Estimating potential-dependent physicochemical properties at metal--electrolyte interfaces using machine learning interatomic potentials

This repository contains links to Zenodo records containing smaple scripts and the training and test datasets and all MACE and AIMD trajectories used in this work: " Estimating potential-dependent physicochemical properties at metal--electrolyte interfaces using machine learning interatomic potentials".

Below we provide links to three Zenodo records for this work

1. Estimating potential-dependent physicochemical properties at metal--electrolyte interfaces using machine learning interatomic potentials (Part 1)

• Training and Test Set (.tar) file contains multiple .extxyz files with all Density Functional Theroy (DFT) Single Point Calculation (SPC) data.
• Within Training data contains .extxyz for all sytems with complete SPCs. Within traj_dir, I provided .extxyz used for training multiple iterations of MACE models with different size of training data.
• Test data contains the DFT SPC.xyz files used for comparison against the MACE_model predicted .extxyz. the # next to model#.extxyz is the random seed used to generate respective model.
• MACE models (.tar) folder contains final MACE models using for generating final trajectories. The folder number corresponds to the random seed. We use the MACE_swa_compiled.model for our work.
• Benchmarking data (.tar) contains all AIMD and MACE trajectories used for benchmarking calculations for Au(111), Cu(111), and Rh(111) surfaces.
• Au111_MACE_MD_2ns_trajectories.tar contains all final 2ns MACE-MD trajectories for Au(111) surfaces with different surface charge densities.
• Au111_1NaCl_MACE_MD_1ns_trajectories.tar contains final 1ns MACE-MD trajectories for Au(111) surfaces with 1NaCl for $\sigma=0$.
• Helping files (.tar) contain sample files used for analysis
• Link: https://zenodo.org/uploads/17754595 or 10.5281/zenodo.17754595

2. Estimating potential-dependent physicochemical properties at metal--electrolyte interfaces using machine learning interatomic potentials (Part 2)

• Cu111_MACE_MD_2ns_trajectories.tar contains all final 2ns MACE-MD trajectories for Cu(111) surfaces with different surface charge densities.
• Cu111_1NaCl_MACE_MD_1ns_trajectories.tar contains final 1ns MACE-MD trajectories for Cu(111) surfaces with 1NaCl for $\sigma=0$.
• Link: https://zenodo.org/uploads/17756911 or 10.5281/zenodo.17756911

3. Estimating potential-dependent physicochemical properties at metal--electrolyte interfaces using machine learning interatomic potentials (Part 3)

• Rh111_MACE_MD_2ns_trajectories.tar contains all final 2ns MACE-MD trajectories for Rh(111) surfaces with different surface charge densities.
• Link: https://zenodo.org/uploads/17757050 or 10.5281/zenodo.17757050