Question Bank

Complex Variables and

Partial Differential Equations
(MAT3003)

School of Advanced Sciences

Vellore Institute of Technology
Vellore-632014
Tamil Nadu, India
 Question Bank

Complex Variables and

Partial Differential Equations
(MAT3003)

School of Advanced Sciences

Vellore Institute of Technology
Vellore-632014
Tamil Nadu, India

July 11, 2020

Contents

1 Analytic functions 1

1.1 Differentiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Harmonic Conjugates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.4 Analytic Functions in Flow Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Conformal Mapping 7

2.1 Geometrical Representation of Complex Functions . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 Conformal Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 The Squared Mapping w = z 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.4 The Exponential Function ez . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Bilinear Transformations 13

3.1 Linear Fractional Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.2 Finding a Blinear Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.3 Cross Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.4 Images under a Bilinear Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4 Complex Integration 17

4.1 Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4.2 Contour integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 Cauchy’s Integral Formula 19

5.1 Simply and Multiply Connected Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

5.2 Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

5.3 Cauchy’s Integral Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Taylor’s Series, Laurent’s Series and Residues 23

6.1 Taylor’s Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

6.2 Taylor’s Series for Rational Polynomial Functions . . . . . . . . . . . . . . . . . . . . . . . . 24

6.3 Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

6.4 Laurent Series about an Isolated Singularity . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

6.5 Laurent Series about a Point of Analyticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

iii
iv CONTENTS

6.6 Types of Isolated Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

6.7 Formula for the Residue at a Pole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

7 Contour Integration through Cauchy’s Residue Theorem 31

7.1 Cauchy’s Residue Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

8 Real Integrals through Residue Theorem 35

8.1 Cauchy’s Residue Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

8.2 Type 1 - Integrals around the Unit Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

8.3 Type 2 - Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

8.4 Type 3 - Fourier Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

8.5 Type 4 - Using Indented Contours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

9 Partial Differential Equations and their Formation 47

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

9.2 Formation of a Partial Differential Equation - Elimination of Arbitrary Constants . . . . . . . 48

9.3 Formation of a Partial Differential Equation - Elimination of Arbitrary Functions . . . . . . . 50

10 Partial Differential Equations of First Order 53

10.1 Solution of a Partial Differential Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

10.2 Quasi-linear Partial Differential Equations of First Order . . . . . . . . . . . . . . . . . . . . 53

10.3 Solution of Lagrange’s Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

10.4 Nonlinear Partial Differential Equations of First Order . . . . . . . . . . . . . . . . . . . . . . 57

10.5 Special forms of Nonlinear First Order Partial Differential Equations . . . . . . . . . . . . . . 58

11 Linear Partial Differential Equations of Higher Order 65

11.1 Linear Equations of Second Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

11.2 Linear Equations of Higher Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

11.3 Finding the Complementary Function zc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

11.4 Finding the Particular Integral z p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

12 Applications of Partial Differential Equations 71

12.1 Vibrating Strings - One dimensional Wave Equation . . . . . . . . . . . . . . . . . . . . . . . 71

12.2 Heat-Flow with Homogeneous Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . 78

12.3 Heat-Flow with nonhomogeneous Boundary Conditions . . . . . . . . . . . . . . . . . . . . 86

Department of Mathematics, Vellore Institute of Technology

 CONTENTS v

12.4 Two Dimensional Steady State Heat Flow in Metal Plates . . . . . . . . . . . . . . . . . . . . 87

13 Complex Fourier Transform 93

13.1 Complex Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
13.2 Properties of the Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
13.3 Fourier Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

14 Fourier Sine and Cosine Transforms 101

14.1 Fourier Sine and Cosine Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

15 Parseval’s Identities 105

15.1 Parseval’s Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Department of Mathematics, Vellore Institute of Technology

vi CONTENTS

Department of Mathematics, Vellore Institute of Technology

Course Title: Complex Variables and Partial Differential Equations
Course Code: MAT3003 - Theory

Syllabus and Organization of Contents

Module 1 Analytic Functions

Complex variables - Analytic functions and Cauchy-Riemann equations - Laplace equation and harmonic
functions - Construction of harmonic conjugate and analytic functions - Applications of analytic functions
to fluid-flow and field problems

Chapter 1 of the Question Bank

Module 2 Conformal Mappings and Bilinear Transformations

Conformal mapping - Elementary transformations: translation, magnification, rotation, inversion, Expo-
nential and squared transformations (ω = ez , z 2 ) - Bilinear transformation and Cross-ratio - Images of the
regions bounded by straight lines under the above transformations

Chapters 2 and 3 of the Question Bank

Module 3 Power Series

Functions given by power series - Taylor’s and Laurent’s series - Singularities - Poles - Residues

Chapter 6 of the Question Bank

Module 4 Complex Integration

Integration of a complex function along a contour - Cauchy-Goursat theorem - Cauchy,s integral formula -
Cauchy’s residue theorem - Evaluation of real integrals - Indented contour integral

Chapters 4, 5, 7 and 8 of the Question Bank

Module 5 Partial Differential equations of First Order

Formation and solution of partial differential equation - General, Particular, Complete and Singular inte-
grals - Partial Differential equations of first order of the forms: F(p, q) = 0, F(z, p, q) = 0, F(x, p) = G(y, q)
and Clairaut’ss form - Lagrange’s equation: Pp + Qq = R

Chapters 9 and 10 of the Question Bank

Module 6 Applications of Partial Differential Equations

Linear partial differential equations of higher order with constant coefficients - Solution of a partial dif-
ferential equation by separation of variables - Boundary Value Problems: One dimensional wave and heat
equations - Fourier series solution

Chapters 11 and 12 of the Question Bank

Module 7 Fourier transforms
Complex Fourier transform and its properties - Relation between Fourier and Laplace transforms - Fourier
sine and cosine transforms - Convolution Theorem and Parseval’s identities

Chapters 13, 14 and 15 of the Question Bank

Chapter 1

Analytic functions

1.1 Differentiability
Let S ⊂ C be a path-connected open set and z0 ∈ S. A mapping f : S → C is said to be differentiable at z0 ,
if f 0(z0 ) = limz→z0 f (z)− f (z0 ) 0
z−z0 , provided this limit exists, and f (z0 ) is called the derivative of f at z0 .
f (z)− f (z0 )
Theorem 1.1.1 (Nonexistence of the derivative). If L = z−z0 is different along different paths of
approach as z → z0 , then the derivative f 0(z0 ) of f does not exist at z0 .

EXERCISE 1.1.1. Show that f (z) = f (x + iy) = x is not differentiable at z = 0.

Ans. Let L = f (z)− f (0)

z−0 = x+iy
x
· Along the real axis L = 1 and along the imaginary axis, L = 0. That is,
L is different along different paths of approach as z → 0. Hence, f 0(0) does not exist. That is f is not
differentiable at z = 0.
Theorem 1.1.2 (Cauchy-Riemann equations). Consider f (z) = u(z) + iv(z) and z0 = (x0, y0 ). If f 0(z0 )
exists, then
∂u ∂v ∂u ∂v
∂x = ∂y and ∂y = − ∂x at (x0, y0 ). (1.1.1)

Conditions (1.1.1) are called the Cauchy-Riemann equations.

Remark 1.1.1. If the Cauchy-Riemann equations are not satisfied at a point, f cannot be differentiable
at that point. Thus the conditions (1.1.1) are necessary for a function f (z) to be differentiable at a point z0 .
Remark 1.1.2. Cauchy-Riemann equations (1.1.1) are not sufficient for the differentiability of f at a
point z0 . That is, a complex function f may satisfy the Cauchy-Riemann equations at a point, without
being differentiable at that point.

EXERCISE 1.1.2. Show that

   
z̄ 2 z̄ 3 x 3 −3xy 2 y 3 −3x 2 y
(
= = x 2 +y 2
+i x 2 +y 2
if z , 0
f (z) = z |z | 2
0, if z = 0

f is not differentiable at z = 0, though the Cauchy-Riemann equations are satisfied there at.

∂u ∂u ∂v ∂v
Ans. Using the definition of the partial derivatives, ∂x (0, 0) = 1, ∂y (0, 0) = 0, ∂x = 0, ∂y (0, 0) = 1. But
f (z)− f (0) z̄ 2
L= z−0 = is different along different linear paths y = mx of approach as z → 0. Therefore, f is
z2
not differentiable at z = 0.

EXERCISE 1.1.3. Show that each of the following functions f (z) satisfies the Cauchy Riemann equations
at the origin, but f 0(0) does not exist:

( xy 2 (x+iy)
x 2 +y 4
, z,0
(a) f (z) =
0, z=0

( x 2 y 5 (x+iy)
x 4 +y 10
, z,0
(b) f (z) =
0, z=0

1
2 1.2. Analytic Functions

( x 3 y(y−ix)
x 6 +y 2
, z,0
(c) f (z) =
0, z=0

x +y
( x 3 −y 3  3 3
x 2 +y 2
+i x 2 +y 2
, z,0
(d) f (z) =
0, z=0

Hints. To examine the differentiability at the origin, choose the path x = my 2 in (a), x 2 = my 5 in (b),
y = mx 3 in (c), y = mx in (d)

EXERCISE 1.1.4. Show that the complex function f (z) = x − iy is nowhere differentiable.

Hint. Verify if the Cauchy-Riemann equations (1.1.1) fail to hold good at any point z = (x, y).
Theorem 1.1.3. Consider f (z) = u(z) + iv(z) and z0 = (x0, y0 ) ∈ C. Suppose that

∂u ∂u ∂v ∂v
(a) the partial derivatives ∂x , ∂y , ∂x and ∂y are continuous at (x0, y0 ), and
∂u ∂v ∂u ∂v
(b) satisfy the Cauchy-Riemann equations: ∂x = ∂y and ∂y = − ∂x .

Then f is differentiable at z0 , and its derivative is given by

∂u ∂v
f 0(z0 ) = ∂x (x0, y0 ) + i ∂x (x0, y0 ). (1.1.2)

1.2 Analytic Functions

A function f is said to be analytic at a point z0 ∈ S, if it is differentiable on some neighborhood of z0 . A
function w = f (z) is analytic on S, if it is analytic at every point of S.
Remark 1.2.1. Everywhere differentiable functions are analytic on C. Conversely, everywhere analytic
functions are everywhere differentiable. A function which is analytic on the complex plane C is called an
entire function.

EXERCISE 1.2.1. Identify the real and imaginary parts of the following functions, show that they are
entire, and find the derivatives at any point z = (x, y) of the complex plane:

(a) f (z) = z 2

(b) f (z) = ez = exp(z)

(c) f (z) = sin z

(d) f (z) = cos z

2
(e) f (z) = ez
z
(f) f (z) = ee

(g) f (z) = z 3

Ans.
The real and imaginary parts are respectively

(a) u = x 2 − y 2 and v = 2xy; f 0(z) = 2z

Department of Mathematics, Vellore Institute of Technology

 Chapter 1. Analytic functions 3

(b) u = e x cos y and v = e x sin y; f 0(z) = ez

(c) u = sin x cosh y and v = cos x sinh y; f 0(z) = cos z

(d) u = cos x cosh y and v = sin x sinh y; f 0(z) = − sin z

2 −y 2 2 −y 2 2
(e) u = e x cos 2xy and v = e x sin 2x y; f 0(z) = 2zez
x x z
(f) u = ee cos y cos(e x sin y) and v = ee cos y sin(e x sin y); f 0(z) = ez ee

(g) u = x 3 − 3x y 2 and v = 3x 2 y − y 3 ; f 0(z) = 3z 2

EXERCISE 1.2.2. Justify why the following functions are analytic at all points of the complex plane
without using the Cauchy-Riemann equations. Also find the derivative f 0 z(z) in each case:

(a) f (z) = (z 2 − 2)e−z

(b) f (z) = 2z 2 − 3 − zez + e−z

(c) f (z) = 3z 2 + 5z − 6i

(d) f (z) = x + sin x cosh y + i(y + cos x sinh y)

(e) f (z) = 4x 2 + 5x − 4y 2 + 9 + i(8xy + 5y − 1)

EXERCISE 1.2.3. Identify the real and imaginary parts of f (z) = e1/z . Whatabout its analyticity?

2 +y 2 )
  2 +y 2 )
 
y y
Ans. The real and imaginary parts are u = e x/(x cos x 2 +y 2
and v = −e x/(x sin x 2 +y 2
. For
z , 0, f is analytic.
Theorem 1.2.1 (Cauchy-Riemann Equations in Polar Form). If f (z) = u(r, θ)+iv(r, θ) is differentiable
at z = reiθ , 0, then
∂u 1 ∂v ∂v
∂r = = − r1 ∂u
r ∂θ , ∂r ∂θ , and (1.2.1)
   
f 0(z) = (cos θ − i sin θ) ∂u∂r + i ∂v
∂r =
cos θ−i sin θ
r
∂v
∂θ − i ∂u
∂θ · (1.2.2)

EXERCISE 1.2.4. If f (z) = z n , show that f 0(z) = nz n−1 where n ≥ 2.

EXERCISE 1.2.5. Show that The principal branch: F(z) = log x 2 + y 2 + i tan−1 (y/x) of the
p
p complex
logarithm f (z) = log z is analytic and hence find its derivative at all points z = x + iy ∈ C with x 2 + y 2 >
0 and − π < tan−1 (y/x) < π.

Ans.
At all the points z = reiθ with r > 0, −π < θ < π, the derivative is
 
f 0(z) = (cos θ − i sin θ) ∂u + ∂v
∂r = e r + i.0 = 1/re = 1/z.
−iθ 1 iθ


∂r i

EXERCISE 1.2.6. Find the choice of the constants (mentioned in brackets) such that each of the following
functions is entire:

(a) f (z) = x + ay + i(bx + cy), (a, b, c)

(b) f (z) = (ax + y) + i(3y − bx), (a, b)

(c) f (z) = e x [cos ay + i sin(y + b)] + c, (a, b, c)

Department of Mathematics, Vellore Institute of Technology

4 1.3. Harmonic Conjugates

(d) f (z) = x 2 + ax y + by 2 + i(cx 2 + dx y + y 2 ), (a, b, c, d)

(e) f (z) = (3x − y + 5) + i(ax + b − 3), (a, b)

EXERCISE 1.2.7. Show that each of the following functions f (z) is analytic in an appropriate domain,
and find its derivative in the domain:

x 3 +xy 2 +x
   
y x 2 y+y 3 −y
(a) x−1
(x−1)2 +y 2
−i (x−1)2 +y 2
[(b)] x 2 +y 2
+i x 2 +y 2

Remark 1.2.2. Analyticity at a point implies the differentiability at a point, but the converse is not true.
In fact, there are functions which are differentiable at a point but are nowhere analytic

EXERCISE 1.2.8. Show that f (z) = |z| 2 = x 2 + y 2 and g(z) = x 2 −ixy are differentiable only at the origin
but nowhere analytic:

EXERCISE 1.2.9. Show that f (z) = eiz̄ , g(z) = 2x + ix y 2 and l(z) = sin (z̄) are nowhere differentiable
and hence nowhere analytic.

EXERCISE 1.2.10. Examine the analyticity of f (z) = cos x − i sinh y and g(z) = (z̄ + 1)3 − 3z̄.
EXERCISE 1.2.11. Show that each of the following functions is nowhere analytic but is differentiable
along the indicated curve(s):

(a) f (z) = x 2 + y 2 + i2xy, the real axis

(b) f (z) = 3x 2 y 2 − i6x 2 y 2 , the coordinate axes

(c) f (z) = x 3 + 3x y 2 − x + i(y 3 + 3x 2 − y), the coordinate axes

(d) f (z) = x 2 y 2 , the coordinate axes

( 4
e−1/z , z , 0
EXERCISE 1.2.12 (HOT). If f (z) = show that the Cauchy-Riemann equations are satis-
0, z = 0,
fied at every point of the complex plane but f is analytic only at all z , 0.

EXERCISE 1.2.13. Prove or disprove the following statements:

(a) Everywhere differentiable functions are analytic everywhere

(b) If f = u + iv satisfies the Cauchy-Riemann equations at every point of the complex plane, then it will
be differentiable at at least one point of the complex plane

(c) Nowhere analytic function is nowhere differentiable

(d) If f is differentiable at a point of the complex plane, then it will be analytic at that point

(e) If f = u + iv is analytic, then f = v + iu is analytic

1.3 Harmonic Conjugates

Theorem 1.3.1. If f (z) is an analytic with constant modulus, that is | f (z)| = constant on a domain S,
then f (z) will also be constant on S.
Theorem 1.3.2. If f (z) is an analytic with f 0(z) = 0 on a domain S, then f (z) will be constant on S.
Definition 1.3.1 (Harmonic function). A function ψ(x, y), having continuous partial derivatives of second
order on a domain S is said to be harmonic, if it satisfies the Laplace’s equation ∇2 ψ ≡ ∂∂xψ2 + ∂∂yψ2 = 0 on S.
2 2

Department of Mathematics, Vellore Institute of Technology

 Chapter 1. Analytic functions 5

Theorem 1.3.3. Let f (z) = u(x, y) +iv(x, y) be analytic on a domain S. Then the real and imaginary parts
∂2 u ∂2 u ∂2 v ∂2 v
u ≡ u(x, y) and v ≡ v(x, y) will be harmonic on S, that is ∂x 2
+ ∂y 2
= 0 and ∂x 2
+ ∂y 2
= 0 on S.
Remark 1.3.1. If f = u + iv is analytic, we say that the imaginary part v is a harmonic conjugate of the
real part u of f . Note that v is a harmonic conjugate of u if and only if u is a harmonic conjugate of −v. This
follows from the observation that i f = i(u + iv) = −v + iu is analytic whenever f = u + iv is analytic. This
property is referred to as the antisymmetry of harmonic conjugates.

Finding a Harmonic Conjugate

Let f be an analytic function with the real part u. Then u is harmonic. The imaginary part v of f is the
conjugate harmonic of u, and is determined by integrating the total derivative
       
∂v ∂v
dv = ∂x dx + ∂y dy = − ∂u∂y dx + ∂u
∂x dy.

EXERCISE 1.3.1. Examine whether each of the following functions can be the rela part of an analytic
function f (z). If so, find its harmonic conjugate v ≡ v(x, y). Then write the analytic function f = u + iv as
a function of the complex variable z:

(a) cos x cosh y

(b) x 2 − y 2 + y + e x cos y
(c) 4x y 3 − 4x 3 y + x
x 2 −y 2
(d) (x 2 +y 2 )2

(e) e2xy sin x 2 − y 2

2 −y 2
(f) e x cos(2xy)
(g) e x (x cos y − y sin y)
sin 2x
(h) cos 2x+cosh 2y

Remark 1.3.2. Let f be an analytic function with the imaginary part v. Then v is harmonic. The real
part u of f is determined by the short-cut formula
∫   ∫  
∂v ∂v
u ≡ u(x, y) = ∂y dx − ∂x dy + constant (1.3.1)
y−constant x−free

∂u
Milne-Thomson method: If u(x, y) is the real part of an analytic function f (z), then f 0(z) == ∂x − i ∂u
∂y .
Replacing x with z and y with 0 in this and integrating with respect to z, we get f (z).

EXERCISE 1.3.2. Examine whether ψ = e x (x sin y + y cos y) can be the imaginary part of an analytic
function f . If so, find f .

EXERCISE 1.3.3. Find the analytic function f (z) = u + iv where u and v satisfy the following relations:
cos x+sin x−e−y
(a) u − v = 2(cos x−cosh y)
x−y
(b) u + v = x 2 +4xy+y 2

Theorem 1.3.4 (Analytic Function and Orthogonal Curves). Suppose that f (z) = u(x, y) + iv(x, y) is
analytic at a point P. Then the curves u(x, y) = a and v(x, y) = b are orthogonal at P.

Department of Mathematics, Vellore Institute of Technology

6 1.4. Analytic Functions in Flow Problems

1.4 Analytic Functions in Flow Problems

Consider the irrotational motion of a frictionless, incompressible fluid moving in planes parallel to the
x y-plane (that is, planar flow). Suppose that

F(x, y) = f1 i + f2 j (1.4.1)

is the two dimensional velocity field of a fluid particle. Then

∇ × F = 0 so that F = grad φ

for some scalar potential φ(x, y). Since the flow is incompressible,

∇.F = 0 or ∇.∇φ = ∇2 φ = 0.

That is φ is harmonic. The scalar potential φ is called a velocity potential.

It follows that there exists a harmonic conjugate ψ(x, y) of φ such that

f (z) = φ(x, y) + iψ(x, y) (1.4.2)

is analytic, and the harmonic conjugate ψ is called the stream function. Since f is analytic, by Theorem
1.3.4, the level curves φ(x, v) = c1 and ψ(x, y) = c2 are orthogonal. The level curves corresponding to the
velocity potential φ are known as equipotential curves, while the level curves corresponding to the stream
function are called the stream lines, along which the fluid particles will move.

In electrostatics and gravitational fields, the curves φ = c1 and ψ = c2 are equipotential lines and lines of
force. In steady state heat flow problems, the curves φ = c1 and ψ = c2 are isotherms and lines of flow or
flux lines.

The flow pattern is fully represented by the analytic function (1.4.2), called the complex potential.

EXERCISE 1.4.1. Be little clever in examining the truth of following statements:

(a) If f (z) is an entire function, then f (z) is analytic at at least one point of the complex plane.

(b) If u and v are harmonic functions, then f = u + iv is analytic.

(c) If f = u + iv is analytic where u is a harmonic conjugate of v, then f is a constant.

(d) If f = u + iv is analytic,then U = eu cos v and V = e x sin v are harmonic conjugates of each other.

(e) If v(x, y) is its harmonic conjugate of a harmonic function u(x, y), then φ(x, y) = u(x, y)v(x, y) and
ψ(x, y) = u2 − v 2 are harmonic.

Department of Mathematics, Vellore Institute of Technology

Chapter 2

Conformal Mapping

2.1 Geometrical Representation of Complex Functions

Let w = f (z) = u(x, y) + iv(x, y) be a complex function of a complex variable z. Since f involves four
variables: two independent variables x, y and two dependent variables u, v, it is convenient to illustrate f
geometrically by two different planes with rectangular coordinates, namely the z-plane and w-plane. The
points z = (x, y) are plotted in the z-plane, the points w = (u, v) are represented in the w-plane and f is
regarded as a transformation of a set of points of the z-plane to the set of points of w-plane. We cannot
draw the graph a complex function.

2.2 Conformal Mapping

Definition 2.2.1 (Path). A path is a continuous function γ : [a, b] → C. A path γ : [a, b] → C is said to be
smooth, if γ is continuously differentiable on (a, b).

Consider a function f : D → C, where D is a domain in C.

Definition 2.2.2 (Angle Preserving Mapping). The mapping f preserves angles at z0 ∈ D, if the angle of
intersection of the curves C1 and C2 at z0 in the domain equals the angle of intersection of the image curves
Γ1 and Γ2 at f (z0 ) in both magnitude and direction (or sense), then f preserves the angles at z0 .

Definition 2.2.3 (Conformal Mapping). A complex mapping or transformation with angle preserving
property is known as a conformal mapping.

Conformal map preserves both angles and shape of infinitesimal small figures but not necessarily their
size. Thus a mapping
w = f (z) (2.2.1)
is conformal at z0 ∈ D, if it preserves the angles between curves through z0 as well as their orientation.
The mapping w = f (z) is conformal on a domain D if and only if it is conformal at every point of D.
Theorem 2.2.1. A mapping (2.2.1) is conformal at a point z0 ∈ D if and only if it is analytic at z0 and

f 0(z0 ) , 0.

The angle of rotation is α = Arg f 0(z0 ) and the scale factor is

 
2
x,y = | f (z0 )| , 0
J = u,v 0

under f at z0 .

7
8 2.3. The Squared Mapping w = z 2

Example 2.2.1. The exponential function ez and the translation w = az + b are conformal at all points of
the complex plane.

EXERCISE 2.2.1. Examine the conformality of the following mappings:

(a) w = 1/z and w = z 2 − 1

(b) w = z2 + 2z

(c) w = sin z

(d) w = f (z) = z + 1/z

(e) w = z̄ and w = x + i(2y)

Ans.

(a) conformal on C − {0}

(b) conformal at all z except z , −1

(c) conformal at z , ±(2n − 1)π/2, where n = 1, 2, 3, ...

(d) conformal at all points of the C-plane except z = 0, ±1

(e) nowhere conformal

EXERCISE 2.2.2. Show that each of the following mappings are conformal at the given points. Also
determine the angle of rotation α = Arg( f 0(z)) and the scale factor | f 0(z)| 2 at those points:

(a) w = cos z; i, 1, π + i

(b) w = 1/z 1; 1 + i,i

EXERCISE 2.2.3. Examine the conformality of the following mappings:

(a) w = z 3 − 3z + 1

(b) z 2 + 2iz − 3

(c) z − e−z + 1 − i
2 −2
(d) zez

2.3 The Squared Mapping w = z2

The square mapping w = f (z) = z 2 is expressed in rectangular coordinates as

w = u + iv = z 2 = (x + iy)2 = (x 2 − y 2 ) + i(2xy). (2.3.1)

The transformation equations give the real and imaginary parts of w

u = u(x, y) = x 2 − y 2 and v = v(x, y) = 2x y. (2.3.2)

Note that
∂u ∂u ∂v ∂v
∂x = 2x, ∂y = −2y, ∂x = 2y, ∂y = 2x (2.3.3)

Department of Mathematics, Vellore Institute of Technology

 Chapter 2. Conformal Mapping 9

That is, the Cauchy-Riemann equations hold good at every point of the complex plane. Since u, v and the
first order partial derivatives are continuous at all z ∈ C, f is differentiable and hence analytic at all points
z of the complex plane (that is, f is an entire function). Also
∂u ∂v
f 0(z) = ∂x + i ∂x = 2x + i2y = 2z. (2.3.4)

Since f 0(z) = 2z , 0 for z , 0, (2.3.1) is conformal at all z , 0, while f 0(0) = 0. The analyticity of f implies
that the level curves

u = x 2 − y 2 = constant (rectangular hyperbolas)

v = 2xy = constant (rectangular hyperbolas)

corresponding to the real and imaginary parts are orthogonal.

If w = z 2 is the complex potential of a fluid-flow, then its real part u = x 2 − y 2 represents the velocity
potential, and the level curves u ≡ x 2 − y 2 = a are equipotential curves. While, the imaginary part v = 2xy
denotes the stream function, and the level curves v ≡ 2xyb are lines of force or stream lines of the fluid-
flow.
Polar Form: Let z = reiθ , and w = Reiϕ . Then, w = z 2 implies that Reiϕ = (reiθ )2 = r 2 ei2θ , which produces
the transformation equations: R = r 2 and ϕ = 2θ.

EXERCISE 2.3.1. Find the images of the following curves or regions under the transformation (2.3.1):

(a) The straight lines x = c and y = d, where c and d are nonzero real numbers

(b) The real and imaginary axes in the xy-plane

(c) The straight line y = x

(d) The square {(x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1}

(e) The rectangular hyperbolas x 2 − y 2 = a2 and xy = b2

(f) The region enclosed by the triangle with vertices i, 1 and 1 + i

(g) The upper half of the xy-plane

(h) The first quadrant {(r, θ) : r ≥ 0, 0 ≤ θ ≤ π/2}

(i) The circle with centre at the origin and radius a

(j) The annulus, enclosed between the concentric circles z = aeiθ and z = beiθ , where a > 0, b > 0 and
0 ≤ θ ≤ 2π

(k) The sector z = aeiθ , a > 0 and − 4π ≤ θ ≤ π

4

(l) The annular sector z = reiθ , a ≤ r ≤ b and 0 ≤ θ ≤ π/3

Ans.

(a) The image of x = c is a negatively oriented parabola v 2 = −4c2 (u − c2 ) with vertex at the point
(c2, 0), and focus at the origin (0, 0); the image of the line y = d is a positively oriented parabola
v 2 = −4c2 (u − c2 ) with vertex at the point (−d 2, 0), and focus at the origin (0, 0) in the uv-plane

(b) The image of the real axis is the nonnegative real axis u ≥ 0, v = 0; the image of the imaginary axis
is the non-positive portion u ≤ 0, v = 0 of the real axis in the in the uv-plane

Department of Mathematics, Vellore Institute of Technology

10 2.4. The Exponential Function ez

(c) The nonnegative imaginary axis u = 0, v ≥ 0 in the uv-plane

(e) The image of x 2 − y 2 = a2 is a straight line u = a2 parallel to the imaginary axis in the right half
uv-plane; the image of x y = b2 is a straight line y = b2 parallel to the real axis in the upper half
uv-plane

(g) The entire uv-plane

(h) The upper half uv-plane {(ρ, ϕ) : ρ ≥ 0, 0 ≤ ϕ ≤ π}

(i) The enlarged circle ρ = a2, 0 ≤ ϕ ≤ 4π with centre at the origin and radius a2

2.4 The Exponential Function e z

The exponential mapping of a complex variable z is defined by the power series:
∞
Õ
zn
w = f (z) = e = z
n! for all z ∈ C. (2.4.1)
n=0

Theorem 2.4.1. The exponential function (2.4.1) is analytic at all points of the complex plane, and
satisfies the following conditions:

(a) ez , 0 for all z ∈ C

(b) ez1 +z2 = ez1 · ez2

(c) For real θ, eiθ = cos θ + i sin θ.

