Energy drift in MD with vdW functional

[eariel99]

I find a substantial failure of energy conservation in molecular dynamics NVE when using Van der Waals functional optB88-vdW.
After touching the ENCUT, LREAL, EDIFF, POTIM, I found that changing the functional to the standard PBE the total energy is much better conserved.

Here is the only difference between both calculations

diff INCAR ../INCAR
11c11
< EDIFF = 1E-7 # Candidato a reducir
---
> EDIFF = 1E-9 # Candidato a reducir
19a20,26
>
> #Options for optB88-vdW functional
> GGA = BO
> PARAM1 = 0.1833333333
> PARAM2 = 0.2200000000
> LUSE_VDW = .TRUE.
> AGGAC = 0.0000


The full INCAR is

GGA = BO
PARAM1 = 0.1833333333
PARAM2 = 0.2200000000
LUSE_VDW = .TRUE.
AGGAC = 0.0000


System = IBBAND
ISMEAR = 0
ENCUT=400 # Cambio
SIGMA = 0.1

!DOS
PREC= Normal
#ALGO = Fast
ISTART = 0
!ICHARG = 2
EDIFF = 1E-9 # Candidato a reducir
#ADDGRID= .TRUE.
#LMAXMIX= 4
!ISPIN=1
#LREAL= AUTO # Cambio
LREAL= .FALSE. # Cambio
KSPACING=0.4
LWAVE=.FALSE.
LCHARG=.FALSE.

#Options for optB88-vdW functional
#Options for MD
ISYM=0
ISIF=1
NELMIN=6
#MAXMIX=40 # reuse mixer from one MD step to next
# MAXMIX should be about three times as larger as the number of iterations required to converge the electronic orbitals in the first iteration.
IBRION=0
NSW=200 ; POTIM=0.2 # 200 fs
#SMASS=-3 # Microcanocnical
#SMASS=0.207162 # Ajustado para T=165.03 fs .T prop 1/sqrt/(SMASS)
#TEBEG=100;TEEND=100
NWRITE= 0
NBLOCK=50 # XDATCAR updated every 50 steps

#Parallelism
NPAR=4
ADDGRID=.TRUE.

Thank you,
Eduardo Menendez
University of Chile

[eariel99]

Let me clarify that rather than a drift it is an oscillation in the energy. It may be a drift for longer simulation time.
I have tested that the functional opt86b-vdW do not cause the oscillation or drift. Maybe this is related to this post about "convergence for optB88-vdW calculations"

http://cms.mpi.univie.ac.at/vasp-forum/ ... B88#p14383

Best regards
Eduardo