Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Is there a closed form for the following infinite product? $$\prod_{n=1}^\infty\sqrt[2^n]{\frac{\Gamma(2^n+\frac{1}{2})}{\Gamma(2^n)}}$$

share|cite|improve this question
up vote 27 down vote accepted

The beautiful idea of Raymond Manzoni can actually be made rigorous. Consider a finite product $\prod_{n=1}^{L}$ and take its logarithm. After using duplication formula for the gamma function and telescoping, it simplifies to the following: $$\sum_{n=1}^{L}\frac{1}{2^n}\ln\frac{\Gamma(2^n+\frac12)}{\Gamma(2^n)}=\left(1-2^{-L}\right)\ln\left(2\sqrt{\pi}\right)-2L\ln2+2\cdot\frac{1}{2^{L+1}}\ln\Gamma(2^{L+1}).$$ This is an exact relation, valid for any $L$. Now it suffices to use Stirling, $$\frac{1}{N}\ln\Gamma(N)=\ln N-1+O\left(\frac{\ln N}{N}\right)\qquad \mathrm{as}\; N\rightarrow\infty$$ to get $$\sum_{n=1}^{\infty}\frac{1}{2^n}\ln\frac{\Gamma(2^n+\frac12)}{\Gamma(2^n)}=\ln\left(2\sqrt{\pi}\right)+2\left(\ln 2-1\right)=\ln\frac{8\sqrt{\pi}}{e^2}.$$ So the answer is indeed $\displaystyle\frac{8\sqrt{\pi}}{e^2}$.

share|cite|improve this answer
+1: Nice and short O.I. ! – Raymond Manzoni May 12 '13 at 20:46

I got $\quad \displaystyle 8\sqrt{\pi}\,e^{-2}$

Now let's search a proof...

I'll use the classical 'duplication formula' for $\Gamma$ : $$\Gamma(z)\Gamma\left(z+\frac 12\right)=\sqrt{2\pi}\ 2^{1/2-2z}\,\Gamma(2z) $$ so that we have : $$\frac{\Gamma(2^n)\Gamma\left(2^n+\frac 12\right)}{2^{2^{n+1}}\Gamma\left(2^{n+1}\right)}=2\sqrt{\pi}\ $$ and I thought at using 'kind of multiplicative telescoping' but it seems that considering telescoping of the logarithms as nicely done by O.L. is less confusing!

share|cite|improve this answer
Did you get this by numeric evaluation? Then it could be just an accidental close value due to a finite precision. – Liu Jin Tsai May 12 '13 at 18:39
@LiuJinTsai: With hundreds of digits no I don't think so (I'll update my answer in nobody gives the right answer faster) – Raymond Manzoni May 12 '13 at 18:43
Hello Raymond, how do you find such guesses? which software do you use? – bluemoon May 12 '13 at 18:49
@RaymondManzoni There exist arbitrarily close pairs of real numbers. – Liu Jin Tsai May 12 '13 at 18:53
@bluemoon: I didn't have the time to elaborate earlier but you may enjoy this thread showing PSLQ at work when... it fails ! :-) (of course its main interest is to discover 'exact' relations). – Raymond Manzoni May 12 '13 at 21:29

Since $$ \Gamma\left(n+\frac12\right)=\frac{(2n)!}{n!4^n}\sqrt\pi\quad\text{and}\quad\Gamma(n)=(n-1)! $$ We have $$ \frac{\Gamma\left(n+\frac12\right)}{\Gamma(n)}=\color{#C00000}{\binom{2n}{n}}\frac{\color{#0000FF}{n}}{\color{#00A000}{4^n}}\sqrt\pi $$ Letting $n=2^k$, and taking logs, we get the log of the partial product to be $$ \begin{align} &\sum_{k=1}^n\left(\color{#C00000}{\frac1{2^k}\log\left(2^{k+1}!\right)-\frac1{2^{k-1}}\log\left(2^k!\right)}\color{#00A000}{-\log(4)}\right)+\sum_{k=1}^\infty\left(\color{#0000FF}{\frac{k}{2^k}\log(2)}+\frac1{2^{k+1}}\log(\pi)\right)\\ &=\color{#C00000}{\frac1{2^n}\log\left(2^{n+1}!\right)-\log(2)}\color{#00A000}{-n\log(4)}\color{#0000FF}{+2\log(2)}+\frac12\log(\pi)\\ &\stackrel{\text{Stirling}}\sim\frac1{2^n}\left(2^{n+1}(n+1)\log(2)-2^{n+1}+\frac12\log(2^{n+2}\pi)\right)+(1-2n)\log(2)+\frac12\log(\pi)\\ &=(2n+2)\log(2)-2+\frac1{2^{n+1}}\log(2^{n+2}\pi)+(1-2n)\log(2)+\frac12\log(\pi)\\ &=\log(8)-2+\frac12\log(\pi)+\frac1{2^{n+1}}\log(2^{n+2}\pi)\\ &\stackrel{n\to\infty}\to\log(8)-2+\frac12\log(\pi) \end{align} $$ Thus, the limit of the product is $$ \frac{8\sqrt\pi}{e^2} $$

