# Why does the Fermi energy change when doing NSCF at different nbnd values?

While doing NSCF, I decided to rerun it at higher nbnd values to incorporated more unoccupied states while plotting PDOS. But the Fermi energy gets changed with each different nbnd. What is the reason for this?

INPUT FILE:

&CONTROL

calculation='nscf'

title='graphene'
prefix='graphene'
restart_mode='from_scratch'
outdir='../tmp'
wf_collect=.true.
pseudo_dir='~/work/pot/'
/
&SYSTEM
ibrav = 4,
a = 14.58 ,
c= 20,
nat = 3
ntyp = 2
vdw_corr = 'dft-d3'
ecutwfc = 50.0 ,
ecutrho = 250.0 ,
occupations='smearing'
smearing='gaussian'
degauss=0.001
nbnd=30
/
&ELECTRONS
electron_maxstep = 200,
conv_thr = 1.0d-10 ,
/
&IONS
ion_dynamics = 'bfgs'
/

ATOMIC_SPECIES
Na 22.989 Na.pbe-spnl-kjpaw_psl.1.0.0.UPF
S 32.065 S.pbe-n-kjpaw_psl.1.0.0.UPF

ATOMIC_POSITIONS (angstrom)
S        0.3560556497      2.9232871880      0.0000000000
Na       2.0049069168      0.9398232344      0.0000000000
Na      -1.8523938267      1.8291379398      0.0000000000

K_POINTS automatic
6 6 1 0 0 0


Here, I tried nbnd=16 which gave Fermi energy 1.8 eV, and then nbnd=30 which gave 2.8 eV.

Also, the system being optimized is a molecule though this change in Fermi energy was observed in crystals too. I have only tried it in insulators(with band gap) which could be the reason for this since the Fermi energy is not well defined for insulators (Other than saying that it's in the middle of the bandgap). But then how do various papers report the DOS of insulators with F.E. as the origin and compare them?

Update : this problem doesn't exist for scf calculations! repeated the same process by using scf and yielded the same Fermi energy. Would the problem be because initially I ran scf without smearing and nscf with smearing?

• May you should show your input cards between two different nbnd calculations. And are you using QE? – Jack May 1 at 0:48
• @Jack Hi ! I am using QE. Literally the only difference I made in the two nscf calculation was the 'nbnd'. Anyways, I will edit the answer with the input file. – Ashique Lal May 1 at 17:34
• Have you compare the eigenvalues? And what's the order of bandgap if it's a gapped system? – Jack May 2 at 1:00