2.3.2 STEM Tomography of Thick Plastic-Embedded
Sections
STEM tomography was first developed for materials science applications, mainly
using the HAADF detector, because it was appreciated that DF contrast provides
“true-projections” of the sample (Sects. 2.2.1, 2.2.2) [59]. HAADF STEM tomography was used also for biological applications to localize ultra-small immuno-gold
labels adsorbed to plastic sections [54, 55] and to localize membranes [60].
A major advantage of performing tomography in STEM mode is the ability to
raster the beam in lines parallel to the tilt axis of the tilted sample, while focusing
the beam according to the height of each line. Thus, by using dynamic focusing
mode [61, 62], focus gradients across highly tilted samples are avoided and all
regions of the resulting image remain in-focus. Dynamic focusing is implemented
in currently available STEM tomography software packages such as SerialEM [63],
Xplore3D [64] (FEI Company, The Netherlands), and others [65, 66]. This is in
contrast to CTEM tomography, in which only a narrow region of the tilted sample
close to the tilt axis is in focus, while the rest of the sample is blurred. This effect is
aggravated as thicker samples are observed due to changes of the contrast transfer
function.
The work of Aoyama et al. [32] demonstrated the unequivocal advantage of
STEM over zero-loss energy-filtered TEM tomography of thick (1 lm) sections
observed with 300 kV STEM. The quality of TEM images was low even at zero tilt
(Fig. 2.4). STEM images showed much more information throughout the tilt series,
and due to dynamic focusing images remained sharp even at high tilt angles (up to
73°). It was found that radiation damage in the STEM mode was considerably lower
than during TEM tilt series acquisition even when the average electron dose was
carefully equalized between the two modes. This is in agreement with our findings
with cryo-STEM tomography [20] (see Sect. 2.4), and could be explained as a
result of the short dwell time of the beam at each scanned point, which probably
allows for energy dissipation, hence to less damage. Images were significantly
improved when the semi-convergence angle was reduced from 30 to 3 mrad.
Interestingly, the authors claimed that due to the thickness of the sample “the
distinct advantages of HAADF imaging could unfortunately not be observed”.
Indeed, the group of Richard Leapman at the NIH optimized conditions for STEM
tomographic acquisition by advocating use of the BF detector instead [31]. The BF
detector was less sensitive to the deleterious effects of multiply-scattered electrons
(see 2.2.3). Thus, BF STEM was found to be more suitable for the imaging of thick
samples as the degradation of resolution due to the top-bottom effect was much
reduced compared to HAADF STEM. In addition, using a small semi-convergence
angle, while increasing the probe size at focus, resulted in less divergence of the
beam through the thick sample (see Sect. 2.2.3), as has been demonstrated elsewhere [27, 32, 58, 67–69].
Further development of STEM tomography for plastic sections include using
microprobe mode (i.e., very small semi-convergent angles) [69], dual-axis data
2 STEM Tomography in Biology
45
Sections
STEM tomography was first developed for materials science applications, mainly
using the HAADF detector, because it was appreciated that DF contrast provides
“true-projections” of the sample (Sects. 2.2.1, 2.2.2) [59]. HAADF STEM tomography was used also for biological applications to localize ultra-small immuno-gold
labels adsorbed to plastic sections [54, 55] and to localize membranes [60].
A major advantage of performing tomography in STEM mode is the ability to
raster the beam in lines parallel to the tilt axis of the tilted sample, while focusing
the beam according to the height of each line. Thus, by using dynamic focusing
mode [61, 62], focus gradients across highly tilted samples are avoided and all
regions of the resulting image remain in-focus. Dynamic focusing is implemented
in currently available STEM tomography software packages such as SerialEM [63],
Xplore3D [64] (FEI Company, The Netherlands), and others [65, 66]. This is in
contrast to CTEM tomography, in which only a narrow region of the tilted sample
close to the tilt axis is in focus, while the rest of the sample is blurred. This effect is
aggravated as thicker samples are observed due to changes of the contrast transfer
function.
The work of Aoyama et al. [32] demonstrated the unequivocal advantage of
STEM over zero-loss energy-filtered TEM tomography of thick (1 lm) sections
observed with 300 kV STEM. The quality of TEM images was low even at zero tilt
(Fig. 2.4). STEM images showed much more information throughout the tilt series,
and due to dynamic focusing images remained sharp even at high tilt angles (up to
73°). It was found that radiation damage in the STEM mode was considerably lower
than during TEM tilt series acquisition even when the average electron dose was
carefully equalized between the two modes. This is in agreement with our findings
with cryo-STEM tomography [20] (see Sect. 2.4), and could be explained as a
result of the short dwell time of the beam at each scanned point, which probably
allows for energy dissipation, hence to less damage. Images were significantly
improved when the semi-convergence angle was reduced from 30 to 3 mrad.
Interestingly, the authors claimed that due to the thickness of the sample “the
distinct advantages of HAADF imaging could unfortunately not be observed”.
Indeed, the group of Richard Leapman at the NIH optimized conditions for STEM
tomographic acquisition by advocating use of the BF detector instead [31]. The BF
detector was less sensitive to the deleterious effects of multiply-scattered electrons
(see 2.2.3). Thus, BF STEM was found to be more suitable for the imaging of thick
samples as the degradation of resolution due to the top-bottom effect was much
reduced compared to HAADF STEM. In addition, using a small semi-convergence
angle, while increasing the probe size at focus, resulted in less divergence of the
beam through the thick sample (see Sect. 2.2.3), as has been demonstrated elsewhere [27, 32, 58, 67–69].
Further development of STEM tomography for plastic sections include using
microprobe mode (i.e., very small semi-convergent angles) [69], dual-axis data
2 STEM Tomography in Biology
45