share|cite|improve this answer
I hope this is not just a duplicate of other answers. I had to go out in the middle of writing it up. – robjohn May 13 '13 at 0:09
It appears that O.L. has the only other complete solution, and I think mine is different enough that I will leave it. I did not use Raymond Manzoni's hint, but did use the telescoping of the red terms. I did limit the blue and black terms after the first equality rather than waiting until the last equality. – robjohn May 13 '13 at 3:11

Firstly, $\Gamma\left(m+\frac{1}{2}\right) = \frac{(2m)!}{4^mm!}\sqrt{\pi}$ when $m\in\mathbb{Z}$, so if we let $m=2^n$, what you have can be written as

$$ \frac{\Gamma\left(m+\frac{1}{2}\right)}{\Gamma\left(m\right)} = \frac{\frac{(2m)!}{4^mm!}\sqrt{\pi}}{(m-1)!} = \frac{(2^{n+1})!\sqrt{\pi}}{4^{2^n}2^n(2^n-1)!^2}. $$

Perhaps someone else can see how to simplify this further because I sure can't.

share|cite|improve this answer

This uses the hints by Raymond Manzoni and Cameron Williams. I am not sure if the answer is correct, but this should give an idea of how to proceed. I believe the answer by Raymond Manzoni is correct.

$$ \dfrac{\Gamma\left(2^n+\frac{1}{2}\right)}{\Gamma\left(2^n\right)} = \dfrac{2^n(2^{n+1})!\sqrt{\pi}}{4^{2^n}(2^n)!^2}$$ Hence, \begin{align} \log \left( \dfrac{\Gamma\left(2^n+\frac{1}{2}\right)}{\Gamma\left(2^n\right)}\right) & = n \log(2) + \dfrac{\log(\pi)}2 - 2^{n+1} \log(2) + \log \left(\dbinom{2^{n+1}}{2^n}\right) \end{align} \begin{align} S_n = \dfrac{\log \left( \dfrac{\Gamma\left(2^n+\frac{1}{2}\right)}{\Gamma\left(2^n\right)}\right)}{2^n} & = \dfrac{n \log(2)}{2^n} + \dfrac{\log(\pi)}{2^{n+1}} - 2 \log(2) + \dfrac1{2^{n}}\log \left(\dbinom{2^{n+1}}{2^n}\right) \end{align} Hence, $$\sum_{n=1}^{\infty}\dfrac{n \log(2)}{2^n} = 2 \log(2)$$ $$\sum_{n=1}^{\infty}\dfrac{\log(\pi)}{2^{n+1}} = \dfrac{\log(\pi)}2$$ Recall that $$\log \left(\dbinom{2k}k\right) \sim 2k \log(2) - \dfrac{\log(\pi)}2 - \dfrac{\log(k)}2$$ This gives us $$\log \left(\dbinom{2 \cdot 2^n}{2^n}\right) \sim 2^{n+1} \log(2) - \dfrac{\log(\pi)}2 - n \dfrac{\log(2)}2$$ $$\dfrac{\log \left(\dbinom{2 \cdot 2^n}{2^n}\right)}{2^n} \sim 2 \log(2) - \dfrac{\log(\pi)}{2^{n+1}} - n \dfrac{\log(2)}{2^{n+1}}$$ This gives us $S_n \sim \dfrac{n}{2^{n+1}} \log(2)$ and hence I would hope (this step needs more justification and is probably wrong in fact) $$\sum_{n=1}^{\infty} S_n \approx \log(2)$$ Hence, the answer you are looking for is approximately $$e^{\log(2)} = 2$$

share|cite|improve this answer
I don't find this numerically – Start wearing purple May 12 '13 at 19:18
@O.L. Yes, it is probably wrong. I have used the $\approx$ sign to indicate this. Essentially, we need a better bound for $$\dfrac1{2^n} \log \left(\dbinom{2n}n \right)$$ – user17762 May 12 '13 at 19:23
@user17762: the asymptotic (and bounds) of this is given here. – Raymond Manzoni May 12 '13 at 19:29
@RaymondManzoni But we need the asymptotic of $$\sum_{n=1}^N \dfrac{\log \left(\dbinom{2 \cdot 2^n}{2^n}\right)}{2^n}$$ The asymptotic of $$\log \left(\dbinom{2 \cdot 2^n}{2^n}\right)$$ alone is not sufficient. – user17762 May 12 '13 at 19:31

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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