Matrix Algebra

Exercise Book

ˆ Practice Problems
ˆ Concept Summaries
ˆ Very Detailed Solution To Each Problem

by Paul Tsopméné

Last Update: December 1, 2022

Matrix Algebra Exercise Book by Paul Tsopméné is licensed under a Creative

Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-
SA 4.0).
ii

About the Author

ˆ Title: Assistant Professor of Teaching
ˆ Since: September 2020
ˆ Institution: University of British Columbia Okanagan
ˆ Department: CMPS (Computer Science, Mathematics, Physics
and Statistics)
ˆ Email: paul.tsopmene@ubc.ca

Source File. To get the source file, please get in touch with the author
at the following email address: paul.tsopmene@ubc.ca
 iii

For Instructors
Dear instructors,

Thank you for your interest in this open exercise book. I hope you find it a
valuable addition to your existing teaching material.

I would love to hear your feedback and adoption decision for your course.
Your feedback will be highly appreciated and will be used for future im-
provement of the book.

Please take a few minutes to fill in the survey, which can be accessed in the
following way:

Survey Link

Thank you for your time and support on this open exercise book develop-
ment.

Best Regards,

Paul Tsopméné

This page and survey is adapted from Introduction to Engineering Thermodynamics by

Dr. Claire Yan which is licensed under a Creative Commons Attribution-NonCommercial-
ShareAlike 4.0 International License, except where otherwise noted.
iv
 Contents

Preface vii

I Exercises 1
1 Systems of Linear Equations 3
1.1 Introduction to Systems of Linear Equations . . . . . . . . . . . . . . . . . 3
1.2 Echelon Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Solving Linear Systems Using Matrices . . . . . . . . . . . . . . . . . . . . 8

2 Vectors in Rn and Linear Systems 11

2.1 Vector Equations and Span . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 The Matrix Equation Ax = b . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Writing the General Solution of a Linear System in Vector Form . . . . . . 16
2.4 Linear Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Linear Transformations 25
3.1 Introduction to Linear Transformations . . . . . . . . . . . . . . . . . . . . 25
3.2 The Matrix of a Linear Transformation . . . . . . . . . . . . . . . . . . . . 28
3.3 One-To-One and Onto Linear Transformations . . . . . . . . . . . . . . . . 31

4 Matrix Algebra 35
4.1 Matrix Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 The Inverse of a Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5 Subspaces of Rn 45
5.1 Subspaces, Column and Null Spaces . . . . . . . . . . . . . . . . . . . . . . 45
5.2 Basis, Dimension, and Rank . . . . . . . . . . . . . . . . . . . . . . . . . . 50

6 Determinants 55
6.1 Introduction to Determinants . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2 Properties of Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

7 Eigenvalues and Eigenvectors 61

7.1 Eigenvectors and Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.2 Finding Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.3 Diagonalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

v
vi Contents

II Solutions to Exercises 67
1 Systems of Linear Equations 69
1.1 Introduction to Systems of Linear Equations . . . . . . . . . . . . . . . . . 69
1.2 Echelon Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1.3 Solving Linear Systems Using Matrices . . . . . . . . . . . . . . . . . . . . 95

2 Vectors in Rn and Linear Systems 111

2.1 Vector Equations and Span . . . . . . . . . . . . . . . . . . . . . . . . . . 111
2.2 The Matrix Equation Ax = b . . . . . . . . . . . . . . . . . . . . . . . . . 127
2.3 Writing the General Solution of a Linear System in Vector Form . . . . . . 136
2.4 Linear Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

3 Linear Transformations 167

3.1 Introduction to Linear Transformations . . . . . . . . . . . . . . . . . . . . 167
3.2 The Matrix of a Linear Transformation . . . . . . . . . . . . . . . . . . . . 182
3.3 One-To-One and Onto Linear Transformations . . . . . . . . . . . . . . . . 199

4 Matrix Algebra 209

4.1 Matrix Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
4.2 The Inverse of a Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

5 Subspaces of Rn 245
5.1 Subspaces, Column and Null Spaces . . . . . . . . . . . . . . . . . . . . . . 245
5.2 Basis, Dimension, and Rank . . . . . . . . . . . . . . . . . . . . . . . . . . 261

6 Determinants 285
6.1 Introduction to Determinants . . . . . . . . . . . . . . . . . . . . . . . . . 285
6.2 Properties of Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

7 Eigenvectors and Eigenvalues 305

7.1 Eigenvectors and Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . 305
7.2 Finding Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
7.3 Diagonalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
 Preface

The author wrote this exercise book when teaching MATH 221 (Matrix Algebra) at the
University of British Columbia Okanagan. The book is for students taking a first course
in linear algebra.
Many students learn math by studying examples/practice problems. While going
through the solution to a problem, students are often faced with several issues. They
may not see the connection between the concept/theory taught in class and the solution.
Others may not understand the solution because a step is missing or there are not enough
explanations.
My main goal in this book is to address these issues to help students learn more effi-
ciently and get better results. To that end, I have included the following features.

ˆ Concept Summaries Boxes: They allow students to review the material and make
the book self-contained. Each concept involved in a problem is summarized in a box
right before the solution. Just the relevant information needed to solve the problem
is given. And the solution clearly shows how that information is used.

ˆ Very Detailed Solutions: They provide an “easy-to-understand” approach. Each

problem has a complete step-by-step solution with careful explanations for better
comprehension. Alternate solutions (to some problems) are also given. Some solutions
have “Help Texts”, which are located to the right of the solution sequence and help
students recognize which rule, property, or procedure is being applied.

The book is divided into two parts. The first part includes a variety of problems. The
idea is not to have as many problems as in a traditional textbook but a fair amount of
problems that cover essential/basic skills in matrix algebra. The second part includes full
solutions to every problem. Hyperlinks are added to ease the navigation of the book
(they are mainly used to go back and forth between the two parts).
To the Student: If you try to solve a problem and have no idea what to do, you can look
at the solution and try to understand it (if there is a box, you should first try to read and
understand its content). Try to solve the subsequent few problems yourself before checking
the solution. It’s not a big deal if you make a mistake; you will learn better from that
mistake.

vii
viii
 PartI

Exercises

1
 Chapter1

Systems of Linear Equations

1.1 Introduction to Systems of Linear Equations

1. Determine whether each equation is linear. If an equation is linear, find the number
of variables, the coefficients, and the constant term. [Solution on page 69]
√
(a) x1 + x2 = 0 (g) x1 − x2 = 19
3x1 +6x3
(b) 3x1 − 6x2 = 11 (h) 4
=1
1
(c) −2x1 + 4x2 − 5x3 = 2
(i) −4x1 − 20x2 + cos(x3 ) = −1
√
(d) 4x1 − x2 + 8x4 = 15 (j) 5x2 + x5 2 = −47
(e) −2x1 − 14x3 + 4x5 = 2021 (k) −x1 + 9x2 x4 − x5 = −3
1
(f ) x1 − x2 + x23 = 0 (l) x1 +x2
− x3 = 24

2. For each equation, three points are given. Determine which one is a solution to the
equation. [Solution on page 70]

(a) 2x1 + 3x2 = 14 Points: (3, 3), (1, 4), and (5, 1).
(b) −4x1 − 5x2 + 21x3 = 7 Points: (−2, 3, 0), (−2, −3, 0), and (2, −3, 0).

3. Determine whether each system is a system of linear equations. If a system is not a

linear system, explain why. [Solution on page 71]

x1 − 3x2 = 0 x1 + 3x2 = 2x3

(a)
4x1 + 8x2 = 9 (c) 4x1 + x2 − x3 = 0
x3
−2x1 + e = −2
1
√
5x1 − 5x2 + x
2 3
= ln 7 34x1 − 5x2 = −4
(b) (d)
x1 + x 2 = −5x3 41x1 + ln(x2 ) = 0

4. For each system, determine whether the given point is a solution to the system.
[Solution on page 71]

x1 − 2x2 = 3
(a) (1, −1)
3x1 + x2 = 2

3
4 Chapter 1. Systems of Linear Equations

−3x1 + 4x2 = 18
(b) (−2, 3)
5x1 − 6x2 = −27

x1 − x2 + x3 = 3
(c) 2x1 + 3x2 − x3 = −5 (0, −1, 2)
−3x1 + 4x2 + 2x3 = 1

x1 + 2x3 − x4 = 7
(d) 2x1 + x2 − 4x3 = −9 (2, −1, 3, 1)
− 5x2 + 10x4 = 15

5. Determine whether each system is consistent. Provide a solution if the system is

consistent. [Do this geometrically by sketching the graphs of the lines involved.]
[Solution on page 73]
x1 + x2 = 1 2x1 + 3x2 = 5
(a) (c)
−x1 + x2 = −1 −3x1 + x2 = 9
−x1 + 2x2 = 3 x1 − x2 = 2
(b) (d)
2x1 − 4x2 = −5 −2x1 + 2x2 = −4

6. Find a linear system in two variables with two different equations that satisfies the
following two conditions: [Solution on page 75]
ˆ Every coefficient is a nonzero number.
ˆ The point (5, −7) is a solution to the system.
7. Find a linear system in three variables with three different equations that satisfies
the following two conditions: [Solution on page 75]
ˆ Every coefficient is a nonzero number.
ˆ The point (−2, 3, −5) is a solution to the system.
8. Suppose we have a linear system in three variables with two equations. Also suppose
that the system is consistent. Is it possible to find a point (a, b) that satisfies the
system? Why or why not? [Solution on page 76]

1.2 Echelon Matrices

1. For each matrix below, fill out the following table. [Solution on page 76]

Zero Row? (Yes or No) Nonzero Row? (Yes or No) Leading Entry
Row 1
Row 2
Row 3
Row 4
1.2. Echelon Matrices 5

   
3 1 4 0 0 0 0 0 0 0
 0 2 0 0 0   8 0 0 0 5 
(a)   (b)  
 0 0 0 0 0   0 0 −7 7 9 
0 0 0 1 0 0 0 0 0 0

2. Determine whether each matrix is in echelon form. [Solution on page 77]

   
3 0 4 −7 0 1 0 2
(a)  0 2 5   0 0 4 3 2 
(d)  
0 0 6  0 6 0 0 6 

−1 1 2
 0 0 0 0 1
(b)  0 −2 4 
0 0 0 
−9 0 3 0 5

 
4 0 0 0  0 8 4 3 2 
(e)  
(c)  0 0 0 0   0 2 0 0 −2 
0 0 8 0 0 0 0 0 0

3. Is it possible to find a matrix in echelon form which is not in reduced echelon form?
Justify your answer. [Solution on page 78]

4. Consider the matrices below. Determine which ones are in echelon form. Also deter-
mine which matrices are in reduced echelon form. [Solution on
page 79]
   
1 0 0 1 0 0 1
(a)  0 1 0  (e)  0 0 5 0 
0 0 1 0 0 0 0
   
1 0 0 0 1 0 0 0
(b)  0 0 1 0  (f )  0 0 0 0 
1 0 0 0 0 0 0 1
   
1 0 1 0 1 7 0 −2 0
(c)  0 0 1 0  (g)  0 0 1 8 0 
0 0 0 1 0 0 0 0 1
   
0 1 0 0 4 1 0 −5 1 0
(d)  0 0 0 1 6  (h)  0 1 0 0 0 
0 0 0 0 0 0 0 0 1 0

5. Determine whether each operation is an elementary row operation. [Solution on

page 80]

(a) R2 + 2R1 → R2 (f ) R2 + 3R1 → R1 (k) 4R3 − R1 → R1

(b) R3 − R1 → R3 (g) R2 + 3R1 → R4 (l) 4R3 + 5R1 → R3
(c) R1 − 4R2 → R1 (h) R2 + 3R1 → R2 (m) R1 + R2 → R3
(d) R4 + 0 · R3 → R4 (i) 3R2 + R1 → R2 (n) R1 ↔ R2
(e) R2 + R1 → R1 (j) 3R2 + R1 → R1 (o) 3R1 ↔ R2
6 Chapter 1. Systems of Linear Equations

(p) R1 ↔ 5R2 (r) −5R2 → R2 (t) 2R1 → R3

(q) 2R1 → R1 (s) 0 · R1 → R1 (u) R3 → R4

6. Consider the matrix  

1 3 −5
A =  −2 0 4 
4 −1 −6
Perform each operation on A. [Solution on page 81]

(a) 2R1 → R1 (d) R2 ↔ R3 (g) R2 + 2R1 → R2

(b) −4R3 → R3 (e) R2 + R1 → R2 (h) R1 − 3R2 → R1
(c) R1 ↔ R2 (f ) R1 + R2 → R1 (i) R3 − 21 R1 → R3

7. In each case, find the single elementary row operation that was performed on A to
get B. [Solution on page 82]
   
1 2 1 2
(a) A = B=
−2 4 0 8
   
−1 3 0 −1 3 −1
(b) A =  1 −2 −3  B =  1 −2 −3 
−1 3 −1 −1 3 0
   
−2 3 −1 −2 3 −1
(c) A =  0 2 −2  B =  0 2 −2 
2 −3 −1 −2 3 1
   
3 5 −4 1 5 2
(d) A =  −1 0 3  B =  −1 0 3 
1 5 2 1 5 2
   
3 −12 −2 3 −12 −2
(e) A =  2 −3 −7  B =  −2 13 −3 
1 −4 −1 1 −4 −1

8. What is wrong with the following? [Solution on page 83]

   
1 1 2 1 1 2
 1 4 2  R2 − R3 → R2  −2 −1 −4 
R3 − R2 → R3
3 5 6 2 1 4

9. Find the pivots, pivot positions, and pivot columns of each matrix. [Solution on
page 85]
   
1 4 7 −3 4 0
(a)  0 9 2  (b)  0 8 2 
0 0 5 0 0 0
1.2. Echelon Matrices 7

   
0 4 0 1 1 0 2 0
(c)
0 0 1 13  0 0 0 1 
(d) 
 0

0 0 0 
0 0 0 0

10. (True or False) Let A be a matrix. Then it is possible to find two echelon forms of
A that are not the same. [Solution on page 86]

11. (True or False) Let A be a matrix. Then it is possible to find two reduced echelon
forms of A that are not the same. [Solution on page 87]

12. (True or False) Let A be a matrix. Let B and C be two echelon forms of A. Then
B and C have the same pivots. [Solution on page 89]

13. (True or False) Let A be a matrix. Let B and C be two echelon forms of A. Then
B and C have the same pivot positions. [Solution on page 89]

14. In each case, show that A and B are row equivalent. (In other words, show that B is
an echelon form of A.) [Solution on page 89]
   
1 0 2 1 0 2
(a) A= 1 3 3  B= 0 3 1 
−1 3 3 0 0 4
   
1 −2 4 1 −2 4
(b) A =  3 −7 15  B =  0 −1 3 
1 0 3 0 0 1
   
0 2 1 1 1 3
(c) A =  −3 1 −2  B= 0 4 7 
1 1 3 0 0 −5
   
1 −6 −4 1 1 0 0 1
(d) A= 0 2 3 5  B =  0 1 0 −2 
0 0 1 3 0 0 1 3

15. Row reduce the following matrices to reduced echelon form.[Solution on page 91]
   
1 2 3 4 −3 1 1 −1
(a)  1 3 1 1  (d)  4 −2 −2 0 
2 7 1 4 5 0 −3 −1
   
3 −5 13 −2 1 2 3 −2 7
(b)  −2 2 −12 14   1 2 3 −2 6 
(e)  
1 −2 3 1  1 2 3 −1 4 
  −1 −2 −4 3 −2
3 9 0 −3  
 −1 −2 2 −2  0 4 −3 −1
(c)  
 4 11 −1 3  (f )  −3 1 1 2 
3 9 0 −3 −2 −4 2 −4
8 Chapter 1. Systems of Linear Equations

16. [Solution on page 95]

Find all possible reduced echelon forms of a 3 × 2 matrix. If an entry can have an
arbitrary value, use an asterisk ∗ symbol.

1.3 Solving Linear Systems Using Matrices

1. Find the coefficient matrix and the augmented matrix of each system. [Solution on
page 95]

3x1 + 4x2 = 5
(a)
6x1 − 7x2 = 8

−x1 − 3x2 + x3 = −2
(b) x1 − x2 − 4x3 = 0
−5x1 + x3 = 6

x2 − x3 + x4 = 1
(c) x1 + 4x2 − 5x3 + 13x4 = −9
−2x1 − x4 = 6

2. Write the linear system corresponding to each augmented matrix. [Solution on

page 96]
   
1 −1 3 9 1 5 0 −2 −3
(a)  −2 4 −8 0  (c)  0 0 1 4 −1 
3 −5 6 12 0 0 0 1 7
 
1 −2 4 10
(b)  0 −3 −9 −41 
0 0 2 23

3. (True or False) Two linear systems are equivalent if and only if they have the same
set of solutions. [Solution on page 97]

4. (True or False) Two linear systems are equivalent if and only if the corresponding
augmented matrices are not row equivalent. [Solution on page 97]

5. Solve each system. [Solution on page 97]

x1 = 5 x1 − x2 + x3 = 3
(a)
x2 = 7 (c) 5x2 − 3x3 = 0
11x3 = 55
x1 + x2 = 7
(b)
3x2 = 12

6. Solve each system. [Solution on page 98]

1.3. Solving Linear Systems Using Matrices 9

x1 + 2x2 = 5 x1 + x2 = 0
(a) (b)
2x1 − x2 = −10 3x1 − 4x2 = 7

2x1 − 3x2 = 7
(c)
5x1 + 7x2 = 3

7. Solve each system. [Solution on page 100]

x1 − 2x2 − x3 = −6
(a) −x1 + 3x2 + 2x3 = 11
−2x1 + 5x2 + 4x3 = 20

x1 − x2 + x3 = 2
(b) 2x1 + x2 = −4
−3x1 + 2x3 = 14

x1 + x2 + x4 = 4
x1 + x3 = 1
(c)
2x2 − x3 + x4 = 4
2x1 − x2 + x3 − 3x4 = −5

8. Solve each system. [Solution on page 102]

x1 + x2 + 2x3 = 3 x1 − 2x2 + x3 = 0
(a) 3x1 + 4x2 + 8x3 = 5 (b) x1 − x2 + 4x3 = 1
− x2 − 2x3 = 2 −x1 + 2x2 − x3 = 5

9. Identify the leading variables and the free variables of each system. [Solution on
page 102]

3x1 + x2 − x3 = 4
(a)
5x2 − 2x3 = −8

x1 − x2 − x3 = 1
(b) −x2 + x3 = 7
2x3 = 5

x1 + 5x2 − 7x3 + x4 = −9
(c)
7x3 − x4 = 9

10. Determine whether each statement is True or False. [Solution on page 103]

(a) If all the variables of a linear system are leading variables, then the system has
a unique solution.
(b) A linear system that has at least one free variable must be consistent.

11. Solve each system. [Solution on page 104]

10 Chapter 1. Systems of Linear Equations

x1 − 2x2 + 4x3 = 5
(a) −x1 + 3x2 − 2x3 = −2
3x1 − 4x2 + 16x3 = 21

x1 + 3x2 + 2x3 = 7
(b) 2x1 + 7x2 + 3x3 = 10
−4x1 − 13x2 − 7x3 = −24

x1 + x2 + 2x3 = 1
(c) x1 + 2x2 + 2x3 = 4
−x1 − 3x2 − 2x3 = −2

x1 − 2x2 − x4 = 2
(d) 2x1 − 4x2 + x3 − 5x4 = 6
x1 − 2x2 − x3 + 2x4 = 0

3x1 + 8x2 − 3x3 − 14x4 = 2

2x1 + 3x2 − x3 − 2x4 = 1
(e)
x1 − 2x2 + x3 + 10x4 = 0
x1 + 5x2 − 2x3 − 12x4 = 1
12. Suppose the general solution to a system of linear equations is

x1 = 2+s−t
x2 = −3 + 7t
s and t arbitrary
x3 = s
x4 = t

Write down a linear system of two equations in four variables that has this general
solution. [Solution on page 107]

13. Consider the following system. [Solution on page 108]

x1 + x2 − x3 = a
x1 + 2x3 = b
−x1 − 2x2 + 4x3 = c

(a) Find a condition on a, b, and c such that the system is consistent.

(b) When the system is consistent, find the general solution (in terms of a, b, and
c).

14. Consider the following system. [Solution on page 108]

−x1 + 3x2 + 2x3 = −8

x1 + x3 = 2
3x1 + 3x2 + ax3 = b

Find conditions on a and b such that the system has no solution, one solution, or
infinitely many solutions.
 Chapter2

Vectors in Rn and Linear Systems

2.1 Vector Equations and Span

     
1 −3 6
1. Let u = ,v= , and w = . Compute each quantity. [Solution on
2 4 −5
page 111]

(a) u + v (c) w − 2u (e) 3u − v − 2w

(b) 10w (d) −3v (f ) −u − 3v + 4w
   
−1 1
2. Consider the vectors u = and v = . Graph the following vectors.
1 2
[Solution on page 112]

(a) 3u (c) u + v (e) −2u + 3v

(b) −2v (d) u − v (f ) −4u − v

3. Let n ≥ 1 be an integer. Algebraic properties of vectors include the following. Let

u, v, and w be vectors in Rn , and let a and b be scalars (real numbers). Let 0 denote
the zero vector in Rn . [Solution on page 115]

(a) (u + v) + w = u + (v + w) (e) a(u + v) = au + av

(b) u + 0 = 0 + u = u (f ) (a + b)u = au + bu
(c) u + (−u) = −u + u = 0 (g) a(bu) = (ab)u
(d) u + v = v + u (h) 1 · u = u

Prove (b) and (e).

4. In each case, write a system of linear equations that is equivalent to the given vector
equation. [Solution on page 117]
     
1 −2 5
(a) x1 2 + x2 −1 = 4 
    
3 0 6

11
12 Chapter 2. Vectors in Rn and Linear Systems

       
0 −2 5 10
(b) x1 + x2 + x3 =
1 1 0 12

5. For each system, write the corresponding vector equation. [Solution on page 118]

x 1 − x2 = 3
(a)
2x1 + 7x2 = −8

2x1 − x2 + x3 = 5
(b) −3x1 + 4x2 + 5x3 = 9
8x1 + 6x3 = 11

6. Consider the following system. [Solution on page 118]

x1 + 2x2 = 4
−4x1 − 3x2 = −1

(a) Graph each equation and solve the system geometrically.

(b) Solve the system algebraically (by row reducing the augmented matrix). And
verify that your answer is the same as the one found in part (a).
(c) Write the vector equation corresponding to the system and verify both alge-
braically and geometrically that the solution found in part (b) satisfies the vector
equation.

7. (True or False). A vector equation x1 a1 + x2 a2 + · · · + xn an = b has the same set


of solutions as the linear system whose augmented matrix is a1 a2 · · · an b .
[Solution on page 120]

8. Solve the vector equation x1 a1 + x2 a2 + x3 a3 = b, where

 
    
1 1 1 −5
a1 =  2  , a2 =  −1  , a3 =  −1  , b =  −4 
−1 0 1 −4

[Solution on page 120]

9. In each case, determine if b is a linear combination of a1 and a2 . [Solution on

page 121]
 
  
1 −2 8
(a) a1 =  −3 , a2 =  4 , b =  −18 
0 5 −15

   
1 4 1
(b) a1 =  −2  , a2 =  −11 , b= 1 
3 −1 29
2.1. Vector Equations and Span 13

   
1 2 0 −10
10. Let A =  1 1 3  and b =  12 . Determine whether b is a linear combina-
0 −1 4 27
tion of the columns of A. [Solution on
page 122]

11. In each case, find three different vectors in Span{a1 , a2 }. [Solution on page 123]
       
1 −2 −1 −4
(a) a1 = , a2 =
2 3 (b) a1 =  0  , a2 =  3 
2 6
     
−1 4 3
12. Let a1 =  2 , a2 =  −5 , and b =  −3 . Determine whether b is in
3 0 4
Span{a1 , a2 }. [Solution on page 123]

13. Consider the vectors.

       
1 3 1 0
 −2   −4   −1   1 
a1 = 
 1 ,
 a2 = 
 0 ,
 a3 = 
 2 ,
 b= 
 6 
0 5 4 7

Determine whether b is in Span{a1 , a2 , a3 }. [Solution on page 124]

     
1 −3 α
14. Let a1 =  0 , a2 =  1 , and b =  1 . For what value(s) of α is b in
−3 6 2
Span{a1 , a2 }. [Solution on page 125]
   
1 3
15. Let u = ,v= . Show that Span{u, v} = R2 . [Solution on page 125]
2 4
 
1 2 8
16. Consider the matrix A =  −2 3 7 . Show that the columns of A span R3 .
3 −1 2
[Solution on page 126]

17. (True or False). If u is a linear combination of v1 and v2 , then

Span{v1 , v2 } = Span{v1 , v2 , u}

[Solution on page 126]

18. (True or False). It is possible to find two vectors, v1 and v2 , of R3 such that

Span{v1 , v2 } = R3

[Solution on page 127]

14 Chapter 2. Vectors in Rn and Linear Systems

2.2 The Matrix Equation Ax = b

1. In each case, compute the product Ax (if it is defined). If a product is not defined,
explain why. [Solution on page 127]
   
1 3 5
(a) A = , x=
2 4 6
 
  3
−1 0 −3
(b) A = , x= 4 

2 4 −5
5
   
1 1 2 10
(c) A = , x=
3 0 6 20
 
1 −1  
7
(d) A =  2 −3  , x=
−2
3 0

2. (True or False). If A is of size 5×8, then the vector b in the matrix equation Ax = b
has 8 entries. [Solution on page 128]

3. In each case, write the system as a matrix equation. [Solution on page 128]

7x1 − 6x2 = 13
(a)
−8x1 + 9x2 = 5

x1 + x2 − x3 = 0
(b) 2x1 + x3 = 1
3x2 − 4x3 = 7

4. Given A, x, and b, write the linear system corresponding to the matrix equation
Ax = b. Also write the vector equation corresponding to Ax = b. [Solution on
page 128]
     
1 3 −4 x1 0
(a) A =  −5 2 1  , x =  x2  , b =  −3 
6 −1 −17 x3 8
 
  x1  
0 0 −2 3 1  x2  4
 2 −1 3 0 4 
 , x =  x3  , b =  23 
   
(b) A =   1 1 12 0 0  
 x4 
  0 
−3 −2 0 −9 9 32
x5

5. Let      
4 −10 10 x1 38
A =  3 −5 15  , x =  x2  , b =  16 
3 −6 12 x3 21
[Solution on page 129]
2.2. The Matrix Equation Ax = b 15

(a) Solve the equation Ax = b for x.

(b) For 1 ≤ i ≤ 3, let ai be the ith column of A. Use part (a) to deduce the solutions
to the equation x1 a1 + x2 a2 + x3 a3 = b.
   
−1 4 b1
6. Let A = and b = . [Solution on page 131]
2 −8 b2

(a) Show that the system Ax = b does not have a solution for all possible b.
(b) Give a geometric description of the set of all b for which the system Ax = b does
have a solution.
   
1 −2 −5 b1
7. Let A =  2 4 22  and b =  b2 . [Solution on page 132]
−3 5 11 b3

(a) Show that the system Ax = b does not have a solution for all possible b.
(b) Describe the set of all b for which the system Ax = b does have a solution.

8. (True or False). If the columns of an m × n matrix A span Rm , then for every

vector b in Rm the equation Ax = b is consistent. [Solution on page 132]

9. (True or False). If every column of an echelon form of a 3 × 3 matrix A is a pivot

column except the right most column, then for every b in R3 the system Ax = b has
a unique solution. [Solution on page 133]

10. Consider the matrices [Solution on page 133]

   
1 2 0 1 1 0 0 0
 0 1 −1 −2   0 1 0 1 
A=
 −1 −2
 B= 
1 −1   −3 0 1 −1 
2 5 −1 0 4 0 0 1

(a) Reduce A to an echelon form, and find the number of pivot positions of A.
(b) Does the equation Ax = b have a solution for every b in R4 ?
(c) Find the number of pivot positions of B.
(d) Does the columns of B span R4 ?
(e) Is it possible to find a vector in R4 which is not a linear combination of the
columns of A?
   
1 2 3 b1
11. Let A =  2 3 5  and b =  b2 . [Solution on page 134]
−1 4 4 b3

(a) Is the system Ax = b consistent for all possible b1 , b2 , b3 ?

(b) Do the columns of A span R3 ?

12. In each case, compute the product Ax using the dot product rule. [Solution on
page 135]
16 Chapter 2. Vectors in Rn and Linear Systems

   
1 3 5
(a) A = , x=
2 4 6
 
  3
−1 0 −3
(b) A = , x= 4 
2 4 −5
5
 
1 −1  
7
(c) A =  2 −3  , x=
−2
3 0

13. Let A be a 2 × 3 matrix. Let u, v be vectors in R3 , and let u0 , v 0 be vectors in R2 such

that Au = u0 and Av = v 0 . Let w be the vector in R3 defined by w = u0 + v 0 . Show
that the system Ax = w is consistent. [Solution on page 136]

14. Let A be a 7 × 4 matrix. Let u be a vector in R4 , and let v be a vector in R7 such

that Au = v. Show that the system Ax = 3v has a solution. [Solution on page 136]

2.3 Writing the General Solution of a Linear System

in Vector Form
1. In each case the solution set of a certain linear system is given. Write it in vector
form (that is, as a linear combination of vectors). [Solution on page 136]

x1 = 3t
(a) x2 = −4t t is arbitrary
x3 = t

x1 = 5 + 6t
(b) x2 = 10 − 8t t is arbitrary
x3 = t

x1 = 1 + s + 9t
x2 = −4 + 13t
(c) s and t are arbitrary
x3 = s
x4 = t

x1 = 4s − 7t
x2 = s
(d) x3 = −2 + 5t s and t are arbitrary
x4 = −3
x5 = t
2.3. Writing the General Solution of a Linear System in Vector Form 17

x1 = 24 + 3s1 − 8s2 + 9s3

x2 = s1
(e) x3 = −7 + 15s2 + 6s3 s1 , s2 , and s3 are arbitrary
x4 = s2
x5 = s3

2. (True or False). A homogeneous system is a system of the form Ax = 0, where A

is an m × n matrix and x is the vector of variables. [Solution on page 138]

3. (True or False). A homogeneous system can be inconsistent. [Solution on

page 138]

4. Describe the solution set to the equation Ax = 0 in vector form. Also find two
nontrivial solutions (if any). [Solution on page 139]
   
−3 5 1 2 −2
(a) A =
6 −9 (b) A =  −1 −1 7 
2 3 −9

5. Let [Solution on page 140]

 
1 −1 3 4
A =  1 −1 4 6 
−3 3 −7 −8

Show that the solution set to the equation Ax = 0 is Span{a1 , a2 }, where

   
1 2
 1   0 
a1 =  0  and a2 =  −2 
  

0 1

6. (True or False). It is possible to write the solution set of a nonhomogeneous linear

system as the span of a set of vectors. [Solution on page 140]

7. (True or False). Let A be an m × n matrix. Then the solution set to Ax = 0 is a

subset of Rm . [Solution on page 141]

8. (True or False). The solution set of Ax = b is the set of all vectors of the form
P +Vh , where P is a particular solution to Ax = b, and Vh is a solution to the equation
Ax = 0. [Solution on page 141]

9. Consider the matrix [Solution on page 141]

 
1 −2
A=
−4 8

(a) Solve the homogeneous system Ax = 0 and describe the solution set geometri-
cally (sketch a graph of the solution set).
18 Chapter 2. Vectors in Rn and Linear Systems

 
3
(b) Solve the system Ax = b, where b = , and describe the solution set
−12
geometrically (sketch a graph of the solution set).

10. Shown is a line L through the origin. [Solution on page 142]

x2
4

−3 4 x1

−3 L

(a) (True or False) The line L is the solution set to a nonhomogeneous linear
system.
(b) From the graph, find two nontrivial solutions to the linear system whose solution
set is L.
(c) Find a linear system with two equations and two variables whose solution set is
L.

11. Let [Solution on page 144]

 
1 1 2 −2 5 0
 0 0 1 −3 8 −13 
M =
 0

0 0 0 1 11 
0 0 0 0 0 0

be the augmented matrix of a system of linear equations. Write the general solution
in vector form.

12. Solve each system and write the general solution in vector form. [Solution on
page 145]
     
1 −1 2 x1 0
(a) Ax = b, where A =  −1 4 −7  , x =  x2 , b = 0 
 
−1 10 −17 x3 0

x1 + 3x2 + 4x3 = 1
(b) −2x1 + x2 + 5x3 = 2
7x1 + 7x2 + 2x3 = −1
2.3. Writing the General Solution of a Linear System in Vector Form 19

x1 + 2x2 + x3 + 5x4 = −4
(c) 2x1 + 4x2 + 3x3 + 7x4 = 5
−3x1 − 6x2 − 5x3 − 9x4 = −14

         
1 −2 −3 4 −9
(d) x1  1  + x2  −1  + x3  −4  + x4  10  =  9 
−2 3 7 −14 0

13. Consider the following system [Solution on page 148]

x1 − 2x2 + 3x3 + x4 = −3
−x1 + 2x2 − 2x3 + x4 = 5
2x1 − 4x2 + 5x3 = −8

(a) Write the system in the form Ax = b.

(b) Reduce the augmented matrix to reduced echelon form.
(c) Solve the system.
(d) Write the general solution in vector form.
(e) Find a particular solution
(f ) Does the corresponding homogeneous system Ax = 0 have a nontrivial solution?
Justify your answer.

14. A zero matrix is a matrix where every entry is 0. Suppose A is a 2 × 2 zero matrix.
Describe the solution set of Ax = 0. [Solution on page 149]

15. A nonzero matrix is a matrix where at   entry is nonzero. Construct a 3 × 3

least one
2
nonzero matrix A such that the vector  3  is a solution to Ax = 0. [Solution on
−4
page 149]

16. Find a homogeneous system with three equations and  three

 variables
  whose general
1 0
solution is the plane in R3 spanned by the vectors  0  and  1 . [Solution on
1 1
page 149]

17. Find a linear system with two equations and three variables that has
   
3 −4
 −1  + s  2 
0 1

as the general solution. [Solution on page 150]

20 Chapter 2. Vectors in Rn and Linear Systems

18. Find a linear system with three equations and four variables that has
     
2 −3 7
 0 
  + s 1  + t 0 
   
 5   0   −1 
0 0 1

as the general solution. [Solution on page 150]

2.4 Linear Independence

1. In each case, determine if the vectors are linearly independent. If the vectors are
dependent then write down a specific linear dependence relation for them. [Solution
on page 151]
   
1 2
(a) v1 = , v2 =
−1 3

   
4 −12
(b) v1 = , v2 =
−5 15

     
0 1 2
(c) v1 =  −1 , v2 =  0 , v3 =  −3 
1 1 5

    
2 4 0
(d) v1 = 3 , v2 =
   0 , v3 = 7 
 
0 −5 6

2. Determine if the statement is True or False, and justify your answer. [Solution on
page 155]

(a) If an echelon form of a matrix A has a pivot in every row, then the columns of
A are linearly independent.
(b) If an echelon form of a matrix A has a pivot in every column, then the columns
of A are linearly independent.

3. In each case, determine if the columns of the matrix form a linearly independent set.
[Solution on page 155]
 
−3 4 27
(a)  2 −1 −13 
4 7 1
2.4. Linear Independence 21

 
1 3 −4
 −1 −2 9 
(b)  
 1 2 −8 
3 11 −1
    
1 0 1
4. Let v1 = 2 , v2 = 1 , v3 = h .
     [Solution on page 157]
3 1 4

(a) For what value(s) of h (if any) is v3 in Span{v1 , v2 }?

(b) For what value(s) of h (if any) is the set {v1 , v2 , v3 } linearly independent?
     
−1 2 1
5. Let v1 =  2 , v2 = −4 , v3 = 1 . For what value(s) of h (if any) is the
   
−3 6 h
set {v1 , v2 , v3 } linearly independent? [Solution on page 157]

6. Determine if the statement is True or False, and justify your answer. [Solution on
page 158]

(a) A set of one vector is always linearly independent.

(b) A set of two vectors is linearly dependent if and only if one vector in the set is
a multiple of the other.
(c) A set S of two or more vectors is linearly dependent if and only if at least one
vector in S is a linear combination of the others.
(d) If S is a set of linearly dependent vectors, then every vector in S is a linear
combination of the other vectors in S.
       
1 −2 1 −8
7. Let v1 =  −1 , v2 =  3 , v3 =  −2 , and v4 =  10 . [Solution on
−2 1 2 10
page 160]

(a) Show that the set {v1 , v2 , v3 , v4 } is linearly dependent.

(b) Is it possible to write v3 as a linear combination of v1 , v2 , and v4 ? Justify your
answer.

8. Determine if the statement is True or False, and justify your answer. [Solution on
page 161]

(a) Let S be a set of vectors in Rn . If one of the vectors in S is the zero vector,
then S is linearly dependent.
(b) A linearly dependent set must contain the zero vector.
(c) Let {v1 , · · · , vp } be a set of vectors in Rn . If p > n, then the set {v1 , · · · , vp } is
linearly dependent.
(d) If a set {v1 , · · · , vp } of vectors in Rn is linearly dependent, then p > n.
22 Chapter 2. Vectors in Rn and Linear Systems

9. Determine by inspection if the given set is linearly dependent. [Solution on

page 162]
 
4
(a) v =  5 , S = {v}.
6

 
0
(b) v = , S = {v}.
0

      
1 1 7 1
(c) v1 =  7  , v2 =  0 , v3 =  8  , v4 =  1 , S = {v1 , v2 , v3 , v4 }.
6 8 9 2

    
−8 0 10
(d) v1 =  −7 , v2 =  0 , v3 =  11 , S = {v1 , v2 , v3 }.
13 0 12

   
−2 4
 4   −8 
(e) v1 =  , S = {v1 , v2 }.
 6 , v2 =  −12
 

10 −21

10. Consider the following vectors in R2 . Let 0 denote the zero vector. [Solution on
page 162]

4 v2

v1

v4
−2 4 x

−2 v3

Determine whether each set is linearly independent.

(a) {v1 } (c) {v1 , v3 } (e) {v2 , v4 }

(b) {v1 , v2 } (d) {v1 , v3 , v4 } (f ) {v2 , v4 , 0}

11. Determine if the statement is True or False, and justify your answer. [Solution on
page 163]
2.4. Linear Independence 23

(a) If every entry in the bottom row of a matrix is 0, then the columns of A are
linearly dependent.
(b) If every entry in the rightmost column of a matrix A is 0, then the columns of
A are linearly dependent.
 
  5
12. Let A = a1 a2 a3 be a 3 × 3 matrix and suppose that x =  −2  is a solution
1
to the homogeneous equation Ax = 0. [Solution on page 163]

(a) Give a linear dependence relation on the columns of A.

(b) Do the columns of A span R3 ? Why or why not?
 
3 5 13
13. Let A =  1 −7 −13 . Observe that a3 = a1 + 2a2 , where ai is the ith column
−2 8 14
of A. Find a nontrivial solution to the equation Ax = 0. [Solution on page 163]

14. Let A be a 2 × 2 matrix with linearly dependent columns. Write out all possible
reduced echelon form of A. For entries that can have an arbitrary value, use an
asterisk ∗ symbol. [Solution on page 164]

15. Determine if the statement is True or False, and justify your answer. [Solution on
page 164]

(a) If the equation Ax = 0 has a nontrivial solution, then the columns of A are
linearly dependent.
(b) If the columns a1 , a2 , · · · , an of an m × n matrix A span Rm , then the set
{a1 , a2 , · · · , an } is linearly independent.
(c) If the columns of an m × n matrix A are linearly independent, then they span
Rm .
(d) Let A be an m × n matrix. If the columns of A are linearly independent, then
for every b in Rm the linear system Ax = b has a unique solution.

16. Let u, v be vectors in Rm such that the set {u, v} is linearly independent. Prove that
the set {u + v, u − v} is also linearly independent. [Solution on page 164]
24 Chapter 2. Vectors in Rn and Linear Systems
 Chapter3

Linear Transformations

3.1 Introduction to Linear Transformations

1. Show that the transformation T : R2 → R2 defined by [Solution on page 167]
   
x1 4x1 − x2
T =
x2 3x1 + 5x2

is linear.

2. Show that the transformation T : R2 → R3 defined by [Solution on page 168]

 
  3x1
x1
T = 0 
x2
x1 − x2

is linear.

3. Determine if the statement is True or False, and justify your answer. [Solution on
page 169]

(a) T : Rn → Rm is a linear transformation if and only if for every vectors x, y in

Rn and for every real numbers c, k,

T (cx + ky) = cT (x) + kT (y)

(b) If T : Rn → Rm is a linear transformation, then T (0) = 0.

(c) If T : Rn → Rm is a transformation such that T (0) = 0, then T is linear.
(d) If T : Rn → Rm is a transformation such that T (0) 6= 0, then T is not linear.

4. In each case, show that the transformation T is not linear. [Solution on page 171]
 
2 x1  √ 
(a) T : R → R defined by T = 3 x1 x2 .
x2
 
  x1
x1
(b) T : R2 → R3 is defined by T =  −x1 + x2 .
x2
5|x1 |

25
26 Chapter 3. Linear Transformations

   
2 2 x1 9x2 + 8
(c) T : R → R defined by T = .
x2 x1

5. Determine whether the following transformation is linear. [Solution on page 172]

 
x1  
7x 1 − x 2
T  x2  =
−8x1 + x2 + ex3
x3

6. Let T : R2 → R2 be a linear transformation, and let u, v be vectors in R2 such that

   
3 5
T (u) = and T (v) =
4 −2

[Solution on page 172]

(a) Find T (4u)

(b) Find T (u + v)
(c) Find T (7u − 6v)

7. Let T : R3 → R2 be a linear transformation and let u, v be vectors in R3 such that

   
2 −4
T (v) = and T (2u + 7v) =
−5 3

Find T (u). [Solution on page 173]

8. Suppose that T : Rn → Rm is a linear transformation, and also that w is in Span{u, v},

where u and v are vectors in Rn . Prove that T (w) is in Span{T (u), T (v)}. [Solution
on page 173]

9. Let [Solution on page 173]

      
−3 5 2 −3
u1 =  2  , u01 =  −1  , u2 =  −1  , and u02 =  −2 
8 2 7 15
3 3 0 0
 transformation T : R → R such that T (u1 ) = u1 and T (u2 ) = u2 .
Considerthe linear
32
Let v = −19 .

1

(a) Show that there exists a unique pair (c1 , c2 ) such that v = c1 u1 + c2 u2 .
(b) Find T (v).
       
1 0 −1 3
10. Let e1 = , e2 = , v1 = , and v2 = . Let T : R2 → R2 be a
0 1 2 4
linear transformation that maps e1 into v1 and e2 into v2 . [Solution on page 174]
3.1. Introduction to Linear Transformations 27

 
12
(a) Find T .
−13
 
x1
(b) Find T (x) for an arbitrary x = .
x2

11. Let T : R3 → R3 be a linear transformation such that [Solution on page 175]

        
  
1 0 1 −4 −1 −2
T  −1  =  1  , T  0  =  1  , and T  3  =  1 
3 2 4 −6 0 5
 
b1
(a) Find T (b) for an arbitrary b =  b2 .
b3
 
−6
(b) Find T (b) for b =  −7 .
13

12. Determine if the statement is True or False, and justify your answer. [Solution on
page 176]

(a) If T : Rn → Rm is a linear transformation, and u and v are two linearly inde-

pendent vectors in Rn , then the set {T (u), T (v)} is linearly independent.
(b) If T : R2 → R2 is a linear transformation, and u and v are two vectors that span
R2 , then the set {T (u), T (v)} spans R2 .
(c) Let u, v be vectors in Rn such that the set {u, v} is linearly independent. Let
T : Rn → Rm be a linear transformation such that the set {T (u), T (v)} is linearly
dependent. Then the equation T (x) = 0 has a nontrivial solution.
(d) If A is an m × n matrix, then the transformation T : Rn → Rm defined by
T (x) = Ax is linear.
(e) If A is a matrix of size 7 × 11, and T : Ra → Rb is the linear transformation
defined by T (x) = Ax, then a = 7 and b = 5.
(f ) If A is an p × q matrix, and T : R6 → R9 is the linear transformation defined by
T (x) = Ax, then p = 6 and q = 9.

13. Let [Solution on page 178]

       
1 −3 2 −7 4
A= , u= , v= , and b =
5 −17 6 8 12

Let T : R2 → R2 be the linear transformation defined by T (x) = Ax.

(a) Find T (u) and T (v).

(b) Find a vector x in R2 such that T (x) = b and determine whether x is unique.
28 Chapter 3. Linear Transformations

14. Let [Solution on page 178]

   
1 −2   7
−1
A= 3 1 , u= , and b =  7 
2
1 7 −11

Let T : R2 → R3 be the linear transformation defined by T (x) = Ax.

 
x1
(a) Find a formula for the image of an arbitrary vector x = under T .
x2
(b) Find the image under T of u.
(c) Find an x in R3 whose image under T is b.
(d) Is there more than one x whose image under T is b? Why or why not?

15. Let [Solution on page 179]

     
1 −1 1 1 1
A =  −2 2 −2  , y =  1 , and y 0 =  −2 
3 −4 −1 1 4

Let T : R3 → R3 be the linear transformation defined by T (x) = Ax.

(a) Find a vector x whose image under T is y.

(b) Find a vector x whose image under T is y 0 .
(c) Is there more than one x whose image under T is y 0 ? Why or why not? If there
is more than one x such that T (x) = y 0 , find three vectors whose image under
T is y 0 .
 
b1
(d) Describe the set of all vectors b =  b2  for which the equation T (x) = b has
b3
a solution. (This set is called the range or image of T .)
 
x1
(e) Describe the set of all x =  x2  for which T (x) = 0. (This set is called the
x3
kernel of T .)

3.2 The Matrix of a Linear Transformation

1. In each case, describe geometrically what the transformation T does on each vector
in R2 . [Solution on page 182]
 
4 0
(a) A = , T : R2 → R2 is defined by T (x) = Ax.
0 4
 
−1 0
(b) A = , T : R2 → R2 is defined by T (x) = Ax.
0 1
3.2. The Matrix of a Linear Transformation 29

 
0 0
(c) A = , T : R2 → R2 is defined by T (x) = Ax.
0 1
 
0 1
(d) A = , T : R2 → R2 is defined by T (x) = Ax.
1 0

2. The unit square in R2 is the set of all points (x1 , x2 ) in R2 such that 0 ≤ x1 ≤ 1 and
0 ≤ x2 ≤ 1. In each case, sketch the graph of the image of the unit square under T .
[Solution on page 184]
 
3 0
(a) T (x) = Ax where A =
0 3
 
−2 0
(b) T (x) = Ax where A=
0 3
 
1 2
(c) T (x) = Ax where A=
0 1
 
1 0
(d) T (x) = Ax where A=
−3 1
 
−2 −1
(e) T (x) = Ax where A=
1 2
"
#
cos π4 − sin π


4
(f ) T (x) = Ax where A=
sin π4 π



cos 4

3. In each case find the standard matrix of the linear transformation T . [Solution on
page 190]
 
2
 
3x1
(a) T : R → R is defined by T x1 =
−5x1
   
x1 7x1 − x2
(b) T : R2 → R2 is defined by T =
x2 3x2
 
x1  
x 2 − 4x 3
(c) T : R3 → R2 is defined by T  x2  =
−x1 + 6x2 − 9x3
x3
 
  5x1
x1
(d) T : R2 → R3 is defined by T = 0 
x2 x1 −x2
3
   
x1 x1 − 3x2 + 6x3
(e) T : R3 → R3 is defined by T  x2  =  −x1 + 4x2 − 7x3 
x3 2x1 + 5x2 + x3

4. Consider the following figure where T : R2 → R2 is a linear transformation. [Solution

on page 193]
30 Chapter 3. Linear Transformations

x2

T (e2 )

e2
e1
−2 5 x1

T (e1 )
−2

(a) Find the standard matrix of T .

 
1
(b) Find the image of the vector u = under T . Graph the vector T (u).
1

5. Consider the following figure where T : R2 → R2 is a linear transformation. [Solution

on page 194]

x2

T (e1 )
e2
T (e2 )

−2 e1 5 x1

−2

(a) Find the standard matrix of T .

 
−2
(b) Graph the vector T (u) where u = .
3

6. In each case, suppose that T is a linear transformation. Find the standard matrix of
T. [Solution on page 195]

(a) T : R2 → R2 is the reflection through the x1 -axis.

π
(b) T : R2 → R2 is the counterclockwise rotation through 2
about the origin.
(c) T : R3 → R3 is the reflection through the x1 x2 -plane.
(d) T : R2 → R2 is the reflection through the line x2 = −x1 .
(e) T : R2 → R2 is the projection onto the x1 -axis.
π
(f ) T : R2 → R2 is the clockwise rotation through 3
about the origin.
3.3. One-To-One and Onto Linear Transformations 31

π
(g) T : R2 → R2 first reflects points through the x1 -axis and then rotates points 2
(clockwise).

7. Describe geometrically the linear transformation T : R2 → R2 whose standard matrix

is " √ #
3 1
2
− 2
A= √
1 3
2 2

[Solution on page 198]

3.3 One-To-One and Onto Linear Transformations

1. Determine if the statement is True or False, and justify your answer. [Solution on
page 199]

(a) A transformation T : Rn → Rm is one-to-one if for every vector b in Rm , there

exists at least one vector x in Rn such that T (x) = b.
(b) A transformation T : Rn → Rm is onto if for every vector b in Rm , the equation
T (x) = b has at most one solution.
(c) Let T : Rn → Rm be a transformation. If there exists a vector b for which the
equation T (x) = b has at most one solution, then T is one-to-one.
(d) Let T : Rn → Rm be a transformation. If there exists a vector b for which the
equation T (x) = b has at least one solution, then T is onto.

2. Determine if the statement is True or False, and justify your answer. [Solution on
page 200]

(a) Let A be an m × n matrix and let T : Rn → Rm be the linear transformation

defined by T (x) = Ax. The transformation T is one-to-one if and only if the
columns of A are linearly independent.
(b) Let A be an m × n matrix and let T : Rn → Rm be the linear transformation
defined by T (x) = Ax. The transformation T is one-to-one if and only if the
equation T (x) = 0 has only the trivial solution.
(c) If a linear transformation T : Rn → Rm is one-to-one, then n ≤ m.
(d) If n ≤ m, then any linear transformation T : Rn → Rm is one-to-one.
(e) It is possible to find a linear transformation T : R3 → R2 which is one-to-one.
(f ) A linear transformation T : Rn → Rm is one-to-one if and only if its kernel is
{0}.

3. Determine if the statement is True or False, and justify your answer. [Solution on
page 202]

(a) Let A be an m × n matrix and let T : Rn → Rm be the linear transformation

defined by T (x) = Ax. The transformation T is onto if and only if the columns
of A span Rm .
32 Chapter 3. Linear Transformations

(b) If a linear transformation T : Rn → Rm is onto, then n ≥ m.

(c) If n ≥ m, then any linear transformation T : Rn → Rm is onto.
(d) It is possible to find a linear transformation T : R2 → R3 which is onto.
(e) If A is a 6×5 matrix, then it is possible to find a linear transformation T (x) = Ax
which is onto.
(f ) If a linear transformation T : Rn → Rm is both one-to-one and onto, then n = m.
(g) If n = m, then any linear transformation T : Rn → Rm is both one-to-one and
onto.
(h) A linear transformation T : Rn → Rm is onto if and only if its range is Rm .

4. In each case determine if the linear transformation T is onto. Also determine if T is

one-to-one. [Solution on page 203]
 
x1  
x 1 − 2x 2 − x 3
(a) T : R3 → R2 is defined by T  x2  =
x1 + 3x2
x3
   
x1 x1 − 3x2 + 6x3
(b) T : R3 → R3 is defined by T  x2  =  −x1 + 4x2 − 7x3 
x3 2x1 + 5x2 + x3
(c) T : R2 → R3 is the linear transformation whose standard matrix is
 
1 4
A =  −3 −12 
2 7

(d) T : R3 → R3 is the linear transformation whose standard matrix is

 
5 −20 10
A =  2 −7 9 
−4 15 −6

(e) T : R5 → R4 is the linear transformation whose standard matrix is

 
1 −2 4 5 0
 0 1 0 −1 2 
A=  −1

4 −3 −4 4 
−2 1 −9 −9 −6

5. [Solution on page 206]

Let T : R3 → R3 be the linear transformation whose standard matrix is
 
1 −2 −5
A =  −4 9 26 
5 −9 −19
3.3. One-To-One and Onto Linear Transformations 33

 
1
(a) Find an x in R3 whose image under T is u =  1 .
1
(b) Is T one-to-one? If not, find three different vectors in the domain that have the
same image.
(c) Is T onto? If not, find three different vectors in the codomain that are not in
the range of T . (The range of T is defined as the set of all vectors b in the
codomain for which the equation T (x) = b is consistent.)
(d) Describe geometrically the range of T .
(e) The kernel of T is the set of all vectors x in the domain such that T (x) = 0.
Describe geometrically the kernel of T .
34 Chapter 3. Linear Transformations
 Chapter4

Matrix Algebra

4.1 Matrix Operations

1. Determine if the statement is True or False, and justify your answer. [Solution on
page 209]
 
2 4 9
(a) The matrix is of size 3 × 2.
3 1 8
 
(b) The matrix 1 9 7 is of size 1 × 3.
 
2 4 9
(c) The (1, 2)-entry of is 3.
3 1 8
 
2 4 9
(d) The (2, 3)-entry of is 8.
3 1 8
 
2 4 9
(e) The diagonal entries of are 2 and 1.
3 1 8
 
2 4 9
(f ) The diagonal entries of 3 1 8  are 9, 1, and 5.
5 7 10
(g) It is possible tofind a 3 × 2 matrix which is a diagonal matrix.
 
2 5
(h) The matrix 1  3 is a square matrix.
0 0
 
06
(i) The matrix is a diagonal matrix.
90
 
1 0 0
(j) The matrix 0  2 0 is a diagonal matrix.
0 0 3
 
0 1
(k) The 2 × 2 identity matrix is the matrix
1 0
 
1 0 0
(l) The 3 × 3 identity matrix is the matrix 0 1 0
0 0 1

35
36 Chapter 4. Matrix Algebra

(m) It is possible to add two matrices of different sizes.

2. Let [Solution on page 210]

   
    9 −7 17 −9
1 7 −2 6
A= ,B = , C =  −3 0  , and D =  −4 1 .
−3 8 5 −14
11 −22 0 19

Compute each of the following expressions if they are defined. If an expression is

undefined, explain why.

(a) A + B (d) C + D + B
(b) C + D (e) 2C
(c) A + C (f ) 3C − 2D

3. Let [Solution on page 212]

 
  3 −3 5
2 −2
A= and B =  −4 7 10 
5 12
11 9 −8

Compute A − 2I2 and 7I3 − B.

4. Consider the matrices A and B from Exercise 2. Compute −5A + 2B − 3I2 . [Solution
on page 212]
   
5 −3 −2 8
5. Solve the the equation X + = −6 , where X is a 2 × 2 matrix.
1 6 1 0
[Solution on page 212]

6. Let α and β be two scalars (real numbers), and let A be a matrix. Prove that
(α + β)A = αA + βA. [Solution on page 213]

7. Determine if the statement is True or False, and justify your answer. [Solution on
page 213]

(a) If A is of size m × n and B is of size m × p, then AB is of size n × p.

(b) If A is of size m × n and B is of size n × p, then AB is of size m × p.
(c) It is always possible to multiply two matrices of different sizes.
(d) If A and B are two matrices such that A is of size 5 × 7 and AB is of size 5 × 11,
then B is of size 5 × 11.
(e) If A and B are two matrices such that B is of size 5 × 9 and AB is of size 3 × 9,
then A is of size 3 × 5.
(f ) If A and B are not square matrices, then AB is not a square matrix.
(g) If A and B are two matrices, then AB and BA have the same size.
4.1. Matrix Operations 37

   
1 2 5 6
(h) If A = and B = , then
3 4 7 8
   
1(5) + 2(7) 3(5) + 4(7) 19 43
AB = =
1(6) + 2(8) 3(6) + 4(8) 22 50
   
1 2 5 6
(i) If A = and B = , then
3 4 7 8
   
1(5) + 2(7) 1(6) + 2(8) 19 22
AB = =
3(5) + 4(7) 3(6) + 4(8) 43 50

8. Let [Solution on page 215]

   
    2 0 −2 1 3
1 2 −3 5
A= ,B = , C =  5 −8  , D =  4 8 9 ,
3 4 4 −7
13 −9 10 −4 5
 
  0 9 4
−2 6 9
E= , and F =  −2 −2 −5 
3 −7 7
1 −10 −1
Compute each of the following expressions if they are defined. If an expression is
undefined, explain why.

(a) AB (d) CB (g) CE

(b) BA (e) CD (h) EC
(c) BC (f ) DC (i) DF

9. Let A and B be matrices such that A is of size 5 × 7 and the number of columns of
AB is 11. What is the size of B? [Solution on page 217]
10. Let A and B be matrices such that the product BA is of size 17 × 32. [Solution on
page 217]
(a) How many rows does B have?
(b) How many columns does A have?
11. Let A be an m × n matrix, and let B be an n × p matrix. Show that every column of
the product AB is a linear combination of the columns of A. [Solution on page 217]
 
1 2
12. Let A = . Let a1 and a2 be the columns of A (of course, a1 is the first column
3 4
and a2 is the second). Consider the following linear combinations. [Solution on
page 217]    
0 2
2a1 − a2 = and − 4a1 + 3a2 =
2 0
 
0 2
Find a 2 × 2 matrix B such that AB = .
2 0
38 Chapter 4. Matrix Algebra

   
1 −2 5 5
13. Let A = . Find a 2 × 2 matrix B such that AB = . [Solution on
3 −1 5 5
page 218]

14. Suppose the third column of B is all zeros. What can you say about the third column
of AB? [Solution on page 219]

15. Let T : Rp → Rn and R : Rn → Rm be two transformations. The composition of T

and R, denoted R ◦ T , is the transformation T ◦ R : Rp → Rm defined as

(R ◦ T )(x) = R(T (x)) for every x in Rp .

Suppose T and R are both linear. [Solution on page 219]

(a) Show that T ◦ R is a linear transformation.

(b) Let A be the standard matrix of T , and let B be the standard matrix of R.
Express the standard matrix of R ◦ T in terms of A and B.
(c) Suppose that T : R2 → R2 and R : R2 → R2 are defined by
       
x1 x1 − x2 x1 −2x1 + 5x2
T = and R =
x2 3x1 + 4x2 x2 6x1 − 7x2

Find the standard matrix of R ◦ T in two different ways.

16. Let A be an m × n matrix, B be an n × p matrix, and let C be an p × q matrix. Prove

that A(BC) = (AB)C. [Solution on page 220]

17. Let A and B be two matrices such that BA = Im , where Im is the m × m identity
matrix. [Solution on page 221]

(a) Show that the equation Ax = 0 has only the trivial solution.
(b) Explain why the number of rows of A is greater than or equal to the number of
columns.
(c) Is the linear transformation T (x) = Ax one-to-one? Why or why not?

18. Determine if the statement is True or False, and justify your answer. In the following,
A, B, and C are arbitrary matrices such that the indicated products and sums are
defined. [Solution on page 221]

(a) AB = BA
(b) AC + BC = (A + B)C
(c) AB + CA = A(B + C)
(d) A(BC) = (AC)B
(e) If AB = AC, then B = C
(f ) If AB = 0, then A = 0 or B = 0
(g) If A = 0 or B = 0, then AB = 0
4.1. Matrix Operations 39

19. Let A, B, and C be n × n matrices. Assume that A and B commute. Also assume
that A and C commute. Show that A commutes with BC. [Solution on page 222]
 
  0 −1 2
1 2
20. Let A = and B =  3 1 0 . Compute A2 and B 3 . [Solution on
3 4
−2 0 1
page 223]
 
  −1
21. If A = −2 3 5 , B =  2 , compute A2 , AB, BA, and B 2 when they are
4
defined. [Solution on page 223]
 
a 0
22. Let A = be a 2 × 2 diagonal matrix. [Solution on page 224]
0 b

(a) Compute A2 and A3 .

(b) If n is a non-negative integer, guess a formula for An .

23. Let [Solution on page 224]

   
−3   −4 2 −2
  0 −1 2
A= 0 6 7 , B =  9 ,C = , D =  3 1 −7  .
3 1 0
−8 10 13 20

Find AT , B T , C T , and DT .
 
4 −1 2
24. Let A = . Compute AAT − 5I2 and 8I3 − AT A. [Solution on
−2 1 3
page 225]

25. Determine if the statement is True or False, and justify your answer. In the following,
A, B, and C are arbitrary matrices such that the indicated products and sums are
defined. [Solution on page 225]

(a) (AT )T = A
(b) (A + B)T = AT + B T
(c) (AB)T = AT B T
(d) (AB)T = B T AT
(e) (ABC)T = C T B T AT
(f ) (AB)2 = A2 B 2
(g) A2 − In = (A − In )(A + In )

26. If A and B are matrices, can you conclude in general that (A+B)2 = A2 +2AB +B 2 ?.
If this is not true in general, give a specific example of A and B that makes the equality
false. [Solution on page 226]
40 Chapter 4. Matrix Algebra

27. A student tries to solve the matrix equation X 2 + BX = XC (where all of these are
n × n matrices) as follows:

X 2 =XC − BX
X 2 =X(C − B)
X 2 − X(C − B) =0
X(X − (C − B)) =0
X = 0 or X = C − B

Explain two different matrix algebra mistakes that the student made. [Solution on
page 227]

28. Recall that a matrix X is symmetric if X T = X. Let A be an m × n matrix. Show

that the product AAT is a symmetric matrix. [Solution on page 227]

4.2 The Inverse of a Matrix

1. Determine if the statement is True or False, and justify your answer. In the following
the identity matrix of any size is denoted I. [Solution on page 227]

(a) If A is an n × n matrix, a matrix B is an inverse of A if AB = I and BA = I,

where I = In is the n × n identity matrix.
(b) An invertible matrix is a matrix that has an inverse.
(c) A 2 × 3 matrix can have an inverse.
   
−1 0 −2 0
(d) The matrix B = is an inverse of A = .
0 −1 0 −2
   
1 0 1 0
(e) The matrix B = is an inverse of A = .
−1 1 1 1
(f ) If B and C are both inverses of A, then B = C.
(g) If A2 = I, then A is invertible and A−1 = A.

2. Let A be a square matrix such that A5 = I. Is A is invertible? Why or why not?

[Solution on page 228]

3. In each case, determine whether B is the inverse of A. [Solution on page 228]

   
1 0 1 1 −1 2
(a) A =  2 2 3  and B =  −1 0 1 
1 1 1 0 1 −2
   
1 0 0 1 0 0
(b) A =  0 1 −1  and B =  0 2 3 
0 1 −2 0 1 3

4. Find the inverse of each matrix. If the inverse does not exist, explain why. [Solution
on page 229]
4.2. The Inverse of a Matrix 41

   
1 2 −2 3
(a) A = (c) A =
2 6 −4 8
   
−6 7 15 −18
(b) A = (d) A =
12 −14 −10 12
 
k −5
5. Let A = . Find the values of k that make A invertible. [Solution on
−4 k + 1
page 230]

6. In each case, solve the linear system by finding and using the inverse of the coefficient
matrix. [Solution on page 230]
−3x1 − 6x2 = 10 4x1 − 5x2 = 11
(a) (b)
−x1 + 3x2 = 10 7x1 + 6x2 = −3

7. Find the inverse of each matrix. If the inverse does not exist, explain why. [Solution
on page 231]
   
1 1 1 1 −6 5
(a) A =  0 1 1  (c) A =  −2 13 −17 
2 3 4 3 −19 22

 
0 2 1
(d) A =  −2 −2 −2 
−1 0 1
 
1 −3 −2 
1 0 0 0

(b) A =  −1 5 2   2 1 0 0 
−3 7 7 (e) A = 
 3

1 1 0 
4 1 1 1

8. [Solution on page 234]

 
2 −1 0
Let A = .
0 1 −1
(a) Find a 3 × 2 matrix B such that AB = I2 .
(b) Compute BA and observe that BA 6= I3 .
(c) Is it possible to find a 3 × 2 matrix C such that CA = I3 ? Justify your answer.

9. Let A and B be square matrices of the same size. Suppose that A and B are both
invertible. [Solution on page 235]

(a) Show that AB is invertible, and that B −1 A−1 is its inverse.

(b) Show that AT is invertible, and that (A−1 )T is its inverse.

10. Determine if the statement is True or False, and justify your answer. In the following
the identity matrix of any size is denoted I. [Solution on page 237]
42 Chapter 4. Matrix Algebra

(a) I is not invertible.

(b) If A is invertible, then (A−1 )−1 = A.
(c) If A and B are both invertible, then (AB)−1 = A−1 B −1 .
(d) If A and B are both invertible, then (AB)−1 = B −1 A−1 .
(e) If A is invertible and α 6= 0 is a scalar, then (αA)−1 = αA−1 .
(f ) If A is invertible, then (AT )−1 = (A−1 )T .
(g) If A, B, C are invertible, then (ABC)−1 = C −1 B −1 A−1 .
(h) If A is invertible, it is possible to find a vector b such that the system Ax = b
has infinitely many solutions.
(i) If A is invertible and AB = 0, then B = 0.

11. Let A, B, C be n × n invertible matrices. Solve each equation for X. [Solution on

page 237]

(a) AX = C (d) A−1 X T B = C T

(b) XB = C (e) A−1 (3X + In )T B −1 = CB
(c) AXB −1 = C (f ) A(5X T − B)C = AX T C

12. Find a 2 × 2 matrix X such that [Solution on page 239]

  −1  
T 0 2 1 1 −1
2X + =
−8 4 6 0 1

13. Let A, B, C, and X be n×n matrices with X and B invertible. Consider the equation
XA = B + XC. [Solution on page 240]

(a) Show that A − C is invertible.

