mathematics

Article
Families of Integrals of Polylogarithmic Functions †
Anthony Sofo
Victoria University, P. O. Box 14428, Melbourne City, Victoria 8001, Australia; anthony.sofo@vu.edu.au
† Lovingly dedicated to Mia Eva Manca.




Received: 30 December 2018; Accepted: 26 January 2019; Published: 3 February 2019


Abstract: We give an overview of the representation and many connections between integrals
of products of polylogarithmic functions and Euler sums. We shall consider polylogarithmic
functions with linear, quadratic, and trigonometric arguments, thereby producing new results and
further reinforcing the well-known connection between Euler sums and polylogarithmic functions.
Many examples of integrals of products of polylogarithmic functions in terms of Riemann zeta values
and Dirichlet values will be given. Suggestions for further research are also suggested, including a
study of polylogarithmic functions with inverse trigonometric functions.

Keywords: Euler sums; zeta functions; Dirichlet functions; series representation; harmonic number

MSC: 11M06; 11M32; 33B15

1. Introduction and Preliminaries

It is well known that integrals of products of polylogarithmic functions can be associated with
Euler sums, see Reference [1]. In this paper, we investigate the representations of integrals of the type

Z1
x m Liq (± x ) Lit (± x ) dx,
0

Z1
1  
Liq (± x ) Lit x2 dx,
x
0

for m ≥ −2, and for integers q and t. For m = −2, −1, 0, we give explicit representations of the integrals
R1
in terms of Euler sums. For the case x m Lit (− x ) Liq (− x ) dx, for m ≥ 0, we show that the integral
0
satisfies a certain recurrence relation. We also mention two specific integrals with a trigonometric
argument in the polylogarithm of the form

Z1
sin πx Li p (sin πx ) dx.
0

Some examples are highlighted, almost none of which are amenable to a computer mathematical
package. This work extends the results given in Reference [1], where the author examined integrals
with positive arguments of the polylogarithm. Devoto and Duke [2] also list many identities of lower
order polylogarithmic integrals and their relations to Euler sums. Some other important sources of
information on polylogarithm functions are the works of References [3] and [4]. In References [5]
and [6], the authors explore the algorithmic and analytic properties of generalized harmonic Euler
sums systematically, in order to compute the massive Feynman integrals which arise in quantum field

Mathematics 2019, 7, 143; doi:10.3390/math7020143 www.mdpi.com/journal/mathematics

Mathematics 2019, 7, 143 2 of 20

theories and in certain combinatorial problems. Identities involving harmonic sums can arise from
their quasi-shuffle algebra or from other properties, such as relations to the Mellin transform

Z1
M [ f (z)]( N ) = dz z N f (z) ,
0

where the basic functions f (z) typically involve polylogarithms and harmonic sums of lower weight.
Applying the latter type of relations, the author in Reference [7] expresses all harmonic sums of the
above type with weight w = 6, in terms of Mellin transforms and combinations of functions and
constants of lower weight. In another interesting and related paper [8], the authors prove several
identities containing infinite sums of values of the Roger’s dilogarithm function. defined on x ∈ [0, 1],
by 
1
 Li2 ( x ) + 2 ln x ln (1 − x ) ; 0 < x < 1

LR (x) = 0; x=0 .

 ζ (2) ; x=1
The Lerch transcendent,
∞
zm
Φ (z, t, a) = ∑ (m + a)t
m =0

is defined for |z| < 1 and < ( a) > 0 and satisfies the recurrence

Φ (z, t, a) = z Φ (z, t, a + 1) + a−t .

The Lerch transcendent generalizes the Hurwitz zeta function at z = 1,

∞
1
Φ (1, t, a) = ∑ (m + a)t
m =0

and the polylogarithm, or de-Jonquière’s function, when a = 1,

∞
zm
Lit (z) := ∑ mt
, t ∈ C when |z| < 1; < (t) > 1 when |z| = 1.
m =1

Let
n ∞
1 − tn
Z 1
1 n
Hn = ∑ r
=
0 1−t
dt = γ + ψ (n + 1) = ∑ j ( j + n)
, H0 := 0
r =1 j =1

be the nth harmonic number, where γ denotes the Euler–Mascheroni constant,

n
(−1)m−1 1 − tn
Z 1 
1
= ∑ m =
(m)
Hn lnm−1 (t) dt (1)
r =1
r ( m − 1) ! 0 1−t

is the mth order harmonic number and ψ(z) is the digamma (or psi) function defined by

d Γ0 (z) 1
ψ(z) := {log Γ(z)} = and ψ(1 + z) = ψ(z) + ,
dz Γ(z) z
moreover,
∞  
1 1
ψ(z) = −γ + ∑ −
n+1 n+z
.
n =0
Mathematics 2019, 7, 143 3 of 20

More generally, a non-linear Euler sum may be expressed as

q j !
(±1)n t r

(α j ) ( β ) mk
 
∑ np ∏ Hn ∏ Jn k
n ≥1 j =1 k =1

where p ≥ 2, t, r, q j , α j , mk , β k are positive integers and

!q !m

(α) q
 n
1 
( β) m
 n
(−1) j+1
Hn = ∑ jα
, Jn = ∑ jβ .
j =1 j =1

If, for a positive integer,

t r
λ= ∑ αj qj + ∑ β j mj + p,
j =1 j =1

then we call it a λ-order Euler sum. The polygamma function

∞
dk 1
ψ(k) ( z ) =
dzk
{ ψ ( z )} = (− 1 ) k +1
k! ∑ k +1
r =0 ( r + z )

and satisfies the recurrence relation

(−1)k k!
ψ ( k ) ( z + 1) = ψ ( k ) ( z ) + .
z k +1
The connection of the polygamma function with harmonic numbers is

( α +1) (−1)α (α)

Hz = ζ ( α + 1) + ψ (z + 1) , z 6= {−1, −2, −3, ...} . (2)
α!
and the multiplication formula is

p −1
1 j
ψ(k) ( pz) = δm,0 ln p +
p k +1
∑ ψ(k) ( z + p ) (3)
j =0

for p is a positive integer and δp,k is the Kronecker delta. We define the alternating zeta function (or
Dirichlet eta function) η (z) as

