Relaxation energy

Hello everyone, I hope this message finds you well. I have a question regarding molecular dynamics simulations. If I were to relax a unit cell and then replicate it to create a supercell, would it be necessary to perform another relaxation of the supercell? And if so, would the energy of the relaxed supercell be lower or higher compared to the unit cell? Thank you in advance for your kind assistance.


I would say no, it is not necessary to relax again, since the inter-atomic distances are the same within the supercell and within the unitcell with periodic boundary conditions (note that this may not be completely true in the extreme limit where the unitcell is much smaller than the cut-off distance of the interaction potential…)

In any case, if you have some doubts, it is easy to run a short test simulation and see for yourself.


That depends a bit on the size of the original cell, its contents and the number of replications.

In general, I would do an additional relaxation, since a larger system can accommodate vibrational modes of lower frequency and thus could find pathways to relax that are not accessible to the smaller cell. But the number of steps required for relaxation or equilibration should be much reduced, since you already did the bulk of the work in the smaller cell.

If your system is liquid, then an additional equilibration is required, since you want the system to “forget” that it was created from replication and thus need to run until the unphysical correlations due to the replication have been eliminated. That can sometimes take longer than what people expect.

1 Like

Thank you very much for your help,

Thank you so much for taking the time to respond. I was a little bit surprised when I found a low energy state with the supercell, but your explanation makes sense. A larger system can have more degrees of freedom, which could certainly result in new relaxed states being discovered.