A Free Analog of Bobkov's Gaussian Isoperimetry Inequality

We prove a one-variable functional inequality which is the free probability analog of Bobkov’s isoperimetry inequality. The inequality involves the $L^1$ norm of the difference quotient of a function $f$ and can be viewed as a non-local isoperimetric inequality. We also prove related inequalities for subsets of an interval as well as for subsets of roots of Hermite polynomials. This paper is also an experiment in AI-based exploration of free analogs of classical probability statements.