16
$\begingroup$

Consider the sum $(1)$

$$\sum_{n=0}^{\infty}{{2n\choose n}^3\over 2^{6n}}={\pi\over \Gamma^4\left({3\over 4}\right)}\tag1$$ How does one prove $(1)$?

An attempt:

Recall

$${2n\choose n}\cdot{\pi\over 2^{2n+1}}=\int_{0}^{\infty}{\mathrm dx\over (1+x^2)^{n+1}}\tag2$$

Choosing $x=\tan{u}$ then $\mathrm dx=\sec^2{u}\mathrm du$

$(2)$ becomes

$${2n\choose n}\cdot{\pi\over 2^{2n+1}}=\int_{0}^{\pi/2}\cos^{2n}{u}\mathrm du\tag3$$

$${2n\choose n}^3\cdot{1\over 2^{6n}}={8\over \pi^3}\cdot\left(\int_{0}^{\pi/2}\cos^{2n}{u}\mathrm du\right)^3\tag4$$

$${2n\choose n}^3\cdot{1\over 2^{6n}}={8\over \pi^3}\cdot\left({(2n-1)!!\over (2n)!!}\right)^3\cdot{\pi^3\over 8}\tag5$$

I am not on the right track here.

How else can we tackle $(1)?$

Note:

$$\sum_{n=0}^{\infty}\left({(2n-1)!!\over (2n)!!}\right)^3={\pi\over \Gamma^4(3/4)}\tag1$$

similar to ramanujan's sum $(6)$

$$\sum_{n=0}^{\infty}(-1)^n(4n+1)\left({(2n-1)!!\over (2n)!!}\right)^5={2\over \Gamma^4(3/4)}\tag6$$

I found similar here, it may be helpful.

$\endgroup$
11
+50
$\begingroup$

The sum is clearly a hypergeometric series. Indeed, defining $a_n=\frac{{2n\choose n}^3}{2^{6n}}$, we have $\frac{a_{n+1}}{a_n}=\left(\frac{n+\frac12}{n+1}\right)^3$. Together with $a_0=1$, this means that $$\sum_{n=0}^{\infty}a_nx^n={}_3F{}_2\biggl[\begin{array}{cc} \frac12,\frac12,\frac12 \\ 1,1\end{array};x\biggr].$$ On the other hand this hypergeometric function is known to be $${}_3F{}_2\biggl[\begin{array}{cc} \frac12,\frac12,\frac12 \\ 1,1\end{array};x\biggr]=\frac{4}{\pi^2}\mathbf{K}^2\left(\frac{1-\sqrt{1-x}}{2}\right).$$ It remains to set $x=1$ in this formula and use first elliptic integral singular value $\mathbf{K}\left(\frac12\right)=\frac{\Gamma^2\left(\frac14\right)}{4\sqrt\pi}$.

| cite | improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ The value of $\vartheta_{3}(e^{-\pi})$ (see math.stackexchange.com/a/1798992/72031) suggests that there is connection with theta functions, elliptic integrals and the like. The expression of ${}_{3}F_{2}$ as square of elliptic integral comes from the Clausen's formula (en.wikipedia.org/wiki/Clausen's_formula). Nice answer +1 $\endgroup$ – Paramanand Singh Feb 19 '17 at 14:47
  • $\begingroup$ @ParamanandSingh Thanks! I did not think about Clausen's formula. By the way it seems to be the best way to address this question. $\endgroup$ – Start wearing purple Feb 19 '17 at 18:21
  • 1
    $\begingroup$ I added an answer based on my comment which gives details of the formula you have used. $\endgroup$ – Paramanand Singh Feb 20 '17 at 0:59
9
$\begingroup$

Hint. One may use the standard evaluation $$ {2n\choose n}\cdot{\pi\over 2^{2n}}=\int_{0}^{\pi}\cos^{2n}{u}\:\mathrm du\tag1 $$ to deduce $$ \sum_{n=0}^{\infty}{{2n\choose n}^3\over 2^{6n}}=\frac1{\pi^3}\cdot\int \!\int\!\int_{\large\left[0,\pi\right]^3}\frac{du\:d\nu\:dw}{1-\cos u \cos \nu \cos w} \tag2 $$ the latter integral is one of Watson's triple integrals which have been evaluated by Watson ("Three Triple Integrals." Quart. J. Math., Oxford Ser. 2 10, 266-276, 1939) using clever and elementary changes of variable (see here): $$ \frac1{\pi^3}\cdot\int \!\int\!\int_{\large\left[0,\pi\right]^3}\frac{du\:d\nu\:dw}{1-\cos u \cos \nu \cos w} = {\Gamma^4\left({1\over 4}\right)\over 4\pi^3}={\pi\over \Gamma^4\left({3\over 4}\right)}.\tag3 $$

| cite | improve this answer | |
$\endgroup$
8
$\begingroup$

This is an extension of my comment to user "Start wearing purple"'s excellent answer and should be considered complementary to that.


