Balanced distribution-energy inequalities and related entropy bounds

Abstract

Let $A$ be a self-adjoint operator acting over a space $X$ endowed with a partition. We give lower bounds on the energy of a mixed state $\rho$ from its distribution in the partition and the spectral density of $A$. These bounds improve with the refinement of the partition, and generalize inequalities by Li-Yau and Lieb--Thirring for the Laplacian in $\R^n$. They imply an uncertainty principle, giving a lower bound on the sum of the spatial entropy of $\rho$, as seen from $X$, and some spectral entropy, with respect to its energy distribution. On $\R^n$, this yields lower bounds on the sum of the entropy of the densities of $\rho$ and its Fourier transform. A general log-Sobolev inequality is also shown. It holds on mixed states, without Markovian or positivity assumption on $A$.