Closed cells

Water, the most interesting of liquids, has traditionally been incompatible with the vacuum constraints of electron microscopy because of its high vapor pressure. Liquids with high vapor pressure can be separated from the microscope vacuum in a closed liquid cell. This is an enclosure that is filled, sealed, and loaded into the microscope on a sample holder. Imaging requires robust, electron-transparent windows that are separated by a controlled, narrow distance, since good image resolution is obtained only through liquid layers thinner than a few micrometers. Although closed cells were considered early in the development of electron microscopy (see the review in Reference 26), development of a practical closed cell required the use of silicon microfabrication techniques.^{Reference Williamson, Tromp, Vereecken, Hull and Ross14} Cells are typically built from two silicon wafers, each of which is coated with a 50–100-nm-thick silicon nitride layer (Figure 1a). A region of Si ∼100 μm across is etched out of each wafer, forming windows of silicon nitride. A spacer material, such as silicon dioxide patterned onto one wafer, keeps the windows at the correct separation.

Figure 1.

(a) Schematic diagram showing a closed liquid cell with two electrodes, on each of which is attached one component of a battery. The cathode is a Si nanowire (NW), and the anode is metallic Li. (b) Series of images during the lithiation of a crystalline Si NW (in this case coated with Cu to increase conductivity), performed by holding the Cu–Si voltage at ∼0.03 V. (c) Schematic diagram showing an open liquid cell with an ionic liquid electrode. (d) Two schematics and (e) an image recorded at a constant voltage of –2 V versus Li metal during lithiation of an (uncoated) Si NW showing the sequential nature of the process. (f) Scanning transmission electron microscopy (STEM) image and electron energy-loss spectroscopy map of element distribution of the process depicted in (d) and (e). (b) and (f) Reproduced with permission from Reference 16. © 2013 American Chemical Society.

Additional functionality can be applied to this basic closed cell. Electrodes can be patterned on the interior surface of the windows and controlled by an external potentiostat to drive electrochemical processes.^{Reference Williamson, Tromp, Vereecken, Hull and Ross14} The interior surfaces can be modified to react with molecules in solution.^{Reference Degen, Dukes, Tanner and Kelly27} Liquids can be pumped through the cell^{Reference Ring and de Jonge28} to examine mixing or to change the solution chemistry during imaging. More recently, heating and cooling capabilities have allowed observations of phase transformations^{Reference Tai, Liu and Dillon29} and crystal growth.^{Reference Liao, Zherebetskyy, Xin, Czarnik, Ercius, Elmlund, Pan, Wang and Zheng30}

Silicon nitride is strong, relatively inert, and amorphous, an almost ideal window material. However, its presence reduces image resolution. Deflection of the windows by the pressure difference between the liquid within the cell and the vacuum within the microscope decreases the resolution still further,^{Reference de Jonge and Ross26,Reference Woehl, Jungjohann, Evans, Arslan, Ristenpart and Browning31–Reference Welch, Faller, Evans and Browning36} as the liquid layer can typically be several times thicker^{Reference Jungjohann, Evans, Aguiar, Arslan and Browning37} in the center of the window compared to the edge. Strategies to improve resolution include the use of thicker, stiffer windows that deflect less and are patterned with small, thinner areas for imaging;^{Reference Creemer, Helveg, Hoveling, Ullmann, Molenbroek, Sarro and Zandbergen38} posts connecting the windows to reduce window deflection;^{Reference Grogan and Bau39} narrow channels;^{Reference Jensen, Burrows and Mølhave40} and graphene as the ultimate window material to enclose small droplets of liquid.^{Reference Yuk, Park, Ercius, Kim, Hellebusch, Crommie, Lee, Zettl and Alivisatos41}

Once a closed liquid cell is assembled on its sample holder with the liquid placed or flowed into the cell, the system is tested for leaks and then loaded into the microscope. Many processes of interest, such as liquid-phase particle growth, are driven by the electron beam through its radiolytic effects on the liquid. We present an example of such an experiment later in this article and discuss electron beam effects. Experiments involving electrochemical processes have more complex requirements (Figure 1a). Diffusion often dominates electrochemical behavior, so the need to set up a “realistic” electrochemical process (where thicker liquid layers allow diffusion fields that are more representative of the bulk process) conflicts with the requirements for high image resolution. However, with appropriate cell design, one can observe and quantify nucleation and growth on and around an electrode or at a sample placed on the electrode. We show some electrochemical results later in this article.

