The ABINIT Users Mailing List ( CLOSED )

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

Hi,

1)When you compare calculations made with different cell parameters it
is always better to use ecutsm. Otherwise you loose some precision /
introduce noise.
2) I would recommand you to use optcell and ionmov in conjunction with
converged calculation parameters. You'll gain a lot of time.

regards

PMA

On 5/2/07, #TAO JUNGUANG# <TAOJ0003@ntu.edu.sg> wrote:
> Thanks for the comments.
>
> The input file is given below. I do forget using ecutsm. Because I am doing
> it manually, meaning I change the lattice constant manually, and looking for
> the one who give us lowest energy. I do not rely on optcell and ionmov. In
> this case, still the ecutsm necessary? The reason I do it manually is
> because I found by using optcell and ionmov give me different results. I
> don't know the mechanism behind for optcell and ionmov, I m kind of more
> prefer manually tune the value. ( because you are kind of control it)
>
> Now, I change the ecut to 30 hartree, the results look much better. But the
> optimized lattice constants changed a lot, around 0.04 angstrom.
>
>
>
>
> Any comment on these? 1, how important role the ecutsm play if I do it by
> manually tuning the value? 2, do ecutsm really help without increasing the
> ecut? Because when increase the ecut, the cpu time increase several times.
> If ecutsm is really reliable for doing the job. I prefer use it instead. I
> don't much experience with this, hope you can give some comment. Thank you.
>
>
>
>
>
> ======================================================
>
>
>
> # Crystalline TiO2
>
>
>
> #Optimization of the lattice parameters
>
>
>
> ecut 10.0
>
>
>
> #acell 2*4.59373 2.95812 angstrom
>
>
>
> acell 2*4.55 2.93   angstrom
>
>
>
>
>
> kptopt 1
>
> ngkpt 4 4 6
>
> toldfe 1.0d-10
>
> nband 48
>
> ntypat 2
>
> znucl 22 8
>
>
>
> natom 6
>
> typat 1 1 2 2 2 2
>
> xred
>
>   0.00000000  0.00000000   0.00000000      #Ti
>
>   0.50000000  0.50000000   0.50000000      #Ti
>
>   0.30200746218   0.30200746218   0.000
>
>   -0.30200746218  -0.30200746218  0.000
>
>   0.80200746218   0.19799253782   0.5000
>
>   0.19799253782   0.80200746218   0.500
>
>
>
> nstep 30
>
> diemac 2.0
>
>
>
>  ________________________________
>
>
> From: Xavier Gonze [mailto:gonze@pcpm.ucl.ac.be]
>  Sent: 2007年5月2日 12:28 PM
>  To: forum@abinit.org
>  Subject: [abinit-forum] Lattice optimization
>
>
>
>
> Dear Tao Junguang,
>
>
>
>
>
> You should have sent your input file, for us to comment correctly on that
> result.
>
>
>
>
>
> Well, in the absence of that file ...
>
>
> This effect might be due to the fact that you did not define ecutsm (well,
> you might
>
>
> as well increase ecut to death and obtain something smooth, but a small
> ecutsm
>
>
> will do the work).
>
>
> If you indeed forgot to define ecutsm, you should follow lesson 3 of the
> tutorial, where the way to obtain
>
>
> the lattice parameter is explained in detail.
>
>
>
>
>
> Best,
>
>
> Xavier
>
>
>
>
>
>
> On 30 Apr 2007, at 14:45, #TAO JUNGUANG# wrote:
>
>
>
>
>
>
> When I optimize my rutile tio2 structure use the procedure I mentioned
> previously. The result looks very strange.
>
>
>
> The figure below is the Etotal/Hartree vs the a/Angstrom, while keep c/a
> fixed as experimental value. The total energy minimized at reasonable value.
> But, is the curve supposed to be smooth, because I got a smooth one when
> using other software like pwscf.
>
>
>
>  I think maybe it's the pseudopotential, because at beginning, when I use
> other ones, I can not even get something like this.  The pseudopotential I
> used is o-ps.abinit.paw.lda.nocore and ti-ps.abinit.paw.lda( sorry the files
> is too large to attach), can any one give some comments?
>
>
>
>
>
>
>
>
> <image001.gif>


--
Pierre-Matthieu Anglade