First-Principles Calculations (DFT)

Density functional theory (DFT) calculates the behavior of electrons in molecules and materials from the fundamental equations of quantum mechanics, without relying on empirical parameters. We use DFT to generate the reference data that trains our machine learning force fields, and to study electronic effects that simpler models cannot capture: bond breaking and formation in supercritical water, how ions coordinate with surrounding water molecules, and the weak attractive forces (dispersion interactions) between molecules that hold liquids and solids together.

Reference data generation: Systematic quantum-mechanical calculations to build training datasets for machine learning force fields, covering aqueous systems, electrolytes, and organic molecules

Electronic structure of reactive systems: Studying how chemical bonds break and form in supercritical water environments, where the rearrangement of electrons during reaction makes classical force fields inadequate

Benchmarking: Validating machine learning force fields against high-accuracy ab initio methods (coupled-cluster theory, perturbation theory) that are too expensive for routine use but provide definitive reference values for critical test cases

Solvation thermodynamics: Calculating how much energy is released or absorbed when ions dissolve in water and how water molecules arrange around them, fundamental quantities for understanding electrolyte chemistry

We use two complementary computational approaches: plane-wave DFT codes (VASP, Quantum ESPRESSO) for periodic systems like bulk liquids and solid-liquid interfaces, and localized-basis codes (Gaussian, ORCA) for molecular-level calculations. Both include dispersion corrections (D3, D4, TS), empirical additions that account for the weak long-range attractions between molecules that standard DFT approximations tend to underestimate.

DFT provides the reference data that propagates upward through our simulation hierarchy: DFT → MLFF → All-atom MD → CG-MD/DPD.