Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Draft changes to Core data names for multiplicity etc.

I am glad that we have managed to get the multiplicity sorted out in time for the new release of SHELX.  If a program needs the multiplicity it should, as George points out, be possible to calculate this from the site symmetry order; all that is needed is the multiplicity of the general position which can be found by counting the number of symmetry elements that are almost always included in the CIF.

I am not sure that I see the problem with the disordered solvent.  The occupancy that SHELX uses is correct, and the site symmetry order can be given for the actual site occupied.  An atom close to a special position could be described as lying on the special position or not depending on the occupancy and the site symmetry order or multiplicity given in the CIF.  Either choice would result in a consistent CIF.  How a program handles this depends on the way it treats the symmetry.  If it tests for special positions by seeing if any symmetry element generates the same initial coordinates within some rounding error there could be a problem if the author and reader of the CIF had different concepts of what a rounding error might be, but there is not much CIF can do about this.  Adopting George's suggestion of assigning groups of atoms to a particular special position is intriguing but would need to be considered very carefully since it involves an entirely new approach to the way CIF describes chemical structures.  It is not just a simple matter of extending existing definitions as George suggests.

There is no problem in adding
_reflns_Friedel_fraction_max  and  _reflns_Friedel_fraction_full to the dictionary.

The change in the way in which the standard uncertainties are calculated is not a concern for CIF.  CIF reports the standard uncertainties given by the author of the CIF and accepts no responsibility for their correctness, nor does it prescribe any particular method for calculating them.  This change in SHELX procedure may be of interest to those who use the CIF, but it seems that in practice this change is small and unlikely to cause many problems.

David Brown

On 5/17/2012 3:59 AM, George Sheldrick wrote:
Dear Brian et al.,

I am planning to release a new version of SHELXL this summer (in time for the ACA Meeting) so your email was timely. I do not like to release a new version more than about once a decade, but I try to debug that version properly before releasing it. To correct my misunderstanding of the 'multiplicity' I have changed '_atom_site_symmetry_multiplicity' to '_atom_site_site_symmetry_order' as you suggest, which will also be consistent with remediated CIFs. However I would like a guarantee that this will not generate an Alert A because  _atom_site_site_symmetry_multiplicity is missing (it really isn't necessary to give both).

However this does not solve a closely related problem, namely how to define these terms when a disordered atom or molecule is close to but not precisely on a special position. A very common example is a toluene solvent molecule trying to sit on an inversion center. In the SHELX world, the solution is to set the occupancy of all the toluene atoms to the value appropriate to the special position, e.g. 0.5. This has worked well without any problems in SHELX refinements for the last 40 years, the unit-cell contents and structure factors are calculated correctly. It is less clear what to do in the CIF file which defines occupancy differently and so also had to introduce the concept of multiplicity that is not used by SHELX.  One needs to take into account that one of the toluene atoms might accidentally be almost exactly on the inversion center, and so might be assigned a different multiplicity and occupancy than the other atoms in the toluene molecule. How close should an atom be to a special position to be assigned its multiplicity? To make it more complicated, a toluene molecule trying to sit on a position of mmm symmetry might have exact mm2 but not mmm symmetry. etc. The situation may be clear to a chemist but not easy to code in a completely robust and general way.

There is a possible (even elegant) solution but I will only implement it if you agree, but if not please suggest a better alternative! In a SHELX refinement, a (solvent) molecule trying to occupy a position that has higher symmetry than the molecule itself is invariably assigned a negative PART number to suppress the generation of meaningless bonds etc. It would be possible for me to use this information to assign the whole (e.g. toluene) molecule the CIF _order (or _multiplicity) appropriate to the special position, the (CIF) occupancy of all its atoms would then be 1. This would merely require a slight extension to the CIF definitions of multiplicity and order, namely that they can be applied either to atoms or to groups of atoms.

I have already implemented two new further CIF names that I did not preface with 'shelx' because I would like COMCIFS to consider them for general use. They are:

_reflns_Friedel_fraction_max  and  _reflns_Friedel_fraction_full

They are defined as the number of Fiedel pairs measured divided by the number theoretically possible (ignoring reflections in centric projections and systematic absences throughout); _max and _full have their usual meanings. In contrast to _reflns_Friedel_coverage they can take values in the full range 0 to 1 for any non-centrosymmetric space group, and so one can see at a glance how completely the Friedel pairs have been measured. For centrosymmetric space groups they would be 0/0 and so are given as '.'.

I have also implemented a change, suggested by Ton Spek, that has far-reaching consequences but also resolves some long-standing problems. It is often difficult to decide when to refine a non-centrosymmetric structure against data averaged according to the Laue group or the point group, especially in view of improvements in data quality. The problem is that the least-squares standard uncertainties are proportional to 1/sqrt(N-P), where N is the number of observations and P the number of parameters refined. The same problem arises in refinements against twinned data, where one unique reflection may contribute to several 'observations'. This even leads to carefully measured experimental data being thrown away, which is clearly not good science. Ton's solution was to always make N equal to the number of unique Laue averaged data (rather than the number of 'observations') when estimating the standard uncertainties. It is important that referees and editors are made aware of this change, which will mean that it will no longer matter whether the data used in the refinement are redundant or not. A possible side-effect is that in some cases the su's may increase a little, but they tended to be underestimated anyway.

In answer to your question, I have no objection to a su being allowed for the wavelength, provided that it is not obligatory (i.e. would generate a CheckCIF Alert A if absent).

