Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Materials Science & Engineering R 139 (2020) 100516

Contents lists available at ScienceDirect

Materials Science & Engineering R


journal homepage: www.elsevier.com/locate/mser

New opportunities in transmission electron microscopy of polymers T


a b b a,b,c,⁎
Brooke Kuei , Melissa P. Aplan , Joshua H. Litofsky , Enrique D. Gomez
a
Materials Science and Engineering, The Pennsylvania State University, University Park, PA 16802, USA
b
Chemical Engineering, The Pennsylvania State University, University Park, PA 16802, USA
c
Materials Research Institute, The Pennsylvania State University, University Park, PA 16802, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Recent advances in instrumentation for transmission electron microscopy have pushed the resolution limit,
Soft materials leading to remarkable instruments capable of imaging at 0.5 Å. But, when imaging soft materials, the resolution
TEM is often limited by the amount of dose the material can handle rather than the instrumental resolution. Despite
STEM the strong constraints placed by radiation sensitivity, recent developments in electron microscopes have the
Radiation damage
potential to advance polymer electron microscopy. For example, the focused ion beam creates opportunities for
In-situ
High resolution imaging
site-specific imaging, recently developed sample holders enable liquid TEM, monochromated sources lead to
spectroscopy and imaging based on the valence electronic structure, and direct electron detectors minimize the
required dose for imaging. Transmission electron microscopy has transformed the field of polymer science, and it
is poised to do so again in the near future.

1. Introduction necessary optimization of dose to achieve good spatial resolution [7].


Mapping of the lithium distribution within nanostructured block co-
Transmission electron microscopy (TEM) has been an invaluable polymers [8] and sulfur in mixtures of conjugated polymers [9–12] has
characterization tool throughout the latter history of polymer science. also been demonstrated.
As early as 1959, electron microscopy elucidated the nucleation me- Despite the tremendous progress that has been achieved in polymer
chanism for the formation of Nylon-6 spherulites [1]. In 1985, electron science with electron microscopy, the resolution limit for imaging of
diffraction coupled with high resolution electron microscopy surpassed soft materials is hindered by their inherently low contrast and their
the accuracy of X-ray diffraction in the structure determination of Cy- beam sensitivity. In 1971, Glaeser expressed Rose’s analysis of the
anine Green and demonstrated the feasibility of direct imaging of or- limits of the image formation process in the context of resolution in the
ganic molecules in crystals [2]. Towards the turn of the twenty-first TEM:
century, cryogenic TEM revealed a new class of tough, thin-shelled SN
block copolymer capsules [3] and toroid formation from triblock co- Cd ≥
fDC (1)
polymers [4]. Electron microscopy has also evolved from two-dimen-
sional (2D) projections to three-dimensional (3D) volume renderings. where C is the contrast, d is the smallest resolvable feature size, SN is
For example, in 2003, electron tomography revealed the coexistence of the minimum acceptable signal-to-noise ratio, DC is the critical electron
three kinds of cylindrical structures within a star ABC terpolymer in 3D dose, and f is the fraction of electrons that actually enter the lens
[5]. More recently, cryogenic electron tomography demonstrated the aperture and contribute to the image [13]. From this equation, it is
self-assembly of crystalline nanotubes from amphiphilic diblock copo- immediately apparent that higher resolution requires an increase in
lymers [6]. Imaging and diffraction at the nanoscale have transformed electron dose. Herein lies the problem: for beam sensitive materials like
our understanding of polymer structure and often function. polymers, beam damage can destroy the specimen such that an image
Analytical techniques within TEM, such as electron energy-loss with high contrast may no longer contain accurate information at the
spectroscopy (EELS), have also been instrumental in polymer science. desired resolution.
EELS has been used to map the spatial distribution of water in a frozen- Nevertheless, recent developments in instrumentation, often not
hydrated polymer despite the complication of distinguishing small designed for soft materials, provide opportunities to overcome many of
variations in water content from noise, a task that highlighted the the challenges of polymer microscopy. The purpose of this Review is to


Corresponding author at: Chemical Engineering, The Pennsylvania State University, University Park, PA 16802, USA.
E-mail address: edg12@psu.edu (E.D. Gomez).

https://doi.org/10.1016/j.mser.2019.100516
Received 19 August 2018; Accepted 23 July 2019
Available online 30 September 2019
0927-796X/ © 2019 Published by Elsevier B.V.
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

summarize the accomplishments of polymer microscopy thus far while sensitive materials, making spherical aberration correctors important
highlighting the evolution towards new opportunities made possible by for extending the point resolution limit in soft materials.
technological advancements such as monochromators, aberration cor- In addition to spherical aberrations, chromatic aberrations arise
rectors, phase plates, direct electron detectors, specimen preparation from the inability of lenses to focus electrons of different energies to the
tools, and in situ sample holders. Altogether, the field of polymer same point; this is particularly problematic for thicker samples. As a
electron microscopy is poised to make significant advances in the near consequence, chromatic aberrations can be limiting for polymeric ma-
future. terials when thicker samples are desired or required, such as in imaging
of multiscale structures or for samples that are challenging to thin
below 100 nm. Furthermore, energy-filtered TEM can be a powerful
2. Pushing the resolution limit with new instrumentation
tool to image polymeric systems with heterogeneity in elemental
composition, such as in salt-containing block copolymers [8] or in
Even with a perfect lens, the resolution of any imaging system must
polymer-fullerene blends used in organic photovoltaics [12,17–20].
be defined in terms of the Rayleigh criterion because diffraction of rays
But, chromatic aberrations may limit the resolution in energy-filtered
at the outermost collection angles of a finite sized lens results in a point
imaging. Chromatic aberration correctors have been demonstrated to
being imaged as an Airy disk. Practically, in addition to this diffraction-
push elemental mapping to atomic resolution in hard materials
limited resolution, resolution in the TEM is governed by inherent
[21–23], to improve resolution at low acceleration voltages (˜80 kV),
aberrations that come from imperfect lenses. Section 2.1 will describe
and for thick samples [24,25], but has not been sufficiently explored for
spherical and chromatic aberrations in the context of imaging of soft
soft materials. We discuss spherical and chromatic aberration correctors
materials, as well as the added constraint of dose-limited resolution in
in the next section, and highlight a few opportunities for soft material
soft materials. Section 2.2 will describe how new advancements in in-
electron microscopy in Sections 5.2 and 7.2.
strumentation have the potential to push the resolution of soft material
Nevertheless, the main limitation in resolution for soft materials
imaging to new limits.
comes from their sensitivity to the beam: contrast and sensitivity to
beam damage are interrelated, as the latter limits the electron dose that
2.1. Theoretical limits of the TEM can be used for imaging. The contrast C between domains must be
larger than the minimum acceptable signal-to-noise ratio SN. Taking DC
The resolution of a TEM is defined as the minimum resolvable dis- to be the maximum dose a sample can handle and d as the smallest
tance within the specimen and arises as a consequence of aberrations in resolvable feature size, the noise in an image is 1/ d 2DC . Thus, the
the microscope. In TEM, the resolution is governed by the quality of the product Cd must be greater than SN / DC , as shown in Eq. (1) [13,26].
objective lens, whereas in STEM resolution is limited by the size of the
probe. In both cases, this resolution is ultimately limited by a combi- 2.2. Development of new instrumentation
nation of spherical and chromatic aberrations in the microscope. For
soft materials, beam sensitivity adds the constraint of dose-limited re- Although often not designed with polymers in mind, new in-
solution (Eq. (1)). strumentation, such as monochromators, aberration correctors, phase
Spherical aberrations arise because electrons that are different dis- plates, and direct electron detectors, can offer significant advantages for
tances from the optical axis are focused to different points, causing imaging of polymeric materials. In this section, we briefly describe
smearing in the image [14–16]. In particular, the rays scattered to high these instrumentation advances in the context of hard materials and
angles, which carry information about the smaller spacings in the ob- polymer microscopy, and discuss opportunities in greater detail in later
ject, are incorrectly focused because they are closer to the electro- sections.
magnetic lenses and deflected more than they should be. If we assume A monochromator is an optical component that can transmit a
that there is no chromatic aberration, the resolution is limited by a chosen narrow band of wavelengths of radiation; in a TEM, its purpose
combination of the aforementioned Rayleigh criterion and spherical is to select electrons of a certain energy emitted from the source. The
aberration which together results in a point resolution limit approxi- four types of monochromators include the single Wien filter (FEI), the
1
mated by (Cs λ3) 4 , where CS is the coefficient of spherical aberration and double Wien filter (JEOL), the omega-shaped electrostatic mono-
λ is the wavelength of the incident electron. Although we can improve chromator (CEOS), and alpha-type magnetic monochromator (NION).
this resolution by decreasing λ with higher accelerating voltages, it can At 100 kV, the energy spread of Tungsten, LaB6, Schottky FEG, and cold
be impractical, it decreases the scattering cross-section (potentially FEG sources are 3 eV, 1.5 eV, 0.7 eV, and 0.3 eV, respectively [27].
requiring thicker samples), and it is potentially harmful for beam Monochromators can thus reduce the energy spread of the beam; for

Fig. 1. (A) Zero loss peaks acquired in vacuum using conventional FEG (Schottky source) and with a monochromator. Reprinted from [28], with permission from
Elsevier. Images and corresponding FFTs of gold particles (B) without a monochromator with an energy resolution of 0.93 eV and (C) with a monochromator with an
energy resolution of 0.10 eV. The distance from the center of an FFT represents the spatial frequencies present in the image; use of a monochomator increases the
achievable resolution. Reprinted from [30], with permission from Elsevier.

2
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 2. (A) Schematic of lens assembly of a hexapole aberration corrected TEM showing the path of intermediate images (axial ray, red) and diffraction images (field
ray, green). The net effect of this lens assembly is the correction of spherical aberrations up to third order. Reprinted from [31], with permission from Elsevier.
Comparison of holograms of a GaAs/AlAs-multilayer recorded by (B) an uncorrected CM30 Special TEM and (C) an aberration corrected Tecnai F20 TEM, showing
the improved contrast achieved through aberration correction. Reprinted with permission from [32]. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)

example, reducing the spread of a 0.7 eV FEG beam to 0.18 eV (Fig. 1A) generated from electrons hitting the detector are integrated over a fixed
[28]. Because a large energy spread is the main contributor to chro- frame rate and then summed for the total exposure time in a single
matic aberration, monochromators can lower the information limit image. In contrast, the high-speed readout capability of direct detectors
from 0.7 Å to 0.5 Å [29]. In addition, monochromators allow for about allows for counting mode, where individual electron events are de-
0.1 eV energy resolution, which has practical applications in low-loss tected such that noise arising from signal readout and scattered electron
electron energy loss spectroscopy (Section 4) and spectrum imaging signal are rejected. The final image is then recorded as a stack of high
(Section 7) of polymers. Use of monochromators has vastly improved signal-to-noise frames. Examples of the application of direct electron
the information limit of imaging in hard materials (Fig. 1B, C) [30], and detectors in imaging techniques for soft materials such as HRTEM, 3D
has the potential to do so in soft materials as well, both for spectroscopy reconstruction, and 4D STEM are discussed in later sections.
and imaging.
Aberration correctors have also advanced resolution limits. 3. Advanced sample preparation techniques for new microscopy
Multipole (non-round) lenses generate a negative spherical aberration opportunities
CS to cancel the positive CS of objective (round) lenses, thus creating a
net zero CS (Fig. 2). In this way, rays scattered at high angles with A key limitation in TEM of many polymeric materials lies in sample
respect to the optical axis (these are the rays carrying information about preparation. Thus, before discussing the applications of new microscope
smaller spacings in the object) are brought to the correct focus, thereby instrumentation to polymers, we first discuss the advanced sample
allowing access to higher frequencies and improving contrast (Fig. 2B, preparation techniques that make polymer microscopy feasible. With a
C). There are two types of correctors, octupole/quadrupole correctors high energy (80–300 keV) electron beam, samples must maintain
and hexapole correctors. Section 5, which discusses high-resolution electron transparency and be thin enough to prevent multiple scattering
TEM, includes examples of aberration correctors in practice. in order to accurately examine composition, microstructure, and phase
Another advancement that has applications in soft materials ima- separation (this is a thickness on the order of 10 s to 100 s of nan-
ging is the advent of the phase plate, a device inserted in the diffraction ometers) [36–39]. Coupled with the need for minimally deformed
plane (Fig. 3A) that modulates the contrast transfer function (CTF) samples devoid of artefacts, sample preparation techniques must be
(Fig. 3B). Traditionally, defocus phase contrast is generated by inherent able to generate a large amount of thin, uniform, and high-quality
(spherical aberrations) or induced (defocus) aberrations (Fig. 3C). samples in a repeatable fashion [40].
Nevertheless, this method results in a loss of information, particularly
in the low frequency region. Alternatively, contrast can be improved
3.1. Focused ion beam
through the use of a phase plate (Fig. 3D). The primary purpose of a
phase plate is to change the sine-type CTF of a conventional TEM to a
One commonly used technique for creating samples suitable for the
cosine-type function, thus improving contrast at low frequencies. The
TEM is the focused ion beam (FIB). The FIB is able to prepare films in
two main types of phase plates are Zernike (generates a circularly
multiple ways, or it can thin existing films, removing possible artefacts
symmetric modulation pattern) and Hilbert (generates an asymmetric
and achieving the necessary sample thickness. Films produced by the
modulation pattern). Section 8 includes several examples of phase
FIB are electron-transparent and often show little or no effect from
plates applied to soft materials.
sample preparation [41–46]. The FIB instrument operates similarly to a
More recently, direct electron detectors have created new oppor-
scanning electron microscope, where the focused beam is capable of
tunities for the imaging of beam sensitive materials. In a traditional
etching away most samples. Often, FIB instruments are equipped with a
charge-coupled device (CCD), a scintillator converts primary electrons
second electron beam, to enable site-selective milling with nanometer
into photons before they hit the detector; this inefficiency limits the
precision. For FIB usage in sample preparation, a gallium source is
resolution and signal-to-noise ratio. On the other hand, a direct electron
contacted with a nanometer-sized tungsten needle, and when coupled
detector bypasses the scintillation step (Fig. 4). Improving the signal-to-
with a heavy electric field, ions are emitted through the tungsten tip
noise ratio is especially useful for beam sensitive materials where, as
[41,42]. Gallium is generally used as the liquid-metal ion source be-
discussed in Section 2.1, there is a limit to how much dose the sample
cause of its low vapor pressure, volatility, and melting temperature
can withstand. In addition, direct electron detectors have the ability to
[47]. Alternatively, noble gas ion sources have recently become avail-
operate in two distinct modes. In linear mode, the accumulated charge
able. As these ions travel down the ion column and raster over the

3
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 3. (A) Schematic of TEM with phase plate.


(B) Moduli of contrast transfer function (CTF)
with and without phase plate. Reprinted from
[33], with permission from Elsevier. Lyophi-
lized GroEL chaperonin protein imaged with
(C) conventional TEM (underfocus 1960 nm)
and (D) with a Zernike phase plate. Reprinted
from [34], with permission from Elsevier.

Fig. 4. Schematics of a direct electron detector


vs a traditional CCD (top). Comparison of
sulfur elemental maps of the polymer/fullerene
blend poly(3-hexylthiophene-2,5-diyl)/ [6,6]-
phenyl-C61-butyric acid methyl ester (P3HT/
PCBM) taken with a direct electron detector
(K2) and traditional CCD (UltraScan) at iden-
tical imaging conditions (bottom). Reprinted
with permission from [35].

4
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 5. H-bar sample preparation for TEM.


Schematic diagram (a) and image (b) detailing
traditional H-bar sample preparation. After the
sample is mechanically cut from the bulk, it is
placed on a TEM grid and further polished and
thinned to reduce time under the ion beam.
Then, the region of interest is milled by the FIB
to an electron-transparent thickness. Reprinted
from [51], with permission from Elsevier.

sample, they are able to precisely remove sample material by sputtering normal tilt at 0° and then thinned further using progressively smaller
it away, allowing for nanoscale milling [48]. ion beams and completely cut free from the bulk specimen. After put-
As stated above, the key advantages of FIB sputtering and milling ting the lifted-out sample under an optical microscope, a micro-
are site-selectivity at the nanoscale coupled with low amounts of in- manipulator tip can place the fully milled and electron transparent
duced damage on the sample or specimen. Due to its nanometer-sized sample onto a TEM grid. The INLO technique allows for thinner milling
tungsten tip and ion column, the ability to control the milling thickness than EXLO or conventional FIB milling, because sections produced by
and location is high [43]. Additionally, because the incident ion beam EXLO cannot be milled further once they are placed on the grid [50].
column mills the sample at an angle that is not perpendicular to the Nevertheless, INLO samples may be slightly less robust than EXLO,
sample, homogeneous thinning can be achieved even within rough requiring great care to ensure the specimen is not damaged during
samples [49]. This allows for the preparation of samples with specific handling or transfer to the TEM.
geometries from film or bulk samples, enabling significant flexibility in Fig. 6e and f show examples of the in-situ lift-out technique used on
sample preparation [50]. block copolymer thin films. Although damage was apparent at the top
Traditional FIB preparation, sometimes referred to as the H-bar film surface, the lamellar structure of poly(styrene-block-2-vinylpyr-
technique, was originally described by Stevie et al. in 1995 and is shown idine) and the spherical morphology of poly(2-vinylpyridine-block-
in Fig. 5 [52–56]. Although not specific to polymers, this technique is styrene-block-2-vinylpyridine) are clearly apparent [55]. As this ap-
nonetheless important as it demonstrates the application of a focused proach can significantly damage soft materials due to the direct impact
ion beam. Thinning of the sample is performed before FIB milling to of the ion beam on the sample, we hypothesize that staining the sample
minimize the time under the ion beam [41]. Tungsten, or a similar with iodine and the high glass transition temperature (near 100 °C for
metal, is deposited around the region of interest to mark and protect the both blocks) were crucial to stabilize the microstructure during FIB
sample. The FIB then thins the sample to a uniform, sub-micron milling. Thus, to allow for general applicability to beam-sensitive soft
thickness, alternating on either side of the sample to reduce re- materials, minimization of damage due to the ion beam is warranted, as
depositing material onto the specimen. This technique cuts a “trench” discussed below.
in the specimen about 5 to 20 microns in size. Because of the high New advances in the field of FIB milling will enable sample pre-
control of sample location within the specimen bulk and low damage, paration of polymeric materials that are sensitive to the ion beam.
multiple TEM samples can be prepared from the same piece of material Traditional FIB milling allows for the ion beam to directly impact the
[57]. sample and can lead to thermally-induced damage or damage from
Other FIB preparation techniques rely on extracting the region of impregnation of ions (e.g. Ga ions). One approach that aims to over-
interest from the sample, such as in the in-situ lift-out (INLO) and the ex- come these limitations is shadow-FIB milling (Fig. 7), a relatively new
situ lift-out (EXLO) approaches. The INLO technique is used to create a technique to prepare freestanding and electron-transparent samples
freestanding film (Fig. 6), which can then be positioned onto a copper [62–64]. This technique relies on milling the sample from the back, in
TEM grid for sample analysis and experimentation [44,51,55,58]. This order to protect the sample of interest. A shard or thin region of a film
process begins by depositing a single-micron scale thick metal line to or bulk sample (on a grid or electron-transparent silicon nitride
mark the region of interest for milling and to protect the sample from window) can be mounted onto a support and loaded into a FIB chamber
ion beam damage and artefact introduction during the sample pre- upside down. The sample is then milled to the desired thickness using
paration process. Following this demarcation, the sample is FIB milled relatively low FIB currents from the back of the sample. Using a mi-
to 10–20 microns in thickness and then extracted from the bulk sample cromanipulator tip, the milled sample, including the grid or silicon
and mounted on a TEM-compatible sample grid [59]. From the grid, the nitride support structure, is moved to the testing chamber. To remove
sample can then be further FIB milled to electron transparency or the the silicon support and create the electron-transparent sample, electron-
desired thickness. beam-assisted etching is used in a xenon difluoride precursor gas to
The EXLO technique begins the same as INLO, depositing a metal prevent deposition of silicon onto the sample. This entire procedure is
protection or guide. Rather than milling the sides of the samples, the done in the shadow geometry, such that the ion beam never directly
front and the rear of the sample are milled away, generally using a impacts the sample, thereby lessening damage and minimizing artefacts
“stair step” milling procedure [41,42,58,60,61]. The sample is then due to ion implantation [63,64]. This approach has been demonstrated
progressively milled to single-micron thickness, and then tilted to at to lead to minimal damage of block copolymers, and consequently has
least 45° to cut away almost all remaining attached material holding the been used to image block copolymer films [46,49,64–66]; an example
sample area to the rest of the bulk. The sample is then returned to the of a hexagonally packed cylindrical morphology of poly(styrene-block-

