63
$\begingroup$

It seems as if no one has asked this here before, unless I don't know how to search.

The Gamma function is $$ \Gamma(\alpha)=\int_0^\infty x^{\alpha-1} e^{-x}\,dx. $$ Why is $$ \Gamma\left(\frac{1}{2}\right)=\sqrt{\pi}\ ? $$ (I'll post my own answer, but I know there are many ways to show this, so post your own!)

$\endgroup$
  • $\begingroup$ Possible duplicate: math.stackexchange.com/questions/9286/… $\endgroup$ – Argon Oct 20 '12 at 17:35
  • 4
    $\begingroup$ @Argon Did you look at the other question? It's not even close to the same thing. And, how on earth did 2 other people vote to close without even looking at it? $\endgroup$ – Graphth Oct 20 '12 at 23:49
  • 6
    $\begingroup$ @Graphth With a simply change of variables ($x^2 = t$), the integrals are virtually identical, except for a constant coefficient. $\endgroup$ – Argon Oct 20 '12 at 23:52
  • $\begingroup$ Maybe not that question, but I am quite sure this has been asked before... $\endgroup$ – J. M. is a poor mathematician Oct 21 '12 at 6:09
  • 1
    $\begingroup$ math.uconn.edu/~kconrad/blurbs/.../gaussianintegral.pdf is a good source of methods to solve this. $\endgroup$ – Ishan Banerjee Mar 17 '13 at 8:07

11 Answers 11

47
$\begingroup$

We only need Euler's formula:

$$\Gamma(1-z) \Gamma(z) = \frac{\pi}{\sin \pi z} \Longrightarrow \Gamma^2\left(\frac{1}{2}\right ) = \pi $$

$\endgroup$
45
$\begingroup$

One can use the trick with spherical coordinates: $$ \Gamma\left(\frac{1}{2}\right) = \int_0^\infty \mathrm{e}^{-x} \frac{\mathrm{d} x}{\sqrt{x}} = \int_0^\infty \mathrm{e}^{-x} \mathrm{d} (2 \sqrt{x}) = \int_{-\infty}^\infty \mathrm{e}^{-u^2} \, \mathrm{d} u $$ Then: $$ \Gamma\left(\frac{1}{2}\right) \Gamma\left(\frac{1}{2}\right) = \int_{-\infty}^\infty \int_{-\infty}^\infty \mathrm{e}^{-x^2} \mathrm{e}^{-y^2} \, \mathrm{d}x \, \mathrm{d} y = \underbrace{\int_0^\infty r \mathrm{e}^{-r^2} \, \mathrm{d} r}_{\frac{1}{2}} \cdot \underbrace{\int_{0}^{2\pi} \, \mathrm{d} \phi}_{2 \pi} = \pi $$

Alternatively one can use the second Euler's integral: $$ \frac{\Gamma\left(\frac{1}{2}\right) \Gamma\left(\frac{1}{2}\right) }{\Gamma(1)} = \int_0^1 t^{-1/2} (1-t)^{-1/2} \mathrm{d}t = \int_0^1 \mathrm{d} \left(2\arcsin\left(\sqrt{t}\right)\right) = \pi $$ Now, using $\Gamma(1) = 1$ the result follows.

Yet another method is to use the duplication identity: $$ \Gamma(2s) = \frac{2^{2s-1}}{\sqrt{\pi}} \Gamma(s) \Gamma\left(s+\frac{1}{2}\right) $$ at $ s= \frac{1}{2}$.

