2
$\begingroup$

All I know of the Gamma function is $$\Gamma(\alpha) = \int\limits_{0}^{\infty}x^{\alpha - 1}e^{-x}\text{ d}x$$ and the recursive formula $$\Gamma(\alpha) = (\alpha-1)\Gamma(\alpha - 1)\text{.}$$ A friend asked me to try to find $\Gamma(2.7)$. I assume this involves complex analysis, which I have absolutely no background on (except for the basic Euler formula). Using Wolfram, I found that $$\Gamma(2.7) = e^{0.434821}$$ which suggests to me perhaps writing out $e^{-x}$ as a Taylor series: $$\Gamma(\alpha) = \int\limits_{0}^{\infty}x^{\alpha - 1}\sum\limits_{i=1}^{\infty}\dfrac{(-x)^{i}}{i!}\text{ d}x$$ but alas, this doesn't look helpful.

How does one derive that $\Gamma(2.7) = e^{0.434821}$?

$\endgroup$
3
$\begingroup$

May be slightly better than Stirling, you could use Gosper approximation $$n! \approx \sqrt{\pi(2n+\frac{1}{3})} \Big(\frac{n}{e}\Big)^n$$ For $n=1.7$, this gives an approximate value of $1.54201$ while Stirling leads to $1.47156$ while the "exact" value is $1.54469$.

If you continue using Stirling series, that is to say $$n!=\sqrt{2 \pi } e^{-n} n^{n+\frac{1}{2}}(1+\frac{1}{12 n}+\frac{1}{288 n^2}-\frac{139}{51840 n^3}-\frac{571}{2488320 n^4}+\cdots+O\left(\left(\frac{1}{n}\right)^6\right))$$ you successively find $1.47156$, $1.54369$, $1.54546$,$1.54466$, $1.54462$ ,$\cdots$.

$\endgroup$
  • 1
    $\begingroup$ If all you have is Gosper (or Stirling) approximation, you can get a few more digits (one or two) out of it by writing $1.7!=\frac{(1.7+m)!}{2.7\times\cdots\times(1.7+m)}$ for a small integer $m$, which only requires a few more operations. $\endgroup$ – Kirill Sep 18 '14 at 7:36
1
$\begingroup$

Hint: Use Stirling's approximation in conjunction with the fact that $e\simeq2.7$.

$\endgroup$
  • 2
    $\begingroup$ It feels like a gross abuse of the word "approximation" to invoke Stirlings formula for $n=2.7$. Unfortunately this is probably the logic wolframalpha used. $\endgroup$ – Alex R. Sep 18 '14 at 4:01
  • $\begingroup$ @AlexR.: It is actually quite surprising how small one can make $|z|$ for use in Stirling's formula for the Gamma function. With maybe 3 or 4 terms in the asymptotic relation one may see in, e.g., dlmf.nist.gov/5.11 , one may get a reasonable accuracy. $\endgroup$ – Ron Gordon Dec 29 '15 at 0:13
  • $\begingroup$ @AlexR. You may also use the relation $\Gamma(z+1)=z\Gamma(z)$ to shift the argument to larger values, where the Stirling approximation is better. $\endgroup$ – Jean-Claude Arbaut May 30 at 0:21
0
$\begingroup$

As stated in the comments, the Legendre duplication formula for the $\Gamma$ function gives: $$ \color{blue}{\Gamma\left(\frac{27}{10}\right)}=\frac{119}{100}\cdot 2^{3/5}\cdot\left(\color{red}{\frac{\Gamma\left(\frac{1}{5}\right)}{\Gamma\left(\frac{2}{5}\right)}}\right)^{-1}\cdot\sqrt{\pi} \tag{1}$$ hence it is enough to compute the red ratio with reasonable precision.

Due to the Euler product for the $\Gamma$ function we have: $$ \color{red}{\frac{\Gamma\left(\frac{1}{5}\right)}{\Gamma\left(\frac{2}{5}\right)}} = \lim_{n\to +\infty}n^{-1/5}\cdot\frac{\frac{2}{5}\cdot\left(\frac{2}{5}+1\right)\cdot\ldots\cdot\left(\frac{2}{5}+n\right)}{\frac{1}{5}\cdot\left(\frac{1}{5}+1\right)\cdot\ldots\cdot\left(\frac{1}{5}+n\right)}\tag{2}$$ and the RHS can be explicitly evaluated at $n=32$ to get an approximation with two accurate digits, or at $n=2^{20}$ to get an approximation with six accurate figures. Since $\frac{1}{5}$ and $\frac{2}{5}$ are quite close to zero and the $\Gamma$ function has a simple pole at the origin, it is reasonable to expect that the red ratio is close to $\color{red}{2}$, and indeed it is: $$ \color{red}{\frac{\Gamma\left(\frac{1}{5}\right)}{\Gamma\left(\frac{2}{5}\right)}}\approx \color{red}{2}.06966344. \tag{3}$$

$\endgroup$
0
$\begingroup$

You could use a Riemann sum to approximate $\Gamma(2.7)$

$\Gamma(2.7)$ $\approx$ $\sum_{x=1}^\infty e^{-x}x^{1.7}$

This comes out to about 1.52485

You can use Riemann sums to evaluate the gamma function of complex numbers too.

$\endgroup$

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.