Anda di halaman 1dari 20

1 Riemann Zeta function 1

1 Riemann Zeta function


On the real line with x > 1, the Riemann Zeta Function can be dened
by the integral
Z∞
1 ux− 1

ζ(x) = du (1)
Γ (x) 0
eu − 1
where Γ (x) is the Gamma Function.

If x = n ∈ Z, then we have the identity

un− 1
e−u un− 1

=
eu − 1 1−e
−u

X

−u n−1
= e u e−ku
k=0
X

= e−ku un− 1
, (2)
k=1
so
Z∞ X ∞ Z∞
un− 1

du = e−ku un− du. 1


(3)
0
eu − 1 k= 1
0

To evaluate ζ(n), let y = ku so that dy = k du and plug in the above

identity to obtain
∞ Z∞
X
1
ζ(n) = e−ku un− du 1

Γ (n) k= 1
0

∞ Z∞
X  y n− dy 1
1
= e−y
Γ (n) k= 1
0
k k
X
∞ Z∞
1 1
= e−y yn− dy. 1
(4)
Γ (n) k= kn 1
0

Integrating the nal expression in (4) gives Γ (n), which cancels the factor
1
and gives the most common form of the Riemann zeta function,
Γ (n)
X

1
ζ(n) = (5)
k=1
kn

@AOsorioC
2

which is sometimes known as a p−Series.

The Riemann zeta function can also be dened in terms of multiple integrals

by
Y
n

Z 1
Z 1
dxi
i=1
ζ(n) = ··· , (6)
Y
n
| {z } 1 −
0
xi
0

n
i=1
and as a Mellin transform by
Z∞
1 ζ(s)
ts− dt = −
1
(7)
0
t s
for 0 < <(s) < 1, where {x} is the fractional part (Balazard and Saias 2000).

It appears in the unit square integral


Z Z
1 1
− lns (xy)
dx dy = Γ (s + 2) ζ(s + 2),
0 0
1 − xy

valid for <[s] > 1 (Guillera and Sondow 2005).

For s∈N 0 , this formula is due to Hadjicostas (2002), and the special cases

s=0 and s=1 are due to Beukers (1979).

Note that the zeta function ζ(s) has a singularity at s = 1, where it reduces
to the divergent harmonic series.

The Riemann zeta function satises the reection functional equation


 
−s 1
ζ(1 − s) = 2 (2 π) cos Γ (s) ζ(s) (8)
2 πs
(Hardy 1999, p. 14; Krantz 1999, p. 160), a similar form of which was

conjectured by Euler for s ∈ R (Euler, read in 1749, published in 1768;

Función ζ
1 Riemann Zeta function 3

Ayoub 1974; Havil 2003, p. 193). A symmetrical form of this functional

equation is given by

s  
− 2s 1 −s 1−s
Γ π ζ(s) = Γ π− 2 ζ(1 − s) (9)
2 2

(Ayoub 1974), which was proved by Riemann for all s ∈ C (Riemann 1859).

As dened above, the zeta function ζ(s) with s = σ + it ∈ C is dened

for <[s] > 1. However, ζ(s) has a unique analytic continuation to the

entire complex plane, excluding the point s= 1, which corresponds to a

simple pole with complex residue 1 (Krantz 1999, p. 160). In particular,

as s −→ 1, ζ(s) obeys
ζ(s) − 1
lim = γ, (10)
s→1 s−1
where γ is the Euler-Mascheroni constant (Whittaker and Watson 1990, p.

271).

To perform the analytic continuation for <[s] > 0, write

X

(−1)n X

1
X

1
+ = 2

n=1
ns n=1
ns n=2, 4,...
ns
X∞
1
= 2

k=1
(2k)s
X

1
1 −s
= 2 , (11)
k=1
ks

so rewriting in terms of ζ(s) immediately gives

X

(−1)n 1 −s
+ ζ(s) = 2 ζ(s). (12)
n=1
ns

Therefore,
1
X

(−1)n− 1

ζ(s) = −s
. (13)
1 −2 1
n=1
ns

@AOsorioC
4

Here, the sum on the right-hand side is exactly the Dirichlet eta function

η(s) (sometimes also called the alternating zeta function). While this for-

mula denes ζ(s) for only the right half-plane <[s] > 0, equation (8) can

be used to analytically continue it to the rest of C. Analytic continuation

can also be performed using Hankel functions. A globally convergent series

for the Riemann zeta function (which provides the analytic continuation of

ζ(s) to the C \ {1}) is given by

X
∞ Xn  
1 1 k n
ζ(s) = −s n+1
(−1) (k + 1)−s (14)
1 −2 1
n=0
2
k= 0
k
 
n
(Havil 2003, p. 206), where is a binomial coecient, which was con-
k
jectured by Knopp around 1930, proved by Hasse (1930), and rediscovered