5
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 6. In-situ lift out for TEM sample preparation. Images taken during the FIB milling process; the region of interest is initially selected in (a), as outlined by the
white rectangle (scale bar, 5 μm). Using a focused ion beam (FIB) trenches are cut around the previously determined region of interest (b) (scale bar, 10 μm). The
wedged TEM sample is extracted from the specimen using a micromanipulator tip (c) (scale bar, 10 μm). The sample is then attached to a copper mesh TEM grid (d)
(scale bar, 5 μm). Reprinted by permission from [54]. TEM images of poly(styrene-b-2-vinylpyridine) are shown in (e) and (f); P2VP is shown in black in the
micrographs. Reprinted from [55], with permission from Elsevier.

isoprene) copolymer film is shown in Fig. 7e. include backscattering detectors or energy dispersive X-ray spectro-
Given that thermal damage is one of the main challenges in ap- scopy detectors, alternating with FIB milling in between every image or
plying the FIB to polymeric or organic materials, the use of a cryogenic elemental map. These milled layers are on the nanometer-scale thick-
stage is transformative for sample preparation (Fig. 8); the cryo-FIB ness, and over the course of hours, around 1000 μm3 can be re-
minimizes damage while maintaining the precise control that comes constructed [61]. Care must be taken as to not redeposit eroded ma-
with FIB milling [43,45,67,68]. As such, the shadow-FIB, EXLO and terial on the sample surface, and drift correction routines must be run to
INLO FIB milling, and H-bar milling can be combined with cryogenic minimize the consequences of ion beam drift.
temperatures to examine a wide array of polymer and biological sam-
ples for TEM, such as cells and cellular structures, polymer films, both
highly crystalline and amorphous, organic semiconductors, and organo- 3.2. Oscillating diamond knives to improve ultramicrotomy
metallic compounds [49,69–73]. This is discussed in more detail in
Section 3.3. A long-standing approach to prepare polymer samples for the TEM
One advantage of milling material away with nanoscale precision is is the ultramicrotome, a tool that can cut very thin slices of a material.
the possibility of creating 3D reconstructions by imaging during the Ultramicrotomes can cut electron-transparent sections, which are on
milling process [44]. Using a dual-beam instrument, slight milling and the order of tens of nanometers to a micron [74,76,77]. Sharp knives,
polishing by the ion beam alternates with imaging by the electron often made from diamond, tungsten carbide, or glass, are crucial to
beam; this technique can be referred to as serial sectioning achieve such thin sections. The ultramicrotome often relies on using
[44,61,74,75]. Once the sample is aligned at the eucentric point, such these sharp knives at an angle of approximately 35–45 degrees to create
that the ion beam and the electron beam interact with the sample at the a small crack in the sample, and then propagating the crack through the
same point, the ion beam is tilted to offset it from the electron beam. material to create a thin section [78]. As discussed in previous reviews,
Images are then gathered using the electron beam system, which can ultramicrotomy can introduce surface roughness or lead to compression
of the sample due to the sectioning process [79,80]. Often, optimization

Fig. 7. Shadow-FIB milling. SEM images of a block copolymer (PS-b-PI) (a–c) during shadow FIB preparation and (d) after sample is prepared for TEM. (e) TEM
micrograph of the section. Reprinted from [46], with permission from Elsevier.

6
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 8. TEM micrographs of PS-b-PI block co-


polymer prepared in a FIB microscope at
cryogenic conditions (−177 °C) and cryo-
transferred to the TEM. (a) High magnification
TEM image with overlaid computer simulated
gyroidal structure. (b) TEM image of two dif-
ferent regions with different thicknesses with
their computer-generated projections as insets.
Reprinted with permission from [73]. Copy-
right 2005 American Chemical Society.

of the cutting speed and cutting temperature, in addition to ensuring Samples sectioned via ultramicrotomy are less uniform in thickness
knife blades are sharp and free of defects, is critical to produce thin than those milled with the FIB [43]; often, knife defects are visible in
sections suitable for TEM experiments. Cryogenic ultramicrotomy, for sections. Furthermore, it can be easier to control the thickness of sec-
example, is often needed to minimize deformation and enable sec- tions with FIB milling. Delamination of layers or particles within the
tioning of soft samples where the glass transition temperature is near or samples from the cutting process and applied mechanical stress can
below room temperature, even if only one component in a multi- introduce artefacts [77,79,87]. Also, ultramicrotomy leads to approxi-
component system is soft, rubbery, or liquid-like [79,81–83]. For in- mately millimeter site-selectivity, as opposed to near nanometer pre-
stance, cryogenic ultramicrotomy followed by cryo-transfer to the TEM cision of FIB sample preparation [43]. A comparison of sections pro-
enabled the imaging of poly(methyl acrylate) (PMMA)-grafted-silica duced by the FIB and ultramicrotomy can be seen in Fig. 10.
nanoparticles in poly(ethylene oxide) (PEO) [84]. Nevertheless, there are also multiple advantages of ultramicrotomy.
One approach to minimize sample compression is to oscillate the Ultramicrotomy typically requires less time to produce sections when
diamond knife (Fig. 9). A diamond knife is attached to a low-voltage compared to the FIB; production of samples for the TEM can be tens of
piezoelectric translator operating at frequencies of 20–25 kHz [77,79]. minutes with the oscillating diamond knife [43,51,79]. Furthermore,
Samples can be frozen prior to sectioning if needed [77,85,86]. Cutting the lack of exposure to an ion or electron beam during the sectioning
speeds are generally on the order of 0.1–1 mm/s [43]. After cutting, the process removes this often limiting cause of sample damage in the FIB.
thin sections can either be floated on water and then transferred to a At the current state-of-the-art for both techniques, thinner samples are
TEM grid, or transferred to the grid directly. With this approach, achievable using ultramicrotomy when compared to FIB milling
electron-transparent samples with less compression or added surface [79,89].
roughness are produced, where the improvement in sample quality has
been attributed to reducing the cutting angle of the knife with respect to 3.3. Vitrification of liquid samples at cryogenic temperatures
the sample to approximately 35 ° [43,76,77,79,87,88]. The oscillating
diamond knife has been demonstrated on a variety of soft materials, Despite the advantages of the above techniques, freezing many
such as organic semiconductors for electronics, polymer membranes, polymer liquids or solutions prior to sectioning through the FIB or ul-
nanopores, and living cells. tramicrotomy leads to damage due to crystallization of the solvent or
Compared to the FIB, ultramicrotomy has various disadvantages. material itself [38]. Alternatively, liquids can be coated on a grid as a

Fig. 9. Comparison of biological samples


(HM20-embedded dinoflagellates) using (a)
conventional sectioning and (b) oscillating
diamond knife sectioning. Compression of the
sample using conventional sectioning is 22.5%,
while compression using the oscillating dia-
mond knife is 7.5%. Scale bar: 100 μm.
Reprinted, by permission, from [79]. Copyright
2000 John Wiley & Sons, Inc.

7
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 10. A comparison of TEM and STEM


images with different sample preparation
techniques: (a, c) FIB and (b, d) ultra-
microtomy. Although sections produced by the
FIB are more uniform, those made by ultra-
microtomy can be slightly thinner.
Additionally, no evidence of mechanical stress
is seen in (a) and (c). Reprinted from [43], with
permission from Elsevier.

thin film and the sample can be quickly vitrified to prevent crystal- used primarily for biological tissue fixation, but can also be used for
lization, after which the frozen section (and grid) can be transferred, polymers and other soft materials [90,93]. In comparison to high-
without allowing the sample to warm up, into a TEM equipped with a pressure freezing, slam-freezing better preserves the integrity and
cryogenic stage. This approach is often termed cryo-TEM. Operating at structure of layers in samples, is simpler in terms of equipment, and can
cryogenic temperatures can also reduce the loss of volatile components make larger area samples. Nevertheless, forcefully slamming samples
from the sample and minimize damage of the sample from the electron against frozen copper blocks can introduce some strain (flattening) and
beam [90–93]. Although many informative reviews of cryo-TEM can be distortion during the sample preparation process.
found in the literature [94–104], we briefly highlight five vitrification Plunge freezing instead relies on smaller amounts of sample to
techniques, high-pressure freezing, slam-freezing, plunge freezing, self- create vitrified films of liquids on TEM grids [110,111]. This procedure
pressurized rapid freezing, and cryogenic-FIB. relies on a liquid solution being first dropped onto a copper grid and
Perhaps the oldest of these cryo-TEM techniques is high-pressure then plunged into a cryogen, usually liquid ethane or propane to ensure
freezing as introduced by Moor and Riehle in 1968 [105,106]. High- fast heat transfer (ethane and propane have higher heat capacities than
pressure freezing generally occurs on the order of a few thousand bar to nitrogen or helium, and are less able to form an insulating vapor layer)
suppress crystallization of solvents such as water [107,108]. The pro- [112]. The simplicity of this approach, coupled with automatic tools to
cess begins with the creation of a sample solution that is dropped onto ensure the proper amount of liquid and electron-transparent films, have
an aluminum husk immersed in n-hexadecane; the hexadecane acts as a led to cryo-TEM imaging of a variety of samples, such as small mac-
pressure transfer medium [108]. Once this frozen drop of sample is romolecules, proteins, and whole cells [113], including 3D re-
within its aluminum half-shell, another half-shell of the same size is constructions [114,115]. Cryo-TEM has also been successfully applied
placed over the first, completely encasing the sample drop. This full- to a range of polymer assemblies, such as toroidal micelles, cylindrical
shell is then transferred into liquid nitrogen at ambient pressure for micelles, and vesicles (Fig. 11). Disadvantages of this technique are
storage. The use of high pressure within the aluminum full-shell greatly associated with creating thin-films of liquid suitable for TEM experi-
reduces the nucleation rate of ice and its subsequent crystal growth, ments, which can induce artefacts due to the water-air interface in-
allowing for samples up to hundreds of microns to be frozen and studied cluding unwanted alignment of particles or molecules, as well as from
[70,92]. This technique can preserve samples in native-like states; due the moderate level of control of the freezing process that can lead to the
to the high pressures, however, there is ample opportunity to introduce formation of ice or sample heterogeneity.
artefacts into the sample [90,108]. A newer cryo-TEM technique is self-pressurized rapid freezing
Another commonly used technique for making cryo-TEM samples is (SPRF), first published by Leunissen in 2009 [117]. Building upon the
slam-freezing, such as described by Escaig in 1982 [109]. This tech- theory behind high-pressure freezing, SPRF takes advantage of the
nique is simpler compared to high pressure freezing, but operates using adaptability of high-pressure freezing but adds the advantages of ease
similar conditions; first, a drop of solution is deposited on a copper of sample preparation [70]. Liquid samples are loaded into a copper
substrate and allowed to stabilize at room temperature but under humid capillary tube and the tubes are then sealed at both ends and submerged
conditions to prevent solvent evaporation [91,107]. These discs are into liquid nitrogen at −210 °C. As the system is kept at constant vo-
attached to a Teflon substrate and then projected at a liquid-helium lume, the increased volume of the solvent freezing (when water is the
cooled 10 K copper block, thus completing the “slam-freezing” process. solvent) will drive an increase of pressure within the capillary, thereby
The frozen sample is then stored in liquid nitrogen. Slam-freezing is achieving high-pressure during the freezing process. This sample can

8
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 11. Cryo-TEM of different types of vi-


trified polymer assemblies. (A) Vitrified fresh
solution of poly(acrylic acid-b-methyl acrylate-
b-styrene) triblock copolymer micelles in 1:2
by volume tetrahydrofuran (THF) to water
showing mixture of cylindrical and spherical
micelles and a small amount of toroids. (B)
Same solution as (A) after evaporation of THF
showing predominantly toroidal micelles.
Scale bar: 100 nm. From [4]. Reprinted with
permission from AAAS. (C) Poly(acrylic acid)-
b-polystyrene cylindrical micelles. Scale bar:
100 nm. Reprinted with permission from
[116]. (D) Polyethyleneoxide-poly-
ethylethylene amphiphilic diblock copolymer
vesicles. Scale bar: 20 nm. From [3]. Reprinted
with permission from AAAS.

then be stored in liquid nitrogen for further sectioning and imaging damage include knock-on damage, radiolysis, and local heating [27].
[117,118]. Nevertheless, previous studies on the beam damage of organic materials
As mentioned in Section 3.1, operating the FIB with a cryogenic have different theories regarding how damage occurs in these beam
stage allows for milling of soft, hydrated or solvated materials. This sensitive materials and suggest that damage depends on many different
technique was first demonstrated on biological samples in 2003 and factors besides the total electron dose, such as dose rate (flux of electron
polymers in 2005, and has since been greatly expanded for both beam) [124,125], probe size [126], temperature [127], and accel-
[73,119]. The sample is created in a way similar to traditional FIB erating voltage [128]. Damage has been examined extensively by
milling, taking place in a vacuum, but with a cryogenic temperature measuring changes in diffraction intensities of crystalline materials
stage that is cooled once the sample is placed in the chamber, or a [125,129–131], but changes in thickness (mass loss) and image contrast
cryogenic-transfer stage that can insert cold samples through a load- have also been quantified [124,127,131,132]. More recently, spectro-
lock stage (similarly to cryo-TEM stages). Samples are prepared through meters provide an opportunity to track damage to chemical bonding,
the same techniques as described in Section 3.1, and once they are either by identifying changes in core-loss spectra associated with spe-
thinned to electron transparency they can be either warmed to room cific bonding orbitals or in the low-loss spectra that is associated with
temperature or transferred cold to the TEM. Cryo-FIB milling reduces the valence electronic structure. Coupled with the development of
thermal damage, local melting, or devitrification that can hinder tra- monochromators, the combination of low-loss EELS and diffraction
ditional FIB sample preparation of soft (or liquid-like) samples experiments will likely add new insights by revealing how chemistry
[67,73,90,120]. As a consequence, tomography of cryogenic samples in and structure change with radiation damage.
the FIB through serial sectioning can be performed as well [121,122].
For example, cryo-FIB cross sections of vitrified liquid electrolyte were
4.1. Mechanisms (knock-on, radiolysis, local heating)
recently used to reconstruct the 3D structure of dendrites [123]. Similar
to how damage from an ion beam is a limitation during FIB sample
This review will not discuss origins in detail, but will provide a brief
preparation, electron beam damage of polymers in the microscope also
explanation of the aforementioned primary beam damage processes
limits resolution and is discussed in the next section.
that can affect polymer samples (knock-on damage, radiolysis, and local
heating). We refer the reader to other excellent resources for more in-
4. New insights on beam damage from low-loss EELS depth discussions [131,133–139].
Knock-on damage is the displacement of atoms from the crystal
Despite the clear impact of beam damage on soft materials in the lattice and occurs when the energy of an incident electron is above
TEM, it is currently not fully understood. In general, mechanisms for some threshold value that scales with atomic number (about 25–80 keV

9
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

for carbon, depends on the bonding type) [135,140]. This can generate Table 1
point defects within the lattice or cause sputtering, the ejection of Summary of reported critical doses for polymers.
atoms from the surface. Low density (soft) materials, primarily made up Polymer Dc (e−/nm2) a
Signal monitored Reference
of low atomic number elements, are particularly susceptible to this
damage mechanism. PE 2031 (2000 kV) Diffraction pattern [145]
268 (100 kV) Diffraction pattern [130]
Radiolysis occurs when inelastic scattering deposits enough energy
749 (125 kV) Diffraction pattern [130]
to break chemical bonds. Polymers are especially susceptible to radi- 936 (500 kV) Diffraction pattern [130]
olysis; the high inelastic scattering cross section can lead to chemical 1498 (1000 kV) Diffraction pattern [130]
changes. Furthermore, because most polymers are electrical insulators, 4484 (2000 kV,100 K) Diffraction pattern [145]
accumulated charge cannot dissipate. Radiolysis damage can be loca- PEO 100 (200 kV) Diffraction pattern [130]
POM 624 (100 kV) Diffraction pattern [130]
lized in either the polymer backbone, leading to mass reduction, or side
Nylon-6 1248 (100 kV) Diffraction pattern [130]
groups, often leading to charged species, free radicals, and cross- PEEK film 2809 (100 kV) Diffraction pattern [130]
linking. PET 1450 (100 kV, 100 K) Diffraction pattern [146]
Another common source of radiation damage in polymers is local 6200 (200 kV) Low-loss EELS (5-10 eV) [147]
PMMA 620 (120 kV, 77 K) EELS (C K-edge) [148]
heating caused by phonons (lattice vibrations). Polymers are thermally
3745 (80 kV) EELS (O K-edge) [149]
insulating, with thermal conductivities of about 0.1-0.5 W/mK, and will 31,208 (80 kV) EELS (C K-edge) [149]
experience a more significant temperature rise than ceramics PS 62,400 (120 kV, 127 K) Low-loss EELS (7 eV) [150]
(1–200 W/mK) or metals (10–400 W/mK) [141]. In addition to thermal iPS 1123 (120 kV, TEPD) b Diffraction pattern [151]
degradation, if heated above the glass transition temperature (Tg), PVP 8000 (200 kV) Low-loss EELS (˜10-40 eV) [152]
PBT 936 (100 kV) Diffraction pattern [153]
chains will start to flow and the structure of the sample will change.
PCTFE 599 (100 kV) EELS (Cl edge) [154]
Fig. 12 illustrates the relative time scales of the different beam da- PC 31,208 (80 kV) EELS (O edge) [149]
mage mechanisms. Electronic processes occur on the order of femto- Collodion 125 (80 kV) EELS (N edge) [149]
seconds and chemical reactions occur on the order of nanoseconds. In 374 (80 kV) EELS (O edge) [149]
insulating materials, radiolysis is typically 103 – 106 times faster than 3745 (80 kV) EELS (C edge) [149]
Formvar 1872 (80 kV) EELS (O edge) [149]
knock-on damage [138]. Molecular vibrations, which lead to local RR P3HT 3200 (80 kV) Low-loss EELS (˜ 2-4 eV) [18]
heating, occur on the femtosecond timescale [142]. Assuming a diffu- 33,000 (80 kV) Diffraction pattern [18]
sion coefficient of 106 cm2/s (similar to that of organic molecules in 10,800 (300 kV,100 K) Diffraction pattern [155]
butyl rubber), molecules take about 10 ns to diffuse 1 nm (about the RRa P3HT 2800 (80 kV) Low-loss EELS (˜ 2-4 eV) [18]
PGeBTBT 4000 (80 kV) Low-loss EELS (˜ 1-3 eV) [18]
lattice spacing of polymer crystals). Thus, although knock-on damage
can disrupt the lattice immediately, damage from radiolysis often relies PE: polyethylene, PEO: poly(ethylene oxide), POM: polyoxymethylene, PEEK:
on solid-state diffusion prior to apparent disruption of the structure. polyether ether ketone, PET: polyethylene terephthalate, PMMA: poly(methyl
methacrylate), PS: polystyrene, iPS: isotactic polystyrene, PVP: poly(vinyl
4.2. Summary of reported critical doses for polymers pyrrolidone), PBT: poly(butylene terephthalate), PCTFE: poly(chlorotri-
fluoroethylene), PC: polycarbonate, RR P3HT: regioregular poly(3-hex-
ylthiophene-2,5-diyl), RRa P3HT: regiorandom poly(3-hexylthiophene-2,5-
Accurate experiments require the effects of beam damage to be
diyl), PGeBTBT: poly[(4,4′-bis(2- ethylhexyl)dithieno[3,2- b :2′,3′- d]germole)-
carefully monitored. This is necessary to ensure the data accurately
2,6-diyl-alt-(2,1,3- benzothiadiazole)-4,7-diyl].
describe the properties of the sample and are not a result of the imaging a
Operating voltage listed in parentheses. All measurements performed at
procedure. Measuring and reporting a critical dose for damage is ne- room temperature unless otherwise stated.
cessary for accurate imaging of polymer morphology. To ensure the b
TEPD = total end point dose.
image is not of a damaged sample, it is imperative the critical dose not
be exceeded during imaging. Historically, beam damage in polymers D
was monitored mostly by the loss of a diffraction peak, such that beam I = Aexp ⎛− ⎞ + Ib
⎜ ⎟