(b) Solve the equation XA = B + XC for X.

14. Let A, B, C, and X be n×n matrices with A, C, and X invertible. Solve the following
equation for X. [Solution on page 241]

BAX T = C + 2AX T

(If you need to invert a matrix, explain why that matrix is invertible.)

15. Suppose that A, B, X, are n × n matrices and assume that A and X are invertible.
Solve the equation (XA)−1 (A + X) = B for the matrix X in terms of A and B,
showing your steps. State clearly which matrix you need to assume is invertible.
Prove that it is invertible. [Solution on page 241]

16. Determine if the statement is True or False, and justify your answer. [Solution on
page 242]

(a) A is invertible if and only if A is row equivalent to the identity matrix.

4.2. The Inverse of a Matrix 43

(b) If the equation Ax = 0 has only the trivial solution, then there is a pivot position
in every row of A.
(c) If there is a pivot position in every column of A, then A is invertible.
(d) If A is a square matrix and if there is a pivot position in every column, then A
is invertible.
(e) If A and C are two matrices such that CA = I, then then A is invertible.
(f ) If A and C are square matrices such that CA = I, then A is invertible.
(g) If a linear transformation T (x) = Ax is one-to-one, then A is invertible.
(h) If an n × n matrix A is invertible, then the columns of A span Rn .
(i) If A is a square matrix, the linear transformation T (x) = Ax is one-to-one if
and only if it is onto.
(j) It is possible that A is invertible and AT is not.
(k) If A and B are square matrices such that AB = I, then A and B commute.
(l) It is possible to find a 4 × 4 invertible matrix whose columns do not span R4 .
(m) If A is a square matrix whose columns are not linearly independent, then A is
not invertible.
(n) If the linear transformation T : Rn → Rn is not onto, then it is not one-to-one.

17. Let A and B be n × n matrices with AB invertible. Show that B is invertible.

[Solution on page 244]

18. Let A be an n × n matrix with two identical columns. Is A invertible? Why or why
not? [Solution on page 244]

19. Let A be an invertible matrix. Show that the columns of A−1 are linearly independent.
[Solution on page 244]

20. Let A be a square n × n matrix. Suppose the columns of A span Rn . Show that the
columns of A3 are linearly independent. [Solution on page 244]
44 Chapter 4. Matrix Algebra
 Chapter5

Subspaces of Rn

5.1 Subspaces, Column and Null Spaces

1. Determine if the statement is True or False, and justify your answer. [Solution on
page 245]

(a) Any subset of Rn is a subspace of Rn .

(b) Any subspace of Rn is a subset of Rn .
(c) If a subset H of R2 contains the zero vector, then H is a subspace of R2 .
(d) If H is a subspace of Rn , then H contains the zero vector.
(e) Any line in R2 is a subspace of R2
(f ) If H is a subspace of Rn , it is possible to find two vectors u, v in H such that
u + v is not in H.
(g) If H is a subspace of Rn , it is possible to find a vector u in H and a scalar c
such that cu is not in H.
(h) The only subspaces of R2 are {0} and R2 .
(i) The subset of R2 consisting of the x1 -axis and x2 -axis is a subspace of R2 .
 
2 x1
2. Let H be the subset of R consisting of vectors such that x1 − 3x2 = 0.
x2
[Solution on page 247]

(a) Find two different vectors that are in H.

(b) Find a vector in R2 that is not in H.
(c) Show that H is a subspace of R2 .
 
a − 2b
3. Let H be the subset of R3 consisting of vectors that have the form  a ,
2a + 3b
where a and b are real numbers. [Solution on page 247]

(a) Find four different vectors that are in H.

(b) Find a vector in R3 that is not in H.

45
46 Chapter 5. Subspaces of Rn

(c) Prove that H is a subspace of R3 .

4. In each case, determine if H is a subspace of R2 . [Solution on page 249]

 
2 x1
(a) H is the subset of R consisting of vectors such that x1 + x2 = 1.
x2
 
2 x1
(b) H is the subset of R consisting of vectors such that x1 ≥ 0 and x2 ≥ 0.
x2

5. Let v1 , · · · , vp be vectors in Rn . The subset of Rn generated by these vectors is the set

of all linear combinations of v1 , · · · , vp . This is nothing but Span{v1 , · · · , vp }. Show
that Span{v1 , · · · , vp } is a subspace of Rn . [Solution on page 249]

6. In each case, determine if H is a subspace of R2 . [Solution on page 249]

(a) H is shown in the following figure.

x2

x1

(b) H is shown in the following figure.

x2

x1

(c) H is shown in the following figure.

5.1. Subspaces, Column and Null Spaces 47

x2

x1

(d) H is shown in the following figure.

x2

x1

(e) H is shown in the following figure.

x2

x1

7. Determine if each subset is a subspace of R2 . [Solution on page 250]

48 Chapter 5. Subspaces of Rn

3
H 2
2
S
1

−2 −1 1 2
1 2 3

8. Determine if the statement is True or False, and justify your answer. [Solution on
page 250]

(a) If A is a matrix, the column space of A is the set of all linear combinations of
the columns of A.
(b) If A is an m × n matrix, the column space of A is a subspace of Rn .
(c) If A is an m × n matrix, the column space of A is a subspace of Rm .
(d) The column space of a 3 × 2 matrix can be R2 .
(e) The column space of a 3 × 2 matrix is always a plane in R3 .
(f ) If the columns of a 3 × 3 matrix A are linearly dependent, then Col(A) = R3 .
(g) If the columns of a 3 × 3 matrix A are linearly independent, then Col(A) = R3 .
(h) If the columns of an m×n matrix A are linearly independent, then Col(A) = Rm .
(i) If A is an m × n matrix and if b is a vector in Rm , then the linear system Ax = b
has at least one solution if and only if b is in Col(A).
(j) Let A be a 2 × 2 matrix such that Col(A) is a line in R2 . Let b be a vector in
R2 . If b does not lie on Col(A), the system Ax = b is consistent.
     
1 2 7
9. Let v1 =  −1 , v2 =  −3 , and u =  −9 . [Solution on page 253]
−2 4 2

(a) Determine if u is in the subspace of R3 generated by v1 and v2 .

 
1 2
(b) Let A =  −1 −3 . Determine if u is in the column space of A.
−2 4
   
1 −2 3 −4
10. Let A =  −1 3 1  and let b =  −1 . Determine if b is in Col(A). [Solu-
2 −3 10 −12
tion on page 253]
 
1 4 −5
11. Let A =  1 5 1 . [Solution on page 254]
5 23 −7
5.1. Subspaces, Column and Null Spaces 49

(a) Find five different vectors in Col(A).

(b) Find two different vectors that are not in Col(A).
 
10
(c) Determine if the vector b =  11  is in Col(A).
12
(d) Describe geometrically the column space of A.
12. Let A be an m × n matrix, and let T : Rn → Rm be the linear transformation defined
by T (x) = Ax. [Solution on page 255]
(a) Show that the column space of A is equal to the range of T .
(b) Show that Col(A) = Rm if and only if for every vector b in Rm the system Ax = b
is consistent.
(c) Show that Col(A) = Rm if and only if T is onto.
(d) Suppose m = n. Show that Col(A) = Rn if and only if the matrix A is invertible.
(e) Let B be an n × p matrix. Explain why every column of AB is in Col(A).
(f ) (True or False.) If Col(A) 6= Rm , then the columns of A are linearly dependent.
(g) (True or False.) If Col(A) 6= Rm , then the columns of A are linearly indepen-
dent.  
1 2 4
(h) Suppose A =  3 7 7 . Describe geometrically the column space of
−4 −9 −11
A.
13. Determine if the statement is True or False, and justify your answer. [Solution on
page 258]
(a) If A is a matrix, the null space of A is the set of all vectors x such that Ax = 0.
(b) If A is an m × n matrix, the null space of A is a subspace of Rn .
(c) If A is a 5 × 3 matrix, the null space of A is a subspace of R5 .
(d) The null space of a 2 × 3 matrix can be R3 .
(e) The null space of a 3 × 2 matrix can be R3 .
(f ) The null space of a 3 × 2 matrix can be a line in R2 .
(g) The columns of a matrix A are linearly independent if and only if Null(A) = {0}.
(h) If A is an n × n matrix, then Null(A) = {0} if and only if A is invertible.
 
1 −3 0 2
 −3 9 1 −2 
14. Let A = 
 −1
. [Solution on page 259]
3 1 2 
2 −6 −1 0
 
0
 0 
(a) Determine if the vector u = 
 −4  is in Null(A).


1
50 Chapter 5. Subspaces of Rn

(b) Write Null(A) in vector form.

(c) Find two nonzero vectors in Null(A).
(d) Find two nonzero vectors in Col(A).

5.2 Basis, Dimension, and Rank

1. Determine if the statement is True or False, and justify your answer. [Solution on
page 261]

(a) Let H be a subspace of Rn . A set S = {v1 , · · · , vp } of vectors in H is a basis for

H if S is linearly independent and spans H.
(b) A set containing only one vector in R2 can be a basis for R2 .
(c) A set containing three vectors in R2 can be a basis for R2 .
(d) If v1 , v2 , and v3 are vectors in R2 such that v1 and v2 are linearly independent
and v2 = v3 , then the set {v1 , v2 , v3 } is a basis for R2 .
   
1 0
(e) The set , is the standard basis for R2 .
0 1
(f ) The standard basis for R3 is
     
 1 0 0 
 0 , 1 , 0 
0 0 2
 

   
1 −1
2. Let v1 = and v2 = . [Solution on page 262]
2 3

(a) Show that the set B = {v1 , v2 } is a basis for R2 .

 
−9
(b) Find the coordinates of the vector b = relative to the basis B.
7
     
1 −1 2
3. Let v1 =  0 , v2 =  2 , and v3 =  −3 . Show that the set B = {v1 , v2 , v3 }
−2 −3 4
is a basis for R3 . [Solution on page 265]

4. In each case, determine if the set B is a basis for Rn . [Solution on page 266]
 
18
(a) n = 2, v = , B = {v}.
−25
   
9 −3
(b) n = 2, v1 = , v2 = , B = {v1 , v2 }.
−12 4
   
1 0
(c) n = 3, v1 = 2 , v2 = 1 , B = {v1 , v2 }.
  
0 3
5.2. Basis, Dimension, and Rank 51

     
−4 −1 −5
(d) n = 3, v1 =  3 , v2 =  2 , v3 =  0 , B = {v1 , v2 , v3 }.
7 6 −4
       
−17 29 2 −51
(e) n = 3, v1 =  2 , v2 =  32 , v3 =  −43 , v4 =  −3 , B =
8 9 −12 −61
{v1 , v2 , v3 , v4 }.
     
2 1 1
5. Let v1 =  0 , v2 =  −1 , x =  3 . Let H = Span{v1 , v2 }, and let
1 3 −7
B = {v1 , v2 }. [Solution on page 266]

(a) Show that B is a basis for H.

(b) Determine if x is in H, and if it is, find the coordinate vector of x relative to B.

6. In each case find a basis for Null(A) and a basis for Col(A). [Solution on page 267]
   
1 2 4 5 −30 40
(a) A =  3 7 7  (b) A =  3 −18 24 
−4 −9 −11 −4 24 −32

7. Determine if the statement is True or False, and justify your answer. [Solution on
page 270]

(a) Any nonzero subspace of Rn has more than one basis.

(b) If H is a nonzero subspace of Rn , it is possible to find two bases of H that do
not have the same number of vectors.
(c) If H is a subspace of Rn and if B is basis for H, then the dimension of H is the
number of vectors in B.
(d) For every nonnegative integer n, dim Rn = n.
(e) If H is a plane in R3 , dim H = 1.
(f ) It is possible to find a subspace of Rn whose dimension is greater than n.
(g) It is possible to find a 6 × 7 matrix A whose dim Null(A) = 0.
   
−2 8
(h) If u = and v = , the dimension of the subspace of R2 spanned
3 −12
by u and v is 2.
   
1 3
(i) If u = and v = , the dimension of the subspace of R2 spanned by
−2 4
u and v is 2.
(j) Let H be a subspace of Rn of dimension p. If a subset B = {v1 , · · · , vp } of H is
linearly independent, then B is a basis for H.
(k) Let H be a subspace of Rn of dimension p. If a subset B = {v1 , · · · , vp } of H
spans H, then B is a basis for H.
52 Chapter 5. Subspaces of Rn

(l) If A is a matrix, the dimension of the null space of A is equal to the number of
free variables of the system Ax = 0.
 
2 x1
8. Let H be the subspace of R consisting of vectors such that x1 − 5x2 = 0.
x2
[Solution on page 271]

(a) Find a basis for H.

(b) What is the dimension of H?

9. Let H = Span{v1 , v2 , v3 }, where

     
1 −3 7
v1 =  −1  , v2 =  5  , and v3 =  −13 
−4 10 −22

Find the dimension of H. [Solution on page 271]

 
2a − 3b
10. Let H be the subspace of R3 consisting of vectors that have the form  −b ,
−5a + 4b
where a and b are any scalars. What is the dimension of H? [Solution on page 272]

11. Determine if the statement is True or False, and justify your answer. [Solution on
page 272]

(a) The rank of a matrix A is the dimension of the column space of A.

(b) If A is an m × n matrix, then dim Null(A) + rank(A) = n.
(c) If A is a 4 × 5 matrix, then dim Null(A) + rank(A) = 4.
(d) If A is a 7 × 11 matrix whose dim Null(A) = 5, then the rank of A is 2.
(e) If A is a 7 × 11 matrix whose dim Null(A) = 5, then the rank of A is 6.
(f ) If A is a 15 × 9 matrix whose rank(A) = 7, then the dimension of the null space
of A is 8.
(g) It is possible to find a 4 × 3 matrix whose rank is 4.
(h) It is possible to find a 4 × 7 matrix whose rank is 5.
(i) It is possible to find a 4 × 3 matrix whose rank is 3.
(j) If A is a 5 × 6 matrix and if the system Ax = 0 has two free variables, then the
rank of A is 2.

12. In each case, find (i) a basis for Null(A); (ii) the dimension of Null(A); (iii) a basis
for Col(A); (iv) the dimension of Col(A); (v) the rank of A. [Solution on page 274]
   
1 −1 1 1 2 −1
(a) A =  0 1 1  (b) A =  −1 −1 4 
−1 2 1 −3 −6 3
5.2. Basis, Dimension, and Rank 53

   
1 3 0 2 1 −5 2 0 −1
 0 0 2 −8  (d) A =  3 −15 6 1 −6 
(c) A = 
 −1 −3

2 −10  −1 5 −2 −1 4
2 6 −2 12

 
3 1 −1 2 1
 −3 −1 1 −4 6 
13. Let A = 
 0
. [Solution on page 280]
0 0 0 0 
−6 −2 2 −4 −2

(a) Find the rank of A.

(b) What is the dimension of Null(A)?

14. Let [Solution on page 280]

 
1 −1 0 1 2 −3
 1 −1 1 −1 2 −4 
A=
 −1

1 1 −2 2 4 
2 −2 −2 6 4 −3

(a) Find a basis for the column space of A.

(b) What is the rank of A?
(c) Find five columns of A that are linearly dependent.
(d) Find four columns of A that are linearly independent.
(e) Fill in the blanks in the following lines.
i. Col(A) is a -dimensional subspace of .
ii. Null(A) is a -dimensional subspace of .

15. Let H be a 3-dimensional subspace of R5 , and let K be subspace of R5 containing

H. Suppose that H 6= K and K 6= R5 . What is the dimension of K? Justify your
answer. [Solution on page 281]

16. Let A be a 9 × 14 matrix. Suppose that an echelon form of A has exactly five zero
rows. [Solution on page 281]

(a) Find dim Null(A).

(b) Find dim Col(A).

17. Suppose A is a 5 × 10 matrix that has four pivot columns. Is Col(A) = R4 ? What is
the dimension of Null(A)? Justify your answers. [Solution on page 281]

18. Determine if the statement is True or False, and justify your answer. [Solution on
page 281]

(a) If A is invertible, then Null(A) = {0}.

(b) If Null(A) = {0}, then A is invertible.
54 Chapter 5. Subspaces of Rn

(c) If A is a square matrix and Null(A) = {0}, then A is invertible.

(d) An n × n matrix A is invertible if and only if rank(A) = n.
(e) If A is an n × n matrix, then dim Col(A) = n if and only if dim Null(A) = {0}.
(f ) If A is an n × n matrix, then the columns of A form a basis for Rn if and only
if Col(A) = Rn .
(g) If A is a 3 × 4 matrix such that Col(A) = R3 , then dim Null(A) = 0.

19. Let A be a 7 × 7 matrix. Suppose that the columns of A are linearly independent.
[Solution on page 282]

(a) What is the dimension of Null(A3 )?

(b) Find the rank of A5 .
(c) Do the columns of AT form a basis for R7 ? Why or why not?
 Chapter6

Determinants

6.1 Introduction to Determinants

1. Compute the determinant of each matrix. [Solution on page 285]
   
(a) A = 1 (c) A = −7
   
(b) A = 23 (d) A = 0

2. Compute the determinant of each matrix. [Solution on page 285]

   
3 4 −2 −5
(a) A = (c) A =
5 7 9 −8
   
−4 3 −10 −8
(b) A = (d) A =
5 2 −15 −12

3. Compute the determinant of each matrix by using the definition. [Solution on

page 286]
   
3 4 2 −2 −5 3
(a) A =  −5 −3 1  (b) A =  2 1 −4 
2 −1 −2 6 −6 −7

4. Calculate the determinant of A using cofactor expansion. [Solution on page 287]

   
1 2 3 15 12 −6
(a) A = 2 4 1
  (c) A =  7 0 0 
3 4 2 20 −1 1
   
2 −3 1 12 −3 0
(b) A = −2 4 1  (d) A =  77 0 −5 
0 −5 −1 −8 −1 0

5. Compute det A by cofactor expansions. [Solution on page 290]

55
56 Chapter 6. Determinants

   
0 4 0 −5 −2 −1 1 2 0
 1 −6 −3 5   1 3 2 1 −4 
(a) A =    
 0 7 0 −2  (c) A = 
 0 0 2 0 0 

4 −6 0 2  −1 1 5 6 0 
2 1 6 3 0
 
  3 0 7 −2 −4
2 8 −1 6 
 −2 0 8 0 9 

 −3 21 4 −4  (d) A =  1 −6 −9 2 −8 
(b) A = 
 0 −2
  
0 0   −5 0 6 −1 −3 
5 19 −7 3 0 0 5 0 0

6. Compute the determinant of each matrix. [Solution on page 293]

   
−5 13 −25 3 0 0 0
(a) A =  0 −4 37   89 −7 0 0 
(b) A = 
 −58

0 0 3 7 −1 0 
61 −6 12 −2

7. Determine if the statement is True or False, and justify your answer. [Solution on
page 294]
 
0 0 0
(a) If A = , the determinant of A is 0.
0 0 0
(b) The determinant of a matrix A is defined if and only if A is a square matrix.
(c) The cofactor expansion of det A across a row is not equal to the cofactor expan-
sion down a column.
(d) The cofactor expansion of det A does not depend on the row or column we
choose.
 
5 2
(e) The determinant of is 30 − 14 = 16.
−7 6
 
0 4 5
(f ) The determinant of  −5 1 2  is
0 2 3

4 5
−5 = −5(12 − 10) = −10
2 3
 
a b
(g) If A = , det(3A) = 9 det A.
c d
 
4 9 −6
(h) The matrix  0 2 −2  is lower triangular.
0 0 −3
(i) The determinant of a triangular matrix is the product of the entries on the main
diagonal.
6.2. Properties of Determinants 57

6.2 Properties of Determinants

1. In each case, the matrix B is obtained from A by performing the indicated elementary
row operation. Express the determinant of B in terms of det A. [Solution on
page 294]

(a) A R2 − R1 → R2 B
(b) A R3 + 5R1 → R3 B
(c) A R2 ↔ R3 B
(d) A 3R2 → R2 B
1
(e) A R
4 1
→ R1 B

2. Let A and B be square matrices. Suppose that B is obtained from A by performing

the following elementary row operations: [Solution on page 295]

4
R1 ↔ R3 , R2 + 6R1 → R2 , −3R2 → R2 , R3 − 7R1 → R3 , R
5 3
→ R3

(a) Express the determinant of A in terms of det B.

(b) Suppose det B = −60. Find det A.

3. Compute the determinant of each matrix by row reduction to echelon form. [Solution
on page 296]
   
5 −10 10 0 −5 −3 −8
(a) A =  −2 8 1   −1 4 0 3 
(b) A =  
3 −10 4  1 −1 2 3 
2 2 4 13

 
0 1 4 −2
 −3 7 −1 −5 
4. Let A =  . Compute det A. [Solution on page 297]
 0 2 3 −2 
3 −7 6 8

5. Compute the determinant of each matrix. [Solution on page 298]

   
5 −10 10 0 1 4 1
(a) A =  0 0 0   −3 7 −1 7 
(c) A =  
3 −10 4  0 2 3 2 
3 −7 6 −7
 
  0 −1 4 −2
4 −11 0  −3 7 −1 −5 
(d) A =  
(b) A =  −7 9 0   0 3 −12 6 
11 −10 0 3 −7 6 8
58 Chapter 6. Determinants

 
a b c
6. Let A =  d e f . Suppose det A = 7. Find the determinant of the following
g h i
matrices. [Solution on page 299]
   
2a 2b 2c 3d 3e 3f
B =  d − a e − b f − c  and C =  a + d b + e c + f 
g + 3a h + 3b i + 3c g − 2d h − 2e i − 2f

7. In each case, use the determinant to determine whether the set S is linearly indepen-
dent. [Solution on
page 300]
    
2 1 4
(a) S = {v1 , v2 , v3 }, where v1 = 0 , v2 = 9 , and v3 = 3 .
    
1 2 3
(b) S = {v1 , v2 , v3 , v4 }, where
       
1 0 3 2
 −1
 , and v4 =  −8 
  5   1   
v1 =   2
 , v2 = 
  4  , v3 =  3
 
  12 
−3 −8 −3 −22
 
k 0 3
8. Let A =  7 1 −8 . Find the values of k for which A is invertible. [Solution
2 0 k+1
on page 301]

9. Let A and B be matrices such that det A = 2 and det B = −3. [Solution on
page 301]

(a) Compute det(AB).

(b) Compute det (A6 ).


(c) Compute det ABAT .
(d) Compute det AT B −1 .

10. Let A be a square matrix such that A2 = A. (Such a matrix is called idempotent.)
Show that det A = 0 or det A = 1. [Solution on page 302]

11. Construct two matrices A and B such that det(A + B) 6= det A + det B. [Solution
on page 302]

12. Determine if the statement is True or False, and justify your answer. [Solution on
page 302]

(a) If B is obtained from A by performing the operation R2 + 5R1 → R1 , then

det B = det A.
6.2. Properties of Determinants 59

(b) If B is obtained from A by performing the operation R1 ↔ R3 , then det B =

− det A.
(c) If B is obtained from A by performing the operation 3R1 → R1 , then det A =
1
3
det B.
(d) If A is a square matrix that has a zero row or a zero column, then det A = 0.
(e) If det A = 0, then A has a zero row or a zero column.
(f ) If A is a square matrix, and if two columns (or rows) of A are identical, then
det A = 0.
(g) If det A = 0, then two columns (or rows) of A must be identical.
(h) If A and B are square matrices, then det(AB) = (det A)(det B).
(i) If A and B are square matrices, then det(A + B) = det A + det B.
(j) If A is invertible, then det (A−1 ) = 1
det A
.
T
(k) If A is a square matrix and if A = A , then det A = 0.
(l) If A is a square matrix and if A3 = A, then det A = 0 or det A = ±1.
(m) If A is a 3 × 3 matrix and if A3 = −A, then det A = 0.
(n) If A is a 3 × 3 matrix and if det A = 0, then for every b in R3 the system Ax = 0
has no solution.
60 Chapter 6. Determinants
 Chapter7

Eigenvalues and Eigenvectors

7.1 Eigenvectors and Eigenvalues

1. In each case, determine if u and v are eigenvectors of A. If so, find the corresponding
eigenvalues. [Solution on page 305]
     
3 −3 −3 1
(a) A = , u= , and v = .
−2 2 2 −1
     
1 0 −1 3 1
(b) A = 1 −3
 0 , u =  0 , and v =  1 .
4 −13 1 2 3

2. Let A be an n × n matrix. [Solution on page 306]

(a) Show that 0 is an eigenvalue of A if and only if A is not invertible.

(b) Let x be an eigenvector of A corresponding to an eigenvalue λ, and let k be a
real number such that k 6= 0. Show that kx is an eigenvector of A corresponding
to λ.
 
3 −3
3. Let A = . An eigenvalue of A is λ = 5. Find the corresponding eigen-
−2 2
vectors. [Solution on
page 306]

4. In each case, find a basis for the eigenspace corresponding to the eigenvalue λ. [So-
lution on page 307]
 
3 −3
(a) A = , λ = 5.
−2 2
 
0 1 0
(b) A =  3 0 1  , λ = −1.
2 0 0
 
0 3 1
(c) A =  1 −2 −1  , λ = 1.
−3 9 4

61
62 Chapter 7. Eigenvalues and Eigenvectors

5. Let A be a square matrix such that A2 = 0. Show that if λ is an eigenvalue of A,

then λ = 0. [Solution on page 309]
6. Let A be a square matrix. Let λ be an eigenvalue of A and let u be an eigenvector
corresponding to λ. Show that for every integer n ≥ 1, λn is an eigenvalue of An
associated with u. [Solution on page 309]
7. Let A be an invertible matrix. Suppose that λ is an eigenvalue of A with corresponding
eigenvector x. Show that λ−1 must be an eigenvalue of A−1 , with the same eigenvector.
[Solution on page 310]
8. Let T : R2 → R2 be the reflection through the line x2 = −x1 . Without writing
the standard matrix of T , find the eigenvalues of T and describe the corresponding
eigenspaces. [Solution on page 310]
9. Let T : R3 → R3 be the projection onto the x1 x2 -plane. Without writing the standard
matrix of T , find an eigenvalue of T and describe the corresponding eigenspace.
[Solution on page 310]
10. Let T : R2 → R2 be the counterclockwise rotation of angle π2 about the origin. With-
out finding the standard matrix, explain why T has no eigenvector. [Solution on
page 311]
11. Determine if the statement is True or False, and justify your answer. [Solution on
page 311]
(a) If A is a square matrix, a nonzero vector x is an eigenvector of A if Ax is a
multiple of x.
(b) An eigenvector can be the zero vector.
   
−2 7 4
(c) The vector u = is an eigenvector of A = .
3 −3 −1
   
−1 3 −4
(d) The vector u = is an eigenvector of A = .
2 −4 5
(e) If λ is an eigenvalue of A, the eigenspace corresponding to λ is Null(A − λI).

7.2 Finding Eigenvalues

1. In each case, determine if λ is an eigenvalue of A. [Solution on page 311]
 
5 4
(a) A = , λ=2
−6 −6
 
−1 2
(b) A = , λ = −4.
2 −1
 
0 3 0
(c) A =  3 0 0 , λ = 7.
4 −5 7

2. In each case, find the eigenvalue(s) of A. [Solution on page 313]

7.2. Finding Eigenvalues 63

   
5 −3 0 4 0
(a) A =
−10 6 (d) A =  1 0 0 
  0 0 1
4 7 −5  
−1 1 4
(b) A =  0 −2 5 
(e) A =  0 2 0 
0 0 3
3 −3 −2
   
3 0 0 0 3 1
(c) A =  −1 2 0  (f ) A =  1 −2 −1 
8 5 3 −3 9 4

3. Construct a 3 × 3 matrix A with eigenvalues 1, 2, and 3. [Solution on page 314]

4. Let A be a 2 × 2 matrix. Define tr(A) as the sum of the entries on the main diagonal
of A. (tr(A) is called the trace of A.) [Solution on page 314]

(a) Show that the characteristic polynomial of A is

det(A − λI) = λ2 − tr(A)λ + det A

(b) Let λ1 and λ2 be the eigenvalues of A. Show that

λ1 + λ2 = tr(A) and λ1 λ2 = det A

(c) If an eigenvalue of A is λ = 3 and tr(A) = −5, what is the other eigenvalue of

A?
(d) If the characteristic polynomial of A is λ2 − λ, what can you say about the
invertibility of A?

5. Let A be a 2 × 2 symmetric matrix. Show that all the eigenvalues of A are real
numbers. [Solution on page 316]

6. Determine if the statement is True or False, and justify your answer. [Solution on
page 317]

(a) An eigenvalue can be 0.

 
4 2 −3
(b) The number λ = 0 is an eigenvalue of A =  −1 5 −2 .
3 7 −5
(c) If A is a square matrix, a number λ is an eigenvalue of A if and only if det(A −
λI) = 0.
(d) The characteristic equation of a square matrix A is the equation det(A−λI) = 0.
(e) If A is a square matrix and if the columns (or rows) of A are linearly dependent,
then λ = 0 is an eigenvalue of A.
(f ) If A is a symmetric matrix, then all the eigenvalues of A are real numbers.
(g) If all the eigenvalues of A are real numbers, then A is a symmetric matrix.
64 Chapter 7. Eigenvalues and Eigenvectors

7. Let A be a square matrix. Suppose that the characteristic polynomial of A is [Solution

on page 317]
det(A − λI) = (λ − 5)(λ − 2)4 (λ + 1)3

(a) What is the size of A?

(b) Find the eigenvalues of A.
(c) Determine whether A is invertible.

7.3 Diagonalization
1. Determine if the statement is True or False, and justify your answer. [Solution on
page 318]
 
2 0 0
(a) The matrix D = 4 0 0 is a diagonal matrix.
0 0 7
 
−3 0 0
(b) The matrix D =  0 −1 0  is a diagonal matrix.
0 0 8
   
1 0 n 1 0
(c) If D = , then D = for every n ≥ 1.
0 5 0 5n
(d) A square matrix A is diagonalizable if there exist a diagonal matrix D and an
invertible matrix P such that A = P DP −1 .
(e) A 2 × 3 matrix can be diagonalizable.
(f ) If A is a diagonal matrix, then A is diagonalizable.

2. Diagonalize each matrix, if possible. [Solution on page 318]

   
9 6 0 3 1
(a) A =
−7 −4 (d) A =  1 −2 −1 

2 −3
 −3 9 4
(b) A =
3 8
   
−2 5 5 2 3 0
(c) A =  0 3 0  (e) A =  3 2 0 
2 −2 1 2 3 −1

 
−1 2 −2
3. Let A =  0 1 0 . [Solution on page 327]
4 −4 5

(a) Show that A is diagonalizable, and find an invertible matrix P and a diagonal
matrix D such that A = P DP −1 .
(b) Compute Ak for an arbitrary integer k ≥ 1.
7.3. Diagonalization 65

4. Find a 2 × 2 matrix A satisfying the following two conditions: [Solution on page 331]
 
4
ˆ λ = 1 is an eigenvalue of A with eigenvector v1 = .
5
 
−1
ˆ λ = −2 is an eigenvalue of A with eigenvector v2 = .
4

5. Suppose that the characteristic polynomial of a matrix A is λ(λ−2)3 (λ+7)4 . [Solution

on page 331]

(a) What is the size of A?

(b) Find the eigenvalues of A and their corresponding multiplicities.
(c) What must the dimension of each of the eigenspaces be in order for A to be
diagonalizable?
(d) Determine whether A is invertible.

6. Let A be a 4 × 4 matrix with two distinct eigenvalues. Each eigenspace if two-

dimensional. Determine if A is diagonalizable. [Solution on
page 331]

7. Determine if the statement is True or False, and justify your answer. [Solution on
page 332]

(a) An n × n matrix that has n distinct eigenvalues is diagonalizable.

(b) An n × n matrix A is diagonalizable if and only if Rn has a basis of eigenvectors
of A.
(c) Let A be a 3 × 3 matrix with eigenvalues λ1 (with multiplicity 2) and λ2 (with
multiplicity 1). If dim Null(A − λ1 I) = 1 and dim Null(A − λ2 I) = 1, then A is
diagonalizable.
(d) If A is invertible and if A is diagonalizable, then A−1 is diagonalizable.
(e) It is possible to find an invertible matrix which is not diagonalizable.
(f ) It is possible to find a diagonalizable matrix which is not invertible.
(g) If A is diagonalizable, then so is AT .
66 Chapter 7. Eigenvalues and Eigenvectors
 PartII

Solutions to Exercises

67
 Chapter1

Systems of Linear Equations

1.1 Introduction to Systems of Linear Equations

1. [Exercise on page 3]

Linear Equations
A linear equation is an equation of the form

a1 x 1 + a2 x 2 + · · · + an x n = b (1.1.1)

where

ˆ n is a positive integer;

ˆ a1 , a2 , · · · , an are real (or complex) numbers called coefficients;

ˆ b is a real (or complex) number called the constant term;

ˆ x1 , x2 , · · · , xn are variables.

For example, the equation 4x1 + 5x2 + 6x3 = 7 is a linear equation in n = 3

variables. The coefficients are a1 = 4, a2 = 5, and a3 = 6, and the constant
term is b = 7.

(a) The equation x1 + x2 = 0 is of the form

a1 x 1 + a2 x 2 = b
where a1 = 1, a2 = 1, and b = 0. So it is a linear equation in n = 2 variables.
The coefficients are 1 and 1. And the constant term is 0.
(b) The equation 3x1 − 6x2 = 11 is linear. The number of variables is 2, the
coefficients are 3 and −6, and the constant term is 11.
(c) The equation −2x1 + 4x2 − 5x3 = 12 is linear. The number of variables is 3, the
coefficients are −2, 4, and −5, and the constant term is 12 .
(d) The equation 4x1 − x2 + 8x4 = 15 is of the form
a1 x 1 + a2 x 2 + a3 x 3 + a4 x 4 = b

69
70 Chapter 1. Systems of Linear Equations

where a1 = 4, a2 = −1, a3 = 0, a4 = 8, and b = 15. So it is a linear equation

in n = 4 variables. The coefficients are 4, −1, 0, and 8, and the constant term is
15.
(e) The equation −2x1 − 14x3 + 4x5 = 2021 is linear. The number of variables is
5, the coefficients are −2, 0 −14, 0, and 4, and the constant term is 2021.
(f ) The equation x1 − x2 + x23 = 0 is not linear. Indeed, it is not of the form (1.1.1)
because of the term x23 . Actually, the term x23 is quadratic.
√
(g) The equation x1 − x2 = 19 is not linear. Indeed, it is not of the form (1.1.1)
√
because of the term x1 .
3x1 +6x3
(h) The equation 4
= 1 can be rewritten as 34 x1 + 64 x3 = 1 or

3 3
x1 + x3 = 1
4 2

So it is linear equation in three variables. The coefficients are 34 , 0, and 32 , and

the constant term is 1.
(i) The equation −4x1 − 20x2 + cos(x3 ) = −1 is not linear. The term that makes
this nonlinear is cos(x3 ).
√
(j) The equation 5x2 + x5 √ 2 = −47 is linear. The number of variables is 5, the
coefficients are 0, 5, 0, 0, 2, and the constant term is −47.
(k) The equation −x1 + 9x2 x4 − x5 = −3 is not linear. The term that makes this
nonlinear is x2 x4 as the definition does not allow products of variables.
1 1
(l) The equation x1 +x2
− x3 = 24 is not linear because of the term x1 +x2
.

2. [Exercise on page 3]

Solution to a Linear Equation

ˆ An n−tuple is an ordered list (s1 , s2 , · · · , sn ) of numbers. Here the term

“ordered” means that if we switch two numbers of the list, we get a
different list. For instance, (1, 2) is not the same as (2, 1). One can think
of this as a point. For example, (1, 2) can be thought of as the point of
the plane whose x-coordinate is 1 and y-coordinate is 2.

ˆ A solution to a linear equation

a1 x1 + a2 x2 + · · · + an xn = b

is an n-tuple (s1 , s2 , · · · , sn ) that satisfies the equation. This means that

if we substitute x1 with s1 , x2 with s2 , · · · , xn with sn into a1 x1 + a2 x2 +
· · · + an xn , we get b.
For example, the point (5, 6) is a solution to (or satisfies) the equation
3x1 + 4x2 = 39 since 3(5) + 4(6) is 39.
1.1. Introduction to Systems of Linear Equations 71

(a) ˆ Substituting x1 with 3 and x2 with 3 into 2x1 + 3x2 , we get

2(3) + 3(3) = 6 + 9 = 15
Since the equation is 2x1 + 3x2 = 14 and since 2(3) + 3(3) is not equal to
14, it follows that (3, 3) is not a solution.
ˆ Substituting x1 with 1 and x2 with 4 into 2x1 + 3x2 , we get
2(1) + 3(4) = 2 + 12 = 14
So (1, 4) is a solution to the equation 2x1 + 3x2 = 14.
ˆ Substituting x1 with 5 and x2 with 1 into 2x1 + 3x2 , we get 2(5) + 3(1) = 13.
So (5, 1) is not a solution to the equation 2x1 + 3x2 = 14.
(b) Consider the following table.

(s1 , s2 , s3 ) Equation Substituting xi with si , 1 ≤ i ≤ 3

(−2, 3, 0) −4x1 − 5x2 + 21x3 = 7 −4(−2) − 5(3) + 21(0) = −7
(−2, −3, 0) −4x1 − 5x2 + 21x3 = 7 −4(−2) − 5(−3) + 21(0) = 23
(2, −3, 0) −4x1 − 5x2 + 21x3 = 7 −4(2) − 5(−3) + 21(0) = 7

From the table, only the point (2, −3, 0) is a solution to the equation −4x1 −
5x2 + 21x3 = 7.

3. [Exercise on page 3]

Systems of Linear Equations

A system of linear equations (or just linear system) is a finite collection

of linear equations. Examples include the following systems.

x1 − x2 + x3 − x4 = 0
3x1 − 4x2 = −1
2x1 − 3x2 − 5x4 = 2
5x1 + 7x2 = 12
x2 − x3 + x4 = −3

The first has two variables and two equations, while the second has four vari-
ables and three equations.
Note. In a linear system every equation is linear. If an equation is not linear,
the system is not a linear system.

(a) Here the system is a linear system since every equation is linear.
(b) The system here is a linear system.
(c) The system here is not a linear system since the third equation, −2x1 + ex3 =
−2, is not linear.
√
(d) Here the system is not a linear system since the first equation, 34x1 − 5x2 =
−4, (as well as the second equation) is not linear.

4. [Exercise on page 3]
72 Chapter 1. Systems of Linear Equations

Solution to a System of Linear Equations

Suppose we have a linear system in n variables with m equations. An n-tuple
(s1 , s2 , · · · , sn ) is a solution to that system if it satisfies every equation of the
system. For example, consider the following system.

x1 − x2 = 1
x1 + x2 = 3

ˆ The point (2, 1) is a solution since it satisfies both equations (2 − 1 = 1

and 2 + 1 = 3).

ˆ The point (4, 3) satisfies the first equation (4 − 3 = 1) but it is not

a solution to the system since it does not satisfy the second equation
(4 + 3 = 7 6= 3).

(a) We need to verify whether (1, −1) is a solution to each equation.

ˆ Substituting x1 with 1 and x2 with −1 into the equation x1 − 2x2 = 3, we
get 1 − 2(−1) = 3, that is, 3 = 3. This implies that (1, −1) is a solution to
the first equation.
ˆ Substituting x1 with 1 and x2 with −1 into the second equation, 3x1 +x2 = 2,
we get 3(1) + (−1) = 2, that is, 2 = 2. This implies that (1, −1) is a solution
to the second equation.
Since (1, −1) is a solution to every equation of the system, it follows that (1, −1)
is a solution to the system.
(b) ˆ Substituting x1 with −2 and x2 with 3 into the first equation, we get
−3(−2) + 4(3) = 18, that is, 18 = 18. This shows that (−2, 3) is a so-
lution to the first equation.
ˆ Substituting x1 with −2 and x2 with 3 into the second equation, we get
5(−2) − 6(3) = −27, that is, −28 = −27, which is not true. So (−2, 3) is
not a solution to the second equation.
Since (−2, 3) does not satisfy all the equations, it follows that (−2, 3) is not a
solution to the system.
(c) Consider the following table.

(s1 , s2 , s3 ) Equation Substituting xi with si , 1 ≤ i ≤ 3

(0, −1, 2) x1 − x2 + x3 = 3 0 − (−1) + 2 = 3
(0, −1, 2) 2x1 + 3x2 − x3 = −5 2(0) + 3(−1) − 2 = −5
(0, −1, 2) −3x1 + 4x2 + 2x3 = 1 −3(0) + 4(−1) + 2(2) = 0

This table shows that (0, −1, 2) satisfy the first two equations, but not the third.
So (0, −1, 2) is not a solution to the system.
(d) Consider the following table.
1.1. Introduction to Systems of Linear Equations 73

(s1 , s2 , s3 , s4 ) Equation Substituting xi with si , 1 ≤ i ≤ 4

(2, −1, 3, 1) x1 + 2x3 − x4 = 7 1(2) + 2(3) − (1) = 7
(2, −1, 3, 1) 2x1 + x2 − 4x3 = −9 2(2) + (−1) − 4(3) = −9
(2, −1, 3, 1) −5x2 + 10x4 = 15 −5(−1) + 10(1) = 15

From the table, we see that (2, −1, 3, 1) satisfies each equation of the system.
So (2, −1, 3, 1) is a solution to the system.

5. [Exercise on page 4]

Consistent and Inconsistent Systems

A linear system is said to be:

ˆ consistent if it has at least one solution;

ˆ inconsistent if it has no solution.

(a) First, observe that each equation of the system

x1 + x2 = 1
−x1 + x2 = −1

is the equation of a line (a linear equation in one or two variables). To sketch

the graph of a line, we need two points.

ˆ Finding two points on the line x1 + x2 = 1. The idea is to assign a number

to one of the variables and find the other variable. For example,
– if x1 = 0, then the equation x1 + x2 = 1 becomes 0 + x2 = 1 or x2 = 1.
So the point (0, 1) belongs to the line x1 + x2 = 1.
– If x2 = 0, then x1 + 0 = 1 or x1 = 1. So the line x1 + x2 = 1 passes
through the point (1, 0).
Using the points (0, 1) and (1, 0), we get the graph of x1 + x2 = 1 as shown
below.
ˆ Finding two points on the line −x1 + x2 = −1.
– If x1 = 0, then x2 = −1. This gives the point (0, −1).
– If x2 = 0, then −x1 = −1 or x1 = 1. This gives the point (1, 0).
Using these points, we get the graph of −x1 + x2 = −1 as shown below.
74 Chapter 1. Systems of Linear Equations

x2

−x1 + x2 = −1

1 3 x1
−1

x1 + x2 = 1

From the figure, the lines intersect at (1, 0), which is the solution to the given
system. Since the system has at least one solution, it is consistent.
(b) Sketching the graphs of −x1 + 2x2 = 3 and 2x1 − 4x2 = −5, we get the following
figure.
x2

−x1 + 2x2 = 3

2x1 − 4x2 = −5

−1 1 x1

Since the two lines are parallel, there is no intersection. This means that the
system has no solution. In other words, the system is inconsistent.
(c) The figure below shows the graphs of 2x1 + 3x2 = 5 and −3x1 + x2 = 9.

x2
10 −3x1 + x2 = 9

3
2x1 + 3x2 = 5

−2 2 4 x1

From the figure, the system is consistent because the lines intersect. The point
of intersection is (−2, 3), which is also the unique solution to the system.
1.1. Introduction to Systems of Linear Equations 75

(d) First, observe that if we divide both sides of the second equation, −2x1 + 2x2 =
−4, by −2, we get the first equation. This means that the two equations have
the same graph as shown in the following figure.

x2

2 x1 − x2 = 2

−2 2 x1

−2
−2x1 + 2x2 = −4

Since the graphs of the two equations coincide, the system has infinitely many
solutions. So it is consistent and a solution is for example the point (2, 0).
Actually, every point on the line is a solution to the system.

6. [Exercise on page 4]
We want to form a system in two variables with two equations whose every coefficient
is nonzero and (5, −7) is a solution. The idea is to take combinations of 5 and −7.

ˆ For example, 5 + (−7) = −2. This means that (5, −7) is a solution to the
equation x1 + x2 = −2.
ˆ From the combination 5(2) + 3(−7) = −11, it follows that (5, −7) is a solution
to the equation 2x1 + 3x2 = −11.

So (5, −7) is a solution to the following system.

x1 + x2 = −2
2x1 + 3x2 = −11

Note. If you consider different combinations, you will end up with a different system
(which is fine).

7. [Exercise on page 4]
As before we need to consider combinations of −2, 3, and −5.

ˆ From the combination (−2) + (3) + (−5) = −4, it follows that (−2, 3, −5) is a
solution to the equation x1 + x2 + x3 = −4.
ˆ From the combination 2(−2) + (3) + (−5) = −6, it follows that (−2, 3, −5) is a
solution to the equation 2x1 + x2 + x3 = −6.
ˆ From the combination 3(−2) + (3) + (−5) = −8, it follows that (−2, 3, −5) is a
solution to the equation 3x1 + x2 + x3 = −8.
76 Chapter 1. Systems of Linear Equations

So (−2, 3, −5) is a solution to the following system.

x1 + x2 + x3 = −4
2x1 + x2 + x3 = −6
3x1 + x2 + x3 = −8
Of course, this is not the unique answer.

8. [Exercise on page 4]
It is not possible to find a point (a, b) that satisfies the system. Indeed, a solution
to the system is a list (s1 , s2 , s3 ) of three numbers (not two numbers!) since the sys-
tem has three variables.

1.2 Echelon Matrices

1. [Exercise on page 4]

What is a Matrix?
A matrix can be defined as rectangular array of numbers organized in rows
and columns. Examples of matrices (plural of matrix) include:
   
1 2 3 0 0 2 15
 4 0 0  and  0 0 0 0 
−1 −3 6 −6 2 1 0

ˆ The first matrix

 has three rows and
 three columns.
 The first row
is 1 2 3 , the second row is 4  0 0 , and the third row is
  1
−1 −3 6 . The first column is  4 , the second column is
    −1
2 3
 0 , and the third column is  0 .
−3 6
ˆ The second matrix has three rows and four columns.

The numbers that appear in a matrix are referred to as entries.

Zero Rows, Nonzero Rows, and Leading Entries

ˆ A zero row in a matrix is a row where every entry is 0.

ˆ A nonzero row in a matrix is a row where at least one entry is a nonzero

number.

ˆ A leading entry of a row is the first nonzero entry from the left.
1.2. Echelon Matrices 77

 
7 2 3 4
For example, consider the matrix 0 0 0 0. The second row is a zero row,
0 5 0 9
while row 1 and row 3 are both nonzero rows. In row 1, the leading entry is
7. In row 2, there is no leading entry because by definition a leading entry is a
nonzero number (and every entry in row 2 is 0). In row 3, the leading entry is
5.

(a) For the matrix

 
3 1 4 0 0
 0 2 0 0 0 
 
 0 0 0 0 0 
0 0 0 1 0

all the rows are nonzero rows except the third row. The leading entry of row
1 is 3, the leading entry of row 2 is 2, and the leading entry of row 3 does not
exist (DNE) since every entry in that row is 0 and by definition a leading entry
is a nonzero number. These are summarized in the following table.

Zero Row? (Yes or No) Nonzero Row? (Yes or No) Leading Entry
Row 1 No Yes 3
Row 2 No Yes 2
Row 3 Yes No DNE
Row 4 No Yes 1

 
0 0 0 0 0
 8 0 0 0 5 
(b) For the matrix  , we have the following table.
 0 0 −7 7 9 
0 0 0 0 0

Zero Row? (Yes or No) Nonzero Row? (Yes or No) Leading Entry
Row 1 Yes No DNE
Row 2 No Yes 8
Row 3 No Yes -7
Row 4 Yes No DNE

2. [Exercise on page 5]
78 Chapter 1. Systems of Linear Equations

Echelon Forms and Echelon Matrices

A matrix is said to be in echelon form (or row echelon form – abbreviated
REF) if it satisfies the following conditions:

(1) Every nonzero row is above any zero rows. (This amounts to saying that
all zero rows are at the bottom.)

(2) Every leading entry is strictly to the right of the leading entry of the row
above it.

A matrix in echelon form is called an echelon matrix.

 
3 0 4
(a) The matrix A =  0 2 5  satisfies conditions (1) and (2). Indeed,
0 0 6
ˆ there is no nonzero rows in A. This implies that (1) is satisfied.
ˆ The leading entries are 3 (in row 1), 2 (in row 2), and 6 (in row 3). The
leading entry 6 is to the right of the leading entry of the row above it (which
is 2), and the leading entry 2 is to the right of the leading entry of the row
above it (which is 3). There is no rows above row 1. So condition (2) is
satisfied.
Thus, the matrix is in echelon form.
 
−1 1 2
(b) Likewise, the matrix  0 −2 4  is in echelon form.
0 0 0
 
4 0 0 0
(c) However, the matrix  0 0 0 0  is not in echelon form because (1) is
0 0 8 0
not satisfied as the zero row (row 2) is not at the bottom.
 
−7 0 1 0 2
 0 0 4 3 2 
(d) The matrix   0 6 0 0 6  is not in echelon form because it violates


0 0 0 0 1
condition (2). Indeed, the leading entry 6 in row 3 is not to the right of the
leading entry of the row above it (which is 4).
 
−9 0 3 0 5
 0 8 4 3 2 
(e) The matrix   0 2 0 0 −2  is not in echelon form because condition


0 0 0 0 0
(2) is violated. Indeed, the leading entry 2 in row 3 is not to the right of the
leading entry of the row above it (which is 8).

3. [Exercise on page 5]
1.2. Echelon Matrices 79

Reduced Echelon Forms and Reduced Echelon Matrices

A matrix is said to be in reduced echelon form (or reduced row echelon
form – abbreviated RREF) if it satisfies conditions (1) and (2) above, and the
following conditions:

(3) Each leading entry is 1.

(4) Each leading entry is the only nonzero entry in the column containing it.

In other words, a matrix is in reduced echelon form if it is an echelon matrix

satisfying (3) and (4).

A matrix in reduced echelon form is called a reduced echelon matrix.

Yes, it is possible to find a matrix in echelon form which is not in reduced echelon
form. For example the matrix  
1 3
0 2
is in echelon form, but not in reduced echelon form because it violates condition (3)
(it also violates condition (4)).

4. [Exercise on page 5]
 
1 0 0
(a) The matrix  0 1 0  is clearly in echelon form and in reduced echelon
0 0 1
form.
 
1 0 0 0
(b) The matrix  0 0 1 0  is not in reduced echelon form because it is
1 0 0 0
not even an echelon matrix (since it violates condition (2)).
 
1 0 1 0
(c) The matrix  0 0 1 0  is in echelon form, but not in reduced echelon
0 0 0 1
form because it violates condition (4). Indeed, the leading 1 in row 2 is not the
only nonzero entry in the column containing it.
 
0 1 0 0 4
(d) One can easily check that the matrix  0 0 0 1 6  is in echelon form
0 0 0 0 0
and in reduced echelon form.
 
1 0 0 1
(e) The matrix  0 0 5 0  is in echelon form, but not in reduced echelon
0 0 0 0
form because it does not satisfy condition (3) as the leading entry in row 2, 5,
is not 1.
80 Chapter 1. Systems of Linear Equations

 
1 0 0 0
(f ) The matrix  0 0 0 0  is not in reduced echelon form since it is not
0 0 0 1
even en echelon matrix (it does not satisfy condition (1)).
 
1 7 0 −2 0
(g) The matrix  0 0 1 8 0  is in echelon form and in reduced echelon
0 0 0 0 1
form since it satisfies conditions (1), (2), (3), and (4).
 
1 0 −5 1 0
(h) The matrix  0 1 0 0 0  violates condition (4) as the leading entry in
0 0 0 1 0
row 3, which is 1, is not the only nonzero entry in column 4. So it is not in
reduced echelon form. But is an echelon matrix.

5. [Exercise on page 5]

Elementary Row Operations

There are three elementary row operations.

(i) Interchange two rows.

(ii) Multiply a row by a nonzero number.

(iii) Replace a row with the sum of itself and a multiple of another row.

Notation. We use the following notation. We write Ri for the ith row. For
example, R1 will represent the first row at each stage, R2 will represent the
second row at each stage, and so on. Using this notation, the elementary row
operations become:

(i) Ri ↔ Rj . This means that we interchange Ri and Rj .

(ii) aRi → Ri , where a is a nonzero number. This means that we multiply

the row Ri by a and we put the answer in row Ri . In other words, Ri is
replaced by aRi .

(iii) Ri + bRj → Ri , where b is a number (not necessarily a nonzero number).

This means that we replace the row Ri with Ri + bRj . For example,
R2 + 3R1 → R2 means that we calculate R2 + 3R1 and we put the result
in R2 .

Warning. Any row operation which is not in one of the forms Ri ↔ Rj , aRi →
Ri , or Ri + bRj → Ri is not an elementary row operation. For example,

ˆ 2R1 ↔ R2 is not an elementary row operation. Actually it is the combi-

nation of two elementary row operations: 2R1 → R1 and R1 ↔ R2 .

ˆ 4R3 → R2 is not a valid operation. According to (ii), if we multiply R3

by a number, we have to put the result in R3 .
1.2. Echelon Matrices 81

ˆ 5R3 + R1 → R3 is not an elementary row operation. Actually, it is the

combination of two elementary row operations: 5R3 → R3 and R3 +R1 →
R3 . According to (iii), aRi + bRj → Rk is an elementary row
operation if and only if either

? i = k and a = 1 or
? j = k and b = 1.

If an operation is an elementary row operation, we will write yes next to it. Otherwise,
we will write no.

(a) R2 + 2R1 → R2 yes (h) R2 + 3R1 → R2 yes (o) 3R1 ↔ R2 no

(b) R3 − R1 → R3 yes (i) 3R2 + R1 → R2 no (p) R1 ↔ 5R2 no
(c) R1 − 4R2 → R1 yes (j) 3R2 + R1 → R1 yes (q) 2R1 → R1 yes
(d) R4 + 0 · R3 → R4 yes (k) 4R3 − R1 → R1 no (r) −5R2 → R2 yes
(e) R2 + R1 → R1 yes (l) 4R3 + 5R1 → R3 no (s) 0 · R1 → R1 no
(f ) R2 + 3R1 → R1 no (m) R1 + R2 → R3 no (t) 2R1 → R3 no
(g) R2 + 3R1 → R4 no (n) R1 ↔ R2 yes (u) R3 → R4 no

6. [Exercise on page 6]

Performing an Operation on a Matrix

How to multiply a row by a number? How to add a multiple of one row to
a different row? Let us answer these questions with an example. Suppose
R1 = [1 2] and R2 = [3 − 5].

• R1 + R2 is [(1 + 3) (2 + (−5))], which is equal to [4 − 3].

• R2 − 3R1 is [(3 − 3(1)) (−5 − 3(2))], which is equal to [0 − 11].

• 5R1 is [(5(1)) (5(2))], which is equal to [5 10].

(a)
   
1 3 −5 2 6 −10
 −2 0 4  2R1 → R1  −2 0 4 
4 −1 −6 4 −1 −6

(b)
   
1 3 −5 1 3 −5
 −2 0 4  −4R3 → R3  −2 0 4 
4 −1 −6 −16 4 24
82 Chapter 1. Systems of Linear Equations

(c)
   
1 3 −5 −2 0 4
 −2 0 4  R1 ↔ R2  1 3 −5 
4 −1 −6 4 −1 −6

(d)
   
1 3 −5 1 3 −5
 −2 0 4  R2 ↔ R3  4 −1 −6 
4 −1 −6 −2 0 4

(e)
   
1 3 −5 1 3 −5
 −2 0 4  R2 + R1 → R2  −1 3 −1 
4 −1 −6 4 −1 −6

(f )
   
1 3 −5 −1 3 −1
 −2 0 4  R1 + R2 → R1  −2 0 4 
4 −1 −6 4 −1 −6

(g)
   
1 3 −5 1 3 −5
 −2 0 4  R2 + 2R1 → R2  0 6 −6 
4 −1 −6 4 −1 −6

(h)
   
1 3 −5 7 3 −17
 −2 0 4  R1 − 3R2 → R1  −2 0 4 
4 −1 −6 4 −1 −6

(i)
   
1 3 −5 1 3 −5
 −2 0 4  R3 − 21 R1 → R3  −2 0 4 
7
4 −1 −6 2
− 2 − 72
5

7. [Exercise on page 6]
If the operation that was performed is Ri ↔ Rj , then one should be able to see this
immediately. If only one operation (aRi → Ri or Ri + bRj → Ri ) is performed on A
to get B, then every row of B will appear in A except one (the one affected by the
operation). In this case, one needs to first find the row of B that does not appear in
A. Then one needs to figure out the correct operation that was performed.
1.2. Echelon Matrices 83

   
1 2 1 2
(a) A = B=
−2 4 0 8

ˆ First, observe that the first row of B appears in A, but the second one does
not. So the second row of A is the only one that changed.
ˆ Since only one row changed, the operation that was performed is either on
the form aRi → Ri or on the form Ri + bRj → Ri .
– Is it possible to multiply the second row of A by a nonzero number to
get the second row of B? Clearly the answer is no. So the required
operation is not on the form aRi → Ri .
– So it is on the form R2 + bR1 → R2 . By inspection, one can see that the
correct operation that was performed on A to get B is R2 +2R1 → R2 .
   
−1 3 0 −1 3 −1
(b) A =  1 −2 −3  B =  1 −2 −3 
−1 3 −1 −1 3 0
The operation that was performed is R1 ↔ R3 .
   
−2 3 −1 −2 3 −1
(c) A =  0 2 −2  B =  0 2 −2 
2 −3 −1 −2 3 1

ˆ Since only one row has changed (row 3), the required operation is either on
the form aRi → Ri or on the form Ri + bRj → Ri .
ˆ Is it possible to multiply the third row of A by a nonzero number a and
obtain the third row of B? Yes, a = −1.
So the required operation is: −R3 → R3 .
(d)
   
3 5 −4 1 5 2
A =  −1 0 3  R1 + 2R2 → R1  −1 0 3  = B
1 5 2 1 5 2

(e)
   
3 −12 −2 3 −12 −2
A =  2 −3 −7  R2 − 4R3 → R2  −2 13 −3  = B
1 −4 −1 1 −4 −1

8. [Exercise on page 6]

Performing Many Operations at the Same Time on a Matrix

Consider the following matrix
 
1 2 1
A= 1 5 3 
2 8 5
84 Chapter 1. Systems of Linear Equations

If we perform the operation R2 − R1 → R2 on A, we get the matrix

 
1 2 1
A1 = 0 3
 2 
2 8 5

Now, if we perform the operation R3 − 2R1 → R3 on A1 , we get the matrix

 
1 2 1
A2 = 0 3
 2 
0 4 3

The point here is that these two operations can be performed at the same time
on A:
   
1 2 1 1 2 1
R2 − R1 → R2
A= 1 5 3   0 3 2  = A2
R3 − 2R1 → R3
2 7 4 0 4 3

Warning. One has to be careful when performing more than one operation at
the same time on a matrix. A common mistake is to do the following:
   
1 2 1 1 2 1
R2 − R1 → R2
A= 1 5 3   0 3 2 
R3 − R2 → R3
2 7 4 1 2 1

This is incorrect because

ˆ R2 is replaced by something when performing the operation R2 −R1 → R2 ,

and

ˆ the operation R3 − R2 → R3 involved R2 .

The second operation should have involved R20 (not R2 ) where R20 = R2 − R1 .
Here is the correct sequence:
     
1 2 1 1 2 1 1 2 1
 1 5 3  R2 − R1 → R2  0 3 2  R3 − R2 → R3  0 3 2 
2 7 4 1 2 1 1 −1 −1

Examples of operations that can be performed at the same time on a matrix

include:
R2 − R1 → R2
R2 + R3 → R2
R3 − 3R1 → R3 etc.
R1 − 5R3 → R1
R4 − 2R1 → R4
1.2. Echelon Matrices 85

The following
   
1 1 2 1 1 2
 1 4 2  R2 − R3 → R2  −2 −1 −4 
R3 − R2 → R3
3 5 6 2 1 4

is incorrect for the same reason as the one mentioned above: R2 is replaced by
something (by performing the first operation) and the same R2 is involved in the
second operation. The correct sequence should be:

     
1 1 2 1 1 2 1 1 2
 1 4 2  R2 − R3 → R2  −2 −1 −4  R3 − R2 → R3  −2 −1 −4 
3 5 6 3 5 6 5 6 10

9. [Exercise on page 6]

Pivots
Let A be an echelon matrix.

ˆ A pivot of A is a leading entry of A.

ˆ A pivot position of A is a position in A containing a leading entry.

ˆ A pivot column of A is a column containing a leading entry.

ˆ If an entry is located in row i and column j, we say that it is in position

(or its position is) (i, j). For example the position of an entry in row 2
and column 3 is (2, 3).

For example, consider the matrix

 
2 1 −4 0
A =  0 0 −5 −2 
0 0 0 7

The pivots of A are 2, −5, and 7. The pivot positions are (1, 1), (2, 3), and
(3, 4). And the pivot columns are column 1, column 3, and column 4.

(a) The pivots of  

1 4 7
 0 9 2 
0 0 5
are 1, 9, and 5. The pivot positions are (1, 1), (2, 2), and (3, 3). And the pivot
columns are column 1, column 2, and column 3.
86 Chapter 1. Systems of Linear Equations

(b) The pivots of  

−3 4 0
 0 8 2 
0 0 0
are −3 and 8. The pivot positions are (1, 1) and (2, 2). And the pivot columns
are columns 1 and 2.
(c) The pivots of  
0 4 0 1
0 0 1 13
are 4 and 1. These correspond to the positions (1, 2) and (2, 3). The pivot
columns are columns 2 and 3.
(d) The pivots of  
1 0 2 0
 0 0 0 1 
 
 0 0 0 0 
0 0 0 0
are 1 and 1. The pivot positions are (1, 1) and (2, 4). And the pivot columns
are columns 1 and 4.

10. [Exercise on page 7]

Let us first recall what me mean by echelon form of a matrix.

Echelon Form of a Matrix and Gaussian Elimination

Let A be a matrix. An echelon form of A is an echelon matrix that is
obtained from A by performing a finite sequence of elementary row operations.
For example, consider the matrix
 
1 −2 3
A =  1 −1 5 
−2 1 −1
To find an echelon form of A, we need to perform elementary row operations
to transform A into an echelon matrix.
ˆ First, we focus on the first column of A.
– The upper left position is 1. So we don’t have to do anything for
this entry, and it can serve as a pivot. (A pivot of 1 or −1 makes
calculations simple.)
– Next, we need to create zeros in the other positions of the first
column. To create a 0 in the position right below the pivot, we can
perform the operation R2 − R1 → R2 . To create a 0 in the lower left
position, we can perform the operation R3 + 2R1 → R3 . We obtain
   
1 −2 3 1 −2 3
 1 −1 R2 − R1 → R2
5   0 1 2 
R3 + 2R1 → R3
−2 1 −1 0 −3 5
1.2. Echelon Matrices 87

This completes the work on the first column.

ˆ Now, we focus on the second column. The 1 in the second position of the
second row makes a good pivot. Using that pivot, we can create a 0 in
the position right below it by performing the operation R3 + 3R2 → R3 .
We obtain
   
1 −2 3 1 −2 3
 0 1 2  R3 + 3R2 → R3  0 1 2 
0 −3 5 0 0 11
 
1 −2 3
The matrix B =  0 1 2  is an echelon matrix. Since it has been
0 0 11
obtained from A by a sequence of elementary row operations, it is an echelon
form of A.

The process above that enabled us to transform A into an echelon matrix is

referred to as Gaussian Elimination.

We come back to the exercise. It is possible to find two echelon forms of a

matrix A that are not the same. For example, consider the matrix A above.
 
1 −2 3
A =  1 −1 5 
−2 1 −1
We saw that an echelon form of A is the matrix
 
1 −2 3
B= 0 1 2 
0 0 11
If we transform B again into an echelon matrix using elementary row operations, we
get an echelon form of B, which is also an echelon form of A. So, for example, if we
1
perform the operation 11 R3 → R3 on B, we get the matrix
 
1 −2 3
C= 0 1 2 
0 0 1
This latter matrix is also an echelon form of A.

Based on this example, we can make the following statement.

A Statement About Echelon Forms

A matrix can have more than one echelon form.

11. [Exercise on page 7]

88 Chapter 1. Systems of Linear Equations

Reduced Echelon Form and Gauss-Jordan Elimination

Let A be a matrix. A reduced echelon form of A is a reduced echelon
matrix that is obtained from A by performing a finite sequence of elementary
row operations. For example, consider the same matrix A as above:
 
1 −2 3
 1 −1 5 
−2 1 −1
To find a reduced echelon form of A, we need to perform elementary row op-
erations to put A into a reduced echelon matrix. There are two phases: the
forward phase and the backward phase.
ˆ Forward Phase. This phase is the one we did above:
   
1 −2 3 1 −2 3
 1 −1 R2 − R1 → R2
5   0 1 2 
R3 + 2R1 → R3
−2 1 −1 0 −3 5
 
1 −2 3
R3 + 3R2 → R3  0 1 2 
0 0 11
 
1 −2 3
1
R → R3
11 3
 0 1 2 
0 0 1
ˆ Backward Phase. The idea is to work backward and create zeros in
each pivot column.
– First, we focus on the rightmost pivot column, which is column 3.
The lower right position is 1. So it can serve as a pivot.
– Next, we need to create zeros in the other positions of the third
column. To create a 0 in the position right above the pivot, we can
perform the operation R2 − 2R3 → R2 . And to create a 0 in the
upper right position, we can perform the operation R1 − 3R3 → R1 .
We obtain the following:
   
1 −2 3 1 −2 0
 0 R2 − 2R3 → R2
1 2   0 1 0 
R1 − 3R3 → R1
0 0 1 0 0 1
This completes the work on the third column.
– Now, we focus on the second column. The 1 in the second position of
the second row makes a good pivot. Using that pivot, we can create
a 0 in the position above it by performing the operation R1 + 2R2 →
R1 :
   
1 −2 0 1 0 0
 0 1 0  R1 + 2R2 → R1  0 1 0 
0 0 1 0 0 1
This completes the backward phase.
1.2. Echelon Matrices 89

 
1 0 0
So a reduced echelon form of A is the matrix  0 1 0 .
0 0 1
The process that enabled us to transform A into a reduced echelon matrix is
referred to as Gauss-Jordan elimination. In other words, the Gauss-Jordan
elimination is the combination of the forward and backward phases. While the
Gaussian elimination is just the forward phase.

