11

Consider a unit square integral $$I=\int _0^1\int _0^1 \frac{\sqrt{x^2+y^2}}{1-xy} dxdy \approx 1.4843079$$ Many unit square integrals admit nice closed forms. But with this one, standard things like expressing $\frac{1}{1-xy}$ as a geometric series or a change to polar coordinates don't seem to work, so I wonder if this integral actually does have a closed form.

Both Mathematica and Maple's identify functions couldn't guess that hypothetical closed form.

Nikitan
  • 801
  • 5
    I was able to boil the problem down to evaluating $\int_{0}^{1}\mathrm{d}x,\frac{x^{2}}{\sqrt{1+x^{4}}}\operatorname{artanh}{\left(x\right)}$. I'm stuck here, but it seems a much friendlier place to be stuck.... – David H Dec 24 '24 at 00:36
  • @DavidH Using the Maclaurin series expansion for $\operatorname{artanh}(x)$ and integrating gives $\sum_{n=1}^{\infty}\frac{2F_1\left(\frac{1}{2},\frac{n+1}{2};\frac{n+3}{2};-1\right)}{2(n+1)(2n-1)}$, then rearranging using the same process starting from equation $(2)$ in my answer below gives $\sum{k=0}^{\infty}\frac{(-1/4)^k(H_{2k+1}+2\ln{2})(2k)!}{2(4k+3)(k!)^2}$. – teadawg1337 Dec 24 '24 at 03:48
  • Mathematica gives the $\ln$ part as $\frac{\ln{2}}{3}{2F_1!\left(\frac{1}{2},\frac{3}{4};\frac{7}{4};-1\right)}$, whereas Sage gives a very different, but numerically equivalent, answer of $-2^{-5/4}(-1)^{5/8}\ln{(2)},\Gamma(3/4),P{-1/4}^{-3/4}(\sqrt{2})$, where $P_n^m(z)$ denotes the associated Legendre function of the first kind. It's not clear to me at all how these two directly relate, which suggests to me that there may be an equivalent, simpler form. – teadawg1337 Dec 24 '24 at 03:49
  • Soltion for the integral It can be expressed by: Srivastava - Daoust multivariable hypergeometric function – Mariusz Iwaniuk Feb 23 '25 at 19:38

3 Answers3

10

This is not a complete answer; instead, it improves upon L. F.'s previous answer by giving a more approachable form and providing a full derivation for it (see below for a progress update). Previously, L. F. derived

$$I=\sum_{n = 1}^{\infty} \frac{2^{1-\frac{n}{2}}}{n(2n + 1)} \, {_2F_1 \Bigl(\frac{n}{2}, \frac{n + 3}{2}; \frac{n}{2} + 1; \frac{1}{2}\Bigr)}.\label{1}\tag{1}$$

Using this hypergeometric identity, we can rewrite the hypergeometric function within $(1)$ as

$${\left(1-\frac{1}{2}\right)^{-n/2}{_2F_1 \Bigg(\frac{n}{2}, \frac{n}{2}+1-\frac{n+3}{2}; \frac{n}{2} + 1; \frac{\frac{1}{2}}{\frac{1}{2}-1}\Bigg)}=2^{n/2}{_2F_1 \Bigl(\frac{n}{2}, -\frac{1}{2}; \frac{n}{2} + 1; -1\Bigr)}.}$$

Substituting this back into $(1)$, we obtain

$$I={\sum_{n=1}^{\infty}\frac{2}{n(2n+1)}{_2F_1 \Bigl(\frac{n}{2}, -\frac{1}{2}; \frac{n}{2} + 1; -1\Bigr)}.}\tag{2}$$

We can rewrite $(2)$ with the series definition for the hypergeometric function.

