Lectures On The Riemann Zeta-Function: K. Chandrasekharan

Lectures on
The Riemann ZetaFunction

By
K. Chandrasekharan
Tata Institute of Fundamental Research, Bombay

1953
Lectures on the Riemann Zeta-Function

By
K. Chandrasekharan
Tata Institute of Fundamental Research

1953
The aim of these lectures is to provide an intorduction to the theory of the Riemann Zeta-function for students who might later want to do research on the subject.
The Prime Number Theorem, Hardys theorem on the
Zeros of (s), and Hamburgers theorem are the principal results proved here. The exposition is self-contained,
and required a preliminary knowledge of only the elements of function theory.
Contents
1 The Maximum Principle
2 The Phragmen-Lindelof principle
3 Schwarzs Lemma
17
4 Entire Functions
25
5 Entire Functions (Contd.)
35
6 The Gamma Function

1
Elementary Properties . . . . . . . . . . . . . . . . . . .
2
Analytic continuation of (z) . . . . . . . . . . . . . . .
3
The Product Formula . . . . . . . . . . . . . . . . . . .
45
45
49
51
7 The Gamma Function: Contd

4
The Bohr-Mollerup-Artin Theorem . . . . . . . . . . . .
5
Gausss Multiplication Formula . . . . . . . . . . . . .
6
Stirlings Formula . . . . . . . . . . . . . . . . . . . . .
55
55
58
59
8 The Zeta Function of Riemann

63
1
Elementary Properties of (s) . . . . . . . . . . . . . . . 63
9 The Zeta Function of Riemann (Contd.)
69
2
Elementary theory of Dirichlet Series . . . . . . . . . . 69
v
vi
Contents
10 The Zeta Function of Riemann (Contd)

75
2
(Contd). Elementary theory of Dirichlet series . . . . . . 75
87
3
Analytic continuation of (s). First method . . . . . . . 87
4
Functional Equation (First method) . . . . . . . . . . . . 90
5
Functional Equation (Second Method) . . . . . . . . . . 92
97
6
Some estimates for (s) . . . . . . . . . . . . . . . . . . 97
7
Functional Equation (Third Method) . . . . . . . . . . . 99
105
8
The zeros of (s) . . . . . . . . . . . . . . . . . . . . . 105
113
9
Riemann-Von Magoldt Formula . . . . . . . . . . . . . 113
121
10 Hardys Theorem . . . . . . . . . . . . . . . . . . . . . 121
127

12 The Prime Number Theorem . . . . . . . . . . . . . . .
13 Prime Number Theorem and the zeros of (s) . . . . . .
14 Prime Number Theorem and the magnitude of pn . . . .
135
135
145
146
Lecture 1
The Maximum Principle

Theorem 1. If D is a domain bounded by a contour C for which Cauchys 1
theorem is valid, and f is continuous on C regular in D, then | f | M
on C implies | f | M in D, and if | f | = M in D, then f is a constant.
Proof.
(a) Let zo D, n a positive integer. Then

1 Z { f (z)}n dz
n

| f (zo )| =
2i
z zo
C
lc M n
,
where lc = length of C, = distance of zo from C. As n

| f (z)| M.
(b) If | f (zo )| = M, then f is a constant. For, applying Cauchys inte
d
gral formula to
{ f (z)}n , we get
dz
1 Z
f n dz

|n{ f (zo )}n1 f (zo )| =
2i
(z zo )2
C
n
M
lC
22
1. The Maximum Principle
2
so that
| f (zo )|
lc M 1
0, as n
22 n
Hence
| f (zo )| = 0.
2
(c) If | f (zo )| = M, and | f (zo )| = 0, then f (zo ) = 0, for

d2
{ f (z)}n = n(n 1){ f (z)}n2 { f (z)}2 + n{ f (z)}n1 f (z).
2
dz
At zo we have
so that

d2
{ f (z)}n z=zo = n f n1 (Z0 ) f (z0 ),
2
dz
2! Z { f (z)}n dz

|nM n1 f (z0 )| =
2i
(z zo )3
C
2!lc n
M ,
23
and letting n , we see that f (zo ) = 0. By a similar reasoning

we prove that all derivatives of f vanish at z0 (an arbitrary point
of D). Thus f is a constant.

Remark. The above proof is due to Landau [12, p.105]. We shall now
show that the restrictions on the nature of the boundary C postulated in
the above theorem cab be dispensed with.
Theorem 2. If f is regular in a domain D and is not a constant, and
M = max | f |, then
zD
| f (zo )| < M,
zo D.
For the proof of this theorem we need a
Lemma. If f is regular in |z z0 | r, r > 0, then

| f (z0 )| Mr ,
3
where Mr = max | f (z)|, and | f (zo )| = Mr only if f (z) is constant for

|z zo | = r.
|zzo |=r
Proof. On using Cauchys integral formula, we get

Z
f (z)
1
dz
f (zo ) =
2i
z z0
|zzo |=n
1
=
2
Z2
f (zo + rei )d.
(1)
Hence
| f (zo )| Mr .
Further, if there is a point (such that) | zo | = r, and | f ()| <
Mr , then by continuity, there exists a neighbourhood of , on the circle
|z zo | = r, in which | f (z)| Mr , > 0 and we should have
| f (zo )| < Mr ; so that | f (zo )| = Mr implies | f (z)| = Mr everywhere on
|z zo | = r. That is | f (zo + rei )| = | f (zo )| for 0 2, or
f (zo + rei ) = f (zo )ei ,
0 2
On substituting this in (1), we get

1
1=
2
Z2
cos d
Since is a continuous function of and cos 1, we get cos 1 for

all , i.e. = 0, hence f (s) is a constant.

Remarks . The Lemma proves the maximum principle in the case of a 4
circular domain. An alternative proof of the lemma is given below [7,
Bd 1, p.117].
Aliter. If f (z0 + rei ) = () (complex valued) then

1
f (z0 ) =
2
Z2
()d
Now, if a and b are two complex numbers, |a| M, |b| M and

a , b, then |a + b| < 2M. Hence if () is not constant for 0 < 2,
then there exist two points 1 , 2 such that
|(1 ) + (2 )| = 2Mr 2, say, where > 0
On the other hand, by regularity,
|(1 + t) (1 )| < /2, for 0 < t <
|(2 + t) (2 )| < /2, for 0 < t <
Hence
|(1 + t) + (2 + t)| < 2Mr , for 0 < t < .

Therefore
Z2
()d =
Z1
0
Z2
1 +
2 +

Z
Z 1 Z2 Z2
+
= +
()d + [(1 + t) + (2 + t)] dt
Hence
1 + Z2
2 + Z2
Z
Z
+
+
+
+
1 +
2 +
() d| Mr (2 2) + (2Mr ) = 2Mr
1
|
2
Z2
0
()d| < Mr

5
Proof of Theorem 2. Let zo D. Consider the set G1 of points z such

that f (z) , f (zo ). This set is not empty, since f is non-constant. It
is a proper subset of D, since zo D, zo < G1 . It is open, because f is
continuous in D. Now D contains at least one boundary point of G1 . For
if it did not, G1 D would be open too, and D = (G1 G1 )D would be
disconnected. Let z1 be the boundary point of G1 such that z1 D. Then
z1 < G1 , since G1 is open. Therefore f (z1 ) = f (zo ). Since z1 BdG1
and z1 D, we can choose a point z2 G1 such that the neighbourhood
of z1 defined by
|z z1 | |z2 z1 |
lies entirely in D. However, f (z2 ) , f (z1 ), since f (z2 ) , f (zo ), and
f (zo ) = f (z1 ). Therefore f (z) is not constant on |z z1 | = r, (see
(1) on page 3) for, if it were, then f (z2 ) = f (z1 ). Hence if M =
max | f (z)|, then
|zz1 |=|z2 z1 |
| f (z1 )| < M M = max | f (z)|

zD
i.e.
| f (zo )| < M
Remarks. The above proof of Theorem 2 [7, Bd I, p.134] does not make
use of the principle of analytic continuation which will of course provide
an immediate alternative proof once the Lemma is established.
Theorem 3. If f is regular in a bounded domain D, and continuous
then f attains its maximum at some point in Bd D, unless f is a
in D,
constant.
Since D is bounded, D is also bounded. And a continuous function
on a compact set attains its maximum, and by Theorem 2 this maximum 6
cannot be attained at an interior point of D. Note that the continuity of
| f | is used.
Theorem 4. If f is regular in a bounded domain D, is non-constant, and
if for every sequence {zn }, zn D, which converges to a point Bd D,
we have
lim sup | f (zn )| M,
n
then | f (z)| < M for all z D. [17, p.111]
Proof. D is an F , for define sets Cn by the property: Cn consists of

all z such that |z| n and such that there exists an open circular neighbourhood of radius 1/n with centre z, which is properly contained in D.
Then Cn Cn+1 , n = 1, 2, . . .; and Cn is compact.

Define
max | f (z)| mn .
ZCn
By Theorem 2, there exists a zn Bd Cn such that | f (zn )| = mn . The

sequence {mn } is monotone increasing, by the previous results; and the
sequence {zn } is bounded, so that a subsequence {zn p } converges to a
limit Bd D. Hence
lim| f (zn p )| M
i.e.
or
i.e.
lim mn p M
mn < M for all n

| f (z)| < M. z D.
N.B. That Bd D can be seen as follows. If D, then there

exists an No such that C No D, and | f ()| < mNo lim mn , whereas
we have f (zn p ) f (), so that | f (zn p )| | f ()|, or lim mn = | f ()|.
Corollaries. (1) If f is regular in a bounded domain D and continuous in D, then by considering e f (z) and ei f (z) instead of f (z),
it can be seen that the real and imaginary parts of f attain their
maxima on Bd D.
(2) If f is an entire function different from a constant, then
M(r) l.u.b | f (z)|, and
|z|=r
A(r) l.u.b Re{ f (z)}

|z|=r
are strictly increasing functions of r. Since f (z) is bounded if M(r)

is, for a non-constant function f , M(r) with r.
(3) If f is regular in D, and f , 0 in D, then f cannot attain a

minimum in D.
and f , 0 in D,
Further if f is regular in D and continuous in D,

and m = min | f (z)|, then | f (z)| > m for all z D unless f is a
zBd D
constant. [We see that m > 0 since f , 0, and apply the maximum
principle to 1/ f ].
and | f | is a
(4) If f is regular and , 0 in D and continuous in D,
constant on Bd D, then f = in D. For, obviously , 0, and,
by Corollary 3, | f (z)| for z D, se that f attains its maximum
in D, and hence f = .
and | f | is a constant on 8
(5) If f is regular in D and continuous in D,
Bd D, then f is either a constant or has a zero in D.
(6) Definition: If f is regular in D, then is said to be a boundary
value of f at Bd D, if for every sequence zn Bd D,
zn D, we have
lim f (zn ) =
n
(a) It follows from Theorem 4 that if f is regular and non - constant in D, and has boundary-values {}, and || H for
each , then | f (z)| < H, z D. Further, if M = max || then
b.v
M = M max | f (z)|
zD
For, || is a boundary value of | f |, so that || M, hence

M M. On the other hand, there exists a sequence {zn },
zn D, such that | f (zn )| M. We may suppose that
zn zo (for, in any case, there exists a subsequence {zn p }
with the limit zo , say, and one can then operate with the
subsequence). Now zo < D, for otherwise by continuity
f (zo ) = lim f (zn ) = M, which contradicts the maximum
n
principle. Hence zo Bd D, and M is the boundary-value
of | f | at zo . Therefore M M , hence M = M .
(b) Let f be regular and non-constant in D, and have boundaryvalues {} on Bd D. Let m = min | f |; m = min || > 0.
zD
b.v.
Then, if f , 0 in D, by Corollary (6)a, we get

max
zD
1
1
1
= = ;
| f (z)| m
m
so that | f | > m = m in D. Thus if there exists a point zo D

such that | f (zo )| m , then f must have a zero in D, so that
m > 0 = m. We therefore conclude that, if m > 0, then
m = m if and only if f , 0 in D [6, Bd. I, p.135].
Lecture 2
The Phragmen-Lindelof
principle
We shall first prove a crude version of the Phragmen-Lindelof theorem 10
and then obtain a refined variant of it. The results may be viewed as
extensions of the maximum-modulus principle to infinite strips.
Theorem 1. [6, p.168] We suppose that
(i) f is regular in the strip < x < ; f is continuous in x
(ii) | f | M on x = and x =
(iii) f is bounded in x
Then, | f | M in < x < ; and | f | = M in < x < only if f is
a constant.
Proof.
(i) If f (x + iy) O as y uniformly in x, x ,

then the proof is simple. Choose a rectangle x , |y|
with sufficiently large to imply | f |(x i) M for x
. Then, by the maximum-modulus principle, for any zo in the
interior of the rectangle,
| f (zo )| M.
9
2. The Phragmen-Lindelof principle
10
(ii) If | f (x + iy)| 6 0, as y , consider the modified function

2
fn (z) = f (z)ez /n . Now
| fn (z)| 0 as y uniformly in x, and
2 /n
| fn (z)| Mec
11
on the boundary of the strip, for a suitable constant c.

Hence
2
| fn (zo )| Mec /n
for an interior point zo of the rectangle, and letting n , we get
| f (zo )| M.
If | f (zo )| = M, then f is a constant. For, if f were not a constant, in the
neighbourhood of zo we would have points z, by the maximum-modulus
principle, such that | f (z)| > M, which is impossible.

Theorem 1 can be restated as:
Theorem 1 : Suppose that
(i) f is continuous and bounded in x
(ii) | f ( + iy)| M1 , | f ( + iy)| M2 for all y
Then
| f (xo + iyo )| M1L(xo ) M21L(xo )
for < xo < , |yo | < , where L(t) is a linear function of t which takes
the value 1 at and 0 at . If equality occurs, then
f (z) = cM1L(z) M21L(z) ;
|c| = 1.
Proof. Consider
f1 (z) =
12
f (z)
M1L(z)
M21L(z)
and apply Theorem 1 to f1 (z).

More generally we have [12, p.107]
11
Theorem 2. Suppose that

(i) f is regular in an open half-strip D defined by:
z = x + iy,
< x < ,
y>
(ii) lim| f | M, for Bd D

z
in D

|z|
(iii) f = O exp e , < D
uniformly in D.
Then, | f | M in D; and | f | < M in D unless f is a constant.
Proof. Without loss of generality we can choose
= ,
2
= +/2,
=0
Set
g(z) = f (z). exp(eikz )
f (z).{(z)} , say,
where > 0, < k < 1. Then, as y +,
"
(
#
k
|z|
ky
g = O exp e e cos
2
uniformly in x, and
13
e|z| eky cos
k
k
e(y+/2) eky cos
2
2
, as y ,
uniformly in x, since < k. Hence

|g| 0 uniformly as y ,
so that
|g| M, for y > y , < x < .
12
Let zo D. We can then choose H so that

|g(z)| M
for y = H, < x < , and we may also suppose that H > y0 = im(zo ).
Consider now the rectangle defined by < x < , y = 0, y = H.
Since | | 1 we have
lim |g| M
z
in D
for every point on the boundary of this rectangle. Hence, by the

maximum-modulus principle,
|g(zo )| < M, unless g is a constant;
ikzo
| f (zo )| < M|ee
or
Letting 0, we get
14
| f (zo )| < M.

Remarks. The choice of in the above proof is suggested by the critical
case = 1 of the theorem when the result is no longer true. For take
= /2,
= 1,
Then
iz
= /2,
iz )
|ee | = eRe(e
yix }
= eRe{e
= 0,
y
= ee
iz
f = ee
cos x
So
| f | = 1 on x = /2,
and f = ee on x = 0.
Corollary 1. If f is regular in D, continuous on Bd D, | f | M on Bd

D, and
|z|
f = O ee , < 1,
then | f | M in D.
13
Corollary 2. Suppose that hypotheses (i) and (iii) of Theorem 2 hold;

suppose further that f is continuous on Bd D,
f = O(ya ) on x = ,
f = O(yb ) on x = . Then
f = O(yc ) on x = ,
uniformly in , where c = p + q, and px + q is the linear function which
equals a at x = and b at x = .
Proof. Let > 0. Define (z) = f (z)(z), where is the single-valued 15
branch of (zi)(pz+q) defined in D, and apply Theorem 2 to . We first
Next
observe that is regular in D and continuous in D.
|(z)| = |(y ix)(px+q)ipy |
log y ix
)
= |y ix|(px+q) exp(py im
y
!
" (
!)#
x
1 O(1)
1
(px+q)
=y
|1 + O
|
exp py + O 2
y
y
y
!
1
},
= y(px+q) epx {1 + O
y
the Os being uniform in the x s.

Thus = O(1) on x = and x = . Since = O(yk ) = O(|z|k ), we
see that condition (iii) of Theorem 2 holds for if is suitably chosen.
Hence = O(1) uniformly in the strip, which proves the corollary.
Theorem 3. [12, p.108] Suppose that conditions (i) and (iii) of Theorem
2 hold. Suppose further that f is continuous in Bd D, and
lim | f | M on x = ,
x=
Then
lim | f | M uniformly in x .
by the previous theorem. Let 0, P > 0. 16

Proof. f is bounded in D,
Let H = H() be the ordinate beyond which | f | < M + on x = , x = .
14
Let h > 0 be a constant. Then

z
<1

z + hi
in the strip. Choose h = h(H) so large that

z
f (z)
< M + on y = H
z + hi

z
| f | < M). Then the function
(possible because f
z + hi
z
g(z) = f
z + hi
satisfies the conditions of Theorem 2 (with M + in place of M) in the
strip above y = H. Thus
|g| M + in this strip
and so lim |g| M + uniformly.
y
That is,
lim| f | M + ,
since
z
1 uniformly in x. Hence
z + hi
lim| f | M.

