Nickel cell optimization

[Alex]

Hey Matt,

Thanks, I will try your version. My only problem at this point is that SCF 
does not converge. 
Now, I thought that UKS (along with MULTIPLICITY) was there to include spin 
polarization, which should be very important for something like nickel. 
Also, why is it not needed if Fermi-Dirac smearing of the wavefunction is 
enabled? I kinda fail to see how one excludes the other qualitatively. In 
other words, by omitting the magnetic part, am I really simulating nickel?

Just to add to the info, the 6x6x6 simulation converged during energy 
minimization (constant cell size) and in fact yielded reasonable cohesive 
energy (the input included smearing and UKS). The convergence issues 
started during cell optimization.

Your help is appreciated.

Thanks,

Alex


On Thursday, October 9, 2014 4:01:38 PM UTC-6, Matt W wrote:
>
> Hi Alex,
>
> here's a slightly tidied version of you original input. It worked for a 
> 3x3x3 supercell. Not sure what problems you were having without seeing more 
> details. I did remove the UKS tag, as with smearing it is not needed - this 
> might make the optimization easier.
>
> Matt
>
> On Thursday, October 9, 2014 8:05:16 PM UTC+1, Alex wrote:
>>
>> Hi Florian,
>>
>> I removed the print statements for moments and tried to play around with 
>> the outer SCF optimization (with BISECT as the optimization scheme). No 
>> luck whatsoever, although the screaming about Berry phase has stopped. I 
>> really think there's something up with the way my cell is set up... I 
>> thought this was one of the simplest possible calculations one could do on 
>> a crystal structure, and here we are.
>>
>> Thanks for your suggestions.
>>
>> Alex
>>
>>
>> On Monday, October 6, 2014 1:56:06 PM UTC-6, Florian Schiffmann wrote:
>>>
>>> Hi Alex,
>>>
>>> try playing with some options in the SCF section, e.g. mixing parameter, 
>>> diis settings, outer_scf loop,... The system might be hard to converge in 
>>> the first step and therefore needs some tweaking.
>>> The second error, as written in the error message has nothing to do with 
>>> the cell optimzation but the point that you have activated the calculation 
>>> of dipole moments (DFT PRINT MOMENTS or HIGH print level). In combination 
>>> with smearing (non uniform occupation) this is not implemented and 
>>> therefore the calcualtion stops. If you remove the moment section the 
>>> calculation will run through even without a perfectly converged 
>>> wavefunction.
>>>
>>> Cheers
>>> Flo
>>>
>>> On Thursday, October 2, 2014 9:39:40 PM UTC+2, Alex wrote:
>>>>
>>>> With the TOPOLOGY subsection fixed, the GEO_OPT run (as defined in 
>>>> input above) produces about 200 SCF updates and quits reporting "SCF not 
>>>> converged." The CELL_OPT run (it is my understanding that it optimizes all 
>>>> coordinates for each given cell size and cycles between the two 
>>>> optimizations) quits with
>>>>
>>>>
>>>>  ***************************************************************************
>>>>  *** 02:51:32 ERRORL2 in qs_moments:qs_moment_berry_phase 
>>>> UNIMPLEMENTED, ***
>>>>  *** Berry phase moments for non uniform MOs' occupation numbers not   
>>>>   ***
>>>>  *** implemented                                                       
>>>>   ***
>>>>
>>>>  ***************************************************************************
>>>>
>>>> Any thoughts? Thanks,
>>>>
>>>> Alex
>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141009/984f9a95/attachment.htm>