I am studying calculus of variation, and I need to prove that

$I[w] = \int_U \frac{1}{2} |Dw|^2 - fw \, dx$ with $f \in L^2(U)$

is weakly lower semicontinuous on $H_0^1(U)$.

In classes, I only learned that

Assume L is bounded below, and in addition $L(p,z,x)$ is convex in $p$, for each $z \in \mathbb{R}$, $x \in U$. Then $I[.]$ is weakly lower semicontinuous on $W^{1,q}(U)$, where $I[w] = \int_U L(Dw, w, x)$.

And I think my $L$ isn't bounded below.

I don't know how I can solve the problem.

  • $\begingroup$ Actually, it is bounded below. Use the Poincare inequality and Holder's inequality. And I'm pretty sure this was asked before. $\endgroup$ – user147263 Jun 5 '15 at 21:08
  • $\begingroup$ Thank you! And I'm sorry, I searched and I didn't find. But this argument is to prove that $L$ is bounded below or $I$? $\endgroup$ – Daniela Jun 5 '15 at 21:45
  • $\begingroup$ I see. But for convex $I$, a simpler argument exists. I'll post as an answer. $\endgroup$ – user147263 Jun 5 '15 at 22:11

It can be proved from basic principles that $$\text{(norm lower semicontinuity)} + \text{(convexity)} \implies \text{(weak lower semicontinuity)}$$ Indeed,

  1. For any topological space $X$ the lower semicontinuity of a function $f:X\to\mathbb{R}$ is equivalent to the epigraph $\{(x,z):z\ge f(x)\}\subset X\times \mathbb{R}$ being closed in the product topology.

  2. In a normed space, a convex closed subset is weakly closed.

The above applies to $I$, which is convex and continuous with respect to the norm on $H_0^1$. Indeed, $$ |I[v]-I[w]|\le \frac12\int_U (|Dv+Dw|\,|Dv-Dw| + |f||v-w|) \\ \le \frac12(\|v\|_{H^1}+\|w\|_{H^1}) \|v-w\|_{H^1} + \frac12 \|f\|_{L^2}\|v-w\|_{L^2} $$

For completeness, I also quote the comment by Chee Han which shows the boundedness of $I$ from below (this does not help with lower semicontinuity, but is needed for the existence of minimizer).

Poincaré's inequality tells you that the $L^2$ norm of the gradient is equivalent to the full $H^1$-norm, so you can bound the first term below by the $H^1$ norm. For the second term, apply Cauchy-Schwarz, then use Young's inequality (some called this inequality by other name but my lecturer always calls it this way) $2ab\le a^2/2\varepsilon + \varepsilon b^2/2$. Then choose $\varepsilon$ so that the constant associated to the $H^1$ norm of $w$ is positive.

  • $\begingroup$ I still don't understand why $I$ is convex :\ $\endgroup$ – Daniela Jun 6 '15 at 7:42
  • $\begingroup$ It's the sum of two things: one is a linear functional of $w$, the other is the integral of a convex function of $w$. $\endgroup$ – user147263 Jun 8 '15 at 3:01

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.