Determination of Resonant State Parameters from Ab Initio Data by Artificial Neural Network and Statistical Pade Approximation
Abstract
Resonant states in electron molecule collisions mediate a variety of energy transfer processes in the plasma edge region, such as vibrational excitation, dissociative recombination, dissociative attachment, associative detachment etc. Here we show that the resonance parameters, in general difficult to obtain, can be computed from standard bound-state ab initio data by means of analytical continuation in the coupling constant.
The procedure uses artificial neural network and statistical Pade approximation to extrapolate from the bound-state region to that of the resonant state by varying the strength of the attractive potential term added. We present benchmark data for the ethylene molecule and demonstrate a reasonable stability of the results over the quantum chemical basis sets employed.