$${\sum_{n=1}^{\infty}\frac{2}{n(2n+1)}{_2F_1 \Bigl(\frac{n}{2}, -\frac{1}{2}; \frac{n}{2} + 1; -1\Bigr)}=\sum_{n=1}^{\infty}\frac{2}{n(2n+1)}\sum_{k=0}^{\infty}\frac{(-1)^k\left(\frac{n}{2}\right)_k\left(-\frac{1}{2}\right)_k}{\left(\frac{n}{2}+1\right)_k k!}}$$

Here, $(x)_n$ denotes the rising Pochhammer symbol. (There are various other notations for the rising factorial with $n$ as a superscript, but I chose this convention to help distinguish $(-1)^k$ as a unique factor.) It immediately follows from its definition that $\frac{(x)_n}{(x+1)_n}=\frac{x}{x+n}$; hence, $\frac{(n/2)_k}{(n/2+1)_k}=\frac{n}{2k+n}$. We can use this to rewrite $(2)$ even further.

$${\sum_{n=1}^{\infty}\frac{2}{n(2n+1)}\sum_{k=0}^{\infty}\frac{(-1)^k\left(\frac{n}{2}\right)_k\left(-\frac{1}{2}\right)_k}{\left(\frac{n}{2}+1\right)_k k!}=\sum_{n=1}^{\infty}\frac{2}{2n+1}\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{(2k+n)k!}}\tag{3}$$

Since there are no issues with absolute convergence, the order of summation can be swapped.

\begin{align*} \sum_{n=1}^{\infty}\frac{2}{2n+1}\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{(2k+n)k!}&{=\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{k!}\sum_{n=1}^{\infty}\frac{1}{\left(n+\frac{1}{2}\right)\!(n+2k)}}\\ &{=2\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{(4k-1)k!}\sum_{n=1}^{\infty}\left(\frac{1}{n+\frac{1}{2}}-\frac{1}{n+2k}\right)}\tag{4} \end{align*}

With this identity for the digamma function, the inner series in $(4)$ can be expressed as $\psi(2k+1)-\psi(3/2)=\psi(2k+1)+\gamma-2+2\ln{2}$, where $\gamma$ denotes the Euler-Mascheroni constant. Additionally, since $\psi(n+1)+\gamma=H_n$, where $H_n$ is the $n$-th harmonic number, we have $\psi(2k+1)+\gamma-2+2\ln{2}=H_{2k}-2+2\ln{2}$. We can now write $(4)$ as a single sum.

$${2\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{(4k-1)k!}\sum_{n=1}^{\infty}\left(\frac{1}{n+\frac{1}{2}}-\frac{1}{n+2k}\right)=2\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{(4k-1)k!}(H_{2k}-2+2\ln{2})}\label{5}\tag{5}$$

Since $(x)_n:=\frac{\Gamma(x+n)}{\Gamma(x)}$, we have $(-1/2)_k=\frac{\Gamma(k-1/2)}{\Gamma(-1/2)}=-\frac{2\sqrt{\pi}(2k-3)!!}{2^{k+1}\sqrt{\pi}}=-2^{-k}(2k-3)!!$. Hence, we can write

$${2\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k}{(4k-1)k!}\left(H_{2k}-2+2\ln{2}\right)=-2\sum_{k=0}^{\infty}\frac{\left(-\frac{1}{2}\right)^k(2k-3)!!}{(4k-1)k!}(H_{2k}-2+2\ln{2}).}$$

Jumping back to $\eqref{1}$, we can conclude

\begin{align*} I &= \sum_{n = 1}^{\infty} \frac{2^{1-\frac{n}{2}}}{n(2n + 1)} \, {_2F_1 \Bigl(\frac{n}{2}, \frac{n + 3}{2}; \frac{n}{2} + 1; \frac{1}{2}\Bigr)} \\ &= -2\sum_{k=0}^{\infty}\frac{\left(-\frac{1}{2}\right)^k(2k-3)!!}{(4k-1)k!}(H_{2k}-2+2\ln{2}).\tag{6} \end{align*}

This is as far as I can go, as I'm not well-versed with series involving the harmonic numbers and don't immediately see a way to approach the non-harmonic terms (see below). However, this form looks much more pliable than a hypergeometric. It converges very slowly (it's only accurate to 14 decimal digits after a million terms), but it is equivalent to $I$, as shown.


