In situ transmission electron microscopy

One of the major challenges in in situ TEM of ferroelectric phenomena is the method by which an electric field is applied across the specimen. In early studies, the electric field was applied by two parallel TEM copper grids.^{Reference Yamamoto, Yagi and Honjo61–Reference Randall, Barber and Whatmore63} In this configuration, the nominal electric field is parallel to the electron beam, and the magnitude of the actual field applied to the imaging area of the specimen is difficult to estimate because of the presence of vacuum gaps between the copper grid electrodes and the specimen. Tan and co-workers modified this electrode configuration by depositing gold films directly onto the TEM specimen surface to generate an electric field perpendicular to the beam direction and to better quantify the actual field.^{Reference Tan, He and Shang64–Reference Qu, Zhao and Tan66} In recent work, a small specimen was cut from a single crystal of lead magnesium niobate-lead titanate and attached between two parallel electrodes on a specially designed TEM grid using a focused ion beam, and the TEM grid was then loaded on an electrical biasing specimen holder.^{Reference Sato, Hirayama and Ikuhara67}

Parallel electrodes can also be made during thin-film growth. For example, Winkler and co-workers prepared samples by growing an epitaxial layer of the electrode material SrRuO₃ on SrTiO₃ and then using lithography and a precision ion-milling step to define the complex planar electrode structure. A second growth step produced an epitaxial layer of BiFeO₃ on top of the device structure.^{Reference Winkler, Damodaran, Karthik, Martin and Taheri44,Reference Winkler, Jablonski, Damodaran, Jambunathan, Martin and Taheri68} These structures permit the application of only in-plane electric fields. Applications, however, generally involve thin sections or thin-film materials with planar surface electrodes where switching occurs through the material thickness, in a direction normal to the surface. The recent integration of a scanning probe into a TEM specimen holder makes in situ experiments much easier.^{Reference Nelson, Gao, Jokisaari, Heikes, Adamo, Melville, Baek, Folkman, Winchester, Gu, Liu, Zhang, Wang, Li, Chen, Eom, Schlom and Pan39,Reference Chang, Kalinin, Yang, Yu, Bhattacharya, Wu, Balke, Jesse, Chen, Ramesh, Pennycook and Borisevich46,Reference Zhang, Xiang, He, Seidel, Zeches, Yu, Yang, Wang, Chu, Martin, Minor and Ramesh69,Reference Gao, Britson, Nelson, Jokisaari, Duan, Trassin, Baek, Guo, Li, Wang, Chu, Minor, Eom, Ramesh, Chen and Pan70} Using such an in situ holder, an electric field can be induced along the film normal by applying a voltage between the surface probe and a planar epitaxial bottom electrode. This geometry can be used for surface probe characterization as depicted schematically in Figure 1c. The results discussed in this review were largely obtained with probe-type in situ holders.

In most in situ TEM observations of domain switching, the domain evolution is observed by diffraction contrast imaging, which shows a strong bright–dark contrast of domains.^{Reference Nelson, Winchester, Zhang, Kim, Melville, Adamo, Folkman, Baek, Eom, Schlom, Chen and Pan40,Reference Chen, Katz, Pan, Das, Kim, Baek and Eom71–Reference Li, Gao, Nelson, Jokisaari, Zhang, Kim, Melville, Adamo, Schlom and Pan74} In this case, the contrast comes from careful choice of the diffraction vector and use of an appropriate aperture to form diffraction contrast images. This imaging mode also lends distinct contrast to defects such as dislocations, which is useful for observing their interactions with moving domain walls. In general, diffraction contrast imaging does not allow atomic-scale resolution. However, a broad range of techniques are usually integrated in advanced TEM instruments. These include high-resolution TEM (HRTEM), high-angle annular dark-field (HAADF), scanning transmission electron microscopy (STEM), and electron energy-loss spectroscopy (EELS). With the appropriate microscope, the determination of structure and chemical composition with atomic resolution is thus possible. Recent advances in spherical aberration-corrected HRTEM/STEM imaging with sub-angstrom resolution make it possible to directly map local strain, atomic displacements, and thus polarization at the atomic scale.^{Reference Nelson, Winchester, Zhang, Kim, Melville, Adamo, Folkman, Baek, Eom, Schlom, Chen and Pan40,Reference Chang, Kalinin, Yang, Yu, Bhattacharya, Wu, Balke, Jesse, Chen, Ramesh, Pennycook and Borisevich46,Reference Li, Gao, Nelson, Jokisaari, Zhang, Kim, Melville, Adamo, Schlom and Pan74–Reference Lubk, Rossell, Seidel, Chu, Ramesh, Hytch and Snoeck77} Used in conjunction with in situ TEM, advanced imaging techniques allow direct measurement of the microscopic mechanism of ferroelectric switching.

