(1) the full angular range would be sampled. There would be no missing wedge or
missing pyramid, leading to (a) simplified interpretation and (b) easier segmentation due to the fact that information is substantially complete and isotropic (Fig. 1.9);
(2) unlike the case for slabs, there would be a negligible focus gradient across the
specimen during ‘rotation’. An imposed defocus of, say, −5 µm would be valid
for all areas, which would simplify CTF correction and would also allow the
use of large-format detectors. Where phase plates are used to provide in-focus
contrast, this benefit would be realised fully;
(3) the thickness would remain constant during rotation, meaning that every projection would have identical SNR characteristics. Each projection would contribute equally to the reconstructed volume. As a corollary, radiation damage
would be uniform, simplifying the dose fractionation task. For example, some
acquisition schemes for plastic sections attempt to compensate for the increased
beam cross-section at higher tilt by calculating the exposure time necessary to
keep the brightness constant for all projections, and increase the exposure time
accordingly. This leads to highly irregular distortions;
(4) constant thickness means that fiducial markers would be equally apparent in
every projection. An inability to track markers precisely or at all sometimes
leads to projections of highly-tilted specimens being discarded;
(5) larger volumes could be reconstructed: if desired, the cylindrical specimen
could be slightly thicker than an untilted slab because all projections will have
the same quality. Proportionally more dose could be expended on each projection. Therefore, for the same nominal resolution, the cylinder allows for a
larger reconstructed volume.
In principle, cylinders of vitreous material can be manufactured by lathe using a
modified cryo- ultramicrotome setup, or by focussed ion beam. Rotation sample
holders for room-temperature tomography have been available for some time but an
equivalent cryo- sample holder has not. The use of a vitreous cylinder was proposed in
2009 [4] and adapted for proof of principle studies of ribosomes, vesicles and bacteria
[123]. Specimens that were plunge-frozen in carbon nanotubes were imaged with
moderate success using a standard (tilting) goniometer, which allowed tilting to ±79°
but as noted above, with equivalent SNR in all projections. In the absence of a
dedicated cryo- rotation holder, the main limit to quality seemed to be the thickness of
between 400 and 800 nm, and not the carbon nanotube wall thickness of 10–20 nm,
which corresponded to the thickness of amorphous carbon films used for plunge
freezing. Nanotubes with an internal diameter of <1 µm impose restrictions on the size
of specimen that can be introduced, which must be in the form of a suspension.
Although progress has been made towards nanofabrication of suitable sample
geometries, targeted milling (preserving the target whilst sputtering less relevant
features above and/or below the target) has only been addressed in part. Methods
for localising features prior to electron tomography have been demonstrated [15,
124, 125]. These methods establish the xy coordinates of features of interest but
they seldom take into account the fact that thinning (as in the case of FIB) can
24
A. Leis
missing pyramid, leading to (a) simplified interpretation and (b) easier segmentation due to the fact that information is substantially complete and isotropic (Fig. 1.9);
(2) unlike the case for slabs, there would be a negligible focus gradient across the
specimen during ‘rotation’. An imposed defocus of, say, −5 µm would be valid
for all areas, which would simplify CTF correction and would also allow the
use of large-format detectors. Where phase plates are used to provide in-focus
contrast, this benefit would be realised fully;
(3) the thickness would remain constant during rotation, meaning that every projection would have identical SNR characteristics. Each projection would contribute equally to the reconstructed volume. As a corollary, radiation damage
would be uniform, simplifying the dose fractionation task. For example, some
acquisition schemes for plastic sections attempt to compensate for the increased
beam cross-section at higher tilt by calculating the exposure time necessary to
keep the brightness constant for all projections, and increase the exposure time
accordingly. This leads to highly irregular distortions;
(4) constant thickness means that fiducial markers would be equally apparent in
every projection. An inability to track markers precisely or at all sometimes
leads to projections of highly-tilted specimens being discarded;
(5) larger volumes could be reconstructed: if desired, the cylindrical specimen
could be slightly thicker than an untilted slab because all projections will have
the same quality. Proportionally more dose could be expended on each projection. Therefore, for the same nominal resolution, the cylinder allows for a
larger reconstructed volume.
In principle, cylinders of vitreous material can be manufactured by lathe using a
modified cryo- ultramicrotome setup, or by focussed ion beam. Rotation sample
holders for room-temperature tomography have been available for some time but an
equivalent cryo- sample holder has not. The use of a vitreous cylinder was proposed in
2009 [4] and adapted for proof of principle studies of ribosomes, vesicles and bacteria
[123]. Specimens that were plunge-frozen in carbon nanotubes were imaged with
moderate success using a standard (tilting) goniometer, which allowed tilting to ±79°
but as noted above, with equivalent SNR in all projections. In the absence of a
dedicated cryo- rotation holder, the main limit to quality seemed to be the thickness of
between 400 and 800 nm, and not the carbon nanotube wall thickness of 10–20 nm,
which corresponded to the thickness of amorphous carbon films used for plunge
freezing. Nanotubes with an internal diameter of <1 µm impose restrictions on the size
of specimen that can be introduced, which must be in the form of a suspension.
Although progress has been made towards nanofabrication of suitable sample
geometries, targeted milling (preserving the target whilst sputtering less relevant
features above and/or below the target) has only been addressed in part. Methods
for localising features prior to electron tomography have been demonstrated [15,
124, 125]. These methods establish the xy coordinates of features of interest but
they seldom take into account the fact that thinning (as in the case of FIB) can
24
A. Leis