Estimating Inverse Scattering Potentials for n-p System Using Variational Monte Carlo & Neural Networks

The Riccati-type nonlinear differential equation, also named the Variable Phase Approach or Phase Function Method, is used to construct local inverse potentials for the ³S₁ and ¹S₀ states of the deuteron. These potentials have been optimized by adjusting Morse potential parameters using the Variational Monte Carlo (VMC) and Multilayer perceptron (MLP) type Neural Networks (NN). The inverse potentials obtained from VMC and NN show almost identical parameters. In VMC all three parameters are varied to obtain the phase shifts while in NN the 3D-parameter problem is converted to a 1D-parameter optimization problem. Recently, the GRANADA group published a comprehensive partial wave analysis of scattering data, which includes 6713 np phase shift data points from 1950 to 2013. Using the final experimental data points from GRANADA, we obtained the parameters for the Morse potential by minimizing the mean square error (MSE) as the cost function. Various quantum functions, such as phase δ(r), amplitude $A(r)$, and wave function $u(r)$, are described up to 5fm with energies Eab=[1-350 MeV] for ³S₁ and ¹S₀ state.