Switching dynamics

Early theoretical calculations by Landauer^{Reference Merz78} analyzed the free energy of nucleating a ferroelectric domain and suggested that the activation energy and consequently the critical nucleus size required to initiate switching were impossibly large. This conclusion was clearly at odds with observations of nucleation-dominated switching and the critical voltage required^{Reference Lee, Baek, Kim, Yoon, Folkman, Eom and Noh79} and became known as “Landauer’s paradox,” which suggests that the switching between ferroelectric orientations is entirely governed by defects that provide a localized decrease in the nucleation barrier. In ferroelectric thin-film heterostructures, the most prominent and voluminous defects are the two interfaces, which not only break the crystal symmetry but also are accompanied by physical and chemical reconstructions, bound charges, space charges, and strain.^{Reference Schlom, Chen, Eom, Rabe, Streiffer and Triscone47,Reference Lee, Baek, Kim, Yoon, Folkman, Eom and Noh80–Reference Okatan, Misirlioglu and Alpay82} These interfaces are the preferential nucleation sites and determine the characteristics of ferroelectric switching,^{Reference Jesse, Rodriguez, Choudhury, Baddorf, Vrejoiu, Hesse, Alexe, Eliseev, Morozovska, Zhang, Chen and Kalinin19,Reference Tagantsev and Gerra83} as was demonstrated recently by in situ TEM observations.

A typical ferroelectric switching process observed by in situ TEM is shown in Figure 2a, where 180° in situ switching of a p-type PZT film is caused by a probe-induced electric field.^{Reference Gao, Nelson, Jokisaari, Baek, Bark, Zhang, Wang, Schlom, Eom and Pan43} Switching follows the contemporary switching model: Nucleation occurs at the surface because of the local high free energy there, followed by rapid forward growth forming sharp needle-like domains. After reaching the opposite interface, the domain walls relax to vertical planes that propagate slowly in the lateral direction. The fast forward growth is due to the high energy of the forward edge of the domain, which carries accumulating bound charge. The slow lateral growth is typically modeled by the formation and growth of atomic-scale steps on the domain wall.^{Reference Miller and Weinreich84,Reference Shin, Grinberg, Chen and Rappe85}

Figure 2.

Domain propagation, pinning, and the origin of asymmetric switching where nucleation occurs at different interfaces of a lead zirconate titanate (PZT) thin film: (a) Growth of the domain during the initial switching cycle. The lines indicate the extents of the domain at evenly spaced time intervals depicted by color. Selected corresponding transmission electron microscope (TEM) images for 69, 414, and 1,345 ms cases are shown. (b) Reverse switching of the domain under reverse bias. Selected corresponding TEM images for 0, 69, and 1,345 ms cases are shown. (c) A schematic model system used to calculate the electric field through the film, showing the distribution of space charge corresponding to the formation of two depletion regions. (d) The corresponding schematic of built-in electric field (black line labeled E _bi) formed by the Schottky junctions and additional external fields, E _ext. (e) The calculated electric field accounting for the tip effect on the external electric field. The curves correspond to 0 V (black), + 0.4 V (red), and –4.8 V (blue). In (d) and (e), the largest positive field and negative field occur on opposite interfaces preferentially nucleating “up” and “down” polarized domains, respectively. Note: P, polarization; SRO, strontium ruthanate (SrRuO₃); STM, scanning tunneling microscopy. Reproduced with permission from Reference 43. © 2011 Nature Publishing Group.