If z = x + iy, from (b) and (c) of Theorem 2.4.1, we see that

w = ez = e x+iy = e x eiy = e x (cos y + i sin y) = u + iv. (2.4.2)

The real and imaginary parts of w = ez are respectively

u = u(x, y) = e x cos y and v = v(x, y) = e x sin y. (2.4.3)

Note that
∂u
∂x = e x cos y, ∂u
∂y = −e sin y, and
x ∂v
∂x
∂v
= e x sin y, ∂y = e x cos y.

That is, the Cauchy-Riemann equations hold good at every point of the complex plane. Since u, v and
their first order partial derivatives are continuous at all points of the complex plane, f is differentiable
and hence analytic at all points z of the complex plane. That is f is an entire function. The derivative of
f (z) = ez is given by
∂u ∂v
f 0(z) = ∂x + i ∂x = e x cos y + ie x sin y = ez for all z ∈ C. (2.4.4)

Since f 0(z) = ez , 0 for all z ∈ C, (2.4.2) is everywhere conformal in C. The analyticity of f implies that
the level curves u = u(x, y) = e x cos y = a and v = v(x, y) = e x sin y = b corresponding to the real and
imaginary parts are orthogonal.

Polar Form: If w = ez = ρeiϕ , then e x (cos y + i sin y) = ρeiϕ . Therefore,

q
ρ = |w| = |e | = e cos2 y + sin2 y = e x > 0 for − ∞ < x < ∞
z x
(2.4.5)

Department of Mathematics, Vellore Institute of Technology

 Chapter 2. Conformal Mapping 11

and ϕ = tan−1 (e x cos y/e x sin y) = tan−1 (tan y) = y for − ∞ < y < ∞. Note that the principal value of
ϕ = y lies in (0, 2π]. Since ρ = e x > 0 and ϕ = y ∈ (0, 2π], the complex plane is mapped into C − {0}.

Periodicity: Note that ei(2nπ) = cos(2nπ) + i sin(2nπ) = 1 + i(0) = 1 for all n = 1, 2, 3, .... Therefore,
ez+i(2nπ) = ez .ei(2nπ) = ez for all z ∈ C. Thus f (z) = ez is periodic with complex period i(2nπ).

EXERCISE 2.4.1. Find the images of the following curves or regions under the under the exponential
mapping (2.4.2):

(a) The straight lines x = c and y = d, where c and d are nonzero real numbers

(b) The real and imaginary axes in the xy-plane

(c) The infinite strip: −∞ < x < ∞, 0 < y < π

(d) The straight line y = x

(e) The curves e x cos y = a and e x sin y = b

(f) The region S = (x, y) : −1 ≤ x ≤ 1, − 4π ≤ y ≤ − 4π



Ans.

(a) Substituting x = c in the polar form, we get ρ = ec > 0 and ϕ = y ∈ R. Taking the principal value
of ϕ ∈ (0, 2π], note that w = ez = ρeiϕ represents a circle with radius ρ = ec > 0 and centre (0, 0).
Now, substituting y = d in the polar form, ρ = e x > 0 for all x and ϕ = y = d, which represent an
open ray from the origin (0, 0) making a constant angle d, with the real axis in the uv-plane.

(b) Substituting y = 0 in the polar transformation, ρ = e x > 0 for all real x and ϕ = 0. This is the
positive real axis in the uv-plane. Substituting x = 0 in the polar form, we get ρ = e0 = 1 and ϕ = y
for all real y. Thus w = eiϕ represents the unit circle in the uv-plane.

(c) Let z = x + iy be any point of the strip. Then −∞ < x < ∞ and 0 < y < π. Now, ez = e x (cos y +
i sin y) = u + iv so that u = e x cos y, v = e x sin y. Now, e x > 0 for all x and e x → 0 as x → −∞. Also,
0 < y < π → 0 < sin y < 1, while −1 < cos y < 1 and hence −∞ < e x cos y < ∞, e x sin y > 0
or −∞ < u < ∞, v > 0, which represents the upper half uv-plane.

(d) The straight line y = x passes through the origin and makes an angle of π/4 with the positive real
axis in the xy-plane. Substituting y = x in the polar transformation, we get ρ = e x and ϕ = x,
where −∞ < x < ∞. Eliminating x between these two relations, we get ρ = eϕ , which represents
the logarithmic spiral in the uv-plane.

(e) The straight lines u = a and v = b in the uv-plane.

Department of Mathematics, Vellore Institute of Technology

12 2.4. The Exponential Function ez

Department of Mathematics, Vellore Institute of Technology

Chapter 3

Bilinear Transformations

3.1 Linear Fractional Transformation

Let a, b, c, d ∈ C with ad − bc , 0. The mapping

w = f (z) = az+b
cz+d (3.1.1)

is called linear fractional, bilinear or Mobius transformation, with determinant ad − bc , 0. Note that

f 0(z) = ad−bc
(cz+d)2
, 0 for all z ∈ C − {−d/c}. (3.1.2)

Thus f is conformal on C − {−d/c}. The Invariant points or fixed points of a bilinear transformation (3.1.1)
are determined by solving the equation f (z) = z.

EXERCISE 3.1.1. Find the fixed points of the following bilinear transformations:

(a) f (z) = z−1

z+1 (b) g(z) = 4z+3
2z−1 (c) h(z) = 6z−9
z

3.2 Finding a Blinear Transformation

Theorem 3.2.1 (Implicit Formula). The unique bilinear transformation that maps three distinct points
z1 , z2 and z3 onto three points w1 , w2 and w3 respectively is given by
z−z1 z2 −z3 w2 −w3
z−z3 · z2 −z1 = w−w1
w−w3 · w2 −w1 · (3.2.1)

Example 3.2.1. Find the bi-linear transformation f , and its invariant points, which maps the points −1, i
and 1 onto the points 1, i and −1 respectively. Discuss the conformality also.
Solution. Substituting z1 = −1, z2 = i, z3 = 1 and w1 = 1, w2 = i, w3 = −1 in (3.2.1), we get
z+1
z−1 · i−1
i+1 = w−1
w+1 · i+1
i−1 or z+1
z−1 · (i − 1)2 = w−1
w+1 · (i + 1)2 or w−1
w+1 = 1+z
1−z

(w−1)+(w+1) (1+z)+(1−z)
Using componendo and dividendo, we get (w−1)−(w+1) = (1+z)−(1−z) . That is

w = f (z) = −1/z. (3.2.2)

Alternate method: We find the constants a, b, c and d in the bi-linear transformation (3.1.1) as follows:

Since f (−1) = 1, from (3.1.1), we get

−a+b
−c+d = 1 or − a + b = −c + d. (3.2.3)

Since f (i) = i, from (3.1.1), we get

i.a+b
i.c+d = i or ai + b = −c + di. (3.2.4)

Finally, f (1) = −1, from (3.1.1), we get

1.a+b
1.c+d = −1 or a + b = −c − d. (3.2.5)

13
14 3.2. Finding a Blinear Transformation

Adding (3.2.5) to (3.2.3), we see that

2b = −2c or b = −c. (3.2.6)
While, subtracting (3.2.3) from (3.2.3), we see that

2a = −2d or a = −d. (3.2.7)

Substituting (3.2.6) to (3.2.7) in (3.2.4), we obtain

−di − c = −c + di or d = −a = 0.

Thus, the choice of the constants is a = 0 = d and b = −c. Inserting these in (3.1.1), we get (3.2.2).

The invariant points of f are given by f (z) = z, that is z 2 = −1 or z = ±i. Finally, w is analytic on
D ≡ C − {0} with f 0(z) = 1/z 2 , 0.
Example 3.2.2. Construct a linear fractional transformation that maps the points 1, i, and −1 on the unit
circle |z| = 1 onto the points −1, 0, 1 on the real axis respectively. What are its invariant points?
Solution. Substituting z1 = 1, z2 = i, z3 = −1 and w1 = −1, w2 = 0, w3 = 1 in (3.2.1), we get
(i+1)2
z−1
z+1 · i+1
i−1 = w+1
w−1 · 0−1
0+1 or z−1
z+1 · −1−1 = − w+1
w−1 or
w+1
w−1 = iz−i
1+z

Using componendo and dividendo, we get

(w+1)+(w−1) (iz−i)+(1+z) (i+1)z+(1−i)
(w+1)−(w−1) = (iz−i)−(1+z) or w = (i−1)z−(1+i) .

Multiplying the numerator and denominator on the right hand side with (1 − i) ,
(1−i 2 )z+(1−i)2
w= −(i−1)2 z−(1−i 2 )
or w = f (z) = iz−1 .
z−i

Alternate method: We find the constants a, b, c and d in the bi-linear transformation (3.1.1) as follows:
Since f (1) = −1, from (3.1.1), we get
a+b
c+d = −1 or a + b = −c − d. (3.2.8)

Since f (i) = 0, from (3.1.1), we get

i.a+b
i.c+d = 0 or ai + b = 0 so that b = −ai. (3.2.9)

Finally, f (−1) = 1, from (3.1.1), we get

−a+b
−c+d = 1 or − a + b = −c + d. (3.2.10)

Adding (3.2.10) to (3.2.8), we see that

2b = −2c or b = −c. (3.2.11)
While, subtracting (3.2.10) from (3.2.8), we see that

2a = −2d or a = −d. (3.2.12)

Thus from (3.2.9), (3.2.11) and (3.2.12), we get

b = −ai, c = −b = ai, d = −a.

Department of Mathematics, Vellore Institute of Technology

 Chapter 3. Bilinear Transformations 15

Substituting these in (3.1.1), we get

w= az−ai
aiz−a or w = iz−1 ,
z−i
(3.2.13)
which is the required bilinear transformation.

The invariant points of f are√given by f (z) = z, that is z − i = iz2 − z, which are the roots of z 2 + 2zi + 1 = 0.
2 √
In fact, the roots are z = −2± 24i −4 = i(−1 ± 2). Since w is a rational function and f 0(z) = −2/(iz − 1)2 , 0
for all z , −i, we conclude that f is analytic on D ≡ C − {−i}.
Example 3.2.3. Construct a linear fractional transformation, that sends the points 0, −i, and −1 onto the
points i, 1, 0 on the real axis respectively. Also find its and its invariant points.
Solution. Substituting z1 = 0, z2 = −i, z3 = −1 and w1 = i, w2 = 1, w3 = 0 in (3.2.1), we get
(1−i)2 z
z−0
z+1 · −i+1
−i−0 = w−i
w−0 · 1−0
1−i w =
or w−i −i z+1

i(z+1)
That is, 1 − i
w = 2z
z+1 or w = The invariant points of f are given by f (z) = z, that is iz + i = z − z 2 ,
1−z . √ √
−(i−1)± (i−1)2 −4i
which are the roots of z 2 + (i − 1)z + i = 0, given by z = 2 = 1−i± −6i
2 . But
√
−i = [cos(3π/2) + i sin(3π/2)]1/2 = cos(3π/4) + i sin(3π/4) = √ .
i−1
2
√ √ √
1−i± 6(i−1)/ 2
Therefore, 2 = 1± 3
2 (1 − i).

EXERCISE 3.2.1. Find the bilinear transformation w = f (z) that maps one triad (z1, z2, z3 ) onto another
triad (w1, w2, w3 ) either by (3.2.1) or by finding an appropriate choice of the constants a, b, c and d in (3.1.1).
Also, determine the derivative, conformality, and the invariant points in each case:

(a) 1, 2 and 3 onto 0, 1 and ∞

(b) 0, 1 and ∞ onto i, 1 and −i
(c) 1, i and −1 onto i, 0 and i
(d) −i, 1 and i onto −1, 0 and 1
(e) −2, −1 − i and 0 −1, 0 and 1
(f) 0, i and −i onto −1, 1 and 0
(g) −i, 0 and i onto the points −1, i and 1
(h) 0, 1 and 2 onto 0, 1 and ∞
(i) 1, i and −1 onto 0, 1 and ∞

3.3 Cross Ratio

Tthe unique bilinear transformation w = f (z), which maps the distinct points z1 , z2 and z3 taken in order,
onto w1 = 0, w2 = 1 and w3 = ∞ respectively is
1−w3 (z−z2 )(z1 −z3 )
w−0
w−w3 · 1−0 = (z−z3 )(z1 −z2 ) .

Dividing the numerator and denominator on the left hand side of this, and using the 1/w3 = 0, we find
(z−z2 )(z1 −z3 )
0−1 · 1−0 = (z−z3 )(z1 −z2 ) or
0−1
that w−0
(z−z1 )(z2 −z3 )
w = f (z) = (z−z3 )(z2 −z1 ) · (3.3.1)

Department of Mathematics, Vellore Institute of Technology

16 3.4. Images under a Bilinear Transformation

The image of z under this transformation (3.3.1) is called the cross ratio of the points z, z1 , z2 and z3 taken
in that order.

EXERCISE 3.3.1. Evaluate the cross ratio [0, 1, i, ∞].

Solution. Here z = 0, z1 = 1, z2 = i and z3 = ∞. From (3.3.1), we have
 z2 
z2 −z3 −1
[z, z1, z2, z3 ] = z2 −z1 · z−z3 = z2 −z1 · zz3
z−z1 z−z1
= zz−z 1
2 −z1
·1= 0−1
i−1 = (i + 1)/2
z3 −1

EXERCISE 3.3.2. Evaluate the following cross ratios:

(a) [i − 1, ∞, 1 + i, 0]

(b) [7 + i, 1, 0, ∞]

(c) [2, 1 − i, 1, 1 + i]

(d) [0, 1, i, −1]

3.4 Images under a Bilinear Transformation

Since ad , bc, the transformation (3.1.1) is invertible, and its inverse is given by

z = f −1 (w) = cw−a ,
b−dw
(3.4.1)

which is also a bilinear transformation with determinant ad − bc , 0. If z = −d/c, then w = f (z) = ∞.

Similarly, If z = a/c, then z = f −1 (w) = ∞.

EXERCISE 3.4.1. Find the images of the points 0, 1+i, i, and ∞ under the linear fractional transformation
f (z) = (2z + 1)/(z − i).

Ans. f (0) = i, f (1 + i) = 3 + 2i, f (i) = ∞ and f (∞) = 2

EXERCISE 3.4.2. Find the images of the points 0, 1, i and ∞ under the following linear fractional trans-
formations:

(a) f (z) = i/z

(b) f (z) = 2/(i − z)

EXERCISE 3.4.3. Find the image of the unit circle |z| = 1 under the linear fractional transformation
f (z) = (z + 2)/(z − 1).

EXERCISE 3.4.4. Find the images of the the disk |z| ≤ 1 under the linear fractional transformations
given below:

(a) f (z) = (z + i/(z − i)

(b) f (z) = z − 1/z

Department of Mathematics, Vellore Institute of Technology

Chapter 4

Complex Integration

4.1 Introductory Concepts

Let γ : [a, b] → C be a continuous function. Then its real and imaginary parts, say x = x(t) and y = y(t)
will be continuous functions on [a, b]. We say that γ traces out a continuous curve in the complex plane
with parametric form
z = x + iy = x(t) + iy(t) for all a ≤ t ≤ b. (4.1.1)
For example, the parametric form of a unit circle |z| = 1 with centre at the origin is

z = cos t + i sin t = eit for 0 ≤ t ≤ 2π. (4.1.2)

The point z(a) = x(a) + i y(a) or A = (x(a), y(a)) is the initial point of the curve γ and z(b) = x(b) + i y(b) or
B = (x(b), y(b)) is its terminal point. The point z = x(t) + i y(t) is also regarded as a two dimensional vector
function.

The derivative of (4.1.1) is given by

z 0(t) = x 0(t) + i y 0(t) for all a < t < b. (4.1.3)

Definition 4.1.1. A curve γ defined by (4.1.1) is smooth, if the tangent z 0(t) is continuous and nonzero in
the interval a ≤ t ≤ b.

A smooth curve will have no sharp corners or cusps.

Definition 4.1.2. A continuous curve γ is said to be piecewise smooth, if it is a finite union of smooth arcs.
A piecewise smooth curve is usually referred to as contour or path.
Definition 4.1.3. A continuous curve γ : [a, b] → C is simple, if z(t1 ) , z(t2 ) for t1 , t2 except at the end
points z = a and z = b. A simple curve does not cross itself in between its initial and terminal points.
Definition 4.1.4. A curve γ : [a, b] → C is closed, if z(a) = z(b).
Definition 4.1.5. A contour is said to be positively oriented, if it is traced in the direction of increasing
values of the parameter t.
Remark 4.1.1. In case of a simple closed curve C, the anticlockwise direction is regarded as its positive
orientation.

4.2 Contour integration

Let γ be a smooth curve in a region S ⊂ C defined by (4.1.1) and f : S → C, continuous. We define the line
integral of f (z) = u(x, y) + iv(x, y) along γ from z = a to z = b by
∫ ∫ ∫ b
f (z) dz = (u + iv)(dx + i dy) = [u(t) + iv(t)] · [x 0(t) + iy 0(t)] dt. (4.2.1)
γ γ t=a

If −γ denotes the curve having the opposite orientation of a curve γ, then

∫ ∫
f (z) dz = − f (z) dz.
−γ γ

17
18 4.2. Contour integration

Linearity: If γ is consists of finite number of smooth arcs γ1 , γ2 , . . . , γn joined end to end, then
∫ ∫ ∫ ∫
f (z) dz = f (z) dz + f (z) dz + · · · + f (z) dz. (4.2.2)
γ γ1 γ2 γn

Line Integral around Unit Circle: Consider the unit circle γ : |z| = 1. We see that z = eiθ on γ so that
dz = eiθ dθ and θ ranges from 0 to 2π. Therefore,
∳ ∫ θ=2π
f (z) dz = f (eiθ ) · ieiθ dθ. (4.2.3)
|z |=1 θ=0

EXERCISE 4.2.1. Find the integrals of f (z) along/around the contours C in Exercises (a) through (e) by
using parametric representations for C:

(a) f (z) = z̄; C is the arc of the paprabola y 2 = x from the point (0, 0) to the point (1, 1)

(b) f (z) = z̄)2 ; C is the real axis from z = 0 to z = 2 and then along parallel to the y-axis from z = 2 to
z = 2+i

(c) f (z) = x 2 + iy 2 ; C is the straight line segment from the origin (0, 0) to the point (1, 1), then to the
point (1, 2)

(d) f (z) = 1/(z − a); C is the circle |z − a| = r

(e) f (z) = 3y 2 + 2iy; C is the unit circle from (1, 0) to (0, 1) in anticlockwise direction
Ans.

(a) 1 − i/3

(b) (14 + 11i)/3

(c) (5i − 7)/3

(0) 2π
∫ 1+i
EXERCISE 4.2.2. Evaluate (x − y + ix 2 ) dz along
0

(a) the straight line y = x;

(b) the real axis from z = 0 to z = 1 and then along parallel to the y-axis from z = 1 to z = 1i.
∫ (2,8)
EXERCISE 4.2.3. Evaluate (x 2 + ix y) dz along
(1,1)

(a) the straight line segment y = x;

(b) the curve x = t, y = t 3 .

∫
EXERCISE 4.2.4. Evaluate |z| dz along

(a) the straight line segment from −i to i

(b) the left-half of the unit circle x 2 + y 2 = 1 from −i to i

Ans. (a) 0 (b) i

Department of Mathematics, Vellore Institute of Technology

Chapter 5

Cauchy’s Integral Formula

5.1 Simply and Multiply Connected Domains

Definition 5.1.1. A domain D is said to be simply connected if every simple closed contour C lying entirely
can be shrunk to a point without leaving D. A domain which is not simply connected is a multiply connected
domain.

A simply (doubly, triply, . . .) connected region will have no (one, two, . . .) holes in it.

5.2 Cauchy’s Theorem

Theorem 5.2.1 (Cauchy’s Theorem). Let D ⊂ C be a simply connected domain. If f : D → C is
analytic on D, then for any simple closed contour γ in D, we have
∳
f (z) dz = 0. (5.2.1)
γ

Example 5.2.1.
∳
3z 2 +7z+1
(a) z+1 dz = 0, where γ is the circle |z| = 1/2
γ
∳
(b) ez dz = 0, where γ is the circle |z| = 1
γ
∳
(c) z−3
z 2 +2z+5
dz = 0
|z |=1

5.3 Cauchy’s Integral Formula

Theorem 5.3.1 (Cauchy’s Integral Formula). Suppose that f is analytic in a simply connected domain
D and γ is any simple closed contour lying entirely within D. Then for any point a within γ,
∳
f (z)
z−a dz = 2πi · f (a) (5.3.1)
γ

Theorem 5.3.2 (Cauchy’s Integral Formula for Derivatives). Suppose that f is analytic in a simply
connected domain D and C is any simple closed contour lying entirely within D. Then for any point a with
in C, we have ∳
f (z)
(z−a) n+1
dz = 2πi
n! · f (n) (a) for n = 1, 2, 3, . . . . (5.3.2)
γ

Example 5.3.1. Evaluate the each of the following contour integrals using (5.3.1) or (5.3.2), where γ is the
circle mentioned against each integral:

∳
(a) cos z
z dz; γ : |z| = 1
γ
∳ ∳
Solution. cos z
z dz = cos z
z−0 dz = 2πi · cos 0 = 2πi
γ γ

19
20 5.3. Cauchy’s Integral Formula

∳
z 2 −z+1
(b) z−1 dz; γ : |z| = 3/2
γ
∳
z 2 −z+1
dz = 2πi ·  z 2 − z + 1z=1 = 2πi · 1 = 2πi
 
Solution. z−1
γ
∳
z 2 −4z+4
(c) z+i dz; γ : |z| = 2
γ
∳
z 2 −4z+4
dz = 2πi ·  z2 − 4z + 4z=−i = 2πi(3 + 4i) = 2(−4 + 3πi)
 
Solution. z+i
γ
∳
e2 z
(d) (z+1)4
dz; γ : |z| = 2
γ
∳  3 
e2 z
dz = 2z 
= =
2πi d  2z  8πi
Solution. (z+1)4 3! ·  dz 3 (e ) πi
3
8e 
z=−1 3e2
γ z=−1
∳
(e) sin z
(z−π/6)3
dz; γ : |z| = 1
γ
∳  2 
e2 z
dz = 2πi
= πi |− sin z| z=π/6 = − 2πi
2 = −πi
d
Solution. ·  dz 2 (sin z)

(z−π/6)3 2! z=π/6
γ
∳
e−z
(f) (z+2)5
dz; γ : |z| = 3
C
Solution. Try yourself

Example 5.3.2. Evaluate the each of the following contour integrals using Cauchy’s integral formula,
where γ is the circle mentioned against each integral:
∳
3z 2 +z
(a) z 2 −1
dz; γ : |z − 1| = 1
γ

Solution. The point z = 1 lies inside, z = −1 lies outside γ. Therefore, we define g(z) = (3z 2 +z)/(z+1)
so that ∳ ∳
3z 2 +z g(z)
z 2 −1
dz = z−1 dz = 2πi · g(1) = 2πi · 1+1 = 4πi
3·1+1
γ γ
∳
(b) 1
z 2 +1
dz; γ : |z| = 2
γ

Solution. Both the points z = ±i lie inside the circle γ : |z| = 2, and z 21+1 = 2i1 z−i
 1 1

− z+i . Therefore,
we define g(z) = 1. Then
∳ ∳ ∳ 
g(z) g(z)
1
z 2 +1
dz = 2i
1
z−i dz − z+i dz = 2i · 2πi[g(i) − g(−i)] = 0
1
γ γ γ

∳
(c) z
z 2 +9
dz; γ : |z − 2i| = 4
γ

Solution. Both the points z = ±3i lie inside the circle γ, and z 2z+9 = 6i1 z−3i + z+3i
 z z

. Therefore, we
define g(z) = z. Then
∳ ∳ ∳ 
z
z 2 +9
dz = 6i
1
z−3i dz +
z
z+3i dz = 6i · 2πi · [g(3i) + g(−3i)] = 3 · [3i − 3i] = 0
z 1 π
γ γ γ

Department of Mathematics, Vellore Institute of Technology

 Chapter 5. Cauchy’s Integral Formula 21

∳
e−z
(d) (z−1)(z−2) dz; γ : |z| = 3.
γ
Solution. Both the points z = 1 and z = 2 lie inside γ. Therefore,
∳ ∳ ∳  
e−z e−z e−z
(z−1)(z−2) dz = z−1 dz − z−2 dz = 2πi e −1
− e −2
= 2πi · e−1
e2
γ γ γ

∳
(e) z−1
(z+1)2 (z−2)
dz; γ : |z − i| = 3
γ
Solution. The points z = −1 and z = 2 lie inside the circle γ, and
z−1
(z+1)2 (z−2)
= − 19 · 1
z+1 + 2
3 · 1
(z+1)2
+ 1
9 · 1
z−2

Therefore,
∳ ∳ ∳ ∳
z−1
dz = − 19 1
dz + 2 1
dz + 1 1
dz = 2πi − 19 · 1 + 2
·0+ 1
·1 =0


(z+1)2 (z−2) z+1 3 (z+1)2 9 z−2 3 9
γ γ γ γ

EXERCISE 5.3.1. Evaluate the each of the following contour integrals using using (5.3.1) or (5.3.2), where
γ is a circle mentioned against each integral:
∳
ez
(a) z(z−1)2 (z−2)
dz; γ : |z| = 3/2
C
∳
e2 z
(b) (z−1)(z−2) dz; γ : |z| = 3
C
∳
ez
(c) 2 +π2 2
dz; γ : |z| = 4
γ ( z )
∳
(d) z
z 4 +2iz 3
dz; γ : |z| = 1
C
∳
EXERCISE 5.3.2. Evaluate cos πz
z 2 −1
dz, where γ is the rectangle whose vertices are 2 ± i, −2 ± i.
γ

Department of Mathematics, Vellore Institute of Technology

22 5.3. Cauchy’s Integral Formula

Department of Mathematics, Vellore Institute of Technology

Chapter 6

Taylor’s Series, Laurent’s Series and Residues

6.1 Taylor’s Series

Let f (z) be analytic in a domain D ⊂ C and z0 ∈ D. If C is the largest circle with center at z0 and radius R
that lies entirely within D, then
∞
Õ
f (z) = an (z − z0 )n = a0 + a1 (z − z0 ) + a2 (z − z0 )2 + · · · an−1 (z − z0 )n−1 + Rn, (6.1.1)
n=0

where
∳
f ( j) (z0 ) f (ξ)
a0 = f (z0 ), a j = j! = 1
2πi (ξ−z0 ) n dξ, j = 01, 2, 3, ..., (n − 1) (6.1.2)
C

and
∳
(z−z0 )2 f (ξ)
Rn = 2πi (ξ−z)(ξ−z0 ) n dξ. (6.1.3)
C

Equation (6.1.1) is called the Taylor’s formula with remainder after n terms. It can be shown that Rn → 0 as
n → ∞. Hence Taylor’s formula (6.1.1) gives the Taylor series of f about z0 :
∞
Õ
f (z) = an (z − z0 )n = a0 + a1 (z − z0 ) + a2 (z − z0 )2 + · · · an (z − z0 )n + · · · , (6.1.4)
n=0

with radius R of convergence, where its coefficients are given by

f (n) (z0 )
a0 = f (z0 ), an = n! , n = 1, 2, 3, .... (6.1.5)

If (6.1.4) converges for all z, we write R = ∞.

f (n) (2) e2
Example 6.1.1. Let f (z) = ez for all z ∈ C and z0 = 2. Then an = n! = n! for all n = 1, 2, .... Therefore,
its Taylor series about z0 = 2 is given by
∞ ∞
f (z) = ez = an (z − 2)n = e2 1
· (z − 2)n ·
Í Í
n! (6.1.6)
n=0 n=0

About z0 = 0, we have
∞ ∞
zn
f (z) = ez = an z n =
Í Í
n! · (6.1.7)
n=0 n=0

EXERCISE 6.1.1. Find the Taylor’s series of

(a) f (z) = sin z about the origin;

(b) f (z) = z cosh z 2 about the origin;
(c) f (z) = cos z about z = π/2.
Ans.
∞ ∞
f (n) (0) (−1) n+1 z 2n
(a) f (z) = sin z = ∞ · zn = (b) f (z) = z =
Í∞ 1
· z 2n+1 · z 2n+1
Í Í Í
n=0 n! n=1 (2n+1)! (2n)! (2n)!
n=0 n=0

23
24 6.2. Taylor’s Series for Rational Polynomial Functions

6.2 Taylor’s Series for Rational Polynomial Functions

We employ the binomial series expansions for obtaining Taylor’s Series for rational polynomial functions:
Example 6.2.1. Find a Taylor series of f (z) = 1
(z−1)(z+2) about z = 0.
Solution. We split f into partial fractions and apply the binomial expansion. Now,
   
h i ∞ ∞
z
−1
n
f (z) = (z−1)(z+2) = 3 z−1 − z+2 = 3 −(1 − z) − 2 1 + 2
1 1 1 1 1 −1 1
=3 −
1
zn − 1
(−1)n u3 .
Í Í
2
n=0 n=0

The first series converges on the region S1 : |z| < 1, while the second series converges on the region
|z/2| < 1 or S2 : |z| < 2. Therefore, the region of convergence of f (z) is S1 ∩ S2 : |z| < 1, which contains
z = 0. Thus the Taylor’s series in its simplified form is
∞ n
n+1 o
f (z) = 1
−1 + − 12 zn .
Í
3
n=0

EXERCISE 6.2.1. Find the Taylor’s series of the following functions about the origin:

1
(a) (z+2)2 (z−3)
;

1
(b) (z 2 +1)(z−1)
;

(c) f (z) = 1
z 2 +z+1
.

