The FT of a function of time $f(t)$ transforms the function from the time domain to the frequency domain, such that: $$\mathscr{F}[f(t)]=F(\omega)=\int ^{\infty}_{-\infty}f(t)e^{-i\omega t}dt$$ And the IFT of a function $F(\omega)$ transforms it from the frequency domain back to the time domain, such that: $$\mathscr{F}^{-1}[F(\omega)]=f(t)=\frac{1}{2\pi}\int ^{\infty}_{-\infty}F(\omega)e^{i\omega t}d\omega$$ So, logically, applying the IFT to the FT should give the original function, such that: $$\mathscr{F}^{-1}[\mathscr{F}[f(t)]]=f(t)$$ But upon trying to solve this, it goes as: $$\mathscr{F}^{-1}[\mathscr{F}[f(t)]]=\frac{1}{2\pi}\int ^{\infty}_{-\infty}\int ^{\infty}_{-\infty}f(t)e^{-i\omega t}e^{i\omega t}dtd\omega=\frac{1}{2\pi}\int ^{\infty}_{-\infty}\int ^{\infty}_{-\infty}f(t)dtd\omega$$ which diverges. Whys is this the case?

  • $\begingroup$ $\mathcal{F}^{-1}(\mathcal{F}(f(t)) = \mathcal{F}^{-1}(\mathcal{F}(\omega)) = f(t)$. $\endgroup$ – Maxtron Aug 11 '18 at 5:00
  • 3
    $\begingroup$ I think you are getting confused by reusing your $t$'s instead of introducing a dummy variable when writing down your expression for the function composition in the inner integral. You can't bring in your "$e^{i\omega t}$" inside the integral here. $\endgroup$ – JessicaK Aug 11 '18 at 5:02

Following the discussion in the comments, one should rather write \begin{aligned} \mathscr{F}^{-1}\mathscr{F} f(t) &= \frac{1}{2\pi} \int \left(\int f (\tau) e^{-\text i \omega \tau}\,\text d\tau\right) e^{\text i \omega t}\,\text d\omega \, , \\ &= \frac{1}{2\pi} \iint f (\tau) e^{\text i \omega (t-\tau)}\,\text d\tau\,\text d\omega \, . \end{aligned} Using the change of variable $\theta = t-\tau$, Fubini's theorem, and properties if the Dirac distribution, \begin{aligned} \mathscr{F}^{-1}\mathscr{F} f(t) &= \frac{1}{2\pi} \iint f (t-\theta) e^{\text i \omega \theta}\,\text d\theta\,\text d\omega \, , \\ &= \int f (t-\theta) \left(\frac{1}{2\pi}\int e^{\text i \omega \theta}\,\text d\omega \right)\text d\theta\, , \\ &= \int f (t-\theta) \delta(\theta)\,\text d\theta\, , \\ &= f (t)\, . \end{aligned} This is a shortened "proof" of the theorem, assuming that the inverse Fourier transform of the constant 1 is the Dirac delta (which is a bit unfair...). The full proof of the Fourier inversion theorem holds for absolutely integrable continuous functions, with absolutely integrable Fourier transforms.


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.