SSCHA-MLIP Library

This repository provides a library for the use of the Stochastic Self-Consistent Harmonic Approximation (SSCHA) method coupled with the active learning procedure used to train Moment Tensor Potentials (MTPs) via the MLIP version 2 code. This method reduces the computational load of performing SSCHA calculations. Notably, we have extended support for self-consistent density functional theory (DFT) calculations performed using both the Quantum ESPRESSO (QE) package and the Vienna ab initio simulation package (VASP). This development makes it possible to investigate anharmonic and quantum nuclear effects on a solid’s properties using a wide variety of exchange-correlation functionals, which can optionally be employed to train quantum-accurate MTPs enabling SSCHA calculations at a low computational cost. The reference work associated with this library can be found at https://doi.org/10.1038/s41524-025-01553-1.

Code Structure

The utils_sscha_mlip directory contains essential utilities for managing the interface between SSCHA, MLIP2, and DFT calculations:

• Cluster_Management.py: Handles input-output operations with the computing cluster, as well as the creation and management of inputs and outputs for self-consistent calculations.
• MTP_environment.py: Provides the interface between SSCHA and MLIP version 2, facilitating the active learning procedure.

Scripts and Examples

The scripts folder contains example scripts demonstrating how to use the code:

• SSCHA.py: Implements the main workflow for the automated SSCHA + MLIP protocol.
• potential.mtp: Example of a level 10 Moment Tensor Potential, one of the various levels available in MLIP version 2.
• utils.py: Includes input scripts for first-principles calculations and cluster handling.
• PdCuH_0.dyn: Dynamical matrices for the PdCuH₂ example

To correctly execute the example, place SSCHA.py, potential.mtp, the PdCuH_0.dyn files, the Pseudo folder, and the utils.py file in the same directory. Make sure to edit the utils.py file according to the specific needs of the cluster.

Code Requirements

This repository requires the SSCHA and mlip-2 codes to be installed on the machine where the code is executed, as well as a version of either Quantum Espresso or VASP on the machine used for self-consistent calculations.

• SSCHA: https://github.com/SSCHAcode
• MLIP version 2: https://gitlab.com/ashapeev/mlip-2
• Quantum Espresso: https://github.com/QEF
• VASP: https://www.vasp.at/

The code requires the following libraries in addition to those required by the SSCHA: numpy, sys, os, time, math, contextlib

Installation

• Make sure to have a version of Wheel installed.

• Enter the SSCHA-MLIP folder where the setup.py file is located.

• Execute the following commands:

   python setup.py sdist bdist_wheel
  
   cd dist
  
   pip install utils_sscha_mlip-0.3-py3-none-any.whl