Correlated X-Ray 3D Ptychography and Diffraction Microscopy Visualize Links between Morphology and Crystal Structure of Lithium-Rich Cathode Materials

Summary

The search for higher performance, improved safety, and lifetime of lithium-ion batteries relies on the understanding of degradation mechanisms. Complementary to methods and studies on primary particles or crystalline structure on bulk materials, here we use spatially correlated ptychographic X-ray computed nanotomography with a 35 nm resolution and scanning X-ray diffraction microscopy with 1 μm resolution to visualize in 3D the hidden morphological and structural degradation processes in individual secondary particles of lithium-rich nickel, cobalt, and manganese oxides. From comparative examination of pristine and cycled particles, we suggest that morphological degradation could have radial dependency and secondary particle size dependency. The same particles were examined to correlate the degradation to crystallinity, which shows surprising core-shell structures. This study reveals the inner 3D structure of the secondary particles while opening up questions on the unexpected crystalline structural distributions, which could offer clues for future studies on this promising cathode material for lithium-ion batteries.

Graphical Abstract

Highlights

• •Correlative ptychographic tomography with 35 nm resolution and diffraction tomography
• •3D micro-cracks in secondary Li-rich NMC particles correlated with crystal lattice
• •Radial and secondary-particle-size-dependent material degradation
• •Intriguing core-shell structure observed in 3D crystalline maps

Materials Characterization Techniques; Energy Materials; Crystallography

Introduction

Lithium-rich nickel, manganese, and cobalt oxides (Li-rich NMC) offer considerable performance advantages over industry standards for cathode materials used in lithium-ion batteries (Thackeray et al., 2007, Rozier and Tarascon, 2015, Wang et al., 2016, Pimenta et al., 2017, Etacheri et al., 2011, Schipper et al., 2017). These advantages include a high average working potential, ca. 3.8 V versus Li⁺/Li, and a specific charge of over 250 mAh/g, which corresponds to an energy density of over 900 Wh/kg, higher than that of LiFePO₄ or LiCoO₂ materials. However, the high operating voltage, first cycle activation, repeated lithiation/delithiation phenomenon, and the anionic redox processes result in structural and chemical degradation, impeding their commercialization. Degradation processes upon cycling and their prevention are thus under extensive investigation (Liu et al., 2017, Croy et al., 2016, Klein et al., 2015, Boulet-Roblin et al., 2015, Watanabe et al., 2014, Amalraj et al., 2013). This includes examination of alternative synthesis procedures, protective coatings or surface treatments, dopings, or modification of the cycling protocol. Commonly, studies examine processes on the primary particle or the bulk level (Kang et al., 2007, Lim et al., 2009, Li et al., 2013, Mohanty et al., 2014, Sasakawa et al., 2015, Gilbert et al., 2017, Kuppan et al., 2017). It was identified that electrochemical activation of the material occurring at the first delithiation is accompanied by oxygen loss from the crystal structure (Lu and Dahn, 2002, Tran et al., 2008, Mu et al., 2018). Continued cycling was shown to not only promote the formation of a new unstable spinel-like phase growing from the surface in the direction of the core but also highlighted the passivation of electrode surfaces by electrolyte decomposition and the degradation of the active material itself by means of transition metal leaching and lithium loss (Gu et al., 2013, Lin et al., 2014, Mukhopadhyay and Sheldon, 2014, Yan et al., 2017, Strehle et al., 2017, Zhan et al., 2018, Tan et al., 2014).

These predominantly chemistry-driven degradation processes are accompanied by well-known but less studied morphological degradation processes in particular on the single secondary particle level (Lee et al., 2014, Yang et al., 2014, Chen et al., 2016, Sun and Manthiram, 2017, Kondrakov et al., 2017, Ryu et al., 2018, Ko et al., 2019, Gent et al., 2016). The Li-rich NMC particles studied in this work are in the form of micron-sized spheres composed of thousands of tightly packed primary particles, 50 to 100 nm in diameter, and possess a synthesis-inherent onion-like structure. During cycling, lithium is extracted from and reinserted into the interlayer, generating stress along the c-axis of the oxide by up to 2% changes (Yoon et al., 2006). Although the inner porosity of such compounds can accommodate part of these volume changes, the changes can eventually generate a significant stress that leads to the formation of cracks on the secondary particles, i.e., micro-cracks, and in turn pulverization, ultimately resulting in battery failures due to the loss of electronic contact.

