of inflection at (0, 4). If x = 0, y = 4. As x -> oo, y -> - oo, and as x -> - oo, y -> oo.
See Figure 2.50.

Figure 2.50

y
242 / Differential Calculus: Functions of One Variable

EXAMPLE

Sketch the curve represented by the function y = x + (4/x).

dy = i _ 4
dx x2

d2y 8x 3
dx2 x4 x3

dy
= 0 if x = ±2
dx

dy
If x < -2, > 0 (increasing)
dx

di
If — 2 < x < 2 and x ^ 0 <0 (decreasing)
dx

If x > 2, (increasing)

so minimum at (2, 4) and maximum at ( — 2, —4). y, —, and discontinuous at x = 0 (if

dx dx \
d2y d2y
x = 2, 9 = 1, and if x = — 2, „ = — 11.
dx' dx'

d2y
<0 if x < 0 (concave downward)
dx2

>0 if x > 0 (concave upward)

No point of inflection at x = 0 because of discontinuity of y = x + (4/x). No x-intercepts or

y-intercepts. As x -► oo, y -*■ co, and as x -► — oo, y -► — oo. See Figure 2.51.

Figure 2.51
 2.11 Applications of Derivatives / 243

EXAMPLE

Find the points at which the curve represented by the function x — y4 + 2y2 —2 = 0 has
vertical tangents and sketch the curve.

3 dy A dy
i - 4.y + 4y -j- = 0
dx dx

dy 1
dx 4 y(y2 — 1)

dy dy
so — ^ 0 for all x. Vertical tangents occur when — is undefined, so there are vertical tangents
dx dx
at (2, 0), (1, — 1), and (1, 1). (Note that x = y4 — 2y2 + 2.)

d2y ~^12y2~4^dx — 4(3y2 — 1)

dx2 [4y(y2 - l)]2 [4y(y2 - 1 )]3

d2y
= 0 if y = ±
dx'

d2y
Ify < -l, > 0 (concave upward)
dx2

if-i<,< -yi, g<o (concave downward)

If -71 < y < 0, 0 > 0 (concave upward)

Figure 2.52
244 / Differential Calculus: Functions of One Variable

so there is an inflection point at y —

d2v
If 0 < y < yj, ^2 < 0 (concave downward)

r d2 y
if y/i < y < l > 0 (concave upward)
WA

so there is an inflection point at y = ^/J.

d2 y
if y > i, -Hz <° (concave downward)
UA

If y = 0, x = 2. As y -* oo, x ->• oo, and as y -* — oo, x -*• oo. See Figure 2.52.

PROBLEMS
For each of the following functions, determine any maxima, minima, and points of inflection;
sketch the curve representing each function. (Refer to problems on pages 230-231 for maxima
and minima.)

1. y — 12 — 12x 4- x:
11. y =
y/x2 ~ 8
2- y =
X + 1 12. y = X4 - 4x3 + 12
X3 X2 3x2
3. y . —-— 6x 13. y =
3 2 Vx2 +"3
4. y = x4 — 32x + 48 1_
14. y =
16 — x2
5- y =
\/X2 + 7 15. y = |x3 — 4x2 -F 6x + 2

2x 8x
6- y = 16. y =
y/x2 ~ T x2 + 4

7. y = x2 - 4x + 3 x -I- 3
17. y = —j-
X
8- y = XV 1 - X2

18. y = x5 + 6
9. y = X3 - 3x2 + 2
19. y = (x — l)1/3(x + 1)2/3
1
10. y =
x2 + 4 20. y = i(x3 — 6x2 + 9x + 6)

21. Sketch a smooth curve y — f(x) having all the following properties:
a./(l) = 0 b./(l) = 0
f'(x) < 0 for x < 1 f'(x) < 0 for x < 1
f (x) > 0 for x > 1 f"(x) > 0 for x > 1
22. Sketch a smooth curve y =/(x) having all the following properties:
/(0) = 10 /'(6) = 0 f"(x) < 0 for x < 9
/(6) = 15 /'(10) = 0 /"(9) = 0
/(10) = 0 /"(x) > 0 for x > 9
23. Sketch a continuous curve y =f{x) having all of the following properties:
/(— 2) = 8 /'(x) > 0 for | x | >2 /"(x) < 0 for x < 0
/(0) = 4 /'(2) =/'(-2) = 0 /"(x) > 0 for x > 0
/(2) = 0
24. Sketch a continuous curve y =/(x) having the properties/'(x) > 0 for x < 2 and/'(x) < 0
for x > 2:
a. If/'(x) is continuous at x = 2
b. If/'(x) -> 1 as x -► 2~ and/'(x) -> - 1 as x -> 2 +
c. If/'(x) = 1 for x < 2 and /'(x) = - 1 for x > 2
 2.12 Applications of Derivatives in Business and Economics / 245

25. Sketch a continuous curve y = fix) having the following properties:

a./(2)=/'(2) = 0 ' b. f(2) =/'(2) = 0
/(0) = 2 /(0) = 2
f"[x) < 0 for x < 1 and x > 3 f"(x) > 0 for all x
/"(x) > 0 for 1 < x < 3
26. Sketch a continuous curve y =f(x) having all of the following properties:
/(0) = 10 /'(x) > 0 for x < 0 /"(x) > 0 for x < 0
/(— 3) = 0 f (x) < 0 for x > 0 /"(x) > 0 for x > 0
/(3) = 0
27. Show that a cubic curve whose equation is of the form

y = ax3 + bx2 + cx + d a, b, c, d f 0

has one (and only one) point of inflection.

28. Sketch a continuous curve y =/(x) for x > 0 if/(l) = 0 and f'(x) = - for x > 0. Is such a
X

curve necessarily concave upward or downward?

The curves representing the functions discussed on pages 221-229 were sketched without
discussion of points of inflection; investigate points of inflection and sketch the curves
accurately.

Answers to Odd-Numbered Problems

1. max ( — 2, 28) 13. no max
min (2, —4) min (0, 0)
pt. inf. (0, 12) pt. inf. (± 76, 6)
3. max (-2, ^) 15. max (1, ^)
min (3, ~¥) min (3, 2)
pt. inf. (i, -yi) pt. inf. (2, ^)
5. no max 17. no max
no min min (-6, -i)
pt. inf. (0, 0) pt. inf. (-9, -in)
7. no max 19. max ( — 1, 0)
min (2, — 1)
no pt. inf.
9. max (0, 2)
min
7)
pt. inf. (1, 0)
min (2, —2)
pt. inf. (1, 0) 27. pt. inf. at x = — ^
11. no max
min (±4, 4^2)
no pt. inf.

2.12 Applications of Derivatives

in Business and Economics
In this section several applications of derivatives in business and economics are
discussed. Marginal cost, elasticity and marginal revenue are considered in much
more detail than in previous sections. In addition, revenue from taxation, profit
under monopoly, the effect of taxation on monopoly and inventory models are
discussed.

COST, AVERAGE COST, AND MARGINAL COST

If the total cost y of producing and marketing x units of a commodity is assumed to
be a function of x only, then the total cost function can be represented by y = /(x).
246 / Differential Calculus: Functions of One Variable

Functions of various types are used to represent total cost curves; in general, cost
curves have the following properties:
1. When no units are produced, total cost is zero or positive—that is,/(0) > 0. If
/(0) > 0, then /(0) is amount of overhead or fixed costs of production.
2. Total cost increases as x increases, so /'(x) is always positive.
3. The cost of producing an extremely large quantity of any commodity usually
reaches a point at which it increases at an increasing rate. Thus usually the total
cost curve eventually is concave upward, that is,/"(x) > 0; however, in a limited
range the total cost curve is frequently concave downward, corresponding to
decreasing marginal cost.

If the total cost function is

y=f(x)
then the average or per unit cost is

and the marginal cost is

dy
dx
=/'(*)

The first derivative of average cost (the marginal average cost) is

dy = xf'jx) -f(x)
dx x2

=0 if and only if x/'(x) —f(x) = 0

that is,/(x) = x/'(x),

® -m
Thus average cost is minimum at the value of x for which average cost equals
marginal cost; that is, the average and marginal cost curves intersect at the point of

minimum average cost. Note that the value of x (if any) for which ^ = 0 is assumed
UA

to be a minimum because of the third property of cost curves mentioned above; for a
particular total cost curve the existence of this minimum can be checked in the usual
way.

EXAMPLE

Linear cost functions:

Total cost: y = ax + b, where a > 0, b > 0

Average cost: y = y/x = a + (b/x)

Marginal cost: ~ — a
dx
w • , dy b
Marginal average cost: — =-y
dx x
 2.12 Applications of Derivatives in Business and Economics / 247

Figure 2.53

y*y

Total cost and marginal cost are represented by straight lines and average cost is repre¬
sented by the first quadrant branch of an equilateral hyperbola with horizontal asymptote
y = a. Average cost is thus a decreasing function of the number of units produced; it has no
minimum value but approaches a, the marginal cost, as the number of units increases (see
Figure 2.53).

EXAMPLE

Quadratic cost functions:

Total cost: y = ax2 + bx -f c, where u > 0, b > 0, c > 0

Average cost: y = y/x — ax -f b + (c/x)

Marginal cost: ^ = 2ax + b

ax
., . i dy c
Marginal average cost: — = a-j
dx x

Total cost is represented by the first quadrant part of a parabola, average cost is repre¬
sented by the first quadrant branch of a hyperbola, and marginal cost is represented by a
straight line (see Figure 2.54).

dy = 0 if ax2 = c
dx

x — + I -, but only x = / C is of interest

a V a
772
dy — 2cx 2c
—-7— — —7 > 0 for x > 0, so minimum at x =
dx' x x

Note that average and marginal cost are equal if x = yfcja, since
248 / Differential Calculus: Functions of One Variable

Figure 2.54
y

and

EXAMPLE

Cubic cost functions:

Total cost: y = ax3 + bx2 -f cx + d, where a > 0, b < 0, c > 0, d > 0, b2 < 3ac

Average cost: y = - = ax2 + bx + c -F -

x x

dy
Marginal cost: 3ax2 T 2bx + c
dx

dy
Marginal average cost: 2 ax + b
dx
 2.12 Applications of Derivatives in Business and Economics / 249

Figure 2.55
y

Total cost is represented by the first quadrant part of a cubic curve, average cost is
represented by the first quadrant branch of a hyperbola, and marginal cost is represented by
the first quadrant part of a parabola (see Figure 2.55).
Cubic cost functions are frequently appropriate when the concavity of the cost function
changes in a specified interval; however, such a function must have no relative maximum or
minimum in the first quadrant:

dy ,
— = 3 ax + 2 bx + c
dx

— 0 if 3ax2 + 2bx + c = 0

— 2b ± a/462 — 12ac —b± yjb2 — 3ac

6a 3a
If b2 — 3ac < 0, y = ax3 + bx2 + cx + d has no relative maximum or minimum.
If by- — 3ac = 0, x = —b/3a and

=a(^)+ m+c(^j+d
~2f>3 bc
3a +

= b(2b> - 9ac) ,
250 / Differential Calculus: Functions of One Variable

2b2 - 9ac < 0, since b2 — 3ac < 0 and thus, since a > 0 and b < 0, b(2b2 — 9ac)/21a2 > 0 and
{-b/3a, [b(2b2 - 9ac)/21a2] + d} lies in the first quadrant.

d2y
= 6 ax + 2b
dx2

d2y
=0
dx2 x = — b/2>a

= 6a^0

So no relative maximum or minimum at x = — b/3a but a point of inflection at x = — b/3a.

Thus, if b2 — 3ac < 0, then y = ax3 + bx2 + cx + d has no relative maximum or minimum in
the first quadrant; a point of inflection occurs at

b(2b2 — 9 ac)
3 a’

Note that b2 — 3ac < 0 is a sufficient but not a necessary condition for y = ax3 +
bx2 + cx + d to have no relative maximum or minimum in the first quadrant. Other
cases are discussed in Technical Note II.
Frequently the tangent at the point of inflection provides a good linear approxi¬
mation of a cubic cost function for a limited range. The tangent at the point of
inflection x = —b/3a has slope

dv -
— = 3 ax2 + 2 bx + c
dx

dy b2 lb2
dx x=-b/3a 3a

b2
= c — —~ which is > 0, since b2 < 3ac
3a

If x = —b/3a,

b
y = a\- + b\ + c - •F d
21a' 9 a' 3a,

2b3 be
Tla1 ~ 3a +

The line through the point ( — b/3a, 2b3/21a2 — (bc/3a) + d) and having slope
c — (b2/3a) is given by

y - yl = m(x- Xj)

2b3 be \ 1 b2\ I b
y - I ■ - — + d i = I c - —-11 x + —
21 a2 3a 3a 3a,

b2\ be b3 2 b3 be ,
y-'C~Ta)X + Ja-9ai + T7^~Ta + d
b2\
= c - — X - + d
3a 21a:
 2.12 Applications of Derivatives in Business and Economics / 251

EXAMPLE

Higher-order polynomial cost functions:

Total cost: y = axb + c, where a > 0, b > 1, c > 0

y
Average cost: v = - = ax
h~ i c
-f -
x x

Marginal cost: ^ = abxb 1

ax

dy ,__
Marginal average cost: — = — l)x
U

Total cost is represented by the first quadrant part of an algebraic curve, average cost is
represented by the first quadrant branch of a hyperbola, and marginal cost is represented by
the first quadrant part of an algebraic curve (see Figure 2.56).

Figure 2.56

y
252 / Differential Calculus: Functions of One Variable

This type of simple polynomial curve is always concave upward for x > 0 and thus in
practice is usually appropriate only for limited ranges of x.

dy
= 0 if a(b — l)xb = c
dx
1 lb

x =
a(b — 1)

dy2 2c
= a(b - 1 )(b ~ 2)xb~3 +
dx" ' '' ' x'

= x~2[a(b — 1 )(b — 2)xb + 2c]

1 lb
x
If x = , then
a(b — 1)
- 3/f>
dy2 c
dx2 a(b — 1) a{b ~l){b ~ + lc

>0
1 /b
so minimum at x =
a(b — 1)
Note that average cost and marginal cost are equal if x = (c/a(b — l))1/b, since
(b-l)/b / ^ \-l lb
y = a\ + c
a(b - 1), a(b — 1);
(■b~ 1 )/b
a + c
a(b — 1), ci(b — l)y
(b - l)/b
[a 4- a(b — 1)]
ya(b — l)y
(b~ 1 )lb
= ab I
a(b — l)y

and
(b- l)/b

d£ = ab[^T),

EXAMPLE

Exponential cost functions:

Total cost: y = aebx, where a > 0, b > 0

v aebx
Average cost: y — - = —-
x x

Marginal cost: ^ = abe bx

dx

dy abxebx — aebx aebx(bx — 1)

Marginal average cost:
dx X X
 2.12 Applications of Derivatives in Business and Economics / 253

Figure 2.57
y

Total cost is represented by the first quadrant part of an exponential curve, average cost is
represented by the first quadrant branch of an exponential hyperbola, and marginal cost is
represented by the first quadrant part of an exponential curve (see Figure 2.57).
Exponential and logarithmic curves are either concave upward or concave downward for
all x > 0 and thus in practice are usually appropriate only for limited ranges of x.

^- = 0 if aeb*(bx - 1) = 0
dx
1
X~b
dy2 (abebx + ab2xebx — abebx)x2 — (abxebx — aebx)( 2x)
dx2 x4
axebx(b2x2 — 2 bx + 2)

If x = 1 /b, then

£*(1-2 + 2)
dy2
dx2 I
¥
= ab2e > 0
254 / Differential Calculus: Functions of One Variable

so minimum at x = l/b. Note that average cost and marginal cost are equal if x = 1 /b, since

y = abe

and

PROBLEMS
1. For each of the following average cost functions, find the minimum average cost and show
that at the minimum average cost, marginal cost and average cost are equal.

a. y = 25 — 8x + x2 e. y — 2x + 5 + -

b. y = 2 + x \n x f. y = 20 + 2x2 +-4x4

c. y = 2ex + e~x g. y = 10 - 4x3 + 3x4

, c
d. y — 3x + 5 H—
6 , _ , _ 36
h. y — 6x + 7 + —
x

2. For each of the following total cost functions, find marginal cost and determine the nature
of marginal cost (increasing, decreasing).
a. y = lOOOx - 180x2 + 3x3 b. y = 220 + 55x - 2x3 + x4
3. Determine the nature of the marginal and average cost functions (increasing, decreasing)
for each of the following total cost functions.
a. y = yJx + 25, 0 < x < 10 b. y = 9x + 5xe~2x
4. For the following total cost function, find the equation of the tangent at the point of
inflection as an approximation of the function near that point.

y = x3 — 6x2 + 14x + 6

5. The Precision Machine Tool Manufacturing Company has a total cost function repre¬
sented by the equation yc = 2x3 — 3x2 — 12x, where y represents cost and x represents
quantity.
a. What equation represents the marginal cost function?
b. What is the equation for the average cost function? At what point is average cost at its
minimum?
c. Is this a set of equations one might realistically expect to find in practice? Why?
6. The Minute Man Colonial Furniture Company’s total revenue function is expressed by the
equation R = 24x — 3x2, where R represents revenue and x represents quantity.
a. What is the maximum revenue the company can expect, assuming this equation is
valid?
b. What equation represents the average revenue function?
c. What equation represents the marginal revenue function?
d. On one graph, plot the total, average, and marginal revenue functions.
7. Arto Company manufactures cabinets for TV sets; the total cost of producing a certain
model is given by y = 4x — x2 4- 2x3, where y represents total cost and x represents
quantity, in thousands of units. The sales department has indicated that between 2 and 6
units (in thousands) should be produced. At what quantity is marginal cost a minimum?
Explain your answer, and graph the marginal cost curve.
8. If the general formula for the total cost function is

TC =/(x) = ax3 + bx2 + cx + d

a. What is the corresponding equation for the marginal cost function?

b. What is the corresponding equation for the average cost function?
 2.12 Applications of Derivatives in Business and Economics / 255

9. An important generalization is often made about the relationships between the total
revenue (TR), average revenue (/!/?), and marginal revenue (MR) functions:

When MR = 0, TR is at a maximum point

When MR > AR, AR is increasing
When MR < AR, AR is decreasing
When MR = AR, AR is unchanging

a. Illustrate this with an example from your reading or an example from the text.
b. Work out the analogous relationships for the total cost (TC), average cost (AC), and
marginal cost (MC) functions.
10. Henderson and Quandt in their book Microeconomic Theory state the following*:

“A monopolist is free to select any price-quantity combination which lies on his nega¬
tively sloped demand curve. Since an expansion of his output results in a reduction of his
price, his MR is less than his price. His first-order condition for profit maximization
requires the equality of MR and MC. His second-order condition requires that MC be
increasing more rapidly than MJR.” (Note that increasing more rapidly is logically equiva¬
lent to decreasing less rapidly.)

Which, if any, of the three diagrams in Figure 2.58 satisfies both conditions?

Figure 2.58

y y y

Answers to Odd-Numbered Problems

1• y min ^

b- y’min 2
e

C. y • = 23/2
d. ymin = 5 + 6/2
e. J’min = 17
f- J’min = 20

§• 3’min ^

h- Jmin = 7 + 12/6

3. a. for 0 < x < 10, marginal and average cost always decreasing
b. marginal cost decreasing for x < 1 and increasing for x > 1; average cost always decreas¬
ing; yc -» 9 as x -> x

* J. M. Henderson and R. E. Quandt, Microeconomic Theory, a Mathematical Approach, McGraw-Hill,

New York, 1958, p. 198.
256 / Differential Calculus: Functions of One Variable

5. a. MC = 6x2 — 6x — 12

b. AC = 2xz — 3x — 12, min at x = j

c. no, TC not in first quadrant for x <

7. in interval 2 < x < 6, MC is minimum for x = 2

9. b. when MC = 0, TC is maximum
when MC > AC, AC is increasing
when MC < TC, TC is decreasing
when MC = TC, AC is unchanging

ELASTICITY
It is frequently of interest to consider the relative change in the dependent variable
which results from a small relative change in the independent variable. The ratio of
the relative change in the dependent variable, y, to the relative change in the indepen¬
dent variable, x, is referred to as the elasticity of y with respect to x. The elasticity of y
with respect to x measures the responsiveness of y to changes in x. Because relative
changes are used in the definition of elasticity, the elasticity of a function is dimen¬
sionless. That is, elasticity is independent of the units in which the variables are
measured. Elasticities are frequently used in measuring the responsiveness of demand
or supply to changes in price or income.
There are two types of measurement of elasticity, arc elasticity and point elasticity.
Arc elasticity measures the elasticity of a function between two points, that is, over an
arc. Unfortunately, the definition of arc elasticity is somewhat ambiguous, as dis¬
cussed below. However, for a continuous function of a continuous variable, point
elasticity, the elasticity of a function at a specific point, can be defined
unambiguously.
If y = /(x), the elasticity of y with respect to x is defined by

Ey _ Ay /Ax x Ay
Ex y y Ax
Ay dy
where, as Ax -* 0, . Elasticity is frequently denoted by t].
Ax dx
In general, the elasticity of a function varies over its range. For example, in Figure
2.59, note that elasticity of the function over the two arcs noted is not the same;
Ay Ay' x xr Ey Ey'
although — = ——, -7^— and thus —y- ^—7. Figure 2.59 also illustrates the
Ax Ax y y Ex Ex
ambiguity in defining arc elasticity since, for example, it is not clear whether
Xj/yq or x2/y2 should be used for x/y.
The following formulas are frequently used for measuring arc elasticity between
the points (xx, yt) and (x2, y2).

El = *1 . ZlTZ1 = . Ay
Ex yx x2 — xx yx Ax

Ry = *2 yi - y 1 _ *2
Ex y2 x2 — Xj y2 Ax
Ey = x! + x2 . y2 - yx _ xx -F x2 _ Ay
Ex yx + y2 x2 - xx yi + y2 Ax
 2.12 Applications of Derivatives in Business and Economics / 257

Figure 2.59
y

The first formula is an approximation of the point elasticity at (x1} yj. The second
formula is an approximation of the point elasticity at (x2, y2). The third formula is
the average of the first two and is probably the most commonly used.
For point elasticity, this ambiguity does not arise. The point elasticity ofy = /(x)
at (xl9 yx) is

Ey = x± dy
Ex Vi
J 1
dx xi>
y yi

and, in general, the point elasticity of y with respect to x is given by

dy

= Ey = -L = x. dy
Ex dx y dx
x

The ratio of the derivative of a function to the function itself is called the logarith¬
mic derivative of the function; that is, if y =f (x), then

dy
f'(x) dx 1 dy d
In y
f(x) y ydx dx

is the logarithmic derivative of /(x). The elasticity of the function y —f{x) at the
point x is the ratio of the logarithmic derivative of y to the logarithmic derivative of
x:

1 dy
Ey y dx x dy
Ex 1 dx y dx
In x
dx x dx

Elasticity is defined as a property of any differentiable function; however, in

economic theory elasticity is most frequently considered for demand, supply, price,
cost, and revenue functions. Elasticity is used, for example, as a measure of the
responsiveness of demand or supply to changes in price or income, and the respon¬
siveness of price, total cost, or total revenue to changes in quantity.
258 / Differential Calculus: Functions of One Variable

Figure 2.60
y y y

y y

Perfectly elastic Perfectly inelastic

A ✓V

The classical and by far most frequent use of the concept of elasticity is in analyz¬
ing the responsiveness of demand for a commodity to changes in its price. Because
the slope of a demand curve is negative, its first derivative is negative, and thus rj < 0.
Since elasticity is dimensionless, the responsiveness of different commodity demands
to price changes can be compared. For this purpose, demand is frequently classified
in general categories as perfectly elastic (rj = — oo), relatively elastic (r/ < — 1), unit
elastic (r/ = — 1), relatively inelastic (— 1 < rj < 0), and perfectly inelastic (rj = 0).
The price elasticity of demand for a commodity is of interest to a firm; for
example, if demand is price elastic at a given price, a price decrease raises quantity
demanded proportionately more than the price decreases, so that total revenue,
which is the product of price and quantity, increases. When demand is unit elastic,
total revenue is unchanged by a decrease in price; when demand is inelastic, total
revenue is decreased by an increase in price. Demand curves having differing elastici¬
ties are shown in Figure 2.60. Note, however, that elasticity is not generally constant
over the range of a function; note also that slope is only one aspect of elasticity and
should not be equated with elasticity.
The concept of cross elasticity is useful as a measure of the relationship between
the demands for two or more commodities. Cross elasticity measures the respon¬
siveness of demand for one commodity to changes in the price of another commo¬
dity. The elasticity of demand for A with respect to price of B is defined as
dx*
Ex* = Xj_ = Ib . dxj
EyB dye xA dyB
yB
and arc cross elasticity as yB changes from ytt[ to yBl and xA changes from xAl to x l2
is given by
Exj = yBi + yB2 . xa2 ~
EyB xAl + xA2 y„2 - yBl
 2.12 Applications of Derivatives in Business and Economics / 259

where the x^’s are the quantities of A, the yB’s are the prices of B, and the price of A is
assumed to be constant. The elasticity of demand for B with respect to price of A is
similarly defined.
When commodities are substitutes for each other, their cross elasticities are posi¬
tive. For example, when the price of butter increases, the consumption of margarine
also increases, assuming that the price of magarine and other relevant factors remain
constant. When commodities are complementary to each other, their cross elastici¬
ties are negative. For example, when the price of lettuce increases, the consumption
of salad dressing decreases, assuming that other relevant factors remain constant.

note: Demand functions are customarily written with price as the dependent
variable and quantity demanded as the independent variable and are graphed in a
corresponding manner. However, in analyses involving demand elasticities, demand
is logically considered to be the dependent variable and price the independent
variable. In the examples below, the demand function is given in one form in some
cases and in the other form in other cases and is represented graphically in both
forms. When a demand function is given in the form y = /(x) and it is not convenient
dx
to solve for x, — can be obtained by implicit differentiation.
dy
Because demand functions are conventionally written y = /(x), where y is price
dx
and x is quantity demanded, and the elasticity of demand involves the derivative —,
dy
equations and graphs must be read carefully to avoid misinterpretation.

EXAMPLE

The demand function for a particular commodity is given by y = yj{20 — x)/2 for 0 < x < 20,
where y is price per unit and x is number of units demanded. Consider the point y = 2, x = 12.
If the price decreases by 6%, determine the corresponding increase in demand and an approxi¬
mation to the elasticity of demand at the point y = 2, x = 12; compare this with the exact
elasticity of demand at the point y = 2, x = 12. Then compute an approximation to the
elasticity of demand and the exact elasticity of demand at the point representing the changed
price and quantity; also compute the arc elasticity based on average price and quantity (see
Figure 2.61).

Figure 2.61
y
260 / Differential Calculus: Functions of One Variable

Figure 2.62
X

Solving the demand equation for x,

/ 20 — x

y2 = i{20 - X)
x = 20 — 2 y2
See Figure 2.62.
y1=2 y2 = 2 - (0.06)(2) = 1.88 Ay = -0.12
xx = 12 x2 = 20 - 2(1.88)2 = 12.94 Ax - 0.94
Arc elasticity of demand at y — 2, x = 12, is
Ex yx Ax
Ey Xx Ay
_ 2 / 0.94\
“12\ Oil)
= -1.31
Point elasticity of demand at y = 2, x = 12, is
Ex y dx
Ey x dy

20 - 2y2 ^ 4y^

2 y2
y2 -10

See Figure 2.63.

Ex _ 8 _
- S = -1.33
Ey y= 2

Arc elasticity of demand at y2 = 1.88, x2 = 12.94, is

Ex y2 Ax
Ey x2 Ay
1.88 /_0.94\
12.94 \ 0.12/
- -1.14
 2.12 Applications of Derivatives in Business and Economics / 261

Figure 2.63
Ex
Ey

Point elasticity of demand at y2 = 1.88, x2 = 12.9, is

Ex _ 2y2
Yy = y2 - 10

Ex
-1.09
Ey y= 1.88

Arc elasticity based on average price and quantity is

Ex _ yi + y2 x2 — 3.88 / 0.94\
Ey Xi + x2 yi ~ yi 24.94\ 0.12/

= -1.22

EXAMPLE

The demand for margarine is related to the price of butter as follows:

q = 5 + l^fp
where q is the demand for margarine (hundreds of pounds) and p is the price of butter (dollars
per pound). Find the equation for cross elasticity of demand for margarine with respect to
price of butter. What is the demand for margarine when butter is $0.81 per pound ? What is the
demand for margarine when butter is $1.00 per pound ? What is the cross elasticity of demand
for margarine as butter goes from $0.81 to $1.00 per pound? (See Figure 2.64.)
Solving the demand equation for p,

q = 5 + 2 y/p

\/P=i(q - 5)

P = Uq2 - 10q + 25)

See Figure 2.65.

262 / Differential Calculus: Functions of One Variable

Figure 2.64
4

1 he elasticity of demand for margarine with respect to price of butter is

Eq dq p 1 p
EP dp q Jp 5 + 2 jp

=
5 + 2 Vp
See Figure 2.66.

Figure 2.66
Eq
Ep

P
 2.12 Applications of Derivatives in Business and Economics / 263

When p = 0.81, q = 5 + 2V/.81 or 680 pounds.

When p = 1.00, q = 5 + 2^/1.00 or 700 pounds.

Cross elasticity is

Eq _ \fp
Ep 5 + 2^/p
For p = 0.81,

Eq _ 0.9 09
= 0.13
Ep p = o.8i _ 5 + 2(0.9) 08

For p = 1.00,

Eq 1 _ 1
0.14
Ep p= i.oo 5 + 2 7

Constant Elasticity of Demand

As noted previously, the elasticity of a function is not, in general, constant over the
range of the function. However, one type of function, the equilateral hyperbola, has
constant elasticity over its range. In many theoretical analyses, the demand function
is represented by an equilateral hyperbola, because it is convenient to have constant
elasticity of demand over the range of the function.
If the demand function is given by the generalized equilateral hyperbola

then

— = —amy 1
dy

Ex y dx
Ey x dy

= —m

That is, elasticity of demand is the constant — m. In terms of approximate changes in

price and quantity demanded, an increase of 1 percent in price results in a decrease of
m percent in demand at any level of price and quantity. Only an equilateral hyper¬
bola has constant elasticity over its range. For any other type of function, elasticity
varies for different points on the curve. (See previous examples.)

EXAMPLE

The demand function is given by

264 / Differential Calculus: Functions of One Variable

Figure 2.67

Determine the elasticity of demand. Consider the point y — 2, x = -fj; if the price increases by
5 percent, determine arc elasticity at the point y = 2, x = f|, and at the point representing the
changed price and quantity (see Figure 2.67).
Solving the demand equation for y,

25
X = -4
y

nr
See Figure 2.68. *

Figure 2.68
v
 2.12 Applications of Derivatives in Business and Economics / 265

Figure 2.69
Ex
Ey

(Note that if 1 < y < 5, then < x < 25.)

dx —100
dy y5

Ex y dx
Ey x dy

-100\
- y f
25 1 V5 )
/
= -4

as could be obtained
2.69.

yi = 2 yi =

25
Xj = 1.56 X2 = = 1.29 Ax = -0.27
(2-1 Y
For the point yx = 2,

Ex _y i . Ax
Ey X! Ay

2 / -0.27
~ 1.56 \ 0.1

= -3.46

For the point y2 - 2. = 1.29, arc elasticity of demand is

Ex _y 2 Ax
Ey x2 Ay

2.1 / -0.27
“ 1.29 ( 0.1 ,

= -4.40

PROBLEMS
1. For each of the following demand functions, i. determine arc elasticity at the point specified,
ii. determine arc elasticity at the point corresponding to the specified change in price or
266 / Differential Calculus: Functions of One Variable

demand, iii. determine point elasticity at the two points, iv. determine arc elasticity based
dy dx
on average quantities and prices, and v. show that — and — are reciprocals of each other.
dx dy
a. x =
60 - 2y2; x = 10, y = 5; price decrease 8%
b. x +
2y = 15; x = 7, y = 4; price increase 5%
c. x =
25 — 5y2; x = 5, y = 2; price increase 5%
d. x =
10 - 5y2; x = 5, y = 1; price increase 10%
e. y = (x — 10)2; 0 < x < 10; x = 8, y = 4; demand decrease 5%
f. x = 19 - 4y2; x = 3, y = 2; price decrease 5%
g. x = 36 - 4y2; x = 20, y = 2; price increase 5%
h. y = (x — 4)2; x = 1, y = 9; demand increase 30%
2. Find elasticity of demand, y, with respect to price, x, for each of the following functions.

3 d. y = 13c (5/4)*
a- y =
1 + 2x2
10
e- x - ^574
b. y = (x — 8)2, 0 < x < 8

c. y ae -kx f. y = 100 - 15x2

Answers to Odd-Numbered Problems

1. a. -9.6, --5.00, --10, -4.79, - 6.73
b. -1.14, -1.27, -1.14, -1.27, -1.21
c. -8.2, --14.59, -8, -14.95, --10.57
d. -2.1, --2.92, --2, -3.06, -2 .46
e. -0.11, -0.17, -0.13, -0.17, -0.14
f.-10.4, -6.5, --10.67, -6.33, -8.05
g- -1.64, -1.88, -1.6, -1.92, -1.75
h. -1.58, -.98, --1.50, -1.04, -1.24

REVENUE, MARGINAL REVENUE, AND ELASTICITY OF DEMAND

For any given demand function y = /(x), the total revenue R, is the product of x, the
number of units demanded, and y, the price per unit quantity demanded,
R = xy — x • f (x)

and the marginal revenue with respect to demand is

dR dy
dx X dx ^
Elasticity of demand with respect to price is

Ex y dx
Ey x dy

, . dx 1
and, since — = —,
dy dy
dx

Ex _ y J_\
Ey x dy
dx I

<fy = _ y
dx Ex
x
Ey 1
 2.12 Applications of Derivatives in Business and Economics / 267

Thus

dR
+ T
dx Ex

j_\
Ex
Ey,

Alternatively,

dR
dx
+y

x dy
y ~“7
ydx
+ 1

j_\
y 1+ Ex
Ey,

That is, marginal revenue is the product of the price per unit quantity demanded and
one plus the reciprocal of the elasticity of demand.
Since both x and y are either zero or positive, the total revenue, R = xy, is also
either zero or positive; however, the marginal revenue may be either positive or
negative.

note: Logically, R could also be considered a function of price

R = xy = y ■ g(y) = G{y)

and marginal revenue with respect to price defined as

dR
+x
dy
This would perhaps be simpler for cases when the demand function is in the form
x = g(y) and it is not easy to solve for y in terms of x. However, to keep the
geometrical representation consistent (that is, to use x as the independent variable), it
dR d2R
is preferable to use implicit differentiation and the chain rule to obtain and
dx dx2
when the demand function is in the form x = g(y).

EXAMPLE

The demand for a particular commodity is given by

y = (12 - x)1/2 for 0 < x < 12

where x is quantity demanded and y is price per unit.

Determine the price and quantity for which revenue is maximum. Show for this demand
function that the relationship between marginal revenue and elasticity of demand holds.

Demand: y = (12 — x)1/2 for 0 < x < 12

268 / Differential Calculus: Functions of One Variable

Revenue: R = x(12 — x)1/2

dR
= (12 — x)1/2 — \x(\2 — x) 1/2
dx
(12 - x) 1/2[12 — x — ix]
3(8 - x)
2(12 - x)1 2
0 if x = 8, R = 16
112
d2R 3 j-(l^-x)1/2 - (8 - x)(-j)(12 - x)
dx2 2 12 - x
3 (12 — x)_1/2[— 12 + x + i(8 — x)]
2 12 — x
3 x — 16
4 (12 - x)3/2
< 0 if x = 8, so maximum at x = 8, y = 16
<0 for 0 < x < 12, so no point of inflection

Figure 2.70
R

dR
dx

X
 2.12 Applications of Derivatives in Business and Economics / 269

Thus R = x( 12 — x)1/2 is concave downward for 0 < x < 12. See Figure 2.70.

Ex y dx
Ey x dy

= -/-
xl dy
\dx

(12-x) 1/2 1

X —i (12 — x) “I/2

-2(12 - x)

dR /, 1 \
1+

\ Ey /

= (12 - x)1/2 1 -
2(12 - x)

(12 - x)1/2(24 - 3x)

2(12 - x)

3(8 — x)
(as above)
2(12-x)1/2

EXAMPLE

The demand for a particular commodity is given by

y = 15c~x/3 for 0 < x < 8

where x is quantity demanded and y is price per unit.

Determine the price and quantity for which revenue is maximum. Show for this demand
function that the relationship between marginal revenue and elasticity of demand stated on
page 267 holds.

Demand: y = \5e~x,2> for 0 < x < 8

Revenue: R = 15xc x/3

dR ■ x/3
= — Sxe x/3 + 15c
dx

= 5e */3(3 — x)

45
= 0 if x = 3, R = —
e

d2R
= -t<T*/3(3 - x) - 5e~xl3
dx2

= 5e Jc/3(^x — 2)

45
< 0 if x = 3, so maximum at x = 3, R — — « 16.6
e

= 0 if x = 6
270 / Differential Calculus: Functions of One Variable

Figure 2.71

0=-^x-2) + |,--
= 5e_x/3(l — ^x) >0 if x = 6

90
so point of inflection at x = 6, R = « 12.2. See Figure 2.71.
e

Ex y dx
Ey x dy

15 e~x/3{ 1 \
x \ — 5e~x/3)

_ 3
x
 2.12 Applications of Derivatives in Business and Economics / 271

dR
dx
y 1 +
Ey

= 15e~x/3

= 5e~x,3(3 - x) (as above)

PROBLEMS
1. When Mr. Smythe’s income was $300 per month he bought 20 qt of milk per month. When
his income increased to $350 per month he bought 24 qt of milk per month. Assuming no
change in the price of milk or other relevant factors, what was Mr. Smythe’s income
elasticity of demand for milk?
2. When the price of A was $5.00, 100 units of B were sold. When the price of A fell to $4.00,
120 units of B were sold. What is the cross elasticity of demand for A in terms of the price of
£? Are A and B substitutes or complements?
3. What is the relation between the slope (positive or negative) of the average cost curve and
the elasticity of total cost?
4. Demonstrate algebraically that the elasticity of total cost is equal to

„ MC
Etc~ac

where MC is marginal cost and AC is average cost.

5. For each of the following demand functions demonstrate the relationship between marginal
revenue and elasticity of demand given by

dR
dx 1 + ~k
Ey

a. y = 550 — 3x — 6x2 d. y = 86 — 25x

e. y = 100 — 6x2

c. y = 17 — 6x

6. Show that if demand is linear and negatively sloped, the elasticity of total revenue is always
less than 1, but to the left of its maximum value the elasticity of total revenue is positive and
to the right of the maximum value the elasticity of total revenue is negative.

Answers to Odd-Numbered Problems

i 13
t- TT
3. when slope of AC is negative, ETC is less than 1
when slope of AC is zero, ETC is equal to 1
when slope of AC is positive, ETC is greater than 1

REVENUE FROM TAXATION

If the government imposes a tax on a given commodity, it is assumed that the price to
the consumer will increase and correspondingly the quantity demanded will
decrease. The discussion in this section concerns the effect of taxation on market
272 / Differential Calculus: Functions of One Variable

Figure 2.72
y> yt

equilibrium under the following conditions: there is pure competition in which the
consumers’ demand depends only on price, the demand function does not change, the
producers adjust the supply curve to the new price which includes the tax, and a tax
of t monetary units is imposed on each unit of quantity produced.
The supply function can be represented by

y = g(x)
where x is the number of units of the commodity supplied and y is the price per unit.
If a tax of t per unit quantity is imposed, the supply function after taxation is

y, = g(x) +1
If the demand function is

y =/M
then the equilibrium point before taxation, E(x, y), is the solution of the equations
Demand: y =f(x)
Supply: y = g(x)
and the equilibrium point after taxation, E(xt, y,), is the solution of the equations
Demand: y = /(x)
Supply: yt = g(x) + t
See Figure 2.72.
note: If the supply function before taxation is in the form
x = G(y)
it may be possible to solve for y in a convenient form. If not, the supply function after
taxation,
y,-t = g(x)

shows that the quantity supplied is

x = G(y, - t)
 2.12 Applications of Derivatives in Business and Economics / 273

and equilibrium after taxation occurs at the solution of the equations

Demand: y =f(x)
Supply: x = G(yt — t)
Geometrically, this is equivalent to moving the original supply curve upward t units.
Except for the case of constant price, where price does not depend on quantity
produced, the increase in equilibrium price is less than the amount of the tax. Note
that a subsidy may be considered a negative tax. The supply curve is moved down¬
ward the amount of the subsidy, price to the consumer decreases, and demand
increases.

The total tax revenue T received by the government from a tax of t per unit
quantity is
T — txt

where x, is the equilibrium quantity after taxation. This total may be represented by
the area of a rectangle of dimensions t and x,, which, if the supply function is linear, is
equivalent to the parallelogram in Figure 2.72. As this diagram indicates, the tax may
be large enough to reduce demand to the vanishing point. If the government is
interested in taxation for revenue purposes, the question is what amount of tax will
maximize revenue—if the tax t is considered as a variable, then there is no revenue if
t = 0 and there is also no revenue if t is large enough so that demand is zero; for some
intermediate value of t the tax revenue T = txt is maximum. Since T may be con¬
sidered a function of either t or x, maximum revenue possible from taxation can be
determined by considering marginal tax revenue with respect to either t or x. It is
usually more convenient to consider T as a function of x, since t occurs linearly in the
relation between t and x.

EXAMPLE

The demand and supply functions for a particular commodity are

Demand: 2y + x = 14
3 x
Supply: y = - + -

Determine the maximum possible revenue from taxation which can be obtained from a tax of t
per unit quantity and the corresponding tax rate.
The supply function after taxation is y = | + (x/3) + t. Thus, after taxation, at equilibrium,

x 3 x
y= 7 — = —i-hr
2 4 3
t =
25
lx
T = tx = ^x — |x2

dT 25 5v
= '~ J*
dx
15
= 0 if x = ~4T

d2T Oc 15
— ~3 < 0, so maximum at x ~sr
dx2
25
If X :- 15 tt —
— ~r
25 — i«) — Ts , and T 375
= ~JT- So the maximum possible revenue from
taxation is Tmax = obtained from a tax of t = ^ per unit quantity. (See Figure 2.73.)
274 / Differential Calculus: Functions of One Variable

Figure 2.73

y> yt

Alternatively,

* = ¥ - jt
r = 1it- ft2

dT _ 15 12,
~dt~ 2 " 5t

=0 if t = f
d2T
-p- = — Ap < 0, so maximum at t = ^

rr _ /15\/25\ 6/25\2 _ 375

Jmax ~ t 2 A 8 / — 5t 8 ) ~ 32

EXAMPLE

The demand and supply functions for a particular commodity are

Demand: y = 30 — 2x2
Supply: y = 3 + x2
Determine the maximum possible revenue from taxation which can be obtained from a tax of t
per unit quantity and the corresponding tax rate.
The supply function after taxation is y = 3 + x2 + t. Thus, after taxation, at equilibrium,
y = 30 - 2x2 = 3 + x2 -l- t
t = 27 - 3x2
T = tx = 21x — 3x3
dT _
27 - 9x2
dx
= 0 if x = x/3
d2T
= — 18x <0 if x = n/3, so maximum at x = x/3
dx2
If x = y/3, t = 27 — 9 = 18, and Tmax - 18^/3. Thus the maximum possible revenue from
taxation is Tmax = 18^/3, obtained from a tax of t = 18 per unit quantity. (See Figure 2.74).
 2.12 Applications of Derivatives in Business and Economics / 275

Figure 2.74
y

Alternatively,

/21 -t
x =
3
27 - t\ 1/2
T=t\
3 )
/27t2 - t3'
( 3 )
(21 - t \ -11/2
(18 -0

=0 if t = 18

d2T I 1 \ /1 \ /27 — t\ ~3/2 1/27 - f\~1/2

(— 1)(18 — t) —
dV 21\2)\ 3 2l 3~
- 3/2

-m
t (21 - t\ ~3/2
= — — I—^—I <0 if t — 18, so maximum at t - 18

T — 11 Q°
27 - 18\1/2
1 max ) = 18^3

PROFIT UNDER MONOPOLY

In the usual forms of imperfect competition, it is assumed that the demand function
y =f{x) is known and that the price the consumer must pay depends only on the
quantity demanded. In a monopolistic situation, the monopolist controls the price by
regulating the supply of the commodity—when supply is limited, price is relatively
high, and when supply increases, price decreases.
276 / Differential Calculus: Functions of One Variable

If yc is the average cost of producing a unit of commodity (as a function of the

quantity produced), then the total cost yc of producing x units is

yc = xyc
(Note that in previous sections y has been used to denote price in some contexts and
cost in others; the subscript c is used here to differentiate cost from price.)
Presumably the monopolist will control the supply x and thus the price y
(determined by the demand function) so as to maximize his profit. The total revenue
he receives is
R = xy
where y =/(x); the total profit P is the difference between total revenue and total
cost
P = R- yc = xy - xyc
P has a relative maximum if and only if
dP dR dyc
= 0 that is
dx dx dx
and
d2P d2R d2yc
—y < 0 that is
dx dx2 dx2
In order to be meaningful this maximum must occur in the interval for which the cost
and demand functions have economic significance.

note: If the demand function is in the form

x = F(y)
and it is not possible or convenient to solve for y in terms of x, profit may be
expressed as a function of y and the problem solved with y as the independent
variable.

EXAMPLE

The demand function for a particular commodity is

y — 26 — 2x — 4x2
and the average cost to the monopolist of producing and marketing the commodity is
yc = x + 8
Determine the maximum profit obtainable by the monopolist.
Revenue: R = xy = 26x — 2x2 - 4x3
Total cost: yc = xyc = x2 + 8x
Profit: P — 26x — 2x2 — 4x3 — x2 — 8x
= 18x — 3x2 — 4x3

dP
—— = 18 — 6x — 12x2
dx
= 0 if 2x2 + x — 3 = 0
(2x + 3)(x — 1) = 0

x = 1
(x = — \ is not meaningful for this problem.)
 2.12 Applications of Derivatives in Business and Economics / 277

d2P
— 6 — 24x
dx2

< 0 if x = 1, so maximum at x = 1

1P max =11
A A

Alternatively,

dR
= 26 — 4x — 12x2
dx

dye
2x + 8
dx

dR dyc
if 26 — 4x — 12x2 = 2x + 8
dx dx

2x2 + x - 3 = 0

(2x + 3)(x — 1) = 0

x = 1

d2R
-4 24x1
dx2 ' d2R d2yc
if x = 1, so maximum at x = 1
I dx2 < dx2

EXAMPLE

The demand function for a particular commodity is

y = 28 — 5x

and the total cost to the monopolist of producing and marketing the commodity is

yc = x2 + 4x

Determine the maximum profit obtainable by the monopolist.

Revenue: R = xy = 28x — 5x2

Total cost: yc = xyc = x2 + 4x

Profit: P = 28x — 5x2 — x2 — 4x

= 24x — 6x2

dP
24 12x
dx

dP
0 if 2 — x = 0
dx

x = 2

d2P
-12 <0, so maximum at x = 2
dx2

P
1 max 24
278 / Differential Calculus: Functions of One Variable

Alternatively,

^ = 28 - lOx
dx

l- = 2*
dx
+4
O' 1 X

if 28 — lOx = 2x + 4
II
'">3

x = 2

2—1 d2R d2yc

, y < —y-, so maximum at x
dx dx
d2yc _ 7
dx2

P =24

or alternatively,

x = ?(28 — y)

R = iy(28 - y)

yc = A(28 — y)2 4-1(28 — y)

= A(28 — y)(28 — y + 20)

= y?(28 - y)(48 - y)

dR 28 2,,
—
— s — ^y
dy

= ~t?(48 — y) — y?(28 — y)

dyc 2 76
25

dR dyc if 18 _ 2 _ _2_ _ 76
11 5 5/ — 254 25
dy dy
y = 18
d2R
dy2
d2R d2yc
< ——2“, so maximum at y = 18
d2yc dy2 dy“
25
dy2

Tmax = R - yc = 36 - 12 = 24

The following two geometric representations of profit under monopoly illustrate

the economic relationships involved and also provide a method for determining
approximate maximum profit in cases for which algebraic solution is difficult or
impossible.
If the revenue and total cost curves are drawn on the same diagram, the vertical
distance between them at any given value of x measures the profit corresponding to
production of x units of the commodity. The maximum vertical distance corresponds
to maximum profit and this occurs for the value of x where the slopes of the curves

are equal, that is, where ^ (see Figure 2.75).

ax ax
 2.12 Applications of Derivatives in Business and Economics / 279

Figure 2.75

Hyc

If the marginal revenue and marginal cost curves are drawn on the same diagram,
they intersect at a point which gives the value of x corresponding to maximum profit.
If the demand curve y =f{x) and the average cost curve yc = g(x) are also drawn on
the diagram, the vertical distance between them is y — yc. For any given value of x,
the area of the rectangle of height y — yc and width x represents profit; this area is
greatest for the value of x where the marginal revenue and marginal cost curves
intersect (see Figure 2.76).

Figure 2.76

_ dy dR
y'y" a
280 / Differential Calculus: Functions of One Variable

EFFECT OF TAXATION ON MONOPOLY

The imposition of a tax t per unit quantity on a commodity produced by a monopo¬
list increases the average cost by t and the total cost by tx. The equilibrium price and
quantity for which the monopolist’s profit is maximum after taxation are thus ob¬
tained by maximizing profit using the cost function after taxation,

yCt = yc + tx

Profit after taxation is

P = R — yCt = R — yc — tx

P has a relative maximum if and only if

dP dR dy^
-0 that is
dx dx dx

and

d2P d2R d2yCt

<0 that is
dx2 dx2 dx2

Since the marginal cost curve after taxation is the marginal cost curve before taxation
translated upward a distance t, the amount produced for maximum profit will be
decreased and the price increased after taxation. (See Figure 2.77).
The total revenue received by the government is

T — tx

where x is the amount produced after the tax is imposed. T is again a function of t
and is zero if t is either zero or large enough to tax the product out of the market.
Thus T has a maximum value obtainable in the usual way. As before, a subsidy may
be considered a negative tax and the general analysis is unchanged.

Figure 2.77

dyc dyc> dR
 2.12 Applications of Derivatives in Business and Economics / 281

EXAMPLE

The demand function for a particular commodity is

y = 20 - 4x

and the average cost to the monopolist is

(a) If a tax of t per unit quantity is imposed on the monopolist, determine his maximum
possible profit and the value of t for which tax revenue is maximized.
(b) Determine the monopolist’s maximum possible profit if a 33^% sales tax is imposed.

Demand: y = 20 — 4x

Revenue: R = xy = x(20 — 4x)

Average cost: yc = 2

Total cost: yc = 2x

Total cost after tax: yCt = 2x + tx

(a) Profit: P = R — yCt = 20x - 4x2 — 2x — tx

= (18 - t)x - 4x2

dP
= 18 — t — 8x
dx

dP „ .f 18 — f 22 + f
0 it x = —-—, y =
dx 8 ’ J 2

d2P 18 - t
= —8, so maximum if x =
dx2 8

(iS - f)2 _ 4/!8 - r\2

■P
* ma y =
8 8

(18-0:
16

r( 18 - t)
Tax revenue: T = tx =

_ 18 - 2r 9 — t
li ~ 8 "

=0 if r = 9

^2T
—i < 0, so maximum if t = 9
dt2
If t = 9,
p _ 81
■* max T5

rp _ 81
••max ”8”

See Figure 2.78.

(b) Demand function: y = 20 - 4x. So the price paid the monopolist after allowing for
33^% sales tax is

yt = |(20 — 4x)
282 / Differential Calculus: Functions of One Variable

Figure 2.78

P
<-*—i

o
Ut
Oh

(18 - /)* - 4x2

- X
Quantity

Profit: P = R — yc = |x(20 — 4x) — 2x

- 13x — 3x2

dP
= 13 — 6x
dx

-0 if x = -t?

d_2p 13
- —6, so maximum if x = “5“
dx2

P _ 169
' max 1 2

EXAMPLE

The demand function for a particular commodity is

y = 14 — 3x

and the total cost to the monopolist is

yc = x2 4- 5x
 2.12 Applications of Derivatives in Business and Economics / 283

(a) If a tax of t per unit quantity is imposed on the monopolist, determine his maximum
possible profit, the change in price, and the tax revenue received by the government as a
function of t.
(b) Determine the maximum tax revenue obtainable by the government.

Demand: y = 14 — 3x

Revenue: R = xy = 14x — 3x2

Total cost: yc = x2 + 5x

Total cost after tax: yCi = x2 + 5x + tx

(a) Profit: P = R — yCi = 14x — 3x2 — x2 — 5x — tx

= (9 — t)x — 4x2

dP
= 9 — t — 8x
dx

n 9 -t 85 3
= o l{x- — ,y=.T +r

d2p q . 9-t
—2 = —8, so maximum n x =
dx' 8

P
1 max Jlzjl. 41 9-t)
g ^1
T
_ (9 ~ tf
16

Figure 2.79

o
u.
CL,

X
284 / Differential Calculus: Functions of One Variable

If x = (9 — t)/8, y = -x + ft, which is a change in price of ft, since if t = 0 the monopolist’s

profit is maximized if x = y = ¥ (see Figure 2.79).

(b) Tax revenue: T = tx = — ^

dT _9 _ i
_ 8 4

—so maximum if r = §
df2

1(9 - f) _ 72
81
T - —
8

PROBLEMS
1. Plot the demand curve y = 120 — x and the supply curve y = x — 10, where x is quantity
demanded and y is price per unit, and indicate the equilibrium price and quantity, a.
Assume that a tax of $15 per unit is imposed on the seller; draw the new supply curve and
indicate the new equilibrium, b. Assume that instead of a tax, a subsidy of $10 is paid to the
producers; draw the appropriate supply curve and indicate the corresponding equilibrium.
2. For each of the following pairs of demand and supply functions find the maximum revenue
that can be obtained by imposing a tax of t per unit and illustrate graphically (by sketching
the demand curve, the supply curves before and after taxation, and the revenue curve).
a. Demand: y = 14 — 3x d. Demand: y = 28 — x2
Supply: y = 4 + 2x Supply: y — 4 + x2
b. Demand: y — 25 — 2x2 e. Demand: y = 45 — x2
Supply: y = 5 + x Supply: y = 6 + 2x
c. Demand: y — 50 — 4x2 f, Demand: y = 30 — 2x2
Supply: y = 5 + x2 Supply: v = 3 + x2
3. For each of the following pairs of demand and (average or total) cost functions find the
maximum profit obtainable by a monopolist.
a. y = 24 — 7x d. y = 12 — 5x
yc = 6 — x yc = 4x 4- 6
b. y = 26 — 3x2 e. y = 26 — 2x 4x:
yc = 3x2 + 2x + 14 yc = x + 8
c. y = 12 — 4x
yc = 8x — x2
4. For each of the following pairs of demand and cost functions find i. change in price,
maximum profit, and maximum revenue from taxation as functions of t if an additive tax of
t per unit is imposed on a monopolist and determine the maximum revenue that can be
obtained by such taxation, and ii. find the maximum revenue that can be obtained by the
specified sales tax.
a. y = 50 — 6x d. y = 33 — 5x2
yc = x2 + 9x 4c = 3x2
Sales tax 20% Sales tax 10%
b. y = 25 — 2x2 e. y = 20 — 4x
yc = 3x 4c = 2
Sales tax 25% Sales tax 33^%
c. y = 12 - 4x f. y = 72 - lx2
yc = 2x 4c = *2
Sales tax 33^% Sales tax 40%
 2.12 Applications of Derivatives in Business and Economics / 285

Answers to Odd-Numbered Problems

7 o lp max
a. _ 27
~r

b p 50
1 max = ^9“
C P 4
= 3
v* 1 max
d P
1 max = 1
e 1P max = 11

INVENTORY MODELS
The objective of inventory control is to minimize total inventory cost. Inventory
costs are of three types: (1) cost of placing an order or starting a production run
(setup cost); (2) cost of holding inventory, including cost of capital or interest and
storage cost (carrying cost); and (3) cost of going short, including loss of goodwill
(shortage cost). Inventory control attempts to balance the economies of large orders
or large production runs against the cost of holding inventory and, in some models,
against the cost of going short.
In practice, various situations occur with respect to the nature of demand, the
procedures involved in ordering or production, the ease of storage, the seriousness of
shortages, and so forth. All the inventory models discussed in this book (including
two in this section and one in Chapter 3) assume that demand is known and uniform.
This assumption may be appropriate, at least as an approximation, in many situa¬
tions involving demand for an input of production; however, it is entirely unrealistic
in many situations involving retail demand. For these latter situations a model
assuming probabilistic demand is more appropriate.
The models in this book also assume that the setup cost, the per unit carrying cost,
and the per unit shortage cost do not depend on the number of items involved.
Neither of the two models discussed in this section permits shortages; each of them
can be derived as a special case of a model that does permit shortages by letting the
shortage cost become very large. The first model assumes that items enter inventory
in batches; the second model assumes that items enter inventory continuously during
the period of ordering or production.
Model A: uniform demand, no shortages, batch arrival
D = demand per period
cx = setup cost
c2 = carrying cost per item per period
q = number of items placed in inventory at one time
(See Figure 2.80.)

Figure 2.80
286 / Differential Calculus: Functions of One Variable

There are D/q batches per period for which the total setup cost is (cx D)/q. The
average inventory is q/2 and the carrying cost per period is (c2 q)/2. The total cost of
inventory per period is thus

Cj D
C=
<7
dC c2 cxD
dq 2 q2

dC
If
dq

d2C _ 2cxD
dq2 q3

>0

so C is minimum if q = yj{2c1 D)/c2 • Thus yj(2c1 D)/c2 items should be placed in

inventory D/q times per period.

Model B: uniform demand, no shortages, continuous arrival

demand per period

D =
rate of arrival of items (number per period)
k =
setup cost
cx =
carrying cost per item per period
c2 —
number of items placed in inventory during buildup of
q =
inventory
= time during which items are placed in inventory
t2 = time during which items are not placed in inventory
t = h + h = Bme required by one inventory cycle

(See Figure 2.81).

Note that tx = q/k and t = D/q. The carrying cost per period is

c2q
c2(q ~ Dti)
2 2

Figure 2.81
 2.12 Applications of Derivatives in Business and Economics / 287

and the setup cost is c^D/q). So the total inventory cost per period is

d2C _ 2c1 D
>0
dq2 q3

So C is minimum if q = ^/(2c1D)/c2[l - (D/k)]. Thus y/(2c1D)/c2[l — {D/k)\ items

should be produced D/q times per period.

EXAMPLE

Alicor needs 2500 widgets per quarter. It costs $3.00 a month to store a widget. The cost of
ordering a supply of widgets (any number of widgets not exceeding 2500) is $0.50. Alicor has
two alternatives: (1) purchase the widgets periodically in lots and allow no shortages to occur
and (2) purchase the widgets periodically from a supplier who sends them continuously at the
rate of 1500 per month until an order is filled and allow no shortages to occur. Determine the
quarterly cost of the optimal policy for each of these alternatives. In order to minimize total
quarterly cost of inventory, what policy should Alicor adopt?
Considering demand and costs on a quarterly basis,

D = 2500

Ci - 0.50

c2 = 9.00

k = 4500

The optimal order quantity for alternative 1 is given by

2c i D (2)(0.50)(2500) _ 50
q =
c2 9 “ 3

and the associated quarterly cost is

c2q ci D 9(Ap) , (i)(2500)

C +-= + 50
= 150
2 q
q 2

The optimal order quantity for alternative 2 is given by

2c i D (2)(0.50)(2500)
q rw i 2 5 00\
= 25
D
c2 1 -

and the associated quarterly cost is

Cj D i(2500)
C + “ =i(9)(25)(l-|)+ 25 - 100

Alternative 2 is preferable. As can be observed by comparing the cost formulas, this is

always the case if the costs of ordering and carrying are the same for the two models.
288 / Differential Calculus: Functions of One Variable

For this problem, demand, supply and costs could be considered on a monthly basis. Since
demand and supply are assumed to be uniform, any convenient common time base can be used
for computation.

PROBLEMS
1. The Ace Manufacturing Company has a contract to supply 4000 refrigerators a year at a
uniform monthly rate. The annual storage cost per refrigerator is $50 and the setup cost for
a production run is $160. If production is instantaneous and shortages are not permitted,
determine the number of refrigerators which should be produced in each run in order to
minimize the total annual cost of inventory.
2. The Harvest Company has a contract to supply 500 tractors a month at a uniform daily
rate. The monthly storage cost per tractor is $10 and the setup cost for a production run is
$200. Production is at a constant rate of 1000 tractors per month. If shortages are not
permitted, determine the number of tractors which should be produced in each run to
minimize the total monthly inventory cost.
3. Profitco Electronics requires 25 transistors per month to make color TV sets. The monthly
storage cost is $2.00 per transistor and every time an order is placed it costs Profitco $4.00.
a. If orders are filled immediately and no shortages are permitted, how many orders of what
size should be placed every month? What is the monthly total cost of this inventory policy?
b. Suppose Profitco can order so that delivery is at the rate of 45 per month (in a uniform
flow, not a block) and that no shortages are permitted. How much inventory should be
accumulated before stopping the flow and how many times per month will the flow be
stopped? What is the total monthly cost of this inventory policy?
4. Fabric Corporation owns a machine which can produce 25 bolts of a particular type of
cloth per day and another machine which can make 10 bolts of this cloth into slipcovers per
day. It costs $1.50 to hold a bolt of cloth in inventory for a day and it costs $2.00 every time
the machine producing the cloth must be stopped, a. What amount of inventory should be
built up before the machine producing the cloth is stopped in order to minimize total
inventory cost ? b. How many times will the machine be shut down in a period of 90 days ? c.
What is the total amount of cloth that will be produced in a period of 90 days? d. If the
demand and cost conditions remain the same, except that the cost of stopping the machine
is interpreted as the cost of ordering, and the cloth must be ordered in batches from another
company, how many orders of what size will be placed in 90 days (assume no shortages)?
5. The yearly demand for a large cable bolt produced by Lanier Corporation is 9000. It costs
Lanier $1000 to put the production line for this cable into operation. The monthly cost of
carrying this item in inventory is $6000. Assuming that Lanier does not permit shortages to
occur, how many cable bolts should be produced per production run and how frequently
should production be stopped?
6. Bookworms Cooperative stocks a very popular gourmet cookbook for which there is a
demand of 1500 copies per month. The publisher pays postage, but the cooperative pays
$2.50 for each order in secretarial and administrative costs. Carrying cost is $0.50 per copy
per month. In order to minimize total inventory cost, how many books should be ordered
at one time? How frequently should orders be placed?

Answers to Odd-Numbered Problems

1. 160
3. a. 10 transistors 2} times per month
cost $20
b. accumulate transistors and stop -f times per month
cost % $13.33
5. 500 cable bolts and stop 18 times per year
 2.13 Indeterminate Forms / 289

2.13 Indeterminate Forms

In previous sections concerning limits, the evaluation of

lim f(x) when/(a) = ^or —

x~>a 0

is discussed for certain particular types of functions f(x). In this section more general
procedures for evaluating these and other indeterminate forms are discussed. These
procedures are based on L'HospitaFs rule:

I /(«) = g(a) = o
or

If Jf{a) = g(a) = 00
U
then r
lim f(x) rhm —/
—= f'(x)(
x-*a g(X) x-+a d(x)

and

r f\x)
lim —T-r exists
>a g'(x)

(this rule holds for finite or infinite a). If lim — is itself an indeterminate form ^ or
x->a g(x) 0
oo f'(x) f"(x)
—, then L’Hospital’s rule is applied again and lim = lim v and so forth.
oo ■a g'(x) g"(x)
CAUTIONS
i 1- f'(x) d (f(x)\ . . .
1. hm —~~ # hm — ——- , that is, the quotient of the derivatives is not in general
x-ag(x) X^adx \g(x)l
equal to the derivative of the quotient.

, r f'(x) f"(x) , , .ri. f'(x) . , . - . f 0 oo

2. hm —f = hm —~ if and only if lim —is the indeterminate form - or —,
x~.(T(x) ^g(x) 0 co’
that is, L’Hospital’s rule applies only to indeterminate forms; thus when the rule is
applied successively, the first determinate answer is correct and, in general, further
applications give incorrect answers.

EXAMPLE

Evaluate lim
y4Th-2
h-*0 h

0 4 + h- 2 r i(4 + h)~1/2 1
type -, so lim
h-> 0 h “ h-0 1 = 4

EXAMPLE

ex — (1 + x)
Evaluate lim
*-*•0

0 ex - (1 + x) O'
type - , so lim -5- lim ——— also type
U v —► o X x~* o 2x 0,

r ex 1
2 2
290 / Differential Calculus: Functions of One Variable

EXAMPLE

In x
Evaluate lim
X~> 00 X

oo .. In x
type —, so Inn lim !f* = 0
oo X y—► m A

EXAMPLE

x
Evaluate lim
X-* 00

1
00 X nx n-
, oo'
type —, so lim — = lim also type
oo x -* oo X-* oo 00;

n(n — l)x" 2 oo
= lim also type
X~* 00 00,

n
= • • • = lim ' = 0
X-00 ^

PROBLEMS
Evaluate the following limits.
?y sin2 x
1. lim 12. lim
oo y x-0 X

X sin x — x
2. lim 13. lim
x-+ oo x->0 X

In x
3. lim (k > 0) 14. lim
x X-* 00 In x
px _ p x
4. lim —— 15. lim
x-o sin x X~* 00 x

ex — In (x 4- 1) — 1 ex — e x — 2 sin x
5. lim 16. lim
x-0 x x->0 3x3

sec x + 1 e2* - 1
6. lim 17. lim
x-*n/2 tan x x-o x — sin x
x2 + x — 1 2 — 3e~x + e~2x
7. lim 18. lim
x-oo er + e x-0 2x"
ex + e~x — 2
8. lim 19. lim
x-0 X-+O0

x4 — 4x3 + 16
9. lim (k > 0) 20. lim
X-* 00 x-2 x3 - 8
In x 4 — 3ex — e~3x
10. lim 21. lim
! X — 1
x-*0 4x2

x3 — 4x
11. lim ln(l~ v) 22. lim
x-+i cot nx x-2 x2 — 2x
 2.13 Indeterminate Forms / 291

1 — cos t — \t
1 4.2
ex — 1
23. lim 33. lim ^2
t->o r x—0 X X

_3
r + 27
24. lim 34. lim -—
X-> 00 ex + x2 —2 x— — 3 x + 3

(sin h)/h — cos h v-2

25. lim 35. lim
h~>0 h2 JC-* 00 ex — 1

26. lim
X-* 00
4x2 + 3x — 6
8x + 2
36. lim COi
x_»o cot 2x
,V
4 y3 + 6y tan x — x
27. lim 37. lim-:-
o 3y3 + 2y x_o x — sin x

In sin x
28. lim 38. lim --
X-* 00 x-jt/2 (n ~ 2x)
sin z2 — sin2 z sin x — sin d>
29. lim 39. lim -—~
r-0 z x-<*> x ~ V

1 - sin 2x ey + sin y - 1
30. lim , t~a\- 40. lim —r—T.-r—
x-*/4 (n/4) - X y-o ln(l+y)
.1 1 + cos X ai v cot x
31. lim —-ry 41. lim —-
x-* (n - x) x-o In x

tan 6 — sin 6 In sin 2x

32. lim 42. lim -:-
0-0 e 2 x_»0 In sin x

Answers to Odd-Numbered Problems

i
1. 00 23. — J4
1
3. 0 25. 3
5. 1 27. 3
i
7. 0 29.
i
9. 0 31. i
11. 0 33. -,1
/
13. “61 35. 0
15. 00 37. 2
17. -2 39. COS 0
19. 00 41. — 00
3
21.

L’Hospital’s rule is also applicable to other types of indeterminate forms, if they

are first put into the form 0/0 or oo/oo, as follows.

Type oo • 0:

If limx^a f(x)g(x) = oo • 0, where limx^a f(x) = oo, limx_+a g(x) = 0, then

limx^a f(x)/[l/g(x)] is of type oo/oo or, alternatively, g(x)/[l/f(x)] is of type 0/0.

Type 0°:

If limx_+a f{x)9(x) = 0°, where limx^a f(x) = 0, lim^,, g(x) = 0, then

g(x) \ . 0
lim is of type -
x —a i
log/(x)
292 / Differential Calculus: Functions of One Variable

or

iiogf(x)\. co
lim ‘ is of type
x~>a 1 oo

9(x)
Type oo°:

If lim*^ f(x)g{x) = oo°, where limx^a f(x) = oo, limx_+a g(x) = 0, then

. g(x) \ . 0
lim /-— J is of type -
x-+a I t I O

log/(x)
or

. ,|og/(x)\. CO
lim I --—- I is of type —
x~+a 1 I oo
9(x)

EXAMPLE

Evaluate lim*-,*, x(el'x — 1).

Mx _ j
type oo • 0, so lim x(ei,x — 1) = lim
X~+ 00 X-* 00 l
X

= lim
X 00

EXAMPLE

Evaluate lim,-^ (1 — x) tan \nx.

Evaluate limx^0 (1 — 2x)3/x

 2.13 Indeterminate Forms / 293

In (1 - 2x)
type lx, so lim (1 — 2x)3/x = exp lim
x-*0 x-*0 x
3

In (1 — 2x)
lim = lim ————- = -6
x-*0 X x->0 1
3
lim (1 — 2x)3/x = e 6
x^O

note: Notation exp/(x) means ef(x).

EXAMPLE

Evaluate limx^0 (sin x + cos x)1/JC.

type 1ix , so ,•lim (sin

/ • x + cos mix ln (sin x +cos x)
x) ' = exp lim -
x-+0 x->0 X

cos x — sin x
ln (sin x + cos x) sin x + cos x 1
lim-- = lim ---= - = 1
x-0 x x-o 1 1

lim (sin x + cos x)1/x = e1 = e

x-0

EXAMPLE

Evaluate limx^0+

ln x
type 0°, so lim xx = exp lim
x-*0 x-*0 +

In x
lim —— = lim = 0
x->0 + x-0

lim x* = e° = 1
x —0 +

EXAMPLE

Evaluate lim*^ (1 — x)tan7tx

294 / Differential Calculus: Functions of One Variable

type 0°, so lim (1 — x)tanJC = exp lim

jc —► 1 1 COt 7ZX

-1
In (1 — x) 1 — x
lim — lim
cot nx esc nx

.. sin2 nx
— lim -—
x-1 X-l

= lim (2 sin nx cos nx)

x-1

= 0

lim (1 - x)tannx = e° = 1
X~>1

EXAMPLE

Evaluate lim*-,*, x1/x.

In x
type oo°, so lim x1/x = exp lim
*-► GO *“► 00

In x
lim lim 1/X = 0
X-» 00 X QQ X

lim x1,x = e° = 1
X-+ oo

EXAMPLE

Evaluate lim*-^ (x + ex)2,x

\
o ,• / v\7/v ,• In (* + ex)
type oo , so lim (x + e ) 1 = exp lim —---
*-► 00 X
2

1 +
In (x + ex) x + eJ
lim lim
X-+ 00 x *-►00 1
2 2

2(1 +
= lim
*-►00 x + ex

2ex
.1,8X
—
II

1 + ex

2ex
= lim - - 2
*-►00 e

lim (x + ex)2lx = e2
*-►00
 2.13 Indeterminate Forms / 295

Type oo — oo:

If \imx^a [f(x) - g(x)] = oo - oo, where lim^ f(x) = oo, limx^a g(x) = oo,
then

Q(x) f(x) 0
lim / 1 1 ^1= lim is of type
x~>a 1 1 x->a 1 1 0

\ /(*) d(x))f a(x) f(x) /

EXAMPLE

Evaluate limx_0 (cot x — -1.

1 x — tan x
type oo — oo, so lim I cot x — lim
x-0 XI x— 0 x tan x

1 — sec2 x
lim
x-0 tan x + x sec2 x

— 2 sec2 x tan x
lim
x—0 sec2 x + sec2 x + 2x sec2 x tan x

— 2 tan x
lim
x-0 2(1 + x tan x)

EXAMPLE

Evaluate lim (I—U.

x— 0 \x ex - l)

QX — 1 — X
type oo - oo, so lim (- lim
x->0 x-> 0 x(ex - 1)
ex — 1
lim
x-* 0 ex + xex — 1
ex 1
lim
x-»0 ex + ex + xex 2

The types of indeterminate forms which can be evaluated using L’Hospital’s rule
are summarized in Table 2.1.
296 / Differential Calculus: Functions of One Variable

Table 2.1 Summary of Application of L'Hospital's Rule

Type of indeterminate form Apply L’Hospital’s rule to

1. lim /(*) 0
lim /W
0

2. lim /(*) 00

g(x) 00

3. lim/(x)g(x) = oo • 0 lim or to lim

4. lim f(x)e(x) = 1“ lim or to lim

5. lim f(x)9(x) = 0° If L’Hospital’s rule applied to one

these gives a constant b for the limit,
6. lim f(x)9(x) =oo° then the original limit is eb

1 1
d(x) f(x)
7. lim [f(x) — g(x)] — oo — oo lim
1 1
g(x) f(x) I

PROBLEMS
Evaluate the following limits.

1. lim x1/(1_*2) 12. lim (sec x — tan x)

X~* 1 - x-tt/2

2. lim x cot x 13. lim (1 + x)llx

x-» 0 X-* 00

3. lim x In sin x 14. lim x" In x (n > 0)

x-> 0 x-+0

4. lim z1/z 15. lim (ex + x)1/jc

Z —*■ oo x-»0

5. lim tan x In sin x 16. lim (1 + ax)blx

X 00
x-*it/2

1
6. lim (1 + x2)1,x2 17. lim ( 1
0
x-* 1 + \x - 1 ^x -

7. lim x2 esc x cot x 1

x~>0 18. lim
In x
X
In xf
\
x-l

8. lim x In x
x-0
1
19. lim X \
x-l In x x — 1/
9. lim (1 + sin x)cotJC
x-0 20. lim (x — 1)<?
X-* 00

10. lim (1 + -\ 21. lim x2ex

x-oo \ x) X-*- — 00

11. lim xe~x 22. lim sin xtanx

X-* 00 X-*7t/2
 2.13 Indeterminate Forms / 297

23. lim (x + sin x)tan * 40. lim tan xC0SX

x-*0 + *-»(*/2)-
24. lim (1 + e~xY 41. lim sin atan 01
x~* co a — (7r/2) “

25. lim xe1/x 42. lim x2e~x

x-*0 + X~* oo

26. lim (x — T^Tx) 43. lim (1 + ct2)c/t

X-+ 00
t-0

27. lim (ex + 2x)1/x 44. lim sin a,an0!

a-*0 +
x~*0

28. lim x1/(1_x2) 45. lim x(1/x2)_1

x-1
x- i -

46. lim (ey - l)1/y

29. Bm y~* 00

x-*-0 ^
47. lim esc xsin x
x->0 +
V* + 1
30. lim
x —0 X 48. lim x2 esc x
x-> 0
31. lim esc 7ix log x
x -* 1 ,.71 71(f)
49. lim — tan —
</>-0 0 2
32. lim (x — 2)e~*2
x —* 00
50. lim (1 + x2)1/x
x-» oo
33. limx2e_3x
x~* oo
51. lim x sin -
34. lim sin x log x X~* 00
x-*0 +

52. lim (n — 2x) tan x

35. lim e~land sec2 9 x-*(rt/2)
0->(nl2)~
53. lim x2 In x
36. lim (x — 4)x2_ 16 x-0
x-4

54. lim (1 — tan 9) sec 26

i™ (y - y.)
37. lim xcot nx 6~*( n/4-)
X- 1

38. lim (x 4- l)cotx 55.

x-*0 +

39. lim (x2 + a)1/x 56. lim (V2 — x2 — l)x_1

x-i -

Answers to Odd-Numbered Problems

1. e 1/2 21. 0 39. 1

3. 0 23. 0 41. 1

5. 0 25. 0 43. 1

7. 1 27. c3 45. e1'

9. e 29. -1 47. 1

11. 0 1 712
31. -- 49.
n 2
13. 1
33. 0 51. £7
15. e2
35. 0 53. 0
17. 00
37. e1/n 55. 11
19. ~11
298 / Differential Calculus: Functions of One Variable

2.14 Sequences and Series

In this section, infinite sequences and series are defined and tests for determining
convergence or divergence of an infinite series are considered. Power series are
defined and the representation of a function by a Taylor’s series is discussed.
A sequence is a succession of terms, u{, u2, .formed according to some fixed
rule; a series is the indicated sum ul + u2 + of the terms of a sequence. A finite
sequence or a finite series has a specified (finite) number of terms; an infinite se¬
quence or an infinite series has an unspecified (infinite) number of terms. The general
term, or nth term, of a sequence or series indicates the rule of formation of the terms.
For example, 1, 3, \ is a finite sequence, and l+^ + ^ + i + ^isa finite series.
The general term is 1/n. However, if continued indefinitely, 1, -§, 3, i, 3,3,7,3,... is an
infinite sequence and l+i + 3 + i + 3 + 3 + 4 + 8 + ‘'- = 1 1A1 *s an infinite
series, still with the general term 1/n.
Another method of specifying the terms of a sequence or series is by a recursion
formula which gives the (n + l)st term as a function of the preceding term or terms.
For example, the infinite sequence 1, 2, 6, 24, 120,... may be specified by the general
term u„ = n! or by the recursion formula un+1 = nu„, where un is the nth term and w„+ x
is the (n -I- l)st term. Many sequences are more conveniently specified by a recursion
formula; other sequences, including the examples above, are as conveniently specified
by a general term, but additional terms are more easily obtained by recursion. In
order to specify a sequence completely by a recursion formula, the first term must be
specified separately. Thus the sum of the sequence above is given by ! un =
m! or by £®=1 un — 1 + nun-1- There are some sequences for which it
is necessary to specify the first term or terms separately, even if the general term is
specified.
Although a finite sequence must have a finite sum, the limit of the sum of an
infinite sequence may be finite or infinite.
Let Sn represent the sum of the first n terms of an infinite sequence ul9 u2,..., un,
• • • 5

S„ = ux + u2 + ■■■ + u„=
i= 1
and let S represent the limit of Sn as n -> 00,
00

S — lim Sn = Y, ui
n -* 00 i= 1

If the limit S exists finitely, the infinite series is convergent and is said to converge to
the value S; if the limit does not exist finitely, the infinite series is divergent. Diver¬
gence may occur because Sn becomes infinite as n -> 00 or because Sn oscillates
without approaching a limit as n ^ 00.
Infinite series are essential in calculating the values of many functions and can also
be used to define a number of useful functions. There are two problems generally
associated with infinite series: determining whether the limit of an infinite series
exists and determining the value of the limit, if it exists.
The problem of convergence is discussed before the use of infinite series in rep¬
resenting functions is considered. Establishing convergence or divergence of a series
is relatively straightforward if an expression for Sn can be obtained.

EXAMPLE

For the geometric series of n terms,

Sn = a -I- ar + ar2 + • • • + arn~1
 2.14 Sequences and Series / 299

it can be shown that

a(l — rn) a(rn — 1)

-
1 — r r - 1

The first form is generally used if I r I < l; the second form is generally used if Irl > 1.

a
If | r | <1, lim rn = 0 and lim Sn - and the series is convergent.
n-+ oo n~* oo 1 -r

If | r I >1, lim rn — oo and lim Sn = oo and the series is divergent.

n~* oo n~* oo

note: If r = — 1, the geometric series is

a — a + a — a + a — a + •••

and if n is odd, Sn = a; if n is even, Sn = 0. Such a series has no limit and is thus divergent; it is
referred to as an oscillating series.

Establishing convergence or divergence is clearly more difficult if an expression

for Sn is not known. In such cases, the following tests are used.

1. Necessary condition for convergence. If an infinite series ^°=1 un is convergent,

then lim^oo un = 0. That is, if the nth term of a series does not approach 0 as n
becomes infinite, the series is divergent. Note that lim„_+00 un = 0 is a necessary but
not a sufficient condition for convergence.
2. Alternating-series test. An alternating series is a series whose terms are alternately
positive and negative. Such a series is convergent if lim„^00 un = 0 and each term is
less in absolute value than the term which precedes it, that is, if | un+, | < \un\ for
all n = 1,2,....

note: It can be shown that the error involved in breaking a convergent alternat¬
ing series at any term does not exceed in absolute value the first of the terms
discarded. That is, if to — u2 + u3 — u4 -f • • • is a convergent alternating series, then
for any k,
00

< UL
n = k+ 1

3. Absolute convergence. A series of some positive and some negative terms is said to
be absolutely convergent if the series formed from it by making all its terms
positive is convergent. Other convergent series of some positive and some negative
terms are said to be conditionally convergent. If a series of some positive and some
negative terms is absolutely convergent, it is conditionally convergent, but the
converse is not necessarily true.
4. Cauchy's test-ratio test. Let

Uy +^2 + ^3 + "' + Un+ i + ' ' *

be an infinite series of positive terms. Using consecutive general terms un and un+1?
form the test ratio

Un+ 1

K
and let

un + 1
p = lim
n~* oo U„
300 / Differential Calculus: Functions of One Variable

Then
If p < 1, the series is convergent
If p > 1, the series is divergent
If p = 1, the test fails
5. Comparison tests. In many cases it is possible to determine whether a given series
is convergent or divergent by comparing it term by term with a series known to be
convergent or divergent. A series of positive terms is convergent if each of its terms
is equal to or less than the corresponding term of a known convergent series. A
series of positive terms is divergent if each of its terms is equal to or greater than
the corresponding term of a known divergent series. The geometric series dis¬
cussed above and the “p-series” are frequently useful in applying the comparison
test. The “p-series” is given by

1 1 1
1i + ^ + ^ + " * + ™ + ‘ * •
2p y nl
This series is convergent if p > 1 and divergent if p < 1. If p = 1, the p-series is the
harmonic series.
note: Since convergence or divergence of a series cannot be affected by the
omission of a finite number of terms, comparison tests may be applied to the terms
uk, uk+1, uk+2, ... rather than to the terms uuu2,u2,_

SUMMARY OF PROCEDURE FOR TESTING CONVERGENCE

OR DIVERGENCE OF AN INFINITE SERIES
In determining convergence or divergence of an infinite series, the tests should gen¬
erally be applied in the order in which they are given above; thus the procedure is:
1. Determine whether lim^^^ un is zero.
If lim un ^ 0, the series is divergent
n~* oo

If lim un = 0, further testing is necessary

n~* oo

Assuming lim,,^ un = 0,
2. Apply the alternating-series test (if appropriate): An alternating series whose
terms decrease in numerical value and for which lim,,^ un = 0 is convergent.

3. Apply the ratio test: If lim = p, then

n~* oo

p | < 1 => series convergent (absolutely)

| p | > 1 => series divergent
\p | = 1 => test fails
4. Apply the comparison test: Compare the series with a series known to be conver¬
gent or divergent, such as the geometric series or the p-series.
Geometric series: a + ar + ar2 + + arn +
Convergent if | < 1
Divergent if | r ■ 1

P-series: 1 + Y” + y + '" + i +
Convergent if p > 1
Divergent if p < 1
 2.14 Sequences and Series / 301

EXAMPLE

Determine the convergence or divergence of the infinite series

1 _ _3 5 _ 7
2 2? + 7? ~ 2* +

General term:

n+l2”-1
"n = (“I)
2"
lim un = 0
n -»> X

and | wn + 1 1 < | | for

/ 2(n + 1) — 1 \
>n + 1
P = lim -= lim
n -► x n-» x 2rc — 1
2 "

2n + 1
= lim ) ~ lim (~i) — _i
X ” 2(2n - 1) / n -► x

IP| = | —i| <1

so the series is absolutely convergent.

EXAMPLE

Determine the convergence or divergence of the infinite series

2 _ 3 , 4 5 ,
3 5 ' 7 9 '
General term:

n+l " + 1
K = (-1) 2n + 1
lim un — \
rt-+ x

so the series is divergent.

EXAMPLE

Determine the convergence or divergence of the infinite series

1 1 1
V3 + ys yi + "'
General term:
i
un (-ir1
sj2n — 1
lim un — 0
rt X
302 / Differential Calculus: Functions of One Variable

and |m„+1 I < \un | for all n, so the alternating series is convergent.

I 1
+1 y/2(n+ 1) —T
p = lim lim
n oo Ur n~* ao 1
x/2n - 1 !

(2n — l)1/2\ _
= lim
n~* oo (2n + 1 )112) '

P |=1

so the ratio test for absolute convergence fails.

l/(2n- l)1 12 > (1 /n) (p-series, p = 1), so the series is divergent as a positive series. Thus the
alternating series is conditionally convergent.

EXAMPLE

Determine the convergence or divergence of the infinite series

1 2! 3] 4-!
^ I ^2 I I I

General term:

n\
Un = gn

lim un = oo
n-^oo

so the series is divergent.

EXAMPLE

Determine the convergence or divergence of the infinite series

!+ 2(|)2 + 3(j)3 + 4(|)4 + ---

General term:

Un = n(if

lim un = 0
n-+ oo

3 + 1
un+ 1 (n+ 08)
= lim
p =
n~* oo Un n-* oo

%in + 1 3
= lim ! %
n~* oo \ "
I3 < 1
P =
so the series is convergent.
 2.14 Sequences and Series / 303

EXAMPLE

Determine the convergence or divergence of the infinite series

52 53 54
5 + 2!+ 3!+4!+'"

General term:

5”

lim u„ = 0
n oo

/ 5”+1
(n + 1)!
p = lim lim -- = 0
n~* oo 5" n-+ oo ^ + 1

\ n! /
|p| = o < 1
so the series is convergent.

PROBLEMS
For each of the following series, write out the first few terms and determine whether the series
is divergent or convergent (conditionally or absolutely for alternating series).

yfy^vVi
00 ft2
12 Y _"_
' „ = ! 10(2n - 1)
00 00
n 1/n
2- I 13.
1=1 n +1 n= 1
oo oo
oo in
n
3- „=i
I n+2 14 n= 1
00 1 oo 3ft — 1
71 + 1
15. I(-l)
4- 71£
= 1 T1 + In n n=0 4”
oo 1 00
71 + 1 3n2/3 - 1
5 y (_ iy»+i- 16. I ( — 1)
nW L) (2n + 1)! n=0 4"
oo oo
10
y (n + 1 )(n + 2) \n+ 1

6' ^ n!- n syit (5.-2,-

00 n2
00
is. £(-ir , (» - 3)!
171- 1
n= 1 + 71 = 0

00 (n + 3)!
00
00 n + 2
8. I 19. Z(-1)"+1^T
«=o 3!«!3" 71= 1 ft -T 1

00 -"JTl
00 n3
20. £
„=1 «:
00 1Q71 00 O 2 71 — 1

10. I 21- I
n= 1 n1 71= ! n +1

oo MI 00
22. I(-l)n + i i
11. V -
,4-, 9" 71= 1 ft
304 / Differential Calculus: Functions of One Variable

Answers to Odd-Numbered Problems

1. absolutely convergent 13. divergent
3. divergent 15. absolutely convergent
5. absolutely convergent 17. conditionally convergent
7. absolutely convergent 19. divergent
9. divergent 21. divergent
11. divergent

POWER SERIES
An infinite series of the form

a0 + a1x + a2x2 -f • • • + anxn + • • • = i>„v

n= 0

where the coefficients a0, au a2,... are independent of x, is called a power series in x.
More generally, an infinite series of the form
00

b0 + b{(x - a) + b2(x - a)2 + • • • -I- bn(x - a)n + • • • = £ “ aT

n=0

where the coefficients b0, bu b2, ... are independent of x, is called a power series in
x — a.
A power series in x (or in x — a) may converge for all values of x, or for no values
of x other than x = 0 (or x = a); or it may converge for some values of x and diverge
for others. If a power series converges for values of x in the interval — R<x<R,
then —R < x < R is the interval of convergence of the power series and R is the radius
of convergence. The interval of convergence is determined by the following procedure,
derived from the Cauchy test-ratio test.
Power series in x:
L = 0=> series converges for all x
L f 0 => series converges for the interval
If lim a"+1 = L, then \ — 1/1L | < x < 1/1L | and diverges outside this inter¬
n~* oo
val; the end points of the interval of convergence
must be examined separately.
Power series in x — a:
M = 0 => series converges for all x

M f 0 => series converges for the interval

’n + 1
If lim = M, then \a — (l/\M \ ) < x < a + {1/ \M ) and diverges out-
00
side this interval; the end points of the interval of
convergence must be examined separately.

EXAMPLE

Find the interval of convergence for the power series

1 + 2x + 3x2 -I- 4x3 + • • •

General term:

un = nxn~ * 1

n + 1
LT = Vlim - = 1,
1

n-* ao n
so the interval of convergence is — 1 < x < 1 and the end points must be tested.
 2.14 Sequences and Series / 305

If x = — 1, the series is 1 — 2-h3 — 4-h-**

u„ = (-\y-ln

lim u„ ^ 0
n -*■ oo

so the series is divergent.

If x = 1, the series is 1 + 2 + 3 + 4 + -- -

un = n

lim u„ =f= 0
n-> oo

so the series is divergent.

Thus the original power series converges for — 1 < x < 1.

EXAMPLE

Find the interval of convergence for the power series

4 x x x3 x4
1 - - + +
2 ' 3 • 22 5 • 23 ' 7 • 24

General term:

(-l)”+1x"_1
un 1
(2n - 3)2 n-

2n - _ i
L = lim I — 2,2n ~1 = lim I — - —i
[2(n + 1) - 3]2", n-> oo An — 21

so the interval of convergence is — 2 < x < 2 and the end points must be tested.
If x = — 2, the series is

2 22 23 24
1 + ~ + —-vy + -+ —-_-j + * * '
2 3-2 5•2 7-2

or

I + I+ 3 + 3 + 7 +

U„ =- (omitting the first term)

" 2n — 1

lim u„ = 0
n-+ oo

1
2n - 1
lim “^=lim — + U - 1 = lim = 1
n~* oo Mn n~* oo 1 n -*■ oo In + T
2n — 1

so the ratio test fails.

un = >— (p-series, p = 1)
2n — 1 2n

so the series is divergent.

note: Multiplying a series by a constant does not affect convergence.

306 / Differential Calculus: Functions of One Variable

If x = 2, the series is

2 22 23 24
* — 2 + 3 . 22 _ 5"* 23 + 1 • 24

or

i-i+4

«»= (-i)1 (omitting first term)

lim un = 0
n -*■ oo

| Un + 1 | ^ | Un |
so the alternating series is convergent.
Thus the original power series converges for — 2 < x < 2.

EXAMPLE

Find the interval of convergence for the power series

l+x + 2!x2 + 3!x3 + -- -

General term:

un = (n — 1)! x"_ 1

n\
L= lim = *lim n = oo
n~> oo 1) • n ■ oo

Thus the power series converges only for x = 0, that is, the series is divergent for all x =/= 0.

EXAMPLE

Find the interval of convergence for the power series

(x - 1) - i(x - l)2 + i(x - l)3 - i(x - l)4 + • • •

General term:

n+1 (X ~ 1)"
“n = (-!)'
n

n
M = lim = -1
n-* oo n + 1

so the interval of convergence is 0 < x < 2 and the end points must be tested.
If x = 0, the series is — 1— \ ^ \ — •••.

l
un
n

lim un = 0
n~* oo

u n+ 1
lim = 1
U,
n-* oo
 2.14 Sequences and Series / 307

so the ratio test fails.

un = - (p-series, p = 1)
n

so the series is divergent.

