Dear Janos,
the only thing that jumps out is the ecutrho - for ultrasoft I'd go for 8 times the ecutwfc, as a starting point. You mention that you raised it to 5 times (presumably from 4) - this is still too little. Second, try decreasing mixing beta below 0.1 . A few extra bands (but not remotely 20% more) and some boradening might indeed help is there are some level-crossing states in the gap. No other suggestions I could think of. Out of curiosity, you could try the same identical system, and converge it using the CP code - either using damped dynamics on the electrons (very robust, a bit slow) or conjugate-gradients. nicola --------------------------------------------------------------------- Prof Nicola Marzari Department of Materials Science and Engineering 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu