[CP2K-user] [CP2K:14857] Re: Severe deformation of structure during geometry optimization

ASSIDUO Network lenardc... at gmail.com
Mon Mar 15 07:46:09 UTC 2021


Just wanted to say that you were right about the dipole correction. You 
need to perform the computation first without the dipole correction, use 
that optimized geometry then with dipole correction included. Just wanted 
to add this in case someone has any problems with it too.

Thanks Matt

On Wednesday, March 3, 2021 at 5:37:50 PM UTC+2 Matt W wrote:

> Wavelet should just work in my experience.
>
> For the dipole correction - In 'other codes' you had to (approximately) 
> converge the electronic stucture without correction and then turn it on. 
> Just starting a fresh calculation tended to diverge. As I say, not tried in 
> CP2K.
>
> On Wednesday, March 3, 2021 at 5:08:55 AM UTC ASSIDUO Network wrote:
>
>> I will look into those things, maybe I will get lucky.
>>
>> In terms of dipole correction, it's been a nightmare for me to work. I 
>> have posted a few questions in this forum about it.
>>
>> On Tue, Mar 2, 2021 at 6:37 PM Matt W <mat... at gmail.com> wrote:
>>
>>> I normally use WAVELET poisson solver - then you need Y as the 
>>> non-periodic direction (so PERIODIC XZ), the system must be in the centre 
>>> of the cell and you need ~5 A vacuum each side to get converged results.
>>>
>>> The ANALYTIC psolver is also usable for 2D poisson, but it requires a 
>>> much bigger cell to converge (~2x), to my knowledge. 
>>>
>>> Do check that you are converged with cell size if you go these routes. 
>>> There is also a dipole correction method available somewhere (I've not 
>>> used).
>>>
>>> Matt
>>>
>>> On Tuesday, March 2, 2021 at 4:04:03 PM UTC Lucas Lodeiro wrote:
>>>
>>>> Yes, and in &CELL too, to avoid interaction between slab replicas and 
>>>> their electrostatic fields.
>>>>
>>>> El mar, 2 mar 2021 a las 12:07, Lenard Carroll (<len... at gmail.com>) 
>>>> escribió:
>>>>
>>>>> Just to be 100% clear, are you saying that for the &POISSON section 
>>>>> which deals with the electrostatic potential, that I select PERIODIC XY?
>>>>>
>>>>> On Tue, Mar 2, 2021 at 4:55 PM Lucas Lodeiro <el... at gmail.com> 
>>>>> wrote:
>>>>>
>>>>>> Just a comment, see your electrostatic potential profile. You are 
>>>>>> using a periodical slab picture, then when you add a polar molecule, there 
>>>>>> is not a flat potential into the vacuum, in that case, disable the 
>>>>>> periodicity in this direction or add an electrostatic field to correct it.
>>>>>> Also, to check the convergence of your quantities with respect to 
>>>>>> your model, you could explore bigger slabs than 6 layers.
>>>>>>
>>>>>> Regards - Lucas Lodeiro
>>>>>>
>>>>>> El mar, 2 mar 2021 a las 6:37, ASSIDUO Network (<len... at gmail.com>) 
>>>>>> escribió:
>>>>>>
>>>>>>> Thank you for the quick response.
>>>>>>>
>>>>>>> I understand what you are saying and it makes total sense to me. 
>>>>>>> Thank you for clarifying.
>>>>>>>
>>>>>>> The 3x3x1 was just to test how my final adsorption energy changes 
>>>>>>> when I change various settings, including alpha, beta, kpoints and cutoff. 
>>>>>>> I got similar results to the the 5x5x1 case, however the 5x5x1 result was 
>>>>>>> more accurate compared to the ltierature.
>>>>>>>
>>>>>>> On Tuesday, March 2, 2021 at 11:33:40 AM UTC+2 fa... at gmail.com 
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Dear Lenard,