If x = 2, the series is 1— i + i — * + •••.

u„ =
n

lim un = 0
n-* oo

| +1 | \U„ |
so the alternating series is convergent.
Thus the original power series converges for 0 < x < 2.

EXAMPLE

Find the interval of convergence for the power series

(x - 1) - i(x - l)2 + 4(x - i)3 - M* - !)“ + • ■ ■

General term:

■ +i (x - 0"
Un = (-1)
n

n
M = lim I — = -1
n-+ oo (n + 1)-

so the interval of convergence is 0 < x < 2 and the end points must be tested.
If x = 0, the series is — 1 — -J — i — rs — • • •.

1
Un j
n

lim un = 0
n~* oo

lim —- = 1
n~* oo Mn

so the ratio test fails.

| un | = 2 (p-series, p — 2)

so the series is convergent.

If x = 2, the series is 1 — \ + \ ~ T5 +

K = (- 1)”+1 -2
n

so the series is absolutely convergent (by procedure above).

Thus the original power series converges for 0 < x < 2.
308 / Differential Calculus: Functions of One Variable

EXAMPLE

Find the interval of convergence for the power series

2(x_2) + 3(^ + 4(-2)! + 5(x-^ + ..

2! 3! 4!

General term:

(n + l)(x - 2)n
un =
n\

I n + 2
{n + 1)! r n + 2
M = lim = lim ^ = 0
n-> oo n + 1 n-> oo (»+ ir
\ n! /
Thus the power series converges for all x.

TAYLOR S THEOREM
For either theoretical or computational purposes, it is frequently convenient to
represent a function of x by a power series; Taylor’s theorem provides appropriate
power series for representing many functions.

taylor’s theorem : The infinite series

* f(k){a) k x -
— a (x - a)2
I (x-af=f(a)+f’(a) — + f"(a) 2, ' + • • •

(x — a)n~ 1
+ Rn
(n - 1)!

converges and represents the function /(x) for those values of x for which all the
derivatives off(x) exist and for which Rn 0 as n -► oo. In this case/(x) is said to be
expanded in a Taylor s series about x = a. For the special case a = 0, the expansion is
a Maclaurins series.
Rn is called the remainder after n terms and it can be shown that

(x — a)n
K=f{n)(i) where a < £ < x
n\

This formula can be used to determine a bound for the error involved in using only the
first n terms of the series if maxa<^<x Rn is obtained.

note: There are some functions for which Taylor’s series converges for values of
x for which the remainder does not approach 0 as n -* oo; for such values of x, the
series does not represent the function. However, in most cases the interval of conver¬
gence of the series is the same as the interval for which Rn -+ 0 as n -> oo, as is the case
for the examples considered here.

A Taylor’s series about x = a is useful for calculating the function it represents for
values of x near a. Similarly, a Maclaurin series is useful for calculating the function it
represents for values of x near zero.
A rigorous proof of Taylor’s theorem is not simple and will not be given. However,
the result can be made plausible by the argument presented in Technical Note III.
 2.14 Sequences and Series / 309

It should be noted that any differentiable function can be expanded in a Taylor’s

series: A function that can be differentiated only a finite number of times can be
expanded in a series having a finite number of terms; a function that can be differen¬
tiated indefinitely can be expanded in an infinite series. However, to be a valid
representation of a function the series must converge for values of x in the range of
interest; in addition, to be useful for computational purposes, the series must con¬
verge fairly rapidly, so a reasonably accurate approximation can be obtained using a
manageable number of terms. Thus, before using a Taylor’s series expansion to
represent a function, its convergence properties should be investigated.

EXAMPLE

Expand the function/(x) = x 1/3 in powers of (x — 1) and determine the interval of conver¬
gence of the series.

f(x) = X"1'3

f'(x) = -ix^'3

/'(*)=■£*-7,3

1-4-7 10/3
/"'(*)

1-4-7-10 13/3
/(,v’(x) .-

34 '

1-4-7-10-13 16/3
/(V) (x) = -

/■(.c) = (-ir(3w ~ 2)(3r_ ~5) "•1 x-<3-3>^

3"

/(x) =f{a) +
/»
v (x - a) +
/» (x - a)2 + • • • + Pn\a) (x — a)n +
2! n\

= 1 - 3^! (x - 1) + jT.yj (x “ *)2 “ y^3l <x - O3 + •

. (3h - 2)(3n - 5) ••• 1

+ (-l) (x - 1)” +
3”n!

(3« + 1)(3n - 2) 1
bn+1 _ 3"+1(n + 1)! _ 3 n+\
(3n - 2)(3n - 5) ••• 1 ~ 3(n + 1)
~ 3 "n!

lim„_ qo bn+1 /bn = — 1, so the interval of convergence is 0 < x < 2 and the end points must be
tested.
If x = 0,

(3n - 2)(3n - 5) 1
Un ~ 3^!

lim un= 1^0

n-> oo

so the series is divergent.

310 / Differential Calculus: Functions of One Variable

note: For notational consistency, the general term is denoted un; actually it is the term
involving the nth derivative and is the (n + l)st term of the series.

If x = 2,

„ (3n — 2)(3n — 5) • • • 1
un = (-1)
3 Hn\

lim un = 1 =/= 0

so the series is divergent.

Thus the power series expansion converges for 0 < x < 2.

EXAMPLE

Expand the function f(x) — 1/(1 + x)2 in powers of (x — 2) and determine the interval of
convergence of the series.

f(x) = (1 + x)~2

/'(*)- -2(1 + x)-3

f"(x) = 2-3(1 + x)“4

f"'(x)= — 2-3-4( 1 + x)-5

/(,v)(x) = 2-3-4- 5( 1 + x)"6

/,V|(x) = — 2-3-4-5-6(l + x)—7

fn(x) = ( — !)"(« + 1)!(1 + xy{n+2)

f(x) = 1! 2! nl

^2 - js(x - 2) + - 2Y - js(x - 2)3 + ■■■ + (-l)"^/(x - 2)" +

n + 2
bn+l = 3^ n + 2
bn n + 1 3(n + 1)
3n + 2

lim„_ bn + l jbn =
oo so the interval of convergence is — 1 < x < 5 and the end points must
be tested.

If x = — 1, un = (n + l)/32.

lim un ~ go 7^ 0,
n~* 00

so the series is divergent.

If x = 5, un = (— l)”(n + l)/32.

lim un = 00 ^ 0,

so the series is divergent.

Thus the power series expansion converges for — 1 < x < 5.
 2.14 Sequences and Series / 311

EXAMPLE

Expand the function /(x) — (1 + xf, where m is any real number, in powers of x and deter¬
mine the interval of convergence of the series. This is known as the binomial expansion.

f(x) = (1 + xf

f'(x) = m( 1 + xf-1

/"(x) = m(m - 1)(1 + x)m~ 2

f"\x) = m(m — 1 )(m — 2)(1 + xf “3

/(1V)(x) = m(m — 1 )(m — 2){m — 3)(1 + x)m~ 4

/<V)(x) = m(m — 1 )(m — 2)(m — 3)(m — 4)(1 4- xf “ 5

ml
/"(x) = (1 + x)m-"
(m — n)\

f(x) = /(a) + /4L)x + /34x2 + ... + ffl;c» + ..

1! 2! n\

m m(m — 1) , m(m — 1 )(m — 2) ,

= 11-1+-Y 4-V--x2
X i- X 4-I--'- J -'
3 Ll 4.
^

m(m — l)(m — 2) • (m — n + 1) „
+ —-+
n!

m(m — 1 )(m — 2) • • • (m — n)
&n+i (n + 1)! m — n
m(m — 1 )(m — 2) • • • (m — n + 1) n + 1
n\

lim,,^ bn+l/bn= — 1, so the interval of convergence is — 1 < x < 1 and the end points must
be tested.

If x = — 1, un 1\ ivml(m — 1) (m — 2) (m — m -h 1)
V
n\

lim u„ = 0
oo

m(m — l)(m — 2) • • • (m — n + 1)
If X = 1, un
n\

lim u„ = 0
00

If x= ±1, \un+l/un\ = |(m —«)/(«+ 1)|, so |mb+1| < || does not hold for unre¬
stricted m.
Thus the power series expansion converges for — 1 < x < 1.

EXAMPLE

Expand the function/(x) = In x in a Taylor’s series about x = 1 and determine the interval of
convergence; use this series to evaluate In 1.04 correct to five decimal places.
312 / Differential Calculus: Functions of One Variable

f(x) = \n x

/'M = l-

/"(x) = -1

24
/,v,(x)

„ + i (« ~ !)!
/"(*) = (-!)'

fin~
In x =f(a) + (x “ fl) + (* ~ a f + •• + 4-A. (x - a)”“' + R„
1! 2! (« - 1)!
1
= 0 + (x - 1) -i(x - l)2 + i(x - l)3 -••'+(- 1)"|
n — 1
'(x-ir1 + *„
i
frn + i = _ n _ _ n - 1
bn 1 n
n — 1

bn + l/bn= — 1, so the interval of convergence is 0 < x < 2 and the end points must be
tested.

If x = 0, un =-—t
ft — 1

lim un = 0
n~* oo

u„ + i •• n ft — 1
lim lim = lim = 1
n~* oo n~*oo 1 n-*ao ^

ft — 1

so the ratio test fails. But

1
un (p-series, p = 1),
ft — i

so the series is divergent.

note:Omitting a finite number of terms from a series does not affect convergence, nor
does multiplying the series by a positive or negative constant.

If x = 2, u„ = (— 1)" + ‘(l/n).

lim u„ — 0
n~+ oo

| +1 | ^ | |

so the series is conditionally convergent.

Thus the power series expansion converges for 0 < x < 2.
 2.14 Sequences and Series / 313

In (1.04) = 0 + 0.04 - i(0.04)2 + i(0.04)3 + RA

= 0.04 0.0008 + 0.000021 + R4

= 0.039221 + /?4

(x — a)n
a < £ < x
nl

1 £ ^ £ 1.04
£4 4!

(0.04)4
4£4

4 I < i(0.04)4 = 0.00000064

so In 1.04 = 0.03922, correct to five decimal places.

note: Since the series is alternating, the error in truncating is less in absolute value than
the next term, which for this example gives the same bound on the error as above.

EXAMPLE

Expand the function f(x) = ex in a Maclaurin’s series about x = 0 and determine the interval
of convergence; use this series to evaluate el/2 correct to four decimal places.

f(x) = ex

f'(x) = e*

/"(*) = e*

fn(x) = e>

/» „, /» r-"( a)
—Y\~x + 2! x2 4- ••• + x"~1 + R„
(n — 1 )!
2 3 ,n- 1
X X

l + x + 'v. + V. + '' + (^T[y. + K

bn +1 nl (n — 1)! 1
1 n n
(n - 1)!

lim bT1 = °
n-+ oo

Thus the power series expansion converges for all x.

,1/2 _
— 1 4* 7 +
(W ^ (W , (i)4 , (i)
+ + + + R,
2! 3! 4! 5!

= 1 + 0.5 + 0.125 + 0.02083333 + 0.00260417 + 0.000260417 + R6

= 1.648697917 + Re
314 / Differential Calculus: Functions of One Variable

a<^<x
n\

0£ t£i

and, since ei/2 <2, \R6 I < 0.000043, so e1/2 = 1.6487, correct to four decimal places.

EXAMPLE

Expand the function f(x) = (1 4 x)1/2 in a Maclaurin’s series about x = 0 and determine the
interval of convergence; use this series to evaluate ^/l.lO correct to six decimal places.

/(x) = (1 + x)1/2

/'(x) = i(l + x)-^2

/"(*)- -^ (! + *)'3/2

f"(x) = ~( 1 + x)-5'2

/,,V) W=-E^2(i + *)-7'2

1-3-5-7

(2rc — 3)(2w — 5) • • • 1 l)/2

/“(*) = (-1) n+ 1 (1 + x)
-(In-
2"

(1 + x^=/(a) + /^,_/^x2 + ... + /^)x»-. +

1! 2! (n —vl)!

1 -1 1 2 3 • 1 ,
= 14- ^—. . x - xi—x 4- o—— xJ - • • •
2 1 • 2 ! 2 21 • 3!
• !

(In - 5)(2n - 1\)•• 1

+ (-!)' + Rn
2n~x\[n - 1) t

(2 n — 3)(2 n — 5)(2 n — !)■•• 1

bn+ 1 2nnl 2n — 3
bn (2n - 5)(2n - 7) ••• 1 2n
2"~1(n - 1)!

lim„^ * bn+1/bn= — 1, so the interval of convergence is - 1 < x < 1 and the end points must
be tested.
If x = — 1 ,/(x) = 0, so expansion is unnecessary.

(2n - 5)(2n - T>- 1

If x = 1, un = (— 1)"
2n~1 (n 1) !

lim u„ = 0
n-+ oo

Un + 1 2n — 3
< 1
2n
 2.14 Sequences and Series / 315

so the alternating series is conditionally convergent.

Thus the power series expansion converges for — 1 < x < 1.

3 • 1 5-3-1
(1.10)1/2 = 1 + Wo)3-
2! (A)2 + >3 • 3! 24 • 4! (M
(to) -
2 • 1!

7 • 5 • 3 • 1 . . „
4-25 . 5,— (tt) + Re

= 1 + 0.05 - 0.00125 + 0.0000625 - 0.000003906 + 0.000000273 + R6

= 1.048808867 + R6

Alternating series, so the error is less than the absolute value of next term,

9 ' 1 (to)6 = 0.0000000205

so correct to six decimal places, (1.10)1/2 = 1.048809.

note: Alternatively, a bound on the error can be obtained using R6:

a < i < x
n!

9 • 7 • 5 • 3 • 1
0 < £ < 175
R*=-26 • 6! (1 + «'n/W

975-3*1
R6 <
26 • 6! (A )6

R6 I < 0.0000000205

Note that the accuracy of the approximation obtained from a power series expan¬
sion increases with the number of terms used in the calculation. Fewer terms are
required for a specified accuracy if the series converges more rapidly. The number of
terms required for a given accuracy must be determined for each power series
approximation by considering the remainder term or, for an alternating series, by
considering the next term. Note that the remainder after n terms

n!

is equal in absolute value to the next term of an alternating series if £ = a. Thus, for
an alternating series, the bound on the error determined by obtaining maxa<^<x Rn is
equal to the bound determined by obtaining the absolute value of the next term of
the alternating series, if maxa<^<x Rn occurs for £ = a.

PROBLEMS
For each of the following power series, determine the general term and the interval of
convergence.

X3 X5 X7
*■ X “ 3! + 5! “ 7! +

2. 2x + 4x2 + 8x3 + 16x4 + • • •

2 3
X X X
3. 1 + + +
V/2 + 73Ty4
2 3 4
X X X X
4. . _ + + O-+
1-2 3^ 4"
316 / Differential Calculus: Functions of One Variable

, , 3x 5x2 7x3 9x4

51-Y + ^--T + l6-"
(x + 3) (x + 3)2 (x + 3)3
6. 1 + -,2 -- + ,2 + .2 + •"
22 32 42

„ 1 2 , 3 5 4 7
7' 2!X + 3!X +4!X + 5!X +"'

„ x + 4 '(x + 4)2 (x + 4)3 (x + 4)4

s/2 y3 + V4 ^5 +"'
„ 1 22 , 32 3 42 . 52 5
9- 2x~ Tx + 8X ~7TX +14x + "

10. ^ (x - 5) + y (x - 5)2 + ~ (x - 5)3 + ^ (x - 5)4 + ^ (x - 5)5 + • • •

Expand the following functions in power series, determine the interval of convergence, and
choose some appropriate value of x for which to estimate the function.

11. f(x) = x^'2 in powers of (x — 1) 17. /(x) = sin x in powers of x

1 18. /(x) = cos x in powers of x

12. f(x) = in powers of x
1 + x2
19. /(x) = tan x in powers of x
13. /(x) = e~(ll2)x in powers of (x — 2)
20. /(x) = cot x in powers of x
14. /(x) — £(<ex + e~x) in powers of x
21. /(x) = sec x in powers of x
15. /(x) = In (a — x) in powers of x
22. /(x) = esc x in powers of x
16. /(x) = In (a 4- x) in powers of x

Answers to Odd-Numbered Problems

,2 n- 1
n+ 1
i. (-1) , conv. all x
{In - 1)!
1
3. , conv. — 1 < x < 1
n

c t i 1 (2w — l)x” 1 . _
5. (-1)" ----, conv. —2 < x < 2

n
7. x2n l, conv. all x
(n + 1)!

n
9. {— 1J"+1 —:—- xn, conv. — 1 < x < 1
3n — l

11- /M = 1 + 2 1! (* - 1) -^7 (* - v
22 • 2!v" l)2 +' 23 • 3! (* - l)3
1-3-5
(x - l)4 + * .5 . ~ 7 (x - l)5 +
24 • 4! v" ' 25 • 5!

»+i (2n - 3)(2" - 5) ••• 1

+ (-l) (x — !)” + •••, conv. 0 < x < 2
2" • n!

13- f(x) = 1 — (x — 2) 4- (x — 2)2 — ‘ 1 (x - 2):

^3
2 1! 2 • 2! 23 • 3!
e 1 , e 1
+ (x - 2)4 - (x - 2)5 + • ■ •
24 • 4!v 7 25 • 5!
1
+ (-!)" 77 (x — 2)" + • • •, conv. all x
2" • n!
 2.14 Sequences and Series / 317

15. f(x) = In a — - X — 4, x2 — — „ X3-— x4 — • • conv. —a<x<a

a 2a‘ 3a' 4a na
3 5 Y2n — 1

i7./w = x-3[+---+... + (-ir*p^+ conv. all x

x3 17x7
2x5 i i 71
19. f(x) —
4~ 4" - 4" * * * 9 conv. x <-
3 15 315 1 1 2
x2 5x4 61x6 i i n
21. f(x) —
4 _ —- + •••, conv. x <—
2 24 720 1 1 2

Technical Note I Derivation of Rules for Differentiation

note: For mathematical convenience, the rules for differentiation are not
derived in the same order as they are given in the text.

Algebraic Functions
Rule 1. If y = c,
Ay = 0

A^o
Ax
dy Ay
= lim ~ = 0
dx “‘“o Ax
Rule 3. If y = cu, where u =f(x) is a differentiable function of x,
y + Ay = c(u + A u)
Ay = c Au
Ay Au
Ax Ax
dy Av du
T= lim ~y- = cyr
dx dx
Rule 2. If y = xn, where n is a positive integer,
y + Ay = (x + Ax)”

= x” + nxn~ 1 Ax + —■———- x”“ 2(Ax)2 + • • • + (Ax)”

Ay = nxn~ 1 Ax -f ^ ^ xn~ 2(Ax)2 + • • • + (Ax)”

„-i , n(n -1Lx”„-22(Ax) + • • • + (Ax)” 1

Ax HX + 2!
dy Ay
= lim —^ = nx n- 1

dx Ax-* 0 Ax
Rule 4. If y = u + v, where u =f(x) and v = g(x) are differentiable functions of

y + Ay = u -f Au + v + Av
Ay = Am + Av
Ay Au Av
Ax Ax Ax
dy Ay du dv
dx Ax-o Ax dx dx
318 / Differential Calculus: Functions of One Variable

Rule 5. If v = ny, where u = f(x) and v = g{x) are differentiable functions of x,

y + Ay = (u + A n)(y + Ay)

= uv + u(Av) 4- v(Au) 4- (An)(Ay)

Ay = u(Av) 4- v(Au) 4- (Au)(Av)

Ay Av Au (Au)(Av)
— u —1_ v -|--

Ax Ax Ax Ax
dy Ay dv du
= lim —— = u ——F v
dx Ajc->0 Ax dx dx

Rule 6. If y = n/y, where u = f(x) and v — g(x) are differentiable functions of x,

u 4- Au
y 4- Ay
y 4- Ay
v(u 4- Au) — u(v 4- Ay)
Ay =
v(v 4- Ay)
y(A u) — u( Av)
v(v 4- Ay)
An Ay
y-— — u
Ay Ax Ax
Ax y(y 4- Ay)
du dv
Vdx U dx
— = lim —
dx 4*^0 ?

Logarithmic Functions
Rule 8. If y -= loga n, where u =f(x) is a differentiable function of x,

y + Ay = loga (n 4- An)
Ay = loga (n + An) - Iogfl n
n 4- An^
= lOga
n
An^
= log,, 1 + n
Ay 1 . L A4
loga 1 +
An Au n
1 n An'
!og„ 1 +
n An u
1 A u\u,Au
= - lOga 1 +
n \ u

Let A u/u = h; then u/Au = 1/h and h —>0 as An —► 0. Thus Ay/An —

(1/n) loga (1 4- h)1,h and, since (by definition) (1 4- h)i,h -> e as h -> 0,

Ay 1
lim / = -logfl lim (1 4 h)],h
Au-^o An n h->0

dy
- loga e
du u
 2.14 Sequences and Series / 319

If u =/(x),

dy 1 , du
- = - log„ e ,
dx u dx

Exponential Functions
Rule 9. If y = au, where u =f(x) is a differentiable function of x,

In y = In au

= u In a

1 dy du
— In a
y dx dx

dy du .
y — In a
dx dx

du
= au In a
dx

Rule 10. If y = uv, where u =f(x) and v = g(x) are differentiable functions of x.

In y = v In u

1 dy 1 du dv
-f In u
y dx u dx dx

dy v- 1
du dv
vu + uv In u
dx dx dx

Rule 7. If y = un, where u = f(x) is a differentiable function of x and n is any real

number,

In y — n In u

1 dy 1 du
y dx u dx

dy du
nu n- 1

dx dx

Trigonometric Functions
Rule 11. If y = sin u, where u = f(x) is a differentiable function of x,

y + Ay = sin (u + Au)

y = sin (u + Au) — sin u

/ Au\ . Au
= 2 cos I u + — \ sin --

since sin x — sin y = 2 cos y) sin i(x -y)

Ay / Au\ sin (Au/2)

— = 2 cos I u + —
Au \ 2 Au

Au\ sin (Au/2)

= cos u -\-
Au/2
320 / Differential Calculus: Functions of One Variable

dy Ay Au\ sin (Aw/2)

lim lim cos I w +
du Au-0 ^W Au->0 T/ Aw/2
sin (Aw/2)
= cos w since lim = 1
Au-> 0 Aw/2

If w = /(x),
dy du
— = cos u —
dx dx

Rule 12. If y = cos w, where w = /(x) is a differentiable function of x,

= sin ^ - w

and

dy
= cos
dx

du
= — sin w
dx

Rule 13. If y = tan w, where u =f(x)

sin w
y =-
cos w

and

7 du . y du
, cos w — + sin w —
wy dx dx
dx cos2 w

but cos2 w + sin2 w = 1, so

du
dy dx du
sec2
dx cos2 w dx

Inverse Functions
Rule 14. If y = f(x) and x = g(y) are inverse differentiable functions,

g(y + Ay) - g(y) = h_ where x = g(y)

Ay Ay x + h = g(y + Ay)

y + Ay = f(x + h) since f[g(x)] = x

and thus
Ay = f(x + h) -f(x)

g(y + Ay) - g(y) =_Ax_

Ay f(x + h)-f(x)

As Ay ->• 0, Ax ->• 0 and

lim y(>’ + Ay) - g(y) Ax

lim
iv-o Ay Ax-> 0 f(x + h) -f \x)
 2.14 Sequences and Series / 321

That is,

dx 1
dy dy
dx

Composite Functions
Rule 15. If y = <p(u) and u = F(x)

Ay Ay Au
Ax Au Ax
Ay dy du
lim 3T = 1~'
du
~T
uX

Technical Note II Conditions for a Cubic Cost Function

h2 — 3ac < 0 is a sufficient but not a necessary condition for

y = ax3 + bx2 + cx + d

to have no relative maximum or minimum in the first quadrant. If b2 — 3ac > 0, then

— = 0 if x = (— b ± Jb2 — 3ac)/3a.
dx
— b + Jb2 — 3ac
x =-->0 since a > 0, b < 0 and b2 — 3ac I < I b
3a

(note that b2 — 3ac > 0 => b f= 0).

/ —b ± Jb2 — 3ac — b± yjb2 — 3ac\ — b ± y/'b2 — 3ac'
+ bl + c + d
y = -3a- 3a ) 3a

Since a>0, h<0, c>0, ^>0 and (— b ± Jb2 — 3ac)/3a > 0, the first term in the
expression for y is positive, the second term is negative, and the third and fourth
terms are positive or zero. Thus y may be positive or negative depending on the
relative value of the coefficients a, b, c, and d, and (x, y) may lie in either the first or
second quadrant. Since

d2y
2( — b ± yjb2 — 3ac + b) = ± 2yjb2 — 3ac
dxz x = [ — b ± fb2 — 3ac]/3a

and b2 — 3ac > 0,

d2y
7^0
dx2 x = [ — b±fb2 — 3ac]/3a

and y = ax3 + bx2 + cx + d has a relative maximum or minimum at the point for

which ~ = 0. Thus, if b2 — 3ac > 0, y = ax3 + bx2 + cx + d has a relative maxi-

dx
mum or minimum in the first quadrant only for certain values of the coefficients. For
simplicity, only equations for which b2 — 3ac < 0 are considered in the text.

Technical Note III Taylor's Theorem

Assume that the function f(x) has a power series expansion and that the power series
can be differentiated term by term.

f(x) = a0 + afx — a) + a2(x — a)2 + ■ ■ • + a„(x — a)n + • • •

322 / Differential Calculus: Functions of One Variable

If x = a,f(a) = a0.

/'(x) = ax + 2a2(x — a) + • • • + nan(x — a)n~ 1 + • • •

If x = a,f'(a) = av

f"(x) = 2a2 + 3 • 2a3(x — a) + • • • + n(n — 1 )an(x — a)n~ 2 + • • •

If x = a,f"(a) = 2a2 .

f"(x) — 3 • 2a3 + 4*3* 2aA(x — a) 4- • • • + n(n — 1 )(n — 2)an(x — a)n~ 3 + • • •

If x - a, f"'(a) = 3 • 2a3.

fk(x) = k(k — 1 )ak + (k + l)(k)(k — 1 )ak+1(x — a) 4- • • •

4- n(n — 1) • • • (n — k + l)u„(x — a)n~k + •

If x = a, fk(a) = k\ak.
Thus

an = fn(a)
— ffor n = 1, 2, rt \
a0 =f(a)
n!

f(x) = f(a) + (x-a) + (x - a)2 + • ■ • + (x _ af + ...

J! l! n!
Differential
Calculus:
Functions of
More Than One
Variable

3.1 Introduction
The discussion and examples in previous sections were concerned with functions of
one variable, that is, functions of the explicit form y =f(x) or the implicit form
/(x, y) = 0. Such functions express a relationship between two variables, x and y, and
implicitly assume that the processes being studied can be represented adequately in
terms of only these two variables. Although in many cases this type of representation
provides a reasonably accurate first approximation to reality, there are clearly many
cases in which such a representation is so inadequate as to be virtually useless. In
these cases it is necessary to express a relationship in terms of several variables or to
express one variable as a function of more than one other variable. For example, in
economic theory supply and demand of a commodity frequently depend not only on
its price but also on the prices of related commodities, on income level, on time, and
on various other factors. Profit depends not just on the output of one commodity,
but on the output levels for several commodities and on the combination of several
inputs.
Relations and functions of several variables can be defined by appropriate exten¬
sions of the corresponding definitions for two variables. Geometric representation is
not possible for functions involving more than three variables, but, with this excep¬
tion, the following discussion and definitions are appropriate for any finite number of
variables. Partial differentiation is defined and several applications of partial deriva¬
tives in business and economics are discussed. Constrained and unconstrained
maxima and minima of functions of two variables are considered and constrained
utility maximization is discussed as an application of constrained maximization and
minimization in business and economics.

FUNCTIONS OF MORE THAN ONE VARIABLE

Just as a point in two-dimensional space is represented by an ordered pair of real
numbers, a point in three-dimensional space is represented by an ordered triple of
real numbers and a point in n-dimensional space is represented by an ordered n-tuple
of real numbers (xj, x2, .x„). Thus a point in rc-dimensional space is an ordered
arrangement of n real numbers. The set of all points in n-dimensional space is
323
324 / Differential Calculus: Functions of More Than One Variable

denoted by Un. For convenience, a point in rc-dimensional space is sometimes

denoted by a single letter, say X.
A set of ordered pairs {(X, z)}, where X e Un, is a relation. If (Xl, z{) e F and
(Xj, z2) e F implies that zl = z2, then {(X, z)} is a function of n variables. The
domain of the function is the set of points {X} in the space Un and the range is the set
of values {z} appearing in the set of pairs in F. Note that the set of ordered pairs is a
relation and, if to each point of the domain there corresponds only one point of the
range, the relation is a function.
When X e U2, the pair (X, z) is an ordered triple, say (x, y, z), which may be
interpreted as a point in three-dimensional space. Functions whose domain is Uk9
where k> 2, cannot be represented geometrically.
A function of several variables is frequently described by an equation expressing
the correspondence between the coordinates of the point X and the functional value
z. For this purpose, the functional notation y =/(x) can be extended for functions
defined on Un. For example, if X e Un, z = F(X) indicates that F is a function of n
variables. If the domain is in U2,
z = F(x, v) or z = F(X)
and z is a function of two variables. In general, if X e Un,
z = F(xx, x2 , ..., xn) or z = F(X)
and z is a function of n variables. The coordinates of X are called independent
variables and z is called a dependent variable. Implicit functions of several variables
can be written, for example, as/(x, y, z) = 0 or, more generally, as/(xl9 x2,..., x„,
z) = 0.
The definition of a function is extended to the case of n independent variables in
the above discussion. Most of the definitions and formulas given subsequently for
functions involving more than two variables are stated explicitly for functions involv¬
ing three or, in some cases, four variables. This is for convenience. These definitions
and formulas can also be extended to the general case of n variables, but additional
notation is required to write the corresponding statements concisely. Several results
concerning maxima and minima of functions of n variables, which are not obtained
easily as generalizations of the corresponding results for two variables, are stated in
matrix notation, after that notation is defined.

Continuity
A function /(x, y) is said to be continuous at x = a, y = b, if the following three
conditions are satisfied:
1. f(a, b) exists.
2. lim^a y^bf(x, y) exists.
3. lim,., ^/lx, y) = f(a, b), no matter in what manner x -*■ a, y -»b.
A function of x and y is said to be continuous in a region of the xy-plane if it is
continuous at every point of the region. Analogous definitions are appropriate for
functions of n variables.

3.2 Partial Differentiation

In this section partial differentiation, the total differential, and the total derivative are
defined and illustrated. Partial differentiation of implicit functions is also considered
and illustrated.
 3.2 Partial Differentiation / 325

Consider the function z of two independent variables x and y,

z =f(x, y)

If y is held constant, z is a function only of x and the derivative of z with respect to x

can be computed. The derivative obtained in this way is the partial derivative of z
with respect to x and is denoted by

dz df
f(x, y) fAx, y) fx
dx dx dx

Similarly, if x is held constant the partial derivative with respect to y can be

computed and is denoted by

dz Sf
f(x, y) fJx, y) f
dy dy dy

The partial derivative of z with respect to x is defined as

£= lim £ = lim /(v + Av- .v)-/(v- y)

dX Ax~*0 AX A v —► 0
Ax~* o Ax

and the partial derivative of z with respect to y is defined as

dz Az f(x, y + Ay) -fjx, y)

— = lim = lim
dy y->0 Ay Ay

note: In general, a function of any number of variables may have a partial

derivative with respect to each of its variables.

EXAMPLE

If z = 2x2 + 3xy — 6y2,

dz
— = 4x -f 3y
dx

~ = 3x - 12y
dy

EXAMPLE

If z — xy + In x,

dz 1
= y +
dx x

dz
— x
dy
326 / Differential Calculus: Functions of More Than One Variable

EXAMPLE

If z = (x + y) sin (x — y),

dz
— = sin (x - y) + (x + y) cos (x - y)

dz
— = sin (x — y) — (x + y) cos (x — y)

EXAMPLE

ir x3 - y3 , , dz dz
It z =-—, show that x ~ + y ~ = z.
xy dx dy

2-1
z = xy - x -12
yz

dz
= 2xy 1 + x 2y2
dx

■=- = —xy
dz 2-2 T - 1
- 2x y
dy

x — 4- y ^ = 2x2y~ 1 + x- 1y2 — x2y~ 1 — 2x~ *y2

2-1
xy — x -12
y = z

Just as a function of one variable can be represented by a curve in a plane, a

function of two variables can be represented by a surface in space. Consider the
surface represented by z — /(x, y) in Figure 3.1. If a plane is passed through any
point of P of the surface parallel to the xz-plane, it cuts the surface in the curve APB,
along which y remains constant; the slope of the tangent line to APB at P represents
the rate at which z changes with respect to x at the point P. Similar statements hold
for a plane parallel to the yz-plane. Thus the partial derivatives have the following
geometrical significance:

dz dz
= tan a = slope of APB at P
dx dx p

dz
tan [I = slope of CPD at P
dy

That is, if z =/(x, y), the partial derivatives — and — are interpreted geometrically

as the slopes of the curves of intersection of the surface z =/(x, y) and the planes
y = constant and x = constant, respectively.
Since, in general, the partial derivatives of a function z =/(x, y) are functions of x
 3.2 Partial Differentiation / 327

Figure 3.1

and y, they may be differentiated with respect to either xory; these derivatives, if
they exist, are called the second partial derivatives of z and are denoted, respectively,
by

5 tdz\ _ d2z _ _ d2f _

dx \c.xj dx2 xx dx2

d Idz\ d2z d2f

dx\dy) dx dy xy dxdy xy

d /8z\ d2z d2f

dy\dx/ dy dx yx dy dx yx

d Idz\ d2z _ d2f __

dy\dy) dy2 yy dy2 yy

A second- or higher-order partial derivative which is obtained by differentiating with

respect to more than one variable is referred to as a mixed partial derivative. Thus fxy
and fyx are mixed partial derivatives. Of the four derivatives above, only three are
generally distinct, since fxy = fyx for all values of x and y for which fxy and fyx are
continuous. Because of this property, it follows that a mixed partial derivative of
second (or higher) order can be obtained by differentiating with respect to the
variables in any order, provided only that the derivatives are continuous.
Similar definitions and equalities hold for higher-order partial derivatives of func¬
tions of two variables and for partial derivatives of functions of more than two
variables.
328 / Differential Calculus: Functions of More Than One Variable

EXAMPLE

If z = (x *12 + y2)312,

dz
?(x2 4- y ) 1 (2x)
dx

= 3x(x -f y 2 2)1/2

d2z
= ?x(x2 4- y2)~ / (2y) 1 2

dy dx

= 3xy(x 4- y2)~1/2
2

dz
= ?(*2 + y2)l,2(2y)
dy

= 3y(x 4- y 2 2)1/2

d2z
= ?y(*2 + y2) 1/2(2x)
dx dy

= 3xy(x 2 + y2)~ 1/2

d2z d2z 2 1/2

3xy(x 4- y2)
2
\~

dy dx dx dy

EXAMPLE

Tr a „ * „ , , x.9 ^ A _ , d3u d3u d3u . d3u

If u — Ax4 4- Bx3y + Cx2y2 4- Dxy3 4- Ey4, find — 3, -—-- , j-j-y , and ^- .
2 3

dx dx dy dx dy dy

-U = 4,4x3 4- 3Bx2y + 2Cxy2 4- Dy3

= 12,4x2 4- 6Bxy 4- 2Cy2

(s A

24y4x 4- 6By

d3u
6 Bx 4- 4 Cy
~dx2 dy
du
Bx3 4- 2Cx2y 4- 3Dxy 2 4 - 4Ey3
dy

d2u
2Cx2 4- Dxy 4- 12£y
6 2

dy2

d3u
4Cx 4- Dy 6

dy'2 dx

d3u
6Dx + 24£y
dy3

PROBLEMS
1. If u = xy — In xy, find ux and uy.

2. If z = (x - y)/(x 4- y), find z* and zy.

 3.2 Partial Differentiation / 329

3. If z — yexly, find zx and zy.

4. If z = In [ ~), find zx and zv.

\* +r/
5. If u = sin xy, find ux and uy.

6. If z = y/xy, find zx and zy.

7. If z = eau + bv2+c”\ find zu, z„, and zw .

8. If u = x2y 4- y2z + z2x, show that ux 4- uy + uz = (x 4- y 4- z)2.

9. If z = x2 sin (y/x) 4- y2 cos (y/x), find xzx 4- yzy.

10. If a = cos xu sin yu, find zx , zy, and zu.

d2z d2z
11. If z = xy + y In xy, show that x
+ ^ dx dy dy2'

du d2u du d2u
12. If M yjx — y2, show that
dx dx dy dy dx2 ’

d2z
13. If z = xy + xelly, show that -——
dx dy dy dx'

x + y , d2v d2v
14. If v = -, show that
x y dx dy dy dx ’

15. If z = 2x2 — 2y2 — 3x — 4xy2, show that ~~ — ^ f .

dx dy dy dx

16. If z = In (x2 4- y2\ show that = 0*

' dx1 dy

17. If/(x, y) = x3ex2+y, find fx,fy, and/xy.

18. If/(x, y) = Ax + By + Cexy, find fxx ,fyy, and fxy.

19. If/(x, y) = x2 cos y + y2 sin x, find fxx,fyy, and/xy.

20. If/(x, y) = x3 + 3x2y + 6xy2 - y3, find/XJC(2, 3),/yy(2, 3), and/xy(2, 3).

21. If/(x, y) = x4 - 4x3y 4- 8xy3 - y\ find/xx(0, l),/yy(0, 1), and fxy(0, 1).

22. If/(x, y) = 2x2 - 3xy + 4y2, find/x(l, -l)and/y(l, -1).

23. lff(x, y) = ~, find fx(X 1) and/,(3, 1).

X 3/

24. If/(x, y) = e sin (x + 2y), find/x|o, and/yjo, ^j.

xv
25. If u -, show that xux 4- yuY = 3u.
x + y

26. If u = In (x 4- y 4- z), show that In ux 4- In uy 4- In uz = — 3w.

Tr Axn 4- By" , .
27. If u = ^ 2 ^ 2 > show that xux 4- yuy = (n — 2)u.

28. If u = /(x, y), y = g(x, 3^), ux = vy, and uy = —vx , show that if x = r cos 9

and v = r sin 0, then ur = - v9 and vr = -ue.

r r

29. If u = /(x, y), x = r cos 0, and y = r sin 9, find w2 4- in terms of x and y.

330 / Differential Calculus: Functions of More Than One Variable

30. If z = f(x, y), x = g(u, v), and y = /i(w, v), find zuu.

31. If u = In ^Jx2 -I- y2, show that uxx = —uyy.

1
32. If u = - and r = yfx1 + y2 + z2, show that wxx + u yy -\- uzz

Answers to Odd-Numbered Problems

1 1
1. ux = y-; uY = x —
x y

3. zx = e”>;z, = e*l4l-?\

5. ux = y cos xy; uy = x cos xy

7. z. = ae"‘+bv2+c"s; z„ = 2bveau + bvl + cwl; zw = 3cw2e°u+bv! +ew3

9. 2z

17. /x = x2(3 + 2x2y,+y;f, = xV’+»;/„ = x2(3 + 2x2)e*, + J

19. /„ = 2 cos y - y2 sin x; f„ = -x2 cos y + 2 sin x;

/x>, = - 2x sin y + 2y cos x

21. fxx(0, 1) = 0;/„(0, 1) = - 12;/xv(0, 1) = 24

23. /x(3, 1)= -i;/,(3, 1) = i

29. u2 -I- w2

THE TOTAL DIFFERENTIAL

The total differential of a function

w =/(*> 33 z)
is defined by
dw dw dw
dw = — ax + — ay -I- — dz
dx dy dz

The separate terms —, — and ~ are sometimes called partial differentials of w

dx dy dz
with respect to x, y, and z, respectively. The sum of the partial differentials of a
function is its total differential.
In general, the total differential of a function

w =/(*!, X2 , ..., X„)

is the sum of all its partial differentials

Av = £ f dx,
i =1 CXi

If the x’s are differentiable functions of another variable, say f, then

dXi = —- dt
dt

and if the x’s are differentiable functions of two variables, say r and s, then

dx^fdr + fds
dr ds
 3.2 Partial Differentiation / 331

EXAMPLE

If z = 2x3 — 4xy2 4- 3y3,

dz = 6x2 dx — 4y2 dx — 8xy dy -f 9y2 dy

= (6x2 — 4y2) dx 4- (9y2 — 8xy) dy

EXAMPLE

If w = x 2 4- y2 4- z2, where x = r cos t, y r sin t, and z r, then

dw . dw ,
dw -r-dx 4- — dy + — dz
dx dy dz

2x dx + 2y dy + 2z dz

But

dx = cos t dr — r sin t dt

dy = sin t dr + r cos t dt

dz = dr

and thus

dw = 2x(cos t dr — r sin t dt) + 2y(sin t dr 4- r cos t dt) + 2z dr

Substituting for x, y, and z,

dw = 2(r cos2 t dr — r2 sin t cos t dt + r sin2 t dr + r2 sin t cos t dt + r dr)

- 2(r 4- r) dr

= 4 r dr

EXAMPLE

If u = (x 4- y)y/'x — y, compute du when x = 6, y = 2, dx = and dy = — 1.

, du 1 du
du = — dx + — dy
dx dy

= [(* - 4)1/2 + (x + y)(i)(x ~ y)~1/2] dx

+ [(x - y)l/2 + (x 4- y)( -i)(x - 1/2] dy

= (x — 1/2[x — y 4 i(x + y)] dx

4 (x - y)~ 1/2[x - y - i(x 4 )] dy

4

= i[(x - y)~ / (3x - y) dx 4- (x - y)_ 1/2(x - 3y) dy]

1 2

= i[(i)(16)(i) + (i)(0)( — 1)]

= 2
332 / Differential Calculus: Functions of More Than One Variable

THE TOTAL DERIVATIVE

dw dw dw
If w = fix, v, z) has continuous partial derivatives —, —and —- in some region
ox dy dz
and x, y, and z are functions of another variable t, then

dw dw dx dw dy dw dz
dt dx dt dy dt dz dt
dw. . dw
and — is said to be the total derivative of w with respect to t. Thus — represents the

rate of change in w as t changes. Heuristically, the total derivative can be obtained by

dividing the total differential by At and letting At approach zero as a limit. Note that
w is actually a function only of t.

EXAMPLE

dw
If w = x2 4- y2 4- z2, where x = e‘ cos t, y = e‘ sin t, and z = e', find
It'
dw dw dx dw dy dw dz
dt dx dt dy dt dz dt

^ = 2x(e' cos t — c' sin t) + 2y(el sin t 4- el cos t) 4- 2ze‘

Substituting for x, y, and z,

dw
= 2[^2r cos2 t — e2t sin t cos t + e2t sin2 t 4- e2t sin t cos t + e2t]
dt
= 2[e2'(cos2 t + sin2 f)] 4- 2e2t

= 4e2t

Alternatively,

w = e2t cos2 t + e2t sin2 t + e2t

dw
= 2e2t cos2 t — 2e2t sin t cos t + 2e2t sin2 t + 2c2' sin t cos t + 2c2'
dt

= 2c2'(cos2 t -f sin2 t) + 2c2'

= 4c2'

EXAMPLE

XV, . 1 ,
If u - + -, where x = 2r, v = -, and z = t ,
y z t

du du dx du dy du dz
dt dx dt dy dt + dz dt

2 x 1 2 yt
4
y y2t2 zt2 z2
 3.2 Partial Differentiation / 333

Substituting for x, y, and z,

du 1 2
= 2t + 2t-
dt ? 7
3
= 4t —
7

Similarly, if w =f (x, y, z) and x, y, and z are differentiable functions of r and s,

dw dw
then — and — can be obtained as follows:
dr ds
dw dw dx dw dy dw dz
— =--f---— -f- ——

dr dx dr dy dr dz dr

dw dw dx dw dy dw dz
ds dx ds dy ds + dz ds

In general, if w is a differentiable function of xlt x2, .x„ and the x’s are
differentiable functions of a second set of variables ul9 u2, .um, then the partial
derivative of w with respect to a variable in the second set, say Uj, is given by

dw dw dxi dw dx2 dw dx„

- =-i H-+•••-}--
duj dxl duj dx2 duj dxn duj

EXAMPLE

If u = x2 — xy + y2, where x = rs, y = r2 + s2,

du du dx du dy
dr dx dr + dy dr

= (2x — y)s + 2(2y — x)r

Substituting for x and y,

= (2rs — r2 — s2)s + ( r + 2s2 — rs)r

2 2 2
dr

= 2rs2 — r2s — s3 + 4r3 -f 4s2r — 2r2s

= 4 r3 — 3 r2s + 6 rs2 — s3

du du dx du dy
ds dx ds dy ds

= (2x — y)r + ( y — x)s

2 2

Substituting for x and y,

~ = (2 rs — r2 — s2)r + 2(2r2 + 2s2 — rs)s

ds

= 2r2s — r3 — rs2 + 4r2s 4- 4s3 — 2rs2

— 4s3 — 3rs2 + 6 r2s — r3

334 / Differential Calculus: Functions of More Than One Variable

EXAMPLE

If w = In (x* 2 + y2 + 2 z), where x = r4-5, y = r — s, and z = 2 rs,

dw dw dx dw dy dw dz
dr dx dr dy dr dz dr

2x 2j
(1) -P 1(25)
x 4- y 4- 2z (1) + + y 4~ 2 z 4- y + 2z

2(x + y + 2s)
x2 + y2 + 2z

2(r + s 4- r — s + 2s)
r2 + 2 rs + s2 + r2 — 2 rs + 52 + 4rs

4(r + 5)
2(r + s)2

2
r + 5
and

dw dw dx dw dy dw dz
ds dx ds + dy ds dz ds

2x 2y
(1) + Kx + y2 + 2z (2 r)
x2 + y2 + 2 z + _y + 2z
2{x - y + 2r)
x2 + y2 + 2z

2(r + s — r + s + 2r)
2(r + s)2

4(r + s)
2(r + s)2

2
r + 5

The composite function rule, which is valid also for partial derivatives, can be used
to obtain the total derivatives of functions whose independent variables are related in
various ways.
For example, if u = /(x, y, z), where x — t, y = y(t), and z = z(t), then
du du dx du dy du dz
dt dx dt dy dt dz dt

du du dy du dz
= — 4 ~- 4 “-

dx dy dt dz dt
dx
since — = 1.
dt
If u = f(x, y, z), where x = x(t), y = y(x), and z — z(f), then
du dudx dudydx dudz
dt dx dt dy dx dt + dz dt *

using the composite function rule.

 3.2 Partial Differentiation / 335

If u—f (x, y, z), where x = x(r, 5), y = y(r), and z = z(y), then
du du dx du dy du dz dy
dr dx dr + dy dr + dz dy dr

using the composite function rule, and

du du dx
ds dx ds
dy
since — = 0.
ds
If u = /(x, y, z), where x = x(r), y = y(x), and z = z(y), then

du du dx du dy dx du dz dy dx
dt dx dt dy dx dt dz dy dx dt

using the composite function rule.

DIFFERENTIATION OF IMPLICIT FUNCTIONS

The process of implicit differentiation discussed in previous sections can be for¬
malized using partial derivatives and can then be generalized to provide formulas for
obtaining partial derivatives of implicit functions of several variables.
The equation/(x, y) = 0 defines x and y as implicit functions of each other. For
purposes of derivation, let

U =f(x, y)

Then

du df df dy
dx dx dy dx
du
But if/(x, y) = 0, then u — 0, — = 0 and
yJ Jk

df + 8£dy
= 0
dx dy dx
Thus

d£
dy dx
dx df
Ty

, dx
and — can be obtained as the reciprocal Note that the equation
dy

-0 represents the process of implicit differentiation discussed

dx dy dx
previously.

EXAMPLE

dy dx
If ex sin y + ey cos x = 1, find and
dx dy
336 / Differential Calculus: Functions of More Than One Variable

/(x, y) = ex sin y + ey cos x — 1 = 0

df
dy dx
dx df
dy
ex sin y — ey sin x
ex cos y + ey cos x

The usual procedure of implicit differentiation can also be used to obtain

dy_ as follows:
dx’

ex sin y + ex cos y — + ey cos x —■ — ey sin x = 0

dx dx
dy ey sin x — ex sin y
dx ex cos y + ey cos x

dy
— can be obtained by either of the methods above or, more easily, as the reciprocal of
dy dx

dx _ 1
dy dy-
dx

ex cos y + ey cos x
ey sin x — ex sin y

EXAMPLE

If/(x, y) = x3 + y3 — 3axy = 0, find and ~.

dx dy
V
dy dx
dx df
dy
3x2 — 3 ay
3y2 — 3ax

ay — x2
y2 — ax

dx 1
dy dy
dx
y2 — ax
ay — x2

, dy
Note that and — could also be obtained using the usual procedure of implicit
dx dy
differentiation.

Similar formulas are appropriate for obtaining the partial derivatives of implicit
functions of any finite number of variables; these formulas can be obtained using the
same method of derivation shown above.
 3.2 Partial Differentiation / 337

If z is defined as an implicit function of x and y by the equation F(x, y, z) = 0, then

dF
az dx
for ^ + 0
dx ~dF dz
dz

dF
dz dy_
dy dF dz
dz

The choice of z as the dependent variable is arbitrary and other partial derivatives
can also be obtained; for example,

dF
dx
forf ^0
dx ~dF dy
dy

similarly, f^, and J can be obtained directly or as rectprocals.

dz dz dy
More generally, if z is defined as an implicit function of xlt x2, . xn by the
equation F(x1? x2, ..., xn, z) = 0, then

dF
dz dxt dF
for i = 1, 2, ..., n and 7^0
dxt ~dF dz
dz

dx.
Again, choice of z as the dependent variable is arbitrary and —1- can also be obtained
dxj
as follows

dF
dxt dxj_ dF
for i = 1, 2, ..., n, i # j and ^0
dxj dF dxt
d X;

EXAMPLE

Find -fZ and ~ for the function zex + ey yez = 0.

dx dy

djy = zex — —ey + ye2

dx

dF
— ey — e2
dy

dF
= ex — ye2
dz
338 / Differential Calculus: Functions of More Than One Variable

dF
dz dx ey — yez
dx dF ex — yez
dz
dF
dz dy ez — ey
dy dF ex — yez
dz

EXAMPLE

Find ^ and for the function z sin x 4- z2 4- x sin y 4- xyz = 0.

dx dx
1 *

dz z cos x + sin y 4- yz
1

dx dF sin x + 2z + xy
dz
^! x

dy _ z cos x 4- sin y 4- yz
1
1

dx dF " x cos y + xz
^4

PROBLEMS
1. If z = x3 4- x2y — y3, find dz.

2. If u = In (x2 4- y2 4- z2)1/2, find du.

3. If u = exyz, find du.

4. If u = ez sin [{x — y)z\ find du.

5. If z = 2x3 — 4xy2 4- 3y3, find dz.

6. If u = xy2z3, find du.

7. If x2 + y2 4- z2 = a2, find dz.

8. If u = x + 4xll2y1/2 —3y, x = t3, and y = find ^.

t dt

9. If x3 4- _y3 — 3bxy = 0, find ~.

dx

10. If x2 + 2xy + 2y = 15, find ^ if x = 2, v = 3.

dx

11. If x3 — y3 — 4x4 = — j, find if x = 2, y = —2.

dx

12. If Ax + By + Cexy = D, find ~ if x — y — 0.

dx

13. If .4x2 + By2 4- Cz2 = D, find zx and zy .

14. If xy 4- yz + zx = 9xyz, find zx and zy.

 3.2 Partial Differentiation / 339

15. If xz = cos yz -I- a, find zx and zy.

16. If ex + ey + ez = axyz, find yx .

17. If F(x, y, z) = 0, show that xyyzzx = — 1.

du
18. If u = x3 — 3xy + y3, x = r2 + s, and y = rs2, find —.

g1 q— 1 du
19. If u = xy + yz, x = —, y = —, and z = t2, find —.
t t dt
0z 02
20. If z = x In y + y In x, x = eu + v, and y = eu~v, find and
du dv

dz
21. If z = In (x2 + y2) + yjx2 + y2, x = eu cos v, and y = eu sin v, find
du

22. If u = xy + yz + zx, x = rs, y = sr, and z = r + s, find ur .

23. If x In yz — y In xz = 0, find zx and zy.

24. If exyz = ex + ey + ez, find zx and zy.

25. If ex + ey + ez = ex+y+z, find zx and zy.

v
26. If u In-w In uv = 0, find uv and uw .
w

Answers to Odd-Numbered Problems

1. dz = (3x2 -(- 2xy) dx 4- (x2 — 3y2) dy

3. du = exyz(yz dx + xz dy F xy dz)

5. dz = (6x2 — 4y2) dx -I- (9y2 — 8xy) dy

7. dz = — (x dx + y dy)/z

9 dy _ by — x2
dx y2 — bx

ii. £-i
dx

n Ax _ By
*3' * Cz ’ 11 Cz

15. zx = —z/(x + y sin yz); zy = — (z sin yz)/(x + y sin yz)

du _t/ v 2
19. Tre *(l-t)-?

21. ~ = 2 + e“
du

yz( 1 - In xz)
^3. / \
x(x - y)

xz(l — In yz)
zy = Hv - x)
ey +
25. zx =
e^h7y

ex + ez
zy = ex -I- ey
340 / Differential Calculus: Functions of More Than One Variable

3.3 Applications of Partial Derivatives

in Business and Economics
In this section several applications of partial derivatives in business and economics
are discussed. These applications include marginal cost, demand surfaces, production
functions, Euler’s theorem, constant product curves, returns to scale, and utility
functions.

MARGINAL COST
If the joint-cost function for producing the quantities x and y of two commodities is
given by

C = Q(x, y)
then the partial derivatives of C are the marginal cost functions:

dC. ,
— is the marginal cost with respect to x
ex

— is the marginal cost with respect to y

dy

In most economic situations marginal costs are positive.

EXAMPLE

If the joint-cost function of producing quantities x and y of two commodities is

C = x In (5 -f y)

then

dC
- — In (5 + y) is the marginal cost with respect to x

and

dC
-- is the marginal cost with respect to y
Ty 5 + y

EXAMPLE

If the joint-cost function for producing quantities x and y of two commodities is

C = 15 + 2x2 + xy + 5y2

then

dC
^ = 4x + y is the marginal cost with respect to x

and

dC = x -I- lOy is the marginal cost with respect to y

Ty
 3.3 Applications of Partial Derivatives in Business and Economics / 341

3C 3C
If x = 3 and y = 6, — = 18 and — = 63. Thus, if y is kept constant at 6, producing an
ex cy
additional unit of x adds $18 to the total cost; if x is kept constant at 3, producing an
additional unit of y adds $63 to the total cost.

DEMAND SURFACES
If there are two related commodities for which the quantities demanded are x and y
and the respective prices are p and q, then the demand functions can be represented
by

x=f(p,q) and y = g(p,q)

assuming that the quantities demanded, x and y, depend only on the prices, p and q,
of the two commodities. If a demand function of two independent variables is contin¬
uous, it can be represented by a surface, referred to as a demand surface.

note : The notation px and py is frequently used for the prices of the two commod¬
ities. It is not used in this section to avoid possible confusion with the subscript
notation for partial derivatives.

In the usual economic situations, the demand functions x =f(p, q) and y = g(p, q)
have the following properties:

1. All the variables, x, y, p, and q, are zero or positive.

2. If q is constant, x is a monotonically decreasing function of p; similarly, if p is
constant, y is a monotonically decreasing function of q.
3. The functions /(x, y) and g(x, y) and the region for which they are defined are
such that it is possible to obtain their inverse functions p = F(x, y) and
q = G(x, y).

For a constant price p, as q increases y decreases, but x may either increase or

decrease; if x increases, the two commodities are said to be competitive, since a
decrease in the demand for one corresponds to an increase in the demand for the
other. For a constant price p, as q decreases, y increases; if x also increases, the
commodities are said to be complementary, since an increase in the demand for one
corresponds to an increase in the demand for the other. Corresponding relationships
hold for changes in p for a constant q.

note: Usually, it is assumed that related commodities which are either competi¬
tive or complementary at one set of prices have the same relationship at other prices;
however, in some cases the relationship between commodities may differ for different
prices and thus must be defined for a particular set of prices.

Marginal Demand
If the demand functions for two related commodities are

x =f(p, q) and y = g(p, q)

then the partial derivatives of x and y are the marginal demand functions:

dx . .
— is the marginal demand of x with respect to p
op
342 / Differential Calculus: Functions of More Than One Variable

dx
is the marginal demand of x with respect to q
dq

dy
is the marginal demand of y with respect to p
dp

dy
is the marginal demand of y with respect to q
dq

For the usual demand functions, x increases if its corresponding price p decreases
Sx dy
and y increases if its corresponding price q decreases, so — and are negative for all
dp dq
economically meaningful values of p and q.
Sx d
If and — are both negative for given (p, q), the commodities are complemen-
dq dp
tary, since then a decrease in either price corresponds to increases in both demands.
Sx d
If — and are both positive for given (p, q), the goods are competitive, since then a
dq dp
decrease in either price corresponds to an increase in one demand and a decrease in

the other. If — and — have opposite signs, the commodities are neither complemen-
dq dp
tary nor competitive; in this case, a decrease in the price of one of the commodities
corresponds to increases in both demands while a decrease in the price of the other
commodity corresponds to an increase in one demand and a decrease in the other.
Such a situation is unusual but might occur, for example, if two different grades of a
material could be used and the better grade material could be obtained only by
processing the lower grade material.
note: Complementary and competitive commodities are defined in terms of the
signs of their cross elasticities in Chapter 2. Those definitions and the definitions
above are equivalent, since the signs of the cross elasticities are the same as the signs
of the appropriate partial derivatives representing marginal demand.

EXAMPLE

If the demand surfaces are linear functions of p and q,

x = fli -f bip + cxq y = a2 + b2p + c2q
then the marginal demand functions are
dx dy
= b2
dp dP
dx dy
T~ = C1 = c2
dq dq
Thus for the usual economic situations, bx < 0 and c2 < 0. If cx and b2 are both positive, the
commodities are competitive; if cx and b2 are both negative, the commodities are
complementary.

EXAMPLE

If the demand functions for two related commodities are given by

 3.3 Applications of Partial Derivatives in Business and Economics / 343

then the marginal demand functions are

dx 2a dy a
dp p3q dp ~ p2q

dx _ a dy a
dq p2q2 dq pq2
dx dy
Since — < 0 and < 0, the commodities are complementary.
dq dq

EXAMPLE

If the demand functions for two related commodities are given by

x = aeq~p y = bep~q a > 0, b > 0

then the marginal demand functions are

dx
= — aeq p d/ =
dp dp

dx
= aeq p di=-be’~*
dq dq
0 X- 3
Since —- > 0 and > 0, the commodities are competitive.
dq dp

EXAMPLE

If the demand functions for two related commodities are given by

x — axq2 — bxpq y = a2p2 — b2pq d\ > 0, a2 > 0, bx > 0, b2 > 0

then the marginal demand functions are

dx dy
= -b^q = 2a2p - b2q
dp dp

dx dy
= 2aIq-b1p = ~b2P
dq dq

dx dy
> 0 if — > hl >0if q-<^
dq 2a i dp P b2

dx bi dy
— < n0 iff -4 < _
dq p 2ai dp p b2

X > 0 if ? > — r>°if l <a~

p tfl P b2

Thus the demand functions are appropriate for values of p and q such that

by k q < di
a\ P b2
and the commodities are competitive for values of p and q such that the demand functions are
appropriate (since /ax < q/p < a2 /b2 => bx /2ax < q/p < 2a2 /b2).
344 / Differential Calculus: Functions of More Than One Variable

EXAMPLE

If the demand functions for two related commodities are given by

x = ae~pq y = bep~q a > 0, b > 0

then the marginal demand functions are

dx pq
dy
= —aqe = bep~q
dp dp
dx
= —ape - pq = -be'-<
dq dq

dx dy
Since — <0 and — > 0, the commodities are neither competitive nor complementary.
dq dp

Partial Elasticities of Demand

If the demand functions for two related commodities are

x=f(P><l) and y = g(p, q)

then the partial elasticities of demand are given by

In x
Ex p dx dx the partial elasticity of demand x with respect to
Yp q=c i
x dp price p, for a constant price q =
In p
dx
d
In x
Ex q dx dx the partial elasticity of demand x with respect to
Yq p = c2 x dq d price q, for a constant price p = c2
In q
dx
d
in y
Ey P . dy dy the partial elasticity of demand y with respect to
Ep q = c 3 y dp price p, for a constant price q = c3
In p
dy
d__
In y
Ey q dy dy the partial elasticity of demand y with respect to
Eq P = C4 y dq price q, for a constant price p = c4
In q
dy

Note that Ex/Eq and Ey/Ep are cross partial elasticities of demand; their signs,
which are the same as the signs of the corresponding marginal demands, can be used
to determine whether commodities are competitive or complementary.

EXAMPLE

If the demand functions for two related commodities are given by

x = aeq~p y = bep~q
then the partial elasticities of demand are
 3.3 Applications of Partial Derivatives in Business and Economics / 345

Ex p dx P Ey _ p dy
(~aeq p) = -p hh^p-q) = p
Ep x dp aeq-p Ep y dp

Ex q dx q Ey _ q dy
(aeq p) = q
Eq x dq aeq~p Eq y dq

and, since Ex/Eq > 0 and Ey/Ep > 0, x and y are competitive commodities (as above).

EXAMPLE

If the demand functions for two related commodities are given by

x = axq2 — bipq y = a2p2 — b2pq a{ > 0, a2 > 0, bx > 0, b2 > 0

then the partial elasticities of demand are

Ex p dx p , u \ bi P
Ep x dp axq —bxpq a^q — b^p

Ex q dx q 2axq - bxp
v- = ' ^- =-2- u-p) =-r—
Eq x dq axq — b^pq a^q-b^p

El_P. dJ_ _ _ 2_n h - laiP~h2q

r- 2 l (2@2 P b2 q) l
Ep y dp CI2P b2pq CI2P ~ ^2<?

Ey = q. dy = q <b , _ M
Eq y dq a2p2 — b2pq 2 a2p — b2q

and, since Ex/Eq > 0 and Ey/Ep > 0, x and y are competitive commodities (as above).

PROBLEMS
1. For the following joint-cost functions, determine the marginal cost with respect to x and the
marginal cost with respect to y.
a. C = x2 In (y +10) c. C = ex + ey + xy + 5
b. C = x3 + 2y2 — xy + 20 d. C = x2y2 — 3xy + y + 8
2. For each of the following pairs of demand functions, determine the four marginal demands,
the nature of the relationship between the two commodities, and the four partial elasticities
of demand. Quantities are denoted by x and y and the corresponding prices are denoted by
p and q, respectively.

a. x = 20 — 2p — q
e. x
p2q
y = 9 — p — 2q
16
b. x = 15 — 2p + q y=
M* 1
y = 16 + p - q
4
f. X =
c. x = 5 — 2p + q pq
y = 8 — 2p — 3q 16
346 / Differential Calculus: Functions of More Than One Variable

Answers to Odd-Numbered Problems

dC dC
1. a. — = 2x In (y + 10) = ex + y
(y A- C‘ dx

dC x2 dC
= ey + x
dy y + 10 dy

dC . dC
-- = 3x2 — y = 2xy2 - 3y
ox ' dx

dC dC
-4 y-x = 2 x2y — 3x + 1
dy dy

PRODUCTION FUNCTIONS
The production of most commodities requires the use of at least two factors of
production, for example, labor, land, capital, materials, or machines. If the quantity z
of a commodity is produced using the amounts x and y, respectively, of two factors of
production, then the production function

Z =f(x, y)
gives the amount of output z when the amounts x and y, respectively, of the inputs are
used simultaneously. For such a representation to be economically meaningful, it is
assumed that the amounts of the inputs can be varied without restriction, at least in
the range of interest, and that the production function is continuous.

Marginal Productivity
dz
If the production function is given by z = /(x, y), then the partial derivative — of z
ox
with respect to x (with y held constant) is the marginal productivity of x or the
dz
marginal product of x; the partial derivative — of z with respect to y (with x held

constant) is the marginal productivity of y or the marginal product of y. Note that the
marginal productivity of either input is the rate of increase of the total product as
that input is increased, assuming that the amount of the other input remains
constant.
Marginal productivity usually is positive for a considerable range, that is, as the
amount of one input increases, with the amount of the other input held constant, the
output also increases. However, as the input of one factor increases with the input of
the other factor held constant, the output usually increases at a decreasing rate until
the point is reached at which there is not further increase in output and, in fact, a
decrease in output occurs with additional inputs of the factor. This characteristic
behavior of production functions is known as the law of eventually diminishing mar¬
ginal productivity.

EXAMPLE

If the production function is given by

z = 4x3/V/4

then the marginal productivity of x is

 3.3 Applications of Partial Derivatives in Business and Economics / 347

and the marginal productivity of y is

dz
dy
= x3/V3/4

dz . . . dz
Note that — is always positive, but decreases as x increases; similarly, — is always positive,
dx dy
but decreases as y increases.

EXAMPLE

If the production function is given by

z = 4xy - x2 — 3y2

then the marginal productivity of x is

dz
= 4y - 2x
dx

and the marginal productivity of y is

dz
— — 4x — 6y
dy
dz dz dz dz
Note that — > 0 for x < 2y, — = 0 for x = 2y, and — <0 for x > 2y; similarly, — >0 for
dx dx dx dy
dz dz
y < |x, — = 0 for y = fx, and — <0 for y > §x. Thus the marginal productivities at first
dy dy
increase and then decrease as input increases.

EXAMPLE

If a production function is given by

z2 + 4x2 + 5y2 — 12 xy = 0
where z is the amount of output and x and y are the amounts of the inputs, find the marginal
productivities.

~ = Sx- 12 y
dx

dF
= lOy - 12x
dy

dF
= 2z
dz
The marginal productivity of x is
dF
dz dx 6 y — 4x
dx dF z
dz
and the marginal productivity of y is
dF
dz dy 6x — 5y
dy dF z
dz
348 / Differential Calculus: Functions of More Than One Variable

dz
Since z is positive for any economically meaningful region of the production function, — is
dx
-» dz f-
positive for x <^y and negative for x > ^y and — is positive for y < §x and negative for
d\

EULER'S THEOREM
If the function z =f(x, y) has the property that, for any constant A,

/(Ax, Ay) = A"f(x, y)

then z is said to be homogeneous of degree n. If n > 0, the function is said to be

positively homogeneous; if n = 1, the function is said to be linear homogeneous.
If z = f(x, y) is positively homogeneous of degree n and the first-order partial
derivatives exist, then it can be shown that

dz dz
nf (x, y)
XFx + yFy
This relationship is known as Euler s theorem.

EXAMPLE

The function z = f(x, y) = 3x4 + 2x2y2 + 7y4 is homogeneous of degree 4, since

/(Ax, Ay) = 3A4x4 + 2A4x2y2 + 7A4y4

= W(x, y)
Thus, according to Euler’s theorem,

dz dz
xrx + ydy = 4f{x’y)
which can be verified as follows:

dz dz
= x(12x3 + 4xy2) + y(4x2y + 28y3)
Xdx + ydy

= 12x4 + 8x2y2 + 28y4

= 4/(x, y)

EXAMPLE

The function z =/(x, y) = -2~y 2 is homogeneous of degree zero, since

x 4- y

A2xy
/(Ax, Ay) =
A2(x2 + y2)

= A°f(x, y)

Thus, according to Euler’s theorem,

 3.3 Applications of Partial Derivatives in Business and Economics / 349

which can be verified as follows:

y(x2 + y2) - xy(2x) x(x2 + y2) - xy(2y)

X (x2 + y2)2 y (.*2+y2)2
2xy(x2 + y2) — 2x3y - 2xy3
(x2 + y2)2

= 0

Linear Homogeneous Production Functions

In economic theory, production functions are frequently assumed to be linear homo¬
geneous, because such functions have convenient properties. If a production func¬
tion z =f(x, y) is linear homogeneous, then according to Euler’s theorem,

dz dz v
x
dx
+ yT=f^y)
total due total due
to to
factor x factor y
That is, total production or output is equal to the product of the amount of one input
and its marginal productivity plus the product of the amount of the other input and
its marginal productivity. The total product is thus allocated to the two factors on
the basis of their marginal productivities.
Euler’s theorem plays an important part in the development of the marginal-
productivity theory of distribution. The basic assumptions of this theory are that
each input is paid the value of its marginal productivity and that total output is just
exhausted. As shown by Euler’s theorem, these conditions are satisfied by linear
homogeneous production functions; thus in the development of the marginal pro¬
ductivity theory of distribution, production functions are generally assumed to be
linear homogeneous. Note that if the production function is homogeneous of degree
n and each input is paid the value of its marginal productivity, then total output
exceeds payments for n > 1 and is less than payments for n < 1.

CONSTANT PRODUCT CURVES

The production function z = f(x, y) is frequently studied by considering the family of
curves

f(x, y) = constant

in the xy-plane. These curves, no two of which intersect, are called constant product
curves, equal product curves, or isoquants. Each curve shows the combinations of
factors x and y which result in a specified output. As discussed below, constant
product curves are such that a decrease in the input of one factor is compensated for
dy
by an increase in the input of the other factor. The slope — of the tangent at a point
dx
on an isoquant is the rate at which x must be substituted for y to maintain the level of

output. The negative of this derivative — ^ is defined as the rate of technical

dx
substitution.
350 / Differential Calculus: Functions of More Than One Variable

Figure 3.2

Equal product curves are of various shapes, depending on the nature of the
production function, but in general they have negative slopes and are convex to the
origin in the area of interest. However, there are some economic situations in which
portions of the equal product curves have positive slopes, that is, there are some
situations in which marginal productivity is diminishing and additional amounts of a
factor of input result in decreased production. For example, if the two factors of
production are land and labor, so many men might be employed (for a fixed amount
of land) that they would interfere with each other and production would decrease.
However, assuming the objective of profit maximization, portions of equal product
curves having positive slopes would not be used in practice, since the given output
could be obtained at less cost by a factor combination on a portion of the curve
having negative slope. Thus, as shown in Figure 3.2, only the portions of the equal
product curves lying between their respective vertical and horizontal tangents are
relevant; the lines through these points of tangency are known as ridge lines. The firm
would never operate above the upper ridge line or to the right of the lower ridge line,
since this would require more input than necessary to produce a specified output.
The assumption that equal product curves are convex to the origin implies that
the marginal significance of one factor with respect to the other factor decreases
along an equal product curve—that is, as the amount of factor x that is used in¬
creases, the amount of factor y that can be given up in exchange for a further unit of
factor x decreases, if the product remains constant; similarly, as the amount of factor
y that is used increases, the amount of factor x that can be given up in exchange for a
further unit of factor y decreases, if the product remains constant. If the marginal
significance of a factor increased as the amount used increased, as would be the case
for a concave equal product curve, then only that one factor should be used, since
each unit of the factor would be progressively more worth purchasing. Thus, in
practice, any portions of equal product curves for which marginal significance of the
factors is not decreasing are not relevant.
If factors x and y are purchased in a perfectly competitive market at constant unit
prices, then the total cost of production is given by

c = PXX + pyy + b
 3.3 Applications of Partial Derivatives in Business and Economics / 351

Figure 3.3
y

where px and py are the respective prices of x and y and b is the cost of the fixed
inputs. Isocost curves give the combinations of input factors that can be purchased
for a specified total cost; for the cost function above, the isocost curves are repre¬
sented by the family of lines

Pxx + Pyy + b = constant

The slopes of the isocost lines are equal to the negative of the input price ratio, since

px b constant
y =-x - —h -
Py Py Py

The greatest total output which can be produced for a specified total cost is given by
the point of tangency of the relevant isocost curve and an isoquant. Similarly, the
lowest possible cost of producing a specified output is given by the point of tangency
of the relevant isoquant and an isocost curve. The curve connecting the points of
tangency of the isocost curves and isoquants is known as the expansion path or scale
line. This line gives the lowest-cost combinations of factors as output increases; the
firm will thus always produce at some point on the scale line (see Figure 3.3). The
expansion path, at all points of which the rate of technical substitution equals
the fixed input-price ratio, is a straight line if the production function is homogen¬
eous of any degree. A straight-line expansion path does not, however, imply a homo¬
geneous production function.

RETURNS TO SCALE
Returns to scale describes the change in output with a proportionate increase in all
inputs. If output increases by the same proportion as the inputs increase, then returns
to scale are constant; if output increases by a greater proportion than inputs, returns
to scale are increasing; if output increases by a smaller proportion than inputs,
352 / Differential Calculus: Functions of More Than One Variable

Figure 3.4
y

returns to scale are decreasing. Very frequently, there are increasing returns to scale
for relatively small inputs (economies of scale), returns to scale are then constant for
a range of inputs, and finally there are decreasing returns to scale for large inputs
(diseconomies of scale).
Returns to scale are easily determined for homogeneous production functions. If a
production function is homogeneous of degree n, returns to scale are increasing if
n > 1, constant if n = 1, and decreasing if n < 1. In practice, a degree of homogeneity
other than 1 is seldom assumed for a production function. If a set of equal product
curves (not necessarily homogeneous) represent successive increases in production
of a given amount, then returns to scale increase as output increases if the distance
along the scale line between successive curves decreases; similarly, returns to scale
decrease as output increases if the distance along the scale line between successive
curves increases. Thus, in Figure 3.4, returns to scale increase from 20 to 80 units of
production and decrease from 80 to 140 units of production.
Note that if the proportions of the amounts of the two factors of production vary,
it is necessary to refer to returns to outlay rather than returns to scale, since the
factors are not increased by the same scale.

UTILITY FUNCTIONS
In the theory of consumer behavior, the consumer is assumed to choose among the
alternatives available to him in such a way that the satisfaction derived from consum¬
ing commodities is maximized. This implies that the consumer is aware of his alter¬
natives and is capable of evaluating them. All relevant information concerning the
satisfaction a consumer derives from various quantities of commodities is contained
in his utility function.
The concept of utility is based not on some type of psychological satisfaction on
the part of the consumer, but on his choice behavior. The assertion that a consumer
derives more satisfaction or utility from A than from B means that he would choose
 3.3 Applications of Partial Derivatives in Business and Economics / 353

A rather than B if A and B were presented as alternatives. In addition, the consumer

is assumed to make consistent choices so that if A is preferred to B and B is preferred
to C, then A is preferred to C.
Modern utility theory assumes only an ordinal utility scale; that is, for all pairs of
alternatives A and B, the consumer is assumed to know whether he prefers A to B,
prefers B to A, or is indifferent between them; he is not assumed capable of assigning
numbers representing amounts of utility to alternatives.
A consumer’s ranking (preference order) of commodities is expressed mathemat¬
ically by his utility function. The utility function associates certain numbers (utility)
with various quantities of commodities consumed, but these numbers represent only
a ranking or ordering of preferences. Thus if the utility of alternative A is 20 and the
utility of alternative B is 5, then A is preferred to B but it is incorrect to say that A is
preferred 4 times as strongly as B.
Consider the simple case in which the consumer’s purchases are limited to two
commodities and the utility function is
u =f(qi. qi)
where q1 and q2 are the quantities of the two commodities Qx and Q2 which are
consumed. It is assumed that f(qx, q2) is continuous and has continuous first-order
and second-order partial derivatives.
The utility function is with respect to consumption during a specified period of
time, a period long enough for the desire for variety to be satisfied but not so long
that tastes change. Utility is derived from the consumption of commodities. By
definition, a commodity is something of which the consumer would rather have more
than less; otherwise a discommodity is involved. In reality, a commodity consumed
in sufficiently large amounts may become a discommodity (for example, too much
candy); the following discussion assumes that such a point of satiation has not been
reached. Thus the partial derivatives of U with respect to qx and q2, denoted/! and
f2, are positive.
Since the utility function is continuous, a given level of utility can be derived from
an infinite number of combinations of ql and q2. The locus of all combinations of
quantities of commodities from which the consumer derives the same utility is an
indifference curve. A collection of indifference curves corresponding to different levels
of utility is an indifference map. (See Figure 3.5.) One indifference curve passes
through every point of the first quadrant; indifference curves corresponding to in¬
creasing levels of utility are farther from the origin. Indifference curves cannot inter¬
sect because this would imply that the same combination of qx and q2 has different
utilities.
Note that U = f(qx, q2) can be represented by a utility surface in three dimen¬
sions. Each indifference curve represents the intersection of that surface and a plane
parallel to the qx q2-plane at a certain value of U. The points on an indifference curve
represent the various combinations of quantities of the commodities that have the
same utility for the consumer and among which, therefore, he is indifferent.
The total differential of the utility function U = f(qlt q2) is
dU = f2 dq1 +f2 dq2
where/i and f2 are the partial derivatives of U with respect to qx and q2, respectively.
Since the change in utility along any indifference curve is zero by definition, aU = 0
and
f\ dqt +f2 dq2 = 0

= f\
dqi h
354 / Differential Calculus: Functions of More Than One Variable

Figure 3.5

<7 2

q2) c4

q2) C3
q 2)
c2
q2)

The slope of an indifference curve dq2/dqx is the rate at which a consumer would be
willing to substitute Qj for Q2 or Q2 for Qx to maintain a given level of utility. The
negative of the slope—dq2 /dql is the rate of commodity substitution or marginal rate of
substitution of Qifor Q2 or Q2for and equals the ratio of the partial derivatives of
the utility function. Note that the rate of commodity substitution at a point on an
indifference curve is the same for movements in either direction, so the verbal
definition can be in terms of substituting Qx for Q2 or Q2 for Qx.
The partial derivatives fx and f2 are said to be marginal utilities of the commod¬
ities Q1 and Q2 . The magnitudes of the marginal utilities are without meaning when
the utility function is ordinal, but their signs and ratios are meaningful. Note that if
the marginal utilities are positive, an increase in the quantity consumed of one
commodity with no change in the quantity consumed of the other commodity will
move the consumer to a higher indifference curve.

PROBLEMS
1. For each of the following production functions, find the marginal productivities. Output is
denoted by z and the inputs are denoted by x and y.

a. z — 25-at x = 1, y = 1
x y

b. z = 5xy — 2x2 — 2y2 at x = 1, y = 1

c. 16z2 — z — 80 + 4(x — 5)2 + 2(y — 4)2 = 0

d. 6z3 — z2 — 6x — 24y + x2 + 4y2 + 50 = 0

2. The Cobb-Douglas production function for the economy as a whole is given by

 3.3 Applications of Partial Derivatives in Business and Economics / 355

where z is total product, x is quantity of labor, y is quantity of capital, and a, b, and c are
constants. It is frequently assumed that fr + c = 1. Is this function homogeneous and, if so,
of what degree?
3. For each of the following production functions, determine the degree of homogeneity and
the nature of the returns to scale. Output is denoted by z and the inputs are denoted by x
and y.

a. z = 3x3 + 5xy2 + y3 c. z = 25y6 — x2y4

, 14 20 25
b. z =- d. Z = —2 d- + -
x y xy y^
4. In addition to the property known as Euler’s theorem, linear homogeneous functions have
the following properties: If z = /(x, y) is a linear homogeneous function, then

z (y\ z y
- that is, - is a function of -
X x) x x

I z
= g2\-\ that is, - is a function of -
Iy \y) y y
dz
= that is, ~ is a function of -
dx \x/ dx x

dz
that is, ^ is a function of -
dy Ky) dy y

dh y d2z
dx2 x dx dy
I
<^z x d2z
dy2 y dx dy

Determine whether each of the following functions is homogeneous; for homogeneous

functions, determine the degree and demonstrate Euler’s theorem; for linear homogeneous
functions, also demonstrate the other three properties given above.

a. z = 3x2 + 4xy + 15y3 i. z = 4x2ey + 3y2ex

b. z = 4x3 + x2y — 3xy2 — 7y3 j. z — 3x2 In al,y2 — 5y3 In blly3

c. z = 3ex + 3ey x2 + xy
k. z
3y
d. z = 6 In 3~2x - In 4“5>’
l. z = 3x2y + 4xy2 + y3 + 10
x4 -I- 3x2y2 4- xy3 + 6y4
e. z =-o-
x xy
m. z =
x + y
f. z = 3x In 5x/y - 9y In 8>/x
n. z = xy — In xy
x2y + y3
g-z' a*1- -

, x2 + 3xy + y2
h. z =-j-
y

5. If a consumer’s utility function is given by U = q^q2 and the consumer purchases 4 units of
Qi and 5 units of Q2, a. what quantity of must he purchase to maintain the same level of
utility if his purchase of Q2 increases to 6 units? b. What quantity of Q2 must he purchase to
maintain the same level of utility if his purchase of^ increases to 6 units ?c. What quantity
of Qx must he purchase to maintain the same level of utility if his purchase of Q2 decreases
to 4 units? d. What quantity of Q2 must he purchase to maintain the same level of utility if
his purchase of Q{ decreases to 2 units?
356 / Differential Calculus: Functions of More Than One Variable

6. Find the marginal utilities for each of the following utility functions. Utility is denoted by U
and the amounts of the two commodities consumed are denoted by qx and q2 respectively.
a. U = q\ q2

b. U = qx q2 + q\

Answers to Odd-Numbered Problems

dz dz x — 3 VO

dx ’ dy z(9z — 1)

dz dz dz My - 3)
dx ’ dy dy z(9z — 1)

dz 8(x — 5)
dx 32 z — 1

dz 4 (y — 4)
~dy= ~ 32z - 1
3. a. degree 3, increasing returns to scale
b. degree — 1, decreasing returns to scale
c. degree 6, increasing returns to scale
d. degree —2, decreasing returns to scale

5. a. qx = 25/9 c. q, = 25/4

d. q2 = 5^2
b.

3.4 Maxima and Minima

of Functions of Two Variables
Just as the first and second derivatives of a function of one variable can be used to
■.

determine its maxima and minima, the first-order and second-order partial deriva¬
tives of a function of two variables can be used to determine its maxima and minima.
The procedure becomes more complicated as the number of variables increases, but
the definitions and methods are essentially the same. The discussion below concerns
functions of two independent variables; maxima and minima of functions of n vari¬
ables are discussed in Chapter 8.
A function f(x, y) of two independent variables is said to have a local maximum
value (or minimum value) for x = a, y = b, iff(a, b) is greater (or less) than/(x, y)
for all values of x and y close to x = a, y = b. If/(x, y) has a maximum (or minimum)
value at x = a, y = b, it follows that the function /(x, b) has a maximum (or mini¬
mum) at x = a and the function/(a, y) has a maximum (or minimum) at y = b. Thus,
for the function /(x, y) to have a maximum (or minimum) at x = a, y = b, it is
necessary that

d
/(*> y) = o
dx x — a, y = b

d
f(x, y) =o
dy x = a, y = b
 3.4 Maxima and Minima of Functions of Two Variables / 357

These two conditions are used to determine the critical points; the procedure below
is then used to determine whether the critical points are in fact local maxima or
minima.

a
f{x, y) = 0
dx x = a, y = b
If

f(x, y) = 0
dy x = a, y = b

and

A = f(x, y) f(x, y) f(x, y)

dx' x = a, y = bl dy' x = a, y = b/ dx dy x = a, y = b)

d2 d2
maximum at x = a, y = b, if /(x, y) < 0 and ^ /(x, y) < 0
A > 0
d2 d2
then minimum at x = a, y = b, if 2 /(x, y) > 0 and ^ /(*» y) > 0
dx' dy'

A < 0 => neither maximum nor minimum at x = a, y = b, but a saddle point at

x = a, y = b
> A = 0 => test fails, function must be investigated near x = a, y = b

These conditions can be stated more concisely as follows:

necessary condition for a critical point: fx = 0,fy = 0

determination of local maxima or minima:

(maximum iffxx < 0, fty < 0

| minimum if fxx > o J„ > 0
saddle point

test fails and the function must be investigated near the critical
point
where the partial derivatives are evaluated for the critical
point

Note that this procedure is used to identify local maxima and minima. In order to
determine global maxima or minima it is necessary not only to compare the values of
the local maxima or minima, but also to determine the values of the function at the
limits of its range, if its range is limited. The procedure for determining global
maxima and minima for functions of two variables is thus analogous to the
procedure for functions of one variable.
Examples of functions having at the origin (a) a minimum, (b) a maximum, and (c)
a saddle point are illustrated in Figure 3.6. Note that at a saddle point a function has
a minimum with respect to one variable and a maximum with respect to the other
variable. For example, in Figure 3.6(c) the origin looks like a minimum to a person
traveling along the surface in the yz-plane, but to a person traveling in the xz-plane
the origin looks like a maximum.
358 / Differential Calculus: Functions of More Than One Variable

Figure 3.6

EXAMPLE

Examine the function

/(x, y) = x2 + xy + y2 - 3x + 2
for maximum and minimum values.

= 2x + y - 3
dx

df
= x + 2y
dy
2x + y = 3
x + 2y = 0

3 y = -3

y = -1, x = 2 so (2, — 1) may be a maximum or minimum

yi = 2
dx2

= 2
dy2
d2f
= 1
dx dy
 3.4 Maxima and Minima of Functions of Two Variables / 359

A = 4 - 1 > 0

d2f d2f
= 2 > 0 = 2 > 0
dx2 dy2

Thus f(x, y) has a minimum value at (2, — 1).

EXAMPLE

Examine the function

f(x, y) = xy - In (x2 + y2)

for maximum and minimum values.

6L- 2x
dx ^ x2 + y2

dl-x_ 2y
dy x2 -I- y2

x2y + y3 — 2x = 0

x3 + xy2 — 2 y = 0

— 2x2 + 2 y2 = 0

xz2 = yz2

Substituting for x2 in the first equation and for y2 in the second equation, y3 = x and x3 = y,
so (1, 1) and (-1, — 1) may be a maximum or minimum.

d2f 2(x2 + y2) — 2x(2x) 2 (y2 — x2)

dx2 (x2 -f y2)2 (x2 + y2)2

d2f _ 2(x2 4- y2) - 2y(2y) _ 2(x2 - y2)

dy2 (x2 + y2)2 (x2 + y2)2

df _ 4 xy
dx dy + (x2 + y2)2

d2j_ syi = 0
dx2 x=y= 1 dx2 x-y-

d2/
-0
dy2 x=y= 1 dy2 x—y— 1

d2f d2f
= 2
dx dy x = y= 1 dx dy x—y— - 1

A =0 — 2 <0

Thus f(x,y) has no maximum or minimum, but does have saddle points at (1, 1) and
(-i. -i)

EXAMPLE

Examine the function

f(x, y) = 1 -f x2 - y2
for maximum and minimum values.
360 / Differential Calculus: Functions of More Than One Variable

dj_
= 2x
dx

df
= -2y
dy

2x = o if x = 0|
so (0, 0) may be a maximum or minimum
-2 y = 0 if y = 0

ay
= 2
dx2 Hi *•

d2/
= -2
dy2

s2f
= 0
dx dy

A = - 4-0 <0

Thus/(x, y) has neither a maximum nor a minimum at (0, 0), but does have a saddle point at
(0, 0).

EXAMPLE

Examine the function

/(x, y) = sin x + sin y + sin (x + y)

for maximum and minimum values.

df
= cos x + cos (x + y)
dx

df
= cos y + cos (x + y)
dy

If — = — = 0, cos x = cos y = — cos (x + y), so (7t/3, 7i/3) may be a maximum or minimum.

dx dy

d*f
= —sin x — sin (x + y)
dx2

d*f
= -sin y — sin (x + y)
dy2

d2f
= —sin (x + y)
dx dy

d_V _ T5 _ 73 = = -A
dx2 x = y = n/3 2 2
d2f
dy2 x = y = n/3 2 2
= -73
d2f 73
dx dy x = y = n/3 2

A = (-73)(-7"3)-(-^)
— 3 —| — | > 0
 3.4 Maxima and Minima of Functions of Two Variables / 361

dV <0 and
d2/
<0
dx2 dy2

Thus f(x, y) has a maximum value for x = y = n/3.

EXAMPLE

Examine the function

f(x, y) = 25 + (x - yf + (y - 1)'

for maximum and minimum values.

df
= 4(x - y):
dx

df
= — 4(x - y)3 + 4(y - l)2
dy

(x — y)3 = 0 if x = y
so (1, 1) may be a maximum or minimum
(X - >’)3 = (y - l)3 if X = y = 1j
d2f
= i2(x - yy
dx2

d2f
- 12(x - y)2 + 12(y - 1)^
d/

d2f
dx dy = 712(x ~ y):

ay
dx2 x = y= 1

dV
dy2 x = y=\

d2f
= o
dx dy x — y= 1

so A = 0 and/(x, y) must be examined further for maximum or minimum values. Suppose h

and k are arbitrarily small positive or negative numbers; then

/(l + h, 1 + k) -/(1, 1) = 25 + [(1 + h) - (1 + k)]4 + (1 + k - l)4 - 25

= (h- k)4 + k4

But (h — /c)4 + /c4 > 0 for all h and k, so/( 1 + h, 1 + k) > /(1, 1) for all h and k, and thus there
is a minimum at (1, 1).

EXAMPLE

Examine the function

/(x, y) = x2 — 6xy + 9y2 4- 3x — 10

for maximum and minimum values.

362 / Differential Calculus: Functions of More Than One Variable

-r- = 2x — 6y 4- 3
dx

~ = —6x + 18 y
dy

2x — 6 y — — 3

— 2x + 6y = 0

no simultaneous solution, so no maximum or minimum or saddle point

EXAMPLE

If the demand functions are

p — 36 — 3x and q = 40 — 5y

and the joint-cost function is

C = x2 + 2 xy + 3 y2

determine the quantities and prices that maximize profit for the monopolist and find the
maximum profit.

P = px + qy — C

- 36x — 3x2 + 40y — 5y2 — x2 — 2xy — 3y2

— — 4x2 — 8y2 — 2 xy + 36x + 40 v

dP
= — 8x — 2y 4- 36
dx

dP
= — 16y — 2x + 40
dy

4x + y = 18

x + 8y = 20

31 y = 62

y= 2(
so P may have a maximum or minimum at (4, 2)
x = 4

d2P
= -8
dx2

d2P
= -16
dy2

d2P
= -2
dx dy

A = 128 — (— 2)2 - 124 > 0

d2P d2P
<0 and <0,
dx2 dy2

so P has a maximum value for x = 4, y = 2. If x = 4, y = 2, then p = 24, q = 30, and

1P max = 1112
 3.4 Maxima and Minima of Functions of Two Variables / 363

EXAMPLE

Suppose the production function is

16z = 65 — 2(x — 5)2 — 4(y — 4)2

The unit prices of the inputs x and y (under pure competition) are 8 and 4, respectively, and the
unit price of the output is 32; determine the maximum profit.

P = 32z - 8x - 4y
= 130 - 4(x - 5)2 - 8{y - 4)2 - 8x - 4y
dP
= ~8(x — 5) — 8 = — 8x -f 32
dx
dP
= — 16(y — 4) — 4 = — 16 y + 60
dy
8x = 32
x = 4
so P may have a maximum or minimum at (4,
16 y = 60
,, — 15
y- t

d2P
= -8
dx2
d2P
= -16
dy2
d2P
-0
dx dy
A = (— 8)( —16) - 0 = 128 > 0
d2P d2P
<0 and <0,
dx2 "w dy2
so P has a maximum value for x = 4, y = ^ and Pmax = 78^.