∞
(−1)n+1  
η (z) := ∑ nz
= 1 − 21 − z ζ ( z ) (4)
n =1

where η (1) = ln 2. If we put

∞ ( p)
(−1)n+1 Hn
S ( p, q) := ∑ nq
,
n =1

in the case where p and q are both positive integers and p + q is an odd integer, Flajolet and Salvy [9]
gave the identity
!
p
p p+i−1
2S ( p, q) = 1 − (−1) ζ ( p) η (q) + 2 (−1) ∑i+2k=q ζ ( p + i ) η (2k )
p−1
! (5)
q+j−1 j
+η ( p + q) − 2 ∑ j+2k= p (−1) η (q + j) η (2k) ,
q−1
Mathematics 2019, 7, 143 4 of 20

where η (0) = 12 , η (1) = ln 2, ζ (1) = 0, and ζ (0) = − 12 in accordance with the analytic continuation of
the Riemann zeta function. We also know, from the work of Reference [10] that for odd weight w = ( p + q)
we have
!
( p) [ 2p ] p + q − 2j − 1
∑∞
Hn p
BW ( p, q) = n =1 n q = (−1) ∑ j =1 ζ ( p + q − 2j) ζ (2j)
p−1
!
p
1

p +1

p [2] p + q − 2j − 1
+ 2 1 + 1
(− ) ζ ( ) ( ) (− ) ∑ j =1
p ζ q + 1 ζ ( p + q − 2j) ζ (2j) (6)
q−1
! !!
ζ ( p+q) p +1 p+q−1 p +1 p+q−1
+ 2 1 + (−1) + (−1) ,
p q

where [z] is the integer part of z. It appears that some isolated cases of BW ( p, q) , for even weight
( p + q) , can be expressed in zeta terms, but in general, almost certainly, for even weight ( p + q) ,
no general closed form expression exits for BW ( p, q) . Two examples with even weight are

∞ (2) ∞ (4)
Hn 1 Hn 37
∑ n4
= ζ 2 (3) − ζ (6) ,
3 ∑ n2
=
12
ζ (6) − ζ 2 (3) .
n =1 n =1

There are also the two general forms of identities

∞ (2p+1)
Hn
2 ∑ n2p+1
= ζ 2 (2p + 1) + ζ (4p + 2) ,
n =1

and
∞ (2p) 2
!
Hn (4p)!B2p
2 ∑ n2p
= 1−
2 ((2p)!)2 B4p
ζ (4p)
n =1

for p ∈ N and where B p are the signed Bernoulli numbers of the first kind. At even weight w = 8,
neither BW (2, 6) nor BW (6, 2) currently have a reduced expression in terms of zeta values and their
products. Using (1) we can express

∞ (2) Z 1
Hn 5 ln x Li6 ( x )
BW (2, 6) = ∑ n6
=
3
ζ (8) −
0 1−x
dx
n =1

and
∞ (6) Z 1 5
Hn 5 1 ln x Li2 ( x )
BW (6, 2) = ∑ n2
=
3
ζ (8) +
5! 0 1−x
dx.
n =1

The work in this paper extends the results of Reference [1] and later Reference [11], in which they gave
identities of products of polylogarithmic functions with positive argument in terms of zeta functions.
Other works, including References [12–23], cite many identities of polylogarithmic integrals and Euler
sums. The following lemma was obtained by Freitas, [1]:

Lemma 1. For q and t positive integers

R1 Lit ( x ) Liq ( x )
I (+,+) (q, t) = x dx
0 (7)
q −1 j +1 q +1
= ∑ j =1 (−1) ζ (t + j) ζ (q − j + 1) + (−1) EU (t + q)

where EU (m) is Euler’s identity given in the next lemma.

Mathematics 2019, 7, 143 5 of 20

Lemma 2. The following identities hold: for m ∈ N. The classical Euler identity, as given by Euler, states

EU (m) = ∑∞
n =1
Hn
nm = ( m2 + 1)ζ (m + 1)
(8)
−2
− 12 ∑m
j =1 ζ ( m − j ) ζ ( j + 1 ) .

The following identities are new. For p a positive even integer,

∞
Hn p
HE ( p) = ∑ p =
2
(ζ ( p + 1) + η ( p + 1)) − (ζ ( p) + η ( p)) ln 2
n =1 (2n + 1)
(9)
p
2 −1
1
−
2 ∑ (ζ ( p + 1 − 2j) + η ( p + 1 − 2j)) (ζ (2j) + η (2j)) .
j =1

For p a positive odd integer,

∞
Hn p
HO ( p) = ∑ p =
4
(ζ ( p + 1) + η ( p + 1)) − (ζ ( p) + η ( p)) ln 2
n =1 (2n + 1)
 p −1

    
1 1 + (−1) 2
 ζ p+1 +η p+1
−  (10)
4 2 2 2

b
1
−
2 ∑ (ζ ( p − 2j) + η ( p − 2j)) (ζ (2j + 1) + η (2j + 1))
j =1

p −1
!
h i
p −1 1+(−1) 2
where η (z) is the Dirichlet eta function, b = 4 − 2 , and [z] is the greatest integer less than

z. For p and t positive integers we have

(−1)n+1
F ( p, t) = ∑∞
n =1 n p ( n +1) t

!
p p −r p+t−r−1
= ∑ r =1 (−1) η (r )
p−r (11)
!
p +1 p+t−s−1
+ ∑ts =1 (−1) (1 − η (s)) ,
t−s

!
p+t−1
∑∞ 1 p +1
G ( p, t) = n =1 n p ( n +1) t = (−1)
p
!
p p −r p+t−r−1
+ ∑ r =2 (−1) ζ (r ) (12)
p−r
!
p p+t−s−1
+ ∑ts =1 (−1) ζ (s) ,
t−s
Mathematics 2019, 7, 143 6 of 20

