Dirichlet Process Mixture of Voigt Profiles with Application to Peak Detection in X-Ray Photoelectron Spectroscopy Data

• Activity Number: 667
• Type: Contributed
• Date/Time: Thursday, August 4, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Anton Lobach* and Gavino Puggioni and David Heskett and Benjamin Young
• Companies: University of Rhode Island and University of Rhode Island and University of Rhode Island and Rhode Island College
• Keywords: Dirichlet Process mixture ; Bayesian Nonparametrics ; Voigt density ; X-ray spectroscopy ; Peak detection ; Convolution of Normal and Cauchy densities

Abstract

In order to increase the life cycle of Li-ion batteries, it is important to study the chemical composition of battery electrodes' surface, described by the electrons energy distribution. The Hard X-ray Photoelectron Spectroscopy (HAXPES) technique was used to measure this distribution which is usually multimodal. The locations of clusters in the data correspond to binding energies of electrons. Since every chemical structure has a particular binding energy, by estimating clusters locations we can infer the chemical composition of materials. In order to identify locations and number of clusters we propose a model based on a Dirichlet Process (DP) mixture of Voigt densities (convolutions of normal and Cauchy distributions). The model is validated with a simulation study and then applied to the HAXPES datasets.