Questions about a system of molecules

Gary gary... at mail.mcgill.ca
Mon Aug 15 07:33:04 UTC 2016


Hi Conrad,

Thanks so much for the reply. By retaining the configuration, I do mean 
keeping the symmetry. I currently have a submitted job with the outer atoms 
of the configuration fixed, and I will see the result.

This issue with the field is not only found in the charge -1 case, but also 
in the charge 0 case. In that case, there is not an unpaired charge, but 
the field is once again concentrated on one of the molecules, but this time 
on the upper side molecule. 

And unfortunately, the system I am using only has CP2K 2.2 working. I will 
try and find a way to use more recent versions.

I also tried submitting another job, but with a differet basis set. I have 
seen in the forum that the basis used can have huge effects on the final 
converged result. I am going to try with DVZP-MOLOPT-SR-GTH. Would this 
make sense for a system like this? 

Thanks,

Gary


On Sunday, August 14, 2016 at 4:36:27 AM UTC-7, Conrad wrote:
>
> Hi Gary,
>
> On Thursday, August 11, 2016 at 7:58:08 PM UTC+1, Gary wrote:
>>
>> Hi,
>>
>> I am trying to calculate the electric field above a system of molecules. 
>> The molecules are arranged on the xy plane. I have done the same for single 
>> molecule case, and that is working fine.
>>
>> My question is, how should I optimise the geometry? If it's a system of 
>> molecules, I want to retain the configuration, but relax the system? There 
>> are 4 molecules, and each molecule has 50 atoms, so there are a total of 
>> 200 atoms. Is this too many for geometry optimisation? 
>>
>
> By retaining the configuration, do you mean the symmetry? 
> Is it too many? No. CP2K scales to thousands of atoms. As a first 
> approximation of computer time, budget for a few hundred self-consistent 
> energy calculations' worth. 
>  
>
>>  
>>
> I did a calculation of the EFIELD_CUBE without optimisation, and the 
>> results are a little strange. The field is asymmetrical and seems to 
>> completely concentrate around one of the 4 molecules. The magnitudei of the 
>> field is also very high. I have attached the input and output files. There 
>> is a total charge of -1. 
>>
>
> Have a look at the SPIN DENSITY cube. What you're doing in this input is 
> effectively introducing an excess electron by having a non-zero overall 
> charge. The SPIN DENSITY cube will show you where the unpaired charge 
> (de)localises and may give you some clues. I've never spent any time 
> looking at EFIELD data so I can't comment on this directly. 
>  
>
>> I also want to point out that the SCF run does not converge in the inner 
>> loop (MAX_SCF 800); I am not sure if this is causing non-sensical data.
>>
> This is okay. You might even want to reduce MAX_SCF for the inner loop. 
> When using OT, preconditioning is repeated before running over the inner 
> loop. In principle, shorter inner loops can improve convergence. 
>
> Anyway, this calculation is converged, so no issue there.
>
>>
>> Thanks,
>>
>> Gary
>>
>
> Finally, I notice that you're using version 2.2 of the code. This is 
> really old, and may have some legacy bugs that haven't been patched.
>
> Best wishes,
> Conrad 
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160815/e15a338a/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: charge0_3nm.jpg
Type: image/jpeg
Size: 40406 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160815/e15a338a/attachment.jpg>


More information about the CP2K-user mailing list