The closed liquid cell is a versatile tool for imaging structures and processes in liquids. The main challenge is to optimize resolution while controlling the liquid environment: thickness, flow, electrical bias, and temperature. Closed cells provide the opportunity to set up processes in a liquid film of well-characterized geometry, with the key advantage of being able to observe liquids with high vapor pressure and flowing liquids.

Open cells

For liquids with low vapor pressure, an alternative approach is possible that avoids the decrease in resolution caused by the windows in a closed liquid cell. In this case, droplets of liquids, particularly ionic liquids (ILs), are placed directly onto a supporting sample that includes a large-area electrode, and this sample is loaded into the microscope on a custom sample holder. The liquid is stable under vacuum, and electrons can pass through the liquid in regions where it is thin (for example, where it forms a wetting layer ∼100 nm in thickness or near the edges of droplets). To examine processes in such free-standing liquid droplets, an electron-transparent probe such as a nanowire is pushed into the liquid, and an electrical bias is applied to drive reactions mediated by the liquid.

Wang et al.^{Reference Wang, Xu, Liu, Choi, Arey, Saraf, Zhang, Yang, Thevuthasan, Baer and Salmon42} and Huang et al.^{Reference Huang, Zhong, Wang, Sullivan, Xu, Zhang, Mao, Hudak, Liu, Subramanian, Fan, Qi, Kushima and Li43} initiated this open cell concept to study reactions in battery materials (Figure 1c). The battery is composed of an ionic liquid electrolyte, a single nanowire electrode, and bulk LiCoO₂ as the counter electrode. A typical IL electrolyte is 10% lithium bis(trifluoromethane sulfonyl)imide (LiTFSI) dissolved in a hydrophobic ionic liquid, 1-butyl-1-methylpyrrolidium (P₁₄) TFSI (P₁₄TFSI). The IL wets the nanowire, forming a thin layer around it. Therefore, this configuration mimics a real battery to the extent that the electrolyte conformally coats the active component in the electrode. Open cells offer the possibility of atomic-level spatial resolution and analytical capability for studying the mechanisms of Li-ion insertion into electrode materials during charge/discharge cycles (Figure 1d–f). This has helped reveal details of the lithiation mechanism and structural evolution, especially in anode materials such as Si,^{Reference Liu, Zheng, Zhong, Huang, Karki, Zhang, Liu, Kushima, Liang, Wang, Cho, Epstein, Dayeh, Picraux, Zhu, Li, Sullivan, Cumings, Wang, Mao, Ye, Zhang and Huang44} Ge,^{Reference Liu, Huang, Picraux, Li, Zhu and Huang45} Al₂O₃,^{Reference Liu, Hudak, Huber, Limmer, Sullivan and Huang46} SnO₂,^{Reference Wang, Xu, Liu, Choi, Arey, Saraf, Zhang, Yang, Thevuthasan, Baer and Salmon42} ZnO,^{Reference Kushima, Liu, Zhu, Wang, Huang and Li47} graphene,^{Reference Liu, Wang, Liu, Zheng, Kushima, Huang, Zhu, Mao, Li, Zhang, Lu, Tour and Huang48} and carbon nanotubes.^{Reference Liu, Zheng, Liu, Huang, Zhu, Wang, Kushima, Hudak, Huang, Zhang, Mao, Qian, Li and Huang49} A variation of this experiment uses lithium oxide as the electrolyte rather than an IL. In this configuration, Li metal is used as the anode, and a single nanowire is used as the cathode, with a Li₂O coating on the Li metal surface serving as the electrolyte.^{Reference Liu, Zheng, Zhong, Huang, Karki, Zhang, Liu, Kushima, Liang, Wang, Cho, Epstein, Dayeh, Picraux, Zhu, Li, Sullivan, Cumings, Wang, Mao, Ye, Zhang and Huang44,Reference Islam and Bredow50,Reference Wang, Yu, Chen, Wu, Pereira, Thornton, Van der Ven, Zhu, Amatucci and Graetz51} In principle, this is a charged battery that should discharge when the circuit is completed. However, the low conductivity of Li ions in Li₂O at room temperature prevents current flow.^{Reference Islam and Bredow50} A negative potential of typically 2–4 V must be applied between Si and Li to drive Li⁺ ions into Si.

