Beruflich Dokumente
Kultur Dokumente
7-1997
This Conference Proceeding is brought to you for free and open access by the School of Computer Science at Research Showcase @ CMU. It has been
accepted for inclusion in Computer Science Department by an authorized administrator of Research Showcase @ CMU. For more information, please
contact research-showcase@andrew.cmu.edu.
A Class of Logarithmic Integrals
Victor Adamchik
Wolfram Research Inc.
100 Trade Center Dr.
Champaign, IL 61820, USA
April 10, 1997
1 Introduction
The aim of the paper is to develop an approach for evaluating a class of integrals
involved cyclotomic polynomials and the nested logarithms log log x. This class of
integrals arose from the research regarding the Potts model on the triangular lattice
(see [1], [2]). The Potts model encompasses a number of problems in statistical
physics and lattice theory. It generalizes the Ising model so that each spin can have
more than two values. It includes the ice-vertex and bond percolation models as
special cases. It is also related to graph-coloring problems. Baxter, Temperley and
Ashley (see [3]) derived the following generating function for the Potts model on the
triangular lattice:
Z
1 sinh(( , y )x) sinh ( 2yx )
P (y) = 3
3
0 x sinh(x) cosh(yx) dx
3
(1)
Performing a logarithmic substitution, the integral can be rewritten in the "alge-
braic" form:
P (y) = 3 z(1 , z(1,)(1z +)(zzy ) log(
,y,
Z 1 y 1 , z y)
2 3 3
3
0
3
z) dz3
Although it isn't known whether the function P (y) has a closed-form expression for
3
all values of y, it can be evaluated explicitly for any y that is a rational multiple of
. Let y = pq , where p and q are positive integers, then
P (y) = 3
Z 1
z,p, (1 , z p)(z q , z p) dz
1 2 3 3
3
0 (1 + z p)(1 , z q ) log(z)
3 3
1
This integral belongs to the more common class of integrals
R(z) dz
Z 1
log(z)
0
where R(z) is a rational function. We assume that the integral is convergent. The
above integral can be envisaged in an alternative form. Performing an integration
by parts, we obtain
Q(z) log log z1 dz
Z 1
(2)
0
where Q(z) is a rational function. It is not known whether the above integral is
doable for any Q(z). However, if the denominator of Q(z) is a cyclotomic polynomial
then the integral can be always expressed in terms of derivatives of the Hurwitz
Zeta function. Using the Graee procedure for determining if a given polynomial
is cyclotomic (see [4]), and then converting a cyclotomic polynomial to the form
1 xn , allows us to reduce the problem of integration of (2) to the following two
classes of integrals:
Z 1
xp, log log 1 dx and
1
xp , 1 , x q
Z 1
0 (1 + x )
n q
x 1,x n 0 x
assuming that p, q, and n provide the convergence of the integrals. A few such
integrals (with p = q = 1 and n = 2; 3) can be found in Gradzhteyn and Ryzhyk's
handbook (see [5], pp. 532, 571-572) and in [6].
@s
(3)
2
s =0
However, if the rst argument of @s (s; z) is not zero no exact formulas were devel-
@
oped. In this section we consider the dierence of derivatives of the Zeta functions
0 s; pq , 0 s; 1 , pq
(4)
where 0(s; z) for ease of notation denotes @s@ (s; z) and p and q are positive integers,
and show that (4) can be represented in nite terms of other functions. Throughout
the paper we will freely use the notation
0 1; pq , 0 1; 1 , pq
2
Proposition 1 Let p and q be positive integers and p < q, then
0 1; pq , 0 1; 1 , pq =
q, (5)
cot q (log(2q) +
) , 2 log , jq sin 2jp
p
1
X
j q =1
Proof. The identity (5) follows straightforwardly from Rademacher's formula (see
[8]):
z; pq = 2,(1 , z)(2q)z, sin z 2jp 1 , z; j
q X
2 + q
1
q (6)
j =1
2 jp
2(log(2q) +
) sin q 0; q , 2 sin 2jp
j 0 0; j
q q
X X
j =1 j q q =1
q j =1
q = 2 q
j =1
we arrive at the identity (5). QED.
