Article Cite This: Acc. Chem. Res. 2018, 51, 2484−2492
pubs.acs.org/accounts
Mesoscale Battery Science: The Behavior of Electrode Particles Caught on a Multispectral X‑ray Camera Chenxi Wei,†,‡ Sihao Xia,†,§ Hai Huang,†,∥ Yuwei Mao,†,⊥ Piero Pianetta,† and Yijin Liu*,† †
Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, Menlo Park, California 94025, United States National Synchrotron Radiation Laboratory, University of Science and Technology of China, Hefei, Anhui 230027, China § School of Electronic and Optical Engineering, Nanjing University of Science and Technology, Nanjing, Jiangsu 210094, China ∥ State Key Laboratory of Infrared Physics, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China ⊥ School of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu 211100, China
Downloaded via REGIS UNIV on October 16, 2018 at 08:53:47 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
‡
CONSPECTUS: Functional materials and devices are usually morphologically complex and chemically heterogeneous. Their structures are often designed to be hierarchical because of the desired functionalities, which usually require many different components to work together in a coherent manner. The lithium ion battery, as an energy storage device, is a very typical example of this kind of structure. In a lithium ion battery, the cathode, anode, and separator are soaked in a liquid electrolyte, facilitating the back and forward shuttling of the lithium ions for energy storage and release. The desired performance of a lithium ion battery has many different aspects that need to be engineered and balanced depending on the targeted applications. In most cases, the cathode material has become the limiting factor for further improvements and, thus, has attracted intense attention from the research community. While the improvement in the overall performance of the lithium ion battery is the ultimate goal of the research in this field, understanding the relationship between the microscopic properties and the macroscopic behaviors of the materials/devices can inform the design of better battery chemistries for practical applications. As a result, it is of great fundamental and practical importance to investigate the electrode materials using experimental probes that can provide good chemical sensitivity and sufficient spatial resolution, ideally, under operating conditions. With this motivation, our group has been focusing on the development of the nanoscale full-field X-ray spectro-microscopy, which has now become a well-recognized tool for imaging battery electrode materials at the particle level. With nanoscale spatial resolution, this technique can effectively and efficiently tackle the intrinsically complicated mesoscale chemistry. It allows us to monitor the particles’ morphological and chemical evolution upon battery operation, providing valuable insights that can be incorporated into the design of new battery chemistries. In this Account, we review a series of our recent studies of battery electrode materials using nanoscale full-field X-ray spectromicroscopy. The materials that are the subjects of our studies, including layer-structured and spinel-structured oxide cathodes, are technically very important as they not only play an important role in today’s devices but also possess promising potential for future developments. We discuss how the subparticle level compositional and state-of-charge heterogeneity can be visualized and linked to the bulk performance through systematic quantification of the imaging data. Subsequently, we highlight recent ex situ and in situ observations of the cathode particles’ response to different reaction conditions, including the spontaneously adjusted reaction pathways and the morphological changes for the mechanical strain release. The important role of surface chemistry in the system is also discussed. While the microscopic investigation at the particle level provides useful insights, the degree to which this represents the overall properties of the battery is always a question for further generalizing the conclusions. In order to address this concern, we finally discuss a high throughput experimental approach, in which a large number of cathode particles are scanned. We discuss a case study that demonstrates the identification and analysis of functionally important minority phases in an operating battery cell through big data mining methods. With an emphasis on the data/information mining aspect of the nanoscale X-ray spectro-microscopic study of battery cathode particles, we anticipate that this Account will attract more research to this field.
1. INTRODUCTION There have been concerns regarding the sustainability of our modern society due to massive energy consumption, which, at the moment, heavily relies on fossil energy sources including oil (33%), coal (30%), and natural gas (24%).1 Such concern has © 2018 American Chemical Society
engendered the rapid growing research efforts that seek for alternative strategies for global energy management. An efficient Received: March 18, 2018 Published: June 11, 2018 2484
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
investigations of battery electrode particles. The machine learning approach and, more specifically, the advanced clustering algorithms are implemented to conduct supervised and unsupervised mining of the imaging data. The combination of the state-of-the-art experimental and computational techniques has facilitated the identification and visualization of unanticipated minority phases in a functional battery cell, which are key evidence of some unwanted side reactions that need to be mitigated properly in the development of next generation batteries. The X-ray microscopic studies of battery particles reviewed in this Account not only highlight the functional importance of the mesoscale chemical heterogeneity but also feature the novel capability of nanoscale X-ray spectro-microscopy and the associated data mining approaches, which are applicable to a wide range of research areas far beyond battery science.