Although this overview effectively describes the switching process in general, there are many significant details that depend sensitively on the structure and electrical details of the system. For example, during 180° switching of a p-type PZT film (Figure 2a–b), nucleation of “down”- and “up”-polarized domains is asymmetric: In the former, nucleation occurs at the top surface (Figure 2a), whereas in the latter, it occurs at the PZT–electrode interface (Figure 2b). Similar effects are also observed when parallel top and bottom electrodes are used.^{Reference Lee, Shin, Song, Choi, Shin, Park, Britson, Cao, Chen, Lee and Oh42} This can be attributed to the built-in electric fields formed by Schottky junctions at the ferroelectric–electrode (semiconductor–metal) interfaces,^{Reference Balke, Gajek, Tagantsev, Martin, Chu, Ramesh and Kalinin86,Reference Sze, Coleman and Loya87} owing to the fact that the work functions of strontium ruthanate (SrRuO₃, SRO) and tungsten electrodes are larger than the electron affinity of PZT.^{Reference Hartmann, Neilson, Lamb, Watanabe and Scott88,Reference Muller89} In this case, the depletion of holes in p-type PZT near the electrodes (Figure 2c) produces a built-in electric field E _bi that has maxima at the interfaces (black line in Figure 2d). Under an external field E _ext (Figure 2d shows a homogenous field, and Figure 2e shows the probe-induced inhomogenous field), the total electrical field results in the largest positive field occurring at the PZT–electrode interface during nucleation of an up domain and the largest negative field occurring at the PZT–probe surface during nucleation of a down domain, which corresponds exactly with the observed nucleation sites. In n-type BFO, the built-in electric fields are opposite in sign to those in p-type PZT.^{Reference Nelson, Gao, Jokisaari, Heikes, Adamo, Melville, Baek, Folkman, Winchester, Gu, Liu, Zhang, Wang, Li, Chen, Eom, Schlom and Pan39,Reference Clark and Robertson90} Thus, nucleation of down-polarized domains during switching occurs at the bottom interface (Figure 3)^{Reference Nelson, Gao, Jokisaari, Heikes, Adamo, Melville, Baek, Folkman, Winchester, Gu, Liu, Zhang, Wang, Li, Chen, Eom, Schlom and Pan39} and vice versa for up-polarized domains. Formation of domain nuclei in BFO was especially anomalous, as the nuclei at the BFO–electrode interface were metastable. As shown in Figure 3, multiple nucleation events occurred, and the nuclei did not continue their growth before surpassing a critical voltage of 2.2 V. This is contrary to conventionally held kinetic growth models^{Reference Dawber, Rabe and Scott21} and points to a thermodynamic restriction on domain growth, specifically a local energy well favoring the down-polarized domains in the vicinity of the electrode.

Figure 3.

Growth of a switched domain in bismuth ferrite (BiFeO₃ or BFO): A chronological transmission electron microscope (TEM) image series depicts the evolution of a “down” polarized domain from an “up” polarized single-domain film. Below each TEM image, a schematic of the corresponding domain configuration of the system is shown. (1–3) Nucleation occurred at the electrode interface at 0.9 V, producing a metastable stationary domain. Additional bias increased the size and number of these nucleated domains (white arrows in micrographs) up to a bias of (4) 2.2 V. (5) At 2.2 V, the domains merged and propagated forward until they stabilized just short of the surface, where the top horizontal domain wall became pinned with the accumulation of negative bound charge. (6) At 4 V, the lateral extent of the domain increased slightly, but the top horizontal domain wall remained pinned. (7) The lateral extent of the domain is reduced slightly with the removal of the bias, and (8) remains unchanged after the tip is removed. Reproduced with permission from Reference 39. © 2011 Association for the Advancement of Science. Note: SPM, scanning probe microscopy; P, polarization; La_0.7Sr_0.3MnO₃, lanthanum strontium manganite; TbScO₃, terbium scandate.