⎝ Dc ⎠ (2)
damage has been mostly quantified for semicrystalline polymers.
For uncorrelated damage events, the signal corresponding to the A is an exponential prefactor and Ib is the background intensity
presence of pristine material (e.g., diffraction peak intensities) decays [143]. Thus, the critical dose can be calculated by monitoring the peak
exponentially with accumulated dose. The critical dose (Dc), typically intensity in the diffraction pattern (I) as a function of electron dose (D)
reported as electrons/area, is then the inverse of the exponential decay [144]. Dc is material-specific and we propose that it can even vary
constant and is taken as the radiation dose above which the material within the same material depending on the type of damage measured
will be significantly changed (intensity of measured signal, such as (damage to crystal vs damage to the chemical structure).
diffraction peak intensity, drops to 1/e of its initial value), as shown in Low-loss EELS provides an alternative method to quantify beam
Eq. (2). damage. The critical dose can be determined by monitoring the decay
of low-loss EELS peaks as a function of dose. EELS spectra will quantify
damage of the electronic structure whereas the diffraction pattern will
quantify loss of crystallinity due to beam damage. Importantly, with
EELS, beam damage of amorphous polymers can be monitored.
A summary of published critical doses (Dc) is reported in Table 1.
The values in Table 1 were obtained by various groups over several
decades. One must exercise caution when interpreting the presented
values, as Dc can vary significantly with experimental conditions in-
cluding accelerating voltage, temperature, and dose rate (the latter is
often not reported). Furthermore, the electron dose required to com-
pletely destroy the signal from a diffraction pattern or EELS peak, or
total end point dose (TEPD), may be used to quantify beam damage
Fig. 12. Different beam damage mechanisms occur on different timescales. In instead of Dc. Nevertheless, one trend evident in the table is that
insulators, radiolysis is typically the most significant. polymers with increased unsaturation appear to be more resistant to

10
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

damage from the electron beam. The Dc values measured at ambient decreased, another peak at ˜5 eV appeared (Fig. 13b). This peak was
temperature with a 100 kV accelerating voltage for PE, POM, PBT, assigned to degraded phenyl rings.
Nylon-6, and PEEK are 268, 624, 936, 1248, and 2809 e/nm2, respec- When the carbon K-edge of the EELS spectrum is measured beyond
tively. PE and POM are both completely saturated polymers and are the critical dose, a broadening of the C=C 1s to π* transition is ob-
reported to have lower Dc values (on the order of 100 s of e/nm2). served while the peak corresponding to the CeH 1s to σ* transition does
Nylon-6 is a polyamide which incorporates a carbonyl group within its not change. Based on the damage being manifested in signals corre-
molecular structure, and PBT and PEEK both incorporate phenyl rings sponding to the phenyl rings in both the EELS and low-loss EELS
within their molecular structures. Nylon-6, PBT, and PEEK have rela- spectra, a crosslinking mechanism is suspected.
tively higher Dc values than PE and POM. In general, polymers with Siangchaew et al. performed similar low-loss EELS beam damage
more degrees of unsaturation, and especially polymers with aromatic experiments in STEM mode to examine the 7 eV peak in PS films [126].
moieties, may be more radiation resistant. Due to enhanced delocali- Unexpectedly, they noticed that as the size of the probe decreased and
zation, these polymers may be able to distribute energy deposited by dose rate increased, the low-loss peak was significantly less susceptible
the electron beam more efficiently. As highlighted above, tabulated to beam damage. They proposed that fast secondary electrons emitted
values can still be useful in extracting general trends, even though roughly perpendicular to the incident beam travel laterally within the
precise characterization of beam damage is sensitive to experimental sample. The seemingly high resistance to beam damage was because the
conditions. fast secondary electrons would travel outside the probe area and thus,
not be monitored by the technique. To support this idea, they per-
formed scans at systematic distances away from the initial probe site.
4.3. Monochromated low-loss EELS beam damage experiments They found that at 80 nm away from the site, each pixel scanned pro-
duced the same peak intensity, suggesting the probed material had not
EELS spectral imaging has significant potential to investigate sys- been previously damaged. When the distance from the initial probe site
tems not easily measured by conventional TEM, particularly polymers. is decreased to 5 nm, the intensity of the 7 eV peak was significantly
Because polymer degradation is the main limitation of EELS, under- reduced, suggesting that the probed material had been previously da-
standing radiation damage is very important maged. Thus, when the probe is small, damage is delocalized beyond
[18,126,135,147,148,150,156–159]. Unlike when beam damage is the sample region being probed by the beam, leading to lower apparent
monitored using the diffraction pattern, low-loss EELS data can reveal radiation damage.
useful information regarding the breakdown of the chemical and elec- In a more recent study, Egerton et al. supported their theory using
tronic structure. Monte-Carlo simulations of fast secondary electron trajectories, al-
Beam damage in the low-loss region was first measured by though the theoretical calculations also indicate that fast secondary
Siangchaew and Libera using low-loss EELS to examine the decay of the electrons cannot be the only explanation for the high resistance to da-
π-π* transition in polystyrene using a field emission source with a full mage that is apparent with small probe sizes [156]. To achieve the
width at half max for the zero-loss peak of 1 eV (Fig. 13) [160,161]. observed beam damage resistance, fast secondary electrons would need
Degradation of the π-π* transition is attributed to chemical effects to travel at least 500 nm away from the probe. Fast secondary electrons
caused by differences in bonding and conjugation as the material is have a mean transport range on the order of 10 nm in organic solids.
damaged. The authors conclude that radiation damage studies provide Thus, fast secondary electrons cannot be the only factor contributing to
important information that defines appropriate conditions for imaging the high resistance of PS. It is suggested that delocalization of in-
polymeric samples. Furthermore, the authors propose that the scientific elastically scattered electrons may also contribute. Thus, the me-
community would benefit if more EELS studies of polymers included chanism for the large Dc with small probes and radiation damage away
quantitative assessment of radiation stability specifically through the from the probe site remains unresolved; we speculate that diffusion of
creation of a database focused on spectral fingerprints and radiation reacting species away from the probe may also contribute to the large
stability of polymers. Since these reports, beam damage in polystyrene Dc value observed.
has been studied extensively using low-loss EELS; it is convenient to Conjugated polymers, often composed of conjugated backbones and
study the peak at an energy-loss of 7 eV, corresponding to the π- π* solubilizing alkyl side chains, have also been examined using low-loss
transition of the phenyl ring. EELS. These materials make up an important class of polymers that can
Varlot et al. used EELS to assign the electronic transition energies at be used as the active layer in many organic electronic devices.
the carbon K-edge for the C]C 1s to π* and CeH 1s to σ* transitions Delocalized π-electron densities impart semiconducting properties, as
[150]. Using the peak in the low-loss region (7 eV), corresponding to well as absorption features in the low-loss region (1–10 eV). Thus, low-
the π- π* transition of the phenyl ring, the authors calculated a critical loss EELS imaging has the potential to significantly enhance imaging of
dose for polystyrene of about 62,000 e/nm2. As the π- π* peak intensity

Fig. 13. Examining beam damage in PS using


low-loss EELS as a function of electron dose.
(a) The exponential decay of the π- π* transi-
tion in polystyrene as a function of electron
dose. Reprinted with permission from [162].
(b) The low-loss spectrum of polystyrene for
different electron doses. As the dose increases
from a) 300 e/nm2 to b) 300,000 e/nm2 to c)
1,000,000 e/nm2 the peak at 7 eV energy-loss
decreases and a new peak at 5 eV energy loss
increases in intensity. Reprinted with permis-
sion from [150].

11
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 14. Quantifying beam damage in rr-P3HT


using monochromated low-loss EELS. (a) rr-
P3HT low-loss EELS spectra at various different
electron doses. (b) Integrated low-loss EELS
intensities as a function of electron dose. As
electron dose increases, the intensity of the
peak at 2.6 eV gradually decreases and shifts to
higher energy, plateauing at 3.2 eV. Reprinted
with permission from [18].

semiconducting polymers, but understanding beam damage will be 5.1. Beam damage minimization approaches
essential.
For materials with excitations near the visible region, the use of a Starting with the relationship between DC, SNR, C, and d from Eq.
monochromated electron source is crucial to minimize the background (1) and taking into account the detective quantum efficiency DQE (a
from the zero-loss peak. In a recent study, the critical dose of re- measure of the signal and noise performance of a detector), the smallest
gioregular poly(3-hexylthiophene-2,5-diyl) (P3HT) was quantified from resolvable feature size can be expressed as [156]
both π-π* transition peak intensities and diffraction intensities using −1
−1 SNR DC 2
monochromated 80 kV electrons with an energy spread of 0.15 eV [18]. d = (DQE ) 2 ⎛ ⎞
Because of the minimal energy dispersion of the probe, the authors C ⎝ e ⎠ (3)
were able to extract the electron absorption spectra near the bandgap at
Assuming DQE = 0.2 (typical for a CCD detector at the Nyquist
2 eV. Degradation of the electronic structure of P3HT as a decrease in
spatial-frequency limit), SNR = 5 (from the Rose criterion), and
peak intensity was measured, and a gradual shift to higher energy of the
C = 0.1 (typical for unstained polymers) as was done by Egerton et al.
peak absorption, from about 2.6 eV to 3.2 eV, was observed (Fig. 14). It
[156], we can estimate the dose required for several desired resolu-
was found that the low-loss EELS critical dose was 3200 e/nm2 com-
tions. These values can be found in Table 2 below. We observe that in
pared to the critical dose from the diffraction pattern of 33,000 e/nm2.
order to achieve the resolution required to resolve chains within crys-
Based on the large difference in magnitude of critical dose measure-
tals (d ˜ 0.4 nm) a dose of nearly 80,000 e/nm2 is needed, and to resolve
ments, the authors conclude that the valence electronic structure is
atomic positions (˜ 0.1 nm resolution) the required dose is 1,250,000 e/
damaged before the crystal structure. This is useful information for
nm2. Clearly, there is a need for minimizing beam damage in order to
measuring the morphology of semiconducting polymers; the optimum
achieve these high doses.
acquisition scheme can be determined depending on what character-
As discussed in Section 4, beam damage studies of organic materials
istics are being measured. Examples of taking advantage of the low-loss
to date have differing theories regarding how damage occurs in these
spectra for imaging are highlighted in Section 7.3.
beam sensitive materials. For example, EELS of the 7 eV π-π* peak of
In a similar study, the beam sensitivity of organic materials was
polystyrene reveals that π bonding is more stable at higher dose rates
measured with STEM-EELS and compared to data from variable angle
[164], but EELS experiments on thin films of collodion show first an
spectroscopic ellipsometry (VASE). Because STEM-EELS has high spa-
increase then decrease in stability with increasing dose rate [124].
tial resolution but causes damage while VASE has low resolution but
Diffraction experiments on P3HT also show an increase then decrease in
does not induce damage, comparison of the complex dielectric function
critical dose with increasing dose rate, which is explained by radiolysis
obtained from the two experiments allows for optimization of acquisi-
followed by the slow diffusion of a reacting species [125]. Conversely,
tion parameters for STEM-EELS [163].
another diffraction study of P3HT/PCBM suggests that there is no dose
Dose-dependent studies have revealed much about the mechanisms
rate dependence on beam damage [165]. Studies on the relationship
behind radiation damage in the TEM, but a complete picture remains
between beam sensitivity and voltage do not follow a clear trend either:
elusive. Often, lower accelerating voltages are useful to reduce knock-
beam sensitivity increases as voltage decreases near 100 kV, but the
on damage, although it remains unclear whether greater ionization
trend is reversed at very low voltages [128]. The effect of other vari-
damage will be a detriment to low-loss EELS measurements. Reducing
ables such as temperature and the presence of water or oxygen during
the probe size appears to significantly reduce the sensitivity to radiation
sample preparation have also been shown to affect stability [165].
dose by allowing damage to diffuse away from the probe, suggesting
Further beam damage studies would be valuable in working towards a
interesting opportunities with focused-probe experiments (e.g., analy-
unified theory for the damage mechanisms in polymers.
tical techniques such as STEM-EELS). Overall, measuring the limits of
Despite an incomplete understanding of how beam damage occurs,
dose tolerance and developing measurement approaches to minimize
some strategies for minimizing damage are known. It is helpful to know
damage is crucial to push the resolution limit, as discussed in the next
how knock-on damage, radiolysis, and local heating affect the polymer
section.
Table 2
Required dose in electrons per squared nm to achieve
5. Making HRTEM possible for polymers desired resolution calculated from Eq. (3).
d (nm) Required dose (e/nm2)
The resolution limit of organic materials is typically set by the dose
the sample can tolerate in the electron microscope. As such, details 1000 0.0125
matter, in the sense that any mitigation of radiation damage can result 100 1.25
in pushing the resolution limits beyond what is currently achievable. 10 125
1 12,500
Here we discuss how beam damage can be minimized, the evolution of
0.4 78,100
HRTEM of polymers thus far, and how aberration correctors and direct 0.1 1,250,000
electron detectors offer new opportunities for the future.

12
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 15. (a) Bright field TEM image of re-


gioregular P3HT grown by directional epitaxial
solidification. (b) Low dose HRTEM image. (c)
Schematic of P3HT chain packing along the b
and c axis. (d) Schematic of flat-on and edge-on
lamellae corresponding to HRTEM image.
Reprinted with permission from [168]. Copy-
right 2009 American Chemical Society.

being imaged, as it can help in choosing experimental parameters to rise to an exit wave ψx. The Fourier transform of ψx is ψk, where k is the
minimize damage. For example, because the scattering cross section can scattering vector. The role of the objective lens is to multiply ψk by a
Ze
be given by σ = πr 2 for r = Vθ where Z is the atomic number, e is the contrast transfer function T = sin χ(k), and then to inverse Fourier
charge of an electron, and V is the accelerating voltage, a sample that is transform ψkT to give the image wave ψi. Thus, sin χ(k) gives the phase
most susceptible to knock-on damage should be imaged with a lower changes of diffracted beams with respect to the direct beam as
accelerating voltage whereas a sample that is most susceptible to 1
radiolysis should be imaged with a higher accelerating voltage. In ad- χ (k ) = πλΔfk 2 + πCS λ3k 4
2 (4)
dition to choice of accelerating voltage, the common practice of using
cryogenic conditions for biological samples is also transferrable to where λ is the wavelength of the incident electron, k is the spatial
polymeric materials; a recent study demonstrated that cryogenic con- frequency, Δf is the defocus, and CS is the coefficient of spherical
ditions increased the beam stability of P3HT/PCBM [165]. Cryogenic aberration, which describes the quality of the objective lens. As a
conditions could also minimize the effects of local heating from the consequence, negative defocus values can be used to offset the positive
electron beam. As mentioned previously, beam damage studies as a spherical aberrations inherent to the lens. Nevertheless, sin χ(k) is
function of dose rate reveal that higher dose rates (smaller beam size) positive only for intermediate values of k, meaning that in this region
could also minimize damage. While this might be an artefact from all information is transferred with positive phase contrast and can be
under-sampling of damaged material, dose rate and beam size optimi- easily interpreted, but once the function crosses the k axis (this is the
zation could be applicable in STEM mode where rastering the probe point resolution) and begins to oscillate strongly, interpretation be-
quickly enough may be able to outrun damage. Additionally, sparse comes more convoluted. The implication here is that for an uncorrected
sampling STEM techniques coupled with reconstruction algorithms microscope, the resolution limit follows the aforementioned point re-
1
could be a useful approach for beam sensitive materials [166]. solution (Cs λ3) 4 (Section 2.1).
Although some combination of the aforementioned approaches may To overcome this resolution limit, non-rotationally symmetric lenses
enable imaging experiments at about 105 e/nm2, thereby resolving the can be used to produce negative aberrations that cancel out the positive
separation of polymer backbones in many materials, achieving doses of aberrations. Eq. (4) is a simplified description of the contrast transfer
106 e/nm2 and atomic resolution seems out of reach. As discussed in the function; in reality, it should contain higher order terms (C5, C7, etc.),
next section, perhaps the application of aberration-corrected micro- with each order corresponding to electrons scattered at higher fre-
scopes and direct electron detectors, in combination with exceptionally quencies. Thus, whereas an uncorrected microscope is C3-limited, a C3-
radiation-hardy soft materials, will enable near atomic resolution corrected microscope is now C5-limited, thereby allowing for higher
imaging. frequency information to be obtained [167]. Aberration correction also
reduces delocalization in the image, which minimizes the spread of
contrast from nonperiodic features in the specimen [31]. At the highest
5.2. Aberration corrected microscopes and direct electron detector for low frequencies, chromatic aberrations, rather than spherical aberrations,
dose HRTEM become the limiting factor.
The large lattice constant of polymer crystals facilitates imaging
We present a brief overview of the physics of image formation in the without aberration correctors. The first published polymer TEM image
TEM to motivate the advantage offered by aberration correctors, and to with lattice resolution was in 1969, when Bassett and Keller imaged the
highlight the potential opportunities for imaging of polymers. Consider 1.8 nm (100) spacings in beta-PPX using two-beam dark-field (DF)
the incident electron beam ψo and the electron potential function of the imaging [169]. Since then, many polymers have been imaged with
sample φ, such that as the beam interacts with the sample, it will give HRTEM, as listed in detail in previous reviews [170]. In 2009, the