EXAMPLE

In simple bivariate regression analysis, the model involves the assumption of a linear relation¬
ship between a predicted variable y and an explanatory variable x and a disturbance term u.
For a sample of n observations of x and y,

yt = a + /?x, + Ui i = 1, 2, ..., n

and the problem is to obtain estimates of a and /?. These estimates are customarily obtained
using the least squares principle. Let a and /? denote estimates of a and (3. Then yt = a +
/be, 4- e{ where e, denotes the ith residual yt —a — ^xt. The least squares principle minimizes
the sum of the squared residuals

i=i
= Z O'* - « - hd2
i=i

Differentiating partially with respect to a and

Mtdh -2Zx,(n-*-**'*
364 / Differential Calculus: Functions of More Than One Variable

Equating the partial derivatives to zero,

Z .V; - not - fifjXi = 0

i= 1 £= 1

Zxiyi -
i=1 £= 1
~/?Zx,2 = o
£= 1

Multiplying the first equation by Z?=i xi and the second equation by n and solving the
equations simultaneously,

*£)>£- Zxi
L_\£= 1_
n in \ 2

nZx?- Z*<
£= 1 \i= 1 /

tv,-fit*
~ £= 1 i= 1
a =-
n

are the least squares estimates for the regression coefficients a and /?.
To show that this solution minimizes Z?=i eh the second order condition for a minimum
must be considered.

Z ei) = 2n

lef)=2l xf
dP2\M i=1

d2
dk dp\i Z
= i ei\ =2 Z xi
i= 1

A = (2n){2 Z"=i x?) ~ (2 Z"=i x<)2 > since n Z«=i ~ (Z"=i x<)2 > 0 and

decS.2
lef
,i= 1
>0 I>,2 >0

Thus a, ^ minimize Z^i and are the least squares estimators of a and (3.

EXAMPLE

A manufacturer has a demand of D units per period T. The storage cost per period per unit is
c2 dollars, the shortage cost is c3 dollars per unit short per period, and it costs cx dollars each
time a production run is started (setup cost). Assuming that production orders are filled with
no delay and demand is at a uniform rate, determine the frequency with which production
should be scheduled, and the quantity which should be produced in each run in order to
minimize the total per period inventory cost K. In Figure 3.7, t — interval between production
orders, q = production quantity, z = planned inventory at completion of a production run,
tl = time (during cycle) when stock is on hand, and t2 — time (during cycle) when no stock is

on hand. Note that shortages are assumed to be backlogged and not lost.

h z h
Note that
t q' t D’

setup cost: cx

c2t iZ C2tz2 c2 Tz2

carrying cost per run:
2 2 q 2D
 3.4 Maxima and Minima of Functions of Two Variables / 365

Figure 3.7

Cjtijq - z) _ c3 t(q - z)2 _ c3 T(q - z):

shortage cost per run
2q 2D

D
since the number of runs per period is —, the total inventory cost per period is
q
r _
CjD + C2 Tz2 + c_3 T(q - z):
V —
q 2q 2q

and, since C is a function of both q and z, partial derivatives must be obtained and set equal to
zero to determine a minimum

dC c2 Tz c3 T(q — z)
dz q q

If ~ = 0,
dz

c2 Tz = c3 T(q - z)

c^q
z =
c2 + c3

dC cxD c2Tz2 + c3T c3Tz*

dq 2T 2T

c±D c2 Tz2 c_3 T(q2 - z2)

2q1 2 q1

if^ = o,
dq

c3 T(q2 — z2) = 2cj D + c2 Tz2

2c i D + c2 Tz2 + c3 Tz2
q2 = c3T

2ci D c2 + c3 ,
+ - z2
c,T Cl

c2q
Substituting z —
C2 + c3

2c i D c3
c3 T(q2 - z2) = +
c3T c2 + c3
366 / Differential Calculus: Functions of More Than One Variable

thus

2 ClD
q2 =
(1 1|c3r
\ c2 T C3J

/ 2ci(c2 + c3)D
q- c2 c3 T

£3 2cx c3D
q
Cl + ^3 Ci(c2 + c3)T

t = Tq = /2c1(c2 + c3)T
D \/ c2c3D

The manufacturer should produce q units at intervals of time r, where q and t are determined
by the formulas above.

PROBLEMS
Determine the maxima, minima, and saddle points (if any) for each of the following functions.

1. g(x, y) = 2x2 — 2xy + y2 + 5x — 3y

2. h(x, y) = 3 + 2x + 2y — 2x2 — 2xy — y2
3. z(x, y) = xy + x — y
4. /(x, y) = x2 + xy + y2 - 6x + 2
5. u(x, y) = 4x + 2y — x2 + xy — y2
6. v(x, y) = x2 + y2 — 2x + 4y -f 6
7. z(x, y) = x3 - 3bxy + y3
8. w(x, y) = x2 — 2xy + 2y2 — 2x + 2y + 1
9. g(x, y) = x2 — y2 - 2x + 4y + 6
10. h(x, y) = x2 4- 2xy
11. z(x, y) = xy - 2y2
12. /(x, y) = axy

Find the quantities and prices that maximize profit and the maximum profit for each of the
following sets of demand functions and joint-cost functions for two commodities. Quantities
are denoted by x and y and the corresponding prices are denoted by p and q, respectively; total
cost is denoted by C.

13. x = 1 — p + 2q 17. p = 26 — x
y = 11 + p - 3q q = 40 — 4y
C = 4x + y C = x2 + 2xy + y2
14. x = 11 — 2p — 2q 18. p = 35 - 2x2
y = 16 — 2p — 3q q = 20 —y
C — 3x + y C — 16 — 2x3 -I- xy + 30x 4- 12y + \x2
15. p = 40-2x2 19. p = 40 - 5x
q = 12 — 3y q = 30 - 3y
C = 8 + 4x -I- 3y C = x2 + 2xy + 3y2
16. p — 16 — x2 20. p = 28 — 3x2
q = 9 - y2 9 = 56 - y2
C = x2 + 3y2 C = 2x2 + y2
21. Determine the maximum profit if the production function is z = 20 - x2 + lOx -
2y2 + 5y, the prices of the inputs x and y are 2 and 1, respectively, and the price of the
output z is 5.
22. Determine the maximum profit if the production function is z = 10 - 2x2 + xy -
y2 + 5y, the prices of the inputs x and y are each 3, and the price of the output z is 6.
 3.5 Maxima and Minima Subject to Constraints / 367

23. The Portland Company has a contract to furnish 1350 cement mixers annually. The yearly
storage cost is $40 per cement mixer, the shortage cost is $50 per cement mixer short per
year, and it costs $150 to start a production run. If production orders are filled without
delay and demand is at a constant rate, determine the frequency with which production
should be scheduled and the quantity which should be produced in each run in order to
minimize the total average annual cost.
24. The Filaway Company has a contract to supply 600 filing cabinets at a uniform rate
during a 9-month period. The storage cost during this period is $30 per cabinet, the
shortage cost is $60 per cabinet, and it costs $20 to start a production run. If production
occurs at a constant rate of 2400 cabinets per 9-month period, determine the frequency
with which production should be scheduled and the quantity which should be produced in
each run in order to minimize the total average annual cost.

Answers to Odd-Numbered Problems

1. minimum (— 1, \) 17. P = 21
3. saddle point (1, — 1) q = 28
5. maximum (^, f) x = 5
7. minimum (b, b) y =3
9. saddle point (1, 2) ^max = 125
11. saddle point (0, 0) 19. p = 25
13. p = 14 q = 24
q= 8 x = 3
x = 3 y=2
y = 1 ^max = 90

Fmax = 37 21. = 229f for X =

-Pmax and y = §
if

15. p = 28 23. Schedule production 10 times a year,

q
/7 —
— t
15
producing 135 cement mixers each run.
X

y=i
Pmax = 24VX6 -

(approximately 57.53)

3.5 Maxima and Minima Subject to Constraints

In many practical applications of maximization and minimization, the problem is to
maximize or minimize a given function subject to certain side conditions or con¬
straints on the variables involved. These constraints may be stated as equalities or
inequalities.
For example, if a manufacturer produces two outputs, he may want to minimize
the joint cost, while producing a specified minimum total amount; a company may
want to maximize sales resulting from use of two advertising media, while keeping
total advertising costs within a specified budget constraint; an engineer may want to
minimize frequency of breakdowns, while keeping total cost of two types of preven¬
tive repairs from exceeding a specified amount; a consumer may want to maximize
utility derived from the consumption of commodities, subject to his budget
constraint; a chemist may want to minimize total sedimentation in a two-stage
process, while keeping the temperatures in the two stages equal.
In some cases equations obtained from the constraints can be substituted into the
function to be maximized or minimized; the problem is thus reduced to one involv¬
ing unconstrained maxima or minima and can be solved by the methods of the
preceding section. However, this procedure is not always feasible, particularly if
368 / Differential Calculus: Functions of More Than One Variable

the function to be maximized or minimized involves more than two variables and
several constraints. The methods discussed in this section are applicable in all cases
and can be generalized for use with any number of variables and constraints.
The method of Lagrange multipliers is used in obtaining maxima or minima of
functions subject to equality constraints, and this method is easily generalized for the
case of one inequality constraint. Maxima or minima subject to multiple equality and
inequality constraints are obtained using the Kuhn-Tucker conditions.

LAGRANGE MULTIPLIERS
The most widely used method of obtaining maxima or minima of functions subject to
equality constraints is Lagrange multipliers. Suppose /(x, y) is to be maximized or
minimized subject to the constraint g(x, y) = 0. Form the objective function

F(x, y, X) =f(x, y) - Xg(x, y)

where X, the Lagrange multiplier, is an unknown. Differentiate F(x, y, X) partially

with respect to x, with respect to y, and with respect to X, and set the results equal to
zero; then the three equations

fJ±-xf = 0
OX OX OX

8F _ _ 3 dJL = n
dy dy dy

8F , . „
Yx = y>= 0
dF
can be solved for the three unknowns x, y, and X. Note that = g(x, y) = 0 is the
Jx
constraint, so F(x, y, X) really need be differentiated partially only with respect to x
and with respect to y. In many cases the values of the X's (Lagrange multipliers) are
not of interest and are not found; for this reason they are sometimes referred to as
“ undetermined multipliers.” The Lagrange multiplier may be preceded by a plus sign
rather than by a minus sign and this form is used by some authors. The only resulting
change in the solution is a change in the sign of X.
The logic of the method of Lagrange multipliers is quite straightforward. The
problem is to determine possible maximum or minimum values of a function that
will satisfy a constraint. If it can be assumed that the points being considered satisfy
the constraint, this problem reduces to the problem of finding an unconstrained
maximum or minimum. Essentially, the method of Lagrange multipliers screens the
possible maximum or minimum values to obtain only those values that satisfy
the constraint.
Note that, if the constraint is satisfied, g(x, y) = 0 and Xg(x, y) = 0, regardless of
the value of X. In that case, the objective function becomes the unconstrained func¬
tion /(x, y). By considering X to be an additional variable and setting the partial
derivative with respect to X equal to zero, the constraint g(x, y) = 0 is incorporated
into the problem and the possible maximum or minimum values satisfy the
constraint.
Solution of the three equations above provides the critical points of the con¬
strained function. These critical points satisfy the constraint, but must be tested as
maxima or minima of the function by a procedure similar to that for unconstrained
maxima or minima.
 3.5 Maxima and Minima Subject to Constraints / 369

For a critical point x = a, y = b,

dF
0
dx x = a, y = b
If
dF
= 0
sy x = a, y — b

d2F (d2F d2F

and A* =
dx2 x = a, y = b) dy1 x = a, y — b) dx dy x = a, y-b)

, .r d2F ^ J d2F ^
maximum at x = a, y = b, if —-^ < 0 and <0
dx4 dy4
A* > 0
then . . d2F d2F
minimum at x = a, v = b, if —T > 0 and ——y > 0
dx1 dyz

A* < 0 test fails, function must be investigated near x = a, y = b

Note that for unconstrained maxima and minima, if A < 0, the critical point is
neither a maximum nor a minimum; however, for constrained maxima and minima,
if A < 0, the critical point may in fact be a maximum or minimum. This corresponds
to the fact that a point may be a maximum or minimum of the constrained function,
although it is not a maximum or minimum of the unconstrained function.
These conditions can be stated more concisely as follows.

necessary condition for a critical point: fx = 0, fy — 0

determination of constrained maxima or minima:

_ 2 (maximum iffxx < 0 Jyy < 0

“v> xv | minimum if fxx > 0, fyy > 0
fxx fyy —fly<0=> test fails and the function must be investigated near the critical
point
where the partial derivatives are evaluated for the critical
point

note: The method of Lagrange multipliers can be extended to a function of n

variables,/(xl5 x2,...» *„), subject to the k constraints gj{xl, x2,x„) = 0,/' = 1,2,
..., k, where k < n. Then
k

F{x i, . . . , Xn , A j, . . . , j (Xj, . .., XM) ^ Xj Qj(x i, . . . , Xn)

j= 1
and partial differentiation results in n + k equations to be solved for n + k un¬
knowns. The method of testing critical points as maxima or minima can also be
extended for the general case.

EXAMPLE

Find the maxima and minima (if any) of/(x, y) = 5x2 -I- 6y2 — xy subject to the constraint
x + 2y = 24.
370 / Differential Calculus: Functions of More Than One Variable

F(x, y, X) — 5x2 + 6y2 — xy — X(x + 2y — 24)

dF
= 10.x — y — X
dx

dF
— — 12 y — x — 2X
ly

dF
= x + 2y — 24
dl

Eliminating X from the first two equations,

20 x — 2y — 2/1 = 0

— x + 12 y — 2A = 0

21x - I4y = 0

3x — 2y = 0

Solving simultaneously with the third equation,

x + 2y = 24

4x = 24

x = 6

Critical point: (6, 9); X = 51.

d2F
= 10
dx2

d2F
= 12
dy2

d2F
= -1
dx dy

A* = (10)( 12) - (-1)2 = 119

d2F d2F
> 0 and > 0
dx2 dy2 %

so (6, 9) is a minimum.

Alternative method: f (x, y) = 5x2 + 6y2 — xy, x + 2y = 24.

G(y) = 5(24 - 2y)2 + 6y2 - y(24 - 2y)

= 2880 - 480y + 20y2 + 6y2 - 24y + 2y2

= 2880 - 504y + 28y2

dG
= -504 + 56y
dy

= 0 if 56y = 504

y = 9

x = 6
d2G
= 56 >0
dy2

so (6, 9) is a minimum.
 3.5 Maxima and Minima Subject to Constraints / 371

EXAMPLE

Find the maxima and minima (if any) of/(x, y) — 12xy — 3y2 -■ x2 subject to the constraint
x + y = 16.

F(2, x, y) = 12xy - 3y2 — x2 — A(x + y — 16)

= 12 y — 2x — 2
ox

dF
— = I2x — 6y — A
dy
k«i |*N

3"
+
II
1

Eliminating A from the first two equations,

14x - 18y = 0

7x — 9y = 0

Solving simultaneously with the third equation,

7x + 7y = 112

16y =112

y = 7

x = 9

Critical point: (9, 7); X = 66.

d2F
3x2 " 2

d2F
dy2

dlF = 12
dx dy

A* = (— 2)( — 6) - (12)2 = -132 <0

so the function must be investigated near (9, 7).

(x + /i) + (y + fc) = 16

9 + /z + 7 + /c=16

k = -h

/(9 + h, 7 - h) -/(9, 7) = 12(9 + h)(7 - It) - 3(7 - n)2 - (9 + h)2 -12(9)(7) - 3(7)2 - 92

= —24* - 12h2 + 42h - 3/i2 - 18* --h2

= -16h2 <0

so maximum is at (9, 7).

Alternative method:/(x, y) = 12xy — 3y2 — x2, x -I- y = 16.

G(y) = 12y( 16 - y) - 3y2 - (16 - y)2

= 192y - 12y2 - 3y2 - 256 + 32y - y2

= - 16y2 + 224y - 256

372 / Differential Calculus: Functions of More Than One Variable

dG
= - 32y + 224
dy

dG
= 0 if 32y = 224
dy

y= ?

x = 9

d2G
= -32
dy2

so maximum is at (9, 7).

EXAMPLE

A factory manufactures two types of heavy-duty machines in quantities x and y. The joint-cost
function is given by

f(x, y) = x2 + 2y2 - xy

To minimize cost, how many machines of each type should be produced if there must be a total
of 8 machines?
The constraint is x + y = 8 and the objective function is

F(x, y, 2) = x2 + 2y2 — xy — 2(x + y — 8)

dF
— 2x — y — 2 = 0
dx

dF
= 4y — x — 2 = 0
dy

dF
= — (x + y — 8)

Eliminating X from the first two equations,

3x — 5y = 0

Solving simultaneously with the third equation,

3x + 3y = 24

8y = 24

y= 3

x = 5

2 = 7

d2F
2
dx2

d2F
= 4
dy2

d2F
= -1
dx dy

A = (2)(4) —( —1)J >0

 3.5 Maxima and Minima Subject to Constraints / 373

—y > 0 and —y > 0

dx ay

so (5, 3) is a minimum.

note: Assuming nonnegative cost, the unconstrained minimum of the function

/ (x, y) = x2 + 2y2 — xy occurs at (0, 0).

Alternative method:/(x, y) = x2 + 2y2 — xy, x + y = 8.

G(x) = x2 + 2(8 — x)2 — x(8 — x)

= x2 + 128 — 32x + 2x2 — 8x + x2

= 4x2 - 40x + 128

dG
= 8x — 40 = 0 if x = 5
dx

y = 3

d2G
= 8 > 0
dx2

so (5, 3) is a minimum.

EXAMPLE

The relationship between sales S and the amounts x and y spent on two advertising media is
given by

200x lOOy
5 + x + 10 T y

Net profit is 5 of sales minus the cost of advertising. The advertising budget is 25; determine
how it should be allocated between the two media in order to maximize net profit.
The constraint is x + y = 25 and the objective function is

1 200x lOOy
F(x, y, A) — x — y — A(x + y — 25)
5 5 + x + 10 + y

40^ 20y
+ — x — y — A(x + y — 25)
5 + x 10 + y

SF _ 40(5 + x) - 40x 200

dx ~ (5 + x)2 ’ (5 + x)2

dF 20(10 + y) — 20y 1 . 200 t .

,.n - ,r '
dF
-=-(x + y- 25) = 0

Eliminating A from the first two equations,

200 _ 200
(5 + x)2 ~ (10 + y)2

(5 + x)2 = (10 + j-)2

Solving simultaneously with the third equation,

x + y = 25
374 / Differential Calculus: Functions of More Than One Variable

so

(5 4- 25 — y)2 = (10 4- y)2

30 — y = 10 + y

y = 10

x = 15

X —

d2F 400
= 200( —2)(5 + x)~3 = -
dx2 (5 + x)3

d2F 400
2 = 200( — 2)( 10 4- y)-3 =
dy " [io + yf
d2F
= 0
dx dy

/ -400 \ / -400 \
A = 0 > 0
1(5 + x) \ (io + y)3/

d2F d2F
<0 and <0
dx2 " w dy2

so (10, 15) is a maximum.

note: The unconstrained maximum of the function

l 200x 100^
p(x, y) = — x — y occurs at (10^/2 — 5, 10x/2 — 10).
5 T x 10 4 y
-

, . 40x 20y
Alternative method: P(x, y) = ---f -x — y, x + y = 25.
v yf 5 + x 10 + y

40x 500 — 20x

P(x)
5 -l- x 35 — x

dP _ 40(5 + x) - 40x -20(35 - x) + (500 - 20x)

dx (5 + x)2 + (35 — x)2

200 200
(5 4- x)2 (35 — x)2

200 _ 200
(5 + x)2 (35 — x)2

5 + x = 35 — x

x = 15

y 10

d2P
200( — 2)(5 + x)“3 4- 200( — 2)(35 - x)~3
dx2

400 400
<0
(5 + x)3 (35 - x)3

so (10, 15) is a maximum.

 3.5 Maxima and Minima Subject to Constraints / 375

EXAMPLE

Maximum profit for a given production function and given prices of the inputs and output can
be determined using Lagrange multipliers. Suppose the production function is

16z = 65 - 2(x - 5)2 - 4(y - 4)2

and the unit prices of the inputs x and y are 8 and 4, respectively, and the unit price of the
output is 32. (See the example on page 363.) Profit is maximized if the net revenue function
32z - 8.x — 4v is maximized subject to the production function constraint
16z - 65 + 2(x - 5)2 + 4(y - 4)2 = 0.

F — 32 z — 8.x — 4 y — 2[16z — 65 + 2(.x — 5)2 + 4 (y — 4)2]

dF
- -32-162 = 0
<3z

2 = 2

g=-8-4(x-5)i-0

X = 4

-7 = -4 - 82(y - 4) = 0

y -~4T

As shown previously, profit is maximized when x = 4, y = F is a function of three variables

x, y, and z. The second-order conditions for maximization of functions of more than two
variables, which are somewhat more complex than the conditions for functions of two var¬
iables, are discussed in Chapter 8.

PROBLEMS
Find the maxima and/or the minima of each of the following functions subject to the given
constraint.

1. /(x, y) = 3x2 + 4y2 — xy if 2x + y = 21.

2. /(x, y) = x2 4- y2 - 2xy if x2 + y2 = 50.
3. f (x, y) = x2 — 10y2 if x — y = 18.
4. /(x, y) = 3xy + 4y2 if x2 + y2 = 10.
5. /(x, y) = x 4- y if x2 + y2 = 1.
6. /(x, y) = x2 + 24xy + 8y2 if x2 + y2 = 25.

Answers to Odd-Numbered Problems

1. minimum (8.5, 4)
3. maximum (20, 2)
5. minimum ( — —
maximum (^/f, Jh)

KUHN-TUCKER CONDITIONS
The method of Lagrange multipliers can be modified to determine the maximum or
minimum of a function of two variables subject to one inequality constraint as
follows: Assume that the inequality constraint holds as an equality constraint and
obtain the maximum or minimum using the method of Lagrange multipliers; if
376 / Differential Calculus: Functions of More Than One Variable

X > 0, this maximum or minimum is also the maximum or minimum subject to the
inequality constraint; ifX < 0, the maximum or minimum determined without regard
to the constraint satisfies the constraint and is thus also the constrained maximum or
minimum.
This procedure can be generalized to include multiple inequality constraints, some
of which may be satisfied as equalities and others as inequalities in any particular
problem. In the more general procedure, the conditions necessary for a maximum or
minimum subject to inequality constraints are known as the Kuhn-Tucker condi¬
tions. These conditions are given for the general case in Chapter 8; for the case of a
function of two variables subject to one inequality constraint, the Kuhn-Tucker
conditions are stated as follows.
A point (x, y) is a local maximum of /(x, y) subject to g(x, y) < 0 only if there
exists a nonnegative X such that X and (x, y) satisfy the following conditions:

df (*> y) _ x 8g(x, y) = 0
dx dx

df (x, y) _ dg(x, y) = Q
dy dy

Xg(x, y) = 0

g{x, y) < 0

These conditions are also sufficient if/ (x, y) is concave and the constraint is concave.
Since a maximum point of/(x, y) is a minimum point of —/(x, y), this result is also
applicable to minimizing a convex function subject to a convex constraint.

note: A function/(x, y) is convex in a region if a line segment drawn through

any two points on the surface does not fall below the surface. A function of two
variables is convex if and only if

/[(l - t)x 1 + tx2, (1 - t)y, + ty2] < (1 - t)f(x!, yY) + tf(x2 , y2)
for 0 < t < 1

The function is strictly convex if < can be replaced by < ; the function is concave if
< can be replaced by > and strictly concave if < can be replaced by >.

A second-degree polynomial of the form

/(x, y) = Ax2 + Bxy + Cy2 + Dx + Ey + F

is convex if 4AC — B2 > 0 and A > 0, C > 0; it is concave if 4AC — B2 > 0 and
A < 0, C < 0; and it is neither convex nor concave everywhere if 4AC — B2 < 0.
Note that this result is a reformulation of the conditions for a maximum or minimum
of two variables.
As shown in the following examples, the method of Lagrange multipliers modified
for one inequality constraint and the Kuhn-Tucker conditions for one inequality
constraint give the same solution. The advantage of the Kuhn-Tucker conditions is
that they can be generalized for more than one inequality constraint. Although the
method of Lagrange multipliers can be generalized for more than one equality
constraint, it cannot conveniently be modified for more than one inequality
constraint. If the function to be maximized or minimized is linear and the constraints
are also linear, the solution may be obtained using linear programming for both
equality and inequality constraints.
Note that the following examples are modifications of some of the examples for
equality constraints.
 3.5 Maxima and Minima Subject to Constraints / 377

EXAMPLE

Find the minimum of/(x, y) = 5x2 + 6y2 — xy subject to the constraint x + 2y > 24.
Assuming the inequality constraint to hold as an equality constraint, 2 = 51 (see page 370).
Since 2 > 0, the minimum, x = 6, y = 9, assuming the equality constraint is also the minimum
assuming the inequality constraint.
Using the Kuhn-Tucker conditions,

d_L _ ydJL 10.x — y — 2 = 0

dx dx

Of _ x 12 y — x — 22 = 0
dy dy

Xg - 2(x + 2y — 24) = 0

g = x + 2y — 24 > 0

Either 2 = 0 or x + 2y — 24 = 0. If 2 = 0, then x = y = 0 in order to satisfy the first two

equations, but then x + 2y > 24 is not satisfied. If x + 2y — 24 = 0, then x = 24 — 2y and,
substituting in the first two equations,

240 — 20y — y — 2 = 0

12y - 24 + 2y - 22 = 0

2 + 21 y = 240

-22 + 14y = 24

56y = 504

y = 9

x = 6

Thus, as above, the minimum of/(x, y) = 5x2 + 6y2 — xy subject to x + 2y > 24 is x = 6,

y = 9.

note: 5x2 + 6y2 — xy is convex, since 4AC — B2 = (4)(5)(6) — (— 1 )2 > 0 and A > 0,
C > 0.

EXAMPLE

Find the maximum of/(x, y) = 12xy — 3y2 — x2 subject to the constraint x + y < 16.
Assuming the inequality constraint to hold as an equality constraint, 2 = 66 (see page 371).
Since 2 > 0, the maximum, x = 9, y = 7, assuming the equality constraint is also the maximum
subject to the inequality constraint.
Using the Kuhn-Tucker conditions,

- 2 CA = 12y — 2x — 2 = 0
ex dx

C[ - X ^ = \2x - 6y - X = 0
dy dy

Xg — 2(x + y — 16) = 0

g = x + y — 16 <0
378 / Differential Calculus: Functions of More Than One Variable

Either 2 = 0 or x + y - 16 = 0. If X = 0, then x = y = 0 in order to satisfy the first two

equations, which satisfies x + y < 16; /(0, 0) = 0. If x + y — 16 = 0, then x = 16 — y and,
substituting in the first two equations,

12 y — 32 + 2y — X = 0

192 - 12y - 6y - 2 = 0

14 y-X = 32

18 y + 2 = 192

32y = 224

y=?
x = 9

The solution x = 9, y = 7 satisfies 12y — 2x — 2 = 0 and 12x — 6y — 2 = 0 if X — 66./(9, 7) =

528, which is larger than /(0, 0) = 0. Thus, as above, the maximum of /(x, y) = 12xy —
3y2 — x2 subject to x -f y < 16 is x = 9, y = 7.
note: 12xy — 3y2 — x2 is concave, since 4AC — B2 = 4( — 1)(— 3) — (12)2 < 0 and A < 0,
C < 0.

EXAMPLE

A factory manufactures two types of heavy-duty machines, x and y. The joint-cost function is
given by

f(x, y) = x2 + 2y2 - xy

To minimize cost, how many machines of each type should be produced if there must be a total
of at least 8 machines?
Assuming the inequality constraint to hold as an equality constraint, 2 = 7 (see page 372).
Since X > 0, the maximum, x = 5, y = 3, assuming the equality constraint is also the maximum
subject to the inequality constraint.
Using the Kuhn-Tucker conditions,

df_
= 2x — y — X = 0
dx

d]_ _ dg
1
= 4y — x — 2 = 0
dy dy

Xg = 2(x + y — 8) = 0

g = x A- y — 8>0

Either 2 = 0 or x + y — 8 = 0. If 2 = 0, then x = y = 0 in order to satisfy the first two equa¬

tions, but then x -f y > 8 is not satisfied. If x + y — 8 = 0, then x = 8 — y and, substituting in
the first two equations,

16 — 2y — y — 2 = 0

4y — 8 T 2 = 0

3y + 2 = 16

5y — 2 = 8

8y = 24

y = 3
 3.5 Maxima and Minima Subject to Constraints / 379

Thus, as above, the minimum of/(x, y) = x2 -I- 2y2 — xy subject to x + y > 8 is x = 5, y = 3.

note: x2 + 2y2 — xy is convex, since 4AC — B1 = 4(1 )(2) — ( — l)2 >0 and A > 0,
C > 0.

EXAMPLE

The relationship between sales S and the amounts x and y spent on two advertising media is
given by

_ 200x lOOv
5 = -+ — —

5 + x 10 + y

Net profit is 5 of sales minus the cost of advertising. The advertising budget has a maximum of
20; determine how it should be allocated between the two media to maximize net profit.

Assuming the inequality constraint to hold as an equality constraint, k = — \ (see page

373). Since k < 0, the unconstrained maximum is also the maximum subject to the inequality
constraint.
1 200x lOOy
/(*, y) = + - x - y
5 + x 10 + y

df = 40(5 + x) - 40x
- 1
dx (5 + x)2

df 20(10 + y)-20y
- 1
dy (10 -f y)2

(5 + x)2 * = 200

x = 10v72 - 5

(10 + y)2 = 200

y = 10./2 - 10

Using the Kuhn-Tucker conditions,

dj_ _ k = 40(5 + x) - 40x _ { ^ = Q

dx dx (5 + x)2

Of _ A dg = 20(10 + y) - 20 y _ _ ^ =Q
dy dy (10 + y)2

kg — k(x + y — 25) = 0

g(x, y) = x + y — 25 <0

200x lOOy
note: /(x, y) = ~ - x -y
5 + x 10 + y
1

Either k = 0 or x + y — 25 = 0. If k — 0, then in order to satisfy the first two equations,

40(5 + x) — 40x = (5 + x)2

(5 + x)2 = 200
1
x = 10^/2 - 5

20(10 + y) — 20y = (10 + y)2

(10 + y)2 = 200

y = 10^2 - 10
380 / Differential Calculus: Functions of More Than One Variable

x = l0y/2 — 5, y = 10sfi — 10, satisfies x + y < 25. /(10v/2 — 5, 10^/2 — 10) = 75 —

40^/2 « 18.44. If x + y — 25 = 0, x = 25 — y, and (see page 374)

(5 4- 25 - y)2 = (10 + y)2

30 — y = 10 + 4

y = 10

x = 15

The solution x = 15, y = 10, satisfies

200
— 1— 2=0
(5 + x)2

200
(10T w -1 - 2=0

if 2 = /(15, 10) = 15, which is less than/(10^/2 — 5, 10^/2 — 10) = 75 — 40N/2. Thus, as
above, the maximum of

200x lOOy
f(x, y) = ^ — x — y
5 + x + 10 + y

subject to x + y < 25 is x = 10^/2 — 5, y = 10^/2 — 10.

3.6 Applications of Constrained Maximization

and Minimization in Business and Economics
There are many applications of constrained maxima and minima in business and
economics, as illustrated in the examples of the preceding section. One of the most
important applications of constrained maximization in economics occurs in utility
theory; the consumer maximizes the utility derived from the purchase of two com¬
modities, where the consumer’s fixed income and the prices of the commodities
constitute a budget constraint. The problem of constrained utility maximization is
considered in this section.

CONSTRAINED UTILITY MAXIMIZATION

If his purchases are limited to two commodities Ql and Q2, the rational consumer
wishes to purchase that combination of QA and Q2 which will maximize his utility,
subject to his budget constraint. Since his purchases are limited to the two commodi¬
ties, the consumer’s budget constraint can be written

/ = Piqi + p2q2
where y° is his fixed income and p{, p2, qx, and q2 are the prices and quantities of
and Q2, respectively.
The budget line y° = pl qx + p2q2 shows all possible combinations of qy and q2
which the consumer can purchase, given his budget constraint. In order to maximize
his utility, the consumer will choose to purchase a combination of qx and q2 lying on
the highest possible indifference curve. As shown in Figure 3.5, this maximizing point
 3.6 Applications of Constrained Maximization and Minimization / 381

is the point at which the budget line is tangent to an indifference curve. Any move¬
ment from this point along the budget constraint puts the consumer on a lower
indifference curve and results in diminished utility.
Mathematically, the problem is one of obtaining a constrained maximum:

maximize U(ql, q2)

subject to y° = p1ql + p2q2

The budget constraint can be written

/ - P i<?i

and the utility function, as a function only of qy, is

y° ~ PiqA
U=f 4i>
Pi )

Note that q2 is determined if qx is specified, so it is sufficient to maximize U with

respect to qx

dU
—/i T f 2 ph=o
dqi pi!

and

fi = Pi
h Pi

note: The utility function may be written

U=f(q!,q2) where q2 = --= g{qi)

P2

Then

du_ _ sf ^qi + ^y, dqi

dqi dq, dqt Sq2 dqy

= fl+f2i~Ppi)

Thus, for a maximum, the ratio of the marginal utilities must equal the ratio of
the prices, or equivalently,

fi = fi
Pi Pi

that is, marginal utility divided by price must be the same for both commodities. This
condition is necessary, but not sufficient, for occurrence of a maximum. The second-
order condition for a maximum requires that

d2U
< 0
dq]
382 / Differential Calculus: Functions of More Than One Variable

But
d2U d ldU\ldqy\ d idU\ldq2
dq\ dq^dqjydqj dq2\dql)\dql
Pi Pi
fll +/l2 + /l2(~L| f22
Pi Pi,
Pi
= /ll + ?/i 2 ( —
P2/ \ P2)
The second-order condition for a maximum requires that

/n + 2/12(-^J+/22(-^J2<0

or

fnPi ~ ?fnPiPi T fnPi < 0

In Technical Note I, it is shown that the second-order condition for a maximum

A f
requires that the utility curves are convex from below. Since —- = —/ =-- and
dqi f2 Pi
prices are positive, the indifference curves are negatively sloped. Thus if the utility
function has a maximum, the indifference curves are of the general form shown in
Figure 3.5.

EXAMPLE

If u = qi q2, Pi = 15, p2 = 5, and the consumer’s income for the period is 150, the budget
constraint is

150 — I5qi — 5q2 = 0

150 - I5qi
= = 30 - 3q,

and
U = 30qi - 3q\

dU
= 30 — 6qi
dq i
dU d2U
If — = 0, then qi = 5, q2 = 15, and —r = —6, so the consumer maximizes utility subject
dqi dq\
to his budget constraint by purchasing 5 units of Qx and 15 units of Q2. Note that

fi = qi = 15 = Pi
h qi 5 p2
as required for a maximum.

EXAMPLE

If U — qiqi — 10qr2, = 2, p2 = 8, and the consumer’s income for the period is 116, the
budget constraint is
116 — 2qi — 8^2 = 0

116 - 8^2

qi = = 58 - 4q2
2
 3.6 Applications of Constrained Maximization and Minimization / 383

and

U = (58-4q2)qi - 10(58 - 4q2)

= 58q\ - 4ql - 580 + 40</2

dV
— = 116i72 — 14- 40
dq2

Iff = 0,
dq 2

3q\ - 29q2 -10 = 0

(3q2 + l)(q2 - 10) = 0

q 2 = —i, 10

But q2 > 0, so the solution is q2 = 10, qx = 18.

d2U
= 116 — 24q2
dql

dHJ
= 116 - 240 = — 124 < 0
dql q 2 =10

So the consumer maximizes utility subject to his budget constraint by purchasing 18 units of
Qi and 10 units of Q2.
Note that

h = d-jo 90 2 Pi
fi 2^i q2 360 8 Pi

as required for a maximum.

Alternatively, the constrained maximum can be obtained using Lagrange

multipliers

Z =/(4l. - l(y° - Pl <5fl - P2<?2>

SZ , .
— = /, + /Ip, = 0

oq i
az
= /2 + ^f>2 = o
5q2

az
si
= y° -Pi<h - Piqi = o
Thus

2 _ /i _ fi
Pl P2
and

a = Pi

fi Pi

and so forth, as above.

384 / Differential Calculus: Functions of More Than One Variable

PROBLEMS

1. Find the minimum of/(x, y) = 4x2 + 5y2 — 6y if x + 2y > 18.

2. Find the maximum of/(x, y) = 16x + 12y — 2x2 — 3y2 if x + y < 11.
3. Find the minimum of/(x, y) = 3x2 + 3y2 if x + y > 10.
4. Find the minimum of/(x, y) = 12x2 + 4y2 — 8xy — 32x if x + y > 1.
5. Find the maximum of/(x, y) = lOxy — 5x2 — ly2 + 40x if x + y < 13.
6. Find the maximum of/(x, y) = 6xy — 3x2 — 4y2 if 3x + y < 19.
7. Production, P, as a function of two inputs x and y, is given by

P = x2 + 5 xy — 4 y2
Find the amounts of x and y which maximize production if a. 2x + 3y — 74, b.
2x + 3y < 74.
8. Sales, S, as a function of the amounts, x and y, spent on two types of promotion is given by

240x 150y
“ 25 + 3x + IU • r
Net profit = tuS - x - y. Find the allocation of x and y which will maximize net profit if
x + y = 15.
9. Cost of production, C, as a function of the numbers produced, x and y, of two types of items
is given by

C = 6x2 + 3y2
To minimize cost, what numbers of the two items should be produced if a. x + y — 18, b.
x + y > 18?
10. The cost of repairs, C, as a function of the numbers, x and y, of inspections at two points in
a process is given by

C = 2x2 + 3y2 + xy — 22x + 5

To minimize repair cost, what number of inspections should be made at each point if
x - y = 2?
11. The number of breakdowns, N, as a function of the numbers, x and y, of replacements of
two parts of a machine is given by

N = 3x2 + y2 + 2xy — 22x + 6

To minimize breakdowns, what numbers of replacements should be made for each part if
2x = y?
12. The cost of repairs, C, as a function of the numbers, x and y, of inspections at two points in
a process is given by

C = 4x2 + 2y2 + 5xy — 20x + 30

To minimize repair costs, what number of inspections should be made at each point if the
total number of inspections is a. 10, b. not fewer than 10?
13. Using Lagrange multipliers, determine the maximum profit if the production function is
z = 20 - x2 + lOx — 2y2 + 5y, the prices of the inputs x and y are 2 and 1, respectively,
and the price of the output is 5. (See Problem 21, page 366.)
14. Using Lagrange multipliers, determine the maximum profit if the production function is
z = 10 — 2x2 + xy — y2 + 5y, the prices of the inputs x and y are each 3, and the price of
the output is 6. (See Problem 22, page 366.)
15. If the consumer’s utility function is U = q\ q2, Pi = 4, p2 = 5, and y° = 120, determine the
quantities and q2 which he should purchase in order to maximize his derived utility.
16. If the consumer’s utility function is U = qx q2 - q\, pi = 3, p2 = 6, and y° = 90, determine
the quantities ql and q2 which he should purchase in order to maximize his derived utility.
17. If the consumer’s utility function is U = q\ + 2q2 + 5q{q2, pl = 5, p2 = 10, and the
consumer’s income for the period is 90, determine the quantities q^ and q2 which he should
purchase in order to maximize his derived utility.
 3.6 Applications of Constrained Maximization and Minimization / 385

18. If the consumer’s utility function is U = q{ q2 — 3q2, P\ = 10, p2 — 15, and the consumer’s
income for the period is 180, determine the quantities q1 and q2 which he should purchase
in order to maximize his derived utility.

Answers to Odd-Numbered Problems

1. (4, 7) li. x = l, y = 2
3- (3,5) 13. Pmax = 229f,x = ^,y = f
5. (8, 5) 15. qx = 20, q2 = 8
7. a. x = 31, y = 4 17. q1=2r, q2=l
b. x = 31, y = 4
9. a. x = 6, y = 12
b. x = 6, y = 12

Technical Note I. Proof of Convexity of Utility Curves

d2q2 . dq2
Obtain the total derivative ——T by further differentiating -—
dq^ dqj fi
1_

d2q2 ' a dq2V ‘ 5 (dq i\ oq2

rH

+
dq] pq\ \dqi) a<ji ,<3<?2 \dqj 3qK
But

fdqi\1 _ 8 | fi 1 fi fi fi 1
8 1 l fl\
dql 1l dqj1 8qdi fi) fl
H
II
t-H

<dq2\|_ 5 |f M f11 fl ~ fi fi 1
5 1
dqf\dqi)1 BqJl fj fi

Sq2 _ fi
Sqi fi

Thus

h
fllfl flfl2 + (fllfl fl fll) I
d2g2 fl,
dq] f:
1
— j-i (fufi If 12 fi fi + fn fi)
fi

Substituting/j = (pjp2)lf2
d2q2 1 2 Pi , , f2Pi
,31/n/l ~2fl2 fl^+fn f i 2
dq] J 2 \ Pi Pi

1
(fnPi ~ 2/12P1P2 +/22P1)
f2Pi

But the second-order condition for a maximum requires that

fnPi - 2/12P1P2 +/22P1 < 0

386 / Differential Calculus: Functions of More Than One Variable

and, since f2 > 0,

d2q2
>0
dq\
and the indifference curves are convex from below.

MU -
Integral
Calculus

4.1 Introduction
Integration has two distinct interpretations; integration is a procedure which is the
inverse of differentiation, and it is a method of determining the area under a curve.
There are many important applications of integration in each of these contexts.
As an operation, integration is the inverse of differentiation. Thus if a function is
differentiated and the resulting function is then integrated, the result is the original
function. As discussed below, this is precisely true only if the constant of integration
is specified in some way; otherwise the result may differ from the original function by
a constant. In this context, integration is the process of finding a function when its
derivative (or rate of change) is known. In economics, integration can be used to find
a total cost function when the marginal cost function is given, to find a total revenue
function when the marginal revenue function is given, and so forth.
Integration can also be defined as the process of finding the limiting value of a sum
of terms when the number of terms increases infinitely and the numerical value of
each term approaches zero. It is in this context that integration is interpreted as
finding the area under a curve. In fact, integral calculus was developed for the
purpose of evaluating areas by supposing them to be divided into an infinite number
of infinitesimally small parts whose sum is the area required; the integral sign is the
elongated S used by early writers to indicate “sum.” In economics total revenue can
be evaluated as the area under the marginal revenue curve, consumer’s surplus and
producer’s surplus can be evaluated as areas under demand and supply curves, and
so forth.
For either application, integration requires operationally that a function be
determined when its derivative is given. Unfortunately, the techniques of integration
are inherently more difficult than those of differentiation and there are functions,
some of them deceptively simple in appearance, whose integrals cannot be expressed
in terms of elementary functions. The simpler cases of integration are accomplished
by reversing the corresponding formulas for differentiation; more complicated cases
are handled by the use of tables of standard forms, by various procedures of substitu¬
tion, and, if necessary, by numerical (approximation) methods.
The two interpretations of integration and some of their applications in business
and economics are discussed first, using functions that are easily integrated by revers-
387
388 / Integral Calculus

ing the formulas for differentiation. Subsequent sections discuss special methods
useful for integrating more difficult functions and several approximate methods of
integration. Multiple integration, the inverse of partial differentiation, is also defined
and illustrated.

4.2 Indefinite Integration

The process of determining a function whose derivative is known is called integration
and the required function is called an integral or an antiderivative of the given
function. Integration of a function is defined and illustrated in this section.
If F(x) is an integral with respect to x of the function /(x), the relationship
between F(x) and f(x) is expressed as follows:

| /(x) dx = F(x) 4- C

where the left-hand member is read “ integral of/of x with respect to x.” The symbol
j is an integral sign,f(x) is the integrand, F(x) is a particular integral, C is the constant
of integration, and F(x) + C is the indefinite integral.
Note that if F(x) is an integral of/(x) with respect to x, then F(x) + C is also such
an integral, where C is any constant, since the derivative of any constant is zero—that
is,

dF(x) dC
^[FW + C] =
dx dx

dF(x)
dx

which is /(x) if F(x) is an integral of /(x) with respect to x. Thus a function whose
derivative is given is not completely determined, since the integral includes an arbi¬
trary additive constant, the constant of integration. It is for this reason that the
function j /(x) dx is referred to as the indefinite integral of f(x).
It can be shown that two functions which have the same derivative differ at most
by a constant; thus if F(x) is an integral of/(x), all integrals of/(x) are included in
the set F(x) + C, where C is any constant. In many applications of integration,
certain information given in the problem, often referred to as an initial or boundary
condition, uniquely specifies the constant of integration.
Geometrically, y = F(x) + C represents a family of curves any one of which can
be obtained by shifting the curve y = F(x) (corresponding to C = 0) through a
vertical displacement C (see Figure 4.1). The curves represented by y = F(x) + C are
parallel to each other in the sense that the slope of the tangent to any one of them at
the point having abscissa x is /(x). Thus this family of curves has the property that,
given any point (x0, y0), there is one and only one curve of the family which passes
through this particular point. For the curve to pass through the point, its equation
must be satisfied by the coordinates of the point, and this uniquely specifies the value
of C, since then

c = y0 ~ F(xo)
 4.2 Indefinite Integration / 389

Figure 4.1

With C thus determined, a definite function expressing y in terms of x is obtained,

that is, the constant of integration is uniquely determined if one point is specified
through which the curve representing the integral passes. This specification is
referred to as an initial condition, because evaluation of the constant of integration
was first done in connection with problems in mechanics where initial velocities or
positions of moving bodies are specified; however, the terminology is usually appro¬
priate for business or behavioral-science applications as well, since the point most
likely to be specified in such cases is the origin or an intercept.

EXAMPLE

The integral of /(x) = 4x — 3 can be obtained by reversing the appropriate formulas for
differentiation:

| (4x — 3) dx = 2x2 — 3x + C

Note that — (2x2 — 3x + C) = 4x — 3, regardless of the vaue of C. Thus, geometrically, the

dx
integral is represented by a family of parabolas each corresponding to a different value of C.
(See Figure 4.2.) Suppose it is specified, for example, that f(x) = 5 when x = 0; then

2x2 — 3x + C = 5

C = 5

and

F(x) = 2x2 — 3x + 5

which specifies a particular curve of the family given by F(x) = 2x2 — 3x + C. (See Figure 4.2.)
390 / Integral Calculus

Figure 4.2
v

The rules (or standard forms) for integration used in the preceding example are
obtained directly by reversing the corresponding rules for differentiation and can be
summarized as follows:

1. j dx = x 4- C.

2. K dx = K dx, where K is any constant.

3. I (du + dv) = du + dv, where u =f(x) and v = g(x) are differentiable functions
J

of x.
r Xn+1
4. xn dx =-- + C, n — 1.
J n+ 1

u n+ 1
5. I un du — 4~ C, n =£ — 1,
n + 1

where u =f(x) is a differentiable function of x.

EXAMPLE

O v-3

(8x2 + 4) dx = ——h 4x + C (rules 1, 2, 3, 4)

2x512
x3/2 dx = + C (rule 4)
5
 4.2 Indefinite Integration / 391

(3x2 + 2)3x dx = - [ (3x2 4- 2)3(6x) dx (rule 2)

u3 du

l^x2
61 4
+ l+,c
2 (rule 5)

(3x2 + 2)4
4- C
24

Alternatively,

| (3x2 + 2)3x dx = | (27x7 H- 54x5 + 36x3 + 8x) dx

27xf
+ 9x6 + 9x4 4- 4x2 4- C
“ 8~~

81x8 4- 216x6 4- 216x4 4- 96x2

+ C
24

(3x2 4- 2)A
4- C
24

Note that C in the last step is not equal to C in the preceding step, since

(3x2 4- 2)4 = 81x8 + 216x6 4- 216x4 + 96x2 + 16

and the value of C in the last step is thus -jJ less than the value of C in the preceding step. It is
customary to make no notational distinction when the value of a constant of integration is
changed by changing the form of the equation.

EXAMPLE

Evaluate the following integrals.

(a) y = j (x3 4- 2x5/2 + 5x3'2 + lOx) dx if y = 0 when x = 0

= ix4 4- 4x7/2 4- 2x5/2 4- 5x2 4- C

If x = 0, y = 0, then C = 0 and

y — ix4 + 4x7/2 4- 2x5/2 + 5x2

(b) y = | (x 4- 2)2 dx if y = 10 when x = 1

_ (x + 2)3
~ 3

If x = 1, y = 10, then C = 1 and

(c) y — j (3x2 + 2x 4- 6) dx if y = 5 when x = 0

= x3 4- x2 4- 6x + C

If x = 0,y = 5, then C = 5 and

y = x3 + x2 4- 6x + 5
392 / Integra! Calculus

EXAMPLE

For each of the following, find the curve having the given slope and passing through the given
point.

(b) = x2Jx, (1, 0)

hi

x2 - 4
(a) dx = dx dx
x

4x“1 „
— x H —- + C

— x H-h C
X

Thus

4 r
y = x H-h C
x

1 = 4 + 1 + C

C = -4

so

y = x H-4
x

x2 - 4x + 4

(x - 2):

is the required curve.

(b) x2sJx dx = x5/2 dx

x 7/2
7 +c
T

= 4*7/2 + C
Thus

y = 4*7/2 + C
2
0 = 4+
7 C

c= -4
so

y = fx7'2 - 4
= +7,2-i)

= +3 i)
is the required curve.
 4.2 Indefinite Integration / 393

PROBLEMS
Evaluate the following integrals.

1. (X- x 4- 4) dx (1 — 2x)2 dx
9.
Jlx

(2 - 7r)2/3 dt dx
10.
tyflx

3. yiTTy dy 11. j (x^x — 5)2 dx

dx (x3 — 1) dx
4. 12.
(3x 4- 2)2 x — 1
3r dr
5. 13. (2x 4- 3) dx

6. x^Jlx2 4- 1 dx 14. I (x2 — y/x) dx

7. x -\-7=1 dx 15. j (72“+~5y) dy

(z 4- 1) dz
8.
3 ^z2 + 2z 4- 2
dy
16. Find the equation of the curve having slope — = 2x — 5 and passing through the point
dx
(5, 4).
dy
17. Find the equation of the curve having slope — = (x 4- l)(x + 2) and passing through the
dx
point (-3, -§).

18. If — 2x — 3 and y = 2 if x = 3, find the value of y if x = 5.

dx

dp
19. If and p = 2a if x = ?a3, find the value of p if x = 2a5.
dx lax

20. Find the equation of the curve for which/" = 2 and which has a slope of —2 at its point of
inflection (1, 3).

21. Find the equation of the curve for which y" = x3 and which is tangent to the line
2x + y = 5 at the point (1, 3).
22. Find the equation of the curve for which y" = 6x2 and which passes through the points
(0, 2) and (— 1, 3).
23. Find the equation of the curve for which y" = x and which passes through the point (1, 2)
with a slope of f.
24. Find the equation of the curve which has a slope of zero at the point (0, 2), has a point of
inflection at (— 1, ^), and for which y" = 4.

Answers to Odd-Numbered Problems

1. -jx3 — jx312 4- 4x 4- C 9. yix(\ -$x + |x2)+ C

3. A(2 4- 5y)312 + C 11. ix4 - 4x5/2 + 25x + C

5. -3^1 - 71 + C 13. x2 + 3x + C

7. jx312 + 2x'/2 + C 15. A(2 + 5y)3'2 + C

394 / Integral Calculus

17. y = i*3 + \x2 + 2x 21 ^ = 2 + 1

19. p(2a3) = 2>a

23. y = ^x3 + 2x —

4.3 Applications of Indefinite Integration

in Business and Economics
As mentioned in previous discussions of differentiation, in economics the variation of
one quantity y with respect to another quantity x is frequently discussed in terms of
two concepts—average variation and marginal variation. Just as its marginal varia¬
tion can be obtained by differentiating a function, the function (apart from a
constant) can be obtained by integrating its marginal variation. This application of
integration is illustrated for cost, revenue, and consumption functions in this section.
A similar application is illustrated for capital formation.

COST
If the total cost y of producing and marketing x units of a commodity is given by the
function

V = /(*)

Then the average cost per unit is

y _ f(x)
X X

and the marginal cost is

That is, marginal cost is the derivative with respect to x, /'(x), of the total cost
function y —f (x). Thus total cost is the integral with respect to x of the marginal cost
function; that is,

f'(x) dx = f (x) 4- C

To obtain a unique total cost function by integrating the corresponding marginal

cost function, an initial condition must be specified. Frequently this specification is in
terms of a fixed cost or initial overhead, that is, the cost when x = 0.

EXAMPLE

Marginal cost y' as a function of units produced x is given by

y - 1.064 - 0.005x

Find the total and average cost functions if fixed cost is 16.3.

y = [ (1.064 - 0.005x) dx

= 1.064x - 0.0025x2 + C
 4.3 Applications of Indefinite Integration in Business and Economics / 395

If x = 0, y = 16.3, and thus C = 16.3 and

y = 16.3 + 1.064x — 0.0025x2 (total cost)

EXAMPLE

Marginal cost y' as a function of units produced x is given by

y' = 2 + 60x — 5x2

Find the total and average cost functions if fixed cost is 65.

y = | (2 + 60x — 5x2) dx

= 2x + 30x2 — -fx3 + C

If x = 0, y = 65, and thus C = 65 and

y = 65 + 2x + 30x2 — |x3 (total cost)

- = — + 2 + 30x — -fx2 (average cost)

X X

REVENUE
For any demand function

y =/(*)
where y is the price per unit and x is the number of units, total revenue, R, is the
product of x and y; that is,

R = xy = x • /(x)

Marginal revenue with respect to demand is the derivative with respect to x of the
total revenue

Thus the total revenue function is the integral with respect to x of the marginal
revenue function; that is,

R = R (x) dx

And, since

R'(x) dx = R{x) + C

an initial condition must be specified to obtain a unique total revenue function by

integrating the corresponding marginal revenue function. The initial condition that
revenue is zero if demand is zero can be used to evaluate the constant of integration.
Note that average revenue or revenue per unit is the price per unit, y, and thus the
average revenue curve and the demand curve are identical.
396 / Integral Calculus

EXAMPLE

If the marginal revenue function is

R'(x) = 8 — 6x — 2x2

determine the total revenue and demand functions.

R(x) = j (8 — 6x — 2x2) dx

= 8x — 3x2 — §x3 + C

If x = 0, R = 0, and thus C = 0 and

R = 8x — 3x2 — -j*3

y = — = 8 — 3x — -jx2
x

EXAMPLE

If the marginal revenue function is

R'(x) = 12 - 8x + x2

determine the revenue and demand functions.

R(x) = j (12 — 8x + x2) dx

— 12x — 4x2 + ^x3 + C

If x = 0, R = 0, and thus C = 0 and

R = 12x — 4x2 + ix3

y = — = 12 — 4x + 3X2
x

= (6 ~ *)2

NATIONAL INCOME, CONSUMPTION, AND SAVINGS

If the consumption function is given by

c =/(x)
where c is total national consumption and x is total national income, then the
marginal propensity to consume is the derivative with respect to x of the consumption
function

and, assuming x = c + s, where s is savings, the marginal propensity to save is

 4.3 Applications of Indefinite Integration in Business and Economics / 397

Total national consumption is the integral with respect to x of the marginal propen¬
sity to consume,

c = ( f(x) dx =f(x) + C

An initial condition must be specified to obtain a unique consumption function by

integrating the corresponding marginal propensity to consume.

EXAMPLE

The marginal propensity to consume (in billions of dollars) is

dc 0.2
— = 0.7 + —f=
dx ./x

When income is zero, consumption is 8 billion dollars. Find the consumption function.

c = 0.7 + 74 | dx

= 0.7x + 0.4^/x + C

If x = 0, c = 8, and thus C = 8 and

c = 8 + 0.7x 4- 0.4^/x

EXAMPLE

Ihe marginal propensity to save is 3. When income is zero, consumption is 11 billion dollars.
Find the consumption function.

dc ds 2
dx dx 3

= ix + C

If x = 0, c = 11, and thus C = 11 and

c = f x + 11

CAPITAL FORMATION
The process of adding to a given stock of capital is referred to as capital formation. If
this process is considered to be continuous over time, capital stock can be expressed
as a function of time, K(t), and the rate of capital formation is then given by

= K'(t). The rate of capital formation at time t is the same as the rate of net
dt
investment flow at time t, denoted by I(t). Thus

K(t) = [ K'(t) dt = [ I(t) dt

398 / Integral Calculus

and capital stock is the integral with respect to time of the rate of capital formation
or, equivalently, the integral with respect to time of net investment. And, since

f /(f) dt = K(t) + C

an initial condition must be specified to obtain a unique function for capital by

integrating the rate of net-investment flow. For example, if the capital at time zero is
a specified amount, K(0), then this condition can be used to evaluate the constant of
integration.

EXAMPLE

If investment flow is given by 7(f) = 5f3/7 and the initial capital stock at t = 0 is K(0), find the
function representing capital, K.

K = 5t3n dt

= ^10/7 + C

If t = 0, K = C, and thus C = K(0) and

K = it1011 + K(0)

EXAMPLE

If investment flow is given by I(t) = 15r1/4 and the initial capital stock at t = 0 is 30, find the
function representing capital, K.

K = 15t1/4 dt

= 12f5/4 + C

If t = 0, K = 30, and thus C = 30 and

K = 12f5/4 + 30

PROBLEMS
1. If marginal revenue is a constant different from zero, show that price is constant.
2. If R'(x) = 0 and R(0) =/= 0, what is the nature of the demand curve?
3. If marginal cost is constant, show that the cost function is a straight line.
4. The marginal propensity to consume (in billions of dollars) is

dc 0.5
= 0.6 +
dx 2X171

When income is zero, consumption is 10 billion dollars. Find the consumption function.
5. The marginal cost function for production is / = 10 + 24x — 3x* 1 2 3 4; if the (total) cost of
producing 1 unit is 25, find the total cost function and the average cost function.
6. The marginal propensity to save is When income is zero, consumption is 6 billion
dollars. Find the consumption function.
7. If marginal revenue is R! = 15 — 9.x — 3x2, find the revenue and demand functions.
3 2
8. If marginal revenue is K = —r-, find the revenue and demand functions if R(l) = 6.
x x
 4.4 Definite Integration / 399

9. If marginal revenue is R' = 10 — 5x, find the revenue and demand functions.
10. If marginal revenue is R' = 20 — 3x2, find the revenue and demand functions.
11. The marginal propensity to consume (in billions of dollars) is

dc 1
0.5 +
dx 3^

When income is zero, consumption is 6 billion dollars. Find the consumption function.
12. The marginal propensity to save (in billions of dollars) is

dc 1
= 1 - 0.4 -
ds 6x2/3

When income is zero, consumption is 9 billion dollars. Find the consumption function.
13. If investment flow is given by I(t) = 20t3/7 and the initial capital stock at t - 0 is 25, find
the function representing capital, K.
14. If investment flow is given by 7(f) = 25f4/11 and the initial capital stock at t = 0 is 22, find
the function representing capital, K.

Answers to Odd-Numbered Problems

5. y = lOx + 12x2 — x3 -1- 4 9. R = lOx — fx2
4 y = 10 — fx
v = 10 + 12x — x2 H—
x
11. c = 0.5x + 0.5x2/3 + 6
7. R = 15x -fx2 - x3
13. K = 14r10/7 + 25
y = 15 — §x — x2

4.4 Definite Integration

As noted above, finding the areas of figures of various shapes was one of the prob¬
lems that led to the development of integral calculus. In elementary geometry, the
area of a rectangle is shown to be equal to the product of its width and its height, and
from this the areas of other figures bounded by straight-line segments are obtained
by elementary geometrical methods. However, these methods are not directly applic¬
able to figures bounded entirely or in part by curved lines. In general, to find areas of
curvilinear figures, the method of limits must be used; for example, the area of a circle
is obtained in geometry by considering it to be the common limit of sets of regular
inscribed and circumscribed polygons as the number of their sides is indefinitely
increased. This use of the method of limits leads to the interpretation of the definite
integral as the area under a curve. In this section, the definite integral is defined and
illustrated.
Consider the problem of determining the area bounded by the continuous positive
curve y = f(x), the x-axis, and the lines x — a and x = b. Divide the base [a, b] into n
subintervals and denote the points of division by a = xl5 x2, •.., xn, xn+1 = b and
the lengths of the n subintervals as Ax, = xI+ j — xf; i = 1, ..., n. Erect ordinates at
the points of division and inscribe rectangles (see Figure 4.3).
The areas of these inscribed rectangles are/^) Axx,/(x2) Ax2,... ,/(*„) Ax„ and
their sum is

Z f(x i) Axi
i= 1
400 / Integral Calculus

Figure 4.3

By increasing the number of inscribed rectangles, that is, by letting n -* oo and

max Axt -> 0, the area under the curve between a and b which is not included in the
rectangles is decreased and, in fact, approaches zero.
definition : The area bounded by the continuous positive function
y=f(x)
the x-axis, and two fixed ordinates x = a, x = b is

A = lim £ f(x,) Ax*

n-> oo i— 1
max AXi~> 0

note: The same limit would be obtained by circumscribing rather than inscrib¬
ing the rectangles, that is,

lim £ f(x,) Ax; = lim £ f(xl+I) Ax;

n~* oo i= 1 n-+oo i=l
max Ax,->0 max Ax/-* 0

If /(x) is nonmonotonic, the argument is modified by using, respectively, the mini¬

mum or maximum values of the function in the intervals rather than/(x,) or f(xi+l).
In fact, any point in an interval can be used without affecting the result.
If/(x) is continuous for the interval [a, b], then

lim £ f(x.) Axi

n-> oo i= 1
max Axi~> 0

exists,/(x) is said to be integrable over the interval [a, b], and it can be shown that

Area = lim £ /(xt) Ax* = /(x) dx

n-> oo i= 1 Ja
max Axi~* 0

by the following argument.

 4.4 Definite Integration / 401

Figure 4.4
y

Consider the area between a fixed ordinate at x = a and a variable (movable)

ordinate at x = b (see Figure 4.4) and denote the area abgh by A. When x takes a
small increment Ax, A takes an increment AA = area bceg. Note that

Area (bcdg) < Area (bceg) < Area (beef)

(bg)( Ax) < A A < (ce)(Ax)

i AA
bg < —— < ce

and let Ax —► 0. (If it happens that the shape off(x) is such that ce < bg, the inequal¬
ity signs can be reversed without affecting the argument.) Then, since y is a contin¬
uous function of x, ce approaches bg and

dA
— bg = y =f(x)
dx

dA = f(x) dx

Integrating,

Denote j f(x) dx by F(x) 4- C. Then A = F(x) + C. Note that A = 0 when x = a

and thus

0 = F(a) + C

C = -F(a)

and
A = F(x) - F(a)

For the area abgh (Figure 4.4), x = b and

A = F(b) - F(a)
402 / Integral Calculus

which is denoted by
b b
y dx or f(x)dx
Ja Ja

and is read “the integral from a to b of y dx.” This operation is referred to as

integration between limits, a being the lower limit of integration and b being the upper
limit of integration. In this context the word “ limit ” means the value of the variable
at one end of its range; this should not be confused with the meaning of the word in
the theory of limits. In the evaluation of \ba f(x) dx, the constant of integration
disappears; the integral thus has a definite value and, for this reason, is referred to as
the definite integral of f(x) from a to b. From the definition of a definite integral, it
can be shown that definite integrals have the following properties.

1. f(x)dx= — j f(x) dx.

d U

2. f(x) dx = 0.

r /
3. f(x)dx= f(x)dx+\ f(x) dx, where a < c < b

note: For some problems, definite integrals occur for which one or both limits
of integration are infinite. An integral of the form
. b 00 00

f(x) dx f(x) dx f(x) dx

— 00 oo

is referred to as an improper integral and is evaluated by taking limits as follows:

f(x)dx= lim f(x)dx

— 00 a-* — oo a

00 b

J f(x)dx= lim j* f(x)dx

a b~* oo a

oo

f (x) dx = lim /(x) dx

— oo a-* — oo a
b~* oo

If the limit exists, the improper integral converges; if the limit does not exist, the
improper integral diverges.

EXAMPLE

(First Property)

f y(y2 - 4)2 dy (2y)(y2 - 4)2 dy

J1
3
[y2 ~ 4)31
6 1
 4.4 Definite Integration / 403

(y2 - 4):
J y(y2 - 4)2 dy =

= -V--41
152 76
- —s- - —r
and

f y(y2 ~ 4)2 dy — - ( y(y2 - 4)2 dy

J1 J3

(Second Property)

JT- z dz = [-|(i - z)3/2]:J

J -1

= - W2-[-!(2)3/2]
= 0

(Third Property)

1 -V dv = I (v1/2 + v~112) dv
i ji

= tfv3'2 + 2vll2]t
= ie+4_2_2

_
-T
20

4
vV T
+ 1 , r 2 v 4- 1 l> + 1
dv = dv
1 x/V 1 x/ t?

[i
- r2„3/2
«3'2 + 2d1'2]? + [V'2 + 2i>1,2]‘
n

= i(2)3'2 + 23'2 -i-2 + ^ + 4- |(2)3'2 - 23,2

_T
— 20

PROBLEMS
Evaluate the following integrals.

1. (x2 — 2x + 3) dx 8. [ (x2 + l)2 dx

J-i
i
r2o
(v + l)2 dv 9. j (a + z) dz
J-i * /I

„ 2 2 X2
v2 - 1
J. (4x + 1)1/2 dx 10. dx
o i x
. 8
dx
11. (w1/3 — w 1/3) du
o (2x + l)3 i
,“2
5. j f(r + l)2 dt 12. (2x + x2 — x3) dx
J-i
la
6. I ( x2 H—, I dx 13. (a3 + 3ax2 + x3) dx

r3
7. (20 + 1 )(3 — 0) 14. j" (y/a — y/x)2 dx
•'i
404 / Integral Calculus

*T

15. (y/z — z)2 dz 16. 4- x)(3x + 1) dx

Answers to Odd-Numbered Problems

7
1. 1 9. W
13
3. “3“ 11.
5. 13. ^a4
7. 26
i5. a
AREA AS A DEFINITE INTEGRAL
The previous result concerning evaluation of area by integration is known as the
fundamental theorem of integral calculus and can be summarized in somewhat more
rigorous form as follows:
Let /(x) be continuous and positive for the interval x = a to x — b. Let this
interval be divided into n subintervals of lengths Ax1, Ax2, .Ax„ and points be
chosen, one in each subinterval, their abscissas being xl9 x2, .x„, respectively.
Then

lim X /(*,) Ax,. = j f(x) dx = F(b) — F(a)

n-*- oo i= 1
max Ax;->0

The importance of this theorem results from the fact that it permits the evaluation
of the limit of a sum of terms by integration and that this limit can be interpreted as
the area under a curve. Specifically, the definite integral \ba f(x) dx can be interpreted
as the area bounded by the continuous positive curve y =/(x), the x-axis, and the
lines x = a and x = b.

EXAMPLE

Find the area bounded by the curve

y = x3 + 3x2

Figure 4.5

x
 4.4 Definite Integration / 405

the x-axis, and the lines x = 0 and x = 2 (see Figure 4.5).

r2
A = (x3 + 3x2) dx = + X- = (4 + 8) - 0 = 12
Jn

EXAMPLE

Find the area bounded by the curve

x2y = x2 — 4

the x-axis, and the lines x = 2 and x = 4 (see Figure 4.6).

Figure 4.6
y

dx = [x + 4x ^2

= (4+1)-(2+ 2)

= 1

EXAMPLE

Find the area in the first quadrant bounded by the x-axis and the curve

y = 6x + x2 — x3

(see Figure 4.7).

2 X3 X4
A = (6x -I- x2 — x3) dx = = (27 + 9 -0 = 15i
*’0 + 3 4
406 / Integral Calculus

Figure 4.7

(1.8, 8.3)

y — 6x + x2 — x3

(-1.1, 6.7)

EXAMPLE

Find the total area between the parabola

y = x2 — 4x

the x-axis, and the line x = —2 (see Figure 4.8).

A — 4.x) dx = 8) = lOf

Figure 4.8
y

= x2 — 4x
 4.4 Definite Integration / 407

Interpretation of Negative Areas

In the definition of area given above,

A — | f(x)dx

f(x) is assumed to be a continuous positive function between a and b. If f(x) is

negative between a and b, that is, if the curve y =f(x) lies below the x-axis between a
and b, then the value of the integral

A = f(x) dx

is negative. Such areas below the x-axis are called negative areas; the total absolute
area between a curve, the x-axis, and two ordinates is given by

Total area = £ (positive areas) — £ (negative areas)

note: This is equivalent to saying that area is equal to the absolute value of the
integral and thus is always positive.

EXAMPLE

Find the area bounded by the curve

y = 2x + x2 — x3

the x-axis, and the lines x = — 1 and x — 1 (see Figure 4.9).

A = (2x + x2 — x3) dx — [ (2x + x2 — x3) dx

- i
_1

i r 3
X

, X X
— —
x 2 -f" --—
4 0 3 4

= [(i+4 — i) - o] — [o — (1-4 — i)]

= y§ — (—yj) (positive area minus negative area)

Figure 4.9

(- 0.55, - 0.63)
408 / Integral Calculus

EXAMPLE_

Find the area bounded by the curve

y = x3 — 4x

and the x-axis (see Figure 4.10).

Figure 4.10
y

A 4x) dx 4x) dx

x4 0 X4 2
— 2x2 —— 2x2
~4 -2 4 0
oo

)]-[( 4 - 8) - 0]
1

4-(-4)

Area Between Two Curves

Suppose that the area to be evaluated is between the curves y{ = f(x) and y2 = g{x)
and the lines x = a and x = b and that (for definiteness) f(x) < g(x) for a < x < b
(see Figure 4.11). Then

A = I [g(x) -fix)] dx
a

Note that this formula includes negative areas (with appropriate signs) in the total
area between the curves.
 4.4 Definite Integration / 409

Figure 4.11
y

EXAMPLE

Find the area bounded by the curves

y = x2 and y = x

(see Figure 4.12).

Finding the points of intersection of the curves,

y = x = x2

x(l — x) = 0

x = 0, 1

Figure 4.12
y
410 / Integral Calculus

If x = 0, y = 0; if x = 1, y = 1. Thus

= (i ~ i) ~ 0
i
= z

EXAMPLE

Find the area bounded by the curves

y = x3 and y = 2x2

(see Figure 4.13).

Finding the points of intersection of the curves,

y = x3 = 2x2

x2(x — 2) = 0

x = 0, 2

If x = 0, y — 0; if x = 2, y = 8. Thus
. 2 [2x3 x4!2
A = (2x2 — x3) dx =
•'o

= (¥ - 4) - 0
_ 34
-

Figure 4.13
y
 4.4 Definite Integration / 411

EXAMPLE_

Find the area bounded by the curves

x + 2y = 2 y — x = [ and 2x + y = 7

(see Figure 4.14).

Figure 4.14
y

Finding the points of intersection of the straight lines,

x + 2y — 2 x + 2y = 2 y — x = 1

y — x = 1 2x 4- y = 7 2x + y = 7

3y = 3 — 3y = 3 3y = 9

1 -1 y = 3
x =0 x = 4 x = 2
2 4
2 — x 2 — x^
A = (1 + x) — dx + j (7 - 2x) - dx

f2 f4 3x2 2 3x2 4
= |x dx + (6 — fx) dx = + 6x--—
•'o "2 4 0 4 2

= 3-0 +(24- 12) — (12 — 3,

= 6

EXAMPLE

Find the area bounded by the curves

y — x2 y = x y — 2x

(see Figure 4.15).

412 / Integral Calculus

Finding the points of intersection of the curves,

y = x = x2 y = 2x = x2 y = x = 2x

x(x — 1) = 0 x(x — 2) = 0 x = 0

x = 0, 1 x = 0, 2 y = 0

If x = 0, y = 0 If x = 0, y = 0

If x = 1, y = 1 If x = 2, y = 4

Thus

A =
, 1
(2x — x) dx +
,2
(2x — x2) dx
J0 Ji

x'
+
* -y

i — 0 + (4 — f) — (1 — j)
7
= 6

note: In evaluating an area it is theoretically of no importance whether horizontal or

vertical elements of area are used (see Figure 4.11); the choice depends on the number and
difficulty of the integrals needed to determine the particular area. In the above examples, the
curves are easily written in the form x = g(y) and horizontal elements of area can conveniently
be used.

EXAMPLE

The area bounded by the curves

y = x2 and y = x
 4.4 Definite Integration / 413

could have been determined using horizontal elements of area as follows:

2y 3/2 r
A = (y12 -y)dy = = a - i) - o=i
2

EXAMPLE

The area bounded by the curves

y = x3 and y = 2x* 2 3

could have been determined using horizontal elements of area as follows:

1/2' 3y4/3 2 1 /2y 3/2
1/3 y
A = y 21/2
dy =

= (12-¥)-0 =

EXAMPLE

The area bounded by the curves

x + 2y = 2 y — x = 1 2x + y = 7

could have been determined using horizontal elements of area as follows:

.i '7-y '7-y
A = - (2 - 2y) dy + - (y - i) dy
-i

I(y + 1) dy + f 1(3 - y) dy
j -1 ji
i
V , 3y 9y 3y: 3

+
4 2 -i

= (I + f) - (I -!) + (¥-¥)-(!- i)
= 6

EXAMPLE

The area bounded by the curves

y = x2 y = x y = 2x

could have been determined using horizontal elements of area as follows:

A = 1/2 y
y -\\ dy + y
~2]dy

r +
2 y3/2 y2
4 3 4

= i - 0 + (4f - 4) - (i - i)
7
=5
414 / Integral Calculus

Probability as an Area
One of the most useful applications of definite integration to obtain the area under a
curve is in evaluating probabilities. The probability of an event can be obtained by
determining the corresponding area under a frequency function. A frequency function
for a continuous variable x is a function f(x) having the following properties

1. /(x) > 0

2. f(x) dx = 1

b
3. j f(x) dx = P(a < x < b)
a

where a and b are anv two values of x and < b.

These properties state that (1) probability is always nonnegative; (2) the probabil¬
ity of an event that is certain to occur is one; and (3) the probability that the value of
x is in the interval (a, b) is equal to the area bounded by y =f (x), y = 0, x = a, and
x = b. (See Figure 4.16).

note: Iff(x) is continuous, P(a < x < b) = P(a < x < b).

Figure 4.16

fix)

EXAMPLE

In an autoparts warehouse, the proportion of orders filled per day has a frequency function
given by

f(x) = 20(x3 — x4) 0< x < 1

(a) What is the probability that less than 20 percent of the orders will be filled in a day?
(b) What is the probability that between 90 and 100 percent of the orders will be filled in a
day?
 4.4 Definite Integration / 415

0.2
(a) P(x < 0.20) = 20 (x3 — x4) dx
•'o
0.2
= 20

0.00032\
= 20
~5/
= 0.008 - 0.00128

= 0.00672

(See area a in Figure 4.17.)

Thus the probability of filling less than 20 percent of the orders is 0.00672, which means
that this is quite unlikely to occur.

(b) F(0.9 < x < 1) = 20 (x3 — x4) dx

*'0.9

x X
i
= 20
T T 0.9

= 20
5 r 0.6561
4
0.59049^
5

= (5 - 4) - (3.2805 - 2.36196)

= 0.08146

(See area b in Figure 4.17.)

Thus the probability of filling between 90 and 100 percent of the orders is 0.08146.
Note that, since/(x) = 20(x3 — x4) is defined only for 0 < x < 1 and is zero elsewhere,

Figure 4,17

fix)
416 / Integral Calculus

So f(x) dx = 1 as required for a frequency function; note also thatf(x) > 0 for 0 < x < 1
and is zero elsewhere.

EXAMPLE

Suppose that measurements of daily temperature are recorded to the nearest tenth of a degree.
Then the difference between the true and recorded temperatures, the rounding error, is be¬
tween — 0.05 and 0.05 degrees. If the error is uniformly distributed in this interval, its
frequency function is represented by

/(x) = -1 =10 -0.05 <x <0.05

Note that, in general, if x is uniformly distributed on the interval a < x < b, then

/M — i- a < x < b
b — a

and is zero elsewhere.

(a) What is the probability that the rounding error is between —0.01 and 0.01?
(b) What is the probability that the rounding error is between 0.04 and 0.05?

o.oi
(a) P(-0.01 < x < 0.01) =| 10 dx
J-o.oi

= [IOxT-Tch

= 0.1 -(-0.1) = 0.2

(See area a in Figure 4.18.)
0.05
(b) P(0.04 < x < 0.05) = 10 dx
*'0.04

= [iOx]Voi
= 0.5 - 0.4 = 0.1

(See area b in Figure 4.18.)

Figure 4.18

/ (x)
 4.4 Definite Integration / 417

Note that
0.05
10 dx = 10 dx = [10x]° o5os
- 0.05

= 0.5 - (-0.5) = 1

So f(x)dx= 1 as required for a frequency function; note also that/(x) > Ofor —0.05 <
x < 0.05 and is zero elsewhere.

PROBLEMS
Sketch each of the following curves and find the area bounded by the curve, the x-axis, and the
given ordinates.

1. y = y/x; x = 1, x = 16
2. y = 2x + l;x = 0, x = 4
3. y = x"; x = 0, x = 1
4. y = 3x2; x = 1, x = 3
5. y = x2 — 3x; x = — 1, x — 4
6. y = — x2 4- 4x (and x-axis)
lx2 x < 2
7. /M = \ — x + 6 x = 0, x = 3
x > 2
|2x + 3 x < 3
8. /(x) = x = 2, x = 5
— x + 12 x > 3
9. Find the area between the curve y = 2x4 — x2, the x-axis, and the two minimum ordinates.
10. Find the area bounded by the coordinate axes and the parabolic arc yfx + y/y — y/a.

Draw a sketch and find the area bounded by the following curves.

11. y2 = 2x 21. y = x2
y = x — 4 y = x + 2
12. x2 = 2ay y = —3x + 18
y — 2a 22. y = 4x — 4
13. y = x — x2 y = W
y = —x y = 6 - x
14. y2 = 4ax 23. y = x3 + 3x2 + 2
x2 = 4 ay y = x3 + 6x2 — 25
15. y2 = x 24. y = 25 - x2
y = x3 y = (5 - *)2
16. y = (x — l)3 25. y = x3 — 3x2 — lOx
>' = x2 - x - 1 y = — 6x
17. y2 = 5a2 — ax 26. y = (x + 2)(x — l)(x — 5)
y2 = 4ax y = (x + 2)(x — 1)
18. x2y = 4 27. y = x(x — 3)(x -I- 3)
y = 7 — 3x y = — 5x
19. y = x2 28. y = x3 4- 3x2 + 6
y = 8 — x2 y = x3 + 4x2 -I- 5x
y = 4x + 12 29. y = x3 — 5x2 — 8x + 12
20. y3 = x2 y = x3 — 6x2 + 21
2x -I- y + 1 = 0
x - y = 4
HINTS
15. y2 = x and y = x3 intersect (0, 0) and (1, 1)
18. x2y = 4 and y = 7 - 3x intersect (1, 4) and (2, 1)
20. y3 = x2 and 2x + y -F 1 = 0 intersect (-1, 1)
y3 = x2 and x - y = 4 intersect (8, 4)
418 / Integral Calculus

29. x3 — 5x2 — 8x + 12 = (x + 2)(x — l)(x — 6)

30. A computer center is open 12 hours a day and repairs, except for emergencies, are
scheduled for the other 12 hours. The frequency function for the number of hours the
center is actually in operation is given by

1 x2
/M 24 + 1152
0 < x < 12

a. Find the probability that the center is in operation between 10 and 12 hours a day. b.
Find the probability that the center is in operation less than 6 hours a day.
31. A firm has a very large number of company cars for use by its employees. Records of the
time each car is out of service for repairs are kept as the basis for deciding when a car
should be sold. The frequency function for the total number of days a car is out of service
before it is considered too expensive to repair and is sold is given by

/(x) = 0.2c 0 2x 0 < x < oo

a. Find the probability that a car is out of service a total of more than 30 days before being
sold. b. Find the probability that a car is out of service a total of less than 5 days before
being sold.
32. Determine whether each of the following is a frequency function.

a. /(x) = x 2, 1 < X < 00 d. /(x) 1 < x < 20

3 x2 4x3
b- f(x) — g 576, 0 < x < 24 e- f(x) 1 < x < 2
U'

c. /(x) = 2e~Ax, 0 < x < oo f- /m 3 < x <j

Answers to Odd-Numbered Problems

1. 42 19. 64
23
3. 0.01 21. 6
49
5. T 23. 108
37
7. "5" 25. 160J
7
9. I2T> 27. 8
11. 18 29. 166§
13. 34 31. a. 0.002
15. T2 b. 0.632
17.

4.5 Applications of Definite Integration

in Business and Economics
Definite integration has a variety of applications in business and economics; the
concepts of consumer's surplus and producers' surplus are two examples of such appli¬
cations. Other examples involve the determination of the revenue produced for a firm
by a factor of production over a period of time in relation to the cost of that factor
over the same period of time; in such examples the point in time at which revenue
produced equals cost of the factor can also be determined. Applications involving
consumers’ surplus, producers’ surplus, and revenue vs. cost are discussed in this
section.

CONSUMERS' SURPLUS
A demand function represents the quantities of a commodity that would be
purchased at various prices. If the market price is y0 and the corresponding market
demand is x0, then those consumers who would be willing to pay more than this
 4.5 Applications of Definite Integration in Business and Economics / 419

Figure 4.19

market price gain from the fact that the price is only y0 (see Figure 4.19). Under
certain economic assumptions, the total consumer gain is represented by the area
below the demand curve and above the line y = y0 • This area is designated by
Marshall as consumers surplus and is evaluated as
r x°
consumers’ surplus = f(x) dx ~ x0y0
Jo

where the demand function is y — f (x), or as

, m0

consumers’ surplus = 9{y) dy

yo

where the demand function is x = g(y) and m0 is the value of y when x = 0, that is, m0
is the y-intercept of the demand function. Thus
mo
rxo
consumers’ surplus = f(x) dx - x0y0 g(y) dy
jo yo

EXAMPLE

If the demand function is y = 32 — 4x — x2, find the consumers’ surplus

(a) if x0 = 3,
(b) if y0 = 27. (See Figure 4.20.)
3

(a ) consumers’ surplus = (32 — 4x — x2) dx — (3)( 11)

•'o

32x — 2x2 — — - 33

= (96 — 18 — 9) — 0 — 33

= 36
420 / Integral Calculus

Figure 4.20

(b) consumers’ surplus (32 — 4x — x2) dx — (1)(27)

Jo

= 32x — 2x2 * — * — 27
3 o

= (32 - 2 - J) - 0 - 27
_ 38
=

EXAMPLE

If the demand function is y = x/9_ — x and x0 = 5, find the consumers’ surplus using two
different methods (see Figure 4.21).

consumers’ surplus =
* 5
(9 — x)1/2 dx — (5)(2) = I
/» (9 — y2) dy
* 0 J2

f5(9 - x)1/2 dx - 10 = [-$(9 - x)3/2]g - 10

* o

= —( — 18) — 10 = f
Alternatively,

y3
2
(9 - y2) dy =
h-y]
8\ 8
= ( 27-9)-
 4.5 Applications of Definite Integration in Business and Economics / 421

Figure 4.21
y

EXAMPLE

The quantity sold and the corresponding price, under a monopoly, are determined by the
demand function y = 16 — x2 and by the marginal cost y' = 6 + x in such a way as to maxi¬
mize profit. Determine the corresponding consumers’ surplus (see Figure 4.22).

revenue = 16x — x3

marginal revenue = 16 - 3x2

Profit is maximized when marginal revenue equals marginal cost—that is, when

16 — 3x2 = 6 + x

3x2 + x — 10 = 0

(3x — 5)(x + 2) = 0

x = -j, -2

Figure 4.22
y

X
422 / Integral Calculus

so x0 = $, yo = H1-
r 5/3
consumers’ surplus = (16 — x2) dx — (iJfH2)

x3l5/3 595
16x —
T o Yf
125 5_9_5
~8rr )-o- 27

3.09

PRODUCERS' SURPLUS
A supply function represents the respective quantities of a commodity that would be
supplied at various prices. If the market price is y0 and the corresponding market
supply is x0, then those producers who would be willing to supply the commodity
below this market price gain from the fact that the price is y0 . Under certain econom¬
ic assumptions the total producer gain is represented by the area above the supply
curve and below the line y = y0 and is known as producers' surplus (see Figure 4.23).
This area is evaluated as

producers’ surplus = x0y0 — f(x) dx

o
where the supply function is y — f(x) and as

producers’ surplus = g(y) dy

J Mo

where the supply function is x = g(y) and M0 is the value of y when x = 0, that is, M0
is the y-intercept of the supply function.
* xo ~ yo
producers’ surplus = x0y0 - /(x) dx = g(y) dy
J0 ‘'Mo

Figure 4.23

y
 4.5 Applications of Definite Integration in Business and Economics / 423

EXAMPLE

If the supply function is y = (x + 2)2 and the price is y0 = 25, find the producers’ surplus using
two different methods (see Figure 4.24).
3 25
producers’ surplus = (3)(25) — (x + 2)2 dx = f (y1/2 - 2) dy
•'o

(x + 2)' = 75 _ (125 _ 8) = 36
75 (x + 2)2 dx = 75 —

Figure 4.24
y

Alternatively,
25 3/2 25
2y = (150 _ 50) _ _ 8) = 36
(.y1/2 “2)dy = - 2y

(as above).

EXAMPLE

The quantity demanded and the corresponding price, under pure competition, are determined
by the demand and supply functions y = 16 — x2 and y = 4 + x, respectively. Determine the
corresponding producers’ surplus (see Figure 4.25).

y = 16 — x2 = 4 + x

x2 + x — 12 = 0

(x + 4)(x — 3) = 0

x = 3, —4

*o = 3, yo = 7
424 / Integral Calculus

Figure 4.25
y

= 21 - [(12 + f) - 0]

note: The area represented by jo (4 + x) dx could have been evaluated also by the
formula for the area of a trapezoid, A = ja(bl + b2) = f(4 + 7) =

EXAMPLE

The quantity demanded and the corresponding price, under pure competition, are determined
by the demand and supply functions y = 36 — x2 and y = 6 + (x2/4), respectively. Determine
the corresponding consumers’ surplus and producers’ surplus (see Figure 4.26).

x
y — 36 — x2 = 6 + -~

5x2 = 120

x2 = 24

x = ±2^6

x0 = 2v/6, y0 = 12

consumers’ surplus — x2) dx — (2n/6)(12)

2V6
36x — ~ — 24^/6
3 o

= (72n/6 - 16n/6) - 0 - 24^/6

= 32n/6 a 78.4
 4.5 Applications of Definite Integration in Business and Economics / 425

Figure 4.26
y

2V6
, X
= 24^/6 —
6X+12 0

= 24^/6 - (12^/6 + 4^/6) - 0

= 8^/6 « 19.6

REVENUE VERSUS COST

Integration can be used to determine total profit or total net earnings in various
contexts. In general, profit is maximized (assuming pure competition) when marginal
revenue equals marginal cost. Total profit is the integral of marginal revenue minus
marginal cost from zero quantity to the quantity for which profit is maximized.

EXAMPLE

Find the profit-maximizing output and the total profit at that point if the marginal revenue
and marginal cost functions are given by

MR = 25 - 5x — 2x2

MC = 15 — 2x — x2

(see Figure 4.27).

If MR - MC = 0,

25 — 5x — 2x2 — 15 + 2x -F x2 = 0
10 — 3x — x2 = 0

(5 + x)(2 — x) = 0

x = —5, 2
426 / Integral Calculus

Figure 4.27
MR, MC

The first derivative of MR — MC is the second derivative of total profit and its sign thus
indicates whether profit is maximized or minimized for a particular value of x.

d2P
^ (MR - MC) = jP2= - 3 - 2x and = -7
UA wA dx2 x= 2

so profit is maximized for x = 2.

total profit = (10 — 3x — x2) dx

= (20 - 6 - f) - 0

_ 34
- ~y

EXAMPLE

A manufacturing company has purchased a machine that has an output representing addi¬
tional earnings (additional revenue minus additional cost of labor and materials) at time t of

E(t) = 225 - if2

 4.5 Applications of Definite Integration in Business and Economics / 427

Figure 4.28

E(t), Rif)

(30, 0) Time

where E(t) is in units of $10,000 and t is years. The additional repair and maintenance cost at
time t is

R(t) = 2t2

where R(t) is in units of $10,000 and t is years. First, suppose that the machine can be disposed
of at any time with no cost or salvage value. Then the machine should be disposed of at the
time when additional earnings equals additional cost of repair and maintenance (see Figure
4.28).
Additional earnings equals additional cost of repair and maintenance if

225 - if2 = 212

225 = 112

t2 = 100

t = 10

Thus the machine should be disposed of after 10 years. The total net earnings (earnings minus
cost of repair and maintenance) after 10 years is

.10 .10
[E(t) - R(t)] dt = (225 - Jr2) dt
*0 ^0

= [225r - Jr3]J0
= (2250 - 750) - 0

= 1500 or $15,000,000

Now suppose the machine has a salvage value at time t of

6480
6 T t

where S(t) is in units of $10,000 and t is years. Then the company will maximize its net earnings
if it disposes of the machine at time T when the net earnings after T equals the salvage value at
T (see Figure 4.29).
428 / Integral Calculus

Figure 4.29

S(/), Nit)

Net earnings after 7 is

10

N(t) = (225 - it2) dt

Jj

10
3t:
2251 -
T

= 1500 - 2257+ |T3

Net earnings after 7 equals salvage value at 7 if

6480
= 1500 - 2257 + |73
6 + 7

6480 = 9000 - 13507 + |73 + 15007 - 22572 + |74

0 = 2520 + 1507 - 225T2 + |73 + |74

= (7 - 4)(|73 4- ^72 - 1957 - 630)

Thus the machine should be disposed of after 4 years.

EXAMPLE

A company is considering adding salesmen to its staff. The cost of employing additional
salesmen is

5y2 = 48x

where y is cost in units of $1,000 and x is number of additional salesmen employed, and the
additional revenue is

(.R - 2)2 = 4(x + 10)

where R is revenue in units of $1,000 and x is number of additional salesmen employed.

