[sdpd] Time for an indexation Round Robin ?

[Armel Le Bail <armel... @fluo.univ-lemans.fr>]

Dear SDPD Subscribers,

At Glasgow IUCr XVIII were shown only successful
studies. Nobody will invite you if you intend to talk
about failure.

So, now that the IUCr meeting is closed, let us discuss
of really difficult cases ;-).

The SDPD Round Robin last year was certainly presenting
really difficult examples, however, cells were given.

It is well known that indexation can be a problem. The
recent CRYSFIRE suite is undoubtedly a big step toward
a more simple life for people facing indexation challenges.
But the example distributed with the package, with a cell
volume of 200A**3 is rather trivial.

Now, the question is : can the indexing programs provide
cells that do not need any confirmation by a subsequent good
WPPF with cell constraint, and ultimately, a successful
structure determination ?

One of the most difficult cases for indexation is when
two parameters are large and the third is small.

As a possible preliminary Indexation Round Robin (which
I will certainly not organize), I have a vicious case to
submit to your attention.

The sample is an organometallic compound containing Bi.

I failed to index quite good data from a Bruker D8 powder
diffractometer producing minimal FWHM lower than 0.06°
(2-theta). In fact, there were some clues that more than
the 20 first reflections could be indexed in a cell with :
  a = 33.163(3) , b= ?, c = 11.4179(9), alpha=?,
beta = 90.637(5), and gamma=?

One year later, I could record a synchrotron pattern (ESRF,
BM02 beamline). The FWHM is now 0.017° (wavelength
= 0.79764 A). Maybe the sample was not happy to wait for
a year since a few weak reflections that were not seen on the
CuKalpha1+2 pattern are now there. Anyway, I continue to
fail, but the above parameters were nicely fitting the data.

Below is a cell parameters refinement by the ERACEL 
software (data in °theta) :

  H   K   L TH(OBS)  TH-ZERO  TH(CALC)   DIFF.

  2   0   0   1.378    1.379    1.378    .000
  1   0  -1   2.109    2.111    2.110    .001
  1   0   1   2.122    2.123    2.124   -.001
  2   0  -1   2.417    2.418    2.418    .000
  2   0   1   2.441    2.442    2.443   -.001
  4   0   0   2.757    2.758    2.757    .000
  3   0  -1   2.861    2.862    2.862    .000
  3   0   1   2.894    2.895    2.895    .000
  4   0  -1   3.389    3.391    3.390    .000
  4   0   1   3.425    3.426    3.426    .000
  5   0  -1   3.968    3.969    3.968    .001
  1   0  -2   4.058    4.059    4.058    .002
  1   0   2   4.071    4.072    4.073   -.001
  6   0  -1   4.577    4.579    4.578    .000
  6   0   1   4.617    4.618    4.619    .000
  8   0   0   5.520    5.521    5.521    .000
  6   0  -2   5.732    5.733    5.732    .000
  8   0  -1   5.853    5.855    5.854    .001
  1   0  -3   6.046    6.047    6.047    .000
  1   0   3   6.061    6.062    6.063   -.001
  3   0   3   6.385    6.386    6.385    .001
  9   0  -1   6.507    6.508    6.510   -.001
  9   0   1   6.552    6.554    6.552    .001
  4   0  -3   6.593    6.594    6.593    .000
  4   0   3   6.649    6.650    6.649    .001
  6   0  -3   7.271    7.272    7.272    .000
  6   0   3   7.346    7.347    7.348   -.001
  9   0   2   7.440    7.441    7.440    .001
 11   0  -1   7.840    7.842    7.843   -.001
  0   0   4   8.031    8.032    8.032    .000
  2   0   4   8.165    8.166    8.166   -.001
 12   0   0   8.297    8.298    8.298   -.001

With such data, you should not use the default values in
CRYSFIRE. If you tolerate an error of +-0.06°, you will
obtain nothing (MATH error with TREOR and so on).
The max error should be +-0.003 or +-0.002, and the
zeropoint becomes quite decisive at the apparently
negligible value of 0.002° (2-theta) ...

OK, I cannot ensure full purity : this is an experimental case,
you know.

OK, I obtained figures of merit M20 >200 for at least 300
different cells ! Which is the good one ? None was really
satisfying at the stage of WPPF with cell constraint.

I would be happy to obtain experts point of view.

If this challenge interests you, the two powder patterns are
available at :

http://www.cristal.org/sdpd/indexrr.zip

Once unzipped, you will obtain two files : biesrf.rit and
bid8.rit in a format that can be read by DMPLOT or other
software.

Good luck !


Armel Le Bail - Universite du Maine, Laboratoire des Fluorures,
CNRS ESA 6010, Av. O. Messiaen, 72085 Le Mans Cedex 9, France
http://www.cristal.org/