by Sondow (1994). This equation is related to renormalization and random

variates (Biane et al. 2001) and can be derived by applying Euler's series

transformation with n=0 to equation (13).

Hasse (1930) also proved the related globally (but more slowly) convergent

series
X
∞ Xn  
1 1 k n 1 −s
ζ(s) = (−1) (k + 1) (15)
s − 1 n= 0
n + 1 k= 0
k

that, unlike (14), can also be extended to a generalization of the Riemann

zeta function known as the Hurwitz zeta function ζ(s, a). ζ(s, a) is dened
such that

ζ(s) = ζ(s, 1). (16)

(If the singular term is excluded from the sum denition of ζ(s, a), then

ζ(s) = ζ(s, 0) as well.) Expanding ζ(s) about s=1 gives

1
X

(−1)n
ζ(s) = + γn (s − 1)n , (17)
s−1 n=0
n!

Función ζ
1 Riemann Zeta function 5

where γn are the so-called Stieltjes constants.

The Riemann zeta function can also be dened in C by the contour integral
I
Γ (1 − z) uz− 1

ζ(z) = du ∀ z 6= 1, (18)
2πi γ e−u − 1
where the contour is illustrated above (Havil 2003, pp. 193 and 249-252).

Zeros of ζ(s) come in (at least) two dierent types. So-called trivial zeros

occur at all negative even integers s = −2, −4, −6, . . ., and nontrivial

zeros at certain

s = σ + it (19)

for s in the critical strip 0 < σ < 1. The Riemann hypothesis asserts

that the nontrivial Riemann zeta function zeros of ζ(s) all have real part
1
σ = <[s] = , a line called the critical line. This is now known to be
2
9
true for the rst 250 × 10 roots.

The Riemann zeta function can be split up into


 
1
ζ + it = Z(t)e−iθ(t) , (20)
2

where Z(t) and θ(t) are the Riemann-Siegel functions. The Riemann zeta

function is related to the Dirichlet lambda function λ(ν) and Dirichlet eta

function η(ν) by
ζ(ν) λ(ν) η(ν)
ν
= ν
= ν (21)
2 2 −1 2 −2

and

ζ(ν) + η(ν) = 2λ(ν) (22)

(Spanier and Oldham 1987). It is related to the Liouville function λ(n) by

ζ(2s) X λ(n)

= (23)
ζ(s) n=
ns 1

@AOsorioC
6

(Lehman 1960, Hardy and Wright 1979). Furthermore,

ζ (s) X 2ω(n)
2

= , (24)
ζ(2s) n= ns 1

where ω(n) is the number of distinct prime factors of n (Hardy and Wright
1979, p. 254).

For −2n a positive even integer −2, −4, . . .,


(−1)n ζ(2n + 1)(2n)!
ζ0 (−2n) = n+ π n
2 1
,
2
2

giving the rst few as

0 ζ(3)
ζ (−2) = −

2

0 3ζ(5)
ζ (−4) =

4

0 45ζ(7)
ζ (−6) = −

6

0 315ζ(9)
ζ (−8) =

8

(Sloane's A117972 and A117973). For n = −1,


0 1
ζ (−1) = − ln(A),
12

where A is the Glaisher-Kinkelin constant.

Using equation (14) gives the derivative

0 1
ζ (0) = − ln 2( π),
2

which can be derived directly from the Wallis formula (Sondow 1994).

0
ζ (0)
= ln(2π)
ζ(0)

Función ζ
1 Riemann Zeta function 7

can also be derived directly from the Euler-Maclaurin summation formula

(Edwards 2001, pp. 134-135).

