Jensen-Mercer two-sided bounds for entropy production rate

Abstract

Bounds on entropy production convert limited observations into constraints on hidden dissipation. The Jensen bound expresses a lower bound on the total and subsystem entropy production rates of continuous stochastic systems in terms of mean coordinate rates alone. We show that the same convexity structure yields a complementary upper bound, inspired by Mercer's reflection inequality. For the quadratic entropy-production functional of overdamped Langevin dynamics, Mercer's reflection inequality is identical to the sharp Bhatia-Davis variance bound and coincides with the convex dual of the smallest affine majorant of the functional over the range of the local current. The two bounds yield a two-sided sandwich controlled solely by the mean drift and the extrema of the current, organized into a hierarchy of optimal bounds in which each further observation tightens the ceiling. The thermodynamic uncertainty relation supplies an independent fluctuation floor. We also treat position-dependent diffusion, underdamped dynamics and the tilted periodic potential, for which the weak-barrier series is obtained in closed form and the strong-barrier regime is solved asymptotically. We check the bounds numerically on the Brownian motor.