I am attempting to solve Newton's Law of Cooling differential equation with Fourier Transforms for a high school math report. Can Fourier Transforms be used to solve first-order ODEs? The equation is: $\frac{dT}{dt} = -kT + kT_a$, where $k$ and $T_a$ are constants. Since the Fourier Transform is a linear operator, I took the transform of each of the three terms: $$F\{T'\} = F\{-kT\} + F\{kT_a\}$$ $$F\{T'\} = -kF\{T\} + kT_aF\{1\}$$ The Fourier transform of 1 should be $\delta(s)$, so: $$F\{T'\} = -kF\{T\} + kT_a\delta(s)$$ The transform of the derivative of a function, using $s$ instead of $\omega$ in the formulas, is $F\{T'\} = 2\pi isF\{T\}$. Substituting this into the original equation, $$2\pi isF\{T\} = -kF\{T\} + kT_a\delta(s)$$ Adding the terms with the transform should yield $$2\pi isF\{T\} +kF\{T\} = kT_a\delta(s)$$ $$F\{T\}(2\pi is +k) = kT_a\delta(s)$$ $$F\{T\} = \frac{kT_a\delta(s)}{2\pi is + k}$$. The formula for the Inverse Fourier Transform is $f(t) = \int_{-\infty}^{\infty} e^{2\pi ist} F(s) ds$. Thus $$T = \int_{-\infty}^{\infty} \frac{e^{2\pi ist} \delta(s) kT_a}{2\pi is + k}ds$$, which I evaluated to $T_a$, dropping the integral and plugging in 0 for $s$ due to the $\delta(s)$ function...however, the actual solution to the differential equation is $T_a + Ae^{-kt}$, where A is a constant. Where did I go wrong?

  • $\begingroup$ The reason is that $F[T-T_a] = F[e^{-kt}] = \int e^{2\pi i s - kt} ds$ does not exist (it is infinite) so your equation $2\pi i sF[T] = -kF[T] + kT_a\delta(s)$ reads $\infty = \infty$. $\endgroup$ – Winther Nov 23 '14 at 0:38

Fourier Transform can certainly be used to solve linear differential equations. Note that if you apply Fourier Transform of its ordinary form with respect to the variable $t$, you implicitly assume that $- \infty < t < + \infty$, i.e. you have no boundary. This is the reason why you cannot include boundary or initial conditions. You can get $T_a + A e^{-kt}$ if you apply Laplace Transform though.

  • $\begingroup$ Thanks for the clarification. I've already solved it with Laplace Transform, but I was hoping there would be a way to apply Fourier Transform as well to add to the diversity of the paper $\endgroup$ – kalkuel343 Nov 23 '14 at 5:03
  • $\begingroup$ If you were to apply different methods, you can derive the solution using the Green's function method for instance. $\endgroup$ – bkosztin Nov 23 '14 at 14:57

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.