Convex Sobolev inequalities derived from entropy dissipation (Q717439): Difference between revisions

From the authors' abstract: This paper is concerned with families of convex Sobolev inequalities of Beckner type, \[ \frac{1}{2-p}\left[\int_{\Omega}\omega^2 d\mu -\left(\int_{\Omega}\omega^p d\mu\right)^{\frac{2}{p}}\right]\leq C_p \int_{\Omega}D|\nabla \omega|^2 d\mu \quad (1\leq p<2), \tag{a} \] which arise as entropy-dissipation relations for certain linear Fokker-Planck equations. Extending the ideas recently developed by the first two authors, a refinement of the Bakry-Émery method is established, which allows us to prove non-trivial inequalities even in situations where the classical Bakry-Émery criterion fails. The main application of our theory concerns the linearized fast diffusion equation in dimensions \(d\geq 1,\) like \[ \partial_tF=\triangle(F^m)+\nabla \cdot (xF)\quad \text{on }\Omega\subset\mathbb R^d, \tag{b} \] which admits a Poincaré, but no logarithmic Sobolev inequality. We calculate bounds on the constants in the interpolating convex Sobolev inequalities, and prove that these bounds are sharp on a specified range. In dimension \(d=1\), our estimates improve the corresponding results that can be obtained by the measure-theoretic techniques of Barthe and Roberto. As a by-product, we give a short and elementary alternative proof of the sharp spectral gap inequality first obtained by Denzler and McCann. In further applications of our method, we prove convex Sobolev inequalities for a mean field model for the redistribution of wealth in a simple market economy, and the Lasota model for blood cell production.