How could one prove that

$$\sum_{k=0}^\infty \frac{2^{1-k} (3-25 k)(2 k)!\,k!}{(3 k)!} = -\pi$$

I've seen similar series, but none like this one...

It seems irreducible in current form, and I have no idea as to what kind of transformation might aid in finding proof of this.

  • 1
    $\begingroup$ This looks like a Generalized Hypergeometric Series evaluated at $n=\frac12$ (thus the $2^{-k}$ term in the numerator); it seems to bear a lot of similarity to some of Chudnovsky or Ramanujan's formulae for $\pi$. Have you tried writing the sum as a hypergeometric function and looking for any relevant identities? $\endgroup$ – Steven Stadnicki Jun 21 '13 at 18:48
  • $\begingroup$ @StevenStadnicki: It's almost certainly a GHS, but I can't prove it... $\endgroup$ – JohnWO Jun 21 '13 at 18:56
  • $\begingroup$ Related: math.stackexchange.com/questions/1652812/… $\endgroup$ – Jaume Oliver Lafont Feb 20 '16 at 22:39

Use Beta function, I guess... for $k \ge 1$, $$ \int_0^1 t^{2k}(1-t)^{k-1}dt = B(k,2k+1) = \frac{(k-1)!(2k)!}{(3k)!} $$ So write $$ f(x) = \sum_{k=0}^\infty \frac{2(3-25k)k!(2k)!}{(3k)!}x^k $$ and compute $f(1/2)$ like this: $$\begin{align} f(x) &= 6+\sum_{k=1}^\infty \frac{(6-50k)k(k-1)!(2k)!}{(3k)!} x^k \\ &= 6+\sum_{k=1}^\infty (6-50k)k x^k\int_0^1 t^{2k}(1-t)^{k-1}dt \\ &= 6+\int_0^1\sum_{k=1}^\infty (6-50k)k x^k t^{2k}(1-t)^{k-1}\;dt \\ &= 6+\int_0^1 \frac{4t^2x(14t^3x-14t^2x-11)}{(t^3x-t^2x+1)^3}\;dt \\ f\left(\frac{1}{2}\right) &= 6+\int_0^1\frac{16t^2(7t^3-7t^2-11)}{(t^3-t^2+2)^3}\;dt = -\pi \end{align}$$


of course any calculus course teaches you how to integrate a rational function...

  • $\begingroup$ Why the rewrite of $f(x)$ such that $f(x) = 6+\sum\cdots$? $\endgroup$ – JohnWO Jun 21 '13 at 20:26
  • $\begingroup$ That Beta integral diverges when $k=0$. $\endgroup$ – GEdgar Jun 21 '13 at 20:30
  • $\begingroup$ Sorry, I'm unable to interpret your answer. $\endgroup$ – JohnWO Jun 21 '13 at 20:47
  • $\begingroup$ +1, this is about what I expected. (Also note that by the rational root theorem and a bit of trial-and-error, $t+1$ is a factor of the cubic in the denominator.) @JohnWO , what part of this is unclear to you? $\endgroup$ – Steven Stadnicki Jun 21 '13 at 21:34
  • 1
    $\begingroup$ @JohnWO: To detail a little GEdgar's fine answer : $\sum_{k=0}\cdots=6+\sum_{k=1}\cdots$ (to avoid later difficulties with $k=0$) $$f(x):=\sum_{k=1}^\infty \left(x\,t^2\,(1-t)\right)^k=\frac{t^2x(1-t)}{1-t^2\,x\,(1-t)}$$ $$xf(x)':=\sum_{k=1}^\infty k\left(x\,t^2\,(1-t)\right)^k=\frac{t^2x(1-t)}{(1-t^2\,x\,(1-t))^2}$$ $$x(xf(x)')':=\sum_{k=1}^\infty k^2\left(x\,t^2\,(1-t)\right)^k=\frac{t^2x(1-t)(1+t^2\,x\,(1-t))}{(1-t^2\,x\,(1-t))^3}$$ so that (rewriting $(1-t)^{k-1}=\frac{(1-t)^k}{1-t}$) : $\endgroup$ – Raymond Manzoni Jun 21 '13 at 21:35

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.