[CPMD-list] Optimization of cell parameters.

[Stefano Piana]

Hi,
I am trying to run an optimization of geometry and cell parameters (I 
have attached at the end of this mail the input file).
The calculation runs fine for the first part, however, when it comes 
close to the end it never converges.
It could simply be the steepest descent algorithm that is not good 
enough, however the variation of the total energy with respect to the 
cell size, in the final part of the calculation worries me a bit.
In practice, if the volume versus total energy is plotted, there is a 
linear relationship between total energy and very small changes in the 
cell size: Etot = E0 + a*(V-V0)
where, in this case, a = 0.05 Hartree/au^3; V0 = 2406.1 au^3; E0 = 
-244.044 Hartree
This behaviour makes me wornder if there is any inconsistency in the 
calculation of the stress tensor. A shrinkage of the cell makes the 
plane waves basis set intrinsically better, however I am quite surprised 
to see an energy difference of 3 kcal/mol for a variation of 0.001% 
(2.0e-4 au) in the cell parameters.
I attach to the mail the input file and the volume with corresponding 
energy calculated for the last step of the geometry optimization (file pip).
Thanks for any comment/suggestion/help,

Stefano

This is the input file I use:
----------------------------------------------------
&CPMD
 OPTIMIZE GEOMETRY
 PARRINELLO-RAHMAN
 TSDC
 LANCZOS  DIAGONALISATION
 LANCZOS PARAMETERS
  1 6 10 1.d-8
 SPLINE RANGE
  6.0
 SPLINE POINTS
  3000
 MAXSTEP
 10000
 RESTFILE
 2
&END
 
&SYSTEM
  POISSON TUCKERMAN
  SYMMETRY
  14
  CHARGE
   0.0
  CELL  DEGREES
    9.811000  0.708694  0.541739  90.000000  91.580000  90.000000
  CUTOFF
  70.0
  ANGSTROM
  KPOINTS MONKHORST-PACK
   1 1 1
  TESR
   4 4 4
&END

[ Atom and System section follows ]
----------------------------

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: pip
Url: http://cpmd.org/pipermail/cpmd-list/attachments/20030826/e9b1330c/attachment.cc