UPDATE: I have found a closed form for the non-harmonic terms and will write my process below.

We want to evaluate

$$4(1-\ln{2})\sum_{k=0}^{\infty}\frac{\left(-\frac{1}{2}\right)^k(2k-3)!!}{(4k-1)k!}.\tag{7}$$

We will write this in terms of the rising factorial to obtain a hypergeometric function again. Note that $\frac{1}{4k-1}=-\frac{-1/4}{k-1/4}=-\frac{(-1/4)_k}{(3/4)_k}$; given that $-2^{-k}(2k-3)!!=(-1/2)_k$ (derived below \eqref{5}), we have

$$\sum_{k=0}^{\infty}\frac{\left(-\frac{1}{2}\right)^k(2k-3)!!}{(4k-1)k!}=\sum_{k=0}^{\infty}\frac{(-1)^k\left(-\frac{1}{2}\right)_k\left(-\frac{1}{4}\right)_k}{\left(\frac{3}{4}\right)_k k!}={}_2F_1\biggl(-\frac{1}{2},-\frac{1}{4};\frac{3}{4};-1\biggr).$$

This is of the form ${}_2F_1(a,b;1+a-b;-1)$, with $a=-1/2$ and $b=-1/4$, so we can use this formula from Kummer to rewrite this as

$${}_2F_1\!\biggl(-\frac{1}{2},-\frac{1}{4};\frac{3}{4};-1\biggr)=\frac{\Gamma\big(1-\frac{1}{2}+\frac{1}{4}\big)\Gamma\big(1-\frac{1}{4}\big)}{\Gamma\big(1-\frac{1}{2}\big)\Gamma\big(1-\frac{1}{4}+\frac{1}{4}\big)}=\frac{\Gamma\big(\frac{3}{4}\big)^2}{\sqrt{\pi}}=\frac{2\pi^{3/2}}{\Gamma\big(\frac{1}{4}\big)^2}.\tag{8}$$

Which form to use comes down to personal preference. We can now return to $(7)$ and write

$$4(1-\ln{2})\sum_{k=0}^{\infty}\frac{\left(-\frac{1}{2}\right)^k(2k-3)!!}{(4k-1)k!}=\frac{8\pi^{3/2}}{\Gamma\big(\frac{1}{4}\big)^2}(1-\ln{2}).$$

Thus, we can conclude

$$\boxed{I=\frac{8\pi^{3/2}}{\Gamma\big(\frac{1}{4}\big)^2}(1-\ln{2})+\sum_{k=1}^{\infty}\frac{\left(-\frac{1}{2}\right)^{k-1}H_{2k}(2k-3)!!}{(4k-1)k!}}\tag{9}$$

(the zeroeth term can be removed since $H_0=0$). The problem has now been reduced to a series involving harmonic numbers. It may be useful to convert it to the rising factorials, as I did above, and try to obtain a closed form that way, but the index of $2k$ has thus far made it resistant to my efforts. There is some literature close to what is needed, e.g. Junesang Choi's and H.M. Srivastava's paper here, but I haven't been able to find anything closer. The identity $H_k+H_{k+1/2}=2H_{2k+1}-2\ln{2}$ strikes me as potentially useful, but the $k+1/2$ index again leads me nowhere. Perhaps someone may have further insight.

