I hope it is OK to ask the following reference request. If my question is not suitable, please let me know and I will do my best to modify it!

Let $N\in\mathbb{N}$, let $q$ be a point in the open unit disc in the complex plane, and let $0\leq m \leq N$ be an integer. Some personal research I have been doing on iterated integrals of Eisenstein series has led me to consider certain series of the following form: $$S_m(N;q):=\sum_{k\geq 1}\frac{\sigma_N(k)}{k^N} k^m q^k,$$ where $\sigma_N(k):=\sum_{d\vert k} d^N$ is the $N$th power sum-of-divisors function.

There are two "extreme" cases that connect to well-known series: $m=0$ and $m=N$. In the $m=0$ case we recover a Lambert series $$S_0(N;q) = \sum_{k\geq 1}\frac{q^k}{k^N(1-q^k)}$$ and in the $m=N$ case (for $N\geq 3$ odd) we recover the nonconstant part of the $\mathbb{Q}$-normalised Eisenstein series $\mathbb{G}_{N+1}(\tau)$, with $q=\exp(2\pi i \tau)$ and $\tau$ in the upper half plane. Another connection to well-known objects comes from taking the limit $$\lim_{q\to 1}S_m(N;q) = \zeta(-m)\zeta(N-m).$$

In the cases $1\leq m\leq N-1$ I do not have a good name for these series, and I am not aware of them in the literature on modular forms/$L$-functions. I hope it is OK to ask: have these series been studied before, and are there any references for them if so? Is anything known about relations between the $S_m(N;q)$ (for a fixed $N$), or the dimension of the $\mathbb{Q}$-vector space generated by $\pi^m S_m(N;q)$ for $0\leq m\leq N$?

In fact, I am interested primarily not in the "functional" case, but with the specific case $q = e^{-2\pi}$. As an example of what I mean about relations in this case, Ramanujan's formula for $\zeta(N)$ in terms of a Lambert series gives a $\mathbb{Q}$-relation between $\zeta(N)$, $\pi^N$ and $S_0(N;e^{-2\pi})$ (at least for $N\equiv 3\pmod{4}$). I have been studying this case in my research and have found more relations between these numbers as well as $\pi^m S_m(N;e^{-2\pi})$ for larger $m$. If anything can be said about this case, I would be very grateful!


The sum which you consider could be rewritten as a double sum of the following:
\begin{align} S_{m}(N,q)=\sum_{i,j\ge 1}i^{m-N}j^me^{-2ij\alpha}:=f_{N,m}(\alpha) \end{align} by setting $q=e^{-2\alpha}$. It is easy to check that $$f_{N,m}(\alpha)=(-1)^m\frac{1}{2^m}\frac{\,d^m}{\,d \alpha^m}\sum_{i,j\ge 1}i^{-N}e^{-2ij\alpha}=\frac{(-1)^m}{2^m}\frac{\,d^m}{\,d \alpha^m}\sum_{k\ge 1}\frac{1}{k^N(e^{2k\alpha}-1)}.$$ For $N=2n+1\ge 3$ be an odd positive integer, we have the Ramanujan’s formula for $\zeta(2n+1)$, (which could be find in [Bruce C Berndt: Ramanujan's Notebooks: Part IV]) \begin{align} \frac{1}{\alpha^{n}}\left(\frac{\zeta(2n+1)}{2} + \sum_{m=1}^\infty \frac{1}{m^{2n+1}(e^{2m\alpha}-1)}\right)\quad- \frac{1}{(-\beta)^{n}}\left(\frac{\zeta(2n+1)}{2} + \sum_{m=1}^\infty \frac{1}{m^{2n+1}(e^{2m\beta}-1)}\right)\\ =2^{2n}\sum_{k=0}^{n+1} (-1)^{k-1}\frac{B_{2k}}{(2k)!}\frac{B_{2n+2-2k}}{(2n+2-2k)!}\alpha^{n+1-k}\beta^k, \end{align} where $\alpha,\beta>0$ such that $\alpha\beta=\pi^2$ and $B_r , r\ge 0$, denote the $r$-th Bernoulli number. Which means $f_{N,m}(\alpha)$ is a quasi-modular function. We also have the modularity for the cases $N=2n$ is even follows the modularity of $$\sum_{k\ge 1}\frac{1}{k^{2n}(e^{2k\alpha}-1)}.$$ However, the above function is not modular or quasi-modular.

  • $\begingroup$ Thanks for your answer! Unfortunately I am a little unclear about your point regarding $f_{N,m}(\alpha)$ being quasi-modular (for $N=2n$ even). You mention that quasi-modularity of $f_{N,m}(\alpha)$ follows from modularity of $\sum k^{-2n}(e^{2k\alpha}-1)^{-1}$, but you then say that this function is not (quasi-)modular. So is $f_{N,m}(\alpha)$ only modular when $N$ is odd? $\endgroup$ – Alex Saad Aug 23 '18 at 12:40
  • 1
    $\begingroup$ Yes, $f_{N,m}(\alpha)$ is modular only $N$ is odd. $\endgroup$ – Zhou Aug 23 '18 at 12:56
  • $\begingroup$ Great - thank you! $\endgroup$ – Alex Saad Aug 23 '18 at 12:57
  • $\begingroup$ In this case, is it possible to say what the weight of $f_{N,m}(\alpha)$ is? $\endgroup$ – Alex Saad Aug 23 '18 at 13:16
  • $\begingroup$ I am not sure, but I think this article [Ramanujan’s Formula for $\zeta(2n+1)$, arxiv.org/pdf/1701.02964.pdf ] gives the answer which you need, possibly. $\endgroup$ – Zhou Aug 23 '18 at 14:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.