[CPMD-list] About geometry optimization

[Juerg Hutter]

Hi

> Dear CPMD users,
>     I am a novice to CPMD. I am trying to optimize the geometry of Bi2Te3 
> and have some questions for the geometry optimization:
> 1. My understanding for the optimization of geometry is that if only 
> "optimize geometry"  is used, the symmetry and the lattice constant of the 
> cell will not change and only the coordinates of atoms in the cell are 
> changed. Is this correct?

Yes

>
> 2. If I want to do the global optimization for both the lattice constant and 
> the coordinates of atoms, should I just add "STEEPEST DESCENT CELL" except 
> "OPTIMIZE GEOMETRY"?

Yes. However, the cell optimization in CPMD is not very good. If your cell
has a simple geometry (1 or 2 independent parameters) you are better
off doing the optimization by hand.
In addition, don't forget about the coonstant cutoff versus constant 
number of plane waves problem (see old posts about this problem or check
the literature).

>
> 3.The atoms in the cell are charged and the electrostatic interaction is 
> important for Bi2Te3. So if I try the geometry optimization, since the 
> cut-off radius of the electrostatic interaction is around 12 Angstrom, do I 
> need to choose a large cell to include the effects of this long-range 
> interactions? or I can just use the primitive cell.
>

Long range Coulomb interactions are calculated by Ewald sums, no cutoff
is used. I don't know where you got the 12 Angstroms from, but if you
refer to the TESR variable, that's only the real space part.
Again, if this is not clear to you, check the literature, e.g. our
review (see www.cpmd.org for a link) or the book by R. Martin or any
other paper on the plane wave method.


> 4. I would like to get the force constants  for each two-body bond with the 
> optimal geometry for later calculations. But the atoms in a primitive cell 
> are always bonded with the atoms in another primitive cell. If I get the 
> force constant matrix, does it correspond the bonded two atoms? or just 
> correponds to the atoms in the same primitive cell?
>

You can calculate the dynamical matrix of your system (at the gamma point
for the phonons) with CPMD. However, what you get are the force constants
of the normal modes. To assume that those force constants can be used
as bond parameters usually fails. Remember, this is DFT not a pair 
potential method.

regards

Juerg Hutter


> Thanks a lot
>
> Baoling Huang
>
> _________________________________________________________________
> i'm making a difference. Make every IM count for the cause of your choice. 
> Join Now. 
> http://clk.atdmt.com/MSN/go/msnnkwme0080000001msn/direct/01/?href=http://im.live.com/messenger/im/home/?source=hmtagline
>
>