Neural Quantum States for the Body-Centered Cubic Lattice

Code for finding ground states of Heisenberg models on the body-centered cubic lattice using neural quantum states. This repository is built on top of NetKet.

Installation

The repository uses uv to manage dependencies. Download the repository and use the command uv run to run any .py scripts in the correct environment. uv will take care of installing the required dependencies.

Contents

A good place to start is the examples folder, which contains:

• nets: examples of how to use the ViT with netket to perform ground state optimizations on the body-centered cubic lattice.
• vmc: examples of how to use the optimization protocols and expectation value computation scripts defined in the vmc package.

Packages in the repository are contained in src:

• nets: The ViT NQS (for use on 1, 2 and 3-dimensional lattices) and some utility functions.
• vmc: Variational Monte Carlo protocols for performing optimization and calculating expectation values from the NQS, with the use of different symmetrization stages.
• nk_extensions: Extra functionality for dealing with space groups, definitions of the BCC lattice and a few other utilities which extend on NetKet.
• data: Defines the SpinCorr class for working with spin-spin correlation data, including computing static structure factors.

A suite of tests are contained in test.