Note on the concentration-compactness principle for generalized Moser-Trudinger inequalities (Q424095): Difference between revisions

Define the funtions \(\log_{[j]}\) and \(\exp_{[j]}\) by \(\log_{[1]}t=\log t\), \(\exp_{[1]}(t)=\exp(t)\) and \(\log_{[j]}t=\log(\log_{[j-1]}t)\), \(\exp_{[j]}t=\exp(\exp_{[j-1]}t)\) for \(j\in\mathbb N\), \(j>1\). The author proves the following analogue of a result of \textit{P.-L. Lions} [Rev. Mat. Iberoam. 1, No. 1, 145--201 (1985; Zbl 0704.49005)] for Orlicz-Sobolev spaces embedded into exponential and multiple-exponential spaces: Theorem. Let \(\ell\in\mathbb N\), \(n\geq2\), \(\alpha<n-1\) and let \(\Omega\subset\mathbb R^n\) be a bounded domain. Let \(\Phi\) be a Young function such that \(\lim\limits_{t\to\infty}\Phi(t)\left[t^n\bigl(\prod^{\ell-1}_{j=1}\log^{n-1}_{[j]}t\bigr)\log^\alpha_{[\ell]}t\right]^{-1}=1\). If \(\{u_k\}\) is a sequence of functions from the Orlicz-Sobolev space \(W_0L^\Phi(\Omega)\) such that \(\int_\Omega\Phi(|\nabla u_k|)\leq1\), \(u_k\rightharpoonup u\) in \(W_0L^\Phi(\Omega)\) and \(u_k\to u\) a.e. in \(\Omega\) for some \(u\in W_0L^\Phi(\Omega)\), then, for every \(p<P=(1-\int_\Omega\Phi(|\nabla u|))^{-1/(n-1-\alpha)}\) (with \(P=\infty\) if \(\int_\Omega\Phi(|\nabla u|)=1\)), the estimate \(\int_\Omega\exp_{[\ell]}(K_{\ell,n,\alpha}p| u_k|^{n/(n-1-\alpha)})\leq C\) holds with \(C\) independent of \(k\). An example shows that, in general, the result does not hold for \(p>P\).