0

How can I find a closed form formula for the following sum?
For $p\in\mathbb{Z}^+\land q\in\mathbb{Z}^+\setminus\{1\}$,
\begin{equation*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^p}{k^q} \end{equation*} Only if $p\leqslant2$ and $q\leqslant2$, this sum is following: \begin{equation*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^p}{k^q}=\dfrac{(pq-q\delta_{p,1})^2+1}{p+q-q\delta_{p,1}}\zeta(p+q) \end{equation*} where $H_k$ is following \begin{equation*} H_k=\displaystyle\sum_{j=1}^{k}\dfrac{1}{j} \end{equation*} When $p=1$ and $q=2$,
\begin{equation*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^1}{k^2}=2\zeta(3) \end{equation*} When $p=2$ and $q=2$,
\begin{equation*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^2}{k^2}=\dfrac{17}{4}\zeta(4) \end{equation*}

However, when ($p\geqslant3\land q\geqslant2$) or ($p\geqslant2\land p\geqslant3$), I've failed.
I think that my formula is wrong for all $p$ and $q$.
I've seen this answer.
Thank you.
p.s.
If $1\leqslant p\leqslant2$ and $2\leqslant q\leqslant4$, \begin{align*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^p}{k^q}=\dfrac{(pq-q\delta_{p,1})^2+1}{p+q-q\delta_{p,1}}\zeta(p+q) -\sum _{k=1}^{q-2} \left(\frac{1}{4} (-1)^{q+1} \binom{q+1}{k} \zeta (k+p) \zeta (q-k)+\frac{\left((p q)^2+3\right) \zeta (p+q)}{2 (p+q)}\right)-\frac{1}{2} q \left(\sum _{k=1}^{p-2} \zeta (p-k) \zeta (k+q)\right)-\sum _{k=1}^{q-3} \frac{(p q+5) \zeta (k+p)^2}{q} \end{align*} when $p=2$ and $q=3$, \begin{equation*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^2}{k^3}=-\zeta(2)\zeta(3)+\dfrac{7}{2}\zeta(5) \end{equation*} when $p=2$ and $q=4$, \begin{equation*} \displaystyle\sum_{k=1}^\infty\dfrac{H_k^2}{k^4}=\dfrac{97\pi^6}{22680}-2\{\zeta(3)\}^2 \end{equation*}

sakura
  • 11
  • There is no known closed form for this general sum –  Jun 01 '25 at 12:08
  • @Dqrksun I think involving MZVs. – sakura Jun 01 '25 at 12:32
  • By the way, given $p$, for certain odd and evenness there is a closed form in terms of lower $p$ –  Jun 01 '25 at 16:14
  • @Dqrksun It's difficult for open? – sakura Jun 01 '25 at 21:43
  • If $q=1$ the sums diverge. – metamorphy Jun 02 '25 at 06:04
  • @Sakura what do you mean by open? –  Jun 02 '25 at 07:00
  • Please show us the case of given $p$, for certain odd and evenness there is a closed form in terms of lower $p$. – sakura Jun 02 '25 at 07:03
  • @metamorphy really? MZSum[HarmonicNumber[n]/n /. {n -> m + 1}, {m, 0, Infinity}] this result is $\dfrac{\zeta(2)}{2}$ with mathematica 14.2. – sakura Jun 02 '25 at 07:06
  • Perhaps as a result of highly optimized calculations, Mathematica is not so good, in that you never know what you're working with, what literature you're using, or what formulas you're using along the way. – sakura Jun 02 '25 at 07:15
  • @metamorphy you're right! It's diverge! I got very large number! – sakura Jun 02 '25 at 07:41
  • $$\sum _{k=1}^{\infty } \frac{\left(H_k\right){}^2}{k^x}=2 \text{MultiZeta}({x,1,1})+2 \text{MultiZeta}({x+1,1})+\text{MultiZeta}({x,2})+\zeta (x+2)$$ MultiZeta see here: https://en.wikipedia.org/wiki/Multiple_zeta_function – Mariusz Iwaniuk Jun 03 '25 at 07:16
  • @MariuszIwaniuk Oh! Thank you so much! – sakura Jun 03 '25 at 07:40
  • @sakura In Mathematica using package: MultipleZetaValues from here: https://www.researchgate.net/publication/357601353_Mathematica_package_MultipleZetaValues .Example code:2 MultiZeta[{x, 1, 1}] + 2 MultiZeta[{x + 1, 1}] + MultiZeta[{x, 2}] + Zeta[x + 2] /. x -> 2 // MZExpand. gives: $\frac{17 \pi ^4}{360}$. – Mariusz Iwaniuk Jun 03 '25 at 09:07
  • @MariuszIwaniuk 'MZSum[(HarmonicNumber[n])^2/n^2 /. {n -> l + 1}, {l, 0, Infinity}]' this result is $\frac{17\pi^2}{360}$. Thank you! – sakura Jun 03 '25 at 23:56

0 Answers0