The high resolution provided by methods such as electron microscopy or atomic force microscopy (Etacheri et al., 2011, Schipper et al., 2017) allows for studying the surface of the material and the primary particles. Complementary to such studies, here we show a study on the morphological and crystalline structural changes for the secondary particles, offering insight into the battery material from a different perspective on a different scale. Through spatially correlated ptychographic X-ray computed tomography (PXCT) and scanning X-ray diffraction microscopy (SXDM), we visualize these changes in 3D within individual secondary particles. Li-rich NMC secondary particles were examined in their pristine state, after the first delithiation, and after long-term cycling, in which the cells were stopped after full lithiation. All particles in this work refer to secondary particles unless stated otherwise. PXCT (Dierolf et al., 2010, Holler et al., 2014) is a lensless imaging technique that provides quantitative electron density tomograms of extended system at spatial resolution levels hardly achievable by common X-ray microscopic techniques, here 35 to 70 nm. Morphological changes were observed in aged particles through PXCT and correlated with the expansion of the unit cell volume through subsequent SXDM measurements with 1 μm resolution on the same particles. X-ray fluorescence (XRF) spectra were also collected with the SXDM measurements simultaneously.

Results and Discussion

Electrochemical Performance and Morphological Characterization

The studied Li-rich NMC particles were obtained from a commercial source (see Transparent Methods) with a composition of Li_y(Ni_0.19Mn_0.54Co_0.1)O_2−x, where y = 1.17 and x = 0 in the pristine stage and, after the first delithiation, y ≈ 0.2 associated with the activation of the Li₂MnO₃ phase (Johnson et al., 2004) and x> 0 accounting for the oxygen loss. This type of material typically loses around 6% oxygen after the first cycle (Tran et al., 2008). Given the gradual structural changes, transition metal leaching, and electrolyte decomposition at potentials above 4.3 V versus Li⁺/Li (Tasaki et al., 2009, Guéguen et al., 2016), fewer lithium ions are reversibly exchanged upon further cycling, giving y ≈ 0.7 for long-term cycled particles. Instead of observing these established phenomena directly, this study aims to determine their effects on the morphological and crystalline structural maps.

Scanning electron micrographs in Figure 1 provide a general picture of the secondary particles. Figure 1A shows the pristine particles; Figure 1B shows the particles once delithiated, i.e., after activation; and Figure 1C shows the particles after long-term cycling, i.e., after 85 charge-discharge cycles at a rate of C/10. These secondary particles are 5–20 μm in diameter. Pristine particles exhibit a relatively spherical and regular surface. Among cycled particles, some particles showed faint cracks on the surface and some revealed the internal onion-like structure composed of alternating pores and primary particles. The electrochemical performance shown in Figure 1D revealed a roughly 20% loss of specific charge after the first cycle and a further decrease after long-term cycling. Figure S1 shows the first cycle activation of the NMC particles. The performance decay implies particle degradation; however, many aged particles in Figures 1B and 1C appeared to be intact when observed through scanning electron microscope. With complementary tomographic methods, the interior 3D structures can be revealed to offer better understanding and control of the material.

Figure 1.

Surface Geometry and Electrochemical Performance of Li-rich NMC

(A–C) Scanning electron micrographs of (A) pristine NMC particles, (B) after first delithiation, and (C) after long-term cycling. Primary particles consist of 50- to 100-nm grains that aggregate to form the secondary particles with diameters of around 5 to 20 μm. A crack on the surface is highlighted in yellow in (B), although the spatial extent of the crack is unknown. Onion-like layered structures are observed from broken particles in (C).

(D) The evolution of the specific charge upon cycling at C/10 after the first cycle activation at C/20. The dashed line gives the specific charge obtained after the first cycle.

See also Figure S1 for the first cycle activation.

Three-Dimensional Electron Density Maps by Ptychographic X-Ray Computed Nanotomography

For the series of PXCT and SXDM measurements, the Li-rich NMC particles were transferred into tapered glass capillaries mounted on tomography pins. Pristine, delithiated, and cycled particles were prepared and handled in inert atmosphere, and the capillaries were sealed airtight. Pytchography is a high-resolution lensless coherent diffractive imaging method (Rodenburg et al., 2007, Faulkner and Rodenburg, 2004), which, when combined with tomography, offers quantitative electron density and absorption tomograms with nanoscopic resolution when incident X-ray energies are away from the absorption edge (Dierolf et al., 2010, Holler et al., 2014, Holler et al., 2017). PXCT experiments (Holler et al., 2012) were performed at the cSAXS beamline (X12SA) of the Swiss Light Source, Paul Scherrer Institut, Switzerland. Experimental details are provided in the Transparent Methods. Multiple high-resolution ptychographic tomograms were collected to account for the varying secondary particle size. In total, 7 tomograms of pristine, 5 tomograms of delithiated, and 12 tomograms of long-term cycled particles were acquired. An overview of the acquired PXCT tomograms is given in Figure S2, where the aforementioned onion-like structure is clearly visible. Spatial resolutions were determined by Fourier shell correlation (Van Heel and Schatz, 2005) and range from 35 to 70 nm.

