Direct X-Ray Imaging as a Tool for Understanding Multiphysics Phenomena in Energy Storage

Storage of energy that is harvested from intermittent sources and consumed in dispersed applications is a critical requirement for developing a more sustainable energy infrastructure. To meet this requirement, a host of advanced battery technologies is emerging for use in mobile electronics, transportation, and grid-based storage. Among these technologies, the Li-ion battery (LIB) has become prevalent, particularly for mobile electronics. As LIB technology continues to improve the range of applications to which this technology can be applied has expanded to include applications as diverse as power tools and unmanned aerial vehicles. Notably, Li-ion batteries have emerged as a key component in reducing the environmental impacts of fossil-based transportation through increased electrification of the global automotive fleet. Sustainable automobile electrification requires development of batteries that achieve sufficient power density for hybrid electric vehicles (HEVs) and plug-in hybrid electric vehicles (PHEVs) and energy density for PHEV and fully electric vehicles while providing lifetimes that support affordable cost of vehicle ownership. As the above applications continue to advance, LIB technology must keep pace. There is a critical need to reduce cost, increase performance, and mitigate degradation limiting battery life. To this end, a diverse array of advanced LIBs is being explored with respect to materials, chemistry, and battery architecture.

Li-ion batteries are inherently multiscale, functional material systems in which the morphology and composition of the constituent materials influence the multiphysics response of the battery to internal and external electrochemical, mechanical, and thermal stimuli. Driven by the need to better understand this coupled multiscale and multiphysics behavior, the field of 3D imaging for LIBs and other energy materials has been expanding rapidly across multiple scales and imaging techniques [1]. For observation of mesoscale features, two methods have emerged in recent years: focused-ion beam/scanning electron microscope (FIB/SEM) serial sectioning [2–5] and X-ray tomography [6–14]. X-ray tomography provides an imaging technique that permits direct 3D observation of material structures at resolutions of microns (microtomography) down to 17 nm (nanotomography) [15]. In contrast to FIB/SEM, X-ray techniques preserve the imaged sample and permit in situ material characterization. When applying synchrotron-based measurements, the capability for high resolution spectroscopic imaging is also afforded by the high intensity, monochromatic X-ray radiation [6,7,16–18].

This paper provides a brief review of recent studies that apply X-ray imaging toward understanding the multiphysics behavior of Li-ion battery operation. The second section provides a basic overview of X-ray tomography technique. The third section addresses specific examples of how X-ray imaging techniques have been applied toward Li-ion batteries. Ex situ, in situ, and in operando studies are addressed first. Discussion of these applications is followed by a brief discussion of the integration of 3D tomographic data with multiphysics modeling. The sources herein are considered to provide a representative cross section of the burgeoning field of 3D X-ray imaging as applied to LIB materials and components. For a more detailed review of radiographic and tomographic imaging using both X-ray and neutron techniques, the reader is referred to the Strobl et al. [19]. A thorough review of 3D imaging methods related to the more general field of energy materials can be found in Ref. [1].

X-ray tomography is a 3D imaging technique that permits the mapping of microstructure in heterogeneous materials. This method has been applied to a broad range of disciplines as diverse as materials, geological, and biological sciences. As noted, X-ray tomography has emerged in recent years as a key technique in observing the structure of energy materials, including battery materials [1]. By providing direct 3D observation of mesoscale material structures at resolutions from nanometers to microns, this technique will continue to play a key role in understanding the multiscale, multiphysics behavior of energy storage devices. X-ray tomography is performed by using transmission X-ray microscope (TXM) to acquire a series of transmission images of a sample during rotation. The incident X-ray beam is focused onto a sample using a condenser or zone plate lens as the focusing optic. A portion of these X-rays are transmitted through the sample and focused onto a scintillator screen. The resulting image is captured using a camera, typically a charge-coupled device (CCD). Transmission images are taken at fixed angular increments while the sample is rotated in the X-ray beam. The series of transmission images is then reconstructed using appropriate algorithms [20,21] to yield a 3D image. The reconstructed data can be segmented to separate distinct materials within the imaged sample, and the resulting digital image data may be characterized using a variety of 3D microstructural characterization methods [22–27]. The digital image data may also be applied as a computational domain in numerical simulations.

Microtomography and nanotomography measurements may be achieved using either laboratory-based or synchrotron X-ray sources. While laboratory-based sources have typically provided microtomography capabilities, recent years have seen the emergence of laboratory-based X-ray nanotomography (XNT) systems capable of resolutions down to 50 nm [28–30]. Applying monochromatic synchrotron radiation can provide higher resolution, faster data acquisition, and the capability to discern solid materials based on elemental and chemical X-ray absorption characteristics using absorption contrast [16,17] and X-ray absorption near edge structure (XANES) imaging [6,7,16–18]. In absorption, contrast nanotomography materials with different elemental composition are mapped by performing X-ray nanotomography above and below an elemental X-ray absorption edge. Below the absorption edge, dense regions in the sample transmit X-rays comparably and permit the less dense phases to be clearly distinguished. Increased X-ray absorption by one material above the absorption edge allows further separation of elementally distinct phases.

Absorption contrast imaging is well-suited for materials with high atomic number, Z, such as active cathode materials and some high capacity anode materials. However, these methods often do not provide good contrast for low atomic number materials, such as polymer binders and carbon additives in LIB electrodes. Such materials with low X-ray attenuation coefficients do not cause variations in X-ray attenuation that are significant enough to produce the high contrast that enables differentiation between the constituents of an imaged sample. In these situations, a Zernike phase contrast technique can be used to enhance the image contrast and discern low Z materials [29,30]. This technique operates on the undiffracted portion of the X-ray signal by implementing a phase ring in between the sample and the detector. The phase ring shifts the phase of the undiffracted wave by −π/2. This shift increases the amplitude between the undiffracted wave and the sample wave. However, a portion of the diffracted wave tends to interact with the phase ring causing a shift in the diffracted wave as well. This shift in the diffracted wave causes bright artifacts to appear on the image where the magnitude of the sample wave is greater than that of the diffraction wave. This challenge has been addressed for laboratory-based systems by Kumar et al. through development of an algorithm that eliminates these artifacts based on the known effects of the phase ring on the diffracted wave [30]. This phase contrast tomography technique has been applied in concert with absorption contrast methods to distinguish the active material and regions containing combined binder and carbon additive in LiCoO₂ cathodes [29].

The elemental mapping capabilities of absorption contrast nanotomography can be expanded to chemical mapping by taking advantage of the unique X-ray absorption behavior of elementally similar but chemically distinct materials, e.g., metals and their oxides [18]. The X-ray absorption behavior of such materials near an elemental X-ray absorption edge provides additional data for distinguishing phases in the microstructure. In this approach, tomographic scans may be taken at a limited set of energy levels where the materials show distinct differences in X-ray absorption [18], or more detailed spectroscopic mapping may be achieved by performing tomographic scans across a range of X-ray energies [7,31]. Nelson et al. demonstrated the former XANES nanotomography approach for Ni and NiO in composite samples [18]. In such cases, the different oxidation states of Ni and NiO resulted in distinct responses to the incident X-ray energy. When variations in valence state in a sample are more subtle detailed XANES scans over a range of X-ray energies can be applied, incrementing the energy level at the single electron volt scale [7,31]. The absorption behavior of individual pixels (2D) or voxels (3D) may then be fit to a linear combination of standard XANES spectra to yield a detailed map of chemical composition.

An additional advantage of synchrotron-based measurements is the capability to perform in situ and in operando imaging studies, particularly for hard X-rays with energy above 5 keV. This advantage is enabled by two key aspects of synchrotron sources. First, the significantly more intense synchrotron X-ray radiation reduces required sample exposure time facilitating real time observation of chemical and structural changes in materials. Second, hard X-ray systems like many current synchrotron-based TXMs do not require vacuum conditions for sufficient X-ray transmission. Thus, active battery cells can be readily integrated into the beam line [7,31–35].

During the operation of Li-ion batteries intercalation related stresses, chemical changes in the active materials, and side reactions can lead to the degradation of the materials that comprise the battery. This degradation typically manifests as variations in the structure and chemical composition of the constituent materials. Structural changes, such as crack formation are readily observed using X-ray methods [6,12,14,33], and 3D data presents the unique capability to observe the impacts of such microstructural changes on transport networks within materials [17]. Ex situ X-ray tomography has been performed to observe the structure of carbon anodes [13] and to assess oxidation and degradation in lithium vanadium oxide (LiVO₂) anodes [14]. However, a majority of recent ex situ X-ray tomography studies have focused on Li-ion cathode materials. The transition metal content of many cathode materials also makes them amenable to synchrotron-based XNT measurements. Taking advantage of distinct elemental absorption behavior, Chen-Wiegart et al. [6] used XNT to observe and analyze the microstructure and composition of Li-ion battery cathodes. In follow-on studies, the structure of a composite LCO/Li(Mn_1/3Ni_1/3Co_1/3)O₂ (NMC) cathode was compared to that of a standard LCO cathode [11]. Microstructural morphology of the composite cathode was also compared for samples imaged using FIB/SEM and XNT. In the LCO electrode, no significant change in microstructural parameters was seen after cycling. The continuous particle size distribution also revealed no change in small particle sizes after cycling and only a small change for larger particle sizes. For the LCO/NMC composite significant variability of the LCO and NMC volume fractions were seen both spatially within the same cathode and at different cycling conditions.

In addition to distinct regions of LCO and NMC, the capability for 3D observation of battery microstructure has enabled investigation of microstructural changes in Li-transition metal oxide particles with functionally graded Ni, Co, and Mn composition [12]. The functionally graded composition provides a method of taking advantage of the desired attributes of these transition metals while mitigating some of the negative effects that have been observed when using only one of these transition metals in a layered oxide. Controlling the composition of the active materials yielded benefits from the high capacity that is associated with nickel-rich materials while maintaining the thermal stability that is seen in manganese-rich materials. The graded composition was also more chemically stable in that detrimental phase transformations that have been recorded in LCO materials have slower kinetic rates for the functionally graded Ni, Co, and Mn composition, which decreases the loss of the reversible capacity of the battery as it is cycled.

At higher length scales, X-ray microtomography has been applied to the study of NMC cathodes by Ebner et al. [36]. The effects of compression on particle structure and distribution were assessed for sixteen cathodes, and the related tomographic and electrochemical data have been made available in an open source format. Fracture of individual particles was observed when the applied load compressed the electrode to a point where particle redistribution could not be accommodated. Packing of particles was found to be increased with higher compressive loads. Furthermore, particle packing was higher near the interior region of the cathode, away from the current collector and separator interfaces. In general, smaller particles dominated the regions near these two interfaces. A mix of large and small particles was observed near the midplane of the electrode. This distribution was attributed to the space filling capability of smaller particles. In addition to the tomographic data, electrochemical characterization was performed on comparable electrodes to establish a connection between the electrode microstructural data and battery performance. Of particular note, high-rate discharge behavior was found to be limited primarily by the quantity of conductive additive and not the applied compressive load.

In situ observation of Li-ion batteries using X-ray imaging techniques is receiving rapidly increasing attention. Several early studies based on 2D transmission imaging focused on to in situ observation of anode materials including Sn and CuO [7,33,34]. Morphological changes in Sn and SnO associated with lithiation and delithiation cycling were studied by Chao et al. [33,34] using 2D transmission X-ray microscopy. Variation in Sn concentration, Sn recrystallization, and pore formation within the active electrode particles was most pronounced during the first cycle of the electrode, suggesting these mechanisms may influence initial capacity fade. Improved kinetics and ductility were found to be key benefits related to smaller particles. The lithiation and delithiation behavior of CuO anodes were studied by Wang et al. using in operando 2D XANES imaging [7]. In these measurements, lithiation and delithiation were tracked by correlating X-ray absorption behavior within the sample to the formation of copper oxides that could be discerned based on XANES standards. Hard X-ray imaging has also permitted in operando 2D studies of cathodes for lithium sulfur batteries [35]. Key morphological changes in these studies were related to dissolution of the sulfur cathode. These changes included a particle size reduction and an increase in the apparent porosity of active sulfur particles.

A fuller picture of variations in electrode structure has been achieved in recent 3D in operando imaging studies of high capacity anode materials. Wang et al. achieved in situ XNT of Sn anodes using custom capillary cells [37]. A carbon substrate with low X-ray attenuation was applied as a current collector within this cell to facilitate imaging. Observations of Sn particles during cycling revealed regions of “stranded” lithium that reduced capacity during cycling. The structure of the electrodes was characterized with respect to feature size, specific surface area, and local curvature. Regions of high local curvature were found to act as stress concentrators within the active material.

Weker et al. [32] studied the morphology and electron density changes in Ge anode particles during operation. In operando X-ray radiography was performed just above the K-absorption edge of Ge to give high contrast between the micron-sized Ge particles and the rest of the battery structure. This contrast allows for the direct correlation of the brightness of the image to the change in particle electron density via the Beer–Lambert law. During lithiation, the particles were found to increase in size and fracture as the lithiation propagated from the edge of the particle toward the center. The fractured particles continue to expand until the cracks were filled and could no longer be seen. During delithiation, the particles reduce in volume, exposing the cracks that were formed by the lithiation process, but do not return to their original shape or size. Similar particles were also imaged in situ using tomographic imaging to show the change in density of the particles and the electrical connectivity of the particles before and after lithiation. The results of their study suggest that large porous Ge particles may be more beneficial in providing electrical contact with the current collector during cycling and thus improving the cycle-life of the battery.

X-ray tomography also allows for the in operando study of the morphological and chemical changes of electrode materials. Yang et al. [31] performed a study in which the chemical composition of NMC particle was determined over 200 cycles using a combination of 3D tomographic TXM and the resulting XANES data. The images provided information not only for the variations in morphology of the particles, but also how the metals react with one another as the battery is cycled. The initial state was found to be primarily NMC, but as the battery was cycled the original NMC began to break down into various components of Ni, Mn, and Co.

At larger length scales, Ebner et al. [38] performed in operando microtomography studies of lithiation and delithiation in SnO electrodes tested in a half cell configuration. A two-step lithiation process was mapped in 3D on the basis of X-ray attenuation coefficients. Conversion of SnO to lithia (Li₂O) and Sn was followed by the alloying reaction between Li and Sn, ultimately yielding Li_4.4Sn. Dealloying back to Sn and lithia was observed during the subsequent delithiation. The SnO conversion process during lithiation followed a shrinking core type behavior, while the alloying and dealloying reactions displayed a spatially homogeneous progression within the particles. For these studies, Ebner et al. synthesized active particles with a well-defined tetragonal morphology, creating in essence a more standard test article for tomographic study. The distinctive particle shape achieved in the SnO synthesis, combined with X-ray diffraction and SEM studies, permitted correlation of fracture behavior to the crystallographic planes within the SnO particles. Fracture due to lithiation was found to propagate sequentially on alternating sides of the particles, running parallel to the rectangular sides of the particles. This behavior suggested that fracture occurred preferentially along the (001) plane of the SnO particles. In addition to the detailed assessment of particle fracture behavior, Ebner et al. presented a multiscale assessment of electrode expansion behavior. The full thickness of the electrode and the volume fraction of active material were tracked during lithiation and delithiation. Swelling of the electrode during lithiation was accompanied by a steady increase in active material volume fraction during later stages of lithiation. This increase indicated that the matrix surrounding the active material could no longer accommodate the particle expansion.

X-ray microtomography has also been applied to study full batteries in operando. Eastwood et al. analyzed a coin cell battery using 3D X-ray computed tomography during standard operation [39]. The full cell analysis of the coin cell was performed to provide a macroscopic perspective of the cell performance in relation to the electrode microstructure. Time was also taken into account by repeated image acquisition as the cell was discharged and charged. Digital volume correlation was applied to track individual particles during the lithiation and delithiation process. The lithiated manganese oxide particles were found to vary in volume depending on the state of charge (SOC) and the distance the particle was from the terminal, with the particles farthest from the terminal showing the lowest change in volume. These results suggested that the lithiation of the electrode was limited by the electronic conductivity of the supporting matrix and not the transport of Li-ions through the electrode. Because of the composition of Li-ion batteries, thermal runaway is a significant risk factor during operation. To help mitigate this risk, Finegan et al. [40] implemented high-speed X-ray microtomography and radiography to track the structural changes in batteries during the initiation and propagation of thermal runaway. The group studied two similar cells with various modifications to analyze the effect of safety features that were implemented on the cells. Since this type of X-ray imaging had not been performed before the basis of the experiments was to establish baseline results and to provide proof of concept. Pre- and post-mortem 3D tomography were performed to examine how thermal runaway the effects of thermal runaway on the cells. The actual propagation of the thermal runaway was captured using high-speed radiography at 1250 fps allowing for the analysis of structural changes and thermal variations that occurred during the runaway.

The imaging studies discussed above yield digitized microstructural data that can be readily adapted for use as computational domains in mesoscale simulations. In this way, a battery model that simulates the multiphysics behavior of the 3D electrode microstructure can be developed. Such direct mesoscale simulation allows for the more detailed study of the interaction between transport, chemistry, mechanics, and microstructural geometry. As with the application of the imaging techniques, multiphysics analysis based on battery microstructures imaged using FIB/SEM and X-ray tomography emerged largely in parallel [4,8–10,41]. These studies were preceded and have been complemented by numerical studies based on simulated structures [42–45].

With respect to FIB/SEM, Less et al. [4] analyzed the behavior of Li concentration in simulated structures and in LiMn₂O₄ microstructure imaged using FIB/SEM. Finite volume modeling results revealed that simulated discharge behavior for the imaged microstructure matched observed discharge to within experimental uncertainty. Wiedemann et al. [41] analyzed both simulated and real microstructures as well, and demonstrated that at high discharge rates (5C) models of both structures yielded comparable predictions of voltage–current behavior. Global discharge characteristics were predicted to be similar at high states of charge and rates of discharge. At a 1C discharge rate and low SOC, the spherical model deviated from the model applying the imaged microstructure as a computational domain. At high discharge rates, where near-surface Li intercalation dominates, the discharge was predicted to be sensitive to the specific electrode–electrolyte surface area. At low discharge rates, where deeper intercalation is more dominant, the results are more sensitive to the shape and size of the electrode particles.

Laboratory-based micro- and nano-tomography measurements of LiCoO₂ (LCO) electrodes have recently been applied to obtain data for simulation of battery charge and discharge, as well as internal heat generation [8–10]. Using X-ray micro- and nano-tomography, Lim et al. [10] applied reconstructed particle structures as computational domains for calculating the diffusion-induced stresses in cathode particles. These stresses were compared at different C rates for both reconstructed particles and spherical particles. Diffusion induced stress in irregular particles was estimated to be significantly higher than stresses in spherical particles of the same volume. The stresses of both negative (Li_xC₆) and positive (Li_yCoO₂) materials were estimated to increase with C rate, reaching a maximum value at higher state of charge (SOC). In case of spherical particles, maximum stress was reached at a lower SOC than the reconstructed particles at the same C rates.

LIB thermal response to electrochemical stimuli is also influenced by the multiscale, heterogeneous nature of battery components. To better understand this influence, Yan et al. [9] have studied microstructural heat generation by simulating isothermal discharge within cathode microstructures imaged using XNT. This heat generation occurred at two critical interfaces: the active material/electrolyte interface and the active material/current collector interface. Entropic heat generation and heat of reaction, both interfacial phenomena, were found to dominate microstructural heat generation.

The 3D microstructural data obtained from the X-ray tomography measurements affords a key opportunity to simulate behavior in real battery structures. Efforts to integrate such data into multiscale studies are emerging for Li-ion batteries [46]. Continued development of such techniques, and attempts at direct validation through in situ and in operando, studies could prove to be a powerful tool for advancing the current understanding of coupled multiphysics phenomena within battery microstructures.

Understanding the multiphysics phenomena that drive battery performance and reliability is critical for the development of the advanced energy storage that will support a more sustainable energy infrastructure. Recent advances in 3D image-based investigations of energy materials present the opportunity to directly observe the mesoscale structure and composition of batteries and their component materials. Both FIB/SEM and X-ray tomography have been applied to achieve this goal for Li-ion batteries and other key energy materials. Both methods have been applied for ex situ imaging of LIB materials, and the 3D digitized microstructures obtained with these methods have been integrated into multiphysics modeling platforms to study the electrochemical, mechanical, and thermal behavior of electrode materials. In recent years, X-ray tomography has emerged for in situ and in operando imaging of battery materials. The nondestructive nature of X-ray imaging and cell-penetrating capabilities of hard X-ray imaging have facilitated this emergence. Such methods provide a unique opportunity for the direct observation of the multiphysics behavior of battery materials. Furthering these methods through multiscale imaging studies and continued integration with, and validation of, numerical simulations will strengthen the capability for advancing energy storage technology in the years to come.

Financial support from an NSF CAREER Award (CBET-1454437) is gratefully acknowledged.