Dear Armel

Happily your UPP-1 dataset did not present a difficult problem and I
expect that most if not all participants will succeed with it.  The
solution which I am proposing with some confidence is as follows:

** Calibration

2Theta zero correction (to be added) = -0.018 deg (std error 0.008 deg)
[despite the use of an Al standard!] (Celref refinement)


** Symmetry

Monoclinic: P2(1)/a (using the setting reported below)

** Cell

Unit cell refined vs 2Theta (using Celref), including 2Theta zero
correction refinement, with standard deviations (i.e. standard errors)
in brackets:

     a         b         c     alpha   beta    gamma       V

  13.8090    8.5019   10.6196    90    98.220    90      1233.96
  (0.0076)  (0.0048)  (0.0065)    -    (0.005)    -       (0.12)

This cell indexes all 28 observed lines using only 48 calculated lines,
with a maximum 2Theta difference of 0.031 deg (line 6) and mean 2Theta
difference of 0.0003 deg, and in my judgement is almost certain to be
correct (assuming that no unreported supercell lines are present).

Its figure of merit M1(20) is 21.68, with M1(28) = 17.81 for all lines,
without allowing for the P2(1)/a absences.

The inclusion of these absences would raise M1(20) to approximately 29
and M1(28) to approximately 22.

Higher figures of merit of 25-28 were reported by the indexing programs
(see the table below), but those were obtained by refining against Q for
the first 20 lines only.  The corresponding cells give higher standard
errors in 2Theta and are a little less physically meaningful than the
refined cell proposed above.

Marginally smaller standard errors can be obtained by refining a specimen
eccentricity correction ("e/R" in Celref) instead of a 2Theta zero
correction, but the resulting improvement is too small to be significant.
Hence, for simplicity, the Celref results from refining a constant 2Theta
zero correction have been reported here.

** Indexing and refinement procedure

Self-calibration for 2Theta zero gave a correction of approx -0.04 deg
(based on 3 line-pairs), which was used for the main indexing searches
(although in this case the cell can be found by e.e. FJZN without
applying the correction).

Indexing, etc. was then carried out within Crysfire 2003 v9.50.03
(32-bit), followed by space group determination (using Chekcell) and
cell refinement (Celref v3 within Chekcell).  The search and refinement
process took rather less than an hour of elapsed time.

** Indexing programs used, with results

To save time, each program was run using its Crysfire defaults, and since
this was sufficient for three programs to find an apparently correct
solution, other options were not explored.  Thus it remains possible that
additional programs might have succeeded with suitably adjusted search
parameters.

1) Using the original dataset:

TAUP v3.3a    Correctly reported no solutions orthorhombic or higher
FJZN v6.22a   Correct cell with I20 = 20, "M20" = 22.76, V = 1230.203

2) Using Z2Th = -0.04 deg from self-calibration (hence volume increase):

FJZN v6.22a   Correct cell with I20 = 20, "M20" = 28.60, V = 1236.890
KOHL v7.01b   Correct cell with I20 = 20, "M20" = 28.53, V = 1236.901
ITO12         Correct cell with I20 = 20, "M20" = 24.6,  V = 1231.06

KOHL also reported a 2V cell with b doubled (I20 = 20, "M20" = 17.09)

TREOR90 reported a non-equivalent 2V cell (I20 = 20, "M20" = 9)

DICVOL91 missed the correct cell and reported two non-equivalent 2V cells
(I20 = 20, "M20" = 9.8 & 9.7)

LZON v6.23b failed to index this dataset using its default settings.

Since a very plausible cell had been found by three programs (with 2V
and 4V derivative cells reported by two more), no further indexing
runs were used.

Crysfire cell data (CE) and indexed pattern (CA) listings are attached
[in file UPP1-CRZ.CFO].  The latter does not allow for P2(1)/a absences.

Also the refined cell, etc., as a Crysfire dataset [in UPP1-CRZ.CDT].

With best wishes

Robin Shirley