Zeros of analytic functions

By a zero of a function f(z) we mean a value of z say z=a such that f(a)=0. If f(a)=0, f (a )  0 ,

f (a )  0, …….., f n 1
(a)  0 but f n (a)  0 , then z=a is called a zero of order or multiplicity n of
f(z). For example the function f ( z )  z n has a zero of order n at z=0 because f ( z )  nz n 1 ,
f ( z )  n(n  1) z n2 ,......, f n1 ( z )  n(n  1)(n  2)......2 z, f n ( z )  n(n  1)(n  2)......2.1 so
that f (0)  0, f (0)  0,......., f n 1
(0)  0 but f n (0)  n! 0 .
Note that a zero of order n is counted as n zeros.
Theorem. The zeros of an analytic function are isolated i.e. if f(z) is an analytic function and z=a is a zero
of f(z) , then there exists a neighborhood of the point a say z  a   which contains no zero of f(z) other
than a.
Proof. Let f(z) be an analytic function of z and let z=a be any zero of order n of f(z). Then by Taylor’s
theorem

( z  a) 2
f ( z )  f (a)  ( z  a) f (a)  f (a)  .......... ,
2!
the expansion being valid in a neighborhood of a.

since z=a is a zero of order n of f(z), f(a)=0= f (a)  f (a)  .......  f n1
(a) , but f n (a)  0 .
Therefore

f n (a) f n1 (a)

f ( z )  ( z  a) [ n
 ( z  a)  ..........]  ( z  a) n  ( z ) ,
n! (n  1)!

f n (a) f n1 (a)

where  ( z )   ( z  a)  .......... is analytic and  ( a )  0 .
n! (n  1)!

Since  (a)  0,  (a) is a positive real number say  (a)   .

Now ,  (z ) being analytic is differentiable and hence continuous at z=a. Therefore for the above   0,
there exists a positive number  such that  ( z )   (a)   for z  a   .

 
Hence  ( z )   ( z )   (a )   (a )   (a )   ( z )   (a )      0 i.e.  ( z )  0 for
2 2
z  a   . In other words,  (z ) has no zero in z  a   . Since the zeros of  (z ) are also the zeros
of f(z), it follows that f(z) has no zero in z  a   other than a. That proves the result.

Identity theorem or Uniqueness theorem

Statement. Let f(z) be analytic in a domain D having zeros at the points z1 , z 2 ,......, z n ,........ in D with
lim n z n  z 0  D . Then f ( z )  0 in D.

Proof. Suppose that f ( z )  0 identically in D. Then f(z) being analytic is continuous in D and therefore

f ( z 0 )  f (lim n z n )  lim n f ( z n )  lim n (0)  0 shows that z0 is a zero of f(z). By the
previous theorem this zero of f(z0 must be isolated. Hence there exists a positive number  such that
f(z) has no zero in z  z 0   other than z0 . In other words , the neighborhood z  z 0   of z0
contains no point of the sequence z1 , z 2 ,......, z n ,........ . But z0 is the limit point of the set of points of
the sequence z1 , z 2 ,......, z n ,........ . Therefore , by definition, every neighborhood of z0 must contain an
infinite number of points of the sequence .Thus, we arrive at a contradiction and this contradiction
proves the result.
Corollary. Let f(z) be analytic inside and on a simple closed curve C and let f(z) vanish on an arc inside
C.
Then f ( z )  0 .

Proof. This follows from the fact that we can choose a sequence z1 , z 2 ,......, z n ,........ on the arc whose
limit belongs to C and then the result follows from the above theorem.
Remark. We usually refer to the above corollary as the identity theorem.

Example . sin 2 z  cos 2 z  1 for all complex z.

Proof. Consider the function f ( z )  sin 2 z  cos 2 z  1 . Then f(z) being the sum of three entire
1 1
functions is an entire function and hence analytic in a domain D containing the points 0,1, , ,....,
2 3
1
,.... .
n

Further , we know that for all real x, sin 2 x  cos 2 x  1  0 so that

 f (1)  sin 2 1  cos 2 1  1  0
1 1 1
f ( )  sin 2  cos 2  1  0
2 2 2
1 1 1
f ( )  sin 2  cos 2  1  0
3 3 3
...............................................
...............................................
...............................................
1 1 1
f ( )  sin 2  cos 2  1  0
n n n
.................................................
1
Since lim n  0  D , it follows , by the identity theorem, that f ( z )  0 in D i.e.
n
sin 2 z  cos 2 z  1  0 or sin 2 z  cos 2 z  1, z .
Argument Principle
Statement. Let f(z) be analytic inside and on a simple closed curve C and let f(z) have no zero on C.
Then
1 f ( z )
2i  f ( z )
dz  number of zeros of f(z) inside C(each zero counted according to its multiplicity).

Proof. Let f(z) have zeros z1 , z 2 ,......, z m of multiplicities k1 , k 2 ,......, k m inside C. Then the number of
zeros of f(z) inside C is n  k1  k 2  ......  k m .

Further

f ( z )  ( z  z1 ) k1 ( z  z 2 ) k 2 ........( z  z m ) km . ( z ) ,

where  (z ) has no zeros has no zeros inside and on C.

Taking logarithm on both sides, we get

log f ( z )  k1 log( z  z1 )  k 2 log( z  z 2 )  ........  k m log( z  z m )  log . ( z ) .

