Nickel cell optimization

[Alex]

That does make sense, thanks. I suppose my comment about smearing and spin 
polarization would be more relevant to a case of a few Kelvin in electronic 
temperature. 
By the way, it is my understanding that the only way to get lattice 
constant as a function of lattice temperature is to run thermostatted MD, 
as CP2K does not include things like Monte Carlo. Is this correct?

Thanks,

Alex


On Friday, October 10, 2014 2:18:42 AM UTC-6, Matt W wrote:
>
> My question about UKS and smearing still stands though... 
>>
>> Without UKS and smearing you'll get a spin averaged structure, no spin 
> polarization obviously. Normally the magnetic effects are actually rather 
> small, so you'll get say energy/volumes that are quite reasonable. 
>
> There will be lower energy solutions that are (anti)-ferromagnetic, spin 
> waves etc depending on the material. In general, you won't find these 
> unless you can construct a good initial guess - by setting a guess based on 
> spin polarized atomic calculations for instance (BS keyword in KIND). The 
> exception is the fully ferromagnetic, which you can just set the 
> multiplicity appropriately.
>
> As you kept the UKS, probably the main change was reducing the ALPHA value 
> (~amount of new solution mixed into the new trial density) to 0.1.
>
> HTH
>
> Matt
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141010/63501c5f/attachment.htm>