Example 6.2.2. To find a Taylor series of f (z) = (z−1)(z−3)

1
about z = 4, we write u = z − 4 or z = u + 4 in
this, and split f into partial fractions and then apply the binomial expansion. Now,
 
 1h
−1 i 1 Í ∞ ∞
n
f (z) = (u+3)(u+1)
1
= 21 u+1 = 2 (1 + u)−1 − 31 1 + u3 =2 (−1)n un − 31 (−1)n u3 .
 1 1 Í
− u+3
n=0 n=0

The first series converges on the region S1 : |u| < 1, while the second series converges on the region
|u/3| < 1 or S2 : |u| < 3. Therefore, the region of convergence of f (z) is S1 ∩ S2 : |u| < 1, which contains
u = 0. In other words, the region of validity is |z − 4| < 1, which is the interior of the circle with centre at
4 and radius 1. Thus the Taylor’s series in its simplified form is
∞  ∞ 
f (z) = 1 1
(−1)n un = 1 1
(−1)n (z − 4)n .
Í Í
2 1− 3 n+1 2 1− 3 n+1
n=0 n=0

EXERCISE 6.2.2. Find the Taylor series about the given point for the following functions about a given
point, and mention the region of validity in each case:

(a) f (z) = 1
(z−3)(z+2)2
about z = −1;

(b) f (z) = 1
z2
about z = i;

z 2 +3z
(c) f (z) = 1−z 2
about z = 2;

(d) f (z) = z−1

z2
about z = 1;

(e) f (z) = 1
z 2 −1
about z = i;

(f) f (z) = 1
z 2 −1
about z = −i.

Department of Mathematics, Vellore Institute of Technology

 Chapter 6. Taylor’s Series, Laurent’s Series and Residues 25

6.3 Singularities
Definition 6.3.1 (Singularity). A singularity of a function w = f (z) is a point z0 at which f is not analytic,
but is analytic at some point in every neighborhood of z0 .
Definition 6.3.2 (Isolated singularity). A singularity z0 such that f is analytic at all points of some deleted
neighborhood N(z0 ) − {z0 } of z0 .
Example 6.3.1.

(a) Let f (z) = 1/z for all z ∈ C. Then f is analytic at all points of the complex plane except at z = 0, and
hence z = 0 is an singularity of f .

(b) The origin z = 0 is an isolated singularity of f (z) = sin z/z 4 .

(c) The function f (z) = 1/(z 2 + 1) is analytic at all points of the complex plane except at z = ±i, and
hence z = ± i are isolated singularities.

(d) For f (z) = 1/sin(π/z), the singularities are z = 0, ±1, ±1/2, ±1/3, ... which lie on the real axis from
z = −1 to z = 1. Each z = 1/n is an isolated singularity. But z = 0 is not an isolated singularity, since
every deleted neighborhood of it contains infinitely many singularities 1/n other than itself.

(e) The functions f (z) = |z| 2 and g(z) = |z| are nowhere analytic and hence the criterion of singularity
is not satisfied at any point of the complex plane. Thus they do not have singularities.

6.4 Laurent Series about an Isolated Singularity

Let f (z) be analytic in the annulus A : 0 < r < |z − z0 | < R, where z0 is an isolated singularity of f . Then
the Laurent series of f about z0 is given by
∞
Õ
f (z) = an (z − z0 )n, (6.4.1)
n=−∞

where
∳
f (ξ)
an = 1
2πi (ξ−z0 ) n+1
dξ, n = 0, ±1, ±2, ... (6.4.2)
C

where C is a simple closed curve that lies entirely within A , containing z0 . We rewrite (6.4.1) as
∞ ∞ ∞
f (z) = an (z − z0 )n = a−n (z − z0 )−n + an (z − z0 )n .
Í Í Í
(6.4.3)
n=−∞ n=1 n=0
| {z } | {z }
Principal Part Analytic Part

Definition 6.4.1 (Residue at an isolated Singularity). The coefficient a−1 of the first negative power of
(z − z0 ) in the expansion (6.4.3), valid in the annulus A is known as the residue Res ( f ; z0 ) of f at z = z0 .
Example 6.4.1. Find Laurent’s series of f (z) = 1/z 2 (1 − z) about the isolated singularity z = 0 in the
regions R1 : 0 < |z| < 1 and R2 : |z| > 1.
Solution. On the punctured region R1 : 0 < |z| < 1, we see that
∞ ∞
1
= 1
zn = z n−2 = 1
+ 1
+ 1 + z + z2 + · · ·
Í Í
z 2 (1−z) z2 z2 z
n=0 n=0

Note that Res ( f ; 0) = the coefficient of 1/z in the Laurent series = 1.

Department of Mathematics, Vellore Institute of Technology

26 6.4. Laurent Series about an Isolated Singularity

While, on R2 : |z| > 1, we see that |1/z| < 1. Therefore,

  −1 ∞ ∞
1
= − z13 1 − 1z = − z13 zn = − z n−3 = − z13 − 1 1
− 1 − z − z2 − · · ·
Í Í
z 2 (1−z) z2
− z
n=0 n=0

Example 6.4.2. Find Laurent’s series of f (z) = 1/z 2 (1 − z) about z = 1 in appropriate regions of validity.
Solution. Let z − 1 = u. Then
∞ ∞
1
= − (u+1)
1
2 u = − u (1 + u)
1 −2
= − u1 (−1)n (n + 1)un = (−1)n+1 (n + 1)un−1,
Í Í
z 2 (1−z)
n=0 n=0

which converges on 0 < |u| < 1. Thus the Laurent series of f about z = 1 is
∞
f (z) = (n + 1)(−1)n+1 (z − 1)n−1,
Í
n=0

valid in R1 : 0 < |z − 1| < 1 and Res ( f ; 1) = −1

Now, on the exterior region R2 : |z − 1| > 1, we see that |u| > 1, and hence
∞ ∞
f (z) = − (u+1)
1
2 u = − u 2 (1 + 1/u)
1 −2
= − u12 (−1)n (n + 1)(1/u)n = (−1)n−1 (n + 1)/un+2 .
Í Í
n=0 n=0

∞
Therefore, the Laurent series of f about z = 1 on R2 is f (z) = (−1)n−1 (n + 1)/(z − 1)n+2 .
Í
n=0
Example 6.4.3. Find the Laurent series and the residues of f (z) = 1/(z + 1)(z − 2) about its isolated
singularities z = −1 and z = 2 in appropriate regions of validity.
∞ ∞ n−1
Solution. Let z + 1 = µ. Then, f (z) = µ(µ−3)
1
= − 3µ
1
(1 − u/3)−1 = − 3µ
1 Í
(µ/3)n = −
Í µ
3 n+1
, which
n=0 n=0
converges on 0 < |µ/3| < 1, that is on R1 : 0 < |z + 1| < 3. Thus the Laurent series of f about z = −1 is
∞
(z+1) n−1
f (z) = − , and Res ( f ; −1) = −1/3.
Í
3 n+1
n=0

While, on the exterior domain R2 : |z + 1| > 3, we see that  z+1  < 1 so that |3/µ| < 1. Therefore,
 3 

  −1 ∞ ∞ ∞
3n 3n
f (z) = 1
= 1 3
= 1
(3/µ)n = = .
Í Í Í
µ(µ−3) µ2
1− µ µ2 µ n+2 (z+1) n+2
n=0 n=0 n=0

EXERCISE 6.4.1. Find the Laurent series and the residues of f (z) = 1/z(z + 5) about z = 0 and z = −5 in
appropriate regions of validity.

Ans. The Laurent series of f about z = 0 converge on the domains R1 : 0 < |z| < 5 and R2 : |z| > 5. Then
identify the coefficient of 1/z in the expansion in the punctured region R1 , which will be the residue of f
at z = 0. Similarly, the Laurent series of f about z = −5 converge on the domains R3 : 0 < |z + 5| < 1
and R2 : |z + 5| > 1. Then identify the coefficient of 1/(z + 5) in the expansion in R3 , which will be the
residue of f at z = −5.

EXERCISE 6.4.2. Find a Laurent’s series and residues of the following functions f (z) at their isolated
singularities, in an appropriate domain of validity:

(a) f (z) = 1
z(z+2) ; z = 0, −2

(b) f (z) = 1
z 2 +4
; z = 2i

Department of Mathematics, Vellore Institute of Technology

 Chapter 6. Taylor’s Series, Laurent’s Series and Residues 27

(c) f (z) = z+4

z 2 (z 2 +3z+2)
; z = −1, −2

(d) f (z) = z
(z−1)(z−3) ; z = 1, 3

7z 2 −9z−18
(e) f (z) = z 3 −9z
; z = 0, ±3

(f) f (z) = 1
(z−1)2 (z+3)
; z = 1, −3

(g) f (z) = 8z+1

z(1−z) ; z = 0, 1

6.5 Laurent Series about a Point of Analyticity

Some times Laurent series can be developed of f about a point of analyticity in an annulus, which does not
contain any of its singularities.
Example 6.5.1. Find the Laurent series of f (z) = 1/z(1− z) about z = 2 in the annulus R : 1 < |z − 2| < 2.
Solution. The isolated singularities z = 0 and z = 1 do not belong to the annulus R. But, f is analytic at
z = 2 ∈ R. Now, let w = z − 2. Then f (z) = − (w+2)(w+1)
1
= 2+w
1 1
− w+1 . In the annulus R : 1 < |w| < 2, we
have  w  < 1 and  2  < 1. Therefore, the Laurent series about z = 2 is given by:
1 w

w −1 1 −1
f (z) = 2−w − w+1 = 2 1 − 2
1 1 1
− w1 1+



w
∞ ∞
n
1 Í w n
= 1 Í
(−1)n w1


2 2 − w
n=0 n=0
∞ n ∞
n+1
= w
+ (−1)n+1 w1
Í Í
2 n+1
n=0 n=0
∞ n ∞
(z−2) (−1) n+1
= +
Í Í
2 n+1 (z−2) n+1
n=0 n=0

EXERCISE 6.5.1. Find the Laurent’s series of the following functions about the indicated points of ana-
lyticity in the given annulus A :

(a) f (z) = 1
z(z+2) about z = −1; A : 1 < |z − 1| < 3

z2
(b) f (z) = z 2 −3z+2
about z = −1; A : 1 < |z| < 2, A : 1 < |z − 3| < 2

(c) f (z) = z
(z−1)(z−3) about the origin; A : 1 < |z| < 3

(d) f (z) = z−1

z2
about z = 2; A : 1 < |z − 2| < 2

(e) f (z) = 1
z(z−1) about z = −1; A : 1 < |z − 2| < 2

6.6 Types of Isolated Singularities

Recall that a singularity z0 is an isolated one, if f is analytic at all points of some deleted neighborhood
N(z0 ) − {z0 } of z0 . Let f (z) be analytic in the annulus 0 < r < |z − z0 | < R, where z0 is an isolated
singularity of f . Then the Laurent series of f about z0 is given by
∞ ∞ ∞
f (z) = an (z − z0 )n = a−n (z − z0 )−n + an (z − z0 )n
Í Í Í
(6.6.1)
n=−∞ n=1 n=0
| {z } | {z }
Principal Part Analytic Part

Department of Mathematics, Vellore Institute of Technology

28 6.6. Types of Isolated Singularities

1
where an0 s are called the coefficients of the series. The coefficient a−1 of the first negative power z−z0 in the
expansion (6.6.1) about is known as the residue of f at z = z0 , and is denoted by Res ( f ; z0 ).
Depending on the nature of the principal part of the Laurent’s series (6.6.1) of f (z) about z0 , isolated sin-
gularities are classified into three types:

Removable singularity: The principal part is absent

Essential singularity: Infinitely many terms in the principal part
Poles: Finite number of terms in the principal part

Removable singularity

∞
Suppose that z0 is a removable singularity of f . Then f (z) = an (z − z0 )n for all 0 < |z − z0 | < δ, and f
Í
n=0
can be made analytic at z0 by properly defining as f (z0 ) = limz→z0 f (z).
Example 6.6.1. The function f (z) = sin z/z is not defined at z = 0, and is analytic at all z , 0. Note that
3 5 7
sin z = z − z3! + z5! − z7! + · · · so that for z , 0,
 3 5 7
 2 4 6
f (z) = sinz z = 1z z − z3! + z5! − z7! + · · · = 1 − z3! + z5! − z7! + · · · ·

Thus the Laurent’s expansion about z = 0 does not contain the principal part. Therefore, f has a removable
singularity at z = 0. Since limz→0 sin z/z = 1, f can be made analytic at z = 0 by redefining as
(
z , z ,0
sin z
f (z) =
1 z = 0.

EXERCISE 6.6.1. Origin z = 0 is a removable singularity of the following functions. Define the functions
appropriately to make it analytic at z = 0:

(a) f (z) = (1 − cos z)/z;

(b) f (z) = (z − sin z)/z.
Ans.
(
1−cos z
, z,0
(a) f (z) = z
0 z=0
(
z ,
z−sin z
z,0
(b) f (z) =
0 z=0

Essential Singularities
2 1/z 3
Example 6.6.2. Let f (z) = e1/z = 1 + 1/z 1/z
1! + 2! + 3! + · · · , which has infinitely many negative powers
of z. Thus z = 0 is an essential singularity of f .

EXERCISE 6.6.2. Explain why z = 0 is an essential singularity of the following functions:

(a) f (z) = z cos(1/z);

(b) f (z) = z sin 1/z 2 .

Department of Mathematics, Vellore Institute of Technology

 Chapter 6. Taylor’s Series, Laurent’s Series and Residues 29

Poles

Definition 6.6.1. If the principal part of the Laurent’s Expansion of f about z = z0 is of the form
a1−n a−n
f (z) = a−1
z−z0 + a−2
(z−z0 )2
+···+ (z−z0 ) n−1
+ (z−z0 ) n , (6.6.2)

where a−n , 0, then we say that z = z0 is a pole of order n for f (z).

Thus if z = z0 is a pole of order n for f , the principal part contains atmost n terms, and the negative powers
1/(z − z0 )m are missing for m > n in the Laurent’s Expansion (6.6.1). Poles of order 1, 2 and 3 are called
simple, double and triple poles respectively.
Definition 6.6.2 (Residue at a pole). Let (z − z0 ) be a pole of order m. Then the coefficient a−1 of the
negative power 1/(z − z0 ) in the expansion (6.6.2), valid in an annulus A enclosing the pole z = z0 is the
residue Res ( f ; z0 ) of f at the pole z0 .
Example 6.6.3. Consider
h  i
(z)2 (z)4 (z)6 z3
f (z) = 1−cosh z
z3
= 1
z3
1− 1+ 2! + 4! + 6! +··· = − 2!1 · 1
z − z
4! z − 6! −··· .

Therefore, z = 0 is a simple pole of f , and Res ( f ; 0) = −1/2.

EXERCISE 6.6.3. Find the order of the pole z = 0, and the residue there at for each of the following
functions:

(a) f (z) = sin2 z/z 3 ;

(b) f (z) = 1 − e2z /z 4 ;

(c) f (z) = sinh z/z 4 .

Ans. (a) 1, Res ( f ; 0) = 1 (b) 3, Res ( f ; 0) = −4/3 (c) 3, Res ( f ; 0) = 1/6

6.7 Formula for the Residue at a Pole

Theorem 6.7.1. An isolated singularity z0 of a complex function f (z) is a pole of order m ≥ 1, if and
only if φ(z) = (z − z0 )m f (z) is analytic at z0 .

(a) At a simple pole z0 : Res ( f ; z0 ) = limz→z0 (z − z0 ) f (z)

(b) At a pole z0 of order m > 1, the residue of f is given by

 m−1 
Res ( f ; z0 ) = (m−1)!
1
 dz m−1 [(z − z0 )m f (z)] for m = 1, 2, 3, ...
d 
(6.7.1)
z=z0

Example 6.7.1. Let f (z) = 1/(z − 1)(z 2 + 1). Then z = 1, ±i are zeros of the denominator of order 1, and
hence are simple poles of f . Therefore,

n o
(a) Res ( f ; 1) = limz→1 (z − 1) f (z) = limz→1 1
z 2 +1
= 1
2
n o
(b) Res ( f ; i) = limz→i (z − i) f (z) = limz→i 1
(z−1)(z+i) = 1
(i−1)(2i) = 1+i
4i
n o
(c) Res ( f ; −i) = limz→−i (z + i) f (z) = limz→−i 1
(z−1)(z−i) = 1
(−i−1)(−2i) = i−1
4i

Department of Mathematics, Vellore Institute of Technology

30 6.7. Formula for the Residue at a Pole

ez
Example 6.7.2. Let f (z) = z 2 +π2
· Then z = ±πi are simple poles of f , and

 
(a) Res ( f ; πi) = limz→πi (z − πi) f (z) = limz→πi eπi
z+πi = i
2π
 
e−πi
(b) Res ( f ; −πi) = limz→−πi (z + πi) f (z) = limz→−πi z−πi = − 2π
i

Example 6.7.3. Let f (z) = z/cos z. Then cos z = 0 only if z = (2n + 1)π/2 = α, where n is any integer.
Also, the numerator is analytic and z , 0 at each α. Therefore, α is a simple pole of f , and

Res ( f ; α) = lim (z − α) f (z) = lim (z−α)z

= − lim 2z−α
= − sinα α ·
z→α z→α cos z z→α sin z

Example 6.7.4. Let f (z) = e2z /(z− 1)2 . Then z = 1 is a double pole of f . Write φ(z) = (z − 1)2 f (z) = e2z .
Then Res ( f ; 1) = (2−1)!
1
· φ0(1) = 2e2z z=1 = 2e2 .
Example 6.7.5. Consider f (z) = [(z + 1)(z − 1)]3 . Note that z = 1 is a triple pole for f . To find the residue
of f at z = 1, write φ(z) = (z − 1)3 f (z) = (z + 1)3 . Then φ 00(z) = 3.2(z + 1) = 3.2.2 = 12 at z = 1. Thus
Res ( f ; 1) = 12/2! = 6.

EXERCISE 6.7.1. Given that βn = (2n + 1)π/2, n = 0, ±1, ±2, ... are simple poles of f (z) = tan z/z, show
that the residue of f at βn is 1/βn .

EXERCISE 6.7.2. Find the poles and the residues there at for the following functions:

(a) f (z) = 1/(z + z 2 )

(b) f (z) = 1/(z 2 + 16)

(c) f (z) = 1/z(z − 1)2

Ans.

(a) z = 0 and −1 are simple poles; Res ( f ; 0) = 1, Res ( f ; −1) = −1

(b) z = ±4i are simple poles; Res ( f ; 4i) = −i/8, Res ( f ; −4i) = i/8

(c) z = 0 is a simple pole and z = 1 is a double pole; Res ( f ; 0) = 1, Res ( f ; 1) = −1

EXERCISE 6.7.3. Given that an = nπ, n = ±1, ±2, ... are simple poles, while 0 is pole of order 5 for
f (z) = cot z/z4 , find the residue of f at each pole.

EXERCISE 6.7.4. Classify all the singularities of the following functions and find the residues at appro-
priate singularities:
cos z sin z
(a) z 2 (z−π)3
(b) z 2 −4
(c) tan z
z 3z+2 sin2 z
(d) z 2 −6z+10
(e) z 4 +1
(f) z3
1 1 log(z+1)
(g) (1−e z ) (h) z 4 +2z 2 +1
(i) z2
2z+3 z2 1+e z
(j) z(z 2 −1)
(k) z 2 +a2
(l) z cos z+sin z

Department of Mathematics, Vellore Institute of Technology

Chapter 7

Contour Integration through Cauchy’s Residue Theorem

7.1 Cauchy’s Residue Theorem

If f is analytic on a simple (positively oriented) closed contour Γ and everywhere inside Γ except the finite
number of isolated singularities z1, z2, · · · zn , then
∳ n
Õ
f (z) dz = 2πi Res ( f ; zk ). (7.1.1)
Γ k=1
∳
1
Example 7.1.1. Evaluate (z−1)(z+2)2
dz, where Γ is the circle
Γ

(a) |z| = 1/2;

(b) |z| = 3/2;

(c) |z| = 5/2.

Solution. Define f (z) = 1/(z − 1)(z + 2)2 . Then z = 1, −2 are simple poles of f .

(a) We note that

 The radius of the circle Γ : |z| = 1/2 is r = 1/2

 Both the poles lie outside the circle Γ
 That is, f is analytic within and on the circle Γ
∳
 By Cauchy’s theorem, 1
(z−1)(z+2)2
dz = 0.
Γ

(b) The radius of the circle Γ : |z| = 3/2 is r = 3/2

 The pole z = 1 lies inside, while the pole z = −2 lies outside Γ

∳
 By Cauchy’s residue theorem, f (z) dz = 2πi Res ( f ; 1).
Γ
 Now Res ( f ; 1) = limz→1 (z − 1) f (z) = limz→1 1
(z+2)2
= 19 ·
∳
 Hence f (z) dz = 2πi/9.
Γ

(c) The radius of the circle Γ : |z| = 5/2 is r = 5/2

 Both the poles z = 1, −2 lie inside Γ

∳
 By Cauchy’s residue theorem, Γ f (z) dz = 2πi [Res ( f ; 1) + Res ( f ; −2)].
 Now, z = 1 is a simple pole and Res ( f ; 1) = 1/9
 while −2 is a double pole and
    
Res ( f ; 1) = (z + 2)2 f (z) z=−2 =  dz = − (z−1) = − 91
1  d
   d 1   1 
(2−1)! dz z−1  2
z=−2 z=−2
∳
f (z) dz = 2πi 1 1
= 0.


 Hence, 9 − 9
Γ

31
32 7.1. Cauchy’s Residue Theorem

∳
ez
Example 7.1.2. Evaluate cos πz dz
|z |=1
Solution. Define f (z) = ez /cos πz. The poles of f are given by cos πz = 0 so that z = ±1/2, ±3/2, ...

 The poles z = ±1/2 lie inside the unit circle Γ : |z| = 1

∳
 By Cauchy’s residue theorem, Γ f (z) dz = 2πi [Res ( f ; 1/2) + Res ( f ; −1/2)].

 Now,
(z−1/2)e z
Res ( f ; 1/2) = lim (z − 1/2) f (z) = lim = −e1/2 /π,
z→1/2 z→1/2 cos(πz)
(z+1/z)e z
Res ( f ; −1/2) = lim (z + 1/2) f (z) = lim cos(πz) = e−1/2 /π.
z→−1/2 z→−1/2

∳  1/2 
e−1/2
 Hence f (z dz = 2πi − e π + π = 4i sinh (1/2).
Γ
∳
Example 7.1.3. Evaluate 1
z 3 (z+4)
dz, where Γ is the circle |z + 2| = 3.
Γ
Solution. Let f (z) = 1/z 3 (z + 4). Then z = 0, −4 are poles of f , lying inside Γ.
∳
 By Cauchy’s residue theorem, f (z) dz = 2πi [Res ( f ; 0) + Res ( f ; −4)]
Γ

 To find the residue at z = 0, we use the Laurent’s expansion

h i
z
−1 z2 z3 z4
3
1
z (z+4)
= 1
4z 3 1 + 4 = 1
4z 3 1 − z
4 + 16 − 64 + 256 − · · · = 1
4z 3
− 1
16z 2
+ 1
64z − z
256 +···

Thus Res ( f ; 0) = 1/64. Also, Res ( f ; −4) = −1/64. Therefore,

∳
f (z) dz == 2πi 64 1 1
= 0.


 − 64
Γ
∳
EXERCISE 7.1.1. Find f (z) dz, where f and Γ are given as follows:
Γ

(a) f (z) = 3z 3 /(z − 1)(z 2 + 9); Γ : |z| = 4;

(b) f (z) = 1/(z − 1)2 (z − 2); Γ : |z| = 3;

(c) f (z) = z 3 /(z − 1)4 (z − 2)(z − 3); Γ : |z| = 7/2;

√
(d) f (z) = z/(z 2 − 1)2 (z 2 + 1); Γ : |z − 1| = 3.

Ans.
∳
f (z) dz = 2πi [Res ( f ; 1) + Res ( f ; −3i) + Res ( f ; 3i)] = 2πi 1
+ 15+49i
+ 15−49i
= 6πi


(a) 2 2 2
Γ
∳
(b) f (z) dz = 2πi [Res ( f ; 1) + Res ( f ; 2)] = 2πi(−1 + 1) = 0
Γ
∳
(c) f (z) dz = 2πi [Res ( f ; 1) + Res ( f ; 2) + Res ( f ; 3)] = −27πi/9
Γ
∳
(d) f (z) dz = 2πi [Res ( f ; i) + Res ( f ; −i) + Res ( f ; 1)] = πi/4.
Γ

Department of Mathematics, Vellore Institute of Technology

 Chapter 7. Contour Integration through Cauchy’s Residue Theorem 33

∳
EXERCISE 7.1.2. Use Cauchy’s residue theorem to find f (z) dz, where f and Γ are given below:
Γ

(a) f (z) = cos z/z 2 (z − π)3 ; Γ is the circle |z| = 5

(b) f (z) = sin z/(z 2 − 4); Γ is the circle |z| = 5π

(c) f (z) = tan z; Γ is the circle |z| = 2π

(d) f (z) = z/(z 2 − 6z + 10); Γ is the circle |z| = 8

(e) f (z) = (3z + 2)(z 2 + 1); Γ is the circle |z| = 2

(f) f (z) = cosh z/z(z 2 + 1); Γ is the circle |z| = 2

Department of Mathematics, Vellore Institute of Technology

34 7.1. Cauchy’s Residue Theorem

Department of Mathematics, Vellore Institute of Technology

Chapter 8

Real Integrals through Residue Theorem

8.1 Cauchy’s Residue Theorem

If f (z) is analytic inside and on a simple (positively oriented) closed contour γ, except at a finite number
of isolated singularities z1, z2, · · · , zn within it, then
∳ n
Õ
f (z) dz = 2πi Res ( f ; zk ). (8.1.1)
γ k=1

We shall find the following four types of real integrals using Cauchy’s Residue Theorem:

∫ 2π
(a) TYPE 1: Integrals around the unit circle |z| = 1 of the form Q(cos t, sin t) dt, where Q is a rational
0
function involving trigonometric sine and cosines.
∫ ∞
(b) TYPE 2: Improper integrals f (x) dx with f (x) = p(x)/q(x), where p and q are polynomials with
−∞
real coefficients and no factors in common; the complex form f (z) has a finite number of poles above
the real axis.
∫ ∞ ∫ ∞
(c) TYPE 3: Fourier integrals Q(x) cos ax dx, Q(x) sin ax dx, with Q(x) = p(x)/q(x), where p
−∞ −∞
and q are polynomials with real coefficients and no factors in common; the complex form f (z) has
a finite number of poles above the real axis.
∫ ∞
(d) TYPE 4: Improper integrals f (x) dx, where some of the poles of the complex form f (z) lie on
−∞
the real axis (indented contours).

8.2 Type 1 - Integrals around the Unit Circle

∫ 2π
Consider I = Q(cos t, sin t) dt, where Q is a rational function involving trigonometric sine and cosines.
0 ∳
We reduce I into a contour integral f (z) dz around the unit circle γ : |z| = 1.
γ

General Procedure

I Let z = eit be any point on γ : |z| = 1 so that

dz = ieit dt or dt = dz/iz, (8.2.1)

where 0 ≤ t ≤ 2π.

I Using Euler’s identity: eit = cos t + i sin t, we get

e i t +e−i t z 2 +1 e i t −e−i t z 2 −1
cos t = 2 = 2z , sin t = 2i = 2zi · (8.2.2)

35
36 8.2. Type 1 - Integrals around the Unit Circle

I Therefore,
∫ 2π ∳  
z 2 +1 z 2 −1
Q(cos t, sin t) dt = 1
iz ·Q 2z , 2zi dz. (8.2.3)
0 γ | {z }
= f (z)

I Find the poles z1, z2, · · · , zn , of f (z) lying within γ, and the residues of f (z) there at.

I Finally, apply (8.1.1).

∫ 2π
sin t
Example 8.2.1. Evaluate 5+4 cos t dt.
0
Solution. Consider the unit circle γ : |z| = 1. Employing (8.2.1) and (8.2.2), we get
∫ 2π ∳ ∳
(z 2 −1)/2iz z 2 −1
I= sin t
5+4 cos t dt = 1
iz · 5+4(z 2 +1)/2z
dz = − 41 z(z+1/2)(z+2) dz.
0 γ γ | {z }
= f (z)

Note that 0, −1/2 and −2 are simple poles of f , of which 0 and −1/2 lie inside γ. Therefore,

z 2 −1
(a) Res ( f ; −1/2) = limz→−1/2 (z + 1/2) f (z) = limz→−1/2 z(z+2) = 1,

z 2 −1
(b) Res ( f ; 0) = limz→0 z f (z) = limz→0 (z+1/2)(z+2) = −1.

In view of (8.1.1), I = − 14 · 2πi [Res ( f ; −1/2) + Res ( f ; 0)] = − 41 · 2πi(−1 + 1) = 0.

