Professional Documents
Culture Documents
David Jekel
June 6, 2013
1
Contents
1 Preliminaries 3
2
1 Preliminaries
Before we get to the zeta function itself, I will state, without proof, some
results which are important for the later discussion. I collect them here so
subsequent proofs will have less clutter. The impatient reader may skip this
section and refer back to it as needed.
In order to define the zeta function, we need the gamma function, which
extends the factorial function to a meromorphic function on C. Like Edwards
and Riemann, I will not use the now standard notation (s) where (n) =
(n 1)!, but instead I will call the function (s) and (n), will be n!. I give
a definition and some identities of , as listed in Edwards section 1.3.
Theorem 3. satisfies
1. (s) = 1 s
Q
n=1 (1 + s/n) (1 + 1/n)
2. (s) = s(s 1)
3. (s)(s) sin s = s
4. (s) = 2 (s/2)(s/2 1/2).
3
2.1 Motivation: The Dirichlet Series
Dirichlet defined (s) = s for Re(s) > 1. Riemann wanted a def-
P
n=1 n
inition valid for all s C, which would be equivalent to Dirichlets for
Re(s) > 1. He found a new formula for the Dirichlet series as follows. For
Re(s) > 1, by Eulers integral formula for (s) 2,
Z Z
nx s1 1 (s 1)
e x dx = s ex xs1 dx = .
0 n 0 ns
Summing over n and applying the formula for a geometric series gives
Z Z x s1 Z s1
X 1 X
nx s1 e x x
(s 1) s
= e x dx = x
dx = dx.
n 0 0 1e 0 ex 1
n=1 n=1
4
cancel as we would expect!) To state the definition quite precisely, the
integral over C is
2 +
es(log zi) dz es(log i+i)es(log z+i) dz
Z Z Z
+ i d + .
+ ez 1 z 0 eei 1
ez 1 z
R
The integrals converge by the same argument given above regarding 0 xs1 /(1
ex ) dx; in fact, they converge uniformly on compact subsets of C.
This definition appears to depend on , but actually does not matter
(so long as we avoid muitiples of 2i). Edwards does not mention this point,
but I will give a short proof for (0, 1). Let A be a curve which moves
in a line from 1 to , then in a semicircle above the real axis from to ,
then in a line from to 1. Let B move in a line from 1 to , then
in a semicircle below the real axis from to , then in a line from to 1.
Then, because the integrals from to 1 and 1 to cancel,
Z 1 Z +
(z)s dz
Z Z
(s) = + + +
+ A B 1 ez 1 z
The first and last integrals obviously do not depend on . In the integral
over A , (z)s is defined by the branch of the log function whose slit is on
the nonnegative imaginary axis, and in the integral over B , (z)s is defined
by the branch of the log function whose slit is on the nonpositive imaginary
axis. Since each branch of the log function is analytic on its domain, the
value of an integral depends only on the endpoints. Hence, the integrals
over A and B do not depend on .
2.3 Definition of
For Re(s) > 0, we can relate the formula for C (z)s /z(ez 1) dz to our
R
5
but /(e 1) 1 and Re(s)1 0. Hence,
!
Z + s(log z+i)
(z)s dz es(log zi) dz
Z Z
e dz
= lim +
C ez 1 z 0+ + ez 1 z ez 1 z
Z s1
z
= (eis eis ) dz.
0 ez 1
(z)s dz
Z
X 1 1
=
ns 2i(sin s)(s 1) C ez 1 z
n=1
6
3.1 First Proof
Fix s < 1. Let DN = {z : < |z| < (2N + 1)} \ (, (2N + 1)). We
consider Dn split along the positive real axis. Let CN be the same contour
as C above except that the starting and ending point is (2N + 1) rather
than +. Then
(z)s dz (z)s dz (z)s dz
Z Z Z
z
= z
z
.
DN e 1 z |z|=(2N +1) e 1 z CN e 1 z
(z)s dz (z)s
Z X
z
= 2i Res ,z .
DN e 1 z z(ez 1)
zDn
In fact, we can argue that this is still the case by spliting Dn into two pieces
and considering an analytic branch of the log function on each piece. This
argument is similar to the one made in the last section and is left to the
reader.
