Electrochemical setup, beam effects, and resolution

Our interest in putting liquids in the TEM came about as a result of the excitement surrounding porous silicon after its optical emission properties were discovered in 1990. ^{Reference Canham17} A liquid-phase (hydrofluoric acid) anodic etching process creates the porous network. To observe this process in situ, we attempted to etch a thin doped Si layer sandwiched between two undoped layers. ^{Reference Ross, Searson, Bailey, Hennigar and Zaluzec18} Problems with this design led to the use of silicon nitride as a more reliable means to enclose a liquid, using a spacer layer to control liquid thickness and a glued vacuum seal. ^{Reference Ross19} This simpler design enabled a more general type of experiment. In particular, electrochemical processes could be examined by incorporating electrodes within the liquid cell. The early cells, glued by hand and with low success rates, have now been superseded by improved designs with various functionalities now available commercially and from academic laboratories. ^{Reference Ring and de Jonge11,Reference Unocic, Sacci, Brown, Veith, Dudney, More, Walden, Gardiner, Damiano and Nackashi12,Reference Grogan, Schneider, Ross and Bau20–Reference Creemer, Helveg, Hoveling, Ullmann, Molenbroek, Sarro and Zandbergen22}

Closed-liquid-cell electron microscopy turns out to be well suited for electrochemical explorations, because it fills an important niche in terms of temporal and spatial resolution that is not addressed by other techniques. It occupies the gap between light microscopy, ^{Reference Gallaway, Desai, Gaikwad, Corredor, Banerjee and Steingart23} which has good temporal resolution but limited spatial resolution, and scanning tunneling microscopy ( Figure 2 ). Scanning probe techniques can have remarkable resolution in liquids, ^{Reference Magnussen, Zitzler, Gleich, Vogt and Behm24} but the frame rate is limited by the physical scan speed, and it is challenging to access rough liquid/solid interfaces.

Figure 2.

Comparison of imaging techniques for electrochemistry indicating the available resolution. (a) Light microscopy: Deposition of Zn in alkaline electrolyte, from a video at 30 images per s. (b) Liquid-phase transmission electron microscopy: Progression of the growth front during potentiostatic Cu deposition from H₂SO₄/CuSO₄. Three images from a video are superimposed showing the advancing growth front. ^{Reference Grogan, Schneider, Ross and Bau20} (c) Scanning tunneling microscopy: Dynamic changes in the structure of the Cu(100) surface in 0.01M HCl at a voltage of –0.23 V_SCE (where V_SCE is the potential with respect to a saturated calomel electrode). One image is shown from a sequence recorded at 10 images per s during “collective” growth of a terrace segment several rows wide. (a) Reprinted with permission from Reference Reference Gallaway, Desai, Gaikwad, Corredor, Banerjee and Steingart23. © 2010 The Electrochemical Society. (c) Reprinted with permission from Reference Reference Magnussen, Zitzler, Gleich, Vogt and Behm24. © 2001 Elsevier.

This leads us to consider the important question of what resolution we should expect from a closed-liquid-cell experiment. In TEM, thicker samples generally reduce the resolution due to multiple scattering and energy loss of the transmitted electrons. ^{Reference de Jonge and Ross1} Resolution is improved by making both the liquid and the windows as thin as possible. However, too thin of a liquid layer may not support realistic physical phenomena. For electrochemical experiments in particular, where diffusion often dominates the process under study, a compromise must be made where the liquid is thick enough to allow the experiment to represent the “real life” process, yet thin enough to provide reasonable resolution.

