RESEARCH NEWS

Analytical Microscopy www.advmat.de

Quantitative Cryo-Scanning Transmission Electron

Microscopy of Biological Materials
Michael Elbaum

of biology will impact studies of other

Electron tomography provides a detailed view into the 3D structure of biolog- organic or hybrid materials as well.
ical cells and tissues. Physical fixation by vitrification of the aqueous medium In this short review, the electron micro-
provides the most faithful preservation of biological specimens in the native, scope is first introduced for nonspecial-
fully hydrated state. Cryo-microscopy is challenging, however, because of the ists, including a brief historical over-
view of scanning transmission electron
sensitivity to electron irradiation and due to the weak electron scattering of
microscopy (STEM). A number of recent
organic material. Tomography is even more challenging because of the depend- achievements are described in analysis
ence on multiple exposures of the same area. Tomographic imaging is typically of cryogenic specimens by STEM, with
performed in wide-field transmission electron microscopy (TEM) mode with emphasis on applications where quantifi-
phase contrast generated by defocus. Scanning transmission electron micro­ cation makes an important contribution to
understanding the system under study.
scopy (STEM) is an alternative mode based on detection of scattering from a
focused probe beam, without imaging optics following the specimen. While
careful configuration of the illumination and detectors is required to generate 1.1. TEM, STEM, and All That
useful contrast, STEM circumvents the major restrictions of phase contrast
TEM to very thin specimens and provides a signal that is more simply inter- The basic electron–optical configura-
preted in terms of local composition and density. STEM has gained popularity tion for electron microscopy is shown in
Figure 1.[1–3] The transmission electron
in recent years for materials science. The extension of STEM to cryo­microscopy
microscope operates very much like a
and tomography of cells and macromolecules is summarized herein. light microscope: a condenser illuminates
a field on the specimen, and an image
is projected by an objective lens onto a
1. Introduction camera. When the sample contains heavy elements that scatter
electrons strongly, or was stained with them as is the case for
Cryo-electron-microscopy (cryo-EM) is a burgeoning field at the most biological EM, image contrast is generated by placing an
forefront of modern science. Its application to structural biology aperture on axis in the objective back focal plane. A bright-field
was recognized by the Nobel Prize in chemistry for 2017. The (BF) image is then formed by electrons that have not under-
developments we report on here represent new extensions of gone strong scattering. Contrast appears as dark density on a
cryo-EM. We have married what had been a relatively neglected bright background. Contrast would be reversed if a stop instead
microscopy method in life science to the modern cryogenic of an aperture were placed on axis, as only the scattered flux
sample preparation and handling. Vitrification of the aqueous would contribute to a dark-field (DF) image. When the sample
medium locks the molecular contents in place, fixing them lacks heavy elements, as is normally the case for unstained
physically without adding or removing any material whatso- biological specimens, the scattering contrast is low and phase
ever. This holds great advantage for compositional analysis as contrast is generated by opening the objective aperture so as to
well as morphology. Perhaps the major potential is to convert permit interference between scattered and unscattered electron
the notion of the image from that of a picture into a quanti- waves. Some degree of defocus is required to produce suitable
tative, 3D map of material properties within the biological phase interference conditions.
context. This has required some reconsideration of a number In the STEM, the microscope lenses are used to converge the
of basic assumptions and traditions. Many important lessons incident beam to a fine spot, which is then rastered across the
are learned from materials science microscopy. We expect that surface. Electrons that pass through the sample reach area detec-
our experience with the very delicate cryogenic specimens tors on the opposite side. A disk-shaped detector positioned on
the optic axis collects the incident beam. To the extent that the
sample scatters electrons away, the BF image goes dark. An
Prof. M. Elbaum
annular dark-field (ADF) detector surrounding the optic axis col-
Department of Chemical and Biological Physics lects electrons that have been scattered away from the illumina-
Weizmann Institute of Science tion direction. In this case, the signal is zero with the sample
Rehovot 7610001, Israel removed; it increases in proportion to the scattering from the
E-mail: [email protected] specimen. The image is built point by point from these signals
DOI: 10.1002/adma.201706681 as the probe beam is scanned. (A conventional scanning EM does

