good enough condition to be put back on the instrument? This leads to the conclusion that it would be better to collect data on several crystals and select the best
result. The main factor against this approach is time: collection of data on multiple
crystals takes time; processing and working up the data takes time. But if a dataset
only takes a couple of minutes to acquire, why not collect on multiple crystals?
Again, the factor against this is the time taken processing and evaluating all the
datasets to find the best one, but if there was an algorithm that could process the data
and then flag the best one to work on, then the problem is solved. If multiple datasets
are collected, there is also the possibility of merging and scaling them into a single
dataset with better averaging statistics and the ability for one collection to cover the
shortfall of another. Data collection on multiple crystals also facilitates assessment
for radiation damage. Some initial crystals could be sacrificed to determine the
experimental conditions that would enable good data to be successfully collected.
Approaches to mitigating radiation damage are likely to be specific and dependent
on the mechanism of decay, but in some circumstances, methods of reducing the
dose could include a better balance of decrease in beam intensity versus longer
counting times, counting for the same time over multiple sweeps of the same strategy
and summing the frames together, merging datasets from multiple crystals, reducing
the beam size to be much smaller than the crystal and translating the crystal during
the data collection so the same area is not continuously illuminated.
An increased speed of data collection also means multiple spheres of data may be
collected. This enables more optimal performance of corrections programs such as
SADABS and for the data to be finer sliced, e.g. 0.1
frame widths rather than 0.3
.
In-house experiments (ALS) indicate that for a combination of SAINT [225] and
SADABS on a compound with a small unit cell (all axes <13 Ǻ), the data correction
performs much better when narrower slices are used. There is also an improved
signal-to-noise ratio when using narrower frame widths. If one considers the case of
a 0.2
wide diffraction peak, for a 1
frame width and 1 s exposure, the background
accumulates for 1 s while for the peak itself only 0.2 s; if collected with a 0.2
frame
width for 0.2 s, the background accumulates for the same time as the peak.
With the move away from commercial instruments to custom ones, there is an
opportunity to better and fully integrate them with the beamline. This allows for the
operation of feedback loops such as the following: if the detector saturates, attenuation can be added or counting time decreased; if the dynamic range of the detector is
being optimally and maximally under attenuation, this can be removed or the
counting time increased. Additional data could also be collected, such as a fluorescence spectrum to provide qualitative element analysis for metals. This would also
allow automated resonant scattering experiments to be run, making them as easy as a
routine data collection. All the metadata for the collection can also be collated with
beamline data such as wavelength, beam positions and intensity after each of the
optical elements on the beamline – making troubleshooting more straightforward.
In the future the ideal detector would be an energy-resolving large area pixel
detector. For monochromatic data collection, this would allow fluorescence background to be discriminated from diffraction and removed, improving the signal-tonoise ratio which is crucial for weakly diffracting crystals. More importantly it
126
S. J. Coles et al.
result. The main factor against this approach is time: collection of data on multiple
crystals takes time; processing and working up the data takes time. But if a dataset
only takes a couple of minutes to acquire, why not collect on multiple crystals?
Again, the factor against this is the time taken processing and evaluating all the
datasets to find the best one, but if there was an algorithm that could process the data
and then flag the best one to work on, then the problem is solved. If multiple datasets
are collected, there is also the possibility of merging and scaling them into a single
dataset with better averaging statistics and the ability for one collection to cover the
shortfall of another. Data collection on multiple crystals also facilitates assessment
for radiation damage. Some initial crystals could be sacrificed to determine the
experimental conditions that would enable good data to be successfully collected.
Approaches to mitigating radiation damage are likely to be specific and dependent
on the mechanism of decay, but in some circumstances, methods of reducing the
dose could include a better balance of decrease in beam intensity versus longer
counting times, counting for the same time over multiple sweeps of the same strategy
and summing the frames together, merging datasets from multiple crystals, reducing
the beam size to be much smaller than the crystal and translating the crystal during
the data collection so the same area is not continuously illuminated.
An increased speed of data collection also means multiple spheres of data may be
collected. This enables more optimal performance of corrections programs such as
SADABS and for the data to be finer sliced, e.g. 0.1
frame widths rather than 0.3
.
In-house experiments (ALS) indicate that for a combination of SAINT [225] and
SADABS on a compound with a small unit cell (all axes <13 Ǻ), the data correction
performs much better when narrower slices are used. There is also an improved
signal-to-noise ratio when using narrower frame widths. If one considers the case of
a 0.2
wide diffraction peak, for a 1
frame width and 1 s exposure, the background
accumulates for 1 s while for the peak itself only 0.2 s; if collected with a 0.2
frame
width for 0.2 s, the background accumulates for the same time as the peak.
With the move away from commercial instruments to custom ones, there is an
opportunity to better and fully integrate them with the beamline. This allows for the
operation of feedback loops such as the following: if the detector saturates, attenuation can be added or counting time decreased; if the dynamic range of the detector is
being optimally and maximally under attenuation, this can be removed or the
counting time increased. Additional data could also be collected, such as a fluorescence spectrum to provide qualitative element analysis for metals. This would also
allow automated resonant scattering experiments to be run, making them as easy as a
routine data collection. All the metadata for the collection can also be collated with
beamline data such as wavelength, beam positions and intensity after each of the
optical elements on the beamline – making troubleshooting more straightforward.
In the future the ideal detector would be an energy-resolving large area pixel
detector. For monochromatic data collection, this would allow fluorescence background to be discriminated from diffraction and removed, improving the signal-tonoise ratio which is crucial for weakly diffracting crystals. More importantly it
126
S. J. Coles et al.