(Assume the cost and revenue functions are continuous, although actually they are meaningful
only for integer values of x.) The company should employ additional salesmen until the cost of
doing so equals the additional revenue obtained (see Figure 4.30).
 4.5 Applications of Definite Integration in Business and Economics / 429

Figure 4.30

y,R

The cost of employing additional salesmen equals the additional revenue obtained if R y-
From the first equation,

V
x =
48

From the second equation,

(R - 2)2 - 40
x =

Solving simultaneously and putting R = y,

5y2 (R - 2)2 - 40
48~ 4

5y2 = 12y2 - 48y + 48 - 480

ly2 48y - 432 = 0

(7y 4- 36)(y — 12) = 0

y = 12
x = 15
and 15 additional salesmen should be employed. The total resulting net revenue (total revenue
minus cost) is

[Ay2 ~ (i(y ~ 2)2 - 10)] dy — | [y + 9 - Ay2] dy

jo Jo

[v2 l12
= ~ +9 y — rAy3
^ o

= (72 + 108 - 84) - 0

= 96 or $96,000
430 / Integral Calculus

PROBLEMS
1. If the demand function is y = 39 — x2, find the consumers’ surplus if a. x0 = | and b. the
commodity is free, that is, y0 = 0.
2. If the demand function is y = 16 - x2 and the supply function is y = 2x + 1, find con¬
sumers’ surplus and producers’ surplus under pure competition.
3. If the supply function is y = -J9 + x and x0 = 7, find the producers’ surplus.
4. If the supply function is y — 4ex/3 and x0 = 3, find the producers’ surplus.
5. The demand and supply functions under pure competition are y = ^(9 — x)2 and
y = i(l + 3x), respectively. If an additive tax of 3 per unit quantity is imposed on the
commodity, determine the decrease in consumers’ surplus.
6. Under monopoly, the quantity sold and the corresponding price are determined by the
demand function y = ^(10 — x)2 and the total cost function y — (x3/4) -I- 5x so that profit is
maximized. Determine the corresponding consumers’ surplus.
7. Under monopoly, the quantity sold and the corresponding price are determined by the
demand function y = 20 — 4x2 and the marginal cost function / = 2x + 6 so that profit is
maximized. Determine the corresponding consumers’ surplus.
8. If the demand function is that part of the equilateral hyperbola y = [8/(x + 1)] — 2 in the
first quadrant, and the supply curve is y = \(x + 3), find the consumers’ surplus and the
producers’ surplus under pure competition.
9. Under monopoly, the quantity sold and the corresponding price are determined by the
demand function y = 45 — x2 and the marginal cost function y' = 6 + (x2/4) so that profit
is maximized. Determine the corresponding consumers’ surplus.
10. The demand and supply functions under pure competition are, respectively, y = 14 — x2
and y = 2x2 + 2; find a. consumers’ surplus, and b. producers’ surplus.
11. The demand function is y = 20 — 3x2 and the supply function is y = 2x2; find the con¬
sumers’ surplus and the producers’ surplus under pure competition.
12. The demand and supply functions under pure competition are, respectively, y = 32 — 2x2
and y = -3X2 + 2x -I- 5; find a. consumers’ surplus, and b. producers’ surplus.
13. Find the profit-maximizing output and the corresponding total profit, assuming pure
competition, if MR = 20 — 2x and MC = 4-1- (x — 4)2.
14. The marginal revenue function is MR = 25 - 3x and the marginal cost function is
MC = 25 — 7x + x2. Find the profit-maximizing output and the corresponding total
profit under pure competition.
15. If MR = 44 — 9x and MC = 20 — 7x + 2x2, find the profit-maximizing output and the
corresponding total profit under pure competition.
16. Assuming pure competition, find the profit-mqximizing output and the corresponding
total profit if MR = 24 — 6x — x2 and MC = 4 — 2x — x2.
17. If MR = 15 — 5x and MC = 10 — 3x + 3x2, find the profit-maximizing output and the
corresponding total profit assuming pure competition.
18. A manufacturing company has purchased a machine whose output represents earnings at
time t given by y2 = 6(t -I- 9), where y is in units of $10,000 and t is years. The repair and
maintenance cost at time t is given by (y + 4)2 = 8(f -F 17), where y is in units of $10,000
and t is years. Assuming the machine can be disposed of at any time with no cost or
salvage value, how many years should it be kept to maximize total net earnings (earnings
minus cost of repair and maintenance)?
19. A company is considering purchasing additional companion units for its computer. The
savings (in time and mistakes) from added units is given by y = 3x2 + 11, where y is in
units of $11,000 and x is number of units added. The cost of repair and maintenance is
given by y = 4x2 + 2, where y is in units of $1000 and x is the number of units added. To
maximize total net revenue (revenue minus cost of repair and maintenance), how many
units should be added and what is the associated savings? Assume the units are added in a
specified order and that the savings and cost curves are continuous.
20. A company is considering adding advertising personnel. The cost of adding such personnel
is given by y = \x, where y is in units of $5000 and x is the number of personel added.
Additional revenue from adding personnel is R2 = 4x, where R is in units of $5000 and x is
number of personnel added. What number of advertising personnel should be added to
 4.6 Special Methods of Integration / 431

maximize profit (revenue minus cost) and what is the associated additional revenue?
Assume continuous functions.
21. If net investment flow is 7(f) = 6r1/2, determine a. capital accumulation during the first
year; b. capital accumulation during the first nine years; and c. total amount of capital
after the first nine years if the initial capital is $500.
22. If net investment flow is 7(f) = 4f1/3, determine a. capital accumulation during the first
year; b. capital accumulation during the first eight years; and c. total amount of capital
after the first eight years if the initial capital is $500.

Answers to Odd-Numbered Problems

1. a. 31i 13. x = 6; Pmax = 36
b. 26^13 15. x = 3; Pmax = 45
2 10
17. x = 1, Pmax = 3
19. 18
21. a. $4.00
9. 16^3 b. $108.00
11. c.s. = 16 c. $608.00
p.s. = ¥

4.6 Special Methods of Integration

In many cases, the integral of a given function cannot be obtained by reversing the
rules of differentiation. A number of special methods are available for determining
the integral in such cases. The special methods considered in this section are standard
forms for integration, integration by parts, integration by partial fractions, integra¬
tion by rationalizing substitution, and integration by miscellaneous substitution.

STANDARD FORMS FOR INTEGRATION

In differential calculus there is a general rule for differentiation, from which particu¬
lar rules are derived. Unfortunately, there is no corresponding general rule for inte¬
gration and the integral of any given expression must be obtained through
knowledge of the results of differentiation. Thus integration is inherently more
difficult than differentiation. In order to expedite the process of integrating a given
expression, tables of known integrals, called standard forms, have been compiled. The
simplest of these standard forms are obtained directly by reversing the corresponding
rules for differentiation; additional standard forms are derived from the results of
more complicated differentiation. Note that a formula for integration can be derived
from the result of any differentiation; the most frequently occurring of such formulas
are given in tables of standard forms.
In the following list of standard forms, the first 14 formulas are obtained by
reversing the corresponding rules for differentiation (the first 5 of these are discussed
at the beginning of the chapter); the last 12 formulas are examples of standard forms
derived from the results of more complicated differentiation.

1. dx = x + C.
.
2. K dx = K dx, where K is any constant.
ft

3. [du + dv] = du dv, where u = f(x) and v = g(x) are differentiable func-

tions of x.
 432 / Integral Calculus

xn + 1

4. x" dx =-- + C, n =/= — 1.

n + 1

f W n+ 1
5. un du =-h C, n + — 1, where u = f(x) is a differentiable function of x.

f 1
6. - du = In u + C, where w =/(x) is a differentiable function of x.
» Vi

7. du = eu + C, where u =f(x) is a differentiable function of x.

8. dw = --h C, where u =f(x) is a differentiable function of x.

J In u

9. j sin u du = —cos u du + C.

10. cos u du = sin n du.

11. j sec2 u du = tan u + C.

12. esc2 u du = —cot u + C.

13. sec w tan u du = sec w + C.

14. j esc u cot u du = —esc w + C.

15. j tan u du — —In cos u + C = In sec u + C.

16. cot u du — In sin u -j- C.

17. sec u du — In (sec u + tan u) + C.

18. esc u du = In (esc u — cot u) + C.

du 1 a + u ^ ,
19. = — In -+ C, u2 < a2.
a — u 2a a — u

du 1 , u —a ^ ,
20. = — In-1- C, ir > a .
u2 — a2 2a u + a

du
21. = In (n -(- yju2 “t- a ) -I- C.
yju2 + a'

u
22. | x/u2 4- u2 dw = + u2 -f In (w + w “I- a ) -f~ C.

23. ueu du = eu(u — 1) + C.

 4.6 Special Methods of Integration / 433

24. j In u du = u In u — u + C.

26. I — = In (In u) + C.
J u In u

The first step in the procedure for integrating a given expression consists of
comparing it with various standard forms. If an expression is identical with a stan¬
dard form, its integral is known; if an expression is not identical with a standard
form, there are various methods by which it may in some cases be reduced to a
standard form. Some of these methods are discussed in later sections.

EXAMPLES

Evaluate the following integrals.

r x dx 1 r 2.x dx
x2 + 1 2 x2 + 1

= -j In (x2 + 1) + C (standard form 6)

-ie~x2 + C (standard form 7)

(standard form 8)

= — -g COS3 3X + C (standard form 5)

(standard form 5)

a + c tan by be J a + c tan by

— In (a -I- c tan by) + C (standard form 6)

be

(standard form 12)

(ex -I- 2x) tan (ex + x2 + 5) dx = —In cos (ex + x2 + 5) + C

= In sec (ex + x2 + 5) + C (standard form 15)

434 / Integral Calculus

sec 4-tzx tan 4nx dx = — sec 4nx 4 C (standard form 13)

J 471

| (0 4 x) esc2 (0 4- x)2 dx = —i cot (0 4- x) + C (standard form 12)

f 7C
7t esc Ox dx = - In (esc 0x — cot 0x) 4- C (standard form 18)
J 0

j* 4 esc 40 cot 40 d6 — — esc 0 + C (standard form 14)

j y2 sec (0 4- y3) dy = 4 In [sec (0 4 y3) 4- tan (0 + y3)] + C (standard form 17)

| x cot (x2 -}- 0x) dx = j In sin (x2 4- 0x) + C (standard form 16)

4 + 6x 1 , (V6 + 4x 4- 3x2\
(standard form 20)
. 6 — (4x -I- 3x2)2 2v/6 176 — 4x — 3x2/

| 4x3cx2 dx = 2ex\x2 — 1) + C (standard form 23)

1 si^dx = Un (ln x3) + c (standard form 26)

| (2x — 3)(x2 — 3x — l)4 In (x2 — 3x — 1) dx

In (x2 — 3x — 1) 1
= (x2 — 3x — l)f + C
5 25

(standard form 25)

| (x2 4- 2x 4- 6)1/2 dx = | 7(x 4- l)2 + (\/5)2 dx

x + 1
(x2 4- 2x 4- 6)1/2 4- | In [(x 4- 1) 4- (x2 4- 2x 4- 6)12] 4 C

(standard form 22)

J (x2 4- 2x) In (x3 4- 3x2 + 14) dx

= i[(x3 4- 3x2 4 14) In (x3 4- 3x2 4 14) — (x3 4- 3x2 4- 14)] + C

(standard form 24)

PROBLEMS
Evaluate the following integrals.

1. j xex2 dx 4. sec - tan dx

a a
2
sec 2x '
2. | sin2 x cos x dx 5. dx
j I 4- tan 2x,

3. sin ax cos ax dx 6. (x2 4 1)5x3 + 3xdx

 4.6 Special Methods of Integration / 435

cos x dx sin 2x
7. 14. | -dx
1 + sin x cos 2x

8. cC0SZ sin z dz 15. I* (sin x + cos x) dx

dx
9. 16. | (3x2 + 5 cos x) dx
sirr ax

du
10. 17. I* (3x + ex) dx
cos2 u

11. sec ax dx 18. j (ex — e~x) dx

dt
12.
1 + cos t
19. J 2x sec x2 tan x2 dx

13. J* 2x cos x2 dx 20. | (x3 -I- 3x)(3x2 + 3)ex3 + 3x dx

Answers to Odd-Numbered Problems

1. jex2 + C
11. - In (sec ax + tan ax) + C
a
sirr ax
3. +C
2a 13. —sin x2 + C

15. sin x — cos x + C

2(1 + tan 2x) + ^
17. fx2 + ex + C
7. In (1 + sin x) + C
19. sec x2 -I- C
„ 1
9.-cot ax 4- C
a

INTEGRATION BY PARTS
When an expression involving products or logarithms cannot be evaluated directly
using standard forms, one of the most useful techniques for transforming it to a
standard form is the formula for integration by parts. This formula is based on the
inverse of the formula for differentiation of a product.
If u and v are both functions of x, then from the formula for the differentiation of a
product,
d dv du
= u - + V
dx dx dx

or
dv d, du
Udx = Tx(Uv)-Vdx

and, integrating with respect to Y

u dv = uv v du
«

This is the formula for integration by parts. The usefulness of this formula depends on
the appropriate choice of u and dv, so that J v du and j dv can be evaluated, even
though j u dv cannot.
436 / Integral Calculus

Unfortunately, there is no general rule for separating a given expression into two
factors u and dv so that the formula for integration by parts can be applied. However,
note that

1. dx is always a part of dv.

2. dv must be integrable.
3. When the expression to be integrated is the product of two functions, it is usually
advisable to choose the most complicated looking one that can be integrated as
part of dv in order to make j v du as easily integrable as possible.

It may be necessary to apply the formula for integration by parts more than once,
as in example (e) below.

EXAMPLES

Evaluate the following integrals.

xe
(a) I xeax dx = dx + C
a a

xeax eax
+ C u — x dv = eax dx
a a‘

eax 1, ^
— I x-1 + C du = dx v =
a \ ai a

(Note that, alternatively, standard form 23 could be used.)

1
(b) | In x dx = x In x — x^-j dx + C

= x In x - x + C u = In x dv = dx

= x(ln x — 1) + C du = - dx V = X
x

(Note that, alternatively, standard form 24 could be used.)

, x f
(c)
J
. x , x
x sin - dx = —ax cos —K
a a
p
J
x
a cos — dx + C
a
_

x ? • x ^
= —ax cos - + a sin - + C u = x dv sin - dx
a a a

, . x X „ x
= a sin-ax cos - + C du — dx v a cos -
a a a

(d) j ex cos x dx = ex sin x — | ex sin x dx + C

= e sin x — — ex cos x + \ ex cos x dx + C

= ex sin x + ex cos x — ex cos x dx + C

ex cos x dx = ^ex(sin x + cos x) + C

u — ex dv = cos x dx

du = e> v = sin x

u = e dv = sin x dx

du = ex v = —cos x
 4.6 Special Methods of Integration / 437

(e) x2e~x dx = —x2e~x + 2 xe~x dx + C

= —x2e x + 2 — xe x + e x dx + C
*

x2e x — 2xe x — 2e X + C

= — e X(x2

U = X dv = e x dx

du = 2x dx v = —e

u — x dv = e x dx

du = dx v = e

(0 ln(X+1)dX = 2(x + irin(x + l)-2r

y/x~-f 1
1
x + 1 v
- (x + 1)1/2 dx + C
'

= 2(x + 1 )1/2 In (x -f 1) — 4(x + 1 )1/2 + C

= 2(x + l)1/2[ln (x + 1) - 2] + C

u = In (x + 1) dv = (x + 1)-1/2 dx

dx
du = i; = 2(x + 1)1/2
x + 1

(Note that, alternatively, standard form 25 could be used.)

As suggested by examples (a), (b), and (f) above, the formula for integration by
parts is the basis of a number of standard forms for integration.

PROBLEMS
Evaluate the following integrals.

xex dx
1. 10. (x + 1) In x dx
(TT^p
2. xe x dx 11. | (x2 + 3x + 4)3(2x + 3) dx

3. x2ex dx 12. | e3x2 + 6x+10(x + ^dx

4. xe2x dx 13. (x1/3 -f x2/3)2 dx

6x2 + 8x + 8
5. t In t dt 14.
? + 2x2 + Ac dX

(x3 + x) dx
6. z2e 3z dz 15.
^/x4 + 2x2 + 1

7. ex(x + l)2 dx 16. x2 -I-1- 6x I dx

1 x /

3x
8. I (x1'2 + x1/4)2 dx 17. xe dx

(x + 1) <ix x3 + 2
9. 18. dx
(x + l)2 + (a + 1): x4 + 8x + 10
438 / Integral Calculus

sin v dv sin x dx
25. j
19-J 1 -1- cos V yj2 — cos X

dx sec2 0 dO
2°.j 26. j
1 4- sin x Vi +2 tan 6

27. j x cos x dx
2LJ u sin u2 du

22. J (x -1- sin 2x) dx 28'J e~ax sin nx dx

23. xax dx 29. 9 sec2 9 d9

.

24. 1* x" In x dx 3°.j y2 sin ny dy

Answers to Odd-Numbered Problems

17. -i<T3x(3x + 1) + C

19. — In (1 4- cos v) 4- C
3. ex(x2 — 2x + 2) 4- C
21. — i cos u2 4- C
5. \t2 In t — \t2 + C
ax / 1
7. ex(x2 4- 1) + C 23.
In a\ In a
9. j In [(x 4- 1 )2 4- (a 4- l)2] 4- C
25. 2VT — cos x 4- C
11. ^.{x2 4- 3x 4- 4)4 4- C
27. x sin x 4- cos x 4- C
13. fx5/3 + x2 4- fx7/3 4- C
29. 0 tan 0 4- In cos 9 4- C
15. i(x4 4- 2x2 + 1)3/4 + C

INTEGRATION BY PARTIAL FRACTIONS

A rational algebraic function, by definition, can be expressed as the quotient of two
polynomials. In theory, every rational function has an integral that can be expressed
in terms of elementary functions. If a rational function cannot be integrated directly,
the method of partial fractions can frequently be used to transform the rational
fraction into a sum of simpler functions which can then be integrated by standard
forms. The method of partial fractions is appropriate only for proper fractions, that
is, for fractions in which the polynomial in the numerator is of lower degree than the
polynomial in the denominator. Any improper fraction, that is, a fraction in which
the polynomial in the numerator is of the same or higher degree than the polynomial
in the denominator, can be reduced by division to the sum of a polynomial (easily
integrated) and a proper fraction (integrated by the method of partial fractions). The
method of integration by partial fractions consists of the following steps:
1. Express the denominator of the fraction as the product of linear factors of the
form ax 4- b and irreducible quadratic factors of the form ax2 + bx 4- c; although
this is not always easy in practice, it is always possible in theory for any polynom¬
ial in x with real coefficients.
 4.6 Special Methods of Integration / 439

2. Determine the form of the partial fractions; several cases arise, depending on the
nature of the factors occurring in the denominator:

Factor Occurring
in Denominator Corresponding Partial Fraction

(a) Distinct linear A

factor: ax + b ax + b
where A is a constant to be determined

(b) Repeated linear

+ + ••• +
factor: (ax + b)n ax + b (ax -f b)d (ax 4- b)n
where Ax An are constants to be determined

(c) Distinct quad¬ Ax + B

ratic factor: ax2 4- bx A c
ax2 + bx + c where A and B are constants to be determined

(d) Repeated quad¬ Axx 4- Bx A2 x B2 Anx + Bn

+ + ••• +
ratic factor: ax2 + bx + c ^ (ax2 + bx + c)2 ' (ax2 + bx 4- c)n
(ax2 + bx -1- c)n where At and , i = 1, 2, ..., n
are constants to be determined

3. Determine the constants occurring in the numerators of the partial fractions.

When a rational fraction is separated into partial fractions, the resulting equation
is an identity, that is, it is true for all meaningful values of the variables. The
method for evaluating the constants occurring in partial fractions, as illustrated in
the example below, is based on the algebraic theorem which states that if two
polynomials of the same degree are identical, the coefficients of like powers of the
variable in both polynomials must be equal.
4. Integrate the partial fractions using standard forms.

EXAMPLES

Evaluate the following integrals.

(x 4- 3) dx (x -I- 3) dx B
(a)j + dx
x2 + 3x + 2 (x + 1 )(x + 2) x 4 1 x 4~ 2

x 4- 3 A(x + 2) 4" B(x 4* 1)

(A 4- B)x + (2A 4 B)

so, equating coefficients,

A + B = 1

2A 4 B = 3

A = 2

B = -1
440 / Integral Calculus

and

(x 4- 3) dx C 2 dx r dx
J (x 3~ 1 )(x 3-2) x 3- 1 x 3- 2

— 2 In (x 3- 1) — In (x 3- 2) + C

= In '(* +1)2 3- C
x 3- 2

(x2 — 3x — 8) x + 9 B
(b) dx = 1 - dx = x — + dx
x2 — 2x 3- 1 1 (x-l);

x + 9 = A(x — 1) 3- B

= Ax — (A — B)

so, equating coefficients,

A = 1

A — B = —9

B = 10

and

• (x2 — 3x — 8) dx dx 10 dx
= x —
x2 — 2x + 1 x — 1

10
= x — In (x — 1) + + C

t dt t dt A21 3 Bi
“ A21 3 B2
~

(c)l P 3- 6f2 3- 5 J (t2 + 5)(f2 3- 1) t2 + 5 t2 3- 1

dt

t — {Ay t 3- By){t2 3" 1) 3~ {Ait 3" B2){t2 3- 5)

= {Ay 3- A2}t2 3- {By 3“ B2)t2 3- {Ay 3- 5A2)t 3- {By 3" 5B2)

Ay 3- A2 = 0

By 3“ B2 = 0

Ay +5A2 = 1

By 3 5B2 — 0
~

From the first and third equations,

A2 — \

Ay =

From the second and fourth equations,

By = B2 = 0

and

t dt t dt 1 r t dt
(PTlM^TT) ? + 5+ 4 J F+T
= -i In {t2 3- 5) + i In {t2 + 1) + C
 4.6 Special Methods of Integration / 441

dz Aq , A{z + By A2z 4 B2
(d) dz
z(z2 4 1): z + z2 4 1 (z2 -l-l)2

1 = 40(z2 4 l)2 4 (AyZ 4- Bl)z(z2 4 1) 4 (A2z 4 B2)z

= 40(z4 4- 2z2 4- 1) 4 Aj(z4 4 z2) 4- By(z3 4 z) 4 42z2 4 52z

= (40 4 41)z4 4 By z3 4 (240 4 4 42)z2 4 (J5j 4 B2)z 4 40

so, equating coefficients,

Aq 4 Ay = 0

By = 0

24.Q 4 Ay 4 42 = 0

By 4 B2 = 0

40 = 1

Substituting 40 = 1 and solving the other equations,

Ay = -1

42 = -1

B2 = 0

and

dz • dz r z dz z dz
J z(z2 4 l)2 z z2 4 1 4 + 4

= In z — j In (z2 4 1) 4 4 C
2(z2 4 1)

1
— i In 4 4 C
z2 4 1 2 4 +1)
dx dx 4 B C
(e) -1-t H-t dx
x3 4 5x2 4 4x J x(x 4 l)(x 4 4) x x 4 1 x 4 4

1 = 4(x 4 l)(x 4 4) 4 Bx(x 4 4) 4 Cx(x 4 1)

— 4(x2 4 5x 4 4) 4 B(x2 4 4x) 4 C(x2 4 x)

= (4464 C)x2 4 (54 4 4B 4 C)x 4 44

so, equating coefficients,

4 4 B 4 C = 0

54 4 46 4 C = 0

44 = 1

4 = 3i

Substituting 4 = i and solving the first two equations,

44 4 3B 0

B
i
'I

C 1
T?
442 / Integral Calculus

and

dx 1 r dx 1 dx 1 dx
4 vx
x(x -I- l)(x + 4) 4 J x 3 x 4- 1 12 . x 4 4

= \ In x — ^ In (x 4 1) 4 yj In (x 4 4) 4 C

x3(x + 4)
= TZ 1° 4c
(* + l)4
(x3 — 2x) dx (x3 — 2x) dx
<f> i 81 (x - 3)(x 4 3)(x2 4- 9)

A B Cx + D
4-x 4 dx
x — 3 ' x 4 3 x2 4 9

x3 — 2x = A(x 4 3)(x2 4 9) 4 B(x - 3)(x2 4 9) 4 (Cx 4 D)(x2 - 9)

= 4(x3 4 3x2 4 9x 4 27) 4 J5(x3 - 3x2 4 9x - 27)

4 C(x3 - 9x) 4 D(x2 - 9)

= (A + B + C)x3 4 {3A - 3B 4 D)x2 4 (9A 4 9B - 9C)x

4 (27A - 27B - 9D)

so, equating coefficients,
A 4 B 4 C = 1

3A - 3B 4 D = 0

9A-\-9B—9C= -2

21A - 21B - 9D = 0

From the first and third equations,

18C = 11
r
c — ii
— rs
From the second and fourth equations,
18D = 0

D = 0
Substituting C = f&, D — 0, and solving the first two equations,

A 4 B = ye

3A - 3B = 0

6A = fg
A = 3^

B = 3^
and
(x3 — 2x) dx dx 1 r dx lip X dx

(x — 3)(x 4 3)(x2 4 9) x^3 + 36 J x 4 3 + 18 J x2 + 9

= ^ in (x - 3) 4 je in (x 4 3) 4 ^ In (x2 4 9) 4 C

= ife in [(x - 3)7(x 4 3)7(x2 4 9)11] 4 C

= ts In [(x2 - 9)7(x2 4 9)11] 4 C

note: In this particular problem the correct solution is also obtained using the factors
(x2 - 9)(x2 4 9); however, in general, irreducible quadratic factors and linear factors must be
used.
 4.6 Special Methods of Integration / 443

PROBLEMS
Evaluate each of the following integrals using partial fractions.

(4x — 2) dx r (4x2 4 6) dx
x3 — x2 — 2x J x3 4 3x

(4x3 -I- 2x2 -I- 1) dx r (xc3 -h 3x) dx

4x3 — x 6‘ J " (x2 4 l)2

z2 dz t5 dt
(T^Tp (t2 -f 4)2

(/ - 8) dy r (2t2 — 8f — 8) dt
y3 4- 2y2 ‘J (t - 2)(t2 4 4)

Answers to Odd-Numbered Problems

x(x — 2)
1. In
(x 4 l)2 5. In [x2(x2 4 3)] 4 C

3 - 4z
3. In (z — 1) + +c 7. — 4 In (t2 + 4) - -JL* + C
W^2 2 v ' t2 4 4

INTEGRATION BY A RATIONALIZING SUBSTITUTION

As discussed in the preceding section, all rational functions are integrable in terms of
elementary functions; however, only a relatively small number of algebraic functions
which are not rational can be integrated in terms of elementary functions. In some
cases, by substitution of a new variable, functions which are not rational can be
transformed into equivalent functions which are rational. These equivalent functions
are integrable either by partial fractions or directly by standard forms. The method of
integrating a function that is not rational by transforming it into a rational function
by substitution is called integration by rationalization. This method is useful for
various classes of functions; two of the most common of these are discussed in the
following sections.

1. An expression involving fractional powers of x only can be transformed into a

rational form by the substitution

x = zn

where n is the least common denominator of the fractional exponents of x.

2. An expression involving fractional powers of (a 4 bx) only can be transformed
into a rational form by the substitution

a 4 bx = zn

where n is the least common denominator of the fractional exponents of (a 4 bx).

EXAMPLES

Evaluate the following integrals.

. . r x1/2 dx
1 4 x3/4
444 / Integral Calculus

involves fractional powers of x whose least common denominator is 4. Substituting,

x = z4

dx = 4z3 dz

x1/2 dx r z2 • 4z3 dz
1 -I- x 3/4 1 + z3

z5 dz
= 4
z3 + 1

= 4 z2 - dz
z3 + 1

= fz3 — 3 In (z3 -f 1) + C
Substituting to obtain a function of x,
.1/2
dx
= fx3/4 - j In (x3/4 + 1) + C
It + x 3/4

x dx
(b)
(a -I- frx)3/2

involves a fractional power of a + bx whose least common denominator is 2.

Substituting,

a + bx = z2

b dx = 2z dz

dx = T z dz
b

x dx 2 r (z2 — a) • z dz
(a + bx)312 b:

2_ (z3 — az) Tiz

F z

a
1 - dz
V2

2_ a
z + - + C
P z

2_ z2 + a
+c
V2
Substituting to obtain a function of x,

f x dx 2 2a + bx
+ C
(a + bx)312 ~ b2 (u + bx)112

(5x + 9) dx
(c) (x - 9)x3/2

involves a fractional power of x whose least common denominator is 2. Substituting,

x = z2

dx = 2z dz

(5x + 9) dx r (5z2 + 9) • 2z dz
(x — 9)x3/2 (z2 - 9)z3

(5z2 + 9) dz
= 2
z2(z - 3)(z + 3)
 4.6 Special Methods of Integration / 445

Integrating using partial fractions,

• (5x + 9) dx ^ x4 J5 C D
dz
J (x - 9)x3/2 = z + z2 + z - 3 + z + 3

5z2 + 9 = .4z(z2 - 9) + B(z2 - 9) + Cz2(z + 3) + Dz2(z - 3)

I
= /l(z3 - 9z) + B(z2 - 9) + C(z3 + 3z2) + D(z3 - 3z2)
SO

A + C + D = 0

B + 3C - 3Z) = 5

-9.4 = 0

-9B = 9

^4 = 0

5 = -1

Substituting A = 0, B = — 1, and solving the first two equations,

C + D = 0

3C - 3D = 6

6C = 6

C = 1

D = -1
and

(5x + 9) dx dz dz dz
= 2 ~2
9)x3/2 Zz “*■
z — 3 J z + 3

1
= 2 + In (z — 3) — In (z + 3) + C
z

z — 3
= 2 1 + ln . , + C
z \z + 3

Substituting to obtain a function of x,

• (5x + 9) dx 1 1/2 - 3'

= 2 in + ln + C
. (x — 9)x3/2 /2 + 3,

(d)r dx
x — x 4/3

involves a fractional power of x whose least common denominator is 3. Substituting,

x = Z'

dx — 3z2 dz

dx r 3z2 dz
,4/3 _3
Z
_4
— Z

dz
= 3
J z( 1 - z)

Integrating using partial fractions,

dx A B
= 3 dz
x — x 4/3 z 1 — z

1 = z4(l — z) + Bz
446 / Integral Calculus

-A + B = 0

A = 1

B = 1

and

dx dz dz
1 x — x 4/3
= 3
Z
+
J 1 — z

= 3[ln z — In (1 — z)] + C

z
= 3 In + C
1 — Zj

Substituting to obtain a function of x,

dx ( x1/3
= 3 !n |--J73 + C
x — x 4/3 1 — X

(e) J y(a + y)1/3 dy

involves a fractional power of (a + y) whose least common denominator is 3

Substituting,

a + y = z3

dy = 3z2 dz

j y(a + y)1/3 dy = j" (z3 — a) • z • 3z2 dz

= 3 J (z6 — az3) dz

z7 az4
= 3 + C
7 4

A[z4(4z3 - 7a)] + C

Substituting to obtain a function of x,

I y(a + y)113 dy = ^[(a + y)4,3(4a + 4y - 7a)] + C

= A[(« + #,3(4y - 3a)] + C

x2 dx
(4x + 1)5/2

involves a fractional power of (4x + 1) whose least common denominator is 2

Substituting,

4x + 1 = z2

4 dx = 2z dz

dx = \z dz
 4.6 Special Methods of Integration / 447

x2 dx Mz2-l)2-iz dz
(4x + 1)5/2

1 r (z4 — 2z2 + 1) dz
32 J
1 i 2 1
1-2 + “4
dz
32 ZZ Z*

1 2 1
z + - + C
32 z 3z3

1 3z4 + 6z2 — 1
+ C
32 3z3

Substituting to obtain a function of x,

f x2 dx 1 3(4x + l)2 -f 6(4x + 1) - T

+ C
J (4x + 1)5/2 ~ 32 3(4x + 1)3/2

6x2 + 6x + 1
+ C
12(4x + 1)3/2

INTEGRATION BY MISCELLANEOUS SUBSTITUTIONS

The substitutions considered in the preceding section made integration possible by
rationalizing the function to be integrated. There are also a number of functions
which can be integrated by means of substitutions that do not rationalize them.
Unfortunately, no general rules can be given for these substitutions and they must be
determined by trial and error, guided by experience. A very useful substitution of this
type is the reciprocal substitution:
1 —dz
x = - dx = —~—

z z

Algebraic functions which cannot be rationalized frequently can be integrated using

this substitution.

EXAMPLES

Evaluate the following integrals.

(x - x3)1/3 dx
(a)
x
1

dx

i yi3dz
r (x — x3)1/3 dx z3] z2
J x^ 1

= — z(z2 — 1)1/3 dz

= -i(z2 - 1)4/3 + C
448 / Integral Calculus

Substituting to obtain a function of x,

(x — x3)1/3 dx 3/1 ,\4/3 ^

i?"1' +

dx
(b) using the standard form
ZyJ 1 4- x 4-

du
= —t= In (2cu 4- b 4- 2y/cy/cT+ bu 4- cu2) 4- C
yja 4- bu 4- cu:

x =
z

dz
dx =

dz
dx 22
cyiTx + xj 1,11
“ / 1 + - + -2
2 V Z ZZ

dz
z2 + z + 1

dz
z2 4 z + 1

— —In (2z 4- 1 4- 27? 4- z 4- 1) 4- C

Substituting to obtain a function of x,

dx
XyJ 1 4- x 4- x"

2 4- x 4- 2-^/1 ■(■ x 4 x'

= —In + C

In 4- C
,2 4- x 4- 2-y/l + x +

PROBLEMS
Evaluate each of the following integrals by rationalization.

y dy A | (V* 4-14-1) dx
1.
yj 2 4- 4y -y/x 4- 1 — 1

dt dx
lj~t 4- 1 4- x 4 2
(x 3/2 _ v 1 /3
) dx dx
6.
6x1/4 1 4-

Evaluate each of the following integrals by reciprocal substitution.

dx ^fx2 — a2 dx
7. 8.
x x2 4- a2
 4.7 Numerical (Approximate) Methods of Integration / 449

— dx
9. 11.
Xy/l + 4x + 5x2
y/4 — x2 dx dx
10. 12.

Answers to Odd-Numbered Problems

1. i(2 + 4y)1/2(v - 1) + C

3- A*9/4-A*13/12 +C

5. |(x + a)2'2 - 3(x + a)113 + 3 In [(x -f a)l/3 4- 1] + C

^x2 4- a2
7. - 2 + C
ax
1/2
2 — x'
9. - +c

1 4- 2x 4- yj\ 4- 4x + 5x2'
11. In 4- C

4.7 Numerical (Approximate)

Methods of Integration
Approximate integration is useful in many practical problems when an integral
cannot be expressed in terms of elementary functions or when an integrand is defined
by an empirical table of values or by an empirical graph. Three methods of approxi¬
mate integration are discussed in the following sections: the trapezoidal rule (linear
approximation), Simpson's rule (quadratic approximation), and Taylor s expansion
(series approximation). In all three of these methods, the function to be integrated is
replaced by an approximation whose integral is known. The trapezoidal rule and
Simpson’s rule (but not Taylor’s expansion) can also be used when the integrand is
defined empirically by a table of values or a graph, rather than by a function.
The methods discussed in this section and several other approximation methods
are used to evaluate integrals using electronic computers. Approximate integration,
even of quite difficult integrals, can be done accurately and efficiently using a com¬
puter and, in practice, a computer is used for almost all approximate integration. The
main purpose of the following sections is to discuss the principles involved in such
approximation. The examples involve fairly crude approximations of relatively
simple functions; the amount of calculation required, even for these cases, indicates
the advantage of computer approximation.

THE TRAPEZOIDAL RULE

The trapezoidal rule uses a series of straight-line segments to approximate the given
function f(x); it is in this sense a method of linear approximation.
Suppose the problem is to evaluate f(x) dx. Divide the interval [a, b] into n
equal parts, each of length Ax, and denote the points of division by a = x0, x1;...,
xn = b. Erect ordinates at the points of division and denote them by y0 = /(x0),
yl =f(xl),..., yn = f(xn). Join the ends of consecutive ordinates to form trapezoids
(see Figure 4.31). Then
450 / Integral Calculus

Figure 4.31

Area of first trapezoid = ^(y0 + yx)Ax

Area of second trapezoid = 2(^1 + ^2)^*

Area of nth trapezoid = i(yn_i 4- yn)Ax

and the total area of the n trapezoids is
n- 1

Ax bo + Z1 yi+ bn
i—

The trapezoidal rule for approximate integration is thus

n— 1

f(x) dx % Ax 2(^0 + yn) + Z1 yi i=

Notes Concerning the Accuracy of the Trapezoidal Rule

1. As the number of divisions n increases, the sum of the areas of the trapezoids more
closely approximates the area under the curve.
2. The trapezoidal rule gives a closer approximation to the area under the curve than
does the sum of the areas of the corresponding rectangles—that is,
n- 1 n- 1

/(x) dx — Ax + y„) + i=
X yi
1
< I f(x) dx - Ax X yi
'a i=0

and
n— 1

f(x) dx — Ax 2<yo + y„) + X yt < f(x) dx - Ax X y,

i= 1 i— 1

3. The integral f(x) dx is bounded by Ax Z^o1 yi an<^ Ax Z"= 1 >0 depending on

which bound is smaller,
n~ 1

Ax X yi < f(x) dx < Ax X yi

i= 0 i — 1
 4.7 Numerical (Approximate) Methods of Integration / 451

or
n b n— 1

Ax £ y, < | f(x) dx < Ax £ y,

i= 1 i=0

The trapezoidal rule averages these two bounds.

EXAMPLE

Evaluate j* x(16 — x2)1/2 dx by the trapezoidal rule using n = 4. (See Figure 4.32).

Figure 4.32
v

i 0 1 2 3 4

Xi 2.0 2.5 3.0 3.5 4.0

f(xt) = yi 4^3 % 6.928 J^/39 % 7.806 3^7 « 7.937 iy/lS % 6.778 0

,*■
j x(16 — x2)1/2 dx « Ax i(yo + n) + Z ^ i= 1

0.5[i(6.928) + (7.806 + 7.937 + 6.778)]

0.5[3.464 + 22.522]

12.993
b n— 1

note: Ax Z < f(x) dx < Ax Z

i= 1 ‘a i—0

,4
11.261 < j x(16 — x2)1/2 dx < 14.725

SIMPSON'S RULE
Simpson’s rule uses a series of parabolic arcs to approximate the given function f(x);
it is in this sense a method of quadratic approximation. Simpson’s rule provides a
closer approximation than the trapezoidal rule for any given number of subdivisions
of the interval over which the integral is to be evaluated.
452 / Integral Calculus

Figure 4.33

A parabola with vertical axis can be passed through any three points. The equa¬
tion of this parabola can be obtained from the coordinates of the points; however, for
the purpose of using Simpson’s rule this is not necessary, since the area under an arc
of the parabola can be evaluated without finding the equation of the parabola.
If a parabola with vertical axis is passed through three points (x0, y0), (xx, yx) and
(x2, y2), where x2 — xx = xt — x0 = Ax, then the area bounded by the parabola, the
x-axis, and the ordinates at x0 and x2 is given by

A = y (yo+ 4y‘ + y^

Again suppose the probiem is to evaluate J'J f(x) dx. Divide the interval [a, b] into
n equal parts, where n must now be an even number, each of length Ax, and denote the
points of division by a = x0, xl,..., xn = b; the corresponding ordinates are y0, yl,
..., yn; consider these ordinates in groups of three: y0, yx, y2, y2 , ^3, y4; y4, y5, y6;
...; yn-2, y„~u yn> and Pass a parabolic arc with vertical axis through each set of
three points (see Figure 4.33). The areas under these parabolic arcs are, respectively,

Ai = y (yo + 4yi + y2)

A2 = y (y2 + 4y3 + y4)

An - y (yn-2 + 4y„-l + y„)

and the sum of these areas is

Ax _
Y [^0 + ^yx + 2y2 + 4v3 + 2y4 + ••* + 2y„_2 + 4yn_1 + yn]

Simpson’s rule for approximate integration is thus

^ Ax
f(x) dx* — [y0 + 4y, + 2y2 + ■ • ■ + 2 y„_2 + 4 y„_ , + y„]
 4.7 Numerical (Approximate) Methods of Integration / 453

As in the case of the trapezoidal rule, as the number of divisions, n, increases, the sum
of the areas under the parabolic arcs more closely approximates the area under the
curve.

EXAMPLE

Evaluate \\ x(16 — x2)112 dx by Simpson’s rule using n = 4. (See Figure 4.33)

i 0 1 2 3 4

X; 2.0 2.5 3.0 3.5 4.0

/(*.) = yi 4^3 % 6.928 1^39 » 7.806 3^7 « 7.937 i^/15 « 6.778 0

| x(16 - x)1'2 dx « — [y0 + 4(j/i + y3) + 2y2 + y4]

« y [6.928 + 4(7.806 + 6.778) + 2(7.937) + 0]

^ -y [6.928 + 58.336 + 15.874]

% 13.523

note:

x(16 - x2)1'2 dx = [—4(16 - x2)3'2]! = 0 + i(12)3'2 = 8^/3 « 13.856

* 2

and, as noted above, Simpson’s rule gives a closer approximation than does the trapezoidal
rule. Note also that the exact answer is in fact within the bounds given by the trapezoidal rule.

TAYLOR S EXPANSION
Taylor’s theorem may be used to expand a function f(x) in an infinite series;
\dc f(x) dx may then be evaluated approximately by integrating the series term by
term and evaluating as many terms as necessary to obtain the required accuracy. For
this procedure to be valid, the series must converge to f(x) for all values of x in the
interval [c, d] over which the integral is to be evaluated. The required accuracy
usually can be obtained with fewer terms by expanding the series about a value of a
such that c < a < d.
Thus, providing the convergence conditions are met,

where, in general, c < a < d.

454 / Integral Calculus

EXAMPLE

1/2 x2 dx
Evaluate using Taylor’s series.
x + 1

1
/M = x + 1 20

2x(x + l)-x2 x2 + 2x T6 + T 9
J 11 (x + l)2 (x+Tp (i)2 25

, (2x + 2)(x + l)2 - (x2 + 2x)(2x + 2) 2 128

' ,X! " ' " (v • I)4 (I)3 “ 125

2
~(TTTf

f,„. . _ ~6(* + O2 = _ _J>_ 6 _ 1536

1 W“ (x + If (x + l)4 (If “ “ 625"

fi'y)ix\ — 24(x + 0 _ 24 24 24576

7 W (x + 1)8 (x + 1)5 f W “ (if “ Tiff

n\ n!
/("’(x) = (-!)" /'”>( i) = (-!)" 5 \n + 1
(x + 1)"+1 (J)

f f(x)dx = /(a)x + /'(«) (x - a)2 + (x - a)3 +

2!

+ ^ (X - u)"+1 +
(« + i)!
Letting a = i,

1/2 x2 dx JL x 4- ^ tv; _ 1 \2 , ^8 (x _ 1 \3
i x + 1 20 + 25 • 2! X ^ 125 • 3! ( ^

1/2
1536 24756
625 • 4!
(x - i)4 +
3125 • 5! (* ~ i)5 0

/ i \jl^ 128 /1\3 24576 /1\5\

\40 + 25^2M>JL + 125 • 3!\4/ ~ W + 3125 • 5!\4/ /

3
128
125 • 3!
1 re

% 0.025 + 0.0053333 + 0.000128

* 0.0304613

or 0.0305 to four decimal places.

note: Using standard forms,

 4.7 Numerical (Approximate) Methods of Integration / 455

1/2 1/2 1/2

1
dx = X — 1 + dx = —— x + In (x + 1)
X + 1 o \ x + 1 0
= (i — i + 1° i) - (0 — 0 + 0)
= -| + ln|
« -0.375 + 0.4055
% 0.0305 to four decimal places

EXAMPLE

Evaluate jo x3 dx using (a) trapezoidal rule, (b) Simpson’s rule, (c) Taylor’s series, and (d)
exactly.
Divide the range into 10 equal parts as summarized in the following table:

i 0 1 2 3 4 5

Xi 0 0.1 0.2 0.3 0.4 0.5

/(*<) = yt 0 0.001 0.008 0.027 0.064 0.125

i 6 7 8 9 10

Xi 0.6 0.7 0.8 0.9 1.0

/(*<) = 0.216 0.343 0.512 0.729 1.00

rb n— 1
(a) trapezoidal rule: f(x) dx % Ax i(yo + yn) + Z yt
a i= 1

1
x3 dx % 0.1 [i(0 + 1.000) + 2.025]

% 0.2525

(b) Simpson’s rule:

b Ax
/(x) dx % — [y0 + 4yt + 2y2 + 4y3 -f • • • + 4y2„_ x + y„]

, 1 01
X3 dx % — [(0 + 1.000) + 2(0.800) + 4(1.224)]
J0 3

« 0.2499
d
(c) J /(x) dx (x — a)3 + • • •
C

Letting a = j,
1
f(i)x + (* - i)2 + -f •••

/(x)-x3 f(i) = i

f'(x) = 3x2 /'(i) = |

/"(*) = 6x /"(i) = 3

/"' (x) = 6 /"'(i) = 6

/,,V|M = 0 /<IV>(i) = 0
456 / Integral Calculus

= 0.125 + 0.125

= 0.250
i
= i - 0 = 0.25
o

note: Both Simpson’s rule and the Taylor’s series expansion provide the exact answer for
this particular problem. Because of the nature of the approximation (quadratic), Simpson’s
rule is exact for any quadratic or cubic function. The Taylor’s series expansion is exact if all
further terms are zero—that is, if n terms are used and the nth derivative of /(x) is zero.
Because Simpson’s rule is exact, the answer could have been obtained by dividing the range
into only two parts:

i 0 1 2

Xi 0 0.5 1

/(*.-) = yi 0 0.125 1

0.5
x3 dx = [It 4(0.125)] = 0.25
'o 3

However, using n = 2 instead of n — 10 would reduce the accuracy of the trapezoidal rule:
i

x3 dx = 0.5[i(l) + 0.125] 0.3125

EXAMPLE

dx
Evaluate using (a) trapezoidal rule, (b) Simpson’s rule, (c) Taylor’s
1

series, and (d) exactly.

Divide the range into four equal parts as summarized in the following table:

i 0 1 2 3 4

Xi 2 3 4 5 6

i i i i
fM = yi i s IT Ti T5

(0.3333) (0.1250) (0.0667) (0.0417) (0.0286)

 4.7 Numerical (Approximate) Methods of Integration / 457

rb
n— 1
(a) trapezoidal rule:
* n
f(x) dx % Ax i(yo + yn)+ Z yt
i= 1

^ dx
--- « l[j(0.3333 + 0.0286) + 0.2334]
j X — I

% 0.4143

b Ax
(b) Simpson’s rule: /(x) dx ^ — [y0 + 4yx + 2y2 + • • • + 4yln-1 + yn]
3

dx
2 t ^ i[(0.3333 + 0.0286) + 4(0.1250 + 0.0417) + 2(0.0667)]

% 0.3874

f'(a) f"(a)
(c) f /(x) dx = f (a)x + —y1 (x — a)2 + J t 7 (x — u)3 + ■ • •
2! 3!

Letting u = 4,

•6 dx r(4)
/(4)x + ^ (x 4)2 + (x - 4)3 + f '(4> (x - 4 Y
J2 x2 - 1 3! 3!

f(X) = (x2 - 1) - 1

/'(*)= ~M*2 - l)_2(2x)

= —2x(x2 — 1)"2

f"(x) = —2(x2 — l)-2 — 2x( — 2)(x2 — l)“3(2x)

= (x2 — l)-3[ —2x2 + 2 + 8x2]

= 2(3x2 + l)(x2 - 1)~3

f"'(x) = 2(6x)(x2 - I)'3

-I- 2(3x2 + 1)(— 3)(x2 — l)~4(2x)

= (x2 — l)_4[12x3 — 12x — 36x3 — 12x]

= — 24x(x2 + l)(x2 - l)"4

/(IV)(x) = —24(3x2 + l)(x2 - 1) 4 — 24(x3 + x)( — 4)(x2 — 1) 5(2x)

= —24(x2 — l)~5[3x4 — 2x2 — 1 — 8x4 — 8x2]

= 24(5x4 + 10x2 + l)(x2 - l)"5

1
/(4) = 15

-8
/'(4) = 2
(15)
98
3
(15)
~ (96)(17)
f"( 4) (15)4

/•••( 4, -
458 / Integral Calculus

6
1 8 98
(x - 4)* 2 + (x - 4)'
^2 (15)2 • 2! (15)3 • 3!
6
(96)(17) (24)( 1441)
(x - 4)4 + (x - 4)5
(15)4 • 4! (15)5 • 5! 2

(24)(1441) \
(15)5 • 5! '' ’ j

^071^ (24)0441)
' (15)5 • 5! 1

« 0.2667 + 0.0774 + 0.0243

0.3684

dx x - 1
(d) 7 In = i[ln 4 - In i]
x2 - 1 X + 1

= i[ln 5 — In 7 + In 3]

«i[ 1.6094 - 1.9459 + 1.0986]

* 0.3811

note: -^ln (- -) + C.
2a \u + a)

It should be noted that the seemingly better results from part (b) are due to the truncation
effects of using only 4 terms in (c). Taylor’s series is more accurate than Simpson’s rule, if a
sufficient number of terms are computed.

PROBLEMS
Evaluate each of the following integrals approximately using a. trapezoidal rule, b. Simpson’s
rule, c. Taylor’s series, and d. exactly.
3 2
1. (x4 + 3x2) dx, n = 6 3. e2x dx, n = 4
•'o J _2

e(1,2)x dx, n = 4

Evaluate each of the following integrals by expanding the function in a Taylor’s series; com¬
ment on the interval of convergence and the accuracy of the evaluation.

5. (x4 — x3 + j) dx + x3 + 2x2 + 1 dx
*'0

6. (5x4 + yx3 + 6x) dx 8. — 6x + 3) dx

Answers to Odd-Numbered Problems

1. a. 78.218, b. 75.625, c. 75.6, d. 75.6
3. a. 35.832, b. 28.904, c. 27.173, d. 27.290
5. (exact, no problem of interval of convergence)
7. 12jf (exact, no problem of interval of convergence)
 4.8 Multiple Integrals / 459

4.8 Multiple Integrals

Multiple integrals are evaluated by successive partial integration. In this section
partial integration, the inverse of partial differentiation, is defined and its use in
evaluating multiple integrals is illustrated. The use of double integrals in evaluating
probabilities associated with joint-frequency functions is also discussed and
illustrated.
The definite integral JJ /(x) dx is defined with respect to the function/(x) over an
interval a < x < b. Similarly, the double integral

| f(x, y) dy dx
R

is defined with respect to the function /(x, y) over a bounded region, R, of the
xy-plane.
Just as the definite integral of f(x) can be interpreted in terms of area, the definite
double integral of /(x, y) can be interpreted in terms of volume. (See Figure 4.34.)
The function f(x, y) is assumed to be positive over the region R and the volume
computed is that below the surface z = f(x, y) and above the region R in the
xy-plane.
The evaluation of double integrals is most easily accomplished by successive
partial integration, which is the inverse of partial differentiation. That is, to evaluate a
double integral, a function of two independent variables is integrated with respect to
one of the variables while the other variable is considered as constant; the result of
this partial integration is then integrated with respect to the other variable. For this
purpose the double integral
b „ 0(x)
/ (x, y) dy dx or /(x, y) dy dx
1a Jh(x)

Figure 4.34
z
460 / Integral Calculus

where a and b are constants, can be written as the iterated integral

g(x)

f{x, y) dy dx
h(x)

To evaluate this expression,/(x, y) is first integrated partially with respect to y and

evaluated for the appropriate limits; the result is a function of x which is then
integrated with respect to x and evaluated for the appropriate limits. Similarly,

, g(y)
f (x, y) dx dy or
h(y)

is first integrated partially with respect to x and then with respect to y.

note: If the order of integration is reversed, new limits of integration must be

determined, that is,
,b g(x) .g(x) b
I /(x, y) dy dx =/= /(x, y) dx dy
Ja Jh( x) Jh(x) Ja

The limits of integration must involve only variables with respect to which integra¬
tion is to be performed subsequently, and the limits of integration must be rewritten
in this manner. Thus

/(x, y) dy dx = /(x, y) dx dy
Ja ' h(x) Jc J n(y)

where the new limits of integration are determined so that, for example, if y = g(x),
then x = m(y).

For functions of more than two independent variables, this process can be gener¬
alized using multiple integrals; there are relatively few applications requiring more
than triple integration. It is important to note that multiple integration is performed
from the inside outward; thus the first integral sign belongs to the last differential and
so forth. Note also that when an integration is performed with respect to a variable,
that variable is eliminated completely from the remaining integral, including the
limits of integration.

EXAMPLE

Evaluate the double integral JJ fF yjx/y dy dx

dy dx = [2x1/2y1/2]F dx
o Jx2 V y

— I (2x — 2x3/2) dx
■'o
4x5/2
x2 -

= d -})-o = i
 4.8 Multiple Integrals / 461

EXAMPLE

Evaluate the double integral jo jSa2 y2x dx dy.

Ja2-y2 s/ai — y2
A

x dx dy = dy
~T

ra a2 - y2...
u y
0 2
a
a2y y3l
2 6 0

o
K fl3i

II
\2 6/ 1

EXAMPLE

Evaluate the double integral jo jo x sin y dy dx.

n x . n
I x sin y dy dx = [ —x cos y]o dx
'0

= ( — x cos x + x) dx
‘'o
X
= — x cos x dx +
*'o

Integrating jo x cos x dx by parts,

u = x dv = cos x

du = dx v = sin x

and

x cos x = [x sin x]5 — sin x dx

= [x sin x + cos x]o

= (0-1)-(0 + 1)
= -2

Thus

71
x sin y dy dx = — (— 2) + —

4 H- re2
2
462 / Integral Calculus

EXAMPLE

Evaluate the double integral JLi jf 1/xy dx dy.

.1
|
J- i Jt
r O’ t
| — dx dy =
xy J
r [f /1 xy
l
dx dy

. i e*
- In X d
-J 1 - y i

dy

= M-i
= 1 - (-1) = 2

EXAMPLE

Evaluate the triple integral J2 y j£2 jonxez dz dx dy.

y2 In x

\*y\ dz dx dy = [ y f
•*1 Jy
[e*]o*x dx dy

2 y2
= j y (x — 1) dx dy
•'l •'v

y2
y dy
y _x

2 xy-y3-y + y2|^

V5 3y:
+ )'2U)'

y
12 8 3

= (-x — 6 + f) ~ (it - i + i)
47
24

Note that the integrand in the example above can also be written
f2 In x

yez dz dx dy
1
and y can be carried as a constant in the first two integrations. However, when an
integrand includes a factor which is constant with respect to one or more integra¬
tions, that factor can be moved outside the integral(s); usually it is more convenient
to separate the integrand in this manner.

EXAMPLE

Evaluate the triple integral jo jo6 josin(,r2 cos 9 sin (f> dr dO d(f>.
 4.8 Multiple Integrals / 463

a sin 0
rn,6>
r2 cos 9 sin </> dr d9 d(f) = sin <i> cos 9 |
COS 6 r2 dr d6 d(f)
Jo J0 J *0 Jn

a sin 0
sin 4> cos 9 d6 d(j)
0

. ft/6
a3
= sin <f) cos 9 sin3 9 d9 d(J)
J no 3

a
- f’ sin (f)[i sin4 9]o/6 d(j)
3 J0

a3 „ TL

sin </>(iV) #
12 J0
„ 71
a3
sin (f> d(f)
192 J0

a3
[-cos (f)]l
192
a3 n v a3
192 1 + ^ “ 96

Most applications of multiple integration involve definite integrals. This is pri¬

marily because any consideration of integration as the reverse process of differentia¬
tion is complicated in the case of multiple integration by the form of the constant of
integration. For example, when integration is with respect to y, x is constant and, in
general, the constant of integration is a function of x. Thus, since multiple integration
is the reverse of successive partial differentiations with respect to different variables,
the constants of integration are functions of one or more variables; determining these
functions requires quite complicated initial conditions. Most problems of this com¬
plexity are more conveniently stated in terms of differential equations, as discussed in
the next chapter.

PROBLEMS
Evaluate the following multiple integrals.
1 .2 a „ b .2 a
1. [ [ (x + 2) dy dx 7. j j x2y2z dz dy dx
Jn
o J•'o
n b J0 Ja

2 . 3 1 . 1 . 1-x
(x -I- y) dx dy 8. x dz dx dy
0 1 •J n
0 J ->,2
Jy2 n
JJ0

l-x l — y2
1
3. j j x112 dx dy z dz dy dx
y 0 •'o Jo
2 . x2 1 x x+y

y dy dx 10. f j” ex + y + z dz dy dx
o 1 Jn Jq Jq

. -1 2y 1 „ 2y
xy dx dy 11. I x2y dx dy
o Jy+i J - i J?

0 y
•1 rx2
6. ey,x dy dx 12. [ [ y2exy dx dy
o Jq J — 3 Jq
464 / Integral Calculus

f2 r x2
13. (xy + 1) dy dx
1 •'o

. 3 , x+ 1
14. (x + y) dy dx 17. x2 f exy dy dx
•'0 JX •'o *'o
1 x+ 1 . x+y
15. (x — y — 1) dy dx (x + y + z) dz dy dx
- 1 'x- 1 o

Answers to Odd-Numbered Problems

11 14
1. 5 11- T5
17 91
3- TO 13. jj
5. ji 15. -4
7. ia2b3(a> - h3) 17. i
9. 11

PROBABILITY AS A DOUBLE INTEGRAL

One of the most frequent applications of double integrals is in determining the

probability that two variables fall within specified ranges. For such applications, the
joint-frequency function f(x, y) must be defined.
A joint-frequency function for continuous variables x and y is a function f(x, y)
having the following properties:

1. f(x, y) > 0
2- J^oo l-oo f(x. y) dx dy = 1
3- 1«2 lo! /(x> y) dx dy = P(aI < x <b^, a2 < y < b2)

These properties are similar to those for a frequency function of one variable and
state that (1) probability is always nonnegative; (2) the probability of an event that is
certain to occur is one; and (3) the probability that the value of x is in the interval
(alt bx) and the value of y is in the interval (a2, b2) is given by the corresponding
integral. Geometrically, /(x, y) is a surface in three dimensions and the volume
under the surface lying above the rectangle determined by al < x < bl and
a2 < y < b2 is the probability that x and y will assume values corresponding to
points in this rectangle. (See Figure 4.29.) If f(x, y) is continuous, P(al < x <b{,
a2 < y < b2) = P(al < x < bu a2 < y < b2).

EXAMPLE

In a study of costs for a certain type of fluorescent tube, a company has determined that the
frequency function for size of order, x, and total cost for the order, y, is approximately

/(x, y) = - 1 < x < 6 0.1 -I- Q.9x < y < 0.1 4- l.lx

where x is in thousands of fluorescent tubes and y is in thousands of dollars. If the company

sets a price of $1.05 per tube, on what proportion of the orders will the company break even or
make a profit?
 4.8 Multiple Integrals / 465

6 . 1.05x
P( 1 < x < 6, 0.1 + 0.9x < y < 1.05x) = dy dx
0.1+ 0.9x 3.5

= 3^ | Moil+0.9* dx

(1.05x - 0.1 - 0.9x) dx

3.5

1 r6
= — (0.15x — 0.1) dx
3.5 Ji

= ~ [0.075x2 - 0.1x]i

= [(2 7 — ° 6) — (0.075 — 0.!)]

2.125
3.5

= 0.6071

6 0.1 + 1.1*
note: dy dx = — | [^]o;i + o;9*
0.1 + 0.9x 3.5

1 r6
= — [(0.1 + 1.1 xc) — (0.1 + 0.9x)] dx
3.5 J i

0.2x dx
3.5 J

1 216
= B[0-1x2]1

= ^[3.6-0.1]

= 1

EXAMPLE

The joint frequency function for the time (in microseconds) required for two switches to close
is given by

/(xj, x2) = 15e“(3xi + 5x2) Xj > 0, x2 > 0

(a) Find the probability that the first switch requires less than 0.1 microsecond and the second
switch requires more than 0.5 microsecond to close.
(b) Find the probability that the first switch requires more than 0.2 microsecond and the
second switch requires more than 0.5 microsecond to close.
466 / integral Calculus

0.1 „ oo

(a) P(0 < Xi < 0.1, x2 > 0.5) =15 f e (3xi + 5x2) dx2 dxi
J•'o
n n *
•'o.s

0.1

= -3 f [e"(3xi + 5x2)]£5 dx1

Jn 0
0.1
= 3 [ e~(3xi + 25) dxt
•'0
= [-e"(3Xl+25)]8 1

= —e~28 + e~25

= -0.061 + 0.082

= 0.021
00 , 00

(b) P(xt > 0.2, x2 > 0.5) = 15 [ f + dx2 dx,

0.2 ^0.5
dn i Ja c

oo
= -3 [ [e~(3xi + 5x2)]q.5 dxt
*'0.2

e-(3x,+2.5) dX]
= 3
0.2

= [-e~(3xi+2-5)]£:

= 0 + e~3A

= 0.045
00 00 00
note: I | 15e~{3x' + 5x2) dxx dx2 — | [ — 5e_(3xi + 5x2)]o dx2
J0 jo *' 0
OO
= 5e 5x2 dx2
*'o

= [-e-5x>]$

= 0 - (- 1)

= 1
Note also that/(xl9 x2) = f(xt)f(x2) = (3c~3xi)(5e_5x2). Thus

5X2
P{x 2 > 0.5) = 5jr«- dx2 -l-e-
0.5

r0'1 e - 3xi
P(x 1 < 0.1) = 3 dx! = [-«'- 3xil0.1 _ .,-0.3
•
0

P(x, > 0.2) = 3J [V

0.2
- 3xi
dxx = [-«•

These probabilities are referred to as marginal probabilities, and the joint probabilities ob¬
tained above can be obtained as the products of the corresponding marginal probabilities.
This is characteristic of probabilities of independent events.

EXAMPLE

Show that/(x, y) = X2e 0 < x < y, is a joint frequency function,

/(x, y) > 0
 4.8 Multiple Integrals / 467

X2e 2y Xy]yo dy

dy

Integrating by parts,
u = y dv = e~ky

= [~Xye~Xy - e~Xy]$

= (0-0) -(0-1)

= i

PROBLEMS
1. Show that/(x, y) = |(x + y)* 1 2 3 4 5 6 7 8, 0 < x < 1, 0 < y < 1 is a frequency function.
2. Show that /(x, y) = ^f(y — x)1/2, 0<x<y<lisa frequency function.
3. Show that /(x, y) = 8xy, 0 < x < 1, 0 < y < x, is a frequency function and find a.
P(0 < x < 0.5, 0 < y < x) and b. P(0.2 < x < 0.5, 0.1 < y < x).
4. Show that/(x, y) = 12(y — x)2, 0<x<y<lisa frequency function and find a. P(0.5 <
x < y < 1) and b. P(0 < x < y < 0.5).
5. Determine the value of k for which

/ (x, y) = k(x + y) 0<x<4, 0<y<6

is a frequency function.
6. Determine the value of c for which

/(x,6y) = — 2<x<4, 5<y<9

c

is a frequency function.
7. The joint frequency function for the errors in measuring two variables is given by

/(x, y) = - 0 < y < x < 1

X

a. Find the probability that the measurement error x is between 0 and 0.5 and the measure¬
ment error y is between 0 and 0.2x. b. Find the probability that the measurement error x is
between 0.8 and 1 and the measurement error y is between 0.8 and 1. c. Show that/(x, y) is
a frequency function.
8. The waste from a chemical plant consists of soluble and nonsoluble substances. The nonsol¬
uble substances are of two types, A and B. The plant’s engineers have studied data and have
determined that the composition of the nonsoluble waste on one day does not depend on its
composition on preceding days. The joint frequency function for the proportion of all
soluble substances (A and B), denoted by x, and type B nonsoluble substances, denoted by
y, is given by

/(x, y) = 15xy2 0 < x < 1, 0 < y < x

468 / Integral Calculus

a. For what proportion of the days does the waste contain more than 20% of type B
nonsoluble substances? b. For what proportion of the days does the waste contain less than
50% of nonsoluble substances? c. For what proportion of the days does the waste contain
more than 50% nonsoluble substances and more than 20% type B substances? d. Show
that f(x, y) is a frequency function.

Answers to Odd-Numbered Problems

3. a. 0.0625
b. 0.0567
5- k =
7. a. 0.1
b. 0.2
Differential
Equations

5.1 Introduction

In many problems the relationships between or among variables are most appro¬
priately stated in terms of rates of change. The rate of change of any given variable
may be expressed as a function of the rates of change or the values of other variables.
For example, in physics equations involving rates of change in energy are used in the
derivation of the laws of conservation of energy; in economics it is assumed that the
rate at which price approaches its equilibrium value depends on the magnitude of
the discrepancy between the quantities supplied and demanded.
Rates of change can be stated in either of two mathematical forms, depending on
whether time (or, more generally, any variable with respect to which changes are
considered) is viewed as continuous or discrete. When changes are considered to
occur continuously or instantaneously, rates of change are stated as derivatives and
equations involving them are differential equations. When changes are considered as
occurring discretely or discontinuously at certain points in time or as average
changes over a period of time, rates of change are stated as differences in the values of
variables at different points in time and equations involving them are difference
equations. Differential equations are the limiting case of difference equations as the
time period between changes or over which an average change is computed
approaches zero.
Differential equations, that is, equations involving derivatives, are discussed in
this chapter; difference equations, that is, equations involving differences between
values of variables at different points in time or with respect to different values of
another variable are discussed in Chapter 6.
In this chapter, the definition and classification of differential equations are dis¬
cussed and types of solutions of ordinary differential equations are considered and
illustrated. A classification of differential equations of first order and first degree is
given and the solutions of these classes of equations are discussed. Applications of
differential equations in economic models are illustrated.

469
470 / Differential Equations

5.2 Definition and Classification

of Differential Equations
A differential equation is an equation that involves derivatives of one or more func¬
tions. Differential equations are classified according to type, order, and degree.
If a differential equation involves derivatives of a function of one independent
variable, it is an ordinary differential equation; if it involves partial derivatives of a
function of two or more independent variables, it is a partial differential equation.
The order of a differential equation is the order of the highest-order derivative
occurring in the equation. The degree of a differential equation is the power of the
highest-order derivative occurring after the differential equation has been ration¬
alized to remove fractional power derivative terms.

EXAMPLE

is an ordinary differential equation of first order

and first degree

is an ordinary differential equation of first order

x dy — y dx = 0
and first degree

d2y is an ordinary differential equation of second

+ y = 0
df2 + order and first degree

is an ordinary differential equation of first order

( and second degree

dz dz is a partial differential equation of first order

x—I- -
ex Cy and first degree

d2u d2u is a partial differential equation of second order

~2 + -r~2 + x + y-uz = Q
dx2 dy" and first degree

d3z ld2zxl is an ordinary differential equation of third

— XZ -y—- + 10 = 0
dx 3
—+xw dx order and first degree

W2 du is a partial differential equation of third order

+ uA — x2u — 15=0
dx3 dx i and second degree

dfy d2y dy is an ordinary differential equation of fourth

+ x' + *y dx lOOy = 0
dx4 dx2 order and first degree

d3y is a partial differential equation of third order

ex> + xy iu8 -3ui=0 and first degree

df is an ordinary differential equation of first order

5xy3 + 2 = 0
dx and third degree

'd*yX4r d2y is an ordinary differential equation of fourth

— 6x
dx" dx' order and fourth degree

Many of the equations in Chapter 4 are differential equations, since any equation
involving at least one derivative is a differential equation. If a differential equation
can be written in the form

its solution is obtained in a straightforward manner by (successive) integration, as in

Chapter 4.
 5.3 Solutions of Ordinary Differential Equations / 471

EXAMPLE

Find the general solution of the differential equation — = cos x + 2x.

ax

If ^ = cos x + 2x,
ax

y — sin x + x2 4- C

EXAMPLE

d2y
Find the general solution of the differential equation = 20x3 - '
dx2

If ^ = 20x3-i.
dxz xz

dl
ax
= 5 x‘ + xi + C,
and
y = x5 + In x + C1x + C2

The methods discussed later in this chapter are appropriate for solving several
types of differential equations whose solutions cannot be obtained by straightfor¬
ward integration. Five types of differential equations of first order and first degree are
discussed in some detail; second-order differential equations and first-degree differ¬
ential equations of higher order are discussed briefly.

5.3 Solutions of Ordinary Differential Equations

A solution of an ordinary differential equation is a function not containing deriva¬

tives or differentials, which satisfies the differential equation. Such a solution may be
expressed as an explicit or implicit function and may be a general or particular
solution, depending on whether the constants of integration have been specified.
The general solution of an nth-order differential equation is a solution containing n
independent arbitrary constants of integration. For example, the general solution of
an nth-order differential equation could be of the implicit form

/(x, y) + Cj + c2 + ••• + c„ = 0

or the explicit form

f(y) =/(*) + c, +c2 + ■■■ + c„

A particular solution of a differential equation is a solution that can be obtained

from the general solution by giving specific values to the arbitrary constants of the
general solution. Thus, in either the implicit or explicit form of the solution above, a
particular solution is obtained if cls c2, .cn have specific values.
Some differential equations also have singular solutions that cannot be obtained as
particular solutions; these exceptional cases are not considered here.
472 / Differential Equations

As in simple integration, the constants of integration of a differential equation are

specified by boundary conditions or initial conditions: a condition of the form y = y0
when x = x0 is a boundary condition; the special case y = y0 when x = 0 is an initial
condition.
Clearly, the problem of the solution or integration of a differential equation is an
extension of the problem of simple integration. As in simple integration, frequently a
particular solution of a differential equation is needed for practical applications. To
obtain such a particular solution it is usually necessary first to find the general
solution of the differential equation and then to determine the arbitrary constants
from the boundary conditions of the problem. A constant of integration may be
written in various forms, such as C, 2C, C* 2, yfc, ec, In C; frequently solutions of
differential equations may be expressed in simpler form by proper choice of the form
of the constant or constants of integration. When the form of a constant of integra¬
tion is changed, the constant frequently continues to be written using the same
notation, with the understanding that it is a “generic constant.” Until the numerical
value of a constant is determined, its form has no real meaning and thus the constant
is written in the simplest way possible.
In each of the following examples, a differential equation and its solution are given
and the solution is shown to satisfy the differential equation. Methods for obtaining
solutions of differential equations are then discussed.

EXAMPLE

x2 C . . dy
Show that y = —- H— is a solution ol x — + y = x 2
3 x ax

If y = x3/3 + C/x,

dy 2x C
dx 3 x2

and thus

dy
xTx + y

EXAMPLE

d2y
Show that y = c^ekx -I- c2e~kx is a solution of - k2y = 0.
dx2
If y — cy ekx + c2 e~kx,

dy
Ci kekx — c2ke -kx
dx

d2y 2 „-kx
= Ci k2ekx + c2k2e
dx2
 5.3 Solutions of Ordinary Differential Equations / 473

and thus

cxk2ekx + c2k2e~kx — k2(cxekx + c2e~kx) = 0

d2y
— k2y
dx“

EXAMPLE

,2 d2y dy 1
Show that 4y = l/(3x) + c^x5 + c2 x is a solution of x1 — 5x — + 5y
dx dx .X
If 4 y = l/(3x) + c^5 + c2x,

4fx = -3? + 5c‘X'4 + C2

41AA + 20c‘x3
and thus

1
\(k+ 20cix5) - *( ” Tx+ 5cix5 + CjX)+ + Ci x5 + C2X X

.2^1 - 5x ^ + 5y
dx2 ~"dx

(A + A + A) - + (5 - + i)ci x5 + (—| + i)c2x = -

X X

1
0 0
X

EXAMPLE

d2y dy
Show that y = c1ex + c2 e 2x is a solution of + — — 2y = 0 and find a particular solution
dx2 dx
dy
that satisfies the condition y = — = 1 when x = 0.
dx
- 2x
I f y = c1ex + c2e

dy
Y~ = cxex -lc2e 2x
dx

d2y 2x
= cx ex + 4c2c
dx2

and thus

cx ex + 4c2e~Zx + cxex — 2c2e-2x — 2cxex — lc2e~2x = 0

d2y dy
dx2
+ dx
-r-.-'iy
474 / Differential Equations

If y = y' = 1 when x = 0,

ci + c2 = 1

c{ — 2c 2 = 1

c2 = 0

Ci = 1

and thus the particular solution is y — ex.

EXAMPLE

1dy\2 dy
Show that y = 2Cx + C is a solution of — + 8xJ = 16x2y and find a particular
\dxj dx
solution that satisfies the condition y = — 1 when x = 1.
If y = 2Cx2 + C2,

dy
4 Cx
dx

and thus

16C2x2 + 8x3(4Cx) = 16x2(2Cx2 -l- C2)

dyV + $xHd^ 16x2y

dx dx,

If y = — 1 when x = 1,

-1 = 2C + C2

C2 + 2C + 1 = 0

(C + l)2 = 0

C= -1,

and thus the particular solution is y = 1 — 2x2.

EXAMPLE

Show that (x — d)2 + y2 = c2 is a solution of

2y , IM2 + 1 0
dx2 \dxj

dy
and find a particular solution that satisfies the condition y 3, and —f when x = 5.
dx
If (x - cj2 + y2 = c2,

2y%= ~2(x ~ ^

dy x — Ci
1

dx
^

>2
1

d2y 2 + (x - Ci)(i)[c2 - (x - Cj)i2] ri/2i(-2) (x - Ci)

<7
1

dx2 c2 - (x - Ci) 4
-1 _ (x - CX)2_
Jc2 - (x - Ci)2 [c2 - (x - Ci)2f12
 5.3 Solutions of Ordinary Differential Equations / 475

and thus

(x - cx): X — Ci

-1 - + + 1=0
c2~ (x- cx)2 Jc2 - (x - Cly

d2y 'dyx 2
y dx' dxi

If v = 3 and ~ = — f when x = 5,
dx

3 = Jc2 - (5 - cx)2

4 _ 5-Cj

3 Jc2 - (5 - c,)2

5 - Ci =4

Ci = 1

c2 = 25

and thus the particular solution is (x — l)2 + y2 — 25.

PROBLEMS
State the type, order, and degree of each of the following differential equations.

d2y d2z dz
1. + 6x + y = 0
dx2 dx2 dy

2. x2 dy = xy2 dx d2y dy
8. + --3x = 0
dx2 dx
d2y^ 2
3 + 5 xy = 0
dx* d2yx 2 dy ,
v dx* ~Tx = Ay
dz dz
4 rx + ¥y = xyz d3y
10. = 10y
dx3

6. - 4x,

11. Verify that y2 = Cx + iC3 is a solution of y = 2x — + y2(—) and find the particular
dx \dxj
solution if / = 1 when x = 0.

d2y (dy \2
12. Verify that (x - cj2 + y2 = c2 is a solution of y +1 — 1 +1=0 and find the particu¬

lar solution if y = 3 and / = — f when x = 5.

ei)
1 d2y
13. Verify that In y = c2c* + c2 e x + x2 + 2 is a solution of = x2 — In y.
y dx2

d2x
14. Verify that x = cos 2r + 2c2 cos 3r + 3c2 sin 31 is a solution of ^2 + 9x = 5 cos 21

r , d3y d2y dy
15. Verify that y = (cl + c2 In x)^x + c3 is a solution of 4xz — + 8x + — = 0.
dx: dx2 dx
476 / Differential Equations

d2y 1 dy
16. Verify that y = x3 + c1x2 + c2 is a solution of - 3x = 0 and find the particu-
dx2 x dx
lar solution if y = 1 when x = 1 and y = 5 when x = 2.

d2y
17. Verify that y = cos x + c2 sin x — 3 cos 2x is a solution of + y = cos 2x.

dy
18. Verify that y = e *(x + C) is a solution of + y = e x.

19. Verify that y = cx x + — + c3 is a solution of + - -r-^ = 0-

20. Verify that y2 = cxx2 + c2 x is a solution of x2y + [X^c~ = ° anC* a particu_

dy
lar solution that satisfies the condition y = 2 and — = 1 when x = 1.
dx

Answers to Odd-Numbered Problems

1. ordinary, second order, first degree 7. partial, second order, first degree
3. ordinary, second order, second degree 9. ordinary, second order, second degree
5. partial, first order, second degree 11. y2 = 2x + 1

5.4 Differential Equations of First Order

and First Degree
In this section differential equations of first order and first degree are classified as
separable, homogeneous, exact, linear, or linear in a function. Methods of solution
for these types of equations are discussed and illustrated.
A differential equation of first order and first degree can be written in the form

di
= F(x, y)
dx

If F(x, y) is a constant or a function only of x, the differential equation is solved by

the usual methods of integration; if F(x, y) is in fact a function of both x and y, the
methods discussed below are appropriate for solving the differential equation.
An alternative form, frequently more useful for classifying differential equations of
the first order and first degree, is

M(x, y) dx -I- N(x, y) dy = 0

There is one difference between these equations which should be noted. In the first
equation y is the dependent variable and x is the independent variable; thus the
solution expresses y as a function of x and an arbitrary constant. In the second
equation the relationship is expressed implicitly and the choice of the dependent and
independent variables is a matter of convenience; the solution is frequently also
expressed as an implicit function.
Methods for the solution of the following types of differential equations of the first
order and first degree are discussed in considerable detail; note that facility in solving
a differential equation of any order and degree depends to a very considerable extent
on classifying it correctly.
 5.4 Differential Equations of First Order and First Degree / 477

1. Separable differential equations: M is a function only of x and AT is a function only

of v, so the equation is of the form M(x) dx + N(y) dy = 0.

2. Homogeneous differential equations: M and N are homogeneous functions of the

same degree of homogeneity.
5 5
3. Exact differential equations: — (M(x, y)) = — (N(x, y)) so the equation is of the

dF cF
form — dx + — dy = 0, where F(x, y) is the solution of the differential equation.

4. Linear differential equations: The equation is of the first degree in y and — or in x

dx

and ~ and thus is of the form ^ + yP(x) = Q(x) or ^ + xP(y) — Q(y).

5. Differential equations linear in a function of y or in a function of x: The equation is

of the first degree in /(y) and ~ f (y) or in x and ~ f (x) and thus is of the form

Yy f(y) +f(y)p(x) = Q(x)or f(x) -fix)P{y) = Q(y)•

SEPARABLE DIFFERENTIAL EQUATIONS

If a differential equation can be written in the form

M(x) dx + N(y) dy = 0

where, as indicated, M is a function only of x and N is a function only of y, then the

variables are said to be separated and the general solution of the differential equation
is obtained by the usual methods of integration:

j M(x) dx + j N(y) dy — C

and
f(x) + giy) = c

EXAMPLE

dy
Solve the equation (1 + x2) + xy = 0.
dx

(1 + x2) dy + xy dx = 0

1
dx + - dy — 0
1 + x2 y
The variables are separated and, integrating,

i In (1 + x2) + In y = C

In (yv/T+ x2) = C
478 / Differential Equations

Note that the constant in the last equation is actually ec in terms of the preceding equations.
As mentioned in the discussion of constants of integration above, until a constant is evaluated
to obtain a particular solution, its form is arbitrary and the constant is thus written in the most
convenient way.

EXAMPLE

Solve the equation y2 dx — (1 — x) dy = 0.

1 , 1 ,
-- dx H—i dy = 0
x - 1 y2

The variables are separated and, integrating,

In (x — 1) — = C

y In (x — 1) — 1 = Cy

y In [C(x — 1)] = 1

or

C(x — 1) = e1,y

EXAMPLE

If the interest rate is lOOz % compounded continuously and A is the amount at any time
(principal plus any accumulated interest), then

dA
= iA
dt

and, separating the variables,

dA
— = idt
A

and, integrating,

r dA r

J ~a J
= idt

In A = it + c

A = eit+c

or

A = celt

If A = T0 at t = 0, then c = A0 and

A = Aq elt

note: The formula for discrete compounding is given by

nk

y = x l +
 5.4 Differential Equations of First Order and First Degree / 479

where y is the amount after n years (principal plus any accumulated interest) and interest rate
is i% per year compounded k times per year. In the notation of this section, the discrete
formula can be written
kt
A — A0\ 1 + - ]

and the formula for the continuous case can be obtained by taking the limit as k oo of the
formula for the discrete case, since

1
e = lim 1 +
k~* oo

EXAMPLE

The relationship between net profit P and advertising expenditure x is such that the rate of
increase of net profit as advertising expenditure increases is proportional to a constant minus
net profit. Find the relationship between net profit and advertising expenditure if P = P0 when
x = 0.

dP
k(a - P)
dx

— In (a — P) = kx + C

In (a — P) = —kx + C

a- P = Ce~kx

P = a - Ce~kx

Figure 5.1

P
480 / Differential Equations

so

C = a — P0

and

P = a — (a — P0)e~kx

Thus net profit is P0 with no advertising expenditure and increases, as advertising expenditure
increases, toward an (asymptotic) maximum of a. (See Figure 5.1.)

HOMOGENEOUS DIFFERENTIAL EQUATIONS

A differential equation of the form

M(x, y) dx + N(x, y) dy = 0

is said to be homogeneous if M(x, y) and N(x, y) are homogeneous functions of the

same degree in x and y. Recall that F(x, y) is a homogeneous function of degree n in x
and y if and only if F(kx, ky) = knF(x, y), where k is any constant. When a differential
equation is homogeneous, its variables can be separated by the substitution

y = vx

dy — v dx + x dv

or, equivalently, by the substitution

x = vy

dx — v dy + y dv

The resulting differential equations can then be written, respectively, in the form
M(x) dx + N(v) dv = 0

or in the form

M(v) dv + N(y) dy = 0

and solved by the usual methods of integration. The general solution of the original
equation is then obtained by substituting, respectively,

y x
v - or v = —
x y

in the solution of the separable differential equation.

note: This procedure can be justified mathematically as follows. If

M(x, y) dx + N(x, y) dy — 0, M(x, y) and N(x, y) are homogeneous of the same
degree n and N(x, y) ± 0, then
dy M(x, y)
dx N(x, y)

But, since M(x, y) = t~nM(tx, ty) and N(x, y) = t~nN(tx, ty),

dy M(tx, ty)
dx N(tx, ty)
 5.4 Differential Equations of First Order and First Degree / 481

for any value of t. If t = 1/x, then

Ml](i, A
dy \ x)
dx
N\(i,y)
X

and the differential equation can be written

dx X,

, dy dv , .
Substituting y vx and — = v + x —, this equation can be written
dx dx

dv . .
v + x— = F(v)
dx

(v — F(v)) dx + x dv — 0
dx dv
— + - = 0
X V F(v)

and the variables x and v are separated. If v — F(v) = 0, then x dv = 0, dy — v dx = 0,

and M(x, y) dx + N(x, y) dy = 0 has the simple form y dx — x dy = 0.
A similar argument shows that the substitution x = vy also results in an equation
in which the variables are separable.

EXAMPLE

Solve the equation (y2 — xy) dx + x2 dy = 0.

The equation is homogeneous (degree 2). Substituting y = vx, dy = v dx + x dv,

(v2x2 — vx2) dx + x2(v dx + x dv) = 0

v2x2 dx + x3 dv = 0

- dx H—~ dv = 0
x v

The variables are separated and, integrating,

In x-= C
v

Substituting v = y/x,

In x — - = C
y
x
^ In x — C

EXAMPLE

Solve the equation (x + y) dx + (x — y) dy = 0.

482 / Differential Equations

The equation is homogeneous (degree 1). Substituting x = vy, dx = v dy + y dv,

{vy + y)(v dy + y dv) + (yy — y ) dy = 0

(vy2 + y2) dv + (v2y + Ivy — y) dy = 0

y2(v + 1) dv + y(v2 + 2v — 1) dy = 0

v + 1 , 1 , A
dv + - dy = 0
tr + 2t> — 1 y

The variables are separated and, integrating,

i In (t?2 + 2t; — 1) + In y = C

y(r2 + 2t> — 1)1/2 = C

y2(u2 +2v — \) = C

Substituting v = x/y,

x2 + 2.xy — y2 = C

EXAMPLE

The relationship between the manufacturing cost per item, M, and the number of types of items
manufactured, N, is such that the rate of increase of manufacturing cost as number of types of
items increases is equal to the ratio of the cost per item plus the number of types of items
divided by the number of types of items. Find the relationship between manufacturing cost per
item and number of types of items manufactured if M = M0 when N = 1.

dM _M + N
dN ~~N-

N dM = (M + N) dN

M = vN

dM = v dN + N dv

vN dN + N2 dv = vN dN + N dN

dN

v = In N + C

Substituting v = M/N,

M
= In N + C
aT
M = N In N + NC

M0 = 0 + C

c = M0

and

M = N In N + NM0

M = N(M0 + In N)

(see Figure 5.2).

 5.4 Differential Equations of First Order and First Degree / 483

Figure 5.2

EXAMPLE

Suppose that the rate of increase in the cost of ordering and holding, y, as the size of the order, s,
increases is equal to the ratio of the sum of the squares of the cost and the size divided by twice
the product of the cost and the size. Find the relationship between the cost of ordering and
holding and the size of the order if y = 3 when s = 1.

dy y2 + s2
ds 2sy

2sy dy — (y2 + s2) ds = 0

y vs

dy v ds + s dv

2vs2(v ds + s dv) — (v2s2 + s2) ds 0

(v2s2 — s2) ds + 2vs3 dv 0

s2(v — 1) ds + 2t;s3 dv 0

ds 2v .
- + -2-r dv= 0
s v — 1
/

In s + In (v2 - 1) = c
s(i;2 — 1) = c

Substituting v = y/s,

y' -1 =c

2 2
yl — sL = Cs

y (Cs + s2)1/2

so

3 = (C + 1)1/2
484 / Differential Equations

Figure 5.3

and

C = 8

y = (8s + s2)1/2

(see Figure 5.3).

PROBLEMS V

Find the general solution for each of the following differential equations.

1. (x2 + y2) dx — 2xy dy — 0 dy 2x — y

14.
dx x + 4y
2. (xy-x2)g = y2
15. dy = 2xy dx

3. y3 dx — x3 dy = 0 16 ldP = P2 tan 6
p dO p2 + 1
4. (y + 3) dx + cot x dy = 0
17. r dO + 0 dr = 2 dr
_ dx .
5. — = 1 — sin 2f
dt 18. dp + p tan 9 d6 = 0

d3y dr
= c 19. (r2 + 1) d9 + = 0
dx3 sec2 9

d2y 1 x
20. (xy — y2) dx + ^y2 ——j dy = 0
dx2 x2

8. sec x cos2 y dx — cos x sin y dy y2

21. — dx + (x — y) dy = 0
9. sin x cos2 y dx + cos2 x dy = 0
22. (xy3 — x) dx + xy2 dy — 0
10. dr + r cos 9 d9 = 0
23. tan x sin2 y dx + cos2 x cot y dy — 0
11 dl = * ~ >’
dx x + y ^, dy y
24. x --y — x sin - = 0
dx x
12. (xy2 — x) dx + (x2y + y) dy = 0
25. sin x cos y dx + cos x sin y dy = 0
dy xy + y
13.
dx x + xy
 5.4 Differentia! Equations of First Order and First Degree / 485

dy y 31. (x3 — y3) dx + xy2 dy = 0

26.
dx X
32. (x + y) dx + x dy = 0

27. x3 - 2y3 + 3xv2 ^ = 0

dx 33. ^
dx

28. dJ_ 34. x(2y — 3) dx + (x2 + 1) dy = 0

dx
35. (x + y) dx = x dy
29. x dy — y dx = 0
36. x dy — y dx = y/xy dx
30. (x2 + 2y2) dx = xy dy

Find the particular solution for each of the following differential equations under the given
conditions.

37. ~ = —, y = 3 when x = 1
dx x

38. xy dx -I- V'l + x2 dy = 0, y = 1 when x = 0

39. (x2 + y2) dx = 2xy dy, y = 0 when x = 1

40. x dy -f- 2y dx — 0, y = 1 when x = 2

41. (xey,x + y) dx = x dy, y = 0 when x = 1

42. y2(y dx — x dy) + x3 dx = 0, y = 3 when x = 1

43. x dx — 4y dy = 0, y = 2 when x = 5

44. x(y + 1) dx -f- y(x -I- 1) dy = 0, y = 1 when x = 0

45. (x — y/xy) dy = y dx, y = 1 when x = 4

46. (x + 4)(2y + 6) dy 4- xy2 dx = 0, y = 16 when x = 0

47. (xy2 + x2y) dy — xy2 dx = 0, y = 1 when x = 6

48. y(x2 4- 6) dy + x(y2 + 1) dx, y = 3 when x = 0

1 ey2
49. — dx H—r-dy = 0, y = 0 when x = 1
xy x — 1

dy y y
50. -7— = —b tan -, y = n when x = 6
dx x x

51. cot y dx + cot x dy = 0, y = 0 when x = 0

52. The relationship between price, p, and quantity demanded, x, is such that the rate of decrease
in demand as price increases is proportional to the quantity demanded and inversely
proportional to the price plus a constant. Find the demand function if p = p0 when x = 1.
Sketch the relationship obtained.
53. The rate of increase of total cost, y, as number of units manufactured, x, increases is
proportional to the number of units manufactured plus a constant and inversely propor¬
tional to the total cost. Find the cost function if y = y0 when x = 0. Sketch the relationship
obtained.
54. The rate of increase in sales, s, as advertising effort, x, increases is equal to a constant plus
advertising effort. Find the relationship between sales and advertising effort if s = s0 when
x = 0. Sketch the relationship obtained.
55. The relationship between revenue, R, and quantity demanded, x, is such that the rate of
increase in revenue as quantity demanded increases is equal to twice the cube of the
revenue minus the cube of the quantity demanded, all divided by three times the product
of the quantity demanded and the square of the revenue. Find the relationship between
revenue and quantity demanded if R = 0 when x = 10. Sketch the relationship obtained.
486 / Differential Equations

56. The relationship between average cost, y, and number of units produced, x, is such that the
change in average cost as the number of units increases is equal to the ratio of the number
of units minus average cost divided by number of units. Find the relationship between
average cost and the number of units produced if y = § when x = 1. Sketch the relation¬
ship obtained.
57. The rate of increase in cost, y, as the number of units manufactured, x, increases is equal to
the ratio of twice the square of the cost minus the square of the number of units divided by
the product of the cost and the number of units. Find the relationship between cost and the
number of units manufactured if y = 3 when x = 1. Sketch the relationship obtained.
58. The rate of increase of sales volume, s, as price, p, decreases is proportional to sales volume
and inversely proportional to price minus a constant. Find the relationship between sales
volume and price if s — s0 when p — p0.
59. If interest is compounded continuously, a. find the amount available after 10 years if $5000
is deposited at 4%; and b. find the amount available after 20 years if $20,000 is deposited at
6%. (Compare the answers with those obtained for the example on page 118 and Problem
2 on page 123, respectively.)
60. If growth is continuous and the rate, r, is proportional to the number, N, present in the
population, then

dN
= rN
dt

If AT = N0 when t = 0, obtain a formula for the number in the population at time t.

(Compare with the discrete case on page 119.)
61. If the growth rate is continuous at the rate of 5 % per year and the original number in the
population is 200, what is the size of the population after 10 years?
62. A manufacturer has found that the change in the cost of distribution, D, as sales, S, increases
is equal to a constant times sales plus another constant. Find the relationship between the
cost of distribution and sales if D = 0 when 5 = 0. Sketch the relationship obtained.
63. The rent for an apartment (two-bedroom, standard furnishings) in a college town varies
with the distance of the apartment from campus. Suppose this relationship is given by

dy k
—V a, 1 < x < 10
dx x

where y is monthly rent (in dollars) and x is distance (in miles) and k and a are constants. If
y = 225 when x = 1, find y as a function of x and sketch the relationship obtained.
64. The relationship between the cost of operating a warehouse and the number of gallons of
oil stored in the warehouse is given by '

dy /
— = kx + a
dx

where y is the monthly cost of operating the warehouse (in dollars) and x is the number of
gallons of oil in storage. If y = y0 (fixed cost) when x = 0, find y as a function of x and
sketch the relationship obtained.

