A Review of Grain Boundary and Heterointerface Characterization in Polycrystalline Oxides by (Scanning) Transmission Electron Microscopy

Interfaces such as grain boundaries (GBs) and heterointerfaces (HIs) are known to play a crucial role in structure-property relationships of polycrystalline materials. While several methods have been used to characterize such interfaces, advanced transmission electron microscopy (TEM) and scanning TEM (STEM) techniques have proven to be uniquely powerful tools, enabling quantification of atomic structure, electronic structure, chemistry, order/disorder, and point defect distributions below the atomic scale. This review focuses on recent progress in characterization of polycrystalline oxide interfaces using S/TEM techniques including imaging, analytical spectroscopies such as energy dispersive X-ray spectroscopy (EDXS) and electron energy-loss spectroscopy (EELS) and scanning diffraction methods such as precession electron nano diffraction (PEND) and 4D-STEM. First, a brief introduction to interfaces, GBs, HIs, and relevant techniques is given. Then, experimental studies which directly correlate GB/HI S/TEM characterization with measured properties of polycrystalline oxides are presented to both strengthen our understanding of these interfaces, and to demonstrate the instrumental capabilities available in the S/TEM. Finally, existing challenges and future development opportunities are discussed. In summary, this article is prepared as a guide for scientists and engineers interested in learning about, and/or using advanced S/TEM techniques to characterize interfaces in polycrystalline materials, particularly ceramic oxides.


How Are GBs and HIs Characterized?
As mentioned earlier, characterizing individual GBs and HIs is a considerable challenge due to the requisite structural and chemical sensitivity on the length of single unit cells. Moreover, sampling a statistically relevant number of GBs/HIs often requires characterizing many tens, hundreds, or thousands of individual GBs/HIs. The former challenge relates to the spatial resolution of characterization techniques, which limit the suitable techniques to electron microscopy (EM), atom probe tomography (APT), X-ray photoelectron spectroscopy (XPS) [70] on intergranular fracture surfaces, electrochemical strain microscopy (ESM) [71], scanning tunneling microscopy (STM) where the GB intersects the surface, and atomic force microscopy (AFM) for studying GB energy using grooves [72]. As useful as the current characterization techniques are, they are still suffering from a tradeoff between spatial resolution and analyzed sample volume within a reasonable time. Among these techniques, EM techniques have for decades offered significant contributions to interfacial characterization [73,74]. For example, TEM phase contrast imaging was the key to the identification of ion-blocking amorphous phases at the GBs, affecting oxygen ion conductivity [75,76]. STEM annular bright field (ABF) imaging upgraded with the help of negative Cs STEM aberration correction was used to directly image oxygen and oxygen vacancies. This information helps shed light on the underlying mechanisms driving GB/HI defect chemistry, paving the way for modeling GBs and their properties in future [8].
In addition to EM, APT is useful for quantitative probing of GBs/HIs, owing to its sufficient spatial resolution and chemical sensitivity. APT is a combination of field ion microscope and mass spectrometer, capable of accurately reconstructing the morphology and chemical composition of the specimen in three dimensions at the nanoscale [77]. The detection efficiency (i.e., the ratio of signals detected to signals produced) of APT is around 30-80%, which is reasonable among other techniques (up to 80% at low count rates for STEM-EELS and <1% for STEM-EDXS) [78]. The APT technique has been used to measure segregations of defects in oxide GBs in addition to learning about charge density and electrostatic potential [79]. This technique is particularly beneficial in ceramics including oxides due to the combination of surface band bending under high electric field and/or surface defects, which make accumulation of charge on the surface possible, and photons of lower energies can be absorbed [80]. However, data reconstruction can be a limiting factor, as it can potentially create artifacts [81]. APT has been coupled with STEM imaging and spectroscopy to provide a complementary understanding of GB structure and chemistry in Hf and La doped-alumina and Y-doped ceria [82,83]. Just like S/TEM, extensive optimization of specimens is necessary to obtain reliable APT results [84]. In a work by Diercks, et al. [79], GB segregation was observed in a 10% and 30% Nb-doped ceria (NDC10 and NDC30) using APT. APT is an inherently destructive method, as illustrated by the overlay of TEM micrograph before and after APT shown in Figure 4a. The Nd% map is shown in Figure 4b, where fraction of Nd ions is illustrated for clarity. There is an evident increase in the Nd concentration at the GB compared with the grain bulk. Figure 4c shows the same volume as in Figure 4b  (a) Overlay of transmission electron microscope images of an atom probe NDC10 specimen before and after atom probe analysis. A GB is observed in the before image. The after image indicates that the atom probe analysis encompassed the entire GB region. Atom probe data reconstructions showing (b) Nd, and (c) a 66 at% oxygen iso-concentration surface [79]. Reproduced from [79] with permission from The Royal Society of Chemistry. Quantitative 3-D maps of the elements as well as segregation of Nd, Al and Si coupled with O depletion at the GB are presented in Figure 5. This information was used to model the interfaces and correlate them with electrochemical impedance spectroscopy data to yield 3-D potentials at the GBs. At 30% Nb, local gaps in the 3-D potentials suggest conduction pathways through the potential barrier at the GB [79]. There is also evidence of Al and Si enrichment at the GB [79]. Reproduced from [79] with permission from The Royal Society of Chemistry.
GBs in polycrystalline oxides have also been investigated using X-ray photoelectron spectroscopy (XPS) [70], a widely used surface analytical method. Researchers used XPS to analyze the fracture surfaces in Mg-doped alumina to assess Mg GB segregation. The concentrations of Mg and Ca at and near the GB are measured and shown in Figure 6. While Mg has much higher concentration in the bulk comparing to Ca, the segregation of Ca to the GB is more significant.
Scanning probe techniques such as ESM [71] and AFM measure local changes in the surface topography of polycrystalline oxides and have proven useful for quantifying GB energies and mapping functional properties. GB energy plays an important role in segregation phenomena. GB segregation generally occurs to lower the energy of the GBs. Both experiments [85] and simulation [86] studies have shown that GB energies can change with temperature and have a strong effect on microstructure development. A common scanning probe technique that allows measurement of relative GB energy ( Scanning probe techniqu surface topography of polycr energies and mapping functio gation phenomena. GB segreg experiments [85] and simula with temperature and have a scanning probe technique tha ϒ GB/ϒS) is AFM. When a so due to preferential removal o formed by etching can then c ergy [87]. Many studies have urements [88][89][90][91].

A Brief Introduction to TEM
The transmission electro ful tools for directly imaging down to the atomic scale. Boo STEM [98][99][100][101], and the vari basic introduction to these te crystalline oxides is provided Scanning probe tech surface topography of p energies and mapping fu gation phenomena. GB s experiments [85] and si with temperature and h scanning probe techniqu ϒ GB/ϒS) is AFM. When due to preferential remo formed by etching can t ergy [87]. Many studies urements [88][89][90][91].

A Brief Introduction t
The transmission el ful tools for directly im down to the atomic scale STEM [98][99][100][101], and the basic introduction to the crystalline oxides is prov S ) is AFM. When a solid surface is polished and etched, the interfaces are revealed due to preferential removal of material near the GBs [72]. The geometry of the grooves formed by etching can then characterized by AFM and used to measure average GB energy [87]. Many studies have been conducted on single phase oxides for GB energy measurements [88][89][90][91].

A Brief Introduction to TEM and STEM
The transmission electron microscope (TEM) and scanning TEM (STEM) are powerful tools for directly imaging atomic arrangements, and quantitative chemical analysis down to the atomic scale. Books and review articles are written focusing on TEM [92][93][94][95][96][97], STEM [98][99][100][101], and the various analytical techniques available in both [93,[102][103][104][105]. A basic introduction to these techniques as they pertain to GB/HI characterization in polycrystalline oxides is provided here. In general, TEMs can be configured to operate in multiple modes. In "TEM mode" (also called conventional TEM) a broad electron beam provides parallel illumination to the specimen. In "STEM mode" the electron beam is converged to a point (or "probe") by pre-specimen electromagnetic focusing lenses and scanned in a raster pattern over the specimen.
Modern S/TEMs routinely demonstrate atomic resolution imaging and chemical analysis in two dimensions, and more recently in three dimensions [79,80,106]. TEM development dates to the era of de Broglie overcoming the resolution of a light microscope by electrons [107]. Later, TEM was built and demonstrated by Max Knoll and Ernst Ruska in 1933 [108]. The developments of spherical and chromatic aberration correctors for TEM and STEM in 1998 [109] and 2002 [110], respectively, extended the resolution limit to sub-Angstrom, creating novel opportunities for the accurate characterization of material defects, such as GBs and His [111]. In addition, enhanced electron and spectroscopic signal detection capabilities [112] have increased the ease of use by removing technical barriers such as resolution and signal-to-noise-ratio improvement [113]. Today, a variety of analysis techniques are available to probe important characteristics and functionalities of organic and inorganic specimens in solid, liquid, and gas phases. This review, highlights imaging, diffraction, and spectroscopy techniques suited to characterize interfaces within polycrystalline ceramics, such as atomic resolution imaging, microstructure analyses (e.g., grain orientation mapping), and chemistry mapping (e.g., elemental concentration and chemical segregation).