Adv. Mater. 2018, 30, 1706681 1706681 (1 of 6) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

the same, but collects electrons emitted toward the illumination

side.) Note that two signals, or possibly even more, may be col- Michael Elbaum studies cell
lected for each point visited. biology from the perspec-
tive of soft matter physics.
Considerations of diffusion
1.2. A Brief History of Scanning Transmission and transport in the crowded
Electron Microscopy environment of the living cell
led him to 3D imaging by cry-
The first scanning electron microscope was in fact a scan- otomography based on soft
ning transmission EM. Invented by von Ardenne in 1938, the X-ray and scanning transmis-
STEM was in many ways a spin-off of his more famous inven- sion electron microscopies.
tion, the television. Perhaps this was the first intersection of
scientific microscopy with the entertainment industry. The
STEM was reinvented by Crewe et al. in the late 1960s.[4] One
of his primary motivations at the time was to study biological
specimens.[5] STEM has had enormous impact in the materials packing density of DNA in various biological contexts.[10] The
sciences, where a recurrent theme has been study of single major impact of quantifiable contrast in STEM emerged for
atoms. In fact the first view of isolated single atoms, uranium, mass measurement of macromolecules.[11] A particularly beau-
and thorium, was obtained by STEM on the basis of high angle tiful example was in the dissection of the nuclear pore complex
scattering.[6] Later developments led even to spectroscopy of to component substructures.[12] Ottensmeyer and co-workers
single atoms.[7] The introduction of aberration correctors pro- used STEM to study the structure of protein assemblies by
vided sub-Angstrom probes, with which it is possible to resolve single particle averaging.[13] The first study of fully hydrated,
and distinguish light atoms in 2D materials.[8] vitrified macromolecules using STEM was reported by Tra-
Carlemalm and Kellenberger showed that STEM could be chtenberg et al. in 1992.[14] Cryo-STEM was also combined with
used to study unstained biological sections.[9] Moreover, the electron energy loss spectroscopy (EELS) to map water in pro-
contrast could be quantified so as to estimate, for example, the tein solutions and erythrocytes in 1993.[15]
General interest in STEM for biological applications then
waned over time. Conventional TEM became progressively
more convenient with digital cameras and user-friendly micro-
scopes. Introduction of the new atomic force microscope also
drew attention away from the vacuum-dependent EM. Another
factor, undoubtedly, was the goal to improve resolution in pro-
tein structures toward the atomic scale. These efforts succeeded
spectacularly using wide-field phase contrast and a new genera-
tion of scintillator-free cameras. Lacking phase contrast in its
normal implementation, STEM was not the method of choice
for this application as the exposure required for high resolution
is prohibitively high.[16]
About a decade ago, several factors contributed to rekindle
interest in STEM for life science microscopy. Among these
were renewed appreciation for convenient chemical ana-
lytics, especially EELS and energy-dispersive X-ray spectros-
copy (EDX),[17,18] in situ microscopy of whole cells in liquid
medium,[19] and demonstrations of electron tomography by
Figure 1. Optical schematic representation of wide-field TEM and STEM. STEM using the conventional S/TEM platform.[20,21] Bright-
A) In TEM, a field of view is illuminated by the condenser, while the objec- field STEM tomography showed clear advantages over TEM for
tive forms an image on the camera. When the sample contains strong thick plastic sections.[22,23]
electron scatterers such as heavy metal stain, an aperture in the objective
back focal plane prevents the scattered electrons from contributing to
the image, resulting in bright-field contrast. For cryo-TEM, where strong
scatterers are typically absent, the objective aperture is opened so as 2. Introduction of Low-Dose Cryo-STEM
to allow phase interference between scattered and unscattered waves.
Contrast is typically generated by defocus, or alternatively by means of a 2.1. Cryo-Scanning Transmission Electron Tomography
phase plate, to produce interference conditions. B) In STEM, the micro-
scope optics focus the illumination to a spot that is scanned across the Cryogenic imaging of radiation-sensitive specimens is a world
specimen. Scattered electrons are lost from the illumination cone, pro- unto itself, often unappreciated by materials scientists. Every
ducing bright-field contrast. Alternatively, scattered electrons collected
inelastic collision deposits energy and has the potential to cause
in an annular detector produce dark-field contrast. Imaging can be opti-
mized by adjustment of the illumination angle α and detector inner and damage. For thin specimens, phase contrast makes more efficient
outer cutoff angles θ as shown. Adapted under the terms of the CC-BY use of the precious exposure than STEM. It had not been appre-
license.[3] Copyright 2015, The Authors, published by AIMS Press. ciated in the cryo-EM community that the opposite is so for thick

