3
$\begingroup$

We consider the problem $$-u''+a(x)u=f, x\in (0,1)=\Omega , u(0)=\alpha,u(1)=\beta$$ with $f \in L^2(\Omega) , a(x) \geq a_0 > 0, a \in L^{\infty}(\Omega)$

Question 1: prove that the variational formulation associate to this problem admits a unique solution in an adequate Hilbert space $V$ (we can use for this proof the function $u_1(x)=\alpha +(\beta - \alpha) x$).

Question 2: prove that $$||u||_V \leq C ||f||_{L^2}$$

For the question 1, I consider $u=w + u_1$ with $w \in V = H^1(\Omega)$ and solution for the equation of the problem, and I prove that $w$ is unique in $H^1_0(\Omega)$ by Lax-Milgram, so the problem admits a unique solution $u$ in $H^1_0.$

My problem is in question 2, I don't know how we prove it. Thank's for the help.

$\endgroup$
4
  • $\begingroup$ Your second question is strange. If $u$ is a solution that belongs to $V$, then it is immediate that $\|u\|_V$ is finite. $\endgroup$ – Tomás Aug 10 '13 at 20:24
  • $\begingroup$ this is a mistake, sorry. We have to prouve that $||u||_V \leq C ||f||_{L^2}$$ $\endgroup$ – jijii Aug 10 '13 at 21:34
  • $\begingroup$ Where did you find this problem? $\endgroup$ – Tomás Aug 10 '13 at 23:13
  • $\begingroup$ this is an old exam, and i haven't a solution $\endgroup$ – jijii Aug 11 '13 at 15:53
1
$\begingroup$

I'll treat the case $\alpha = \beta = 0$ (so that I do not need to jump throught the $u_1$ hoops); from the bound required in part 2, I deduce that we assume $f \in L^2(\Omega)$.

  1. If you test your original equation with functions from $C^\infty_0(\Omega)$, you arrive at $$ \int_0^1 [ -u'' + au ]v = \int_0^1 fv$$ Integration by parts turns this into $$ u'(1)v(1) -u'(0)v(0)+ \int_0^1 [u'v' + au v] = \int_0^1 fv$$ where the first two terms cancel since we have $v(0) = v(1) = 0$. The final variational problem reads $$ \int_0^1 [u'v' + au v] = \int_0^1 fv \quad \forall v \in C^\infty_0(\Omega)$$ or more concisely $$ a(u,v) = \ell(v) $$ The right space to pose this problem over is $H^1_0(\Omega)$, which is a Hilbert space w.r.t. the scalar product $(v,w) = \int_0^1 uv + \int_0^1 u'v'$.

    The Lax-Milgram theorem can now be used to prove that the variational problem has exactly one solution in $H^1_0$. To that end, observe that $\ell$ is continuous; this is true since we assumed $f \in L^2(\Omega)$. We also need the bilinear form to be coercive and continuous with respect to the aforementioned scalar product: For ellipticity, we can use the fact that $a \ge a_0 > 0$, so that clearly (without loss of generality, $a_0 \le 1$), we have $a(v,v) \ge a_0 (v,v)$. We also have $a(v,w) \le \|a\|_\infty (v,w)$ and thus continuity.

    Note that because of Poincare's inequality, we could handle the case $a_0 = 0$ as well.

  2. Since $a(u,v) = \ell(v)$ holds for all $v \in C^\infty_0$ functions, it also holds for all $H^1_0$ functions by density. In particular, we can choose $v = u$ and obtain $$ a_0 \|u\|_{H^1}^2 = a_0 (u,u) \le a(u,u) = \ell(u) \le \|f\|_{L^2}\|u\|_{L^2} \le \|f\|_{L^2}\|u\|_{H^1} $$ where our earlier thoughts on the bilinear form and the Cauchy-Schwarz inequality were used. Dividing by $\|u\|_{H^1}$ (if $u = 0$, the claim would be trivial) yields $$ \|u\|_{H^1} \le \frac 1{a_0} \|f\|_{L^2} $$ as desired.

$\endgroup$

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.