>>>>>>>>
>>>>>>>> Yes, you have correctly answered your question. For metals you need 
>>>>>>>> a quite good sampling of the brillouin zone. Therefore a gamma-point 
>>>>>>>> calculation of a 5x5 angstrom unit cell is expected to yield unreasonable 
>>>>>>>> results. 
>>>>>>>> The fact that you managed to get a good structure with a cutoff of 
>>>>>>>> 300 is probably due to a coincidental error cancelling. But as soon as you 
>>>>>>>> alter the configuration by adding a CO the error cancelling does not work 
>>>>>>>> anymore and the structure becomes unreasonable again.
>>>>>>>>
>>>>>>>> You should make sure that your quantity of interest is converged 
>>>>>>>> w.r.t the size of the k-point grid. 3x3x1still seems small to me, but maybe 
>>>>>>>> its already enough for you.
>>>>>>>>
>>>>>>>> Cheers,
>>>>>>>> Fabian
>>>>>>>> On Tuesday, 2 March 2021 at 06:42:45 UTC+1 ASSIDUO Network wrote:
>>>>>>>>
>>>>>>>>> Hi there everyone, I am currently trying to geometry optimize a 
>>>>>>>>> stack of Au (111) 2x2 surfaces, specifically by fixing the bottom 3 
>>>>>>>>> surfaces and relaxing the top 3.
>>>>>>>>>
>>>>>>>>> Unfortunately, the top surfaces completely mess up and the final 
>>>>>>>>> structure looks bad. When I turn KPOINTS on  (3x3x1 or 5x5x1), it no longer 
>>>>>>>>> happens.
>>>>>>>>> Finally, when I lower the cutoff to 300, but I leave KPOINTS off, 
>>>>>>>>> I get a good optimized geometry, but when I try to perform a CO adsorption 
>>>>>>>>> with that optimized structure, it severely deforms as well.
>>>>>>>>>
>>>>>>>>> Does anyone have a solution to my problem, particularly one that 
>>>>>>>>> doesn't change the ELECTRONIC_TEMPERATURE, KPOINTS or CUTOFF. Is it the 
>>>>>>>>> need for KPOINTS in a small structure like this that messes up the results?
>>>>>>>>>
>>>>>>>>> I have attached all my relevant files.
>>>>>>>>>
>>>>>>>>> Kind regards,
>>>>>>>>> Lenard
>>>>>>>>>
>>>>>>>> -- 
>>>>>>> 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 cp... at googlegroups.com.
>>>>>>> To view this discussion on the web visit 
>>>>>>> https://groups.google.com/d/msgid/cp2k/d1100c7c-514e-4fbc-82e5-528bcdad45abn%40googlegroups.com 
>>>>>>> <https://groups.google.com/d/msgid/cp2k/d1100c7c-514e-4fbc-82e5-528bcdad45abn%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 cp... at googlegroups.com.
>>>>>> To view this discussion on the web visit 
>>>>>> https://groups.google.com/d/msgid/cp2k/CAOFT4PJSHNzMu9F08E0UO-rBJZJO7puF6Z9UMUxQDxzu10L_bA%40mail.gmail.com 
>>>>>> <https://groups.google.com/d/msgid/cp2k/CAOFT4PJSHNzMu9F08E0UO-rBJZJO7puF6Z9UMUxQDxzu10L_bA%40mail.gmail.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 cp... at googlegroups.com.
>>>>>
>>>> To view this discussion on the web visit 
>>>>> https://groups.google.com/d/msgid/cp2k/CAKur158Q1dvprpLA5QyzBVSMy1qJd14oxfFPUaFygaKF2TcYBA%40mail.gmail.com 
>>>>> <https://groups.google.com/d/msgid/cp2k/CAKur158Q1dvprpLA5QyzBVSMy1qJd14oxfFPUaFygaKF2TcYBA%40mail.gmail.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 cp... at googlegroups.com.
>>>
>> To view this discussion on the web visit 
>>> https://groups.google.com/d/msgid/cp2k/31940cf9-268f-443d-a5ff-17f5759e0969n%40googlegroups.com 
>>> <https://groups.google.com/d/msgid/cp2k/31940cf9-268f-443d-a5ff-17f5759e0969n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210315/2ab8cf57/attachment.htm>


More information about the CP2K-user mailing list