Hi Mark,
Estimating twin fractions is best done on the basis of refining it with
a model. This can be done with phenix.refine or other programs like
shelx or CNS.
The ML estimate ofd the twin fraction relies on the correctness of the
sigmas. The 0.02 is actually the lower limit the ML procedure will
produce, I should fix that/be more clear when reporting the results I guess.
The fact that the twin fractions vary this much is not surprising I
think. I guess this depends on the type of twinning (contact vs
penetration), although I am not sure about this.
Peter
Mark Mayer wrote:
For cases where people have had merohedral twinning, did the twin fraction vary substantially
between individual crystals grown under indentical conditions? I have no prior experience with
merohedral twinning, and was surprised to see that the twin fraction varied substantially as detailed
below, and that by screening we were able to get untwinned xtals.
The project started with a weak home data set for which the twin fraction was 0.478, and which
scaled in both H3 and H32. We just came back from APS with data sets from another three crystals,
for which the ML twin fraction, estimated using phenix.xtriage with scalepack merged intensities as
input, varied from 0.335, 0.219 and 0.02. The latter is refining very nicely, in H3 and will not scale in
H32.
Thanks - Mark