Orthoslices and volume renderings of representative electron density tomograms obtained through PXCT are shown in Figure 2. The measured electron density (Diaz et al., 2012), n_e, can be related to the material composition through $n_e=\left(\rho Z{N}_{\text{A}}\right)/A$, where ρ is the mass density, Z is the number of electrons of the average molecule in the voxel, N_A is the Avogadro's number, and A is the molar mass. The expected theoretical value of the NMC is roughly 1 $e/{\text{Å}}^3$ for ρ = 3.6 g/cm³. Owing to the partial volume effects, i.e., the intermixture of unresolved pores with NMC material, measured values that are lower than the theoretical intrinsically indicate the porosity.

Figure 2.

Ptychographic X-ray Computed Tomography (PXCT) for Li-rich NMC Secondary Particles

(A–E) Volume rendering of (A) pristine, (B) delithiated, and (C–E) cycled particles. Cracks originated from the particle cores can be seen in (B), (D), and (E).

(F–H) Exterior of cracked particles. Surface cracks are seen on particles in (F) and (G) while particle in (H) appears to be intact from the surface. Common to all subfigures is a linear color scale ranging from 0.5 to 1.1 $e/{\text{Å}}^3$ for the electron density.

(I–M) Cross-sections of the particles. The lower left half of each image shows the pore diameter maps given in color scale ranging from 1 to 150 nm. The upper right parts give the electron density in grayscale. Pristine particle in (I) has in general smaller pores, seen as dark blue, compared to the larger pores in particles (J-M), seen as teal or orange.

See also Figure S2 for high-resolution tomograms of all particles and Figure S3 for delithiated particles imaged with near-field ptychography.

Pristine particles display alternating concentric layers of rather uniform thickness and uniform electron density from the core to the edge of the particle, as observed in Figure 2A. After the first charge, i.e., on delithiated particles, we observe an overall widening of the pores as well as the first signs of crack formation, as shown in Figure 2B. These internal morphological changes continue upon prolonged cycling with particles decreasing in electron density, cracking, or fracturing entirely, as observed in Figures 2C–2E. A video of the 3D rendering for the particle in Figure 2D is given in Video S1. Particle exteriors are given in grayscale electron density in Figures 2F–2H with cracks highlighted. Their corresponding volume renderings on the first row show that major fracturing begins at the core and in some cases extends to the surface. We have complemented our observation in regard to morphology change using lower resolution measurements with near-field ptychography on the same ptychography setup with a modified reconstruction algorithm (Odstrčil et al., 2018), showing several particles with cracks from the core, as shown in Figure S3.

Figures 2I–2M show the cross-sections of each particle, giving the pore networks obtained through image segmentation (see Transparent Methods). For the upper right half of each subfigure, the electron density is shown as a reference to show the accuracy of the segmentation. The pore size distribution ranges from a few nanometers to around 200 nm, calculated based on the segmentation. The corresponding histograms of this distribution are given in Figure 3A, showing increases in both the number of pores and the pore diameter upon aging.

Figure 3.

Pore Analysis for PXCT Tomograms

(A) The pore diameter distributions for particles shown in Figure 2, showing an increase in pore size and number upon activation and aging.

(B) Electron density with respect to normalized particle radius, reflecting the porosity distribution. Each curve was obtained by averaging over all particles in the category and the error bars indicate half of the standard deviation.

Alternatively, we also use a direct measure for the porosity without image segmentation as a way to circumvent the influence of partial volume effects and possible inaccuracy in segmentation by calculating the radially averaged electron density to compare particles, as shown in Figure 3B. The line plots were fitted quadratically and averaged over several tomograms with error bars indicating half of the standard deviation. Pristine particles display a relatively constant density from the surface to the core of around 0.95 $e/{\text{Å}}^3$, which approximates the expected Li-rich NMC electron density and thus indicates a low porosity. Delithiated particles possess an equally homogeneous electron density distribution at a lower averaged electron density of roughly 0.85 $e/{\text{Å}}^3$. Long-term cycled particles also show a lower averaged electron density compared with the pristine particle and, moreover, suggest a size dependency.

