Total DOS is missing electrons

[xamnidar]

Sometimes the DOS columns in the DOSCAR seem to be missing electrons. Below is one example. There are six electrons with energies around -50 eV, as can be seen from the total DOS in columns 4 and 5. The DOS shown in columns 2 and 3, however, is zero. This calculation was run on vasp.5.3.5 with IBRION = -1 and ISMEAR = -5.

Code: Select all

    -50.272  0.0000E+00  0.0000E+00  0.0000E+00  0.0000E+00
    -50.202  0.0000E+00  0.0000E+00  0.0000E+00  0.0000E+00
    -50.132  0.0000E+00  0.0000E+00  0.2000E+01  0.2000E+01
    -50.062  0.0000E+00  0.0000E+00  0.4000E+01  0.4000E+01
    -49.992  0.0000E+00  0.0000E+00  0.6000E+01  0.6000E+01
    -49.922  0.0000E+00  0.0000E+00  0.6000E+01  0.6000E+01

[admin]

the integrated DOS clearly shows that the electrons are not missing, but that the
energy grid along which the DOS is evaluated and plotted is too coarse to resolve the
dispersion of the bands at this energy. Please increase NEDOS such that the grid points of the DOS
are dense enough to cover this dispersion

[xamnidar]

Ok thanks for your response. If this the intended behavior of vasp, then I would suggest updating the description of the DOSCAR file in the vasp manual. The equation in the manual says that the DOS is calculated as the difference between subsequent values of the integrated DOS. In this case the states would appear in the DOS regardless of how coarse the energy grid is.

[admin]

one can see from the integrated DOS that there must be a narrow peek in this energy range.
alternatively, one can use a different BZ integration method (eg gaussian integration, with an appropriate smearing parameter) to resolve
narrow peaks