13
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 16. (a) HRTEM image of rr-P3HT:PCBM


showing the (100) fringes of rr-P3HT crystals
and the rr-P3HT/PCBM phase boundaries (red
lines). (b) HRTEM of pure rr-P3HT showing
(100) fringes of lamellar crystals. Images were
taken on a Cs-corrected FEI Titan operating at
300 kV with a low-dose system. Reprinted with
permission from [19]. Copyright 2011 Amer-
ican Chemical Society. (For interpretation of
the references to color in this figure legend, the
reader is referred to the web version of this
article.)

semicrystalline structure of regioregular poly(3-hexylthiophene-2,5- Fig. 17, comparison of expected imaging results and experiments sug-
diyl) (rr-P3HT) thin films grown by directional epitaxial solidification gest that atomic-scale information is revealed. Thus, although earlier
were imaged with HRTEM by Brinkmann (Fig. 15) [168]. Crystalline work with a conventional electron microscope relied upon electron
order was aligned through directional epitaxy, and images were en- diffraction to reveal molecular information about polyvinylidene
hanced using a Fourier filter that emphasized frequencies corre- fluoride (PVDF) electrospun nanofibers [172], imaging with the CS-
sponding to the lattice spacing. While HRTEM has been accomplished corrected TEAM 0.5 revealed features that are interpreted to be rows of
on a number of polymer systems before the advent of aberration cor- CF2 groups [173]. The HRTEM image shown in Fig. 17 was acquired
rectors, aberration correctors have improved resolution and opened up with 2000 e/nm2 after exposing the sample to a dose of 170,000 e/nm2.
opportunities for imaging of polymers. Previous work with electron diffraction of chain folded crystals of PVDF
For example, Drummy et al. combined EFTEM with low dose had demonstrated a loss of diffraction spots at about 200 e/nm2 (ori-
HRTEM on a CS-corrected FEI Titan to image rr-P3HT crystals and their ginally reported as 38 C/m2) [174]. The authors of the study that
orientation within rr-P3HT-rich domains of rr-P3HT/PCBM blends produced the images in Fig. 17 suggest that their observed larger ra-
(Fig. 16) [171]. Distinct regions corresponding to rr-P3HT and PCBM diation tolerance is due to the presence of heavy metals (Pt, used to
crystallites are apparent without any image processing, in contrast to enhance contrast) in earlier work, because these heavy atoms could
previous work on HRTEM of rr-P3HT [168]. Although the improvement cause secondary radiation events that promote damage. Nevertheless,
may be due to a better detector, we speculate that the image quality is achieving near angstrom resolution with 2000 e/nm2 for a single image
also enhanced by extending the frequency at which the CTF crosses zero suggests that aberration-corrected microscopes are moving beyond the
using Cs correctors, given that finite defocus is needed to enhance phase limitations outlined in Table 2, likely due to a higher contrast and lower
contrast. This could effectively increase the signal-to-noise ratio. acceptable signal-to-noise ratio than our assumptions used in Eq. (3).
Recent work on a CS-corrected microscope has also imaged poly- The combination of direct electron detectors with aberration cor-
vinylidene fluoride (PVDF) fibers at high resolution. As shown in rectors has the potential to push resolution limits for polymers even
further. The advantage of direct electron detectors is that they have an
improved modulation transfer function (MTF), which describes the ef-
fect of an electron being detected as signal in multiple pixels, as well as
better detective quantum efficiency (DQE), which describes how the
detector affects the signal-to-noise ratio in an image [175]. Thus, the
application of direct electron detectors will reduce the dose required to
image at a given resolution when compared to traditional CCD detec-
tors. In addition to better MTF and DQE, it has also been demonstrated
that direct detectors operating in counting mode, as opposed to in-
tegrating mode, offer improved resolution as well. For example, Stach
et al. took advantage of the Gatan K2-IS direct electron detector oper-
ating in counting mode to achieve aberration corrected HRTEM images
of P3HT by taking multiple images that can be drift corrected prior to
summing (Fig. 18) [176].
Another technique with growing potential due to direct electron
detectors is known as scanning nanodiffraction, also called 4D STEM.
When the electron beam is focused into a probe, a convergent beam
electron diffraction (CBED) pattern is formed in the far field. This
pattern is rich in crystallographic and scattering data as well as in-
formation regarding thermal vibrations and energy losses and is parti-
cularly amenable to soft materials because of the low dose required for
a large amount of diffraction information. Traditional STEM imaging,
however, simply uses a monolithic detector such as an annular dark
Fig. 17. (A) HRTEM image of a segment of a thin nanofiber of PVDF compared field (ADF) or bright field (BF) detector, throwing away most of the
with (B) a simulation of PVDF. Images are overlaid by a stick model in which diffracted signal information. By placing a high-speed pixelated direct
carbon-carbon bonds are green, CH2 groups are white, and CF2 groups are red. electron detector in the far field, every pixel of the CBED pattern can be
Reprinted with permission from [173]. (For interpretation of the references to recorded at each probe location with millisecond dwell times, resulting
color in this figure legend, the reader is referred to the web version of this in a 2D CBED pattern (reciprocal space) at each point of a 2D STEM
article.) image (real space) — hence the name 4D STEM [177]. The recent

14
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

crucial parameters such as the contrast and signal-to-noise ratio and


assumes that a traditional CCD is used. We propose that revisiting the
resolution limits for soft materials in modern electron microscopes is
needed, and that new instrumentation likely enables near atomic re-
solution for many polymers.

5.3. Improved modeling software and computational tools

Image interpretation can often be a challenge for materials with


significant disorder, as is often the case for polymers. Fig. 17 highlights
one example of how computational tools can aid in analysis of HRTEM
images. Another example is a recently published software called GRATE
(GRaph based Analysis of TEM images) which converts HRTEM images
of polymer films into easily interpreted line drawings from which mi-
crostructural information can be extracted [182,183]. Software has also
been helpful for analyzing 3D reconstructions, where a discrete alge-
braic reconstruction technique (DART) [184–186] has been demon-
strated to be a useful tool in the interpretation of STEM-tomography tilt
series of P3HT/PCBM (Fig. 20) [187]. In a typical tomography re-
construction procedure, many gray-scale levels are allowed for every
voxel, such that eventual segmentation (the assignment of dark or
bright domains to different materials) is done manually. DART, on the
Fig. 18. HRTEM images of P3HT domains. Aligned summation of 20 images,
each acquired at 12,500 e/nm2. Images were taken on an aberration corrected other hand, automatically segments reconstructions through the use of
microscope using a Gatan K2-IS direct electron detector operating in counting constant gray-level assignations to each material. The addition of this
mode. constraint adds information to the reconstruction, and thereby im-
proves reconstructions for a given exposure or number of images in a
tilt series.
development of an electron microscope pixel array detector (EMPAD)
Besides software that is used for analysis post-data acquisition, re-
has allowed for ptychographic reconstructions capable of revealing
cent work on automating sample preparation, image acquisition, and
deep sub-angstrom resolution (0.39 Å) in 2D and beam-sensitive ma-
image analysis all at once has enabled high-throughput TEM capable of
terials [178]. Although various different types of 4D STEM experiments
generating phase diagrams of block copolymer amphiphiles [188]. In
can be done, including position-averaged convergent beam electron
addition to advancements in software, developments in instrumentation
diffraction (PACBED), virtual dark field imaging, fluctuation electron
and techniques are clearly warranted to minimize the effects of radia-
microscopy, strain measurements, and phase contrast imaging methods
tion damage in the TEM. The next section overviews some approaches
such as ptychography and MIDI STEM (see discussion in Section 7.2),
to address this central challenge in polymer electron microscopy.
4D STEM has only been demonstrated on polymers very recently. In
earlier studies, 4D STEM was used to map islands of P3HT in a PS
matrix [179], whereas more recent studies have used 4D STEM to 6. Low-dose techniques for polymer microscopy
create orientation maps of π-π stacking in a small molecule, 7,7′-(4,4-
bis(2-ethylhexyl)-4H-silolo[3,2-b:4,5-b′] dithiophene-2,6-diyl)bis(6- Although the beam sensitivity of soft materials appears prohibitive
fluoro-4-(5′-hexyl[2,2′-bithiophen]- 5-yl)benzo[c][1,2,5]-thiadiazole), for high resolution imaging as exemplified in Table 2, advances in in-
and a conjugated polymer, poly[2,5-bis(3-tetradecylthiophen-2-yl) struments and techniques aim to push the limits established by Glaeser.
thieno[3,2-b]thiophene] (PBTTT) (Fig. 19) [180]. Pair distribution This Section highlights a few examples of instrument and technique
function analysis of 4D STEM data has also been proposed as a method development that attempt to move beyond Eq. (1); another example,
for characterizing short and medium range order in aperiodically single-particle reconstructions, is described in Section 8.1.
packed organic molecules [181].
Aberration-correctors and direct electron detectors are pushing the 6.1. Monochromators for low-dose imaging
resolution in images of polymers, as shown by the examples highlighted
in this Section. Our simple calculations of the dose required for imaging Most modern microscopes are designed to maximize flux. As a
at specific length scales shown in Table 2 relies on rough estimates of consequence, controlling the dose rate to achieve low-dose imaging for
soft materials can be challenging. Monochromators significantly reduce

Fig. 19. π-π stacking orientation maps obtained from 4D STEM. (a) Orientation map of as-cast PBTTT thin film showing no mesoscale order. (b) Orientation map of
annealed PBTTT thin film showing enhanced ordering with domains several nanometers in size. Reprinted with permission from [180].

15
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 20. Volume reconstructions of as-cast and thermally-annealed blends of P3HT and the endohedral fullerene Lu3 N@C80-PCBM. Reproduced with permission
from [187].

Fig. 21. (A) High-dose (2000 e/Å2s) TEM


image of carbon nanothreads with FFT (inset)
showing amorphous structure despite (C) pre-
dicted hexagonal packing from X-ray diffrac-
tion data. (B) Low-dose (50 e/Å2s) TEM image
and (D) line profile showing thread-like fea-
tures that correlate with the predicted model.
Reprinted with permission from [189].

flux by filtering electrons from the gun based on energy, providing an [125,126] may be more applicable when slow, irreversible reactions
opportunity to control flux by increasing the energy resolution or even dominate.
by misaligning the monochromator with respect to the emission max-
imum. For example, Juhl et al. tuned the monochromator to lower the
6.2. Single shot dynamic TEM
dose rate and thereby achieved low-dose imaging conditions, which
was crucial to detect the hexagonally packed carbon nanothread
In a traditional TEM, only a single electron is traveling down the
structure that agreed with density functional theory predictions
microscope column at any given time (to put this into perspective, a
(Fig. 21) [189,190]. A similar strategy enabled HRTEM of highly-or-
200 keV electron travels at approximately 2/3 the speed of light). On
iented polyacetylene [191]. Simply by affording the operator more
the other hand, in dynamic TEM (DTEM), an ultrafast laser pulse illu-
time, we propose that controlling the electron flux is useful for imaging
minates a photocathode source that can photoemit a billion electrons in
of radiation-sensitive samples. Furthermore, dose rate may affect ra-
a single packet. For example, at the Pegasus facility at UCLA, long
diation damage [192]. While low dose rate techniques may minimize
electron pulses produced by an electron source will enter a linear ac-
damage in the case of reversible reactions induced by the electron
celerator whose electromagnetic fields compress the pulse several me-
beam, other studies that demonstrate an advantage of high dose rate
ters downstream into 10 fs pulses (Fig. 22) [193].

16
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 22. At the Pegasus facility at UCLA, an


electron pulse (green) passes through a linear
accelerator that compresses the pulse to below
10 fs. This duration time is so short that the
pulse can “outrun” all atomic motion in mole-
cules, thereby enabling an image to be captured
before damage occurs. (For interpretation of the
references to color in this figure legend, the
reader is referred to the web version of this ar-
ticle.)

At time scales this short, all atomic motion is essentially frozen. One the particles, the low-dose rate method avoids sample degradation.
of the appeals of ultrafast DTEM is thus the ability to make “molecular While the single recorded image with a dose of 10 e/Å2 has poor con-
movies” [194]. From a polymer microscopy standpoint, another ad- trast, the accumulated dose of the focus series comes out to 1360 e/Å2.
vance of DTEM is the ability to capture atomic images beyond the The images can then be aligned to obtain the complex exit wave
sample’s damage threshold. In other words, we can “diffract and de- function, creating an in-line hologram with high contrast and no sample
stroy” or “outrun damage” – electrons will interact with and pass degradation. This method has been used on many beam sensitive sys-
through the sample before damage propagates and affects the structure. tems, such as graphene [196], halide perovskites [197], Ziegler-Natta
catalysts [198], and gallium nitride [199]. The innovation of a re-
sonator capable of pulsing an electron beam with picosecond resolution
6.3. Low dose rate in-line holography has enabled HRTEM of the beam sensitive Ziegler-Natta catalyst mag-
nesium chloride; this study also suggests that phonons play a role in
While the “diffract and destroy” method is one solution to the ra- beam damage, such that by pulsing the electron beam at the time scale
diation damage problem, another strategy is to “divide and conquer”. In of phonon vibrations, the sample can “heal” between pulses [200]. By
other words, we can deliver electrons in such small quantities at once systematically increasing dose rates, in-line holography can also be
that the distortion of structure is negligible as long as the excitation is used to investigate dynamics.
reversible and decays before the next probing electron hits the mole- Whether through instrumental optimization or computational ana-
cule. This can be repeated over and over again, resulting in a large lyses, from Eq. (3) it is apparent that enhancing contrast in the TEM is
image series that when averaged, reveal a high resolution image that crucial to maximize resolution. Polymeric systems often contain ele-
remains faithful to the undamaged structure [195]. Such an approach ments other than carbon and hydrogen, creating opportunities through
uses a Nelsonian illumination scheme that bypasses the conventional analytical TEM. Furthermore, the inner potential of domains within
collimator setup and is able to deliver electrons in a controlled manner different polymers often varies, such that phase interference can be
exclusively to the imaged area. used to enhance imaging contrast. These strategies are discussed in the
This technique has been demonstrated for imaging gold particles next two Sections.
with minimal beam-induced damage (Fig. 23) [195]. Whereas the tra-
ditional method of imaging with high dose rates shows alterations in

Fig. 23. A comparison of high dose rate images of gold particles to the averaging of low dose rate acquisitions. The low dose rate approach simultaneously avoids
beam damage and boosts resolution. Reprinted with permission from [195].

17
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

7. From elemental mapping to mapping local electronic structure Recent work has highlighted the unique set of challenges for EFTEM
when applied to polymeric materials, which are largely due to the in-
Morphologies of multicomponent polymer systems are inherently herently low contrast and similar electron densities between domains.
challenging to examine using electron microscopy. Soft materials are An optimized procedure for elemental mapping of polymeric materials
often amorphous and have similar densities and elastic scattering cross- was demonstrated with the polymer/fullerene mixture poly(3-hex-
sections. Because of this, conventional TEM, where imaging is based on ylthiophene-2,5-diyl)/ [6,6]-phenyl-C61-butyric acid methyl ester
elastic scattering, is typically limited to samples that are chemically or (P3HT/PCBM) and emphasizes three factors: focusing at zero-loss with
physically modified with heavy stains or to systems with high contrast the aid of Fast Fourier Transforms (FFTs), using an objective aperture,
(such as organic-inorganic nanocomposites). and ensuring sufficient signal-to-noise and counts. Additionally, gen-
Spectroscopic imaging exploits contrast mechanisms that conven- erating a set of images that include a bright field image, a thickness
tional TEM cannot. Inelastic scattering of electrons by soft materials map, and an elemental map at the same region is useful for minimizing
generates contrast based on chemical and electronic structure. The misinterpretation of elemental maps [35].
development of monochromators that can reduce the energy spread of
the electron beam to 0.1 eV (see Section 2.2) as well as spectrometers 7.2. Recent elemental mapping and spectrum imaging examples
with detection resolution as low as 0.04 eV have facilitated the devel-
opment of electron energy loss spectroscopy (EELS) and low-loss EELS. Elemental mapping by energy-filtered TEM is a convenient tech-
As the electron passes through a sample, it can interact with core or nique to determine the structure of polymer systems with significant
valence shell electrons, causing it to lose energy. Electrons detected by compositional contrast [5,9,12,203–212]. This technique has been ap-
the spectrometer can be recorded as a function of energy-loss; inter- plied to image ions within polymer electrolytes [8,213], to generate
actions with core electrons result in high energy loss (100 s of eV) and elemental maps of solid-electrolyte interfaces [123], and to characterize
interactions with valence electrons result in low energy loss (1–10 s of nanoscale phase separation in mixtures of organic semiconductors
eV). Thus, at high energies, the EELS spectrum can be used to quantify [9,11,12,17,19,20,214]. Energy-filtered electron tomography has also
elemental composition and at low energies, low-loss EELS can be used revealed the 3D structure of conjugated polymer/fullerene mixtures
to characterize the electronic structure of the sample. Analogously, [215]. Most examples of elemental mapping require very high doses,
energy-filtered TEM (EFTEM) can be used for elemental or valence such that this approach relies on minimal atomic diffusion. As a con-
mapping. This can be done in TEM mode, where images are collected at sequence, EFTEM and elemental mapping have been demonstrated to
specific energy-losses, or in STEM mode, where an EELS spectrum is reveal the mesoscale structure of materials with domains larger than
collected at each pixel. about 5 nm.
The distribution of ions within a polymer matrix affects the per-
7.1. Strategies for optimized EFTEM of polymers formance of electrolytes used in lithium batteries [216,217]. EFTEM
was used to image bis(trifluoromethane)sulfonimide lithium salt
EFTEM for elemental mapping may be performed using the two (LiTFSI) within a poly(styrene-block-ethylene oxide) (PS-b-PEO) matrix.
window method or three window method [139,201]. At a minimum, LiTFSI ions are selectively imaged using F and Li elemental maps and
two images must be taken to allow for background subtraction. The two the PEO block is imaged using the O map (Fig. 24). While LiTFSI does
window method involves taking an image at an energy before (pre- contain O, the low ion concentration ensured the signal was dominated
edge) and after (post-edge) the ionization edge. The pre-edge image is by PEO; using oxygen maps to characterize the PEO regions agreed with
subtracted from the post-edge image. This method is insufficient for both bright-field TEM and small angle X-ray scattering (SAXS) data.
quantitative analysis [202]. EFTEM results demonstrate that the lithium salt preferentially ag-
The three window method enables quantitative elemental mapping; gregates in the center of the PEO lamellae as the PEO chain length
two pre-edge images are taken to generate an averaged image for increases, and this localization has been attributed to increases in ionic
background subtraction from the post-edge image. The resulting map is conductivity [8].
directly proportional to the total amount of the specific element, such Although elemental maps are quantitative in terms of the total
that variations in composition and thickness are convoluted. While the amount of a specific element, variations in thickness can confound
three window method will allow for images with intensities that are compositional information. A simple approach to account for variations
proportional to elemental content, taking three separate exposures to in film thickness was demonstrated using a mixture of a conjugated
create one elemental map can be costly in terms of beam damage to polymer and PCBM. By dividing elemental maps by thickness maps that
radiation-sensitive polymers. Nevertheless, taking multiple images can measure the thickness in units of the mean free path, the local com-
be useful to properly account for the background or overlapping re- position of polymer can be extracted. Mean free paths for the individual
sonances [139]. components were predicted from elemental compositions. As shown in
Moreover, the inelastic scattering cross section of electrons used for Fig. 25, this approach leads to maps of the volume fraction of polymer
elemental mapping, core-shell electrons, is relatively low; this ne- by essentially accounting for inelastic scattering at a given edge and the
cessitates high exposure times for sufficient intensity. The inelastic total amount of inelastic scattering at every pixel [9].
scattering cross section decreases with increasing energy-loss to the EFTEM can also generate images based on differences in plasmon
fourth power, such that the inelastic scattering cross section of low-loss resonances. An advantage lies in that the zero-loss and plasmon peaks
energies resonant with valence band transitions is greater than the io- typically dominate the scattering cross-section of a material, although
nization K-edge by several orders of magnitude [136,137]. generating contrast requires distinct plasmon peaks. Unfortunately,
Thus, one strategy to obtain high-contrast energy-filtered images for plasmon resonances in polymers are generally broad and overlapping
quantitative analysis while minimizing beam damage is to use low-loss [218]. Nevertheless, due to their high intensity and slight shifts in peak
energies for imaging. Contrast is generated by differences in valence position, plasmon resonances have been used to image various polymer
band electronic structure. Spectroscopic imaging has benefitted from systems [19,20,206,212–214,218–222].
the development of monochromators (Section 2.2). The energy re- Recent work demonstrates the value of using direct electron de-
solution associated with spectral imaging by EELS is limited by the tectors for EELS and spectrum imaging. The improved point spread
energy spread of the electron source, quantified by the FWHM of the function and pixel density of the direct electron detector provides both
zero-loss peak. Newer monochromators can achieve energy resolutions improved energy resolution and energy field of view, suggesting that
small enough to resolve the valence band structure of polymeric sam- direct electron detectors could facilitate elemental mapping [223]. In-
ples, as low as about 0.1 eV [136]. deed, direct detection has allowed for faster acquisition of energy-

