Professional Documents
Culture Documents
Rsos.191204 LCTEM
Rsos.191204 LCTEM
royalsocietypublishing.org/journal/rsos
electron microscopy and
its applications
Shengda Pu†, Chen Gong† and Alex W. Robertson
Review
Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, UK
Cite this article: Pu S, Gong C, Robertson AW.
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
AWR, 0000-0002-9521-6482
2020 Liquid cell transmission electron microscopy
and its applications. R. Soc. open sci. 7: 191204. Transmission electron microscopy (TEM) has long been an
http://dx.doi.org/10.1098/rsos.191204 essential tool for understanding the structure of materials. Over
the past couple of decades, this venerable technique has
Received: 10 July 2019 undergone a number of revolutions, such as the development
Accepted: 19 November 2019 of aberration correction for atomic level imaging, the realization
of cryogenic TEM for imaging biological specimens, and new
Subject Category: instrumentation permitting the observation of dynamic systems
Chemistry in situ. Research in the latter has rapidly accelerated in recent
years, based on a silicon-chip architecture that permits a
Subject Areas: versatile array of experiments to be performed under the high
materials science/electron microscopy/ vacuum of the TEM. Of particular interest is using these silicon
electron microscopy chips to enclose fluids safely inside the TEM, allowing us to
observe liquid dynamics at the nanoscale. In situ imaging of
Keywords: liquid phase reactions under TEM can greatly enhance our
understanding of fundamental processes in fields from
transmission electron microscopy,
electrochemistry to cell biology. Here, we review how in situ
in situ transmission electron microscopy,
TEM experiments of liquids can be performed, with a
liquid cell transmission electron microscopy particular focus on microchip-encapsulated liquid cell TEM.
We will cover the basics of the technique, and its strengths and
Author for correspondence: weaknesses with respect to related in situ TEM methods for
Alex W. Robertson characterizing liquid systems. We will show how this technique
e-mail: [email protected] has provided unique insights into nanomaterial synthesis and
manipulation, battery science and biological cells. A discussion
on the main challenges of the technique, and potential means
to mitigate and overcome them, will also be presented.
© 2020 The Authors. Published by the Royal Society under the terms of the Creative
Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits
unrestricted use, provided the original author and source are credited.
Liquid cell TEM allows the observation of processes that cannot be imaged with conventional TEM or 2
other techniques. For example, biological specimens are destroyed within seconds in conventional TEM due
royalsocietypublishing.org/journal/rsos
to the high vacuum. And while cryo-electron microscopy can be used to safely image and determine their
structures [18–20], the specimens are no longer in their native state, with the potential for specimen damage
and artefacts being introduced during sample freezing. Liquid cell TEM not only preserves the liquid state
of the specimens inside the TEM vacuum, but also allows the in situ observation of biological process [9].
In situ TEM has become a popular technique for observing reactions at the nanoscale [21]. Compared
with other in situ techniques like atomic force microscopy (AFM), scanning tunnelling microscopy (STM)
and various X-ray methods, TEM not only has the advantage of having both high temporal and spatial
resolution, it also provides direct visualization of any changes in structural, morphological [22] or
elemental distribution [23,24] at the nanoscale. The development of enclosed liquid cells has allowed
the in situ TEM imaging of liquid phase reactions, opening up fields from electrochemistry to cell
biology for study. This paper reviews in situ TEM experiments of liquid samples, with a focus on
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
microchip encapsulated liquid cell TEM. We will cover the basics of the technique, its strengths and
royalsocietypublishing.org/journal/rsos
open cell closed cell
(iv)
(i) (iii)
(viii)
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
(vii)
(vi)
Figure 1. Comparison of different techniques for liquid phase TEM imaging. (i) Ionic liquid films maintained inside micrometre-sized
holes from surface tension, suitable for TEM imaging [26]. (ii) Design schematic of a miniature battery that may be imaged inside TEM,
using an ionic-liquid-based electrolyte [27]. (iii) Schematic of differentially pumped TEM column with three different pumping stages,
indicated by different coloured pipes, permitting a high pressure sample chamber [28]. (iv) Schematic of an in situ liquid TEM stage on
the tip of a TEM holder [29]. (v) Schematic of an assembled liquid cell from two silicon/silicon nitride chips [23]. (vi) Silicon chip with
microfabricated electrodes [14]. (vii) Magnified SEM image of the red box region in (v), showing electrodes partially over the SiNx viewing
window [14]. (viii) Schematic of a graphene liquid cell for TEM. The specimen is encapsulated between two graphene sheets [30].
experiments were successfully conducted within such systems [31,37], it is mainly used for gas phase
experiments [38–42] and biological specimens [7,43], which do not require very high pressure to maintain.
Another approach can be defined as the ‘closed environmental cell’. Unlike the ‘open environmental cell’
approach with differential pumps, the pressure of the closed cell is maintained by encapsulating the sample
volume in screening membranes. Such a specimen can be maintained at a much higher pressure. The
electron beam penetrates and images the specimen through robust and electron-transparent windows on the
top and bottom of the encapsulated specimen cell. The cell can be in-built within the TEM specimen holder
and is therefore compatible with conventional TEMs. A closed environmental cell can be used for both gas
phase [44] and liquid phase imaging [1,3]. When used for liquid phase, it is usually called a ‘liquid cell’.