17
Corollary 1. If conditions (i) and (iii) of Theorem 2 hold, if f is continuous on Bd D, and if
a on x = ,
lim| f |
b on x = ,
where a , 0, b , 0, then
lim | f | e px+q ,
uniformly, p and q being so chosen that e px+q = a for x = and = b for

x = .

Proof. Apply Theorem 3 to g = f e(pz+q)
15
Corollary 2. If f = O(1) on x = , f = o(1) on x = , then f = o(1) on

x = , < .
[if conditions (i) and (iii) of Theorem 2 are satisfied].
Proof. Take b = in Corollary 1, and note that e p+q 0 as 0 for
fixed , provided p and q are chosen as specified in Corollary 1.
Lecture 3
Schwarzs Lemma
Theorem 1. Let f be regular in |z| < 1, f (o) = 0.
If, for |z| < 1, we have | f (z)| 1, then
| f (z)| |z|,
for |z| < 1.
Here, equality holds only if f (z) c z and |c| = 1. We further have

| f (o)| 1.
f (z)
is regular for |z| < 1. Given o < r < 1 choose such that
z
r < < 1; then since | f (z)| 1 for |z| = , it follows by the maximum
modulus principle that
f (z) 1
,

z
Proof.
also for |z| = r. Since the L.H.S is independent of 1, we let and

obtain | f (z)| |z|, for |z| < 1.
f (z0 )
= 1, (by the
If for z0 , (|z0 | < 1), we have | f (z0 )| = |z0 |, then
z0
f (z)
) hence f = cz, |c| = 1.
maximum principle applied to
z
Since f (z) = f (o)z + f (o)z2 + in a neighbourhood of the origin,
f (z)
1 in z 1, we get | f (o)| 1. More generally, we
and since
z
have

17
18
3. Schwarzs Lemma
18
19
Theorem 1 : Let f be regular and | f | M in |z| < R, and f (o) = 0.

Then
M|z|
, |z| < R.
| f (z)|
R
In particular, this holds if f is regular in |z| < R, and continuous in
|z| R, and | f | M on |z| = R.
Theorem 2 (Caratheodorys Inequality). [12, p.112] Suppose that
(i) f is regular in |z| < R f is non-constant
(ii) f (o) = 0
(iii) Re f in |z| < R. (Thus > 0, since f is not a constant).
Then,
|f|
2
, for |z| = < R.
R
Proof. Consider the function

(z) =
f (z)
.
f (z) 2
We have Re{ f 2} < 0, so that is regular in |z| < R.

If f = u + iv, we get
s
u2 + v2
|| =
(2 u)2 + v2
so that
|| 1, since 2 u |u|.
But (o) = o, since f (o) = o. Hence by Theorem 1,
|(z)|
20
|z|
,
R
|z| < R.
2
But f =
. Now take |z| = < R.
1
Then, we have,
3. Schwarzs Lemma
19
|f|
N.B.
If
2 || 2
1 || R
2
, for |zo | < R, then
R
2U cz
f (z) =
, |c| = 1.
1 cz
| f (zo )| =
More generally, we have
Theorem 2 : Let f be regular in |z| < R; f non-constant

f (o) = ao = + i
Re f (|z| < R), so that
Then
2( )
| f (z) ao |
R
for |z| = < R
Remark . If f is a constant, then Theorem 2 is trivial. We shall now

prove Borels inequality which is sharper than Theorem 2.
Theorem 3 (Borels Inequality). [12, p.114] Let f (z) =
regular in |z| < 1. Let Re f . f (o) = ao = + i.
Then
|an | 2( ) 2 1 , say, n > 0.
an zn , be
n=0
Proof. We shall first prove that |a1 | < 1 , and then for general an .
(i) Let f1
n=1
an zn . Then Re f1 1 .
For |z| = < 1, we have, by Theorem 2,

2 1
| f1 |
1

21
f1 (z)

i.e.
z 1
21
3. Schwarzs Lemma
20
Now letting z 0, we get
|a1 | 21
(ii) Define = e2i/k , k a +ve integer.
Then
k
X
v=1
We have
if m , 0
=o
& if m , a multiple of k
if m = 0
=k
or if m = a multiple of k.
k1
X
1X
f1 (r z) =
ank znk = g1 (zk ) g1 (), say.
k r=0
n=1
22
The series for g1 is convergent for |z| < 1, and so for || < 1. Hence
g1 is regular for || < 1. Since Re g1 1 , we have
| coefficient of | 21 by the first part of the proof i.e. |ak |
21 .

Corollary 1. Let f be regular in |z| < R, and Re{ f (z) f (o)} 1 . Then
| f (z)|
2 1 R
,
(R )2
|z| = < R
Proof. Suppose R = 1. Then

X
X
| f (z)|
n|an |n1 2 1
nn1
This argument can be extended to the nth derivative.
Corollary 2. If f is regular for every finite z, and

k
e f (z) = O(e|z| ), as |z| ,

then f (z) is a polynomial of degree k.
3. Schwarzs Lemma
Proof. Let f (z) =
21
an zn . For large R, and a fixed , || < 1, we have

Re{ f (R)} < cRk .
By Theorem 3, we have
|an Rn | 2(2Rk Re ao ),
n > 0.
Letting R , we get
an = 0 if n > k.
Apropos Schwarzs Lemma we give here a formula and an inequality which are useful.

Theorem 4. Let f be regular in |z zo | r
23
f = P(r, ) + Q(r, ).
Then
1
f (zo ) =
r
Z2
P(r, )ei d.
Proof. By Cauchys formula,

(i)
1
f (zo ) =
2i
|zzo |=r
f (z)dz
1
=
2
2r
(z zo )
Z2
(P + iQ)ei d
0
By Cauchys theorem,
(ii)
1 1
0= 2
r 2i
1
f (z)dz =
2r
|zzo |=r
Z2
(P + iQ)ei d
0
We may change i to i in this relation, and add (i) and (ii).

Then
Z2
1
f (zo ) =
P(r, )ei d
r
0
3. Schwarzs Lemma
22
Corollary. If f is regular, and | Re f M in |z z0 | r, then

| f (zo )|
24
2M
.
r
Aliter: We have obtained a series of results each of which depended

on the preceding. We can reverse this procedure, and state one general
result from which the rest follow as consequences.
P
Theorem 5. [11, p.50] Let f (z)
cn (zz0 )n be regular in |zz0 | < R,
n=0
and Re f < . Then
2( Re co )
,
Rn
and in |z zo | < R, we have
|en |
n = 1, 2, 3, . . .
2
| f (z) f (z0 )|
{ Re f (zo )}
R
(n)
f (z)
2R
{ Re f (zo )} n = 1, 2, 3, . . .

n!
(R )n+1
(5.1)
(5.2)
(5.3)
Proof. We may suppose zo = 0.

Set
(z) = f (z) = c0
X
0
bn zn ,
cn zn
|z| < R.
Let denote the circle |z| = < R. Then

1
bn =
2i
1
(z)dz
=
n+1
2n
z
(P + iQ)ein d, n 0,
(5.4)
where (r, ) = P(r, ) + iQ(r, ). Now, if n 1, then (z)zn1 is regular

is , so that
n
0=
2r
(P + iQ)ein d, n 1
(5.5)
3. Schwarzs Lemma
23
Changing i to i in (5.5) and adding this to (5.4) we get

1
bn n =
Peni d,
n 1.
But P = Re f 0 in |z| < R and so in . Hence, if n 1,

1
|bn |n
1
|Peni |d =
Pd = 2 Re bo ,
25
(5.6)
on using (5.4) with n = 0. Now letting R, we get

|bn |Rn 2o 2 Re bo .
Since bo = co , and bn = cn , n 1, we at once obtain (5.1). We
then deduce, for |z| < R
|(z) (o)| = |
X
1
bn z |
20
n
R
2o
R
(5.7)
and if n 1,
(n)
| (z)|
X
r=n
r(r 1) . . . (r n + 1)
d
=
d
!n X
20
n=0
!n
2e R
d
d R
o R n!
=2
(R )n+1
20 rn
Rr
r
R
(5.8)
(5.7) and (5.8) yield the required results on substituting = f and

0 = Re f (0).
Lecture 4
Entire Functions
An entire function is an analytic function (in the complex plane) which 26
has no singularities except at . A polynomial is a simple example. A
polynomial f (z) which has zeros at z1 , . . . , zn can be factorized as
!
!
z
z
f (z) = f (0) 1
... 1
z1
zn
The analogy holds for entire functions in general. Before we prove this,
we wish to observe that if G(z) is an entire function with no zeros it can
G (z)
; every
be written in the form eg(z) where g is entire. For consider
G(z)
(finite) point is an ordinary point for this function, and so it is entire
and equals g1 (z), say. Then we get
) Zz1
G(z)
log
=
g1 ()d = g(z) g(z0 ), say,
G(z0 )
(
z0
so that
G(z) = G(z0 )eg(z)g(z0 ) = eg(z)g(z0 )+log G(zo )
As a corollary we see that if G(z) is an entire function with n zeros,
distinct or not, then
G(z) = (z z1 ) . . . (z zn )eg(z)
25
4. Entire Functions
26
where g is entire. We wish to uphold this in the case when G has an

infinity of zeros.
27
Theorem 1 (Weierstrass). [15, p.246] Given a sequence of complex

numbers an such that
0 < |a1 | |a2 | . . . |an | . . .
whose sole limit point is , there exists an entire function with zeros at
these points and these points only.
Proof. Consider the function
1
!
z Q (z)
e
,
an
where Q (z) is a polynomial of degree q. This is an entire function which

does not vanish except for z = an . Rewrite it as
!

z
z Q (z)
1
e
= eQ (z)+log 1 an
an
azn
=e
and choose
Q (z) =
z2
...+Q (z)
2a2
z
z
+ ... +
an
a n
so that
+1
!
azn 1 ...
z Q (z)
+1
1
e
=e
an
1 + n (z), say.
We wish to determine in such a way that
Y
1
28
!
z Q (z)
e
1
an
is absolutely and uniformly convergent for |z| < R, however large R may
4. Entire Functions
27
be.
Choose an R > 1, and an such that 0 < < 1; then there exists a
q (+ ve integer) such that
|aq |
R
,
R
.
|aq+1 | >
Then the partial product

q
Y
1
!
z Q (Z)
e
1
an
is trivially an entire function of z. Consider the remainder
Y
q+1
!
z Q (z)
e
1
an
for |z| R.

R
z
We have, for n > q, |an | >
or < < 1. Using this fact we
an
shall estimate each of the factors {1 + n (z)} in the product, for n > q.

+1
z
1
| n (z)| = e +1 an ... 1

1 z +1 ...
+1 a

Since |em 1| e|m| 1, for em 1 = m +

|m| +
|m2 |
. . . = e|m| 1.
2
+1
z
1+ azn +...
an
|Un (z)| e
1,
+1
z (1++2 + )
an
=e
+1
1 z
1 an

m2
+ . . . or |em 1| 29
2!
4. Entire Functions
28
real,
ex
) = xe x
Hence
xe x ,

1
1 z +1 1
e
1 an
for
ex
, since if x is
!
x2
x2
x
+ x(1 + x +
+
1 = x 1+ +
2! 3!
2!
|Un (z)|
30
+1
z
an
1

e 1 z +1

1 an
Now arise two cases: (i) either there exists an integer p > 0 such
P
1
< or (ii) there does not. In case (i) take = p 1, so
that
p
n=1 |an |
that
1
e 1a
Rp
, since |z| R;
|Un (z)|
|an | p 1
and hence
|Un (z)| < for |z| R, i.e.
uniformly convergent for |z| R.

In case (ii) take = n 1, so that
|Un (z)| <
(1 + Un (z)) is absolutely and
q+1
1

e 1 z n
,n>q
1 an

z < 1
an
|z| R
|an |
P
Then by the root-test |Un (z)| < , and the same result follows
as before. Hence the product
Y
1
31
!
z Q (z)
e
an
is analytic in |z| R; since R is arbitrary and does not depend on R we

see that the above product represents an entire function.
4. Entire Functions
29
Remark. If in addition z = 0 is also a zero of G(z) then
G(z)
has no
zmG(z)
zeros and equals eg(z) , say, so that

g(z)
G(z) = e
Y
n=1
!
z Q (z)
1
e
an
In the above expression for G(z), the function g(z) is an arbitrary

entire function. If G is subject to further restrictions it should be possible
to say more about g(z); this we shall now proceed to do. The class of
entire functions which we shall consider will be called entire functions
of finite order.
Definition. An entire function f (z) is of finite order if there exists a con
stant such that | f (z)| < er for |z| = r > r0 . For a non-constant f , of
finite order, we have > 0. If the above inequality is true for a certain
, it is also true for > . Thus there are an infinity of s o satisfying
this. The lower bound of such s is called the order of f . Let us
denote it by . Then, given > 0, there exists an ro such that
| f (z)| < er
for |z| = r > r0 .

This would imply that
+
M(r) = max | f (z)| < er , r > r0 ,

|z|=r
while
32
r
M(r) > e
for an infinity of values of r tending
to +
Taking logs. twice, we get
log log M(r)
<+
log r
and
log log M(r)
> for an infinity of values of r
log r
4. Entire Functions
30
Hence
log log M(r)
r
log R
= lim
Theorem 2. If f (z) is an entire function of order < , and has an

infinity of zeros, f (0) , 0, then given > 0, there exists an R such that
for R Ro , we have
+
eR
1
n
log
3
log 2
| f (o)|
R
Here n(R) denotes the number of zeros of f (z) in |z| R.

Proof. We first observe that if f (z) is analytic in |z| R, and a1 . . . an
are the zeros of f inside |z| < R/3, then for
33
we get the inequality
g(z) =
1
z
a1
f (z)

... 1
|g(z)|
z
an
M
2n
where M is defined by: | f (z)| M for |z| = R. For,

if |z| = R, then since
z

R
z
|a p | , p = 1 . . . n, we get 3 or 1 2
ap

3
ap
By the maximum modules theorem,
|g(o)|
M
M
, i.e. | f (o)|
2n
2n
or
1
M
log
nn
3
log 2
f (o)
R
If, further, f is of order , then for r > r , we have M < er

gives the required results.
N.B. The result is trivial if the number of zeros is finite.
which
4. Entire Functions
31
Corollary.
n(R) = O(R+ ).
For

1
R+ log | f (o)| ,
3
log 2
log | f (0)| < R+ , say.
n
and
R
Then for R R
n
and hence the result.
R
3
2 +
R
log 2
<
Theorem 3. If f (z) is of order < , has an infinity of zeros

X 1
<
a1 , a2 , . . . , f (o) , 0, > , then
|an |
34
Proof. Arrange the zeros in a sequence:
|a1 | |a2 | . . .
Let
p = |a p |
Then in the circle |z| r = n , there are exactly n zeros. Hence
+
n < c n , for n N
or
1
1 1
>
n c +
n
Let > + (i.e. choose 0 < < ). Then
1
n/+
or
Hence
>
1
c/+
1
,
n
1
1
< c/+ /+ for n N
n
n
P 1
< .
n
32
35
4. Entire Functions
Remark. (i) There cannot be too dense a distribution of zeros, since

+
n < c < n . Nor can their moduli increase too slowly, since, for
P
1
does not converge.
instance,
(log n) p
(ii) The result is of course trivial if there are only a finite number of
zeros.
1
< ,
|an |
is called the exponent of convergence of {an }. We shall denote it by 1 .
Then
X 1
X 1
<
,
= , > 0
|an |1 +
|an |1
Definition. The lower bound of the numbers for which
By Theorem 1, we have 1
N.B. If the an s are finite in number or nil, then 1 = 0. Thus 1 > 0

implies that f has an infinity of zeros. Let f (z) be entire, of order < ;
f (o) , 0, and f (zn ) = 0, n = 1, 2, 3, . . .. Then there exists an integer
P 1
(p + 1) such that
<
|zn |+1
(By Theorem 1, any integer > will serve for + 1). Thus, by
Weierstrasss theorem, we have
!
Y
z zzn + 2zz2n 2 ++ zz
Q(z)
n,
e
f (z) = e
1
z
n
1
36
where = p (cf. proof of Weierstrasss theorem).

P
Definition. The smallest integer p for which |z |1p+1 < is the genus
n
!
p
z zzn ++ pzz p
n , with zeros zn . If z s
e
of the canonical product 1
n
zn
are finite in number, !we (including nil) define p = 0, and we define the
z
product as 1
zn
Examples. (i) zn = n, p = 1 (ii) zn = en , p = 0
(iii) z1 = 12 log 2, zn = log n, n 2, no finite p.
4. Entire Functions
Remark. If > 1 , then
Also,
But
33
P 1
<
|zn |
X 1
<
|zn | p+1
X 1
=
|zn | p
Thus, if 1 is not an integer, p = [1 ], and if 1 is an integer, then either

P 1
P 1
< or
= ; in the first case, p + 1 = 1 , which in the
1
|zn |
|zn |1
second case p = 1 . Hence,
1
But 1 . Thus
p 1
Also
p 1 p + 1, since
1
< .
|zn | p+1
Lecture 5
Entire Functions (Contd.)

