[Ape-users] Problems when performing all-electron calculation using MGGA functional

Tue Nov 23 14:19:28 WET 2010

Dear Yi,

  The trick is to shift the potential so that it goes to zero at 
infinity. This works very well for the Becke-Jonhson potential and 
should be easy to do for the Tran-Blaha potential.

  Nevertheless, if your goal is to generate pseudopotentials with the TB 
potential, you should be careful. In fact, there is a fundamental 
problem in this case. The TB uses a quantity that is averaged over the 
unit cell (see eq. 3 of the TB paper). The value of this quantity is 
quite different for an isolated atom and for a solid (not to mention 
that it was not at all meant to be used for finite systems). This means 
that the pseudopotential will not be truly transferable and it will very 
likely produce wrong results.

  Regards,

Micael

On 11/19/2010 04:29 PM, gaoy at aphy.iphy.ac.cn wrote:
>
> Dear Micael,
>
> Thank you very much for your help. I tried BJ06 functional so it works smoothly and pseudopotential can be generated.
>
> But if we want to study the band gap problem, it seems TB09 can give more better results compared to experiments. Maybe still we want to use that functional in further calculations.
>
> It is encouraging that you guys solved the similar problem in BJ06, but could we know it is solved in APE package or libxc package? Could the problem in TB09 be solved using exactly the same method as in BJ06?
>
> I will be grateful if you enlighten us.
>
> Best wishes!
>
> Yi
>
> 2010.11.20
>
>
>
>
>> ----------------------------------------------------------------------
>>
>> Message: 1
>> Date: Mon, 8 Nov 2010 21:04:18 +0800 (CST)
>> From: gaoy at aphy.iphy.ac.cn
>> To: ape-users at tddft.org
>> Subject: [Ape-users] Problems when performing all-electron calculation
>> 	using MGGA functional
>> Message-ID:<477b3e.11a.12c2b964aeb.Coremail.gaoy at aphy.iphy.ac.cn>
>> Content-Type: text/plain; charset=GBK
>>
>> Dear APE developers and users,
>>
>> I am very sorry that the last mail concerning the same problem is not complete. Following is the complete description of my problem.
>>
>> It is amazing and awesome work that ape-1.1.0 incorporated more functionals including MGGA. However, I got a problem when performing all-electron calculation using MGGA functional.
>>
>> For calculating element Silicon, the following "inp-ape" file was prepared:
>>
>> ###########################################################################
>> Title = 'Silicon'
>> CalculationMode = ae
>> Verbose = 40
>> TypeOfMixing = 1
>> Mixing = 0.05
>> EigenSolverTolerance = 1.0e-5
>> XCFunctional = 208012
>>
>> NuclearCharge = 14
>> %Orbitals
>>   "Ne"
>>   3 | 0 | 2 | 0
>>   3 | 1 | 2 | 0
>>   4 | 0 | 0 | 0
>> %
>> ###########################################################################
>>
>> The calculation can be started but after a few iteration, the program terminated, presenting the error message:
>>
>> ###########################################################################
>> Fatal Error (description follows)
>>    Unable to bracket eigenvalue for state: 3p
>> ###########################################################################
>>
>> Does anyone have any idea on that?
>>
>> Thank you very much for attentions.
>>
>> Best wishes!
>>
>> Yi
>>
>> 2010.11.08
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> ------------------------------
>>
>> Message: 2
>> Date: Mon, 08 Nov 2010 20:48:27 +0000
>> From: Micael Oliveira<micael at teor.fis.uc.pt>
>> To: ape-users at tddft.org
>> Subject: Re: [Ape-users] Problems when performing all-electron
>> 	calculation using MGGA functional
>> Message-ID:<4CD8621B.6040806 at teor.fis.uc.pt>
>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>
>> Dear Yi,
>>
>>    It seems that at a certain point during the SCF cycle the code is
>> unable to find the eigenvalue for a given state and this usually means
>> that the state is unbound. In this case, my guess is that the problem
>> comes from the asymptotic limit of the Tran-Blaha MGGA, which will be a
>> problem for APE if it implies that the eigenvalue of a given state
>> becomes positive.
>>
>>    In the case of the Becke-Johnson MGGA we found a way of getting around
>> this problem, but unfortunately we did not implemented it for the Tran
>> -Blaha functional. In principle this should be fixed in the next release
>> of the APE. To be sure this is the problem, maybe you could try the
>> Becke-Johnson functional and see if it works. Also, this functional
>> should give better results for atoms than the Tran-Blaha one.
>>
>>    Best regards,
>>
>> Micael
>>
>>
>> On 11/08/2010 01:04 PM, gaoy at aphy.iphy.ac.cn wrote:
>>> Dear APE developers and users,
>>>
>>> I am very sorry that the last mail concerning the same problem is not complete. Following is the complete description of my problem.
>>>
>>> It is amazing and awesome work that ape-1.1.0 incorporated more functionals including MGGA. However, I got a problem when performing all-electron calculation using MGGA functional.
>>>
>>> For calculating element Silicon, the following "inp-ape" file was prepared:
>>>
>>> ###########################################################################
>>> Title = 'Silicon'
>>> CalculationMode = ae
>>> Verbose = 40
>>> TypeOfMixing = 1
>>> Mixing = 0.05
>>> EigenSolverTolerance = 1.0e-5
>>> XCFunctional = 208012
>>>
>>> NuclearCharge = 14
>>> %Orbitals
>>>    "Ne"
>>>    3 | 0 | 2 | 0
>>>    3 | 1 | 2 | 0
>>>    4 | 0 | 0 | 0
>>> %
>>> ###########################################################################
>>>
>>> The calculation can be started but after a few iteration, the program terminated, presenting the error message:
>>>
>>> ###########################################################################
>>> Fatal Error (description follows)
>>>     Unable to bracket eigenvalue for state: 3p
>>> ###########################################################################
>>>
>>> Does anyone have any idea on that?
>>>
>>> Thank you very much for attentions.
>>>
>>> Best wishes!
>>>
>>> Yi
>>>
>>> 2010.11.08
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Ape-users mailing list
>>> Ape-users at www.tddft.org
>>> http://www.tddft.org/mailman/listinfo/ape-users
>>>
>>
>>
>>
>> ------------------------------
>>
>> _______________________________________________
>> Ape-users mailing list
>> Ape-users at www.tddft.org
>> http://www.tddft.org/mailman/listinfo/ape-users
>>
>>
>> End of Ape-users Digest, Vol 7, Issue 3
>> ***************************************
>
>
>
>
>
>
>
>
>
> _______________________________________________
> Ape-users mailing list
> Ape-users at www.tddft.org
> http://www.tddft.org/mailman/listinfo/ape-users
>