Imaging and Selected Area Electron Diffraction in the TEM
Due to the wavelike nature of electrons and periodic arrangement of atoms within a crystalline specimen, electrons scattered from the specimen generate a diffraction pattern that contains rich crystal structure information. From the position and symmetry of the diffraction spots, it is possible to determine the crystal structure, unit-cell parameters, lattice type, and defects such as twinning, while the intensities of the diffraction spots are related to the arrangement of atoms within the unit cell [114]. Selected-area electron diffraction (SAED) is the most used method to acquire this information, and it requires the microscope to operate in "diffraction mode". The ray diagram presented in Figure 7 explains the differences between diffraction and imaging modes in a TEM [92]. As in light optics, ray diagrams are often used in TEM to show how electron lenses control the electron beam by depicting the path electrons take as they traverse the electron column. As can be seen from Figure 7, the information contained in the diffraction pattern is also contained in the image However, the TEM image displays this information in real space while the diffraction pattern represents this information into reciprocal space, the two are related through a Fourier transform. Depending on the application, one or the other representation may prove to be more useful. Figure 7. Ray diagrams demonstrating two basic operation modes in a conventional TEM. In (a) diffraction mode, the DP is projected onto the viewing screen and (b) image mode, projecting the image onto the screen [92]. Switching between the modes requires changing the strength of the intermediate lens. A SAED pattern is acquired by isolating a small area of the specimen for diffraction analysis with an area-limiting aperture (usually 0.1 um or greater). In addition to crystal structure determination, the SAED pattern also contains information about overall sample crystallinity and can be used to establish orientation relationships between multiple diffracting crystals (e.g., between precipitates and a host matrix, or at an interface like GB). Further structural information is encoded into the diffraction pattern through the presence of features such as extra spots, the spot splitting, satellite spots, streaks and diffuse scattering that can indicate the presence of superlattice [115], point defect ordering, or extended defects [3], or can indicate particle shape [92]. Widely used techniques such as bright-field (BF) and dark-field (DF) TEM imaging, demonstrated in Figure 8, rely on filtering information from the SAED pattern to either exclude (BF-TEM) or preferentially select (DF-TEM) diffracted beams for image formation, generating contrast based on whether a particular region does or does not meet the Bragg condition such as examples in [116]. In the case of apertureless imaging, the transmitted and diffracted intensities recombine, and the diffraction contrast is suppressed, while the BF and DF TEM images show much better diffraction contrast due to the coexistence of large intensity in transmitted beam and large loss of intensity in the diffracted beam, and vice versa. For example, heavier phases tend to be darker in BF and brighter in the DF mode. While the BF-TEM helps learn about the morphology of the sample, DF can provide information about nanocrystal size and distribution in addition to defects such as dislocations, stacking faults and twinning.
Another useful technique is weak beam imaging that allows the formation of diffraction contrast in BF and DF modes using a weakly excited beam. The weak beam dark-field imaging (WBDF) approach is popular as it offers advantages over BF, such as having a stronger contrast that is easier to understand using simple physical models. WBDF is routinely used for highlighting the detailed defect structure, and is hence very suitable for lattice defect observation, including dislocations, stacking faults and grain boundaries [117]. Furthermore, high resolution TEM (HRTEM) images demonstrate the atomic arrangement at interfaces and enable observation of individual defects, which can be used to analyze secondary phases or impurities and can probe local crystallinity (degree of structural ordering) at the GBs [118]. One key consideration is that high resolution images require complementary image simulation before the image contrast can be conclusively assigned to an atomic arrangement.

Imaging Techniques in the STEM
In STEM mode, the electron beam is converged to form a probe and scanned across the specimen [98,99,119]. Imaging in the STEM is done primarily using elastically and inelastically scattered electrons which have interacted with, and passed through a thinned section of material, usually <300 nm. The scatter collection angle is critically important for image interpretation. The main imaging modes in STEM are presented in Figure 9 and explained below in the order of high to low scattering collection angles. They are often collected simultaneously and can be coupled with EELS and EDXS analysis. . Ray diagrams demonstrating the formation of BF, DF, and centered DF images by selecting specific beams in the diffraction plane with the objective aperture and reforming the filtered image at the imaging plane. (a) BF images are produced by selecting the direct beam while both (b) DF and (c) centered DF images are formed by filtering out the direct beam, using only one or more diffracted beams to form the image. Centered DF images are formed by tilting the incident beam so that the diffracted beam aligns with and is centered on the optic axis, which reduces aberrations. The bottom row shows the region of the diffraction pattern that is selected by the aperture in each case [92]. When high angle scattered electrons are collected which are incoherent, the image would be a high-angle annular dark-field (HAADF) STEM image. The contrast is generated based on the efficiency of the Rutherford scattering process at that angle of collection, and so is roughly proportional to the square of the atomic number. In addition, atomic positions are very clear since there are no contrast reversals. This technique is the most widely used but is not ideal for imaging light elements since the Rutherford scattering processing is inefficient for light atoms with smaller atomic nuclei.
Medium and or low-angle annular dark-field (MAADF or LAADF) STEM is another commonly used condition that collects lower angle coherent scatter. These images retain some of the characteristics of HAADF images (e.g., no contrast inversion, interpretability of atomic columns over a large focus range) and is also good at showing features that diffract strongly (strained areas such as dislocations, nanosized coherent or semi-coherent precipitates etc.).
Annular bright field (ABF) is when the inner part of the collection disc is collected which is dominated by phase contrast (thickness and defocus dependent). This method works very well with the aid of simulations and seems to show that the outer part of the BF disc produces images that are more incoherent and reveal all atoms as dark shadows, including light atoms such as oxygen [120] or hydrogen [121]. ABF pairs well with analytical STEM techniques where localized chemical information is desired, typically preferred to operate in STEM mode for chemical analysis. Bright-field (BF) STEM is analogous to bright-field TEM images attained in C-TEM. The angle of scatter is small, and the electrons are coherent. Just as in HRTEM, the contrast is critically dependent on sample thickness, microscope defocus and can show contrast inversions. The benefit of BF-STEM compared with HR-TEM is that it can often be collected simultaneously with ADF-STEM images of varying collection conditions (HAADF, MAADF/LAADF).
A comparison of STEM BF, STEM DF and HAADF STEM is given in Figure 10 for Hexoloy SiC.  Figure 10 reprinted from [122] with permission.

Specimen Preparation and Requirements for S/TEM
As discussed in the introduction, S/TEM specimens are in the form of a thin film or particles typically around <300 nm. There are several methods for preparing such thin specimens in oxide; the most useful ones are briefly explained below.

Focused Ion Beam (FIB) Lift-Out and Milling
A common way of preparing TEM specimens for materials-particularly ceramics and oxides-is using focused ion beams (FIB) [123]. In this method, a thin lamella of a specific thickness is etched and extracted from a specific area of the sample. First, a platinum or carbon layer is deposited on the surface to protect the material during ion milling. Then, an ion beam is used to mill two parallel trenches on both sides of the sample ( Figure 11). The FIB is usually equipped with an internal micromanipulator needle for extracting the lamella after milling and transferring to a TEM gird. This process can take place externally using a fine electrostatically charged glass needle. The thin lamella extracted from the bulk is then ready for final thinning and cleaning, before the TEM analysis [124]. To prepare an electron beam sensitive specimen such as biological materials, cryogenic-FIB (cryo-FIB) was developed to allow sample preparation at lower temperatures to reduce redeposition, thermal damages and preserve materials structure for subsequent analyses. Recently, cryo-FIB techniques have been extended to oxides for lithium-ion batteries, for example to characterize lithium dendrites within a frozen liquid electrolyte [125,126]. (c) welding platinum between the slice and the needle (micromanipulator) [124]. Reprinted from [124] with permission from Springer Science & Business Media.