and ecofriendly energy storage solution is a vital component in the landscape of our energy outlook, because it could dramatically change how energy is stored, transported, and consumed. The applications of lithium ion batteries (LIBs) in consumer electronics and electronic vehicles are good examples of how a new energy storage technology could expedite ground breaking changes in our lifestyle and could open vast opportunities with enormous market potential. Based on the targeted applications, the desired functionality of LIBs has many aspects that need to be considered and balanced. For example, while it is generally desirable to achieve high energy density, high power density, long cycle life, enhanced safety, and reduced cost, the application of grid energy storage emphasizes the need to accommodate highly intermittent renewable energy sources in a cost efficient manner, whereas the application of electrical vehicles more critically accentuates the power/energy density for the sake of accelerating power and cruising distance. Therefore, we have witnessed the devotion of a huge amount of effort into the research for new battery materials as well as for deeper understanding of the fundamental mechanisms underlying the battery performance improvement/degradation. LIBs are designed and engineered to be hierarchically complex in their structure. Comprehensive understanding of the fundamental reaction mechanisms requires systematic investigation of the electrode material over a wide range of length scales. We have seen impressive efforts on both the bulk scale and on the atomic scales through advanced analytical techniques, e.g., X-ray diffraction and electron microscopy.2 These efforts synergistically advance our understanding of the relationship between the performance at the device level and the materials’ microscopic structural and chemical properties. The particle/subparticle level investigation, i.e. the nanoscale to mesoscale, is however well regarded as an area of great importance but not very well understood due to the intrinsic complexity at this length scale as well as the lack of suitable analytical tools. With the motivation described above, we have been focusing on the development of nanoscale full-field X-ray spectromicroscopy,3 which has now become a well-recognized tool for imaging battery electrodes at the particle level. Combined with energy tunable X-rays sources, e.g., synchrotrons, we have demonstrated chemically sensitive X-ray imaging of battery materials with nanoscale resolution under both ex situ4 and in situ conditions.5 This technique has been adopted to monitor the morphological and chemical evolution of electrode particles in both 2D and 3D upon battery operation, providing valuable insights that could be compiled into design rules toward reaching the theoretical limits of the battery electrodes. A number of battery electrode materials were subjected to our spectro-microscopic investigations, including layer-structured (LiCoO2,6 LiNi1−x−yMnxCoyO2,4,7−9 and Li2Ru0.5Mn0.5O310) and spinel-structured (Li1+xMn2−xO411nd LiMn1.5Ni0.5O412) oxide cathodes. These materials are technically very important as they not only act as key players in today’s market but also possess promising potential for future developments. Our imaging results are linked to the material performance through systematic quantification of the spatially resolved spectroscopic data that probes the mesoscale chemistry, which often differs from the general expectations at the bulk level. Visualization of the reaction fronts and their migration is also conducted by analyzing the local chemical gradient, which causes the local overpotential that drives the diffusion of lithium ions and affects the spatial pathways. Finally, we look into the application of the novel developments in data science to the spectro-microscopic
2. MESOSCALE COMPOSITIONAL AND CHEMICAL HETEROGENEITY IN BATTERY PARTICLES Compositional and chemical heterogeneity ubiquitously exist in battery electrodes across a wide range of length scales. At the mesoscale, namely, at the particle and subparticle level, the heterogeneity is of fundamental and practical importance as its evolution could lead to the degradation and even failure of the battery. Understanding of the mesoscale phenomenon can, therefore, inform the design of new materials and chemistries for battery applications. While the mesoscale heterogeneity is often developed upon cycling of the battery cells, it could also be intentionally introduced through material engineering. Systematic and quantitative evaluation of the mesoscale heterogeneity in the battery electrode particles not only can help to understand the working principle of the materials but also can serve as key descriptors for material characterization. One example of successful engineering of particle level compositional heterogeneity is the synthesis of hierarchically structured LiNi0.4Mn0.4Co0.2O2 electrode by spray pyrolysis.7 The R3m LiNi1−x−yMnxCoyO2 (NMC) is a family of stoichiometric layered cathode materials that possess significant advantages in high power applications. By varying the relative concentration of the three transition metals (TM), the characteristics of the cathode material can be systematically tuned.13 On one hand, it is desirable for the auto industry to increase the Ni content in the NMC compounds, because Ni can empower the material with higher gravimetric energy density. On the other hand, high Ni content increases the tendency of the unwanted structural reconstruction from layered structure to a mixed spinel/rock-salt structure, which often take place on the particle surface and is related to the high-voltage failure. Consider the spray pyrolysis synthesized NMC electrode (Figure 1), which has demonstrated superior cyclability, although the global composition of the product can be well controlled, the three TM elements interestingly self-assemble, resulting in selective particle level segregation of the TMs. As shown in Figure 1, transmission X-ray microscopy (TXM) with nominal resolution at 30 nm was used to reveal the subparticle level elemental distribution through energy resolved tomographic measurements (Figure 1a). By conducting the 3D differential imaging above and below the absorption K-edges of the three TMs in the system, the distributions of Mn, Co and Ni were visualized by virtual slices through the center of the particle (Figure 1b−d), presenting considerable composition variation throughout the hollow spherical particle. The spatial correlation of Mn, Co and Ni can also be visualized by 3D rendering the elemental association maps (Figure 1e). Further quantification of 2485
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