Differentiating both sides w.r.to z, we get

f ( z ) k k k  ( z )
 1  2  .........  m  .
f ( z ) z  z1 z  z 2 z  z m  ( z)

1
Multiplying both sides by and integrating over C, we get
2i
 1 f ( z ) 1 k1 1 k2 1 km 1  ( z )
2i C f ( z ) 2i C z  z1 2i C z  z 2 2i C z  z m 2i C  ( z )
dz  dz  dz .......  dz  dz

1 1 1 1 1 1 1  ( z )
 k1 . 
2i C z  z1
dz k 2 . 
2i C z  z 2
dz .......  k m . 
2i C z  z m
dz 
2i C  ( z )
dz

Since  (z ) is analytic and not equal to zero inside and on C,  (z ) is also analytic inside and on C and
 ( z )  ( z )
hence so is . Therefore, by Cauchy’s theorem,  dz  0 .
 ( z) C
 ( z )

1
Also , for j=1,2,…..,m, zz
C j
dz  2i , as proved earlier in connection with Cauchy’s integral formula.

Hence, it follows from the above equation that

1 f ( z )
2i C f ( z )
dz k1  k 2  ......  k m  0  n  number of zeros of f(z) inside C and the proof is

complete.

f ( z )
Remark. Note that 
C
f ( z)
dz   C log f ( z ) , where  C log f ( z ) denotes the variation of logf(z) over

C so that  C log f ( z )  2in , where n is the number of zeros of f(z) inside C.

We can use the argument principle to evaluate certain integrals.

Examples

1. Let C be the circle z  a  r and f(z)=z-a. Then f(z) has only one zero z=a inside C and no zero on
C. Also f(z) is analytic inside and on C with f ( z )  1 .

1 f ( z )
Therefore,  z  a dz  
C C
f ( z)
dz  2i.1  2i .n

2z  3
2. Consider the integral z
C
2
 3z  2
dz , where C is the circle z  1  2 . Taking f ( z )  z 2  3z  2 ,

we see that f(z) is analytic inside and on C having 2 simple zeros z=1 and z=2 inside C and no zero on C.
Also f ( z )  2 z  3 . Therefore, by argument principle,

2z  3
z
C
2
 3z  2
dz  2i.2  4i.

Rouche’s Theorem
Statement. Let f(z) and g(z) be any two functions analytic inside and on a simple closed curve C such
that g ( z)  f ( z) on C. Then f (z ) and f ( z )  g ( z ) have the same number of zeros inside C.
Proof. We first note that f ( z )  0 on C. For, if f ( z )  0 on C, then according to the hypothesis

g ( z )  f ( z)  0 i.e. g ( z )  0 on C, which is absurd.

Also , f ( z )  g ( z )  0 on C, because otherwise g ( z )   f ( z ) so that g ( z)   f ( z)  f ( z)

on C, a contradiction to the hypothesis that g ( z)  f ( z) on C.

Since f(z) and f(z)+g(z) are analytic inside and on C, we have, therefore, by the argument principle,

1 f ( z )
2i C f ( z )
dz  number of zeros of f(z) inside C

1 f ( z )  g ( z )
2i C f ( z )  g ( z )
dz  number of zeros of f(z)+g(z) inside C.

Therefore, to prove that f (z ) and f ( z )  g ( z ) have the same number of zeros inside C, we must show
that
1 f ( z ) 1 f ( z )  g ( z )
2i C f ( z ) 2i  f ( z )  g ( z )
dz  dz (1)

Now, let  be any number such that 0    1 i.e.   [0,1] .

Then , f ( z )  g ( z ) is analytic inside and on C and

f ( z)  g ( z)  f ( z)  g ( z)  f ( z)   g ( z)  f ( z)  g ( z)  0 on C

i.e. f ( z )  g ( z )  0 on C.

Hence, by the argument principle,

1 f ( z )  g ( z )
2i C f ( z )  g ( z )
dz  number of zeros of f ( z )  g ( z ) inside C. (2)

1 f ( z )  g ( z )
2i C f ( z )  g ( z )
Let dz  I ( ) .

Then I ( ) is a continuous function of  in [0,1], which assumes only positive integral values by (2).

1 f ( z ) 1 f ( z )  g ( z )

2i C f ( z )
dz 
2i  f ( z )  g ( z )
dz .

That proves (1) .and thus f(z) and f(z)+g(z) have the same number of zeros inside C..
Replacing g(z) by –g(z) in the above development and noting that –g(z) is analytic and noting that
 g ( z)  g ( z) , it follows that f(z) and f(z)-g(z) also have the same number of zeros inside C.
Applications of Rouche’s theorem
1 (Fundamental Theorem of Algebra). Any complex polynomial of degree n has exactly n zeros in the
complex plane C.

Proof. Let P( z )  a n z n  a n 1 z n 1  ......  a1 z  a 0 be any complex polynomial of degree n so that

an  0 and hence an is affixed positive number.

Take f ( z )  a n z n and g ( z )  a n 1 z n 1  a n  2 z n  2  .......  a1 z  a 0 . Then f(z) and g(z) being entire

functions are analytic for z  R , for any R  0 howsoever large.

Now, for z  R , f ( z )  an z n  an z  an R n and

n

g ( z )  a n 1 z n 1  a n  2 z n  2  .....  a1 z  a 0
n 1 n2
 a n 1 z  an2 z  .......  a1 z  a0
 a n 1 R n 1  a n 2 R n 2  .......  a1 R  a0

Thus, for z  R ,

g ( z) an1 R n1  an2 R n2  .......  a1 R  a0


f ( z) an R n

an1 1 an2 1 a1 1 a0 1
 .  . 2  ......  . n1  . .
an R an R an R an R n

Since R is at our disposal, we let R   and noting that the RHS of of the above inequality tends to
zero as R   , we can chose R sufficiently large to make the RHS less than 1.

