Smoothing a constructed response

This proposition smooths an already defined distribution. It is not a construction of an otherwise undefined graph product, and finite-$ϵ$ smoothing need not preserve exact causal support.

The benchmark is an exact first derivative. No nonperturbative bound on the $O({\lambda }^2)$ term in (eq:source-expansion) is claimed. Such a bound, calibration uncertainty, and apparatus noise are separate from (eq:smoothing-rate) and (eq:annihilation-error).