∫ 2π
dt
Example 8.2.2. Evaluate 2+sin t ·
0
Solution. Consider the unit circle γ : |z| = 1. Employing (8.2.1) and (8.2.2), we get
∫ 2π ∳ ∳
I= dt
2+sin t = 1
iz · 1
2+(z 2 −1)/2iz
dz = 2 1
z 2 +4iz−1
dz
0 γ γ | {z }
= f (z)
√ √
The poles of f are given by z 2 + 4iz − 1 = 0, which are the simple poles −2i ± 3. But a = −2i + 3 lies
inside γ. Therefore,

Res ( f ; a) = lim (z + a) f (z) = lim √1 √

= √1
z→a z→−2i+ 3 z−(−2i− 3) 2 3i

In view of (8.1.1), √ √ √
I = 2 · 2πi Res ( f ; a) = 2 · 2πi(1/2 3 i) = 2π/ 3 = 2 3 π/3.
∫ 2π
dt
Example 8.2.3. Evaluate sin2 t+4 cos2 t
·
0
Solution. Using (8.2.1) and (8.2.2), we have
∫ 2π ∳
I= dt
sin2 t+4 cos2 t
= 1
iz · 
(z 2 −1)2
1
(z 2 +1)2
 dz
0 γ − 2 +4 2
4z 4z
∳ ∳
= 4
i
z
4(z 2 +1)2 −(z 2 −1)2
dz = 4
i
z
(3z 2 +1)(z 2 +3)
dz.
γ γ | {z }
= f (z)

Department of Mathematics, Vellore Institute of Technology

 Chapter 8. Real Integrals through Residue Theorem 37

√ √ √
Note that ± i/ 3 and ± i 3 are simple poles of f (z); where ± i/ 3 lie inside γ. Therefore,

√ √
Res ( f ; i/ 3) = limz→i/√3 (z − i/ 3) f (z) = limz→i/√3 √ √ z = 1
16 ,
3( 3z+i)(z 2 +3)
√ √
Res ( f ; −i/ 3) = limz→−i/√3 (z + i/ 3) f (z) = limz→−i/√3 √ √ z = 1
16 .
3( 3z−i)(z 2 +3)

√ √
In view of (8.1.1), I =
· 2πi [Res ( f ; i/ 3) + Res ( f ; −i/ 3)] = 8πi
4 1
+ 1
= π.


i 16 16
∫ 2π
dt
Example 8.2.4. Compute a+cos t ·
0
Solution. In view of (8.2.1) and (8.2.2), we have
∫ 2π ∳ ∳
I= a+cos t =
dt 1
iz · 1
z 2 +1
dz = 2
i f (z) dz,
0 γ a+ 2z γ

where f (z) = 1
z 2 +2az+1
·

√ √
The poles of f are α = −a ± + a2 − 1 and β = −a − a2 − 1.

 √ 
Only, α lies inside γ , since −a ± + a2 − 1 < 1 (why?) Therefore,
 

Res ( f ; α) = lim (z − α) f (z) = lim 1

= 1
= √1 ·
z→α z→α z−β α−β 2 a2 −1

Hence,
∳
f (z) dz = 2πi Res ( f ; α) = 2πi · √1 = √ πi
γ 2 a2 −1 a2 −1

so that I = 2
i · √ πi = √ 2π .
a2 −1 a2 −1

EXERCISE 8.2.1. Evaluate the following real integrals using residue theorem:
∫ 2π
dθ
(a) 5+4 cos θ
0
∫ 2π
dθ
(b) 13+12 sin θ
0
∫ 2π
1
(c) 5+3 cos t dt
0
∫ 2π
cos 2t
(d) 5−4 cos t dt
0
∫ 2π
cos t
(e) 1+3 sin t dt
0

Answers

(a) 8π/3 (b) 2π/5 (c) π/2 (d) π/6 (e) 0

Department of Mathematics, Vellore Institute of Technology

38 8.3. Type 2 - Improper Integrals

8.3 Type 2 - Improper Integrals

Definition 8.3.1 (Improper Integrals). Let y = f (x) be a real-valued function, continuous on the interval
[0, ∞). We define
∫ ∞ ∫ R
I1 = f (x) dx = lim f (x) dx. (8.3.1)
0 R→∞ 0

If the limit in (8.3.1) exists, then the integral I1 is said to be convergent; otherwise, it is divergent. Similarly,
we define
∫ 0 ∫ 0
I2 = f (x) dx = lim f (x) dx. (8.3.2)
−∞ R→∞ R

If f is continuous on (−∞, ∞), we define

∫ ∞ ∫ ∞ ∫ 0
I= f (x) dx = f (x) dx + f (x) dx = I1 + I2 . (8.3.3)
−∞ 0 −∞

The integral I is convergent if and only if both I1 and I2 are convergent. We find

∫ ∞ ∫ R
I= f (x) dx = lim f (x) dx.
−∞ R→∞ −R

The limit in this, if it exists, is called the Cauchy principal value of the integral I and is written as
∫ ∞ ∫ R
P.V . f (x) dx = lim f (x) dx. (8.3.4)
−∞ R→∞ −R

About Cauchy principal value: Suppose that f (x) is continuous on (−∞, ∞) and is an even function, that is,
f (−x) = f (x). Then
∫ R ∫ R
f (x) dx = 2 f (x) dx.
−R 0
∫∞ ∫∞
Therefore, if the Cauchy principal value (8.3.4) exists, then both 0 f (x) dx and −∞ f (x) dx converge.
Hence the values of the integrals are
∫ ∞ ∫ ∞ ∫ ∞ ∫ ∞
f (x) dx = 2 P.V .
1
f (x) dx and f (x) dx = P.V . f (x) dx. (8.3.5)
0 0 −∞ −∞
∫ ∞
Consider an improper integral of the form I = f (x) dx with f (x) = p(x)/q(x), where p and q are
−∞
polynomials with real coefficients and no factors in common.

General Procedure of Evaluation: Choose the closed contour γ = [−R, R] ∪ ΓR , where

(a) [−R, R] is the interval of length 2R on the real axis, R >> 0 and

(b) ΓR is the arc of the semicircle |z| = R from (R, 0) to (−R, 0), in the anticlockwise direction in the
upper half plane. Along ΓR , z = eit , 0 ≤ t ≤ π.

Department of Mathematics, Vellore Institute of Technology

 Chapter 8. Real Integrals through Residue Theorem 39

Then
∳ ∫ R ∫
f (z) dz = f (x) dx + f (z) dz. (8.3.6)
γ −R ΓR

We find the singularities z1 , z2 , ..., zn of f , which lie inside γ (of course above the real line - why?), and
hence the residues there at. Then by residue theorem,
∳ n
Õ
f (z) dz = 2πi · Res ( f ; z j ). (8.3.7)
γ j=1
∫ 
 
Employing the limit as R → ∞ in (8.3.6), showing that  f (z) dz  → 0 as R → ∞ and using (8.3.7), we
ΓR
have
∫ ∞ ∫ R n
Õ
P.V . f (x) dx = lim f (x) dx = 2πi Res ( f ; z j ).
−∞ R→∞ −R j=1
∫ ∞
dx
Example 8.3.1. Find the Cauchy principal value of x 2 +1
·
−∞
Solution. Write f (z) = 1/(z2 + 1). Then z = ± i are simple poles of f . Choose the following paths:

I [−R, R]: a straight line along the real axis from −R to R

I ΓR : the semi-circular arc in the upper half plane from R to −R.

I γ = [−R, R] ∪ γR

Then
∫ R ∫ ∳
f (x) dx + f (z) dz = f (z) dz = 2πi Res ( f ; i) = 2πi lim (z+i)(z−i)
1
= 2πi
i+i = π. (8.3.8)
−R ΓR γ z→i

Note that |z 2 + 1| ≥  |z| 2 − 1 = R2 − 1, where R > 1, we have

 

∫  ∫ 
f (z) dz  = 
   dz 

  
z 2 +1 
≤ RπR
2 −1 → 0 as R → ∞.
γR

ΓR

Therefore proceeding the limit as R → ∞ in (8.3.8), and using this

∫ ∞ ∫ ∞
P.V . dx
x 2 +1
= π so that dx
x 2 +1
= 2π ·
−∞ 0
∫ ∞
Example 8.3.2. Verify that P.V . dx
(x 2 +a2 )2
= π
4a3
, a > 0.
0
Solution. Write f (z) = 1/(z2 + a2 )2 . Note that z = ± ai are double poles of f . Choose the following paths:

I [−R, R]: a straight line along the real axis from −R to R

I γR : the semi-circular arc in the upper half plane from R to −R.

I ΓR = [−R, R] ∪ γR

Department of Mathematics, Vellore Institute of Technology

40 8.3. Type 2 - Improper Integrals

Since the pole α = ai lies inside ΓR above the real axis, we have
∫ R ∫ ∳
f (x) dx + f (z) dz = f (z) dz = 2πi · B,
−R γR ΓR

where
n o n o
B = Res ( f ; ai) = lim d
(z − ai)2 · 1
(z−ai)2 (z+ai)2
= lim d 1
(z+ai)2
= − (2ai)
2
3 =
1
4a3 i
·
z→ai dz z→ai dz

Thus
∫ R ∫ ∫ R ∫
f (x) dx + f (z) dz = 2πi · 1
4a3 i
= π
2a3
or f (x) dx = π
2a3
− f (z) dz, R >> 0. (8.3.9)
−R γR −R γR

Now,
∫  ∫  ∫
f (z) dz  = 
   1
 1

(z 2 +a2 )2
dz  ≤ 2 +a 2 2
|dz|

γR γ γR | z |
∫ R ∫
1 1
≤ 2 2
|dz| ≤ (R2 −a2 )2 |dz|
γ R ( | z | −a )
2 2
γR
= πR
(R 2 −a2 )2
→ 0 as R → ∞.

Proceeding the limit as R → ∞ in (8.3.9), and then using this, we get

∫ ∞ ∫ ∞
P.V . dx
(x 2 +a2 )2
= π
2a3
so that P.V . dx
(x 2 +a2 )2
= π
4a3
·
−∞ 0
∫ ∞
Example 8.3.3. Verify that P.V . dx
x 4 +1
= π
√ ·
2 2
0
Solution. Write f (x) = 1/(x 4 + 1). The poles of f are given by z 4 + 1 = 0 or z = (−1)1/4 , which are the
fourth roots of −1, namely h i h i
(2k−1)π (2k−1)π
cos 4 + i sin 4 , k = 1, 2, 3, 4

That is,

z1 = cos 4 + i sin 4 = 2 , = cos + i sin 3π4 =

√ ,
π

π 1+i

3π


−1+i
√ z2 4 2
z3 = cos 4 + i sin 4 =
5π 5π √ ,
−1−i
z4 = cos 7π + 7π
= 1−i





i sin √ ·
2 4 4 2

Choose the following paths:

I [−R, R]: a straight line along the real axis from −R to R

I γR : the semi-circular arc in the upper half plane from R to −R

to trace the closed contour ΓR = [−R, R] ∪ γR . The poles z1 and z2 only lie above the real axis and inside
∫ R ∫ ∳
ΓR . Thus f (x) dx + f (z) dz = f (z) dz = 2πi · (B1 + B2 ), where
−R γR ΓR
 
B1 = Res ( f ; z1 ) = lim (z − z1 ) 1
z 4 +1
= lim 1
3 = 1
4z13
= z1
4z14
= − z41 ,
z→z1 z→z1 4z
 
B2 = Res ( f ; z2 ) = lim (z − z2 ) 1
z 4 +1
= lim 1
3 = 1
4z23
= z2
4z24
= − z42 ·
z→z2 z→z2 4z

Department of Mathematics, Vellore Institute of Technology

 Chapter 8. Real Integrals through Residue Theorem 41

Further,
h i
B1 + B2 = − z1 +z
4
2
= − 14 1+i
√ + −1+i
√ =− i
√ ·
2 2 2 2

Hence
∫ R ∫   ∫ R ∫
f (x) dx + f (z) dz = 2πi − √i = √π or f (x) dx = √π − f (z) dz, R >> 0. (8.3.10)
2 2 2 2
−R γR −R γR

Note that
∫  ∫ 
f (z) dz  =  dz  =
   1
 πR

z 4 +1 R 4 −1
→ 0 as R → ∞.
γR γR


Proceeding the limit as R → ∞ in (8.3.10), and then using this, we get

∫ ∞ ∫ ∞
P.V . dx
x 4 +1
= √ so that P.V .
π dx
x 4 +1
= π
√ ·
2 2 2
−∞ 0

EXERCISE 8.3.1. Find the Cauchy principal value of the following improper integrals:
∫ ∞
x2
(a) (x 2 +1)(x 2 +4)
dx
−∞
∫ ∞
x2
(b) (x 2 +1)2
dx
0
∫ ∞
1
(c) x 2 −2x+2
dx
−∞
∫ ∞
x 2 −x+2
(d) x 4 +10x 2 +9
dx
−∞
∫ ∞
x2
(e) x 4 +1
dx
0
∫ ∞
1
(f) (x 2 +1)3
dx
−∞
∫ ∞
1
(g) x 6 +1
dx
0

Ans.
√
(a) π/2 2

(b) π/2

(c) π

(d) 5π/12
√
(e) π/2 2

(f) 3π/8

(g) π/3

Department of Mathematics, Vellore Institute of Technology

42 8.4. Type 3 - Fourier Integrals

8.4 Type 3 - Fourier Integrals

∫ ∞ ∫ ∞
Consider Fourier integrals Q(x) cos ax dx and Q(x) sin ax dx, with Q(x) = p(x)/q(x), where p
−∞ −∞
and q are polynomials with real coefficients and no factors in common; the complex form f (z) has a finite
number of poles above the real axis.

∫ ∞
Fourier integrals appear as the real and imaginary parts in the improper integral Q(x)eiax dx as given
−∞
below:
∫ ∞ ∫ ∞ ∫ ∞
Q(x)e iax
dx = Q(x) cos ax dx + i Q(x) sin ax dx,
−∞ −∞ −∞

whenever both integrals on the right-hand side converge. The Fourier integrals are evaluated as in Type 2
integrals.
∫ ∞
Example 8.4.1. Verify that P.V . x sin mx dx
x 2 +a2
= e−am π/2, a > 0, m > 0.
0
Solution. Let f (z) = + 1). Then α = ai is a simple pole lying above the real axis. Suppose that
zeimz /(z 2
γ = [−R, R] ∪ ΓR as in the previous problems. Then
∳
iz −a m
f (z) dz = 2πi Res ( f ; ai) = 2πi lim (z − ai) ze2 +1 = 2πi · iae2ia = πi.e−am
γ z→ai

or
∫ R ∫
f (x) dx + f (z) dz = πi.e−am · (8.4.1)
−R ΓR

Any point on ΓR is of the form: z = Reiψ = R(cos ψ + i sin ψ) so that

eiz = ei (R(cos ψ)+i sin ψ) = e−R sin ψ .ei R cos ψ so that eiz  = e−R sin ψ
 

∫  −R sin ψ
f (z) dz  ≤ πRe → 0 as R → ∞. In the limit as R → ∞, (8.4.1) becomes
 
and hence  γ R2 −a2
R

∫ ∞ h i
x(cos mx+i sin mx)
P.V . x 2 +1
dx = πi e−am ·
−∞

Comparing the imaginary parts on both sides,

∫ ∞  ∫ ∞  
P.V . x sin mx
x 2 +1
dx = π e −am
so that P.V . x sin mx
x 2 +1
dx = π
2 · e−am
−∞ 0
∫ ∞
Example 8.4.2. Verify that P.V . cos 3x
(x 2 +1)2
dx = π/e3 .
0
Solution. Note that
∫ ∞ ∫ ∞ ∫ ∞
ei3x
cos 3x
(x 2 +1)2
dx = 1
2 · cos 3x
(x 2 +1)2
dx = 1
2 · Re (x 2 +1)2
dx .
0 −∞ −∞
| {z }
=I

We evaluate the Cauchy principal value of I.

Department of Mathematics, Vellore Institute of Technology

 Chapter 8. Real Integrals through Residue Theorem 43

Let f (z) = ei3z /(z 2 + 1)2 . Then f has a double pole at z = i in the upper half plane. Therefore,
∫ ∞  
ei3x  = 2πi d  e i 3z2  = 2π3 .
= = 2
d 
 
(x +1)
2 2 dx 2πi Res ( f ; i) 2πi dx (z − i) f (z) z=i dx (z+i) e
−∞ z=i
∫ ∞ ∫ 
Thus P.V . cos 3x
dx = 1
· 2π
= π
provided  γ f (z) dz  → 0 as R → ∞.
 
(x 2 +1)2 2 e3 e3
,
R
0

EXERCISE 8.4.1. Find the Cauchy principal value of the following improper integrals:
∫ ∞
cos mx
(a) x 4 +4a4
dx
0
∫ ∞
x sin x
(b) (x 2 +1)(x 2 +4)
dx
0
∫ ∞
sin x
(c) (x 2 +4x+5)
dx
−∞
∫ ∞
cos 3x
(d) (x 2 +4x+5)
dx
0

Ans.

(a) π −am
8a3
e (cos am + sin am), a > 0, m ≥ 0
π −1
(b) 6 (e − e−2 )
π
(c) − e sin 2
2π
(d) e3

8.5 Type 4 - Using Indented Contours

First we prove the following result which is useful in evaluating the integrals through indented contours:
Theorem 8.5.1. Suppose that f (z) has a simple pole z = c on the real axis. If Γ is the contour in the
parametric form z = c +  eiθ for 0 ≤ θ ≤ π, then
∫
lim f (z) dz = π Res ( f ; c). (8.5.1)
 →0 Γ

Proof. Since z = c is a simple pole of f , its Laurent’s series is given by

f (z) = a−1
z−c + g(z),

where g(z) gives its analytic part, and a−1 is the residue of f at z = c. Hence
∫ ∫ ∫
f (z) dz = a−1 z−c dz +
1
g(z) dz. (8.5.2)
Γ Γ Γ

Now, on Γ : z − c =  eiθ so that dz = ieiθ dθ and θ varies from 0 to π. Therefore,

∫ ∫ π ∫ π
1
z−c dz = 1
 ei θ
· ie iθ
dθ = i dθ = πi. (8.5.3)
Γ 0 0

Department of Mathematics, Vellore Institute of Technology

44 8.5. Type 4 - Using Indented Contours

And g(z) is analytic at c so that it is differentiable and hence continuous at c, and is bounded in a small
neighborhood Nδ (c) = {z : |z − c| < δ} of c. That is there exists an M > 0 such that |g(z)| ≤ M on Nδ (c).
Hence
∫  ∫
dz = M · π → 0 as  → 0.
 

 g(z) dz ≤ M
 (8.5.4)
Γ Γ

Finally, employing the limit as  → 0 in (8.5.2), and then using (8.5.3) and (8.5.4), we obtain (8.5.1). 
∫ ∞
Example 8.5.1. Show that P.V . sin x
x dx = π/2
0
Solution. Note that
∫ ∞ ∫ ∞  ∫ ∞ 
ei x
P.V . sin x
x dx = 1
2 · P.V . sin x
x dx = 1
2 · Im P.V . x dx .
0 −∞ −∞

Let f (z) = eiz /z. Then f has a simple pole at z = 0. Consider the closed contour

C = γR ∪ [−R, −] ∪ γ ∪ [, R],

where γR is the circle |z| = R with centre at the origin ans radius R, while γ is the circle |z| =  with
centre at the origin ans radius . Therefore,
∳ ∫ ∫ − ∫ ∫ R
f (z) dz = f (z) dz + f (x) dx + f (z) dz + f (x) dx. (8.5.5)
C γR −R γ 

As  → 0, in view of Theorem 8.5.1, the third integral tends to

−πi · Res( f , 0) = −πi lim z f (z) = −πi lim eiz = −πi · e0 = −πi. (8.5.6)
z→0 z→0

Now we estimate IR =
∫
γR
f (z) dz.

  √
Let z = x + iy so that eiz = ei(x+iy) = eix e−y . Since eix  = cos2 x + sin2 x = 1, it follows that eiz  = e−y .
 

Further, on the arc γR , from the parametric form of z, we see that

z = Reiφ so that dz = iReiφ dφ and |dz| = Rdφ

y = R sin φ, φ : 0 → π, |z| = R

Therefore,
∫  ∫ ∫ π ∫ π
| ei z | e−y
Rdφ = e−y dφ.
 
 f (z) dz  ≤ |z | |dz| ≤ R
γR γR

0 0

But sin φ ≥ 2φ/π so that −R sin φ ≤ −2Rφ/π and e−R sin φ ≤ e−2Rφ/π . With this substitution, the above
inequality becomes
∫  ∫ π ∫ π/2
−2Rφ/π
dφ = 2 e−2Rφ/π dφ
 
 f (z) dz  ≤
 e
γR

0 0
π(1−e−R )
= R → 0 as R → ∞. (8.5.7)

Department of Mathematics, Vellore Institute of Technology

 Chapter 8. Real Integrals through Residue Theorem 45

Moreover, since there is no pole lying inside the contour C, by Cauchy’s theorem,
∳
f (z) dz = 0. (8.5.8)
C

Proceeding the limit as R → ∞ and  → 0 in (8.5.5), and then using (8.5.6), (8.5.7) and (8.5.8),
∫ ∞
ei x
P.V . x dx = πi.
∞

∫ ∞
Comparing the imaginary parts both sides, we finally get P.V . sin x
x dx = π/2.
0

EXERCISE 8.5.1. Find the Cauchy principal value of the following improper integrals through indented
contours:
∫ ∞ ∫ ∞ ∫ ∞
sin x sin x 1−cos x
(a) x(x 2 −2x+2)
dx (b) x(x 2 +1)
dx (c) x2
dx
−∞ −∞ 0

 
1+ sin 1−cos 1 1


Ans. (a) π
2 e (b) π 1 − e (c) π/2

Department of Mathematics, Vellore Institute of Technology

46 8.5. Type 4 - Using Indented Contours

Department of Mathematics, Vellore Institute of Technology

Chapter 9

Partial Differential Equations and their Formation

9.1 Introduction
Definition 9.1.1. A relation which binds a function of two or more independent variables, and its partial
derivatives upto some order is known as a partial differential equation. The order of the highest derivative
which appears in a partial differential equation is its order.

Linear and Nonlinear Partial Differential Equations

Definition 9.1.2 (Linear Equation). A partial differential equation is said to be linear, if

(a) it has no product terms of the dependent variable with its partial derivatives,
(b) the dependent variable and its partial derivatives are not in transcendental form (trigonometric,
hyperbolic, exponential, logarithmic etc.),
(c) there are no square roots, cube roots and radicals of some order of the dependent variable and its
partial derivatives so that their degree is only one.
Example 9.1.1. The following are linear:

∂2 f ∂2 f
(a) Laplace equation: ∂x 2
+ ∂y 2
=0
∂2 f ∂ f
2
(b) One dimensional wave equation: ∂t 2
= ν2 ∂x 2

∂f ∂ f
2
(c) One dimensional heat equation or Diffusion Equation: ∂t = ν2 ∂x 2

∂f ∂f
(d) Transport equation: ∂t + 2 ∂x
Definition 9.1.3 (Quasi-linear and Almost Linear Equations). A partial differential equation is said to
be quasi-linear, if it is linear with respect to highest order derivatives and almost linear, if the degree of
dependent variable is not linear, while all the other terms involving the partial derivatives are linear.
Example 9.1.2. The following are quasi-linear:

∂u
(a) Equation for shock waves: ∂x + u ∂u
∂y = 0

(b) Burger’s equation: ∂u ∂u ∂ u 2

∂t + cu ∂x =  ∂x 2
  2 2   2
∂f ∂ f ∂ f ∂ f ∂2 f ∂f ∂2 f
(c) 1 − ∂t ∂x 2
+ 2 ∂x ∂t ∂x∂t − 1 + ∂x ∂t 2
=0

∂2 u ∂2 u
Example 9.1.3. The equation ∂t 2
− ∂x 2
+ u3 = 0 is almost linear.
Definition 9.1.4 (Fully Noninear Equation). A partial differential equation is said to be fully nonlinear, if it
is nonlinear with respect to highest order derivatives.
 2  2  2 2
Example 9.1.4. The equations ∂u ∂u ∂ u
− ∂∂xu2 ∂∂yu2 = x 2 + y 2 are fully nonlinear.
2 2
∂x + ∂y = 1 and ∂x∂y

∂z
Notation: We regard z as a function of independent variables x and y and employ the notation: zx = ∂x =p
∂z ∂2 z ∂2 z ∂2 z
and zy = ∂y = q, zxx = ∂x 2 , zyy = ∂y 2 , z xy = ∂y∂x .

47
48 9.2. Formation of a Partial Differential Equation - Elimination of Arbitrary Constants

EXERCISE 9.1.1. Classify each of the following partial differential equations with respect to the linearity
and identify the order in each case:

∂2 u ∂u
(a) ∂x 2
− ∂t = cos(ax + bt)
∂2 u ∂2 u
(b) x 2 · ∂x 2
+ y2 · ∂y 2
= u1/3

(c) x 2 ∂∂xu2 − y 3 ∂∂yu2 = ∂3 u

2 2
∂x 3

∂2 f ∂f
(d) ∂x∂t + 2u2 · ∂y − 4t = 0

∂ u
− 3t ∂u
2
(e) 5x y ∂x∂y ∂y + 2u = 0
3
∂f ∂3 f

(f) ∂t − ∂x 3
+ t4 = 0

∂f
3
(g) x ∂u
∂t − y ∂y =1

Ans.

(a) linear, second order

(b) almost linear, second order

(c) linear, third order

(d) quasi-linear, second order

(e) linear, second order

(f) quasi-linear, third order

(g) nonlinear, first order

9.2 Formation of a Partial Differential Equation - Elimination of Arbitrary

Constants
EXERCISE 9.2.1. Obtain a partial differential equation from each of the following relations, where arbi-
trary constants are mentioned in braces:

(a) z = ax 2 − by 2 (a, b)

(b) z = ax + by + a2 + b2 (a, b)

(c) z = a(x + y) + b(x − y) + abt + c (a, b, c)

(d) z = ax 2 + bx y + cy 2 (a, b, c)

(e) z = (x − a)2 + (y − b)2 + 1 (a, b)

(f) a sin x + b cos y = 2z (a, b)

(g) ax 2 + by 2 + z 2 = 1 (a, b)

(h) 2z = (ax + y)2 + b (a, b)

(j) ax 2 + by 2 + cz2 = 1 (a, b, c)

Department of Mathematics, Vellore Institute of Technology

 Chapter 9. Partial Differential Equations and their Formation 49

Ans.
(a) 2z = px + qy
(b) z = px + qy + p2 + q2 , other possibility is z = p2 y 2 + q2 x 2 + 2x 2 y 2 (px + qy) = 4x 2 y 2 z
∂z
(c) p2 − q2 = 4 · ∂t

(d) 2z = px + qy
(e) p2 + q2 = 4(z − 1)
(f) p tan x − q cot y = z
∂2 z ∂z ∂z
(g) (px + qy)z = z2 − 1, other possibility is z · ∂x∂y + ∂x · ∂y =0
∂2 z ∂2 z ∂z ∂z
(h) ∂y 2
= 1other possibility is z · ∂x∂y + ∂x · ∂y =0
∂2 z ∂z ∂z
(j) z · ∂x∂y + ∂x · ∂y =0

EXERCISE 9.2.2. Obtain a partial differential equation from each of the following relations, where arbi-
trary constants are mentioned in braces:

(a) (x − a)2 + (y − a)2 + (z − b)2 = 1 (a, b)

(b) z = ax + by + cxy (a, b, c)
(c) log(az − 1) = x + ay + b (a, b)
√
(d) z = (x + x 2 − a2 )y + b (a, b)
(e) z = (x 2 + a)(y 2 − b) (a, b)
Ans.
p−q
2
(a) p2 + q2 + 1 = y−x

∂2 z ∂2 z ∂2 z
(b) ∂x 2
= 0, other possibilities are ∂y 2
= 0 and z = px + qy + ∂x∂y

(c) p(q + 1) = qz
(d) px + qy = pq
(e) pq = 4x yz

EXERCISE 9.2.3. Obtain a partial differential equation from each of the following families of surfaces:

(a) z = aebx sin by

(b) family of all planes, which are at a constant distance k units from the origin
(c) family of all spheres with centres lying on the z-axis
(d) family of all spheres with unit radius, and centres lying on the line y = x in the x y-plane
Ans.
∂2 z ∂2 z
(a) ∂x 2
+ ∂y 2
=0

(b) z = px + qy − k p2 + q2 + 1 = z
p

(c) qx = py
(d) (p2 + q2 + 1)z 2 = 1

Department of Mathematics, Vellore Institute of Technology

50 9.3. Formation of a Partial Differential Equation - Elimination of Arbitrary Functions

9.3 Formation of a Partial Differential Equation - Elimination of Arbitrary

Functions
Consider the relation

φ(u, v) = 0, (9.3.1)

where u = u(x, y, z) and v = v(x, y, z) are functions of x, y and z. We wish to derive a partial differential
equation by eliminating the arbitrary function φ from the relation (9.3.1). Indeed, differentiating partially
with respect to x and y and using chain rule of partial differentiation, (9.3.1) gives

∂φ ∂u ∂u ∂z ∂φ ∂v ∂v ∂z ∂φ ∂u ∂u ∂z ∂φ ∂v ∂v ∂z
       
+ + + = 0 and + + + =0
∂u ∂ x ∂z ∂ x ∂v ∂ x ∂z ∂ x ∂u ∂ y ∂z ∂ y ∂v ∂ y ∂z ∂ y
or
∂φ ∂u ∂u ∂φ ∂v ∂v ∂φ ∂u ∂u ∂φ ∂v ∂v
       
+ p + + p = 0 and + q + + q = 0.
∂u ∂ x ∂z ∂v ∂ x ∂z ∂u ∂ y ∂z ∂v ∂ y ∂z

Solving these simultaneous equations for ∂φ/∂u and ∂φ/∂v , we get the determinant relation:

   ∂u ∂u ∂v ∂v 
 ∂x + ∂z p ∂x + ∂z p
 
J x,y =  ∂u ∂u
u,v
 = 0. (9.3.2)
 + q ∂v + ∂v q 
∂y ∂z ∂y ∂z
 
We realize that u and v are functionally dependent, and hence from the theory of Jacobians, J u,v x,y = 0.
 
x,y contains the partial derivatives p = ∂z/∂ x and q = ∂z/∂ y . Hence,
Since z is a function of x and y, J u,v
(9.3.2) gives a partial differential equation of first order.

