Multiple polylogarithms, a regularisation process and an admissible open domain of convergence

Pawan Singh Mehta and Biswajyoti Saha · 2025

Read Multiple polylogarithms, a regularisation process and an admissible open domain of convergence by Pawan Singh Mehta and Biswajyoti Saha (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-00889.txt.

Introduction

Pawan Singh Mehta and Biswajyoti Saha study multiple polylogarithms, a family of functions built from nested series that appear across number theory and physics. The region where the defining series converges absolutely is known, but the wider region of conditional convergence has been little explored, especially at depth two and beyond.

Using translation formulas satisfied by these functions, they exhibit a larger open domain where the series converges and defines a holomorphic function. They then build a regularisation process, extending earlier work by Saha, which needs a generalized Euler-Boole summation formula proved in an appendix. The process extends the domain further at integer points, setting up a study of local behavior there.

Catalog record

CREATORPawan Singh Mehta and Biswajyoti Saha
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-00889

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Multiple polylogarithms, a regularisation process and an admissible open domain of convergence Pawan Singh Mehta and Biswajyoti Saha Abstract . In this article, we study the analytic properties of the multiple polylogarithms in the s s -aspect. Although the domain of absolute convergence of the series defining the multiple polylogarithms is well-known, the study towards a larger open domain of (conditional) convergence has been limited, particularly when the depth is ≥ 2 \geq 2 . Here, we exhibit a larger open domain of (conditional) convergence for this series by writing certain translation formulas satisfied by them. The series moreover defines a holomorphic function in this open set. We then introduce a regularisation process for the multiple polylogarithms, extending an earlier work of the second author. This regularisation process requires a generalisation of the Euler-Boole summation formula that we derive in the appendix of this article. The regularisation process leads to a larger open domain, where the series (conditionally) converges at integer points. The holomorphicity at such points is a more delicate question and this regularisation process is to be used to study the local behaviour of the multiple polylogarithms around such points. 2020 Mathematics Subject Classification 11M32 1. Introduction Throughout this article, we denote the set of all non-negative integers by {\mathbb{N}} . Let a be a positive integer and z be a complex number with \lvert z\rvert<1 . The classical polylogarithm function is defined by the convergent power series \mathrm{Li}(a;z):=\sum_{n>0}\frac{z^{n}}{n^{a}}. Convergence of such a series depends on z and a both. For example, we can take \lvert z\rvert\leq 1 , if a\geq 2 . However, for \lvert z\rvert>1 , the series is not convergent. For a fixed value of z with \lvert z\rvert\leq 1 , the polylogarithm functions can be considered as the Dirichlet series with the integer a replaced by a complex variable s such that \Re(s)>1 , i.e. \mathrm{Li}_{z}(s):=\sum_{n>0}\frac{z^{n}}{n^{s}}. Note that the above Dirichlet series converges (absolutely) for every complex number s if |z|<1 . Now if |z|=1 , the domain of convergence of this Dirichlet series depends on the value of z . For z=1 , we get back the ubiquitous Riemann zeta function, which converges (absolutely) for \Re(s)>1 . However, for |z|=1 with z\neq 1 , the Dirichlet series \mathrm{Li}_{z}(s) converges for \Re(s)>0 . This is immediate as the partial sum \sum_{N>n>0}z^{n}=\frac{z-z^{N}}{1-z} is bounded. This also shows that in this case, the Dirichlet series \mathrm{Li}_{z}(s) does not converge for any s with \Re(s)<0 . The multiple polylogarithms are defined by considering several variable analogue of the classical polylogarithm functions. For positive integers r , a_{1},\ldots,a_{r} and complex number z_{1},\ldots,z_{r} with |z_{i}|<1 for all 1\leq i\leq r , the multiple polylogarithm (of depth r ) is defined by the convergent power series \mathrm{Li}(a_{1},\ldots,a_{r};z_{1},\ldots,z_{r}):=\sum_{n_{1}>\cdots>n_{r}>0}\frac{z_{1}^{n_{1}}\cdots z_{r}^{n_{r}}}{n_{1}^{a_{1}}\cdots n_{r}^{a_{r}}}. Again, this definition can be extended for |z_{i}|\leq 1 for all 1\leq i\leq r , if a_{1}\geq 2 . Further, this series can be considered as a multiple Dirichlet series by fixing complex numbers z_{1},\ldots,z_{r} and replacing a_{1},\ldots,a_{r} by complex variables s_{1},\ldots,s_{r} with \Re(s_{1}+\cdots+s_{i})>i for 1\leq i\leq r , i.e., (1) \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}):=\sum_{n_{1}>\cdots>n_{r}>0}\frac{z_{1}^{n_{1}}\cdots z_{r}^{n_{r}}}{n_{1}^{s_{1}}\cdots n_{r}^{s_{r}}}, where (s_{1},\ldots,s_{r})\in U_{r} with