Adv. Mater. 2018, 30, 1706681 1706681 (2 of 6) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Image simulation for STEM indicated

that the bulk of the scattered flux would be
directed to low angles.[24] This is consistent
with the composition of light elements in
biological material, primarily hydrogen,
carbon, nitrogen, and oxygen. Detecting the
scattering to low angles requires a narrow
convergence angle for the illumination. This
puts a constraint on resolution since the dif-
fraction limit relates the convergence angle
to the probe diameter inversely. On the other
hand, the depth of field, defined for STEM as
the distance from ideal focus over which the
probe beam dia­meter varies negligibly, varies
inversely with the square of the convergence
angle. Thus, a small compromise in resolu-
tion may bring a large benefit in accessible
volume.[2]
STEM imaging on such thick samples
turned out to be more tolerant to radiation
exposure than TEM.[24] For light elements, the
cross section for inelastic scattering is larger
than that for elastic scattering.[18] Plasmon
excitations in the specimen create a broad
energy spread of the transmitted electrons.
Due to chromatic aberration in the magnetic
objective lens, each energy loss comes to
focus at a different plane, resulting in a hazy
image. The conventional solution is to use an
imaging spectrometer to remove the inelas-
tically scattered electrons before they reach
the camera. Damage is still done to the spec-
imen, however. The inelastic mean free path
in vitreous ice is about 200 nm (for 200 kV
acceleration), so for thicker specimens only a
small fraction of the image signal reaches the
camera. STEM, on the other hand, is relatively
Figure 2. Tomography of whole bacteria. A) Tomographic slices through 3D reconstructions
blind to inelastic scattering because there are
of Agrobacterium tumefaciens, shown in STEM annular dark-field, bright-field, and TEM as indi-
cated. The texture of STEM and TEM images differs qualitatively because of the different optical no image forming optics after the specimen.
transfer functions. The STEM images are less sharp but better represent specimen density than Inelastic scattering events divert the probe
the TEM. The prominent round feature at the upper pole of the bacterium is a polyphosphate electrons only by very small angles (small
body. B) Intensity profile along the lines indicated by white arrows. STEM preserves low spatial momentum transfer), so the signal recorded
frequencies and avoids the bright fringes surrounding dark features. C) Contrast in projection by either bright- or dark-field STEM detectors
through the polyphosphate body varies as a function of cutoff angle, from which an estimate of reflects primarily elastic scattering.
the phosphate concentration can be made. The figure is a composite image reproduced from
ref. [2], which was adapted from ref. [24] and [26]. Reproduced with permission.[2] Copyright
2016, Materials Research Society. Original images: A,B) Adapted with permission.[24] Copyright
2014, Springer Nature. C) Adapted with permission.[26] Copyright 2015 The Authors, Journal of 2.2. Quantitative Contrast in STEM
Microscopy, and Royal Microscopical Society.
If we consider the incoherent STEM image
specimens. Our first publication demonstrated cryo­ -scanning formation on a pixel by pixel basis, the elec-
transmission electron tomography (CSTET) using samples of tron flux reaching a detector is just the total scattering to the
the soil bacterium Agrobacterium tumefaciens.[24] This organism relevant directions along the beam path through the specimen.
is of special interest for its use in gene transfer to plants.[25] A The scattering per volume element depends on the atomic
rod-shaped bacterium, its diameter ranges from 600 to 800 nm. density, weighted by the relevant cross section, and integrated
This is too thick for conventional TEM tomography, especially at according to the detector geometry. Changes in composition
200 kV. Contrast in the TEM tomogram showed strong Fresnel or density produce contrast with magnitude proportional to
fringes due to defocus. Much more internal detail was visible in the relative material fractions. Underlying this simple picture
the STEM data. Bright- and dark-field reconstructions showed is an assumption that the specimen has no long range order
complementary, inverse contrast as shown in Figure 2. that would generate Bragg diffraction. Typical materials science

