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

I always see various definitions of Fourier transform. A standard form is: $$\hat{f}(\xi)=\int_{\mathbb{R}^d}f(x)e^{-2\pi ix\cdot\xi}dx$$ and its attached inversion is $$f(x)=\int_{\mathbb{R}^d}\hat{f}(\xi)e^{2\pi ix\cdot\xi}d\xi$$ Another form is like this: $$\hat{f}(\xi)=\int_{\mathbb{R}^d}f(x)e^{-ix\cdot\xi}dx$$ and the inversion formula is $$f(x)=\frac{1}{(2\pi)^d}\int_{\mathbb{R}^d}\hat{f}(\xi)e^{ix\cdot\xi}d\xi$$

I believe they are actually the same and I try to find their relationship.

An article on ProofWiki says

There exist several slightly different definitions of the Fourier transform which are commonly used; they differ in the choice of the constant 2π inside the exponential and/or a multiplicative constant before the integral. Their properties are essentially the same, and by a simple change of variable one can always translate statements using one of the definitions into statements using another one.

So I tried change of variable: $$ \int_{\mathbb{R}^d}f(x)e^{-ix\cdot\xi}dx=(2\pi)^d\int_{\mathbb{R}^d}f(2\pi x)e^{-2\pi ix\cdot\xi}dx$$ But then since the variable in $f$ is $2\pi ix$ rather than $x$, I don't know how to deal with it. Can you please help? Thank you.

EDIT: According to James Edward Lewis, the change of variable should be $2\pi x$. I revised this.

share|cite|improve this question
up vote 5 down vote accepted

The difference between the first and the second form is in the definition of $\xi$ (so you should change the name of the variables correspondingly). Let me introduce the new notation $\tilde\xi$ and $\tilde f(\tilde \xi)$ such that the confusion is (hopefully) lifted.

You want to proof that $$ \hat{f}(\xi)=\int_{\mathbb{R}^d}f(x)e^{-2\pi ix\cdot\xi}dx \qquad f(x)=\int_{\mathbb{R}^d}\hat{f}(\xi)e^{2\pi ix\cdot\xi}d\xi$$ is equivalent to $$ \tilde{f}(\tilde\xi)=\int_{\mathbb{R}^d}f(x)e^{-ix\cdot\tilde\xi}dx \qquad f(x)=\frac{1}{(2\pi)^d}\int_{\mathbb{R}^d}\tilde{f}(\tilde\xi)e^{ix\cdot\tilde\xi}d\tilde\xi.$$

Let's take the first form as given and try to derive the second form. The Fourier transform is by definition the same with $\tilde\xi = 2\pi \xi$ and $\tilde f(\tilde \xi) = \hat{f} (\xi)$. For the inversion formula, we have to work a bit harder (and use substitution). $$f(x)=\frac{1}{(2\pi)^d}\int_{\mathbb{R}^d}\tilde{f}(\tilde\xi)e^{ix\cdot\tilde\xi}d\tilde\xi = \frac{1}{(2\pi)^d}\int_{\mathbb{R}^d}\tilde{f}(\tilde\xi)e^{2\pi ix\cdot\xi}d(2 \pi \xi) = \int_{\mathbb{R}^d}\hat{f}(\xi)e^{2\pi ix\cdot\tilde\xi}d\xi.$$

share|cite|improve this answer

The correct change of variable is x=2pi*X, and in the end you actually will be using a different function; to put it another way, the first form of Fourier transform on a function f is the same as the second form on its pullback by the linear map consisting of multiplication by 2pi, divided by (2pi)^d.

The properties are essentially the same, but the transforms are not exactly the same; the trade-off is between simplicity and asymmetry of inversion.

share|cite|improve this answer

Sorry Fabian, but i think James Edward Lewis II is absolutely right with his Statement, that $\tilde f(\tilde \xi) \neq \hat{f} (\xi)$ and there is an easy way of getting all definition of the Fourier transform:

  1. My favourite definition: $$ \hat f_1(k_1)=\int_{\mathbb{R}}f_1(x_1)e^{-i2\pi k_1 x_1}dx_1 \hspace{10mm} f_1(x_1)=\int_{\mathbb{R}}\hat f_1(k_1)e^{-i2\pi k_1 x_1}dk_1 \hspace{10mm} (1)$$

  2. Let $k_2=2\pi k_1$, so we can define for this new variable also the corresponding $\hat f_2(k_2)$, which is also a Fourier-transform of $f_1(x_1)$ with just other units, "radians/meter" if you like, in contrast to $\hat f_1(k_1)$, which has "circles/meter". The function $\hat f_2(k_2)$ is just a bit stretched, but with invariant area, therefore: $$ \hat f_2(k_2)|dk_2|=\hat f_1(k_1)|dk_1| \hspace{5mm} => \hspace{5mm} \hat f_2(k_2) = \hat f_1(k_1) \left|\frac{dk_1}{dk_2}\right|=\frac{1}{2\pi}\hat f_1(k_1)$$ $$=> \hspace{5mm}\hat f_2(k_2)=\frac{1}{2\pi}\int_{\mathbb{R}}f_1(x_1)e^{-ik_2 x_1}dx_1 \hspace{10mm}f_1(x_1)=\int_{\mathbb{R}}\hat f_2(k_2)e^{-ik_2 x_1}dk_2 \hspace{5mm}(2)$$

  3. Let $x_3 = 2\pi x_1$, and with the same argument, we have a new $f_3(x_3)$ with $$ f_3(x_3) = f_1(x_1) \left|\frac{dx_1}{dx_3}\right|=\frac{1}{2\pi} f_1(x_1)$$ $$=> \hspace{5mm}\hat f_1(k_1)=\int_{\mathbb{R}}f_3(x_3)e^{-ik_1 x_3}dx_3 \hspace{5mm}f_3(x_3)=\frac{1}{2\pi}\int_{\mathbb{R}}\hat f_1(k_1)e^{-ik_1 x_3}dk_1 \hspace{5mm}(3)$$

  4. Let $x_4 = \sqrt{2\pi} x_1$ and $k_4 = \sqrt{2\pi} k_1$, therefore we have a new function $f_4(x_4)$ and a new function $\hat f_4(k_4)$ with $$ f_4(x_4) = f_1(x_1) \left|\frac{dx_1}{dx_4}\right|=\frac{1}{\sqrt{2\pi}} f_1(x_1)$$ $$ \hat f_4(k_4) = \hat f_1(k_1) \left|\frac{dk_1}{dk_4}\right|=\frac{1}{\sqrt{2\pi}} \hat f_1(k_1)$$ $$=> \hspace{5mm}\hat f_4(k_4)=\frac{1}{\sqrt{2\pi}}\int_{\mathbb{R}}f_4(x_4)e^{-ik_4 x_4}dx_4 \hspace{5mm}f_4(x_4)=\frac{1}{\sqrt{2\pi}}\int_{\mathbb{R}}\hat f_4(k_4)e^{-ik_4 x_4}dk_4 \hspace{5mm}(4)$$

The first definition of the Fourier transform is much more symmetric and pretty, and therefore i want to show you a little math joke:

$$ \hat f(k)=\int_{\mathbb{R}}f(x)e^{-i2\pi k x}dx \stackrel{?}{=} \int_{\mathbb{R}}f(x)\left(e^{-i2\pi}\right)^{k x}dx = \int_{\mathbb{R}}f(x)1^{k x}dx = \hat f(0)$$

Does anyone get it?

share|cite|improve this answer
Velcome to the site! – kjetil b halvorsen May 30 '14 at 13:00

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.