Tables of convolution integrals

A.B. Arbuzov
Bogoliubov Laboratory for Theoretical Physics,
JINR, Dubna, 141980, Russia
e-mail: [email protected]
arXiv:hep-ph/0304063v1 7 Apr 2003

Abstract

An analytical approach to convolution of functions, which appear in perturbative

calculations, is discussed. An extended list of integrals is presented.

1 Introduction

There are many situations, where one can describe a certain probability distribution of
a complicated process in a form of a conditional probability involving two or more sub
processes. Typically in quantum physics, a factorization of subprocesses occurs due to
the presence of a small (or large) parameter, which allows to suppress the interference
of amplitudes, describing different subprocesses. We will consider the dependence on
one continuous variable 0 x 1, and assume that the conditional probability can be
presented as a convolution of the corresponding distributions for the subprocesses. In
particular, convolution appear in the socalled evolution equations, arising in the renor-
malization group approach.

Quite often the convolution is performed by using the Mellin transformation. This ap-
proach is very transparent and powerful. On the other hand, it requires more steps (direct
and inverse transformations) and involves a considerable number of auxiliary functions in
the moment space. Moreover, in a realistic application, one might be interested to change
the limits of convolution integrals to separate a certain contribution with a particular
physical meaning.

Here I am going to discuss the direct analytical convolution, which is known to work
well with a rather wide class of functions, which appear in perturbative calculations. The
paper is organized as follows. In the next section I introduce the notation. The tables
of convolution integrals of singular and nonsingular functions are given in Sect. 3 and
Sect. 4, respectively. Possible applications of the Tables are discussed in Conclusions.
Properties of polylogarithmic functions are sketched in Appendixes.

Preprint submitted to Elsevier Science 7 February 2008

2 Preliminaries

Let us consider two functions f (x) and g(y), defined on the interval 0 x, y 1. Their
convolution is given by

Z1 Z1 Z1
dx z
 
[f g](z) = dx dy (z xy)f (x)g(y) = f (x)g , 0 z 1. (1)
z
x x
0 0

Sometimes we need to perform a convolution of special functions. In particular, func-

tions with the socalled plus prescription can be used. The prescription regularizes a pole
singularity at x = 1. It is defined as follows:

Z1 Z1
dx [f (x)]+ g(x) = dx f (x)[g(x)(x xmin ) g(1)], (2)
xmin 0

1

for x 0
(x) = , 0 xmin < 1.
0 for x < 0

Integrals of functions, which have poles at x = z or x = 1, are divergent. They can be

regularized by introducing a small auxiliary parameter 1. In the final result of a
particular calculation, one has than look for the cancellation of the parameter between
different contributions. The parameter can get also a physical meaning as a separator
between soft and hard radiation. There is a onetoone correspondence between the -
regularization and the plus prescription. In fact, the following definition is equivalent to
the prescription (2):

[f (x)]+ = lim [(1 x)f + (1 x)f (x)], (3)

0
1
Z 

f = dx f (x), f (x) = f (x) .
x<1
0

We will call f and f (x) as the -part and the -part of the special function f (x).
The above realization of the plus prescription is convenient in analytical calculations and
especially in numerical computations, where the direct implementation of Eq. (2) is prob-
lematic. In what follows we will systematically use -regularization for divergent integrals,
keeping in mind that the cancellation of the parameter will happen after summing with
the corresponding contribution of the -part of the relevant functions as in Eq. (3).

To define the -part for a given function, one requires a table of definite integrals over the

2
interval 0 < x < 1 or 0 < x < 1 for nonsingular functions. They can be found in
numerous sources (see, for instance, Ref. [3]).

