11

$$K(k)=\int_{0}^{\pi/2}\frac{dt}{\sqrt{1-k^2\sin^2t}}\tag{1}$$ is the complete elliptic of the first kind and $k'=\sqrt{1-k^2}$ the complementary modulus. We are interested in solving: $$\frac{K(k')}{K(k)}=\sqrt{n}\tag{2}$$ with $n$ a natural number. This question concerns the case: $$\frac{K(k')}{K(k)}=\sqrt{7}\tag{3} > \implies k=k_{7}=\frac{\sqrt{2}}{8}(3-\sqrt{7}).$$

Question

Is it possible to prove $(3)$ with elementary methods (no modular forms or elliptic functions) and without evaluating $K(k_{7})$? Thanks in advance for your efforts.

Updated

For esceptics. This formula is very useful: $$|K(k)(k+1)|=|-K(\frac{1-k}{1+k})+\frac{1+k}{1-k}K(\frac{1+k}{1-k})|\tag{4}$$ You have to choose the appropiate $k$. You can do calculations for the easier cases $k_{1}$,$k_{2}$,$k_{3}$,$k_{4}$,$k_{5}$ and $k_{6}$ first and you will see that $k_{3}$ and $k_{7}$ are the same in complexity.


We show the cases $n=3,12$ for getting out doubts. We put $k=i^{1/3}=\frac{\sqrt{3}}{2}+\frac{i}{2}$ in $(4)$, then: $$|K(i^{1/3})|\frac{\sqrt{6}+\sqrt{2}}{2}=|-K(i(3-\sqrt{2})+i(3+\sqrt{2})K(i(3+\sqrt{2}))|.\tag{5}$$ Now we use: $$K(ik)=\frac{K(\frac{k}{\sqrt{1+k^2}})}{\sqrt{1+k^2}}\tag{6}.$$ We have: $$K(i(3-\sqrt{2}))=\frac{\sqrt{6}+\sqrt{2}}{4}K(\frac{\sqrt{6}-\sqrt{2}}{4})\tag{7}$$ and $$K(i(3+\sqrt{2}))=\frac{\sqrt{6}-\sqrt{2}}{4}K(\frac{\sqrt{6}+\sqrt{2}}{4})\tag{8}.$$ Replacing this two in $(5)$ we have:

$$\begin{align} |K(i^{1/3})| \frac{\sqrt{6} + \sqrt{2}}{2} &= \left| - \left( \frac{\sqrt{6} + \sqrt{2}}{4} \right) K \left( \frac{\sqrt{6} - \sqrt{2}}{4} \right) \right. \nonumber \\ &\quad \left. + i \left( \frac{\sqrt{6} + \sqrt{2}}{4} \right) K \left( \frac{\sqrt{6} + \sqrt{2}}{4} \right) \right| \end{align}\tag{9} .$$

We use now Landen ascending transformation for $K(k)$: $$K(k)=\frac{2}{1+k'}K(\frac{1-k'}{1+k'})\tag{10}.$$ And we apply it to $K(i^{1/3})$ then: $$K(i^{1/3})=(1+i(2-\sqrt{3}))K(i(2-\sqrt{3}))\tag{11}.$$ And we use again $(6)$ in this last to obtain: $$K(i^{1/3})=(1+i(2-\sqrt{3}))(\frac{\sqrt{6}+\sqrt{2}}{4})K(\frac{\sqrt{6}-\sqrt{2}}{4})\tag{12}.$$ Replacing in $(9)$ we have:

$$\begin{align} \frac{\sqrt{2} (\sqrt{3} + 1) K \left( \frac{\sqrt{6} - \sqrt{2}}{4} \right)}{2} &= \left| - \left( \frac{\sqrt{6} + \sqrt{2}}{4} \right) K \left( \frac{\sqrt{6} - \sqrt{2}}{4} \right) \right. \nonumber \\ &\quad + i \left( \frac{\sqrt{6} + \sqrt{2}}{4} \right) K \left( \frac{\sqrt{6} + \sqrt{2}}{4} \right) \left. \right|. \tag{13} \end{align}$$

