Hi all,
After spending a good day writing letters to european representatives (h=
ope=20
some of you did the same, hum ?), I have put the last touch to the next maj=
or=20
version of Fox.
This version really needs serious testing before release, as there are a=
=20
number of major changes, as you can see below. I'd very much like to see=20
tested:
- your xml files, to see if there is no problem switching to the new globa=
l=20
optimization algorithm
- any organic sample, as this should be much better handled by the new=20
Molecule object. You can convert any ZScatterer to the new Molecule object=
=20
using the import/export menu of the ZScatterer object. This rellay needs=20
testing
- and of course the supplied Fox examples for the new Molecule format (see=
=20
the file below)
This version is roughly 50% faster on powder patterns, and the number of=
=20
trials has also diminished by a factor 2 for the cimetidine and potassium=20
tartrate examples. I hope it can be used to solve more complex samples,=20
especially organic ones.
Many thanks go to Brian Toby for its help on the 3D display of crystal=20
structure & fourier maps, as well as polishing the CIF output function.
The beta version of Fox can be obtained:
=2D for people compiling it, by updating to the latest CVS and re-compiling.
=2D for people under windows and no compiler, get it from the usual places:
ftp://ftp.unige.ch/pub/soft/crystal/fox/test
ftp://ftp.ccp14.ac.uk/ccp14/ftp-mirror/objcryst/test
http://ccp14.sims.nrc.ca/ccp/ccp14/ftp-mirror/objcryst/test/
http://ccp14.semo.edu/ccp/ccp14/ftp-mirror/objcryst/test/
You should get the Fox-20030625CVS.zip, and put the .exe in the same=20
directory as your old Fox.exe.
You can also get the updated examples in examples-molecule.zip.
Note that the new documentation should appear by friday on the Fox websit=
e=20
(delay due to sourceforge), at http://objcryst.sf.net/Fox/
Report to me any success or failure, I intend to do the official release=
=20
sometime next week if everything goes all right.
Cheers,
Vincent
Changelog:
NEW FEATURES-MAJOR CHANGES
* The optimization will now use log(likelihood) as a criterion:
in practice, this means that the Chi^2 statistics are used
rather than the user-chosen cost functions. The likelihood
can sometimes be parametrized in the objects (crystal, pattern),
but not any more in the Global Optimization objects.
* Added a Molecule class, which suprsedes the ZScatterer approach
The scatterer is defined by a list of atoms, and restraints
(bond lengths, angles and dihedral angles) to specify the
geometry. This gives more flexibility, allows a better=20
convergence and allows to correctly define cyclic molecules.
* Added the ability to define a statistical positionnal error
for atoms, which will be taken into account following Maximum
Likelihood principles (ML field in ScatteringPowers), of course
with a significant amount of approximations to be computationnaly
friendly.
* CIF output (with the help of Brian Toby).
OTHER NEW FEATURES
* Faster computation when using integrated profiles for powder
pattern objects
* Error bars are now displayed on the Powder Pattern graphical=20
display.
* Further improvements of the Fourier maps display (mostly from
Brian Toby)
* A cursor can be displayed in the center of the 3D Crystal view,
its coordinates are displayed at the bottom of the window, with
the value of Fourier maps (if any)
* Added atom labels in 3D Crystal view (right-click menu to
deactivate)
* In OpenGL view of a Crystal structure, shift-dragging with left
mouse button now correctly changes the center of the view without
changing the rotation center.
* Changed lighting model in 3D Crystal view
* Added the ability to save structure factors from a
DiffractionSingleCrystal object, including detailed contributions
from different elements (useful for anomalous diffraction
experiments)
=20
BUG FIXES
* Fixed bug when exporting to a Fenske-Hall z-matrix, the 4th atom
was always repeated twice
* Overall ZScatterer occupancy is now correctly taken into account.
* Auto-save of xml files will now use local time rather than gmt.
=2D-=20
Vincent Favre-Nicolin
Universit=E9 Joseph Fourier
http://v.favrenicolin.free.fr
ObjCryst & Fox : http://objcryst.sourceforge.net
|
