[Pw_forum] vc-relax time

[Gabriele Sclauzero]

On 07/13/2010 12:49 PM, Masoud Nahali wrote:
>   Dear  Gabriele Sclauzero and pwscf Users
>    Many thanks for your attentions. The cell dimension and 
> the positions of the atoms are exactly correct.

Are you completely sure? You can find as an attachment a picture (from 
xcrysden) of the central graphene layer according to the positions given 
in your input. You can now judge if it looks like a honeycomb lattice.

> Yes, indeed the high ecutrho is important for ultrasoft 
> pseudopotentials. About the rippling : It was my mistake in selecting 
> a wrong pseudopotential which has a hole. Nicola had explained it 
> before and accordingly I solved it. The graphite surface is not 
> ripple. When I use the "max_second=6000 and dt=150" the job completes 
> very fast as the example of  pwscf. Is the using of such keyboards 
> plausible?

Please spend some time reading Doc/INPUT_PW to understand the meaning of 
the _keywords_. dt is used only for molecular dynamics runs, not 
relaxations, while max_seconds has nothing to do with how fast your job 
is completed.


GS
>  As I mentioned before I had used the optimized cell parameters 
> of (1*1 slab) for vc-relaxing the (2*2 slab) and I expected to see the 
> results very soon but the calculation was time consuming while there 
> was only a very very bit change of the cell dimensions during this 16 
> hours. There is only 1-3 iterations per each step in the output file 
> and each of them was time consuming. The job was completed after about 
> 50 steps.
> -------------------------------------------------------------------------
>
>     Dear Masoud,
>
>       first I would suggest you to use bfgs as the algorithm for both
>     ions and cell dynamics. Excepted particular cases, it should reach
>     the minimum much faster.
>
>     Also, why do you specify the cell with such an unusual way. You
>     simply need celldm(1) and celldm(3) with ibrav=4 if you want to
>     describe an hexagonal lattice. Other suggestions: your ecutrho
>     looks really large to me, do you really need it. On the other
>     hand, degauss might be too large to describe a spin-polarized system.
>
>     Then, are you sure that you have built correctly your supercell?
>     It looks like there are some C-C bonds much shorter that others in
>     the central graphene plane (1.2 instead of 1.4 angs). Please check
>     again your structure.
>     In general, you can expect that if you relax the atoms in the
>     supercell some kind of surface-reconstruction may appear, since
>     you leave more freedom to atoms to rearrange in structures with
>     larger periodicity. I don't think this is the case for graphite,
>     but you may find some ripples (as you mentioned in your earlier
>     emails, if I am not wrong) if the C-C bonds are at a distance
>     shorter than the theoretical equilibrium distance (I guess).
>
>     HTH
>
> ------------------------------------------------------------------------------------------------
>  Dear Quantum Espresso Users
> >
> > I vc-relaxed a (1*1) slab of graphite surface with 3 layers; It 
> takes 20 minutes with parallel running by 4 CPUs. Then I used the 
> exact optimized cell parameters (obtained from vc-relaxed calculation) 
> to make a (2*2) slab of graphite with 3 layers and I expected to see 
> the results in a few minutes. But amazingly it took 17 hours to 
> complete. 48 steps were done in the calculation for vc-relaxing the 
> cell which have the parameters that had been optimized before. The 
> cell parameters only change a very bit in the current vc-relaxing the 
> (2*2) slab. I appreciate if one explain the physical procedure of 
> vc-relaxing and the reason of the time needed for the computation.
> >
> > input file:
> >
> > CONTROL
> >   calculation  = "vc-relax",
> >   pseudo_dir   = "/home/koa/soft/qe4.2/
> > espresso-4.2/pseudo",
> >   outdir       = "/home/koa/tmp",
> >   etot_conv_thr= 1.0D-4,
> >   forc_conv_thr= 1.0D-3,
> >   dt=80,
> >     /
> > &SYSTEM
> >   ibrav     = 4,
> >   a         = 2.4579,
> >   b         = 2.4579,
> >   c         = 16.3069,
> >   cosab     = -0.5,
> >   cosac     = 1.0,
> >   cosbc     = 1.0,
> >   nat       = 6,
> >   ntyp      = 1,
> >   ecutwfc   = 40.D0,
> >   ecutrho   = 480.D0,
> >   occupations = 'smearing'
> >   smearing ='mp',
> >   degauss = 0.03,
> >   nspin = 2,
> >   starting_magnetization(1)= 0.003,
> >   london=.true.,
> >   /
> > &ELECTRONS
> >   conv_thr    = 1.D-6,
> >   mixing_beta = 0.7D0,
> >   diagonalization = "david",
> > /
> > &IONS
> >  ion_dynamics="cg"
> > /
> > &CELL
> > cell_dynamics = 'damp-w',
> > press = 0.0,
> > /
> > ATOMIC_SPECIES
> > C  12.0107  C.pbe-rrkjus.UPF
> > ATOMIC_POSITIONS {angstrom}
> >  C                  0.00000000    0.00000000    0.00000000  1 1 0
> >  C                  0.00000000    1.41908472    0.00000000
> >  C                  0.00000000    0.00000000    3.15347111
> >  C                  11.22896342    0.70954236    3.15347111
> >  C                  0.00000000    0.00000000    6.30694222
> >  C                  0.00000000    1.41908472    6.30694222
> > K_POINTS {automatic}
> > 4 4 1 1 1 1
> >
>  Sincerely Yours
>  Masoud Nahali
>  SUT
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>    


-- 

Gabriele Sclauzero, EPFL SB ITP CSEA
PH H2 462, Station 3, CH-1015 Lausanne

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20100713/3ac318a1/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: layer2.png
Type: image/png
Size: 92726 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20100713/3ac318a1/attachment.png>