See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281103415

Evaluating Euler sums and deriving series for $\zeta(3)$

Research · August 2015

DOI: 10.13140/RG.2.1.2813.6160

CITATIONS READS

0 293

1 author:

Amrik Singh Nimbran

Government of Bihar, India
56 PUBLICATIONS   12 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Euler sums for odd denominators View project

All content following this page was uploaded by Amrik Singh Nimbran on 20 August 2015.

The user has requested enhancement of the downloaded file.

 Evaluating Euler sums and deriving series for ζ(3)
Amrik Singh Nimbran

August 20, 2015

Abstract
The author evaluates two De Doelder like Euler sums and two sums of
the type found by Bailey, Borwein and Girgensohn based on their results by
a simple method. He further also derives classes of series for ζ(3).

1 Introduction
In April 1993, Enrico Au-Yeung, a graduate student of Jonathan Borwein at
Waterloo, conjectured on the basis of his computations that
∞ 
Hn 2 17π 4

X 17
= = ζ(4). (1)
n 360 4
n=1

It was was soon established by Borweins[3]. Au-Yeung was unaware of the

fact that De Doelder[4] had in 1991 evaluated the following sum
∞ 
Hn 2 11π 4

X 11
= = ζ(4). (2)
n+1 360 4
n=1

Formula (1) initiated a fruitful research on the experimental evaluation of

“Euler sums” involving heavy use of M aple. I reproduce one result so discovered[2]:
∞
X Hn2 2 1 1 37 6
4
= ζ(6) − ζ(4)ζ(2) − ζ 2 (3) + ζ 3 (2) = π − ζ 2 (3). (3)
(n + 1) 3 3 3 22680
n=1

This aroused interest in Euler’s 46-page long paper[10] published in 1776. A

few papers[1, 14, 16] on the topic have been published since 1991. I intend to
evaluate Euler sums like (1) and (3) as well as sums of other types for deriving
classes of series for ζ(3) in terms of only ζ(2) or/and ζ(4). But before that, I
consider it appropriate to give some necessary (and interesting) background.

1
2 Gamma function, Psi function and Zeta function
Leonhard Euler (1707-1783), the great Swiss mathematician, will always be re-
membered for his outstanding work on infinite series and products. He evaluated
the series of reciprocals of even powers of the positive integers, known as the
Euler zeta function ζ(2m) and discovered a beautiful relationship between ζ(s)
and the prime numbers
∞  −s
X 1 Y 1
ζ(s) = = 1− , <(s) > 1.
ns p
n=1 p−prime

Euler extended the concept of factorial, the product of all natural numbers
less than or equal to a given natural number, to positive rational numbers. He
recorded in a letter dated October, 13, 1729 to Goldbach and in his paper[5]
presented to the St. Petersburg Academy on November 28, 1729:
∞
Y n1−m (n + 1)m
m! = . (4)
n+m
n=1
√
Setting m = 12 , he obtained Γ( 32 ) = 2π . (The notation for the generalized
Factorial function, the Gamma function, Γ(z), defined for all complex numbers
except non-positive integers, was given in 1839 by Legendre.)
In [5] and his letter of January, 8, 1730 to Goldbach, Euler took up the series
n! and found the general term for the factorial as (− ln x)n dx. Therefore,
P R
R1
the terms of the series transform into a definite integral 0 (ln x1 )n dx , which on
R∞
substituting x = e−t , converts into 0 tn e−t dt. In fact, he noted for the ‘hyper-
geometric curve’, y = Γ(x + 1), a definite integral in Art.11 of his paper [9] pre-
R∞
sented to the St. Petersburg Academy on December 19, 1765: y = 0 e−v v x dv.
Extended to non-integral values, it becomes in modern notation:
Z ∞
Γ(z) = tz−1 e−t dt, R(z) > 0. (5)
0

In Art.10 of [9], Euler recorded an infinite product representation:

√
2π(x + n) x + n x+n s
 
y = Qn e , (6)
k=1 (x + k) e

where
∞
X c2m (2m − 2)!
s= (−1)m−1 ,
22m−1 (x + n)2m−1
m=1

2
c2m being the coefficient of π 2m occurring in Euler’s zeta function ζ(2m).
In Art.12, he noted two more product representations (including one recorded
in his letter to Goldbach):
∞
n+1 x
 
Yn
y = (7)
n+x n
n=1
∞
x Y
2n + 1 + x x
  
1+x n
= . (8)
2 n + x 2n − 1 + x
n=1

Gauss [12] defined the Gamma function, denoting it by Πz as:

k!k z
Πz = lim Qk . (9)
j=1 (z + j)
k→∞

He uses a different functional equation: Π(z + 1) = (z + 1)Πz.

