[CP2K-user] [CP2K:17224] Best practices for large non periodic molecules

[Matt Watkins]

Quick note on the POISSON solver:
The wavelet solver you have chosen is great because it only needs the 
density to drop to zero at the box edges - so size of molecule + about 10A. 
However, it only works for cubic cells so it needs to be largest molecule 
dimension + 10A in all 3 directions. This could be inefficient for very 
long or disc shaped molecules.
Alternatively the MT or analytic solvers can be used but in each direction 
they should be at least twice the minimum box size (I believe). This could 
still be beneficial in some cases compared to above...
Matt

On Wednesday, 22 June 2022 at 16:01:28 UTC+1 marcos... at gmail.com wrote:

> Thank you very much, I will try as soon as possible.
>
> Best,
> Marco
>
> Il giorno mercoledì 22 giugno 2022 alle 16:45:14 UTC+2 hehn... at gmail.com 
> ha scritto:
>
>> Dear Marco, 
>>
>> regarding convergence I would recommend to use the OUTER_SCF keyword and 
>> SCF_GUESS RESTART, giving a relatively small number of inner SCF cycles, 
>> maybe 20 to 50. Also choosing DIIS as minimizer can help to speed up 
>> convergence.
>>
>> Best regards,
>>
>> Anna
>>
>> Am Mi., 22. Juni 2022 um 14:11 Uhr schrieb Marco Severi <
>> marcos... at gmail.com>:
>>
>>> Good morning, 
>>>
>>> I am a fairly new CP2K user and I am trying to perform some calculations 
>>> on a non-periodic molecular system of 657 atoms.  
>>>
>>> I am writing because I do not fully understand which keywords I should 
>>> use and I have the feeling I am doing some rookie mistakes.
>>>
>>> At the moment I am trying to do some DFT calculations with the PBE 
>>> functional and the dispersion correction. The molecule is about 35 Angstrom 
>>> long and so I chose a 50x50x50 cell. The system is organometallic, contains 
>>> 8 Fe atoms and the BF4- counterions. I tried to use the DZVP basis set but 
>>> I quickly reached 100GB of occupied memory, is it normal? I am using the 
>>> non "PERIODIC NONE" and "PSOLVER WAVELET" keywords  and the FULL_KINETIC 
>>> preconditioner. I am using 5 grids with cutoff equal to 400 and rel_cutoff 
>>> equal to 60. I tried to maintain all the other options as “default” as 
>>> possible. I enclose the input and the output of the calculation. (I am 
>>> aware that the attached example did not converge and that I can change the 
>>> number of SCF iterations)
>>>
>>> I would like to know if those options are adequate for the type of 
>>> calculation I am performing and if there is room for any improvement in the 
>>> efficiency or in the convergence, since this calculation seems to converge 
>>> particularly slowly. Any help or reference will be greatly appreciated. 
>>>
>>> Thanks in advance for your time and patience. 
>>>
>>> Kind regards, 
>>>
>>> Marco Severi 
>>>
>>> -- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "cp2k" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to cp2k+uns... at googlegroups.com.
>>> To view this discussion on the web visit 
>>> https://groups.google.com/d/msgid/cp2k/838e8bc8-a443-4cd2-a8fa-b591ddf1d8ffn%40googlegroups.com 
>>> <https://groups.google.com/d/msgid/cp2k/838e8bc8-a443-4cd2-a8fa-b591ddf1d8ffn%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/742d8096-20da-411c-a9fc-7cc040b9a92an%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20220623/1791ad33/attachment-0001.htm>