teadawg1337
  • 1,294
7

Polar integration shows that $$ I = 2 \int_0^{\pi/4} \frac{\operatorname{artanh} (\sqrt{\tan \theta}) - \sqrt{\tan \theta}}{(\cos \theta \sin \theta)^{3/2}} \, d\theta. $$ Using the series expansion $$ \operatorname{artanh} x - x = \sum_{n = 1}^{\infty} \frac{x^{2n + 1}}{2n + 1}, $$ we have \begin{align*} I &= \sum_{n = 1}^{\infty} \frac{2}{2n + 1} \int_0^{\pi/4} \tan^{n - 1} \theta \sec^3 \theta \, d \theta \\ &= \boxed{\sum_{n = 1}^{\infty} \frac{2^{1-\frac{n}{2}}}{n(2n + 1)} \, {_2F_1 \Bigl(\frac{n}{2}, \frac{n + 3}{2}; \frac{n}{2} + 1; \frac{1}{2}\Bigr)}}, \end{align*} where $_2F_1(a, b; c; z)$ is the hypergeometric function.


This is the closest I can get to a closed form. Experimentation with Wolfram Mathematica (attached below) seems to show that \begin{align*} f_n &= {_2F_1 \Bigl(\frac{n}{2}, \frac{n + 3}{2}; \frac{n}{2} + 1; \frac{1}{2}\Bigr)} \\ &= \begin{cases} a_m \sqrt{2} \operatorname{arsinh}(1) + b_m & \text{for } n = 2m - 1 \\ c_m \sqrt{2} + d_m & \text{for } n = 2m \end{cases} \end{align*} for $a_m, b_m, c_m, d_m \in \mathbb{Q}$. FindSequenceFunction returns \begin{align*} a_m &= \frac{(-1)^{m + 1} 2^{m - 2} \bigl( \frac{3}{2} \bigr)^{\overline{m - 1}}}{m!}, \\ d_m &= \frac{(-2)^m m!}{\bigl( \frac{3}{2} \bigr)^{\overline{m}}}, \end{align*} where $$ x^{\overline{n}} = x(x + 1)(x + 2) \dotsm (x + n - 1) $$ is the rising factorial (a.k.a. Pochhammer symbol). Similarly, FindGeneratingFunction returns \begin{align*} \sum_{m = 1}^{\infty} a_m x^m &= \frac{1}{2} - \frac{1}{2 \sqrt{1 + 2x}}, \\ \sum_{m = 1}^{\infty} d_m x^m &= \frac{\operatorname{arsinh} \sqrt{2x}}{\sqrt{2x(2x + 1)}} - 1. \end{align*} For $b_m$ and $c_m$, FindSequenceFunction and FindGeneratingFunction either freeze or refer back to $_2F_1$.


