UPSC Civil Services Main 1979 - Mathematics

Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) If a function f (z) is analytic and bounded in the whole plane, show that
f (z) reduces to a constant. Hence show that every polynomial has a root.

Solution. See 1989, question 2(b) for the first part. See 1996 question 2(a) for the second
part.

Question 1(b) Evaluate the following integrals by the method of residues.

1.
2π
sin2 θ
Z
dθ (a > b > 0)
0 a + b cos θ
2. 1
Z ∞
x 6 log x
dx
0 (1 + x)2

Solution.

1. Let
2π 2π
sin2 θ 1 − cos 2θ
Z Z
1
I= dθ = dθ
0 a + b cos θ 2 0 a + b cos θ
Z 2π
1 1
Let I1 = dθ. Put z = eiθ so that
2 0 a + b cos θ
Z Z
1 dz 1 dz
I1 = =
2 |z|=1 iz(a + 2b (z + z1 )) i |z|=1 bz 2 + 2az + b

1
 √
1 −a + a2 − b2
The integrand has two simple poles at z1 = , z2 =
√ bz 2 + 2az + b b
−a − a2 − b2
. Since a > b > 0, |z2 | > 1, but |z1 z2 | = 1 so |z1 | < 1 i.e. the
b
pole at z = z1 lies within |z| ≤ 1.
z − z1 1 1 1 1
Residue at z1 is lim 2 = = √ . Thus I1 = 2πi √ =
z→z1 bz + 2az + b 2bz1 + 2a 2 a2 − b 2 i 2 a2 − b 2
π
√ . Let
a − b2
2

1 2π cos 2θ
Z 2π
e2iθ dθ
Z
1
I2 = dθ = Re
2 0 a + b cos θ 2 0 a + b cos θ
Z 2
1 1 2z dz
= Re
2 i |z|=1 bz 2 + 2az + b
1 z2
= Re × 2πi Residue of 2 at z = z1
i bz + 2az + b
1 z12
= 2π
b z1 − z2
Thus
2π 2πz12
I1 − I2 = −
b(z1 − z2 ) b(z1 − z2 )
2π
= √ (1 − z12 )
2 a2 − b 2
√
π  a2 − 2a a2 − b2 + (a2 − b2 ) 
= √ 1−
a2 − b 2 b2
 √ √
π a − a2 − b 2 
= √ 2 a2 − b 2
a2 − b 2 b2
2π
Thus I = √ .
a + a2 − b 2
1
γR
log z
2. Let f (z) = z(1+z)
6
2 and the contour C as

shown. γr is a circle of radius r oriented

clockwise, and γR a circle of radius R ori-
A B
ented anticlockwise. AB is along x-axis
on which z = x, CD is the line on which O
z = xe2πi . To avoid the branch point D C
of the multiple valued function log z, we γr
consider C− positive side of the x-axis.
We choose the branch of log z for which
log z = log |z| + iθ, 0 < θ ≤ 2π.

2
 (a) Clearly f (z) has a double pole at z = −1. Residue of f (z) at z = −1 is
1
1 d (z + 1)2 z 6 log z
 

1! dz (z + 1)2 at z=1
 1 
z6 1 5 log z + 6
= + z − 6 log z = 5 at z = eiπ
z 6 at z=−1=e iπ 6z 6

iπ
 
log e + 6 iπ + 6 5π 5π
= 5iπ = cos − i sin
6e 6 6 6 6
 √
√

iπ + 6 3 1 1
= − − i = − (6 + iπ)( 3 + i)
6 2 2 12
(b) On γR , z = Reiθ , |z + 1| ≥ |z| − 1 = R − 1 and | log z| = | log Reiθ | = | log R + iθ| ≤
log R + θ ≤ log R + 2π as 0 ≤ θ ≤ 2π. Thus
Z 1 Z 2π 1 7
z 6 log z R 6 (log R + 2π) R6
2
dz ≤ R dθ = 2π (log R + 2π)
γR (1 + z) 0 (R − 1)2 (R − 1)2
 7 7 
R 6 log R 2πR 6
Clearly lim + = 0, and therefore
R→∞ (R − 1)2 (R − 1)2
Z 1
z 6 log z
lim dz = 0
R→∞ γ (1 + z)2
R

(c) On γr , z = reiθ , |z + 1| ≥ 1 − |z| = 1 − r and | log z| = | log reiθ | = | log r + iθ| ≤

log r + θ ≤ log r + 2π as 0 ≤ θ ≤ 2π. Thus
Z 1 Z 2π 1 7
z 6 log z r 6 (log r + 2π) r6
2
dz ≤ r dθ = 2π (log r + 2π)
γr (1 + z) 0 (1 − r)2 (1 − r)2
 7 7 
r 6 log r 2πr 6
But lim + = 0, and therefore
r→0 (1 − r)2 (1 − r)2
Z 1
z 6 log z
lim dz = 0
r→0 γ (1 + z)2
r

By Cauchy’s residue theorem, using 1, 2, 3, we get

Z ∞ 1 Z 0 1
(xe2πi ) 6 log(xe2πi )
Z
x 6 log x
lim f (z) dz = dx + dx
R→∞ C 0 (1 + x)2 ∞ (1 + x)2
r→0

because on AB, z = x and on CD, z = xe2πi . Therefore

Z ∞ 1 Z ∞ 1 2πi √
x 6 log x x 6 e 6 (log x + 2πi) 2πi
dx − dx = − (6 + iπ)( 3 + i)
0 (1 + x)2 0 (1 + x)2 12
 1 √3  Z ∞ x 16 log x Z ∞ 1 1 √3
x 6 ( 2 + 2 i)2πi πh √ √ i
⇒ − i dx − dx = − −(6 + π 3) + i(6 3 − π)
2 2 0 (1 + x)2 0 (1 + x)2 6

3
Equating real and imaginary parts, we get
1
1 ∞ x 6 log x
Z √ Z ∞ x 16 π √
dx + 3π dx = (6 + π 3) (1)
2 0 (1 + x)2 0 (1 + x)2 6
√ Z ∞ 1 1
3 x 6 log x
Z ∞
x6 π √
− 2
dx − π 2
dx = (π − 6 3) (2)
2 0 (1 + x) 0 (1 + x) 6

√
Multiplying (1) by 3 and adding
1
Z ∞
x 6 log x π √ √  π √ 
− 2
dx = 6 + π 3 + 3π − 18 = 2π 3 − 12
0 (1 + x) 6 6

Thus 1
∞
π2
Z
x 6 log x
dx = 2π − √
0 (1 + x)2 3
√
In addition, multiplying (2) by 3 and adding, we get
1
∞
π √ √ 
Z
x6
2π dx = 6 3 + 3π + π − 6 3
0 (1 + x)2 6
giving us
Z ∞ 1
x6 2π
dx =
0 (1 + x)2 3

4
 UPSC Civil Services Main 1980 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

June 20, 2010

1
Question 1(a) Find the expansion in powers of z of in the region 0 < |z| <
z(z − 1)(z + 3)
4.
Solution. It can easily be seen that
1 1 1 1
f (z) = =− + +
z(z − 1)(z + 3) 3z 4(z − 1) 12(z + 3)
1. Region 0 < |z| < 1.
1 1 −1 1 z −1
f (z) = − − (1 − z) + 1+
3z 4 36 3
∞ ∞
1 1X n 1 X  z n
= − − z + (−1)n
3z 4 n=0 36 n=0 3
∞
1 2 X n  1 (−1)n 
= − − + z − +
3z 9 n=1 4 3n
This is the Laurent expansion of f (z) in the region 0 < |z| < 1. The given function
satisfies the requirements of Laurent’s theorem.
2. Region 1 < |z| < 3.
1 1 1 −1 1 z −1
f (z) = − + 1− + 1+
3z 4z z 36 3
∞ ∞  
1 1 X 1 1 X z n
= − + n
+ (−1)n
3z 4z n=0 z 36 n=0 3
∞ ∞
1 X  z n 1 1X 1
= (−1)n − +
36 n=0 3 12z 4 n=1 z n+1

1
 This is again the Laurent expansion valid in the annular region 1 < |z| < 3.
3. Region |z| > 3
1 1 1 −1 1 3 −1
f (z) = − + 1− + 1+
3z 4z z 3z z
∞ ∞  
1 1 X 1 1 X 3 n
= − + n
+ (−1)n
3z 4z n=0 z 3z n=0 z
∞
1 X 1 1 n−1 n

= + + 3 (−1)
4z n=1 z n+1 4

The is Taylor’s expansion of f (z) around ∞.

Question 1(b) Evaluate by contour integration

Z ∞
dx
1. 4
0 x +1
Z 2π
cos 2θ
2. dθ
0 5 + 4 sin θ
Solution.
1. See 2001 question 2(b).
2. The given integral is the real part of
2π
e2iθ dθ
Z
I=
0 5 + 4 sin θ
Put z = eiθ , dz = ieiθ dθ so that
z2 z 2 dz
Z Z
dz
I= 4 1 = 2
|z|=1 5 + 2i (z − z ) iz |z|=1 5iz + 2z − 2

z2
The integrand 2
has two simple poles, which are given by 2z 2 + 5iz − 2 = 0
5iz + 2z − 2
or 2(z + 2i )(z + 2i) = 0. Out of the two poles z = −2i, − 2i , only z = − 2i is inside the
( 2i )2 i2
unit disc |z| ≤ 1. Residue at this pole is given by i = 4(3i) = 12i . Thus by
2(− 2 + 2i)
Cauchy’s residue theorem
Z 2π
z 2 dz e2iθ
Z
i π
2
= dθ = 2πi = −
|z|=1 5iz + 2z − 2 0 5 + 4 sin θ 12 6

2
Equating real and imaginary parts, we get
Z 2π Z 2π
cos 2θ π sin 2θ
dθ = − , dθ = 0
0 5 + 4 sin θ 6 0 5 + 4 sin θ

3
 UPSC Civil Services Main 1981 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) State and prove Cauchy’s integral formula.

Solution. See 1986 question 1(a).

Question 1(b) Evaluate

Z ∞ −k
x
1. dx, 0 < k < 1.
0 x+1
Z ∞
sin2 x
2. dx
0 x2

Solution.

1. We shall show that for 0 < a < 1

∞
xa−1
Z
π
dx =
0 1+x sin aπ
Now let a − 1 = −k so that a = 1 − k and 0 < a < 1 ⇔ 0 < k < 1. Thus for 0 < k < 1,
Z ∞ −k
x π π
dx = =
0 x+1 sin(1 − k)π sin kπ

We consider first of all C − {positive real axis} i.e. there is a cut along the real axis
for which x ≥ 0 to make log z single valued. We choose that branch of log z for which
log z = log x when z = x, x > 0.

1
 R ∞ a−1 a−1
For 0 x1+x dx we take f (z) = z1+z and γR
the contour C as shown in the figure. γr
is a circle of radius r oriented clockwise, B C
and γR is a circle of radius R oriented an-
ticlockwise. BC is the line joining (r, 0) O
E D
to (R, 0), so is DE. We finally make γr
r → 0 and R → ∞. Note that on BC
z a−1 = xa−1 and on DE z a−1 = (xe2πi )a−1 .

z a−1
(a) Clearly f (z) = has a simple pole at z = −1 inside the contour. Residue at
1+z
z = −1 = eπi of f (z) is (eπi )a−1 . Thus
z a−1
Z
lim dz = 2πi(−eπia )
R→∞ C 1 + z
r→0

Note that z = 0 is excluded by the cut.

(b)
2π
z a−1 Ra−1 eiθ(a−1) Ra−1 R
Z Z
iθ
dz = Rie dθ ≤ 2π
γR 1+z 0 1 + Reiθ R−1
z a−1
Z
Here we use |z + 1| ≥ |z| − 1. Thus lim dz = 0 as a < 1.
R→∞ γ 1 + z
R

(c) Similarly
z a−1 ra
Z
dz ≤ 2π
γr 1 + z 1−r
z a−1
Z
because |z + 1| ≥ 1 − |z|. Thus lim dz = 0.
r→0 γ 1 + z
r

Thus
z a−1 xa−1 xa−1 e2πi(a−1)
Z Z Z
lim dz = lim dx + lim dx
R→∞ C 1+z R→∞ BC 1 + x R→∞ DE 1+x
r→0 r→0 r→0
Z ∞ a−1 Z 0 a−1
x x
= dx + e2πia dx
0 1+x ∞ 1+x

as on BC, z = x and on DE, z = xe2πi . Thus

Z ∞ a−1
x
(1 − e2πia ) dx = −2πieπia
0 1 + x
or ∞
xa−1 eπia −2i
Z
π
dx = −2πi = π =
0 1+x 1 − e2πia e−πia − eπia sin aπ

2
 Alternate proof: This avoids the use of multiple valued functions. In 1991, question
2(c), we proved Z ∞
eax π
x
dx = for 0 < a < 1
−∞ 1 + e sin aπ
Put ex = t, then
Z ∞ Z ∞ a Z ∞ a−1
eax t dt t
x
dx = = dt
−∞ 1 + e 0 1+t t 0 1+t
Z ∞ a−1
t π
Thus dt = , 0 < a < 1.
0Z 1+t sin aπ
∞
sin2 x 1 ∞ sin2 x
Z
2. Clearly dx = dx
0 x2 2 −∞ x2 γR
2 2ix
sin x 1−e
and 2
is the real part of ,
x 2x2
2iz
1−e
therefore we take f (z) = and γr
2z 2
the contour C as shown. Finally we let
R → ∞, r → 0.
A(−R, 0) B(−r, 0) C(r, 0) D(R, 0)

(a) On γR , z = Reiθ and

|1 − e2iz | = |1 − e2i(R cos θ+iR sin θ) | ≤ 1 + |e2i(R cos θ+iR sin θ) | ≤ 2
because |e2iR cos θ | = 1 and |e−2R sin θ | ≤ 1 as sin θ > 0 for 0 < θ < π. Therefore
1 − e2iz
Z
2 π
dz ≤ πR =
γR 2z 2 2R2 R
1 − e2iz
Z
and hence lim dz = 0.
R→∞ γ
R
2z 2
(b) Residue of f (z) at z = 0: Note that z = 0 is a simple pole, so the residue is
1 − e2iz 1 − e2iz −2ie2iz
lim z = lim = lim = −i. Thus
z→0 2z 2 z→0 2z z→0 2
1 − e2iz
Z
lim dz = i(−i)(0 − π) = −π
r→0 γ
r
2z 2
Here we have used the following property: If f (z) has a simple pole at z = a and
γr is a circular arc (part of a circle with center a and radius r), from θ1 to θ2 , then
Z
lim f (z) dz = ia−1 (θ2 − θ1 )
r→0 γr

where a−1 is the residue of f (z) at z = a. See 1985, question 1(c) for more details
and proof.

3
 Thus ∞
1 − e2πiz 1 − e2πix
Z Z
lim dz = dx − π = 0
R→∞ C 2z 2 −∞ 2x2
r→0

as there is no singularity inside C. Taking real parts, we get

Z ∞ Z ∞
sin2 x sin2 x π
2
dx = π =⇒ dx =
−∞ x 0 x2 2

Question 1(c) Obtain the Laurent expansion in powers of z of

1 sinh z
z+ +
z−1 z2
Solution.
1 1
1. z−1
is analytic in the annular region 0 ≤ |z| < 1, so we have the Taylor series for z−1
valid in 0 ≤ |z| < 1. In fact for |z| < 1,
∞
1 −1
X
= −(1 − z) = − zn
z−1 n=0

sinh z
2. z2
has a simple pole at z = 0 and is analytic everywhere else. We have Laurent
series valid in |z| > 0:
∞
sinh z 1 X z 2n+1
=
z2 z 2 n=0 (2n + 1)!
ez + e−z
Note that sinh z = , which gives us the desired expansion.
2
Thus ∞ ∞
1 sinh z X
n 1 X z 2n+1
z+ + = z − z +
z−1 z2 n=0
z 2 n=0 (2n + 1)!
or ∞   X ∞
1 sinh z 1 z X 2n+1 1
z+ + = − 1 + + z − 1 − z 2n
z−1 z2 z 3! n=1 (2n + 3)! n=1

and this expansion is valid in 0 < |z| < 1.

4
 UPSC Civil Services Main 1982 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) Evaluate by contour integration

Z ∞
eax
x
dx, 0 < a < 1
−∞ 1 + e

Solution. See 1991, question 2(c).

Question 1(b) Find the function f (z), holomorphic within the unit circle, which takes the
values
a − cos θ + i sin θ
a2 − 2a cos θ + 1
on the circle.

Solution. See 1997, question 2(c).

Question 1(c) Find by contour integration the value of

Z π
x sin x dx
2
0 a − 2a cos x + 1

if a > 1.

Solution. Note: Even though the question restricts us to a > 1, we shall also consider the
case 0 < a < 1 for completeness.

1
 D(−π, R) C(π, R)
We take f (z) = 1−aez −iz and the con-
tour C is the rectangle ABCD where A =
(−π, 0), B = (π, 0), C = (π, R), D = (−π, R) C
oriented in the anticlockwise direction. We
let R → ∞ eventually. A(−π, 0) B(π, 0)

Clearly f (z) has simple poles at points z given by eiz = a = elog a+2nπi , n ∈ Z. Thus
z = −i log a + 2nπ, n ∈ Z.
Thus when a > 1 i.e. log a > 0, f (z) has no pole in the vertical strip bounded by
x = −π, x = π, y > 0. When 0 < a < 1, f (z) has a simple pole at z = −i log a inside C.
Residue at z = −i log a is given by limz→−i log a (z+i log a)z
1−ae−iz
−i log a
= aie − log a = − log a.

Thus by Cauchy’s residue theorem,

Z (
z 0, when a > 1
lim −iz
dz =
R→∞ C 1 − ae −2πi log a, when 0 < a < 1

1. On CD, z = x + iR, x varies from π to −π.

Z Z −π
z x + iR
−iz
dz = dx
CD 1 − ae π 1 − ae−i(x+iR)

Now |x + iR| ≤ |x| + |R|, |1 − ae−ix eR | ≥ |aeR e−ix | − 1 and therefore

Z π Z π
|x| + |R| π + |R|
Z
z
−iz
dz ≤ R
dx ≤ R
dx
CD 1 − ae −π ae − 1 −π ae − 1

2π(π + |R|) R
Now limR→∞ = 0, so CD
f (z) dz = 0.
aeR − 1
2. Z Z π Z 0 Z π
x dx x dx x dx
f (z) dz = = +
AB −π 1 − ae−ix −π 1 − ae−ix 0 1 − ae−ix
Changing x to −x in the first integral, we get
Z Z π Z π
x dx x dx
f (z) dz = −
AB 1 − ae−ix 0 1 − ae
ix
Z0 π ix −ix
ax(−e + e )
= ix −ix ) + a2
dx
0 1 − a(e + e
Z π
−2iax sin x
= 2
dx
0 1 − 2a cos x + a

2
 3. On BC, z = π + iy and on DA, z = −π + iy, dz = i dy, and
Z Z 
lim f (z) dz + f (z) dz
R→∞ BC DA
Z ∞ Z 0
π + iy −π + iy
= i y−iπ
dy + i y+iπ
dy
0 1 − ae ∞ 1 − ae
Z ∞ 
π + iy −π + iy
= i − dy
0 1 + aey 1 + aey
Z ∞ ∞
e−y −y a+1
= 2πi −y
dy = −2πi log(e + a) = 2πi log
0 e +a 0 a

Thus
(
π
−2iax sin x
Z Z
a+1 0, when a > 1
lim f (z) dz = dx + 2πi log =
R→∞ C 0 1 − 2a cos x + a2 a −2πi log a, when 0 < a < 1

showing that when a > 1,

−2πi log a+1

Z π
x sin x a π  1
dx = = log 1 +
0 1 − 2a cos x + a2 −2ia a a
and when 0 < a < 1,

−2πi log a − 2πi log a+1

Z π
x sin x a π
2
dx = = log(1 + a)
0 1 − 2a cos x + a −2ia a

3
 UPSC Civil Services Main 1983 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) Obtain the Taylor and Laurent series expansions which represent the func-
z2 − 1
tion in the regions
(z + 2)(z + 3)
(i) |z| < 2 (ii) 2 < |z| < 3 (iii) |z| > 3.

Solution. The only singularities of the function are at z = −2 and z = −3.

1. |z| < 2. In this region f (z) is analytic and therefore will have Taylor series. It can be
checked easily using partial fractions that
3 8
f (z) = 1 + −
z+2 z+3
Therefore
 −1  −1
3 z 8 z
f (z) = 1 + 1+ − 1+
2 2 3 3
∞   n ∞  n
3X n z 8X n z
= 1+ (−1) − (−1)
2 n=0 2 3 n=0 3
∞  
1 X 3 8
= − + (−1)n n+1 − n+1 z n
6 n=1 2 3

is the required Taylor series valid in |z| < 2.

1
 2. 2 < |z| < 3: In this case we shall have a Laurent series.
 −1  −1
3 2 8 z
f (z) = 1 + 1+ − 1+
z z 3 3
∞  n ∞  n
3 X
n 2 8 X
n z
= 1+ (−1) − (−1)
z n=0 z 3 n=0 3
∞ ∞
X
n 2n 5 8X zn
= 3 (−1) − − (−1)n n
n=1
z n+1 3 3 n=1 3

This is valid in 2 < |z| < 3.

3. |z| > 3. We have a Taylor series around ∞ given by

−1 −1 ∞
(−1)n
 
3 2 8 3 X
f (z) = 1 + 1+ − 1+ =1+ n+1
(3 · 2n − 8 · 3n )
z z z z n=0
z

Question 1(b) Use the method of contour integration to evaluate

Z ∞ a−1
x
dx, 0<a<2
0 1 + x2

Solution.
za−1
We take f (z) = 1+z 2 and the contour

C as shown in the figure. We choose the Γ

xa−1
branch of z a−1 which results in f (x) = 1+x 2

on the real axis. The only pole of f (z) in-

side C is at z = i. The residue at z = i γ
(z − i)z a−1 ia−1 1 πi a−1
is lim 2
= = (e 2 ) =
z→i 1 + z 2i 2i
1 π(a − 1) π(a − 1) 
cos + i sin . A(−R, 0) B(−r, 0) C(r, 0) D(R, 0)
2i 2 2

Now π
z a−1 Ra−1 πRa
Z Z
dz ≤ R dθ ≤ → 0 as R → ∞ ∵ 0 < a < 2
Γ 1 + z2 0 R2 − 1 R2 − 1
and π
z a−1 ra−1 πra
Z Z
dz ≤ r dθ ≤ → 0 as r → 0 ∵ a > 0
γ 1 + z2 0 1 − r2 r2 − 1

2
Here we use |1 + z 2 | ≥ 1 − |z|2 = 1 − r2 . Thus
Z Z 0 Z ∞
iπ
lim f (z) dz = f (xe )(− dx) + f (x) dx
R→∞ C −∞ 0
r→0
∞ Z ∞ a−1
xa−1 eiπ(a−1)
Z
x
= 2
dx + dx
0 1+x 0 1 + x2
Z ∞ a−1
x

= 1 + cos π(a − 1) + i sin π(a − 1) dx
0 1 + x2
1 π(a − 1) π(a − 1) 
= 2πi · cos + i sin
2i 2 2
Equating the real parts on both sides,
∞
xa−1 π(a − 1)
Z


1 + cos π(a − 1) dx = π cos
0 1 + x2 2
or ∞
xa−1 π(a − 1)
Z
2
dx = π sec
0 1+x 2
Equating the imaginary parts also gives us the same answer.
Alternate solution: In 1984, question 1(b), we obtained
Z ∞ a−1 Z ∞ a−1
2 πa t log t t π2 πa
2 sin 2
dt + π sin πa 2
dt = cos
2 0 1+t 1+t 2 2
Z ∞ a−1 Z0 ∞ a−1 2
t log t t π πa
− sin πa 2
dt − π cos πa 2
dt = sin
0 1+t 0 1+t 2 2
Multiplying the first by cos πa
2
and the second by sin πa
2
and adding gives us

πa
∞ ta−1 π 2  2 πa
Z
πa 2 πa

π sin πa cos − π cos πa sin dt = cos + sin
2 2 1 + t2 2 2 2
Z0
 aπ  ∞ ta−1
π 2
=⇒ π sin aπ − dt =
2 1 + t2 2
Z0 ∞ a−1
t π 1 π π π
=⇒ 2
dt = aπ = π aπ

= sec(a − 1)
0 1+t 2 sin 2 2 cos 2 − 2 2 2

as calculated before.
Note: In this solution the advantage is that we avoid the use of the multiple valued
function log z, however it is much longer.

3
 UPSC Civil Services Main 1984 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Z ∞
x sin mx
Question 1(a) Evaluate by contour integration method dx.
0 x 4 + a4

Solution. See 1998 question 2(b).

∞
xa−1 log x
Z
Question 1(b) Evaluate by contour integration method dx, 0 < a < 2.
0 1 + x2

Solution. D(−R, π) C(R, π)

y=π
zeaz
We take f (z) = and the con-
1 + e2z x = −R x=R
tour C as the rectangle ABCD where A = C
(−R, 0), B = (R, 0), C = (R, π), D =
(−R, π) oriented positively. y=0
A(−R, 0) (0, 0) B(R, 0)

1. On BC, z = R + iy and therefore

Z π Z π
(R + iy)eaR+iay (R + y)eaR π(R + π)eaR
Z
f (z) dz = i dy ≤ dy ≤
BC 0 1 + e2(R+iy) 0 e2R − 1 e2R − 1

because |1 + e2R | ≥ e2R − 1 and R + y ≤ R + π on 0 ≤ y ≤ π.

