Integrals Over Polytopes, Multiple Zeta Values and Polylogarithms, and Euler's Constant
Jonathan Sondow (New York), Sergey Zlobin (Moscow)

TL;DR
This paper investigates integrals over a specific triangle, expressing them via multiple zeta values and polynomials, deriving asymptotics, and exploring connections to Euler's constant and higher-dimensional polylogarithm integrals.
Contribution
It provides new explicit formulas for integrals over polytopes in terms of multiple zeta values and extends these results to higher-dimensional cases involving multiple polylogarithms.
Findings
Expressions for integrals as linear combinations of multiple zeta values.
Asymptotic expansions of integrals and zeta sums.
Relations between multiple polylogarithms and zeta values.
Abstract
Let be the triangle with vertices (1,0), (0,1), (1,1). We study certain integrals over , one of which was computed by Euler. We give expressions for them both as a linear combination of multiple zeta values, and as a polynomial in single zeta values. We obtain asymptotic expansions of the integrals, and of sums of certain multiple zeta values with constant weight. We also give related expressions for Euler's constant. In the final section, we evaluate more general integrals -- one is a Chen (Drinfeld-Kontsevich) iterated integral -- over some polytopes that are higher-dimensional analogs of . This leads to a relation between certain multiple polylogarithm values and multiple zeta values.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Integrals Over Polytopes, Multiple Zeta Values and Polylogarithms, and Euler’s Constant
Jonathan Sondow
209 West 97th Street
New York City
New York 10025 USA
jsondow(at)alumni.princeton.edu
and
Sergey Zlobin
Faculty of Mechanics and Mathematics
Moscow State University
Leninskie Gory, Moscow 119899 RUSSIA
sirg_zlobin(at)mail.ru
Abstract
Let be the triangle with vertices (1,0), (0,1), (1,1). We study certain integrals over , one of which was computed by Euler. We give expressions for them both as a linear combination of multiple zeta values, and as a polynomial in single zeta values. We obtain asymptotic expansions of the integrals, and of sums of certain multiple zeta values with constant weight. We also give related expressions for Euler’s constant. In the final section, we evaluate more general integrals – one is a Chen (Drinfeld-Kontsevich) iterated integral – over some polytopes that are higher-dimensional analogs of . This leads to a relation between certain multiple polylogarithm values and multiple zeta values.
Contents
- 1 Introduction
- 2 The integral for
- 3 Asymptotic expansion of
- 4 Applications to multiple zeta values
- 5 The integral and Euler’s constant
- 6 Integrals over higher-dimensional analogs of
- 7 References
1 Introduction
Let be the triangle defined by
[TABLE]
with vertices (1,0), (0,1), (1,1). In this paper we study the integral over
[TABLE]
for . We also consider integrals over several polytopes that are higher-dimensional analogs of .
Euler computed an iterated integral equivalent to , and found that
[TABLE]
Using integration by parts he derived formula (22), and used it to calculate correctly to six decimals—see [5, Section 1.2], [7, pp. 43-45].
We generalize Euler’s result to by showing that is equal to an integer linear combination of multiple zeta values
[TABLE]
of weight . We also express as a polynomial in single zeta values.
Theorem 1 Let * be an integer.*
(i) Then
[TABLE]
where denotes (k times).
(ii) Moreover, is equal to an explicit polynomial of several variables with rational coefficients in the values of the Riemann zeta function
(Theorem 1, Corollary 1, and Lemma 1 were obtained by the second author in [17].) The proof is given in Section 2, along with the explicit formula. Examples are
[TABLE]
[TABLE]
The cases and 2 are particularly simple.
Corollary 1 For , and 2, the integral is a rational multiple of .
For and 1, this also follows from Beukers’ [2] formulas for and as integrals over the unit square
[TABLE]
Namely, the change of variables transforms both into
[TABLE]
and into
[TABLE]
Here is an outline of the proof of Theorem 1. We first prove
Lemma 1 If * and*
- are integers, then*
[TABLE]
If in addition , then
[TABLE]
Expanding , part (i) follows immediately. To prove the lemma, we show that , and then evaluate the integral . Part (ii) of Theorem 1 follows, using a formula in [9] for in terms of single zeta values.