Answers to Odd-Numbered Problems

1. x2 — y2 = Cx V
13. y - x + In c
X
3. x2 — y2 = Cx2y2
15. y = Cexl
5. X == t + i cos 2r + C
17. r(0 -2) = C
7. y == —In x + C! x + c2
19. tan 6 + In + 1 = c
9. sec: x + tan y = C
21. y = Cey/X
11. y2 + 2xy — x2 = C
23. tan:2 x — cot2 y = c
 5.4 Differential Equations of First Order and First Degree / 487

25. cos y = C’ sec x , x

47. In y-(-6 = 0
27. x3 + y3 =- Cx2 y
49. 2x3 — 6 In x + ?>ey2 — 5 = 0
29. y = Cx
51. cos y = sec x
31. x V3/x3 =C
53. y = (ax2 + 2 abx + yo)1/2
33. ey - ex = c
55. F = (10x2 - x3)1/3
35. x = Ceylx
57. y = ^/8x4 + X2
37. y = 3x

39. yz2 2 -X 59. a. $7460

= x-

= l b. $5421.69
41. In x 4- e~
61. 329.8
43. x2 — 4y2 = 9
63. y = 225 4- a — ax — k In x
45. 2 / - - In v = 4
V y

EXACT DIFFERENTIAL EQUATIONS

The total derivative of a function of x and y, say F(x, y), is given by

j , x dF J dF J
dF(x, y) = — dx + — dy

Thus the differential equation

dF dF
dx + dy = 0
dx dy

has the general solution F(x, y) = C and is said to be an exact differential equation.
A differential equation of the general form

M(x, y) dx + N(x, y) dy — 0

is said to be an exact differential equation if M(x, y) dx + N(x, y) dy is the total

derivative of some function F(x, y); M(x, y) and N(x, y) are then the partial deriva¬
tives of F(x, y) with respect to x and y, respectively.
If the second-order mixed partial derivatives of F(x, y) exist and are continuous,
then

d_(dF\ _ _d/dF\
dy \ dx / ex \ cy J
Thus if a differential equation of the form M(x, y) dx -I- N(x, y) dy = 0 is exact,

3 d
- M(x, y) = - N(x, y)

It can be shown that this is also a sufficient condition for exactness; that is,

5
M(x, y) N(x, y)oM(x, y) dx + N(x, y) dy — 0 is exact
dy dx

If a differential equation is exact, its solution can be found by the following

method.
488 / Differential Equations

1. Integrate M(x, y) with respect to x replacing the usual constant of integration

with a function f(y) of y.

F(x, y) = M(x, y) dx = G(x, y) + /(y)

ft

2. Differentiate F(x, y) = G(x, y) + f (y) obtained from step 1 with respect to y and
compare this with N(x, y) from the differential equation to be solved to obtain the

Va‘Ue °f Jy /

8G d ,, . dN
dy + dy y dy

8 dN dG
Tyf[y) = ^y dy

3. Integrate — f(y) with respect to y to obtain f(y).

f(y ) dy =/(y)
dy

It is not necessary to include the usual constant of integration, since it is in¬

troduced in the final step of the solution.
4. The solution, from steps 1 and 3, is

F(x, y) = G(x, y) +f(y) + C = 0

clearly the solution may also be obtained by integrating first with respect to y.

EXAMPLE

Solve the equation [y(2 + x3y)/x3] dx = [(1 — 2x3y)/x2] dy.

/y(2 + x3y)\ _ 2 -I- 2x3y _ d I 1 — 2x3y^

, so the equation is exact.
dy\ X' dx

dF y(2 + x3y)
dx x3

F(x, y) = -~2 + y2x + f(y) (step 1)

A

8F 1 ^^
-=--2 a i + ,2xy
+ 2xy + -f(y)=--2

SO

Ty'W-O (step 2)

f(y) = c (step 3)

and

F(x, y) = ~ ^2 + y2x + C = 0 (step 4)

X
 5.4 Differential Equations of First Order and First Degree / 489

Thus

x3y2 — y = Cx2

is the solution.

EXAMPLE

Solve the equation (x2 — x + y2) dx — (yey — 2xy) dy = 0.

d d
— (x2 — x + y2) = 2y = — (— yey + 2xy), so the equation is exact.

dF
= x2 - x + y4
dx

F(x> y) = y - y + xy2 +f(y) (step 1)

dF d
ey=2xy + Tyf{y}= ~ye, + 2xy

so

fyfiy) = ~ye" (step2)

f(y) = ~(y - iy (step 3)

and
3 2
= y ~ y + xy2 - (y - l)ey + C = 0 (step 4)

Thus

2x3 — 3x2 + 6xy2 — 6(y — 1 )ey — C

is the solution.

note: Alternatively, the terms of the original equation can be considered in two sets:
x2 dx — x dx — yey dy is exact by inspection and can be integrated by the usual methods;
y2 dx + 2xy dy can be shown to be exact and can be integrated by the method above. When
this method is used,/(y) is a constant, as illustrated in the following example.

EXAMPLE

Solve the equation [(2xy — l)/y] dx + [(x + 3y)/y2] dy = 0 and find the particular solution if
y = 1 when x = 2.

3 1 x
2x dx + - dy-dx -\—y dy = 0
_____ j_ y y2

exact, integrable d l
by usual methods dy \ y) y2 ~ dx\y2J
to obtain x2 + 3 In y

so the equation is exact.

490 / Differential Equations

Hx> y)= ~^+f(y) (step 1)

dF x d r. x x
dy y2 + dy y y2

(step 2)

f(y) = c (step 3)

F(x, y)= - - + C (step 4)

Thus

x2 + 3 In y — - = C
y
and

x2y — x + 3y In y = Cy

is the solution.
If y = 1 when x = 2,

4 — 2 = C

C = 2

and

x2y — x — 2y + 3y In y = 0

is the particular solution for x = 2, y = 1.

EXAMPLE

The change in price, y, with change in quantity demanded, x, of a particular commodity is given
by

dy 2 xy + 24x
dx x2 + 16

Find the relationship between price and quantity demanded if the price is 7.5 when the
quantity demanded is 4.

(2xy + 24x) dx + (x2 + 16) dy — 0

— (2xy + 24x) = 2x = — (x2 + 16)

cy dx

so the equation is exact.

dF = .
— 2xv -f 24x
ex

F(x, y) = x2y + 12x2 +f(y) (step 1)

 5.4 Differential Equations of First Order and First Degree / 491

Figure 5.4

so

^/M = 16 (step 2)
3

VO

(step 3)
li

and

F(x, y) = x2y 4- 12x2 4- 16y + C = 0 (step 4)

thus

x2y + I2x2 + 16y = C

is the solution.
If y = 7.5 when x = 4,

120 + 192 + 120 = C

C = 432

and

x2y 4- 12x2 4- 16y — 432 = 0

is the particular solution for x = 4, y = 7.5. (See Figure 5.4.)

LINEAR DIFFERENTIAL EQUATIONS

In some cases a differential equation which is not exact can be made exact by
multiplication by a factor; such a factor is called an integrating factor, since it permits
the equation to be integrated. In general, determination of the appropriate integrat¬
ing factor for a given differential equation may not be an easy matter; however, it can
492 / Differential Equations

be shown that any linear first-order differential equation has the integrating factor
e$ P(x)dx

An equation which can be written in the form

^ + yP(x) = Q(x)

is a linear differential equation of first order. If this equation is multiplied by the

integrating factor esp(x)dx, the resulting equation

e, PM dx^l + yP(x)eI PM dx = PM dx
dx

can be integrated with respect to x to obtain

yeS P(x)dx _ | ei p(x)dxQ(x) dx + C

since

— (yeSPMdx) = eIP(X,dxd2 + yp(X)eIP(x)dx

dx dx

Thus the equation

^ + yP(x) = Q(x)

has as its solution

y _ e~ 5 P(x)dx
es P(x)dxQ(x) dx + C

Similarly, the equation

dx
+ xP(y) = Q(y)
dy

has eSP(y)dy as an integrating factor and its solution is

x = e ~f P(y)dy efp(y)dyQ(y) dy + C

note: Every differential equation of first order and first degree, which has a
solution, can be shown to have an infinite number of integrating factors. However,
determining these integrating factors may not be easy; in general, only linear first-
order differential equations are solved using integrating factors. As discussed in the
following section, first-order differential equations which are linear in a function of a
variable can also be integrated using integrating factors.

EXAMPLE

Solve the equation x2 dy + (y — 2xy — 2x2) dx = 0.

Writing the equation in the form ~ + yP{x) = Q(x),

tt

dy
dx
+y
 5.4 Differential Equations of First Order and First Degree / 493

The integrating factor is

p~
[(1 - 2x)lx2] dx _ £-(l/x)-ln *2 _ _

x2
and

2e 1/x
dx

= 2e~1,x + C

y = 2x2 + Cx2ellx

= x2(2 + Ce1,x)

EXAMPLE

dx
Solve the equation y — = 2ye3y + x(3y + 2).

dx
Writing the equation in the form = Q(y)»
dy +

dx
dy -m
The integrating factor is
- -

^J-[(3y+ 2)/y] dy _ ^-3y-2 1ny _

y
and

xe~3y
2
y

-- + C
y
X -2 ye3y + Cy2e3y

ye3y(Cy - 2)

EXAMPLE

Solve the equation 2y dx = (yA + x) dy and find the particular solution if y 1 when x = 0.
dx
Writing the equation in the form -- + xP(y) = Q(y),

dx / 1 \ y3
dy~X\Tyj=i:

The integrating factor is

eS~dy/2y _ (1/2) In y _ - 1/2
494 / Differential Equations

xy 1/2 = ^ j y5/2 dy

= W2+c
x = 4/ + c//2
7x = / + Cy1/2

If y = 1 when x = 0,

0 = 1 + C

C = -1

and

7x = / - y1/2

is the particular solution for x = 0, y = 1.

EXAMPLE

A manufacturing company has found that the cost, c, of operating and maintaining its
equipment is related to the length, x, of the interval between overhauls by the equation

dc b — 1 ba
dx x x2

where a and b are constants. Find c as a function of x if c = c0 when x = x0.

dc
Writing the equation in the form -—b cP(X) = Q(x),
It A

dc l(b — 1)\ ba
dx \ x / x2

The integrating factor is

eS - [(ft - 1 )/x] dx _ e - (ft - 1) In x _ x - (ft - 1)

and

cx (b = —ba j x ft-1

= ax b + C

a ^ i
c = - 4- Cxb 1
x

If c = c0 when x = x0,

a
Cq — — + Cx0ft- i
x0

Co-^o —
C =
xbo

a Cq Xq — a -1
c = - + ^- .ft

X xbo

(See Figure 5.5.)

 5.4 Differential Equations of First Order and First Degree I 495

Figure 5.5

DIFFERENTIAL EQUATIONS LINEAR IN

A FUNCTION OF y OR IN A FUNCTION OF *
An equation which can be written in the form

dx
f(y) +f(y)p(x) = Q(x)

is a linear differential equation in the variable f(y); that is, it is of the first degree in

/(y) and ^ f (y). Its solution is thus given by

eS P(x)dxQ(xj dx -f C
f(y) —
= a~S
e P(x) dx

Similarly, an equation which can be written in the form

jyf(x) +f(x)p(y) = Q(y)

is a linear differential equation in the variable f(x); that is, it is of the first degree in

/(x) and ~ f (x). Its solution is thus given by

wA

_ e~S P(y) dy eSPiy)dyQ(y) dy + C

EXAMPLE

Solve the equation y dx -1- x(l — x2y4) dy = 0.

496 / Differential Equations

Writing the equation in the form ^f(x)+f(x)P(y) = Q(y),

dx
x3y 3
dy

, dx ,/T
x-3 — + x 2 -
dy \y

dx 2^
-2x~3~ + x_21 - -2y3
dy y>
d_
-2 y3
dy

The integrating factor is

eS~(2/y)dy __ e~2\n y _ y-

and

x -2 y V dy + C

y dy + C

= y2[-y2 + c]
1 = — x2y4 + Cx2y2

1 + x2y4 = Cx2y2

EXAMPLE

dy 2
Solve the equation-1- xy In y = xye x and find the particular solution if y = 1 when x = 0.
dx

Writing the equation in the form ~~ f(y) + f(y)P(x) = Q(x),

if r\

1 dy
+ (In y)(x) = xe~
y dx

- v2
(In y) + (In >)(x) = xc
dx

The integrating factor is

e5 xdx _ ex2/2

and

In y = e *2/2 I xex2/1e x2 dx + C

= e~x212 xe x2/1 dx + C

= e-x2l2[-e~x212 + C]

= —e~x2 + Ce~xl/2
 5.4 Differential Equations of First Order and First Degree / 497

If y = 1 when x = 0,

0 = -1 + C

C = 1

and

In y = —e~xl + e~x2/1

= e~x2l2(l - e~x2/2)

is the particular solution for x = 0, y = 1.

EXAMPLE

The cost control unit of a large public accounting firm has found, as the company has
increased in size, that the average monthly cost, y, of office supplies is related to the number, x, of
employees (in addition to the unit director) by the equation

+ 2y = y2e x
dx

Find y as a function of x if y = 3 when x = 0.

Writing the equation in the form f(y) + f(y)P(x) = Q(x),

UA

y 2 + 2y 1 = e x
dx

-y-2t~2y-'= ~e'

{y x) - 2y 1 = -e
dx

The integrating factor is

gS~ 2 dx _ 2x

and

y 1 = e2x e 2xe x dx + C

= e 2x e 2x dx + C

— 3x

= e 2x
+c

+c
3ex = y + Cyex

If y = 3 when x = 0,

3 = 3 + 3C

C = 0
and

y = 3ex

is the particular solution for x = 0, y = 3. (See Figure 5.6.)

498 / Differential Equations

Figure 5.6

note: In general, there is no rule for determining the function f(y) or f(x) in terms of
which a differential equation is, or can be made, linear. The equation must be put in the form

— + f(y)P(x) = Q(x) or the form g(x) — +f(x)P(y) = Q(y) so that it can be determined
dx dy
by inspection whether the differential equation can be made linear in f(y) or f(x). However,
there is one class of differential equations for which the appropriate function is known. An
equation of the form

d-r. + yPM = y"Q(x)

or

- + xP(y) = x"S(y)

can be made linear in yn~1 or xn~1 by multiplication by (1 — n)y~n or (1 — n)x~n, respectively.

Equations of this type are known as Bernoulli equations. The first and last examples of this
section involve Bernoulli equations.

5.5 Differential Equations of Higher Order

and/or Higher Degree

Except for special cases, methods of solution of differential equations of the second
order or degree are somewhat complicated and, in general, solution becomes more
difficult as the order or degree of the differential equation increases. In many cases,
methods for obtaining an exact solution are not known and numerical methods for
 5.5 Differential Equations of Higher Order and/or Higher Degree / 499

obtaining approximate solutions must be used. It should be noted that there are, in
fact, some types of differential equations of first order and first degree for which exact
solutions are not known. With electronic computers, highly accurate approximate
solutions can be obtained; these solutions may sometimes be preferred, for reasons of
convenience, even though methods for exact solution are known.

PROBLEMS
Find the general solution for each of the following differential equations.

1. x dy — 3y dx = x2 dx

2 . + 2xy — 2xe~xl = 0
dx

3. yj~ + { 1 + y)x — ey = 0
dy

4. y -—b 2x — 3y = 0
dy

5. x(6xy + 5) dx + (2x3 + 3y) dy = 0

6 . (aj x + bx y + c^) dx + (bi x + b2y 4- c2) dy = 0

7. y dx + 3x2ex dx = dy

8. (yexy + 2xy) dx + (xexy + x2) dy = 0

9. (x + 1)^ + 2y = ex(x 4- l)-1

dx

10 . (y cos x — 2 sin y) dx = (2x cos y — sin x) dy

11. (3x2y -I- xy2 + ex) dx 4- (x3 4- x2y + sin y) dy = 0

12 . 2y sin xy dx -f (2x sin xy + y3) dy — 0

13. (3y sin x — cos y) dx + (x sin y — 3 cos x) dy = 0

14. (y2 esc2 x + 6xy — 2) dx — (2y cot x — 3x2) dy

15. cos y dx — (x sin y — y2) dy = 0

16- + p) dx -(?
(4 + ~3rl
=
r
dy
3 dy 4 -
17. ^ + 2y-x = 0 23. yJ -—K xy = xc
dx dx

24. xy dy = (x2 — y2) dx

18. x^ = x3 +y
dx
25. (x + y) dx + (x — 2y) dy = 0
x
3 - + . — 2 dy
dx x + 1 y 26. + |y — —|x = 0
dx
dx
20. ~T + X = e y ds
dy 27.
dt
dy in x2 ds s sin t
--xy = y ^xe
dx 28.
df t t
dy x
22. dy
dx y 29. + y = cos x — sin x
dx
500 / Differential Equations

30. (2x cos y — ex) dx — x2 sin y dy = 0

31. (y sin x + xy cos x) dx 4- (sin y + ey) dy — 0

32. — — xy = e{ll2)x2 cos x 33. cos 9 ^ = 2 + 2r sin 9

dx dd

34. cos x sec y dx -f- sin x sin y sec2 y dy = 0

ds . 7
35. — 4- s tan t = 2t + t2 tan t
dt

36. cos y dy 4- (sin y — 1) cos x^x = 0

37. x tan2 y dy + x dy = (2x2 + tan y) dx

2
38. 3y2 -xy3 _= ^(1/2)jc2
e COS X
dx

39. (In y 4- y3 4- yexy) dx 4- 4- 3xy2 4- xc*yj dy = 0

40. dx 4- (2xy — 4xy3 — 2y3ey4) dy = 0

41. |y2 + - + ycyj dx + ^2xy — + xe* 4- x.yeyj dy = 0

,2 dy ,
42. y2 ^ + (y2 4- 2y)x — 1=0 47. x2 -—I- y2? _
= xy
dx

43. x dy = (5y 4- x 4- 1) dx 48. sin 9 d9 + cos 9 dt = tel dt

xy 4- 1 2y — x 49. esc y cot y dy = (esc y 4- ex) dx

44. dx 4- dy = 0

y
50. x + v = v2x2 cos x
dx
45. ^ = x + e”
dy
dx . .
51. --x cot y = ey( 1 — cot y)
dy
46. x^ +2y = 3xV/3
dx

Find the particular solution for each of the following differential equations under the given
conditions. v

dy
52. -—by tan x = sec x, y = — 1 when x = 0
dx

53. dy 4- (y cot x — sec x) dx = 0, y = 1 when x = 0

dx
54. (y2 4- 1) —4- 2xy = y2, y = — 1 when x = 0

55. dy = x(l — e2*-*2) dx, y = 0 when x = 0

56. |x2 4- ^ 4- yexyj dx 4- (ey 4- 3y2 4- xe**') dy = 0, y = 0 when x = 1

57. (1 — x2)^ + xy = x(l — x2)y1/2, y = 1 when x = 0

58. 4- y = y2e~x, y = 2 when x = 0

U A

59. y dx 4- 2(x — 2y2) dy = 0, y = — 1 when x = 2

60. (y 4- y3) dx + 4(xy2 — 1) dy = 0, y = 1 when x = 0

 5.5 Differential Equations of Higher Order and/or Higher Degree / 501

61. 2y dx = (x2y4 4- x) dy, y = 1 when x = 1

62. (x2 — 1) ^ + (x2 — 1 )2 + 4y = 0, y = — 6 when x = 0

63. (yex — 2x) dx + ex dy = 0, y = 6 when x = 0

64. dx + dy = 0, y = 2 when x = 1

65. (2y — xy — 3) dx 4- x dy = 0, y = 1 when x = 1

66 . (yey + 2x + y) dx + (xey + xyey 4- 3y2 4- x) dy = 0, y = 0 when x = 2

67. |3x2y — ~ — 6x4j dx 4- dy = 0, y = 0 when x = — 1

68 . dr — (1 4- 2r cot 6) d6, r = 3 when 6 = tc/2

69. The change in net profit, P, as advertising expenditure, x, changes is given by the equation

— = k — a(P 4- x)
dx

where a and k are constants. Find P as a function of x if P = P0 when x = 0.

70. The change in the cost of ordering and holding, C, as quantity, x, changes is given by the
equation

,
dC
dx
=
C
a-
x

where a is a constant. Find C as a function of x if C = C0 when x = x0.

71. Manufacturing and marketing costs, C, are related to the number of items, x, by the equation

dC _ , .
——K aC = h 4- kx
dx

where a, b, and k are constants. Find C as a function of x if C = 0 when x = 0.

72. The change in consumption, c, of a particular commodity as income, /, changes is given by
the equation

dc
= c + ke1
dl

where k is a constant. Find c as a function of / if c = c0 when 1 = 0.

Answers to Odd-Numbered Problems

1. y = —x2 + Cx3 21. 3y1/2 =ex2 + Cex2/4

3. 2xyey — e2y = C 23. e2x2y4 = 2ex2 + C

5. 4x3y 4- 5x2 + 3y2 = C 25. x2 4- 2xy — 2y2 = C

1. y = x3ex + Ce* 27. s = t 4~ 2 4“ C sin t

9. y(x 4- l)2 = ex + C 29. y = cos x + O-*

11. 2x3y 4- x2y2 4- 2ex — 2 cos y = C 31. xy sin x — cos y 4- = C

13. 3y cos x 4- x cos y = C 33. r cos2 6 = 2 sin 6 -b C

15. y3 4- 3x cos y = C 35. s = r2 + C cos t

17. y = ^x — i 4- Cc-2x 37. tan y = 2x2 4- Cx

19. 12(x -b l)2y3 = 3x4 4- 8x3 -b 6x2 + C 39. x In y -b xy3 4- exy = C

502 / Differential Equations

57. 3y1/2 + 1 - x2 = 4(1 - x2)1/4

41. x|y2 4- - 4- yeyj = C
59. y2(x — y2) = 1
43. 20y 4- 5x 4- 4 = Cx5
61. lOx = (9 + x/V/2
45. x = yey 4- Cey
63. yex — x2 = 6

47. x = y In Cx
65. x2y 4- 3(x + 1) = lex~1
49. 2ex esc y 4- e2x = C
67. y = 2x2 — 2x2e_*3_1
51. x = ey + C sin y k + 1
k + 1 t-ax
69. P = - x 4- P0 -
53. y sin x 4- In cos x = 0 a a

55. x2+ l = ex2~2y

71. c = ab 2 * (1 -e'“) + -x
a a

5.6 Applications of Differential

Equations in Economic Models
There are two general types of economic models, static and dynamic. Static models
concern equilibrium situations, that is, situations which if attained will be main¬
tained. Dynamic models concern situations which change over time. In dynamic
models, time enters either explicitly as a variable or implicitly in the form of lagged
variables. The models discussed in this chapter are very simple dynamic economic
models expressed in terms of differential equations.
There are two general classes of variables in economic models, referred to as
endogenous and exogenous variables. Endogenous variables are those variables
whose values or levels are to be predicted or explained; exogenous variables are
assumed to be determined and known in advance and can be regarded as constant in
the model. Endogenous and exogenous come from Greek words meaning “ generated
from inside ” and “ generated from outside,” respectively. Endogenous variables are
predicted from the model; exogenous variables are determined outside the model.
Usually a model is first written in terms of structural equations which express
relationships among endogenous and exogenous variables. This set of structural
equations is then solved, if possible, for what are called reduced form equations, each
of which expresses an endogenous variable as a function of exogenous variables and
parameters. A model is solved by obtaining a reduced form equation for each
endogenous variable in the model. Solution of economic models is illustrated in this
section for the Domar macro model, two Domar debt models, the Evans price
adjustment model, and an income-consumption-investment model.

DOMAR MACRO MODEL

The following very simple macro model was proposed by E. D. Domar.

S(t) = a y(t)

«■>-'£
s(t) = i(t)
y(0) = y0
a>0 f>0
 5.6 Applications of Differential Equations in Economic Models / 503

where S is saving, I is investment, y is income, and each of these endogenous vari¬

ables is a function of time.
The first equation states that saving is a fixed proportion of income; the second
equation states that investment is proportional to the rate of change of income over
time; the third equation states that saving equals investment; the fourth equation
states the initial condition. From these relationships, specific functions expressing
variations in the variables over time can be obtained.
Since S(t) = 7(f), the differential equation

dy a
dt ~P
is obtained for solution.

1 dy a
y dt p

ln y=+c
y = CeMn

If y — yo at t = 0, then

= C
and the particular solution is

y — y0 e(alp)t

Note that this solution gives income, y, as a function of time, t. Since a > 0,/?> 0, the
function has an increasing positive slope, the rate of increase depending on a/p.
Solutions for the remaining variables of the model, / and S, are as follows:

I = S = (xy = ay0 e(*lp)t

DOMAR DEBT MODEL

Domar uses a set of models similar to the macro model above to express relation¬
ships between national income and national debt. Consider the model

y(0) = y0
0(0) = D0
a >0 p>0
where D is national debt and y is national income (both endogenous). In this model,
national income increases at a constant rate p over time and the rate of increase of
national debt is a fixed proportion of national income. The third and fourth equa¬
tions state the initial conditions. Integrating the second equation,

y — pt -V C
504 / Differential Equations

Since y = y0 when t = 0, C = y0 and

y = pt + y0
Substituting in the first equation of the model,

dD
—- = <xpt + Gcy0
at

D = {apt2 + a y0t + C

Since D = D0 when t = 0, C = D0 and

D — fit2 + ay0t + D0

The solution of the model is thus

D(t) = joc fit2 + ay 0t + D0

y(t) = fit + y0

Domar was interested in the ratio of national debt to national income

D(t) _ jeepy + oty0t + Dp

y(t) pt + y0

or

D(t) = Do aypt ja^t2

y(t) pt + y0 pt + y0 pt + y0

As t -*■ oo,

(a constant)
pt + yo P
{cf.pt2
—-► 00
Pt + yo
\

Thus, as t -+ oo, D(t)/y(t) -> oo and, for this model, the ratio of national debt to
national income increases over time without limit.

A SECOND DOMAR DEBT MODEL

A second version of the Domar debt model is obtained by modifying the second
equation in the above model, so that income increases by a constant proportion of
income, and leaving the other equations unchanged.
dD

dy
dt

y(0) = ya
0(0) = D0
a > 0 p > 0
 5.6 Applications of Differential Equations in Economic Models / 505

Integrating the second equation,

1 dy
= P
y dt

In y = fit -I- C

y = Cept

Since y = y0 when t = 0, C = y0 and

y = y0 e"'
Substituting in the first equation of the model,

dD
= a y0efit
dt

a
D = -y0eV + C

oe
Since D = D0 when t — 0, C = D0 — - y0 and

y. a nt
D = Do~^y0 + gy«ee

or

a
D = D0 + jy0(e»- 1)

The solution of the model is thus

D(t) = D0 +^y0(eIH - 1)

y(t) = y<>e0'

The ratio of national debt to national income is

m _ Do .«/. _ m
At) yOe0’ n e>1}

As t -*■ oo,

D,
0
fit

a 1 a
1 -
,fit

Thus, as t -> oo, D(t)/y(t) -► a/j8, a (finite) constant.

EVANS PRICE ADJUSTMENT MODEL

A model of a particular market for some commodity was proposed by G. C. Evans.
The demand and supply equations are the same as those for the usual simple linear
model and can be solved for the equilibrium price in the usual way. In addition, there
is an equation stating that the rate of change of price over time is proportional to the
506 / Differential Equations

excess demand (d — 5). This factor of proportionality is positive, implying that a

positive excess demand causes a rise in price and a negative excess demand causes a
fall in price.

d(t) = a0 + alp(t)

s(t) = Po + PiP(t)

fry(d-s)

oq < 0 /?i>0 y > 0

where d is demand, s is supply, and p is price. Substituting the first two equations into
the third equation,

Yt = - Po + («1 -PM

= y(<xl - p,)(p - pe)

where pe — (a0 — Po)/(Pi — ai) is the equilibrium price in the model obtained in the
usual way by solving d(t) = s(t) for p(t), the equilibrium price. Letting X — y(a1 — Pi),

jt = x{p~ p‘]

= x

p - pedt

In (p - pe) = At + C

P - Pe = Ceu

P = Pe + Cex'

Since p = p0 when t — 0, C = p0 ~ pe and

P = Pe + (PO ~ Pe)eM

where, as above, pe — (a0 — p0)/(Pi — ax) and X = y(ax — Pi). Since X < 0, p -> pe as
t -> 00.

INCOME-CONSUMPTION-INVESTMENT MODEL
Consider a differential-equations form of the income-consumption-investment model
where current consumption and investment are linear functions of current income
and income changes at a rate that is proportional to excess demand, that is, to
consumption plus investment minus income.

Cd(t) = a Y,(t)

hit) = 7 hit)
dY
-jf = Hcs + Is — Ys)

»i(0) =Y0-Ye

0<a<l 0 < y < 1 0 <A < 1

 5.6 Applications of Differential Equations in Economic Models / 507

where Cs, Id, and Yd are deviations of consumption, investment, and income, respec¬
tively, from their equilibrium values Ce, Ie, and Ye. Substituting the first two equa¬
tions into the third equation,

^ = X(a + y - l)y4

= X(a + y — 1) dt

In Yd — A (a + y — l)t + C

Yd = CeMa+y~l)t

Since Ys = Y0 - Ye when t = 0, C = Y0 - Ye and

Yd = (Y0-Ye)e^ + y~l)t

That is,

Y = Ye + (Y0 - Ye)eMa + y'l)t

And, if a + y < 1, eMa + y~1)1 0 and Y -> Ye as t -► oo.

PROBLEMS
1. Demand, x, and supply, y, of a product are given as a function of unit price, p, by the
following equations:

x = ap + b y = cp + d

Suppose that price changes in such a way that the excess of demand over supply is de¬
creased at a rate proportional to the excess. Show that

a' 7t^x ~ ^= ~k^x ~ ^

b. ~ 4- k(p — p) = 0, where p = -—-

c. The unit price tends to an equilibrium value p and

P = P + (Po ~ P)e

where p0 is the initial price at t = 0.

2. Consider the model

y(t) + P

*= yy{t)

y(°) = yo
0(0) = D0

a > 0 (3 > 0 y > 0

where D is national debt and y is national income.

a. Solve the model.
b. Determine the limit as t -* oo of the ratio of national debt to national income.
508 / Differential Equations

3. Consider the model

dR a
dt S(t)

?■- -As(t>

R( 0) = 0

S(0) = S0 (original cost)

a > 0 A > 0

where JR is running cost of a car and S is resale value. Solve the model.
4. The increase in the number of new products tested, y, as amount, x, allocated to research and
development increases equals a constant times the product of the number of new products
tested and the amount allocated to research and development plus another constant times
the ratio of the amount allocated to research and development divided by the number of
new products tested. Find the relationship between the number of new products tested and
the amount allocated to research and development if y = y0 when x = 0.
5. Consider the model

dR a
dt S(t)

ft = -m

R( 0) = R0

S(0) = S0

a > 0 /? > 0 A > 0

where R is the cost of repairing and operating a machine and S is salvage value of the
machine. Find R and S as functions of time, t.
6. Consider the model

dw 1
+ kept
dt ccy(t)

dy
Mt)
dt

w(0) = w0

)>(0) =
k
0 > 0 a >
Pyo
where w is per capita consumption of wheat products and y is per capita income. Find w
and y as functions of time and determine the limit as t —► oo of the ratio of per capita
consumption of wheat to per capita income.

Answers to Odd-Numbered Problems

3. S{t) = S0e~kt 5. S = S0e~xt

R(t) = ~ (ei‘- 1) R = R0 + pt + ^(e1’- 1)

AOQ Auq
Difference
Equations

6.1 Introduction
When variables are considered as changing discretely or discontinuously rather than
continuously or instantaneously, difference equations rather than differential equa¬
tions are appropriate for expressing relationships among the changes. Difference
equations are frequently useful in business and economic analyses, since many eco¬
nomic data are recorded for uniformly spaced periods of time, for example, gross
national product may be given for a year; net profit for a quarter; quantities
produced, bought, or sold for a month.
In many econmic analyses, time is the independent variable and changes in other
variables over time are studied. Such studies of variables over discrete sets of time
values are referred to as period analyses, and difference equations provide the basis
for such analyses. Recall that when the time periods (or, more generally, the changes
in the independent variable) are made smaller and smaller, difference equations
approach differential equations as a limiting case.
For generality, the following discussion of the classification and solution of differ¬
ence equations is in terms of an independent variable x. In many economic and
business analyses, including the models discussed later, the independent variable
time is denoted by t.
After the definition and classification of difference equations are discussed, the
solution of a difference equation is defined. The method of solution and the behavior
of the solution sequence are considered for linear first-order difference equations
with constant coefficients and applications of these difference equations in economic
models are illustrated. The methods of solution, the behavior of the solution
sequence, and applications in economic models are then discussed for second-order
difference equations.
509
510 / Difference Equations

6.2 Definition and Classification

of Difference Equations
In this section, a difference equation is defined and the classification of difference
equations by order and degree is discussed. Since most difference equations used in
applications are linear, particular attention is given to the definition and
classification of linear difference equations.
Suppose that y is a function of x, y = /(x), where y is defined for integer values of
x, x = 0, 1, 2, 3,.... In the context of difference equations, the functional relationship
y = /(x) is frequently indicated by yx. The change in y as x changes from x to x + 1 is
the first difference of yx and is written

Ay* = y* +1 - y* (read “ delta ” yx)

Note that Ayx is also a function of x; A is an operator and provides the rule for
computing Ayx from the sequence y0, yl9 y2,... . Similarly, higher-order differences
are obtained as differences of differences by applying the operator A:

the second difference of yx is A2yx = A(Ayx) = Ayx+1 — Ayx

= (y*+2 - y*+i) - (y*+i - y*)

= y*+2 - 2y*+i + y*
the third difference of yx is A3yx = A(A2yx) = Ayx+2 - 2Ayx+1 + A>-x

= (y*+3 — y*+2)— z(y*+2 — y*+1)

+ (y*+i -y*)
y* + 3 +2 yx

the kth difference of yx is Akyx = A(A‘ 1 vx) = Y. n, k\, ■■

i=0 Ij.l.

EXAMPLE==========

If y = 2x2 -3,
A2y = Ayx+1 - Ayx

= [2(x + 2)2 - 3] - [2(x + l)2 - 3] - {[2(x + l)2 - 3] - [2x2 - 3]}

= 4

Alternatively,

A2y* = yx + 2 - 2yx+l + yx = [2(x + 2)2 - 3] - 2[2(x + l)2 - 3] 4- [2x2 - 3]

= 4

EXAMPLE

If y = 3x + 5,
 6.2 Definition and Classification of Difference Equations / 511

A3y* = Ayx + 2 - 2Ayx+1 + Ayx

= [3(x + 3) + 5] - [3(x + 2) + 5] - 2{[3(x + 2) + 5] - [3(x + 1) + 5]}

+ [3(x + 1) + 5] — [3x + 5]

= 0

Alternatively,

A3y* = yx + s ~ 3y* + 2 + 3yx+1 - y*

= [3(x + 3) + 5] - 3[3(x + 2) + 5] + 3[3(x + 1) + 5] - [3x + 5]

= 0

A difference equation states a relationship involving differences or, equivalently, a

relationship involving the values of a dependent variable for a discrete set of (lagged)
values of the independent variable. In the following discussion, the values of the
independent variable are assumed to be equally spaced; for convenience, the
independent variable is redefined, if necessary, so that its values are successive posi¬
tive integers.

LINEAR DIFFERENCE EQUATIONS

A difference equation is said to be linear if the dependent variable occurs only in
expressions of the first degree, that is, if the dependent variable does not occur raised
to a power higher than one and does not occur in cross-product terms. Thus a
difference equation is linear if it can be written in the form

a0{x)yx + „ + a1(x)yx+„.1 + ■■■ + an-1(x)yx+l + a„(x)yx = g(x)

where a0, al9..., a„_ l5 an and g are functions of x (but not of yx) defined for x = 0,1,
2
Jm* J • • • •

A linear difference equation is of order n if, when written in this form, both a0 and
an are unequal to zero for all values of x under consideration; that is, a linear
difference equation is of order n if it involves values of y corresponding to values of x
differing by n but by not more than n. Difference equations that are not linear are, in
general, very difficult to solve and are seldom used. The following discussion is
confined to linear difference equations.

EXAMPLE

Each of the following difference equations is linear and of the order indicated.

y3c + 2 - 7yx+1 + 5yx = 3x order 2

3yx + 2 + 4yx+1 = 2x order 1

18yx + 2 - by* = order 2

%xyx + 3 - 3xyx + 2 + 9xyx+1 + 2yx = 3 order 3

An nth-order difference equation can be written as an implicit function of the

values of the variable y at n different values of x, that is, n lagged values of y,

f[yx+n. yx+n-1. •••> yx] = o

512 / Difference Equations

Alternatively, since knowledge of n 4* 1 adjacent values of y permits calculation of

the value of y and its first n differences, an nth-order difference equation can also be
written as a function of y and its first n differences,

F[A"yx , A"~ 1yx , ..., Ayx , yx] = 0

EXAMPLE

Ayx = 0 can be written

yx + i-yx = Q (order 1)

Ayx — 2yx = 5 can be written

y*+i - yx - 2yx = 5

y*+i - 3yx - 5 = 0 (order 1)

A2yx — 3Ayx — 3yx = x can be written

A(y*+1 - yx) ~ 3(yx+1 - yx) -3yx = x

yx+i - yx+1 - yx+i +yx- 3y*+i + 3yx - 3yx = x

yx+2 - 5yx+i + yx - * = 0 (order 2)

A3yx + A2yx + Ayx + yx = 0 can be written

A2(yx+1 - yx) + A(yx+1 - y) + yx+1 - yx + yx = 0

A(yx+2 ~ yx+1 ~ yx +1 + yx) + yx+2 — yx+1 — yx +1 + yx + yx+1 =0

yx+3 — yx + 2 — 2yx+2 + 2 yx+i + yx+i — yx + yx+2 — yx+1 + yx — 0

yx + 3 - 2yx + 2 + 2yx+1 = 0
(order 2)

6.3 Solutions of Difference Equations

A solution of a difference equation is a functional relation, not involving differences,
which is defined for all nonnegative integers and which satisfies the difference
equation.
The general solution of an nth-order difference equation is a solution containing n
arbitrary constants. It can be shown that the general solution of a difference equation
is unique.
A particular solution of a difference equation is a solution that can be obtained
from the general solution by giving particular values to the arbitrary constants of the
general solution. As in the case of differential equations, the arbitrary constants of a
difference equation are specified by boundary conditions or initial conditions. The
general solution of an nth-order difference equation includes n arbitrary constants,
and a particular solution thus requires the specification of n boundary conditions.
In each of the following examples a proposed solution is shown to satisfy a given
difference equation and a particular solution is determined for the specified bound¬
ary condition(s). Methods for obtaining solutions of linear first- and second-order
difference equations with constant coefficients are then discussed. Similar methods of
solution are appropriate for higher-order linear difference equations.
 6.3 Solutions of Difference Equations / 513

EXAMPLE

Show that yx = x + c is a solution of yx + 1 — yx = 1 and find a particular solution if y0 = 1.

If yx = x + c,

yx+ i - >x = 1

(x + 1 + c) — (x + c) = 1

so

y* = x + c

is a solution.
Ifyo = L
1=0 + c

c = 1

and

yx = x + 1

is the particular solution when y0 = 1.

EXAMPLE

Show that yx = [x(x — l)/2] + c is a solution of yx+1 — yx = x and find a particular solution if
yo = 2.
If = x(x - l)/2,

>y+i ~ yx = *

= X

x(x + 1 — X + 1)
= X

2x
= X
T
so

x(x — 1)
yx = —- + c

is a solution.
ify0 = 2,

2 = 0 + c

c = 2

and

x(x - 1)
yx = —- + 2

is the particular solution when y0 = 2.

514 / Difference Equations

EXAMPLE

Show that yx = Ci + c2( — l)x is a solution of yx + 2 — yx = 0 and find a particular solution if

yo = 2, yt = 5.
if ^ = ci + c2(-i)x,

yx+2 - yx = o

Ci + c2(-l)x + 2 - [d + c2(-i)x] = o

c2[(-i)*+2- (-1)1=0
so

yx = Ci + C2(~l)x

is a solution.
If yo = 2, yx = 5,

2 = d + c2

5 = Cj - c2

2d = 7

C\ = \

c2 = — i
and

yx = J-l(-ir
is the particular solution for y0 = 2, yx = 5.

Analogies Between Difference Equations and Differential Equations

There are many analogies in the classification and solution of difference and differen¬
tial equations; these arise because the derivative of a function is defined as a limit of a
difference quotient
dy Ay, yx+Ax yx
lim lim
dx Ax-+ 0 Ax Ax-> 0 Ax

and, as Ax -► 0, the discrete case (difference equations) approaches as a limit the

continuous case (differential equations).

PROBLEMS
1. Determine the order of each of the following linear difference equations,
a. 3yx + 2 - 3yx+1 = 3x c. 7yx+1 - 5yx = 3
b- 8yx + 3 - yx = 4 d. 6yx + 2 - 7yx = 5x
2. If y — x2 + 2x, find A2yx.
3. If y = ex, find A2yx.
4. If y = x3 -I- 3, find A2yx and A3yx.
C y
5. Show that yx =- is a solution of yx+1 =--— and find a particular solution if
1 + cx 1 + yx
>’o = — 4.
6. Show that yx = cx -f c2 2X is a solution of yx + 2 — 3yx+1 + 2yx = 0 and find a particular
solution if y0 = 1, yt = 2.
 6.4 Linear First-Order Difference Equations with Constant Coefficients / 515

7. Show that yx = c t + c2 2X — x is a solution of yx + 2 — 3yx +1 + 2yx = 1 and find a particu¬

lar solution if y0 = 0, y{ = 3.
8. Show that yx = + c2 2X 4- c33x is a solution of yx + 3 — 6 yx + 2 4 ll^+i — 6yx = 0 and
find a particular solution if y0 = 1, yi = 1, y2 = — 1.
9. Show that yx — Ci 4- c2 x 4 c3 x2 4 c4x3 is a solution of yx + 4 — 4yx + 3 4 6yx + 2 —
4yx+1 4 yx = 0 and find a particular solution if y0 = 1, yx = 5, y2 = 9, y3 = 7.
10. Write each of the following difference equations in terms of lagged values of y.
a. Ayx = 10 c. A2yx - 4yx = 2
b. A2yx - 3Ayx -5=0 d. A3yx 4 5Ayx = yx

Answers to Odd-Numbered Problems

1. a. order 1
b. order 3
c. order 1 7. yx = 4(2* — 1) — x
d. order 2
9. yx = 1 4 2x 4 3x2 — x3
3. ex(e - l)2

6.4 Linear First-Order Difference Equations

with Constant Coefficients
In this section, the method of solution for linear first-order difference equations with
constant coefficients is discussed. The behavior of the solution sequence is considered
and equilibrium and stability are denned.
A linear first-order difference equation can be written in the form

a0(x)yx + i + Oi(x)yx = g(x) x = 0, 1, 2, ...

where a0(x) =£ 0 and a^x) ± 0 or, alternatively, as

ai(x) , g(x)
1 / \ ' / \
a0(x) a0(x)

If a0(x), ai(x)> an(l dix) are constants, that is, are not in fact functions of x, then

y*+i =Ayx + B

where A and B are constants and A ± 0. (B = 0 if and only if g — 0 in the original

equation.) The difference equation yx+l = Ayx 4 B is the general first-order differ¬
ence equation with constant coefficients.
The solution of this equation can be obtained by induction as follows:

4 i = Ay0 4 B
y2 = Ayx = A(Ay0 4 B) 4 B
— A2y0 4 AB 4 B

y3 = Ay2 = A(A2y0 4 AB 4 B) 4 B

= A3y0 4 A2B 4 AB 4 B
516 / Difference Equations

y4 = Ay3 = A[A3y0 4- A2B 4- AB + B) 4- B

= A4y0 4- A3B + 4l2£ 4- AB + B

yx = Axy0 4- B(1 + A + A2 + A3 + • ■ • + Ax~1)

and thus, noting that 1 4- A 4- A2 4- A3 4- • • • 4- Ax 1 is a geometric series with sum

(1 — Ax)/( 1 — A), the solution of yx+l = Ayx + B is

1 - A:
>4 = Axy0 + B for A 1 x = 0, 1,2,...
1 - A

yx = y0 4. Bx for A — 1 x = 0, 1, 2, ..

This solution, obtained by induction, does in fact satisfy the equation

Tc+i = Ayx 4- B, since

For For A —— 1,

=Ayx + B yx+i = yx + B

(1 — Ax
Axy0 + B j—j I + B = (Fo + Bx) + B

IX+
. 1, —
,A - Ax+1 + 1 - A"
1
= A*+1y0 + B = y0 + B(x 4- 1)
1 - A

1 - Ax+1
= Ax+1y0 + B
1 - A

There are three special cases of the difference equation yx+l = Ayx + B which
occur frequently in descriptions of business and economic data.

1. First-order difference a constant:

>4+i ~ yx = B (special case: A = 1)

Solution: yx = y0 + Bx

2. First-order difference proportional to the variable:

>4+i -34 = 0)4+!| special case: A = B = 0

1
Solution: yx = >4
1 — or

3. First-order difference a linear function of the variable:

Examples of solutions of linear first-order difference equations with constant

coefficients and of economic models in which they occur are given in the following
sections.
 6.4 Linear First-Order Difference Equations with Constant Coefficients / 517

EXAMPLE

Solve the equation 2yx+l = Ayx + 3 and find the particular solution if y0 = i-

2 Vx +1 = 4yx + 3

y*+i = 2yx + i

y* = 2*y0 -1(1 - 2*)

yx = (yo + !)2" -1

if y0 = i

yx = 2x+1

EXAMPLE

Solve the equation 3yx+1 = 3yx — 7 and find the particular solution if y0 = 3.

3y* + i = 2>yx-l

y^+i = y, - ix

yx = yo - ix

If y0 = 3,

yx = 3 - ix

EXAMPLE

Solve the equation 6yx+1 + 2yx = 0 and find the particular solution if y0 = 2.

6yx +1 +2yx = 0

yx+i = —

yx = (~i)xyo

If y0 = 2,

y* = 2(-i)*

EXAMPLE

Solve the equation 3yx+1 — 9yx + 8 = 0 and find the particular solution if y0 —
518 / Difference Equations

3y*+i - 9yx + 8 = 0

yx+i = 3y,-f
8 /I - 3J
= 3xy0
3 \ -2

yx = 3*y0 + Hi- 3*)

yx = (yo — i)3* + f
if yo =
4 3x
yx = i

BEHAVIOR OF THE SOLUTION SEQUENCE

A sequence is a succession of terms formed according to some rule or, equivalently, a
sequence is a function defined for positive integer values of the independent variable.
The solution of a difference equation is a sequence. When the independent variable is
time, this sequence is sometimes referred to as the time path of the dependent
variable. For a linear first-order difference equation, the specification of y0 deter¬
mines or generates the solution sequence y0, yi? y2, y3, • •each term is found from
the difference equation

yx+l = Ayx + B x = 0, 1, 2, ...

or, equivalently, from its solution,

y* = Axy0 + B j~_Aa for A ± 1 x = 0, 1, 2, ...

yx = y0 + Bx for A = 1 x = 0, 1, 2, ...

When A and B are given, specification of y0 thus determines a solution sequence of

real numbers.
The behavior of the solution sequence of a difference equation is of interest for
many applications; this behavior depends on the values of y0, A, and £, as shown in
Table 6.1. A sketch typical of each type of behavior is given in Figure 6.1, labeled to
match the corresponding case in the table.

Figure 6.1

y* yx y*

J L J 1 I L -*► x ► X
0 1 2 3 4 5 6 7 8 • • • • • «

(a) (b)
6.4 Linear First-Order Difference Equations with Constant Coefficients / 519

(j) (k)
520 / Difference Equations

Table 6.1 Behavior of the Solution Sequence for yx + i = Ayx + B

Behavior of the
Case A B y0 yx, x = 1, 2, ... solution sequence

Constant: yx = y*

it
(a) A± 1 yx = y

II
o
(b) A > 1 yo > y* yx > y* Diverges to + oo
(monotone increasing)
(c) A > 1 yo < y* yx < y* Diverges to — oo
(monotone decreasing)
(d) 0 <A < 1 yo > y* yx > y* Converges to y*
(monotone decreasing)
(e) 0 <A < 1 y0 < y* yx < y* Converges to y*
(monotone increasing)
(f) -1 <A <0 y0 + y* Converges to y*
(damped oscillatory)
(g) A = —1 yo + y* Diverges
(oscillates finitely)
(h) A < —1 yo t4 y* Diverges
(oscillates infinitely)
(i) A = 1 B = 0 yx = yo Constant: yx = y0
(j) A = 1 B > 0 yx > yo Diverges to + oo
(monotone increasing)
00 A = 1 B <0 yx < yo Diverges to — oo
(monotone decreasing)

The results given in the table can be summarized in the following theorem.

theorem: The linear first-order difference equation

y*+i = Ayx + B x = 0, 1, 2, ...

has the solution

y* = Ax(y0 - y*) + y* if A + 1 x = 0, 1, 2, ...

yx = y0 + Bx if A = 1 x = 0, 1, 2, ...
where y* = B/( 1 — A). If — 1 < A < 1, the solution sequence converges to y*; other¬
wise it diverges, unless yx = y0.

EXAMPLE

Solve each of the following difference equations, determine the behavior of the solution
sequence, and calculate the first few values of the solution sequence.
(a) 6yx+1 + 2yx - 3 = 0, y0 = 1.

yx+ i = ~iyx + i

= (-1)^0+i(i-(-in
= (~i)*(yo -1) +1
If y0 = l,
yx = ■§( ~i)* + I
 6.4 Linear First-Order Difference Equations with Constant Coefficients / 521

A = —3, B = 3, y* = case (f), so converges to y*, damped oscillatory. y0 = 1, yx = f,

^2 = f, y3 = fi y4 = It, and so forth.
(b) 5yx+l - 4yx - 15 = 0, y0 = 10.

y*+i = hx + 3

y* - (?) yo + 3I yz|'
~

= (f)*yo + 15(1 - (f)*)

= (f)x(y0 - 15) + 15

If y0 = 10,

y* = 15 - 5(f)*

A = f, B = 3, y* = 15; case (d), so yx converges to y*, monotone increasing. y0 = 10, = 11,

y2 = Ilf, y3 = 12yf, y4 = 12jf§, and so forth.
(c) 5yx+! - 25yx - 1 = 0, y0 = 0.

yx+1 = 5y* + f

^ = 5>>’»+! (r:-y)
= 5*y0 — ys(l -5*)

= 5x(y0 + tu) - 27)

If yo = 0,

yx - 2t>(5x - 1)

A = 5, B = f, y* = —yo; case (b), so yx diverges to + 00, monotone increasing. y0 = 0, yi = f,

y2 = f, ys = nr, y4 = -4*, and so forth.

EXAMPLE

If money earns simple interest, the amount on deposit at any interest date is equal to the
amount on deposit one period (usually a year) previously plus the interest earned in that
period on the initial principal invested. If S0 is the initial deposit, St is the amount on deposit
after t periods, and r is the interest rate per period, then

St+1 = St + rS0 t = 0, 1, 2, ...

which has the solution

S, - 50(1 + tr) t = 0, 1, 2, ...

When the initial amount accumulates at compound interest, the interest earned during any
period is computed on the total sum on deposit at the beginning of that period, rather than on
the initial deposit, and

St + i = St 4- iSt = (1 4 i)St t — 0, 1, 2, ...

where i is the interest rate per conversion period; thus

Sf = (1 4 i)*So t = 0,1,2,...

The amount on deposit diverges to + 00 linearly for simple interest [case (j)] and more rapidly
for compound interest [case (b)].
522 / Difference Equations

EQUILIBRIUM AND STABILITY

If a difference equation has a constant function yx = y* as its solution, then the value
y* of this function is called an equilibrium or stationary value of yx. An equilibrium
value y*, or the corresponding difference equation, is said to be stable if every
solution of the difference equation converges to y*, independent of the initial condi¬
tions. This is the type of stability referred to in economic literature as “perfect
stability of the first kind.”
A displacement from the equilibrium value is equivalent to a new solution with
different initial conditions, and thus a stable equilibrium can be defined as one for
which any displacement from equilibrium is followed by a sequence of values of y
which again converge to equilibrium.

theorem: For the difference equation

b*+i =Ayx + B
an equilibrium value of y is given by

y* = y~Ta if A^ 1

and y* is stable if and only if — 1 < A < 1, unless yx is constant.

Note that this theorem is closely related to the preceding theorem, which states
that the solution sequence converges to y* if — 1 < A < 1 and diverges otherwise,
unless yx = y0. Stable equilibrium occurs only for cases (a), (d), (e), (f), and (i) of
Table 6.1.

PROBLEMS
Solve each of the following difference equations.

3- i + 3yx = 0 1. 3yx + x = 2yx + 3

4. VX +1 + y* — 2 = 0 2. 2yx +1 + yx — 3 =0

Solve each of the following difference equations, determine the behavior of the solution
sequence, and calculate the first few values of the solution sequence.

5. Tx+ i -- yx -- 10 = 0, y0 = 2 15. lyx + i + 2yx -- 7 = 0, y0 = 1

6. Tx+i == 7y* + 6, y0 = 1 _
16. y^+i + yx + 2 0, y0 = 3
7. 8yx+1 + 4 yx - 3 = 0, y0 = i 17. 15yx + i ~ 10y;V 3 = 0, y0 =
8. 16.Vx + i ~ 6yx ~ 1, y0 = 18. 5yx + i -yx - 60 = 0, y0 = 15
9. 3y^+i ~ 2y x ~ 3 = 0, y0 = 5 19. yx+1 + 4yx + 12 = 0, y0 = 6
<-n|K>

10. 3yx+i ~ Tx 20. 8yx + i + yx - 4 ==

V;
II

II
o

yo = T
11. Tx+1 == 3yx - b To = T 21. 4yx + i ~Tx - 3 == 0, yo = j
12. Tx+ i + 3yx + 1 = 0, y0 = 1 22. 4yx + i + 3yx -- 4 = 0, y0 = 1
13. 2 yx+1 - yx = 2, yo = 4 23. 3yx + i - 2yx -- 6 = 0, y0 = 4
14. yx+i == Vx -- b y0 = 5 24. 9yx + i + 5yx -- 18 = 0, y0 = 1

Answers to Odd-Numbered Problems

i- yx = (i)x(yo - 3) + 3
3. yx = (-3)*y0
5. yx = 2 + lOx; monotone increasing, diverges to + oo; y0 = 2, yx = 12, y2 = 22, y3 = 32,
y4 = 42, y5 = 52
7. yx = i(-i)x + i; damped oscillatory, converges to y* = i; y0 = i yi=i y2 = 1^,
>’3 = ST y4 = H, >’5 = ITS
9. yx = 2(i)x -I- 3; monotone decreasing, converges to y* = 3; y0 = 5, yx = 4-J, y2 = 3f,
ys = 3#, y4 = 3jf, y5 = 3^
 6.5 Applications of Difference Equations in Economic Models / 523

11. yx = i; constant
9
13. yx = 2ft)x + 2; monotone decreasing, converges to y* = 2;y0 = 4, yx = 3, y2 = -f, y3 = T,
17 _ 33
y4 = y 5 — T6
15- yx = i{~i)x + 7 damped oscillatory, converges to y* = J; Vo = 1, ^1
,, _ 5
— 7? ,, _ 39
y 2 — T9,
265 1871 13065
y 3 = JTS, y 4- = TTUT y s = 16807
17. y* = + i; monotone decreasing, converges to y* = f; y0 = 1, yi = ■$, y2 = i,
,, _ 97 ,, _ 55 ,, _ 29
y3 — T75> y4 — ¥T> ^5 — *5
19. yx = ¥( — 4)x — 4^; divergent, oscillates infinitely; y0 = 6, yx — -36, y2 = 132,
y3 — 540, y4 = 2148, y5 = -8604
31
21. yx — (i)(i)x + 1; monotone increasing, converges to y* = 1; y0 = i, yi = J, y2 T2>
127 _ 511 ,, _ 2047
y3 TTS> y4 ~ JTJ, y5 — 2048
_ 14 46
23. y* = -2ft) + 6; monotone increasing, converges to y* = 6; y0 = 4, yx “ T> y2 =
146 ,, _ 454 1394
t lOi
.V 3 y 4 - “8T~ >’5 = TTT

6.5 Applications of Difference

Equations in Economic Models
The following very simple economic models are examples of the use of first-difference
equations to express the dependence of the value of a variable on its value in the
preceding period. The Harrod model, the general cobweb model, a consumption
model, and an income-consumption-investment model are discussed.

THE HARROD MODEL

The following model for national income analysis was proposed by R. F. Harrod.
St = ocYt

I, = P(Y,-Y,_,)

S, = I,
Y0 = Y0 (known value at t = 0)

a>0 p>0
where S is saving, I is investment, Y is income, and each of these variables is a
function of time t.
From the first three equations of the model, the difference equation
aY, = P(Y,-Y,.l)
is obtained for solution and

P
Y,-
p — a.

I, = S, = aj
P
lo
P — a,
Assuming Y0 > 0, the behavior of the solution sequence depends on the value of the
constant p/(p — a). Since Y represents income and is assumed to be nonnegative,

P >0
p — a
524 / Difference Equations

and from the model, a > 0, p > 0, and thus

Since Y* = 0, the sequence is monotone increasing, diverging to + go [case (b)].

Since a > 0, the sequences {/,} and {St} are also monotone increasing, diverging to
+ oo. None of the variables of this model have equilibrium values.
note: The Harrod model above is one of the classical models used to study the
growth of national income in an expanding economy; models for which the solution
sequences do not diverge have also been considered.

THE GENERAL COBWEB MODEL

The adjustment of supply and demand can be studied using the following model.

Supply: q, = a + ppt.i

Demand: pt = y + Sqt

q0 = q0 (known value at t — 0)

where p is price, q is quantity, and both are functions of time t. Substituting the
second equation into the first equation,

q, = cc + py + Pdq.-t

and, substituting the first equation into the second equation,

p, = y + aS + Pdp,-!

Difference equations can be solved in the usual way; however, it is the behavior of
the solution sequence that is of interest. Since p > 0, S < 0, their product pS < 0, and
the solution sequences are thus always oscillatory. The equilibrium point is

y + ocS a + py\
(P*5 q*) =
i-ps'

If — 1 < PS < 0, the sequences {pt} and {qt} are damped and converge to (p*, q*); if
PS = —1, the sequences oscillate finitely; if pS < —1, the sequences oscillate
infinitely. Thus the equilibrium is stable only if — 1 < p3 < 0.

note: If the supply curve is written so that its slope is with respect to the
quantity variable, the slope is 1/p. Thus the equilibrium is stable if and only if the
slope of the demand curve is less (in numerical value) than the slope of the supply
curve.

CONSUMPTION MODEL
The following simple model for consumption has been proposed.

Ct + St = Yt

Yt = aSf _ i

Y0 = Y0 (known value at t = 0)

cc > 0 0 < y < 1

 6.5 Applications of Difference Equations in Economic Models / 525

where C is consumption, S is saving, Y is income, and each of these variables is a

function of time, t, and y is the marginal propensity to consume.
Substituting the second and third equations in the first equation of the model, the
difference equation

15+1 = “(i - y)Y,

is obtained for solution and

Yt = (a - ay)%

C, = yY,

so

C, = y(a - ay)' Y0

and, since C0 = yY0>

c, = (a - ay)'C0
S:=Yt-C,

so

s, = (1 - y)(a - ay)%

and, since S0 = Y0 - C0 = (1 — y) Y0,

St = (oc- ocy)‘S0

Since 0 < y < 1, a — ay>0; the sequences {1J}, {Cj, and {,St} are monotone increas¬
ing and diverge to + 00 if a(l — y) > 1, monotone decreasing and converge to y* = 0
if a(l — y) < 1, and constant at Y* = 0 if a(l — y) = 1.

INCOME-CONSUMPTION-INVESTMENT MODEL
When changes over time are considered to occur periodically rather than contin¬
uously, the income-consumption-investment model stated in terms of differential
equations in Chapter 5 can be stated in terms of difference equations, as follows.

Ct = <xYt + p

it = yYt + g
A Yt_l=6[Ct.1Ylt-l-Yt_1]

Y0=Y0

0 <a < 1 0<y<l O<0<1

Rewriting the third equation and substituting the first and second equations into it,

Y,-Y,-1= 0[C,-i + /,_! -

Y, = e[C,-l+I,-l] + (l-0)Yt-l
+P + yY,.l+g] + (l-0)Y,.l
= [6{a + y) + (1 — 0)]V)_ 1 + 0(/? + g)

1 - [% + y) + (1 - 0)J
= [% + y) + (i - 0)]% + 0(P + a)
1 - [9(a + y) + (1 - 0)]
526 / Difference Equations

The solution is stable if

— 1 < 0(a + y) + (1 — 0) < 1

__<a+?+__i<_
1 1 i 1

2
1—- < a + y < 1
0
Since 0 < 0 < 1 and a + y > 0, the left-hand part of the inequality 1 - (2/0) < a + y is
satisfied. Thus the stability condition is a + y < 1, as for the differential equations
form of the model.

PROBLEMS
1. In the Cobweb model, if the supply curve has negative slope, state the conditions for stable
equilibrium.
2. Derive the solution of the Kahn model:

Ct = «Yt.l + p

Yt = Ct + /,
where C is consumption, Y is income, and / is investment.
3. Solve the following model, determine the behavior of the solution sequences, and state any
additional “ logical ” restrictions on the parameters.

St = (xYt + (J

It = y(Yt-Yt.l)

St = SIt

Y0= Y0

a >0 p > 0 y>0 <5 > 0

where S is savings, Y is income, and / is investment.

4. Solve the following simplified Metzler inventory model:

yt = ut + v0

ut = pyt-1

yo = y0

0 < p < 1

where y is income produced, u is the number of units produced for sale, with units of
measurement appropriately chosen, v0 is the constant noninduced investment, and p is the
marginal propensity to consume of one year’s consumption with respect to the previous
year’s income.
5. Consider the model

Pt = 2 - qt

qt + i = pTV i + 1

P*+ i = aPr+i + (1 - a)pt

Po = Po
0 < a < 1

where q is the quantity supplied, p is the actual price, p* is the supplier’s estimated price for
the next period, and p is a public forecast for the price. Note that a = 1 implies that the
 6.5 Applications of Difference Equations in Economic Models / 527

suppliers have perfect confidence in the public forecast; a = 0 implies they have no
confidence in the public forecast. Defining At+1 by the equation

+1 = P*+1 — pt +1

show that A, = 1 — 2pt.

6. Solve the following model of the growth of national income in an expanding economy:

Yt = Ct + It

Ct = a + pYt

Yt +1 - Yt = ylt

Y0 =Y0 C0 = C0 /„ = I0

a > 0 0 <p < 1 y > 0

where Y is income, C is consumption, and / is investment.

7. Consider the following model of a market:

Qdt = a - pPt (demand)

Qst = ~y + ?>Pt (supply)

Qdt = Qst

a>0 P > 0 y > 0 <5>0

where Pf denotes the expected price in period t and sellers have an “adaptive” price
expectation so that

P? = P?-1 +>l(Pt-i-Pr-i) 0 < r] < 1

a. Show that the model can be represented by the first-order difference equation

P t +1 >?(« + y)
—
P
Hint: Solve the supply equation for Pf and note that Qst = Qdt = a - pPt-
b. Determine the nature of the solution sequence.
2 S
c. Show that the solution sequence, if oscillatory, converges only if 1-<-.
V P
8. In the Harrod model, graph {7f} for each of the following cases and comment on the
plausibility of each.

Case a P ^0
i 2 100
(a) 3
1 2 100
(b) ~3
(c) -4 2 100
i 2 -100
(d) 3

Answers to Odd-Numbered Problems

1. If S < 0 and p < 0, the approach to equilibrium will not be oscillatory and \dp\ <1 (the
stability condition) indicates that the demand curve must be flatter than the supply curve.
528 / Difference Equations

3.
P approaches + oo
cc

s, = \aY0 + j8) approaches + oo

approaches + oo

Sy > cc

rjd
7. b. converges (nonoscillatory)

-l <1 - rj - j <0 converges (oscillatory)

i t]S - t diverges (oscillatory)

*

i r]S i diverges (oscillatory)

6.8 Linear Second-Order Difference

Equations with Constant Coefficients
The method of solution for linear second-order difference equations with constant
coefficients and the behavior of the solution sequence are discussed in this section.
Special methods of solution are required when second-order difference equations are
nonhomogeneous; these methods are considered and equilibrium and stability are
defined.
The general second-order difference equation with constant coefficients can be
written in the form

yx+2 + Aiy*+i + Ai>’x = g(x)

Consider first the special case g(x) = 0:

yx+2 + Aiyx+i + A2 y x = 0
An equation whose constant term is zero is sometimes said to be homogeneous.
Thus yx+ 2 + yx+1 + A2yx = 0 is the general homogeneous second-order linear
difference equation with constant coefficients. The definition of a homogeneous
equation should not be confused with the definition of a homogeneous function; the
two uses of the word homogeneous are not related.
In order to obtain the solution of

yx+2 + Aiyx+i + a2 yx = o
form the auxiliary equation

m2 -f At m + A2 = 0

and solve for its roots, using the quadratic formula (if necessary):

_ -At+y/Al- 4A2
m\ 2

-Ai - JA\ - AA2

2
 6.6 Linear Second-Order Difference Equations with Constant Coefficients / 529

These roots mx and m2 may be real and unequal, real and equal, or complex
(involving the square root of a negative number). The form of the solution of the
equation yx+2 + yx+1 + A2yx = 0 depends on the nature of the roots mx and m2,
as summarized below. The general solution of a second-order difference equation
includes two arbitrary constants and a particular solution is specified by two boun¬
dary conditions, that is, by two consecutive values of y.

Case 1. ml and m2 are real and unequal (m, ^ m2)

solution: yx = Cx mx + C2m2

Case 2. ml and m2 are real and equal (ml = ra2 = m)

solution: yx = Clmx + C2 xmx

Case 3. m1 and m2 are complex (ml = a + bi, m2 = a — hi where i = — 1)

solution :yx = rx(CA cos Ox + C2 sin 0x), where r = ^/u* 2 + fi2 and 6 is the
angle having tan = a/b; alternatively, 6 is the angle having sin 6 = a/r,
cos 6 = b/r.

Note that a and b can be interchanged in these definitions, since the trigonometric
functions are periodic. For convenience, the solution is usually written for 0 in the
first quadrant.

EXAMPLE

Obtain the general solution for the difference equation

y^ + 2 - 5yx+i + 6y* = 0

and the particular solution if y0 = 2, = 5.

The auxiliary equation is

m2 — 5m + 6 = 0

and

5 + 25 - 24
m{ =- = 3

5 - v/25“T4 ,
m2 =-^-= 2

Thus the general solution is

yx = Cl3x + C22x

If y0 = 2, yx = 5,

2 = Cj + C2

5 = 3 Cx + 2C2

C2 = 1

C, = 1

and the particular solution is

yx = 3X + 2X
530 / Difference Equations

EXAMPLE

Obtain the general solution for the difference equation

yx+i - 4yx+1 + 4yx = 0

and the particular solution if y0 =--1, yi = 6.

The auxiliary equation is

m2 — 4m + 4 = 0
and

4 + VI6 - 16 ,
m,= V2 =2

4-^16-16 ,
m2 = 2 =2

Thus the general solution is

yx = C{2X + C2x 2X

If y0 = Uy i = 6,

1 = Cj + 0

6 = 2CX + 2 C2

Ci = 1

C2 = 2

and the particular solution is

yx = 2x + x2x + 1

EXAMPLE

Obtain the general solution for the difference equation

Tc+2 ~ Tc+1 + bx = 0
OJ

and the particular solution if y0 =

m

II

The auxiliary equation is

m2 — m + j = 0

and

1 +s/l -2 , ,
mi = ^ = I + l1

l-v/T-2 , i.
m2 = ^ —IV

r=yifFMip=

tan 9 = 1

0 =b
 6.6 Linear Second-Order Difference Equations with Constant Coefficients / 531

Thus the general solution is

C

(Ci cos inx + C2 sin kx)

If y0 = 3, yi = i

3 = me Ml) + (C2)(0)]

Cj = 3

5-iZ? [,/>/? + C: 'v^

i = I + tC:
C, = 2
and the particular solution is
C

(3 cos i;r:x -I- 2 sin inx)

BEHAVIOR OF THE SOLUTION SEQUENCE

The behavior of the solution sequence depends on both the difference equation and
the initial conditions; the roots of the auxiliary equation indicate the limiting be¬
havior of the solution sequence as follows.

Case 1. Real roots, ml ± mi

If ml is the root with the greater absolute value, that is, m > m- then the
i

limiting behavior of the solution sequence {Clmxl + C2m2) is the same as that of
{Ci mj), provided Cx ± 0.
This can be shown by writing

Cjm? 4- C2mx2 = mx1[C1 + C2(m2/m1)x]

Since — 1 < m2/m1 < 1, (m2/mx)x -> 0 as x -*■ oo and

Cxm\ Cx Cj
= 1 as x 00
C1m^-fC2m2 Ci + C2(m2/m1)x Cl

Thus the limiting behavior of {Ct mx -f C2 m2} is the same as the limiting behavior of
{Ci mi} if | mi | > \m2\. The limiting behavior of {Ci mx} is discussed in the preced¬
ing sections concerning solution sequences of first-order difference equations. This
limiting behavior can be summarized in the notation of the present section as follows.

If Wj <1, the sequence converges

If mi > 1» the sequence diverges
If — 1 < m] <0, the sequence is damped oscillatory
If m < — 1, the sequence oscillates infinitely
If Ci = 0, the solution sequence is {C2 m2} and similar considerations are applicable
to this sequence. Note that if Cx = 0, y0 = C2 and yx = C2m2. Thus Cx = 0 implies
boundary conditions which specify a particular solution.

Case 2. Real roots, mx = m2 = m

The solution sequence is {(Cj + C2 x)mx]y which diverges if m > 1 unless
Ci = C2 = 0 and also diverges if m |=1 unless C2 = 0. If m < 1, the sequence
{xmx} converges to zero and {(Cx + C2 x)mx} also converges to zero. If m is negative,
the sequence is oscillatory.
532 / Difference Equations

Case 3. Complex roots, ml = a + bi, m2 = a — bi

The solution sequence is oscillatory; it converges to zero ifO < Ja1 + b2 < 1 and
diverges if Ja2 + b2 > 1.

A complete classification of the solution sequences of the second-order linear

homogeneous difference equation yx+2 + At yx+l + A2yx = 0 is somewhat in¬
volved, since for exceptional initial values Cl or C2 or both are zero. There is one
case in which the solution sequence for a homogeneous second-order linear differ¬
ence equation converges to zero for every possible pair of initial values, as sum¬
marized in the following theorem.

theorem: If p = max m m ), where and m2 are the roots of the auxi¬

liary equation of the homogeneous second-order difference equation,

yx+2 + + A2yx — 0
then p < 1 is a necessary and sufficient condition for the solution sequence {yx} to
converge with limit 0 for all initial values y0 and yx.

More specifically, the definition of p is as follows.

Case 1. ml and m2 real and unequal

p — max ( m ™2 I)

Case 2. ml and m2 real and equal, ml = m2 = m

m

Case 3. ml and m2 complex, ml = a + bi, m2 = a — bi

p = J a2 + b2

EXAMPLE

In the preceding example the general solution for the difference equation yx + 2 ~ $yx + i +
6yx = 0 is found to be
yx = CX3X + C22x
and, since — 3 and m2 = 2, p > 1 and the solution sequence diverges, as can be seen from
the first few values of the particular solution yx = 3* + 2X: y0 = 2, yx = 5, y2 = 13, y3 = 35,
y4 = 97, y5 = 275, and so forth.
In another previous example, the general solution for the difference equation
yx + 2 — 4yx+l + 4yx = 0 is found to be
yx = Ci2x + C2x2x

and, since = m2 = 2, p > 1 and the solution sequence diverges, as can be seen from the first
few values of the particular solution yx — 2X + x2x + 1: y0 = 1, y i = 6, y2 = 20, y3 = 56,
y4 = 144, y5 = 352, and so forth.
In another previous example, the general solution for the difference equation
y* + 2 - yx+1 + iyx = o is found to be

T* = (Ci cos inx + C2 sin ?nx)

and p = yj2/2 < 1, so the solution sequence is oscillatory and converges to zero, as can be seen
from the first few values of the particular solution

/v/2\x
yx = \— |‘ (3 cos \nx -f 2 sin lynx)
 6.6 Linear Second-Order Difference Equations with Constant Coefficients / 533

NONHOMOGENEOUS SECOND-ORDER DIFFERENCE EQUATIONS

If the second-order linear homogeneous difference equation

y*+2 + ^iy*+i + A2yx = 0

has the general solution yx, then the second-order linear nonhomogeneous difference
equation
^+2 + ^iy«+i + a2 y, = g{x)
has the general solution yx 4- yp, where yp is any particular solution of the nonho¬
mogeneous equation.
Unfortunately, there is no general rule for determining yp. The form of yp depends
on the form of g(x); for simpler forms, yp is similar to g(x), but there are many special
cases.
When g(x) is a constant, yp can be determined quite easily. Suppose g(x) is a
constant, K, that is,

>\+2 + ^i^+i + A2 yx = K

First try a solution of the form: yp equals a constant; say, yp = k. If the solution
yp = k satisfies the difference equation, then

k Ayk A2k = K
and

1 4- Ay + A2

Thus yp — K/( 1 + Ax + A2) is a particular solution of the difference equation

yx+ 2 + ^i^x+i + A2yx = K, unless 1 -I- Al 4- A2 = 0.
If 1 + At + A2 = 0, try a solution of the form yp = kx. Then

k(x + 2) + y4j/c(x -+* 1) -f- A2kx — K

k(l + A± -f- A^x ■+■ k(2 + Ax) — K

But 1 + Ax + A2 = 0, and thus

k = J^2
and yp = F^xJ(A1 + 2), unless 1 + + A2 = 0 and A1 -1-2 = 0.
If both 1 + Ax + A2 ~ 0 and Ax — —2, that is, if A1 — —2 and A2 = 1, then try a
solution of the form yp = kx2. It can be verified that, in this case, yp = (K/2)x2. Note
that this solution is appropriate only for the difference equation yx + 2 — 2yx+1 +
yx = K. Except for this one case, if g(x) is a constant, the difference equation
yx + 2 + Ai yx+l + A2yx — ^(x) has a particular solution, yp, which is a constant.
In summary, when g(x) is constant, the difference equation yx + 2 -I- Atyx+1 +
A2 yx — g{x) has a particular solution as follows:
K
if 1 + -|- A2 ^ 0
p 1 + Ai + A2
Kx if 1 -f- Ax + A2 — 0
Vp = A, + 2 Al + 2 =£ 0

if 1 —F Ay -f- A2 = 0
Ay 4-2 = 0
(that is, if Ay — —2, A2 = 1)
534 / Difference Equations

EXAMPLE

Find the general solution of the equation

y* + 2 ~ 6y*+1 4- 8y*= 9
and a particular solution if y0 = 10, yx = 25.
The auxiliary equation of the homogeneous equation is
m2 — 6m + 8 = 0

(m - 4)(m — 2) = 0

—4

m2 = 2
The general solution of the homogeneous equation is thus

yx = Ct4x + C22x
and the general solution of the nonhomogeneous equation is

y, = C, 4X + C2 2X +
where yp = K/( 1 + + A2) = 9/(1 — 6 + 8) = 3 and thus

yx = Cl 4* + C22x + 3
If y0 = 10, yx = 25,
Ci+C2 = 10-3 = 7

4Ci + 2C2 = 25 - 3 = 22

2 Ci = 8

C\ =4
C2 = 3
and the particular solution is

yx = 4x+l + (3)2* + 3

EXAMPLE

Find the general solution of the equation

y* + 2 + 3yx+i - 4yx = 10
and a particular solution if y0 = 5, yx = 20.
The auxiliary equation of the homogeneous equation is
m2 + 3m — 4 = 0

(m + 4)(m - 1) = 0
mj = —4

m2 = 1
The general solution of the homogeneous equation is thus

y« = c,(-4)* + c2(i)*
and the general solution of the nonhomogeneous equation is

y, = c,(-4)* + Cj + y,
 6.6 Linear Second-Order Difference Equations with Constant Coefficients / 535

where yp = Kx/(AX + 2) = 10/5 = 2, since l+Al+A2=:l+3 — 4 = 0. Thus

yx = Ct(-4 )* + C2 + 2x
If y0 = 5, yx = 20,
Ci + C2 = 5

— 4C1 + C2 = 20 - 2 = 18

-5Cj = 13

Ci = ~2.6

C2 = 7.6
and the particular solution is

yx = — 2.6 (— 4)x + 2x + 7.6

Particular solutions of nonhomogeneous difference equations for which g(x) is not

constant, but is in fact a function of x, can be obtained similarly by examining trial
solutions. However, this procedure may be quite tedious when g(x) is not constant;
these cases are not discussed here.

EQUILIBRIUM AND STABILITY

If a nonhomogeneous difference equation is of the form ^ + 2 + ^1^ + 1 +
A2 yx — K, it can be shown that the solution sequence, of the corresponding ho¬
mogeneous equation converges to zero if and only if 2 ± Al > 1 — A2 > 0, which
implies that 1 + At + A2 f 0. Thus the solution sequence of the homogeneous
equation converges to zero only when the nonhomogeneous equation has a particu¬
lar solution yp = K/( 1 + A1 T- ^42)-
If the solution sequence of the corresponding homogeneous equation converges to
zero, the solution of the nonhomogeneous equation yx + 2 A Axyx+l -f A2yx = K
converges to K/( 1 -I- Ax + A2); if the solution of the corresponding homogeneous
equation diverges, the solution of the nonhomogeneous equation also diverges.
If the equation yx+2 + Ax yx+ Y + A2 yx = K has a particular solution
yp = K/( 1 + At + A2), then yp is an equilibrium or stationary value of y, denoted y*.
An equilibrium value has the property that if two consecutive values of a solution
of the equation yx+2 4- AAyx+l A A2yx = K are equal to y*, then all succeeding
values are equal to y*. Substituting yx = yx+1 = y* = K/( 1 + Al + A2) into
Tx+2 + 1 + A2yx = K,
y x+2 + Ai y* + a2 y* = K
yx+ 2 = K-(AX+ A2)y*
= K_ K(A,AA2)
1 + Aj T A2

1 + Ai A A2
The equilibrium value is said to be stable or the difference equation
y*+2 + Aiyx+i + A2 yx = K is said t0 t>e stable if every solution of the difference
equation converges to y* for every possible set of initial conditions y0 and yl9 that is,
if
lim yx = y* for all y0 and yx
536 / Difference Equations

It can be shown that a necessary and sufficient condition for the equilibrium value
y* = K/( 1 4- Ax + A2) to be stable is p < 1, where,as above, p = max (\ml \ m- )
and Wj and m2 are the roots of the auxiliary equation m2 + A1 m + A2 = 0.
It can be shown that p < 1 if and only if2±Al>l — A2>0. Thus these inequal¬
ities are necessary and sufficient for stability of the equilibrium value y* and
y* = K/( 1 4- Ax 4- A2) is a stable equilibrium for the difference equation
y* + 2 + ^iFx+i + ^2 y x = K if and only if 2 ± At > 1 - A2 > 0.

PROBLEMS
Obtain the general solution for each of the following difference equations.

1. yx + 2 + 2yx+l 4- yx = 0
2- y* + 2 - y* = 0
3. 2yx + 2 - 5yx+1 4- 2yx = 0
4. 3yx + 2 - 6yx +, 4- 4yx = 0
5. Find the general solution for the difference equation yx + 2 + 2yx = 0 and the particular
solution if y0 = 1, yi = J2.
6. Find the general solution for the difference equation yx + 2 4- 3yx+1 4- 3yx = 0 and the
particular solution if y0 = 3, yx = 0.

For each of the following difference equations, obtain the general solution and the particular
solution for the specified initial values.

7- y* + 2 + 4yx+1 4- 8yx = 26, y0 = 6, yi = 3

8. yx + 2 + 8yx+1 + 16yx = 25, y0 = 0, yx = 4
9- y* + 2 - 8yx+i - 9yx = 24, y0 = 2, y{ = 0
10. 3yx + 2 - 10yx+1 + 3yx = 8, y0 = 5, yi = 3

Solve each of the following difference equations, determine the behavior of the solution
sequence, and calculate the first few values of the solution sequence.

11- ^ + 2 - 3yx+1 + 3yx = 5, y0 = 5, = 8

12. 9yx + 2 - 6yx+1 4- yx = 16, y0 = 0, y{ = 3
13. 6yx + 2 + 5yx+l - yx = 20, y0 = 3, yt = 8
14. 4yx + 2 - yx = 15, y0 = 15, yx = 10
15. 8yx + 2 - 6yx+1 4- yx = 9, y0 = 10, yi = 15
16. yx + 2 - 4yx+1 + 4yx = 1, y0 = 0, yx = 1
17. yx + 2 ~ 5yx+1 + 6yx = 4, y0 = 0, = 1
18. yx + 2 ~ 7yx+ 1 + 12yx = 2, y0 = 0, yx = 1
19. yx + 2 - 2yx+1 + 2yx = 3, y0 = 5, yx = 6
20. yx + 2 ~ 4yx = 9, y0 = 0, yt = 1
21. 12yx + 2 - 7yx+i + yx = 18, y0 = 0, yx = 3
22. 3yx + 2 + 5yx+1 + 2yx = 4, y0 = 0, y 1 = 1

Answers to Odd-Numbered Problems

1. yx = ( — l)x (Cj +C2x)
3- yx = Ci(i)x + C2{2)x
5- yx — (v//2)x(cos jtzx + sin jnx)
7. yx = (2V/2)JC(4 cos \nx — \ sin ^nx) 4- 2
9. yx = 3( — l)x 4- i(9)x — \
11. yx = 2(^/3)x(sin \nx) + 5, divergent (oscillatory), y0 = 5, yx = 8, y2 = 5 4- 3^/3, y3 = 5,
y4 = 5 - 9^3, y5 = -22
13. yx = — 5( — 1)* + (i)x_1 + 2, divergent (oscillatory), y0 = 3, yx = 8, y2 = -2|, y3 = 7-&,
_ ^215 ,, _ 7 1
y4 — ~z216> y5 — 1 1 296
15. yx = 6(i)x 4- (i)x + 3, convergent, y0 = 10, yx = 5, y2 = 3|, y3 = 3^, y4 = 3^, y5 = 3^
17. yx = 3X+ 1 — 5(2)x + 2, divergent, y0 = 0, yt = 1, y2 = 9, y3 = 43, y4 = 165, y5 = 571
 6.7 Applications of Second-Order Difference Equations in Economic Models / 537

19. yx = (v/2)*(2 cos \nx + sin \nx) + 3, divergent (oscillatory), y0 = 5, yl = 6, y2 — 5,

v3 = 1, = -5, y5 = -9
21. = - 12(i)x + 9(i)x + 3, convergent, y0 = 0, y^ = 3, y2 = 3i y3 = 3^g, y4 = 3^,
y 5 = 3 W^T2

6.7 Applications of Second-Order Difference

Equations in Economic Models

Second-order difference equations are used in many economic models to indicate the
dependence of the value of a variable on its value for two preceding periods. Second-
order difference equations may be associated with various types of relationships
among variables in the model; some of these relationships are illustrated in Samuel-
son’s interaction model and Metzler’s inventory model, which are discussed in this
section.

SAMUELSON'S INTERACTION MODEL

The following model was proposed by P. A. Samuelson for national income analysis.

Yt = Ct + It + Gt
Ct = a yr_!

It = p[Ct-Ct_l\
Yo=Y0

Yi = Yl

a >0 p >0
where Y is national income, C is consumption, / is investment, and G is government
expenditure. G is assumed constant (exogenous) from period to period and, for
convenience, the units are assumed to be such that G = 1. Note that investment in
any period is equal to a constant times the increase in consumption of that period
over the preceding period; this is referred to as the acceleration principle.
Substituting the second and third equations into the first equation,

Y, = «Y,~i + ~ +1
Yt- a(l +P)Y,.l +cgSy,_2 -1=0
The three types of general solution of the homogeneous equation, depending on the
nature of the roots of the auxiliary equation, are as follows:

(a) Yt = Clm\ + C2 , where

a(l + p) + + P)2 - 4oe/?

m’=- - 2

a(l
+ — - P) v/a2(l + P)2 4a/?
m =--—--
2 2
(b) Yt = Cl mx + C2 tm\ where

m= + P)
538 / Difference Equations

(c) Yt = rt(C1 cos 6t + C2 sin 6t), where

g2(l + p2) | 4ap a2(l +/?2)

r =

a(l + /?)
sin 0 =
2v^
cos 6 =
4a p — a2(l + p2) «(1 + P)
4ap 4/J
A particular solution of the nonhomogeneous equation

Yt-a(l+P)Yt-1+aPYt-2-l=0

is given by
1
^p 1 — a(a + P) + ap
1
1 —a
assuming that 1 — a ^ 0. Since a is the marginal propensity to consume, a < 1 and
1 — a =/= 0.
Thus the solution of the nonhomogeneous equation has the form

(a) Yt = C1m\ + C2mr2 + --

1 —a

(b) Yt = Ct mf + C2 tm1 + --
1 —a

(c) Yt = cos Ot 4- C2 sin Ot) + —^—

where ml9 m2, m, r, and 6 are defined as above and the arbitrary constants can be
determined from the initial conditions T0 and Yl. Whenever the sum of the first two
terms of the solution (in any of the three possible forms) tends to zero, y* = yp =
1/(1 — a) is an equilibrium solution.
The stability conditions 1 + T1+T2>0, 1 — ^ + T2 > 0 and 1 — A2 > 0
become
1 — a(l + p) + ap > 0 or 1 — a > 0

1 + a(l + P) T aP > 0 or 1 + a T 2aP > 0

1 — aP > 0
Since a > 0, p > 0, the second condition is satisfied and the necessary and sufficient
conditions for y* to be a stable equilibrium value are

a < 1

ap < 1

That is, both the marginal propensity to consume (one year’s consumption with
respect to the previous year’s income) and its product with the accelerator parameter
must be less than 1 for the sequence of income values to converge to y* for all
possible initial conditions. The convergence to y* is oscillatory if the roots of the
auxiliary equation are complex numbers. This is the case if A\ — 4A2 is negative, that
is, if a2(l + p)2 — 4aP < 0.
 6.7 Applications of Second-Order Difference Equations in Economic Models / 539

METZLER'S INVENTORY MODEL

The following model was proposed by L. A. Metzler for analyzing inventory cycles.

yt = Mr + St + V0

u, = Py,-i

s, = P(y,-i - >’,-2)
0 < p<1
where yt is total income produced in period t, ut is consumers’ goods produced for
sale in period t, st is consumers’ goods produced for inventories in period t, and v0 is
the constant noninduced net investment in each period.
The total income produced in any period is equal to the total production of
consumers’ goods plus net investment. Sales in any period are a constant proportion
of the income in the preceding period. Production for inventory is equal to the
difference between actual and anticipated sales of the preceding period; that is, there
is an attempt to keep inventory at a constant level. It is assumed that inventories are
sufficient to meet differences between production and consumer demand.
Substituting the second and third equations into the first equation,

y, = Py,- 1 + P(y,-i -y,-i) + v0

y, - + py,-2 = v0
equivalently, so the f-values can begin at t = 0,

y,+2 - 2py,+1 + Py, = v0

The auxiliary equation of the homogeneous difference equation corresponding to
the equation yt + 2 - 2fiyt+l + f}yt = v0 is

m2 — 2 pm + P = 0

m =
2p ± jW - 4^

Since 0 < p<ip2-p < 0 and the roots are complex numbers

m = P ± iy/ftl - P)
Thus

r = V/;2 + M - J) = VP

cos v =
Sp

sin 6 =
^pj^p) = VTzr~p
■Jp
The general solution of the homogeneous equation

y,+2 - 2pyt+i + Py, = °

is thus given by

y, = (sTpyiCi cos 6t + C2 sin 6t)

A particular solution of the nonhomogeneous equation

y,+2 - 2Py,+i + Py, = v0

540 / Difference Equations

is given by

Vo

since 0 < ft < 1 and thus 1 — p ± 0. The general solution of the nonhomogeneous
equation thus has the form

y, = U/MCt cos Ot + C2 sin 6t) +

The term involving sines and cosines results in cyclical fluctuations because of its
oscillation between positive and negative values. These fluctuations are damped by
the factor (v//?)f since 0 < p < 1. Thus

(yfmCx cos Ot + C2 sin Ot) -» 0 as t —► oo

and

Vq
yt-+ as t 00
1-p
as an equilibrium value.

PROBLEMS
1. For the Samuelson model, verify the expressions for ml5 m2, m, r, and 9 given in the text.
2. For the Samuelson model, show that the roots of the auxiliary equation are complex if and
only if

a2(l + p)2 - 4ap < 0.

3. For the Samuelson model, obtain the solution and determine the behavior of the solution
sequence if a. a = 0.8, p = 0.3 and b. a = 0.5, P = 0.1.
4. For the price-adjustment equation

Pt + 2 = Pro + P^iPt + i ~ P,)

where p is price and p, a, and r0 are positive constants, determine the limiting behavior of
the solution if a. pet = 1, b. pa = j, c. pa = 2, d. pa = |.
5. Consider the following difference equation discussed by Baumol:

y,+2--yi+, + -y, = o
S S

where Y is warranted income and C and s are positive constants. Show that for C > 4s this
equation gives an explosive (divergent) solution for warranted income, but for C < 4s
warranted income is cyclic.
6. For the Metzler model, note that (total) inventory, denoted by i, is given by

it = it_1 + st + ut - yt

Obtain an expression for change in inventory, it — it-u in terms of yt and determine the
behavior of this solution sequence.

Answers to Odd-Numbered Problems

3. a. yt - Ci(2)r + C2(f)' + 5, divergent
b. y, = (V2)‘(C, cos int + C2 sin ?nt) + 2, convergent (oscillatory) to 2
Matrix
Algebra

7.1 Introduction
In many economic analyses, variables are assumed to be related by sets of linear
equations. Matrix algebra provides a clear and concise notation for the formulation
and solution of such problems, many of which would be almost impossibly com¬
plicated in conventional algebraic notation.
In this chapter, vectors, matrices, and operations involving them are defined.
Special types of matrices, the transpose of a matrix, partitioned matrices, and the
determinant of a matrix are considered. Linear dependence of a set of vectors and the
rank and inverse of a matrix are also discussed and are applied to the solution of
simultaneous equations. Vector differentiation is defined and illustrated. In Chapter
8 further applications of matrix algebra are considered.

7.2 Definition of a Matrix

In this section matrices are defined and illustrated. A matrix is a rectangular array of
numbers written in the form

I alx al2 • <*1„ all a12

a2\ a22 a2n a21 a22 ®2n
A = or A =
.
\ am1 am2 ®mn j _
^ml ^m2 Qmn
The letters ai} stand for real numbers, the elements of the matrix. Note that ai} is
the element in the iih row and the jth column of the matrix A; the matrix A is
sometimes denoted by (ai}) or by {n(j}. A matrix that has m rows and n columns is said
to be an m x n (read “m by «”) matrix or a matrix of order m x n. If m = n, the
matrix is said to be square. Especially when various operations are being performed
on them, the order of matrices is frequently denoted by subscripts, for example, AmXn
or (djj)m x n •
541
542 / Matrix Algebra

EXAMPLE

10-26 is a 2 x 4 matrix
4 8 3 -9

6 6 3
3 8 -2 is a 3 x 3 (square) matrix
-1 0 0

5 -8 -2
12 10 -1
13 9 -3 is a 5 x 3 matrix
2 7 6
6 4 10

1 -1
is a 2 x 2 (square) matrix
-1 1

Two matrices of the same order are said to be equal if and only if all the corre¬
sponding elements are equal, that is, if the matrices are identical.

EXAMPLE

A = D, but A ^ B, A ^ C, B ^ C, B ^ D, and C ^ D.