18
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 24. Using elemental maps to quantify


domain spacing of a lithium salt distributed in
a PS-b-PEO matrix. (a) Chemical structures of
block copolymer and lithium salt. (b)
Elemental maps of F, Li, and O. Lithium ions
localize near the center of PEO lamellae. (c)
Lamellae thickness of lithium and fluorine
using EFTEM elemental maps. Lamellae thick-
ness of PEO quantified using O EFTEM ele-
mental map (O), bright-field TEM (BF), and
SAXS (d x fEO). Scale bar: 50 nm. Reprinted
with permission from [8]. Copyright 2009
American Chemical Society.

filtered images of labeled biological specimens, thereby overcoming the at higher energies was collected at higher doses. The resulting com-
problem of drift [224]. posite maps of P3HT and PCBM were generated using a principle
component analysis algorithm, and are representative of local varia-
tions in the electronic structure of the film (Fig. 27). Further work in
7.3. Monochromated low-loss EELS SI to highlight differences in valence STEM mode (e.g., STEM-EELS) could achieve higher energy resolution
electronic structure to resolve subtle features associated with vibronic states or inter-
molecular coupling. Overall, this work exemplifies the potential of low-
Monochromated low-loss EELS spectrum imaging (SI) uses elec- loss EELS spectrum imaging to take advantage of contrast arising from
tronic transitions as an additional source of contrast for samples with the characteristic valence band structure of semiconducting polymers
similar mass densities and elemental composition. In 2000, Varlot et al. and how advances in instrumentation can drive the development of
used low-loss EELS SI to characterize a triphase polymer composite of new techniques.
PS, PMA, and PB [209]. The PS phase was mapped using the low-loss π- Despite the tremendous potential for analytical TEM, most energy-
π* transition of the aromatic ring. A slit width of 5 eV, corresponding to filtered experiments require very high doses. Low-loss spectrum ima-
the limit of the instrument, was used. Using such a small slit width ging and EELS does have the potential to image at low doses because of
made obtaining sufficient signal-to-noise without severely damaging the higher inelastic scattering cross-section at low energy losses, as
the sample challenging. Nevertheless, contrast was significantly en- exemplified in Fig. 27. Nevertheless, developing alternative approaches
hanced relative to the zero-loss image as seen in Fig. 26. to enhance contrast is crucial for imaging polymers, as discussed in the
In 2015, Guo et al. used monochromated low-loss EELS to char- next section.
acterize P3HT blended with PCBM [18]. These materials exhibit ab-
sorption features at low loss (2–8 eV) that are a result of the conjugated
electronic structure. Advances in instrumentation enabled a slit width 8. Enhancing contrast in polymers with phase plates
of 1.5 eV for spectrum imaging, resulting in acquisition of images with
good signal-to-noise ratio at 1 eV increments. Experiments were de- In the past, one method of increasing contrast in polymeric systems
signed to collect data at energies corresponding to P3HT prior to has been to use heavy element stains such as osmium tetroxide, which
reaching the critical dose for this material, although the “background” reacts with unsaturated but non-aromatic carbon-carbon bonds, or

Fig. 25. Sulfur map and corresponding com-


position map of 1:3 poly[(4,4′-bis(2-ethyl-
hexyl)dithieno[3,2-b:2′,3′-d]germole)-2,6-
diyl-alt-(2,1,3-benzothiadiazole)-4,7-diyl]
(PGeBTBT) mixed with [6,6]-phenyl-C 71
-butyric acid methyl ester (PC71BM) annealed
at 160 °C for 20 min. The color legend re-
presents the scale for the volume fraction of
PGeBTBT. Reprinted with permission from [9].

19
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 26. Low-loss EELS SI of a triphase polymer


composite incorporating PS, PMA, and PB. (a)
Zero-loss image of elastically scattered elec-
trons. (b) PS is mapped using the low-loss
image at an energy-loss of 7 eV, corresponding
to the π to π* transition of the aromatic ring.
Scale bar in both images is 30 nm. Reprinted
from [209], with permission from Elsevier.

Fig. 27. Monochromated low-loss EELS spectrum imaging using the valence electronic structure of P3HT/PCBM blends. (a–d) image slices taken at 3, 4, 5, and 6 eV,
respectively. (e) P3HT and (f) PCBM phase maps generated using principal component analysis to deconvolute micrographs shown in (a–d). In image (e), P3HT fibers
are the bright areas and PCBM-rich domains are bright in image (f). (g) Quantitative composite image generated from phase maps in images (e) and (f). P3HT-rich
domains are red and PCBM-rich domains are green. Reprinted with permission from [18]. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)

ruthenium tetroxide, which reacts with most carbon-carbon double 8.1. Phase plate contrast enhancement in TEM mode
bonds, including aromatic rings. Several shortcomings exist in the
practice of polymer staining, such as the fact that contrast is qualitative In organic materials, the phase contrast ratio is often approximately
(the way in which the stain interacts with the polymer is often not 5%, making them weak phase objects. As described in Section 5.2, in-
completely characterized) and the fact that stains can introduce arte- teractions between the incident beam and the sample produce a phase
facts such as inorganic nanostructures [136]. A recently developed shift in the exit wave. Consequently, if the amplitude of the scattered
approach, in which the contrast in weak phase objects can be enhanced, wave is small compared to the unscattered wave, the image contrast
relies on phase plates. While the use of phase plates has been useful in a will be small. While phase contrast can be enhanced via spherical
number of organic and biological materials, such as in the aid of 3D aberrations under defocused conditions, this leads to oscillating con-
reconstructions of proteins [34], the use of phase plates to image trast at high spatial frequencies and difficulty in image interpretation.
polymers remains underexplored. This method also results in a loss of phase information, particularly at

20
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 28. TEM images of NR-CB with conventional TEM at focus (a), with conventional TEM at underfocus (b), and with a Zernike phase plate (c). Reprinted with
permission from [226]. Copyright 2005 American Chemical Society.

smaller frequencies. Nevertheless, defocusing to enhance contrast based high-angle annular dark field detector (collects mostly incoherently
on differences in the inner potential of two components of a block co- scattered electrons and leads to Z-contrast). In any of these cases, the
polymer has been demonstrated to reveal the microstructure [225]. STEM probe prevents interference between scattered and forward-
Here we highlight a few successful examples of polymer microscopy scattered electrons, thereby eliminating the limitations in resolution
using phase plates. For example, a Zernike phase plate was used to due to the contrast transfer function of the objective lens. But, this also
enhance contrast in natural rubber filled with carbon black (NR-CB) eliminates contrast enhancement due to defocus and aberrations of the
(Fig. 28) [226]. The phase plate consists of a thin carbon film with a objective lens. In theory, interference between scattered and forward-
small hole in the center placed at the back focal plane of the objective scattered electrons to create phase contrast is possible in STEM mode.
lens, such that the phase of the scattered beams is shifted by the carbon In order to achieve this, however, a very small detector is needed to
film while the unscattered beams pass through the center hole without collect a perfect parallel beam, leading to low signal-to-noise. The
being phase shifted. The result is that at the image plane, interference combination of spherical aberration correctors and phase plates holds
between the scattered and unscattered beam enhance phase contrast. promise in enabling phase-contrast STEM.
Another type of phase plate, the semicircular Hilbert phase plate, Ophus et al. demonstrated a new kind of phase-contrast electron
also shows promise for revealing structure in polymeric materials. microscopy called matched illumination and detector interferometry
Whereas the Zernike phase plate must be precisely aligned in order for (MIDI)-STEM. In a MIDI-STEM set up, a ring pattern phase plate with
its center hole to be on the optical axis, the semicircular Hilbert phase alternating thicknesses of Si3N4 placed at the probe-forming aperture
plate covers one half of the back focal plane such that the central beam generates a probe with a built-in reference wave; this probe can be
passes through an open area close to the edge of the phase plate and rastered across the sample as in traditional STEM imaging. The elec-
only electrons passing through the plate experience an additional phase trons scattered to high angles can then be collected by a standard ADF
shift. The lamellar structure of polystyrene-block-polyisoprene (PS-b-PI) detector while a pixelated direct electron detector records an image of
was clearly seen in an image taken with a Hilbert phase plate (Fig. 29A) the forward-scattered beam at each scanned position. The images
without the use of a stain [226]. More recently, PS-b-PI was also imaged generated from the forward-scattered beam are then processed by fit-
successfully with another type of hole-free phase plate that uses a ting a virtual detector that matches the geometry of the phase plate.
uniform thin film in the back focal plane (Fig. 29B) [227]. Here, the Aberration correction is crucial for MIDI-STEM (and any other phase
difference in phase shift arises from the fact that charging occurs only at contrast method) because the q4 dependence of third order spherical
the beam crossover. Because the primary beam will induce the emission aberrations causes phase shifts past π/4, where contrast gets worse or
of secondary electrons from the film, the consequential local bias in the flips sign, in the large scattering vector (q = 1 to 2 1/Å) range. Thus,
film creates a phase shift in the diffracted beams with respect to the while the MIDI-STEM phase plates fill in most of the low spatial fre-
primary beam. quencies, C3 spherical aberration correction is necessary to image with
the high spatial frequencies. MIDI-STEM, which combines phase plates,
aberration correction, direct electron detectors, and phase reconstruc-
8.2. Spherical aberration corrected STEM with phase plates
tion with an interference pattern, produces almost ideal linear phase-
contrast images over a wide range of spatial frequencies, making it an
Image acquisition in STEM mode can occur via a bright field de-
attractive method for imaging soft materials [228].
tector (collects forward-scattered electrons), an annular dark field de-
tector (collects scattered electrons similarly to dark-field imaging), or a

Fig. 29. (A) Phase contrast image of an un-


stained lamellar polystyrene-polyisoprene di-
block copolymer (PS-b-PI) using a 300 kV mi-
croscope equipped with a Hilbert phase plate.
Reprinted with permission from [226]. Copy-
right 2005 American Chemical Society. (B)
TEM image of PS-b-PI taken with a hole-free
phase plate (inset was taken with a conven-
tional TEM at defocus). Reprinted with per-
mission from [227].

21
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 30. (a) 2D image of free standing film of poly(styrenesulfonate-b-methyl-


butylene) after exposure to air with RH = 98% for 24 h. (b) 3D electron mi-
crotomography image obtained from 53 tilt series images of sample in (a).
Reprinted with permission from [66]. Fig. 31. (a) Bright-field cryo TEM (2D projection) of a frozen-hydrated, as-cast
100 nm Nafion membrane. (b) Cryo TEM 3D reconstruction showing two per-
pendicular slices through the tomogram. Isosurface rendering is used to mark
9. Three-dimensional imaging opportunities using direct electron
the spatial distribution of the central region of the dark hydrophilic phase in
detectors
yellow. Reprinted with permission from [229]. Copyright 2014 American
Chemical Society. (For interpretation of the references to color in this figure
Understanding the three-dimensional (3D) structure of a system can legend, the reader is referred to the web version of this article.)
offer invaluable insights on the morphology that is otherwise difficult to
interpret through two-dimensional (2D) projections. Electron tomo-
and opportunities have been summarized in various reviews
graphy is often performed by taking many images at different sample
[232–236]. We thus focus our discussion on techniques for 3D re-
tilt angles to elucidate the 3D morphology of various polymer systems.
constructions beyond the acquisition of tilt series.
For example, the orientation of domains in polymer electrolyte mem-
branes made from poly(styrenesulfonate-b-methylbutylene) copolymers
with different mol% of sulfonated polystyrene moieties in contact with 9.1. 3D structure from single particle reconstructions
humid air was imaged via electron tomography (Fig. 30) [66]. Fifty-
three TEM images were obtained using tilt angles ranging from -52° to While electron tomography of polymeric materials has been
+52° and then aligned with the aid of gold nanoparticles that served as achieved through tilt series of TEM images, the repetitive imaging at
fiducial markers. While the conventional 2D image can only suggest the one location deposits high doses to the imaged area. As such, single-
presence of perpendicular cylinders, the 3D structure demonstrates that particle cryo-electron microscopy, which does not require tilting of the
the vertically oriented cylinders span the entire thickness of the film. sample and instead uses 2D images of individual particles, is a pro-
Electron tomography has also been used to study the hydrated form mising method for beam-sensitive materials.
of Nafion membranes, revealing an interconnected channel-type net- Single-particle cryo-electron microscopy has proved to be useful in
work with a domain spacing of about 5 nm (Fig. 31) [229]. Recent work determining the macromolecular structures of many biological mate-
has demonstrated the need for 3D tomographic reconstructions to rials, such as proteins and viruses. The workflow for this process gen-
properly characterize closed voids and surface area (surface roughness) erally involves flash-freezing the biological sample, collecting 2D pro-
in water filtration membranes [230]. Cryo-tomography of P3HT as- jections via cryo-electron microscopy, aligning and averaging the
semblies in vitrified organic solvents has enabled even higher resolu- images, and then finally constructing the 3D model (Fig. 32) [237].
tion, revealing a 3D lamellar structure of the 1.7 nm stacking of con- Unlike in tomography, the collection of 2D images does not involve
jugated backbones showing increased order in the bulk of nanowires a tilt series at a single location on the sample. Instead, 2D images of
[231]. Like in the previous examples, the 3D reconstruction was ob- individual particles dispersed throughout the sample grid are taken.
tained with a TEM tilt series. This approach has been successfully ap- Although each 2D projection alone is too noisy to discern atomic detail,
plied to many polymeric systems, and progress, applications, challenges the signal can be improved by averaging over thousands of particles
using stages that can hold multiple samples and acquire images

22
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

automatically. With the development of direct electron detectors, 2D


projections can be collected with much greater sensitivity. Moreover,
because particles are frozen in random orientations throughout the
sample grid, the 2D projections of such a large number of individual
particles provides the distribution of 2D views necessary to reconstruct
the 3D structure. Once the 2D images are aligned and averaged, soft-
ware can be used to construct a 3D map that is iteratively refined.
While this method has been successful for biological samples where
each particle is identical, the conformational freedom of most polymers
and polymer nanostructures makes averaging over multiple regions and
samples essentially impossible. Nevertheless, a combination of crystal-
lographic and single-particle methods has been successful on a highly
ordered, synthetic polymer: a peptoid [238,239]. As seen in Fig. 33,
these images revealed the V-shaped motif of a peptoid nanosheet with
atomic resolution. We predict that single particle reconstructions may
reveal 3D reconstruction of other highly ordered polymeric materials
that are challenging to achieve through other means, such as for the
unit cell of nanocrystalline polymers.

9.2. Exit wave reconstructions

Exit wave reconstruction is a well-established technique for HRTEM


in which a series of images taken at different defocus can be used to
improve image interpretation and increase resolution. Phase informa-
tion is lost when recording an image or scattering intensities, although
in principle the information is present due to interference; exit wave
reconstructions take advantage of phase interference from different
values of defocus to recover phase information of the exit wave. More
recently, exit wave reconstructions have been used to extract 3D in-
formation at the atomic level. This is possible because although images
acquired in HRTEM are 2D projections, the reconstructed exit wave
contains 3D information due to the sample thickness, relative position
of atomic layers, and multiple scattering events that occur as the
electrons pass through the sample. In fact, simulations have demon-
strated that the use of direct detectors allow for accurate exit wave
reconstructions that can lead to extraction of 3D information [240].
This method is especially attractive for polymers because it is compa-
tible with low-dose imaging, such that beam damage is minimized
without compromising atomic resolution.

9.3. MicroED tomography

Electron crystallography has been a powerful tool for studying


atomic structure for many years, but it has traditionally been restricted
to 2D patterns from crystals and has only modest resolution (4–10 Å). In
2013, electron diffraction was used to determine the structure of a
protein in 3D using a method called MicroED [241]. The overall idea of
this technique is that if many diffraction patterns are taken from a
single crystal, a large enough region of reciprocal space would be re-
presented such that the crystal data set could be properly indexed. This
is analogous to X-ray crystallography, but has two important distinc-
tions. First, electrons interact more strongly with matter despite less
energy onto the sample, thereby allowing electron diffraction to extract
high-resolution data from very small crystals. Second, electrons have a
much smaller wavelength and are therefore able to probe smaller length
scales.
The first iteration of MicroED involved a series of still diffraction
patterns as the crystal was rotated in discrete angles between exposures,
with the entire process conducted under cryogenic conditions and a
dose rate on the order of 0.01 e/Å2s. This resulted in a 3D lysozyme
Fig. 32. Workflow of single-particle cryo-EM of a protein sample. Reprinted structure with 2.9 Å resolution. Following this proof of concept, a more
with permission from [237]. sophisticated version of MicroED was developed. This involved con-
tinuous-rotation of the crystal and was made possible by the high frame
rates achievable by direct electron detectors [242]. Continuous-rotation
MicroED not only resulted in improved data quality, but allowed data
processing to be done with standard X-ray crystallographic software.

23
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 33. A combination of crystallographic and


single-particle methods, originally developed
for cryo-electron microscopy of biological
macromolecules, was successfully im-
plemented on self-assembled nanosheets of a
peptoid polymer. The atomic length scales
achieved with this technique represents a new
level of resolution for polymer microscopy.
Reprinted with permission from [238]. Copy-
right 2019 American Chemical Society.

Fig. 34. Workflow for microED (top) and example structures determined by this method (a–d). Structures are shown with their resolution in parentheses, the full
model on the left, and the representative region of the model and density map on the right. Reprinted from [243], with permission from Elsevier.