A second consideration in liquid-cell electron microscopy—for all experiments, not just electrochemical investigations—is the effect of the electron beam. ^{Reference Abellan, Woehl, Parent, Browning, Evans and Arslan25} Sample–beam interactions in TEM have been well studied for solid samples, but for liquids, we can use the literature from medical and reactor physics. The electron beam deposits energy that breaks bonds to create radiolytic species that then react to generate a cascade of other products. In water, a several hundred keV electron beam can produce strong effects at the high dose rates typical of TEM imaging. ^{Reference Peckys, Veith, Joy and de Jonge26}

Models for water radiolysis, adapted to apply to TEM, ^{Reference Schneider, Norton, Mendel, Grogan, Ross and Bau27,Reference Grogan, Schneider, Ross and Bau28} show that radiolysis products quickly reach a steady-state concentration that depends on the dose rate, and can be rather different from the initial state of the sample. The pH can be changed by irradiation (generally becoming more acidic), radiolytic hydrogen can form bubbles, and radiolytic hydrated electrons, which are highly reactive electrons surrounded by a shell of water, can react with metal ions in solution to form solid particles within the irradiated area. Because of the sensitivity of electrochemical processes to solution chemistry, beam effects must be modeled, and low-dose imaging and efficient detectors are important.

Examples in electrochemistry

Closed-liquid-cell TEM can address a variety of electrochemical questions, with Figure 3 illustrating a few of the possibilities. The deposition of copper from an acidified copper sulfate solution is shown in Figure 3a. This process is used in the microelectronics industry to form metallic interconnects, and details of nucleation and grain structure are important in optimizing the results. To image nucleation, we use a liquid cell with electrodes that are electron transparent and also polycrystalline (making them somewhat more relevant to the industrial substrates). ^{Reference Williamson, Tromp, Vereecken, Hull and Ross7,Reference Ross19} The single image shown in Figure 3a is taken from a set of movies recorded under constant potential. ^{Reference Radisic, Vereecken, Hannon, Searson and Ross29} Nucleation, growth, crystal shape, and coalescence are visible as a function of time.

Figure 3.

Electrochemical processes imaged using a closed-liquid cell in transmission electron microscopy. (a) Bright-field image from a movie obtained during potentiostatic deposition of Cu islands (dark features) on polycrystalline Au (background), and the complete current (i) versus time (t) for the movie. The electrolyte is 0.1 M CuSO₄ + 0.18 M H₂SO₄, and deposition took place at a constant potential of –70 mV with respect to a Cu reference electrode. ^{Reference Radisic, Vereecken, Hannon, Searson and Ross29} (b) Growth instabilities forming during galvanostatic deposition of Zn onto polycrystalline Au. In a pulsed mode, 40 nA was applied (10 s on, 10 s off for 20 cycles). The electrolyte is 0.5 M ZnSO₄ + 0.1 M H₂SO₄ for both images, but with the addition of 0.25 mM Bi₂O₃ in the left-hand image. The Bi does not suppress dendrites, but it delays their development by favoring Zn growth between the asperities. (c) Formation of intragranular pits in a coarse-grained Al film after exposure to distilled water in a closed-liquid cell for the times indicated. ^{Reference Chee, Ross, Duquette and Hull33} (d) Galvanic displacement of Ni by Pd. The right-hand image shows Pd nanostructures (dark features) that formed after flow of 0.18 M H₂SO₄ + 0.025 M PdSO₄.

We simultaneously recorded the current flowing through the liquid cell at each time (see the graph in Figure 3a) and correlated this electrochemical signature with the movie. Such information is helpful in developing models of nucleation and growth, especially as compared to standard analysis where the surface is imaged only ex situ after deposition is complete or at fixed times after interrupting growth. The data in Figure 3a were used to identify different pathways by which atoms arrive at the growing nuclei. ^{Reference Radisic, Vereecken, Hannon, Searson and Ross29,Reference Radisic, Vereecken, Searson and Ross30}

Once the surface is covered with metal, it is still possible to track deposition by imaging the edge of the electrode as the growth front propagates into the liquid. An example is shown in Figure 3b for the growth of zinc. ^{Reference Park, Steingart, Schneider, Kodambaka and Ross31} The growth front has developed an irregular, dendritic appearance. Analysis of the growth front evolution can be carried out over a range of conditions, measuring for example, roughness, local deposition rate, and spatial correlations. This provides the opportunity to examine the physical phenomena that determine growth front stability. Analysis based on fitting classical models for growth front dynamics, such as kinetic roughening and diffusion limited models, can help in understanding what triggers the growth front to become unstable.