The proposition was rst proved by G. Almkvist and A. Meurman [9].
Let us consider several particular cases:
0 1; 41 , 0 1; 43 =
+ 4 log(2) + 3 log() , 4 log , 14
(7)
(10)
p 2
+ log 32 + 8 log(2) , 12 log , 13
3
Proposition 2 Let n be a positive integer and 0 < x < 1, then
0(,n; x) + (,1)n 0(,n; 1 , x) = i Bnn +(1x) + e, in2 (2n!)n Lin (e ix)
+1
+1
2
(11)
Proof. From Lerch's transformation formula for the function (z; s; v) (see [7]):
(z; s; v) = iz,v (2)s, ,(1 , s) 1
) s
(s; 1 , x) + e (s; x) = e ,(s) Li ,s (e ix)
is is (2
2 (12) 2
1
where we assume that 0 < x < 1 and s is real. Dierentiating the functional
equation (12) with respect to s, setting s to ,n , where n is a positive integer, and
making use of
lim (x) = (,1)n n!
x!,n ,(x)
+1
where Bn (x) denotes the Bernoulli polynomials, we complete the proof. QED.
+1
The identity (11) can be rewritten in the alternative form by means of the
Clausen function that is dened by
Cln(z) = ,= <(Lin (e,iz )); n is odd
0 , 2; 14 + 0 , 2; 34 = 364 (3)
2
4
where G is Catalan's constant (see [7] and [10]). Moreover, using the multiplication
property of the Zeta function
,1
s; z + ki
kX
(s; kz) = k,s
i=0
and the Proposition 2 (or Rademacher's formula (6)), one can easily deduce that
0 , 1; 61 = log(12)
, p + p0( ) + 1 0(,1)
1
3
(14)
144 12 3 8 3 6
0 , 1; 41 = 4G , 18 0(,1)
(15)
p + p 0( )
0 , 1; 31 = , log(3) 1 0(,1)
1
72 , 18 3 12 3 3
, 3
(16)
0 , 1; 21 = , log(2) 1 0(,1)
24 , 2 (17)
p , p 0( )
0 , 1; 23 = , log(3) 1 0(,1)
1
72 +
18 3 12 3 3
, 3
(18)
0 , 1; 34 = , 4G , 18 0(,1)
(19)
0 , 1; 65 = log(12)
+ p , p0( ) + 1 0(,1)
1
3
(20)
144 12 3 8 3 6
Remark1. The special case of (11) when n = 1 was also obtained by W. Gosper
(see [11]).
Remark2. Notice that, 0(,1) is related to Glaisher's constant A (see [12]) as
0(,1) = 121 , log(A)
3 Integrals
Proposition 3 Let <(p) > 0 and <(n) > 0, then
Z 1
xp, log log 1 dx =
1
01 + xn x (21)
+ log(2n) ( ( p ) , ( n + p )) + 1 ( 0(1; p ) , 0(1; n + p ))
2n 2n 2n 2n 2n 2n
where 0(s; z) denotes @s@ (s; z).
5
Proof. We shall proceed with the well-known identity
x p, 1 (,1)k Z 1 1
dx =
X
+ xn (22)
k (nk + p)
k 0
=0
Dierentiating both sides of the new identity (23) with respect to q and then setting
q = 1, we nd that
Z 1
xp, log log 1 dx = ,
xp, dx , 1 1 (,1)k log(kn + p) (24)
1 Z 1 1 X
+ xn
0 x + xn k k (kn + p)
0
=0
Now we consider the limiting case ! 1. It is clear that the innite sum in the right-
hand side of (24) is divergent when = 1. However, the sum can be analytically
continued to the whole complex plane of the parameter by means of the Lerch
function. We have
1 (,1)k log(kn + p)
X
= , lim @ 1 ( , 1) k
= , lim
X @ ( , ; s; p
n ) 1
k =0
(kn + p)
k s! @s
k (kn + p)
k s
1 s! @s
=0
ns 1
=0
since
( , 1; s; np ) = 2,s (s; 2pn ) , (s; 12 + 2pn )
Thus,
xp, log log 1 dx = ,
xp, dx + lim @ (s; pn ) , (s; + pn )
!
