[CP2K-user] [CP2K:20290] Re: Dipole error with liquid metal system: "Berry phase moments for non uniform MOs' occupation numbers not implemented"

[Michela Benazzi]

Hello everyone, I was just hoping to post an update. 

After trying all of the following and more and seeing minor/no improvements:

- lowering MIXING ALPHA (sped up convergence times)
- changing electronic temperature
- switching from CP2K v. 6.1.0 to the newest container available
- testing out ASPC, PS, previous wave function extrapolators
- 2 vs 3 extrapolation order
- GAPW method
- changing CSVR TIMECON value

My issue was actually my particle density! I was operating with a system 6x 
less dense than liquid aluminum, essentially a metal gas (just calling it 
for the laughs!). I decreased my unit cell dimensions and adjusted the 
number density - my MD job has been running for over 4 h (~5 h at the 
moment now) and has computed many more MD steps than usual (currently at 
#40, while it would normally crash after 4 h at step #10).

Thank you everyone for all the help! I learned a lot of useful literature 
in the process, and I feel like my understanding of CP2K got better.

Michela
On Wednesday, May 29, 2024 at 9:27:25 AM UTC-4 Michela Benazzi wrote:

> Hello Marcella, thank you so much for your thorough response!!
>
> I do not get any error message when my calculations stop without 
> PRINT_LEVEL HIGH - there is just a line at the end of the output saying 
> that the job timed out. At first, I thought that the issue might be on the 
> computing side, but it's not. I was able to run other jobs (Al + methanol, 
> instead of Al + CO2) for the runtime that I allocated, and I was using the 
> same bash file settings, same basis sets for the same elements, same 
> smearing temperature, etc. The only thing that changed were the coordinates 
> and the # ADDED_MOs. 
>
> Could you explain what you mean by 'previous wavefunction' (regarding 
> diagonalization)?
>
> I might respond to this thread again once I try changing other parameters 
> - grazie!
>
> Michela
> On Wednesday, May 29, 2024 at 3:52:37 AM UTC-4 Marcella Iannuzzi wrote:
>
>> Dear Michela,
>>
>> The dipole moment error  is triggered by the PRINT_LEVEL HIGH 
>> The calculation of the dipole moment is not  explicitly requested in the 
>> input, but it is activated by the PRINT_LEVEL HIGH 
>> Though, it is correct that the Berry phase approach cannot be applied to 
>> MOS with different occupation numbers.
>>
>> This does not explain what happens when the calculation stops without 
>> PRINT_LEVEL HIGH . 
>> Do you have an error message in that case too. 
>>
>> Just a few more comments concerning the computational settings. 
>> The convergence of the SCF is very slow. Probably it could be improved 
>> adjusting some parameters, first of all reducing the mixing  parameter 
>> ALPHA (e.g. 0.005), but also  considering smearing temperature, 
>> EPS_DEFAULT,  number of added mos and maybe cutoff.
>> You should also consider to use more accurate basis sets and VdW 
>> corrections. 
>> When running MD with diagonalization, the ASPC extrapolator is a bad 
>> choice, better use the previous wave function , 
>>
>> Regards 
>> Marcella
>>
>>
>> On Wednesday, May 29, 2024 at 12:06:05 AM UTC+2 bnzmi... at gmail.com wrote:
>>
>>> Hello everyone,
>>>
>>> I am using CP2K 6.1.0 (the cluster I have access to is not updated). My 
>>> MD jobs have been timing out after 4 hours despite the CPU time allocation 
>>> I specify (24 h). Other files run normally, so I am pretty sure that the 
>>> issue is due to my input file. 
>>>
>>> I tried adding PRINT_LEVEL HIGH to get more information on the issue, 
>>> and this is what comes up: Berry phase moments for non uniform MOs' 
>>> occupation numbers not implemented. SCF converged successfully, and I am 
>>> not sure if the issue is ADDED_MOs (I went by 20% of all available MOs), or 
>>> if it intrinsic to SMEAR, which I need because I have a metallic system.
>>>
>>> Please kindly refer to my attached input and output for more 
>>> information. I would genuinely appreciate the help - this has been a 
>>> roadblock for some time! Thank you very much and have a nice day,
>>>
>>> Michela
>>>
>>

-- 
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/4086b78d-b0e9-4c2c-ac29-3c418569446fn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20240610/90aa4a81/attachment.htm>