search

WolfgangdJ / Manuwolf

Manuscript development with manubot
0 stars 0 forks source link

center of mass sampling #1

Open WolfgangdJ opened 4 years ago

WolfgangdJ commented 4 years ago

The actual numerical problem lies in the necessity of extremely long runs compared to those with fixed center of mass. The statistical reliability of the Monte-Carlo scheme suffers from a variable step width that extremely depends on the steepness of the lateral slope in contrast to the flatness of the longitudinal slope of the wave function for a correct sampling of the
distribution. A failure in the sampling shows in decreasing step acceptance rates occurring together with a loss of kinetic energy and a misleading gain of potential energy close to the tips, i.e. at the classical centers of attraction.

WolfgangdJ commented 4 years ago

The strong lateral COM localization by the value of COMZETARHO agrees with the balancing between the kinetic and potential energy. In contrast the longitudinal nuclear motion which is controlled by the decay constant COMZETAX depends also strongly on NUZETAX. The high gain in potential energy with increasing NUZETAX cannot be balanced by the nuclear kinetic energy which is small by 1/NUMASS. What is going on there, error in code or are the electron parameters significantly to be changed?