A Neural Network Approach for the Peak Profile Characterization

TL;DR

This paper introduces a neural network method for analyzing peak profiles in data like x-ray diffraction, accurately predicting characteristics such as width and asymmetry, aiding in initial parameter estimation for peak decomposition.

Contribution

The paper presents a novel neural network architecture specifically designed for Voigt-type peak profile analysis, applicable across various scientific fields.

Findings

Successfully tested on experimental x-ray diffraction data.

Accurately predicts peak profile parameters such as width and asymmetry.

Applicable to various types of peak distributions beyond the tested case.

Abstract

The neural network-based approach, presented in this paper, was developed for the analysis of peak profiles and for the prediction of base profile characteristics, such as width, asymmetry, asymptotic ("peak tales"), etc. of the observed distributions. The obtained parameters can be used as the initial parameters in the peak decomposition applications. The neural network architecture, presented here, was designed for the analysis of one particular type of peak profiles, the Voigt type distributions (symmetrical and asymmetrical), and is suitable for a variety of applications, such as x-ray and neutron powder diffraction, x-ray spectroscopy, etc. The approach itself, however, is not limited to the demonstrated case, but is applicable to other types of peak profile distributions. The approach was successfully tested on experimentally collected x-ray powder diffraction data.