Although a matrix can have many echelon forms as we saw above, the reduced
echelon form is unique as stated in the following theorem.

Theorem 1.2.1. Every matrix has a unique reduced echelon form.

In other words, every matrix is row equivalent to a unique reduced echelon

matrix.

We come back to the exercise. Using Theorem 1.2.1, the answer to the question is:
False.

12. [Exercise on page 7]

 
1 2
False. Indeed, consider the matrix A = . Performing the operation R2 −
3 10
3R1 → R2 , we get the matrix
 
1 2
B=
0 4

Now, by performing the operation 41 R2 → R2 , we get the matrix

 
1 2
C=
0 1

We thus obtain two echelon forms of A, namely B and C. Clearly, they don’t have
the same pivots as the pivots of B are 1 and 4 while the pivots of C are 1 and 1.

13. [Exercise on page 7]

True. This follows from the following important fact.

Pivot Positions are Invariant

Theorem 1.2.2. Let A be a matrix. Let B and C be two echelon forms of A.
Then B and C have the same pivot positions.

14. [Exercise on page 7]

90 Chapter 1. Systems of Linear Equations

Row Equivalence
Two matrices are said to be row equivalent if one can be obtained from the
other by performing a finite sequence of elementary row operations.
Notation. If two matrices A and B are row equivalent, we write A ∼ B.

(a) We want to show that the matrices

   
1 0 2 1 0 2
A= 1 3 3  and B= 0 3 1 
−1 3 3 0 0 4
are row equivalent. To do this, we will try to transform A into B by using
elementary row operations.

   
1 0 2 1 0 2
R2 − R1 → R2
A= 1 3 3   0 3 1 
R3 + R1 → R3
−1 3 3 0 3 5
 
1 0 2
R3 − R2 → R3  0 3 1  = B
0 0 4
This shows that A and B are row equivalent (or A is row equivalent to B).
(b)
   
1 −2 4 1 −2 4
R2 − 3R1 → R2
A =  3 −7 15   0 −1 3 
R3 − R1 → R3
1 0 3 0 2 −1
 
1 −2 4
R3 + 2R2 → R3  0 −1 3 
0 0 5
 
1 −2 4
1
R → R3
5 3
 0 −1 3  = B
0 0 1
This shows that A and B are row equivalent.
(c) Since the upper left position of A is 0, it can’t serve as a pivot. So we first need
to interchange row 1 with a row below it in order to place a nonzero number in
the upper left position. There are two possibilities: either we use the operation
R1 ↔ R2 or we use R1 ↔ R3 . The first operation will place a −3 in the pivot
while the second will place a 1. Since it is easier to work with 1 as a pivot, we
will use the operation R1 ↔ R3 first.

   
0 2 1 1 1 3
A =  −3 1 −2  R1 ↔ R3  −3 1 −2 
1 1 3 0 2 1
1.2. Echelon Matrices 91

 
1 1 3
R2 + 3R1 → R2  0 4 7 
0 2 1
 
1 1 3
2R3 → R3  0 4 7 
0 4 2
 
1 1 3
R3 − R2 → R3  0 4 7 =B
0 0 −5

This shows that A and B are row equivalent.

(d) We need to create zeros in each position above each pivot of
 
1 −6 −4 1
A= 0 2 3 5 
0 0 1 3

ˆ First, the rightmost pivot is 1 which is located in the third row and third
column. We need to create zeros in each position above that pivot. Per-
forming the operations R2 − 3R3 → R2 and R1 + 4R3 → R1 , we get the
matrix  
1 −6 0 13
 0 2 0 −4 
0 0 1 3
ˆ Next, we create a zero above the pivot located in the second row and second
column. Performing the operation R1 + 3R2 → R1 , we get the matrix
 
1 0 0 1
 0 2 0 −4 
0 0 1 3

ˆ To finish, we perform the operation 12 R2 → R2 . We obtain

 
1 0 0 1
 0 1 0 −2  = B
0 0 1 3

This shows that A and B are row equivalent.

15. [Exercise on page 7]

(a) Performing the operations R2 − R1 → R2 and R3 − 2R1 → R3 on the matrix

 
1 2 3 4
 1 3 1 1 
2 7 1 4
92 Chapter 1. Systems of Linear Equations

we get  
1 2 3 4
 0 1 −2 −3 
0 3 −5 −4

Performing R3 − 3R2 → R3 , we get

 
1 2 3 4
 0 1 −2 −3 
0 0 1 5

This completes the forward phase.

Now, Performing the operations R1 − 3R3 → R1 and R2 + 2R3 → R2 , we get
 
1 2 0 −11
 0 1 0 7 
0 0 1 5

Finally, we perform the operation R1 − 2R2 → R1 . We obtain the matrix

 
1 0 0 −25
 0 1 0 7 
0 0 1 5

(b)
   
3 −5 13 −2 1 −2 3 1
 −2 2 −12 14  R1 ↔ R3  −2 2 −12 14 
1 −2 3 1 3 −5 13 −2
 
1 −2 3 1
R2 + 2R1 → R2  0 −2 −6 16 
R3 − 3R1 → R3
0 1 4 −5
 
1 −2 3 1
− 21 R2 → R2  0 1 3 −8 
0 1 4 −5
 
1 −2 3 1
Forward Phase Ends R3 − R2 → R3  0 1 3 −8 
0 0 1 3

 
1 −2 0 −8
R2 − 3R3 → R2
Backward Phase Starts  0 1 0 −17 
R1 − 3R3 → R1
0 0 1 3
 
1 0 0 −42
R1 + 2R2 → R1  0 1 0 −17 
0 0 1 3
1.2. Echelon Matrices 93

(c)
   
3 9 0 −3 1 3 0 −1
1
 −1 −2
 2 −2 
 R
3 1
→ R1  −1 −2
 2 −2 

 4 11 −1 3  1
R → R4  4 11 −1 3 
3 4
3 9 0 −3 1 3 0 −1
 
1 3 0 −1
R2 + R1 → R2  0 1 2 −3 
R3 − 4R1 → R3 
 0 −1 −1

7 
R4 − R1 → R4
0 0 0 0
 
1 3 0 −1
 0 1 2 −3 
Forward Phase Ends R3 + R2 → R3 
 0 0 1

4 
0 0 0 0
 
1 3 0 −1
 0 1 0 −11 
Backward Phase Starts R2 − 2R3 → R2 
 0

0 1 4 
0 0 0 0
 
1 0 0 32
 0 1 0 −11 
R1 − 3R2 → R1 
 0

0 1 4 
0 0 0 0
(d)
   
−3 1 1 −1 1 −1 −1 −1
 4 −2 −2 0  R1 + R2 → R1  4 −2 −2 0 
5 0 −3 −1 5 0 −3 −1
 
1 −1 −1 −1
R2 − 4R1 → R2  0 2 2 4 
R3 − 5R1 → R3
0 5 2 4
 
1 −1 −1 −1
1
R
2 2
→ R2  0 1 1 2 
0 5 2 4
 
1 −1 −1 −1
R3 − 5R2 → R3  0 1 1 2 
0 0 −3 −6
 
1 −1 −1 −1
Forward Phase Ends − 13 R3  0 1 1 2 
0 0 1 2
 
1 −1 0 1
R2 − R3 → R2  0
Backward Phase Starts 1 0 0 
R1 + R3 → R1
0 0 1 2
94 Chapter 1. Systems of Linear Equations

 
1 0 0 1
R1 + R2 → R1  0 1 0 0 
0 0 1 2

(e)
   
1 2 3 −2 7 1 2 3 −2 7
R2 − R1 → R2
 1 2 3 −2 6   0 0 0 0 −1 
  R3 − R1 → R3  
 1 2 3 −1 4   0 0 0 1 −3 
R4 + R1 → R4
−1 −2 −4 3 −2 0 0 −1 1 5
 
1 2 3 −2 7
 0 0 −1 1 5 
R2 ↔ R4  
 0 0 0 1 −3 
0 0 0 0 −1

 
1 2 3 −2 7
 0 0 −1 1 5 
Forward Phase Ends −R4 → R4  
 0 0 0 1 −3 
0 0 0 0 1

 
1 2 3 −2 0
R3 + 3R4 → R3  0 0 −1 1 0 
Backward Phase Starts R2 − 5R4 → R2 
 0

0 0 1 0 
R1 − 7R4 → R1
0 0 0 0 1
 
1 2 3 0 0
R2 − R3 → R2  0 0 −1 0 0 
 
R1 + 2R3 → R1  0 0 0 1 0 
0 0 0 0 1
 
1 2 0 0 0
 0 0 −1 0 0 
R1 + 3R2 → R1 
 0 0

0 1 0 
0 0 0 0 1
 
1 2 0 0 0
 0 0 1 0 0 
−R2 → R2 
 0 0 0 1 0 


0 0 0 0 1

(f )
   
0 4 −3 −1 0 4 −3 −1
 −3 1 1 2  − 21 R3 → R3  −3 1 1 2 
−2 −4 2 −4 1 2 −1 2
 
1 2 −1 2
R1 ↔ R3  −3 1 1 2 
0 4 −3 −1
1.3. Solving Linear Systems Using Matrices 95

 
1 2 −1 2
R2 + 3R1 → R2  0 7 −2 8 
0 4 −3 −1
 
1 2 −1 2
R2 − 2R3 → R2  0 −1 4 10 
0 8 −6 −2
 
1 2 −1 2
R3 + 8R2 → R3  0 −1 4 10 
0 0 26 78
 
1 2 −1 2
1
Forward Phase Ends R
26 3
→ R3  0 −1 4 10 
0 0 1 3
 
1 2 0 5
R2 − 4R3 → R2  0 −1 0 −2
Backward Phase Starts 
R1 + R3 → R1
0 0 1 3
 
1 0 0 1
R1 + 2R2 → R1  0 −1 0 −2 
0 0 1 3
 
1 0 0 1
−R2 → R2  0 1 0 2 
0 0 1 3

16. [Exercise on page 8]

We have the following possibilities:
       
1 0 1 ∗ 0 1 0 0
0 1 , 0 0 , 0 0 , and 0 0
0 0 0 0 0 0 0 0

1.3 Solving Linear Systems Using Matrices

1. [Exercise on page 8]

Coefficient Matrix and Augmented Matrix

Let (S) be a linear system.

ˆ The coefficient matrix (or matrix of coefficients) of (S) is the matrix

containing all the coefficients of (S).

ˆ The augmented matrix of (S) is the matrix containing all the coeffi-
cients and all the constant terms of (S).
96 Chapter 1. Systems of Linear Equations

3x1 + 4x2 = 5
(a) For the system
6x1 − 7x2 = 8
ˆ the coefficient matrix is  
3 4
6 −7
ˆ the augmented matrix is  
3 4 5
6 −7 8
−x1 − 3x2 + x3 = −2
(b) For the system x1 − x2 − 4x3 = 0
−5x1 + x3 = 6
ˆ The coefficient matrix is
 
−1 −3 1
 1 −1 −4 
−5 0 1

ˆ The augmented matrix is

 
−1 −3 1 −2
 1 −1 −4 0 
−5 0 1 6
x2 − x3 + x4 = 1
(c) For the system x1 + 4x2 − 5x3 + 13x4 = −9
−2x1 − x4 = 6
ˆ The coefficient matrix is
 
0 1 −1 1
 1 4 −5 13 
−2 0 0 −1

ˆ The augmented matrix is

 
0 1 −1 1 1
 1 4 −5 13 −9 
−2 0 0 −1 6

2. [Exercise on page 8]
 
1 −1 3 9
(a) The linear system corresponding to the augmented matrix  −2 4 −8 0 
3 −5 6 12
is
x1 − x2 + 3x3 = 9
−2x1 + 4x2 − 8x3 = 0
3x1 − 5x2 + 6x3 = 12
1.3. Solving Linear Systems Using Matrices 97

 
1 −2 4 10
(b) The linear system corresponding to the augmented matrix  0 −3 −9 −41 
0 0 2 23
is
x1 − 2x2 + 4x3 = 10
− 3x2 − 9x3 = −41
2x3 = 23
 
1 5 0 −2 −3
(c) The linear system corresponding to the augmented matrix  0 0 1 4 −1 
0 0 0 1 7
is
x1 + 5x2 − 2x4 = −3
x3 + 4x4 = −1
x4 = 7

3. [Exercise on page 8]
True. This follows from the following definition.

Equivalence of Systems – Definition

Two linear systems are said to be equivalent if they have the same set of
solutions.

4. [Exercise on page 8]
False. This is an immediate consequence of the following important theorem.

Equivalence of Systems – Theorem

Theorem 1.3.1. Two linear systems are equivalent if and only if the corre-
sponding augmented matrices are row equivalent.

Thanks to this theorem, one can solve a linear system as follows:

Step 1. Write the augmented matrix.

Step 2. Transform the augmented matrix to reduced echelon form.

Step 3. Solve the linear system corresponding to the reduced echelon form.
(This system is pretty easy to solve.)

5. [Exercise on page 8]

(a) The solution is (5, 7).

(b) We need to reduce the augmented matrix.
   
1 1 7 1 1 1 7
R → R2
3 2
0 3 12 0 1 4
98 Chapter 1. Systems of Linear Equations

 
1 0 3
R1 − R2 → R1
0 1 4
The corresponding system is
x1 = 3
x2 = 4
So the solution is (3, 4).

Alternate Solution
x1 +x2 = 7
The system can be solved by using the so called
3x2 = 12
backward substitution:

ˆ First, by dividing both sides of the third equation by 3, we get

x2 = 4.

ˆ Substituting x2 with 4 into the first equation, we get x1 + 4 = 7.

Solving this latter equation, we get x1 = 3.

So the solution is (3, 4)

(c) We reduce the augmented matrix.

   
1 −1 1 3 1 −1 1 3
1
 0 5 −3 0  R
11 3
→ R3  0 5 −3 0 
0 0 11 55 0 0 1 5
 
1 −1 0 −2
R2 + 3R3 → R2  0 5 0 15 
R1 − R3 → R1
0 0 1 5
 
1 −1 0 −2
1
R
5 2
→ R2  0 1 0 3 
0 0 1 5
 
1 0 0 1
R1 + R2 → R1  0 1 0 3 
0 0 1 5
So the solution is (1, 3, 5).

6. [Exercise on page 8]
(a) We reduce the augmented matrix.

   
1 2 5 1 2 5
R2 − 2R1 → R2
2 −1 −10 0 −5 −20
 
1 2 5
− 15 R2 → R2
0 1 4
1.3. Solving Linear Systems Using Matrices 99

 
1 0 −3
R1 − 2R2 → R1
0 1 4

The system corresponding to the reduced echelon form is

x1 = −3
x2 = 4

So the solution is (−3, 4).

(b) We reduce the augmented matrix.

   
1 1 0 1 1 0
R2 − 3R1 → R2
3 −4 7 0 −7 7
 
1 1 0
− 17 R2 → R2
0 1 −1
 
1 0 1
R1 − R2 → R1
0 1 −1

The system corresponding to the reduced echelon form is

x1 = 1
x2 = −1

So the solution is (1, −1).

(c) We reduce the augmented matrix.

   
2 −3 7 2 −3 7
R2 − 2R1 → R2
5 7 3 1 13 −11
 
1 13 −11
R1 ↔ R2
2 −3 7
 
1 13 −11
R2 − 2R1 → R2
0 −29 29
 
1 1 13 −11
− 29 R2 → R2
0 1 −1
 
1 0 2
R1 − 13R2 → R1
0 1 −1

The system corresponding to the reduced echelon form is

x1 = 2
x2 = −1

So the solution is (2, −1).

100 Chapter 1. Systems of Linear Equations

7. [Exercise on page 9]

(a) As usual, we reduce the augmented matrix.

   
1 −2 −1 −6 1 −2 −1 −6
 −1 R2 + R1 → R2
3 2 11   0 1 1 5 
R3 + 2R1 → R3
−2 5 4 20 0 1 2 8
 
1 −2 −1 −6
R3 − R2 → R3  0 1 1 5 
0 0 1 3
 
1 −2 0 −3
R2 − R3 → R2  0 1 0 2 
R1 + R3 → R1
0 0 1 3
 
1 0 0 1
R1 + 2R2 → R1  0 1 0 2 
0 0 1 3
The system corresponding to the reduced echelon form is
x1 = 1
x2 = 2
x3 = 3

So the solution is (1, 2, 3).

(b) We reduce the augmented matrix.
   
1 −1 1 2 1 −1 1 2
R2 − 2R1 → R2
 2 1 0 −4   0 3 −2 −8 
R3 + 3R1 → R3
−3 0 2 14 0 −3 5 20
 
1 −1 1 2
R3 + R2 → R3  0 3 −2 −8 
0 0 3 12
 
1 −1 1 2
1
R
3 3
→ R3  0 3 −2 −8 
0 0 1 4

 
1 −1 0 −2
R2 + 2R3 → R2  0 3 0 0 
R1 − R3 → R1
0 0 1 4
 
1 −1 0 −2
1
R
3 2
→ R2  0 1 0 0 
0 0 1 4
 
1 0 0 −2
R1 + R2 → R1  0 1 0 0 
0 0 1 4
1.3. Solving Linear Systems Using Matrices 101

The system corresponding to the reduced echelon form is

x1 = −2
x2 = 0
x3 = 4

So the solution is (−2, 0, 4).

(c) We reduce the augmented matrix.
   
1 1 0 1 4 1 1 0 1 4
 1
 0 1 0 1 
 R2 − R1 → R2  0 −1
 1 −1 −3 

 0 2 −1 1 4  R4 − 2R1 → R4  0 2 −1 1 4 
2 −1 1 −3 −5 0 −3 1 −5 −13
 
1 1 0 1 4
 0 1 −1 1 3 
−R2 → R2  
 0 2 −1 1 4 
0 −3 1 −5 −13
 
1 1 0 1 4
R3 − 2R2 → R3  0 1 −1
 1 3 

R4 + 3R2 → R4  0 0 1 −1 −2 
0 0 −2 −2 −4
 
1 1 0 1 4
 0 1 −1 1 3 
R4 + 2R3 → R4  
 0 0 1 −1 −2 
0 0 0 −4 −8
 
1 1 0 1 4
 0 1 −1 1 3 
− 41 R4 → R4  
 0 0 1 −1 −2 
0 0 0 1 2

 
1 1 0 0 2
R3 + R4 → R3  0 1 −1 0 1 
R2 − R4 → R2 
 0

0 1 0 0 
R1 − R4 → R1
0 0 0 1 2
 
1 1 0 0 2
 0 1 0 0 1 
R2 + R3 → R2  
 0 0 1 0 0 
0 0 0 1 2
 
1 0 0 0 1
 0 1 0 0 1 
R1 − R2 → R1  
 0 0 1 0 0 
0 0 0 1 2
102 Chapter 1. Systems of Linear Equations

The system corresponding to the reduced echelon form is

x1 = 1
x2 = 1
x3 = 0
x4 = 2

So the solution is (1, 1, 0, 2).

8. [Exercise on page 9]

(a) We reduce the augmented matrix.

   
1 1 2 3 1 1 2 3
 3 4 8 5  R2 − 3R1 → R2  0 1 2 −4 
0 −1 −2 2 0 −1 −2 2
 
1 1 2 3
R3 + R2 → R3  0 1 2 −4 
0 0 0 −2
At this stage we observe that the third equation is 0 = −2, which is not true.
So the system has no solution.

How to Know That a Linear System has No Solution?

While performing elementary row operations on the augmented matrix of
a linear system, if at some point a row of the form
 
0 0 ··· 0 b ,

with b 6= 0, occurs, then the system has no solution.

(b) We reduce the augmented matrix.

   
1 −2 1 0 1 −2 1 0
 1 −1 R2 − R1 → R2
4 1   0 1 3 1 
R3 + R1 → R3
−1 2 −1 5 0 0 0 5
 
6 0,
Since the third row is of the form 0 0 0 b with b = 5 = it follows that
the system has no solution.

9. [Exercise on page 9]

Leading Variables and Free Variables

Let (S) be a linear system whose corresponding augmented matrix, N , is in

echelon form. (We say that (S) is an echelon system.)
ˆ A leading variable of (S) is a variable corresponding to a leading entry
of N .
1.3. Solving Linear Systems Using Matrices 103

ˆ Any variable which is not a leading variable is referred to as a free vari-

able.

3x1 + x2 − x3 = 4
(a) The augmented matrix of the system is
5x2 − 2x3 = −8
 
3 1 −1 4
N=
0 5 −2 −8

ˆ Observe that N is in echelon form. The leading entries are 3 and 5. And
the variables corresponding to these entries are x1 and x2 respectively. So
the leading variables are x1 and x2 .
ˆ The other variable, x3 , is a free variable.
(b) For the system
x1 − x2 − x3 = 1
−x2 + x3 = 7
2x3 = 5
the leading variables are x1 , x2 , and x3 . There is no free variable.
(c) For the system
x1 + 5x2 − 7x3 + x4 = −9
7x3 − x4 = 9
the leading variables are x1 and x3 . The other variables, x2 and x4 , are
free.

10. [Exercise on page 9]

(a) False. For example, consider the system

x1 = 3
x2 = 5
0 = 7

Clearly, in this system, all the variables (x1 and x2 ) are leading variables. But
the system has no solution because the third equation is not satisfied.

For a system to have a unique solution, all the variables has to be leading
variables and one more condition has to be satisfied as one can see in the following
theorem.

Existence and Uniqueness Theorem

Theorem 1.3.2. Let (S) be a linear system, and let M be the augmented
matrix of (S). Let N be an echelon form of M .

(A) The system (S) is inconsistent if and only if the matrix N has a row
104 Chapter 1. Systems of Linear Equations

of the form  
0 0 ··· 0 b
with b 6= 0. (This amounts to saying that (S) is inconsistent if and
only if the rightmost column of N is a pivot column.)

(B) The system (S) has a unique solution if and only if the following two
conditions are satisfied:
 
(1) The matrix N has no row of the form 0 0 · · · 0 b with
b 6= 0. (This amounts to saying that the system (S) is consistent
thanks to part (I).)
(2) All the variables of (S) are leading variables.

(C) The system (S) has infinitely many solutions if and only if the fol-
lowing two conditions are satisfied:
 
(i) The matrix N has no row of the form 0 0 · · · 0 b with
b 6= 0. (This amounts to saying that the system (S) is consis-
tent.)
(ii) At least one variable of (S) is a free variable.

(b) False. For example, consider the system

x1 + x2 = 6
0 = 1

This system has a free variable (x2 ), but it is not consistent as the second
equation is not satisfied.

11. [Exercise on page 9]

How to Solve a System of Linear Equations?

To solve a system of linear equations, one can proceed as follows.

Step 1. Transform the augmented matrix into a reduced echelon form using
elementary row operations.
 
Step 2. If there is a row of the form 0 0 · · · 0 b , with b 6= 0, the
system is inconsistent.

Step 3. If the system is consistent, solve the system corresponding to the re-
duced echelon form for leading variables (if there are free variables,
express each leading variable in terms of the free variables).
1.3. Solving Linear Systems Using Matrices 105

(a) We want to solve the following system.

x1 − 2x2 + 4x3 = 5
−x1 + 3x2 − 2x3 = −2
3x1 − 4x2 + 16x3 = 21

First, we need to reduce the augmented matrix.

   
1 −2 4 5 1 −2 4 5
 −1 R2 + R1 → R2
3 −2 −2   0 1 2 3 
R3 − 3R1 → R3
3 −4 16 21 0 2 4 6
 
1 −2 4 5
R3 − 2R2 → R3  0 1 2 3 
0 0 0 0
 
1 0 8 11
R1 + 2R2 → R1  0 1 2 3 
0 0 0 0

The linear system corresponding to the reduced echelon form is

x1 + 8x3 = 11
x2 + 2x3 = 3
In this latter system x3 is a free variable. So we set x3 equal to a parameter:

x3 = s

Now we solve each equation for the leading variables (x1 and x2 ) in terms of s.
ˆ Replacing x3 with s into the first equation, we get x1 + 8s = 11. Solving
this for x1 , we get x1 = 11 − 8s.
ˆ Replacing x3 with s into the second equation, we get x2 + 2s = 3. This
implies that x2 = 3 − 2s.
Thus, the general solution to the system is
x1 = 11 − 8s
x2 = 3 − 2s
x3 = s
where s can be any real number.
Warning. Do not forget to write the sentence “where s can be any real number”.
(b) First, we reduce the augmented matrix.
   
1 3 2 7 1 3 2 7
R2 − 2R1 → R2
 2 7 3 10   0 1 −1 −4 
R3 + 4R1 → R3
−4 −13 −7 −24 0 −1 1 4
 
1 3 2 7
R3 + R2 → R3  0 1 −1 −4 
0 0 0 0
106 Chapter 1. Systems of Linear Equations

 
1 0 5 19
R1 − 3R2 → R1  0 1 −1 −4 
0 0 0 0
The linear system corresponding to the reduced echelon form is
x1 + 5x3 = 19
x2 − x3 = −4
In this latter system x3 is a free variable. So we set x3 equal to a parameter:
x3 = s. Solving the the first and second equations for the leading variables (x1
and x2 ), we get x1 = 19 − 5s and x2 = −4 + s. So the general solution is
x1 = 19 − 5s
x2 = −4 + s
x3 = s
where s can be any real number.
(c) We reduce the augmented matrix.
   
1 1 2 1 1 1 2 1
 1 R2 − R1 → R2
2 2 4   0 1 0 3 
R3 + R1 → R3
−1 −3 −2 −2 0 −2 0 −1
 
1 1 2 1
R3 + 2R2 → R3  0 1 0 3 
0 0 0 5
We do not need to go further because the third equation is 0 = 5, which is not
true. So the system has no solution.
(d) Transforming the augmented matrix:
   
1 −2 0 −1 2 1 −2 0 −1 2
 2 −4 R2 − 2R1 → R2
1 −5 6   0 0 1 −3 2 
R3 − R1 → R3
1 −2 −1 2 0 0 0 −1 3 −2
 
1 −2 0 −1 2
R3 + R2 → R3  0 0 1 −3 2 
0 0 0 0 0
The linear system corresponding to the reduced echelon form is
x1 − 2x2 − x4 = 2
x3 − 3x4 = 2
In this latter system x2 and x4 are free variables. So we set them equal to
parameters: x2 = t and x4 = s. Solving the the first and second equations for
the leading variables (x1 and x3 ), we get x1 = 2 + 2t + s and x3 = 2 + 3s. So
the general solution is
x1 = 2 + 2t + s
x2 = t
x3 = 2 + 3s
x4 = s
where s and t can be any real numbers.
1.3. Solving Linear Systems Using Matrices 107

(e) As usual we reduce the augmented matrix.

   
3 8 −3 −14 2 1 −2 1 10 0
 2 3 −1 −2 1   2 3 −1 −2 1 

 1 −2
 R1 ↔ R3  
1 10 0   3 8 −3 −14 2 
1 5 −2 −12 1 1 5 −2 −12 1
 
1 −2 1 10 0
R2 − 2R1 → R2  0 7 −3 −22 1 
R3 − 3R1 → R3 
 0 14 −6 −44 2


R4 − R1 → R4
0 7 −3 −22 1
 
1 −2 1 10 0
R3 − 2R2 → R3  0
 7 −3 −22 1 

R4 − R2 → R4  0 0 0 0 0 
0 0 0 0 0
 
1 −2 1 10 0
1
 0 3
1 − 7 − 22 1 
R
7 2
→ R2 
 0
7 7 
0 0 0 0 
0 0 0 0 0
1 26 2
 
1 0 7 7 7
 0 1 − 73 − 22 1
 
R1 + 2R2 → R1 7 7

 
 0 0 0 0 0 
0 0 0 0 0

(Note. Here we can’t avoid fractions.) The linear system corresponding to the
reduced echelon form is
1 26 2
x1 + x
7 3
+ 7 4
x = 7
3 22 1
x2 − x
7 3
− 7 4
x = 7

The leading variables are x1 and x2 , and the free ones are x3 and x4 . Set x3 = s and
x4 = t . Then the general solution is

2
x1 = 7
− 17 s − 26
7
t
1
x2 = + 37 s +
7
22
7
t
x3 = s
x4 = t
where s and t can be any real numbers.
12. [Exercise on page 10]

ˆ Substituting s with x3 and t with x4 into the first equation, we get x1 = 2 +

x3 − x4 . This is equivalent to x1 − x3 + x4 = 2.
ˆ Substituting t with x4 into the second equation, we get x2 = −3 + 7x4 . This is
equivalent to x2 − 7x4 = −3.
108 Chapter 1. Systems of Linear Equations

So the following linear system works:

x1 − x3 + x4 = 2
x2 − 7x4 = −3

13. Consider the following system. [Exercise on page 10]

(a) Remember that a system is consistent if it has at least one solution. To answer
this question, we first need to transform the augmented matrix.
   
1 1 −1 a 1 1 −1 a
R2 − R1 → R2  0 −1
 1 0 2 b  3 b−a 
R3 + R1 → R1
−1 −2 4 c 0 −1 3 c+a
 
1 1 −1 a
R3 − R2 → R3  0 −1 3 b−a 
0 0 0 2a − b + c
 
1 0 2 b
R1 + R2 → R1  0 −1 3 b−a 
0 0 0 2a − b + c
 
1 0 2 b
−R2 → R2  0 1 −3 a−b 
0 0 0 2a − b + c

The third equation of the corresponding system is 0 = 2a − b + c. So the system

is consistent if and only if 2a − b + c = 0.
(b) The system corresponding to the reduced echelon form above is
x1 + 2x3 = b
x2 − 3x3 = a−b
0 = 2a − b + c
If 2a − b + c = 0, taking x3 = s, we get the general solution:
x1 = b − 2s
x2 = a − b + 3s
x3 = s
where s can be any real number.
14. [Exercise on page 10]
As usual, we first need to transform the augmented matrix.
   
−1 3 2 −8 −1 3 2 −8
R2 + R1 → R2
 1 0 1 2   0 3 3 −6 
R3 + 3R1 → R3
3 3 a b 0 12 a + 6 b − 24
 
−1 3 2 −8
1
R → R2
3 2
 0 1 1 −2 
0 12 a + 6 b − 24
1.3. Solving Linear Systems Using Matrices 109

 
−1 3 2 −8
R3 − 12R2 → R3  0 1 1 −2 
0 0 a−6 b
 
1 −3 −2 8
−R1 → R1  0 1 1 −2 
0 0 a−6 b

At this stage, the matrix is just in an echelon form, which is enough to solve the
problem.

ˆ If a = 6 and b 6= 0, then the third row becomes 0 0 0 b with b 6= 0.

 

Therefore the system has no solution (thanks to Theorem 1.3.2–Part (A)).

ˆ If a = 6 and b = 0, then the system becomes

x1 − 3x2 − 2x3 = 8
x2 + x3 = −2
0 = 0

Since this latter (echelon) system has a free variable (x3 ) and since there is no
equation of the form 0 = d with d 6= 0, it follows that the system has infinitely
many solutions (thanks to Theorem 1.3.2–Part (C)).
ˆ If a 6= 6, then the system becomes

x1 − 3x2 − 2x3 = 8
x2 + x3 = −2
(a − 6)x3 = b

Since all the variables of this latter (echelon) system are leading variables and
since there is no equation of the form 0 = d with d 6= 0, it follows that the
system has a unique solution (thanks to Theorem 1.3.2–Part (B)).
110 Chapter 1. Systems of Linear Equations
 Chapter2

Vectors in Rn and Linear Systems

2.1 Vector Equations and Span

1. [Exercise on page 11]

Vectors – Definition
Let n ≥ 1 be an integer.

ˆ A vector in Rn is a matrix of size n × 1, that is, a matrix with n rows

and 1 column.
 
u1
 u2 
ˆ If u =  ..  is a vector in Rn , for 1 ≤ i ≤ n, the entry ui is referred to
 
 . 
un
as the ith entry (or the ith component ) of u.

For example,
 
4
ˆ u= is a vector in R2 with entries 4 (first entry) and −3 (second
−3
entry).
 
−2
ˆ The vector  5  is in R3 with entries −2 (first entry), 5 (second entry),
0
and 0 (third entry).

Vector Addition and Scalar Multiplication

   
u1 v1
 u2   v2 
Let n ≥ 1 be an integer. Let u =  ..  and v =  ..  be vectors in Rn ,
   
 .   . 
un vn
and let c be a scalar (a real number).

111
112 Chapter 2. Vectors in Rn and Linear Systems

ˆ Then the sum u + v is the vector of Rn whose ith entry is the sum of the
ith entry of u and the ith entry of v. That is,
 
u1 + v1
 u2 + v2 
u+v =
 
.. 
 . 
un + vn

ˆ The scalar multiplication of c and u, denoted cu, is the vector of Rn

whose ith entry is the product of c and the ith entry of u. That is,
 
cu1
 cu2 
cu =  .. 
 
 . 
cun

       
1 −3 1 + (−3) −2
(a) u + v = + = =
2 4 2+4 6
     
6 10(6) 60
(b) 10w = 10 = =
−5 10(−5) −50
         
6 1 6 −2 4
(c) w − 2u = −2 = + =
−5 2 −5 −4 −9
     
−3 −3(−3) 9
(d) −3v = −3 = =
4 −3(4) −12

(e) The vector 3u − v − 2w is

     
1 −3 6
3u − v − 2w = 3 − −2
2 4 −5
         
3 −3 12 3 − (−3) − 12 −6
= − − = =
6 4 −10 6 − 4 − (−10) 12

(f ) The vector −u − 3v + 4w is
     
1 −3 6
−u − 3v + 4w = − −3 +4
2 4 −5
       
−1 9 24 32
= + + =
−2 −12 −20 −34

2. [Exercise on page 11]

2.1. Vector Equations and Span 113

Geometry of Vector Addition and Scalar Multiplication

ˆ Let u and v be two vectors as shown below.

y
4

−1 4 x
−1

To find the sum of u and v geometrically, first we need to construct the

parallelogram generated by u and v:

y
4

−1 4 x
−1

Then the sum u + v is the diagonal of the parallelogram as shown below:

y
4 u+v

−1 4 x
−1

This construction is referred to as the parallelogram rule for addition.

114 Chapter 2. Vectors in Rn and Linear Systems

ˆ For the multiplication of a vector u by a real number c, there are three

possibilities:

– If c is positive, then cu goes in the same direction as u.

– If c is negative, then cu and u go in opposite directions.
– c = 0, then cu is the zero vector.

All this is illustrated in the following figure.

y
3 3u

−2 3 x

−2u −2

(a) and (b) The vectors 3u and −2v are shown below.

y
3u
v
u

−4 5 x

−2v −4

(c) and (d) The vectors u + v and u − v are shown below.

2.1. Vector Equations and Span 115

y
3 u+v
v
u

−3 3 x
u−v
−v

(e) and (f ) The vectors −2u + 3v and −4u − v are shown below.

y
6 3v

−2u + 3v
3
v
u

−6 −3 3 6 x
−v −2u
−3
−4u

−6
−4u − v

3. [Exercise on page 11]

The Zero Vector and Equality of Vectors

Let n ≥ 1 be an integer.

ˆ The zero vector of Rn , denoted 0, is the vector whose every entry is 0,

that is,  
0
 0 
0 =  .. 
 
 . 
0
   
u1 v1
 u2   v2 
ˆ Two vectors u =   and v =   of Rn are equal if for every
   
.. ..
 .   . 
un vn
116 Chapter 2. Vectors in Rn and Linear Systems

1 ≤ i ≤ n, the ith entry of u is equal to the ith entry of v; that is,

u1 = v1 , u2 = v2 , · · · , un = vn

(a) (u + v) + w = u + (v + w) (e) a(u + v) = au + av

(b) u + 0 = 0 + u = u (f ) (a + b)u = au + bu
(c) u + (−u) = −u + u = 0 (g) a(bu) = (ab)u
(d) u + v = v + u (h) 1 · u = u
 
u1
 u2 
Proof of (b). Let u =   be a vector of Rn . We want to show that u + 0 =
 
..
 . 
un
0 + u = u.

   
u1 0
 u2   0 
u+0= +
   
.. .. 
 .   . 
un 0
 
u1 + 0
 u2 + 0 
= By definition of vector addition
 
.. 
 . 
un + 0

 
u1
 u2 
= =u
 
..
 . 
un

Similarly, one can show that 0 + u = u.

   
u1 v1
 u2   v2 
Proof of (e). Let u =  ..  and v =  ..  be vectors in Rn . Let a be a real
   
 .   . 
un vn
number. We want to prove that a(u + v) = au + av.
   
u1 v1
 u2   v2 
a(u + v) = a  ..  +  .. 
   
 .   . 
un vn
2.1. Vector Equations and Span 117

 
u1 + v1
 u2 + v2 
= a By definition of vector addition
 
.. 
 . 
un + vn
 
au1 + av1
 au2 + av2 
= By definition of scalar multiplication
 
.. 
 . 
aun + avn
   
au1 av1
 au2   av2 
= + By definition of vector addition
   
.. .. 
 .   . 
aun avn
   
u1 v1
 u2   v2 
= a  ..  + a  .. By definition of scalar multiplication
   

 .   . 
un vn
= au + av

4. [Exercise on page 11]

Vector Equations
Let n, m ≥ 1 be integers. A vector equation in n variables x1 , x2 , · · · , xn is
an equation of the form

x1 a1 + x2 a2 + · · · xn an = b,

where a1 , · · · , an , and b are vectors in Rm .

(a) Using the definition of scalar multiplication, the vector equation

     
1 −2 5
x1  2  + x2  −1  =  4 
3 0 6
is equivalent to      
x1 −2x2 5
 2x1  +  −x2  =  4 
3x1 0 6
Using the definition of vector addition, this latter equation becomes
   
x1 − 2x2 5
 2x1 − x2  =  4 
3x1 6
118 Chapter 2. Vectors in Rn and Linear Systems

By the definition of equality of two vectors, the latter equation gives the following
linear system:
x1 − 2x2 = 5
2x1 − x2 = 4
3x1 = 6

(b) The linear system corresponding to the vector equation

       
0 −2 5 10
x1 + x2 + x3 =
1 1 0 12

is
− 2x2 + 5x3 = 10
x1 + x2 = 12

5. [Exercise on page 12]

(a) The vector equation corresponding to the system

x1 − x2 = 3
2x1 + 7x2 = −8

is      
1 −1 3
x1 + x2 =
2 7 −8

(b) The vector equation corresponding to the system

2x1 − x2 + x3 = 5
−3x1 + 4x2 + 5x3 = 9
8x1 + 6x3 = 11

is        
2 −1 1 5
x1  −3  + x2  4  + x3  5  =  9 
8 0 6 11

6. [Exercise on page 12]

x1 + 2x2 = 4
−4x1 − 3x2 = −1

(a) Solution. The graphs of x1 + 2x2 = 4 and −4x1 − 3x2 = −1 are shown below.
2.1. Vector Equations and Span 119

x2
x1 + 2x2 = 4
3 −4x1 − 3x2 = −1

−2 3 x1

−2

Geometrically, we see that the lines intersect at (−2, 3). This means that the
solution to the system is (−2, 3). The figure above is referred to as the row
picture of the system.
(b) To solve the system algebraically, we need to reduce the augmented matrix.
   
1 2 4 1 2 4
R2 + 4R1 → R2
−4 −3 −1 0 5 15
 
1 1 2 4
R
5 2
→ R2
0 1 3
 
1 0 −2
R1 − 2R2 → R1
0 1 3

So the solution is (−2, 3), which is, as expected, the same as the one found in
part (a).
(c) The vector equation corresponding to the system is
     
1 2 4
x1 + x2 =
−4 −3 −1

ˆ Verifying algebraically that (−2, 3) is a solution to the vector equation.

       
1 2 −2 6
−2 +3 = +
−4 −3 8 −9
   
−2 + 6 4
= =
8−9 −1

ˆ Verifying geometrically that (−2,3) is asolution to the vector

 equation. We
1 2
need to graph the vectors u = and v = , and verify that
  −4 −3
4
−2u + 3v = b, where b = .
−1
120 Chapter 2. Vectors in Rn and Linear Systems

y
−2u 8

−9 b 8 x

v
u

−9 3v
Geometrically we see that −2u + 3v is indeed equal to b. The figure here is
referred to as the column picture of the system above.

7. [Exercise on page 12]

True. Actually this is a fact, which follows immediately from the definition of vector
equation:

Linear Systems and Vector Equations - Solutions

A vector equation x1 a1 + x2 a2 + · · · + xn an = b has the same set of solutions
 as
the linear system whose augmented matrix is a1 a2 · · · an b .

8. [Exercise on page 12]

 
We need to reduce the augmented matrix a1 a2 · · · an b .
   
1 1 1 −5 1 1 1 −5
 2 −1 −1 −4  R2 − 2R1 → R2  0 −3 −3 6 
R3 + R1 → R3
−1 0 1 −4 0 1 2 −9
 
1 1 1 −5
− 31 R2 → R2  0 1 1 −2 
0 1 2 −9
 
1 1 1 −5
R3 − R2 → R3  0 1 1 −2 
0 0 1 −7
 
1 1 0 2
R1 − R3 → R1  0 1 0 5 
R2 − R3 → R3
0 0 1 −7
 
1 0 0 −3
R1 − R2 → R1  0 1 0 5 
0 0 1 −7
2.1. Vector Equations and Span 121

So the solution is (−3, 5, −7), that is, x1 = −3, x2 = 5, and x3 = −7.

9. [Exercise on page 12]

Linear Combinations
Let n, p ≥ 1 be integers, and let a1 , · · · , ap be vectors in Rn . A linear
combination of a1 , · · · , ap is a vector of the form x1 a1 + · · · + xn an , where
x1 , · · · , xn are real numbers.

Note. The coefficients x1 , x2 , · · · , xn can be any real numbers (including zero).

(a) We want to determine if there exists numbers x1 and x2 such that

     
1 −2 8
x1  −3  + x2  4  =  −18 
0 5 −15

x1 − 2x2 = 8
This amounts to determining if the system −3x1 + 4x2 = −18 has a so-
5x2 = −15
lution. To solve this, we need to reduce the augmented matrix.
   
1 −2 8 1 −2 8
 −3 4 −18  R2 + 3R1 → R2  0 −2 6 
0 5 −15 0 5 −15
 
1 1 −2 8
− 2 R2 → R2
1
 0 1 −3 
R3 → R 3
5 0 1 −3

 
1 −2 8
R3 − R2 → R3  0 1 −3 
0 0 0
 
1 0 2
R1 + 2R2 → R1  0 1 −3 
0 0 0

The corresponding system is x1 = 2 and x2 = −3. So 2a1 − 3a2 = b, which

shows that b is a linear combination of a1 and a2 .

(b) We want to determine whether there exist x1 and x2 such that

     
1 4 1
x1  −2  + x2  −11  =  1  .
3 −1 29
122 Chapter 2. Vectors in Rn and Linear Systems

x1 + 4x2 = 1
This amounts to determining if the system −2x1 − 11x2 = 1 has a so-
3x1 − x2 = 29
lution. To solve this, we need to reduce the augmented matrix.
   
1 4 1 1 4 1
 −2 −11 1  R2 + 2R1 → R2  0 −3 3 
R3 − 3R1 → R3
3 −1 29 0 −13 26
 
1 1 4 1
− 3 R2 → R2  0 1 −1 
1
− 13 R3 → R3
0 1 −2
 
1 4 1
R3 − R2 → R3  0 1 −1 
0 0 −1
Since the third equation of the corresponding system is of the form 0 = d with
d = −1 6= 0, the system has no solution. This implies that b is not a linear
combination of a1 and a2 .

10. [Exercise on page 13]

Let a1 be the first column of A, let a2 be the second column, and let a3 be the third
column. We want to determine if there exist x1 , x2 , and x3 such that
       
1 2 0 −10
x1  1  + x2  1  + x3  3  =  12  .
0 −1 4 27
 
We need to reduce the augmented matrix a1 a2 a3 b of the corresponding
system.

   
1 2 0 −10 1 2 0 −10
 1 1 3 12  R2 − R1 → R2  0 −1 3 22 
0 −1 4 27 0 −1 4 27
 
1 2 0 −10
(−1)R2 → R2  0 1 −3 −22 
0 −1 4 27
 
1 2 0 −10
R3 + R2 → R3  0 1 −3 −22 
0 0 1 5
 
1 2 0 −10
R2 + 3R3 → R2  0 1 0 −7 
0 0 1 5
 
1 0 0 4
R1 − 2R2 → R1  0 1 0 −7 
0 0 1 5
2.1. Vector Equations and Span 123

The corresponding system is x1 = 4, x2 = −7, and x3 = 5. So 4a1 − 7a2 + 5a3 = b.

Thus, b is a linear combination of the columns of A.

11. [Exercise on page 13]

Spanning Set
Let a1 , a2 , · · · , ap be vectors in Rn . The span of a1 , a2 , · · · , ap , denoted
Span{a1 , · · · , ap }, is the set of all linear combinations of a1 , · · · , ap .

ˆ So, a vector b is in Span{a1 , · · · , ap } if and only if b is a linear combination

of a1 , · · · , ap .

ˆ The set Span{a1 , · · · , ap } is also called the subset of Rn generated (or

spanned) by a1 , a2 , · · · , ap .

(a) ˆ The vector a1 is in Span{a1 , a2 } since a1 can be written as a linear combi-

nation of a1 and a2 :
a1 = 1a1 + 0a2
ˆ Likewise, a2 is a vector in Span{a1 , a2 }.
 
−1
ˆ The vector a1 + a2 = is clearly in Span{a1 , a2 }.
5
(b) As before, the vectors
    
−1 −4 −5
a1 =  0  , a2 =  3  , and a1 + a2 =  3 
2 6 8

are in Span{a1 , a2 }.

12. [Exercise on page 13]

We want to determine whether b is in Span{a1 , a2 }. This amounts to determining
whether there exist x1 and x2 such that x1 a1 + x2 a2 = b. To solve this, we need to
reduce the corresponding augmented matrix.

   
−1 4 3 1 −4 −3
 2 −5 −3  (−1)R1 → R1  2 −5 −3 
3 0 4 3 0 4
 
1 −4 −3
R2 − 2R1 → R2  0 3 3 
R3 − 3R1 → R3
0 12 13
 
1 −4 −3
1
R
3 2
→ R2  0 1 1 
0 12 13
124 Chapter 2. Vectors in Rn and Linear Systems

 
1 −4 −3
R3 − 12R2 → R3  0 1 1 
0 0 1

The third equation of the corresponding system is of the form 0 = d with d = 1 6= 0.

This means that b is not a linear combination of a1 and a2 . In other words, b is not
in Span{a1 , a2 }.

13. [Exercise on page 13]

 
We need to reduce the augmented matrix a1 a2 a3 b .
   
1 3 1 0 1 3 1 0
 −2 −4 −1 1  R1 + 2R1 → R1  0 2 1 1
  
 
 1 0 2 6  R3 − R1 → R3  0 −3 1 6 
0 5 4 7 0 5 4 7
 
1 3 1 0
 0 2 1 1 
R4 − 2R2 → R4 
 0 −3 1 6


0 1 2 5
 
1 3 1 0
 0 1 2 5 
R2 ↔ R4 
 0 −3 1 6


0 2 1 1
 
1 3 1 0
R3 + 3R2 → R3  0 1
 2 5 

R4 − 2R2 → R4  0 0 7 21 
0 0 −3 −9

 
1 30 1
1
R
7 3
→ R3  0
 15 
 2
1  0 03  1
− 3 R4 → R4
0 03 1
 
1 3 1 0
 0 1 2 5 
R4 − R3 → R4 
 0

0 1 3 
0 0 0 0
 
1 3 0 −3
R1 − R3 → R1  0
 1 0 −1 

R2 − 2R3 → R2  0 0 1 3 
0 0 0 0
 
1 0 0 0
 0 1 0 −1 
R1 − 3R2 → R2 
 0

0 1 3 
0 0 0 0
2.1. Vector Equations and Span 125

So x1 = 0, x2 = −1, and x3 = 3. This implies that 0a1 − a2 + 3a3 = b, and therefore

b is in Span{a1 , a2 , a3 }.

14. [Exercise on page 13]

We need to reduce the augmented matrix
 
1 −3 α
 0 1 1 
−3 6 2

   
1 −3 α 1 −3 α
 0 1 1  R3 + 3R1 → R3  0 1 1 
−3 6 2 0 −3 2 + 3α
 
1 −3 α
R3 + 3R2 → R3  0 1 1 
0 0 5 + 3α

The corresponding system is

x1 − 3x2 = α
x2 = 1
0 = 5 + 3α

This latter system has a solution if and only if 3α + 5 = 0, or α = − 53 . So b is in

Span{a1 , a2 } if and only if α = − 35 .

15. [Exercise on page 13]

To show that Span{u, v} = R2 , we need to show that every vector in R2 is a linear
h
combination of u and v. Let w = be an arbitrary vector of R2 . We need to
k
find x1 and x2 such that x1 u + x2 v = w. To do this, we first need to reduce the
corresponding augmented matrix.
   
1 3 h 1 3 h
R2 − 2R1 → R2
2 4 k 0 −2 k − 2h
 
1 1 3 h
− 2 R2 → R2
0 1 − 21 k + h
1 0 32 k − 2h
 
R1 − 3R2 → R1
0 1 − 21 k + h

The reduced echelon form indicates that x1 = 32 k − 2h and x2 = − 21 k + h. So every

vector w of R2 can be written as a linear combination of u and v. This proves that
Span{u, v} = R2 .
126 Chapter 2. Vectors in Rn and Linear Systems

16. [Exercise on page 13]

Let a1 be the firstcolumn
 of A, let a2 be the second column, and let a3 be the third
h
column. Let b =  k  be an arbitrary vector of R3 . We need to show that there
l
exist real numbers x1 , x2 , and x3 suchthat x1 a1 + x2 a2 + x3 a3 = b. To do this, we
need to reduce the augmented matrix a1 a2 a3 b .

   
1 2 8 h 1 2 8 h
 −2 R2 + 2R1 → R2
3 7 k   0 7 23 k + 2h 
R3 − 3R1 → R3
3 −1 2 l 0 −7 −22 l − 3h
 
1 2 8 h
R3 + R2 → R3  0 7 23 k + 2h 
0 0 1 l+k−h
 
1 2 0 −8l − 8k + 9h
R1 − 8R3 → R1  0 7 0 −23l − 22k + 25h 
R2 − 23R3 → R2
0 0 1 l+k−h
 
1 2 0 −8l − 8k + 9h
1
R → R2  0 1 0 − 23 l − 22 k + 25 h 
 
7 2 7 7 7
0 0 1 l+k−h
0 0 − 10 l − 12 k + 13
 
1 7 7 7
h
R1 − 2R2 → R1  0 1 0 − 23 l − 22 k + 25 h 
 
7 7 7
0 0 1 l+k−h

Since the corresponding linear system has a solution,

x1 = − 10
7
l− 12
7
k + 13
7
h
x2 = − 23
7
l− 22
7
k + 25
7
h
x3 = l + k − h,

it follows that b is a linear combination of the columns of A, and since this holds for
an arbitrary b in R3 , it follows that the columns of A span R3 .

17. (True or False). [Exercise on page 13]

True. This follows from the following theorem.

Spanning Sets and Linear Combinations

Theorem 2.1.1. Let v1 , v2 , · · · , vp be vectors in Rn , and let u be a linear com-
bination of these vectors. Then

Span{v1 , v2 , · · · , vp } = Span{v1 , v2 , · · · , vp , u}
2.2. The Matrix Equation Ax = b 127

18. [Exercise on page 13]

False. This follows from the following theorem.

Number of Vectors Needed to Span Rn

Theorem 2.1.2. Let v1 , v2 , · · · , vp be vectors in Rn . If p < n, then

6 Rn
Span{v1 , v2 , · · · , vp } =

In words, this theorem says that if the number of vectors in a set is less than
n, that set cannot span Rn . So, to span Rn one needs at least n vectors.

2.2 The Matrix Equation Ax = b

1. [Exercise on page 14]

Multiplication of a Matrix by a Vector

Let n, m ≥ 1 be integers. Let a1 , a2 , · · · , an be vectors in Rm , and
 let 
A =
x1
   x2 
a1 a2 · · · an be the matrix whose ith column is ai . Let x =  ..  be
 
 . 
xn
a vector in R . The product of A and x, denoted Ax, is the vector of Rm
n

defined as
Ax = x1 a1 + x2 a2 + · · · + xn an
Note. The product Ax is defined if and only if the number of columns of A is
equal to the number of entries (or rows) of x.

(a)
      
1 3 5 1 3
Ax = =5 +6 By definition
2 4 6 2 4
   
5 + 18 23
= =
10 + 24 34

(b)
 
  3      
−1 0 −3  4 =3 −1 0 −3
Ax = +4 +5
2 4 −5 2 4 −5
5
   
−3 + 0 − 15 −18
= =
6 + 16 − 25 −3

(c) The product Ax is undefined because the number of columns of A, which is 3,

is not equal to the number of entries of x, which is 2.
128 Chapter 2. Vectors in Rn and Linear Systems

(d)
         
1 −1   1 −1 7+2 9
7
Ax =  2 −3  = 7  2  − 2  −3  =  14 + 6  =  20 
−2
3 0 3 0 21 − 0 21

2. [Exercise on page 14]

False. If A has 5 rows and 8 columns, the vector b will have 5 entries (not 8!). In
general, in the equation Ax = b, the number of columns of A corresponds to the
number of entries of x, while the number of rows of A corresponds to the number of
entries of b.

3. [Exercise on page 14]

Matrix Equation
 
x1
A matrix equation in one variable x =  ...  is an equation of the form
 
xn

Ax = b

where

ˆ A is an m × n matrix called matrix of coefficients.

ˆ b is a vector in Rm called vector of constant terms.

7x1 − 6x2 = 13
(a) For the system , the corresponding matrix equation is
−8x1 + 9x2 = 5
Ax = b, where
     
7 −6 x1 13
A= , x= , b=
−8 9 x2 5

(b) The matrix equation corresponding to the system

x1 + x2 − x3 = 0
2x1 + x3 = 1
3x2 − 4x3 = 7

is Ax = b, where
     
1 1 −1 x1 0
A= 2 0 1 , x =  x2  , b= 1 
0 3 −4 x3 7

4. [Exercise on page 14]

2.2. The Matrix Equation Ax = b 129

(a) ˆ The linear system corresponding to the matrix equation Ax = b is

x1 + 3x2 − 4x3 = 0
−5x1 + 2x2 + x3 = −3
6x1 − x2 − 17x3 = 8

ˆ The associated vector equation is

       
1 3 −4 0
x1  −5  + x2  2  + x3  1  =  −3 
6 −1 −17 8

(b) ˆ The linear system corresponding to the matrix equation Ax = b is

− 2x3 + 3x4 + x5 = 4
2x1 − x2 + 3x3 + 4x5 = 23
x1 + x2 + 12x3 = 0
−3x1 − 2x2 − 9x4 + 9x5 = 32

ˆ The associated vector equation is

           
0 0 −2 3 1 4
 2   −1   3
 + x4  0  + x5  4 
     23 
x1 
 1  + x2  1  + x3  12
   
 0 
= 
  0   0 
−3 −2 0 −9 9 32

5. [Exercise on page 14]

Three Different Ways to View a Linear System

A linear system can be viewed as

ˆ a finite collection of linear equations, or as

ˆ a vector equation, or as

ˆ a matrix equation.

The following theorem says that whatever point of view you choose, you will
get the same solution set.

Theorem 2.2.1. Let n, m ≥ 1 be integers. Let a1 , a2 , · · · , an be vectors in

Rm ,   A = a1 a2 · · ·
and let an be the matrix whose ith column is ai . Let
x1 b1
 x2   b2 
x =  .. , and let b =  ..  be a vector in Rm . Then
   
 .   . 
xn bm

ˆ the linear system whose augmented matrix is

 
a1 a2 · · · an b ,
130 Chapter 2. Vectors in Rn and Linear Systems

ˆ the vector equation

x1 a1 + x2 a2 + · · · + xn an = b,

and

ˆ the matrix equation Ax = b

have the same solution set.

(a) Thanks to Theorem 2.2.1, solving the equation Ax = b amounts to solving the
linear system whose augmented matrix is
 
  4 −10 10 38
A b =  3 −5 15 16 
3 −6 12 21
Reducing this, we get
   
4 −10 10 38 4 −10 10 38
 3 −5 15 16  1
R
3 3
→ R3  3 −5 15 16 
3 −6 12 21 1 −2 4 7
 
1 −2 4 7
R1 ↔ R3  3 −5 15 16 
4 −10 10 38
 
1 −2 4 7
R2 − 3R1 → R1  0 1 3 −5 
R3 − 4R1 → R1
0 −2 −6 10
 
1 −2 4 7
R3 + 2R2 → R3  0 1 3 −5 
0 0 0 0
 
1 0 10 −3
R1 + 2R2 → R1  0 1 3 −5 
0 0 0 0
The linear system corresponding to the reduced echelon form is
x1 10x3 = −3
x2 + 3x3 = −5
Set x3 = s. Then x1 = −3 − 10s and x2 = −5 − 3s. So the solutions to the
matrix equation Ax = b are vectors of the form
 
−3 − 10s
x =  −5 − 3s 
s
where s is a real number. (So the equation Ax = b has infinitely many solutions.)
2.2. The Matrix Equation Ax = b 131

(b) According to Theorem 2.2.1, the vector equation x1 a1 + x2 a2 + x3 a3 = b has the 
same solution set as the linear system whose augmented matrix is a1 a2 a3 b .
So, using the result from part (a), the general solution to the equation x1 a1 +
x2 a2 + x3 a3 = b is
x1 = −3 − 10s
x2 = −5 − 3s
x3 = s
where s can be any real number.
6. [Exercise on page 15]
(a) We first need to reduce the augmented matrix.
   
−1 4 b1 1 −4 −b1
(−1)R1 → R1
2 −8 b2 2 −8 b2
 
1 −4 −b1
R2 − 2R1 → R2
0 0 b2 + 2b1
If the variables are x1 and x2 , the corresponding linear system will be
x1 − 4x2 = −b1
0 = b2 + 2b1
So the system has a solution if and only if 2b1 + b2 = 0. Thus, the system does
not have a solution for all possible b1 and b2 because some choices of b1 and b2 can
make 2b1 + b2 nonzero. For example, if b1 = 0 and b2 = 1, then 2b1 + b2 = 1 6= 0.
(b) From part (a), the system Ax = b does have a solution if and only if the entries
of b satisfy 2b1 + b2 = 0. This is the equation of a line through the origin in R2 .
To be more precise, we have the following description:
   
b1 b1
= Because b2 = −2b1
b2 −2b1
 
1
= b1 Factoring out b1
−2
So the system Ax = b has a solution
 if and only if the vector b lies on the line
1
in R2 spanned by the vector . (That line is shown below.)
−2
b2

−2 −1 1 2 b1

−2

−4
132 Chapter 2. Vectors in Rn and Linear Systems

7. [Exercise on page 15]

(a) We need to reduce the augmented matrix.

   
1 −2 −5 b1 1 −2 −5 b1
R2 − 2R1 → R2
 2 4 22 b2   0 8 32 b2 − 2b1 
R3 + 3R1 → R3
−3 5 11 b3 0 −1 −4 b3 + 3b1
1 −2 −5
 
b1
1
1
R
8 2
→ R2  0 1 4 b − 14 b1 
8 2
0 −1 −4 b3 + 3b1
1 −2 −5
 
b1
1
R3 + R2 → R3  0 1 4 b − 14 b1 
8 2
11
0 0 0 b + 18 b2 + b3
4 1

So the system has a solution if and only if 11

4 1
b + 81 b2 + b3 = 0. Multiplying both
sides by 8, this latter equation becomes 22b1 + b2 + 8b3 = 0. This shows that
the system is consistent only for the values of b1 , b2 , and b3 satisfying this latter
equation.
(b) From part (a), the system Ax = b has a solution if and only if the entries of b
satisfy the equation
22b1 + b2 + 8b3 = 0

This is the equation of a plane through the origin in R3 . To be more precise, we

have the following description:
   
b1 b1
 b2  =  −22b1 − 8b3  Since b2 = −22b1 − 8b3
b3 b3
   
b1 0
=  −22b1  +  −8b3 
0 b3
   
1 0
= b1  −22  + b3  −8 
0 1

So the system Ax = b has a solution

 ifand only
 if the
 vector b lies on the plane
1 0
in R3 spanned by the vectors  −22  and  −8 .
0 1

8. [Exercise on page 15]

True. This follows from the following nice result.
2.2. The Matrix Equation Ax = b 133

Existence of Solutions – Equivalent Statements

Three statements P, Q and R are equivalent means that if one is true, so are
the other two (or if one is false, so are the other two).
m
Theorem 2.2.2. Let n, m ≥ 1 be
 integers. Let a1 , a2 , · · · , an be vectors in R ,
and let A = a1 a2 · · · an be the matrix whose ith column is ai . Then
the following three statements are equivalent.

(i) For every b in Rm , the equation Ax = b has at least one solution.

(ii) Every vector b in Rm can be written as a linear combination of the columns

of A. (This means that the columns of A span Rm .)

(iii) Every row of A has a pivot position.

9. [Exercise on page 15]

False. For example, consider the matrix
 
1 0 0
A= 0 1 0 
0 0 0

Clearly,
 every
 column of A is a pivot column except the right most one. But, for
0
b =  0 , the system Ax = b does not have a solution.
1

10. [Exercise on page 15]

   
1 2 0 1 1 0 0 0
 0 1 −1 −2   0 1 0 1 
A=
 −1 −2
 B= 
1 −1   −3 0 1 −1 
2 5 −1 0 4 0 0 1

(a) Reducing A to an echelon form, we get

   
1 2 0 1 1 2 0 1
 0
 1 −1 −2   R3 + R1 → R3  0
 1 −1 −2 

 −1 −2 1 −1  R4 − 2R1 → R4  0 0 1 0 
2 5 −1 0 0 1 −1 −2
 
1 2 0 1
 0 1 −1 −2 
R4 − R2 → R4 
 0

0 1 0 
0 0 0 0

Since this latter matrix has three pivot columns, it follows that the matrix A
has three pivot positions ((1, 1), (2, 2), and (3, 3)).
134 Chapter 2. Vectors in Rn and Linear Systems

(b) No. Indeed, from part (a), an echelon form of A does not have a pivot in
every row (actually there is a pivot in every row except the fourth one). So the
statement (iii) from Theorem 2.2.2 is false for this matrix A. This implies that
the statement (i) is also false, that is, the equation Ax = b does not have a
solution for every b in R4 .
(c) To find the number of pivot positions of the matrix B, we need to put it into an
echelon form.
   
1 0 0 0 1 0 0 0
 0 1 0
 1 
 R3 + 3R1 → R3  0
 1 0 1 

 −3 0 1 −1  R4 − 4R1 → R4  0 0 1 −1 
4 0 0 1 0 0 0 1

From this latter matrix, it follows that the number of pivot positions of B is 4.
(d) Yes. Since any two echelon forms of B have the same pivot positions (thanks
to Theorem 1.2.2) and since the echelon form above has four pivot positions, it
follows that any echelon form of B has four pivot positions. Since this number
is equal to the number of rows of B, it follows that any echelon form of B has
a pivot in every row. So the statement (iii) from Theorem 2.2.2 is true for the
matrix B. Therefore the statement (ii) from the same theorem is also true. So
the columns of B span R4 .
(e) Yes. As we saw above, any echelon form of A does not have a pivot in every row
(as there is no pivot in the fourth row). So the columns of A do not span R4 .
Therefore, there exists at least one vector in R4 which is not a linear combination
of the columns of A.

11. [Exercise on page 15]

(a) We need to reduce the augmented matrix.

   
1 2 3 b1 1 2 3 b1
R2 − 2R1 → R2  0 −1
 2 3 5 b2  −1 b2 − 2b1 
R3 + R1 → R3
−1 4 4 b3 0 6 7 b3 + b1
 
1 2 3 b1
(−1)R2 → R2  0 1 1 −b2 + 2b1 
0 6 7 b3 + b1
 
1 2 3 b1
R3 − 6R2 → R3  0 1 1 −b2 + 2b1 
0 0 1 −11b1 + 6b2 + b3
 
1 2 0 34b1 − 18b2 − 3b3
R1 − 3R3 → R1  0 1 0 13b1 − 7b2 − b3 
R2 − R3 → R2
0 0 1 −11b1 + 6b2 + b3
 
1 0 0 8b1 − 4b2 − b3
R1 − 2R2 → R1  0 1 0 13b1 − 7b2 − b3 
0 0 1 −11b1 + 6b2 + b3
2.2. The Matrix Equation Ax = b 135

The corresponding system is

x1 = 8b1 − 4b2 − b3
x2 = 13b1 − 7b2 − b3
x3 = −11b1 + 6b2 + b3
This shows that the system has a solution for all possible b1 , b2 , and b3 .