EXERCISE 9.3.1. Obtain a partial differential equation from each of the following relations by eliminat-
ing the arbitrary function f :

(a) f (x 2 + y 2 + z 2, x yz) = 0

(b) f (x 2 + y 2 ) = z − x y

(c) f (x 2 + y 2, y + z 2 ) = 0

(d) z = f (x 2 + y 2 )

(e) f (x 2 − y 2 ) = z/(x + y)

(f) f (x 2 + y 2 + z 2 ) = z

(g) z − x − y = f (x y)

(h) (x + y) f (x y + yz + zx) = z

(i) f (x 2 + y 2 + z 2 ) = y/x

(j) f (x y + z 2 ) = x + y + z

(k) f (z sin x, z cos y) = 0

Ans.

(a) px(y 2 − z 2 ) + qy(z2 − x 2 ) = z(x 2 − y 2 )

Department of Mathematics, Vellore Institute of Technology

 Chapter 9. Partial Differential Equations and their Formation 51

(b) py − qx = y 2 − x 2

(c) py − qx = x y/z

(d) py − qx = 0

(e) px + qy = z

(f) py − qx = 0

(g) px − qy = x − y

(h) [p(x + 2z) − q(y + 2z)](x + y) = z(x − y)

(i) (px + qy)z + x 2 + y 2 = 0

(j) (2z − x)p − (2z − y)q = x − y

(k) p tan x − q cot y + z = 0

EXERCISE 9.3.2. Obtain a partial differential equation from each of the following relations by eliminat-
ing the arbitrary functions µ and δ:

(a) z = xµ(y/x) + xδ(y/x)

(b) z = yµ(x) + xδ(y)

(c) z = µ(x) + ey δ(x)

(d) z = µ(x + ct) + δ(x − ct)

(e) z = µ(y − ax) + δ(y − bx)

Ans.
∂ z 2 ∂ z 2
2∂ z 2
(a) x 2 ∂x 2 + 2x y ∂x∂y + y ∂y 2 = 0

∂z ∂z ∂ z 2
(b) x ∂x + y ∂y = x y ∂x∂y +z

∂2 z ∂z
(c) ∂y 2
= ∂y

∂2 z ∂ z 2
(d) ∂t 2
= c2 ∂x 2

∂2 z ∂ z
2 ∂ z
2
(e) ∂x 2
− (a − b) ∂x∂y + ab ∂y 2 = 0

Department of Mathematics, Vellore Institute of Technology

52 9.3. Formation of a Partial Differential Equation - Elimination of Arbitrary Functions

Department of Mathematics, Vellore Institute of Technology

Chapter 10

Partial Differential Equations of First Order

10.1 Solution of a Partial Differential Equation

we begin with
Definition 10.1.1 (General Integral). A solution of a partial differential equation, which has the maximum
number of arbitrary functions is called its general integral or general solution.

We shall discuss the solutions of quasi-linear and non-linear partial differential equations of first order.

Notation: We regard z as a function of independent variables x and y and employ the notation:

∂z ∂z
zx = ∂x = p, zy = ∂y = q,
∂2 z ∂2 z ∂2 z
zxx = ,z
∂x 2 yy
= ,z
∂y 2 xy
= ∂y∂x .

10.2 Quasi-linear Partial Differential Equations of First Order

Consider the Lagrange’s quasi-linear equation of first order:

P(x, y, z)p + Q(x, y, z)q = R(x, y, z) or Pp + Qq − R = 0. (10.2.1)

Let z = z(x, y) be a solution of (10.2.1). Then

f (x, y, z) ≡ z(x, y) − z = 0 (10.2.2)

represents an integral surface. We recall from vector calculus that the normal at a point P(x, y, z) on the
integral surface (10.2.2) is given by its gradient function:

∇ f = (zx , zy , −1) = (p, q, −1). (10.2.3)

Note that the left hand side of (10.2.1) is written as

Pp + Qq − R = (Pi + Qj + Rk) · (pi + qj + k) = V · ∇ f ,

where

V = Pi + Qj + Rk. (10.2.4)

Inserting (10.2.4) in (10.2.1), we see that

V · ∇ f = 0.

Thus V is perpendicular ∇ f . Since ∇ f is normal at P, V = Pi + Qj + Rk is tangent at P. Geometrically,

V defines a direction field, called the characteristic field.

Again, let C be a space curve on the surface (10.2.2), parametrized as

r = x(t)i + y(t)j + z(t)k, t ∈ [t1, t2 ] (10.2.5)

53
54 10.3. Solution of Lagrange’s Equation

dy
then dx
dt i + dt j + dz
dt k represents its tangent. Comparing this with (10.2.4), we find that
dx/dt dy/dt dz/dt dy
P = Q = R or dx
P = Q = dz
R = dt. (10.2.6)

Solving any pair of the auxiliary equations (10.2.6), we obtain two linearly independent solutions

u(x, y, z) = a and v(x, y, z) = b, (10.2.7)

where a and b are arbitrary constants. Each pair of the level surfaces u = a and v = b represents a unique
integral curve, called the characteristic. The locus of all characteristics (10.2.7), obtained by assigning an
arbitrary functional relation Φ(a, b) = 0 between a and b, that is

Φ(a, b) = 0 or Φ u(x, y, z), v(x, y, z) = 0

(10.2.8)

is also an integral surface, and gives the general integral or general solution of the partial differential equation
(10.2.1).

10.3 Solution of Lagrange’s Equation

We follow the working rule, given below for finding the general integral of (10.2.1):

STEP 1. Write the auxiliary equations (10.2.6)

STEP 2. Find any two linearly independent solutions (10.2.7) of (10.2.6)

STEP 3. The general integral of (10.2.1) is given by (10.2.8).

Linearly independent solutions in STEP 2 are obtained either in two ways:

(a) Method of Grouping: any pair of fractions in (10.2.6)

(b) Method of multipliers: Let l, m, n be one set of multipliers in (10.2.6). Then each ratio in it equals the
pooled ratio
l dx+mdy+ndz
lP+mQ+nR · (10.3.1)

The multipliers l, m and n are chosen such that

lP + mQ + nR = 0 (10.3.2)

and the numerator of the pooled ratio can be grouped as the total differential of some function
u(x, y, z). That is

d[u(x, y, z)] = 0, (10.3.3)

which on integration yields a solution u(x, y, z) = a. Similarly, choose another set of multipliers to
get a second solution v(x, y, z) = b.

EXERCISE 10.3.1. Find the general integral of each of the following first order partial differential equa-
tions:

√ √ √
(a) p x + q y = z

(b) pyz + qzx = xy

Department of Mathematics, Vellore Institute of Technology

 Chapter 10. Partial Differential Equations of First Order 55

(c) py 2 z + qx 2 z = y 2 x

(d) py 2 − qx y = x(z − 2y)

(e) p tan x + q tan y = tan z

Ans.
√ √
(a) Auxiliary equations are √dxx = √dyy = √dzz . Grouping the first two ratios and solving, we get x − y = a.
√ √
Grouping the second and the third ratios and solving, y − z = b. Therefore, the general integral
√ √ √ √ √ √ √ √
is f ( x − y, y − z) = 0. Other form of general integral is g( x − y, x − z) = 0.
dy
yz = zx = xy . Grouping the first two ratios and solving, we get x − y = a.
dz
(b) Auxiliary equations are dx 2 2

Grouping the second and the third ratios and solving, y − z = b. Therefore, the general integral is
2 2

f (x 2 − y 2, y 2 − z 2 ) = 0. Other form of general integral is g(y 2 − z 2, x 2 − z 2 ) = 0.

dy
2 z = zx 2 = xy 2 . Grouping the first two ratios and solving, x − y
(c) Auxiliary equations are ydx dz 3 3 = a.

Grouping the first and the third ratios and solving, x 2 − z 2 = b. Therefore, the general integral is
f (x 3 − y 3, x 2 − z 2 ) = 0.
dy
(d) Auxiliary equations are dx
y2
= −xy = x(z−2y)
dz
. Grouping the first two ratios and solving, x 2 + y 2 = a.
From the second and the third ratios, y dz + z dy − 2y dy = 0. This on integration gives the second
solution yz − y 2 = b. Therefore, the general integral is f (x 2 + y 2, yz − y 2 ) = 0.
dy
x = tan y = tan z . Grouping the first two ratios, cos y
= a. From the
dx dz cos x
(e) Auxiliary equations are tan
second and the third ratios, cos y cos x cos y
cos z = b. Therefore, the general integral is f cos y , cos z = 0.

EXERCISE 10.3.2. Find the general integral of each of the following first order partial differential equa-
tions, using appropriate multipliers:

(a) p(y + z) − q(z + x) = x + y

(b) x(y − z)p + y(z − x)q = z(x − y)

(c) x 2 (y − z)p + y 2 (z − x)q = z2 (x − y)

(d) x(y 2 − z 2 )p + y(z 2 − x 2 )q = z(y 2 − x 2 )

     
(e) y−z
yz p + z−x
zx q = x−y
xy

Ans.
dy
(a) Auxiliary equations are y+z dx
= −z−x = x+y
dz
. Choosing (1, 1, 1) as multipliers, the numerator of the
pooled ratio gives dx + dy + dz = 0 or d(x + y + z) = 0. Integrating this total differential, one
solution is x + y + z = a. Choosing (x, y, −z) as multipliers, the numerator of the pooled ratio gives
x dx + y dy + z dz = 0 or d(x 2 + y 2 − z 2 ) = 0. Integrating this total differential, second solution is
x 2 + y 2 − z 2 = b. Therefore, the general integral is f (x + y + z, x 2 + y 2 − z 2 ) = 0.
dy
dx
(b) Auxiliary equations are x(y−z) = y(z−x) = z(x−y)
dz
. Choosing (1, 1, 1) as multipliers, the numerator
of the pooled ratio gives dx + dy + dz = 0 or d(x + y + z) = 0. Integrating this total differential,
one solution is x + y + z = a. Choosing (1/x, 1/y, 1/z) as multipliers, the numerator of the pooled
ratio gives x −1 dx + y −1 dy + z −1 dz = 0. Integrating this, one more solution is log(x yz) = log b or
xyz = b. This can be achieved with multipliers (yz, zx, xy) also. Therefore, the general integral is
f (x + y + z, x yz) = 0.

Department of Mathematics, Vellore Institute of Technology

56 10.3. Solution of Lagrange’s Equation

dy
dx
(c) Auxiliary equations are x 2 (y−z) = y 2 (z−x) = z 2 (x−y)
dz
. Choosing (1/x 2, 1/y 2, 1/z 2 ) as multipliers, the
numerator of the pooled ratio gives x −2 dx + y −2 dy + z −2 dz = 0. Integrating this, one solution is
− 1x − y1 − 1z = −a or 1x + y1 + 1z = a. Choosing (yz, zx, xy) as multipliers, the numerator of the pooled
ratio gives yz dx + zx dy + xy dz = 0. Integrating this, the second solution is xyz = b. Therefore,
the general integral is f x + y + z , x yz = 0.
1 1 1

dy
2 −z 2 ) = y(z 2 −x 2 ) = z(y 2 −x 2 ) . With one set of multipliers (1/x, 1/y, −1/z),
dz
(d) Auxiliary equations are x(ydx
one solution is xy
z = a. Choosing (x, y, −z) as multipliers, the second solution is x 2 + y 2 − z 2 = b.
xy 2
Therefore, the general integral is f z , x + y 2 − z 2 = 0.

dy
dx
(e) Auxiliary equations are 1/z−1/y = 1/x−1/z = 1/y−1/x
dz
. With one set of multipliers (1, 1, 1), one solution
is x + y + z = a. Choosing (yz, zx, x
y) as multipliers, the second solution is xyz = b. Therefore, the
general integral is f x + y + z, x yz = 0.

EXERCISE 10.3.3. Find the general integral of each of the following first order partial differential equa-
tions:

(a) (x 2 − yz)p + (y 2 − zx)q = z 2 − x y

(b) p cos(x + y) + q sin(x + y) = z

(c) (z 2 − 2yz − y 2 )p + x(y + z)q = x(y − z)

(d) px(z − 2y 2 ) + qy(z − y 2 − 2x 3 ) = z(z − y 2 − 2x 3 )

Ans.
dy
(a) Auxiliary equations are dx
x 2 −yz
=
y 2 −zx
= z 2dz
−xy
. With two sets of multipliers (1, −1, 0) and (0, 1, −1),
dx−dy dy−dz
we get the combined fractions: x 2 −yz−y 2 +zx = y 2 −zx−z 2 +xy . Canceling the common factor x + y + z
dx−dy dy−dz
from the denominators, we get x−y = y−z · This gives one solution: x−y y−z = a. Similarly, by
y−z
symmetry, the second solution will be z−x = b. Therefore, the general solution is f x−y , y−z

y−z z−x = 0.

dy
dx
(b) Auxiliary equations are cos(x+y) = sin(x+y) = dz
z . With two sets of multipliers (1, 1, 0) and (1, −1, 0),
we get the combined fractions:
dx+dy dx−dy
cos(x+y)+sin(x+y) = cos(x+y)−sin(x+y) ,

which can be written as h i

cos(x+y)−sin(x+y)
sin(x+y)+cos(x+y) d(x + y) + d(y − x) = 0.

Integrating this, one solution is

log[sin(x + y) + cos(x + y)] + y − x = a.

Now from the first and third ratios,

dx+dy dx+dy
√
cos(x+y)+sin(x+y) = dz
z or cos(π/4) cos(x+y)+sin(π/4) sin(x+y) = 2 dz
z
√
so that sec(x + y − π/4) d(x + y) − 2 dz/z = 0. Integrating this, the second solution is
n  o √
log tan x+y−π/4
2 + π
4 − 2 log z = b.

Department of Mathematics, Vellore Institute of Technology

 Chapter 10. Partial Differential Equations of First Order 57

Then the general integral is


x+y
√ 
f log[sin(x + y) + cos(x + y)] + y − x, log tan + − 2 log z = 0.
 π


2 8

dy dy
2 = x(y+z) = x(y−z) . From the last two fractions: y+z = y−z , which
dx dz dz
(c) Auxiliary equations are z 2 −2yz−y
on rearranging 2y dy − 2(z dy + y dz) − 2z dz = 0. The first solution is y 2 − 2yz − z 2 = a. Now,
with multipliers (x, y, z), the second solution is x 2 + y 2 + z 2 = b. Hence the general integral is
f (y 2 − 2yz − z 2, x 2 + y 2 + z 2 ) = 0.
dy dy
2 ) = y(z−y 2 −2x 3 ) = z(z−y 2 −2x 3 ) . From the last two fractions: y = z
dx dz dz
(d) Auxiliary equations are x(z−2y
which gives the first solution z/y = a. Now, substitute z = ay in the first two fractions and simplify
dx
to get x(a−2y) = ay−ydy
2 −2x 3 . This implies, −a(x dy − y dx) + 2x y dy − y dx − 2x dx = 0. Dividing
2 3
2
by x 2 and grouping the terms, −a d yx + yx − x 2 = 0. Integrating this, we get the second solution


y y2 z y y2
x + x − x = b. Hence the general integral is f y , x + x − x
2 2 = 0.

10.4 Nonlinear Partial Differential Equations of First Order

Definition 10.4.1 (Complete Integral). A solution of a partial differential equation, in which the number of
arbitrary constants equals the number of independent variables is called its complete integral or complete
solution.

For a nonlinear partial differential equation of first order:

f (x, y, z, p, q) = 0, (10.4.1)

the complete integral is of the form

ω(x, y, z, a, b) = 0. (10.4.2)

The complete solution (10.4.2) represents a two-parameter family of surfaces. If the envelope of the system
(10.4.2) exsts, it is also a solution called the singular integral of the equation (10.4.1). Note that an envelope
of the system (10.4.2) touches a member-surface of the system. The singular solution of (10.4.1) is obtained
by eliminating the arbitrary constants a and b from the relations: ω ≡ 0, ∂ω ∂ω
∂a ≡ 0 and ∂b ≡ 0.

Charpit’s Auxiliary Equations

Given below are the Charpit’s auxiliary equations, which are employed to get the complete solution (10.4.2)
of (10.4.1):
dy dp dq
dx
fp = fq = dz
p fp +q fq = −( fx +p fz ) = −( fy +q fz ) · (10.4.3)

In fact, we solve equations (10.4.1) and (10.4.3) for p and q, and then substitute these expressions in the
total differential
∂z ∂z
dz = dx + dy = p dx + q dy. (10.4.4)
∂x ∂x
Integrating (10.4.4), we get the complete integral of the form (10.4.2).

EXERCISE 10.4.1. Find the complete integrals of the following equations:

(a) p2 x + q2 y = z

Department of Mathematics, Vellore Institute of Technology

58 10.5. Special forms of Nonlinear First Order Partial Differential Equations

(b) (p2 + q2 )y = qz

(c) p = (z + qy)2

(d) px 5 − 4q3 x 2 + 6x 2 z − 2 = 0

(e) 2(z + xp + yq) = yp2 2

Ans.
√ √
(1 + a)z = ax + y + b
p
(a)

(b) (x + b)2 + y 2 = az2

(c) z = bx a y 1/a
2
(d) z = 32 (y + a)3/2 + be3/x + 1
3x 2
+ 1
9

a2
(e) z = ax
y2
+ b
y − 4y 3

10.5 Special forms of Nonlinear First Order Partial Differential Equations

TYPE 1: Equations involving only p and q

Consider an equation of the form

f (p, q) = 0. (10.5.1)
For this, the Charpit’s auxiliary equations reduce to
dy dp dq
dx
fp = fq = dz
p fp +q fq = 0 = 0 ·

Note that p = a or q = a is an obvious solution of these relations. Inserting p = a in (10.5.1),

f (a, q) = 0 or q = φ(a). (10.5.2)

Using p = a and (10.5.2) in (10.4.4),

dz = a dx + φ(a) dy,
which on integration gives the complete integral of (10.5.1) as

z = ax + φ(a)y + b. (10.5.3)

Sometimes the substitution q = a reduces computations in obtaining the complete solution.

Non-existence of Singular Integral: Partially differentiating (10.5.3) with respect to b, we het a contradiction
that 0 = 1. Therefore, the envelope of the 2-parameter family of surfaces represented by the complete
integral (10.5.3) does not exist. Thus Type 1 equations do not have singular integrals.

EXERCISE 10.5.1. Find the complete integrals of the following equations:

√ √
(a) p+ q =1

(b) p + q + pq = 0

(c) p2 + q2 = 4

(d) p(p + 1) = q2

Department of Mathematics, Vellore Institute of Technology

 Chapter 10. Partial Differential Equations of First Order 59

(e) q2 − 3q + p = 2

Ans.
√ √
(a) z = ax + (1 − a)2 y + b or z = (1 − a)2 x + ay + b

(b) z = ax + b − ay/(a + 1)
√
(c) z = ax + 4 − a2 y + b

(d) z = ax + a(a + 1)y + b

p

(e) z = (2 + 3a − a2 )x + ay + b

TYPE 2: Equations involving p and q and only one of the variables x, y and z

(a) For an equation involving p, q and x:

f (p, q, x) = 0. (10.5.4)
Charpit’s auxiliary equations become
dy dp dq
dx
fp = fq = dz
p fp +q fq = −( fx +p fz ) = 0 ·

Note that q = a is a solution of these ratios. Inserting q = a in (10.5.4),

f (p, a) = 0 or p = φ(a). (10.5.5)

Using q = a and (10.5.5) in (10.4.4),

dz = φ(a) dx + a dy,

which on integration gives the complete integral of (10.5.7) as

z = φ(a)x + ay + b. (10.5.6)

Non-existence of Singular Integral: Type 2(a) equations do not have singular integrals.

EXERCISE 10.5.2. Find the complete integrals of the following equations:

(a) p2 + px = q
√ √
(b) p + q = x
(c) pq = x

Ans.
2 √
(a) z = − x4 + x
4 x
2 + 4a2 ) + a2 sinh−1 (x/2a) + a2 y + b
(b) z − (x − a)3 /3 = a2 y + b
(c) z − x 2 /2a = ay + b

(b) For an equation involving p, q and y:

f (p, q, y) = 0. (10.5.7)
Charpit’s auxiliary equations become
dy dp dq
dx
fp = fq = dz
p fp +q fq = 0 = −( fy +q fz ) ·

Department of Mathematics, Vellore Institute of Technology

60 10.5. Special forms of Nonlinear First Order Partial Differential Equations

Since p = a is a solution, inserting p = a in (10.5.7),

f (a, q) = 0 or q = φ(a). (10.5.8)

Using p = a and (10.5.8) in (10.4.4),

dz = a dx + φ(a) dy,

which on integration gives the complete integral of (10.5.7) as

z = ax + φ(a)y + b. (10.5.9)

Non-existence of Singular Integral: Type 2(b) equations do not have singular integrals.

EXERCISE 10.5.3. Find the complete integrals of the following equations:

(a) q2 = yp4
√ √
(b) p + q = y
(c) p2 q3 = y

Ans.
2a2 y 3/2
(a) z = ax + 3 +b
(y−a)3
(b) z = a2 x + 3 +b
3y 4/3
(c) z = a3 x + 4y 2
+b

(c) For an equation involving p, q and z:

f (p, q, z) = 0. (10.5.10)
Charpit’s auxiliary equations become
dy dp dq
dx
fp = fq = dz
p fp +q fq = −p fz = −q fz ·

From the last two ratios, we see that p = aq, Inserting this in (10.5.10),

f (aq, q, z) = 0 or q = φ(z, a). (10.5.11)

Using p = aq and (10.5.11) in (10.4.4),

dz = q(a dx + dy) or {φ(z, a)}−1 dz = a dx + dy

which on integration gives the complete integral of (10.5.10):

Φ(z, a) = (ax + y) + b. (10.5.12)

We may employ the substitution q = ap also, and obtain the complete integral.

Non-existence of Singular Integral: Type 2(c) equations do not have singular integrals.

EXERCISE 10.5.4. Find the complete integrals of the following equations:

(a) z = p2 + q2

(b) p3 = qz

Department of Mathematics, Vellore Institute of Technology

 Chapter 10. Partial Differential Equations of First Order 61

(c) z = p2 + q2 + 1
(d) z(p2 + q2 + 1) = 1
(e) 4(z 3 + 1) = 9z 4 pq
(f) q2 = p2 z 2 (1 − p2 )
(g) p2 z2 + q2 = 1
(h) 9(p2 z + q2 ) = 4
Ans.
(a) 4(a2 + 1)z = (x + ay + b)2
√ √
(b) 2 z = a(x + ay + b)
√
(c) a2 + 1 cosh−1 z = x + ay + b
(d) (a2 + 1)(1 − z 2 ) = (x + ay + b)2
(e) a2 (z 3 + 1) = (x + a2 y + b)2
(f) z = a2 + (x + a2 y + b)2
√
(g) 2z z 2 + a2 ) + a2 sinh−1 (z/a) = x + ay + b
(h) (z + a)3/2 = x + ay + b

TYPE 3: Separable Equations of the form

Consider an equation of the form

f (p, x) = g(q, y). (10.5.13)
For this, the Charpit’s auxiliary equations reduce to
dy dp dq
dx
fp = −gq = dz
p fp −qgq = − fx = −gy ·

dp fx
From these, we have an ordinary differential equation: dx + fp = 0, which can be written as

fp dp + fx dx = 0 or d f (p, x) = 0.

Integrating this total differential, we get the solution

f (p, x) = a.

Using this in (10.5.13),

f (p, x) = a, g(q, y) = a.
Solving these for p and q,
p = µ(x, a), q = ν(y, a). (10.5.14)
Using (10.5.14) in (10.4.4),
dz = µ(x, a) dx + ν(y, a) dy,
which on integration gives the complete integral of (10.5.13) as
∫ ∫
z= µ(x, a) dx + ν(y, a) dy + b. (10.5.15)

Department of Mathematics, Vellore Institute of Technology

62 10.5. Special forms of Nonlinear First Order Partial Differential Equations

Non-existence of Singular Integral: Type 3 equations also do not have singular integrals.

EXERCISE 10.5.5. Find the complete integrals of the following equations:

(a) px 2 = qy 2

(b) pq + qx = y
p2 q2
(c) x2
− y2
=1

(d) p2 − q2 = x − y
√ √
(e) p + q = x + y

(f) (p + q)x + pq = 0

(g) p2 = q/x y

(h) px − y 2 q2 = 1

Ans.

(a) xyz + a(x + y) = bxy

y 2
(b) z = ax − x 2 + 2a +b
3 2 √
(c) z = x 3a + 2 a2 − 1y + b

(d) 3z = 2(x + a)3/2 + 2(y + a)3/2 + κ

(e) 3z = (x + a)3 + (y − a)3 + κ

(f) 2a(a + 1)z = −(a + 1)x 2 + ay 2 + b

√
(g) 2z = (4 x + y 2 )a2 + b

(h) z = (a2 + 1) log x + a log y + b

EXERCISE 10.5.6. Find the general and complete integrals of the following equations:

(a) px − qy = y 2 − x 2

(b) p + q = sin x + sin y

(c) px 2 − 2y 3 q = 1

(d) 2p − 3q = z

(e) p + q = 1

Ans.
dy
x = −y = y 2 −x 2 . Grouping the first two ratios, we get x y = a. Choosing
dz
(a) Auxiliary equations are dx
(x, y, 1) as multipliers, the second solution is x 2 + y 2 + 2z = b. Therefore, the general integral is
f xy, x 2 + y 2 + 2z = 0. Also, the complete integral is x 2 + y 2 + 2z = 2a log x y + b.

(b) The general and complete integrals are f (z−cos x−cos y, x−y) = 0 and z = a(x−y)−(cos x+cos y)+b
respectively.

Department of Mathematics, Vellore Institute of Technology

 Chapter 10. Partial Differential Equations of First Order 63

1
+ 1
,z + 1
= 0 and z = − ax + a−1
+ b respectively.


(c) The general and complete integrals are f x 4y 2 x 4y 2
√

(d) The general and complete integrals are f 2x + 3y, x − log z = 0 and (2 − 3a) log z = 4(ax + y) + b
respectively.

(e) The general and complete integrals are f x − y, y − z = 0 and z = ax + (1 − a)y + b respectively.

Department of Mathematics, Vellore Institute of Technology

64 10.5. Special forms of Nonlinear First Order Partial Differential Equations

TYPE 4: Clairaut’s equation

Consider an equation of the form

z = px + qy + f (p, q). (10.5.16)
For this, the Charpit’s auxiliary equations reduce to
dy dp dq
dx
x+ fp = y+ fq = dz
px+qy+p fp +q fq = 0 = 0 ·

From the last two ratios, obviously p = a and q = b are solutions. Using these in (10.4.4) and then inte-
grating, the complete integral of (10.5.16) is

z = ax + by + f (a, b). (10.5.17)

Singular Integral: Partially differentiating (10.5.17) with respect to a and b,

0 = x + fa and 0 = y + fb . (10.5.18)

Eliminating a and b from (10.5.17) and (10.5.18), we obtain the singular solution of (10.5.16).

EXERCISE 10.5.7. Find the complete and singular integrals of the following equations:

(a) z = px + qy + p2 q2
√
(b) z = px + qy − 2 pq

(c) z = px + qy + pq

(d) q(z − px − qy) = p(1 − q)

(e) z = px + qy + p2 − q2

Ans.