Thus , for that R, we have g ( z)  f ( z) for z  R . Hence, by Rouche’s theorem , it follows that f(z)
and f(z)+g(z) have the same number of zeros in z  R . But, f ( z )  a n z n has exactly n zeros in
z  R all at z=0 . Therefore f(z)+g(z)=P(z) also has exactly n zeros in z  R and hence in C.

2. If a>e, then the equation e z  az n has exactly n roots inside the circle z  1.

Proof. Let f ( z )  az n and g ( z )  e z . Then f(z0 and g(z) are entire functions and as such analytic inside
and on the circle z  1.

Further , for z  1, since a>e>0,

 z2 z
2
1 z   ...... 1 z   .....
g ( z) ez ez ez ez 2! 2!
z
e e
   n
     1
f ( z) az n az n az a a a a a

i.e. g ( z)  f ( z) for z  1.

Hence, by Rouche’s theorem, f(z) and f(z)-g(z) have the same number of zeros in z  1 . Since

f ( z )  az n has exactly n zeros in z  1(all at z=0), it follows that f(z)-g(z) i.e. az n  e z has exactly n
zeros in z  1 . In other words, the equation az n  e z  0 or az n  e z has exactly n zeros in z  1 .

Remark. Taking n=1 in the above result, it follows that the equation az  e z has exactly one zero
in z  1 for a>e.. For z=x real, if we consider the function f ( x)  ax  e x , then f(x) is continuous in
[0,1] with f (0)  1  0 and f (1)  a  e  0 . Hence, f(x) has a zero between 0 and 1.This shows that
the equation az  e z  0 i.e. az  e z has exactly one zero in z  1 which is real and lies between 0 and
1.

3. All the zeros of the polynomial z 7  5 z 3  12 lie between the circles z  1 and z  2 .

Proof. Let f(z)=12 and g(z)= z 7  5z 3 . Then f(z) and g(z) are analytic for z  1 .

Now, for z  1, g ( z )  z 7  5z 3  z 7   5z 3  z  5 z  1  5  6 and f ( z )  12  12 .

7 3

Since 6<12, we have g ( z)  f ( z) for z  1. Hence, by Rouche’s theorem, f(z) and f(z)+g(z) have the

same number of zeros in z  1 . Since f(z)=12 has no zero in z  1 , it follows that f(z) +g(z) i.e.
z 7  5 z 3  12 has no zero in z  1 . In other words, all the zeros of the given polynomial lie in z  1 .

Again, if we take f(z)= z 7 and g(z)=  5 z 3  12 .Then, f(z) and g(z) are analytic for z  2 and for

z  2 , g ( z)   5z 3  12   5z 3  12  5 z  12  40  12  52 and f ( z )  z 7  z
3 7

 2 7  128 . Since 52<128, we have g ( z)  f ( z) for z  2 . Therefore , by Rouche’s theorem, f(z)

and f(z)+g(z) have the same number of zeros in z  2 .But f(z)= z 7 has all its seven zeros in z  2 (all
at z=0). Hence, f(z)+g(z) i.e. the given polynomial has all its seven zeros in z  2 . Thus all the zeros of
the given polynomial lie in 1  z  2 i.e. between the circles z  1 and z  2 .

4. The equation e z  2 z  1 has one root within z  1.

Proof. Choose f (z)=2z and g ( z )  (e z  1) .Then f(z) and g(z) are both entire and hence analytic for

z  1.
Also , for z  1, f ( z )  2 z  2 z  2.1  2 and since 2<e<3,

2
z2 z2 z
g ( z )   (e  1)  (e  1)  1  z 
z z
 ...... _ 1  z   ......  z   ......
2! 2! 2!

 e  1  e  1  2  f ( z)
z

i.e, g ( z)  f ( z) for z  1.

Hence, by Rouche’s theorem, f(z) and f(z)+ g(z) have the same number of zeros in z  1 . Since f (z)=2z
has only one zero z=0 in z  1 , it follows that f(z)+ g(z) i.e 2z  (e z  1) also has only one zero in

z  1.This implies that the equation e z  2 z  1 has one root within z  1.

Remark. If instead of 2z , we have 2 z n , where n is a positive integer, in the above result, then the
equation e z  2 z n  1 has n roots in z  1 .

n
z  zk
5. let a  1 and z k  1 ,k=1,2,…..,n. Then the function  ( z )   (1  z
k 1 z
) assumes the value a n-
k

times within z  1.

n
z  zk 1
Proof. The possible singularities of the function  ( z )   (1  z
k 1 z
) are given by z 
zk
,
k

1 1
k  1,2,...., n so that z    1 . Thus the possible singularities of  (z ) lie outside z  1.
zk zk
Hence  (z ) is analytic for z  1 .

z  zk
Also, for z  1,  1, k  1,2,...., n.
1  zk z

n
z  zk
Therefore, for z  1,  ( z )   1 z
k 1 z
1 a .
k

Thus , if we choose g(z)= -a and f(z)=  (z ) , then f(z) and g(z) are analytic for z  1 and
g ( z)  f ( z) for z  1.