Alternate Solution
By row reducing the coefficient matrix (not the augmented matrix), we
get  
1 0 0
 0 1 0 
0 0 1
Since this matrix has a pivot in every row, it follows that the system
Ax = b has a solution for all possible b1 , b2 , b3 (thanks to Theorem 2.2.2
and Theorem 1.2.2).

(b) Yes. Indeed, from part (a), the system Ax = b has a solution for each b in R3 .
This implies (by Theorem 2.2.2) that the columns of A span R3 .

12. [Exercise on page 15]

The Dot Product Rule

Another way to compute the product of a matrix A and a vector x is given
by the
 following
 rule called dot product rule (or row–vector rule): Let
x1
 x2 
x =  .. . Suppose that the number of columns of A is equal to the number
 
 . 
xn
 
of entries of x. If the ith row of A is ai1 ai2 · · · ain , then the ith entry
of Ax is the sum
ai1 x1 + ai2 x2 + · · · + ain xn

(a) Using this rule, we have

        
1 3 5 1(5) + 3(6) 5 + 18 23
= = =
2 4 6 2(5) + 4(6) 10 + 24 34

(b)
 
  3  
−1 0 −3  4 = (−1)(3) + 0(4) + (−3)(5)
2 4 −5 2(3) + 4(4) + (−5)(5)
5
   
−3 + 0 − 15 −18
= =
6 + 16 − 25 −3
136 Chapter 2. Vectors in Rn and Linear Systems

(c)      
1 −1   1(7) + (−1)(−2) 9
 2 −3  7
=  2(7) + (−3)(−2)  =  20 
−2
3 0 3(7) + 0(−2) 21

13. [Exercise on page 16]

Properties of the Product of a Matrix and a Vector

Theorem 2.2.3. Let A be an m × n matrix. Let u, v be vectors in Rn , and let
c be a real number (or scalar). Then we have the following properties.

(i) A(u + v) = Au + Av

(ii) A(cu) = cAu

To show that the system Ax = w is consistent, we need to find x such that Ax = w.

This amounts to finding x such that Ax = u0 + v 0 because w = u0 + v 0 . Since
u0 = Au and v 0 = Av, the system Ax = w is equivalent to Ax = Au + Av. By using
now property (i) of Theorem 2.2.3, the system Ax = w is equivalent to the system
Ax = A(u + v). Since x = u + v is a solution to the latter system, it follows that
x = u + v is also a solution to the original system Ax = w. This proves that the
system Ax = w is consistent.

14. [Exercise on page 16]

Multiplying both sides of the equation Au = v by 3, we get 3Au = 3v. Using
property (ii) of Theorem 2.2.3, this latter equation becomes A(3u) = 3v. So x = 3u
is a solution to the system Ax = 3v.

2.3 Writing the General Solution of a Linear System

in Vector Form
1. [Exercise on page 16]

Vector Form
Let x be the general solution to a linear system. Writing x in vector form
amounts to writing x as a linear combination of a set of vectors.

x1 = 3t
(a) A vector form of x2 = −4t is
x3 = t
   
x1 3t
x =  x2  =  −4t 
x3 t
2.3. Writing the General Solution of a Linear System in Vector Form 137

 
3
= t  −4  Factor out t
1
(b)
   
x1 5 + 6t
x =  x2  =  10 − 8t 
x3 t
   
5 6t
=  10  +  −8t  Split x into two vectors
0 t
   
5 6
=  10  + t  −8  Factor out t
0 1
(c)
   
x1 1 + s + 9t
 x2   −4 + 13t 
 x3  = 
x=   
s 
x4 t
     
1 s 9t
 −4   0   13t 
= 0 + s + 0 
     Split x into three vectors
0 0 t
     
1 1 9
 −4 
 + s  0  + t  13 
   
= 0   1   0  Factor out s and t
0 0 1
(d)
   
x1 4s − 7t

 x2  
  s 

x=
 x3 =
  −2 + 5t 

 x4   −3 
x5 t
     
0 4s −7t

 0   s  
   0 
=
 −2  +  0  +  5t 
   
 Split x into three vectors
 −3   0   0 
0 0 t
     
0 4 −7

 0 
 1 
 
 0 
 
=
 −2  + s  0  + t 
  
 5 
 Factor out s and t
 −3   0   0 
0 0 1
138 Chapter 2. Vectors in Rn and Linear Systems

(e)
   
x1 24 + 3s1 − 8s2 + 9s3

 x2  
  s1 

x
 x3 =
  −7 + 15s2 + 6s3  
 x4   s2 
x5 s3
       
24 3s1 −8s2 9s3

 0 
  s1  
   0   0 
 
=
 −7  +  0  +  15s2  +  6s3 
     
 0   0   s2   0 
0 0 0 s3
       
24 3 −8 9

 0 

 1 
 
 0 
 
 0 
 
=
 −7  + s1  0  + s2  15  + s3  6 
     
 0   0   1   0 
0 0 0 1

2. [Exercise on page 17]

True. This is the definition of a homogeneous system.

Homogeneous System
A linear system Ax = b is called homogeneous if b is the zero vector. So, a
homogeneous linear system is an equation of the form Ax = 0.

3. [Exercise on page 17]

False. Any homogeneous linear system has a solution (the trivial solution)

Trivial Solution
Let Ax = 0 be a homogeneous system. Clearly, if x is the zero vector, then
Ax = 0. So the zero vector is a solution to the equation
 Ax = 0. In other
x1
 x2 
words, if A is an m × n matrix, the vector x =  ..  where
 
 . 
xn

x1 = 0, x2 = 0, · · · , xn = 0

is a solution to the equation Ax = 0. This solution is called the trivial

solution.

If at least one entry of x is a nonzero number and x is a solution to the equation

Ax = 0, then x is called a nontrivial solution.
2.3. Writing the General Solution of a Linear System in Vector Form 139

4. [Exercise on page 17]

(a) We need to reduce the augmented matrix.
   
−3 5 0 −3 5 0
R2 + 2R1 → R2
6 −9 0 0 1 0
 
−3 0 0
R1 − 5R2 → R1
0 1 0
 
1 0 0
− 13 R1 → R1
0 1 0
The reduced echelon
 form indicates that x1 = 0 and x2 = 0. So the solution is
0
the vector x = . Since the trivial solution is the sole solution, it follows
0
that there is no nontrivial solution.
(b) We need to reduce the augmented matrix.
   
1 2 −2 0 1 2 −2 0
 −1 −1 R2 + R1 → R2
7 0   0 1 5 0 
R3 − 2R1 → R3
2 3 −9 0 0 −1 −5 0
 
1 2 −2 0
R3 + R2 → R3  0 1 5 0 
0 0 0 0
 
1 0 −12 0
R1 − 2R2 → R1  0 1 5 0 
0 0 0 0
The linear system corresponding to the reduced echelon form is
x1 − 12x3 = 0
x2 + 5x3 = 0
Set x3 = s. Then x1 = 12s and x2 = −5s, and the general solution is
     
x1 12s 12
x =  x2  =  −5s  = s  −5 
x3 s 1
where s can be any real number.

Since the system has infinitely many solutions, we can find nontrivial solutions.
For example,
   
12 12
ˆ if s = 1, then x = 1  −5  =  −5  is a nontrivial solution.
1 1
   
12 24
ˆ If s = 2, then x = 2  −5  =  −10  is also a nontrivial solution.
1 2
140 Chapter 2. Vectors in Rn and Linear Systems

Actually, each nonzero value of s gives a nontrivial solution.

5. [Exercise on page 17]
As usual, we need to reduce the augmented matrix.
   
1 −1 3 4 0 1 −1 3 4 0
 1 −1 R2 − R1 → R2
4 6 0   0 0 1 2 0 
R3 + 3R1 → R3
−3 3 −7 −8 0 0 0 2 4 0
 
1 −1 3 4 0
R3 − 2R2 → R3  0 0 1 2 0 
0 0 0 0 0
 
1 −1 0 −2 0
R1 − 3R2 → R1  0 0 1 2 0 
0 0 0 0 0
The corresponding linear system is
x1 − x2 − 2x4 = 0
x3 + 2x4 = 0
Set x2 = s and x4 = t. Then x1 = s + 2t and x3 = −2t, and the general solution is
       
x1 s + 2t s 2t
 x2   s 
= s + 0 
  
x= =
 x3   −2t   0   −2t 


x4 t 0 t
   
1 2
 1   0 
= s
 0  + t  −2 
  

0 1
where s and t can be any real numbers. This shows that the solution set is the set of
linear combinations of the vectors
   
1 2
 1   0 
a1 =  0  and a2 =  −2 
  

0 1
In other words, the solution set is Span{a1 , a2 }.
6. [Exercise on page 17]
False. Suppose the solution set to a nonhomogeneous linear system Ax = b (b 6= 0)
is the span of some vectors a1 , a2 , · · · , ap . Then every linear combination

x 1 a1 + x 2 a2 + · · · + x p ap

is a solution. In particular, the linear combination

x0 = 0a1 + 0a2 + · · · 0ap = 0,

2.3. Writing the General Solution of a Linear System in Vector Form 141

where every coefficient is 0, is a solution. So Ax0 = b. But Ax0 = 0 since x0 is the

zero vector. So the equation Ax0 = b becomes 0 = b. This contradicts the fact that
b 6= 0. So it is not possible to write the solution set of a nonhomogeneous
linear system as the span of a set of vectors.
7. [Exercise on page 17]
False. For example, consider the linear system
x1 + +x3 = 0
x2 +x3 = 0
 
1 0 1
The coefficient matrix, A = , is an m × n matrix with m = 2 and n = 3.
0 1 1
Setting x3 = s, we find that the general solution is
     
x1 −s −1
x =  x2  =  −s  = s  −1 
x3 s 1

where s can be any real number. For each value of s, the vector x lies in Rn = R3
(since the vector x has three entries). So the solution set is not a subset of Rm = R2 .
Note. In general the solution set to a system Ax = b, where A is an m × n matrix,
is a subset of Rn .
8. [Exercise on page 17]
True. It is true because of the following theorem.

General Solution of a Linear System

A solution P to Ax = b is sometimes called a particular solution.

Theorem 2.3.1. Let A be an m × n matrix, and let b be a vector in Rm .

Suppose the system Ax = b is consistent. Then every solution x to the linear
system Ax = b has the form
x = P + Vh
where

ˆ P is a solution to Ax = b, and

ˆ Vh is a solution to the associated homogeneous equation Ax = 0.

9. [Exercise on page 17]

 
1 −2
A=
−4 8
(a) Reducing the augmented matrix, we get
   
1 −2 0 1 −2 0
R2 + 4R1 → R2
−4 8 0 0 0 0
142 Chapter 2. Vectors in Rn and Linear Systems

The corresponding system is x1 − 2x2 = 0. Setting x2 = s, we find x1 = 2s. And

the general solution is
     
x1 2s 2
x= = =s
x2 s 1

where s can be any real number. So the solution set to the

 equation
 Ax = 0
2
is the line through the origin spanned by the vector u = . This is shown
1
below.
(b) We need to reduce the augmented matrix.

   
1 −2 3 1 −2 3
R2 + 4R1 → R2
−4 8 −12 0 0 0

Set x2 = s. Then x1 = 3 + 2s, and the general solution is

           
x1 3 + 2s 3 2s 3 2
x= = = + = +s = P + su
x2 s 0 s 0 1

   
3 2
where P = and u = . Geometrically, this means that a solution to
0 1
the equation Ax = b is the vector P plus a multiple of the vector u. In other
words, the solution set to Ax = b is the line through the point (3, 0) parallel
to the solution to Ax = 0 as shown below. This amounts to saying that the
solution set to Ax = b is obtained by translating the solution set to Ax = 0 to
the right 3.

x2
Ax = 0

Ax = b
su
P + su
u

−3 P 7 x1

−3

10. [Exercise on page 18]

2.3. Writing the General Solution of a Linear System in Vector Form 143

x2
4

−3 4 x1

−3 L

 
0
(a) False. Since the line L passes through the origin, it contains the vector .
  0
0
So if (S) is a linear system whose solution set is L, then will be a solution
0
to (S). Therefore, (S) will be a homogeneous system.
(b) To find two nontrivial solutions,it suffices
 to find 
two nonzero vectors that lie
−2 2
in L. For example, the vectors and are nontrivial solutions to
3 −3
the linear system whose solution set is L.
 
−2
(c) In part (b), we saw that the vector u = is a solution. Actually, any
3
other solution is alinearcombination of u (or a multiple of u), that is, a vector
−2 x1
of the form x = s , where s is a real number. If x = , then
3 x2

x1 = −2s
x2 = 3s
x2
Solving the second equation for s, we find s = 3
. Substituting this into the first
equation, we get
x 
2
x1 = −2
3
3x1 = −2x2 Multiply both sides by 3
3x1 + 2x2 = 0 Adding 2x2 to both sides

So the solution set to the system

3x1 + 2x2 = 0
0 = 0

is L.
144 Chapter 2. Vectors in Rn and Linear Systems

Alternate Approach

From the graph, one can see that the slope of the line L is m = − 32 . So
an equation of L is of the form
3
x2 = mx1 + b = − x1 + b
2
Since L passes through the origin, it follows that b = 0. So the equation
of L is x2 = − 32 x1 . Multiplying both sides by 2, we get 2x2 = −3x1 or
3x1 + 2x2 = 0. We thus obtain the same equation as above.

11. [Exercise on page 18]

Let x1 , x2 , x3 , x4 , x5 be the variables. Then the linear system whose augmented matrix
is M is
x1 + x2 + 2x3 − 2x4 + 5x5 = 0
x3 − 3x4 + 8x5 = −13
x5 = 11
This system has 3 leading variables (x1 , x3 and x5 ) and 2 free variables (x2 and x4 ).
Set x2 = s and x4 = t.

ˆ From the third equation, we have x5 = 11.

ˆ Substituting this into the second equation, we get x3 −3t+8(11) = −13. Solving
this for x3 , we get x3 = −101 + 3t.
ˆ Substituting all this into the first equation, we get x1 + s + 2(−101 + 3t) − 2t +
5(11) = 0. Solving this for x1 , we get x1 = 147 − s − 4t.

So the general solution is

x1 = 147 − s − 4t
x2 = s
x3 = −101 + 3t
x4 = t
x5 = 11
where s and t can be any real numbers. Writing this in vector form, we get
   
x1 147 − s − 4t
 x2   s 
   
x=  x3  =  −101 + 3t 
  
 x4   t 
x5 11
     
147 −1 −4

 0 

 1 
 
 0 
 
=  −101  + s  0  + t 
   
 3 

 0   0   1 
11 0 0
= P + su + tv
2.3. Writing the General Solution of a Linear System in Vector Form 145

     
147 −1 −4

 0 


 1 


 0 

where P = 
 −101 , u = 
  0 , and v = 
  3 .

 0   0   1 
11 0 0
12. [Exercise on page 18]

(a) We need to reduce the augmented matrix.

   
1 −1 2 0 1 −1 2 0
 −1 R2 + R1 → R2
4 −7 0   0 3 −5 0 
R3 + R1 → R3
−1 10 −17 0 0 9 −15 0
 
1 −1 2 0
R3 − 3R2 → R3  0 3 −5 0 
0 0 0 0
 
1 −1 2 0
1
R
3 2
→ R2  0 1 − 35 0 
0 0 0 0

 1

1 0 3
0
R1 + R2 → R1  0 1 − 35 0 
 
0 0 0 0

The corresponding system is

1
x1 + x
3 3
= 0
5
x2 − x
3 3
= 0
0 = 0.
The leading variables are x1 and x2 , and x3 is free. Set x3 = t. Then the general
solution is
x1 = − 13 t
5
x2 = 3
t
x3 = t
where t can be any real number.

A vector form is
  1   1 
−3t −3

x1
x =  x2  =  3 t  = t  53  = tv,
5
   
x3 t 1
 
− 13
5
where v =  .
 
3
1
146 Chapter 2. Vectors in Rn and Linear Systems

(b) Solution. We first need to reduce the augmented matrix.

   
1 3 4 1 1 3 4 1
 −2 1 5 R2 + 2R1 → R2
2   0 7 13 4 
R3 − 7R1 → R3
7 7 2 −1 0 −14 −26 −8
 
1 3 4 1
R3 + 2R2 → R3  0 7 13 4 
0 0 0 0
 
1 3 4 1
1
R
7 2
→ R2  0 1 13 7
4 
7
0 0 0 0
 
1 0 − 11
7
− 5
7
13 4 
R1 − 3R2 → R1  0 1

7 7 
0 0 0 0

The corresponding system is

11
x1 − 7 3
x= − 75
13 4
x2 + 7 3
x= 7
0 = 0

The variables x1 and x2 are the leading variables while x3 is free. Set x3 = t.
Then the general solution is
x1 = − 57 + 11
7
t
4 13
x2 = 7
− 7
t
x3 = t

where t can be any real number.

A vector form is
  5 11   5   11 
−7 + 7 t −7

x1 7
x =  x2  =  74 − 13
  4   13 
t = + t − = P + tv,

7   7   7 
x3 t 0 1
 5   11 
−7 7
where P =  47  and v =  − 13 .
   
7 
0 1
(c) Solution. As usual, we first need to reduce the augmented matrix.

   
1 2 1 5 −4 1 2 1 5 −4
R2 − 2R1 → R2
 2 4 3 7 5   0 0 1 −3 13 
R3 + 3R1 → R3
−3 −6 −5 −9 −14 0 0 −2 6 −26
2.3. Writing the General Solution of a Linear System in Vector Form 147

 
1 2 1 5 −4
R3 + 2R2 → R3  0 0 1 −3 13 
0 0 0 0 0
 
1 2 0 8 −17
R1 − R2 → R1  0 0 1 −3 13 
0 0 0 0 0
The corresponding system is
x1 + 2x2 + 8x4 = −17
x3 − 3x4 = 13
0 = 0.
The leading variables are x1 and x3 , and the others are free. Set x2 = s and
x4 = t. Then the general solution in vector form is
       
−17 − 2s − 8t −17 −2 −8
 s   0 
 + s  1  + t  0  = P + su + tv,
   
x= =
 13 + 3t   13   0   3 
t 0 0 1
     
−17 −2 −8
 0 

 1 , v =  0 , and s, t can be any real
   
where P =   13  , u =  0   3 
0 0 1
numbers.
(d) Reducing the augmented matrix, we get
   
1 −2 −3 4 −9 1 −2 −3 4 −9
 1 −1 −4 R2 − R1 → R2
10 9   0 1 −1 6 18 
R3 + 2R1 → R3
−2 3 7 −14 0 0 −1 1 −6 −18
 
1 −2 −3 4 −9
R3 + R2 → R3  0 1 −1 6 18 
0 0 0 0 0
 
1 0 −5 16 27
R1 + 2R2 → R1  0 1 −1 6 18 
0 0 0 0 0
The corresponding system is
x1 − 5x3 + 16x4 = 27
x2 − x3 + 6x4 = 18
0 = 0
The leading variables are x1 and x2 , and the free ones are x3 and x4 . Set x3 = s
and x4 = t. Then the general solution in vector form is
       
27 + 5s − 16t 27 5 −16
 18 + s − 6t   18 
 + s  1  + t  −6  = P + su + tv,
   
x= =
 s   0   1   0 
t 0 0 1
148 Chapter 2. Vectors in Rn and Linear Systems

     
27 5 −16
, v =  −6 , and s, t can be any real numbers.
 18   1   
where P = 
 0 
, u = 
 1   0 
0 0 1

13. [Exercise on page 19]

(a) The matrix equation is Ax = b, where

 
  x1  
1 −2 3 1  x2  −3
A = −1
 2 −2 1  , x=
 x3  ,
 and b =  5 
2 −4 5 0 −8
x4

(b)
   
1 −2 3 1 −3 1 −2 3 1 −3
 −1 R2 + R1 → R2
2 −2 1 5   0 0 1 2 2 
R3 − 2R1 → R3
2 −4 5 0 −8 0 0 −1 −2 −2
 
1 −2 3 1 −3
R3 + R2 → R3  0 0 1 2 2 
0 0 0 0 0
 
1 −2 0 −5 −9
R1 − 3R2 → R1  0 0 1 2 2 
0 0 0 0 0

(c) The linear system corresponding to the reduced echelon form is

x1 − 2x2 − 5x4 = −9
x3 + 2x4 = 2
0 = 0
The leading variables are x1 and x3 , and the free ones are x2 and x4 . Set x2 = s
and x4 = t. Then the general solution is

x1 = −9 + 2s + 5t
x2 = s
x3 = 2 − 2t
x4 = t

where s and t can be any real numbers.

(d) A vector form of the general solution is
         
x1 −9 + 2s + 5t −9 2s 5t
 x2 s = 0 + s 0 
       
x= = + 
 x3   2 − 2t   2   0   −2t 
x4 t 0 0 t
2.3. Writing the General Solution of a Linear System in Vector Form 149

     
−9 2 5
 0 
 + s  1  + t  0  = P + su + tv,
   
= 2   0   −2 
0 0 1
     
−9 2 5
 0   1   0 
where P = 
 2  , u =  0 , and v =  −2 .
    

0 0 1


−9
 0 
(e) A particular solution to the equation Ax = b is the vector P = 
 2 .


0
(f ) Yes. Indeed, a particular solution is P , and the general solution to the associated
homogeneous system Ax = 0 is su + tv where s and t can be any real numbers.
When s = 1 and t = 0, the vector 1u + 0v = u is a nontrivial solution to the
equation Ax = 0.

14. [Exercise on page 19]

 
0 0
Since A is a 2 × 2 zero matrix, A = , it follows that A multiplied by any vector
  0 0
0
in R2 is the zero vector 0 = . This implies that every vector in R2 is a solution
0
to Ax = 0. So the solution set to the equation Ax = 0 is the set of all vectors in R2 .
In other words, the solution set is R2 .
15. [Exercise on page 19]
 
2 0 1
The matrix A = 0
 0 0 works because
0 0 0
        
2 2 0 1 2 4+0−4 0
A  3  = 0 0 0  3  =  0 + 0 + 0  =  0  .
−4 0 0 0 −4 0+0+0 0

16. [Exercise on page 19]

   
1 0
A vector x in the plane spanned by the vectors u = 0 and v = 1  is a linear
  
1 1
combination
  of u and v, that is, x = su + tv, where s and t are real numbers. If
x1
x =  x2 , then the equation x = su + tv becomes
x3
           
x1 1 0 s 0 s
 x2  = s  0  + t  1  =  0  +  t  =  t .
x3 1 1 s t s+t
150 Chapter 2. Vectors in Rn and Linear Systems

So
x1 = s
x2 = t
x3 = s + t

Substituting s with x1 and t with x2 into the third equation, we get x3 = x1 + x2 ,

or x1 + x2 − x3 = 0. (Note that the idea here is to eliminate the parameters and
get one equation (or more) involving only the variables.) This gives the following
homogeneous system
x 1 + x2 − x3 = 0
0 =0
0 =0

One can easily verify that this homogeneous system has the desired properties (three
equations and three variables, and the solution set is the plane spanned by u and v).

17. [Exercise on page 19]

 
x1
Let x =  x2  be a solution. Then
x3
           
x1 3 −4 3 −4s 3 − 4s
 x2  =  −1  + s  2  =  −1  +  2s  =  −1 + 2s 
x3 0 1 0 s s

So
x1 = 3 − 4s
x2 = −1 + 2s
x3 = s

As before, the idea is to find one equation (or more) involving only the variables.

ˆ Substituting s with x3 into the first equation, we get x1 = 3−4x3 or x1 +4x3 = 3.

This gives one equation.
ˆ Now, substituting s with x3 into the second equation, we get x2 = −1 + 2x3 or
x2 − 2x3 = −1. This gives another equation.

So the following linear system has the desired properties.

x1 + 4x3 = 3
x2 − 2x3 = −1
0 = 0

18. [Exercise on page 20]

2.4. Linear Independence 151

 
x1
 x2 
Let x = 
 x3  be a solution. Then


x4
         
x1 2 −3 7 2 − 3s + 7t
 x2   0   1   0   s 
 x3  =  5  + s  0
 + t
 −1  = 
       
 5−t 
x4 0 0 1 t
So
x1 = 2 − 3s + 7t
x2 = s
x3 = 5−t
x4 = t
ˆ Substituting s with x2 and t with x4 into the first equation, we have x1 =
2 − 3x2 + 7x4 . This latter equation is equivalent to x1 + 3x2 − 7x4 = 2.
ˆ Substituting t with x4 into the third equation, we get x3 = 5 − x4 . This gives
the equation x3 + x4 = 5.
So the following linear system has the desired properties.
x1 + 3x2 − 7x4 = 2
x3 + x4 = 5
0 = 0

2.4 Linear Independence

1. [Exercise on page 20]

Linear Independence

ˆ A set of vectors {v1 , v2 , · · · , vp } in Rn is said to be linearly independent

if the only solution to the vector equation

x1 v1 + x2 v2 + · · · xp vp = 0

is the trivial solution (x1 = x2 = · · · = xp = 0).

ˆ If the vector equation x1 v1 + x2 v2 + · · · xp vp = 0 has a nontrivial solution,

the set {v1 , v2 , · · · , vp } is said to be linearly dependent.

ˆ If (c1 , · · · , cp ) is a nontrivial solution to x1 v1 + · · · xp vp = 0, then the

relation c1 v1 + · · · cp vp = 0 is called linear dependence relation among
v1 , · · · , vp .

Warning. From the definition, the trivial combination 0v1 + · · · + 0vp = 0,

where every coefficient is 0, is not a dependence relation on v1 , · · · , vp .
152 Chapter 2. Vectors in Rn and Linear Systems

 
1
(a) We want to determine whether the set {v1 , v2 }, where v1 = and v2 =
  −1
2
, is linearly independent. Let x1 and x2 be scalars (real numbers) such
3
that x1 v1 + x2 v2 = 0. Then
     
1 2 0
x1 + x2 =
−1 3 0
   
x1 + 2x2 0
This is equivalent to = , which gives the following system
−x1 + 3x2 0
of linear equations
x1 + 2x2 = 0
−x1 + 3x2 = 0
To solve this, we need to reduce the augmented matrix as usual.
   
1 2 0 1 2 0
R2 + R1 → R2
−1 3 0 0 5 0
 
1 1 2 0
R → R2
5 2 0 1 0
 
1 0 0
R1 − 2R2 → R1
0 1 0

The corresponding system is

x1 = 0
x2 = 0

So the system has only the trivial solution (0, 0). This means that the set {v1 , v2 }
is linearly independent.

(b) Let x1 and x2 be scalars such that x1 v1 + x2 v2 = 0. Then

     
4 −12 0
x1 + x2 =
−5 15 0

This gives the system

4x1 − 12x2 = 0
−5x1 + 15x2 = 0
To solve this we need to reduce the augmented matrix.
1
→ R1
   
4 −12 0 4 1
R 1 −3 0
−5 15 0 − 51 R2 → R2 1 −3 0
 
1 −3 0
R2 − R1 → R2
0 0 0
2.4. Linear Independence 153

The corresponding system is x1 − 3x2 = 0 and 0 = 0. The variable x2 is free,

and x1 is a leading variable. Set x2 = s. Then the general solution is
x1 = 3s
x2 = s
where s can be any real number. Since there are infinitely many solutions, it
follows that the trivial solution is not the only solution. (For example, (3, 1)
is a nontrivial solution. This corresponds to s = 1.) Thus, the set {v1 , v2 } is
linearly dependent.

We now want to find a specific linear dependence relation among v1 and v2 . Any
(nonzero) choice for s results in a choice of a dependence relation for v1 and
v2 . For example, if s = 1, then x1 = 3 and x2 = 1. This gives the relation
3v1 + v2 = 0.

Alternate Solution.
By inspection, we see that v2 is a multiple of v1 (v2 = −3v1 ). This provides
a specific relation among v1 and v2 , and shows that the set {v1 , v2 } is
linearly dependent.

(c) Let x1 , x2 , and x3 be scalars such that x1 v1 + x2 v2 + x3 v3 = 0. Then

       
0 1 2 0
x1 −1 + x2 0 + x3 −3 = 0 
      
1 1 5 0
This gives the system
x2 + 2x3 = 0
−x1 −3x3 = 0
x1 + x2 + 5x3 = 0
To solve this, we need to reduce the augmented matrix.
   
0 1 2 0 1 1 5 0
 −1 0 −3 0  R1 ↔ R3  −1 0 −3 0 
1 1 5 0 0 1 2 0
 
1 1 5 0
R2 + R1 → R2  0 1 2 0 
0 1 2 0
 
1 1 5 0
R3 − R2 → R3  0 1 2 0 
0 0 0 0
 
1 0 3 0
R1 − R2 → R1  0 1 2 0 
0 0 0 0
154 Chapter 2. Vectors in Rn and Linear Systems

The corresponding system is

x1 + 3x3 = 0
x2 + 2x3 = 0

The variables x1 and x2 are leading variables, while x3 is a free variable. Set
x3 = s. Then the general solution is

x1 = −3s
x2 = −2s
x3 = s

where s can be any real number. If s = 1, then x1 = −3, x2 = −2, and

x3 = 1. And this gives a nontrivial solution. So the set {v1 , v2 , v3 } is linearly
dependent.
To get a dependence relation , take for example s = 1. Then x1 = −3, x2 = −2,
and x3 = 1. This gives the relation −3v1 − 2v2 + v3 = 0.

(d) Let x1 , x2 , and x3 be scalars such that x1 v1 + x2 v2 + x3 v3 = 0. Then

       
2 4 0 0
x1 3 + x2
   0 + x3 7 = 0 
   
0 −5 6 0

This gives the system

2x1 + 4x2 = 0
3x1 + 7x3 = 0
−5x2 + 6x3 = 0

To solve this, we need to reduce the augmented matrix.

   
2 4 0 0 1 2 0 0
1
 3 0 7 0  R
2 1
→ R1  3 0 7 0 
0 −5 6 0 0 −5 6 0
 
1 2 0 0
R2 − 3R1 → R2  0 −6 7 0 
0 −5 6 0
 
1 2 0 0
R2 − R3 → R2  0 −1 1 0 
0 −5 6 0
 
1 2 0 0
(−1)R2 → R2  0 1 −1 0 
0 −5 6 0
 
1 2 0 0
R3 + 5R2 → R3  0 1 −1 0 
0 0 1 0
2.4. Linear Independence 155

 
1 2 0 0
R2 + R3 → R2  0 1 0 0 
0 0 1 0
 
1 0 0 0
R1 − 2R2 → R1  0 1 0 0 
0 0 1 0

The corresponding system is x1 = 0, x2 = 0, and x3 = 0. So the system has only

the trivial solution. This implies that the set {v1 , v2 , v3 } is linearly indepen-
dent.

2. [Exercise on page 20]

(a) False. For example, consider the following matrix.

 
1 0 1
A=
0 1 1

This matrix has a pivot in every row, but its columns are linearly dependent
since the third column is the sum of the first two columns.
(b) True. This is true because of the following theorem.

Linear Independence and Pivot Columns

Theorem 2.4.1. Let A be an m × n matrix, and let B be an echelon form
of A. Then the columns of A are linearly independent if and only if every
column of B is a pivot column.
Proof. Every column of B is a pivot column if and only if the homogeneous
equation Bx = 0 has only the trivial solution. (This is essentially because
every column of B is a pivot column if and only if every variable in the
the system Bx = 0 is a leading variable.) Since B is an echelon form of
A, the system Ax = 0 has the same solution set as the system Bx = 0.
So every column of B is a pivot column if and only if the homogeneous
equation Ax = 0 has only the trivial solution, that is, the columns of A
are linearly independent.
Thanks to this theorem, to determine whether the columns of a matrix A
are linearly independent, one can proceed as follows.

Step 1. Transform A into an echelon matrix, say B.

Step 2. If every column (not row!) of B is a pivot column, then

the columns of A are linearly independent. Otherwise, the
columns of A are linearly dependent.

3. [Exercise on page 20]

156 Chapter 2. Vectors in Rn and Linear Systems

(a) We want to determine whether the columns of the following matrix are linearly
independent.  
−3 4 27
A =  2 −1 −13 
4 7 1
First, we need to put A into an echelon form.

   
−3 4 27 1 11 28
 2 −1 −13  R1 + R3 → R1  2 −1 −13 
4 7 1 4 7 1
 
1 11 28
R2 − 2R1 → R2  0 −23 −69 
R3 − 4R1 → R3
0 −37 −111
 
1 11 28
1
− 23 R2 → R2  0 1 3 
0 −37 −111
 
1 11 28
R3 + 37R2 → R3  0 1 3 =B
0 0 0
ˆ The upper left entry of B is a leading entry. So the first column is a pivot
column.
ˆ The entry in the second row and second column is also a leading entry. So
the second column is also a pivot column.
ˆ The third column of B is not a pivot column because it does not contain a
leading entry.
Since there is a column of B which is not a pivot column, it follows that the
columns of A are not linearly independent (so they are linearly dependent)
thanks to Theorem 2.4.1.

(b)
   
1 3 −4 1 3 −4
 −1 −2 R2 + R1 → R2
9   0 1 5 
  R3 − R1 → R3  
 1 2 −8   0 −1 −4 
R4 − 3R1 → R4
3 11 −1 0 2 11
 
1 3 −4
R3 + R2 → R3  0 1
 5 

R4 − 2R2 → R4  0 0 1 
0 0 1
 
1 3 −4
 0 1 5 
R4 − R3 → R4 
 0 0

1 
0 0 0
2.4. Linear Independence 157

Since every column of this latter matrix is a pivot column, it follows that the
columns of A are linearly independent (thanks to Theorem 2.4.1).

4. [Exercise on page 21]

(a) We want to find h (if any) such that v3 = x1 v1 + x2 v2 for some scalars x1 and
x2 . This amounts to finding h such that the following system has at least one
solution.
x1 = 1
2x1 + x2 = h
3x1 + x2 = 4
We reduce the augmented matrix:
   
1 0 1 1 0 1
 2 1 h  R2 − 2R1 → R2  0 1 h−2 
R3 − 3R1 → R3
3 1 4 0 1 1
 
1 0 1
R3 − R2 → R3  0 1 h−2 
0 0 −h + 3

The system has at least one solution if and only if −h + 3 = 0. Thus, v3 is in

Span{v1 , v2 } if and only if h = 3.
 
(b) We need to reduce the matrix A = v1 v2 v3 whose columns are v1 , v2 , and
v3 . Using the same operations as in part (a), an echelon form of A is
 
1 0 1
 0 1 h−2 
0 0 −h + 3

Every column of this latter matrix is a pivot column if and only if −h + 3 6= 0,

that is, if and only if h 6= 3. So the set {v1 , v2 , v3 } is linearly independent if and
only if h 6= 3.

5. [Exercise on page 21]

 
As before, we need to reduce the matrix A = v1 v2 v3 whose columns are v1 , v2 ,
and v3 .
   
−1 2 1 1 −2 −1
 2 −4 1  (−1)R1 → R1  2 −4 1 
−3 6 h −3 6 h
 
1 −2 −1
R2 − 2R1 → R2  0 0 3 
R3 + 3R1 → R3
0 0 h−3
 
1 −2 −1
1
R → R2
3 2
 0 0 1 
0 0 h−3
158 Chapter 2. Vectors in Rn and Linear Systems

 
1 −2 −1
R3 − (h − 3)R2 → R3  0 0 1 
0 0 0
Since there is a non-pivot column (the second column of this latter matrix is not a
pivot column) for every h, it follows that the set {v1 , v2 , v3 } is linearly dependent
for every h.
6. [Exercise on page 21]
(a) False. This is false because of the following result.

Set of One Vector

Theorem 2.4.2. Let n ≥ 1 be an integer. A set {v} of one vector in Rn
is linearly independent if and only if v 6= 0.
Proof. Let v be a vector in Rn . By definition, the set {v} is linearly
independent if and only if the equation x1 v = 0 has only the trivial solution
x1 = 0, that is, if and only if v 6= 0. (v = 0 if and only if the equation
x1 v = 0 has nontrivial solutions (for example x1 = 5).)

(b) True. This is true because of the following theorem.

Set of Two Vectors

Theorem 2.4.3. Let n ≥ 1 be an integer. A set of two vectors in Rn is
linearly dependent if and only if one vector in the set is a multiple of the
other.
Proof. Let S = {v1 , v2 } be a set of two vectors in Rn .
ˆ Suppose that the set {v1 , v2 } is linearly dependent. Then there exist
scalars x1 , x2 such that x1 v1 +x2 v2 = 0 and at least one of the scalars
is a nonzero number. Two possibilities:
– If x1 6= 0, then the equation x1 v1 + x2 v2 = 0 becomes v1 =
− xx12 v2 . And therefore v1 is a multiple of v2 .
– If x2 6= 0, then the equation x1 v1 + x2 v2 = 0 becomes v2 =
− xx12 v1 . And therefore v2 is a multiple of v1 .
This proves that if S = {v1 , v2 } is linearly dependent, then one of
the vectors in S is a multiple of the other.
ˆ Conversely, suppose that one of the vectors in {v1 , v2 } is a multiple
of the other. Without loss of generality, suppose that v2 is a multiple
of v1 . Then there exists a real number α such that v2 =  αv1 , that
α
is, αv1 − v2 = 0. This latter equation shows that x = is a
−1
nontrivial solution to the vector equation x1 v1 + x2 v2 = 0. So the
set {v1 , v2 } is linearly dependent.
2.4. Linear Independence 159

This ends the proof.

(c) True. This is true because of the following theorem.

Set of Two or More Vectors – Characterization of Linearly De-

pendent Sets
The following theorem is a generalization of Theorem 2.4.3.
Theorem 2.4.4. Let p ≥ 2 be an integer. A set S = {v1 , v2 , · · · , vp } of
vectors in Rn is linearly dependent if and only if at least one vector in S
can be written as a linear combination of the other vectors.
Warning. This theorem does not say that if a set S of vectors is linearly
dependent, then every vector in S can be written as a linear combination
of the others. An illustration of this is provided in part (d) and also in
Exercise 7.

(d) False. For example, consider the set {v1 , v2 , v3 } where

     
1 0 1
v1 = , v2 = , v3 =
0 1 0

ˆ This set is linearly dependent because there is a non-pivot column in the

matrix  
1 0 1
0 1 0
(The third column is not a pivot column.)
ˆ But the vector v2 cannot be written as a linear combination of v1 and v3 .
Indeed, since v1 = v3 and since v1 lies along the x-axis in R2 , it follows that
any linear combination of v1 and v3 lies in the x-axis. Since the vector v2 is
not in the x-axis (see figure below), it is not possible to write it as a linear
combination of v1 and v3 .
y
3

v2
v1
−1 v3 3 x

−1

Note. According to Theorem 2.4.4, at least one vector in {v1 , v2 , v3 } is a linear

combination of the others. This is true because v3 is a linear combination of v1
and v2 as v3 = 1v1 + 0v2 .
160 Chapter 2. Vectors in Rn and Linear Systems

7. [Exercise on page 21]

 
(a) Let A = v1 v2 v3 v4 be the matrix whose columns are v1 , v2 , v3 , and v4 . We
need to reduce A.

   
1 −2 1 −8 1 −2 1 −8
 −1 R2 + R1 → R2
3 −2 10   0 1 −1 2 
R3 + 2R1 → R3
−2 1 2 10 0 −3 4 −6
 
1 −2 1 −8
R3 + 3R2 → R3  0 1 −1 2 
0 0 1 0
 
1 −2 0 −8
R2 + R3 → R2  0 1 0 2 
R1 − R3 → R1
0 0 1 0
 
1 0 0 −4
R1 + 2R2 → R1  0 1 0 2 
0 0 1 0

Since there is a non-pivot column (the fourth column is not a pivot column), it
follows that the columns of A are linearly dependent.
(b) To answer this question, we need to find dependence relationships among v1 , v2 , v3 ,
and v4 . From part (a), the reduced echelon form of the augmented matrix of the
homogeneous equation Ax = 0 is
 
1 0 0 −4 0
 0 1 0 2 0 
0 0 1 0 0

The corresponding linear system is

x1 −4x4 = 0
x2 + 2x4 = 0
x3 = 0

The variables x1 , x2 , x3 are leading variables and x4 is free. Set x4 = t. Then

the general solution is
x1 = 4t
x2 = −2t
x3 = 0
x4 = t
where t can be any real number. So every dependence relation among v1 , v2 , v3 , v4
has the form
4tv1 − 2tv2 + 0v3 + tv4 = 0
where t is a nonzero real number. Since the coefficient of the vector v3 is 0, it is
not possible to express it as a linear combination of v1 , v2 , and v4 .
2.4. Linear Independence 161

8. [Exercise on page 21]

(a) True. This is true because of the following result.

Sets Containing the Zero Vector

Theorem 2.4.5. Let n ≥ 1 be an integer, and let S be a set of vectors in
Rn . If S contains the zero vector, then S is linearly dependent.
Proof. Let S = {v1 , v2 , · · · , vp } be a set of vectors in Rn . Suppose that one
of  vj , in S is the zero vector, and consider the matrix A =
 the vectors, say
v1 v2 · · · vp whose columns are v1 , v2 , · · · , vp . Then every echelon
form of A has a non-pivot column (the zero column vj is a non-pivot
column). This implies that the set {v1 , v2 , · · · , vp } is linearly dependent
(thanks to Theorem 2.4.1).

  
1 2
(b) False. Indeed, the set {v1 , v2 }, where v1 = and v2 = , is linearly
0 0
dependent because v2 is a multiple of v1 (v2 = 2v1 ), but it does not contain the
zero vector. This example shows that the converse of Theorem 2.4.5 is not true.
(c) True. This is true because of the following theorem.

Sets Containing More Vectors Than the Dimension

We know that Rn is the set of all vectors of the form
 
x1
 x2 
 
 .. 
 . 
xn

where x1 , x2 , · · · , xn are real numbers. So every vector in Rn has n entries.

The integer n is referred to as the dimension of Rn . For example, the
dimension of R2 is 2, the dimension of R3 is 3, etc.
Theorem 2.4.6. Let n ≥ 1 be an integer. Let {v1 , · · · , vp } be a set of
vectors in Rn . If p > n, then the set {v1 , · · · , vp } is linearly dependent.
In other words, this theorem says that if the number of vectors in a set S
is greater than the number of entries of each vector in S, then S is linearly
dependent.
 
Proof. Suppose p > n. Let A = v1 · · · vp be the matrix whose
columns are v1 , · · · , vp . Consider the homogeneous equation Ax = 0.
ˆ Since Ax = 0 is consistent (as the zero vector x = 0 is a solution),
and

ˆ since the system Ax = 0 has at least one free variable (because the
number of variables, p, is greater than the number of equations, n),
162 Chapter 2. Vectors in Rn and Linear Systems

it follows (by Theorem 1.3.2) that Ax = 0 has infinitely many solutions.

So the trivial solution is not the only solution, which implies that the
columns of A, v1 , · · · , vp , are linearly dependent.

 
0
(d) Solution. False. Indeed, the set {v1 }, where v1 = is the zero vector in
0
R2 , is linearly dependent (thanks to Theorem 2.4.5), but p = 1 is not greater
than n = 2. This example shows that the converse of Theorem 2.4.6 is not true.

9. [Exercise on page 22]

(a) Since v 6= 0, it follows that S = {v} is linearly independent thanks to Theo-

rem 2.4.2.
(b) The set S is linearly dependent because it contains the zero vector.
(c) The set S has p = 4 vectors, and each vector has n = 3 entries. Since p > n, it
follows that S is linearly dependent thanks to Theorem 2.4.6.
(d) Since the set S contains the zero vector, it follows (by Theorem 2.4.5) that S is
linearly dependent.
(e) It seems like v2 is −2v1 . This is true for the first three entries, but not for the
fourth one. So v2 is not a multiple of v1 , and therefore S is linearly indepen-
dent.

10. [Exercise on page 22]

4 v2

v1

v4
−2 4 x

−2 v3

(a) The set {v1 } is linearly independent because v1 is a nonzero vector.

(b) Geometrically, we see that v2 is a multiple of v1 . So the set {v1 , v2 } is linearly
dependent.
(c) Since the vectors v1 and v3 do not lie along the same line, they are linearly
independent.
(d) The set {v1 , v3 , v4 } contains p = 3 vectors and each vector is in Rn with n = 2.
Since p > n, it follows (by Theorem 2.4.6) that {v1 , v3 , v4 } is linearly depen-
dent.
2.4. Linear Independence 163

(e) Geometrically, we see that v4 is not a multiple of v2 . So the set {v2 , v4 } is

linearly independent.
(f ) Since the set {v2 , v4 , 0} contains the zero vector, it follows (by Theorem 2.4.5)
that it is linearly dependent.

11. [Exercise on page 22]

(a) False. Indeed, consider the following matrix.

 
1 0
A= 0 1 
0 0

The bottom row of A is a zero row, but the columns of A are not linearly
dependent because each of them is a pivot column.
(b) True. Indeed, let {v1 , v2 , · · · , vp } be the columns of A. Since the rightmost
column is a zero column, it follows that vp is the zero vector. So the set
{v1 , v2 , · · · , vp } is linearly dependent.

12. [Exercise on page 23]

(a) By definition of the product of a matrix by a vector, we have

 
  5
Ax = a1 a2 a3  −2  = 5a1 − 2a2 + a3
1

Since Ax = 0, it follows that 5a1 − 2a2 + a3 = 0, which is a linear dependence

relation on the columns of A.
(b) The relation 5a1 − 2a2 + a3 = 0 implies that a3 = −5a1 + 2a2 . Since a3 is a linear
combination of a1 and a2 , it follows (by Theorem 2.1.1) that

Span{a1 , a2 , a3 } = Span{a1 , a2 }

But Span{a1 , a2 } cannot be equal to R3 because two vectors are not enough to
span the 3-dimensional space R3 . Thus, the columns of A do not span R3 .

13. [Exercise on page 23]

The relation a3 = a1 + 2a2is equivalent
 to a1 + 2a2 − a3 = 0, which in turn is
  1 1
equivalent to a1 a2 a3  2  = 0. So A  2  = 0, and this shows that the
  −1 −1
1
vector  2  is a nontrivial solution to Ax = 0.
−1
164 Chapter 2. Vectors in Rn and Linear Systems

14. [Exercise on page 23]

All possible reduced echelon forms of a 2 × 2 matrix include the following.
       
1 0 1 ∗ 0 1 0 0
, , , and
0 1 0 0 0 0 0 0

Only the first matrix does not have the required property. So all possible reduced
echelon form of a 2 × 2 matrix whose columns are linearly dependent are
     
1 ∗ 0 1 0 0
, , and .
0 0 0 0 0 0

15. [Exercise on page 23]

(a) True. By definition, if the equation Ax = 0 has a nontrivial solution, then the
columns of A are linearly dependent.
(b) False. Indeed, consider the matrix
 
1 0 0
A=
0 1 0

The columns of A span R2 since A has a pivot in every row. But the columns
of A are not linearly independent because one of the columns is the zero vector
(or because the number of columns is greater than the number of rows).
(c) False. Indeed, consider the following 3 × 2 matrix.
 
1 0
A= 0 1 
0 0

The columns of A are linearly independent since each of them is a pivot column.
But they do not span R3 because there is a row (the third row) that does not
contain a pivot (or because two vectors are not enough to span R3 ).
(d) False. Indeed, consider the following.
   
1 0 0
A= 0 1  b= 0 
0 0 1

As we saw in part (c), the columns of A are linearly independent. But the linear
system Ax = b has no solution since the third equation is of the form 0 = d with
d = 1 6= 0.

16. [Exercise on page 23]

Let u, v be vectors in Rm such that the set {u, v} is linearly independent. We want
to prove that the set {u + v, u − v} is also linearly independent. Let x1 , x2 be scalars
such that
x1 (u + v) + x2 (u − v) = 0
2.4. Linear Independence 165

Then

x1 u + x1 v + x2 u − x2 v = 0
(x1 + x2 )u + (x1 − x2 )v = 0

Since u and v are linearly independent, the latter equation implies that x1 + x2 = 0
and x1 − x2 = 0. This gives the following system.

x 1 + x2 = 0
x 1 − x2 = 0

To solve this, we need to reduce the augmented matrix.

   
1 1 0 1 1 0
R2 − R1 → R2
1 −1 0 0 −2 0
 
1 1 0
− 12 R2 → R2
0 1 0
 
1 0 0
R1 − R2 → R1
0 1 0

So x1 = 0 and x2 = 0. This implies that the equation x1 (u + v) + x2 (u − v) = 0 has

only the trivial solution, which means that the vectors u + v and u − v are linearly
independent.
166 Chapter 2. Vectors in Rn and Linear Systems
 Chapter3

Linear Transformations

3.1 Introduction to Linear Transformations

1. [Exercise on page 25]

Definition of a Linear Transformation

ˆ A function T : Rn → Rm , from Rn to Rm , is a rule that associates each
input in Rn to exactly one output in Rm . If the input is x, the output is
denoted T (x).

– The set Rn is called the domain of T .

– The set Rm is called the codomain of T .

Examples of functions include the following.

– T : R → R, T (x) = x2
 
2 3x
– T : R → R , T (x) =
4x + 5
 
  −x2
x1
– T : R2 → R3 , T =  x1 − x2 
x2
−3x2 + 7x3
ˆ A linear transformation is a function T : Rn → Rm that satisfies the
following two conditions:

(i) For all x, y in Rn ,

T (x + y) = T (x) + T (y)

(ii) For all x in Rn , for every scalar c,

T (cx) = cT (x)

167
168 Chapter 3. Linear Transformations

We want to show that the transformation T : R2 → R2 defined by

   
x1 4x1 − x2
T =
x2 3x1 + 5x2
   
x1 y1
is linear. Let x = and y = be vectors in R2 , and let c be a scalar. We
x2 y2
want to show that T (x + y) = T (x) + T (y) and T (cx) = cT (x).

ˆ Proving that T (x + y) = T (x) + T (y).

     
x1 y1 x1 + y1
T (x + y) = T + =T
x2 y2 x2 + y2
 
4(x1 + y1 ) − (x2 + y2 )
= By definition of T
3(x1 + y1 ) + 5(x2 + y2 )
 
4x1 + 4y1 − x2 − y2
=
3x1 + 3y1 + 5x2 + 5y2
 
(4x1 − x2 ) + (4y1 − y2 )
=
(3x1 + 5x2 ) + (3y1 + 5y2 )
   
4x1 − x2 4y1 − y2
= +
3x1 + 5x2 3y1 + 5y2
= T (x) + T (y) By definition of T

ˆ Proving that T (cx) = cT (x).

    
x1 cx1
T (cx) = T c =T
x2 cx2
 
4(cx1 ) − (cx2 )
= By definition of T
3(cx1 ) + 5(cx2 )
 
c(4x1 − x2 )
= Factor out c
c(3x1 + 5x2 )
 
4x1 − x2
=c
3x1 + 5x2
= cT (x) By definition of T

So T is a linear transformation.
2. [Exercise on page 25]
   
x1 y1
ˆ Condition (i). Let x = ,y = be vectors in R2 . Then
x2 y2
     
x1 y1 x1 + y 1
T (x + y) = T + =T
x2 y2 x2 + y 2
3.1. Introduction to Linear Transformations 169

   
3(x1 + y1 ) 3x1 + 3y1
= 0 = 0 
(x1 + y1 ) − (x2 + y2 ) x1 + y 1 − x2 − y 2
 
3x1 + 3y1
= 0+0 
(x1 − x2 ) + (y1 − y2 )
   
3x1 3y1
= 0 + 0  = T (x) + T (y)
x1 − x2 y1 − y2

Thus, condition (i) is satisfied.

 
x1
ˆ Condition (ii). Let c be a real number, and let x = be a vector in R2 .
x2
Then
 
     3(cx1 )
x1 cx1
T (cx) = T c =T = 0 
x2 cx2
cx1 − cx2
   
c(3x1 ) 3x1
= 0  = c 0  = cT (x)
c(x1 − x2 ) x1 − x2
So condition (ii) is satisfied.

Since the two conditions of the definition are met, T is a linear transformation.
3. [Exercise on page 25]
(a) True. This is true because of the following theorem.

Alternate Definition of a Linear Transformation

Theorem 3.1.1. A function T : Rn → Rm is a linear transformation if
and only if for every vectors x, y in Rn and for every scalars c, k, we have
T (cx + ky) = cT (x) + kT (y) (3.1.1)
Proof. Let T : Rn → Rm be a function.
ˆ Suppose that T : Rn → Rm is a linear transformation. We want to
show that the equation (3.1.1) holds for every vectors x, y in Rn and
for every scalars c and k. Let x, y be vectors in Rn . And let c, k be
scalars. Then
T (cx + cy) = T (cx) + T (cy) Since T is linear
= cT (x) + cT (y) Since T is linear

ˆ Conversely, suppose that the equation (3.1.1) holds for every vectors
x, y in Rn and for every scalars c and k. We want to show that T is
linear.
170 Chapter 3. Linear Transformations

– Let x, y be vectors in Rn . Then

T (x + y) = T (1x + 1y) = 1T (x) + 1T (y) By (3.1.1)

= T (x) + T (y)

– Let x be a vector in Rn and let c be a scalar. Then

Where y = 0 is the
T (cx) = T (cx + 0y)
zero vector in Rn
= cT (x) + 0T (y) By (3.1.1)
= cT (x) + 0 = cT (x)

So T is a linear transformation.

This ends the proof.

(b) True. This is true because of the following result.

The Image of the Zero Vector Under a Linear Transformation

Theorem 3.1.2. If T : Rn → Rm is a linear transformation, then

T (0) = 0

Proof.

T (0) = T (0 + 0) 0=0+0
T (0) = T (0) + T (0) Since T is linear
T (0) = 2T (0)
T (0) − 2T (0) = 0 Subtract 2T (0) from both sides
−T (0) = 0
T (0) = 0 Multiply both sides by −1

(c) False. Indeed, consider the transformation T : R → R defined by T (x) = x2 .

ˆ Clearly, T (0) = 0.
ˆ But T is not linear because condition (i) (as well as condition (ii)) of the
definition is not satisfied. Indeed,

T (1 + 1) 6= T (1) + T (1)

as T (1 + 1) = T (2) = 22 = 4 and T (1) + T (1) = 12 + 12 = 2.

(d) True. This is true because of the following fact.
3.1. Introduction to Linear Transformations 171

If T (0) 6= 0, Then T is Not Linear

Theorem 3.1.3. Let T : Rn → Rm be a function. If T (0) 6= 0, then T is

not linear.
Proof. Suppose that T (0) 6= 0. If T is linear, then (by Theorem 3.1.2)
the image of 0 under T will be 0, and this will contradict the fact that
T (0) 6= 0. So T is not linear.
This theorem suggests that if we want to check whether a transformation
T is linear, we can first find T (0) and see if it is zero or not.

ˆ If T (0) 6= 0, then T is not linear.

ˆ If T (0) = 0, then T may or may not be linear.

4. [Exercise on page 25]

(a) To show that T is not linear, we need to find a specific counterexample to one
of the
 conditions
 ofthe definition of a linear transformation. Take for example
1 1
x= and y = . Then
1 1
   
1 1
  h
2 p i h√ i
3 3
T (x + y) = T + =T = 2(2) = 4
1 1 2
On the other hand, we have
h√ i h√ i
3 3
T (x) + T (y) = 1 + 1 = [1 + 1] = [2]
√
Since 3 4 6= 2, it follows that T (x + y) 6= T (x) + T (y). This shows that T is not
linear.
Note. Of course there are infinitely many counterexamples:
         
2 2 1 3
x= and y = or x = and y = , etc.
2 2 3 2
 
1
(b) Let c = −2 and x = . Then
1
    
1 −2
T (cx) = T (−2x) = T −2 =T
1 −2
     
−2 −2 −2
=  −(−2) − 2  =  0 = 0 
5| − 2| 5(2) 10
On the other hand, we have
     
  1 1 −2
1
−2T (x) = −2T = −2  −1 + 1  = −2  0  =  0 
1
5|1| 5 −10
172 Chapter 3. Linear Transformations

Since T (−2x) 6= −2T (x), it follows that T is not linear.

(c)        
0 9(0) + 8 8 0
T (0) = T = = 6=
0 0 0 0
Since T (0) 6= 0, it follows that T is not linear by Theorem 3.1.3.

5. [Exercise on page 26]

The transformation
 
x1  
7x1 − x2
T  x2  =
−8x1 + x2 + ex3
x3

is not linear because T (0) 6= 0. Indeed,

 
0  
7(0) − (0)
T  0  =
−8(0) + 0 + e0
0
 
0
= e0 = 1
0+1
   
0 0
= 6=
1 0

6. [Exercise on page 26]

(a)

T (4u) = 4T (u) Since T is linear

   
3 12
=4 =
4 16

(b)

T (u + v) = T (u) + T (v) Since T is linear

       
3 5 3+5 8
= + = =
4 −2 4 + (−2) 2

(c)

T (7u − 6v) = 7T (u) − 6T (v) Since T is linear

       
3 5 21 −30
=7 −6 = +
4 −2 28 12
 
−9
=
40
3.1. Introduction to Linear Transformations 173

7. [Exercise on page 26]

T (2u + 7v) = 2T (u) + 7T (v) Since T is linear

T (2u + 7v) − 7T (v) = 2T (u) Subtract 7T (v) from both sides
1
(T (2u + 7v) − 7T (v)) = T (u) Divide both sides by 2
2
       
1 −4 2 −4 2
−7 = T (u) T (2u + 7v) = and T (v) =
2 3 −5 3 −5
 
1 −18
= T (u)
2 38
 
−9
So T (u) = .
19

8. [Exercise on page 26]

Since w is in Span{u, v}, there exist scalars c1 and c2 such that w = c1 u + c2 v.
Applying T to both sides, we get

T (w) =T (c1 u + c2 v)
=T (c1 u) + T (c2 v) Because T is linear
=c1 T (u) + c2 T (v) Again because T is linear

This shows that T (w) is a linear combination of T (u) and T (v). In other words, T (w)
is in Span{T (u), T (v)}.

9. [Exercise on page 26]

(a) We need to show that the equation c1 u1 +c2 u2 = v hasa unique

 solution.
 This
−3 2 32
amounts to showing that the equation c1  2  + c2  −1  =  −19  has
8 7 1
a unique solution. This latter equation gives the system

−3c1 + 2c2 = 32
2c1 − c2 = −19
8c1 + 7c2 = 1

To solve this, we need to reduce the augmented matrix:

   
−3 2 32 −1 1 13
 2 −1 −19  R1 + R2 → R1  2 −1 −19 
8 7 1 8 7 1
 
1 −1 −13
(−1)R1 → R1  2 −1 −19 
8 7 1
174 Chapter 3. Linear Transformations

 
1 −1 −13
R2 − 2R1 → R2  0 1 7 
R3 − 8R1 → R3
0 15 105
 
1 −1 −13
R3 − 15R2 → R3  0 1 7 
0 0 0
 
1 0 −6
R1 + R2 → R1  0 1 7 
0 0 0
So the system has a unique solution: c1 = −6 and c2 = 7.
(b) Find T (v).
Solution. From part (a), v = −6u1 + 7u2 . Taking T of both sides, we get
T (v) = T (−6u1 + 7u2 )
= −6T (u1 ) + 7T (u2 ) Because T is linear
= −6u01 + 7u02 Because T (u1 ) = u01 and T (u2 ) = u02
   
5 −3
= −6  −1  + 7  −2 
2 15
   
−30 − 21 −51
= 6 − 14  =  −8 
−12 + 105 93
10. [Exercise on page 26]
(a) The sentence “T is a linear transformation that maps e1 into v1 and e2 into v2 ”
means that T (e1 ) = v1 and T (e2 ) = v2 . So we know what the transformation
does on e1 and e2 .
   
12 12
ˆ To find T , we first need to write as a linear combina-
−13 −13
tion of e1 and e2 . We have
         
12 12 0 1 0
= + = 12 − 13 = 12e1 − 13e2
−13 0 −13 0 1
ˆ Now we have
 
12
T = T (12e1 − 13e2 ) = 12T (e1 ) − 13T (e2 )
−13
       
−1 3 −12 − 39 −51
= 12 − 13 = =
2 4 24 − 52 −28
(b)
      
x1 1 0
T (x) = T =T x1 + x2 = T (x1 e1 + x2 e2 )
x2 0 1
     
−1 3 −x1 + 3x2
= x1 T (e1 ) + x2 T (e2 ) = x1 + x2 =
2 4 2x1 + 4x2
3.1. Introduction to Linear Transformations 175

11. [Exercise on page 27]

 
b1
(a) To find T (b) for an arbitrary b = b2 , we first need to write b as a linear

b3
combination of the vectors
     
1 1 −1
a1 =  −1  , a2 =  0  , and a3 =  3  .
3 4 0
 
To do this, we need to reduce the augmented matrix a1 a2 a3 b .
   
1 1 −1 b1 1 1 −1 b1
 −1 0 R2 + R1 → R2
3 b2   0 1 2 b2 + b1 
R3 − 3R1 → R3
3 4 0 b3 0 1 3 b3 − 3b1
 
1 1 −1 b1
R3 − R2 → R3  0 1 2 b2 + b1 
0 0 1 −4b1 − b2 + b3
 
1 1 0 −3b1 − b2 + b3
R2 − 2R3 → R2  0 1 0 9b1 + 3b2 − 2b3 
R1 + R3 → R1
0 0 1 −4b1 − b2 + b3
 
1 0 0 −12b1 − 4b2 + 3b3
R1 − R2 → R1  0 1 0 9b1 + 3b2 − 2b3 
0 0 1 −4b1 − b2 + b3
So the solution to the equation x1 a1 + x2 a2 + x3 a3 = b is
x1 = −12b1 − 4b2 + 3b3
x2 = 9b1 + 3b2 − 2b3
x3 = −4b1 − b2 + b3
Therefore
b = (−12b1 − 4b2 + 3b3 )a1 + (9b1 + 3b2 − 2b3 )a2 + (−4b1 − b2 + b3 )a3
Using the fact that T is a linear transformation, we find
T (b) = (−12b1 − 4b2 + 3b3 )T (a1 ) + (9b1 + 3b2 − 2b3 )T (a2 ) + (−4b1 − b2 + b3 )T (a3 )
     
0 −4 −2
= (−12b1 − 4b2 + 3b3 )  1  + (9b1 + 3b2 − 2b3 )  1  + (−4b1 − b2 + b3 )  1 
2 −6 5
 
0 − 4(9b1 + 3b2 − 2b3 ) − 2(−4b1 − b2 + b3 )
=  (−12b1 − 4b2 + 3b3 ) + (9b1 + 3b2 − 2b3 ) + (−4b1 − b2 + b3 ) 
2(−12b1 − 4b2 + 3b3 ) + (9b1 + 3b2 − 2b3 ) + 5(−4b1 − b2 + b3 )
 
−28b1 − 10b2 + 6b3
=  −7b1 − 2b2 + 2b3 
−98b1 − 31b2 + 23b3
176 Chapter 3. Linear Transformations

 
b1
(b) Here b =  b2  with b1 = −6, b2 = −7, and b3 = 13. Using the formula we
b3
obtained in part (a), we have
 
−28b1 − 10b2 + 6b3
T (b) =  −7b1 − 2b2 + 2b3 
−98b1 − 31b2 + 23b3
   
−28(−6) − 10(−7) + 6(13) 316
= −7(−6) − 2(−7) + 2(13)  =  82 
−98(−6) − 31(−7) + 23(13) 1104

12. [Exercise on page 27]

(a) False. Indeed, let T : R2 → R2 be the linear transformation defined by

   
x1 x1
T =
x2 0
   
1 0
Let u = and v = .
0 1
ˆ The vectors u and v are linearly independent because v is not a multiple of
u.
ˆ But T (u) and T (v) are not linearly independent. Indeed,
       
1 1 0 0
T (u) = T = and T (v) = T =
0 0 1 0
Since T (v) is the zero vector, the set {T (u), T (v)} is linearly dependent.
(b) False. Indeed, consider the linear transformation T and the vectors u and v
from part (a).
 
1 0
ˆ The vectors u and v span R because the matrix
2
(whose columns are
0 1
u and v) have a pivot in every row.
 
1 0
ˆ But the set {T (u), T (v)} does not span R since the matrix
2
(whose
0 0
columns are T (u) and T (v)) does not have a pivot in every row as the there
is no pivot in the second row.
(c) True. The fact that {T (u), T (v)} is linearly dependent means that there exist
scalars c1 and c2 (with c1 6= 0 or c2 6= 0) such that c1 T (u) + c2 T (v) = 0. Since
T is linear, this latter equality is equivalent to T (c1 u + c2 v) = 0. So the vector
c1 u + c2 v is a solution to the equation T (x) = 0.
Now, we need to show that c1 u + c2 v is a nonzero vector. We will proceed by
contradiction. Suppose c1 u + c2 v = 0. Then c1 = 0 and c2 = 0 because the set
{u, v} is linearly independent by hypothesis. But this contradicts the fact that
the scalars c1 and c2 are not all zero. Thus c1 u + c2 v 6= 0, and therefore the
equation T (x) = 0 has a nontrivial solution.
(d) True. This is true because of the following theorem.
3.1. Introduction to Linear Transformations 177

Any Matrix Transformation is Linear

Theorem 3.1.4. Let A be an m × n matrix. Let T : Rn → Rm be the
transformation defined by
T (x) = Ax
Then T is a linear transformation.
Proof. Let x, y be vectors in Rn and let c be a scalar.

ˆ Using the property A(x + y) = Ax + Ay, we have

T (x + y) = A(x + y) = Ax + Ay = T (x) + T (y)

ˆ Using the property A(cx) = cA(x), we have

T (cx) = A(cx) = cA(x) = cT (x)

So T is a linear transformation.
The transformation T defined by T (x) = Ax is called a matrix trans-
formation.

