Nuclear <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>β</mml:mi> </mml:math> -decay half-life predictions and <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>r</mml:mi> </mml:math> -process nucleosynthesis using machine learning models

Abstract

This study investigates the predictive capabilities of machine learning models in nuclear $\ensuremath{\beta}$-decay half-life predictions and their application to $r$-process nucleosynthesis. The research explores the intricacies of statistical modeling using support vector machines (SVM), focusing on understanding the learning and prediction of various nuclear configurations and features that influence $\ensuremath{\beta}$-decay half-lives. By considering a comprehensive dataset spanning from light to heavy mass nuclei, the SVM demonstrates remarkable accuracy in reproducing experimentally known half-lives across diverse nuclear structures. Evaluations reveal the effectiveness of this model across different nuclear classes, with notable improvements observed in even-even nuclei predictions. Furthermore, this study demonstrates the extrapolative capabilities of SVM predictions in solar $r$-process nucleosynthesis, emphasizing its ability to accurately predict $r$-process abundances. The SVM model, particularly when utilizing the radial basis function kernel, exhibits strong agreement with experimental data, providing valuable insights into the behavior of highly neutron-rich nuclei. These findings underscore the significance of machine learning as a powerful tool in nuclear physics research, offering promising avenues for advancing our understanding of $\ensuremath{\beta}$-decay processes and $r$-process nucleosynthesis.

Topics & Concepts