Discussion: sigma value in NSCF calculation

[Andrew-S-Rosen]

I want to open up a discussion here.

In the NonSCFSetGenerator, the following logic is applied:

atomate2/src/atomate2/vasp/sets/core.py

Lines 305 to 310 in 22b2fa0

	elif self.mode in ("line", "boltztrap"):
	# if line mode or explicit k-points (boltztrap) can't use ISMEAR=-5
	# use small sigma to avoid partial occupancies for small band gap materials
	# use a larger sigma if the material is a metal
	sigma = 0.2 if bandgap == 0 else 0.01
	updates.update({"ISMEAR": 0, "SIGMA": sigma})

For metals, does it make sense to be using SIGMA = 0.2 in combination with ISMEAR = 0? My understanding is that if you're using ISMEAR =0, one should really be using a small SIGMA (e.g. <= 0.05). Should this be SIGMA = 0.2 and ISMEAR = 1 for metals vs. SIGMA = 0.01 and ISMEAR = 0 for non-metals?

I am confident that @utf has opinions on this.