I am trying to prove the 3D Diffusion Equation $$\begin{cases}u_t(\vec x,t) &= c\nabla u\\ u(\vec x,0)&=g(\vec x)\end{cases}$$ From the 3D Fourier Transform, where $\vec k,\vec x\in\mathbb R^3$.

I took the Fourier Transform of both sides of the 3D Diffusion to get:

$$\widehat u_t(k,t)+c|\vec k|^2\widehat u(k,t) = 0$$

Now, this gives that the solution to the ODE is: $$\widehat u(k,t) = \widehat g(k) e^{-ct|\vec k|^2} = \widehat g(\vec k)\cdot\widehat f(\vec k)$$

Now, to find the solution, I want to find the Inverse FT of $f$, whcih leads to: $$\frac{1}{(2\pi)^3}\iiint_{\mathbb R^3}e^{-ct|\vec k|^2+ikx} d\vec k$$ So isn't $|\vec k|^2$ a scalar? If not, how would I take the integral of the integrand?


$|k|^2$ is a scalar, but $\vec k$ is a vector. Note that also, in your last equation, $x$ should be $\vec x$, so the last expression is $$\frac{1}{2\pi}\iiint_{\mathbb R^3}e^{-ct|\vec k|^2+i\vec k\vec x} d\vec k$$ You can either solve the integral in Cartesian coordinates, or change it to spherical coordinates. $|\vec k|^2$ will transform into $k^2$, $\vec k \vec x$ is $kx\cos\theta$, and $d\vec k$ is $k^2\sin\theta dk d\theta d\phi$. The integration limits for $\phi$ is from $0$ to $2\pi$, $theta$ goes from $0$ to $\pi$, and $k$ is from $0$ to $\infty$

  • $\begingroup$ Wouldn't it be a lot harder in spherical...? And so wouldn't I just have to integrate with respect to the vector $k$? Or do I have to consider the scalar $|\vec k|$ there as well? $\endgroup$ Nov 6 '17 at 2:37
  • $\begingroup$ The scalar $k$ is the "radius" in spherical coordinates. In Cartesian coordinates $|\vec k|^2=k_x^2+k_y^2+k_z^2$, and $\vec k\vec x=k_x x$ if you choose the appropriate coordinate system. And $d\vec k=dk_x dk_y dk_z$ $\endgroup$
    – Andrei
    Nov 6 '17 at 2:41
  • $\begingroup$ Which one is the easier way to integrate this? Both of them seem like a lot of work... The Cartesian you would have you integrate over $k_1, k_2,k_3$ given $\vec k=<k_1,k_2,k_3>$. But Spherical you would have to do crazy integration by parts to solve this, both of them seem very ridiculous. Is there a trick I am missing here? $\endgroup$ Nov 6 '17 at 2:51
  • $\begingroup$ Both are relatively easy. See for example math.stackexchange.com/questions/307019/3d-fourier-transform $\endgroup$
    – Andrei
    Nov 6 '17 at 2:53
  • $\begingroup$ I would also reccomend my question from a yearish ago, it's a more general version of this problem: math.stackexchange.com/questions/1281047/… $\endgroup$
    – DaveNine
    Nov 6 '17 at 4:51

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.