$$ \begin{array}{cc} n & f_n \\ \hline 1 & \frac{1}{2} \left(2+\sqrt{2} \operatorname{arsinh}(1)\right) \\ 2 & \frac{4}{3} \left(2 \sqrt{2}-1\right) \\ 3 & -\frac{3}{4} \left(\sqrt{2} \operatorname{arsinh}(1)-6\right) \\ 4 & \frac{32}{15} \left(1+\sqrt{2}\right) \\ 5 & \frac{5}{12} \left(14+3 \sqrt{2} \operatorname{arsinh}(1)\right) \\ 6 & \frac{32}{35} \left(11 \sqrt{2}-4\right) \\ 7 & -\frac{7}{48} \left(15 \sqrt{2} \operatorname{arsinh}(1)-122\right) \\ 8 & \frac{256}{315} \left(8+13 \sqrt{2}\right) \\ 9 & \frac{3}{80} \left(682+105 \sqrt{2} \operatorname{arsinh}(1)\right) \\ 10 & \frac{128}{693} \left(211 \sqrt{2}-64\right) \\ 11 & -\frac{11}{480} \left(315 \sqrt{2} \operatorname{arsinh}(1)-3074\right) \\ 12 & \frac{512 \left(128+271 \sqrt{2}\right)}{3003} \\ 13 & \frac{13 \left(27626+3465 \sqrt{2} \operatorname{arsinh}(1)\right)}{3360} \\ 14 & \frac{512 \left(1919 \sqrt{2}-512\right)}{6435} \\ 15 & \frac{501022-45045 \sqrt{2} \operatorname{arsinh}(1)}{1792} \\ 16 & \frac{8192 \left(1024+2597 \sqrt{2}\right)}{109395} \\ 17 & \frac{187 \left(37862+4095 \sqrt{2} \operatorname{arsinh}(1)\right)}{16128} \\ 18 & \frac{2048 \left(67843 \sqrt{2}-16384\right)}{230945} \\ 19 & -\frac{19 \left(765765 \sqrt{2} \operatorname{arsinh}(1)-9434878\right)}{161280} \\ 20 & \frac{8192 \left(32768+95259 \sqrt{2}\right)}{969969} \\ 21 & \frac{13 \left(11618366+1119195 \sqrt{2} \operatorname{arsinh}(1)\right)}{84480} \\ 22 & \frac{8192 \left(588933 \sqrt{2}-131072\right)}{2028117} \\ 23 & -\frac{23 \left(14549535 \sqrt{2} \operatorname{arsinh}(1)-194991322\right)}{1013760} \\ 24 & \frac{65536 \left(262144+850251 \sqrt{2}\right)}{16900975} \\ 25 & \frac{5 \left(3819921514+334639305 \sqrt{2} \operatorname{arsinh}(1)\right)}{2635776} \\ 26 & \frac{32768 \left(10098967 \sqrt{2}-2097152\right)}{35102025} \\ 27 & -\frac{3 \left(1673196525 \sqrt{2} \operatorname{arsinh}(1)-24084946414\right)}{4100096} \\ 28 & \frac{131072 \left(4194304+14904091 \sqrt{2}\right)}{145422675} \\ 29 & \frac{29 \left(62097106486+5019589575 \sqrt{2} \operatorname{arsinh}(1)\right)}{61501440} \\ 30 & \frac{131072 \left(85806311 \sqrt{2}-16777216\right)}{300540195} \\ 31 & -\frac{31 \left(145568097675 \sqrt{2} \operatorname{arsinh}(1)-2229742283746\right)}{984023040} \\ 32 & \frac{4194304 \left(33554432+128927573 \sqrt{2}\right)}{9917826435} \\ 33 & \frac{19 \left(3150308268566+237505843575 \sqrt{2} \operatorname{arsinh}(1)\right)}{506920960} \\ 34 & \frac{524288 \left(5792144099 \sqrt{2}-1073741824\right)}{20419054425} \\ 35 & \frac{73030431035486-4512611027925 \sqrt{2} \operatorname{arsinh}(1)}{260702208} \\ 36 & \frac{2097152 \left(2147483648+8834766227 \sqrt{2}\right)}{83945001525} \\ 37 & \frac{259 \left(9093229741774+644658718275 \sqrt{2} \operatorname{arsinh}(1)\right)}{4953341952} \\ 38 & \frac{2097152 \left(48605936617 \sqrt{2}-8589934592\right)}{172308161025} \\ 39 & \frac{2838808987540886-166966608033225 \sqrt{2} \operatorname{arsinh}(1)}{2540175360} \\ 40 & \frac{16777216 \left(17179869184+75096287791 \sqrt{2}\right)}{1412926920405} \\ 41 & \frac{41 \left(2488324849033834+166966608033225 \sqrt{2} \operatorname{arsinh}(1)\right)}{53343682560} \\ 42 & \frac{8388608 \left(812156618077 \sqrt{2}-137438953472\right)}{2893136075115} \\ 43 & -\frac{43 \left(6845630929362225 \sqrt{2} \operatorname{arsinh}(1)-121718302325751046\right)}{117356101 6320} \\ 44 & \frac{33554432 \left(274877906944+1268822838961 \sqrt{2}\right)}{11835556670925} \\ 45 & \frac{4610656329982787582+294362129962575675 \sqrt{2} \operatorname{arsinh}(1)}{599820075008} \\ 46 & \frac{33554432 \left(6760265315081 \sqrt{2}-1099511627776\right)}{24185702762325 } \\ 47 & -\frac{423 \left(10902301109725025 \sqrt{2} \operatorname{arsinh}(1)-201950171465282598\right)}{479856060 0064} \\ 48 & \frac{536870912 \left(2199023255552+10665172132163 \sqrt{2}\right)}{395033145117975} \\ 49 & \frac{7 \left(75588052822912144694+461167336941368557 5 \sqrt{2} \operatorname{arsinh}(1)\right)}{17137716428800} \\ 50 & 33554432 \, _2F_1\left(-\frac{1}{2},25;26;-1\right) \\ \end{array} $$