!
(−1)n+1 Hn q + t − 2 1 
∑∞
t +1 2
H A (t, q) = = (−1)
n =1 n t ( n +1) q 2 ζ ( 2 ) − ln 2
t−1
!
q+t−r−1
+ ∑rt =2 (−1)t−r S (1, r ) (13)
t−r
!
q t q+t−s−1
+ ∑s =2 (−1) (η (s + 1) − S (1, s)) ,
q−s

and !
p+t−2
∑∞
Hn p +1
HG ( p, t) = = (−1)
n =1 n p ( n +1) t ζ (2)
p−1
!
p p+t−r−1
+ ∑r =2 (−1) p−r EU (r ) (14)
p−r
!
p+t−s−1
+ ∑ts =2 (−1) p ( EU (s) − ζ (s + 1)) .
t−s

Proof. The identity (8) is the Euler relation and by manipulation we arrive at (9) and (10). The results (9)
and (10) are closely related to those given by Nakamura and Tasaka [24]. For the proof of (11) we
notice that
p
!
1 p+t−r−1 1
t
= ∑ (−1) p −r
r
p
n ( n + 1) r =1 p − r n
!
t
p+t−s−1 1
+ ∑ (−1) p

s =1 t−s ( n + 1) s

therefore, summing over the integers n,

∞ p
!
(−1)n+1 p+t−r−1
F ( p, t) = ∑ t
= ∑ (−1) p−r η (r )
p
n =1 n ( n + 1 ) r =1 p−r
!
t
p+t−s−1
+ ∑ (−1) p
(1 − η (s))
s =1 t−s

and hence (11) follows. Consider

!
p
!
1 (−1) p+1 p+t−2 p+t−r−1 1
np ( n + 1) t
=
n ( n + 1) p−1
+ ∑ (−1) p −r
p−r nr
r =2
!
t
p+t−s−1 1
+ ∑ (−1) p
t−s ( n + 1) s
,
s =2

and summing over the integers n produces the result (12). The proof of (13) and (14) follows by
summing ∑∞
n =1 p
Hn
t in partial fraction form. Here is an example from (10):
n ( n +1)

∞
Hn 9207 961 2 889 511
∑ (2n + 1) 9
=
2048
ζ (10) −
1024
ζ (5) −
512
ζ (7) ζ (3) −
256
ζ (9) ln 2
n =1
Mathematics 2019, 7, 143 7 of 20

and from (9):

∞
Hn 381 93 105 63
∑ (2n + 1) 6
=
64
ζ (7) − ζ (5) ζ (2) −
64 64
ζ (4) ζ (3) − ζ (6) ln 2.
32
n =1

2. Summation Identity
We now prove the following theorems:

Theorem 1. For positive integers q and t, the integral of the product of two polylogarithmic functions with
negative arguments

(−,−) R1 R0
I0 (q, t) = Lit (− x ) Liq (− x ) dx = Lit ( x ) Liq ( x ) dx
0 −1
q −1 (15)
= (−1) j+1 η (q − j + 1) F (t, j)
∑ j =1
+ (−1)q ( F (t, q + 1) − ( F (t, q) − G (t, q)) ln 2) + (−1)q Wn (q, t)

where the sum

∞
!
1 1 1
Wn (q, t) = ∑ Hn t
(2n) (2n + 1) q
− t
n ( n + 1 ) q +
(2n + 1) (2n + 2)q
t
(16)
n =1

is obtained from (8), (9), (10), and the terms F (·, ·) , G (·, ·) are obtained from (11) and (12), respectively.

Proof. By the definition of the polylogarithmic function, we have

Z1 ∞ ∞
(−,−) (−1)n+r
I0 (q, t) = Lit (− x ) Liq (− x ) dx = ∑ ∑ n t r q ( n + r + 1)
0 n =1 r =1

∞ ∞ q
!
(−1)n+r (−1)q (−1) j+1
= ∑ ∑ + ∑
n =1 r =1
nt ( n + r + 1 ) ( n + 1 ) q j =1 ( n + 1 ) j r q − j +1

∞ q
!
(−1)n+r (−1)q+1 (−1) j+1 η (q − j + 1)
 
1 1
= ∑ H n +1 − H n + ∑
n =1
nt ( n + 1) q 2 2 2 2 j =1 ( n + 1) j

q −1 ∞ ∞
(−1)n+1 (−1)n+1
= ∑ (−1) j+1 η (q − j + 1) ∑ + (−1)q ∑ n t ( n + 1 ) q +1
j =1 n =1 n t ( n + 1) j n =1
∞
(−1)n+1
 
1 1
+ (−1) ∑ t q
q H n+1 − H n − ln 2 .
n =1 n ( n + 1 )
2 2 2 2

Now we utilize the double argument identity (3) together with (11), we obtain

q −1
(−,−)
I0 (q, t) = ∑ (−1) j+1 η (q − j + 1) F (t, j) + (−1)q F (t, q + 1)
j =1

∞
(−1)n+1  
+ (−1)q ∑ t q Hn − H n − 2 ln 2 ,
n =1 n ( n + 1 )
2
Mathematics 2019, 7, 143 8 of 20

and we can use the alternating harmonic number sum identity (5) to simplify the last sum. However,
we shall simplify further as follows:

q −1
(−,−)
I0 (q, t) = ∑ (−1) j+1 η (q − j + 1) F (t, j) + (−1)q F (t, q + 1)
j =1

∞
(−1)n+1    
+ (−1)q ∑ nt (n + 1)q (−1)n+1 H[ n2 ] − Hn − 1 + (−1)n ln 2


n =1

where [z] is the integer part of z. Now,

q −1
(−,−)
I0 (q, t) = ∑ (−1) j+1 η (q − j + 1) F (t, j) + (−1)q F (t, q + 1)
j =1

− (−1)q ( F (t, q) − G (t, q)) ln 2 + (−1)q Wn (q, t)

where
∞
!
1 1 1
Wn (q, t) = ∑ Hn t
(2n) (2n + 1) q
− t
n ( n + 1) q +
(2n + 1) (2n + 2)q
t
n =1

and the infinite positive harmonic number sums are easily obtainable from (8), (9), (10), hence the
identity (15) is achieved.

The next theorem investigates the integral of the product of polylogarithmic functions divided by
a linear function.