We can divide by $K(\frac{\sqrt{6}-\sqrt{2}}{4})$ and then: $$\frac{\sqrt{2}(\sqrt{3}+1)}{2}=|-\left(\frac{\sqrt{6}+\sqrt{2}}{4}\right)+i(\frac{\sqrt{6}+\sqrt{2}}{4})\frac{K(\frac{\sqrt{6}+\sqrt{2}}{4})}{K(\frac{\sqrt{6}-\sqrt{2}}{4})}| \\ \implies \frac{K(\frac{\sqrt{6}+\sqrt{2}}{4})}{K(\frac{\sqrt{6}-\sqrt{2}}{4})}=\sqrt{3}\hspace{.5cm}and{}\hspace{.5cm}k_{3}=\frac{\sqrt{6}-\sqrt{2}}{4}.$$ And again applying $(10)$ then $k_{12}=\frac{1-k'_{3}}{1+k'_{3}}$: $$\frac{K(k'_{12})}{K(k_{12})}=\sqrt{12} \implies k_{12}=(\sqrt{3}-\sqrt{2})^2(\sqrt{2}-1)^2.$$

User-Refolio
  • 1,513
  • You can have a good approximation of it : $k_7 \sim 4 e^{-\frac{\sqrt{7} \pi }{2}}$ using the fist term of the series expansion of $\left(\frac{K(k')}{K(k)}\right)^2$; This is a relative error of $0.10$%. – Claude Leibovici May 23 '25 at 09:48
  • @ClaudeLeibovici good approximation. – User-Refolio May 23 '25 at 09:57
  • @ClaudeLeibovici It seems that $(3)$ can be proven without using modular equations. :) – User-Refolio May 23 '25 at 09:57
  • 1
    Still better $k_7 \sim \sqrt{2 W\left(8 e^{-\sqrt{7} \pi }\right)}$ (relative error of $0.00016$% obtained without squaring and using two terms of the series expansion – Claude Leibovici May 23 '25 at 10:06
  • W Lambert function? – User-Refolio May 23 '25 at 10:27
  • Could you give some details about how you computed $k_7$? – user May 23 '25 at 10:56
  • @user the question is compute $k_{7}$, if I share my computation then is an answer to the question itself. – User-Refolio May 23 '25 at 10:59
  • Probably you found it by "non-elementary" methods, so it is not quite an answer. – user May 23 '25 at 11:01
  • An elementary argument for this solution being unique is that, since $K(k)$ is positive and monotonically-decreasing for $k\in (0,1)$, one can show that $K'(k)/K(k)$ is monotonically-decreasing as well. So if $K'(k_7)/K(k_7)=\sqrt{7}$, it is the only such solution and thus $K'(k)/K(k)=\sqrt{7}\implies k=k_7$. (That such a solution exists follows from $K'(0)/K(0)=0$ and $K'(k)/K(k)\to\infty$ as $k\to 0^+$.) But this sheds no light on $k_7=\sqrt{2}(3-\sqrt{7})/8$ and so is another not-an-answer. – Semiclassical May 23 '25 at 23:14
  • 1
    Just an idea, we are trying to show that $\frac{\theta_2(e^{-\pi \sqrt{7}})}{\theta_3(e^{-\pi\sqrt{7})}}=\sqrt{\frac{3-\sqrt{7}}{2\sqrt{2}}}$ I wonder if we can relate $\theta_2$ to $\theta_3$ with smth like Poisson summation –  May 24 '25 at 04:01
  • @Dqrksun there is no need of theta functions, only with complex numbers the result is obtained. – User-Refolio May 24 '25 at 10:58
  • 1
    @Dqrksun Jacobi developed his theory using as one of the main tools Legendre's relation. I think that working with Eisenstein series and complex numbers in modern approach will arrive also to the value of $k_{7}$. – User-Refolio May 24 '25 at 11:41
  • 1
    @user I showed $k_{3}$ for getting you out doubts. – User-Refolio May 25 '25 at 09:43
  • 1
    Thank you! As soon as I have time I will try to apply this to $k_7$. – user May 25 '25 at 10:27
  • Your equation $(4)$ is obtained from $K(1/k)=k(K(k)-iK(k'))$ for $k\in(0,1)$ by replacing $k$ with $(1-k)/(1+k)$ and thereby replacing $k'$ with $2\sqrt{k}/(1+k)$ and then applying Landen transformation. – Paramanand Singh May 30 '25 at 10:55
  • This identity gives a direct explanation of $k_7$: $\sqrt{\vartheta_2(q)\vartheta_2(q^7)} +\sqrt{\vartheta_4(q)\vartheta_4(q^7)} =\sqrt{\vartheta_3(q)\vartheta_3(q^7)},$ but proving it requires efforts and not that elementary though. – Setness Ramesory May 30 '25 at 13:52
  • The only thing I can say at this momment is that I know how to prove Ramanujan series for $1/\pi$ indendently of $k_{n}$. This includes Chudnovsky $n=163$. I hope I will have news as soon as possible. I know that is hard to believe but time will say. Till this momment I will remain a little time reserved. – User-Refolio May 30 '25 at 14:47
  • @ParamanandSingh Ramanujan obtained all his series with rigor included that for $n=58$ and Borwein's where mistaken because he arrived to the famous number 1103 with rigor and no intuition. Ramanujan left something for whoever saw it. And seems that I have seen it. Best. – User-Refolio May 30 '25 at 14:57
  • 1
    Take your time to verify your calculations and publish it when you are ready. Best of luck! – Paramanand Singh May 30 '25 at 15:34
  • @ParamanandSingh Finally I have good feedback of the below answer. Translating this proof to elliptic functions for the cases 2,4 a modular equation of degree two is enough and for the cases 3,7 a modular equation of degree 2 and the change q to -q in the nome is enough. For the case n=5 we need a modular equationd of degree 3 and the change q to -q. Seems that using complex numbers reduces the modular equation needed. – User-Refolio Jul 15 '25 at 06:12

