[DFTB-Plus-User] Spin constants and Hubbard Us parameters for iron

[José Manuel Vásquez Pérez]

Dear Bálint Aradi

Hi, I'm trying to reproduce the magnetic moments of iron clusters
calculated by Köhler in his paper (C. Köhler et al., Chem. Phys., vol. 309,
pp. 23, 2005) but I noticed that the values of spin constants and Hubbard
Us used by him correspond to a different electronic configuration than the
values used by default in DFTB+. When I use the parameters given in the
manual which I think correspond to the configuration 3d74s1 I get very
different magnetic moments than Köhler. When use spin and Hubbard constants
for 3d64s2 (which I think is what Köhler used) my calculated magnetic
moments are then in good agreement with his results except for a few
cluster sizes.

In order to resolve this issues I have some questions:
¿Why are only spin and Hubbards U's constants for 3d74s1 configuration used
in DFTB+, and not for 3d64s2 configuration?
In Köhler thesis there are Hubbards U's and spin constants for LDA and GGA,
but I'm not sure if he is using LDA or GGA parameters for his magnetic
moment calculations. I understand that Slater-Köster files for Iron in the
trans3d set are for GGA, but then, do Slater-Köster files for LDA also
exist? In that the case where can I find the LDA parameters?

As an example, for Fe15 the energies that I get  with different magnetic
moments after relaxing the cluster (starting from the same symmetry than
Köhler), have a minimum at 46 unpaired electrons, but according to the
magnetic moment reported by him the minimum should be at 48.
The following are my DFTB+ energies:

# Unpaired e / Energy (eV)
40 -593.2981
42 -593.6837
44 -594.0011
46 -594.0268
48 -593.9829
50 -592.7165

My hamiltonian is as follows, maybe there is some option that I missed to
fix?

Hamiltonian = DFTB {
  SCC = Yes
  MaxSCCIterations = 50
  Eigensolver = RelativelyRobust {}
  SlaterKosterFiles = Type2FileNames {
    Prefix = "./"
    Separator = "-"
    Suffix = ".skf"
    LowerCaseTypeName = No
  }
  OrbitalResolvedSCC = Yes
  CustomisedHubbards {
#        U4s    U4p    U3d
    Fe = 0.237  0.214  0.447
  }
  MaxAngularMomentum {
    Fe = "d"
  }
  SpinPolarisation = Colinear {
    UnpairedElectrons = 6
    RelaxTotalSpin = Yes
  }
  SpinConstants = {
    Fe = {
#      Wss     Wsp     Wsd     Wps     Wpp     Wpd     Wds     Wdp     Wdd
      -0.015  -0.012  -0.003  -0.012  -0.019  -0.001  -0.003  -0.001  -0.017
    }
#    ShellResolvedSpin = Yes
  }
  Mixer = Broyden {}
}

Thank you in advance,

José Manuel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20190308/b2c5c9e8/attachment.htm>