Hence , by Rouche’s theorem, f(z) and f(z)+ g(z) have the same number of zeros in z  1 . Since

 (z ) has n zeros in z  1namely z  z k , k  1,2,...., n , it follows that f(z)+ g(z) i.e.  (z ) -a also has n
zeros in z  1. In other words,  (z ) assumes the value a n-times within z  1.
 z2 zn
6. Consider the polynomial p n ( z )  1  z   .......  . Prove that howsoever large R may be,
2! n!
pn (z ) does not vanish within z  R if n is chosen sufficiently large. Does it contradict the fundamental
theorem of algebra.
Proof. We know that

z2 zn
ez  1 z   .......   ..... for z  R ,
2! n!
The series being uniformly convergent for these values of z because for z  R ,

n
zn z Rn zn
 n!

n!

n!
 e R and thus by Weierstrass’s M-test  n!
is uniformly convergent for

z R.

Now, min z R
e z  min 0  2 e R cos iR sin  min 0  2 e R cos  e  R .

(This incidently shows that e z is never equal to zero).

We choose   e  R . For this   0 , we can find a +ve integer N depending only on  i.e. depending
only on R (because of uniform convergence) such that

p n ( z )  e z  e  R for n  N ( ), z  R .

Therefore , pn ( z )  e z  e  R  e for z  R, n  N ( )
z

i.e. p n ( z )  e z  e z for z  R, n  N ( ) .

Since both e z and pn (z ) are entire functions, they are analytic for z  R and ,therefore, by Rouche’s
theorem, e z and e z  p n ( z )  e z have the same number of zeros within z  R for n  N ( ) i.e. e z
and pn (z ) have the same number of zeros within z  R for n  N ( ) . But e z is never equal to zero.
Therefore , pn (z ) does not vanish for z  R provided n  N ( ) , where N depends on R. It does not
contradict the fundamental theorem of algebra because in the present case the degree of the polynomial
varies with R.

7. Let f(z) be analytic for z  1 and f ( z )  1 for z  1. Then f(z) has exactly one fixed point within

z  1.

Exercises
1. Prove that all the zeros of the polynomial 8 z 11  2 z  5 lie in the interior of z  1, but no zero of the
1
polynomial lies in z  .
2

2. Prove that the polynomial 16 z 17  3z 11  4 z 7  2 z 3  z  3  5 has all its zeros within z  1and no
1
zeros within z  .
2
3. Find the number of zeros of the polynomial P(z) inside the circle z  1, where P(z)=

(i) z 10  6 z 7  3z 3  1

(ii) z 8  4 z 5  z 2  1
Maximum Modulus Theorem
Let f(z) be analytic inside and on a simple closed curve C, then f(z) is continuous there and hence so is
f (z ) because for any complex number h, f ( z  h)  f ( z )  f ( z  h)  f ( z )  0 as h  0 .

Since the interior and boundary of C is a closed and bounded set, f (z ) being continuous there must
attain its maximum somewhere inside or on C. The Maximum Modulus Theorem says that this maximum
will be attained on the boundary of C and never inside C. Formally, we state the theorem as follows:
Statement of Maximum Modulus Theorem (MM -Theorem)

Let f(z) be analytic inside and on a simple closed curve C of finite length l. If f ( z)  M on C, then

f ( z)  M inside C also and if f ( z)  M for some point z inside C, then f (z ) and hence f(z) is a
constant inside and on C of absolute value M.
In other words, a non-constant function analytic inside and on a simple closed curve of finite length
attains its maximum modulus on the boundary of the curve and never inside the curve.

Proof .Suppose that f ( z)  M on C. For any positive integer n, consider the function

F ( z )  { f ( z )}n .

Then , F ( z )  n{ f ( z )}n1 . f ( z ) .

Since f(z) is analytic inside and on C, f (z ) exists inside and on C. Therefore F (z ) also exists inside and
on C.Hence F(z) is analytic inside and on C.
Now, let a be any point inside C and let  be the distance of a from C i.e  is the minimum distance of a
from the points on C.

Then   0 and z  a   , z on C so that

 1 1
 , .  z on C (1)
za 

Since a is an interior point of C, we have by Cauchy’s integral formula

1 F ( z )dz
2i C z  a
F (a) 

1 { f ( z )}n dz
2i C z  a
i.e. { f (a)}n  . (2)

Taking modulus on both sides, we get

1 { f ( z )}n dz
f (a)   za
n

2 C

Using ML-Theorem and (1), this gives

1 M nl
f (a) 
n
.
2 
1
l
 f (a )  M .( )n . (3)
2
The L.H.S. of (3) is independent of n and also l and  are independent of n. Therefore letting n   and
1
l
noting that ( )  1 as n   , we get
n
2

 f ( a)  M

for an arbitrary point a inside C.

Thus, f ( z)  M inside C. This proves the first part of the theorem.

To prove the second part, suppose that for some point a inside C f (a)  M . We will show that f(z) is a
constant with f ( z)  M .

We have from (2) , on differentiating both sides w.r.to a,

1 { f ( z )}n dz
2i C ( z  a) 2
n{ f (a)} n 1
f (a)  . (4)

Taking modulus on both sides and using (1) and the ML-theorem, we get

1 M nl
 .
2  2
 1 M nl
Or nM n f (a)  .
2  2
l 1
 f (a )  . .
2 n 2

1
Letting n   and noting that the L.H.S .is independent of n and  0 as n   , we get
n

f (a)  0  f (a)  0.

Similarly differentiating (4) w.r.to a and using f ( a )  0 , we get f (a)  0 . In general by using
induction on n, we get f n (a)  0, n  1,2,3,........ .