L. F.
  • 2,703
  • 1
    Very, interesting approach! So, seems like there are 4 sequences involved.

    It looks like we can get the factor in front of $\operatorname{arcsinh}(1)$ corresponding to $a_{2m+1}$ by taking the factor corresponding to $a_{2m-1}$ and multiplying it by the numerator of the fraction in front of $a_{2m-1}$ which seems to be A051417 in OEIS.

    – Nikitan Dec 23 '24 at 19:41
  • 1
    Hi @Nikitan, I was able to find (conjectured) closed forms for two of the four sequences using Mathematica, which should agree with your observation. See updated answer. I'm not sure how helpful A051417 $$ \frac{\operatorname{lcm} {1, 3, 5, \dotsc, 2n + 1}}{\operatorname{lcm} {1, 3, 5, \dotsc, 2n - 1}} $$ is, though, since it is probably the result of accidental cancellations in the fractions. – L. F. Dec 23 '24 at 20:30
  • 2
    Using this hypergeometric identity, along with rewriting it as a double sum and rearranging/simplifying, this can be expressed as ${-2\sum_{k=0}^{\infty}\frac{(-1/2)^k (H_{2k}-2+2\ln{2})(2k-3)!!}{(4k-1)k!}}$. This certainly looks more approachable, but I don't see where to start. – teadawg1337 Dec 23 '24 at 20:35
  • @teadawg1337 I think going from $_2F_1$ to the harmonic numbers is an improvement. It would nice if you could write up an answer for that. – L. F. Dec 23 '24 at 20:43
  • 1
    I'll work on writing one up right now. – teadawg1337 Dec 23 '24 at 20:45
3

