[CPMD-list] test calcualtion

[Axel Kohlmeyer]

On 7/17/06, Yi LUO <yi_luo at riken.jp> wrote:
> Dear All,
>
>   I am a begnner of CPMD. I would like to perform MD on Titanocene system.
> Before the MD job, I did test calculation (MD and Geometry Optimization,
> seperately) on TiO molecule. It is found that the optimized Ti-O bond length
> is about 2.0 angstrom, which is longer than that (about 1.6 angstrom)
> obtained from DFT calculation based on the Gaussion Basis function, such as
> B3LYP/6-31G*, LanL2DZ in Gaussion Program. The longer bond length was also

if you _really_ want to compare results, you have to use the same
DFT functional, i.e. BLYP (and not B3LYP). in your case the differences
are significant beyond the differences between functionals...

> obtained for TiO2. The MD on TiO showed the seperation of Ti and O atoms,
> and the Ti-O distance is 4.8 angstrom at 2000 step.
>
>  My questions are:
>
>  1. Are the inputed parameters wrong ? (Input file is attached for your kind
> referece)

there are indeed some issues with your input. see below.

>  2. For a MD job, is the test calcualtion of small molecules the geometrical
> optimization run or MD run?

i don't quite understand what you mean. can you please rephrase.

> &DFT
>   FUNCTIONAL HCTH
>   GC-CUTOFF
>    1.0D-06

note, that raising GC-CUTOFF for HCTH has some
implications on the amount of echange and correlation
included in your calculation. please see the manual for
more details.

>   CUTOFF
>    70.0

for oxides and norm conserving pseudopotentials, a
plane wave cutoff of 70ry is pushing the limit. please
try with about 85ry.

>   ANGSTROM
> &END
>
> &ATOMS
> Oxygen
> *O_MT_HCTH.psp
      KLEINMAN-BYLANDER is missing here ^^^^
CPMD will default to gauss-hermite integration for the non-local
parts which not as efficient as KB-separation, especially if you
do not instruct cpmd to search for more roots.

>    LMAX=P LOC=P
>    1
>    0.096625    0.010126    1.642619
>
> Tatanium
> *Ti_MT_HCTH_NLCC.psp

KB-separation is missing again. but also the pseudopotential should
be used with LMAX=D LOC=S SKIP=P (see the info section and the
CPMD output). but there is another issue. this pseudopotential is a
'minimal' pseudopotential with only 4 valence electrons and as such
best suited for metallic titanium or 'soft' compounds. in a 'hard' oxide
the oxygens will strip off all electrons, and it is difficult to tell, whether
the NLCC will provide enough electron density at the position of the
Ti atoms so that DFT will still work properly. i suggest to you try the
changes from above and see if the result is more reasonable. if it does
not work, you will have to resolve to a semi-core pseudopotential
with more valence electrons, which will (unfortunately) need a much
higher plane wave cutoff (easily about 120ry).

>    LMAX=P LOC=D
>    1
>    0.003375    0.289874    0.057381
>
> &END
> -----------input end for optimization---------


> -----------------input for MD on TiO--------------
>      &CPMD
>   MOLECULAR DYNAMICS
>   restart accumulators wavefunction coordinates velocities cell

if you want to do CP-molecular dynamics, you _have_ to restart
from a converged wavefunction (unless you want to use CP in
combination with annealing to optimize the wavefunction).

>   restart nosec nosep nosee
>   restart latest
>   quench bo
>   ODIIS
>    5
>   MAXSTEP
>    2000
>   STORE
>    5000
>   TIMESTEP
>    5.0
>   EMASS
>    500.0
>   TRAJECTORY XYZ
> &END
>
> &DFT
>   FUNCTIONAL HCTH
>   GC-CUTOFF
>    1.0D-06
> &END
>
> &SYSTEM
>   SYMMETRY
>    0
>   CELL
>    15.000  1.0  1.0  0.0  0.0  0.0
>   CUTOFF
>    40.0

a cutoff of 40.ry is _far_ too low for this kind of system.

>   ANGSTROM
> &END
>
> &ATOMS
> Oxygen
> *O_MT_HCTH.psp
>    LMAX=P LOC=P
>    1
>    0.000000000      0.000000000     -1.174581000
>
> Tatanium
> *Ti_MT_HCTH_NLCC.psp
>    LMAX=P LOC=D
>    1
>    0.000000000      0.000000000      0.427120000
>
> &END
> ----------------------input end for MD----------------
>
> Although I know there are successful MD runs of Ti system. I have no idear
> on the above questions.  Any comments would be very appreciated.

please check out the extensive publication list on the cpmd.org homepage.

ciao,
   axel.

>
> Best Regards
> Yi LUO
>
>
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://cpmd.org/mailman/listinfo/cpmd-list
>
>


-- 
=======================================================================
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.