3 Answers3

5

This is not an answer (just for the fun of an approximate solution).

We can have a good approximation of the solution of $$\frac{K(k')}{K(k)}=\sqrt{n}$$ combining the series expansions of the elliptic integrals around $k=0$. This gives $$\frac{K(k')}{K(k)}=\frac{2 (2 \log (2)-\log (k))}{\pi}-\frac{k^2}{2 \pi }+O\left(k^4 \right)$$ and the solution is given in terms of Lambert function (let $k=\sqrt t$). This gives $$\large\color{blue}{k_n \sim \sqrt{2 W\left(8 e^{-\pi\sqrt{n} }\right)}}\tag 1$$ where $W(.)$ is the principal branch of Lambert function.

Expanding for "large" values of $n$, using the above estimate $$\frac{K(k')}{K(k)}-\sqrt{n}=-\frac{52}{\pi }\,e^{-2 \pi \sqrt{n}}\Big( 1-\frac{256}{39} e^{-\pi \sqrt{n}}+\cdots\Big)$$

Trying for a few values of $n$

$$\left( \begin{array}{ccc} n & \text{approximation} & \text{solution} \\ 2 & 0.41548680 & 0.41421356 \\ 3 & 0.25893793 & 0.25881902 \\ 4 & 0.17158802 & 0.17157287 \\ 5 & 0.11887936 & 0.11887695 \\ 6 & 0.08516469 & 0.08516423 \\ 7 & 0.06262301 & 0.06262301 \\ 8 & 0.04702192 & 0.04702192 \\ 9 & 0.03592157 & 0.03592157 \\ 10 & 0.02784245 & 0.02784245 \\ \end{array} \right)$$

Edit

We can improve the estimate using the first iterate of any Newton-like method of order $m$ starting with $(1)$. The formulae remain fully explicit.

