The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

Hi,
I'm not used to RF calculation and I don't know why the SCF cycle
convergency should be linked with it.
Yet your questions seems to be some kind of FAQ so here are the usual
standard answers:
- Your preconditioner (diemac+dielng) is not smart enough. Since you
use a very small cell you may easily use the RPA approximation of the
dielectric matrix. Try something like "iprcel=[3-4][1-9]" (usualy 4 is
best for the first number and the number of the best converged step
among the 9 first of your calculation (with iprcel 49) is best for the
second) .
If this works no need to worry any more. If it doesn't but seems to
improve greatly the results (this would mean that your system is very
complicated) you might have to tune the code a little bit for a third
computation of the dielectric matrix.
- iscf=5 is usualy a nice algorithm. Yet on some systems it fails
miserably. you may want to try either something more efficient "iscf
7" or something more regular "iscf 2(-3-4)" (I personnaly trust best
iscf 2 at least until reaching a reasonnable convergency).
-If all the previous fails it may mean that  your system has some so
called dangling bonds then it is no wonder that you can not achieve
convergency. Your only hope is to had a "tsmear " and "occopt"
eventhougth this might not be very physical.

regards

PMA


On 9/22/06, zmq.lyao@gmail.com <zmq.lyao@gmail.com> wrote:
> Dear abinit users:
>
> We encounter a server problem when we perform the RF calculation.
> The first step for Ground state calculation is OK. But in the second step for 
> Response Function calculation d/dk, it is very difficulty to converge even 
> after sever hundred cycles. The value of residm varied around 10-4 and 10-5, 
> and is very far from tolwfr=1.0d-22. We do not know how to resolve the 
> problem. Generally, what factors can affect the convergence for the Response 
> Function calculation?
> We appreciated for anyone who could give some help suggestion to figure out 
> the problem.
>
> Bo Liu
>
>
> The input file is attached below.
>
> ndtset  3
> #Response Function calculation for LaBr3
> #Ground state calculation
>   kptopt1   1             # Automatic generation of k points, taking
>                           # into account the symmetry
>   tolvrs1   1.0d-18       # SCF stopping criterion
>     iscf1   5             # Self-consistent calculation, using algorithm 5
>
> #Response Function calculation : d/dk
>   rfelfd2   2
>    rfdir2   1 1 1
>     nqpt2   1
>      qpt2   0.0 0.0 0.0   # This is a calculation at the Gamma point
>   getwfk2   -1            # Uses as input the output wf of the previous 
> dataset
>   kptopt2   2             # Automatic generation of k points,
>     iscf2  -3             # The d/dk perturbation must be treated
>                           # in a non-self-consistent way
>   tolwfr2   1.0d-22       # Must use tolwfr for non-self-consistent 
> calculations
>
> #Response Function calculation : electric field perturbation and phonons
>   rfphon3   1
>  rfatpol3   1 8
>   rfelfd3   3
>    rfdir3   1 1 1
>     nqpt3   1
>      qpt3   0.0 0.0 0.0   # This is a calculation at the Gamma point
>   getwfk3   -2            # Uses as input wfs the output wfs of the dataset 1
>   getddk3   -1            # Uses as input ddk wfs the output of the dataset 2
>   kptopt3   2             # Automatic generation of k points,
>   tolvrs3   1.0d-8
>     iscf3   5             # Self-consistent calculation, using algorithm 5
>
> #Definition of the unit cell
> acell  1.5788197427E+01  1.5788197427E+01  8.9547236657E+00
> angdeg  90 90 120
>
> #Definition of the atom types
> ntypat 2
> znucl 35 57
>
> #Definition of the atoms
> spgroup 176
> natom 8
> #natrd 4
> typat 1 1 1 1 1 1 2 2
> #typat 1 2
>
>       xred    2.8849128174E-01  9.1452153238E-01  7.5000000000E-01
>               6.2603025064E-01  7.1150871826E-01  7.5000000000E-01
>               9.1452153238E-01  6.2603025064E-01  2.5000000000E-01
>               8.5478467616E-02  3.7396974936E-01  7.5000000000E-01
>               3.7396974936E-01  2.8849128174E-01  2.5000000000E-01
>               7.1150871826E-01  8.5478467616E-02  2.5000000000E-01
>               3.3333333333E-01  6.6666666667E-01  2.5000000000E-01
>               6.6666666667E-01  3.3333333333E-01  7.5000000000E-01
>
> #Definition of the planewave basis set
> ecut 40.0
>
> #Definition of the k-point grid
> kptopt 1
> ngkpt    4 4 6
> nshiftk 1
> shiftk 0.0 0.0 0.5
>
> #Definition of the SCF procedure
> nstep 20000
> diemac 12.0


--
Pierre-Matthieu Anglade