[Pw_forum] SCF converges to different total magnetization values.

[Yury Vishnevskiy]

Dear QE community,
I am doing vc-relax calculations (pw.x version 5.0.1) of copper 
phthalocyanine polymer with gaussian smearing for electrons. When the 
first such calculation finished I've started a second one with the 
latest geometry and cell parameters in order to be sure that they are 
indeed optimal. After the initial SCF procedure the program said that 
the BFGS is converged, did no updates to the geometry and started the 
final SCF as usually. Then I've found that the total energies after the 
first and last (here second) SCF procedures are very different: 
-513.10435104 vs. -513.10131945 Ry. The difference is that the first SCF 
has converged to the total magnetization 0.94 (I have odd number of 
electrons in my system) while the second to 0.00. Why is this happening? 
The part of my input is as follows:

  &SYSTEM
     ibrav = 0,
     celldm(1) = 20.158,
     nat = 33,
     ntyp = 4,
     ecutwfc = 35 , ecutrho = 420,
     london = .true.
     nspin = 2 ,
     occupations = 'smearing' , degauss = 0.02 , smearing = 'gaussian' ,
     starting_magnetization(4) = 0.5,
  /
  &ELECTRONS
     conv_thr=1.0d-8
  /
ATOMIC_SPECIES
     N   14.00700  N.pbe-van_ak.UPF
     H    1.00800  H.pbe-van_ak.UPF
     C   12.01100  C.pbe-van_ak.UPF
    Cu   63.54600  Cu.pbe-n-van_ak.UPF



Anyway, I see that the electronic solution with the total magnetization 
equal almost to 1.0 is lower in energy and want to do calculations with 
this value fixed. I remove gaussian smearing and set 
tot_magnetization=1.0. But with such settings (see part of input below) 
the SCF does not converge. The program in the very start of log file 
says "the system is metallic, specify occupations" (how is decides that 
the system is metallic?) but runs further as usually. How should I 
"specify occupations"? I know from experiment that my object is a 
semiconductor and would like to keep occupations fixed in best case.

  &SYSTEM
     ibrav = 6,
     celldm(1) = 20.158,
     celldm(3) =  0.316328,
     nat = 33,
     ntyp = 4,
     ecutwfc = 35 ,
     ecutrho = 420,
     london = .true.
     nspin = 2 ,
     tot_magnetization = 1.0,
  /
ATOMIC_SPECIES
     N   14.00700  N.pbe-van_ak.UPF
     H    1.00800  H.pbe-van_ak.UPF
     C   12.01100  C.pbe-van_ak.UPF
    Cu   63.54600  Cu.pbe-n-van_ak.UPF
K_POINTS automatic
   8 8 8   0 0 0


Sincerely,
Yura V. Vishnevskiy
Uni-Bielefeld, Germany