3
$\begingroup$

(This is partly recreational math, playing with this question originally looking for a useful and fast converging representation for $\pi$ ).

A bit aside of fiddling with series which approximate $\pi$ I came to that following variant $$ s(x) = \sum_{k=0}^\infty {x^k\over \binom{2k+1}{k} \cdot \binom{2k+3}{2}} $$ which comes out to be related to $\pi$ (by 100 and more digits precision) in the following way: $$ \begin{array} {} s(1) &= \frac 41(1- { \pi \over 2 \sqrt 3}) \\ s(2) &=\frac 42(1- {\pi \over 4}) &=2 - { \pi \over 2} \\ s(3) &= \frac 43(1 - { \pi \over 3 \sqrt 3 } )\\ \end{array} $$
which looks somehow intriguing for me.


Background: I arrived at this series, when I invented a series for computation of $\pi$ by collecting partial expressions from the Leibniz-series, coming to $$ \pi=4 - 2( {0! \over 3!!}+{ 1!\over5!! }+{2! \over 7!! }+{ 3! \over 9!! } + ...) $$ (which is of course well known, accordingly to mathworld, even since Euler).

To have this series more smooth I converted the factorial and doublefactorial expressions into binomial-expressions, getting increasing powers of 2 in the numerators which gave me the idea to define this as $s(2)$ and to play with other $x$, giving a nice value for $s(3)$. (But I've nothing so far for, say $s(2.5)$ etc.)


For $s(4)$ the series converges very slowly, I summed up to 140 000 and to 140 200 terms and get the results

%3078 = 0.99763 1473716   \\ 140 000 terms
%3079 = 0.99763 3163689   \\ 140 200 terms

and I guess, that it might eventually run to 1.0 . Because it seems that there will be no index $k$ from where the rate of convergence improves, much likely an analytical approach is required. So my question:

Q: What would be an analytical expression for the sum? Or could we improve the serial summation with some convergence-acceleration? And of course: is the final value $s(4)=1$ ?

$\endgroup$
  • 1
    $\begingroup$ Hint: This is merely another play on the lame old, same old $\displaystyle\sum_{n=1}^\infty\frac{(2x)^{2n}}{n^2~\displaystyle{2n\choose n}}=2\arcsin^2x$. Notice that $\displaystyle{2n+3\choose2}$ is a humble polynomial in n. $\endgroup$ – Lucian Sep 5 '15 at 21:06
  • $\begingroup$ @Lucian 's comment tells pretty everything. $\endgroup$ – Jack D'Aurizio Sep 5 '15 at 21:08
6
$\begingroup$

That kind of series are strictly related with the arcsine function. See, for instance, this well-known question. By using the Euler beta function for converting the series into an integral, we have:

$$ s(x)=\frac{4}{x}-\frac{4\sqrt{4-x}}{x^{3/2}}\,\arcsin\left(\frac{\sqrt{x}}{2}\right) \tag{1}$$ hence $$ \lim_{x\to 4^-} s(x) = \color{red}{1} \tag{2}$$ as conjectured.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Wow, that's neat. Thank you very much! $\endgroup$ – Gottfried Helms Sep 5 '15 at 21:08
  • $\begingroup$ @GottfriedHelms: you're welcome! $\endgroup$ – Jack D'Aurizio Sep 5 '15 at 21:13
  • $\begingroup$ Ah, and after some simplification and substituting $y=4/x$ I get, feeding wolframalpha: the above series when expanded around $y=\infty$... I surely should get more experience with that "around infinity" thing... $\endgroup$ – Gottfried Helms Sep 5 '15 at 21:19

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.