ML_FF for water

[suojiang_zhang1]

Hi,
I want to train a accurate FF for water by vasp ML_FF.
I referred to the https://www.vasp.at/wiki/index.php/Best ... Monitoring the INCAR, as follows:

ENCUT = 700 #larger cutoff
LASPH = .True.
GGA = RP
IVDW = 11
ALGO = Normal
LREAL = Auto
ISYM = 0
IBRION = 0
MDALGO = 3
ISIF = 3
POTIM = 1.5
TEBEG = 200
TEEND = 500
LANGEVIN_GAMMA = 10.0 10.0
LANGEVIN_GAMMA_L = 3.0
PMASS = 100
PSTRESS = 0.001
NSW = 20000
POMASS = 8.0 16.0
ML_LMLFF = .TRUE.
ML_ISTART = 0
I constructed a cubic box with a=6.6A and angle=90 degree that is closed to the experimental density.
After training of 20000 steps, I grep ERR ML_LOGFILE, the energy and force both are good, but checked the CONTCAR and found the box is changing the larger with a=8.4A, and density is smaller than the initial POSCAR.
it is very strange?
all of processes are normal.
I do not know where is problem.

[ferenc_karsai]

Did you use an ICONST file?

[suojiang_zhang1]

Yes, I used the ICONST:
LA 1 2 0
LA 1 3 0
LA 2 3 0
LR 1 0
LR 2 0
LR 3 0
S 1 0 0 0 0 0 0
S 0 1 0 0 0 0 0
S 0 0 1 0 0 0 0
S 0 0 0 1 -1 0 0
S 0 0 0 1 0 -1 0
S 0 0 0 0 1 -1 0

[ferenc_karsai]

Ok that's good.
Please upload your POSCAR, POTCAR, INCAR, KPOINTS, OUTCAR and ML_LOGFILE.

[suojiang_zhang1]

The files were attached as reference

[ferenc_karsai]

I looked into your calculation. The number of local reference configurations look ok and the error of your force field is also good.

Concerning the density:
First you use a temperature ramp 200-500 K. You also allow full volume fluctuations (NpT ensemble), so your cell will try to change its volume to the given temperature.
Second you are using a given exchange-correlation functional whose equilibrium structure differs from experiment.