Calculation using Primitive cell or Conventional Cell

[Steven]

Hi everyone,
                         I tried to find the bandstructure of a primitive cell and i got an expected result. But when i tried using a conventional cell of the same crystal, i got a different  bandstructure and the bandgap. Furthermore , the brilioun zone route reduces to only two points G and X. Is it because the number of atoms increased??.Please explain. And which cell should we use (primitive/conventional) for calculation of properties?

[Daniele Stradi]

The electronic band structure of the conventional unit cell differs from that of the primitive because their lattice vectors, and the associated Brillouin zones determined from these lattice vectors, are different. As a consequence, the folding of the electronic bands in the Brilluoin zone is different. This is why you observe two different band structures.

The Brillouin zone reduces to only 2 points in the conventional unit cell because when you transform the cell from conventional to primitive, the lattice type is changed by default to "unit cell". You can select an appropriate lattice type by clicking on the Lattice Parameters... option in the Builder.

Regarding the band gap, are you sure you are using exactly the same parameters for the two calculations?

The unit cell to be used really depends on the type of properties you are interested in.

[Steven]

Thanks a lot

[Steven]

One one question!! i want to optimize a crystal structure by plotting Volume vs Energy and find its equilibrium lattice parameters . do i have to use conventional or primitive.. please guide me

[Daniele Stradi]

The energy vs. Volume curve should not depend on the cell size, but for reason of numerical precision, I suggest that you use the primitive cell. In this way, calculations are faster and you just need to change one lattice parameter:

lattice = FaceCenteredCubic(5.4306*Angstrom)

[Steven]

Thanks
I want to optimize a tetragonal crystal structure which has both lattice parameters 'a' and 'b' . By varying values of 'a' i want to plot the energy vs  volume graph. Do i have to keep c/a ratio constant? if yes, why? please guide.

[Steven]

Should i use the primitive cell for a tetragonal crystal structure during such a structural optimization? please advise

[Jess Wellendorff]

Yes, use the primitive cell, it is computationally most efficient. And yes, if you want an energy vs. volume graph by just changing the value of a, you should fix the c/a ratio. However, in principle, the c/a ratio should also be optimized, so it's a 2D optimization problem. Using the BFGS algorithm to minimize forces and stress would be a more straight forward way to relax the cell.

[Steven]

        Thanks for the reply.
Sir,
                 I built a ZnGa2Te4(Space Group-82) crystal with the atomic positions  as provided in the paper- http://www.sciencedirect.com/science/article/pii/S0025540816306730  . I tried  to optimize the primitive cell using ATK-DFT using GGA-PBE exchange correlation and the pseudopotential SG15 with basis set High  ( also in attached input file).  So i start the optimization  with 'a'=5.72 Angstrom c=13.043810 Angstrom ( expt. values a=5.92 and c=11. .The optimization( Using OPTIMIZE GEOMETRY)  was done without constraining any of the lattice vectors but the bravias lattice type was constrained to preserve the symmetry. It was completed and i got an optimized lattice parameters a= 5.964880 Angstrom and c=13.043810 Angstrom. Though  'a' value is close to the expt. value but 'c' value is quite far . Unexpectedly, the space group of the optimized structure  is 1 . It means the structure has not optimized correctly. So could you please tell me what could be the error?

[Steven]

expt. value of c= 11.8 Angstrom

[Ulrik G. Vej-Hansen]

I agree, 13.04 Å is a rather big deviation compared to 11.8 Å. Looking at your script, I will suggest that you increase the mesh cutoff quite a bit, to at least 100 Hartree, and also check whether it should be higher, perhaps as high as 200 Hartree for some elements. For some systems it makes a difference whether or not Gamma is included in the k-point grid, so you should also check that. The easiest way would be to try with an odd number of k-points.

[Steven]

Thanks for the reply
As  advised ,
      I performed  optimization of the ZnGa2Te4 crystal taking density cut_off as 100 Hartree ( also in attached input file:opt-4) and 200 Hartree ( also in attached input file :opt-5) respectively each for 13*13*13 k_point sampling  and  even included Gamma point for both the optimization. The first optimization (with 100 Hartree) gives   a=6.006947 angstrom and c=12.053359 angstrom, and the latter(with 200 Hartree) gives a=6.006978 angstrom, and c=12.052855 angstrom. The lattice parameter 'a' changes from a= 5.964880 Angstrom as in my previous optimization (earlier posted in the same Subject) even the parameter c also changes. Nevertheless these are far from the experimental values(a=5.92 angstrom and c=11.8 angstrom) . So what can be done to achieve a better optimization. Do basis sets also matters ? I have used SGI5 pseudopotential with the HIGH basis sets.
     Another thing which I have mentioned in my earlier post that the space group changes from 82 to 1 after optimization. Does it should? If yes, then Why it changes?
                Please guide me.

[Ulrik G. Vej-Hansen]

Previously you reported a value for c of 13.04 but now you write 12.05 Å. The latter is a 2% error in the lattice constant, which I think is within expectations for PBE.

I cannot say why the space group changes without looking further at your system, so if you could post also the log-file and the trajectory from the optimization, I can have a look at it.