The workflow for MicroED involves identification of a suitable damage under the electron beam during imaging, it is difficult to dis-
crystal, continuous collection of diffraction data as the sample is rotated tinguish between morphological changes caused by the beam versus
in the beam, and then indexing by standard X-ray crystallography changes caused by the applied external stimuli. A better understanding
programs (Fig. 34) [243]. So far, MicroED has solved the crystal of beam damage combined with direct electron detectors that will allow
structure for a variety of biological samples, such as lysozyme, trypsin, for low-dose imaging makes it an ideal time to revisit in situ TEM ex-
and thermolysin, among others. Nevertheless, to our knowledge, it has periments for polymers.
not yet been applied to synthetic polymers.
Most of the samples discussed in this Section have been captured in
their frozen state. Nevertheless, in order to study solutions in their true, 10.1. Heating/cooling stages
liquid form or to study solid samples under different external stimuli, in
situ TEM methods, which are discussed in the next Section, are needed. Perhaps the most common of in situ TEM studies involves tem-
perature changes during experimentation. The earliest use of tem-
perature in situ TEM comes from Easterling in 1970 to study iron and
10. In situ TEM using new sample holders iron-nickel composites at the atomic level [245]. The ability to look at
changes in the microscopy images and their Fourier transforms during
Demonstrations of in situ electron microscopy go back to work by real-time heating events allows for monitoring growth processes at the
Pashley in 1956, who examined material epitaxy in thin films [244]. atomic scale to understand growth and growth mechanisms [246–249].
Combining TEM and external pressures, such as temperature changes, Recent developments in precise measurement control of TEM holders
mechanical stresses, electrical bias, and chemical exposure, allows have further pushed the boundaries of temperature-dependent mor-
microscopists to study instantaneous material changes at the atomic phological resolution [248,250]; sample holders with very small
level. Using various techniques to study samples while they undergo thermal mass allow for increased temperature control.
these phenomena is critical to understanding certain dynamics of these A schematic of in situ TEM heating can be seen in Fig. 35, along with
materials. One of the major challenges of in situ TEM of polymers is electron diffraction patterns of the thermolysis of ammonium tetra-
again electron beam damage. Because polymers are susceptible to thiomolybdate [246]. As the sample is heated from room temperature

24
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 35. An in-situ heating TEM experiment is shown. (a) A schematic of the heating stage is presented, with the red “solid precursor” being (NH4)2MoS4. (b–e)
Selected-area electron diffraction of the thermolysis of (NH4)2MoS4 is shown at various temperatures (25 °C, 400 °C, 780 °C and 900 °C), demonstrating the crys-
tallization of MoS2 up to 800 °C followed by decomposition of MoS2 into metallic Mo at about 800 °C. This work is licensed under the Creative Commons Attribution
4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ or send a letter to Creative Commons, PO Box 1866,
Mountain View, CA 94042, USA. [246]. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

to 900 °C, crystallization is observed, followed by metallic decomposi- developments in TEM sample holders have focused on the miniatur-
tion. By observing these processes in situ, not only can the existence of ization of mechanical testing instrumentation that is compatible with
crystallization at a specific temperature be confirmed, but the grain size sample stage space limitations while maintaining the ability to accu-
of these crystals can also be determined. Phase changes can be directly rately control and measure applied stress, such as in situ nanoindenters
observed in real-time as well, further increasing the flexibility of this [256,257] and on-chip microelectromechanical systems (MEMS)
technique. [258,259].
In situ heating in the TEM has been applied to many inorganic Since its inception a few decades ago, in situ TEM mechanical testing
systems, such as for observing the growth mechanisms of 2D mo- has made important advances in materials science, by revealing the
lybdenum disulfide flakes [246], the rotation of grains in nanocrystal- nanoscale origins for deformation [260], tearing [261], and compres-
line platinum [247], the martensitic transformation of nickel titanium sion [262] of molybdenum disulfide, the compression of metal-organic
shape memory alloys [249], and grain growth in nanocrystalline copper framework microcrystals [263], tensile deformation of carbon nanofi-
thin films [251]. To our knowledge, its application to organic materials bers [264], and the nucleation and propagation of dislocations in alu-
has not been widely published. Nevertheless, in situ heating in the TEM minum films [265]. In situ mechanical studies of polymers would be
has offered valuable contributions to organic-inorganic systems, such as valuable because there is a wide range of micromechanical processes
for monitoring the degradation of organometallic halide perovskite that occurs in polymers under load, such as nanometer scale changes to
solar cells under heating [248] and observing the loss of a capping individual macromolecular segments and micrometer scale plastic
polymer in the heating of polymer-capped platinum nanocrystals [252]. yielding (e.g., crazing and shear bands). Unfortunately, the main lim-
Even minimal damage below the critical dose as measured via itation, in addition to aforementioned interpretation difficulties that
electron diffraction could cause chemical changes that alter the dy- arise as a consequence of beam damage, is the challenge of conducting
namics and phase behavior of organic materials, thereby limiting in situ mechanical tests on polymer films that are thin enough to be electron
heating TEM studies. As a consequence, a series of in situ heating images transparent. One solution has been high-voltage electron microscopy
that are taken with a total dose below which damage to the chemical (HVTEM), which allows for thicker specimens. For example, in situ
structure is significant is needed; we propose that low-loss EELS may be straining of poly(styrene-co-acrylonitrile) (PSAN) blended with poly
useful to define this acceptable dose by measuring damage to the va- (methyl methacrylate) (PMMA)/acrylate-rubber/PMMA core-shell
lence electronic structure. particles in a HVTEM (1000 keV) at room temperature revealed si-
On the other hand, materials can also be studied in the TEM while multaneous crazing and shear yielding of the PSAN and fibrillation of
undergoing cooling (in situ cooling). Using TEM cooling stages, various the rubber particles with drawing of material from the PMMA cores
phase and magnetic transformations can be accessed that cannot gen- (Fig. 36) [266]. For this study, microtomed polymer sections were
erally be observed with isothermal TEM stages [253,254]. While the glued between two metallic films to create a “sample/metallic-film
effects of heating are often detrimental to polymeric materials, cooling sandwich” that could be mounted into a tensile holder.
slows down transient processes and has enabled imaging of dynamic Continued advances of in situ mechanical testing systems will likely
processes, for example, in the cooling of nanolaminates [255]. We be necessary in order to overcome current limitations in the field. For
predict that the slowing of reactions due to cooling could be applicable example, current MEMS chips are not amenable for soft materials, as
to polymer electron microscopy as well. Unlike the aforementioned specimens are often attached to MEMS devices inside a FIB. The advent
cryo-TEM, where motion is essentially frozen, variable cooling could of the cryo-FIB, discussed in sections 3.1 and 3.3, could help mitigate
enable imaging of dynamic processes at select temperatures, such as this issue. Furthermore, as increasingly creative methods for measuring
polymer crystallization upon quenching. Such a study could also be the mechanical properties of polymer thin-films are explored, such as
useful for visualizing processes associated with Tg for polymers with Tg tensile testing thin-films on a liquid surface [267] or an elastomer
below room temperature. Care must be taken to use low dose rates such [268], in situ mechanical testing of thin films in the TEM could be more
that local heating from the electron beam does not alter the desired widely adopted and used to explore phenomena such as shear-induced
imaging condition temperature. crystallization [269,270].

10.2. Mechanical deformations 10.3. Liquid cell TEM

Although mechanical properties of materials are directly linked to Recent development in sample holders that can encase a droplet of
their structure, ex situ mechanical testing complemented by imaging of liquid have enabled liquid cell TEM [271–274]. The use of 100 nm
stress-induced microstructural evolution struggles to connect nanoscale Si3N4 films to create liquid sample holders was demonstrated for the
phenomena to macroscopic responses to applied stress. Recent TEM in 2003 by Williamson while at the same time Thiberge developed

25
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 36. (A) TEM image of stained core-shell particles in PSAN matrix. (B) Deformation of core-shell particles at room temperature. Direction of deformation
indicated by arrow. Reprinted with permission from [266].

a similar cell for the scanning electron microscope (SEM) [273,274].


Because the window material adds background scattering, Si3N4 films
have been further thinned to 25 nm [275]. Liquid cells have enabled
studies of nanoparticle and nanocrystal synthesis growth, as well as
growth of lithium dendrites at the interface between metal electrodes
and organic electrolytes [276–281]. Controlling the sample and
window thickness to minimize background scattering, as well as im-
provements in detectors, continue to push the limits of liquid cell TEM
[282]. Two other limitations of liquid TEM, electrical charging and
radiolysis of the liquid under the electron beam, have also recently been
addressed by the development of encapsulating graphene sheets that
alleviate charging and using deuterated water to prolong the sample
lifetime [283].
Liquid cell TEM has been demonstrated on polymer solutions.
Fig. 37A–C shows a small amount of liquid in between two graphene
sheets where solubilized polymer chains are apparent [284]. The au-
thors could discern changes in polymer conformations as a function of
time at the timescale of seconds. Given that polymer segmental dy-
namics are expected to be much faster, the authors speculated that
polystyrene sulfonate chains were absorbed onto graphene, thereby
slowing down motion. Liquid cell TEM has also been conducted on a
solution of amphiphilic block copolymer micelles, revealing the evo-
lution of micelle fusion and growth in real time with nanometer spatial
resolution (Fig. 37D) [285]. These studies clearly demonstrate the
possibility of examining dynamical processes of polymers in solutions.

10.4. Electrical bias

The application of electric fields is possible either in the solid state Fig. 37. Liquid cell TEM. (a) Schematic of liquid cell TEM, showing a polymer
solution between two electron transparent substrates (graphene sheets). (b)
or in solution through microfabricated chips and sample holders.
Low magnification TEM image that shows the polymer solution channel in
Electrical bias experiments in the TEM have elucidated field-induced
between graphene sheets. (c) TEM image of polymer molecules in solution.
phenomena such as oxygen vacancy migration in cerium oxides [286], Reprinted, by permission, from [284]. Copyright 2017 by John Wiley & Sons,
reversible formation of lamellar domains in a piezoelectric [287], Inc. (D) Direct observation of growth and evolution processes in block copo-
structural damage of multiwalled nanotubes [288], and nanostructure lymer micelles. Reprinted with permission from [285]. Copyright 2017 Amer-
changes in perovskite solar cells as a function of a current-voltage sti- ican Chemical Society.
mulus [289]. In the case of polymers, resolving phase transformations,
structural rearrangement, or time-resolved processes due to an applied studies reveal that particulates are formed in solution as the deposition
field requires enough radiation hardness to acquire multiple images. As
proceeds, followed by deposition of these polymer particles. This re-
a consequence, no direct application to polymers has been demon- fines previous descriptions of electrochemical polymerization, which
strated.
assumed that only monomers approach the surface prior to chain
The development of liquid cell TEM allowed for the development of growth.
electrochemical sample holders by adding electrodes to Si3N4 windows
[258,274,275,277,280,290–293]. In situ TEM electrochemistry allowed
for studies of dendrite growth, intercalation and interfacial changes in 11. Summary
electrochemical cells such as batteries [277,280,291,293]. The elec-
trochemical polymerization of poly(3,4-ethylenedioxythiophene) TEM has transformed our understanding of the structure and mor-
(PEDOT) has also been visualized, as shown in Fig. 38 [294]. TEM phology of polymeric materials. Throughout the past few decades,

26
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

Fig. 38. Direct imaging of the in situ electrochemical deposition of PEDOT from an aqueous solution of EDOT monomer using TEM with an electrochemical liquid
flow cell. Captured frames from live video of growth of PEDOT clusters during electrochemical deposition (a–f) and projected areas of individual PEDOT clusters
(g).Reprinted with permission from [294]. Copyright 2015 American Chemical Society.

polymer microscopy has evolved to include high-resolution imaging, Acknowledgements


spectrum imaging, tomography, and in-situ characterization.
Nevertheless, growth in the field has always been hampered by the Funding for this work was provided by NSFDMR-1609417. BK also
inherent lack of contrast and beam sensitivity of polymers. The ac- acknowledges support from the DOE Office of Science Graduate Student
complishments of polymer microscopy despite these constraints have Research (SCGSR) Award, and the ALS Doctoral Fellowship in
been impressive, although new advancements in instrumentation such Residence.
as specimen preparation tools, monochromators, aberration correctors,
phase plates, direct electron detectors, and in situ sample holders can References
push imaging limits even further.
For example, the development of sample preparation techniques [1] E.W.F.R. Eppe, H.A. Stuart, J. Polym. Sci. XXXIV (1959) 721–740.
based on the FIB, new microtome technology, and cryo-preparation [2] N. Uyeda, Dye. Pigment. 6 (1985) 115–134.
[3] B.M. Discher, Y.-Y. Won, D.S. Ege, J.C.-M. Lee, F.S. Bates, D.E. Discher,
methods will enable new imaging experiments. The application of D.A. Hammer, Sci. 284 (1999) 1143–1146.
monochromators to polymer microscopy provide a means to control [4] D.J. Pochan, Z. Chen, H. Cui, K. Hales, K. Qi, K.L. Wooley, Science 306 (2004)
beam flux, and a way to probe beam damage through monochromated 94–97.
[5] K. Yamauchi, K. Takahashi, H. Hasegawa, H. Iatrou, N. Hadjichristidis, T. Kaneko,
EELS, which in turn paves the way for optimized EFTEM and spectrum Y. Nishikawa, H. Jinnai, T. Matsui, H. Nishioka, M. Shimizu, H. Furukawa,
imaging using differences in valence electronic structure. Various ex- Macromolecules 36 (2003) 6962–6966.
periments have been made possible by new sample holders, such as [6] J. Sun, X. Jiang, R. Lund, K.H. Downing, N.P. Balsara, R.N. Zuckermann, Proc.
Natl. Acad. Sci. 113 (2019) 3954–3959.
liquid TEM and in situ electrochemistry. By highlighting the limitations
[7] A. Sousa, A. Aitouchen, M. Libera, Ultramicroscopy 106 (2006) 130–145.
and potential of HRTEM of polymers, we establish that a combination [8] E.D. Gomez, A. Panday, E.H. Feng, V. Chen, G.M. Stone, A.M. Minor,
of protocols to minimize the effects of beam damage with aberration C. Kisielowski, K.H. Downing, O. Borodin, G.D. Smith, N.P. Balsara, Nano Lett. 9
(2009) 1212–1216.
correctors and direct electron detectors will push the resolution limits
[9] S.V. Kesava, Z. Fei, A.D. Rimshaw, C. Wang, A. Hexemer, J.B. Asbury, M. Heeney,
of polymer microscopy despite beam sensitivity. We propose that E.D. Gomez, Adv. Energy Mater. 4 (2014) 1400116.
electron microscopy is poised to leap forward and continue to transform [10] C. Guo, D.R. Kozub, S.V. Kesava, C. Wang, A. Hexemer, E.D. Gomez, ACS Macro
our understanding of structure and structure-property relationships in Lett. 2 (2013) 185–189.
[11] K. Vakhshouri, D.R. Kozub, C. Wang, A. Salleo, E.D. Gomez, Phys. Rev. Lett. 108
polymers. Indeed, various techniques highlighted here are currently (2012) 026601.
underexplored in polymer microscopy, such as 3D reconstructions be- [12] D.R. Kozub, K. Vakhshouri, L.M. Orme, C. Wang, A. Hexemer, E.D. Gomez,
yond tilt series and 4D STEM. Macromolecules 44 (2011) 5722–5726.
[13] R.M. Glaeser, J. Ultrastruct. Res. 36 (1971) 466.
[14] M. Haider, S. Uhlemann, E. Schwan, H. Rose, B. Kabius, K. Urban, Nature 392
(1998) 768.
Declaration of Competing Interest [15] O. Scherzer, Optik 2 (1947) 114–132.
[16] H. Rose, Optik 85 (1990) 19–24.
[17] D.R. Kozub, K. Vakhshouri, S.V. Kesava, C. Wang, A. Hexemer, E.D. Gomez, Chem.
The authors declare that they have no known competing financial Commun. 48 (2012) 5859–5861.
interests or personal relationships that could have appeared to influ- [18] C. Guo, F.I. Allen, Y. Lee, T.P. Le, C. Song, J. Ciston, A.M. Minor, E.D. Gomez, Adv.
Funct. Mater. 25 (2015) 6071–6076.
ence the work reported in this paper.
[19] L.F. Drummy, R.J. Davis, D.L. Moore, M. Durstock, R.A. Vaia, J.W.P. Hsu, Chem.
Mater. 23 (2011) 907–912.
[20] A.A. Herzing, L.J. Richter, I.M. Anderson, J. Phys. Chem. C 114 (2010)
17501–17508.

