[DFTB-Plus-User] dftb+ 22.2 raman spectra bornderiv not parallel

Auguste TETENOIRE auguste.tetenoire at univ-rennes1.fr
Mon Jul 3 17:17:35 CEST 2023


Good morning,


I am currently using the DFTB+software to make structural calculations 
of gold with a grafted organic molecule, in a periodic cell.
I am specifically interested in the new feature of DFTB+ 22.2 release 
"Born charges and polarizability derivatives from finite difference 
derivatives." for Raman spectra calculation.


My question is:

-It seems that the calculation of the hessian, and born charges are 
parallelized with OpenMP while the calculation of the derivative of the 
born charges is not.Am I right ?


Icheckedperforming them separately, thatthe calculationof the Hessian 
and of the born chargeswereparallelized, and Iconfirmed that they are 
running on all procs.
Then, I found that the calculation of the the derivative of the born 
charges run on only one proc.

As technical information, we installed on our cluster of calculations 
the version 22.2 of DFTB+ with D3 dispersion and time dependent options.


Thank you for your time and consideration.


Best regards.


PS: Here, I give you my dftb_in.hsd file:


Geometry = GenFormat {
   <<< "geom.out.gen"
}


Driver = SecondDerivatives {
   Atoms = 225:-1
   Delta = 1e-4
}

ElectronDynamics = {
   Steps = 200
   TimeStep [fs] = 0.01
   WriteEnergyAndCharges = Yes
   FieldStrength [v/a] = 0.001
   Perturbation = Kick {PolarizationDirection=all}
   Populations = No
   WriteFrequency = 50
   WriteRestart = Yes
   EnvelopeShape = Constant {}
}


Hamiltonian = DFTB {
   charge = 0
   SCC = Yes
   SCCtolerance = 1e-8
   MaxSCCIterations = 500
   shellresolvedSCC = Yes

   SlaterKosterFiles = Type2FileNames {
     prefix = "/home/users/atetenoire/dftb_parameters/"
     Separator = "-"
     Suffix = ".skf"
   }

   MaxAngularMomentum = {
     C = "p"
     O = "p"
     H = "s"
     Au = "d"
   }


   Filling = Fermi {
     Temperature [Kelvin] = 300.0
   }


   KPointsAndWeights = SupercellFolding {
     1 0 0
     0 1 0
     0 0 1
     0 0 0
   }

}


Options = {
   WriteDetailedXML = Yes
   WriteResultsTag = Yes
}


Analysis = {
Polarisability = {} # static enabled by default
PertubDegenTol = 128.0
MaxPerturbIter = 100
PerturbSccTol = 1.0000000000000001E-005
ConvergedPerturb = Yes
WriteEigenvectors = Yes
CalculateForces = Yes
}

Parallel = {
UseOmpThreads = Yes
}


-- 

TETENOIRE Auguste.

Doctor in "Physics of Nanostructures and Advanced Materials" at ISCR Beaulieu - Building 10B - 213 Rennes France.

Website:https://sites.google.com/view/auguste-tetenoire/publications

Contact:
auguste.tetenoire at univ-rennes1.fr
auguste.tetenoire at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20230703/b4eef3cf/attachment.htm>


More information about the DFTB-Plus-User mailing list