1
$\begingroup$

[Cross posting https://math.stackexchange.com/questions/1374384/fundamental-solution-for-a-parabolic-pde-with-costant-coefficents ]

I don't know if this question is more appropriate in Mathematics and not here, in this case I will delete it.

As it is well known, the fundamental solution of the heat equation is the function

$G(t,x)=\frac{1}{(4\pi t)^{n/2}}e^{-\frac{|x|^2}{4t}}$,

for all $t>0,x\in\mathbb{R}^n$.

I wonder if exists (and if you have same references) a similar explicit formula for the fundamental solution for a parabolic PDE with constant coefficents. It is possible that it can be found in "Linear and quasilinear equation of parabolic type" by Ladyzenskaja, Solonnikov and Ural'ceva, but as I cant' consulte the book by now, I don't know.

$\endgroup$
2
  • $\begingroup$ Could you write out the precise form of the PDE you want to solve? Is it $u_t = a^{ij}\partial_i\partial_ju$? $\endgroup$
    – Deane Yang
    Jul 26, 2015 at 14:57
  • $\begingroup$ Yes, it is like you wrote. first I'm interested in second order equation, then in higher order $\endgroup$
    – foo90
    Jul 26, 2015 at 21:47

3 Answers 3

6
$\begingroup$

Yes, it can be written out explicitly. Namely, for the operator $$ \partial_t-\sum_{i,j=1}^na_{ij}\partial_{ij}-\sum_{i=1}^nb_i\partial_i-c, $$ where $A=(a_{ij})>0$, $b=(b_1,\ldots,b_n)$, $$ G(t,x)=\frac{1}{(4\pi t)^{n/2}|A|^{1/2}}\exp\left\{-\frac{(A^{-1}(x-bt),x-bt)}{4t}+ct\right\}. $$ It can be obtained either via the Fourier transform wrt $x$ as in the case of the heat equation or making an affine change of variables in $\mathbb R^{n}$, reducing again to the heat equation.

$\endgroup$
4
  • $\begingroup$ And for an higher order parabolic equation there exists a similar formula? $\endgroup$
    – foo90
    Jul 26, 2015 at 21:54
  • $\begingroup$ Thank you very much. Do you have any reference? $\endgroup$
    – foo90
    Jul 27, 2015 at 9:47
  • 1
    $\begingroup$ @foo90 Not off top of my head. $\endgroup$
    – Andrew
    Jul 27, 2015 at 14:16
  • $\begingroup$ @foo90 For the equation $u_t+(-1)^m\Delta^{m}u=0$ the fundamental solution is the inverse Fourier transform of $e^{-|y|^{2m}t}$. For $m>1$ it can be expressed through hypergeometric functions. Say, for $n=1$ and $m=2$ Mathematica gives $$ \frac1{\sqrt{2\pi}}\left(\frac{2 \Gamma \left(\frac{5}{4}\right) \, _0F_2\left(;\frac{1}{2},\frac{3}{4};\frac{x ^4}{256 t}\right)}{\sqrt[4]{t}}-\frac{\Gamma \left(\frac{3}{4}\right) x ^2 \, _0F_2\left(;\frac{5}{4},\frac{3}{2};\frac{x ^4}{256 t}\right)}{4 t^{3/4}}\right). $$ $\endgroup$
    – Andrew
    Jul 28, 2015 at 6:33
6
$\begingroup$

For an operator of the form $L=\partial_t-\sum A_{jk}\partial_j\partial_k$, the fundamental solution is computed in Section 3.3 of Volume I of Hormander's treatise (The Analysis of Linear PDOs). I think one may try to extend the formula to include lower order terms by playing with it a little; for instance if $u$ solves $Lu=0$, then $v = e^{-at}u$ solves $Lv+av=0$, and similarly $v=(a_1x_1+\dots+a_nx_n)u$ solves an equation with additional first order terms.

$\endgroup$
4
  • $\begingroup$ Thank you. And for higher orde parabolic pde do you have some reference? $\endgroup$
    – foo90
    Jul 26, 2015 at 21:59
  • 1
    $\begingroup$ Sorry, I have no reference for the higher order case. I guess it should be possible to do something along the lines of Hormander's computations $\endgroup$ Jul 27, 2015 at 20:57
  • $\begingroup$ I have tried your approach. For the zero order term it works, for the first order terms I haven't succeded, can you write something more? $\endgroup$
    – foo90
    Aug 17, 2015 at 16:31
  • 1
    $\begingroup$ You are right, I think that part of my answer is not useful (it leads to first order terms but with a special form). However you have the full formula in Andrew's answer, no? from that answer one might guess that the right transformation to use is $v(x,t)=u(x-bt,t)$ $\endgroup$ Aug 18, 2015 at 21:17
4
$\begingroup$

A good place to start is the series of articles by P. Wagner and N. Ortner. In addition to their constructive proof of the Malgrange-Ehrenpreis theorem on the existence of fundamental solutions for pde's with constant coefficients (AMS 116---MR 2510844), they have, in a series of previous papers, computed explicit solutions for many concrete examples.

$\endgroup$
3
  • $\begingroup$ I could be wrong since I'm relying on distant memories, but I believe that the theorems of Malgrange and Ehrenpreis apply only to PDE's where the top order term is nondegenerate (i.e., PDE's whose real analytics solutions can be obtained via Cauchy-Kovalevski). Without this assumption there might not be any solutions at all. The heat equation has a degenerate top order term, so I'm not sure that the theorems you cite apply. $\endgroup$
    – Deane Yang
    Jul 26, 2015 at 14:57
  • $\begingroup$ The theorem of Malgrange-Ehrenpreis applies to all linear pde's with constant coefficients. I had assumed that the OP was referring to this situaton (implicitly implied by the phrase "with constant coefficients"---I'm not sure what this phrase can mean when applied to non-linear pde's). If not, then this is, of course, a completely new ball game. $\endgroup$
    – priel
    Jul 26, 2015 at 15:13
  • $\begingroup$ No, you're right. I stand corrected. The Malgrange-Ehrenpreis theorem does not make any assumptions about the top order term at all. $\endgroup$
    – Deane Yang
    Jul 26, 2015 at 15:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.