27
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

[21] B.D. Forbes, L. Houben, J. Mayer, R.E. Dunin-Borkowski, L.J. Allen, O.D. Lavrentovich, M.G. Tamba, A. Kohlmeier, Microsc. Res. Tech. 77 (2014)
Ultramicroscopy 147 (2014) 98–105. 754–772.
[22] Z. Wang, A.H. Tavabi, L. Jin, J. Rusz, D. Tyutyunnikov, H. Jiang, Y. Moritomo, [72] T.F. Kelly, O. Nishikawa, J.A. Panitz, T.J. Prosa, MRS Bull. 34 (2009) 744–750.
J. Mayer, R.E. Dunin-Borkowski, R. Yu, J. Zhu, X. Zhong, Nat. Mater. 17 (2018) [73] K. Niihara, T. Kaneko, T. Suzuki, Y. Sato, H. Nishioka, Y. Nishikawa, T. Nishi,
221–225. H. Jinnai, Macromolecules 38 (2005) 3048–3050.
[23] F.F. Krause, A. Rosenauer, J. Barthel, J. Mayer, K. Urban, R.E. Dunin-Borkowski, [74] A. Zankel, J. Wagner, P. Poelt, Micron 62 (2014) 66–78.
H.G. Brown, B.D. Forbes, L.J. Allen, Ultramicroscopy 181 (2017) 173–177. [75] K. Mangipudi, V. Radisch, L. Holzer, C. Volkert, Ultramicroscopy 163 (2016)
[24] B. Kabius, P. Hartel, M. Haider, H. Müller, S. Uhlemann, U. Loebau, J. Zach, 38–47.
H. Rose, J. Electron Microsc. 58 (2009) 147–155. [76] J. Jésior, Scanning Microsc. (Suppl. 3) (1989) 147–152 discussion 152-143.
[25] B. Kabius, H. Rose, CHAPTER 7 - Novel Aberration Correction Concepts, in: [77] A. Al‐Amoudi, J. Dubochet, H. Gnaegi, W. Lüthi, D. Studer, J. Microsc. 212 (2003)
P.W. Hawkes (Ed.), Advances in Imaging and Electron Physics, vol. 153, Elsevier, 26–33.
2008, pp. 261–281. [78] T. Matzelle, H. Gnaegi, A. Ricker, R. Reichelt, J. Microsc. 209 (2003) 113–117.
[26] D.C. Martin, J. Chen, J. Yang, L.F. Drummy, C. Kübel, J. Polym. Sci. Part B: Polym. [79] D. Studer, H. Gnaegi, J. Microsc. 197 (2000) 94–100.
Phys. 43 (2005) 1749–1778. [80] Q. Xu, R.M. Rioux, M.D. Dickey, G.M. Whitesides, Acc. Chem. Res. 41 (2008)
[27] C.B.C.D.B. Williams, Transmission Electron Microscopy, Springer, 2019. 1566–1577.
[28] J. Wang, Q. Li, C. Ronning, D. Stichtenoth, S. Muller, D. Tang, Micron 39 (2008) [81] S. Jain, F.S. Bates, Science 300 (2003) 460–464.
703–708. [82] Y. Won, A.K. Brannan, H.T. Davis, F.S. Bates, J. Phys. Chem. B 106 (2002)
[29] P.C. Tiemeijer, M. Bischoff, B. Freitag, C. Kisielowski, Ultramicroscopy 114 (2012) 3354–3364.
72–81. [83] A. Napoli, M. Valentini, N. Tirelli, M. Müller, J.A. Hubbell, Nat. Mater. 3 (2004)
[30] K. Kimoto, K. Kurashima, T. Nagai, M. Ohwada, K. Ishizuka, Ultramicroscopy 121 183.
(2012) 31–37. [84] D. Zhao, V. Gimenez-Pinto, A.M. Jimenez, L. Zhao, J. Jestin, S.K. Kumar, B. Kuei,
[31] C. Hetherington, Mater. Today (2004). E.D. Gomez, A.S. Prasad, L.S. Schadler, M.M. Khani, B.C. Benicewicz, ACS Cent.
[32] H. Lichte, D. Geiger, M. Linck, Philos. Trans. A Math. Phys. Eng. Sci. 367 (2009) Sci. 3 (2017) 751–758.
3773–3793. [85] J.S. Vastenhout, H. Gnägi, Microsc. Microanal. 8 (2002) 324–325.
[33] R. Danev, K. Nagayama, Ultramicroscopy 88 (2001) 243–252. [86] P. Walther, J. Microsc. 212 (2003) 34–43.
[34] R. Danev, K. Nagayama, J. Struct. Biol. 161 (2008) 211–218. [87] T. Hashimoto, G.E. Thompson, X. Zhou, P.J. Withers, Ultramicroscopy 163 (2016)
[35] B. Kuei, B. Kabius, J.L. Gray, E.D. Gomez, MRS Commun. (2018). 6–18.
[36] D.S. Rao, K. Muraleedharan, C. Humphreys, Microscopy: Science, Technology, [88] M. Nasiri, A. Bertrand, T.M. Reineke, M.A. Hillmyer, ACS Appl. Mater. Interfaces 6
Applications and Education, (2010), pp. 1232–1244. (2014) 16283–16288.
[37] A. Barna, B. Pécz, M. Menyhard, Micron 30 (1999) 267–276. [89] Q. Xu, R. Perez-Castillejos, Z. Li, G.M. Whitesides, Nano Lett. 6 (2006) 2163–2165.
[38] L. Sawyer, D. Grubb, G.F. Meyers, Polymer microscopy, Springer Science & [90] R.F. Thompson, M. Walker, C.A. Siebert, S.P. Muench, N.A. Ranson, Methods 100
Business Media, (2008). (2016) 3–15.
[39] E. Roche, R. Stein, E. Thomas, J. Polym. Sci. Part B: Polym. Phys. 18 (1980) [91] D. Studer, B.M. Humbel, M. Chiquet, Histochem. Cell Biol. 130 (2008) 877–889.
1145–1158. [92] J.C. Gilkey, L.A. Staehelin, Microsc. Res. Technol. 3 (1986) 177–210.
[40] T. Uragami, Science and Technology of Separation Membranes, 2 Volume Set, [93] M. Winey, J.B. Meehl, E.T. O’Toole, T.H. Giddings, Mol. Biol. Cell 25 (2014)
John Wiley & Sons, 2017. 319–323.
[41] L.A. Giannuzzi, F.A. Stevie, Micron 30 (1999) 197–204. [94] X. Bai, G. McMullan, S.H.W. Scheres, Trends Biochem. Sci. 40 (2015) 49–57.
[42] C.A. Volkert, A.M. Minor, MRS Bull. 32 (2007) 389–399. [95] R. Fernandez-Leiro, S.H.W. Scheres, Nature 537 (2016) 339.
[43] M. Corazza, S.B. Simonsen, H. Gnaegi, K.T. Thydén, F.C. Krebs, S.A. Gevorgyan, [96] R. Henderson, Q. Rev. Biophys. 37 (2004) 3–13.
Appl. Surf. Sci. 389 (2016) 462–468. [97] E. Nogales, S.H.W. Scheres, Mol. Cell 58 (2015) 677–689.
[44] R. Wirth, Chem. Geol. 261 (2009) 217–229. [98] J.L.S. Milne, M.J. Borgnia, A. Bartesaghi, E.E.H. Tran, L.A. Earl, D.M. Schauder,
[45] A. Rigort, J.M. Plitzko, Arch. Biochem. Biophys. 581 (2015) 122–130. J. Lengyel, J. Pierson, A. Patwardhan, S. Subramaniam, FEBS J. 280 (2013) 28–45.
[46] S. Kim, M.J. Park, N.P. Balsara, G. Liu, A.M. Minor, Ultramicroscopy 111 (2011) [99] D. Danino, Curr. Opin. Colloid Interface Sci. 17 (2012) 316–329.
191–199. [100] H. Cui, T.K. Hodgdon, E.W. Kaler, L. Abezgauz, D. Danino, M. Lubovsky,
[47] P.D. Prewett, G.L.R. Mair, Focused Ion Beams From Liquid Metal Ion Sources, Y. Talmon, D.J. Pochan, Soft Matter 3 (2007) 945–955.
Research Studies PressLtd, 1991. [101] K.R. Vinothkumar, R. Henderson, Q. Rev. Biophys. 49 (2016).
[48] S. Reyntjens, R. Puers, J. Micromech. Microeng. 11 (2001) 287. [102] R.M. Glaeser, Rev. Sci. Instrum. 84 (2013) 312.
[49] N. Bassim, B. De Gregorio, A. Kilcoyne, K. Scott, T. Chou, S. Wirick, G. Cody, [103] R.M. Glaeser, J. Struct. Biol. 128 (1999) 3–14.
R. Stroud, J. Microsc. 245 (2012) 288–301. [104] L. Gan, G.J. Jensen, Q. Rev. Biophys. 45 (2012) 27–56.
[50] L.A. Giannuzzi, B. Kempshall, S. Schwarz, J. Lomness, B. Prenitzer, F. Stevie, FIB [105] H. Moor, U. Riehle, Snap-freezing under high pressure: a new fixation technique
Lift-Out Specimen Preparation Techniques, Introduction to Focused Ion Beams, for freeze-etching, Proc. Fourth Eur. Reg. Conf. Electr. Microsc, (1968), pp.
Springer, 2005, pp. 201–228. 445–446.
[51] J. Mayer, L.A. Giannuzzi, T. Kamino, J. Michael, MRS Bull. 32 (2007) 400–407. [106] H. Moor, Cryotechniques in Biological Electron Microscopy Vol. 17 (1987), pp.
[52] F. Stevie, T. Shane, P. Kahora, R. Hull, D. Bahnck, V. Kannan, E. David, Surf. 175–191.
Interface Anal. 23 (1995) 61–68. [107] A. Leforestier, K. Richter, F. Livolant, J. Dubochet, J. Microsc. 184 (1996) 4–13.
[53] D.M. Longo, J.M. Howe, W.C. Johnson, Ultramicroscopy 80 (1999) 69–84. [108] D. Studer, M. Michel, M. Müller, Scanning Microsc. (Suppl. 3) (1989) 253–268
[54] K.M. Reddy, P. Liu, A. Hirata, T. Fujita, M. Chen, Nat. Commun. 4 (2013) 2483. discussion 268-259.
[55] H. White, Y. Pu, M. Rafailovich, J. Sokolov, A. King, L. Giannuzzi, C. Urbanik- [109] J. Escaig, J. Microsc. 126 (1982) 221–229.
Shannon, B. Kempshall, A. Eisenberg, S. Schwarz, Polymer 42 (2001) 1613–1619. [110] P. Brüggeller, E. Mayer, Nature 288 (1980) 569–571.
[56] W. Brostow, B.P. Gorman, O. Olea-Mejia, Mater. Lett. 61 (2007) 1333–1336. [111] M.J. Costello, Ultrastruct. Pathol. 30 (2006) 361–371.
[57] R. Anderson, S.J. Klepeis, MRS Online Proceedings Library Archive, (1997), p. [112] W.F. Tivol, A. Briegel, G.J. Jensen, Microsc. Microanal. 14 (2008) 375–379.
480. [113] H.X. Sui, K.H. Downing, Nature 442 (2006) 475–478.
[58] J. Loos, J.K. van Duren, F. Morrissey, R.A. Janssen, Polymer 43 (2002) [114] D. Tennant, J. Microsc. 197 (2000) 68–79.
7493–7496. [115] J.C. Taveau, D. Nguyen, A. Perro, S. Ravaine, E. Duguet, O. Lambert, Soft Matter 4
[59] T. Kamino, T. Yaguchi, T. Ohnishi, T. Ishitani, M. Osumi, J. Electron Microsc. 53 (2008) 311–315.
(2004) 563–566. [116] J. Zhu, S. Zhang, K. Zhang, X. Wang, J.W. Mays, K.L. Wooley, D.J. Pochan, Nat.
[60] H. Luo, S. Yin, G. Zhang, C. Liu, Q. Tang, M. Guo, Ultramicroscopy (2017). Commun. 4 (2013) 2297.
[61] M. Schaffer, J. Wagner, B. Schaffer, M. Schmied, H. Mulders, Ultramicroscopy 107 [117] J. Leunissen, H. Yi, J. Microsc. 235 (2009) 25–35.
(2007) 587–597. [118] H. Han, J. Huebinger, M. Grabenbauer, J. Struct. Biol. 178 (2012) 84–87.
[62] J. Liebig, M. Göken, G. Richter, M. Mačković, T. Przybilla, E. Spiecker, O. Pierron, [119] H. Mulders, GIT Imaging Microsc. 2 (2003) 8–10.
B. Merle, Ultramicroscopy 171 (2016) 82–88. [120] M. Marko, C. Hsieh, W. Moberlychan, C. Mannella, J. Frank, J. Microsc. 222
[63] S. Welz, N. Browning, A. Minor, Microscopy and Microanalysis-New York- 11, (2006) 42–47.
(2005) 834CD. [121] M. Marko, C. Hsieh, R. Schalek, J. Frank, C. Mannella, Nat. Methods 4 (2007)
[64] S. Kim, G. Liu, A.M. Minor, Micros. Today 17 (2009) 20–23. 215–218.
[65] B.F. Vieweg, B. Butz, W. Peukert, R.N.K. Taylor, E. Spiecker, Ultramicroscopy 113 [122] A. Rigort, F.J. Bäuerlein, A. Leis, M. Gruska, C. Hoffmann, T. Laugks, U. Böhm,
(2012) 165–170. M. Eibauer, H. Gnaegi, W. Baumeister, J. Struct. Biol. 172 (2010) 169–179.
[66] M.J. Park, S. Kim, A.M. Minor, A. Hexemer, N.P. Balsara, Adv. Mater. 21 (2009) [123] M.J. Zachman, T. Zengyuan, S. Choudhury, L.A. Archer, L.F. Kourkoutis, Nature
203–208. 560 (2018) 345–349.
[67] J. He, C. Hsieh, Y. Wu, T. Schmelzer, P. Wang, Y. Lin, M. Marko, H. Sui, J. Struct. [124] R.F. Egerton, I. Rauf, Ultramicroscopy 80 (1999) 247–254.
Biol. (2017). [125] B. Kuei, E.D. Gomez, Microsc. Microanal. 24 (2018) 1988–1989.
[68] J. Mahamid, R. Schampers, H. Persoon, A.A. Hyman, W. Baumeister, J.M. Plitzko, [126] K. Siangchaew, M. Libera, Philos. Mag. A 80 (2000) 1001–1016.
J. Struct. Biol. 192 (2015) 262–269. [127] Z.J.W.A. Leijten, A.D.A. Keizer, Gd. With, H. Friedrich, J. Phys. Chem. B 121
[69] K.M. Strunk, K. Wang, D. Ke, J. Gray, P. Zhang, J. Microsc. 247 (2012) 220–227. (2017) 10552–10561.
[70] L. Mielanczyk, N. Matysiak, M. Michalski, R. Buldak, R. Wojnicz, Folia Histochem. [128] L.F. Drummy, J. Yang, D.C. Martin, Ultramicroscopy 99 (2004) 247–256.
Cytobiol. 52 (2014) 1–17. [129] D.C. Martin, E.L. Thomas, Polymer 36 (1995) 1743–1759.
[71] M. Gao, Y.K. Kim, C. Zhang, V. Borshch, S. Zhou, H.S. Park, A. Jákli, [130] S. Kumar, W.W. Adams, Polymer 31 (1990) 15–19.