In nonferroelastic 180° switching of PZT films, multiple metastable nuclei are not observed.^{Reference Gao, Nelson, Jokisaari, Baek, Bark, Zhang, Wang, Schlom, Eom and Pan43} The orientation of the switched domain, however, is strongly affected by the applied fields and built-in fields at the interfaces, resulting in the formation of metastable inclined domain walls under bias.^{Reference Gao, Nelson, Jokisaari, Zhang, Baek, Bark, Wang, Liu, Li, Eom and Pan41} A typical example of this process is shown in Figure 4. Under bias, the inhomogeneous electric field (Figure 4c) results in the inclination of the lateral domain walls of the switched domain (Figure 4a–b). The head-to-head polarization at this domain wall is stable only when a field is actively applied, and upon the reduction (or removal) of the field, the domain walls relax to the equilibrium vertical orientation. If the domain is sufficiently small (e.g., ∼75-nm midfilm width, as in Figure 4a), this process brings the opposite domain walls into close proximity, and they annihilate one another. Another switching event within the same region, using a longer voltage pulse as shown in Figure 4b, results in a larger lateral domain size. The inclined lateral edges still undergo the same loss in area from reorientation; however, their additional separation prevents complete back-switching from occurring. Therefore, in writing nanosized domains, the domain wall relaxation can drive back switching, resulting in a predictable area loss at the domain edge and imposing a thickness-dependent critical size on the stable domain width.

Figure 4.

Effects of the applied and internal electric fields on the orientation of switched domain: (a–b) Chronological transmission electron microscope (TEM) image series depict a domain switching in lead zirconate titanate (PZT). Below each TEM image is a schematic of the corresponding domain configuration of the system. (a) For lower bias (–6.6 V), a triangular domain with charged walls forms. When the bias is reduced to –3.1 V, the charged domain walls relax and meet, annihilating the switched domain, showing the domain to be unstable. (b) At higher bias (–8.7 V), the domain grows larger, forming a trapezoid. The angled, charged domain walls again relax to the vertical, but the core of the domain remains stable as the bias is removed. (c) Contour plots of the calculated distribution of (left) the built-in electric field in a fully depleted PZT film resulting from the Schottky barrier between PZT and SRO, (middle) the vertical component of the external electric field in PZT with a tip contact width of 10 nm and voltage of –6 V, and (right) the combination of the built-in and external electric fields, corresponding closely to the triangular domain shape initially adopted during switching. Note: P, polarization; SRO, strontium ruthanate (SrRuO₃); θ, domain wall inclination angle. Reproduced with permission from Reference 41. © 2012 Wiley-VCH.

The applied fields and intrinsic electric fields at the interfaces also influence the growth rates of ferroelectric domains and result in a significant asymmetry in domain wall mobility.^{Reference Gao, Nelson, Jokisaari, Baek, Bark, Zhang, Wang, Schlom, Eom and Pan43}Figure 5a–b shows domain wall velocities along the vertical (black) and lateral (left, red; right, blue) directions for the negative switching branch (top surface is positively biased, switching from upward to downward polarization) and the positive switching branch, respectively, which are measured for the switching of Figure 2. Vertical switching during the positive branch is a minimum of 3.6 times faster than that during the negative branch. This trend is reversed for the lateral creep rates, which favor the negative switching branch. The disparity in switching rates can be explained by inhomogeneity in the electric field along the film depth. Specifically, with positive bias, the down-polarized domain is opposed by the positive field from the Schottky contact at the bottom electrode (red line in Figure 2e), which slows its progress. In contrast, with negative bias, the negative built-in field at the tip contact is countered by the strong tip-induced field, and switching occurs rapidly (blue line in Figure 2e). Similar effects are also observed when the electric field is induced by parallel top and bottom electrodes.^{Reference Lee, Shin, Song, Choi, Shin, Park, Britson, Cao, Chen, Lee and Oh42} The underlying mechanism, however, is different. In the latter case, the two electrodes are distinguished by their interface structures, stemming from different film deposition methods. This leads to a differing charge screening capability at the two electrodes and thus a variation in the depolarization field, resulting in asymmetric domain wall mobility.^{Reference Lee, Shin, Song, Choi, Shin, Park, Britson, Cao, Chen, Lee and Oh42}

Figure 5.