A convolution of two singular functions regularized by the plus prescription can be rep-
resented as

1
dx z
   Z   
[f ]+ [g]+ (z) = lim f (x)g + f g (z) + f (z)g . (4)
0 x x
z/(1)

3 Integrals of singular functions

1
lnn (1 x) 1 1
Z
dx = lnn+1 (1 z) lnn+1 , n = 0, 1, 2, . . . (5)
1x n+1 n+1
z/(1)
1
1
Z
dx = ln + ln(1 z) ln z, (6)
xz
z/(1)
1
ln x 1
Z
dx = ln ln z + Li2 (1 z) + ln(1 z) ln z ln2 z, (7)
xz 2
z/(1)
1
Z
ln(1 x)
dx = ln ln(1 z) ln(1 z) ln z + ln2 (1 z) (2), (8)
xz
z/(1)
1
ln2 x
Z
dx = ln ln2 z + 2 S1,2 (1 z) + 2 Li2 (1 z) ln z
xz
z/(1)
1 3
+ ln(1 z) ln2 z ln z, (9)
3
1
ln2 (1 x)
Z
dx = ln ln2 (1 z) + ln3 (1 z) ln2 (1 z) ln z
xz
z/(1)

2(2) ln(1 z) + 2(3), (10)

1
ln(1 x) ln x
Z
dx = ln ln(1 z) ln z + 2 S1,2 (1 z) Li3 (1 z)
xz
z/(1)

+ Li2 (1 z) ln(1 z) + Li2 (1 z) ln z + ln2 (1 z) ln z

1
ln(1 z) ln2 z (2) ln z, (11)
2

3
 1
Li2 (1 x)
Z
dx = Li2 (1 z) (ln(1 z) ln z ln )
xz
z/(1)

S1,2 (1 z) Li3 (1 z) , (12)

1
Z
Li3 (1 x)
dx = Li3 (1 z) (ln(1 z) ln z ln )
xz
z/(1)

S2,2 (1 z) Li4 (1 z) , (13)

1
S1,2 (1 x)
Z
dx = S1,2 (1 z) (ln(1 z) ln z ln ) + S2,2 (1 z)
xz
z/(1)
1
2 S1,3 (1 z) (Li2 (1 z))2 , (14)
2
1
Z
Li2 (1 x) ln(1 x)
dx = Li2 (1 z) ln(1 z)(ln(1 z) ln z ln )
xz
z/(1)
1
+ Li4 (1 z) 2 S2,2 (1 z) + (Li2 (1 z))2 Li3 (1 z) ln(1 z)
2
S1,2 (1 z) ln(1 z) (2) Li2 (1 z) , (15)
1
Li2 (1 x) ln x
Z
dx = Li2 (1 z) ln z(ln(1 z) ln z ln )
xz
z/(1)

+4 S1,3 (1 z) 4 S2,2 (1 z) + (Li2 (1 z))2 Li3 (1 z) ln z

1
+ S1,2 (1 z) ln z + Li2 (1 z) ln2 z, (16)
2
1
ln2 (1 x) ln x
Z
dx = ln ln2 (1 z) ln z + 2 Li4 (1 z)
xz
z/(1)

(Li2 (1 z))2 2 Li3 (1 z) (ln(1 z) + ln z) + 4 S1,2 (1 z) ln(1 z)

+2 Li2 (1 z) ln(1 z) ln z + Li2 (1 z) ln2 (1 z) + ln3 (1 z) ln z
1
ln2 (1 z) ln2 z 2(2) ln(1 z) ln z + 2(3) ln z, (17)
2
1
ln(1 x) ln2 x
Z
dx = ln ln(1 z) ln2 z + 2 S2,2 (1 z) (Li2 (1 z))2
xz
z/(1)

2 Li3 (1 z) ln z + 2 S1,2 (1 z) (ln(1 z) + ln z)

+2 Li2 (1 z) ln(1 z) ln z + Li2 (1 z) ln2 z + ln2 (1 z) ln2 z
1
ln(1 z) ln3 z (2) ln2 z, (18)
3
1
ln3 (1 x)
Z
dx = ln3 (1 z)(ln(1 z) ln z ln ) 3(2) ln2 (1 z)
xz
z/(1)

4
 +6(3) ln(1 z) 6(4), (19)
1
ln3 x
Z
dx = ln ln3 z + 6 S1,3 (1 z) + 6 S1,2 (1 z) ln z
xz
z/(1)
1 4
+3 Li2 (1 z) ln2 z + ln(1 z) ln3 z ln z. (20)
4

In the above integrals I omitted terms, which are vanishing in the limit 0.

