The unit cell you have defined in your input file you just sent cannot b
right. Think about it- you are defining a simulation cell with dimensions

LatticeConstant     2.46 Ang
%block LatticeVectors
  0.866   0.5   0.000
  -0.866  0.5   0.000
  0.0     0.0   1.0
%endblock LatticeVectors

that means the basis functions of all your atoms will be overlapping in the
z-direction since they will be 2.46 Angstroms away from each other! You
need to increase the size of the z-dimension of your unit cell to at least
12 Angstroms to avoid atom-image interactions. Specifically, the
z-dimension for a 2D material should be more than double the longest basis
set cutoff radius. Additionally, you have defined atomic coordinates that
are greater than your lattice vectors in the xy-directions. That means they
will be translated back onto the unit cell you've defined. This cell you've
defined looks like it might be good for a 2-atom cell, but you've got 54
atoms packed into it, so it's no wonder why your simulation isn't
converging!

There was another message on the list very recently that referenced
graphene. That individual had another issue with his output, but (s)he had
defined the unit cell correctly. Maybe take a look at that for starters.
After you figure out what simulation cell and atomic coordinates to use,
you should optimize the lattice constant using the

MD.VariableCell         .true.

keyword- The cell size will be slightly different for different basis set
sizes and XC-functionals. If you want to find the global minimum in the
potential energy surface, you need to allow the cell size to relax, using
the experimental lattice constant as a starting point. The converged cell
vectors should not differ by more than 2-4% or so compared to the
experimental ones.

You also need to make sure the k-point sampling is very dense around the K
points of the Brillouin zone, since graphene is a semimetal and the bands
cross the Fermi Level at a single point. Use at least a 30x30x1
Monkhorst-Pack k-point mesh.

Best,

On Thu, Nov 8, 2012 at 9:36 AM, 毛飞 <[email protected]> wrote:

> Hi, Heiko Dumlich
>
> Thank you for your kind reply. I have modified the input file for Siesta,
> and it is attached. I am testing the input file if it works. I have two
> questions.
>
> ****
>
> a)    Because graphene is a 2D material, I just want to know whether the
> unit cell defined in the input file is right or not?****
>
> b)    The variable LatticeConstant is set by an experimental value. Maybe
> when I have successfully optimized the geometry,
>
>      the LatticeConstant is changed, then I should set which one to be the
> LatticeConstant, the experimental one, or the optimized one?****
>
> ** **
>
> Any comment or suggestion is appreciated.****
>
> Best
>
>
>
> ****
>
> Fei Mao****
>
>
>
>
> -----原始邮件-----
> *发件人:* "Heiko Dumlich" <[email protected]>
> *发送时间:* 2012年11月8日 星期四
> *收件人:* [email protected]
> *抄送:*
> *主题:* Re: [SIESTA-L] Problem of graphene sheet optimization
>
>
> Dear Fei Mao,
>
> your unit cell does not periodically repeat the graphene structure.
> Further none of your calculations is converged. You should think about your
> unit cell
>
> LatticeConstant     2.44 Ang
> %block LatticeVectors
>   13.000  0.000   0.000
>    0.000  13.000  0.000
>    0.000  0.000   4.100
> %endblock LatticeVectors
>
> and optimize the values to create an infinite graphene sheet. Then the 
> optimization becomes far less demanding, as no dangling bonds exist at the 
> edge. If you desire dangling bonds at the edge, you have to at least converge 
> your calculations to a decent value, which "siesta:   99    -8321.6798    
> -8318.5587    -8318.5639  0.9802 -4.8893" definitely is not true for 
> dDmax=0.9802.
>
>
> Good Luck
>
> Heiko Dumlich
>
> On 11/08/2012 09:50 AM, 毛飞 wrote:
>
> Hi, all****
>
> ** **
>
> Has anyone ever tried to get a well converged geometry optimization of
> pure graphene, which is not passivated with any hydrogen atoms in the edge
> of graphene layer. In that case, the carbon atoms in the edge of graphene
> are kept with dangling bonds. ****
>
> ** **
>
> I have done this job within many codes, and all failed. I find it is very
> hard to do this, but in many published scientific papers, there are the
> energy band analysis of pure graphene (and sometimes adsorbed with adatom),
> such as DOS, or PDOS. It is very amazing, it seems that these authors have
> the pure graphene relaxed fully. Or they obtain the band structure with not
> fully relaxed graphene sheet. Do you have any ideas?****
>
> ** **
>
> I attach the input and output files for optimizing the pure graphene by
> using Siesta codes (in order to obtain the converged geometry optimization,
> I have increased the MeshCutoff to 400 Ry, it still does not work).****
>
> ** **
>
> Any comment or suggestion is appreciated. Thank you.****
>
> ** **
>
> Fei Mao****
>
> Beijing Normal University****
>
>
>
>
> --
> Heiko Dumlich
> Freie Universität Berlin
> Fachbereich Physik - AG Reich
> Arnimallee 14
> D-14195 Berlin
> GERMANY
>
> Phone:   +49-30-838 56157
>
>


-- 
*Abraham Hmiel*
Katherine Belz Groves Fellow in Nanoscience
Xue Group, College of Nanoscale Science and Engineering at SUNY Albany
http://abehmiel.net/about

Responder a