tips for finalising refinements

The importance of even routine crystal structure analyses is highlighted by Hui et al. [Acta Cryst. C70, 250–255] who say "structural `anomalies', even when first seen in flawed X-ray diffraction data sets obtained under non-ideal (but real world) conditions, carry important messages to anyone willing to listen. Outliers are sometimes truly `noise', but in select cases can be the invitation to discovery."

The following suggestions are intended to help authors produce refinement results and a CIF in which the atoms of the structure, the atom naming and the lists of atomic coordinate and geometric parameters are organised in a logical way. While these matters are mostly cosmetic and do not change the scientific results, they are `good housekeeping' and allow any user of the resultant CIF and publication to easily find an entry and follow the structure description. These concepts require little effort to implement if most of them are attended to during the model building and refinement or at the latest prior to the final refinement. It is much easier to implement this in the instructions input to the refinement program (e.g. in the SHELXL .ins file) than by trying to edit the final CIF. If it is later discovered that another refinement is needed, there will be no need to have to edit the CIF all over again.

atom labels

The symmetry-unique atoms in the refinement model should be identified by unique labels composed of a number appended to the IUPAC chemical symbol without the use of parentheses (e.g. Zn1, C7 etc.). Chemical and crystallographic numbering should be in agreement wherever possible. Atom labels should be as simple and concise as possible and not contain superfluous characters, e.g. C2 is better than C02. In principle, there is no need to specify the O atom of a water molecule as OW or O5W; just any numbered O label, like O5, will do. Where there is just one atom of a given element, include the digit, i.e. it is preferable to use Zn1 rather than just Zn.

[ellipsoid plot]

H-atom numbers should relate to the atom to which they are bonded, unless this leads to naming ambiguities; chemically ambiguous or complex labels such as HO1 (could mean holmium), H1N4, H1W2 etc., should be avoided.

If a structure has more than one chemically identical entity in the asymmetric unit, it is sensible to label both using similar labels and in the same sequence. This might be C1A, C2A... and C1B, C2B... for two molecules, or to start the labels for the second molecule at the next decade, e.g. if one molecule has C1 through to C33, the second molecule could start at C41.

Two isostructural compounds, even if one is in another report, should be labelled in the same way and the atomic coordinates of both structures defined using the same asymmetric units. This simplifies comparisons of geometry, coordination and packing parameters. The same applies for labelling of atoms in polymorphs.

atomic coordinates

Atomic coordinates for molecular and the unique part of extended species should be supplied as connected sets, i.e. the entire species is shown as one connected entity if the CIF is displayed in PLATON with the NOMOVE option turned on (note that the NOMOVE default is on for a CIF, but off for a .res or .ins file). If this test displays the molecule as two or more disconnected fragments, one or more of these fragments needs to be moved to another symmetry-related position so that all fragments are connected. This can easily be achieved by reading the CIF into PLATON, turning NOMOVE off and then clicking the option to generate a new SHELXL .res file. PLATON automatically moves the atoms into a connected set with their centre of gravity within the principle unit cell.

Whenever structure geometry permits, it is good practice to choose the set of connected coordinates which specify the asymmetric unit to have their centre of gravity within the primary unit cell (see the above tip on using PLATON).

In systems with hydrogen-bonded networks, it is helpful to choose the asymmetric unit so that the minimum number of symmetry operators is required to specify the hydrogen-bond network. Among other things, these concepts help simplify the labelling of packing diagrams.

sorting of atoms

Before the final refinement, the list of atoms should be sorted into a logical sequence. As a consequence, the geometry lists in the CIF and any tables generated therefrom will then also be in a logical order. A reasonable sort order might be: (i) group atoms belonging to discrete components together (e.g. multiple molecules in the asymmetric unit, ions or solvent molecules), (ii) sort those groups of atoms into decreasing atomic weight and (iii) sort atoms of the same element into numerically increasing order. The exception to this is that H atoms in calculated positions should immediately follow their parent atom, rather than being sorted to the end of the list.

[ellipsoid plot]

H-atom treatment

Authors should note the advice on H-atom treatment given in the SHELXL97 manual, section 4.6: `For most purposes it is preferable to calculate the hydrogen positions according to well-established geometrical criteria and then adopt a refinement procedure which ensures that a sensible geometry is retained'. While this is usually good practice, care must be taken not to assume that a particular site has a certain geometry when defining the H atoms. Allowing methyl groups to rotate is often better than assuming a staggered conformation (i.e. in SHELXL, the AFIX 137 instruction may be preferable to AFIX 33). Except for amides, NH and NH2 groups are not always planar (sp2 N), even when a pi-system is nearby; e.g. if there are two NH substituents on a phenyl ring, the ring cannot delocalise with the lone pair on both N atoms, so at least one of the groups will be pyramidal (sp3 N).

With good quality data, especially low-temperature data, it should be possible to locate and even freely refine H atoms on any heteroatom, at least in a test refinement, thus alleviating any risk of ambiguity about the H-atom positions on groups such as amines, hydroxy groups and water molecules. The free refinement of Uiso values for such H atoms can act as a further confirmation of the correct positioning. Plotting contoured difference maps, e.g. using the `ContourDif' option of PLATON and omitting the relevant H atoms (`OmitFromSFC' option), can also aid in confirming the correctness of the H-atom positions.

absolute structure

It is no longer necessary to merge Friedel opposites for noncentrosymmetric structures in cases where the anomalous scattering power is too low to allow a definitive value of the Flack parameter to be obtained. With advances ongoing in this area, it is better to archive the Friedel-unmerged reflection file, so that future users of the data might be able to extract useful information from the Friedel pairs as techniques develop. Reporting the Hooft parameter (calculated by PLATON) or the Parson's parameter may also be useful (the latter is calculated in SHELXL2012 or later versions and reported in the CIF if more precise than the Flack parameter).

extinction parameter

If the refined extinction parameter is less than three times its s.u. from zero, the value is meaningless and the parameter should not be included in the refinement.

Special and virtual issues

Acta Crystallographica Section C has recently published special issues on

NMR Crystallography (March 2017)

Scorpionates: a golden anniversary (November 2016)

Full details are available on the special issues page.

The latest virtual issue features Coordination polymers, with an introduction by Len Barbour.

What are the 'most read' articles from the recent special issues?

Follow Acta Cryst. C
Sign up for e-alerts
Follow Acta Cryst. on Twitter
Follow us on facebook
Sign up for RSS feeds