Let $u_n \to u$ in $C^0([0,T];H^{-1}(\Omega))$ and suppose $\lVert u_n \rVert_{L^\infty(0,T;L^\infty(\Omega))} \leq C$ for all $n$.

It follows that for almost all $t$, $u_n(t)$ is bounded in $L^\infty(\Omega)$, so we can extract a weak-* convergent subsequence and after a bit of work we can show that $$\text{for almost all $t$} \qquad u_n(t) \rightharpoonup u(t) \text{ in $L^q(\Omega)$} \qquad\text{for all $q < \infty$.}$$

I want to show that this weak convergence holds for all $t$.

From personal correspondence, I have been told that

The key point is that one already has convergence in the larger space. So one already knows that pointwise, i.e. for all $t$, $u(t)$ converges in this larger space. But since it is also bounded in $L^\infty$, $u(t)$ [for almost all $t$ -- riem's note] will also converge weak-* up to a subsequence, and the limit is the same are identified as distributions. The same is true for any $L^q$, $q<\infty$.

Simply put, I don't get why it must hold for all $t$. I think we need to exploit the continuity of $u_n$ and $u$ wrt. $t$ in the larger space $H^{-1}$ but once again $L^q$ is a stronger space than this..

Originally posted on MSE.


The explanation is that by the hypotheses you in fact have $\lVert u_n \rVert_{\ell^{+\infty}([0,T],L^{+\infty}(\Omega))} \leq C$ for all $n$ as for arbitrarily fixed $t\in[0,T]$ you can take a sequence of $t_i\to t$ with $\lVert u_n(t_i) \rVert_{L^{+\infty}(\Omega)} \leq C$ and $u_n(t_i)\to u_n(t)$ in $H^{-1}(\Omega)$ . Then a subsequence converges also weakly$^*$ in $L^{+\infty}(\Omega)$ and hence $u_n(t)$ is also there with norm at most $C$ . So everything you have written "for almost all $t$ " in fact holds for all $t\in[0,T]$ .


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.