[CP2K:6359] Re: Electronic structure of CeO2 greatly depends on initial guess of wave function in OT method

[Matt W]

Hi again,
 

> Thanks for your great suggestion. 
>
> Core correction indeed can help localize electron density correctly. But 
> next step I will run molecular dynamics simulation with adsorbed molecules. 
> In that case, I will not be able to use the core correction.
>
>  

> So question is: if the electronic properties can be correctly predicted by 
> restarting from the wavefunction obtained from the calculation with core 
> correction, are the parameters (basis set, potential, U value, etc.) good 
> enough to be used in MD simulation?
>
>
sorry can't answer that. This would allow you to systematically get 
different solutions and compare them. To be clear, I only suggest this 
"hack" with the core correction to get an initial guess - then you remove 
it. At that point, when the correction is removed, you get what your 
density functional gives you.

For MD you need to decide a starting point - then what wants to happen -> 
happens, if you see what I mean, e.g. 
http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.105.146405

Whether the basis sets/pseudopotentials/density functional you are using 
are good enough to describe what you want ...you need to decide.

Matt

Sorry for so many questions since I don't want to start a big project with 
> some unknown questions. 
>
> Thanks great!
>
> On Tue, Apr 7, 2015 at 12:10 PM, Matt W <Matt... at gmail.com <javascript:>
> > wrote:
>
>> Hi,
>>
>> you should check in the atomic output at the beginning of the run whether 
>> the changes to the atomic states have worked as you want - it shows the 
>> orbital occupations of both spin channels for each atomic kind. Also, 
>> because of .. well it is a feature .. the changes in electron number will 
>> be divided by two (there is a post from Marcella about this somewhere).
>>
>> I'd still go with what I suggested a while back - add a CORE_CORRECTION 
>> to your desired Ce3+ to make the electrons localize, then restart with that 
>> wavefunction without the extra core charge. 
>>
>> Run in a triplet (or maximum multiplicity for all aligned Ce3+) state - 
>> there should be no noticable exchange interaction between pairs of Ce3+, if 
>> you get splittings between singlet/triplet spin states etc then something 
>> is wrong. Running in max spin state automatically breaks some symmetry and 
>> makes it easier to get a proper solution.
>>
>> Matt
>>
>>
>> On Tuesday, April 7, 2015 at 6:00:17 PM UTC+2, stclair wrote:
>>
>>> Hi, Florian,
>>>
>>> Thanks great for your suggestion.
>>>
>>> I have tried broken symmetry for Ce with 'scf_guess' set as 'atomic'. 
>>> But the situation is same to what I got from calculation without Ce (total 
>>> energy ~0.86 eV higher than the restart situation)
>>> Here are the parameters I used. Could you check if they look good? First 
>>> time to use broken symmetry. Also, I will try to add it for O.
>>> For two Ce atoms next-nearby the oxygen vacancy, I tried the following 
>>> setting to make it Ce3+. 
>>>        &BS
>>>          &ALPHA
>>>            NEL -2   -1  1
>>>            L    0   2   3
>>>            N    6   5   4
>>>          &END
>>>          &BETA
>>>            NEL -2  -1   -1
>>>            L    0   2   3
>>>            N    6   5   4
>>>          &END
>>>        &END
>>> But for other Ce atoms, I used the following setting to make it Ce4+.
>>>        &BS
>>>          &ALPHA
>>>            NEL -2   -1  -1
>>>            L    0   2   3
>>>            N    6   5   4
>>>          &END
>>>          &BETA
>>>            NEL -2  -1   -1
>>>            L    0   2   3
>>>            N    6   5   4
>>>          &END
>>>        &END
>>>
>>> Thanks!
>>>
>>> Best.
>>>
>>>
>>> On Tue, Apr 7, 2015 at 12:37 AM, Florian Schiffmann <flos... at gmail.com
>>> > wrote:
>>>
>>>> Hi,
>>>>
>>>> that makes it more clear. As I am not 100% sure what is going on I will 
>>>> have to do some guesswork.
>>>>
>>>> From your input you run an unrestricted calculation. In the atomic 
>>>> guess, CP2K still uses the restricted atomic orbitals as initial guess for 
>>>> both alpha and beta. This might lead to a situation where you have 0.5 of 
>>>> an electron initially sitting in the orbital affected by the U correction. 
>>>> As far as I remember te U correction is symmetric at this point and it 
>>>> might be arbitrary whether the orbital gets further filled or emptied due 
>>>> to the penalty function (therefore there is some arbitraryness in your 
>>>> result). I think this is somewhat general that the U term can lock you in a 
>>>> given state depending on your initial guess.
>>>>
>>>> Could you try to use a broken symmetry guess in your calculation where 
>>>> you initialize O as 2-  and Ce as 4+ and see how the energy is affected?
>>>> For O this chould look like 
>>>> &BS
>>>> &ALPHA
>>>>    L 1
>>>>    N 2
>>>>    NEL 2
>>>> &END
>>>>  &BETA
>>>>    L 1
>>>>    N 2
>>>>    NEL 2
>>>> &END
>>>> &END
>>>>
>>>> For Ce it is a bit more complicated. simply remeber, that cp2k puts a 
>>>> closed shell wfn in alpha and beta. Therefore NEL in this case is 2 as the 
>>>> closed shell representation of a 2p3 is a 2p6 wfn.
>>>>
>>>> Hope that helps
>>>>
>>>> Flo 
>>>>
>>>> -- 
>>>> 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+... at googlegroups.com.
>>>> To post to this group, send email to cp... at googlegroups.com.
>>>> Visit this group at http://groups.google.com/group/cp2k.
>>>> For more options, visit https://groups.google.com/d/optout.
>>>>
>>>
>>>
>>>
>>> -- 
>>> Best,
>>> Lei
>>>
>>> University of Nebraska - Lincoln,
>>> Department of Chemistry,
>>> 536 Hamilton Hall.
>>> Lincoln, NE 68588
>>> Phone: 402-853-8119
>>>
>>>   -- 
>> 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+... at googlegroups.com <javascript:>.
>> To post to this group, send email to cp... at googlegroups.com <javascript:>
>> .
>> Visit this group at http://groups.google.com/group/cp2k.
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>
>
> -- 
> Best,
> Lei
>
> University of Nebraska - Lincoln,
> Department of Chemistry,
> 536 Hamilton Hall.
> Lincoln, NE 68588
> Phone: 402-853-8119
>
>  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150407/8508735e/attachment.htm>