Natural orbitals for many-body expansion methods

TL;DR

This paper demonstrates that natural orbitals, constructed from large single-particle bases, significantly improve the efficiency and accuracy of many-body nuclear calculations, enabling larger and more precise simulations.

Contribution

It systematically benchmarks natural orbitals in nonperturbative many-body methods, showing their effectiveness in reducing computational cost and improving convergence.

Findings

Natural orbitals enable reduced model spaces with maintained accuracy.

Significant computational savings in large-scale nuclear calculations.

Enhanced convergence and lower sensitivity to basis parameters.

Abstract

The nuclear many-body problem for medium-mass systems is commonly addressed using wave-function expansion methods that build upon a second-quantized representation of many-body operators with respect to a chosen computational basis. While various options for the computational basis are available, perturbatively constructed natural orbitals recently have been shown to lead to significant improvement in many-body applications yielding faster model-space convergence and lower sensitivity to basis set parameters in large-scale no-core shell model diagonalizations. This work provides a detailed comparison of single-particle basis sets and a systematic benchmark of natural orbitals in nonperturbative many-body calculations using the in-medium similarity renormalization group approach. As a key outcome we find that the construction of natural orbitals in a large single-particle basis enables for performing the many-body calculation in a reduced space of much lower dimension, thus offering significant computational savings in practice that help extend the reach of ab initio methods towards heavier masses and higher accuracy.