Weierstrass came up with this infinite product expression:
∞ 
1 γz
Y z  −z
= ze 1+ e n, (10)
Γ(z) n
n=1

where γ is Euler’s constant (the symbol occurs nowhere in his writings) introduced
by Euler as the constant of integration C in a paper of 1734 stating that if
1 + 21 + 13 + · · · + n1 is taken as s, then ds = n+1
dn
, and thus s = C + ln(n + 1).
1 1 1
In a paper of 1776, he wrote 1 + 2 + 3 + · · · + x = C + ln x. Thus ln(n + 1) was
replaced with ln n.
The constant γ crops up when one evaluates the partial sums of the har-
monic series Hn = nk=1 k1 which were first investigated by Euler in 1729 [6]. He
P
n
found the general summation term of the progression to be 1−x
R
1−x dx and stated
that from this term it is possible for that progression to be interpolated. He
further recorded that the summation term for the progression a1 , a+b 1 1
, a+2b ,...,
nb
R 1−(ax) a
was b dx.
1−(ax) a
In Art.13 of his paper [9], Euler gave the rudiments of the Digamma function,
the logarithmic derivative of the Gamma function:
∞
1 dy Γ0 (x + 1) X x
= = −∆ + , (11)
y dx Γ(x + 1) n(n + x)
n=1

Γ0 (1)
where ∆ = γ. Putting x = 0, he obtained (Art.14 of [9]) Γ(1) = γ. Then in
1 dy
Art.17, he set y dx = 0 to deduce that this is true if x satisfies the equation

3
 P∞ x
∆= n=1 n(n+x) and followed it with:
∞
X
∆= (−1)n−1 xn ζ(n + 1). (12)
n=1

More series involving γ and ζ(n) appear in Euler’s paper[11].

Gauss studied the Gamma function and the Digamma function while dealing
with the hypergeometric series in 1812 and gave the modern notation of the Psi
function, using the capital letter Ψ(z) in [12]. He gave two integral expressions:
Z 1
zλ

1
Ψ(λ) = − − dz, (13)
0 ln z 1 − z
Z ∞  −u
e−λu

e
= − u du. (14)
0 u e −1
Since Gauss starts with a slightly different definition (from that of Euler) of
the Gamma function, his Psi function also differs. Euler gives ψ(1) = −γ, Gauss
takes Ψ(0) = −γ. Euler noted in Art.14 of [9], taking y = 1 · 2 · 3 · · · n,
 
dy 1 1 1
= 1 · 2 · 3···n 1 + + + ··· + − ∆
dx 2 3 n
which in modern notation becomes:

ψ(n + 1) = Hn − γ.

Taking the log of the two sides of (10) and then differentiating yields
∞  
1 X 1 1
ψ(z) = −γ − + − , z 6= 0, −1, −2, . . . (15)
z n z+n
n=1

This series representation

∞  
X 1 1
ψ(z) = −γ + − , (16)
n z+n−1
n=1

yields the telescoping sum

∞  
X 1 1
ψ(1 + z) − ψ(z) = − ,
z+n−1 z+n
n=1

which in turn leads to the recurrence relation derivable from Γ(1 + z) = zΓ(z)
1
ψ(1 + z) = ψ(z) + . (17)
z

4
 In fact, we have this general formula for n ∈ N
n
X 1
ψ(n + z) = ψ(z) + . (18)
z+k−1
k=1

Further, for m ∈ N:
m−1  
1 X j
ψ(mz) = ln m + ψ z+ . (19)
m m
j=0

Most of the properties of the Psi function can be obtained directly from the
corresponding properties of the Gamma function. The reflection formula comes
from Γ(1 − z)Γ(z) = π csc πz

