Flack parameter reporting enquiry

[Mike Hoyland]

Dear All,

We have had the following correspondence regarding the reporting of Flack 
parameters which we thought would be good to be discussed in this forum. It 
was sent to us by David Watkin.

===========================================================================
We are updating the cif-generator in CRYSTALS and are faced with a 
problem when reporting the Flack parameter.

The current dictionary terms seem to be:

_refine_ls_abs_structure_Flack
_refine_ls_abs_structure_Rogers
_refine_ls_abs_structure_details


Following the publication of the Parsons, Flack & Wagner paper
(Simon Parsons et al.  Intensity quotients and absolute structure 
refinement Acta Cryst. (2013). B69, 249–259) there has been a drift, 
encouraged by recent releases of SHELXL, to determine the Flack 
parameter by a post refinement method.  Though this method does involve 
least squares, it is not part of the structure refinement.

We think that, if the cif is to reflect the actual analysis, some 
additional (optional) terms need defining - something along the lines of:

_flack_parameter_post-refinement_value      value, su

The value determined by any of the post-refinement method (e.g. Parsons 
Quotient, Sheldrick Hole-in-One, Bijvoet Difference, Hooft 'y')

_flack_parameter_post-refinement_details
;
reference etc to actual method used
;

_flack_parameter_in_model       value (no su)

The actual value used in computing structure factors, defaulting to 
zero.  This is needed because by default SHELXL does a Parsons Quotient 
calculation at the end but sticks with a value of zero unless the user 
opts to do a BASF/TWIN refinement (in which case the Flack parameter 
appears as the twin-fraction).

We believe that
_refine_ls_abs_structure_Flack
  and/or
_flack_parameter_post-refinement_value

should be in the cif because they provide the evidence for the reported 
absolute structure, even though the value used in the model may be 
different from the determined value. I would give both values because 
significant differences are indications that there is an issue with the 
observed data.

Example, where the user or program decides the sample is enantiopure:

_refine_ls_abs_structure_Flack         0.2(4)
_refine_ls_abs_structure_details
;
Refinement together with other structural parameters, Flack(1983)
;
_flack_parameter_post-refinement_value 0.0(7)
_flack_parameter_post-refinement_details
;0
Parsons Quotient, Parsons, Flack & Wagner, (2013)
;
_flack_parameter_in_model  0.0

Take care reading SHELX output - GMS now calls his old Hole-in-One the 
'Classical Method', while I would think Flack's 1983 paper really gives 
the original 'Classical' method.

If you have to Self Isolate in the coming weeks, this paper may help 
pass the time:
Cooper & Watkin, Acta Cryst. (2016). B72, 661-683
All is explained therein.
======================================================================================

-- 

With thanks and best wishes,
Mike Hoyland
Systems Developer
IUCr, Chester
==================================