In general, ζ(n) (0) can be expressed analytically in terms of π, ζ(n), the

Euler-Mascheroni constant γ, and the Stieltjes constants γi , with the rst

few examples being

00 1 2
1 2
1 2
ζ ( 0) = γ + γ − 1 π − ln (2π)
2 24 2
000 3 1 3
1 2
3 2 3
ζ ( 0) = 3 ln 2 ( π)γ + 3γγ + γ − ζ(3) −
1 1 2 ln (2π) − π ( π) + γ
ln 2 ( π) + γ
ln 2 .
2 2 8 2

(25)
 
1
Derivatives ζn can also be given in closed form, for example,
2
    
0 1 1 n 1
ζ = π + 2γ + 6 ln(2) + 2 ln(π)ζ (26)
2 4 2

= −3.92264613 . . . (27)

(Sloane's A114875).

The derivative of the Riemann zeta function for <(s) > 1 is dened by

0
X

ln(k)
ζ (s) = − (28)
k=1
ks
X∞
(k)
ln
= − . (29)
k=2
ks
0
ζ ( 2) can be given in closed form as

0 1
ζ (2) = (γ + ln(2π) − 12 ln(A)) (30)
π 6
2

= −0.93754825431 . . . (31)

(Sloane's A073002), where A is the Glaisher-Kinkelin constant (given in

series form by Glaisher 1894).

@AOsorioC
8

0
The series for ζ (s) about s=1 is

0 1 1 2
ζ (s) = − − γ + γ (s − 1) − γ (s − 1) + · · ·
1 2 3 , (32)
(s − 1) 2
2

where γi are Stieltjes constants.

2n
In 1739, Euler found the rational coecients C in ζ(2n) = C π in terms

of the Bernoulli numbers. Which, when combined with the 1882 proof by

Lindemann that π is transcendental, eectively proves that ζ(2n) is trans-

cendental. The study of ζ(2n + 1) is signicantly more dicult. Apéry

(1979) nally proved ζ(3) to be irrational, but no similar results are known
for other odd n. As a result of Apéry's important discovery, ζ(3) is some-

times called Apéry's constant. Rivoal (2000) and Ball and Rivoal (2001)

proved that there are innitely many integers n such that ζ(2n + 1) is

irrational, and subsequently that at least one of ζ(5), ζ(7), . . . , ζ(21) is

irrational (Rivoal 2001). This result was subsequently tightened by Zu-

dilin (2001), who showed that at least one of ζ(5), ζ(7), ζ(9) or ζ(11) is

irrational.

A number of interesting sums for ζ(n), with n ∈ N, can be written in terms


of binomial coecients as the binomial sums

X

1
ζ(2) = 3   (33)
k
2
k=1 k 2

k
X∞
(−1)k−
5
1

ζ(3) =   (34)
2 2k
k= k 1 2

k
36
X∞
1
ζ(4) =   (35)
17 2k
k= k 1 4

k
(Guy 1994, p. 257; Bailey et al. 2007, p. 70). Apéry arrived at his result

Función ζ
1 Riemann Zeta function 9

with the aid of the k− 3


sum formula above. A relation of the form

X∞
(−1)k− 1

ζ(5) = Z  5 (36)
2k
k= k 1 5

has been searched for with Z 5 a rational or algebraic number, but if Z 5 is

a root of a polynomial of degree 25 or less, then the Euclidean norm of


383
the coecients must be larger than 1.24 × 10 , and if ζ(5) is algebraic
380
of degree 25 or less, then the norm of coecients must exceed 1.98 × 10
(Bailey et al. 2007, pp. 70-71, updating Bailey and Ploue). Therefore,

no such sums for ζ(n) are known for n > 5.

The identity

X

1
X

2 n
= ζ(2n + 2)x (37)
k=1
k −x
2 2
n=0
1 − πx cot(πx)
= (38)
2x
2

2
x 4
X

1
Y
k−1 1 −
= 3    m 2
(39)
2 2
k
2 x x
k=1 k2
1− m=1 1−
k k 2
m 2

 
3 1
34 F
3 1, 2, 1 − 2x, 1 + 2x; , 2 − x, 2 + x;
2 4
= (40)
2 (1 − x ) 2

for x ∈ C \ {0} is complex number not equal to a nonzero integer gives an

Apéry-like formula for even positive n (Bailey et al. 2006, pp. 72-77).