Z 1 1 Z 1 1 1
2 2 2
0 1 + xn x 0 1 + xn s! @s 1 (2n)s
Performing further evaluations and taking into account the asymptotic expansion
(s; n) ! ,
, (n) + (s); s ! 1
we nally arrive at (21). QED.
Note if np is a positive integer, then the right-hand side of (21) can be expressed
in terms of elementary functions. Let p = nr , r 2 N , then
xnr, log log 1 dx =
Z 1 1
1 + xn
0 x (25)
1 (log(2n) +
) ( r ) , r + 1 + 0(1; r ) , 0 1; r + 1
2n 2 2 2 2
6
Now making use of the following reduction formulas
0(s; v) = 0(s; v , 1) + log( v , 1)
(v , 1)s
(s) = (s , 1) + s ,1 1
and the identity
0 1; 12 , 0(1; 1) = (log(2) +
) log(4) , log (2)
2
it is easy to see that the right-hand side of (25) can be transformed to the combi-
nation of logarithmic functions. Thus,
Corollary 2 If p = n and <(n) > 0, then
Z 1
xn, log log 1 dx = , log(2) log(2n )
1 2
(26)
0 1 + xn x 2n
2
0
1+x 0 x 2
+ log(4) 1 , 3 + 1 0 1; 1 , 0 1; 3
4 4 4 4 4 4
By means of (7), we have
p
1 log log 1 dx = log 2,( )
!
Z 1 3
1+x 2
x 2 ,( ) 1
4
(28)
0
4
3.2
Z 1
x log log 1 dx
0 1+x 4
x
7
Taking p = 2 and n = 4 in (21), we have
x log log 1 dx =
Z 1
1+x 0 x 4
1 (log(8) + ) 1 , 3 + 1 0 1; 1 , 0 1; 3
8 4 4 8 4 4
Using the identity (7), we obtain
x log log 1 dx = log p,( )
Z 1 !
3
1+x x 04 ,( ) 4 1
4
(29)
4
3.3
Z
1 log log 1 dx
1
1+x
0 x
3
1+x 0 x 3
1 (log(6) + ) 1 , 2 + 1 0 1; 1 , 0 1; 2
6 6 3 6 6 3
Performing further simplications and using the formula (10), we nd that
1 log log 1 dx =
Z 1
1+x x 3
(30)
0
2
0
log log 1 dx =
1,x+x 0 x 2
1 log log 1 dx +
Z 1
x log log 1 dx Z
1
1+x 0 x 1+x 3
x 0
3
log log 1 dx =
1,x+x 0 x 2
1 0 1; 1 , 0 1; 2 + 0 1; 1 , 0 1; 5 , 4(log(6)
p +
)
6 3 3 6 6 3
Then taking into account formulas (8) and (9), we nally nd that
1 Z 1
8
Proposition 4 Let <(p) > 0 and <(n) 1, then
xp, 11,,xxn log log x1 dx = n1 (log(n) +
)( ( np ) , ( p +n 1 ))+
Z
1
1
0
(32)
1 0(1; p ) , 0(1; p + 1 )
n n n
where 0(s; z) denotes @s@ (s; z):
If n = 1, then the formula (32) simplies to
xp,1 log log 1 dx = , log(p) +
Z
1
x p (33)
0
k (nk + p + 1)
k (34)
0
k=0 k=0
since
xp, dx = 1 1
Z 1 1 X
,x n
k (nk + p)
k 0
=0
k (kn + p)
k q
k (kn + p + 1)q