The residue at each point 2in for integer n 6= 0 is computed by inte-
grating over a circle of radius r < 2:
(z)s dz z 2in
Z Z
1 1 dz
z
= (z)s1 z .
2i |z2in|=r e 1 z 2i |z2in|=r e 1 z 2in
7
and |(z)s /z| (2N + 1)s1 , so the whole integral is less than a constant
times (2N + 1)s , which goes to zero as N because we assumed s < 0.
Therefore, taking N gives
(z)s dz
Z X
s1 s1 s1
z
= 2i(2) (i + (i) ) ns1 .
C e 1 z n=1
8
Both sides can be shown to be analytic, so this relation holds not only for
Re(s) > 1, but for all s. The right side is unchanged by the substitution of
1 s for s, so the theorem is proved again.
where ranges over the roots of . This type of product expansion is guar-
anteed for polynomials by the fundamental theorem of algebra, but not for
functions with possibly infinitely many roots. The convergence of the prod-
uct depends on the ordering of the terms and the frequency of the roots.
By the argument principle, {0 Re(s) 1, 0 Im(s) T } is equal to the
integral of 0 /2i around the boundary of the region (assuming there are
no roots on the boundary). Riemann estimates this integral as
T T
log 1
2 2
with an error on the order of T 1 . This estimate on the frequency of the roots
of would guarantee that the product expansion of converges, provided
each root is paired with its symmetric twin 1 . But no one else could
prove this estimate until 1905.
9
4.2 Proof by Hadamards Theorem
Hadamard proved that the product representation of was valid in 1893.
His methods gave rise to a more general result, the Hadamard factorization
theorem. Edwards devotes chapter 2 to Hadamards proof of the product
formula for . But for the sake of space I will simply quote Hadamards
theorem to justify that formula. The reader may consult Taylor chapter
8 for a succinct development of the general theorem, which I merely state
here, as given in [2]:
1 s(1 s)
Z
dx
(s) = (x)(xs/2 + x1/2s/2 ) ,
2 2 1 x
which Edwards derives in section 1.8. Using more integration by parts and
change of variables, we can show that
Z
d h 3/2 0 i 1/4
cosh 12 (s 21 ) log x dx.
(s) = 4 x (x) x
1 dx
Direct evaluation of (d/dx)[x3/2 0 (x)] will show that the integrand is posi-
tive. Hence, has a power series about 12 with real, positive coefficients.
This is all we need to derive a simple estimate of as Edwards does
in section 2.3:
10
Theorem 10. For sufficiently large R, |( 12 + w)| RR for |w| < R.
Proof. Since the power series of around 12 has positive coefficients, |( 12 +
w)| achieves its maximum value for |w| R at R itself. Choose N such that
1 1
2 +R 2N 2 +R+2. Since is entire, the power series expansion is valid
for all s, and the fact that is has positive coefficients shows is increasing
on [ 21 , ). Thus,
Since the left hand side and the product are both even functions, eh(w) must
be even, which implies it is constant. We undo the change of variables by
letting 21 + w = s and 12 + = . Thus, we have
Y s 1/2
(s) = c 1 .
1/2
11
Each term in the product is a linear function of s which is 0 when s = and
1 when s = 0. Hence, each term is 1 s/ and we have
Theorem 11. has a product expansion
Y s
(s) = (0) 1 ,
12
where the last equality holds because SN \ TN +1 = SN +1 and TN +1 SN .
This completes the induction.
The proof will be complete if we show that nSN ns approaches 1 as
P
N . Notice that if 1 < n N < PN , then n 6 SN . This is because any
prime factor of n has to be less than or equal to PN , but all multiples of all
primes pn PN have been removed from SN . Hence,
X 1 X 1 X 1 X 1
1 = ns ns .
ns ns
nSN N nSN N nSN n=N
We are justified in applying the triangle inequality here because the right-
most sum converges absolutely. Indeed, the sum on the right isPsmaller than
any positive for N sufficiently large. Hence, the limit of nSN ns is
1.