Adv. Mater. 2018, 30, 1706681 1706681 (3 of 6) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

specimens would have such long-range order, but typical bio- phosphorus. Since the granules showed no evidence of Bragg
logical specimens do not. diffraction, we could identify them as amorphous calcium
In examining the original tomograms of Agrobacterium, a phosphate. On the basis of a standard elemental composition,
prominent feature appears at the cell pole. In comparison with we could then estimate the density at about half that of the crys-
the cell membrane of dense carbon, its intensity is more promi- talline form. Given the spongy nature and the likelihood of pro-
nent in both BF and DF images. This suggests a different ele- tein inclusions this is quite plausible. In solid form, the gran-
mental composition. EDX identified the feature as a polyphos- ules can store large concentrations of ions, thereby buffering
phate body (PPB). By recording a series of dark-field signals the solution for availability to adenosine triphosphate synthesis,
with different cutoff angles, it was found that maximum con- signaling, or other biochemical functions. Notably, the granules
trast between the PPB and the cytoplasm occurred for an inner were lost in preparation of the cells for conventional embedded
cutoff of 40 mrad (Figure 2). By comparison with the predicted thin sections. Again this justifies the cryofixation in providing
scattering per atom, the image contrast could be explained by a an unmanipulated snapshot, frozen in time, of the local mate-
phosphorus concentration 3% that of carbon.[26] The strength rial distributions. The advantage of looking at the cell intact,
of cryoimaging for compositional studies lies in the simple without elaborate cryosectioning, is also quite obvious. Even if
sample preparation, with nothing added and nothing removed. such procedures were simple, the chance to catch sparse fea-
Tomography reconstructs the scattering distribution through tures is much enhanced when looking at large volumes.
the sample volume from projection images taken at different
angular views. Thus, the image contrast reflects the distribu-
tion of material density and composition in 3D. The bright- and 2.3. Protein-Bound Metals
dark-field reconstructions provide complementary informa-
tion to the extent that heavier elements scatter more strongly The sensitivity of STEM image contrast to such “heavy” ele-
to higher angles. According to the simplest picture, the BF ments as phosphorus and calcium, with Z = 15 and 20, respec-
image represents essentially all scattering (above a small cutoff tively, begs the question where is the limit. Does it extend to
that includes the incident beam), while the ADF image reflects single atoms? To address this question, we studied the common
preferentially the heavy elements. Ribosomes appear darker iron storage protein ferritin.[28] Zinc binds (and blocks) spe-
in BF and brighter in ADF than protein structures because cific sites on the protein where Fe undergoes oxidation, at a
of the denser oxygen and phosphorus in RNA. Strong signals stoichiometry of two atoms per site.[29] Cryo-STEM data were
in both BF and ADF virtual sections were seen from dense collected, this time with emphasis on resolution rather than
granules in the mitochondria of mammalian tissue culture depth of field, and reconstructed to 3D by single particle
cells[27] (Figure 3). EDX showed the presence of calcium and averaging methods (Figure 4). Peaks in the 3D distribution

Figure 3. Tomography of human tissue culture cells reveals the richness of organelle content in 3D. A) A slice through the tomogram in a region of
the cell ≈800 nm thick shows the plasma membrane (pm), filamentous actin (a), lipid droplets (dr), large vesicles (v), microtubules (mt), endoplasmic
reticulum (ER), and mitochondria (M). Within the mitochondria are large, dense matrix granules. Scale bar 1 µm. B) Manual segmentation renders the
same features through the 3D volume. Note the microtubules passing above and below the junction where the endoplasmic reticulum meets the mito-
chondrion, which is apparently undergoing fission. C) Matrix granules are represented by volume rendering; false color shows the density variations within
the granules. D) EDX over the yellow squares shows the excess of Ca and P in the granule-rich mitochondrion, indicative of calcium phosphate. Scale
bar 400 nm. E) Intensity thresholds provide quantitative measurements from which the density of amorphous calcium phosphate in the granules may be
extracted. Scale bar 400 nm. Adapted under the terms of the CC-BY license.[27] Copyright 2017, The Authors, published by eLife Sciences Publications, Ltd.

Adv. Mater. 2018, 30, 1706681 1706681 (4 of 6) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

Figure 4. Detection of isolated metal atoms on ferritin by cryo-STEM and single particle analysis. A) The low-resolution (20 Å) protein surface from
STEM reconstruction is shown in blue mesh, overlaid onto the atomic structure previously determined by X-ray crystallography[29] (PDB: 2CEI) shown
in gray. Yellow spheres mark Zn atoms in the X-ray structure. Solid blue surfaces show the contours of strong scattering in the STEM reconstruction,
which corresponds well to the Zn locations in the crystal structure. B) A central slice through the reconstructed 3D structure shows different material
features at distinctly different intensity contours, with the Zn atoms representing the strongest scattering. Adapted with permission.[28] Copyright 2017,
The Authors. Published by National Academy of Science, USA.

