Hi Magnus and others,
I do not know about the troubles with PLATON, but I used this case for a
test using the symmetry determination method implemented in SUPERFLIP.
Magnus' first test structure gives in SUPERFLIP (use the 'perform
symmetry' option with a manually generated CIF-file) P21/m as does PLATON.
Interesting is that SUPERFLIP's logfile gives an indication how good or
wrong symmetry elements are with respect to the electronic density. This
is called the 'symmetry agreement factor' (phi_sym). If phi_sym=0 then
the density is perfectly symmetrical with respect to the symmetry
operator. If phi_sym=1.0 then the density is completely random with
respect to the symmetry operator.
For the first test structure phi_sym is situated in between 0.35 and
0.70 for the non-trivial symmetry operators in Cmcm. Is this sufficient
to consider that the operators are present? No, but they are indeed in
the 'grey' area between present and non-present, which we call 'pseudo'.
SUPERFLIP determines Pbnm for Magnus' second structure, with again quite
a number of symops in the grey area between 0.37 and 0.58. The third
structure gives also Pbnm, but now with 'grey' phi_sym values between
0.26 and 0.51 (the c-glide being at 0.26). Note that the SUPERFLIP
threshold for including a symop is set at 0.25.
Note that the C-centering is not detected by SUPERFLIP, although it
should do so if the atoms are placed at the ideal positions. This is
probably a bug.
The interesting point about the SUPERFLIP approach is, which is by the
way fundamentally different from the PLATON approach, that the notion
'pseudo' can now be quantified.
best wishes, Arie
Magnus H. Sørby wrote the following on 23/07/2008 17:31:
Hi!
I was getting some unexpected results from Addsym (Platon version
230608) so I ran some tests to try to clear things up. Now I’m confused
to the point where I need help!
The test structure I have used is:
Space group P-1
a = 3.5 Å b = 12.3 Å c = 4.5 alpha = beta = gamma = 90 deg
Al1 in 0.06 0.85 0.75
Al2 in 0.49 0.33 0.77
This is pseudo-/Cmcm/ symmetry. True /Cmcm/ positions (0 x ¾ , ½ x ¾)
are only 0.24 Å away.
There are three points that are about to drive me nuts:
1) Addsym suggests /P2_1 /m/ instead of /Cmcm/. It has no effect to
increase the criteria parameters d1, d2 and d3 from the default 0.25 Å
to 0.5 Å.
2) If I move the structure somewhat closer to /Cmcm /symmetry by putting
Al1 in, say, 0.03 0.85 0.75, I get no new space group suggestion at
all! Addsym produces some lines that say
“E: SYMM TRANSLATION n/a: 0.958 .500 0”
during the run and writes
“Higher (pseudo)symmetry not compatible,(current symmetry retained)”
in the lis-file.
3) When I move the atoms very close to the Cmcm positions, e.g.
Al1 in 0.01 0.84 0.75
Al2 in 0.49 0.33 0.74
then Addsym suggests /Cmcm/ if I launch it from the command line. If I
use the graphical interface, it suggests /Pmma/ instead. That happens
even if I place the atoms in ideal /Cmcm/ positions.
Are these things user errors or bugs? If they are user errors I would be
very grateful if anybody could explain what I’m doing wrong or point out
something I can read. I’m aware of the Platon/Addsym resources on the
ccp14, but I didn’t find anything there that helped me.
I see on the Platon web page that the newest version should be from July
13 2008, but the version I can download is still from June.
Best wishes,
Magnus
--
***************************************************************************
A. van der Lee
Institut Européen des Membranes
CNRS - UMR 5635
Université de Montpellier II - Case Courrier 047
Place E. Bataillon
34095 MONTPELLIER Cedex 5 - FRANCE
Tel : 33 (0) 4 67 14 91 35
Fax : 33 (0) 4 67 14 91 19
Website X-ray scattering facility ICG/IEM:
http://www.iemm.univ-montp2.fr/xrayweb/main_uk.html
****************************************************************************