Theorem 1 (Hadamard). [8] Let f be an entire function with zeros
|z1 | |z2 | . . . {zn } such that
Let f (o) , 0. Let f be of order < . Let G(z) be the canonical product
with the given zeros.
Then
f (z) = eQ(z)G(z),
where Q is a polynomial of degree .
We shall give a proof without using Weierstrasss theorem. The proof
f (z)
1
will depend on integrating the function
.
over a sequence
f (z) z (z x)
of expending circles.
Given R > 0, let N be defined by the property:
|zN | R,
|zN+1 | > R.
We need the following

Lemma. Let 0 < < 1. Then

N
f (z) X
1

< cR1+
f (z) n+1 z zn
for |z| =
R
R
> R0 , and |z| = is free from any zn .
2
2
35
37
5. Entire Functions (Contd.)
36
For we can choose an R0 such that for |z| = R > R0 ,

38
(i) | f (z)| < eR ,

(ii) no zn lies on |z| =
(iii) log
R
, and
2
1
< (2R)+
| f (0)|
For such an R > R0 , define

N
f (z) Y
z
gR (z) =
1
f (0) n=1
zn
!1
Then for |z| = 2R, we have

|gR (z)| <
1 (2R)+
e
| f (o)|
z
| 1 for 1 n N.
zn
Hence, by the maximum modulus principle,
+
since | f | < e(2R) , and |1
|gR (z)| <
1 (2R)+
e
for |z| = R.
f (0)
That is
log |gR (z)| < 2+2 R+ . (0 < < 1)
If hR (z) log gR (z), and the log vanishes for z = 0, then, since gR (z)
is analytic and gR (z) , 0 for |z| R, we see that hR (z) is analytic for
|z| R, and hR (0) = 0. Further, for |z| = R,
Re hR (z) = log |gR (z)| < 2+2 R+
39
Hence, by the Borel-Caratheodory inequality, we have for |z| = R/2,

|hR (z)| < cR1+ ,
which gives the required lemma.
37
Proof of theorem. Let = [] Then consider

I
1
f (z)
dz
f (z) z (z x)
R
taken along |z| = R/2, where |x| < . The only poles of the inte4
grand are: 0, x and those zn s which lie inside |z| = R/2. Calculating the
residues, we get
1
2i
|z|=R/2
X
1
1
f (z)
dz
=
f (z) z (z x)
z (zn x)
|Z |<R/2 n
n
f (x) 1
f (x) x
!

1
1
f (z)
1
D
+
z=0 f (z)
( 1)!
zx
+
We shall prove that the l.h.s tends to o as R in such a way that

|z| = R/2 is free from any zn . Rewriting I as
Z X
N
N Z
X
f
1
dz
dz
1
+

I=

f
z zn z (z x)
z zn z (z x)
1
1
I1 + I2 ,
we get for |z| = R/2 (by the Lemma),
40
|I1 | = O(R1+ ) = O(1).

Let k denote the number of zn s lying inside |z| = R/2
Then
I2 =
Z
k
X
... +
n=1|z|=R/2
I2,1 + I2,2 , say.
N Z
X
n=k+1
...
38
The value of I2,1 remains unaltered when we integrate along |z| =

3R/4, and along the new path, |z zn | R/4 since n k. Since N =
O(R+ ) we get
I2,1 = O(R1+ ) = O(1)
Similarly integrating I2,2 over |z| = R/4, we get
I2,2 = O(1)
Thus I = o(1). Hence
!
f (x) 1 X
1 h i1 f 1
1
D
=0
+
z=0
f (x) x n=1 zn (z x) ( 1)!
f zx
i.e.
i.e.
41
f (x)
f (x)
f (x)
=
f (x)
1
X
0
1
X
0
bn xn
x1
zn (z
n=1
= 0,
bn = an , an =
x)
X
x1 x2
1
n
+
+ +
cn x +
1
x zn
zn
zn
n=1
h f i(n)
f z=0
( n)!
Integrating and raising to the power of e we get the result:
f (x) = e 0
dn x n
Y
1
!
x x/z p ++x /zn
cn1
1
, dn =
e
.
zn
n
If f is an entire functions of order , which is not a constant, and

such that f (o) , 0, and f (zn ) = 0, n = 1, 2, 3, . . . |zn | |zn+1 |, then
f (z) = eQ1 (z)G1 (z),
where Q1 (z) is a polynomial of degree q and
!
Y
z zz ++ 1p zz p
n ,
G1 (z) =
1
en
zn
n=1
where p + 1 is the smallest integer for which
existing, since f is of finite order.
1
< , this integer
|zn | p+1
39
Remark. This follows at once from Weierstrasss theorem but can also
be deduced from the infinite-product representation derived in the proof
of Hadamards theorem (without the use of Weierstrasss theorem). One
has only to notice that
Y
e p+1

z p+1
... 1 zzn
zn
is convergent, where p is the genus of the entire function in question, so 42

that f (z) = eQ(z)G(z) may be multiplied by this product yielding f (z) =
eQ1 (z)G1 (z), where Q1 is again of degree .
We recall the convention that if the zn s are finite in number (or if
there are no zn s), then p = 0 and 1 = 0.
We have also seen that
q , 1 i.e. max(q, 1 )
We shall now prove the following
Theorem 2. max(q, 1 )
Let
up
u2
++
p
E(u) (1 u)e 2
u+
If |u| 21 , then
p+1 p+2

u
up
u
|E(u)| = elog(1u)+u+...+ p = e p+1 p+2
e|u|
p+1 (1+ 1 + 1 + )
2 4
= e2|u|
p+1
e|2u|
p+1
e|2u| , if p + 1
If |u| > 12 , then
|E(u)| (1 + |u|)e|u|++
(1 + |u|)e|u|
(1 + |u|)e|u|
|u| p
p
p (|u|1p ++1)
p (2 p1 ++1)
40
< e|2u|
p +log(1+|u|)
+log(1+|u|)
e|2u|
, if p , since |2u| > 1.
43
Hence, for all u, we have

+log(1+|u|)
|E(u)| e|2u|
if p p + 1.
It is possible to choose in such a way that
(i) p p + 1,
(ii) > 0,
(iii)
P 1
< , (if there are an infinity of zn s) and
|zn |
(iv) 1 1 +
P 1
< (if the series is infinite this would imply that
|zn |1
1 > 0), we have only to take = > 0. And in all other cases we must
P 1
< ] and so we
have p 1 < p + 1, [for, if 1 = p + 1, then
|zn |1
P 1
choose such that 1 < < p + 1; this implies again that
<
|zn |
and > 0. For such a we can write the above inequality as
For, if
|E(u)| ec1 |u| , c1 is a constant.

44
Hence
| f (z)| e|Q(z)|+c1
c2 rq +c3 r
i.e. M(r) < e

Hence

X
z
zn
n=1
, |z| = r > 1, and c1
|1/zn | = c3 .
log M(r) = O(rq ) + O(r ), as r ,
41
= O(rmax(q,) ), as r
(2.1)
so that
max(q, )
Since is arbitrarily near 1 , we get
max(q, 1 ),
and this proves the theorem.
Theorem 3. If the order of the canonical product G(z) is , then = 1
Proof. As in the proof of Theorem 2, we have
p p + 1;

>0
|E(u)| ec1 |u|
P 1
<
|zn |
1 + ,
>0
Hence

X
z
c1
zn
n=1
|G(z)| e
ec3 r , |z| = r.
If M(r) = max |G(z)| , then
45
|z|=r
M(r) < ec3 r
Hence

which implies 1 ; but 1 . Hence = 1
Theorem 4. If, either
42
(i) 1 < q or
(ii)
then
P 1
< (the series being infinite),
|zn |1
log M(r) = O(r ), for = max(1 , q).
Proof.
(i) If 1 < q, then we take < q and from (2.1) we get

log M(r) = O(rq ) = O(rmax(1 ,q) )
1
< , then we may take = 1 > 0 and from (2.1) we
n=1 |zn |1
get
(ii) If
log M(r) = O(rq ) + O(r1 )

= O(r )

Theorem 5.
(i) We have
= max(1 , q)
(The existence of either side implies the other)

46
(ii) If > 0, and if log M(r) = O(r ) does not hold, then 1 = , f (z)
P
has an infinity of zeros zn , and |zn | is divergent.
Proof. The first part is merely an affirmation of Hadamards theorem,

1 and Theorem 2.
For the second part, we must have 1 = ; for if 1 < , then = q
by the first part, so that 1 < q, which implies, by Theorem 4, that
logM(r) = O(rq ) = O(r ) in contradiction with our hypothesis. Since
> 0, and 1 = we have therefore 1 > 0, which implies that f (z) has
P
P
an infinity of zeros. Finally |zn |1 is divergent, for if |zn |1 < ,
1
then by Theorem 4, we would have log M(r) = O(r ) which contradicts

the hypothesis.
43
Remarks. (1) 1 is called the real order of f (z), and the apparent
order Max (q, p) is called the genus of f (z).
(2) If is not an integer, then = r , for q is an integer and =
max(1 , q). In particular, if is non-integral, then f must have an
infinity of zeros.
Lecture 6
The Gamma Function

1 Elementary Properties [15, p.148]
47
Define the function by the relation.

Z
(x) =
t x1 et dt, x real.
0
(i) This integral converges at the upper limit , for all x, since
t x1 et = t2 t x+1 et = O(t2 ) as t ; it converges at the lower
limit o for x > 0.
(ii) The integral converges uniformly for 0 < a x b. For
1
Z
Z1 Z
t x1 et dt =
+
= O ta1 dt + O tb1 et dt
= O(1), independently of x.
Hence the integral represents a continuous function for x > 0.

(iii) If z is complex,
tz1 et dt is again uniformly convergent over any
finite region in which Re z a > 0. For if z = x + iy, then

|tz1 | = t x1 ,
45
6. The Gamma Function
46
and we use (ii). Hence (z) is an analytic function for Re z > 0

(iv) If x > 1, we integrate by parts and get
(x) =
i
t x1 et
0
+ (x 1)
t x2 et dt
= (x 1)(x 1).
48
However,
Z
(1) =
et dt = 1; hence
(n) = (n 1)!,
if n is a +ve integer. Thus (x) is a generalization of n!
(v) We have
1
log (n) = (n ) log n n + c + O(1),
2
where c is a constant.
log (n) = log(n 1)! =
n1
X
log r.
r=1
We estimate log r by an integral: we have

1
r+ 2
Z
log t dt =
r 12
Z2
Z2
log(s + r)ds =
21
Z1/2

0
Z0
12

log(t + r) + log(r t) dt
1. Elementary Properties
47
!#
Z1/2"
t2
2
log r + log 1 2 dt
=
r
0
!
1
= log r + cr , where cr = O 2 .
r
Hence
log (n) =
r+ 12
Z
n1
X
r=1
r 12
n 2
Z
log t dt cr
log t dt
1
2
n1
X
cr
r=1
!
!
)
1
1
1
1
1
1
log
(n ) +
n log n
2
2
2
2
2
2
X
X
cr +
Cr
n
r=1
1
= (n ) log n n + c + o(1), where c is a constant.
2
49
Hence, also
(n!) = a.en nn+ 2 eo(1) ,

where a is a constant such that c = log a
(vi) We have
(x)(y)
=
(x + y)
Z
O
ty1
dt, x > 0, y > 0.
(1 + t) x+y
For
(x)(y) =
Z
0
x1 t
e dt
Z
0
sy1 es ds, x > o, y > 0
48
Put s = tv; then
(x)(y) =
Z
0
dt
vy1 dv
t x+y1 et(1+v) dt
vy1 dv
u x+y1 eu
du
(1 + v) x+y
ty vy1 etv dv
= (x + y)
Z
0
50
x1 t
vy1
dv.
(1 + v) x+y
The inversion of the repeated integral is justified by the use of the

fact that the individual integrals converge uniformly for x > 0,
y > 0.
(vii) Putting x = y =
tan2 ]
1
in (vii), we get [using the substitution t =
2
1
{( )}2 = 2(1)
2
Z/2
d = .
0
Since ( 12 ) > 0, we get ( 21 ) =

Putting y = 1 x, on the other hand, we get the important relation
(x)(1 x) =
Z
0
ux
du =
, 0 < x < 1,
1+u
sin(1 x)
since
Z
0
xa1
dx =
for 0 < a < 1, by contour
1+x
sin a
2. Analytic continuation of (z)
49
integration. Hence
(x)(1 x) =
,
sin x
0 < x < 1.
2 Analytic continuation of (z) [15, p. 148]

We have seen that the function
(z) =
et tz1 dt
is a regular function for Re z > 0. We now seek to extend it analytically

into the rest of the complex z-plane. Consider
Z
I(z)
e ()z1 d,
C
where C consists of the real axis from to > 0, the circle || = in 51

the positive direction x and again the real axis from to .
We define
()z1 = e(z1) log()
by choosing log() to be real when =
The integral I(z) is uniformly convergent in any finite region of the
Z-plane, for the only possible difficulty is at = , but this was covered
by case (iii) in 1. Thus I(z) is regular for all finite values of z.
We shall evaluate I(z). For this, set
= ei
so that
log() = log + i( ) on the contour
[so as to conform with the requirement that log() is real when = ]

Now the integrals on the portion of C corresponding to (, ) and
(, ) give [since = 0 in the first case, and = 2 in the second]:
Z
+(z1)(log i)
d +
e+(z1)(log +i) d
50
Z h
i
e+(z1)(log i) + e+(z1)(log +i) d
h
i
e+(z1) log e(z1)i e(z1)i d
= 2i sin z
e z1 d
On the circle || = , we have

|()z1 | = |e(z1) log()| = |e(z1)[log +i()] |
= e(x1) log y()
= O( x1 ).
52
The integral round the circle || = therefore gives

O( x ) = O(1), as 0, if x > 0.
Hence, letting 0, we get
I(z) = 2i sin z
e z1 d, Re z > 0
= 2i sin z(z), Re z > 0.

We have already noted that I(z) is regular for all finite z; so
1
iI(z) cosec z
2
is regular for all finite z, except (possibly) for the poles of cosec z
namely, z = 0, 1, 2, . . .; and it equals (z) for Re z > 0.
Hence - 21 iI(z) cosec z is the analytic continuation of (z) all over
the z-plane.
3. The Product Formula
51
We know, however, by (iii) of 1, that (z) is regular for z =

1
1, 2, 3, . . .. Hence the only possible poles of iI(z) cosec x are z =
2
0, 1, 2, . . .. These are actually poles of (z), for if z is one of these
numbers, say n, then ()z1 is single-valued in 0 and I(z) can be evaluated directly by Cauchys theorem. Actually
Z
e
2i
2i
n+1
(1)
(1)n+n+1 =
d =
n+1
n!
n!
i.e. I(n) =
2i
.
n!
So the poles of cosec z, at z = 0, n, are actually poles of (z). The 53

residue at z = n is therefore
!
2i z + n (1)n
lim
=
zn
n!
2i sin z
n!
The formula in (viii) of 1 can therefore be upheld for complex values.
Thus
(z) (1 z) = cosec z
for all non-integral values of z. Hence, also,
1
is an entire function.
(z)
(Since the poles of (1 z) are cancelled by the zeros of sin z)
3 The Product Formula

1
is an entire function of order 1.
(z)
Since I(z) = 2i sin z (z), and
We shall prove that
(z) (1 z) =
,
sin z
we get
I(1 z) = 2i sin( z)(1 z)
52
= 2i sin z
or
54
Now
sin z (z)
2i
(z)
1
I(1 z)
=
.
(z)
2i
Z1
Z
|z|
t x1
t x1
e t dt + e t dt
I(z) = O e
0
1
|z|
t |z|
= O e
1 + e t dt
1+|z| Z
+
1
+
= O e|z|
1
1+|z|
i
h
= O e|z| (|z| + 1)|z|+1
h 1+ i
= O e|z| , > 0.
1
is of order 1. However, the exponent of convergence
(z)
of its zeros equals 1. Hence the order is = 1, and the genus of the
canonical product is also equal to 1.
Thus
Y
z z
1
az+b
= z, e
(1 + )e n
(z)
n
n=1
Hence
1
= 1 and
z0+ z(z)
by Hadamards theorem. However, we know that lim

(1) = 1. These substitutions give b = 0, and
!
1 1/n
1 = ea 1 +
e
or,
!
#
"
X
1
1
0=a+
log 1 +
n
n
1
3. The Product Formula
53
m "
!
#
1
1
= a lim
log
1
+
m
n
n
1
m
X 1
X
= a + lim
log(n + 1) log n
m
n
1
1
m
X
= a + lim log(m + 1)
m
n
1
= a , where is Eulers constant.
55
Hence

Y
1
z
= ez z
1 + ez/n
(z)
n
n=1
As a consequence of this we can derive Gausss expression for (z).

