Issue with Ge PAW atomic datasets

[fudong]

Hello all,

I calculated the band structure of Ge with the GGA-PBE exchange-correlation functional and its corresponding PAW atomic dataset available from http://www.abinit.org/downloads/PAW2/ATOMICDATA/032-ge/Ge.GGA_PBE-JTH.xml. However, not only did the band structure lack a band gap, but the bands were very inconsistent with what is illustrated in Papaconstantopoulos' Handbook of the Band Structure of Elemental Solids (p. 352). To verify that this discrepancy was not due to the exchange-correlation functional itself, but rather the PAW atomic dataset, I calculated the band structures of diamond and silicon. Thankfully, the band structures for both were very consistent with those of Papaconstantopoulos (p. 342, 347).

Because the PAW atomic dataset corresponding with GGA-PBE for Ge produced inaccurate results, I tried recalculating the band structure with that corresponding with LDA-PW. I selected ixc 5 for the LDA-HL exchange-correlation functional, and again, I obtained a band structure that wholly disagreed with Papaconstantopoulos'. Therefore, since I have tried both PAW atomic datasets for Ge and obtained subpar results in both instances, I have reached the conclusion that they require significant improvement.

Attached are my band structures, for reference. An electronic version of Handbook of the Band Structure of Elemental Solids was found on http://download.springer.com/static/pdf/184/bok%253A978-1-4419-8264-3.pdf?originUrl=http%3A%2F%2Flink.springer.com%2Fbook%2F10.1007%2F978-1-4419-8264-3&token2=exp=1435763802~acl=%2Fstatic%2Fpdf%2F184%2Fbok%25253A978-1-4419-8264-3.pdf%3ForiginUrl%3Dhttp%253A%252F%252Flink.springer.com%252Fbook%252F10.1007%252F978-1-4419-8264-3*~hmac=757a025bbc40b5c48f285d30427adb643421ec9b9fc782c2417d04454c02ae47

[jzwanzig]

elemental germanium is a case where it is well-known that standard DFT fails. It is not a case of a poor pseudopotential or PAW dataset, rather, the approximations in Kohn-Sham DFT and the current exchange/correlation functionals simply give a poor description of germanium.

[fudong]

Hello Dr. Zwanziger,

I agree with your point that it is well-known that DFT fails with elemental germanium, but only with regards to its band gap. In particular, if you would direct your attention to the lower valence bands of the graphs for germanium, it is readily apparent that the discrepancy is not only with the band gap. The lowest valence band should be at its lowest point at Gamma 1, but it is not. My Abinit calculation gives X, W, and L lower than Gamma. This is a much more serious error than the issue of the gap. We have done an LAPW calculation that does not have this problem, although it doesn't work for the gap. We suspect that another pseudopotential would fix this problem regarding the low valence bands, although it is doubtful to do anything about the gap.

What is your view?

Fudong Fan