Struggling with relaxation of PbS slab (triclinic unit cell)

Running Abinit Ground state
1

Hi

I have been struggling to get this run to converge. There is vacuum in the y [0 1 0] direction. The unit cell is triclinic. It seems like the following is working now though, thoughts?
acell 7.31929 42.88705 4.22579 angstrom
angdeg 78.6345 90.0000 99.8266
chksymtnons 0

pseudos “H.psp8, O.psp8, S.psp8, Pb.psp8”

natom 30
xred
0.82715 0.49383 0.08333
0.32715 0.49383 0.58333
0.17284 0.00617 0.41667
0.67283 0.00617 0.91667
0.85082 0.51749 0.08333
0.35082 0.51749 0.58333
0.14917 0.98251 0.41667
0.64917 0.98251 0.91667
0.04166 0.20833 0.08333
0.79166 0.45833 0.08333
0.62500 0.12500 0.25000
0.37499 0.37500 0.25000
0.20833 0.04167 0.41667
0.95833 0.29167 0.41667
0.54166 0.20833 0.58333
0.29166 0.45833 0.58333
0.12500 0.12500 0.75000
0.87499 0.37500 0.75000
0.70833 0.04167 0.91667
0.45833 0.29167 0.91667
0.33333 0.16667 0.16667
0.08332 0.41667 0.16667
0.91667 0.08333 0.33333
0.66666 0.33333 0.33333
0.24999 0.25000 0.50000
0.83333 0.16667 0.66667
0.58333 0.41667 0.66667
0.41666 0.08333 0.83333
0.16666 0.33333 0.83333
0.74999 0.25000 0.00000
ntypat 4
znucl 1 8 16 82
typat 42 41 124 103

kptopt 0
nkpt 1
kpt 0.0 0.0 0.0

occopt 3 # smearing to help convergence
tsmear 0.01
nstep 150

iscf 17
diemix 0.15 # Added cause didnt converge earlier
npulayit 7

ecut 30
ecutsm 0.5

nsppol 1 #non-spin-polarization
optcell 0
ionmov 2
ntime 200

#natfix 2 #because slab
#iatfix 13 19 #Bottom two Pb atoms
strtarget 0 1 0 0 0 0 #fix b direction

tolmxf 5.0d-5
tolvrs 1.0d-12

2

What is not converging? The SCF or the relaxation?

  • you should probably use occopt 4,5 or 7. This Fermi temperature is insane and may distort things
  • Try iprcel 45, or at least modify diemac, to improve preconditioning and SCF convergence
  • the strtarget is expressed in Ha/Bohr**3, you want a stress 22 of 1 Ha/Bohr**3 = 29421.033 GPa? This is not a logical flag. You should also look at the different constraint options, e.g. the other values of optcell (requires 1 axis orthogonal to the other 2 - you only have 1 angle of 90 degrees, which is between a and c!)
  • You have few significant digits for the xred. Either up them or use tolsym to make sure you have the right symmetry.
  • You may well need some more k points, esp in the plane.
  • If you say it is working, which thoughts are you looking for?
3

Thank you for your response. The SCF converges but so far not the relaxation. I was looking for general advice as I am new to DFT and Abinit - your comment is very helpful!

4

So you have tried the stuff above and the relaxation is not converging the structure with ionmov 2 (you might try 14 as well)? Are the forces/stress decreasing at least? You might try to pre-relax with a universal machine learning potential (Mattersim, ESEN…)