A q-logarithmic analogue of Euler's sine integral (*)
NOBUSHIGEKUROKAWA(**) - MASATOWAKAYAMA(***)
ABSTRACT- We study aq-logarithmic analogue of Euler's sine integral proved in 1769. It is evaluated by the quantum dilogarithm function.
1. Introduction.
In [E] 1769, Euler calculated the famous integral Zp2
0
log (sinx)dx p 2log 2:
1:1
This appears frequently in calculus texts as an example of somewhat tricky calculation. We remark that a refined formulation of (1.1) is given by the formula
Zp2
0
log (1 e2ix)dx p2i 8 : 1:2
(*) Partially supported by Grant-in-Aid for Scientific Research (B) No. 15340012, and by Grant-in-Aid for Exploratory Research No. 15654003
Mathematics Subject Classification 2000: 11M36
(**) Indirizzo dell'A.: Department of Mathematics, Tokyo Institute of Technol- ogy, Oh-okayama Meguro, Tokyo, 152-0033 JAPAN. e-mail: [email protected] tech.ac.jp
(***) Indirizzo dell'A.: Faculty of Mathematics, Kyushu University, Hakozaki Fukuoka, 812-8581, JAPAN. e-mail: [email protected]
In fact, taking the real part of (1.2), we have (1.1) because
Re Zp2
0
log (1 e2ix)dx Zp2
0
logj1 e2ixjdx
Zp2
0
log (2 sinx)dx
Zp2
0
log (sinx)dxp 2log 2 by noticing that Re p82i 0.
We formulate a slightly generalized version. We use Euler's diloga- rithm function
Li2(x)X1
n1
xn n2 and the double sine function
F(x)exY1
n1
1 nx 1nx n
e2x
due to HoÈlder [H]. We define
F1(x)F(x) 2pi(1 e2pix)2pixep2(x2 16) for Im (x)0. Then we have the
THEOREM1.1. For0<a1 Zp2
0
log (1 ae2ix)dx 1
2ifLi2( a) Li2(a)g
1
2ilog F1 loga 2pi 12
F1 loga 2pi
0
@
1 A:
In the case of a1 in Theorem 1.1 we obtain (1.2) from the facts F1(0)e p62andF1 12 ep122. We can easily show these values are obtained from the definition of F1(x). Actually, the value ofF1(0) is obvious from
F(0)1. Also, using the resultF 12
p2
of HoÈlder in [H], we have F1 1
2 F 1 2
2pi2piep122ep122:
Now we present aq-analogue of Theorem 1.1 or (1.2). Letq>1. We put
`q(x)(q 1)X1
m1
x 1 x 1qm
forx2C, which defines a meromorphic function inx. The function`q(x) is considered as aq-analogue of the usual logarithm
logxX1
m1
( 1)n 1(x 1)n n
expanded injx 1j<1, as the following calculation shows:
`q(x)(q 1)X1
m1
(x 1)q m 1(x 1)q m
(q 1)X1
m1
X1
n1
( 1)n 1(x 1)nq nm
(q 1)X
n1
( 1)n 1(x 1)n
qn 1 X
n1
( 1)n 1(x 1)n [n]q ; where [n]qqqn11and limq#1[n]qn.
We show that aq-analogue of (1.2) is expressed as Zp2
0
`q(1 e2ix)dxq 1
2i log Y1
n1
1q n 1 q n
! : 1:3
To formulate the result neatly, we recall the quantum dilogarithm (see Kirillov [K]);
Li2;q(x)X
n1
xn n[n]q: We also introduce
Fq(x)Y1
n1
(1 q ne2pix)q 1: Then we have the following theorem.
THEOREM1.2. For0<a1, it holds that Zp2
0
`q(1 ae2ix)dx 1
2i
Li2;q( a) Li2;q(a)
1
2ilog Fq loga 2pi 12
Fqlog2pia 0
@
1 A:
If we puta1 in Theorem 1.2, the equation (1.3) follows immediately from the second expression. Moreover, we have the following limit for- mulas which allow us to obtain (1.2) as the limitq#1 of (1.3).
PROPOSITION1.3.
limq#1(q 1) log Y1
n1
(1q n)
!
p2 1:4 12
and
limq#1(q 1) log Y1
n1
(1 q n)
!
p2 6 : 1:5
Furthermore, we prove the following
THEOREM1.4. SupposeImx0. Then we have limq#1Fq(x)F1(x):
This result indeed shows that ``limq#1 (Theorem 1.2) = Theorem 1.1''.