DEFINITION OF A VECTOR
A matrix that consists of a single column, that is, an m x 1 matrix, is said to be a
column vector and is written

ui «i
u2 u2
u = or u =

Um

The letters u{ stand for real numbers, the components of the vector; is the ith
component of the vector u. A column vector that has m rows is said to be an
m-component or an m-dimensional vector.
 7.2 Definition of a Matrix / 543

Similarly, a matrix that contains only a single row, that is, a 1 x n matrix, is said
to be a row vector and is written

v = v2, vn) or v = [vl9 v2 , ..., vn]

The letters Vj stand for real numbers, the components of the vector; Vj is the jth
component of the vector v. A row vector that has n columns is said to be an n-
component or an n-dimensional vector.

EXAMPLE

-1
is a 2 x 1 matrix, a 2-dimensional column vector
-1
'O'

0
3 is a 5 x 1 matrix, a 5-dimensional column vector
0
_2_

[0, 3, 0] is a 1 x 3 matrix, a 3-dimensional row vector

[-1, -1, 5, -1] is a 1 x 4 matrix, a 4-dimensional row vector

Two row vectors having the same number of rows or two column vectors having
the same number of columns are said to be equal if and only if all the corresponding
elements are equal, that is, if the vectors are identical.

EXAMPLE

u = [1, 3]

T
V=1 J 3
w = [1, 3]

x = [3, 1]

u = w, but u ^ v, u ^ x, and v ± w, v ± x, and w =f= x.

note: Frequently it is useful to regard a matrix as being composed of a series of

row or column vectors; for example, the matrix

1 -6 "

2 2
-3 4
can be regarded as consisting of the two column vectors

r "-6"
-

L._
4^ N)

2 and
-3
i

or, alternatively, of the three row vectors

[1,-6] [2,2] [-3,4].

544 / Matrix Algebra

7.3 Matrix Operations

Operations analogous to the operations of addition, subtraction, multiplication, and

division of real numbers can be defined for matrices. Since a matrix is an array of
numbers, rather than a single number, some of the properties of operations for real
numbers do not hold for the analogous matrix operations; specific examples are
noted in following sections. In this section addition and subtraction of matrices,
multiplication of a matrix by a scalar, and multiplication of matrices are defined and
illustrated.

ADDITION AND SUBTRACTION OF MATRICES

Matrices can be added or subtracted if and only if they are of the same order. The
sum or difference of two m x n matrices is another m x n matrix whose elements are
the sums or differences of the corresponding elements in the component matrices;
thus, if

ai i • a In

am 1 • a mn

b\ l
B=
_bml b mn

then

A + B = C
where
• • •
a\ 1 ± &11 •" Ain ± bln cn c 1n

... r,
_ aml ± bml •' ‘ amn i bmn _ Cm 1 ^mn

that is, (au) + (bI7) = (c0), where ci;- = atj + btJ for all i and j.

EXAMPLE
i

i
oo
O

m
_

"3 2 -4 ‘ “3 5 4"
o o

H-
4^ O

(a) 5 6 8 + -5 -6 2 = 0
_3 0 0 _ 0 0 —4 _ _3
1
i

3 --10 -6 -4' 9 -6
(b)
-11 25. 22 -21. -33 46.

(c) [4, 6, 12] + [--3, 2, - 12] = [1, 8, 0]

' 1 " " 1 “ " 0“
1 1 0
(d) 2 — 2 = 0
4 4 0
_ 1 . _ 1 _ 0 _
o
■

2 3 1 1 0 3 4
(e) +
6 4 -1 2 6 4 -1 2.
 7.3 Matrix Operations / 545

' 1' " 3" " 6 ' ' O' " -2“

_1

_1

i-

i_
— —

<N rf

oo O

^ -p*.
(f) 1 + + 1

1 1
_ 1. _ 0_

i
i
(g) [2, 1] + [3, 4] + [6, 7] -[11, 12] = [0,0]

' 1 4 " " 0 5 " " 10 13 " -9 -14 "

2 6 — 7 9 — 20 0 = -25 -3
_ 3 8 _ .11 12 _ 1 0_ -9 -4 _

PROBLEMS
Obtain the matrix resulting from each of the following operations.

" 1 " "2 ' " 3“

_1

—
O <N

3. 3 + 1
_ 4 . _ —2 _
i

2 1 -1 0 '2 2
'

1 2 . 0 -1 2 2.

5. [1, 3, -1, 2] - [0, 1, -2, 3]

6. [-1, 2]-[3, 4] + [l, -2]-[6, 5]

Answers to Odd-Numbered Problems

' 1 0 2 “ " 2 “
O

1. 5 6 3. 4
OO

_ -1 -1 _ 0 _
1

5. [1,2, 1, -1]

MULTIPLICATION OF A MATRIX BY A SCALAR

A real number (a 1 x 1 matrix) is referred to as a scalar when it occurs in operations
involving matrices. When a matrix is multiplied by scalar, every element in the
matrix is multiplied by that scalar (number); thus, if

aj ^ * a In
A = and k is any scalar (constant)
ami a mn
then
kalx ••• kaln
t-
n. x A
a rkmXn — Is A k(aij)m x n ~ (kau)
ijfm x n

kami kamn

EXAMPLE

4 ' -3 “ ‘ 12 -9 '
(a) 3 8 -2 — 24 -6
_-l 0 _ _ -3 0 .
546 / Matrix Algebra

0 ' 0 "
-1 = -5
0 _ 0 _

(c) — 1 [6, -2, -3] = [-6,2, 3]

5 6 2 4 5a 6a 2a 4a
(d) a
-3 -1 0 -6 — 3a —a 0 — 6a

0 “ 0 "
-2 2b
3 = -3 b
-1 b
5 _ — 5b _

(f) c[0, 0, 0, 0, 16] = [0, 0, 0, 0, 16c]

MULTIPLICATION OF MATRICES
Two matrices can be multiplied if and only if the number of columns in one matrix is
equal to the number of rows in the other matrix. In particular, the matrix product AB
is defined if and only if the number of columns in A is the same as the number of rows
in B, in this case the matrices A and B are said to be conformable for multiplication
and the product matrix has the same number of rows as A and the same number of
columns as B. Thus an m x n matrix may be multiplied times an n x p matrix to
obtain an m x p matrix.

definition: When a 1 x n row vector multiplies an n x 1 column vector, the

result is a scalar, the inner product of the two vectors; the value of the inner product is
the sum of products of the components of the vectors. Thus if

u = [m19 ..., uj and

v.

then ulx„vnxl — w (a scalar), where w = u1vl + u2 v2 + • •• + unvn = £”=1 ui vi •

When two matrices are multiplied, the element in the ith row and jth column of
the product matrix is the inner product of the ith row vector of the first matrix with
the jth column vector of the second matrix. Thus the product of two matrices may be
written as a matrix of their inner products: If A = (aij)mXn and B = (,bjk)nXp, then
AB = C, where
n

X aljbjm
j= 1 j= i
C = (cik) m y p
n

I amjbj\ X amjbjn
j= 1 j= 1

that is, cik = i aijbjk ■

EXAMPLE

" 1 3 -1 " _ 1 0 ■ ' -3 3 '

2 0 0 -1 2 = 2 0
_ 0 -1 6 _ 3x3 1 3 _ 3x2 7 16 .
 7.3 Matrix Operations / 547

where

(l)(l) + (3)(-l)+ (—1)(1)= -3

(2)(1) + (0)(— 1) + (0)(1) = 2

(0)(1) + ( — !)( —1) + (6)(1) = 7

(1)(0) + (3)(2) + ( —1)(3) = 3

(2)(0) + (0)(2) + (0)(3) = 0

(0)(0) + ( —1)(2) + (6)(3) = 16

3 0
4 0
(b) [-1,0,6, 3, 2]1x5 2 3 = [-12, 38]i x2
1 8
0 -2

where

(-1)(3) + (0)(4) + (6)( — 2) + (3)(1) + (2)(0) = -12

(—1)(0) + (0)(0) + (6)(3) + (3)(8) + (2)( —2) = 38

In matrix multiplication, the sequence in which multiplication is performed is very

important. If A is m x n and B is n x m, then it is possible to obtain both of the
product matrices AB and BA; however, in general, AB # BA. In the matrix product
AB, A is said to premultiply B or, alternatively, B is said to postmultiply A. Since, in
general, premultiplication and postmultiplication give different results, even when
both are defined, care must be taken to preserve the appropriate sequence in all
matrix multiplications. Note that this precaution is not necessary in the multiplica¬
tion of numbers.

EXAMPLE

(a) If
1 2
4 6-13 -1 1
and B =
0-121 1 6
2 3

then

1 2"
4 6 -1 3 -1 1 3 17
AB =
0 -1 2 1 2x4 1 6 5 14 2 x 2
2 3 _ 4x2

1 2 “ 4 4 3 5 '
-1 1 4 6 -1 3 ' -4 -7 3 -2
BA =
1 6 0 -1 2 1 2x4 4 0 11 9
oo

2 3 _ 4x2 9 4 9 __ 4 x 4

(b) If

5 -6
-1 8 3
A = 1 0 and B =
0 10 4
0 3
548 / Matrix Algebra

then

i_i
i
-20

vo O m
" 5 9"

1
1
-1 8 -3
-8

t-h o
AB = -1 3
0 3x2
0 10 —4 2x3 30 12 J 3x3

i
" 5 -6 _

▼-H
m i—H
-1 8 -3 -3

1 1
BA = -1 0

O
0 10 -4 2x3 -12 2x2
0 3 _ 3x2

(c) If
-1 3 1 ' 0 2 3 "
A = 2 0 -2 and B = 8 -1 9
-
0 4 5 _ . -2 0 5 _

' -1 3 1 0 ' 2 3 ' ' 22 -5 29 ‘

AB = 2 0 -2 8 -1 9 =
4 4 -4
0 4 5 3x3 _ —2 0 5 _ 3x3 _ 22 -4 61 _

0 " 2 3
-
~ -1 3 1 “ 4 12 11 '
BA = 8 -1 9 2 0 -2 = — 10 60 55
_ -2 0 5 3x3 0 4 5 _ 3x3 2 14 23 _

note: When a row vector premultiplies a column vector, the result is their inner
product, a scalar whose value is the sum of the products of the elements of the two
vectors. When an n x 1 column vector premultiplies a 1 x n row vector, the result is
an n x n square matrix whose elements are the products of the elements of the
vectors; thus if

u = [iq, ..., H„] and

then, as above, ux x„v„xl = w (a scalar) where w = u{ v{, and v„x jiij x„ = xnXn
(a square matrix), where xtj =

EXAMPLE

(a) [1, 3, 6] — 2 + 12 — 6 = 4

' -2 " " -2 —6 12 ‘

4 [1, 3, 6] = 4 12 24
_ -1 _ _ -1 -3 -6 .
0
-1
(b) [1,0,0, 2, -3] -4 0 +0+0—6+6=0
-3
-2
0" 0 0 0 0 0 “
-1 -1 0 0 -2 3
-4 [1,0, 0, 2,-3] = -4 0 0 -8 12
-3 -3 0 0 -6 9
-2 _ -2 0 0 -4 6
 7.3 Matrix Operations / 549

Although the sequence in which two matrices are multiplied affects the result, the
order in which three or more matrices are multiplied does not affect the result,
provided the sequence is preserved. That is,

x n x p ^p x q Am x n(Bn x p Cp x ( Am x n x p)C p*q

A corresponding property holds for multiplication of numbers. In summary, addition

of matrices is commutative, that is, A + B = B + A and both addition and subtrac¬
tion are associative, that is, A ± B ± C = A ± (B ± C) = (A ± B) ± C; multiplica¬
tion of matrices is not commutative, that is, AB ^ BA, but is associative, that is,
ABC = A(BC) = (AB)C. For numbers, addition, subtraction, and multiplication are
both associative and commutative.

EXAMPLE

B BC

" 1 -1 O' ' 3~ "1 -1 O' '0 6 -3“

2 -3 4 1 [0, 2, -l]lx3 = 2 -3 4 0 2 -1
_0 0 1_ 3x3 _-l_ 3 x 1 _0 0 1_ 3x3 _0 -2 1_-13x3

ABC

0 4-2
0 -2 1
0 -2 1 J 3 x 3

Alternatively, by associativity,

B = A B
_ 3 _
' 1 -1 0“ " 2 ■
2 -3 4 1 [0, 2, -l]lx3 = -1 [0, 2, — 1] 1 X 3

_ 0 0 1 . 3x3 _-l 3 x 1 _ -1 _

ABC

0 4 2
0 -2 1
0 -2 1 _ 3 x 3

EXAMPLE

= ABC
-1
-262 2 x 1
550 / Matrix Algebra

Alternatively, by associativity,

A B c — A B c
"- 1 " ~-l"
1 - i 1 5 -2 3 13 -2 -3

o
_
0
6 10 j 2 x 2 -8 0 6 2x3 -50 -12 78 . 2x3
_ —4_

1
3 x l

i
= ABC
-1
-262 2 x l

EXAMPLE

A B C A B C
3 4 0 1 -6 -8 10 3 4 -5 5 4
4 3. 2x2 -1 0 2x2 -5 5 4 2x3 4 3 2x2 6 8 10 2 x 3

ABC
9 47 28
-2 44 14 2 x 3

Alternatively, by associativity,

A B = A B
3 4 0 1 -6 -8 10] -4 3 -6 -8 10
4 3 2x2 -1 0 2x2 -5 5 4]2x 3 _ . -3 4. 2x2 -5 5 4. 2x3

ABC
9 47 -28
-2 44 —14 J 2x3

EXAMPLE

A B c = A B C

0“ r o 0 01
2 -2 0 12
L—i

o
'sO
r—i

[L — 1, 4, 2] ! x 4 = [1, -1,4, 2]1x4

1

-1 1 0 -6
0_ 4x 1 0 0 0

= ABC
= [6, 0, -36]! x 3
Alternatively, by associativity,

A B C =A B C

0
2
[1, -1, 4, 2]
1
[ — 1, 0, 6] 1 x 3 -6 X [ — 1, 0, 6]! x 3

0 4x 1

= ABC
= [6, 0, — 36] i x 3
 7.4 Special Types of Matrices / 551

PROBLEMS
Obtain the matrix resulting from each of the following operations.

-1 2 3 4
1. 3 0 - 1-1 2
12 0 3

2 -r
2 1
2. i o
1 0
3 -3

2 1
1 3
3. [1, -1,0,2]
-1 0
0 2

1 -f
4. [1, -2,0] 3 0
2 1
-i r '2 0 T 1 1
5.
i -1 0 2. 111
2 O'
1 0
6. [1, -1, 1] 0 2
0 1
2 2

Answers to Odd-Numbered Problems

6 9 12 '

1. -3 -3 6
6 0 9

3. [1/2]

rooo
*"000

7.4 Special Types of Matrices

In this section three special types of matrices are discussed. These are diagonal
matrices, identity matrices, and null matrices.

DIAGONAL MATRICES
A diagonal matrix is a square matrix that has zeros everywhere except on the main
diagonal, that is, the diagonal running from upper left to lower right; thus

A = ij/n x n
= (aij)n

_ani

is a diagonal matrix if and only if

au = 0 for i ± j

ai} ± 0 for at least one i = j (if all the elements of a matrix are zero, it is a null
matrix discussed below)
552 / Matrix Algebra

An n x n diagonal matrix may be indicated by the notation

a\i 0 ~ a\ 0"
A = • or

-1s:
0 0 an_

c
or by

~d1 0

0 ' dn

EXAMPLE

-10 0 0
"1 0 0“

0 0 0 0
0-3 0 and
0 0 0 0
_0 0 10 _
0 0 0 7

are diagonal matrices.

IDENTITY MATRICES
An identity matrix is a diagonal matrix each of whose diagonal elements is postive
one;thus

rfl i i aln
A = ij/n x n
= (aij)n

L^ni ®nn

is an identity matrix if and only if

au = 0 for i ^ j

ai} = 1 for i = j

or equivalently, a diagonal matrix D„ is an identity matrix if and only if

di— 1 for i — 1, 2, ..., n

An n x n identity matrix is denoted by I„.

EXAMPLE

1 0 0 0 0 0
"1 0 0"
0 1 0 0 0 0
0 1 0 and I6 0 0 1 0 0 0
I3
0 0 1 0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
are the 3x3 and 6x6 identity matrices.

Note that premultiplying or postmultiplying a matrix by the appropriately sized

identity matrix leaves it unchanged—that is,
Am x n
=IA
*m x n
— A x n
l=A x n
 7.4 Special Types of Matrices / 553

EXAMPLE

A2 x 4 = I2 A2x4
2 -3 -6 4 0 2 -3 -6 4
-1 -1 0 3. 2x4 1 2x2 -1 -1 0 3 2x4

A2 X 4 I4

1 0 0 0"
2 -3 -6 4 0 1 0 0
- 1-1 0 3 2x4 0 0 1 0
0 0

A2 4 X

2 -3 -6 4
- 1-1 0 3 2x4

EXAMPLE

A3 X 3 — 13 A3 x 3
‘3 0 o' "l 0 0 ■ "3 0 0 '
0 -2 0 = 0 1 0 0 -2 0
_0 0 6 _ 3x3 .0 0 1. 3x3 _0 0 6_

= A3 3 X i3
[3 0
°1
ri 0 01
= 0 -2 0 0 1 0
_0 0 6_ 3x3
_o 0 1_

= A3 3 X

r3 0 O'
= 0 -2 0
_0 0 6. 3x3

NULL MATRICES
A null matrix is an m x n matrix all of whose elements are zeros; it is denoted by O.
When the appropriately sized null matrix is added to or subtracted from another
matrix that matrix is unchanged, that is,
a +0
x n —— x n ~
— A x n

Premultiplying or postmultiplying a matrix by the appropriately sized null matrix

results in another null matrix—that is,

Ofc x m Am X n ~ ^k x n and Am x n 0„ x 1 = xj

EXAMPLE

A2 X 3 ± O2 X 3 =
A2 3 X

0 1 0 0 0 0 0 1 0
+
10 0 -3J 2x3 [0 0 oj 2x3 [10 0 -3J
554 / Matrix Algebra

EXAMPLE

O 3x2 l2 x 4 o 3x4
“0 0 " 0 0 0 0
1 6 4 -10
0 0 0 0 0 0
-1 20-13 11 2x4
_0 0 _
3x2 0 0 0 0J 3x4

l2 x 4 o 4X1 =o 2 X 1

1-
1
O
_
O
1 6 4 -10 0
-1 20 -13 11 2x4 0 0 2x1
0 4 x 1

7.5 The Transpose of a Matrix

In many applications, the transpose of a matrix is used. The transpose of matrix is

defined in this section and the transpose of a sum or difference of matrices and the
transpose of a product of matrices are discussed.
The transpose of an m x n matrix A is an n x m matrix, denoted by A', whose rows
are the columns of A and whose columns are the rows of A. Thus if

” al 1 " ' al n~

x n ! ! \®ij)m x n

- 1 &mn-

then the transpose of A is

all ain~ " al 1 am 1

A' x m - =z
~ {@ji)n x m j)m x n
- ^ml ®mn - - ^1 n ^mn _

Note that the transpose of an rc-dimensional row vector is an rc-dimensional column

vector and, similarly, the transpose of an n-dimensional column vector is an n-
dimensional row vector. The transpose of a diagonal matrix is that same diagonal
matrix.

EXAMPLES

"3
2
If A = then A'= 2
0 2x3
Lo 3x2

If A = [6, 2, -1, 0, -5]ix5, then A' =

5 x 1

~ 6 -8 -10" 6 -2 30 "
If A = -2 0 7 , then A'= -8 0 40
30 40 0 _ 3x3 _-io 7 0_
 7.5 The Transpose of a Matrix / 555

3
0
If A = , then A' = [3, 0, -1, 4] 1 X 4
1
4 J 4 x 1

r-l 5 -3 “I r-i 5 -3 "

If A = 5 0 4 ,then A'= 5 0 4
_ -3 4 9 . 3x3 _ —3 4 9 _ 3x3

"6 0 0 O' 6 0 0 0
0 -2 0 0 0 -2 0 0
If A = ,then A'
0 0 3 0 0 0 3 0
_0 0 0 5. 4x4 - 0 0 0 5J 4 x4

If a (square) matrix and its transpose are equal, that is, if a0 = ayi for all i and
then the matrix is said to be symmetric (about its main diagonal). In the last two
examples above, the matrices are symmetric.

A symmetric matrix that reproduces itself when it is multiplied by itself is said to

be idempotent. Thus A is idempotent if and only if

A' = A

and

A A = A

EXAMPLE

The identity matrix (of any size) is idempotent, since

I„ = In
and

In In

EXAMPLE

The matrix
i
3 i
4
3

is idempotent, since
1 2 I /
3 3 3 3
2 4
17 3j -3 3.

and

3 ilfy 3 3 3

3 3JI3 3 3 3
556 / Matrix Algebra

TRANSPOSE OF A SUM OR DIFFERENCE OF MATRICES

The transpose of a sum or difference of matrices is equal to the sum or difference of
the transposes of the matrices; thus
(\ + n + c
m x n— ** m x n— v-' m x r/
Y = A' + R'
4- C'
^ n x m — ** n x m — ^ n X m

that is,

i^ij)m X n {dji)n x m

where dtj = atj ± bu ± and djt = ajt ± bji ± Cji.

EXAMPLE

If

0 2 0 0 6
A =
8 -1 -2 5 25

then

"2 -5"
2 6 6 f

[A + B + C]' = 6 11
-5 11 25
_6 25 _

or, alternatively,

"3 -6“ ‘-I 3“ "0 -2" "2 -5"

A' + B' + C' = 0 8 + 6 -2 + 0 5 = 6 11
_2 -1 _ _ —2 1_ _6 25 _ _6 25 _

EXAMPLE

If

6 8 0
B = C =
.10 --12. 0
then
f
-4 -10 —4 - 10
[A - B + C]' =
-10 11 -10 11
or, alternatively,

2 0 6 10 0 0 -4 -10
A - B' + C' = +
-1 -1 8 -12 -1 0 -10 11

TRANSPOSE OF A PRODUCT OF MATRICES

The transpose of a product of matrices is equal to the product of the transposes of the
matrices in reverse sequence; thus

x n x p Cp x J ^q x P x « An x m
 7.5 The Transpose of a Matrix / 557

EXAMPLE

If
" 6 ‘
3 0 ' 3 5 -7'
A = B = C = -1
-4 -1 0 -1 8
0_
then
39
ABC =
-53

39
[ABC]' =
-53

= [39, -53]

or, alternatively,
3 0
3 -4
[ABC]' = C'B'A' = [6, -1, 0] 5 1
0 -1
7 8J

= [39, -53]

EXAMPLE

If
6 O'
0
[3, -1,0] B = -7 2 c =
1
1_
o
i

then ABC = — 2 and [ABC]' = — 2 (the transpose of a scalar is that scalar) or, alternatively,

3"
6-7 0
[ABC]' = C'B'A' = [0, 1] -1 = -2
0 2 3
o
i

PROBLEMS
Obtain the transpose of each of the following matrices.

2 6
1.
5 -3

-3'
2. 6
5
-2

-7 -5 6
3.
8 10 -3

r 1 -6 3
4. -1 4 -8
0 5 7

5. [3, -2,4, -9]

-1 3
6. 2 1
L-4 2
558 / Matrix Algebra

1 2 2
7. If A find:
2 1 4 0

a. [A + B]'
b. [B + C]'
c. [AB]'
d. [ABC]'

‘ 2 -1“
0 2
8. If A = 1 2 B =
-i 2 -4
_ 0 1_

and D =

a. [AB]'
b. [C + D]'
c. [D'C]'
d. A + D' + C'

Answers to Odd-Numbered Problems

2
1.
6

8
3. 10
-3

5.

6
7. a.
1

6
b.
3

6 0
c.
3 6
6 12
d.
3 18

7.6 Partitioned Matrices

It is frequently convenient to partition a matrix into submatrices. These submatrices
can then be treated as scalars in performing operations on the original matrix.
Partitioning of a matrix is indicated by horizontal and vertical lines between rows
and columns.
The m x n matrix A may be partitioned, for example, as follows

A = ; A2)

where Aj is m x nx, A2 is m x n2, and n1 + n2 = n. The transpose of a partitioned

matrix can be written in terms of the transposes of its submatrices; thus
 7.6 Partitioned Matrices / 559

EXAMPLE

If

'4 -3 i 5 0 "

A = [Aii A2] = 2 -1 : 1 6
_8 -2 j 3 -7.

then

If they are partitioned conformably, partitioned matrices can be added, sub¬

tracted, or multiplied. If an m x n matrix A is partitioned A = [Ax j A2], where A1 is
m x nu A2 is m x n2, and nx + n2 = n, and an m x n matrix B is partitioned
B = [Bi i B2], where B1 is m x nl9 B2 is m x n2, and n1 + n2 = n, then
A + B = [Aj + Bj j A2 + B2]
Similarly, if

where A is m x n, Al is mi x n, A2 is m2 x n, and + m2 = m, and

B
[b2]
where B is m x n, Bj is mi x n, B2 is m2 x n, and + m2 = m, then
Ai ± Bi
A+ B
A2 ± B2

EXAMPLE

If

‘ 3 4 -2 0“ 3 0 2 4
A= -1 0 5 6 and B = — 1 -1 -2 5
7 3 3 2_ 5 4 3 1

then

A + B = (A i: A2) + (Bi i b2)

‘ 3 4 ; -2 0" '-3 0 2 4"

= -1 0 5 6 -f -1 -1 -2 5
7 3 3 2_ 5 4 3 1 _

0 4 0 4
-2 -1 3 11 — [Ai -}- Bx! A2 + B2]
12 7 6 3
560 / Matrix Algebra

A + B= +
B2
' 3 4 -2 0“ "-3 0 2 4
_ -1 0 6 -1 -1 -2 5
5 +
_ 7 3 3 2_ _ 5 4 3 1

r o 4 0 4 1 Ai -f Bj
=
-2 -1 3 11 =
At Bi
L 12 7 6 3 _

Many other partitionings are possible and will give the same result.

Partitioning is frequently conceptually convenient when matrices are to be added

or subtracted, but the computational advantages of matrix partitioning are asso¬
ciated primarily with multiplication and other more complex operations. Partitioned
matrices must be partitioned conformably for multiplication. If an m x n matrix A is
partitioned A = [Ax | A2], where Aj ism x wl5 A2 is m x n2, and n1 + n2 = n, and an
n x p matrix B is partitioned

where Bx is nl x p and B2 is n2 x p, then

B,
AB = [Ai A2] — A i B j -f- A2B2
B2

EXAMPLE

If

3 0 -1 7 r 2 i i
A = -2 4 1 and B = i 3
1 -1 2_ -l 1 _
then

AB =
r -2 3
0
4
-1 1
1
r 2
l
17
3
L i -1 2_ L-i 1J
3 0 1
2 1
2 4 + 1 [-1, 1]
1 3
1 -1 2
6 3 1 -1
0 10 + -1 1
1 -2 J -2 2.

7 2
-1 11
-1 0

which can be verified by direct matrix multiplication.

 7.6 Partitioned Matrices / 561

Matrices may be partitioned into more than two submatrices. In fact, it is possible
to partition an m x n matrix into a maximum of mn submatrices; note that maximum
partitioning is equivalent to no partitioning, since each element is treated as a scalar
matrix. Unless there is a logical partitioning in terms of the variables in a problem,
matrices should be partitioned for ease of computation; this involves some reason¬
able compromise between minimizing the number of submatrices and minimizing
their maximum size.
Frequently matrices are partitioned once horizontally and once vertically. Thus
the m x n matrix A may be partitioned

Al 1 Ai2
A =
^21 A22.
where An is mi x nx, A12 is ml x n2, A21 is m2 x nlf A22 is m2 x n2, and
mi + m2 = m, nx + n2 = n. Then

If the n x p matrix B is partitioned

[Bn ®12
B =
®21 B22
where Bn is nx x p1? B12 is nx x p2, B21 is n2 x px, B22 is n2 x p2, and nx + n2 = n,
px + p2 = p, then A and B are partitioned conformably for multiplication and

^11
i Ai2.l lBu..i.. ®12
^21 i ^22 -1 Ib21 i ®22

^11 ®11 F ^12 ®21 ^11 ®12 "h ^12 ®22

^21 1 F ^22 ®21 A2i B12 ■+■ A22 B22

EXAMPLE

If
1 3 l'
3 -2 1 0 -1
-1 0 i
A = and B = 5 -1 4 3 2
2 -1 4
3 -2 0 1 -1
0 2 -3 _
then

AB =
An ; An 1 [Bn b12
A2i | A22| IB2i i b22J
r i 3 i 1
0 1 1
-2 ; 1
i ■
-

-l 0 1 r3
5 -1 i 4 -3 2
2 -i i 4
-3 --2 | 0 1 - 1_

. 0 2 ; 3 J
and
1 3“ " r
[3 -2
»i i + a12 B2 i = -1 0 + i [3, -2]
5 -1
2 - 1 _ _ 4_

18 -5" 3 -2“
-3 2 + 3 -2
. 1 —3 J L 12 — 8J

21 -7
0 0
13 -11 J
562 / Matrix Algebra

1 3 " T
1 0-1
An ®i2 T Ai->
12 B->?
D2 2 — —
-1 0 + 1 [0, 1, -1]
4-3 2
2 -1 _ _4_

-1
13 -9
0
51 r° 1
1 + 01
-1
-1
-2 3 -4J |_0 4 -4

13 -8 4
-1 1 0
-2 7 -8

-2
A21 Bi 1 4- A22 B2i — [0, 2] + [ —3][3, -2]
-1
= [10, —2] 4- [ — 9, 6]

= [1, 4]

0
A2i Bi2 4- A22 B22 — [0, 2] + [ — 3] [0, 1, -1]
-3

= [8, —6, 4] 4- [0, -3, 3]

= [8, -9, 7]

Thus

An Bn 4- Ai2 B2i ; An Bi2 4- A12 B12 1

AB
A21 Bn 4- A22 B21 ! A2j Bj2 + A22 B22 J
-7 i 13 -8 41
r21
0 0 i -1 1 0
13 -11 : -2 7 -8
_ 1 4 8
-9 7 .

which can be verified by direct matrix multiplication.

PROBLEMS
1. Compute the following.

6 r 4 2"
a. 2 0 -3 -3 0 1
_ -1 2. _—5 -1.

Use partitioned matrices to check the results.

"5~ "-1 "

2. If U = [1, -1, 4], X = [0, 1, 2], V = 0 , and Y = -1 , find
_1 _ 2 _

a UV + XY
b. 5UV 4 10[X(2V - Y)]

3. If A is 2 x 3, B is 4 x 3, C is 3 x 3, and D is 3 x 2, determine the shape of

a. AC d BC
b DA e. DAC
c. AD f. BCDA
 7.6 Partitioned Matrices / 563

4. If

' 1 2 3“ ' 3 1 O'

A = 0 -1 1 B = 1 -1 2
_ 2 3 0. _0 2 1.

find 2(AB — BA) and use partitioned matrices to check the result.

5. If

" -1 -2 -2" "-3 -6 2“ '-5 -8 O'

A = 1 2 1 B = 2 4 — 1 c = 3 5 0
_ -1 -1 0. 2 3 0. 1 2 -1_
show that
a. A2 = B2 = C2 = I c. BC = CB = A
b AB = BA = C d. AC = CA = B
Use partitioned matrices to check the results.

6. If

show that A2 — 11A + 101 = O. Use partitioned matrices to check the result.

7. If

2 2 3
2 5

find a. AB — 2CD
b. A2
c. (BC)2

8. If

■ 1" ' 4 2 -r "1 0 0"

U = [1, 0, 1] V = 2 X = -1 3 0 Y = 0 1 0
_3 0 1 1_ _0 0 1_
find a. UV
b. VU + X
c. XY
Use partitioned matrices to check the results.

9. If

' 2 -2 -4' "-1 2 4"

U = -1 3 4 V = 1 -2 -4
1 -2 —3 _ -1 2 4_

show that
a. U2 = U and V2 = V
b. UV = VU = O
c. U + V = I
Use partitioned matrices to check the results.

10. If

2 3"
X = 2 3
-2 -3.

show that X2 = O. Use partitioned matrices to check the result.

564 / Matrix Algebra

11. If

' 1 -2 1 " " 2 5 -1 -7"

A = 2 1 -3 X = -2 1 3 4
_ —5 2 3 _ 3 2 1 2_
" 3 6 0 -6“
Y = -1 2 4 5
4 3 2 3. >

show that AX = AY (although X ^ Y) and use partitioned matrices to check the result.

12. If
"-1

i
1 UJ to
_1
" 1 -2"

1
0 -2 3
0 3 B = C = D = 1
0 1 3
_2 -1_ [_— 3 J 2_
find AB - CD .

Answers to Odd-Numbered Problems

0 — 4“
0 —9
_ 13 7_

29 13
-6 -3

3. a. 2 x 3
b. 3 x 3
c. 2 x 2
d. 4 x 3
e. 3 x 3
f. 4 x 3

' -24 -32'

7. a.
-24 -32

1 0
b.
1 0
0 0
16 16

7.7 The Determinant of a Matrix

The determinant of a matrix is a scalar (number), obtained from the elements of a

matrix by specified operations, which is characteristic of the matrix. Determinants
are defined only for square matrices. Methods for obtaining determinants and
properties of determinants are discussed in this section.
The determinant of a 2 x 2 matrix

^ _ a\ 1 al2

a2\ a22

is given by

det A = I A ~ a\1 a22 a12a21

 7.7 The Determinant of a Matrix / 565

EXAMPLE

3 -6 3 -6
If A then A| = 3 - (-24) = 27
4 1 4 1
-1 0 -1 0
If A then I A = -10 - 0 = -10
6 10 6 10

Similarly, the determinant of a 3 x 3 matrix,

-an a\2 al3~
A= a2l a22 a23
- a3l a32 a33-
is given by

| A | = i ^22^33 “E ^12^23^31 + ^13^32^21 al3a22a3l a23a32all

— a33a2la12
The terms can be obtained by the rule illustrated in Figure 7.1, where positive
product terms are formed of elements connected by solid lines and negative product
terms are formed of elements connected by dashed lines. Note that similar rules are
not valid for higher-order matrices.

By definition, a determinant of order n, | A | = | af;- | with i,j = 1, 2,..., n, is the

algebraic sum of n\ terms, each representing the product of n elements selected one
from each row and one from each column in all possible and distinct combinations,
having either a positive or negative sign depending on whether the number of inver¬
sions in the order of j (after the order of i in the product term has been arranged in
numerical ascendency) is even or odd.
Using this rule to obtain the determinant of a matrix larger than, say, 3 x 3 is
tedious. For example, the determinant of a 4 x 4 matrix is the algebraic sum of
4! = 24 terms, each of which is the product of four elements; the determinant of a
5x5 matrix is the algebraic sum of 5! = 120 terms, each of which is the product of
five elements.
Determinants of matrices larger than 3x3 are frequently computed by a
procedure known as expansion by cofactors. For example, the determinant of the
3x3 matrix above can be written
— a{ j ia22a33 — a 23a32) al2 (a2 l a33 a23 a m) + a13(a21 a32

a22 a23 a21 a23 a2l a22

= an — a12 + al3
a32 a33 a31 a33 a3l a32

Figure 7.1
566 / Matrix Algebra

Note that each determinant in the sum is the determinant ot a submatrix of A

obtained by deleting a particular row and column of A. These determinants are
called minors.
definition: Let MtJ denote the (n — 1) x (n — 1) matrix obtained by deleting
the ith row and the jth column of A„ x „. The determinant

I My |
is a minor of the matrix A. The scalar

Ctj = (— 1)I + J | My |
is referred to as the cofactor or signed minor of the element of the matrix A. The
n x n matrix

(<+)'
is referred to as the adjoint of A and is denoted by adj A.
As noted above, the determinant of a matrix can be obtained by a procedure
known as expansion by cofactors. The determinant of A can be expanded in terms of
row i by the formula
n

IA | = £ CLijCij for any row i= 1, 2, ..., n

j= i
and in terms of column j by the formula
n

A | = Y, aijCu f°r any column j = 1, 2, ..., n

i= 1

Thus the determinant of A3 x 3 given above as

a 22 a 23 a 21 a 23 a2l a22
= ai i — a 12 a 13
a32 a33 a 31 a 33 a3l a32

can be written
3

A | — Q\iC n + al2C 12 + ai3C 13 — ZfliiCi,

j= i

EXAMPLE

Find the determinant of the matrix

"3 0 -2
A = 6 -8 1
0 3 4

A| = -96 - 36 + 0 -(0 + 9 + 0)

= -141

Alternatively, expanding in terms of the first row.

1 6
+ 0-2
4 0
= 3( —32 — 3) — 2(18 — 0)

= -105 - 36 = -141
 7.7 The Determinant of a Matrix / 567

Alternatively, expanding in terms of the second column,

3-2 3-2
A = 0 — 8 -3 , *
11 0 4 6 1

= 0 - 8(12 - 0) - 3(3 + 12)

= -96 - 45 = -141

Note that cofactors are essentially determinants and can be evaluated by further
expansion in terms of determinants of lower order. By repeated expansion, an nth-
order determinant can be written in terms of second- or third-order determinants
which can be evaluated easily. To simplify computation, a determinant should be
expanded in terms of the row or column having the largest number of zero elements.

PROPERTIES OF DETERMINANTS
The following properties of determinants are frequently useful in their evaluation:

1. Interchanging the corresponding rows and columns of a determinant does not

change its value, that is, A | = | A' |.
2. If every element of a row (or of a column) of a determinant is zero, the value of the
determinant is zero.
3. If every element of a row (or of a column) of a determinant is multiplied by the
same constant, the value of the determinant is multiplied by that constant.
4. If two rows (or two columns) of a determinant are interchanged, the sign of the
determinant is changed, but its absolute value is unchanged.
5. If two rows (or two columns) of a determinant are identical, the value of the
determinant is zero.
6. The value of a determinant is not changed if each element of any row (or of any
column), or each element multiplied by the same constant, is added to or sub¬
tracted from the corresponding element of any other row (or column).
7. The determinant of the product of two matrices is equal to the product of the
determinants of the two matrices, that is, | AB | = | A | | B |.
8. The determinant of a diagonal matrix is equal to the product of its diagonal
elements.

EXAMPLES

" 1 3 —3“
If A = 2 0 1 9

-1 4 -2

(Property 1)

1 3 -3
= 201
-1 4 -2

= 0 - 24 - 3 - (0 + 4 - 12)

= -19
568 / Matrix Algebra

1 2-1
A'| = 3 0 4
-3 1 -2

= 0 - 3 - 24 - (0 + 4 - 12)

= -19

(Property 3)

3 3 -3
M*l = 6 0 1
-3 4 -2

= 0 - 9 - 72 - (0 + 12 - 36)

= -57

= 3|A|

(Note: First column of A is multiplied by 3 to form A*.)

(Property 4)

1 3 -3
B 1 4 -2 = 4 + 0 - 12 - (-24 + 0-3)
2 0 1
= 19

= - | A|

(Note: Second and third rows of A are interchanged to form B)

(Property 6)

1 3 -3
C| = 3 4 0 = 4 + 0 + 0 - (-24 + 0 + 9)
2 0 1
= 19

= \B\

(Note: Second row of C is equal to second row of B plus two times the third row of B.)

(Property 2)

1 3 -3
If A = 0 0 0 , then
-1 4 -2

1 3 ■3
AI = 0 0 0
-1 4 2

0 +0+ 0-(0 + 0 + 0)
0

(Property 5)

1 1 -3
If A = 2 2 1 then
-1 1 -2 J
 7.8 The Inverse of a Matrix / 569

1 1 -3
A! = 2 2 1
1 -1 -2

= -4 + 6 -- 1 - (6-1-4)

= 0

(Property 7)

‘ 1 3 -3" "-1 0 2“
If A = 2 0 1 and B = 3 -2 -2 , then
_-l 4 -2_ 1 0 —! _

5 -6 -1
| AB = -1 0 3
11 -8 -8

= 0 - 8 - 198 --(o- 120 - 48)

= --38

1 3 -3
|A| = 2 0 1 = -19
-1 4 -2

-1 0 2
|B| = 3 -2 -2
1 0 -1

= —2 + 0 + 0 — (-4 + 0 + 0)

= 2

and | AB | = | A j | B |.

(Property 8)

3 0 0
If A = 0 -2 0 , then
0 0 1

3 0 0
AI = 0 -2 0
0 0 1

= -6+ 0 + 0-(0 + 0 + 0)
= -6

7.8 The Inverse of a Matrix

The inverse of a matrix is used in solving simultaneous linear equations and in other
applications. In this section the inverse of a matrix is defined and several methods of
inversion are discussed; these methods include inversion by Gaussian elimination,
inversion using adjoints and determinants, and inversion of partitioned matrices.
Properties of inverses are also considered.
If for an n x n (square) matrix A there is another n x n (square) matrix B such
that their product is the identity matrix of size n, that is, if

B,
x n "n x n = I„ = B
U/1 X n
A X n
570 / Matrix Algebra

then B is said to be the reciprocal or inverse of A and is written

B = A~1 = (<jy)- 1 = (aiJ) = (b:J)

(it can be shown that AB = IoBA = I).

note: In ordinary algebra, division of a quantity x by a quantity y is equivalent

to multiplication of x by the reciprocal of y—that is,

Finding the reciprocal or inverse of a matrix is an operation analogous to division

in ordinary algebra; however, although every nonzero number has a reciprocal, there
are square matrices, in addition to the null matrix, which do not have inverses. A
matrix that has an inverse is said to be nonsingular; a matrix that has no inverse is
said to be singular.
A matrix which is not square may have a left or right inverse. Ann x s matrix A is
said to have B as a left inverse if BA = I; in this case I must be s x s and B must be
s x n. Similarly, if AC = I, C is said to be a right inverse of A; in this case I must be
n x n and C must be s x n. It can be shown that if A has both a left inverse B and a
right inverse C, then B = C = A-1 and A is square (and nonsingular).
Obtaining the inverse of a matrix, referred to as matrix inversion, is essential for
the matrix solution of sets of simultaneous linear equations; such equations arise, for
example, in the solution of economic models and in linear programming problems.
Any set of simultaneous linear equations can be written in matrix notation; for
example,

011*1 + 012*2 + 013*3 + •*• + alnXn = C1

021*1 + 022*2 + 023*3 + + 02n*« =

0m 1 *1 + 0m 2*2 + 0m3*3 + **• + amnXn = Cm

can be written

If m = n and A has an inverse, that is, if A is square and nonsingular, then the set of
simultaneous linear equations represented by

xn x 1 C„xi

has the solution

Xnx 1 Anx n x 1

Solution of simultaneous linear equations is considered in detail in section 7.9.

INVERSION OF 2X2 MATRICES

If n = 2, that is, if the set of simultaneous equations consists of two equations in two
unknowns,
 7.8 The Inverse of a Matrix / 571

or, in matrix notation,

A2 x 2 ^2 x 1 = ^2 x 1

then A2xX2 can be obtained directly from its definition, as follows. If

ai i fli12
A =
a21 a22

and A 1 is denoted by

^11 ^12
B =
b2 i b 22
then, be definition, AB = I, that is,

al 1 fl121f ^11 b 12 1 0

a21 a22 J1621 b2 2 0 1.

and thus

an bn “E d\2 b2 1 anb\2 + a12b22 1 0

a2ib\\ + a22b2i d2ibl2 “E d22b22 0 1
and

#11^11 "t~ ^12^21 = 1

^11 b 12 “ “1 ^ 12^22 = 0

a2ib\\ + a2 2 b21 = 0

d2 lb 12 “E a22b22 = 1
Solving these four equations for the four 6,/s,

a22
bU =
d\l d22 do21i w12
Cl

a 12
b 12 —
^11^22 a21a12

~fl21
bo i —
al 1 a22 ~ a21a12

al 1
^22 —
al 1 a22 ' a21a12

Note that the denominator of each of these expressions is the determinant of A

ai i a 12
det A = — $11 d22 do21i a
w12
a 21 a 22

If det A = 0, the 6,/s are not defined and A -1 cannot be obtained. It can be shown
that a square matrix of any size has an inverse if and only if its determinant is
nonzero.

EXAMPLE

Determine, if it exists, the inverse of the matrix

572 / Matrix Algebra

-1 6
4 3

-1 6
— 3 - 24 — 27, so the inverse exists.
4 3

1
b 11
9

-6 _2
b 12
— 27 " 9

-4 _ 4
bn
^27 27

-1 _ 1
b22
^27”27

and thus

-1 6 -1 1 21
4 1
4 3 17 17

Note that

1 0
0 1

EXAMPLE

Determine, if it exists, the inverse of the matrix

-3 15
1 —5

-3 15
= 15 — 15 = 0, so the inverse does not exist, that is, the matrix is singular.

PROBLEMS
For each of the following matrices, evaluate the determinant and find the inverse, if it exists.
Use partitioned matrices to check the results.

1 6 -1 3
1. 7.
0 1 2 7.

1 0 -1 -1
2. 8.
0 1 -1 -1

0 1 ■ o -2 -r
3.
1 0 9. 1 -3 4
-1 -1 -1
0 1
4. "3 1 3“
1 6
10. 3 3 1
'2 -3 _2 0 3_
5.
4 -6

2 2
6
-4 -4
 7.8 The Inverse of a Matrix / 573

" 1 2 — r “2 3 4"
12. 0 -3 2 14. 4 3 1
_4 1 0 _ _1 2 4_

‘ 2 4 6"
13. -1 2 3
1 4 9_

Answers to Odd-Numbered Problems

T -6 " 14 -10 r
1. det = 1,
0 1. 11. det = 1, -15 11 -l
11 -8 l.
3. det = -1, i i
1 0 * i 0
13. det - 24, i i i
i i
5. det = 0, so no inverse i i i
7 3
s 7 J
—17
7. det = -13, 2 1
T7 17 J
7 1 11 7
TT> TO ~T7
3 1 1
9. det = 10, ~ TO -iry T7
2 i 1
~7 7 5 J

INVERSION OF LARGER MATRICES

In principle, the inverse of any nonsingular matrix can be obtained from its definition
in the same way as the inverse of a 2 x 2 matrix is obtained above. That is, if

a 11 ^12
A = a21 a22 a2n
m
anl “n2 @nn

and B = A"1, then AB = I

1

<2l x Ui2 b 12 •" bu 1 C12 clB

▼Hc

_^n
a21 ai„ ••• h2„ ••• /-»
a22 ^21 ^22 = C21 C22 c2n

•4' an„ _ _hnl *•• bnn ... /-«

an2 hn2 C«1 Cn2 nn

1 0

0 ' 1

where cik = Yj=i aijbjk and the n2 elements of the inverse matrix (bjk) can be ob¬
tained by solution of the n2 equations
n

L au bJk = 1 if i = k
i= i

Z aa bjk = 0 if i^k
1= i
where i = 1, 2, ..., n and k — 1, 2, ..., n.
For a matrix of size 3 x 3 or larger, this method of inversion is tedious. For¬
tunately, numerous alternative computational procedures for inverting a matrix are
574 / Matrix Algebra

available and inversion is also a standard program for most electronic computers.
Two methods of inversion are discussed in following sections; both of these are
feasible unless the matrix is quite large or its elements consist of several digits.

INVERSION USING ROW OR COLUMN OPERATIONS

To solve a system of simultaneous linear equations, certain simple operations are
commonly used to convert the original system to an equivalent system—that is, to a
system having the same solution as the original system—which is easier to solve. An
equivalent system is obtained if (1) two equations are interchanged, (2) an equation is
multiplied by any nonzero constant, and (3) the ith equation is replaced by the sum
of the ith equation and k times the jth equation, where k is any nonzero constant.
Similarly, three elementary row operations on matrices are defined by (1) inter¬
change of two rows, (2) multiplication of a row by any nonzero scalar, and (3)
replacement of the ith row by the sum of the ith row and k times the jth row, where k
is any nonzero scalar.
The corresponding operations on the columns of a matrix are elementary column
operations.

theorem: If a matrix A is converted to the identity matrix by a series of row

operations or by a series of column operations, then the same series of operations
performed on the identity matrix will convert it to A-1.

There is a standard method, invariably appropriate, for converting a nonsingular

matrix to the identity matrix or for establishing the singularity of a matrix which
cannot be converted to the identity matrix. This standard method is outlined and
illustrated below. The generality of the method, which is appropriate for any matrix
and which can be programmed for machine computation, makes it especially useful,
even though for particular cases shorter methods are available.

Steps in Converting a Square Matrix to the Identity

Matrix (Gaussian Elimination)
1. Divide the first row of the matrix by the element in its first column; use the
resulting row to obtain zeros in the first column of each of the other rows.
2. Divide the second row by the element in its second column; use the resulting row
to obtain zeros in the second column of each of the other rows.

n. Divide the nth row by the element in its nth column; use the resulting row to
obtain zeros in the nth column of each of the other rows.
note: In the above procedure, if at the kth step the element in the kth column is
zero, interchange that row with a subsequent row having a nonzero element in the
kth column. Then proceed with the kth step.

To find the inverse of a matrix A, it is customary to work with a tableau of the

form

[A 11]
and change it by the standard procedure given above to the tableau

[I|B]
Then, by the theorem above, R is A-1.
 7.8 The Inverse of a Matrix / 575

Even if A has no inverse, this procedure can be started. At some point a tableau
will be obtained which is not of the appropriate form and which cannot be changed
in the manner outlined above. It can be shown that if the above procedure cannot be
completed to obtain [I I B] from [A 11], then A does not have an inverse.

EXAMPLE

Find the inverse, if it exists, of the matrix

‘ 0 -2 -3"
1 3 3
_-l -2 -2.

Following the standard procedure outlined above,

0 -2 -3 1 0 0
1 3 3 0 1 0
_-l -2 -2 0 0 1 _

" 1 3 3 0 1 0 " interchanging

0 -2 -3 1 0 0 first and
-1 -2 -2 0 0 1 _ second rows

" 1 3 3 0 1 0 ~
0 -2 -3 1 0 0 Step 1
0 1 1 0 1 1.

" 1 3 3 0 1 O'
0 1 3 1 0 0
I
0 1 1 0 1 1 _
fStep 2
3 3
' 1 0 ~I I 1 O'
0 1 3 1 0 0
1
o 0 1 1 1 1.
1
3 3
' 1 0 —? 1 1 0 ~
0 1 3 1 0 0
1 ~J
o 0 1 -1 -2 -2 _
VStep 3
■ 1 0 0 0 -2 -3 '
0 1 0 1 3 3
0 0 1 -1 -2 —2 _

Thus

0 -2 -3 ' - i ' 0 -2 -3'

1 3 3 = 1 3 3
_ -1 —2 -2 _ -1 -2 ~2_

and this matrix is equal to its inverse. Note that

i—

—i
-1

o
_1
_1

o
OJ
O

K>
CM

1
1
1

1 3 3 1 3 3 = 0 1 0
_1

_i

_1
o
o
CM

CM
CM

<N

1
1

1
1

EXAMPLE

Find the inverse, if it exists, of the matrix

576 / Matrix Algebra

■ 1 2 3"
-1 0 4
0 2 2_

■ 1 2 3 1 0 0 '
-1 0 4 0 1 0
0 2 2 0 0 1 .

1 2 3 1 0 0 ~
0 2 7 1 1 0 Step 1
. o 2 2 0 0 1.

" 1 2 3 1 0 0 "
7 1 1 0
0 1 I I 1
0 2 2 0 0 1 _
^Step 2
" 1 0 -4 0 -1 0 "
7 1 1 0
0 1 I 1 I
0 0 -5 -1 -1 1 _

■ 1 0 -4 0 -1 0 “
7 1 1 0
0 1 I 1 T
1 1 1
0 0 y y -y J
J^Step 3
4 i 4•
" 1 0 0 y -y -1
i i 1
0 1 0 —y -y IT)
1
— i*
i i
0 0 1 y y
Thus
" 2 3 “ -1 r 4
-y
i
-y
4 “1
1 y
i i 7
-1 0 4 = -y —y TO
i i 1
0 2 2 . L y y -y J
Note that
1 4 1 "1
^*|v>

2 3 " 3 0 O'
1

-y
-i 0 4
1
"3 -y
i 7
TO
= 0 1 0
L o 2 2 J 1 i
LO 0 1J
1

3 y
1

and
r 4
y -y
i
--y
4
1 2 3 ' "i 0 0 "
i
-y -y
i 7
TO -1 0 4 = 0 1 0

L y
i
y
i
--y
1
L o 2 2 j LO 0 1J

EXAMPLE

Find the inverse, if it exists, of the matrix

■ i 2 -r
-3 4 5
i—
-4 2 6 —*

Following the standard procedure outlined above,

" 1 2 -1 1 0 0 '
-3 4 5 0 1 0
-4 2 6 0 0 1 .

" 1 2 -1 1 0 0'
0 10 2 3 1 0 Step 1
0 10 2 4 0 1 .
 7.8 The Inverse of a Matrix / 577

1 2-1 1 0 01
0 1 i A A o
_ 0 10 2 4 0 1 J
\ Step 2
" 1 2-1 1 0 01
0 1 i 3
TO
i
TO
n
u
0 0 0 1-1 1 J
.

The procedure cannot be continued and thus the matrix has no inverse. Note that the
procedure can be discontinued at the end of Step 1; at the end of Step 1 the left-hand matrix
has two identical rows and thus the determinant of the matrix is zero (Property 5) and it has
no inverse. The determinant of the original matrix is also zero (Property 6) and it has no
inverse.

INVERSION USING ADJOINTS AND DETERMINANTS

An alternative method of matrix inversion involves obtaining the adjoint and deter¬
minant of the matrix to be inverted. If A is nonsingular, that is, if | A j A 0, then

A- i adj A

EXAMPLE

Find the inverse, if it exists, of the matrix

0 2 -3
A = 1 3 3
-1 2 -2

| A | = 0-f6-l-6 — (9 + 0 + 4)= -1

Alternatively, expanding by elements of the first row,

1 3 1 +3 1 3
A | = 0 + (— 2)( — 1) 1 + 2 + (— 3)( — 1)
-1 -2 1 -2

= 0 + 2( —2 + 3) + ( —3)( —2 + 3) = -1

or, expanding by elements of the second column,

1 3 0 3 0 -3
A | = (— 2)( — 1)
1 +2
+ (3)( —1)
2 +2
+( 2)( 1) 3 + 2
-1 -2 -1 2 1 3

= 2( — 2 + 3) + 3(0 - 3) + 2(0 + 3) = -1

Obtaining the adjoint,

3 3
Cu = (-l)1 + 1 =-6+6=0
-2 -2

1 3
c12 i
= (- )1 + 2
-1 -2
= -(-2 + 3) = -1

1 3
C13 = (-l)1 + 3 =-2+3=1
-1 -2

-2 -3
c21 l
= (- )2+1
-2 -2
= -(4-6) = 2
578 / Matrix Algebra

0 -

c22 = (- l)2 + 2 -1 -
> = 0 - 3 = -3

0 -

c23 = (- l)2 + 3 -1 -
; = -(0-2) = 2

-2 -

c31 =(-l)3+1
3
* =-6+9=3

0 -3
c32 = (- l)3 + 2 1 3
= -(0 + 3) = -3

0 -2
c33 = (- l)3 + 3 1 3
=0+2=2

Thus

" 0 2
adj A = (Cy)' = -1 -3
1 2

0
A"1 1
-1

(as obtained in the example above by the standard procedure for row operations).

EXAMPLE

Find the inverse, if it exists, of the matrix

'12 3
A = -10 4
0 2 2

| A | = 0- 6 + 0-(0 + 8-4)= -10

Alternatively, expanding by elements of the first column,

0 4 2 3
(~1)1 + 1 (-1)2 + 1 +0
2 2 2 2
= (0 - 8) + (4 - 6) = -10

or, expanding by elements of the second row,

1 2
I A | =(-l)(-l)2 + 1 2 3 + 0 + 4( — 1)2 + 3
0 2
= (4 — 6) + 0 + ( —4)(2 - 0) = -10

Obtaining the adjoint,

Cu (-1 r1 - 8 = -8

-1 4
Cl2 (-1)1+2 — (— 2 + 0) = 2
0 2

-1 0
C13 (-1)1+3 -2 + 0 = -2
0 2
 7.8 The Inverse of a Matrix / 579

2+ 1
2 3
C21 — (~ 1) -(4-6) = 2
2 2

2+2
1 3
C22 — (— 1) 2 + 0 = 2
0 2

2+3 1 2
c23 = (- )i 0 2
-(2 + 0)= -2

3+ 1 2 3
C31 — ( 1) 8 + 0 = 8
0 4

3+2 1 3
^32 — ( 1 ) = — (4 + 3) = -7
-1 4

3+3 1 2
C33 = (“I) =0+2=2
-1 0

Thus

-8 2 8
adj A = (Cuy = 2 2 7
-2 2 2
4 1 4
7 y -y
1 1 1 7
A 1 = adj A = T y TO
A 1 1 1
y y -y

(as obtained in the example above by the standard procedure for row operations).

EXAMPLE

Find the inverse, if it exists, of the matrix

2 - 1 "

4 5
2 6_

40 - (16 + 10 - 36) = 0

Thus A is singular.

In general, inversion by row or column operations is less laborious if the elements

of a matrix are small whole numbers; inversion by determinant and adjoint is less
laborious if the elements of a matrix are large numbers or fractions. For large
matrices both of these methods are tedious, and electronic computers should be used,
if available.
Unless there is good reason to believe that a matrix is nonsingular, as in some
practical applications, it is advisable first to find its determinant, even if inversion by
row or column operations is intended. A number of row or column operations may
be necessary before the singularity of a matrix becomes apparent using this method
of inversion, and much useless computation is thus avoided by establishing the
existence of an inverse before attempting (by any method) to compute it.
580 / Matrix Algebra

INVERSION OF PARTITIONED MATRICES

It is sometimes convenient to obtain the inverse of a matrix in partitioned form. If an
n x n matrix A is partitioned

where An is n{ x nl, A12 is nl x n2, A21 is n2 x nl5 A22 is n2 x w2, and + w2 = n,

then

V[I + A12B *A2i An1] ~An1A12 B

“l —B A^AfV B1
where B = A22 - A21 A^A12 and An and B are nonsingular. This result can be
verified by multiplication:

An A12lfA1-11[I + A12B-‘A21 Af,1] -AuA12 B~‘ I 0

• A2i A22J[ -B-'AjjAjV B’1 0 I

since

An An^I + A12 B xA21 An1] — A12 B JA21 An1 = I

-An An1A12B-1+A12B"1=0
A2i Ax/[I + A12 B 2A21 An1] - A22 B xA21 Aj/
= [I + A21 A111A12 B 1 — A22 B x]A21 An1

[A2i A111A12 — A22]B

= [I — BB_1]A21 An1

= 0
— A21 A111A12 B 1 + A22 B 1 = [ —A21 A111A12 + A22]B
= BB_1

= I

EXAMPLE

Find the inverse, if it exists, of the matrix

" 0 -2 -3"
1 3 3
-1 -2 -2 _

Partitioning the matrix,

 7.8 The Inverse of a Matrix / 581

and

J3 1 /[ 0 ’ -3 3
1
An1 [I + A12 B !A2i Aj/] - 1 [ — 2] E
~2 0. (to 1 3 —T 0
3 0 -6 —12
I 11 ([1
+
i i
1 1 6
“I 0] llo 12. — 7 0
3 0 -3
J 11 1 ~61\
1 +
oj llo 1 3
3 n ' -2 -6
I
1 3 7
—2 0]
0 ■2
1 3
where
4 11 ' —3

»■
-1
B = A22 [ — 2] — [ — 1, -2]

O
3
1
-3
= [-2]-[-i -1]
3

= -2 - (-1)
1
I
and
B -1 = [-2]
3
■1 -3
-Ar/A^B"1 = 1 [-2]
1 3
3
i
1
1 ill.
-3
3
3
1 1
-B-^, An1 [2][ — 1, -2] 1
I 0 = [-1, -2]
3
1 1
= [ —2, -4] 1
1 0

= [-l, -2]

Thus
" 0 -2 -3 “ -1 " 0 -2 -3“
1 3 3 = 1 3 3
_-l -2 -2 _ _-l -2 —2 _

as above.

EXAMPLE

Find the inverse, if it exists, of the matrix

"12 3"
-10 4
0 2 2_
582 / Matrix Algebra

Partitioning the matrix,

r * 1 22 334- : Ai i ; A12
-1 : 0 4 A21 ;
L 0 : 2 2J A22.

and
1 7
-1
t TO
An1 [I + A12 B xA2i An1] — [1]^[1] + [2, 3] 1
3 S
1 0 [1]
-1
= [!]([!] +[it] 0

-i(i-i)
_ 4
=T
where
0 4 -1
B = A22 — A21 A! I1 A1 2 = [1][2, 3]
2 2 0
0 4 -2 3
2 2 0 0
2 7
2 2
and
7
TO
B1 = i 1
5 “5
7
i- i TO
-Ar/AnB*1 = [-l][2, 3] i 1

= [-i -f]
i
-i = J
~B_ 1A21 A^1 i
■J

Thus
1 2 3 " -1 —T
1
-J^ 1
-1 0 4 = 1 1 7
-J —J TO
. 0 2 2 _ 1
L ?
1 1
J —1 J
as above.

PROPERTIES OF INVERSES
The following properties of inverses are frequently useful in their evaluation.
1. The inverse of the inverse of a matrix is the original matrix; that is, [A - *] "1 = A.
2. The determinant of the inverse of a matrix is equal to the reciprocal of the
determinant of the matrix; that is, | A ~11 = 1/1A |.
3. The inverse of the transpose of a matrix is equal to the transpose of the inverse of
the matrix; that is, [A']-1 = [A-1]'.
4. The inverse of the product of two matrices is equal to the product of their inverses
in reverse order; that is, [AB]1 = B_1A“‘.
 7.8 The Inverse of a Matrix / 583

EXAMPLES

" 1 1 4 -i

—i
2 3 " “7 5
If A = -1 0 4 , then A 1 = i 1 7
“7 ~7 115
0 2 2_ 1 1 1
L 7 7 ~7 J
(Property 1)
4 1 4 -1
7 7 “7
[A"1]"1 = 1 1 7
7 iry
1 1 i
7 5
12 3
-10 4
0 2 2
= A

Thus [A~*]~1 = A.

(Property 2)

1 2 3
AI = -1 0 4
0 2 2

0-6 + 0-(0 + 8 -4)

-10
4 1 4
I '7 ~5
-1 1 1 7
'7 '7 ITT
1 1 1
7 7 ~7

= ttt[4 + 4-1-(4+ 14- 1)]

= (tt?)(~¥)

1
IT)

Thus I A -1

(Property 3)
1 -i 0 T~1
(a r1- 2 0 2
3 4 2
4 1 1
? “7 7
1 l 1
5 -7 7
4 7 1
J 115 t
4 1 4
7 ‘7 “7
1 1 7
[A-J]' = ■7 •7 115
l 1 1
7 7 “7
4 1 1
7 “7 7
1 1 1
■7 ~7 7
4 7 1
7 TO ‘7

Thus [A']'1 = [A-1]'.

584 / Matrix Algebra

(Property 4)

"-1 3 r ' -2 3 r
If B = 0 2 0 ,then B i _ 0 i o
_ —2 0 4_ _-l 3 1
T IJ
1 2 3" — 1 3 r
AB = -1 0 4 0 2 0
_ 0 2 2_ 2 0 4_

"-7 7 13"
= -7 -3 15
_—4 4 8.
- i
7 7 13
[AB]-1 = 7 3 15
_ —4 4 8 -

i 18 “I
-to "T
i 7
= “TO TO TO
7
_ -1 0

1 n 4 1 4 ^
"-2 3 2 5 —1 ~5
1 7
B -1 A =
A 'I - 0
i
t 0 — 11 —7 TO
3 1 1 1 1
-1 T I J - I 3 “3 J

1 18 -1
’-ft ~T0
i 1 7
= “TO “TO TO
7
-1 0 4-i

!- 1 A -1
Thus [AB]"1 = BXA

PROBLEMS
1. If
" 3 2 -2~ "1 0“
111
A = 0 1 4 B= 0 1 c= 1 1 1
_-l 0 5_ _0 1_

find a. del [[AB]'C' - D]

b. [BC]"1
c. [D2]-1
If
3 -1" 2 1 "
3 4
o

A = B = -1 -2 C =
1

2 2
2 1_ 1 1 .
find a. [A - BC]'
b. [A'B]’1
3. If
"1 2 “
'2 1 1 0 0 1 0
1 2.
B=
1
c= 3 4 D =
0 .
1 1 1
_ 1 2.

find a. A + B - 2DC
b. A^B1
c. C[A + B]B“1
d. CA“1B”1D
 7.9 Simultaneous Linear Equations / 585

4. If
'0 1 -1 “ " 4 -3 3"
A= 3 -2 3 B= 2 -1 2
_2 -2 3. _ —3 3 -2.

show that a. A2 = B2 = [*[A + B]]2 = I

b. [A - B]2 = O

Answers to Odd-Numbered Problems

1. a. -39
b. inverse does not exist
r 5 12
T5"9
c. 12 5
TSS
-3 r " 5 7“
3. a.
-9 -13 c. 13 15
2 _ 5 7.
3 —3
b. l 1 1 11
“3 i\ 5 7 5
d. 3 3 3
1 1 1

7.9 Simultaneous Linear Equations

This section concerns simultaneous linear equation systems. Linear dependence of a
set of vectors and the related concept of rank of a matrix are defined and applied to
the problem of determining whether a set of simultaneous linear equations has a
unique solution. Methods of obtaining the solution of a set of simultaneous linear
equations are also discussed.

LINEAR DEPENDENCE AND RANK

A set of m vectors al9 a2,..., am each with n elements is said to be linearly dependent if
there is a nontrivial linear combination of the vectors that is equal to the zero vector
with n elements. That is, if there is a set of numbers X2,..., Am (not all zero) such
that
m

^1«1 + A2 a2 + •• • + Xm am = Xxi ai = 0
i= 1
then the set of vectors at, a2,..., am is said to be linearly dependent. If there is no set
of A’s (except all zeros) such that 1 A,-a= 0, the set of vectors is said to be linearly
independent.
A matrix can be thought of as a set of row vectors or a set of column vectors. It can
be shown that, for any matrix, the number of linearly independent rows is equal to
the number of linearly independent columns; this number is the rank of the matrix.
Thus if a matrix is m x n and its rank is denoted by r, then r < min(m, n).

EXAMPLE

Determine the rank of the matrix

" 1 4"
5 10
_3 2_
586 / Matrix Algebra

Since 2(first row) -I- (third row) — (second row) = 0, the three rows are linearly dependent.
However, each pair of two rows is linearly independent; the two columns are also linearly
independent. The rank of the matrix is 2.

EXAMPLE

Determine the rank of the matrix

"3 6"
1 2
_2 4_

The three rows are linearly dependent, since

(first row) — (second row) — (third row) = 0

Each pair of two rows is linearly dependent, since

(first row) — 3(second row) = 0

2(first row) — 3(third row) = 0

2(second row) — (third row) = 0

Note that the two columns are also linearly dependent, since

2(first column) — (second column) = 0

The rank of the matrix is 1.

The following very useful result provides a systematic method of testing for linear
dependence: Consider all the square submatrices of A whose determinants are non¬
zero. The rank of A is the order of the determinant of largest order. Thus one method
of computing the rank of a matrix is to look for the largest nonzero determinant; the
order of this determinant is the rank of the matrix.
The following properties of rank are useful in determining the rank of a matrix:
1. Since the determinant of a diagonal matrix is equal to the product of its diagonal
elements, if A is a diagonal matrix, then r(A) is the number of nonzero diagonal
elements in A. In particular, r(I„) = n.
2. Since any submatrix of A' is the transpose of a submatrix of A and since B =
| B' |, r(A') = r(A).
3. The rank of the product of two matrices cannot exceed the smaller rank of the two
matrices; that is, r(AB) < min{r (A), r(B)}.
4. If A is an n x n (square) matrix, then r(A) = n if and only if A is nonsingular;
r(A) < n if and only if A is singular. Thus if a square matrix is singular, its rows
(and also its columns) are linearly dependent; if a square matrix is nonsingular, its
rows (and also its columns) are linearly independent.

Consider the solution of a general system of n linear equations in n variables x1?

X2, . .. , Xn ,

y = Ax

where y is n x 1, A is n x n, and x is n x 1. If A is nonsingular, the unique solution is

x = A_1y

and, conversely, if y = Ax has a unique solution, A is nonsingular.

 7.9 Simultaneous Linear Equations / 587

More generally, if y = Ax, then the whole set of equations is consistent and has at
least one solution if r(A) = r(A:y). The solution is unique if and only if r(A) =
r(A:y) = n, that is, if and only if A is nonsingular.
In the special (homogeneous) case y = 0, if A is nonsingular, the unique solution is
x = 0 and there cannot be a nonzero solution. Thus, when the set of equations
Ax = 0 does have a nonzero solution, A must be singular.
Suppose that y = Ax and A is m x n (not square), that is, there are m equations in
n variables where m may be less than, equal to, or greater than n; then
r(A) < min(m, n).
The following summary concerning solution of simultaneous linear equations
applies to the general case of m equations in n variables and also to the special case of
n equations in n variables.

If r(A i y) = r(A), the equations in the set are logically consistent and there is at
least one solution

If r(A: y) = r(A) = n, there is a unique solution

If r(Ai y) = r(A) < n, there are infinitely many solutions and the rows (and
columns) of A are linearly dependent

If r(Aj y) ^ r(A), the equations in the set are not logically consistent and
there is no solution

EXAMPLE

Consider the following set of simultaneous linear equations

x + 2y — z = 10

2x + 4y — 2z — 5

x + y + z — 6

The determinant of the coefficient matrix A is zero

1 2 -1
2 4 -2 = 4- 2- 4-(-4- 2 +4) = 0
1 1 1

so the rank of the matrix is not 3; the rank is 2 since, for example,

2 4
=2-4=-2^0
1 1

The augmented coefficient matrix [A: Y]

i

"1 2 -1 10 "

2 4-2 5
1116

has rank 3 since, for example,

2-1 10
4-2 5 -24 4- 40 - 5 - (-20 + 10 - 24) = 45 =£ 0
1 1 6
1 hus r(A) ± r(A; y) and the set of equations has no solution since they are not all consistent.
Note that the first and second equations are clearly inconsistent.
588 / Matrix Algebra

EXAMPLE

Consider the following set of simultaneous linear equations

x + 2y — z = 10

2x 4- 4y — 2z = 20

x + y + z = 6

The coefficient matrix is the same as that for the preceding example and its rank is 2. The
augmented coefficient matrix (A j y)

"1 2 -1 10”

2 4 -2 20
_1 1 1 6_
is also rank 2, since

1 2 10 1 -1 10 2 -1 10
2 4 20 —
2 -2 20 = 4 -2 20
1 1 6 1 1 6 1 1 6
Thus r(A) = r(A; y) = 2 ^ n = 3, and the set of equations has infinitely many solutions; the
rows (and columns) of the coefficient matrix are linearly dependent. Note that the first and
second equations are linearly dependent.

EXAMPLE

Consider the following set of simultaneous linear equations

x + 2y — z = 10

2x — 4y — 2z — 5

x + y 4- z = 6

The determinant of the coefficient matrix A is nonzero

1 2-1
2 -4 -2 = -4-2-4 -(4-2 + 4)= -16
1 1 1
and the rank of the matrix is 3; the rank of the augmented coefficient matrix is also 3. Thus
r(A) = r(A:y) = 3 = n and the set of equations has a unique solution.

PROBLEMS
Determine the rank of each of the following matrices.

T 2 " 1 -1 0
1.
0 1 4. -1 1 0
2 -2 0
1 2 3"
o -1 0 1 T 3
J.
0 4 8_ 2 6.

' 1 -1 O' ”—2 1 4

3. -1 2 3 6. 0 -1 5
o 1 2 _ .-2 0 -9
 7.9 Simultaneous Linear Equations / 589

3 -1 0 2'
7. 9. -2 1 5 -2
0 -4 6 -3
-3 5 -6 1
8. 2 -1 3 0
10. -3 3 4 2
4 -5 -11 -4
-1 2 7 2

Answers to Odd-Numbered Problems

1. rank 2 7. rank 3
3. rank 3 9. rank 3
5. rank 1

SOLUTION OF SIMULTANEOUS LINEAR EQUATIONS

There are various methods for obtaining the solution of a set of simultaneous linear
equations which have a unique solution. Several of these methods are discussed in
this section. If one of the methods is applied to a set of linear equations not having a
unique solution, the method cannot be completed, thus indicating that there is no
unique solution.
As noted above, a set of n simultaneous linear equations in n unknowns can be
written in matrix notation as

xAnxl = Cnx l

and the solution obtained by inversion of A can be denoted by

Alternatively, the standard procedure described previously can be used to change

the tableau

(A 111 c)

into the tableau

(I | A"1 | x)

from which the solution can be read directly.

A third matrix method for the solution of simultaneous linear equations is given
by Cramer's rule: The solution of

An xn i ==CnXl
590 / Matrix Algebra

can be obtained as the ratios of determinants

Cl a 12
••• al n

C2 a22 a2n

cn an2
••• ann
x, =
A|

an CT
... ain
c'2
... a2n
a21

Cn
••• ann
x? =
|A
l

a 12
... Cl
al 1

a21 a 22 C2

anl a n2 ••• Cn
=

Aj

ch Xf, i = i. 2 ^ • • ., n, the
matrix and the numerator is the determinant of the coefficient matrix with the ith
column replaced by the column of constant terms from the right-hand side of the
equations. If there is no unique solution for a set of linear equations, | A 0 and
these quotients are not defined.

EXAMPLE

Solve the set of simultaneous linear equations

3xx + x2 — x3 = 2

xv — 2.x 2 + x3 = —9

4xt -I- 3x2 + 2x3 = 1

Apply the standard procedure to the tableau

(A 111 c)

to obtain the tableau

(I I A -11 x)

Then the solution can be read directly from the tableau

 7.9 Simultaneous Linear Equations / 591

or obtained as

x = A- xc

"3 1 -1 1 0 0 2 '
1 -2 1 0 1 0 — 9
_4 3 2 0 0 1 1 _

"1 -2 1 0 1 0 _9 "
3 1 -1 1 0 0 2
4 3 2 0 0 1 1 .
"1 -2 1 0 1 0 _9 "
0 7 -4 1 -3 0 29
_0 11 -2 0 -■4 1 37 _
i i
"1 0 ~7 4 7 0 "
4 1 3 29
0 1 “7 7 —7 0 ~T
30 11 60
_o 0 ~T -T 4 1 T J
7
"1 0 0 TO i A -1
1 i
0 1 0 ~T5 “T T5 3
11 1
_0 0 1 — TO 0 TO -2

Thus, reading directly from the tableau,

'Xj " ■ -r
x2 3
_*3_ _ -2

or, using the inverse matrix, x = A 1c,

r 7
TO
1
0 TO ■ 2~ ' -r
1 1
*2
—
~TT T A -9 —
3
11 1
-*3- L —TO 0 TO . 1_ _ -2.

In practice, the solution would be obtained from x = A-1c only if A-1 had been obtained
without obtaining the tableau (I | A“11 x).
Using Cramer’s rule,

3 1 -1
1 -2 1 = —12 — 3 + 4 — (8 + 9 + 2)= -30
4 3 2

2 1 -1
-9 -2 1
1 3 2 — 8 + 27+1— (2 + 6 — 18) 30
= 1
-30 - 30 ~ - 30 ~

3 2 -1
1 -9 1
4 1 2 -54 - 1 + 8 - (36 + 3 + 4) -90
*2 =
-30 -30 _ — 30 _

3 1 2
1 -2 -9
4 3 1 -6 + 6 - 36 - (-16 - 81 + 1) 60
*3 =
-30 -30 ~ -30
592 / Matrix Algebra

EXAMPLE

Solve the set of simultaneous linear equations

Xl + x2 — *3=6

3xi — 4x2 + 2x3 = —2

2xx 4- 5x2 + x3 = 0

Apply the standard procedure to the tableau

(A|I|c)
to obtain the tableau
(I | A-11 x)
1 1 1 1 0 0 6
3 -4 2 0 1 0 -2
_ 2 5 1 0 0 1 o .

" 1 1 -1 1 0 0 6 “
0 -7 5 -3 1 0 -20
0 3 3 -2 0 1 —12 _
2 4 1 22
1 0 “7 7 7 0 ~T

0 5 3 l 20
1 “7 7 '7 0 ^T
36 23 3 144
0 0 ~T 7 1 T~
7 1 1
1 0 0 TS 6 TS 2
1 1 5
0 1 0 76 12 77 0
23 1 7
0 0 1 77 17 77 -4

thus, reading directly from the tableau, x = A lc

"*i " ' 2“

*2 = 0

_*3 . L —4

or, using the inverse matrix,

7 i
1

_1
On

~*r TS 7 TS ' 2"

1 1 5
*2
=
'77 17 77 -2 = 0
_i

—i

__i

23 1 7
o

■77
1

_*3_

Using Cramer’s rule,

1 1 -1
AI = 3-4 2 = -4 - 15 + 4 - (8 + 10 + 3) = -36
2 5 1
6 1 -1
2 4 2
0 5 1 -24 + 10 + 0 - (0 + 60 - 2) -72
Xl = = 2
-36 -36 -36

1 6
3 2
2 0 -2 + 0 + 24 - (4 + 0 + 18)
*2 =0
-36 ’ -36

1 1 6
3 4 -2
2 5 0 0 + 90 - 4 - (-48 - 10 + 0) 144
*3 = = -4
-36 -36 ^36
 7.10 Vector Differentiation / 593

PROBLEMS
For each of the following sets of simultaneous linear equations, determine whether there is a
unique solution and obtain the solution if there is one.

1. x 4 2y = 1 11. 4xj — 5x2 — 7x3 = 15

3x 4 4y = 2 3xj + 2x2 — 6x3 = 8
2. x + y — 2z = 5 Xi — 7x2 — x3 = 6
2x — 4y 4 3z = 6 12. 5Xi — 2x2 + x3 = 12
3x — 3y + z = 11 2xi 4- 2x2 — 3x3 = 7
3. Xi + x2 + x3 = 3 Xi — 6x2 4 7x3 = — 2
2xj — x2 — x3 = 0 13. 2xi 4 3x2 4 x3 = 0
3x, — 4x2 — x3 = 8 4xi — 8x2 — 6x3 = 2
4. 3x — y + 2z = —2 6xi 4 x2 — x3 = 0
x 4 y 4 z = 5 14. xj 4 5x2 — 4x3 = 0
2x — 2y 4 z = 3 3xi — x2 4 4x3 = —4
5. x — 5y 4 6z = 7 2xx 4 3x2 - 8x3 = 0
3x 4 3y — z = 8 15. 3xi — 3x2 4 4x3 = —18
2x 4 8y — 7z = 1 4Xi — 4x2 4 4x3 = — 24
6. Xi — 3x2 = —2 2x2 — 2x2 4 4x3 = 6
2xi 4 7x2 = 3 16. x^ 4 x2 4 6x3 = —2
7. Xi — 4x2 = — 1 3xi 4 2x2 4 x3 = 0
3xx — 2x2 — x3 = 0 2xi 4 x2 4 5x3 = 2
Xi 4 x3 = 3 17. Xi 4 2x2 = 0
8. xj 4 x2 -l- x3 = 9 x, 4 x2 4 x3 = 2
3xi 4 2x3 = 17 2Xi 4 2x2 4 3x3 = 7
x2 + x3 = 10 18. Xj 4 x2 4 x3 = 4
9. 3x — y 0 2xi 4 3x2 4 2x3 = 5
2x + 4 y = — 14 3xj 4 4x2 — 3x3 = — 3
10. 3xj — 4x2 = — 3
6xj + x2 = 3

Answers to Odd-Numbered Problems

1. x = 0, y = j 11. inconsistent, no solution
3. Xi = 1, x2 = 1, x3 = 1 13. Xi = — i, x2 = 1, x3 = —2
5. linearly dependent, no unique solution 15. inconsistent, no solution
7. Xi = 1, x2 = i, x3 = 2 17. Xi = —2, x2 = 1, x3 = 3
9. x = -1, y = -3

7.10 Vector Differentiation

In some maximization and minimization problems, expressions involving vectors
and matrices must be differentiated. The derivatives of simple expressions involving
vectors and matrices can be obtained directly from the definitions. The vector deriva¬
tives of linear functions, quadratic forms, and bilinear forms are defined and il¬
lustrated in this section. Second- and higher-order vector derivatives are not
discussed explicitly, but can be obtained by successive differentiation in the usual
manner.

VECTOR DIFFERENTIATION OF A LINEAR FUNCTION

An expression of the form a'x where a is n x 1 and x is n x 1 is a linear function. For
the linear function
594 / Matrix Algebra

xl
X2
a'x = [al9 a2 , an\

x„

= aixl + a2x2 + ••• + a„x

n n

the partial derivatives of a'x with respect to the scalar xi9 i = 1,2, .., n, are

d(a'x)
dxx
a,
d(*'x)
= <3-
dx->

d(a'x)

that is, the partial derivatives are the elements of the vector a. Thus if the n partial
derivatives are arranged as a vector a, the process of vector differentiation is defined
by
d(a'x)
——- = a
^x

a(a'x).
where indicates the operation of differentiating a'x with respect to the elements
dx
of the vector x

EXAMPLE

a'x = 2 axx — ax2 + 3 ax3 + ax4

and

(a'x) = 2 a
dx{

_a_
(a'x) = —a
dx2

(a'x) = 3 a
dx3

_d_
(a'x) = a
dx4

Thus
2a
—a
(a'x) = = a
dx 3a
a
 7.10 Vector Differentiation / 595

Vector Differentiation of a Vector of Functions

If y denotes an rc-dimensional column vector each element of which is a function of
the m elements of x, that is, if

yi fi (* i» *2 » • • •» ) i 1, 2, ..., n

then each yt can be differentiated partially with respect to each Xj and these partial
can be arranged in an

5>’i Sy2 Sy„

dxt dx± dxx

Syi 8y2 Sy„

dx dx2 dx2 dx2
. •
>•>

sy2 8yn
i-H

dxm
Tn Sxm 8x„

EXAMPLE

If

x\ + 2x2 — 3xi x3
3xixj - x\ + 4x2x]

then

2xj — 3x3 3x2

dy_
4x2 6xj x2 — 2x2 + 4x3
dx
— 3xi 8x2x3

VECTOR DIFFERENTIATION OF A QUADRATIC FORM

An expression of the form x'Ax, where A is an n x n symmetric matrix, is a quadratic
form. The quadratic form x'Ax can be expanded as follows

ai i 012 **• 01n" "*1 "

012 022 ••• 02« *2

x'Ax = [x1? x2 , ..., xj

01n 02 n 0/i«

= atlXi + 2 a12xlx2 + 2a13x1x3 + ••• + 2alnx1xn

+ 022*2 + 2 a23x2x3 + ••• + 2 a2nx2n

_|_ • • ♦ •

+ a„nx2n
596 / Matrix Algebra

Taking partial derivatives with respect to the elements of x,

i (x'Ax) = 2(a11x1 4- a12x2 + al3x3 4- ••• + alnx„)

^ (x'Ax)
~ 2(Ui2Xi T U22X2 T U23X3 4" T ^2ti-^n)

— 2-I- fl2„x2 T ^3^X3 4- T cinnxnj

^„(x'Ax)
Apart from the factor 2, the right-hand side of this set of equations contains the
elements of Ax, which is an rc-dimensional column vector; alternatively, the right-
hand side contains the elements of x'A, which is an n-dimensional row vector. Thus

rs
— (x'Ax) = 2Ax
dxx '

or

~ (x'Ax) = 2x'A
dxx '

In practice, the choice between these two forms usually depends on the context in
which differentiation occurs, since matrices can be equated only if they are the same
order and, specifically, a row vector cannot be set equal to a column vector.

EXAMPLE

"Xi“ ‘ 3 1 -2
If x = x2 and A = 1 0 3 then
_* *3_ -2 3 2

‘ 3 1 -2" "xi"
* = [xl5 x2 , x3] 1 0 3 *2

-2 3 2 _ -*3_
Xi
= [3*! 4- x2 — 2x3 , xl 4- 3x3 , — 2xt + 3x2 4- 2x3] x2
*3

= 3xj 4- xxx2 — 2xxx3 + XjX2 4- 3x2x3 — 2xtx3 4- 3x2x3 4- 2x3

= 3x? 4- 2x3 + 2x^2 — 4xjX3 -I- 6x2x3

and

— (x'Ax) = 6x, 4- 2x2 — 4x3

8x1

— (x'Ax) = 2xi 4- 6x3

cx2

(x'Ax) = 4x3 — 4x! 4- 6x2

dx-
 7.10 Vector Differentiation / 597

Thus
3xi + x2 — 2x3
_d_
(x'Ax) = 2 x, + 3x3 2Ax
dx
— 2xj + 3x2 + 2x3

or

(x'Ax) = 2[3xj + x2 — 2x3 , x, + 3x3, — 2x} + 3x2 + 2x3] = 2x'A

dx

VECTOR DIFFERENTIATION OF A BILINEAR FORM

An expression of the form x'Bz, where x' is 1 x m, B is m x n and z is n x 1, is a
bilinear form. The following derivatives for a bilinear form can be verified using the
procedures above.
ps
— (x'Bz) = Bz
ax' '

~ (x'Bz) = B'z
dzK '

EXAMPLE

If
"xr ~ 2b, b i + b2
X — x2 B= b2 + 2b3 3b2 ' z=
-•^3 -
_ b, + b3 4b32 _
then
b, T b2
^z,1
x Bz — [*1> x2, x3] b2 + 2 b3 3b2 ‘ =
Z2_
_ b, + b3 4 b\ _
Zl
— [2b,x, -I- b2x2 + 2b3x2 + />1x3 + b3x3, blxl + b2x, + 3 b2x2 + 4fr3x3]
z2
= 2 b,x,z, + b2 x2 z i + 2 b3x2z, + b,x3z, + b3x3z, + b,x,z2
+ b2 x i z2 + 3 b2x2z2 + 4 b\x3z2
arid

a— (x'Bz) = 2b1z1 + b,z2 + b2z2

OX j

-— (x'Bz) = b2z, + 2 b3z, + 3 b2z2

OX 2

—— (x'Bz) = b,z, + b3z, + 4 b\z2

cx3

Thus
2 b,z, + b,z2 + b2z2
b2z, + 2b3z, + 3b2z2 = Bz
b,z, + b3z, + 4fr3z2
598 / Matrix Algebra

and

(x'Bz) — 2bxxx + b2x2 T 2/73X2 4~ biX3 4- b3X3

dzx

(x'Bz) = bxxx 4- b2xl 4- 3b2x2 4- 4/73X3

dz2

Thus

d 2h i X i + b2 X2 + 2/>3 X2 + /7jX3 + /73X3

(x'Bz) =
dz />1*1 + Z?2 X1 + 3/72X2 + 4/7 3 x3

~xf
2/?! /? 2 + 2/73 /?! + Z?3
x2
bx 4 h2 3b2 4/?|
-*3_

= B'x

APPLICATIONS OF VECTOR DIFFERENTIATION

IN MAXIMIZATION AND MINIMIZATION
Vector differentiation is used in many areas of applications for problems of maximi¬
zation and minimization. In particular, the derivations of most of the estimators in
multivariate analysis involve vector differentiation. As an example of these applica¬
tions, the derivation of multivariate regressions estimators is discussed in this section.
In multivariate regression analysis, the model involves the assumption of a linear
relationship between a variable y and k explanatory variables xl9 x2, xk and a
disturbance term u. For a sample of n observations on y and the x’s,

4; = A) + PiXu + P2xn + "* + Pkxki + ui 1= 1> 2, ..., n

and the n equations can be written in matrix notation as
y = xp + u

where

’ yi' " 1 *21 xkl

1
y.2 X =
X22 Xk2

. _
i X2n Xkn

'Pi' ux

Pi u = U2

Jk_
The intercept p0 is represented by the column of ones in the x matrix. As a result of
the convention of denoting the ith observation on the kth variable by xki, the sub¬
scripts in the x matrix are the reverse of the usual pattern, which indicates the row by
the first subscript and the column by the second subscript.
The problem is to obtain estimates of the p's. These estimates are customarily
obtained using the least squares principle. Let
 7.10 Vector Differentiation / 599

denote a vector of estimates of p. Then y = xj} + e where e denotes the column

vector of n residuals (y — xp). The least squares principle minimizes the sum of
squared residuals
n

i= 1

= (y - xP) (y - xp)
= y'y - 2p'x'y + p'x'xp

Note that (y — xp)' = y' — P'x' and p x'y is a scalar and thus equal to its transpose
yxp. To obtain the estimate p which minimizes the sum of squared residuals, differ¬
entiate e'e with respect to p,

— 2x'y + x'xp + (p'x'x)'

— 2x'y + 2x'xp

If | ^ = °-

x'xp = x'y

and

p = (x'x) Vy

Alternatively,

4 (e'e) = - 2y'x + 2p'x'x

op

If jp {e'e) = °’
p x X = y'x
p'^y'xtx'x]1

and

p = [x x] ' x'y

is the vector of least squares estimates for the regression coefficients.

note: To show that p = [x'x]~ *x'y is a minimizing solution, the second deriva¬
tive must be examined

^ 2x'y + 2xxl*J
= 2x'x

Thus the second order condition for minimization requires that x'x is positive
definite, that is, that y'[x'x]y' > 0 for all y > 0. In practical examples, x'x is positive
definite.
600 / Matrix Algebra

EXAMPLE

"1 -1" 5 "

1 0 3
1 1 y= -2
1 —2 _ 8 _
find the least squares estimate p for the regression y = xp + u.

1 -1
1 1 1 1 1 0
x'x 1 1
-1 0 1 -2
1 -2

4 -2
-2 6

x'x = 24 - 4 = 20
3 1
ITT TO
[x'x]1 = 1
TO

and

3 1 1 1 1
TO TO
p = [x'xJ-Vy = 1 1
TO 5 0 1 -2

19
TO
16
"TJ

Thus the estimated regression is y = ytj — ^x.

PROBLEMS
" 3 ‘ ' Xi "
1. If b = -2 and x = X2 find — [b'x].
_ 1 - - *3_

—C yi
2. If c
0
and y =
y2 find
2c y3
— 3c _

■
c2 4- 1 Xi

3. Ifc and find -f- (c'x).

d\{ }

2x\ — 3x2x3
4. If y = Xi — 4*! X2 X3 find
dx'
xj — 2x2 + x3 _

x\ + 3x2
2xj x2 — x2
5. If y find
2xj + XjX2 — 3x2 dx ’
3xj — x\x2 — x2
 7.10 Vector Differentiation / 601

3 + bi b! + 2b2 4b2
6. If x = B =
*2 b2 4b i bxb2

yr
and y= yi find — [x'By] and — [x'By].
ox dy
ya
~*r " 3 0 -1"
7. If x = x2 A = 2 1 4
_*3_ _-l 0 3_

yr
and y = y2 find — [x'Ay] and — [x'Ay].
ox dy
y3
"1 -r ’-3“
8. If x = 1 i and y= 5
.1 2. . 11.
find the least squares estimate p for the regression y = xp + u.

Answers to Odd-Numbered Problems

c2 + 1
3c
4c - 5

2xj 2x2 2 + x2 6x1 — 2xx x2

5.
3 2xx — 1 Xi — 6x2 —Xi — 3x2

3yi - y3
7. 2yi + y2 + 4y3
- -yi + 3y3 _
Applications
of Matrix
Algebra

8.1 Introduction
There are many applications of matrix algebra in various areas. The applications
discussed in this chapter are maxima and minima of functions of n variables, input-
output analysis, linear programming, game theory, and Markov processes.

