MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Hello everyone. I'm about to use a comparison principle that I belive is true, but I can't find any precise reference to be sure of it. Here is what I would like to say : I have a parabolic equation $u_t - Lu = 0$ in $(t,x,y) \in \mathbb R \times \mathbb R \times \omega$ where $\omega$ is a bounded domain.
Note : L has a $0$-order term with unknown sign, but I would like to compare with $0$ so this is not a problem.

I have a function $U$ smooth enough (that is $\mathcal C^1$ in time, $\mathcal C^2$ in space in the whole closed cylinder $\mathbb R\times \overline\omega$) that verifies :

$U_t - LU \geqslant 0$
$\partial_\nu U = 0$ on $\partial\omega$ (so $\geqslant 0$...)
$U(0,x,y) \geqslant 0$

I would like to conclude $U \geqslant 0$.

But every maximum principle I know "works with Dirichlet conditions" : $U$ must also be non-negative at any time on the boundary of $\mathbb R\times \omega$. In Smoller, Shock Waves and Reaction-Diffusion Equations p.93 I found a principle comparison that is almost exactly what I want (far more general in fact, because it tolerates nonlinearities, and works with mixed conditions), but with $L$ as follows :

$Lu = \sum_{i,j = 1}^n (a_{ij}(x,t)u_{x_i})_{x_j})$

what is somehow restrictive : in my case, the equation is an advection-reaction-diffusion one, of the type $u_t - \Delta u + q(x,y)\cdot\nabla u -k(t,x,y)u = 0$ that is not like the one above.

So my question is : does anyone knows about such a comparison principle ? Does the one in Smoller's book extend to more general operators ?

Thank you very much.

share|cite|improve this question
may I ask if you solved this problem and indeed what field do you work in? – lost1 Dec 3 '13 at 1:00

Some kind of a maximum principle for the Neumann problem, as well as some comparison result for rather general nonlinear equations and even the third (Robin) boundary condition are proved in Chapter 2 of the book by A. Friedman, "Partial Differential Equations of Parabolic Type", Prentice-Hall, 1964.

share|cite|improve this answer
Thank you. I did not really found what I wanted, but by reading one of his proofs, I think I understood a little more what's going on. If I understand correctly, it states that the max cannot be reached at a point where there is Neumann condition, since at this point, the normal derivative should be > 0, so the maximum on the frontier is the maximum on the other part of the frontier. What I have a hard time dealing with now, is the 0-order term. I believe everything falls down if it is arbitrary, (except if the sup is 0 - which is kind of what I want to show !), doesn't it ? – Laurent Mar 20 '12 at 17:21

In Gary M. Lieberman, Second Order Parabolic Partial Differential Equations p.13 I found a comparison principle that gives exactly what I want for a large class of boundary operators, but needs a parabolic operator with non-positive $0$-order term. I'm starting to think that maybe there is an error in the paper I'm reading !

On the other hand, Lieberman says just after, that the theorem still holds with the $0$-order term only bounded from above, but with stronger hypotheses on the parabolic boundary and the boundary operator. Still, mine is not really pretty since it is of type $a\cdot\nabla u + bu$ with vector field $a$ being either $-\vec n$ or $0$ and scalar function being either $-1$ or $0$, depending on the part of the parabolic boundary.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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