A liquid cell is typically only hundreds of nanometres thick, confined by two thin but robust electron-
transparent membranes. The commercially available and by-far most widely adopted liquid cell system is
(i) (ii) 4
royalsocietypublishing.org/journal/rsos
Si substrate
deposit electrodes
flow liquid
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
deposit spacer
back-etching
Figure 2. (i) Simplified schematics of the production process of a liquid cell microchip via photolithography. (ii) Schematics of the
assembly of the microchips for liquid cell TEM imaging.
based on silicon microchips, as shown in figure 1(iv–vii). A thin layer of liquid phase specimen of
between tens of nanometres to a micrometre can be confined between two microchips using O-rings
and spacer materials. A thin layer of amorphous silicon nitride (SiN) is most commonly used as the
membrane window material, through which the electron beam can penetrate and be used to image.
(Detailed fabrication and application processes of this system is shown in the next section.) Such
liquid cells can safely contain almost all types of liquid, independent of vapour pressure. In addition,
microfabrication allows additional capabilities to be added, including allowing liquid flow [45],
temperature control (heating and cooling [46]) and applying electrical bias using patterned electrodes.
Avariant of the liquid cell is the ‘graphene liquid cell’, which has been used since 2012. Microencapsulation
using graphene [10,47], or 3 nm thick amorphous carbon films [5], have been shown to well preserve volatile
specimens like biological cells and liquid crystals within the ultra-high vacuum in TEM. This technique can also
be used to image liquid specimens [48–50]. Compared with the microchip cell, the graphene cell can normally
achieve better resolution due to less scattering from the window material and smaller liquid thicknesses. In
addition, it does not require an expensive dedicated sample holder [51–55]. However, this technique is
currently limited in several ways. First, encapsulation of a liquid sample with graphene requires very careful
cell assembly due to the difficulties in handling graphene. Second, only a very small amount of liquid can
be encapsulated and imaged, less than 0.01 picolitre, compared with the nanolitre capacity of a microchip
liquid cell. Third, unlike with microchips, microfabrication cannot be used in this cell, which means that it is
not compatible with functions like liquid circulation or applying electrical biasing. The lack of liquid flow,
combined with the low volume of liquid between the graphene membranes, means studying dynamic
processes is hindered by the limited availability of reactants. This review focuses on the microchip-based
liquid cell, with a detailed discussion of the graphene liquid cell beyond the scope of this review.
royalsocietypublishing.org/journal/rsos
silicon oxide [56] and amorphous carbon [9]. Electrodes and spacers may then be deposited. The spacer is
used to provide enough space between the two chips to ensure enough liquid remains in the cell during
imaging and to allow for liquid flow. Electrodes can be employed to apply an electric potential to
manipulate particle movement within the liquid or to induce electrochemical reactions during imaging.
Finally, the Si is etched away from the back, leaving a small window of only the SiN membrane. The
window is usually in a rectangular shape, as shown in figure 1 (vi,vii), with widths of tens of micrometres
and lengths of over a hundred micrometres. While this limits the viewing area, wider windows are more
susceptible to breaking and bowing, common issues associated with the liquid cell, which will be
discussed in detail later. The electron beam is able to transmit through the windows to allow imaging.
A comprehensive description of the production process of a liquid cell was given by Grogan & Bau [57].
The assembly of a liquid cell is shown in figure 2(ii). A second, flipped, chip is aligned on top of the
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
bottom chip, normally fixed in place by clamping tight with O-rings and metal frames on the holder tip
As we can see from both equations (3.1) and (3.2), thicker liquid means worse resolution. Another
critical factor limiting the resolution is the dose rate. For both TEM and STEM, the resolution is
dependent on the dose rate as follows [60]:
Dose rate should be carefully monitored and controlled during imaging to avoid beam damage, as
will be discussed in detail later. A comprehensive study of all factors affecting liquid cell resolution in
TEM and STEM was done by de Jonge et al. [60].
royalsocietypublishing.org/journal/rsos
(iv)
11.2°
viewing
direction
–
(110)
p
h
– –
(121) 23.3°
(v)
(i) (ii) (iii) –
(211)
––
(112)
(b)
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
(ii) (iv)
(i)
(c)
Li+
conducting epoxy SnO2 nanowires
LiCoO2
200 nm
(ii)
Au rod IL Au rod
(206)
(201)
–– (110)
(110)
(006)
– +
(i) potentiostat (101)
(011) (000)
(200)
(211)
5 nm
(iii) (11–1) (iv) (v) (vi) (301)
(vii)
Figure 3. (a) Growth of saw-tooth faceting Si nanowires using liquid Au-Si droplet as the catalyst [2]. (i–iii) Three Si wires of different
width grown from Au-Si droplets of three different sizes. Scale bar is 100 nm. (iv) Schematic of the three-dimensional structure of the
nanowire. (v) The surface structure with facet angles indicated. p and h denote sawtooth period and amplitude. Scale bar is 50 nm.
(b) Crystalline ordering at the surface of Au-Si liquid [61]. (i) A crystalline bilayer visualized on a 50 nm diameter droplet at 350°C.
(ii) Surface ordering on a 12 nm diameter droplet at 350°C. (iii) A solid Au particle heated on Si at 360°C where the eutectic liquid
starts to cover its surface. Scale bars are 2 nm for (i–iii). (iv) The surface ordering evolution at different temperature. (c) Lithiation of a
single SnO2 nanowire anode at −3.5 V against a LiCoO2 cathode [34]. (i) Schematic of the experimental set-up. (ii) The reacted
(‘amorphous’) and non-reacted (‘single-crystal SnO2’) wire separated by the lithiation interface ‘dislocation cloud’. (iii–vi) Diffraction
patterns taken from different sections of the nanowire in (ii) showing the amorphization upon lithiation. (vii) Sn nanoparticles
dispersed in an amorphous matrix from the lithiated part of the wire.
illustrates how in situ TEM is able to provide unique insights into fundamental materials processes,
which in this case was the relationship between the surface energy and morphology of the
nanomaterial, and how to manipulate the growth of nanomaterials to achieve accurate and direct
nanowire growth in devices [63]. Since then, this system was widely adopted to study in situ the
three-phase growth (solid formed from gas, catalysed by a liquid alloy) of different nanowires [64–66].
