Hi,
I have some questions regarding the parameters for AIMD and MLFF simulations.
After relaxing my initial structure and confirming its stability at 0K, I set SMASS=-1 to preheat the system from 0K to 300K. I would like to know how to set other parameters such as ISIF (2 or 3?) after setting SMASS=-1. Additionally, do I need to set a thermostat?
Regarding the Langevin thermostat, how do I choose Langevin_gamma, Langevin_gamma_L, and PMASS?
If I trained an MLFF system containing H elements and set the POMASS of H to 8, can I use this force field to run production simulations on the same structure using POMASS = 1 for hydrogen?
Thank you for your help.
Questions about parameters in AIMD and MLFF
Moderators: Global Moderator, Moderator
-
jun_yin2
- Newbie
- Posts: 25
- Joined: Wed Jul 20, 2022 7:18 am
-
jonathan_lahnsteiner2
- Global Moderator
- Posts: 215
- Joined: Fri Jul 01, 2022 2:17 pm
Re: Questions about parameters in AIMD and MLFF
Dear jun_yin2,
With the ISIF tag you can decide if you want do allow the box to change the shape. For more information check the
vasp wiki https://www.vasp.at/wiki/index.php/ISIF.
To use variable cell shape with ISIF=3 during a heating run you have to switch to the NPT ensemble.
You can find more information and how to switch on this ensemble on the wiki https://www.vasp.at/wiki/index.php/NpT_ensemble.
Regarding the LANGEVIN_GAMMA, Langevin_gamma_L. These parameters have units of inverse time. One usually chooses those
by setting them to the lowest frequency mode in the simulated system. But in practice this is often done by a trial and error procedure,
where you try different sets of parameters and check the stability of your temperature and pressure. The same applies to PMASS.
To your last question. In principle you can use the force field if you are only interested in the configurational part of the partition function. But you can not
rely on the hydrogen dynamics predicted by such a force field. The safest way would be to train a second force field with the original hydrogen masses and
compare the results that you are interested for both force fields.
I hope this helps
All the best Jonathan
With the ISIF tag you can decide if you want do allow the box to change the shape. For more information check the
vasp wiki https://www.vasp.at/wiki/index.php/ISIF.
To use variable cell shape with ISIF=3 during a heating run you have to switch to the NPT ensemble.
You can find more information and how to switch on this ensemble on the wiki https://www.vasp.at/wiki/index.php/NpT_ensemble.
Regarding the LANGEVIN_GAMMA, Langevin_gamma_L. These parameters have units of inverse time. One usually chooses those
by setting them to the lowest frequency mode in the simulated system. But in practice this is often done by a trial and error procedure,
where you try different sets of parameters and check the stability of your temperature and pressure. The same applies to PMASS.
To your last question. In principle you can use the force field if you are only interested in the configurational part of the partition function. But you can not
rely on the hydrogen dynamics predicted by such a force field. The safest way would be to train a second force field with the original hydrogen masses and
compare the results that you are interested for both force fields.
I hope this helps
All the best Jonathan