Cell Optimization problem

Emile · Post by **Emile** » Tue May 26, 2015 2:58 pm

Dear abiniters,

I am encountering a little but quite annoying problem.
I am performing a cell optimization within a HCP structure, that provides me relaxed cell parameters with a zeroed pressure (-1E-6 GPa).

When I reuse these cell parameters and perform a static ground state calculation (just turn off the optcell value and related parameters), I get a non negligible pressure (about -1 GPa)

I thought it could be due to Pulay stress or inaccuracies, but the initial cell values are very closed to the final ones and I increased k-point grid and cutoff energy without success.

Can anyone have a suggestion to fix this?

Here is the cell relaxation input file:

Code: Select all

# Ni Primitive Cell 

autoparal 1

prtdos 0
prtden 0
prtwf  0

#Definition of the relaxation parameters
optcell 2            # kind of cell relaxation
tolmxf  1.0d-6       # cell convergence criteria
dilatmx 1.1          # used to avoid Pulay stress
ecutsm  0.1          # used to avoid Pulay stress
strfact 10000.0      # stress forces x 100 (accelerate cell relaxation)
ionmov  2            # kind of atom displacement
ntime   200          # number of steps


#Definition of the unit cell
charge   0.0
acell    2*4.7202993505  7.8814711929
angdeg   90.0  90.0  120.0

#Definition of the atom types
ntypat 1            # There is only one type of atom
znucl 28            # The keyword "znucl" refers to the atomic number of the 
                    # possible type(s) of atom. The pseudopotential(s) 
                    # mentioned in the "files" file must correspond
                    # to the type(s) of atom. Here, the only type is Hydrogen.
                         

#Definition of the atoms
natom 2           
typat 2*1  
xred             
#Ni
    0.0 0.0 0.0
    1/3 2/3 1/2

#Definition of the planewave basis set
ecut  60.0         # Maximal plane-wave kinetic energy cut-off, in Hartree

#Definition of the k-point grid
kptopt 1
ngkpt  24 24 16      # Only one k point is needed for isolated system,
                     # taken by default to be 0.0 0.0 0.0

nshiftk 1
shiftk
  0.0 0.0 0.5

occopt 3
nband  20
tsmear 0.001

#Definition of the SCF procedurei
nstep  100        # Maximal number of SCF cycles
tolvrs 1.0d-20    # Will stop when, twice in a row, the difference 
#diemac 1.0       # Although this is not mandatory, it is worth to
                  # precondition the SCF cycle. The model dielectric
#diemix 0.5       # function used as the standard preconditioner
                  # is described in the "dielng" input variable section.
                  # Here, we follow the prescriptions for molecules 
                  # in a big box

Optimized cell parameters : 4.7204534418 7.8790390028

jzwanzig · Post by **jzwanzig** » Thu May 28, 2015 8:28 pm

I ran your input file using PAW for nickel--

some comments:
at least for the PAW set I used, the valence space was 18 electrons, so 36 in your unit cell. Thus 20 bands worked but barely, I increased it to nband 32, which worked better and gave some empty bands even after smearing (which usually helps with convergence).

secondly, this system is magnetic, so I ran it with nsppol 2.

third, after optimization I also got 1E-6 for stress, and then re-running the new cell I got 1E-2. This is normal and indicates usually that you have to go to pretty dense k point sampling (typical of stress and very typical for metals). You say you got -1 upon re-running, that shows a real problem, maybe because you didn't treat it as magnetic? Anyway, from 1E-6 to 1E-2 is not atypical but shows that very tight convergence parameters are needed if you need to sample very small stress changes.

Emile · Post by **Emile** » Mon Jun 01, 2015 10:24 am

Dear Pr. Zwanziger,

Thank you for having taking the time to work on my problem.

First of all, I want to precise that I use a 10 electron GGA pseudopotential for Ni, so 10 bands per atom was fine, I increased this number without seeing improvement so I came back to that initial number.

I am not really interested by magnetic properties, I know it can affect the cell parameter but if I neglect it in the cell optimization run and in the static run, well, I suppose I should not see any effect. Actually, I neglected it in HCP and FCC phase also, and as long as I do not have SCF convergence issue, I am ok with this.

These calculations are starting points for a study within a warmed electronic subsystem, and tsmear is both a convergence parameter and the electronic temperature parameter, so it's use is ambiguous to me, and the most important is to get the lowest pressure for a low tsmear value.

So, according to these points and from what you tell concerning these stress differences, my best option is to work with a much denser k-point grid. For the FCC cell I worked with a 32x32x32 kpoint grid, got converged up to 1E-6 GPa and static calculation provide something like 1E-2 GPa, this is ok. Maybe I had a too low reciprocal mesh with the HCP structure.

ABINIT Discussion Forums

Cell Optimization problem

Cell Optimization problem

Re: Cell Optimization problem

Re: Cell Optimization problem