systematic and quantitative evaluation of the mesoscale structural and chemical characteristics is very informative and valuable. We conducted systematic study of the electrochemical reaction induced morphological and chemical changes in the layer structured cathode particles to search for the key factors that are responsible for the performance degradation. In this work, the technologically relevant Li-rich Li2Ru0.5Mn0.5O3 were subjected to our investigation because it is a model Li rich layered oxide with 4d transition metal element.16 The presence of the Ru improves the electronic conductivity of the material and, thus, enhances its electrochemical performance. Although it has been reported that the substitution of Ru in the layered oxide could significantly suppress unwanted oxygen release,17 the voltage fade is still noticeable in Li2Ru0.5Mn0.5O3 upon cycling. Our goal is to understand the mesoscale structural and chemical characteristics and their roles in affecting the overall performance. It is evident from our result that the morphological degradation (Figure 2a−d) and the chemical evolution (Figures 2e−h) happened concurrently at the particle level as the battery is cycled. The pore space develops from isolated voids into a large and interconnected network. While the overall oxidation state of Mn reduces upon cycling, the diversity of the local chemical state increases, suggesting the development of significant mesoscale inhomogeneity in the local state of charge (SOC). At the early stage of the evolution, the depth dependence of the Mn oxidation state was developed because the oxygen release preferably takes place on the particle surface. As the interconnected pore network is developed upon further cycling, the liquid electrolyte wets into the open space, forming new reaction active interfaces. It is not surprising that the new interfaces exhibit different local chemistry compared to that of the original particle surface because they had a shorter exposure time to the electrolyte. In this scenario, the chemical heterogeneity is enhanced while the depth dependence of the Mn’s valence state breaks down. A comprehensive summary of the morphological and chemical characteristics is shown in Figure 2i, pointing to the structural integrity as a vital factor that influences the performance of the battery by critically affecting the mesoscale chemical and structural characteristics. More examples of the systematic quantification of the mesoscale compositional and chemical heterogeneity include but are not limited to (1) the study of the lithium−manganese rich NMC cathode (Li1.2Mn0.525Ni0.175Co0.1O2),4 (2) the evaluation of the persistent SOC heterogeneity in partially charged NMC (Li1−xNi1/3Mn1/3Co1/3O2) after long-term relaxation,18 and (3) the study of the synthesis−microstructure−performance relationship of Li1.5Ni0.25Mn0.75O2.5.19
Figure 1. Compositional heterogeneity in NMC secondary particle synthesized by spray pyrolysis. Panels (b), (c), and (d) are elemental distribution over the virtual slice through the center of the particle (panel a). Panel (e) shows the 3D rendering of the elemental association maps, which is color coded to the corresponding inset. Panel (f) shows the depth profile of the compositional distribution. The scale bar in panel (e) is 5 μm. Reprinted with permission from ref 7. Copyright 2016 Springer Nature.
the compositional depth profile (Figure 1f) suggests that the surface layer of the particle is relatively poor in Ni/Co while rich in Mn, forming a robust surface layer that protects the interior of the particle against the well-known surface reconstruction effect.14 In this case study, the X-ray spectro-tomography technique critically contributed to understanding the mechanism behind the improved cyclability of the spray pyrolysis synthesized NMC cathode through resolving the mesoscale compositional heterogeneity. The quantification of the elemental association and the depth profile was critical in connecting the experimental observation at the mesoscale to the bulk performance. We point out here that the engineering of particle level compositional heterogeneity has also been achieved using other sophisticated synthesis methods.15 The above-discussed X-ray spectrotomography technique is well applicable to these systems as well. As discussed, the mesoscale chemical heterogeneity could also be induced through normal operation of the battery. In this circumstance, it is often useful to correlate the mesoscale morphological and chemical features with electrochemical performance. The battery electrode cycled under realistic conditions could experience many different processes including local over charge, local over discharge, surface reconstruction, solid electrolyte interphase (SEI) formation, oxygen release, detouring of the ionic diffusion, change of electronic contact, change of electrolyte wetting, buildup of mechanical strain, crack formation/propagation and cation dissolution/precipitation. All of these side reactions contribute to the mesoscale heterogeneity, which may eventually lead to device failure. As a result,
3. MESOSCALE REACTION PATHWAYS IN BATTERY PARTICLES As discussed above, the mesoscale compositional and chemical heterogeneity can be visualized and quantified using the nanoscale X-ray spectro-microscopy technique, providing important insights into the complicated interplays of fine structural and chemical components within and between the battery particles. Going beyond the quantification analysis discussed above, the inhomogeneous distribution of the valence states of the TMs can also serve as key descriptors for evaluating the local electrochemical potential. The local chemical gradient, therefore, can be used to represent the local over potential, which drives the ionic diffusion at the mesoscale. For better understanding of the dynamic evolution of the state of charge (SOC) distribution at the particle level, which is directly linked to the ionic diffusion kinetics and pathways, we 2486
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
Figure 2. Nanoscale X-ray spectro-microscopic study of Li2Ru0.5Mn0.5O3 particles that have gone through different cycling history (pristine, 01-cycled, 10-cycled and 20-cycled). Panels (a)−(d) show the 3D rendering of the particles, highlighting the internal pore space (black) that develops from isolated voids into a large and interconnected network. Panels (e)−(h) are two-dimensional maps of Mn’s valence state revealed by the nanoscale full-field transmission X-ray spectro-microscopy. Panel (i) is the summary of the mesoscale morphological and chemical changes, including the porosity, the morphological complexity, the pore space connectivity, the oxidation state of Mn as well as the diversity in optical density and the valence state. The scale bar in panel (e) is 5 μm. Reprinted with permission from ref 10. Copyright 2016 Elsevier.