(e) False. Indeed, the product of A by a vector x is defined if the number of columns
of A equals the number of entries of x. In that case, the number of entries of Ax
corresponds to the number of rows of A. So, if A has 7 rows and 11 columns,
the product Ax is defined if the vector x has 11 entries, that is, x is a vector in
R11 . Thus, if A is of size 7 × 11 and T : Ra → Rb is defined by T (x) = Ax, then
a = 11 and b = 7. In general, we have the following.

Matrix Transformation – Dimensions of the Domain and

Codomain
Let A be an m × n matrix (that is, a matrix with m rows and n columns),
and let T : Ra → Rb be the linear transformation defined by T (x) = Ax.

ˆ The number of rows of A corresponds to the dimension of

the codomain of T , that is, m = b.

ˆ The number of columns of A corresponds to the dimension

of the domain of T , that is, n = a.

In other words, the domain of T is Rnumber of columns of A and the codomain

is Rnumber of rows of A :

T : Rnumber of columns of A → Rnumber of rows of A

(f ) False. Indeed,
ˆ since the number of rows of A corresponds to the the dimension of the
178 Chapter 3. Linear Transformations

codomain of T , it follows that p = 9.

ˆ Since the number of columns of A corresponds to the the dimension of the
domain of T , it follows that q = 6.

13. [Exercise on page 27]

(a)
   
    
1 −3 1(2) − 3(6)
2 2 − 18 −16
T (u) = Au = = = =
5 −17 5(2) − 17(6)
6 10 − 102 −92
      
1 −3 −7 1(−7) − 3(8) −31
T (v) = Av = = =
5 −17 8 5(−7) − 17(8) −171

(b) To find a vector x in R2 such that T (x) = b, we need to solve the system Ax = b.
   
1 −3 4 1 −3 4
R2 − 5R1 → R2
5 −17 12 0 −2 −8
 
1 1 −3 4
− 2 R2 → R2
0 1 4
 
1 0 16
R1 + 3R2 → R1
0 1 4
 
16
So the system Ax = b has a unique solution: x = .
4

14. [Exercise on page 28]

(a) By definition,
   
    1 −2   x1 − 2x2
x1 x1 x1
T (x) = T =A = 3 1  =  3x1 + x2 
x2 x2 x2
1 7 x1 + 7x2

(b) Using the formula from part (a), we get

   
  −1 − 2(2) −5
−1
T (u) = T =  3(−1) + 2  =  −1 
2
−1 + 7(2) 13

(c) We need to solve theequation

 T(x) = b for x. This amounts to solving the
x1 − 2x2 7
equation  3x1 + x2  =  7 . This gives the system
x1 + 7x2 −11

x1 − 2x2 = 7
3x1 + x2 = 7
x1 + 7x2 = −11
3.1. Introduction to Linear Transformations 179

As usual we need to reduce the augmented matrix.

   
1 −2 7 1 −2 7
R2 − 3R1 → R2
 3 1 7   0 7 −14 
R3 − R1 → R3
1 7 −11 0 9 −18
 
1
R → R2 1 −2 7
7 2
1
 0 1 −2 
R
9 3
→ R 3 0 1 −2
 
1 −2 7
R3 − R2 → R3  0 1 −2 
0 0 0
 
1 0 3
R1 + 2R2 → R1  0 1 −2 
0 0 0
 
3
The solution is x1 = 3 and x2 = −2. So the image of the vector under
  −2
7
T is the vector b =  7 .
−11
(d) From part (c), the equation T (x) = b has a unique solution. So there is only
one x whose image under T is b.

15. [Exercise on page 28]

(a) To find a vector x whose image under T is y, we need to solve the equation
T (x) = y, that is, the equation Ax = y.
   
1 −1 1 1 1 −1 1 1
 −2 R2 + 2R1 → R2
2 −2 1   0 0 0 3 
R3 − 3R1 → R3
3 −4 −1 1 0 −1 −4 −2
 
1 −1 1 1
R2 ↔ R3  0 −1 −4 −2 
0 0 0 3

The third equation is of the form 0 = d with d = 3 6= 0. So the equation

T (x) = y has no solution. This means that there is no vector x whose image
under T is y.
 
x1
(b) To find a vector x =  x2  whose image under T is y 0 , we need to solve the
x3
equation T (x) = y , that is, the equation Ax = y 0 .
0

   
1 −1 1 1 1 −1 1 1
 −2 R2 + 2R1 → R2
2 −2 −2   0 0 0 0 
R3 − 3R1 → R3
3 −4 −1 4 0 −1 −4 1
180 Chapter 3. Linear Transformations

 
1 −1 1 1
R2 ↔ R3  0 −1 −4 1 
0 0 0 0
 
1 −1 1 1
−R2 → R2  0 1 4 −1 
0 0 0 0
 
1 0 5 0
R1 + R2 → R1  0 1 4 −1 
0 0 0 0

The corresponding linear system is

x1 + 5x3 = 0
x2 + 4x3 = −1

Let x3 = s. Then the general solution is

x1 = −5s
x2 = −1 − 4s
x3 = s

where s can be any real

 number.
 If s = 1 (you can choose any
 value
 for s), we
−5 −5
have the solution x = −5 . So the image of the vector −5  under T is
  
1 1
0
y.
(c) Yes because from part (b), the equation T (x) = y 0 has infinitely many solutions.
To find three vectors whose image under T is y 0 , all we have to do is to choose
three values for s.
   
−5s 0
ˆ If s = 0, we have the vector x =  −1 − 4s  =  −1 .
s 0
   
−5s −5
ˆ If s = 1, we have the vector x =  −1 − 4s  =  −5 .
s 1
   
−5s −10
ˆ If s = 2, we have the vector x =  −1 − 4s  =  −9 .
s 2
 
b1
(d) To describe the set of all vectors b = b2  for which the equation T (x) = b

b3
has a solution, we need to solve the equation Ax = b.
   
1 −1 1 b1 1 −1 1 b1
 −2 R2 + 2R1 → R2
2 −2 b2   0 0 0 b2 + 2b1 
R3 − 3R1 → R3
3 −4 −1 b3 0 −1 −4 b3 − 3b1
3.1. Introduction to Linear Transformations 181

 
1 −1 1 b1
R2 ↔ R3  0 −1 −4 b3 − 3b1 
0 0 0 b2 + 2b1
 
1 −1 1 b1
−R2 → R2  0 1 4 −b3 + 3b1 
0 0 0 b2 + 2b1
 
1 0 5 4b1 − b3
R1 + R2 → R1  0 1 4 3b1 − b3 
0 0 0 b2 + 2b1
So the equation T (x) = b is consistent if and only if
b2 + 2b1 = 0
Solving this latter equation for b2 , we get b2 = −2b1 . So
       
b1 b1 b1 0
b =  b2  =  −2b1  =  −2b1  +  0 
b3 b3 0 b3
   
1 0
= b1 −2 + b3 0 
  
0 1
 
b1
So the set of all vectors b =  b2  for which the equation T (x) = b has a
b3
3
solution is the plane in R generated by the vectors
   
1 0
 −2  and  0 
0 1
 
x1
(e) To describe the set of all x =  x2  for which T (x) = 0, we need to solve the
x3
equation Ax = 0. Using the same elementary operations as in the part (d), the
reduced echelon form of the augmented matrix is
 
1 0 5 0
 0 1 4 0 
0 0 0 0
Let x3 = s. Then the general solution to the equation Ax = 0 is
     
x1 −5s −5
x =  x2  =  −4s  = s  −4 
x3 s 1
where s can be any real number. So the set
 of all
 vectors x such that T (x) = 0
−5
3
is the line in R generated by the vector −4 .

1
182 Chapter 3. Linear Transformations

3.2 The Matrix of a Linear Transformation

1. [Exercise on page 28]
 
x1
(a) Let x = be an arbitrary vector in R2 . Then
x2
        
4 0 x1 4x1 + 0x2 4x1 x1
T (x) = Ax = = = =4 = 4x
0 4 x2 0x1 + 4x2 4x2 x2
So, geometrically, the transformation T multiplies every vector by 4. Such trans-
formation is called a dilation.

Dilation and Contraction

Let r be a real number, and let T : Rn → Rn be the transformation defined
by T (x) = rx. Then T is a linear transformation. Indeed, let x, y be
vectors in Rn and let c be a scalar. Then

T (x + y) = r(x + y) = rx + ry = T (x) + T (y)

T (cx) = r(cx) = c(rx) = cT (x)

ˆ If r > 1, T is called a dilation.

ˆ If 0 ≤ r ≤ 1, T is called a contraction.

The effect of the transformation T (x) = 4x on some vectors is shown below.

x2

T (v) T (u)

v u
T (x) T (w)
x w x1

 
x1
(b) Let x = be a vector in R2 . Then
x2
        
x1 −1 0 x1 −x1 + 0x2 −x1
T (x) = T = Ax = = =
x2 0 1 x2 0x1 + 1x2 x2
   
x1 −x1
So the transformation T maps to . Geometrically, this means
x2 x2
that T is the reflection through the x2 -axis. The effect of T on some vectors is
shown below.
3.2. The Matrix of a Linear Transformation 183

x2

w T (w)

T (v)
v

T (u) u x1

T (x) x

 
x1
(c) Let x = be a vector in R2 . Then
x2

      
0 0 x1 0x1 + 0x2 0
T (x) = Ax = = =
0 1 x2 0x1 + 1x2 x2

   
x1 0
So the transformation T maps to . Geometrically, this is the
x2 x2
projection onto the x2 -axis. The effect of T on some vectors is shown below.

x2

v
T (v)

T (u) u

x1

w T (w)

 
x1
(d) Let x = be a vector in R2 . Then
x2

      
0 1 x1 0x1 + 1x2 x2
T (x) = Ax = = =
1 0 x2 1x1 + 0x2 x1

   
x1 x2
So T is maps to . Geometrically, this is the reflection through
x2 x1
the line x2 = x1 . The effect of T on some vectors is shown below.
184 Chapter 3. Linear Transformations

x2
T (w) x2 = x1

v
w
T (u)
u T (v) x1
T (x) x

2. [Exercise on page 29]

Sketching the Graph of the Image of the Unit Square Under a Linear
Transformation
   
u1 v1
Let u = and v = be two vectors in R2 . The line L connecting
u2 v2
u and v is the line through the points (u1 , u2 ) and (v1 , v2 ) as shown in the
following figure.
x2

v L
v2
u
u2

u1 v1 x1

If A = (x1 , x2 ) and B = (y1 , y2 ) are two points in the plane, the vector from
A to B, denoted AB, ~ is defined by
 
~ y 1 − x1
AB =
y 2 − x2
Theorem 3.2.1. We have the following.
ˆ The set of points on the line L is the same as the set of points

(1 − t)u + tv, where t is a real number

ˆ The set of points on the line segment connecting u and v is the same as
the set of points

(1 − t)u + tv, where 0 ≤ t ≤ 1

3.2. The Matrix of a Linear Transformation 185

Proof. Let M = (x1 , x2 ) be a point on the line L. Let A = (u1 , u2 ) be the point
corresponding to the vector u, and let B = (v1 , v2 ) be the point corresponding
to the vector v. (See figure below.)
x2

M
v
v2
B
u
u2
A

u1 v1 x1

~ is parallel to the vector AB.

Then the vector AM ~ This means that there exists
~ = tAB,
a number t such that AM ~ that is,

x1 − u1 = t(v1 − u1 )
x2 − u2 = t(v2 − u2 )
ˆ From the equality x1 − u1 = t(v1 − u1 ), it follows that

x1 = tv1 − tu1 + u1 = (1 − t)u1 + tv1

ˆ From the equality x2 − u2 = t(v2 − u2 ), it follows that

x2 = tv2 − tu2 + u2 = (1 − t)u2 + tv2

So
       
x1 (1 − t)u1 + tv1 u1 v1
= = (1 − t) +t = (1 − t)u + tv
x2 (1 − t)u2 + tv2 u2 v2
Clearly, the point M = (x1 , x2 ) is between A and B if and only if 0 ≤ t ≤ 1.

Thanks to Theorem 3.2.1, we can prove the following result, which says that a
linear transformation maps a line to a line.
Theorem 3.2.2. Let T : R2 → R2 be a linear transformation. Let u and v be
two vectors in R2 , and let L be the line segment connecting u and v. Then the
image of L under T , T (L), is the line segment connecting T (u) and T (v).
 
x1
Proof. Let x = be a vector in R2 such that the corresponding point
x2
(x1 , x2 ) is on the line segment connecting u and v. Then, by Theorem 3.2.1,
there exists a number 0 ≤ t ≤ 1 such that
x = (1 − t)u + tv
186 Chapter 3. Linear Transformations

Applying T to this, we get

T (x) = T ((1 − t)u + tv)

= (1 − t)T (u) + tT (v) Since T is linear

So the point corresponding to the vector T (x) is on the line segment connecting
the vectors T (u) and T (v). This ends the proof.
Thanks to Theorem 3.2.2, to find the image of the unit square under a linear
transformation, all we have to do is to find the image of each corner.


3 0
(a) T (x) = Ax where A =
0 3
Let S be the
 unit
square
as shown
  below.
To find the
imageS under T , we need
0 1 1 0
to find T ,T ,T , and T .
0 0 1 1
     
0 0 0
T =A =
0 0 0
        
1 1 3 0 1 3
T =A = =
0 0 0 3 0 0
        
1 1 3 0 1 3
T =A = =
1 1 0 3 1 3
        
0 0 3 0 0 0
T =A = =
1 1 0 3 1 3
So T (S) is the square whose corners are (0, 0), (3, 0), (3, 3), and (0, 3) as shown
below.
T (S)
3 3

2 2
S T
1 1

1 2 3 1 2 3
 
−2 0
(b) T (x) = Ax where A =
0 3
     
0 0 0
T =A =
0 0 0
        
1 1 −2 0 1 −2
T =A = =
0 0 0 3 0 0
3.2. The Matrix of a Linear Transformation 187

        
1 −21 −2 0 1
T =A = =
1 3
1 0 3 1
        
0 0 −2 0 0 0
T =A = =
1 1 0 3 1 3

So T (S) is the rectangle whose corners are (0, 0), (−2, 0), (−2, 3), and (0, 3) as
shown below.
T (S)
3 3

2 2
S T
1 1

1 2 3 −2 −1
 
1 2
(c) T (x) = Ax where A =
0 1

     
0 0 0
T =A =
0 0 0
        
1 1 1 2 1 1
T =A = =
0 0 0 1 0 0

        
1 1 1 2 1 3
T =A = =
1 1 0 1 1 1
        
0 0 1 2 0 2
T =A = =
1 1 0 1 1 1

So T (S) is the parallelogram whose corners are (0, 0), (1, 0), (3, 1), and (2, 1) as
shown below.
3 3

2 2
S T T (S)
1 1

1 2 3 1 2 3
This transformation is called a shear transformation.

Shear Transformations
Let A be an n × n matrix (a square matrix). Let In be the identity matrix.
188 Chapter 3. Linear Transformations

For example,
 
  1 0 0 0
  1 0 0
1 0  0 1 0 0 
I1 = [1], I2 = , I3 = 0 1 0  ,
 I4 =  
0 1  0 0 1 0 
0 0 1
0 0 0 1

ˆ A shear matrix is a matrix obtained from In by performing an

elementary operation of the form Ri + aRj → Ri . For example,
shear matrices of size 2 × 2 are of the form
   
1 a 1 0
or
0 1 a 1

where a is a real number.

ˆ If A is a shear matrix, then the linear transformation T : Rn → Rn

defined by T (x) = Ax is called a shear transformation.

 
1 0
(d) T (x) = Ax where A =
−3 1
     
0 0 0
T =A =
0 0 0
        
1 1 1 0 1 1
T =A = =
0 0 −3 1 0 −3

        
1 1 1 1 0 1
T =A = =
1 −2 1 −3 1 1
        
0 0 1 0 0 0
T =A = =
1 1 −3 1 1 1

So T (S) is the parallelogram whose corners are (0, 0), (1, −3), (1, −2), and (0, 1)
as shown below.
1
3

2 1 2 3
S T −1 T (S)
1
−2
1 2 3 −3
 
−2 −1
(e) T (x) = Ax where A =
1 2
3.2. The Matrix of a Linear Transformation 189

     
0 0 0
T =A =
0 0 0
        
1 1 −2 −1 1 −2
T =A = =
0 0 1 2 0 1
        
1 1 −2 −1 1 −3
T =A = =
1 1 1 2 1 3
        
0 0 −2 −1 0 −1
T =A = =
1 1 1 2 1 2
So T (S) is the parallelogram whose corners are (0, 0), (−2, 1), (−3, 3), and (−1, 2)
as shown below.

3 3
T (S)
2 2
S T
1 1

1 2 3 −3 −2 −1 1
" π

π

#
cos 4
− sin 4
(f ) T (x) = Ax where A = π

π


sin 4
cos 4
     
0 0 0
T =A =
0 0 0
  "
# "
# " √ #
π π

π 2
cos − sin
  
1 1 4 4 1 cos 4

T =A = = = √2
0 0 π

π


0 π 2
sin 4
cos 4
sin 4 2

  " π

π

#
cos − sin
    
T
1
=A
1
= 4 4 1
= √0
1 1 π π 1 2



sin 4
cos 4
  "
# " √ #
π π

− 2
cos − sin
  
0 0 4 4 0 2
T =A = = √
1 1 π π 1


2
sin 4
cos 4 2

The image of S under T is shown below.

2 2
S T T (S)
1

1 2 3 −2 −1 1 2
190 Chapter 3. Linear Transformations

π
Here T is the counterclockwise rotation through 4
about the origin.

3. [Exercise on page 29]

Standard Matrix of a Linear Transformation

For n ≥ 1, 1 ≤ i ≤ n, define ej to be the vector in Rn whose  every entry is 0
0 if i 6= j
except the jth entry which is 1. That is, the ith entry of ej is
1 if i = j.
For example,

ˆ when n = 1, e1 = 1 .
 

   
1 0
ˆ For n = 2, e1 = and e2 = .
0 1
     
1 0 0
ˆ For n = 3, e1 = 0 , e2 = 1 , and e3 = 0 .
    
0 0 1
ˆ Etc.

Theorem 3.2.3. Let T : Rn → Rm be a linear transformation. Then there

exists a unique matrix A such that

T (x) = Ax for every x in Rn

 
x1
 x2 
Proof. Let x =  ..  be a vector in Rn . Then
 
 . 
xn
     
x1 0 0
 0   x2   0 
x= +  + ··· + 
     
.. .. .. 
 .   .   . 
0 0 xn
     
1 0 0
 0   1   0 
= x1   + x2   + · · · + xn 
     
.. .. .. 
 .   .   . 
0 0 1
= x1 e1 + x2 e2 + · · · + xn en

So

T (x) = T (x1 e1 + x2 e2 + · · · + xn en )
= x1 T (e1 ) + x2 T (e2 ) + · · · + xn T (en ) Since T is linear
3.2. The Matrix of a Linear Transformation 191

 
x1
  x2  By definition of the
= T (e1 ) T (e2 ) · · · T (en )   product of a matrix by
 
..
.
a vector
 
xn
= Ax Where A = [T (e1 ) T (e2 ) · · · T (en )]

Uniqueness. Let B be another matrix such that T (x) = Bx for every x in Rn .

We want to prove that B = A. Let b1 , b2 , · · · , bn be the columns of B.

ˆ Multiplying B by e1 , we get
 
1
  0 
Be1 = b1 b2 · · · bn  = 1b1 + 0b2 + · · · + 0bn = b1
 
 ..
 . 
0

So Be1 = b1 . Since T (e1 ) = Be1 , it follows that b1 = T (e1 ).

ˆ Likewise, b2 = T (e2 ), · · · , bn = T (en ).

This proves that B = A.

From the proof of Theorem 3.2.3, it follows that A is the m × n matrix whose
jth column is the vector T (ej ):
 
A = T (e1 ) T (e2 ) · · · T (en )

The matrix A is referred to as the standard matrix of T .

 
 
3x1
(a) For the linear transformation T : R → R2 defined by T x1 = ,
    −5x1
n = 1 and e1 = 1 . So the standard matrix is A = T (e1 ) . But
   
 
3(1) 3
T (e1 ) = T 1 = = .
−5(1) −5
 
3
Thus, the standard matrix of T is A = .
−5
(b) Since the domain of this is R2 , we have that n = 2 and
   
1 0
e1 = and e2 =
0 1

To get the standard matrix, we need to find T (e1 ) and T (e2 ).

     
1 7(1) − 0 7
T (e1 ) = T = =
0 3(0) 0
192 Chapter 3. Linear Transformations

     
0 7(0) − 1 −1
T (e2 ) = T = =
1 3(1) 3

So the standard matrix is

 
  7 −1
A= T (e1 ) T (e2 ) =
0 3
     
1 0 0
(c) Here n = 3, e1 =  0  , e2 =  1  , and e3 =  0 . To get the standard
0 0 1
matrix, we need to find T (e1 ), T (e2 ), and T (e3 ).
 
1    
0 − 4(0) 0
T (e1 ) = T   0   = =
−1 + 6(0) − 9(0) −1
0

0    
1 − 4(0) 1
T (e2 ) = T  1  = =
−0 + 6(1) − 9(0) 6
0

0    
0 − 4(1) −4
T (e3 ) = T  0  = =
−0 + 6(0) − 9(1) −9
1
So the standard matrix is
 
  0 1 −4
A= T (e1 ) T (e2 ) T (e3 ) =
−1 6 −9

(d) We need to find T (e1 ) and T (e2 ).

   
  5(1) 5
1
T (e1 ) = T =  0  = 0 

0 1−0 1
3 3

   
  5(0) 0
0
T (e2 ) = T = 0 = 0 
1 0−1
3
− 31
 
5 0
Thus, the standard matrix is A = 0
 0 
1
3
− 13
(e) We need to find T (e1 ), T (e2 ), and T (e3 ), where
     
1 0 0
e1 = 0 , e2 = 1 , and e3 = 0 
    
0 0 1
3.2. The Matrix of a Linear Transformation 193

     
1 1 − 3(0) + 6(0) 1
T (e1 ) = T  0  =  −1 + 4(0) − 7(0)  =  −1 
0 2(1) + 5(0) + 0 2
     
0 0 − 3(1) + 6(0) −3
T (e2 ) = T  1  =  −(0) + 4(1) − 7(0)  =  4 
0 2(0) + 5(1) + 0 5
     
0 0 − 3(0) + 6(1) 6
T (e3 ) = T  0  =  −(0) + 4(0) − 7(1)  =  −7 
1 2(0) + 5(0) + 1 1

So the standard matrix of T is

 
1 −3 6
A =  −1 4 −7 
2 5 1

4. [Exercise on page 29]

x2

T (e2 )

e2
e1
−2 5 x1

T (e1 )
−2

   
4 1
(a) From the figure, we see that T (e1 ) = and T (e2 ) = . So the standard
−1 3
matrix of T is  
4 1
A=
−1 3
     
1 1 0
(b) First, observe that u = = + = e1 + e2 . So, by using the fact
1 0 1
that T is linear, we find that
     
4 1 5
T (u) = T (e1 + e2 ) = T (e1 ) + T (e2 ) = + =
−1 3 2
194 Chapter 3. Linear Transformations

Alternate Solution
 
1
The image of the vector u = under T is
1
    
4 1 1 5
T (u) = Au = =
−1 3 1 2

The graph of T (u) is shown below.

x2

T (e2 )

T (u)
e2
e1
−2 5 x1

T (e1 )
−2

5. [Exercise on page 30]

x2

T (e1 )
e2
T (e2 )

−2 e1 5 x1

−2
   
−1 1
(a) From the figure, we see that T (e1 ) = and T (e2 ) = . So the standard
2 1
matrix of T is  
−1 1
A=
2 1
(b) First, observe that
         
−2 −2 0 1 0
u= = + = −2 +3 = −2e1 + 3e2
3 0 3 0 1
So, by using the fact that T is linear, we get

T (u) = T (−2e1 + 3e2 ) = −2T (e1 ) + 3T (e2 )

3.2. The Matrix of a Linear Transformation 195

     
−1 1 5
= −2 +3 =
2 1 −1

The graph of T (u) is shown below.

x2
u 3T (e2 )

T (e1 ) e2
T (e2 )
e1 x1
T (u)

−2T (e1 )

6. [Exercise on page 30]

(a) T : R2 → R2 is the reflection through the x1 -axis.

Solution. To get the standard matrix,
  all we have to do is to find T (e1 ) and
1 0
T (e2 ), where e1 = and e2 = .
0 1
ˆ Since the reflection through the x1 -axis maps every point
 x on the x1 -axis
1
into the same point x, it follows that T (e1 ) = e1 = .
0
ˆ The reflection through
  the  maps e2 into −e2 . This implies that
 x1 -axis
0 0
T (e2 ) = −e2 = − = . (See the figure below.)
1 −1
x2
2

e2
T (e1 )
−2 e1 2 x1
T (e2 )

−2

Thus, the standard matrix of T is

 
  1 0
A= T (e1 ) T (e2 ) =
0 −1

(b) We need to find T (e1 ) and T (e2 ).

196 Chapter 3. Linear Transformations

ˆ If we rotate e1through

π
2
(counterclockwise) about the origin, we get e2 . So
0
T (e1 ) = e2 = as shown in the following figure.
1
x2
2

e2 T (e1 )

−2 T (e2 ) e1 2 x1

−2

ˆ If we rotate e2 through

π
2
(counterclockwise) about the origin, we get −e1 .
−1
So T (e2 ) = −e1 = .
0
Thus, the standard matrix is
 
  0 −1
A= T (e1 ) T (e2 ) =
1 0

   
1 0
(c) We need to find T (e1 ), T (e2 ), and T (e3 ), where e1 =  0  , e2 =  1  , and
  0 0
0
e3 =  0 . The reflection through the x1 x2 -plane maps every point in the
1
x1 x2 -plane into
 the same point. Since e1 and  e2 lie in the x1 x2 -plane, we have
1 0
T (e1 ) = e1 =  0  and T (e2 ) = e2 =  1  .
0 0
 
0
On the other hand, T (e3 ) = −e3 =  0 . Thus, the matrix of T is
−1

 
  1 0 0
A = T (e1 ) T (e2 ) T (e3 ) =  0 1 0 
0 0 −1

   
1 0
(d) We need to find T (e1 ) and T (e2 ), where e1 = and e2 = . The images
0 1
of e1 and e2 under T are shown in the following figure.
3.2. The Matrix of a Linear Transformation 197

x2

e2

T (e2 ) e1 x1

T (e1 ) x2 = −x1

From the figure, we find that

   
0 −1
T (e1 ) = −e2 = and T (e2 ) = −e1 =
−1 0
Thus, the standard matrix of T is
 
  0 −1
A = T (e1 ) T (e2 ) =
−1 0
(e) Remember that the projection onto the x1 -axis is the transformation T : R2 →
R2 defined by    
x1 x1
T =
x2 0
So        
1 1 0 0
T (e1 ) = T = and T (e2 ) = T =
0 0 1 0
 
1 0
Thus, the standard matrix of T is A = . (By the way, the projection
0 0  
2 2 x1
onto the x2 -axis is the transformation T : R → R defined by T =
  x 2
0
.)
x2
(f ) Solution. The images of e1 and e2 under T , the clockwise rotation through π3
about the origin, are shown in the following figure.
x2

e2
T (e2 )
π
3
π
3
π
3 e1 x1

π
3

T (e1 )
198 Chapter 3. Linear Transformations

From the figure, we have

" # " 1 # " # " √ #
cos π3 2√ sin π3 2
3
T (e1 ) = = and T (e2 ) = =
− sin π3 − 23 cos π3 1
2

Thus, the standard matrix of T is

" √ #
1 3
2 2
 
A= T (e1 ) T (e2 ) = √
3 1
− 2 2

In general, we have the following theorem.

Matrix of a Rotation
Theorem 3.2.4. We have the following two statements.

ˆ The matrix of the transformation T : R2 → R2 that rotates each point

in R2 about the origin through an angle θ > 0 (counterclockwise)
is  
cos θ − sin θ
A=
sin θ cos θ

ˆ The matrix of the transformation R : R2 → R2 that rotates each

point in R2 about the origin through an angle ϕ > 0 (clockwise) is
 
cos ϕ sin ϕ
A=
− sin ϕ cos ϕ

(g) As usual, we need to find T (e1 ) and T (e2 ).

ˆ If we reflect e1 through the x1 -axis, we still get e1 . Now if we rotate e1
0
through π2 (clockwise), we get −e2 . So T (e1 ) = −e2 = .
−1
ˆ If we reflect e2 through the x1 -axis, we get −e2 . Now if we rotate −e2
−1
through π2 (clockwise), we get −e1 . So T (e2 ) = −e1 = . Thus, the
  0
0 −1
matrix of T is A = .
−1 0
Note. This matrix is the same as the matrix we found in part (d). So the linear
transformation from (d) is the same as the linear transformation from (g).

7. [Exercise on page 31]

Since
" √ # " #
3
2
− 12 cos π6 − sin π6
√ = ,
1 3 sin π6 cos π6
2 2
it follows by the first part of Theorem 3.2.4 that T is the counterclockwise rotation
through π6 about the origin.
3.3. One-To-One and Onto Linear Transformations 199

3.3 One-To-One and Onto Linear Transformations

1. [Exercise on page 31]

One-To-One and Onto Linear Transformations – Definitions

Let T : Rn → Rm be a function (not necessarily a linear transformation).

ˆ T is one-to-one if for every b in the codomain Rm , there exists at most

one x in the domain Rn such that T (x) = b. (In other words, T is one-to-
one if for every b in Rm the equation T (x) = b has at most one solution.)

ˆ T is onto if for every b in the codomain Rm , there exists at least one x

in the domain Rn such that T (x) = b. (In other words, T is onto if for
every b in Rm the equation T (x) = b has at least one solution.)

Note. If T is linear, the term “at most one solution” means a unique solution
or no solution. “At least one solution” means a unique solution or infinitely
many solutions.
The term “one-to-one” is also referred to as “injective”. And the term “onto”
is also referred to as “surjective”.

(a) False. This is false by definition.

(b) False. This is false by definition.
(c) False. Indeed, consider the transformation T : R2 → R2 defined by
   
x1 x1
T =
x2 0
 
1
Let b = .
1
   
x1 1
ˆ The equation T (x) = b is equivalent to = , that is, x1 = 0 and
0 1
0 = 1. Since 0 = 1 is not true, the equation T (x) = b has no solution. So
the equation T (x) = b has at most one solution.
 
1
ˆ But T is not one-to-one because the equation T (x) = does not have
0
at most one solution as it has infinitely many solutions:
   
x1 1
T =
x2 0
   
x1 1
=
0 0
x1 = 0 and x2 is free

(d) False. Consider the same transformation as in part (c).

200 Chapter 3. Linear Transformations

 
1
ˆ From part (c), the equation T (x) = b has at least one solution for b = .
0
 
1
ˆ But T is not onto because, as we saw in part (c), the equation T (x) =
1
does not have at least one solution.

2. [Exercise on page 31]

One-To-One Linear Transformations – Some Useful Results

The following theorem gives an alternate definition of one-to-one.

Theorem 3.3.1 (alternate definition of one-to-one). A function T : Rn →

Rm is one-to-one if and only if for every u, v in Rn , T (u) = T (v) implies that
u = v.

Proof. ˆ Suppose that T is one-to-one and let u, v be elements in Rn such

that T (u) = T (v). We want to show that u = v. Let b = T (u) = T (v).
Then u and v are both solutions to the equation T (x) = b. Since T is
one-to-one, the equation T (x) = b has at most one solution. So it is not
possible to have two different solutions. This implies that u = v.

ˆ Conversely, suppose that for every u, v in Rn , T (u) = T (v) implies that

u = v. Let b be a vector in Rm . If u and v are solutions to the equation
T (x) = b, then T (u) = b = T (v), and this would imply that u = v. So
the equation T (x) = b has at most one solution, which means that T is
one-to-one.

Theorem 3.3.2. A linear transformation T : Rn → Rm is one-to-one if and

only if the equation T (x) = 0 has only the trivial solution.

Proof. ˆ Suppose that T is one-to-one. Since T (0) = 0, the equation T (x) =

0 becomes T (x) = T (0). This latter equation implies that x = 0 thanks
to Theorem 3.3.1. So the equation T (x) = 0 has only the trivial solution.

ˆ Conversely, suppose that the equation T (x) = 0 has only the trivial
solution and let u, v be vectors in Rn such that T (u) = T (v). Then
T (u) − T (v) = 0, that is, T (u − v) = 0 since T is linear. This latter
equation implies that u − v = 0 or u = v. So, by Theorem 3.3.1, T is
one-to-one.
This ends the proof.

Theorem 3.3.3. Let T : Rn → Rm be a linear transformation and let A be the

standard matrix of T .

(a) T is one-to-one if and only if the columns of A are linearly independent.

(b) If B is an echelon form of A, T is one-to-one if and only if every column

of B is a pivot column.
3.3. One-To-One and Onto Linear Transformations 201

(c) If T is one-to-one, then n ≤ m. (This basically means that if T is one-

to-one, then the domain of T is smaller than or equal to its codomain.)

(d) If n > m, then T is not one-to-one. (In other words, if the number of
columns of A is greater than the number of rows, then T is not one-to-
one.)

Proof. (a) Since A is the standard matrix of T , it follows that T (x) = Ax for
every x in Rn .

ˆ Suppose that T is one-to-one. Then, by Theorem 3.3.2, the homoge-

neous equation Ax = 0 has only the trivial solution. This implies (by
definition of linear independence) that the columns of A are linearly
independent.
ˆ Conversely, suppose that the columns of A are linearly independent.
Then, by definition, the equation Ax = 0 has only the trivial solu-
tion, which implies that T is one-to-one again by Theorem 3.3.2.

This proves (a).

(b) Let B be an echelon form of A.

ˆ From part (a), T is one-to-one if and only if the columns of A are

linearly independent.
ˆ Using Theorem 2.4.1, the columns of A are linearly independent if
and only if every column of B is a pivot column.

Putting these two points together, we get (b).

(c) Suppose that T is one-to-one. Then, from part (a), the columns of A
are linearly independent. This implies (by Theorem 2.4.6) that n (the
number of columns of A) is less than or equal to m (the number of rows
of A).

(d) This is the contrapositive of the statement from part (c).

(a) True. This is true because of Theorem 3.3.3 -(a).

(b) True. This is true because of Theorem 3.3.2.
(c) True. This is true because of Theorem 3.3.3 -(c).
(d) False. Indeed, let n = 2 and m = 3 and consider the linear transformation
T : Rn → Rm defined by T (x) = 0 for every x in Rn .
ˆ Clearly, n ≤ m.
ˆ But T is not one-to-one because the equation T (x) = 0 does not have only
thetrivial
solution (as the image of every vector under T is 0; for example
1
T = 0).
1
202 Chapter 3. Linear Transformations

(e) False. Because the dimension of the domain, n = 3, is greater than the dimen-
sion of the codomain, m = 2 (by Theorem 3.3.3 -(d), if n > m, then T is not
one-to-one).
(f ) True. This is true because of the following result.

Kernel of a Linear Transformation

Let T : Rn → Rm be a linear transformation. The kernel of T , denoted
ker T , is the set of all vectors x in Rn such that T (x) = 0.
Theorem 3.3.4. The linear transformation T is one-to-one if and only
if ker T = {0}.
Proof. This follows immediately from Theorem 3.3.2.

3. [Exercise on page 31]

Onto Linear Transformations – Some Useful Results

Theorem 3.3.5. Let T : Rn → Rm be a linear transformation, and let A be the
standard matrix of T .
(a) T is onto if and only if the columns of A span Rm .
(b) If B is an echelon form of A, T is onto if and only if every row of B
contains a pivot.
(c) If T is onto, then n ≥ m. (This basically means that if T is onto, then
the domain of T is bigger than or equal to its codomain.)
(d) If n < m, then T is not onto. (In other words, if the number of columns
of A is less than the number of rows, then T is not onto.)
Proof. (a) Since A is the standard matrix of T , we have T (x) = Ax for every
x in Rn .
ˆ Suppose that T is onto. Then for every b in Rm , the equation Ax = b
has at least one solution. This implies (by Theorem 2.2.2) that the
columns of A span Rm .
ˆ Conversely, suppose that the columns of A span Rm . Then, again by
Theorem 2.2.2, for every b in Rm the equation Ax = b has at least
one solution. This implies that T is onto.
(b) This follows immediately from part (a) and Theorem 2.2.2.
(c) Suppose that T is onto. Then, from part (a), the columns of A span Rm .
This implies (by Theorem 2.1.2) that n (the number of columns of A) is
greater than or equal to m (the number of rows of A).
(d) This is the contrapositive of the statement from part (c).
3.3. One-To-One and Onto Linear Transformations 203

(a) True. This is true because of Theorem 3.3.5 -(a).

(b) True. This is true because of Theorem 3.3.5 -(c).
(c) False. Indeed, let n = 3, m = 2 and let T : Rn → Rm be the linear transforma-
tion defined by T (x) = 0 for every x in Rn . Then the standard matrix of T is
the zero matrix:  
0 0 0
A=
0 0 0
ˆ Clearly, n ≥ m.
ˆ But T is not onto because the columns of A do not span R2 .
(d) False because the dimension of the domain, n = 2, is less than the dimension
of the codomain, m = 3.
(e) False because the number of columns of A, n = 5, is less than the number of
rows, m = 6.
(f ) True. Indeed, if T is both one-to-one and onto, then n ≤ m (by Theorem 3.3.3)
and n ≥ m (by Theorem 3.3.5). This implies that n = m.
(g) False. Indeed, let n = m = 2 and let T : Rn → Rm be the linear transformation
defined by the matrix  
0 0
A=
0 0
Clearly, n = m. But T is neither one-to-one (because the columns of A are not
linearly independent) nor onto (because the columns of A do not span R2 ).
(h) True. This is true because of the following result.

Range of a Linear Transformation

Let T : Rn → Rm be a linear transformation. The range (or image) of T

is the set of all vectors b in Rm such that the equation T (x) = b has at least
one solution. By definition, the range of T is a subset of the codomain
Rm .
Theorem 3.3.6. The linear transformation T is onto if and only if its
range is Rm .
Proof. This follows immediately from the definitions.

4. [Exercise on page 32]

 
b1
(a) Let b = be a vector in R2 . We need to solve the equation T (x) = b.
b2    
x1 − 2x2 − x3 b1
This amounts to solving the equation = . This latter
x1 + 3x2 b2
204 Chapter 3. Linear Transformations

equation gives the following system.

x1 − 2x2 − x3 = b1
x1 + 3x2 = b2

Reducing the augmented matrix, we have

   
1 −2 −1 b1 1 −2 −1 b1
R2 − R1 → R2
1 3 0 b2 0 5 1 b2 − b1
 
1 1 −2 −1 b1
R
5 2
→ R2 1 b2 b1
0 1 5 5
− 5
" #
3b1
1 0 − 53 5
+ 2b52
R1 + 2R2 → R1 b2
0 1 1
5 5
− b51

The variables x1 and x2 are leading variables and x3 is free. Because there is
a free variable and no equation of the form 0 = d with d 6= 0, the system has
infinitely many solutions, and this holds for every b in R2 . Therefore, according
to the definitions, the transformation T is onto, but not one-to-one.

Some Alternate Solutions

 
x1  
x1 − 2x2 − x3
The standard matrix of T  x2  = is
x1 + 3x2
x3
 
1 −2 −1
A=
1 3 0

ˆ A is of size m × n with m = 2 and n = 3. Since n > m, it follows

(by Theorem 3.3.3 -(d)) that T is not one-to-one.

ˆ Performing the operation R2 − R1 → R2 on A, we get an echelon

form:  
1 −2 −1
B=
0 5 1
Since there is a non-pivot column (as the third column is not a
pivot column), it follows that the columns of B are not linearly
independent. This implies (by Theorem 3.3.3 -(a)) that T is not
one-to-one.

ˆ Since every row of B contains a pivot, it follows that T is onto

thanks to Theorem 3.3.5 -(b).

(b) The standard matrix of the linear transformation T is

 
1 −3 6
A =  −1 4 −7 
2 5 1
3.3. One-To-One and Onto Linear Transformations 205

Row reducing A, we get

   
1 −3 6 1 −3 6
 −1 R2 + R1 → R2
4 −7   0 1 −1 
R3 − 2R1 → R3
2 5 1 0 11 −11
 
1 −3 6
R3 − 11R2 → R3  0 1 −1  =B
0 0 0

ˆ Since there is a non-pivot column (the third column of B is not a pivot

column), it follows (by Theorem 3.3.3 -(b)) that T is not one-to-one.
ˆ Since there is a row that does not contain a pivot (the third row of B), it
follows (by Theorem 3.3.5 -(b)) that T is not onto.
(c) The domain of T is Rn with n = 2 and the codomain is Rm with m = 3.
ˆ Since the dimension of the domain (n = 2) is less than the dimension of the
codomain (m = 3), it follows (by Theorem 3.3.5 -(d)) that T is not onto.
ˆ Reducing A, we get
   
1 4 1 4
 −3 −12  R2 + 3R1 → R2  0 0 
R3 − 2R1 → R3
2 7 0 −1
 
1 4
R2 ↔ R3  0 −1 
0 0

Since every column is a pivot column, it follows (by Theorem 3.3.3 -(b))
that T is one-to-one.
(d)
   
5 −20 10 1 −4 2
 2 −7 1
9  R
5 1
→ R1  2 −7 9 
−4 15 −6 −4 15 −6
 
1 −4 2
R2 − 2R1 → R2  0 1 5 
R3 + 4R1 → R3
0 −1 2
 
1 −4 2
R3 + R2 → R3  0 1 5  =B
0 0 7

ˆ Since every column of B is a pivot column, it follows (by Theorem 3.3.3 -(b))
that T is one-to-one.
ˆ Since every row of B contains a pivot, it follows that T is onto thanks to
Theorem 3.3.5 -(b).
206 Chapter 3. Linear Transformations

(e)
   
1 −2 4 5 0 1 −2 4 5 0
 0
 1 0 −1 2 
 R3 + R1 → R3  0
 1 0 −1 2 

 −1 4 −3 −4 4  R4 + 2R1 → R4  0 2 1 1 4 
−2 1 −9 −9 −6 0 −3 −1 1 −6
 
1 −2 4 5 0
R3 − 2R2 → R3  0
 1 0 −1 2 

R4 + 3R2 → R4  0 0 1 3 0 
0 0 −1 −2 0
 
1 −2 4 5 0
 0 1 0 −1 2 
R4 + R3 → R4 
 0
 =B
0 1 3 0 
0 0 0 1 0
ˆ Since there is a non-pivot column (the rightmost column is not a pivot
column), it follows that T is not one-to-one thanks to Theorem 3.3.3 -
(b).

Alternate Solution
Since the number of columns of A is greater than the number of rows,
it follows that T is not one-to-one thanks to Theorem 3.3.3 -(d).

ˆ Since every row of B has a pivot, it follows that T is onto thanks to

Theorem 3.3.5 -(b).

5. [Exercise on page 32]

(a) We need to solve the equation T (x) = u.

   
1 −2 −5 1 1 −2 −5 1
 −4 R2 + 4R1 → R2
9 26 1   0 1 6 5 
R3 − 5R1 → R3
5 −9 −19 1 0 1 6 −4
 
1 −2 −5 1
R3 − R2 → R3  0 1 6 5 
0 0 0 −9
 
1 0 7 11
R1 + 2R2 → R1  0 1 6 5 
0 0 0 −9
The third equation is 0 = −9, which is not true. So there is no vector x
whose image under T is u.
 
b1
(b) Let b =  b2  be a vector in the codomain R3 . To determine whether T is
b3
one-to-one, we need to solve the equation T (x) = b. The augmented matrix is
3.3. One-To-One and Onto Linear Transformations 207

 
1 −2 −5 b1
 −4 9 26 b2 . Performing the same operations as before, we get
5 −9 −19 b3
 
1 0 7 9b1 + 2b2
 0 1 6 b2 + 4b1  (3.3.1)
0 0 0 −9b1 − b2 + b3
This system does not have at most one solution for every b. For example, if
b1 = 0, b2 = 0, b3 = 0, the system has infinitely many solutions:
x1 = −7s
x2 = −6s
x3 = s
where s is a real number. So T is not one-to-one.
Finding
 three
 different vectors in the domain that have the same image. Let
0
b =  0 . Then, as we saw above, the equation T (x) = 0 has infinitely many
0
solutions.
 
0
ˆ For s = 0, we have the vector u1 = 0 .

0
 
−7
ˆ For s = 1, we have the vector u2 =  −6 .
1
 
−14
ˆ For s = 2, we have the vector u3 =  −12 .
2
The vectors u1 , u2 , u3 have the same image: T (u1 ) = T (u2 ) = T (u3 ) = 0.
(c) The transformation T is not onto since the equation  T(x) = b does
 not have at
b1 2
least one solution for every b. For example, for b =  b2  =  0 , the equation
b3 0
T (x) = b has no solution since −9b1 − b2 + b3 = −9(2) − 0 + 0 = −18 6= 0.
A vector b is not in the range of T is the equation T (x) = b has no solution.
From the augmented matrix (3.3.1), the equation T (x) = b has no solution if
and only if
−9b1 − b2 + b3 6= 0
ˆ For b1 = 1,
 b2 = 0, b3 = 0, we have −9b1 − b2 + b3 = −9 6= 0. So the vector
1
v1 =  0  is not in the range of T .
0
   
0 0
ˆ Likewise, the vectors v2 = 1 and v3 = 0  are not in the range of T .
  
0 1
208 Chapter 3. Linear Transformations

 
b1
(d) From (3.3.1), a vector b =  b2  is in the range of T if and only if
b3

−9b1 − b2 + b3 = 0

Solving this for b3 , we get b3 = 9b1 + b2 ( it is correct if the equation is solved

for b1 or for b2 ). So,
       
b1 b1 b1 0
b =  b2  =  b2  =  0  +  b2 
b3 9b1 + b2 9b1 b2
   
1 0
= b1 0 + b2 1 
  
9 1
   
1 0
So the range of T is the plane in R3 generated by the vectors  0  and  1 .
9 1
(e) By definition, thekernel
 ofT is the set of solutions to the equation T (x) = 0.

x1
From part (b), T  x2  = 0 if and only if x1 = −7s, x2 = −6s, and x3 = s
x3  
x1
where s is a real number. So x =  x2  is in the kernel of T if and only if x is
x3
of the form      
x1 −7s −7
x =  x2  =  −6s  = s  −6 
x3 s 1
 
−7
Thus, the kernel of T is the line in R3 generated by the vector  −6 .
1
 Chapter4

Matrix Algebra

4.1 Matrix Operations

1. [Exercise on page 35]

Size of a Matrix, Diagonal Entries, Diagonal Matrix, Identity Matrix,

Zero Matrix
ˆ A matrix is a rectangular array of numbers organized in rows and
columns.

ˆ A matrix is of size m × n if it has m rows and n columns.

ˆ If A is a matrix, the entry in the ith row and jth column of A is denoted
aij and is called the (i, j)-entry of A. Notation:
 
a11 a12 · · · a1n
 a21 a22 · · · a2n 
A =  ..
 
.. .. 
 . . . 
am1 am2 · · · amn

The compact notation for this is A = [aij ], where 1 ≤ i ≤ m and 1 ≤ j ≤

n, or simply A = [aij ].

ˆ A square matrix is a matrix whose number of rows equals the number

of columns.

ˆ If A = [aij ] is an m × n matrix, the diagonal entries of A are the entries

a11 , a22 , a33 , · · · . The diagonal entries form the main diagonal of A.

ˆ A matrix A of size m × n is called diagonal matrix if m = n (that is, it

is a square matrix) and all the nondiagonal entries are zero.

ˆ The n × n identity matrix, denoted In , is the n × n diagonal matrix

where every diagonal entry is 1.

ˆ The m × n zero matrix, denoted 0, is the m × n matrix where every

entry is 0.

209
210 Chapter 4. Matrix Algebra

 
2 4 9
(a) False. Indeed, the matrix A = has 2 rows and 3 columns. So it is of
3 1 8
size 2 × 3.
 
(b) True. Because the matrix 1 9 7 has 1 row and 3 columns.
 
2 4 9
(c) False. Indeed, the (1, 2)-entry of is the entry in the first row, second
3 1 8
column, that is, 4.
 
2 4 9
(d) True. The (2, 3)-entry of is the entry in the second row, third column,
3 1 8
that is, 8.
 
2 4 9
(e) True. Indeed, the diagonal entries of are a11 = 2 and a22 = 1.
3 1 8
 
2 4 9
(f ) False. The diagonal entries of 3 1 8  are a11 = 2, a22 = 1, and a33 = 10.
5 7 10
(g) False. Indeed, by definition, a diagonal matrix is a square matrix whose every
nondiagonal entry is 0.
 
2 5
(h) False. The matrix 1 3 is not a square matrix because the number of rows,
0 0
m = 3, is not the same as the number of columns, n = 2.
 
0 6
(i) False. The matrix is not a diagonal matrix because there is a nonzero
9 0
nondiagonal entry, a12 = 6 6= 0 (also, a21 = 9 6= 0).
 
1 0 0
(j) True. The matrix 0 2 0 is a diagonal matrix since it is a square matrix in
0 0 3
which every nondiagonal entry is 0.
 
0 1
(k) False. Indeed, the matrix is not even a diagonal matrix.
1 0
 
1 0 0
(l) True. Indeed, the matrix 0 1 0 is a 3 × 3 diagonal matrix in which every
0 0 1
diagonal entry is 1.
(m) False. By definition, the sum of two matrices is defined if and only if they have
the same size.

2. [Exercise on page 36]

4.1. Matrix Operations 211

Matrix Addition
Let A = [aij ] and B = [bij ] be matrices of the same size. Then the sum A + B
is the matrix whose (i, j)-entry is aij + bij . That is,

A + B = [aij + bij ]

As an immediate consequence of the definition, if A and B have the same size,

say m × n, then the sum A + B is of size m × n.

(a)
   
1 7 −2 6
A+B = +
−3 8 5 −14
   
1 + (−2) 7+6 −1 13
= =
−3 + 5 8 + (−14) 2 −6

(b)
   
9 −7 17 −9
C + D =  −3 0  +  −4 1 
11 −22 0 19
   
9 + 17 −7 − 9 26 −16
=  −3 − 4 0 + 1  =  −7 1 
11 + 0 −22 + 19 11 −3

(c) The sum A + C is undefined because A and C do not have the same size.
(d) The sum C + D + B is undefined because D and B (as well as C and B) do not
have the same size.
(e)
Scalar Multiplication

Let A = [aij ] be an m × n matrix, and let c be a scalar. Then the scalar

multiplication cA is the matrix of size m × n defined by

cA = [caij ]

In other words, cA is the matrix obtained from A by multiplying every

entry of A by c.

 
9 −7
For c = 2 and C =  −3 0 , we have
11 −22
     
9 −7 2(9) 2(−7) 18 −14
2C = 2  −3 0  =  2(−3) 2(0)  =  −6 0 
11 −22 2(11) 2(−22) 22 −44
212 Chapter 4. Matrix Algebra

(f )
   
9 −7 17 −9
3C − 2D = 3  −3 0  − 2  −4 1 
11 −22 0 19
   
27 −21 −34 18
=  −9 0 + 8 −2 
33 −66 0 −38
   
27 − 34 −21 + 18 −7 −3
=  −9 + 8 0 − 2  =  −1 −2 
33 + 0 −66 − 38 33 −104

3. [Exercise on page 36]

 
1 0
ˆ Remembering that I2 = , we have
0 1
         
2 −2 1 0 2 −2 −2 0 0 −2
A − 2I2 = −2 = + =
5 12 0 1 5 12 0 −2 5 10
 
1 0 0
ˆ Remembering that I3 = 0 1 0, we have
0 0 1
   
1 0 0 3 −3 5
7I3 − B = 7  0 1 0  −  −4 7 10 
0 0 1 11 9 −8
     
7 0 0 −3 3 −5 4 3 −5
= 0 7 0 + 4 −7 −10  =  4 0 −10 
0 0 7 −11 −9 8 −11 −9 15

4. [Exercise on page 36]

    
1 7−2 6 1 0
−5A + 2B − 3I2 = −5 +2 −3
−3 8 5 −14 0 1
     
−5 −35 −4 12 −3 0
= + +
15 −40 10 −28 0 −3
   
−5 − 4 − 3 −35 + 12 + 0 −12 −23
= =
15 + 10 + 0 −40 − 28 − 3 25 −71

5. [Exercise on page 36]

   
5 −3 −2 8
We want to solve the equation X + = −6 .
1 6 1 0
4.1. Matrix Operations 213

 
5 −3
Subtracting the matrix from both sides, we get
1 6
       
−2 8 5 −3 12 − 5 −48 + 3 7 −45
X = −6 − = =
1 0 1 6 −6 − 1 0−6 −7 −6

6. [Exercise on page 36]

 
Let A = aij . Then
 
(α + β)A = (α + β)aij By definition of scalar multiplication
 
= αaij + βaij Property of real numbers
   
= αaij + βaij By definition of matrix addition
   
= α aij + β aij By definition of scalar multiplication
= αA + βA

This is a property of matrix addition and scalar multiplication. Other properties are
listed in the following theorem.

Properties of Matrix Addition and Scalar Multiplication

Theorem 4.1.1. Let A, B, and C be matrices of the same size, and let α and
β be scalars.

(a) A + B = B + A (commutativity of addition)

(b) A + (B + C) = (A + B) + C (associativity of addition)

(c) A + 0 = A

(d) α(A + B) = αA + αB (distributivity of scalar multiplication)

(e) (α + β)A = αA + βA (distributivity of scalar multiplication)

(f ) (αβ)A = α(βA) (associativity of scalar multiplication)

(g) A + (−A) = 0

7. [Exercise on page 36]

Matrix Multiplication
Let A be an m × n matrix, and let B be an n × p matrix. Let b1 , b2 , · · · , bp be
the columns of B. The product AB is the m × p matrix whose columns are
Ab1 , Ab2 , · · · , Abp , that is,
   
AB = A b1 b2 · · · bp = Ab1 Ab2 · · · Abp
Note.
ˆ The product AB is defined if and only if the number of columns of A
214 Chapter 4. Matrix Algebra

equals the number of rows of B.

ˆ If A is of size m × n and B is of size n × p, then AB is of size m × p. In

other words,

– the number of rows of AB corresponds to the number of rows of A.

– The number of columns of AB corresponds to the number of columns
of B.

(a) False. Indeed, let m = 2, n = 3, and p = 3. Then A is of size m × n = 2 × 3 and

B is of size m × p = 2 × 3. But AB is not defined since the number of columns
of A does not match the number of rows of B.
(b) True by definition.
(c) False. Indeed, let A be a 2 × 3 matrix and let B be a 4 × 5 matrix. Then A
and B do not have the same size. But the product AB is undefined because the
number of columns of A, 3, is not equal to the number of rows of B, 4.
(d) False. Indeed, let n × p be the size of B.
ˆ Since the number of columns of A corresponds to the number of rows of B,
and since the number of columns of A is 7, it follows that n = 7.
ˆ Since the number of columns of AB corresponds to the number of columns
of B, and since the number of columns of AB is 11, it follows that p = 11.
So B is of size 7 × 11 (not 5 × 11).
(e) True. Indeed, suppose A has m rows and n columns.
ˆ Since the number of rows of AB corresponds to the number of rows of A,
and since the number of rows of AB is 3, it follows that m = 3.
ˆ Since the columns of A corresponds to the number of rows of B, and since
B has 5 rows, it follows that n = 5.
So the size of A is 3 × 5.
(f ) False. Indeed, let A be a 2 × 3 matrix and let B be a 3 × 2 matrix. Clearly,
both A and B are not square matrices. But the product AB is a square matrix
since it is of size 2 × 2.
(g) False. Indeed, let A be a 2 × 3 matrix and let B be a 3 × 2 matrix. Then AB
and BA do not have the same size since AB is of size 2 × 2, and BA is of size
3 × 3.
   
1 2 5 6
(h) If A = and B = , then
3 4 7 8
 
False. Let b1 and b2 be the columns of B. Then, by definition, AB = Ab1 Ab2 .
But         
5 1 2 5 1(5) + 2(7) 19
Ab1 = A = = =
7 3 4 7 3(5) + 4(7) 43
        
6 1 2 6 1(6) + 2(8) 22
Ab2 = A = = =
8 3 4 8 3(6) + 4(8) 50
4.1. Matrix Operations 215

 
19 22
So AB = .
43 50
(i) True by the result obtained in part (h).

8. [Exercise on page 37]

   
1 2 −3 5  
(a) For A = ,B = , we have B = b1 b2 , where b1 is the first
3 4 4 −7
column, and b2 is the second of B.
ˆ The product AB is defined because the number of columns of A equals the
number of rows of B.
ˆ By definition, the product AB is Ab1 Ab2 .
 
        
1 2 −3 1(−3) + 2(4) −3 + 8 5
ˆ Ab1 = = = = .
3 4 4 3(−3) + 4(4) −9 + 16 7
        
1 2 5 1(5) + 2(−7) 5 − 14 −9
ˆ Ab2 = = = = .
3 4 −7 3(5) + 4(−7) 15 − 28 −13
ˆ So  
  5 −9
AB = Ab1 Ab2 =
7 −13

Alternate Solution
One can compute AB directly as follows.
    
1 2 −3 5 1(−3) + 2(4) 1(5) + 2(−7)
AB = =
3 4 4 −7 3(−3) + 4(4) 3(5) + 4(−7)
   
−3 + 8 5 − 14 5 −9
= =
−9 + 16 15 − 28 7 −13

From now on, we will compute the product of two matrices this way.

(b)
    
−3 5 −3(1) + 5(3) −3(2) + 5(4)
1 2
BA = =
4 −7 3 44(1) − 7(3) 4(2) − 7(4)
   
−3 + 15 −6 + 20 12 14
= =
4 − 21 8 − 28 −17 −20

(c) The product BC is undefined because the number of columns of B is not equal
to the number of rows of C.
(d)
   
2 0   2(−3) + 0(4) 2(5) + 0(−7)
−3 5
CB =  5 −8  =  5(−3) − 8(4) 5(5) − 8(−7) 
4 −7
13 −9 13(−3) − 9(4) 13(5) − 9(−7)
216 Chapter 4. Matrix Algebra

   
−6 + 0 10 + 0 −6 10
=  −15 − 32 25 + 56  =  −47 81 
−39 − 36 65 + 63 −75 128

(e) The product CD is undefined because the number of columns of C is not equal
to the number of rows of D.
(f )
  
−2 1 3 2 0
DC =  4 8 9   5 −8 
10 −4 5 13 −9
 
−2(2) + 1(5) + 3(13) −2(0) + 1(−8) + 3(−9)
=  4(2) + 8(5) + 9(13) 4(0) + 8(−8) + 9(−9) 
10(2) − 4(5) + 5(13) 10(0) − 4(−8) + 5(−9)
   
−4 + 5 + 39 0 − 8 − 27 40 −35
=  8 + 40 + 117 0 − 64 − 81  =  165 −145 
20 − 20 + 65 0 + 32 − 45 65 −13

(g)
 
2 0  
−2 6 9
CE =  5 −8 
3 −7 7
13 −9
 
2(−2) + 0(3) 2(6) + 0(−7) 2(9) + 0(7)
=  5(−2) − 8(3) 5(6) − 8(−7) 5(9) − 8(7) 
13(−2) − 9(3) 13(6) − 9(−7) 13(9) − 9(7)
   
−4 + 0 12 + 0 18 + 0 −4 12 18
=  −10 − 24 30 + 56 45 − 56  =  −34 86 −11 
−26 − 27 78 + 63 117 − 63 −53 141 54

(h)
 
  2 0
−2 6 9  5 −8 
EC =
3 −7 7
13 −9
 
−2(2) + 6(5) + 9(13) −2(0) + 6(−8) + 9(−9)
=
3(2) − 7(5) + 7(13) 3(0) − 7(−8) + 7(−9)
   
−4 + 30 + 117 0 − 48 − 81 143 −129
= =
6 − 35 + 91 0 + 56 − 63 62 −7

(i)
  
−2 1 3 0 9 4
DF =  4 8 9   −2 −2 −5 
10 −4 5 1 −10 −1
4.1. Matrix Operations 217

 
−2(0) + 1(−2) + 3(1) −2(9) + 1(−2) + 3(−10) −2(4) + 1(−5) + 3(−1)
=  4(0) + 8(−2) + 9(1) 4(9) + 8(−2) + 9(−10) 4(4) + 8(−5) + 9(−1) 
10(0) − 4(−2) + 5(1) 10(9) − 4(−2) + 5(−10) 10(4) − 4(−5) + 5(−1)
   
0 − 2 + 3 −18 − 2 − 30 −8 − 5 − 3 1 −50 −16
=  0 − 16 + 9 36 − 16 − 90 16 − 40 − 9  =  −7 −70 −33 
0+8+5 90 + 8 − 50 40 + 20 − 5 13 48 55

9. [Exercise on page 37]

We know that the product AB of a matrix A of size m × n by a matrix B of size
n × p is a matrix of size m × p. Here m = 5, n = 7 and p = 11. So the size of B is
n × p = 7 × 11.

10. [Exercise on page 37]

(a) If the product BA is of size 17 × 32, then the size of B is of the form 17 × n,
and the size of A is of the form n × 32 where n is a positive integer. So B has
17 rows.
(b) Since the size of A is of the form n × 32, the number of columns of A is 32.

11. [Exercise on page 37]

 
b1j
 .. 
Let ai be the ith column of A, and let bj =  .  be the jth column of B. By
bnj
 
definition, AB = Ab1 · · · Abp . For 1 ≤ j ≤ p, the jth column of AB is

 
b1j
an  ...  = b1j a1 + · · · + bnj an
  
Abj = a1 · · ·
bnj
This shows that every column of AB is a linear combination of the columns of A.
This proves the following result.

Columns of the Product as Linear Combinations

Theorem 4.1.2. Let A and B be two matrices such that the product AB is
defined. Then every column of AB is a linear combination of the columns of
A.

12. [Exercise on page 37]

 
Let b1 and b2 be
 the columns of B. Then
  , by definition,
  AB = Ab1 Ab2 . Since
 
0 2 0 2 b11
AB = , it follows that Ab1 = and Ab2 = . Let b1 = and
 2 0 2 0 b21
b12
b2 = . Then
b22
218 Chapter 4. Matrix Algebra

     
b11 0 0
ˆ Ab1 = a1 a2
 
= b11 a1 +b21 a2 . Since Ab1 = , b11 a1 +b21 a2 = .
b21  2 2
0
From the equation 2a1 − a2 = , it follows that b11 a1 + b21 a2 = 2a1 − a2 . So
2
b11 = 2 and b21 = −1.
ˆ Similarly, b12 = −4 and b22 = 3.
 
2 −4
So B = .
−1 3

13. [Exercise on page 38]

 
a b
Let B = . Then
c d
    
1 −2 a b a − 2c b − 2d
AB = =
3 −1 c d 3a − c 3b − d

ˆ Observe that the first column of AB is the vector

     
a − 2c 1 −2
=a +c
3a − c 3 −1
 
5 5
Since AB = , it follows that
5 5
     
1 −2 5
a +c =
3 −1 5

This gives the following linear system (in two variables)

a − 2c = 5
3a − c = 5

Row reducing the augmented matrix, we have

   
1 −2 5 1 −2 5
R2 − 3R1 → R2
3 −1 5 0 5 −10
 
1 1 −2 5
R
5 2
→ R2
0 1 −2
 
1 0 1
R1 + 2R2 → R1
0 1 −2

So a = 1 and c = −2.
 
b − 2d
ˆ Using the second column of AB, , and the same approach as before,
3b − d
we find b = 1 and d = −2.
 
1 1
Thus, B =
−2 −2
4.1. Matrix Operations 219

14. [Exercise on page 38]

Let b3 be the third column of AB and suppose that b3 = 0. By the definition of
matrix multiplication, the third column of AB is Ab3 = A0 = 0.
15. [Exercise on page 38]
(a) ˆ Let x, y be vectors in Rp . Then

(R ◦ T )(x + y) = R(T (x + y)) By definition of the composition

= R(T (x) + T (y)) Because T is linear
= R(T (x)) + R(T (y)) Because R is linear
= (R ◦ T )(x) + (R ◦ T )(y)
ˆ Similarly, for every scalar c and any vector x is Rp , we have (R ◦ T )(cx) =
R(T (cx)) = R(cT (x)) = cR(T (x)) = c(R ◦ T )(x).
Thus, R ◦ T is a linear transformation.
(b) ˆ Since the matrix of T is A, we have T (x) = Ax for every x in Rp .
ˆ Since the matrix of R is B, we have R(y) = By for every y in Rn .
ˆ For any x in Rp , we have

(R ◦ T )(x) = R(T (x)) = R(Ax) = B(Ax) = (BA)x

This shows that the standard matrix of R ◦ T is the product BA. This
proves the following theorem.

Standard Matrix of the Composition of two Linear Transfor-

mations
Theorem 4.1.3. Let T : Rp → Rn and R : Rn → Rm be linear trans-
formations. If A is the standard matrix of T , and if B is the standard
matrix of R, then the standard matrix of the composition R ◦T is BA.
In other words, this says that the standard matrix of the composition
R ◦ T is the standard matrix of R times the standard matrix of T .