8.2 Maxima and Minima of Functions of n Variables

The necessary and sufficient conditions for a point to be a maximum or minimum of
a function of two variables are discussed in previous sections. These results are a
special case (n — 2) of the conditions for a maximum or minimum of a function of n
variables; the general results are given below in matrix notation. Unconstrained
maximization and minimization and constrained maximization and minimization
using Lagrange multipliers and the Kuhn-Tucker conditions are discussed and
illustrated.

UNCONSTRAINED MAXIMA AND MINIMA

Consider a function of n variables f(xl, x2, ..., xn) at a point x* = (x*, xf
such that the n partial derivatives are zero:

d£ df_ df
= 0 = 0, = 0
dxx dx. dx„

Define the determinant An of second-order partial derivatives

d2f d2f 82f

dx\ dxi dx2 dxx dx

s2f 82f 82f

dx2 dxl dxl dx2 dxt

82f 82f 82f

dxn dxx dxn dx2 8x2n

sometimes referred to as a Hessian determinant.

 8.2 Maxima and Minima of Functions of n Variables / 603

The principal minors of A„ are

37
A, =
dx\

37 37
dxj dxl dx2
A, =
82f 37
dx2 dx1 dxl

37 37 37
dxl dx± dx2 dx1 dx

A, =
37 37 37
dx2 dxx dx\ dx2 dx

37 37 37
dx3 dxx dx3 dx2 dxl

A„ =
The stationary point x* = (xf, xf, ..., x*) is a

Local maximum if Ai < 0, A2 > 0, A3 < 0, .

Local minimum if Ax > 0, A2 > 0, A3 > 0, ...

If neither of these conditions holds, the function must be examined in the region of
the stationary point. Note that the conditions given previously for maxima and
minima of a function of two variables are a special case of those given above for n
variables.

EXAMPLE

Determine the maxima or minima (if any) of the function

f(xi, x2 , x3) = xj + 2x2 + x3 + xxx2 — 2x3 — lx1 + 12A„

~~ = 2xi + x2 - 7
CX 1

~ = 4x2 + xt
0X2

If 7 = 0, then
CX3

2x1 + x2 = 7

Xi + 4x2 = 0

2x3 = 2
604 / Applications of Matrix Algebra

Thus

x2 — — 1

X! = 4

x3 = 1

d2f a2/
= 2 = 1 = 0
dx2 dxi dx2 dx3

W = 4 d2f
= 0
dxj dx2 dx3

dV = 2
dxj

Aj = 2
2 1
A2 — 1 =8-1=7
4

2 1 0
A3 — 1 4 0 = 16+ 0 + 0-(0 + 0 + 2) = 14
0 0 2

Ai > 0, A2 > 0, A3 > 0, so the point (4, — 1, 1) is a local minimum of the function f(xu x2,
x3) = x2 + 2xj + x3 + xi x2 — 2x3 — 7xi + 12.

EXAMPLE

Determine the maxima and minima (if any) of the function

f(x, y, Z) = e-x2-,2-22+2y + xz

df
= ( —2x + z)e~x2~y2~z2 + 2y+xz
dx

df
= (-2y + 2)e~x2~y2~z2 + 2y+xz
dy

df
-f- = (-2 z + X)e-x2-y2~z2 + 2y+xz
oz

If ~ = 0, ^ = 0, f = 0, then
dx dy dz

— 2x + z = 0

2y = 2

x — 2z = 0

Thus

x = z =0

y = i
d2f
= [( —2x + z)2 — 2]e — x2 — y2 — z2 + 2y + xz = -2c
dx2 dx2 (0, 1.0)

d2f d2f
= ( — 2x + z)( — 2y + 2)c -x2— y2 — z2 + 2y + xz = 0
dx dy dx dy (0. 1, 0)
 8.2 Maxima and Minima of Functions of n Variables / 605

d2f
= [(lx + z)( — 2z + x) + 1 ]e-*2-y2-*2+2y+xz = c
dx dz dx dz (0, 1, 0)

d^f = 2 _ 9]x2-y2-z2 +2y +xz

[(-2y + 2)2-2]e = -2e
dy2 dy2 (0.1, 0)

d2f - x2 - y2 — z2 + 2y + xz d2f
= ( — 2y + 2)( — 2z + x)e = 0
dy dz dy dz (0, 1. 0)

d2f
= [( —2z + x)2 - 2]e-x2~y2-z2+2y+- = -2e
dz2 dz2 (0, 1.0)

Aj = -2c

-2e 0
A2 = = 4c
0 — 2e

-2c 0 c
a3 = 0 — 2c 0 = -Se3 + 2e3 = -6c3
e 0 — 2c
Ai < 0, A2 >0, A 3 < 0, SO the point (0, 1, 0)
f(x, y, z) = e~x2~yl — z2 + 2y + xz

EXAMPLE

Total profit is equal to total revenue minus total cost. In a perfectly competitive market, total
revenue is equal to the product of the number of units sold and the fixed unit price; thus

P = pq-C

where P is profit, p is price per unit, q is quantity sold, and C is total cost. If the production
function is

q =/(xl5 x2)

where and x2 are the variable inputs, and the cost function is

C = p^Xi + p2x2 + b

where px and p2 are the unit prices of xx and x2, respectively, and b is the cost of the fixed
inputs, then

P = Pf(xi, x2) - piXi - p2x2 - b

Profit is thus a function of x1 and x2 and is maximized with respect to these variables as
follows.

dP
^ = pfl-p' = 0

g = P/2-P2 = °
or

Pfi = Pi

Pf2 = Pi

Since fi and f2 are the marginal products of the two inputs, pf{ and pf2 are the values of these
marginal products, that is, the rates at which revenue would increase with increase of the
respective inputs. Thus the first-order conditions for profit maximization require that each
input be increased until the value of its marginal product equals its price. Note that the
606 / Applications of Matrix Algebra

maximum profit-input combination lies on the expansion path since the conditions pfi = P\
and pf2 = p2 are a special case of the condition jf2 = p 1 /pi.
Second-order conditions require that the principal minors of the Hessian determinant

d* 1 2 3p d2P
dxj dx i dx

d2P d2P
dx2 dx{ dx2

alternate in sign. That is, for a maximum it is required that

^2 D

2 <°
dxj

d2P d2P
dxj dxi dx
>0
d2P d2P
dx2 dxi dx 2

Expanding the determinant, the second condition is

(d2P\ id2P\ d2P

>0
dxj ,dx2 \dxidx-

d2P d2P
Since —■ < 0 (first condition) and >0,
dxj [dx1dx:

d2P
<0
dx7

The second-order partials can be evaluated as follows:

d2P
= Pf n < 0
dxj

d2P
= Pfl2 < 0
dx2

and, since p > 0, the second-order conditions for maximization require that

/u <0

f22 < 0

Thus the second-order conditions for maximization require that both marginal products be
decreasing. This is reasonable, since the price of a variable is constant under free competition
and thus, if the marginal product of a variable were increasing, profit could be increased by
increasing the quantity of the variable.

PROBLEMS
1. Determine the values of xl, x2, and x3 (if any) that maximize or minimize the function

f(xi, x2 , x3) = Xi x2 + 10*! — xj — x\ — x3.

2. Determine the values of x, y, and z (if any) that maximize or minimize the function
f(x, y, z) = £4*2 + 2>>2 + z2-5xy-AZ'

3. Determine the values of x, y, and z (if any) that maximize or minimize the function
/(x, y, z) = g-*2-2>*-z2-2*y
 8.2 Maxima and Minima of Functions of n Variables / 607

4. Determine the values of x1? x2, and x3 (if any) that maximize or minimize the function

f(xu X2 , X3) = x\ + x\ + 7x3 - XjX2 •

Answers to Odd-Numbered Problems

1. Xi = x2 = -x, x3 = 0 (maximize)
3. x = v = z = 0 (maximize)

LAGRANGE MULTIPLIERS
Consider a function of n variables f(xu x2,..., xj subject to the constraint g(xu x2,
x„) = 0at a point x* = (xf, x*, ..., x*) which satisfies the n + 1 equations

Sf x dg
= 0
dxx dxx

--- X — = 0
dx2 dx2

df dg
- 2. C = 0
oxn oxn

dX

Note that for x* = (x?, X?, ..., x*), X = (fXi/gXi), i = 1, 2, ..., n. Define the
determinant

0 9xi 9x2 • •
9x„
9xi fx 1*1 X9X \ X1 fx 1X2 X9x 1X2 fxiXn ~ XgxlXn
^n+ 1 — 9x2 fx 2X1 X9x 2x1 fx 2X2 — Xgx 2X2 fx 2Xn — XgX2xn

9xn fxnXl X9xnx 1 fx„X 2 X9xnx2 fx „xn X9x„x„

sometimes referred to as a bordered Hessian determinant. In order to determine

whether x* = (xf, xf,..., xj) is a maximum or minimum, the n — 1 principal minors
of An+ j: A3, A4,..., A„+15 must be evaluated for x* = (x*, x*,..., xj). Note that A,
consists of the first i rows and i columns of A„+ The point x* = (xj, xf,..., x*) is a

Local (constrained) maximum if A3 > 0, A4 < 0, A 5 > 0, ...

Local (constrained) minimum if A3 < 0, A4 < 0, A5 < 0, ...

If neither of these conditions holds, the function must be examined in the region of
the stationary point. Note that the conditions given previously for constrained
maxima and minima of a function of two variables are a special case of those given
above for n variables.

EXAMPLE

Determine the minimum of the function

608 / Applications of Matrix Algebra

f(Xu *2 , X3) = XXX2 if xi x2x3 = 125.

F(xu x2, x3, X) = xxx2 — X(xx x2x3 — 125)

dF
= x2 + x3 — Ax2x3
dxx

dF
— Xi + X3 — AxjX3
dx2

5F
— Xi + x2 — Xxxx2
dx3

dF_
= — (x!x2x3 — 125)
Jx

dF dF dF
If r = 0, *— = 0, — = 0, — = 0, then
UX\ OX2 DX3 dA

Xi = x2 = x3 = ^/l25 = 5 X = \

fx 1*1 = 0 fx2x2 = 0 fx 3x3 = 0

fX1 X2 = 1
fx 1,3 = 1 fx 2x3 = 1

9xi = x2x3= 25 9x 2 = ^x3 = 25 9x3 = XXX2 =

9x\X\ = 0 9x2X2 = 0 9x3X3 = 0

= xx = 5
X

x2 5
II
II
u>

9x\X2 9xix3 9x2X3

0 25 25
A, = 25 0 -1 = 0 - 625 - 625 - (0 + 0 + 0) = - 1250
25 -1 0

0 25 25 25
25 0 -1 -1
Aa = 25 -1 0 -1
25 -1 -1 0

25 -1 -1 25 0 -1 25 0 -1
0-25 25 0 -1 + 25 25 -1 -1 -25 25 -1 0
25 -1 0 25 -1 0 25 -1 -1

= 0 - 25[0 + 25 + 25 - (0 + 25 + 0)] + 25[0 + 25 + 0 - (25 + 25 + 0)]

- 25[25 + 25 + 0 - (25 + 0 + 0)]

= 0 - 625 - 625 - 625 = — 1875

A3 < 0, A4 < 0, so the point (5, 5, 5) is a local minimum of the function f(xu x2, x3) = Xi x2 +
X! x3 + x2 x3 subject to xx x2 x3 = 125.

EXAMPLE

Determine the maximum of the function

 8.2 Maxima and Minima of Functions of n Variables / 609

/(x, y, z) = — x2 — 2y2 — z2 + xy + z if x + y + z = 35.

F(x, y, z, X) = —x2 — 2j;2 — z2 + xy 4- z — 2.(x + y + z — 35)

dF
— 2x + y — X
dx

dF
— 4y + x — X
dy

dF_
— 2z+ 1 — A
dz

M = -(x + y + z-35)

^ 17
If^-0 ^ = 0,
0 d-L = 0, — = 0, then
dx <7y <5z 0/1

— 2x + y = k

— 4y + x = X

— 2z + 1 = X

From the first and second equations,

3x = 5 y

x = iy

From the second and third equations,

-2z+ 1 = y-1ry= ~iy

1 + iy

l + iy
■fy + y + 2
35

+ 2y + 1 + iy = 70

¥>’ = 69
y = 9

x = 15

z = 11

A = -21

-2 = -2
II
1

fxx = fzz

fXy= 1 fxz =0 fyz= 0

gx = i Qy = 1 = 1
O

o 0
II

gxx = Gyy =
N
N

Qxy =0 9XZ 0 9yz ^

0 1 1
A3 1 -2 1 = (0 + 1 + 1) - (-2 + 0 - 4) = 8
1 1 -4
610 / Applications of Matrix Algebra

0 1 1 1
A* 1 -2 1 1
1 1 -4 0
1 1 0 -2

1 1 1 1 -2 1 1 -2 1
0 - 1 -4 0 + 1 1 0 — 1 1 -4
1 0 -2 1 1 -2 1 1 0

= 0 - [8 + 0 + 0 - (-4 + 0 - 2)] + [-2 + 1 + 0 - (1 + 0 + 4)]

-[0+ 1 + 8 - (1 - 4 + 0)]

= 0-14-6-11= -31

A3 > 0, A4 < 0, so the point (15, 9, 11) is a local maximum of the function f(x, y, z) = — x2
— 2y2 — z2 + xy + z subject to x + y + z = 35.

EXAMPLE

If utility analysis is generalized to the case of n commodities, the utility function is given by

U =f{qi, q2 , •••> qn)

where qu q2,..., qn are the quantities of the n commodities, and the budget constraint is given
by

y - ILPiVi= 0
i=i

where pu p2, p„ are the unit prices of the n commodities. Then the function to be
maximized is
\
n

f = f(q^qi, • • •, qn) - M y - Z Pi <?,

i= 1

and, for maximization,

a- = fi + tyi = 0 i = 1, 2, ..., n
cqi
n
dF
y- LPiQi = 0
~dX i= 1

The demand curves for the n commodities can be obtained by solving for the q s. The
conditions

dF
=f + = o i = 1, 2, ..., n
dq,

can be written as
 8.2 Maxima and Minima of Functions of n Variables / 611

That is, the rate of substitution of commodity i for commodity j must equal the price ratio
Pj/Pi. The second-order conditions for maximization require that the principal minors of the
bordered Hessian determinant must alternate in sign; that is,

0 -Pi Pi 0 -p 1 -Pi -Pi

-p 1 fll fll >0 -Pi fl 1 fll fll

_ -P2 /21 fll -Pi fll fll fll

P3 fll fll fll

0 -Pi ~Pl -Pn

-Pi fl 1 .111 fin

-Pi fll fll fin > 0

-Pn fnl fnl ••‘ fnn

KUHN-TUCKER CONDITIONS
Consider a function of n variables/(x1? x2, ..., xn) subject to the constraint g(xu x2,
..., x„) < 0. A point x* = (x*, x2, ..., x*) is a local maximum of/(xl5 x2, ..., x„)
subject to g(x1, x2, ..., x„) only if there exists a nonnegative X such that X and (x*, x2,
.x*) satisfy the Kuhn-Tucker conditions:

, _ df , dg
X~ = 0 i = 1, 2, ..., n
h‘ dXi dxt

Xg(xu x2, . x„) = 0

g(xx, x2 , . • •, x„) < 0

These conditions are also sufficient if/(x1} x2,..., x„) is concave and the constraint is
concave. Since a maximum point of/(xl9 x2, ..., x„) is a minimum point of —f{xl,
x2,..., x„), this result is also applicable when a convex function is minimized subject
to a convex constraint.

note: A function f(xl9 x2, ..., x„) is convex in a region if for any two points (xl5
x2, ..., x„) and (xl9 x2, ..., x„),

/[(l - t)x1 + txu ..., (1 ~ t)xn + txn]

< (1 - t)f(xu x2 , ..., x„) 4- tf(xu x2 , ..., x„)

The function is strictly convex if < can be replaced by < ; the function is concave if
< can be replaced by > and strictly concave if < can be replaced by >.
The method of Lagrange multipliers can be modified to determine the maximum
or minimum of a function of n variables subject to one inequality constraint in a
manner similar to the modification for a function of two variables subject to one
inequality constraint. Assume that the inequality constraint holds as an equality
constraint and obtain the maximum (or minimum) using the method of Lagrange
multipliers. Then, if X > 0, this maximum (or minimum) is also the maximum (or
minimum) subject to the inequality constraint; if^ < 0, the maximum (or minimum)
determined without regard to the constraint satisfies the constraint and is thus also
the constrained maximum (or minimum).

EXAMPLE

Find the minimum of f(x lf x2, x3) = X! x2 + Xi x3 + x2 x3 subject to the constraint

Xi x2 x3 > 125.
612 / Applications of Matrix Algebra

Assuming the inequality constraint to hold as an equality constraint, X = f (see page 608).
Since X > 0, the minimum (5, 5, 5) assuming the equality constraint is also the minimum
assuming the inequality constraint.
Alternatively, using the Kuhn-Tucker conditions,
df_
— x2 + x3 — Xx2 x3 = 0
dxi dxi

df ;t dQ = Xi + X3 — Xx! x3 = 0
dx2 dx2

x dg
= x{ + x2 — Xx! x2 = 0
dx2 dx2
A(x! x2x3 — 125) = 0
Xix2x3 > 125

Either X — 0 or X! x2 x3 — 125 = 0. If X = 0,
x2 + x3 = 0

Xi + x3 = 0

xi -I- x2 — 0
and Xi = x2 = x3 = 0; but then xx x2 x3 > 125 is not satisfied. If xx x2 x3 — 125 = 0,
x2 + x3 _Xi + x3 _Xj + x2
X2X3 X'i X3 Xj X2

and xx = x2 = x3 = 5; this solution satisfies the Kuhn-Tucker conditions. Thus, as above, the
minimum of /(xl5 x2, x3) = XjX2 + xtx3 + x2x3 subject to XjX3x3>25 is
Xi = x2 = x3 = 5.

EXAMPLE

Find the maximum of f(x, y, z) = — x2 — 2y2 - z2 + xy + z subject to the constraint

x + y + z < 35.
Assuming the inequality constraint to hold as an equality constraint, X = — 21 (see page
609). Since X < 0, the unconstrained maximum is also the maximum subject to the inequality
constraint.

8f d2f = d2f
2 = 1
Xx=~2x + y dx2 dx dy
CO

d2f
'Tt

f=-4y + X = 0
II
1

dy dx dz

5f , _L1 d2f
— — — 2z -f 1 = 2 = 0
cz 5z2 dy dz

lf^ = 0 £ = 0 d± = 0, then
8x ' dy ’ dz

2x — y = 0

x — 4y = 0

and thus

x = y = 0

2 =j

A! = -2
 8.2 Maxima and Minima of Functions of n Variables / 613

2 1
A2 — =8-1=7
1 -4

-2 1 0
A3 = 1 -4 0 = -16 - (-2) = -14
0 0 ■2

Ai < 0, A2 > 0, A3 < 0 so the point (0, 0, J) is a maximum.

Alternatively, using the Kuhn-Tucker conditions,

8f_ _ xdl _
2x 4 y — A = 0
dx dx

df-Xd^=-Ay + X-X=0
cy oy

dj_ _ , dg -2z 4 1 - A = 0
dz dz

A(x 4 y 4 z — 35) = 0 x 4 y 4 z < 35

Either 2 = 0 or x + y + z — 35 = 0. If X = 0,

x = v = 0 z =j

and x + y + z < 35 is satisfied. /(0, 0, i) = i- If x + y + z = 35, solving the first three

equations,

— 2x 4- y — A = 0

x — 4_y — 2 = 0

2z 4 X = 1

From the first two equations,

3x — 5y = 0

and from the second two equations,

x — 4y 4 2z = 1

and, since x 4 y 4 z = 35,

x 4 6y = 69

3x — 5y = 0

23y = 207

y = 9

x = 15

z = 11

The solution x = 15, y = 9, z = 11 satisfies — 2x4y — 2 = 0, x — Ay — 2 = 0 and 2z 4 A = 1

if A = — 21./(1, 9, 11) = —362, which is less than/(0, 0, i). Thus, as above, the maximum of
/(x, y, z) = —x2 — 2 y2 — z2 4 xy 4 z subject to the constraint x4y4z<35isx = y = 0,

PROBLEMS
1. Determine the values of x1? x2, and x3 that minimize the function

f(xi, x2 , x3) = x\ 4 4x2 4 x3 — 4x! x2 - 6x3

subject to the constraint xx 4 x2 4 x3 = 15.

614 / Applications of Matrix Algebra

2. Determine the values of x1? x2, and x3 that maximize the function

f(xi, x2 , x3) = 6*! x2 4- 4x2x3 + 6x2 — 3x3 — x2

subject to the constraint xx 4- 2x2 + x3 = 75.

3. Determine the values of xu x2, and x3 that minimize the function

f(xu X2 , X3) = x\ + x\ + x\ + X!X2X3

subject to the constraint Xi x2x3 = 1000.

4. Determine the values of xl5 x2, and x3 that minimize the function

f(xu x2 , x3) = 5x? 4- 10x2 + x3 - 4x*x2 - 2xjX3 - 36x2

subject to the constraint xx 4- x2 4- x3 = 3.

5. Determine the values of xl5 x2, and x3 that minimize the function

f(xi, x2 , x3) = x? 4- 4x! 4- x§ - 4x! x2 - 6x3

subject to the constraint *i 4- x2 4 x3 > 15 (see Problem 1).

6. Determine the values of xx, x2, and x3 that maximize the function

f (x!, x2 , x3) = 6x! x2 4- 4x2x3 4 6x2 - 3x] - x\

subject to the constraint xx 4- 2x2 4- x3 < 75 (see Problem 2).

7. Determine the values of xl9 x2, and x3 that minimize the function

f(xu X2 , X3) = Xl + Xj + X\ + Xj x2x3

subject to the constraint x: x2x3 < 1000 (see Problem 3).

8. Determine the values of xl9 x2, and x3 that minimize the function

/(xl9 x2 , x3) = 5x1 4- 10x2 4 x3 — 4xt x2 — 2xt x3 — 36x2

subject to the constraint Xj 4 x2 I x3 > 3 (see Problem 4).

Answers to Odd-Numbered Problems

1. Xi = 8, x2 = 4, x3 = 3 5. x{ = 8, x2 = 4, x3 = 3
3. X! = x2 = x3 = 10 7. no finite maximum

8.3 Input-Output Analysis

Input-output analysis traces the flow of production in order to study the effects of a
change in the demand for final foods on production of primary, intermediate, and
finished goods. The primary purpose of input-output analysis is to calculate the
output levels in various industries that would be required by particular levels of final
goods demand. Input-output analysis can also be applied in studying sectors of the
economy, either using a closed model where the output for a sector is equal to
the sum of its inputs in the other sectors or using an open model which includes final
demand in addition to demands from other sectors.
Assume that an economy is divided into n industries and that each industry
produces only one type of output. Industries are usually related in the sense that each
must use some of the others’ products in order to operate. In addition, an economy
must usually produce some finished products for final demand. Input-output analysis
determines the production of each of the industries if final demand changes, assum¬
ing the structure of the economy does not change.
It is convenient to tabulate the data for input-output analysis as shown in Table
8.1, where bi} is the dollar amount of the products of industry i used by industry j, h,: is
 8.3 Input-Output Analysis / 615

Table 8.1

User
Final Total
Producer 1 2 ••• n demand output

1 b 12 •• b i„ by Xi
2 ^2 1 bn bjn h2 *2

n b»i bn 2 ■ •• bnn bn xn

the final demand for the products of industry i, and xt = ]T"= j + ht is the total
output of industry i.
The structure of the economy can be described by the technological matrix

A = (fly)

where = (b^/xj) = dollar value of the output of industry i that industry j must
purchase to produce one dollar’s worth of its own products. Note that this definition
of a{j assumes that the purchases of intermediate product of an industry are propor¬
tional to the level of output of the industry. This assumption of constant proportion¬
ality between inputs and outputs is standard for input-output analysis.
The ith industry must produce outputs

ailx1 + ai2x2 + + ainxn for i = 1, 2, ..., n

in order to supply the needs of all industries. The interindustry demand vector X can
thus be written AX, where

a\i al2 "• «ln" "xf

a21 a22 «2n x2

A= and X =

an 1 an2 ’ • • ann_ Xn

the production of the economy must be adjusted to fulfill both interindustry needs
and final demand. If the final demand vector is

K
h2
H = ht> 0 i = 1, 2, ..., n
h„

this requirement can be written

X = AX + H
Thus

[I - A]X - H

and

X = [I — A] XH

the matrix I — A is known as the Leontief matrix.

There are a number of mathematical and practical problems in input-output
analysis. For example, the technological matrix must be such that each of the x-s is
616 / Applications of Matrix Algebra

nonnegative; otherwise, the solution is not economically meaningful. There are also
problems concerning industry classification and stability of the technological matrix.
Given a set of positive final demands, consider the problem of determining the
conditions under which there is a unique set of positive production levels consistent
with the set of final demands. First, since [I-A]'1 must exist, 11 — A | must be
nonzero. In addition, each element of [I — A] 1 must be nonnegative, since other¬
wise an increase in final demand would result in a decrease in production at some
point in the production process. It can be shown that, in order for positive gross-
output levels to be associated with any given set of positive demands, the following
conditions are required

1 > >0 all i and j

I — A| >0
Since any /c-industry subset of the n industries must be able to satisfy interindustry
demands with some surplus to satisfy demands external to the k industries, all
principal minors of the n industry Leontief determinant

— a1! — al2 -01 n

~ a2l 1 — d22 ~ a2n
I - A
— ani ~ an2 1
must be positive.

EXAMPLE

Consider a very simple hypothetical economy of two industries, A and B, represented in Table
8.2, where data are in millions of dollars of products.
Determine the output vector of the economy if final demand changes to 200 for A and 100 for
B.
i 7 '
A =
i T5
8 9
15 TO -

i 7
T ~T0
I - A = 8 11
"15 T0\

11 7
200 TO TO
[I-A]"1 = 17" 8 1
15 1
• ll A I 1.11 7.8 27
NOTE. | ! - A | —-j T0~ TO T5 ~ TTT0-
As must be the case,
1 1 1000
1[15011
X = [I-A]-‘H = ^
TO ■re] -
8
15 il 120 800

Table 8.2

User
Final Total
Producer A B demand output

A 500 350 150 1000

B 320 360 120 800
 8.3 Input-Output Analysis / 617

And, if H = , then

11 7 200 1138|
175 T3
X = [I-A]‘1H = ^ 8 1
33 3 100 8444

That is, industry A must have an output of 1138f and industry B must have an output of 844|,
where output is given in millions of dollars of products.

EXAMPLE

Consider a very simple hypothetical economy of three industries, A, B, and C, represented in

Table 8.3, where data are in millions of dollars of products.
Determine the output vector of the economy if final demand changes to:
(a) 50 for A, 10 for B, and 100 for C
(b) 100 for A, 20 for B, and 60 for C
(c) 80 for A, 100 for B, and 120 for C.
r 1 1 1 -1
6 4 4

A = 1 1 1
4 4 3
1 -5«-1 1
L3 3 3 J
j— 5 1 1■
6 — cf '4

[I - A] = 1 3 1
4 P ’3
1 1 2
L 3 3.

r 7 13 ■
T8 4 45

[I-A]"1 = 864 1 31 49
109 3 72 T44
11 29 9
l_34 72 T3 .

note : = 35 . 43 . 32
11 - A I| - 1.1.1
4 3 4 ~ 3
1.1.1
3 4

/I . 3 . 1 , 1 . 1 . 5 , 2 . 1 . 1\
“ (i
“IT 4 3+3 3 3 + 3 4 4j
1 09
864

3 1 5 1
1+1 4 —3 7 2+3 3 'T 29
ci i — (~ 1) 1 2 = T5 C23 — (~ 1) 1 1 73
3 3 ’3 -3

1 — -31 1 1
1 +2 4 1 3+ 1 'T T 13
C12 — ( — 1) 1 2 3 C31 — ( — 1) 3 1 45
1 3 T -3

1 3 5 1
1+3 4 4 11 3+2 3 '4 49
C13 — (~ 1) 1 1 34 C32 - ( — 1) 1 1 T44
3 'T '3

1 1 5 1
2+ 1 4 —4 1 3+3 3 4 9
C21 ~ 1) 1 2 4 C33 — ( — 1) 1 3 T3
3 1 "T 4

5 1
2+2 6 —4 31
c22 ~ ( ~ 1) 1 2 73
2 3

Table 8.3

User
Final Total
Producer A B C demand output

A 90 150 225 75 540

B 135 150 300 15 600
C 270 200 300 130 900
618 / Applications of Matrix Algebra

As must be the case,

r 7 1 13 i
1" 757 " 5401
T7 4 47
1 31 49
15 = 600
X = [I — A] ~1H = to! 7 72 144
11
1-74
29
72
9
17 J Ll 30 _ L 900 _

" 50 "
(a) If H = 10
100
r 7 1 13 -|
17 4 47 r 5o~
Ai-ln _ 864 1 31 49
X = [i -- AJ H — 7 72 144 10
11
l_74
29
72
9
17 J Lioo J
"388.62"
435.96
_659.45_

100 '

(b) IfH = 20
60
r 7 1 13 i
O
o

17 4 47
1 31 49
X = [I — A]- 'H = 7 72 144 20
11 29 9
L 24 72 T7 J . 60 _

"476.70"
494.31
694.68

' 80'
(c) If H = 100
120
r 7 1 13 -i
17 4 47 r 80]
at-Iij _ 864 1 31 49
X = [I AJ H — JU7 7 77 144 100
11 29 9
L74 72 17 J |_120_
" 702.39"
876.33
1144.95

PROBLEMS
1. A simple hypothetical economy of three industries, A, B, and C, is represented in the
following table (data are in millions of dollars of products):

User
Final Total
Producer A B c demand output

A 80 100 100 40 320

B 80 200 60 60 400
C 80 100 100 20 300

Determine the output vector for the economy if the final demand changes to a. 120 for A, 40
for B, and 10 for C; b. 60 for A, 60 for B, and 60 for C.
2. A simple hypothetical economy of two industries, A and 5, is represented in the following
table (data are in tens of millions of dollars of products):
 8.4 Linear Programming / 619

User
Final Total
Producer A B demand output

A 14 6 8 28
B 7 18 11 36

Determine the output vector for the economy if the final demand changes to a. 16 for A and
3 for B; b. 2 for A and 4 for B.
3. A simple hypothetical economy of two industries, A and B, is represented in the following
table (data are in millions of dollars of products):

User
Final Total
Producer A B demand output

A 150 240 210 600

B 200 120 160 480

Determine the output vector for the economy if the final demand changes to a. 100 for A
and 200 for B,b. 50 for A and 60 for B.
4. A simple hypothetical economy of three industries, A, J3, and C, is represented in the
following table (data are in tens of millions of dollars of products):

User
Final Total
Producer A B C demand output

A 5 4 3 3 15
B 3 10 6 1 20
C 3 4 4 1 12

Determine the output vector for the economy if the final demand changes to a. 1 for A, 2 for
B, and 10 for C; b. 2 for A, 3 for B, and 2 for C.

Answers to Odd-Numbered Problems

1. a. 481.74 for A, 469.57 for 5, and 371.74 for C
b. 469.56 for A, 542.61 for 5, and 469.56 for C
3. a. 442.11 for A and 463.16 for B
b. 170.53 for A and 155.79 for B

8.4 Linear Programming

Many problems in business and economics are concerned essentially with the alloca¬
tion of limited resources—money, personnel, materials, machines, space, time—in
order to maximize some measure of performance or minimize some measure of cost.
620 / Applications of Matrix Algebra

The mathematical techniques for planning such allocations are referred to as math¬
ematical programming; the special case in which the measure of performance or cost
is a linear function and the restrictions on the availability or utilization of resources
are expressible as linear equations or inequalities is referred to as linear programming.
More specifically, the general linear programming problem involves maximizing
or minimizing a linear function of several primary variables, referred to as an objective
function, subject to a set of linear equalities or inequalities referred to as constraints.
None of the variables may be negative. (Note, however, that a negative variable may
be written as the difference of two positive variables.)
Mathematically, the linear programming problem for maximization is written

Maximize an objective function Z = c1Xl + c2X2 + • * • + cnXn

subject to the constraints

#11 + #12^2 + * *' + ainXn ^ ^1

#21^1 + #22^2 + T- a2n^n — b2

®mi Xi T ^m2^2 T ~b amnXn < bm

Xj> 0 7=1,2,...,«
This may be written more compactly as

Maximize Z = ]T CjXj
j= 1
n

subject to Yj aijXj < bt for i = 1, 2, ..m

7=1

Xj>0 for j = 1, 2, .... n

and, in matrix notation, as

Maximize Z = CX

subject to AX < B, X > O, where

C (cj, C2 , ..., c„)

xo
x2
X =

#11 a12

a21 a22 a2 n
A =
l a mi am2 a mn J

b1
^2

B=
lb m
Similarly, a linear programming problem involving minimization may be stated in
matrix notation as

Minimize Z = CX
 8.4 Linear Programming / 621

subject to AX > B, X > 0. These formulations are referred to as the standard forms
for the problems of linear maximization and minimization, respectively, because the
constraint inequalities are all expressed in the form < for the maximization problem
and in the form > for the minimization problem. In order to write a problem in
standard form, it may be necessary to reverse the direction of an inequality by
multiplying by a negative one. For example,

anxi + a 12x2 ^ b

can be written

-allxl - a12x2 > -b

Although the standard form may be convenient, it is not necessary for the solution
of a linear programming problem. The standard form is necessary for obtaining the
dual, as discussed later in this section.
It can be shown that when there are m constraints in a linear programming
problem, there are at most m variables or allocations in its solution. Various methods
are available for determining which combination of m of the X/s maximizes or
minimizes the objective function Z. Linear programming problems involving no
more than two primary variables can be solved geometrically; although algebraic
solution is usually more efficient, even for these problems, geometric solution is
intuitively appealing and helps develop a basic understanding which is applicable to
more complicated problems. For this reason, geometric solution is discussed below.
There are several algebraic methods for solving linear programming problems; for
more complicated problems, involving many variables and many constraints, the
simplex method is the most efficient of these. The simplex method is an iterative
computational routine, based on matrix algebra, which develops successively better
solutions until the optimal solution is obtained. Most computer packages include a
program for simplex solution of linear programming problems. Simpler solution is
discussed and illustrated in this section. The interpretation of the dual of a linear
programming problem is also considered.

GEOMETRIC SOLUTION
Linear programming problems involving two primary variables are solved geome¬
trically by graphing the inequality constraints as equalities and thus determining a
polygon of feasible solutions. A solution is said to be feasible if it satisfies all the
constraints of a linear programming problem. Once the polygon of feasible solutions
has been obtained, the next step is to determine which feasible solution optimizes
(maximizes or minimizes) the objective function and is thus the solution of the
problem. The following theorem is used to reduce the number of feasible solutions to
be checked.

theorem: If there is a unique solution which maximizes or minimizes a linear

objective function, then that solution must correspond to a vertex (or corner) of the
polygon of feasible solutions; if there is more than one solution, at least two of the
solutions must correspond to adjacent vertices of the polygon of feasible solutions.

Thus the value of the objective function need be computed only for solutions
corresponding to vertices of the polygon of feasible solutions in order to determine
the optimal solution. (In some cases certain of these solutions are clearly inferior to
others and the corresponding values of the objective function need not be computed.)
622 / Applications of Matrix Algebra

EXAMPLE

A manufacturer produces bicycles and motor scooters, each of which must be processed
through two machine centers. Machine center 1 has a maximum of 120 hours available and
machine center 2 has a maximum of 180 hours available. Manufacturing a bicycle requires 6
hours in machine center 1 and 3 hours in machine center 2; manufacturing a motor scooter
requires 4 hours in machine center 1 and 10 hours in machine center 2. If profit is $45 for a
bicycle and $55 for a motor scooter, determine the number of bicycles and the number of
motor scooters that should be manufactured in order to maximize profit.
If X i = number of bicycles and X2 — number of motor scooters, the problem can be stated
as follows:

Maximize Z = 45*! + 55*2

subject to 6Xx + 4X2 < 120, 3*: + 10*2 < 180, and Xu *2 > 0. The polygon of feasible
solutions is obtained by graphing the equations

6*x + 4*2 = 120

3*! + 10Jr2 = 180

*i =0

*2=0

Any point lying within or on the edge of the shaded polygon in Figure 8.1 corresponds to a
pair of values *x and X2 which satisfy the constraints of the problem and thus are a feasible

Figure 8.1

*2

solution. The optimal solution corresponds to one of the vertices of the polygon; the objective
function thus must be evaluated for the solutions (0, 0), (0, 18), (10, 15), and (20, 0).

Solution Z = 45*! + 55*2

(0, 18) $ 990

(10, 15) $1275
(20, 0) $ 900

Since (0, 0) is clearly inferior to (0, 18), it need not be considered. Thus the optimal solution
is (10, 15); that is, manufacture 10 bicycles and 15 motor scooters.
 8.4 Linear Programming / 623

EXAMPLE

A company produces two commodities in quantities Xx and X2, respectively, and wishes to
minimize Cost = 2Xx + 10X2 subject to the constraints

2X, + A2 < 6

5Xx + 4X2 > 20

Xu X2 >0

Determine the optimal quantities of each commodity to be produced and the associated cost.

Figure 8.2

A2

The polygon of feasible solutions is obtained by graphing the equations

2Xt + X2 = 6

5Xx + 4X2 = 20

Xx =0

X2 =0

Any point lying in or on the edge of the shaded polygon in Figure 8.2 corresponds to a feasible
solution. The optimal solution is one of the vertices (0, 5), (0, 6), and (3, -x).

Solution Z = 2Xt + 10X2

(0, 5) $50
& $36

Note that (0, 6) need not be evaluated, since clearly it is associated with a higher cost than
(0, 5). The optimal solution is (4 ^); that is, produce 3 units of the first commodity and 1y
units of the second commodity. The associated cost is $36.
624 / Applications of Matrix Algebra

Figure 8.3

*2

Note that the constraints determine the polygon of feasible solutions. The objec¬
tive function is used in determining which of these feasible solutions is optimal. It is
possible to write sets of constraints for which there is no feasible solution. For
example, there is no set of values for X1 and X2 which satisfies the constraints

2X1 +X2<6

5X1 + 4X2 > 40

(See Figure 8.3.)

Figure 8.4

*2
 8.4 Linear Programming / 625

There are also some sets of constraints for which the polygon of feasible solutions
is not bounded. For example, if the inequality sign of the first constraint above is
reversed and the second constraint is unchanged,
2Xx +X2>6

5Xx + 4X2 > 40

the polygon of feasible solutions is unbounded. (See Figure 8.4.) In this case, the
linear programming problem has a solution if it involves minimization, but not if it
involves maximization.

THE SIMPLEX METHOD

The simplex method solves linear programming problems by obtaining a feasible
solution and, by an iterative procedure, improving this solution until the optimal
solution is obtained. The computational routine of the simplex method is based on
matrix algebra and consists essentially of obtaining an inverse matrix in order to
solve a set of simultaneous linear equations.
For simplex solution it is necessary to write the constraint inequalities as equali¬
ties (that is, equations) by adding positive variables, referred to as slack variables, to
the left-hand side of the inequalities. For a maximization problem written in stan¬
dard form, one slack variable is added to the left-hand side of each inequality. If the
slack variable added to the z'th equation is denoted by Xn + i, the problem can be
rewritten as follows:

Maximize Z = clXl 4- c2X2 4- ••• 4- cnXn

subject to

auXi +tfi2^2 + ‘“+ alnXn + Xn+1 = bl

a2iX± a22X2 4- F a2nXn 4~ Xnjr2 b2

®mi ^i "F ^m22 "F 4~ amnXn 4~ Xn + m bm

Xj> 0 j = 1, 2, ..., n 4- m

In this problem there are n primary variables and m constraints. The addition of
slack variables results in m equations involving m + n unknowns. Note that the slack
variables are omitted from the objective function or, equivalently, are given a value of
zero in the objective function.
If an inequality is of the form >, a positive slack variable is subtracted from the
left-hand side of the inequality; this slack variable is given a value of zero in the
objective function. In order to obtain a feasible solution, an artificial variable is
added to the left-hand side of the inequality; this artificial variable is given a value of
— M in the objective function for a maximization problem and a value of 4- M in the
objective function for a minimization problem, where M is very large in absolute
value. The artificial variable is thus assigned a value in the objective function that
precludes its appearance in the final solution. For example, if an inequality is of the
form

®iiX y + an^ 2 + 4- alnXn > b{

626 / Applications of Matrix Algebra

Table 8.4
Profits per unit
Variables in the solution
Quantities of the variables
{ | in the solution

cx C2 • •• cn 0 0 • • 0
Variables
Cj allocated Quantity X, X 2 ■ •• Xn Xn+1 Xn + 2 • Xn + m

0 xH+1 bi all «12 • ® In 1 0 • • 0

0 Xn + 2 b2 a21 a2 2 ®2n 0 1 • •• 0

0 Xn + m bm am 1 am2 ®mn 0 0 • i

x n *m x m

(body matrix) (identity matrix)

a slack variable Xn+1 is subtracted from and an artificial variable Xn + 2 is added to

the left-hand side

auX i + a12X 2 + • • ■ + alnXn — Xn+l + Xn + 2 = bx

where the value of Xn+x in the objective function is zero and the value of Xn + 2 in the
objective function is — M for maximization and +M for minimization.
For convenience, the data are arranged in a simplex tableau on which the compu¬
tational routine is performed. The initial tableau for the maximization problem
stated above is of the form shown in Table 8.4. In the simplex method, two rows are
usually included in addition to the rows representing the constraints: these are the Zj
row, representing the total profit from the solution, and the C} — Z7 row, represent¬
ing the net profit from adding one unit of a variable.
The simplex method consists of the following basic steps, which are illustrated in
detail in the examples:

1. Add the necessary slack and artificial variables to convert inequalities to

equalities.
2. Arrange the data in a simplex tableau.
3. Determine a feasible solution from the simplex tableau.
4. Check the solution for optimality.
5. If the solution is not optimal, determine from the tableau the entering variable and
the departing variable for the next solution.
6. Compute the entries for the revised tableau.
7. Check the solution of the revised tableau for optimality.
8. Repeat this procedure (steps 5 through 7) until the optimal solution is obtained.
These steps are illustrated in detail below.
Consider the problem of the manufacturer of bicycles and motor scooters (see
page 622). Adding the slack variables X3 and X4, the linear programming problem
can be stated as follows:

Maximize Z = 45Xt + 55X2

subject to

6Xx + 4X2 + X3 = 120

3ATj + 10X2 + X4= 180

 8.4 Linear Programming / 627

Table 8.5

$45 $55 $0 $0
Variables
Cj allocated Quantity Xt X2 X3 X4

$0 x3 120 6 4 1 0
$0 X4 180 3 10 0 1
2, $0 $0 $0 $0 $0
Cj - Zj $45 $55 $0 $0

T
Optimal column

With the necessary slack variables included, the data are arranged in a simplex
tableau, and an initial feasible solution is indicated in the tableau (steps 1 through 3).
The initial simplex tableau is shown in Table 8.5.
Note the following characteristics of the initial simplex tableau:

a. Any unknown that occurs in one equation occurs (implicitly at least) in all equa¬
tions ; the unknowns that do not appear explicitly in an equation are considered to
have zero coefficients in that equation. Thus the zeros in the X4 column of the first
row and the X3 column of the second row correspond to those in the equations

6X, + 4X 2 + + 0X4 = 120

3Xx + 10X2 + 0X3 + X4 = 180

b. The initial tableau is constructed for the solution having zero profit. The objective
function can be written

Z = 45X1 + 55X2 + OX3 + 0X4

If X3 = 120 and X4 = 180, then X1 = X2 = 0 and Z is also zero. This solution is

obviously not optimal, but provides a convenient initial feasible solution. The
solution is then checked for optimality.

Step 4. Check for optimality: The C; — Z-} row gives the net profit from adding one
unit of a variable. For a maximization problem, the presence of at least one positive
number in the C7 — Zj row indicates that profit can be improved; absence of positive
numbers in the Cj — Zj row indicates that profit cannot be improved, that is, the
optimal solution has been obtained. For a minimization problem, the presence of at
least one negative number in the Cj — Zj row indicates that cost can be decreased;
absence of negative numbers in the C7 — Zj row indicates that cost cannot be de¬
creased, that is, the optimal solution has been obtained. In this example, there are
positive numbers in the C, — Zj row of the initial tableau, so the corresponding
solution is not optimal.
Step 5. Entering and departing variables: The largest contribution to profit per
unit ($55) is made by variable X2, so it is the entering or replacing variable. To
determine the departing variable, the quantities of X3 and X4 in the initial tableau
are divided respectively by their corresponding entries in the optimal column:

120 hours available

X3 row: —-— -; = 30 units ol X2
4 hours per unit

180 hours available

X4 row: ■-— = 18 units of X2
10 hours per unit
628 / Applications of Matrix Algebra

Table 8.6
$45 $55 $0 $0
Variables
Cj allocated Quantity x2 X3 X 4.

$0 X3 48 fl4\ 0 1 2
—1
$55 T2 18
© 1 0 TO

Zj $990 $-¥ $55 $0 $J2L

Cj ~ Zj $¥ $0 $0 $ -¥

The variable corresponding to the smaller positive ratio is the departing or replaced
variable; thus X4 is the replaced variable in the example. (Note that 18 is the largest
number of units of X2 than can be added without violating the constraints.)
Step 6. Recompute the simplex tableau: The X2 row of the revised tableau is
computed by dividing each number in the replaced row X4 by the element in the
optimal column of the replaced row, which is 10 in this example (Table 8.6).
The entries in the X2 row are
18 & 1 0 ^
All the remaining rows corresponding to variables in the tableau are computed using
the formula

element of old row / corresponding

/ element in \ _ / element in
in the optimal x element in
\ new row / — \ old row
column \ replacing row
In this example, the only remaining variable row is X3; its entries in the revised
tableau are

120 - (4 x 18) = 48

6 - (4 x i) = ^
4 — (4 x 1) = 0

1 - (4 x 0) = 1
0 - (4 x -[L) =

is the total profit for the solution = 0 • 48 + 55 • 18 = 990

for r, = (0)(^) + (55)(-A) = ¥ I

for X2 = (0)(0) + (55)(1) = 55 profit given up by adding

for X3 = (0)(l) + (55)(0) = 0 one unit of the variable

for X4 = (0)(—|) + (55)(xs) = -x
The Cj — Zj row gives the net profit per unit (profit per unit minus profit given up by
adding one unit). The positive entry in the revised tableau indicates that the
solution can be improved by the inclusion of variable X v Checking to see which
variable is to be omitted (although it is obviously X3 in this example),

10

60
 8.4 Linear Programming / 629

Table 8.7

$45 $55 $0 $0
Variables -
Cj allocated Quantity X2 *3 Z4

5
$45 10 1 0 14 T2
1 1
$55 X2 15 0 1 ~T6 ~5
<p 95 <r 25
2, $1275 $45 $55 4>T5
<r 95 <r 25
Cj - Zj $0 $0 ~ T5 -8“

The smallest positive ratio corresponds to X3, and X3 is thus the departing variable.
The simplex tableau is recomputed the same way as in the preceding stage (see Table
8.7). The absence of positive entries in the Cj — Zj row indicates that the optimal
solution has been obtained: Manufacture 10 bicycles and 15 motor scooters, as was
also determined above using the geometric method.
Note the following characteristics of the simplex tableau:

a. Slack and artificial variables are included in the constraint equations as follows: If
an inequality constraint is of the form <, a slack variable is added and is given a
value of zero in the objective function for both maximization and minimization
problems. If an inequality constraint is of the form >, a slack variable is sub¬
tracted and is given a value of zero in the objective function for both maximization
and minimization problems; in addition, an artificial variable is added and is
given a value in the objective function of — M for maximization problems and
+ M for minimization problems.
b. The elements of the body and identity matrices of the simplex tableau represent
marginal rates of substitution between the variables in the solution and the var¬
iables heading the columns. For example, in Table 8.6, X3 must be decreased by ^
units if 1 unit of X1 is added; X3 must be increased by f units if 1 unit of X4 is
added, and so forth. Note that a variable always has a marginal rate of substitu¬
tion of 1 with itself and a marginal rate of substitution of 0 with a variable with
which it is not substitutable.
c. The body matrix of the initial tableau has become an identity matrix in the final
tableau; the identity matrix of the initial tableau has become the inverse of the
initial body matrix in the final tableau.
d. In the final tableau, the C} - Z; row consists of zero or negative numbers for a
maximization problem and of zero or positive numbers for a minimization
problem.

EXAMPLE

Consider the problem of cost minimization discussed above (see page 623). Adding the slack
variables X3 and X4 and the artificial variable X5, the linear programming problem can be
stated as follows.

Minimize Z = 2Xx + 10X2 + MX5

subject to

2 X i + X2 + X3 = 6

5X i -f- 4X2 — X 4 -f- X$ = 20

X 1, X 2 , X 3 , X 4. , X 5 >0

The initial simplex tableau is shown in Table 8.8.

630 / Applications of Matrix Algebra

Table 3.8

$2 $10 $0 $0 $M
Variables
Cj allocated Quantity Xi X2 *3 *4 *5

$0 *3 6 2 1 1 0 0
$M V5 20 5 4 0 -1 1
Zj $20M $5 M $4M $0 $ - M $M
Cj - Zj $(2 - 5M) $(10 - 4M) $0 $M $0

Optimal column

The variable to enter at the next step is and the variable to be replaced is X3 since

X3:f = 3
*5:^ = 4

The revised simplex tableau is shown in Table 8.9.

Table 8.9

$2 $10 $0 $0 $M
Variables
Cj allocated Quantity X1 X2 *3 *5

$2 *1 3 1 h 0 0
3 5
$M *5 5 0 —? -1 1
Zj $(6 + 5 M) $2 $(X + 3 M/2) $(1 5 M/2) $ - M $M
Cj - Zj $0 $(9 - 3M/2) $( - 1 + 5M/2) $M $0

T
Optimal column

The variable to enter at the next step is X2 and the variable to be replaced is X5 since

X,:~ = 6
1

The revised simplex tableau is shown in Table 8.10. The absence of negative entries in the
Cj — Zj row indicates that the optimal solution has been obtained: manufacture § units of the
first commodity and ^ units of the second commodity at a total cost of $36, as was also
determined above by the graphical method.

Table 8.10

$2 $10 $0 $0 $M
Variables
allocated Quantity Xx X2 X3 X4 X5

$2 4 1 0
$10 10 0 1
X2 T
Zj $36 $2 $10 $-14 $-6 $6
~Zj $0 $0 $14 $6 $(M - 6)
 8.4 Linear Programming / 631

There are several variations of the standard simplex method, although none differs
substantially from the one discussed above. In addition, there are some linear pro¬
gramming problems in which the coefficients in the constraints have special forms
and special cases of the simplex method, which require less computation, can be used.
The simplex method must be modified if the solution variables are required to be
integer-valued.

THE DUAL OF A LINEAR PROGRAMMING PROBLEM

Corresponding to every linear programming problem is a second linear program¬
ming problem referred to as the dual. When the initial or primal problem involves
maximization (minimization) of an objective function, the dual problem involves
minimization (maximization) of an objective function. The number of variables in
the dual problem is equal to the number of constraints in the primal problem and the
number of constraints in the dual problem is equal to the number of variables in the
primal problem.
For the linear programming problem stated above (page 620), the dual can be
stated as follows:
Minimize Zdual = b{ V, + b2Y2 + ■ ■ ■ + bmYm

subject to

allY1 + a2i Y2 + • • • + aml Yt > c1

a12 Yl + a22 Y2 + • - T am2 Ym > c2

@ln ^1 f @2n ^2 F @mn

Yj> 0 j = 1, 2, ..., m

Thus if a linear programming problem, written in standard form, is

Maximize CX

subject to

AX < B

X>0
then its dual, written in standard form, is

Minimize BY

subject to

AY > C

Y > 0
and, similarly, if a linear programming problem, written in standard form, is

Minimize CX

subject to

AX > B

X > 0
632 / Applications of Matrix Algebra

then its dual, written in standard form, is

Maximize BY

subject to

AY < C

Y > 0

The correspondence between the primal and dual problems can be summarized as
follows:
1. The constants in the constraints of the primal are the coefficients in the objective
function of the dual.
2. The coefficients in the objective function of the primal are the constants in the
constraints of the dual.
3. The inequality signs in the constraints of the primal and dual are reversed.
4. The matrices of coefficients in the constraints of the primal and dual are trans¬
poses of each other; for example, the ith row of coefficients in the constraints of
the primal is the ith column of coefficients in the constraints of the dual.

EXAMPLE

Write the dual of each of the following linear programming problems.

(a) Min 5Xi + 9X2

subject to

3Xx + 2X2 < 6

5Xi + X2 > 10

X! + 10Y2 > 9

*!>() *2 >0
Writing the problem in standard form, v

Min 5Aj + 9X2

subject to

-3Xi - 2X2 > -6

+ X2 > 10

Xl + 10X2 > 9

X,>0 X2>0

and its dual is

Max -6T + 1072 + 9y3

subject to

-37! + 5y2 + y3 < 5

-27! + y2 + ioy3 < 9

7i > o y2 > o y3 > o

(b) Max Xi + 9X2 + 15X3

 8.4 Linear Programming / 633

subject to

3Xi 4 2X2 > 11

X^ 4 X 2 “I- X 3 > 15

8X2 4 IX3 < 25

X{>0 X2>0 X3>0

Writing the problem in standard form,

Max Xl 4- 9X2 4- 15^3

subject to

-3Xi - 2X2 < -11

-Xt - x2- x3< -15

8X2 4 7X3 < 25

*!>() x2>0 X3>0

and its dual is

Min -117! - 1572 4 2573

subject to

-37i - 72 > 1
— 2Y1 - Y2 4 873 > 9

-Y2 4 7y3 > 15

y{ > 0 y2 > 0 y3 > 0

The solution of the dual problem also provides the solution of the primal problem.
The number of iterations required for simplex solution depends in part on the
number of rows of variables in the simplex tableau; thus, if m < n, solution of the
dual problem usually requires less computation and is therefore preferable.
The correspondence between the primal and dual solutions can be summarized as
follows:

1. The value of the objective function is the same for the primal and dual solutions.
2. The criteria for the primary variables of the primal are the solutions for the slack
variables of the dual.
3. The criteria for the slack variables of the primal are the solutions for the primary
variables of the dual.
4. The solutions for the primary variables of the primal are the negative values of the
criteria for the slack variables of the dual.
5. The solutions for the slack variables of the primal are the negative values of the
criteria for the primary variables of the dual.

EXAMPLE

For the maximization problem on page 622,

Max Z = 45X: 4 55*2

634 / Applications of Matrix Algebra

Table 8.11
$120 $180 $0 $M $0 $M
Variables
Cj allocated Quantity Yx y2 >3 Ya Y5 Y6

95 5 1
$120 T6 1 0 — J4 16 — T5
1 1 1
$180 y2 ¥ 0 1 ~8

1
T2 —5

27
Zj $1275 $120 $180 $ - 10 $ - 15 $15

©
Cj - Zj $0 $0 $10 $15 $(M - 15)

subject to

6Xl + 4X2 < 120

3*! + 10X2 < 180

Xu X2>0
the dual is given by

Min 120Yi -f 180Y2

subject to

6Y, 4- 3Y2 > 45

4Yx + 10Y2 > 55

Yu Y2>0
With slack and artificial variables, the dual problem is

Min 120Yx 4- 180Y2 4- MY4 4- MY6

subject to

6Yx 4- 3Y2 - Y3 + Y4 = 45

4Yx 4- 10Y2 - Y5 4- Y6 = 55

Yu Y2,Y3,Y4,Y5,Y6>0
\

The final tableau of the simplex solution of the dual is given in Table 8.11. Note the correspon¬
dence between Tables 8.7 and 8.11 with respect to the properties listed above.

EXAMPLE

For the minimization problem on page 623,

Min 2Xx + 10X2

subject to

2Xx + X2<6

5Xx 4- 4X2 > 20

Xu X2 >0
the dual is given by

Max — 6yj 4- 20Y2

 8.4 Linear Programming / 635

Table 8.12

$ - 6 $20 $0 $0
Variables
Cj allocated Quantity yi y2 y3 Y4

$20 6 0 1 -4 2
^2 3
4 5
$ - 6 V! 14 1 0 ~7 3
Z, $36 $ - 6 $20 $4 $^
Cj - Zj $0 $0 S-j $-^

subject to

-2Yt + 5Y2 < 2

-Yt + 4Y2 < 10

Yu Y2 > 0

With slack and artificial variables, the dual problem is

Max —6 Yx + 20 72

subject to

-2 Yx + 5Y2 + Y3 = 2

- Yi + 4Y2 + Y4 = 10

Yu Y2 , Y3, Y4 > 0

The final tableau of the simplex solution of the dual is given in Table 8.12. Note the correspond¬
ence between Tables 8.10 and 8.12 with respect to the properties listed above.

Thus the dual problem is completely symmetric to the primal problem and the
solution of one problem provides the solution of the other problem. Since the
number of iterations required for simplex solution depends in part on the number of
rows in the simplex tableau, the dual can provide a simpler computational problem
than the primal when the simplex tableau of the dual has fewer rows than the simplex
tableau of the primal.
The solution of the dual provides implicit, imputed, or shadow prices that can be
used as the basis for efficient allocation of resources. These shadow prices are also
used in sensitivity analysis, which determines the effect on the optimal solution if
constraints or objective function coefficients change. These analyses are discussed in
economics and are only mentioned here to indicate some of the applications of the
dual.

PROBLEMS
1. A company manufactures two types of novelty souvenirs made of plywood. Souvenirs of
type A require 5 minutes each for cutting and 10 minutes each for assembling; souvenirs of
type B require 8 minutes each for cutting and 8 minutes each for assembling. There are 3
hours and 20 minutes available for cutting and 4 hours available for assembling. The profit
is 50^ for each type A souvenir and 60^ for each type B souvenir. How many souvenirs of
each type should the company manufacture in order to maximize profit?
636 / Applications of Matrix Algebra

2. Determine the values of x and y that maximize the function

/(x, y) = 0.4x + 0.9y

subject to the constraints

5x + 3y < 30 lx + 2y < 28 x >0 y>0

3. Determine the values of xx and x2 that maximize the function

/(x, x2) = 3xi + 5x2

subject to the constraints

X! + 2x2 < 10 3xi + x2 < 10 Xi > 0 x2 > 0

4. Determine the values of Xi and x2 that maximize the function

f(xu x2) = 3xi + 6x2

subject to the constraints

2xx + 5x2 < 20 2xi + 2x2 <10 xt > 0 x2 > 0

5. Determine the values of x and y that maximize the function

/(x, y) = 3x + 2y

subject to the constraints

x + y <7 2x + 3y < 16 x > 0 y > 0

6. A manufacturer of a line of hair shampoos is planning a production run for products A

(for dry hair) and B (for oily hair). There are sufficient ingredients on hand for 60,000
bottles of each shampoo but there are only 60,000 bottles into which either shampoo can
be put. It takes 4 hours to prepare enough shampoo to fill 1000 bottles of shampoo A and
3 hours to prepare enough shampoo to fill 1000 bottles of shampoo B ; there are 200 hours
available for preparation. Profit is 9 cents a bottle for shampoo A and 7 cents a bottle for
shampoo B. How should production be scheduled to maximize profit?
7. The Tr-Ply Company manufactures two types of heavy-duty wood crates. The profit on
each crate of type 1 is $9 and the profit on each crate of type 2 is $12. Each crate must go
through two production lines; a total of 10 hours are available on production line A and a
total of 12 hours are available on production line B. Each crate of type 1 requires 2 hours
on line A and 4 hours on line B. Each crate of type 2 requires 5 hours on line A and 3 hours
on line B. Determine the number of crates of each type that should be produced in order to
maximize the associated profit.
8. The UP-DOWN Ladder Company can manufacture three types of ladders; profit is $5 per
ladder of type 1, $7 per ladder of type 2, and $8 per ladder of type 3. Each ladder must be
processed through three centers according to the following requirements:

Center 1 Center 2 Center 3

(min) (min) (min)

Type 1 4 5 6
Type 2 5 7 9
Type 3 6 7 7
Total available 80 100 120

In order to maximize profit, determine the number of ladders of each type that should be
manufactured.
9. The Capsize Yachting Company produces two models of racing boats. Their profit is $520
for Model 1 and $450 for Model 2. Model 1 requires 40 hours for cutting and assembling
 8.4 Linear Programming / 637

and 24 hours for finishing. Model 2 requires 25 hours for cutting and assembling and 30
hours for finishing. There are 400 hours available for cutting and assembling and 360
hours available for finishing. Determine the optimal number of each model for the com¬
pany to produce and the resulting profit.
10. The Duoply Company can make three products; their profit is $10 per unit for Product 1,
$14 per unit for Product 2, and $15 per unit for Product 3. Each of these products is
manufactured using three raw materials according to the following requirements:

Material 1 Material 2 Material 3

Product 1 3 lb/unit 4 lb/unit 5 lb/unit

Product 2 5 lb/unit 7 lb/unit 7 lb/unit
Product 3 4 lb/unit 8 lb/unit 6 lb/unit
Total lb
available 220 280 320

Determine the optimal product mix.

11. College Publishers, Inc., plans to use one section of its plant to produce two textbooks.
The profit is $2 for Textbook 1 and $3 for Textbook 2. Textbook 1 requires 4 hours for
printing and 6 hours for binding. Textbook 2 requires 5 hours for printing and 3 hours for
binding. There are 200 hours available for printing and 210 hours available for binding.
Determine the optimal number of each textbook for the publishers to produce and their
resulting profit.
12. Obtain the solution for the following linear programming problem.

Maximize 6*: + 3*2

subject to

4Xi + X2 < 12

2Xl + 2X2 < 10

2Xi + 4*2 > 8

*ls*2 >0

13. Obtain the solution for the following linear programming problem.

Minimize 12Xx — 5X2

subject to

Xx - 2*2 > 3

X, - X2 > 4

X! > 1

*l5 X2 >0

Write the duals of the following linear programming problems.

14. Max Xi X2 + 5*3

subject to 4X x + 3*2 + *3 < 10

2Xx + 10*2 + 3*3 > 15

XU *2, *3 >0

15. Min 3*! + 5*2 + *3

638 / Applications of Matrix Algebra

subject to X 1 + X2 + X3 > 6

3Xx + 8X2 + 9X3 < 50

6 *! + 1X3 > 12

\2X2 + 4X3 > 15

Xu X2 , X3 > 0

16. Max 5*! - 6X2 + 10X3

subject to X 1 + X2 + X3 > 15

2Xx + 3X2 + 4X3 < 35

3Xx -- 4X2 + 6X3 < 30

x,-x2 > 0

X1, X 2 , X 3 > 0

17. Min 2Xx T 10X2

subject to X 1 + x2 + X3 > 5

3*! -- 2X2 + 6X3 < 20

X2 + 3^3 > 10

X1, X 2 , x3 > 0

18. Max 8Xx - 3X 2

subject to 9Xx + 2X2 < 20

10Xi + 3X2 < 32

Xu X2>0

19. Max 5Xx + 3X2 + 14* 3

subject to + X2 + 3X3 < 14

Xt + 3X2 + 2X3 < 15

Xx + X2 + X3 > 8

*1, X2,X3>0

20. Min 6Xi - 3X2 + 4X3

subject to 3Xt + 6X2 + 2X3 < 30

5Xx + X2 + 6X3 > 25

xu X2,X3>0

Answers to Odd-Numbered Problems

1. 8 type A and 20 type B
3. xx = 2, x2 = 4
15. Max 6^ - 50Y2 + 1273 + 1574
5. x = 7, y = 0
7. V type 1, 7 type 2, profit $33 subject to Yi — 372 + 673 < 3
9. 5 Model 1, 8 Model 2, profit $6200 Yx - 8 Y2 + 1274 < 5
11. 0 Textbook 1, 40 Textbook 2, profit $120 Yi - 972 + 773 + 474 <l
13. Xi = 4, X2 = 0, min = 48 Yu Y2,73,74 >0
 8.5 Game Theory / 639

17. Max 57x - 2072 + 1073 19. Min 147! + 15 72 - 873

subject to 7i — 372 < 2 subject to 27i + 72 - 73 > 5

Yl + 272 + 73 < 10 Yl + 372 - 73 > 3
7i - 672 + 373 < 0 37! + 272 - 73 > 14
Yu Y2,73 > 0 Yi, Y2,Y3> 0

8.5 Game Theory

Game theory was developed for the purpose of analyzing competitive situations
involving conflicting interests. In the situations considered in game theory, there are
two or more persons with different objectives, each of whose action influences, but
does not completely determine, the outcome of the game; furthermore, each person is
assumed to know his opponent’s objectives. Game theory provides solutions to such
games, assuming that each of the players wishes to maximize his minimum expected
profit or, equivalently, minimize his maximum expected loss. This criterion, which is
based on a somewhat conservative view of the problem, is referred to as the minimax
or maximin criterion. It is the basis for the theory of games of strategy, originally
developed by John von Neumann and Oskar Morganstern and subsequently applied
in various areas of research.
Note that the theory of games assumes a particular type of situation involving
maximizing the expected value of a decision made under uncertainty. Decisions
made under uncertainty sometimes involve only one person making a decision; the
events which, with the decision, influence the result are thought of as being controlled
by some random nonrational device. Unless they are reformulated, these decision
problems do not involve games of strategy, since the decision maker does not have a
rational opponent. However, a minimax or game theory solution may be appropriate
even for these decision problems if the decision maker wishes to protect himself
against the worst possible occurrence, even though it results from a random event
rather than from the decision of a rational opponent.
Most recreational games such as tick-tack-toe, checkers, backgammon, chess,
poker, bridge, and other card games can be analyzed as games of strategy. As usually
formulated, gambling games such as dice and roulette are not games of strategy, since
a person playing one of these games is “playing against the odds,” not against a
rational opponent.
The applications of game theory are not limited to parlor games but include
competitive situations in economics, business, warfare, and social behavior. There
are several fundamental characteristics by which games are classified for solution.
The most important of these are the number of persons, the nature of the payoff, and
the number of available strategies. These characteristics are discussed and methods
of solution of games are then considered.

Number of Persons
Games are classified according to the number of distinct sets of interests or objectives
present in the game. From the game theory point of view, the number of persons in
the game is not necessarily the same as the number of people playing the game—that
is, if two or more players form a coalition in which they agree to pool their winnings
or losses, game-theory analysis treats them as a single person.
Most of the work done thus far in game theory deals with two-person games and
the discussion in the following sections is confined to this case. Analysis of games
640 / Applications of Matrix Algebra

involving more than two persons is usually very difficult. Note that many situations
that are not strictly two-person games may be analyzed as though they were—for
example, the interests in a game of cards can be considered as “his” and “everybody
else’s.”

The Payoff
Games are also classified with respect to the nature of the payoff, that is, what
happens at the end of the game. The distinction in this respect is between zero-sum
games and nonzero-sum games. If the sum of the payoffs to all players of a game is
zero, counting winnings as positive and losses as negative, then the game is zero-sum;
otherwise, it is nonzero-sum. In a zero-sum game, anything won by one player is lost
by another player. The importance of this distinction lies in the fact that a zero-sum
game is a closed system and a nonzero-sum game is not. Almost all parlor games are
zero-sum, and many other situations can be analyzed as zero-sum games. Most of the
work in game theory has concerned zero-sum games and the discussion in the
following sections is confined to such games. Note that a nonzero-sum game may be
made zero-sum by adding a fictitious player, say Nature, but this necessitates a more
difficult analysis, especially if the original game was a two-person game.

Strategies
In game theory, a strategy for a particular player is a plan which specifies his action
for every possible action of his opponent, that is, a strategy is a complete plan for
playing the game, without any connotation of skillfulness on the part of the player. In
a game completely amenable to analysis, it is possible, at least conceptually, to
foresee all eventualities and thus to catalogue all possible strategies. Games are
classified according to the number of strategies available to each player: if player 1
has m possible strategies and player 2 has n possible strategies and they are the only
players, then the game is m x n, that is, m by n. The important distinction in classify¬
ing games on the basis of strategies is between finite games and infinite games. If the
greatest number of strategies available to any player is finite, then the game is finite; if
at least one player has an infinite number of available strategies, then the game is
infinite. The theory of infinite games is very difficult and is not discussed in the
following sections. For the analysis of finite games, it is convenient to distinguish
three cases: those in which the player having the least number of strategies has two,
three, or more than three.
In summary, the discussion in following sections will generally concern finite,
zero-sum, two-person games.

The Game Matrix

A problem is usually set up for game-theory analysis in the form of a game matrix. A
game matrix or payoff matrix is a rectangular array of the payoffs, where the rows
represent the strategies of one player and the columns represent the strategies of the
other player; thus an m x n game is represented by an m x n game matrix. It is
conventional to write the payoffs from the point of view of the player whose
strategies are associated with the rows of the matrix; the payoffs for the other player,
in a zero-sum game, are then given by the negative of this matrix.
Game theory thus assumes that the strategies available to each player can be
enumerated and that the corresponding payoffs can be expressed in meaningful,
although not necessarily monetary, units. This information is sufficient for solution
 8.5 Game Theory / 641

of the game—that is, for determining which choice of strategies each player should
make—assuming that each player wishes to maximize his minimum expected profit
or minimize his maximum expected loss. The minimax theorem, the key result of the
theory of games, states that such a minimax solution exists for every finite, zero-sum,
two-person game. Note that minimax is not the only possible criterion for solving a
game matrix and that its use leads to a conservative theory, since the opponent is
assumed to be skillful and to use his best strategy.
The value of a game is the expected or average payoff per play over a long series of
plays, assuming that both players consistently use their optimum strategies. It is
conventional to regard value from the point of view of the player whose strategies
correspond to the rows of the payoff matrix. A game is said to be fair if its value is
zero; in a fair game neither player has an advantage. In a game that is not fair, one
player will win from the other in the long run if both play their optimum strategies; if
the value of the game is positive, the row player has the advantage; if the value of the
game is negative, the column player has the advantage.

Saddle Points
If a game matrix contains an entry that is simultaneously a maximum of row minima
and a minimum of column maxima, this minimax entry is said to be a saddle point of
the game and the game is said to be strictly determined. In this case, according to the
criterion of game theory, the optimum strategies for the respective players are repre¬
sented by the row and column whose intersection is the saddle point. The value of a
strictly determined game is the value of its saddle point.
The first step in the solution of a matrix game is to check for the existence of a
saddle point; if one is found, the game is solved; if not, further analysis is necessary.
Checking for a saddle point is usually done by writing the row minimum beside each
row and the column maximum at the bottom of each column and then determining
the maximum of the minima and the minimum of the maxima. A saddle point can
also be determined by checking for an entry which is simultaneously the minimum of
the row in which it occurs and the maximum of the column in which it occurs (see
Table 8.13).

Table 8.13

Player Y
1 2 ••• n

1 a\i a12 aln min a{j

j

2 a2\ a22 a2 n min a2j

j
•
Player X • aij
• max min a,-;
< j
m am i U m2 ^mn min amj
j

max an max ai2 max ain

i i i
min max atj
642 / Applications of Matrix Algebra

EXAMPLE

Check the following games for saddle points.

12 2 25 -10 -10
16 (D 4 10 3 . ,
max mm = 3
-2 -1 26 0 -2
14 -4 8 6 -4

16 3 26 10

min max = 3

There is a saddle point at the intersection of the second row (player X's optimum strategy) and
the second column (player Ys optimum strategy); the value of the game is 3.

(b) -15 22 10 8 6 -14 -8 -15

-3 4 -6 0 -4 22 -10 —10 max min — —6
-2 3 4 10 -1 0 o> -6

-2 22 10 10 6 22 -6

min max = — 6

There is a saddle point at the intersection of the third row (player 3Ts optimum strategy) and
the seventh column (player Ys optimum strategy); the value of the game is —6.

(c) -3 2 4 -3
6 1 3 1
3 10 12 3 max min = 3
5 0 -2 -2
0 -4 6 -4

6 10 12

min max = 6

There is no saddle point.

2 0 6 -4 8 10 25 \ -10 12 14 — 10 max min = — 10

-14 -2 0 14 9 12 15 0 -10 -3 -14

2 0 6 14 9 12 25 0 12 14

min max = 0

There is no saddle point.

SOLUTION OF TWO-PERSON, TWO-STRATEGY GAMES

The basic concepts of game-theory analysis discussed in previous sections are il¬
lustrated in the following sections for two-person, two-strategy games, that is, for
games in which at least one of the two players has only two strategies. By convention,
the strategies for player X are listed and indexed in a column along the left edge of
the game matrix and the strategies for player Y are listed and indexed in a row along
the top edge. The payoffs are to player X, that is, a positive number indicates a payoff
from player Y to player X and a negative number indicates a payoff from player X to
player Y.
 8.5 Game Theory / 643

2x2 Games
The most easily analyzed two-strategy games are 2x2, that is, games in which each
player has only two possible strategies. Solution of 2 x 2 subgames is also frequently
necessary as a step in the solution of larger (2 x n or m x 2) two-strategy games.
The existence or nonexistence of a saddle point is easily determined for a 2 x 2
game, either directly by the method above or using the following theorem.

theorem: The 2x2 matrix game

a b
c d

is nonstrictly determined if and only if one of the following two conditions is

satisfied:

1. a < b, a < c, d < b, and d < c.

2. a > b, a > c, d > b, and d > c.

That is, a 2 x 2 matrix game is nonstrictly determined if and only if the two entries
on one diagonal of the matrix are each greater than each of the two entries on the
other diagonal.

EXAMPLE

The game

Player Y

1 2

1 1
Player X
-3 10

is strictly determined and fair. Player X’s optimum strategy is 1; player Ts optimum strategy
is 1.

EXAMPLE

The game

Player Y

1 2

1 5 ©
Player X
-7 -4

is strictly determined, but not fair (its value is 2). Player X's optimum strategy is 1; player P’s
optimum strategy is 2.
644 / Applications of Matrix Algebra

EXAMPLE

The game
Player Y
1 2

1 0 1
Player X
2 0

is not strictly determined.

In a nonstrictly determined game, there is no clearly optimum strategy for either

player to use consistently and, furthermore, consistent use of any particular strategy
by either player can be capitalized on by the other player. There is thus an important
difference between strictly determined and nonstrictly determined games: In a
strictly determined game there is an optimum strategy for each player and no “ secur¬
ity measures” are necessary; in a nonstrictly determined game, optimal play involves
preventing the opponent from knowing what strategy is to be used on a given play.
This is accomplished by selecting the strategy to be used for each play at random,
according to probabilities that can be computed from the game matrix. Such a
strategy, which consists of a probability mixture of more than one (pure) strategy, is
called a mixed strategy.
The solution of a 2 x 2 nonstrictly determined game consists of a pair of probabil¬
ities px and p2 = 1 — Px with which player X selects at random his strategies 1 and 2,
respectively, and a pair of probabilities ql and q2 = 1 — q1 with which player Y
selects at random his strategies 1 and 2, respectively. These probabilities provide the
optimum mixed strategies, that is, the strategies by which each player, respectively,
can maximize his minimum expected profits or minimize his maximum expected
losses against optimal play by the other player.
The value of px can be obtained by equating the two possible expected payoffs for
player X and solving for pu since for any other pl one of the two expected payoffs is
less than the other and thus the minimax criterion is violated. Similarly, ql can be
obtained by equating the two possible expected payoffs for player Y and solving.
If
a b
c d

represents the game matrix, then the expected payoff to player X if player Y uses
strategy 1 is

opl + c(l - Pl)

and if player Y uses strategy 2 is

bpi + d( 1 - p,)

Thus, equating the expected payoffs,

api + c(l - pt) bpi + d{ 1 -

px{a — b — c + d) d — c
d — c
Pi =
a — b — c + d
a -b
Pi = i - Pi =
a — b — c + d
 8.5 Game Theory / 645

Similarly, equating the (negative) expected payoffs to player Y,

aq{ +*>(1 -q1) = cq1 + d( 1 - qx)

a{(a — b — c + d) = d — b
d-b
<?i =
a — b — c + d
a — c
q2 = i - q. a — b — c + d

The value of a game has the same meaning for strictly and nonstrictly determined
games: The value of a game is the payoff that a player can expect to obtain per play;
on the average, a player cannot win more than the value of a game unless his
opponent plays poorly, nor can he win less than the value of a game unless he plays
poorly.
The value of the game
a b
G =
c d

(to player X) is

v — apt + c( 1 - Pi) = bpi + d( 1 - Pi)

= ~[aq 1 + H1 - <?i)] = -[«h + d{ 1 - )]

ad — be
a — b - c + d
The negative of this quantity is the payoff to player Y.

EXAMPLE

2 0
The game is nonstrictly determined. The optimal mixed strategies are
0 2

Pi
2 _
4-1
1

n _ 2 _ 1
Pi — * — 1
— 2 — 1
qn i —t —i

q2 - * — i
and the value is f = 1; the game is biased in favor of player X.

EXAMPLE

-1 0
The game is nonstrictly determined. The optimal mixed strategies are
0 -2

and the value is v = trj = — the game is biased in favor of player Y.

646 / Applications of Matrix Algebra

EXAMPLE

7 -6
The game is nonstrictly determined. The optimal mixed strategies are
5 8
Pi — TZ

Pl=U

■g
14 7
<h T6

„ _ 2 _ 1
<?2 - T5 - S

and the value is v = ft = 5f; the game is biased in favor of player X.

EXAMPLE

10 -30
The game is nonstrictly determined. The optimal mixed strategies are
-10 20
_ _ 30 _ 3
Pi — TO — 7

n _ 40 _ 4
Pi ~ TO ~ 1
„ _ 50 _ 5
<7l ~ 7(5 — 7

<?2 = ^8 = T

and the value is v = —^ = — 1?; the game is biased in favor of player Y.

The play of a game is not affected by adding a constant to all payoffs or by

multiplying all payoffs by a positive constant. The value of the game is affected by the
same transformation as that applied to the payoffs of the game matrix.

EXAMPLE

For each of the games

8 1
4 6

11 4
7 9

16 2
8 12

Pi = l, P2 = i, qi = q2 = f. Note that G2 = Gj + 3; the value of Gi is 4| and the value of G2

is 7f. G3 = 2Gj; the value of Gi is 4| and the value of G3 is 9^.

Solution of 2 x 2 Games by Matrix Algebra

The optimum strategies and the value of a nonstrictly determined 2x2 game can be
obtained using matrix algebra as follows: If the payoff matrix is represented by

_ all a12

a2l a22-
 8.5 Game Theory / 647

then X’s optimal strategies are given by

[I, l][adj A]
[pi. Pi\ =
[1, 1 ] [adj A]

Y’s optimal strategies are given by

[1, 13[adj A]'

[<Jl> =
[1, l][adj A]

and the value of the game is given by

v =
[1, 1] [adj A]

Alternatively, the value of the game is given by

all a12 <h

V = [p 1, p2]
al2 a22

The examples on pages 645 and 646 can be used to illustrate this method of solution.

EXAMPLE

2 0
The game is nonstrictly determined.
0 2
[1, 1] [adj A]
Ip i, Pi\ =
[1, 1] [adj A]

2 0
[1, 1]]
[° 2 [2, 2] 1 1
2 0 1 2’2
[1, 1]
0 2 1

[1, 1][adj A]'

[<7i> 42] =
[1, 1] [adj A]

2 0
[1, 1]
0 2 [2,2] 1 1
4
n
o

T 2 2
’

[L 1] 1
0 2

[1» 1] [adj A]

4-
A- I1
648 / Applications of Matrix Algebra

Alternatively,
a 11 a 12 91
v = [pl, Pl\
a2i a22 92

01 12
= a i] 1
2

= [i, i]

= 1
(as obtained above using algebraic methods).

EXAMPLE

-1 0
The game is nonstrictly determined.
0 -2
[1, 1] [adj A]
[Pi> P2J =
[1, 1] [adj A]

‘-2 0
[1, 1]
0 -1 [-2, -1] 2 1
-2 0] 1] -3 3’ 3
[1, 1]
0 -1

[1, 1][adj A]'

[9i> 92] =
[1, 1] [adj A]

■2 0
[1, 1] 0 1 2 1
[-2, -1]
-2 0 -3 3’3
[1, 1]
0 1

v =
[1, 1] [adj A]

2 2
^3 3

Alternatively,

a 12 91
v = [pi, P2]
a21 a 22 92
2
i o 3
= ti ti 0 -2 i
3.

[2
3
[-4 -f] 1
.3 .

_ _ 2.
3

(as obtained above using algebraic methods).

 8.5 Game Theory / 649

EXAMPLE

7 -6
The game is nonstrictly determined.
5 8

[1, 1 ] [adj A]
[PoP2] =
[1, 1] [adj A]

' 8 6
[i, i;]
-5 7 [3, 13] 3 13
8 6] 1] 16 !6’ 16

[<?i> <?2] =

[L 1] 7 [14, 2] 7 r
6
8 6 T 16 8’ 8
[1, 1] -5 7 1

v =
[1, 1] [adj A]

86 _ 43
16 ” ¥

Alternatively,
all al2 <h
v = Ipi, P2]
a21 a22 <?2

7 -6
r_3_ 13t
L1 6’ 1 6 J
5 8
■i
r43 43-1 8
L 8’ 8 J 1
.8 .
_ 43
— S’

(as obtained above using algebraic methods).

EXAMPLE

10 -30
The game is nonstrictly determined.
-10 20

[L 1] [adj A]
[Po P2]
[1, l][adjA]

'20 30
[i, i;]
10 10 [30, 40] [3 4
'20 30 1 70 [7 ’ 7
[i, i]
10 10 1
650 / Applications of Matrix Algebra

[1, l][adj A]'

[4i> 42] =
[1, l][adj A] ]

20 10
[1,1]
30 10 [50, 20] 5 2
'20 30 r 70 7’7
[1,1]
10 10 1

A
v=
[1, 1] [adj A]

-100 _ 10
^ 70“ _T
Alternatively,

(as obtained above using algebraic methods).

2 x n Games and m x 2 Games

In 2 x n games and m x 2 games, one player has two strategies and the other player
has more than two strategies; the solution of a 2 x n game or an m x 2 game can be
reduced to the solution of a 2 x 2 subgame.
As in the solution of 2 x 2 games, the first step is to check for a saddle point if one
exists; the optimum (pure) strategies and the value of the game are thus determined.

EXAMPLE

The game 4 4
5 3
6 ©
1 3
5 4

has a saddle point at the intersection of the third row (player X’s optimal strategy) and the
second column (player Ts optimal strategy); the value of the game is 5.

EXAMPLE

The game 1 7 3
4 8 -1 6
 8.5 Game Theory / 651

has a saddle point at the intersection of the first row (player X's optimal strategy) and the third
column (player Y’s optimal strategy); the value of the game is zero.

Dominance
In an m x n matrix game, row i is said to majorize or to dominate row h if every entry
in row i is as large or larger than the corresponding entry in row h. Similarly, column
j is said to minorize or to dominate column k if every entry in column j is as small as or
smaller than the corresponding entry in column k. Note that any dominated (ma¬
jorized) row or dominated (minorized) column can be omitted from the matrix game
without affecting its solution, since such strategies are clearly not optimal. If a 2 x n
or an m x 2 game does not have a saddle point, that is, is not strictly determined, all
majorized rows and minorized columns should be eliminated as the next step in the
solution.
The solution of a 2 x n game consists of probabilities and p2 = 1 — with
which player X selects, at random, his strategies 1 and 2, respectively, and probabili¬
ties ql9 q2, qn, where Ya=i qt = 1, with which player Y selects, at random, his
strategies 1,2respectively. Similarly, the solution of an m x 2 game consists of
probabilities pl9 p2,..., pm, where i Pi = 1, for player X and probabilities ql and
q2 — 1 — q± for player Y.

EXAMPLE

The game z C
J

A -5
J

3 6
5 4
A A

has no saddle point. However, row 3 dominates row 1 and row 4 dominates rows 2 and 5. Thus
the game is reduced, for calculation, to the subgame

3 6
“TT"

EXAMPLE

The game -6 -1 <1 1 4 1

7 -2 (> 1
/
4 \ -5 f

has no saddle point. However, columns 3, 4, 5, and 7 dominate column 2. Thus the game is
reduced, for calculation, to the subgame

-6 -1 4
7 -2 -5

After dominance has been used to reduce a2xrcoranrax2 game for calcula¬
tion, all possible 2x2 games derived from the matrix of this reduced game can be
solved. The value of the original game is the value of one of these derived 2x2
652 / Applications of Matrix Algebra

games and the optimal strategies of the original game are also those of that derived
2x2 game, extended for one player by the addition of zeros. Which of the derived
games provides the solution of the original game can be determined either by trial
and error or graphically.
The trial-and-error procedure consists of solving the derived 2x2 games until
one is found for which the two-strategy player does at least as well (usually better)
against all his opponent’s other strategies as he does against the pair appearing in the
2x2 subgame. When such a game is determined, its solution provides the solution
of the original game.

EXAMPLE

In the first example above, only one 2x2 subgame remained

3 6
_5__4_

Its solution is p2 = f; q i = i, qi = and the solution of the 5x2 game is thus = 0,

Pi = 0, p3 = i p4 = j, p5 = 0; q! = i, q2 = For both games, v = f.

EXAMPLE

In the second example above, a 2 x 3 subgame remained

has the solution pj = A> P2 = Ai qi = A> = H; its value is — {J.

Against the other remaining strategy of player Y, column 3 of the reduced game, px = fi,
Pi = A has the value

A(4) + A(~ 5) = A
which is greater than — ff, so the solution of this 2x2 game extended is the solution of the
original game:

Pi = A Pi =A
<?i = A qi =A <?3 = <?4 = <?5 = <?6 = qi = o
note: The 2x2 game

-1 4
-2 -5

has a saddle point; thus p{ = 1, p2 = 0, and the value of the game is —1. Against the other
remaining strategy of player Y, column 1 of the reduced game, p1 = 1, p2 = 0 has the value
— 6, which is not greater than — 1, so the solution of this game is not the solution of the original
game. The 2x2 game

-6 4
7 -5
 8.5 Game Theory / 653

has the solution px = fi, P2 = rn and the value of the game is — yy Against the other remain¬
ing strategy of player Y, column 2 of the reduced game, = rr, P2 = tt has value — {f, which
is not greater than —yy so the solution of this game is not the solution of the original game.

Graphical Solution
Graphically, the 2x2 game whose solution is the solution of a 2 x n game is
determined as follows. Plot the pairs of payoffs of the n strategies of player Y on two
vertical axes and connect the pairs of points by straight lines; locate the highest point
on the line segments that form the lower boundary of the graph. The lines that
intersect at this point identify the strategies player Y should use in his optimum
strategy. Note that it may be possible to reduce a game by dominance before doing
graphical analysis.

EXAMPLE

In the game > -2 3 r 6

1
1 5 1 1 0

column 5 dominates column 4 and column 2 dominates column 1. The 2x3 game to be
solved is

-2 3 6
5 1 0

which can be solved by solving the 2x2 game

-2 3
5 1

(See Figure 8.5.)

Figure 8.5

The solution of this 2x2 subgame is px = p2 = f; qx = f, q2 = its value is J^z. Thus the
solution of the original game is p2 = I; q\ = 0, q2 = f, <?3 = l, q4 = 0, q5 = 0; its value

Similarly, for an m x 2 game, the payoffs of the m strategies of player X are plotted
and the lines that intersect at the lowest point on the line segments that form the
upper boundary of the figure identify the strategies player X should use in his
optimum strategy.
654 / Applications of Matrix Algebra

EXAMPLE

In the game -3 6
6 3
8 -2

the 2x2 game to be solved is (see Figure 8.6)

-3 6
6 3

The solution of this 2x2 subgame is px = i, p2 = J; <7i = i, 4 2 = 15 its value is Thus the
solution of the original game is px = i, P2 = i P3 = 0; = i, <72 = 15 its value is

Figure 8.6

EXAMPLE

An automobile manufacturer has five proposed designs for next year’s new cars. Which of
these is likely to sell best depends largely on whether his competitor’s standard model is
excellent, good, fair, or poor. If the model is excellent, his net profit (millions of dollars) will be
100, 150, 50, 125, and 90, respectively; if the model is good, his profit will be 80, 55, 55, 60, and
70, respectively; if the model is fair, his profit will be 150, 100,100,100, and 125, respectively; if
the model is poor, his profit will be 50, 80, 25, 80, and 75, respectively. What design should he
choose in order to maximize his minimum expected profit?

Competitor’s model

Excellent Good Fair Poor

1 1( X) 80 1 ! 50 50
2 1 i 50 55 1( K) 80
Design 3 i 50 55 1 ( K) 25
4 i: !5 60 1 )0( 80
5 5 '0 70 i: !5 75

The 2x2 game to be solved is (see Figure 8.7)

80 50
70 75

The solution of this 2x2 subgame is px = 7, p2 = 7 and qx = 7, <72 = -7* Thus the manufacturer
should produce model 1 with probability 7 and model 5 with probability 7; his competitor
 8.5 Game Theory / 655

Figure 8.7

should produce a good model with probability 7 and a poor model with probability 7; the
value is Note that the payoffs in this problem are all positive and the competitor must
therefore have some external reason (such as trying to enter the market) for manufacturing at
an inevitable loss. This is assuming that the matrix literally represents net profits and not some
type of additional revenue.

EXAMPLE

The Defense Department plans to award a contract for a new missile range at one of two
locations, A or B. A real estate speculator intends to invest $5000 in land—all at location A, all
at location B, or half at each location. If he buys at location A, the land will be worth $10,000 if
the missile range is built there, but $3000 if it is built at location B. If he buys at location B, the
land will be worth $4000 if the missile range is built at location A and $8000 if it is built at
location B. If he buys at both locations, the land will be worth $6000 if the missile range is built
at location A and $5000 if it is built at location B. In order to maximize his minimum expected
profits, what should the speculator do?

Missile range

A B

A 10,000 3,000
Investment B 4,000 8,000

A and B 6,000 5,000

The 2x2 subgame to be solved is (see Figure 8.8)

10,000 3,000
4,000 8,000

The solution of this 2x2 subgame is px = yj, p2 — it- Thus the speculator should invest his
money in property A with probability yj and property B with probability yx\ the value is
$6181.82. If the Defense Department is considered to be somehow plotting against the specula¬
tor, it should choose location A with probability yx and location B with probability yx.
656 / Applications of Matrix Algebra

Figure 8.8

Alternatively, the speculator can be considered to be choosing in order to protect himself

against the worst possible occurrence, whether that occurrence is in fact random or is the
result of Defense Department strategy against him.

SOLUTION OF LARGER GAMES

For 3 x 3 and larger games, if there is no saddle point and if the original game
cannot be reduced to a smaller game by dominance, linear programming provides
the most efficient method of solution. Linear programming is concerned with the
problem of maximizing or minimizing a linear function whose variables are restricted
to values satisfying a system of linear constraints. A matrix game can be expressed as
a problem of this type, since each player is concerned with maximizing the value of
the game to himself but is subject to the constraints imposed by the payoff matrix.
If, as above, the payoff matrix is represented by

0i i 012 ••• 0in“

021 a22 a2n
A =
a ml ®m2 ®mn

and Ts optimum strategies are represented by (q{, q2,..., qn), the inequalities which
express the expectations of Y are given by

0ntfi + «i202 + + 0inqn < v

02101 + «2202 + + a2nqn < v

0»il01 "h ^m2 02 "h "h ^mnQn < V

01 + 02 + " ‘ + 0n = 1
 8.5 Game Theory / 657

Substituting q{ = qjv,

^11^1 "b ^12^2 "b "b ^InQn — ^

fl2l?l + ^22^2 + •■' + < 1

^ml ^1 "b Um2^2 "b "b ®mnQn — ^

1
Qi + 0.2 + "' + On — ~

Player Fs objective is to minimize v or, equivalently, maximize 1/v. Thus the

solution of the matrix game can be written as a linear programming problem as
follows:
Maximize ql 4- q2 + ’’' + On
subject to

all<!l + al2Q2 + *“ + alnOn ^ 1

a2i0i + a22^2 + ••• + a2nqn < 1

amlOl + ^mlOl + "' + amnOn ^ 1

1
Oi + o2 + ' ■ ■ + on — ~
V

or, more concisely,

n

Maximize YOi
i= 1
subject to
n

Yj QijQj < 1 for i = 1, 2, ..., m

7=1
1

PROBLEMS
Find the optimal strategies for the players of each of the following matrix games using the
minimax criterion.

