Neural networks and determination of diatomic molecule interatomic potential of cadmium dimer

A new method for obtaining a pointwise potential energy curve of diatomic molecule using Neural Network is reported. The method is employed to generate new characteristics of the B11u(51P1) electronic state of Cd2 based on LIF excitation spectrum previously recorded using the B11u←X10g+(51S0) transition. The obtained potential provides better simulation-to-experiment agreement than those obtained using other methods. Correctness of the method is additionally tested on artificially generated LIF excitation spectra based on well characterized b30u+(53P1)←X10g+ transition in Cd2. A method for obtaining parameters of an analytical form of the potential using Neural Network applied on artificially generated Cd2 spectra is also presented.