Re: Bi2Te3 topological insulator ( No.1 ) 
 Date: 2014/06/19 22:33
 Name: T. Ozaki
 Hi,
Your first three results may be correct, while the result with "Bi10.0 Te9.0" is due to overcompleteness.
Similar calculations by OpenMX have been reported in http://www.pnas.org/content/108/1/24.full http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.110.016403
They, who published the first paper, used 40 quintuplelayers to avoid the surface and surface interaction.
Your system seems to be too thin compared to their system, which may be the reason why you didn't get what you expect.
Regards,
TO

Re: Bi2Te3 topological insulator ( No.2 ) 
 Date: 2014/06/20 18:03
 Name: Artem <artem.pulkin@epfl.ch>
 Dear Taisuke,
Thank you for the references. I understand the problem with interacting surface states. In Quantum Espresso, however, it is enough to have the unit cell as large as I demonstrated: the gap is of order of 1 meV. Instead, with the localized basis set I obtain ~100 meV. I wonder if I can obtain desired result in the same unit cell by introducing ghost atoms instead of increasing the cell size?
As for overcompleteness, I do not get it: the basis size is the same, isn't it? We just change matrix elements and obtain something very different. Or you meant that the initial basis set is too large? Then why does this problem reveal itself only for this combination of basis functions?
Thank you in advance,
Yours faithfully,
Artem

Re: Bi2Te3 topological insulator ( No.3 ) 
 Date: 2014/06/21 10:11
 Name: T. Ozaki
 Hi,
> In Quantum Espresso, however, it is enough to have the unit cell as large as > I demonstrated: the gap is of order of 1 meV.
A VASP calculation shows a band gap of 148 and 36 meV at Gamma point for 2 and 3 QLs, respectively, published in https://journals.aps.org/prb/abstract/10.1103/PhysRevB.81.041307
There seems to be a large difference even among PW codes.
> Instead, with the localized basis set I obtain ~100 meV. I wonder if I can obtain > desired result in the same unit cell by introducing ghost atoms instead of increasing the > cell size?
It is easy to see how the band gap at Gamma point is affected by introducing ghost atoms. Why don't you try it?
> As for overcompleteness, I do not get it: the basis size is the same, isn't it? We just > change matrix elements and obtain something very different. Or you meant that the initial > basis set is too large? Then why does this problem reveal itself only for this combination > of basis functions?
The overcompleteness means a condition that some of the eigenvalues of kdependent overlap matrix become negative. You changed cutoff radii of the basis functions, thus leading to change of the spectrum of overlap matrix. The setting you made for basis functions must be close to the condition, since your basis functions look very rich.
Regards,
TO