Growth front stability is important in battery charge/discharge cycling, where formation of the type of structure shown in Figure 3b could potentially short out the battery. ^{Reference Tarascon and Armand32} For zinc-based batteries, which otherwise have many environmental and cost benefits, dendrite formation is a particular problem, reducing cycle life. ^{Reference Gallaway, Desai, Gaikwad, Corredor, Banerjee and Steingart23} Liquid-cell TEM allows us to evaluate strategies that have been developed to maintain growth-front planarity, such as the use of additives and pulse plating. As an example, the two images in Figure 3b compare similar growth with and without the presence of an inorganic additive, Bi₂O₃. The additive does not suppress instabilities, but delays their onset (Figure 3b, left). ^{Reference Park, Steingart, Schneider, Kodambaka and Ross31} It would allow a slightly faster charge rate for a battery—if we first complete the important work of showing that this result also applies to much larger practical systems.

Another electrochemical example where liquid-cell microscopy can be helpful is corrosion. This hugely important problem starts at small features—pinholes, grain boundaries, crevices—making corrosion challenging to analyze using spatially averaged techniques. The images in Figure 3c are from an Al thin film exposed to water in the liquid cell and imaged every few hours. ^{Reference Chee, Ross, Duquette and Hull33} Intragranular pit formation is followed by the formation of voids beneath the passivating oxide layer. The possibilities are exciting, especially if combined with compositional measurements. ^{Reference Zaluzec, Burke, Haigh and Kulzick34}

Figure 3d shows a related example of the process of galvanic displacement. A solution containing palladium ions is flowed over a nickel surface, and Pd is deposited as the “less noble” metal as Ni goes into solution. Liquid-cell TEM provides a good way to access the microstructural details of this type of process. ^{Reference Sutter, Jungjohann, Bliznakov, Courty, Maisonhaute, Tenney and Sutter35} Here, Pd is seen in the second image to deposit in the form of nanoparticles, creating a nanostructured Pd/Ni surface that has useful catalytic properties, for example, for subsequent electroless deposition of magnetic materials.

These examples illustrate some of the electrochemical processes that are accessible with closed-liquid-cell TEM. The key benefit is the combination of electrochemical measurements with structural information obtained with a temporal and spatial resolution that cannot be achieved using other techniques. The challenge in using the results to refine electrochemical models is to relate the liquid-cell data to the larger-scale process. We need to control the experimental geometry, compare with ex situ data, and understand and model the sample–beam interactions and liquid flow patterns.

With careful quantification, and continuing improvements in the closed-cell experimental technique, we anticipate exciting opportunities involving heating, cooling, different electrode geometries and materials, and well-controlled liquid flow. Advanced electron microscopy techniques including high-speed image acquisition, analytical microscopy, aberration-corrected microscopy, and correlative microscopy, which we discuss next in the context of biological materials, will extend the reach of these in situ investigations even further.

The fundamental concept of open-cell TEM

Energy storage systems such as rechargeable batteries include multiple components, and these components are closely integrated to allow concordant ionic and electronic transport during functioning. ^{Reference Tarascon and Armand32} Cyclic charging and discharging of a battery causes structural and chemical evolution of the cathode, anode, electrolyte, and the interfaces between the liquid electrolyte and electrode. It is this evolution of the battery components that, in most cases, directly contributes to capacity fading and the eventual failure of the battery. Designing and discovering new or improved materials for batteries therefore critically relies on a fundamental understanding and direct correlation between structural and chemical evolution and performance. Electron microscopy and spectroscopy in ex situ mode is one of the most powerful tools to probe the relevant characteristics of battery materials, comparing microstructural features before and after cycling to yield insights into the relationship between microstructure and battery failure. However, the dynamic nature of cycling and the continuous evolution of the battery components mean that ex situ microscopy cannot answer key questions regarding the timing and mechanism of structural and chemical evolution. Direct observation under dynamic operating conditions is preferable.

