README.md

Data repository for "Mechanistic insight on water dissociation on pristine low-index TiO$_2$ surfaces from machine learning molecular dynamics simulations"

arXiv preprint: https://arxiv.org/abs/2303.07433

MLPs:

Contains machine learning potentials trained on optb88vdw (./optb88vdw/), and the Delta-learning potentials (./delta-pbe/, ./delta-scan-1200Ry/, ./delta-scan-350Ry/). The Delta-learning potentials need to be used together with the optb88vdw MLP.

cp2k-input:

The DFT input files for CP2K.

fig1-data:

Source data for generating Fig.1. Contains the free energy surfaces, water density profiles, and water orientations.

fig2-data:

Source data for generating the kPCA maps in Fig.2.

fig3-data:

Source data for generating the kPCA maps in Fig.3a,b, and the transition matrix in Fig.3d.

metad-input:

The example input files for running metadynamics simulations using LAMMPS and PLUMED.

training-data:

The training data for the optb88vdw MLPs, in N2P2 format. Units are in hartree/Bohr.

delta-PBE-training-data:

The training data for the Delta-learning PBE MLPs, in N2P2 format. Units are in hartree/Bohr.

delta-SCAN-training-data:

The training data for the Delta-learning SCAN MLPs, in N2P2 format. Units are in hartree/Bohr.

example-analysis:

Python script and notebook to generate features for hydrogen environments, perform classification and kernal PCA.