Effects of the applied and internal electric fields on switching dynamics: (a–b) Domain wall velocities in the vertical (black) and lateral (left, red; right, blue) directions, derived from the switching as in Figure 2, for (a) forward (negative) and (b) reverse (positive) bias segments (positive and negative are defined with respect to the film polarization not the bias.) The vertical domain wall velocity at the beginning of the reverse switching (black arrow in b) is greater than 2900 nm/s and thus has not been incorporated in the plot. (c) Hysteresis loop derived from measuring the switched area at a given bias, the shift of which coincides well with the electric field in the film. The line colors in the sequence correspond to the initial negative switching (black), negative saturation (red), positive switching (yellow), positive saturation (blue), and second negative switching event (green). Note: (c) “a” refers to the switching process in (a), and “b” refers to the switching process in (b). Reproduced with permission from Reference 43. © 2011 Nature Publishing Group.

In summary, characteristics of ferroelectric switching in thin films, such as nucleation sites, domain growth rates, and the orientation and mobility of switched domain walls, are determined by applied electric fields and intrinsic fields formed at ferroelectric–electrode interfaces. These microscopic characteristics are manifested in typical electrical measurements, such as hysteresis loops. Using in situ TEM, a hysteresis loop can also be measured in terms of the total switched domain area, rather than the classical polarization derived from electrical measurements. This method is advantageous because it directly measures the property of interest, the region undergoing a change in spontaneous polarization, while ignoring leakage currents, dielectric contributions, and electronic noise. The hysteresis loop in Figure 5c summarizes how the applied and built-in fields create several asymmetries in this switching behavior, including asymmetric nucleation bias (horizontal shift of –2.2 V) and switching rates (disparity in the slopes of the positive [yellow] and negative [black and green] switching branches).

Domain wall pinning in ferroelectrics

During the switching of ferroelectrics, defects serve not only as nucleation sites but also as pinning centers for moving domain walls. As discussed earlier, in thin films, the concentration of the electric field in the vicinity of the interfaces typically results in nucleation there. Hence, the interactions between switched domains and other types of defects, such as dislocations, oxygen vacancies, and preexisting immobile ferroelastic domain walls, are usually manifested as pinning effects. A fundamental understanding of the mechanisms of pinning requires the microstructures of the defects to be known. Atomic resolution afforded by aberration-corrected TEM is ideal for characterizing such defects. In current in situ TEM work, observations at relatively low magnification (less than 100k) of domain evolution are often combined with ex situ high-resolution STEM imaging of atomic details to uncover the microscopic mechanisms underlying the interactions of defects and domain dynamics.

When the thickness of a ferroelectric thin film is above a critical value, the epitaxial strain can be relieved through the formation of misfit dislocations or ferroelastic domain structures. In such systems, the interactions between the switched domains and these dislocations or preexisting ferroelastic domains dominate the domain wall propagation kinetics. Figure 6a shows an example of a laterally moving 180° domain wall affected by a mirrored pair of edge dislocations in PZT, where the pinning effect is relatively weak.^{Reference Gao, Nelson, Jokisaari, Baek, Bark, Zhang, Wang, Schlom, Eom and Pan43} This is evidenced by an increase in the domain wall velocity toward the dislocation and a subsequent decrease. The atomic structure surrounding the dislocations can be viewed in a high-resolution HAADF image (Figure 6b). The lattice distortion bounded by two misfit dislocations is shown in a Fourier-filtered image using (200) reflections, with the dislocation cores labeled in Figure 6c–e. The regions of compressive and tensile stress are mirrored by the pair of dislocations and cancel when averaged along the domain wall plane, ruling out significant net piezoelectric or flexoelectric effects. The vertical 180° domain wall itself is neither ferroelastic nor charged, precluding strong elastic or electrostatic interactions with the dislocation. This suggests that the pinning is the result of a reduction in energy caused by the presence of coincident regions of disorder.^{Reference Gao, Nelson, Jokisaari, Baek, Bark, Zhang, Wang, Schlom, Eom and Pan43}

Figure 6.