Now, by Taylor’s theorem, we have

( z  a) 2 ( z  a) n n
f ( z )  f (a)  ( z  a) f (a)  f (a)  ......  f (a)  ........
2! n!
in a neighborhood of the point a.
This gives f ( z )  f (a ) i.e., f(z) is a constant in a neighborhood of a. By identity theorem it follows
that f(z) is a constant inside and on C with f ( z )  f (a)  M . That proves the second part of the
theorem and hence the theorem.
Remark 1. The first part of the theorem can be deduced from the Rouche’s theorem also. To see this let

f( z) be analytic inside and on a simple closed curve C and let f ( z)  M on C . We have to show that

f ( z)  M within C i.e. we have to prove that for any point z 0 inside C, f ( z 0 )  M .

If possible, suppose f ( z 0 )  M . Then on C, f ( z )  M  f ( z 0 ) . We choose g(z)= f ( z 0 ) ,

h( z )   f ( z ) , then g(z) and h(z) are analytic within and on C and

h( z)   f ( z)  f ( z)  f ( z0 )  g ( z) on C.

Therefore, by Rouche’s theorem, g(z) and g(z)=h(z) i.e. f ( z 0 ) and f ( z 0 )  f ( z ) have the same number
of zeros within C. Since f ( z 0 )  M  0 , f ( z 0 ) has no zero within C. Hence f ( z 0 )  f ( z ) has no
zero no within C, which is a contradiction since it has a zero at z  z 0 within C. This contradiction
proves the result.
Remark 2. The finiteness of the length l of the curve C plays an important role in the proof of the above
theorem. If l is not finite , the proof does not hold. In fact, MM-theorem does not hold for unbounded
regions. For example, consider the function f ( z )  e z in the right half plane Re( z )  x  0 . It is an
unbounded region having imaginary axix as its boundary which is of infinite length. On the boundary, we
have z=iy so that f ( z )  e z  e iy  1 , whereas at an interior point z=1, f ( z )  e1  e  1 .As

another example , consider the function f ( z)  e e in the region bounded by the straight lines
z



x i

z  xi . On the boundary of this region f ( z )  e ez

e  e ie  1 , whereas at the point z=0
2 x
e

2
inside the region f ( z )  e e  e e  4  1 . In fact, on the real axis which lies inside the region

f ( z )  e e  e e   as n   .
x x

Applications of MM-theorem

1. Let f(z) be analytic for z  r , f ( z)  m  0 for z  r and f (0)  m , then f(z) must vanish

Proof. If possible suppose f ( z )  0 for any z with z  r . Since by hypothesis f ( z)  m  0 for

z  r , f(z) does not vanish for z  r also. Therefore, it follows that f ( z )  0 for z  r . Hence

1 1 1
the function F ( z )  is analytic for z  r . Further, for z  r , F ( z )   by
f ( z) f ( z) m
1 1
hypothesis.Therefore , by MM-theorem, F ( z )  for z  r . In particular F (0)  i.e.
m m
1 1
 . This implies f (0)  m , which is a contradiction to the hypothesis that f (0)  m .
f (0) m

This contradiction proves the result..

2.( Fundamental Theorem of Algebra) Every complex polynomial has at least one zero in C.

Proof. Let P( z )  a n z n  a n 1 z n 1  ......  a1 z  a 0 , a n  0 be any complex polynomial of degree n.

Then P(z) is an entire function and hence analytic for z  R , howsoever large R may be.

Also for z  R ,

P( z )  a n z n  a n 1 z n 1  ......  a1 z  a 0

n 1
 a n z  ( a n 1 z  ......  a1 z  a 0 )
n

a n 1 1 a n  2 1 a1 1 a0 1
 a n R n {1  ( .  . 2  ......  . n 1  . )}
an R an R an R an R n
 a n R n (1  E )
 an1 1 an2 1 a1 1 a0 1
where E .  . 2  ......  . n1  . .
an R an R an R an R n

Since E  0 as R   , therefore by choosing R sufficiently large we can make E as small as we

1
please and hence less than .
2
Thus for z  R ,

an R n
P( z )   m  0.
2

an R n
Also P(0)  an is a fixed positive real number. Either it is already less than or we can choose
2
an R n
R still larger to make it less than  m and thus the conditions of the above result are satisfied.
2
Hence P(z)=0 for some z in z  R i.e. P(z) has a zero in the z-plane. That completes the proof.

3. Let f(z) be analytic inside and on a simple closed curve C and let f (z ) be constant on C. Then unless
f(z) vanishes inside C, f(z) must be constant throughout.

Proof. Suppose that f(z) is not constant. Then , since f (z ) is constant on C, say f ( z )  k , where k is
a positive constant, we have by MM-theorem, f ( z )  k inside C(because f(z) is not a constant) .

1
Now, suppose f ( z )  0 inside C. Then F ( z )  is analytic inside and on C and on C,
f ( z)

1 1 1
F ( z)   . Therefore F ( z )  inside C(strict inequality because f(z) and hence F(z) is not a
f ( z) k k
1 1
constant). This gives  or f ( z)  k inside C, which is a contradiction since f ( z)  k inside
f ( z) k
C . This contradiction shows that f(z) must be a constant under the given conditions.

Remark. If f(z) vanishes inside C, the above result is not true. To see this consider f ( z )  z 2 . Let C be

z  1. Then f(z) is analytic for z  1 and for z  1, f ( z )  z 2  z  1. Thus f (z ) is constant on

2

C , but f(z) is not constant throughout z  1 .

4. Let f(z) be analytic inside and on a simple closed curve C. Then each of Ref(z), - Ref(z), Imf(z),
- Imf(z) attains its maximum on C.
Proof . Let Ref(z)=u(x,y) and let g ( z )  e f ( z ) . Then g(z) is analytic inside and on C and by MM-
theorem, g ( z )  e f ( z )  e Re f ( z )  e u ( x , y ) attains its maximum on C. Since e u is maximum when u is
maximum, it follows that u attains its maximum on C. The other cases follow similarly.