.
(c) First, we can find the standard matrix
 of R ◦ T by using Theorem 4.1.3. Since
1 −1
the standard matrix of T is A = , and since the standard matrix of R
  3 4
−2 5
is B = , it follows that the standard matrix of R ◦ T is
6 −7
      
−2 5 1 −1 −2 + 15 2 + 20 13 22
BA = = =
6 −7 3 4 6 − 21 −6 − 28 −15 −34

Another way to find this is to first find a formula for the composition R ◦ T .
      
x1 x1 x1 − x2
(R ◦ T ) =R T =R
x2 x2 3x1 + 4x2
220 Chapter 4. Matrix Algebra

 
−2(x1 − x2 ) + 5(3x1 + 4x2 )
=
6(x1 − x2 ) − 7(3x1 + 4x2 )
   
−2x1 + 2x2 + 15x1 + 20x2 13x1 + 22x2
= =
6x1 − 6x2 − 21x1 − 28x2 −15x1 − 34x2
 
13 22
So the standard matrix of R ◦ T is .
−15 −34

16. [Exercise on page 38]

Let c1 , c2 , · · · , cq be the columns of C. Then, by definition,
 
BC = Bc1 Bc2 · · · Bcq

and  
A(BC) = A(Bc1 ) A(Bc2 ) · · · A(Bcq )
p
To finish,
  we need to show that for every vector x in R , A(Bx) = (AB)x. Let
x1
 x2 
x =  ..  be a vector in Rp , and let b1 , b2 , · · · , bp be the columns of B. Then
 
.
xp
  
x1
x2 
  
A(Bx) = A  b1 b2 · · · bp  .. 

  . 
xp
By the definition of the product
= A (x1 b1 + x2 b2 + · · · + xp bp )
of a matrix by a vector
= A(x1 b1 ) + A(x2 b2 ) + · · · + A(xp bp )
= x1 Ab1 + x2 Ab2 + · · · + xp Abp
 
x1
   x2 

By the definition of the product
= Ab1 Ab2 · · · Abp  .. 

. of a matrix by a vector
xp
By definition of matrix multipli-
= (AB)x
cation

So
 
A(BC) = A(Bc1 ) A(Bc2 ) · · · A(Bcq )
 
= (AB)c1 (AB)c2 · · · (AB)cq
= (AB)C

This proves the first property of the following theorem.

4.1. Matrix Operations 221

Properties of Matrix Multiplication

Theorem 4.1.4. Let A, B, C be three matrices, and let α be a scalar.

(a) A(BC) = (AB)C (associativity of the product)

(b) A(B + C) = AB + AC (left distributive property)

(c) (A + B)C = AC + BC (right distributive property)

(d) α(AB) = (αA)B = A(αB)

(e) Im A = A = AIn (here A is of size m × n).

17. [Exercise on page 38]

(a) Multiplying both sides of the equation Ax = 0 by B, we get BAx = 0, which

implies that Im x = 0 (since BA = Im by hypothesis). But, by Theorem 4.1.4 -
(e), Im x = x. So the equation Im x = 0 becomes x = 0. This shows that the the
equation Ax = 0 has only the trivial solution.
(b) From part (a), the equation Ax = 0 has only the trivial solution. This means
that the columns of A are linearly independent, which in turn implies that
the number of rows of A is greater than or equal to the number of columns.
(Indeed, if the number of columns of A is greater than the number of rows, then
(by Theorem 2.4.6) the columns of A would be linearly dependent.)
(c) Yes, the linear transformation T (x) = Ax is one-to-one since the columns of A
are linearly independent.

18. [Exercise on page 38]

   
1 1 0 0
(a) False. Indeed, let A = and let B = . Then
0 0 1 1
      
1 1 0 0 0+1 0+1 1 1
AB = = =
0 0 1 1 0+0 0+0 0 0
      
0 0 1 1 0+0 0+0 0 0
BA = = =
1 1 0 0 1+0 1+0 1 1
We see that AB 6= BA. This gives the following.

Warning
In general, AB 6= BA.

(b) True by Theorem 4.1.4-(c).

(c) False. Indeed, by Theorem 4.1.4-(b), we have A(B + C) = AB + AC. But AC
is not equal to CA in general as we saw in part (a).
222 Chapter 4. Matrix Algebra

     
1 0 1 1 0 0
(d) False. Indeed, let A = ,B = , and C = . Then
0 0 0 0 1 1
      
1 1 0 0 1 1 1 1
BC = = and A(BC) =
0 0 1 1 0 0 0 0

On the other hand,

   
0 0 0 0
AC = and (AC)B =
0 0 0 0

We see that A(BC) 6= (AC)B.

     
1 1 0 0 1 0
(e) False. Indeed, let A = ,B = , and C = = I2 . Then, from
0 0 1 1 0 1
part (a), AB = A. Since AI2 = A, it follows that AB = AI2 . But B 6= I2 . This
gives the following.

Warning
If AB = AC, it is not true in general that B = C. This means that the
cancellation properties do not hold for matrix multiplication.

   
0 0 1 0
(f ) False. Indeed, let A = and let B = . Clearly, we have AB =
  0 1 0 0
0 0
, but A 6= 0 and B 6= 0. This gives the following.
0 0

Warning
If AB = 0, it is not true in general that either A = 0 or B = 0. This means
that the zero product property does not hold for matrix multiplication.

(g) True. Indeed, if A = 0, then 0B = 0. If B = 0, then A0 = 0.

19. [Exercise on page 39]

Commutative Matrices
Let A and B be matrices. We say that A and B commute if AB = BA.
Warning: In general, AB 6= BA as we saw in Exercise 18-(a).

We want to show that A commutes with BC.

A(BC) =(AB)C Associativity of the product

=(BA)C Because A and B commute
=B(AC) Associativity of the product
=B(CA) Because A and C commute
4.1. Matrix Operations 223

=(BC)A Associativity of the product

This shows that A commutes with BC.

20. [Exercise on page 39]

Power of a Matrix.
Let A be a square matrix, and let k be a positive number. Define Ak as

Ak = A
| ·{z
· · A}
k

 
1 2
ˆ For A = , we have
3 4
      
2 1 2 1 2 1(1) + 2(3) 1(2) + 2(4) 7 10
A = AA = = =
3 4 3 4 3(1) + 4(3) 3(2) + 4(4) 15 22

ˆ To find B 3 , we first need to find B 2 .

  
0 −1 2 0 −1 2
B2 =  3 1 0  3 1 0 
−2 0 1 −2 0 1
 
0(0) − 1(3) + 2(−2) 0(−1) − 1(1) + 2(0) 0(2) − 1(0) + 2(1)
=  3(0) + 1(3) + 0(−2) 3(−1) + 1(1) + 0(0) 3(2) + 1(0) + 0(1) 
−2(0) + 0(3) + 1(−2) −2(−1) + 0(1) + 1(0) −2(2) + 0(0) + 1(1)
   
0−3−4 0−1+0 0−0+2 −7 −1 2
=  0 + 3 + 0 −3 + 1 + 0 6 + 0 + 0  =  3 −2 6 
0+0−2 2 + 0 + 0 −4 + 0 + 1 −2 2 −3

We now have
  
−7 −1 2 0 −1 2
B 3 = BBB = B 2 B =  3 −2 6  3 1 0 
−2 2 −3 −2 0 1
 
−7(0) − 1(3) + 2(−2) −7(−1) − 1(1) + 2(0) −7(2) − 1(0) + 2(1)
=  3(0) − 2(3) + 6(−2) 3(−1) − 2(1) + 6(0) 3(2) − 2(0) + 6(1) 
−2(0) + 2(3) − 3(−2) −2(−1) + 2(1) − 3(0) −2(2) + 2(0) − 3(1)
   
0−3−4 7 − 1 + 0 −14 + 0 + 2 −7 6 −12
=  0 − 6 − 12 −3 − 2 + 0 6 − 0 + 6  =  −18 −5 12 
0 + 6 + 6 +2 + 2 − 0 −4 + 0 − 3 12 4 −7

21. [Exercise on page 39]

ˆ A2 is not defined because A is not a square matrix.

224 Chapter 4. Matrix Algebra

ˆ  
  −1    
AB = −2 3 5  2  = −2(−1) + 3(2) + 5(4) = 28
4
ˆ
   
−1   −1(−2) −1(3) −1(5)
BA =  2  −2 3 5 =  2(−2) 2(3) 2(5) 
4 4(−2) 4(3) 4(5)
 
2 −3 −5
= −4
 6 10 
−8 12 20

ˆ B 2 is undefined because B is not a square matrix.

 
a 0
22. Let A = be a 2 × 2 diagonal matrix. [Exercise on page 39]
0 b
(a)

   2 
2 a 0 a 0 a 0
A = AA = =
0 b 0 b 0 b2
 2    3 
3 2 a 0 a 0 a 0
A =A A= =
0 b2 0 b 0 b3

(b) From part (a), we can make the guess that

 n 
n a 0
A =
0 bn

23. [Exercise on page 39]

Transpose of a Matrix

Let A = [aij ] be an m × n matrix. The transpose of A, denoted AT , is the

n × m matrix whose (i, j)-entry is aji . In other words, the transpose of A is
the n × m matrix whose rows are just the columns of A in the same order.


0
ˆ For A = 0 6 7 , AT =  6 .
 

7
 
−3
ˆ For B =  9 , B T = −3 9 −8 .
 

−8
 
  0 3
0 −1 2
ˆ For C = , C T =  −1 1 .
3 1 0
2 0
4.1. Matrix Operations 225

   
−4 2 −2 −4 3 10
ˆ For D =  3 1 −7  , DT =  2 1 13 .
10 13 20 −2 −7 20

24. [Exercise on page 39]

ˆ

 4 −2   
T 4 −1 2  1 0
AA − 5I2 = −1 1 −5
−2 1 3 0 1
2 3
   
16 + 1 + 4 −8 − 1 + 6 1 0
= −5
−8 − 1 + 6 4+1+9 0 1
     
21 −3 −5 0 16 −3
= + =
−3 14 0 −5 −3 9
ˆ
   
1 0 0 4 −2  
T 4 −1 2
8I3 − A A = 8 0 1 0 − −1
   1 
−2 1 3
0 0 1 2 3
   
1 0 0 16 + 4 −4 − 2 8−6
= 8  0 1 0  −  −4 − 2 1 + 1 −2 + 3 
0 0 1 8 − 6 −2 + 3 4+9
     
8 0 0 20 −6 2 −12 6 −2
=  0 8 0  −  −6 2 1 = 6 6 −1 
0 0 8 2 1 13 −2 −1 −5

25. [Exercise on page 39]

Some Properties of Transposes

Theorem 4.1.5. Let α be a scalar, and let A and B be two matrices. Then

(a) (AT )T = A

(b) (A + B)T = AT + B T

(c) (αA)T = αAT

(d) (AB)T = B T AT

The generalization of (d) is that if A1 , · · · , Ak are matrices, then

(A1 A2 · · · Ak−1 Ak )T = ATk ATk−1 · · · AT2 AT1

(a) True by Theorem 4.1.5-(a).

(b) True by Theorem 4.1.5-(b).
226 Chapter 4. Matrix Algebra

   
1 1 1 0
(c) False. Indeed, let A = and B = .
0 1 1 1
      
1 1 1 0 2 1 2 1 T
AB = = and (AB) =
0 1 1 1 1 1 1 1

On the other hand, we have

    
T1 0T 1 1 1 1
A B = =
1 1 0 1 1 2

We see that (AB)T 6= AT B T .

(d) True by Theorem 4.1.5-(d).
(e) True. Indeed,

(ABC)T = ((AB)C)T = C T (AB)T = C T (B T AT ) = C T B T AT

(f ) False. Indeed, (AB)2 = (AB)(AB) = A(BA)B.  This

 is not equal
 toA(AB)B
1 1 1 0
in general. Here is a counterexample. Let A = and B = .
0 1 1 1
         
1 1 1 0 2 1 2 1 2 2 1 5 3
AB = = and (AB) = =
0 1 1 1 1 1 1 1 1 1 3 2

On the other hand, we have

         
2 1 1 1 1 1 2 2 1 0 1 0 1 0
A = = , B = =
0 1 0 1 0 1 1 1 1 1 2 1
    
1 2 1 0 5 2
and A2 B 2 = = . We see that (AB)2 6= A2 B 2 .
0 1 2 1 2 1
(g) True. Indeed,

(A − In )(A + In ) = A2 + AIn − In A − In2 = A2 + A − A − In = A2 − In

26. [Exercise on page 39]

In general, we have

(A + B)2 = (A + B)(A + B) = A2 + AB + BA + B 2
2 2 2
But AB 6= BA in general, so the equality
 (A + B) = A + 2AB + B is not true in
1 1 0 0
general. For example, take A = and B = . Then
0 0 1 1
      
1 1 1 1 1 1 2 2
ˆ A+B = and (A + B) = 2
= .
1 1 1 1 1 1 2 2
4.2. The Inverse of a Matrix 227

ˆ On the other hand,

     
2 1 1 2 0 0 1 1
A = ,B = , AB = , and
0 0 1 1 0 0
       
2 2 1 1 1 1 0 0 3 3
A + 2AB + B = +2 + =
0 0 0 0 1 1 1 1

One can see that (A + B)2 6= A2 + 2AB + B 2 .

27. [Exercise on page 40]

ˆ The first mistake is at the second line because XC−BX is not equal to X(C−B).
Indeed, X(C − B) = XC − XB, but XB 6= BX in general (see Exercise 18a).
ˆ The second mistake is at the last line. To understand why it is a mistake, see
Exercise 18f.

28. [Exercise on page 40]

We need to show that (AAT )T = AAT . We have

(AAT )T =(AT )T AT By property (d) of Theorem 4.1.5

T
=AA By property (a) of Theorem 4.1.5

This shows that AAT is a symmetric matrix.

4.2 The Inverse of a Matrix

1. [Exercise on page 40]

The Inverse of a Matrix – Definition

Let A be an n×n matrix (a square matrix), and let I = In be the n×n identity
matrix.
ˆ A matrix B is called an inverse of A if

AB = I and BA = I

ˆ If A has an inverse, we say that A is invertible.

Theorem 4.2.1. Let A be an n × n matrix. If B and C are both inverses of
A, then B = C.
Proof. Let B and C be inverses of A. Then, by definition,
AB = I and BA = I, and CA = I and AC = I
Using these, we have
B = BI = BAC I = AC
228 Chapter 4. Matrix Algebra

= B(AC) = (BA)C Associativity of the product

= IC = C I = BA

Notation. Thanks to Theorem 4.2.1, we can talk about the inverse of A. If A

is invertible, the inverse of A is denoted A−1 . So

AA−1 = I and A−1 A = I

(a) True by definition.

(b) True by definition.
(c) False because only square matrices have inverses.
(d) False. Indeed,
    
−2 0 −1 0 2 0
AB = = 6= I
0 −2 0 −1 0 2

(e) True. Indeed,

    
1 0 1 0 1 0
AB = = =I
1 1 −1 1 0 1
and     
1 0 1 0 1 0
BA = = =I
−1 1 1 1 0 1
(f ) True by Theorem 4.2.1.
(g) True. Indeed, the equality A2 = I is equivalent to AB = I and BA = I, where
B = A. This implies (by definition and Theorem 4.2.1) that A is invertible and
A−1 = B = A.

2. [Exercise on page 40]

The equation A5 = I implies that AA4 = I and A4 A = I. Therefore, A4 is an inverse
of A, and this shows that A is invertible. Since the inverse is unique by Theorem 4.2.1,
it follows that A−1 = A4 .
3. [Exercise on page 40]
(a) Since
  
1 0 1 1 −1 2
AB =  2 2 3   −1 0 1 
1 1 1 0 1 −2
   
1 + 0 + 0 −1 + 0 + 1 2 + 0 − 2 1 0 0
=  2 − 2 + 0 −2 + 0 + 3 4 + 2 − 6  =  0 1 0  = I,
1 − 1 + 0 −1 + 0 + 1 2 + 1 − 2 0 0 1
4.2. The Inverse of a Matrix 229

and
    
1 −1 2 1 0 1 1 0 0
BA =  −1 0 1   2 2 3  =  0 1 0  = I,
0 1 −2 1 1 1 0 0 1

it follows that B is the inverse of A. In other words, A−1 = B.

(b) We have
    
1 0 0 1 0 0 1 0 0
AB =  0 1 −1   0 2 3  =  0 1 0 =6 I
0 1 −2 0 1 3 0 0 −3

Since AB is not the identity matrix, it follows that B is not the inverse of A.

4. [Exercise on page 40]

The Inverse of a 2 × 2 Matrix

 
a b
Let A = be a 2 × 2 matrix. The determinant of A, denoted det A,
c d
is the quantity
det A = ad − bc

Theorem 4.2.2. (a) The matrix A is invertible if and only if det A 6= 0.

(b) If det A 6= 0, the inverse of A is given by

 
−1 1 d −b
A =
ad − bc −c a

 
1 2
(a) For the matrix A = , the determinant is
2 6

det A = (1)(6) − 2(2) = 6 − 4 = 2

Since det A 6= 0, A is invertible and

" # " #
6
− 22 3 −1
 
1 6 −2
A−1 = = 2
=
2 −2 1 − 22 1
−1 1
2 2

 
−6 7
(b) ˆ The determinant of A = is
12 −14

det A = (−6)(−14) − (12)(7) = 84 − 84 = 0

ˆ Since det A = 0, A is not invertible. So A−1 does not exist.

230 Chapter 4. Matrix Algebra

 
−2 3
(c) The determinant of A = is
−4 8

det A = (−2)(8) − (−4)(3) = −16 + 12 = −4

Since det A 6= 0, it follows that A is invertible and

 " 8 −3 # " 3
#
−2

1 8 −3
−1
A = = −4 −4
−2
= 4
−4 −(−4) −2 4
−4 −4 −1 1
2

 
15 −18
(d) The determinant of A = is
−10 12

det A = 15(12) − (−10)(−18) = 180 − 180 = 0

Since det A = 0, it follows that A is not invertible. So A−1 does not exist.

5. [Exercise on page 41]

 
k −5
The determinant of A = is
−4 k + 1

det A = k(k + 1) − (−4)(−5) = k 2 + k − 20 = (k − 4)(k + 5)

So A is invertible if and only if (k − 4)(k + 5) 6= 0, that is, k 6= 4 and k 6= −5.

6. [Exercise on page 41]

Matrix Inverses and Linear Systems

Theorem 4.2.3. Let Ax = b be a linear system with n equations in n variables.
If the coefficient matrix A is invertible, the system has a unique solution. (In
fact, the only solution is x = A−1 b.)

Proof. Suppose that A is invertible.

Ax = b
A−1 Ax = A−1 b Left-multiply both sides by A−1
Ix = A−1 b Since A−1 A = I
x = A−1 b Since Ix = x

−3x1 − 6x2 = 10
(a) ˆ The matrix equation for the system is Ax = b with
−x1 + 3x2 = 10
     
−3 −6 x1 10
A= , x= , and b =
−1 3 x2 10
4.2. The Inverse of a Matrix 231

ˆ The determinant of A is det A = (−3)(3) − (−1)(−6) = −9 − 6 = −15.

ˆ Since det A 6= 0, A is invertible and
 " 3 6
# " #
− 15 − 52

−1 1 3 6 −15 −15
A = = −3
=
−15 1 −3 1
−15 −15
1
− 15 1
5

ˆ The solution to the system is

" #  " #
−1 − 1
5
− 2
5 10 − 10
5
− 20
5
x=A b= 1 1 =
− 15 5
10 − 10 + 10
15 5
" # " # " #
− 10
5
− 20
5 − 30
5
−6
= 10 30
= 20 = 4
− 15 + 15 15 3

(b) ˆ Here the matrix equation is Ax = b with

     
4 −5 x 11
A= , x= 1 , and b =
7 6 x2 −3

ˆ The determinant of A is det A = (4)(6) − (7)(−5) = 24 + 35 = 59.

ˆ Since det A 6= 0, A is invertible and
  " 6 5 #
−1 1 6 5 59 59
A = = −7
59 −7 4 4
59 59

ˆ The solution to the system is

" #  " # " #
6 5 66 15 51
11 −
x = A−1 b = −7 59 59
= 59 59
= 59
4 −3 −77
− 12 89
− 59
59 59 59 59

7. [Exercise on page 41]

An Inversion Method
Let A be a square matrix.

ˆ If A can be reduced to the identity matrix I by performing a sequence

O1 , · · · , Or of elementary row operations, then A is invertible. And A−1 is
the matrix obtained by performing exactly the same sequence O1 , · · · , Or
of operations on the identity matrix I.

ˆ If A cannot be reduced to the identity matrix, A−1 does not exist.

So to find
 the inverse of A (if any), we start off with the augmented matrix
A I . The goal is to use elementary row operations to transform A into I
(if this is possible). These operations are performed on A and I simultaneously.
If A is invertible, we will end up with the augmented matrix I A−1 .

232 Chapter 4. Matrix Algebra

 
1 1 1
(a) We find the inverse of A =  0 1 1  (if any). The augmented matrix
  2 3 4
A I is
 
1 1 1 1 0 0
 0 1 1 0 1 0 
2 3 4 0 0 1
We perform elementary row operations, the idea being to transform A into the
identity matrix (if it is possible).
   
1 1 1 1 0 0 1 1 1 1 0 0
 0 1 1 0 1 0  R3 − 2R1 → R3  0 1 1 0 1 0 
2 3 4 0 0 1 0 1 2 −2 0 1
 
1 1 1 1 0 0
R3 − R2 → R3  0 1 1 0 1 0 
0 0 1 −2 −1 1
 
1 1 0 3 1 −1
R1 − R3 → R1  0 1 0 2 2 −1 
R2 − R3 → R2
0 0 1 −2 −1 1
 
1 0 0 1 −1 0
R1 − R2 → R1  0 1 0 2 2 −1 
0 0 1 −2 −1 1
Since we were able to transform A into the identity matrix using elementary row
operations, it follows that A is invertible and
 
1 −1 0
A−1 =  2 2 −1 
−2 −1 1
It is a good idea to check that AA−1 = I:
    
1 1 1 1 −1 0 1 0 0
AA−1 =  0 1 1  2 2 −1  =  0 1 0  = I
2 3 4 −2 −1 1 0 0 1
It is not necessary to check that A−1 A = I because A is invertible.
 
1 −3 −2
(b) A =  −1 5 2 
−3 7 7
 
We reduce the augmented matrix A I .

   
1 −3 −2 1 0 0 1 −3 −2 1 0 0
 −1 R2 + R1 → R2
5 2 0 1 0   0 2 0 1 1 0 
R3 + 3R1 → R3
−3 7 7 0 0 1 0 −2 1 3 0 1
4.2. The Inverse of a Matrix 233

 
1 −3 −2 1 0 0
R3 + R2 → R3  0 2 0 1 1 0 
0 0 1 4 1 1
 
1 −3 −2 1 0 0
1 1 1
R
2 2
→ R2  0 1 0 2 2
0 
0 0 1 4 1 1
 
1 −3 0 9 2 2
1 1
R1 + 2R3 → R1  0 1 0 0 
 
2 2
0 0 1 4 1 1
21 7
 
1 0 0 2 2
2
1 1
R1 + 3R2 → R1  0 1 0 0 
 
2 2
0 0 1 4 1 1

So A is invertible and
21 7
 
2 2
2
A−1 =  1 1
0 
 
2 2
4 1 1
One can easily check that AA−1 = I.
 
(c) We reduce the augmented matrix A I .

   
1 −6 5 1 0 0 1 −6 5 1 0 0
 −2 R2 + 2R1 → R2
13 −17 0 1 0   0 1 −7 2 1 0 
R3 − 3R1 → R3
3 −19 22 0 0 1 0 −1 7 −3 0 1
 
1 −6 5 1 0 0
R3 + R2 → R3  0 1 −7 2 1 0 
0 0 0 −1 1 1
 
1 −6 5
So an echelon form of A is  0 1 −7 . Since every entry in the third row of
0 0 0
this latter matrix is 0, it is not possible to transform A into the identity matrix.
Therefore A is not invertible, that is, A−1 does not exist.
 
(d) We reduce the augmented matrix A I .

   
0 2 1 1 0 0 0 2 1 1 0 0
− 12 R2
→ R2
 −2 −2 −2 0 1 0   1 1 1 0 − 12 0 
(−1)R3 → R3
−1 0 1 0 0 1 1 0 −1 0 0 −1
 
1 0 −1 0 0 −1
R1 ↔ R3  1 1 1 0 − 12 0 
0 2 1 1 0 0
234 Chapter 4. Matrix Algebra

 
1 0 −1 0 0 −1
R2 − R1 → R2  0 1 2 0 − 12 1 
0 2 1 1 0 0
 
1 0 −1 0 0 −1
R3 − 2R2 → R3  0 1 2 0 − 12 1 
0 0 −3 1 1 −2
 
1 0 −1 0 0 −1
− 31 R3 → R3  0 1 2 0 − 12 1 
1 1 2
0 0 1 −3 −3 3

1 0 0 − 13 − 13 − 13
 
R1 + R3 → R1 2 1
 0 1 0 − 13 
 
R2 − 2R3 → R2 3 6
0 0 1 − 13 − 13 2
3

− 13 − 31 − 13
 
2 1
Thus A−1 =  − 13  .
 
3 6
1
−3 − 31 2
3

 
(e) We reduce the augmented matrix A I .

   
1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0
R2 − 2R1 → R2
 2 1 0 0 0 1 0 0   0 1 0 0 −2 1 0 0 
  R3 − 3R1 → R3  
 3 1 1 0 0 0 1 0   0 1 1 0 −3 0 1 0 
R4 − 4R1 → R4
4 1 1 1 0 0 0 1 0 1 1 1 −4 0 0 1
 
1 0 0 0 1 0 0 0
R3 − R2 → R3  0 1
 0 0 −2 1 0 0 

R4 − R2 → R4  0 0 1 0 −1 −1 1 0 
0 0 1 1 −2 −1 0 1
 
1 0 0 0 1 0 0 0
 0 1 0 0 −2 1 0 0 
R4 − R3 → R4  
 0 0 1 0 −1 −1 1 0 
0 0 0 1 −1 0 −1 1
 
1 0 0 0
 −2 1 0 0 
Thus, A−1 =
 −1 −1
.
1 0 
−1 0 −1 1

8. [Exercise on page 41]

 
a x
(a) Let B =  b y . We want to find a, b, c, x, y, and z such that AB = I2 =
c z
4.2. The Inverse of a Matrix 235

 
1 0
. First, we have
0 1
 
  a x  
2 −1 0  b y = 2a − b 2x − y
AB =
0 1 −1 b−c y−z
c z

So AB = I2 if and only if

2a − b = 1 2x − y = 0
and
b − c =0 y − z =1

We have two systems. The goal here is not to find the set of all solutions, but just
to find one solution. So we don’t need to solve those systems by row reducing
the augmented matrix as usual. The idea is to assign values to some variables
and find other variables.
ˆ Finding a solution to the first system. Let b = 1 (you can assign any value
to b). Then the first equation becomes 2a − 1 = 1. This implies that a = 1.
Substituting b with 1 into the second equation, we get 1 − c = 0, or c = 1.
Thus, if b = 1, then a = 1 and c = 1.
ˆ Finding a solution to the second system. We proceed in the same way. Let
y = 2. Then 2x − 2 = 0 and 2 − z = 1. These imply that x = 1 and z = 1.
Thus, if y = 2, then x = 1 and z = 1.
 
1 1
So B =  1 2 . (Of course, this is not the only solution to the problem.
1 1
Actually, there are infinitely many matrices B satisfying the equation AB = I2 .)
(b)
    
1 1   2 + 0 −1 + 1 0 − 1 2 0 −1
2 −1 0
BA =  1 2  =  2 + 0 −1 + 2 0 − 2  =  2 1 −2 
0 1 −1
1 1 2 + 0 −1 + 1 0 − 1 2 0 −1

We observe that BA 6= I3 .
(c) Suppose that there is a 3 × 2 matrix C such that CA = I3 . Then

CAB = (CA)B = I3 B = B
CAB = C(AB) = CI2 = C

So C = B. And the equation CA = I3 becomes BA = I3 , which contradicts the

fact that BA 6= I3 (from part (b)). Thus, it is not possible to find a 3 × 2 matrix
C such that CA = I3 .

9. [Exercise on page 41]

(a) We want to show that B −1 A−1 is the inverse of AB. Since A and B are both
invertible, AA−1 = I, A−1 A = I, BB −1 = I, and B −1 B = I.
236 Chapter 4. Matrix Algebra

ˆ Using these, we get

(AB)(B −1 A−1 ) = A(BB −1 )A−1 Associativity of the product

= AIA−1 Because BB −1 = I
= AA−1 Because AI = A
=I Because AA−1 = I

ˆ Likewise, (B −1 A−1 )(AB) = B −1 (A−1 A)B = B −1 IB = B −1 B = I.

This proves that AB is invertible and (AB)−1 = B −1 A−1 .
(b) Since A is invertible, AA−1 = I and A−1 A = I. Multiplying AT by (A−1 )T , we
get

AT (A−1 )T = (A−1 A)T Property of the transpose: (XY )T = Y T X T

= IT Because A−1 A = I
=I The transpose of I is I

Similarly, one has (A−1 )T AT = (AA−1 )T = I T = I. This proves that AT is

invertible and (AT )−1 = (A−1 )T .

We just proved two properties of the following theorem.

Properties of Inverses
Theorem 4.2.4. Let A, B be square matrices of the same size.

(a) The identity matrix I is invertible and I −1 = I.

(b) If A is invertible, so is A−1 , and (A−1 )−1 = A.

(c) If A and B are invertible, so is AB, and (AB)−1 = B −1 A−1 .

(d) If A is invertible and α 6= 0 is a scalar, then αA is invertible, and

(αA)−1 = α1 A−1 .

(e) If A is invertible, so is AT , and (AT )−1 = (A−1 )T .

(f ) If A1 , A2 , · · · , Ak are all invertible, so is their product A1 A2 · · · Ak−1 Ak ,

and
(A1 A2 · · · Ak−1 Ak )−1 = A−1 −1 −1 −1
k Ak−1 · · · A2 A1

In other words, the inverse of a product of matrices is the product of their

inverses in the reverse order.

(g) If A is invertible, and k ≥ 1 is an integer, then Ak is invertible and

−1 k
Ak = (A−1 ) .

We proved (c) and (e) in Exercise 9. The proofs of (a), (b), and (d) are similar.
The properties (f) and (g) follow immediately from (c).
4.2. The Inverse of a Matrix 237

10. [Exercise on page 41]

(a) False by Theorem 4.2.4 (a).

(b) True by Theorem 4.2.4 (b).
   
1 0 1 1
(c) False. Indeed, let A = and B = .
1 −1 0 −1
      
1 0 1 1 1 1 −1 2 −1
AB = = and (AB) =
1 −1 0 −1 1 2 −1 1

On the other hand,

     
−1 1 0 −1 1 1 −1 −1 1 1
A = , B = , A B =
1 −1 0 −1 1 2

We see that (AB)−1 6= A−1 B −1 .

(d) True by Theorem 4.2.4 (c).
 
1 0
(e) False. Indeed, let α = 2 and A = I = .
0 1

1
   
2 0 −1 2
0
αA = , (αA) = 1
0 2 0 2

On the other hand,

   
−1 1 0 −1 2 0
A =A= , αA =
0 1 0 2

We see that (αA)−1 6= αA−1 .

(f ) True by Theorem 4.2.4-(e).
(g) True since

(ABC)−1 = (A(BC))−1 = (BC)−1 A−1 = (C −1 B −1 )A−1 = C −1 B −1 A−1

(h) False by Theorem 4.2.3.

(i) True. Indeed, suppose that A is invertible. Then AA−1 = I and A−1 A = I.

AB = 0
A−1 AB = A−1 0 Left-multiply both sides by A−1
IB = 0 A−1 A = I
B=0 IB = B

11. [Exercise on page 42]

238 Chapter 4. Matrix Algebra

(a)

AX = C
A−1 AX = A−1 C Left-multiply both sides by A−1
IX = A−1 C A−1 A = I since A is invertible
X = A−1 C IX=X

(b)

XB = C
XBB −1 = CB −1 Right-multiply both sides by B −1
XI = CB −1 BB −1 = I since B is invertible
X = CB −1 XI = X

(c)

AXB −1 = C
A−1 AXB −1 = A−1 C Left-multiply both sides by A−1
IXB −1 = A−1 C A−1 A = I since A is invertible
XB −1 = A−1 C IX = X
XB −1 B = A−1 CB Right-multiply both sides by B
XI = A−1 CB B −1 B = I
X = A−1 CB XI = X

(d)

A−1 X T B = C T
Left-multiply both sides by A and
AA−1 X T BB −1 = AC T B −1
Right-multiply both sides by B −1
IX T I = AC T B −1 AA−1 = I and BB −1 = I
X T = AC T B −1

T
(X T )T = AC T B −1 Take the transpose of both sides
X = (B −1 )T (C T )T AT Use the property (X T )T = X
X = (B −1 )T CAT Use the property (C T )T = C

(e)

A−1 (3X + In )T B −1 = CB
4.2. The Inverse of a Matrix 239

Left-multiply both sides by A and

AA−1 (3X + In )T B −1 B = ACBB
Right-multiply both sides by B
I(3X + In )T I = ACB 2 AA−1 = I and B −1 B = I
(3X + In )T = ACB 2

T
(3X + In )T = (ACB 2 )T Take the transpose of both sides
Use the properties ((A)T )T = A
3X + In = (B 2 )T C T AT
and (ABC)T = C T B T AT
3X = (B 2 )T C T AT − In Subtract In from both sides
1 1
X= (B 2 )T C T AT − In Divide both sides by 3
3 3
(f )
A(5X T − B)C = AX T C
Left-multiply both sides by A−1
−1 T −1 −1 T −1
A A(5X − B)CC = A AX CC and Right-multiply both sides by
C −1
I(5X T − B)I = IX T I A−1 A = I and CC −1 = I
5X T − B = X T

4X T = B Subtract X T from both sides and

add B to both sides
1
XT = B Divide both sides by 4
4
 T
1
(X T )T = B Take the transpose of both sides
4
1 Use the properties (X T )T = X
X = BT
4 and (kB)T = kB T
12. [Exercise on page 42]

  −1  
T 0 2 1 1 −1
2X + =
−8 4 6 0 1
"  −1 −1  
# −1
T 0 2 1 1 −1
2X + = Take the inverse of both sides
−8 4 6 0 1
    −1
0 2 1 1 −1
T
2X + = Use the property (A−1 )−1 = A
−8 4 6 0 1
   −1  −1
T 0 2 1 1 −1 Use the property
2X + =
−8 4 6 0 1 (kA)−1 = k −1 A−1
   −1
0 2 1 −1
2X T + =6
−8 4 0 1
240 Chapter 4. Matrix Algebra

   
1 −1 1 −1
Since the determinant of is 1 − 0 = 1, it follows that the matrix
0 1 0 1
is invertible and  −1    
1 −1 1 1 1 1 1
= =
0 1 1 0 1 0 1

Therefore, the equation above becomes

   
T 0 2 1 1
2X + =6
−8 4 0 1
   
T 0 2 6 6
2X + =
−8 4 0 6
   
T 6 6 0 2
2X = −
0 6 −8 4
 
T 6 4
2X =
8 2
 
T 3 2
X = Divide both sides by 2
4 1
 T
T T 3 2
(X ) = Take the transpose of both sides
4 1
 
3 4
X= Use the property (AT )T = A
2 1

13. [Exercise on page 42]

(a) We want to show that A − C is invertible.

XA = B + XC
XA − XC = B Subtract XC from both sides
X(A − C) = B Left-factor out X
A − C = X −1 B Left-multiply both sides by X −1

This shows that A − C = X −1 B. Since X is invertible (by hypothesis), it follows

that X −1 is invertible thanks to Theorem 4.2.4-(b). Moreover, B is invertible by
hypothesis. Since the product of two invertible matrices is again an invertible
matrix (by Theorem 4.2.4-(c)), it follows that X −1 B is invertible, that is, A − C
is invertible.
(b)

XA = B + XC
4.2. The Inverse of a Matrix 241

XA − XC = B Subtract XC from both sides

X(A − C) = B Left-factor out X
X = B(A − C)−1 Right-multiply both sides by (A − C)−1

14. [Exercise on page 42]

BAX T = C + 2AX T
BAX T − 2AX T = C Subtract 2AX T from both sides
(BA − 2A)X T = C Right-factor out X T

The next thing we want to do is to left-multiply both sides of the latter equation by
(BA − 2A)−1 . But this requires BA − 2A to be invertible. So, before we continue, we
need to explain why BA − 2A is invertible. Consider the equation (BA − 2A)X T = C
above. Since X is invertible by hypothesis, it follows (by Theorem 4.2.4-(e)) that X T
is invertible. So we can right-multiply both sides of the equation (BA−2A)X T = C by
(X T )−1 . We obtain BA − 2A = C(X T )−1 . Since both C and (X T )−1 are invertible, it
follows that the product C(X T )−1 is invetible. This implies that BA−2A is invertible.
We come back to the equation.

BAX T = C + 2AX T
BAX T − 2AX T = C Subtract 2AX T from both sides
(BA − 2A)X T = C Right-factor out X T
Left-multiply both sides by
X T = (BA − 2A)−1 C
(BA − 2A)−1

T
(X T )T = (BA − 2A)−1 C Take the transpose of both sides

T Use the properties (AT )T =
X = C T (BA − 2A)−1
A and (AB)T = B T AT

15. [Exercise on page 42]

(XA)−1 (A + X) = B
Use the property
A−1 X −1 (A + X) = B
(AB)−1 = B −1 A−1
A−1 X −1 A + A−1 X −1 X = B
A−1 X −1 A + A−1 I = B Use the property X −1 X = I
A−1 X −1 A + A−1 = B Use the property A−1 I = A−1
A−1 X −1 A = B − A−1 Subtract A−1 from both sides
A A−1 X −1 A = A(B − A−1 )


Left multiply both sides by A
242 Chapter 4. Matrix Algebra

A A−1 X −1 A A−1 = A(B − A−1 )A−1 Right multiply both sides by A−1

IX −1 I = A(B − A−1 )A−1

X −1 = A(B − A−1 )A−1
−1
(X −1 )−1 = A(B − A−1 )A−1

Take the inverse of both sides
Use the properties (A−1 )−1 = A
X = (A−1 )−1 (B − A−1 )−1 A−1
and (ABC)−1 = C −1 B −1 A−1
X = A(B − A−1 )−1 A−1

ˆ Here we assumed that the matrix B − A−1 is invertible.

ˆ Proving that B − A−1 is invertible. The sixth line of what we did above says
that A−1 X −1 A = B − A−1 . So proving that B − A−1 is invertible amounts to
proving that the matrix A−1 X −1 A is invertible, which is the case since it is a
product of invertible matrices.

16. [Exercise on page 42]

The Invertible Matrix Theorem

Theorem 4.2.5. Let A be an n × n matrix. The following statements are
equivalent.

IMT1. A is invertible.

IMT2. The reduced echelon form of A is the n × n identity matrix I.

IMT3. A has n pivot positions.

IMT4. The homogeneous equation Ax = 0 has a unique solution (the trivial

solution).

IMT5. The columns of A are linearly independent.

IMT6. The linear transformation T : Rn → Rn defined by T (x) = Ax is

one-to-one.

IMT7. The equation Ax = b has a unique solution for every b in Rn .

IMT8. The columns of A span Rn .

IMT9. The linear transformation T : Rn → Rn defined by T (x) = Ax is

onto.

IMT10. There is an n × n matrix B such that BA = I.

IMT11. There is an n × n matrix C such that AC = I.

IMT12. AT is invertible.

Definition and Notation. Let P and Q be two statements.

4.2. The Inverse of a Matrix 243

ˆ We say that P implies Q, and we write P =⇒ Q, if whenever P is

true, Q is also true.

ˆ We say that P is equivalent to Q, and we write P ⇐⇒ Q, if P is true

if and only if Q is true. (Or, equivalently, P is false if and only if Q is
false.) In other words,

P ⇐⇒ Q if (P =⇒ Q and Q =⇒ P )

(a) True. This is the equivalence IMT1 ⇐⇒ IMT2 from Theorem 4.2.5.
 
1 0
(b) False. Indeed, let A = 0 1. Clearly, the system Ax = 0 has only the trivial
0 0
solution. But there is not a pivot position in every row as the third row does
not have a pivot.
 
1 0
(c) False. The matrix A = 0 1 has a pivot in every column, but A is not
0 0
invertible as it is not even a square matrix. (Remember that only square matrices
have inverses.)
(d) True. Indeed, suppose A is a square n × n matrix and that there is a pivot
position in every column. Then there is a pivot position in every row. There-
fore, the statement IMT3 from Theorem 4.2.5 is true. Using the equivalence
IMT3 ⇐⇒ IMT1 in Theorem 4.2.5, it follows that A is invertible.
(e) False. Indeed, let
 
  1 0
1 0 0
C= and A = 0 1
0 1 0
0 0


1 0
We have CA = = I. But A is not invertible as it is not even a square
0 1
matrix.
(f ) True. The sentence “A and C are square matrices such that CA = I” is precisely
the statement IMT10 from Theorem 4.2.5. Using the equivalence IMT10 ⇐⇒
IMT1 in Theorem 4.2.5, we can conclude that A is invertible.
 
1 0
(g) False. Indeed, let A = 0 1 and let T : R2 → R3 be the linear transformation
0 0
defined by T (x) = Ax. Since every column of A is a pivot column, T is one-
to-one. However, the matrix A is not invertible since it is not even a square
matrix.
(h) True. The sentence “an n×n matrix A is invertible” is the statement IMT1 from
Theorem 4.2.5. Using the equivalence IMT1 ⇐⇒ IMT8 from Theorem 4.2.5,
it follows that the columns of A span Rn .
244 Chapter 4. Matrix Algebra

(i) True. This is the equivalence IMT6 ⇐⇒ IMT9 from Thereom 4.2.5.
(j) False because the equivalence IMT1 ⇐⇒ IMT12 from Theorem 4.2.5 says that
A is invertible if and only if AT is invertible.
(k) True. Indeed, suppose A and B are square matrices such that AB = I. Then,
by the equivalence IMT11 ⇐⇒ IMT1 from Theorem 4.2.5, A is invertible. So
there exists a matrix X such that AX = I and XA = I. Since AB = I, it
follows that AB = AX. Left-multiplying both sides of this latter equation by
A−1 , we get B = X. So AB = I and BA = I, which implies that AB = BA.
This proves that A and B commute.
(l) False because the equivalence IMT1 ⇐⇒ IMT8 from Theorem 4.2.5 says that
an n × n matrix A is invertible if and only if its columns span Rn .
(m) True. Indeed, the sentence “A is a square matrix whose columns are not linearly
independent” means that the statement IMT5 from Theorem 4.2.5 is false. Since
IMT5 is equivalent to IMT1 and since IMT5 is false, it follows that IMT1 is also
false. So A is not invertible.
(n) True. This follows from the equivalence IMT6 ⇐⇒ IMT9 from Theorem 4.2.5.

17. [Exercise on page 43]

Since AB is invertible, there is a matrix X such that (AB)X = I and X(AB) = I.
Since the product is associative, it follows that (XA)B = I. Using the equivalence
IMT10 ⇐⇒ IMT1 from Theorem 4.2.5, we can conclude that B is invertible.

18. [Exercise on page 43]

Since A has two identical columns, the columns of A are not linearly independent.
Therefore, by the equivalence IMT1 ⇐⇒ IMT5 from Theorem 4.2.5, A is not
invertible.

19. [Exercise on page 43]

Since A is invertible, A−1 is invertible by the property (b) of Theorem 4.2.4. Using the
equivalence IMT1 ⇐⇒ IMT5 from Theorem 4.2.5, the columns of A−1 are linearly
independent.

20. [Exercise on page 43]

Since A is a square matrix whose columns span Rn , it follows that A is invertible
by the equivalence IMT8 ⇐⇒ IMT1 from Theorem 4.2.5. This implies that A3 is
invertible as well by the property (g) of Theorem 4.2.4. Using now the equivalence
IMT1 ⇐⇒ IMT5 from Theorem 4.2.5, we can conclude that the columns of A3 are
linearly independent (L.I). We can summarize this as follows:

The columns of A span Rn =⇒ A is invertible

=⇒ A3 is invertible
=⇒ The columns of A3 are L. I.
 Chapter5

Subspaces of Rn

5.1 Subspaces, Column and Null Spaces

Definition of Subspace
Let H be a subset of Rn . We say that H is a subspace of Rn if the following three
conditions are satisfied:

(1) H contains the zero vector.

(2) For every vectors u, v in H, the sum u + v is also in H. (We say that H is
closed under addition.)

(3) For every u in H, for every scalar c, the vector cu is in H. (We say that H is
closed under scalar multiplication.)

Example of subspaces of Rn include {0} and Rn itself. The subspace {0} is called
the zero subspace of Rn .

1. [Exercise on page 45]

 
x1
 x2 
n
(a) False. Indeed, consider the subset H of R consisting of vectors x =  .. 
 
 . 
xn
such that
x1 + x2 + · · · + xn = 1
The zero vector of Rn is not in H since 0 + 0 + · · · + 0 = 0 6= 1. So the first
condition of the definition of subspace is not satisfied. Therefore, H is not a
subspace of Rn .
(b) True. By definition, a subspace of Rn is a subset of Rn satisfying certain
conditions.
(c) False. Let H be the set of points (x1 , x2 ) in R2 such that x2 = x21 . This set
0
contains the zero vector because 0 = 02 . But it is not a subspace of R2 as
0

245
246 Chapter 5. Subspaces of Rn

   
1 2
it is not closed under addition. Indeed, let u = and v = be vectors
  1 4
3
in H. Then the sum u + v = is not in H since 5 6= 32 .
5
(d) True by definition.
(e) False because the line of equation x2 = 2x1 + 3 is not a suspace of R2 as it does
not pass through the origin (since 0 6= 2(0) + 3).
(f ) False by definition.
(g) False by definition.
(h) False because any line in R2 through the origin is a subspace of R2 . Indeed,
the equation of a line in R2 is of the form x2 = mx1 , where m is a real number.
Let m be a real number and let H be the set of points (x1 , x2 ) in R2 such that
x2 = mx1 . We want to prove that H is a subspace of R2 , which amounts to
proving that H satisfies conditions (1), (2), and (3) of the definition of subspace.
ˆ Condition (1) is satisfied because 0 = m(0).
   
u1 v1
ˆ Checking condition (2). Let u = and v = be vectors in H.
u2 v2
Then
u2 = mu1 and v2 = mv1
 
u1 + v1
These imply that u2 + v2 = m(u1 + v1 ). So the sum u + v = is
u2 + v2
in H.  
u1
ˆ Checking condition (3). Let u = be a vector in H and let c be a
u2
scalar. Then u2 = mu1 . Multiplying both sides of thislatter equation by c,
cu1
we get u2 = m(cu1 ). This shows that the vector cu = is in H.
cu2
Thus, any line in R2 through the origin is a subspace of R2 .

Subspaces of R2 and R3

Theorem 5.1.1. The subspaces of R2 are

ˆ {0},

ˆ any line through the origin, and

ˆ R2

Theorem 5.1.2. The subspaces of R3 are

ˆ {0},

ˆ any line through the origin,

ˆ any plane through the origin,

ˆ R3
5.1. Subspaces, Column and Null Spaces 247

   
1 0
(i) False because the vector u = is on the x1 -axis and the vector v =
0
  1
1
is on the x2 -axis, but the sum u + v = is neither on the x1 -axis nor on the
1
x2 -axis.

2. [Exercise on page 45]

 
x1
(a) A vector is in H if x1 − 3x2 = 0.
x2
ˆ Let x2 = 1. Substituting this into the equation x1 −3x  2 , we get x1 −3(1) = 0,
3
which implies that x1 = 3. So the vector u1 = is in H.
1
 
6
ˆ If x2 = 2, x1 = 3(2) = 6. So the vector u2 = is in H.
2
 
x1
(b) It suffices to find a vector such that x1 − 3x2 6= 0. For example, the
  x 2
1
vector is not in H because 1 − 3(0) = 1 6= 0.
0
 
0
(c) ˆ The zero vector in R2 , , is in H because 0 − 3(0) = 0.
0
   
x1 y1
ˆ Let u = ,v = be vectors in R2 . Then
x2 y2

x1 − 3x2 = 0 and y1 − 3y2 = 0.

 
x1 + y 1
The vector u + v is . Since
x2 + y 2

(x1 + y1 ) − 3(x2 + y2 ) =x1 + y1 − 3x2 − 3y2

=(x1 − 3x2 ) + (y1 − 3y2 )
=0 + 0 = 0

the vector u +v is in H.

x1
ˆ Let u = be a vector in H, and let c be scalar. Then x1 − 3x2 = 0.
x2
Multiplying both sides
 by c, we get cx1 − 3(cx2 ) = 0. This shows that the
cx1
vector cu = is in H.
cx2
Thus, H is a subspace of R2 .

3. [Exercise on page 45]

(a) To get a vector in H, it suffices to assign a value to a and a value to b.

248 Chapter 5. Subspaces of Rn

ˆ Let a = 0 and b = 0. Then the vector

     
a − 2b 0 − 2(0) 0
 a  =  0  = 0 

2a + 3b 2(0) + 3(0) 0
is in H.
ˆ Let a = 0 and b = 1. Then the vector
     
a − 2b 0 − 2(1) −2
 a = 0 = 0 
2a + 3b 2(0) + 3(1) 3
is in H.
ˆ Let a = 1 and b = 0. Then the vector
     
a − 2b 1 − 2(0) 1
 a = 1 = 1 
2a + 3b 2(1) + 3(0) 2
is in H.
ˆ Let a = 1 and b = 1. Then the vector
     
a − 2b 1 − 2(1) −1
 a = 1 = 1 
2a + 3b 2(1) + 3(1) 5
is in H. 
2
(b) By inspection, the vector u =  0  is not in H because if u is of the form
  0
a − 2b
 a , then
2a + 3b
a − 2b = 2, a = 0, and 2a + 3b = 0.
The first two equations imply that a = 0 and b = −1. But the third equation is
not satisfied because 2a + 3b = 2(0) + 3(−1) = −3 6= 0.
(c) ˆ As we saw in part (a), the zero vector is in H.
   
a1 − 2b1 a2 − 2b2
ˆ Let u1 =  a1  , u2 =  a2  be vectors in H. We want to
2a1 + 3b1 2a2 + 3b2
show that the sum u1 + u2 is in H. We have
   
(a1 − 2b1 ) + (a2 − 2b2 ) (a1 + a2 ) − 2(b1 + b2 )
u1 + u2 =  a1 + a2 = a1 + a2 
(2a1 + 3b1 ) + (2a2 + 3b2 ) 2(a1 + a2 ) + 3(b1 + b2 )
 
a − 2b
=  a 
2a + 3b
where a = a1 + a2 and b = b1 + b2 . This shows that u1 + u2 is in H.
5.1. Subspaces, Column and Null Spaces 249

 
a − 2b
ˆ Let u =  a  be a vector in H, and let c be a scalar. Then
2a + 3b
   
ca − 2cb α − 2β
cu =  ca = α ,
2ca + 3cb 2α + 3β

where α = ca and β = cb. This shows that cu is in H.

Hence, H is a subspace of R3 .

4. [Exercise on page 46]

2
(a) The
 subset H is not a subspace of R because it does not contain the zero vector
0
. Indeed, 0 + 0 = 0 6= 1.
0
(b) This is not a subspace of R2 because the third
 condition of the definition
 of
 a
1 −2
subspace is not satisfied as the vector u = is in H, but −2u = is
1 −2
not in H.

5. [Exercise on page 46]

Let H = Span{v1 , · · · , vp }. A vector is in H if it can be written as a linear combina-
tion of v1 , · · · , vp . We want to show that H is a subspace of Rn .

ˆ The zero vector of Rn is in H because 0 = 0v1 + · · · + 0vp .

ˆ Let u be a vector in H. Then there exist scalars c1 , · · · , cp such that u =
c1 v1 + · · · cp vp . Let v be a vector in H. Then there exist scalars d1 , · · · , dp such
that v = d1 v1 + · · · dp vp . The sum u + v is

u + v = (c1 v1 + · · · cp vp ) + (d1 v1 + · · · dp vp )
= (c1 + d1 )v1 + · · · + (cp + dp )vp = α1 v1 + · · · + αp vp ,

where αi = ci + di , 1 ≤ i ≤ p. This shows that u + v is in H.

ˆ Let u = c1 v1 + · · · cp vp be a vector in H, and let d be a scalar. Then

du = dc1 v1 + · · · dcp vp = β1 v1 + · · · + βp vp ,

where βi = dci , 1 ≤ i ≤ p. This shows that du is in H.

Hence, Span{v1 , · · · , vp } is a subspace of Rn . This proves the following theorem.

Spanning Sets are Subspaces

Theorem 5.1.3. Let v1 , · · · , vp be vectors in Rn . Then Span{v1 , · · · , vp } is a

subspace of Rn .

6. [Exercise on page 46]

250 Chapter 5. Subspaces of Rn

(a) The subset H is not a subspace of R2 since it does not contain the zero vector.
(b) Here H is a line through the origin. So, by Theorem 5.1.1, it is a subspace of
R2 .

Alternate Solution
Let u be a nonzero vector in H. Then, since H is a line through the origin,
every vector in H is a multiple of u. This means that H = Span{u}. So,
by Theorem 5.1.3, H is a subspace of R2 .

(c) This is not a subspace of R2 because it is not closed under scalar multiplication.
Indeed, if we multiply a nonzero vector in H by a negative number, we get vector
that does not lie in H.
(d) For the same reason as in part (c), H is not a subspace of R2 .
(e) From the graph, we see that the subset H does not contain the origin. So it is
not a subspace of R2 .

7. [Exercise on page 47]

3
H 2
2
S
1

−2 −1 1 2
1 2 3

ˆ The subset H is not a subspace of R2 because it does not contain the origin.
ˆ The subset S is not a subspace either
 since
 it is not closed under
 scalar
 mul-
0.5 2
tiplication. Indeed, the vector u = is in S, but 4u = is not in
0 0
S.

8. [Exercise on page 48]

Column Spaces
 
Let a1 , a2 , · · · , an be vectors in Rm . Let A = a1 a2 · · · an be the matrix
whose columns are a1 , a2 , · · · , an . The column space of A, denoted Col(A),
is the set of all linear combinations of the columns of A, that is,

Col(A) = Span{a1 , a2 , · · · , an }

Theorem 5.1.4. The column space of an m × n matrix A is a subspace of Rm .

Proof. This follows immediately from the definition and Theorem 5.1.3.
5.1. Subspaces, Column and Null Spaces 251

Theorem 5.1.5. Let A be an m × n matrix and let b be a vector in Rm . Then

the linear system Ax = b is consistent if and only if b is in Col(A).
 
Proof. Let A = a1 a2 · · · an , where a1 , a2 , · · · , an are the columns of A.

ˆ Suppose
 that
 the system Ax = b is consistent. Then there exists a vector
x1
 x2 
x =  ..  in Rn such that Ax = b. By definition of the multiplication
 
 . 
xn
of a matrix by a vector, we have
 
x1
  x2 
Ax = a1 a2 · · · an   = x 1 a1 + x 2 a2 + · · · + x n an
 
..
 . 
xn

So the equation Ax = b becomes

x 1 a1 + x 2 a2 + · · · + x n an = b

This latter equation shows that b is a linear combination of the columns

of A, which means that b is in Col(A).

ˆ Conversely, suppose that b is in Col(A). Then, by definition of the column

space, b is a linear combination of the columns of A. This implies that
there exist scalars c1 , c2 , · · · , cn such that b = c1 a1 + c2 a2 + · · · + cn an .
But  
c1
 c2 
 
c1 a1 + c2 a2 + · · · cn an = a1 a2 · · · an  ..  = Ac,
 . 
cn
 
c1
 c2 
where c =  .. . So the equation c1 a1 + c2 a2 + · · · cn an = b becomes
 
 . 
cn
Ac = b, which means that the vector c is a solution to the equation
Ax = b. Thus, the system Ax = b is consistent.

This ends the proof.

(a) True by definition.

(b) False. By definition the column space of an m × n matrix is a subspace of Rm .
(c) True by Theorem 5.1.4.
252 Chapter 5. Subspaces of Rn

(d) False. The column space of a 3 × 2 matrix is a subspace of R3 , so it can never

be R2 .
 
1 1
(e) False. Indeed, let A = 0 0 and let a1 and a2 be the columns of A. Since
0 0
a1 = a2 , it follows that a1 and a2 are linearly dependent. So Span{a1 , a2 } =
Span{a1 }. But Span{a1 } is a line in R3 (not a plane).
(f ) False. Indeed, suppose the columns a1 , a2 , and a3 of a 3 × 3 matrix are linearly
dependent. Then one of the vectors in the set {a1 , a2 , a3 } is a linear combination
of the others. Without loss of generality, suppose that a3 is a linear combination
of a1 and a2 . Then
Span{a1 , a2 , a3 } = Span{a1 , a2 }
6 R3 since two vectors can’t span R3 .
But Span{a1 , a2 } =
(g) True. Let A be a 3 × 3 matrix. Suppose that the columns of A are linearly
independent. Since Col(A) is a subset of R3 , to prove that Col(A) = R3 , all
we have to do is to prove that every vector in R3 is also in Col(A). Let b be a
vector in R3 . Consider the system Ax = b. Since the columns of A are linearly
independent, it follows that every column of A has a pivot position. This implies
that every row of A also has a pivot position (since A is a square matrix). So,
if B is an echelon form of A, the system Bx = b has no equation of the form
0 = d with d 6= 0. This means that the system Bx = b is consistent. Since
the systems Ax = b and Bx = b have the same solution set, it follows that
 the
c1
equation Ax = b is consistent. Therefore, there exists a vector c =  c2  in R3
c3
such that Ac = b. But, if the columns of A are a1 , a2 , a3 ,
 
  c1
Ax = a1 a2 a3  c2  = c1 a1 + c2 a2 + c3 a3
c3

So the equation Ac = b becomes c1 a1 + c2 a2 + c3 a3 = b, which shows that b is a

linear combination of the columns of A. This means that b is in Col(A).

Alternate Solution (using the Invertible Matrix Theorem)

Let b be a vector in R3 . Since the columns of A are linearly indepen-

dent, and since A is a square matrix, it follows (by the Invertible Matrix
Theorem) that the system Ax = b is consistent. This implies that b is in
Col(A).

 
1 0
(h) False. Indeed, let A = 0 1 be an m × n matrix with m = 3 and n = 2.
0 0
The columns of A are linearly independent (since every column has a pivot), but
they do not span R3 since two vectors cannot span R3 .
5.1. Subspaces, Column and Null Spaces 253

(i) True by Theorem 5.1.5.

(j) False. Indeed, if b does not lie on Col(A), then the system Ax = b has no
solution thanks to Theorem 5.1.5.

9. [Exercise on page 48]

(a) We want to determine if there exist scalars c1 and c2 such that c1 v1 + c2 v2 = u.
This amounts to determining if the following system is consistent.
c1 + 2c2 = 7
−c1 − 3c2 = −9
−2c1 + 4c2 = 2
We need to reduce the augmented matrix.
   
1 2 7 1 2 7
 −1 −3 −9  R2 + R1 → R2  0 −1 −2 
R3 + 2R1 → R3
−2 4 2 0 8 16
 
1 2 7
(−1)R2 → R2  0 1 2
1

R → R3
8 3 0 1 2
 
1 2 7
R3 − R2 → R3  0 1 2 
0 0 0
 
1 0 3
R1 − 2R2 → R1  0 1 2 
0 0 0

The corresponding system is c1 = 3 and c2 = 2. So u = 3v1 + 2v2 . This shows

that u is in the subspace of R3 generated by v1 and v2 .
(b) From part (a), u is a linear combination of the columns of A. So u is in Col(A).
10. [Exercise on page 48]
Let ai be the ith column of A. We need to determine if there exist scalars c1 , c2 , and
c3 such that c1 a1 + c2 a2 + c3 a3 = b. This amounts to determining if the following
system is consistent.
c1 − 2c2 + 3c3 = −4
−c1 + 3c2 + c3 = −1
2c1 − 3c2 + 10c3 = −12
As usual, we need to reduce the augmented matrix.
   
1 −2 3 −4 1 −2 3 −4
 −1 R2 + R1 → R2
3 1 −1   0 1 4 −5 
R3 − 2R1 → R3
2 −3 10 −12 0 1 4 −4
 
1 −2 3 −4
R3 − R2 → R3  0 1 4 −5 
0 0 0 1
254 Chapter 5. Subspaces of Rn

The third equation of the corresponding system, 0 = 1, is inconsistent. So the system

has no solution, and therefore b is not in Col(A).
 
1 4 −5
11. Let A =  1 5 1 . [Exercise on page 48]
5 23 −7
     
1 4 −5
(a) Let a1 =  1  , a2 =  5  , and a3 =  1  be the columns of A. By
5 23 −7
definition, Col(A) is the set of all linear combinations of the vectors a1 , a2 , and
a3 . So a vector u in Col(A) is of the form

u = c1 a1 + c2 a2 + c3 a3

where c1 , c2 , and c3 are real numbers.

 
0
ˆ Let c1 = c2 = c3 = 0. Then the vector 0a1 + 0a2 + 0a3 =  0  is in Col(A).
0
 
1
ˆ Let c1 = 1, c2 = 0, c3 = 0. Then the vector 1a1 + 0a2 + 0a3 = a1 = 1  is

5
in Col(A).
 
4
ˆ Let c1 = 0, c2 = 1, c3 = 0. Then the vector 0a1 + 1a2 + 0a3 = a2 =  5 
23
is in Col(A).
 
−5
ˆ Let c1 = 0, c2 = 0, c3 = 1. Then the vector 0a1 + 0a2 + 1a3 = a3 =  1 
−7
is in Col(A).
 
5
ˆ Let c1 = 1, c2 = 1, c3 = 0. Then the vector 1a1 +1a2 +0a3 = a1 +a2 =  6 
28
is in Col(A).
 
b1
(b) Let b =  b2  be a vector in R3 , and consider the system Ax = b. Row reducing
b3
the augmented matrix, we have
   
1 4 −5 b1 1 4 −5 b1
R2 − R1 → R2
 1 5 1 b2   0 1 6 b 2 − b1 
R3 − 5R1 → R3
5 23 −7 b3 0 3 18 b3 − 5b1
 
1 4 −5 b1
R3 − 3R2 → R3  0 1 6 b 2 − b1 
0 0 0 −2b1 − 3b2 + b3
5.1. Subspaces, Column and Null Spaces 255

So the system Ax = b is consistent if and only if

−2b1 − 3b2 + b3 = 0

This means that b is not in Col(A) if and only if −2b1 − 3b2 + b3 6= 0.

ˆ For b1 = 1, b2 = 0, b3 = 0, the equation −2b1 −  3b2+ b3 = 0 is not satisfied
1
since −2(1) + 0 + 0 = −2 6= 0. So the vector 0  is not in Col(A).

0
 
0
ˆ Similarly, the vector 1  is not in Col(A).

0
 
10
(c) The vector b =  11  is not in Col(A) since it does not satisfy the equation
12
−2b1 − 3b2 + b3 = 0.
 
b1
(d) From part (b), a vector b =  b2  is in Col(A) if and only if −2b1 −3b2 +b3 = 0.
b3
Solving this latter equation for b3 , we get b3 = 2b1 + 3b2 . So b is in Col(A) if and
only if
       
b1 b1 1 0
b = b2 =
   b2  = b1 0 + b2 1 
  
b3 2b1 + 3b2 2 3
 
 
1 0
3
Thus, Col(A) is the plane in R spanned by the vectors 0 and 1 .
  
2 3

12. [Exercise on page 49]

(a) First, recall the definition of the range. The range of T : Rn → Rm is the set
of all vectors b in Rm for which the equation Ax = b is consistent. Let b be a
vector in Rm . The following statements are equivalent.
(i) The vector b is in Col(A).
(ii) The linear system Ax = b is consistent.
(iii) The vector b is in the range of T .
The statement (i) is equivalent to (ii) by Theorem 5.1.5, while (ii) is equivalent
to (iii) by the definition of the range.
Since (i) is equivalent to (ii) and (ii) is equivalent to (iii), it follows that (i) is
equivalent to (iii). This shows that a vector b is in the column space of A if and
only if b is in the range of T . So Col(A) equals the range of T .
256 Chapter 5. Subspaces of Rn

Alternate Solution
The following alternate solution does not use Theorem 5.1.5. Let
a1 , a2 , · · · , an be the columns of A.

Let Range(T ) denote the range of T . To prove that Col(A) = Range(T ),

we need to prove that Col(A) is a subset of Range(T ) and that Range(T )
is a subset of Col(A).

ˆ Proving that Col(A) is a subset of Range(T ). Let b be a vector in

Col(A). We want to prove that b is in the range of T . The vector
b is in Col(A) means that there exist scalars c1 , c2 , · · · , cn such that
c1 a1 + c2 a2 + · · · + cn an = b. This implies that the system Ax = b
has a solution. Since T (x) = Ax, the equation T (x) = b also has a
solution, which means that b is in the range of T .

ˆ Proving that Range(T ) is a subset of Col(A). Let b be a vector in

Range(T ). Then the equation T (x) = b is consistent, and therefore
the equation Ax = b is also consistent (since T (x) = Ax). This
implies that there exist scalars c1 , c2 , · · · , cn such that

c 1 a1 + c 2 a2 + · · · + c n an = b

So b is in Col(A).

This proves that Col(A) = Range(T ).

(b) The following statements are equivalent.

(i) Col(A) = Rm .
(ii) Every vector b in Rm is in Col(A).
(iii) For every vector b in Rm , the system Ax = b is consistent.
The equivalence (i) ⇐⇒ (ii) is obvious. And the equivalence (ii) ⇐⇒ (iii)
follows immediately from Theorem 5.1.5. So (i) is equivalent to (iii).

Alternate Solution
The following alternate solution does not use Theorem 5.1.5. Let
a1 , a2 , · · · , an be the columns of A.

ˆ Suppose Col(A) = Rm . Let b be a vector in Rm . Then b is in

Col(A). This implies that there exist scalars c1 , c2 , · · · , cn such that
c1 a1 + · · · , cn an = b, which in turn implies that the system Ax = b
is consistent.

