Complex-Analytic Proof of Fourier Transform of Gaussian

Many weeks ago, I wrote a post on how to compute the Fourier transform of the Gaussian function defined by f(x):=e^{-\pi x^{2}} using the methods of ordinary differential equations. Specifically, we showed that

\widehat{f}(\xi)=\int_{\mathbb{R}}e^{-\pi x^{2}}e^{-2\pi i\xi x}dx=e^{-\pi\xi^{2}},\indent\forall\xi\in\mathbb{R}

In that post, I promised to later show how to prove the same result by means of complex analysis–Cauchy’s theorem for closed contours. As I keep my promises, though not necessarily with the promptness others might like, I will now give this second proof.

Consider the Fourier transform of the function f(x):=e^{-ax^{2}}, for a > 0 fixed:

\displaystyle\widehat{f}(\xi)=\int_{\mathbb{R}}e^{-ax^{2}}e^{-2\pi i \xi x}dx=\int_{\mathbb{R}}e^{-a(x^{2}+2\pi i \xi a^{-1}x)}dx,\indent\forall \xi\in\mathbb{R}

We can complete the square in the exponent to obtain

\displaystyle\widehat{f}(\xi)=\int_{\mathbb{R}}e^{-a(x+\pi i \xi a^{-1})^{2}-\pi^{2}\xi^{2}a^{-2}}dx=e^{-\pi^{2}\xi^{2}a^{-1}}\int_{\mathbb{R}}e^{-a(x+i\pi\xi a^{-1})^{2}}dx,\indent\forall\xi\in\mathbb{R}

Consider the entire function f(z)=e^{-az^{2}}, where a>0 is a fixed real number. Let \mathcal{C}_{R} denote the positively oriented rectangular contour with vertices \pm R,\pm R+i\pi\xi a^{-1}, for R>0. By Cauchy’s theorem,

\displaystyle\int_{\mathcal{C}_{R}}f(z)dz=0

By the dominated convergence theorem,

\begin{array}{lcl}\displaystyle\int_{\mathbb{R}}e^{-a(x+i\pi\xi a^{-1})^{2}}dx&=&\displaystyle\lim_{R\rightarrow\infty}\int_{-R}^{R}e^{-a(x+i\pi\xi a^{-1})^{2}}dx\\&=&\displaystyle\lim_{R\rightarrow\infty}\int_{-R}^{R}e^{-a x^{2}}dx+i\int_{0}^{\pi\xi a^{-1}}e^{-a(R+it)^{2}}dx-i\int_{0}^{\pi\xi a^{-1}}e^{-a(-R+it)^{2}}dt\end{array}

Let’s examine the growth of the second and third integral as R\rightarrow\infty.

\displaystyle\left|\int_{0}^{\pi\xi a^{-1}}e^{-a(\pm R+it)^{2}}idt\right|\leq e^{-aR^{2}}\int_{0}^{\pi\xi a^{-1}}e^{at^{2}}dt\rightarrow 0, R\rightarrow\infty

It is well-known identity that \int_{\mathbb{R}}e^{-x^{2}}dx=\sqrt{\pi}, so by making the change of variable y=\sqrt{a}x, we obtain

\begin{array}{lcl}\displaystyle\lim_{R\rightarrow\infty}\int_{-R}^{R}e^{-ax^{2}}dx=a^{-\frac{1}{2}}\lim_{R\rightarrow\infty}\int_{-\sqrt{a}R}^{\sqrt{a}R}e^{-y^{2}}dy=\displaystyle a^{-\frac{1}{2}}\int_{\mathbb{R}}e^{-y^{2}}dy&=&\sqrt{\dfrac{\pi}{a}}\end{array},

where we use the dominated convergence theorem to obtain that penultimate equality. We conclude that

\displaystyle\widehat{f}(\xi)=\sqrt{\dfrac{\pi}{a}}e^{-\pi^{2}\xi^{2}a^{-1}},\indent\forall \xi\in\mathbb{R}

In particular, if a=\pi, then f is a fixed point of the Fourier transform.

Advertisements
This entry was posted in math.CA, math.CV and tagged , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s