[Pw_forum] SCF_Calculation_Zeolite

[Giovanni Cantele]

> On 20 Oct 2015, at 20:55, Mofrad, Amir Mehdi (MU-Student) <amzf5 at mail.missouri.edu> wrote:
> 
> Dear all, 
> 
> I want to do an SCF calculation on sodalite (a kind of zeolite) unit cell (consisting 36 atoms) but I'm having some difficulties to do so. First off, I don't know what the vital and least parameters needed for these types of calculations are. So I used virtual nanolab to generate an input file. Following a tutorial on Quantum Wise website, I manipulated the input files, for instance they said to include nbnd=10, change the ecutwfc=30, etc.However, when I stared running the simulation, I kept getting an error "too few bands". So I started increasing the number by the order of 10 (without having any idea what it is!) and at nbnd=100 my simulations is now being run. I left other variables (such as K_POINTS) unchanged. I would really appreciate it if someone would tell me what should be and should not be in my input file for scf calculations and give me a sense of the range of numbers that I'm allowed to assign to the variables. Here is my input file: 
> 
> 
> &control
>  calculation = 'scf'
>  verbosity = 'high'
>  restart_mode = 'from_scratch'
>  prefix = 'SOD'
>  tstress = .false.
>  tprnfor = .false.
>  pseudo_dir = './'
>  outdir = '/tmp'
> /
> 
> &system
>  ibrav=0
>  nat=36
>  ntyp=2
>  ecutwfc=30
>  ecutrho=300
>  nbnd=100
>  input_dft='PBE'
> /
> 
> &electrons
>  diagonalization='david'
>  mixing_mode = 'plain'
>  mixing_beta = 0.7
>  conv_thr =  1.0d-08
> /
> 
> ATOMIC_SPECIES
>  O 15.999400 O.pbe-kjpaw.UPF
>  Si 28.085500 Si.pbe-rrkj.UPF
> 
> ATOMIC_POSITIONS bohr 
>  O 4.223852 10.209245 16.553687
>  O 12.694549 1.738548 8.082990
>  O 16.194639 10.209245 16.553687
>  O 7.723942 1.738548 8.082990
>  O 16.194639 10.209245 4.582900
>  O 7.723942 1.738548 13.053597
>  O 4.223852 10.209245 4.582900
>  O 12.694549 1.738548 13.053597
>  O 16.194639 4.223852 10.568293
>  O 7.723942 12.694549 19.038991
>  O 16.194639 16.194639 10.568293
>  O 7.723942 7.723942 2.097596
>  O 4.223852 16.194639 10.568293
>  O 12.694549 7.723942 2.097596
>  O 4.223852 4.223852 10.568293
>  O 12.694549 12.694549 2.097596
>  O 10.209245 16.194639 4.582900
>  O 1.738548 7.723942 13.053597
>  O 10.209245 16.194639 16.553687
>  O 1.738548 7.723942 8.082990
>  O 10.209245 4.223852 16.553687
>  O 1.738548 12.694549 8.082990
>  O 10.209245 4.223852 4.582900
>  O 1.738548 12.694549 13.053597
>  Si 5.973897 10.209245 19.038991
>  Si 14.444594 1.738548 10.568293
>  Si 14.444594 10.209245 2.097596
>  Si 5.973897 1.738548 10.568293
>  Si 1.738548 5.973897 10.568293
>  Si 10.209245 14.444594 19.038991
>  Si 1.738548 14.444594 10.568293
>  Si 10.209245 5.973897 2.097596
>  Si 10.209245 1.738548 6.332945
>  Si 1.738548 10.209245 14.803642
>  Si 10.209245 1.738548 14.803642
>  Si 1.738548 10.209245 6.332945
> 
> K_POINTS gamma
> 
> CELL_PARAMETERS bohr
> 17.933187 0.000000 0.000000
> 0.000000 17.933187 0.000000
> 0.000000 0.000000 21.136587                                                                                                                                                                                               
> 
> 
> 
> Best regards,
> 
> Amir M. Mofrad    
> Graduate Research Assistant     
> Chemical Engineering Department
> University of Missouri
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum


12 Si atoms (valence: 4), 24 O atoms (valence: 6) correspond to 4*12 + 6*24 = 192 electrons, that is, 96 filled bands (if your system is semiconducting, so you cannot include less than 96 bands to correctly reproduce the ground state charge density. This is the reason why the first run complained about the number of bands.

Concerning ecutwfc, please follow/read the tutorials on the web site. The wave function cutoff must be chosen so as any desired property, to be computed, is converged with respect to this parameter. Usually, norm conserving pseudo potentials require cutoffs that may be higher to much higher than 30 Ry, but you can easily choose all ultrasoft or PAW pseudo potentials, that  require lower cutoffs. However, in the latter case, consider that you must increase the ecutrho, that by default is set to 4*ecutwfc (exact result for norm conserving pseudo potentials), to values of 8-12 times ecutwfc.

Instead of ibrav=0, because you have a tetragonal unit cell, you can specify ibrav=4 (if I well remember, please see Doc/INPUT_PW.txt+ with a= 17.933187, c=21.136587

If your system is a crystal, so you do expect band dispersion along one or more directions in k space, K_POINTS gamma sampling might be insufficient. Try to use
K_POINTS { automatic } specifying an n x n x m grid . Again the convergence of any property to be computed with respect to the size of the k-point sampling should be carefully checked.

I would never use the input_ft parameter (which is deduced from the pseudo files), unless one needs to override the default value. In your case you have chosen the correct value, but it is not needed.


Hope this helps,

    Giovanni