where $H_n$ is the harmonic number and can be defined as:



I managed to prove $\quad\displaystyle\sum_{n=1}^\infty\frac{(-1)^nH_n^{(3)}}{n^2}=\frac{21}{32}\zeta(5)-\frac34\zeta(2)\zeta(3)\quad$ using logarithmic integral and couple harmonic identities. other approaches are much appreciated.

you can find a different solution by Cornel in his book (Almost impossible integrals, sum and series).

  • $\begingroup$ @ Ali Shather Over the last couple of weeks you have provided interesting results for special cases. What about explaining your general method and showing its capabiities in solving a class of problems, e.g. for $S(p,q) = \sum_{k=1}^\infty (-1)^k H_{k}^{(p)}/k^q$ and answer questins such as "Which combinations of $,p$ and $q$ permit closed expressions?" and "What is new in these results?" $\endgroup$ Jun 8, 2019 at 16:40
  • $\begingroup$ @ Dr. Wolfgang Hintze thank you. actually I solved these sums over a year ago and I posted some of them on my facebook group ( harmonic series) but I've been posting them here lately as I've just learned LaTex. about your question, I know that Bastien derived the generalization for S(1,q) here arxiv.org/pdf/… where q is even. $\endgroup$ Jun 8, 2019 at 18:04
  • 1
    $\begingroup$ @ Dr. Wolfgang and couple days ago Cornel derived the generalization for S( p,1) here researchgate.net/publication/…. The generalization of S(p,q) is extremely hard and its not in the mathematical literature yet. I do not see myself good enough to investigate it and I am just focused on providing new solutions to particular cases. $\endgroup$ Jun 8, 2019 at 18:04
  • $\begingroup$ @ Ali Shather From the work of Bastien (thanks for the hint) I conclude that $S(1,5)=\sum_{k=1}^{\infty} (-1)^k \frac{H_{k}}{k^5}$ is the first sum which seems not to have a closed form (in terms). So this is a real challenge. $\endgroup$ Jun 8, 2019 at 21:56
  • $\begingroup$ @Dr.WolfgangHintze no problem. Yes it's pretty tough and that's why mathematicians still struggle with it. $\endgroup$ Jun 8, 2019 at 22:14

1 Answer 1


using the fact that $\displaystyle\sum_{n=1}^\infty z^nH_n^{(3)}=\frac{\operatorname{Li}_3(z)}{1-z}\ $ divide both sides by $z$ then integrate from $z=0$ to $x$, we get $$\sum_{n=1}^\infty \frac{x^nH_n^{(3)}}{n}=\operatorname{Li}_4(x)-\ln(1-x)\operatorname{Li}_3(x)-\frac12\operatorname{Li}_2^2(x)\tag{1}$$

replace $x$ with $-x$ in(1), then divide both sides by $x$ and integrate from $0$ to $1$, we get

\begin{align} \sum_{n=1}^\infty\frac{(-1)^nH_n^{(3)}}{n^2}&=\operatorname{Li}_5(-1)-\underbrace{\int_0^1\frac{\ln(1+x)\operatorname{Li}_3(-x)}{x}\ dx}_{IBP}-\frac12\int_0^1\frac{\operatorname{Li}_2^2(-x)}{x}\ dx\\ &=\operatorname{Li}_5(-1)+\operatorname{Li}_2(-1)\operatorname{Li}_3(-1)-\frac32\int_0^1\frac{\operatorname{Li}_2^2(-x)}{x}\ dx\\ &=\frac38\zeta(2)\zeta(3)-\frac{15}{16}\zeta(5)-\frac32\int_0^1\frac{\operatorname{Li}_2^2(-x)}{x}\ dx\\ \end{align} I proved here \begin{align} \int_0^1\frac{\operatorname{Li}_2^2(-x)}{x}\ dx=\frac34\zeta(2)\zeta(3)-\frac{17}{16}\zeta(5) \end{align} which follows that



By setting $x=-1$ in (1) we have



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.