s
P
Taking the log of this formula gives
P n log (s) = p log(1 p ) for
Re(s) > 1. Since log(1 x) = n=1 x /n, we have
XX 1 sn
log (s) = p .
p
n
n=1
13
J(x) x and J(x) = 0 for x < 2, we know that for Re(s) > 1,
R Since s1
0 J(x)x dx converges. Hence, by Theorem 4, we can integrate term
by term and
Z
1 s1
Z
s1
XX 1 X X 1 sn
J(x)x dx = x dx = p ,
0 p
n pn s p n
n=1 n=1
Applying (s) = (s/2)(s 1) s/2 (s) together with the product for-
mula for gives:
s s
log (s) = log (s) + log log(s 1) log
2 2
X s s s
= log (0) + log 1 + log log(s 1) log .
2 2
He substitutes this into the integral for J(x), then evaluates the integral
term by term as
Z
X dt
J(x) = Li(x) [Li(x ) + Li(x1 )] + 2 1) log t
+ log (0),
x t(t
Im>0
Rx
where Li(x) is the Cauchy principal value of 0 1/ log x dx. The evaluation
requires too much work to explain it completely here (Edwards sections
1.13-1.16), but the term-by-term integration itself is even more difficult to
justify, and Riemanns main formula was not proved until Von Mangoldts
work in 1905.
14
7 Von Mangoldts Formula
Von Mangoldts formula is essentially the same as Riemanns, except that
instead of considering log (s), he considers its derivative 0 (s)/(s). By
differentiating the formula for log (s) as a sum over primes, we have
0 (s) XX
= pns log p, Re(s) > 1.
(s) p n=1
(In the sum over , we pair with 1 and sum in order of increasing Im().
In the integral from a i to a + i we take the Cauchy principal value.)
His proof, like Riemanns, depends on several tricky termwise integrations.
15
We want to integrate term by term. This is justifiable on a finite interval
because the series converges uniformly for Re(s) a > 1 (since (m)
log m and x is constant). Thus, we have
Z a+ih X ! Z a+ih s
1 xs ds 1 X x ds
(m) s = (m) s
.
2i aih m s 2i aih m s
m=1 m=1
To take the limit as h and evaluate the integral, we will need the
following lemma, which can be proved by straightforward estimates and
integration by parts.
Lemma 13. For t > 0,
1
Z a+ih 0
t<1
s ds 1
lim t = u(t 1) = , t=1
h+ 2i aih s 2
1, t > 1.
The right hand side is summable over m because (log m)/| log x log m|
approaches 1 as m and (x/m)a is summable. Thus, for fixed x,
Z a+ih s
X x ds K
(m)
s
u(x m) ,
aih m s h
m=1
for some constant K. Hence, the limit of the sum as h is zero. This
argument assumes x 6= m, but if x is an integer, we can simply throw away
the terms where m x, since finitely many terms do not affect convergence.
This implies
Z a+ih s
1 X x ds X
lim (m) s
= (m)u(x m),
h 2i aih m s
m=1 m=1
or Z a+i 0
1 (s) s ds
x = (x).
2i ai (s) s
This completes the first evaluation of the integral.
16
7.2 Second Evaluation of the Integral
For the second evaluation, we use a different formula for 0 /. Differentiate
X s s s
log (s) = log (0) + log 1 + log log(s 1) log
2 2
to obtain
0 (s) X 1 1 1 d s
= log + + log
(s)
s 2 s 1 ds 2
for any fixed s, for Im() sufficiently large. In fact, on the set {s : Re(s) >
1 + , a < Im(s) < a}, we can bound |1/( s)| uniformly by |1/( s0 )|
where s0 = 1 + + ai. This shows that the series converges uniformly on
compact sets and term-by-term differentiation is justified.
QIn the preceding formula, substitute the product formula for (s) =
1 (1 + 1/n)s from Theorem 3:
n=1 (1 + s/n)
d s d X 1 s
log = s log 1 + log 1 +
ds 2 ds 2n 2n
n=1
X 1 1
= log 1 + +
2n 2n + s
n=1
17
This is the formula for 0 / we intend to integrate term by term. To do this,
we need another lemma. The reader may try deducing this lemma from the
previous one using a change of variables and Cauchys theorem.