(a) The complete integral is z = ax + by + a2 b2 ; the singular integral is 16z 3 + 27x 2 y 2 = 0

√
(b) The complete integral is z = ax + by − 2 ab; the singular integral is x y = 1

(c) The complete integral is z = ax + by + ab; the singular integral is z = x y

(d) The complete integral is z = ax + by + a

b − a; the singular integral is (1 − x) = y

(e) The complete integral is z = ax + by + a2 − b2 ; the singular integral is 4z = y 2 − x 2

EXERCISE 10.5.8. Find the complete, singular and general integrals of (1 − x)p + (2 − y)q = 3 − z.
Ans.
The complete integral is z = ax + by + (3 − a − 2b); the singular integral is z = 3; the general integral is
f y−2 , z−3

x−1 y−2 = 0

Department of Mathematics, Vellore Institute of Technology

 Chapter 10. Partial Differential Equations of First Order 65

Table 10.1: Reference Table

Type Description Finding the Complete and Singular Integrals

Substitute p = a or q = a. Write p = a in the given
Equations having equation and solve it for q, say q = φ(a). Then the total
Type I: f (p, q) = 0
only p and q differential dz = a dx + φ(a) dy on integration gives the
complete integral z = ax + φ(a)y + b.
Substitute q = a in the given equation and solve it for
Type II(a): Equations with p,
p, say p = φ(a). Then dz = φ(a) dx + a dy gives the
f (p, q, x) = 0 q and x
complete integral z = φ(a)x + ay + b.
Write p = a in the given equation and solve it for q, say
Type II(b): Equations having
q = φ(a). Then dz = a dx + φ(a) dy gives the complete
f (p, q, y) = 0 p, q and y
integral z = ax + φ(a)y + b.
Insert p = aq in the given equation, and then solve it for
Type II(c): Equations with p, q and z, say q = φ(z, a). With these substitutions, the
f (p, q, z) = 0 q and z total differential {φ(z, a)}−1 dz = a dx +dy is integrated
to get the complete integral Φ(z, a) = (ax + y) + b.
Equate each side to a constant a and solve for p and q, say
Separables form p = µ(x, a), q = ν(y, a). Use these in the total differential
containing p and x
Type III
on one side, and q dz = µ(x, a) dx + ν(y, a) dy,
f (p, x) = g(q, y)
and y on the other
side which
∫ on integration
∫ then gives the complete integral
z = µ(x, a) dx + ν(y, a) dy + b.
Substituting p = a and q = b in the given equation, its
complete integral is

Type IV σ = z − ax − by − f (a, b) = 0.
Clairaut’s equation
z = px+qy+ f (p, q)
Elimination of the arbitrary constants a and b from the
relations σ ≡ 0, ∂σ/∂a ≡ 0 and ∂σ/∂b ≡ 0 results in
the singular solution.

Remark 10.5.1. The singular solutions do not exist for the equations of Types I, II and III.

Department of Mathematics, Vellore Institute of Technology

66 10.5. Special forms of Nonlinear First Order Partial Differential Equations

Department of Mathematics, Vellore Institute of Technology

Chapter 11

Linear Partial Differential Equations of Higher Order

11.1 Linear Equations of Second Order

Second order linear partial differential equations are frequently encountered in the applications of mathe-
matics. They are classified into three types: hyperbolic, parabolic and elliptic. Each type describes a different
physical phenomenon. Hyperbolic equations describe wave phenomena, parabolic equations describe dif-
fusion processes, and elliptic equations describe equilibrium conditions.

A linear partial differential equation of 2nd order of the form

∂ z
2 2∂ z ∂ z
2 ∂z ∂z
a ∂x 2 + b ∂x∂y + c ∂y 2 + d ∂x + e ∂y + f z = 0 (11.1.1)

is called hyperbolic if b2 − 4ac > 0, parabolic if b2 − 4ac = 0, elliptic if b2 − 4ac < 0.

EXERCISE 11.1.1. Classify the following equations as hyperbolic, parabolic or elliptic:

(a) zxx + 6zxy + 12zyy = 0

Ans. Elliptic

(b) zxx + 20zxy + 64zyy = 0

Ans. Hyperbolic

(c) zxx + 2zxy + zyy = 0

Ans. Parabolic

(d) 6zxx − zxy − zyy = 0

Ans. Hyperbolic

(e) zxx + 4zxy + 5zyy = 0

Ans. Elliptic

(f) zxx + 2zxy + 5zyy = 0

Ans. Elliptic

(g) zxx − ztt = 0

Ans. Hyperbolic

(h) zxx + 6zxy + 9zyy = 0

Ans. Parabolic

67
68 11.2. Linear Equations of Higher Order

11.2 Linear Equations of Higher Order

The general form of a linear partial differential equation of nth order with constant coefficients is
∂n z
+ a1 ∂x∂n−1z∂y + a2 ∂x n−
∂ z ∂ z ∂ z
n n n n
∂x n 2 ∂y 2 + · · · + an−1 ∂x∂y n−1 + an ∂y n = Q(x, y), (11.2.1)

where all ai0s are constants. If Q(x, y) = 0, (11.2.1) is homogeneous, otherwise non-homogeneous. Introduc-
ing the operators notation:
∂r ∂s ∂n
∂x r = Dr , ∂y s = D 0s , ∂x r ∂y s = Dr D 0s , 1 ≤ r, s ≤ n, r + s = n, (11.2.2)

then (11.2.1) is written as

 
D n + a1 D n−1 D 0 + a2 D n−2 D 02 + · · · + an−1 DD 0n−1 + an D 0n z = Q(x, y) or f (D, D 0)z = Q(x, y), (11.2.3)

where D n + a1 D n−1 D 0 + a2 D n−2 D 02 + · · · + an−1 DD 0n−1 + an D 0n = f (D, D 0).

The general solution of (11.2.1) or (11.2.3) consists of two parts:

(a) The Complementary Function (zc ): The general solution of the homogeneous part f (D, D 0)z = 0, with
n arbitrary functions;
n o
(b) The Particular Integral z p = f (D,D
1
0 ) Q(x, y), without arbitrary constants and arbitrary functions.

The general solutions is given by z = zc + z p .

11.3 Finding the Complementary Function zc

Replace D with m and D 0 with 1 in f (D, D 0), we get the auxiliary equation f (m, 1) = 0 of (11.2.3).

(a) Let m1 , m2 , ..., mn be distinct roots of the auxiliary equation, then

zc = φ1 (y + m1 x) + φ2 (y + m2 x) + · · · + φn (y + mn x).

(b) Let m1 = m2 = κ be a double root of the auxiliary equation, then the corresponding part of zc is
φ1 (y + κx) + xφ2 (y + κx). In general, if all the roots are coincident, say m1 = m2 = · · · = mn κ, then

zc = φ1 (y + κx) + xφ2 (y + κx) + · · · + x n−1 φn (y + κx).

Remark 11.3.1. The roots of the auxiliary equation are real or complex. Complex roots occur as conju-
gate pair α ± iβ.
Remark 11.3.2. The part of zc corresponding to the root z = a/b may be written as φ(by + ax).
EXERCISE 11.3.1 (Reasoning). What is the motive behind the replacement of D with m and of D 0 with
1 in f (D, D 0) to get the auxiliary equation f (m, 1) = 0?

The general solution of the homogeneous form f (D, D 0)z = 0 is z = zc .

EXERCISE 11.3.2. Find the general solutions of the following linear partial differential equations of
higher order:

(a) (2D2 + 5DD 0 + 2D 02 )z = 0

Department of Mathematics, Vellore Institute of Technology

 Chapter 11. Linear Partial Differential Equations of Higher Order 69

(b) (D2 + 6DD 0 + 9D 02 )z = 0

(c) (25D2 − 40DD 0 + 16D 02 )z = 0

(d) (D2 + DD 0 − 2D 02 )z = 0

(e) (D2 + 9DD 0 + 25D 02 )z = 0

(f) (D3 − 4D2 D 0 + 4DD 02 )z = 0

(g) (D4 − D 04 )z = 0

(h) (D4 − 2D3 D 0 + 2DD 03 − D 04 )z = 0

Ans.

(a) z = φ1 (2y − x) + φ2 (y − 2x)

(b) z = φ1 (y − 3x) + xφ2 (y − 3x)

(c) z = φ1 (5y + x) + xφ2 (5y + x)

(d) z = φ1 (y + 2x) + φ2 (y − 2x)

(e) z = φ1 [2y + (8 − 3i)x] + φ2 [2y − (8 + 3i)x]

(f) z = φ1 (y) + φ2 (y + 2x) + xφ3 (y + 2x)

(g) z = φ1 (y + x) + φ2 (y − x) + φ3 (y + ix) + φ4 (y − ix)

(h) z = φ1 (y + x) + xφ2 (y + x) + x 2 φ3 (y + x) + φ4 (y − x)

11.4 Finding the Particular Integral z p

Case (a): Q(x, y) = e ax+by

zp = 1
f (a,b) e
ax+by
, provided f (a, b) , 0.

Case (b): Q(x, y) = sin(ax + by) or cos(ax + by)

Replace D2 with −a2 , D 02 with −b2 and DD 0 with −ab in f (D, D 0) to get φ(a2, −ab, −b2 ) , 0. Then
sin(ax+by) cos(ax+by)
zp = φ(a2 ,−ab,−b 2 )
or φ(a2 ,−ab,−b 2 )

respectively.

Failure Case: If f (D, D 0) = 0 with the relevant substitutions in Case (a) and Case (b), then differentiate
h with ∂ f /∂Diover Q(x, y) using Case (a) or Case (b). Then
f (D, D 0) partially with respect to D and operate
pre-multiply the quantity by x. Thus z p = x { ∂ f /∂D
1
}Q(x, y) .

This technique is employed until the denominator quantity after the substitutions becomes non-zero.

Particular Integral associated with a Linear Factor D − mD 0:

∫
z p = D−mD0 Q(x, y) =
1
Q(x, c − mx) dx, where y = c − mx. (11.4.1)

Department of Mathematics, Vellore Institute of Technology

70 11.4. Finding the Particular Integral z p

After integration, again substitute c = y + mx.

Case (c): Q(x, y) = x r y s

(a) If r < s, expand [ f (D, D 0)]−1 in ascending powers of D/D 0 and operate term-wise on x r y s ;

(b) If r > s, expand [ f (D, D 0)]−1 in ascending powers of D 0/D and operate term-wise on x r y s ;

(c) If r = s, employ (11.4.1) with each linear factor.

EXERCISE 11.4.1. Find the general solutions of the following equations:

(a) (D2 − 5DD 0 + 6D 02 )z = e x+y

(b) (D3 − 3D2 D 0 + 4D 03 )z = e x+2y

(c) (4D2 + 12DD 0 + 9D 02 )z = e3x−2y

(d) (D3 − 4D2 D 0 + 5DD 02 − 2D 03 )z = e2x+y

(e) (D2 − 2DD 0 + 2D 02 )z = sin x

(f) (D2 − α2 D 02 )z = E sin px

(g) (D2 + DD 0 − 6D 02 )z = cos(2x + y)

(h) (D3 − 4D2 D 0 + 4DD 02 )z = 2 sin(3x + 2y) + 6 sin(3x + 2y)

(i) (D2 − DD 0)z = sin x cos y

(j) (D2 − D 02 )z = cos 2x cos 3y

(k) (D2 − DD 0)z = cos x cos 2y

(l) (D2 − 2DD 0 + D 02 )z = cos(x − 3y)

(m) (D2 − 3DD 0 + D 02 )z = sin x sin y

(n) (D3 + D2 D 0 − DD 02 − D 03 )z = 3 sin(x + 2y)

(o) (D3 − 2D2 D 0)z = 2e2x + 3x 2 y

(p) (D2 − DD 0 − 2D 02 )z = (y − 1)e x

(q) (D2 + DD 0 − 6D 02 )z = y cos x

(r) (D2 + 3DD 0 + 2D 02 )z = x + y

(s) (D3 − 7DD 02 − 6D 02 )z = sin(x + 2y) + x 2 y

Ans.

(a) z = µ1 (y + 2x) + µ2 (y + 3x) + e x+y /2

(b) z = φ1 (y − x) + φ2 (y + 2x) + xφ3 (y + 2x) − e x+2y /27

(c) z = φ1 (2y − 3x) + xφ2 (2y − 3x) + x 2 e3x−2y /8

(d) z = φ1 (y + x) + xφ2 (y + x) + φ2 (y + 3x) + xe2x+y

Department of Mathematics, Vellore Institute of Technology

 Chapter 11. Linear Partial Differential Equations of Higher Order 71

(e) z = φ1 (y + x) + xφ2 (y + x) − sin x

(f) z = φ1 (y + αx) + φ2 (y − αx) − E sin px/p2

(g) z = φ1 (y + 2x) + φ2 (y − 3x) + x sin(2x + y)/5

(h) z = f1 (y) + f2 (y + 2x) + x f3 (y + 2x) + 23 sin(3x + 2y) + 2

45 cos(3x + 2y)

(i) z = f1 (y) + f2 (y + 2x) + 16 sin(x + 2y) − 1

10 sin(x − 2y)

(j) z = f1 (y + x) + f2 (y − x) − 51 cos 2x cos 3y

(k) z = f1 (y) + f2 (y + x) − 21 cos(x + 2y) − 16 cos(x − 2y)

(l) z = f1 (y + x) + x f2 (y + x) − 16
1
cos(x − 3y)
h  √  i h  √  i
(m) z = g1 y + 3+2 5 x + g2 y + 3−2 5 x + 21 sin(x + y) − 1
10 sin(x − y)

(n) z = g1 (y − x) + xg2 (y − x) + g3 (y + x) − 3x sin(x + y)/4

(o) z = g1 (y) + xg2 (y) + g3 (y + 2x) + 41 e2x 20 x y + x 6 /60

1 5

(p) z = ξ1 (y − x) + ξ2 (y + 2x) + ye x

(q) z = ξ1 (y + 2x) + ξ2 (y − 3x) + sin x − y cos x

x2 y x3
(r) z = ξ1 (y − x) + ξ2 (y − 2x) + 2 − 3

x5 y
(s) z = ω1 (y − x) + ω2 (y − 2x) + +ω3 (y + 3x) + 1
75 cos(x + 2y) − 60

Department of Mathematics, Vellore Institute of Technology

72 11.4. Finding the Particular Integral z p

Department of Mathematics, Vellore Institute of Technology

Chapter 12

Applications of Partial Differential Equations

12.1 Vibrating Strings - One dimensional Wave Equation

Consider a thin copper string of length L cm, stretched between the fixed ends x = 0 and x = L. Suppose
that it is released from its initial position f (x) cm by giving a velocity g(x) cm per second.

The subsequent transverse displacement u(x, t) at any point x in vibrating string is governed by the homo-
geneous partial differential equation of second order:
∂2 u
= ν2 ∂∂xu2 ,
2
∂t 2
(12.1.1)

where ν is a fixed constant depending on the physical properties of the string. In fact, ν = T/ρ, where T
p

is the constant horizontal component of the tension and ρ is the mass per unit length of the string.
Since both the ends x = 0 and x = L are fixed, there will not be any vertical displacement at the ends at
any time t. Thus the boundary conditions are:

u(0, t) = 0 for all t > 0, (12.1.2)

and u(L, t) = 0 for all t > 0. (12.1.3)

We assume a trial solution of the product form

u(x, t) = X(x)T(t), (12.1.4)

where X(x) = X is a function of x only and T(t) = T is a function of t alone.

Differentiating (12.1.4) partially with respect to x and t,

∂2 u d2 X ∂2 u d2 T
∂x 2
= dx 2
· T = X 00T and ∂t 2
=X· dt 2
= XT 00 .

Substituting these in the wave equation (12.1.1) and then separating the variables, we get
T 00 X 00
XT 00 = ν2 X 00T or ν2 T
= X · (12.1.5)

Since x and t are independent variables, each of the fractions in (12.1.5) reduces to a constant, say λ. This
results in two ordinary differential equations:

X 00 + λX = 0, (12.1.6)
T 00 + λν2T = 0. (12.1.7)

We solve (12.1.6) in three cases:

73
74 12.1. Vibrating Strings - One dimensional Wave Equation

Case (a): λ < 0, say λ = −µ2, where µ , 0

The general solution of (12.1.6), X(x) = Aeµx + Be−µx satisfies the boundary conditions (12.1.2) and (12.1.3)
only if A + B = AeµL + Be−µL = 0, that is only if A = 0 = B. This leads to the trivial solution: X ≡ 0 for
all x ∈ [0, L] of (12.1.6).

Case (b): λ = 0

Then the general solution of (12.1.6) is X(x) = Ax + B, which satisfies the boundary conditions (12.1.2)
and (12.1.3) only if A = 0 = B. This again is the trivial solution X ≡ 0 of (12.1.6).

Case (c): λ > 0, say λ = µ2, where µ , 0

Vibrations in strings are periodic and the general solution of (12.1.6) is given by

X(x) = A cos µx + B sin µx. (12.1.8)

Applying the first boundary condition (12.1.2), we see that X(0) = A = 0.

Thus (12.1.8) reduces to

X(x) = B sin µx.
But then using the second boundary condition (12.1.3), this implies that X(L) = B sin µL = 0. In order to
get a nontrivial solution, we must have B , 0 so that

sin µL = 0 ⇒ µL = nπ or µ = L,
nπ
n = ±1, ±2, . . . . (12.1.9)

Thus all the nontrivial solutions of (12.1.6) are of the form:

Xn (x) = Bn sin nπx n = ±1, ±2, . . . ,

L

where Bn0 s are arbitrary constants depending on n.

   
Since sin − kπx
L = − sin kπx
L for all k, we see that X−k is a constant multiple of Xk for each k. Therefore,
we shall discard the indices n = −1, −2, −3, ... so that the linearly independent solutions are given by

Xn (x) = Bn sin nπx L , n = 1, 2, 3, . . . .

(12.1.10)

Now the general solution of (12.1.7) is T(t) = C cos µνt + D sin µνt. Inserting the values of µ obtained in
(12.1.9) into this, we get

Tn (t) = Cn cos nπνt + Dn sin nπνt , n = 1, 2, 3, . . . ,

L L (12.1.11)

where Cn and Dn are the arbitrary constants, which depend on the index n.

Using (12.1.10) and (12.1.11), the nontrivial solutions of (12.1.1) are given by

un (x, t) = Xn (x)Tn (t) = Bn sin nπx Cn cos nπνt + Dn sin nπνt




L L L

or un (x, t) = sin nπx bn cos nπνt + cn sin nπνt , n = 1, 2, 3, . . ., where bn = Bn Cn and cn = Bn Dn .




L L L

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 75

In view of the superposition principle, the general solution of the wave equation (12.1.1) is given by
∞
Õ
u(x, t) = nπx nπνt
+ cn sin nπνt
,




sin L bn cos L L (12.1.12)
n=1

where 0 < x < L. The constants bn and cn are determined by employing the following initial displacement
and initial velocity conditions:

u(x, 0) = f (x) for 0 < x < L (12.1.13)

and
∂u
∂t (x, 0) = g(x) for 0 < x < L. (12.1.14)

In fact, writing t = 0 in (12.1.12) and then using (12.1.13), we see that

∞
Õ ∞
Õ
f (x) = u(x, 0) = nπx
[bn cos 0 + cn sin 0] = nπx
,



sin L bn sin L
n=1 n=1

which may be regarded as the half-range sine series of f (x) in the interval 0 ≤ x ≤ L. Therefore, bn is the
Fourier coefficient obtained by the formula

∫L
bn = 2 nπx
n = 1, 2, . . . .


L f (x) sin L dx, (12.1.15)
0

On the other hand, differentiating (12.1.12) partially with respect to t term-by-term, we have
∞
∂u
Õ
= nπνt nπx nπνt
+ cn cos nπνt
.




∂t (x, t) L sin L −bn sin L L
n=1

Writing t = 0 in this and then using (12.1.14), we get

∞
Õ
ncn πν

g(x) = nπx
, 0 ≤ x ≤ L.


L sin L
n=1

This is the half-range sine series of g(x) in 0 ≤ x ≤ L so that its Fourier coefficients are found by

∫L ∫L
ncn πν
= 2 nπx
or cn = 2 nπx
dx, n = 1, 2, . . . .



L L g(x) sin L dx, nπν g(x) sin L (12.1.16)
0 0

If the string is placed along the x-axis between the fixed ends, then

u(x, 0) = f (x) = 0 for 0 ≤ x ≤ L

so that bn = 0 for all n. Thus the complete solution (12.1.12) becomes

∞
Õ
u(x, t) = nπx nπνt
,



cn sin L sin L (12.1.17)
n=1

where cn is given by (12.1.16).

Department of Mathematics, Vellore Institute of Technology

76 12.1. Vibrating Strings - One dimensional Wave Equation

∂u(x,0)
If the string is released from rest, its initial velocity is zero. That is, ∂t = g(x) = 0 so that cn = 0 for all
n from (12.1.16). The complete solution (12.1.12) will be
∞
Õ
u(x, t) = nπx nπνt
,



bn sin L cos l (12.1.18)
n=1

where bn is given by (12.1.15).

Example 12.1.1. A thin copper string, stretched along the x-axis is fastened to the fixed
ends x = 0 and
x = l. If it is plucked from the mean position by giving an initial velocity g(x) = sin3 πx
l , find the vertical
displacement u(x, t) of the string.
Solution. The transverse displacement u(x, t) of the string is described by
∞
Õ
u(x, t) = nπx nπνt
,



cn sin l sin l (12.1.19)
n=1

where cn0 s are determined by applying the initial condition:

∂u
∂t (x, 0) = g(x) = sin l = 4 3 sin .
3 πx
1 3πx
 πx



l − sin l (12.1.20)

Now partially differentiating (12.1.19) with respect to t term-wise, we get

∞
∂u
Õ
= nπx nπνt
.
πν



∂t (x, t) l · ncn sin l cos l
n=1

Writing t = 0 in this and then using (12.1.20), we get

∞
Õ
1 3πx
= nπx
.
 πx


 πa


4 3 sin l − sin l l · ncn sin l
n=1

Comparing the coefficients of the like terms both sides, we get

πν
l · c1 = 34 , c2 = 0, 3πν
l · c3 = − 41 c4 = c5 = · · · = 0

or
c1 = 4πν ,
3l
c2 = 0, c3 = − 12πν
l
, c4 = c5 = · · · = 0.
With these values, the complete solution of the boundary value problem can be written as:

u(x, t) = 4πν
l
− sin 3πx sin 3πνt





9 sin πx πνt
l sin l l l ·

Example 12.1.2. A thin metal string of length l cm is tightly stretched along the x-axis between the
fixed ends x = 0 and x = l. If it is disturbed from its mean position by giving an initial velocity g(x) =
k x(l − x), 0 < x < l, determine the vertical displacement of the string at any point x and at any time t.
∞
Solution. We see that u(x, t) = cn sin nπx sin nπνt
Í


l l , where
n=1

∫l ∫l
cn = 2 nπx
dx = 2k
(l x − x 2 ) sin nπx



nπν g(x) sin l nπν l dx
0 0
nπx  l
 ( ) ( ) 
nπx nπx
  
 cos l sin l cos
= nπν
2k 
(l x − x 2 ) − nπ/l − (l − 2x) − n2 π2 /l + (−2) n3 π3 /ll 3 

2
 
x=0

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 77


 lx(l−x) l 2 (l−2x) 2l 3

 l
= 2k nπx
+ nπx nπx



nπν − nπ cos l n 2 π2
sin l − n 3 π3
cos l 
x=0
2l 3 8kl 3
= 2k
nπν · n 3 π3
[cos 0 − cos nπ] = (2m−1)4 π4 ν
for m = 1, 2, 3, ...
   
∞
8kl 3 (2m−1)πx (2m−1)πνt
Thus u(x, t) = 1
Í
π4 ν (2m−1)4
sin l sin l ·
m=1
Example 12.1.3. A tightly stretched string is placed along the x-axis between the fixed ends x = 0 and
x = l. Determine its transverse displacement at any point x and at any time t if the string is disturbed from
its initial position by giving an a velocity
(
k x, 0 ≤ x ≤ 2l
g(x) = (12.1.21)
k(l − x), 2l ≤ x ≤ l

where k is a positive constant.

Solution. The transverse displacement u(x, t) of the string is given by
∞
Õ
u(x, t) = nπx nπνt
,



cn sin l sin l
n=1

where

∫l ∫l/2 ∫l 
 
cn = 2 nπx
dx = 2k nπx
dx + nπx




g(x) sin  x sin (l − x) sin dx 
 
nπν l nπν  l l 
0 0
 l/2 

= 2k
nπν [I1 + I2 ] , say.

Then
∫l/2  ( 
nπx
) ( 
nπx
 )  l/2
 cos l sin l 
I1 = nπx
dx =  x − nπ/l


x sin − (1) − n2 π2 /l2 
 
l  
0 x=0

l2

 l/2 l2 l2
= − nπ + = − 2nπ + ,
 lx nπx
sin nπx nπ nπ




cos l n2 π2 l  cos 2 n2 π2
sin 2
x=0
∫l
I2 = nπx


(l − x) sin l dx,
l/2
 ( ) (  ) l
nπx nπx
 
 cos l sin l 
= (l − x) − nπ/l − (−1) − n2 π2 /l2 
 
 
x=l/2
 2
 l
= −  l(l−x) nπx
+ n2l π2 sin nπx



nπ cos
 
l l  x=l/2

l2 l2
= 2nπ cos nπ 2 + n2 π2 sin 2 ·
nπ

2l 2 4kl 2
Therefore, cn = 2k nπ
= nπ
for n = 1, 2, ....



nπν · n2 π2
sin 2 n3 π3 ν
sin 2

Hence the Fourier solution of the boundary value problem is

∞
Õ
4kl 2
u(x, t) = 1 nπ nπx nπνt
.




π3 a n3
sin 2 sin l sin l
n=1

Department of Mathematics, Vellore Institute of Technology

78 12.1. Vibrating Strings - One dimensional Wave Equation

Example 12.1.4. A tightly stetched string is placed along the x-axis between the fixed ends x = 0 and
x = l. If it is disturbed its initial position by giving an a velocity
(
x, 0 ≤ x ≤ 2
l l
g(x) = l−x
l , 2 ≤ x ≤ l
l

determine its transverse displacement at any point x and at any time t.

Solution. Taking k = 1/l in the previous Example, we get
∞
Õ
u(x, t) = 4l 1 nπ nπx nπνt
.




π3 ν n3
sin 2 sin l sin l
n=1

Example 12.1.5. A tightly
stretched string of length l cm with fixed ends x = 0 and x = l is initially of
the form: f (x) = k sin πx
l , where k > 0. Show that the transverse displacement of any point x from one
end at time t > 0 is

u(x, t) = k sin πx .

πνt


l cos l

Also show that each point of the string executes simple harmonic motion and find its period.
Solution. From the general formulation, we have
∞
Õ
u(x, t) = nπx nπνt
.



bn sin l cos l (12.1.22)
n=1

We evaluate the constants bn0 s by using the initial displacement condition that

u(x, 0) = f (x) = k sin πx for 0 < x < l.

l

Substituting t = 0 in (12.1.22) and then using this, we see that

∞
Õ
= nπx
.
πx



k sin l bn sin l
n=1

Comparing the coefficients of the like terms on both sides of this, we get

b1 = k, b2 = b3 = · · · = 0.

Therefore
u(x, t) = k sin πx

πνt


l cos l

describes the transverse displacement of any point x of the string from one end, say from x = 0 at time
t > 0.

Now partially differentiating two times with respect to t, this gives

∂2 u 2 2
= −k π l2ν sin πx

πνt


∂t 2 l cos l

or
∂2 u πν 2 ∂2 u
=− + ω2 u = 0, ω = l .

πν
∂t 2 l u ⇒ ∂t 2
This shows that each point of the string executes simple harmonic motion and its time period is

T= 2π
ω = 2π
πν/l = ν.
2l

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 79

Example 12.1.6. A tightly stretched string of length l cm with fixed ends x = 0 and x = l is initially
positioned as
u(x, 0) = u0 sin3 πx
l .

If it is released from the rest from this position, find the subsequent transverse displacement u(x, t) of the
string.
Solution. From the general formulation, transverse displace ment u(x, t) of the string is given by
∞
Õ
u(x, t) = nπx nπνt
.



bn sin l cos l (12.1.23)
n=1

We evaluate the constants bn0 s by using the initial displacement condition

u(x, 0) = f (x) = u0 sin3 πx for 0 < x < l.

l

Substituting t = 0 in (12.1.23) and then using this, we see that

∞
Õ
u0 3πx
= sin3 = nπx
.
 πx


 πx



4 3 sin l − sin l l bn sin l
n=1

Comparing the coefficients of the like terms on both sides of this, we get
3u0
b1 = 4 , b2 = 0, b3 = − u40 , b4 = b5 = · · · = 0.

With these values, the transverse displacement at any point x and at any time t > 0 is given by

u(x, t) = u40 3 sin πx − sin 3πx cos 3πνt


πνt




l cos l l l ·

Example 12.1.7. A thin tightly stetched metal string of length l cm is fastened at to the ends x = 0 and
x = l. Vibrations are allowed by disturbing from the position: f (x) = k x(l − x), k > 0. Find the vertical
displacement of the string at any point x and at any time t, if it is released from rest.
Solution. From the general formulation, transverse displacement u(x, t) of the string is
∞
Õ
u(x, t) = nπx nπνt
,



bn sin l cos l
n=1

where
∫l ∫l
bn = 2 nπx
dx = 2k
(l x − x 2 ) sin nπx



l f (x) sin l l l dx
0 0
nπx l
 ( ) ( ) 
nπx nπx
  
 cos l sin l cos l 
= 2k
(l x − x 2 ) − nπ/l − (l − 2x) − n2 π2 /l2 + (−2) n3 π3 /l3 

l  
x=0
 2 3
 l
+ l n(l−2x)
 lx(l−x)
= 2k
− nπ cos nπx sin nπx − n2l3 π3 cos nπx




l l 2 π2 l l  x=0
2l 3 4kl 2
= 2k
l · n3 π3
[cos 0 − cos nπ] = n3 π3
[1 − (−1) ] . n

Thus
∞
Õ
8kl 2
u(x, t) = 1 nπx nπνt



π3 n3
sin l cos l
n=1

describes the subsequent displacement of the vibrating string of the problem.