k
=0 =0
0
Z 1
1 , x 1 1 log(kn + p) 1 1 log(kn + p + 1) (36)
,
x , xn dx , k (kn + p) + k (kn + p + 1)
, X X
p 1
0
k k =0 =0
The innite sums in the right side of (36) can be represented in terms of the Lerch
function. We have
1
log(kn + p) = ,lim @ 1 1 @ ( ; s; np ) 1
,
X X
k (kn + p) s! @s
= lim
k (kn + p)
k s s! @s ns 1 1
k =0 =0
kn + p + 1 , kn + p = , lim
s! @s ns ns 1
k =0
1 log(n) ( p ) , p + 1 + 0(1; p ) , 0 1; p + 1
n n n n n
9
since
(1; s; np ) = (s; np )
and
(s; n) ! ,
, (n) + (s); s ! 1
Thus,
xp, 11,,xxn log log x1 dx = ,
1x, xn (1 , x)dx+
Z 1 p, Z 1 1
1
0 0
1 log(n) ( p ) , p + 1 + 0(1; p ) , 0 1; p + 1
n n n n n
Evaluating the elementary integral,
Z 1
xp, (1 , x)dx = 1 ( p + 1 ) , ( p )
1
0 1 , xn n n n
we complete the proof. QED.
Here are some particular cases.
3.5 1 + x1+ x log log x1 dx
Z 1
2
0
Since
Z 1
1+x+x 0 x 1,x x2
0
3
log 1
log 1 dx = ( , 3 log(3) + 8 log(2
p
) , 12 log (,( )))
1
3
0 1+x+x x 2
6 3
3.6
Z 1
x log log 1 dx
0 1+x+x +x +x +x 2
x 3 4 5
In view of
Z 1
1+x+x +x +x
0 x2
1,x
3
x 4
0
5
log log 1 dx =
1+x+x +x +x0 x 2 3 4
1 (log(5) +
) 2 , 3 + 0 1; 2 , 0 1; 3
5 5 5 5 5
10
Applying Proposition 1, we obtain
Z 1
x
log log 1 dx =
0 1+x+x +x +x 2
x 3 4
(38)
where 0(s; z) denotes @s@ (s; z)
If p = n, then
xn, log log 1 dx = , 1
+ log 2n
Z 1 1
(39)
(1 + xn) 0 x 2n 2
Proof. Dierentiating both sides of (24) with respect to , we obtain
xp, log log 1 dx =
Z 1 1
( + xn)
0 x 2
(40)
,
xp, dx , 1 1 (,1)k (k + 1) log(kn + p)
Z
1 1 X
( + xn) k
0 k (kn + p) 2 2
=0
We express the innite sum in the left-hand side of (40) by means of the Lerch
function
1 (,1)k (k + 1) log(kn + p) @ 1 (,1)k (k + 1) =
,
X X
k (kn + p) = lim
s! @s k (kn + p)s 1
k =0 k =0
lim @ 1 ( p , 1) , 1 ; s; p , , 1 ; s , 1; p
s!1 @s ns n n n
Therefore, the limiting value of the sum when approaches 1 is
lim
1
X ( , 1) k
(k + 1) log(kn + p) =
! 1
k =0
k (kn + p)
log(2n) + (n , p) log(2n) ( ( n + p ) , ( p ))+
2n 2n 2n 2n2
11
Then taking into account the identity (3) and
xp, dx = 1 + (n , p) ( ( n + p ) , ( p ))
Z 1 1
(1 + xn)
0 2n 2n 2n
2
2n 2
(1 + x)
0
2
(41)
, 3 log(2) + log 4,( ) + log 2p,( )
! !