As an application, we obtain an explicit version of a result in [3].
Corollary 2 If and , then the multiple zeta value
- can be explicitly represented as a polynomial of several variables with rational coefficients in the single zeta values*
Lemma 1 also affords a simple proof of a special case of the duality theorem for multiple zeta values (see, for example, [5, Section 2.8]).
Corollary 3 If * and* , then
For instance, and . Using these equalities, we give a second proof of Corollary 1. However, unlike the cases and 1, we do not have a proof of the case of Corollary 1 that does not use Theorem 1.
On the basis of numerical evidence and examples such as (3), we make the
Conjecture 1 The integral is not* a rational multiple of* * when* .
This has not been proved for a single value of . However, using Theorem 1 (ii), we give a conditional proof for all , assuming a standard conjecture (see, for example, [16, Introduction]).
Theorem 2 If the numbers
- are algebraically independent over the rationals, then Conjecture* 1* is true.*
Using a lemma we prove which gives an asymptotic expansion for the coefficients of the Taylor series of certain meromorphic functions (Lemma 2), we estimate for large.
Theorem 3 The asymptotic equivalence
[TABLE]
holds. More precisely, the following asymptotic expansion is valid:
[TABLE]
where the numerator of the k-th term is , for .
This in turn gives an estimate for the sum of the multiple zeta values of constant weight .
Corollary 4* The average of the multiple zeta values*
- is asymptotic to* * as m tends to infinity*.
- In fact, the following asymptotic expansion holds*:
[TABLE]
Another application of Theorem 3 is a curious result.
Corollary 5 The series
[TABLE]
diverges, but is Abel summable to
Let us now go "down" from to .
Question Can one evaluate the integral
[TABLE]
in terms of more familiar constants?
Surprisingly, it turns out that involves all the integrals (hence all multiple zeta values for and .
Theorem 4 If is the logarithmic integral function, then
[TABLE]
(Compare the convergent series here with the divergent series in Corollary 5.)
We now transform the double integral into single integrals, one involving the generalized binomial coefficient
[TABLE]
Proposition 1 The following integral formulas for are valid:
[TABLE]
Expanding the first integrand in a power series, we find that the th coefficient involves the integral .
Theorem 5 If , then
[TABLE]
An application is Corollary 5.
We now relate to Euler’s constant , which is defined as the limit
[TABLE]
If one thinks of as "," then from the formulas , , and one might expect that involves . This is also suggested by the similarity between the double integral (7) for and the double integral for Euler’s constant [12], [14]
[TABLE]
Formula (8) leads to another, related connection between and . Namely, when is a positive integer, is the central binomial coefficient , which figures in the formulas for Euler’s constant
[TABLE]
and
[TABLE]
where is a certain rational number and is a particular linear form in logarithms [12].
If in (10) we perform the change of variables , we obtain an integral over the triangle for Euler’s constant,
[TABLE]
analogous to the triangle integral (7) for .
We find an analog for of the first integral for in (8), which involves the generalized binomial coefficient . (There exist classical analogs for of the second integral, which involves logarithms.)
Proposition 2 The following formula for Euler’s constant is valid:
[TABLE]
As an application, if we integrate termwise, and exponentiate the resulting series, we recover Ser’s infinite product for [11] (rediscovered in [13], [15]):
[TABLE]
The rest of the paper is organized as follows. In Sections 2 and 3 we establish the non-asymptotic and asymptotic results, respectively, on for . The applications to multiple zeta values are proved in Section 4, and in Section 5 we prove the formulas for and . The final section is devoted to generalizing to integrals over higher-dimensional analogs of the triangle ; one is a Chen (Drinfeld-Kontsevich) iterated integral (see Remark 3). An application is a relation between certain multiple polylogarithm values and multiple zeta values (Corollary 7).
2 The integral for
We prove the non-asymptotic results on stated in the Introduction.
Lemma 1 If * and* * are integers, then*
[TABLE]
If in addition , then
[TABLE]
Proof. We have
[TABLE]
Replacing with , we see that
[TABLE]
Thus (4) follows from (5). To prove (5), we multiply the formula [16, Section 1]
[TABLE]
by , and substitute the resulting series
[TABLE]
into the integral (5) for . We then integrate termwise, using the fact that
[TABLE]
The result is
[TABLE]
and the lemma follows.
