[CPMD-list] Problem with Al pseudopotential

[Axel Kohlmeyer]

On Mon, 12 Feb 2007, Julen Larrucea wrote:

JL> Dear CPMD users,
JL> 
JL> I'm trying to perform a simulation with H, O, C, N and Al. First of
JL> all I started by checking the pseudopotentials by comparing the
JL> results of a geometry optimization with some other geometry
JL> optimization. In this case I optimized an Al(OH)3 with Gaussian 03, I
JL> took the last geometry and started a geometry optimization with CPMD
JL> (input below).

julen,

i am not sure if the Al potential that you are using is suitable 
for compound where Al has a 3+ configuration. to make sure that
that uspp file is not broken, you may try using an equivalent
norm-conserving pseudopotential. a 3electron NC Al pseudo should 
be 'soft' enough that you can use it with 50ry (or less).
if it also does not work, you may need an 11-e pseudo (which will
be very hard if norm conserving, or has to be built from scratch
in case of ultra-soft).

JL> 
JL> In the result the molecule breaks its Al···O bonds, which doesn't make sense!
JL> 
JL> I have no problem with H and O pseudopotentials. I already have used
JL> both ASCII and binary ones, but the problems start when I add Al.
JL> 
JL> If I just optimize the wave function and then run the Molecular
JL> Dynamics, the system explodes. I set the cutoff to 50 Ry to ensure
JL> maximum accuracy.

for ultra-soft pseudopotentials this does not always result in
much higher accruracy. however, since CPMD does not have a (working)
dual-grid scheme, you should try cranking up the density cutoff,
e.g. by setting DUAL to 6.0 or so.

JL> I'm using CPMD-3.11.1 and I have tried it on Xeon, Itanium 2 and Core
JL> 2 Duo machines and I got the same (bad) results in all of them.

the problem that you are reporting is unlikely to be machine specific.

cheers,
   axel.

JL> 
JL> Thanks in advance,
JL> 
JL>      Julen
JL> 
JL> ------------------------------------------------------------------------------------------------------------
JL> &CPMD
JL>   OPTIMIZE {GEOMETRY,XYZ}
JL>   MEMORY BIG
JL>   CONVERGENCE
JL>   1.0e-5  1.0e-4
JL> &END
JL> 
JL> &SYSTEM
JL>   ANGSTROM
JL>   SYMMETRY
JL>   1
JL>   CELL
JL>   5.0  1.0  1.0  0  0  0
JL>   CUTOFF
JL>   50.0
JL> &END
JL> 
JL> &ATOMS
JL> 
JL> *H_VDB_BLYP.psp   NEWF FORMATTED
JL>    LMAX=S
JL>     3
JL>     0.808295     0.000000     2.236960
JL>     1.512917     0.007121    -1.825982
JL>    -2.340615    -0.025809    -0.418771
JL> 
JL> *O_VDB_BLYP.psp   NEWF FORMATTED
JL>    LMAX=P
JL>    3
JL>     0.000000     0.000000     1.709564
JL>    -1.480118    -0.014026    -0.855939
JL>     1.477794     0.014800    -0.861083
JL> 
JL> *Al_VDB_PBE_NLCC.psp NEWF FORMATTED
JL>   LMAX=D
JL>    1
JL>      0.000000     0.000000     0.000000
JL> 
JL> &END
JL> 
JL> &DFT
JL>   NEWCODE
JL>   FUNCTIONAL BLYP
JL> &END
JL> ------------------------------------------------------------------------------------------------------------
JL> 
JL> 
JL> 

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