[CPMD-list] Cell optimization / symmetry

[BOUYER Frédéric 153746]

Dear Axel,

I have noticed that the SYMMETRY and CELL VECTORS are redundant.

Specifying CELL VECTORS with SPEEPEST DESCENT CELL generates the following
error: LATTICE CONSTANT IS NULL ! and stops the program.

Furthermore, do you mean that with ISOTROPIC CELL, the "global shape" of the
cell is conserved during a cell optimization? (~stress tensor diagonal?)

More, during a cell optimization, are the atoms/ions replicated according to
the SYMMETRY used in the input file ? 

Regards,

Frederic Bouyer

-----Message d'origine-----
De : Axel Kohlmeyer [mailto:axel.kohlmeyer at theochem.ruhr-uni-bochum.de] 
Envoyé : mardi 6 avril 2004 14:57
À : 'cpmd-list at cpmd.org'; BOUYER Frédéric 153746
Objet : Re: [CPMD-list] Cell optimization / symmetry

On Mon, 5 Apr 2004, BOUYER Frédéric 153746 wrote:


FB> Dear all,
FB>  
FB> Is the keyword CELL VECTORS (given in &SYSTEM) compatible with the
STEEPEST
FB> DESCENT CELL (given in &CPMD) ?
FB>  
FB> One more question : is the STEEPEST DESCENT CELL keyword works with
SYMMETRY
FB> ?
FB>  
FB> Many thanks in advance for any reply about these keywords.

dear frederic,

do you have any specific problems with those flags?

it 'should' work (i just tried it).
please note, using CELL VECTORS and SYMMETRY are
mutually exclusive (in fact the symmetry is implicitely
set to triclinic IBRAV=14).
 
also STEEPEST DESCEND CELL implicitely sets the symmetry 
to 14 unless ISOTROPIC CELL is also specified.

STEPEST DESCEND CELL has to be used with some care, as cpmd 
does not change the number of plane waves during the calculation, 
so the effective cutoff changes. you therefor have to make sure
that your stress tensor is converged with respect to the
cutoff (i.e. you have to use a rather large cutoff). the situation
is similar to performing constant pressure MD simulations,
which have been discussed before (CUTOFF, use of REFERENCE CELL, etc...
check out the mailing list archives). if you only want to 
optimize the lattice constant, performing a series of single
point calculations with different lattice constants,
is usually much faster and more reliable in that case.


hope this helps,

	axel kohlmeyer.

FB>  
FB> Regards
FB> 

-- 


=======================================================================
Dr. Axel Kohlmeyer                        e-mail: axel.kohlmeyer at rub.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum  http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/
=======================================================================