Theorem 1 If , then * can be expressed both*
(i) in terms of multiple zeta values as
[TABLE]
(ii) and in terms of single zeta values as
[TABLE]
where the integral ,* defined in* (5),* is given by the formula* [9]
[TABLE]
the sum on * being taken over all sets of integers* * with*
[TABLE]
and the sum on * over all sets of integers* * with*
[TABLE]
Proof. Expanding in the definition (1) of , and applying (4), gives (12). Hence, using (5), formula (13) holds. Finally, the evaluation (14) of the integral (5) for is proved in [9].
Corollary 1 For ,* and* 2, the integral is a rational multiple of .
We give two proofs. The first is short, but uses Theorem 1 (ii), whose proof depends on [9]. The second is longer, but is self-contained (except for a formula due to Euler): it uses Theorem 1 (i) and Corollary 3, whose proofs do not rely on other papers.
Proof 1. For , and 2, formulas (13) and (14) yield and and .
Proof 2. In the Introduction, we showed that . Using the same method, together with the formula , we obtain
[TABLE]
Alternatively, and by Theorem 1 (i), and by Corollary 3.
In order to prove that , it suffices, by Theorem 1 (i) and Corollary 3, to apply Euler’s formula . (For the latter, take in his equation (9.5) of [1, p. 252].)
Theorem 2 If the numbers
- are algebraically independent over the rationals, then* is not a rational multiple of
- when*
Proof. First take the case , with . The integral is equal to a linear combination (13) of integrals with positive coefficients. Each is equal to a polynomial (14) of several variables in single zeta values. Now in (14) the monomial appears only when , and then its coefficient is nonzero and has sign . Hence in the expression for the coefficient of is nonzero. It follows, using the hypothesis, that cannot be a rational multiple of .
The cases and , with , are similar: consider the monomials and , respectively. The remaining cases and can be handled by direct calculation, completing the proof.
Theorem 5 If , then
[TABLE]
Proof. The generating function
[TABLE]
(compare (19)) is derived in [3]. If , then the series converges when . Setting , we obtain
[TABLE]
Applying (2), the theorem follows.
3 Asymptotic expansion of
Using Theorem 5 and the next lemma, we estimate the integral when is large.
Lemma 2* Suppose that the function* * is meromorphic in the complex plane and has only simple poles* , with residues , respectively. If , then the coefficients of the Taylor series
[TABLE]
have the asymptotic expansion
[TABLE]
Remark 1 Recall [6, Section 1.5] that the last formula means that, for every fixed positive integer ,
[TABLE]
Proof of Lemma 2. A meromorphic function has only finitely many poles in any bounded region, so for each there exists with . Note that the only singularities of the function
[TABLE]
are . It follows, using the inequalities , that the radius of convergence of the series (18) is equal to . As , we may substitute into the series. Therefore,
[TABLE]
Since , the limit implies the asymptotic formula (17). This proves the lemma.
Theorem 3 We have * as n tends to infinity. More precisely, the following asymptotic expansion is valid*:
[TABLE]
where the numerator of the k-th term is for .
Proof. Denote the function on the left side of (9) by . Aside from a removable singularity at , the singularities of are simple poles at . A calculation shows that the residue at is equal to
[TABLE]
for . Using Theorem 5 and Lemma 2, the second statement (which implies the first) follows.
Corollary 5 The series
[TABLE]
diverges, but is Abel summable to
Proof. The divergence follows from (6). Letting in (15), we obtain the desired Abel summation.
4 Applications to multiple zeta values
Using the results obtained on , we study multiple zeta values of the form .
Corollary 2 If and , * then the multiple zeta value* * can be explicitly represented as a polynomial of several variables with rational coefficients in the single zeta values*
Remark 2 This result, including the polynomial formula (at least implicitly), was first obtained in [3], using the generating function (see [3] for the equivalence with (16))
[TABLE]
Corollary 3 If * and* , then
Proof. Making the change of variables in the integral (4), the symmetry of the triangle yields . Using Lemma 1, the result follows.
