I've tried to calculate this as well as been looking around online and I can't seem to find the calculation of this. If $\mathfrak{F}$ is the Fourier transform: $$ f\mapsto \int_{-\infty}^{\infty} f(x)e^{ix}dx, $$ then what would the functional derivative $$ \frac{\delta \mathfrak{F}}{\delta f} $$ be?

  • $\begingroup$ The Fourier transform is a linear operator so the derivative is? $\endgroup$ – Bananach Jul 3 '17 at 6:04
  • $\begingroup$ It is itself, if I'm correct? $\endgroup$ – AIM_BLB Jul 3 '17 at 6:07
  • $\begingroup$ Yoh are correct $\endgroup$ – Bananach Jul 3 '17 at 6:35

The functional derivative is defined as the linear mapping $$\left\langle \frac{\delta}{\delta f(x)} F[f](\xi), \phi(x) \right\rangle := \left. \frac{d}{d\lambda} F[f+\lambda\phi](\xi) \right|_{\lambda=0}$$

For $\mathfrak F[f](\xi) := \int_{-\infty}^{\infty} f(x)e^{ix\xi} \, dx$ we get $$\begin{align} \left. \frac{d}{d\lambda} \int_{-\infty}^{\infty} (f+\lambda\phi)(x)e^{ix\xi} \, dx \right|_{\lambda=0} & = \left. \frac{d}{d\lambda} \int_{-\infty}^{\infty} \left(f(x)+\lambda\phi(x)\right)e^{ix\xi} \, dx \right|_{\lambda=0} \\ & = \left. \frac{d}{d\lambda} \left( \int_{-\infty}^{\infty} f(x) e^{ix\xi} \, dx + \lambda \int_{-\infty}^{\infty} \phi(x) e^{ix\xi} \, dx \right) \right|_{\lambda=0} \\ & = \left. \int_{-\infty}^{\infty} \phi(x) e^{ix\xi} \, dx \right|_{\lambda=0} \\ & = \int_{-\infty}^{\infty} \phi(x) e^{ix\xi} \, dx \\ & = \left\langle e^{ix\xi}, \phi(x) \right\rangle \end{align}$$ Thus, $$\frac{\delta}{\delta f(x)} \mathfrak F[f](\xi) = e^{ix\xi}$$

| cite | improve this answer | |
  • $\begingroup$ This answer uses a wrong definition of the Fourier transform. It is an operator $L^2(\mathbb{R})\to L^2(\mathbb{R})$, not a functional $L^2(\mathbb{R})\to\mathbb{R}$ $\endgroup$ – Bananach Jul 4 '17 at 13:12
  • $\begingroup$ @Bananach. Look at the question! $\endgroup$ – md2perpe Jul 4 '17 at 14:04
  • $\begingroup$ I've edited my answer to make it a Fourier transform and not just the value at $\xi=1$. I haven't taken $\mathfrak F$ to act on distributions though. $\endgroup$ – md2perpe Jul 4 '17 at 14:33
  • $\begingroup$ It is still incorrect as you cannot consider pointwise derivatives. But this is just a matter of rewriting and presenting your calculations in a more formal manner $\endgroup$ – Bananach Jul 4 '17 at 17:24
  • $\begingroup$ I know that it is formally incorrect, but I use this notation in a similar way as one sometimes writes $u(x)$ for a distribution; it's not meant as the value at $x$ but rather to denote what variable belongs to $u$. An example of such notation: $\langle u * v, \phi \rangle = \langle u(x) \otimes v(y), \phi(x+y) \rangle$. $\endgroup$ – md2perpe Jul 4 '17 at 17:44

Are you looking for derivative of Fourier transform ?

considering $\omega$ insted of $f$ we can write accroding to the definition

$$\begin{align}\frac{d \mathcal{F}(\omega)}{d\omega}&=\int_{-\infty}^{+\infty}f(t)\frac{d}{d\omega}\cdot e^{i\omega t}\,dt\\&=-i\int_{-\infty}^{+\infty}t.f(t)e^{i\omega t}\,dt\\&=-i \cdot\mathcal{IFT}\{t \,.f(t)\}\end{align}$$

And the reverse process can be

$$\mathcal{FT}\left\{\frac{d f^n(t)}{d^nt}\right\}=(i\omega)^n.F(\omega)$$

Please let me know if somehow i can improve my answer , Thanks !

| cite | improve this answer | |
  • $\begingroup$ Quick question, what are $\mathfrak{I},msthfrak{T}$ symbolise? $\endgroup$ – AIM_BLB Jul 3 '17 at 6:06
  • $\begingroup$ @CSA sorry i'm not very much adept with latex formatting i'm just learning those , the first one seems to be for imaginary part , i don't know about others $\endgroup$ – Siddhartha Jul 3 '17 at 6:17
  • $\begingroup$ He was asking for the functional derivative of the Fourier transform, not the derivative of the Fourier transform. en.wikipedia.org/wiki/Functional_derivative $\endgroup$ – md2perpe Jul 3 '17 at 11:19
  • $\begingroup$ sorry for wrong answer , i'm new to integral transforms and couldn't interpret his query correctly , sorry for that @md2perpe $\endgroup$ – Siddhartha Jul 3 '17 at 12:55
  • 1
    $\begingroup$ @Lelouch.D.Light. No problem. It's easy to misunderstand a question. $\endgroup$ – md2perpe Jul 3 '17 at 15:42

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.