A significant advantage of the open-cell system is the high resolution achievable due to the high 7
vacuum and no interceding encapsulating layers. A good example of this is shown in figure 3b, which
royalsocietypublishing.org/journal/rsos
shows a stable atomically thin crystalline phase formed from liquid Au-Si and its changes with respect
to different temperatures [61]. The imaged solid–liquid interface atomic structure is of interest to trace
the individual atoms during solidification/melting to help better understand and model these processes.
Ionic liquids are liquid salts consisting of ions and ion pairs [67], which can dissolve many kinds of
substances. They can withstand high vacuum and electron beam irradiation [68,69]. They are also good
ionic conductors, which makes them perfect for use as alternative electrolytes for battery and other
energy-related in situ TEM research. In 2010, the first experimental design for performing biasing
under high vacuum with an ionic liquid in a TEM was published [27]. Within the same year, with
this design, Huang et al. [34] studied the in situ lithiation of SnO2 nanowires within TEM, as shown in
figure 3c. (Lithiation/delithiation are the incorporation/removal process of lithium from an electrode
in a lithium ion battery.) Upon lithiation, the crystalline SnO2 nanowire was converted to amorphous
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
Li2O with nanocrystalline Sn and LixSn dispersed within it (as shown in figure 3c(vi)). The interface
Table 1. Selected list of liquid cell experiments on nanomaterials performed since 2003.
Cu cluster nucleation, growth Williamson et al. 2003 [1] biasing via electrodes first liquid cell experiment
Pt nanoparticle nucleation, growth Zheng et al. 2009 [4] E beam illumination sub-nm resolution allows clear visualization of growth mechanism
PbS nanoparticle nucleation, growth Evans et al. 2011 [6] E beam illumination different morphology from different solution; first atomic resolution
in liquid cell
Ag nanoparticle nucleation, growth Woehl et al. 2012 [80] E beam illumination different morphology formed from different beam current.
iron oxyhydroxide oriented attachment Li et al. 2012 [81] premade in situ crystal movement and attachment with atomic resolution
nanoparticles
Pt3Fe nanorod nucleation, growth and Liao et al. 2012 [82] E beam illumination first in situ nanorod formation
assembly
Au-Pd core–shell growth of the shell onto Jungjohann et al. 2013 E beam illumination first in situ core–shell synthesis by coating the core with shell
nanoparticle the core [83] material
Pt nanoparticle nucleation, growth Liao et al. 2014 [84] E beam illumination first in situ crystal growth with atomic resolution
Pd nanoparticle dissolution Jiang et al. 2014 [85] oxidative etching by E beam first in situ oxidative etching of nanoparticle
Ag-Pd nanocage formation Sutter et al. 2014 [86] galvanic replacement of Ag using a Pd first in situ formation of nanocage
solution
Au nanoparticle dissolve, merge Hermannsdörfer et al. premade particles change pH, NaCl in situ change of particle with respect to environment
2015 [87] concentration, beam dose
Au nanoparticle nucleation, growth Park et al. 2015 [88] E beam illumination quantify the relationship between beam dose with growth rate
Au nanoparticle nucleation, growth Alloyeau et al. 2015 [89] E beam illumination
Fe3Pt–Fe2O3 core–shell nucleation and growth Liang et al. 2015 [90] E beam illumination first in situ growth of the complete core–shell structure
nanoparticle
Pt-Au core–shell growth of the shell onto Wu et al. 2015 [91] E beam illumination quantify the growth kinetics of core–shell structure, showing the
nanoparticle the core three different stages during growth
(Continued.)
8
Table 1. (Continued.)
UiO-66(Zr) & ZIF-8 MOF nucleation, growth and Patterson et al. 2015 [92] mixing solutions and heating first liquid cell in situ experiment on MOF
assembly
Pd nanoparticle nucleation, growth, size Abellan et al. 2016 [93] E beam illumination precise size control growth to sub-3 nm using capping agent
control
Ag nanoparticle nucleation, growth Ge et al. 2017 [94] E beam illumination catalytic growth due to the presence of Pt
Ag particle capped with oriented attachment Zhu et al. 2018 [95] premade the role of capping ligands on the oriented attachment process
ligands
Pd nanoparticle nucleation, growth Yang et al. 2019 [96] electrochemical biasing significant effect of HCl on the growth mechanism and resultant
morphology
9
I (10–7 A)
royalsocietypublishing.org/journal/rsos
f
0
a f h j
cb
d
e
–5
–0.2 –0.1 0 0.1 0.2 0.3 200 nm
U (V)
(b) 4.0
(i) t (ii)
3.5
D (nm)
3.0
2.5 a: simple growth (iii)
2.0 b: with coalescence
0 20 40 60 80 100 120
t (s)
(c)
(i) (ii) (iii) (iv)
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
2 nm
139.5 s 145.5 s 797.5 s 797.5 s
2D projection
22 s 55 s 70 s 85 s 107 s 143 s
{111} {111}
100
d100
2 nm
d111
d011
010 (i) (iii)
(ii) 4.0 {111}
{011}
3.5 {100}
D (nm)
3.0
2.5
2.0
1.0
(iv)
0 20 40 60 80 100 120 140
time (s)
(e) (i) 0:00 0:39 1:40 2:34 3:46 9:32
10 nm
10 nm
Figure 4. (a) Images recorded during the electrochemical deposition and stripping process of copper; a, b, e were recorded during
deposition; f, h, j were recorded during stripping [97]. The cyclic voltammogram on the left-hand side shows the point in the cycle
at which each image was recorded. (b) Platinum nanocrystal growth trajectories. (i) Particle size versus growth time for these
particles with two different types of growth trajectories. (ii) Coalescence growth. (iii) Monomer addition growth (simple
growth) [3]. (c) Direction-specific oriented attachment of iron oxyhydroxide nanoparticles. (i–vii) Sequential dynamics of the
attachment process of two particles ( particle I and II), the surfaces of particles I and II made transient contact at many points
and orientations (1–1, 1–2, 2–3 and 3–4) before final attachment and growing together (point 3–5), scale bar 5 nm.
