crystal – and requires large amounts of material, which limits its practical application. This method also leads to unavoidable shot-to-shot variation in both the
diffraction quality and power, which must be accounted for by scaling and merging
during processing. For photocrystallographic studies in particular, given the strong
dependence on the size and morphology of the crystal, one would also expect
considerable variation in photoconversion level.
Fixed-target methods instead attach crystals to a target on a translation stage that
allows the X-ray beam to scan the surface during the experiment. Mehrabi et al.
developed a system specifically for TR studies in which 30 mm
2 patterned silicon
chips, each containing 20,736 sample wells, are mounted on a nano-translation stage
that enables fast raster scanning across the chip surface to expose different crystals to
the X-ray beam [81]. The chip is usually also scanned prior to the experiment to
identify which grid cells contain crystals, thereby removing an element of uncertainty and drastically improving the hit rate compared to injection methods. However, it is important to ensure during mounting that crystals are well separated to
avoid contamination between measurements. The better control of the crystal sampling afforded by fixed-target methods makes them potentially better suited to
chemical crystallography in general and photocrystallography in particular.
4.6 Data Processing
Another important consideration in a TR-SCXRD experiment is the data processing.
Whereas most conventional spectroscopic techniques generate data that is straightforward to visualise – if not to interpret – the raw data collected in a SCXRD
experiment requires extensive processing to analyse.
At a basic level, a traditional SCXRD dataset consists of a series of 2D diffraction
patterns captured as the crystal is rotated to bring different regions of its reciprocal
space onto the detector. The series of images is first subject to a peak-finding
procedure, by which all of the reflections are identified, and the reflections are
then indexed to obtain a unit cell, space group and an orientation matrix describing
the orientation of the crystal with respect to the diffractometer axes. Once this is
complete, the diffraction images are integrated to obtain the intensities of the
diffraction spots, resulting in a list of relative intensities with assigned Miller indices
(hkl). This information is then used to obtain an initial crystallographic model – a
unit cell and a set of atom positions – in a structure solution, and the model is
iteratively refined to obtain a best fit to the data.
There are a number of commercial and open-source software programs for
SCXRD data processing. The vast majority have been developed for laboratory
experiments and as such use graphical user interfaces (GUIs) that require considerable user interaction in their standard operation, although some programs offer
automatic processing routes that aim to automate part or all of the workflow. As
TR-SCXRD experiments typically produce vast amounts of data very rapidly by
design, it is desirable, even essential, to obtain an initial set of solutions and
232
L. E. Hatcher et al.
diffraction quality and power, which must be accounted for by scaling and merging
during processing. For photocrystallographic studies in particular, given the strong
dependence on the size and morphology of the crystal, one would also expect
considerable variation in photoconversion level.
Fixed-target methods instead attach crystals to a target on a translation stage that
allows the X-ray beam to scan the surface during the experiment. Mehrabi et al.
developed a system specifically for TR studies in which 30 mm
2 patterned silicon
chips, each containing 20,736 sample wells, are mounted on a nano-translation stage
that enables fast raster scanning across the chip surface to expose different crystals to
the X-ray beam [81]. The chip is usually also scanned prior to the experiment to
identify which grid cells contain crystals, thereby removing an element of uncertainty and drastically improving the hit rate compared to injection methods. However, it is important to ensure during mounting that crystals are well separated to
avoid contamination between measurements. The better control of the crystal sampling afforded by fixed-target methods makes them potentially better suited to
chemical crystallography in general and photocrystallography in particular.
4.6 Data Processing
Another important consideration in a TR-SCXRD experiment is the data processing.
Whereas most conventional spectroscopic techniques generate data that is straightforward to visualise – if not to interpret – the raw data collected in a SCXRD
experiment requires extensive processing to analyse.
At a basic level, a traditional SCXRD dataset consists of a series of 2D diffraction
patterns captured as the crystal is rotated to bring different regions of its reciprocal
space onto the detector. The series of images is first subject to a peak-finding
procedure, by which all of the reflections are identified, and the reflections are
then indexed to obtain a unit cell, space group and an orientation matrix describing
the orientation of the crystal with respect to the diffractometer axes. Once this is
complete, the diffraction images are integrated to obtain the intensities of the
diffraction spots, resulting in a list of relative intensities with assigned Miller indices
(hkl). This information is then used to obtain an initial crystallographic model – a
unit cell and a set of atom positions – in a structure solution, and the model is
iteratively refined to obtain a best fit to the data.
There are a number of commercial and open-source software programs for
SCXRD data processing. The vast majority have been developed for laboratory
experiments and as such use graphical user interfaces (GUIs) that require considerable user interaction in their standard operation, although some programs offer
automatic processing routes that aim to automate part or all of the workflow. As
TR-SCXRD experiments typically produce vast amounts of data very rapidly by
design, it is desirable, even essential, to obtain an initial set of solutions and
232
L. E. Hatcher et al.