(R + π)eaR (R + π)aeaR + eaR (R + π)a + 1
Since lim = lim = lim = 0 if 2−a > 0
R→∞ e2R − 1 R→∞ Z 2e2R R→∞ 2e2R−aR
i.e. a < 2, it follows that lim f (z) dz = 0.
R→∞ BC

1
 2.
π π
(−R + iy)e−aR+iay (−R + y)e−aR π(R + π)e−aR
Z Z Z
f (z) dz = i dy ≤ dy ≤
DA 0 1 + e2(−R+iy) 0 1 − e−2R 1 − e−2R
Z
But limR→∞ Re−aR = 0 (note that e−aR 1
≤ a2 R2 ), therefore lim f (z) dz = 0.
R→∞ DA
∞
xeax
Z Z
3. lim f (z) dz = dx.
R→∞ AB −∞ 1 + e2x
−∞
(x + iπ)ea(x+iπ)
Z Z
4. lim f (z) dz = dx as z = x + iπ.
R→∞ CD ∞ 1 + e2x+2iπ

Thus
∞ Z ∞
xeax (x + iπ)ea(x+iπ)
Z Z
lim f (z) dz = dx − dx
R→∞ C −∞ 1 + e2x −∞ 1 + e2x
∞ Z ∞ ax iπa
xeax (1 − eiπa )
Z
e e
= dx − iπ dx
−∞ 1 + e2x −∞ 1 + e
2x

πi
The poles of f (z) are given by e2z = e(2n+1)πi . The only pole in the strip 0 ≤ y ≤ π is z = 2
and it is a simple pole.
πi πia
πi
zeaz (z − πi2
) 2
e 2 πi πia
Residue at z = 2 is limπi 2z
= πi
= − e 2 . Thus
z→ 2 1+e 2e 4
∞ ∞
xeax (1 − eiπa ) eax eiπa
Z Z  
πi πia
dx − iπ dx = 2πi − e 2 (1)
−∞ 1 + e2x −∞ 1 + e2x 4

Equating the real part of both sides, we get

Z ∞ ax Z ∞
xe (1 − cos πa) eax π2 πa
2x
dx + π sin aπ 2x
dx = cos
−∞ 1+e −∞ 1 + e 2 2
Z ∞ ax Z ∞ ax 2
πa xe e π πa
⇒ 2 sin2 dx + π sin πa dx = cos
2 −∞ 1 + e2x −∞ 1 + e
2x 2 2

Putting ex = t so that x = log t, dx = dt/t, we get

Z ∞ a−1 Z ∞ a−1
2 πa t log t t π2 πa
2 sin 2
dt + π sin πa 2
dt = cos (2)
2 0 1+t 0 1+t 2 2

Equating the imaginary parts in (1), we get

Z ∞ Z ∞
xeax eax π2 πa
− sin πa 2x
dx − π cos πa 2x
dx = sin
−∞ 1 + e −∞ 1 + e 2 2

2
Put ex = t as before, to get
Z ∞ a−1 Z ∞ a−1
t log t t π2 πa
− sin πa 2
dt − π cos πa 2
dt = sin (3)
0 1+t 0 1+t 2 2
Multiplying (2) by cos πa and (3) by sin πa and adding we get

∞ ta−1 log t π2
Z  
2 πa 2 πa πa
2 sin cos πa − sin πa dt = cos πa cos + sin πa sin
2 0 1 + t2 2 2 2
2 2
 
π πa π πa
= cos πa − = cos
2 2 2 2
πa πa h πa i
Now 2 sin2 cos πa − sin2 πa = 2 sin2 cos πa − 2 cos2
2 2 h 2
πa πa πa i
= 2 sin2 2 cos2 − 1 − 2 cos2
2 2 2
2 πa
= −2 sin
Z ∞ a−1 2 
2
t log t π πa πa
⇒ 2
dt = − cos 2 sin2
0 1+t 2 2 2
2
π πa πa
= − cot csc
4 2 2
as required.

Question 1(c) Distinguish clearly between a pole and an essential singularity. If z = a is

an essential singularity of a function f (z), prove that for any positive numbers η, ρ,  there
exists a point z such that 0 < |z − a| < ρ for which |f (z) − η| < .

Solution. If f (z) has an isolated singularity at z0 , which is not a removable singularity,

then f (z) has a pole at z = z0 if limz→z0 f (z) = ∞. In this case if f (z) has a pole of order k
at z = z0 , then
∞
X
−k −1
f (z) = a−k (z − z0 ) + . . . + a−1 (z − z0 ) + an (z − z0 )n
n=0

and this Laurent expansion is valid in some deleted neighborhood 0 < |z − z0 | < δ of z0 .
If limz→z0 f (z) does not exist, then f (z) has an essential singularity at z = z0 . (Note
that limz→z0 f (z) is not finite as z0 is not a removable singularity). In this case
∞
X
f (z) = an (z − z0 )n
n=−∞

and a−n 6= 0 for infinitely many n. Again this Laurent expansion is valid in some deleted
neighborhood 0 < |z − z0 | < δ of z0 .

3
 The second part is Casorati-Weierstrass theorem. Let f (z) be analytic in some deleted
neighborhood N of a. Suppose that there exists  > 0 such that |f (z) − η| <  is not satisfied
1
for any z ∈ N i.e. |f (z) − η| ≥  for every z ∈ N . Let g(z) = f (z)−η . Then g(z) is analytic
in N and g(z) is bounded in N , therefore g(z) has a removable singularity at a. Since g(z)
is not constant as f (z) is not constant, either g(a) 6= 0 or g(z) has a zero of order k > 0 at
z = a. This means that f (z) − η is either analytic at z = a or f (z) − η has a pole of order k
at z = a. But this is not true, because f (z) has an essential singularity at z = a. Thus our
assumption is false i.e. we must have z ∈ N for which |f (z) − η| < . Note that we could
take our deleted neighborhood N of the type 0 < |z − a| < δ ≤ ρ.

4
 UPSC Civil Services Main 1985 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) Prove that every power series represents an analytic function within its
circle of convergence.
∞
X
Solution. Let f (z) = an z n have R as its radius of convergence. We shall show that for
n=0
∞
X
0
any z in the region C = {z : |z| < R}, f (z) = nan z n−1 . We first of all note that the
n=1
∞
1
X
radius of convergence of the series nan z n−1 is also R as lim n n = 1.
n→∞
n=1
Let z ∈ C and |z| < ρ < R and let h be chosen so small that |z| + |h| ≤ ρ < R. Thus
(z + h)n − z n
≤ (|z| + |h|)n−1 + |z|(|z| + |h|)n−2 + . . . + |z|n−1 ≤ nρn−1 (1)
(z + h) − z
∞
X
Since the series nan ρn−1 is convergent, given  > 0 ∃N1 > 0 such that
n=1
∞
X 
r|ar |ρr−1 < for n ≥ N1
r=n+1
3
∞
X 
and in particular r|ar |ρr−1 < . (2)
r=N +1
3
1
(z − h)n − z n
 
n−1
Since lim an − nan z = 0, given  > 0 there exists δ > 0 such that
h→0 h
N1 
(z − h)n − z n

X
n−1 
an − nan z < for |h| < δ (3)
n=1
h 3

1
Now
∞
f (z + h) − f (z) X
− nan z n−1
h n=1
N ∞
∞
1
(z + h)n − z n |an (z + h)n − z n |
X   X X
n−1
≤ an − nan z + + |nan z n−1 |
n=1
h n=N +1
h n=N +1
1 1
∞ ∞
 X X
= + |an |nρn−1 + |an |nρn−1 for |h| < δ
3 n=N
1 +1 n=N1 +1
< 
∞
f (z + h) − f (z) X
Thus lim = nan z n−1 = f 0 (z), so f (z) is analytic in C.
h→0 h n=1

Question 1(b) Prove that the derivative of a function analytic in a domain is itself an
analytic function.

Solution. Cauchy’s integral formula states that if f (z) is analytic within and on a simple
closedZ countour C oriented positively and if z0 is any interior point of C, then f (z0 ) =
1 f (z) dz
.
2πi C z − z0
Let f (z) be differentiable in a domain D and z0 ∈ D. Let C be a circle with center
z0 , the boundary of which is positively oriented, such that f (z) is differentiable within
and on C, and Z C along with its interior lies in D. Then by Cauchy’s integral formula,
1 f (z) dz
f (z0 ) = .
2πi C z − z0
Let h ∈ C be so small that z0 + h also lies in the interior of C.
Z  
f (z0 + h) − f (z0 ) 1 f (z) f (z)
= − dz
h 2πih C z − z0 − h z − z0
Z
1 hf (z) dz
=
2πih C (z − z0 − h)(z − z0 )
Z
1 f (z) dz
=
2πi C (z − z0 − h)(z − z0 )
Now
f (z0 + h) − f (z0 )
Z
1 f (z) dz
−
h 2πi C (z − z0 )2
Z  
1 f (z) f (z)
= − dz
2πi C (z − z0 − h)(z − z0 ) (z − z0 )2
Z
1 hf (z) dz
=
2πi C (z − z0 − h)(z − z0 )2

2
Let M = supz∈C |f (z)|, l = length of C, d = minz∈C |z − z0 |, d > 0. Since we are interested
in h → 0, we could have assumed in the beginning itself that 0 < |h| < d. Thus we get
f (z0 + h) − f (z0 ) M |h|l
Z
1 f (z) dz
− 2
≤
h 2πi C (z − z0 ) 2πd2 (d − |h|)
Since the right hand side of the above inequality tends to 0 as h → 0, it follows that
f (z0 + h) − f (z0 )
Z
1 f (z) dz
lim =
h→0 h 2πi C (z − z0 )2
i.e. f (z) is differentiable at z0 and since z0 is an arbitrary point of D, it follows that
Z
0 1 f (ζ) dζ
f (z) =
2πi C (ζ − z)2
where C is any positively oriented circle containing z in its interior.
We shall now prove that Z
00 2! f (z) dz
f (z0 ) =
2πi C (z − z0 )3
where z0 , C are as chosen above. Let h be also chosen as above. Then
f 0 (z0 + h) − f 0 (z0 )
Z
2! f (z) dz
−
h 2πi C (z − z0 )3
Z  
1 1 1 2h
= f (z) − − dz
2πih C (z − z0 − h)2 (z − z0 )2 (z − z0 )3
(z − z0 )3 − (z − z0 − h)2 (z − z0 ) − 2h(z − z0 − h)2
Z
1
= f (z) dz
2πih C (z − z0 − h)2 (z − z0 )3
Now (z − z0 )3 − (z − z0 − h)2 (z − z0 ) − 2h(z − z0 − h)2
= (z − z0 )[(z − z0 )2 − (z − z0 − h)2 ] − 2h[(z − z0 )2 − 2h(z − z0 ) + h2 ]
= (z − z0 )h[2(z − z0 ) − h] − 2h(z − z0 )2 + 4h2 (z − z0 ) − 2h3
= h[2(z − z0 )2 − h(z − z0 ) − 2(z − z0 )2 + 4h(z − z0 ) − 2h2 ]
= h2 [3(z − z0 ) − 2h]
Thus we get
f 0 (z0 + h) − f 0 (z0 ) M |h|(3ρ + 2|h|2 )l
Z
2! f (z) dz
− ≤
h 2πi C (z − z0 )3 2πd3 (d − |h|)2
where M, d, ρ are as before. Since the right hand side of the above inequality tends to 0 as
h → 0, it follows that
f 0 (z0 + h) − f 0 (z0 )
Z
00 2! f (z) dz
f (z0 ) = lim =
h→0 h 2πi C (z − z0 )3
i.e. f 0 (z) is also analytic in D.

3
 Z ∞
x sin ax
Question 1(c) Evaluate by the method of contour integration dx.
0 x 2 − b2
zeiaz
Solution. We take f (z) = z 2 −b2
and the contour C consisting of the following

1. The line AB joining A = (−R, 0) and

B = (−b − r1 , 0).
2. γ1 , the semicircle (x+b)2 +y 2 = r12 lying Γ
in the upper half plane.
3. Line CD joining C = (−b + r1 , 0) and
D = (b − r2 , 0).
4. γ2 , the semicircle (x−b)2 +y 2 = r22 lying
in the upper half plane. γ1 γ2
5. Line EF joining E = (b + r2 , 0) and
F = (R, 0).
6. Γ, the semicircle x2 + y 2 = R2 lying in A(−R, 0) B C D E F (R, 0)
the upper half plane.

Eventually we will let R → ∞, r1 , r2 → 0. Now the integrand has no pole in the upper half
plane, therefore
zeiaz dz
Z
lim =0
R→∞ C (z 2 − b2 )
r1 →0
r2 →0

1. On Γ,
π iθ
zeiaz dz Rei θeiaRe
Z Z
≤ Rieiθ dθ
Γ (z 2 − b2 ) 0
2
R −b 2

because of Γ, |z − b | ≥ |z| − b = R − b2 .
2 2 2 2 2

π
π
zeiaz dz R2 2R2
Z Z Z
2
−aR sin θ
≤ e dθ = 2 e−aR sin θ dθ
Γ (z 2 − b2 ) R 2 − b2 0 R − b2 0

(We can double the integral and halve the limit, because sin(π − θ) = sin θ). Using
Jordan’s inequality sin θ ≥ 2θ
π
for 0 ≤ θ ≤ π2 we get
Z π
zeiaz dz 2R2 2R2 1 − e−aR πR(1 − e−aR )
Z  
2
−aR 2θ
2 2
≤ e π dθ = =
Γ (z − b ) R 2 − b2 0 R2 − b2 2aR/π a(R2 − b2 )

zeiaz dz
Z
showing that lim = 0.
R→∞ Γ (z 2 − b2 )

2.

4
 To get the value of the integral along γ1 , γ2
we observe that if f (z) has a simple pole at γr
z = a and γr is a part of a circle of radius
r with center a, then
Z
lim f (z) dz = ia−1 (θ2 − θ1 ) θ2
r→0 γr

where a−1 is the residue of f (z) at a. θ1

Proof: Let
a−1 a−1
f (z) = + a0 + a1 (z − a) + a2 (z − a)2 + . . . = + φ(z)
z−a z−a
where φ(z) is analytic in the circle |z − a| ≤ r. Thus
Z
φ(z) dz ≤ M r(θ2 − θ1 )
γr
Z
where M = sup|z−a|=r |φ(z)|. Thus lim φ(z) dz = 0 and
r→0 γr
Z Z Z θ2
a−1 dz
lim f (z) dz = =i a−1 dθ = ia−1 (θ2 − θ1 )
r→0 γr γr z−a θ1

zeiaz
Now the residue of 2 at z = b is 21 eiab , and the residue at z = −b is 12 e−iab .
z − b2
Thus limr1 →0 γ1 f (z) dz = 12 ie−iab (0−π) = − iπ2 e−iab and limr2 →0 γ2 f (z) dz = 21 ieiab (0−
R R

π) = − iπ2 eiab .
Using the above data we get
Z ∞
zeiaz dz xeiax dx
Z
iπ iπ
0 = lim 2 2
= 2 2
− e−iab − eiab
R→∞ C (z − b ) −∞ (x − b ) 2 2
r1 →0
r2 →0

or ∞
xeiax dx
Z
= πi cos(ab)
−∞ (x2 − b2 )
Taking imaginary parts, we get
Z ∞
x sin ax dx
= π cos(ab)
−∞ (x2 − b2 )
or Z ∞
x sin ax dx π cos(ab)
2 2
=
0 (x − b ) 2

5
 UPSC Civil Services Main 1986 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) Let f (z) be single valued and analytic within and on a simple closed curve
C. If z0 is any point in the interior of C, then show that
Z
1 f (z) dz
f (z0 ) =
2πi C z − z0
where the integral in taken in the positive sense around C.

Solution. This is known as the Cauchy integral formula. We shall show that given  > 0
Z
1 f (z) dz
− f (z0 ) < 
2πi C z − z0

which implies the result as  is arbitrary.

Since f (z) is analytic at z0 , it is continuous at z0 , therefore given  > 0 as above, there
exists a δ > 0 such that |z − z0 | ≤ δ =⇒ |f (z) − f (z0 )| < . We choose δ > 0 so small that
the disc |z − z0 | < δ lies within the interior of C. Then by Cauchy-Goursat’s theorem (See
1987, 1(b)) we have Z Z
1 f (z) dz 1 f (z) dz
=
2πi C z − z0 2πi γ z − z0
where γ is the circle |z − z0 | = ρ < δ and is positively oriented.
Now put z − z0 = ρeiθ to get
Z 2π
iρeiθ dθ
Z
dz
= = 2πi
γ z − z0 0 ρeiθ

1
Therefore
Z Z Z 
1 f (z) dz 1 f (z) dz dz
− f (z0 ) = − f (z0 )
2πi C z − z0 2πi C z − z0 γ z − z0
f (z) − f (z0 )
Z
1
= dz
2πi γ z − z0
Z
1 
≤  |dz| = length of γ = 
2πρ γ 2πρ
Z
1 f (z) dz
Thus − f (z0 ) = 0 and the proof is complete.
2πi C z − z0
Question 1(b) By the contour integration method show that
Z ∞ √
dx π 2
1. = where a > 0.
0 x 4 + a4 4a2
Z ∞
sin x π
2. dx =
0 x 2
Solution.
1
1. We take f (z) = z4 +a 4 and the contour C Γ
consisting of Γ a semicircle of radius R
with center (0, 0) lying in the upper half
plane, and the line AB joining (−R, 0) and
(R, 0). C is positively oriented. A(−R, 0) (0, 0) B(R, 0)
πi
(a) Poles of f (z) are given by z = ±ae± 4 = ±a[cos π4 ± i sin π4 ], out of which z =
πi πi
ae 4 , z = −ae− 4 are in the upper half plane.
 −1
πi 1 1 1 i
Residue at z = ae 4 is 3πi = −√ + √ .
4a3 e 4 4a3 2 2
 −1
− πi −1 −1 1 i
Residue at z = −ae 4 is 3πi = −√ − √ .
4a 3 e− 4 4a3 2 2
√   √
2 1 1 i 2
Sum of residues is 3 + = − 3 . Thus
4a i − 1 1 + i 4a
√ √
2πi · −i 2
Z
dz 2 2π π
lim 4 4
= 3
= 3
=√
R→∞ C z + a 4a 4a 2a3
(b)
π
Rieiθ
Z Z
dz πR
≤ ≤ 4
Γ z + a4
4 4
0 R −a
4 R − a4
Z
4 4 4 4 4 4 dz
because on Γ |z + a | ≥ |z | − a = R − a . Thus lim = 0.
R→∞ Γ z 4 + a4

2
 Z Z ∞ Z ∞
dz dx dx
(c) lim = =2 .
R→∞ AB z + a4
4
−∞ x + a4
4
0 x4 + a4

Using (a), (b) and (c) we get

Z Z ∞
dz dx π
lim =2 =√
R→∞ C z + a4
4
0 x4 +a 4
2a3
Thus √
Z ∞
dx π π 2
= √ =
0 x 4 + a4 2 2a3 4a3
as required.
iz
2. We take f (z) = ez and the contour C con-
sisting of the line AB joining (−R, 0) to Γ
(−r, 0), the semicircle γ of radius r with
center (0, 0), the line CD joining (r, 0) to
(R, 0) and Γ a semicircle of radius R with γ
center (0, 0). The contour lies in the up-
per half plane and is oriented anticlock-
wise. We took γ as part of the contour to
A(−R, 0) B(−r, 0) C(r, 0) D(R, 0)
avoid the pole at (0, 0).

We will eventually make R → ∞ and r → 0.

eiz
(a) Since the integrand has no poles in the upper half plane, it follows that
z
Z iz
e
lim dz = 0
ZR→∞,r→0 C z
eiz
(b) In order to prove that lim dz = y = sin θ
R→∞ Γ z
0, we use Jordan inequality, which
states that sin θ ≥ 2θ
π
for 0 ≤ θ ≤ π2
— compare the graphs as shown in the y = 2θπ
figure. θ
θ = π2

π
π
eiz e−R sin θ
Z Z Z
2
dz ≤ R dθ = 2 e−R sin θ dθ (∵ sin(π − θ) = sin θ)
Γ z 0 R 0
Z π
2 2θ π
≤ 2 e−R π dθ = 2 · [1 − e−R ]
0 2R
eiz 1 − e−R
Z
showing that lim dz = 0 as lim = 0.
R→∞ Γ z R→∞ R

3
 iz
(c) Now ez = z1 + φ(z) where φ(z) is analytic atR z = 0. Thus given  > 0, ∃δ >
0 such that |z| < δ ⇒ |φ(z)| < . Thus | γ φ(z) dz| ≤  · (length of γ) ⇒
R
limr→0 γ φ(z) dz = 0.
Z Z 0 iθ
dz re i dθ
= = −iπ
γ z π reiθ
(d)
0 ∞
eiz eix eiz eix
Z Z Z Z
lim dz = dx, lim dz = dx
R→∞,r→0 AB z −∞ x R→∞,r→0 CD z 0 x

Using these results, we get

Z iz Z 0 ix Z ∞ ix
e e e
0 = lim dz = dx − iπ + dx (∗)
R→∞,r→0 C z −∞ x 0 x

Since
0 ∞
eix e−iy
Z Z
dx = − dy
−∞ x 0 y
we get
∞ ∞
eix e−ix
Z Z
dx − dx = iπ
0 x0 x
or Z ∞ ix Z ∞
e − e−ix 1 π sin x π
dx = ⇒ dx =
0 2i x 2 0 x 2
Z ∞ ix Z ∞
e sin x
Note that in (*) we cannot write dx = πi and conclude that dx =
Z ∞ Z ∞ −∞ x −∞ x
cos x cos x
π, dx = 0, because dx has convergence problem at x = 0.
−∞ x −∞ x

4
 UPSC Civil Services Main 1987 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

April 12, 2010

1
Question 1(a) By considering the Laurent series for f (z) = prove that if C
(1 − z)(z
Z − 2)
is a closed contour oriented in the counter clockwise direction, then f (z) dz = 2πi.
C

Solution. Laurent’s theorem states that if f (z) is analytic throughout the annular region
R1 < |z − z0 | < R2 and C is any positively oriented simple closed curve lying in the annular
region and having z0 in its interior, then
∞ Z
X
n 1 f (z) dz
f (z) = an (z − z0 ) where an =
n=−∞
2πi C (z − z0 )n+1
Z
Thus f (z) dz = 2πia−1 .
C
Now in our case: R1 = 1, R2 = 2, z0 = 0 i.e. f (z) is analytic in 1 < |z| < 2. Moreover
 −1  −1
1 1 1 1 1 z
f (z) = − = 1− + 1−
z−1 z−2 z z 2 2
∞ ∞
X 1 X zn
Since | z1 | < 1 and | z2 | < 1, we get f (z) = n+1
+ n+1
as the Laurent expansion of
n=0
z n=0
2
f (z) valid in 1 < |z| < 2. Thus
Z if C is a simple closed contour lying in 1 < |z| < 2 with the
origin in its interior, then f (z) dz = 2πi, since a−1 = 1.
C
Note that in this question, the curve C has not been Z clearly specified. If C is in the
region 1 < |z| < 2, but does not contain the origin, then f (z) dz = 0.
C
Note: What about the other cases — C lies in |z| < 1 or |z| > 2

1
Question 1(b) State and prove Cauchy’s residue theorem.

Solution. Statement: If C is a simple closed contour oriented anticlockwise and f (z)

is a complex valued function which is analytic on and within the interior of C except for a
finite number of poles z1 , . . . zn in the interior of C, then
Z n
X
f (z) dz = 2πi Residue of f (z) at z = zr
C r=1

Proof: We enclose each zj , 1 ≤ j ≤ n is a small disc Cj such that Cj along with its
boundary lies within C and, and these discs are small enough that Cj and Ck , j 6= k do not
overlap, or even touch on their boundaries. We now use Cauchy-Goursat theorem:
Theorem (Caucy-Goursat): If C is a simple closed positively oriented contour and
C1 , . . . Cn are simple closed positively oriented contours which lie within C, and whose inte-
riors have no points in common, and if f (z) is a function which is analytic within and on C
except for the interiors of Cj , 1 ≤ j ≤ n, then
Z n Z
X
f (z) dz = f (z) dz
C j=1 Cj

Using this theorem we see that in our case

Z n Z
X
f (z) dz = f (z) dz
C j=1 Cj

But in the last question we have seen that

Z
f (z) dz = 2πi × Residue of f (z) at z = zj
Cj

(This is Laurent’s theorem.) Thus

Z n
X
f (z) dz = 2πi Residue of f (z) at z = zr
C r=1

Question 1(c) By the method of contour integration show that

Z ∞
cos x πe−a
dx = , a>0
0 x 2 + a2 2a

Solution. See 2002, question 2(b).

2
 UPSC Civil Services Main 1988 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Z
dz
Question 1(a) By evaluating over a suitable contour C prove that
z+2
Z π
1 + 2 cos θ
dθ = 0
0 5 + 4 cos θ
Z
dz
Solution. By using the unit circle |z| = 1 as contour, and integrating , we have
|z|=1 z + 2
Z 2π Z 2π
1 + 2 cos θ 1 + 2 cos θ
proved dθ = 0 — see 1997, question 1(b). Now in dθ put
0 5 + 4 cos θ π 5 + 4 cos θ
θ = 2π − φ so that
Z 2π Z 0 Z π
1 + 2 cos θ 1 + 2 cos(2π − φ) 1 + 2 cos φ
dθ = (−dφ) = dφ
π 5 + 4 cos θ π 5 + 4 cos(2π − φ) 0 5 + 4 cos φ
Z 2π Z π
1 + 2 cos θ 1 + 2 cos θ
Thus dθ = 2 dθ showing that
0 5 + 4 cos θ 0 5 + 4 cos θ
Z π
1 + 2 cos θ
dθ = 0
0 5 + 4 cos θ

.
Note: If the contour was not prescribed, we could have put z = eiθ to get
Z 2π
z2 + z + 1
Z
1 + 2 cos θ 1
dθ = dz
0 5 + 4 cos θ i |z|=1 z(5z + 2z 2 + 2)

1
The integrand has two poles at z = 0, z = − 21 inside |z| = 1, which are simple poles. The
residue at z = 0 is 12 and the residue at z = − 21 is − 12 , so we get
2π
z2 + z + 1
Z Z
1 + 2 cos θ
dz = 0 ⇒ dθ = 0
|z|=1 z(5z + 2z 2 + 2) 0 5 + 4 cos θ

Question Z 1(b) If f (z) is analytic in |z| ≤ R and x, y lie inside the disc, evaluate the
f (z) dz
integral and deduce that a function analytic and bounded for all finite
|z|=R (z − x)(z − y)
z is a constant.