For example, for $n=7$, we shall have

$$\left( \begin{array}{ccc} m & k_7^{(m)} & \text{method} \\ 2 & \color{red}{0.062622912543}0918265780130914496 & \text{Newton} \\ 3 & \color{red}{0.062622912543167970}0824424230563 & \text{Halley} \\ 4 & \color{red}{0.0626229125431679701026646}644880 & \text{Householder} \\ 5 & \color{red}{0.0626229125431679701026646800391} & \text{unnamed} \\ \end{array} \right)$$

Update

The series expansion is $$\frac{4 \log (2)}{\pi }-\frac{\log \left(k^2\right)}{\pi }-\frac 1{2\pi}\sum_{n=1}^\infty a_n\,k^{2n}$$ where the first coefficients are $$\left\{1,\frac{13}{32},\frac{23}{96},\frac{2701}{16384}, \frac{5057}{40960},\frac{76715}{786432},\frac{146749}{1835008},\frac{144644749}{2147483648},\frac{279805685}{4831838208}\right\}$$ which do not seem to correspond to any known sequence in OEIS.

  • 1
    +1 thanks for the computation. – User-Refolio May 23 '25 at 17:14
  • (+1) This can be turned into an exact computation of $k_7$ once we know (by the theory of modular forms) that $k_7$ is an algebraic number with low degree. Mathematica's command $\text{RootApproximant}$, when applied to Halley or Householder's approximations, correctly returns $1-256x^2+256x^4$ as the minimal polynomial of $k_7$. – Jack D'Aurizio Jul 16 '25 at 15:29
3

We are going to consider: $l_{7}=\frac{3\sqrt{7}}{8}-\frac{i}{8}$. Then using formula $(4)$ of the question:

$$|K(\frac{3\sqrt{7}}{8}-\frac{i}{8})(1+\frac{3\sqrt{7}}{8}-\frac{i}{8})|=|-K(i(8-3\sqrt{7}))-i(3\sqrt{7}+8)K(i(3\sqrt{7}+8))|\tag{1}.$$ We consider now Landen ascending transformation: $$K(k)=\frac{2}{1+k'}K(\frac{1-k'}{1+k'})\tag{2},$$ where $k'=\sqrt{1-k^2}$ is the complementary modulus. And the imaginary transformation: $$K(ik)=\frac{1}{\sqrt{1+k^2}}K(\frac{k}{\sqrt{1+k^2}})\tag{3}.$$ Then we have: $$K(\frac{3\sqrt{7}}{8}-\frac{i}{8})\overset{(2)}=(\frac{11}{8}-\frac{\sqrt{7}i}{8})K(\frac{3}{8}-\frac{\sqrt{7}i}{8}) \overset{(2)}=(\frac{11}{8}-\frac{\sqrt{7}i}{8})(1+i(3\sqrt{7}-8))K(i(3\sqrt{7}-8))\\ \overset{(3)}=(\frac{5\sqrt{2}+\sqrt{14}}{8}+i(\frac{\sqrt{14}-5\sqrt{2}}{8}))K(\frac{3\sqrt{2}-\sqrt{14}}{8})\tag{4},$$ $$K(i(8-3\sqrt{7}))\overset{(3)}=\frac{3\sqrt{2}+\sqrt{14}}{8}K(\frac{3\sqrt{2}-\sqrt{14}}{8}),\tag{5}$$ and: $$K(i(3\sqrt{7}+8))\overset{(3)}=\frac{3\sqrt{2}-\sqrt{14}}{8}K(\frac{3\sqrt{2}+\sqrt{14}}{8}).\tag{6}$$ Replacing $(4)$, $(5)$ and $(6)$ in $(1)$: $$\frac{3+\sqrt{7}}{2}K(\frac{3\sqrt{2}-\sqrt{14}}{8})\\=|-\frac{3\sqrt{2}+\sqrt{14}}{8}K(\frac{3\sqrt{2}-\sqrt{14}}{8})-i(\frac{(3\sqrt{7}+8)(3\sqrt{2}-\sqrt{14})}{8}K(\frac{3\sqrt{2}+\sqrt{14}}{8}))|$$ And from this last it follows that: $$\frac{K(\frac{3\sqrt{2}+\sqrt{14}}{8})}{K(\frac{3\sqrt{2}-\sqrt{14}}{8})}=\sqrt{7}\hspace{.5cm}and\hspace{.5cm}k_{7}=\frac{3\sqrt{2}-\sqrt{14}}{8}.$$