The development of open-cell TEM techniques has therefore been a major advance for TEM studies of rechargeable batteries. ^{Reference Huang, Zhong, Wang, Sullivan, Xu, Zhang, Mao, Hudak, Liu, Subramanian, Fan, Qi, Kushima and Li15,Reference Wang, Xu, Liu, Choi, Arey, Saraf, Zhang, Yang, Thevuthasan, Baer and Salmon36} Open cells can be used because electrolytes based on ionic liquids with low-vapor pressure can be loaded directly into the TEM vacuum and imaged without the need for windows. At the same time, to make the battery small enough to fit into the TEM, a single nanowire can be used to form the observable electrode. The arrangement of such a nanobattery is schematically illustrated in Figure 4a. ^{Reference Wang37} Here we show SnO₂ for the anode and LiCoO₂ for the cathode; the electrolyte is 10% lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) dissolved in a hydrophobic ionic liquid, 1-butyl-1-methylpyrrolidium (P₁₄) TFSI (P₁₄TFSI). This electrolyte wets the nanowire surface to form a conformal coating (Figure 4b), which is equivalent to immersion of the electrode material in the electrolyte. In this sense, the open-cell nanobattery configuration can mimic the processes that occur in a real battery. A variation of the open-cell configuration is achieved by using lithium oxide as a solid electrolyte. ^{Reference Liu and Huang38,Reference Wang, Yu, Chen, Wu, Pereira, Thornton, Van der Ven, Zhu, Amatucci and Graetz39}

Figure 4.

Open-cell transmission electron microscope (TEM) used to examine operation of a nanobattery formed using an ionic liquid as the electrolyte and a 100 nm diameter nanowire as the anode. ^{Reference Wang37} (a) Schematic drawing showing the experimental configuration. (b) Typical TEM image of the anode of a working nanobattery visualized during cycling.

Although the open-cell nanobattery works well for direct observation of electrode material evolution, we find some deficiencies if we compare it with real batteries. These include modification of kinetics (Li-ion transport and electrode response to ion motion) and the lack of a solid-electrolyte interphase layer. Closed cells allow use of electrolytes that are closer to real battery operation, mitigating the previously discussed shortcomings, but the presence of the windows lowers the resolution. A combination of open- and closed-cell TEM is the best way to achieve an accurate understanding of the operation of these complex materials systems.

Direct observations of lithiation-induced structural and chemical evolution

The open-cell configuration has provided a wealth of information regarding the behavior of metal ions such as Li, Na, and Ca in electrode materials during charge/discharge cycles, ^{Reference He, Gu, Xiao, Luo, Shao, Gao, Du, Mao and Wang40} directly revealing structural and morphological evolution in anode materials such as Si, Ge, Al₂O₃, SnO₂, ZnO, Sn, graphene, and carbon nanotubes. The lithiation behavior of Si illustrates the complexity of these reactions, as it features lithiation-induced solid-state amorphization, anisotropic volume expansion, a two-phase reaction process, a self-regulated lithiation rate due to stress generation, a phase transition of amorphous Li _x Si to Li₁₅Si₄ at a critical composition x, and a critical size for fracture of the Si particles. ^{Reference Gu, Wang, Connell, Perea, Lauhon, Gao and Wang41} The lithiation of a silicon nanowire is schematically illustrated in Figure 5 .

Figure 5.

Lithiation process studied with scanning transmission electron microscopy (STEM). (a) STEM annular dark-field image of a Si nanowire anode. Electron energy-loss spectroscopy maps of elemental distribution, (b) Si, (c) Li, and (d) superimposed Li and Si, revealing the core–shell lithiation of the silicon nanowire. (e) Schematic of the silicon nanowire lithiation process and the spontaneous congruent phase transition of amorphous a-Li _x Si to crystalline c-Li₁₅Si₄ when x reaches the critical value of 3.75 in the amorphous phase. ^{Reference Gu, Wang, Connell, Perea, Lauhon, Gao and Wang41}

To illustrate the power of open-cell TEM, consider one stage of this process, the phase transition of amorphous Li _x Si (a-Li _x Si) to crystalline (c-) Li₁₅Si₄. A variety of in situ and ex situ spectroscopy measurements have shown that lithiated a-Li _x Si crystallizes instantaneously to either Li₁₅Si₄ or Li₂₂Si₅. ^{Reference Hatchard and Dahn42} In situ S/TEM provided a detailed atomistic mechanism for this phase transition, revealing that a-Li _x Si transforms to c-Li₁₅Si₄ when x reaches a critical value of 3.75, and this phase transition occurs via a congruent process that leads to neither phase segregation nor formation of a residual phase (Figure 5). ^{Reference Gu, Wang, Connell, Perea, Lauhon, Gao and Wang41} The congruent phase transformation process does not involve long-range diffusion of the chemical species, only local atomic rearrangement, in contrast to general nucleation and growth processes in which local chemical composition fluctuations lead to phase separation. In situ TEM imaging, with its resolution and compositional capability, is a powerful approach for understanding and optimizing the reactions taking place during battery operation.

Principles of studying membrane proteins

Liquid-phase electron microscopy has great potential for the study of membrane proteins in mammalian cells. ^{Reference de Jonge, Peckys, Kremers and Piston10,Reference Kinoshita, Mori, Hirano, Sugimoto, Okuda, Matsumoto, Namiki, Ebihara, Kawata, Nishiyama, Sato, Suga, Higashiyama, Sonomoto, Mizunoe, Nishihara and Sato43–Reference Nishiyama, Suga, Ogura, Maruyama, Koizumi, Mio, Kitamura and Sato47} Membrane proteins (receptors) in the plasma membrane of the cell respond to the presence of small proteins, specifically binding to the receptor (so-called ligands), and then decide what action the cell should take, for example, cell growth. But despite the fact that membrane proteins represent almost one-third of all cellular proteins and two-thirds of current drug targets, many factors governing their function are still unknown.

An example of a membrane protein is the epidermal growth factor receptor (EGFR) ( Figure 6 ). ⁴⁸ This receptor becomes activated upon binding of its ligand (EGF) and then forms a protein complex together with another activated EGFR. This dimerization is followed by a conformational change of the intracellular part of the dimer, which triggers a signal cascade eventually leading to cell growth. To find out exactly what happens at the cell membrane once EGF is present, one would be interested in knowing, for example, in which cellular region or in which cells dimerization takes place, or under which chemical conditions and at which time points. But no one has observed individual proteins in the intact cell. Most of our knowledge about these proteins has been obtained from indirect biochemical methods where material from thousands of cells is pooled, and cell heterogeneity is thus not taken into account. ^{Reference Peckys, Korf and de Jonge46}

Figure 6.

Side view of the HER1 dimer in the plasma membrane. Left: Non-activated HER1 monomer and unbound ligand epidermal growth factor (EGF). Right: Activated HER1 dimer with bound EGF ligand. ⁴⁸

To study membrane proteins with liquid-phase STEM, mammalian cells in liquid are enclosed between thin and electron transparent silicon nitride windows separating the sample from the vacuum of the electron microscope ( Figure 7a). In order to obtain contrast through water and cell material that is several micrometers in thickness, nanoparticles of a material of high density and atomic number (Z) such as gold or fluorescent quantum dots (QDs) are used as specific protein labels. Since the labels may influence the proteins, extensive control experiments are needed. The atomic number contrast in STEM imaging allows these nanoparticles to be detected above the background signal produced by the thick layer of low-Z material. ^{Reference de Jonge, Peckys, Kremers and Piston10,Reference de Jonge, Poirier-Demers, Demers, Peckys and Drouin49}

Figure 7.

Principles of liquid-phase scanning transmission electron microscopy (STEM) for whole cells. The cell is grown on a supporting silicon nitride (SiN) thin film. Proteins labeled with nanoparticles (NPs) reside in the plasma membrane. As the focused electron beam scans over the cell, the transmitted electrons are recorded using a STEM detector beneath the sample. (a) The cell is fully enclosed in a microfluidic chamber with two windows. (b) The cell is maintained in an atmosphere of saturated water vapor, while a thin layer of cooled water covers the cell, allowing STEM imaging in an environmental scanning electron microscope. ^{Reference Peckys and de Jonge45} Note: ADF, annular dark field.

