Coercivity of $B[u,v] = \int_\Omega a^{ij}\partial u_{x_i} \partial u_{x_j} + \int_\Omega c(x)uv $, with no sign assumption on $c(x)$

111 Views Asked by At

I am reviewing previous qualifying exams on PDE's and here is a problem that I am not confident in my solution.

Assume that $\Omega \subset \mathbb{R}^n$ is open, bounded, connected, and $\partial\Omega \in C^1$. Assume that $a^{ij} \in L^\infty(\Omega)$ is a uniformly elliptic matrix in the sense that there exists two constants $\lambda>0$ and $\Lambda>0$ so that $$ \forall \xi \in \mathbb{R}^n, \,\,\, \lambda|\xi|^2 \leq a^{ij}(x)\xi_i\xi_j \leq \Lambda|\xi|^2 $$ and that $c\in L^\infty(\Omega)$ is given. The bilinear form associated to $a^{ij}$ and $c$ is: $$ B[u,v] = \int_\Omega a^{ij}\frac{\partial u}{\partial x_i}\frac{\partial u}{\partial x_j} \,dx + \int_\Omega c(x)uv \,dx $$ Assume no assumption on the sign of $c(x)$. Prove there exists a choice of $\lambda$ depending on $\Omega, n$, and $\|c\|_{L^\infty}$, so the that the bilinear form, $B$ is coercive on $H^1_0(\Omega)$.

Here is my thinking and work so far. Using Poincare inequality, we find that

$$ \|u\|^2_{H^1_0(\Omega)} = \int_\Omega u^2 \,dx+\int_\Omega|\nabla u|^2\,dx \leq (C_p+1)\int_\Omega |\nabla u|^2\,dx. $$ Manipulating the bilinear form, $$ B[u,u] - \int_\Omega c(x) u^2\,dx = \int_\Omega a^{ij}\frac{\partial u}{\partial x_i}\frac{\partial u}{\partial x_j} \,dx $$ and using the estimate $$ \int_\Omega c(x) u^2 \,dx \leq \|c\|_{L^\infty}\int_\Omega u^2 \, dx $$ we get $$ B[u,u] +\|c\|_{L^\infty}\int_\Omega u^2 \, dx \geq B[u,u] -\int_\Omega c(x) u^2 \,dx = \int_\Omega a^{ij}\frac{\partial u}{\partial x_i}\frac{\partial u}{\partial x_j} \,dx \geq \lambda\int_\Omega |\nabla u|^2\,dx. $$ Combining this with the Poincare inequality $$ B[u,u] + \|c\|_{L^\infty} \int_\Omega u^2 \, dx \geq \frac{\lambda}{C_p +1}\left( \int_\Omega u^2 \,dx+\int_\Omega|\nabla u|^2\,dx \right). $$ then $$ B[u,u] \geq \frac{\lambda}{C_p +1}\int_\Omega|\nabla u|^2\,dx + \left( \frac{\lambda}{C_p +1}-\|c\|_{L^\infty} \right) \int_\Omega u^2 \, dx. $$

From here we could could find a bound on $\lambda$ so the coefficient on the $\int u^2 $ term is some fraction. I personally have difficulty with these types of inequalities. Is my process and logic sound?

1

There are 1 best solutions below

0
On BEST ANSWER

This solution is correct, but others are possible depending on how gross of an estimate is acceptable or if the estimate should take a particular form for a given application.