Now we are going to derive: $$28E(k_{7})K(k_{7})-(4\sqrt{7}+14)K^2(k_{7})=\sqrt{7}\pi,\hspace{1cm}k_{7}=\frac{3\sqrt{2}-\sqrt{14}}{8}\tag{7}.$$

We follow the same approach as we have done for $k_{3}$: https://math.stackexchange.com/a/5080646/1187943.

We need also Landen ascending transformation and the imaginary transformation for $E$: $$\frac{E(k)}{1+k'}+\frac{k'K(k)}{1+k'}=E(\frac{1-k'}{1+k'}),\tag{8}$$ and $$E(ik)=\sqrt{1+k^2}E(\frac{k}{\sqrt{1+k^2}})\tag{9}.$$

Let be $s_{7}=\frac{3}{8}-\frac{\sqrt{7}i}{8}$ then we have to observe that: $$E(s_{7})K(s_{7})=2E(k_{7})K(k_{7})-K^2(k_{7}).\tag{10}$$ To confirm it we have: $$K(s_{7})=K(\frac{3}{8}-\frac{\sqrt{7}i}{8})\overset{(2)}=(1+i(3\sqrt{7}-8))K(i(3\sqrt{7}-8))\\ \overset{(3)}=(1+i(3\sqrt{7}-8))(\frac{3\sqrt{2}+\sqrt{14}}{8})K(k_{7}),\tag{11}$$ and $$E(s_{7})=E(\frac{3}{8}-\frac{\sqrt{7}i}{8})\overset{(8)}=(1+\frac{3\sqrt{7}}{8}+\frac{i}{8})E(i(3\sqrt{7}-8))-(\frac{3\sqrt{7}}{8}+\frac{i}{8})K(\frac{3}{8}-\frac{\sqrt{7}i}{8})\\ \overset{(9)}=(1+\frac{3\sqrt{7}}{8}+\frac{i}{8})(6\sqrt{2}-2\sqrt{14})E(k_{7})-(\frac{3\sqrt{7}}{8}+\frac{i}{8})K(\frac{3}{8}-\frac{\sqrt{7}i}{8}) \overset{(2)}=(1+\frac{3\sqrt{7}}{8}+\frac{i}{8})(6\sqrt{2}-2\sqrt{14})E(k_{7})-(\frac{3\sqrt{7}}{8}+\frac{i}{8})(1+i(3\sqrt{7}-8))K(i(3\sqrt{7}-8)) \\\overset{(3)} =(1+\frac{3\sqrt{7}}{8}+\frac{i}{8})(6\sqrt{2}-2\sqrt{14})E(k_{7})-(\frac{3\sqrt{7}}{8}+\frac{i}{8})(1+i(3\sqrt{7}-8))(\frac{3\sqrt{2}+\sqrt{14}}{8})K(k_{7}) \tag{12}.$$ And using $(11)$ and $(12)$ we confirm $(10)$. Also we have to note that: $$K(s'_{7})=K(\frac{3\sqrt{7}}{8}+\frac{i}{8})\overset{(2)}=(\frac{11}{8}+\frac{\sqrt{7}i}{8})K(\frac{3}{8}+\frac{\sqrt{7}i}{8})\overset{(2)}= (\frac{11}{8}+\frac{\sqrt{7}i}{8})(1+i(8-3\sqrt{7}))K(i(8-3\sqrt{7})) \\ \nonumber \overset{(3)}=(\frac{11}{8}+\frac{\sqrt{7}i}{8})(1+i(8-3\sqrt{7}))\left(\frac{3\sqrt{2}+\sqrt{14}}{8}\right)K(k_{7}).\tag{13}$$ Now we use: $$\frac{K(k')}{K(k)}\left(K^2(k)-\frac{K(k)E(k)}{1+k}\right)=-\frac{\pi}{2(k+1)}+K(k)E(\frac{1-k}{1+k})\tag{14}.$$ that is a consequence of Legendre's relation and Landen's descending transformations. We replace $s_{7}=\frac{3}{8}-\frac{\sqrt{7}i}{8}$ in $(14)$ and using $(10)$, $(11)$ and $(13)$ we have: $$\left(\frac{\sqrt{7}}{2}+\frac{i}{2}\right)\left(K^2(s_{7})-\frac{2E(k_{7})K(k_{7})-K^2(k_{7})}{1+s_{7}} \right)=-\frac{\pi}{2(1+s_{7})}+K(s_{7})E(\frac{1-s_{7}}{1+s_{7}}).