and 100 nm samples when they are highly tilted. The mentioned limit is less
deteriorative than the limit associated to defocus determination.
10.3.3 STEM Tomography
Atomic resolution of 2.4 Å was obtained for tomograms of a *10 nm thick gold
nanoparticle [51]. It was recorded in ADF-STEM modality with a maximum tilt
angle of 72.6° and up to 69 images per tilt series, pixel size 0.42 Å and a total
electron dose of 7.6 Â 10e6 e
− /Å
2 . Tilt series were aligned using centres of mass;
tomographic reconstruction was generated with equally sloped tomography [52]. As
described in Chap. 2 of this book, STEM tomography may also be applied to
cryo-preserved samples. The maximal tolerable electron dose that may be applied
on a cryo sample preserving near atomic resolution is *200,000 times lower. This
will result in noisier tomograms however may be compensated for using subtomogram averaging. In case of the ideally aligned particles *200,000 asymmetric
units could be enough, however in practice much more particles may be needed.
STEM imaging gives an advantage over TEM in obtaining higher resolution for
cryo-samples thicker than 500 nm (using 300 kV electrons) [53]. One of the limitations during recording STEM tomograms of radiation sensitive samples is to have
the sample thickness at high tilts match the depth of field. For a 1-micron thick sample
the effective probe size is 3–4 nm [53] and a sampling limit for resolution of 6–8 nm.
It is not clear if STEM tomography in combination with StA is beneficial in terms of
obtainable resolution over TEM tomography for thin cryo-preserved samples.
10.3.4 Measuring the Resolution of Tomograms
As discussed earlier in this chapter several factors limit the resolution of electron
tomograms; therefore, additionally to the theoretical estimates several practical
ways to measure the resolution were suggested. Knowing the resolution is useful
for interpretation and visualization of the data—band pass or other filters may be
designed based on the maximum resolution. Additionally, it is important for
subtomogram averaging (below) to understand up to which resolution the particles
have reliable signal. An intuitive way to estimate the resolution is to see the defined
features inside the tomograms—resolving a lipid bilayer means that the resolution
of a tomogram is at least 4 nm.
Two quantitative criteria have been proposed based on signal-to-noise levels of
the tomogram [54]: the first method includes splitting the tilt series into even and
odd projections, generating two tomograms and calculating Fourier shell correlation
between them, which has the downside of losing half of the signal during FSC
calculation and depending on interpolated Fourier voxels. Second approach is to
compare the original projection images and the corresponding re-projections of the
tomogram lacking the given projections. The advantage of the second method is the
272
M. Kudryashev
deteriorative than the limit associated to defocus determination.
10.3.3 STEM Tomography
Atomic resolution of 2.4 Å was obtained for tomograms of a *10 nm thick gold
nanoparticle [51]. It was recorded in ADF-STEM modality with a maximum tilt
angle of 72.6° and up to 69 images per tilt series, pixel size 0.42 Å and a total
electron dose of 7.6 Â 10e6 e
− /Å
2 . Tilt series were aligned using centres of mass;
tomographic reconstruction was generated with equally sloped tomography [52]. As
described in Chap. 2 of this book, STEM tomography may also be applied to
cryo-preserved samples. The maximal tolerable electron dose that may be applied
on a cryo sample preserving near atomic resolution is *200,000 times lower. This
will result in noisier tomograms however may be compensated for using subtomogram averaging. In case of the ideally aligned particles *200,000 asymmetric
units could be enough, however in practice much more particles may be needed.
STEM imaging gives an advantage over TEM in obtaining higher resolution for
cryo-samples thicker than 500 nm (using 300 kV electrons) [53]. One of the limitations during recording STEM tomograms of radiation sensitive samples is to have
the sample thickness at high tilts match the depth of field. For a 1-micron thick sample
the effective probe size is 3–4 nm [53] and a sampling limit for resolution of 6–8 nm.
It is not clear if STEM tomography in combination with StA is beneficial in terms of
obtainable resolution over TEM tomography for thin cryo-preserved samples.
10.3.4 Measuring the Resolution of Tomograms
As discussed earlier in this chapter several factors limit the resolution of electron
tomograms; therefore, additionally to the theoretical estimates several practical
ways to measure the resolution were suggested. Knowing the resolution is useful
for interpretation and visualization of the data—band pass or other filters may be
designed based on the maximum resolution. Additionally, it is important for
subtomogram averaging (below) to understand up to which resolution the particles
have reliable signal. An intuitive way to estimate the resolution is to see the defined
features inside the tomograms—resolving a lipid bilayer means that the resolution
of a tomogram is at least 4 nm.
Two quantitative criteria have been proposed based on signal-to-noise levels of
the tomogram [54]: the first method includes splitting the tilt series into even and
odd projections, generating two tomograms and calculating Fourier shell correlation
between them, which has the downside of losing half of the signal during FSC
calculation and depending on interpolated Fourier voxels. Second approach is to
compare the original projection images and the corresponding re-projections of the
tomogram lacking the given projections. The advantage of the second method is the
272
M. Kudryashev