Non-strict singularity of optimal Sobolev embeddings

TL;DR

This paper studies the non-strict singularity of optimal Sobolev embeddings in rearrangement-invariant spaces, showing they are not strictly singular in many cases by constructing specific spike-function sequences.

Contribution

It introduces a framework for proving non-strict singularity of optimal Sobolev embeddings using spike-function sequences, especially in weighted Lambda spaces.

Findings

Optimal Sobolev embeddings are not strictly singular in a large class of r.i. spaces.

Constructs spike-function sequences to demonstrate non-strict singularity.

Shows non-strict singularity for embeddings in weighted Lambda spaces, excluding the endpoint case.

Abstract

We investigate the operator-theoretic property of strict singularity for optimal Sobolev embeddings within the general framework of rearrangement-invariant function spaces (r.i. spaces). More specifically, we focus on studying the ``quality'' of non-compactness for optimal Sobolev embeddings $V^m_0X(\Omega)\to Y_X(\Omega)$, where $X$ is a given r.i. space and $Y_X$ is the corresponding optimal target r.i. space (i.e., the smallest among all r.i. spaces). For the class of sub-limiting norms (i.e., the norms whose fundamental function satisfies $\varphi_{Y_X}(t)\approx t^{-m/n}\varphi_X(t)$ as $t\to0^+$), we construct suitable spike-function sequences that establish a general framework for proving non-strict singularity of optimal (and thus non-compact) sublimiting Sobolev embeddings. As an application, we show that optimal sublimiting Sobolev embeddings are not strictly singular in a rather large subclass of r.i. spaces, namely weighted Lambda spaces $X=\Lambda^q_w$, $q\in[1, \infty)$. Except for the endpoint case $X=L^{n/m,1}$, our spike-function construction enables us to construct a subspace of $V^m_0X$ that is isomorphic to $\ell_q$, which we then leverage to prove the non-strict singularity of the corresponding optimal Sobolev embedding.