0
$\begingroup$

I am attempting to prove the Lax-Milgram lemma for the weak formulation of the finite element method. However I first need to prove continuity of the bilinear and linear forms ($a(u,v)$ and $l(v)$), plus coercivity of the bilinear form. I am struggling to understand how they connect to the norm. I know the $u$ and $v$ lie in the hilbert space $H^1(\Omega)$ with the $L^2$ norm.

I have that $a(u,v)=\int_{\Omega}\bigtriangledown u\cdot\bigtriangledown v$

so does that make the norm associated with it $||\bigtriangledown u\cdot\bigtriangledown v||$?

$\endgroup$
0
$\begingroup$

If you wish to use Lax-Milgram lemma, you must show continuity of the bilinear form $a$ in the space which is relevant to the posed problem. Suppose that you are solving $$- \Delta u = f$$ in domain $\Omega$ with the boundary condition $u|_{\partial \Omega}=0$. Then the weak formulation reads: find $u \in H^1_0(\Omega)$ such that $$(\nabla u, \nabla v) = (f,v)$$ holds for every $v \in H^1_0(\Omega)$. Now $H^1_0(\Omega)$ is a Hilbert space with the norm $$\|w\|_{H^1_0(\Omega)}:=\|\nabla w\|_{L^2(\Omega)}$$ and, therefore, the continuity follows immediately from the Cauchy-Schwarz inequality: $$(\nabla u, \nabla v) \leq \|\nabla u\|_{L^2(\Omega)} \|\nabla v\|_{L^2(\Omega)}.$$ Moreover, coercivity is obvious: $$(\nabla u, \nabla u) = \|\nabla u\|_{L^2(\Omega)}^2.$$

Another example. Suppose that you are solving a pure Neumann problem: $$- \Delta u = f$$ in domain $\Omega$ with the boundary condition $(\nabla u \cdot n)|_{\partial \Omega}=0$. Then the weak formulation is: find $u \in H^1(\Omega)$ such that $$(\nabla u, \nabla v) = (f,v)$$ holds for every $v \in H^1(\Omega)$. Now the respective norm is $$\|w\|_{H^1(\Omega)}^2 := \|\nabla w\|_{L^2(\Omega)}^2 +\|w\|_{L^2(\Omega)}^2$$ Continuity can be proven somewhat similarly as before but now you cannot show the coercivity: $$(\nabla u, \nabla u) = \|\nabla u\|_{L^2(\Omega)}^2 \geq~???$$ Thus, the solution to this problem is not unique. In order to make the solution unique, you can for example include the constraint $$\int_\Omega u \,\mathrm{d}x = 0.$$ With this constraint you can use the Poincaré-Friedrich's inequality to get $$(\nabla u, \nabla u) = \frac 12 \|\nabla u\|_{L^2(\Omega)}^2+\frac 12 \|\nabla u\|_{L^2(\Omega)}^2 \geq \frac 12 \|\nabla u\|_{L^2(\Omega)}^2 + \frac{1}{2C}\|u\|_{L^2(\Omega)}^2 \geq \frac{1}{2 \max(1,C)} \|u\|_{H^1(\Omega)}^2. $$

In conclusion, the correct norm depends on the space where the weak problem is posed in. You cannot deduce the correct norm by just looking at the bilinear form; you also have to know the boundary conditions.

$\endgroup$
  • $\begingroup$ square missing after "Moreover, coercivity is obvious" $\endgroup$ – user66081 Feb 23 '15 at 13:59

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.