conducted a nanoscale X-ray spectro-microscopic study of a LiCoO2 particle inside an operating cell as it was cycled at different rates.6 We developed a robust battery pouch setup that enables the direct operando observation of the local SOC distribution as well as the particle’s recovery rate, both are dependent on the cycling rate (Figure 3a). More interestingly, the operando SOC maps revealed the redistribution of the inactive domains (Figure 3b and c) within the particle upon repeated cycling under stabilized conditions. The change in the distribution of the inactive domains is accompanied by the rearrangement of the local chemical gradient, which was concentrated on the particle surface in early stage of the battery operation and gradually migrated to the center of the particle during the repeated electrochemical activation process. The above-discussed in situ observation reveals a single particle’s response to different local chemical environment, which is governed by the cycling rate and history. The rearrangement of the local chemical gradient illustrated the detouring of the ionic diffusion pathway over long-term cycling (Figure 3b and c). The rearrangement of the ionic diffusion pathway could lead to further consequences including deactivation of local domains, accumulation of mechanical strain, and formation of morphological or
chemical defects. In a separate in situ study of the Li1+xMn2−xO4 single crystal particles,11 we clearly observed the spatial correlation of the large local chemical gradient and the formation of fractures and cracks. It can be envisaged that a high local chemical gradient causes large local ionic currents, building up the mechanical strain, which, eventually, is released through cracking of the material. More sophisticated 3D propagation of the reaction fronts in a LiFePO4 particle has also been visualized by Wang et al. using nanoscale X-ray spectro-tomography.20 Although it is reported that the ionic diffusion in the olivine structure of LiFePO4 could exhibit a 2D behavior,21 it primarily favors one direction along the 1D channels.22 It is very interesting to see that the 3D anisotropic and isotropic propagation of the reaction fronts coexist in the same particle, but dominate the system at different SOC respectively. While in situ observation of the mesoscale reaction pathways within battery particles can offer valuable insights relevant to real world battery operation, it is necessary to point out that the observed phenomenon might not be exclusively related to the intrinsic behavior of the particles. The local electrochemical environment in a real battery cell is rather complex. Many factors, including the electronic contact and electrolyte wetting, could 2487
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
Figure 3. Operando observation of a LiCoO2 particle’s response to different cycling conditions. Panel (a) shows the evolution of the mesoscale distribution of state of charge (SOC) as the particle is cycled at different rates. Panels (b) and (c) show the distribution of the inactive phase as well as the chemical gradient within the active particle when it was cycled at 0.2 C for the first time (panel b) and the 20th time (panel c). The black arrows in panels (b) and (c) highlight the redistribution of the inactive domains upon repeated cycling at 0.2 C. The scale bar in panel a is 5 μm. Reprinted with permission from ref 6. Copyright 2017 American Chemical Society.
Figure 4. Three dimensional visualization of the Li diffusion pathways in octahedron-shaped spinel LiMn1.5Ni0.5O4 single crystal particles upon chemical delithiation. The shape of the particle is presented as the transparent surface with the internal oxidation state heterogeneity illustrated using the diagonal slices (panels a−d) and the surfaces of the 3D Ni oxidation state map (panels e−j). Panel (k) shows a high resolution scanning electron microscopic image, highlighting the truncation of some corners of the crystal, which exposes the (100) facet (panel m) instead of the (111) facet (panel l). The scale bar in panel (k) is 1 μm. Reprinted with permission from ref 12. Copyright 2017 Springer Nature.
significantly affect the mesoscale diffusion pathways. In the in situ experiments discussed above,6,20,21 it is likely that nonuniform electrical contact and electrolyte wetting were the dominating factors that dictated the electronic and ionic diffusion within the particle. As a result, an alternative approach is needed to eliminate these effects and to gain more fundamental understanding of the particle’s intrinsic behaviors. The solution-based chemical oxidation process has emerged as a promising method for preparing partially reacted battery electrode materials free of inactive components. In this process, the solid active electrode materials are immersed in the oxidizing solution, ensuring a homogeneous oxidizing environment. Depending on the choice of the oxidant and the molar ratio between the active material and the oxidant, the over potential and the final SOC can be effectively tuned. The chemical delithiation shows admirable advantages for certain types of studies and has been widely adapted in battery research. Our recent study has also shown that the chemical oxidation and the electrochemical cycling produce similar degree of chemical heterogeneity at particle level, justifying the X-ray spectromicroscopic study of chemically oxidized battery materials, although the surface (up to 10 nm) chemistry may differ.8 We, therefore, turned to studying the chemically delithiated single crystal particle of spinel LiMn1.5Ni0.5O4, which is technologically significant as it possesses promising high-rate capability. We aim to investigate how particles intrinsically manage its internal ionic diffusion. Since the particles are exposed to a homogeneously applied driving force for the reaction, common wisdom would suggest a classic shrinking-core process. Indeed, the core−shell separation has been observed in the chemically oxidized LiFePO4 particles,23 in good agreement with the general expectation. However, we observed a clear trend of the Ni’s valence state across the particle (Figure 4b−e), suggesting that the oxidation reaction initiates at two of the corner vertices and propagate through the particle in a rather
complicated pattern (Figure 4f−j). This phenomenon is attributed to the truncation of some of the crystal corners (Figure 4k), which exposes the (100) crystal facet (Figure 4m) that is more favorable to the Li extraction/insertion than the (111) facet (Figure 4l). While unambiguously invalidating the notion of the core−shell model for the octahedron-shaped spinel LiMn1.5Ni0.5O4 single crystal particles, our results suggest that the exposure of different crystal facets to the reaction driving force will result in different optimal diffusion pathways. It is, therefore, important to consider the different kinetics associated with crystal facets in the design of the battery materials. The single crystal particle of LiMn1.5Ni0.5O4 is certainly an excellent model system for studying the reaction kinetic. In the more general case, however, the solid state phase transformation can spontaneously initiate and develop in polycrystalline ensembles, which is a more common form for battery electrodes. The mesoscale reaction pathways are further complicated due to the coexistence of grain boundaries, compositional variation, anisotropic volume change, local strain, fractures and cracks. Our recent study of a thermally driven solid state phase transformation in polycrystalline LiNi0.4Mn0.4Co0.2O2 reveals that it follows a complex pathway that was defined by the local curvature of the valence state topographies.9 We, therefore, speculated that good management of the primary particles agglomeration could be an effective method for tuning the mesoscale battery chemistry.
