Some problems regarding isothermal-isobaric ensembles

icet
1

Hello everyone, I am working with a high-pressure system (~300 GPa) where pressure effects cannot be neglected. The structures used for training have undergone relaxation at a given pressure. By fitting the enthalpy with icet, I have obtained an effective cluster interaction (ECI) to predict enthalpy. However, I am uncertain about the following aspects:

  1. To calculate the enthalpy of the random alloy at a given pressure and temperaute (isothermal-isobaric ensembles), I perform Monte Carlo swaps to change alloy configurations, the Metropolis acceptance criterion typically relies on the potential energy difference. Does this imply that I should train an ECI based on potential energy instead? Or would using enthalpy differences in the Metropolis criterion be a valid approach? I have the same question regarding semi-grand-canonical Monte Carlo simulations.
  2. All training structures have undergone structural relaxation at the same pressure (~300 GPa), resulting in different equilibrium volumes. Based on my understanding, cluster expansion does not explicitly incorporate volume. If I estimate a structure with a low enthalpy using my fitted ECI, does this mean that the enthalpy given by the ECI corresponds to that structure at my training pressure (~300 GPa)?
    Thanking you in advance for your help.
2

Hi,

  1. I would think you should like to use the enthalpy, H=E+PV, in the MC swaps to capture the fact the two structures have different volumes.
  2. Yes, the cluster expansion you train on enthalpys at 300GPa will only work at this pressure as far as I understand. If you give the CE training data for enthaplys at 300GPa, that is what it will predict.