[DFTB-Plus-User] Failure in diagonalisation

[Zack Terranova]

Thank you for the reply Bálint!  That is what we thought too, so we tried switching to SteepestDescent.  It took longer to fail (Geom step 135), but failed nonetheless.  
********************************************************************************
** Geometry step: 135
********************************************************************************

  iSCC Total electronic   Diff electronic      SCC error    
ERROR!
-> Failure in diagonalisation routine zhegvd, non-positive definite overlap! Minor    886 responsible.

Here is some background on how I set this up. I am testing different parameter sets for the same material, TiO2.  I set up all three with the same initial geometries with one water molecule on top, and the other two finished fine.  The Zheng, et al. set seem to be giving me the trouble.  So what I have done is optimized the bare slab of TiO2, which completed successfully, took that relaxed geometry and put a water on top.  This is where it is failing me at steps 53, 82(CG) and 135 (SD).  Any more insight you could give would be greatly appreciated.

Thank you so much for your time!

Here is a copy of my dftb_in.gen:
Geometry = GenFormat {
 <<<"halfmolec.gen"
}

Driver = SteepestDescent {
  MovedAtoms = Range{1 -1}
  OutputPrefix = "geom"
  MaxSteps = 5000
  AppendGeometries = Yes
}

Hamiltonian = DFTB {
  KPointsAndWeights= SupercellFolding {
    5 0 0
    0 5 0
    0 0 1
    0.0 0.0 0.0
}
  SCC = Yes
  SCCTolerance = 1.0e-5
  MaxSCCIterations = 9500
  OrbitalResolvedSCC = Yes
  Mixer = Broyden {
  }
  SlaterKosterFiles = {
 (my home directory)
}
  MaxAngularMomentum = {
    H = "s"
    O = "p"
    Ti = "d"
  }
  Charge = 0.0
}

Options = {}

ParserOptions = {
  ParserVersion = 3


On May 20, 2010, at 8:24 AM, Bálint Aradi wrote:

> Dear Zack,
> 
> On 05/18/2010 03:47 PM, Zack Terranova wrote:
>> I understand that DFTB+ will fail with this error: "ERROR!
>> -> Failure in diagonalisation routine zhegvd, non-positive definite overlap! Minor XXX responsible"
>> when two atoms become too close together, or are on top of each other.  But why are they failing in the middle of the geometry optimization job, for example at Geom step 53 or 82?
> 
> Hard to say, without knowing any details. Keep in mind, that also
> conjugate gradient optimisation can lead to non-physical geometries, as
> it does not optimize along the forces as steepest descent, or an MD with
> simulating anealing would do. If you system is very unstable (with huge
> forces), the best practice is too take steepest descent with a small
> step size, until forces go down enough, so that you can safely use CG.
> 
> Best regards
> 
>  Bálint
> 
> -- 
> Dr. Bálint Aradi
> Bremen Center for Computational Materials Science, University of Bremen
> http://www.bccms.uni-bremen.de/en/~b_aradi
> 
> 
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at dftb-plus.info
> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user