Mechanical Polishing
This conventional method is still widely used today due to its simplicity and ability to create large areas of electron-transparent specimens. In most cases, mechanical polishing is used to obtain a flat surface in bulk materials, compacts, thin films and even fine embedded particles [124]. However, mechanical polishing needs to be followed by other final thinning methods such as Ar ion milling [127] and twin-jet electropolishing [128] to produce an electron transparent specimen. First, a small disc of~3 mm in diameter (i.e., the standard size accepted by TEM specimen holders) is extracted from the material of interest using a tool called a "dimpler". The disc is mounted and polished to a thickness of~100 µm (perpendicular to the disc radius) using a combination of rotary polisher and grinding wheel and turntable ( Figure 12). This is followed by either twin-jet electropolishing [128] or Ar ion-milling [127] to produce damage-free, vanishingly thin and electron transparent regions. Twin-jet electropolishing (an extension of electropolishing) is a fast, damage-free and cheap method for preparing TEM specimens from electrically conductive bulk materi-als such as metals and alloys. In this method the material is electrochemically removed from both sides of a mechanically thinned disc until sample perforation occurs [129]. Ion-milling involves bombarding the sample with relatively low energy (e.g., <1-2 keV) ions or neutral atoms to remove material from the surface until a very thin electron transparent specimen is formed. Variables including accelerating voltage, ion energy and incident angle of irradiation can be changed depending on the specimen requirements and polishing rate. Ar is widely used in ion milling due to its inert and heavy nature, though other ions are utilized. While ion milling is a common and useful technique, it is good to be aware of its artifacts and possibilities of redepositing material from one side to the other side of your specimen.

Ultramicrotome
The ultramicrotome technique uses a sharp diamond or glass knife to produce a nanoscale cross-section through manipulating six degrees of freedom (X-axis, Y-axis, rotational angle and clearance angle of the knife holder, the pitch and rotation of the block sample). Thereby, the knife-block relative position can be accurately regulated and adjusted according to the observation from the stereomicroscope attached at the ultramicrotome [130]. To obtain desired a TEM specimen, the sample requires to be sheared off by a clean single-side razor blade to form a trimmed surface (generally <1 mm 2 ). Then, the well sheared sample goes through trimming and sectioning at high and/or low speeds, to acquire ultra-thin slices (40-100 nm) which are easily transferred to TEM grids by a loop [131]. Figure 13 presents a schematic of ultramicrotome and images showing the sample cross sectioning. Due to the need to mechanically slice the specimen, it is not as practical in hard materials, where slicing can potentially lead to breaking and failure.

Nanomilling
Nanomilling is an effective tool for final thinning and cleaning of TEM specimens following other preparations by low voltage Ga + FIB milling or mechanical polishing. Nanomilling uses a focused Ar beam to polish both sides of the TEM specimen [54,133]. The spot size of an Ar beam is as small as 1 µm-smaller than the size of TEM lamella-so there will be no milling or redeposition of the specimen grid material (e.g., copper or molybdenum) if correct milling procedures are used [134]. Nanomill operates at sub-kV acceleration voltage and low current (<150 pA) under liquid N 2 environment to reduce ion beam damage and achieving effective thinning and cleaning [135]. Therefore, as a post-FIB processing step, nanomilling can not only remove a surface amorphous layer and implanted Ga + layer effectively but avoids additional damage and contamination [136]. The final TEM specimen will be clean, thin (<100 nm), and flat [137,138]. Figure 14 shows a clear comparison between two TEM specimens with and without the nanomilling as post-FIB processing step. The specimen that went through 500 V nanomilling exhibited clear lattice fringes even at the edge of the sample, while the specimen that went through 5 kV FIB polishing showed obvious redeposition and damaged layers which impeded imaging and energy spectrum analysis [134].

Specimen Contamination in S/TEM
Residual hydrocarbons in the microscope column coming from pump oil, specimen air lock or even user contact with holder can be a resource of contamination [92,139]. Carbonaceous materials then deposit on the surface of perfectly thinned specimen and significantly affect high resolution imaging and microanalysis. This can be minimized by heating the specimen to >100 • C in a heating holder to burn off the carbonaceous contamination or cooling the specimen to liquid-N 2 temperature in a cooling holder. Plasma cleaning of specimen and holder prior to operation is a very successful approach in reducing contaminations [140].
2.6. Analytical Techniques in TEM/STEM 2.6.1. Energy Dispersive X-ray Spectroscopy (EDXS) EDXS is a powerful qualitative and quantitative analytical technique available in S/TEM used for chemical analysis of a desired specimen. The technique relies on the X-rays generated in the microscope due to the interaction between the electron beam and sample. Two different types of X-ray signals are generated: Bremsstrahlung X-rays and Characteristic X-rays. Bremsstrahlung or braking radiation results from the deceleration of primary beam electrons when deflected by the atomic nuclei in the sample. The characteristic X-rays are a result of sample atoms being ionized by the primary electron beam. Simply, a core-shell electron is excited and ejected from the atom, while an outer-shell electron replaces it; this energy difference is released as an X-ray. These characteristics X-rays are like fingerprints for each element and as a result, EDXS spectra are very useful to determine elemental composition of any sample [102]. Figure 15 exhibits a summarized illustration of X-ray signal generation, collection, data display and resulting quantification [102]. The methods for quantitative analysis of EDXS data is discussed below. EDXS is a standard analytical technique in S/TEM for determining elemental concentrations in each sample. However, the quantification of low Z elements, including oxygen and lighter elements, has been widely known to be challenging in EDSX. This has to do with poor detection efficiency of these elements, severe absorption of X-rays within the specimen etc. Thus, even with very thin TEM samples, absorption and fluorescence, correction cannot be ignored for lighter elements. However, with detector advancements in recent years, particularly use of multiple silicon-drift detectors (SDDs), improved efficiency in signal collection, etc., it has become possible to quantify elements like oxygen with good accuracy [6,141].

