dependence of DFTB-results on the input-sequence of heavy atoms

[jgh]

Hi

I just committed a fix for this bug. I hope this is the correct solution to 
the problem.
The new code passes all the cross checks I have with other implementations.
The fix will most likely affect results of your previous calculations.

best regards

Juerg Hutter

Am Freitag, 8. März 2013 06:44:55 UTC+1 schrieb X_Hatakeyama:
>
> With respect to the dependence of DFTB-forces on the input-sequence of 
> atoms,
>
> I found a new problem that "dependence of DFTB-forces" also depends on the 
> input-structure.
>
> For example, when I analyzed dimer of carbon-monoxide (CO),
>
> the dependence of DFTB-forces was changed depending on the distance 
> between two CO.
>
> When two CO were separated with a long distance,
>
> only little dependence was obtained.
>
> While, when two CO were contacted, an apparent dependence was obtained.
>
> At last, DFTB-forces of single CO have showed no dependence on the 
> input-sequence.
>
> 2013年2月13日水曜日 20時36分07秒 UTC+9 X_Hatakeyama:
>>
>> Dear all,
>>
>> I'm trying DFTB-calculation of CP2K for molecules consisted of more than 
>> three heavy atoms (e.g. CO2).
>>
>> I have one problem that the atomic-forces depend on the sequence of heavy 
>> atoms in the input-file.
>>
>> For example, the following input coordinates-(A) and (B) showed different 
>> results at the calculated forces,
>>
>> (A),,,{at angstrom-unit}
>>
>> O  0.0  0.0  +1.2
>>
>> C  0.0   0.0    0.0
>>
>> O   0.0  0.0  -1.2
>>
>> (B),,,
>>
>> C  0.0  0.0   0.0
>>
>> O  0.0  0.0  +1.2
>>
>> O  0.0  0.0  -1.2
>>
>> Could anyone tell me whether the this problem has been occurred or 
>> reported?
>>
>> To be more precise, the above dependency was not resolved by the 
>> following approaches
>>
>> {Periodicity, SCC-calculation and changing compile condition (e.g. 
>> gfortran to ifort within Linux_x86)}.
>>
>> The only solution was to copy&paste the Slater-Koster file of "co.spl" to 
>> "oc.spl",
>>
>> or to focus on hydrogen compounds (e.g. CH4, H2O, etc.).
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20130308/5b40e1eb/attachment.htm>