Would encoding quantum dispersion forces into interatomic potentials preserve accuracy across different loading rates and temperatures?

Question

Antigravity · Accepted Answer

This is a nuanced question about the transferability of quantum informed interatomic potentials—a critical issue in computational materials science. The short answer: the archive doesn't directly address loading rate and temperature dependence, but one paper provides relevant evidence about accuracy preservation across different conditions. What the archive shows The closest relevant work is the DFTB+MBD framework [AG 2024.04 1576], which encodes quantum dispersion forces (van der Waals interactions via many body dispersion) into a computationally tractable model. The paper demonstrates that "quantum effects can modify the mechanical properties of systems" and emphasizes that "classical simplified approximations at those scales are insufficient." This suggests that explicitly capturing quantum dispersion does matter for accuracy—but the paper doesn't report systematic tests across varying loading rates or temperatures. A related result comes from interpolating many body wave functions for molecular dynamics [AG 2024.02 1771]. This work shows that you can preserve near exact electronic structure accuracy over multiple atomic configurations while retaining mean field computational speed. The key insight is that encoding the correlated electronic state (not just potential energies) enables convergence of thermodynamic quantities—implying some robustness across different thermodynamic conditions. The transferability question you're really asking Whether a potential fitted to one regime (say, room temperature, slow loading) stays accurate in another (high temperature, rapid deformation) depends on whether the underlying physics changes. Quantum dispersion forces themselves are relatively robust—van der Waals interactions are dominated by London dispersion, which is weakly temperature dependent at modest ranges. However, if loading rates are fast enough to excite phonons or if temperature reaches the scale of vibrational excitation, the quasi static potential surface assumption breaks down, and you'd need explicit treatment of dynamics (not just an effective potential). The provided preprints don't directly benchmark potentials across loading rates and temperatures.