Corollary 4
- The average of the multiple zeta values* * is asymptotic to* * as m tends to infinity*.* In fact, the following asymptotic expansion holds*:
[TABLE]
Proof. Setting in Theorems 1 and 3 gives the desired expansion. It follows that the average in question is asymptotic to as tends to infinity.
5 The integral and Euler’s constant
We prove the results on and stated in the Introduction.
Proposition 1 The following single integral formulas for the double integral are valid:
[TABLE]
Proof. In (7) make the substitution
[TABLE]
and change the order of integration, obtaining
[TABLE]
using Euler’s integral for the beta function. Replacing with , the first equality follows. To see that the second integral is also equal to , integrate with respect to in (7).
Theorem 4 If * is the logarithmic integral, then*
[TABLE]
Proof. Make the substitution
[TABLE]
in (7). Using the proof of Proposition 1, we get
[TABLE]
Substituting the series (15) into the first integral, we integrate termwise and obtain the series in the desired formula. Letting in the second integral gives
[TABLE]
and the following calculation (see [8, Section 6.212]) completes the proof:
[TABLE]
Proposition 2 The following formula for Euler’s constant is valid:
[TABLE]
Proof. In (20) we replace xy with . Substituting the result into (11), we change the order of integration and get
[TABLE]
Using the proof of Proposition 1, we obtain the desired formula.
6 Integrals over higher-dimensional analogs of
There are several ways to generalize the triangle and the integral . The simplest generalization of is the polytope
[TABLE]
Theorem 6 For * and* , the integral
[TABLE]
is equal to an integer linear combination of multiple zeta values of weight , namely,
[TABLE]
It is also equal to a polynomial of several variables with rational coefficients in values of the Riemann zeta function at integers.
Proof. Expanding gives
[TABLE]
Since
[TABLE]
we get
[TABLE]
Evaluating the last integral using formulas (5) and (14), the theorem follows.
Taking , the polytope is the triangle , the integral is the triangle integral , and Theorem 6 reduces to Theorem 1. In particular, formula (21) for is a weighted version of formula (2) for .
Corollary 6 If , then
Proof. Taking in Theorem 6 forces in (21).
There is another, more natural proof of Corollary 6, one that does not use Theorem 6.
Second proof of Corollary 6. We use the representation
[TABLE]
(To prove this formula, expand the integrand in a geometric series and integrate termwise.) We perform the change of variables
[TABLE]
(For , compare this with the transformation of into Beukers’ integral for in the Introduction.) We get
[TABLE]
where the integral is over the polytope defined by , , . By symmetry, we may interchange the variables and if . Using all permutations of , we arrive at
[TABLE]
where the integral is over
[TABLE]
This is the integral , and the second proof is complete.
Remark 3 The integral is equivalent to the Chen (Drinfeld-Kontsevich) integral [16, Section 1]
[TABLE]
over the polytope : set and for .
If we rewrite (21) as
[TABLE]
where denotes the sum of the multinomial coefficients
[TABLE]
then for any fixed one can derive a closed expression for . For example,
[TABLE]
In general, we have
Proposition 3 Fix * and* . Then the integers , ,…, satisfy the recurrence
[TABLE]
Proof. First note that if we denote
[TABLE]
then
[TABLE]
Now note that if is the subset of consisting of those -tuples with exactly numbers among the equal to , then
[TABLE]
Finally, since is the disjoint union
[TABLE]
the proposition follows.
Another generalization of the triangle is the polytope
[TABLE]
Note that it is symmetric in all variables, unlike .
We first generalize the triangle integral to an integral over . Then we extend to a generalization of over for all .
Recall that, for all complex and all with , the polylogarithm is defined by the convergent series
[TABLE]
Theorem 7 If , then the integral
[TABLE]
is equal to a polynomial of several variables with integer coefficients in the values , , and
- for* , namely,
[TABLE]
Proof. The symmetry of yields
[TABLE]
where is the polytope defined by and . Integrating consecutively with respect to , , …, , and setting , , we obtain
[TABLE]
where is the triangle defined by and . (Thus is the upper half of the triangle when bisected by the line .) Since is also defined by and , we see that
[TABLE]
The series expansion
[TABLE]
yields
[TABLE]
Now consider the identity
[TABLE]
If we differentiate times with respect to , and then set , we obtain
[TABLE]
Putting , the theorem follows.