\tag{15}$$ In the other hand: $$ E(\frac{1-s_{7}}{1+s_{7}})=E(\frac{3}{8}+\frac{\sqrt{7}i}{8}) \\ \overset{(8)}=\left( (1+\frac{3\sqrt{7}}{8})E(i(8-3\sqrt{7}))-\frac{3\sqrt{7}}{8}K(\frac{3}{8}+\frac{\sqrt{7}i}{8})\right)+ i\left(\frac{K(\frac{3}{8}+\frac{\sqrt{7}i}{8})}{8}-\frac{E(i(8-3\sqrt{7}))}{8} \right) \\ \overset{(9)}=\left( \frac{3\sqrt{2}+\sqrt{14}}{4}E(k_{7})-\frac{3\sqrt{7}}{8}K(\frac{3}{8}+\frac{\sqrt{7}i}{8})\right)+ i\left(\frac{K(\frac{3}{8}+\frac{\sqrt{7}i}{8})}{8}+\frac{\sqrt{14}-3\sqrt{2}}{4}E(k_{7}) \right) \\ \overset{(2)}=\left( \frac{3\sqrt{2}+\sqrt{14}}{4}E(k_{7})-K(i(8-3\sqrt{7}))\right)+ i\left((8-3\sqrt{7})K(i(8-3\sqrt{7}))+\frac{\sqrt{14}-3\sqrt{2}}{4}E(k_{7}) \right) \\ \overset{(3)}=\left( \frac{3\sqrt{2}+\sqrt{14}}{4}E(k_{7})-\frac{3\sqrt{2}+\sqrt{14}}{8}K(k_{7})\right)+ i\left(\frac{3\sqrt{2}-\sqrt{14}}{8}K(k_{7})+\frac{\sqrt{14}-3\sqrt{2}}{4}E(k_{7}) \right).\tag{16}$$ Finally using $(11)$, $(15)$ and $(16)$: $$ (\frac{5\sqrt{7}}{16}+\frac{11}{8})K^2(k_{7}) -\frac{5\sqrt{7}K(k_{7})E(k_{7})}{8}+i\left((\frac{\sqrt{7}}{8}+\frac{9}{16})K^2(k_{7}) -\frac{9K(k_{7})E(k_{7})}{8}\right) \\ =-\frac{3\sqrt{7}K^2(k_{7})}{8}+\frac{3\sqrt{7}K(k_{7})E(k_{7})}{4}-\frac{11\pi}{32}+i\left(\frac{K^2(k_{7})}{8}-\frac{K(k_{7})E(k_{7})}{4}-\frac{\sqrt{7}\pi}{32}\right) $$ We take real parts in this last equality (with the imaginary is the same) and: $$ E(k_{7})=(\frac{\sqrt{7}}{7}+\frac{1}{2})K(k_{7})+\frac{\sqrt{7}\pi}{28K(k_{7})}.$$ Or which is the same: $$28E(k_{7})K(k_{7})-(4\sqrt{7}+14)K^2(k_{7})=\sqrt{7}\pi.\tag{17}$$ For obtain the famous series by Ramanujan we invoke Clausen's formula: $$\frac{4K^2(s)}{\pi^2}=\sum_{n=0}^{\infty}\frac{(2n)!^3}{2^{4n}n!^6}(s^2(1-s^2))^{n}\hspace{.5cm}0<|s| \leq \frac{1}{\sqrt{2}}\tag{18},$$ and: $$\frac{4K(s)E(s)}{\pi^2}=\sum_{n=0}^{\infty}\frac{(2n)!^3}{2^{4n}n!^6}(s^2(1-s^2))^{n}(n(1-2s^2)+(1-s^2))\hspace{.5cm}0<|s|<\frac{1}{\sqrt{2}},\tag{19}$$ which is obtained from $(18)$ after derivating and using: $\frac{\mathrm{dK}}{\mathrm{d}k}=\frac{E(k)}{k(1-k^2)} - \frac{K(k)}{k}$.

