many more microscope parameters, which will again be important here (e.g.
apertures, spot size and use of high current modes). It is difficult to give figures for
which parameters should be used with particular sample types as this varies based
on the processing protocol and the microscope used. With such variability it is
necessary to do a lot of testing with a system to gain knowledge of what will work
with a particular sample type. However, it is important to have a place to start and
reviewing literature can provide this (see Table 5.1 for a summary of examples).
The figures in this table also show the variability in parameters that can be used to
achieve results.
As mentioned previously, it is important to use low voltages and as a standard
this is in the range of 1–3 kV, although voltages as high as 4 kV have been used
[57]. This lowers the interaction volume so it is less than the section thickness but
also lessens the charging. Of course hand-in-hand with this is lower backscatter
signal production from the sample and so deterioration of SNR. Small changes in
voltage can make substantial changes in the image quality and the beam damage on
the sample. Figure 5.5 shows that with only a change in voltage from 1.5 to 1.8 kV
there is a considerable increase in signal but also an increase in the effects of
charging (Fig. 5.5b, c).
Vacuum conditions can help and many groups use low vacuum mode in order to
dissipate the charges. However, this also leads to a drop in image quality with a
lower SNR: the poorer the vacuum, the poorer the image. To compensate, it may be
necessary to use a higher voltage or a longer dwell time. While high vacuum has
been used successfully [9, 58], a wide range in low vacuum conditions have also
been found to be useful from 5 Pa [9] to 40–50 Pa [57, 59, 60]. An increase in
dwell time can improve the SNR, but the compromise here is that this can lead to
sample damage and will of course extend the acquisition time. The figures in the
literature show an incredibly varied range from 2 ls [9] to 15–20 ls [61].
The section thickness dictates the resolution in the z direction and cutting as thin
as 10 nm can be achieved on some biological samples [9]. However, this is not
usually possible. If high resolution is not the object of the study then sections of
100 nm may be enough and allow acquisition of more data in z in a shorter time, for
example by Starborg et al. and Scheuring et al. [57, 60].
When setting up a SBEM run it must also be taken into consideration how long to
dedicate the instrument to this one data collection. It is a relatively time-consuming
process. It can take 11 h to section and then capture 500 images at 4 k  4 k pixel
resolution with a 4 ls dwell time, a full volume size of 46 Â 46 Â 25 lm. This
would give 500 images with a resolution of 11 Â 11 Â 50 nm. By many standards
this is a small amount of data. A SBEM is a big investment of money and to tie it up
for many hours or days for one data collection is possibly false economy. Montages
make it possible to collect information over a large area but with no drop off in
resolution. For example, Helmstaedter et al. [62] collected images at 16.5 Â 23 nm
with a 4 Â 4 montage to acquire a total volume of 1,200,000 mm
3 , while Briggman
et al. [63] using similar resolutions and a 1 Â 7 mosaic acquired 6,300,000 mm
3 of
data, both from the mouse retina.
132
R. I. Webb and N. L. Schieber
apertures, spot size and use of high current modes). It is difficult to give figures for
which parameters should be used with particular sample types as this varies based
on the processing protocol and the microscope used. With such variability it is
necessary to do a lot of testing with a system to gain knowledge of what will work
with a particular sample type. However, it is important to have a place to start and
reviewing literature can provide this (see Table 5.1 for a summary of examples).
The figures in this table also show the variability in parameters that can be used to
achieve results.
As mentioned previously, it is important to use low voltages and as a standard
this is in the range of 1–3 kV, although voltages as high as 4 kV have been used
[57]. This lowers the interaction volume so it is less than the section thickness but
also lessens the charging. Of course hand-in-hand with this is lower backscatter
signal production from the sample and so deterioration of SNR. Small changes in
voltage can make substantial changes in the image quality and the beam damage on
the sample. Figure 5.5 shows that with only a change in voltage from 1.5 to 1.8 kV
there is a considerable increase in signal but also an increase in the effects of
charging (Fig. 5.5b, c).
Vacuum conditions can help and many groups use low vacuum mode in order to
dissipate the charges. However, this also leads to a drop in image quality with a
lower SNR: the poorer the vacuum, the poorer the image. To compensate, it may be
necessary to use a higher voltage or a longer dwell time. While high vacuum has
been used successfully [9, 58], a wide range in low vacuum conditions have also
been found to be useful from 5 Pa [9] to 40–50 Pa [57, 59, 60]. An increase in
dwell time can improve the SNR, but the compromise here is that this can lead to
sample damage and will of course extend the acquisition time. The figures in the
literature show an incredibly varied range from 2 ls [9] to 15–20 ls [61].
The section thickness dictates the resolution in the z direction and cutting as thin
as 10 nm can be achieved on some biological samples [9]. However, this is not
usually possible. If high resolution is not the object of the study then sections of
100 nm may be enough and allow acquisition of more data in z in a shorter time, for
example by Starborg et al. and Scheuring et al. [57, 60].
When setting up a SBEM run it must also be taken into consideration how long to
dedicate the instrument to this one data collection. It is a relatively time-consuming
process. It can take 11 h to section and then capture 500 images at 4 k  4 k pixel
resolution with a 4 ls dwell time, a full volume size of 46 Â 46 Â 25 lm. This
would give 500 images with a resolution of 11 Â 11 Â 50 nm. By many standards
this is a small amount of data. A SBEM is a big investment of money and to tie it up
for many hours or days for one data collection is possibly false economy. Montages
make it possible to collect information over a large area but with no drop off in
resolution. For example, Helmstaedter et al. [62] collected images at 16.5 Â 23 nm
with a 4 Â 4 montage to acquire a total volume of 1,200,000 mm
3 , while Briggman
et al. [63] using similar resolutions and a 1 Â 7 mosaic acquired 6,300,000 mm
3 of
data, both from the mouse retina.
132
R. I. Webb and N. L. Schieber