I will contribute a simpler way to derive the harmonic sums and the following closed form: $$\int_0^1 \int_0^1 \frac{\sqrt{x^2+y^2}}{1-xy} dxdy = 2\left(-5-\sqrt{2} + \operatorname{arcsinh}(1)\right.$$ $$\left.+\frac{2\pi^{3/2}(-8+\operatorname{arcsinh}(1))}{\Gamma \left(\frac14\right)^2}+8\times 2^{1/4} \mathcal{H}_1 + 4\sqrt{\sqrt{2}-1} \mathcal{H}_2\right)$$ $$=1.4843079462309457237929418228902024\!\ldots$$ where $$\mathcal{H}_1 = {}_3F_2\left(-\frac14,-\frac14,\frac14;\,\frac34,\frac34;\,\frac12\right) = \frac{1}{\Gamma \left(\frac14\right)}\sum_{k=0}^\infty \frac{\Gamma \left(k+\frac14\right)}{2^k k! (4k-1)^2} $$ $$\mathcal{H}_2 = {}_3F_2\left(\frac14,\frac12,\frac12;\,\frac32,\frac32;\,2(\sqrt{2}-1)\right) = \frac{1}{\Gamma \left(\frac14\right)}\sum_{k=0}^\infty \frac{\Gamma \left(k+\frac14\right)2^k(\sqrt{2}-1)^k}{k! (2k+1)^2}$$ Since the integrand is invariant with respect to exchange $x \leftrightarrow y$ we can only consider the case $x \leq y$. Also, by the change of variables $v^2=xy$, $u^2=y/x$ and integrating with respect to $v$ we can reduce the dimension of the integral immediately: $$\int_0^1 \int_0^1 \frac{\sqrt{x^2+y^2}dxdy}{1-xy} = 2\int_{x=0}^1 \int_{y=0}^x \frac{\sqrt{x^2+y^2}dxdy}{1-xy}$$ $$=4\int_{u=0}^1 \int_{v=0}^u \frac{\sqrt{1+u^4}}{1-v^2} \frac{v^2 du dv}{u^2}=4\int_0^1 \frac{\sqrt{1+u^4}}{u^2} (\operatorname{arctanh} u - u) du$$ Next, we expand the integrand $$=4\int_0^1 u \left[ \begin{pmatrix}1/2\\0\end{pmatrix} + \begin{pmatrix}1/2\\1\end{pmatrix} u^4 + \begin{pmatrix}1/2\\2\end{pmatrix} u^8 + \cdots\right] \left[ \frac13 + \frac{u^2}{5} + \frac{u^4}{7} + \cdots \right] du$$ $$=2\sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \sum_{j=0}^{\infty} \frac{1}{(2j+3)(j+2k+1)} = 2\sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \frac{1}{4k-1} \sum_{j=0}^\infty \left( \frac{2}{2j+3} - \frac{1}{j+2k+1}\right)$$ $$=2\sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \frac{\psi(2k+1) - \psi(3/2)}{4k-1} = 2\sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \frac{H_{2k}}{4k-1} - 2\psi(3/2) \sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \frac{1}{4k-1}$$ The first sum can be decomposed into $$\sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \frac{H_{2k}}{4k-1} = \sum_{k=0}^\infty \frac{(-1)^{k+1}}{4^k(2k-1)} \begin{pmatrix}2k\\k\end{pmatrix} \frac{H_{2k}}{4k-1} = -\underbrace{\sum_{k=0}^\infty \frac{(-1)^{k}}{4^k} \begin{pmatrix}2k\\k\end{pmatrix} \frac{H_{2k}}{2k-1}}_{S_1} + 2\underbrace{\sum_{k=0}^\infty \frac{(-1)^{k}}{4^k} \begin{pmatrix}2k\\k\end{pmatrix} \frac{H_{2k}}{4k-1}}_{{S}_2}$$ where $S_1$ and $S_2$ are evaluated in another question.

The second sum is relatively easy to find: $$\sum_{k=0}^\infty \begin{pmatrix}1/2\\k\end{pmatrix} \frac{1}{4k-1} = {}_2F_1 \left(-\frac12,-\frac14;\,\frac34;\,-1\right) = \frac{2\pi^{3/2}}{\Gamma \left(\frac14\right)^2}$$

Efim Mazhnik
  • 1,735
  • I tried convert your latex code: $2\left((-5-\sqrt{2} + \operatorname{arcsinh}(1)\right.$$ $$\left.+\frac{2\pi^{3/2}(-8+\operatorname{arcsinh}(1))}{\Gamma \left(\frac14\right)^2}+8\times 2^{1/4} \mathcal{H}_1 + 4\sqrt{\sqrt{2}-1} \mathcal{H}_2\right)$ to Wolfram language and can't get correct value? – Mariusz Iwaniuk Feb 26 '25 at 09:43
  • 2(-5-Sqrt[2]+ArcSinh[1] +2 Pi^(3/2) (-8 + ArcSinh[1])/Gamma[1/4]^2 +8*2^(1/4) HypergeometricPFQ[{-1/4, -1/4, 1/4}, {3/4, 3/4}, 1/2] +4 Sqrt[Sqrt[2]-1]HypergeometricPFQ[{1/4, 1/2, 1/2}, {3/2, 3/2},2 (Sqrt[2]-1)]) – Efim Mazhnik Feb 26 '25 at 10:05
  • +1) Thank you very much . – Mariusz Iwaniuk Feb 26 '25 at 10:08