Closed band gap in semiconductors

[pedro]

I observed that in germanium band gap is closing and VASP predicting a qualitatively wrong metallic ground state. (I used PAW potentials with and without d state as valance in LDA approx.)

Any one could explain why the band gap is closing? And how to "open" it?

Maybe someone had the same problem in other semiconductors?

<span class='smallblacktext'>[ Edited ]</span>

[grassman]

This is a result of the well-known "band gap problem" in DFT due to the approximation of the exchange potential, which causes the underestimation of basically all band gaps (semiconductor and insulator), as well as the prediction of quasi-metallic small-gap SCs (InAs, GaSb, Ge, etc.). You'll find copious amounts of literature on this issue. To "open" it properly (i.e. theoretically rigorous) you need to introduce "higher-level" methods into your calculations. Exact-exchange (hybrid functionals) helps fix the gap problems, or better yet, GW. Unfortunately, those aren't available in VASP till version 5, which I'm sure many of us are anxiously waiting for (I know I am). There are a couple other available codes that already include such methods (e.g. CPMD, CRYSTAL, ABINIT), but they just usually aren't as fast or user-friendly or full-featured as VASP.