It is not always necessary to enclose the cells fully in liquid in a microfluidic chamber. For many studies, it is sufficient to obtain information from the thin outer regions of the cells, and those can be imaged with high resolution using ESEM with a STEM detector (Figure 7b). ^{Reference Bogner, Thollet, Basset, Jouneau and Gauthier16,Reference Peckys, Baudoin, Eder, Werner and de Jonge50} ESEM works for thinner samples than standard TEM/STEM because of the lower electron-beam energy (30 keV versus 200 keV), but the same principle of Z-contrast applies using the STEM detector in ESEM. Cells in water are kept at 3°C in a sample stage equipped with a Peltier cooling element, and the vacuum level in the specimen chamber is carefully adjusted to around 740 Pa to maintain a thin liquid layer on the sample. Imaging is accomplished with an electron dose below the damage threshold. ^{Reference Hermannsdörfer, Tinnemann, Peckys and de Jonge51}

For many biological experiments, it is highly desirable to examine not just a randomly selected cellular region, but rather to navigate through many cells and select regions of interest. For this reason, correlative light microscopy and liquid-phase STEM are powerful tools, ^{Reference Dukes, Peckys and de Jonge52} especially since the microchip with cells presents a simple platform for transferring a sample between the two microscopes.

Research on HER2 in breast cancer cells

As an example of the capabilities of liquid-phase STEM, we describe results obtained for HER2 receptors in breast cancer cells. ^{Reference Peckys, Korf and de Jonge46} This receptor of the EGFR family is special because it does not need a ligand to become active; it is an orphan receptor. As a consequence, this protein can trigger cell growth without being controlled by an external signal. This can lead to dysregulation of cell growth in the tumor type that overexpresses HER2. ^{Reference Brennan, Kumagai, Berezov, Murali and Greene53} Particularly problematic is that drug resistance develops in many patients, ^{Reference Vu and Claret54} with an origin that is not fully understood.

Figure 8a shows an overview fluorescence image of QD-labeled HER2 in breast cancer cells. These cells differ from each other in morphology and HER2 membrane concentration. The fluorescence and STEM images of a region of interest are shown in Figure 8b–c. The QD labels become visible at higher magnification (Figure 8d). The high spatial resolution of 3 nm allows measurement of the individual positions of all labeled HER2 proteins. The HER2 distances were statistically analyzed and it was found that an interlabel distance of 20 nm occurred more frequently than expected at random. Since the expected center-to-center label distance of two labels attached to a HER2 homodimer amounts to 20 nm, it was deduced that HER2 was present as a homodimer. ^{Reference Peckys, Korf and de Jonge46} It was thus possible to quantify the stoichiometry of the HER2 complexes, distinguishing between monomers, dimers, and higher-order clusters, and resolving this information per cellular region.

Figure 8.

Correlative light and electron microscope overview images of quantum dot-labeled HER2 proteins in SKBR3 human breast cancer cells. (a) Fluorescence overview image showing several dozens of cells. (b) Fluorescence image of the cells within the boxed area in (a). (c) Liquid-phase scanning transmission electron microscope (STEM) image of the boxed region in (b) recorded at 15,000× magnification using environmental scanning electron microscopy. (d) STEM image recorded in the boxed region shown in (c) at 75,000× magnification. The localization of individual HER2 receptors is visible as the bright, bullet-shaped labels. The brighter background represents membrane ruffles. Many pairs (homodimers) are visible; two are indicated with arrowheads. ^{Reference Peckys, Korf and de Jonge46}

A key finding was the discovery of a small (few percent of the total of cells) sub-population of cells with a different phenotype compared to the average cell. ^{Reference Peckys, Korf and de Jonge46} This group of cells was characterized by a flat cytoplasmic border and was identified as resting cells. HER2 homodimers were found to be absent from this subpopulation of cells, even though the membrane concentration of HER2 was similar to that of the majority of cancer cells. We hypothesize that this sub-population of cancer cells plays a critical role in the development of drug resistance because these small sub-populations of cells could possibly respond differently or not at all to the drugs, while they would remain undetected using biochemical methods or conventional microscopy techniques. This example demonstrates both the need for single-cell studies to address cancer cell heterogeneity, and the unique capabilities of liquid phase STEM.