ψ(1 − z) − ψ(z) = π cot πz, z 6= 0, −1, −2, . . . (20)

1 1
Setting z = a and z = b in the formula (16), and taking their difference yields
∞
X 1 ψ( a1 ) − ψ( 1b )
= , a 6= b, a, b ∈ N. (21)
(an + 1)(bn + 1) b−a
n=0

1
Expanding 1−t by the Binomial theorem, integrating term by term and using
∞
X 1
ψ(z) = −γ + (z − 1) ,
(n + 1)(z + 1)
n=0

we obtain the following integral representation

Z 1
1 − tz−1
ψ(z) + γ = dt, <(z) > 0. (22)
0 1−t
We immediately infer from this integral that in the case of integral arguments
n−1
X 1
ψ(n) + γ = = Hn−1 .
k
k=1

The following formula relates ψ(z) and ζ(z)

∞
X
ψ(z) = −γ − π cot πz − z m−1 ζ(m), |z| < 1. (23)
m=2

Euler introduced his zeta function in the wake of his solution of the Basel
problem, finding the the exact sum of the reciprocals of the squares of all natu-
ral numbers. First posed by the Italian mathematician Pietro Mengoli in 1644,

5
the problem was referred to in Jakob Bernoulli’s 1689 treatise Tractatus de se-
riebus infinitis published in Basel(hometown of the Bernoullis and Euler). Jacob’s
younger brother Johann tried unsuccessfully and urged his student Euler to at-
tempt it. Euler solved it in 1735 by extending the logic valid for a polynomial of
finite degree to an infinite series treated as a polynomial of infinite degree.
Recall that every polynomial can be expressed as a product of linear factors in-
volving its roots. Taking the power series expansion sinx x = ∞ n−1 x2n−2
P
n=1 (−1) (2n−1)!
as a polynomial in x2 = y and knowing its zeros to be (±π)2 , (±2π)2 , (±3π)2 , . . . ,
Euler applied the rule that the sum of the reciprocals of all the individual roots
equals the coefficient of the linear term (y = x2 here). He thus obtained the
2
formula ζ(2) = π6 that immortalized him.
Euler generalized the Basel problem in a paper[7] read in the St. Peters-
burg Academy on December 5, 1735 (published in 1740) and evaluated the sums
P∞ 1
n=1 n2m for m up to 6. He kept returning to the topic time and again. In a
paper[8] published in 1743, he does the sum of the reciprocals of odd squares by
an arcsine integral and records the formulas for m up to 13 which were repro-
duced in chapter X of his most famous book Introductio in analysin infinitorum,
volume 1 (written 1745, published 1748). His paper contains the formula
∞
X 1 22m−1 c2m π 2m
=
n2m (2m + 1)!2
n=1

2m−2
while in his book he writes down the coefficient of π 2m in the form 2(2m+1)! c2m

removing 2 from the denominator. Apparently, he had not yet recognised the
connection of his coefficient with the Bernoulli numbers. The connection was
clear to him by the time of the publication of his Institutiones calculi differen-
tialis (1755), wherein he (chapter V, p. 340, 3rd line) uses the term ‘numeris
Bernoullians’ for the coefficients. So
∞
X 1 |B2m |(2π)2m
= , m ≥ 1,
n2m 2 · (2m)!
n=1

B2m appear as coefficients in the expansion

∞
X Bk z k
z
= , |z| < 2π.
ez − 1 k!
k=o

All B4m+2 are positive while all B4m , m > 0, are negative. Some of the first
few numbers are: B0 = 1, B1 = − 12 , B2 = 16 , B4 = − 30 1 1
, B6 = 42 , B8 =

6
 1 5
− 30 , B10 = 66 . Bk = 0 for all odd values of k > 1. These numbers satisfy the
following recurrence relations:
m  
X m+1
Bk = 0, m ≥ 1.
k
k=0

Euler extended his list of ζ(2m) up to m = 17 in his 1775 paper[13].