We categorize large secondary particles as those around or greater than 15 μm in diameter, and the rest are categorized as small secondary particles. Based on our collected data, we suggest that the extent of the damage is particle size dependent. As seen in Figures 2C–2E, large particles seem to be less affected by fracture mechanisms than the small particles. Moreover, an increase in pore frequency and size, i.e., material depletion is noticeable in the direction of the particle core for larger particles. This is shown quantitatively by the light violet curve in Figure 3B. For smaller cracked particles a higher porosity is observed closer to the surface, shown by the black curve in Figure 3B. The dashed white lines in Figures 2C and 2E highlight the boundaries at which the electron density or porosity changes. The size dependency we observed here offers clues for the degradation mechanism.

The surface of secondary/primary particles with short diffusion pathways, such as those close to pores or diffusion highways, reacts first with the lithium ions during lithiation/delithiation, while the reaction front progresses toward the center of the secondary particle. Assuming the obedience of such surface diffusion model and that all secondary particles have a uniform porosity at their pristine stage regardless of diameter, repeated volume expansion/contraction of the primary particles can cause a strain differential inside the secondary particle whose magnitude is governed by, for example, the secondary particle diameter, material chemical composition and porosity, and cycling conditions. The strain differential due to expansion results in a tendency for a spherical particle to crack from the core to the surface once a critical strain value is surpassed (Kondrakov et al., 2017, Sun and Manthiram, 2017, Ryu et al., 2018, Ko et al., 2019, Christensen and Newman, 2006). This can lead to the observed difference in cracking behavior for different sizes of particles. The larger particles retain their structural integrity, although they display higher porosity at the core due to their size and longer diffusion pathways when compared with the smaller particles at a set cycling rate. Smaller secondary particles are subjected to an increased level of surface damage as they are cycled faster to completion and to a deeper extent because of the higher surface-area-to-volume ratio and shorter diffusion paths. Cracks also create more channels for electrolyte infiltration and interfaces for electrolyte decomposition. With this cracking mechanism in mind, we turned to X-ray diffraction microscopy to examine the crystallographic structure of the material and its correlation to the observed morphology.

Structural and Chemical Distribution within Particles at a 1 μm Resolution

We performed a series of SXDM measurements combined with XRF spectroscopy measurements on the same particles to infer how and if these morphological changes observed from PXCT results relate to changes in chemical composition and thus different extent of lithiation. Experiments were performed at the microXAS beamline of the Swiss Light Source, Paul Scherrer Institut, Switzerland. Through XRF tomography, a homogeneous distribution of Mn, Co, and Ni is observed for these same particles, as shown in Figure S4. However, SXDM measurements reveal unexpected findings.

Typically X-ray diffraction measurements are performed on bulk particles to obtain averaged X-ray diffractograms, e.g., Figure S5, whereas here we scanned samples with a 1-μm beam to probe each Li-rich NMC secondary particle individually. Diffraction maps highlighting the normalized integrated intensity of the selected (101), (104), and (003) Bragg peaks are given in Figure 4A in red, green, and blue, respectively. These selected peaks reflect the interlayer spacing where the Li ions are located, and they shift upon lithiation and delithiation (Mohanty et al., 2013). For visualization, the ratio of the intensity of these Bragg reflections is given in the form of a red-green-blue color map, in which each peak intensity is normalized with respect to its maximum in the projection. To compare between the particles at different stages, the dashed lines in insets in Figure 4A (A2 and A3) show the intensities normalized with respect to the corresponding peaks in the pristine projection. The reduced brightness from Figure 4A (A1) to the insets in Figure 4A (A2 and A3) reflects a reduced crystallinity upon cycling. Particles regardless of age and diameter possess an intriguing core-shell structure. A dominant (003) peak in blue is observed on the outermost shell, dominant (104) peak in green is seen on the inner shell, and dominant (101) peak in red is seen at the core. We attribute this core-shell crystal structure to a disorder in the structure and potentially lithium gradient from the shell to the core. With delithiation and prolonged cycling, the spatial extent of the core slightly increases and, in the case of some cracked particle, the shell is disappearing.

Figure 4.

Scanning X-ray Diffraction Microscopy (SXDM) for Secondary NMC Particles

(A) X-ray diffraction intensity maps of (A1) the pristine, (A2) delithiated particles, and (A3) particles after long-term cycling. Each projection or tomogram gives the normalized intensity with the (101) Bragg reflection distribution map shown in red, (104) in green, and (003) in blue. Dashed lines in insets show the normalization with respect to pristine in (A1), showing a reduced crystallinity upon cycling.

(B) Ratio of the FWHM (003) peak over the FWHM (104) peak, showing a higher anisotropic disorder in some regions.

