Total energy changes by several eV at 10^-5 smearing parameter

[jerjorg]

I ran some tests to see how the total energy (the last line where energy(sigma->0) in the OUTCAR) depends on the smearing parameter. For all 12 systems I tested, the total energy changes discontinuously at exactly 10^-5 eV (this is the value of the smearing parameter in the INCAR). You can observe this in the plots I attached. The difference in energy with a smearing parameter greater than 10^-5 eV to a smearing parameter less than 10^-5 eV is several eV for many of the systems I tested. I've also the input files I used to generate the data. I only include one set of input files with a single smearing parameter; the rest of the data in the plots is obtained from the same input files with the only change being the value of the smearing parameter in the INCAR. Is this a bug?

[martin.schlipf]

Thank you for reporting this. I reproduced the issue and will investigate its cause.

[martin.schlipf]

Okay, the issue is identified. In dos.F in the routine DENMP, there is a hard exit if SIGMA < 1e-5. For now you can manually adjust this value if you really want to investigate very small values of sigma. We will discuss internally to decide how to deal with this in the long run.

[jerjorg]

Thanks for looking into this. I'll just limit small smearing parameters to 1e-5 for now.