Parameter-Space Heat Flow, Gaussian Density Ratios, and Sharp Hermite Truncation Rates

We reinterpret the classical Hermite generating function as a Gaussian density ratio: relative to the unit Gaussian reference, it is the density ratio of a Gaussian with shifted mean and unchanged covariance. Applying the heat semigroup in the mean-parameter variable to this generating function produces the corresponding temperature variation. Thus the heat-semigroup time variable is reinterpreted as the temperature variation of the Gaussian density ratio. This parameter-space formulation also gives a parabolic control principle for Hermite approximation errors. Since Hermite projections act in the velocity variable and the heat flow acts in the mean variable, Hermite block energies and truncation tails are subsolutions of the same parameter-space heat equation. This remains useful for heat-evolved non-Gaussian perturbations where no usable closed coefficient formula is available. For Gaussian density ratios with general covariance, the Hermite coefficients satisfy a weighted homogeneity in the mean and covariance-defect parameters. This yields Ornstein–Uhlenbeck covariance, an exact generating function for total-degree Hermite block energies, and the sharp geometric Hermite truncation rate, equal to the square root of the largest absolute covariance defect. We also derive precise isotropic block and tail asymptotics and interpret the estimates for near-Gaussian kinetic distributions.