The Riemann zeta function ζ(2n) may be computed analytically for even n
using either contour integration or Parseval's theorem with the appropriate

Fourier series. An unexpected and important formula involving a product

@AOsorioC
10

over the primes was rst discovered by Euler in 1737,


  
−s 1 1 1
ζ(s) (1 − 2 ) = 1 + s
+ s
+ ··· 1 − s
2 3 2

(41)
 
1 1
= 1 + s
+ s
+ ···
2 3
 
1 1 1
− s
+ s
+ s
+ ··· (42)
2 4 6
 
−s −s 1 1 1
ζ(s) (1 − 2 ) (1 − 3 ) = 1 + s
+ s
+ s
+ ···
3 5 7
 
1 1 1
− s
+ s
+ s
+ ··· (43)
3 9 15

Y

−s −s
ζ(s) (1 − 2 ) (1 − 3 ) · · · (1 − p−s
n )··· = ζ(s) (1 − p−s
n ) (44)
n=1
= 1. (45)

Here, each subsequent multiplication by the nth prime pn leaves only terms
that are powers of p−s . Therefore,

!−
Y
∞ 1

ζ(s) = (1 − p−s
n ) , (46)
n=1

which is known as the Euler product formula (Hardy 1999, p. 18; Krantz

1999, p. 159), and called the golden key; by Derbyshire (2004, pp. 104-

106). The formula can also be written

Y Y
ζ(s) = (1 − 2−s )− 1
(1 − q−s )− 1
(1 − r−s )− 1
(47)
q=1 r=3
(mod 4) (mod 4)

where q and r are the primes congruent to 1 and 3 modulo 4, respectively.

For even n > 2,


2
n−1
|Bn |πn
ζ(n) = , (48)
n!
Función ζ
1 Riemann Zeta function 11

where Bn is a Bernoulli number (Mathews and Walker 1970, pp. 50-53; Ha-

vil 2003, p. 194). Another intimate connection with the Bernoulli numbers

is provided by

Bn = (−1)n+ nζ(1 − n) 1
(49)

for n > 1, which can be written

Bn = −nζ(1 − n) (50)

for n> 2. (In both cases, only the even cases are of interest since Bn = 0

trivially for odd n.) Rewriting (50),

Bn+ 1
ζ(−n) = − (51)
n+1
for n = 1, 3, . . . (Havil 2003, p. 194), where Bn is a Bernoulli number, the
−1 1 −1 1
rst few values of which are , , , ,... (Sloane's A001067
12 120 252 240
and A006953).

Although no analytic form for ζ(n) is known for odd n,

1
X

Hk
ζ(3) = , (52)
2
k=1
k 2

where Hk is a harmonic number (Stark 1974). In addition, ζ(n) can be

expressed as the sum limit

1
X
x  
k
n
ζ(n) = lim cot (53)
x→∞ (2x + 1) n k=1
2x + 1

for n = 3, 5, . . . (Apostol 1973, given incorrectly in Stark 1974).

For µ(n) the Möbius function,

1
X

µ(n)
= (54)
ζ(s) n=1
ns

@AOsorioC
12

(Havil 2003, p. 209).

The values of ζ(n) for small positive integer values of n are

ζ(1) = ∞ (55)
2
π
ζ(2) = (56)
6

ζ(3) = 1.2020569032 . . . (57)


4
π
ζ(4) = (58)
90

ζ(5) = 1.0369277551 . . . (59)


6
π
ζ(6) = (60)
945

ζ(7) = 1.0083492774 . . . (61)


8
π
ζ(8) = (62)
9450

ζ(9) = 1.0020083928 . . . (63)


10
π
ζ(10) = (64)
93555

Euler gave ζ(2) to ζ(26) for even n (Wells 1986, p. 54), and Stieltjes (1993)

determined the values of ζ(2), . . . , ζ(70) to 30 digits of accuracy in 1887.

