Checking the Structure
During a refinement, you need to make a number of checks after each refinement cycle.
- The Cambridge Structural Database is an excellent resource to use to check that your bond lengths and angles are sensible and within expected ranges. If they are not, investigate why not. -Check that all of the ellipsoids are of a similar size throughout the structure and look sensible (the ellipsoid axes should not appear significantly different).
- Check that the correct number of hydrogen atoms are present and that they are in the right place.
-Go to the
Info
tab and open theRefinement-Indicators
header tab. If there are any errors, some or all of the parameters will be orange (potentially moderate problem) or red (potentially serious problem). These should improve as the refinement proceeds. Attempts should always be made to correct these or be able to explain why they are not ideal. - Check that the bond lengths and angles are sensible. There are several ways to do this:
- Hover over a bond to see the bond length. Left click on the bond to select it.
- Select 2 atoms for a bond length, 3 atoms for a bonded angle or 4 atoms to get a torsion angle. Then go to the
View
tab, under theGeometry
header tab click on Distance and Angles (of selection). - Select all bond that you are interested in, then type
>>labels
. This will display the bond distances as labels on the bonds. You can also type>>sel -l
to print a list of the distances or>>sel -l -c
to then place this list on the clipboard, which you can then paste elsewhere.
The
>>sel
command can be always used in the command-line to print information regarding the current selection.
What To Look For In a Good Refinement
A summary of the refinement indicators is provided under Info|Refinement-Indicators
. These can be used to check the refinement of your structure. Red colour indicates there is a problem with the refinement; green colour indicates a good refinement.
All of the ADPs are chemically sensible
The thermal ellipsoids of the atoms should all look similar and of approximately the same dimensions. If there are some really large or really small ones, then there is something wrong. If there are some that are very elongated or very “squashed”, then that is also not right.
No residual least squares (L.S.) shift
Check that the Max Shift is very close to 0 in the top panel, if it is not, the refinement has not converged, and further cycles of refinement should be carried out. If convergence cannot be achieved using further cycles for refinement, then the reasons for this need to be investigated, for example, overuse of restraints. Often you find that the largest shifts are associated with hydrogen atoms that have not been appropriately constrained. Click on the Max Shift label that tells you about the largest shift and the “culprit” will be selected.
Good $R$ factors: $R_{1}$ and $wR_{2}$
$R_1$ and $wR_2$ should be as small as possible. $wR_{2}$ is always larger than $R_{1}$ due to the way it is calculated. The expected values depend both on the type of structure that is being solved and the quality of the data e.g. for a molecular organic compound an R1 ~ 5–7% and wR2 ~ 10–15% is reasonable. If heavy elements are involved, the final R-factors will be considerably lower. This is because a larger proportion of the electrons in the molecule are in the metal, and we can typically pinpoint the position of these very accurately. Associated with wR2 (and also the weighting scheme) is the Goodness of Fit (GooF) parameter, which should be close to 1. If it deviates a lot (say, larger than 1.2 or smaller than 0.8), there is a problem.
The Highest Peak and Deepest Hole
The Highest Electron Density Peak and Deepest Electron Density Hole should be small and approximately equal for a molecular organic compound ~0.5 e/A. If the structure contains a heavy atom e.g. a transition metal, Cl or Br, slightly larger peaks might be expected particularly in the vicinity of the heavy element. You can click on the Q-peak and type >>envi
, or right-click on it and then choose BANG (which stands for Bonds ANGgles) to calculate the distance of any suspect residual peak to the heavy element.
Data/Parameter Ratio
This one is a tricky one because there is not much you can do to fix it if this should become a problem. The ratio of the number of reflections (observed parameters) to refinement variables (model parameters) should be no less than 8 to 1. If enough data have not been recorded, e.g. a very weakly diffracting crystal, very small sample or high-pressure data – it may be necessary to reduce the number of refinement parameters using constraints.