Evaluating a Crystal Structure
The IUCr has developed a format for communicating information about a
crystal structure using Crystallographic
Information File. These guidelines direct the user to ways of evaluating the
correctness and quality of a crystal structure report. The user must remember
that the precision of results is at best a fuzzy concept with generally no sharp
dividing lines. The International Union of Crystallography
does have a list of criteria for acceptance of crystal structure reports
to the journal Acta Crystallographica C. These
acceptance criteria are being used by nearly all journals
for identifying problem areas in small molecule crystal structures.
All small molecule crystal structures should be
validated using the checkcif program.
Tables of data as a pdf document may be prepared by submitting a cif via the
printcif web page.
Be sure to select the type of output; usually this should be
Several validation suites for macromolecular crystal structures have been developed. These include software from the Protein Data Bank server, the Joint Center for Structural Genomics, and the UCLA server.
The general criteria used to evaluate small-molecule crystal structures are as follows:
Geometry The bonding geometry must be chemically reasonable. The values of bond lengths, angles, and torsion angles must agree with similar values from related compounds. A variety of excellent database resources are available. The Cambridge Structural Database is especially useful in finding structures with particular molecular fragments. Also, similar chemical bonds within the given crystal structure must have similar bond lengths and angles. Non-bonded lengths should be chemically-reasonable. In this day when so many crystal structures have been determined, it is extremely rare to see a valid new crystal structure with
Precision The final unweighted R should be < 0.10. This value should not be used as a determining factor of the overall quality of the structure, but unfortunately it often is. A better gauge of the precision of the structure is found in the standard deviations (now called standard uncertainties) of the refined results. If the goodness of fit parameter S is significantly different from 1.0, then consider multipling the reported standard uncertainties by S to get a better idea of the true precision of the parameters.
In addition to the standard uncertainties, good quality structures have a large, > 10.0, data-to-parameter ratio; reasonably spherical anisotropic displacement parameters for all non-hydrogen atoms; and flat (no peaks > 0.25 e/Å3 or valleys < -0.25 e/Å3) electron density difference maps.
Cell From the cell parameters and the chemical formula the program confirms that the proper cell was selected as well as checking the cell volume, formula weight, Z, density, and linear absorption coefficient (also using the wavelength). The conventional cell parameter constraints should match the reported space group.
Resolution of Data The data should be collected out to an appropriate resolution limit (d < 0.67 Å-1). For Mo radiation this limit is θ ≤ 25.2 °. For Cu radiation this limit is θ ≤ 67 °. When possible, data should be collected beyond these upper limits. Although the IUCr says it will relax this standard in extraodinary circumstances, they seldom do. Most journals other than Acta Cryst. C will accept structures with data collected to lower resolution values if appropriate explanations, such as the data were weak beyond the reported cutoff, [R(int) > 0.25 for this or higher resolution shells of data or I/σ(I) < 2 for all higher resolution shells]. Whenever possible, all collected data should be used in refinement.
Absorption Correction An appropriate absorption correction must be applied. The IUCr uses the value of μ × the middle crystal dimension as a guide of the type of absorption correction needed for a particular sample and radiation. The Fourier method of absorption correction is never recommended. Note that the data should contain a sufficient number of redundant measurements to verify through the merging R that the absorption correction was sufficient and appropriate for the problem.
|μ × mid dim||Correction type|
|> 3.0||Analytical required|
|> 1.0||Analytical recommended|
|> 0.1||Some correction required|
|< 0.1||None required|
Refinement Either full-matrix or blocked full-matrix refinements should be used. If the data are weak, (F2 / σ) < 6.0, then the refinement should be based on F2. Strong data sets may be refined on either by minimizing differences between observe and calculated F2 or F. All atoms, including hydrogens should be included in the model. All non-hydrogen atoms should be refined with anisotropic displacement parameters, and all of these ellipsoids should appear to be nearly spherical. The refinement should be converged, the largest shift / σ should be < 0.05. Any disorder should be handled with the fewest number of parameters needed to fit the problem. The absolute structure of samples with non-centrosymmetric space groups should be properly determined.