BNN-HC: Bosonic Neural Networks Helium Clusters

Bosonic Neural Networks, or Bosenet for short, is an trial wave function based on neural networks and particullarly adapted for Helium Clusters interacting through the Aziz87 potential[1].

The main results extracted from this version of the algorithm are reported in "Synergy between deep neural networks and the variational Monte Carlo method for small (⁴He_N) clusters", William Freitas and S.A.Vitiello, arXiv:2302.00599

Intallation of requiriments

The code was mostly tested using python3.10 and python3.8. You also should have installed git. We recommend the installation of the requiriments inside a python virtual environment. For more information visit: https://docs.python.org/3/library/venv.html

First, to create the environment use:

python3.10 -m venv ./venv/bnnhc

To activate the environment

source ./venv/bnnhc/bin/activate

The versions specified in the requiriments file are the ones that the tests were performed, change it carefully. To install the required python libraries, execute:

pip install -r requiriments -f https://storage.googleapis.com/jax-releases/jax_releases.html

If you have a GPU, and cuda installed, it is recommended to install jaxlib with cuda support. For instance

pip install jaxlib==0.1.75+cuda11.cudnn82 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

Usage

The example config file is under the scripts directory. To see what kind of parameters you can change, you should look into the input.py file or the bnnhc/base_config.py file. Running the codes looks like:

python3.10 bose.py --config scripts/he02n/input.py
python3.10 vmcbose.py --config scripts/he02n/input.py

The first line executes the optimisation process, while the second uses the optimised wave function to compute estimations of the total, kinetic and potential energy. The outputs are the files train_stats.csv and vmc_stats.csv.

A simple analysis of the data can be done by executing

cd scripts/he02n/
python3.10 ../analysis.py

The outputs are an image called optimisation.png and a text file estimations.out

Acknowledgements

The BNN-HC Ansatz is inspired in the FermiNet[2].

Bibliography

[1] A new determination of the ground state interatomic potential for He₂, Ronald A. Aziz, Frederick R.W. McCourt, and Clement C.K. Wong, Molecular Physics, 1987

[2] FermiNet github, James S. Spencer, David Pfau and FermiNet Contributors, http://github.com/deepmind/ferminet, 2020

Giving Credit

If you want to use this code or your work is based/inspired by this code, please cite the associated paper mentioned in the beginning.