Open-cell experiments have provided useful information on structural and chemical evolution of electrodes upon lithiation/delithiation. However, three typical deficiencies are associated with the open-cell configuration. First, the electrolyte is in only point contact with the electrode. This presumably modifies the diffusion pattern of Li ions in the electrode, such that the results are not necessarily representative of a real battery in which the electrode is fully immersed in the liquid electrolyte. Second, when Li₂O is used as the electrolyte, the large overpotential required might change the kinetics and phase behavior of the solid-state electrode lithiation. Third, the use of IL and Li₂O electrolytes excludes the study of many processes in real electrolytes and operating conditions, such as the interaction between electrolyte and electrode and solid–electrolyte interphase (SEI) formation; furthermore, ILs can polymerize under an electron beam, so the cell can be cycled only a limited number of times, short of the performance required for revealing structural evolution as a battery ages. Studies of one material system in open and closed cells can help address the shortcomings of each cell, as we illustrate later in this article.

Nanoparticle nucleation and growth kinetics in closed cells

Liquid-cell TEM provides a unique platform for the study of colloidal nanoparticle formation. Through the tracking of single-particle growth trajectories, details of the growth mechanisms have been revealed, including previously unrecognized growth pathways.^{Reference Zheng, Smith, Jun, Kisielowski, Dahmen and Alivisatos10,Reference Liao, Cui, Whitelam and Zheng11,Reference Yuk, Park, Ercius, Kim, Hellebusch, Crommie, Lee, Zettl and Alivisatos41,Reference Liao and Zheng52,Reference Jungjohann, Bliznakov, Sutter, Stach and Sutter53} Topics of study have included nanoparticle coalescence or oriented attachment,^{Reference Liao, Cui, Whitelam and Zheng11,Reference Li, Nielsen, Lee, Frandsen, Banfield and De Yoreo20,Reference Yuk, Park, Ercius, Kim, Hellebusch, Crommie, Lee, Zettl and Alivisatos41} growth from monomers,^{Reference Woehl, Evans, Arslan, Ristenpart and Browning35,Reference Jungjohann, Bliznakov, Sutter, Stach and Sutter53} shape evolution and control,^{Reference Liao, Cui, Whitelam and Zheng11,Reference Liao, Zherebetskyy, Xin, Czarnik, Ercius, Elmlund, Pan, Wang and Zheng30,Reference Liao and Zheng52} and surfactant effects.^{Reference Liao and Zheng52}Figure 2 shows an example in which the nucleation and propagation of individual facets in a nanoparticle were captured using an advanced single-electron detector (a Gatan K2-IS camera providing 400 images per second) during the formation of Pt nanocubes in a liquid precursor solution containing oleylamine surfactant.^{Reference Liao, Zherebetskyy, Xin, Czarnik, Ercius, Elmlund, Pan, Wang and Zheng30} Direct observation showed that the surface energy minimization rule breaks down at the nanoscale for colloidal synthesis. Instead, a selective facet-arrested growth mechanism was identified, arising from the mobility difference of the surfactants on the facets.