Domain pinning at dislocations in lead zirconate titanate: (a) A contour map of the domain outline at each time step with the color corresponding to the domain wall velocity. Line profiles for the initial vertical velocity and lateral velocity versus time of the left domain wall are overlaid. In both cases, a pinning event is highlighted by a white arrow. The inset shows the transmission electron microscope image of this region prior to switching. (b) High-resolution high-angle annular dark-field image and (c) the same image filtered in Fourier space (inset) by including only the (h00) lattice plane frequencies, showing the details of the pinning defect, composed of two edge dislocations. (d–e) Magnifications of the dislocation cores are pointed out with black arrows. Note: For (c–e), the original grayscale contrast was converted to a false color scale. Reproduced with permission from Reference 43. © 2011 Nature Publishing Group.

In epitaxial thin films, another type of defect, stress-relieving ferroelastic domains, is usually nearly immobile because of the strong coupling between charge and strain in the system. An in situ switching study provided direct experimental evidence that domain switching can be hindered by these ferroelastic domains in PZT thin films.^{Reference Gao, Britson, Jokisaari, Nelson, Baek, Wang, Eom, Chen and Pan45}Figure 7a shows results of an in situ experiment in which a 180° domain grew as the voltage was increased from 0 to 10 V. At this point, with the voltage held constant, the domain grew slowly, terminating once the created 180° domain reached the edge of the 90° ferroelastic domain. These results show that the ferroelastic domain wall acted as an obstacle to the propagation of the ferroelectric domain wall. The switched domain was stable after removal of the electric field, and the hindering effect could be overcome by applying a higher bias, allowing the ferroelectric domain to pass through the ferroelastic one.^{Reference Gao, Britson, Jokisaari, Nelson, Baek, Wang, Eom, Chen and Pan45}

Figure 7.

Ferroelastic and ferroelectric domain interactions: (a) A chronological series of transmission electron microscope (TEM) images depicts growth of a 180° ferroelectric domain intersecting with a 90° ferroelastic domain. Below each TEM image is a schematic of the corresponding domain configuration of the system. During this process, the 180° domain is eventually pinned against the 90° domain, resulting in a charged domain wall. (b) Map of lattice tilt angle based on geometric phase analysis calculations overlaid on the corresponding atomic-resolution high-angle annular dark-field image. (c) Map of lattice ratios of out-of-plane to in-plane Pb atoms calculated from the Pb sublattice of the same region. (d) Polarization angle map showing the transition zone, highlighted by the white dashed line, between the 180° domain and the 90° domain. (e) Polarization vector map corresponding to the squares numbered 1–3 in (d), respectively. Analysis of (b–c) reveals that the strains around the charged boundary show sharp boundaries, whereas in (d), the polarization shows the formation of a dipole-glass-like structure at the charged interface, which stabilizes the domain wall. Note: P, polarization; PZT, lead zirconate titanate; SRO, strontium ruthanate (SrRuO₃). Reproduced with permission from Reference 45. © 2013 Nature Publishing Group.

Atomically resolved HAADF images of the domain wall between the ferroelastic and ferroelectric domains provided an explanation of the hindering mechanism.^{Reference Gao, Britson, Jokisaari, Nelson, Baek, Wang, Eom, Chen and Pan45} Mapping of the ferroelastic variants around the domain wall using both geometric phase analysis (Figure 7b) and lattice plane spacing determined from the Pb atom positions (Figure 7c) showed that the change in lattice orientation remained sharp across both domain walls during the switching process and indicated that no ferroelastic switching or crystal structure change occurred around the 90° domain wall during this process. The corresponding polarization angle map (Figure 7d) calculated from the Pb sublattice offset, however, showed the upper wall to be diffuse. A transition zone with a thickness of ∼10 unit cells was identified in which the polarization vectors were randomly oriented and had substantially reduced magnitudes, resulting in a glassy polarization configuration, or dipole glass (Figure 7e, middle). The uncharged 90° domain wall in the bottom-left corner, however, was unaffected, and the polarization angle map showed a sharply defined boundary similar to that seen before switching (Figure 7e, right).^{Reference Gao, Britson, Jokisaari, Nelson, Baek, Wang, Eom, Chen and Pan45}