For
ez/n
ez/1
.
.
.
(z) = lim e
n
z 1 + 1z
1+ z
"
# n
1
1
2
n
= lim nz
...
n
z z+1 z+2
z+n
"
#
nz n!
= lim
n z(z + 1) . . . (z + n)
"
z(log n1 12 ... n1 ) 1
We shall denote
nz n!
n (z),
z(z + 1) . . . (z + n)
so that
56
(z) = lim n (z).

n
Further, we get by logarithmic-differentiation,

!
(z)
1 X 1
1
=
.
(z)
z n=1 z + n n
54
Thus
Hence
X
1
d2
[log
(z)]
=
;
2
dz2
(n
+
z)
n=0
d2
[log (z)] > 0 for real, positive z.
dz2
Lecture 7
The Gamma Function:

Contd
4 The Bohr-Mollerup-Artin Theorem
[7, Bd.I, p.276]
We shall prove that the functional equation of (x), namely (x + 1) = 57
d2
x(x), together with the fact that 2 [log (x)] > 0 for x > 0, deterdx
mines (x) essentially uniquely-essentially, that is, except for a constant of proportionality. If we add the normalization condition (1) = 1,
then these three properties determine uniquely. N.B. The functional
equation alone does not define uniquely since g(x) = p(x)(x), where
p is any analytic function of period 1 also satisfies the same functional
equation.
We shall briefly recall the definition of Convex functions. A realvalued function f (x), defined for x > 0, is convex, if the corresponding
function
f (x + y) f (x)
(y) =
,
y
defined for all y > x, y , 0, is monotone increasing throughout the
range of its definition.
If 0 < x1 < x < x2 are given, then by choosing y1 = x1 x and
55
7. The Gamma Function: Contd
56
y2 = x2 x, we can express the condition of convexity as:

f (x)
58
x2 x
x x1
f (x1 ) +
f (x2 )
x2 x1
x2 x1
Letting x x1 , we get f (x1 + 0) f (x1 ); and letting x2 x, we

get f (x) f (x + 0) and hence f (x1 + 0) = f (x1 ) for all x1 . Similarly
f (x1 0) = f (x1 ) for all x1 .
Thus a convex function is continuous.
Next, if f (x) is twice (continuously) differentiable, we have
(y) =
y f (x + y) f (x + y) + f (x)
y2
and writing x + y = u, we get

f (x) = f (u y) = f (u) y f (u) +
y2
f [u (1 )y], 0 1
2
which we substitute in the formula for (y) so as to obtain

(y) =
1
f [x + y].
2
Thus if f (x) 0 for all x > 0, then (y) is monotone increasing and f
is convex. [The converse is also true].
Thus log (x) is a convex function; by the last formula of the previous section we may say that (x) is logarithmically convex for x > 0.
Further (x) > 0 for x > 0.
Theorem. (x) is the positive function uniquely defined for x > 0 by the
conditions:
(1.) (x + 1) = x(x)
(2.) (x) is logarithmically convex
(3.) (1) = 1.
4. The Bohr-Mollerup-Artin Theorem
59
57
Proof. Let f (x) > 0 be any function satisfying the above three conditions satisfied by (x).
Choose an integer n > 2, and 0 < x 1.
Let
n 1 < n < n + x n + 1.
By logarithmic convexity, we get
log f (n 1) log f (n) log f (n + x) log f (n)
<
(n 1) n
(n + x) n
log f (n + 1) log f (n)
(n + 1) n
Because of conditions 1 and 3, we get
f (n 1) = (n 2)!, f (n) = (n 1)!, f (n + 1) = n!

and
f (n + x) = x(x + 1) . . . (x + n 1) f (x).
Substituting these in the above inequalities we get
log(n 1) x < log
f (n + x)
log n x ,
(n 1)!
which implies, since log is monotone,
f (n + x)
nx
(n 1)!
(n 1)!(n 1) x
(n 1)!n x
< f (x)
x(x + 1) . . . (x + n 1)
x(x + 1) . . . (x + n 1)
or
(n 1) x <
Replacing (n 1) by n in the first inequality, we get

x+n
, n = 2, 3, . . .
n (x) < f (x) n (x)
n
where n is defined as in Gausss expression for .
Letting n , we get
f (x) = (x), 0 < x 1.
For values of x > 1, we uphold this relation by the functional equation.

58
5 Gausss Multiplication Formula. [7, Bd.I, p.281]

60
Let p be a positive integer, and

!
!
!
x
x+1
x+ p1
x
f (x) = p
...
p
p
p
[If p = 1, then f (x) = (x)]. For x > 0, we then have
dx
[log f (x)] > 0.
dx2
Further
f (x + 1) = x f (x).
Hence by the Bohr-Artin theorem, we get
f (x) = a p (x).
Put x = 1. Then we get
!
!
!
2
p
1
...
a p = p
p
p
p
(5.1)
[a1 = 1, by definition]. Put k = 1, 2, . . . p, in the relation

!
k
nk/p n!pn+1
n
,
=
p
k(k + p) . . . (k + np)
and multiply out and take the limit as n . We then get from (5.1),
a p = p. lim
p+1
2
(n!) p pnp+p
(np + p)!
However
!
!
!#
"
2
p
1
1+
... 1 +
(np + p)! = (np)!(np) p 1 +
np
np
np
Thus, for fixed p, as n , we get
(n!) p pnp
n (np)!n(p1)/2
a p = p lim
6. Stirlings Formula
61
59
Now by the asymptotic formula obtained in (v) of 1, [see p.49]

(n!) p = a p nnp+p/2 enp eo(1)
1
(np)! = annp+ 2 pnp+ 2 enp eo(1)

Hence
ap =
p a p1 .
Putting p = 2 and using (5.1) we get

!
1
1
1
(1) = 2
a = , a2 = 2
2
2
2
Hence
ap =
(5.2)
p(2)(p1)/2 .
Thus
!
!
x
x+ p1
p
...
= p(2)(p1)/2 (x)
p
p
x
For p = 2 and p = 3 we get:
x x + 1! x + 2!
x x + 1!
(x).
= x1 (x);
=
1
2
2
3
3
3
2
3 x 2
6 Stirlings Formula [15, p.150]

From (v) of 1, p.49 and (5.2) we get
1
(n!) = 2 en nn+ 2 eo(1) ,
which is the same thing as
!
1
log(n 1)! = n log n n + log 2 + o(1)
2
(6.1)
Further
62
1+
1
1
+ +
log N = + 0(1), as N
2
N1
(6.2)
60
and

z
log(N + z) = log N + log 1 +
N !
z
1
= log N + + O 2 , as N .
N
N
(6.3)
We need to use (6.1)-(6.3) for obtaining the asymptotic formula for

(z) where z is complex.
From the product-formula we get
log (z) =

X
z
n=1

z
z log z,
log 1 +
n
n
each logarithm having its principal value.

Now
ZN
0
N1
X
[u] u + 21
du =
u+z
n=0
N1
X
Zn+1
1
n + 2 + z
1 du
u+z
n
1
+ z)[log(n + 1 + z) log(n z)] N
2
n=0
! N1
N1
X
1 X
=
n[log(n + 1 + z) log(n + z)] + z +
2 n=0
n=0

log(n + 1 + z) log(n + z) N
!
N1
X
1
= (N 1) log(N + z)
log(n + z) + z + log(N + z)
2
n=1
!
1
z + log z N
2
!
!
1
1
= N + z log(N + z) z + log z
2
2
N1

X
z
N
log n + log 1 +
n
n=1
=
(n +
(6.4)
6. Stirlings Formula
61
!
!
1
1
= N + z log(N + z) z + log z N log(N 1)!
2
2
N1
N1

X1
X z
z
z
log 1 +
+
n
n
n
1
n=1
63
Now substituting for log(N + z) etc., from (6.1)-(6.3), we get

ZN
0
!)
!
(
[u] u + 21
1
1
z
1
du = (N + z) log N + + O 2 z +
u+z
2
N
2
N
!
1
log z N N log N + N c + O(1)+
2
N1
N1

X1
Xz
z
log 1 +
z
n
n
n
1
n=1
N1
X 1
+ z + O(1) log 2
= z log N
n
1
N1

Xz
1
z
log z log z +
log 1 +
2
n
n
n=1
Now letting N , and using (6.4), and (6.2), we get

Z
0
!
[u] u + 12
1
1
du + log 2 z + z log z = log (z)
u+z
2
2
Ru
If (u) = ([v] v + 12 )dv, then
0
(u) = O(1), since (n + 1) = (n),
if n is an integer.
Hence
Z
Z
Z
[u] u + 21
d(u)
(u)
du =
=
u+z
u+z
(u + z)2
0
64
(6.5)
62
Z
dv
= O
|u + z|2
0
If we write z = rei , and u = ru1 , then
Z
[u] u + 21
1
du
1
du = O
r
u+z
|u1 + i |2
0
The last integral is finite if , . Hence

Z
0
!
[u] u + 21
1
du = O
, uniformly for |arg z| <
u+z
r
Thus we get from (6.5):

!
!
1
1
1
log (z) = z log z z + log 2 + O
2
2
|z|
for |arg z| <
65
Corollaries. (1) log (z + ) = (z + 12 ) log z z + 12 log 2 + O

as |z| uniformly for |arg z| < . bounded.
(2) For any fixed x, as y
1
|(x + iy)| e 2 |y| |y| x 2

1
|z|
!
(z)
1
1
(3)
= log z
+ O 2 , |arg z| < [11, p.57]
(z)
2z
|z|
This may be deduced from (1) by using
Z
1
f ()
d,
f (z) =
2i
( z)2
C

where f (z) = log (z) z 21 log z + z 21 log 2 and C is a circle
with centre at = z and radius |z| sin /2.
Lecture 8
The Zeta Function of

Riemann
1 Elementary Properties of (s) [16, pp.1-9]
We define (s), for s complex, by the relation
(s) =
X
1
, s = + it, > 1.
ns
n=1
66
(1.1)
We define x s , for x > 0, as e s log x , where log x has its real determination.
Then |n s | = n , and the series converges for > 1, and uniformly
in any finite region in which 1 + > 1. Hence (s) is regular
for 1 + > 1. Its derivatives may be calculated by termwise
differentiation of the series.
We could express (s) also as an infinite product called the Euler
Product:
!1
Y
1
(s) =
1 s
, > 1,
(1.2)
p
p
where p runs through all the primes (known to be infinite in number!).
It is the Euler product which connects the properties of with the properties of primes. The infinite product is absolutely convergent for > 1,
63
64
8. The Zeta Function of Riemann
since the corresponding series

X 1 X 1
=
< , > 1
ps
p
p
p
Expanding each factor of the product, we can write it as

!
Y
1
1
1 + s + 2s +
p
p
p
67
Multiplying out formally we get the expression (1.1) by the unique

factorization theorem. This expansion can be justified as follows:
!
Y
1
1
1
1
1 + s + 2s + = 1 + s + s + ,
p
n
n
p
1
2
pp
where n1 , n2 , . . . are those integers none of whose prime factors exceed
P. Since all integers P are of this form, we get

!1 X
X
Y
1

1
1
1
1

=

1

n s pP
n1s n2s
p2 n=1 n s

n=1
1
1
+
+
(P + 1) (P + 2)
0, as P , if > 1.
Hence (1.2) has been established for > 1, on the basis of (1.1).
As an immediate consequence of (1.2) we observe that (s) has no zeros
for > 1, since a convergent infinite product of non-zero factors is
non-zero.
We shall now obtain a few formulae involving (s) which will be of
use in the sequel. We first have
!
X
1
(1.3)
log (s) =
log 1 2 , > 1.
p
p
1. Elementary Properties of (s)

If we write (x) =
65
1, then
px
!
1
log (s) = {(n) (n 1)} log 1 s
n
n=2
"
!
!#
X
1
1
=
(n) log 1 s log 1
n
(n + 1) s
n=2
=
X
n=2
(n)
Zn+1
n
s
dx
x(x s 1)
68
Hence
log (s) = s
Z
2
(x)
dx, > 1.
x(x s 1)
(1.4)
It should be noted that the rearrangement of the series preceding the

above formula is permitted because
!
1
(n) n and log 1 s = O(n ).
n
A slightly different version of (1.3) would be
log (s) =
XX
p
1
,
mpms
>1
(1.3)
where p runs through all primes, and m through all positive integers.
Differentiating (1.3), we get
(s) X ps log p X X log

=
=
,
(s)
1 ps
pms
p
p,m
or
(s) X (n)
=
, >1
s
(s)
n
n=1
(1.5)
66
log p, if n is a + ve power of a prime p

where (n) =
0, otherwise.
Again
!
Y
1
1
=
1 s
(s)
p
p
=
X
(n)
ns
n=1
>1
69
(1.6)
(1.6) where (1) = 1, (n) = (1)k if n is the product of k different

primes, (n) = 0 if n contains a factor to a power higher than the first.
We also have
X
d(n)
, >1
2 (s) =
ns
n=1
where d(n) denotes the number of divisors of n, including 1 and n. For
2 (s) =
X
1 X 1 X 1 X
1
=
s
s
s
m
n
mn=
m=1
n=1
=1
More generally
k
(s) =
X
dk (n)
ns
n=1
70
>1
(1.7)
where k = 2, 3, 4, . . ., and dk (n) is the number of ways of expressing n

as a product of k factors.
Further
(s) (s a) =
=
X
1 X na
,
m s n=1 n s
m=1
X
1 X a
n ,
s mn=
=1
so that
(s) (s a) =
X
2 ()
=1
(1.8)
1. Elementary Properties of (s)
67
where a () denotes the sum of the ath powers of the divisors of .

If a , 0, we get, from expanding the respective terms of the Euler
products,
!
!
Y
pa p2a
1
1
1 + s + 2s + 1 + s + 2a +
(a)(s a) =
p
p
p
p
p
!
Y
1 + pa 1 + pa + p2a
1+
=
+
+
ps
p2s
p
!
Y
1 p2a 1
+
1+
=
1 pa p s
p
Using (1.8) we thus get
1 +1)a
1 p(m
1 pr(mr +1)a
1
2 (n) =
.
.
.
1 pa1
1 par
(1.9)
1
m2
mr
1
if n = pm
1 , p2 . . . pr by comparison of the coefficients of n s .
More generally, we have
1 p2s+a+b
(s) (s a) (s b) (s a b) Y
=
(2s a b)
(1 ps )(1 ps+a )(1 ps+b )
p
(1 ps+a+b )
for > max {1, Re a + 1, Re b + 1, Re(a + b) + 1}. Putting ps = z, we

get the general term in the right hand side equal to
1 pa+b z2
(1 z) (1 pa z) (1 pb z)(1 pa+b z)
(
)
pa+b
1
1
pa
pb
+
=
(1 pa )(1 pb ) 1 z 1 pa z 1 pb z 1 pa+b z
n
X
o
1
(m+1)a
(m+1)b
(m+1)(a+b) m
1
p
+
p
z
=
(1 pa )(1 pb ) m=0
=
n
X
on
o
1
(m+1)a
(m+1)b m
1
p
1
p
z
(1 pa )(1 pb ) m=o
71
68
Hence
72
(s) (s a)(s b) (s a h)
(2s a b)
YX
1 p(m+1)a 1 p(m+1)b 1
=
ms
1 pa
p
1 pb
p m=0
Now using (1.9) we get
(s) (s a) (s b)(s a b) X a (n)b (n)

=
(2s a b)
ns
n=1
(1.10)
> max{1, Re a + 1, Re b + 1, Re(a + b) + 1}

If a = b = 0, then
{(s)}4 X {d(n)2 }
=
,
(2s)
ns
n=1
get
> 1.
(1.11)
If is real, and , 0, we write i for a and i for b in (1.10), and
2 (s)(s i)(s + i) X |i (n)|2

=
, > 1,
(2s)
ns
n=1
P i
d .
where i (n) =
d/n
(1.12)
Lecture 9

Riemann (Contd.)
2 Elementary theory of Dirichlet Series [10]
73
The Zeta function of Riemann is the sum-function associated with the

P 1
. We shall now study, very briefly, some of the
Dirichlet series
ns
elementary properties of general Dirichlet series of the form
X
n=1
an en s , o 1 < 2 < . . .
Special cases are: n = log n; n = n, es = x. We shall first prove a

lemma applicable to the summation of Dirichlet series.
Lemma. Let A(x) =
n x
an , where an may be real or complex. Then, if
x 1 , and (x) has a continuous derivative in (0, ), we have

X
an (n ) =
A(x) (x)dx + A()().
69
9. The Zeta Function of Riemann (Contd.)
70
If A()() 0 as , then
X
n=1
an (n ) =
A(x) (x)dx,
provided that either side is convergent.

Proof.
A()()
=
an (n )
an {() (n )}
XZ
an (x)dx
A(x) (x)dx
74
Theorem 1. If
n=1
an en s converges for s = s0 0 + ito , then it
converges uniformly in the angular region defined by

|am(s s0 )| < /2,
for
0 < < /2
Proof. We may suppose that so = 0. For

X
X
X
an en s =
an en s0 en(ss0 )
bn en s , say,
and the new series converges at s = 0. By the above lemma, we get
X
+1
n s
an e
X X
s
= an en
1
2. Elementary theory of Dirichlet Series
71
A(x) exs sdx + A( )e s A( )e s
=s
{A(x) A( )}exs dx + [A( ) A( )]e s
P
We have assumed that an converges, therefore, given > 0, there 75
exists an n0 such that for x > n0 , we have
|A(x) A( )| <
Hence, for such , we have

Z
X

an en s |s| ex dx + e

+1
|s|
+
< 2 (sec + 1),
|s|
< sec , and this proves the theorem. As a conseif , 0, since
quence of Theorem 1, we deduce

P
s
Theorem 2. If an en converges for s = s0 , then it converges for
Re s > o , and uniformly in any bounded closed domain contained in
the half-plane > o . We also have
P
Theorem 3. If an en s converges for s = s0 to the sum f (s0 ), then
f (s) f (s0 ) as s s0 along any path in the region |am(s s0 )| <
/2.
A Dirichlet series may converge for all values of s, or some values
of s, or no values of s.
Ex.1.
an ns , an =
1
converges for all values of s.
n!
72
Ex.2.
Ex.3.
76
an ns , an = n! converges for no values of s.

ns converges for Re s > 1, and diverges for Re s 1.
If a Dirichlet series converges for some, but not all, values of s, and
if s1 is a point of convergence while s2 is a point of divergence, then,
on account of the above theorems, we have 1 > 2 . All points on the
real axis can then be divided into two classes L and U; U if the
series converges at , otherwise L. Then any point of U lies to the
right of any point of L, and this classification defines a number o such
that the series converges for every > o and diverges for < o ,
the cases = 0 begin undecided. Thus the region of convergence is a
half-plane. The line = 0 is called the line of convergence, and the
number o is called the abscissa of convergence. We have seen that 0
may be . On the basis of Theorem 2 we can establish
Theorem 4. A Dirichlet series represents in its half-plane of convergence a regular function of s whose successive derivatives are obtained
by termwise differentiation.
We shall now prove a theorem which implies the uniqueness of
Dirichlet series.
P
s
Theorem 5. If an en converges for s = 0, and its sum f (s) = 0 for
an infinity of values of s lying in the region:
> 0,
|am s| < /2,
then an = 0 for all n.
77
Proof. f (s) cannot have an infinite number of zeros in the neighbourhood of any finite point of the giver region, since it is regular there.
Hence there exists an infinity of valves sn = n + itn , say, with n+1 >
, lim n = such that f (sn ) = 0.
However,
g(s) e1 s f (s) = a1 +
X
2
an e(n 1 )s ,
2. Elementary theory of Dirichlet Series
73
(here we are assuming 1 > 0) converges for s = 0, and is therefore

uniformly convergent in the region given; each term of the series on the
right 0, as s and hence the right hand side, as a whole, tends
to a1 as s . Thus g(s) a1 as s along any path in the given
region. This would contradict the fact that g(sn ) = 0 for an infinity of
sn , unless a1 = 0. Similarly a2 = 0, and so on.