(viii) Attachment interface from (vii), showing an inclined twin plane [81]. (d ) In situ facet development of a Pt nanoparticle.
(i) The atomic model of a Pt nanoparticle and its projection along the [011] zone axis. (ii) Average growth profile along three
different directions (100), (011) and (111). (iii) Sequential growth of the Pt nanoparticle. (iv) Simulated TEM images of the Pt
nanoparticle in (iii). (e) [82]: Real-time imaging of Pt3Fe nanorod formation. (i) Nucleation and growth of Pt3Fe particles
followed by their assembly into a nanorod. Particles contributed to the nanorod are highlighted in green. (ii) The growth of a
long Pt3Fe nanorod from the assembly of several short nanorods [84].
By improving the liquid cell design and using intensive electron beam radiation instead as the means to
instigate nucleation and growth of nanoparticles, Zheng et al. [3] managed to improve the resolution of the
liquid cell from 5 nm to sub-nanometre. A high electron beam dose can lead to radiolysis of the liquid,
yielding radiolysis species like hydrated electrons, eh [100], which act as a strong reducing agent and
thus cause the reduction of the solution and the formation of the nanoparticles. Using this technique,
the nucleation and growth of nanosized Pt particles was clearly visualized. With this high-resolution
technique, Zheng et al. revealed two different growth mechanisms during the early stage of nanocrystal 11
growth (coalescence versus monomer attachment), as shown in figure 4b. In addition, by tracing the
royalsocietypublishing.org/journal/rsos
growth rate of individual particles, it was found that a monodisperse size distribution could always be
reached regardless of the initial growth path, as shown in figure 4b(i). Zheng et al.’s method of
instigating nanomaterials growth by beam illumination has been widely adopted to study nanomaterials
synthesis with liquid cell TEM. For example, using such method, Evans et al. [6] grew PbS nanoparticles
of different morphologies with precursor solutions of different chemical compositions. In the same
paper, Evans et al., for the first time, showed that lattice resolution could be achieved within liquid cell TEM.
Apart from nanoparticle nucleation and growth, many other processes can also be visualized in situ
with atomic resolution. For example, in 2012, the oriented attachment process of iron oxyhydroxide
nanoparticles was visualized by Li et al. [81], as shown in figure 4c. The nanocrystal kept moving and
rotating around another bigger particle until a point where lattice matching with the adjacent
nanocrystal was achieved. Then they merged into a single crystal. Such oriented attachment process
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
was later also visualized in Au nanoparticles capped with organic ligands by Zhu et al. [95], who
As shown before, applying bias and electron beam irradiation are the most common ways to grow
royalsocietypublishing.org/journal/rsos
nanostructures in liquid cell TEM, although temperature control can also be used to induce growth
and dissolution [107]. Regardless of the growth method, two key factors always dominate the growth
kinetics and the resultant material: precursor solution composition and electron beam radiation. Many
works have been performed to assess their effects.
Different precursor solutions can lead to different growth mechanisms. For example, Yang et al. [96]
showed that by adding HCl into the precursor solution, the Pd particles switched from three-
dimensional island growth to aggregative growth. Such visualization helped explain the change in the
resultant morphology (from smooth to porous and open morphology). Abellan et al. [93] visualized
how, by adding a capping agent tri-n-octylphosphine (TOP) to the solution, precise size control of the
grown Pd nanoparticle to sub-3 nm can be achieved. Ge et al. [94] showed anomalously fast growth rates
of Ag nanoparticles when the growth took place adjacent to Pt, providing experimental proof of the
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
royalsocietypublishing.org/journal/rsos
to the surface charge of the silicon nitride windows.
Particle motions due to various interactions can also be studied using liquid cell, including
nanoparticle–nanoparticle interactions [111–113], particle–beam interactions [114], particle–bubble
interaction [115] etc. For example, Park et al. [113] showed how Au nanoparticles may self-assemble
into a superlattice. By tracking the movement of individual nanoparticles with respect to time, he was
able to quantify the driving force for the process [116]: strong long-range anisotropic forces first drive
the formation of chains from particles, then these chains fold to form the closed-pack superlattice due
to close-range van der Waals forces. Using the same technique, processes like nanomaterial growth by
particle attachment [81,117] can also be tracked and better understood.