Figure 2.

Sequential transmission electron microscope images and corresponding simulated images show the facet development of a Pt nanocube in a liquid cell. A growth solution of platinum bis(acetylacetonate) (20 mg/mL) in a solvent mixture of oleylamine, oleic acid, and pentadecane (4.5:1.5:4) was used. The particles were primarily imaged along the [110] zone axis. Reproduced with permission from Reference 30. © 2014 American Association for the Advancement of Science.

A diverse set of synthetic approaches exists for colloidal nanocrystals. Nanoparticle synthesis is often driven by heating but can also be stimulated by UV−visible light irradiation,^{Reference Akhavan and Ghaderi54} microwave heating,^{Reference Panda, Glaspell and El-Shall55,Reference Liao, Chen, Xu, Yang and Zhu56} electrons,^{Reference Li, Peng, Fang, Chen, Pan and Wu57,Reference Zhou, Zhou, Hao, Wu, Rao, Chen and Gao58} x-rays,^{Reference Divan, Ma, Mancini and Keane59} gamma rays,^{Reference Rojas and Castano60} and ion irradiation.^{Reference Mishra, Avasthi, Kulriya, Singh, Kabiraj, Tripathi, Pivin, Bayer and Biswas61} Most liquid-cell studies on nanoparticle synthesis have used the electron beam to reduce metal precursors. Recently, however, heating effects are starting to be examined, with observations of oscillatory particle growth at 180°C.^{Reference Xin and Zheng62} Heating extends the range of synthesis techniques available for in situ observations and suggests that the study of nanocrystal growth using liquid-cell TEM will further enrich the field of colloidal nanocrystals and materials chemistry.

Electrochemical deposition in closed cells

Electrochemical experiments can be carried out in closed cells with great flexibility, as electrodes can be patterned with arbitrary geometry and using a range of materials. Figure 3a shows closed-cell data obtained during the nucleation and growth of copper from an acidic copper sulfate electrolyte.^{Reference Radisic, Vereecken, Hannon, Searson and Ross63–Reference Radisic, Vereecken, Searson and Ross65} Copper electro-deposition is used to fabricate conducting tracks in integrated circuits. A continuous yet thin Cu film is required, implying a high initial nucleation density on the polycrystalline liner (the material on which the Cu is deposited). In Figure 3a, Cu growth takes place on polycrystalline Au under conditions of constant voltage. The electrode is transparent to the imaging electrons, allowing the spatial distribution and growth rates of individual Cu islands to be measured. A second key feature of the experiment is the ability to correlate images with the electrochemical response. In Figure 3a, the time-varying current through the whole cell is correlated with the nucleation times and growth rates of the Cu islands. Growth rates, measured from video frames, are then matched with expectations from electrochemical models,^{Reference Radisic, Vereecken, Hannon, Searson and Ross63–Reference Radisic, Vereecken, Searson and Ross65} calculated from the current–time curves, to explore pathways for Cu diffusion and to understand nucleus density. The imaged area is 2 μm², while the total electrode area is 2000 μm². If we measure the island volumes and hence charge passed for the imaged area and then compare these results with the total charge passed for the total area, the comparison suggests that all areas of the electrode behave uniformly, even those out of view. This type of calibration is important for quantitative analysis of electrochemical processes. Other experiments allow the effects of additives to be examined and the effects of the electron beam to be quantified.^{Reference den Heijer, Shao, Radisic, Reuter and Ross66}

Figure 3.