ˆ Conversely, suppose that for every vector b in Rm the system Ax = b

is consistent. Since Col(A) is a subset of Rm , to prove that Col(A) =
Rm , all we have to do is to prove that Rm is a subset of Col(A). Let
b be a vector in Rm . Then the system Ax = b is consistent by
5.1. Subspaces, Column and Null Spaces 257

hypothesis. So there exists a vector c such that Ac = b. Since Ac

is a linear combination of the columns of A (by definition of the
product of a matrix by a vector), it follows that b is in Col(A). So
Col(A) = Rm .

Thus, Col(A) = Rm if and only if for every vector b in Rm the system

Ax = b is consistent.

(c) The following statements are equivalent.

(i) Col(A) = Rm .
(ii) For every vector b in Rm , the system Ax = b is consistent.
(iii) The linear transformation T (x) = Ax is onto.
The equivalence (i) ⇐⇒ (ii) follows from part (b), while (ii) ⇐⇒ (iii) follows
immediately from the definitions.
(d) Using the equivalence (i) ⇐⇒ (ii) from part (c) and the invertible matrix
theorem, we can conclude that if A is an n × n matrix, Col(A) = Rn if and only
if the matrix A is invertible.
(e) Since every column of AB is a linear combination of the columns of A, it follows
that every column of AB is in Col(A).
 
1 0
(f ) False. Indeed, let A = 0 1. The column space of A is spanned by two
0 0
vectors (the columns of A) in R3 . Since two vectors can’t span R3 , it follows
that Col(A) 6= R3 . However, the columns of A are not linearly dependent since
every column is a pivot column.
 
1 1
(g) False. Indeed, let A = 0 0. The column space of A is spanned by two
0 0
vectors (the columns of A) in R3 . Since two vectors can’t span R3 , it follows
that Col(A) 6= R3 . However, the columns of A are not linearly independent since
there is a non-pivot row (the second row for example has no pivot).
 
1 2 4
(h) Suppose A =  3 7 7 .
−4 −9 −11
Let a1 , a2 , and a3 be the columns of A. Let b be a vector in Col(A), then by
definition there exist scalars x1 , x2 , and x3 such that x1 a1 + x2 a2 + x3 a3 = b.
This gives the following system.
x1 + 2x2 + 4x3 = b1
3x1 + 7x2 + 7x3 = b2
−4x1 − 9x2 − 11x3 = b3
To solve this, we need to reduce the augmented matrix as usual.
   
1 2 4 b1 1 2 4 b1
R2 − 3R1 → R2
 3 7 7 b2   0 1 −5 b2 − 3b1 
R3 + 4R1 → R3
−4 −9 −11 b3 0 −1 5 b3 + 4b1
258 Chapter 5. Subspaces of Rn

 
1 2 4 b1
R3 + R2 → R3  0 1 −5 b2 − 3b1 
0 0 0 b1 + b2 + b3

So the equation Ax = b is consistent if and only if b1 + b2 + b3 = 0. Solving this

for b1 , we get b1 = −b2 − b3 . So
       
b1 −b2 − b3 −1 −1
b =  b2  =  b2  = b2  1  + b3  0 
b3 b3 0 1

Thus, a vector
 b is in Col(A)if and
 only if b is a linear combination of the
−1 −1
vectors u =  1  and v =  0 . This means that Col(A) is the plane in
0 1
3
R spanned by the vectors u and v.

13. [Exercise on page 49]

Null Space of a Matrix

Let A be an m × n matrix. The null space of A, denoted Null(A), is the set

of all vectors x in Rn such that Ax = 0. In other words,

Null(A) = set of solutions to the homogeneous system Ax = 0

Theorem 5.1.6. If A is an m × n matrix, the null space of A is a subspace of

Rn .

Proof. We need to show that Null(A) satisfies the three conditions of the defi-
nition of a subspace.

ˆ Null(A) contains the zero vector since A0 = 0.

ˆ Let u and v be vectors in Null(A). Then Au = 0 and Av = 0. We want

to show that u + v is in Null(A).

A(u + v) = Au + Av Property: A(u + v) = Au + Av

=0+0 Since Au = 0 and Av = 0
=0

This shows that u + v is in Null(A).

ˆ Let u be a vector in Null(A) and let c be a scalar. Then Au = 0. Multiply-

ing both sides of this latter equation by c, we get cAu = 0, or A(cu) = 0.
This shows that cu is in Null(A).

So Null(A) is a subspace of Rn
5.1. Subspaces, Column and Null Spaces 259

(a) True by definition.

(b) True by Theorem 5.1.6.
(c) False. By Theorem 5.1.6, if A is a 5 × 3 matrix, the null space of A will be a
subspace of R3 .
 
0 0 0
(d) True. Indeed, let A = be the zero 2 × 3 matrix. For every vector x
0 0 0
in R3 , Ax = 0. So Null(A) is the set of all vectors in R3 , that is, Null(A) = R3 .
(e) False. By Theorem 5.1.6, if A is a 3 × 2 matrix, the null space of A will be a
subspace of R2 . So Null(A) cannot be R3 .
 
1 −1
(f ) True. Indeed, let A = . The system Ax = 0 is
0 0

x1 − x2 = 0
= 0

The variable x1 is a leading variable, while x2 is a free variable. Let x2 = s.

Then x1 = s, and the general solution is

x1 = s
x2 = s

where s can be any real number. So

     
x1 s 1
x= = =s
x2 s 1

Thus, the
 solution
 set to the equation Ax = 0 is the line in R2 spanned by the
1
vector .
1
(g) True. By definition, the columns of A are linearly independent if and only if the
system Ax = 0 has only the trivial solution, that is, if and only if Null(A) = {0}.
(h) True. Indeed, by definition Null(A) = {0} if and only if the equation Ax = 0
has only the trivial solution. By the Invertible Matrix Theorem, the system
Ax = 0 has only the trivial solution if and only if A is invertible.

14. [Exercise on page 49]

(a) We compute Au:

      
1 −3 0 2 0 0+0+0+2 2
 −3 9 1 −2   0
    0 + 0 − 4 − 2   −6 
=
  0 + 0 − 4 + 2  =  −2 
   
 −1 3 1 2   −4
2 −6 −1 0 1 0+0+4+0 4

Since Au 6= 0, the vector u is not in Null(A).

260 Chapter 5. Subspaces of Rn

(b) In other to determine Null(A), we need to solve the equation Ax = 0. As usual,

we reduce the augmented matrix:
   
1 −3 0 2 0 1 −3 0 2 0
 −3 R2 + 3R1 → R2
9 1 −2 0   0 0 1 4 0 

 −1
 R3 + R1 → R3  
3 1 2 0   0 0 1 4 0 
R4 − 2R1 → R4
2 −6 −1 0 0 0 0 −1 −4 0
 
1 −3 0 2 0
R3 − R2 → R3  0
 0 1 4 0 

R4 + R2 → R4  0 0 0 0 0 
0 0 0 0 0

The corresponding system is

x1 − 3x2 + 2x4 = 0
x3 + 4x4 = 0
The variable x1 and x3 are leading variables, while x2 and x4 are free. Set x2 = s
and x4 = t. Then the solution set is
x1 = 3s − 2t
x2 = s
x3 = −4t
x4 = t
 
x1
x2 
Thus, Null(A) is the set of vectors x = x3  of the form


x4
       
x1 3s − 2t 3 −2
 x2 s  = s 1  + t 0 
      
x=
 x3
=
  −4t   0   −4 
x4 t 0 1
where s and t are real numbers.
(c) To find a vector in Null(A), all we have to do is to assign numbers to s and t.
ˆ Let s = 1 and t = 0. Substituting s with 1 and t with 0 into the solution
3
 1 
obtained in part (b), we get the vector 
 0  in Null(A).


0
 
−2
 0 
ˆ When s = 0 and t = 1, we obtain the vector   −4  in Null(A).


1
(d) Let ai , 1 ≤ i ≤ 4, be the ith column of A. By definition, a vector in Col(A) is a
linear combination of the columns of A. For example,
5.2. Basis, Dimension, and Rank 261

 
1
 −3 
ˆ 1a1 + 0a2 + 0a3 + 0a4 = a1 = 
 −1  is in Col(A).


2
 
−3
 9 
ˆ The vector 0a1 + 1a2 + 0a3 + 0a4 = a2 = 
 3  is in Col(A) as well.


−6

5.2 Basis, Dimension, and Rank

1. [Exercise on page 50]

Basis for a Subspace

Let H be a subspace of Rn , and let B = {v1 , · · · , vp } be a set of vectors in H.

We say that B is a basis for H if the following two conditions are satisfied.

(1) B is linearly independent. (This means that the vector equation x1 v1 +

· · · + xp vp = 0 has only the trivial solution.)

(2) B spans H. (This means that every vector in H can be written as a linear
combination of v1 , · · · , vp .)

(a) True by definition.

(b) False because a set containing one vector in R2 cannot span R2 as one vector is
not enough to span R2 (we need at least two vectors).
(c) False. Indeed, a set containing three vectors in R2 is not linearly independent
thanks to Theorem 2.4.6.
(d) False for the same reason as in part (c).
(e) True by the following theorem and definition.

Standard Basis for Rn

Let n ≥ 1 be an integer. For 1 ≤ i ≤ n, let ei be the vector of Rn defined
as follows. The jth entry of ei , denoted (ei )j , is defined by

0 if j 6= i
(ei )j =
1 if j = i

In other words, all the entries of ei are 0 except the ith entry, which is 1.
For example,
ˆ if n = 1, e1 = 1 .
 

   
1 0
ˆ If n = 2, we have e1 = and e2 = .
0 1
262 Chapter 5. Subspaces of Rn


    
1 0 0
ˆ If n = 3, we have e1 = 0 , e2 = 1 , and e3 = 0 .
    
0 0 1

Theorem 5.2.1. The set {e1 , · · · , en } forms a basis for Rn .

Proof. Let A be the matrix whose columns are e1 , e2 , · · · , en .
 
1 0 ··· 0
 0 1 ··· 0 
A =  ..
 
.. . . .. 
 . . . . 
0 0 ··· 1

This is nothing but the n × n identity matrix In . So by the Invertible

Matrix Theorem the columns of A are linearly independent and span Rn .

The set {e1 , · · · , en } is referred to as the standard basis for Rn .

(f ) False by definition.

2. [Exercise on page 50]

(a) Showing that B = {v1 , v2 } is linearly
 independent.
   Let c1 , c2 be scalars such that
1 −1 0
c1 v1 + c2 v2 = 0. Then c1 + c2 = . This gives the following
2 3 0
system.
c1 − c2 = 0
2c1 + 3c2 = 0
To solve this, we need to reduce the augmented matrix:
   
1 −1 0 1 −1 0
R2 − 2R1 → R2
2 3 0 0 5 0
 
1 1 −1 0
R → R2
5 2 0 1 0
 
1 0 0
R1 + R2 → R1
0 1 0
The corresponding system is c1 = 0 and c2 = 0. This implies that B is linearly
independent.
 
2 b1
Showing that B spans R . Let b = be an arbitrary vector in R2 . We want
b2
to show that there exist scalars c1 and c2 such that c1 v1 + c2 v2 = b. To do this,
we need to prove that the following system has a solution.
c 1 − c 2 = b1
2c1 + 3c2 = b2
5.2. Basis, Dimension, and Rank 263

Using the same elementary row operations as above, the augmented matrix of
the latter system can be transformed into

" 3
#
1 0 b
5 1
+ 15 b2
1
0 1 b
5 2
− 52 b1

So c1 = 35 b1 + 51 b2 and c2 = − 25 b1 + 15 b2 . This shows that B spans R2 .

Conclusion: Since B is a linearly independent set in R2 that spans R2 , we can

conclude that B is a basis for R2 .

Alternate Solution 1 (Using the Concept of Pivot)

 
1
We want to prove that the set B = {v1 , v2 }, where v1 = and v2 =
  2
−1  
, forms a basis for R2 . Let A = v1 v2 be the matrix whose
3  
1 −1
columns are v1 and v2 . Then A = . Row reducing A, we have
2 3
   
1 −1 1 −1
R2 − 2R1 → R2 =B
2 3 0 5

The matrix B is an echelon form of A.

ˆ Since every column of B is a pivot column, it follows that the

columns of A are linearly independent.

ˆ Since B has a pivot in every row, it follows that the columns of A

span R2 .

So the columns of A form a basis for R2 .

264 Chapter 5. Subspaces of Rn

Alternate Solution 2 (Dimension and The Basis Theorem).

Dimension and The Basis Theorem

Theorem 5.2.2. Let H be a subspace of Rn , and let B =
{v1 , · · · , vp } be a basis for H. Then any other basis for H has exactly
p vectors.
The number p in Theorem 5.2.2 is called the dimension of H, and
we write dim H = p. The zero subspace, {0}, has no basis and its
dimension is defined to be zero. That is, dim{0} = 0.

Dimension of Rn . In Theorem 5.2.1, we proved that the set

{e1 , · · · , en } is a basis for Rn . Since that set has n vectors, it follows
that dim Rn = n.
Theorem 5.2.3 (The Basis Theorem). Let H be a subspace of
dimension p. Then the following two statements hold.

ˆ Every set {v1 , · · · , vp } of p vectors in H that is linearly inde-

pendent is a basis for H.

ˆ Every set {v1 , · · · , vp } of p vectors in H that spans H is a

basis for H.

Note. This theorem says that if the dimension of a subspace H is

known to be p, to prove that a set {v1 , · · · , vp } of p vectors in H
is a basis for H, we just need to prove that {v1 , · · · , vp } is linearly
independent. Or we just need to prove that {v1 , · · · , vp } spans H.

We now explain how to use Theorem 5.2.3 to show that the set B = {v1 , v2 }
above is a basis for R2 . Since dim R2 = 2 and since B has two vectors, we
just need to show that B is linearly independent (thanks to Theorem 5.2.3)
to conclude that it is a basis for R2 . But we already proved this above.

(b) Let’s first recall the notion of coordinates of a vector relative to a basis.

Coordinates of a Vector Relative to a Basis

If B = {v1 , · · · , vp } is a basis for a subspace H, then every vector in H can
be written as a linear combination of v1 , · · · , vp . The following theorem
says that such a linear combination is unique.
Theorem 5.2.4. Let B = {v1 , · · · , vp } be a basis for a subspace H. Then
for every vector x in H there exist a unique sequence c1 , · · · , cp of scalars
such that x = c1 v1 + · · · cp vp .
The coefficients c1 , · · · , cp are called the coordinates of x relative to
5.2. Basis, Dimension, and Rank 265

 
c1
 .. 
the basis B. The notation for this is [x]B =  . .
cp

 
−9
To find the coordinates of b = relative to the basis B = {v1 , v2 }, we
7
need to write b asa linear
 combination of v1 and v2 . In part (a), we showed that
b1
every vector b = can be written as
b2
   
3 1 2 1
b= b1 + b2 v1 + − b1 + b2 v2 .
5 5 5 5
Substituting b1 with −9 and b2 with 7, we get
   
3 1 2 1
b= (−9) + (7) v1 + − (−9) + (7) v2
5 5 5 5
   
27 7 18 7
= − + v1 + + v2 = −4v1 + 5v2
5 5 5 5
Hence, the coordinates
 of b relative to the basis {v1 , v2 } are −4 and 5, that is,
−4
[x]B = .
5

3. [Exercise on page 50]

 
Solution. Let A = v1 v2 v3 be the matrix whose columns are v1 , v2 , and v3 .
1 −1 2
Then A =  0 2 −3 . Row reducing A, we have
−2 −3 4
   
1 −1 2 1 −1 2
 0 2 −3  R3 + 2R1 → R3  0 2 −3 
−2 −3 4 0 −5 8
 
1 −1 2
1
R → R2  0 1 − 23 
 
2 2
0 −5 8
 
1 −1 2
R3 + 5R2 → R3  0 1 − 23  =B
 
1
0 0 2

The matrix B is an echelon form of A. Since every column of B is a pivot column, it

follows that the columns of A are linearly independent. Since dim R3 = 3 and since
the set B = {v1 , v2 , v3 } is linearly independent and has exactly 3 vectors, we can
conclude that B is a basis for R3 thanks to Theorem 5.2.3.
266 Chapter 5. Subspaces of Rn

4. [Exercise on page 50]

(a) The set B has one nonzero vector.

 So
 it is linearly independent. But it does not
−18
span R2 because the vector is in R2 , but not in Span{v} as it is not a
−25
multiple of v. Thus, B is not a basis for R2 .

Alternate Solution
Since dim R2 = 2, any basis for R2 must have 2 vectors. Since the set
B = {v} has only one vector, it cannot be a basis for R2 .

(b) By inspection, one can see that v1 = −3v2 . This implies that the set B = {v1 , v2 }
is not linearly independent, and therefore is is not a basis for R2 .
(c) The set B is not a basis for R3 because any basis for R3 must have 3 vectors
since dim R3 = 3. And B does not contain 3 vectors.
 
(d) First, we need to determine if B is linearly independent. Let A = v1 v2 v3
be the matrix whose columns are v1 , v2 , and v3 . We need to reduce A.
   
−4 −1 −5 −1 1 −5
 3 2 0  R1 + R2 → R1  3 2 0 
7 6 −4 7 6 −4
 
1 −1 5
(−1)R1 → R1  3 2 0 
7 6 −4
 
1 −1 5
R2 − 3R1 → R2  0 5 −15 
R3 − 7R1 → R3
0 13 −39
 
1
R → R 1 −1 5
5 2 2
1
 0 1 −3 
13
R 3 → R 3 0 1 −3
 
1 −1 5
R3 − R2 → R3  0 1 −3  =B
0 0 0

The matrix B is an echelon form of A. Since B has a non-pivot column (the

third column is not a pivot column), it follows that the columns of A are not
linearly independent. So the set B is not a basis for R3 .
(e) Since dim R3 = 3 and the number of vectors in B is 4 6= 3, it follows that B is
not a basis for R3 .

5. [Exercise on page 51]

(a) Since H = Span{v1 , v2 }, the set B = {v1 , v2 } spans H. To show that B is linearly
independent, let c1 , c2 be scalars such that c1 v1 + c2 v2 = 0. Then we have the
5.2. Basis, Dimension, and Rank 267

following system.
2c1 + c2 = 0
− c2 = 0
c1 + 3c2 = 0
Solving this, we get c1 = 0 and c2 = 0. This implies that B is linearly indepen-
dent. Hence, B is a basis for H.
(b) We want to determine if there exist scalars c1 and c2 such tha c1 v1 + c2 v2 = x.
This amounts to determining if the following system is consistent.

2c1 + c2 = 1
− c2 = 3
c1 + 3c2 = −7
We reduce the augmented matrix:
   
2 1 1 1 3 −7
 0 −1 3  R1 ↔ R3  0 −1 3 
1 3 −7 2 1 1
 
1 3 −7
R3 − 2R1 → R3  0 −1 3 
0 −5 15
 
1 3 −7
(−1)R2 → R2  0 1 −3 
− 15 R3 → R3
0 1 −3
 
1 3 −7
R3 − R2 → R3  0 1 −3 
0 0 0
 
1 0 2
R1 − 3R2 → R1  0 1 −3 
0 0 0

So c1 = 2 and c3 = −3, and x = 2v1 − 3v2 is a vector in H.

Since x= 2v1− 3v2 , the coordinates of x relative to B are 2 and −3, that is,
2
[x]B = .
−3

6. [Exercise on page 51]

(a) To find a basis for Null(A), we first need to find Null(A). By definition, Null(A)
is the set of solutions to the equation Ax = 0. So, to find Null(A), we first need
to solve the system Ax = 0. Row reducing the augmented matrix, we have
   
1 2 4 0 1 2 4 0
R2 − 3R1 → R2
 3 7 7 0   0 1 −5 0 
R3 + 4R1 → R3
−4 −9 −11 0 0 −1 5 0
268 Chapter 5. Subspaces of Rn

 
1 2 4 0
R3 + R2 → R3  0 1 −5 0 
0 0 0 0
 
1 0 14 0
R1 − 2R2 → R1  0 1 −5 0 
0 0 0 0

The corresponding system is

x1 + 14x3 = 0
x2 − 5x3 = 0

Let x3 = s. Then x1 = −14s and x2 = 5s. So x is a solution to Ax = 0 if and

only if x is of the form
     
x1 −14s −14
x =  x2  =  5s  = s  5 
x3 s 1

where s is a real number.  This implies

 that Null(A) is the subspace in R3
−14
spanned by the vector v =  5 , that is,
1

Null(A) = Span{v}

So the set B = {v} spans Null(A). Moreover, B is linearly independent since

v 6= 0. (Remember that a set of one vector v is linearly independent if and only
if v 6= 0.) Since B spans Null(A) and is linearly independent, it follows that B
is a basis for Null(A).

Now, we find a basis for Col(A). We will need the following theorem.

Basis for Col(A)

Theorem 5.2.5. Let A be a matrix, and let B be an echelon form of A.

Then the columns of A corresponding to the pivot columns of B form a
basis for Col(A).
So to find a basis for Col(A), we can proceed as follows.

Step 1. Row reduce A to an echelon matrix, say B.

Step 2. Identify the pivot columns of B.

Step 3. The columns of A corresponding to the pivot columns of B

form a basis for Col(A).
5.2. Basis, Dimension, and Rank 269

 
1 2 4
From the row reduction above, an echelon form of A =  3 7 7  is
−4 −9 −11
 
1 2 4
B= 0 1 −5 
0 0 0

The pivot columns of B are column 1 and column 2. The corresponding columns
in A are the vectors
   
1 2
u1 =  3  and u2 =  7 
−4 −9

By Theorem 5.2.5, the set {u1 , u2 } forms a basis for Col(A).

(b) To find a basis for Null(A), we first need to solve the system Ax = 0.
1
R → R1
   
5 −30 40 5 1 1 −6 8
 3 −18 1
24  R
3 2
→ R2  1 −6 8 
−4 24 −32 1
− 4 R3 → R3 1 −6 8
 
1 −6 8
R2 − R1 → R2  0 0 0 
R3 − R1 → R3
0 0 0

The corresponding system is x1 − 6x

2 + 8x3 = 0. Let x2 = s and x3 = t. Then
x1
x1 = 6s − 8t. So a vector x =  x2  is a solution to Ax = 0 if and only if x is
x3
of the form
       
x1 6s − 8t 6s −8t
x =  x2  =  s = s + 0 
x3 t 0 t
   
6 −8
=s 1 +t
   0 
0 1

where s and t are real numbers. This  implies

 that every
 vector
 in Null(A) is a
6 −8
linear combination of the vectors v1 =  1  and v2 =  0 . In other words,
0 1

Null(A) = Span{v1 , v2 }

So the set {v1 , v2 } spans Null(A). Moreover, the vectors v1 and v2 are linearly in-
dependent since v2 is not a multiple of v1 (otherwise, the third entry of v2 would
270 Chapter 5. Subspaces of Rn

be zero, which is not the case). Thus, the set {v1 , v2 } is a basis for Null(A).

Now, we find a basis for Col(A). From the row reduction above, an echelon form
of A is  
1 −6 8
B= 0 0 0 
0 0 0
The matrix B has only one pivot
 column:
 the first column. The corresponding
5
column in A is the vector u =  3 . By Theorem 5.2.5, the set {u} is a basis
−4
for Col(A).

7. [Exercise on page 51]

(a) True. Indeed, let H be a nonzero subspace of Rn and let B = {v1 , · · · , vp } be a

basis for H. Then the set {5v1 , · · · , 5vp } is clearly another basis for H. Actually,
for every nonzero number c, the set {cv1 , · · · , cvp } is a basis for H.
(b) False by Theorem 5.2.2.
(c) True by the definition of dimension of a subspace.
(d) True. Indeed, by Theorem 5.2.1, the vectors e1 , · · · , en form a basis for Rn .
(e) False because a plane in R3 is spanned by two linearly independent vectors. So
the dimension of a plane is 2.
(f ) False because the dimension of a subspace of Rn is always less than or equal to
n.
(g) False. Indeed, let A be a 6 × 7 matrix. Then the homogeneous system Ax = 0
has at least one free variable since the number of columns of A is greater than
the number of rows. So the system Ax = 0 has infinitely many solutions. This
implies that Null(A) 6= {0}.
(h) False. Indeed, by inspection, the vectors u and v are linearly dependent as
v = −4u. So
Span{u, v} = Span{v}
But Span{v} is the line in R2 generated by v. Since the dimension of a line is
1, it follows that dim Span{u, v} = 1.
(i) True. Indeed, let H = Span{u, v}.
ˆ Then the set {u, v} spans H.
 
1 3
ˆ Let A be the matrix whose columns are u and v. Then A = .
−2 4
Row reducing A, we get
   
1 3 1 3
R2 + 2R1 → R2 =B
−2 4 0 10
Since every column of B is a pivot column, it follows that the columns of A
are linearly independent.
5.2. Basis, Dimension, and Rank 271

So the set {u, v} is a basis for H. Since this basis has two vectors, it follows
that dim H = 2.
(j) True by the Basis Theorem.
(k) True by the Basis Theorem.
(l) True. This is true because of the following theorem.

Dimension of the Null Space of a Matrix or Nullity of a Matrix

Theorem 5.2.6. Let A be a matrix. Then the dimension of the null space
of A is equal to the number of free variables of the system Ax = 0.
The dimension of the null space of a matrix A is sometimes referred to as
the nullity of A.

8. [Exercise on page 52]

 
x1
(a) Let x = be a vector in H. Then x1 −5x2 = 0. This implies that x1 = 5x2 .
x2
So     
x1 5x25
x= = = x2
x2 x21
 
5
This shows that every vector in H is a multiple of v = . So the set B = {v}
1
spans H. Moreover, this set is linearly independent because v 6= 0. (Remember:
any set containing one nonzero vector is linearly independent.) Thus, B is a
basis for H.
(b) Since the basis B contains one vector, dim H = 1.

9. [Exercise on page 52]

We want to find a basis for H = Span{v1 , v2 , v3 }. Let A be the matrix whose columns
are v1 , v2 , and v3 . Observe that by definition, Col(A) = Span{v1 , v2 , v3 }. So to find
a basis for H = Span{v1 , v2 , v3 }, all we have to do is to find the pivot columns of A
and use Theorem 5.2.5. So we first need to transform A into an echelon matrix.
   
1 −3 7 1 −3 7
 −1 R2 + R1 → R2
5 −13   0 2 −6 
R3 + 4R1 → R3
−4 10 −22 0 −2 6
 
1 −3 7
R3 + R2 → R3  0 2 −6  =B
0 0 0

The first
 two  columns of B are pivot columns. The corresponding columns in A are
1 −3
u1 = −1 and u2 =
   5 . Thanks to Theorem 5.2.5, the set {u1 , u2 } is a basis
−4 10
for Span{v1 , v2 , v3 }.
272 Chapter 5. Subspaces of Rn

10. [Exercise on page 52]

 
2a − 3b
ˆ We first need to find a basis for H. Let x =  −b  be a vector in H.
−5a + 4b
Then
         
2a − 3b 2a −3b 2 −3
x= −b = 0  +  −b  = a  0  + b  −1 
−5a + 4b −5a 4b −5 4
 
2
This shows that every vector in H is a linear combination of v1 =  0  and
  −5
−3
v2 =  −1 . So the set B = {v1 , v2 } spans H. Furthermore, B is linearly
4
independent. Indeed, let c1 and c2 be scalars such that c1 v1 + c2 v2 = 0. Then
we have the following system.
2c1 − 3c2 = 0
− c2 = 0
−5c1 + 4c2 = 0.
From the second equation, we get c2 = 0. Substituting this into the first equa-
tion, we get 2c1 − 0 = 0, which implies that c1 = 0. Substituting these into
the third equation, we get 0 = 0. So the system has only the trivial solution.
Therefore, the set B is linearly independent.
Conclusion. The set B = {v1 , v2 } is a basis for H.
ˆ Since B has two vectors, dim H = 2.

11. [Exercise on page 52]

Rank of a Matrix and The Rank-Nullity Theorem

Let A be a matrix. The rank of A, denoted rank(A), is defined to be the

dimension of the column space of A. That is,

rank(A) = dim Col(A)

The following important theorem is referred to as the Rank-Nullity Theorem

or just as the Rank Theorem.

Theorem 5.2.7 (Rank-Nullity Theorem). Let A be a matrix. Then the

rank of A plus the dimension of the null space of A is equal to the number of
columns of A. That is, if n is the number of columns of A, then

rank(A) + dim Null(A) = n,

Proof. Let B be an echelon form of A.

5.2. Basis, Dimension, and Rank 273

ˆ By Theorem 5.2.5, the columns of A corresponding to the pivot columns

of B form a basis for Col(A). So the dimension of Col(A), that is, the
rank of A, is equal to the number of pivot columns of B:

rank(A) = number of pivot columns of B (5.2.1)

ˆ By definition, Null(A) is the set of solutions to the system Ax = 0. Since

B is an echelon form of A, it follows that the systems Ax = 0 and Bx = 0
have the same set of solutions. Since

– every non-pivot column of B corresponds to a free variable of the

system Ax = 0, and
– since the number of free variables of the system Ax = 0 equals the
dimension of the null space of A (thanks to Theorem 5.2.6),

it follows that

dim Null(A) = number of non-pivot columns of B (5.2.2)

Combining (5.2.1) and (5.2.2), we get

rank(A) + dim Null(A) = number of columns of B

= number of colums of A

This ends the proof.

Remark. Let A be an m × n matrix. Then

rank(A) ≤ n and rank(A) ≤ m

The fact that rank(A) ≤ n follows immediately from the Rank-Nullity The-
orem. The inequality rank(A) ≤ m follows from the fact that the rank of A
corresponds to the number of pivot positions of A and the fact that the number
of pivot positions of A cannot be greater than the number of rows of A.

(a) True by the definition of rank.

(b) True by the Rank-Nullity Theorem.
(c) False. By the Rank-Nullity Theorem, dim Null(A) + rank(A) = 5 if A has 5
columns.
(d) False. If A is a 7 × 11 matrix with dim Null(A) = 5, then the rank of A is
11 − 5 = 6 (not 2!) by the Rank-Nullity Theorem.
(e) True. If A is a 7 × 11 matrix with dim Null(A) = 5, then the rank of A is
11 − 5 = 6 by the Rank-Nullity Theorem.
(f ) False. Indeed, if A is a 15 × 9 matrix whose rank(A) = 7, then by the Rank-
Nullity Theorem the dimension of the null space of A is 9 − 7 = 2 (not 8!).
274 Chapter 5. Subspaces of Rn

(g) False. Indeed, let A be a 4 × 3 matrix. Then, by the Rank-Nullity Theorem,

rank(A)+dim Null(A) = 3. So the rank of A is less than or equal to 3. Therefore,
it is not possible to find a 3 × 4 matrix whose rank is 4.
(h) False. Indeed, let A be a 4 × 7. Since the rank of A corresponds to the number
of pivot positions of A, and since the number of pivot positions of A cannot be
greater than the number of rows of A, it follows that rank(A) ≤ 4. So it is not
possible to find a 4 × 7 matrix whose rank is 5.
 
1 0 0
0 1 0
(i) True. Indeed, let A =  0 0 1. Since A has three pivot positions, it follows


0 0 0
that rank(A) = 3.
(j) False. Indeed, if the system Ax = 0 has two free variables, then dim Null(A) = 2
(by Theorem 5.2.6). Therefore, by the Rank-Nullity Theorem, rank(A) = 6−2 =
4.

12. [Exercise on page 52]

(a) (i) By definition, Null(A) is the set of solutions to the homogeneous equation
Ax = 0. So to find a basis for Null(A), we first need to solve the system Ax = 0:
   
1 −1 1 0 1 −1 1 0
 0 1 1 0  R3 + R1 → R3  0 1 1 0 
−1 2 1 0 0 1 2 0
 
1 −1 1 0
R3 − R2 → R3  0 1 1 0 
0 0 1 0
 
1 −1 0 0
R1 − R3 → R1  0 1 0 0 
R2 − R3 → R2
0 0 1 0
 
1 0 0 0
R1 + R2 → R1  0 1 0 0 
0 0 1 0

So the equation Ax = 0 has only the trivial solution. Therefore, Null(A) is

the zero subspace and its basis is the empty set {}. (Remember that the zero
subspace has no basis.)

(ii) Since Null(A) is the zero subspace, dim Null(A) = 0.

(iii) To find a basis for the column space of A, we will Theorem 5.2.5. From the
5.2. Basis, Dimension, and Rank 275

 
1 −1 1
row reduction above, an echelon form of A =  0 1 1  is
−1 2 1
 
1 −1 1
B= 0 1 1 
0 0 1

The pivot columns of B are column 1, column 2, and Column 3. The corre-
sponding columns in A are the vectors
     
1 −1 1
u1 =  0 , u2 =
  1 , and u3 = 1 
 
−1 2 1

By Theorem 5.2.5, the set {u1 , u2 , u3 } is a basis for Col(A).

(iv) Since a basis for Col(A) has three vectors, dim Col(A) = 3.

(v) By definition, rank(A) = dim Col(A). Since dim Col(A) = 3, it follows that
rank(A) = 3.
(b) (i) To find a basis for Null(A), we first need to solve the equation Ax = 0. We
reduce the augmented matrix:

   
1 2 −1 0 1 2 −1 0
 −1 −1 R2 + R1 → R2
4 0   0 1 3 0 
R3 + 3R1 → R3
−3 −6 3 0 0 0 0 0
 
1 0 −7 0
R1 − 2R2 → R1  0 1 3 0 
0 0 0 0

The corresponding system is

x1 − 7x3 = 0
x2 + 3x3 = 0

Set x3 = s. Then the general solution to the equation Ax = 0 is

x1 = 7s
x2 = −3s s is an arbitrary number.
x3 = s

So Null(A) is the set of vectors x of the form

   
7s 7
x =  −3s  = s  −3  = sv,
s 1
276 Chapter 5. Subspaces of Rn

 
7
where v =  −3 . Since every vector of Null(A) is a multiple of v, the set
1
B = {v} spans Null(A). Moreover, B is linearly independent because v 6= 0.
(Remember that a set containing one vector v is linearly independent if and
only if v 6= 0.) Thus, B is a basis for Null(A).

(ii) Since B has one vector, dim Null(A) = 1.

 
1 2 −1
(iii) From the row reduction above, an echelon form of A =  −1 −1 4 
  −3 −6 3
1 2 −1
is B =  0 1 3 . The pivot columns of B are columns 1 and 2. The
0 0 0
corresponding columns in A are the vectors
   
1 2
u1 =  −1  and u2 =  −1 
−3 −6

By Theorem 5.2.5, the set {u1 , u2 } forms a basis for Col(A).

(iv) Since the latter basis has two vectors, dim Col(A) = 2.

(v) By definition, rank(A) = dim Col(A). Since dim Col(A) = 2 (from part (iv)),
it follows that rank(A) = 2.
 
1 3 0 2
 0 0 2 −8 
(c) A =   −1 −3

2 −10 
2 6 −2 12
(i) To find a basis for Null(A), we first need to solve the equation Ax = 0.
   
1 3 0 2 0 1 3 0 2 0
 0
 0 2 −8 0  R3 + R1 → R3  0
 0 2 −8 0 

 −1 −3 2 −10 0  R4 − 2R1 → R4  0 0 2 −8 0 
2 6 −2 12 0 0 0 −2 8 0
 
1 3 0 2 0
R3 − R2 → R3 
 0 0 2 −8 0 

R4 + R2 → R4  0 0 0 0 0 
0 0 0 0 0
 
1 3 0 2 0
1
 0 0 1 −4 0 
R
2 2
→ R2  
 0 0 0 0 0 
0 0 0 0 0
5.2. Basis, Dimension, and Rank 277

The corresponding system is

x1 + 3x2 + 2x4 = 0
x3 − 4x4 = 0

Set x2 = s and x4 = t. Then the general solution is

x1 = −3s − 2t
x2 = s
where s and t can be any real numbers
x3 = 4t
x4 = t
 
x1
 x2 
So every vector x =  x3  in Null(A) is of the form


x4
     
−3s − 2t −3 −2
 s  = s  1  + t  0  = sv1 + tv2 ,
    
x=  4t   0   4 
t 0 1
   
−3 −2
 1 
 and v2 =  0 . This shows that Null(A) is spanned
 
where v1 =   0   4 
0 1
by the set B = {v1 , v2 }. Moreover, B is linearly independent since v2 is not a
multiple of v1 (otherwise the third entry of v2 would be 0, which is not the case).
Hence, B is a basis for Null(A).

Alternate Way to Show That B is Linearly Independent

Remember that a set {v1 , · · · , vp } is linearly independent if the vectors

equation
x1 v1 + · · · + xp vp = 0
has
 only  the trivial   We want to show that {v1 , v2 }, where v1 =
solution.
−3 −2
 1   0 
 0  and v2 =  4 , is linearly independent. Let x1 and x2 be
   

0 1
scalars such that x1 v1 + x2 v2 = 0. Then
   
−3x1 − 2x2 0

 x1 = 0 
  
 4x2   0 
x2 0
This implies that
−3x1 − 2x2 = 0, x1 = 0, 4x2 = 0, and x2 = 0
278 Chapter 5. Subspaces of Rn

So x1 = 0 and x2 = 0. Thus, the set {v1 , v2 } is linearly independent.

(ii) Since B contains two vectors, dim Null(A) = 2.

(iii) An echelon form of A is the matrix

 
1 3 0 2
 0 0 2 −8 
B=  0 0

0 0 
0 0 0 0
The pivot columns of B are columns 1 and 3. And the corresponding columns
in A are the vectors
   
1 0
 0   2 
u1 =  
 −1  and u2 = 
 2 


2 −2
By Theorem 5.2.5, the set {u1 , u2 } forms a basis for Col(A).

(iv) Since this latter basis contains two vectors, dim Col(A) = 2.

(v) By definition, rank(A) = dim Col(A). Since dim Col(A) = 2 (from part (iv)),
it follows that rank(A) = 2.
(d) (i) To find a basis for Null(A), we first need to solve the system Ax = 0. Row
reducing the augmented matrix, we have
   
1 −5 2 0 −1 1 −5 2 0 −1
 3 −15 R2 − 3R1 → R2 
6 1 −6  0 0 0 1 −3 
R3 + R1 → R3
−1 5 −2 −1 4 0 0 0 −1 3
 
1 −5 2 0 −1
R3 + R2 → R3  0 0 0 1 −3 
0 0 0 0 0
The corresponding system is
x1 − 5x2 + 2x3 − x5 = 0
x4 − 3x5 = 0
Set x2 = s1 , x3 = s2 , and x5 = s3 . Then
x1 = 5s1 − 2s2 + s3 and x4 = 3s3
So every vector x in Null(A) is of the form
         
x1 5s1 − 2s2 + s3 5s1 −2s2 s3
 x2   s 1   s1 0 0
      
       
x =  x3  = 
   s2 = 0
 
+
  s2 +
  0 

 x4   3s3   0   0   3s3 
x5 s3 0 0 1
5.2. Basis, Dimension, and Rank 279

     
5 −2 1

 1 


 0 


 0 

= s1 
 0  + s2 
  1  + s3 
  0  = s1 v1 + s2 v2 + s3 v3

 0   0   3 
0 0 1
where      
5 −2 1

 1 


 0 


 0 

v1 = 
 0 ,
 v2 = 
 1 ,
 and v3 = 
 0 

 0   0   3 
0 0 1
This shows that a vector x is in Null(A) if and only if x is a linear combination of
the vectors v1 , v2 , and v3 . In other words, the set B = {v1 , v2 , v3 } spans Null(A).
To determine whether B is linearly independent, let c1 , c2 , c3 be scalars such that
c1 v1 + c2 v2 + c3 v3 = 0. Then
   
5c1 − 2c2 + c3 0

 c1   0 
  

 c2 = 0 
  
 3c3   0 
c3 0
This implies that

5c1 − 2c2 + c3 = 0, c1 = 0, c2 = 0, 3c3 = 0, and c3 = 0

Since the equation c1 v1 + c2 v2 + c3 v3 = 0 has only the trivial solution, the set
{v1 , v2 , v3 } is linearly independent.
Conclusion. Since the set B = {v1 , v2 , v3 } is linearly independent and spans
Null(A), it follows that B is a basis for Null(A).

(ii) Since B has three vectors, it follows that dim Null(A) = 3.

(iii) Finding a basis for Col(A). From the row reduction above, an echelon form
of A is  
1 −5 2 0 −1
B= 0 0 0 1 −3 
0 0 0 0 0
The pivot columns of B are columns 1 and 4. The corresponding columns in A
are the vectors    
1 0
u1 =  3  and u2 =  1 
−1 −1
By Theorem 5.2.5, the set {u1 , u2 } is a basis for Col(A).

(iv) From part (iii), a basis for Col(A) has two vectors. So dim Col(A) = 2.
280 Chapter 5. Subspaces of Rn

(v) By definition, rank(A) = dim Col(A). Since dim Col(A) = 2 from part (iv),
it follows that rank(A) = 2.

13. [Exercise on page 53]

(a) To find the rank of A, we first need to put A into echelon form.
   
3 1 −1 2 1 3 1 −1 2 1
 −3 −1 1 −4 6  R2 + R1 → R2  0 0 0 −2 7 
   =B
 0 0 0 0 0  R4 + 2R1 → R4  0 0 0 0 0 
−6 −2 2 −4 −2 0 0 0 0 0

Since B has two pivot columns (columns 1 and 4), it follows that the dimension
of the column space is 2. So
rank(A) = dim Col(A) = 2

(b) Using the Rank-Nullity Theorem, we have rank(A) + dim Null(A) = 5. This
implies that dim Null(A) = 5 − 2 = 3.

14. [Exercise on page 53]

(a) We first need to row reduce A.
   
1 −1 0 1 2 −3 1 −1 0 1 2 −3
 1 −1 R2 − R1 → R2
1 −1 2 −4   0 0 1 −2 0 −1 

 −1
 R3 + R1 → R3  
1 1 −2 2 4   0 0 1 −1 4 1 
R4 − 2R1 → R4
2 −2 −2 6 4 −3 0 0 −2 4 0 3
 
1 −1 0 1 2 −3
R3 − R2 → R3  0
 0 1 −2 0 −1 

R4 + 2R4 → R4  0 0 0 1 4 2 
0 0 0 0 0 1
The columns of A corresponding to the pivot columns of the echelon form are
the vectors
       
1 0 1 −3
 , a3 =  1  , a4 =  −1  , and a6 =  −4 
 1       
a1 = 
 −1   1   −2   4 
2 −2 6 −3
So a basis for Col(A) is B = {a1 , a3 , a4 , a6 }.
(b) By definition, the rank of A is the dimension of Col(A). Since a basis for Col(A)
contains four vectors, dim Col(A) = 4. So rank(A) = 4.
(c) The idea is to add one vector to the basis B. Let a2 be the second column of
A, and let S = {a1 , a2 , a3 , a4 , a6 }. Since S contains B and since B already spans
Col(A) (because B is a basis for Col(A)), it follows that S spans Col(A). This
implies that S must be linearly dependent (otherwise, S would be a basis for
Col(A) and the dimension of Col(A) would be 5, which is not the case). Thus,
the columns a1 , a2 , a3 , a4 , a6 are linearly dependent.
5.2. Basis, Dimension, and Rank 281

Alternate Solution
Let a5 be the fifth column of A. Using the same reasoning as above, the
columns a1 , a3 , a4 , a5 , and a6 are linearly dependent.

(d) By definition, a basis for a subspace H is a set that is linearly independent and
spans H. From part (a), the set B = {a1 , a3 , a4 , a6 } is a basis for Col(A). So B is
linearly independent and spans Col(A). This implies that the columns a1 , a3 , a4 ,
and a6 are linearly independent.
(e) i. Col(A) is a 4-dimensional subspace of R4 . Indeed, from part (a), dim Col(A) =
4. Since A has 4 rows, every column of A is a vector in R4 . So Col(A) is a
subspace of R4 .
ii. Null(A) is a 2-dimensional subspace of R6 . Indeed, by the Rank-Nullity
Theorem,

6 represents the number of

rank(A) + dim Null(A) = 6
columns of A
4 + dim Null(A) = 6 Since rank(A) = 4
dim Null(A) = 6 − 4 = 2

Since A has 6 columns, Null(A) is a subspace of R6 .

15. [Exercise on page 53]

ˆ Since K contains a 3-dimensional subpace H, and since K is not equal to H,

dim K > 3.
ˆ Since R5 contains K, and since R5 is not equal to K, 5 > dim K.
ˆ Combining these inequalities, we get 3 < dim K < 5. This implies dim K = 4.

16. [Exercise on page 53]

(a) Since an echelon form of a 9 × 14 matrix A has exactly five rows of all 0’s, there
are 9 − 5 = 4 pivot columns. This implies that dim Col(A) = 4.
(b) Using the Rank-Nullity Theorem, dim Null(A) = 14 − 4 = 10.

17. [Exercise on page 53]

ˆ The column space of A is not equal to R4 because every vector in Col(A) has 5
entries (the number of entries corresponds to the number of rows of A), while
every vector in R4 has 4.
ˆ Since A has four pivot columns, dim Col(A) = 4. By the Rank-Nullity Theorem,
we deduce that dim Null(A) = 10 − 4 = 6.

18. [Exercise on page 53]

282 Chapter 5. Subspaces of Rn

The Invertible Matrix Theorem (continued)

We stated the Invertible Matrix Theorem (IMT) in Theorem 4.2.5. The theo-
rem says that 12 statements are equivalent. The following theorem provides 6
more statements to the IMT.

Theorem 5.2.8. Let A be an n × n matrix. The following statements are

equivalent.

IMT1. A is invertible.

IMT13. The columns of A form a basis of Rn .

IMT14. Col(A) = Rn

IMT15. dim Col(A) = n

IMT16. rank(A) = n

IMT17. Null(A) = {0}

IMT18. dim Null(A) = 0

(a) True. This follows from the implication IMT1 =⇒ IMT17 in Theorem 5.2.8.
 
1 0
(b) False. Indeed, let A = 0 1. Consider the system Ax = 0:
0 0

x1 = 0
x2 = 0
0 = 0

The solution is x1 = 0 and x2 = 0. So Null(A) = {0}. But A is not invertible

as it is not even a square matrix.
(c) True. This follows from the equivalence IMT1 ⇐⇒ IMT17 in Theorem 5.2.8.
(d) True. This is the equivalence IMT1 ⇐⇒ IMT16 in Theorem 5.2.8.
(e) True. This is the equivalence IMT15 ⇐⇒ IMT18 in Theorem 5.2.8.
(f ) True. This is the equivalence IMT13 ⇐⇒ IMT14 in Theorem 5.2.8.
(g) False. Indeed, let A be a 3×4 matrix such that Col(A) = R3 . Then dim Col(A) =
3 = rank(A). By the Rank-Nullity Theorem, rank(A) + dim Null(A) = 4. This
implies that dim Null(A) = 4 − 3 = 1.

19. [Exercise on page 54]

(a) First, observe that A is a square matrix. So, since the columns of A are linearly
independent by hypothesis, it follows that A is invertible (by the Invertible
Matrix Theorem). This implies that A3 is invertible (using one of the properties
5.2. Basis, Dimension, and Rank 283

of inverses: A invertible implies that An is invertible). Using now the equivalence

IMT1 ⇐⇒ IMT18 in Theorem 5.2.8, it follows that dim Null(A3 ) = 0.
(b) Using a similar reasoning as before, A5 is invertible. And this implies that
rank(A5 ) = 7 thanks to the equivalence IMT1 ⇐⇒ IMT16 in Theorem 5.2.8.
(c) By hypothesis, the columns of A are linearly independent. So, by the Invertible
Matrix Theorem, AT is invertible. This implies that the columns of AT form a
basis for R7 thanks to the equivalence IMT1 ⇐⇒ IMT13 in Theorem 5.2.8.
284 Chapter 5. Subspaces of Rn
 Chapter6

Determinants

6.1 Introduction to Determinants

1. [Exercise on page 55]

Determinant of a 1 × 1 Matrix
 
Let A = a11 be a 1 × 1 matrix. The determinant of A, denoted det A, is
defined by
det A = a11

 
(a) det 1 = 1
 
(b) det 23 = 23
 
(c) det −7 = −7
 
(d) det 0 = 0

2. [Exercise on page 55]

Determinant of a 2 × 2 Matrix
 
a11 a12
Let A = be a 2 × 2 matrix. The determinant of A, denoted det A,
a21 a22
is defined by
det A = a11 a22 − a12 a21 (6.1.1)

 
3 4
(a) det = (3)(7) − (4)(5) = 21 − 20 = 1.
5 7
 
−4 3
(b) det = (−4)(2) − (3)(5) = −8 − 15 = −23.
5 2
 
−2 −5
(c) det = (−2)(−8) − (−5)(9) = 16 + 45 = 61.
9 −8
 
−10 −8
(d) det = (−10)(−12) − (−8)(−15) = 120 − 120 = 0.
−15 −12

285
286 Chapter 6. Determinants

3. [Exercise on page 55]

Determinant of an n × n Matrix – Definition

Before defining the determinant, we need to introduce some notation. Let A
be an n × n matrix with n ≥ 2. For 1 ≤ i ≤ n and 1 ≤ j ≤ n, define Aij as the
(n − 1) × (n − 1) matrix obtained
 from A by deleting the ith row and the jth
1 2 3
column. For example, if A = 4 5 6,

7 8 9
       
5 6 4 6 1 2 1 3
A11 = , A12 = , A23 = , A32 =
8 9 7 9 7 8 4 6
Now, we give a definition of the determinant.

Definition of the Determinant

ˆ Let n = 1. From the definition above, we know how to calculate
the determinant of every 1 × 1 matrix.

ˆ Let n ≥ 2 be an integer. Suppose that we know how to compute

 
the determinant of every (n − 1) × (n − 1) matrix. Let A = aij be
an n × n matrix. The determinant of A, denoted det A, is defined
as
det A = a11 C11 + a12 C12 + · · · + a1n C1n , (6.1.2)
where Cij = (−1)i+j det Aij .

The number Cij is called the (i, j)-cofactor of A.

Note.
ˆ Each Aij in the definition above is an (n − 1) × (n − 1) matrix. So the
determinant of an n × n matrix is defined by determinants of (n − 1) ×
(n − 1) matrices. Such a definition is said to be recursive.
ˆ Equation (6.1.2) can be rewritten as
det A = a11 det A11 − a12 det A12 + · · · + (−1)1+n det A1n
where the signs alternate.
ˆ Note that
 the equation
 (6.1.2) coincides with (6.1.1) when n = 2. Indeed,
a a
let A = 11 12 . Then, by (6.1.2), det A = a11 C11 + a12 C12 . But
a21 a22
C11 = (−1)1+1 det A11 = (−1)2 det a22 = (1)a22 = a22
 

C12 = (−1)1+2 det A12 = (−1)3 det a21 = (−1)a21 = −a21

 

So
det A = a11 C11 + a12 C12 = a11 (a22 ) + a12 (−a21 ) = a11 a22 − a12 a21
6.1. Introduction to Determinants 287

ˆ Alternate Notation. The determinant of A is also denoted by |A|.

 
3 4 2
(a) We want to compute the determinant of A =  −5 −3 1  using the defi-
2 −1 −2
nition. Here n = 3, a11 = 3, a12 = 4, and a13 = 2. So, by definition,

det A = a11 C11 + a12 C12 + a13 C13 = 3C11 + 4C12 + 2C13

We need to calculate the cofactors C11 , C12 , and C13 .

−3 1
ˆ C11 = (−1)1+1 det A11 = (−1)2 = (1) (6 − (−1)) = 6 + 1 = 7.
−1 −2

−5 1
ˆ C12 = (−1)1+2 det A12 = (−1)3 = (−1) (10 − 2) = −8.
2 −2

−5 −3
ˆ C13 = (−1)1+3 det A13 = (−1)4 = (1) (5 − (−6)) = 5 + 6 = 11.
2 −1
So the determinant of A is

det A = a11 C11 + a12 C12 + a13 C13 = 3C11 + 4C12 + 2C13
= 3(7) + 4(−8) + 2(11) = 21 − 32 + 22 = 11
 
−2 −5 3
(b) First we need to find the cofactors C11 , C12 , and C13 of A =  2 1 −4 .
6 −6 −7
1 −4
ˆ C11 = (−1)1+1 det A11 = (−1)2 = (1) (−7 − 24) = −31.
−6 −7

2 −4
ˆ C12 = (−1)1+2 det A12 = (−1)3 = (−1) (−14 + 24) = −10.
6 −7

2 1
ˆ C13 = (−1)1+3 det A13 = (−1)4 = (1) (−12 − 6) = −18.
6 −6
Using (6.1.2), we have

det A = a11 C11 + a12 C12 + a13 C13 = −2C11 + (−5)C12 + 3C13
= −2(−31) − 5(−10) + 3(−18) = 62 + 50 − 54 = 58

4. [Exercise on page 55]

Cofactor Expansions
Let A be an n × n matrix.
ˆ The sum (6.1.2) is called the cofactor expansion across the first row
288 Chapter 6. Determinants

of A.

ˆ For 1 ≤ i ≤ n, the cofactor expansion across the ith row of A is

defined by the sum

ai1 Ci1 + ai2 Ci2 + · · · + ain Cin

ˆ For every 1 ≤ j ≤ n, the cofactor expansion down the jth column

of A is defined by the sum

a1j C1j + a2j C2j + · · · + anj Cnj

The following theorem says that the determinant does not depend on the row
or column we choose for the cofactor expansion.

Theorem 6.1.1. Let A be an n × n matrix. Then for every 1 ≤ i ≤ n, for

every 1 ≤ j ≤ n, we have

det A = ai1 Ci1 + ai2 Ci2 + · · · + ain Cin

= a1j C1j + a2j C2j + · · · + anj Cnj

So the determinant of A can be computed by using the cofactor expansion

across any row or down any column of A.



1 2 3
(a) We want to compute the determinant of A =  2 4 1  using cofactor expan-
3 4 2
sion.
Step 1. Select one row or one column, preferably the one that contains
the most zeros. Let’s select row 1. Then
 
1 2 3
A= 2 4 1 
3 4 2

Step 2. Add signs to the row or column selected. The rule is that
the sign we add to the upper left corner is always +. And the signs
alternate across each row and down each column. Another way to
add the sign is to use the formula (−1)i+j : the sign we  add to the

1 2 3
 + − + 
entry aij is (−1)i+j . Adding the signs to row 1, we get  2 4 1 
3 4 2
Step 3. Expand det A across the row or down the column selected.
Here we selected row 1, so we are going to expand det A across that
row.
6.1. Introduction to Determinants 289

ˆ Deleting
  the first row and the first column, we get the matrix
4 1
. Taking the determinant of this latter matrix, we get
4 2
4 1
. Multiplying this by the entry at the intersection be-
4 2
4 1
tween the row and column we deleted, we get (1) . Mul-
4 2
4 1
tiplying all this by the sign we added, we get (+)(1) ,
4 2
which is the term obtained by deleting row 1 and column 1.
ˆ Similarly, the term obtained by deleting row 1 and column 2 is
2 1
(−)(2) .
3 2
ˆ The term obtained by deleting row 1 and column 3 is
2 4
(+)(3) .
3 4
The determinant of A is the sum of those three terms. That is,
4 1 2 1 2 4
det A = (+)(1) + (−)(2) + (+)(3)
4 2 3 2 3 4
= 1(8 − 4) − 2(4 − 3) + 3(8 − 12)
= 1(4) − 2(1) + 3(−4) = 4 − 2 − 12 = −10
Alternate Solution
We compute det A by expanding down a column. Let’s select column 2.
 
1 2 3
−
ˆ Adding the signs to column 2, we get  2 4 1 
 

+ 
3 4 2
−

ˆ – The term obtained by deleting column 2 and row 1 is

2 1
(−)(2) .
3 2
– The term obtained by deleting column 2 and row 2 is
1 3
(+)(4) .
3 2
– The term obtained by deleting column 2 and row 3 is
1 3
(−)(4) .
2 1
ˆ The determinant of A is
2 1 1 3 1 3
det A =(−)(2) + (+)(4) + (−)(4)
3 2 3 2 2 1
= − 2(4 − 3) + 4(2 − 9) − 4(1 − 6)
= − 2 − 28 + 20 = −10
290 Chapter 6. Determinants

(b) ˆ To take advantage of the presence of zero, we can expand det A across row 3
or downcolumn 1. Let’sselect column 1. Adding the signs to that column,
2 −3 1
+
−2 4 1 
 
we get 
 − 
0 −5 −1
+
ˆ Expanding det A down column 1, we have

4 1 −3 1 −3 1
det A = (+)(2) + (−)(−2) + (+)(0)
−5 −1 −5 −1 4 1
= 2[−4 − (−5)] + 2[3 − (−5)] + 0(−3 − 4)
= 2(−4 + 5) + 2(3 + 5) + 0 = 2 + 16 = 18
 
15 12 −6
(c) We want to find det A, where A =  7 0 0 
20 −1 1
ˆ To takeadvantage of the
 zeros, we select row 2. Adding the signs to row 2,
15 12 −6
we get  7 0 0 .

− + −
20 −1 1
ˆ Expanding det A across row 2, we get

12 −6 15 −6 15 12
det A = (−)(7) + (+)0 + (−)(0)
−1 1 20 1 20 −1
= −7(12 − 6) + 0 + 0 = −42
 
12 −3 0
(d) We want to find det A, where A =  77 0 −5 . Expending det A down
−8 −1 0
column 3, we have
77 0 12 −3 12 −3
det A = (+)(0) − (−5) + (+)(0)
−8 −1 −8 −1 77 0
= 0 + 5(−12 − 24) + 0 = −180

5. [Exercise on page 55]

(a) ˆ To take advantage of the zeros, we select column 3. Adding the signs to
that column, we get  
0 4 0 −5
+
 1 −6 −3 5 
 
 − 
 0 7 0 −2 
 
 + 
4 −6 0 2
−
6.1. Introduction to Determinants 291

ˆ Expanding det A down column 3, we get

0 4 −5
det A = 0 + (−)(−3) 0 7 −2 + 0 − 0
4 −6 2
0 4 −5
0 4 −5 +
0 7 −2 Select column 1 and
=3 0 7 −2 =3 −
add signs
4 −6 2 4 −6 2
+
 
4 −5
= 3 0 − 0 + (+)(4) Expand down column 1
7 −2
= 3 (4(−8 + 35)) = 3(4)(27) = 324

(b)

2 8 −1 6
−3 21 4 −4 Select row 3 and add
det A = 0 −2 0 0
+ − + − signs
5 19 −7 3
2 −1 6
= (−)(−2) −3 4 −4 Expand across row 3
5 −7 3

2 −1 6 2 −1 6
+ − + Select row 1 and add
= 2 −3 4 −4 = 2 −3 4 −4 signs
5 −7 3 5 −7 3

4 −4
= 2 (+)(2) +
−7 3

−3 −4 −3 4
(−)(−1) + (+)(6) Expand across row 1
5 3 5 −7
= 2 [2(12 − 28) + 1(−9 + 20) + 6(21 − 20)]
= 2 (−32 + 11 + 6) = 2(−15) = −30

(c)

−2 −1 1 2 0
+
1 3 2 1 −4
−
0 0 2 0 0 Select column 5 and
det A = + add signs
−1 1 5 6 0
−
2 1 6 3 0
+
292 Chapter 6. Determinants

−2 −1 1 2
0 0 2 0
= (−)(−4) Expand down column 5
−1 1 5 6
2 1 6 3

−2 −1 1 2 −2 −1 1 2
0 0 2 0 0 0 2 0 Select row 2 and add
=4 =4 − + − +
−1 1 5 6 −1 1 5 6 signs
2 1 6 3 2 1 6 3
 
−2 −1 2
= 4 0 + 0 + (−)(2) −1
 1 6 + 0 Expand across row 2
2 1 3

−2 −1 2 −2 −1 2
+ − + Select row 1 and add
= −8 −1 1 6 = −8 −1 1 6 signs
2 1 3 2 1 3

1 6
= −8 (+)(−2) +
1 3

−1 6 −1 1
(−)(−1) + (+)(2) Expand across row 1
2 3 2 1
= −8 [−2(3 − 6) + 1(−3 − 12) + 2(−1 − 2)]
= −8(6 − 15 − 6) = 120

(d)

3 0 7 −2 −4
−2 0 8 0 9
1 −6 −9 2 −8 Select row 5 and add
det A =
−5 0 6 −1 −3 signs
0 0 5 0 0
+ − + − +

3 0 −2 −4
−2 0 0 9
= (+)(5) Expand across row 5
1 −6 2 −8
−5 0 −1 −3
3 0 −2 −4
−
−2 0 0 9
+ Select column 2 and
=5
1 −6 2 −8 add signs
−
−5 0 −1 −3
+
 
3 −2 −4
= 5 (−)(−6) −2 0 9  Expand down column 2
−5 −1 −3
6.1. Introduction to Determinants 293

3 −2 −4
Select row 2 and add
= 30 −2 0
+
9
−
− signs
−5 −1 −3

−2 −4
= 30 (−)(−2) + 0+
−1 −3

3 −2
(−)(9) Expand across row 2
−5 −1
= 30 [2(6 − 4) − 9(−3 − 10)]
= 30(4 + 117) = 3630

6. [Exercise on page 56]

Determinants of Triangular Matrices

ˆ An upper triangular matrix is a square matrix where every entry

below the main diagonal is 0. For example, the following matrices are
upper triangular.
 
    −1 1 2 3
4 7 9 −5 6 0
 0 −2 −3  ,  0 0 −7  ,  0 −2 3 4 
 
 0 0 9 5 
0 0 8 0 0 −1
0 0 0 16

ˆ A lower triangular matrix is a square matrix where every entry above

the main diagonal is 0. For example, the following matrices are lower
triangular.
 
    10 0 0 0
4 0 0 0 0 0  4 13
 2 −2 0  ,  2 1 0 0 
0 ,   5

6 17 0 
−4 4 8 −7 0 −1
−3 −6 −19 23

ˆ A triangular matrix is a matrix that is upper triangular or lower tri-

angular.

The determinant of a triangular matrix is pretty easy to calculate as we can

see in the following theorem.

Theorem 6.1.2. Let A = [aij ] be an n × n matrix. Suppose A is a triangular

matrix. Then the determinant of A is the product of the diagonal entries. That
is,
det A = a11 a22 · · · ann
294 Chapter 6. Determinants

 
−5 13 −25
(a) Since A =  0 −4 37  is a triangular matrix, it follows (by Theorem 6.1.2)
0 0 3
that
det A = (−5)(−4)(3) = 60
 
3 0 0 0
 89 −7 0 0 
(b) Since A =  −58
 is a triangular matrix, it follows (by Theo-
7 −1 0 
61 −6 12 −2
rem 6.1.2) that
det A = (3)(−7)(−1)(−2) = −42

7. [Exercise on page 56]

(a) False because the determinant of a non-square matrix is undefined.

(b) True by definition.
(c) False by Theorem 6.1.1.
(d) True by Theorem 6.1.1
 
5 2
(e) False. Indeed, the determinant of is (5)(6) −(2)(−7) = 30+14 = 44.
−7 6
 
0 4 5
(f ) False. Indeed, expanding the determinant of  −5 1 2  down the first col-
0 2 3
umn, we get
4 5
(−)(−5) = 5(12 − 10) = 10
2 3
 
a b
(g) True. Indeed, if A = ,
c d

3a 3b
det(3A) = = (3a)(3d) − (3b)(3c)
3c 3d
= 9ad − 9bc = 9(ad − bc) = 9 det A
 
4 9 −6
(h) False. By definition, the matrix  0 2 −2  is upper triangular (not lower
0 0 −3
triangular!).
(i) True by Theorem 6.1.2.

6.2 Properties of Determinants

1. [Exercise on page 57]
6.2. Properties of Determinants 295

Determinants and Elementary Row Operations

Theorem 6.2.1. Let A be a square matrix.

ˆ If B is obtained from A by performing an elementary row operation of

the form Ri + αRj → Ri , then det B = det A. That is,
Ri +αRj →Ri
if A −→ B, then det B = det A

ˆ If B is obtained from A by performing an elementary row operation of

the form Ri ↔ Rj , then det B = − det A. That is,
Ri ↔Rj
if A −→ B, then det B = − det A

ˆ If B is obtained from A by performing an elementary row operation of

the form kRi , then det B = k det A. That is,
kRi →Ri
if A −→ B, then det B = k det A

Note. The only operation that does not affect the determinant is Ri + αRj →
Ri .

(a) If B is obtained from A by performing the operation R2 − R1 → R2 , then

det B = det A by Theorem 6.2.1.
(b) det B = det A.
(c) det B = − det A.
(d) det B = 3 det A.
(e) det B = 41 det A.

2. [Exercise on page 57]

(a) We have

A R1 ↔ R3 A1 det A1 = − det A
R2 + 6R1 → R2 A2 det A2 = det A1
−3R2 → R2 A3 det A3 = −3 det A2
R3 − 7R1 → R3 A4 det A4 = det A3
4 4
R
5 3
→ R3 B det B = det A4
5
Putting these together, we have
4 4
det B = det A4 Since det B = det A4
5 5
296 Chapter 6. Determinants

4
= det A3 Since det A4 = det A3
5
4
= (−3 det A2 ) Since det A3 = −3 det A2
5
12
= − det A1 Since det A2 = det A1
5
12
= − (− det A) Since det A1 = − det A
5
12 12
So det B = 5
det A. Dividing both sides of this latter equation by 5
, we get

5
det A = det B
12
5
(b) If det B = −60, then det A = 12
(−60) = −25.

3. [Exercise on page 57]

5 −10 10
(a) We want to compute −2 8 1 by using elementary row operations. The
3 −10 4
idea is to transform A into an echelon matrix.

