ISIF=8 with MDALGO=3 ?

[kirk3141]

Hi

I am training a MLFF using ISIF=8 with MDALGO=3 and it runs during the first temperature interval from T=10 to 50K, but when I use that CONTCAR as a POSCAR and copy the ML_ABN to ML_AB in the next temperature interval from 50 to 100K, I get this error:

positions in direct lattice
velocities in cartesian coordinates
-----------------------------------------------------------------------------
| |
| EEEEEEE RRRRRR RRRRRR OOOOOOO RRRRRR ### ### ### |
| E R R R R O O R R ### ### ### |
| E R R R R O O R R ### ### ### |
| EEEEE RRRRRR RRRRRR O O RRRRRR # # # |
| E R R R R O O R R |
| E R R R R O O R R ### ### ### |
| EEEEEEE R R R R OOOOOOO R R ### ### ### |
| |
| Inconsistent initialization of ion. and lat. DOFs |
| |
| ----> I REFUSE TO CONTINUE WITH THIS SICK JOB ... BYE!!! <---- |
| |
-----------------------------------------------------------------------------

Is ISIF=8 compatible with MDALGO=3?

Thanks!

[ferenc_karsai]

The CONTCAR/POSCAR saves the velocities and predictor/corrector values used by MD. The velocities are a separate block after the ionic positions and the predictor/corrector values are an additional block. Here is also some explanation on the structure of the POSCAR file: https://www.vasp.at/wiki/index.php/POSCAR.

When you use the CONTCAR as a new POSCAR please delete the predictor/corrector values. If that also does not help then delete everything after the ionic positions. That should let you start the MD.

[kirk3141]

Hi

In the CONTCAR removed those predictor/corrector values and indeed I can continue training now from the CONTCAR by copying it to the POSCAR.

However, I noticed though that the values of the axes are not changing at all...they are constant, even though the temperature has increased by hundreds of degrees. This is in sharp contrast with the same problem run using ISIF=3.

Is there something else I need to be doing when using ISIF=8 to allow the volume to change? Is there some issue using it with MDALGO=3? I am using:
POTIM = 1.0
MDALGO = 3 ! Langevin thermostat
LANGEVIN_GAMMA = 3*5 ! friction
LANGEVIN_GAMMA_L = 1*5 ! lattice friction
PMASS = 10 ! lattice mass

Thanks much
Kirk

[kirk3141]

Oh and also I am using VASP v6.5.0 .

[ferenc_karsai]

I have never tried ISIF=8 with MD.

I think the solution for you is to use ISIF=3 and the ICONST file (https://www.vasp.at/wiki/index.php/ICONST).

You can use that file to fix a given shape of the cell (for example the lattice angles and the ratios of the lengths of lattice constants) while being able to allow volume fluctuations.

There are examples for cubic and orthorombic cells on the wiki page.

[kirk3141]

Hi

That is a really good point. I implemented that and it seems to be working fine.

Thanks so much for your suggestion!!

Cheers
Kirk