3 3
4 4
2 2 ,( ) 2 ,( ) 1 1
px 4 4
log log x1 dx =
Z
1
(1 + x ) 0 3 2
p (42) q
1 6,( ) 1 2 ,( ) 3
! 3 !
, 6 , 3 log ,( ) + 6 log ,( ) 1
4 3
1
4
4 4
Z 1
x log log x1 dx =
0 (1 , x + x ) 2 2
p p (43)
1 6 3
, 3 , 3 log + 273 5 log(2) , 6 log , 16
!
0 (1 + xn) x3
,( n np ) +
2
8n 2n 2 2 +
2
1 ( 0( , 1; p ) , 0( , 1; n + p ))+
n 2n 2n
(n , p)(2n , p) ( 0(1; p ) , 0(1; n + p ))
4n 2n
3
2n
If p = n, then
Z 1
xn, log log 1 dx = ,10 log(2) , 9 log(n) + 6 log() , 9
, 36 0(,1)
1
0 (1 + xn) 3
x 24n
(45)
12
Proof. The proof is similar to that for Proposition 5. QED.
The following integrals follow immediately from (44):
px p
log log x1 dx = , 2G + 8 log 2 ,()
)
!
Z 1 3
4
(46)
0 (1 + x) 3
,( 1
4
Z 1
1 log log 1 dx =
0 (1 + x ) 2 3x
(47)
1 , 16 log(2) + 3 log(2) + 8G , 8
+ 2(3 , 8) log ,( )
!!
3
4
32 ,( ) 1
4
2 0 (1 + x ) x
2 3 4 z 0
Let us consider the particular case of the generating function for the Potts model
on the triangular lattice when y = . Other special cases of the integral (1) are
2
described in [2]. From (1) we have
P ( 2 ) = 3 tanh(x)
1 Z
dx (49)
x (1 , 2 cosh(2 x))
3
0
2
Originally the integral was calculated by R.J. Baxter (see [3]) by using the Fourier
analysis and the residue theorem.
Performing an exponential substitution and then integrating a correspondent
integral one time by parts, we transform (49) to
P ( 2 ) = 6 x (1 + x) (1 , x , x , x + x ) log log( x1 ) dx
Z 1 2 3 4
3
0 (1 + x ) 3 3
References
[1] R. M. Zi, S. R. Finch and V. Adamchik, Number of clusters in 2D percola-
tion:Values, nite-size corrections,
uctuations, and explicit evaluation of exact
results , Phys. Rev. Letters, (submitted for publication).
13
[2] V. Adamchik, S. R. Finch and R. M. Zi, The Potts model on the triangular lat-
tice , in Web site http://www.wolfram.com/~victor/articles/percolation.html.
[3] R. J. Baxter, H. N. V. Temperley and S. E. Ashley, Triangular Potts model at
its transition temperature, and related models , Proc. Royal Soc. London A 358
(1978) 535-559.
[4] R. J. Bradford and J. H. Davenport, Eective Tests for Cyclotomic Polynomi-
als , Proceedings of ISSAC '88, p.244-245.
[5] I. S. Gradzhteyn and I. M. Ryzhyk, Table of Integrals, Series, and Products ,
Academic Press, New York, 1980.
[6] I. Vardi, Integrals, an Introduction to Analytic Number Theory , Amer. Math.
Monthly, 95(1988).
[7] H. Bateman and A. Erdelyi, Higher Transcendental Functions , Vol. 1, McGraw-
Hill, 1953.
[8] T. M. Apostol, Introduction to Analytic Number Theory , Springer-Verlag, 1976.
[9] G. Almkvist and A. Meurman, private communication.
[10] V. Adamchik, Integral and Series Representations for Catalan's Constant , in
Web site http://www.wolfram.com/~victor/articles/catalan.html.
m
[11] R. W. Gosper: log ,(z)dz, Amer. Math. Soc. 14(1997).
R
n6
4
14