Department of Mathematics, Vellore Institute of Technology

80 12.2. Heat-Flow with Homogeneous Boundary Conditions

EXERCISE 12.1.1. Solve the following boundary value problems related to vibrating strings:

(a) utt = 4u xx ; u(0, t) = u(π, t) = 0 for t > 0, u(x, 0) = (sin 2x)/10, ut (x, 0) = 0 for 0 < x < π
(b) utt = 4u xx ; u(0, t) = u(π, t) = 0 for t > 0, u(x, 0) = 3
40 sin x − 1
40 sin 3x, ut (x, 0) = 0 for 0 < x < π
(c) utt = 25u xx ; u(0, t) = u(3, t) = 0 for t > 0, u(x, 0) = 1
4 sin πx, ut (x, 0) = 10 sin 2πx for 0 < x < 3
(d) utt = 100u xx ; u(0, t) = u(1, t) = 0 for t > 0, u(x, 0) = 0, ut (x, 0) = x for 0 < x < 1
(e) utt = 4u xx ; u(0, t) = u(π, t) = 0 for t > 0, u(x, 0) = sin x, ut (x, 0) = 1 for 0 < x < π
Ans.
(a) u(x, t) = 1
10 cos 4t sin 2x
(b) u(x, t) = 3
40 cos 2t sin x − 1
40 cos 6t sin 3x
(c) u(x, t) = 1
5 cos 5πt sin πx − 1π sin 10πt sin 2πx
(−1) n+1
(d) u(x, t) =
Í∞
sin 10nπt sin nπx
n=1 5π2 n2

(e) u(x, t) = cos2t sin x + 4π n odd n1 sin 2nπt sin nx

Í

12.2 Heat-Flow with Homogeneous Boundary Conditions

Consider a metal rod of length l cm, heated to a temperature f (x) which is then placed along the x-axis
with one end at x = 0. Suppose that it is laterally thermally well insulated. We may assume that the rod is
very thin as if the heat flow can take place along only one direction, namely along the x-axis.

The temperature distribution in the rod is determined by the one-dimensional heat equation:
∂u
= a2 ∂∂xu2 ,
2
∂t (12.2.1)

where a2 is a physical constant, called the heat constant or diffusivity of the metal with which the rod is made.

Given that the initial temperature is

u(x, 0) = f (x) for 0 < x < l, (12.2.2)

we assume a solution of (12.2.1) of the form

u(x, t) ≡ XT = X(x)T(t).

Substituting this expression into (12.2.1),

XT 0 = a2 X 00T .

Now separating the variables by dividing by XT, we get

T0 X 00
a2T
= X = −λ, say.

Since x and t are independent variables, λ must be a constant, and we have two ordinary differential equa-
tions:

X 00 + λX = 0, (12.2.3)

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 81

T 0 + λa2T = 0. (12.2.4)

We shall discuss the zero-boundary conditions in the following different situations:

(a) Both the ends are at zero temperature

(b) Both the ends are thermally insulated

(c) One end is at zero temperature and the other end thermally insulated

Case (a) - The ends x = 0 and x = l are suddenly cooled to zero temperature and maintained thereat.

The boundary conditions are:

u(0, t) = 0, (12.2.5)
and u(l, t) = 0. (12.2.6)

First, suppose that

λ < 0, say λ = −µ2, µ , 0.
The general solution of (12.2.3) will then be given by

X(x) = Aeµx + Be−µx .

This satisfies the boundary conditions (12.2.5) and (12.2.6) only if A = 0 = B. This leads to the trivial
solution : X ≡ X(x) = 0 for all x ∈ [0, l] of (12.2.3).

Similarly if λ = 0, the general solution of (12.2.3) is X(x) = Ax + B, which satisfies the boundary conditions
(12.2.5) and (12.2.6) only if A = 0 = B. This again is the trivial solution X ≡ 0 of (12.2.3).

Therefore, we expore for the nontrivial solutions when

λ > 0, say λ = µ2, µ , 0.

The general solution of (12.2.3) will be given by

X(x) = A cos µx + B sin µx. (12.2.7)

Applying the boundary condition of (12.2.5), we see that X(0) = A = 0.

Thus (12.2.7) reduces to
X(x) = B sin µx.
But then using the boundary condition of (12.2.6), this implies that

X(l) = B sin µl = 0.

In order to get a nontrivial solution, we must have B , 0 so that

sin µl = 0 ⇒ µl = nπ or µ = l ,
nπ
n = ±1, ±2, . . . . (12.2.8)

Thus all the nontrivial solutions of (12.2.3) are of the form:

Xn (x) = Bn sin nπx n = ±1, ±2, . . . ,

l

Department of Mathematics, Vellore Institute of Technology

82 12.2. Heat-Flow with Homogeneous Boundary Conditions

where Bn0 s are arbitrary constants depending on n.

Since    
sin − kπx
l = − sin kπx
l for all k = 1, 2, 3, ...,

we see that X−k is a constant multiple of Xk for each k.

Therefore, we shall discard the indices n = −1, −2, −3, ... so that the linearly independent solutions of
(12.2.3) are given by
Xn (x) = Bn sin nπx , n = 1, 2, 3, . . . .


l (12.2.9)
Now the general solution of (12.2.4) is
2 a2 t
T(t) = Ae−µ .

Inserting the values of µ obtained in (12.2.8), into this, we get

2 π2 a 2 t/l 2
Tn (t) = Cn e−n , n = 1, 2, 3, . . . , (12.2.10)

where Cn0 s are the arbitrary constants, which depend on the index n.

Using (12.2.9) and (12.2.10), the nontrivial solutions of (12.2.1) are given by
2 π2 a 2 t/l 2
u( x, t) = Xn (x)Tn (t) = Bn Cn sin nπx
e−n


l

or
2 π2 a 2 t/l 2
u( x, t) = bn sin nπx
e−n , n = 1, 2, 3, . . . ,


l

where bn = Bn Cn . By the superposition principle, the general solution of the wave equation (12.2.1) wll be
of the form
∞
Õ 2 π2 a 2 t/l 2
u(x, t) = nπx
e−n .


bn sin l (12.2.11)
n=1

Finally the constants bn are determined through the initial condition (12.2.2):

In fact, writing t = 0 in (12.2.11) and then using (12.2.2), we see that

∞
Õ
f (x) = u(x, 0) = nπx
,


bn sin l
n=1

which may be regarded as the half-range sine series of f (x) in the interval 0 ≤ x ≤ l. Therefore the bn is the
Fourier coefficient given by
∫ l
bn = 2 nπx
n = 1, 2, . . . .


l f (x) sin l dx, (12.2.12)
0

Equation (12.2.11) reveals that the heat-flow is is transient, that is, the temperature u decreases with in-
crease of time t.

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 83

Case (b) - The ends x = 0 and x = l are thermally insulated.

∂u
There will be no heat-flow through the cross sections where the temperature gradient ∂x will be zero.

Thus we impose two boundary conditions

∂u
∂x (0, t) = 0, (12.2.13)
∂u
and ∂x (l, t) = 0. (12.2.14)

As in the previous case, here also λ < 0 leads to the trivial solution X ≡ X(x) = 0 for all x ∈ [0, l] of (12.2.3).

If λ = 0, the general solution of (12.2.3) will be given by

X(x) = Ax + B. (12.2.15)

Applying the first boundary condition (12.2.13), we see that X(0) = A = 0,

Thus (12.2.15) reduces to X(x) = B, which obviously satisfies other boundary condition (12.2.14). In order
to get a nontrivial solution, we must have B , 0.

With λ = 0, (12.2.4) becomes T 0 = 0 whose solution is T(t) = C.

In this way, we get a nontrivial solution of the heat equation (12.2.1) as

u(x, t) = X(x)T(t) = BC , 0. (12.2.16)

We may write BC = a0
2 which can be determined with the help of the initial condition (12.2.2).

We now consider
λ > 0, say λ = µ2, µ , 0.

The general solution of (12.2.3) will be given by

X(x) = A cos(µx) + B sin(µx),

and the general solution of (12.2.4) will be given by

2 a2 t
T(t) = Ce−µ .

Therefore the general solution of (12.2.1) is given by

2 a2 t
u(x, t) = [A sin(µx) + B cos(µx)] Ce−µ . (12.2.17)

Differentiating (12.2.17) partially with respect to x,

∂u 2 a2 t
∂x (x, t) = −Aµ [sin(µx) + B cos(µx)] Ce−µ .

Department of Mathematics, Vellore Institute of Technology

84 12.2. Heat-Flow with Homogeneous Boundary Conditions

Writing x = 0 in this and applying the boundary condition (12.2.13),

2 a2 t
Bµe−µ = 0 for all t ⇒ B = 0.

With this value, (12.2.17) reduces to

2 a2 t
u(x, t) = AC cos(µx)e−µ .

But then differentiating this partially with respect to x,

∂u 2 a2 t
∂x (x, t) = −ACµ sin(µx)e−µ .

Taking x = l in this and using the other boundary condition (12.2.14), this implies
2 a2 t
−ACµ sin(µl)e−µ = 0 for all t

Since AC , 0 and µ , 0, this implies

sin(µl) = 0 ⇒ µl = nπ or µ = l ,
nπ
n = ±1, ±2, . . . . (12.2.18)

Thus the nontrivial solutions (12.2.17) are given by

2 π2 a 2 t/l 2
un (x, t) = an cos nπx
e−n n = ±1, ±2, . . . ,


l

where an0 s are arbitrary constants depending on n.

   
Since cos − kπx l = cos kπx
l for all k = 1, 2, 3, ..., we find that u−n (x, t) is a constant multiple of un (x, t), n =
1, 2, . . . .

Therefore the general solution of (12.2.1) is determined from the expression:

∞
Õ 2 π2 a 2 t/l 2
u(x, t) = a0
+ nπx
e−n , 0 ≤ x ≤ l.


2 an cos l (12.2.19)
n=1

Finally the constants an0 s are obtained through the initial condition (12.2.2):

In fact, writing t = 0 in (12.2.19) and then using (12.2.2), we see that

∞
Õ
f (x) = a0
+ nπx


2 an cos l
n=1

which may be regarded as the half-range cosine series of f (x) in 0 ≤ x ≤ l.

Therefore, the Fourier coefficients are given by

∫ l
a0 = 2
l f (x) dx (12.2.20)
0

and
∫ l
an = 2 nπx
n = 1, 2, . . . .


l f (x) cos l dx, (12.2.21)
0

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 85

Inserting (12.2.20) and (12.2.21) into (12.2.19), we can write

u(x, t) = us (x) + ut (x, t), (12.2.22)

where
∫ l
us (x) = 1
l f (x) dx
0

and
∞  ∫
Õ l 
2 π2 a 2 t/l 2
ut (x, t) = 2 nπx nπx
e−n .



l f (x) cos l dx cos l
n=1 0

The first term us (x) in (12.2.22) is independent of the time t and ∂u

∂t = 0. This is known as the steady
state temperature. While the second term ut (x) in (12.2.22) is called the transient solution, that is the
temperature ut decreases with increase of time t.

Case (c) - The end x = 0 is at zero temperature and the end x = l is thermally insulated.

Then the boundary conditions are

∂u
u(0, t) = 0 = ∂x (l, t)

and the general solution will be given by

∞  1


1 2 π2 a 2 t


Õ n− 2 πx − n− 2 · 2
u(x, t) = bn sin l e l .
n=1

Example 12.2.1. A thin metal rod length l = 20 cm with insulated sides is initially at a constant tem-
perature θ 0 . Its ends are suddenly cooled to 0◦C and kept at that temperature throughout. Describe the
temperature distribution in it.
Solution. From Case (a), the temperature at any point x and at any time t is
∞
Õ 2 π2 a 2 t/l 2
u(x, t) = nπx
e−n ,


bn sin l
n=1

∫ l ∫ 20
2×θ0
bn = 2
f (x) sin nπx = nπx



l l dx 20 sin 20 dx
0 0
 cos  nπx   20
 
θ0  20 
= 10 − nπ/20  = 2θ
nπ [1 − cos(nπ)] =
0 2θ0
nπ [1 − (−1)n ] .
 
x=0

∞
2θ0 1−(−1) n 2 π2 a 2 t/l 2
Therefore, u(x, t) = nπx
e−n .
Í

π n sin l
n=1
Example 12.2.2.
A laterally insulated thin metal rod of length l = 25 cm is initially at a temperature
f (x) = sin 50 cos 50 . Its ends are suddenly cooled to 0◦C and kept at that temperature throughout.
πx πx

Find the temperature u(x, t) in it at any time t > 0.

Department of Mathematics, Vellore Institute of Technology

86 12.2. Heat-Flow with Homogeneous Boundary Conditions

Solution. From Case (a), the temperature at any point x and at any time t is
∞
Õ 2 π2 a 2 t/625
u(x, t) = nπx
e−n .


bn sin 25
n=1

Writing t = 0 in this and using the initial condition

u(x, 0) = sin πx = 1
,

πx

πx


50 cos 50 2 sin 25

we get
∞
Õ
1
= nπx
.
πx



2 sin 25 bn sin 25
n=1

Comparing the coefficients of like powers both sides, b1 = 1/2, bn = 0 for n ≥ 2. Hence,u(x, t) =
1 πx

−π2 ν2 t/625
2 sin 25 e .
Example 12.2.3. A homogeneous copper rod of length l = 100 cm with diffusivity a = 1 with insulated
sides has its ends maintained at 0◦C and is initially at a temperature
(
x, 0 ≤ x ≤ 50
f (x) =
100 − x, 50 ≤ x ≤ 100.

Determine the temperature distribution in the rod.

Solution. The temperature at any point x and at time t > 0 is
∞ 
nπ 2

− 100 t
Õ
u(x, t) = nπx
,


bn sin 100 e
n=1

where
∫ 100 ∫ 50 ∫ 100 
bn = 2 nπx
dx = 1 nπx
dx + nπx




100 f (x) sin 100 50 x sin 100 (100 − x) sin 100 dx
0 0 50
= 1
50 [I1 + I2 ] , say.

Now
 ( ) (  )  50
nπx nπx
 
∫ 50  cos sin 
100 100
I1 = x sin 100 dx =  x − nπ/100
nπx


− (1) − n2 π2 /10000
 

0  
x=0

100 2

50

2
= − nπ cos 100 + nπ sin 100  = − nπ cos 2 + 100 ,
 100x nπx


nπx  100×50 nπ

nπ


nπ sin 2
x=0

 ( ) (  )  100
nπx nπx
 
∫ 100  cos 100 sin 100 
I2 = (100 − x) sin 100 dx = (100 − x) − nπ/100 − (−1) − (nπ/l)2 
nπx


50  
x=50

 100(100−x) 100 2

100

100 2
= − cos 100 + nπ sin 100 
nπx nπx 
= nπ cos 2 + nπ sin nπ
100×50 nπ






nπ 2 ·x=50

Therefore, (
h
100 2

i 0, if n is even
bn = 1 nπ
=


· 2 sin
50 nπ 2 400
sin nπ , if n is odd.


n2 π2 2
nπ 2
 
∞ − 100 t
u(x, t) = 400 1 nπ nπx
.
Í


⇒ π2 n2
sin 2 · sin 100 e
n=1 (n is odd)

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 87

Example 12.2.4. A homogeneous copper rod of length l cm with diffusivity a has insulated sides. Its ends
x = 0 and x = l are thermally insulated. If its initial temperature is f (x) = k x, 0 ≤ x ≤ l, where k is a
positive constant, then find the temperature u(x, t) at any time t inside the rod.
Solution. From Case (b),
∞
Õ 2 π2 a 2 t/l 2
u(x, t) = a0
+ nπx
e−n , 0 ≤ x ≤ l,


2 an cos l
n=1

where
∫ l ∫ l  l
x2 
a0 = 2
f (x) dx = 2k
x dx =  2kl · = kl,

l l 2  x=0
0 0

∫ l
an = 2 nπx


l f (x) cos l dx
0
∫ l
= 2k nπx


l x cos l dx
0
 (  ) (  ) l
 sin nπx  cos

nπx 
l l
= 2k
x − (1) −
 
l  nπ/l n2 π2 /l 2 

x=0
  2  l
=  sin nπx + l cos nπx 
2k
 lx


l  nπ l nπ l 
  2  x=0   2 
2
= 2k
l nπ sin(nπ) + nπ
l l
cos(nπ) − 0 + nπ cos(0) l

= 2k
l [(−1)n − 1] , n = 1, 2, ...
2
nπν

∞
(−1) n −1 − t
u(x, t) = kl
+ 2kl nπx
.
Í

⇒ cos e l
2 π2 n2 l
n=1
Example 12.2.5. A homogeneous copper rod of length l cm has perfectly insulated sides. Its ends x = 0
and x = l are also thermally insulated. If its initial temperature is

f (x) = x(l − x), 0 ≤ x ≤ l,

where k is a positive constant, then find the temperature u(x, t) at any time t inside the rod.
Solution. From Case (b),
∞
Õ 2 π2 a 2 t/l 2
u(x, t) = a0
+ nπx
e−n , 0 ≤ x ≤ l,


2 an cos l
n=1

where
∫ l ∫ l
a0 = 2
l f (x) dx = 2l (l x − x 2 ) dx
0 0
 3 l
 h i l
2  lx 2 l3 l3 l3 l2
= l  2 − x3  = 2
l 2 − 3 = 2
l · 6 = 3,
x=0 x=0

∫ l ∫ l
an = 2 nπx
dx = 2
(l x − x 2 ) cos nπx



l f (x) cos l l l dx
0 0

Department of Mathematics, Vellore Institute of Technology

88 12.3. Heat-Flow with nonhomogeneous Boundary Conditions

 (  nπx  ) ( ) (  ) l
nπx nπx
 
 sin l cos l sin l 
= 2l (l x − x 2 ) − (l − 2x) − + (−2) −
 
nπ/l n 2 π 2 /l 2 n 3 π3 /l 3 
 
x=0
hn 2 3
o n 2
oi
= 2l 0 − l · n2l π2 cos(nπ) + n2l3 π3 sin(nπ) − 0 + l · n2l π2 cos 0 + 0
2
= − n2l π2 [1 + (−1)n ], n = 1, 2, . . . .
2
nπν

∞
l2 4l 2 1+(−1) n − t
u(x, t) = nπx
.
Í

⇒ − cos e l
6 π2 n2 l
n=1

EXERCISE 12.2.1. Solve the following boundary value problems related to one dimensional heat flow in
metal rods:

(a) ut = u xx ; u(0, t) = u(π, t) = 0 for t > 0, u(x, 0) = 60 sin x − 20 sin 3x for 0 < x < π
(b) 3ut = u xx ; u x (0, t) = u x (2, t) = 0 for t > 0, u(x, 0) = cos2 πx for 0 < x < 2
(c) 5ut = u xx ; u x (0, t) = u x (10, t) = 0 for t > 0, u(x, 0) = 4x for 0 < x < 10
Ans.
(a) u(x, t) = 60e−t sin x − 20e−9t sin 3x
h 2 t/3
i
(b) u(x, t) = 2 1 + e
1 −16π cos 4πx

(−1) n+1 −n2 π2 t/500

(c) u(x, t) = 80 Í∞
π n=1 n e sin(nπx/10)

12.3 Heat-Flow with nonhomogeneous Boundary Conditions

Suppose that the ends x = 0 and x = l are maintained at nonzero temperatures, say at θ 1 and θ 2 respec-
tively. Then the nonhomogeneous boundary conditions are given by

u(0, t) = θ 1, u(l, t) = θ 2 for all t > 0. (12.3.1)

Since the boundary conditions are nonzero, the separation of variables cannot be applied directly to solve
(12.2.1). The general solution of (12.2.1) is given by

u(x, t) = us (x) + ut (x, t). (12.3.2)

where us (x) is the steady-state solution and ut (x, t) is the transient part.
We may assume that the temperature function u(x, t) satisfies the condition that

lim u(x, t) = us (x), (12.3.3)

n→∞

where us (x) is the steady-state temperature which is independent of time t.

In fact, us (x) is given by

∂2 u s
∂x 2
=0 ⇒ us (x) = Ax + B.
and satisfies (12.3.1), that is

us (0) = θ 1 = A.0 + B, u(l) = θ 2 = A.l + B

which imply that

θ2 −B θ2 −θ1
B = θ 1, A = l = l ·

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 89

Thus we get the steady-state solution of (12.2.1) is

(θ2 −θ1 )x
us (x) = l + θ1 . (12.3.4)

We then find the transient part ut (x, t) by solving (12.2.1) with the homogeneous boundary conditions:

ut (0, t) = u(0, t) − us (0) = 0, u(l, t) = u(l, t) − us (l) = 0 for all t > 0, (12.3.5)

and the initial condition:

h i
(θ2 −θ1 )x
ut (x, 0) = u(x, 0) − us (x) = f (x) − l + θ 1 for 0 < x < l. (12.3.6)

EXERCISE 12.3.1. Suppose that a laterally insulated rod with length L = 50 centi meters and thermal
diffusivity κ = 1 has initial temperature u(x, 0) = 0 and endpoint temperatures u(0, t) = 0, u(50, t) =
Í∞ (−1)n+1 −n2 π2 t/2500
100◦C. Show that u(x, t) = 200
π n=1 n e sin(nπx/50).

12.4 Two Dimensional Steady State Heat Flow in Metal Plates

Consider a heated thin rectangular metal plate with dimensions

0 ≤ x ≤ a, 0 ≤ y ≤ b where a, b > 0.

Suppose its upper and lower surfaces are thermally insulated so that the heat flow can take place only in
the x and y directions.

The heat-flow in the plate is governed by the two dimensional heat equation:
 
∂u 2 ∂2 u ∂2 u
∂t = a ∂x 2 + ∂y 2 , (12.4.1)

where a is a physical constant, called the heat constant or diffusivity of the metal with which the rod is
made.

In the steady state, the temperature u will be independent of time t. Thus (12.4.1) reduces to
∂2 u ∂2 u
∂x 2
+ ∂y 2
= 0. (12.4.2)

This is called the Laplace equation in two dimensions.

We wish to determine the steady state temperature u(x, y) at any point (x, y) under four boundary condi-
tions:

• Two along the edges x = 0 and x = a

• Two along the edges y = 0 and y = b

Three Forms of Solution of Laplace Equation

By the method of variables-separables, one can obtain the following three possible forms of the solution
for (12.4.2):

u(x, y) = (Ae px + Be−px ) (C cos py + D sin py) (12.4.3)

Department of Mathematics, Vellore Institute of Technology

90 12.4. Two Dimensional Steady State Heat Flow in Metal Plates

u(x, y) = (A cos px + B sin px) (Ce py + De−py ) (12.4.4)

u(x, y) = (Ax + B) (C y + D) (12.4.5)

where A, B, C and D are arbitrary constants, which can be determined by using the boundary conditions.

If
u(x, 0) = 0 and u(x, b) = 0 for all 0 < x < a (12.4.6)
or
u(0, y) = 0 and u(a, y) = 0 for all 0 < y < b, (12.4.7)
then form (12.4.5) reduces to the trivial solution u ≡ 0.

Horizontal Edges at Zero Temperature

Case (a): Homogeneous boundary conditions (12.4.6) prevail along the edges y = 0 and y = b and the
boundary conditions along the x-edges are

u(0, y) = r(y) and u(a, y) = 0 for all 0 < y < b, (12.4.8)

then the Fourier solution of (12.4.2) is given by

∞
Õ h i
nπ(a−x) nπy

u(x, y) = bn sinh b sin b , (12.4.9)
n=1

where bn is evaluated from the condition:

u(0, y) = r(y) for 0 < y < b.

In fact, writing x = 0 in (12.4.9) and using this we get

∞
Õ
nπy

u(0, y) = r(y) = nπa
b , 0<y<b


bn sinh b sin
n=1

which is the half-range sine series of

u(0, y) = r(y) in 0 < y < b.

Thus
∫b
nπy

nπa
= 2
for n = 1, 2, . . . .


bn sinh b b r(y) sin b dy,
0

or
∫b
nπy

bn = 2
nπa
 r(y) sin b dy, n = 1, 2, . . . . (12.4.10)
b sinh b
0

Example 12.4.1. Consider a thin square plate of length 30 cm with thermally insulated lateral surfaces.
If the edges y = 0, y = 30 and x = 30 are maintained at zero temperature, while the edge x = 0 at 100◦ C
until the steady state is attained. Determine the steady state temperatuture in the plate. Also find the
temperature at the centre of the plate.

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 91

Solution. Here u(0, y) = r(y) = 100 for 0 < y < 30. Then with a = b = 30, we get from (12.4.10) of Case (a)

∫30
nπy

bn = 2
30 sinh(nπ) 100 sin 30 dy
0
 cos  nπy  30
 
30  1−(−1) n
= 15 sinh(nπ) − nπ/30 
100
= 200
· n sinh(nπ) , n = 1, 2, ...

π
y=0
 

Therefore, the complete solution (12.4.9) of the Laplace equation (12.4.2) is

∞
Õ h i
nπ(30−x) nπy

u(x, y) = bn sinh 30 sin 30
n=1
∞ n o h i
1−(−1) n
Õ
nπ(30−x) nπy

= 200
π n sinh(nπ) sinh 30 sin 30 ·
n=1

Horizontal Edges at Zero Temperature

Case (a*): Homogeneous boundary conditions (12.4.6) prevail along the edges y = 0 and y = b and the
boundary conditions along the x-edges are

u(0, y) = 0 and u(a, y) = r(y) for all 0 < y < b, (12.4.11)

then the Fourier solution of (12.4.2) is given by

∞
Õ
nπy

u(x, y) = nπx
b ,


bn sinh b sin (12.4.12)
n=1

where bn is evaluated from the condition:

u(a, y) = r(y) for 0 < y < b.

In fact, writing x = a in (12.4.12) and using this we get

∞
Õ
nπy

u(a, y) = r(y) = nπa
b , 0<y<b


bn sinh b sin
n=1

which is the half-range sine series of

u(a, y) = r(y) in 0 < y < b.

Thus
∫b
nπy

nπa
= 2
for n = 1, 2, . . . .


bn sinh b b r(y) sin b dy,
0

or
∫b
nπy

bn = 2
nπa
 r(y) sin b dy, n = 1, 2, . . . . (12.4.13)
b sinh b
0

Department of Mathematics, Vellore Institute of Technology

92 12.4. Two Dimensional Steady State Heat Flow in Metal Plates

Vertical Edges at Zero Temperature

Case (b): Homogeneous boundary conditions (12.4.7) prevail along the edges x = 0 and x = a and the
boundary conditions along the y-edges are

u(x, 0) = f (x) and u(x, b) = 0 for all 0 < x < a, (12.4.14)

then the Fourier solution of (12.4.2) is given by

∞
Õ h i
nπ(b−y)
u(x, y) = b∗n sin nπx
,


a sinh a (12.4.15)
n=1

where b∗n is evaluated from the condition:

u(x, 0) = f (x) for 0 < x < a.

In fact, writing y = 0 in (12.4.15) and using this we get

∞ h
Õ  i
u(x, 0) = f (x) = b∗n sinh nπb nπx
,0<x<a


a sin a
n=1

which is the half-range sine series of

u(x, 0) = f (x) in 0 < x < a.

Thus
  ∫a
b∗n nπb
= 2 nπx
for n = 1, 2, . . . .


sinh a a f (x) sin a dx,
0

or
∫a
b∗n = 2 nπx
n = 1, 2, . . . .


nπb
 f (x) sin a dx, (12.4.16)
a sinh a
0

Example 12.4.2. Consider a thin square plate of length 100 cm with thermally insulated lateral surfaces.
If the edges x = 0, x = 100 and y = 100 are maintained at zero temperature, while the edge y = 0 at a
constant temperature u0 until the steady state is attained. Determine the steady state temperatuture in
the plate. Also find the temperature at the centre of the plate.
Solution. Here u(x, 0) = f (x) = u0 for 0 < x < 100. Then by (12.4.16) of Case (b),

∫100
b∗n = 2 nπx


100 sinh(nπ) u0 sin 100 dx
0
 cos  nπx  100
 
100  2u0 1−(−1) n
= 50 sinh(nπ)
u0
− nπ/100  = · n sinh(nπ) , n = 1, 2, ...

π
x=0
 

Therefore, the complete solution (12.4.15) of the Laplace equation (12.4.2) is

∞ h
Õ n oi
nπ(100−y)
u(x, y) = b∗n sinh nπx


100 sin 100
n=1

Department of Mathematics, Vellore Institute of Technology

 Chapter 12. Applications of Partial Differential Equations 93

∞ n o h i
1−(−1) n
Õ
2u0 nπ(100−y)
= nπx


π n sinh(nπ) sin 100 sinh 100 ·
n=1

The temperature at the centre (50, 50) is obtained from by writing x = y = 50 in the above expression,
that is
∞ n o
1−(−1) n
Õ
2u0
u(50, 50) = nπ

 nπ 
π n sinh(nπ) sin 2 sinh 2 ·
n=1

Department of Mathematics, Vellore Institute of Technology

94 12.4. Two Dimensional Steady State Heat Flow in Metal Plates

Department of Mathematics, Vellore Institute of Technology

Chapter 13

Complex Fourier Transform

13.1 Complex Fourier Transform

Let f (x) be defined for all real −∞ < x < ∞. Then its Fourier transform is given by
∫ ∞
F f (x) = F(ω) = √ 1
f (x)e−iωx dx

(13.1.1)
2π
−∞

Let f (x) represent a signal. Then

(a) its Fourier transform F (ω) is known as complex frequency spectrum,

(b) the graph of the magnitude |F (ω)| is called its amplitude spectrum,
(c) the graph of the argument arg F (ω) is called its phase spectrum.