As a corollary using $(17)$, $(18)$ and $(19)$: $$\sqrt{7}\pi=28E(k_{7})K(k_{7})-(4\sqrt{7}+14)K^2(k_{7})=\frac{\sqrt{7}\pi^2}{16}\sum_{n=0}^{\infty}\frac{(2n)!^3}{2^{12n}n!^6} (42n+5) \\ \implies \sum_{n=0}^{\infty}\frac{(2n)!^3}{2^{12n}n!^6} (42n+5)=\frac{16}{\pi}.$$

Wich is formula $(29)$ of Ramanujan's master paper.


We have to note that using the hypergeometric approach and $(17)$ we can derive the following:

$$ \sum_{n=1}^{\infty}\binom{2n}{n}^3\frac{q(n)H_{n}}{2^{12n}(2n-1)^3}=\frac{16}{\pi}\tag{20} $$ where $q(n)=21168n^5-18816n^4+272n^3-180n^2+17n + 5$ and $H_{n}=\sum_{i=1}^{n}\frac{1}{i}$ is the n-th harmonic number. $$ \sum_{n=0}^{\infty}\frac{(-1)^n(4n)!}{63^{2n}n!^{4}}(65n+8)=\frac{9\sqrt{7}}{\pi}.\tag{21} $$ And: $$ \sum_{n=0}^{\infty}\frac{(-1)^n(6n)!}{15^{3n}n!^{3}(3n)!}(63n+8)=\frac{5\sqrt{15}}{\pi}.\tag{22} $$


Here is some feedback of Zudilin.

You certainly master the techniques. This type of proof is based on the Legendre relation between periods and quasi-periods, it goes back to the Borweins' book "Pi and the AGM". I am impressed that you do such things by hand. You may feel that it looks new computationally, but mathematically there is exploitation of the very same ideas that have been already used for such derivations. These ideas are generalised in several directions; for example, James Wan had other extensions to 4F3 series in his work around 2013 in the Integral Transforms and Special Functions.

Several new ideas, also in executing modular proofs, originate from Jesús Guillera.


Borwein's books is the seed of the work of many others but in their book this case is not showed in detail. I believe this proof has pedagogical value.

User-Refolio
  • 1,513
1

Using Mathematica notations, we need to use $$K(k^2)=\frac{\pi }{2\, \text{agm}\left(1,\sqrt{1-k^2}\right)} \quad \implies \quad \frac{K(1-k^2)}{K(k^2)}=\frac{\text{agm}\left(1,\sqrt{1-k^2}\right)}{\text {agm}(1,k)}$$ If $$k=\frac{3-\sqrt{7}}{4 \sqrt{2}}$$ then $$\frac{\text{agm}\left(1,\sqrt{1-k^2}\right)}{\text{agm}(1,k)}=\frac{\text{agm}\left(1,\frac{1}{4} \sqrt{8+3 \sqrt{7}}\right)}{\text{agm}\left(1,\frac{1}{4} \sqrt{8-3\sqrt{7}}\right)}$$

For sure, we could have use Jacobi theta functions to try to go further from a formal point of view.

Numerically, this is

$$2.6457513110645905905016157536392604257102591830825\cdots=\sqrt 7$$