5. Let f(z) be analytic for z  r , f ( z)  M for z  r and f (0)  0 . Then the number of zeros of f(z)
r 1 M
in z  does not exceed log .
3 log 2 f (0)

M M
Proof. We first note that f (0)  M by MM-theorem so that  1 and hence log  0.
f (0) f (0)

r
Now, let f(z) have n zeros z1 , z 2 ,....., z n in z  so that
3
r
z p  , p  1,2,......, n . (1)
3
f ( z)
Consider F ( z)  n
in z  r .
z

p 1
(1  )
zp

Apparently F(z) is not analytic at the points z  z p , p  1,2,....., n . But z p are the zeros of f(z)so that

z  z p are the factors of f(z) for p  1,2,....., n and hence F(z) is analytic in z  r .

Further, for z  r ,

f ( z)
F ( z) 
n
z
 1 z
p 1 p

f ( z)
 .
n
z

p 1 zp
1

z z r r n
z
Since
zp
1 
zp
1 
zp
 1   1  3  1  2 by using (1), it follows that
r z
p 1
 1  2n .
p
3
M
Hence, for z  r , F ( z)  . By MM-theorem the same inequality holds for z  r also i.e.
2n
 M M M
F ( z)  n
for z  r . In particular F (0)  n . But F(0)=f(0). Therefore f (0)  n giving
2 2 2
M M 1 M
2n   n log 2  log n log . That proves the result.
f (0) f (0) log 2 f (0)

6 (Minimum Modulus Theorem ). Let a non-constant function f(z) be analytic inside and on a simple
closed curve C such that f(z) does not vanish inside or on C. Then the minimum modulus of f(z) is
attained on C.
Proof. Suppose that the non-constant function f(z) analytic inside and on a simple closed curve C does
1
not vanish inside or on C. Then the function F ( z )  is analytic inside and on C and is not a
f ( z)
1
constant. Hence by MM-theorem F (z ) will attain its maximum on C. In other words becomes
f ( z)
maximum when z is on C which in turn implies that f (z ) is minimum when z is on C.

Remark. Consider f ( z )  z 2 which is analytic for z  1 . Then the minimum modulus of f(z) for z  1
is equal to zero attained at z-0 which is not a boundary point. The reason is that here f(z) vanishes at z=0
which is an interior point of the circle z  1.

Examples

1. Consider the function f ( z )  e z . Then f(z) is entire and hence analytic for z  r for any r>0
howsoever large. By MM-theorem f(z) will attain its maximum modulus on z  r . Also since f ( z )  0
for any z, the minimum modulus of f(z) will also be attained on z  r .

Now, for z  r i.e. z  re i , dz  rie i d ,  varying from 0 to 2 , we have

i
f ( z)  e z  e re  e r cos ir sin  e r cos .e ir sin  e r cos , e ir sin  e r cos .1  e r cos  p( ) (say)

Therefore, max z r
f ( z )  max 0 2 p( ) . For maximum or minimum of p ( ) , p ( )  0 i.e.

e r cos .  sin   0  sin   0    0,  ,2 .

Now p ( )  e r cos .(r sin  ) 2  re r cos cos  re r  0 for   0, ,2 and p ( )  re r  0

for    . Hence , p ( ) is maximum for   0, ,2 and minimum for    , the maximum value
being p(0)  p(2 )  e r and the minimum value p( )  e  r . Therefore

M (r , f )  max z r
f ( z )  e r and m(r , f )  min z r
f ( z )  e  r and both are attained at the
boundary points z=r and z= -r of the circle z  r .
2. Consider the function f(z)=cos z. It is an entire function and hence analytic for z  r for any r>0
howsoever large. By MM-theorem f(z) will attain its maximum modulus on z  r . For z  r , where
z=x+iy, we have

e iz  e iz 1 iz 1 1 1
f ( z )  cos z   e  e iz  ( e iz  e iz )  (e  y  e y )  (e r  e r )  cos(ir )
2 2 2 2 2

 cosh r .
Therefore , M (r , f )  max z r
f ( z )  cosh r attained at the boundary point z=ir of z  r .

3. Consider the function f ( z )  z 2 in z  1  i  1 . Then f(z)is analytic inside and on the the circle

z  1  i  1 . Therefore by MM-theorem the function will attain its maximum modulus on the boundary
of the circle.

Now , on the the circle z  1  i  1 , z 1  i  e i i.e. z  1  i  e i  (1  cos  )  i(1  sin  ) ,

  [0,2 ] .
Therefore ,
f ( z )  (1  cos  )  i (1  sin  )  (1  cos  ) 2  (1  sin  ) 2  3  2(cos   sin  )  p( ) .
2

For maximum of p ( ) , p ( )  0

 3
i.e. 2(cos   sin  )  0  sin   cos   tan   1    , .
4 4
1 1 
Further , p ( )  2(sin   cos  )  2(  )  2 2  0 at   .
2 2 4

 
Therefore , p ( ) is maximum for   and the maximum value of p ( ) is p ( )  3  2 2 .
4 4

i
Hence the maximum value of f (z ) is 3  2 2 attained at the boundary point z  1  i  e 4 .

4. Consider the function f ( z )  z 2  3z  1 in z  1 . The maximum of f (z ) will be attained on the

boundary z  1i.e. for z  e i . For such points , we have

f ( z )  e 2i  3e i  1  e i  3  e i  3  2i sin   9  4 sin 2 


which is maximum for    i.e. z  i , the points on the boundary of the unit circle. The maximum
2
value of f (z ) is 13 .
Exercises

1. Find M (r , f )  max
z

z r
f ( z ) , where f(z)= (i) sin z (ii) z (iii) z 2 (iv) e e .