In situ switching during TEM has also provided direct experimental support for the contemporary theory of domain pinning at defects as a source of ferroelectric fatigue. Significant defect–domain wall interactions were observed in BFO films,^{Reference Nelson, Gao, Jokisaari, Heikes, Adamo, Melville, Baek, Folkman, Winchester, Gu, Liu, Zhang, Wang, Li, Chen, Eom, Schlom and Pan39} where a switched domain often became pinned along (001)_P (the subscript P represents pseudocubic indices) planes after several cycles and no longer extended through the full thickness between the two electrodes, resulting in incomplete switching (see panels 5–8 in Figure 3). Once formed, such pinning sites exhibited perfect repeatability during cyclic switching, pointing to immobile defects as pinning sites. High-resolution STEM images showed that these sites had no apparent crystalline defects, or cation disorder, suggesting ordered planes of charged oxygen vacancies, which are common in perovskite films,^{Reference Scott and Dawber91} as the probable cause for pinning of the negatively charged domain walls. It should be noted that this pinning could be overcome by applying a sufficiently large bias or repositioning the probe, confirming that the switched domain was thermodynamically stable beyond the pinning site.^{Reference Nelson, Gao, Jokisaari, Heikes, Adamo, Melville, Baek, Folkman, Winchester, Gu, Liu, Zhang, Wang, Li, Chen, Eom, Schlom and Pan39}

To summarize, domain interactions with various types of defects have been shown to influence, and in many cases control, the switching kinetics and the configuration of the switched domain wall through pinning effects. In situ TEM studies allow direct interrogation of these effects and provide insight into the underlying physical processes correlated with specific defective microstructures. Such fundamental understanding is critical in modifying the properties of ferroelectric materials and engineering applicable devices.

Outlook

Development in the field of ferroelectric materials and devices continues apace, with improvements in the quality of thin-film growth on complex oxide substrates,^{Reference Schlom, Chen, Fennie, Gopalan, Muller, Pan, Ramesh and Uecker92} device designs leveraging multiferroic materials,^{Reference Chu, Martin, Holcomb, Gajek, Han, He, Balke, Yang, Lee, Hu, Zhan, Yang, Fraile-Rodriguez, Scholl, Wang and Ramesh16,Reference Krieger93} and even new materials such as organic/perovskite hybrid systems.^{Reference Grinberg, West, Torres, Gou, Stein, Wu, Chen, Gallo, Akbashev, Davies, Spanier and Rappe94} In conjunction with these developments, in situ TEM methods provide direct access to atomic-scale details of ferroelectric structure and properties with spatial and time resolution unavailable by any other methods. Further refinements of in situ methods, including new holders for biasing, scanning tunneling microscopy, force application, and even atmospheric control are becoming available (see the Crozier and Hansen article in this issue), opening the door to examining the behavior of these materials under many different conditions, all within the transmission electron microscope. Concurrently, much progress has been made in improving TEM instruments, with ultrahigh-aberration-corrected microscopes with sub-angstrom resolution (see the Sun et al. article in this issue) becoming available from many vendors, and coincidental improvements in the sensitivity and resolution of analytical tools such as energy-dispersive spectroscopy and EELS. Progress is also being made in time resolution, with ultrafast dynamic TEM methods becoming available. The combination of new TEM methods with further development of in situ platforms provides an immense opportunity to develop a deeper understanding of the fundamental properties of ferroelectrics and aid in the continued evolution of ferroelectric materials and devices.