Nuclear Reaction Theory in the Precision Era: Probing New Physics in the Early Universe

Project Overview

In recent years, experimental measurements of nuclear reactions have further pushed the limits of precision. Theoretical calculations, however, are just now beginning to follow suit, with robust uncertainty quantification (UQ) remaining out of reach, in most cases due to the increased amount of computational effort required. To remedy this situation, we combined first-principles calculations of nuclear reactions based on state-of-the-art nucleon–nucleon and three-nucleon interactions derived from chiral effective field theory (EFT) with efficient Gaussian process (GP) emulators to perform Markov-chain Monte Carlo estimations of theoretical uncertainties. A key input for constructing GP emulators is a modest number of training runs (typically in the tens or hundreds) that will map out the chiral EFT input parameter space. For complex reactions, even this may be computationally infeasible. Therefore, we construct a mixed-fidelity approach that leverages multiple computationally cheap low-fidelity calculations, merging them with a handful of high-fidelity ones to minimize the GP emulator error. This mixed-fidelity approach dramatically reduces the amount of time required to generate the training points for the GP. Furthermore, we were able to use such GP emulators to conduct a sensitivity study for constraining the chiral EFT parameters on experimental data. This study clarified the enhanced impact of certain terms in the chiral expansion, previously thought to be of reduced importance, and demonstrated the viability of GP emulators for UQ in nuclear reactions.

Finally, we devised a novel approach that exploits correlations in low-energy observables to produce reaction cross section evaluations with greatly reduced theoretical uncertainties. By combining runs at various chiral EFT input parameters, we were able to uncover underlying correlations between nuclear reaction observables connecting observations that have been made in the lab to difficult-to-measure cross sections at astrophysically relevant energies. Using this approach, combined with experimental data, we arrived at an estimate for the zero-energy proton radiative capture cross section on beryllium-7 that has significantly smaller uncertainty than the currently recommended value.

Mission Impact

This project opened a new avenue for evaluating nuclear data for light-ion nuclear reactions and deriving correlations and covariances for nuclear observables using first-principles calculations as a basis. Furthermore, the approach is coupled with an efficient mixed-fidelity Gaussian process emulator for a robust estimate of theoretical uncertainties. As a result, by providing access to uncertainty-quantified nuclear data, this research enhances our fundamental understanding of nuclear processes that help maintain and enhance the safety, security, and effectiveness of the U.S. nuclear weapons stockpile, results that are fundamental to Lawrence Livermore National Laboratory's stockpile stewardship mission. The outcome of this research can support the DOE Office of Science and NNSA's Office of Defense Nuclear Nonproliferation program missions by providing crucial theoretical input for interpreting experimental data, as well as identifying reactions where new experiments are needed.

Publications, Presentations, and Patents

"Predictive theory for light-ion reactions, nuclear physics at the edge of stability," ECT* (Virtual), June 2021 (LLNL-PRES-824063).

"First Principles Calculations of Light Ion Reactions," Reaction Seminar (Virtual), June 2021 (LLNL-PRES-823550).

"First Principles Calculations of Atomic Nuclei and Their Interactions," Grand Challenge Seminar Series, LLNL (Virtual), May 2021 (LLNL-PRES-822625).

"First Principles Calculations of Light Ion Reactions," Theory Seminar (Virtual) Michigan State University, April 2021 (LLNL-PRES-821202).

"Predictive Calculations of Nuclear Reactions with The No-Core Shell Model with Continuum," American Physical Society April Meeting, April 2021 (LLNL-PRES-821573).

"Light-ion reaction calculations from first principles," Uncertainties in Calculations of Nuclear reactions of Astrophysical Interest workshop, Mainz ITP (Virtual), December 2020 (LLNL-PRES-817357).

"No-Core Shell Model with Continuum Approach to Alpha Clustering and Alpha Induced Reactions," APS-DNP Fall Meeting, October 2020 (LLNL-PRES-815996).

"Uncertainty quantification in ab initio nucleon-alpha scattering," APS April Meeting, April 2020 (LLNL-PRES-808818).

"Ab initio calculations of alpha-Clustering and alpha-induced reactions," Progress in Ab Initio techniques in nuclear physics workshop, TRIUMF, March 2020 (LLNL-PRES- 806181).

"Alpha clustering and alpha induced reactions," Vth topical workshop on modern aspects of nuclear structure, Bormio, February 2020 (LLNL-PRES- 804658).