Report

Visualization and Quantification of Electrochemical and Mechanical Degradation in Li Ion Batteries

See allHide authors and affiliations

Science  08 Nov 2013:
Vol. 342, Issue 6159, pp. 716-720
DOI: 10.1126/science.1241882

Battery Breakdown

Although a range of materials can be used for chemically storing electrical charge, many cannot be made into batteries that retain their capacity over many cycles. Failure may be because of secondary reactions, poisoning through the formation of surface coatings, or volumetric changes leading to fracture. Ebner et al. (p. 716, published online 17 October) studied this last scenario in an operating battery using synchrotron x-ray tomographic microscopy, tracking both the chemical changes in the battery and the resulting mechanical changes in a tin oxide model system, which is known to undergo large volume changes.

Abstract

High–energy-density materials that undergo conversion and/or alloying reactions hold promise for next-generation lithium (Li) ion batteries. However, these materials experience substantial volume change during electrochemical operation, which causes mechanical fracture of the material and structural disintegration of the electrode, leading to capacity loss. In this work, we use x-ray tomography during battery operation to visualize and quantify the origins and evolution of electrochemical and mechanical degradation. Tomography provides the time-resolved, three-dimensional chemical composition and morphology within individual particles and throughout the electrode. In the model material tin(II) oxide, we witness distributions in onset and rate of core-shell lithiation, crack initiation and growth along preexisting defects, and irreversible distortion of the electrode, highlighting tomography as a tool to guide the development of durable materials and strain-tolerant electrodes.

Volume change of the active material during battery operation is the primary cause of short life in lithium ion batteries containing high–energy-density materials that undergo conversion and/or alloying reactions (1, 2). This class of materials—which includes pure metals, their alloys, oxides, fluorides, sulfates, nitrides, phosphates, and hydrides—offers up to an order of magnitude increase in specific capacity compared with the intercalation compounds that are used in commercial lithium ion batteries today (1). This potential has spurred considerable activities in material synthesis (16) and characterization (712), as well as theoretical studies and modeling efforts (1316). However, due to the complexity of the interactions between electrochemistry and mechanical deformation (1719), no comprehensive framework for rational material design exists to date.

An experimental technique providing quantitative chemical and morphological insight with three-dimensional (3D) spatial and temporal resolution that could advance the development of such a framework is needed. In situ x-ray diffraction (XRD) and Mössbauer spectroscopy quantify crystal structure evolution, whereas acoustic emission and dilatometry can be used to monitor fracture and expansion of entire electrodes (2022). However, none of these approaches is able to directly relate the observed behavior of the ensemble to effects at the single-particle scale. On the other hand, techniques that enable visualization of single particles during electrochemical cycling—such as in situ transmission electron microscopy, scanning electron microscopy (SEM), and atomic force microscopy—are surface-sensitive techniques that cannot probe into the bulk of realistic porous battery electrodes (7, 8, 23, 24). X-ray transmission microscopy has enabled observation of active particles, including Sn and SnO (911), as well as chemical phase mapping when used in combination with x-ray absorption near-edge structure spectroscopy (12). However, during electrochemical cycling, the 2D nature of these studies did not allow for quantitative assessment of particle phase evolution, volume expansion, or fracture.

Here, we show that synchrotron radiation x-ray tomographic microscopy (SRXTM) is a tool to simultaneously visualize and quantify the factors that affect battery performance. In SRXTM, monochromatic x-rays are directed onto the sample under investigation, which is rotated through 180° (Fig. 1A). 3D microstructure representations are then calculated numerically from recorded projection images by a tomographic reconstruction algorithm (25, 26). Previously, ex situ SRXTM and x-ray nanotomography have been used to characterize porous lithium ion battery electrodes (27, 28). We fabricated an electrochemical cell, compatible with SRXTM, to study electrodes during battery operation (Fig. 1B and fig. S1F) in a realistic environment (29). The cell is designed to be x-ray–transparent, to offer mechanical stability against vibrations during acquisition (to prevent reconstruction artifacts), and to provide good sealing and light pressure on the electrode for stable electrochemistry.

Fig. 1 X-ray tomography.

(A) Sketch and (B) photograph of the x-ray tomography setup. CMOS, complementary metal-oxide semiconductor. (C) Scanning electron micrograph of SnO particles. (D) 3D visualization of x-ray tomograms recorded during battery operation. (E) Unprocessed cross-sectional tomogram showing individual SnO particles in the electrode with high resolution and good contrast against a low-attenuating carbon black, binder, and electrolyte phase. Att. coeff, attenuation coefficient. (F) A series of cross sections through two particles demonstrates a core-shell process, volume expansion, and particle fracture during the initial reduction and particle redensification during subsequent oxidation. mAhg−1, milliampere hours per gram.