4 Integrals of nonsingular functions

Z1
1
 
dx xn = 1 z n+1 , n 6= 1, (21)
z
n+1
Z1
lnn x 1
dx = lnn+1 z, n = 0, 1, 2, . . . (22)
z
x n+1
Z1
lnn (1 x)
dx = (1)n n! [(n + 1) S1,n (z)], n = 1, 2, 3, . . . (23)
z
x
Z1
ln(1 x) ln x 1
dx = S1,2 (1 z) ln(1 z) ln2 z, (24)
z
x 2
Z1
z n+1 1
 
n
dx x ln x = ln z 2
1 z n+1 , n = 0, 1, 2, . . . (25)
z
n+1 (n + 1)
Z1
ln x 1
 
dx n = n1 (n 1) ln z + 1 z n1 , n = 2, 3, 4, . . . (26)
z
x z (n 1)2
Z1
lnn x
dx = (1)n n! S1,n (1 z) , n = 1, 2, 3, . . . (27)
z
1x
Z1
ln(1 x) ln x
dx = Li3 (1 z) Li2 (1 z) ln(1 z), (28)
z
1x
Z1 n+1
1 z n+1 1 X zk 
n
dx x ln(1 x) = ln(1 z) + S1 (n + 1) ,
z
n+1 n + 1 k=1 k
n = 0, 1, 2, . . . (29)
Z1
ln(1 x) 1z
dx 2
= ln z + ln(1 z), (30)
z
x z

5
Z1 n2
ln(1 x) 1 1 z n1 X 1
 
dx n
= ln z + n1
ln(1 z) k
+ S1 (n 2) ,
z
x n1 z k=1 kz
n = 3, 4, 5, . . . (31)
Z1
2 2z n+1 z n+1 2
dx xn ln2 x = (1 z n+1
) + ln z ln z, n 6= 1, (32)
z
(n + 1)3 (n + 1)2 n+1
Z1 n
1 z n+1 2 (1 z)k+1
!
n 2

n 2
(1)k
X
dx x ln (1 x) = ln (1 z) + 2
z
n+1 k=0 k (k + 1) k+1

2 ln(1 z) , n = 0, 1, 2, . . . (33)
Z1
1 z n+1
dx xn ln(1 x) ln x = Li2 (1 z) ln(1 z) ln z
z
n+1 n+1
n+1
1 z n+1 1 X zk
ln(1 z) + ln z
(n + 1)2 n+1 k=1 k
n+1
1 X
k n+k+1
+ (1 z ) , n = 0, 1, 2, . . . (34)
(n + 1)2 k=1 k2
Z1
ln x ln(1 x) 1 1
dx = Li2 (1 z) + ln z ln(1 z)
z
xn n1 z n1 (n 1)
1 1 n2
X 1 1 z n1
 
+ ln z + ln(1 z)
(n 1)2 n 1 k=1 z k k z n1 (n 1)2
n2
X (1 z k )(n + k 1)
1 1
+ ln2 z , n = 2, 3, 4, . . . (35)
2(n 1) (n 1)2 k=1 zk k2
Z1
ln2 (1 x) 1z 2
dx 2
= ln (1 z) + 2 Li2 (1 z) + 2 ln(1 z) ln z, (36)
z
x z
Z1
ln2 (1 x) 1 z n1 2
 
dx n
= n1
ln2 (1 z) + Li2 (1 z) + ln(1 z) ln z
z
x (n 1)z n1
Z1
2 n1
X ln(1 x)
dx n = 3, 4, 5, . . . (37)
n 1 k=2 z xk

Z1
dx ln2 (1 x) ln x = 2 Li3 (1 z) 2 Li2 (1 z) ln(1 z)
z
z ln2 (1 z) ln z + 2 Li2 (1 z) + 2z ln(1 z) ln z (1 z) ln2 (1 z)
+4(1 z) ln(1 z) 2z ln z + 6z 6, (38)