No formula exists for ζ(2m + 1). In 1978, However, Euler did find this general
formula for the alternating series:
∞
X 1 |E2m |(2π)2m+1
(−1)n−1 = , m ≥ 0,
(2n − 1)2m+1 22m+2 · (2m)!
n=1

where E2m denote the Euler numbers. These numbers appear as coefficients in
this power series expansion:
∞
X E2m x2m
sec x = (−1)m .
(2m)!
m=0

All E4m+2 are negative while all E4m , m > 0, are positive. The first few
numbers are: E0 = 1, E2 = −1, E4 = 5, E6 = −61, E8 = 1385, E10 = −50521.
In 1978, Apéry proved that ζ(3) is irrational and gave an alternating series
∞
5X 1
ζ(3) = (−1)n−1
.
2
n=1
n 2n
3
n

With this detailed background, I now come to the subject of my paper. But
we need to introduce here the polylogarithm functions defined by
∞
X zm
Lin (z) = , |z| ≤ 1, n ∈ N \ {1},
mn
m=1
Z z
Lin−1 (z)
= dz.
0 z

These are treated superbly in Lewin’s book[15]. De Doelder begins his pa-
per with the statement that it was the discussion on integrals related with
dilogarithms and polylogarithms appearing in this book that led him to his re-
sults. Jung, Cho and Choi[14] too follow Lewin’s method and use integrals from
Lewin’s book to evaluate different Euler sums. In particular, they show that
P∞ Hn−1 1 1
n=2 n3 = 4 ζ(4) = 4 Li4 (1) can be obtained from the integral (7.65) recorded
on page 204 of Lewin’s book.

7
 In his paper [10], Euler treated the sum
     
1 1 1 1 1 1 1 1 1
Sm, n = 1 + m 1 + n + m 1 + n + n + m 1 + n + n + n + . . . .
2 2 3 2 3 4 2 3 4
He gave formulas for many integral values of m + n ≥ 2 with both m, n being
positive integers. But his notation is a little cumbersome.
We divide our exposition in two parts — dealing with the series having Hn
and Hn2 in the numerator. The series in the first part are directly deducible from
Euler’s general formula while the series in the second part are deduced from De
Doelder’s formula and the formula found by Bailey/associates.

3 Euler sums with numerator Hn

A shift of index shows that
∞ ∞
X Hn−1 X Hn
= .
nm (n + 1)m
n=2 n=1
It is easy to see that
∞ ∞ ∞
X Hn X Hn X 1
− = .
nm (n + 1) m nm+1
n=1 n=1 n=1
My method of evaluating Euler sums is founded on these two simple obser-
vations.
We begin with the simplest series ∞
P Hn P∞ Hn
n=1 n . Obviously it diverges as n=1 n =
P∞ 1 P∞ Hn −1 P∞ 1
n=1 n + n=1 n and the first series n=1 n diverges while the general term
Hn −1
n of the second is positive for all n > 1.
However, the series of the differences of the partial sums converges:
∞   X ∞ ∞
X Hn Hn Hn X 1
− = = = ζ(2). (24)
n n+1 n(n + 1) n2
n=1 n=1 n=1
We find this formula (in Art. 12) in Euler’s paper[10]:
∞
X Hn
= 2ζ(3). (25)
n2
n=1
Following the method used above, we obtain
∞
X Hn
= ζ(3); (26)
(n + 1)2
n=1
∞
X Hn
= ζ(3) + ζ(2) − 2. (27)
(n + 2)2
n=1

8
 Combining the above three series, we obtain
∞
π2 1 X Hn
ζ(3) = + ; (28)
9 3 n (n + 1)2
2
n=1
∞
π2 X 4Hn
ζ(3) = −4+ . (29)
2 n (n + 1)(n + 2)2
2
n=1

We can go on like this and keep increasing the number of factors in the
denominator.
Formula (25) is a special case of this general formula for m ≥ 2
∞ m−2
X Hn X
2 = (m + 2)ζ(m + 1) − ζ(m − k) ζ(k + 1), m = 2, 3, 4, . . . , (30)
nm
n=1 k=1

or, using the observations noted above,

∞ m−2
X Hn X
2 = mζ(m + 1) − ζ(m − k) ζ(k + 1), m = 2, 3, 4, . . . . (31)
(n + 1)m
n=1 k=1

Putting m = 3 in (30) yields

∞
X 4Hn
= 5ζ(4). (32)
n3
n=1

Treated (removing the last term of Hn ) like the earlier series, it yields
∞
X 4Hn
= ζ(4). (33)
(n + 1)3
n=1

