Influence of electronic entropy on Hellmann-Feynman forces in ab initio molecular dynamics with large temperature changes

TL;DR

This paper investigates how electronic entropy affects Hellmann-Feynman forces in ab initio molecular dynamics during large temperature changes, emphasizing the importance of choosing the correct electronic temperature for accurate melting point calculations.

Contribution

It quantifies the impact of electronic temperature choices on melting temperature estimates in ab initio MD simulations using the Z method.

Findings

Melting temperature varies by 200-300 K with electronic temperature choice.

Higher electronic temperature accelerates melting and stabilizes the liquid phase.

Careful selection of electronic temperature is crucial for accurate high-temperature MD simulations.

Abstract

The Z method is a popular atomistic simulation method for determining the melting temperature where a sequence of molecular dynamics runs are carried out to target the lowest system energy where the solid always melts. Homogeneous melting at the limit of critical superheating, Th, is accompanied by a drop in temperature as kinetic energy is converted to potential energy and the equilibrium melting temperature, Tm, can be calculated directly from the liquid state. Implementation of the Z method interfaced with modern ab initio electronic structure packages use Hellmann-Feynman forces to propagate the ions in the microcanonical(NVE) ensemble where the Mermin free energy plus the ionic kinetic energy is conserved. The electronic temperature, Tel, is kept fixed along the trajectory which may introduce some spurious ion-electron interactions in MD runs with large temperature changes such as often seen in homogeneous melting and freezing processes in the NVE ensemble. We estimate systematic errors in the calculated melting temperature to choice of Tel for two main mantle components, SiO2 and CaSiO3 at high pressure. Comparison of the calculated melting temperature from runs where the Tel=Th and Tel=Tm representing reasonable upper and lower boundaries respectively to choice of Tel shows that the difference in melting temperature is 200-300 K for our two test systems. The melting temperature decreases with increasing Tel due to the increasing entropic stabilisation of the liquid and the systems melts typically about 3 times faster in MD runs with Tel = Th compared to runs where Tel = Tm. A careful choice of electron temperature in BOMD simulations where the ions are propagated using Hellmann-Feynamn forces with the Mermin free energy + the ionic kinetic energy being conserved is therefore essential for the critical evaluation of the Z method and in particular at very high temperatures.