   
5 −10 10 1 −2 2
1
A =  −2 8 1  R
5 1
→ R1  −2 8 1  = A1
3 −10 4 3 −10 4
 
1 −2 2
R2 + 2R1 → R2  0 4 5  = A2
R3 − 3R1 → R3
0 −4 −2
 
1 −2 2
R3 + R2 → R3  0 4 5  = A3
0 0 3

So
1
det A3 = det A2 = det A1 = det A
5
This implies that det A3 = 15 det A. Since A3 is a triangular matrix, det A3 is the
product of the entries on the main diagonal, that is, det A3 = (1)(4)(3) = 12.
Therefore, the equation det A3 = 15 det A becomes 51 det A = 12. Multiplying
both sides of this latter equation by 5, we get det A = 60.
(b) We transform A into an echelon matrix.
   
0 −5 −3 −8 1 −1 2 3
 −1 4 0 3   −1 4 0 3 
A=  1 −1
 R1 ↔ R3  0 −5 −3 −8  = A1
 
2 3 
2 2 4 13 2 2 4 13
6.2. Properties of Determinants 297

 
1 −1 2 3
R2 + R1 → R2  0
 3 2 6 
 = A2
R4 − 2R1 → R4  0 −5 −3 −8 
0 4 0 7
 
1 −1 2 3
 0 2
1
1 3
2 
R → R2  = A3
 
3 2
 0 −5 −3 −8 


0 4 0 7
 
1 −1 2 3
 0 2
R3 + 5R2 → R3 1 3
2 
 = A4
 
R4 − 4R2 → R4 1

 0 0 3
2 
8
0 0 −3 −1
 
1 −1 2 3
 0 1 23 2 
R4 + 8R3 → R4  = A5
 
0 13

 0 2 
0 0 0 15

So

1 1 1 1
det A5 = det A4 = det A3 = det A2 = det A1 = (− det A) = − det A
3 3 3 3

1
(15) = 5, it follows that 5 = − 13 det A. This implies


Since det A5 = (1)(1) 3
that det A = −15.

 
0 1 4 −2
 −3 7 −1 −5 
4. Let A =  . Compute det A. [Exercise on page 57]
 0 2 3 −2 
3 −7 6 8
 
0 1 4 −2
 −3 7 −1 −5 
To create more zeros in A =  , we perform the operation R4 +
 0 2 3 −2 
3 −7 6 8
R2 → R4 . We obtain
 
0 1 4 −2
 −3 7 −1 −5 
 =B
 0 2 3 −2 
0 0 5 3

Since the operation R4 + R2 → R4 does not affect the determinant, it follows that
det A = det B. We now compute the determinant of B by expanding down column
298 Chapter 6. Determinants

1.
0 1 4 −2
+
−3 7 −1 −5 1 4 −2
−
det B = = (−)(−3) 2 3 −2
0 2 3 −2
+ 0 5 3
0 0 5 3
−

1 4 −2
+
2 3 −2 Select column 1 again
=3 −
and add signs
0 5 3
+
 
3 −2 4 −2
= 3 (+)(1) + (−)(2) +0 Expand down column 1
5 3 5 3
= 3 (1(9 + 10) − 2(12 + 10)) = 3(19 − 44) = −75

Thus, det A = −75.

5. [Exercise on page 57]

Matrices with a Zero Determinant

Theorem 6.2.2. Let A be a square matrix that has a zero row (or a zero
column). Then det A = 0.
Proof. ˆ Suppose A has a row of zeros. By expanding det A across that
row, we get 0.
ˆ Likewise, if A has a zero column, we expand det A down that column.
We obtain det A = 0.

Theorem 6.2.3. Let A be a square matrix.

(a) If A has two identical rows. Then det A = 0.
(b) If A has two identical columns. Then det A = 0.
Proof. (a) Suppose A has two identical rows, say Ri1 and Ri2 . Let B be the
matrix obtained from A by performing the operation Ri2 − Ri1 → Ri2 .
Then B has a zero row since Ri2 and Ri1 are identical. By Theorem 6.2.2,
det B = 0. Since the operation Ri2 − Ri1 → Ri2 does not affect the
determinant, it follows that det B = det A. So det A = 0.
(b) Suppose A has two identical

columns. Then AT has two identical

rows.
This implies that det AT = 0 by part (a). Since det AT = det A by
Theorem 6.2.6, it follows that det A = 0.

In general we have the following result, which follows from Theorem 6.2.5 and
the Invertible Matrix Theorem.
6.2. Properties of Determinants 299

Theorem 6.2.4. Let A be a square matrix. Then det A = 0 if and only if the
columns (or rows) of A are linearly dependent.

 
5 −10 10
(a) Since the matrix A =  0 0 0  has a zero row (the second row), it follows
3 −10 4
that det A = 0.
 
4 −11 0
(b) Since the matrix A =  −7 9 0  has a zero column (the third column),
11 −10 0
it follows that det A = 0.
 
0 1 4 1
 −3 7 −1 7 
(c) Since the second and fourth columns of A = 
 0
 are identical,
2 3 2 
3 −7 6 −7
it follows that det A = 0
(d) Observe that the third row of A is a multiple of the first row. Using that we can
create a zero row:
   
0 −1 4 −2 0 −1 4 −2
 −3 7 −1 −5   R3 − 3R1 → R3  −3 7 −1 −5 

A= =B
 0 3 −12 6   0 0 0 0 
3 −7 6 8 3 −7 6 8

Since det B = 0 (because B has a zero row) and since det A = det B (because
the operation R3 − 3R1 → R3 does not affect the determinant), it follows that
det A = 0.

Alternate Solution
Since the third row of A is a multiple of the first row, it follows that the rows
of A are linearly dependent. So, by Theorem 6.2.4, det A = 0.

6. [Exercise on page 58]

ˆ To find det B, we will use elementary row operations. The idea is to carry B to
A:
   
2a 2b 2c a b c
1
B = d−a e−b f −c  R → R1  d − a e − b f − c  = B1
2 1
g + 3a h + 3b i + 3c g + 3a h + 3b i + 3c
 
a b c
R2 + R1 → R2 
d e f =A
R3 − 3R1 → R3
g h i
300 Chapter 6. Determinants

So det A = det B1 = 12 det B. This gives the equation 1

2
det B = det A. Since
det A = 7, it follows that

det B = 2 det A = 2(7) = 14

ˆ To find det C, we use the same strategy.

   
3d 3e 3f d e f
1
C = a+d b+e c+f  R → R1  a + d b + e c + f  = C1
3 1
g − 2d h − 2e i − 2f g − 2d h − 2e i − 2f
 
d e f
R2 − R1 → R2 
a b c  = C2
R3 + 2R1 → R3
g h i
 
a b c
R1 ↔ R2  d e f  = A
g h i

So  
1 1
det A = − det C2 = − det C1 = − det C = − det C.
3 3
This gives the equation det A = − 13 det C. Since det A = 7, it follows that

det C = −3 det A = −3(7) = −21

7. [Exercise on page 58]

Determinants and Invertible Matrices

Theorem 6.2.5. Let A be a square matrix. Then A is invertible if and only if
det A 6= 0.

 set S ={v1 , v2 , v3 } is linearly independent, we will consider

(a) To determine if the
the matrix A = v1 v2 v3 and use the Invertible Matrix Theorem. Let A =

2 1 4
 0 9 3  be the matrix whose jth column is vj . We compute det A:
1 2 3

2 1 4
+
0 9 3 9 3 1 4
det A = −
= (+)(2) − 0 + (+)(1)
2 3 9 3
1 2 3
+

= 2(27 − 6) + 1(3 − 36) = 2(21) − 33 = 9

Since det A 6= 0, it follows that A is invertible (thanks to Theorem 6.2.5). There-

fore, by the Invertible Matrix Theorem, the columns of A are linearly indepen-
dent. In other words, the set S is linearly independent.
6.2. Properties of Determinants 301

 
1 0 3 2
 −1 5 1 −8 
(b) Let A =  2
 be the matrix whose columns are v1 , v2 , v3 , and
4 3 12 
−3 −8 −3 −22
v4 . We need to find the determinant of A.
   
1 0 3 2 1 0 3 2
 −1 R2 + R1 → R2 
5 1 −8  0 5 4 −6 
A=   R3 − 2R1 → R3   = A1
2 4 3 12   0 4 −3 8 
R4 + 3R1 → R4
−3 −8 −3 −22 0 −8 6 −16
 
1 0 3 2
 0 5 4 −6 
R4 + 2R3 → R4 
 0 4 −3
 = A2
8 
0 0 0 0

So det A2 = det A1 = det A. Since A2 has a row of zeros, det A2 = 0. This

implies that det A = 0. So A is not invertible, and by the Invertible Matrix
Theorem, the columns of A are not linearly independent. In other words, the
set S = {v1 , v2 , v3 , v3 } is linearly dependent.

8. [Exercise on page 58]

 
k 0 3
To find the values of k for which A =  7 1 −8  is invertible, we need to find
2 0 k+1
det A. Expanding down column 2, we have

k 3
det A = (+)(1) = k(k + 1) − 6 = k 2 + k − 6 = (k − 2)(k + 3)
2 k+1

So det A = 0 if and only if k = 2 or k = −3. This implies (by Theorem 6.2.5) that A
is invertible if and only if k 6= 2 and k 6= −3.

9. [Exercise on page 58]

Multiplicative Property and Determinant of the Transpose and the

Inverse of A
Theorem 6.2.6. Let A and B be a square matrices. Then

(a) det(AB) = (det A)(det B).

(b) det AT = det A.

(c) If A is invertible, det (A−1 ) = 1

det A
.

As an immediate consequence of (a), we have

det (An ) = (det A)n for every integer n ≥ 1.

302 Chapter 6. Determinants

(a) Using Theorem 6.2.6-(a), we have

det(AB) = (det A)(det B) = (2)(−3) = −6

(b) Using Theorem 6.2.6-(a), we have

det A6 = (det A)6 = (2)6 = 64

(c)

det ABAT = det(A) det(B) det(AT )

det(AB) = (det A)(det B)
= det(A) det(B) det(A) det(AT ) = det A
= (2)(−3)(2) = −12

(d)

det(AT B −1 ) = det(AT ) det(B −1 ) det(AB) = (det A)(det B)

 
1 det AT = det A and
= det(A)
det B det (B −1 ) = det1 B
 
1 2
= (2) =−
−3 3

10. [Exercise on page 58]

The equation A2 = A implies that

det A2 = det A

(det A)(det A) = det A

(det A)(det A − 1) = 0

This latter equation implies that det A = 0 or det A = 1.

11. [Exercise on page 58]

   
1 1 1 1
Let A = and B = A = . (This is just one example among many others.)
0 1 0 1
 
2 2
ˆ A+B = and det(A + B) = (2)(2) = 4.
0 2
ˆ On the other hand, det A = 1, det B = 1, and det A + det B = 2.

We see that det(A + B) 6= det A + det B.

12. [Exercise on page 58]

(a) True by Theorem 6.2.1.

(b) True by Theorem 6.2.1.
(c) True. By Theorem 6.2.1, det B = 3 det A, which implies that det A = 13 det B.
6.2. Properties of Determinants 303

(d) True by Theorem 6.2.2.

 
1 1
(e) False. Indeed, let A = . The determinant of A is 1 − 1 = 0, but A has
1 1
neither a zero row nor a zero column.
(f ) True by Theorem 6.2.4.
 
1 2
(g) False. Indeed, the determinant of A = is det A = 4 − 4 = 0, but the
2 4
rows of A, as well as the columns, are not identical.
(h) True by Theorem 6.2.6-(a).
(i) False. A counterexample is provided in Exercise 10.
(j) True by Theorem 6.2.6-(c).
 
1 2
(k) False. Indeed, let A = . Clearly, AT = A. But det A is not equal to 0 as
2 3
det A = 3 − 4 = −1.
(l) True. Indeed, suppose A3 = A. Then det(A3 ) = det A or (det A)3 = det A.
Subtracting det A from both sides, we get (det A)3 − det A = 0. Factoring out
det A, we have det A ((det A)2 − 1) = 0. Factoring further, we get

det A(det A − 1)(det A + 1) = 0

This implies that det A = 0 or det A = 1 or det A = −1.

(m) True. Indeed, let A be a 3 × 3 matrix such that A3 = −A. Then det(A3 ) =
det(−A) or (det A)3 = det(−A). Since A is a 3 × 3 matrix, det(−A) =
(−1)3 det A (this fact can be proved by using cofactor expansion). So the equa-
tion (det A)3 = det(−A) becomes (det A)3 = − det(A). Adding det A to both
sides and factoring out det A, we have

det A[(det A)2 + 1] = 0

This latter equation implies that det A = 0.

Here we used the fact that det(−A) = (−1)3 det A. In general, we have the
following.

The Determinant of kA
Fact. Let A be an n × n matrix. Then one can prove that for every real
number k,
det(kA) = k n det A

 
1 0 −2
(n) False. Indeed, let A = 0 1 −4. The determinant of A is 0 since A has a
0 0 0
row of zeros. But the statement “for every b in R3 the system Ax = 0 has no
304 Chapter 6. Determinants

 
0
solution” is false. Indeed, if b =  0 , the system Ax = b has infinitely many
0
solutions (x1 = 2s, x2 = 4s, x3 = s, where s can be any real number).
 Chapter7

Eigenvectors and Eigenvalues

7.1 Eigenvectors and Eigenvalues

1. [Exercise on page 61]

Eigenvalues and Eigenvectors

Let A be a square matrix. Geometrically, a vector x is an eigenvector of A if Ax
is parallel to x, that is, Ax is a multiple of x. The factor by which x is stretched
is called eigenvalue. To be more precise, we have the following definition.

Definition
Let A be an n × n matrix. A vector x in Rn is called an eigenvector of
A if x 6= 0 and if there exists a scalar λ such that

Ax = λx

The scalar λ is referred to as an eigenvalue of A.

If Ax = λx, we say that x is an eigenvector corresponding to λ. We also say

that λ is the eigenvalue associated with x.

(a) ˆ To determine if u is an eigenvector of A, we need to compute Au, and see if

it is a multiple of u.
        
3 −3 −3 −9 − 6 −15 −3
Au = = = =5 = 5u
−2 2 2 6+4 10 2
Since Au = 5u, it follows that u is an eigenvector of A corresponding to the
eigenvalue λ = 5.
ˆ We now compute Av.
      
3 −3 1 3+3 6
Av = = =
−2 2 −1 −2 − 2 −4
   
6 1
Clearly, the vector Av = is not a multiple of v = . This
−4 −1
implies that v is not an eigenvector of A.

305
306 Chapter 7. Eigenvectors and Eigenvalues

(b) ˆ Computing Au, we get

      
1 0 −1 3 3+0−2 1
Au =  1 −3 0  0  =  3 + 0 + 0  =  3 
4 −13 1 2 12 + 0 + 2 14
Since Au is not a multiple of u, it follows that u is not an eigenvector of A.
ˆ Computing Av, we get
        
1 0 −1 1 1+0−3 −2 1
Av =  1 −3 0  1  =  1 − 3 + 0  =  −2  = −2  1 
4 −13 1 3 4 − 13 + 3 −6 3
Since Av = −2v, the vector v is an eigenvector of A corresponding to the
eigenvalue λ = −2.

2. Let A be an n × n matrix. [Exercise on page 61]

(a) The scalar λ = 0 is an eigenvalue of A if and only if there exists a nonzero vector
u such that Au = 0. This is equivalent to the fact that the equation Ax = 0
has a non trivial solution, which in turn is equivalent to the fact that A is not
invertible (thanks to the Invertible Matrix Theorem).
(b) We need to show that A(kx) = λ(kx). Since x is an eigenvector of A corre-
sponding to λ, Ax = λx. Using this, we have

A(kx) = kAx Property: A(kx) = kA(x)

= k(λx) Ax = λx
= λ(kx)

This proves that kx is an eigenvector of A corresponding to λ.

3. [Exercise on page 61]

Finding the Eigenvectors

Let A be a square matrix, and let λ be an eigenvalue of A. To find the eigen-
vectors of A corresponding to λ, we need to solve the equation Ax = λx.
Subtracting λx from both sides, we get Ax − λx = 0. Factoring out x, we get
(A − λI)x = 0, where I is the identity matrix with the same size as A. So the
eigenvectors of A corresponding to λ are the nontrivial solutions to
the homogeneous equation (A − λI)x = 0.

 
3 −3
ˆ First, we form the matrix A − λI, where A = , λ = 5, and I =
  −2 2
1 0
0 1
       
3 −3 1 0 3 − 5 −3 − 0 −2 −3
A − λI = −5 = = .
−2 2 0 1 −2 − 0 2−5 −2 −3
7.1. Eigenvectors and Eigenvalues 307

ˆ Consider the equation (A − 5I)x = 0. To solve this, we need to reduce the

augmented matrix.
   
−2 −3 0 −2 −3 0
R2 − R1 → R2
−2 −3 0 0 0 0
" #
3
1 0
− 12 R1 → R1 2
0 0 0

The corresponding system is x1 + 32 x2 = 0 and 0 = 0. Set s = x2 . Then the first

equation becomes x1 = − 32 s, and the general solution is
 " 3 #
−2s
  3 
x1 −2
x= = =s = sv,
x2 s 1
" #
− 32
where v = .
1
ˆ Thus, the eigenvectors of A corresponding to the eigenvalue λ = 5 are the
nonzero multiples of v.

4. [Exercise on page 61]

Eigenspaces
Let A be a square matrix, and let λ be an eigenvalue of A. The eigenspace
of A corresponding to λ is the set of all solutions to the homogeneous equation
(A − λI)x = 0. In other words, the eigenspace of A corresponding to λ is the
null space of the matrix A − λI, that is, Null(A − λI).
Note. Every vector in the eigenspace, except the zero vector, is an eigenvector
of A.

 
3 −3
(a) A = , λ = 5.
−2 2
 
3 −3
In Exercise 3 we considered the matrix A = and the eigenvalue
−2 2
λ = 5. And we showed that the  solutions to the equation (A − 5I)x = 0
 set3 of
−2
is the set of multiples of v = . This implies that the eigenspace of A,
1
Null(A − 5I), is the set of vectors of the form sv, where s is a real number.
So the set B = {v} spans Null(A − 5I). Moreover, since v 6= 0, B is linearly
independent. Thus, B is a basis for Null(A − 5I).
(b) ˆ First, we form the matrix A − (−1)I = A + I:
     
0 1 0 1 0 0 1 1 0
A+I = 3 0 1 + 0 1 0 = 3 1 1 
2 0 0 0 0 1 2 0 1
308 Chapter 7. Eigenvectors and Eigenvalues

ˆ Then we reduce the augmented matrix of the system (A + I)x = 0.

   
1 1 0 0 1 1 0 0
 3 1 1 0  R2 − 3R1 → R2  0 −2 1 0 
R3 − 2R1 → R3
2 0 1 0 0 −2 1 0
 
1 1 0 0
R3 − R2 → R3  0 −2 1 0 
0 0 0 0
 
1 1 0 0
− 12 R2 → R2  0 1 −1 0 
2
0 0 0 0
1
 
1 0 2
0
R1 − R2 → R1  0 1 −1 0 
2
0 0 0 0

ˆ The corresponding system is x1 + 12 x3 = 0 and x2 − 12 x3 = 0. Set x3 = s.

Then x1 = − 12 s, x2 = 12 s, and the general solution is
  −1s   1 
−2

x1 2
 1   1 
x = x2 =  2 s  = s  2 
 
x3 s 1
 1  
 −2 
 
ˆ So a basis for the eigenspace corresponding to λ = −1 is B =  2  .
 1 

 

1
(c) ˆ The matrix A − I is
     
0 3 1 1 0 0 −1 3 1
A − I =  1 −2 −1  −  0 1 0  =  1 −3 −1 
−3 9 4 0 0 1 −3 9 3

ˆ We reduce the augmented matrix of the homogeneous equation (A−I)x = 0:

   
−1 3 1 0 1 −3 −1 0
 1 −3 −1 0  R1 ↔ R2  −1 3 1 0 
−3 9 3 0 −3 9 3 0
 
1 −3 −1 0
R2 + R1 → R2  0 0 0 0 
R3 + 3R1 → R3
0 0 0 0

ˆ The corresponding system is x1 − 3x2 − x3 = 0. Set x2 = s and x3 = t.

Then x1 = 3s + t, and the general solution is
       
x1 3s + t 3 1
x = x2 =
   s  =s 1 +t 0 
  
x3 t 0 1
7.1. Eigenvectors and Eigenvalues 309

ˆ Thus, a basis for the eigenspace corresponding to λ = 1 is

   
 3 1 
B =  1 + 0 
0 1
 

5. [Exercise on page 62]

Let λ be an eigenvalue of A. Then there exists a nonzero vector x such that Ax = λx.
Multiplying both sides by A, we get

A(Ax) =A(λx)
=λAx
=λ(λx) = λ2 x

So A2 x = λ2 x. Since A2 = 0, it follows that λ2 x = 0. Since x 6= 0, this latter equation

implies that λ2 = 0, which in turn implies that λ = 0.

6. Let A be a square matrix. Let λ be an eigenvalue of A and let u be an eigenvector

corresponding to λ. Show that for every integer n ≥ 1, λn is an eigenvalue of An
associated with u. [Exercise on page 62]

Proof by Induction

Let P(n) be a statement depending on an integer n ≥ 1. To prove that P(n)

is true for all n ≥ 1, one can proceed as follows:

Step 1. Prove that P(1) is true.

Step 2. Assume that P(n − 1) is true and prove that P(n) is also true.

This process is called proof by induction.

Here P(n) is
λn is an eigenvalue of An associated with u

ˆ By hypothesis, λ is an eigenvalue of A associated with u. So P(1) is true.

ˆ Suppose that P(n − 1) is true. Then λn−1 is an eigenvalue of An−1 associated
with u. That is,
An−1 u = λn−1 u
We want to prove that λn is an eigenvalue of An associated with u. By left-
multiplying both sides of the equation An−1 u = λn−1 u by A, we get

AAn−1 u = A(λn−1 u)
An = λn−1 (Au)
An = λn−1 (λu) Since Au = λu
An = λn u

So P(n) is true.
310 Chapter 7. Eigenvectors and Eigenvalues

Thus, if λ is an eigenvalue of A associated with u, then for every n ≥ 1, λn is an

eigenvalue of An associated with u.
7. [Exercise on page 62]
Suppose that λ is an eigenvalue of A with corresponding eigenvector x. Since A is
invertible, it follows that λ 6= 0 (thanks to Exercise 2). Now, we have
Ax = λx
A Ax = A−1 (λx)
−1
Left-multiply both sides by A−1
x = λA−1 x A−1 A = I
Since λ 6= 0, we can
λ−1 x = A−1 x
divide both sides by λ

This proves that λ−1 is an eigenvalue of A−1 with the same eigenvector x.
8. [Exercise on page 62]
Consider the following figure.

x2
x2 = x1

u v
x1

x2 = −x1

ˆ Geometrically, if x is vector that lies along the line x2 = −x1 , then T (x) = x
(since T is the reflection through the line x2 = −x1 ). So every nonzero vector
on the line x2 = −x1 is an eigenvector and the associated eigenvalue is λ = 1.
ˆ Again from the figure, if x is perpendicular to u, then T (x) = −x. (For example,
T (v) = −v.) So every nonzero vector on the line x2 = x1 is an eigenvector and
the associated eigenvalue is λ = −1.
ˆ If x is neither on the line x2 = −x1 nor on the line x2 = x1 , then T (x) is not a
multiple of x.

Thus, the eigenvalues of T are 1 and −1. The eigenspace corresponding to λ = 1 is

the line of equation x2 = −x1 . And the eigenspace corresponding to λ = −1 is the
line of equation x2 = x1 .
9. [Exercise on page 62]
Since T is the projection onto the x1 x2 -plane, T maps every vector u in the x1 x2 -plane
to itself, that is, T (u) = u. So an eigenvalue of T is λ = 1. And the corresponding
eigenspace is nothing but the x1 x2 -plane.
7.2. Finding Eigenvalues 311

10. [Exercise on page 62]

π
Since T is the counterclockwise rotation of angle ,
T (x) is not a multiple of x for
2
2
every vector x in R . (Every vector will get rotated.) So T has no eigenvector.
11. [Exercise on page 62]
(a) True by definition.
(b) False by definition.
(c) True. Indeed,
      
7 4 −2 −14 + 12 −2
Au = = = =u
−3 −1 3 6−3 3
Since Au is a multiple of u, it follows that u is an eigenvector of A.
   
−11 −1
(d) False because Au = is not a multiple of u = .
14 2
(e) True by the definition of eigenspace.

7.2 Finding Eigenvalues

1. [Exercise on page 62]

Finding Eigenvalues – Characteristic Equation

The key result to find the eigenvalues is the following theorem.
Theorem 7.2.1. Let A be an n × n matrix. Then a scalar λ is an eigenvalue
of A if and only if
det(A − λI) = 0
Proof. ˆ Suppose that λ is an eigenvalue of A. Then there exists a nonzero
vector u in Rn such that Au = λu. This implies that Au − λu = 0. Since
λu = λIu, the latter equation becomes

Au − λIu = 0
(A − λI) u = 0 Right-factor out u

So the homogeneous equation (A − λI)x = 0 has a nontrivial solution.

This implies (by the Invertible Matrix Theorem) that the matrix A − λI
is not invertible, which in turn implies that det(A − λI) = 0.
ˆ Conversely, suppose that det(A − λI) = 0. Then A − λI in not invertible,
which implies that the system (A − λI)x = 0 has a nontrivial solution,
say u. So

(A − λI)u = 0
Au − λu = 0
Au = λu
312 Chapter 7. Eigenvectors and Eigenvalues

Since u 6= 0, the latter equation tells us that λ is an eigenvalue of A

associated with u.
This ends the proof.
It turns out that det(A−λI) is a polynomial in λ of degree n. This polynomial is
called the characteristic polynomial of A. And the equation det(A−λI) = 0
is called the characteristic equation of A.
So to find the eigenvalues of A, we first need to find the characteristic polyno-
mial. Then we solve the equation det(A − λI) = 0 for λ. The solutions are the
eigenvalues of A.

 
5 4
(a) To determine if λ = 2 is an eigenvalue of A = , we need to find
−6 −6
det(A − λI). First, we form the matrix A − λI:
       
5 4 1 0 5 4 2 0
A − λI = A − 2I = −2 = −
−6 −6 0 1 −6 −6 0 2
   
5−2 4 3 4
= =
−6 −6 − 2 −6 −8

Taking the determinant of this, we have

3 4
det(A − 2I) = = (3)(−8) − (4)(−6) = −24 + 24 = 0
−6 −8

Since det(A − 2I) = 0, it follows that λ = 2 is an eigenvalue of A.

(b) First, we have
     
−1 2 1 0 3 2
A − λI = A − (−4)I = +4 =
2 −1 0 1 2 3

Taking the determinant of this, we get

3 2
det(A − (−4)I) = =9−4=5
2 3

Since det(A − (−4)I) 6= 0, it follows that λ = −4 is not an eigenvalue of A.

(c)

0−7 3 0
det(A − λI) = det(A − 7I) = 3 0−7 0
4 −5 7 − 7
−7 3 0
= 3 −7 0 =0 Expand down column 3
4 −5 0

Since det(A − 7I) = 0, it follows that λ = 7 is an eigenvalue of A.

7.2. Finding Eigenvalues 313

2. [Exercise on page 62]

(a) ˆ First, we form the matrix A − λI.

     
5 −3 1 0 5 − λ −3
A − λI = −λ =
−10 6 0 1 −10 6 − λ

ˆ The characteristic polynomial is

5 − λ −3
det(A − λI) = = (5 − λ)(6 − λ) − (−3)(−10)
−10 6 − λ
= 30 − 5λ − 6λ + λ2 − 30 = λ2 − 11λ = λ(λ − 11)

ˆ Set det(A − λI) = 0. Then λ(λ − 11) = 0. This implies that λ = 0 or

λ = 11. So the eigenvalues of A are λ1 = 0 and λ2 = 11.
(b) We will use the following theorem.

The Eigenvalues of a Triangular Matrix

Theorem 7.2.2. The eigenvalues of a triangular matrix A are the diag-
onal entries of A.
Proof. The proof follows from the fact that the determinant of a triangular
matrix is the product of the entries on the main diagonal.

 
4 7 −5
Since A =  0 −2 5  is a triangular matrix, the eigenvalues are the diag-
0 0 3
onal entries, that is, 4, −2, and 3.
 
3 0 0
(c) Since A =  −1 2 0  is a triangular matrix, the eigenvalues are the entries
8 5 3
on the main diagonal of A, that is, 2 and 3.
(d) ˆ First, we find A − λI:
     
0 4 0 1 0 0 −λ 4 0
A − λI =  1 0 0  − λ  0 1 0  =  1 −λ 0 
0 0 1 0 0 1 0 0 1−λ

ˆ Using the cofactor expansion across the third row, we get

−λ 4 0
1 −λ 0 −λ 4
det(A − λI) = = 0 − 0 + (+)(1 − λ)
0 0 1−λ 1 −λ
+ − +

= (1 − λ) λ2 − 4 = (1 − λ)(λ − 2)(λ + 2)

ˆ So the eigenvalues of A are −2, 1, and 2.

314 Chapter 7. Eigenvectors and Eigenvalues

(e) The characteristic polynomial is

−1 − λ 1 4
det(A − λI) = 0 2−λ 0
3 −3 −2 − λ
−1 − λ 4 Cofactor expansion
= (2 − λ)
3 −2 − λ across row 2
= (2 − λ)[(−1 − λ)(−2 − λ) − 12]
= (2 − λ)(2 + λ + 2λ + λ2 − 12)
= (2 − λ)(λ2 + 3λ − 10)
= (2 − λ)(λ − 2)(λ + 5) = −(λ − 2)2 (λ + 5)
So the eigenvalues are 2 and −5.
(f ) The characteristic polynomial is
−λ 3 1
+ − +
det(A − λI) = 1 −2 − λ −1
−3 9 4−λ
−2 − λ −1 1 −1 1 −2 − λ
= −λ −3 +
9 4−λ −3 4 − λ −3 9
= −λ(−8 + 2λ − 4λ + λ2 + 9) − 3(4 − λ − 3) + (9 − 6 − 3λ)
= −λ(λ2 − 2λ + 1) − 3(−λ + 1) + (3 − 3λ)
= −λ3 + 2λ2 − λ + 3λ − 3 + 3 − 3λ
= −λ3 + 2λ2 − λ
= −λ(λ2 − 2λ + 1)
= −λ(λ − 1)2
So the eigenvalues of A are 0 and 1.

3. [Exercise on page 63]

Because the eigenvalues of a triangular matrix are the diagonal entries (by Theo-
rem 7.2.2), the matrix  
1 0 0
A= 0 2 0 
0 0 3
has the required property.
4. [Exercise on page 63]
 
a b
(a) Let A = be a 2 × 2 matrix. We want to show that the characteristic
c d
polynomial of A is
det(A − λI) = λ2 − tr(A)λ + det A
7.2. Finding Eigenvalues 315

By definition, the characteristic polynomial of A is

a−λ b
det(A − λI) =
c d−λ
= (a − λ)(d − λ) − bc = ad − λa − λd + λ2 − bc
= λ2 − (a + d)λ + ad − bc
Since tr(A) = a + d
= λ2 − tr(A)λ + det A
and det A = ad − bc
(b) Let λ1 and λ2 be the eigenvalues of A. Then λ1 and λ2 are solutions to the
equation det(A − λI) = 0. This implies that the characteristic polynomial of A
is of the form
det(A − λI) = (λ − λ1 )(λ − λ2 )
Distributing this, we get
det(A − λI) = λ2 − (λ1 + λ2 )λ + λ1 λ2
In part (a), we proved that det(A − λI) = λ2 − tr(A)λ + det A. So
λ2 − (λ1 + λ2 )λ + λ1 λ2 = λ2 − tr(A)λ + det A
This implies that λ1 + λ2 = tr(A) and λ1 λ2 = det A. (Here we used the fact that
two polynomials are equal if they have the same coefficients.)

The results we just proved can be generalized to larger matrices.

Sum and Product of Eigenvalues

Let A = [aij ] be an n × n matrix. The trace of A, denoted tr(A), is the

sum of the entries on the main diagonal of A. That is,

tr(A) = a11 + a22 + · · · + ann

Theorem 7.2.3. Let λ1 , · · · , λn be the eigenvalues of A. Then

λ1 + λ2 + · · · + λn = tr(A) and (λ1 )(λ2 ) · · · (λn ) = det A

Some Facts About the Characteristic Polynomial

Theorem 7.2.4. Let A be an n × n matrix. Then

i. the coefficient of λn in the characteristic polynomial of A is (−1)n .

ii. The coefficient of λn−1 in the characteristic polynomial of A is

(−1)n−1 tr(A).

iii. The constant term of the characteristic polynomial of A is det A.

316 Chapter 7. Eigenvectors and Eigenvalues

(c) Let λ1 and λ2 be the eigenvalues of A. If λ1 = 3 and tr(A) = −5, then (by
part (b)) λ1 + λ2 = −5, that is, 3 + λ2 = −5. Subtracting 3 from both sides of
the latter equation, we find λ2 = −8.
(d) If the characteristic polynomial of A is λ2 − λ, then det(A − λI) = λ2 − λ + 0.
This implies that det A = 0 since the determinant of A is the constant term of
the characteristic polynomial (by part (a)). So A is not invertible.

5. [Exercise on page 63]

Symmetric Matrices
Let A be a square matrix. We say that A is a symmetric matrix if it is equal
to its transpose, that is, if A = AT .

 
a b
Let A = . Suppose A is symmetric. Then A = AT , that is
c d
   
a b a c
=
c d b d
 
a b
This implies that b = c. So A = . We want to show that all the eigenvalues of
b d
A are real numbers. First, the characteristic polynomial of A is
a−λ b
det(A − λI) =
b d−λ
= (a − λ)(d − λ) − b2 = ad − (a + d)λ + λ2 − b2
= λ2 − (a + d)λ + ad − b2

Using the quadratic formula, the solutions to the equation det(A − λI) = 0 are
p √
(a + d) ± (a + d)2 − 4(ad − b2 ) (a + d) ± a2 + 2ad + d2 − 4ad + 4b2
=
2 2
√
(a + d) ± a − 2ad + d2 + 4b2
2
=
2
p
(a + d) ± (a − d)2 + 4b2
=
2
p
Since (a − d)2 + 4b2 is always greater than or equal to 0, (a − d)2 + 4b2 is always
a real number. This implies that the solutions to the equation det(A − λI) = 0 are
real numbers. In other words, the eigenvalues of A are real numbers.

Caution. We just proved that if A is a 2 × 2 symmetric matrix, then the eigenvalues

of A are real numbers. The converse to this statement is not true. That is, if all
the eigenvalues of A are real numbers, the matrix A is not necessary symmetric. In-
2 4
deed, let A = . Since A is a triangular matrix, its eigenvalues are the diagonal
0 3
7.2. Finding Eigenvalues 317

entries; that is, 2 and 3. So the eigenvalues of A are real numbers. But A is not a
symmetric matrix since AT 6= A.

The result we proved above can be generalized to larger matrices.

Eigenvalues of Symmetric Matrices are Real Numbers

Theorem 7.2.5. Let A be an n × n matrix. If A is a symmetric matrix, then
all the eigenvalues of A are real numbers.

6. [Exercise on page 63]

 
0 0
(a) True. Indeed, the eigenvalues of A = are 0 and 1. So an eigenvalue can
0 1
be 0.  
4 2 −3
(b) True. Indeed, observe that the rows of A =  −1 5 −2  are linearly de-
3 7 −5
pendent (as row 3 = row 2 + row 1). This implies that det A = 0, that is,
det(A − 0I) = 0. Therefore, λ = 0 is an eigenvalue of A.
(c) True by Theorem 7.2.1.
(d) True by definition.
(e) True. Indeed, if the columns (or rows) of A are linearly dependent, then det A =
0. This implies that det(A − 0I) = 0, that is, λ = 0 is an eigenvalue of A.
(f ) True by Theorem 7.2.5.
 
3 4
(g) False. Indeed, all the eigenvalues of A = are real numbers (λ1 = 3 and
0 5
λ2 = 5). But A is not a symmetric matrix since AT 6= A.
7. [Exercise on page 64]

Multiplicity of an Eigenvalue
Let A be a square matrix and let λ be an eigenvalue of A. The multiplicity
of λ is m if λ occurs m times as a root of det(A − λI). For example, if
det(A − λI) = (λ − 1)3 (λ − 2)5 , then the eigenvalue 1 has multiplicity 3 and
the eigenvalue 2 has multiplicity 5.

(a) We know that if A is of size n × n, the degree of the characteristic polynomial is

n. Since the degree of the polynomial (λ − 5)1 (λ − 2)4 (λ + 1)3 is 1 + 4 + 3 = 8,
it follows that A is of size 8 × 8.
(b) Set det(A − λI) = 0. Then (λ − 5)(λ − 2)4 (λ + 1)3 = 0. This implies that λ = 5
or λ = 2 or λ = −1. So the eigenvalues of A are 5, 2, and −1.
(c) Since 0 is not an eigenvalue of A, it follows that det(A − 0I) 6= 0, that is,
det(A) 6= 0. This implies that A is invertible.
318 Chapter 7. Eigenvectors and Eigenvalues

7.3 Diagonalization
1. [Exercise on page 64]

Diagonal Matrices
A square matrix is called diagonal if all its non-diagonal entries are 0. It is
very easy to do computations with diagonal matrices.
 
a11 0 ··· 0
 0 a22 · · · 0 
Theorem 7.3.1. Let D =  .. ..  be a diagonal matrix. Then
 
.. . .
 . . . . 
0 0 · · · ann
for every integer k ≥ 1, we have
 
ak11 0 ··· 0
 0 ak · · · 0 
k 22
D =  ..
 
.. . . .. 
 . . . . 
k
0 0 · · · ann

Diagonalization – Definition
Let A be an n × n matrix. We say that A is diagonalizable if there exist a
diagonal matrix D and an invertible matrix P such that

A = P DP −1

(a) False because there is a nonzero entry off the main diagonal (the (2, 1)-entry,
4, is nonzero).
(b) True because all the entries of D off the main diagonal are zeros.
(c) True by Theorem 7.3.1.
(d) True by definition.
(e) False since only square matrices can be diagonalizable.
(f ) True. Indeed, suppose A is a diagonal matrix and let D = A. Let P = I be the
identity matrix of the same size as A. Since I −1 = I, it follows that

P DP −1 = IDI = D = A

So A is diagonalizable.

2. [Exercise on page 64]

7.3. Diagonalization 319

Multiplicity of an Eigenvalue
Let A be a square matrix and let λ be an eigenvalue of A. The multiplicity
of λ is m if λ occurs m times as a root of det(A − λI). For example, if
det(A − λI) = (λ − 2)5 (λ − 3)6 (λ − 4), then the eigenvalue 2 has multiplicity
5, the eigenvalue 3 has multiplicity 6, and the eigenvalue 4 has multiplicity 1.

Diagonalization–Theorem
To determine whether a matrix is diagonalizable, we will use the following
theorem.

Theorem 7.3.2. Let A be a square matrix that has only real eigenvalues. Then
A is diagonalizable if and only if for each eigenvalue λ,

dim Null(A − λI) = multiplicity of λ (7.3.1)

The following is a consequence of Theorem 7.3.2.

Theorem 7.3.3. An n × n matrix that has n distinct eigenvalues is diagonal-

izable.

 
9 6
(a) A =
−7 −4
 
9 6
We want to diagonalize A = , if possible.
−7 −4
ˆ Step 1. Find the characteristic polynomial and the eigenvalues.
The characteristic polynomial is

9−λ 6
det(A − λI) = = (9 − λ)(−4 − λ) − (6)(−7)
−7 −4 − λ
= −36 − 9λ + 4λ + λ2 + 42 = λ2 − 5λ + 6 = (λ − 2)(λ − 3)

So the eigenvalues are λ1 = 2 and λ2 = 3.

ˆ Step 2. Find a basis and the dimension of each eigenspace. Recall
that the eigenspace corresponding to an eigenvalue λ is Null(A − λI), that
is, the set of solutions to the homogeneous equation (A − λI)x = 0.
– Finding the eigenspace corresponding to λ1 = 2. Consider the equation
(A − 2I)x = 0. The coefficient matrix of this system is
     
9 6 1 0 7 6
A − 2I = −2 =
−7 −4 0 1 −7 −6

We reduce the augmented matrix.

   
7 6 0 7 6 0
R2 + R1 → R2
−7 −6 0 0 0 0
320 Chapter 7. Eigenvectors and Eigenvalues

" #
1
1 76 0
R
7 1
→ R1
0 0 0

The corresponding system is x1 + 76 x2 = 0. Set x2 = s. Then x1 = − 67 s,

and the general solution is
" #
− 67
   6 
x1 −7s
x= = =s
x2 s 1

where s can be any real number. So a basis

" for #
the eigenspace corre-
6
−7
sponding to λ1 is the set {v}, where v = . This implies that
1
dim Null(A − 2I) = 1.

Using a basis involving no fraction

We know that if B = {u1 , · · · , up } is a basis for a subspace H,

then for every nonzero scalar k, {ku1 , · · · , kup } is also a basis for
H. Here the set {v} is a basis for Null(A − 2I). So any set of the
form {kv}, where k 6= 0, is also a basis for Null(A − 2I). Taking
for example k = 7, we have
 6   
−7 −6
7v = 7 =
1 7
 
−6
So the set B1 = {v1 }, where v1 = is another basis for
7
Null(A − 2I). We will use this basis since computations are easier
with integers than with fractions.

– Finding the eigenspace corresponding to λ2 = 3. First we need to form

the matrix A − 3I.
     
9 6 1 0 6 6
A − 3I = −3 =
−7 −4 0 1 −7 −7

We now solve the system (A − 3I)x = 0.

1
R → R1
   
6 6 0 6 1 1 1 0
−7 −7 0 − 71 R2 → R2 1 1 0
 
1 1 0
R2 − R1 → R2
0 0 0

The corresponding system is x1 + x2 = 0. Set x2 = s. Then x1 = −s

and the general solution to the equation (A − 3I)x = 0 is
     
x1 −s −1
x= = =s = sv2 ,
x2 s 1
7.3. Diagonalization 321

 
−1
where v2 = . So a basis for Null(A − 3I) is B2 = {v2 }, and
1
dim Null(A − 3I) = 1.
ˆ Step 3. Check if condition (7.3.1) from Theorem 7.3.2 is satisfied
for each λ, and decide whether A is diagonalizable. For every eigen-
value λ, we need to check if its multiplicity is equal to the dimension of the
corresponding eigenspace. Recall the characteristic polynomial:
det(A − λI) = (λ − 2)(λ − 3) = (λ − 2)1 (λ − 3)1
– For λ1 = 2, the multiplicity is m1 = 1. From Step 2, dim Null(A−I) = 1.
So condition (7.3.1) is satisfied for λ1 .
– Clearly, condition (7.3.1) is also satisfied for λ2 = 3.
Thus, A is diagonalizable.

Alternate way to see that A is diagonalizable

Since A is a 2 × 2 matrix that has two distinct eigenvalues, it follows
that A is diagonalizable (thanks to Theorem 7.3.3).

ˆ Step 4. If A is diagonalizable, find P and D. (Do not compute

P −1 , unless you are asked to do so.)
– Constructing P . The columns of P are the basic eigenvectors
found in Step 2. That is, the columns of P are v1 and v2 . (The order
of the vectors in not important.) Using the order chosen in Step 2, P is
the matrix whose first column is v1 and second column is v2 , that is,
 
  −6 −1
P = v1 v2 =
7 1
– Constructing D. The entries on the main diagonal of D are
the eigenvalues. Here the order of the eigenvalues corresponds
 to the
a11 0
order chosen for the columns of P . That is, if D = , then
0 a22
a11 is the eigenvalue associated with the eigenvector v1 . And a22 is the
eigenvalue associated with the eigenvector v2 . So
   
λ1 0 2 0
D= =
0 λ2 0 3
Let’s check if A = P DP −1 (this step is not mandatory). Using the formula
 −1  
a b 1 d −b
= ,
c d ad − bc −c a
 
1 1
we have P −1 = 1 1
. Computing P DP −1 , we get
−7 −6
     
−1 −6 −1 2 0 1 1 9 6
P DP = = =A
7 1 0 3 −7 −6 −7 −4
322 Chapter 7. Eigenvectors and Eigenvalues

 
2 −3
(b) A =
3 8
ˆ Finding the eigenvalues. The characteristic polynomial is

2 − λ −3
det(A − λI) = = (2 − λ)(8 − λ) − (−3)(3)
3 8−λ
= 16 − 2λ − 8λ + λ2 + 9 = λ2 − 10λ + 25 = (λ − 5)2

So A has only one eigenvalue: λ = 5.

ˆ Finding the eigenspace corresponding to λ = 5. First, we form the matrix
A − 5I.      
2 −3 1 0 −3 −3
A − 5I = −5 =
3 8 0 1 3 3
We now solve the system (A − 5I)x = 0.

− 13 R1 → R1
   
−3 −3 0 1 1 0
3 3 0 1 1 1 0
R → R2
3 2
 
1 1 0
R2 − R1 → R2
0 0 0

The corresponding system is x1 + x2 = 0. Set x2 = s. Then x1 = −s, and

the general solution is
     
x1 −s −1
x= = =s = sv,
x2 s 1
 
−1
where v = . So a basis for Null(A − 5I) is B = {v}. This implies
1
that dim Null(A − 5I) = 1.
ˆ Since λ = 5 has multiplicity m = 2, and since dim Null(A − 5I) 6= m, it
follows that A is not diagonalizable. Therefore it is not possible to find
an invertible matrix P and a diagonal matrix D such that A = P DP −1 .
 
−2 5 5
(c) A =  0 3 0 
2 −2 1
ˆ Finding the eigenvalues of A. The characteristic polynomial is

−2 − λ 5 5
det(A − λI) = 0 3−λ 0
2 −2 1 − λ
−2 − λ 5
= (3 − λ) Expand across row 2
2 1−λ
= (3 − λ) [(−2 − λ)(1 − λ) − (5)(2)]
= (3 − λ)(−2 + 2λ − λ + λ2 − 10)
7.3. Diagonalization 323

= (3 − λ)(λ2 + λ − 12)
= (3 − λ)(λ − 3)(λ + 4) = −(λ − 3)2 (λ + 4)
So the eigenvalues of A are λ1 = 3 and λ2 = −4.
ˆ Finding the eigenspaces.
– Finding the eigenspace corresponding to the eigenvalue λ1 = 3. First,
we need to form the matrix A − 3I:
     
−2 5 5 1 0 0 −5 5 5
A − 3I =  0 3 0  − 3 0 1 0  =  0 0 0 
2 −2 1 0 0 1 2 −2 −2
Then we need to solve the system (A − 3I)x = 0. Row reducing the
augmented matrix, we have
   
−5 5 5 0 1
− 5 R1 → R1 1 −1 −1 0
 0 0 0 0   0 0 0 0 
1
2 −2 −2 0 2
R 3 → R3 1 −1 −1 0
 
1 −1 −1 0
R3 − R1 → R3  0 0 0 0 
0 0 0 0
The corresponding system is x1 − x2 − x3 = 0. Set x2 = s and x3 = t.
Then x1 = s + t, and therefore the general solution is
       
x1 s+t 1 1
x = x2 =
   s  = s 1 + t 0  = sv1 + tv2 ,
  
x3 t 0 1
   
1 1
where v1 =  1  and v2 =  0 . So a basis for Null(A − 3I) is
0 1
B1 = {v1 , v2 }.
– Finding the eigenspace corresponding to the eigenvalue λ2 = −4. First,
we have
     
−2 5 5 1 0 0 2 5 5
A−(−4)I = A+4I =  0 3 0  +4  0 1 0  =  0 7 0 
2 −2 1 0 0 1 2 −2 5
Now, we solve the system (A + 4I)x = 0. Row reducing the augmented
matrix, we have

   
2 5 5 0 1 2 5 5 0
 0 → R2
R
7 2
7 0 0   0 1 0 0 
R3 − R1 → R3
2 −2 5 0 0 −7 0 0
 
2 5 5 0
R3 + 7R2 → R3  0 1 0 0 
0 0 0 0
324 Chapter 7. Eigenvectors and Eigenvalues

1 25 25
 
0
1
R
2 1
→ R1  0 1 0 0 
0 0 0 0
1 0 25
 
0
R1 − 52 R2 → R1  0 1 0 0 
0 0 0 0

Set x3 = s. Then x1 = − 52 s and x2 = 0. So the general solution is

 5 
−2
   5 
x1 −2s
x = x2 =
   0  =s  0 
x3 s 1

− 52
 

To avoid fractions, we can multiply the vector  0  by 2. We obtain

  1
−5
the vector v3 =  0 . So a basis for Null(A + 4I) is B2 = {v3 }.
2
ˆ Since
dim Null(A − λ1 I) = 2 = multiplicity of λ1 ,
and since
dim Null(A − λ2 I) = 1 = multiplicity of λ2
it follows that A is diagonalizable.
ˆ The matrices P and D are
 
  1 1 −5
P = v1 v2 v3 = 1 0 0 
0 1 2
   
λ1 0 0 3 0 0
D =  0 λ1 0  =  0 3 0 
0 0 λ2 0 0 −4
 
0 3 1
(d) A =  1 −2 −1 
−3 9 4
ˆ Finding the eigenvalues of A. From Exercise 2f from Section 7.2, the char-
acteristic polynomial is

det(A − λI) = −λ(λ − 1)2

So the eigenvalues are λ1 = 0 and λ2 = 1.

ˆ Finding the corresponding eigenspaces.
7.3. Diagonalization 325

– Finding the eigenspace corresponding to λ1 = 0. We first need to solve

the system (A − 0I)x = 0, that is, Ax = 0.
   
0 3 1 0 1 −2 −1 0
 1 −2 −1 0  R1 ↔ R2  0 3 1 0 
−3 9 4 0 −3 9 4 0
 
1 −2 −1 0
R3 + 3R1 → R3  0 3 1 0 
0 3 1 0
 
1 −2 −1 0
R3 − R2 → R3  0 3 1 0 
0 0 0 0
1 −2 −1
 
0
1 1
R → R2
3 2
 0 1 3
0 
0 0 0 0
 
1 0 − 31 0
1
R1 + 2R2 → R1  0 1 0 
 
3
0 0 0 0

The corresponding system is x1 − 13 x3 = 0 and x2 + 13 x3 = 0. Set x3 = s.

Then x1 = 13 s and x2 = − 31 s, and the general solution is
  1   1 
s

x1 3 3
x =  x2  =  − 31 s  = s  − 31  = sv,
   
x3 s 1
 1

3
where v =  − 13 . To avoid fractions, we can multiply v by 3. We
 
1
 
1
obtain v1 =  −1 . So a basis for Null(A − 0I) is B1 = {v1 }. This
3
implies that dim Null(A − 0I) = 1.
– Finding the eigenspace corresponding to λ2 = 1. We first form A − I.
     
0 3 1 1 0 0 −1 3 1
A − I =  1 −2 −1  −  0 1 0  =  1 −3 −1 
−3 9 4 0 0 1 −3 9 3

We now solve the system (A − I)x = 0.

   
−1 3 1 0 −1 3 1 0
 1 −3 −1 0  R2 + R1 → R2  0 0 0 0 
R3 − 3R1 → R3
−3 9 3 0 0 0 0 0
326 Chapter 7. Eigenvectors and Eigenvalues

 
1 −3 −1 0
(−1)R1 → R1  0 0 0 0 
0 0 0 0
The corresponding system is x1 − 3x2 − x3 = 0. Set x2 = s and x3 = t.
Then x1 = 3s + t, and the general solution is
       
x1 3s + t 3 1
x =  x2  =  s  = s  1  + t  0  = sv2 + tv3 ,
x3 t 0 1
   
3 1
where v2 =  1  and v3 =  0 . So a basis for Null(A − I) is
0 1
B2 = {v2 , v3 }. This implies that dim Null(A − I) = 2.
ˆ Since for each eigenvalue λ the dimension of the corresponding eigenspace
is equal to the multiplicity of λ, it follows that A is diagonalizable.
ˆ The matrices P and D are
 
  1 3 1
P = v1 v2 v3 =  −1 1 0 
3 0 1
   
λ1 0 0 0 0 0
D =  0 λ2 0  =  0 1 0 
0 0 λ2 0 0 1
 
2 3 0
(e) A = 3 2
 0 
2 3 −1
ˆ Finding the eigenvalues of A.
2−λ 3 0
det(A − λI) = 3 2−λ 0
2 3 −1 − λ
2−λ 3
= (−1 − λ) Expand down column 3
3 2−λ
= (−1 − λ) (2 − λ)2 − 9
 

= (−1 − λ) [(2 − λ − 3)(2 − λ + 3)] a2 − b2 = (a − b)(a + b)

= (−1 − λ)(−λ − 1)(−λ + 5)
− (λ + 1)2 (λ − 5)
So the eigenvalues are λ1 = −1 and λ2 = 5.
ˆ Finding the eigenspaces. To find the eigenspace corresponding to the eigen-
value λ1 = −1, we need to solve the system (A − (−1)I)x = 0, that is,
(A + I)x = 0. The coefficient matrix is
     
2 3 0 1 0 0 3 3 0
A+I = 3 2 0  + 0 1 0 =  3 3 0 
2 3 −1 0 0 1 2 3 0
7.3. Diagonalization 327

Row reducing the augmented matrix, we have

   
3 3 0 0 1
R → R1 1 1 0 0
 3 3 3 1
0 0  1
 1 1 0 0 
2 3 0 0 R → R2
3 2 2 3 0 0
 
1 1 0 0
R2 − R1 → R2  0 0 0 0 
R3 − 2R1 → R3
0 1 0 0
 
1 1 0 0
R2 ↔ R3  0 1 0 0 
0 0 0 0
 
1 0 0 0
R1 − R2 → R1  0 1 0 0 
0 0 0 0

The corresponding system is x1 = 0 and x2 = 0. Let x3 = s. Then the

general solution is
     
x1 0 0
x = x2 = 0 = s 0 
    
x3 s 1
 
0
So a basis for Null(A + I) is the {v1 }, where v1 = 0 .

1
ˆ Since a basis for Null(A+I) contains one vector, it follows that dim Null(A+
I) = 1. On the other hand, the multiplicity of λ1 is 2. So

dim Null(A − λ1 I) 6= multiplicity of λ1

This implies that A is not diagonalizable. (We don’t need to find the
eigenspace corresponding to λ2 = 5 since A is not diagonalizable.)
ˆ Since A is not diagonalizable, it is not possible to find P and D such that
A = P DP −1 .

3. [Exercise on page 64]

(a) ˆ Finding the characteristic polynomial and eigenvalues.

−1 − λ 2 −2
det(A − λI) = 0 1−λ 0
− + −
4 −4 5 − λ
−1 − λ −2
=(1 − λ)
4 5−λ
=(1 − λ) [(−1 − λ)(5 − λ) − (−2)(4)]
328 Chapter 7. Eigenvectors and Eigenvalues

=(1 − λ)(−5 + λ − 5λ + λ2 + 8)
=(1 − λ)(λ2 − 4λ + 3)
=(1 − λ)(λ − 1)(λ − 3) = −(λ − 1)2 (λ − 3)

So the eigenvalues are λ1 = 1 and λ2 = 3.

ˆ Finding the corresponding eigenspaces.
– Finding the eigenspace corresponding to λ1 = 1. First, we form the
matrix A − I.
     
−1 2 −2 1 0 0 −2 2 −2
A−I = 0 1 0 − 0 1 0 = 0 0 0 
4 −4 5 0 0 1 4 −4 4

We now solve the system (A − I)x = 0. Row reducing the augmented

matrix, we get

   
−2 2 −2 0 − 21 R1 → R1 1 −1 1 0
 0 0 0 0   0 0 0 0 
1
4 −4 4 0 4 3
R → R3 1 −1 1 0
 
1 −1 1 0
R3 − R1 → R3  0 0 0 0 
0 0 0 0
The corresponding system is x1 − x2 + x3 = 0. Set x2 = s and x3 = t.
Then x1 = s − t, and the general solution is
       
x1 s−t 1 −1
x =  x2  =  s  = s  1  + t  0  = sv1 + tv2 ,
x3 t 0 1
   
1 −1
where v1 =  1  and v2 =  0 . So a basis for Null(A − I) is
0 1
{v1 , v2 }. This implies that dim Null(A − I) = 2.
– Finding the eigenspace corresponding to λ2 = 3. The matrix A − 3I is
     
−1 2 −2 1 0 0 −4 2 −2
A − 3I =  0 1 0  − 3  0 1 0  =  0 −2 0 
4 −4 5 0 0 1 4 −4 2

We solve the system (A − 3I)x = 0.

   
−4 2 −2 0 −4 2 −2 0
 0 −2 R3 + R1 → R3
0 0   0 1 0 0 
− 12 R2 → R2
4 −4 2 0 0 −2 0 0
 
−4 2 −2 0
R3 + 2R2 → R3  0 1 0 0 
0 0 0 0
7.3. Diagonalization 329

1 − 12 21
 
0
− 14 R1 → R1  0 1 0 0 
0 0 0 0
1 0 21
 
0
R1 + 21 R2 → R1  0 1 0 0 
0 0 0 0

The corresponding system is x1 + 12 x3 = 0 and x2 = 0. Set x3 = s. Then

x1 = − 12 s, and the general solution is
 1 
−2
   1 
x1 −2s
x = x2 =
   0 =s
  0  = sv,
x3 s 1
 1 
−2
where v =  0 . To avoid fraction, we can multiply v by 2. We
 1 
−1
obtain v3 =  0 . So a basis for Null(A − 3I) is {v3 }. This implies
2
that dim Null(A − 3I) = 1.
ˆ Since the dimension of the eigenspace corresponding to each eigenvalue λ is
equal to the multiplicity of λ, it follows that A is diagonalizable.
The matrices P and D are
   
  1 −1 −1 1 0 0
P = v1 v2 v3 =  1 0 0  and D =  0 1 0 
0 1 2 0 0 3

(b) We will use the following result.

Powers of a Matrix
Theorem 7.3.4. Let P be an invertible matrix and let
 
a11 0 · · · 0
 0 a22 · · · 0 
D =  ..
 
.. 
 . . 
0 0 · · · ann

be a diagonal matrix Then for every k ≥ 1,

ˆ Dk is the diagonal matrix whose ith diagonal entry is akii , that is,
 
ak11 0 · · · 0
 0 ak · · · 0 
k 22
D =  .. and
 
.. 
 . . 
0 0 · · · aknn
330 Chapter 7. Eigenvectors and Eigenvalues

k
ˆ (P DP −1 ) = P Dk P −1 .

Theorem 7.3.4 is used to find the powers of a matrix. If A is diagonalizable,

then there exist an invertible matrix P and a diagonal matrix D such that
A = P DP −1 . This implies that

Ak = P Dk P −1 for every k ≥ 1

Using Theorem 7.3.4, we have Ak = P Dk P −1 , where

 k  k   
1 0 0 1 0 0 1 0 0
D k =  0 1 0  =  0 1k 0  =  0 1 0 
0 0 3 0 0 3k 0 0 3k
To find P −1 , we reduce the augmented matrix P I .
 
   
1 −1 −1 1 0 0 1 −1 −1 1 0 0
 1 0 0 0 1 0  R2 − R1 → R2  0 1 1 −1 1 0 
0 1 2 0 0 1 0 1 2 0 0 1
 
1 −1 −1 1 0 0
R3 − R2 → R3  0 1 1 −1 1 0 
0 0 1 1 −1 1
 
1 −1 0 2 −1 1
R2 − R3 → R2 
0 1 0 −2 2 −1 
R1 + R3 → R1
0 0 1 1 −1 1
 
1 0 0 0 1 0
R1 + R2 → R1  0 1 0 −2 2 −1 
0 0 1 1 −1 1
So  
0 1 0
P −1 =  −2 2 −1 
1 −1 1
We can now compute Ak = P Dk P −1 . First, we compute P Dk :
    
1 −1 −1 1 0 0 1 −1 −3k
P Dk =  1 0 0  0 1 0  =  1 0 0 
k
0 1 2 0 0 3 0 1 2 · 3k
We now have
 k
 
1 −1 −3 0 1 0
Ak = P Dk P −1 = P Dk P −1 =  1


0 0   −2 2 −1 
0 1 2 · 3k 1 −1 1
 
2 − 3k −1 + 3k 1 − 3k
= 0 1 0 
k k k
−2 + 2 · 3 2 − 2 · 3 −1 + 2 · 3
7.3. Diagonalization 331

4. [Exercise on page 65]

Since A is of size 2 × 2, and since A has two distinct eigenvalues, A is diagonalizable
(by Theorem 7.3.3), and
   
−1
  4 −1 1 0
A = P DP , where P = v1 v2 = and D =
5 4 0 −2

ˆ The product P D is
    
4 −1 1 0 4 2
PD = =
5 4 0 −2 5 −8

ˆ Finding the inverse of P . The determinant of P is det P = 16 + 5 = 21, and the

inverse of P is   " 4 1 #
1 4 1 21 21
P −1 = = −5
21 −5 4 4
21 21

ˆ The matrix A is
 " 4 1
# " 2 4
#
4 2 21 21 7 7
A = P DP −1 = =
5 −8 −5 4 20 −9
21 21 7 7

5. [Exercise on page 65]

(a) The size of A is n × n, where n is the degree of the characteristic polynomial.

Since the degree of λ1 (λ − 2)3 (λ + 7)4 is 1 + 3 + 4 = 8, A is of size 8 × 8.
(b) The eigenvalues of A are λ1 = 0 (with multiplicity m1 = 1), λ2 = 2 (with
multiplicity m2 = 3), and λ3 = −7 (with multiplicity m3 = 4).
(c) Let Ei , 1 ≤ i ≤ 3, be the eigenspace corresponding to λi . In order for A to be
diagonalizable, the dimension of Ei has to be equal to mi for every i. So A is
diagonalizable if and only if dim E1 = 1, dim E2 = 3, and dim E3 = 4.
(d) Since λ = 0 is an eigenvalue of A, it follows (by definition) that there exists a
nonzero vector u such that Au = 0u = 0. This implies that the system Ax = 0
has a nontrivial solution, which in turn implies that A is not invertible (thanks
to the Invertible Matrix Theorem).

6. [Exercise on page 65]

We will use the following result.

Dimension of Eigenspaces
Theorem 7.3.5. Let A be a matrix. Let λ be an eigenvalue of A with multi-
plicity m. Then the dimension of the eigenspace corresponding to λ is greater
than or equal to 1 and less than or equal to m. That is,

1 ≤ dim Null(A − λI) ≤ m

332 Chapter 7. Eigenvectors and Eigenvalues

Let A be a 4 × 4 matrix with two distinct eigenvalues, say λ1 and λ2 . Suppose that
each eigenspace is two-dimensional. We want to determine if A is diagonalizable. Let
m1 (respectively m2 ) be the multiplicity of λ1 (respectively λ2 ). Then, since each
eigenspace is two-dimensional,
dim Null(A − λ1 I) = 2 and dim Null(A − λ2 I) = 2
Using Theorem 7.3.5, we have
1 ≤ 2 ≤ m1 and 1 ≤ 2 ≤ m2
So
m1 ≥ 2 and m2 ≥ 2 (7.3.2)
On the other hand, the degree of the characteristic polynomial is 4 since A is a 4 × 4
matrix. Since the sum of the multiplicities is equal to the degree of the characteristic
polynomial, it follows that
m1 + m2 = 4 (7.3.3)
Combining (7.3.2) and (7.3.3), we get m1 = 2 and m2 = 2. So
dim Null(A − λ1 I) = m1 and dim Null(A − λ2 I) = m2
This implies that A is diagonalizable.
7. [Exercise on page 65]
(a) True by Theorem 7.3.3.
(b) True. This is true because of the following theorem.

Eigenvector Bases of Rn
Theorem 7.3.6. Let A be an n × n matrix. Then A is diagonalizable if
and only if Rn has a basis of eigenvectors of A.

(c) False because the dimension of Null(A − λ1 I) is not equal to the multiplicity of
λ1 .
(d) True. Indeed, suppose that A is invertible and diagonalizable. Then, by def-
inition, there exist an invertible matrix P and a diagonal matrix D such that
A = P DP −1 . Taking the inverse of both sides of this latter equation, we get

−1
A−1 = P DP −1

−1 −1 −1
A−1 = P −1 D P (ABC)−1 = C −1 B −1 A−1
A−1 = P D−1 P −1 (P −1 )−1 = P
A−1 = P QP −1 Where Q = D−1
Since the inverse of a diagonal matrix is still a diagonal matrix, it follows that Q
is a diagonal matrix. Therefore, the equation A−1 = P QP −1 implies that A−1
is diagonalizable.
7.3. Diagonalization 333

 
2 −3
(e) True. For example, the matrix A = from Exercise 2b is invertible
3 8
(since det A = 16 + 9 = 25 6= 0), but not diagonalizable.
 
1 0
(f ) True. Indeed, let A = . The matrix A is a diagonal matrix. So it is
0 0
diagonalizable (in Exercise 1f–page 318, we proved that any diagonal matrix is
diagonalizable). But A is not invertible since det A = 0.
(g) True. Indeed, suppose that A is diagonalizable. Then, by definition, there exist
an invertible matrix P and a diagonal matrix D such that A = P DP −1 . Taking
the transpose of both sides of this latter equation, we get

T
AT = P DP −1

T
AT = P −1 DT P T (ABC)T = C T B T AT
T −1
h
i−1
T −1

−1
T
T T
PT = P −1


A = P D P

−1
AT = RQR−1 Where R = P T and Q = DT

Since the transpose of a diagonal matrix is still a diagonal matrix,
it follows that
−1
Q is a diagonal matrix. Since P is invertible, it follows that P T = R is also
invertible. Thus, the equation AT = RQR−1 implies that AT is diagonalizable.