[Pw_forum] NEB Calculation with Ti and Graphene

Eyvaz Isaev eyvaz_isaev at yahoo.com
Wed Dec 22 14:41:55 CET 2010


Dear   Ettore,


> From: Ettore Baldini Neto <neto.baldini at gmail.com>
 
>I understood that but since you mentioned another diagonalisation procedure I 
>have a naive >question (I'm sorry I'm not a native from solid state physics and 
>all these codes are new to me).
>Before I used the atomic positions for my NEB calculation I performed a vc-relax 
>calculation using >the cg scheme and all went well. So the question is, by 
>default the  NEB option uses what >scheme? 
>

I suppose, the diagonalization procedure is not related to the NEB or other 
structure optimization scheme, as always we have to solve an eigenvalue problem 
Hx=\lambda*x  to find eigenvalues and eigenvectors.

The default diagonalization procedure is "david", i.e. Davidson scheme.

>Does it matter or could it be the error cause, I mean different diagonalisations 
>schemes?

No.

>To my experience it depends on the calculation. Sometimes a given algorithm is 
>more suitable >than others for a given problem. I read the NEB tutorials, saw 
>some examples, but although they >are good starting points, they do not mention 
>these particular issues. 
>

Results should not depend on the diagonalization procedure used, though, 
sometime it might fail. But this depends on  libraries used, numerical errors, 
etc. In fact, the error means that diagonalization of a matrix  failed within 
specific machine precision (that depends on a computer).    


If you suppose "david" or "cg" gives you different results, please provide your 
script to test.

Bests,
Eyvaz.

On Dec 22, 2010, at 9:34 AM, Eyvaz Isaev wrote:

> Dear Ettore,
> 
> Of course, I meant the error message, not NEB calculations for Ti and 
>Graphene.  
>
> 
> 
>> *******************************************
>>    from cdiaghg : error #       434
>>    diagonalization (ZHEGV*) failed
>> ********************************************
> 
> 
> If you look at page 54 of User's Guide then you can find that this can be fixed 
>
> using another 
> 
> diagonalization procedure ('cg' or "conjugated gradient" instead of davidson), 

> but there might be another reason.
> 
> Hope this helps.
> 
> Bests,
> Eyvaz.
> 
> 
> 
> ----- Original Message ----
> From: Ettore Baldini Neto <neto.baldini at gmail.com>
> To: PWSCF Forum <pw_forum at pwscf.org>
> Sent: Wed, December 22, 2010 12:18:33 PM
> Subject: Re: [Pw_forum] NEB Calculation with Ti and Graphene
> 
> Dear Eyvaz,
> 
> Thanks for the answer. Although, I could not find the similar topic in the 
>forum 
>
> archives. Maybe I'm not searching well.
> The system relaxed very well. Reading the tutorial I thought it could be the 
> pseudopotential but since the structure is fine I do not think so.
> 
> Cheers,
> 
> Ettore
> 
> 
> ***************************************************************************
> Dr. Ettore Baldini-Neto
> Researcher. Wernher von Braun Center for Advanced Research,
> Campinas, Brazil
> ****************************************************************************
> 
> 
> 
> On Dec 17, 2010, at 8:50 PM, Eyvaz Isaev wrote:
> 
>> Dear Ettore,
>> 
>> 
>> The best hints are 
>> 
>> 1) Searching the forum archive, as the same question was submitted just  few 
>> days ago.
>> 2) Read User's Guide, page 54 (for QE 4.2)
>> 
>> Hope this will help you to find the answer to fix your problem.
>> 
>> Bests,
>> Eyvaz.
>> -------------------------------------------------------------------
>> Prof. Eyvaz Isaev, 
>> Department of Physics, Chemistry, and Biology (IFM), Linkoping University, 
>> Sweden 
>> 
>> Theoretical Physics Department, Moscow State Institute of Steel & Alloys, 
>> Russia, 
>> 
>> isaev at ifm.liu.se, eyvaz_isaev at yahoo.com
>> 
>> 
>> 
>> ----- Original Message ----
>> From: Ettore Baldini Neto <neto.baldini at gmail.com>
>> To: PWSCF Forum <pw_forum at pwscf.org>
>> Sent: Fri, December 17, 2010 11:15:58 PM
>> Subject: [Pw_forum] NEB Calculation with Ti and Graphene
>> 
>> Dear users,
>> 
>> I'm playing with a NEB calculation involving one Ti atom in a graphene 
>> monolayer.
>> The calculation crashes due to diagonalization problems according to the 
>> message 
>> 
>> below.
>> *******************************************
>>  task #         8
>>    from cdiaghg : error #       434
>>    diagonalization (ZHEGV*) failed
>> ********************************************
>> 
>> In what follows I copied my input file (without the atomic positions in order 

>> to 
>> 
>> make it small for the readers)
>> Any hint? Am I doing something very wrong? 
>> I'm not using for now polarization for Ti for now in order to make things 
>> simpler. I'm planning to include it soon.
>> Also, I'm using a large ecut but I figured out that for graphene and for this 

>> pseudo 35-40Ry might work fine.
>> 
>> title = 'Graphene mono layer'
>>  calculation =   "neb"   ,
>>  restart_mode = 'from_scratch' ,
>>  outdir='/home/usuarios/baldini/espresso-4.2.1/tmp/' ,
>>  pseudo_dir = '/home/usuarios/baldini/espresso-4.2.1/pseudo' ,
>>              etot_conv_thr = 1.0E-3  ,
>>              forc_conv_thr = 1.0D-4 ,
>>              nstep=10,
>>              prefix='1MLG_Ti_NEB',
>> /
>> &SYSTEM
>>                      ibrav = 0.,
>>                      ntyp = 2 ,
>>                      nat = 33 ,
>>                      ecutwfc = 50.0 ,
>>                      ecutrho = 400.0,
>>                      occupations = 'smearing' ,
>>                      degauss = 0.02,
>>                      smearing = 'mv' ,
>> /
>> &ELECTRONS
>>            conv_thr  =  1.0d-7  ,
>>            mixing_beta = 0.4 ,
>> /
>> &IONS
>> num_of_images     = 4,
>> opt_scheme        = "broyden",
>> path_thr          = 0.03D0,
>> /
>> ATOMIC_SPECIES
>> C   12.0107   C.pbe-van_ak.UPF
>> Ti  47.8670   Ti.pbe-sp-van_ak.UPF
>> CELL_PARAMETERS
>> 16.12131954   9.383047808   0.0000000
>> 0.00000000  18.658162370   0.0000000
>> 0.0000000    0.000000000  56.6893424
>> ATOMIC_POSITIONS  angstrom 
>> first_image
>> ...
>> last image
>> .......
>> K_POINTS {automatic} 
>> 6.0 6.0 1.0 1.0 1.0 1.0
>> 
>> 
>> Thanks a lot and best regards
>> 
>> ************************************************************************
>> Dr. Ettore Baldini-Neto,
>> Researcher at Wernher von Braun Center of Advanced Research 
>> Campinas, Brazil
>> ***********************************************************************
>> 
>> 
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum
>> 
>> 
>> 
>> 
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://www.democritos.it/mailman/listinfo/pw_forum
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
> 
> 
> 
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

_______________________________________________
Pw_forum mailing list
Pw_forum at pwscf.org
http://www.democritos.it/mailman/listinfo/pw_forum



      



More information about the users mailing list