Let $\Omega=\mathbb{R}^n\times \mathbb{R}^+$ and $a_0,\ldots, a_n\in \mathbb C$, I must find a fundamental solution of the PDE $$(*)\qquad\qquad\frac{\partial u}{\partial t}(x, t)+\sum_{k=1}^n a_k\frac{\partial u}{\partial x_k}(x, t)+a_0 u(x, t)=f(x, t),\qquad (x, t)\in \Omega.$$ If we call $$L=\frac{\partial }{\partial t} +\sum_{k=1}^n a_k\frac{\partial }{\partial x_k} +a_0,$$ then it is enough to find a fundamental solution for $L$, say $T$ (i.e., $LT=\delta$, and afterwards we just convolute with $f$)

How should I proceed here? I know that every non-zero linear differential operator with constant coefficients has a fundamental solution (and one can even show an explicit formula by means of one of the constructive proofs of the Malgrange-Ehrenpreis theorem), but this is rather long and cheap.

Is there an easier(direct) way to find such solution? Any help or reference is highly appreciated

  • $\begingroup$ How about using Fourier transform (in $x$) of $LT$? that gives you that $\hat{T}$ is the solution of a linear ODE of order 1 in $t$. Then you take the inverse Fourier transform of $\hat{T}$ $\endgroup$
    – Tryss
    Nov 21, 2017 at 1:14
  • $\begingroup$ I've already did it, as you say, one must solve a first order equation like $\frac{\partial \hat{T}}{\partial t}(\xi, t)=\frac{1}{(2\pi)^{n/2}}+\alpha(\xi)\hat{T}(\xi, t)$. It can be solved after one tricks or two, and then one can compute its inverse transform (I haven't done this yet), but I don't see a clean way to prove that such $T$ is indeed a fundamental solution (by clean, I mean without actually doing a lot of calculations again) $\endgroup$ Nov 21, 2017 at 1:23
  • $\begingroup$ It isn't enough to solve for T via a fourier transform and show that indeed, it does satisfy $LT=\delta$? I guess the problem I'm having is that our problem is of first order, I don't think I've ever encountered fundamental solutions for first order PDE. $\endgroup$
    – DaveNine
    Nov 21, 2017 at 6:26

1 Answer 1


For the computation of the fourier transform method, some progress:

$LT=\delta$ as you've written, and then we can find $T$ via Fourier transform. I'm going to rename your equation slightly,

$u_t + \textbf{a}\cdot \nabla_x u(x,t) + bu(x,t) = f(x,t)$

I take the convention that(as in Evan's PDE book)

$$\mathcal{F}[u](\xi)=\frac{1}{(2 \pi)^{\frac{n}{2}}}\int_{\mathbb{R}^n}u(x)e^{-ix \cdot \xi}\,dx$$

$$\mathcal{F}^{-1}[u](x)=\frac{1}{(2 \pi)^{\frac{n}{2}}}\int_{\mathbb{R}^n}u(\xi)e^{i\xi \cdot x}\,d\xi$$


$\mathcal{F}[\partial_t u]=\hat{u}_t$

$\mathcal{F}[a \cdot \nabla u] = i(a \cdot \xi) \hat{u}$

$\mathcal{F}[\delta(x)] = 1$

Hence, we arrive to the equation

$$\hat{T}_t + (i(a \cdot \xi)+b)\hat{T} = 1$$

so the Fourier transform of the fundamental solution satisfies

$$\hat{T}(\xi,t) = C e^{-t(i(a \cdot \xi)+b)} + (i(a \cdot \xi)+b)^{-1}$$

where $C$ is yet to be found. However from here, I've been having some trouble writing out the inverse transform of each of those functions.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .