[CP2K-user] MD calculation of NPT_F ensemble including baro- and thermostats at various T

[Lukas C]

Hi!

I just found out that it works with an NVE ensemble. but this seems to be a 
bad choice in my understanding.

Best regards,
Lukas

Lukas C schrieb am Mittwoch, 18. November 2020 um 13:42:02 UTC+1:

> Hi Marcella! Hi everyone!
>
> I tried to run the annealing as suggested. So I turned off the thermostats 
> and choose an NPE_F ensemble. The temperature is changed very gently: 
> starting from 2K, with a rescaling factor for the velocities of 1.01, it 
> should reach 3.127K after 50 steps. Unfortunately, I get unexplained MPI 
> errors at the end of the first SCF step which apart from that converges 
> nicely and does not show any signs of malfunction. I do get a WARNING 
> before the SCF cycle starts, namely when calculating the PDOS at iteration 
> step 0. All the input and output files (except for the .wfn file which is 
> too large) are attached. I tried running it with different preconditioners 
> but this did not change anything. I even ran the simulation on another 
> cluster where it got stuck at the same point.
> Any thoughts on this? Thank you!
>
> Kind regards,
> Lukas
>
> Lukas C schrieb am Dienstag, 17. November 2020 um 10:08:26 UTC+1:
>
>> Dear Marcella!
>>
>> Thank you so much for your help! I understand a lot more now. I will try 
>> to incorporate your advices and keep you updated about any progress.
>>
>> Best regards,
>> Lukas
>>
>> Marcella Iannuzzi schrieb am Montag, 16. November 2020 um 16:27:03 UTC+1:
>>
>>>
>>> Dear Lukas, 
>>>
>>> It is recommendable to monitor the behaviour of the system at conditions 
>>> closer to equilibrium. 
>>> Jumping from 0K to 3000K seems quite ambitious. 
>>>
>>> The conserved quantity has to be conserved also with the NPT ensemble, 
>>> where it is not simply the sum of pot. and kin. energy, but it includes all 
>>> the due terms of the extended ensemble. If it is not conserved there is a 
>>> problem.  
>>> I would be conservative with the choice of the thermostat, Nose-Hoover 
>>> and CSVR are for sure reliable. 
>>> Adding a thermostat to the barostat helps in controlling the cell 
>>> fluctuations, that indeed are quite wild according to the posted example. 
>>> Anyway, monitoring the stress is also useful to check on the sanity of 
>>> the system. 
>>>
>>> For the annealing, I would rather suggest a step-wise procedure, i.e. 
>>> running a series of run at increasing temperature  (\Delta T ~ 200 K ), 
>>> starting always from the previous one. 
>>>
>>> Kind regards
>>> Marcella
>>>
>>>
>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20201118/157c2c1f/attachment.htm>