4. DATA MINING FOR IDENTIFICATION OF THE FUNCTIONALLY IMPORTANT MINORITY PHASES AT MESOSCALE As discussed above, X-ray spectro-microscopy has been successful in visualizing and quantifying the mesoscale 2488
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
Figure 5. Supervised machine learning approach for extracting the 2D and 3D mesoscale structural and chemical heterogeneity from the full-field X-ray spectro-microscopy data. Panel (a) illustrates the data structure. Panel (b) shows the prior knowledge of a list of anticipated principal chemical species in the sample, which is indispensable for the chemical mapping in this approach. Panel (c) illustrates the quantification procedures including the linear combination fitting, the color coding, and the tomographic reconstruction. Panel (d) shows the 3D mesoscale structural and chemical heterogeneity of a cluster of partially reduced NiO electrode particles. The scale bar in panel (d) is 5 μm. Reprinted with permission from ref 24. Copyright 2011 International Union of Crystallography.
spectroscopic fingerprints quantified by means of linear combination fitting (Figure 5c). The corresponding pixels are, subsequently, color coded to illustrate the distribution of the local valence state. This process can be repeated as a function of the viewing angle, enabling the tomographic reconstruction of the 3D chemical heterogeneity (Figure 5d). The results shown in Figure 5 suggest that, upon cycling, the micron-sized NiO electrode particle agglomerate is falling apart while being reduced to metallic form. The above-described supervised machine learning approach has been very popular and successful. It has been applied to the X-ray spectro-microscopic data collected using different experimental configurations, including the full-field transmission X-ray microscopy,24 the scanning transmission X-ray microscopy,25 and the soft X-ray spectro-ptychography.26 However, two questions are naturally raised: (1) Are the investigated particles representative? (2) Is the prior knowledge that facilitated the supervised data mining complete and accurate? Indeed, if there is too much reliance on the prior knowledge, the opportunity to discover unknown chemical species, which could actually be a critical piece of information for understanding the mesoscale chemistry, may be hindered. As a result, we looked into the novel computational developments in data science. We adapted and modified the novel clustering algorithms to conduct a search through the massive X-ray spectro-microscopic data. Clusters identified by the
compositional/chemical heterogeneity and, subsequently, in revealing the reaction’s spatial pathways. While we are productively harvesting the scientifically valuable structural and chemical information, the limitations of the established workflow have caught our attention. In full-field X-ray spectro-microscopy, as we are collecting spatially resolved spectroscopic data using an area detector of 1k × 1k or 2k × 2k pixels, the data rate is 6 orders of magnitude higher than that of the traditional bulk X-ray spectroscopic measurements, making it practically impossible to interact with every single spectrum for detailed analysis. A good degree of automation is clearly needed and has been implemented for several data reduction steps. Going beyond the data reduction, the conversion of “data” into “knowledge” requires comprehensive interpretation of the spectroscopic features, which is usually nontrivial and, thus, has been identified as the bottleneck in the procedure. The mesoscale spectro-microscopic studies of battery particles discussed in the previous sections rely heavily on a supervised machine learning approach, which is illustrated in Figure 5. This approach is straightforward and, yet, very effective and efficient. Based on the prior knowledge of a list of anticipated principal chemical species in the sample (Figure 5b, which is NiO and metallic Ni in this case), the XANES spectra from every single pixel (Figure 5a) are compared against the know standards for quantification of the local chemical state. The local chemical composition is determined based on the similarity in the 2489
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
Figure 6. Unsupervised machine learning aided full-field spectro-microscopic study of LiCoO2 battery. The pouch cell was raster scanned to collect the spectro-microscopic data in more than a hundred active particles with over 10 million spatially resolved XANES spectra. The clustering analysis suggested that while majority of the particles went through the anticipated lithiation and delithiation, evidence of two unwanted side reactions was also observed. Reprinted with permission from ref 28. Copyright 2017 American Chemical Society.
