[DFTB-Plus-User] Cannot reproduce published geometries

[Ben Hourahine]

Hello Hannah,

I made a couple of quick tests with your input, what may be the cause
of the difference is that you are using a non-orbitally resolved
Hamiltonian, which is often not appropriate for most d orbital systems.
You'd need to add

OrbitalResolvedSCC = Yes

in the Hamiltonian block (check dftb_pin.hsd as to where this is in the
file) of dftb_in.hsd, which I ran to give 1.734 AA:

> cat geom.gen
    2  C
 Ti  H
    1 1      0.00000000      0.00000000      0.00304995
    2 2      0.00000000      0.00000000      1.73695005

The authors don't discuss whether they use an orbitally resolved
Hamiltonian, so this may not be the correct explanation.

A slight complication is that the results in the paper were generated
using the Gaussian implementation of DFTB which uses integer electron
filling, unlike DFTB+ which uses a Fermi distribution (by default of
around 10-8 a.u.), hence the two codes could in principle produce
slightly different electronic ground states.

I did manage to reproduce the exact geometry of 1.7400000000 AA by
using an excessively large electronic temperature (so a self consistent
solution but nothing like the ground state) and making a sign error
when adding up the geometry:

> cat geom.gen
    2  C
 Ti  H
    1 1      0.00000000      0.00000000     -0.60828328
    2 2      0.00000000      0.00000000      2.34828328

i.e., 2.34828328-0.60828328 = 1.74000000, and I find the same geometry
also when using OrbitalResolvedSCC = No too. This suggests an
alternative explanation that there might be a mistake in data given in
the paper.

Unfortunately, the paper doesn't contain enough information to know
exactly what the ground state they found for this case was, so I would
suggest contacting the corresponding author.

Regards

Ben

Hannah Fox wrote:
> Hello,
> 
> I am trying to reproduce some geometries using trans3d parameters in the JTCT
> (3, 1357)
> 2007 paper (that is quoted in the skf files).
> For TiH, for example, I take the optimised geometry from the paper, run a static
> calculation (SCC, spin polarised), and find that the H-Ti distance is 1.68A,
> instead of the reported 1.74A.
> Below is my dftb_in.hsd.
> Can anybody see if I'm doing anything wrong?
> 
> Thanks,
> Hannah Fox
> 
> 
> 
> Geometry = GenFormat {
>  <<<"in.gen"
> }
> 
> Driver = ConjugateGradient {
>   MovedAtoms = Range{1 -1}
>   MaxSteps = 1000
>   MaxForceComponent = 1.000000000000000E-006
>   OutputPrefix = geom
>   AppendGeometries = Yes
> }
> 
> Hamiltonian = DFTB {
>   SpinPolarisation = Colinear {
>     UnpairedElectrons = 1
>   }
>   SpinConstants = {
>     Ti = {
>     -0.014 -0.012 -0.004 -0.012 -0.014 -0.001 -0.004 -0.001 -0.014
>     }
>     H = {
>     -0.072
>     }
>   }
>   SCC = Yes
>   SCCTolerance = 1.0e-5
>   MaxSCCIterations = 100
>   Mixer = Broyden {
>   }
>   SlaterKosterFiles = {
>     H-H = "/net/maul/Volumes/Corcelli/hfox1/dftb/H-H.skf"
>     H-Ti = "/net/maul/Volumes/Corcelli/hfox1/dftb/H-Ti.skf"
>     Ti-H = "/net/maul/Volumes/Corcelli/hfox1/dftb/Ti-H.skf"
>     Ti-Ti = "/net/maul/Volumes/Corcelli/hfox1/dftb/Ti-Ti.skf"
>   }
>   MaxAngularMomentum = {
>     H = "s"
>     Ti = "d"
>   }
>   Charge = 0.0
> }
> 
> Options = {
>   WriteDetailedXML = Yes
> }
> 
> ParserOptions = {
>   ParserVersion = 3
> }
> 
> 
> 
> 
> 
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at dftb-plus.info
> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user

-- 
     Dr. B. Hourahine, SUPA, Department of Physics,
   University of Strathclyde, John Anderson Building,
          107 Rottenrow, Glasgow G4 0NG, UK.
    +44 141 548 2325, benjamin.hourahine at strath.ac.uk
   The University of Strathclyde is a charitable body,
registered in Scotland, with registration number SC015263