(C) Unit cell volume calculated based on the rhombohedral $\mathrm{R}\overline{3}\text{m}$ model. Insets given with a different color scale show slight variation between particles within a projection. Comparison between (C1) and (C3) shows obvious increase of the unit cell volume for some particles. The pixel size is roughly 1 μm. The corresponding elemental distribution of Mn, Co, and Ni obtained from the XRF spectra is shown in Figure S4.

The ratio of the (003) and (104) reflects the cation intermixing and related structural disorder (Sasakawa et al., 2015, Fujii et al., 2007). Figure 4B (B1 to B3) gives the ratio of the full-width-half-maximum (FWHM) of the (003) peak over the FWHM of the (104) peak, showing in some regions a higher disorder, potentially attributed to less amount of Li ions in the lattice.

Although pristine NMC has a combination of rhombohedral and monoclinic phases, the rhombohedral model gives a relatively better fit to the data here compared with other models. The extrapolated interpretation presented here is an attempt to explain the surprising core-shell structure, inviting further investigation. Figure 4C (C1 to C3) gives the unit cell volume maps assuming the rhombohedral $\mathrm{R}\overline{3}\text{m}$ model, showing a volume expansion of 5% to 10%, compatible with the usually reported volumetric expansion upon cycling (Ates et al., 2013, Mohanty et al., 2013). Correlating this result with PXCT shows a relation between the extent of cracking with the degree of unit cell expansion. Multiple cracks near the core of L3 in Figure 2E correlate with reduced electron density and the most dramatic change in the crystal structure with over 10% volume expansion. Particles with a single crack extended to the surface show a lesser degree of changes. This type of crack formation leads to new exposed surface area that generates parasitic reactions that could possibly slow the aging of the crystal structure as some charge is consumed at each cycle to passivate the newly created surfaces.

Average peaks of projections in each state are given in Figure 5A, where prominent peaks bear high similarity with the X-ray diffractogram of bulk particles in Figure S5, giving validity to our measurements. Figure 5A also shows the shifting and broadening of the (003) peak, as expected for aged Li-rich NCM (Tran et al., 2006, Croy et al., 2013, Mohanty et al., 2013). The shifting is noted by the black arrow, and the broadening, by the gray line in the inset. Shown in Figure 5B is a comparison between typical single 0.5-μm-voxel powder diffraction patterns of particles L2 and L3, from their core and shell regions. A lower crystallinity is observed for the particle L3 in Figure 5B, and higher disorder along the c crystalline direction is observed in the core region for both these particles. This surprising core-shell structure deserves further investigation and could offer a different perspective for future studies.

Figure 5.

X-ray Diffraction Peaks in Projections and Voxels

(A) Averaged peaks for projections, showing shifting and broadening of the (003) peak upon aging, noted by the black arrow and the gray line, respectively.

(B) Typical single voxel (0.5 × 0.5 × 0.5 μm³) powder diffraction patterns from the shell, i.e., near the particle surface, and the core of particles L2 and L3 retrieved through SXDM tomography. Typical X-ray diffractograms of bulk pristine and cycled NMC are given in Figure S5.

Conclusions

Our examination of Li-rich NMC secondary particles revealed a series of morphological degradation processes on the secondary particle level that could complement current understandings of battery failure mechanisms. Acquired electron density tomograms of the spherical NMC particles with nanoscale resolution (35–70 nm) through PXCT revealed that the uniform onion-like layer-by-layer structure, common to most particles, changes upon going through a size-dependent aging mechanism. After long-term cycling, we observed that particles greater than 15 μm in diameter revealed an enhanced material depletion near the core, whereas smaller particles tend to be subjected to crack formation and depletion close to the surface. This size dependency gives insight into the aging mechanism and also draws attention to its possible influence on the performance. To better understand the degradation mechanism, we turned to SXDM with a 1 μm resolution to explore the relation between morphological and crystallographic change. A correlation between the spatial extent of the cracks and degree of unit cell volume expansion was observed. The SXDM results also present an intriguing spatially dependent crystal structure, seen as core shell, whose extent changes upon aging. This core-shell structure is inferred from the intensity ratio of selected Bragg peaks. Aspects of the SXDM result that are different from typical observations on bulk NMC material could give a different perspective while inviting for further studies. Complementary to studies on the primary particles or on a bulk level, this work reveals in 3D the cathode material on a secondary particle level with regard to both its morphology and crystalline structure, a connection that could lead to a thorough fundamental understanding of the aging mechanism as well as present a method for studying battery materials in general.

Limitations of the Study

Particles were extracted from cells after cycling for imaging in this work, with resolution limited by the sample size and the experimental time. With ongoing instrumentation advances, future studies will seek fast high-resolution imaging of cells operando.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: January 25, 2019