engineered or spontaneously developed upon cycling of the battery. Many processes can contribute to the development of the mesoscale structural and chemical heterogeneity. Systematic investigation of the mesoscale phenomenon can lead to valuable understandings that could critically inform the design of next generation battery materials. X-ray spectro-microscopy has emerged as a promising tool for studying battery electrodes at mesoscale. The combination of nanoscale spatial resolution and chemical sensitivity has been successful and productive in the studies of a number of technologically significant battery cathode materials. In this Account, we first reviewed a few case studies, in which the systematic quantification of the imaging data has yielded critical understanding of the fundamental mechanisms behind the battery performance improvement/degradation. In the second part, the visualization of the local SOC distribution was discussed, highlighting the battery particles’ dynamic response to different reaction conditions and to different crystal facet kinetics. Finally, we featured the application of novel machine learning approaches in X-ray spectro-microscopic investigations of massive imaging data. The supervised and unsupervised data mining approaches can efficiently and effectively extract scientifically relevant information from the big data and, thus, can greatly complement the researchers with domain expertise. While we are excited about the above presented capabilities, we see vast opportunities associated with the further technical developments. For example, with the developments in the X-ray optics and the next generation synchrotron facilities with unprecedented coherence properties, the spatial and temporal resolution of the X-ray imaging techniques could be further improved. The X-ray free electron lasers (XFELs) can even facilitate the investigation of the material’s ultrafast behavior at a state far away from the equilibrium. Going beyond the spectroscopic contrast, it is also desirable to incorporate the other image contrast mechanisms, e.g., the phase contrast and the diffraction contrast. By working at different X-ray energy range, we can also probe the structural and chemical heterogeneity on the surface or in the bulk. Indeed, this is a growing area with very active scientific and technological activities. We wish that this Account will call for
numerical search can be further evaluated by researchers with domain expertise (in our case, domain expertise indicates the knowledge about the X-ray spectroscopic features and the battery chemistry) to interpret the scientific significance of the result.27 We adapted this approach to investigate a functional LiCoO2 pouch cell after it has gone through a designed cycling sequence.6 By raster scanning the cell, we conducted spectro-imaging over several selected field-of-views, covering more than a hundred active particles with over 10 million spatially resolved XANES spectra.28 Due to the limited signal-to-noise ratio in the data, direct application of the existing clustering algorithms had limited success. As a result, we first extracted several key spectroscopic data attributes as features and evaluated the correlation among them to make sure that we do not over emphasis certain aspects of the data more than intended. These features were further grouped and weighted, before they were fed into the computing engine for clustering. In this work, the above-described machine learning approach identified four different chemical clusters. Clusters 1 and 2 are normal LixCoO2 particles, which have experienced the anticipated lithiation/delithiation. The unanticipated clusters 3 and 4 point to two different side reactions by which the cell capacity may degrade upon operation (Figure 6). The direct observation of metallic Co (cluster 4) confirms the common belief that at high charging voltages the transition metal dissolves out from the layered LiCoO2 cathode and is subsequently reduced at the anode due to side reactions with electrolyte. Discovery of cluster 3, which is at lower oxidation state than Co3+, in the core region of a particle is attributed to local overlithiation caused domain deactivation (mechanical failure). This study not only features the identification of unwanted side reactions in the LiCoO2 battery, but also highlights the capability of the combined state-of-the-art experimental and computational techniques, which facilitated the discoveries that were otherwise inaccessible.
5. CONCLUDING REMARKS Mesoscale structural and chemical heterogeneity plays a critical role in affecting device level battery performance. The mesoscale compositional and chemical heterogeneity may be intentionally 2490
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research