Solution. Cauchy’s integral formula states that if f (z) is analytic on and within the disc
|z| ≤ R, then for any ζ which lies within the disc
Z
1 f (z) dz
f (ζ) =
2πi |z|=R ζ − z

Thus
Z Z Z 
f (z) dz 1 f (z) dz f (z) dz 2πi  
= − = f (x) − f (y)
|z|=R (z − x)(z − y) x−y |z|=R z−x |z|=R z−y x−y

We now prove the remaining part, which is Liouville’s theorem.

Let |f (z)| ≤ M for every z. Clearly |z − x| ≥ |z| − |x| = R − |x| and similarly |z − y| ≥
R − |y| on |z| = R, and therefore

M · 2πR
Z
f (z) dz
≤
|z|=R (z − x)(z − y) (R − |x|)(R − |y|)

1 |x − y| · M · 2πR R
Thus |f (x) − f (y)| ≤ . Since → 0 as R → ∞,
2πi (R − |x|)(R − |y|) (R − |x|)(R − |y|)
it follows that |f (x) − f (y)| = 0 or f (x) = f (y), so f is constant.
P∞
Question 1(c) If f (z) = n=0 an z n has radius of convergence R and 0 < r < R, prove
that ∞
Z 2π
1 iθ 2
X
|f (re )| dθ = |an |2 r2n
2π 0 n=0

Solution.
∞
X ∞
X ∞ X
X
|f (z)|2 = f (z) · f (z) = an z n am z m = ap aq z p z q
n=0 m=0 n=0 p+q=n

2
 We know that if a power series has a radius of convergence R, then it is uniformly and
absolutely convergent in |z| ≤ r where 0 < r < R, therefore
Z 2π ∞ Z 2π X
1 X
2 1
|f (z)| dθ = ap aq rp rq ei(p−q)θ dθ
2π 0 n=0
2π 0 p+q=n

R 2π
Since 0
ei(p−q)θ dθ = 0 when p 6= q, we get
Z 2π ∞
1 2
X
|f (z)| dθ = |an |2 r2n
2π 0 n=0
P∞
(Note: This shows that if |f (z)| ≤ M on |z| = r, then n=0 |an |2 r2n ≤ M 2 .)

zez dz
Z
Question 2(a) Evaluate if a lies inside the closed contour C.
C (z − a)3

zez
Solution. Clearly the only pole of is of order 3 at z = a. The residue at this pole
(z − a)3
is
1 d2 (z − a)3 zez
 
1 d z z
 1 z z z

a
 a
= ze + e = ze + e + e = e 1 +
2! dz 2 (z − a)3 z=a 2 dz z=a 2 z=a 2

Thus by Cauchy’s residue theorem,

zez dz
Z
a
 a
3
= 2πi · e 1 + = πiea (2 + a)
C (z − a) 2

Question 2(b) Prove

Z ∞ √
−x2 π −b2
e cos(2bx) dx = e (b > 0)
0 2
2
by integrating e−z along the boundary of the rectangle |x| ≤ R, 0 ≤ y ≤ b.

Solution. D(−R, b) C(R, b)

y=b
Let the rectangle be ABCD where A =
(−R, 0), B = (R, 0), C = (R, b), D = (−R, b) x = −R x=R
2
e−z has no pole in-
oriented positively. Since Z C
2
side ABCD, we get lim e−z dz = 0. y=0
R→∞ ABCD
A(−R, 0) (0, 0) B(R, 0)
3
 2
(Note that e−z has no pole in the entire complex plane.)

1. On BC, z = R + iy and 0 ≤ y ≤ b, therefore

Z Z b Z b
−z 2 −R2 −2Riy −i2 y 2 −R2 2 2
e dz = e e e i dy ≤ e ey dy = (constant)e−R
BC 0 0
Z
2 2
Clearly e−R → 0 as R → ∞, so lim e−z dz = 0.
R→∞ BC

2. On DA, z = −R + iy and 0 ≤ y ≤ b, therefore

Z Z 0 Z b
−z 2 −R2 2Riy −i2 y 2 −R2 2
e dz = e e e i dy ≤ e ey dy
DA b 0
Z
2
Thus lim e−z dz = 0.
R→∞ DA
Z Z ∞ √
−z 2 2
3. On AB, z = x so lim e dz = e−x dx = π.
R→∞ AB −∞

4. On CD, z = x + ib, therefore

Z Z −∞ Z ∞
−z 2 −x2 −i2 b2 −2ibx b2 2
lim e dz = e e e dx = −e e−x [cos 2bx − i sin 2bx] dx
R→∞ CD ∞ −∞

Using the above calculations, we get

Z
√
Z ∞
−z 2 b2 2
0 = lim e dz = π − e e−x [cos 2bx − i sin 2bx] dx
R→∞ C −∞

Equating real and imaginary parts,

Z ∞
2
e−x sin 2bx dx = 0
−∞

and ∞ √ −b2
Z
2
e−x cos 2bx dx = πe
−∞

Thus √ 2
∞ ∞
πe−b
Z Z
−x2 1 −x2
e cos 2bx dx = e cos 2bx dx =
0 2 −∞ 2

4
Question 2(c) Prove that the coefficients cn of the expansion
∞
1 X
= cn z n
1 − z − z2 n=0

satisfy cn = cn−1 + cn−2 , n ≥ 2. Determine cn .

√ √ √
Solution. z 2 + z − 1 = 0 ⇒ z = −1±2 5 . Let λ = −1+2 5 , µ = −1−2 5 . Thus f (z) = 1−z−z
1
2 is

analytic in the disc |z| < λ as both the singularities at z = λ and z = µ lie outside it. Thus
f (z) has Taylor P
series expansion with center zP= 0.
Let f (z) = n=0 cn z , then (1 − z − z ) ∞
∞ n 2 n
n=0 cn z = 1. Equating coefficients of like
powers we get
c0 = 1
c1 − c0 = 0
c2 − c1 − c0 = 0
...
cn − cn−1 − cn−2 = 0
Thus cn = cn−1 + cn−2 , n ≥ 2. The cn ’s are Fibonacci numbers.
Now
−1
f (z) =
(z − λ)(z − µ)
 
−1 1 1
= −
λ−µ z−λ z−µ
  −1  −1 
−1 1 z −1 z
= √ − 1− − 1−
5 λ λ µ µ
If we confine z to the disc |z| < λ, then | λz | < 1, | µz | < 1 and we have
∞ ∞ ∞
1 X zn zn
X  X
f (z) = √ n+1
− n+1
= cn z n
5 n=0 λ n=0
µ n=0

where cn are given as above. But the Taylor series of a function is unique, therefore we have
 
1 1 1
cn = √ −
5 λn+1 µn+1
 n+1  n+1 
1 2 −2
= √ √ − √
5 5−1 5+1
 √ n+1  √ n+1 
1 2( 5 + 1) −2( 5 − 1)
= √ −
5 5−1 5−1
 √ n+1 √ n+1 
1 5+1 n 5−1
= √ + (−1)
5 2 2

5
 UPSC Civil Services Main 1989 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

July 19, 2010

Question 1(a) Evaluate the integral

ez dz
Z
1
2πi C z(1 − z)3
if
1. the point 0 lies inside C and the point 1 lies outside C.

2. both 0 and 1 lie inside C.

3. the point 1 lies inside C and the point 0 lies outside C.

ez
Solution. The only possible poles of are z = 0 and z = 1. Clearly z = 0 is a
z(z − 1)3
zez
simple pole, and residue at z = 0 is lim = 1.
z→0 z(1 − z)3
1 d2 (z − 1)3 ez
 
We have a triple pole at z = 1, and the residue at z = 1 is =
2! dz 2 z(1 − z)3 z=1
1 d ze − ez
 z
1 (z 2 (zez + ez − ez ) − 2z(zez − ez )
  
e
− 2
= − 4
=− .
2 dz z z=1 2 z z=1 2
ez dz
Z
1
By Cauchy’s residue theorem, = Sum of the residues at poles of inte-
2πi C z(1 − z)3
grand within C.
1. The only pole inside C is 0, so
ez dz
Z
1
= Residue at 0 = 1
2πi C z(1 − z)3

1
 2. Both poles are in C, so
ez dz
Z
1 e
3
= Residue at 0 + Residue at 1 = 1 −
2πi C z(1 − z) 2

3. The only pole inside C is 1, so

ez dz
Z
1 e
3
= Residue at 1 = −
2πi C z(1 − z) 2

∞
X
Question 1(b) Let f have the Taylor expansion f (z) = an z n in |z| < R and let sn (z) =
n=0
m
X
ak z k . If 0 < r < R and if |z| < r show that
k=0

f (w) wn+1 − z n+1

Z
1
sn (z) = dw
2πi γ wn+1 w−z

where γ is the circle |w| = r oriented positively.

wn+1 − z n+1
= wn + wn−1 z + . . . + wz n−1 + z n = nk=0 wn−k z k , it follows
P
Solution. Since
w−z
that
 n
f (w) wn+1 − z n+1
Z Z 
1 1 f (w) X n−k k
n+1
dw = n+1
w z dw
2πi γ w w−z 2πi γ w k=0
Z X n 
1 k f (w)
= z dw
2πi γ k=0 wk+1
n Z
1 X k f (w)
= z k+1
dw
2πi k=0 γ w

But from Cauchy’s integral formula we know that

f (k) (0)
Z
1 f (w)
= dw
k! 2πi γ wk+1

Therefore n
f (w) wn+1 − z n+1 X f (k) (0)
Z
1
dw = zk
2πi γ wn+1 w−z k=0
k!

2
 P∞ n
Since f (z) = n=0 an z and the series is uniformly and absolutely convergent within
|z| ≤ r, we can differentiate it termwise. Thus we obtain f (0) = a0 , f 0 (0) = a1 , f 00 (0) =
2a2 , . . . , f (k) (0) = k!ak , . . .. Substituting above, we get
n
f (w) wn+1 − z n+1
Z
1 X
dw = ak z k = sn (z)
2πi γ wn+1 w−z k=0

1
Question 1(c) By integrating around a suitable contour, prove that
1 + z5
Z ∞ 
dx π π
5
= sin
0 1+x 5 5
Solution. We shall present two proofs.
D(−R, 2π
5
) C(R, 2π
5
)
e z
Proof 1: Let f (z) = 1+e and the 2π
5z y= 5
contour be C, the rectangle ABCD where
A = (−R, 0), B = (R, 0), C = (R, 2π 5
), D = x = −R x=R
(−R, 2π
5
) oriented positively. We let R → C
∞ eventually. The only pole in the strip
bounded by y = 0 and y = 2π is z = πi y=0
5 5
and it is a simple pole. A(−R, 0) (0, 0) B(R, 0)
πi πi
πi
(z − πi
5
)ez e5 e5
Residue of f (z) at z = 5
is limπi 5z
= iπ = − . Thus
z→ 5 1+e 5e 5
πi
ez dz
Z
2πie 5
lim = −
R→∞ C 1 + e5z 5
Now we evaluate the integral on all 4 sides of the rectangle.
1. 2π 2π
ez dz eR+iy eR 2π eR
Z Z Z
5 5
= i dy ≤ dy ≤
BC 1 + e5z 0 e5z + 1 0 e5R − 1 5 e5R − 1
5z 5z 5R+5iy
Z |e z + 1| ≥ |e | − 1 = |e
because | − 1 = e5R − 1. as on BC, z = R + iy. Thus
e dz
lim = 0.
R→∞ BC 1 + e5z

2. On DA, z = −R + iy and therefore |e5z + 1| ≥ 1 − |e5z | = 1 − e−5R . This shows that

ez dz 2π e−R
Z
≤
DA 1 + e
5z 5 1 − e−5R
e−R ez dz
Z
As → 0 as R → ∞, it follows that lim = 0.
1 − e−5R R→∞ DA 1 + e5z

3
 3. On AB, z = x so
∞
ez dz ex dx
Z Z
lim =
R→∞ AB 1 + e5z −∞ 1 + e5x
2πi
4. On CD, z = x + 5
, so
−∞ 2πi
ez dz ex e 5 dx
Z Z
lim =
R→∞ CD 1 + e5z ∞ 1 + e5x

Using the above, we get

∞ ∞ πi
ez dz ex dx ex dx
Z Z Z
2πi 2πie 5
lim = −e 5 =−
R→∞ C 1 + e5z −∞ 1+e 5x
−∞ 1+e 5x 5
or πi
∞
ex dx
Z 
2πi e 5 π 2i π π
5x
= − 2πi = πi πi = sin
−∞ 1 + e 5 1−e 5 5e5 −e5 5 5
Z ∞ 
dt π π
We now put ex = t to get 5
= sin as desired.
0 1+t 5 5
1 2π
Proof 2: Let f (z) = 1+z 5 and the con- B(Rei 5 )
tour be C, the angular region OABO where
OA is the line joining (0, 0), (R, 0), AB is the
arc of the circle |z| = R and B is on the circle
such that angle ∠AOB = 2π 5
. C is oriented
C
positively. We let R → ∞ eventually. The
only pole in the sector is z = πi 5
and it is a 2π
simple pole. 5 y=0
O(0, 0) A(R, 0)
Using Cauchy’s residue theorem, we get
πi
z−e5
Z
dz πi 2πi
lim = 2πi × Residue at e 5 = 2πi lim = 4πi
R→∞ C 1 + z5 z→e 5 1 + z
πi 5
5e 5
2π
1. On AB, z = Reiθ , |z 5 + 1| ≥ |z|5 − 1 = R5 − 1, 0 ≤ θ ≤ 5
and therefore
2π
Rieiθ dθ
Z Z
dz 5 2π R
≤ ≤
AB 1 + z5 0
5
R −1 5 R5 − 1
Z
dz
showing that lim = 0.
R→∞ AB 1 + z 5
Z Z ∞
dz dx
2. On OA, z = x and therefore lim = .
R→∞ OA 1 + z5 0 1 + x5

4
 2πi
3. On BO, z = Re 5 and R varies from ∞ to 0. Therefore
Z Z 0 2πi Z ∞
dz e 5 dR 2πi dR
lim = 2πi = −e 5
R→∞ BO 1 + z 5 ∞ 1 + (Re 5 ) 5 0 1 + R5

Thus
Z Z ∞ Z ∞
dz dx 2πi dR 2πi
lim = −e 5 = 4πi
R→∞ C 1 + z5 0 1+x 5
0 1+R 5
5e 5
∞ πi
e− 5
Z
dx 2πi 1 2πi
⇒ 5
= 4πi 2πi = 4πi
− πi πi
0 1+x 5e 5 1 − e 5 5e 5 e 5 − e 5

π 2i π π
= πi πi = sin
5 e5 −e 5 − 5 5
Note: We have provided both proofs because sometimes the examiner prescribes the contour.

Question 2(a) Let f (z) = ∞ n

P
n=0 an z be analytic for |z| < 1 + δ, (δ > 0). Prove that the
polynomial pk (z) of degree k which minimizes the integral
Z 2π
1
|f (eiθ ) − pk (eiθ )|2 dθ
2π 0
is pk (z) = n=0 an z n . Prove that the minimum value is given by ∞
Pk P 2
n=k+1 |an | .

Solution. On |z| = 1, z = eiθ and

Z 2π Z 2π ∞
X
f (z)f (z) dθ = an am ei(n−m)θ dθ
0 0 n,m=0

Now termwise integration is justified because the series ∞ n

P
n=0 an z is uniformly convergent
in |z| ≤ 1 as the given series is convergent in |z| < 1 + δ with δ > 0. Thus
Z 2π ∞ Z 2π ∞
1 2 1 X i(n−m)θ
X
|f (z)| dθ = an am e dθ = |an |2
2π 0 2π n,m=0 0 0
R 2π
as 0
ei(n−m)θ dθ = 0 or 2π according as n 6= m or n = m.
Let pk (z) = kn=0 bn z n , then as above
P

Z 2π k ∞
1 2
X
2
X
|f (z) − pk (z)| dθ = |an − bn | + |an |2
2π 0 n=0 n=k+1

Clearly the right hand side is minimum if and only if kn=0 |an − bn |2 = 0 ⇒ an = bn for
P

sum are non-negative. Thus pk (z) = kn=0 an z n and the

P
n = 1, . . . , k, as all terms in the P
minimum value of the integral is ∞ 2
n=k+1 |an | .

5
Question 2(b) If f is regular in the whole plane and the values of f (z) do not lie in the
disc with center w0 and radius δ, show that f is constant.

Solution. Liouville’s Theorem: If f (z) is entire, i.e. regular in the whole plane, and
bounded, then f (z) is constant.
1
Consider the function F (z) = f (z)−w 0
. Since f (z) is entire and f (z) 6= w0 (note that if
f (z) = w0 for some z then one of its values would lie inside the disc with center w0 and
radius δ). it follows that F (z) is an entire function. Since |f (z) − w0 | > δ for every z,
|F (z)| < 1δ for every z, thus by Liouville’s theorem F (z) ≡ c a constant, and therefore f (z)
is a constant.
Proof of Liouville’s theorem: From Cauchy’s integral formula, we have for any z0 and ρ
however large Z
0 1 f (z) dz
f (z0 ) =
2πi |z−z0 |=ρ (z − z0 )2
Now f (z) is bounded, say |f (z)| ≤ M and |z − z0 | = ρ, so let z − z0 = ρeiθ , dz = ρieiθ dθ
which gives us
M M
|f 0 (z0 )| ≤ 2
2πρ =
2πρ ρ
Letting ρ → ∞, we get f 0 (z0 ) = 0 for any z0 , thus f 0 (z) = 0 so f is a constant.

1
Question 2(c) Find the singularities of sin( 1−z ) in the complex plane.

1 1
Solution. Since 1−z is analytic everywhere except z = 1, sin( 1−z ) is regular everywhere
1
except z = 1. At z = 1 the function has an essential singularity — Clearly sin( 1−z )=0⇔
1 1 1
1−z
= nπ, n 6= 0 ⇔ z = 1 − nπ , n ∈ Z, n 6= 0. Thus 1 is a limit point of zeros of sin( 1−z ) and
1
therefore sin( 1−z ) has an essential singularity at z = 1.
1 ζ
Note that sin( 1−z ) is regular at ∞ as sin( 1−ζ ) is regular at ζ = 0.

6
 UPSC Civil Services Main 1990 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

March 5, 2010

Question 1(a) Let f be regular for |z| < R, prove that, if 0 < r < R,
Z 2π
0 1
f (0) = u(θ) exp(−iθ) dθ
πr 0

where u(θ) = Re f (reiθ ).

Solution. Using Cauchy’s integral formula, it is easily deduced that for any z in the interior
of {CR : |z| = R}, we have

f (t) (z)
Z
1 f (ζ)
= dζ
t! 2πi CR (ζ − z)t+1
Z
0 1 f (ζ)
In particular, f (0) = dζ.
2πi CR ζ 2
Putting ζ = Reiθ , dζ = Rieiθ dθ, we get
Z 2π Z 2π
0 1 f (Reiθ ) 1
f (0) = iθ
Rie dθ = f (Reiθ )e−iθ dθ (1)
2πi 0 R2 e2iθ 2πR 0
We now consider the integral

f (ζ)ζ t−1
Z
1
dζ
2πi CR (R − Rz ζ)t+1

By Cauchy’s residue theorem, the above integral is equal to 2πi(sum of residues of the
2
integrand within CR ). If t ≥ 1, the only possibility of a pole could be at the point ζ = Rz ,

1
 2 2 2
but |z| = |z| < R, therefore | Rz | > RR = R, so Rz lies outside CR and hence the integrand
has no pole inside CR , so
f (ζ)ζ t−1
Z
1
dζ = 0 for t ≥ 1
2πi CR (R − Rz ζ)t+1
In particular, taking t = 1, z = 0,
Z
1 f (ζ)
dζ = 0
2πi CR R2
Thus we get
Z 2π
1
0 = f (Reiθ )eiθ dθ
2πR 0
Z 2π
1
⇒0 = f (Reiθ )e−iθ dθ (2)
2πR 0
Adding (1), (2), we get
Z 2π Z 2π
0 1 iθ iθ −iθ 1
f (0) = (f (Re ) + f (Re ))e dθ = u(θ) exp(−iθ) dθ
2πR 0 πR 0
as required.
Note 1: To get the desired form, we could have considered the integral over {Cr : |z| =
r < R} instead of CR and in that case ζ = reiθ and instead of R, we would have got r i.e.
Z 2π
0 1
f (0) = u(θ) exp(−iθ) dθ
πr 0
Note 2: The integral
f (ζ)ζ t−1
Z
1
dζ
2πi CR (R − Rz ζ)t+1
plays an important role in questions of this type, and has to be kept in mind.
Question 1(b) Prove that the distance from the origin to the nearest zero of f (z) = ∞ n
P
n=0 an z
r|ao |
is at least where r is any number not exceeding the radius of convergence of the
M + |a0 |
series, and M = M (r) = sup|z|=r |f (z)|.
Solution. By Cauchy’s integral formula,
Z Z
1 f (ζ) dζ 1 f (ζ) dζ
f (z) − f (0) = −
2πi |ζ|=r ζ − z 2πi |ζ|=r ζ
where |z| < r ≤ R, R is the radius of convergence. If f (z) = 0, then
Z 2π
rieiθ dθ M |z|
Z
1  1 1 M
|f (0)| ≤ M − dζ ≤ |z| iθ
=
2π |ζ|=r ζ − z ζ 2π 0 re (r − |z|) r − |z|
|f (0)|r
because |ζ −z| ≥ |ζ|−|z| = r−|z| on |ζ| = r. Thus r|f (0)| ≤ |z|(M +|f (0)|) ⇒ |z| ≥ M +|f (0)|
.
Here f (0) = a0 , and the result follows.

2
Question 1(c) If f = u + iv is regular throughout the complex plane, and au + bv − c ≥ 0
for suitable constants a, b, c then f is constant.

Solution. Theorem: If f (z) = u + iv is entire, and u ≤ 0, then f is constant.

Proof: Consider F (z) = ef (z) , then F (z) is also entire. Moreover

|F (z)| = |eu+iv | = |eu | ≤ 1 ∵ u ≤ 0

Thus F (z) is entire and bounded, hence is a constant by Liouville’s theorem. Now F 0 (z) =
f 0 (z)ef (z) = 0 ⇒ f 0 (z) = 0 because ef (z) 6= 0, so f (z) is constant.
Corollary: If f (z) = u + iv is entire, and u ≥ 0, then f is constant. Proof: Consider
−f (z) = −u − iv, then −u ≤ 0 and −f (z) is constant.
Now consider F (z) = (a − ib)f (z) − c = (au + bv − c) + i(av − bu). Now F (z) is entire,
and Re F (z) = au + bv − c ≥ 0, so F (z) is constant, hence f (z) is constant.

∞
x4 dx
Z
π π
Question 2(a) Prove that = √ sin using residue calculus.
−∞ 1 + x8 2 8

Solution.
4
We take f (z) = 1+z z
8 and the contour C
γ
consisting of γ a semicircle of radius R with
center (0, 0) lying in the upper half plane, and
the line joining (−R, 0) and (R, 0). Finally we
will let R → ∞.
(−R, 0) (0, 0) (R, 0)
By Cauchy’s residue theorem
Z ∞ 4
z 4 dz
Z Z 4
x dx z dz
lim 8
= 8
+ lim
R→∞ C 1 + z −∞ 1 + x R→∞ γ 1 + z 8

= 2πi(sum of residues at poles of f (z) in the upper half plane)

Now π
z 4 dz R4 e4iθ Rieiθ dθ πR5
Z Z
≤ ≤
γ 1 + z8 0 R8 − 1 R8 − 1
because |z 8 + 1| ≥ |z 8 | − 1 = R8 − 1 on |z| = R. Therefore
Z 4
z dz
lim =0
R→∞ γ 1 + z 8

(2n+1)πi
f (z) has poles at zeros of z 8 + 1 = 0 ⇒ z 8 = −1 ⇒ z 8 = e(2n+1)πi ⇒ z = e 8 , n ∈ Z.
πi 3πi 5πi 7πi
Clearly z = e 8 , e 8 , e 8 , e 8 are the only poles of f (z) in the upper half plane and all these

3
 z04 1
are simple poles. The residue at any simple pole z0 is 8z07
= 8z03
,

sum of residues at poles of f (z) in the upper half plane

1 −3πi/8
+ e−9πi/8 + e−15πi/8 + e−21πi/8


= e
8
1 −3πi/8
− e−πi/8 + eπi/8 − e3πi/8


= e
8
1 π 3π

= 2i sin − 2i sin
8 8 8
i π π

= sin − cos
4√ 8 8
i 2 π π π π

= cos sin − cos sin
4 4 8 8 4
i π
= − √ sin
2 2 8
Thus ∞
x4 dx
Z
i π π π
8
= 2πi(− √ sin ) = √ sin
−∞ 1+x 2 2 8 2 8
as required.

Question 2(b) Derive a series expansion of log(1 + ez ) in powers of z.