Example 1 Taking gives
[TABLE]
On the other hand, . This proves Euler’s formula for the dilogarithm at [5, Section 1.2], [7, pp. 43-45], [10, Section 1.4]:
[TABLE]
Now take . Using Landen’s formula for the trilogarithm at [10, Equation 6.12],
[TABLE]
we get
[TABLE]
Thus, surprisingly, and are both rational multiples of .
Finally, setting and using the formulas for and , we obtain
[TABLE]
We now generalize to an integral over the polytope . First, we extend the definition of the polylogarithm by defining the multiple polylogarithm
[TABLE]
Theorem 8 If * and* , then the integral
[TABLE]
is equal to a polynomial of several variables with rational coefficients in the values ,
- with* , and * with*
Explicitly, if * and if*
[TABLE]
for , then
[TABLE]
Proof. The proof is similar to that of Theorem 7 (the case .
Notice the equality of the integrals .
As an application of Theorem 8, we obtain the following relation between certain multiple polylogarithm values and multiple zeta values. (The relation can also be deduced from the Hölder convolution formula in [4, Equation (7.2)].)
Corollary 7 If , then
[TABLE]
Proof. In Theorem 8, take and set , so that . Then and
[TABLE]
Now substitute , and apply Theorem 1 (i) and the identity
[TABLE]
Example 2 The case is Euler’s formula (22) for . Taking and substituting gives the relation
[TABLE]
which is a special case of [4, Equation (7.3)]. Using the values of and in (22) and (23), we get the formula
[TABLE]
Finally, adding recovers Ramanujan’s summation [1, p. 258]
[TABLE]
7 References
-
B. C. Berndt, Ramanujan’s Notebooks, Part I, Springer-Verlag, New York, 1985.
-
F. Beukers, A note on the irrationality of and , Bull. London Math. Soc. 11 (1979) 268-272.
-
J. M. Borwein, D. M. Bradley, and D. J. Broadhurst, Evaluations of -fold Euler/Zagier sums: a compendium of results for arbitrary , Electron. J. Combin. 4 (1997) no. 2, R5.
-
J. M. Borwein, D. M. Bradley, D. J. Broadhurst, and P. Lisonek, Special values of multiple polylogarithms, Trans. Amer. Math. Soc. 353 (2001) 907-941.
-
P. Cartier, Fonctions polylogarithmes, nombres polyzêtas et groupes pro-unipotents, Astérisque 282 (2002) 137-173.
-
N. G. de Bruijn, Asymptotic Methods in Analysis, Dover Publications, New York, 1981.
-
W. Dunham, Euler: The Master of Us All, Dolciani Mathematical Expositions No. 22, Mathematical Association of America, Washington, D.C., 1999.
-
S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, 6th edition, A. Jeffrey and D. Zwillinger, editors, Academic Press, San Diego, 2000.
-
K. S. Kölbig, J. A. Mignaco, and E. Remiddi, On Nielsen’s generalized polylogarithms and their numerical calculation, BIT 10 (1970) 38-74.
-
L. Lewin, Polylogarithms and Associated Functions, North-Holland, New York, 1981.
-
J. Ser, Sur une expression de la fonction de Riemann, C. R. Acad. Sci. Paris Sér. I Math. 182 (1926) 1075-1077.
-
J. Sondow, Criteria for irrationality of Euler’s constant, Proc. Amer. Math. Soc. 131 (2003) 3335-3344.
-
J. Sondow, An infinite product for via hypergeometric formulas for Euler’s constant, (2003, preprint); available at http://arXiv.org/abs/math/0306008.
-
J. Sondow, Double integrals for Euler’s constant and and an analog of Hadjicostas’s formula, Amer. Math. Monthly 112 (2005) 61-65.
-
J. Sondow, A faster product for and a new integral for , Amer. Math. Monthly 112 (2005) 729-734.
-
M. Waldschmidt, Multiple polylogarithms: an introduction, in Number Theory and Discrete Mathematics, Proceedings of the International Conference in Honour of Srinivasa Ramanujan, Chandigarh, 2000, A. K. Agarwal et al., editors, Birkhäuser, Basel, 2002, pp. 1-12.
-
S. Zlobin, On a certain integral over a triangle (2005, preprint); available at