Techniques for Studying Materials Electrochemistry in Rechargeable Batteries. Chem. Rev. 2017, 117, 13123−13186. (3) Liu, Y.; Andrews, J. C.; Meirer, F.; Mehta, A.; Gil, S. C.; Sciau, P.; Mester, Z.; Pianetta, P. Applications of Hard X-ray Full-Field Transmission X-ray Microscopy at SSRL. AIP Conf. Proc. 2010, 1365, 357−360. (4) Yang, F.; Liu, Y.; Martha, S. K.; Wu, Z.; Andrews, J. C.; Ice, G. E.; Pianetta, P.; Nanda, J. Nanoscale Morphological and Chemical Changes of High Voltage Lithium−Manganese Rich NMC Composite Cathodes with Cycling. Nano Lett. 2014, 14, 4334−4341. (5) Nelson Weker, J.; Li, Y.; Shanmugam, R.; Lai, W.; Chueh, W. C. Tracking Non-Uniform Mesoscale Transport in LiFePO4 Agglomerates During Electrochemical Cycling. ChemElectroChem 2015, 2, 1576− 1581. (6) Xu, Y.; Hu, E.; Zhang, K.; Wang, X.; Borzenets, V.; Sun, Z.; Pianetta, P.; Yu, X.; Liu, Y.; Yang, X.-Q.; Li, H. In Situ Visualization of State-of-Charge Heterogeneity within a LiCoO2 Particle That Evolves upon Cycling at Different Rates. ACS Energy Lett. 2017, 2, 1240−1245. (7) Lin, F.; Nordlund, D.; Li, Y.; Quan, M. K.; Cheng, L.; Weng, T.-C.; Liu, Y.; Xin, H. L.; Doeff, M. M. Metal Segregation in Hierarchically Structured Cathode Materials for High-Energy Lithium Batteries. Nat. Energy 2016, 1, 15004. (8) Tian, C.; Xu, Y.; Nordlund, D.; Lin, F.; Liu, J.; Sun, Z.; Liu, Y.; Doeff, M. Charge Heterogeneity and Surface Chemistry in Polycrystalline Cathode Materials. Joule 2018, 2, 464−477. (9) Mu, L.; Yuan, Q.; Tian, C.; Zhang, K.; Liu, J.; Pianetta, P.; Doeff, M. M.; Liu, Y.; Lin, F. Propagation Topography of Redox Phase Transformations in Heterogeneous Layered Oxide Cathode Materials. Nat. Commun. 2018, submitted for publication. (10) Xu, Y.; Hu, E.; Yang, F.; Corbett, J.; Sun, Z.; Lyu, Y.; Yu, X.; Liu, Y.; Yang, X.-Q.; Li, H. Structural IntegritySearching the Key Factor to Suppress the Voltage Fade of Li-Rich Layered Cathode Materials through 3D X-Ray Imaging and Spectroscopy Techniques. Nano Energy 2016, 28, 164−171. (11) Yu, Y.-S.; Kim, C.; Liu, Y.; van der Ven, A.; Meng, Y. S.; Kostecki, R.; Cabana, J. Nonequilibrium Pathways during Electrochemical Phase Transformations in Single Crystals Revealed by Dynamic Chemical Imaging at Nanoscale Resolution. Adv. Energy Mater. 2015, 5, 1402040. (12) Kuppan, S.; Xu, Y.; Liu, Y.; Chen, G. Phase Transformation Mechanism in Lithium Manganese Nickel Oxide Revealed by SingleCrystal Hard X-Ray Microscopy. Nat. Commun. 2017, 8, 14309. (13) Xu, J.; Lin, F.; Doeff, M. M.; Tong, W. A Review of Ni-Based Layered Oxides for Rechargeable Li-Ion Batteries. J. Mater. Chem. A 2017, 5, 874−901. (14) Lin, F.; Markus, I. M.; Nordlund, D.; Weng, T.-C.; Asta, M. D.; Xin, H. L.; Doeff, M. M. Surface Reconstruction and Chemical Evolution of Stoichiometric Layered Cathode Materials for Lithium-Ion Batteries. Nat. Commun. 2014, 5, 3529. (15) Sun, Y.-K.; Myung, S.-T.; Park, B.-C.; Prakash, J.; Belharouak, I.; Amine, K. High-Energy Cathode Material for Long-Life and Safe Lithium Batteries. Nat. Mater. 2009, 8, 320−324. (16) Mori, D.; Sakaebe, H.; Shikano, M.; Kojitani, H.; Tatsumi, K.; Inaguma, Y. Synthesis, Phase Relation and Electrical and Electrochemical Properties of Ruthenium-Substituted Li2MnO3 as a Novel Cathode Material. J. Power Sources 2011, 196, 6934−6938. (17) Sathiya, M.; Ramesha, K.; Rousse, G.; Foix, D.; Gonbeau, D.; Prakash, A. S.; Doublet, M. L.; Hemalatha, K.; Tarascon, J.-M. High Performance Li2Ru1−yMnyO3 (0.2 ≤ y ≤ 0.8) Cathode Materials for Rechargeable Lithium-Ion Batteries: Their Understanding. Chem. Mater. 2013, 25, 1121−1131. (18) Gent, W. E.; Li, Y.; Ahn, S.; Lim, J.; Liu, Y.; Wise, A. M.; Gopal, C. B.; Mueller, D. N.; Davis, R.; Nelson Weker, J.; Park, J.-H.; Doo, S.-K.; Chueh, W. C. Persistent State-of-Charge Heterogeneity in Relaxed, Partially Charged Li1−xNi1/3Co1/3Mn1/3 O2 Secondary Particles. Adv. Mater. 2016, 28, 6631−6638. (19) Chen, W.-C.; Song, Y.-F.; Wang, C.-C.; Liu, Y.; Morris, D. T.; Pianetta, P. A.; Andrews, J. C.; Wu, H.-C.; Wu, N.-L. Study on the Synthesis−Microstructure-Performance Relationship of Layered LiExcess Nickel−Manganese Oxide as a Li-Ion Battery Cathode Prepared
even more endeavors into mesoscale science, where the structural and chemical complexity at the fine scale and the overall performance at the macroscopic level meet.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. ORCID
Yijin Liu: 0000-0002-8417-2488 Author Contributions
C.W., S.X., H.H., and Y.M. contributed equally to this work. Notes
The authors declare no competing financial interest. Biographies Chenxi Wei is a PhD candidate at University of Science and Technology of China and a visiting student at Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory. Her research interest focuses on multimodality X-ray imaging methods and their application in the research of energy materials. Sihao Xia is a PhD candidate at Nanjing University of Science and Technology and a visiting student at Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory. His research interests are related to semiconductor based optoelectronic devices, first-principles study of III−V nanomaterials, and data mining in X-ray microscopy. Hai Huang is a PhD candidate at Shanghai Institute of Technical Physics and a visiting student at Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory. His research interest focuses on the study of optoelectronic materials and superconductors using synchrotron techniques. Yuwei Mao is a postgraduate student at Nanjing University of Aeronautics and Astronautics. Her research interests include machine learning, deep learning, and data mining methods. Piero Pianetta is a professor of electronic engineering at Stanford University. He is also the deputy director of Stanford Synchrotron Radiation Lightsource, the chair of the photon science department at SLAC National Accelerator Laboratory. His research covers synchrotron based surface science and material science. Yijin Liu is a staff scientist at Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory. With over 10 years of experience in X-ray microscopy and the associated big data mining methods, his scientific passion lies in the research of energy materials.