corresponded precisely to the Zn-binding sites known from as macromolecules, where the objective collects essentially all
X-ray crystallo­graphy. Detection of metals on metalloproteins the scattering and phase shifts are small. This limitation is
is in general a difficult problem in structural biology, so the much more restrictive than often appreciated.[16,37] Unsectioned
initial work on ferritin carries great potential for extension to cells are nor weak phase objects, nor are even single heavy
other systems. atoms in a conventional TEM. In such cases, STEM extends the
possibilities where TEM becomes ineffective. Practically, the
instrumentation required for STEM is of very low cost in com-
2.4. Spectroscopic Imaging of Radiation-Sensitive Materials parison with the camera and energy filter used in cryo-TEM.
This should make cryotomography available in contexts where
Aside from morphological preservation, the benefits of cryomi- cryo-TEM would be prohibitively expensive. Perhaps, ironically,
croscopy for suppression of radiation damage have long been the major impediment to cryo-STEM lies in its flexibility. With
recognized. Cryogenic conditions suppress sublimation of all the angles to keep track of, there are too many parameters
volatile components, as well as diffusion of radicals generated to adjust. Efforts to systematize those choices are underway,[38]
by irradiation. Intrinsically sparse materials, such as hydrogels optimizing the necessary trade-offs between large volume, high
and polyelectrolytes, present challenges similar to biological resolution, and elemental sensitivity.
materials, including the tendency to collapse catastrophically
under the beam. Cryo-STEM imaging was used, for example,
to study a stimulus-responsive hydrogel that would not likely Acknowledgements
have survived imaging at room temperature.[30] Tomography
The author thanks colleagues Sharon Wolf, Lothar Houben, Nadav
is an especially demanding enterprise because of the need for Elad, and Peter Rez, with whom the developments reported here have
multiple exposures of the same region on the specimen. Even been made. Work in the lab has been supported in part by a grant from
in hard materials, cryogenic sample cooling was justified on the the Israel Science Foundation (1285/14) and by the Irving and Cherna
basis of damage mitigation.[31] Another major benefit is seen Moskowitz Center for Bio and Nano-bio Imaging. The lab has benefited
in spectroscopy. STEM is very convenient for spatially resolved from the historical generosity of the Harold Perlman family.
elemental imaging by EELS or EDX, where long exposures are
required to collect weak signals.[32,33] Specifically for hydrated
Conflict of Interest
materials, cryo-STEM EELS has been used to map water in
polymer blends,[34] to characterize biphasic polymer nanocol- The authors declare no conflict of interest.
loids suspended in water,[35] and to evaluate hydrogen evolution
from biological material as a result of radiation damage in TEM
imaging.[36] Keywords
analytical microscopy, cryo-electron microscopy, cryotomography, cryo-
scanning transmission electron tomography, STEM
3. Conclusions
Received: November 15, 2017
Where is the impact of cryo-STEM most likely to be felt? Cryo- Revised: January 1, 2018
TEM is well optimized for the realm of weak phase objects such Published online: May 11, 2018

Adv. Mater. 2018, 30, 1706681 1706681 (5 of 6) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.advmat.de

