In the proof of a compactness theorem involving fractional derivatives in Temam's Navier-Stokes Equations, an argument as the following is made.

Suppose $X_0,X,X_1$ are Hilbert spaces such that $ X_0\overset{\textrm{cpt}}{\hookrightarrow} X\hookrightarrow X_1. $ Suppose $\gamma>0$. Suppose $v_m$ is a sequence all supported in a bounded subset $K$ of $\mathbb{R}$ such that

  • $v_m\to 0$ in $L^2(\mathbb{R};X_0)$ weakly; (2.29)
  • $|\xi|^\gamma\hat{v_m}\to 0$ in $L^2(\mathbb{R};X_1)$ weakly. (2.30)

Then for (fixed large enough) $M>0$, $$ \lim_{m\to\infty}\int_{|\xi|\leq M}\|\hat{v_m}(\xi)\|_{X_1}^2\ d\xi\to 0.\tag{1} $$

Here is my question: Would anybody elaborate why this is true?

[Added:] For $u:\mathbb{R}\to X_1$, $\hat{u}$ denotes the Fourier transform (when exists) $$ \hat{u}(\xi)=\int_\mathbb{R}e^{-2\pi i \xi x}u(x)\ dx. $$

[Added later 11/26/16] In order to show (1) by the Lebesgue dominated convergence theorem, one essentially needs the following:

  • Show that for each $\xi$ with $|\xi|\leq M$, $\|\hat{v_m}(\xi)\|_{X_1}^2\to 0$ as $m\to\infty$. This is the same as saying that $\hat{v_m}(\xi)\to 0$ in $X_1$ strongly and it is done by showing that $\hat{v_m}(\xi)\to 0$ in $X_0$ weakly thanks to the elaboration in Fabian Wirth's answer.
  • Show that the map $\xi\mapsto 1_{|\xi|\leq M}\|\hat{v_m}(\xi)\|^2_{X_1}$ is dominated by some integrable function. (This step has not been elaborated yet.) The argument for this given by the proof is as follows:

    The assumption regarding the supports of $\{v_m\}$, we have $v_m=1_Kv_m$ and thus $$ \hat{v_m}(\xi)=\int_{\mathbb{R}}e^{-2\pi it\xi}1_K(t)v_m(t)\ dt, $$ which implies by Cauchy-Schwartz that for each $\xi$, $$ \|\hat{v_m}(\xi)\|_{X_1}\leq\|v_m\|_{L^2(\mathbb{R};X_1)}\|e^{-2\pi it\xi}1_K\|_{L^2(\mathbb{R};X_1)}. $$ The proof in the book immediately concludes that $$ \|\hat{v_m}(\xi)\|_{X_1}\leq C $$ for some constant $C$. Here is my follow-up question: How does one know that $C$ is independent of $m$?

[Added:] Here is the author's original argument:

enter image description here

  • $\begingroup$ Is the hat a Fourier transform, or something else? $\endgroup$ – Nate Eldredge Oct 11 '16 at 4:35
  • $\begingroup$ @NateEldredge: Yes, it is a Fourier transform. $\endgroup$ – Jack Oct 11 '16 at 12:05

The underlined sentence is a general statement about your triple $ X_0\overset{\textrm{cpt}}{\hookrightarrow} X\hookrightarrow X_1 $. If I understand the question correctly, you are satisfied that you have weak convergence of the sequence $\hat{\nu}_\mu$. The rest plays on general properties.

You have a weakly convergent sequence $\{\nu_\mu\}$ in $X_0$. By the principle of uniform boundedness this sequence is bounded in $X_0$. As the embedding into $X$ is compact, the embedded sequence has subsequences which are Cauchy in the norm of $X$. With this is mind it is not hard to show that $\{\nu_\mu\}$ cannot have a subsequence which does not converge to $0$ in the norm of $X$. (Take a subsequence which does not, this has a subsequence converging in $X$-norm to some $x\neq 0$. This gives a contradiction to weak convergence to $0$ in $X_0$.) In other words, the sequence converges to $0$ in the norm of $X$.

Finally, as the emebdding from $X$ to $X_1$ is continuous the sequence will also converge strongly to $0$ in $X_1$ because it already does in $X$.

| cite | improve this answer | |
  • $\begingroup$ Thank you for your answer. Now I see that (2.30) is not needed at all in this argument. $\endgroup$ – Jack Oct 16 '16 at 14:24
  • $\begingroup$ I am very embarrassed that I just found later your helpful answer explains only one part of the argument in order to get (1). By any chance, would you bother to take a look at the other part of it: why is $\{\|v_\mu\|_{X_1}\}$ bounded? $\endgroup$ – Jack Nov 26 '16 at 16:04
  • 1
    $\begingroup$ Hi Jack. Not sure what the question is here: ${\nu_\mu}$ is already bounded in $X_0$. As the embeddings are continuous, bounded sequences are mapped to bounded sequences. Does this solve your problem? $\endgroup$ – Fabian Wirth Nov 28 '16 at 10:16
  • $\begingroup$ Thank you very much for your comment! I think you meant "$v_{\mu}$ is already bounded in $L^2(\mathbb{R};X_0)$", which is implied by the assumption of weak convergence. (And I should have asked instead in the previous comment "why is $\{\|v_\mu\|_{L^2(\mathbb{R};X_1)}\}$ bounded") $\endgroup$ – Jack Nov 28 '16 at 13:25

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.