The denominators of ζ(2n) for n= 1, 2, . . . are 6, 90, 945, 9,450, 93,555,

638,512,875, . . . (Sloane's A002432). The numbers of decimal digits in the

denominators of ζ (10n ) for n = 0, 1, . . . are 1, 5, 133, 2,277, 32,660,

426,486, 5,264,705, . . . (Sloane's A114474).

An integral for positive even integers is given by

Z
(−1)n+ 2 n− π n 1 2 3 2 1

ζ(2n) = E (n−1) (x)


2 dx, (65)
(2 n − 1)(2n − 2)!
2
0

Función ζ
1 Riemann Zeta function 13

and integrals for positive odd integers are given by


Z
(−1)n 2 n− π n+
2 1 2  πx 1 1

ζ(2n + 1) = E n (x) tan dx 2 (66)


(2 n+ − 1)(2n)!
2 1
2
Z
0

(−1)n 2 n− π n+
2 1 2  πx 1 1

= E n (x) cot dx 2 (67)


(2 n+ − 1)(2n)!
2 1
2
Z
0

(−1)n 2 n π n+
2 2 1
 πx  1

= B n+ (x) tan dx 2 1 (68)


(2n + 1)! 2
Z
0

(−1)n+ 2 n π n+
1 2 2 1
 πx  1

= B n+ (x) cot dx, 2 1 (69)


(2n + 1)! 2 0

where En (x) is an Euler polynomial and Bn (x) is a Bernoulli polynomial

(Cvijovic and Klinowski 2002; J. Crepps, pers. comm., Apr. 2002).

The value of ζ(0) can be computed by performing the inner sum in equation
(14) with s = 0,
X
∞ Xn  
1 k n
ζ(0) = − n+1
(−1) , (70)
n=0
2
k= 0
k
to obtain
X

δ n 0, 1 1
ζ(0) = − n+1
=− +1
0
=− , (71)
2 2 2
n=0

where δ n is the Kronecker delta.


0, Similarly, the value of ζ(−1) can be

computed by performing the inner sum in equation (14) with s = −1,


X
∞ Xn  
1 1 k n
ζ(−1) = − n+1
(−1) (k + 1), (72)
3
n=0
2
k= 0
k

which gives

1
X

δ 0, n − nδ n
1,
ζ(−1) = − n+1
(73)
3 2
n=0
 
1 1 1
= − +1
0
− 1 +1
(74)
3 2 2

1
= − (75)
12

@AOsorioC
14

This value is related to a deep result in renormalization theory (Elizalde et

al. 1994, 1995, Bloch 1996, Lepowski 1999).

It is apparently not known if the value

 
1
ζ = −1.46035450880 . . . (76)
2

(Sloane's A059750) can be expressed in terms of known mathematical cons-

tants. This constant appears, for example, in Knuth's series.

Rapidly converging series for ζ(n) for n odd were rst discovered by Rama-
nujan (Zucker 1979, 1984, Berndt 1988, Bailey et al. 1997, Cohen 2000).

For n>1 and n ≡ 3 (mod 4),


n+1
π X X
2n−1 n
  ∞
2 k− n+1 1
1
ζ(n) = (−1) Bn+ 1−2k B2k −2 ,
(n + 1)! k= 2k
0 k=1
kn (e 2 πk − 1)
(77)
 
n
where Bk is again a Bernoulli number and is a binomial coecient.
k
n
The values of the left-hand sums (divided by π ) in (77) for n = 3, 7, 11, . . .
7 19 1453 13687 7708537
are , , , , , . . . (Sloa-
180 56700 425675250 390769879500 21438612514068750
ne's A057866 and A057867). For n > 5 and n ≡ 1 (mod 4), the correspon-

ding formula is slightly messier,

n+1
(2π)n X4 
n+1

k
ζ(n) = (−1) (n + 1 − 4k) Bn+ 1 −2k B2k
(n + 1)!(n − 1) k= 0
2k
 
4πk
πk 2
X∞ e 1 + −1
n−1
−2 ,
2
(78)
k n (e πk − 1) 2
k= 1

(Cohen 2000).