The lemma allows us immediately to integrate the first and last terms:
Z a+i 0
0 (0)
Z a+i
1 s s ds 1 (0) s ds
x = x, x = .
2i ai s 1 s 2i ai (0) s (0)
The lemma also tells us what the value of other integrals will be, if we can
justify term-by-term integration.
First, consider
Z a+i X !
1 s ds
xs .
2i ai 2n(s + 2n) s
n=1
Because the series converges uniformly on compact subsets of {Re(s) > 1},
we can integrate term by term over a finite interval. We only have to worry
about taking the limit
Z a+ih
X 1 1 xs
lim ds.
h 2i 2n aih s + 2n
n=1
To justify the termwise limit, we begin by showing that the sum of the
limits converges. By Lemma 14,
x2n
Z a+i
1 1 xs
ds = .
2i 2n ai s + 2n 2n
18
By change of variables,
a+ih a+2n+ih a+2nih
xs xs xs
Z Z Z
ds = x2n ds = 2x2n ds.
aih s + 2n a+2nih s a+2n s
2Kx2n K0
Z a+2nih s
2n 1 x
2x ds .
2i 0 s (a + 2n) log x n2
Hence, the sum of the partial integrals is summable. This implies that
evaluating the limit termwise is valid. This is because for any > 0, we
can choose N large enough that the sum of the first N infinite integrals is
within of the complete sum. By choosing N larger if necessary, we can
make the sum of the first N finite integrals within /3 of its limit, uniformly
in h. Then by choosing h large enough, we can make the first N integrals
be within /3N of their limits. Thus, the two partial sums are within /3
of each other, and so the infinite sum of the infinite integrals is within of
the infinite sum of the finite integrals. P
All that remains is to evaluate the integral of s/( s).
Lemma 16. For T sufficiently large, has fewer than 2 log T roots in T
Im() T + 1.
We want to evaluate
!
Z a+i
1 X s ds
xs .
2i ai
(s ) s
We first note that we can exchange summation and integration for a finite
integral:
X 1 Z a+ih xs
lim ds.
h
2i aih (s )
19
show that each term of the series is less than a constant times | 12 |2 .
The proof is merely a manipulation of fractions and is left to the reader.
We already know the limit exists because we have shown that all the
other terms in von Mangoldts formula have limits as h . Von Mangoldt
considers the diagonal limit
X x Z a+ih xs
lim ds.
h aih s
|Im()|h
He shows that
X x Z a+ih a+ih
xs x xs
X Z
ds ds
aih s aih s
|Im()|h
approaches zero, so that the diagonal limit exists and is equal to the original
limit. If is written as + i, then we can estimate the sum of roots with
positive as
X x 1 Z a+ih xs X x 1 Z a+i(+h) xt
ds ds ,
aih s
2i 2i a+i(h) t
>h >h
20
This sum converges and as h , it approaches zero; the proof of this is
left to the reader.
We now only have to evaluate the diagonal limit, and we know that it
exists. We will show that
X x Z a+ih xs X x
ds
aih s
|Im()|h |Im()|h
21
7.4 Von Mangoldts and Riemanns Formulae
We have now evaluated term by term
!
a+i
0 (0)
Z
1 s X s X s ds
+ xs ,
2i ai s1
(s ) 2n(s + 2n) (0) s
n=1
showing that
X x x2n 0 (0) a+i
0 (s) s ds
Z
X 1
x + = x = (x).
2n (0) 2i ai (s) s
n=1
The one thing missing from von Mangoldts formula is the evaluation of
0 (0)/(0) which Edwards performs in section 3.8. It is log 2.
Von Mangoldt used his formula to prove Riemanns main formula; Ed-
wards explains this proof in section 3.7. Von Mangoldt used a different
method from Riemanns and did not directly justify Riemanns term-by-
term integration, although three years later in 1908, Landau did just that.
Von Mangoldts formula can also be used to prove the prime number theo-
rem, which says essentially that (x)/x approaches 1 as x .
References
[1] Harold M. Edwards. Riemanns Zeta Function. 1974. Dover Publica-
tions.
22