A three-fold twin in B2₁

7/7: Solve using B2₁

Although B2₁ is unconventional, a SHELXD ins file to solve the structure can easily be made from the file used to solve in P2₁. All it takes is to change the number on the LATT command from '-1' to '-6', which specifies a non-centrosymmetric B-centered cell. Nevertheless, the solution is likely to run faster and be more complete using the detwinned dataset. Subsequent refinement should, of course, use the full dataset, and include three-fold twinning.

SHELXD easily solves the structure ...

... and finds all of the non-hydrogen atoms. The only drawback is SHELXD assigned atoms exclusively as carbon, which is apparent if we view it in Mercury:

That is easily fixed, e.g., using ShelXle:

To refine as a three-component twin, the TWIN matrix used for the original P2₁ model will work. Just remember to specify three components, a BASF with two fractions, and include 'LATT -6' to ensure a B-centered cell. The model refines smoothly, with no need for constraints or restraints, even when fully anisotropic. Here's a look at the complete asymmetric unit:

It still needs a sensible numbering scheme and since Z' = 6, it's best to number them all the same way and make each a separate 'RESI'. That's easy enough to do, but not necessary for this tutorial. For all intents and purposes, the problem is solved.

An entertaining postscript for 9mj is that B2₁ caused the validation routines in Platon and checkCIF to throw a massive wobbler. Those issues were quickly patched (thanks to Ton Spek). CheckCIF still returns an 'A-level' squawk about space group type B2₁, but it is trivial to justify.