Función ζ
1 Riemann Zeta function 15

Dening
X

1
S± (n) = , (79)
k=1
kn (e 2πk ± 1)
the rst few values can then be written

7 3
ζ(3) = π − 2S− (3) (80)
180
1 5
72 2
ζ(5) = π − S− (5) − S+ (5) (81)
294 35 35
19 7
ζ(7) = π − 2S− (7) (82)
56700
125 9
992 2
ζ(9) = π − S− (9) − S+ (9) (83)
3704778 495 495
1453 11
ζ(11) = π − 2S− (11) (84)
425675250
89 13
16512 2
ζ(13) = π − S− (13) − S+ (13) (85)
257432175 8255 8255
13687 15
ζ(15) = π − 2S− (15) (86)
390769879500
397549 17
261632 2
ζ(17) = π − S− (17) − S+ (17) (87)
112024529867250 130815 130815
7708537 19
ζ(19) = π − 2S− (19) (88)
21438612514068750
68529640373 21
4196352
ζ(21) = π − S− (21)
1881063815762259253125 2098175
2
− S+ (21) (89)
2098175

(Ploue 1998).

Another set of related formulas are

π 3
16
X

1 2
X

1
ζ(3) = + − (90)
28 7
n=1
n 3
(enπ + 1) 7
n=1
n (e3 2πn + 1)
X

1 259
X

1 1
X

1
ζ(5) = 24 − −
n=1
n 5
(enπ − 1) 10
n=1
n (e
5 2 πn − 1) 10
n=1
n (e
5 πn
4
− 1)
(91)

@AOsorioC
16

7 π 5
328
X

1 419
X

1
ζ(5) = − + −
1840 115
n=1
n 5
(eπn − 1) 460
n=1
n (e5 πn
2
− 1)
9
X

1 261
X

1
− πn
+ πn
115
n=1
n (e5 3
− 1) 1840
n=1
n (e5 6
− 1)
9
X∞
1
− πn
(92)
1840
n=1
n (e5 12
− 1)
304
X

1 103
X

1
ζ(7) = −
13
n=1
n 7
(eπn − 1) 4
n=1
n (e 7 2 πn − 1)
19
X

1
− πn
(93)
52
n=1
n (e
7 4
− 1)
64
X

1 441
X

1
ζ(9) = +
3
n=1
n 9
(eπn − 1) 20
n=1
n (e9 2 πn − 1)
X∞
1 4763
X

1
−32 πn
− πn
n=1
n (e9 3
− 1) 60
n=1
n (e9 4
− 1)
529
X∞
1 1
X

1
+ πn
− πn
(94)
8
n=1
n (e9 6
− 1) 8
n=1
n (e9 12
− 1)

(Ploue 2006).

Multiterm sums for odd ζ(n) include

X∞
(−1)k+ 5
X∞
(−1)k+ Hk−
( ) 1 1 2

1
ζ(5) = 2  −   (95)
2k 2 2k 5 3
k= k k= k 1 1

k k
5
X (−1)k+

25
X (−1)k+ H( )
∞ 1 1 4

ζ(7) =  +  k− 1
(96)
2 2k 2 2k 7 3
k= k k= k 1 1

k k

Función ζ
1 Riemann Zeta function 17

9
X∞
(−1)k+ 5
X1

(−1)k+ Hk−
( ) X∞ ( )
(−1)k+ Hk− 1 2 1 4

1 1
ζ(9) =  −   +5  
4 2k
9
4 2k 7
2k 5
k= k
1 k= k 1k= k 1

k k k
45
X∞
(−1)k+ Hk−
( ) 1
25
X∞ 6 ( )
(−1)k+ Hk− Hk−
( ) 1 2 4

1 1 1
+   −   (97)
4 2k 3
4 2k 3
k= 1k k= k 1

k k

X∞
5 (−1)k+ 1
25
X∞ ( )
(−1)k+ Hk− 1 4

1
ζ(11) =  +  
2 2k 11
2 2k 7
k= k 1 k= k
1

k k
2
 
( ) 4

75
X

(−1) k+1 (8)
Hk−1 125
X
∞ (−1) Hk− k+1
1
−   +   (98)
4 3
2 k 4 2k 3
k=1 k k=1 k
k k
(r)
(Borwein and Bradley 1996, 1997; Bailey et al. 2007, p. 71), where Hn
is a generalized harmonic number.