2. Find max z 1
f ( z ) ,where f(z)= (i) z 2  4 z  1 (ii) z 2  z  1 (iii) z 2  z  1 .

Schwarz Lemma

Mz
Statement . Let f(z) be analytic for z  r , f ( z )  M for z  r and f(0)=0. Then f ( z)  for
r
z r.

f ( z)
Proof . Consider the function F ( z )  . Then a possible singularity of F(z) is the point z=0. But
z
since
f(0)=0, z=0 is a zero of f(z) and hence z-0=z is a factor of f(z) so that the term z in the denominator of
F(z) cancels with the same factor of f(z) in the numerator. Therefore F(z) is analytic wherever f(z) is
analytic. Since f(z) is analytic for z  r by hypothesis , it follows that F(z) is analytic for z  r .

Also for z  r , since f ( z)  M by hypothesis,

f ( z) f ( z) f ( z) M
F ( z)     .
z z r r

M f ( z) M
Therefore , by M M-theorem, F ( z )  for z  r also i.e.  for z  r or
r z r

Mz
f ( z)  for z  r .
r
Mz
Since for z  r ,  M and f ( z )  M for z  r by hypothesis, it follows that
r
Mz
f ( z)  for z  r .
r
Mz
Thus , f ( z )  for z  r and the result follows.
r
Remark 1. While applying MM-theorem to F(z), it follows that if sign of equality holds in
M M
F ( z)  at an interior point of z  r ,then F(z) is a constant of absolute value i.e.
r r
 M i M i
F ( z)  e ,  real or f ( z )  e z ,  real.
r r
Mz
Remark 2. Since  M for z  r, the bound M for f (z ) given by MM-theorem is reduced to a
r
smaller quantity inside the circle z  r . This means that Schwarz Lemma is a refinement of the MM-
theorem in case of a circle . But for this , we have to pay a price namely that f(0)=0.

Remark 3. The condition of the analyticity of f(z) on z  r is a superfluous condition . In fact, the
conclusion holds without this condition as is proved in the following .
Stronger form of Schwarz lemma

Mz
Statement . Let f(z) be analytic and f ( z)  M for z  r and f(0)=0. Then f ( z )  for
r
z r.

Proof . Let  be any positive number less than r. Then according to the hypothesis f(z) is analytic ,

Mz
f ( z)  M for z   and f(0)=0. Therefore , by Schwarz lemma, f ( z )  for z   . Letting


Mz
  r , it follows that f ( z )  for z  r .
r
M
Corollary . Under the conditions of the weaker or stronger form of Schwarz lemma, f (0)  with
r
M i
equality only for the function f ( z )  e z ,  being any real number.
r
Mz
Proof. Under the given conditions, we have f ( z )  for z  r (or z  r ) . This implies
r
f ( z) M f ( z )  f (0) M f ( z )  f ( 0) M
 or  or  . Letting z  0 and noting that
z r z 0 r z0 r

f ( z )  f (0) M
lim z 0  f (0) , it follows that f (0)  with equality only for the function
z0 r
M i
f ( z)  e z ,  being any real number, as before.
r
Generalizations of Schwarz lemma for a circle

za
1. Let f(z) be analytic for z  1, f ( z )  M for z  1and f(a)=0, a  1 . Then f ( z)  M for
1  az
 z  1.

(For a=0, it reduces to Schwarz lemma).

1  az
Proof. Let F ( z )  f ( z ).( ) . Then, since f(a)=0, z-a divides f(z) and hence F(z) is analytic
za
whenever f(z) is i.e. for z  1 .

We first prove that for a  1 ,

za
 1 for z  1
1  az

<1 for z  1 . (1)

Let z=x+iy and a    i so that z  x 2  y 2 , a   2   2 and a    i . Then

2 2

2
za ( x  iy )  (  i ) (x   )2  ( y   )2 x 2  y 2   2   2  2x  2y
2

  
1  az 1  \(  i )( x  iy ) (1  x  y ) 2  ( x  y ) 2 1  ( 2   2 )( x 2  y 2 )  2x  2y
z  a  2x  2y
2 2


1  a z  2x  2y
2 2

=1 for z  1.

z  a  2x  2y
2 2
za
2

Also ,  1 if 1
1  az 1  a z  2x  2y
2 2

i.e. if z  a  2x  2y  1  a z  2x  2y

2 2 2 2

or if (1  a ) z  1  a or z  1 i.e. if z  1 .
2 2 2 2

That proves (1).

1  az
Now, for z  1, F ( z )  f ( z ) .  f ( z ) .1  f ( z )  M by using (1) and the hypothesis that
za
f ( z)  M for z  1.Therefore , by MM-theorem, F ( z)  M for z  1 also i.e.

1  az za
f ( z) .  M for z  1 which implies f ( z )  M for z  1 . Since for z  1,
za 1  az

za za
 1 and f ( z )  M for z  1 by hypothesis, it follows that f ( z )  M for z  1.
1  az 1  az
 za
Thus , f ( z )  M for z  1 and the result follows.
1  az

Remark . Note that while applying MM-theorem to F(z), equality holds in F ( z)  M for z  1 only
when F(z) is a constant with modulus equal to M i.e. if F ( z )  Me i ,  real or

1  az za
f ( z ).( )  Me i i.e. f ( z )  Me i ( ),  real.
za 1  az
za
Thus, equality holds in the concluding inequality only for the function f ( z )  Me i ( ),  real.
1  az
M
Corollary. Under the conditions of the above result f (a )  with equality only for the function
1 a
2

za
f ( z )  Me i ( ),  real.
1  az
z a
2. Let f(z) be analytic and f ( z )  1 for z  1 with f(0)=a, 0  a  1 . Then f ( z )  for
1 a z
z  1.