Remark . In the hypothesis it is essential that > 0, for if = 0, the
origin itself may be a limit point of the zeros of f , and a contradiction
does not result in the manner in which it is derived in the above proof.
P
The above arguments can be applied to an en s so as to yield the
existence of an abscissa of absolute convergence ,
etc. In general,
0 . The strip that separates

from o may comprise the whole
plane, or may be vacuous or may be a half-plane.
Ex.1. 0 = 0,
=1
1s
(1 2
Ex.2.
!
!
1
1
1
1
1
)(s) = s + s + s + 2 s + s +
1
2
3
2
4
!
1
1
1
1
= s s + s s +
1
2
3
4
P (1)n
(log n)s converges for all s but never absolutely. In the 78
n
simple case n = log n, we have
Theorem 6.
0 1.
P
P |an |
For if an ns < , then |an | n = O(1), so that
<
n s+1+
for > 0
While we have observed that the sum-function of a Dirichlet series
is regular in the half-plane of convergence, there is no reason to assume
that the line of convergence contains at least one singularity of the function (see Ex 1. above!). In the special case where the coefficients are
positive, however we can assert the following
Theorem 7. If an 0, then the point s = 0 is a singularity of the
function f (s).
74
Proof. Since an 0, we have 0 = ,

and we may assume, without
loss of generality, that 0 = 0. Then, if s = 0 is a regular point, the
Taylor series for f (s) at s = 1 will have a radius of convergence > 1.
(since the circle of convergence of a power series must have at least one
singularity). Hence we can find a negative value of s for which
f (s) =
X
(s 1)
=0
79
f () (1) =
X
(1 s) X
=0
an n en
n=1
Here every term is positive, so the summations can be interchanged and

we get
X
X
(1 s) n X
s
f (s) =
an en
=
an en
!
n=1
=0
n=1
Hence the series converges for a negative value of s which is a contradiction.
Lecture 10

Riemann (Contd)
2 (Contd). Elementary theory of Dirichlet series
80
We wish to prove a formula for the partial sum of a Dirichlet series. We

shall give two proofs, one based on an estimate of the order of the sumfunction [10], and another [14, Lec. 4] which is independent of such an
estimate. We first need the following
Lemma 1. If > 0, then
+i
Z
1
2i
where
+i
R
= lim
+iT
R
1,
eus
ds =
0,
1,
2
if u > 0
if u < 0
if u = 0
iT
Proof. The case u = 0 is obvious. We shall therefore study only the

cases u 0. Now
+iT
Z
iT
eus ds eus
=
s
us
+iT
Z
iT
75
+iT
Z
iT
eus
ds
us2
10. The Zeta Function of Riemann (Contd)
76
However
|eu(+iT ) | = eu
and |s| > T ; hence, letting T , we get

+i
Z
eus ds
=
s
eus
ds I, say.
us2
81
+i
Z
We shall evaluate I on the contour suggested by the diagram. If

u < 0 , we use the contour L + CR ; and if u > 0, we use L + C L .
If u < 0, we have, on CR ,
us
u
e e , since Re s > .
s2
r2
If u > 0, we have, on C L ,
us
u
e e , since Re s < .
s2
r2
Hence

Z
us
eu
e ds

0, as r .
2r

2
2
us
|u|

R ,C L
C
By Cauchys theorem, however, we have

Z
Z
= ;
L
CR
2. (Contd). Elementary theory of Dirichlet series
77
hence
1
2i
(1)
+i
Z
eus
ds = 0,
us2
if u < 0.
If, however, u > 0, the contour L + C L encloses a pole of the second

order at the origin, and we get
Z
Z
= +1,
L
CL
so that
1
2i
(2)
+i
Z
eus
ds = 1,
us2
if u > 0.

Remarks. We can rewrite the Lemma as: if a > 0,
a+i
0, if < n
Z (n )s
e
1
ds =
1, if > n
2i
s
ai
if = n
2,
Formally, therefore, we should have
1
2i
a+i
Z
ai
a+i
Z (n )s
1 X
e
es
ds
f (s) ds =
an
s
2i 1
s
ai
X
an
=
n
the dash denoting that if = n the last term should be halved. We wish
now to establish this on a rigorous basis.
We proceed to prove another lemma first.
82
78
Lemma 2. Let
an en s converge for s = real. Then

n
X
a e s = o(|t|),
the estimate holding uniformly both in + > and in n. In

particular, for n = , f (s) = o(|t|) uniformly for + >
Proof. Assume that = 0, without loss of generality. Then
a = O(1) and A( ) A( ) = O(1) for all and
If 1 < N < n, then
n
X
a e
N
n
X
X
=
+
a e s
1
N+1
N
X
a e
+s
Zn
{A(x) A(N )}exs dx
83
{A(n ) A(N )}en
Z n
= O(N) + O(1) + O |s| ex dx
N
!
|s| N
= O(N) + O
e

= O(N) + O |t|eN
since |s|2 = t2 + 2 and .

Hence
n
X
a e s = O(N) + O(|t|eN ), if 1 < N < n
1
If N n, then, trivially,
n
X
1
a e s = O(N).
79
If we choose N as a function of |t|, tending to more slowly than |t|, we

get
n
X
a e s = O(|t|)
1
in either case.
84
Theorem (Perrons Formula). Let 0 be the abscissa of convergence of

P
an en s whose sum-function is f (s). Then for > 0 and > 0, we
have
+i
Z
X
f (s) s
1
e ds,
an =
2i
s

where the dash denotes the fact that if = n the last term is to be
halved.
Proof. Let m < m+1 , and
n s
f
(s)
a
e
g(s) = e
(This has a meaning since Re s > 0 ). Now

1
2i
+i
Z
g(s)
ds
s
1
=
2i
+i
Z
X
f (s)es
ds
an
s

n
by the foregoing lemma. We have to show that the last expression is

zero.
Applying Cauchys theorem to the rectangle
iT 1 , + iT 2 , + iT 2 , iT 1 ,
80
where T 1 , T 2 > 0, and > , we get

1
2i
+i
Z
g(s)
ds = 0,
s
85
for
1
2i
+iT
Z 2
iT 1
iT +iT +iT
Z 1 Z 2 Z 2
1
=
+
+
2i
iT 1
iT 1
+iT 2
= I1 + I2 + I3 , say.
Now, if we fix T 1 and T 2 and let , then we see that I2 0,
P
s
for g(s) is the sum-function of bn n , where bn = am+n , n = m+n ,
1
with 1 > 0, and hence g(s) = o(1) as s in the angle |ams| .

2
Thus
+iT
Z 1 +iT
Z 2
Z 2
iT
1
1
g(s)
ds =
2i
s
2i
iT 1
iT 1
+iT 2
if the two integrals on the right converge.

Now if
h(s) g(s)e(m+1 )s = am+1 + am+2 e(m+2 m+1 )s +

then by lemma 2, we have
|h(s)| < T 2
86
for s = + iT 2 , 0 + , T 2 T 0 . Therefore for the second integral

on the right we have

Z 2
Z

+iT
T 2
g(s)

ds < q
e1 di < /1

s

2
2

+T
+iT 2
1 > 0. This proves not only that the integral converges for a finite T 2 ,
but also that as T 2 , the second integral tends to zero. Similarly for
the first integral on the right.
81
N.B. An estimate for g(s) alone (instead of h(s)) will not work!
Remarks. More generally we have for > 0 and >
X
an en
for
an en S
+i
Z
f (s) (ss )
e
ds,
s s
i
X
n (ss )
e
=
an en s = f (s),
1
=
2i
and writing s = s s and bn = an en s , we have f (s) f (s ) =

P
bn en s .
We shall now give an alternative proof of Persons formula without
using the order of f (s) as s [14, Lec. 4]
Aliter. If f (s) =
P
1
an en s has 0 , as its abscissa of convergence,
and if a > 0, a > 0 , then for , n , we have

1
an =
2i
<
a+i
Z
f (s)
es
ds
s
ai
Proof. Define
87
fm (s) =
m
X
an en s
and
rm (s) =
m+1
an .en s = f (s) fm (s)
Then
(A)
1
2i
a+iT
Z
aiT
1
es
ds =
f (s)
s
2i
a+iT
Z
aiT
1
es
ds +
fm (s)
s
2i
a+iT
Z
rm (s)
es
ds
s
aiT
Here the integral on the left exists since f (s) is regular on = a > 0 .
By Lemma 1, as applied to the first integral on the right hand side, we
82
get
(B)
1
lim
T 2i
a+iT
Z
aiT
a+iT
Z
X
1
es
es
ds
an = lim
ds
f (s)
rm (s)
T 2i
s
s
<
n
aiT
the limits existing.

Since n , we can choose m0 such that for all m m0 we have
m > . For all such m the last relation holds.
P
Now write bn = an en , (x) = ex(s ) , so that an en s =

P
P
bn (). Write B(x) =
bn .
nx
Then applying the Lemma of the 9th lecture, we get

N
X
m+1
n s
an e
N
m
X
X
=
= (s )
ZN
B(x)ex(s ) dx
+ B(N )eN (s ) B(m)em (s )
= {B(N ) B(m )}eN (s )
ZN
+ (s ) {B(x) B(m )}ex(s ) dx.

m
88
Now choose > 0 and a > > 0 . [If o 0 the second

condition implies the first; if 0 < 0, then since a > 0, choose 0 < <
a].
Then B(x) tends to a limit as x , so that B(x) = o(1) for all x;
while |ex(s ) | = ex( ) 0 as x for > . Hence, letting

N , we get
rm (s) = (s )
{B(x) B(m )}ex(s ) dx,

or
|rm (s)| c
Now consider
1
2i
83
|s | m ( )
e

rm (s)
es
ds
s
L+CR
where L + CR is the contour indicated in the diagram.
rm (s) is regular in this contour, and the integral is therefore zero. 89

Hence
Z
Z
=
L
CR
On CR , however, s = + rei and a, so that

r
|rm (s)| c
em r cos
a
|es | = e( +r cos ) , |s| r.

Hence

Z/2
Z

s
cr
e
1

rm (s)
e(m )r cos d
ds
e
2i
)
s
2(a

/2
CR
cr
e
(a )
Z/2
e(m )r sin d
84
2
for 0 /2; hence

a+iT
Z/2
Z

s
2
e
cre

1
e (m )r d
ds
rm (s)
2i
s
(a )
aiT
0
However, sin
cre
,
2(a )(m )r
for all T . Hence

a+iT

Z
s
c1
e

ds
lim
rm (s)
T
s
(m )
aiT
(C)
90
for all m mo .
Now reverting to relation (A), we get

a+iT
a+iT
Z
Z

s
s
1
e
1
e

(D) lim
f (s)
ds
fm (s)
ds
T 2i
s
2i
s

aiT
aiT

a+iT
Z

s
1
e

ds
rm (s)
= lim
T 2i
s

aiT
for every m. However,
1
lim
T 2i
a+iT
Z
aiT
fm (s)
X
es
ds =
an
s
<
independently of m. Hence1
1
We use the fact that if

lim | f (T ) g(T )| =
and lim g(T ) = , then

T
lim | f (T ) |
85

a+iT
Z

s
X
1
e

lim
ds
an left hand side in (D)
f (s)
T 2i
s

n <
aiT
c1
for m > mo .
m
Letting m we get the result.
Remarks.
(i) If = n the last term in the sum
1
multiplied by .
2
91
P
an has to be
n <
(ii) If a < 0, (and a > 0 ) then in the contour-integration the residue

at the origin has to be taken, and this will contribute a term - f (0).
In the proof, the relation between |s| and r has to be modified.
Lecture 11

Riemann (Contd)
3 Analytic continuation of (s). First method [16,
p.18]
92
We have, for > 0,

(s) =
x s1 ex dx.
Writing nx for x, and summing over n, we get
X
(s)
n=1
ns
Z
X
n=1n=0
Z X
nx
=
e x s1 dx, if > 1.
0
since
x s1 enx dx
n=1

Z
Z
X
X
X

()
s1 nx
1 nx
< ,
x
e
dx
x
e
dx
=

n

n=1
n=1
n=1
0
87
88
if > 1. Hence
X
(s)
n=1
ns
Z
0
ex s1
x dx =
1 ex
Z
0
x s1 dx
ex 1
or
(s)(s) =
Z
0
x s1
dx, > 1
ex 1
In order to continue (s) analytically all over the s-plane, consider

the complex integral
Z s1
z
dz
I(s) =
e1
C
93
where C is a contour consisting of the real axis from + to , 0 < <

2, the circle |z| = ; and the real axis from to . I(s), if convergent,
is independent of , by Cauchy s theorem.
Now, on the circle |z| = , we have
|z s1 | = |e(s1) log z | = |e{(1)+it}{log |z|+i arg z} |
= e(1) log |z|t arg z
= |z|1 e2|t| ,
while
|ez 1| > A|z|;
Hence, for fixed s,
Z
2 1 2|t|
|
|
e
0 as 0, if > 1.
A
|z|=
Thus, on letting 0, we get, if > 1,

I(s) =
Z
0
x s1
dx +
ex 1
Z
0
(xe2i ) s1
dx
ex 1
3. Analytic continuation of (s). First method
89
= (s)(s) + e2is (s)(s)

= (s)(s)(e2s 1).
Using the result

(s)(1 s) =
sin s
we get
94
e2is
1
(s)
2i. is
(1 s)
e eis
(s)
=
2i eis ,
(1 s)
I(s) =
or
eis (1 s)
(s) =
2i
Z
C
z s1 dz
, > 1.
ez 1
The integral on the right is uniformly convergent in any finite region of

the s-plane (by obvious majorization of the integrand), and so defines
an entire function. Hence the above formula, proved first for > 1,
defines (s), as a meromorphic function, all over the s-plane. This only
possible poles are the poles of (1 s), namely s = 1, 2, 3, . . .. We know
that (s) is regular for s = 2, 3, . . . (As a matter of fact, I(s) vanishes at
these points). Hence the only possible pole is at s = 1.
Hence
Z
dz
I(1) =
= 2i,
z
e 1
C
while
1
+
s1
Hence the residue of (s) at s = 1 is 1.
We see in passing, since
(1 s) =
that
1
1 1
z
z3
=
+
B
B
+
1
2
ez 1 z 2
2!
4!
1
(0) = , (2m) = 0, (1 2m) =
2
(1)m Bm
2m
, m = 1, 2, 3, . . .
95
90
4 Functional Equation (First method) [16, p.18]

Consider the integral
Z
taken along Cn as in the diagram.
z s1 dz
ez 1
Between C and Cn , the integrand has poles at

2i, . . . , 2ni.
The residue at
2mi is (2me 2 i ) s1
while the residue at 2mi is (2me3/2i ) s1 ; taken together they amount
to
h i
i
i
(2m)ei(s1) e 2 (s 1) + e 2 (s1)
= (2m) s1 ei(s1) 2 cos (s 1)

2
s1 is
= 2(2m) e sin s
2
4. Functional Equation (First method)

96
Hence
I(s) =
91
n
sin s is X
z s1 dz
+ 4i
e
(2m) s1
ez 1
2
m=1
Cn
by the theorem of residues.

Now let < 0, and n . Then, on Cn ,
|z s1 | = O(|z|1 ) = O(n1 ),
and
1
= O(1),
ez 1
for
|ez 1|2 = |e x+iy 1|2 = |e x (cos y + i sin y) 1|2

= e2x 2e x cos y + 1,
which, on the vertical lines, is (e x 1)2 and, on the horizontal lines,

= (e x + 1)2 . (since cos y = 1 there).
Also the length of the square-path is O(n). Hence the integral round
the square 0 as n .
Hence
is sin s
I(s) = 4ie
is
= 4ie
(2m) s1
m=1
s
(2) s1 (1 s), if < 0.
sin
2
or
97
(s)(s)(e2is 1) = (4i)(2) s1 eis sin

= 2ieis
s
(1 s)
2
(s)
(1 s)
Thus
(s) = (1 s)(1 s)2 s s1 sin
s
,
2
92
for < 0, and hence, by analytic continuation, for all values of s (each
side is regular except for poles!).
This is the functional equation.
Since
x x + 1!
= x1 (x),
2
2
2
we get, on writing x = 1 s,
!
1s
s
s
2
1
= (1 s);
2
2
also
Hence
s
s
1
=
2
2
sin s
2
(1 s) = 2s
Thus
!
s
1 s s 1
{sin }1
2
2
2
s/2 (s/2)(s) =
98
(1s)
2
!
1s
(1 s)
2
If
1
s(s 1)s/2 (s/2)(s)
2
1
s(s 1)(s),
2
(s)
then (s) = (1 s) and (s) = (1 s). If (z) = ( 12 + iz), then

(z) (z).
5 Functional Equation (Second Method) [16, p.13]

Consider the lemma given in Lecture 9, and write n = n, an = 1,
(x) = xs in it. We then get
X
nx
=s
ZX
1
[X]
[x]
dx + s , if X 1.
X
x s+1
5. Functional Equation (Second Method)

s
s
s
=
s 1 (s 1)X s1
ZX
93
1
X [X]
x [x]
dx + s1
s+1
Xs
x
X
Since

1 = 1/X 1 , and X [X] 1/X ,

X s1
Xs
we deduce, on making X ,
s
(s) =
s
s1
x [x]
dx, if > 1
x s+1
or
(s) = s
[x] x +
1
2
x s+1
dx +
1
1
+ , if > 1
s1 2
1
5.1 Since [x] x + is bounded, the integral on the right hand side 99
2
converges for > 0, and uniformly in any finite region to the right of
= 0. Hence it represents an analytic function of s regular for > 0,
and so provides the continuation of (s) up to = 0, and s = 1 is clearly
a simple pole with residue 1.
For 0 < < 1, we have, however,
Z1
[x] x
dx =
x s+1
xs dx =
and
s
=
2
Z
1
Hence
Z1
dx
1
=
s+1
2
x
1
,
s1
94
5.2
(s) = s
[x] x
dx,
x s+1
0<<1
We have seen that (5.1) gives the analytic continuation up to = 0.

