Hi all,

I am doing a calculation on bulk InAs, and I am using LDA pseudo- potentials downloaded from the SIESTA website -- the ones imported from Abinit. I am using In pseudo-potential with semicore.

After I converged the mesh cut-off (200 Ry) and the kpoints (Monkhorst pack 20*20*16) -- I am using a 4 atom cell (for the diamond structure), hence longer lattice vector in z-direction.

I relaxed the lattice constant variationally, and found it to be far too small: 3.4951 Ang, as compared with the experimental value of 4.2840 Ang. Also the gap found from the DOS plot is about 4eV, which is also way too large.

I am using automatically generated DZP basis for As, with energy shift of 100 meV. The pseudopotential of As gives a ghost at l=3, so I have restricted PS.lmax for As to 2 (d shells.).

For In, since it has semicore, I need to define the basis by hand, and here are the parameters:

PAO.BasisSize       DZP
PAO.EnergyShift     100 meV

%block PS.lmax
In  3
As  2
%endblock  PS.lmax

%block PAO.Basis
In   3
 n=4 2 2 P
     0.00  0.00
     1.00  1.00
 n=5 0 2
     0.00  0.00
     1.00  1.00
 n=5 1 2
     0.00  0.00
     1.00  1.00
%endblock PAO.Basis

I left the cut-off radii to 0.0 so that they will be generated by the code automatically. And I have added a polarisation orbital to the 4d shell.

I thought these parameters were reasonable... but the results show otherwise. Either the pseudo-potential is bad or I am using a bad basis. Please could anyone give me any suggestions in how to improve the calculations?

Thank you very much!

Lianheng

Reply via email to