6
 Z1
ln2 (1 x) ln x 1
dx = 2 S1,2 (1 z) ln(1 z) 2 S2,2 (1 z) ln2 (1 z) ln2 z,(39)
z
x 2
Z1
ln2 (1 x) ln x
dx = 2 S1,2 (1 z) 2 Li3 (1 z) + 2 Li2 (1 z) ln(1 z)
z
x2
1
+ ln2 (1 z) ln z + ln(1 z) ln2 z + 2 Li2 (1 z) + 2 ln(1 z) ln z
z
1z 2
+ ln (1 z), (40)
z
Z1
ln2 (1 x) ln x
dx = 2 Li4 (1 z) + 2 Li3 (1 z) ln(1 z)
z
1x
Li2 (1 z) ln2 (1 z), (41)
Z1
dx ln(1 x) ln2 x = 2 S1,2 (1 z) z ln(1 z) ln2 z + 2 Li2 (1 z)
z
+2z ln(1 z) ln z + z ln2 z + 2(1 z) ln(1 z) 4z ln z 6 + 6z, (42)
Z1
ln(1 x) ln2 x 1
dx = 2 S1,3 (1 z) ln(1 z) ln3 z, (43)
z
x 3
Z1
ln(1 x) ln2 x 1 1
dx 2
= 2 S1,2 (1 z) + ln(1 z) ln2 z + ln3 z + 2 Li2 (1 z)
z
x z 3
2 1z
+ ln(1 z) ln z + ln2 z + 2 ln(1 z) + 2 ln z, (44)
z z
Z1
ln(1 x) ln2 x
dx = 2 S2,2 (1 z) + 2 S1,2 (1 z) ln(1 z), (45)
z
1x
Z1
dx ln3 (1 x) = (1 z)(ln3 (1 z) 3 ln2 (1 z) + 6 ln(1 z) 6), (46)
z
Z1
ln3 (1 x)
dx = 6 Li3 (1 z) + 6 Li2 (1 z) ln(1 z) + 3 ln2 (1 z) ln z
z
x2
1z 3
+ ln (1 z), (47)
z
Z1
dx ln3 x = z ln3 z + 3z ln2 z 6z ln z + 6z 6, (48)
z
Z1
ln3 x 1
dx 2 = (ln3 z + 3 ln2 z + 6 ln z 6z + 6). (49)
z
x z

By means of identical relations (see Appendix B) we reduce the arguments of the polylog-

7
arithm functions to (1 x). On the right hand side of the integrals we perform the same
reduction of arguments. Its worth to note, that there are certain physical arguments in
favor of the (1 x) argument with respect to the simple x. Namely, the point x = 1
corresponds usually to a singularity of a fragmentation or structure function, remind e.g.
the common lowest order splitting function

1 + x2
" #
P (0) (x) = . (50)
1x +

So in analytical results of a certain convolution, the limit x 1 has usually a principal

importance, and the choice of the argument of polylogarithms helps to analyze this limit.
Anyway, conversion of functions depending on (1 x) into the ones depending on x
can be performed using simple formulae from Appendix B. The relevant integrals of
polylogarithmic functions are

Z1
1 z n+1 1 n+1
X zk 1 zk
 
n
dx x Li2 (1 x) = Li2 (1 z) ln z + ,
z
n+1 n + 1 k=1 k k2
n = 0, 1, 2, . . . (51)
Z1
Li2 (1 x)
dx = Li3 (1 z) , (52)
z
1x
Z1
Li2 (1 x)
dx = 2 S1,2 (1 z) Li2 (1 z) ln z, (53)
z
x
Z1
Li2 (1 x) 1 z n1
dx = Li2 (1 z)
z
xn z n1 (n 1)
1 n2 1 z k ln z 1 2
X  
+ + ln z , n = 2, 3, 4, . . . (54)
n 1 k=1 z k k 2 zk k 2

Z1
dx Li3 (1 x) = (1 z) Li3 (1 z) (1 z) Li2 (1 z) + z ln z + 1 z, (55)
z
Z1
Li3 (1 x) 1
dx = Li3 (1 z) ln z (Li2 (1 z))2 , (56)
z
x 2
Z1
Li3 (1 x) 1z
dx 2
= Li3 (1 z) + 2 S1,2 (1 z) + Li2 (1 z) ln z, (57)
z
x z
Z1
Li3 (1 x)
dx = Li4 (1 z) , (58)
z
1x

