Missing DOS peak

[hhe]

Dear VASP developers and users,

I am facing one problem in calculating DOS for a system. In the calculated band structure, one can clearly see a band near -0.5 eV. However in the calculated DOS, there is no peak at ~-0.5 eV. An even more interesting thing is that although there is no peak in DOS, there is a sudden increment of one electron in the accumulated DOS (total number of electron).

The output of eigenvalue of energy levels from OUTCAR is as follow:
98 -1.5740 1.00000
99 -1.5739 1.00000
100 -0.5627 0.00000
101 2.3351 0.00000
102 2.4353 0.00000


The DOSCAR is as follows (see the spin-down component, please):
-0.600 0.0000E+00 0.0000E+00 0.1000E+03 0.9900E+02
-0.589 0.0000E+00 0.0000E+00 0.1000E+03 0.9900E+02
-0.577 0.0000E+00 0.0000E+00 0.1000E+03 0.9900E+02
-0.566 0.0000E+00 0.0000E+00 0.1000E+03 0.9900E+02
-0.554 0.0000E+00 0.0000E+00 0.1000E+03 0.1000E+03
-0.543 0.0000E+00 0.0000E+00 0.1000E+03 0.1000E+03
-0.532 0.0000E+00 0.0000E+00 0.1000E+03 0.1000E+03
-0.520 0.0000E+00 0.0000E+00 0.1000E+03 0.1000E+03
-0.509 0.0000E+00 0.0000E+00 0.1000E+03 0.1000E+03


I am using PAW pseudo with proper optimized structure. I have repeated the calculation using different versions of VASP code (VASP4.6 and VASP5.2), but it did not make a difference. The input parameters are as follows:

SYSTEM = A
NWRITE = 2
PREC = A
ISTART = 0
ICHARG = 2
ISPIN = 2
ENCUT = 500

NELM = 90
NELMIN = 8
NELMDL = 10
EDIFF = .1E-07
LREAL = .FALSE.

LREAL = Auto
ISMEAR = -5
LWAVE = T
LCHARG = T
LORBIT=11
EMIN=-20
EMAX=20
NEDOS=3500

Thank you in advance for the help.

Haiying

[hhe]

If I put ISMEAR=0 (Gaussian smearing), the problem is eliminated.