8 10 13 16 9 14 3 5 8 4
10 12 6 15 10 -3 -1 6 -2 5
16 18 9 13 25 2 12 10 13 16
4 9 18 20 6
15 -20 -12 3
5 8 7 4 2 -10 -6
-1 -3 10 20 -18 -15 -8
2 12 -6 -12 8 -10 6
10 9 -11 4
10 8 6 2
15 12 2 4 3 10 20 18 0
-4 6 -3 1 -9 -8 46 10 4
12 -2 8 -6 2 0 17 18 0
16 13 7 12 -1 1 13 5 3
658 / Applications of Matrix Algebra

Find the optimal strategies for the players of each of the following matrix games using the
minimax criterion and determine the value of each game.

17. A contractor is going to build a large number of houses for a housing development. Four
types of houses have been discussed: colonial, ranch, split-level, and contemporary; the
development committee will choose two of these types for the contractor to construct. The
contractor has the opportunity to buy materials in carload lots, thus saving considerable
money, but he must order in advance of the committee’s decision and can order only one
type of material. If the committee chooses colonial and ranch, the contractor will make an
(extra) profit (in thousands of dollars) of 125, 120, 60, and 50 if he orders colonial, ranch,
split-level, and contemporary materials, respectively. If the committee chooses colonial
and split-level, he will make 90, 40, 80, 75, respectively; if the committee chooses colonial
and contemporary, he will make 150, 30, 75, 100, respectively; if the committee chooses
ranch and split level, he will make 70, 70, 75, 65, respectively; if the committee chooses
ranch and contemporary, he will make 90, 80, 80, 120, respectively; if the committee
chooses split-level and contemporary, he will make 80, 40, 130, and 80, respectively. How
should the contractor order to maximize his minimum expected extra profit?
18. A plant manager must set up his reactors to produce a certain type of polythene using
either process 1, 2, 3, 4, or 5. Unfortunately, the chemical raw material varies in nitrogen
 8.5 Game Theory / 659

content and may contain 3, 4, 5, or 6% nitrogen. The nitrogen content affects the relative
efficiency of the five processes. With 3% nitrogen, the five processes have an output of 50,
45, 60, 50, and 30 tons, respectively. With 4% nitrogen, the outputs are 60, 70, 75, 90, and
60 tons, respectively; with 5% nitrogen, the outputs are 30, 55, 60, 45, and 70 tons,
respectively; with 6% nitrogen, the outputs are 45, 80, 80, 65, and 85 tons, respectively.
Testing for nitrogen content is too expensive to be practical. What process should the
plant manager use to maximize the minimum expected output?
19. A firm builds construction machinery whose performance depends on the reliability of a
gasoline motor. The firm can buy an expensive motor, fully guaranteed including replace¬
ment costs, for $500; the firm can also buy a moderately expensive motor for $400,
guaranteed for half its cost and the cost of replacement, so that a failure means the firm
pays $600 for the motor; or the firm can buy a cheaper motor for $300; if it buys this
cheaper motor it can either guarantee replacement for a total cost of $700, or it can pay
$50 to have the motor examined before installation, so total cost, if the motor is defective,
is $650. To minimize maximum expected costs, what should the firm do?
20. Mr. Smith and Mr. Jones each have a small hothouse in which they grow plants to sell.
Each of them grows either tomatoes, flowers, or strawberries in any particular year but
they are very secretive about their plans. If Mr. Smith grows tomatoes, his profit is $100 if
Mr. Jones grows tomatoes, $150 if Mr. Jones grows strawberries, and $200 if Mr. Jones
grows flowers. If Mr. Smith grows strawberries, his profit is $180, $125, $200 if Mr. Jones
grows tomatoes, strawberries, flowers, respectively. If Mr. Smith grows flowers, his profit is
$140, $125, $100 if Mr. Jones grows tomatoes, strawberries, flowers, respectively. For each
to maximize his minimum expected profit, what should Mr. Smith and Mr. Jones grow?
21. Mr. Reno has decided to bet $5 on a basketball game between Anglewood and Barleyville.
Mr. Las and Mr. Vegas have each offered Mr. Reno a bet. Mr. Las wants to bet on
Barleyville and is willing to pay Mr. Reno $10 if Anglewood wins and collect $5 if
Barleyville wins. Mr. Vegas wants to bet on Anglewood and is willing to pay Mr. Reno $4
if Barleyville wins and collect $5 if Anglewood wins. If he wants to maximize his minimum
expected winnings, what should Mr. Reno do?
22. Casey has a concession at the Yankee Stadium for the sale of sunglasses and umbrellas. He
has observed that he can sell about 500 umbrellas when it rains, and about 100 when it
shines; and in the latter case he can also dispose of 1000 sunglasses. Umbrellas cost $0.50
and sell for $1.00; glasses cost $0.20 and sell for $0.50. He is willing to invest $250 in the
project. Assuming that everything that isn’t sold is a total loss, what should he purchase to
maximise his minimum expected profit?
23. Mr. Hilton Conrad is faced with the following decision: The State Medical Society would
like to use his hotel for their annual convention; they would essentially take over all
facilities of the hotel (so business from other sources would be negligible) and would pay
$20,000 for this privilege. Ordinarily the hotel would do $10,000 business during the time
the convention would be in session; however, there is a chance that the World Series will
be played in the city at that time and, in this event, the hotel would do $50,000 business.
Should Mr. Hilton Conrad accept or decline the Medical Society’s offer in order to
maximize his minimum expected business?
24. An investor has a choice of buying one of three stocks: A, B, or C. The outcome of his
purchase depends on whether a particular company executes a merger, divests itself of a
subsidiary, or maintains the status quo. In case of a merger, stock A results in a gain of 20,
stock B in a loss of 25, and stock C in a gain of 12. In case of divesting, stock A results in a
loss of 5, stock B in a loss of 10, and stock C in a loss of 12. If the status quo is maintained,
stock A results in a gain of 5, stock B in a gain of 30, and stock C in a loss of 4 (amounts are
in thousands of dollars), a. In order to maximize his minimum expected gains, what stock
should the investor buy? b. If divesting were impossible, but the gains and losses from the
other possibilities were unchanged, what stock should he buy to maximize his minimum
expected gains?
25. A commuter must decide what to do about buying insurance for his car. He will definitely
carry liability insurance, but the car is rather old and he is not sure whether collision
insurance is worthwhile. He can either carry no collision insurance, buy a $50 deductible
660 / Applications of Matrix Algebra

policy for $60, or buy a full-coverage policy for $70. The commuter drives this car only to
and from work and he thinks one of three things will happen during the year: He will have
no accident, he will have one minor accident not exceeding $50 damage, or he will have
one more serious accident not exceeding $250 damage (the car is worth only about $400 so
this is not unreasonable optimism). The commuter realizes he could possibly have more
than one accident during the year, but on the basis of his past performances he is willing to
assume that he won’t. What should he do to minimize his maximum expected cost?
26. In the summer Mr. Smith is considering the winter coal problem. During a normal winter,
it takes about 15 tons of coal to heat his shop, but he has observed extremes where as little
as 10 tons or as much as 20 tons were used. The price per ton seems to fluctuate with the
weather, being $10, $15, and $20 a ton during mild, normal, and severe winters, respec¬
tively. He can buy now at $10 a ton. Mr. Smith considers three possible alternatives: to
buy 10, 15, or 20 tons now and the rest, if any, later. Assuming that all coal not used is a
total loss (Mr. Smith plans to sell the shop in the spring), what should he do to minimize
his maximum expected cost?
27. A student must decide how to study for a final exam in history. He studies differently for
true-false, multiple-choice, and essay exams, and he doesn’t know which type this exam
will be. The student thinks that if he studies for a true-false test, he will score 85 on a
true-false test, 80 on a multiple-choice test, and 75 on an essay test. If he studies for a
multiple-choice test, he will score 85 on a true-false test, 90 on a multiple-choice test, and
85 on an essay test. If he studies for an essay test, he will score 80 on a true-false test, 90 on
a multiple-choice test, and 90 on an essay test. To maximize his minimum expected score,
for what type of test should the student study?
28. A service station must have its service and parking area plowed after every heavy snowfall.
The manager can pay $5 every time the area needs plowing; he can buy a contract for $25
which provides blowing for up to and including six snowfalls, with each additional plow¬
ing at a cost of $3; or he can buy a contract for $30 which provides plowing as many times
as it is needed. What should the manager do to minimize his maximum expected cost of
plowing if he is willing to assume (based on past records) that there will be between 3 and 8
(inclusive) snowfalls during the winter?

Answers to Odd-Numbered Problems

1. Pl = p2 = 0, p3 = i Pa = 3
<h = i <72 = 0, q3 = 7, qA = q5 = 0
3. Pl = p2 = p3 = 0, p5 = jf
qi = qi = o, q3 = If, <?4 = A
5. Pi = 0, p2 = 1, p3 = Pa = P5 = 0
qi = q2 = 0, q3 = 1, qA = 0
7. Pl = p2 = o, p3 = 1
q\ = 0, q2 = 1, g3 = 0
v= 5
9- P\ = P2 = o, Pi = a, P4. = a
qi = 0, q2 = if, qi = if, =0
„V _
—
252
-zg-
11. Pi = p2 = p3 = 0, p4 = 1, p5 = 0
q\ = qi = qi = o, qA = 1, q5 = q6 = 0
V = 1

13. Pi = 0, p2 = i p3 = o, p4 = i p5 = 0
q\ = ?, <?2 = i
v = 3^
15. Pi = 0, p2 = i, p3 = 0, p4 = i
q\ = I, q2 = i
,, _ 3

17. buy for colonial with probability ^ and for split-level with probability
19. buy expensive motor
 8.6 First-Order Markov Processes / 661

21. bet Las with probability J and Vegas with probability f

23. accept offer of a convention
25. buy full coverage
27. study for multiple-choice test

8.6 First-Order Markov Processes

Sequences of observations or experimental outcomes are sometimes considered to be
independent; that is, the probability of observing any particular outcome is assumed
to be constant. The simplest generalization of this model permits the probability of
the outcome for any experiment or observation to depend on the outcome of the
immediately preceding observation, but not on the outcomes of other prior observa¬
tions. A process or sequence of this type is said to be a first-order Markov chain
process, first-order Markov chain, or first-order Markov process. First-order Markov
processes and their steady or equilibrium states are defined and illustrated in this
section.

DEFINITION OF A FIRST-ORDER MARKOV PROCESS

Suppose that the outcome of each of a sequence of experiments or observations is
one of a finite number of possible outcomes ax, a2, ar. The probability of
outcome aj for any given experiment or observation depends on at most the outcome
of the immediately preceding observation. These probabilities are denoted by pu,
i = 1, 2,..., r and j = 1, 2,..., r, where pu represents the probability of outcome a, for
any particular observation given that outcome at occurred for the immediately
preceding observation. The outcomes ax, a2, .ar are called states and the ptj are
called transition probabilities of a first-order Markov chain. If it is assumed that the
process begins in some particular state, the probabilities of various sequences of
observations can be calculated. Thus a first-order Markov chain is specified by
defining its possible states, specifying the initial probability distribution for these
states, and specifying the transition matrix.
The transition probabilities can be summarized in a square matrix. For a Markov
process with states au a2, ..., ar, the matrix of transition probabilities is

P11 P12 Plr

P21 P22 *'* P2r

p = {Pa) =
Prl Pr2 Prr

Note that the sum of the elements in each row of the matrix P is 1, since the elements
of the zth row represent the probabilities for all possible transitions when the process
is in state a,. That is,
r

I Pij =1 for i = 1, 2, ..., r

j= 1

Thus if the probability distribution of the states on trial n is [pls p2,..., pr], the
probability distribution of the states on trial n + 1 is

p 11 P12 ••• Plr "

P21 P22 P 2r r r r
• • • =
[Pi, P2 > •••> Pr] Zp.P.i X>iPi2 . • * • » Xi PiPir
i= 1 i= 1 i= 1

Prl Pr2 ••• Prr

662 / Applications of Matrix Algebra

STEADY STATE OR EQUILIBRIUM

The probability distribution of the outcomes for the nth observation of a first-order
Markov process is the product of the initial probability vector and the nth power of
the transition matrix. This can be shown by extending the arguments of the previous
section. If the vector of initial probabilities is denoted by p0 and the vector of
probabilities at step n is denoted by pn, then Pi=p0P, P2 = Pi P = PoP2>
P3 = P2P = PoP3> •••» P„ = PoP" Under some rather nonrestrictive mathematical
assumptions, it can be shown that a first-order Markov process approaches a steady
state or equilibrium as the number of observations increases, that is, as n approaches
infinity.
By definition, when a first-order Markov process is in equilibrium, the probability
of each possible state or outcome is constant from observation to observation. The
first-order Markov processes used to describe situations in business and economics
almost invariably fulfill the mathematical assumptions for equilibrium, and the prob¬
ability distribution of the states in equilibrium is frequently the most interesting
property of these processes.
It can be shown that if a first-order Markov chain approaches a steady or equili¬
brium state as the number of observations or transitions approaches infinity, the
stationary (equilibrium) probability distribution of its states is unique and depends
only on the transition matrix and not on the initial probability distribution of the
states. Any particular Markov chain is, of course, in one particular state for any given
observation. The physical and computational significance of stationarity thus be¬
comes apparent only if a number of Markov processes are in the same equilibrium
state simultaneously. In this case if there are N such processes and p, is the equili¬
brium probability of state i, p, N of the processes are expected to be in state i for any
given observation. Thus, if N is large, a state of macroscopic equilibrium is main¬
tained by a large number of transitions in opposite directions. Most statistical equilib¬
ria in physics are of this kind.
Computationally, the stationary probability distribution [p1? p2, ..., pr] of the
states of a Markov chain is obtained by solving the equations given in matrix form by

P11 P12 ••• Plr

P21 P22 ’ ’ ‘ P2r

[Pi? P2 » * • • > Pr] ~ lPb P2 > •••, Prl

Prl Pr2 ’ " Prr

where pt is the probability of being in state i, i = 1, ..., r and £j=1 pf = 1. This

involves solution of the r independent linear equations

X PiPu = Pj for j = 1, 2, ..., r (r — 1 of which are independent)

i= 1

I>;= 1
i= 1
Note that these equations actually define equilibrium: If [pls p2, ..., pr] is the
probability distribution of the states for a given observation, the product

P11 Pl2 ••• Plr

P21 P2 2 • ‘ • P2r
[P 1, Pi , - Pri

Prl Pr2 Prr

 8.6 First-Order Markov Processes / 663

gives the probability distribution of the states for the next observation. If that distrib¬
ution is also [pi, p2, •.., pr], the same as for the preceding observation, the system is
in equilibrium.

EXAMPLE

Suppose that, on September 1, of the subscribers in a fixed area, the Herald hasf, the Tribune
has i, and the Gazette has During the month of September, the Herald retains t of its
subscribers and loses f of them to the Tribune; the Tribune retains yj of its subscribers and
loses i of them to the Herald and t of them to the Gazette; the Gazette retains f of its
subscribers and loses f of them to the Herald and t of them to the Tribune. Assume there are
no new subscribers and that none quit subscribing.
(a) What proportion of the subscribers does each paper have on October 1 ?
(b) If the same pattern of gains and losses continues for October, what proportion of the
subscribers does each paper have on November 1 ?
(c) If the same pattern of gains and losses continues each month, what proportion of the
subscribers will each paper have in the long run (that is, in equilibrium)?
The transition matrix is

Herald Tribune Gazette

r 1
Herald 37 3 0

Tribune 33 1
TJ
1
E

Gazette L 1
1 E
1 1
3

(a)
r
37 31 0"
[i, i, i] 3
3 TZ
1 E1 = [f, i, i]
Li
1
E
1 31 J

Thus on October 1, of the subscribers, the Herald has f, the Tribune has -g, and the Gazette has
i-
(b)
1
3
1 1
TZ E = [ft, 3, fs]
1 1
E 3 J

Thus on November 1, of the subscribers, the Herald has ft, the Tribune has i, and the Gazette
has ye-
(c)
1
3
1
[Pi, Pi, Pa] T2 [Pi, Pi , P3]
1
E

ip 1 + IPi + ip 3 = Pi

3P1 + T2P2 + ip 3 = Pi

zPi + ip 3 = P3

Pi + P2 + P3 = 1
664 / Applications of Matrix Algebra

Multiplying the first equation by 8, the second equation by 24, and the third equation by 6,

~P i + 6 p2 + 4 p3 — 0

3pi - 22p2 + 4p3 = 0

Pi ~ 4p3 = 0

Pi + Pi + Pi — 1
Solving the first, third, and fourth equations simultaneously (any three of the four equations
are independent),

1 1 1 1

0 1 -4 0

_ - 1 6 4 0

'1 1 '
1 1

0 1 -4 0

_0 7 5 1 .

'1 0 5 1 "
0 1 -4 0

_0 0 33 1

"10 0
4
0 1 0 33
1
0 0 1 33 J
Thus in the long run, of the subscribers, the Herald will have f§, the Tribune will have-33, and
the Gazette will have 33.

EXAMPLE

In the community of Gardenville, each year 5 percent of the residents in the city proper move
to the suburbs and 2 percent of the people in the suburbs move to the city. Assuming that the
total number of people in the community remains constant, determine the long-run propor¬
tions of city and suburban residents.
The transition matrix is

City Suburbs

City 0.95 0.05

Suburbs 0.02 0.98

and the equilibrium probabilities are determined by

0.95 0.05
lPu Pi\ = [Pi, P2I
0.02 0.98

0.95 pi + 0.02p2 = Pi

0.05pj + 0.98 p2 = p2

Pi + Pi = 1
Multiplying the first two equations by 100,

-5 Pi + 2p2 = 0

5pi ~2p2=0

Pi + Pi = 1
 8.6 First-Order Markov Processes / 665

Solving the first and third equations simultaneously (any two of the three equations are
independent),

1 1 1

-5 2 0

1 1 1

0 7 5 .

1 1 1

0 1
5

2
1 0 7
5
0 1 7

Thus, eventually, of the people in the community, 7 will be city residents and 7 will be residents
of the suburbs. Note that in this equilibrium state each year 0.05 x 7 = ^ of the people move
from the city to the suburbs and 0.02 x 7 = 73 of the people move from the suburbs to the
city—the numbers of city and suburban residents thus are unchanged or stable.

EXAMPLE

A broker is studying the price movements of various stocks on the market and is particularly
interested in a company called Astronaut Instruments. He has observed that if this stock goes
up on a given day, then the next day it has a 50: 50 chance of going up again, a 3 chance of
staying the same price, and a 3 chance of going down. If the stock stays the same on a given
day, then it is equally likely to go up, stay the same, or go down the next day. If the stock
goes down on a given day, then the next day it has a 50 : 50 chance of going down again, a 3
chance of staying the same price, and a 3 chance of going up. What proportion of the time (in the
long run) does the stock go up, stay the same, and go down?
The transition matrix is

Up Same Down
1 1 1
Up 7 7 7

1 1 1
Same 7 7 7

1 1 1
Down 7 7 7

and the equilibrium probabilities are determined by

1 1 1
777
1 1 1
[p 1, p2, p3] 777 = [Pi, Pi , Pa]
1 1 1
L7 7 7

7P1 + hpi + 7P3 = Pi

7P1 + 7P2 + 7P3 = Pi

7Pl + 7P2 + 7P3 = P3

Pi + P2 + P3 = 1

Multiplying the first equation by 6, the second equation by 2, and the third equation by 6,

— 3pi + 2p2 + p3 = 0

Pi - 2p2 + p3 = 0

Pi + 2p2 — 3p3 = 0

Pi + P2 + P3 = 1
666 / Applications of Matrix Algebra

Solving the second, third, and fourth equations simultaneously (any three of the four equations
are independent),

" 1 1 1 1 "
1 2 -3 0
_ 1 -2 1 0.
" 1 1 1 r
0 1 4 -1
_0 -3 0 -1 _

" 1 0 5 2
0 1 -4 1
_0 0 - 12 -4

"1 0 0 1 1
3

0 1 0 1
3

_0 0 1 1
3 J
Thus, in the long run, the stock goes up 3 of the time, stays the same 3 of the time, and goes
down 3 of the time.

PROBLEMS
1. A country has a three-party political system and the results of elections follow a definite
pattern: If a party wins an election, its chance of winning the next election are 50 : 50 and if
it loses the next election, each of the other two parties has a 50 : 50 chance of winning.
What proportion of the elections does each party win over a long period of time?
2. A trucking company offers its drivers three approved routes between two cities: over the
Mystic Bridge, on the Interstate Parkway, and on Route 1. If a trucker goes on the Mystic
Bridge, the chance of his getting into a traffic jam is 3; if he does get in a traffic jam, the
next day he takes the Interstate Parkway with probability 3 and Route 1 with probability
3; if he doesn’t get in a traffic jam, the next day he takes the Mystic Bridge again with

probability 3, the Interstate Parkway with probability 4 and Route 1 with probability 3. If
he takes the Interstate Parkway the chance of his getting into a traffic jam is 3; if he does
get in a traffic jam, the next day he takes the Mystic Bridge; if he doesn’t get in a traffic
jam, the next day he takes the Mystic Bridge, the Interstate Parkway, and Route 1 with
equal probability. If he takes Route 1, the trucker is invariably late, so he never takes
Route 1 two days in a row and the next day he takes the Mystic Bridge with probability 3
and the Interstate Parkway with probability 3. What proportion of the time does the
trucker go on the Mystic Bridge, the Interstate Parkway, and Route 1, respectively?
3. Every year the Smith family goes on a vacation—a camping trip (preferred by the chil¬
dren), a visit to the city (preferred by Mrs. Smith), or a winter vacation (preferred by Mr.
Smith). They never take the same kind of vacation 2 years in a row. Each year they flip a
coin to decide which of the two types of vacation they did not take the previous year they
will take that year. What proportion of the time do the Smiths go on a camping trip, visit
the city, and take a winter vacation, respectively?
4. A mechanic has a very unreliable automobile. Every morning he goes out to the garage
hoping to start it; some days it starts by itself, some days it starts if he gets a push from a
neighbor, other days no amount of pushing is effective and the service station must be
called. If it starts one day, the chances are 3 it will start, 3 it must be pushed, and 3 the
service station must be called the next day. If it is pushed one day, the chances are 3 it will
start and 3 the service station must be called the next day (it is never pushed 2 days in a
row). If the service station is called one day, the chances are | it will start and \ it must be
pushed the next day (the service station is never needed 2 days in a row). In the long run,
what proportion of the days will the auto start, be pushed, and need the service station,
respectively?
 8.6 First-Order Markov Processes / 667

5. The students of Professor Geology never know what’s going to happen in class; the
professor may give them a surprise quiz, take them on a field trip, discuss the daily
assignment, or deliver a lecture. If he gives a quiz one day, the next day there is always a
field trip. If there is a field trip one day, there is never a quiz or a field trip the next day, but
discussing the assignment and a lecture are equally likely. If he discusses the assignment
one day, there is \ chance of a quiz, \ chance of a field trip, \ chance of discussing the
assignment, and 3 chance of a lecture the next day. If he lectures one day, there is \ chance
of a quiz, 3 chance of a field trip, 3 chance of discussing the assignment, and \ chance of a
lecture the next day. One of the students has calculated that they can expect quizzes of
the days, field trips 3^7 of the days, discussion of the assignment ^7 of the days, and
lectures 337 of the days. Is he correct?
6. Mr. S (who prefers steak), Mr. C (who prefers chicken), and Mr. H (who prefers ham) are

invited frequently to the home of a friend who serves only these three entrees. They have
decided to bet before each dinner engagement on which entree will be served. Suppose that
if the hostess serves steak on one occasion, she never serves it next time; she tosses a fair
coin and serves chicken if it comes up heads and ham if it comes up tails. If she serves
chicken on one occasion, then the next time she tosses two dice and serves chicken if the
dice come up the same; if the dice come up different on the first toss, she tosses them again
and serves steak if they come up the same and ham otherwise. If she serves ham on one
occasion, then the next time she draws a card randomly from a deck and serves chicken if
it is a diamond and steak otherwise. If each man always bets on his preference, what
proportion of the time, in the long run, will each win?
7. Every summer the Stormy Lakes Yachting Association must decide whether to hold its
annual regatta in June, July, or August. If the regatta is in June, the probability of good
weather is -J; if there is good weather, the next year the regatta will be held in June with
probability 3, in July with probability 4 and in August with probability i; if there is bad
weather, the next year the regatta will be held in July and August with equal probabilities.
If the regatta is in July, good and bad weather are equally probable; if there is good
weather, the next year the regatta will be held in July; if there is bad weather, the regatta
will be held in August with probability 3 and in June with probability 3. If the regatta is in
August, the probability of good weather is 3; if there is good weather, the next year the
regatta will be held in July and August with equal probabilities; if there is bad weather, the
regatta will be held in June with probability 3 and in July with probability 3. What
proportion of the time is the regatta held in June, July, and August, respectively?
8. A scientific book club has three mailing lists: a selected list consisting of members of 3

years or more of active membership, a membership list consisting of current members, and
an augmented list consisting of current members and others who have shown interest in
the club’s activities in the past. For each new publication, it must be decided which list to
use for mailing the sales literature. If after the preceding mailing there were too few copies
ordered, the secretary uses the member list with probability \ and the augmented list with
probability j; if after the preceding mailing there was a satisfactory response, the secretary
uses the member list and the augmented list with equal probability; if after the preceding
mailing, the publication was oversold, the secretary uses the selected list with probability 3
and the member list with probability 3. If the selected list is used, there are too few sold, or
a satisfactory number sold with equal probability; if the member list is used, there are too
few sold with probability \ and an adequate sale with probability |; if the augmented list is
used, there is an adequate sale with probability 3 and an oversale with probability 3. What
proportion of the time are the sales too small, adequate, and an oversale, respectively?
9. A housewife always buys one of three brands of detergent: A, B, or C. Which brand she
buys depends partly on which, if any, of the three companies are having a promotional
campaign (free combs, plastic roses, etc.). The companies time these campaigns at random,
paying no attention to whether the competition is running one at the same time. Company
A runs a campaign 3 of the time, company B runs a campaign 3 of the time, and company
C runs a campaign 3 of the time. If the housewife buys brand A one time, the next time she
buys brand A if brand A is running a campaign or if none of the brands is running a
campaign, she buys brand B if brand B is running a campaign but brand A is not, and she
buys brand C if brand C is running the only campaign at the time. If she buys brand B one
668 / Applications of Matrix Algebra

time, the next time she buys brand A if it is running the only campaign at the time and
brand B otherwise. If she buys brand C one time, the next time she buys brand A if it is
running the only campaign at the time, brand C if it is running a campaign and B is not,
and brand B otherwise. What proportion of the time does the housewife buy brands A, B,
and C, respectively?

helpful hint: From P(A) = P(B) = 3, P(C) = 3 given above, the following probabili¬
ties can be obtained with a little arithmetic:

P(A only) = | P(A and C, not B) = ?

P(B only) = ^ P(B and C, not A) = yg

P(C only) = i P(A, B and C) = A

P(A and £, not C) = i P(no campaign) = %

10. Computeronics Corporation orders 2-foot lengths of wire for use in its experimental
models from a supplier who claims that the wire has a specified strength desired by the
Computeronics engineers. Each shipment is classified by the engineers who use it as
satisfactory in strength (no report is made to the supplier), below standard (the supplier is
notified by letter that the shipment was not as claimed), or unacceptable (the shipment is
returned and a replacement which has been subjected to a 100% test is sent by the supplier
in accordance with the guarantee). The classification depends upon the extent to which a
shipment fulfills or fails to fulfill the supplier’s claim regarding strength. The engineers
have observed that if a shipment is classified as satisfactory, the next shipment is satisfac¬
tory 3 of the time, below standard ^5 of the time, and unacceptable ^ of the time; if a
shipment is classified as below standard, the next shipment is satisfactory of the time,
below standard ^ of the time, and never unacceptable. What proportion of the shipments
is satisfactory, below standard, and unacceptable in the long run, respectively?

Answers to Odd-Numbered Problems

Algebra
Review
Problems

EXPONENTS
I. PROBLEMS

Write each of the following in radical form.

1. k312 2. (ab)2'4 3. (m — n2)~1/2 4. 6//5 5. -4xy1/2 6. (xy)2'3

7. (x — y)5/6 8. (a + 5y)l/2 9. /c1/2 — m12 10. (4x — 6y)3/7 .
11 y“2'3
.
12 (a2 -b2)-112

Answers

1. yjk* 2. a3b3 3. ——_ 4. 6^^ 5. —4x%/y

^/m — n2

7. ^(x — y)5 8. ^/a + 5y 9. — ^fm 10. N/(4x — 6y)3

12. ,_
A“b2

II. PROBLEMS

Write each of the following with positive fractional exponents.

1
1. 2.
Vm - 1
7. x^a2b3> 8. v^x — 5,/y

669
670 / Algebra Review Problems

Answers

1. 43'2 2. 1-,„ 3. (5x2)’'3 = 51/3x2'3 4. (ab)113 5.x4'5 6.-^

(m - l)1'2 y

7. xa2'3b 8. x1'2 - 5y1'2 9. -2-ui 10. -L 11. (x - y)1'3

(x + yy12 x1M
— 6x
12. (lab2)115 = 71/5a1/5h2/5 13. 8m1/2 14. 15. -4(x7)1/5 = -4x3/5y1/5

III. PROBLEMS

Determine each of the following indicated roots.

1. ^32 2. y/l 3. 4. ~yflx2 5. -yi6 6. -Jiff

1. 8. y^Y 9. -vW 10. ^27a6

Answers
1. -2 2. 3 3. x2y 4. — 3x 5. -4 6. -a4y2 7. -3
8. jxy3 9. — a2b3 10. 3n2

IV. PROBLEMS
Change each of the following to simplest form.

1. 7300 2. 7Sxyy/5x 3. NJ2x5y2 4. —7300 5. yit4 6. —yin3

7. yy 8. -yy4 9. -yy io. yy 11. .Jab^a'b
i2. y?y i3. yy 14. yoo is. yoo 16. 70 7

Answers

1. 710073 = 1073 2. 725x7 - 725x77 = 5*7^ 3. 77772* - x^T^c

4. —710073 = —1073 5. k2 6. —nujm 7. 3x7* 8. — 3x2
9. ~7(4)(6)*6 = -2x7^ 10. 7(8)(2) = 4 11. 7^7 = a3b 12. T^4 = a
13. 757 = 3 14. 74(1°)2 = 20 15. 73(107 = 1073 16. a2yfay

V. PROBLEMS

Rationalize the denominator in each of the following.

— ab
1. 7.
y

Answers

j y _ yy _ y 2.
yy
yy = xy
‘ y yy 6 y
_ j_ .y = _y
y y 3 m

— ab^/b o >y y
7.
yy ■yy 2
 Algebra Review Problems / 671

VI. PROBLEMS

Simplify each of the following.

4l2
8 a%
1. (2a2b)(2ab2) 2. ( —3x3y4)(7y3xz) 3. (x1/2)(yx1/2) 4. 5.
4 a2b

12x3y3z3 48 x3,2y315
6. 8. (3x2)2 9. (3ah1/2)2 10. (x2/3y4/3)3
— 6 xyz 16x1/2y1/5

(x1/2 • y4)2 (x")3/2

11. (x2/3y4/3 )3/2 12. 13. rn/2
14. (/ • y*/2)4
(x2/3 • y)3
xn\ 1/2 / \ 3/2 ^mj^2m\ 1/m

15. (x2)"/2 • (y2n)2,n 16. 17. ,4m

18. x4(8a6x3)~1/3

4a3b (x3/"1j/2/m)m 6a 3 x2
19. 20. 21. 22. (a2/?-3)"1 23. x° • y"
(a4b6)112 xmy1_w 3(y2c 4)1/2

4 “VlT3 x~2y 1
2 1
24. 25. 26. a~1 + /T2 27. (n 1 + n 2) \ —

T2 aby4

'n1-flX“2
28. 29. (x)(x2/3) 30. n(n 3/5 — n 2/5)
<n 2 — a

31. (a + h)1/2[(a + b)1/2 — (a + h)] 32. a- 1/2(a3/2 — a + 6)

Answers

1. 4a3fr3 2. —21x4y7z 3. xy 4. x 3 = —t 5. 2a2h 6. — 2x2y2z2

x

7. 3xy215 8. 9x4 9. 9a2h 10. x2y4 11. xy2 12. r 13. x*

X

x3 4a
14. y6fc 15. x"y4 !7. ^ 18. 19.
16- X- b2 2a2

2c 2„2
x
20
3 — m ,,1 + m
. xJ_my 21. 22. 23. y" 24. 4a5b 25.
a3v a 3x 2,y,2

26. * + « 27. 28. n2 29. x5/3 30. n2/5 - n3/5

ah2 n+ 1

31. (a + b) — (a + b)3/2 32. a — a112 + 6a~ 1/2

LOGARITHMS
I. PROBLEMS

Write each of the following as the sum or difference of simpler logarithmic quantities.

/ ttiTi \ v3 (a\ /—
1. logJ —I 2. logx(ahc) 3. logJ-l 4. \ogxJy 5. loga n5

6. logfl x5y2 7. loga — 8 . log4 Va 9. log3 ^a(a - h)

v n

10 . log10^
672 / Algebra Review Problems

Answers

1. i(logfl m + logfl n - loga x) 2. logx a + logx b + logx c 3. log„ a - log„ b

4. ^ logx y 5. 5 logfl n 6. 5 loga x+ 2 loga 3; 7. \\oga m - \\oga n
8. ilog4 a + §log4 b 9. ilog3 a + ilog3 (a - b) 10. ilog10 a + flog10 b - ylog10 c

EXPONENTIAL FUNCTIONS
I. PROBLEMS
1. Find the second component of each of the following ordered pairs (x, y) if y = 2X:
( — 3, ), (0, ), (3, ), (6, ). 2. Find the second component of each of the following
ordered pairs (x,/(x)) if,f(x) = (j)*: ( — 3, ), (0, ), (3, ), (5, ). 3. State the equations

, 1
a = bx if b = 10 ,
e

Answers

1. y = 2-3 = i;y = 2°= 1; y = 23 = 8; y = 26 = 64

2. f(x) = (ir3 = 8; f(x) = (i)° = 1; /(*) = (i)3 = *; f(x) = (i)5 = *
3. a = 10x; a — ex; a = (ik))x = 10_JC; a = l-j - e~x

EXPONENTIAL AND LOGARITHMIC FUNCTIONS

I. PROBLEMS

Express each of the following exponential statements in logarithmic notation.

1. 52 = 25 2. 3~2=i 3. 161/4 = 2 4. 32 = 9 5. 23 = 8 6. (^)2 = i
7. 102 = 100 8. 103 = 1000 9. 10“1 = 0.1 10. 10° 1

Answers

1. log5 25 = 2 2. log3 ^ = -2 3. log16 2 = \ 4. log3 9 = 2

5. log2 8 = 3 6. log1/3 i = 2 7. logi0 100 = 2 8. log10 1000 = 3
9. log! o 0.1 = 10. log10 1 = 0

II. PROBLEMS

Express each of the following logarithmic statements in exponential notation.

1. log1/5 125 = -3 2. log4 16 = 2 3. log3 27 = 3 4. log6 36 = 2

5. log10 1000 = 3 6. log1/3 i = 2 7. log12 144 = 2 8. log, e = 1

Answers

1. (i)-3 = 125 2. 42 = 16 3. 33 = 27 4. 62 = 36 5. 103 = 1000

6. (i)2 = i 7. 122 = 144 8. c1 = c

III. PROBLEMS

Find the value of x, y, or z for each of the following.

1. log4 16 = x 2. log6 216 = y 3. logi0 1 = x 4. log4 4 = z

5. log!o 100 =y 6. logic 0-1 = y 7. logi0 0.01 = x 8. log41 = z
9. log3 27 = x 10. logj2 144 = y 11. log4 x = 2 12. log3 x = 2
13. log2 4 = y 14. log6 36 = z 15. logx 10 = \ 16. log3 27 = y
 Algebra Review Problems / 673

Answers

1. x = 2 2. y = 3 3. x = 0 4. z = 1 5. y = 2 6. y = — 1
1. x= -2 8. z = — 1 9. x = 3 10. y = 2 .
11 x= 16
12. x = 9 13. y = 2 14. z = 2 15. x = 100 16. y = 3
Tables

Exponential Functions

X ex e x X ex e x

0.00 1.000 1.000 3.00 20.086 0.050

0.10 1.105 0.905 3.10 22.198 0.045
0.20 1.221 0.819 3.20 24.533 0.041
0.30 1.350 0.741 3.30 27.113 0.037
0.40 1.492 0.670 3.40 29.964 0.033
0.50 1.649 0.607 3.50 33.115 0.030
0.60 1.822 0.549 3.60 36.598 0.027
0.70 2.014 0.497 3.70 40.447 0.025
0.80 2.226 0.449 3.80 44.701 0.022
0.90 2.460 0.407 3.90 49.402 0.020
1.00 2.718 0.368 4.00 54.598 0.018
1.10 3.004 0.333 4.10 60.340 0.017
1.20 3.320 0.301 4.20 66.686 0.015
1.30 3.669 0.273 4.30 73.700 0.014
1.40 4.055 0.247 4.40 81.451 0.012
1.50 4.482 0.223 4.50 90.017 0.011
1.60 4.953 0.202 4.60 99.484 0.010
1.70 5.474 0.183 4.70 109.95 0.009
1.80 6.050 0.165 4.80 121.51 0.008
1.90 6.686 0.150 4.90 134.29 0.007
2.00 7.389 0.135 5.00 148.41 0.007
2.10 8.166 0.122 5.10 164.02 0.006
2.20 9.025 0.111 5.20 181.27 0.006
2.30 9.974 0.100 5.30 200.34 0.005
2.40 11.023 0.091 5.40 221.41 0.005
2.50 12.182 0.082 5.50 244.69 0.004
2.60 13.464 0.074 5.60 270.43 0.004
2.70 14.880 0.067 5.70 298.87 0.003
2.80 16.445 0.061 5.80 330.30 0.003
2.90 18.174 0.055 5.90 365.04 0.003
3.00 20.086 0.050 6.00 403.43 0.002

675
676 / Tables

Common Logarithms

N 0 1 2 3 4 5 6 7 8 9

10 0000 0043 0086 0128 0170 0212 0253 0294 0334 0374
11 0414 0453 0492 0531 0569 0607 0645 0682 0719 0755
12 0792 0828 0864 0899 0934 0969 1004 1038 1072 1106
13 1139 1173 1206 1239 1271 1303 1335 1367 1399 1430
14 1461 1492 1523 1553 1584 1614 1644 1673 1703 1732

15 1761 1790 1818 1847 1875 1903 1931 1959 1987 2014
16 2041 2068 2095 2122 2148 2175 2201 2227 2253 2279
17 2304 2330 2355 2380 2405 2430 2455 2480 2504 2529
18 2553 2577 2601 2625 2648 2672 2695 2718 2742 2765
19 2788 2810 2833 2856 2878 2900 2923 2945 2967 2989

20 3010 3032 3054 3075 3096 3118 3139 3160 3181 3201
21 3222 3243 3263 3284 3304 3324 3345 3365 3385 3404
22 3424 3444 3464 3483 3502 3522 3541 3560 3579 3598
23 3617 3636 3655 3674 3692 3711 3729 3747 3766 3784
24 3802 3820 3838 3856 3874 3892 3909 3927 3945 3962

25 3979 3997 4014 4031 4048 4065 4082 4099 4116 4133
26 4150 4166 4183 4200 4216 4232 4249 4265 4281 4298
27 4314 4330 4346 4362 4378 4393 4409 4425 4440 4456
28 4472 4487 4502 4518 4533 4548 4564 4579 4594 4609
29 4624 4639 4654 4669 4683 4698 4713 4728 4742 4757

30 4771 4786 4800 4814 4829 4843 4857 4871 4886 4900
31 4914 4928 4942 4955 4969 4983 4997 5011 5024 5038
32 5051 5065 5079 5092 5105 5119 5132 5145 5159 5172
33 5185 5198 5211 5224 5237 5250 5263 5276 5289 5302
34 5315 5328 5340 5353 5366 5378 5391 5403 5416 5428

35 5441 5453 5465 5478 5490 5502 5514 5527 5539 5551
36 5563 5575 5587 5599 5611 5623 5635 5647 5658 5670
37 5682 5694 5705 5717 5729 5740 5752 5763 5775 5786
38 5798 5809 5821 5832 5843 5855 5866 5877 5888 5899
39 5911 5922 5933 5944 5955 5966 5977 5988 5999 6010

40 6021 6031 6042 6053 6064 6075 6085 6096 6107 6117
41 6128 6138 6149 6160 6170 6180 6191 6201 6212 6222
42 6232 6243 6253 6263 6274 6284 6294 6304 6314 6325
43 6335 6345 6355 6365 6375 6385 6395 6405 6415 6425
44 6435 6444 6454 6464 6474 6484 6493 6503 6513 6522

45 6532 6542 6551 6561 6571 6580 6590 6599 6609 6618
46 6628 6637 6646 6656 6665 6675 6684 6693 6702 6712
47 6721 6730 6739 6749 6758 6767 6776 6785 6694 6803
48 6812 6821 6830 6839 6848 6857 6866 6875 6884 6893
49 6902 6911 6920 6928 6937 6946 6955 6964 6972 6981

50 6990 6998 7007 7016 7024 7033 7042 7050 7059 7067
51 7076 7084 7093 7101 7110 7118 7126 7135 7143 7152
52 7160 7168 7177 7185 7193 7202 7210 7218 7226 7235
53 7243 7251 7259 7267 7275 7284 7292 7300 7308 7316
54 7324 7332 7340 7348 7356 7364 7372 7380 7388 7396
 Tables / 677

Common Logarithms

N 0 1 2 3 4 5 6 7 8 9

55 7404 7412 7419 7427 7435 7443 7451 7459 7466 7474
56 7482 7490 7497 7505 7513 7520 7528 7536 7543 7551
57 7559 7566 7574 7582 7589 7597 7604 7612 7619 7627
58 7634 7642 7649 7657 7664 7672 7679 7686 7694 7701
59 7709 7716 7723 7731 7738 7745 7752 7760 7767 7774

60 7782 7789 7796 7803 7810 7818 7825 7832 7839 7846
61 7853 7860 7868 7875 7882 7889 7896 7903 7910 7917
62 7924 7931 7938 7945 7952 7959 7966 7973 7980 7987
63 7993 8000 8007 8014 8021 8028 8035 8041 8048 8055
64 8062 8069 8075 8082 8089 8096 8102 8109 8116 8122

65 8129 8136 8142 8149 8156 8162 8169 8176 8182 8189
66 8195 8202 8209 8215 8222 8228 8235 8241 8248 8254
67 8261 8267 8274 8280 8287 8293 8299 8306 8312 8319
68 8325 8331 8338 8344 8351 8357 8363 8370 8376 8382
69 8388 8395 8401 8407 8414 8420 8426 8432 8439 8445

70 8451 8457 8463 8470 8476 8482 8488 8494 8500 8506
71 8513 8519 8525 8531 8537 8543 8549 8555 8561 8567
72 8573 8579 8585 8591 8597 8603 8609 8615 8621 8627
73 8633 8639 8645 8651 8657 8663 8669 8675 8681 8686
74 8692 8698 8704 8710 8716 8722 8727 8733 8739 8745

75 8751 8756 8762 8768 8774 8779 8785 8791 8797 8802
76 8808 8814 8820 8825 8831 8837 8842 8848 8854 8859
77 8865 8871 8876 8882 8887 8893 8899 8904 8910 8915
78 8921 8927 8932 8938 8943 8949 8954 8960 8965 8971
79 8976 8982 8987 8993 8998 9004 9009 9015 9020 9025

80 9031 9036 9042 9047 9053 9058 9063 9069 9074 9079
81 9085 9090 9096 9101 9106 9112 9117 9122 9128 9133
82 9138 9143 9149 9154 9159 9165 9170 9175 9180 9186
83 9191 9196 9201 9206 9212 9217 9222 9227 9232 9238
84 9243 9248 9253 9258 9263 9269 9274 9279 9284 9289

85 9294 9299 9304 9309 9315 9320 9325 9330 9335 9340
86 9345 9350 9355 9360 9365 9370 9375 9380 9385 9390
87 9395 9400 9405 9410 9415 9420 9425 9430 9435 9440
88 9445 9450 9455 9460 9465 9469 9474 9479 9484 9489
89 9494 9499 9504 9509 9513 9518 9523 9528 9533 9538

90 9542 9547 9552 9557 9562 9566 9571 9576 9581 9586
91 9590 9595 9600 9605 9609 9614 9619 9624 9628 9633
92 9638 9643 9647 9652 9657 9661 9666 9671 9675 9680
93 9685 9689 9694 9699 9703 9708 9713 9717 9722 9727
94 9731 9736 9741 9745 9750 9754 9759 9763 9768 9773

95 9777 9782 9786 9791 9795 9800 9805 9809 9814 9818
96 9823 9827 9832 9836 9841 9845 9850 9854 9859 9863
97 9868 9872 9877 9881 9886 9890 9894 9899 9903 9908
98 9912 9917 9921 9926 9930 9934 9939 9943 9948 9952
99 9956 9961 9965 9969 9974 9978 9983 9987 9991 9996
,
Selected
References

MATHEMATICS

Andree, Richard V. Introduction to Calculus with Analytic Geometry. New York: McGraw-
Hill, 1962.
Birkhoff, G., and MacLane, S. A Survey of Modern Algebra. 3d ed. New York: Macmillan,
1965.
Britton, J. R., Kriegh, R. B., and Rutland, Leon W. Calculus and Analytic Geometry. San
Francisco, Calif.: Freeman, 1966.
Buck, R. C. Advanced Calculus. 2d ed. New York: McGraw-Hill, 1965.
Burkill, J. C. The Theory of Ordinary Differential Equations. New York: Wiley, Halstead Press,
1956.
Coddington, E. A., and Levinson, N. Theory of Ordinary Differential Equations. New York:
McGraw-Hill, 1955.
Courant, R. Differential and Integral Calculus. Translated by E. J. McShane. Vols. I and II.
New York: Interscience, 1955 printing.
Cullen, Charles. Matrices and Linear Transformations. 2d ed. Reading, Mass.: Addison-Wesley,
1972.
Garvin, W. W. Introduction to Linear Programming. New York: McGraw-Hill, 1960.
Glicksman, A. M. Linear Programming and The Theory of Games. New York: Wiley, 1963.
Goldberg, S. Introduction to Difference Equations. New York: Wiley, 1958.
Golomb, M., and Shanks, M. Elements of Ordinary Differential Equations. 2d ed. New York:
McGraw-Hill, 1965.
Hadley, G. Linear Algebra. Reading, Mass.: Addison-Wesley, 1961.
Hadley, G. Linear Programming. Reading, Mass.: Addison-Wesley, 1962.
Ince, E. L. Ordinary Differential Equations. New York: Dover, 1953.
Jolley, L. B. W. Summation of Series, 2d ed. New York: Dover, 1961.
Kaplan, W. Advanced Calculus. 2d ed. Reading, Mass.: Addison-Wesley, 1973.
Kemeny, J. G. Snell, J. L., and Thompson, G. L. Finite Markov Chains. Englewood Cliffs, N.J.:
Prentice-Hall, 1960.
Kuhn, H. W., and Tucker, A. W., eds. Linear Inequalities and Related Systems. Princeton, N.J.:
Princeton University Press, 1956.
Lang, S. A First Course in Calculus. Reading, Mass.: Addison-Wesley, 1964.
Llewellyn, R. W. Linear Programming. New York: Holt, Rinehart & Winston, 1964.
Luce, R. D., and RaifTa, H. Games and Decisions. New York: Wiley, 1957.
McKinsey, J. C. C. Introduction to the Theory of Games. New York: McGraw-Hill, 1952.
Mirsky, L. An Introduction to Linear Algebra. 2d ed. New York: Oxford University Press,
1961.
679
680 / Selected References

Nelson, A. L., Folley, K. W., and Coral, M. Differential Equations. 3d ed. Boston: D. C. Heath,
1964.
Nemhauser, G. L. Introduction to Dynamic Programming. New York: Wiley, 1966.
Niven, Ivan. Calculus, An Introductory Approach. 2d ed. Princeton, N.J.: D. Van Nostrand,
1966.
Orchard-Hays, W. Advanced Linear Programming Computer Techniques. New York: McGraw-
Hill, 1968.
Perlis, S. Theory of Matrices. Reading, Mass.: Addison-Wesley, 1952.
Peterson, T. S. Elements of Calculus. 2d ed. New York: Harper & Row, 1960.
Richardson, C. H. An Introduction to the Calculus of Finite Differences. New York: Van
Nostrand, Reinhold, 1954.
Sasaki, Kyohei. Introduction to Finite Mathematics and Linear Programming. Belmont, Calif.:
Wadsworth, 1970.
Sisam, C. H., and Atchison, W. F. Analytic Geometry. New York: Holt, Rinehart & Winston,
1955.
Smail, L. L. Calculus. New York: Appleton-Century-Crofts, 1949.
Spitzbart, A., and Bardell, R. H. College Algebra and Plane Trigonometry. 2d. ed. Reading,
Mass.: Addison-Wesley, 1964.
Steen, F. H., and Ballou, D. H. Analytic Geometry. 3d ed. Lexington, Mass.: Xerox College
Publishing, 1963.
Stoll, R. F. Linear Algebra and Matrix Theory. New York: Dover, 1969.
Thomas, G. B., Jr. Calculus and Analytic Geometry. 4th ed. Reading, Mass.: Addison-Wesley,
1968.
Thrall, R. M., and Tornheim, L. Vector Spaces and Matrices. New York: Dover, 1970.
Whipkey, K. L., and Whipkey, M. N. The Power of Calculus. 2d ed. New York: Wiley, 1975.
Widder, D. V. Advanced Calculus. 2d ed. Englewood Cliffs, N.J.: Prentice-Hall, 1961.
Wooton, W., and Drooyan, I. Intermediate Algebra. 3d ed. Belmont, Calif.: Wadsworth, 1972.
Wylie, C. R., Jr. Calculus. New York: McGraw-Hill, 1953.

APPLICATIONS OF MATHEMATICS
IN BUSINESS AND ECONOMICS

Arrow, K., and others. Studies in Linear and Non-Linear Programming. Stanford, Calif.: Stan¬
ford University Press, 1958.
Arrow, K., and Karlin, S. Studies in the Mathematical Theory of Inventory and Production.
Stanford, Calif.: Stanford University Press, 1958.
Beatty, William E. Mathematical Relationships in Business and Economics. Morristown, N.J.:
General Learning Corporation, 1970. '
Beatty, William E. and Gage, R. William. Introductory Calculus for Business and Economics.
Morristown, N.J.: General Learning Corporation, 1973.
Bierman, H., Bonini, C. P., Jr., and Hausman, W. H. Quantitative Analysis for Business Deci¬
sions. 4th ed. Homewood, Ill.: Irwin, 1973.
Bowen, E. K. Mathematics—With Applications in Management and Economics. 3d ed. Home-
wood, Ill.: Irwin, 1972.
Buffa, E. S. Production Inventory System: Planning and Control. Rev. ed. Homewood, Ill.:
Irwin, 1972.
Carr, C. R., and Howe, C. W. Introduction to Quantitative Decision Procedures in Management
and Economics. New York: McGraw-Hill, 1964.
Chiang, Alpha C. Fundamental Methods of Mathematical Economics. New York: McGraw-
Hill, 1974.
Churchman, C. W., Ackoff, R. L., and Arnoff, E. L. Introduction to Operations Research. New
York: Wiley, 1957.
Daus, P. H., and Whyburn, W. M. Algebra With Applications to Business and Economics.
Reading, Mass.: Addison-Wesley, 1961.
Daus, P. H., and Whyburn, W. H. Introduction to Mathematical Analysis. Reading, Mass.:
Addison-Wesley, 1958.
Dorfman, R., Samuelson, P. A., and Solow, R. Linear Programming and Economic Analysis.
New York: McGraw-Hill, 1958.
 Selected References / 681

Fowler, F. P., Jr., and Sandberg, E. W. Basic Mathematics for Administration. New York:
Wiley, 1962.
Horowitz, I. An Introduction to Quantitative Business Analysis. 2d ed. New York: McGraw-
Hill, 1972.
Horst, P. Matrix Algebra for Social Scientists. New York: Holt, Rinehart & Winston, 1963.
Howell, J. E., and Teichroew, D. Mathematical Analysis for Business Decisions. Homewood,
Ill.: Irwin, 1963.
Intriligator, M. D. Mathematical Optimization and Economic Theory. Englewood Cliffs, N.J.:
Prentice-Hall, 1971.
Karlin, S. Mathematical Methods and Theory in Games, Programming and Economics. Reading,
Mass.: Addison-Wesley, 1959.
Kemeny, J. G., Snell, J. L., and Thompson, G. L. Introduction to Finite Mathematics. 2d ed.
Englewood Cliffs, N.J.: Prentice-Hall, 1966.
Kooros, A. Elements of Mathematical Economics. Boston: Houghton Mifflin, 1965.
Levin, R. I., and Kirkpatrick, C. A. Quantitative Approaches to Management. 3d ed. New York:
McGraw-Hill, 1971.
Naddor, E. Inventory Systems. New York: Wiley, 1966.
Sasieni, M., Yaspan, A., and Friedman, L. Operations Research—Methods and Problems. New
York: Wiley, 1959.
Teichroew, D. An Introduction to Management Science, Deterministic Models. New York:
Wiley, 1964.
Tintner, G., and Millham, C. B. Mathematics and Statistics for Economists. 2d ed. New York:
Holt, Rinehart & Winston, 1970.

ECONOMICS
Ackley, G. Macroeconomic Theory. New York: Macmillan, 1961.
Allen, C. L. Elementary Mathematics of Price Theory. Belmont, Calif.: Wadsworth, 1962.
Baumol, W. J. Economic Dynamics. 3d ed. New York: Macmillan, 1970.
Beach, E. F. Economic Models. New York: Wiley, 1957.
Boulding, K. E. Economic Analysis, 4th ed. New York: Harper & Row, 1966.
Cyert, R. M., and March, J. G. A Behavioral Theory of the Firm. Englewood Cliffs, N.J.:
Prentice-Hall, 1963.
Dernburg, T. F., and Dernburg, J. D. Macroeconomic Analysis. Reading, Mass.: Addison-
Wesley, 1969.
Dernburg, T. F., and McDougall, D. M. Macroeconomics. New York: McGraw-Hill, 1968.
Gale, David. The Theory of Linear Economic Models. New York: McGraw-Hill, 1960.
Henderson, J. M., and Quandt, R. E. Microeconomic Theory, A Mathematical Approach, 2d ed.
New York: McGraw-Hill, 1971.
Hibdon, James E. Price and Welfare Theory. New York: McGraw-Hill, 1969.
Keynes, J. M. The General Theory of Employment Interest and Money. New York: Harcourt,
Brace, Jovanovich, 1935.
Neumann, J. von, and Morgenstern, O. Theory of Games and Economic Behavior. 3d ed.
Princeton, N.J.: Princeton University Press, 1954.
Samuelson, P. A. Foundations of Economic Analysis. Cambridge, Mass.: Harvard University
Press, 1947.
Samuelson, P. A. Economics: Introductory Analysis, 9th ed. New York: McGraw-Hill, 1973.
Schultz, H. The Theory and Measurement of Demand. Chicago: University of Chicago Press,
1957.
Shapiro, Edward. Macroeconomics: Selected Readings. New York: Harcourt, Brace and
World, 1970.
Stonier, A. W., and Hague, D. C. A Textbook of Economic Theory. 4th ed. New York: Wiley,
Halstead Press, 1973.
.

'
Index

Abscissa, 16. See also 607. See also Hessian Compound interest, 118, 521.
Rectangular coordinates determinant; Maximum See also Exponential
Adjoint of a matrix, 566, 577, of a function of n function; Interest; Simple
647-650. See also variables interest
Determinant of a matrix; Boundary conditions. See also Concave function, 611
Inverse of a matrix; Constant of integration; Concavity, 231-236, 240. See
Matrix Initial conditions; also Convexity; Point of
Algebraic function, 50, 174-177, Particular solution inflection; Second
201, 317-318. See also for a difference equation, 512 derivative
Function for a differential equation, 472 concave downward, 231-232,
Alternating series, 299-300. See for integration, 389 240
also Infinite series Break-even analysis, 44-46 concave upward, 231-232, 240
Angle of inclination, 19-20 Conic section, 71-90. See also
Antiderivative, 388 Capital formation, 397-398. See Circle; Ellipse;
Approximate integration, 449- also Integration; Net Hyperbola; Parabola;
458. See also Integration, investment Quadratic equation
approximate methods of Cauchy’s test ratio, 299-300. circular, 72-74
Approximation, 204-206. See See also Convergence of degenerate, 71, 73, 75, 79, 83
also Differential an infinite series directrix of, 71
Area. See also Integration; Proba¬ Circle, 72-74. See also Conic eccentricity of, 71
bility beneath a curve, section elliptic, 72, 74-76
404-408, 414-417 center of, 72 foci of, 71
between two curves, 408-413 radius of, 72-73 hyperbolic, 72, 81-90
negative, 407-408 standard form of, 72-73 parabolic, 72, 77-81
Asymptote, 57-59 Cobweb model, 524 standard forms for, 72-73, 74-
horizontal, 58-59 Cofactor of a matrix, 566. See 75, 77-79, 82-83, 86-87
vertical, 58-59 also Determinant of a Constant. See also Constant of
Average cost, 168, 245-246, 394- matrix; Matrix integration
395. See also Cost Column vector, 542. See also absolute, 6
Average revenue, 170, 266-267, Vector arbitrary, 6
395-396. See also Competitive commodities, 341— numerical, 6
Revenue 344. See also parametric, 6, 33
Complementary Constant of integration, 388-
Base e, 114-115. See also commodities; Demand 389. See also Boundary
Logarithm; Logarithmic surface conditions; Initial
function Complementary commodities, conditions; Integration
Bernoulli equations, 498 341-344. See also Constant product curve, 349-
Bilinear form, 597-598. See also Competive commodities; 351. See also Equal
Vector differentiation Demand surface product curve; Isoquant;
Bivariate regression analysis, Completing the square, 72, 134— Production function
363-364. See also Least 137 Constrained optima, 367-383,
squares estimates; Composite function, 10, 196-197, 607-613. See also Kuhn-
Regression analysis 202, 321, 334-335. Sec Tucker conditions;
Bordered Hessian determinant, also Function Lagrange multipliers

683
684 / Index

Constraints. See also linear, 246-247 first, 161-168, 219-230, 240-

Constrained optima; polynomial, 251-252 244
Linear programming quadratic, 247-248 geometric interpretation of,
in the dual problem, 631-632 Cotangent, 125. See also 161-162, 219-221
in the primal problem, Trigonometric function higher-order, 207-212
620-621, 624-625, 629 Cramer’s Rule, 589-590. See also partial, 324-328
Consumers’ surplus, 418-422. Simultaneous linear rules for obtaining, 174-202,
See also Producers’ equations, solution of 317-321
surplus Curve. See also Family, of algebraic function, 174-177,
Consumption function, 46-48, curves; Function 201, 317-318
172-174, 506-507. See algebraic, 50 composite function, 196—
also Consumption model asymptotes of, 57-59 197, 202, 321, 334-
marginal propensity to exponential, 111-114 - 335
consume, 172, 396-397 higher plane, 50 exponential function, 187—
marginal propensity to save, intercepts of, 51 190, 202, 319
172, 396-397 logarithmic, 114-117 implicit function, 213-215,
multiplier, 172 nonlinear, 50-133 335-338
Consumption model, 506-507, polynomial, 50 inverse function, 194-195,
525-526. See also transcendental, 50 202, 320-321
Consumption function logarithmic function, 185—
Continuity, 151-159, 324. See 187, 202, 318-319
also Discontinuity; Decreasing function, 219-220, polynomial function, 177,
Smooth function 230. See also First 201, 317
and differentiability, 217-219 derivative; Function trigonometric function, 191—
in an interval, 151 Definite integral, 399-417. See 193, 202, 319-320
at a point, 151 also Integration second, 231-244
Convergence of an infinite series, geometric interpretation of, total, 332-335
298-299. See also Infinite 399-401 Descartes, Rene, 16
series; Interval of Degenerate conic, 71, 73, 75, 79, Determinant of a matrix, 564—
convergence; Maclaurin’s 83. See also Conic 569. See also Cramer’s
series; Power series; section; Quadratic Rule; Inverse of a
Radius of convergence; equation matrix; Simultaneous
Taylor’s series Degree. See also Difference linear equations
absolute, 299-300 equation; Differential rules for obtaining
conditional, 299-300 equation; Equation diagrammatic, 565
tests for of an equation expansion by cofactors,
alternating series, 299-300 algebraic, 18, 50 565-566
Cauchy ratio, 299-300 difference, 511 Diagonal matrix, 551-552. See
comparison series, 299-300 differential, 470 also Identity matrix
necessary condition, 299- of a polynomial, 50 Difference equation, 509-540.
300 of a term, 18, 50 See also Cobweb model;
Convergence of an integral, 402 of a variable, 18 Consumption model;
Convex function, 611 Degree measure, 124-128. See Harrod model; Metzler
Coordinates of a point, 16. See also Trigonometric inventory model;
also Polar coordinates; function v Samuelson’s interaction
Rectangular coordinates Demand function, 37-40, 43-44, model
abscissa, 17 92-98, 170, 257-265, classification of
ordinate, 17 341-344. See also degree, 510-511
polar, 129-131 Demand surface; Market order, 510-511
Cosine, 125, 127, 130. See also equilibrium; Supply general solution of, 512-514
Trigonometric function function linear, 511-512
Cost, 44-46, 168-169, 245-254, linear, 37-40, 43-44 particular solution of, 512-514
340-341, 394-395, 425- quadratic, 92-98 solution of linear first order,
430, 605-606. See also Demand surface, 341-344. See 515-518
Cost function also Competitive solution sequence, behavior
average, 168, 245-246, 394- commodities; of, 518-521;
395 Complementary equilibrium of, 522;
fixed, 44-46 commodities; Partial stability of, 522;
joint, 340-341 demand; Partial stationary value of,
marginal, 168-169, 245-246, differentiation 522
340-341, 394-395 Dependent variable, 8-9, 141, solution of linear second
total, 44-46, 168, 245-246, 324. See also Function; order, 528-531
394-395, 605 Variable nonhomogeneous, 533-535
variable, 44-46 Derivative. See also solution sequence, behavior
Cost function, 44-46, 168-169, Differentiable function; of, 531-532;
245-254, 425-430, Differentiation; equilibrium of, 535—
605-606. See also Cost Maximum; Minimum; 536; stability of,
cubic, 248-250 Point of inflection; Rate 535-536; stationary
exponential, 252-254 of change of a function value of, 535-536
 Index / 685

Differentiability and continuity, infinite, 152 Ellipse, 72, 74-76

217-219 missing point, 152 center of, 74
Differentiable function, 162. Discontinuous function, 151- major axis of, 74-75
See also Differential; 159. See also minor axis of, 74-75
Differentiation Discontinuity; Discrete standard form of, 74-75
Differential, 203-206. See also function Endogenous variable, 502. See
Derivative; Discrete function, 155-159. See also Economic model;
Differentiation also Discontinuity; Exogenous variable
partial, 330-331 Discontinuous function Equal product curve, 349-351.
total, 330-331 Distribution, marginal See also Constant
of x, 203-204 productivity theory of, product curve; Isoquant
of y, 203-204 346-349 Equation. See also Curve; Dif¬
Differential equation, 470-507. Distribution of income, Pareto’s ference equation; Differ¬
See also Consumption law, 107-109 ential equation; Function;
model; Domar debt Disturbance term, 363, 598. See Simultaneous linear
models; Domar macro also Regression analysis equations
model; Evans price Domain. See also Function; graphical representation of,
adjustment model Relation 16-133
classification of, 470-471 of a function, 8 locus of, 18
degree, 470 of a relation, 8 Equations. See also Economic
order, 470 Domar debt models, 503-505 model; Simultaneous
of first order and first degree, Domar macro model, 502- linear equations
476-498 503 consistent, 27-28
exact, 477, 487-491 Dominance, 651-653. See also independent, 27-28
homogeneous, 477, 480-484 Game theory; Matrix reduced form, 502
linear, 477, 491-495 game; Minimax solution; simultaneous linear, 27-32,
linear in a function of x or Saddle point 589-592
y, 477, 495-498; Double integral, 459-460. See structural, 502
Bernoulli equations, also Integration; Multiple Equilibrium. See also Demand
498 integral; Probability function; Difference
separable, 477-480 Dual, 631-635. See also Linear equation; Market
general solution of, 471-475 programming; Primal equilibrium; Stationary
of higher order and/or degree, value; Supply function
498-499 of a difference equation, 522,
ordinary, 470 535-536
partial, 470 in purely competitive markets,
particular solution of, 471— Eccentricity of a conic, 71. See 42-44, 93-98
475 also Conic section Error, 204—206. See also
Differentiation, rules for, 174— Economic model, 502-507, 537— Approximation;
202, 332-338. See also 540. See also Cobweb Differential equation
Derivative; Differentiable model; Consumption percentage, 206
function; Implicit model; Domar debt relative, 206
differentiation models; Domar macro Euler’s theorem, 348-349. See
algebraic function, 174-177, model; Evans price also Marginal
201, 317-318 adjustment model; productivity theory of
composite function, 196-197, Harrod model; Metzler distribution; Partial
202, 321, 334-335 inventory model; differentiation
exponential function, 187-190, Samuelson’s interaction Evans price adjustment model,
202, 319 model; Simultaneous 505-506
implicit function, 213-215, linear equations Exact differential equation, 477,
335-338 dynamic, 502 487-491
inverse function, 194-195, static, 502 Exogenous variable, 502. See
202, 320-321 Elasticity, 171, 256-265, 266- also Economic model;
logarithmic function, 185-187, 271, 344-345. See also Endogenous variable
202, 318-319 Logarithmic derivative; Expansion path, 351. See also
polynomial function, 177, 201, Marginal analysis, of Returns to outlay;
317 revenue; Partial Returns to scale; Ridge
summary of, 201-202 differentiation line; Scale line
summary of procedure of, arc, 256, 258-259 Explicit function, 351. See also
199-200 constant, 263-265 Function
trigonometric function, 191— cross, 258-263 Exponential cost function, 252-
193, 202, 319-320 of demand, 266-271, 344-350 254. See also Cost
Directrix of a conic, 71. See partial, 344-350 function
also Conic section point, 171, 256-258 Exponential function, 111-114,
Discontinuity, 151-159. See also Elementary operations. See also 117-122, 187-190, 202,
Continuity; Determinant of a matrix; 252-254, 319. See also
Discontinuous function; Inverse of a matrix Function; Power
Discrete function on columns, 574 function; Scientific
finite, 152 on rows, 574 notation
686 / Index

Extent, 56-57. See also inverse, 13-14, 55, 117, 194— Harrod model, 523-524
Graphing, general 195, 202, 320-321 Hessian determinant, 605, 607.
methods of linear, 17-25 See also Bordered
logarithmic, 114-117, 185— Hessian determinant;
187, 202, 318-319 Maximum of a function
Factorization, 60-61, 62. See also nonlinear, 50-133 of n variables
Graphing, general polynomial, 50, 177, 201, 317 Higher-order derivative, 207-
methods of power, 108 212. See also Derivative
Fair game, 641. See also Matrix quadratic, 70-90 Homogeneous differential
game; Value of a game range of, 8 equation, 477, 480-484.
Family. See also Parameter smooth, 218-219 See also Differential
of curves, 90 step, 155-159 equation
of straight lines, 33-36 transcendental, 50 Homogeneous function, 348-349,
Finite discontinuity, 152. See trigonometric, 123-131, 133, 477, 480-484
also Discontinuity 191-193, 202 Hyperbola, 72, 81-90. See also
Finite game, 640. See also Game value of, 9 Conic section
theory; Matrix game Function value, 9. See also asymptotes of, 82-83, 86
First derivative, 161-230. See Function center of, 81
also Derivative; equilateral, 82, 86-90
Maximum; Minimum generalized equilateral, 87-90
interpretation of, 165-168 Game matrix, 640-641. See also of Fermat, 87
First difference, 510. See also Dominance; Game rectangular, 82, 86-90
Difference equation theory; Matrix game; standard form of, 82-83, 86
Foci of a conic, 71. See also Minimax solution; Payoff transverse axis of, 81
Conic section matrix; Saddle point;
Frequency function, 414-417, Strategies; Value of a Identity, trigonometric, 129. See
464-467. See also Area; game also Trigonometric
Joint frequency function; Game theory, 639-657. See also function
Probability Dominance; Game Identity matrix, 552-553. See
Function, 8-10, 323-324. See matrix; Matrix game; also Diagonal matrix;
also Consumption Minimax solution; Payoff Inverse of a matrix
function; Cost function; matrix; Saddle point; Imaginary locus, 61-62. See also
Curve; Demand function; Strategies; Value of a Locus
Growth function; game Implicit differentiation, 213-215.
Joint-cost function; Gaussian elimination, 574-577. See also Differentiation
Production function; See also Inverse of a Implicit function, 213-215, 335—
Rate of change of a matrix 338. See also Function;
function; Relation; General solution. See also Implicit differentiation
Revenue function; Constant of integration; Implicit prices, 635. See also
Supply function; Particular solution Dual; Imputed prices;
Variable of a difference equation, Shadow prices
algebraic, 50, 174-177, 201, 512-514 Improper integral, 402. See also
317-318 of a differential equation, Convergence;
composite, 10, 196-197, 202, 471-475 Divergence; Integral
321, 334-335 Generic constant, 472. See also Imputed prices, 635. See also
concave, 611 Constant of integration; Dual; Implicit prices;
continuous, 151, 159 Differential equation Shadow prices
convex, 611 Gompertz function, 120-121. Income, 46-48, 172-174, 396-
decreasing, 219-220, 230 See also Function; 397, 506-507, 525, 526.
dependent variable of, 8-9, 141, Growth function See also Consumption
324 Graph, 18. See also Equation; function; Consumption
differentiable, 162 Locus model
discontinuous, 151-159 Graphical solution of a matrix Increasing function, 219-220,
discrete, 155-159 game, 653-656. See also 230. See also First
domain of, 8 Game theory; Matrix derivative; Function
explicit, 213 game; Minimax solution Indefinite integral, 388-398.
exponential, 111-114, 117-122, Graphing. See also Asymptote; See also Constant of
187-190, 202, 252-254, Extent; Factorization; integration; Integration
319 Intercepts; Maximum; interpretation of, 388-389
frequency, 414-417, 464-467 Minimum; Point of Independent variable, 8-9, 141,
Gompertz, 120-121 inflection; Symmetry 324. See also Function;
growth, 119-122 general methods of, 50-70 Variable
homogeneous, 348-349, 477, using first and second Indeterminate expression, 7, 20.
480-484 derivatives, 219-244 See also Indeterminate
implicit, 213-215, 335-338 Growth function, 119-122. See form; Undefined
increasing, 219-220, 230 also Curve; Function expression
independent variable of, 8-9, biological, 119-120 Indeterminate form, 148-149,
141, 324 Gompertz, 120-121 289-296. See also
integrable, 400 learning, 121-122 Discontinuity;
 Index / 687

Indeterminate form (Continued) Simpson’s rule, 449, 451- Kuhn-Tucker conditions, 375—
Indeterminate expression; 453, 455-458 380, 611-613. See also
L’Hospital’s rule; Limit Taylor’s expansion, 449, Constrained optima
Type & 148-149, 289-290, 296 453-458
Type §, 148-149, 289-290, 296 trapezoidal rule, 449-451,
Type oo-0, 291-295, 296 455-458 Lagrange multipliers, 368-375,
Type 1®, 291-295, 296 definite, 399-429 607-611. See also
Type 0°, 291-295, 296 double, 459-461, 464-467 Constrained optima
Type oo°, 291-295, 296 fundamental theorem of, Law of diminishing marginal
Type oo — oo, 295, 296 404 productivity, 346
Indifference curve, 353. See also indefinite, 388-392 Least squares estimates, 363-364,
Indifference map; Utility multiple, 459-463 598-599. See also
Indifference map, 353. See also partial, 459 Regression analysis
Indifference curve; by partial fractions, 438-442 Leibnitz, Gottfried, 139
Utility by parts, 435-437 Leontief matrix, 615. See also
Infinite discontinuity, 152. See by rationalization, 443-447 Input-output analysis
also Discontinuity by standard forms, 390-392, L’Hospital’s rule, 289-296. See
Infinite games, 640. See also 431-434 also Indeterminate form;
Game theory; Matrix by substitutions, 443-448 Limit
game Intercepts Limit, 141-149. See also
Infinite sequence, 298. See also of a curve, 51 Asymptote; Convergence;
Difference equation, of a straight line, 18-19 Divergence;
solution sequence of; Interest. See also Compound Indeterminate form;
Infinite series; interest; Simple interest L’Hospital’s rule
Recursion formula compound, 118, 521 of a function, 141-149
Infinite series, 298-315. See also simple, 521 left-hand, 144-146
Convergence of an infinite Interval of convergence, 304. See properties of, 146-149
series; Infinite sequence; also Infinite series; Power right-hand, 144-146
Maclaurin’s series; series; Radius of of a sequence, 298
Recursion formula; convergence of a series, 298
Taylor’s series Inventory models, 285-288, of a variable, 141-142
alternating, 299-300 539-540 Limits of integration, 402. See
convergent, 298-299 Inverse of a function, 13-14, also Definite integral
divergent, 298-299 55, 117, 194-195, 202, Linear dependence, 585-588.
oscillating, 298, 520 320-321. See also See also Rank;
Inflection, point of, 236-239, 240. Function Simultaneous linear
See also Second Inverse of a matrix, 569-584. See equations
derivative also Determinant of a Linear difference equation, 511—
Initial conditions. See also matrix; Elementary 540. See also Difference
Boundary conditions; operations; Game equation
Constant of integration; matrix; Gaussian Linear differential equation, 477,
Particular solution elimination; Identity 491-498. See also
for a difference equation, 512 matrix; Input-output Differential equation
for a differential equation, analysis; Linear Linear equation, 17-25. See also
472 programming; Markov Linear function;
for integration, 389 chain process; Simultaneous linear
Inner product of two vectors, Simultaneous linear equations; Straight line
548. See also Vector equations, solutions of Linear function, 17-25. See also
Input-output analysis, 614-618 by adjoint and determinant, Function; Linear
Integrable function, 400. See 577-579 equation; Maximization;
also Integration by canonical procedure using Minimization; Straight
Integral, 388. See also row or column line; Vector
Integration operations, 574-577 differentiation
definite, 402 of a partitioned matrix, Linear programming, 619-635.
double, 459-461, 464-467 580-582 See also Constrained
improper, 402 Inverse of a relation, 13-14. See optima; Inverse of a
indefinite, 388 also Relation matrix; Simplex method
multiple, 459-463 Isocost curve, 351 artificial variables of, 625-626,
particular, 388 Isoquant, 349-351. See also 629, 633
Integrand, 388. See also Constant product curve; constraints of, 620-621, 624-
Integration Equal product curve 625, 629, 631-632
Integrating factor, 491-492. See dual of, 631-635
also Differential equation feasible solution of, 621, 624-
Integration, 387-467. See also Joint-cost function, 340-341. See 625
Constant of integration; also Cost function geometric solution of, 621-625
Integral; Multiple Joint frequency function, 414— of a matrix game, 656-657
integration 417, 464-467. See also objective function of, 620-621,
approximate methods of, 449- Area; Frequency 631-632
458 function; Probability primal of, 631-632
688 / Index

Linear programming See also Inverse of a Maximum. See also Kuhn-

(Continued) matrix; Matrix Tucker conditions;
primary variables of, 620 equilibrium of, 662-666 Lagrange multipliers
simplex solution of, 625-631 nonstationary, 662 of a function of n variables
slack variables of, 625-626, states of, 661 constrained, 607-613
629, 633 stationary, 662 unconstrained, 602-606
Locus, 18. See also Equation; steady state of, 662-666 of a function of one variable
Graph; Imaginary locus; transition matrix of, 661 absolute, 228-230
Point locus transition probabilities of, 661 relative, 220-228, 230, 235
Logarithm, 114-115. See also Mathematical programming, of a function of two variables
Base e; Logarithmic 620. See also Linear constrained, 367-383
function programming unconstrained, 356-366
Briggsian, 114 Matrix, 541-542. See also Metzler inventory model, 539—
common, 114 Determinant of a matrix; 540
Naperian, 114 Elementary operations; Minimax solution, 639. See also
natural, 114 Inverse of a matrix; Dominance; Game
Logarithmic derivative, 257. Matrix operations theory; Matrix game;
See also Elasticity; adjoint of, 566, 577, 647-650 Maximin solution;
Logarithm cofactors of, 566 Mixed strategy; Saddle
Logarithmic function, 114-117, determinant of, 564-569 point
117-122, 185-187, 202. diagonal, 551-552 Minimum
See also Function; elements of, 541 of a function of n variables
Logarithm idempotent, 555 constrained, 607-613
identity, 552-553 unconstrained, 602-606
Maclaurin’s series, 308. See also inverse of, 569-584 of a function of one variable
Infinite series; Taylor’s left, 570 absolute, 228-230
series right, 570 relative, 220-228, 230, 235
Majorize, 651. See also minors of, 566 of a function of two variables
Dominance nonsingular, 570 constrained, 367-383
Marginal analysis. See also null, 553-554 unconstrained, 356-366
First derivative order of, 541-542 Minorize, 651. See also
of consumption, 172, 396-397 partitioned, 558-562, 580-582 Dominance
of cost, 168-169, 245-246, principal minors of, 603 Minors of a matrix, 566. See
340-341, 394-395 rank of, 585-588 also Matrix; Principal
of demand, 341-344 singular, 570 minors of a matrix
of production, 346-349, 605 square, 541 Missing-point discontinuity, 152.
of revenue, 170, 266-271, symmetric, 555 See also Discontinuity
395-396, 425-429 transpose of, 554-557, 558- Mixed partial derivative, 327.
of utility, 352-354, 380-383 559 See also Partial
Marginal product, 346-349, 602. Matrix game, 639-657. See also derivative; Second
See also Euler’s theorem; Game matrix; Game derivative
Marginal productivity theory; Minimax Mixed strategy, 644. See also
theory of distribution; solution; Mixed strategy; Dominance; Game
Production function Saddle point theory; Matrix game;
Marginal productivity theory of dominance in, 651-653 Minimax solution;
distribution, 349. See also graphical solution of, 653-656 Saddle point
Euler’s theorem; linear programming solution Monopoly
Marginal product of, 656-657 profit under, 275-279
Marginal propensity to consume, saddle point of, 641-642 taxation under, 280-284
172-174, 396-397. See solution of 2 x 2 games, Morganstern, Oskar, 639
also Consumption 642-650 Multiple integral, 459-463. See
function; Consumption solution of 2 x n and m x 2 also Integral; Multiple
model games, 650-656 integration
Marginal propensity to save, solution of larger games, Multiple integration, 459-463.
172-174, 396-397. See 656-657 See also Integration;
also Consumption Matrix operations, 544-551, Multiple integral
function; Consumption 558-562. See also Multiplier, 172. See also
model Elementary operations; Consumption function;
Marginal rate of substitution, Inner product of two Consumption model
354. See also Equal vectors; Matrix Multivariate regression analysis,
product curve addition, 544-545, 559-560 598-600. See also Least
Marginal utility, 352-354, multiplication, 546-551, 560- squares estimates;
380-383. See also Utility 562 Regression analysis
function multiplication by a scalar,
Market equilibrium, 42-44, 545-546 Napier, John, 114
93-98. See also Demand subtraction, 544-545, 559-560 Net investment, 397-398. See
function; Supply Maximin solution, 639. See also also Capital formation;
function Game theory; Matrix Integration
Markov chain process, 661-666. game; Minimax solution Newton, Sir Isaac, 139
 Index / 689

Nonlinear curve, 50-133. See Initial conditions general equation of, 70-71
also Curve; Function of a difference equation, 512—
Nonsingular matrix, 570. See 514 Radian measure, 124-128. See
also Determinant of a of a differential equation, 471— also Trigonometric
matrix; Inverse of a 475 function
matrix; Singular matrix Payoff matrix, 640-641. See also Radius of convergence, 304. See
Nonstrictly determined game, Game matrix; Matrix also Infinite series;
643-645. See also Game game Interval of convergence;
matrix; Matrix game; Period analysis, 509. See also Power series
Minimax solution; Mixed Difference equation Range. See also Function;
strategy Point locus, 61-62. See also Relation; Variable
Non-zero-sum game, 640. See Locus of a function, 8
also Game theory; Point of inflection, 236-244. See of a relation, 8
Matrix game; Zero-sum also Second derivative of a variable, 6
game Polar coordinates, 129-131. See Rank of a matrix, 585-588. See
Null matrix, 553-554. See also also Coordinates of a also Linear dependence;
Matrix point; Function; Matrix; Simultaneous
Numbers Trigonometric function linear equations
complex, 7 Polynomial, 50. See also Rate of change of a function. See
imaginary, 7 Polynomial function also Derivative
real, 7 Polynomial function, 50, 177, average, 167-168
201, 317. See also instantaneous, 167-168
Objective function, 620, 631-632. Function, polynomial Rate of technical substitution,
See also Linear Postmultiply, 547. See also 349. See also Equal
programming Matrix operations, product curve
Operator, 510 multiplication; Reciprocal substitution, 447-448.
Ordinate, 17. See also Premultiply See also Integration, by
Rectangular coordinates Power function. 111. See also substitution
Oscillating series, 298, 520. See Exponential function; Rectangular coordinates, 16-17.
also Difference equation, Function See also Coordinates of a
solution sequence of; Power series, 304-315. See also point; Polar coordinates
Infinite series Infinite series; axes of, 16-17
Maclaurin’s series; origin of, 16
Parabola, 72, 77-81. See also Taylor’s series plotting points in, 17
Conic section Premultiply, 547. See also quadrants of, 16-17
axis of, 77 Matrix operations, Recursion formula, 298. See also
orientation of, 77-79 multiplication; Infinite sequence; Infinite
standard form of, 77-79 Postmultiply series
vertex of, 77 Primal, 631-632. See also Reduced form equations, 502.
Parameter, 6, 33. See also Dual; Linear See also Economic model;
Constant, parametric; programming Structural equations
Family, of curves Principal minors of a matrix, Regression analysis, 363-364,
Pareto, Vilfredo, 107 603. See also Matrix; 598-599. See also
Pareto’s law, 107-109. See also Minors of a matrix Bivariate regression
Power function; Scientific Probability, 414-417, 464-467. analysis; Least squares
notation See also Area; Double estimates; Multivariate
Partial demand, 341-344. See integral; Frequency regression analysis
also Demand surface function; Integral Relation, 8. See also Function;
Partial derivative, 324-328. See Producers’ surplus, 418, 422-425. Variable
also Derivative; Mixed See also Consumers’ binary, 8
partial derivative; surplus dependent variable of, 8
Partial differentiation Production function, 346-349, domain of, 8
Partial differential, 330-331. 351-352, 605. See also independent variable of, 8
See also Differential Euler’s theorem; inverse, 8
Partial differentiation, 324-328. Function; Marginal range of, 8
See also Differentiation; product Relative maximum, 220-228,
Partial derivative linear homogeneous, 348-349 230, 235. See also
Partial elasticities of demand, Product transformation curve, Maximum
344-345. See also 98-104 Relative minimum, 220-228,
Demand surface; Properties of limits, 146-149. 230, 235. See also
Elasticity; Partial See also Indeterminate Minimum
differentiation form; L'Hospital’s rule; Returns to outlay, 352. See also
Particular integral, 388. See also Limit Equal product curve;
Boundary conditions; Expansion path; Returns
Constant of integration; Quadratic form, 595-597. See to scale; Ridge line; Scale
Initial conditions also Vector differentiation line
Particular solution. See also Quadratic function, 70-90. See Returns to scale, 351-352. See
Boundary conditions; also Conic section; also Equal product curve;
Constant of integration; Function Expansion path; Returns
690 / Index

Returns to scale (Continued) Simplex method, 625-631. See process, 661. See also
to outlay; Ridge line; also Inverse of a matrix; Markov chain process
Scale line Linear programming Stationary probability
Revenue, from taxation, 271-275. steps in solution by, 626 distribution, 662. See also
See also Revenue function tableau of, 625-626 Markov chain process
Revenue, 44-46, 170-171, 266- variations in, 631 Stationary value, 535-536. See
271, 395-396, 425-429, Simpson’s rule, 449, 451-453, also Difference equation;
605. See also Demand; 455-458. See also Equilibrium
Elasticity Integration, approximate of a difference equation,
average, 170, 266-267, 395-396 methods of 535-536
and elasticity of demand, 266- Simultaneous linear equations, Steady state, 662-666 See also
271 26-32, 585-592. See also Equilibrium; Markov
marginal, 170, 266-271, 395- Economic model; Linear chain process
396 dependence; Rank Step function, 155-159. See also
total, 170, 266, 395-396, 605 consistent, 27, 28, 587-588 Function
Ridge line, 350. See also Equal dependent, 27, 28, 585-588 Straight line. See also Linear
product curve; Expansion inconsistent, 27, 28, 587-588 equation; Straight lines
path; Returns to outlay; independent, 27, 28, 585-588 general equation of, 18, 25
Returns to scale; Scale solution of, 28-32, 589-592 horizontal, 25
line by elimination, 29-32 intercept form of, 22-23, 25
by matrix algebra, 29-32, intercepts of, 18-19
Saddle point, 357-358, 641-642. 589-592; Cramer’s rule point-slope form of, 23-24, 25
See also Game matrix; 589-590; inversion slope-intercept form of, 24,
Matrix game; Maximum; procedures, 589 25
Minimax solution; by substitution 29-32 two-point form of, 21-22, 25
Minimum; Second Sine, 125, 127, 130. See also vertical, 24, 25
derivative Trigonometric function Straight lines, 17-36. See also
Samuelson’s interaction model, Singular matrix, 570. See also Simultaneous linear
537-538 Determinant of a matrix; equations; Straight line
Scalar, 545. See also Matrix Inverse of a matrix; families of, 33-35
Scale line, 351-352. See also Nonsingular matrix intersecting, 27-32
Equal product curve; Slack variables, 625-626, 629, parallel, 26
Expansion path; Returns 633. See also Linear perpendicular, 26'
to outlay; Returns to programming; Simplex Strategies, 640. See also Game
scale; Ridge line method matrix; Matrix game;
Scientific notation, 112 Slope of a line, 19-20, 161. See Minimax solution; Saddle
Secant, 125, 161-162. See also also First derivative; point
Trigonometric function Rate of change of a mixed, 641-642
Second derivative, 207, 231-244. function; Tangent to a pure, 644-646
See also Concavity; curve Strictly determined game, 641.
Derivative; Partial Smooth function, 218-219. See See also Game matrix;
derivative; Point of also Continuity; Matrix game; Minimax
inflection Function; Second solution; Saddle point
interpretation of, 231-232 derivative Structural equations, 502. See
Separable differential equation, Solution. See also Difference also Economic model;
477-480. See also equation; Differential Reduced form equations
Differential equation equation; Equation; Supply function, 40-42, 43-44,
Sequence, 298. See also Infinite Simultaneous linear 92-98, 170, 257-265. See
sequence equations also Demand function;
Series, 298-315. See also Infinite of a difference equation Market equilibrium
series general, 512-514 linear, 40-42, 43-44
Set, 2-5. See also Venn Diagram particular, 512-514 quadratic, 92-98
complement, 4 of a differential equation Symmetric matrix, 555. See also
defining relation of, 2 general, 471-475 Matrix
difference, 4 particular, 471-475 Symmetry
element of, 2 of simultaneous linear with respect to an axis, 51-55
empty, 3 equations, 28-32, 589- with respect to the origin,
intersection, 3 592 51-55
proper subset, 3 Square matrix, 541. See also with respect to a point, 55-56
subset, 3 Matrix with respect to a straight line,
union, 3 Standard forms for conics, 72- 55-56
universal, 4 73, 74-75, 77-79, 82-83, with respect to the x-axis,
universe, 4 86-87. See also Circle; 51-55
Shadow prices, 635. See also Conic section; Ellipse; with respect to the y-axis,
Dual; Implicit prices; Hyperbola; Parabola 51-55
Imputed prices Standard forms for integration, Tangent of an angle, 19-21, 125,
Simple interest, 521. See also 390-392, 431-434. Sec 127, 130, 161, 204, 326.
Compound interest; also Integration See also Angle of
Interest States of a Markov chain inclination; First
 Index / 691

Tangent of an angle (Continued) 458. See also Integration, game; Minimax solution;
derivative; Slope of a approximate methods of Strategies
line; Trigonometric Trigonometric function, 123-131, Variable, 6-7, 8. See also
function 133, 191-193, 202. See Function; Relation
Tangent to a curve, 162. See also Function continuous, 6
also First derivative; cosecant, 125-129 dependent, 8-9, 141, 324
Rate of change of a cosine, 125-129 discrete, 6
function; Slope of a line identities, 129 independent, 8-9, 141, 324
Taxation polar coordinates, 129-131 range of, 6
maximization of monopolistic radian measure of, 124-125 Vector, 542-543. See also Inner
profit under, 280-284 secant, 125-129 product of two vectors;
maximization of revenue from, sine, 125-129 Matrix
271-275 tangent, 125-129, 326 column, 542
Taylor’s series, 308-315. See also components of, 542-543
Infinite series; dimension of, 542-543
Maclaurin’s series of functions, 595
Taylor’s theorem, 308-315, 321- Undefined expression, 7, 20. See row, 543
322, See also Taylor’s also Indeterminate Vector differentiation, 593-600.
series expression; Indeterminate See also Maximization;
Technological matrix, 615. See form Minimization; Partial
also Input-output Undetermined multipliers, 368- differentiation; Vector
analysis; Leontief matrix 375, 607-611. See also of a bilinear form, 597-598
Total derivative, 332-335. See Lagrange multipliers of a linear function, 593-595
also Derivative Utility function, 352-354, of a quadratic form, 595-597
Total differential, 330-331. See 610-611. See also of a vector of functions, 595
also Differential Marginal analysis, of Venn Diagram, 4. See also Set
Transcendental function, 50. See utility Von Neumann, John, 639
also Function constrained maximization of,
Transition probabilities, 661. See 380-383
also Markov chain
process Zero-sum game, 640. See also
Transpose of a matrix, 554-557, Game theory; Matrix
558-559. See also Matrix Value of a game, 641, 645. See game; Non-zero-sum
Trapezoidal rule, 449-451, 455- also Game matrix; Matrix game
 1

'

81 82 83 9 8 7
SBN 06-041762-