Bayesian Neural Network improvements to nuclear mass formulae and predictions in the SuperHeavy Elements region

Abstract

A systematic study based on the Bayesian Neural Network (BNN) statistical approach is introduced to improve the predictive power of current nuclear mass formulae when applied to nuclides not yet experimentally detected. In a previous work by the present authors, the methodology was introduced considering only the Duflo-Zuker mass model (Duflo J. and Zuker A., Phys. Rev. C, 52 (1995) R23) to explore the S uperH eavy E lements (SHE) region, with focus on the α-decay process. Due to the discrepancy among different mass formulae we decided to extend in the present calculation the application of the Bayesian Neural Network methodology to other mass formula models and to discussing their implications on predictions of SHE α-decay half-lives. The -value prediction using a set of ten different mass models has been greatly improved for all models when compared to the available experimental data. In addition, we have used the improved -value to determine the SHE α-decay half-lives with a well-succeeded model in the literature, currently employed for different hadronic nuclear decay modes of heavy nuclei, the E ffective L iquid D rop M odel (ELDM). Possible SHE candidates recently investigated are explicitly calculated (specially the 298, 299,300120 isotopes, and results present a promising via of research for these nuclei through α-decay process.