Training Dataset and Machine-Learned Force Field for Spatially Functionalized α-SiO₂ Surfaces

This dataset accompanies the publication "Interfacial Behavior from the Atomic Blueprint: Machine Learning-Guided Design of Spatially Functionalized α-SiO₂ Surfaces", available online: https://doi.org/10.1016/j.jcis.2025.138943 ML_AB: Ab initio molecular dynamics (AIMD) training structures and corresponding DFT energies, forces, and stresses. ML_FF: Trained machine learning force field parameters for α-SiO₂ surfaces functionalized with hydroxyl (OH) and methyl (CH₃) groups. Number of structures: 2330 System: α-SiO₂ (0001) surface with varying OH/CH₃ coverages and spatial arrangements Purpose: Development and validation of a machine-learned force field capable of capturing interfacial energetics, hydrogen-bond networks, and vibrational signatures. DFT calculation parameters: Ab initio molecular dynamics simulations were performed using the Vienna Ab initio Simulation Package (VASP 6.4.3) with the PBE-GGA exchange-correlation functional and projector augmented wave (PAW) pseudopotentials. Dispersion interactions were treated with Grimme’s DFT-D3 correction. A plane-wave energy cutoff of 500 eV and Γ-point sampling were used.