You have stumbled onto a famous series from the classical theory of elliptic integrals, elliptic functions and theta functions. I provide a brief outline here. Let $0 < k < 1$ and $k'=\sqrt{1 - k^{2}}$ then we define a function $$K(k) = \int_{0}^{\pi/2}\frac{dx}{\sqrt{1 - k^{2}\sin^{2}x}}\tag{1}$$ which is normally called complete elliptic integral of first kind and mostly we drop the parameter $k$ and just use $K$ for the elliptic integral.

By expanding the integrand into an infinite series (using binomial theorem for general index) and integrating term by term we can obtain $$K = \frac{\pi}{2}\left\{1 + \left(\frac{1}{2}\right)^{2}k^{2} + \left(\frac{1\cdot 3}{2\cdot 4}\right)^{2}k^{4} + \cdots\right\}$$ or $$\frac{2K}{\pi} = {}_{2}F_{1}\left(\frac{1}{2}, \frac{1}{2}; 1; k^{2}\right)\tag{2}$$ Next we can transform the hypergeometric series on right into another form by using the formula $${}_{2}F_{1}\left(a, b; a + b + \frac{1}{2}; 4x(1 - x)\right) = {}_{2}F_{1}\left(2a, 2b; a + b + \frac{1}{2}; x\right)\tag{3}$$ which holds if $|x| < 1, |4x(1 - x)| < 1$ and $(a + b + (1/2))$ is neither zero nor a negative integer (see this blog post for proof). Applying $(3)$ on $(2)$ (using $a = 1/4, b = 1/4, x = k^{2}$) we get $$\frac{2K}{\pi} = {}_{2}F_{1}\left(\frac{1}{4},\frac{1}{4}; 1; (2kk')^{2}\right)\tag{4}$$ or in explicit form $$\frac{2K}{\pi} = 1 + \left(\frac{1}{4}\right)^{2}(2kk')^{2} + \left(\frac{1\cdot 5}{4\cdot 8}\right)^{2}(2kk')^{4} + \left(\frac{1\cdot 5\cdot 9}{4\cdot 8\cdot 12}\right)^{2}(2kk')^{6} + \cdots$$ which holds for $0 \leq k \leq 1/\sqrt{2}$. Now we need another piece of magic called Clausen's formula (proof available in the blog post linked earlier) $$\left({}_{2}F_{1}\left(a, b; a + b + \frac{1}{2}; z\right)\right)^{2} = \,{}_{3}F_{2}\left(2a, 2b, a + b; 2a + 2b, a + b + \frac{1}{2}; z\right)\tag{5}$$ Using $(4), (5)$ together (with $a = 1/4, b = 1/4, z = (2kk')^{2}$) we get $$\left(\frac{2K}{\pi}\right)^{2} =\,{}_{3}F_{2}\left(\frac{1}{2}, \frac{1}{2}, \frac{1}{2}; 1, 1; (2kk')^{2}\right)$$ or in explicit form $$\left(\frac{2K}{\pi}\right)^{2} = 1 + \left(\frac{1}{2}\right)^{3}(2kk')^{2} + \left(\frac{1\cdot 3}{2\cdot 4}\right)^{3}(2kk')^{4} + \left(\frac{1\cdot 3\cdot 5}{2\cdot 4\cdot 6}\right)^{3}(2kk')^{6} + \cdots\tag{6}$$ for $0 \leq k\leq 1/\sqrt{2}$. Your sum in question is the above series on right with $k = 1/\sqrt{2}$ so that $2kk' = 1$ and thus the value of the series is $(2K(1/\sqrt{2})/\pi)^{2}$.


Equation $(6)$ was sitting idly in the classical theory for a long time until Ramanujan appeared on the scene and in 1914 he differentiated the series $(6)$ (and some more series like it) with respect to $k$ (plus some highly non-obvious stuff) to obtain a class of series for $1/\pi$ the simplest of which is $$\frac{4}{\pi} = 1 + \frac{7}{4}\left(\frac{1}{2}\right)^{3} + \frac{13}{4^{2}}\left(\frac{1\cdot 3}{2\cdot 4}\right)^{3} + \frac{19}{4^{3}}\left(\frac{1\cdot 3\cdot 5}{2\cdot 4\cdot 6}\right)^{3} + \cdots\tag{7}$$ (see this blog post for details).

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.