Can CP2K fixed part of atoms when perform CI-NEB?

J. Ye jingyu... at gmail.com
Tue Nov 8 19:13:32 CET 2016


Hi Matthias, 

Thanks very much! I miss understand the point you mention "You should 
rather check if fixing atoms is really needed as a boundary for a realistic 
modelling of the reaction you are interested in." Sorry about this. 

Here I would like to reported the performance for my current calculation 
(~600 atoms). Fix atoms does not improve the speed of BAND calculation. The 
observation is based on two ionic steps.  

Best regards, 
J. Ye


On Tuesday, November 8, 2016 at 10:40:24 AM UTC-6, Matthias Krack wrote:
>
> Hi
>
> the default values might be tight, but these defaults should be considered 
> like all other default values only as a first guess or proposal which has 
> to be adapted to the actual application.
>
> Best regards
>
> Matthias
>
> On Tuesday, 8 November 2016 17:12:41 UTC+1, J. Ye wrote:
>>
>> I have another question. The default value for CONVERGENCE_CONTROL 
>> <https://manual.cp2k.org/cp2k-4_1-branch/CP2K_INPUT/MOTION/BAND/CONVERGENCE_CONTROL.html> for 
>> BAND optimization is very difficult to converge. I was curious is why the 
>> criterias for BAND optimization is so high? What is the reasonable 
>> criteria?   
>>     &CONVERGENCE_CONTROL
>>       MAX_DR 0.0002
>>       MAX_FORCE 0.00045
>>       RMS_DR 0.0001
>>       RMS_FORCE 0.0003
>>     &END CONVERGENCE_CONTROL
>>
>>
>>
>> On Tuesday, November 8, 2016 at 9:32:49 AM UTC-6, J. Ye wrote:
>>>
>>> Hi Matthias, 
>>>
>>> I am working on a catalyst material that has metal nanoparticles 
>>> supported on a porous material.  This porous material which is a big system 
>>> that has about 600 atoms, however the catalytic reaction process only on 
>>> the metal clusters. We want to study the effect of the pore size on the 
>>> catalytic reaction. So I want to perform a periodic calculations and 
>>> compare with cluster model. If the barriers are similar, then I could 
>>> perform for cluster model instead. 
>>>
>>> Your suggestion is peroform QM/MM? 
>>>
>>> What's the difference between DIIS and MD? Why DIIS is more efficient? 
>>> Do you have any books or papers recommend for us to read? 
>>>
>>> Thanks very much!
>>>
>>> Best regards,
>>>  
>>> J. Ye
>>>
>>> On Tuesday, November 8, 2016 at 2:12:42 AM UTC-6, Matthias Krack wrote:
>>>>
>>>> Hi
>>>>
>>>> do you have any special reason to use the OPT_TYPE MD? Otherwise I 
>>>> suggest to try the default OPT_TYPE DIIS which should be much more 
>>>> efficient. Moreover, if you want to fix some atoms only because of the 
>>>> computational cost caused by the large size of your system then there will 
>>>> be most likely not much benefit doing so, because the forces for all atoms 
>>>> are calculated anyway and then set to zero for the fixed atoms afterwards. 
>>>> You should rather check if fixing atoms is really needed as a boundary for 
>>>> a realistic modelling of the reaction you are interested in.
>>>>
>>>> Best regards
>>>>
>>>> Matthias
>>>>
>>>> On Tuesday, 8 November 2016 07:01:18 UTC+1, J. Ye wrote:
>>>>>
>>>>>  Hi Fabio, 
>>>>>
>>>>> I indeed included both "ALIGN_FRAMES FALSE" and "ROTATE_FRAMES FALSE" 
>>>>> in my input(Please see the attachment). But I still found the atoms moves. 
>>>>> The methods we use are different. I don't know how does this affect the 
>>>>> fixed atoms. Do you have any idea? 
>>>>>
>>>>> I have tested your input. Atom 1 and 3 indeed do not move. I also 
>>>>> tried to switch TI-NEB to CI-NEB, the atoms are not move. So this should 
>>>>> not be affected by BAND_TYPE. 
>>>>>
>>>>> Can any expert of CP2K give us some hints on this problem? 
>>>>>
>>>>> Thanks very much!
>>>>> Best,
>>>>>
>>>>> J. Ye
>>>>>
>>>>> On Monday, November 7, 2016 at 2:02:05 PM UTC-6, elb... at gmail.com 
>>>>> wrote:
>>>>>>
>>>>>> Hello J. Ye!
>>>>>>
>>>>>> Yes, I have checked. If you use only the constraint flag, indeed the 
>>>>>> fixed atoms move for the other images. But if you include both 
>>>>>> "ALIGN_FRAMES FALSE" and "ROTATE_FRAMES FALSE" flags in the BAND section, 
>>>>>> that does not happen anymore, at least in the calculations I have performed 
>>>>>> so far. Check for example the NEB input attached to this e-mail. It is very 
>>>>>> simple and it runs very fast (it has no meaning, it is just an example), 
>>>>>> but you can check that atoms 1 and 3, that are asked to be fixed, do not 
>>>>>> move for all NEB images. Let me know if you get the same behavior as me!
>>>>>>
>>>>>> All the best!
>>>>>>
>>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20161108/67f93fcc/attachment.html>


More information about the CP2K-user mailing list