We add one remark. Since Zp2
0
log (sinx)dx1 4
Z2p
0
logjsinxjdx1 4Re1
4 Z2p
0
log (sinx)dx;
it may be also reasonable to think the quantity 1
4Re1 4
Z2p
0
`q(sinx)dx
as an analogue of (1.1). From this viewpoint, we obtain the following
THEOREM1.5. For q>2 1
4 Re1 4
Z2p
0
`q(sinx)dx p
2 X1
n1 2nn
2 n [n]q
(q 1)p 2
X1
n1
1 1 2q m
p 1
( )
:
We note that, in the formula above, it is impossible to take the limitq#1 directly.
2. Proof of Theorem 1.1.
The former part of the theorem is straightforward. In fact, we calculate Zp2
0
log (1 ae2ix)dx Zp2
0
X1
n1
ane2inx n
! dx
X1
n1
an n
Zp2
0
e2inxdx X1
n1
an n
e2inx 2in p2
0
1 2i
X1
n1
( a)n an n2
1 2i
Li2( a) Li2(a) :
To show the latter part of the theorem, it is sufficient to prove the following lemma.
LEMMA2.1. ForImx0
F1(x)exp ( Li2(e2pix)):
This is essentially proved in [KOW] or [KK], where the theory of multiple sine functions is developed in extending the results to the double sine functions due to HoÈlder [H] (for the period (1;1)) and Shintani [S] (for the period (v1;v2)). For reader's convenience, we sketch the proof. See also ``Miscellaneous Examples'' Chap. VII-20 of Whittaker-Watson [WW], which treats indeed the double sine function F(x) of HoÈlder [H] (but without mentioning to this reference).
PROOF OFLEMMA2.1. We first show that F(x)exp
Zx
0
ptcot (pt)dt 0
@
1 A: 2:1
In fact, from the expression logF(x)xX1
n1
n log 1 x n
log 1x n
2x
n o
we obtain
F0(x)
F(x) 1X1
n1
n 1
x n 1 nx
2
1X1
n1
2x2
x2 n2pxcot (px):
Hence, usingF(0)1, we obtain (2.1).
It follows from (2.1) that logF(x)
Zx
0
ptcot (pt)dt:
Suppose Im (x)0. Then we have logF(x)
Zx
0
pitepite pit epit e pitdt
Zx
0
pit 12 e2pit 1 e2pit
dt
pix2
2 2piX1
n1
Zx
0
te2pintdt
pi
2 x2 2piX1
n1
xe2pinx 2pin
e2pinx 1 (2pin)2
pi
2 x2 xX1
n1
e2pinx n 1
2pi X1
n1
e2pinx 1 n2
pi
2 x2xlog (1 e2pix) 1
2piLi2(e2pix)pi 12:
SinceF1(x)F(x) 2pi(1 e2pix)2pixep2(x2 16), this shows the claim in Lem-
ma 2.1. p
Thus Theorem 1.1 follows.
3. Proof of Theorem 1.2
The former part is exactly similar to the case of Theorem 1.1. Actually, we have
Zp2
0
`q(1 ae2ix)dx Zp2
0
X1
n1
ane2inx [n]q
! dx
X1
n1
an [n]q
Zp2
0
e2inxdx 1 2i
X1
n1
( a)n an [n]qn
1 2i
Li2;q( a) Li2;q(a) :
To show the latter part, it suffices to use the following result.
LEMMA3.1. ForIm (x)0
Fq(x)exp ( Li2;q(e2pix)):
PROOF. This is easily seen as follows (see [KW] also):
logFq(x)(q 1)X1
n1
log (1 q ne2pix)
(q 1)X1
n1
X1
m1
q nme2pimx
m (q 1)X1
m1
e2pimx m(qm 1)
X1
m1
e2pimx
m[m]q Li2;q(e2pix):
Hence the claim follows. p
Thus, we obtain Theorem 1.2.
4. Proof of Theorem 1.4.
This is obvious from Lemma 2.1 and Lemma 3.1, because the formula limq#1Li2;q(x)Li2(x):
holds forjxj<1.
5. Calculations of special values.
First, we notice that one can directly see that HoÈlder's result is equivalent to Euler's integral, that is,
F 1 2
p2
() (1:1) as follows. From (1.2) we see that
F 1
2 exp Z12
0
ptcot (pt)dt
!
exp
tlog (sinpt)12
0
Z12
0
log (sinpt)dt
!
exp Z12
0
log (sinpt)dt
! :
Hence we observe that F 1
2
p2 ()
Z12
0
log (sinpt)dt 1 2log 2
() Zp2
0
log (sinx)dx p 2log 2:
We now look at the casea1 of Theorem 1.2 and take the limitq#1.
Then we have (1.2) again by
limq#1(q 1) log Y1
n1
1q n 1 q n
!
p2 4 :
To see this limit formula, it is enough to show that (1.4) and (1.5) in Pro- position 1.3.