Theorem 2. Let (t, q) be positive integers, then for t + q an odd integer

Z1 0
Lit (− x ) Liq (− x ) Z
Lit ( x ) Liq ( x )
I (−,−) (q, t) = dx = − dx
x x
0 −1
q −1 (17)
= ∑ (−1) j+1 η (t + j) η (q − j + 1) + (−1)q+1 (ζ (t + q) + η (t + q)) ln 2
j =1

+ (−1)q+1 2−t−q − 1 EU (q + t) + (−1)q+1 HO (q + t) .

For t + q an even integer

q −1
I (−,−) (q, t) = ∑ j =1 (−1) j+1 η (t + j) η (q − j + 1)
+ (−1)q+1 (ζ (t + q) + η (t + q)) ln 2 + (−1)q+1 2−t−q − 1 EU (q + t) (18)

+ (−1)q+1 HE (q + t) .

Proof. Consider

Z1
(−,−) Lit (− x ) Liq (− x )
I (q, t) = dx
x
0
Z1
(−1)n
= ∑ x n−1 Liq (− x ) dx,
n ≥1
nt
0
Mathematics 2019, 7, 143 9 of 20

and successively integrating by parts leads to

q −1 Z1
(−1)n (−1)n+q+1
I (−,−) (q, t) = ∑ t+ j ∑ η ( q − j + 1) + ∑ n t + q −1 x n−1 Li1 (− x ) dx.
n ≥1 n j =1 n ≥1 0

Evaluating the inner integral,

Z1 Z1
x n−1 Li1 (− x ) dx = − x n−1 ln (1 + x ) dx
0 0
 
1 1 1
= H n − H n−1 − ln 2 ,
n 2 2 2 2
so that
q −1
(−1)n
I (−,−) (q, t) = ∑ t+ j ∑ (−1) j η (q − j + 1)
n ≥1 n j =1

(−1)n+q+1
 
1 1
+ ∑ nt+q H n − H n−1 − ln 2
2 2 2 2
n ≥1

q −1
= ∑ (−1) j+1 η (q − j + 1) η (t + j)
j =1

(−1)n+q+1
 
1 1
+∑ H n − H n−1 − ln 2 .
n ≥1 nt+q 2 2 2 2

If we now utilize the multiplication formula (3), we can write

q −1
(−1)n+1  
I (−,−) (q, t) = ∑ (−1) j+1 η (q − j + 1) η (t + j) + (−1)q+1 ∑ nt+q Hn − H n2 .
j =1 n ≥1

Now consider the harmonic number sum

1 − (−1)n ln 2


n +1 n +1
(−1)   (−1)
∑ Hn − H n = ∑
 
nt+q nt+q
2
   
n ≥1 n ≥1 n +1
+ (−1) H[ n ] − Hn
2

!
(−1)n+1 n
(−1) j
= ∑ 1 − (−1) n

ln 2 + (−1) n +1
∑ j
n ≥1 nt+q j =1

(−1)n+1
 
1 Hn Hn
= ∑ nt+q 1 − (− 1 ) n

ln 2 + ∑ 2t + q − 1
n t + q
+∑ t+q
n ≥1 n ≥1 n≥1 (2n + 1)
 
1 Hn Hn
= (ζ (t + q) + η (t + q)) ln 2 + ∑ t + q
−1 t + q
+∑ t+q
n ≥1 2 n n≥1 (2n + 1)
Mathematics 2019, 7, 143 10 of 20

where [z] is the integer part of z. Hence,

q −1
I (−,−) (q, t) = ∑ (−1) j+1 η (q − j + 1) η (t + j) + (−1)q+1 (ζ (t + q) + η (t + q)) ln 2
j =1
 
1 Hn Hn
+ (−1)q+1 ∑ 2t + q
−1
nt+q
+ (−1)q+1 ∑ t+q
n ≥1 n≥1 (2n + 1)

q −1
= ∑ (−1) j+1 η (q − j + 1) η (t + j) + (−1)q+1 (ζ (t + q) + η (t + q)) ln 2
j =1
 
1
+ (−1)q+1 − 1 EU (q + t)
2t + q

 HO (q + t) , for t + q odd

+ (−1)q+1 ,

 HE (q + t) , for t + q even

hence (17) and (18) follow.

Remark 1. It is interesting to note that for m ∈ R,

Z1 Z1
Lit (− x m ) Liq (− x m ) 1 Lit (− x ) Liq (− x )
dx = dx.
x m x
0 0

The next theorem investigates the integral of the product of polylogarithmic functions divided by
a quadratic factor.

Theorem 3. For positive integers q and t, the integral of the product of two polylogarithmic functions with
negative arguments

(−,−) R1 Lit (− x ) Liq (− x ) R0 Lit ( x ) Liq ( x )

I2 (q, t) = x2
dx = x2
dx
0 −1
q −1 (19)
= η (q + 1) + ∑ j =1 (−1) j η (q − j + 1) F ( j, t)
+ (−1)q ( F (q, t) + G (q, t)) ln 2 + (−1)q Wn (t, q)

where the sum

∞
!
1 1 1
Wn (t, q) = ∑ Hn
(2n)q (2n + 1)t
−
n q ( n + 1) t
+
(2n + 1)q (2n + 2)t
(20)
n =1

is obtained from (8), (9), (10), and the terms F (·, ·) , G (·, ·) are obtained from (11) and (12), respectively.

Proof. Follows the same process as in Theorem 2.

The following recurrence relation holds for the reduction of the integral of the product of
polylogarithmic functions multiplied by the power of its argument.
Mathematics 2019, 7, 143 11 of 20

Lemma 3. For (q, t) ∈ N and m ≥ 0, let

Z1 Z0
m m
J (m, q, t) = x Lit (− x ) Liq (− x ) dx = (−1) x m Lit ( x ) Liq ( x ) dx
0 −1

then
(m + 1) J (m, q, t) = η (q) η (t) − J (m, q, t − 1) − J (m, q − 1, t) .
For q = 1,

(m + 1) J (m, 1, t) = η (t) + mJ (m − 1, 1, t) + J (m − 1, 1, t − 1) − J (m, 1, t − 1)

−mK (m, t) − K (m, t − 1)

where
Z1
K (m, t) = x m Lit (− x ) dx.
0

Proof. The proof of the lemma follows in a straightforward manner after integration by parts.

