Abstract

We extend the neural network backflow (NNBF) approach to ab-initio solid-state materials using a two-stage pruning strategy to efficiently optimize large configuration spaces in periodic systems. The method is demonstrated on benchmark calculations for 1D hydrogen chains, 2D graphene, and 3D silicon, showing competitive accuracy and improved computational efficiency.

Abstract

We extend the neural network backflow (NNBF) approach to ab-initio solid-state materials using a two-stage pruning strategy to efficiently optimize large configuration spaces in periodic systems. The method is demonstrated on benchmark calculations for 1D hydrogen chains, 2D graphene, and 3D silicon, showing competitive accuracy and improved computational efficiency.

Abstract

We extend the neural network backflow (NNBF) approach to ab-initio solid-state materials using a two-stage pruning strategy to efficiently optimize large configuration spaces in periodic systems. The method is demonstrated on benchmark calculations for 1D hydrogen chains, 2D graphene, and 3D silicon, showing competitive accuracy and improved computational efficiency.

Abstract

We extend the neural network backflow (NNBF) approach to ab-initio solid-state materials using a two-stage pruning strategy to efficiently optimize large configuration spaces in periodic systems. The method is demonstrated on benchmark calculations for 1D hydrogen chains, 2D graphene, and 3D silicon, showing competitive accuracy and improved computational efficiency.