royalsocietypublishing.org/journal/rsos
0
(v) (vi) (vii) (viii) (x) b0
1 mm 1 mm
–
(b) cyclic voltammetry Li deposition Li dissolution
100
80 1st cycle BF STEM <4 S 4:20 min 7:00 min
60
40
J/mA cm–2
20 Li dissolution
0
–20
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
–40 Li deposition
–60 1 mm
20
0
–20
–40 Li deposition
–60 2 mm
–80
–100
100
80 3rd cycle BF STEM <4 S 5:00 min 7:00 min
60
40
J/mA cm–2
20 Li dissolution
0
–20
–40 Li deposition
–60
–80
2 mm
–100
–4 –3 –2 –1 0
E/V versus Pt
(c)
(i) (ii) (iii) 1.29 mm (iv) 1.93 mm
920 nm
120 nm 536 nm 714 nm
280 nm 440 nm
571 nm
400 nm 464 nm
2 mm 2 mm 2 mm 2 mm
(d) 18
normalized volume of
40 12
stripping of Li
10
0 8
deposition of Li
6
–40
4
2 5 mm
–80 5 mm
0
–4.0 –3.5 –3.0 –2.5 –2.0 –1.5 –1.0 –0.5 0 200 250 300 350 400
E versus Pt/V time (s)
0
(ii) (iii) (iv) (v) (vi) (vii) (viii)
Figure 5. (a) In situ study of lead dendrite plating and stripping from lead ions in aqueous solution [120]. Sequential dendrite plating
at −1.3 V (i–iv) and stripping at +0.3 V (iv–vii). (viii) Specimen surface after two cycles of plating and stripping. (ix–x) The change in Pb2+
concentration in the liquid visualized by subtracting two consecutive frames from the in situ movie, lighter or darker regions indicate where
the intensity increases or decreases. (b) In situ Li plating and stripping on Pt working electrode in LiPF6/PC [29]. The first, second and third
rows correspond to the first, second and third charge/discharge cycles, showing the cyclic voltammogram, electrode before plating, after
plating and after stripping for each cycle. (c) The interface between the Pt working electrode and the LiPF6/PC electrolyte after cycles 5–8,
the red line shows the outside of the bright contrast SEI layer while the blue line shows the outline of the Pt electrode [29]. (d) Effect of HF
formed from water as an additive to the Li plating and stripping process in LiPF6/PC electrolyte [122]. (i) Cyclic voltammograms of electrolyte
with 10 and 50 ppm water. (ii) Quantified total area of Li deposited and stripped for the electrolyte with 10 and 50 ppm of water. (iii) Li
deposited on Pt working electrode in electrolyte with 10 ppm water. (iv) Li deposited on Pt working electrode in electrolyte with 50 ppm
water. (e) Sequential evolution of a LiFePO4/FePO4 cluster during one cycle [123]. (i) The charge/discharge voltage profile. (ii–viii) 5 eV
EFTEM images, brighter particles indicated by the arrows are FePO4 particles formed from delithiation of LiFePO4 during charging. Scale bar
200 nm.
additive, rather than the typical dendrite morphology, a smooth and dense layer was deposited. In 15
addition, more Li was deposited and stripped during the cycle and a better coulombic efficiency was
achieved. This is due to the increased LiF concentration within the SEI, which helped sustain faster Li+
royalsocietypublishing.org/journal/rsos
cation diffusion through the highly conductive LiF channels. The effect of LiNO3 as an additive was
also studied [128], which has been shown to induce distinctive difference on the dendrite morphology.
Other battery systems beyond Li-ion can also be tested with liquid cell TEM. Mg plating and SEI
formation were visualized by Singh et al. [129] in 2018. The non-dendrite growth and the formation of
the highly functional SEI, which allows continuous deposition and dissolution of Mg metal, have
shown Mg as a promising candidate for future batteries. Liquid TEM provides us a means for the
direct visualization of time-resolved dendrite morphology change [130], allowing such processes to be
quantified [131] and for the growth mechanism to be better understood. This can help us find good
ways for morphology control to mitigate the dendrite problem in batteries, via tuning of the
electrolyte and electrode chemistry and electrode topography.
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
royalsocietypublishing.org/journal/rsos
7 ± 1 µm. Since then, combining Au-tagging and phase contrast in STEM mode became a popular
technique to identify and trace the in situ changes of specific proteins within live biological cells in
liquid state [138]. For example, in 2011, using such a technique, Peckys et al. [11] were able to identify
nanoparticles around vesicles which helped trace and quantify vesicles’ accumulation process.
Another advantage of using liquid cell for biological specimen imaging is that cells can be easily
grown onto the SiN membrane. And by labelling certain proteins using both affibody and quantum
dots, the same cell can be imaged with light microscopy [139], SEM and STEM [140] using a
technique called correlative fluorescence and electron microscopy [139,141,142]. This is demonstrated
in figure 6. HER2 within a SKBR3 cell were labelled with quantum dots, which can emit bright
fluorescence signals under conventional light microscopy. Quantum dots’ electron-dense core can also
be easily distinguished under STEM. In this case, the STEM was performed with an environmental
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
SEM with saturated water vapour atmosphere. However, the same chip is also compatible with
6.1. Resolution
Liquid cell TEM suffers from poor resolution mainly due to electron scattering from the SiN window and
the liquid layer. Liquid thicknesses of up to several micrometres can be imaged, but resolution will be
heavily compromised [8,11]. A quantitative study showing the correlation between the resolution and
the liquid layer thickness was done by de Jonge et al. [58].
Unnecessarily thick liquid can be formed due to bowing of the membrane, leading to reduced
resolution [11,57]. Such bowing is caused by the pressure difference between the inside and outside
the liquid cell. There are several possible ways to avoid bowing. Increasing the membrane thickness
can reduce bowing and make the window more robust, but resolution will decrease due to the
electron scattering with the membrane material itself. Another way is to incorporate micrometre-sized
pillars pinning the top and bottom window together [156]. However, this requires the manufacture of
top and bottom chips as a single piece, which is more expensive and prone to breaking. The most
popular way to reduce bowing is to simply reduce the size of the window.
A good way to mitigate electron scattering by the window is to use alternative membrane materials.