Electron Energy-Loss Spectroscopy (EELS)
EELS is the analysis of the energy distribution of electrons that have lost energy through inelastic scattering while passing through an electron transparent sample [98,[142][143][144]. There are two general types of EELS in S/TEM, both of which rely on the facts that inelastically scattered electrons have kinetic energy corresponding to the energy lost, and that they can be dispersed in space by passing through a magnetic field, i.e., an energy-loss spectrometer. The first is energy-filtered imaging, known as energy-filtered TEM (EFTEM) [145], in which electrons having lost a relatively narrow range of energy (e.g.,~5 eV) are filtered out of the transmitted beam and used to form a real space image. The specific amounts of energy form a real space image. The second is parallel EELS, where all electrons having lost a relatively large range of energy (e.g., <2000 eV) are dispersed onto a detector plane and typically analyzed further in terms of a one-dimensional spectrum of counts versus energy loss. Both provide spatially resolved information about chemical composition and electronic structure, though parallel EELS is currently more commonly used because an entire EEL spectrum can be recorded at specified locations in a S/TEM specimen, potentially providing a rich set of data containing information about specimen thickness, chemical composition, and electronic structure of multiple elements simultaneously.
EELS is now incorporated into S/TEMs that typically operate using high energy electron (60-300 kV) sources. The interactions between electrons and matter are either elastic (approximately zero energy loss) or inelastic (finite energy loss), providing various information about the sample. Elastic scattering involves Coulomb interactions with the atomic nuclei, while inelastic scattering refers to the interactions between a fast incident electron and the atomic electrons that surround the nucleus. Inelastic processes can be understood in terms energy band theory. Figure 16 shows the energy level diagram in a solid and several signals generated after electron interaction and excitation in the specimen [142].
The energy loss spectrum consists of three main components: (i) the zero-loss or "elastic" peak (ZLP), (ii) the low energy-loss region (∆E <50 eV) where optical information can be determined by examining excitations into low-lying states above the Fermi energy [98], and (iii) the high energy-loss or core-loss region, where compositional and electronic information can be obtained from the electron beam interaction with deep core states (∆E >50 eV). These regions are indicated in Figure 17 for YBa 2 Cu 3 O 7 .
Electron energy-loss spectrum of YBa 2 Cu 3 O 7 is a high-temperature superconductor. The electron intensity is depicted on a logarithmic scale, showing zero-loss, plasmon peaks and ionization edges arising from each element [142,146].  The ZLP represents electrons that are transmitted without losing measurable energy, including electrons scattered elastically and those that excite phonon modes, for which the energy loss is less than the experimental energy resolution (in conventional EELS systems). While the ZLP does not contain spectroscopic information about the specimen, it is useful for energy calibration of the loss spectrum, deconvolution of plural scattering in thicker specimens and determining local thickness [145,147]. The width of the ZLP is related to the energy spread of the incident beam emitted from the electron source and is typically 0.25-1 eV [148,149]. These spreads are sufficient for applications such as core-loss elemental mapping, bulk plasmon analysis and analyses of wide-band gap materials but can mask multiple narrow excitations such as surface plasmons, near-edge fine structures, or vibrational excitations at very low losses. The tails of the ZLP ultimately define the spectral resolution and can obscure the optical excitations [150][151][152][153][154][155]. Spectral resolution can be improved to less than 10 meV in advanced monochromated electron microscopes that include specialized equipment to reduce the electron energy spread of incident beam [148]. In a typical monochromator, the electron beam is dispersed by energy (like a pre-specimen EEL spectrometer) and a narrow band of the dispersed beam is selected by an energyselecting slit which defines the energy spread of the electron beam, which is incident on the specimen. This technique has proven to be useful in high spatial resolution vibrational and low-loss spectroscopy in the electron microscopy, with monochromated STEMs having now provided impressive results for several years [92,149,[154][155][156][157][158][159].
The lower loss energies (<50 eV) are related to the formation and destruction of phonons and electrons of the outermost orbitals, including both single and collective collections, which are plasmons and electrons transmitting from valence band to empty conduction band, allowing band gap energy measurements ( Figure 18) [160,161]. The most common application of energy losses in this low range is the measurement of sample thickness, simply by comparing zero loss intensity with inelastic scattering intensity.  The high-loss electrons are related to the core binding energy, and therefore provide both qualitative and quantitative chemical information, as the intensity of each edge corresponds to the quantity of electrons generating a specific energy. In addition to the core-loss edges, the first 30-40 eV beyond each edge is related to the change in density of unoccupied states, which affects the electron loss near edge structure (ELNES), or fine structure [162][163][164].
The detection efficiency of EELS is generally much higher (up to 80% at low count rates) than EDXS systems (<1%) [78]. This is because a large proportion of inelastically scattered electrons are forward scattered into the EEL spectrometer, whereas X-ray emission is isotropic and only a small fraction is intercepted by finite-sized detectors; in recent years, additional (e.g., 4) X-ray detectors have been introduced to the S/TEM column to enhance efficiency. Remember that the detector efficiency will be affected by the specimen thickness, the probe current, and the solid angle of collection of the detector. In EELS, a thin specimen is preferred as more electrons transmit the specimen and can pass through the spectrometer entrance aperture, which can be expanded in diameter to increase collection efficiency.
The first step for analyzing EELS data is to subtract the background and any potential plural scattering effects that occur in a thick specimen when a significant fraction of incident electrons passing through a sample scatter inelastically more than once [158,165]. When the background is removed, intensities of elements can be converted to concentrations.
STEM-EELS analysis on an interface in a Y-doped ZrO 2 bicrystal is shown in Figure 19 [166]. In this case, STEM was used to observe the interface ( Figure 19a) and EELS was used to generate signal maps of each element relative to the bulk (Figure 19b-d). The line profile of elemental signals is shown in Figure 19e that shows depletion of 0.5-1 nm for O and Zr and enrichment of Y at this synthetic interface. Therefore, the combination of STEM with EELS is a powerful method, allowing the investigation of defect chemistry at atomic level, paving the way for understanding the relationship between GB and ionic conductivity [166]. In an electron microscope, the energy of electrons is transferred to specimen, bringing it to an exited state. As the excited electrons relax to the ground state, they release both high and low energy photons. The higher energy core-shell transitions generate X-ray emission, while the lower energy photons that are associated with valence band energy states and transitions between free electrons, generate cathodoluminescence (CL) which has applications in S/TEM (CL-S/TEM), particularly in the semiconductors, where an incident electron beam generates electron-hole pairs during the imaging mode. It is possible to separate electron hole pairs using an external voltage to generate a signal of electron charge pulse onto the STEM screen. The strength or weakness of the signal is related to the status of electron-hole pairs. When electron-hole pairs are separated, the signal is strong and when they recombine in presence of defects such as dislocations, the signal is weak. If the electron-hole recombines, they give off visible light in form of a spectrum which contains information about band gaps and dopants of the system. According to a recent study [167], it is impossible to interpret the emitted light spectrum to the transitions excited by the electrons directly. This is due to the existence of Cherenkov effect [168] in addition to ratio-induced transitions in the spectra. Therefore, the spectra need to be corrected for each sample shape, geometry, thickness, and beam energy.  23 , and Zr-L 23 EELS spectra from the bulk (black) and the dislocation core (red). Figure 19 reprinted from [166] under the terms of the Creative Commons CC BY license.

Electron Beam Induced Current (EBIC-S/TEM)
In EBIC, high energy electrons induce a current that is scanned across the specimen surface and variations in the EBIC induced current are used to map the electronic activity of the specimen. In this technique, the specimen needs to be coated to create ohmic contacts for applying an external voltage. EBIC has been used to identify defects such as GBs [169] or minority carrier properties [170] in semiconductors. It has also proven useful in characterizing oxides such as GBs and structure of zinc oxide varistors [171][172][173].  [165]. This technique has become a major crystallographic investigation tool to identify crystal structure and relevant parameters down to the nanometer fine scale. It is also used to map the local orientation to investigate crystal texture, rotation and strain and virtual STEM imaging at the nanometer scale, in connection with the scanning system of the microscope [166]. In this system, the incident electron beam is tilted and precessed using the pre-specimen optics to form a coni-cal electron probe. The result of this process is a higher number of spots in the diffraction patterns acquired and integrated over the precession cycle ( Figure 20). Compared to the case for an unprecessed beam, these PEND reflections have intensities that are determined by integrating through the Bragg condition of each reflection. Therefore, PEND enables better indexing of high order diffraction patterns, allowing complementary studies to X-ray diffraction, especially in multi-phase samples, disordered crystals or other systems that are not suitable for conventional X-ray crystallographic analysis [174]. PEND is now a powerful tool for looking at the twinning behavior in hexagonal alloys [149] and phase changes in oxide structures like Ni-YSZ electrodes during reduction and oxidation [146]. The sensitivity of this technique to local orientations has been the key to strain analysis in semiconductors [157].

D-STEM
In this method, the probe is scanned over the specimen in a two-dimensional (2D) array and at each probe position a 2D diffraction pattern is recorded, thus generating 4D data that can be further analyzed. This technique enables orientation, strain, and phase mapping down to the atomic scale, as well as virtual STEM imaging. Due to the volume of obtained data during 4S-STEM, data analysis can be complicated via requiring capable algorithms and simulation, though these are expected to become more automated over the next few years. In addition to providing spatially resolved diffraction patterns, 4D-STEM data can also be used for STEM imaging using "virtual" detectors. This is accomplished in post processing by integrating the scattered intensity within arbitrary segments of the diffraction pattern at each probe position, thus recreating a STEM image [175]. Figure 21 [175] shows imaging of Y-doped ZrO 2 using 4D-STEM. Two methods of image interpretation are shown: (i) generation of a virtual detector and (ii) selecting diffraction patterns from various regions of constant contrast over the probe position in the real space. As all the diffraction patterns for each position in the raster scan are stored, a comprehensive analysis of the GB crystallography in addition to local strain mapping can be performed (Figure 21a). Additionally, virtual STEM images can be created, such as BF, ADF and HAADF can be reconstructed from a single dataset showing in a by selecting angular ranges of interest (Figure 21b). For this purpose, computer image processing methods are used to classify the crystal orientations automatically, which is called "automated crystal orientation mapping" (ACOM). This method allows tracking orientation changes of hundreds of grains simultaneously and has been used significantly for in-situ mechanical testing measurements in nanocrystalline metals [177][178][179]. In a work by Garner, et al. [180], the orientation map of monoclinic zirconia reveals a fine level of detail with grains as small as 5 nm in the outer region. This was also done for tetragonal zirconia grains and therefore can be used for mapping any type of phase distribution [180].

In-Situ S/TEM
In-situ S/TEM is a technique where an electron microscope is used to watch and capture specimen's response to one or more external stimuli in real time. Operando S/TEM is an extension of in-situ analysis whereby specimen properties are measured in addition to simultaneous direct observation, thus shedding light on the specimen functionality.
Both techniques have emerged as a powerful field of nanomaterials characterization. Now, previously mysterious reactions and processes can be dynamically monitored under realistic or close-to-realistic conditions. The past decades have witnessed continuous development in both in-situ technology and the applications in a wide field of research, particularly in catalysis, energy storage and conversion [181,182], gas-solid interactions, semiconductor growth, nanomechanical testing, electrical properties and phenomena, and corrosion [183] (Figure 22).