Electrochemical deposition in the liquid cell. (a) Time-varying current recorded during the deposition of Cu on polycrystalline Au from 0.1 M CuSO₄ + 1 vol% H₂SO₄ at a potential of –0.07 V, with images extracted from a video recorded simultaneously. The times corresponding to the images are indicated by vertical marks. Nucleation and growth of individual Cu crystallites can be measured. (b) Deposition of Cu onto polycrystalline Pt from the same electrolyte, at –1.2 V relative to the open-circuit potential. Time zero corresponds to application of the potential. Initial rapid coverage of the electrode is followed by unstable lateral growth. The field of view was moved to track the growth front. (c) Pb dendrite growth and collapse from 1.5 M PbNO₃ during application of square-wave voltage pulses. (a) Reproduced with permission from Reference 63. © 2006 American Chemical Society. (b) Reproduced with permission from Reference 67. © 2012 Indian Institute of Science. (c) Reproduced with permission from Reference 15. © 2012 American Chemical Society.

If larger volumes of Cu are grown, the nuclei merge and cover the electrode (Figure 3b), and deposition then proceeds outward from the electrode edge.^{Reference Mirsaidov, Ohl and Matsudaira18,Reference Grogan, Schneider, Ross and Bau67} Depending on the growth rate and electrolyte chemistry, such lateral deposition can occur uniformly, or the growth front can become unstable. Instability occurs when growth is controlled by diffusion through the electrolyte, such that random asperities collect more ions and grow even further. Lateral growth experiments provide an interesting model for the charging process in secondary batteries such as those based on Zn or Pb, where nonuniform deposition onto the anode during charging can short the battery. Figure 3b shows the development of an unstable Cu growth front, and Figure 3c shows Pb dendrites grown from lead nitrate,^{Reference White, Singer, Augustyn, Hubbard, Mecklenburg, Dunn and Regan15} relevant to conventional lead-acid batteries.

Lithium-ion battery anode reactions in closed and open cells

Charge and discharge processes in lithium-ion batteries (LIBs), involving insertion/extraction of ions into/from a host lattice, can be examined in both closed and open cells if the electrolyte has appropriate stability. Figure 1b shows closed-cell results from Gu et al.^{Reference Gu, Parent, Mehdi, Unocic, McDowell, Sacci, Xu, Connell, Xu, Abellan, Chen, Zhang, Perea, Evans, Lauhon, Zhang, Liu, Browning, Cui, Arslan and Wang16} obtained during lithiation and delithiation of a Si nanowire in a battery-relevant electrolyte; similar devices have been demonstrated by other groups.^{Reference Gu, Parent, Mehdi, Unocic, McDowell, Sacci, Xu, Connell, Xu, Abellan, Chen, Zhang, Perea, Evans, Lauhon, Zhang, Liu, Browning, Cui, Arslan and Wang16,Reference Holtz, Yu, Gunceler, Gao, Sundararaman, Schwarz, Arias, Abruña and Muller68–Reference Unocic, Sacci, Brown, Veith, Dudney, More, Walden, Gardiner, Damiano and Nackashi71} It is interesting to compare Figure 1b with nanowire lithiation observed in an open cell, Figure 1e. In the open cell, lithiation proceeds in a sequential manner, and the experiment provides excellent resolution of the structure changes, dislocations, and mechanical deformation that occur during the process. In the closed cell, lithiation and delithiation progress in a core–shell fashion; in other words, the exterior surface of the nanowire reacts first, and more central regions react subsequently. The difference in reaction mechanism can be understood in terms of the diffusion pathways available. In the closed cell, ions arrive from the large volume of electrolyte that surrounds the nanowire, whereas in the open cell, ions diffuse through the thin layer that wets the nanowire. Different reaction interfaces and stress relaxation processes are expected. Thus, comparing results from the two types of liquid cells is the best way to understand the true behavior of the active materials in an LIB electrode.

