[DFTB-Plus-User] Scalapack error in MD run

[Bálint Aradi]

Dear Luca,

the failure you get happens in the Cholesky-decomposition. That usually
means, that your overlap is not positive definite any more. There could
be two reasons for that:

- something is weird with the SK-files you use

- the geometry, you obtain is unphysical, having atoms so close to each
other, that the basis gets linearly dependent (more likely).

Investigate your geometry, whether it is physically meaningful or not.

I also have doubts, whether the 3obw set is supposed to cover your use
casse. It contains a reparametrized repulsive potential in order to
improve on the description of bulk water. This is achieved (if I
remember correctly) by making the repulsive potential rather long, which
according to our experience usually destroys transferability completely.
So, whether it works for anything apart of bulk water reliably, is an
open question.

Best regards,

Bálint


On 20.07.22 12:06, Luca Babetto wrote:
> Dear all,
> 
> I'm running a MD simulation on a urea molecule explicitly solvated with
> 200 water molecules. For this simulation I am using the modified 3OBw
> parameters for water developed in the paper J. Phys. Chem. B 2014, 118,
> 38, 11007–11027. In particular, these modified parameters contain
> Slater-Koster files for O-O, H-H, and O-H pairs. All other parameters
> are taken from the standard 3OB set.
> 
> I am regularly encountering a hard crash due to Scalapack when carrying
> out the simulation on the solvated urea after a few hundred steps. I'm
> using the latest mpi-parallel DFTB+ version on a Ubuntu 20.04 machine.
> The analogous simulation in which only water is present (box with 200
> molecules which only uses the 3OBw parameters) is running normally.
> 
> I don't see any red flag in this simulation, the temperature is
> oscillating around the correct value and the energies do not seem to be
> diverging. Looking at the geo_end.xyz file directly, atoms do not seem
> to be behaving in a weird way.
> 
> I suspect, reading back at other topics in the mailing list, that this
> could be due to a mismatch in the SK files, even though these modified
> 3OBw parameters should not differ significantly from the standard 3OB
> parameters, and indeed the simulation is able to run for a few hundred
> steps before crashing.
> 
> Does anyone have any idea on why this issue is occurring? Let me know if
> I should try something else to diagnose the problem!
> 
> Please find below the .zip with all the input/output files:
> 
> files: https://we.tl/t-3IpuqXehoL <https://we.tl/t-3IpuqXehoL>
> 
> For completeness, here is the dftb_in.hsd file you can also find in the
> archive above:
> 
> Geometry = GenFormat {
>   <<< "urea_solvated.gen"
> }
> 
> Driver = VelocityVerlet{
>   TimeStep [fs] = 1.0
>   Thermostat = NoseHoover {
>     Temperature [Kelvin] = 298
>     CouplingStrength [cm^-1] = 3200
>   }
>   Steps = 4000000
>   MovedAtoms = 1:-1
>   MDRestartFrequency = 1
> }
> 
> Hamiltonian = DFTB {
>   Scc = Yes
>   SlaterKosterFiles = Type2FileNames {
>     Prefix = "3ob/3obw/"
>     Separator = "-"              
>     Suffix = ".skf"            
>   }
>   ThirdOrderFull = Yes
>   HubbardDerivs {
>     C = -0.1492
>     O = -0.1575
>     H = -0.1857
>     N = -0.1535
>   }
>   HCorrection = Damping {
>     Exponent = 4.00
>   }
>   MaxAngularMomentum {
>     C = "p"
>     O = "p"
>     N = "p"
>     H = "s"
>   }
>   kPointsAndWeights = { 0.0 0.0 0.0 1.0 }
> }
> 
> Options {}
> 
> ParserOptions {
>   ParserVersion = 11
> }
> 
> As always, thanks to all the developers for their amazing work on this
> piece of software! :)
> 
> Kind regards
> 
> Luca
> 
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at mailman.zfn.uni-bremen.de
> https://mailman.zfn.uni-bremen.de/cgi-bin/mailman/listinfo/dftb-plus-user


-- 
Dr. Bálint Aradi
Bremen Center for Computational Materials Science, University of Bremen
http://www.bccms.uni-bremen.de/cms/people/b-aradi/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20220720/cf3d1e35/attachment-0001.sig>