Theorem 4. For positive integers q and t, the integral

Z1 Z0
(+,−)
I0 (q, t) = Liq ( x ) Lit (− x ) dx = Liq (− x ) Lit ( x ) dx
0 −1
q −1
= ∑ (−1) j ζ (q − j + 1) F (t, j)
j =1

+ (−1)q F (t, q + 1) + (−1)q H A (t, q) .

Let (q, t) be positive integers, then for t + q an even integer,

Z1 0
Liq ( x ) Lit (− x ) Z
Liq (− x ) Lit ( x )
I (+,−) (q, t) = dx = − dx
x x
0 −1
(21)
q −1
= ∑ (−1) j η (t + j) ζ (q − j + 1) + (−1)q S (1, q + t) ,
j =1

Z1 0
(−,+) Liq (− x ) Lit ( x ) Z
Liq ( x ) Lit (− x )
I (q, t) = dx = − dx
x x
0 −1
(22)
t −1
= ∑ (−1) j η (q + j) ζ (t − j + 1) + (−1)t S (1, q + t)
j =1
Mathematics 2019, 7, 143 12 of 20

For positive integers q and t,

Z1 Z0
(+,−) Liq ( x ) Lit (− x ) Liq (− x ) Lit ( x )
I2 (q, t) = 2
dx = dx
x x2
0 −1
q −1
= − ζ ( q + 1) + ∑ (−1) j+1 ζ (q − j + 1) F ( j, t)
j =1

+ (−1)q+1 H A (q, t) ,

where the terms F (·, ·) , H A (·, ·) and S (·, ·) are obtained from (11) (13) and (5), respectively.

Proof. The proof is similar to the one outlined in the previous theorems.

Theorem 5. For positive integers q and t, the integral of the product of two polylogarithmic functions,

R1 Liq ( x ) Lit ( x2 )
U (+,+) (q, t) = x dx
0
(23)
 
q −1
= 2t−1 ∑ j =1 (−1) j+1 ζ (t + j) ζ (q − j + 1) + (−1)q+1 EU (t + q)
 
q −1
+2t−1 ∑ j =1 (−1) j η (t + j) ζ (q − j + 1) + (−1)q S (1, q + t)

where the terms EU (·, ·) , S (·, ·) are obtained from (8) and (5), respectively.

Proof. By the definition of the polylogarithmic function, we have

Z1 ∞ Z1
Liq ( x ) Lit x2


1
U (+,+)
(q, t) = dx = ∑ t x2n−1 Liq ( x ) dx.
x n =1
n
0 0

Successively integrating by parts leads to

q −1
U (+,+) (q, t) = ∑ j =1 (−1) j+1 ζ (q − j + 1) ∑n≥1 t 1 j
n (2n)
1
(−1)q+1 R 2n−1
+ ∑ n ≥1 t q −1 x Li1 ( x ) dx,
n (2n)
0 (24)
q −1 (−1)q+1 H2n
= ∑ j =1 (−1) j+1 ζ (q − j + 1) ∑n≥1 t
1
j + ∑ n ≥1 nt (2n)q
n (2n)
q−1 (−1) j+1 (−1)q+1
= ∑ j =1 2j
ζ ( q − j + 1) ζ ( t + j) + 2q ∑n≥1 nHt+2nq .

Alternatively, we have the relation

 
Lit x2 = 2t−1 (Lit ( x ) + Lit (− x )) , (25)

from which we obtain

R1 Liq ( x ) Lit ( x2 ) R1 Liq ( x ) (Lit ( x )+Lit (− x ))

x dx = 2t −1 x dx
0 0
 
= 2t−1 I (+,+) (q, t) + I (+,−) (q, t) (26)
 
q −1
= 2t−1 ∑ j =1 (−1) j+1 ζ (t + j) ζ (q − j + 1) + (−1)q+1 EU (t + q)
 
q −1
+2t−1 ∑ j =1 (−1) j η (t + j) ζ (q − j + 1) + (−1)q S (1, q + t) ,
Mathematics 2019, 7, 143 13 of 20

hence the identity (23) is achieved. An alternative expansion of the integral

Z1
Liq ( x ) Lit x2


1 1
U (+,+)
(q, t) = dx = ∑ q ∑ rt (n + 2r) .
x n ≥1
n r ≥1
0

By partial fraction expansion,

Hn
U (+,+) (q, t) = ∑ n ≥1 1
nq ∑ r ≥1 1
r t (n+2r )
= (−1)t−1 2t−1 ∑n≥1 2
nt+q
(27)
+ ∑tj −=11 (−1) j+1 2 j−1 ζ ( q + j ) ζ ( t − j + 1) .

Remark 2. We note that from (23) and (24), we are able to isolate a new Euler identity of weight q + t + 1,
in the form
Han
W ( a, k ) = ∑ k .
n ≥1 n

For a = 2, k = q + t
 
W (2, q + t) = ∑ n ≥1 H2n
nt+q
= (−1)q+1 2q+t−1 I (+,+) (q, t) + I (+,−) (q, t)
(28)
q −1
− ∑ j =1 (−1) j+q 2q− j ζ ( q − j + 1) ζ ( t + j ) .

Similarly, from (26) and (27), we obtain, for a = 21 , k = q + t, another new Euler identity
  Hn  
W 1
2, q + t = ∑ n ≥1 2
nt+q
= (−1)t+1 I (+,+) (q, t) + I (+,−) (q, t)
(29)
t j +1 j −1
+ (−1) ∑tj − 1
=1 (−1) 2 ζ ( q + j ) ζ ( t − j + 1) .

The difference of (28) and (29) delivers the delightful identity

q −1
(−1)q (−1) j
 
1
∑ nt+q
(−1)t 2t−1 H n2 −
2q
H2n = ∑ j ζ ( q − j + 1) ζ ( t + j )
n ≥1 j =1 2

t −1
+ ∑ (−1) j+1 2j−1 ζ (q + j) ζ (t − j + 1) .
j =1

The next theorem investigates the integral of the product of polylogarithmic functions with one
positive and one negative argument.

