Given the nth harmonic number $ H_n = \sum_{j=1}^{n} \frac{1}{j}$. In this post it asks for the evaluation,


while this post and this answer discusses the next one,

$$ \sum_{n=1}^{\infty}\frac{H_n}{n^4} = -\zeta(2)\zeta(3)+3\zeta(5)$$

Given the more general sum,

$$F_k(z) = \sum_{n=1}^{\infty}\frac{H_n}{n^k}z^n\tag1$$

It seems the special cases $z=1$ is,

$$F_k(1)= \sum_{n=1}^{\infty}\frac{H_n}{n^k} = S_{k-1,2}(1)+\zeta(k+1)\tag2$$

while $z=-1$ is,

$$F_k(-1)= \sum_{n=1}^{\infty}\frac{H_n}{n^k}(-1)^n = S_{k-1,2}(-1)-\frac{2^k-1}{2^k}\zeta(k+1)\tag3$$

where $S_{n,p}(z)$ is the Nielsen generalized polylogarithm,

$$S_{n,p}(z) = \frac{(-1)^{n+p-1}}{(n-1)!\,p!}\int_0^1\frac{(\ln t)^{n-1}\big(\ln(1-z\,t)\big)^p}{t}dt$$

However, for the range $-1\leq z \leq 1$, the closely related sum in this answer has a simple formula,

$$G_k(z) = \sum_{n=1}^{\infty}\frac{H_n}{(n+1)^k}\,z^{n+1} = S_{k-1,2}(z)\tag4$$

Q: Like $G_k(z)$, does $F_k(z) = \sum_{n=1}^{\infty}\frac{H_n}{n^k}z^n$ have a common closed-form in the range $-1\leq z \leq 1$?

  • $\begingroup$ @ Tito Piezas III I have enumerated $F_k(+1)$ for $k=1,\cdots,11$ and also got some results on $F_k(z)$ in math.stackexchange.com/questions/2169507/… . I guess you ponder on what is the relationship between Nielsen poly-logs and poly-logs. In general I do not know that. I lack motivation to further investigate those things. Can i ask you why are you interested in those things? $\endgroup$
    – Przemo
    May 31 '19 at 11:01
  • $\begingroup$ @Przemo:I must confess it is pure curiosity. I came across the Nielsen polylogs while trying to solve this log sine integral. $\endgroup$ May 31 '19 at 14:16

After my last edit, I figured out a way to partially answer my question. The trick is to test,

$$F_k(z) - G_k(z) = \sum_{n=1}^{\infty}\frac{H_n}{n^k}z^n - \sum_{n=1}^{\infty}\frac{H_n}{(n+1)^k}z^{n+1}$$

for various values of $k,z$ to see if it yields something familiar. For $k=2$ and value $z = 1/3$, the Inverse Symbolic Calculator was able to recognize it as,

$$F_2\big(\tfrac13\big) - G_2\big(\tfrac13\big) = \rm{Li}_3\big(\tfrac13\big)$$

A little more testing showed that for $-1\leq z\leq 1$, apparently,

$$F_k(z) = \sum_{n=1}^{\infty}\frac{H_n}{n^k}z^n= S_{k-1,2}(z) + S_{k,1}(z)$$

with Nielsen generalized polylogarithm $S_{n,p}(z)$. Equivalently, in terms of polylogarithm $\rm{Li}_n(z)$,

$$F_k(z) = \sum_{n=1}^{\infty}\frac{H_n}{n^k}z^n= S_{k-1,2}(z) + \rm{Li}_{k+1}(z)$$

For the special cases $z=1$ and $z=-1$, the polylogarithm reduces to formulas $(2)$ and $(3)$ in the post.

P.S. Of course, what remains is to rigorously prove the proposed formula.

  • $\begingroup$ The case when $z=\frac12$ for $F_2(z)$ and $F_3(z)$ is discussed in this and this post. $\endgroup$ Jun 1 '19 at 6:37
  • $\begingroup$ Hmm in fact you simply noticed that $$\sum_{n=1}^{\infty}\frac{H_n}{(n+1)^k}z^{n+1}=\sum_{m=2}^{\infty}\frac{H_{m-1}}{m^k}z^m=\sum_{m=2}^{\infty}\frac{H_{m}-\frac 1m}{m^k}z^m=\sum_{m=1}^{\infty}\frac{H_{m}}{m^k}z^m-\sum_{m=1}^{\infty}\frac 1{m^{k+1}}z^m$$ (sorry...) $\endgroup$ Jun 1 '19 at 18:52
  • $\begingroup$ @RaymondManzoni: No problem. The important thing is the observation is correct. Furthermore, it makes explicit the connection between those harmonic sums and the Nielsen polylogs. $\endgroup$ Jun 3 '19 at 15:49

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.