[Pw_forum] Pw_forum Digest, Vol 28, Issue 38

[Q.J.Wang]

>Message: 2
>Date: Thu, 15 Oct 2009 15:42:34 +0200
>From: "Lorenzo Paulatto" <paulatto at sissa.it>
>Subject: Re: [Pw_forum] total force ascend in the course of VC-relax
>To: "PWSCF Forum" <pw_forum at pwscf.org>
>Message-ID: <op.u1ugc8i1a8x26q at paulax>
>Content-Type: text/plain; charset=utf-8; format=flowed; delsp=yes
>
>In data 15 ottobre 2009 alle ore 15:38:39, Q.J.Wang <wangqj1 at 126.com> ha  
>scritto:
>
>> Dear all
>>     I did geometry optimization with vc-relax in a supercell. When I  
>> increase the concentration of doped atoms ,the value of total force  
>> ascend instead of declining.Before I did not increase the concentration  
>> of doped atoms ,It converges well .Anyone who encounter the same problem  
>> as me ?
>
>I sounds normal to me, when you perturb the system you will have to redo  
>the relaxation... why do you find it puzzling?
The problem is when I redo the relaxation ,the value of 'Total force'and the 'P' become large in every bfgs cycle .And at last shows error :
                            from bfgs : error #         1
           bfgs history already reset at previous step
The input file as following :
   &CONTROL
                 calculation = 'vc-relax' ,
                restart_mode = 'from_scratch' ,
                      outdir = '/root/work3/' ,
                  pseudo_dir = '/root/pseudo/' ,
                      prefix = 'GaAs' ,
                        nstep = 250 ,
                     tstress = .true. ,
                     tprnfor = .true. ,
 /
 &SYSTEM                 
                         nat = 48,
                        ntyp = 4,
                     ecutwfc = 30 ,
                     ecutrho = 300 ,
                       nosym = .true. ,
                 occupations = 'smearing' ,
                     degauss = 0.005 ,
                    smearing = 'gaussian' ,
                       nspin = 2 ,
   starting_magnetization(1) = 0.5,
   starting_magnetization(2) = 0.5,
   starting_magnetization(3) = 0.5,
   starting_magnetization(4) = 0.5,
                  lda_plus_u = .false. ,
 /
 &ELECTRONS
                    conv_thr = 1.D-7 ,
                 mixing_mode = 'plain' ,
                 mixing_beta = 0.3 ,
 /
 &IONS
                ion_dynamics = 'bfgs' ,
                 phase_space = 'full' ,
           pot_extrapolation = 'second_order' ,
           wfc_extrapolation = 'second_order' ,
                     upscale = 100 ,
 /
 &CELL
               cell_dynamics = 'bfgs' ,
 /
ATOMIC_SPECIES
Ga   ...........
ATOMIC_POSITIONS angstrom 
...............
  K_POINTS automatic 
  3 3 3   1 1 1 


>
>
>>     The second problem is when I increase the relative distance of the  
>> two doped atoms in the supercell and did geometry optimization with  
>> vc-relax,and the error turn up as like : from c_bands : error #         1
>>                         too many bands are not converged
>> So I am puzzled why before I increase relative position of two doped  
>> atoms the vc-relax converged and the latter turns up error in the same  
>> parameter .
>
>It happens with peculiar configurations, the easiest way to get rid of it  
>is normally to increase the number of bands.
>
>cheers

>-- 
Best regards
 
Q.J.Wang
 
XiangTan University 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20091016/a21b3671/attachment-0001.htm