7

Being: $$R(k)=\left(-\frac{K(k)^2}{k}+K(k)E(k)\left(\frac{1}{k}-\frac{k}{2(k^2-1)}\right)\right)$$

We have these evaluations:

$$\int_{0}^{1/\sqrt{2}}R(k)dk=\frac{3\Gamma{(1/4)}^4}{128\pi}-\frac{\pi^2}{8},\tag{1}$$

$$\int_{0}^{2^{5/4}(\sqrt{2}-1)}R(k)dk=\frac{3\Gamma{(1/4)}^4}{64\pi}-\frac{\pi^2}{8},\tag{2}$$

$$\int_{0}^{\sqrt{2}-1}R(k)dk=\frac{\Gamma(1/8)^2\Gamma(3/8)^2\left(\frac{3\sqrt{2}}{512}+\frac{1}{256} \right)}{\pi}-\frac{\pi^2}{8},\tag{3}$$ $$\int_{0}^{\sqrt{2\sqrt{2}-2}}R(k)dk=\frac{\Gamma(1/8)^2\Gamma(3/8)^2}{64\pi}-\frac{\pi^2}{8}.\tag{4}$$ Where $K(k)$ and $E(k)$ are the complete elliptic integrals of the first and second kind respectively. Are particular cases of: $$\int_{0}^{l}R(k)dk=\frac{K(l)^2(2-l^2)}{4}-\frac{\pi^2}{8},\tag{5}$$ and: $$\int_{0}^{\frac{2\sqrt{l}}{1+l}}R(k)dk=\frac{K(l)^2(1+l^2)}{2}-\frac{\pi^2}{8}.\tag{6}$$ Questions: Are these integrals known? Can we improve this result with Clausen's transformation and derive some series?

User-Refolio
  • 1,088
  • 4
  • 29
  • 1
    The function $R(k)$ seems quite arbitrary to me. Does $R(k)$ have much importance in your work and if so, could you explain why? And if not, what makes you think its integral would be a well known result? – Luca Armstrong Jun 14 '23 at 10:31
  • 6
    The result follows from fundamental theorem of calculus. Your integrand $R(k) $ is nothing but derivative of $K(k) ^2(2-k^2)/4$. – Paramanand Singh Jun 14 '23 at 17:24
  • @ParamanandSingh once one knows the primitive the result is easy. I'm looking for a manipulation of the integrand $R(k)$ in some way that we can obtain some associated series. – User-Refolio Jun 17 '23 at 07:28
  • @LucaArmstrong Yes, the integrand is quite arbitrary but it has a primitive. The integrand has not much importance in my work. Since has been written a lot of elliptic integrals I wonder if someone has derived it before. – User-Refolio Jun 17 '23 at 07:31
  • 1
    Both $E, K$ can be expressed as hypergeometric series in $k^2$ and further $K^2$ can also be written like that so there is a series form for $R(k) $. – Paramanand Singh Jun 17 '23 at 09:28
  • 1
    In case you wish to obtain some Ramanujan type series for $1/\pi$ you should try to get a series for $(2K/\pi)^2$ in terms of $k$ (or some function of $k$). – Paramanand Singh Jun 17 '23 at 09:29
  • @ParamanandSingh Thank's for the hint. I would like to obtain series involving gamma functions rather than series for $1/\pi$ . – User-Refolio Jun 17 '23 at 09:37
  • 1
    @User: I wasn't aware of it. I am waiting for your derivation of the series. – Paramanand Singh Sep 16 '23 at 15:20

2 Answers2

4

After some work something has come. The answer to the question is yes, but the sums are not obvius and with not beatiful forms. We have: $$\frac{8K(k)E(k)}{\pi^2}=\frac{2-k^2}{\sqrt{1-k^2}}\sum_{n=0}^{\infty}\frac{(2n+1)(2n)!^3}{2^{8n}n!^6}{\left(\frac{k^4}{ k^2-1}\right)}^n,\tag{1}$$ with $-\frac{1}{\sqrt{2}}\leq k\leq\frac{1}{\sqrt{2}}.$ Where: $$K(k)=\int_{0}^{\pi/2}\frac{dx}{\sqrt{1-k^2\sin^2{(x)}}},\tag{2}$$ and $$E(k)=\int_{0}^{\pi/2}\sqrt{1-k^2\sin^2{(x)}}dx,\tag{3}$$ are the complete elliptic of the first and second kind respectively.

As suggested by Bob Dobbs and Paramanand Singh we provide a proof of $(1)$.

Since we have: $$K(k)(1-k^2)=E(k)-K'(k)(k-k^3),\tag{4}$$ where $K'(k)$ here is the derivative of $K(k)$. Multiplying both sides by $K(k)$: $$K^2(k)(1-k^2)+K(k)K'(k)(k-k^3)=K(k)E(k),\tag{5}$$ We recognize $K(k)K'(k)$ as $\left(\frac{K^2(k)}{2}\right)'$. Then: $$K^2(k)(1-k^2)+\left(\frac{K^2(k)}{2}\right)'(k-k^3)=K(k)E(k).\tag{6}$$ Now we invoke Landen's transformation: $$K(k)(1+k)=K(\frac{2\sqrt{k}}{1+k}),\tag{7}$$ and Clausen's formula (1828) (https://arxiv.org/pdf/1302.5984.pdf) formula as in the version of linked papers. $$_3F_2(1/2,1/2,1/2;1,1;4k^2(1-k^2))=\frac{4K^2(k)}{\pi^2}.\tag{8}$$ Some calculus shows that the combination of both gives: $$\frac{4K^{2}(k)}{\pi^2}=\frac{1}{\sqrt{1-k^2}}\sum_{n=0}^{\infty}\frac{(2n)!^3}{2^{8n}n!^6}\left(\frac{k^4}{k^2-1} \right)^{n},\tag{9}$$ with $|k|\leq \frac{1}{\sqrt{2}}$. Finally, derivation of $(9)$ and $(8)$ applied to $(6)$ gives: $$\frac{8K(k)E(k)}{\pi^2}=\frac{2-k^2}{\sqrt{1-k^2}}\sum_{n=0}^{\infty}\frac{(2n+1)(2n)!^3}{2^{8n}n!^6}{\left(\frac{k^4}{ k^2-1}\right)}^n.$$

