** UPPW-11 **


Dear Armel


With the new datasets for UPPW-11, all the previous difficulties
evaporated and both the synchrotron and lab d8 datasets were solved
quickly (5-10 minutes each) using Crysfire's default settings, giving
the same monoclinic cell with volume c.1038 A**3.  Without exception,
this cell correctly predicts various minor profile peaks that were not
included in the peak lists supplied, so I'm very confident that it's
correct.




** Cell **


Synchrotron data (39 WinPLOTR peaks in CELREF v3) - esd's on line 2:


  Zero    Lambda      a       b       c      alpha  beta   gamma   volume
  -0.003 0.69950   8.8805  16.4077   7.1379  90.00  93.86  90.00  1037.691
  0.0025 0.00000   0.0035   0.0063   0.0029   0.000  0.002  0.000    0.036


The same cell was found quickly by KOHL, ITO, FJZN, DICVOL and TREOR,
often several times, with figures of merit of up to 160 (KOHL).



Lab Bruker d8 data (40 WinPLOTR peaks in CELREF v3) - esd's on line 2:


  Zero    Lambda      a       b       c      alpha  beta   gamma   volume
   0.004 1.54060   8.8828  16.4206   7.1489  90.00  93.86  90.00  1040.386
  0.0021 0.00000   0.0012   0.0022   0.0016   0.000  0.002  0.000    0.034


This cell was also found quite quickly, but only by KOHL and DICVOL, with
figures of merit of 50 and 49 respectively. 


Presumably the synchrotron-based version of the refined cell is to be
preferred, though in fact the lab-based cell has smaller esd'd.



** Spacegroup **


Although Chekcell's "best group" module recommended P2(1)/n, P2(1) gave
only slightly calc lines and would be preferred if the molecule is chiral
(I don't have that information).



** Why the previous datasets failed **


With these new results, it immediately became apparent why the original
two datasets failed to index:  both samples contained additional
low-angle lines (presumably from another phase), and were also only
moderately accurate.


Thus, for the 35-line ICDD 43-1748, Bernstein & Zevin (1992) dataset,
only 12 of the first 20 lines belonged to the correct phase, and all the
first 4 lines were extraneous.


Similarly, for the 24-line ICDD 46-1964, Kountourellis (1995) dataset, 3
of the first 4 lines were extraneous.


In addition to these mixed-phase data problems, the cell itself is not
quite straightforward, since it exhibits a dominant row (b is
approximately twice as long as a and c).  That makes it a bit more
difficult for most indexing programs, which generally don't incorporate
special routines to handle this type of case (as most now do for
dominant zones).



With best wishes


Robin Shirley