Best wishes, George

On 05/16/2012 12:32 PM, Brian McMahon wrote:
Dear Colleagues

This message is being sent to members of the CIF Core Dictionary
Management Group, who are requested to respond to it at your
earliest convenience. A copy will also be sent to other software
developers for their early information.

Some programs have output values for the core data item
_atom_site_symmetry_multiplicity that are actually the order
of the space-group site symmetry and not the multiplicity as
defined in International Tables. As a result, the values given
in CIFs in the IUCr archive and elsewhere are ambiguous.

It is proposed to remove this ambiguity by deprecating the
existing data name _atom_site_symmetry_multiplicity and replacing it
with two new names,
formal definitions of which are given below.

This will allow future "remediation" of the IUCr CIF archive to
remove this ambiguity; and software authors may cleanly update their
output in the next release cycle of their programs. In practice,
we think that the usage of existing CIFs will not be greatly affected,
since most programs do not import and use the site multiplicities
directly; but by making this formal change we will also be drawing
attention to the need for end-users who do rely on this value to
validate it before use.

Please let us know whether you agree to, or have any objections to
this approach, as soon as possible.

For your information, it is proposed also to make two small editorial
changes to other data items (clarifying that F(000) may be given
in the case of neutron diffraction experiments, and sanctioning the
use of standard uncertainties when quoting the wavelength, as is
practice in powder diffraction). The details are also included below.

Best wishes
Brian McMahon
COMCIFS Coordinating Secretary


    2012-05-15 BMcM: Deprecated _atom_site_symmetry_multiplicity;
                     replaced with _atom_site_site_symmetry_multiplicity
                     and added _atom_site_site_symmetry_order (IDB)
                Reworded _exptl_crystal_F_000 definition to take
                         account of neutron diffraction and removed
                         _enumeration_range (can be negative for neutrons)
                Added '_type_conditions su' to _diffrn_radiation_wavelength

     _name                      '_atom_site_symmetry_multiplicity'
     _category                    atom_site
     _type                        numb
     _list                        yes
     _list_reference            '_atom_site_label'
     _related_item              '_atom_site_site_symmetry_multiplicity'
     _related_function            replace
     _enumeration_range           1:192
;              The multiplicity of a site due to the space-group symmetry as
               given in International Tables for Crystallography Vol. A (2002).

                Use of this data name is deprecated because of
                inconsistencies in practice among structure refinement
                software packages. The number of positions given for
                this Wyckoff site in International Tables for
                Crystallography Vol. A (2002). should now be expressed
                using the data name _atom_site_site_symmetry_multiplicity.
                In the historic archive some CIFs use this item to give values
                that belong in _atom_site_site_symmetry_order.

     _name                      '_atom_site_site_symmetry_multiplicity'
     _category                    atom_site
     _type                        numb
     _list                        yes
     _list_reference            '_atom_site_label'
     _related_item              '_atom_site_symmetry_multiplicity'
     _related_function            alternate
     _enumeration_range           1:192
;              The number of different sites that are generated by the
                application of the space-group symmetry to the
                coordinates given for this site. It is equal to the
                multiplicity given for this Wyckoff site
                in International Tables for Crystallography Vol. A (2002).
                It is equal to the multiplicity of the general position
                divided by the order of the site symmetry given in

     _name                      '_atom_site_site_symmetry_order'
     _category                    atom_site
     _type                        numb
     _list                        yes
     _list_reference            '_atom_site_label'
     _enumeration_range           1:48
;              The order of the site symmetry of the site identified by

                This is the number of times application of the
                crystallographic symmetry to the coordinates given for
                this site generates the same set of coordinates.
                It is equal to:

                           multiplicity of the general position
                               multiplicity of this site

                 where 'multiplicity of this site' is
                 given in _atom_site_site_symmetry_multiplicity.

     _name                      '_diffrn_radiation_wavelength'
     _category                    diffrn_radiation_wavelength
     _type                        numb
     _type_conditions             su
     _list                        both
     _list_reference            '_diffrn_radiation_wavelength_id'
     _enumeration_range           0.0:
     _units                       A
     _units_detail              'angstroms'
;              The radiation wavelength in angstroms.

     _name                      '_exptl_crystal_F_000'
     _category                    exptl_crystal
     _type                        numb
     _list                        both
     _list_reference            '_exptl_crystal_id'
;              The expression for a structure factor evaluated in the
                zeroth-order case h = k = l = 0, F(000). This may contain
                dispersion contributions and is calculated as

                F(000) = [ (sum f~r~)^2^ + (sum f~i~)^2^ ]^1/2^

                f~r~   = real part of the scattering factors at theta = 0
                f~i~   = imaginary part of the scattering factors at theta = 0

                         the sum is taken over each atom in the unit cell

                For X-rays, non-dispersive F(000) is a positive number
                and counts the effective number of electrons in the unit cell;
                for neutrons, non-dispersive F(000) (which may be negative)
                counts the total nuclear scattering power in the unit cell. See

Brian McMahon                                       tel: +44 1244 342878
Research and Development Officer                    fax: +44 1244 314888
International Union of Crystallography            e-mail:  bm@iucr.org
5 Abbey Square, Chester CH1 2HU, England
coreDMG mailing list

fn:I.David Brown
org:McMaster University;Brockhouse Institute for Materials Research
adr:;;King St. W;Hamilton;Ontario;L8S 4M1;Canada
title:Professor Emeritus
tel;work:+905 525 9140 x 24710
tel;fax:+905 521 2773

coreDMG mailing list

[Send comment to list secretary]
[Reply to list (subscribers only)]