4/4) Refinement - TWIN in SHELXL


Before inserting a twin law, we'll backtrack a bit to a fully isotropic model. You could manually edit the '.res' file, but it is easier to use the ISOT command in XP, and re-save as an '.ins' file. When the '.ins' file is run through SHELXL, the tail end of the screen output looks like this:


As you can see, it does not look very good. Hopefully, adding the twin law and occupancy factor (TWIN and BASF instructions) will help.


A few cycles of refinement with SHELXL reveal the following:


That's a whole lot better. The R-values and GooF are much improved, and the difference map is considerably flatter. Here's what it looks like in Mercury:


The model is dramatically better, but a nitrogen and carbon atom on the upper right ring seem odd. The N looks too big, and the C looks too small. This illustrates the effect of wrongly assigned atom types. In this case, the N has been smeared out by the least-squares in order to fit the experimental electron density of carbon. Likewise, the C was 'sharpened' to better fit the scattering power of nitrogen. Manual editing of the next '.ins' file is sufficient to switch their atom types. In SHELXL we only need to change the scattering factor number in the '.ins' file, but for aesthetics we'll also change the atom labels). While we're at it, we'll also make the model anisotropic (ANIS).


Yet more improvement. It is now time to re-check atom assignments prior to applying a sensible numbering scheme and adding hydrogen atoms to complete the structure.


We'll renumber the structure using XP, as usual ...


As with any structure, it helps to use a sensible numbering scheme, and to sort the atoms in a logical order (SORT in XP). We'll save this model, run a round of refinement and see if hydrogen atoms show in the difference map. This is the result:


For sure we can add hydrogen atoms to the carbons ...


After another round of refinement, nitrogen atoms N2, N4, N6, N8 also clearly have hydrogen attached. We'll add those to the '.res' file, refine, modify the weighting scheme (WGHT instruction), check to see that the moieties are placed well in the unit cell and test for extinction effects (EXTI).


A final round of refinement shows that in spite of the twinning, the structure is rather high quality. The '.res' file is available here.


Conclusion: The final refined structure model is good, with no unusual features.



1) Assign a space group using XPREP.
2) Direct methods structure solution.
3) Decipher the twin law.
4) Twin refinement with SHELXL.