Hi Emilio,
Thanks for your answer. I suppose they are; it's a metallic structure, so
what I did was to first perform an SCF calculation with the atoms and cell
at their ideal values (not relaxing any of them). The first MP mesh I tried
was 6x6x4, and when I saw the high values for the forces, I went through
12x12x8, 18x18x12 and 24x24x16, with the following results for the
monoclinic cell:
6x6x4:
siesta: Atomic forces (eV/Ang):
siesta: 1 -0.004511 -0.031247 2.256165
siesta: 2 0.002122 0.031237 -2.254896
siesta: 3 0.031266 0.004508 2.256000
siesta: 4 -0.031249 -0.002104 -2.254720
siesta: ----------------------------------------
siesta: Tot -0.002373 0.002394 0.002548
12x12x8:
siesta: Atomic forces (eV/Ang):
siesta: 1 -0.001038 -0.002666 2.212205
siesta: 2 -0.001360 0.002672 -2.210901
siesta: 3 0.002669 0.001034 2.212189
siesta: 4 -0.002661 0.001349 -2.210885
siesta: ----------------------------------------
siesta: Tot -0.002390 0.002389 0.002609
18x18x12:
siesta: Atomic forces (eV/Ang):
siesta: 1 -0.001144 -0.008778 2.162036
siesta: 2 -0.001200 0.008782 -2.160760
siesta: 3 0.008796 0.001146 2.162037
siesta: 4 -0.008774 0.001192 -2.160751
siesta: ----------------------------------------
siesta: Tot -0.002323 0.002342 0.002563
24x24x16:
siesta: Atomic forces (eV/Ang):
1 -0.001123 -0.000006 2.124617
2 -0.001151 0.000006 -2.123333
3 0.000019 0.001122 2.124619
4 0.000000 0.001144 -2.123328
----------------------------------------
Tot -0.002255 0.002265 0.002574
----------------------------------------
Max 2.124619
Res 1.226277 sqrt( Sum f_i^2 / 3N )
----------------------------------------
Max 2.124619 constrained
However, when I do an scf for the orthorhombic cell, a 6x6x2 sampling
already gives me much smaller forces in the basal plane than any of the
denser meshes I tried for the monoclinic cell:
siesta: Atomic forces (eV/Ang):
siesta: 1 0.000002 0.000007 2.239400
siesta: 2 -0.000005 0.000005 -2.237123
siesta: 3 0.000000 -0.000009 2.239404
siesta: 4 0.000000 -0.000004 -2.237616
siesta: 5 -0.000001 0.000006 2.238852
siesta: 6 -0.000001 -0.000004 -2.237679
siesta: 7 -0.000006 0.000002 2.239290
siesta: 8 0.000006 -0.000001 -2.237632
siesta: ----------------------------------------
siesta: Tot -0.000003 0.000002 0.006896
I have never really worked with monoclinic cells before, and while I was
already aware that samplings could be different, I never thought I would
have to go to such dense meshes for the smaller monoclinic cell - and still
not get a result as good as the one for the orthorhombic cell. If you'd
like, I can send you my input files - maybe I'm messing up with something
here inadvertently.
Cheers,
Marcos
On Fri, Aug 21, 2015 at 5:29 AM, Emilio Artacho <[email protected]> wrote:
> Hi Marcos
>
> The k-point sampling would in general be different. Are
> both calculations well converged in k points?
>
> Emilio
>
> On Aug 21, 2015, at 1:33 AM, Marcos Veríssimo Alves wrote:
>
> Hi all,
>
> I am running some calculations for a 3D structure made of carbon, using
> Siesta 3.2-pl-5. The executable was compiled with
>
> gfortran using mpif90 -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2
> -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64
> -mtune=generic
>
> however, I am running the code serially, since the structures I'm working
> with have 4 or 8 atoms.
>
> My structure has a monoclinic unit cell of 4 atoms, and when I run an SCF
> for the atomic positions, I get appreciable forces along the basal (xy)
> plane of the cell, which seemed a bot strange to me. When I transformed the
> 4-atom monoclinic cell into an 8-atom orthorhombic one, with exactly the
> same structure, the forces along the basal plane were around 10^-5 eV/Ang,
> which was what I expected.
>
> Has anyone experienced that kind of problem? I'd like to stress that the
> problem is *not* the atomic positions: upon replicating the 4-atom
> supercell properly, I get exactly the same geometry of the 8-atom
> orthorhombic cell. Would anyone have a hint as to why this happens?
>
> Best regards,
>
> Marcos
>
>
>
> --
> Emilio Artacho
>
> CIC nanoGUNE Consolider, and Cavendish Laboratory, University of Cambridge
> Tolosa Hiribidea 76, E-20018 Donostia - San Sebastián, Spain,
> [email protected], +34 943 574039, http://theory.nanogune.eu
>
>