Electron Beam Damage
The bombardment of thin TEM specimens by intense electron beams can result in irreversible atomic displacements and chemical reactions generally termed "electron beam damage". This can particularly affect any quantitative analysis performed on samples exposed to such intense, energetic electron beams [185,186]. The beam damage in S/TEMs can be a result of inelastic scattering caused due to the interaction between the energetic incident electrons and sample atoms/electrons. This beam damage can thus result in various phenomena in TEM samples including mass loss, phase decomposition, crystallization, amorphization, atom transport, etc. [185].
Two main mechanisms of radiation damage in TEMs are knock-on damage (elastic beam-atom collision) and radiolysis (ionization damage), or a combination of both [185,186]. In knock-on beam damage, the sample-beam interactions can cause preferential surface sputtering in oxides, while radiolysis can generate point defects and break chemical bonds, etc. [185] While radiolysis mainly occurs in insulating oxides, knock-on damage can occur in both conducting and insulating oxides. Furthermore, knock-on damage is significant during prolonged exposure to e-beam and can become predominant at higher accelerating voltages ≥200 kV. Other mechanisms include thermal effects for phonon excitation and electric-fielddriven migration [185]. Jiang [185] reviewed that while majority of the observed TEM/STEM beam damage in reported studies have been interpreted as knock-on damage or radiolysis, most observations appeared to involve collective atomic displacements in which atoms of the same element move together in synchronization.
The electron beam damage phenomena have been studied extensively in different oxides, both for bulk and GB characterization, such as spinel [187][188][189], alumina [190,191], perovskites [192][193][194], etc. In oxides, electron beam damage phenomena are often noticeable when using aberration-correction STEMs, which by design provide very small and intense. With EDXS particularly, longer acquisition times are preferred for better counting statistics, but this increases the likelihood of e-beam damage, as well as contamination [195]. This makes it difficult to interpret any chemical information obtained from an area damaged by e-beam. Electron beam damage was shown to influence the oxidation state of cations such as Ce 4+ and reduce it to Ce 3+ , particularly near GBs, providing a methodology for further investigating oxygen-vacancy concentrations in solid oxide solutions [196]. Figure 23 shows the Ce 3+ fraction across two GBs in Gd-and Fe-doped ceria. In both plots, the bulk contains larger than expected amounts of Ce 3+ due to the e-beam damage (30% and 20%, respectively). The Ce 3+ drops to zero or negative values, with a slight rise at the GB plane followed by another drop near the GB on the other side. After that, Ce 3+ increases again to high values in the bulk. It is noteworthy that the slight increase of Ce 3+ at the GB can be due to the beam damage, intrinsic boundary properties or both.

Imaging Atomic and Nano Structures
Atomic and nanoscale imaging by HRTEM is a common approach to assess GBs in many oxide systems, such as alumina [197][198][199][200], zirconia [74,201,202], spinel [203,204], ceria [76,205], etc. [206][207][208] In addition, aberration-corrected TEM is widely used in oxides [27,[209][210][211]. HRTEM micrographs and schematic models of GBs in a rare-earth (RE) doped alumina and a pure alumina are shown in Figure 24 [195]. Lattice fringes joining at the interface are observed in the RE-doped GB. This can indicate relaxation in the crystallographic structure near this GB and is consistent with observations of interface broadening in RE-doped GB, which itself is consistent with observations of interface broadening in RE-doped GBs and bicrystals compared to pure ones. It is concluded from the micrographs that the segregants are confined in a~1 nm region at the GB and that the GB structure is sensitive to GB type/character.  [195]. Note the wider GB region in RE doped alumina shown in (b,d). Spherical aberration (Cs)-corrected scanning transmission electron microscopy (STEM), with a sub-angstrom-sized electron probe can is another powerful technique for understanding structures of interfaces and segregation of dopants [212][213][214][215]. In a study by Schustertisch, et al. [215], HR-STEM, complementary EELS analysis and density functional theory (DFT) were used to accurately determine GBs structures in TiO 2 at atomic scale ( Figure 25). The existence of a unique nanoscale phase like bulk anatase TiO 2 was discovered in a fabricated Σ13 (221) 110 rutile TiO 2 bicrystals. This work show that it is possible to embed regions of TiO 2 anatase in rutile TiO 2 structure by GB fabrication and design. In other words, the power of GB analysis and engineering in oxides allows for further tuning the GB structure and properties in such materials. Most ceramics are composed of a mixture heavy and light element, the latter of which are hard to detect. Annular bright-field (ABF)-STEM combined with HAADF imaging is a powerful tool for imaging both light and heavy elements. Such studies shed light on the origin of structure property relationships in oxides and improve even more through the incorporation of theoretical calculations [122,217]. This technique was proved to be useful in imaging cation and oxygen sublattices in a (210) Σ5 CeO 2 GB with a tilt axis of [001] ( Figure 26) [33]. In addition, the oxygen non-stoichiometry used to reveal the role of the oxygen vacancies on the GB atomic structure. Modeling of the GB structure in this case was done using a static lattice calculation with the GULP program [218]. Figure 26a,c shows HAADF and ABF images taken from this GB, along with simulated HAADF and ABF images of the following in Figure 26b,d.
The bright spots in the HAADF image correspond to Ce column locations and the black and gray spots in ABF correspond to Ce and O column locations, respectively. The presence of O columns in the GB shown by the gray contrast is weaker compared to the O columns in the bulk. This is reasonable due to lower density of O column in GB compared to the bulk region.

Measuring GB Character
To determine the GB character in oxides, grain orientation mapping with a spatially resolved transmission electron diffraction techniques are used. In cases where the character of many GBs is of interest, researchers measure the GB misorientation angle distribution, which indicates the fractional length of boundaries with a given misorientation angle between adjacent grains. Such technique can employ PEND, 4D-STEM, or atomic resolution imaging to quantify GB character and orientation by automated acquisition and indexing of the diffraction data sets [7,147,219]. Figure 27 shows a of grain orientation overlay map on the BF-TEM image in polycrystalline Pr and Gd doped-ceria. Each grain is colored based on the crystallographic direction oriented parallel to the electron beam direction (see the color-coded stereographic inset). The misorientation angle distribution measured agrees well with polycrystalline cubic structures full of randomly oriented grains [220].

Chemical Analysis by EDXS and EELS
Analytical electron microscopy (AEM) techniques such as EDXS and EELS are powerful methods for qualitative and quantitative chemical and electronic structural analysis of the interfaces in oxides down to atomic scale. Modern AEMs readily provide spatially resolved spectroscopic data from points, lines, and 2D areas (also called "maps"). These methods have been applied readily for several decades, as described in books and reviews [99,[103][104][105]142]. STEM-EDXS is used to characterize GB and HI segregations in YSZ-Al 2 O 3 -MgAl 2 O 4 ceramic oxides ( Figure 28) and correlate them with thermal properties of these composites [41]. In a work by Ikuhara et al. [211], STEM-EDXS was used to study segregation behavior in YSZ bicrystal at atomic scale. Figure 29 represents EDXS maps and line profiles of Zr and Y, respectively. From this data, chemically ordered segregation around the GB core is detected. Y atoms are strongly segregated to the immediate left and right planes of the cation mirror plane (M c in Figure 29c,d), whereas Y is depleted in the mirror plane and Zr tends to segregate on the mirror plane (Figure 29b). Such detailed observations in GBs of oxides can lead to a better understanding about defect chemistry and related properties. In this case, the interactions between Y and oxygen vacancies govern segregation behavior by inducing a phase transition near the GB. Such characterization method capable of atomically mapping elements enables GB design at atomic scale, paving new avenues towards controlling and optimizing properties of oxides, particularly when supported by theoretical calculations such as Monte Carlo (MC), DFT, or phase field simulations [18,22,221].
Chemical doping is fundamental method of control and design GB functionality, and dopants are often added to ceramics improve sintering and control grain growth. Dopants or unwanted impurities in the system can segregate and modify GB properties. For instance, intergranular glassy phases are often observed in sintered ceramics [74,200]. It was elucidated in a work by Ishihara et al. [222], that small differences in the amounts of Si impurity led to distinct GB atomic structures in a Ti-doped Σ13 α-Al 2 O 3 GB (Figure 30). In GB type-I, Ti 3+ is replacing Al 3+ while in type-II, Ti 3+ oxidizes to Ti 4+ and the Al vacancies were introduced to the GB to compensate for that leading to three-times larger structure units (oxidation states confirmed by EELS). EELS is regularly employed in both TEM and STEM, and in addition to elemental concentrations can provide information about the electronic structure, bonding environment, chemistry, and oxidation states of atoms and ions at interfaces. While application of EELS can be very similar to EDXS, it acquires additional information mentioned, has higher detection efficiency, making it more promising for analysis and quantifications in e-beam sensitive materials and light weight elements. In a study by Saito et al. [223], segregation behavior of impurities to an MgO bicrystal was investigated. Some impurities tend to segregate to GB and interact with native structural defects, dominating the structures of the GB and hence determining many of its properties. A case study of the Cliff-Lorimer method for quantitative GB analysis in oxides is a work by Bowman et al. [8] that analyzes the cation concentrations at the GBs of a polycrystalline, oxygen conducting material, Pr 0.04 Gd 0.11 Ce 0.85 O 2−δ . Using the quantitative EELS analysis along with other characterization techniques, the authors concluded that the boundaries with higher solute concentrations have lower activation energy of ion conduction across GBs.  and (c,d). Intensity profiles by summing the X-ray counts in the maps in the direction parallel to the GB for (c) Zr-K and b. Y-K (normalized to the total Zr and Y counts). Reprinted from [212] under the terms of the Creative Commons CC BY license.
Yang et al. [224] published a study on GB segregation in Mn-doped SrTiO 3 . This material has promising magnetic and electric properties, which can be improved by engineering the GBs. It was found that Mn 2+ segregates inside GBs both in Sr and interstitial sites, while Mn 4+ is found to substitute Ti in bulk SrTiO 3 . Figure 31 represents the HAADF image, EELS elemental maps of (Figure 31a Figure 31e is extracted by averaging the compositions across the images shows an enrichment in Mn and Sr is accompanied by Sr deficiency at GB core. It is evident that Sr is roughly 50 at% away from the GB while the sum of Ti and Mn is also roughly 50 at%. This would suggest that Mn substitutes for Ti and should, therefore, have a +4 charge state [224]. This interface was also investigated by EDXS (Figure 31f,g). It is evident that there is significant difference between qualitative elemental maps (Figure 31b,d) and quantitative maps (Figure 31e,g) which proves the importance of quantitative analysis.

