Riemann Zeta (2k) Using Fourier Analysis
Riemann Zeta (2k) Using Fourier Analysis
Zeta(2k)
Introduction
where Z
2 2πnx
an = f (x)cos( )dx(n ≥ 0)
p I p
and Z
2 2πnx
an = f (x)sin( )dx(n ≥ 0)
p I p
where I is any interval of length p.
Bernoulli polynomials
Usual definition: Lets look at the case of the usual way how the bernoulli’s
polynomail is defined
∞
(text ) X tn
= B n (x)
(et − 1) n=0 n!
23
24 CHAPTER 7. ZETA(2K)
Let
1
pn (x) = Bn (x)
n!
then ∞
(text ) X
= pn (x)tn
(et − 1) n=0
Proposition 1:
po (x) = 1
pn (0) = pn (1)∀n ≥ 2
Proof:
∞
X
= p0n (x)tn
n=1
0
because p (x) = 0. Dividing both sides by t and using the defining
relation again, we obtain
∞
X ∞
X
n
pn (x)t = p0n (x)tn−1
n=0 n=1
∞
X
= p0n+1 (x)tn
n=0
3)
∞
X (tet )
pn (1)tn =
n=0
(et − 1)
(tet − 1) + t
=
(et − 1)
t
=t+ t
(e − 1)
X∞
=t+ pn (0)tn
n=0
po (x) = 1
Note also that it is obvious that properties (1)’, (2)’, and (3)’ define a
family of polynomials pn (x)n≥0 . One could therefore use these proper-
ties to define the pn (x).
Proposition 2:
That is ∀n ≥ 0
pn (1 − x) = (−1)n (pn (x))
Proof ∞
X (te(1−x)t )
pn (1 − x)tn =
n=0
(et − 1)
(tet e−xt )
= t
(e − 1)
(−ue−u eux )
=
(e−u − 1)
where u = -t
(ueux )
=
(eu − 1)
26 CHAPTER 7. ZETA(2K)
∞
X
= pn (x)un
n=0
∞
X
= (−1)n pn (x)tn
n=0
so fn is continuous on R.
We will compute the fourier series of pn (x) where n is even .More precisely
we extend pn (x) to a periodic function fn such that
fn (x) = pn (x)
if 0 ≤ x < 1 ,
fn (x + 1) = fn (x)∀x ∈ R
if
n 6= 1, fn (0) = pn (0) = pn (1) = fn (1)
therefore fn is continuous.
Lemma 3
Proof :
Let {a} = a − bac denote the fractional part of a real number a, and note
that {−a} = 1 - {a} if a ∈ R − Z. Then if x ∈ R − Z,
fn (−x) = pn ({−x})
= pn ({1 − x})
= pn ({x})
27
by Proposition 2
= fn (x)
Fourier Series
Z 1
2 pn (x)cos(2πmx)dx =
0
1 Z 1
2 2
[pn (x)sin(2πmx)] −
pn−1 (x)sin(2πmx)dx =
2πm 0 2πm 0
1 Z 1
−2 1 1
−[ [pn−1 (x)cox(2πmx)] + pn−2 (x)cos(2πmx)dx =
2πm 2πm 0 0 2πm
1
−1 2
Z
2( ) pn−2 (x)cos(2πmx)dx =
2πm 0
1
1 2
Z
2( ) pn−2 (x)cos(2πmx)dx
2πim 0
1 Z 1 !
1 1 1
= )[p2 (x)sin(2πmx)] − p1 (x)sin(2πmx)dx
(2πim)n−2 2πm 0 2πm 0
1 Z 1 !
1 −1 −1 1
= )[p1 (x)cos(2πmx)] + cos(2πmx)dx
(2πim)n−2 2πm 2πm 0 2πm 0
1 −1 −1
= +0
(2πim)n−2 2πm 2πm
1
=
(2πim)n
Thus
1
1 2
Z
p2 (x)cos(2πmx)dx = −( )
0 2πim
−2
an,m =
(2πim)n
Now an,m = 0 when n is even and when n ≥ 4 and m ≥ 1, and the equality
holds when n = 2 as well. The remaining case where n = 2 and m = 0 is
immediate from the fact that pn (0) = pn (1) when n ≥ 2; specifically, an,0 = 0.
Hence, when n ≥ 2 is even. By fourier convergence
∞
X
fn (x) = an,m cos(2πmx) =
m=1
∞
X −2 cos(2πmx)
m=1
(2πi)n mn
Hence now, as we have already remarked, fn is continuous when n 6= 1, so
we may evaluate the Fourier series at x = 0 to find that
29
∞
X −2
fn (0) =
m=1
(2πi m)n
as cos(0) = 1
∞
−2 X 1 −2
= ζ(n)
(2πi)n m=1 mn (2πi)n
Rearranging gives us
−1
ζ(n) = (2πi)n fn (0) =
2
−1
(2πi)n pn (0) =
2
−1 B2k
(2πi)2k =
2 2k!
(−1)k+1 (2π)2k B2k
2(2k!)
where k = n2 .
Hence Done
QED