is it possible to reproduce a very very tiny bandgap~2mev

[njuxyh]

Hi Sir:
i want to reproduce the bandgap ~2meV, which is calculated by PAW pp with VASP,
so i am wondering is it possible to reproduce it,  because at present, i used DZP, and try as many as possible K points,  in on direction i used 101 K along high symmetry line, it seems not work, i think maybe LCAO have not such precision ?
please give me a hint.

thanks very much .

 
 

[Petr Khomyakov]

You may use  LCAO basis set combined with SG15 pseudopotential (or pseudoDojo that will be available in O-2018.06), you will  then have an option to choose between Medium, High and Ultra accuracy for the LCAO basis set functions.

The exact answer to your question depends on the specific system of study, and requires actual calculation for verification of the LCAO basis set accuracy. However, I would like to notice that the DFT-PAW approach is also an approximation to all-electron DFT calculations,  meaning that 2 meV band gap obtained in VASP can be very much within the accuracy of the method. In that sense, one would have to do accurate all-electron calculations for having a reference.

From the physical point of view, band gap of 2 meV is of relevance for ultra-low temperatures < 30 K only. You have to be aware that DFT may significantly underestimate band gaps of semiconductors, or even predict no gap, e.g., for bulk Ge.

Note that FHI pseudopotential and LCAO basis sets associated with it are less accurate than SG15 pseudoptential/LCAO basis sets.

[njuxyh]

You may use  LCAO basis set combined with SG15 pseudopotential (or pseudoDojo that will be available in O-2018.06), you will  then have an option to choose between Medium, High and Ultra accuracy for the LCAO basis set functions.

The exact answer to your question depends on the specific system of study, and requires actual calculation for verification of the LCAO basis set accuracy. However, I would like to notice that the DFT-PAW approach is also an approximation to all-electron DFT calculations,  meaning that 2 meV band gap obtained in VASP can be very much within the accuracy of the method. In that sense, one would have to do accurate all-electron calculations for having a reference.

From the physical point of view, band gap of 2 meV is of relevance for ultra-low temperatures < 30 K only. You have to be aware that DFT may significantly underestimate band gaps of semiconductors, or even predict no gap, e.g., for bulk Ge.

Note that FHI pseudopotential and LCAO basis sets associated with it are less accurate than SG15 pseudoptential/LCAO basis sets.

thank you for your detailed help .

indeed, with DFT, using  LCAO, i get the zero gap. so i have no ideas what to do.

as you mentioned, SG15 is only available in the coming version 2018??

because i have not seen this functional in 2017.12.

the results i want to reproduce is Nanoscale, 2011, 3, 3883,and want to get the bandstructure as figures enclosed.

[mlee]

The SG15 has been implemented in 2017 version. You can check Pseudopotential in Basis set/exchange correlation in New Calculator.

[Petr Khomyakov]

as you mentioned, SG15 is only available in the coming version 2018??

because i have not seen this functional in 2017.12.

SG15 is not a density functional, but pseudopotential, and it is available in 2017.12 (as mentioned by mlee) with three different basis sets Medium, High and Ultra. You should use the same density functional as in the paper.

[Anders Blom]

Although it's possible there could be some details of the calculation that are not share properly in the paper, and you need very high accuracy or some special tricks, I find some parts of this paper unclear. Look at the dispersion close to M: why is  the branching at -1 eV happening not exactly at M but after? Normally the bands branch at the symmetry points... This could indicate an index error in the plotting, which also would affect the interpretation of the bands at K. I would also like to see the points at K, not just the lines.  I am not saying the paper is wrong, but they are making it hard to understand their data, and for such small values as 2 meV one should be very careful. I would like to see e.g. the effect of the k-point sampling which is know to affect the band structure of graphene very much, unless you sample the K point properly, which I'm not sure is done with a 5x5 sampling on a 4x4 or 5x5 supercell.

[Petr Khomyakov]

To add to Anders' post, the k-point sampling issue is discussed in detail in the Appendix of Phys. Rev. B 87, 075414 (2013). Note that the conical point of graphene should be included in the k-grid for accurate calculation of the graphene's band gap.

[njuxyh]

Although it's possible there could be some details of the calculation that are not share properly in the paper, and you need very high accuracy or some special tricks, I find some parts of this paper unclear. Look at the dispersion close to M: why is  the branching at -1 eV happening not exactly at M but after? Normally the bands branch at the symmetry points... This could indicate an index error in the plotting, which also would affect the interpretation of the bands at K. I would also like to see the points at K, not just the lines.  I am not saying the paper is wrong, but they are making it hard to understand their data, and for such small values as 2 meV one should be very careful. I would like to see e.g. the effect of the k-point sampling which is know to affect the band structure of graphene very much, unless you sample the K point properly, which I'm not sure is done with a 5x5 sampling on a 4x4 or 5x5 supercell.

Dear Sir:

thank you for your kindly relpy.

in last few days, i was searching other literature  about the tiny gap issue about Graphene/Mos2. and found a paper  APPLIED PHYSICS LETTERS 105, 031603 (2014), in which they used SIESTA and dzp basis  to model the  heterostructures, and found zero bandgap,so i though the results we used ATK-DFT would be similar to that paper.
on the other hand, i am woundering it is necessary to concern such tiny gap? as  Petr Khomyakov said, 2 meV is of relevance for ultra-low temperatures < 30 K only

thanks again .

[Petr Khomyakov]

Opening a band gap in graphene is an old idea, see Phys. Rev. B 76, 073103 (2007) and references therein, where a graphene sheet was put on the h-BN substrate, and a gap of 50 meV was found in the DFT-LDA calculations. But one needs much larger gap for any practical applications. Also, it is not obvious that opening a sufficiently large gap in graphene would preserve its beneficial electron transport properties.  I do not know what is the-state-of-the-art in that regard, but 2 meV gap seems to be of no relevance at all, in my opinion.