The ABINIT Users Mailing List ( CLOSED )

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

Hi,

I see (or is it my mistake) that you are using the default Abinit preconditioner. This is very efficient for bulk metal calculation. But for surface or non metal calculation you have to perform some tuning work because other wise you prevent the actual convergency of your calculation.
Your best tuning oprion may be to use keyword "iprcel". The RPA approximation of the dielectric matrix should perform fine. You can also use the other related keywords: diemac, dielng. But you won't get better result with that. 
Also, on very long cell, it happens that the CG on WF does not perform perfectly. Then you might get better convergency with "nnsclo 2" and "nline 6".

Last but not least (yet I assumed this is already done) you must be sure of your:
a) ecut convergency (Hydrogen usualy require a higher ecut I think)
a)k-point convergency at every SCF cycle (check the change of kptrlen in the output).

regards

PMA

On 11/23/06, Yunhee Chang <yunhee.chang@kriss.re.kr> wrote:

Dear Xhang,

Thank you for your comments.
But I used same stop-criterion for SCF cycle.
And the step is stopped at nstep=28.

Is it a problem related on SCF methods?
I'm checking..^^

Thank you again.

Best regards,
Yunhee Chang

-----Original Message-----
From: Jin Zhang [mailto:jin.zhang.pku@gmail.com]
Sent: Thursday, November 23, 2006 1:39 PM
To: forum@abinit.org
Subject: Re: [abinit-forum] SCF calculation from atomic relaxed coordinates

Dear chang,

Did you use different stop-criterions for scf cycle. I suspect the
difference comes from these settings:
-----
toldff 2.3d-6
# tolvrs 1.0d-6
----

BTW: maybe nstep=30 is a little small for a slab to reach convergence.

Best,

--
Jin Zhang
Dep. of Physics,
Peking University

On 11/23/06, yunhee.chang@kriss.re.kr <yunhee.chang@kriss.re.kr> wrote:
> Dear Abinit users,
>
> I'm calculating the In on Si(111) surface system.
>
> When I calculate the total energy using the fully relaxed atomic
> coordinates,
> I cannot reach the same total energy and force.
> I didn't use the previous WFK files,
> I just do SCF cycle to get ground state
> using relaxed atomic coordinates.
>
> The relaxed force is reached about 0.001 eV/A,
> but in SCF calculation using this coordinates,
> the force is reaching about 0.05 eV/A.
>
> That means electronic SCF minimization has some problem..??
> I increased the nband, but it's still same.
> Do I have to choose another SCF cycle?
> Would you give me any comments?
>
> This is my input file.
>
>
============================================================================
==
> # Si(111)-4x1: 3BL slab
> # A first step of atomic relaxation and total energy
>
> # Definition of occupation number
>   occopt 3
>   tsmear 0.0007
>
>   nband 80
>
> #Definition of the unit cell
>   acell  2.105671378027334         3.647129810786902
> 1.488934510353547
>   rprim 12.000000000000000         0.000000000000000
> 0.000000000000000
>          0.000000000000000         2.000000000000000
> 0.000000000000000
>          0.000000000000000          0.000000000000000
> 40.000000000000000
>   chkprim 0              # This input variable allows to use non-primitive
> unit
>                          # cells. Please, do not use it in other cases,
>                          # you might miss a primitive cell, faster to
> handle.
>
> #Definition of the atom types
>   ntypat 4           # There is only one type of atom
>   znucl  49 14 14 1  # The keyword "znucl" refers to the atomic number of
> the
>                      # possible type(s) of atom. The pseudopotential(s)
>                      # mentioned in the "files" file must correspond
>                      # to the type(s) of atom. Here, the only type is
> Aluminum
>
>
> #Definition of the atoms
>   natom 34                  # There are two atoms in this doubled cell
>   typat 4*1 2*2 24*3 4*4    # These atoms are of type 1, that is, Aluminum
>
> #Define the methodology to find the minimum
> # ionmov  4              # Use the conjugate gradient algorithm
> # ntime   30             # Maximum number of Broyden "timesteps"
> # tolmxf   2.3d-5         # Stopping criterion for the geometry
optimization
> : when
>                          # the residual forces are less than tolmxf, the
> Broyden
>                          # algorithm can stop
>   natfix   12            # Fix the position of two symmetry-equivalent
atoms
>                          #  in doing the structural optimization
>   iatfix   23 24 25 26 27 28 29 30 31 32 33 34
>                          # Choose atoms 1 and 2 as the fixed atoms (see
> discussion)
>   xangst                 # This keyword indicate that the location of the
> atoms
>                          # will follow, one triplet of number for each
atom
>           4.81208922716606     0.00000000000000    18.96242987264197
>            7.06126514599433     1.92997798069774    18.44619525314767
>            9.54107109659771     0.00000000000000    18.50244627453926
>           11.79858364903775     1.92997798069774    18.89701241902430
>            0.93752037965351     1.92997798069774    17.98049761864093
>            2.32517430108428     0.00000000000000    18.02460434084392
>             0.06883434180141     1.92997798069774    15.70707828157864
>            3.24669481604957     0.00000000000000    15.78259586066670
>            6.69653954661645     1.92997798069774    15.76397531371200
>           10.00849002402831     0.00000000000000    15.82662540146863
>            1.09655439061929     0.00000000000000    14.82273027715486
>            4.42974022967583     1.92997798069774    15.06176351602779
>             7.80109204091853     0.00000000000000    14.94996234174407
>           11.13930017846617     1.92997798069774    15.05651503941083
>            1.10445225490822     0.00000000000000    12.50168835550348
>             4.45211632475764     1.92997798069774    12.65811154209729
>            7.79905934629338     0.00000000000000    12.59472103448601
>           11.14927953093290     1.92997798069774    12.65600080591594
>             2.24677893930407     1.92997798069774    11.79467792107055
>            5.56327379886284     0.00000000000000    11.83351245642332
>            8.92328674704478     1.92997798069774    11.82214274533964
>           12.23084192705557     0.00000000000000    11.81163195386842
>            2.22854661337181     1.92997798069774     9.45492253503307
>            5.57136653342950     0.00000000000000     9.45492253503307
>             8.91418645348719     1.92997798069774     9.45492253503307
>           12.25700637354488     0.00000000000000     9.45492253503307
>            0.00000000000000     1.92997798069774     8.66701232378028
>             3.34281992005769     0.00000000000000     8.66701232378028
>            6.68563984011538     1.92997798069774     8.66701232378028
>           10.02845976017307     0.00000000000000     8.66701232378028
>             0.00000000000000     1.92997798069774     7.20149933085016
>            3.34281992005769     0.00000000000000     7.20149933085016
>            6.68563984011538     1.92997798069774     7.20149933085016
>           10.02845976017307     0.00000000000000     7.20149933085016
>
> # getwfk -1
>
> #Definition of the k-point grids
>   kptopt 1             # Option for the automatic generation of k points,
>   ngkpt  2 8 1
>   nshiftk 1
>   shiftk  0.5 0.5 0.0
>
> #Exchange-correlation functional
>   ixc 11               # GGA Perdew-Burke-Ernzerhof GGA functional
>
> #Definition of the planewave basis set
>   ecut 16.0            # Maximal kinetic energy cut-off, in Hartree
>
> #Definition of the SCF procedure
>   nstep 30             # Maximal number of SCF cycles
>   toldff 2.3d-6
> # tolvrs 1.0d-6
>
>   prtden 1
>   prtdos 1
>   prtgeo 1
>   prtwf 1
>
============================================================================
>
>
> Best regards,
> Yunhee Chang
>

-- 
Pierre-Matthieu Anglade