DOS mismatched with ICHARG=1 and ICHARG=11 calculations

[carhorse]

Dear Admin,
We are studying electronic structures of Mott-insulator Sm2CuO4 (modeled by 2×2×1 supercell containing 32 O, 8 Cu and 16 Sm atoms) with PBE+U scheme. U value varies from 0 to 8eV, k-points are sampled by 12×12×8 grids and the POSCAR is in the following. To study nearest-neighbor antiferromagnetic states with self-consistent calculation, we can see the metal-insulator transition from the DOS with increasing U value, and the band gap is around 1.27eV with U=8eV and J=0.88eV, which is similar to the experiment value, and the spin moment of Cu atom is about 0.65μB. However, when we keep the CHGCAR file to calculate the DOS (or Band structures) with non-self consistent method, we found and the spin magnetism of Cu atoms is about 0.34μB. More confusingly, we found it is always metallic property. Please kindly give us some comments.

POSCAR
1.00000000000000
7.8754127424485594 -0.0000005176053014 0.0000000000000000
-0.0000005176053014 7.8754127424485594 0.0000000000000000
0.0000000000000000 0.0000000000000000 12.0642549575775728
O Cu Sm
32 8 16
Direct
0.2500000000000000 0.0000000000000000 0.0000259469653823
0.7500000000000000 0.0000000000000000 0.0000259469653823
0.0000000000000000 0.2500000000000000 0.0000259469653823
0.5000000000000000 0.2500000000000000 0.0000259469653823
0.2500000000000000 0.0000000000000000 0.2500671939321513
0.7500000000000000 0.0000000000000000 0.2500671939321513
0.5000000000000000 0.2500000000000000 0.2500671939321513
0.0000000000000000 0.2500000000000000 0.2500671939321513
0.5000000000000000 0.2500000000000000 0.7500026281623094
0.0000000000000000 0.2500000000000000 0.7500026281623094
0.2500000000000000 0.0000000000000000 0.7500026281623094
0.7500000000000000 0.0000000000000000 0.7500026281623094
0.2500000000000000 0.0000000000000000 0.4999971994038258
0.7500000000000000 0.0000000000000000 0.4999971994038258
0.0000000000000000 0.2500000000000000 0.4999971994038258
0.5000000000000000 0.2500000000000000 0.4999971994038258
0.2500000000000000 0.5000000000000000 0.0000259469653823
0.7500000000000000 0.5000000000000000 0.0000259469653823
0.0000000000000000 0.7500000000000000 0.0000259469653823
0.5000000000000000 0.7500000000000000 0.0000259469653823
0.2500000000000000 0.5000000000000000 0.2500671939321513
0.7500000000000000 0.5000000000000000 0.2500671939321513
0.5000000000000000 0.7500000000000000 0.2500671939321513
0.0000000000000000 0.7500000000000000 0.2500671939321513
0.5000000000000000 0.7500000000000000 0.7500026281623094
0.0000000000000000 0.7500000000000000 0.7500026281623094
0.2500000000000000 0.5000000000000000 0.7500026281623094
0.7500000000000000 0.5000000000000000 0.7500026281623094
0.2500000000000000 0.5000000000000000 0.4999971994038258
0.7500000000000000 0.5000000000000000 0.4999971994038258
0.0000000000000000 0.7500000000000000 0.4999971994038258
0.5000000000000000 0.7500000000000000 0.4999971994038258
0.2500000000000000 0.2500000000000000 0.0000096882896442
0.7500000000000000 0.2500000000000000 0.0000096882896442
0.0000000000000000 0.0000000000000000 0.4999908697039460
0.5000000000000000 0.0000000000000000 0.4999908697039460
0.2500000000000000 0.7500000000000000 0.0000096882896442
0.7500000000000000 0.7500000000000000 0.0000096882896442
0.0000000000000000 0.5000000000000000 0.4999908697039460
0.5000000000000000 0.5000000000000000 0.4999908697039460
0.0000000000000000 0.0000000000000000 0.1501394339118112
0.5000000000000000 0.0000000000000000 0.1501394339118112
0.2500000000000000 0.2500000000000000 0.3499585562188336
0.7500000000000000 0.2500000000000000 0.3499585562188336
0.2500000000000000 0.2500000000000000 0.6500914065244953
0.7500000000000000 0.2500000000000000 0.6500914065244953

[admin]

Do you use LMAXMIX=6, necessary for f-electrons? If not then the default value is 2,
and electron densities in PAW spheres, written in the scf run, can considerably differ
from scf electron densities. The discrepancies are particularly large for
LDA+U method.