(For a=0, the result reduces to Shwarz lemma)

f ( z)  a
Proof. Consider the function F ( z )  (1)
1  af ( z )
1 1 1
Since f ( z )  1 for z  1, f ( z )  for z  1 because   1 . In other words, 1-af(z)  0 for
a a a

z  1. Hence F(z) is analytic whenever f(z) is i.e. for z  1 .

f (0)  a a  a 0
Also, F (0)     0 since a<1.
1  af (0) 1  a 2
1 a2

Now, f ( z )  1 for z  1 and 0  a  1 so that a  1 .

f ( z)  a
Hence, F ( z )   1 for z  1 .
1  af ( z )
Therefore, by Schwarz lemma,

F ( z)  z for z  1(here M=1,r=1).

Let F(z)= e i so that   z for z  1 . (2)

Now , solving (1) for f(z) in terms of F(z), we get
F ( z)  a
f ( z) 
1  aF ( z )
giving

e i  a  cos   i sin   a (  cos   a)  i sin 

f ( z)    .
1  ae i
1  a (cos   i sin  ) (1  a cos  )  ia sin 
Therefore ,

(  cos   a) 2   2 sin 2   2  a 2  2a cos 

f ( z)  
2
.
(1  a cos  ) 2  a 2  2 sin 2  1  a 2  2  2a cos 

 2  a 2  2a cos 
Now , the maximum value of is attained for   0 and the maximum value is
1  a 2  2  2a cos 

 2  a 2  2a  a 2
( ) .
1  a   2a
2 2
1  a

a 2
f ( z)  (
2
Therefore, ) giving
1  a
a
f ( z)  ( ). (3)
1  a

a dp 1  a  a  a 2 1 a2
Let p  .Then    0 since a<1.
1  a d (1  a ) 2 (1  a ) 2

Therefore p is an increasing function of  . Hence for   z  1 ( by (2)), we have p(  )  p( z ) i.e.

 a z a z a
 for z  1 . Combining this with (3), we get f ( z )  for z  1 and the result
1  a 1  a z 1 a z
follows.

z a
Corollary. Under the conditions of the above result, f ( z )  for z  1 .
1 a z

 2  a 2  2a cos 
Proof. This follows from the fact that the minimum value of is attained for
1  a 2  2  2a cos 
 2  a 2  2a  a 2  a z a
   and the minimum value is ( ) and thus f ( z )  
1  a   2a
2 2
1  a 1  a 1  a z

for z  1 (since   z by (2)).

Applications of Schwarz lemma

1 (Bound for the derivative). Let f(z) be analytic and f ( z )  1 for z  1 . Then

1  f ( z)
2

f ( z )  for z  1 .
1 z
2

1  f (a)
2

Proof. We take an arbitrary point a with a  1 and prove that f (a)  . That will prove the
1 a
2

result.

za za
Since z  1 ,  1 also and hence according to the hypothesis f ( )  1.
1  az 1  az
Now, consider the function

za
f( )  f (a)
F ( z)  1  a z .
za
1  f (a) f ( )
1  az

1 1
We note that   1 . Therefore, the denominator of F(z) does not vanish for z  1 . In other
f (a) f (a)
words F(z) is analytic whenever f(z) is analytic i.e. for z  1 .

f (a)  f (a)
Also F (0)   0.
1  f (a)
2

za
Further, since f ( )  1 and f (a)  1 , it follows that F ( z )  1 for z  1 .
1  az

Hence , by the corollary of Schwarz lemma, F (0)  1 (1)

1 za z  a (1  az )  az  a a
But, F ( z )  [{1  f (a) f ( )}. f ( ).
za 2 1  a z 1  a z (1  az ) 2
{1  f (a) f ( )}
1  az
za d za
{ f ( )  f (a)}. {1  f (a) f ( }] .
1  az dz 1  az
Therefore ,
 (1  f (a) ). f (a).(1  a ) (1  a )
2 2 2

F (0)   f (a). .
(1  f (a) ) 2 (1  f (a) )
2 2

Using this in (1), we get

(1  a )
2

f (a) . 1
(1  f (a) )
2

1  f (a)
2

 f (a)  , thereby proving the result.

1 a
2

2. Let f(z) be analytic and f ( z )  1 for z  1 withf(a)=a, f(b)=b, a  b, a  1, b  1 . Then f(z0=z.

Proof. We note that a and b are interior points of the unit circle. Therefore, we may choose b=0 so that
a  0.
Since f(0)=0, the stronger form of Schwarz lemma gives

f ( z)  z for z  1 (1)

f ( z)
and if equality holds at an interior point in (1), then is a constant of absolute value 1 i.e.
z
f ( z)
 e i ,  real or f ( z )  e i z,  real.
z
Since f(a)=a, equality holds in (1) at an interior point. Hence we must have f ( z )  e i z,  real.

Putting z=a, we get a  e i a . Since a  0 , it follows that e i  1 and , therefore, f(z)=z as required.
Remark .This very important result shows that the two points have fixed the transformation.

1 1 1
3. If f(z) is analytic and f ( z )  1 for z  1 with f(0)=0, f ( )  , find f( ).
5 5 3
1 1
Sol. By the previous result f(z)=z. Therefore f( )= .
3 3
.