[DFTB-Plus-User] DFTB xTB-GFN2

[Sebastian Ehlert]

Hi James,

this doesn't sound right. I just gave it a try taking the water dimer
from the WATER27 benchmark set, but couldn't really reproduce this
behaviour with any of my GFN2-xTB implementations:

implementation     | E(dimer)/H     | E(monomer)/H  | ΔE / kcal/mol
------------------ | -------------- | ------------- | -------------
GFN2-xTB in xtb    | -10.1487623509 | -5.0704513546 | +4.93
GFN2-xTB in tblite | -10.1487624308 | -5.0704513658 | +4.93
GFN2-xTB in DTFB+  | -10.1487623437 | -5.0704513509 | +4.93

Could you share an input which is able to reproduce this spurious
behaviour?

All the best,
Sebastian


On Fri, 2021-07-16 at 14:32 +0000, James A Charles wrote:
> Hi,
> 
> I recently saw that you were adding the GFN2 tight binding model to
> DFTB+ and was quite excited to try it. I did some basic tests with
> water dimers and trimers and got strange results. Perhaps there is a
> bug? 
> 
> Here is for the dimer but similar for the trimer, GFN2 gives a
> negative, unphysical, binding energy : 
> 
> model
> Calc. binding energy [kcal/mol]
> Ref. binding energy* [kcal/mol]
> Relative diff. %
> mio
> 4.99247
> 4.99
> 0.05
> 3ob3-1
> 4.59272
> 4.99
> -7.96
> GFN1
> 4.80922
> 4.99
> -3.62
> GFN2
> -53.46331
> 4.99
> -1171
> 
> *J. Phys. Chem. B 2005, 109, 33, 15677–15683
> 
> 
> Is there anything I'm missing? 
> 
> James Charles, PhD
> Senior Research Associate
> School of Electrical and Computer Engineering 
> Purdue University 
> 465 Northwestern Ave, 
> West Lafayette, IN 47907
> _______________________________________________
> 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