The ABINIT Users Mailing List ( CLOSED )

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

Hi,

OK, it looks like your system is a difficult one. Please send the
various input and relevant part of the output so that we can give you
a better diagnose (just the change with respect to the first one you
send will be sufficient). I'm quite astonished that you do not get
perfect convergency following my previous advices. Is it possible that
you did a mistake ? Or are you using PAW ?

About parallel computing:
- If you use a single k-point, you must rely on the band FFT. It
requires a fast network. Forget about it if you use ethernet.
-If you have a few k-point use as much processor as k-point .
Parallelism will decrease the time per step. But never enhance the
convergence you get.

Regards

PMA
On 8/29/07, Hua Bao <hbao@purdue.edu> wrote:
> Dear all,
>
> Does parallel computing help on this kind of calculation? I mean the SCF
> calculation of a molecule/cluster in a large box.
>
> Thanks,
> Hua
>
> Anglade Pierre-Matthieu wrote:
> > Don't worry too much for positive deltaE as long ad you get some signs
> > of convergence in less than "npulayit" steps.
> > If you really require a steady convergence you can just use:
> > iprcel 141
> > The computation will converge steadily but every step will last twice
> > as long as before.
> >
> > On 8/28/07, Hua Bao <hbao@purdue.edu> wrote:
> >
> >> Dear Anglade and all,
> >>
> >> Thank you for your suggestion.
> >> Here I've got some positive deltaE while doing SCF calculation (using
> >> iscf 5 or 7) . I know this happens. How to deal with it? It seems it
> >> will fluctuating for ever?
> >>
> >> Thanks,
> >> Hua
> >>
> >>
> >>
> >>
> >> Anglade Pierre-Matthieu wrote:
> >>
> >>> Hi,
> >>>
> >>> To get faster convergence within Abinit you can use:
> >>> 1) better mixing algorithm like pulay:
> >>> iscf 7
> >>> npulayit 20
> >>>
> >>> 2) preconditionners
> >>> a) try
> >>> iprcel 45
> >>> (& diecut 1  if you feel preconditionning is too long).
> >>> It's based on KM Ho preconditioner (prb 25 4260 1982)  and is very 
> >>> efficient
> >>> b) also you can just tune the default preconditioner based on Kerker's
> >>> method (PRB 23 3082 1981)
> >>> by setting more accurate value for diemac (try something around diemac 
> >>> 50).
> >>>
> >>> 3) Also make sure you get both convergence for the potential (vres2)
> >>> and for the wave functions (residm). With the suggested change you'll
> >>> get a very good convergence of the first one and it is likely that you
> >>> need to increase "nline" to get the other.
> >>>
> >>> regards
> >>>
> >>> PMA
> >>>
> >>> On 8/28/07, Hua Bao <hbao@purdue.edu> wrote:
> >>>
> >>>
> >>>> Dear all,
> >>>>
> >>>> First, I have a general question. How to get a faster SCF
> >>>> calculation? I appreciate if you can point out which literatures to
> >>>> refer.
> >>>>
> >>>> I am doing a geometry optimization of TiO2 cluster contains less than
> >>>> 40 atoms. The SCF calculation is really slow. It take a few hundreds
> >>>> steps to get convergent. The energy sometimes is fluctuating. So I
> >>>> expect it would take a really long time to achieve geometry
> >>>> optimization. Parallel running does not help too much, either. Since
> >>>> my initial geometry is a little far from optimization. Is there any
> >>>> faster way to get a optimized structure?
> >>>>
> >>>> Here is my input file, NCpp is used.
> >>>>
> >>>> ionmov  3              # Use the modified Broyden algorithm
> >>>> ntime   1000             # Maximum number of Broyden "timesteps"
> >>>> tolmxf  1.0d-4         # Stopping criterion for the geometry
> >>>> optimization : when
> >>>> iscf    5
> >>>> chkprim 1
> >>>>
> >>>> toldff  1.0d-4         # Will stop the SCF cycle when, twice in a row,
> >>>>                         # the difference between two consecutive
> >>>> evaluations of
> >>>>                         # forces differ by less than toldff (in
> >>>> Hartree/Bohr)
> >>>>
> >>>>
> >>>> xangst   9.186000    9.186000   11.779326
> >>>> .......
> >>>> acell  15 15 15  angstrom  # The keyword "acell" refers to the
> >>>> angdeg 90 90 90
> >>>>
> >>>> ntypat 2          #
> >>>> znucl 22 8      # The keyword "znucl" refers to the atomic number of 
> >>>> the
> >>>>
> >>>>
> >>>> #Definition of the atoms
> >>>> natom 33
> >>>> typat 1 2 2 1 2 1 2 2 2 1 2 2 2 1 2 2 1 2 2 2 1 2 1 2 2 1 2 1 2 1 2 2 2
> >>>>
> >>>> #Definition of the planewave basis set
> >>>> ecut 30.0         # Maximal kinetic energy cut-off, in Hartree
> >>>>
> >>>> #Definition of the k-point grid
> >>>> nkpt 1
> >>>> #Definition of the SCF procedure
> >>>> nstep 500          # Maximal number of SCF cycles
> >>>>
> >>>>
> >>>> Thanks,
> >>>>
> >>>> Hua Bao
> >>>> Graduate Student
> >>>> School of Mechanical Engineering,
> >>>> Purdue University
> >>>> O: 765-494-6944
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>>
> >>
> >
> >
> >
>
>


-- 
Pierre-Matthieu Anglade