$\endgroup$
21
$\begingroup$

Method 1: $$\Gamma(1/2) = \int_0^{\infty} x^{-1/2} \exp(-x) dx$$ Set $x = t^2$, to get $$\Gamma(1/2) = \int_0^{\infty} 2\exp(-t^2) dt = \sqrt{\pi}$$

Method 2: $$\beta(x,y) = \dfrac{\Gamma(x) \Gamma(y)}{\Gamma(x+y)}$$ Take $y=x=1/2$, to get $$\beta(1/2,1/2) = \dfrac{\Gamma^2(1/2)}{\Gamma(1)} \implies \Gamma(1/2) = \sqrt{\beta(1/2,1/2)}$$

$$\beta(x,y) = 2 \int_0^{\pi/2} \sin^{2x-1}(\theta) \cos^{2y-1}(\theta) d \theta$$ Hence, $$\beta(1/2,1/2) = 2 \int_0^{\pi/2} d \theta = \pi$$ $$\Gamma(1/2) = \sqrt{\pi}$$

$\endgroup$
  • 2
    $\begingroup$ Sorry for my ignorance, but how did $\frac 1 t$ become 2 in method 1? $\endgroup$ – Hassan Jun 23 '15 at 19:09
  • 3
    $\begingroup$ @Hassan dx = 2t dt ; this term cancels out 1/t and leaves 2. $\endgroup$ – Luciano Oct 25 '15 at 9:06
19
$\begingroup$

If there's any justice in the universe, someone must have asked here how to show that $$ \int_{-\infty}^\infty e^{-x^2/2}\,dx = \sqrt{2\pi}. $$ Let's suppose that has been answered here. Let (capital) $X$ be a random variable whose probability distribution is $$ \frac{e^{-x^2/2}}{\sqrt{2\pi}}\,dx. $$ Consider the problem of finding $\operatorname{E}(X^2)$. It is $$ \frac{1}{\sqrt{2\pi}}\int_{-\infty}^\infty x^2 e^{-x^2/2}\,dx = \text{(by symmetry)} \frac{2}{\sqrt{2\pi}} \int_0^\infty x^2 e^{-x^2/2}\,dx $$ $$ \sqrt{\frac2\pi}\int_0^\infty xe^{-x^2/2}\Big(x\,dx\Big) = \sqrt{\frac2\pi}\int_0^\infty \sqrt{u}\ e^{-u}\,du = \sqrt{\frac2\pi}\ \Gamma\left(\frac32\right) = \frac12\sqrt{\frac2\pi} \Gamma\left(\frac12\right). $$ So it is enough to show that this expected value is $1$. That is true if the sum of two independent copies of it has expected value $2$. So: $$ \Pr\left(X^2+Y^2<w\right) = \frac{1}{2\pi}\iint\limits_\mathrm{disk} e^{-(x^2+y^2)/2}\,dx\,dy $$ where the disk has radius $\sqrt{w}$. This equals $$ \frac{1}{2\pi}\int_0^{2\pi}\int_0^{\sqrt{w}} e^{-\rho^2/2} \,\rho\,d\rho\,d\theta = \int_0^{\sqrt{w}} e^{-\rho^2/2} \,\rho\,d\rho. $$ This last equality holds because we are integrating with respect to $\theta$ something not depending on $\theta$. Differentiating this with respect to $w$ gives the probability density function of the random variable $X^2+Y^2$: $$ e^{-w/2}\sqrt{w}\frac{1}{2\sqrt{w}} = \frac{e^{-w/2}}{2}\text{ for }w>0. $$ So $$ \operatorname{E}(X^2+Y^2) = \int_0^\infty w \frac{e^{-w/2}}{2}\,dw =2. $$

$\endgroup$
  • 13
    $\begingroup$ If $\int_{-\infty}^\infty e^{-u^2/2}\,du = \sqrt{2\pi}$ is given, then it seems much simpler to take $\Gamma(1/2) = \int_0^{\infty} x^{-1/2} \exp(-x) dx$, substitute $u^2=2x$, and we are done. $\endgroup$ – leonbloy Jun 17 '13 at 1:13
16
$\begingroup$

I'm surprised that no one has mentioned that this $\sqrt{\pi}$ is also "the same one" as the $\sqrt{\pi}$ in Stirling's formula.

