Sobolev-type inequalities for Dunkl operators (Q785889)

The author studies the Sobolev inequality in the Dunkl setting (as well as other estimates, like Nash inequality or Besov space embeddings), providing a simpler elementary proof of the classical case \(p = 2\) and an extension to the end-point case \(p = 1\), that was previously unknown. The main result shows that if \(N+2\gamma>2\), then there exists \(C>0\) such that \[ \|f\|_{L^q(\mu_k)}\le C\|\nabla_kf\|_{L^2(\mu_k)}, \] for all \(f\in C_c^\infty(\mathbb R^N)\), where \(\mu_k\) is the Dunkl measure, \(\nabla_k\) is the Dunkl gradient and \(q=2(N+2\gamma)/(N+2\gamma-2).\) Using some ideas from [\textit {B. Bakry} et al., Indiana Univ. Math. J. 44, No. 4, 1032--1074 (1995; Zbl 0857.26006)], this estimate follows in fact as a consequence of the Nash inequality: \[ \|f\|^{1+2/(2\gamma+N)}_{L^2(\mu_k)}\le C\|\nabla_kf\|_{L^2(\mu_k)}\|f\|^{2/(2\gamma+N)}_{L^1(\mu_k)}. \] A key ingredient in the proofs is the carré-du-champ operator, defined as \[\Gamma(f)=\frac12(\Delta_k(f^2)-2\Delta_k f). \] Further extensions are also considered involving Besov spaces. In the last section, explicit expressions for the best constant are obtained as a consequence of an isoperimetric inequality.