0

What are the steps to calculate Fourier transform of $e^{-4t^2} $ ?

If it was a form of $|t|$ maybe it would be easier, but as it is i cannot find a way.

Jean Marie
  • 88,997
  • You mean fourier transform? This is a function defined on all reals, so it'll have a continuous spectrum. Anyway, the Gaussian function is famous for being its own fourier transform (up to some factors in the exponent). – orion May 10 '16 at 09:03
  • yes, fourier transform. So you mean that the fourier transform of this is itself? – Christos I. May 10 '16 at 09:09
  • I think if i write $t^2=|t^2|$ i can proceed in a way. – Christos I. May 10 '16 at 09:10
  • The fourier transform is another Gaussian function, something like $A e^{-k\omega^2}$ where $A$ and $k$ actually depend on what definition you take for the fourier transform. There are different conventions. – orion May 10 '16 at 09:10
  • If you have selected the tag "differential equations", there is a reason (an hint that has been given to you ?). So, what elementary differential equation can we find having your function as solution (compare the function to its derivative)? Then apply Fourier transform to both sides of the diff. equ. – Jean Marie May 10 '16 at 09:15
  • We have the definition that $ F(w)= \int_{-\infty}^\infty f(t) e^{-iwt} }dt $ – Christos I. May 10 '16 at 09:17
  • You wrote differential-equation, then you can look at http://math.stackexchange.com/questions/1637994/fourier-transform-of-ft-4te-t2/1638023#1638023 . – reuns May 10 '16 at 09:17

2 Answers2

1

Let be $f(t) = \mathrm e^{-at^2}$ and $$\mathcal F\{f(t)\}=\hat{f}(\omega)= \int_{-\infty}^{\infty} f(t)\mathrm e^{- i\omega t} \,\mathrm dt = \int_{-\infty}^{\infty} \mathrm e^{-at^2}\mathrm e^{-i\omega t}\,\mathrm dt $$ Differentiating with respect to $\omega$ yields $$\frac{\mathrm d}{\mathrm d\omega} \hat{f}(\omega) = \int_{-\infty}^{\infty} \mathrm e^{-at^2}(-i t)\mathrm e^{-i\omega t}\,\mathrm dt= \frac{i}{2a} \int_{-\infty}^{\infty} \left(\frac{\mathrm d}{\mathrm dt} \mathrm e^{-at^2} \right) \mathrm e^{- i\omega t} \,\mathrm dt$$ Integrating by parts, we obtain

$$\hat{f}'(\omega) = - \frac{\omega}{2a} \int_{-\infty}^{\infty} \mathrm e^{-at^2}\mathrm e^{-i\omega t}\,\mathrm dt= - \frac{\omega}{2a} \hat{f}(\omega)$$

The unique solution to this ordinary differential equation is given by

$$\hat{f}(\omega) =\beta \cdot \exp \left(- \frac{\omega^2}{4a} \right)$$

where the constant $\beta$ is $$\beta=\hat{f}(0) = \int_{-\infty}^{\infty} \mathrm e^{-at^2}\,\mathrm dt=\sqrt{\frac{\pi}{a}}$$ It follows that $$\mathcal F\left\{\mathrm e^{-at^2}\right\}=\hat{f}(\omega) =\sqrt{\frac{\pi}{a}} \exp \left(- \frac{\omega^2}{4a} \right)$$ and for $a=4$ we have $$\mathcal F\left\{\mathrm e^{-4t^2}\right\}=\sqrt{\frac{\pi}{4}} \mathrm e^{- \left(\omega/4\right)^2}$$

alexjo
  • 15,328
0

Assuming $t \in \mathbb{R^1},$ the task is to compute $F(e^{-4t}) \equiv \int_{-\infty}^\infty e^{i\xi t} e^{-4t^2} \; dt = {\hat f}(\xi).$

Here are the steps:

0) If the problem was n-Dimensional, you could write it as the 1D problem multiplied together n times. So dealing only with the 1D problem,

1) Complete the square in the exponential.

2) Use a contour integral to simplify the problem.

The details, $(-4t^2 + i\xi t) = -4(t - \frac{i\xi}{8})^2 - \frac{\xi^2}{16}.$

Thus the problem becomes, ${\hat f} = e^{- \frac{\xi^2}{16}}\int_R e^{-4(t - \frac{i\xi}{8})^2} \; dt.$ Use the rectangle contour where the long sides lay on the real axis and on $z = x + i\frac{\xi}{8},$ and the short sides are $z = \pm R + iy$ as $R \to \infty$ and $y$ ranges between $[0, \frac{\xi}{8}]$ to connect the contour. The function is analytic in this region so the full contour integral is zero, the short sides vanish in the limit and one finds that, \begin{equation} \int_R e^{-4(t - \frac{i\xi}{8})^2} \; dt = \int_R e^{-4t^2}\; dt = \sqrt{\frac{\pi}{4}}. \end{equation} Thus, ${\hat f}(\xi) = \sqrt{\frac{\pi}{4}} e^{- \frac{\xi^2}{16}},$ when the Fourier transform is defined as $ \int_{-\infty}^\infty e^{i\xi t} e^{-4t^2} \; dt.$ The details such as generalizing it to higher dimension and doing the Gaussian integral at the end are not difficult to fill in yourself.

Merkh
  • 3,700
  • 1
    do you know the differential equation derivation ? and with complex analysis you can also say that with $F(a) = \int_{\infty}^\infty e^{-t^2} e^{a t} dt$, when $a$ is real, you can complete the square, obtaining an entire function of $a$, hence by analytic continuation the equality stays true when $a$ is complex. – reuns May 10 '16 at 11:56
  • I hadn't seen the differential equation derivation before, that's a neat way to do things, thanks! This way remains my standard way of computing such integrals, I like it because its a very clear presentation and its generalizable to arbitrary dimension without thinking too hard. – Merkh May 10 '16 at 13:51