Similarly, we find that

∞ ∞ ∞ ∞ ∞
X Hn X Hn X Hn X 1 X 1
3
− 3
= 3
= 2
−
(n + 1) (n + 2) n(n + 1) n(n + 1) (n + 1)3
n=1 n=1 n=1 n=1 n=1
∞ ∞ ∞
X 1 X 1 X 1
= − 2
− = 3 − ζ(2) − ζ(3).
n(n + 1) n n3
n=1 n=2 n=2

Hence, we obtain
∞
X Hn 1
3
= ζ(4) + ζ(3) + ζ(2) − 3. (34)
(n + 2) 4
n=1

We can go on like this.

9
 Subtracting (33) from (32), we get
∞ ∞
X 3Hn X Hn
2 2
+ = ζ(4).
n (n + 1) n (n + 1)3
3
n=1 n=1

Using the value found earlier, we obtain

∞
X Hn
= ζ(4) − 9ζ(3) + 6ζ(2), (35)
n3 (n+ 1)3
n=1

that is,
∞
π4 π2 1 X Hn
ζ(3) = + − .
810 9 9 n3 (n + 1)3
n=1
Now we find that
∞
X Hn
= 2ζ(3) + ζ(2) − 4. (36)
(n + 1)2 (n + 2)2
n=1

Using various values derived earlier, one gets

∞
X Hn
= 15 − 4ζ(2) − 7ζ(3),
(n + 1)3 (n + 2)3
n=1

yielding a series for ζ(3) with only ζ(2) and two cubic factors in the denominator:
∞
15 2π 2 1 X Hn
ζ(3) = − − . (37)
7 21 7 (n + 1)3 (n + 2)3
n=1

Subtracting (37) from (35) and shifting terms, we get

∞ ∞
1 15 X 3Hn X Hn
ζ(3) = ζ(4) + 5ζ(2) − − 3 3
− .
2 2 n (n + 1)(n + 2) n (n + 1)3 (n + 2)3
3
n=1 n=1

Now we find on decomposition of the first series that

3Hn 3Hn 15Hn 15Hn 3Hn
= − + +
n3 (n + 1)(n + 2) 3 8n 3 16n 2 16(n + 2) 2 8(n + 2)3
3Hn 3Hn
+ − .
2n(n + 1) 2(n + 1)(n + 2)
Putting the values of sums on the right hand side, we get
3Hn 15 15 15
= ζ(4) − 2ζ(3) + (ζ(3) + ζ(2) − 2)
n3 (n
+ 1)(n + 2) 3 32 16 16
 
3 1 3 3 9 9 45 9
+ ζ(4) + ζ(3) + ζ(2) − 3 + ζ(2) − = ζ(4) − ζ(3) + ζ(2) − ,
8 4 2 2 16 16 16 2

10
that is,
∞
5π 2 π4 48 X 16Hn
ζ(3) = − − − . (38)
6 630 7 7n (n + 1)3 (n + 2)3
3
n=1
Deriving the next sum and combining that with its predecessor would give
∞
329 31π 2 X 2Hn
ζ(3) = − − . (39)
104 156 13(n + 2)3 (n + 3)3
n=1

This can be combined with (37) and the sum so found may then be combined
with (38) getting four quadratic factors in the denominator. We can thus go on
increasing the number of factors in the denominator.

4 Euler sums with numerator Hn2

Euler also obtained the relation
2
2Sm, m = Sm + S2m .

Thus    
1 1 1 1 1 7
1+ 2 1 + 2 + 2 1 + 2 + 2 + · · · = ζ(4). (40)
2 2 3 2 3 4
But Euler did not evaluate series with Hn2 in the numerator. We shall now
derive such formulas.
The comparison of the terms in the two series in the following expression
makes it clear that
∞ ∞ ∞ X Hn2 − Hn−1
2 ∞
X Hn2 X H2 X H2
n n
= − = .
n(n + 1) n (n + 1) n
n=1 n=1 n=1 n=1

Now Hn2 − Hn−1

2 = (Hn − Hn−1 )(Hn + Hn−1 ) = n1 (2Hn − n1 ) = 2 Hnn − n12 .
This relation is the basis of our evaluation of Euler sums with Hn2 in the
numerator.
Hence, we get
∞ ∞ ∞
X Hn2 X Hn X 1
=2 − = 3ζ(3). (41)
n(n + 1) n2 n3
n=1 n=1 n=1

We can derive the series conjectured by Au-Yeung in the following manner.