For example, graphene-membrane cells achieve better resolution than SiN [47–52]. Recently, MoS2 [157]
was also successfully used as the membrane material to see Pt nanoparticle growth. However, as
17
(a)
royalsocietypublishing.org/journal/rsos
nm
HER2-monomer
(extracellular domains)
4
~1
streptavidin-QDs
4 nm
~8 nm
(b)
(i) imaging buffer (ii) 3°C, 720–740 Pa saturated H2O vapour
electron
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
beam
63× oil
immersion scattered electrons (towards
objective dark-field STEM detector)
(iii)
silicon
microchip electron
beam
cell liquid
flow
SiN membrane
scattered
electrons
(c)
(i) (ii) (iii) (iv)
(v) (vi)
Figure 6. (a) Schematics of HER2 labelling with affibody-quantum dots [141]. (b) Schematics showing correlative microscopy of
whole cells in liquid cell. (i) Fluorescence imaging [141]. (ii) For STEM imaging within environmental SEM [141]. (iii) For STEM
imaging inside TEM [139]. (c) Correlative light and electron microscopic images of HER2-labelled SKBR3 cell [141].
(i) Fluorescence overview image showing several dozen cells. (ii) Higher-resolution fluorescence image from marked solid line
rectangle in (i). (iii) Even higher-resolution fluorescence image from selected area in (ii). (iv) STEM image of the same area as
(iii). (v) STEM image of the boxed region shown in (iii) and (iv). (vi) Automatically detected labels outlined in light green.
Scale bars, 100 µm (i), 20 µm (ii); 2 µm (iii and iv); 200 nm (v and vi).
Table 2. Different types of electron beam damage and their effects on liquid cell TEM (dotted lines indicate relatively 18
insignificant contribution).
royalsocietypublishing.org/journal/rsos
electron beam damage
mentioned, these membrane materials can only encapsulate a very small amount of liquid, and they are
not compatible with microfabrication and therefore functions like electrical biasing and liquid flow.
Using correlative techniques like AFM, traction force microscopy (TFM) [158] can also help obtain
more information and potentially better resolution.
royalsocietypublishing.org/journal/rsos
Heating is another type of beam damage due to inelastic scattering of electrons where part of their
kinetic energy was converted to thermal energy. The relatively large liquid volume within the liquid
cell and possible liquid flow system allows good heat exchange to prevent localized heating, which
makes heating a relatively insignificant beam damage mechanism in the liquid cell. However, this can
still contribute to the degradation of biological cells that are generally very sensitive to localized
heating and are normally characterized within an enclosed cell without liquid flow.
A bigger contributor to biological cell degradation is radiolysis due to the electron beam, which
destroys the cells [12,138] and induces structural change to biological materials [54]. Keeping a low
electron dose is critical to limit the cell damage during TEM imaging. A systematic damage study on
the cell by beam dose was performed by Hermannsdörfer et al. [162] in 2016. They found a beam
dose limit of around 102 electrons nm−2 should not be exceeded during imaging [59,162]. It is almost
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
impossible to have good contrast or resolution of organic biological features at such low dose [137].
6.3. Representativeness
Due to the limited solution volume and confined space, in addition to the beam damage mechanisms
discussed above, liquid cell TEM results’ representativeness is always in doubt. For example, studies of
battery electrode materials using liquid cell are currently limited. The electrode materials were either
assembled onto the microchips by picolitre drop casting [123] or by flowing a suspension [133]. Not only
do these methods have a low success rate, the amount of material deposited onto the chip is too small,
which makes such systems far from being representative of their real battery counterparts. To improve its
representativeness, a larger amount of material with better attachment to the electrode should be achieved
by either nanoscale thin-film patterning [153,154,171] or by focused ion beam (FIB) and nanoscale
welding [172]. Membrane materials can also cause problems, making liquid cell TEM less representative.
Donev et al. [173] showed that for electron-beam-induced deposition, the membrane material can have a great 20
influence on the morphology and size of the nanoparticles formed. For the Cu deposition and stripping
royalsocietypublishing.org/journal/rsos
experiment [1], the cyclic voltammogram in liquid cell is quite different from that performed in bulk
solution. This is due to the hindered diffusion of liquid and particle within the liquid cell. Such hindered
diffusion can also contribute to the unexpected size and number distribution of the deposited material
[99]. Nanoparticle diffusion in liquid cell was reported to be up to seven to nine orders slower than
expected for a bulk solution [155]. This slowing is due to increase in liquid viscosity (by five orders of
magnitude) [110] caused by the formation of an ordered liquid layer due to the surface charge [174] of the
silicon nitride windows. A solution for this problem is to either find a way to cancel the charge on the SiN
membrane or to use a different membrane material that does not interact or charge as easily.
Apart from the three main challenges mentioned above, there are other common issues like trapped
References
1. Williamson MJ, Tromp RM, Vereecken PM, Hull electron microscopy and their application to 9. Daulton TL, Little BJ, Lowe K, Jones-Meehan J.
R, Ross FM. 2003 Dynamic microscopy of dynamical observation of liquid crystals in 2001 In situ environmental cell-transmission
nanoscale cluster growth at the solid–liquid transmission electron microscopy. electron microscopy study of microbial
interface. Nat. Mater. 2, 532–536. (doi:10.1038/ Nanotechnology 15, S329–S332. (doi:10.1088/ reduction of chromium(VI) using electron
nmat944) 0957-4484/15/6/001) energy loss spectroscopy. Microsc.
