Bayesian predictions for A=6 nuclei using eigenvector continuation emulators

Artikel i vetenskaplig tidskrift, 2022

We make ab initio predictions for the A=6 nuclear level scheme based on two- and three-nucleon interactions up to next-to-next-to-leading order in chiral effective field theory (χEFT). We utilize eigenvector continuation and Bayesian methods to quantify uncertainties stemming from the many-body method, the χEFT truncation, and the low-energy constants of the nuclear interaction. The construction and validation of emulators is made possible via the development of jupiterncsm - a new M-scheme no-core shell model code that uses on-the-fly Hamiltonian matrix construction for efficient, single-node computations up to Nmax=10 for Li6. We find a slight underbinding of He6 and Li6, although consistent with experimental data given our theoretical error bars. As a result of incorporating correlated χEFT-truncation errors we find more precise predictions (smaller error bars) for separation energies: Sd(Li6)=0.89±0.44MeV, S2n(He6)=0.20±0.60MeV, and for the beta decay Q value: Qβ-(He6)=3.71±0.65MeV. We conclude that our error bars can potentially be reduced further by extending the model space used by jupiterncsm.