Pseudo-tetragonal twinning

2/6: Analysis and file set up using XPREP

As stated in the introduction, initial indexing gave a triclinic unit cell, which was used for data collection so as to not impose any unwarranted symmetry constraints. On reading into XPREP, the program suggests a number of related I and C-centered cells:

Notice that the R (sym) for tetragonal-I (option A, above) is poor – almost twice that of any other suggestion. As noted by Hao et al. (2005), between crossed polarizers the crystals did not show optical extinctions characteristic of tetragonal symmetry. Thus, even prior to data collection, tetragonal seemed unlikely. Lower symmetry and twinned was suspected. The next highest symmetry is orthorhombic-I (option B), for which XPREP warns of a B-superlattice:

Nevertheless, it goes on to suggest a space group of type I bca (#73 in the International Tables, vol. A):

I bca was a dead end. All attempts to find a chemically reasonable solution using this symmetry failed (feel free to try if you like). The next step is a further lowering of the symmetry to monoclinic-I. Since all the cell angles are about 90°, XPREP suggests three alternatives, all having similar R (sym). The transformation matrices that map the original triclinic-P cell to these monoclinic-I settings (options D, F, H) are:

Note that option F gives the same transformation matrix as option B (orthorhombic-I ). It also has the lowest R(sym), though not by much. The worst case scenario is that we’ll need to try each of these three settings. The right choice should be obvious because it will be the only one that gives a recognizable solution. Since option F has the lowest R(sym), we’ll try it first:

XPREP tries to force orthorhombic again, so be sure to override it to monoclinic. It should then suggest I2/a. Go ahead and generate files to solve the structure, but be sure to specify SHELXD, not SHELXS.

XPREP suggests writing a new file with data transformed for the chosen setting. You could do that if you like, and it would work. However, XPREP scales down the F² and σF², with potential loss of precision. It's better to transform the data yourself. The unix-utility awk (which is also available for Windows via the Cygwin project ) does the trick:

If you compare such a file with one written by XPREP, what do you notice ? Ten points to Gryffindor if you spot the difference.

In section 3 we'll 'solve' the structure in I 2/a using SHELXD .