3/7: 'Solve' using P21
The SHELXD ins file written by XPREP needs a bit of manual editing. Firstly,
the wavelength is wrong (this is a CuKα dataset) and the ZERR line has incorrect
values. Although SHELXD can solve this structure without being told of the twinning,
the success rate is higher if a TWIN matrix is given. Also, the success rate is dramatically
improved when WEED is specified. With these changes we can safely reduce NTRY to 20. Lastly, a
SEED instruction is added so that repeat runs on the same hardware give reproducible results.
After editing, the ins file looks like this:
This
SHELXD instructions file is downloadable using the following link:
9mj-shelxd.ins
The filename needs to be changed to
9mj.ins to work with the
9mj.hkl dataset. On
running
SHELXD, the following summary is written to the screen:
From the looks of it, the structure was solved well before the 20th try. In Mercury,
it looks like this:
Aside from a few disconnected parts, this is not so bad. SHELXD has assigned a few
of the atoms as oxygen (red) and the rest as carbon (grey). It is straightforward, if a bit
tedious, to manually build this into twelve discrete molecules. To proceed, we do need a
more complete model, but it does not have to be fully refined. That's just as well, because
anisotropic refinement of this P21 model is not stable, even with an
extensive battery of constraints and restraints. Indeed, it was refinement instability that
prompted a closer look. Luckily, for its intended purpose here, an isotropic model without
hydrogen atoms will suffice. Similarly, it won't matter if some of the atom types are wrong;
it just needs to be good enough. A few cycles of model building/refinement gave the
following, in which all non-hydrogen atoms have been left as carbon:
Since this tutorial is not intended to be an exercise in laborious model building, the above
model is available for download, here:
9mj-good-enough.res
In part 4 we'll use this model to create a 'detwinned' dataset for diagnostic purposes.