The concentration-compactness principle in the calculus of variations. The locally compact case. II (Q917004): Difference between revisions

[For part I see the author, ibid. 109-145 (1984; Zbl 0541.49009) which is also covered by the following review.] Let H be a function space on \({\mathbb{R}}^ N\), and let J,\({\mathcal E}:\) \(H\to {\mathbb{R}}\), \[ {\mathcal E}(u)=\int e(x,Au(x))dx,\quad J(u)=\int j(x,Bu(x))dx, \] where e: \({\mathbb{R}}^ N\times {\mathbb{R}}^ m\to {\mathbb{R}}\), j: \({\mathbb{R}}^ N\times {\mathbb{R}}^ n\to [0,\infty [\), and A: \(H\to E\), B: \(H\to F\) (E,F are function spaces defined on \({\mathbb{R}}^ N\) with values in \({\mathbb{R}}^ m\), \({\mathbb{R}}^ n\), respectively) commute with a translation of \({\mathbb{R}}^ N\); we consider the minimization problem inf\(\{\) \({\mathcal E}(u):\) \(u\in H\), \(J(u)=1\}\). Because of the loss of boundedness of domains, the classical convexity-compactness methods fail to treat the problem and thus the author presents a new method to solve it. He derives a general principle in a heuristic form and it is rigorously justified on all problems studied in the paper. He first imbeds the problem into a one-parameter family of problems \[ I_{\lambda}=\inf \{{\mathcal E}(u):\;u\in H,\quad J(u)=\lambda \},\quad \lambda >0; \] he supposes \(j(x,q)\to j^{\infty}(q)\), \(e(x,p)\to e^{\infty}(p)\) as \(x\to \infty\) for all \(p\in {\mathbb{R}}^ m\), \(q\in {\mathbb{R}}^ n\); and he considers \[ I^{\infty}_{\lambda}=\inf \{{\mathcal E}^{\infty}(u):\;u\in H,\quad J^{\infty}(u)=\lambda \},\text{ where } {\mathcal E}^{\infty}(u)=\int e^{\infty}(Au(x))dx,\quad J^{\infty}(u)=\int j^{\infty}(B(u(x))dx; \] he assumes \(\{\) \(u\in H:\) \(J(u)=\lambda \}\neq \emptyset\), \(I_{\lambda}>-\infty\) for \(\lambda\in]0,1]\) and that minimizing sequences for \(I_{\lambda}\), \(I^{\infty}_{\lambda}\) are bounded in H. The concentration-compactness principle is the following: In the case when e and j depend on the first variable, for each \(\lambda >0\) all minimizing sequences for I are relatively compact if and only if the strict subadditivity condition \(I_{\lambda}<I_{\alpha}+I^{\infty}_{\lambda -\alpha}\) holds for all \(\alpha\in [0,\lambda [\); in the case when e and j do not depend on the first variable, for each \(\lambda >0\) all minimizing sequences for I are relatively compact up to a translation if and only if the strict subadditivity condition \(I_{\lambda}<I_{\alpha}+I^{\infty}_{\lambda -\alpha}\) holds for all \(\alpha\in]0,\lambda [\) (he remarks that the weak subadditivity condition is always satisfied). The proof is based upon a compactness lemma obtained with the help of the notion of the concentration function of a measure. The author gives a rigorous proof of the previous principle in several examples: The rotating star problem: \[ \inf \{\int [j(\rho (x))+k(x)\rho (x)]dx-(1/2)\int \rho (x)\rho (y)f(x-y)dxdy:\;\rho \geq 0,\quad \rho \in L^ 1({\mathbb{R}}^ 3),\quad \int \rho (x)dx=\lambda \} \] where K, f are given, j is a convex function, \(\lambda >0\); the Choquard-Pekar problem: \[ \inf \{\int [(1/2)| \nabla u(x)|^ 2+(1/2)V(x)u(x)^ 2]dx- (1/4)\int u(x)^ 2u(y)^ 2(1/| x-y|)dxdy\}, \] \[ subject\quad to\quad u\in H^ 1({\mathbb{R}}^ N)\text{ and } \int u(x)^ 2dx=1; \] the standing waves in nonlinear Schrödinger equations: \[ \inf \{\int [| \nabla u(x)|^ 2-F(x,u(x))]dx:\;u\in H^ 1({\mathbb{R}}^ N),\quad | u|^ 2_{L^ 2({\mathbb{R}}^ N)}=1\} \] (e.g. \(F(x,t)=| t|^ p)\) and inf\(\{\int [| \nabla u(x)|^ 2+V(x)u(x)^ 2]dx:\) \(u\in H^ 1({\mathbb{R}}^ N)\), \(\int K(x)| u(x)|^ pdx=1\}\), \(p>1\); nonlinear field equations: \[ \inf \{\int | \nabla u(x)|^ 2dx:\;F(x,u(x))dx=\lambda \}; \] unconstrained problems (e.g., Hartree-Fock problems); Euler equations and minimization over manifolds; problems with multiple constraints; problems in unbounded domains other than \({\mathbb{R}}^ N\) (strips, half-spaces, exterior domains, etc.); problems invariant by translation only in some particular directions (e.g., vortex rings, rotating stars).