complications for sample preparation. Once the path a probing
electron takes through the sample becomes too long, multiple
scattering events or absorption will either contaminate the diffracted intensities or prevent their observation altogether
[13]. MicroED data are now routinely collected under continuous
rotation of the sample, which mitigates dynamical artifacts due to
multiple scattering [14, 15]. As crystal milling is becoming more
commonplace, it is increasingly possible to carefully prepare the
sample to its desired thickness [16, 17]. Even so, owing to their
unique chemical composition and unit cell size, the exposure will
have to be tuned to each sample depending on its susceptibility to
radiation damage.
3 Data Collection
Data acquisition is the final experimental stage in crystal structure
determination; all subsequent stages involve calculations that are
ultimately limited by the quality of the data. Computational steps
may be modified and repeated as necessary, whereas data collection
is constrained by sample availability and access to the instrument.
Since no data processing program can rescue an irredeemably bad
dataset, it is necessary to plan data collection such that it yields the
most suitable data for the purpose at hand.
To recover the information in a diffraction pattern, the spots
must first be indexed, which entails finding their coordinates in the
corresponding three-dimensional reciprocal lattice. Once the identity of each spot is known, their intensities are calculated, either by
summing the pixel values in a region around their predicted central
location, or by fitting a suitable spot profile, often derived from
strong, nearby spots. This reduces a stack of hundreds, or even
thousands, of diffraction images to a single indexed table of intensities and associated error estimates. Low-dose data collection
inherently implies that data processing software must deal with
weak reflections, barely visible over the background.
MicroED often requires data collected from multiple crystals.
This is in part because crystal rotation is limited to a single, fixed
axis, and its accessible rotation range may be further constrained by
obstructing grid bars or other crystals. When processing multicrystal data, each individual dataset must be strong enough to be
indexed on its own; otherwise, it will not be possible to register the
datasets with respect to each other. If the unit cell is known before
data collection, it has recently become possible to index each frame
individually [18]. For a sample with unknown unit cell, indexing
usually requires several good diffraction patterns, all at different
orientations [19, 20]. It follows that the exposure must generally
be adjusted such that each crystal gives sufficiently many reflections
distributed over a wide enough rotation range for a dataset to be
312
Johan Hattne
electron takes through the sample becomes too long, multiple
scattering events or absorption will either contaminate the diffracted intensities or prevent their observation altogether
[13]. MicroED data are now routinely collected under continuous
rotation of the sample, which mitigates dynamical artifacts due to
multiple scattering [14, 15]. As crystal milling is becoming more
commonplace, it is increasingly possible to carefully prepare the
sample to its desired thickness [16, 17]. Even so, owing to their
unique chemical composition and unit cell size, the exposure will
have to be tuned to each sample depending on its susceptibility to
radiation damage.
3 Data Collection
Data acquisition is the final experimental stage in crystal structure
determination; all subsequent stages involve calculations that are
ultimately limited by the quality of the data. Computational steps
may be modified and repeated as necessary, whereas data collection
is constrained by sample availability and access to the instrument.
Since no data processing program can rescue an irredeemably bad
dataset, it is necessary to plan data collection such that it yields the
most suitable data for the purpose at hand.
To recover the information in a diffraction pattern, the spots
must first be indexed, which entails finding their coordinates in the
corresponding three-dimensional reciprocal lattice. Once the identity of each spot is known, their intensities are calculated, either by
summing the pixel values in a region around their predicted central
location, or by fitting a suitable spot profile, often derived from
strong, nearby spots. This reduces a stack of hundreds, or even
thousands, of diffraction images to a single indexed table of intensities and associated error estimates. Low-dose data collection
inherently implies that data processing software must deal with
weak reflections, barely visible over the background.
MicroED often requires data collected from multiple crystals.
This is in part because crystal rotation is limited to a single, fixed
axis, and its accessible rotation range may be further constrained by
obstructing grid bars or other crystals. When processing multicrystal data, each individual dataset must be strong enough to be
indexed on its own; otherwise, it will not be possible to register the
datasets with respect to each other. If the unit cell is known before
data collection, it has recently become possible to index each frame
individually [18]. For a sample with unknown unit cell, indexing
usually requires several good diffraction patterns, all at different
orientations [19, 20]. It follows that the exposure must generally
be adjusted such that each crystal gives sufficiently many reflections
distributed over a wide enough rotation range for a dataset to be
312
Johan Hattne