By refining the argument dealing with the integral on the right-hand side
of (5.1) we can get the continuation all over the s-plane. For, if
1
f (x) [x] x + ,
2
f1 (x)
Zx
f (y)dy,
then f1 (x) is also bounded, since
n+1
R
f (y)dy = 0 for any integer n.
100
Hence
Zx2
x1
x2
Zx2
f1 (x)
f (x)
f1 (x)
dx = s+1 + (s + 1)
dx
s+1
x
x
x s+2
x1
x1
0, as x1 , x2 ,
if > 1.
Hence the integral in (5.1) converges for > 1.

Further
s
Z1
[x] x +
x s+1
1
2
dx =
(s) = s
1
1
+ , for < 0;
s1 2
[x] x +
1
2
x s+1
dx, 1 < < 0
Now the function [x] x +
1
2
has the Fourier expansion
X
sin 2nx
n=1
(5.3)
5. Functional Equation (Second Method)
95
if x is not an integer. The series is boundedly convergent. If we substitute it in (5.3), we get
sX1
(s) =
n=1 n
s
=
X
n=1
sin 2nx
dx
x s+1
(2n) s
n
sin y
dy
y s+1
s
s
(s) = (2) s {(1 s)} sin (1 s),
5.4 If termwise integration is permitted, for 1 < < 0. The right 101
hand side is, however, analytic for all values of s such that < 0. Hence
(5.4) provides the analytic continuation (not only for 1 < < 0) all
over the s-plane.
The term-wise integration preceding (5.4) is certainly justified over
any finite range since the concerned series is boundedly convergent. We
have therefore only to prove that
X
1
sin 2nx
dx = 0, 1 < < 0
lim
X
n
x s+1
n=1
X
Now
Z
X
"
#
Z
cos 2nx
s+1
sin 2nx
2nx
dx =
dx
s+1
s+1
2n
x
2nx
x s+2
X
X

!
Z
1
dx
1
=O
+
O
n
nX +1
x+2
X
!
1
=O
nX +1
96
and hence
X
1
sin 2nx
dx = 0, if 1 < < 0
lim
X
n
x s+1
n=1
X
This completes the analytic continuation and the proof of the Functional equation by the second method.
As a consequence of (5.1), we get
(
lim (s)
s1
) Z
[x] x + 12
1
1
dx +
=
2
s1
2
x
1
= lim
Zn
[x] x
dx + 1
x2
m+1
n1
dx
log
n
+
1
= lim
m
n
x
m=1 m
n1
X 1
= lim
+
1
log
n
n
m+1
m=1
102
Hence, near s = 1, we have

(s) =
1
+ + O(|s 1|)
s1
Lecture 12

Riemann (Contd)
6 Some estimates for (s) [11, p.27]
103
In any fixed half-plane 1 + > 1, we know that (s) is bounded,

since
|(s)| () (1 + ).
We shall now use the formulae of 5 to estimate |(s)| for large t, where
s = + it.
Theorem.
|(s)| < A log t,
|(s)| < A()t1 ,
1,
t 2.
t 1, if 0 < < 1.
Proof. From (5.1) we get

s
(s) =
s
s1
x [x]
dx,
x s+1
97
>0
98
Also from 5 of the 11th Lecture,

ZX
X 1
[X]
[x]
dx + s , if X 1, s , 1
=s
s
s+1
n
X
x
nX
1
If > 0, t 1, X 1, we get
Z
X 1
x [x]
1
X [X]
=
s
(s)
dx +
+
s
s+1
s1
n
Xs
x
(s 1)X
nX
X
Hence
Z
X 1
1
dx
1
+ 1 + + |s|
|(s)| <
n
X
tX
x+1
nX
X

X 1
1
t 1
1
+
+
1
+
, since |s| < + t.
<
+
n tx1 X
X
nX
104
If 1,
|(s)| <
X1
nX
1 1 1+t
+ +
t X
X
(log X + 1) + 3 +
t
, since t 1,
X
Taking X = t, we get the first result.

If , where 0 < < 1,
!
X 1
1 1
1
+ 2+
|(s)| <
+
n tX 1
x
nX
<
Z[X]
3t
dx X 1
+
+
x
t
X
3t
X 1
+ X 1 +
1
X
X 1.
7. Functional Equation (Third Method)
99
Taking X = t, we deduce
|(s)| < t
!
3
1
+1+ ,
1
t 1.
7 Functional Equation (Third Method) [16, p.21]

The third method is based on the theta-relation
!
1
1
(x) =
,
x x
where
(x) =
en x , Re x > 0
n=
This relation can be proved directly, or obtained as a special case of 105

Poissons formula. We shall here give a proof of that formula [2, p.37]
Let f (x) be continuous in (, ). Let
() =
2ix
f (x)e
dx lim
ZT
f (x)e2ix dx,
if the integral exists. Then, for and pro-positive integers, we have

1
p+ 2
Z
p 12
Z2
p
2ix
X
2ix
e
dx.
f
(x
+
k)
f (x)e
dx =
12
If we assume that
lim
p
X
p
f (x + k) = g(x)
(7.1)
100
uniformly in 12 x 21 , then g(x) is continuous, and on letting p

in (7.1), we get
1
Z2
() =
g(x)e2ix dx.
21
Thus () is the Fourier coefficient of the continuous periodic function

g(x). Hence, by Fejers theorem,
X
g(o) =
()
(C,1)
If we assume, however, that
Hence
lim
106
or
() < , then
p
X
f (k) = lim
f (k) =
p
X
(C,1)
() =
()
()
(7.2)
()
This is Poissons formula. We may summarize our result as follows.

Poissons Formula. If
(i) f (x) is continuous in (, ),
(ii)
(iii)
f (x + k) converges uniformly in 21 x <
1
2
and
() converges,
then
f (k) =
()
Remarks . This result can be improved by a relaxation of the hypotheses.
101
2
The Theta Relation. If we take f (x) = ex t , t > 0, in the above

discussion, we obtain the required relation. However, we have to show
that for this function the conditions of Poissons formula are satisfied.
This is easily done. f (x) is obviously continuous in (, ). Secondly
2
e(x+k) t e|k|t for |k| k0 (t)

This proves the uniform convergence of
e(x+k) t ,
k=
t > 0.
1
1
x<
2
2
Thirdly we show that, for t > 0

()
2 t+2yi
ey
For
() = e
2
t
dy =
1
t1/2
2 /t
107

2
et y i dy, and
t
If 0, we consider the contour integral

Z
2
ets ds, s = + i
C
taken over C as indicated. Then on the vertical lines, we have
102
2 2 )
et Re(+i) = et(
t(2 2 )
Letting , we thus see that the integrals along the vertical lines
vanish. Hence we have, for all real ,
Z
t2
d =
et( t i) d
Thus
2 /t
() = e
et d
2 /t
=e
2
e /t
eu du
e /t 1
dv =
t
=
108
2
e /t
ez z 2 dz
1 1
( ) =
2
2
e /t
Hence we have the desired relation:
k2 t
1 X k2 /t
=
e
, for t > 0,
t
(7.3)
or, writing
(t) =
ek t , we get
!
)
(
1
1
2(t) + 1 = 2
+1
t
t
Remark. (7.3) holds for t complex, with Re(t) > 0.
(7.4)
103
Functional Equation (Third method).

We have, for > 0,
s
ex x s/21 dx.
If > 0, n > 0, we then have on writing n2 x for x,
s
2
en x (n2 x) 2 1 n2 dx
s/2 s
en
2 dx
x s/21 dx
or
1
s/2
=
n s (s/2)
en x x s/21 dx.
Now if > 1, we have
109
X
2
s/2 X
1
=
(s) =
en x x s/21 dx
s
n
(s/2)
n=1
1
0
s/2
(s/2)
Z X
s/21
n2 x
x
dx
e
0
the inversion being justified by absolute convergence. Hence, for > 1,

we have
Z
s/2
(s) =
(x)x s/21 dx.
(s/2)
0
Using (7.4) we rewrite this as

s/2
(s/2)(s) =
Z1
0
s/21
(x)dx +
Z
1
(x)x s/21 dx
104
Z1
s/21
110
!
)
Z
1
1
1
1

+
dx + (x)x s/21 dx
2 x 2
x x
1
1
1
+
s1 s
1
+
s(s 1)
Z1
0
Z
1
x s/23/2
1
dx +
x

1
xs/2 2 + x s/21 (x)dx,
(7.5)
for > 1.
Now the integral on the right converges uniformly in a s b, for
if x 1, we have
s/21/2

x
+ x s/21 xb/21 + xa/21/2
while
(x) <
enx =
1
,
1
ex
and hence the integral on the right hand side of (7.5) is an entire function. Hence (7.5) provides the analytic continuation to the left of = 1.
It also yields the functional equation directly. We have also deduced that
s/2 (s/2)(s)
1
s(s 1)
is an entire function; so is s/2 {(s/2)}1 . Hence

(s)
1
s/2
1
s/2
= (s)
s(s 1) (s/2)
s 1 2(s/2 + 1)
is an entire function. But

entire function.
1/2
1
= 1. Hence (s)
is an
2(1/2 + 1)
s1
Lecture 13

Riemann (Contd)
8 The zeros of (s)
111
In the 11th Lecture, (p.98) we defined

(s)
1
1
s(s 1)s/2 (s/2)(s) s(s 1)(s).
2
2
The following theorem about (s) is a consequence of the results which

we have already proved.
Theorem 1. [11, p.45]
(i) (s) is an entire function, and
(s) = (1 s);
(ii) (s) is real on t = 0 and = 12 ;
(iii) (0) = (1) = 12 .
Proof. That (s) = (1 s) is immediate from the functional equation
of (s) (cf. p.97, Lecture 11) Clearly (s) is regular for > 0, and since
(s) = (1 s) it is also regular for < 1, and hence it is entire.
105
106
This proves (i)

(s) is obviously real for real s. Hence by a general theorem it takes
conjugate values at conjugate points. Thus ( 21 + ti) and ( 12 ti) are
conjugate; they are equal by the functional equation. So they are real.
This proves (ii).
To prove (iii) we observe that
s/2
(s) =
s

+ 1 (s 1)(s),
so that
1
(1) = 1/2 (3/2) 1 = ,
2
and by the functional equation, (0) = 21 .
112
We shall next examine the location of the zeros of (s) and of (s).
Theorem 2. [11, p.48] (i) The zeros of (s) (if any !) are all situated in
the strip 0 1, and lie symmetrical about the lines t = 0 and = 21
(ii) The zeros of (s) are identical (in position and order of multiplicity) with those of (s), except that (s) has a simple zero at each of
the points s = 2, 4. 6, . . .
(iii) (s) has no zeros on the real axis.
Proof. (i) (s) = (s 1)s/2
s
2

+ 1 (s)
h(s)(s), say.
Now (s) , 0; for > 1; (Euler Product!) also h(s) , 0, for
> 1; hence (s) , 0, for > 1; hence (s) , 0, for < 0; since
(s) = (1 s).
8. The zeros of (s)
107
The zeros, if any, are symmetrical about the real axis since ( ti)
are conjugates, and about the point s = 21 , since (s) = (1 s); they are
therefore also symmetrical about the line = 12 .
(ii) The zeros of (s) differ from those of (s) only in so far as h(s)
has zeros or poles. The only zero of h(s) is at s = 1. But this is not a
zero of (s) since (1) = 12 , nor of (s) for it is a pole of the latter.
The poles of h(s) are simple ones at s = 2, 4, 6, . . .. Since these
are points at which (s) is regular and not zero, they must be simple zeros 113
of (s). We have already seen this by a different argument on p.95.
(iii) Since (s) , 0 for < 0 and > 1, it is enough, in view of (ii),
to show that
() , 0, for 0 < < 1.
Now
(1 21s )(s) = (1 2s ) + (3s 4s ) + , for > 0.
To prove this we first notice that the relation is obvious for > 1,
and secondly each side is regular for > 0; the left side is obviously
regular for > 0 (in fact it is entire); and, as for the right side, we have

Z2n
1
dx
1

s
=
(2n 1) s (2n) s
s+1
x

2n1
108
|s|
<
+1
(2n 1)
(2n 1)+1
if > and |s| < for any fixed and

But, if 0 < < 1, the above relation gives
(1 21 )() > 0, or () < 0.
This establishes (iii).
114
Remarks . The strip 0 1 is called the critical strip; the line

= 12 the critical line; the zeros at 2, 4, 6, . . . are the trivial zeros
of (s). We still have to show that (s) actually has zeros, i.e. (s) has
non-trivial zeros. This is done by an appeal to the theory of entire
functions gives in the first few lectures.
Theorem 3. [11, p.56] If M(r) max |(s)|, then
|s|=r
1
log M(r) r log r, as r
2
Proof. For 2, we have
|(s)| (2),
and for 12 , |t| 1, we have
|(s)| < c1 |t|1/2 ,
so that
1
|(s)| < c2 |s|1/2 , for , |s| > 3.
2
Applying Stirlingls formula to (s/2), we get, for
1
, |s| = r > 3,
2
1
1
|(s)| < e| 2 s log 2 s|+c3 |s|
1
< e 2 r log r+c4 r ,
8. The zeros of (s)
109
s
since log = log |s| log 2 + i arg s, | arg s| <
2
(s) = (1 s), we infer that
2.
From the equation
|(s)| < e 2 |1s| log |1s|+c4 |1s|

1
< e 2 r log r+c5 r
for 12 , |s| = r > 4. Combining the results we get

1
M(r) < e 2 r log r+c6 r , r > 4.

On the other hand, if r > 2,
115
1
1
r e 2 r log rc7 r
2
M(r) (r) > 1. 2 r

Thus we have, for r > 4,
1
1
r log r C7 r < log M(r) < r log r + c6 r,
2
2
and hence the result.
Theorem 4. [11, p.57]

(i) (s) has an infinity of zeros;
(ii) If they are denoted by , then
verges for 1;
|| converges for > 1, di-
!
)
s s/
1 e
,
(iii)
b0 +b1 s
(s) = e
where b0 , b1 are constants;

!
P 1
1
(s)
= b1 +
+
(iv)
(s)
s
!
P 1
s
(s)
1
1
1
(v)
=b
+1 +
+ ,
(s)
s1 2 2
110
where b = b1 + 21 log 2.
Proof. (0) = 12 , 0, and we apply the theory of entire functions as

developed in the earlier lectures. By Theorem 3, (s) is of order 1, and
the relation log M(r) = O(r) does not hold. Hence 1 = 1, (s) has
P 1
diverges (by a previous theorem). This
an infinity of zeros and
||
proves (i), (ii) and (iii). (iv) follows by logarithmic differentiation. (v)
is obtained from (s) = (s 1)s/2 ( 2s + 1)(s).
We shall now show that (s) has no zeros on the line = 1.

Theorem 5. [11, p.28] (1 + it) , 0.
116
First Proof. 2(1 + cos t)2 = 3 + 4 cos + cos 2 0 for real . Since
XX 1
log (s) =
, for > 1,
mpms
m p
we have
log |(s)| = Re
=
cn nti
n=2
cn n cos(t log n),
n=2
1/m if n is the mth power of a prime

where cn =
0
otherwise.
Hence

log 3 () ( + ti)( + 2ti)
X

=
cn n 3 + 4 cos(t log n) + cos(2t log n)
0,
because cn 0 and we have the trigonometric identity above.

Thus

4
1
3 ( + ti)
|( + 2ti)|
,
{( 1)()}

1
1
8. The zeros of (s)
111
if > 1.
Now suppose 1 + ti(t 0) were a zero of (s); then, on letting
1 + 0, we see that the left hand side in the above inequality would
tend to a finite limit, viz. | (1 + ti)|4 |(1 + 2ti)|, while the right hand side 117
tends to . Hence (1 + it) , 0.
Second Proof. [11, p.89] We know from the 8th Lecture (1.12) that if
is real, and , 0, then
f (s)
2 (s)(s i)(s + i) X |i (n)|2

=
, > 1
s
(2s)
n
n=1
(8.1)
Let 0 be the abscissa of convergence of the series on the right hand

side. Then 0 1. The sum-function of the series is then regular in
the half-plane > 0 , and hence, by analytic continuation, (8.1) can be
upheld for > 0 . And since the coefficients of the series are positive,
the point s = 0 is a singularity of f (s).
Now, if 1 + i is a zero of (s), then so is (1 i), and these two
zeros cancel the double pole of 2 (s) at s = 1. Hence f (s) is regular
on the real axis, as far to the left as s = 1, where (2s) = 0. Hence
0 = 1. This is, however, impossible since (8.1) gives
!
1
|i (1)|2 = 1,
f
2
while f ( 21 ) = 0. Thus (1 + i) , 0.
Remarks. (i) We shall show later on that this fact ((1 + it) , 0) is
equivalent to the Prime Number Theorem.
(ii) If = + i is a zero of (s), then we have seen that 0 1,
and since (1 + it) , 0, we actually have 0 < 1, and by
symmetry about the line = 21 , we have 0 < < 1.
Lecture 14

Riemann (Contd)
9 Riemann-Von Magoldt Formula [11, p.68]
Let N(T ) denote the number of zeros (necessarily finite) of (s) in 0 118
1, 0 t T . Then, by what we have proved, N(T ) is the number
of zeros = + i of (s), or of (s), for which 0 < T
Theorem 1. N(T ) =
T
T
T
log
+ O(log T ) as T .
2n
2 2
Proof. Suppose first that T is not equal to any . Also let T > 3. Consider the rectangle R as indicated. (s) has 2N(T ) zeros inside it, and
none on the boundary. Hence, by the principle of the argument,
2N(T ) =
1
[arg (s)]R ,
2
where [arg (s)]R denotes the increase in arg (s), as s describes the
boundary of R.
113
114
Now
1
[arg (s)]R = [arg s(s 1)]R + [arg (s)]R ,
2
where (s) = s/2 (s/2)(s). Here
1
[arg s(s 1)]R = 4,
2
119
and (s) has the properties: (s) = (1 s), and ( ti) are conjugates.
Hence
[arg (s)]R = 4[arg (s)]ABC
Hence
N(T ) = + arg[s/2 ] + [arg (s/2)]ABC
+ [arg (s)]ABC
(9.1)
Now

t
T
= log .
[arg s/2 ]ABC = log
2
2
ABC
(9.2)
Next, since (z+) = (z+ 12 ) log zz+ 12 log 2+O(|z|1 ) as |z| ,

in any angle | arg z| < , we have
9. Riemann-Von Magoldt Formula
115
!#
"

1
= im log (s/2) ABC
arg
2 ABC
!
1 1
= im log + iT im log (1)
4 2
(
!
!
)
1 1
1
1
1
1
= im + iT log iT iT + log 2 + O(T )
4 2
2
2
2
!

