questions on parametes in DOS analysis

[luke419]

I'd like to know keywords for the analysis of DOS in detail.

1. Should I always use NPAR = 1 for DOS analysis? Is there any case of DOS analysis in which NPAR = 1 is not required?

2. If I analyze semiconducting materials, what parameters should I use for ISMEAR and SIGMA commonly?
How about SIGMA = 0.2 for semiconductor or insulator? It is supposed that narrow sigma (e.g., 0.1) should require much more computation costs.

3. It seems that ISMEAR -5 requires much computational costs but gives more accurate result. Is it right?
In what case is the use of "ISMEAR -5" required espeically?
Can I use it generally for any material(e.g. semiconductor, insulator etc.)?

4. When should I assign the parameters of NBANDS and IMIN, IMAX specifically?


With best regards,


Luke

[pkroll]

Many of these questions are answered at least in parts in the manual.
However, you are always advised to gain experience yourself, rather than to rely on some written statements.
So go ahead and test: try out all what you want to do and analyze the differences. This way you may even learn to understand WHY certain parameters have to be chosen as the are.

[panda]

I generally vary sigma decreasing from 0.5 to 0.05 or until I reach required force accuracy.

ISMEAR = -5 ISMEAR = 0 ISMEAR = 1 are always tested in my simulations, as are all of the pseudopotentials

I agree that it is best to just keep a well documented notebook and test out many cases and analyze for each case to really understand what is going on rather than just selecting the same parameters for every run and assuming you are doing things correctly