8
 Z1
z 2
dx S1,2 (1 x) = (1 z) S1,2 (1 z) + ln z z ln z + z 1, (59)
z
2
Z1
S1,2 (1 x)
dx = 3 S1,3 (1 z) S1,2 (1 z) ln z, (60)
z
x
Z1
S1,2 (1 x) 1z 1
dx 2
= S1,2 (1 z) + ln3 z, (61)
z
x z 6
Z1
S1,2 (1 x)
dx = S2,2 (1 z) , (62)
z
1x
Z1
dx Li2 (1 x) ln(1 x) = (1 z) Li2 (1 z) ln(1 z) + (z 2) Li2 (1 z)
z
z ln(1 z) ln z (1 z) ln(1 z) + 2z ln z + 3 3z, (63)
Z1
Li2 (1 x) ln(1 x) 1
dx = 2 S2,2 (1 z) (Li2 (1 z))2
z
x 2
2 S1,2 (1 z) ln(1 z) Li2 (1 z) ln(1 z) ln z, (64)
Z1
Li2 (1 x) ln(1 x)
dx = 3 S1,2 (1 z) + Li2 (1 z) ln z
z
x2
1z 1
+ Li2 (1 z) ln(1 z) ln(1 z) ln2 z, (65)
z 2
Z1
Li2 (1 x) ln(1 x)
dx = Li4 (1 z) + Li3 (1 z) ln(1 z), (66)
z
1x
Z1
dx Li2 (1 x) ln x = 2 S1,2 (1 z) (z ln z z + 1) Li2 (1 z) z ln2 z
z
+3z ln z 3z + 3, (67)
Z1
Li2 (1 x) ln x 1
dx = 3 S1,3 (1 z) Li2 (1 z) ln2 z, (68)
z
x 2
Z1
Li2 (1 x) ln x 1
dx 2
= 2 S1,2 (1 z) + (ln z z + 1) Li2 (1 z)
z
x z
1 1
ln3 z ln2 z, (69)
3 2
Z1
Li2 (1 x) ln x 1
dx = (Li2 (1 z))2 . (70)
z
1x 2

Integrals of some functions, which depend on (1 + x), are required in certain cases (we

9
consider only the real part of the corresponding functions):

Z1
1
dx = ln 2 ln(1 + z), (71)
z
1+x
Z1
ln x 3
dx = Li2 (1 + z) (2), (72)
z
1+x 2
Z1
ln2 x 7
dx = (3) 2 S1,2 (1 + z) , (73)
z
1+x 2
Z1
dx Li2 (1 + x) = (1 + z) Li2 (1 + z) z ln z 1 + z + 3(2), (74)
z
Z1
Li2 (1 + x) 7 3
dx = Li3 (1 + z) + (3) + (2) ln 2, (75)
z
1+x 8 2
Z1
Li2 (1 + x) 7
dx = (3) 2 S1,2 (1 + z) Li2 (1 + z) ln z, (76)
z
x 2
Z1
Li2 (1 + x) 3 1
dx n
= (2) + n1 Li2 (1 + z)
z
x 2(n 1) z (n 1)
(1)n1 1 3
 
Li2 (1 + z) + ln2 z (2)
n1 2 2
n2 k
1 X 1z ln z
 
(1)n+k k 2
+ k , n = 2, 3, 4, . . . (77)
n 1 k=1 z k z k

5 Conclusions

The tables of integrals were implemented in a FORM [4] subroutine and used to perform
analytical calculations of various convolutions in Refs. [5,6,7,8]. The subroutine can be
used to construct an automated program for convolution of a rather wide class of functions,
which appear in perturbative QED and QCD calculations. In particular, a possibility to
include the leading and nexttoleading logarithmic corrections into the SANC project [1]
is considered. It can be done by means of an automated convolution of perturbative
coefficient functions with relevant structure and fragmentation functions.

Most of the presented integrals can be found in other sources, including automatic inte-
grators in MATHEMATICA [9] and other packages. But I hope that the tables can come in
handy in further analytical calculations.

10
Acknowledgements

This work is supported by the RFBR grant 03-02-17077.

Appendix A
Notation for polylogarithm and other functions

The short notation for partial sums is

n
X 1
Sk (n) = k
, Sk (0) = 0. (A.1)
j=1 j

Binomial coefficients are

!
n n!
= . (A.2)
k k!(n k)!