Lithium-ion battery cathode reactions in closed and open cells

Lithium-ion transport in LiFePO₄ serves as the prototypical example to address scientific questions related to ion transport in cathode materials. It is known that the removal and insertion of Li ions lead to a reversible phase transformation, LiFePO₄ ↔ FePO₄. Three models have been proposed to describe Li-ion transport, namely, the shrinking-core,^{Reference Morgan, Van der Ven and Ceder72} phase-boundary migration,^{Reference Gibot, Casas-Cabanas, Laffont, Levasseur, Carlach, Hamelet, Tarascon and Masquelierk73} and solid-solution^{Reference Malik, Zhou and Ceder74} models; distinguishing these models requires correlation of ionic transport with structure. Figure 4 shows Li-ion transport kinetics and degradation mechanisms in LiFePO₄ particles during charge/discharge cycles^{Reference Holtz, Yu, Gunceler, Gao, Sundararaman, Schwarz, Arias, Abruña and Muller68} obtained in a closed cell using energy-filtered TEM imaging. A valence energy loss peak at 5 eV was selected because it is unique for FePO₄ but not LiFePO₄. This allows determination of the lithiation state of a LiFePO₄ electrode and the surrounding aqueous electrolyte in real time and nanoscale resolution during electrochemical charge and discharge. In Figure 4, delithiation occurs particle-by-particle, with slow nucleation and growth starting at the edge and moving through each particle. This behavior supports an anisotropic growth model. Open-cell results^{Reference Zhu, Wang, Liu, Liu, Kushima, Liu, Xu, Mao, Li, Wang and Huang75} are consistent with this model in that many of the particles appear inactive at any moment. Li enrichment and depletion captured during lithiation and delithiation highlight the advantages of nanoscale imaging under operando conditions.

Figure 4.

Energy-filtered transmission electron microscopy (EFTEM) data showing structural and chemical evolution of a LiFePO₄ cathode, with an activated carbon counter electrode and an aqueous 0.5 M Li₂SO₄ electrolyte. (a) EFTEM images (5 eV) of charging and discharging at indicated times. Scale bar = 400 nm. Bright regions are delithiated FePO₄, and dark regions are LiFePO₄. White arrows point toward “bright” charged particles, and black arrows point toward “dark” discharged particles. (b) Current profile corresponding to a charge rate of 10 C. Time t = 0 s is the start of the third charge cycle after assembly. Arrows on the top axis indicate the times of the images in (a). (c–d) Integrated intensity over various regions, shown as voltage, from the regions indicated by boxes in (a). The solution becomes dark during discharge and returns to the background level during charge. Regions of the particle light up and disappear, potentially upon delithiating and fracturing off the particle cluster. Reproduced with permission from Reference 68. © 2014 American Chemical Society.

Solid–electrolyte interphase formation in closed cells

The complex reactions occurring at the electrode–electrolyte interface lead to the formation of a passive layer, SEI, at the electrode surface.^{Reference Kang Xu, von Cresce and Lee76} The SEI is fundamental to battery performance and lifetime because it controls the passivation, stability, and impedance of the electrode. The chemistry and evolution of the SEI with cycling are therefore key topics for liquid-cell studies. Two recent closed-cell reports highlighted several aspects of SEI formation. During electrochemical lithiation and delithiation of Au anodes in LiPF₆/ethylene carbonate (EC)/diethyl carbonate electrolyte,^{Reference Zeng, Liang, Liao, Xin, Chu and Zheng70} lithiation occurs inhomogeneously, showing lithium metal dendritic growth, electrolyte decomposition, and SEI formation during cyclic voltammetry. The SEI layer develops uniformly on one Au electrode, concurrently with Li dendrite growth on the other electrode. In another observation using LiPF₆/EC/dimethyl carbonate,^{Reference Sacci, Dudney, More, Parent, Arslan, Browning and Unocic69} a dendritic SEI appeared on an Au electrode, forming prior to Li deposition and remaining after Li dissolution. Neither of these studies included a lithium metal source; thus, all of the phenomena observed were the result of electrolyte decomposition. Experiments that use a Li source and a reference electrode would provide a systematic and quantitative set of data addressing the key questions of SEI formation kinetics and stability.