CUTOFF convergence problem

[Guanna Li]

I also met the same problem for my zeolite system. I used an optimized 
structure as input and only calculated the single point energy with 
different CUTOFF values.
The energy differences were bigger than acceptable. 


On Wednesday, November 11, 2015 at 5:40:48 PM UTC+1, Matt W wrote:
>
> Dear Katrina,
>
>
> You are right, I am comparing the energy differences after geometry 
>> optimizations and it seems that with different cutoff values I am finding 
>> different minima. What is bothering me is that even CUTOFF 1200 Ry is 
>> obviously not enough as I found another minima for structure 1 with cutoff 
>> 1400 Ry. The using of such high cutoff is not very convinient for zeolite 
>> structures as far as I know. So I am looking for some "bug" in my setup.
>>
>>
> energy minimisation can be essentially chaotic, tiny differences in the 
> potential energy surface / initial conditions etc can rapidly lead you to 
> minimise to different minima, especially if you start from very high energy 
> structures. I think the 'bug' in your set up is that you start from a very 
> high energy state and then minimize. If you did this for many different 
> starting positions you should locate all the possible minima eventually. 
> But better to start from close to a local minima and not ask so much from 
> the optimiser.
>
> It is not really a question of one being right and the other wrong in this 
> case if you find different minima. Do both minima exist for both 1200 and 
> 1400 Ry cutoff with consistent energy differences?
>
> HTH,
>
> Matt
>  
>
>> Katarina
>>
>> Dňa utorok, 10. novembra 2015 10:13:10 UTC+1 Matt W napísal(-a):
>>>
>>> Hi Katrina,
>>>
>>> At a given geometry (the one in your input) the total energy of the 
>>> system converges quite OK with cutoff after about 800 Ry, according to me.
>>>
>>> Are you comparing the energy differences after performing geometry 
>>> optimization? Is it possible that a some of your runs are optimizing to 
>>> subtly different structures. Even a tiny difference in the shape of the 
>>> potential energy surface can lead you into a different minima, are you are 
>>> starting from quite high energy structures it seems. Your structure 2 seems 
>>> well behaved, but structure 1 gives a weird energy at the highest cutoff 
>>> you used.
>>>
>>> Matt
>>>
>>> BTW, the basis sets you are using DZV-GTH are probably not really 
>>> production quality.
>>>
>>> On Monday, November 9, 2015 at 3:33:32 PM UTC, katarína stančiaková 
>>> wrote:
>>>>
>>>> Dear CP2K users,
>>>> I am running my calculations and I have CUTOFF convergence problem - in 
>>>> fact I can not see any convergence in the total energy. I know that the 
>>>> relative differences in energies are important (not the absolute values of 
>>>> energies), therefore I am comparing the energy differences between two 
>>>> structures (I am working with zeolites, and my two structures differs in 
>>>> the position of hydrogen). Even with very high cutoff values I can not 
>>>> observe any convergence of the total energies as well as energy differences.
>>>> Do you have any idea where is the problem? I will be thankful for any 
>>>> help!
>>>>
>>>> Here are my data:
>>>>
>>>> CUTOFF [Ry]            Total Energy [a.u.]
>>>>                                 structure 1                structure 2
>>>> 600                           -3467.0403965424    -3467.0619253888
>>>> 800                           -3467.0521866655    -3467.0622253178
>>>> 1000                         -3467.0838714739    -3467.1096147317
>>>> 1200                         -3467.0833499728    -3467.1103018678
>>>> 1400                         -3467.0975445683    -3467.1102723936
>>>>
>>>>
>>>> ENERGY DIFFERENCES
>>>>
>>>>
>>>> CUTOFF [Ry]    E_str1 - E_str2 [kJ/mol]    
>>>>
>>>> 600                55.9750007029
>>>> 800                26.1004960059
>>>> 1000              66.932470253
>>>> 1200              70.0749269477
>>>> 1400              33.0923456248
>>>>     
>>>> Thank you
>>>> Katarina
>>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20151112/f1a3e8c6/attachment.htm>