[CPMD-list] Ag PP

[Alexander Martins Silva]

       Hi,

       My imput is the following:

&CPMD

  OPTIMIZE WAVEFUNCTION

  MAXSTEP

  1000

  STRUCTURE BONDS ANGLES

&END

&DFT

  NEWCODE

  FUNCTIONAL BLYP

&END

&SYSTEM

  ANGSTROM

  SYMMETRY

  1

  CELL

  8.2554  1.0000 1.0000  0  0  0

  CUTOFF

  70.0

&END

&ATOMS

*AG_MT_BLYP_1.8     KLEINMAN-BYLANDER

  LMAX=D

  3

    .....

*H_MT_BLYP

  LMAX=S

  1

    ......

&END


       I don't know much about some keywords like LMAX: what are the 
criteria used? I'm trying to carry out another calculation and I'm 
seeing the same problems... I dont' know why...  where are the mistakes?

       Alexander.


Ari P Seitsonen wrote:

>Dear Alexander,
>
>  How many silver atoms do you now have in your system? Like Axel wrote it 
>might be that there are so many states close to the Fermi energy - ehmm, 
>better said that there might be many occupied and unoccupied states 
>closeby, and during the optimisation they change the ordering. If you are 
>sure that there is a gap (e.g. you have calculated the same system with a 
>cluster code) there might be some parameters which could help, otherwise 
>you might want to switch to fractional occupation numbers (FREE ENERGY 
>FUNCTIONAL) and a broadening of the occupation numbers.
>
>  Just a thought...
>
>    Greetings from Karelia/Finland and Zurich,
>
>       apsi
>
>PS Just a detail, do you use the s or p as the local component of the 
>  pseudo? And how high is the total energy at the first step?
>
>-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
> Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
> Tel +41 1 635 44 97 / Fax +41 1 635 68 38 / GSM +41 79 719 09 35      
> Anschrift: Physikalisch Chemisches Institut (PCI), Universität Zürich (UniZh)
> Indirizzo: Winterthurerstraße 190, CH-8057 Zürich
> Address:   Schweiz / Svizzera / Suisse / Svizra / Switzerland
>
>On Sat, 22 May 2004, Alexander Martins Silva wrote:
>
>  
>
>>       Hi,
>>
>>          I'm using a blyp PP from a colleague of this list ( ...Please, 
>>forgive me, I have deleted your email accidentally... Did you have these 
>>same problems? ...), and I've realized a little improvement, but there 
>>is oscilations and the excessive number of iteractions in the 
>>wavefunction optimization yet. As suggested for you, I change the 
>>minimizer from ODIIS to PCG, and with this approach the oscillations 
>>have finished, but the convergence stils slow:
>>...
>>
>> 407  1.293E-03   5.452E-05    -131.750308   -2.013E-05     54.93
>>
>> LINE SEARCH : LAMBDA=.176           ENERGY =          -131.75035
>>
>> 408  1.302E-03   5.516E-05    -131.750328   -2.013E-05     55.04
>>
>> LINE SEARCH : LAMBDA=.165           ENERGY =          -131.75037
>>
>> 409  1.150E-03   5.651E-05    -131.750350   -2.155E-05     54.81
>>
>> LINE SEARCH : LAMBDA=.171           ENERGY =          -131.75039
>>
>> 410  1.486E-03   5.603E-05    -131.750371   -2.137E-05     54.76
>>
>> LINE SEARCH : LAMBDA=.172           ENERGY =          -131.75041
>>...
>>
>>
>>       What others parameters should I change to resolve this problem?  
>>Another question: Is there another code for TM pseudopotentials besides 
>>the Gianozzi's one? 
>>
>>      Alexander.
>>
>>       Thanks,
>>
>>          Alexander.
>>
>>
>>Axel Kohlmeyer wrote:
>>
>>    
>>
>>>On Fri, 21 May 2004, alex.msilva wrote:
>>>
>>>AM> Hi,
>>>
>>>hello alexander,
>>>
>>>AM> 
>>>AM>    I´m trying to obtain some results for a 
>>>AM> little Ag cluster, so I must to generate a PP 
>>>AM> for this atom, using the Gianozzi´s code. After 
>>>AM> that, I have done some tests to validate this 
>>>AM> PP. I found that a simple optimization 
>>>AM> wavefunction calculation were evaluated after a 
>>>AM> slow convergence of more than 100 steps. Is this 
>>>AM> wrong? What can I do to improve this PP?
>>>
>>>i have noticed, that systems containing transition
>>>metal atoms frequently have convergence issues,
>>>especially for larger systems.
>>>
>>>you may want to try PCG MINIMZE instead of the
>>>default ODIIS minimizer. also sometimes the 
>>>convergence improves with increasing the cutoff
>>>a bit.
>>>
>>>as for the quality of your pseudopotential,
>>>there are two Ag pseudos (one with 1 valence electron
>>>and one with 11 valence electrons) in the 
>>>goedecker-hartwigsen pseudopotential library for LDA.
>>>if your pseudo is using LDA, or you recreate it
>>>for LDA, you could compare your results.
>>>
>>>hope this helps,
>>>	axel.
>>>
>>>AM> 
>>>AM> 
>>>AM> Alexander.
>>>AM>  
>>>AM> __________________________________________________________________________
>>>AM> Acabe com aquelas janelinhas que pulam na sua tela.
>>>AM> AntiPop-up UOL - É grátis!
>>>AM> http://antipopup.uol.com.br/
>>>AM> 
>>>AM> 
>>>AM> _______________________________________________
>>>AM> CPMD-list mailing list
>>>AM> CPMD-list at cpmd.org
>>>AM> http://www.cpmd.org/mailman/listinfo/cpmd-list
>>>AM> 
>>>AM> 
>>>
>>> 
>>>
>>>      
>>>
>>_______________________________________________
>>CPMD-list mailing list
>>CPMD-list at cpmd.org
>>http://www.cpmd.org/mailman/listinfo/cpmd-list
>>
>>    
>>
>_______________________________________________
>CPMD-list mailing list
>CPMD-list at cpmd.org
>http://www.cpmd.org/mailman/listinfo/cpmd-list
>
>  
>