Solution. Let f (z) = log(1 + ez ), then
ez 1 z 2 1 z 1
f 0 (z) = = e2 z z = e2
1+e z −
2 e2 + e 2 2 cosh z2
Let g(z) = cosh z2 , then (
1
2n
sinh z2 , n odd
g (n) (z) = 1
2n
cosh z2 , n even
1
In particular, g (n) (0) = 0 when n is odd, and g (n) (0) = 2n
when n is even. Moreover
z 1 z
f 0 (z) cosh = f 0 (z)g(z) = e 2
2 2
Using Leibnitz rule for the derivative of the product of two functions, we get
z n  
dn  1 z  e2 X n (n−p)
n
e 2 = n+1 = g (z)f (p+1) (z)
dz 2 2 p=0
p

Thus when z = 0, we get

n  
(
X n n−p (p+1) 1 0, n odd
n−p
f (0) = n+1 where n =
p=0
p 2 2 1, n even

4
and therefore
n−1  
n+1 (n+1)
X n
2 f (0) = 1 − 2p+1 n−p f (p+1) (0)
p=0
p

Case (1) : When n is even

  n−2  
n+1 (n+1) n 0
X n p+1
2 f (0) = 1 − 2f (0) − 2 n−p f (p+1) (0)
0 p=1
p

Note that odd p do not contribute anything to the summation, as n−p = 0 for odd p. Now we
can see by induction that f (n) (0) = 0 whenever n is odd and n > 1. f 0 (0) = 12 . 23 f (3) (0) =
1 − 2 · 12 = 0. Assume by induction hypothesis that f (3) (0) = f (5) (0) = . . . = f (2m−1) (0) = 0,
then letting n = 2m in the above formula,
m−1
X 
2m+1 (2m+1) 2m 2p+1 (2p+1)
2 f (0) = − 2 f (0) = 0
p=1
2p

Case (2): When n is odd: The terms with even p in the formula above do not make any
contribution. Thus letting n = 2m + 1,
m−1
X  m  
2m+2 (2m+2) 2m + 1 2r+2 (2r+2) X 2m + 1 2r (2r)
2 f (0) = 1 − 2 f (0) = 1 − 2 f (0) (∗)
r=0
2r + 1 r=1
2r − 1

We can now see that f 00 (0) = 41 , f (4) (0) = − 18 , f (6) (0) = 41 .

Thus
z 11 2 11 4 11 6
log(1 + ez ) = log 2 + + z − z + z + ...
2 4 2! 8 4! 4 6!
∞
z X f (2n) (0)z 2n
= log 2 + +
2 n=1 (2n)!

where f (2n) (0) is given by (∗) for n ≥ 1.

Note: We now present an alternative solution, where we use Leibnitz rule for the n-th
derivative of the quotient of two functions. It is a good exercise in itself and is usually
missing from textbooks.
Theorem: Let y = uv , where u, v are functions with derivatives up to order n. Then

v 0 0 ... u
v1
v 0 ... u1
1 2
yn = v2 1 1
v v ... u2
v n+1
... .
. . .
. . ... ...
n n
vn 1 n−1
v v
2 n−2
... un

5
 dn y
Here the determinant is (n + 1) × (n + 1), and yn = n .
dx
Proof: vy = u, therefore, by taking successive derivatives using Leibnitz product rule
we get
vy = u
v1 y + vy1 = u1
v2 y + 2v1 y1 + vy2 = u2
... ...
n


vn y + 1 vn−1 y1 + . . . + vyn = un
These are n + 1 equations in n + 1 unknowns y, y1 , . . . , yn , and the determinant of the
coefficient matrix is v n+1 . Thus by Cramer’s rule
v 0 0 ... u
v1 v
0 ... u1
1 2
yn = v2 1 1
v v ... u2
v n+1
...
. . .
. . . ... ...
n n
vn 1 n−1
v v
2 n−2
... un
as required.
ez
Now f (z) = log(1 + ez ), f (0) = log 2. f 0 (z) = 1+e 0 1 z z
z , f (0) = 2 . Let u = e , v = 1 + e .

Then un (0) = 1 for every n, and v(0) = 2, vn (0) = 1 for n ≥ 1. Let F (z) = uv , then
2 0 0 ... 0 1
1 2 0 ... 0 1
1
F (n) (0) = f (n+1) (0) = 1 2 2 ... 0 1
2n+1 ...
n

n

..
.
n

..
.
1 1 2
... n−1
1

1 2 1 1
F (1) (0) = f (2) (0) = =
4 1 1 4
2 0 1
1
F (2) (0) = f (3) (0) = 1 2 1 = 0
8
1 2 1
2 0 0 1 2 0 0 1
1 1 2 0 1 1 −1 2 0 0 −2 1
F (3) (0) = f (4) (0) = = = =−
16 1 2 2 1 16 −1 2 2 0 16 8
1 3 3 1 −1 3 3 0
2 0 0 0 1
1 2 0 0 1
1
F (4) (0) = f (5) (0) = 1 2 2 0 1 = 0
32
1 3 3 2 1
1 4 6 4 1
Thus log(1 + ez ) has the expansion as given above.

6
  1 
Question 2(c) Determine the nature of singular points of sin and investigate its
cos z1
behavior at z = ∞.

Solution.
 1 
1. Let ζ = z1 , and φ(ζ) = f ( ζ1 ) = sin . Therefore lim φ(ζ) = sin 1, showing that
cos ζ ζ→0
φ(ζ) has a removable singularity at ζ = 0. In fact φ(ζ) is analytic at ζ = 0 if φ(0) is
defined to be sin 1. Note that

φ(ζ) − φ(0) sin( cos1 ζ ) − sin 1  1 

lim = lim = lim cos sec ζ tan ζ = 0
ζ→0 ζ ζ→0 ζ ζ→0 cos ζ
 1 
Thus sin is regular at ∞.
cos z1

1 2  1 
2. At all zeros of cos z i.e. z = the function sin has essential singu-
(2n + 1)π cos z1
larities because limx→∞ sin x does not exist — if it did, then given  > 0, we would
have N such that x1 > N, x2 > N ⇒ | sin x1 − sin x2 | < . But for any N we can take
x1 = 2nπ + π2 > x2 = 2nπ > N , then | sin x1 − sin x2 | = 1 6<  if  < 1.

3. z = 0 is also an essential singularity of the given function as it is a limit point of

2
essential singularities z = .
(2n + 1)π

7
 UPSC Civil Services Main 1991 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

March 5, 2010

Question 1(a) A function f (z) is defined for all finite values of z by f (0) = 0 and f (z) =
−4
e−z everywhere else. Show that the Cauchy-Riemann equations are satisfied at the origin.
Show also that f (z) is not analytic at the origin.

Solution. Let f (z) = u + iv. By definition

∂u u(x, 0) − u(0, 0) ∂u u(0, y) − u(0, 0)

(0, 0) = lim , (0, 0) = lim
∂x x→0 x ∂y y→0 y
−4 −4
Now u(x, 0) = Re f (x, 0) = e−x , u(0, y) = Re f (0, y) = e−(iy) , therefore
−4 −4
∂u e−x − 0 4 ∂u e−y
(0, 0) = lim = lim te−t = 0, (0, 0) lim =0
∂x x→0 x t→∞ ∂y y→0 y
4 4 4
(Note that et > t4 ⇒ te−t < t13 ⇒ limt→∞ te−t = 0).
−4
It is obvious that v(x, 0) =Imaginary part of e−x = 0, and v(0, y) =Imaginary part
−4
of e−(iy) = 0, and therefore vx (0, 0) = vy (0, 0) = 0. Thus ∂u ∂x
∂v
(0, 0) = ∂y (0, 0), ∂u
∂y
(0, 0) =
∂v
− ∂x (0, 0), i.e. the Cauchy-Riemann equations are satisfied at (0, 0).
However f (z) is not analytic at z = 0 because it is not even continuous at z = 0: if we take
iπ iπ −4 πi −4
z = re 4 , then z → 0 ⇔ r → 0, but limr→0 f (re 4 ) = limr→0 e−r e = limr→0 er = ∞, so
limz→0 f (z) 6= f (0).

Question 1(b) If |a| =

6 R, show that

|dz|
Z
2πR
< 2
|z|=R |(z − a)(z + a)| |R − |a|2 |

1
 Solution. On |z| = R, z = Reiθ , 0 ≤ θ ≤ 2π, |dz| = |Rieiθ dθ| = R dθ. |z 2 − a2 | ≥
|z|2 −|a|2 = R2 −|a|2 and |z 2 −a2 | ≥ |a|2 −|z|2 = |a|2 −R2 , showing that |z 2 −a2 | ≥ |R2 −|a|2 |,
with the strict inequality occurring when a = |a|eiθ , z 6= Reiθ . Thus
Z 2π
|dz|
Z
R dθ 2πR
< 2 2
= 2
|z|=R |(z − a)(z + a)| 0 |R − |a| | |R − |a|2 |

as required.

Question 1(c) If Z 2π
1
Jn (t) = cos(nθ − t sin θ) dθ
2π 0
show that
t 1 1 1 1
e 2 (z− z ) = J0 (t) + zJ1 (t) + z 2 J2 (t) + . . . − J1 (t) + 2 J2 (t) − 3 J3 (t) + . . .
z z z
t 1
Solution. The function f (z) = e 2 (z− z ) is analytic in 0 < |z| < ∞ and therefore by
Laurent’s theorem — If f (z) is analytic is the annular region D : R1 < |z − z0 | < R2 and
if C is any positively oriented simple closed contour lying within the region D, then for any
z ∈ D, we have
∞
X ∞
X
f (z) = n
an (z − z0 ) + bn (z − z0 )−n
n=0 n=1

where Z Z
1 f (z) dz 1 f (z) dz
an = , bn =
2πi C (z − z0 )n+1 2πi C (z − z0 )−n+1
or ∞ Z
X 1
n f (z) dz
f (z) = cn (z − z0 ) where cn =
n=−∞
2πi C (z − z0 )n+1
t 1
If we take f (z) = e 2 (z− z ) , R1 = 0, R2 = ∞, z0 = 0, then
∞ t 1
e 2 (z− z ) dz
Z
t
(z− z1 )
X
n 1
f (z) = e 2 = cn z where cn =
n=−∞
2πi C z n+1

We now take C as |z| = 1. Then z = eiθ and

Z 2π t (eiθ −e−iθ )
1 e2
cn = ieiθ dθ
2πi 0 ei(n+1)θ
Z 2π
1
= eti sin θ e−inθ dθ
2π 0
Z 2π Z 2π
1 i
= cos(nθ − t sin θ) dθ + sin(−nθ + t sin θ) dθ
2π 0 2π 0

2
 Z 2π Z 2π
But sin(−nθ + t sin θ) dθ = 0, because if we put θ = 2π − η, then sin(−nθ +
0 Z 0 0

t sin θ) dθ = sin(−nη + t sin η) dη. Therefore

2π
Z 2π
1
cn = cos(nθ − t sin θ) dθ = Jn (t)
2π 0

Hence ∞
t 1
X
e 2 (z− z ) = Jn (t)z n
n=−∞

Since on replacing z by − z1 , the function f (z) remains unaltered, we get J−n (t) = Jn (t) if n
is even, and J−n (t) = −Jn (t) if n is odd. Thus
t 1 1 1 1
e 2 (z− z ) = J0 (t) + zJ1 (t) + z 2 J2 (t) + . . . − J1 (t) + 2 J2 (t) − 3 J3 (t) + . . .
z z z
as required.

Question 2(a) Examine the nature of the singularity of ez at ∞.

Solution. ez has an essential singularity at ∞. We examine the nature of the singularity

1 1
of e ζ at ζ = 0. Taking ζ = n1 , lim e ζ = lim en = ∞.
ζ→0 n→∞
1
−n
Taking ζ = − n1 , lim e ζ = lim e = 0.
ζ→0 n→∞
1 1
Thus lim e ζ does not exist and therefore e ζ has an essential singularity at ζ = 0, proving
ζ→0
that ez has an essential singularity at ∞.
∞
1 X 1 1 1
Alternately e =
ζ is the Laurent expansion of e ζ having infinitely many negative

n=0
n! ζ n
powers, showing the same result.

z3
Question 2(b) Evaluate the residues of the function at all its singu-
(z − 2)(z − 3)(z − 5)
larities and show that their sum is 0.

Solution. The given function has simple poles at z = 2, 3, 5.

z 3 (z − 2) 8
Residue at z = 2 is lim = .
z→2 (z − 2)(z − 3)(z − 5) 3
z 3 (z − 3) 27
Residue at z = 3 is lim =− .
z→3 (z − 2)(z − 3)(z − 5) 2
z 3 (z − 5) 125
Residue at z = 5 is lim = .
z→5 (z − 2)(z − 3)(z − 5) 6

3
 Residue at ∞ is = − coefficient of z1 in the expansion of f (z) around ∞.
 −1  −1  −1
2 3 5
f (z) = 1− 1− 1−
z z z
  
2 1 3 1
= 1 + + Higher powers of 1 + + Higher powers of
z z z z
 
5 1
1 + + Higher powers of
z z
10 1
= 1+ + Higher powers of
z z
Thus the residue at ∞ is −10.
16 − 81 + 125 − 60
Sum of all residues is = 0.
6
Note: The function f (z) has no singularity as such at ∞, but the residue at ∞ is always
defined as such. The function is actually analytic at ∞ as f ( z1 ) is analytic at z = 0.
Question 2(c) By integrating along a suitable contour show that
Z ∞
eax π
x
dx =
−∞ 1 + e sin aπ
where 0 < a < 1.
D(−R, 2π) C(R, 2π)
Solution.
y = 2π
eaz
Our f (z) = 1+ez
and the contour is C, the
rectangle ABCD where A = (−R, 0), B = x = −R x=R
(R, 0), C = (R, 2π), D = (−R, 2π) oriented in C
the anticlockwise direction. We let R → ∞
eventually. y=0
A(−R, 0) (0, 0) B(R, 0)

The function f (z) has only a simple pole at z = πi in the strip bounded by y = 0 and
z − πi az eaπi
y = 2π. Residue of f (z) at πi is lim e = πi = −eaπi .
z→πi 1 + ez Z e
eaz dz
Thus by Cauchy’s residue theorem lim = −2πieπia .
R→∞ C 1 + ez
We now evaluate the integral along the four lines.
1. On the line BC i.e. x = R, z = R + iy, dz = i dy and
Z 2π a(R+iy) Z 2π aR
eaz dz 2πeaR
Z
e i dy e dy
z
= ≤ =
BC 1 + e 0 1 + eR+iy 0 eR − 1 eR − 1
eaR
because |ez + 1| ≥ |ez | − 1 = |eR+iy | − 1 = eR − 1 on BC. Since limR→∞ eR −1
= 0 as
eaz dz
Z
0 < a < 1 using L’Hospital’s Rule, it follows that lim = 0.
R→∞ BC 1 + ez

4
 2. On the line DA i.e. x = −R, z = −R + iy, dz = i dy. Since |ez + 1| ≥ 1 − |ez | =
1 − |e−R+iy | = 1 − e−R

eaz dz 2πe−aR
Z
≤ (→ 0 as R → ∞)
DA 1 + e
z 1 − e−R

eaz dz
Z
thus lim = 0.
R→∞ DA 1 + ez

3. On the line AB, z = x, so

∞
eaz dz eax
Z Z
lim = dx
R→∞ AB 1 + ez −∞ 1 + ex

4. On the line CD, z = x + 2πi, so

Z −∞ a(x+2πi) Z ∞
eaz dz eax
Z
e 2πia
lim = dx = −e dx
R→∞ CD 1 + ez ∞ 1 + ex+2πi −∞ 1 + e
x

Thus ∞
eaz dz eax
Z Z
lim = (1 − e2πia ) dx = −2πieπia
R→∞ C 1 + ez −∞ 1 + ex
so ∞
eax −2πieπia −2πi
Z
π
x
dx = 2πia
= −πia πia
=
−∞ 1+e 1−e e −e sin aπ
as required.

5
 UPSC Civil Services Main 1992 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

March 1, 2010

Question 1(a) If u = e−x (x sin y − y cos y), find v such that f (z) = u + iv is analytic. Also
find f (z) explicitly as a function of z.

Solution. See 1993, question 2(b).

Question 1(b) Let f (z) be analytic inside and on the circle C defined by |z| = R and let
reiθ be any point inside C. Prove that
Z 2π
iθ 1 (R2 − r2 )f (Reiφ )
f (re ) = dφ
2π 0 R2 − 2Rr cos(θ − φ) + r2

Solution. By Cauchy’s integral formula

Z
iθ 1 f (ζ)
f (z) = f (re ) = dζ (1)
2πi CR :|ζ|=r ζ − z

f (ζ)
We note that the function 2 has no singularity within and on CR , because f (ζ) is
ζ − Rz
2 2
analytic within and on CR and (ζ − Rz )−1 is also analytic within and on CR as Rz lies
2
outside CR and therefore ζ − Rz 6= 0 (Note that R2 = R · R > R|z|, because |z| = r < R,
2
thus | Rz | > R. Thus by Cauchy’s theorem
Z
f (ζ)
0= R2
dζ (2)
CR ζ − z

1
Using (1), (2) we get
Z  
1 1 1
f (z) = f (ζ) − dζ
2πi |ζ|=r ζ − z ζ − Rz2
2
z − Rz
Z  
1
= f (ζ) 2 dζ
2πi |ζ|=r (ζ − z)(ζ − Rz )
zz − R2
Z  
1
= f (ζ) dζ
2πi |ζ|=r (ζ − z)(ζz − R2 )
Z 2π
r 2 − R2
 
1
⇒ f (reiθ ) = iφ
f (Re ) Reiφ i dφ
2πi 0 (Reiφ − reiθ )(rRei(φ−θ) − R2 )
Z 2π
r 2 − R2
 
1 iφ
= f (Re ) dφ
2π 0 (R − rei(θ−φ) )(rei(φ−θ) − R)
Z 2π
r 2 − R2
 
1 iφ
= f (Re ) dφ
2π 0 −R2 − r2 + rR(ei(θ−φ) + ei(φ−θ) )
Z 2π
R2 − r 2
 
1 iφ
= f (Re ) 2 dφ
2π 0 R + r2 + 2rR cos(θ − φ)

as required.

Question 1(c) Prove that all the roots of z 7 − 5z 3 + 12 = 0 lie between the circles |z| = 1
and |z| = 2.

Solution. See 2006 question 2(b).

(−1)n−1 z 2n−1
Question 2(a) Find the region of convergence of the series whose n-th term is .
(2n − 1)!

Solution. Clearly

Coefficient of the (n + 1)-th term (2n − 1)!

= → 0 as n → ∞
Coefficient of the n-th term (2n + 1)!
1
Thus lim |Coefficient of the n-th term| n = 0. So the radius of convergence of the power
n→∞
∞
X (−1)n−1 z 2n−1
series is ∞, i.e. the region of convergence is the entire complex plane.
n=1
(2n − 1)!

2
 1
Question 2(b) Expand f (z) = in a Laurent series valid for (i) |z| > 3, (ii) 1 <
(z + 1)(z + 3)
|z| < 3, (iii) |z| < 1.
Solution. (i) |z| > 3.
   
1 1 1 1  1 −1  3 −1
f (z) = − = 1+ − 1+
2 z+1 z+3 2z z z
Since | z1 | < 13 , | z3 | < 1, we have
∞ ∞
1 X (−1)n X (−1)n 3n

f (z) = −
2z n=0 z n n=0
zn
∞
1 X (−1)n (1 − 3n )
=
2z n=0 zn
∞
X (−1)n (1 − 3n ) 1
=
n=0
2 z n+1

(ii) 1 < |z| < 3.

1 1 −1 1 1  z −1
f (z) = 1+ − 1+
2z z 23 3
Since | z1 | < 1, | z3 | < 1, we get
∞ ∞
1 X (−1)n 1 X (−1)n z n
f (z) = −
2z n=0 z n 6 n=0 3n
∞ ∞
1 X (−1)n X (−1)n+1 z n

= +
2 n=0 z n+1 n=0
3n+1

(iii) |z| < 1.

1 −1 1 1  z −1
f (z) = 1+z − 1+
2 23 3
As |z| < 1, | z3 | < 1, we get
∞ ∞
1X 1 X (−1)n z n
f (z) = (−1)n z n −
2 n=0 6 n=0 3n
∞
1X  1 
= (−1)n 1 − n+1 z n
2 n=0 3
These are the Laurent or Taylor series in the required three cases.
Z ∞
cos mx
Question 2(c) By integrating along a suitable contour evaluate dx
0 x2 + 1
Solution. See 1995, question 2(a).

3
 UPSC Civil Services Main 1993 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

March 1, 2010

1
Question 1(a) In the finite plane, show that the function f (z) = sec has infinitely many
z
isolated singularities in a finite interval which includes zero.

1 1 π 2
Solution. We know that cos = 0 if and only if = (2n + 1) , n ∈ Z, or z = .
z z 2 (2n + 1)π
1
Moreover all these zeros are simple zeros of cos and are isolated singular points. Thus
z
2
the given function has infinitely many simple poles at the points z = . Since
(2n + 1)π
2
→ 0 as n → ∞, it follows that any finite interval containing 0 will have all but
(2n + 1)π
2
finitely many points of the type z = . Thus any finite interval containing 0 will
(2n + 1)π
1
have infinitely many isolated singularities (simple poles) of sec .
z

Question 1(b) Find the orthogonal trajectories of the family of curves in the xy-plane
defined by e−x (x sin y − y cos y) = α, where α is a real constant.

Solution. If f (z) = u + iv is an analytic function, then u =constant, v =constant repre-

sent families of curves which are orthogonal to each other, because of the Cauchy-Riemann
equations:      
∂u ∂u ∂v ∂v
− × − = −1
∂x ∂y ∂x ∂y
i.e. tangents to the curve u = c and v = c0 respectively cut each other at right angles. Thus
given u = ex (x sin y − y cos y) we have to find v so that f = u + iv is analytic.

1
 We use Milne Thompson’s method. We know
   
0 ∂u z + z z − z ∂u z + z z − z
f (z) = , −i ,
∂x 2 2i ∂y 2 2i
is an identity.
f 0 (z) = e−x sin y − e−x (x sin y − y cos y) − ie−x (x cos y − cos y + y sin y)
Putting z = z ⇒ x = z, y = 0 ⇒ f 0 (z) = −ie−z (z − 1), or f (z) = −i e−z (z − 1) dz = ize−z .
R

Thus
u + iv = i(x + iy)e−x (cos y − i sin y)
= e−x (x sin y − y cos y) + ie−x (x cos y + y sin y)
Thus v = e−x (x cos y + y sin y) = β is the required family of curves.

Question 1(c) Prove by applying Cauchy’s integral formula or otherwise that

Z 2π
1 · 3 · 5 · . . . · (2n − 1)
cos2n θ dθ = 2π
0 2 · 4 · 6 · . . . · (2n)
where n = 1, 2, 3, . . ..
Solution. We put z = eiθ so that dz = ieiθ dθ and the integral is along the curve |z| = 1.
We get
2n
z + z1
Z 2π
(1 + z 2 )2n
Z Z
2n dz 1
cos θ dθ = = dz
0 |z|=1 22n iz i |z|=1 22n z 2n+1
R 2π 1
Thus by Cauchy’s residue theorem 0 cos2n θ dθ = 2πi 2n (sum of residues at poles of
2 i
(1 + z 2 )2n
inside |z| = 1). Clearly z = 0 is the only pole of the integrand in |z| = 1, and it
z 2n+1
is of order 2n + 1.
(1 + z 2 )2n
Residue of 2n+1
at z = 0 is the coefficient of z1 in the Laurent expansion.
z
Now (1 + z 2 )2n =sum of powers of z with exponent < 2n + 2n

2n
n
z + sum of powers with
(1 + z 2 )2n
exponent > 2n. Thus coefficient of z1 in the Laurent expansion of around z = 0
z 2n+1
is 2n


n
. Thus
Z 2π
1 2n!
cos2n θ dθ = 2π 2n
0 2 n!n!
2n!
Now 2n n! = 2n(2n − 2)(2n − 4) . . . · 6 · 4 · 2, so n = (2n − 1)(2n − 3)(2n − 5) . . . · 5 · 3 · 1.
2 n!
Hence Z 2π
1 · 3 · 5 · . . . · (2n − 1)
cos2n θ dθ = 2π
0 2 · 4 · 6 · . . . · (2n)
as required.

2
 3 2
Question 2(a) If Z C is the curve y = x − 3x + 4x − 1 joining the points (1, 1) and (2, 3),
find the value of (12z 2 − 4iz) dz.
C

Solution. If C1 is any curve joining (1, 1) and (2, 3), then C and C1 form a closed contour.
Since 12z 2 − 4iz is analytic, by Cauchy’s theorem
Z Z
2
− (12z − 4iz) dz + (12z 2 − 4iz) dz = 0
C C1

so the integral is independent of the path between (1, 1) and (2, 3). Thus
Z  2+3i
2 3 2
(12z − 4iz) dz = 4z − 2iz
C 1+i
= 4[(2 + 3i) − (1 + i)3 ] − 2i[(2 + 3i)2 − (1 + i)2 ]
3

= 4[8 + 36i − 54 − 27i − 1 − 3i + 3 + i] − 2i[4 + 12i − 9 − 1 − 2i + 1]

= 4[−44 + 7i] − 2i[−5 + 10i] = −156 + 38i
R R
Note that calculating C by C udx + vdy would be more work.

