There is a significant problem that affects the autoindexing in the
latest release of iMosflm (Version 1.0.6, March 2012).
The indexing solutions themselves are not affected, but the suggested
solution (highlighted in blue) will, in some cases, be a solution with
a higher symmetry than the true solution. This will be indicated by
the fact that the penalty for the suggested solution will be
significantly higher than for lower symmetry solutions.
An example is shown below, where iMosflm is suggesting the
orthorhombic face centred solution (number 6, oF) which has a penalty
of 20, while the correct solution is in fact the C face centred
monoclinic solution (number 2). This happens because of a bug in the
code and because the rms error on spot positions (sigma(x,y)) for the
oF solution is not much higher than that for the triclinic solution.
Therefore the solution suggested by the latest iMosflm release needs
to be examined critically. In the example below, the jump in penalty
in going from solution 2 (penalty 1) to solution 3 (penalty 15)
strongly suggests that the correct solution is number 2.
Indexing in background jobs with ipmosflm is not affected by this
error.
Apologies to all those that have downloaded this new version. A bug
release, that corrects this problem, will be available soon.
Thanks to Frank von Delft for bringing this to our attention, mea culpa.
Andrew Leslie
|