G. Huvent (2002) found the beautiful formula

16
X

(2(−1)n + 1)Hn
ζ(5) = − . (99)
11
n=1
n 4

A number of sum identities involving ζ(n) include

X

(ζ(n) − 1) = 1 (100)
n=2
X

3
(ζ(n) − 1) = (101)
4
n=2, 4,...
X

1
(ζ(n) − 1) = (102)
4
n=3, 5,...

X

1
(−1)n (ζ(n) − 1) = . (103)
2
n=2

@AOsorioC
18

Sums involving integers multiples of the argument include

X

3
(ζ(2n) − 1) = (104)
4
n=1
X∞
1
 2 1

(ζ(3n) − 1) = −(−1) 3 H 3−i√3 + (−1) 3 H 3+i√3 (105)
3 2 2
n=1
X∞
1
(ζ(4n) − 1) = (7 − 2 coth(π)), (106)
8
n=1

where Hn is a harmonic number.

Two surprising sums involving ζ(x) are given by

X

(−1)k ζ(k)
= γ (107)
k=2
k
X

ζ(k) − 1
= 1 − γ, (108)
k=2
k

where γ is the Euler-Mascheroni constant (Havil 2003, pp. 109 and 111-

112). Equation (107) can be generalized to

X

(−x)k ζ(k)
= xγ + ln(x!) (109)
k=2
k

(T. Drane, pers. comm., Jul. 7, 2006) for −1 < x 6 1. Other unexpected

sums are
X

ζ(2n)
n
= ln(π) − 1 (110)
n=1
n(2n + 1)2 2

(Tyler and Chernho 1985; Boros and Moll 2004, p. 248) and

X

ζ(2n)
= ln(2π) − 1. (111)
n=1
n(2n + 1)

(110) is a special case of

X

ζ(2k, z)

1

k
= (2z− 1) ln z − − 2z+ 1 + ln(2π)− 2 ln (Γ (z)) , (112)
k=1
k(2k + 1)2 2
2

Función ζ
1 Riemann Zeta function 19

where ζ(s, a) is a Hurwitz zeta function (Danese 1967; Boros and Moll

2004, p. 248).

Considering the sum

X2
n−
ζ(k)ζ(n − k)
Sn = k
, (113)
2
k=2

then

lim Sn = ln(2), (114)


n→∞

where ln(2) is the natural logarithm of 2, which is a particular case of

X2
n−
lim ζ(k)ζ(n − k)xk− = x− − ψ (−x) − γ,
1 1
0 (115)
n→∞
k=2

where ψ (z)
0 is the digamma function and γ is the Euler-Mascheroni cons-

tant, which can be derived from

X

ζ(k)xk− = −ψ (1 − x) − γ
1
0 (116)
k=2

(B. Cloitre, pers. comm., Dec. 11, 2005; cf. Borwein et al. 2000, eqn. 27).

A generalization of a result of Ramanujan (who gave the m= 1 case) is

given by

X

1
X
m
ζ(2k)(2m + 1 − 2k
m+1
= −2 (117)
k=1
(k(k + 1)) 2
k=0
2 m + i − 2k

(B. Cloitre, pers. comm., Sep. 20, 2005).

@AOsorioC
20

An additional set of sums over ζ(n) is given by

X

ζ(n)
C 1, 1 = (118)
n=2
n!
Z∞ √ √
I (2 u) − u
1
= √ du (119)
0 (eu − 1) u
Z∞ ~
0 F (; 2; u) − 1
1
= du (120)
0
eu − 1
≈ 1.078189 (121)
X

ζ(2n)
C 2, 1 = (122)
n=1
n!
X∞
1
= e n2 − 1 (123)
n=1
 4

3 u
Z u F
1 0 2 ; , 2;
2 4
= du (124)
0
eu −1
≈ 2.407447 (125)
X

ζ(2n)
C 2, 2 = (126)
n=1
(2n)!
Z∞
u + ~F (; 0 1 2;−u) − ~F (; 0 1 2; u)
= u
du (127)
0
2(1 − e )

≈ 0.869002. (128)

(Sloane's A093720, A076813, and A093721), where In (z) is a modied Bes-


sel function of the rst kind, p ~
Fq is a regularized hypergeometric function.
These sums have no known closed-form expression.

Función ζ

Anda mungkin juga menyukai