---

---

Supplementary data for "Accurate nuclear quantum statistics on machine-learned classical effective potentials"

Scripts, input files, codes and models required to reproduce the results of "I. Zaporozhets, F. Musil, V. Kapil, & C.Clementi (2024). Accurate nuclear quantum statistics on machine-learned classical effective potentials. arXiv: 2407.03448.

Systems

• 0_morse_potential : particle in 1D Morse potential
• 1_h2o_molecule : single H2O molecule in vacuum
• 2_zundel_cation : Zundel cation
• 3_bulk_h2o: Water box (256 molecules)

Setup instructions

1. Setup conda environment:

cd setup
./setup_environment.sh
activate picg

2. Download and install te-pigs with

pip install git+https://github.com/felixmusil/te-pigs.git@v0.1
pip install git+https://github.com/felixmusil/mace.git@develop

4. Download and install i-pi: https://github.com/venkatkapil24/i-pi.git

to use i-pi you will need to setup the environment variables using source $PATH2IPI/env.sh

5. For bulk water only: Download interface for MBPol calculations, MBX-pythonic: https://github.com/venkatkapil24/MBX-pythonic.git and setup enfironmental variable $MBX_HOME, eg, export $MBX_HOME=$HOME/software/MBX-pythonic/

Training data

Training data are available on Zenodo (https://doi.org/10.5281/zenodo.12684727). To use the datasets, first, download and unpack CG_quantum_statistics.zip. Then, copy the dataset/datasets corresponding to the system of interest to the corresnonding training data generation folder (e.g., copy CG_quantum_statistics/1_h2o_molecule/h2o.h5 to 1_h2o_molecule/01_training_data_generation/)

License

The content of this repository is licensed under the CC-BY-SA-4.0 license. See the file LICENSE for details.