The ABINIT Users Mailing List ( CLOSED )

["Chun Li " <chun.li@mf.mpg.de>]

Dear Marek,

Thanks a lot for your kind help.

I have tried the modified input file you sent to me. The convergence is
much better than before indeed. However, after the dataset 4 came to
Broyden step 6, the similar BUG appeared:

cgwf (electric field) : BUG -
  For k-point #    8 and band #   1,
  the determinant of the overlap matrix is found to be 0.
  Action : contact ABINIT group.

I tried using abinit-4.4.3, the similar BUG appeared after Broyden step
8:

cgwf (electric field) : BUG -
  For k-point #    1 and band #  12,
  the determinant of the overlap matrix is found to be 0.
  Action : contact ABINIT group.

I also tried adding more datasets and making the increase of the
amplitude smoother (add 0.0001 per dataset), the BUG still appeared
after step 7:

cgwf (electric field) : BUG -
  For k-point #    1 and band #  14,
  the determinant of the overlap matrix is found to be 0.
  Action : contact ABINIT group.

I don't think it is because the amplitude of the field is too large, for
the code estimated that the band gap should be larger than 0.64eV, which
is normal for ZnO. Do you have some ideas or just give me some
suggestions? Thanks again.

Wish you a merry Christmas and a happy new year.

Chun Li

-----Original Message-----
From: Marek Veithen [mailto:Marek.Veithen@ulg.ac.be] 
Sent: Wednesday, December 22, 2004 2:53 PM
To: forum@abinit.org
Subject: Re: [abinit-forum] finite electric field calculation

Dear Chun Li,

ABINIT can do structural optimizations at finite electric field.
However, you have to distinguish 2 cases. First, if you relax the
atomic positions keeping constant the lattice parameters, the
calculation
is quite straightforward. Second, if you relax both the atomic positions
and the lattice parameters you have to distiguish between two kinds
of electrical boundary conditions: during the structural optimization,
you
can keep constant the electric field or the potential drop across the
unit
cell as explained in the paper of I. Souza et al, PRL 89, 117602 (2002).
For the moment, the optimization of lattice parameters at finite
electric field is not properly implemented and should be avoided.

Concerning the calculation you did, as a general rule, you must
initialize
a finite field calculation from a calculation at zero field and increase
the amplitude of the electric field in small steps. Moreover, the
electric field must be smaller than the critical field. Please find
enclosed a modified version of your input file.

Best regards,
Marek



Chun Li wrote:

>Dear abinit users,
>
>When I tried to optimize the cell with electric field applied using
>ABINIT 4.4.2, I got the following BUG in the end of the log file:
>
>cgwf (electric field): WARNING -
>  New trial energy at line       4 =  -3.274174E-03
>     is higher than former: -3.304785E-03
>
> cgwf (electric field) : BUG -
>  For k-point #   21 and band #  18,
>  the determinant of the overlap matrix is found to be 0.
>  Action : contact ABINIT group.
>
> Delivered   32 WARNINGs and   2 COMMENTs to log file.
>
> leave_new : decision taken to exit ...
>
>I wonder if ABINIT can perform such kind of calculations. If yes, how
to
>overcome this BUG then? Thanks you. I am eager to receive your reply.
>
>My input:
>  optcell    2
>   ionmov    2
>    ntime    100
>  dilatmx    1.1
>   ecutsm    0.5
>   prtgeo    1
>   efield    0.0 0.0 0.006
>    nband    18         
>   kptopt    3        
>    ngkpt    6 6 4
> berryopt    4
>
>#Common data
>    shiftk    0.0 0.0 0.5
>   nshiftk    1
>     acell    6.15956186 6.15956186 9.83980322
>    diemac    9.0
>      ecut    35
>      iscf    3
>     natom    4 
>     nstep    500
>    ntypat    2
>    occopt    1
>    angdeg    90 90 120
>      xred    
>1.6667200028E-12 -1.6666737435E-12  1.4682831241E-03
>1.6667292546E-12 -1.6666829953E-12  3.8048074342E-01
>6.6666666667E-01  3.3333333333E-01  5.0146828312E-01
>6.6666666667E-01  3.3333333333E-01  8.8048074342E-01
>    toldfe    1.0d-12
>     typat    1 2 1 2
>     znucl    30 8
>
>Best regards.
>
>Chun Li
>
>
>
>  
>


-- 
Marek VEITHEN
Universite de Liege
Institut de Physique, Bat. B5
Allee du 6 aout, 17
B- 4000 Sart Tilman
BELGIUM
 

Phone : ++(32) (0)4-366.36.12
Fax   : ++(32) (0)4-366.29.90
E-mail: Marek.Veithen@ulg.ac.be