Concentration-Compactness Principle: The loss of mass at infinity in the subcritical case

May 13, 2010

Concentration-Compactness Principle: The loss of mass at infinity in the subcritical case

Filed under: PDEs — Tags: Concentration-Compactness — Ngô Quốc Anh @ 16:18

One of the main problems encountered in solving variational problems is to show the convergence of minimizing sequence (the direct method). In many problems, it is relatively easy to show the boundedness of a minimizing sequence in an appropriate Sobolev space.

Th aim of this section is to derive the concentration-compactness at infinity (just for subcritical case). Roughly speaking, for unbounded domains due to the lack of compact embedding in the Sobolev embedding theorem, the loss of mass of a weakly convergent sequence may occur. In fact, this can only occur at infinity.

Let us firstly consider the case $\Omega \subset \mathbb R^n$ is a bounded domain. Let $\{u_m\}$ be a sequence in $W_0^{1,2}(\Omega)$ such that

$u_m \rightharpoonup u$

weakly in $W_0^{1,2}(\Omega)$ (i.e. $u \in W_0^{1,2}(\Omega)$ and $f(u_m) \to f(u)$ for any $f \in W_0^{-1,2}(\Omega)$ ). Assume $2<p<2^\star$ . The Sobolev embedding tells us that

$W_0^{1,2}(\Omega) \hookrightarrow L^p(\Omega)$

is compact. Thus

$u_m \to u$

strongly in $L^p(\Omega)$ in the sense that

$\displaystyle\mathop {\lim }\limits_{m \to \infty } \int_\Omega {{{\left| {{u_m} - u} \right|}^p}dx} = 0$ .

It now follows from the Brezis-Lieb lemma or the use of the Minkowski inequality

$\displaystyle {\left( {\int_\Omega {{{\left| {{u_m}} \right|}^p}dx} } \right)^{\frac{1}{p}}} \leqslant {\left( {\int_\Omega {{{\left| {{u_m} - u} \right|}^p}dx} } \right)^{\frac{1}{p}}} + {\left( {\int_\Omega {{{\left| u \right|}^p}dx} } \right)^{\frac{1}{p}}}$

and

$\displaystyle {\left( {\int_\Omega {{{\left| u \right|}^p}dx} } \right)^{\frac{1}{p}}} \leqslant {\left( {\int_\Omega {{{\left| {u - {u_m}} \right|}^p}dx} } \right)^{\frac{1}{p}}} + {\left( {\int_\Omega {{{\left| {{u_m}} \right|}^p}dx} } \right)^{\frac{1}{p}}}$

that

$\displaystyle \mathop {\lim }\limits_{m \to \infty } \int_\Omega | {u_m}{|^p}dx = \int_\Omega | u{|^p}dx$

$\displaystyle\mathop {\lim }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | {u_m}{|^p}dx = \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | u{|^p}dx$ .

However, when $\Omega$ is unbounded, the situation is quite different. Actually, we are going to prove the following

Theorem. Let $\Omega \in \mathbb R^n$ be an unbounded domain and $2<p<2^\star$ . Let $\{u_m\}$ be a sequence in $W_0^{1,2}(\Omega)$ such that $u_m \rightharpoonup u$ in $W_0^{1,2}(\Omega)$ and define

$\displaystyle {\alpha_\infty } = \mathop {\lim }\limits_{R \to \infty } \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | {u_m}{|^p}dx$

and

$\displaystyle {\beta _\infty } = \mathop {\lim }\limits_{R \to \infty } \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | \nabla {u_m}{|^2}dx$ .

Then these quantities are well-defined and satisfy

$\displaystyle\mathop {\lim \sup }\limits_{m \to \infty } \int_\Omega {{{\left| {{u_m}} \right|}^p}dx} = \int_\Omega {{{\left| u \right|}^p}dx} + {\alpha _\infty }$

and

$\displaystyle\mathop {\lim \sup }\limits_{m \to \infty } \int_\Omega {{{\left| {\nabla {u_m}} \right|}^2}dx} \geqslant \int_\Omega {{{\left| {\nabla u} \right|}^2}dx} + {\beta _\infty }$ .

Proof. For each $R>0$

$\displaystyle\begin{gathered} \mathop {\lim \sup }\limits_{m \to \infty } \int_\Omega {{{\left| {{u_m}} \right|}^p}dx} = \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | {u_m}{|^p}dx + \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | {u_m}{|^p}dx \hfill \\ \qquad= \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | u{|^p}dx + \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | {u_m}{|^p}dx. \hfill \\ \end{gathered}$

Letting $R \to \infty$ gives us the first relation.

Similarly,

$\displaystyle\mathop {\lim \sup }\limits_{m \to \infty } \int_\Omega {{{\left| {\nabla {u_m}} \right|}^2}dx} = \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | \nabla {u_m}{|^2}dx + \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | \nabla {u_m}{|^2}dx$ .

Note that

$\displaystyle \nabla u_m \rightharpoonup \nabla u$

in $L^2(\Omega\cap\{|x|<R\})$ . Using the weak lower semicontinuity of norms we get

$\displaystyle\mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | \nabla {u_m}{|^2}dx \geqslant \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | \nabla u{|^2}dx$

which implies

$\displaystyle\mathop {\lim \sup }\limits_{m \to \infty } \int_\Omega {{{\left| {\nabla {u_m}} \right|}^2}dx} \geqslant \int\limits_{\Omega \cap \left\{ {|x| < R} \right\}} | \nabla u{|^2}dx + \mathop {\lim \sup }\limits_{m \to \infty } \int\limits_{\Omega \cap \left\{ {|x| > R} \right\}} | \nabla {u_m}{|^2}dx$ .

Thus by sending $R\to\infty$ we obtain the second identity.

About these ads

Leave a Comment »

No comments yet.

RSS feed for comments on this post. TrackBack URI

M	T	W	T	F	S	S
« Apr				Jun »
					1	2
3	4	5	6	7	8	9
10	11	12	13	14	15	16
17	18	19	20	21	22	23
24	25	26	27	28	29	30
31

	Ngô Quốc Anh on The Poincaré inequality: W^{1,…
	Paul on The Poincaré inequality: W^{1,…
	Ngô Quốc Anh on A lower bound for solutions in…
	fanicolas on A lower bound for solutions in…
	Ngô Quốc Anh on The Alexandrov-Bol inequa…
	Fran on A lower bound for solutions in…
	Fab on The Alexandrov-Bol inequa…
	Trần Lượng on “Tích phân suy rộng…
	Trần Lượng on “Tích phân suy rộng…
	Ngô Quốc Anh on Contact me

Ngô Quốc Anh

May 13, 2010