U_{r}:=\{(s_{1},\ldots,s_{r})\in{\mathbb{C}}^{r}~:~\Re(s_{1}+\cdots+s_{i})>i~\text{ for all }~1\leq i\leq r\}. The above series converges normally on compact subsets of U_{r} and therefore defines a holomorphic function there (see [ 7 , Proposition 2] ). The second author further showed in [ 7 , Theorem 9] (albeit using different notations) that the multiple Dirichlet series ( 1 ) can be extended to a meromorphic function on the whole of {\mathbb{C}}^{r} and the set of singularities depends on the values of the product z_{[1,i]}:=z_{1}\cdots z_{i} for 1\leq i\leq r . Theorem 1 (Saha) . The function \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}) extends to a meromorphic function on the whole of {\mathbb{C}}^{r} . If z_{[1,i]}\neq 1 for all 1\leq i\leq r , then \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}) is holomorphic on {\mathbb{C}}^{r} . Otherwise, if i_{1}<\cdots<i_{m} denote all the indices such that z_{[1,i_{j}]}=1 for all 1\leq j\leq m , then the set of all possible singularities of \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}) can be given as follows: (a) If i_{1}=1 , then \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}) is holomorphic outside the union of the hyperplanes given by the equations s_{1}=1;~s_{1}+\cdots+s_{i_{j}}=n\ \text{ for all }n\in{\mathbb{Z}}_{\leq j}\ \text{and }2\leq j\leq m. (b) If i_{1}\not=1 , then \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}) is holomorphic outside the union of the hyperplanes given by the equations s_{1}+\cdots+s_{i_{j}}=n\text{ for all }n\in{\mathbb{Z}}_{\leq j}\text{ and }1\leq j\leq m. This indicates that there could be an open domain of convergence bigger than U_{r} for the series in ( 1 ), while treating it as the limit (2) \lim_{N\to\infty}\sum_{N>n_{1}>\cdots>n_{r}>0}\frac{z_{1}^{n_{1}}\cdots z_{r}^{n_{r}}}{n_{1}^{s_{1}}\cdots n_{r}^{s_{r}}}, depending on the values of z_{i} , exactly as in the case of depth r=1 . This question does not seem to have been addressed before. If all the z_{i} ’s are of unit modulus, the convergence of the series in ( 1 ) outside U_{r} will be conditional convergence and hence we follow the above convention for the convergence going forward. It is easy to see that for s\in{\mathbb{C}} with \Re(s)>1 , the series \sum_{n_{1}>n_{2}>0}\frac{(-1)^{n_{2}}}{n_{1}^{s}} converges and the sum is -2^{-s}\zeta(s) . One can also see that the series \sum_{n_{1}>n_{2}>0}\frac{(-1)^{n_{1}}}{n_{1}} does not converge, although the multiple Dirichlet series \mathrm{Li}_{(-1,1)}(s_{1},s_{2}) is holomorphic everywhere in {\mathbb{C}}^{2} . Hence there are some inherent intricacies. This brings us to our first theorem. We need some notations. For r\geq 1 an integer and {\bf z}:=(z_{1},\ldots,z_{r})\in{\mathbb{C}}^{r} where |z_{i}|\leq 1 for all 1\leq i\leq r , let q({\bf z}) be the smallest positive integer such that 1\leq q({\bf z})\leq r and the product z_{[1,q({\bf z})]}\neq 1 . If no such q({\bf z}) exists, then we set q({\bf z})=r+1 . We define an open set U_{r}(\mathbf{z}) of \mathbb{C}^{r} as follows: (3) \begin{split}U_{r}(\mathbf{z}):=\{&(s_{1},\ldots,s_{r})\in\mathbb{C}^{r}:\Re(s_{1}+\cdots+s_{i})>i\text{ if }1\leq i<q({\bf z})\\ &\text{ and }\Re(s_{1}+\cdots+s_{i})>i-1\text{ if }q({\bf z})\leq i\leq r\}.\end{split} Note that in general, we have U_{r}\subseteq U_{r}(\mathbf{z}) . Further, U_{r}(\mathbf{z})=U_{r} if and only if z_{i}=1 for all 1\leq i\leq r (i.e., q({\bf z})=r+1 ). In this context, we prove the following theorem. Theorem 2 . With the notations and convention as above, the multiple Dirichlet series in ( 1 ) converges on U_{r}(\mathbf{z}) . Moreover, it defines a holomorphic function on U_{r}(\mathbf{z}) . This theorem therefore extends the known result for the depth 1 Dirichlet series \mathrm{Li}_{z}(s) . Also, in the last example, note that the point (1,0) lies on the boundary of U_{2}(-1,1) . However, this open domain of convergence is not optimal in general and we can do better. For example, as N tends to \infty , one has (4) \sum_{N>n_{1}>n_{2}>0}\frac{(-1)^{n_{2}}}{n_{1}^{2}n_{2}^{-1}}=\frac{\log 2}{2}-\frac{\pi^{2}}{16}+o(1). Clearly, (2,-1)\notin U_{2}(1,-1) . In fact, as we can see, for s\in{\mathbb{C}} with \Re(s)>1 , the limit \lim_{N\to\infty}\sum_{N>n_{1}>n_{2}>0}\frac{(-1)^{n_{2}}}{n_{1}^{s}n_{2}^{-1}} exists. Our next aim in this article is therefore to establish a more accurate and larger open domain where the series in ( 1 ) (viewed as the limit in ( 2 )) converges at such interesting integer points. To do this we need to set up a regularisation process for the multiple polylogarithms (at roots of unity), following the work of the second author [ 8 ] . This is done in the next section. Using this regularisation process, for roots of unity z_{1},\ldots,z_{r} , we can give a more accurate and larger open domain where the multiple Dirichlet series \mathrm{Li}_{(z_{1},\ldots,z_{r})}(s_{1},\ldots,s_{r}) converges at integer points. The optimality of

Read the complete text