Deep learning on nuclear mass and $\alpha$ decay half-lives

TL;DR

This paper demonstrates that deep neural networks can effectively predict nuclear masses and alpha decay half-lives, overcoming computational challenges in heavy nuclei, and highlights the importance of physical features like shell structure.

Contribution

The study introduces a deep learning approach for nuclear mass and decay predictions, incorporating physical priors and high-dimensional features, achieving competitive accuracy.

Findings

Achieved a standard deviation of 0.263 MeV in nuclear mass prediction.

Obtained a 0.797 standard deviation in alpha decay half-life prediction.

Identified the significance of shell structure and magic numbers in the model.

Abstract

Ab-initio calculations of nuclear masses, the binding energy and the $\alpha$ decay half-lives are intractable for heavy nucleus, because of the curse of dimensionality in many body quantum simulations as proton number($\mathrm{N}$) and neutron number($\mathrm{Z}$) grow. We take advantage of the powerful non-linear transformation and feature representation ability of deep neural network(DNN) to predict the nuclear masses and $\alpha$ decay half-lives. For nuclear binding energy prediction problem we achieve standard deviation $\sigma=0.263$ MeV on 10-fold cross validation on 2149 nuclei. Word-vectors which are high dimensional representation of nuclei from the hidden layers of mass-regression DNN help us to calculate $\alpha$ decay half-lives. For this task, we get $\sigma=0.797$ on 100 times 10-fold cross validation on 350 nuclei on $log_{10}T_{1/2}$ and $\sigma=0.731 $ on 486 nuclei. We also find physical a priori such as shell structure, magic numbers and augmented inputs inspired by Finite Range Droplet Model are important for this small data regression task.