6
$\begingroup$

Question:

Evaluate

$$\sum_{k\geq 1} \frac{(H'_k)^2}{k^2}$$

Where we define the alternating harmonic number

$$H'_k=\sum_{n=1}^k\frac{(-1)^n}{n}$$

I remember seeing a closed form involving a quadrilogarithm.

I am interested in knowing the full solution, if possible.

$\endgroup$
0

2 Answers 2

3
$\begingroup$

This is going to be almost the full solution because I lack time know. We start from the following identity: \begin{equation} H_k^{'} = (-\log(2)) (1-(-1)^k) + \int\limits_0^{-1} Li_1(\xi) k \xi^{k-1} d\xi \end{equation} Now we square the above and divide by $k^2$ and sum from one to infinity. We end up with three terms. The first one is trivial the second one is a one dimensional integral which is doable because it contains only a product of logs and the last term is a two dimensional integral. We have: \begin{eqnarray} &&\sum\limits_{k=1}^\infty \frac{[H_k^{'}]^2}{k^2} = \\ &&\log(2)^2 \frac{\pi^2}{2} - 2 \log(2) \frac{7}{8} \zeta(3) + \int\limits_{[0,-1]^2} \frac{\log(1-\xi_1) \log(1-\xi_2) }{1-\xi_1 \xi_2} d\xi_1 d\xi_2=\\ &&3 \text{Li}_4\left(\frac{1}{2}\right)+\frac{7}{4} \zeta (3) \log (2)-\frac{43 \pi ^4}{1440}+\frac{\log ^4(2)}{8}+\frac{3}{8} \pi ^2 \log ^2(2)-\int\limits_0^{-1} \frac{\log(1-\xi_1) Li_2(\frac{2 \xi_1}{-1+\xi_1})}{\xi_1} d\xi_1 =\\ &&3 \text{Li}_4\left(\frac{1}{2}\right)+\frac{7}{4} \zeta (3) \log (2)-\frac{43 \pi ^4}{1440}+\frac{\log ^4(2)}{8}+\frac{3}{8} \pi ^2 \log ^2(2)+\\ &&\int\limits_0^1 \log(1-\frac{u}{2}) Li_2(u) \left(\frac{1}{2-u} + \frac{1}{u}\right)d u=\\ &&3 \text{Li}_4\left(\frac{1}{2}\right)+\frac{7}{4} \zeta (3) \log (2)-\frac{7 \pi ^4}{160}+\frac{\log ^4(2)}{8}+\frac{3}{8} \pi ^2 \log ^2(2) + \sum\limits_{m=1}^\infty \frac{H_m^2}{m^2} \frac{1}{2^m} \end{eqnarray} Now in the third line from the top we integrated over $\xi_2$ and then used Landen's identity and then carried out the doable integrals. In the fourth line from the top we substituted for $2 \xi_1/(-1+\xi_1)$ and in the last line we evaluated the remaining integral by expanding the integrand in a series and integrating term by term. Now we use my answer to this question Double harmonic sum $\sum_{n\geq 1}\frac{H^{(p)}_nH_n}{n^q}$ . We have: \begin{eqnarray} \sum\limits_{m=1}^\infty \frac{H_m^2}{m^2} \frac{1}{2^m} &=& -\frac{1}{3} \int\limits_0^{\frac{1}{2}} \frac{[\log(1-\xi)]^3}{\xi}d\xi + \frac{1}{2} [Li_2(\frac{1}{2})]^2 + Li_4(\frac{1}{2}) \\ &=&-\frac{1}{3} \int\limits_{\frac{1}{2}}^1 \frac{Li_0(\xi)}{\xi} \log(\xi)^3d\xi + \frac{1}{2} [Li_2(\frac{1}{2})]^2 + Li_4(\frac{1}{2}) \\ &=& -\text{Li}_4\left(\frac{1}{2}\right)-\frac{7}{4} \zeta (3) \log (2)+\frac{37 \pi ^4}{1440}-\frac{\log ^4(2)}{24}+\frac{1}{24} \pi ^2 \log ^2(2) \end{eqnarray} Bringing everything together we have: \begin{eqnarray} \sum\limits_{k=1}^\infty \frac{[H_k^{'}]^2}{k^2} =2 \text{Li}_4\left(\frac{1}{2}\right)-\frac{13 \pi ^4}{720}+\frac{\log ^4(2)}{12}+\frac{5}{12} \pi ^2 \log ^2(2) \end{eqnarray} as expected.

Note: As usual in those cases the consecutive terms correspond to decompositions of the weight (in this case number four) into sums of integers. Apart from the first term on the rhs we have $(4)$, $(1,1,1,1)$ and $(2,1,1)$. Oddly enough the term $(3,1)$ has canceled out in this case. Why?!

$\endgroup$
3
  • $\begingroup$ Could you give a reference for the first identity? $\endgroup$ Commented Jun 8, 2017 at 10:58
  • $\begingroup$ @user153012: Do you mean the very first identity in this body of text? This follows just from integration by parts. $H_k^{'} = \int\limits_0^{-1} \frac{1-\xi^k}{1-\xi} d\xi = \left. Li_1(\xi) (1-\xi^k) \right|_0^{-1} + \int\limits_0^{-1} Li_1(\xi) k \xi^{k-1} d\xi$ $\endgroup$ Commented Jun 8, 2017 at 11:03
  • $\begingroup$ I thought you could give some good literature on the topic of alternating harmonic numbers and its representations, but thank you for the details. Nice answer. $\endgroup$ Commented Jun 8, 2017 at 11:15
2
$\begingroup$

Numerically, this is $$ -\frac{13}{8}\zeta(4)+\frac5{2}\zeta(2)\log^22+\frac1{12}\log^42+2\mathrm{Li}_4({\textstyle\frac12}), $$ from "Experimental Evaluation of Euler Sums" by Bailey, Borwein and Girgensohn.

This can be found with an integer relation algorithm applied to the sum evaluated to high precision using this integral: $$ \int_0^1\frac{\log(\frac1z)dz}{z(1-z)}\left(-\zeta(2)+\log^22+2\log(1-z)\log\left(\frac{1+z}{2}\right)+2\mathrm{Li}_2\left(\frac{1-z}{2}\right)+\mathrm{Li}_2(z)\right). $$

$\endgroup$
4
  • $\begingroup$ Thanks , for your input. So the value has not been proved analytically? $\endgroup$ Commented Jun 19, 2014 at 15:55
  • $\begingroup$ @ZaidAlyafeai In the linked paper, they say they proved it (your sum is almost the same as $s_a(2,2)$). $\endgroup$ Commented Jun 19, 2014 at 16:04
  • $\begingroup$ But there is no solution in the paper. $\endgroup$ Commented Jun 19, 2014 at 17:07
  • $\begingroup$ @ZaidAlyafeai Yes, I know, but they marked it as such in the table, so I assume they proved it. They cite Lewin in the text. $\endgroup$ Commented Jun 19, 2014 at 17:08

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.