2. Ross FM, Tersoff J, Reuter MC. 2005 Sawtooth 6. Evans JE, Jungjohann KL, Browning ND, Microanal. 7, 470–485. (doi:10.1017.
faceting in silicon nanowires. Phys. Rev. Lett. 95, Arslan I. 2011 Controlled growth of nanoparticles S1431927601010467)
146104. (doi:10.1103/PhysRevLett.95.146104) from solution with in situ liquid transmission 10. Mohanty N, Fahrenholtz M, Nagaraja A, Boyle D,
3. Zheng H, Smith RK, Jun Y-W, Kisielowski C, electron microscopy. Nano Lett. 11, Berry V. 2011 Impermeable graphenic
Dahmen U, Alivisatos AP. 2009 Observation of 2809–2813. (doi:10.1021/nl201166k) encasement of bacteria. Nano Lett. 11,
single colloidal platinum nanocrystal growth 7. Parsons DF, Matricardi VR, Moretz RC, Turner JN. 1270–1275. (doi:10.1021/nl104292k)
trajectories. Science 324, 1309–1312. (doi:10. 1974 Electron microscopy and diffraction of wet 11. Peckys DB, De Jonge N. 2011 Visualizing gold
1126/science.1172104) unstained and unfixed biological objects. Adv. nanoparticle uptake in live cells with liquid
4. Zheng H, Claridge SA, Minor AM, Alivisatos AP, Biol. Med. Phys. 15, 161–270. (doi:10.1016/ scanning transmission electron microscopy.
Dahmen U. 2009 Nanocrystal diffusion in a B978-0-12-005215-8.50012-7) Nano Lett. 11, 1733–1738. (doi:10.1021/
liquid thin film observed by in situ transmission 8. de Jonge N, Peckys DB, Kremers GJ, Piston DW. nl200285r)
electron microscopy. Nano Lett. 9, 2460–2465. 2009 Electron microscopy of whole cells in 12. Peckys DB, Mazur P, Gould KL, de Jonge N. 2011
(doi:10.1021/nl9012369) liquid with nanometer resolution. Proc. Natl Fully hydrated yeast cells imaged with electron
5. Nishijima K, Yamasaki J, Orihara H, Tanaka N. Acad. Sci. USA 106, 2159–2164. (doi:10.1073/ microscopy. Biophys. J. 100, 2522–2529.
2004 Development of microcapsules for pnas.0809567106) (doi:10.1016/j.bpj.2011.03.045)
13. Wang C-M. 2015 In situ transmission electron 27. Wang CM et al. 2010 In situ transmission nanoscale. Microsc. Microanal. 8, 21–28. 21
microscopy and spectroscopy studies of electron microscopy and spectroscopy studies of (doi:10.1017/S143192760201005X)
rechargeable batteries under dynamic operating interfaces in Li ion batteries: challenges and 42. Peng X et al. 2019 In situ TEM study of the
royalsocietypublishing.org/journal/rsos
conditions: a retrospective and perspective view. opportunities. J. Mater. Res. 25, 1541–1547. degradation of PbSe nanocrystals in air. Chem.
J. Mater. Res. 30, 326–339. (doi:10.1557/jmr. (doi:10.1557/jmr.2010.0198) Mater. 31, 190–199. (doi:10.1021/acs.
2014.281) 28. Hansen TW, Wagner JB, Dunin-Borkowski RE. chemmater.8b04052)
14. Sacci RL, Dudney NJ, More KL, Parent LR, Arslan 2010 Aberration corrected and monochromated 43. Sugi H, Akimoto T, Sutoh K, Chaen S, Oishi N,
I, Browning ND, Unocic RR. 2014 Direct environmental transmission electron microscopy: Suzuki S. 2002 Dynamic electron microscopy of
visualization of initial SEI morphology and challenges and prospects for materials science. ATP-induced myosin head movement in
growth kinetics during lithium deposition by in Mater. Sci. Technol. 26, 1338–1344. (doi:10. living muscle thick filaments. Proc. Natl Acad.
situ electrochemical transmission electron 1179/026708310X12756557336355) Sci. USA 94, 4378–4382. (doi:10.1073/pnas.94.
microscopy. Chem. Commun. 50, 2104–2107. 29. Mehdi BL et al. 2015 Observation and 9.4378)
(doi:10.1039/c3cc49029g) quantification of nanoscale processes in lithium 44. Creemer JF, Helveg S, Hoveling GH, Ullmann S,
15. Zeng Z, Liang WI, Liao HG, Xin HL, Chu YH, batteries by operando electrochemical (S)TEM. Molenbroek AM, Sarro PM, Zandbergen HW.
Zheng H. 2014 Visualization of electrode- Nano Lett. 15, 2168–2173. (doi:10.1021/acs. 2008 Atomic-scale electron microscopy at
electrolyte interfaces in LiPF 6/EC/DEC electrolyte nanolett.5b00175) ambient pressure. Ultramicroscopy 108,
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
for lithium ion batteries via in situ TEM. Nano 30. Yuk JM, Seo HK, Choi JW, Lee JY. 2014 993–998. (doi:10.1016/j.ultramic.2008.04.014)
royalsocietypublishing.org/journal/rsos
Chang P, Hsu L, Yew T-R. 2008 Novel microchip embrittlement of multiwalled carbon Maisonhaute E, Tenney S, Sutter P. 2014 In situ
for in situ TEM imaging of living organisms and nanotubes. ACS Nano 5, 7245–7253. (doi:10. liquid-cell electron microscopy of silver–
bio-reactions in aqueous conditions. Lab Chip 8, 1021/nn202071y) palladium galvanic replacement reactions on
1915. (doi:10.1039/b804986f) 72. Liu XH et al. 2012 In situ transmission electron silver nanoparticles. Nat. Commun. 5, 4946.
57. Grogan JM, Bau HH. 2010 The nanoaquarium: a microscopy of electrochemical lithiation, (doi:10.1038/ncomms5946)
platform for in situ transmission electron delithiation and deformation of individual 87. Hermannsdörfer J, de Jonge N, Verch A. 2015
microscopy in liquid media. J. Microelectromech. graphene nanoribbons. Carbon N. Y. 50, Electron beam induced chemistry of gold
Syst. 19, 885–894. (doi:10.1109/JMEMS.2010. 3836–3844. (doi:10.1016/j.carbon.2012.04.025) nanoparticles in saline solution. Chem. Commun.