∞
X zn
Question 2(b) Prove that the series converges absolutely for |z| ≤ 1.
n=1
n(n + 1)

∞
X
1 1 1
Pn 1
Solution. Consider the series an where an = n(n+1)
= n
− n+1
. Now sn = r=1 ( r −
n=1
∞
1 1
X zn
r+1
) = 1− n+1
. Thus sn → 1 as n → ∞, so an is convergent. Now ≤ an
n=1
n(n + 1)
∞
X zn
for |z| ≤ 1, therefore by Weierstrass M-test, the series converges absolutely (in
n=1
n(n + 1)
fact uniformly) in the region |z| ≤ 1.
Z ∞
dx
Question 2(c) Evaluate by choosing an appropriate contour.
0 x6 + 1

Solution.
1
We take f (z) = 1+z 6 and the contour Γ
γ consisting of Γ a semicircle of radius R
with center (0, 0) lying in the upper half
plane, and the line joining (−R, 0) and (R, 0).
(−R, 0) (0, 0) (R, 0)

3
 Z
dz
By Cauchy’s residue theorem = 2πi(sum of residues at poles of f (z) in the upper
γ 1 + z6
half plane).
1 πi πi 5πi
Clearly 1+z 6 has simple poles at z = e
6 , z = e 2 and z = e 6 inside the contour.
1
Residue at z = ζ is 5 .
6ζ
 
1 1 1 1
Sum of residues = + 15πi + 25πi
6 e 5πi6 e 6 e 6
1  5π 5π π π  −2i
= cos − i sin − i + cos − i sin =
6 6 6 6 6 6
−i
Z
dz 2π
as cos 5π
6
= − cos π6 , sin 5π
6
= sin π6 = 12 . Thus lim 6
= 2πi = .
R→∞ γ 1 + z 3 3
Now Z Z π
dz R πR
6
≤ 6
dθ = 6
Γ 1+z 0 R −1 R −1
iθ 6 6
Z z = Re and using |z + 1| ≥ R − 1 on Γ.
on putting
dz
Thus 6
→ 0 as R → ∞. Consequently,
Γ 1+z
Z Z ∞
dz dx 2π
lim = =
R→∞ γ 1 + z 6 −∞ 1 + x
6 3
Z ∞
dx π
and hence 6
= .
0 1+x 3

4
 UPSC Civil Services Main 1994 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

March 1, 2010

Question 1(a) Suppose that z is the position vector of a particle moving on the ellipse
C : z = a cos ωt+ib sin ωt where ω, a, b are positive constants, a > b and t is time. Determine
where

1. the velocity has the greatest magnitude.

2. the acceleration has the least magnitude.

Solution. See 1996, question 1(a).

Question 1(b) How many zeroes does the polynomial p(z) = z 4 + 2z 3 + 3z + 4 possess (i)
in the first quadrant, (ii) in the fourth quadrant.

Solution.

1. p(−1) = 0. p(−2) = −2 < 0, p(−3) = 22 > 0, therefore the intermediate value theorem
shows that there exists x, −3 < x < −2 such that p(x) = 0. Thus we have determined
that to zeros of p(z) lie on the negative real axis, and since p is a polynomial of degree
4 and hence has 4 zeros, we are left with the task of locating the the remaining two
zeros.

2. p(z) has no zeros on the positive real axis because p(x) > 0 when x ≥ 0.

3. p(z) has has no zero on the imaginary axis because p(iy) = y 4 + 4 − 2iy 3 + 3iy = 0 ⇒
y 4 + 4 = 0, 2y 3 − 3y = 0, but y 4 + 4 = 0 has no real zeros, so p(iy) 6= 0.

1
 B(0, R)
We now consider the contour OABO
where OA is straight line joining (0, 0) and
(R, 0), AB is the arc of the circle x2 +y 2 = R2
in the first quadrant, and BO is the line join-
ing (0, R) to (0, 0). O(0, 0) A(R, 0)
1
By the Argument Principle, the number of zeros of p(z) in the first quadrant = 2π ×
(the change in the argument of p(z) when z moves along the contour OABO oriented anti-
clockwise as R → ∞).
Change in the argument along OA: On OA, p(z) = x4 + 2x3 + 3x + 4 > 0 ⇒ arg p(z) = 0
for every x on OA. Therefore as z moves from O to A, the change in the argument of p(z)
i.e. ∆OA arg p(z) = 0.
Change in the argument along BO: On BO, z = iy and p(z) = y 4 + 4 + i(3y − 2y 3 ).
3
−1 3y − 2y

Therefore arg p(z) = tan .
y4 + 4
3 0

−1 3y − 2y

∆BO arg p(z) = tan =0−0=0
y4 + 4 ∞

Change in argument along AB: On arc AB, z = Reiθ , 0 ≤ θ ≤ π2 , so that

h 2 3 4 i
p(z) = R4 e4iθ + 2R3 e3iθ + 3Reiθ + 4 = R4 e4iθ 1 + + + −→ R4 e4iθ
Reiθ R3 e3iθ R4 e4iθ
π
as R → ∞. Thus ∆AB arg p(z) = 4θ 02 = 2π.1
Hence ∆OABO arg p(z) = 2π as R → ∞, so p(z) has exactly one zero in the first quadrant.
Since p(z) is a polynomial with real coefficients, it follows that if ζ is a zero of p(z) and
it lies in the first quadrant, then ζ is also a zero of p(z) and it lies in the fourth quadrant.
Thus p(z) has one zero in each of the first and the fourth quadrants.

Question 1(c) Test for uniform convergence in the region |z| ≤ 1 the series
∞
X cos nz
n=1
n3

Solution. By definition

einz + e−inz e−ny einx + eny e−inx

cos nz = =
2 2
 
1
Alternately, p(z) = z 4 1 + z2 + z33 + z44 = z 4 (1 + w) where w = 2
z + z33 + z44 . Clearly w → 0 as R → ∞.
Therefore |1+w−1| <  for |z| large. This means 1+w remains inside a circle of radius 1 as z moves along AB
and R → ∞. Therefore ∆AB arg(1 + w) = 0 and ∆AB p(z) = ∆AB z 4 + ∆AB (1 + w) = 4∆AB z = 4 · π2 = 2π.

2
and therefore ∞ ∞ ∞
X cos nz X e−ny einx X eny e−inx
= +
n=1
n3 n=1
2n3 n=1
2n3
Case 1: y > 0.
∞ ∞
X e−ny einx X 1
3
≤
n=1
2n n=1
2n3
showing that the first term is absolutely convergent.
eny e−inx
But the second term is not convergent, because its n-th term 6→ 0 as n → ∞
2n3
eny e−inx
— in fact → ∞ as n → ∞ when y > 0.
2n3
∞
X cos nz
Therefore is not even convergent when y > 0.
n=1
n3
Case 2: y < 0. This case is entirely analogous to the above case — the first term
∞ ∞
X e−ny einx X cos nz
3
is not convergent, so 3
is not convergent.
n=1
2n n=1
n
∞
X cos nx
Case 3: y = 0. 3
is uniformly and absolutely convergent, because of Weierstrass
n=1
n
M-test, which states that if ∞
P
n=1 fn (z) P
is a series and there exist positive constants Mn such
that |fn (z)| < Mn for every z ∈ Ω and n Mn is convergent, then ∞
P
n=1 n (z) is absolutely
f
1
and uniformly convergent in Ω. Here Mn = n3 for all x.
Thus the given series converges uniformly only on the real axis in |z| ≤ 1.

Question 2(a) Find the Laurent series for

e2z
1. about z = 1.
(z − 1)3
1
2. about z = 3.
z 2 (z − 3)2

Solution.

1. The function e2z is analytic everywhere in the complex plane. The Taylor series of e2z
with center z = 1 is given by
∞ dn e2z ∞
X 2n e2
2z
X
dz n
at z = 1
e = (z − 1)n = (z − 1)n
n=0
n! n=0
n!

3
 dn e2z
because = 2n e2z . Thus
dz n
∞
e2z e2 2e2 4e2 X 2n e2
3
= 3
+ 2
+ + (z − 1)n−3
(z − 1) (z − 1) (z − 1) 2!(z − 1) n=3 n!
∞
e2 2e2 4e2 X 2n+3 e2
= + + + (z − 1)n
(z − 1)3 (z − 1)2 2!(z − 1) n=0 (n + 3)!

e2z
which is the required Laurent series of with center z = 1. It is valid in the
(z − 1)3
ring 1 < |z| < ∞.
1
2. Let f (z) = z2
then
2 (−2)(−3) (−2)(−3) . . . (−n − 1)
f 0 (z) = − 3
, f 00 (z) = 4
, . . . , f (n) (z) =
z z z n+2
and therefore
1 0 2 (n) (−1)n (n + 1)!
f (3) = , f (3) = − , . . . , f (3) =
32 33 3n+2
Thus the Taylor series of f (z) with center z = 3 is given by
∞
X (−1)n (n + 1)! ∞
X (−1)n (n + 1)
1 n
2
= n+2
(z − 3) = n+2
(z − 3)n
z n=0
3 n! n=0
3

Thus
∞ ∞
1 X (−1)n (n + 1) n−2 1 2 X (−1)m (m + 3)
2 2
= n+2
(z−3) = 2 2
− 3 + m+4
(z−3)m
z (z − 3) n=0
3 3 (z − 3) 3 (z − 3) m=0 3
1
is the required Laurent series of z 2 (z−3)2
with center z = 3 valid in 0 < |z| < 3.

Question 2(b) Find the residues of f (z) = ez csc2 z at all its poles in the finite plane.

Solution. The poles are at zeros of sin2 z, and sin2 z = 0 iff z = nπ, n ∈ Z, the set of
integers. All these poles are double poles.
1 d (z − nπ)2 ez
 
Residue at z = nπ of f (z) is . Now
1! dz sin2 z z=nπ

d (z − nπ)2 ez sin2 z[(z − nπ)2 ez + 2(z − nπ)ez ] − (z − nπ)2 ez 2 sin z cos z

 
=
dz sin2 z sin4 z
ez (z − nπ)

= 3 (z − nπ) sin z + 2 sin z − 2(z − nπ) cos z
sin z

4
 z − nπ 1
Using lim = = (−1)n , we get
z→nπ sin z cos nπ
2 z
 
d (z − nπ) e (z − nπ)
= enπ lim


2 3 (z − nπ) sin z + 2 sin z − 2(z − nπ) cos z
dz sin z z=nπ
z→nπ sin z

(z − nπ)(sin z − 2 cos z) + 2 sin z

= enπ (−1)n lim
z→nπ sin2 z
sin z − 2 cos z + (z − nπ)(cos z + 2 sin z) + 2 cos z
= enπ (−1)n lim
z→nπ 2 sin z cos z
sin z + (z − nπ)(cos z + 2 sin z)
= enπ lim
z→nπ 2 sin z
cos z + cos z + 2 sin z + (z − nπ)(− sin z + 2 cos z)
= enπ lim
z→nπ 2 cos z
= enπ
Thus the residue at z = nπ of ez csc2 z is enπ .
∞
(loge u)2
Z
Question 2(c) By means of contour integration evaluate du.
0 u2 + 1
Solution.
2
We take f (z) = (log z)
z 2 +1
and the contour
C consisting of the line joining (−R, 0) to Γ
(−r, 0), the semicircle γ of radius r with cen-
ter (0, 0), the line joining (r, 0) to (R, 0) and
Γ a semicircle of radius R with center (0, 0). γ
The contour lies in the upper half plane and
is oriented anticlockwise. We have avoided
the branch point z = 0 of the multiple valued
A(−R, 0) B(−r, 0) C(r, 0) D(R, 0)
function log z.

(Eventually we shall let R → ∞, r → 0).

(1) On Γ, z = Reiθ and |1 + z 2 | ≥ |z|2 − 1 = R2 − 1. Thus

2
log(Reiθ )
Z Z π
f (z) dz ≤ 2−1
iReiθ dθ
Γ R
Z 0π
| log R + iθ|2
≤ R dθ
0 R2 − 1
Z π
R 2 2 R  2 π3 
= ((log R) + θ ) dθ = π(log R) +
R2 − 1 0 R2 − 1 3
3
But R2R−1 π(log R)2 + π3 → 0 as R → ∞, therefore


Z
lim f (z) dz = 0
R→∞ Γ

5
 (2) On γ, z = reiθ , |z|2 + 1 ≥ 1 − |z|2 = 1 − r2 . Thus
Z 0
(log r)2 + θ2 π3 
Z
r  2
f (z) dz ≤ r dθ = π(log r) +
γ π 1 − r2 1 − r2 3
Z
But the right side → 0 as r → 0, it follows that lim f (z) dz = 0.
r→0 γ
(3) f (z) has a simple pole at z = i in the the upper half plane (inside C) and the residue
(log i)2 1  πi 2 π 2 i
at z = i of f (z) is = = . Thus
2i 2i 2 8
Z R Z r
π2i
Z
iπ iπ
lim f (z) dz = lim f (x) dx + f (xe ) dxe = 2πi
R→∞,r→0 C R→∞,r→0 r R 8

because on the line CD, z = x, and on the line AB, z = xeiπ . Hence
Z 0 Z ∞
(log(xeiπ ))2 (log x)2 π3
− 2 2πi
dx + dx = −
∞ 1+x e 0 1 + x2 4

Now (log(xeiπ ))2 = (log x)2 − π 2 + 2iπ log x, so

Z ∞ Z ∞ Z ∞
(log x)2 2 dx log x π3
2 dx − π + 2iπ dx = −
0 1 + x2 0 1 + x2 0 1 + x2 4
Z ∞ i∞ π
dx −1
Equating real parts, and noting that = tan x = , we get
0 1 + x2 0 2
Z ∞
(log x)2 π3 π3 π3
2 dx = − =
0 1 + x2 2 4 4
Z ∞
(log x)2 π3
so that dx = .
0 1 + x2 8 Z ∞
log x
Note that by equating imaginary parts, we get dx = 0.
0 1 + x2

6
 UPSC Civil Services Main 1995 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

February 24, 2010

Question 1(a) Let u(x, y) = 3x2 y + 2x2 − y 3 − 2y 2 . Prove that u is a harmonic function.
Find a harmonic function v such that u + iv is an analytic function of z.

Solution. Clearly
∂u ∂u
= 6xy + 4x , = 3x2 − 3y 2 − 4y
∂x ∂y
∂ 2u ∂ 2u
= 6y + 4 , = −6y − 4
∂x2 ∂y 2

∂ 2u ∂ 2u
Thus + = 0, showing that u is a harmonic function.
∂x2 ∂y 2
Let f (z) = u + iv, where v is to be so determined that f (z) is analytic and v is harmonic.
∂u
Such a function v along with u would have to satisfy the Cauchy-Riemann equations =
∂x
∂v ∂u ∂v
, = − . Now
∂y ∂y ∂x
∂u ∂v ∂u ∂u
f 0 (z) = +i = −i
∂x ∂x ∂x ∂y
= 6xy + 4x − i(3x − 3y 2 − 4y)
2

= −3i(x2 − y 2 + 2ixy) + 4(x + iy)

= −3iz 2 + 4z

1
Thus

f (z) = 2z 2 − iz 3
= 2(x + iy)2 − i(x + iy)3
= 2x2 − 2y 2 + 4ixy − ix3 + 3x2 y + 3ixy 2 − y 3
= 3x2 y + 2x2 − y 3 − 2y 2 + i(4xy − x3 + 3xy 2 )

Thus v = 4xy − x3 + 3xy 2 . Clearly

∂v ∂v
= 4y − 3x2 + 3y 2 , = 4x + 6xy
∂x ∂y
∂ 2v ∂ 2v
= −6x , = 6x
∂x2 ∂y 2

∂ 2v ∂ 2v
so that + = 0, showing that v is a harmonic function.
∂x2 ∂y 2

z
Question 1(b) Find the Taylor series expansion of f (z) = around z = 0. Find also
z4 + 9
the radius of convergence.

Solution. It is obvious that

z z 4 −1 z z4 z8 z 12 
f (z) = 1+ = 1− + − + ...
9 9 9 9 81 729
∞ 4 ∞
zX  z n X
 z 4n+1
= (−1)n = (−1)n n+1
9 n=0 9 n=0
9
4
provided | z9 | < 1. This indeed is Taylor’s series representation of f (z) which to start
X∞
4
with is valid for | z9 | < 1. The radius of convergence of a power series an z n is given
n=0
−1 1 
 1
  1  4n+1 −1
by lim sup |an | n . In this case the radius of convergence is lim n+1 =
n→∞ 9
n+1 1 √
lim 9 = 9 = 3.
4n+1 4
n→∞
Note: We did not get the radius of convergence √ greater than the disc of validity namely
4
| z9 | < 1 as we have a singularity of f (z) on |z| = 3, namely those z for which z 4 = −9 = i2 9
or z 2 = ±3i.

Question 1(c) Let C be a circle |z| = 2 oriented counter-clockwise. Evaluate the integral
Z
cosh πz
2
dz with the aid of residues.
C z(z + 1)

2
 Z
cosh πz
Solution. By Cauchy’s residue theorem, dz = 2πi(sum of residues at poles of
C z(z 2 + 1)
cosh πz
inside C).
z(z 2 + 1)
cosh πz
The only poles of are at z = 0, ±i all within |z| = 2. All these are simple poles.
z(z 2 + 1)
z cosh πz
Residue at z = 0 is lim = 1.
z→0 z(z 2 + 1)
(z − i) cosh πz cosh πi cos π 1
Residue at z = i is lim 2
= =− = .
z→i z(z + 1) i · 2i 2 2
(z + i) cosh πz cosh(−πi) 1
Residue at z = −i is lim = = .
Z z→−i z(z 2 + 1) (−i) · (−2i) 2
cosh πz h 1 1i
Thus 2
dz = 2πi 1 + + = 4πi.
C z(z + 1) 2 2
Z ∞
cos ax
Question 2(a) Evaluate the integral dx, a ≥ 0.
0 x2 + 1
Solution.
eiaz
Let f (z) = . Let γ be the con- Γ
z2 + 1
tour consisting of the line joining (−R, 0) and
(R, 0) and Γ, which is the arc of the circle of
radius R and center (0, 0) lying in the upper
half plane. γ is oriented counter-clockwise. (−R, 0) (0, 0) (R, 0)
Z ∞ iax
eiaz
Z Z
e
lim f (z) dz = 2
dx + lim dz
R→∞ γ −∞ x + 1 R→∞ Γ z 2 + 1
iθ
Since |z 2 + 1| ≥ R2 − 1 on Γ and |eiaz | = |eiaRe | = |e−aR sin θ | ≤ 1 because sin θ ≥ 0 in
0 ≤ θ ≤ π, so
eiaz
Z
πR
2
dz ≤ 2
Γ z +1 R −1
Z iaz
e
as dz = iReiθ dθ, showing that lim 2
dz = 0.
R→∞ Γ z + 1
R
By Cauchy’s residue theorem, limR→∞ γ f (z) dz = 2πi( Sum of residues at poles of
iaz
e
2
in the upper half plane). z = i is the only pole of f (z) in the upper half plane, and
z +1
(z − i)eiaz e−a
the residue there is given by lim = .
Z ∞ iax z→i z2 + 1 2i
e
Thus 2
dx = πe−a , so
−∞ x + 1
Z ∞ Z ∞
cos ax −a sin ax
2
dx = πe , 2
dx = 0
−∞ x + 1 −∞ x + 1

3
 cos ax
Since is an even function of x,
x2 + 1
Z ∞
1 ∞ cos ax πe−a
Z
cos ax
dx = dx =
0 x2 + 1 2 −∞ x2 + 1 2

Question 2(b) Let f be analytic in the entire complex plane. Suppose that there exists a
constant A > 0, such that |f (z)| ≤ A|z| for all z. Prove that there is a complex number a
such that f (z) = az for all z.

Solution. We first prove (Cauchy’s inequality) that if f (z) is analytic in a domain G and
if the disc |z − z0 | ≤ ρ ⊆ G then

n!M (ρ)
|f (n) (z0 )| ≤
ρn

where M (ρ) = max |f (z)| on |z − z0 | = ρ — this follows from Cauchy’s Integral formula:
Z
(n) n! f (z)
f (z0 ) = dz
2πi |z−z0 |=ρ (z − z0 )n+1

and therefore
n! M (ρ) n!M (ρ)
|f (n) (z0 )| ≤ n+1
2πρ =
2π ρ ρn
We now prove that if f (z) is entire i.e. analytic over the whole complex plane, and
|f (z)| ≤ G|z|m for all |z| > R, then f (z) is a polynomial of degree ≤ m.
f (n) (0)
Let f (z) = ∞ n
P
n=0 a n z be a Taylor series of f (z) around z = 0. Then a n = .
n!
f (n) (0) M (r)
By Cauchy’s inequality proved above, |an | = ≤ where M (r) is maximum
n! rn
Grm G
of |f (z)| on |z| = r. Let r > R, then M (r) ≤ Grm and we get |an | ≤ n = n−m . and
r r
G
therefore as r → ∞, n−m → 0 for n > m i.e. |an | = 0 for n > m. Hence f (z) = m r
P
r=0 ar z
r
i.e. f (z) is a polynomial of degree ≤ m.
Now we are given |f (z)| ≤ A|z|. This means that f (z) = a0 + a1 z. But 0 ≤ |f (0)| ≤
A · 0 ⇒ f (0) = 0 ⇒ a0 = 0, so f (z) = a1 z, where a1 is a constant.
Note: An alternative statement of the above question is: If f (z) is an entire transcen-
dental function, then whatever G > 0, R > 0, m > 0 are prescribed, there exist points z such
that |f (z)| > G|z|m and |z| > R.

4
Alternate solution: Consider the function g(z) = f (z) z
, z 6= 0 and g(0) = f 0 (0). Note that
|f (z)| ≤ A|z| ⇒ f (0) = 0. Then g is continuous at 0, because
f (z) f (z) − f (0)
lim |g(z) − g(0)| = lim − f 0 (0) = lim − f 0 (0) = 0
z→0 z→0 z z→0 z
Let f = u + iv, where u, v satisfy the Cauchy Riemann equations, since f is entire. Then
u + iv (ux + yv) + i(vx − uy)
g(z) = =
x + iy x2 + y 2
ux + yv vx − uy
Writing g(z) = U + iV , we get U = 2 2
,V = 2 . Now it is clear that g is analytic
x +y x + y2
over the entire complex plane except possibly at z = 0. We now check the Cauchy Riemann
equations for U, V at z = 0. Note that f (0) = 0 ⇒ u(0, 0) = v(0, 0) = 0.
u(h,0)
∂U U (h, 0) − U (0, 0) − ux (0, 0) u(h, 0) − hux (0, 0)
(0, 0) = lim = lim h = lim
∂x h→0 h h→0 h h→0 h2
ux (h, 0) − ux (0, 0) 1
= lim = uxx (0, 0)
h→0 2h 2
v(0,k)
∂U U (0, k) − U (0, 0) − ux (0, 0) v(0, k) − kux (0, 0)
(0, 0) = lim = lim k = lim
∂y k→0 k k→0 k k→0 k2
vy (0, k) − ux (0, 0) 1
= lim = vyy (0, 0)
k→0 2k 2
v(h,0)
∂V V (h, 0) − V (0, 0) − vx (0, 0) v(h, 0) − hvx (0, 0)
(0, 0) = lim = lim h = lim
∂x h→0 h h→0 h h→0 h2
vx (h, 0) − vx (0, 0) 1
= lim = vxx (0, 0)
h→0 2h 2
−u(0,k)
∂V V (0, k) − V (0, 0) − vx (0, 0) −u(0, k) − kvx (0, 0)
(0, 0) = lim = lim k = lim
∂y k→0 k k→0 k k→0 k2
−uy (0, k) − vx (0, 0) 1
= lim = − uyy (0, 0)
k→0 2k 2
Now by the Cauchy Riemann equations for u, v, ux = vy ⇒ uxx = vxy and uy = −vx ⇒
uyy = −vyx . Hence Ux (0, 0) = 12 uxx (0, 0) = 12 vxy (0, 0) = − 12 uyy (0, 0) = Vy (0, 0).
Also, vx = −uy ⇒ vxx = −uxy , and vy = ux ⇒ vyy = uyx . So Uy (0, 0) = 21 vyy (0, 0) =
1
u = − 12 vxx (0, 0) = −Vx (0, 0). Thus the Cauchy Riemann equations hold at (0, 0) also, so
2 yx
g(z) is analytic at 0, as it is continuous at 0. Thus g(z) is an entire function.
But |g(z)| = | f (z)
z
| ≤ A|z|
|z|
= A, so g is bounded over the complex plane. Hence by
Liouville’s theorem, g is a constant, say a. Thus f (z) = az, as required.
X∞
Question 2(c) Suppose a power series an z n converges at a point z0 6= 0. Let z1 be
n=0
such that |z1 | < |z0 | and z1 6= 0. Show that the series converges uniformly in the disc
{z : |z| ≤ |z1 |}.

5
 Solution. Let | zz10 | = ρ, then ρ < 1. Since ∞ n n
P
n=0 an z0 is convergent, an z0 → 0 as n → ∞,
n
therefore there exists M such that |an z0 | < M for n ≥ 0. Now let z be any point such that
|z| ≤ |z1 |, then
r+p r+p r+p  z n r+p r+p
X
n
X
n
X X z n X
an z ≤ |an z | = an z0n ≤M =M ρn
n=r n=r n=r
z0 z
n=r 0 n=r

∞ r+p
X
n  X
Since the series ρ is convergent, given  > 0 there exists N such that ρn <
n=0 n=r
M
for all r ≥ N and p = 1, 2, . . .. Clearly this N is independent of z. Thus given  > 0 there
exists N independent of z such that
r+p
X
an z n <  for n ≥ N, p = 1, 2, 3, . . .
n=r

∞
X
i.e. the series an z n is uniformly convergent for all z with |z| ≤ |z1 |.
n=0

6
 UPSC Civil Services Main 1996 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

February 20, 2010

Question 1(a) Sketch the ellipse C described in the complex plane by

z = A cos λt + iB sin λt, A > B

where t is a real variable and A, B, λ are positive constants.

If C is the trajectory of a particle with z(t) as the position vector of the particle at time
t, identify with justification

1. the two positions where the velocity is minimum.

