[Pw_forum] relaxation of even-layer slab

Stefano de Gironcoli degironc at sissa.it
Thu Oct 12 10:14:25 CEST 2006


you could define the positions simmetrically w.r.t the origin, in this way 
there will be no fractional translation along z but only some f.t. along x 
and y, and take care that the nr1 and nr2 are multiples of 3 so that also 
these f.t are allowed.

In any case the fact that a symmetry that should be there is not enforced
(due to grid problems) affects the number of k-points used but should not
change the physics of the system. 
If you see diffrences between the two surfaces this can be due to
1) insufficient self consistency that lead to some noise in the forces, 
hence in the relaxed geometry.
2) charge and potential FFT grid which is not dense enough, so that the
energy slightly depends on the grid origin making the two surfaces 
inequivalent.

stefano



On Thu, 12 Oct 2006, lucky8121 wrote:

> Dear Prof. Baroni£¬
>
> ¡¡Thanks for your prompt reply.
>
> (1) The four-layer slab
>  I did get warnings about the setup of the four-layer slab in the output file:
>     warning: symmetry operation #  4 not allowed.   fractional translation:
>       0.3333333 -0.3333333  0.0000000  in crystal coordinates
>     warning: symmetry operation # 11 not allowed.   fractional translation:
>       0.3333333 -0.3333333  0.0000000  in crystal coordinates
>     warning: symmetry operation # 12 not allowed.   fractional translation:
>       0.3333333 -0.3333333  0.0000000  in crystal coordinates
>     warning: symmetry operation # 13 not allowed.   fractional translation:
>       0.3333333 -0.3333333  0.0000000  in crystal coordinates
>     warning: symmetry operation # 19 not allowed.   fractional translation:
>       0.3333333 -0.3333333  0.0000000  in crystal coordinates
>     warning: symmetry operation # 20 not allowed.   fractional translation:
>       0.3333333 -0.3333333  0.0000000  in crystal coordinates
>
>  The real-space FFT grids in this calculation were:
>     G cutoff =  146.8225  (  47267 G-vectors)     FFT grid: ( 25, 25,180)
>     G cutoff =   73.4112  (  16847 G-vectors)  smooth grid: ( 18, 18,125)
>
> (2) The six-layer slab
>  Then I performed a calculation of a six-layer slab. The initial positions were:
>     Cu   0.5000000      0.2886751     -1.632993
>     Cu   0.0000000      0.5773505     -0.816497
>     Cu   0.0000000      0.0000000      0.000000
>     Cu   0.5000000      0.2886751      0.816497
>     Cu   0.0000000      0.5773503      1.632993
>     Cu   0.0000000      0.0000000      2.449490
>
>  After relaxation, the final positions are:
>     Cu       0.500000000   0.288675100  -1.626940097
>     Cu       0.000000000   0.577350500  -0.819973440
>     Cu       0.000000000   0.000000000  -0.003573650
>     Cu       0.500000000   0.288675100   0.818345337
>     Cu       0.000000000   0.577350300   1.634784508
>     Cu       0.000000000   0.000000000   2.446847342
>
>  The interlayer spacings on the two sides of the slab are 0.812062834 and 0.806966657, respectively. This time the asymmetry becomes larger than the case of four-layer slab.
>
>  I noticed that in the output file there are warnings as below:
>     warning: symmetry operation #  4 not allowed.   fractional translation:
>       0.3333334 -0.3333333 -0.0909088  in crystal coordinates
>     warning: symmetry operation # 11 not allowed.   fractional translation:
>       0.3333334 -0.3333333 -0.0909088  in crystal coordinates
>     warning: symmetry operation # 12 not allowed.   fractional translation:
>       0.3333334 -0.3333333 -0.0909088  in crystal coordinates
>     warning: symmetry operation # 13 not allowed.   fractional translation:
>       0.3333334 -0.3333333 -0.0909088  in crystal coordinates
>     warning: symmetry operation # 19 not allowed.   fractional translation:
>       0.3333334 -0.3333333 -0.0909088  in crystal coordinates
>     warning: symmetry operation # 20 not allowed.   fractional translation:
>       0.3333334 -0.3333333 -0.0909088  in crystal coordinates
>  And the FFT grids are:
>     G cutoff =  146.8225  (  57757 G-vectors)     FFT grid: ( 25, 25,225)
>     G cutoff =   73.4112  (  20595 G-vectors)  smooth grid: ( 18, 18,160)
>
>  Since I have six atom layers and six vacuume layers, it seems that the code translated the FFT grids along the z direction about eleventh of the z-axis length to judge whether the grid (nr3) can coincide with itself. It failed due to 225 can't be divided by 11. I have tried to set "nr1=25, nr2=25, nr3 =220" manually (also nr3=231 and 242), but these FFT grids were not allowed.
>  A solution to this may be to change the FFT library which allows factor of 11, and recompile the PWSCF code. Another solution to this problem may be to reduce or increase the vacuum layers so that the FFT grids are appropriate for the symmetric operations. But since I want to see the dependence of relaxations of surface layers on the slab thickness, I have to fix the vacuum thickness for different slabs. Is there anyway to get symmetric relaxations of different even-layer slabs? For example, is it possible to shift the FFT grid at the setup so that it coincides with symmetry of different slabs?
>
> Best regards,
> Lucky
>
>
> Dear Lucky:
>
>
> the non-symmetric part of the relaxation is very small and possibly not very meaningful. Nevertheless, I agree that if reflection were an exact symmetry of your system, this asymmetry should vanish. One possible explanation could be that reflection symmetry is (slightly) violated by your numerical setup, e.g. the real-space fft grid could be not symmetryc.?/DIV>
>
>
> Stefano
>
>
> On Oct 11, 2006, at 3:14 AM, lucky8121 wrote:
>
>
> Dear users,
>
>
> I am a new user of PWSCF. I am doing the structure relaxation of metal slabs.
> The relaxations (interlayer spacings) on the two sides of the odd-layer slab are symmetric. However, I met problem with the relaxation of even-layer slabs.
>
>
> The initial positions of the Cu(111) slab were:
> Cu ???0.000000000 ?0.577350269190?-1.224744871392
> Cu ???0.000000000 ?0.000000000000?-0.408248290464
> Cu ???0.500000000 ?0.288675134595 ?0.408248290464
> Cu ???0.000000000 ?0.577350269190 ?1.224744871392
>
>
> After relaxation (forc_conv_thr = 4.0D-4), the final positions are:
> Cu ???0.000000000 ?0.577350269?-1.207031882
> Cu ???0.000000000 ?0.000000000?-0.404842681
> Cu ???0.500000000 ?0.288675135 ?0.404671325
> Cu ???0.000000000 ?0.577350269 ?1.207203237
>
>
> As you can see, the relaxations on the two sides of the slab are different. The same case occured whether I set "nosym" as ".true." or ".false.". This is strange. I wonder whether someone else has met the same problem and how he or she solved it. Thanks.
>
>
> By the way, I have used a vacuum of 12.7 angstrom.
>
>
> Best regards
>
>
> lucky
> lucky8121 at sohu.com
> 2006-10-11
>
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
>
> ---
> Stefano Baroni - SISSA?&?DEMOCRITOS National Simulation Center - Trieste
> [+39] 040 3787 406 (tel) -528 (fax) / stefanobaroni (skype)
>
>
> Please, if possible, don't?send me MS Word or PowerPoint attachments
> Why? See:?http://www.gnu.org/philosophy/no-word-attachments.html
>


More information about the users mailing list