In Situ S/TEM of GBs and HIs
In-situ mechanical testing inside electron microscopes has been an effective approach to investigate interfacial deformation and fracture modes in ceramics. Among different techniques, in situ nanopillar compression is widely used to explore the mechanical behavior and deformation mechanisms, including dislocation motion, stacking fault formation and GB sliding, in ceramic materials at small scale [225,226]. Dynamic interactions between individual dislocations and well-controlled GBs in SrTiO 3 can be directly observed under TEM during nanoindentation [227]. The dislocation impediment effect induced by high-angle or low-angle GBs was found to be related to both the geometric and structural stabilization effects of the GBs [227]. In another example, Kondo, et al. [228] conducted nanoindentation at a Zr-segregated ∑13 GB in Al 2 O 3 and observed the crack nucleation and propagation along the GB inside TEM. Post-mortem HAADF STEM revealed that atomic-scale cleavage path is along the center layer of the triple Zr-rich layers, pointing to the importance of local oxygen coordination at the dopant-enriched GB core in determining the fracture path and therefore, toughness, of oxide ceramics. Combining localized laser heating inside TEM, in-situ mechanical testing at ultra-high temperatures up to the melting point of ceramics could be realized. Using this approach, Grosso et al. [229], investigated the temperature dependency of yield strength in nanocrystalline Sc 2 O 3 -stablized ZrO 2 from room temperature to 2050 • C. Their results cast doubt on the previous hypothesis that GB diffusion is the reason for Hall-Petch break down in nanocrystalline ceramics, at least in this material system. This approach has also been used to investigate GB and surface diffusivities in ZrO 2 bicrystal by the same group of authors [230]. Figure 31. Atomic resolution mapping of the Mn-doped SrTiO 3 GB. EELS compositional maps of (a) Sr, (b) Ti, (c) Mn, and (d) composite maps were obtained from the MSA with RGB false colors applied. The scale bar is 0.5 nm. A few nm away from the GB core, Mn has a much lower but detectable concentration, and substitutes the Ti site as indicated by the coincident dashed circles in (b,c). The relative concentration profile (at%) across the GB in (e) is obtained by concerting the vertically summed up EELS signal of the two-dimensional map using the Cliff-Lorimer factors of a 200 kV microscope. Atomic EDXS mapping of the Mn-doped SrTiO 3 GB. (f) shows count maps of Sr, Ti and Mn, respectively. (g) shows the atomic fraction of elements normalized to1 (or 100%) using the Cliff-Lorimer factors. As seen clearly in these maps, Ti shows reduced counts but increased concentration (or atomic fraction) as the concentration of Sr drops inside the GB. Reprinted from [224] under the terms of the Creative Commons CC BY license.
In addition to mechanical testing, in-situ characterization of other crystalline defects, such as oxygen vacancies under thermal or electrical stimulus is of great importance for a wide range of functional properties in ceramics. Klie, et al. [3] performed in-situ heating coupled with STEM-EELS on a series of perovskite oxides to capture changes in the structure, composition, and valence states. An excess of oxygen vacancies was observed at the GB before heating. At 724 K, the amount of oxygen vacancies increased both in the bulk and at the GB, even though the GB cation structure did not change upon heating. According to the EELS data collected from the same GB upon cooling, oxygen diffusion occurs from the bulk into the GB plane which agrees with the increase of oxygen vacancies upon heating. In other words, the oxygen diffuses to the bulk upon heating and goes back to the GB upon cooling. The impact of electrical biasing on oxygen vacancy migration in CeO 2 was studied by Gao, et al. [231] using STEM-EELS. The formation and ordering of oxygen vacancies in the CeO 2 thin film within an Au/CeO 2 /Nb-SrTiO 3 junction was directly observed by applying a positive voltage. These promising results have proven that the oxygen vacancy migration mechanism is responsible for the resistance switching in CeO 2 and possibly other oxides with similar properties [231]. Such material systems can be used for next generation memories due to their small power consumption, high density, and other properties. It is important to understand and control the switching behavior of nanoscale oxide devices.
In-situ electron microscopy combined with electron loss near edge structure (ELNES) can be used to monitor changes in the electronic structure of many elements such as 3dtransition and rare earth metal oxides, including Fe and Ce oxide [232,233]. Numerous experiments have suggested a strong correlation between changes in valence states of cations and the L 3 /L 2 or the ionization edges (white lines) ratio and O-K edge fine structure. In addition, the O-K pre-peak, which is caused due to transitions from the O1s state to the unoccupied hybridized O 2p in transition metal 3d orbitals, helps quantify the transition metal valence and the concentration of mobile holes and the transition metal spin states [234][235][236][237][238][239][240].
In 2000, Wang, et al. [241], reported the L 3 /L 2 ratio for direct measurement of Co and Mn valence states. Significant differences between O-K edges in Co 3 O 4 and CoO allowed to distinguish Co oxidation state in these two structures [232,233].
In an in-situ heating experiment coupled with STEM-EELS, metallic cobalt was distinguished from cobalt in 2+ and 3+ oxidation states [237]. The ELNES used for this study are shown in Figure 32. A clear energy shift in the Co-L 3 and L 2 peaks of CoO (contains Co 2+ ) compared to Co 3 O 4 (contains Co 2+ and Co 3+ ) is observed which is related to transition of 2p 3/2 to 2p 1/2 core-shell electrons into unoccupied 3d orbitals. In summary, the following techniques can be used to measure the oxidation state of each cation at the GBs and HIs, where defect chemistry at such interfaces plays an important role and it directly relevant to the concentration of such defects at the interface.
In-situ EM has proved to be beneficial in research of battery materials, particularly observing material degradation during electrochemical performance and cycling [242]. While lithium-ion batteries with Ni-rich layered oxides such as LiNi 0.8 Co 0.15 Al 0.05 O 2 (NCA) are of interest as cathodes, they suffer from capacity fade hypothesized to be correlated to microcracks that initiate and grow through GBs [243,244]. These microcracks serve as pathways for electrolyte penetration into secondary NCA particle and grow with the electrolyte corrosion and mechanical strain during lithiation/delithiation [245]. These hypotheses are validated by the comparison ELNES in the bulk and GB. According to Figure 33b, Ni shows an increase in the L 3 /L 2 ratio indicates the change of Ni oxidation state from the bulk to the GB, while the same thing did not occur for cobalt (Figure 33e). In addition, GB were observed to be preferential cites for in situ exsolution of metallic particles [246]. This phenomenon is related to the lower formation energy for vacancy and nucleation sites at GBs, which is ultimately important for controlling particle density and distribution.  (i) Normalized Ni concentration (relative to Co) from the GB to the bulk derived from EELS spectra [245]. Figure 33 reprinted with permission from the Royal Society of Chemistry Copyright 2021.

Electrical Properties
In polycrystalline oxides, GBs and HIs play an important role in functional properties such as ionic conductivity [247,248]. To measure these properties, techniques such as electrochemical impedance spectroscopy (EIS) are used that can measure conductivity of ions and distinguish the contribution of each component of the system to the measured conductivities [249,250]. It is common for GBs to behave different than the bulk and affect the conductivity values, whether reduce or increase them. For example, in many oxide systems GBs tend to hinder oxygen vacancy movement thereby drastically reducing the conductivity values. To better understand the mechanisms responsible for these observations, a comprehensive analysis of charge carriers responsible for conductivity or in other words defect chemistry at the GB is required and has been studied to good extent [71]. Previously, an intrinsic space charge-layer was discovered that is localized at the GB which induces local redistribution of charge carriers in the vicinity of the GB core. This results in the depletion of charge carriers, reducing the conductivity at the GB [251]. In the following example, STEM-EELS was used to link nanoscale GB defect chemistry to the oxygen ion conductivity across the GBs Ca-doped ceria [21]. Figure 34a presents an ADF image of a GB where both grains are oriented near a zone axis orientation ( Figure 34) [21]. The contrast is sensitive to the Z or atomic number of the elements. The decrease in ADF intensity at the GB is an indication of a decrease in high angle scattering of the electrons. This agrees with the EELS results shown on the side, presenting an increase in Ca 2+ concentration at and near GB core compared to the bulk, as Ca 2+ is lighter than Ce 4+ therefore we expect lower signals where it segregates. The amount of Ca segregated at the GB was measured using EELS and is shown in Figure 34b. The segregation is maximized in the case of 5 mol% Ca added to the system, which is also followed by a 2 times higher conductivity value (Figure 34c). Another major application of S/TEM characterization technique is the field of all solidstate lithium batteries, where ceramic oxides are employed as solid electrolytes [252]. These batteries are of interest for electrochemical energy storage applications due to their high energy density, safety, long cycle life and wide range of operation temperature. The candidate oxides include cubic garnet-type Li 7 La 3 Zr 2 O 12 (LLZO), perovskite-type Li 3 xLa 2/3−x TiO 3 (LLTO) and NASICON NLiM 2 (PO 4 ) 3 (M = Ti, Ge, Zr, Hf) [253][254][255][256][257][258]. These structures are stable in air and with respect to Li metal and other high voltage cathodes. However, there are a few downsides including low ionic conductivity at various interfaces such as the anode/electrolyte/cathode, and GBs in the bulk. For example, the perovskite Li 0.34 La 0.51 TiO 2.94 has shown bulk ionic conductivity of 1 × 10 −3 S cm −1 and total ionic conductivity of higher than 2 × 10 −5 S cm −1 at room temperature, meaning GBs tend to decrease conductivity [257]. Therefore, researchers began investigating this phenomenon to design new interfaces with higher efficiencies. Doping has been effective in improving total conductivity values in some oxides [259]. In general, doping of elements such as Al 3+ /Ge 4+ at the Li sites and Ta 5+ /Nb 5+ at the Zr sites has stabilized the cubic garnet and reduced Li ion migration activation energy thereby increasing the Li ion conductivity [260,261]. The substitution of the Zr 4+ sites by Ge 4+ improved ionic conductivity to be 4.78 × 10 −4 S cm −1 at 20 • C with an activation energy of 0.29 eV which are among the best in the field [262,263]. This dopant has also improved the sintering process by reducing the sintering temperature and increasing the density.
Another example is the poor GB conductivity in the Li-ion conducting solid electrolyte LLTO. Figure 35 shows HAADF-STEM image of GBs with dark contrast denoted as GB type I (upper portion) and normal contrast denoted as type II (lower portion). The difference between atomic structure of these types is evident in Figure 35b,c. While the GB type I has significantly deviated from the bulk and perovskite atomic structure disappeared, the GB type II experienced a much smaller lattice mismatch compared to that in type I and the perovskite structure remained. It is discovered using TEM that most GBs have severe structural and chemical changes from normal perovskite to compensate for the random orientation of the adjacent grains including such as reduction of Ti 4+ to Ti 3+, depletion of Li, decrease in La content and deformation of Ti-O polyhedral ( Figure 36). In addition, the structure of type I GB was schematically depicted in Figure 37 based on the HAADF-STEM and EELS analysis. The intensities of each element were defined by the atomic column intensity in the bulk. As seen in EELS results, the GB was filled with O and Ti and possessed a non-perovskite crystal structure. It was concluded that the existing GB is not energetically preferred for either Li accommodation or transport, giving rise to the poor GB conductivity. Papers including one by Ma et al. [20] perfectly demonstrates the correlation between atomic-scale GB structure and ionic conductivity in Li superionic conductors [27,261,264].
The atomic structures and the band gaps of four typical GBs in α-Al 2 O 3 by stem and valence electron energy-loss spectroscopy. It was found that the band gaps of the GBs are narrowed by 0.5-2.1 eV compared with that of 8.8 eV in the bulk [161].  [20]. Reproduced from [20] with permission from the Royal Society of Chemistry, Copyright 2014.   [20]. Reproduced from [20] with permission from the Royal Society of Chemistry, Copyright 2014.

Thermal Properties
Another functional property that can be affected by the segregations at the GBs are thermal conductivity and thermoelectric characteristics. While most of the literature in this area is related to metals and non-oxide ceramics [265][266][267], there still exist a few studies that correlate thermal conductivity to segregation at the GB. In a work done by Boyle, et al. [268], it was proven that the addition of bismuth dopant to the calcium cobaltite structure, has improved Seebeck factor (a measure of the magnitude of an induced thermoelectric voltage in response to a temperature difference across the material [269]) and thermal conductivity, increasing the power factor in this material that can potentially be used as thermoelectric generators, coolers, etc. Segregation of bismuth at the GB plane is evident in Figure 38 captured using TEM and HAADF imaging [268].
Furthermore, it was shown previously that the combination of lattice dopants substitution and secondary phase segregation at the GB has dramatically increased the energy conversion efficiency in Bi-doped CaMnO 3-δ . In this case CuO segregates at the GB, leading to a significant increase of both the Seebeck coefficient and electrical conductivity ( Figure 39) [268].
Using S/TEM techniques, it is possible to learn about grain texture and orientations to correlate that to the properties, as done in a work by Shi et al. [271]. It has been demonstrated in the literature that GB has a huge impact on thermal conductivities of polycrystalline ceramics, including oxides [270,272,273]. GB engineering has shown to be a better way for improving thermal resistance of materials such as nanocrystalline yttria-stabilized zirconia and alumina ceramics and refractories [273]. While the impact of GBs is evident, a more comprehensive study of the nature of the structure and chemistry of the boundaries and their interaction with phonons is required to optimize such properties.

Mechanical Properties
As a general type of planar defect, GBs in ceramics play an important role in their deformation mechanisms, such as GB sliding and intergranular fracture. Changes in the structure and chemistry of GBs could strongly influence the mechanical properties of ceramics. Tremendous research efforts have been devoted to enhancing the mechanical properties via GB engineering approaches that include controlling the grain size [90,[274][275][276][277]; doping with dilute amount of rare earth elements [90,278,279]; forming glassy nanolayer films at GBs [200,280]; and optimizing the GB character distribution by introducing a high frequency of low-angle GBs [281]. Despite these important improvements, fundamental breakthroughs in the understanding of the correlations between GB structure and mechanical response is needed for the development of ceramic components with superior mechanical reliability.
Accurately determining the structure, chemical composition and bonding state at GBs is critical to unveil the structure-mechanical property relationships. GB non-stoichiometry often strongly influences the mechanical properties. As one of the most widely used structural ceramics, Al 2 O 3 has a weak resistance to creep at high temperatures via GB sliding, while its creep resistance could be effectively improved by the addition of rare earth elements. Ikuhara, et al. [278,279] observed a segregation layer rich in Ln cations at GBs using HRTEM, EDXS and EELS. Such non-stoichiometry at GBs contributes to the dopant-dependent creep resistance in Ln oxide-doped Al 2 O 3 , although questions remain regarding its atomistic mechanisms ( Figure 40). Later studies identified the atomicscale location of rare earth atoms at the expanded sites along GBs, which leads to an increase in the number of bonds and bond strength at GBs [200]. This finding suggests the enhancement in creep resistance could be affected by the ionic radii of rare earth dopants. Towards three-dimensional characterization of the GB segregation, a buried GB in Y-doped Al 2 O 3 was studied under off-axis illumination condition using HAADF STEM, which revealed an ordered two-dimensional structure of Y dopants on the GB with local disordering. In addition to dopant segregation, locations of oxygen vacancies on GBs were determined by HAADF and ADF STEM and EELS, underlining the importance of oxygen non-stoichiometry on the mechanical properties in oxides [7,282].
Furthermore, microstructural characterization of mechanically tested ceramics could shed light on their deformation mechanisms. Using HR-TEM, the atomic structure of Al 2 O 3 recovered from shock loading was characterized, and a transition in deformation mechanisms from intergranular fracture and dislocation activities near GBs at low impact pressure to deformation twinning at high impact pressure has been observed [283]. In B 4 C, which is an important super-hard lightweight ceramic, the deformation and failure mechanism has long been controversial. Acquiring experimental evidence on atomic structure of B 4 C is limited by its complex crystal structure with a C-B-C chain and B 11 C icosahedra and low Z-contrast of boron and carbon. Recent developments in aberration-corrected scanning transmission electron microscopy enables microstructural characterization with sub angstrom spatial resolution, particularly for light elements. Using ABF STEM imaging, detailed structure within the amorphous shear bands in nanoindentation tested B 4 C was investigated. Distorted icosahedra structures were found randomly distributed in the amorphous bands, offering insights into the deformation-induced amorphization mechanisms of B 4 C [284]. Guo et al. characterized the GB structure in nanocrystalline B 4 C after quasi-static bending tests of cantilever nanobeams. They reported 2 nm wide amorphous phase at GBs, which is likely induced by GB sliding and corresponds to the breakdown of Hall-Petch relationship in ceramics [181].

Magnetic Properties
GBs and HIs can affect magnetic behavior of materials such as oxides. For example, lower magnetoresistance at T amorphous bands, offering insights into the deformation-induced amor nisms of B4C [284]. Guo et al. characterized the GB structure in nanocry quasi-static bending tests of cantilever nanobeams. They reported 2 nm phase at GBs, which is likely induced by GB sliding and corresponds to t Hall-Petch relationship in ceramics [181].

Magnetic Properties
GBs and HIs can affect magnetic behavior of materials such as oxid lower magnetoresistance at T ⪡ TC in a polycrystalline La0.67Ba0.33MnO2.99 compared to it is single c lated to high volume of GBs present in the former [285]. In another exa netic properties of nanocrystalline Mn-doped ZnO thin films were impr istence of greater GB specific area compared to the critical values [286] show that the wurtzite lattice of ZnO is separated by amorphous layers increases with the Mn concentration which is controllable (Figure 41). Am at GBs and HIs can potentially affect other magnetic properties such sistance of a magnetic material to demagnetization). This is evident in a amin et al. [287], where B-enriched amorphous intergranular phases c wall pinning that is correlated to high coercivity observed in B-doped Lattice distortion and strain in HIs can affect magnetic flux pinning cente critical density in SrZrO3 [288]. Figure 41 represents the BF-TEM image of Mn-doped ZnO films w 15 at% Mn concentration. The direct resolution of the lattice allows the o crystalline ZnO grains in wurtzite structure, and it is clearly visible phases lie at the areas between the grains or in other words, GBs [289]. T C in a polycrystalline La 0.67 Ba 0.33 MnO 2.99 compared to it is single crystal form is related to high volume of GBs present in the former [285]. In another example, ferromagnetic properties of nanocrystalline Mn-doped ZnO thin films were improved due the existence of greater GB specific area compared to the critical values [286].BF-TEM studies show that the wurtzite lattice of ZnO is separated by amorphous layers whose thickness increases with the Mn concentration which is controllable ( Figure 41). Amorphous phases at GBs and HIs can potentially affect other magnetic properties such as coercivity (resistance of a magnetic material to demagnetization). This is evident in a work by Sepehri-amin et al. [287], where B-enriched amorphous intergranular phases cause the domain wall pinning that is correlated to high coercivity observed in B-doped Sm (Fe 0.8 Co 0.2 ) 12 . Lattice distortion and strain in HIs can affect magnetic flux pinning centers leading to high critical density in SrZrO 3 [288].  15 at% Mn concentration. The direct resolution of the lattice allows the observation of the crystalline ZnO grains in wurtzite structure, and it is clearly visible that amorphous phases lie at the areas between the grains or in other words, GBs [289].
The examples mentioned above and more in literature [290,291] prove the importance of GBs and HIs in magnetic behavior of oxides, and therefore highlight the importance of their characterization.

Optical Properties
Polycrystalline ceramics can be transparent, translucent, or opaque [40]. Transparent polycrystalline ceramics have become of interest since Neodymium-doped yttrium aluminum garnet (Nd:YAG) was introduced as a lasing medium for solid state lasers in 1995 [201] and have applications in armors, opto-electrical devices such as solar cells, displays, and circuitries. Compared to single crystals, polycrystalline ceramics are costeffective, easier to produce, exhibit good mechanical, thermal and chemical stability, and shape-control availability [292,293]. However, certain defects such as impurities and porosity inside their microstructure scatter and refract light leading to optical losses, thus limiting wide employment of transparent polycrystalline ceramics [203,294]. GBs are one main area accommodating such defects and can impact properties of these materials. The GB structure can also control grain size and phase composition, thus playing a significant role in transmittance of polycrystalline ceramics. Some boundaries with higher doping concentrations show a low reflectance, whereas other boundaries with secondary phases show a bold difference in refractive index with respect to the bulk, becoming the major scattering sources [295]. In addition to impurities, grain size related birefringence also affects the transparency of polycrystalline ceramics based on Rayleigh approximation [296,297]. Therefore, it is important to characterize such defects and interfaces to achieve optimal properties in polycrystals [12][13][14] For example, S/TEM techniques can help detect the origin of light scattering at interface to improve optical quality. Segregation of elements at the GBs in polycrystalline oxides affect optical properties such as transmittance [148]. In a work by Trunec, et al. [298] polycrystalline alumina was doped with Zr and spinel nanoparticles to improve in-line transmittance. According to STEM EDXS results, Zr segregated at the GBs shown in Figure 42. According to this study, the segregation has impeded GB motion to achieve grain size refinement in dense alumina structure, thereby lowering light loss by birefringence. In addition to intentional doping, small amounts of impurities in raw material can impact the microstructure and properties of transparent magnesium aluminate spinel (MgAl 2 O 4 ) [299]. Impurities such as carbon, sulfur and silicon tend to segregate and form amorphous phases at GBs and triple junctions (Figure 43a-c) and restrict grain growth by GB pinning and solute dragging mechanisms. Hence, sub-micrometer grains increase GB volume fraction and impurity segregation at GBs discontinues light transmittance by increase in light scattering, leading to opacity (Figure 43f). LiF reacts with impurities, forming volatile species. These products can be easily removed in high temperature to clean GB and facilitate grain growth. Clean GBs greatly improve transparency of the MgAl 2 O 4 that can be used as transparent armor and IR-seeking missile domes (Figure 43d-g).

Conclusions and Future Perspective
Recent and continual advancements in electron microscopy hardware, data acquisition and processing algorithms, three-dimensional analyses, and innovative workflows provide enormous potential to the field of interfacial characterization. An overarching goal of high-resolution analysis is to such fine-scale data to develop models capable of predicting macroscopic structure-property relationships. However, it is indeed challenging to collect a statistically significant amount of observations/data using S/TEM, simply because one needs to prepare many samples and analyze tens, hundreds or even thousands of interfaces, all of which might vary slightly in their atomic scale nature (e.g., atomic structure and chemistry). Therefore, there is significant potential in automation of such measurements using machine learning and artificial intelligence.
It is imperative to consider the limitations of S/TEM while using and planning to use it. Direct observations made with the help of powerful electron microscopes give us valuable insights to the underlying science of interfaces. But these techniques focus on a small volume of material, likely subject to multiple specimen preparation steps as mentioned earlier, and by the time the specimen is observed, the material may be somewhat different than the material in bulk form. Thus, the authors strongly suggest that (potential) users keep this in mind and be aware of possible impacts of preparation steps.
There is a clear lack in literature on heterointerface characterization in multiphase polycrystalline ceramics. The added complexity of such systems due to existence of several phases and defects makes such studies more challenging and resource intensive. While the majority ceramics community focuses on single phase oxides due to their promising properties, mixed phase counterparts have proven to be of importance in various fields such as structural materials, thermoelectric materials, and mixed ion electron conductors (where one or more phases dominate the electron conduction while the other/others maintain ion conduction). It is expected that advancements in high-throughput automated EM, including the ones mentioned in this paper, will enable more investigation of GBs/HIs in multiphase polycrystalline systems.
Lastly, in-situ and in-operando studies of interfaces in functional materials, and those sensitive to our atmosphere and electron beam irradiation are very promising, even though there are limited studies at present. This will likely change with recent and future technological and workflow advancements in EM, such as sample preparation and cryo-EM methods for sensitive materials, such as alkali ion conductors. Such dynamic experiments on interfaces (and nanomaterials more generally) are particularly exciting due to the possibility of observing reactions, degradation, and mechanisms underlying interface functionality.
In closing, while there are many experimental and practical considerations to make when undertaking such a study, the authors hope this article paves the way for more scientists and engineers interested in using S/TEM techniques to study and develop interfacial sciences, particularly in ceramic oxides.