On comparing the terms of the two series, we notice that
∞  ∞  ∞
Hn 2 X
2 X
Hn2 − Hn−1
2

X Hn
− = .
n (n + 1) n2
n=1 n=1 n=1

11
 Therefore,
∞  2 ∞  2 ∞ ∞
X Hn X Hn X Hn X 1 3
− =2 − = ζ(4).
n (n + 1) n3 n4 2
n=1 n=1 n=1 n=1

Thus knowing either sum yields the other one.

Now consider the simple formula
X
(a1 + a2 + a3 + · · · + an )2 = a21 + a22 + a23 + . . . a2n + 2ai · aj ,
i, j≤n; i6=j

where the total number of terms is n(n+1)

2 and so the number of terms within the
n(n−1)
sigma sign is 2 . We deduce from two of the results recorded earlier, this new
kind of sum:
∞ 1
P
X 1≤i, j≤n; i6=j i·j 5
2
= ζ(4). (42)
n 4
n=2
We see that
∞  ∞
!2 ∞
!2 ∞
Hn 2 2Hn2

X Hn X Hn X Hn X
− = + −
n n+1 n n+1 n(n + 1)
n=1 n=1 n=1 n=1

So using the values from (1), (2) and (41), we get

∞  2
X Hn
= 7ζ(4) − 6ζ(3). (43)
n(n + 1)
n=1

Using the technique used above, I derived this De Doelder like sum
∞  ∞ 
Hn 2 X Hn 2
 
X 11
= + 2ζ(3) − 3 = ζ(4) + 2ζ(3) − 3. (44)
n+2 n+1 4
n=1 n=1

We can combine the two formulas (2) and (44) using

∞
X Hn2
= ζ(2) + 1, (45)
(n + 1)(n + 2)
n=1

which is obtained from (41). We thus get

∞  2
X Hn 11
= ζ(4) + 2ζ(3) − 2ζ(2) − 5. (46)
(n + 1)(n + 2) 2
n=1

Combining the two formulas (43) and (46), we obtain

∞
X Hn2 3 1 5
= ζ(4) − 2ζ(3) + ζ(2) + ,
n2 (n + 1)(n + 2)2 8 2 4
n=1

12
that is, we get a fast converging formula for ζ(3) :
∞
5 π2 π4 X Hn2
ζ(3) = + + − . (47)
8 24 480 2n (n + 1)(n + 2)2
2
n=1

To derive the next series in the chain, we proceed as follows.

∞  ∞  ∞ ∞ ∞
Hn 2 X Hn 2 X Hn2 − Hn−1
2
 
X X 2Hn X 1
− = 2
= 2
− .
n+2 n+3 (n + 2) n(n + 2) n (n + 2)2
2
n=1 n=1 n=1 n=1 n=1

This can be expressed as

∞  ∞  ∞ ∞ ∞
Hn 2 X Hn 2 X Hn
 
X X Hn X 1
− = − 2
− .
n+2 n+3 n(n + 2) (n + 2) n (n + 2)2
2
n=1 n=1 n=1 n=1 n=1

We evaluated the middle sum earlier and found the sum of the last series:
∞
X 1 π 2 11
= − . (48)
n2 (n + 2)2 12 16
n=1

To derive the first series, we put in an intervening series and its negative:
∞ ∞ ∞
!
X Hn 1 X Hn X Hn
= −
n(n + 2) 2 n n+2
n=1 n=1 n=1
∞ ∞ ∞ ∞
!
1 X Hn X Hn X Hn X Hn
= − + −
2 n n+1 n+1 n+2
n=1 n=1 n=1 n=1
∞ ∞
!
1 X Hn X Hn 1
= + = (ζ(2) + 1) .
2 n(n + 1) (n + 1)(n + 2) 2
n=1 n=1