2051321) 73. Wang JW et al. 2013 Two-phase electrochemical 51, 16 393–16 396. (doi:10.1039/c5cc06812f )
58. de Jonge N, Poirier-Demers N, Demers H, Peckys lithiation in amorphous silicon. Nano Lett. 13, 88. Park JH, Schneider NM, Grogan JM, Reuter MC,
DB, Drouin D. 2010 Nanometer-resolution 709–715. (doi:10.1021/nl304379k) Bau HH, Kodambaka S, Ross FM. 2015 Control
electron microscopy through micrometers-thick 74. Li X et al. 2014 Mesoporous silicon sponge as of electron beam-induced Au nanocrystal
water layers. Ultramicroscopy 110, 1114–1119. an anti-pulverization structure for high- growth kinetics through solution chemistry.
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
(doi:10.1016/j.ultramic.2010.04.001) performance lithium-ion battery anodes. Nat. Nano Lett. 15, 5314–5320. (doi:10.1021/acs.
royalsocietypublishing.org/journal/rsos
110, 7862–7868. (doi:10.1021/jp057549a) Cent. Sci. 1, 33–39. (doi:10.1021/acscentsci. 127. Mehdi BL, Cao R, Park C, Henderson WA, Xu W,
100. Gupta T, Schneider NM, Park JH, Steingart D, 5b00001) Liu J, Zhang J-G, Mueller KT, Browning ND.
Ross FM. 2018 Spatially dependent dose rate in 113. Park J, Zheng H, Lee WC, Geissler PL, Rabani E, 2016 Understanding the effect of additives in
liquid cell transmission electron microscopy. Alivisatos AP. 2012 Direct observation of Li-sulfur batteries by operando ec- (S)TEM.
Nanoscale 10, 7702–7710. (doi:10.1039/ nanoparticle superlattice formation by using Microsc. Microanal. 22, 836–837. (doi:10.1017/
c8nr01935e) liquid cell transmission electron microscopy. ACS S143192761600502X)
101. Wu J, Gao W, Yang H, Zuo J-M. 2014 Imaging Nano 6, 2078–2085. (doi:10.1021/nn203837m) 128. Mehdi BL et al. 2017 Multi-modal
shape-dependent corrosion behavior of Pt 114. White ER, Mecklenburg M, Shevitski B, Singer characterization of new battery technologies by
nanoparticles over extended time using a liquid SB, Regan BC. 2012 Charged nanoparticle operando ec-STEM. Microsc. Microanal. 23,
flow cell and TEM. Microsc. Microanal. 20, dynamics in water induced by scanning 886–887. (doi:10.1017/s1431927617005098)
1508–1509. (doi:10.1017/S1431927614009271) transmission electron microscopy. Langmuir 28, 129. Singh N, Arthur TS, Tutusaus O, Li J, Kisslinger
102. Wu J, Gao W, Yang H, Zuo J-M. 2017 3695–3698. (doi:10.1021/la2048486) K, Xin HL, Stach EA, Fan X, Mohtadi R. 2018
Dissolution kinetics of oxidative etching of cubic 115. Zhou Y, Powers AS, Zhang X, Xu T, Bustillo K, Achieving high cycling rates via in situ
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
and icosahedral platinum nanoparticles revealed Sun L, Zheng H. 2017 Growth and assembly of generation of active nanocomposite metal
royalsocietypublishing.org/journal/rsos
quantum-dot-labeled proteins in whole cells in TEM imaging of CaCO3 nucleation reveals Appl. Phys. Express 4, 055201. (doi:10.1143/
liquid. ACS Nano 4, 4110–4116. (doi:10.1021/ coexistence of direct and indirect pathways. APEX.4.055201)
nn1010232) Science 345, 1158–1162. (doi:10.1126/science. 165. Grogan JM, Schneider NM, Ross FM, Bau HH.
140. Nishiyama H, Suga M, Ogura T, Maruyama Y, 1169434) 2014 Bubble and pattern formation in liquid
Koizumi M, Mio K, Kitamura S, Sato C. 2010 152. Smeets PJM, Cho KR, Kempen RGE, Sommerdijk induced by an electron beam. Nano Lett. 14,
Reprint of: Atmospheric scanning electron NAJM, De Yoreo JJ. 2015 Calcium carbonate 359–364. (doi:10.1021/nl404169a)
microscope observes cells and tissues in open nucleation driven by ion binding in a 166. Schneider NM, Norton MM, Mendel BJ, Grogan
medium through silicon nitride film. J. Struct. Biol. biomimetic matrix revealed by in situ electron JM, Ross FM, Bau HH. 2014 Electron-water
172, 191–202. (doi:10.1016/j.jsb.2010.08.006) microscopy. Nat. Mater. 14, 394–399. (doi:10. interactions and implications for liquid cell
141. Peckys DB, Korf U, de Jonge N. 2015 Local 1038/nmat4193) electron microscopy. J. Phys. Chem. C 118,
variations of HER2 dimerization in breast cancer 153. van de Put MWP, Carcouët CCMC, Bomans PHH, 22 373–22 382. (doi:10.1021/jp507400n)
cells discovered by correlative fluorescence and Friedrich H, de Jonge N, Sommerdijk NAJM. 167. Jiang N. 2017 Note on in situ (scanning)
liquid electron microscopy. Sci. Adv. 1, 2015 Writing silica structures in liquid with transmission electron microscopy study of liquid
Downloaded from https://royalsocietypublishing.org/ on 08 September 2023
e1500165. (doi:10.1126/sciadv.1500165) scanning transmission electron microscopy. samples. Ultramicroscopy 179, 81–83. (doi:10.