With $(1)$ and $(9)$ we can now rewrite the $R(k)$ in the question in terms of an $_3F_2$ function and perform integration term by term. The first case in question is $\int_{0}^{k_{1}}R(k)dk$, where $k_{1}=\frac{1}{\sqrt{2}}$ which maybe can be interesting for first instance but the resulting sum involves also Beta functions in the general terms.

Note : Putting $k_{1}=\frac{1}{\sqrt{2}}$ in $(9)$ and $(1)$ we get: $$\frac{4K^2(k_{1})}{\sqrt{2}\pi^2}=\sum_{n=0}^{\infty}\frac{(-1)^n(2n)!^3}{2^{9n}n!^6}=\frac{\Gamma{(1/4)}^4}{4\sqrt{2}\pi^3},\tag{10}$$ and $$\frac{8\sqrt{2}E(k_{1})K(k_{1})}{3\pi^2}\\=\sum_{n=0}^{\infty}\frac{(-1)^n(2n)!^3(2n+1)}{2^{9n}n!^6}=\frac{2\sqrt{2}}{3\pi}+\frac{\Gamma(\frac{1}{4})^4}{6\sqrt{2}\pi^3},\tag{11}$$ and we deduce: $$-\frac{8K^2(k_{1})}{\sqrt{2}\pi^2}+\frac{8\sqrt{2}E(k_{1})K(k_{1})}{\pi^2}\\=\sum_{n=0}^{\infty}\frac{(-1)^n(6n+1)(2n)!^3}{2^{9n}n!^6}=\frac{2\sqrt{2}}{\pi}.\tag{12}$$

Another way to obtain $(12)$ without using the explicit values of $K(\frac{1}{\sqrt{2}})$ and $E(\frac{1}{\sqrt{2}})$ is using Legendre's relation: $$\frac{\pi}{2}=2E(\frac{1}{\sqrt{2}})K(\frac{1}{\sqrt{2}})-K^2(\frac{1}{\sqrt{2}})\tag{13}$$ Translating to the language of series $(13)$ using $(1)$ and $(9)$ when $k=\frac{1}{\sqrt{2}}$ gives the right hand of $(12)$.

It's curious that Ramanujan didn't list this sum in his famous paper "Modular equations and approximations to $\pi$". The proof of this sum is attributed to Berndt and Baruah using elliptic functions and Jesús Guillera using "creative telescoping".

But we have showed that this sum is the easiest to obtain because Legendre's relation works fine in this case giving a direct link between $K(k)$ and $E(k)$ when $k=\frac{1}{\sqrt{2}}$. Bauer's series is a little bit complicated but with Landen's transformation (related to the modular equation of degree 2 in the language of elliptic functions) also Legendre's relation is able to link $K(\sqrt{2}-1)$ and $E(\sqrt{2}-1)$.

User-Refolio
  • 1,088
  • 4
  • 29
1

For the first question (Are these integrals known?), Altough OP is not interested integrals now, turned his attention to series:

First we need to compute the derivative of $K=K(k)=\int_0^{\pi/2}\frac{1}{\Delta}d\phi$ where $\Delta=\sqrt{1-k^2\sin^2\phi}\,\,\,$: $$\frac{dK}{dk}=\int_0^{\pi/2}-\frac{\Delta'}{\Delta^2}d\phi=\int_0^{\pi/2}\frac{k\sin^2\phi}{\Delta^3}d\phi=\frac1k\int_0^{\pi/2}(\frac{1}{\Delta^3}-\frac1\Delta)\,d\phi=\frac{E-(1-k^2)K}{k(1-k^2)}$$ due to the reduction formula $$\int\frac1{\Delta^3}d\phi=-\frac{k^2\sin\phi\cos\phi}{(1-k^2)\Delta}+\frac{1}{(1-k^2)}\int\Delta\,d\phi.$$ Note that $E=E(k)=\int_0^{\pi/2}\Delta\,d\phi$.

Now we can check: $$\frac{d}{dk}\left(\frac{K^2(2-k^2)}{4}\right)=\frac12K\frac{dK}{dk}(2-k^2)-\frac12kK^2=\frac12K\frac{E-(1-k^2)K}{k(1-k^2)}(2-k^2)-\frac12kK^2\\ =-\frac{K^2}{k}+\frac{2-k^2}{2k(1-k^2)}EK=R(k).$$ Therefore, $\int_0^{l}R(k)dk=\frac{K(l)^2(2-l^2)}{4}-\frac{K(0)^2(2-0^2)}{4}=\frac{K(l)^2(2-l^2)}{4}-\frac{\pi^2}{8}$, since $K(0)=\frac\pi 2$.

Bob Dobbs
  • 15,712
  • 2
    Hi Bob. My question is not to prove the integral which I know to prove. The question is if they are known or if we can improve this result to derive some associated series. Best all. – User-Refolio Jun 20 '23 at 12:22
  • I've answered the question. Did you know $(1)$? – User-Refolio Sep 16 '23 at 11:21