2. the two positions where the acceleration is maximum.

x2 y2
Solution. We are given that x = A cos λt, y = B sin λt which implies that 2 + 2 = 1.
A B
Since A > B, it follows that it is the standard ellipse with 2A as the major axis and 2B as
the minor axis.
B(0, B)

A0
(−A, 0) O A(A, 0)

B 0 (0, −B)

1
 dz
1. The velocity v = = −Aλ sin λt + iBλ cos λt.
dt

dz
Speed = magnitude of velocity =
p dt
2 2 2 2 2 2
= A λ sin λt + B λ cos λt
q
= λ (A2 − B 2 ) sin2 λt + B 2

Since A2 − B 2 > 0, the speed is minimum when sin2 λt = 0 i.e. when x(t) = ±A, y(t) =
0 i.e. when the particle is at the two ends of the major axis, the points A and A0 in
the figure.
d2 z
2. Acceleration = = −Aλ2 cos λt − iBλ2 sin λt.
dt2
√ p
Magnitude of acceleration = λ2 A2 cos2 λt + B 2 sin2 λt = λ2 (A2 − B 2 ) cos2 λt + B 2 .
Since A2 − B 2 > 0, acceleration is maximum when cos2 λt = 1 ⇒ cos λt = ±1 i.e. the
particle is at either end of the major axis, A or A0 . (Note that acceleration is minimum
when cos2 λt = 0 i.e. the particle is at either end of the minor axis).

1 − cos z
Question 1(b) Evaluate lim .
z→0 sin(z 2 )

Solution.
sin2 z2
1 − cos z 2 sin2 z2 2 ( z2 )2 1
lim = lim = lim sin(z 2 )
=
z→0 sin(z 2 ) z→0 sin(z 2 ) z→0 4 2
z2
Note that sin z has a simple zero at z = 0 and sin z = zφ(z) where φ(z) is analytic and
sin z
φ(0) = 1, so lim = 1.
z→0 z

√
sin z
Question 1(c) Show that z = 0 is not a branch point for the function f (z) = √ . Is it
z
a removable singularity?
√
Solution. We know that √ w = √z is a multiple valued function and has two branches. Once
we fix a branch of w = z, sin z is analytic, and
√ √
√ √ ( z)3 ( z)5
sin z = z − + + ...
3! 5!
or √
sin z z z2 z3
√ =1− + − + ...
z 3! 5! 7!

2
 √ √
sin z sin z
Thus lim √ = 1, so z = 0 is not a branch point of the function f (z) = √ . In fact
z→0 z z
z = 0 is a removable singularity of f (z). In fact
( √
sin z
√ , z 6= 0
z
F (z) =
1, z=0
√
is analytic everywhere once a branch of z is specified.

Question 2(a) Prove that every polynomial equation a0 + a1 z + a2 z 2 + . . . + an z n = 0, an 6=

0, n ≥ 1 has exactly n roots.

Solution. Let P (z) = a0 + a1 z + a2 z 2 + . . . + an z n . Suppose, if possible, that P (z) 6= 0

1
for any z ∈ C. Let f (z) = , then f (z) is an entire function i.e. f (z) is analytic in the
P (z)
whole complex plane. We shall now show that f (z) is bounded.
 
n an−1 an−2 a0
P (z) = z an + + 2 + ... + n
z z z
aj
Since n−j → 0 as z → ∞, for 0 ≤ j < n, is follows that given  = |a2nn | there exists R > 0
z
aj |an |
such that |z| > R ⇒ n−j < for 0 ≤ j < n. Thus
z 2n
an−1 an−2 a0 an an
an + + 2 + . . . + n ≥ |an | − n =
z z z 2n 2
and therefore
1 1 2
|f (z)| = = n an−1 an−2 a0

≤ for |z| > R
P (z) z an + z
+ z2
+ ... + zn
|an |Rn

Since |z| ≤ R is a compact set and f (z) is analytic on it, f (z) is bounded on |z| ≤ R.
Consequently f (z) is bounded on the whole complex plane. Now we use Liouville’s theorem
— If an entire function is bounded on the whole complex plane, then it is a constant. Thus
f (z) and therefore P (z) is a constant, which is not true, hence our assumption that P (z) 6= 0
for all z ∈ C is false. So there is at least one z1 ∈ C where P (z1 ) = 0. (This result is called
the fundamental theorem of algebra.)
We now prove by induction on n that P (z) has n zeros. If n = 1, P (z) = a0 + a1 z has
one zero namely z = − aa10 .
Assume as induction hypothesis that any polynomial of degree n − 1 has n − 1 zeros. By
Euclid’s algorithm, , we get P1 (z) and R(z) such that P (z) = (z − z1 )P1 (z) + R(z), where
R(z) ≡ 0 or deg R(z) < 1 i.e. R(z) is a constant. Putting z = z1 we get R(z) ≡ 0, so
P (z) = (z − z1 )P1 (z). Since P1 (z) is a polynomial of degree n − 1, by induction hypothesis
it has n − 1 roots in C, and therefore P (z) has n roots in C.

3
 We now prove that P (z) has exactly n roots. Let z1 , z2 , . . . , zn be the (not necessarily
P (z)
distinct) roots of P (z). Let g(z) = . Clearly g(z) is analytic in
(z − z1 )(z − z2 ) . . . (z − zn )
the whole complex plane. Since

P (z) an + an−1z
+ an−2
z2
+ . . . + zan0
lim g(z) = lim = = an
z→∞ z→∞ (z − z1 )(z − z2 ) . . . (z − zn ) (1 − zz1 )(1 − zz2 ) . . . (1 − zzn )

it follows that given  > 0 there exists R such that |g(z) − an | <  for |z| > R, so g(z) is
bounded in the region |z| > R. The function g(z) being analytic is bounded in the compact
region |z| ≤ R. Thus by Liouville’s theorem g(z) is a constant, in fact g(z) = an , and
therefore
P (z) = an (z − z1 )(z − z2 ) . . . (z − zn )
Thus if ζ is a zero of P (z), then ζ = zj for some j, 1 ≤ j ≤ n. Thus P (z) has exactly n
zeroes.
Alternate Proof: We shall use Rouche’s theorem — Let γ be a simple closed rectifiable
curve. Let f (z), g(z) be analytic on and within γ. Suppose |g(z)| < |f (z)| on γ, then f (z)
and f (z) ± g(z) have the same number of zeroes inside γ.
Let f (z) = an z n and g(z) = an−1 z n−1 + . . . + a0 . Let R be so large that |g(z)| < |f (z)|
on |z| = R. Then f (z) and f (z) + g(z) = P (z) have the same number of zeroes within
|z| = R. But whatever R > 0 we take, f (z) has exactly n zeroes in |z| = R, therefore P (z)
has exactly n zeroes in C.
Note: Rouche’s theorem follows from the Argument Principle — Note that ∆γ (arg(f (z)+
g(z))) =change in argument of f (z)+g(z) as z moves along γ = ∆γ arg f (z)+∆γ arg(1+ fg(z) (z)
)
as f (z) 6= 0 along γ. But ∆γ arg(1 + fg(z)
(z)
) = 0 because | fg(z)
(z)
| < 1 and therefore fg(z)
(z)
continues
to lie in the disc |w − 1| < 1 as z moves on γ i.e. does not go around the origin.

Question 2(b) By using the residue theorem, evaluate

Z ∞
loge (x2 + 1)
dx
0 x2 + 1
Solution.
log(z + i)
Let f (z) = and we consider
1 + z2
log(z + i) in C − {z | z = iy, y ≤ −1},
where it is single-valued. Let γ be the con- Γ
tour consisting of the line joining (−R, 0) and
(R, 0) and Γ, which is the arc of the circle of
radius R and center (0, 0) lying in the upper
half plane. γ is oriented counter-clockwise. (−R, 0) (0, 0) (R, 0)

4
Clearly f (z) has a simple pole at z = i in the upper half plane. The residue at z = i is
(z + i) log(z + i) log 2i 1 πi 1 π π 1
lim = = log 2e 2 = log 2 + i = − i log 2
z→i 1 + z2 2i 2i 2i 2 4 2
Thus by Cauchy’s residue theorem
Z Z Z ∞
log(z + i) log(z + i) log(x + i) π 1 
lim 2
= lim 2
+ 2
dx = 2πi − i log 2
R→∞ γ 1+z R→∞ Γ 1+z −∞ 1+x 4 2
as z = x on the real axis. Z
log(z + i)
We shall now show that lim = 0. On Γ, z = Reiθ , so
R→∞ Γ 1 + z2
Z π
log(Reiθ + i)Rieiθ
Z
log(z + i)
= dθ
Γ 1 + z2 0 R2 e2iθ + 1
Now |R2 e2iθ + 1| ≥ R2 − 1, log(Reiθ + i) = log Reiθ + log(1 + Reiiθ ). Clearly | log Reiθ | =
| log R + iθ| ≤ log R + π and therefore
R| log(1 + Reiiθ )|
Z Z π Z π
log(z + i) (π + log R)R
≤ dθ + dθ
Γ 1 + z2 0 R2 − 1 0 R2 − 1
Z
(π + log R)R R| log(1+ iiθ )| log(z + i)
Since 2
→ 0 and Re
R2 −1
→ 0 as R → ∞, it follows that lim =
R −1 R→∞ Γ 1 + z2
0.
Thus Z ∞
log(x + i) π2
dx = π log 2 + i
−∞ 1 + x2 2
Equating real and imaginary parts, we get
Z ∞
log(1 + x2 ) 1 ∞ log(1 + x2 ) 1 ∞ log(x + i) + log(x − i)
Z Z
1
2
dx = 2
dx = 2
dx = [2π log 2] = π log 2
0 1+x 2 −∞ 1 + x 2 −∞ 1+x 2

 1 
Question 2(c) Find the Laurent expansion of f (z) = (z − 3) sin about the singu-
z+2
larity z = −2. Specify the region of convergence and the nature of the singularity at z = −2.
Solution. It is well known that
∞
 1  X (−1)k−1  1 2k−1
sin =
z+2 k=1
(2k − 1)! z + 2
 1   1   1 
⇒ (z − 3) sin = (z + 2) sin − 5 sin
z+2 z+2 z+2
∞ k−1  ∞
X (−1) 1 2k−1 X (−1)k−1  1 2k−1
= (z + 2) −5
k=1
(2k − 1)! z + 2 k=1
(2k − 1)! z + 2
∞
X ak (−1)k−1 5(−1)k−1
= , a 2k−2 = , a 2k−1 =
k=0
(z + 2)k (2k − 1)! (2k − 1)!

5
The region of convergence of the series is 0 < |z + 2| < ∞. The Laurent expansion shows
that the function has an essential singularity at z = −2 — this also follows from the fact
that limz→0 sin z1 does not exist.

6
 UPSC Civil Services Main 1997 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

February 20, 2010

Question 1(a) Prove that u = ex (x cos y−y sin y) is harmonic and find the analytic function
whose real part is u.
Solution.
∂u
= ex (x cos y − y sin y) + ex cos y
∂x
∂ 2u
= ex (x cos y − y sin y) + 2ex cos y
∂x2
∂u
= ex (−x sin y − sin y − y cos y)
∂y
∂ 2u
= ex (−x cos y − 2 cos y + y sin y)
∂y 2
∂ 2u ∂ 2u
Clearly + 2 = 0, showing that u is harmonic.
∂x2 ∂y z + z z − z  z + z z − z 
Let f (z) = u , + iv , . Then
2 2i 2 2i
f 0 (z) = ux + ivx = ux − iuy because of the C-R equations
z + z z − z  z + z z − z 
= ux (x, y) − iuy (x, y) = ux , − iuy ,
2 2i 2 2i
Since the above is an identity, we take z = z, so x = z, y = 0. Thus f 0 (z) = ux (z, 0) −
iuy (z, 0) = zez + ez . Then
Z Z
0
f (z) = f (z) dz = (z + 1)ez dz = zez + C

Hence f (z) = zez is the required function.

1
 I Z 2π
dz 1 + 2 cos θ
Question 1(b) Evaluate where C is the unit circle. Deduce that dθ =
C z+2 0 5 + 4 cos θ
0.
I
dz 1
Solution. Cauchy’s theorem implies that = 0 because z+2
has no pole inside
C z+2
|z| = 1.
Putting z = eiθ , we get
2π
(i cos θ − sin θ) dθ
Z Z
dz
I = =
|z|=1 z+2 0 cos θ + 2 + i sin θ
2π
(i cos θ − sin θ)(cos θ + 2 − i sin θ)
Z
= dθ
0 (cos θ + 2)2 + sin2 θ
Z 2π
i(cos2 θ + sin2 θ + 2 cos θ) − 2 sin θ
= dθ
0 cos2 θ + 4 cos θ + 4 + sin2 θ
Z 2π Z 2π
2 sin θ 1 + 2 cos θ
= − dθ + i dθ
0 5 + 4 cos θ 0 5 + 4 cos θ
Since I = 0, it follows that
Z 2π Z 2π
2 sin θ 1 + 2 cos θ
dθ = 0, dθ = 0
0 5 + 4 cos θ 0 5 + 4 cos θ

A1 A2 An f (z)
Question 1(c) If f (z) = + + . . . + , find the residue at a for
z − a (z − a)2 (z − a)n z−b
where A1 , A2 , . . . An , a, b are constants. What is the residue at infinity?

Solution. Case (1): a 6= b.

f (z) A1 (z − a)n−1 + A2 (z − a)n−2 + . . . + An

=
z−b (z − b)(z − a)n

f (z)
showing that has a pole of order n at z = a. The residue at z = a is the coefficient of
z−b
1 f (z)
z−a
in the Laurent expansion of around a. Now
z−b
1 1  z − a −1
= (z − a + a − b)−1 = 1+
z−b  a−b a − b  
f (z) A1 A2 An 1 z − a  z − a 2  z − a 3
= + + ... + 1− + − + ...
z−b z − a (z − a)2 (z − a)n a − b a−b a−b a−b

2
 1 f (z)
Thus the coefficient of i.e. the residue of is given by
z−a z−b
A1 A2 A3 (−1)n−1 An
− + + . . . + = −f (b)
a − b (a − b)2 (a − b)3 (a − b)n

Note: The same  residue could  be computed by using the formula — Residue at z = a is
1 dn−1 f (z)
(z − a)n but this calculation would be much more complicated.
(n − 1)! dz n−1 z−b
f (z)
Case (2): a = b. In this case has a pole of order n + 1 at z = a. Residue at z = a
n
z−b
1 d
is given by [(z − a)n f (z)] = 0.
n! dz n
Residue at ∞:
A1  a −1 A2  a −2 An  a −n
f (z) = 1− + 2 1− + ... + n 1 −
z z z z z z 
f (z) f (z)  b −1 A1 1 1 1 b
= 1− = + 2 (. . .) + 3 (. . .) + . . . + + ...
z−b z z z z z z z2

1 f (z)
Since the term z
is not present in the Laurent expansion of the residue at ∞ is 0.
(z − b)
1
Question 2(a) Find the Laurent series for the function e z in 0 < |z| < ∞. Deduce that

1 π cos θ
Z
1
e cos(sin θ − nθ) dθ =
π 0 n!
for all n = 0, 1, 2, . . ..

Solution. See 2001 question 2(a).

2
Question 2(b) Integrating e−z along a suitable rectangular contour show that
Z ∞ √
−x2 π −b2
e cos 2bx dx = e
0 2

Solution. More generally, we shall prove that

Z ∞ √
−λx2 π − 1 −λa2
e cos 2aλx dx = λ 2e
0 2
then λ = 1, a = b will give us the desired result.

3
 D(−R, a) C(R, a)
y=a

2
Our f (z) = e−λz and the contour is Γ, the x = −R x=R
rectangle ABCD where A = (−R, 0), B = Γ
(R, 0), C = (R, a), D = (−R, a) oriented in
the anticlockwise direction. y=0
A(−R, 0) (0, 0) B(R, 0)

I
2
−λz 2
Since e is an entire function, and therefore has no poles, e−λz dz = 0.
Γ
Now we compute the integrals along the four sides.
1. Z Z a
−λz 2 2
e dz = e−λ(R+iy) i dy
BC 0
because z = R + iy on BC and 0 ≤ y ≤ a. Thus
Z Z a
−λz 2 −λR2 2 2
e dz ≤ e eλy dy = constant × e−λR
BC 0
Z
2 2
Thus since e−λR → 0 as R → ∞, e−λz dz → 0 as R → ∞.
BC
Z
2
2. A similar argument shows that e−λz dz → 0 as R → ∞.
DA

3. Z Z ∞
−λz 2 2
lim e dz = e−λx dx
R→∞ AB −∞
as z = x on AB.
4. Z Z −∞
−λz 2 2
lim e dz = e−λ(x+ia) dx
R→∞ CD ∞
as z = x + ia on CD and orientation is from C to D. Thus
Z Z −∞
−λz 2 2 2
lim e dz = e−λ(x −a ) e−2iaλx dx
R→∞ CD
Z∞∞ Z ∞
−λ(x2 −a2 ) 2 2
= −e cos(2aλx) dx + i e−λ(x −a ) sin(2aλx) dx
−∞ −∞

Now
Z Z
2
0 = limf (z) dz = lim e−λz dz
R→∞ Γ R→∞ Γ
Z Z Z Z 
−λz 2 −λz 2 −λz 2 −λz 2
= lim e dz + e dz + e dz + e dz
R→∞ AB BC CD DA
Z ∞ Z ∞ Z ∞
−λx2 λa2 −λx2 λa2 2
= e dx − e e cos(2aλx) dx + ie e−λx sin(2aλx) dx
−∞ −∞ −∞

4
Equating real and imaginary parts, we get
Z ∞
2
e−λx sin(2aλx) dx = 0
Z −∞
∞ Z ∞
−λx2 −λa2 2
e cos(2aλx) dx = e e−λx dx
−∞ −∞
√
(Substituting X = λx)
Z ∞
2 1 √ 1
= 2 e−X λ− 2 dX = πλ− 2
Z ∞ Z0 ∞
2 1 2 1 2 1√
⇒ e−λx cos(2aλx) dx = e−λx cos(2aλx) dx = e−λa λ− 2 π
0 2 −∞ 2
This completes the proof.

Question 2(c) Find the function f (z) analytic within the unit circle which takes the values
a − cos θ + i sin θ
, 0 ≤ θ ≤ 2π on the circle.
a2 − 2a cos θ + 1
Solution. Since f (z) is analytic within |z| < 1, the Maclaurin series of f (z) is given by
∞
f (n) (0)z n
Z
X
(n) n! f (z) dz
f (z) = , where f (0) =
n=0
n! 2πi |z|=1 z n+1
a − cos θ + i sin θ
We are given that on |z| = 1, f (z) = and we know that on |z| = 1,
a2 − 2a cos θ + 1
z = eiθ , dz = ieiθ dθ, 0 ≤ θ ≤ 2π, therefore
a − z1 a − z1 1
f (z) = 2 1 = 1 = on |z| = 1
a − a(z + z ) + 1 (a − z )(a − z) a−z
Now we use the Maclaurin series to compute the value of f inside the unit circle.
Z
(n) n! dz
f (0) =
2πi |z|=1 (a − z)z n+1
Z 2π iθ −(n+1)iθ
n! ie e
= dθ
2πi 0 a − eiθ
Z 2π
n!  eiθ −1
= e−niθ 1 − dθ
2πa 0 a
Z 2π
n!  eiθ e2iθ einθ 
= e−niθ 1 + + 2 + . . . n + . . . dθ
2πa 0 a a a
Z 2π 2π Z 2π
ikθ eikθ
Since e dθ = = 0 for k 6= 0, and dθ = 2π, it follows that
0 ik 0 0

n! 1 n!
f (n) (0) = · n · 2π = n+1
2πa a a

5
Consequently
∞
X zn 1 z 1 z −1 1
f (z) = = + + . . . = 1 − =
n=0
an+1 a a2 a a a−z

We need a > 1 so that | az | < 1 on |z| ≤ 1.

6
 UPSC Civil Services Main 1998 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

February 20, 2010

Question 1(a) Show that the function

 3 3
 x (1 + i) − y (1 − i) , z 6= 0
f (z) = x2 + y 2
0, z=0


is continuous and C-R conditions are satisfied at z = 0, but f 0 (z) does not exist at z = 0.
x3 − y 3 x3 + y 3
Solution. Let f (z) = u + iv, then u = , v = for z 6= 0, and u(0, 0) =
x2 + y 2 x2 + y 2
v(0, 0) = 0.
h3
∂u u(h, 0) − u(0, 0) h2
−0
(0, 0) = lim = lim =1
∂x h→0 h h→0 h
−k3
∂u u(0, k) − u(0, 0) k2
−0
(0, 0) = lim = lim = −1
∂y k→0 k k→0 k
h3
∂v v(h, 0) − v(0, 0) 2 − 0
(0, 0) = lim = lim h =1
∂x h→0 h h→0 h
k3
∂v v(0, k) − v(0, 0) k2
−0
(0, 0) = lim = lim =1
∂y k→0 k k→0 k
∂u ∂v ∂u ∂v
Thus = , =− at (0, 0), i.e. the Cauchy Riemann equations are satisfied at
∂x ∂y ∂y ∂x
(0, 0).
f (z) is clearly continuous at z = 0, because
x3 − y 3 r3 (cos3 θ − sin3 θ) p
|u(x, y) − u(0, 0)| = = ≤ 2 x2 + y 2
x2 + y 2 r2
p
|v(x, y) − v(0, 0)| ≤ 2 x2 + y 2

1
Thus u, v are continuous at (0, 0), so f (z) is continuous at (0, 0).
If f (z) is to be differentiable at 0, then

f (z) − 0 (x3 − y 3 ) + i(x3 + y 3 ) (x3 + iy 3 )(1 + i)(x − iy)

lim = lim = lim
z→0 z x→0,y→0 (x2 + y 2 )(x + iy) x→0,y→0 (x2 + y 2 )2

∂u ∂v
should exist and it should be equal to (0, 0) + i (0, 0) = 1 + i.
∂x ∂x
But if we take the limit along y = x, then

f (z) − 0 (x3 + ix3 )(1 + i)(x − ix) 1+i

lim = lim 2 2
=
z→0 z x→0 (2x ) 2

Therefore f (z) is not differentiable at z = 0.

z
Question 1(b) Find the Laurent expansion of about the singularity z = −2.
(z + 1)(z + 2)
Specify the region of convergence and nature of singularity at z = −2.

Solution. Clearly
z 2 1 2 1
f (z) = = − = +
(z + 1)(z + 2) z+2 z+1 z + 2 1 − (z + 2)
∞
2 X
= + (z + 2)n for |z + 2| < 1 (∗)
z + 2 n=0

The function satisfies the requirements of Laurent’s theorem in the region 0 < |z + 2| < 1
and the right hand side of (∗) represents the Laurent series of f (z), which converges for
|z + 2| < 1, because we have a singularity at z = −1 which lies on |z + 2| = 1. The Laurent
series expansion (∗) shows that f (z) has a simple pole at z = −2, where its residue is 2.

Question 1(c) By using the integral representation of f (n) (0), prove that
 x n 2 1
I
xn exz
= dz
n! 2πi C n!z n+1
Hence show that ∞  n Z 2π
X x 2 1
= e2x cos θ dθ
n=0
n! 2π 0

Solution. It is easily deducible from Cauchy’s Integral formula that if f (z) is analytic
within and on a simple closed contour C and z0 is a point in the interior of C, then
I
(n) n! f (z)
f (z0 ) = , dz
2πi C (z − z0 )n+1

2
Let f (z) = exz (Here x is not Re x but a parameter), then f (z) is an entire function and
therefore
exz
I
(n) n n!
f (0) = x = , dz
2πi C z n+1
where C is any closed contour containing 0 in its interior. Hence
 x n 2 xn n!
I
exz 1
I
xn exz
= , dz = dz
n! (n!)2 2πi C z n+1 2πi C n!z n+1

as required.
We take C to be the unit circle for convenience. Then
∞  n ∞ I ∞
xn exz xn exz
I X
X x 2 1 X 1
= n+1
dz = n+1
dz
n=0
n! 2πi n=0 C n!z 2πi C n=0
n!z

Interchange of summation and integral is justified. Thus

∞  n ∞
exz X ( xz )n exz x
Z Z
X x 2 1 1
= dz = e z dz
n=0
n! 2πi |z|=1 z n=0 n! 2πi |z|=1 z

Put z = eiθ so that dz = ieiθ dθ and

∞  n Z 2π x(eiθ +e−iθ ) Z 2π
X x 2 1 e iθ 1
= iθ
ie dθ = e2x cos θ dθ
n=0
n! 2πi 0 e 2π 0

as required.

Question 2(a) Prove that all roots of z 7 − 5z 3 + 12 = 0 lie between the circles |z| = 1 and
|z| = 2.

Solution. See 2006 question 2(b).

Question 2(b) By integrating around a suitable contour show that

Z ∞
x sin mx π
4 4
dx = 2 e−mb sin mb
0 x +a 4b
where b = √a .
2

3
Solution.
zeimz
Let f (z) = 4 . We consider the in-
z + a4
R
tegral γ f (z) dz where γ is the contour con- Γ
sisting of the line joining (−R, 0) and (R, 0)
and Γ, the arc of the circle of radius R and
center (0, 0) lying in the upper half plane.
(−R, 0) (0, 0) (R, 0)
π
Reiθ eimR(cos θ+i sin θ) R2
Z Z
iθ
f (z) dz = Rie dθ ≤ π
Γ 0 z 4 + a4 R 4 − a4
because |zZ + a | ≥ |z| − |a | = R4 − a4 on Γ, and e−mR sin θ ≤ 1 as sin θ > 0 for 0 < θ < π.
4 4 4 4

Thus f (z) dz → 0 as R → ∞ and

Γ
∞
xeimx
Z Z
lim f (z) dz = dx
R→∞ γ −∞ x 4 + a4
Z
But by Cauchy’s residue theorem f (z) dz = 2πi × (the sum of the residues of poles of
γ
πi 3πi πi 3πi
f (z) inside γ). The poles of f (z) are simple poles at ±ae 4 , ±ae 4 , out of which ae 4 , ae 4

are inside γ.
πi πi 3πi 3πi
πi ae 4 eima e 4 3πi ae 4 eima e 4
Residue at z = ae 4 is 3πi . Residue at z = ae 4 is 9πi .
4a3 e 4 4a3 e 4
i h ima(cos π +i sin π ) ima(cos 3π +i sin 3π )
i
Sum of residues = −e 4 4 +e 4 4
4a2
i h ma
√ (i−1) ma
√ (−i−1)
i
= −e 2 +e 2
4a2
ma
−√ − ma
√
ie 2  ma  e 2 ma
= 2
−2i sin √ = 2
sin √
4a 2 2a 2
Thus Z ∞ − ma
√
xeimx e 2 ma
4 4
dx = 2πi 2 sin √
−∞ x + a 2a 2
Taking imaginary parts of both sides, we get
Z ∞ Z ∞ − ma
√
x sin mx x sin mx πe 2 ma πe−mb
4 4
dx = 2 dx = sin √ = sin mb
−∞ x + a 0 x 4 + a4 a2 2 2b2
where b = √a . Thus
2
∞
πe−mb
Z
x sin mx
dx = sin mb
0 x 4 + a4 4b2
as required.