1
1
T
1
= T log T + O T 1 , as T
(9.3)
2
2
8
2
Using (9.2) and (9.3) in (9.1), we get
T
T
T
7 1
1
N(T ) =
log
+ + [arg (s)]ABC + 0
2
2 2 8
T
(9.4)
Consider the broken line ABC exclusive of the end-points A and C.

If m is the number of distinct points s on ABC at which Re (s) = 0,
then
[arg (s)]ABC (m + 1),
(9.5)
since arg (s) cannot vary by more than (since Re does not change 120
sign here) on any one of the m + 1 segments into which ABC is divided
by the point s
Now no point s can be on AB, for
X
1
1
du 1
Re (2 + iT ) 1
>
1
=
2
2
2
4
n
2
u
2
(9.6)
Thus m is the number of distinct points of 12 < < 2 at which

Re ( + iT ) = 0, i.e. the number of distinct zeros of
1
g(s) = {(s + iT ) + (s iT )}
2
on the real axis for which 12 < < 2, because g() = Re ( + iT ) for
real , since ( iT ) are conjugate.
116
Now since g(s) is regular except for s = 1 iT , m must be finite,

and we can get an upper bound for m by using a theorem proved earlier.
3
7
Consider the two circles: |s 2| , |s 2| . Since T > 3, g(s)
4
2
is regular in the larger circle. At all points s of this circle, we have
1
,
4
1 < |t T | < 2 + T.
Hence, by an appeal to the other of (s) obtained in the 12th Lecture

(with = 14 ), we get
1
|g(s)| < c1 (|t + T |3/4 + |t T |3/4 )
2
< c1 (T + 2)3/4 ,
on the larger circle. At the centre, we have
g(2) = Re (2 + iT ) >
121
1
by (9.6)
4
Hence, by using the theorem (given in the Appendix), we get

!m
c1 (T + 2)3/4
7
< T, for T > T 0 3.
<
1
6
4
Thus
m < c2 log T, for T > T 0 .
Substituting this into (9.5) and (9.4) we get

N(T ) T log T + T c log T,
3

2
2 2
T > T0,
provided that T , for any . The last restriction may be removed

by first taking T larger than T and distinct from the s and then letting
T T + 0.

Remarks . The above theorem was stated by Riemann but proved by
Von Mangoldt in 1894. The proof given here is due to Backlund (1914).
117
Corollary 1. If h > 0 (fixed), then

N(T + h) N(T ) = O(log T ), as T
Proof. If we write
f (t) =
t
t
t
log
,
2
2 2
we get
f (t + h) f (t) = h f (t + h), 0 < < 1,
where f (t) =
t
1
log . Now use the Theorem.
2
2
Corollary 2.
!
X 1
X 1
log T
2
=O
= O(log T ); S
S
T
2
>T
0T
(summed over all zeros whose imaginary parts satisfy the given conditions).
Let
122
X1
, m < m + 1, and
sm =
X 1
, m < m + 1.
sm =
2
Then
S
[T ]
X
sm ,
m=0
sm
m=[T ]
If m 1, the number of terms in sm (or sm ) is

N(m + 1) N(m) = m , say.
By Corollary 1, m = O(log m) as m ; and
sm
m
,
m
sm
m
m2
118
Hence, as T ,
[T ]
X log m
= O(log2 T )
S = O(1) + O
m
2
!
X log m
log T
= O
S = O
T
m2
[T ]
Corollary 3. If the zeros for which > 0 are arranged as a sequence

n = n + in , so that n+1 n , then
|n | n
2n
, as n
log n
Proof. Since N(n 1) < n N(n + 1), and

2N(n 1) (n 1) log(n 1)
n log n ,
123
we have first
2n n log n
whence
log n log n ,
and
2n
2n
log n log n
n |n | < n + 1.
n
And
Remarks. (i) The main theorem proves incidentally the existence of

an infinity of complex zeros of (s), without the use of the theory
of entire functions.
(ii) We see from corollary 3 that
X
1
||(log ||)
converges for > 2 and diverges for 2.
119
(iii) Corollary 1 may be obtained directly by applying the theorem

(on the number of zeros) to (s) and to two circles with centre at
c + iT and passing through 12 + (T + 2h)i, 12 + (T + h)i respectively,
where c = c(h) is sufficiently large, and using the symmetry about
= 1/2
(iv) If N0 (T ) denotes the number of complex zeros of (s) with real
part 21 and imaginary part between 0 and T , then Hardy and Littlewood proved that N0 (T ) > cT in comparison with N(T )
1
T log A. Selberg has shown, however, that N0 (T ) > cT log T .
2
Appendix
Theorem. Let f (z) be regular in |z z0 | R and have n zeros (at least) 124
in |z z0 | r < R. Then, if f (z0 ) , 0.
R n
M
,
r
| f (z0 )|
where M = max | f (z)|

|zz0 |=R
Proof. Multiple zeros are counted according to multiplicity. Suppose

z0 = 0 (for otherwise put z = z0 + z ). Let a1 , . . . an be the zeros of f (z)
in |z| r.
Then
n
Y
R(z a )
f (z) = (z)
R2 z
=1
where a , a are conjugates and is regular in |z| R. On |z| = R, each

factor has modulus 1; hence
|(z)| = | f (z)| M on |z| = R
Since is regular in |z| R, we have
|(0)| M
Hence
result.
| f (0)| = |(0)|
n
n |a |
Q
M Rr and f (0) , 0, hence the
=1 R
120
Theorem. If f is analytic inside and on C and N is the number of zero

inside C, then
Z
1
f (z)
dz = N
2i
f (z)
C
= c log{ f (z)}
so that
N=
1
c arg f (z).
2
Lecture 15

Riemann (Contd)
10 Hardys Theorem [16, p.214]
We have proved that (s) has an infinity of complex zeros in o < < 1. 125
Riemann conjectured that all these zeros lie on the line = 21 . Though
this conjecture is yet to be proved, Hardy showed in 1914 that an infinity
of these zeros do lie on the critical line. We shall prove this result.
In obtaining the functional equation by the third method, we have
used the formula:
Z
s
1
s/2
(s/2)(s) =
+ (x)(x 2 1 + xs/21/2 )dx,
s(s 1)
1
where (x) =
x=
1
2
en
2 x
, x > 0. Multiplying by
n=1
1
s(s 1), and writing
2
+ it, we get
!
!
! Z
1
1
1
1
2
3/4
+ it = (t) = t +
(x)x
cos t log x dx
2
2
4
2
1
It is a question of showing that (t), which is an even, entire function

of t, real for real t, has an infinity of real zeros.
121
122
Writing x = e4u , we get
1
1
(t) = 4 t2 +
2
4
! Z
(04u )eu cos(2ut)du.
If we write
(u) = (e4u )eu ,
126
we obtain
1
= (t2 + 1)
2
2
t
(u) cos ut du
We wish now to get rid of the factor

+ 1) and the additive constant 12
so as to obtain (t/2) as a cosine transform of . This we can achieve
by two integrations (by parts). However, to get rid of the extra term
arising out of partial integration we need to know (0). We shall show
that (0) = 21 . For, by the functional equation of the theta-function,
!#
"
1
1
1 + 2(x) = 1 + 2
x
x
(t2
or
1 + 2(e4u ) = e2u [1 + 2(e4u )]
or
eu + 2eu (e4u ) = eu + 2eu (e4u )
or
so that
eu + 2(u) = eu + 2(u)

d u
e + 2(u) u=0 = 0
du
i.e.
1
(0) = .
2
We also know that (x) = O(ex ) as x . Hence

t
Z
0
1
(u) cos ut du =
2
Z
0
(u) cos ut du
10. Hardys Theorem

127
123
Hence
(t/2) =
[ (u) (u)] cos ut du
(10.1)
By actual computation it may be verified that

(t) (t) = 4[6e5t (e4t ) + 4e9t (e4t )]
Again if we write
1
1
(u) = eu (e4u ) + eu = (u) + eu
2
2
1 u 4u
= e (e ),
2
then by the theta relation, we have
(u) = (u), and (u) (u) = (u) (u)
Let us now write (10.1) in the form
(t) = 2
f (u) cos ut du
(t) cos ut dt,
We wish to deduce that

1
f (u) =
(10.2)
by Fouriers integral theorem. We need to establish this result.

Fourier Inversion Formula.[5, p.10] If f (x) is absolutely integrable in
(, ), then
Z
()
f (x)eix dx
124
exists for every real , < < , is continuous and bounded in

(, ). We wish to find out when
1
f (x) =
2
()eix d
128
for a given x. The conditions will bear on the behaviour of f in a neighbourhood of x.

If
1
S R (x) =
2i
ZR
eix ()d
sin Rt
f (x + t) dt
t
sin Rt f (x + t) + f (x t)
dt,
t
2
and
g x (t) =
then

1
f (x + t) + f (x t) f (x),
2
Z
sin Rt
g x (t) dt
t
0

Z Z
2
= + = I1 + I2 , say,
2
S R (x) f (x) =
for a fixed > 0.

g x (t)
I2 0 as R , by the Riemann-Lebesgue Lemma. So, if
t
is absolutely integrable in (0, ), then I1 0 as 0. Hence we have
the
10. Hardys Theorem
125
Theorem . [5, p.10] If

lim S R (x) = f (x).
g x (t)
is absolutely integrable in (0, ), then
t
If, in particular, f is absolutely integrable in (, ) and differen- 129

tiable in (, ) then lim S R (x) = f (x) for every x in (, ). It
R
should be noted that sufficient conditions for the validity of the theorem are: (i) f (x) is of bounded variation in a neighbourhood of x,
and (ii) f (x) is absolutely integrable in (, ). Then lim S R (x) =
R
1
2 [ f (x
+ 0) + f (x 0)].
We now return to (10.2) which is an immediate deduction from the
formula for (t), on using Fouriers inversion formula.
Now we know that (t) = ( 12 + it) = O(t1 e 4 t ); and (t) is an entire function of t. Hence we are justified in obtaining, by differentiation
of (10.2),
Z
(1)n
(2n)
(t)t2n cos ut dt.
f (u) =
Since (x) is regular for Re x > 0, f (u) is regular for /4 < Im u <
(Note that x = e2u ). Let
f (iu) = c0 + c1 u2 + c2 u4 +
Then
1
(1)n f (2n) (0)
=
cn =
(2n)!
(2n)!
(t)t2n dt
If (t) had only a finite number of zeros, then (t) would be of constant
sign for t > T . Let (t) > 0 for t > T . Then cn > 0 for n > 2n0 , since
Z
0
(t)t
2n
dt >
T +2
Z
(t)t
2n
dt
T +1
> (T + 1)
ZT
| (t)|t2n dt
2n
T +2
Z
T +1
(t)dt T
2n
ZT
0
| (t)| dt
126
> 0 for n > 2n0 , say.

130
Hence f (n) (iu) increases steadily with u for n > 2n0 . However we
i
along
shall se that f (u), and all its derivatives, tend to zero as u
4
the imaginary axis.
We know that
!
1 1 1
1
1/2
(x) = x
+ x 2
x
2
2
Further
(i + ) =
en
n=1
=2
i.e.
1
2
2 (i+)
e(2n)
(1)n en
n=1
en
= 2(4) ()
!
!
1
1
1 1
1
=

2
4
!
!
1
1
1
1
1
+ (i + ) =

2
4

+ (x) 0 as 0, and so do its derivatives (by a similar

i
argument). Hence 12 + (u) 0 as u
along the imaginary axis,
4
i
and so do its derivatives. Thus f (u) and all its derivatives 0 as u .
4
Thus
Lecture 16

Riemann (Contd)
Hamburgers Theorem.[13]
131
We have seen that the functional equation of the Gamma function,
together with its logarithmic convexity, determine the Gamma function
essentially uniquely (i.e. up to a constant). A corresponding result can
be proved for a the zeta function.
G(s)
, where G(s) is an entire function of finite
P(s)
order and P(s) is a polynomial, and let
Theorem 1. Let f (s) =
f (s) =
X
an
n=1
ns
the series on the right converging absolutely for > 1.

Let
!
s
1 s (1s)
s/2
= g(1 s) 2
f (s)
2
2 2
where
g(1 s) =
X
bn
,
n1s
n=1
127
(1.1)
(1.2)
(1.3)
128
The series on the right converging absolutely for < < 0. Then
f (s) = a1 (s)(= g(s)).
For the proof we need to evaluate two integrals:
y
1
=
2i
1+i
Z
ys (s)ds,
y>0
(1)
1i
a2 x bx
2ab
dx
=
e
,
a
x
a > 0,
b 0.
(2)
132
The first formula is a classic example of Mellins inversion formula

which can be obtained as an instance of Fouriers inversion formula discussed in the last lecture. We have seen that if f (x) L1 (, ), then
its Fourier transform () is defined in < < , and if in a neighbourhood of x0 , f (x) is differentiable, then
1
f (x0 ) =
2
ix0
F (s) =
f (y)eiy dy.
If we define
f (x)x s1 dx,
(3)
for s = c + it, c > 0, where f (x)xc1 L1 (0, ), so that the integral on

the right exists absolutely, we can look upon this as a Fourier transform
by a change of variable: x ey , for
F (c + it) =
f (ey )eyc eit y dy
Hence, provided that f (ey )eyc satisfies a suitable local condition, such
as differentiability in the neighbourhood of a point, we can invert this
129
relation and obtain

1
f (e ) e =
2
y
yc
or
f (ey ) =
or
f (x) =
1
2
F (c + it)eiyt dt
133
F (c + it)e(c+it)y dt
c+i
Z
1
2i
1
2i
F (s)eys ds
ci
c+i
Z
F (s)xs ds, c > 0.
(4)
ci
Relations (3) and (4) give the Mellin inversion formulae. Since
R
(s) = ex x s1 dx, Re s > 0, we get (1) as in immediate application.
0
We know that, for k > 0

ZA
e(ki)x dx =
so that
2

1
1 ekA eiA
k i
ekx cos xdx =
2.k
k 2 + 2
and by Fouriers inversion formula,

Z
cos x
ekx
d
=
,
2k
k 2 + 2
k>0
On the other hand, we have for k 0

Z
2
2
(1)
=
e(x +k )y dy,
x2 + k 2
0
(5)
130
134
so that
Z
0
Z
2
cos x
12 (k2 y+ 4y )
dx
=
dy
e
y
2
x2 + k 2
0
= e||k , from (5).

2k
Setting k2 = a2 , 2 = 4b2 , we get
Z
2ab
dx
=
e
,a > 0 b 0
a
x
a2 x bx
which proves formula (2).

Proof of Theorem 1. For any x > 0, we have by (1.1) and (1.2), (Here
f (s) = O(1))
1
S1
2i
2+i
Z
f (s)
(s) (s/2) s/2
x
ds
2
2i
2+i
Z
X
an
=
(s/2)(n2 x)s/2 ds
2i
n=1
2i
=2
an en
2x
n=1
We have, however, by (1.2), S 1 = S 2 , where

1
S2 =
2i
2+i
Z
2i
135
!
1 s (1s) s/2
ds.
2 x
g(1 s)
2
Now we wish to move the line of integration here from = 2 to =

1 .
By the hypothesis on the order of f (s) and (1.2), it follows that
there exist two numbers T > 0, > 0, such that for |t| T and 1
131
2, the function g(1 s) is regular and O(e|t| ). By (1.3), we see that

g(1 s) = O(1) on = 1 ; and, since f (s) = O(1) on = 2, while
(s/2)
= O(|t|3/2 ), on = 2,
( 1s
)
2
it follows that g(1 s) = O(|t|3/2 ) on = 2. Hence, by the principle of
Phragmen-Lindelof, we observe that
g(1 s) = O(|t|3/2 )
for |t| T , 1 2. Taking a suitable rectangle, and applying
Cauchys theorem, we get
1
S2 =
2i
1+i
Z
1i
!
m
X
1 s (1s) s/2
2
x
ds +
R
g(1 s)
2
v=1
where R is the residue of the integrand at the pole s lying in the

region 1 < < 2. The integrand, however, equals
136
132
f (s)
s
2
s/2 xs/2
The residue of this at a pole s of order q is

s
()
()
q 1
log
x
+
+
A
log
x
+
A
x 2 A()
q 1
1
0
Hence
m
X
R =
=1
m
X
xs /2 Q (log x) = Q(x), say, where
=1
Q is a polynomial
Here Re s 2 , > 0. (i.e. we are using the absolute convergence

P
of an ns only for > 2 ).
Hence
1 X bn
S2 =
x n=1 2i
2 X bn
=
x n=1 2i
1+i
Z
(1 s) n2
2
x
1i
2 +1+i
2 +1i
n2
(s)
x
!s
! (1s)
2
ds + Q(x)
ds + Q(x)
2
2 X
=
bn en /x + Q(x)
x n=1
137
Hence (since S 1 = S 2 ) we have

2
an en
2x
2
2 X
bn en /x + Q(x)
=
x n=1
Multiplying this by et x , t > 0, and integrating over (0, ) w.r.t. x, we

get
Z
X
X
2
bn 2nt
an
e
+
Q(x)et x dx
=2
2
2 + n2 )
t
(t
n=1
n=1
0
133
by (2).
The last integral is convergent. Since Re s 2 , = 1, 2, . . . m,
each term of Q(x) is O(x1+ 4 ) as x 0.

