0

With the input

f[t_] := Cos[2 Pi t]
FourierTransform[f[t], t, w, FourierParameters -> {0, -2*Pi}]

I get 1/2 DiracDelta[-1 + w] + 1/2 DiracDelta[1 + w] as expected. But unless I'm having a brain fart, this implies that

Integrate[f[t]*Exp[-2 Pi I t 3], {t, -Infinity, Infinity}]

is zero. Mathematica, however, claims that this integral doesn't converge. I could understand if FourierTransform used other methods internally and thus Integrate doesn't find a solution. But I'm a bit surprised that I get the apodictic claim that the integral doesn't converge. What am I missing? Are there any options to Integrate or any assumptions I could use to make Mathematica change its mind?

BTW, I am aware of this, but although it's related it doesn't answer my question.

Frunobulax
  • 215
  • 1
  • 3
  • 4
    What's strange here? The integral diverge? period. Fourier transform command deals with distributions and action of the FT on those are not integrating generally. It's a more widely defined mapping. – Andrew Dec 28 '18 at 17:49
  • 3
    Stated differently, not every FT can be recast as a convergent integral. – Daniel Lichtblau Dec 28 '18 at 21:23

1 Answers1

3

The reason is that Integrate will not produce generalized functions like DiracDelta as results. I don't believe there is a strategy that will produce consistent results in general for such (formally divergent) integrals. FourierTransform is restricted to Fourier integrals, a relatively safe domain.

John Doty
  • 13,712
  • 1
  • 22
  • 42