4
 Z 2π
dθ
Question 2(c) Using the residue theorem evaluate .
0 3 − 2 cos θ + sin θ
dz
Solution. We put z = eiθ , so that dθ = iz
, cos θ = 21 (z + z1 ), sin θ = 1
2i
(z − z1 ). Thus
Z 2π
dθ
I =
3 − 2 cos θ + sin θ
I0
dz
= 1 1 1
|z|=1 iz[3 − (z + z ) + 2i (z − z )]
I
dz
= 2 2 2
|z|=1 6iz − 2iz − 2i + z − 1
I
dz
= 2 i 5i
|z|=1 (1 − 2i)(z + 1−2i )(z + 1−2i )

Clearly (6iz − 2iz 2 − 2i + z 2 − 1)−1 has two simple poles − 1−2i

i 5i
and − 1−2i i
of which only − 1−2i
i
z + 1−2i 1
lies inside |z| = 1. The residue at this pole is limi i 5i = . Thus
z→− 1−2i (1 − 2i)(z + 1−2i )(z + 1−2i ) 4i
by Cauchy’s residue theorem
Z 2π
dθ 1
I= = 2 · 2πi · =π
0 3 − 2 cos θ + sin θ 4i

5
 UPSC Civil Services Main 1999 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 30, 2010

Question 1(a) Examine the nature of the function

x2 y 5 (x + iy)
f (z) = , z 6= 0, f (0) = 0
x4 + y 10
in a region including the origin and hence show that the Cauchy-Riemann equations are
satisfied at the origin, but f (z) is not analytic there.

Solution.
(
x3 y 5
x4 +y 10
, (x, y) 6= (0, 0)
u(x, y) = Re f (z) =
0, (x, y) = (0, 0)
( 2
x y6
x4 +y 10
, (x, y) 6= (0, 0)
v(x, y) = Im f (z) =
0, (x, y) = (0, 0)

u(x, 0) − u(0, 0) v(0, y) − v(0, 0)

Now =0= , therefore ux (0, 0) = vy (0, 0) = 0. Similarly
x y
uy (0, 0) = 0 = −vx (0, 0). Thus the Cauchy-Riemann equations are satisfied at (0, 0).
f (z) − f (0) x2 y 5
However f (z) is not analytic at (0, 0) because lim = lim 4 does not
z→0 z z→0 x + y 10
f (z) − f (0) m
exist — when we take y 5 = mx2 , then lim = which is different for
z→0 z 1 + m2
different values of m.

1
 Additional notes: Let z 6= 0. It can be calculated that

∂u 3x2 y 15 − x6 y 5 ∂v −4x2 y 15 + 6x6 y 5

= =
∂x (x4 + y 10 )2 ∂y (x4 + y 10 )2
∂v 2xy 16 − 2x5 y 6 ∂u 5x7 y 4 − 5x3 y 14
= =
∂x (x4 + y 10 )2 ∂y (x4 + y 10 )2

Now ∂u∂x
= ∂y∂v
⇔ 3x2 y 15 − x6 y 5 = −4x2 y 15 + 6x6 y 5 ⇔ x2 y 15 = x6 y 5 ⇔ x4 = y 10 or x = 0 or
y = 0.
Also, ∂u∂y
∂v
= − ∂x when x4 = y 10 or x = 0 or y = 0. Thus the Cauchy-Riemann equations
are satisfied at all those z for which x4 = y 10 or x = 0 or y = 0. But f (z) is not analytic at
any of these points because f (z) is not differentiable in any neighborhood of these points,
as we can find points in every neighborhood which are not of this kind, so there are no
neighborhoods in which the Cauchy Riemann equations are satisfied everywhere.
−1
Question 1(b) For the function f (z) = , find the Laurent series for the domain
z2 − 3z + 2
(i) 1 < |z| < 2 (ii) |z|I> 2.
Show further that f (z) dz = 0 where C is any closed contour enclosing the points z = 1
C
and z = 2.
1 1
Solution. f (z) = −
z−1 z−2
(i) 1 < |z| < 2 ⇒ | z1 | < 1, | z2 | < 1.

1 1 −1 1  z −1
f (z) = 1− + 1−
z z 2 2
∞ ∞ n
1 X 1 1 X z
= n
+
z n=0 z 2 n=0 2n
∞ ∞
X 1 X zn
= +
n=1
z n n=0 2n+1

(ii) |z| > 2 ⇒ | z1 | < 1, | z2 | < 1

1 1 −1 1  2 −1
f (z) = 1− − 1−
z z z z
∞ ∞ n
1 X 1 1 X 2
= n
−
z n=0 z z n=0 z n
∞
X 1 − 2n
=
n=0
z n+1

2
 I I 
1 1 
f (z) dz = − dz
C C z −1 z−2
1 1
 
= 2πi residue of z−1 at z = 1 − residue of z−2
at z = 2
= 2πi[1 − 1] = 0

2z + 3
Question 1(c) Show that the transformation w = transforms the circle x2 +y 2 −4x =
z−4
0 into the straight line 4u + 3 = 0 where w = u + iv.
Solution. The point z = 4 goes to the point at ∞, showing that the given circle 0 =
x2 + y 2 − 4x = zz − 4( z+z
2
) = zz − 2z − 2z = 0 is mapped onto a line, as z = 4 lies on it.
3 + 4w
Now zw−4w = 2z+3 ⇒ zw−2z = 3+4w ⇒ z = . Thus the circle zz−2z−2z = 0
w−2
goes to

3 + 4w 3 + 4w 3 + 4w 3 + 4w
0 = −2 −2 =0
w−2 w−2 w−2 w−2
⇒0 = 9 + 12w + 12w + 16ww − 2(3 + 4w)(w − 2) − 2(3 + 4w)(w − 2)
⇒0 = 9 + 12w + 12w + 16ww − 6w + 12 + 16w − 8ww − 6w + 12 + 16w − 8ww
= 33 + 22w + 22w
0 = 2(w + w) + 3
Thus 4u + 3 = 0, as required.
Alternate solution: The given circle is |z − 2| = 2 ⇒ z = 2 + 2eiθ . Substituting in
transformation expression,
2z + 3 4 + 4eiθ + 3 7 + 4eiθ (7 + 4eiθ )(e−iθ − 1)
w = = = =
z−4 2 + 2eiθ − 4 2(eiθ − 1) 2(eiθ − 1)(e−iθ − 1)
7e−iθ − 4eiθ − 3 7(cos θ − i sin θ) − 4(cos θ + i sin θ) − 3
= iθ −iθ
=
2(2 − e − e ) 2(2 − 2 cos θ)
3 cos θ − 3 − 11i sin θ 3 11 sin θ
= = − −i
4(1 − cos θ) 4 4(1 − cos θ)
Thus u = − 34 ⇒ 4u + 3 = 0, hence all points on the circle |z − 2| = 2 are mapped onto the
line 4u + 3 = 0.
Question 2(a) Using the Residue Theorem show that
Z ∞
x sin ax π −a
4
dx = e sin a (a > 0)
−∞ x + 4 2
Solution.

3
 Z
We consider I = f (z) dz where f (z) = as shown.
γ
zeiaz
and the contour γ consists of Γ a semi- Γ
z4 + 4
circle of radius R with center (0, 0) lying in
the upper half plane bounded by the real axis

Z (−R, 0) (0, 0) (R, 0)

Thus by Cauchy’s residue theorem, f (z) dz = 2πi(sum of residues at poles of f (z) inside
γ
γ). √ πi √ 3πi
4 (2n+1)πi
Clearly f
√ 5πi √ 7πi (z) has simple poles at z = 4e for
√ πi √ 3πi n = 0, 1, 2, 3, or z = 2e 4 , 2e 4 ,
2e 4 , 2e 4 . Out of these only the  poles 2e 4 , 2e 4 lie inside γ.
√ πi ze iaz √ πi αeiaα √ πi
Residue at 2e 4 is d 4 at z = 2e 4 , which is where α = 2e 4 = 1+i.
dz
(z + 4) 4α3
√ 3πi eiaβ √ 3πi
Residue at 2e 4 is where β = 2e 4 = −1 + i.
4β 2
Sum of these residues is
1 h eiaα eiaβ i 1 h eia(1+i) eia(−1+i) i
+ 2 = +
4 α2 β 4 2i (−2i)
−a   e−a sin a
e
= eia − e−ia =
8i 4
zeiaz dz e−a sin a
Z
Thus = 2πi . Now
γ z4 + 4 4
π iθ π
zeiaz dz Reiθ eiaRe R2 πR2
Z Z Z
= iRe iθ
dθ ≤ e−aR sin θ dθ ≤
Γ z4 + 4 0 z4 + 4 R4 − 4 0 R4 − 4

because
R |z 4 + 4| ≥ |z 4 | − 4 = R4 − 4 on Γ, and e−aR sin θ ≤ 1 as sin θ ≥ 0 on [0, π]. Thus
Γ
f (z) dz → 0 as R → ∞. Thus
∞
xeiax zeiaz e−a sin a
Z Z
dx = lim dz = 2πi
−∞ x4 + 4 R→∞ γ z4 + 4 4

Taking the imaginary parts of both sides, we get

Z ∞
x sin ax πe−a sin a
4
dx =
−∞ x + 4 2

as required.

4
Question 2(b) The function f (z) has a double pole at z = 0 with residue 2, a simple pole
at z = 1 with residue 2, is analytic at all other finite points of the plane and is bounded as
|z| → ∞. If f (2) = 5 and f (−1) = 2, find f (z).

Solution. Since f (z) has only poles as singularities in the extended complex plane, it is
well known that f (z) has to be a rational function. Since f (z) has a double pole at z = 0
φ(z)
and a simple pole at z = 1, it has to be of the form f (z) = 2 . where φ(z) is a
z (z − 1)
polynomial such that φ(0) 6= 0, φ(1) 6= 0. Moreover degree of φ(z) is ≤ 3 as we are given
that f (z) is bounded as z → ∞. Let φ(z) = a0 + a1 z + a2 z 2 + a3 z 3 . Then
a0 + 2a1 + 4a2 + 8a3
f (2) = 5 ⇒ =5 (1)
4
a0 − a1 + a2 − a3
f (−1) = 2 ⇒ =2 (2)
−2
(z − 1)φ(z)
Residue of f (z) at z = 1 is lim = φ(1). This value is given to be 2, so
z→1 z 2 (z − 1)

a0 + a1 + a2 + a3 = 2 (3)

(z − 1)(φ0 (z)) − φ(z)

 
1 d φ(z)
Residue of f (z) at z = 0 is given by at z = 0, or =
1! dz z − 1 (z − 1)2
−a1 − a0 . Since this is given to be 2,

−a0 − a1 = 2 (4)

Adding (2), (3) we get 2a0 + 2a2 = −2 ⇒ a2 = −1 − a0 . Substituting a2 = −1 − a0 , a1 =

−a0 −2 in (1), we get a0 −2a0 −4−4−4a0 +8a3 = 20 ⇒ 8a3 = 5a0 +28. Substituting in (3), we
have a0 − a0 − 2 − 1 − a0 + 5a08+28 = 2 ⇒ −3a0 + 28 = 40 ⇒ a0 = −4 ⇒ a1 = 2, a2 = 3, a3 = 1.
−4 + 2z + 3z 2 + z 3
Hence f (z) = is the desired function.
z 2 (z − 1)
Note: If f (z) has only poles in C∪∞, then it is a rational function. If φ1 (z), φ2 (z), . . . , φr (z)
are principal parts ofPfr (z) at the polesz1 , z2 , . . . , zr and ψ(z) is the principal part of f (z)
at ∞, then Prf (z) − j=1 φj (z) − ψ(z) being bounded and analytic in C ∪ ∞ is constant
⇒ f (z) = j=1 φj (z) + ψ(z) + C. Thus f (z) is a rational function, as each φj (z) is a ratio-
nal function and ψ(z) is a polynomial.

Question 2(c) What kind of singularities do the following functions have?

1
1. at z = 2πi.
1 − ez
1
2. at z = π4 .
sin z − cos z

5
 cot πz
3. at z = a and z = ∞. What happens when a is an integer (including a = 0)?
(z − a)2

Solution.
(z − 2πi)2 (z − 2πi)3
1. Clearly ez − 1 = ez−2πi − 1 = (z − 2πi) + + + . . ., showing that
2! 3!
1
ez − 1 has a simple zero at z = 2πi. Thus the given function 1−e z has a simple pole at

z = 2πi. Now residue at z = 2πi is given by

z − 2πi
lim = −1
z→2πi 1 − ez

1
2. f (z) = . We know that
sin z − cos z

2
1  π 1 z − π4 1  π
sin z = √ + z − √ − √ + . . . + Higher powers of z −
2 4 2 2! 2 4
π 2


1  π 1
 z− 4 1  π
cos z = √ − z − √ − √ + . . . + Higher powers of z −
2 4 2 2! 2 4

3
√  π  √ z− 4 π  π
⇒ sin z − cos z = 2 z− − 2 + . . . + Higher powers of z −
4 3! 4
π 1
Since sin z − cos z has a simple zero at z = , the given function has a
4 sin z − cos z
π
simple pole at z = .
4
π z − π4 1
Residue at z = is given by limπ =√ .
4 z→ 4 sin z − cos z 2
cot πz
3. f (z) = . f (z) has a simple pole at each z = n, n ∈ Z, n 6= a, with residue
(z − a)2
1
(n−a)2
. f (z) also has a pole at z = a, whose nature is as follows:

(a) a is not an integer and a 6= n + 12 .

In this case, cos πa 6= 0, sin πa 6= 0 and therefore f (z) has a double pole at z = a.
d
(The residue at z = a is [(z − a)2 f (z)]z=a = −π csc2 πa.)
dz
(b) a is not an integer and a = n + 21 .
In this case cos πz has a simple zero at a, and sin πz = ±1, therefore f (z) has a
cos πz 1 −π sin πa
simple pole at z = a. (The residue at z = a is lim = =
z→a z − a sin πa sin πa
−π.)

6
 (c) a is an integer.
sin πz has a simple zero at z = a and cos πa 6= 0, then f (z) has a triple pole at
z = a. The residue in this case is − π3 , because
3
 
a 3 (z − a)
sin πz = (−1) π(z − a) − π + Higher powers of(z − a)
3!
2
 
a 2 (z − a)
cos πz = (−1) 1 − π + Higher powers of(z − a)
2!
2
1 1 − π 2 (z−a)
2!
+ Higher powers of(z − a)
f (z) =
(z − a)2 π(z − a)[1 − π 2 (z−a)2 + Higher powers of(z − a)]
3!
2 2
  
1 2 (z − a) 2 (z − a)
= 1−π + ... 1 + π + ...
π(z − a)3 2! 3!
1
The coefficient of z−a in the Laurent series of f (z) (formed by multiplying the
π2 2
1
above series) is π − 2 + π6 = − π3 , which is the required residue.


(Note that the computation of residues was not required for this problem.)
Finally, f (z) has an essential singularity at ∞, because f (z) has zeros at z = n + 21 , a 6=
n + 12 whose limit point is ∞.

7
 UPSC Civil Services Main 2000 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 30, 2010

Question 1(a) Show that any four given points of the complex plane can be carried by a
bilinear transformation to positions 1, −1, k, −k where the value of k depends on the given
points.

Solution. It is known that a bilinear transformation mapping z1 , z2 , z3 to w1 , w2 , w3 is

given by the crossratio(z1 , z2 , z3 , z) = crossratio(w1 , w2 , w3 , w), i.e.

(z − z1 )(z2 − z3 ) (w − w1 )(w2 − w3 )
=
(z1 − z2 )(z3 − z) (w1 − w2 )(w3 − w)

Now w1 = 1, w2 = −1, w3 = k, so

(z − z1 )(z2 − z3 ) (w − 1)(−1 − k)
=
(z1 − z2 )(z3 − z) (2)(k − w)

It will map z4 to −k provided k is given by

(z4 − z1 )(z2 − z3 ) (−k − 1)(−1 − k) (k + 1)2

= =
(z1 − z2 )(z3 − z4 ) (2)(k − (−k) 4k

Clearly k depends on the points z1 , z2 , z3 , z4 .

Z
f (s) ds
Question 2(a) Suppose f (ζ) is continuous on a circle C. Show that as z varies
C s−z
inside C is differentiable under the integral sign. Find the derivative. Hence or otherwise
derive as integral representation for f 0 (z) if f (z) is analytic on and inside of C.

1
 Solution. If f (z) is analytic on and inside C, then by Cauchy’s integral formula
Z
1 f (s) ds
f (z) =
2πi C s − z
Let h be a complex number so chosen that z + h also lies in the interior of C. Then
 Z 
f (z + h) − f (z) 1 1 h f (s) f (s) i
= − ds
h 2πi h C s − z − h s − z
 Z 
1 1 hf (s)
= ds
2πi h C (s − z)(s − z − h)
f (z + h) − f (z)
Z Z h
1 f (s) 1 f (s) f (s) i
=⇒ − ds = − ds
h 2πi C (s − z)2 2πi C (s − z − h)(s − z) (s − z)2
Z
1 hf (s)
= ds
2πi C (s − z − h)(s − z)2

Let M = supz∈C |f (z)|, l = arc length of C, d = mins∈C |z − s| > 0. Since we are interested
in h → 0, we can assume that 0 < |h| < d. Now |s−z| ≥ d, |s−z −h| ≥ |s−z|−|h| ≥ d−|h|,
and therefore
f (z + h) − f (z) |h|
Z
1 f (s) M
− ds ≤ ·l
h 2πi C (s − z)2 2π d2 (d − |h|)

Since the right hand side of the above inequality → 0 as h → 0, it follows that

f (z + h) − f (z)
Z
1 f (s)
lim = ds
h→0 h 2πi C (s − z)2
i.e. Z
d(f (z)) 0 1 d  f (s) 
= f (z) = ds
dz 2πi C dz s − z
or Z Z
d 1 f (s) ds 1 d  f (s) 
= ds
dz 2πi C s − z 2πi C dz s − z
i.e. differentiation under the integral sign is valid. The representation for f 0 (z) is given by
Z
0 1 f (s) ds
f (z) =
2πi C (s − z)2

2
 UPSC Civil Services Main 2001 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 30, 2010

Question 1(a) Prove that the Riemann zeta function ζ defined by ζ(z) = ∞ −z
P
n=1 n con-
verges for Re z > 1 and converges uniformly for Re z > 1 +  where  is arbitrarily small.

Solution.
1 1 1 1 1
z
= x iy = x ∵ iy
= iy log n = 1
n n ·n n n e
∞
X 1
converges for x > 1, it follows that ∞ −z
P
Since x n=1 n converges absolutely for Re z > 1.
n=1
n
If Re z ≥ 1 + , then n1x ≤ n1+
1
and
∞ ∞
X
−z
X 1
|n | ≤ 1+
n=1 n=1
n

for Re z ≥ 1 + . Weierstrass’ M-test gives that the given series converges uniformly and
absolutely for Re z ≥ 1 + .
1
Question 2(a) Find the Laurent series for the function e z in 0 < |z| < ∞. Using the
expansion show that
1 π cos θ
Z
1
e cos(sin θ − nθ) dθ =
π 0 n!
n = 1, 2, . . ..
1
Solution. Clearly e z is analytic in 0 < |z| < ∞ and satisfies requirements of Laurent’s
expansion, and we have
∞ Z 1
1
X
n 1 ez
ez = an z , where an = n+1
dz (∗)
n=−∞
2πi |z|=1 z

1
Note — z = 0 is an essential singularity, therefore we have infinitely many terms with
negative exponents. In the expression for an we could have taken any disc, we have taken
|z| = 1 for convenience.
Put z = eiθ in (∗), dz = ieiθ dθ, we get
Z 2π cos θ−i sin θ
1 e
an = ieiθ dθ
2πi 0 ei(n+1)θ
Z 2π
1
= ecos θ e−i sin θ−inθ dθ
2π 0
Z 2π Z 2π
1 cos θ i
= e [cos(sin θ + nθ)] dθ − ecos θ [sin(sin θ + nθ)] dθ
2π 0 2π 0
cos θ cos θ
Z 2π g(θ) = e [sin(sin θ + nθ)], then g(2π − θ) = −e [sin(sin θ + nθ)] = −g(θ). Thus
Let
ecos θ [sin(sin θ + nθ)] dθ = 0.
0 Z 2π
1
Thus an = ecos θ [cos(sin θ + nθ)] dθ.
2π 0 Z 2π
1
In particular, a−n = ecos θ [cos(sin θ − nθ)] dθ for n = 1, 2, . . ..
2π 0
∞
1
X 1
But we know that e z = 1 + n
.
n=1
n!z
Therefore, comparing the two expansions we get for n = 1, 2, . . .,
Z 2π
1 1
ecos θ [cos(sin θ − nθ)] dθ =
2π 0 n!

Since ecos 2π−θ cos(sin(2π − θ) − n(2π − θ)) = ecos θ [cos(sin θ − nθ)], we can double the integral
and halve the limit to obtain
1 π cos θ
Z
1
e cos(sin θ − nθ) dθ =
π 0 n!

Z ∞
dx π
Question 2(b) Show that 4
=√ .
−∞ 1+x 2

Solution.
1
We take f (z) = 1+z 4 and the contour Γ
γ consisting of Γ a semicircle of radius R
with center (0, 0) lying in the upper half
plane, and the line joining (−R, 0) and (R, 0).
(−R, 0) (0, 0) (R, 0)
2
 Z
dz
By Cauchy’s residue theorem = 2πi(sum of residues at poles of f (z) in the upper
γ 1 + z4
half plane).
1 πi 3πi
Clearly 1+z 4 has two simple poles at z = e
4 and z = e 4 inside the contour.
πi 1 1
Residue at z = e 4 is d(z4 +1) = 3πi .
dz
4e 4
3πi 1 1
Residue at z = e 4 is 9πi = πi .
4e 4 4e 4

1 3π 3π π π
Sum of residues = cos − i sin + cos − i sin
4 4 4 4 4
1 π π π π
= − cos − i sin + cos − i sin
4 4 4 4 4
i 2 i
= − √ =− √
4 2 2 2
−i
Z
dz π
Thus lim = 2πi √ = √ .
R→∞ γ 1 + z4 2 2 2
Now Z Z π
dz R πR
4
≤ 4
dθ =
Γ 1+z 0 R −1 R4 − 1
iθ 4 4
Z z = Re and using |z + 1| ≥ R − 1 on Γ.
on putting
dz
Thus 4
→ 0 as R → ∞. Consequently,
Γ 1+z
Z Z ∞
dz dx π
lim 4
= 4
=√
R→∞ γ 1 + z −∞ 1 + x 2

as required.

3
 UPSC Civil Services Main 2002 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 30, 2010

Question 1(a) Suppose that f and g are two analytic functions on the set C of all complex
numbers with f ( n1 ) = g( n1 ) for n = 1, 2, 3, . . ., then show that f (z) = g(z) for all z ∈ C.

Solution. Let G(z) = f (z) − g(z), then G( n1 ) = 0 for n = 1, 2, . . .. We shall show that
G(z) ≡ 0 for z ∈PC whichn would prove the result.
Let G(z) = ∞ n=0 an z be the power series of G(z) with center 0 and radius of convergence
R, clearly R > 0. We shall now prove that an = 0 for every n.
If an 6= 0 for some n, let ak be the first non-zero coefficient. Then

G(z) = z k (ak + ak+1 z + . . .) = z k H(z)

Clearly H(z) is analytic in |z| < R, and H(0) 6= 0. We now claim that H(z) 6= 0 in a
neighborhood |z| < δ of 0. Let  = |H(0)| 2
, then continuity of H(z) at z = 0 implies that there
exists a δ > 0 such that |z| < δ ⇒ |H(z) − H(0)| <  or |H(0)| −  < |H(z)| < |H(0)| + 
for |z| < δ. Thus |H(z)| > |H(0)| 2
> 0 for |z| < δ. Consequently, G(z) 6= 0 for any z in
0 < |z| < δ. But this is not possible, as |z| < δ contains all but finitely many n1 , at which
G(z) vanishes. Thus our assumption that an 6= 0 for some n is false, thus G(z) ≡ 0 in
|z| < R.
Let z 0 be any point in C, and let r(t), a ≤ t ≤ b be a continuous curve joining 0 and z 0 .
Using uniform continuity of r(t), we get a partition a = t0 < t1 < . . . < tn = b of [a, b] such
that r(t0 ) = 0, r(t1 ) = z1 , . . . , r(tn ) = r(b) = z 0 , and |zj − zj−1 | < R.
Now the disc K0 = |z − 0| < R contains z1 , the center of disc K1 = |z − z1 | < R. Since
G(z1 ) = 0 as z1 ∈ K0 ∩ K1 , and K0 ∩ K1 contains a sequence of points yn such that yn → z1
and G(yn ) = 0, we can prove as before that G(z) ≡ 0 in K1 . Proceeding in this way, in
n steps we get G(z) ≡ 0 in Kn , or G(z 0 ) = 0. Since z 0 is an arbitrary point of C, we get
G(z) ≡ 0 in C.