28
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

[131] D.T. Grubb, J. Mater. Sci. 9 (1974) 1715–1736. [190] S.J. Juhl, T. Wang, B. Vermilyea, X. Li, V.H. Crespi, J.V. Badding, N. Alem, J. Am.
[132] E.D. Gomez, M.L. Ruegg, A.M. Minor, C. Kisielowski, K.H. Downing, R.M. Glaeser, Chem. Soc. 141 (2019) 6937–6945.
N.P. Balsara, Macromolecules 41 (2008) 156–162. [191] H. Huang, M.D. Ward, S.J. Juhl, A. Biswas, N. Alem, J.V. Badding, T.A. Strobel,
[133] R.F. Egerton, Microsc. Res. Tech. 75 (2012) 1550–1556. Microsc. Microanal. 24 (2018) 2030–2031.
[134] R.F. Egerton, Ultramicroscopy 127 (2013) 100–108. [192] C. Kisielowski, S. Helveg, L. Hansen, P. Specht, Microsc. Microanal. 21 (2015)
[135] R.F. Egerton, P. Li, M. Malac, Micron 35 (2004) 399–409. 1321–1322.
[136] M.R. Libera, R.F. Egerton, Polym. Rev. 50 (2010) 321–339. [193] B. Siwick, American Physical Society Viewpoint Vol. 10 (2017), p. 39.
[137] D.B. Williams, C.B. Carter, Transmission Electron Microscopy: A Textbook for [194] G. Sciaini, R.J.D. Miller, Rep. Prog. Phys. 74 (2011) 096101.
Materials Science, Springer, 2009. [195] C. Kisielowski, Adv. Mater. 27 (2015) 5838–5844.
[138] R.F. Egerton, Adv. Struct. Chem. Imaging 1 (2015) 5. [196] J.C. Meyer, J. Kotakoski, C. Mangler, Ultramicroscopy 145 (2014) 13–21.
[139] R.F. Egerton, Electron Energy-Loss Spectroscopy in the Electron Microscope, [197] Y. Yu, D. Zhang, C. Kisielowski, L. Dou, N. Kornienko, Y. Bekenstein, A.B. Wong,
Springer Verlag, 2011. A.P. Alivisatos, P. Yang, Nano Lett. 16 (2016) 7530–7535.
[140] J.C. Meyer, F. Eder, S. Kurasch, V. Skakalova, J. Kotakoski, H.J. Park, S. Roth, [198] D.F. Yancey, C. Kisielowski, P. Specht, S. Rozeveld, J. Kang, P. Nickias, Microsc.
A. Chuvilin, S. Eyhusen, G. Benner, A.V. Krasheninnikov, U. Kaiser, Phys. Rev. Microanal. 24 (2018) 1968–1969.
Lett. 110 (2013) 239902. [199] P. Specht, M. Luysberg, J. Chavez, T.R. Weatherford, T.J. Anderson, A.D. Koehler,
[141] S.M. Marcus, R.L. Blaine, Thermochim. Acta 243 (1994) 231–239. C. Kisielowki, Microsc. Microanal. 24 (2018) 1972–1973.
[142] A. Mozumder, Fundamentals of Radiation Chemistry, Academic Press, 1999. [200] C. Kisielowski, P. Specht, B. Freitag, E.R. Kieft, W. Verhoeven, J.F.M. van Rens,
[143] D.C. Martin, E.L. Thomas, Polymer 36 (1995) 1743–1759. P. Mutsaers, J. Luiten, S. Rozeveld, J. Kang, A.J. McKenna, P. Nickias, D.F. Yancey,
[144] R. Henderson, R.M. Glaeser, Ultramicroscopy 16 (1985) 139–150. Adv. Funct. Mater. 29 (2019) 1807818.
[145] A. Boudet, J. Phys. France 47 (1986) 1043–1052. [201] J. Verbeeck, D. Van Dyck, G. Van Tendeloo, Spectrochim. Acta Part B At.
[146] E.G. Rightor, A.P. Hitchcock, H. Ade, R.D. Leapman, S.G. Urquhart, A.P. Smith, Spectrosc. 59 (2004) 1529–1534.
G. Mitchell, D. Fischer, H.J. Shin, T. Warwick, J. Phys. Chem. B 101 (1997) [202] M.A. Society, M.A.S. Conference, R. Gooley, Microbeam Analysis 1983,
1950–1960. Proceedings of the 18th Annual Conference of the Microbeam Analysis Society :
[147] K. Varlot, J.M. Martin, C. Quet, Y. Kihn, Ultramicroscopy 68 (1997) 123–133. Phoenix, Arizona, 6-12 August 1983 (1983).
[148] K. Varlot, J.M. Martin, D. Gonbeau, C. Quet, Polymer 40 (1999) 5691–5697. [203] C.A. Correa, B.C. Bonse, C.R. Chinaglia, E. Hage, L.A. Pessan, Polym. Test. 23
[149] R.F. Egerton, P.A. Crozier, P. Rice, Ultramicroscopy 23 (1987) 305–312. (2004) 775–778.
[150] K. Varlot, J.M. Martin, C. Quet, J. Microsc. 191 (1998) 187–194. [204] T. Hayakawa, R. Goseki, M. Kakimoto, M. Tokita, J. Watanabe, Y. Liao,
[151] M. Tsuji, S.K. Roy, R. St. John Manley, J. Polym. Sci.: Polym. Phys. Ed. 23 (1985) S. Horiuchi, Org. Lett. 8 (2006) 5453–5456.
1127–1137. [205] S. Horiuchi, T. Hanada, M. Ebisawa, Y. Matsuda, M. Kobayashi, A. Takahara, ACS
[152] A. Sousa, A. Aitouchen, M. Libera, Ultramicroscopy 106 (2006) 130–145. Nano 3 (2009) 1297–1304.
[153] R.M. Briber, E.L. Thomas, Polymer 27 (1986) 66–70. [206] E.M. Linares, C.A.P. Leite, L.F. Valadares, C.A. Silva, C.A. Rezende, F. Galembeck,
[154] R.M. Briber, F. Khoury, J. Polym. Sci. Part B: Polym. Phys. 26 (1988) 621–636. Anal. Chem. 81 (2009) 2317–2324.
[155] Z.J.W.A. Leijten, A.D.A. Keizer, G. de With, H. Friedrich, J. Phys. Chem. C 121 [207] A.E. Ribbe, M. Hayashi, M. Weber, T. Hashimoto, Macromolecules 33 (2000)
(2017) 10552–10561. 2786–2789.
[156] R.F. Egerton, S. Lazar, M. Libera, Micron 43 (2012) 2–7. [208] M.M. Rippel, C.A. Paula Leite, F. Galembeck, Anal. Chem. 74 (2002) 2541–2546.
[157] K. Varlot, J.-M. Martin, C. Quet, Y. Kihn, Macromol. Symp. 119 (1997) 317–324. [209] K. Varlot, J.M. Martin, C. Quet, Polymer 41 (2000) 4599–4605.
[158] S. Yakovlev, M. Libera, Micron 39 (2008) 734–740. [210] S. Venkatesan, J. Chen, E.C. Ngo, A. Dubey, D. Khatiwada, C. Zhang, Q. Qiao,
[159] R. Pal, A.K. Sikder, K. Saito, A.M. Funston, J.R. Bellare, Polym. Chem. 8 (2017) Nano Energy 12 (2015) 457–467.
6927–6937. [211] K. Zhao, O. Wodo, D. Ren, H.U. Khan, M.R. Niazi, H. Hu, M. Abdelsamie, R. Li,
[160] K. Siangchaew, M. Libera, Microsc. Microanal. 3 (1997) 530–539. E.Q. Li, L. Yu, B. Yan, M.M. Payne, J. Smith, J.E. Anthony, T.D. Anthopoulos,
[161] K. Siangchaew, D. Arayasantiparb, M. Libera, Mrs Proc. 461 (1996). S.T. Thoroddsen, B. Ganapathysubramanian, A. Amassian, Adv. Funct. Mater. 26
[162] K. Siangchaew, M. Libera, Microsc. Microanal. 3 (2003) 530–539. (2016) 1737–1746.
[163] J.A. Alexander, F.J. Scheltens, L.F. Drummy, M.F. Durstock, J.B. Gilchrist, [212] M.M. Rippel, C.A.P. Leite, L.-T. Lee, F. Galembeck, Colloid Polym. Sci. 283 (2005)
S. Heutz, D.W. McComb, J. Mater. Chem. A 4 (2016) 13636. 570–574.
[164] M.L.K. Siangchaew, Philos. Mag. A 80 (2000) 1001–1016. [213] F.I. Allen, M. Watanabe, Z. Lee, N.P. Balsara, A.M. Minor, Ultramicroscopy 111
[165] A.D.A.K. Zino, J.W.A. Leijten, With Gijsbertus de, Heiner Friedrich, J. Phys. Chem. (2011) 239–244.
C 121 (2017) 10552–10561. [214] D. Leman, M.A. Kelly, S. Ness, S. Engmann, A. Herzing, C. Snyder, H.W. Ro,
[166] L. Kovarik, A. Stevens, A. Liyu, N.D. Browning, Appl. Phys. Lett. 109 (2016) R.J. Kline, D.M. DeLongchamp, L.J. Richter, Macromolecules 48 (2015) 383–392.
164102. [215] M.R. Hammond, R.J. Kline, A.A. Herzing, L.J. Richter, D.S. Germack, H. Ro,
[167] V. Intaraprasonk, H.L. Xin, D.A. Muller, Ultramicroscopy 108 (2008) 1454–1466. C.L. Soles, D.A. Fischer, T. Xu, L. Yu, M.F. Toney, D.M. DeLongchamp, ACS Nano 5
[168] M. Brinkmann, P. Rannou, Macromolecules 42 (2009) 1125–1130. (2011) 8248–8257.
[169] G.A. Bassett, A. Keller, KolloidZ 231 (1969) 386. [216] J.M. Tarascon, M. Armand, Nature 414 (2001) 359–367.
[170] D.C. Martin, E.L. Thomas, Polymer 36 (1995) 1743–1759. [217] W.H. Meyer, Adv. Mater. 10 (1998) 439–448.
[171] L.F. Drummy, R.J. Davis, D.L. Moore, M. Durstock, R.A. Vaia, J.W.P. Hsu, Chem. [218] M. Pfannmöller, H. Flügge, G. Benner, I. Wacker, C. Sommer, M. Hanselmann,
Mater. 23 (2011) 907–912. S. Schmale, H. Schmidt, F.A. Hamprecht, T. Rabe, W. Kowalsky, R.R. Schröder,
[172] X. Ma, J. Liu, C. Ni, D.C. Martin, D.B. Chase, J.F. Rabolt, ACS Macro Lett. 1 (2012) Nano Lett. 11 (2011) 3099–3107.
428–431. [219] J.A. Amonoo, A. Li, G.E. Purdum, M.E. Sykes, B. Huang, E.F. Palermo, A.J. McNeil,
[173] D. Lolla, J. Gorse, C. Kisielowski, J. Miao, P.L. Taylor, G.G. Chase, D.H. Reneker, M. Shtein, Y. Loo, P.F. Green, J. Mater. Chem. A 3 (2015) 20174–20184.
Nanoscale 8 (2016) 120–128. [220] S. Venkatesan, N. Adhikari, J. Chen, E.C. Ngo, A. Dubey, D.W. Galipeau, Q. Qiao,
[174] A.J. Lovinger, Macromolecules 18 (1985) 910–918. Nanoscale 6 (2014) 1011–1019.
[175] R.S. Ruskin, Z. Yu, N. Grigorieff, J. Struct. Biol. 184 (2013) 385–393. [221] J. You, L. Dou, K. Yoshimura, T. Kato, K. Ohya, T. Moriarty, K. Emery, C. Chen,
[176] E.A. Stach, J. Li, H. Xin, D. Zakharov, Y.H. Kwon, E. Reichmanis, Microsc. J. Gao, G. Li, Y. Yang 4 (2013) 1446.
Microanal. 22 (2016) 1924–1925. [222] A.A. Herzing, H.W. Ro, C.L. Soles, D.M. DeLongchamp, ACS Nano 7 (2013)
[177] C. Ophus, P. Ercius, M. Sarahan, C. Czarnik, J. Ciston, Microsc. Microanal. 20 7937–7944.
(2014) 62–63. [223] J.L. Hart, A.C. Lang, A.C. Leff, P. Longo, C. Trevor, R.D. Twesten, M.L. Taheri, Sci.
[178] Y. Jiang, Z. Chen, Y. Han, P. Deb, H. Gao, S. Xie, P. Purohit, M.W. Tate, J. Park, Rep. 7 (2017) 8243.
S.M. Gruner, V. Elser, D.A. Muller, Nature 559 (2018) 343. [224] R. Ramachandra, J.C. Bouwer, M.R. Mackey, E. Bushong, S.T. Peltier, N.H. Xuong,
[179] O. Panova, X.C. Chen, K.C. Bustillo, C. Ophus, M.P. Bhatt, N. Balsara, A.M. Minor, M.H. Ellisman, Microsc. Microanal. 20 (2014) 706–714.
Micron 88 (2016) 30–36. [225] D.L. Handlin, E.L. Thomas, Macromolecules 16 (1983) 1514–1525.
[180] O. Panova, C. Ophus, C.J. Takacs, K.C. Bustillo, L. Balhorn, A. Salleo, N. Balsara, [226] M. Tosaka, R. Danev, K. Nagayama, Macromolecules 38 (2005) 7884–7886.
A.M. Minor, Nat. Mater. 18 (2019) 860–865. [227] Y. Konyuba, H. Iijima, P. Donnadieu, T. Higuchi, H. Jinnai, N. Hosogi, European
[181] X. Mu, A. Mazilkin, C. Sprau, A. Colsmann, C. Kubel, Microscopy (2019) 301–309. Microscopy Congress 2016: Proceedings (2016) 265–266.
[182] B.S.S. Pokuri, J. Stimes, K. O’Hara, M.L. Chabinyc, B. Ganapathysubramanian, [228] C. Ophus, J. Ciston, J. Pierce, T.R. Harvey, J. Chess, B.J. McMorran, C. Czarnik,
Comput. Mater. Sci. 163 (2019) 1–10. H.H. Rose, P. Ercius, Nat. Commun. 7 (2016) 10719.
[183] K. O’Hara, C.J. Takacs, S. Liu, F. Cruciani, P. Beaujuge, C.J. Hawker, [229] F.I. Allen, L.R. Comolli, A. Kusoglu, M.A. Modestino, A.M. Minor, A.Z. Weber, ACS
M.L. Chabinyc, Macromolecules 52 (2019) 2853–2862. Macro Lett. 4 (2014) 1–5.
[184] K.J. Batenburg, S. Bals, J. Sijbers, C. Kubel, P.A. Midgley, J.C. Hernandez, [230] T.E. Culp, Y. Shen, M. Geitner, M. Paul, A. Roy, M. Behr, R. S, G. J, M. Kumar,
U. Kaiser, E.R. Encina, E.A. Coronado, G. Van Tendeloo, Ultramicroscopy 109 E. Gomez, PNAS 115 (2018) 8694–8699.
(2009) 730–740. [231] M.J. Wirix, P.H. Bomans, H. Friedrich, N.A. Sommerdijk, G. de With, Nano Lett. 14
[185] K.J. Batenburg, J. Sijbers, IEEE IV (2007) 133–136. (2014) 2033–2038.
[186] K.J. Batenburg, J. Sijbers, IDDD Trans. Image Process. 20 (2011) 2542–2553. [232] H. Jinnai, R.J. Spontak, Polymer 50 (2009) 1067–1087.
[187] J.D. Roehling, K.J. Batenburg, F.B. Swain, A.J. Moulé, I. Arslan, Adv. Funct. Mater. [233] H. Jinnai, R.J. Spontak, T. Nishi, Macromolecules 43 (2010) 1675–1688.
23 (2013) 2115–2122. [234] H. Jinnai, T. Higuchi, X.D. Zhuge, A. Kumamoto, K.J. Batenburg, Y. Ikuhara, Acc.
[188] M.A. Touve, D.B. Wright, C. Mu, H. Sun, C. Park, N.C. Gianneschi, Macromolecules Chem. Res. 50 (2017) 1293–1302.
(2019). [235] H. Jinnai, X. Jiang, Curr. Opin. Solid State Mater. Sci. 17 (2013) 135–142.
[189] S. Juhl, X. Li, J. Badding, N. Alem, Microsc. Microanal. 22 (2016) 1840–1841. [236] H. Jinnai, H. Morita, K. Niihara, Kobunshi Ronbunshu 65 (2008) 547–561.

29
B. Kuei, et al. Materials Science & Engineering R 139 (2020) 100516

[237] A. Doerr, Nat. Methods 13 (2015) 23-23. [266] J.U. Starke, R. Godehardt, G.H. Michler, C.B. Bucknall, J. Mater. Sci. 32 (1997)
[238] X. Jiang, D.R. Greer, J. Kundu, C. Ophus, A.M. Minor, D. Prendergast, 1855–1860.
R.N. Zuckermann, N.P. Balsara, K.H. Downing, Macromolecules 51 (2018) [267] J. Kim, A. Nizami, Y. Hwangbo, B. Jang, H. Lee, C. Woo, S. Hyun, T. Kim, Nat.
7794–7799. Commun. 4 (2013) 2520.
[239] X. Jiang, S. Xuan, J. Kundu, D. Prendergast, R.N. Zuckermann, N.P. Balsara, Soft [268] D. Rodriquez, J. Kim, S.E. Root, Z. Fei, P. Boufflet, M. Heeney, T. Kim, D. Lipomi,
Matter 15 (2019) 4723–4736. ACS Appl. Mater. Interfaces 9 (2017) 8855–8862.
[240] K.B. Borisenko, G. Moldovan, A.I. Kirkland, A. Wang, D. Van Dyck, F. Chen, J. [269] L. Li, W.H. de Jeu, Macromolecules 37 (2004) 5646–5652.
Phys. Conf. Ser. 371 (2012) 012057. [270] P.K. Agarwal, R.H. Somani, W. Weng, A. Mehta, L. Yang, S. Ran, L. Liu, B.S. Hsiao,
[241] D. Shi, B.L. Nannenga, M.G. Iadanza, T. Gonen, Elife 2 (2013) e01345. Macromolecules 36 (2003) 5226–5235.
[242] B.L. Nannenga, D. Shi, A.G. Leslie, T. Gonen, Nat. Methods 11 (2014) 927–930. [271] H. Zheng, S.A. Claridge, A.M. Minor, A.P. Alivisatos, U. Dahmen, Nano Lett. 9
[243] B.L. Nannenga, T. Gonen, Curr. Opin. Struct. Biol. 40 (2016) 128–135. (2009) 2460–2465.
[244] D. Pashley, Adv. Phys. 5 (1956) 173–240. [272] N. De Jonge, F.M. Ross, Nat. Nanotechnol. 6 (2011) 695–704.
[245] K. Easterling, A. Thölen, Met. Sci. J. 4 (1970) 130–135. [273] S. Thiberge, A. Nechushtan, D. Sprinzak, O. Gileadi, V. Behar, O. Zik, Y. Chowers,
[246] L. Fei, S. Lei, W. Zhang, W. Lu, Z. Lin, C. Lam, Y. Chai, Y. Wang, Nat. Commun. 7 S. Michaeli, J. Schlessinger, E. Moses, Proc. Natl. Acad. Sci. U. S. A. 101 (2004)
(2016). 3346–3351.
[247] L. Wang, J. Teng, P. Liu, A. Hirata, E. Ma, Z. Zhang, M. Chen, X. Han, Nat. [274] M. Williamson, R. Tromp, P. Vereecken, R. Hull, F. Ross, Nat. Mater. 2 (2003)
Commun. 5 (2014). 532–536.
[248] G. Divitini, S. Cacovich, F. Matteocci, L. Cina, A. Di Carlo, C. Ducati, Nat. Energy 1 [275] H. Zheng, R.K. Smith, Y. Jun, C. Kisielowski, U. Dahmen, A.P. Alivisatos, Science
(2016) 15012. 324 (2009) 1309–1312.
[249] T.D. Anthopoulos, C. Tanase, S. Setayesh, E.J. Meijer, J.C. Hummelen, [276] H.M. Zheng, Y.S. Meng, Y.M. Zhu, MRS Bull. 40 (2015) 12–18.
P.W.M. Blom, D.M. de Leeuw, Adv. Mater. 16 (2004) 2174-+. [277] Z.Y. Zeng, W.I. Liang, H.G. Liao, H.L.L. Xin, Y.H. Chu, H.M. Zheng, Nano Lett. 14
[250] H. Saka, T. Kamino, S. Ara, K. Sasaki, MRS Bull. 33 (2008) 93–100. (2014) 1745–1750.
[251] S. Simões, R. Calinas, M. Vieira, M. Vieira, P. Ferreira, Nanotechnology 21 (2010) [278] H.G. Liao, K.Y. Niu, H.M. Zheng, Chem. Commun. 49 (2013) 11720–11727.
145701. [279] J. Park, H.M. Zheng, W.C. Lee, P.L. Geissler, E. Rabani, A.P. Alivisatos, ACS Nano 6
[252] M.A. Asoro, D. Kovar, P.J. Ferreira, ACS Nano 7 (2013) 7844–7852. (2012) 2078–2085.
[253] A. Kröger, S. Dziaszyk, J. Frenzel, C. Somsen, A. Dlouhy, G. Eggeler, Mater. Sci. [280] H.G. Liao, L.K. Cui, S. Whitelam, H.M. Zheng, Science 336 (2012) 1011–1014.
Eng. A 481 (2008) 452–456. [281] K. Jungjohann, S. Bliznakov, P. Sutter, E. Stach, E. Sutter, Nano Lett. 13 (2013)
[254] M. De Graef, M.A. Willard, M.E. McHenry, Y. Zhu, IEEE Trans. Magn. 37 (2001) 2964–2970.
2663–2665. [282] H. Liao, H. Zheng, Annu. Rev. Phys. Chem. 67 (2016) 719–747.
[255] J.S. Kim, T. LaGrange, B.W. Reed, M.L. Taheri, M.R. Armstrong, W.E. King, [283] V. Cleave, G. Yahioglu, P. Le Barny, D.H. Hwang, A.B. Holmes, R.H. Friend,
N.D. Browning, G.H. Campbell, Science 321 (2008) 1472–1475. N. Tessler, Adv. Mater. 13 (2001) 44-+.
[256] E.A. Stach, T. Freeman, A.M. Minor, D.K. Owen, J. Cumings, M.A. Wall, [284] K.H. Nagamanasa, H. Wang, S. Granick, Adv. Mater. 29 (2017).
T. Chraska, R. Hull, J. Morris, A. Zettl, Microsc. Microanal. 7 (2001) 507–517. [285] L.R. Parent, E. Bakalis, A. Ramirez-Hernandez, J.K. Kammeyer, C. Park, J. de
[257] M. Bobji, C. Ramanujan, J. Pethica, B. Inkson, Meas. Sci. Technol. 17 (2006) 1324. Pablo, F. Zerbetto, J.P. Patterson, N.C. Gianneschi, J. Am. Chem. Soc. 139 (2017)
[258] R.A. Bernal, R. Ramachandramoorthy, H.D. Espinosa, Ultramicroscopy 156 (2015) 17140–17151.
23–28. [286] P. Gao, Z. Wang, W. Fu, Z. Liao, K. Liu, W. Wang, X. Bai, E. Wang, Micron 41
[259] H.D. Espinosa, Y. Zhu, N. Moldovan, J. Microelectromechanical Syst. 16 (2007) (2010) 301–305.
1219–1231. [287] J. Kling, X. Tan, W. Jo, H. Kleebe, H. Fuess, J. Rodel, J. Am. Ceram. Soc. 93 (2010)
[260] G. Casillas, U. Santiago, H. Barrón, D. Alducin, A. Ponce, M. José-Yacamán, J. 2452–2455.
Phys. Chem. C 119 (2014) 710–715. [288] Z.L. Wang, R.P. Gao, W.A. de Heer, P. Poncharal, Appl. Phys. Lett. 80 (2002).
[261] J.P. Oviedo, S. KC, N. Lu, J. Wang, K. Cho, R.M. Wallace, M.J. Kim, ACS Nano 9 [289] Q. Jeangros, M. Duchamp, J. Werner, M. Kruth, R.E. Dunin-Borkowski, B. Niesen,
(2014) 1543–1551. C. Ballif, A. Hessler-Wyser, Nano Lett. 16 (2016) 7013–7018.
[262] A.A. Tedstone, D.J. Lewis, R. Hao, S. Mao, P. Bellon, R.S. Averback, C.P. Warrens, [290] F.M. Ross, Science 350 (2015) aaa9886.
K.R. West, P. Howard, S. Gaemers, ACS Appl. Mater. Interfaces 7 (2015) [291] J.Y. Huang, L. Zhong, C.M. Wang, J.P. Sullivan, W. Xu, L.Q. Zhang, S.X. Mao,
20829–20834. N.S. Hudak, X.H. Liu, A. Subramanian, Science 330 (2010) 1515–1520.
[263] Z. Su, Y. Miao, S. Mao, G. Zhang, S. Dillon, J.T. Miller, K.S. Suslick, J. Am. Chem. [292] Á.K. Kiss, E.F. Rauch, J.L. Lábár, Ultramicroscopy 163 (2016) 31–37.
Soc. 137 (2015) 1750–1753. [293] M. Sun, H. Liao, K. Niu, H. Zheng, Sci. Rep. 3 (2013).
[264] A.M. Beese, D. Papkov, L. S, Y. Dzenis, H.D. Espinosa, Carbon 60 (2013) 246–253. [294] J. Liu, B. Wei, J.D. Sloppy, L. Ouyang, C. Ni, D.C. Martin, ACS Macro Lett. 4 (2015)
[265] A.M. Minor, S.A.S. Asif, Z.W. Shan, E.A. Stach, E. Cyrankowski, T.J. Wyrobek, 897–900.
O.L. Warren, Nat. Mater. 5 (2006) 697–702.

30

You might also like