[CPMD-list] too long Si-O bond-length from geometry optimization

[Axel Kohlmeyer]

On Fri, 4 May 2007, liu liuxiandong wrote:

LX> Dear CPMDers,
LX> 
LX> I used CPMD to optimize a mineral containing Si, Al, O and H within a fixed
LX> cell. The bond lenths of Al-O, O-H are very close to the experimental
LX> values, whereas the Si-O lengths are about 1.72 angstrom, much longer than
LX> the experimental value (about 1.61 angstrom). Could you give some
LX> suggestions ?

dear ???,

did you check for basis set (cutoff) convergence?

please note, 70ry is the absolute minimal cutoff for 
an MD with oxygen atoms at room temperature, where 
the accurate position of the potential minimum does 
not matter that much.

did you check for k-point convergence?
you unit cell is rather small.

did you check whether a Si potential containing
the d-channel makes a difference? if my memory of
undergraduate chemistry serves me right, it should
in this case (hi ari!).

bala already noted the typo(?). in that case CPMD will 
fall back to (more costly) gauss-hermite integration
for which recent versions should have reasonable defaults.

greetings from sunny philadelphia,
    axel.

LX> The following is part of the input file.
LX> &CPMD
LX> 
LX> 
LX> 
LX>   OPTIMIZE GEOMETRY
LX> XYZ
LX>   HESSIAN
LX> UNIT
LX> 
LX> 
LX> 
LX>   CONVERGENCE
LX> ADAPT
LX> 
LX>   0.02
LX> 
LX>   CONVERGENCE
LX> INITIAL
LX> 
LX>   1.0D-4
LX> 
LX>   CONVERGENCE
LX> ORBITALS
LX>   5.0D-5
LX> 
LX>   CONVERGENCE
LX> GEOMETRY
LX>   4.0D-4
LX> 
LX> 
LX> 
LX>     ODIIS
LX> NO_RESET=40
LX> 
LX> 
LX> 5
LX> 
LX> 
LX> 
LX>     LBFGS
LX> NREM
LX> 
LX> 
LX> 20
LX> 
LX> 
LX> &END
LX> 
LX> 
LX> 
LX> &DFT
LX> 
LX>   FUNCTIONAL
LX> BLYP
LX> 
LX> 
LX> 
LX> 
LX> 
LX> &END
LX> 
LX> 
LX> 
LX> &SYSTEM
LX> 
LX> 
LX> 
LX> 
LX> SYMMETRY
LX> 
LX> 
LX> 14
LX> 
LX> 
LX> ANGSTROM
LX> 
LX>   CELL ABSOLUTE
LX> DEGREE
LX> 
LX>    5.1535 8.9419 7.3906 91.9260 105.0460 89.7970
LX> 
LX> 
LX> CUTOFF
LX> 
LX>    70.0
LX> 
LX> &END
LX> 
LX> 
LX> 
LX> &ATOMS
LX> 
LX> *O_MT_BLYP.psp
LX> KLEINMAN-BYLANDER
LX> 
LX>  LMAX=P
LX> 
LX>  18
LX> ...
LX> 
LX> 
LX> *H_MT_BLYP.psp
LX> KLEINMAN-BYLANDER
LX> 
LX>  LMAX=S
LX> 
LX>  8
LX> ...
LX> 
LX> 
LX> *Si_MT_BLYP.psp
LX> KLEINMAN-BYLANDER
LX> 
LX>  LMAX=P
LX> 
LX>  4
LX> ...
LX> 
LX> 
LX> *Al_MT_GIA_BLYP.psp
LX> KLEINMAN-YLANDER
LX>  LMAX=D
LX> LOC=0
LX> 
LX>  4
LX> ...
LX> 
LX> 
LX> &END
LX> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.