1
 z
Question 2(a) Show that when 0 < |z − 1| < 2, the function f (z) = has the
(z − 1)(z − 3)
Laurent series expansion in powers of (z − 1) as
∞
−1 X (z − 1)n
−3
2(z − 1) n=0
2n+2

Solution. Let ζ = z − 1, so that

z ζ +1 1 3
f (z) = = =− +
(z − 1)(z − 3) ζ(ζ − 2) 2ζ 2(ζ − 2)

3 3 −1  ζ −1 ζ
Now for 0 < |ζ| < 2, = · · 1− and < 1. Consequently,
2(ζ − 2) 2 2 2 2
∞
3 3 X  ζ n
=−
2(ζ − 2) 4 n=0 2

and ∞ ∞
1 3 X  ζ n 1 X (z − 1)n
f (z) = − − =− −3
2ζ 4 n=0 2 2(z − 1) n=0
2n+2
which is the desired Laurent series expansion.

Question 2(b) Establish by contour integration

Z ∞
cos(ax) π
2
dx = e−a , where a ≥ 0
0 x +1 2

Solution. Let I be the given integral. Put ax = t, so that

Z ∞ Z ∞
cos t dt cos t
I= t2 = a 2 + a2
dt
0 a 2 + 1 a 0 t
Z ∞
cos t π
We shall now prove that 2 2
dt = e−a , which will show that I = π2 e−a as required.
0 t +a 2a
ix
cos x e
Clearly 2 2
is the real part of 2 .
x +a R x + a2
We consider the integral γ f (z) dz where
eiz Γ
f (z) = 2 2
and γ is the contour consist-
z +a
ing of the line joining (−R, 0) and (R, 0) and
Γ, which is the arc of the circle of radius R
and center (0, 0) lying in the upper half plane. (−R, 0) (0, 0) (R, 0)

2
 Clearly on Γ, if we put z = Reiθ , then 0 ≤ θ ≤ π and
π iθ π iθ
eiz dz Rieiθ eiRe dθ Rieiθ eiRe
Z Z Z
= ≤ dθ
Γ z 2 + a2 0 R2 e2iθ + a2 0 R2 e2iθ + a2
iθ
But |eiRe | = |eiR cos θ e−R sin θ | = e−R sin θ ≤ 1 as sin θ ≥ 0 for 0 ≤ θ ≤ π. |z 2 +a2 | ≥ |z|2 −a2 =
R2 − a2 . Therefore Z π
eiz dz
Z
R πR
2 2
≤ 2 2
dθ = 2
Γ z +a 0 R −a R − a2
eiz dz
Z
Hence 2 2
→ 0 as R → ∞.
ΓZz + a
eiz dz
Now 2 2
= 2πi(sum of residues at poles inside γ).
γ z +a
But the only pole in the upper half plane is z = ia, (a > 0) and the residue at z = ia is
ei(ia) e−a
= . Thus
2ia 2ia
Z ∞ ix
eiz dz e−a πe−a
Z
e dx
lim = = 2πi · =
R→∞ γ z 2 + a2 2
−∞ x + a
2 2ia a
Z ∞ −a Z ∞
cos x dx πe sin x dx
=⇒ 2 2
= , 2 2
=0
−∞ x + a a −∞ x + a
Z ∞
cos x dx πe−a
=⇒ = ∵ cos x = cos(−x)
0 x 2 + a2 2a

This completes the proof.

3
 UPSC Civil Services Main 2003 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 16, 2010

Question 1(a) Determine all the bilinear transformations which transform the unit circle
|z| ≤ 1 into the unit circle |w| ≤ 1.
az + b
Solution. Let the required transformation be w = . Clearly z = − ab ⇒ w = 0 and
cz + d
z = − dc ⇒ w = ∞. Since 0, ∞ are inverse points with respect to the circle |w| = 1, then
− ab , − dc are inverse points with respect to the circle |z| = 1 (note that R, S different from 0
are said to be inverse points with respect to |z| = 1 if O, R, S are collinear and OR · OS = 1).
Thus if we set − ab = α, then − dc = α1 and we get
a z−α aα z − α
w= 1 =
c z−α c αz − 1

aα 1 − α
Since |z| = 1 maps onto |w| = 1, we take z = 1 to get = 1. But |1 − α| = |1 − α|,
c α−1
aα aα
therefore = 1. Let = eiθ , θ ∈ R, so that
c c
z−α
w = eiθ
αz − 1
We now check that when |z| = 1, we have |w| = 1.
z−α
|w| = |eiθ |
αz − 1
z−α
= |z| (∵ zz = 1)
α−z
= 1 (∵ |z − α| = |α − z|)

1
Now let |z| < 1. Then
z−α z−α
ww − 1 = eiθ · e−iθ −1
αz − 1 αz − 1
zz − αz − αz + αα
= −1
(αz − 1)(αz − 1)
zz − αz − αz + αα − ααzz + αz + αz − 1
=
(αz − 1)(αz − 1)
zz + αα − ααzz − 1
=
|αz − 1|2
(zz − 1)(1 − αα)
=
|αz − 1|2

Thus if |α| < 1, then |w| < 1. This shows that the transformation
z−α
w = eiθ , θ ∈ R, |α| < 1
αz − 1
maps the interior of |z| = 1 onto the interior of |w| = 1 and the boundary of |z| = 1 onto
the boundary of |w| = 1. Thus all bilinear transforms which map |z| ≤ 1 onto |w| ≤ 1 are
given by
z−α
w = eiθ , θ ∈ R, |α| < 1
αz − 1
Note: If |α| > 1, then the interior of |z| = 1 would map onto the exterior of |w| = 1. The
boundary will map onto the boundary, as before.
 z − ic 2
Question 2(a) 1. Discuss the transformation W = , c real, showing that the
z + ic
upper half of the W -plane corresponds to the interior of a semicircle lying to the right
of the imaginary axis in the z-plane.

2. Using the method of contour integration prove that

Z π
a dθ π
2 2 =√ (a > 0)
0 a + sin θ 1 + a2
Solution.

1. We need to assume c > 0 as otherwise the question is incorrect.

2
 Let W = U + iV , so that
 2
x + i(y − c)
U + iV =
x + i(y + c)
 2
(x + i(y − c))(x − i(y + c))
=
x2 + (y + c)2
 2 2
x + y 2 − c2 − 2icx
=
x2 + (y + c)2
(x2 + y 2 − c2 )2 − 4c2 x2 − 4icx(x2 + y 2 − c2 )
=
[x2 + (y + c)2 ]2
(x2 + y 2 − c2 )2 − 4c2 x2
⇒U =
[x2 + (y + c)2 ]2
−4cx(x2 + y 2 − c2 ) 4cx(c2 − x2 − y 2 )
V = =
[x2 + (y + c)2 ]2 [x2 + (y + c)2 ]2

Thus if z belongs to the interior of the semicircle given by x2 + y 2 = c2 , x ≥ 0, then

V > 0, which means that U + iV is in the upper half plane.
(y 2 − c2 )2  y − c 2
For any point on the line x = 0, we have V = 0 and U = = .
(y + c)4 y+c
Clearly when y changes from −c to c, U changes from ∞ to 0.
As z moves over the circle x2 + y 2 = c2 , we have V = 0 and

−4c2 x2 −4c2 x2 −4c2 x2 −x2 c2 − y 2 c−y

U= 2 2 2
= 2 2 2 2
= 2 2
= 2
= − 2
=−
(x + (y + c) ) (x + y + c + 2yc) (2c + 2yc) (y + c) (y + c) c+y

1 − cos θ
Let y = c cos θ, then U = − = − tan2 2θ . When y moves from −c to c, i.e.z
1 + cos θ
traverses the boundary of the semicircle, θ varies from π to 0, and U varies from −∞
to 0. Thus the boundary of the semicircle x2 + y 2 = c2 with x ≥ 0 is mapped onto the
U -axis. Hence the semicircle x2 + y 2 = c2 with x ≥ 0 is mapped onto W = U + iV
with V ≥ 0.

2. Let the given integral be I. Then

Z π Z π Z 2π
a dθ 2a dθ a dφ
I= 2 = 2
= 2
0 2a + (1 − cos 2θ) 2a + 1 − cos φ
2
0 a + sin θ 0

on putting 2θ = φ. We now let z = eiφ to obtain

Z Z Z
a dz 1 2a dz 2ai dz
I= = =
iz(2a2 + 1 − 12 (z + z1 )) i 2(2a2 + 1)z − (z 2 + 1) z2 − 2(2a2 + 1)z + 1
|z|=1 |z|=1 |z|=1

3
 p √
Now z 2 − 2(2a2 + 1)z + 1 = 0 ⇒ z = 2a2 + 1 ± (2a2 + 1)2 − 1 = 2a2 + 1 ± 2a a2 + 1.
√ √
Clearly |2a2 + 1 + 2a a2 + 1| > 1 showing that |2a2 + 1 − 2a a2 + 1| < 1 because√ the
product of the roots is 1. Thus the only pole inside |z| = 1 is z = 2a2 + 1 − 2a a2 + 1.
√ 1
Residue at z = 2a2 +1−2a a2 + 1 is √ √ =
(2a + 1 − 2a a + 1) − (2a2 + 1 + 2a a2 + 1)
2 2
1
√ .
−4a a2 + 1
1 π
Thus I = 2ai · 2πi · √ =√ .
−4a a2 + 1 a2 + 1

4
 UPSC Civil Services Main 2004 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 16, 2010

4
Question 1(a) Find the image of the line y = x under the mapping w = and draw
z2 + 1
it. Find the points where this transformation ceases to be conformal.
Solution. Let z = x + iy. Then
4 4
w = = 2
z2
+1 2
x − y + 1 + 2ixy
4
= 2 2 2 2 2
[x2 − y 2 + 1 − 2ixy]
(x − y + 1) + 4x y
4 − 8ix2 4 −8x2 1
So if x = y, w = 4
. Let u = 4
, v = 4
⇒ u2 + v 2 = 16 = 4u ⇒
1 + 4x 1 + 4x 1 + 4x 1 + 4x4
1
(u − 2)2 + v 2 = 4, v ≤ 0. So the image of the line x = y under the mapping w = 2 is a
z +1
semicircle with center (2, 0), radius 2 and below the x-axis.

(2, 0)

dw 8z dw
Conformality: =− 2 2
when z 6= ±i. Clearly 6= 0 when z 6= 0. Thus the
dz (z + 1) dz
dw
mapping is conformal at all points which are different from z = 0, ±i (as does not exist
dz
at ±i).

1
Question 2(a) If all zeros of a polynomial P (z) lie in a half plane, then show that zeros of
the derivative P 0 (z) also lie in the same half plane.

Solution. We can assume Q without loss of generality that the zeros of P (z) lie in the half
plane Re z < 0. Let P (z) = nj=1 (z − αj ) where αj = xj + iyj , xj < 0.
If Re z ≥ 0, then P (z) 6= 0 and
n
P 0 (z) X 1
=
P (z) j=1
z − αj
n
X 1
=
j=1
x − xj + i(y − yj )
n
X x − xj − i(y − yj )
=
j=1
(x − xj )2 + (y − yj )2

Since xj < 0, 1 ≤ j ≤ n, it follows that

 0  X n
P (z) x − xj
Re = 2 + (y − y )2
>0
P (z) j=1
(x − x j ) j

P 0 (z)
whenever Re z = x ≥ 0. Thus and therefore P 0 (z) has no zeros in the right half plane
P (z)
Re z ≥ 0. Hence all zeros of P 0 (z) lie in the same half plane in which the zeros of P (z) lie.

Question 2(b) Using Contour integration, evaluate

Z 2π
cos2 3θ
dθ, 0 < p < 1
0 1 − 2p cos 2θ + p2
Solution. Clearly
2π 2π
cos2 3θ
Z Z
1 1 + cos 6θ
2
dθ = dθ
0 1 − 2p cos 2θ + p 2 0 1 − 2p cos 2θ + p2
1 + ei6θ dz
The integrand is the real part of 2
. Put z = eiθ , dz = ieiθ dθ or dθ = .
1 + 2p cos 2θ + p iz
2π
1 + ei6θ 1 + z6
Z Z
1 dz
2
dθ = 2 1 2
0 1 + 2p cos 2θ + p i |z|=1 1 − p(z + z2 ) + p z
z(1 + z 6 )
Z
1
= dz
i |z|=1 −pz 4 + z 2 (1 + p2 ) − p
z(1 + z 6 )
Z
1
= dz
i |z|=1 (1 − pz 2 )(z 2 − p)

2
 √
Now the integrand has simple poles at z = ± p, ± √1p . Since 0 < p < 1, the only poles inside
√ √
|z| = 1 are z = ± p. The residue at z = p is
√ √
(z − p)z(1 + z 6 ) p(1 + p3 ) 1 + p3
lim
√ = √ =
z→ p (1 − pz 2 )(z 2 − p) (1 − p2 )2 p 2(1 − p2 )
√
Similarly residue at z = − p is
√ √
(z + p)z(1 + z 6 ) − p(1 + p3 ) 1 + p3
lim√ = √ =
z→− p (1 − pz 2 )(z 2 − p) (1 − p2 )(−2 p) 2(1 − p2 )

Thus
2π Z 2π
cos2 3θ 1 + ei6θ
Z 
1
dθ = Re dθ
0 1 − 2p cos 2θ + p2 2 0 1 + 2p cos 2θ + p2
 
1 √
= Re 2πi[Sum of residues at z = ± p]
2i
1 + p3 1 − p + p2
= π = π
1 − p2 1−p

3
 UPSC Civil Services Main 2005 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 16, 2010

Paper II

Question 1(a) If f (z) = u + iv is an analytic function of the complex variable z and

u − v = ex (cos y − sin y), determine f (z) in terms of z.

Solution. Let F (z) = (1 + i)f (z) = (1 + i)(u + iv) = (u − v) + i(u + v). Now

∂ 2 (u − v) ∂ 2 (u − v)
+ = ex (cos y − sin y) + ex (− cos y + sin y) = 0
∂x2 ∂y 2

Let F (z) = U + iV , where U = u − v is harmonic. If f (z) is analytic, then so is F (z) and

dF ∂U ∂V ∂U ∂U
= +i = −i
dz ∂x ∂x ∂x ∂y
∂V ∂U
as =− by the Cauchy-Riemann equations. Thus
∂x ∂y

F 0 (z) = ex (cos y − sin y) − iex (− sin y − cos y)

= ex (cos y + i sin y) + iex (cos y + i sin y)
= (1 + i)ex · eiy = (1 + i)ez

Thus F (z) = (1 + i)ez and f (z) = ez , which is the required function.

1
 1
Question 2(a) Expand f (z) = in Laurent’s series which is valid for (i)
(z + 1)(z + 3)
1 < |z| < 3 (ii) |z| > 3 and (iii) |z| < 1.
1 1 1 
Solution. Clearly f (z) = −
2 z+1 z+3
(i) 1 < |z| < 3. In this region
1h1 1 −1 1  z −1 i
f (z) = 1+ − 1+
2 z z 3 3
Since | z1 | < 1 and | z3 | < 1, we get
∞ ∞
1 X n 1 1X  z n
f (z) = (−1) n − (−1)n
2z n=0 z 6 n=0 3
∞ ∞
X (−1)n 1 X (−1)n z n
= −
n=0
2 z n+1 n=0 2 3n+1

as Laurent’s expansion in the region 1 < |z| < 3.

(ii) |z| > 3. In this region
1 h 1 −1  3 −1 i
f (z) = 1+ − 1+
2z z z
Now | z1 | < 1 and | z3 | < 1, so we get
∞ ∞
1 X 1 1 X 3n
f (z) = (−1)n n − (−1)n n
2z n=0 z 2z n=0 z
∞
X (−1)n 1 − 3n
=
n=0
2 z n+1

as Laurent’s expansion in the region |z| > 3. This is Taylor’s expansion of f (z) around ∞.
(iii) |z| < 1. In this region
1h 1 z −1 i
f (z) = (1 + z)−1 − 1+
2 3 3
Now |z| < 1, | z3 | < 1, so we get
∞ ∞
1X n n 1X zn
f (z) = (−1) z − (−1)n n
2 n=0 6 n=0 3
∞
X (−1)n  1 
= 1 − n+1 z n
n=0
2 3

as Laurent’s expansion valid in |z| < 1. This has no negative powers of z as f (z) is analytic
in |z| < 1.

2
 UPSC Civil Services Main 2006 - Mathematics
Complex Analysis
Sunder Lal
Retired Professor of Mathematics
Panjab University
Chandigarh

January 16, 2010

Question 1(a) Determine all the bilinear transformations which map the half plane Im(z) ≥
0 into the unit circle |w| ≤ 1.

Solution. Since the x-axis is to be mapped onto the circle |w| = 1, we determine conditions
on a, b, c, d where
az + b
w=
cz + d
is the desired bilinear transformation, such that points z = 0, z = 1, z = ∞ are mapped onto
points with modulus 1.
First of all, c 6= 0, because if c = 0 then the image of z = ∞ would be w = ∞, which is
not possible.
Clearly z = ∞ is mapped onto w = ac , thus we must have | ac | = 1 or |a| = |c| = 6 0.
b
When z = 0, w = d (note that d 6= 0, because otherwise z = 0 would be mapped onto
∞), thus |w| = 1 gives us |b| = |d| = 6 0.
a z + ab
 
az + b z − z0 a
Since a 6= 0, c 6= 0, we can write w = = d
= eiα with eiα = , z0 =
cz + d c z+ c z − z1 c
b d b d
− , z1 = − . But = , so |z0 | = |z1 |. Thus we have proved that w can be written in
a c a c
the form  
iα z − z0
w=e with α ∈ R, |z0 | = |z1 |
z − z1

1
We now use the fact that the image of z = 1 has modulus 1, and get

|1 − z0 | = |1 − z1 |
⇒ (1 − z0 )(1 − z0 ) = (1 − z1 )(1 − z1 )
⇒ 1 − z0 − z0 + z0 z0 = 1 − z1 − z1 + z1 z1
⇒ z0 + z0 = z1 + z1 ∵ |z0 | = |z1 | ⇒ z0 z0 = z1 z1
⇒ Re(z0 ) = Re(z1 )

This gives us z0 = z1 or z0 = z1 because if z0 = x + iy0 , z1 = x + iy1 , then x2 + y02 = x2 + y12 ⇒

y02 = y12 ⇒ y1 = ±y0 . If z0 = z1 , then w = eiα , a constant,  which isnot possible, therefore
z − z0
z1 = z0 and the transformation w can be written as w = eiα .
z − z0
This transformation maps z0 to w = 0. Since we require the upper half plane Im(z) > 0 to
be mapped onto the interior of |w| = 1, we must have Im(z0 ) > 0. Thus any transformation
which maps the real axis onto |w| = 1 and the region Im(z) > 0 to the interior of |w| = 1
can be written in the form
 
iα z − z0
w=e , α ∈ R, Im(z0 ) > 0
z − z0
 
iα z − z0
We now prove the converse — any bilinear transformation w = e , α ∈
z − z0
R, Im(z0 ) > 0 maps Im(z) > 0 to |w| ≤ 1.
If z is such that Im(z) ≥ 0, then it can be seen easily that |z − z0 | < |z − z0 |, therefore
|w| < 1. Similarly if we assume that Im(z) < 0, then |z − z0 | > |z − z0 | and therefore |w| > 1.
Clearly when z lies on the real axis, then |w| = 1 as |z − z0 | = |z − z0 |. This proves the
result.
Hence all bilinear transformations of the required type are of the form
 
iα z − z0
n o
w=e , α ∈ R, Im(z0 ) > 0
z − z0
Alternate Solution: Since the x-axis is mapped to the unit circle |w| = 1, the
reflection of the image of z in |w| = 1 is the same as the image of the reflection of z in the
x-axis i.e. z. Thus
  
az + b az + b
1 =
cz + d cz + d
⇒ (cz + d)(cz + d) = (az + b)(az + b)
⇒ ccz 2 + (cd + cd)z + dd = aaz 2 + (ab + ab)z + bb

Comparing coefficients of the powers of z we have |a| = |c|, |b| = |d|, Re (ba) = Re (dc), thus
since ad − bc 6= 0, we have |a| = |c| =
6 0, |b| = |d| = ba
6 0, and Re ( ab ) = Re ( aa ) = Re ( dc
cc
)=
d
Re ( c ).

2
 a z + ab
 
az + b iα z − z0 iα a b
Thus we can write w = = = e with e = , z0 = − , z1 =
cz + d c z + dc z − z1 c a
d
− . Thus Re z0 = Re z1 , |z0 | = |z1 |, hence z0 = z1 or z0 = z1 . The former is not possible as
c
it would make ad − bc = 0, hence z0 = z1 . For the same reason, Im z0 6= 0. So we have
 
iα z − z0
w=e , α∈R
z − z0
Now in addition, since the upper half plane Im(z) > 0 is mapped onto the interior of |w| = 1,
and the image of z0 is 0, and thus inside the unit circle, so z0 is in the upper half plane,
hence Im(z0 ) > 0.
Hence the required set of bilinear transformations is
 
iα z − z0
n o
w=e , α ∈ R, Im(z0 ) > 0
z − z0
The converse is proved as above.

Question 2(a) With the aid of residues, evaluate

Z π
cos 2θ
2
dθ, −1 < a < 1
0 1 − 2a cos θ + a

Solution. Let
Z π
cos 2θ
I = dθ
1 − 2a cos θ + a2
Z0 π
(cos 2θ)(1 + 2a cos θ + a2 )
= dθ
(1 + a2 )2 − 4a2 cos2 θ
Z0 π Z π
(cos 2θ)(1 + a2 ) 2a cos θ cos 2θ
= 2 2 2 2
dθ + 2 2 2 2
dθ
0 (1 + a ) − 4a cos θ 0 (1 + a ) − 4a cos θ

Since cos(π − θ) = − cos θ, on puuting π − θ = α we get

Z π Z 0
2a cos θ cos 2θ 2a(− cos α) cos 2α
2 2 2 2
dθ = 2 2 2 2
(−dα)
0 (1 + a ) − 4a cos θ π (1 + a ) − 4a cos α
Z π
2a cos θ cos 2θ
showing that 2 2 2 2
dθ = 0. Thus
0 (1 + a ) − 4a cos θ
Z π
(cos 2θ)(1 + a2 )
I= 2 2 2
dθ
0 (1 + a ) − 2a (1 + cos 2θ)

Putting 2θ = β, we get
1 2π (cos β)(1 + a2 ) 1 2π (1 + a2 ) cos β
Z Z
I= dβ = dβ
2 0 (1 + a2 )2 − 2a2 (1 + cos β) 2 0 1 + a4 − 2a2 cos β

3
 dz
We now put z = eiβ , so that dz = iz dβ or dβ = .
iz
(1 + a2 ) 12 (z + z1 ) dz
Z
1
I = 4 2 1
2 |z|=1 1 + a − a (z + z ) iz
1 + a2 z2 + 1
Z
= 
4 2 2 2
 dz
4i |z|=1 z (1 + a )z − a z − a
1 + a2 −(z 2 + 1)
Z
=  dz
4a2 i |z|=1 z z 2 − (a2 + a12 )z + 1


(1 + a2 )i z2 + 1
Z
= 1 dz
4a2 2
|z|=1 z(z − a )(z − a2 )

Clearly the integrand has simple poles at z = 0, z = a2 , z = a12 , out of which z = 0 and
z = a2 lie inside |z| = 1 as −1 < a < 1.
(z 2 + 1)z
Residue at z = 0 is lim = 1.
z→0 z(z − a2 )(z − 12 )
a
(z 2 + 1)(z − a2 ) a4 + 1
Residue at z = a2 is lim = .
z→0 z(z − a2 )(z − 12 ) a 4−1
a
Cauchy’s residue theorem (the integral around a curve = 2πi· sum of residues at poles
inside the curve) now gives us

(1 + a2 )i a4 + 1 (1 + a2 ) · 2a4 πa2
 
I= · 2πi 1 + = −2π =
4a2 a4 − 1 4a2 (a4 − 1) 1 − a2

Question 2(b) Prove that all the roots of z 7 − 5z 3 + 12 = 0 lie between the circles |z| = 1
and |z| = 2.

Solution. Let g(z) = z 7 − 5z 3 , f (z) = 12, then

1. On |z| = 1, |g(z)| ≤ |z 7 | + 5|z 3 | = 6 < 12 = |f (z)|.

2. Both g(z) and f (z) are analytic on and within |z| < 1

3. Both g(z) and f (z) have no zeros on |z| = 1

By Rouche’s theorem, f (z) + g(z) = z 7 − 5z 3 + 12 and f (z) have the same number of zeros
inside |z| = 1. But f (z) = 12 has no zeros anywhere and in particular in the region |z| < 1,
therefore z 7 − 5z 3 + 12 has no zeros inside the unit circle.
Now we take g(z) = 12 − 5z 3 , f (z) = z 7 .

1. On |z| = 2, |g(z)| ≤ 12 + 5|z 3 | = 52 < 27 = |f (z)|.

4
 2. Both g(z) and f (z) are analytic on and within |z| < 2

Therefore by Rouche’s theorem, g(z) + f (z) = z 7 − 5z 3 + 12 and f (z) have the same number
of zeros inside |z| = 2. Since f (z) = z 7 has 7 zeros (z = 0 is a zero of order 7 of f (z)) inside
|z| = 2, the given polynomial has seven zeros inside |z| = 2 i.e. all its zeros lie inside |z| = 2.
Since z 7 − 5z 3 + 12 has no zeros inside and on |z| = 1, therefore all zeros lie in the ring
1 < |z| < 2.