[1] L. Reimer, H. Kohl, Transmission Electron Microscopy: Physics of [20] K. Aoyama, T. Takago, A. Hirase, A. Miyazawa, Ultramicr. 2008, 109, 70.
Image Formation, Springer Science & Business Media, New York [21] A. E. Yakushevska, M. N. Lebbink, W. J. C. Geerts, L. Spek,
2008. E. G. van Donselaar, K. A. Jansen, B. M. Humbel, J. A. Post,
[2] M. Elbaum, S. G. Wolf, L. Houben, MRS Bull. 2016, 41, 542. A. J. Verkleij, A. J. Koster, J. Struct. Biol. 2007, 157, 381.
[3] D. Kirchenbuechler, Y. Mutsafi, B. Horowitz, S. Levin-Zaidman, [22] M. F. Hohmann-Marriott, A. A. Sousa, A. A. Azari, S. Glushakova,
D. Fass, S. G. Wolf, M. Elbaum, AIMS Biophys. 2015, 2, 259. G. Zhang, J. Zimmerberg, R. D. Leapman, Nat. Methods 2009,
[4] A. V. Crewe, J. Wall, L. M. Welter, J. Appl. Phys. 1968, 39, 5861. 6, 729.
[5] A. V. Crewe, Phil. Trans. R. Soc. Lond. B 1971, 261, 61. [23] W. J. Hickey, A. R. Shetty, R. J. Massey, D. B. Toso, J. Austin II,
[6] A. V. Crewe, J. Wall, J. Langmore, Science 1970, 168, 1338. J. Microsc. 2017, 265, 3.
[7] C. Colliex, A. Gloter, K. March, C. Mory, O. Stéphan, K. Suenaga, [24] S. G. Wolf, L. Houben, M. Elbaum, Nat. Methods 2014, 11, 423.
M. Tencé, Ultramicroscopy 2012, 123, 80. [25] S. B. Gelvin, Microbiol. Mol. Biol. Rev. 2003, 67, 16.
[8] O. L. Krivanek, M. F. Chisholm, V. Nicolosi, T. J. Pennycook, [26] S. G. Wolf, P. Rez, M. Elbaum, J. Microsc. 2015, 260, 227.
G. J. Corbin, N. Dellby, M. F. Murfitt, C. S. Own, Z. S. Szilagyi, [27] S. G. Wolf, Y. Mutsafi, T. Dadosh, T. Ilani, Z. Lansky, B. Horowitz,
M. P. Oxley, S. T. Pantelides, S. J. Pennycook, Nature 2010, 464, 571. S. Rubin, M. Elbaum, D. Fass, eLife 2017, 6, e29929.
[9] E. Carlemalm, E. Kellenberger, EMBO J. 1982, 1, 63. [28] N. Elad, G. Bellapadrona, L. Houben, I. Sagi, M. Elbaum,
[10] B. Bohrmann, M. Haider, E. Kellenberger, Ultramicroscopy 1993, 49, Proc. Natl. Acad. Sci. USA 2017, 114, 11139.
235. [29] L. Toussaint, L. Bertrand, L. Hue, R. R. Crichton, J.-P. Declercq,
[11] J. Wall, J. Hainfield, Ann. Rev. Biophys. Biophys. Chem. 1986, 15, 335. J. Mol. Biol. 2007, 365, 440.
[12] R. Reichelt, A. Holzenburg, E. L. Buhle, M. Jarnik, A. Engel, U. Aebi, [30] K. Shikinaka, K. Kaneda, S. Mori, T. Maki, H. Masunaga, Y. Osada,
J. Cell Biol. 1990, 110, 883. K. Shigehara, Small 2014, 10, 1813.
[13] R. Z.-T. Luo, D. R. Beniac, A. Fernandes, C. C. Yip, F. P. Ottensmeyer, [31] K. Yoshida, Y. H. Ikuhara, S. Takahashi, T. Hirayama, T. Saito,
Science 1999, 285, 1077. S. Sueda, N. Tanaka, P. L. Gai, Nanotechnology 2009, 20, 315703.
[14] S. Trachtenberg, K. Leonard, W. Tichelaar, Ultramicroscopy 1992, 45, [32] Y. Kubo, A. Mizoguchi, J. Fujita, Anal. Chem. 2016, 88, 5225.
307. [33] R. Sahore, B. D. A. Levin, M. Pan, D. A. Muller, F. J. DiSalvo,
[15] S. Sun, S. Shi, R. Leapman, Ultramicroscopy 1993, 50, 127. E. P. Giannelis, Adv. Energy Mater. 2016, 6, 1600134.
[16] P. Rez, Ultramicroscopy 2003, 96, 117. [34] A. Sousa, A. Aitouchen, M. Libera, Ultramicroscopy 2006, 106, 130.
[17] M. A. Aronova, R. D. Leapman, MRS Bull. Mater. Res. Soc. 2012, 37, 53. [35] G. Kim, A. Sousa, D. Meyers, M. Libera, Microsc. Microanal. 2008,
[18] A. Engel, in Adv. Imaging Electron Phys. (Ed.: Peter W. Hawkes), 14, 459.
Elsevier, 2009, pp. 357–386. [36] S. Yakovlev, M. Misra, S. Shi, M. Libera, J. Microsc. 2009, 236, 174.
[19] N. de Jonge, D. B. Peckys, G. J. Kremers, D. W. Piston, Proc. Natl. [37] P. J. Koeck, A. Karshikoff, J. Microsc. 2015, 259, 197.
Acad. Sci. U. S. A. 2009, 106, 2159. [38] P. Rez, T. Larsen, M. Elbaum, J. Struct. Biol. 2016, 196, 466.

Adv. Mater. 2018, 30, 1706681 1706681 (6 of 6) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim