Equilibrating lattice constant: minimize, fix npt and convergence

LAMMPS LAMMPS General Discussion
,
1

Dear all,
I was thinking about a problem I need to deal with, which is energy minimization to find the equilibrium lattice constant of a solid or a binary alloy.
using fix minimize.
I see that minimization with minimize is uniquely at 0K.
If I want to apply minimization to a binary alloy with a certain composition that is above the miscibility gap only for a determined temperature, then minimize gets stuck in a local minimum and results are unphysical. only above the miscibility gap any permutation of the atom positions is an equilibrium state.

All this to ask: is there a way to minimize the free energy at specific pressure and temperature? Is that way running dynamics with fix npt until the system equilibrates after initializing the system at the desired temperature?

by the way, I noticed that fixed npt doesn’t converge at all for some values of the temperature to the desired pressure and for others it converges very fast. I’m not sure I understand this behaviour.

Thank you,
Mattia Siviero

2

Not really. If you read up on this in a statistical mechanics text book, you should realize that at finite temperature there not a single configuration but a whole ensemble. So all you can do is equilibrate, collect, and do statistical analysis.

It is built in. The method couples fictitious degrees of freedom to your system to adjust the temperature and volume. Their characteristic frequencies are determined by the coupling time. If that matches a frequency of you system, you get feedback and resonances. Ideally you want to couple with a frequency that is in the tail of a wide peak of your system’s spectrum so you do have coupling but only weakly.