■
ACKNOWLEDGMENTS The authors gratefully thank Johanna Weker for valuable discussions. Use of the Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, is supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC0276SF00515.
■
REFERENCES
(1) Chu, S.; Majumdar, A. Opportunities and Challenges for a Sustainable Energy Future. Nature 2012, 488, 294−303. (2) Lin, F.; Liu, Y.; Yu, X.; Cheng, L.; Singer, A.; Shpyrko, O. G.; Xin, H. L.; Tamura, N.; Tian, C.; Weng, T.-C.; Yang, X.-Q.; Meng, Y.-S.; Nordlund, D.; Yang, W.; Doeff, M. M. Synchrotron X-Ray Analytical 2491
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
Article
Accounts of Chemical Research by High-Temperature Calcination. J. Mater. Chem. A 2013, 1, 10847− 10856. (20) Wang, J.; Chen-Wiegart, Y.-K.; Eng, C.; Shen, Q.; Wang, J. Visualization of Anisotropic-Isotropic Phase Transformation Dynamics in Battery Electrode Particles. Nat. Commun. 2016, 7, 12372. (21) Hong, L.; Li, L.; Chen-Wiegart, Y.-K.; Wang, J.; Xiang, K.; Gan, L.; Li, W.; Meng, F.; Wang, F.; Wang, J.; Chiang, Y.-M.; Jin, S.; Tang, M. Two-Dimensional Lithium Diffusion Behavior and Probable Hybrid Phase Transformation Kinetics in Olivine Lithium Iron Phosphate. Nat. Commun. 2017, 8, 1194. (22) Chen, G.; Song, X.; Richardson, T. J. Electron Microscopy Study of the LiFePO4 to FePO4 Phase Transition. Electrochem. Solid-State Lett. 2006, 9, A295−A298. (23) Lachal, M.; Bouchet, R.; Boulineau, A.; Surblé, S.; Rossignol, C.; Alloin, F.; Obbade, S. Remarkable Impact of Grains Boundaries on the Chemical Delithiation Kinetics of LiFePO4. Solid State Ionics 2017, 300, 187−194. (24) Meirer, F.; Cabana, J.; Liu, Y.; Mehta, A.; Andrews, J. C.; Pianetta, P. Three-Dimensional Imaging of Chemical Phase Transformations at the Nanoscale with Full-Field Transmission X-Ray Microscopy. J. Synchrotron Radiat. 2011, 18, 773−781. (25) Lim, J.; Li, Y.; Alsem, D. H.; So, H.; Lee, S. C.; Bai, P.; Cogswell, D. A.; Liu, X.; Jin, N.; Yu, Y.-S.; Salmon, N. J.; Shapiro, D. A.; Bazant, M. Z.; Tyliszczak, T.; Chueh, W. C. Origin and hysteresis of lithium compositional spatiodynamics within battery primary particles. Science 2016, 353, 566−571. (26) Yu, Y.-S.; Farmand, M.; Kim, C.; Liu, Y.; Grey, C. P.; Strobridge, F. C.; Tyliszczak, T.; Celestre, R.; Denes, P.; Joseph, J.; Krishnan, H.; Maia, F. R. N. C.; Kilcoyne, A. L. D.; Marchesini, S.; Leite, T. P. C.; Warwick, T.; Padmore, H.; Cabana, J.; Shapiro, D. A. Three Dimensional Localization of Nanoscale Battery Reactions Using Soft X-Ray Tomography. Nat. Commun. 2018, 9, 921. (27) Duan, X.; Yang, F.; Antono, E.; Yang, W.; Pianetta, P.; Ermon, S.; Mehta, A.; Liu, Y. Unsupervised Data Mining in Nanoscale X-Ray Spectro-Microscopic Study of NdFeB Magnet. Sci. Rep. 2016, 6, 34406. (28) Zhang, K.; Ren, F.; Wang, X.; Hu, E.; Xu, Y.; Yang, X.-Q.; Li, H.; Chen, L.; Pianetta, P.; Mehta, A.; Yu, X.; Liu, Y. Finding a Needle in the Haystack: Identification of Functionally Important Minority Phases in an Operating Battery. Nano Lett. 2017, 17, 7782−7788.
2492
DOI: 10.1021/acs.accounts.8b00123 Acc. Chem. Res. 2018, 51, 2484−2492