Theorem 6. Let (q, t) be positive integers, then for t + q an even integer,

R1 Liq (− x ) Lit ( x2 ) R0 Liq ( x ) Lit ( x2 )

U (−,+) (q, t) = x dx = x dx
0  −1 
= 2t−1 ∑tj −=11 (−1) j η (q + j) ζ (t − j + 1) + S (1, q + t)
 
q −1
+2t−1 ∑ j =1 (−1) j+1 η (t + j) η (q − j + 1) + (−1)q+1 (ζ (t + q) + η (t + q)) ln 2 (30)
  

 HO (q + t) , for q + t odd
+2t−1 (−1)q+1 2−t−q − 1 EU (t + q) + (−1)q+1


.
 

 HE (q + t) , for q + t even
Mathematics 2019, 7, 143 14 of 20

Proof. From (25), we have

R1 Liq (− x ) Lit ( x2 ) R1 Liq (− x ) (Lit ( x )+Lit (− x ))

U (−,+) (q, t) = x dx = 2t −1 x dx
0  0  (31)
= 2t−1 I (−,+) (q, t) + I (−,−) (q, t) ,

replacing I (−,+) (q, t), and I (−,−) (q, t), with (22) and (17), respectively, leads to (17), as required. We can
also consider

Z1
Liq (− x ) Lit x2


(−,+)
U (q, t) = dx
x
0
Z1
(−1)n  
= ∑ x n−1 Lit x2 dx,
n ≥1
nq
0

and successively integrating by parts, as in Theorem 1, leads to

t −1 (−1)n+1 H n2
U (−,+) (q, t) = ∑ (−1) j ζ (t − j + 1) η (q + j) + (−1)t ∑ nq+t
. (32)
j =1 n ≥1

Another expansion of

Z1 Z1
Liq (− x ) Lit x2 (−1)n x n+2r−1


U (−,+)
(q, t) =
x
dx = ∑ ∑ nq r t
dx
0 r ≥1 n ≥1 0
n
(−1)
= ∑∑ n q r t ( n + 2r ) .
r ≥1 n ≥1

Partial fraction expansion, and summing

q −1
U (−,+) (q, t) = (−1)q ∑ (−1) j 2j−q η (1 + j) ζ (q + t − j)
j =1

(−1)q+1
 
Hn H
+
2q ∑ nq+t
− q2n
n +t
,
n ≥1

q −1
(−1)q+1
= (−1)q ∑ (−1) j 2 j−q η (1 + j) ζ (q + t − j) +
2q
( EU (q + t) + W (2, q + t)) .
j =1

hence (17) follows. We note that from (18) and (32), we obtain another new Euler sum, namely

(−1)n+1 H n  
∑ n ≥1 nq+t
2
= (−1)t I (−,+) (q, t) + I (−,−) (q, t)
(33)
+ (−1)t ∑tj −=11 (−1) j+1 2 j−t ζ (t − j + 1) η (q + j) .

Example 1. Some examples follow.

Z1
Li3 ( x ) Li5 x2


(+,+) 521 1
U (3, 5) = dx = ζ (9) − 8ζ (2) ζ (7) − 2ζ (4) ζ (5) − ζ (6) ζ (3) .
x 32 2
0
Mathematics 2019, 7, 143 15 of 20

Z1
Li4 (− x ) Li4 x2


(−,+) 501 11
U (4, 4) = dx = ζ (9) − 4ζ (2) ζ (7) − ζ (4) ζ (5) .
x 64 8
0

H2n 2059
W (2, 6 + 4) = ∑ n10
=
4
ζ (11) − 256ζ (9) ζ (2)
n ≥1

−4ζ (8) ζ (3) − 64ζ (7) ζ (4) − 16ζ (6) ζ (5) .

H2n
W (2, 2 + 1) = ∑ n3
= 5ζ (4) − 4L (3)
n ≥1

where  
11 7 1 2 1 4 1
L (3) = ζ (4) − ζ (3) ln 2 + ζ (2) ln 2 − ln 2 − 2Li4 .
4 4 2 12 2
 
1 1 17 517
∑ n9 2 H 2 + 25 H2n = 4 ζ (3) ζ (7) + ζ 2 (5) − 32 ζ (10) .
3 n
n ≥1

Hn
 
1 521 1
W
2
,5+3 = ∑ n8
2
=
512
ζ (9) − ζ (2) ζ (7)
64
n ≥1

1 1
− ζ (4) ζ (5) − ζ (3) ζ (6) .
16 4

(−1)n+1 H n2 49 31 635
∑ n6
=
32
ζ (4) ζ (3) + ζ (5) ζ (2) −
32 128
ζ (7) .
n ≥1

The next two results deal with Polylog integrals involving the inverse of the argument.

Theorem 7. For positive integers q and t, the integral of the product of two polylogarithmic functions,

R1 Liq (− 1x ) Lit ( x2 )
I3 (q, t) = x dx
0
(34)
= − 2q1+1 ζ (q + t + 1) + 1
( EU (q + t) − W (2, q + t))
2q
q −1 1
− ∑ j =1 2j
ζ ( q + t ) η ( q − j + 1)

where the terms EU (·) , W (·, ·) are obtained from (8) and (28), respectively. For positive integers q and t,
the integral of the product of two polylogarithmic functions,
 
R1 Liq 1
x2
Lit ( x )
I4 (q, t) = x dx
0
(35)
 
= − 2q ζ ( q + t + 1 ) + 2q −1 W 1
2, q + t
q −1
+ ∑ j =1 2 j−1 ζ ( j + t) ζ (q − j + 1) − 2q−1 iπζ (q + t)

1
 √
where W 2, q + t is obtained from (29), and i = −1.