Now
Z
0
t2 x
Q(x)e
1
dx = 2
t
Z
0
x
t2
ex dx =
m
X
t s 2 H (log t),
where H is a polynomial, = H(t), say.

Hence
!
X
X
1
1
an
tH(t) = 2
+
bn e2nt
t
+
ni
t
ni
n=1
n=1
The series on the left hand side is uniformly convergent in any finite
region of the t-plane which excludes t = ki, k = 1, 2, . . . ; It is a meromorphic function with poles of the first order at t = ki.
H(t) is regular and single-valued in the t-plane with the negative real 138
axis deleted and t , 0.
The right hand side is a periodic function of t with period i for
Re t > 0. Hence, by analytical continuation, so is the function on the
left. Hence the residues at ki, (k + 1)i are equal. So ak = ak+1
Lecture 17

Riemann (Contd)
12 The Prime Number Theorem [11, pp.1-18]
The number of primes is infinite. For consider any finite set of primes; 139
let P denote their product, and let Q = P + 1. Then (P, Q) = 1, since any
common prime factor would divide Q P, which would be impossible
if (P, Q) , 1. But Q > 1, and so divisible by some prime. Hence there
exists at least one prime distinct from those occurring in P. If there were
only a finite number of primes, by taking P to be their product we would
arrive at a contradiction.
Actually we get a little more by this argument: if pn is the nth prime,
the integer Qn = p1 . . . pn + 1 is divisible by some pm with m > n, so
that
pn+1 pm Qn
n
from which we get, by induction, pn < 22 . For if this inequality was

true for n = 1, 2, . . . N, then
N
pN+1 p1 . . . pN + 1 < 22+4++2 + 1 < 22

and the inequality is known to be true for n = 1.
135
N+1
136
The fact that the number of primes is infinite is equivalent to the

P
statement that (x) =
1 as x . Actually far more is true;
px
x
n
. Again the relation pn < 22 is far
log x
weaker than the asymptotic formula: pn n log n, which is know to be
x
(The Prime Number Theorem!).
equivalent to the formula: (x)
log x
We shall prove the Prime Number Theorem, and show that it is equivalent to the non-vanishing of the zeta function on the line = 1. We
shall state a few preliminary results on primes.
!1
P1
1
Theorem 1. The series
and the product 1
are divergent.
p
p
!1
Q
P 1
1
, P(x) =
, x > 2.
1
Proof. Let S (x) =
p
px
px p
Then, for 0 < u < 1,
we shall show that (x)
140
Hence
1
1 um+1
>
= 1 + u + + um
1u
1u
P(x)
Y
(1 + p1 + + pm ),
px
P1
summed
n
over a certain set of positive integers n, and if m is so chosen that 2m+1 >
x, this set will certainly include all integers from 1 to [x]. Hence
where m is any positive integer. The product on the right is
P(x) >
[x]
X
1
1
141
>
we have
du
> log x.
u
Since
log(1 u) u <
[x]+1
Z
1 2
2u
1u
log P(x) S (x) <
X
px
, for 0 < u < 1,
p2
2(1 p1 )
(1)
12. The Prime Number Theorem
137
<
X
n=2
1
1
=
2n(n 1) 2
Hence, by (1),
1
S (x) > log log x .
2
(1) and (2) prove the theorem.
(2)
The Chebyschev Functions. Let

X
X
log p, (x) =
log p, x > 0.
(x) =
px
pm x
Grouping together terms of (x) for which m has the same value, we
get
(x) = (x) + (x1/2 ) + (x1/3 ) + ,
(3)
this series having only a finite number of non-zero terms, since (x) = 0
if x < 2.
Grouping together terms for which p has the same value ( x), we
obtain
X " log x #
log p,
(4)
(x) =
log p
px
since the number of values of m associated with a given p is equal to
the
of positive integers m satisfying m log p x, and this is 142
#
" number
log x

log p
Theorem 2. The functions
(x)
,
x/ log x
(x)
,
x
(x)
x
have the same limits of indetermination as x .

Proof. Let the upper limits (may be ) be L1 , L2 , L3 and the lower
limits be l1 , l2 , l3 , respectively. Then by (3) and (4),
X log x
log p = (x) log x.
(x) (x)
log p
px
138
Hence
L2 L3 L1 .
(5)
On the other hand, if 0 < < 1, x > 1, then

X
log p {(x) (x )} log x
(x)
x <px
and (x ) < x , so that

(x)
(x) log x log x
1
x
x
x
143
log x
Keep fixed, and let x ; then 1 0
x
Hence
L2 L1 ,
i.e.
L2 L1 , as 1 0.
This combined with (5) gives L2 = L3 = L1 , and similarly for the ls.
(x)
(x)
(x)
or
or
tends to a limit, then so do all,
x/ log x
x
x
and the limits are equal.
log p, if n is a +ve power of a prime p

Remarks. If (n) =
0, otherwise,
then
X
X
(x) =
log p =
(n), and
Corollary . If
(s)
(s)
pm x
Also
(s)
=s
(s)
nx
(n)
, > 1 [p.69]
ns
Z
1
(x)
dx, > 1.
x s+1
139
The Wiener-Ikehara theorem. [3, 4] Let A(u) be a non-negative nondecreasing function defined for 0 u < . Let
f (s)
A(u)eus du,
s = + i,
converge for > 1, and be analytic for 1 except at s = 1, where it 144

has a simple pole with residue 1. Then eu A(u) 1 as u
Proof. If > 1,
1
=
s1
e(s1)u du
Therefore
1
=
g(s) f (s)
s1
{A(u) eu }eus du
{B(u) 1}e(s1)u du,
where
B(u) A(u)eu
Take
s = 1 + + i.
Then
g(1 + + i) g () =
{B(u) 1}eu eiu du
Now g(s) is analytic for 1; therefore

g (t) g(1 + it) as 0.
uniformly in any finite interval 2 t 2.
140
We should like to form the Fourier transform of g (), but since we

do not know that it is bounded on the whole line, we shall introduce a
smoothing kernel. Thus
1
2
Z2
!
|| iy
e d
g () 1
2
Z2
1
=
2
iy
145
! Z
||
d {B(u) 1}euiu du.
1
2
0
For a fixed y, a finite -interval, and an infinite u-interval we wish to

change the order of integration. This is permitted if
Z
{B(u) 1}eu eiu du
converges uniformly in 2 2. This is so, because it is equal to

Z
u iu
B(u)e
du
eu eu du.
For a fixed > 0, the second converges absolutely and uniformly in .

In the first we have
B(u) e(/2)u = A(u) e(1+/2)u 0 as u
Hence B(u) = O(e(/2)u ), which implies the first integral converges uniformly and absolutely.
Thus
1
2
Z2
iy
!
||
g ()d
1
2
!
Z2
Z
||
1 i(yu)
= {B(u) 1}eu du
e
d
1
2
2
0
[B(u) 1]eu
141
sin2 (y u)
du
(y u)2
Now consider the limit as 0. The left hand side tends to

!
Z2
1
||
iy
g(1 + i)d
e 1
2
2
2
since g is analytic, or rather, since g () g(1 + i) uniformly. On the

right side, we have
Z 2
Z
2
sin (y u)
u sin (y u)
du
du.
e
2
(y u)
(y u)2
0
Hence
lim
Z
0
eu B(u)
sin2 (y u)
du =
(y u)2
Z
0
sin2 (y u)
du +
(y u)2
Z2
iyt
!
|t|
g(1 + it) dt
1
2
The integrand of the left hand side increases monotonically as 0; it

is positive. Hence by the monotone-convergence theorem,
Z
0
sin2 (y u)
B(u)
du =
(y u)2
Z
0
sin2 (y u)
du
(y u)2
+
Z2
iyt
!
|t|
1
g(1 + it) dt
2
Now let y ; then the second term on the right-hand side tends to
zero by the Riemann-Lebesgue lemma, while the first term is
lim
Zy
sin2
dv =
v2
sin2 v
dv =
v2
146
142
147
Hence
Zy
v sin2 v
B y
lim
dv = for every .
y
v2
To prove that
lim B(u) = 1.
Second Part.
(i) limB(u) 1.
For a fixed a such that 0 < a < y, we have

lim
Za

v sin2 v
dv
B y
v2
By definition B(u) = A(u)eu , where A(u) , so that

v
d
a
v
ey B y
ey B y
or

v
a
(va)/
B y
e
B y
.
Hence
Za
a (va)/ sin2 v
B y
lim
e
dv
y
v2
or
Za
2a
or
2a/

a
sin2 v
limB
y
dv
v2
Za
148
sin2 v
lim B(y) dv
v2 y
Now let a , in such a way that

Then
lim B(y)
y
a
0.
143
(ii) lim B(u) 1.

u
Za Za Zy
Zy
v sin2 v
dv =
+ +
B y
v2
()
We know that |B(u)| c, so that

a
Zy
Z 2
Z
Za
2
sin
v
sin
v
c
dv +
dv +
v2
v2
We know that |B(u)| c, so that
a
Z 2
Zy
Za
Z
2
sin v
sin v
dv +
dv +
c
v2
v2
As before we get

a
v
ey+a/ B y +
eyv/ B y
or

a 2a/
v
a a+v
B y
B y+
e B y+
e
Therefore
Za
Za

a 2a/ sin2 v
v sin2 v
dv
dv
B y+
e
B y
v2
v2
a
Za

a
sin2 v
2a/
e
B y+
dv
v2
a
Hence taking the lim in () we get, since
lim
Zy
= lim =
y
149
144
and
lim(c + (y)) c + lim(y),

that
a
Za
Z Z 2
sin v
c + 2 dv + lim

a 2a/
e
{ } + limB y +
{ } + e2a/
Za
limB(y)
Za
sin2 v
dv
v2
sin2 v
dv
v2
a
Let a , in such a way that 0.
Then
limB(y)
Thus
lim B(u) = lim A(u)eu = 1.
150
Proof of the Prime Number Theorem.

We have seen [p.143] that
(s)
=
s(s)
(x)
dx,
x s+1
est (et )dt,
>1
> 1.
Since (s) is analytic for 1 except for s = 1, where it has a

simple pole with residue 1, and has no zeros for 1, and (et ) 0
and non-decreasing, we can appeal to the Wiener-Ikehara theorem with
(s)
f (s) =
, and obtain
s(s)
(et ) et as t
13. Prime Number Theorem and the zeros of (s)
145
(x) x.
or
13 Prime Number Theorem and the zeros of (s)

We have seen that the Prime Number Theorem follows from the Wiener- 151
Ikehara Theorem if we assume that (1+it) , 0. On the other hand, if we
assume the Prime Number Theorem, it is easy to deduce that (1 + it) ,
0. If
X
X
(x) =
log p =
(n),
pm x
nx
then, for > 1, we have

Z
(s)
1
(x) x
dx
=
(s), say.
s+1
s(s) 1 s
x
1
Now (s) is regular for > 0 except (possibly) for simple poles at the
zeros of (s). Now, if (x) = x + O(x), [which is a consequence of the
p.n. theorem], then, given > 0,
|(x) x| < x, for x x0 () > 1.
Hence, for > 1,
|(s)| <
Zx0
|(x) x|
dx +
x2
Z
x0
dx < K +
x
1
where K = K(x0 ) = K(). Thus

|( 1)( + it)| < K( 1) + < 2,
for 1 < < 0 (, K) = 0 (). Hence, for any fixed t,
( 1)( + it) 0 as 1 + 0.
This shows that 1 + it cannot be a zero of (s), for in that case
( 1)( + it)
would tend to a limit different from zero, namely the residue of (s) at
the simple pole 1 + it.
146
14 Prime Number Theorem and the magnitude of

pn
152
It is easy to see that the p.n. theorem is equivalent to the result: pn

n log n. For if
(x) log x
1
(1)
x
then
log (x) + log log x log x 0,
hence
log (x)
1
log x
(2)
Now (1) and (2) give

(x) log (x)
1
x
Taking x = pn , we get pn n log n, since (pn ) = n.
Conversely, if n is defined by : pn x < pn+1 , then pn n log n
implies
pn+1 (n + 1) log(n + 1) n log n,
as x . Hence
or
x n log n
x y log y,
y = (x) = n.
Hence
log x log y, as above, so that
y x/ log x
Bibliography
[1] L.Bieberbach: Lehrbuch der Funktionentheorie I & II, Chelsea, 153
1945
[2] S.Bochner: Vorlesungen u ber Fouriersche Integrale, Leipzig, 1932
[3] S.Bochner: Notes on Fourier Analysis, Princeton, 1936-37
[4] S.Bochner: Math. Zeit. 37(1933), pp.1-9
[5] S.Bochner & K.Chandrasekharan: Fourier Transforms, Princeton,
1949
[6] A.P.Calderon & A.Zygmund: Contributions to Fourier Analysis,
Princeton, 1950, pp. 166-188
[7] C.Caratheodory: Funktionentheorie, I & II, Basel, 1950
[8] K.Chandrasekharan: J.Indian Math. Soc. 5(1941), pp. 128-132
[9] K.Chandrasekharan & S.Minakshisundaram:
Bombay, 1952
Typical Means,
[10] G.H.Hardy & M.Riesz: The General Theory of Dirichlet Series,

Cambridge, 1915
[11] A.E.Ingham: The Distribution of Prime Numbers, Cambridge,
1932
[12] J.E.Littlewood: The Theory of Functions, Oxford, 1945
147
148
BIBLIOGRAPHY
[13] C.L.Siegel: Math. Annalen, 86 (1922), pp.276-9

[14] C.L.Siegel: Lectures on Analytic Number Theory, New York,
1945
[15] E.C.Titchmarsh: The Theory of Functions, Oxford, 1932
[16] E.C.Titchmarsh: The Theory of the Riemann Zeta-Funcation, Oxford, 1951
[17] W.Thron: Introduction to the Theory of Functions of a Complex
Variable, Wiley, 1953

Lectures On The Riemann Zeta-Function: K. Chandrasekharan

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lectures On The Riemann Zeta-Function: K. Chandrasekharan

Uploaded by

Copyright:

Available Formats

Lectures on

The Riemann ZetaFunction

Tata Institute of Fundamental Research, Bombay

Lectures on the Riemann Zeta-Function

Tata Institute of Fundamental Research

2 The Phragmen-Lindelof principle

5 Entire Functions (Contd.)

6 The Gamma Function

7 The Gamma Function: Contd

8 The Zeta Function of Riemann

10 The Zeta Function of Riemann (Contd)

17 The Zeta Function of Riemann (Contd)

The Maximum Principle

(a) Let zo D, n a positive integer. Then

where lc = length of C, = distance of zo from C. As n

1. The Maximum Principle

(c) If | f (zo )| = M, and | f (zo )| = 0, then f (zo ) = 0, for

and letting n , we see that f (zo ) = 0. By a similar reasoning

For the proof of this theorem we need a

1. The Maximum Principle

Lemma. If f is regular in |z z0 | r, r > 0, then

where Mr = max | f (z)|, and | f (zo )| = Mr only if f (z) is constant for

Proof. On using Cauchys integral formula, we get

f (zo + rei )d.

On substituting this in (1), we get

Since is a continuous function of and cos 1, we get cos 1 for

1. The Maximum Principle

Aliter. If f (z0 + rei ) = () (complex valued) then

Now, if a and b are two complex numbers, |a| M, |b| M and

|(1 + t) (1 )| < /2, for 0 < t <

|(2 + t) (2 )| < /2, for 0 < t <

|(1 + t) + (2 + t)| < 2Mr , for 0 < t < .

1. The Maximum Principle

Proof of Theorem 2. Let zo D. Consider the set G1 of points z such

| f (z1 )| < M M = max | f (z)|

then | f (z)| < M for all z D. [17, p.111]

1. The Maximum Principle

Proof. D is an F , for define sets Cn by the property: Cn consists of

By Theorem 2, there exists a zn Bd Cn such that | f (zn )| = mn . The

mn < M for all n

N.B. That Bd D can be seen as follows. If D, then there

A(r) l.u.b Re{ f (z)}

are strictly increasing functions of r. Since f (z) is bounded if M(r)

1. The Maximum Principle

(3) If f is regular in D, and f , 0 in D, then f cannot attain a

Further if f is regular in D and continuous in D,

For, || is a boundary value of | f |, so that || M, hence

1. The Maximum Principle

Then, if f , 0 in D, by Corollary (6)a, we get

so that | f | > m = m in D. Thus if there exists a point zo D

(i) If f (x + iy) O as y uniformly in x, x ,

2. The Phragmen-Lindelof principle

(ii) If | f (x + iy)| 6 0, as y , consider the modified function

on the boundary of the strip, for a suitable constant c.

and apply Theorem 1 to f1 (z).

2. The Phragmen-Lindelof principle

Theorem 2. Suppose that

(ii) lim| f | M, for Bd D

e|z| eky cos

uniformly in x, since < k. Hence

2. The Phragmen-Lindelof principle

Let zo D. We can then choose H so that

for every point on the boundary of this rectangle. Hence, by the