It is a known fact that $\Gamma$ is uniquely characterized as the function that satisfies the equation $\Gamma(z+1) = z \Gamma(z)$ and whose logarithm "has nice asymptotics at $+\infty$". The conventional way to explain what "nice asymptotics" should mean in this context involves logarithmic convexity, but I prefer to leave it like this: any function that satisfies that functional equation must coincide with $\Gamma(z)$ up to a periodic factor, which, if nontrivial, would prevent it from having an asymptotic expansion at $+\infty$ in terms of powers and logs.

Now let's turn to Stirling's formula and assume that it holds both for integer and half-integer values of $\Gamma$. Clearly,

$$\Gamma(n+1/2) = \Gamma(1/2) \cdot \frac{1}{2} \cdot \frac{3}{2} \cdot \dots \cdot \frac{2n-1}{2} = \Gamma(1/2) \cdot \frac{(2n)!}{2^{2n} n!},$$

which is basically the duplication formula. Now if we plug it into the Stirling's formula, we will find out that $\Gamma(1/2) = \sqrt{\pi}$.

$\endgroup$
11
$\begingroup$

It also follows from the famous Riemann functional equation with $s=1/2$: $$ \zeta(s) = 2^s\pi^{s-1}\ \sin\left(\frac{\pi s}{2}\right)\ \Gamma(1-s)\ \zeta(1-s) $$ but Euler's reflection formula is probably more basic.

$\endgroup$
11
$\begingroup$

This is a "proof". We know that the surface area of the $n-1$ dimensional unit sphere is $$ |S^{n-1}| = \frac{2\pi^{\frac{n}2}}{\Gamma(\frac{n}2)}. $$ On the other hand, we know that $|S^2|=4\pi$, which gives $$ 4\pi = \frac{2\pi^{\frac32}}{\Gamma(\frac32)} = \frac{2\pi^{\frac32}}{\frac12\Gamma(\frac12)}. $$

$\endgroup$
  • 3
    $\begingroup$ This formula also holds for $n = 1$ if you take counting measure as your $0$-dimensional measure. $\endgroup$ – JavaMan Dec 23 '12 at 18:09
7
$\begingroup$

There are a few details missing (showing the integral converges, justifying the differentiation under the integral sign, and so forth), but the main idea should be clear.

Consider the Fourier transform of $f(x)=e^{-x^2}$.

$$\hat f(\xi)= \int_{-\infty}^{\infty} e^{-x^2} e^{-2\pi i x\xi} \ dx =\int_{-\infty}^{\infty} e^{-x^2} (\cos(2\pi x \xi) - i \sin(2\pi x \xi)) dx =\int_{-\infty}^{\infty} e^{-x^2} \cos(2\pi x \xi) \ dx$$

Differentiate under the integral sign and integrate by parts.

$$\hat f'(\xi) =-2\pi \int e^{-x^2}\sin(2\pi \xi x) \ dx = -2\pi^2 \xi \int e^{-x^2} \cos(2\pi x \xi) \ dx = -2\pi^2\xi \hat f(\xi)$$

Separate variables.

$$\frac{\hat f'(\xi)}{\hat f(\xi) }= -2\pi^2 \xi \Rightarrow \hat f(\xi) =Ce^{-\pi^2 \xi^2}.$$

Use the inversion formula and change variables .

$$e^{-x^2} = C\int e^{-\pi^2 \xi^2} \cos(2\pi x \xi) \ d\xi = \frac{C}{\pi} \int e^{- \xi^2} \cos(2 x \xi) \ d\xi $$

$$e^{-\pi^2 x^2}\frac{C}{\pi} =\int e^{- \xi^2} \cos(2 x\pi \xi) \ d\xi =\frac{C^2}{\pi}e^{-\pi^2x^2}\Rightarrow C=\sqrt{\pi}$$

$$\hat f (0) = \sqrt{\pi} = \int e^{-x^2} \cos(0) \ dx = \int e^{-x^2} \ dx.$$

$\endgroup$
6
$\begingroup$

$$B\left(\frac{1}{2},\frac{1}{2}\right)=\frac{\left[\Gamma\left(\frac{1}{2}\right)\right]^{2}}{\Gamma{(1)}}=\left[\Gamma\left(\frac{1}{2}\right)\right]^{2}$$ $$B\left(\frac{1}{2},\frac{1}{2}\right)=\frac{\pi}{\sin{\frac{\pi}{2}}}=\pi$$

Chris.