Proof. Following the same process as in Theorem 2, we have

 
Z1 Liq − 1 Lit x2 Z1

 
x 1 1
I3 (q, t) =
x
dx = ∑ nt x2n−1 Liq −
x
dx.
0 n ≥1 0
Mathematics 2019, 7, 143 16 of 20

Integrating by parts, as in Theorem 2, we have

q −1 Z1  
1 1 1
I3 (q, t) = − ∑ j η ( q − j + 1) ζ ( t + j ) + ∑ q −1
x 2n−1
Li1 − dx
j =1 2
t
n≥1 n (2n )
x
0
q −1  
1 1 1 1 1 1
= − ∑ 2 j η ( q − j + 1 ) ζ ( t + j ) − 2q ζ ( t + q ) + 2q ∑ nt+q 2
Hn − Hn− 1 − ln 2 ,
2 2
j =1 n ≥1

from the multiplication Formula (3), we can simplify so that

q −1
1 1 1 1
I3 (q, t) = − ∑ j
η ( q − j + 1) ζ ( t + j ) − q ζ ( t + q ) − q
2 2 ∑ nt+q
( H2n − Hn )
j =1 2 n ≥1

and the proof of the first part of the theorem is finalized. The second result is proved as in Theorem 2.

Example 2.
 
Z1 Li4 − 1 Li2 x2


x 119 7
I3 (4, 2) = dx = ζ (7) − ζ (2) ζ (5) − ζ (4) ζ (3) .
x 64 16
0  
Z1 Li3 1
Li5 ( x )
x2 31 503
I4 (3, 5) = dx = ζ (2) ζ (7) − ζ (9)
x 16 128
0
1
− ζ (4) ζ (5) − iπζ (8) .
4

3. Some Extensions
Some other valuable research on the representation of integrals of polylogarithmic functions with
a trigonometric argument has been carried out, and we highlight the following results: Espinosa and
Moll [25,26] give for p ∈ N0
h i
p +1
Z1 2
!
p+1 (2j)!
( p + 1) p
x ln (sin (πx )) dx = − ln 2 + ∑ (−1) j
ζ (2j + 1)
0 j =1
2j (2π )2j

where [m] is the integer part of m. For a non-trigonometric argument, the work of Espinosa and Moll
may be extended in the following way: For ( p, q) ∈ N, let

Z1  q
ln x Li p ( x )
M ( p, q) = dx.
1−x x
0

Rewriting !
1 n+q−1
= ∑ xn ,
(1 − x ) q n ≥0 n

then after integration and simplification

!
(−1)q n+q−1 1
M ( p, q) = ζ ( p + q + 1) + ∑ ∑ ,
q! j ≥1 n ≥1
n j p ( j + n ) q +1
Mathematics 2019, 7, 143 17 of 20

which results in the product of zeta functions. For the case of q = 2, we may write

Z1  2
ln x Li p ( x )
M ( p, 2) = dx = 2ζ ( p + 3) − 2ζ ( p + 3) ζ ( p − 1)
1−x x
0

+2ζ ( p) (ζ (2) + ζ (3)) − 2BW (2, p) − 2BW (3, p) + 2BW (3, p − 1) ,

where BW (·, ·) is given by (6). For p = 4, we find

Z1  2
ln x Li4 ( x ) 31
M (4, 2) = dx = ζ (6) − 3ζ 2 (3)
1−x x 16
0

+20ζ (2) ζ (5) + 2ζ (3) ζ (4) − 34ζ (7) .

Choi [27,28] has also produced many interesting examples of log-sine and log-cosine integrals.
Mezo [29] continues the study and considers the class of integrals

Z1 Z1
Ip = Li p (sin πx ) dx and J p = Li p (cos πx ) dx,
0 0

such that !
1 2j 1
Jp = p
2 ∑ j 4j j p
j ≥1

and
2 (2j − 2)!! 1
Ip = Jp +
π ∑ ( 2j − 1 ) !! (2j − 1) p
,
j ≥1

where 

 1, for m = −1, 0
m!! = m (m − 2) . . . 6.4.2, for m > 0 even .
m (m − 2) . . . 5.3.1, for m > 0 odd



Mezo [29] gives the following examples:

Z1
π3
 
π 2
Li3 (sec πx ) dx = i ln 2 +
4 48
0

and
Z1 Z1
    4
2
Li3 sin πx dx = Li cos2 πx dx = 2ζ (3) − 2ζ (2) ln 2 + ln3 2. (36)
3
0 0

It is possible to generalize (36), so that for m ∈ N

Z1
!
  1 2mn
Li p sin 2m
πx dx = ∑ 22mn n p mn
,
0 n ≥1

and
Z1   9 2
Li4 sin2 πx dx = ζ (4) + 2ζ (2) ln2 2 − 4ζ (3) ln 2 − ln4 2.
4 3
0
Mathematics 2019, 7, 143 18 of 20

There are a number of extensions that may be very interesting to investigate further in regard to the
representation of integrals of polylogarithmic functions. It would be beneficial to study the integrals

Z1 Z1  a
arcsin x  
(sin πx ) a Libp (sin πx ) dx, Libp x2 dx
x
0 0

where, for instance

1
Z1 Z4
 πx   
π sin πx Li4 tan4 dx = 4π sin 4πy Li4 tan4 πy dy
4
0 0
= 6ζ (4) + 32ζ (2) ln 2 − 32ζ (3) ,

Z1
π 1 π
sin πx Li4 (sin πx ) dx = − 1 + ζ (2) + (2 + ln 2) ln 2
2 8 4
0

1 2n n!
+
4π ∑ (n + 1)3 (2n + 3)!!
,
n ≥0

Z1
!
arcsin x   π π 1 2n
x
Li p x2 dx = ζ ( p + 1) −
4 4 ∑ 22n n p+1 n
,
0 n ≥1

and

Z1
1 arcsin x   9 1 1 5
Li4 x2 dx = ζ (4) ln 2 + ζ (2) ln3 2 − ln 2
π x 8 3 15
0

1   5
+ ζ (3) ζ (2) − 2 ln2 2 − ζ (5) .
2 4
We may also study
Z1
Liap (sin πx ) Libq (cos πx ) dx,
0

and the moments

Z1
x a Libp (sin πx ) dx,
0

Z1
x a Libp (arcsin x ) dx,
0

where ( a, b, p, q) ∈ N.
In this paper we have highlighted various new identities of general Euler type sums, which
represent the analytical solution of integrals with polylogarithm functions containing linear, quadratic,
and trigonometric type arguments. In a series of papers [7,30,31], the authors explore linear combinations
of associated harmonic polylogarithms and nested harmonic numbers. The multiple zeta value data mine,
computed by Blumlein et al. [32], is an invaluable tool for the evaluation of harmonic numbers, in which
values with weights of twelve, for alternating sums, and weights above twenty for non-alternating sums
are presented. Further areas of fruitful research are those related to trigonometric, inverse trigonometric,
and linear arguments for products of integrals of polylogarithmic functions.
Mathematics 2019, 7, 143 19 of 20

Funding: This research received no external funding.