Thus we obtain
∞  2 ∞ 
Hn 2 51

X Hn X
− = − ζ(2) − ζ(3).
n+2 n+3 16
n=1 n=1

Hence,
∞ 
Hn 2 11

X 99
= ζ(4) + 3ζ(3) + ζ(2) − . (49)
n+3 4 16
n=1

We can go on like this. We can use

∞
X Hn2 1 11
= ζ(2) + , (50)
(n + 2)(n + 3) 2 8
n=1

13
which is derived from (45), to combine the last two formulas to get a series for
ζ(3) involving only ζ(4)
∞  2
X Hn 11 191
= ζ(4) + 5ζ(3) − . (51)
(n + 2)(n + 3) 2 16
n=1

We can also combine the three series, viz. (41), (45) and (50) having linear
factors in the denominator to obtain a series involving ζ(2) only
∞
1 5 X 2Hn2
ζ(3) = ζ(2) + + . (52)
2 24 n(n + 1)(n + 2)(n + 3)
n=1

Using (43), (51) and (52), we can derive series with four quadratic factors in
the denominator as
 2  2
Hn Hn Hn
− =
n(n + 1) (n + 2)(n + 3) n(n + 1)
2
2Hn2

Hn
+ − .
(n + 2)(n + 3) n(n + 1)(n + 2)(n + 3)

We thus have
25 1 5 (2n + 3)Hn2
ζ(3) = ζ(4) + ζ(2) + − . (53)
4 4 48 [n(n + 1)(n + 2)(n + 3)]2

This two parallel classes of series can go on indefinitely with more and more
linear/quadratic factors in the denominator.
Let us derive a formula with two cubic factors in the denominator and see
what we get. For this, we begin by observing that
∞ ∞ ∞ ∞
X H2 n
X Hn2 X 2Hn X 1
− = − ,
n3 (n + 1)3 n4 n5
n=1 n=1 n=1 n=1

that is,
∞ ∞ ∞ ∞
X 3Hn2 X Hn2 X 2Hn X 1
+ = − ,
n2 (n + 1)2 n3 (n + 1)3 n4 n5
n=1 n=1 n=1 n=1
or,
∞
X Hn2
= 18ζ(3) − 21ζ(4) + 5ζ(5) − 2ζ(3)ζ(2). (54)
n3 (n + 1)3
n=1

This suggests that for higher powers in the denominators we are going to get
series involving ζ(5) or product like ζ(3)ζ(2).

14
 Since our method of computing difference shows that
∞ ∞ ∞ ∞
X H2 n
X Hn2 X 2Hn X 1
− = − ,
n4 (n + 1)4 n5 n6
n=1 n=1 n=1 n=1

using the formula (3) and the values of other sums, I derived
∞
X H2n 20 7 1 97
4
= ζ(6) − ζ(4)ζ(2) − 2ζ 2 (3) + ζ 3 (2) = π 6 − 2ζ 2 (3). (55)
n 3 3 3 22680
n=1

Further, we see that

∞ ∞ ∞ ∞
X Hn2 X H2
n
X 2Hn X 1
4
− 4
= 4
− .
(n + 1) (n + 2) n(n + 1) n (n + 1)4
2
n=1 n=1 n=1 n=1

Now
1 1 2 2 2 1
= 2 − + + + .
n2 (n + 1) 4 n (n + 1) 2 n(n + 1) (n + 1) 2 (n + 1) 3 (n + 1)4
Therefore,
∞ ∞ ∞ ∞
X Hn2 X H2
n
X 2Hn X 2Hn
− = −
(n + 1)4 (n + 2)4 n(n + 1) (n + 1)2
n=1 n=1 n=1 n=1
∞ ∞ ∞
X 2Hn X 2Hn X 1
− − −
(n + 1)3 (n + 1)4 n2 (n + 1)2
n=1 n=1 n=1
∞ ∞ ∞ ∞
X 1 X 2 X 2 X 2
− − − + ,
(n + 1)4 (n + 1)3 (n + 1)2 n(n + 1)
n=1 n=1 n=1 n=1

that is,
∞ ∞
X Hn2 X H2
n 1
4
− 4
= 2ζ(2) − 2ζ(3) − ζ(4)
(n + 1) (n + 2) 2
n=1 n=1

−[4ζ(5) − 2ζ(3)ζ(2)] − (2ζ(2) − 3) − (ζ(4) − 1) − 2(ζ(3) − 1) − (2ζ(2) − 1) + 2.

