Automation of hyperfine interaction parameters analysis using machine learning techniques

Abstract

The time-differential perturbed angular correlation (TDPAC) spectra often exhibit complex interactions between multiple hyperfine fields, making it difficult to isolate individual parameter contributions and accurately model the system. This study introduces a Python-based software designed to efficiently process files containing hyperfine parameters, which are essential for analyzing results obtained from fitting TDPAC spectra. The primary objective is to investigate hyperfine interactions in a variety of materials. The software organizes hyperfine parameters measured at different temperatures according to their respective sites and thermal regimes, enabling the automatic separation of behaviors associated with heating and cooling cycles. During the exploratory data analysis, unsupervised machine learning techniques, such as cluster analysis and Principal Component Analysis (PCA), are employed. These techniques facilitate the identification of cluster formation as a function of temperature, enabling the examination of correlations among parameters and the resulting groupings. In addition, the software implements a rule-based decision tree to automatically assign the most appropriate correlation coefficient to each pair of variables, taking into account the statistical characteristics of the dataset. This methodology represents a comprehensive and adaptable approach for the statistical analysis of TDPAC data.