$\endgroup$
  • 3
    $\begingroup$ For those who find the second line not obvious, here is an elementary explanation: by definition, $B(\frac{3}{2},\frac{3}{2}) = \intop_0^1 x^{1/2} (1-x)^{1/2} dx$, and the graph of $x^{1/2} (1-x)^{1/2}$ is a semicircle, so this directly relates to $\pi$ being defined via the area of the circle. This allows to calculate $\Gamma(\frac{3}{2})$, which is just $\frac{1}{2} \Gamma(\frac{1}{2})$ $\endgroup$ – Alexander Shamov Jun 16 '13 at 22:27
3
$\begingroup$

The Gamma function is $$\Gamma(\alpha)=\int_0^\infty x^{\alpha-1} e^{-x}\,dx$$

Therefore

$$\Gamma\left( \frac{1}{2}\right) =\int_0^\infty \frac{1}{\sqrt{x}} e^{-x}\,dx$$

Thus, after the change of variable $t=\sqrt{x}$, this turns into the Euler integral $$t=\sqrt{x} \implies x = t^2$$ $$\frac{dt}{dx} = \frac{1}{2\sqrt{x}} = \frac{1}{2t} \iff dx = 2t\,dt$$

We have $$\int_0^\infty \frac{1}{t} e^{-t^2}\,2t\,dt = 2\int_0^\infty e^{-t^2}\,dt $$

And following holds: $$\Gamma\left( \frac{1}{2}\right) = 2\int_0^\infty e^{-t^2}\,dt = \int_{-\infty}^\infty e^{-t^2}\,dt$$

Which $\int_{-\infty}^\infty e^{-t^2}\,dt$ is the Gaussian integral. It follows:

$$\left[ \Gamma\left( \frac{1}{2}\right)\right]^2 = \bigg(\int_{-\infty}^{+\infty} e^{-x^2} dx\bigg)\bigg(\int_{-\infty}^{+\infty} e^{-y^2} dy\bigg)$$ $$\int_{-\infty}^{+\infty}\int_{-\infty}^{+\infty}e^{-(x^2+y^2)} \,dx\,dy$$

Now change to polar coordinates
$$\int_0^{+2 \pi}\int_0^{+\infty}e^{-r^2} r\,dr\,d\theta$$

The $\theta$ integral just gives $2\pi$, while the $r$ integral succumbs to the substitution $u=r^2$
$$\left[ \Gamma\left( \frac{1}{2}\right)\right]^2 = 2\pi\int_{0}^{+\infty}e^{-u} \,du/2=\pi$$ $$\Gamma\left( \frac{1}{2}\right)=\sqrt{\pi}$$

$\endgroup$
  • $\begingroup$ You are 5 years late my friend. Sasha already has this in his/her answer. $\endgroup$ – Simply Beautiful Art Aug 8 '17 at 23:59
  • $\begingroup$ Rather than computing $\dfrac{dt}{dx}$ and then deducing $2t\,dt=dx$ from that, I would differentiate both sides of $t^2 = x$ to get $2t\,dt = dx. \qquad$ $\endgroup$ – Michael Hardy Aug 11 '17 at 1:50
2
$\begingroup$

By the Bohr-Mollerup theorem,

$$\Gamma(1/2)=\lim_{n\to\infty}\frac{\sqrt n(n!)}{(1/2)(-1/2)\dots(1/2-n)}=\lim_{n\to\infty}\frac{\sqrt n4^{n+1}(n!)^2}{(2n)!(n+\frac12)}$$

Apply the Stirling approximation and watch almost everything simplify!

$$\Gamma(1/2)=\sqrt\pi\lim_{n\to\infty}\frac n{n+\frac12}=\sqrt\pi$$

$\endgroup$

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.