maximising the electron dose per image for a good signal-to-noise ratio for
alignment, and to increase the acquisition time. Because specimens are on planar
grids and are thicker when tilted, it is possible to compensate for this by sampling
higher tilt angles at finer increments. The Saxton tilt scheme [61] where the tilt
increment is proportional to the cosine of the tilt angle provides even sampling in
Fourier space and optimally divides the electron dose into the minimum number of
images, although in practice is not generally performed, particularly as the most
information-rich images are those taken of the thinnest sample at lower tilt [55]. It is
important to note that the Crowther criterion does not apply to subtomogram
averaging projects, as data missing from one subtomogram will be filled in by
another, the only limitation being the requirement to accurately align particles,
which is dependent mainly upon high-contrast, low spatial frequency data.
The structural information in an image depends on tilt angle (a shallow tilt angle
provides a thinner sample, and thus less inelastic or multiple scattering) and
cumulative electron dose (the less pre-exposure the sample has seen, the less
damaged it will be)—so at which tilt angle should the user collect their first projection image—the image with the least electron damage? While more traditional
unidirectional strategies start at one extreme of the tilt range and collect the entire
tilt series in a single sweep tilt over to the other extreme, bidirectional schemes start
at 0°, seeking to expend the first electrons on the thinnest view of the sample by
collecting one half of the tilt series, before returning to 0° to collect the other side.
Both schemes, however, have problems. Unidirectional schemes use the first
electrons at high tilts when the specimen is thickest, losing high resolution features.
While the bidirectional scheme uses its first electrons when the undamaged specimen is thinnest, artefacts can be introduced into the reconstruction as a result of the
large structural differences due to damage from cumulative electron exposure
between corresponding negative and positive angles. The specimen will have
undergone structural changes due to electron damage and will not be the same in
the first set of low angle images as in the second set. Recently a dose symmetric
scheme has been developed with the goal of preserving the high-resolution data,
producing near-symmetric information in Fourier space and evenly distributing
electron damage throughout the tilt series [62]. The dose symmetric scheme starts at
zero and oscillates in tilt direction (for example, 0°, −3º, 3º, 6º, −6º, −9º, 9º…).
When using an unstable side-entry holder, an intermediate scheme, the asymmetric
scheme, can be used. Here, the tilt scheme starts at an intermediary angle (say 24º),
sweeps through to one end (−60º) and then returns to collect the remaining images
(+24º to +60º) [63]. This is a good compromise and expends the ‘best’ electrons
across the entirety of the relatively low-tilt range, with higher tilts (with thicker
specimen and therefore lower signal) being collected only after the specimen has
received much of the electron dose [52]. It is also possible to collect two tilt series
of the same specimen at different defoci with the intention of CTF correcting the
data, with the higher spatial frequency components being derived from the first tilt
series [64].
76
J. L. Ferreira et al.
alignment, and to increase the acquisition time. Because specimens are on planar
grids and are thicker when tilted, it is possible to compensate for this by sampling
higher tilt angles at finer increments. The Saxton tilt scheme [61] where the tilt
increment is proportional to the cosine of the tilt angle provides even sampling in
Fourier space and optimally divides the electron dose into the minimum number of
images, although in practice is not generally performed, particularly as the most
information-rich images are those taken of the thinnest sample at lower tilt [55]. It is
important to note that the Crowther criterion does not apply to subtomogram
averaging projects, as data missing from one subtomogram will be filled in by
another, the only limitation being the requirement to accurately align particles,
which is dependent mainly upon high-contrast, low spatial frequency data.
The structural information in an image depends on tilt angle (a shallow tilt angle
provides a thinner sample, and thus less inelastic or multiple scattering) and
cumulative electron dose (the less pre-exposure the sample has seen, the less
damaged it will be)—so at which tilt angle should the user collect their first projection image—the image with the least electron damage? While more traditional
unidirectional strategies start at one extreme of the tilt range and collect the entire
tilt series in a single sweep tilt over to the other extreme, bidirectional schemes start
at 0°, seeking to expend the first electrons on the thinnest view of the sample by
collecting one half of the tilt series, before returning to 0° to collect the other side.
Both schemes, however, have problems. Unidirectional schemes use the first
electrons at high tilts when the specimen is thickest, losing high resolution features.
While the bidirectional scheme uses its first electrons when the undamaged specimen is thinnest, artefacts can be introduced into the reconstruction as a result of the
large structural differences due to damage from cumulative electron exposure
between corresponding negative and positive angles. The specimen will have
undergone structural changes due to electron damage and will not be the same in
the first set of low angle images as in the second set. Recently a dose symmetric
scheme has been developed with the goal of preserving the high-resolution data,
producing near-symmetric information in Fourier space and evenly distributing
electron damage throughout the tilt series [62]. The dose symmetric scheme starts at
zero and oscillates in tilt direction (for example, 0°, −3º, 3º, 6º, −6º, −9º, 9º…).
When using an unstable side-entry holder, an intermediate scheme, the asymmetric
scheme, can be used. Here, the tilt scheme starts at an intermediary angle (say 24º),
sweeps through to one end (−60º) and then returns to collect the remaining images
(+24º to +60º) [63]. This is a good compromise and expends the ‘best’ electrons
across the entirety of the relatively low-tilt range, with higher tilts (with thicker
specimen and therefore lower signal) being collected only after the specimen has
received much of the electron dose [52]. It is also possible to collect two tilt series
of the same specimen at different defoci with the intention of CTF correcting the
data, with the higher spatial frequency components being derived from the first tilt
series [64].
76
J. L. Ferreira et al.