Fourier Inversion formula:

∫ ∞
f (x) = F −1
F(ω) = √1 F(ω)eiωx dω.

(13.1.2)
2π
−∞

We say that h f (x), F(ω)i is a Fourier pair.

13.2 Properties of the Fourier Transform

Theorem
 13.2.1 (Linearity). Let F(ω) and G(ω) be the Fourier transforms of f (x) and g(x) respectively.
Then F a f (x) + bg(x) = aF(ω) + bG(ω), where a and b are scalars, not both zeros.

Example 13.2.1. Find the Fourier transform of the negative exponential signal
(
e−ax , x ≥ 0,
f (x) =
0, elsewhere.

Hence find the amplitude and phase spectra of the signal f (x).
Solution.
∫ ∞ ∫ ∞
F f (x) = F(ω) = √1 −ax −iωx
dx = √1 e−(a+iω)x dx

e e
2π 2π
0 0
 −(a+iω)x ∞
= √1 · − e a+iω  = √1 a+iω
1
 
2π x=0 2π
= 1
.
 a  ω 
√
a2 +ω2
− i a2 +ω 2
2π

(a) Amplitude spectrum of f (x) is |F(ω)| = √1 · √ 1 , and

2π a2 +ω2

(b) Phase spectrum of f (x) is arg F(ω) = tan−1 ω

a ·
Example 13.2.2. Find the Fourier transform of the saw-tooth signal
(
x, |x| < a,
f (x) =
0, elsewhere.

95
96 13.2. Properties of the Fourier Transform

Solution.
∫ a
F(ω) = √1 xe−iωx dx
2π
∫−aa
= √1 x[cos ωx − i sin ωx] dx
2π
−a
odd function even function
∫ a z }| { ∫ a z }| {
= √1 x cos ωx −i · √1 x sin ωx dx
2π 2π
−a −a
| {z } | {z }
=0
∫a
=2 0
x sin ωx dx
q   a
= −i 2π  x − cosωωx − (1) − sinωωx

2

x=0
q  
= i 2π a cos
ω
ωa
− sinωωa
2 ,ω , 0

Example 13.2.3. Find the Fourier transform of the parabolic signal

(
1 − x 2, |x| < 1,
f (x) =
0, elsewhere.
∫ ∞
dx =
 sin x−x cos x x

3π
Hence deduce that x3
cos 2 16 ·
0
Solution.
∫ 1
F(ω) = √1 (1 − x 2 )e−iωx dx
2π
−1
∫ 1
= √1 (1 − x 2 )[cos ωx − i sin ωx] dx
2π
−1
even function odd function
∫ 1z
}| { ∫ 1z }| {
=√1 2
(1 − x ) cos ωx −i · √1 2
(1 − x ) sin ωx dx
2π 2π
−1 −1
| {z } | {z }
=2
∫1
(1−x 2 ) cos ωx dx =0
0
q  1
= 2π (1 − x 2 ) − sinωωx − (−2x) − cosω2ωx + (−2) − sinωωx
  
3

x=0
q 
1
= 2π  0 − 2 cos + sin
ω ω

ω2 ω3
− (0 − 0 − 0)x=0
q 
= −2 2π cos − sin ,ω , 0
ω ω

ω2 ω3

By the inversion formula (13.1.2),

∫ ∞
f (x) = F −1
F(ω) = √1 F(ω)eiωx dω

2π
−∞
∫ ∞ q  
e dω = f (x)
iωx
⇒ √1 −2 2 cos ω sin ω
− ω3
2π π ω2
−∞
∫ ∞  
⇒ cos ω
ω2
− sin ω
ω3
eiωx dω = − 2π · f (x)·
−∞

Writing x = 1/2 in this, we see that

∫ ∞ 
cos ω
+ sin ω
eiω/2 dω = − 2π f 1


ω2 ω3 2
−∞

Department of Mathematics, Vellore Institute of Technology

 Chapter 13. Complex Fourier Transform 97

or
∫ ∞  
cos ω
+ sin ω
dω = − 2π · 3
= − 3π
ω

ω


ω2 ω3
cos 2 − i sin 2 4 8 ·
−∞

Since the imaginary part on the left hand side is odd function of ω, we get
∫ ∞ 
cos ω sin ω
cos ω2 dω = − 16
3π


ω2
− ω3
·
0

Multiplying with −1, and replacing ω with x in the integral, we get

∫ ∞
cos 2x dx = 16
 sin x−x cos x
3π
x3
·
0

EXERCISE 13.2.1. Find the complex Fourier transform of the triangular pulse
(
1 − |x|, |x| < 1,
f (x) =
0, elsewhere.

q
2 1−cos ω
Ans. π · ω2
, for ω , 0
Theorem 13.2.2 (Duality). h f (x), F(ω)i is a Fourier transform pair if and only if hF(x), f (−ω)i is a
Fourier pair.
Example 13.2.4. Find the Fourier transform of the rectangular pulse
(
k, |x| < a,
f (x) =
0, elsewhere,

where k and a are positive numbers. Hence show that

∫ ∞
(a) sin aω cos ωx
ω dω = π
2 · f (x),
0
∫ ∞
(b) sin x
x dx = 2π ,
0

Then, obtain the Fourier transform of g(x) = sin x

x ·
Solution.
∫ a  −iωx a
F(ω) = √1 dx =
−iωx √k
 e
k·e

2π 2π
− iω 
x=−a
 −a
iωa −iωa
  
= √k e −e iω = √k 2 sin ωa
2π 2π ω
q  
= k 2π sinωωa .

(a) By the inversion formula (13.1.2),

∫ ∞
f (x) = F F(ω) =
−1
F(ω)eiωx dω√1

2π
−∞
∫ ∞ q  
⇒ √ 1
k 2π sinωωa eiωx dω = f (x)
2π
−∞

Department of Mathematics, Vellore Institute of Technology

98 13.2. Properties of the Fourier Transform

∫ ∞   ∫ ∞  
⇒ sin ωa
ω eiωx dω = π
k · f (x) ⇒ sin ωa
ω (cos ωx + i sin ωx) dω = π
k · f (x).
−∞ −∞

Comparing the real parts on both sides,

∫ ∞ 
sin ωa cos ωx
ω dω = π
k · f (x), −∞ < x < ∞.
−∞

(b) In particular, for x = 0, a = 1, this gives

∫ ∞  ∫ ∞  
sin ω
ω dω = kπ · f (0) = π
k · k = π or sin ω
ω dω = 2π ·
−∞ 0

Replacing ω with x in the integral, we get

∫ ∞  
sin x
x dx = 2π ·
0

To find the Fourier transform of g(x) = sinx x , we employ the duality property that hF(x), f (−ω)i is a Fourier
q  
pair. That is, F F(x) = f (−ω). But F(x) = k 2π sinxax for any a > 0. Thus


(
k, |ω| < a,
 q 
F k 2π · sin ax
x = f (ω) =
0, elsewhere
(p
2, |ω| < 1,
π
or with a = 1, we getF =
 sin x
x
0, elsewhere.
2 /2
Example 13.2.5. Show that the Gaussian signal f (x) = e−x , −∞ < x < ∞ is self-reciprocal.
Solution. We say that f (x) is self-reciprocal if F(ω) = f (ω).
∫ ∞ ∫ ∞
e−(x +2x.iω)/2 dx
2 2
F(ω) = √1 e−x /2 · e−iωx dx = √1
2π 2π
∫−∞
∞
−∞
2
∫ ∞
2 2 2
e √ /2
−ω
= √1
e −[(x+iω) −(iω) ]/2
dx = e−(x+iω) /2 dx
2π 2π
−∞ −∞
∫ ∞
2 2 2
= e−ω /2 · √1 e−z /2 dz = e−ω /2 ·
2π
−∞
| {z }
=1

2 /2
Thus f (x) = F(ω). Hence f (x) = e−x is self-reciprocal.
Theorem 13.2.3 (Change of Scale). Let F(ω) be the Fourier transform of f (x), and a , 0. Then

F f (ax) = 1
 ω


|a | ·F a · (13.2.1)

∫ ∞
Proof. Let I = F f (ax) = √1 f (ax)e−iωx dx.

2π
−∞

Case (a): If a > 0, write ax = u so that dx = du/a and u ranges from −∞ to ∞ as x ranges from −∞ to ∞.
Therefore,
∫ ∞
I = F f (ax) = a · √ f (u)e−i(ω/a)u du = a1 · F ωa ·
 1 1

2π
−∞

Department of Mathematics, Vellore Institute of Technology

 Chapter 13. Complex Fourier Transform 99

Case (b): If a < 0 so that −a > 0. Then write −ax = v so that dx = − dv/a and v ranges from ∞ to −∞ as
x ranges from −∞ to ∞. Therefore,
∫ −∞
I = F f (ax) = a · √ f (−v)ei(ω/a)v dv.
 1 1
2π
∞

Now, write −v = p so that dv = − dp and p ranges from −∞ to ∞ as v ranges from ∞ to −∞.

Therefore, the above equation becomes
∫ ∞
I = − a1 · √1 f (p)e−i(ω/a)p dp = − a1 · F ωa ·


2π
−∞

This completes the proof. 

2 /2 2 /2 2
Example 13.2.6. Given that the Fourier transform of e−x is e−ω , find the Fourier transform of e−bx ,
where b > 0.
2 /2
Solution. Let f (x) = e−x . Then
2
F f (ax) = F(ω) = e−ω /2 .


Since
2 2 /2
√
2bx 2 /2
√
e−bx = e−2bx = e− = f ( 2bx),

by (13.2.1), we see that

 √   2 /4b
F f ( 2bx) = √1 = e−ω .

·F √ω
2b 2b

Theorem 13.2.4 (Spacial-Shifting). Let F(ω) be the Fourier transform of f (x), and a , 0. Then

F f (x − a) = e−iaω F(ω).

(13.2.2)

Theorem 13.2.5 (Frequency-Shifting). Let F(ω) be the Fourier transform of f (x), and a , 0. Then

F e−iax f (x) = F(ω + a).


(13.2.3)

Theorem 13.2.6 (Modulation). Let F(ω) be the Fourier transform of f (x), and a , 0. Then

F f (x) cos ax = 12 [F(ω − a) + F(ω − a)].


(13.2.4)
2 /2 2 /2
EXERCISE 13.2.2. Given that the Fourier transform of e−x is e−ω , find the Fourier transform of
2 2
e−4(x−3) and e−x cos 3x.

Example 13.2.7. Find the Fourier transform of

(
1, 0 < x < 1,
f (x) =
0, elsewhere.

Hence obtain the Fourier transform of

(
e−iax , 0 < x < 1,
g(x) =
0, elsewhere.

Department of Mathematics, Vellore Institute of Technology

100 13.2. Properties of the Fourier Transform

Solution.
∫ 1  −iωx 1  
F f (x) = √1 −iωx
dx = √1 · − e iω  = − √i −iω
.

e 1−e
 
2π 2π x=0 ω 2π
0

Then by the frequency-shifting property (13.2.3), we get

i (1−e−i(ω+a) )
F g(x) = F e−iax f (x) = F(ω + a) =
 
√ ·
(ω+a) 2π

Theorem 13.2.7 (Multiplication by x n ). Let F(ω) be the Fourier transform of f (x). Then

F x n f (x) = (i)n F (n) (ω), for n = 1, 2, ...


(13.2.5)

Example 13.2.8. Find the Fourier transform of the decaying exponential signal f (x) = e−|x | for all −∞ <
x < ∞. Hence
∫ ∞
(a) derive that cos ωx
1+ω2
dω = π
2 · e− |x | , and
0

(b) obtain the Fourier transform of xe− |x | .

Solution.
∫ ∞ ∫ ∞
F(ω) = √1 e − |x | −iωx
e dx = √1 e−|x | (cos ωx − i sin ωx) dx
2π 2π
−∞ −∞
even function odd function
∫ ∞z }| { ∫ ∞z }| {
= √1 − |x |
e cos ωx dx −i √1 e −|x |
sin ωx) dx
2π 2π
−∞ −∞
| {z } | {z }
=0
∫∞
=2 0
e−| x | cos ωx dx
∫ ∞ q  −x ∞ q
= √2 e−x cos ωx dx = 2
·  e (− cos1+ω
ωx+ω sin ωx) 
= 2
π ·
1
.

2  1+ω2
2π π x=0
0

(a) By the inversion formula (13.1.2),

∫ ∞
f (x) = F −1
F(ω) = √1 F(ω)eiωx dω

2π
∫−∞
∞ q 
= √1 2
· 1
π 1+ω2 e
iωx
dω
2π
−∞
∫ ∞ 
= 1 1
eiωx dω

π 1+ω2
∫−∞
∞
= 1 1
(cos ωx + i sin ωx) dω
 
π 1+ω2
∫−∞
∞  
= 2
π
cos ωx
1+ω2
dω
0

or ∫ ∞
cos ωx dω
1+ω2
= π
2 · e−|x | , −∞ < x < ∞.
0

(b) By the multiplication by x property (13.2.5) with n = 1, we have

q  q
F x f (x) = i · F 0(ω) = i · dω
d 2 1
= 2 2iω

·
π 1+ω2 − π · (1+ω2 )2

Department of Mathematics, Vellore Institute of Technology

 Chapter 13. Complex Fourier Transform 101

or
q
F xe−|x | = − 2π · 2iω

(1+ω2 )2

13.3 Fourier Convolution

∫ ∞
We define ( f ∗ g)(x) = f (u)g(x − u) du, −∞ < x < ∞.
√1
2π
−∞
Theorem 13.3.1 (Convolution Theorem). F ( f ∗ g)(x) = F(ω) · G(ω).


Proof. By definition, we have

∫ ∞
F ( f ∗ g)(x) = √1 ( f ∗ g)(x)e−iωx dx

2π
∫−∞
∞ ∫ ∞ 
= √ 1 √ 1
f (u)g(x − u) du e−iωx dx
2π 2π
−∞ −∞
∫ ∞  ∫ ∞ 
= √1 f (u)e−iωu √1 g(x − u)e−iω(x−u) dx du
2π 2π
−∞
 ∫ ∞  −∞ ∫ ∞ 
= √ 1
f (u)e −iωu
du √ 1
g(x − u)e −iω(x−u)
dx
2π 2π
−∞ −∞
 ∫ ∞  ∫ ∞ 
= √1 f (u)e −iωu
du √1
g(v)e −iωv
dv
2π 2π
−∞ −∞
= F(ω) · G(ω).

That is, the Fourier transform of the convolution of two functions is equal to the product of the respective
Fourier transforms. 

Theorem 13.3.2 (Fourier transform of Derivatives). Suppose that

(a) f , f 0, f 00, ..., f (n−1) all tend to 0 as |x| → ∞, and

∫ ∞ ∫ ∞ 
| f (x)| dx < ∞,  f (x) dx < ∞, k = 1, 2, ..., n.
 (k) 
(b)
−∞ −∞

If F(ω) be the Fourier transform of f (x), then

F f (n) (x) = (iω)n F(ω), for n = 1, 2, ...



Department of Mathematics, Vellore Institute of Technology

102 13.3. Fourier Convolution

Department of Mathematics, Vellore Institute of Technology

Chapter 14

Fourier Sine and Cosine Transforms

14.1 Fourier Sine and Cosine Transforms

Let f (x) be defined for all x > 0. Then

(a) The Fourier Sine transform of f is given by

q ∫ ∞
Fs f (x) = Fs (ω) = 2π

f (x) sin ωx dx
0
The Sine Inversion Formula of f is given by
q2 ∞
∫
f (x) = Fs Fs (ω) = π
−1

Fs (ω) sin ωx dω
0
(b) The Fourier Cosine transform of f is given by
q ∫ ∞
Fc f (x) = Fc (ω) = 2π

f (x) cos ωx dx
0
The Cosine Inversion Formula of f is given by
∫ ∞
q
f (x) = Fc−1 Fc (ω) = 2π

Fc (ω) cos ωx dω
0

Example 14.1.1. Find the cosine transform of

(
1, 0 ≤ x ≤ a
f (x) =
0, elsewhere,

where a > 0.
Solution.
q ∫ ∞ q ∫ a q  q
2  sin ωx  a
Fc (ω) = f (x) cos ωx dx = cos ωx dx = =
2 2
 2 sin ωa
π π π ω x=0 π · ω
0 0

Example 14.1.2. Find the cosine transform of f (x) = e−ax , a > 0, x > 0.
Solution.
q ∫ ∞ q ∫ ∞
Fc (ω) = 2
π f (x) cos ωx dx = e−ax · cos ωx dx
2
π
q 0 0
∞ q
2  e−a x
= π · a2 +ω2 (−a cos ωx + ω sin ωx)x=0 = 2π · a2 +ω
a
2,

Example
∫ ∞ 14.1.3. Find the sine transform of f (x) = e−ax , a > 0, x > 0, and and hence deduce that
−a
x sin mx
1+x 2
dx = πe2 ·
0
Solution.
q ∫ ∞ q ∫ ∞
Fs (ω) = 2
π f (x) sin ωx dx = e−ax · sin ωx dx
2
π
q 0 0
∞ q
2  e−a x
= π · a2 +ω2 (−a sin ωx − ω cos ωx)x=0 = 2π · a2 +ω
ω
2

103
104 14.1. Fourier Sine and Cosine Transforms

By the inversion formula,

q ∫ ∞ q ∫ ∞ q 
f (x) = 2
π Fs (ω) sin ωx dω = 2
π
2
π · ω
a2 +ω2
sin ωx dω
0 0

or ∫ ∞
πe−a x
ω sin ωx
a2 +ω2
dω = π
2 f (x) = 2 ·
0
Changing x to m, ω to x and a = 1 we get
∫ ∞
πe−m
x sin mx
1+x 2
dx = 2 ·
0

Example 14.1.4. Find the sine transform of f (x) = 1/x, x > 0.

Solution.
q ∫ ∞ q q
Fs (ω) = 2
π
sin ωx
x dx = 2
π · π
2 = π
2·
0
2 /2
Example 14.1.5. Show that f (x) = e−x , −∞ < x < ∞ is self-reciprocal under the cosine transform.
2
Solution. We recall that f (x) = e−x /2 is self-reciprocal under the Fourier transform. That is
∫ ∞ ∫ ∞
−ω2 /2 −x 2 /2 2
e =√ 1
e ·e −iωx
dx = √1
e−x /2 (cos ωx − i sin ωx) dx
2π 2π
−∞ −∞
even function odd function
∫ ∞
z }| { ∫ ∞
z }| {
−x 2 /2 −x 2 /2
= √1 e cos ωx dx − i √1 e sin ωx dx
2π 2π
−∞ −∞
| {z }
=0
∫ ∞ q ∫ ∞
−x 2 /2 −x 2 /2
= √2 e cos ωx dx = 2
e cos ωx dx.
2π π
0 0
 2 2 2
In other words, Fc e−x /2 = e−ω /2 . Hence f (x) = e−x /2 is self-reciprocal.
Theorem 14.1.1 (Cosine transform from the Sine transform). If Fs f (x) = Fs (ω) and f (x) → 0


as x → ∞, then

Fc f 0(x) = ωFs (ω)


(14.1.1)
Fc x f (x) = Fs (ω)
0

(14.1.2)

Theorem 14.1.2 (Sine transform from the Cosine transform). If Fc f (x) = Fc (ω) and f (x) → 0


as x → ∞, then

Fs f 0(x) = −ωFc (ω)


(14.1.3)
Fs x f (x) = −Fc (ω)
0

(14.1.4)
2
Example 14.1.6. Given that f (x) = e−x /2 is self-reciprocal under the cosine transform, find the sine
2 2
transform of g(x) = xe−x /2 and the cosine transform of h(x) = x 2 e−x /2 .
 2 2 2 2
Solution. Given that Fc e−x /2 = e−ω /2 . Let f (x) = e−x /2 and F(ω) = e−ω /2 . Then,

Fs f 0(x) = −ωFc (ω)


2
⇒ Fs − xe−x /2 = −ωFw (ω)

2 2
⇒ Fs xe−x /2 = ωFc (ω) = ωe−ω /2 = G s (ω), say.


Department of Mathematics, Vellore Institute of Technology

 Chapter 14. Fourier Sine and Cosine Transforms 105

2 /2
In other words, g(x) = xe−x is self-reciprocal under the sine transform.

Now by the multiplication by x property, we have

2
n 2 /2
o 2 /2
Fc x(xe−x /2 ) = d
= d
ωe−ω = (1 − ω2 )e−ω

di f f ω {G s (ω)} dω

2
Thus Fc h(x) = (1 − ω2 )e−ω /2 .

−a x
Example 14.1.7. Find the sine transform of f (x) = e x and hence the cosine transform of g(x) = e−ax ·
q ∫ ∞ 
e−a x
Solution. Let I = Fs (ω) = 2π x sin ωx dx. Differentiating w. r. t. ω under the integral sign, this
0
gives
q ∫ ∞   q ∫ ∞ q
e−a x
dI
dω = 2
π (x cos ωx) x dx = 2
π e −ax
cos ωx dx = 2
π · a
a2 +ω2
·
0 0
On one hand, this gives ∫ ∞ q
Fc e −ax
= e−ax cos ωx dx = 2 a
.

π · a2 +ω2
0
On the other hand, separating the variables in the differential equation
q
dI
dω = 2 a
π · a2 +ω2

and then integrating, we get the general solution

q
I = 2π · tan−1 + A.
ω


a

Then writing ω = 0 in this so that

q q ∫ ∞  
e−a x
2
π · tan−1 0 + A = 2
π x sin 0 dx = 0
0
q
or A = 0. Thus Fs (ω) = 2
· tan−1 ω


π a ·

EXERCISE 14.1.1. Find the sine transform of f (x) = 1

x(x 2 +a2 )
and hence the cosine transform of g(x) =
1
x 2 +a2
·

Ans.
Fs (ω) = 1
· (1 − e−aω ), G c (ω) = 1
· e−aω
pπ pπ
a2 2 a 2

EXERCISE 14.1.2. Find the cosine transform of f (x) = 1

x 2 +1
and hence the sine transform of g(x) = x
x 2 +1
·

Ans.
Fc (ω) = G s (ω) = 2π · e−ω .
p

Department of Mathematics, Vellore Institute of Technology

106 14.1. Fourier Sine and Cosine Transforms

Department of Mathematics, Vellore Institute of Technology

Chapter 15

Parseval’s Identities

15.1 Parseval’s Identities

Theorem 15.1.1 (Parseval’s Identity for Complex Fourier Transform).
∫ ∞ ∫ ∞
( f (x)) dx =
2
|F(ω)| 2 dω. (15.1.1)
−∞ −∞

Proof. By definition, we note that

∫ ∞ ∫ ∞  ∫ ∞ 
( f (x)) dx =
2
f (x) √ 1
F(ω) e iωx
dω dx
2π
−∞ −∞ −∞
∫ ∞  ∫ ∞ 
= F(ω) √1 iωx
f (x) e dx dω
2π
−∞ −∞
∫ ∞  ∫ ∞ 
= F(ω) √ 1
f (x) e −iωx dx dω
2π
−∞ −∞
∫ ∞  ∫ ∞ 
= F(ω) √1
f (x) dx dω e−iωx
2π
∫−∞
∞ ∫ ∞ −∞

= F(ω) F(ω) dω = |F(ω)| 2 dω

−∞ −∞

Theorem 15.1.2 (Parseval’s Identity for Sine and Cosine Transforms).

∫ ∞ ∫ ∞
(a) f (x) g(x) dx = Fs (ω) G s (ω) dω and
0 0
∫ ∞ ∫ ∞
| f (x)| dx =
2
|Fs (ω)| 2 dω, in particular,
∫0 ∞ 0
∫ ∞
(b) f (x) g(x) dx = Fc (ω) G c (ω) dω and
∫0 ∞ ∫ ∞0
| f (x)| 2 dx = |Fc (ω)| 2 dω, in particular.
0 0
∫ ∞
Remark 15.1.1. The integral |h(ξ)| 2 di f f ξ of the squared magnitude of a function h(ξ) is known
−∞
as the energy of the signal h. Thus the energy of the signal f (x) is the same as the the energy contained in
its transform.
 −ax q 2
Example 15.1.1. Given that Fc e = π · ω2 +aa
2 , prove that

∫ ∞
dx
(x 2 +a2 )(x 2 +b 2 )
= π
2ab(a+b) ·
0

Solution. Let f (x) = e−ax , g(x) = e−bx . Then

q
Fc f (x) = 2π · a
= Fc (ω),

ω2 +a2

107
108 15.1. Parseval’s Identities

q
Fc g(x) = 2π · b
= G c (ω).

ω2 +b 2

Hence by Parseval’s identity for cosine transform, we have

∫ ∞ ∫ ∞
Fc (ω) G c (ω) dω = f (x) g(x) dx
0 0
∫ ∞ ∫ ∞
⇒ 2π · ab
(ω2 +a2 )(ω2 +b 2 )
dω = e−ax e−bx dx = 1
a+b
0 0

or ∫ ∞
dω
(ω2 +a2 )(ω2 +b 2 )
= π
2ab(a+b) ·
0
Changing ω to x, this gives
∫ ∞
dx
(x 2 +a2 )(x 2 +b 2 )
= π
2ab(a+b) ·
0

Example 15.1.2. Given that

q q
Fc e −ax
= 2 a
and Fc g(x) = 2 sin ωa
 
π · ω2 +a2 π · ω

where (
1 (0 < x < a)
g(x) =
0, elsewhere,
∫ ∞ 2
π (1−e−a )
prove that sin ax
x(x 2 +a2 )
dx = 2a2
·
0
Solution. From the Parseval’s identity for cosine transform, we have
∫ ∞ ∫ ∞
Fc (ω) G c (ω) dω = f (x) g(x) dx
0 0
q ∫ ∞ ∫ a 2
1−e−a
⇒ 2
π · a sin ωa
ω(ω2 +a2 )
dω = e−ax · 1 dx = a2
0 0
∫ ∞ 2
∫ ∞ 2
π (1−e−a ) π (1−e−a )
or sin ωa dω
ω(ω2 +a2 )
= 2a2
. Changing ω to x, this gives sin ax
x(x 2 +a2 )
dx = 2a2
·
0 0
Example 15.1.3. Find the sine and cosine transforms of e−x , and use Parseval’s identities to prove that
∫ ∞ ∫ ∞
x2
dx
(x 2 +1)2
=4=
π
(x 2 +1)2
dx·
0 0

Solution. Note that

q ∫ ∞
Fc (ω) + i Fs (ω) = 2
π e−x (cos ωx + i sin ωx) dx
·
0
q ∫ ∞ q ∫ ∞
= π·
2 −x iωx
e e dx = 2π · e−(1−iω)x dx
q 0 ∞ q 0
−( 1 −iω)x
= 2π · − e 1−iω  = 2π · 1−iω
1
 
x=0
q h  i
= 2π · 1+ω 1
2 + i ω
1+ω2
·

Department of Mathematics, Vellore Institute of Technology

 Chapter 15. Parseval’s Identities 109

Comparing the real and imaginary parts on both sides, we get

q q
Fc (ω) = π · 1+ω2 , Fs (ω) = 2π ·
2 1 ω
1+ω2
·

Then from Parseval’s identity for cosine transform, we get

∫ ∞ ∫ ∞ ∫ ∞
2
π
dω
(ω2 +1)2
= e −2x
dx = 2 or
1 dω
(ω2 +1)2
= 4π ,
0 0 0

from which the first result follows by replacing ω with x. Similarly, from Parseval’s identity for sine trans-
form, the second result follows.
q
Example 15.1.4. Given that F f (x) = 2π · sinωωa = F(ω)


where (
1 (−a ≤ x ≤ a)
f (x) =
0, elsewhere,
∫ ∞
sin2 ax
prove that x2
dx = πa2 ·
0
Solution. From the Parseval’s identity for the complex Fourier transform,
∫ ∞ ∫ ∞ ∫ ∞ ∫ a
sin2 ωa
|F(ω)| 2 dω = ( f (x))2 dx ⇒ 2π · ω2
dω = 1. dx
−∞ −∞ −∞ −a
∫ ∞ ∫ ∞
sin2 ωa sin2 ωa
⇒ ω2
dω = πa ⇒ ω2
dω = πa
2 ·
−∞ 0

Replacing ω with x, the result follows.

q 4 sin2 (ωa/2)
EXERCISE 15.1.1. If F f (x) = 2π · = F(ω), where

ω2
(
a − |x| (−a ≤ x ≤ a)
f (x) =
0, elsewhere,
∫ ∞  2 ∫ ∞  4
show that sin x
x dx = π
2 and sin x
x dx = 3π ·
0 0

Department of Mathematics, Vellore Institute of Technology