[ADF-LIST] Calculation of NMR Parameters

Erik van Lenthe vanlenthe at scm.com
Thu Feb 18 10:54:46 CET 2021

Dear Michael,

Thanks for the bug report.
The failing convergence of the NMR calculation is due to the auxiliary 
fit used in calculation of Hartree-Fock exchange like integrals needed 
in the GIAO formalism that is used in NMR chemical shielding calculations.
Workaround for this molecule is to use a different Quality than Quality 
Good for the RIHartreeFock method in the ADF calculation, include:
RIHartreeFock
   Quality Normal
End
or
RIHartreeFock
   Quality VeryGood
End

or to use the old RI Hartree-Fock method (that was used in ADF2016):
RIHartreeFock
   Useme False
End

Best regards,
Erik
SCM

> Thank you Erik for your reply. I have attached an input of a related 
> smaller compound. The NMR calculation also fails to converge in this 
> case, but the fluctuations are much smaller than in the original 
> compound.
>
> Again, I appreciate all suggestions!
>
> Best regards,
> Michael
>
>
>
>
>> On 16. Feb 2021, at 13:18, Erik van Lenthe <vanlenthe at scm.com 
>> <mailto:vanlenthe at scm.com>> wrote:
>>
>> Dear Michael Patzschke,
>>
>> If it is not a secret Th(IV) complex could you send the full input of 
>> the calculation to me or to support at scm.com ?
>> There have been several changes between ADF2016 and ADF2020.
>> At the moment I have no idea which of these changes are responsible 
>> for why you get such bad convergence in the NMR part.
>>
>> Best regards,
>> Erik van Lenthe
>> SCM
>>> Hello,
>>>
>>> I have a closed shell Th(IV) complex that I try to simulate the NMR 
>>> spectrum for. The calculation is not at all converging. To see the 
>>> influence of relativistic effects I tried the calculation both 
>>> non-relativistically, scalar-relativistically and including SO 
>>> coupling.
>>>
>>> All calculations were done with the PBE0 functional and a TZ2P basis 
>>> set for all atoms, no core and a “good” integration grid and COSMO 
>>> for THF as a solvent. The SCF goes through fine, the NMR part starts 
>>> and then utterly fails to converge. I have calculated similar 
>>> systems in the past with ADF2016 and convergence was achieved within 
>>> 6-10 iteration steps. Now the calculations have been going through 
>>> more than 90 steps and show no sign of convergence.
>>>
>>> Are there any new parameters that have to be set in ADF2020 to 
>>> achieve better convergence? The structure was optimised scalar 
>>> relativistically at the same level. I am at a loss of what to do and 
>>> I’d appreciate any advise on the matter.
>>>
>>> This is the relevant part of the input file:
>>>
>>> Engine ADF
>>>     Basis
>>>         Type TZ2P
>>>         Core None
>>>     End
>>>     Save TAPE10
>>>     XC
>>>         Hybrid PBE0
>>>     End
>>>     Solvation
>>>         Surf Delley
>>>         Solv name=Tetrahydrofuran cav0=0.0 cav1=0.0067639
>>>         Charged method=CONJ
>>>         C-Mat POT
>>>         SCF VAR ALL
>>>         CSMRSP
>>>     End
>>> NumericalQuality Good
>>> EndEngine
>>> eor
>>>
>>> # =======
>>> # NMR NMR
>>> # =======
>>>
>>> cp "${AMS_RESULTSDIR-${AMS_JOBNAME-ams}.results}/adf.rkf" TAPE21
>>> cp "${AMS_RESULTSDIR-${AMS_JOBNAME-ams}.results}/TAPE10" TAPE10
>>>
>>> "$AMSBIN/nmr" << eor
>>> NMR
>>>     out    iso
>>>     adfgui
>>>     Atoms 102 104 106 108 110 112 114 116 118 120
>>>     Atoms 122 124 126 128 130 132 134 136 138 140
>>>     Atoms 50 52 54 56 58 60 62 64 66 68
>>>     Atoms 70 72 74 76 78 80 82 84 86 88
>>>     Atoms 90 92 94 96 98 100
>>>     u1k best
>>>     calc all
>>> End
>>>
>>> eor
>>>
>>> Best regards,
>>> Michael
>>>
>>> _______________________________________________
>>> ADFlist mailing list
>>> ADFlist at scm.com
>>> https://lists.scm.com/mailman/listinfo/adflist
>>
>
>
> _______________________________________________
> ADFlist mailing list
> ADFlist at scm.com
> https://lists.scm.com/mailman/listinfo/adflist

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scm.com/pipermail/adflist/attachments/20210218/bdc2c1cb/attachment.html>


More information about the ADFlist mailing list