Hence, we obtain
∞
X H2n 2 3
4
= ζ(6) + 4ζ(5) + ζ(4) + 4ζ(3) + 2ζ(2)
(n + 2) 3 2
n=1 (56)
1 1
− ζ(4)ζ(2) − 2ζ(3)ζ(2) − ζ 2 (3) + ζ 3 (2) − 10.
3 3
Acknowledgement: The author wishes to thank Prof. Paul Levrie, University
of Antwerp, Belgium for a thorough reading of the draft that helped in removing
typographical errors.

15
References
[1] David Borwein, Jonathan M. Borwein And Roland Girgensohn, Explicit Eval-
uation of Euler Sums, Proceedings of the Edinburgh Mathematical Society
(1995) 38, 277–294.

[2] David H. Bailey, Jonathan M. Borwein and Roland Girgensohn, Experimental

Evaluation of Euler sums, Experimental Mathematics, vol.3(1994), No.1, 17–
30.

[3] David Borwein and Jonathan M. Borwein, On an intriguing integral and some
series related to ζ(4), Proc. Amer. Math. Soc. 123 (1995), 1191–1198.

[4] P. J. De Doelder, On some series containing ψ(x) − ψ(y) and (ψ(x) − ψ(y))2
for certain values of x and y, J. Comput. Appl. Math. 37 (1991), 125–141.

[5] L. Euler, De progressionibus transcendentibus seu quarum termini generales

algebraice dari nequeunt, Commentarii academiae scientiarum Petropolitanae
5, 1738, 36–57; Opera Omnia: Series 1, Volume 14, 1–24. Available as E19 at
http://eulerarchive.maa.org/

[6] L. Euler, De summatione innumerabilium progressionum, Commentarii

academiae scientiarum Petropolitanae 5 (1738), 91–105; Opera Omnia: Series
1, Volume 14, 25–41. E20.

[7] L. Euler, De summis serierum reciprocarum, Commentarii academiae scien-

tiarum Petropolitanae 7, 1740, 123-134. E41.

[8] L. Euler, Demonstration de la somme de cette suite 1 + 41 + 19 + 16 1

+ etc.,
Journ. lit. d’Allemange, de Suisse et du Nord, 2:1, 1743, 115–127. E63.

[9] L. Euler, De curva hypergeometrica hac aequatione expressa y=1.2.3.x, Novi

Commentarii academiae scientiarum Petropolitanae 13, 1769, 3–66; Opera
Omnia: Series 1, Volume 28, 41–98. E368.

[10] L. Euler, Meditationes circa singulare serierum genus originally published in

Novi Commentarii academiae scientiarum Petropolitanae 20, 1776, 140–186.
E477.

[11] L. Euler, De numero memorabili in summatione progressionis harmonicae

naturalis occurrente, Acta Academiae Scientarum Imperialis Petropolitinae
5, 1785, 45–75; Opera Omnia: Series 1, Volume 15, 569–603. E583.

16
 [12] C. F. Gauss, Disquisitiones generales circa seriem infinitam, Commenta-
tiones societatis regiae scientiarum, Gottingen, Vol. II (1813), 1–46; Werke,
Band III, 1876, 122–162, p.157, formulas (74), (75).

[13] L. Euler, De seriebus potestatum reciprocis methodo nova et facillima sum-

mandis, originally published in Opuscula Analytica 2, 1785, 257–274.

[14] Myungho Jung, Young Joon Cho and Junesang Choi, Euler Sums Evaluat-
able From Integrals, Commun. Korean Math. Soc. 19 (2004), No. 3, 545–555.

[15] Leonard Lewin, Polylogarithms and Associated Functions, Elsevier North

Holland, New York, 1981.

[16] Odd Magne Ogreid and Per Osland, More series related to the Euler series,
Journal of Computational and Applied Mathematics 136 (2001), 389–403.

AMRIK SINGH NIMBRAN, I.P.S.(Retd.)

Former Director General of Police, Bihar (INDIA)
Email: simnimas@yahoo.co.in / amrikn622@gmail.com

17

View publication stats