Conflicts of Interest: The author declares no conflict of interest.

References
1. Freitas, P. Integrals of polylogarithmic functions, recurrence relations, and associated Euler sums.
Math. Comput. 2005, 74, 1425–1440. [CrossRef]
2. Devoto, A.; Duke, D.W. Table of integrals and formulae for Feynman diagram calculations.
Riv. Nuovo Cimento 1984, 7, 1–39. [CrossRef]
3. Kölbig, K.S. Nielsen’s generalized polylogarithms. SIAM J. Math. Anal. 1986, 17, 1232–1258.
4. Lewin, R. Polylogarithms and Associated Functions; Elsevier Science Ltd.: North Holland, The Netherland;
New York, NY, USA, 1981.
5. Ablinger, J.; Blümlein, J.; Schneider, C. Analytic and algorithmic aspects of generalized harmonic sums and
polylogarithms. J. Math. Phys. 2013, 54. [CrossRef]
6. Chavez, F.; Duhr, C. Three-mass triangle integrals and single-valued polylogarithms. J. High Energy Phys.
2012. . [CrossRef]
7. Blümlein, J. Structural Relations of Harmonic Sums and Mellin Transforms at Weight w = 6. Motives, Quantum
Field Theory, And Pseudodifferential Operators; American Mathematical Society: Providence, RI, USA, 2010;
pp. 167–187.
8. Hoorfar, A.; Qi, F. Sums of series of Rogers dilogarithm functions. Ramanujan J. 2009, 18, 231–238. [CrossRef]
9. Flajolet, P.; Salvy, B. Euler sums and contour integral representations. Exp. Math. 1998, 7, 15–35. [CrossRef]
10. Borwein, D.; Borwein, J.M.; Girgensohn, R. Explicit evaluation of Euler sums. Proc. Edinburgh Math. Soc.
1995, 38, 277–294. [CrossRef]
11. Xu, C.; Yan, Y.; Shi, Z. Euler sums and integrals of polylogarithm functions. J. Number Theory 2016, 165,
84–108. [CrossRef]
12. Ablinger, J.; Blümlein, J. Harmonic sums, polylogarithms, special numbers, and their generalizations.
In Computer Algebra in Quantum Field Theory; Texts & Monographs in Symbolic Computation; Springer:
Vienna, Austria, 2013; pp. 1–32.
13. Bailey, D.H.; Borwein, J.M. Computation and structure of character polylogarithms with applications to
character Mordell-Tornheim-Witten sums. Math. Comput. 2016, 85, 295–324. [CrossRef]
14. Borwein, D.; Borwein, J.M.; Bradley, D.M. Parametric Euler sum identities. J. Math. Anal. Appl. 2006, 316,
328–338. [CrossRef]
15. Borwein, J.M.; Bradley, D.M.; Broadhurst, D.J.; Lisoněk, P. Special values of multiple polylogarithms.
Trans. Am. Math. Soc. 2001, 353, 907–941. [CrossRef]
16. Borwein, J.M.; Zucker, I.J.; Boersma, J. The evaluation of character Euler double sums. Ramanujan J. 2008, 15,
377–405. [CrossRef]
17. Davydychev, A.I.; Kalmykov, M.Y. Massive Feynman diagrams and inverse binomial sums. Nuclear Phys. B
2004, 699, 3–64. [CrossRef]
18. Kalmykov, M.Y.; Veretin, O. Single-scale diagrams and multiple binomial sums. Phys. Lett. B 2000, 483,
315–323. [CrossRef]
19. Sofo, A. Polylogarithmic connections with Euler sums. Sarajevo J. Math. 2016, 12, 17–32. [CrossRef]
20. Sofo, A. Integrals of logarithmic and hypergeometric functions. Commun. Math. 2016, 24, 7–22. [CrossRef]
21. Sofo, A. Quadratic alternating harmonic number sums. J. Number Theory 2015, 154, 144–159. [CrossRef]
22. Sofo, A. Integrals of polylogarithmic functions with negative argument. Acta Univ. Sapientiae Math. 2018, 10,
347–367.
23. Sofo, A. Integral identities for sums. Math. Commun. 2008, 13, 303–309.
24. Nakamura, T.; Tasaka, K. Remarks on double zeta values of level 2. J. Number Theory 2013, 133, 48–54.
[CrossRef]
25. Espinosa, O.; Moll, V. On Some Integrals Involving the Hurwitz Zeta Function: Part 2. Ramanujan J. 2002, 6,
440–468.
26. Espinosa, O.; Moll, V. On Some Integrals Involving the Hurwitz Zeta Function: Part 1. Ramanujan J. 2002, 6,
159–188. [CrossRef]
27. Choi, J. Log-sine and log-cosine integrals. Honam Math. J. 2013, 35, 137–146. [CrossRef]
Mathematics 2019, 7, 143 20 of 20

28. Choi, J.; Srivastava, H.M. The Clausen function Cl2 ( x ) and its related integrals. Thai J. Math. 2014, 12,
251–264.
29. Mezo, I. A family of polylog-trigonometric integrals. Ramanujan J. 2018, 46, 161–171. [CrossRef]
30. Ablinger, J.; Blümlein, J.; Schneider, C. Harmonic sums and polylogarithms generated by cyclotomic
polynomials. J. Math. Phys. 2011, 52, 102301. [CrossRef]
31. Blümlein, J. Algebraic relations between harmonic sums and associated quantities. Comput. Phys. Commun.
2004, 159, 19–54.
32. Blümlein, J.; Broadhurst, D.J.; Vermaseren, J.A.M. The multiple zeta value data mine. Comput. Phys. Commun.
2010, 181, 582–625.

c 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).