The Riemann functions are defined as usual:

X 1 2
(n) = n
, (2) = , (3) 1.20205690315959,
k=1 k 6
4
(4) = , (5) 1.03692775514337 . (A.3)
90

Following the notation of Refs. [2,3], the general Nielsens polylogarithm is

Z1
(1)n+m1 lnn1 (x) lnm (1 xz)
Sn,m (z) = dx ,
(n 1)!m! x
0
n = 1, 2, 3 . . . , m = 1, 2, 3 . . . (A.4)

In particular,

Z1 1
ln(1 xz) 1Z ln2 (1 xz)
Li2 (z) S1,1 (z) = dx , S1,2 (z) = dx ,
x 2 x
0 0

11
 Z1 z
ln(x) ln(1 xz) Z Li2 (x)
Li3 (z) S2,1 (z) = dx = dx ,
x x
0 0
Z1 2 Z1
1 ln (x) ln(1 xz) 1 ln3 (1 xz)
Li4 (z) S3,1 (z) = dx , S1,3 (z) = dx ,
2 x 6 x
0 0
Z1
1 ln(x) ln2 (1 xz)
S2,2 (z) = dx . (A.5)
2 x
0

Appendix B
Relations between polylogarithms

To convert polylogarithms of z into the ones of (1 z), one can use the following relations:

Li2 (z) = Li2 (1 z) ln z ln(1 z) + (2),

1 2
S1,2 (z) = Li3 (1 z) + Li2 (1 z) ln(1 z) + ln (1 z) ln z + (3),
2
1
Li3 (z) = S1,2 (1 z) Li2 (1 z) ln z ln(1 z) ln2 z + (2) ln z + (3),
2
1
S1,3 (z) = Li4 (1 z) + Li3 (1 z) ln(1 z) Li2 (1 z) ln2 (1 z)
2
1 3
ln (1 z) ln z + (4),
6
S2,2 (z) = S2,2 (1 z) Li3 (1 z) ln z + S1,2 (1 z) ln(1 z)
1 1
+ Li2 (1 z) ln(1 z) ln z + ln2 (1 z) ln2 z + (3) ln z + (4),
4 4
1
Li4 (z) = S1,3 (1 z) S1,2 (1 z) ln z Li2 (1 z) ln2 z
2
1 1
ln(1 z) ln3 z + (2) ln2 z + (3) ln z + (4). (B.1)
6 2

Some explicit expressions for polylogarithms of constant arguments can be useful:

Sn,m (0) = 0, n = 1, 2, 3 . . . , m = 1, 2, 3 . . .
Lin (1) = (n), n = 2, 3, 4, . . .
1
S2,2 (1) = (4), S1,n (1) = (n + 1), n = 1, 2, 3 . . .
4
3 1 1 1 1
 
Re Li2 (2) = (2), Li2 (1) = (2), Li2 = (2) ln2 2,
2 2 2 2 2
7 3 3
Re Li3 (2) = (3) + ln 2(2), Re Li3 (1) = (3),
8 2 4

12
 1 7 1 1
 
Li3 = (3) (2) ln 2 + ln3 2,
2 8 2 6
7 1 1 1 1
 
Re S1,2 (2) = (3), S1,2 (1) = (3), S1,2 = (3) ln3 2. (B.2)
4 8 2 8 6

References

[1] A. Andonov, D. Bardin, S. Bondarenko, P. Christova, L. Kalinovskaya, G. Nanava, and

G. Passarino, hep-ph/0209297.

[2] K.S. Kolbig, J.A. Mignaco, and E. Remiddi, B. I. T. 10 (1970) 38.

[3] A. Devoto and D.W. Duke, Riv. Nuovo Cim. 7 (1984) 1.

[4] J. A. Vermaseren, math-ph/0010025.

[5] A. B. Arbuzov, Phys. Lett. B 470 (1999) 252.

[6] A. Arbuzov, A. Czarnecki, and A. Gaponenko, Phys. Rev. D 65 (2002) 113006.

[7] A. Arbuzov and K. Melnikov, Phys. Rev. D 66 (2002) 093003.

[8] A. Arbuzov, JHEP 03 (2003) 063.

[9] S. Wolfram, Mathematica - A system for doing mathematics by computer, Addison - Wesley
Publishing Company Inc., 1998.

13