PROOF OFPROPOSITION 1.3. Let us show these limit formulas by em- ploying the modular formD(t) of Ramanujan defined by
D(t)e2pitY1
n1
(1 e2pint)24
for Imt>0. This has the functional equation, i.e. the automorphy ofD(see, e.g., [We]);
D 1
t
t12D(t):
Forq>1, taketlog (q2pi1). Then, sinceD(t)q 1Q1
n1(1 q n)24, we have log Y1
n1
(1 q n)
!
1
24logD(t) 1 24logq
1
24log t 12D 1 t
1 24logq
1
2logt 1 24
2pi
t log Y1
n1
(1 e 2pint )
!
1 24logq:
Hence it follows that (q 1) log Y1
n1
(1 q n)
!
p2 6
q 1 logq
q 1
2 logt(q 1) log Y1
n1
(1 e 2pint )
!
q 1 24 logq:
Takingq#1, we have q 1
logq !1; (q 1) logt!0; e 2pit !0 and (q 1) logq!0:
Therefore we obtain
limq#1(q 1) log Y1
n1
(1 q n)
!
p2 6 :
(This process shows that one may have more detailed asymptotics.) Then, the relation
(q 1) log Y1
n1
(1q n)
!
(q 1) log Y1
n1
1 q 2n 1 q n
!
1
q1(q2 1) log Y1
n1
(1 q 2n)
!
(q 1) log Y1
n1
(1 q n)
!
yields
limq#1(q 1) log Y1
n1
(1q n)
!
1 2
p2 6
p2 6
p2 12:
This completes the proof of Proposition 1.3. p
6. Proof of Theorem 1.5.
Forq>2 we calculate Z2p
0
`q( sinx)dxX1
n1
( 1)n 1 [n]q
Z2p
0
(sinx 1)ndx
X1
n1
1 [n]q
Z2p
0
(1 sinx)ndx:
A simple calculation shows the following lemma.
LEMMA6.1.
Z2p
0
(1 sinx)ndx2p 2n n 2 n: 6:1
PROOF. We calculate as Z2p
0
(1 sinx)ndx Z2p
0
sinx 2 cosx
2
2n
dx Z2p
0
2 p sin x
2 p 4
2n
dx
2n1 Z3p4
p4
sin2nudu2n2 Zp2
0
sin2nudu:
Therefore, by the well-known evaluation Zp2
0
sin2nudup 2
2n n 2 2n;
we get (6.1). p
From this lemma we get the former part of Theorem 1.4. We next show X1
n1 2nn
2 n
[n]q (q 1)X1
m1
n 1 1 2q m p 1o
:
Recalling the binomial expansion (1 x) 12X1
n0
2n
n 2 2nxn (jxj<1);
we can calculate as X1
n1 2nn
2 n
[n]q (q 1)X1
n1
2n
n 2 n(qn 1) 1
(q 1)X1
n1
X1
m1
2n
n 2 nq nm(q 1)X1
n1
X1
m1
2n
n 2 2n(2q m)n
(q 1)X1
m1
n 1 1 2q m p 1o
: Thus the theorem follows.
REFERENCES
[E] L. EULER, De summis serierum numeros Bernoullianos involventium.
Novi comm. acad. scient. Petropolitanae, 14 (1769), pp. 129-167 (Opera Omnia I-15, pp.91-130).
[H] O. HOÈLDER,Ueber eine transcendente Function. GoÈttingen Nachrichten, 16(1886), pp. 514-522.
[K] A. N. KIRILLOV,Dilogarithm identities, Progr. Theoret. Phys. Suppl.,118 (1995), pp. 61-142.
[KK] N. KUROKAWA - S. KOYAMA, Multiple sine functions. Forum Math., 15 (2003), pp. 839-876.
[KOW] N. KUROKAWA- H. OCHIAI - M. WAKAYAMA,Multiple trigonometryand zeta funcitons.J. Ramanujan Math. Soc.,17(2002), pp. 101-113.
[KW] N. KUROKAWA- M. WAKAYAMA,Absolute tensor products.Internat. Math.
Res. Notices,5(2004), pp. 249-260.
[S] T. SHINTANI,On a Kronecker limit formula for real quadratic fields. J.
Fac. Sci. Univ. Tokyo,24(1977), pp. 167-199.
[We] A. WEIL,Sur une formule classique. J. Math. Soc. Japan,20(1968), pp.
400-402.
[WW] E.T. WHITTAKER - G. N. WATSON, A Course of Modern Analysis,Cam- bridge Univ. Press (Third ed.) 1920.
Manoscritto pervenuto in redazione l'8 dicembre 2004, modificato il 20 aprile 2005.