To demonstrate the complex interplay between electrochemistry and mechanical degradation in batteries, we opt to study tin(II) oxide (SnO) as a representative material that undergoes a conversion reaction and subsequent alloying with lithium (20, 3032). We select a SnO synthesis route that offers size and shape control to create uniform particles, as shown in the SEM image in Fig. 1C (33). A porous electrode is fabricated from the SnO particles, carbon black, and polymeric binder and is inserted into the electrochemical cell.

Tomograms are recorded every 15 min during galvanostatic reduction (lithiation) at 110 mA·hour g−1 over a period of 12 hours and during oxidation (delithiation) at 167 mA·hour g−1 over 5 hours. Throughout the experiment, microstructural data of the entire electrode (diameter of 1.6 mm, initial thickness of 50 μm) are collected. The full electrode fits into the field of view to allow tracking of all particles during battery operation. To highlight the 3D nature of the collected data, we provide a rendering of a subsection of the electrode (Fig. 1D) corresponding to a volume of 333 by 33 by 50 μm3, or 1/25th of a full 3D data set. The voxel size is (0.65 μm)3, providing a true resolution of 2.0 μm, as determined from step profiles. Particles can be identified due to differences in x-ray attenuation coefficients (μ). A tomographic reconstruction without any image processing of an electrode cross section, recorded before electrochemical reduction, is shown in false-color in Fig. 1E. Yellow-red cs correspond to regions of high attenuation coefficients and can be associated with SnO particles, whereas the blue region corresponds to the minimally attenuating carbon black, polymer, and electrolyte phase that fills the space between the SnO particles. To obtain absolute values for the x-ray attenuation coefficient, a normalization procedure based on SnO and the background as internal standards is implemented, as discussed in the supplementary text.

Visualization and quantification of phase transformations within individual particles are possible due to changes in the attenuation coefficient, which is intimately linked to the composition and mass density of a material. In Fig. 1F, 12 snapshots selected from the 74 tomograms track two particles through their reactions. During reduction, we can qualitatively observe two consecutive processes consistent with the conversion reaction of SnO, in which nanosized Sn clusters form in a growing amorphous Li2O (lithia) matrix, and the subsequent alloying reaction of the Sn clusters with lithium to form LixSn with 0 ≤ x ≤ 4.4 (20, 3032). The Sn cluster alloying reaction is expected to traverse the equilibrium crystal phases present in the Li-Sn phase diagram at room temperature: Sn, Li0.4Sn, LiSn, Li2.33Sn, Li2.5Sn, Li2.6Sn, Li3.5Sn, and finally, Li4.4Sn. In Fig. 1F, we first observe a low-attenuating (yellow-green) front appear and progressively penetrate the particles, consuming the highly attenuating (yellow-red) phase. This core-shell process is the conversion reaction. Subsequently, the low-attenuating phase homogenously transitions to an even lower-attenuating phase (dark green). This is the alloying reaction. During oxidation, we observe a homogeneous increase in attenuation coefficient from dark green to green-yellow. This process is consistent with the dealloying of the LixSn clusters in the lithia matrix. Accompanying the conversion and alloying reactions, we observe expansion of the particles as well as crack initiation and growth; during dealloying, we observe redensification of the particles. Because the voxel size is larger than the size and spacing of the LixSn clusters (both < 10 nm) in lithia (32), the attenuation coefficient obtained by our technique represents a weighted average over the attenuation coefficients of the LixSn clusters and lithia matrix.

As indicated by the two particles shown in Fig. 1F and, more generally, in movie S1, particles exhibit a distribution in onset, rate, and completion of the core-shell reaction. To quantify the extent of the conversion reaction and enable direct comparison to electrochemical data, analysis must be performed considering a 3D electrode volume.

The chemical composition of the active particles within the electrode can be quantified during cell operation by analyzing the distribution of the x-ray attenuation coefficients. The leftmost peak in the attenuation coefficient histograms (μ = 0.4 cm−1) from a central electrode subvolume (Fig. 2, A and B) corresponds to the carbon black, binder, and electrolyte phase, whereas the rightmost peak corresponds to the SnO particles (μ = 45 cm−1). During electrochemical reduction (Fig. 2A), the SnO peak decreases in magnitude and vanishes after ~500 mA·hour g−1. Simultaneously, a center peak associated with the formation of a lithia matrix studded with Sn clusters emerges at μ = 25 cm−1 and then progressively shifts to 14 cm−1 as it increases in magnitude. Careful inspection of the SnO peak (close-up in fig. S2B) shows that it slowly shifts from 45 to 50 cm−1, which can be explained by a disproportionation reaction of SnO to SnO2 and Sn (μ = 49 cm−1) (20). The shift toward higher attenuation coefficients can be seen visually in Fig. 1F and movie S1 as a shift to a deeper red color. This SnO2 also undergoes a conversion reaction to form LixSn clusters in lithia that alloy to Li4.4Sn by the end of reduction (30).

Fig. 2 Evolution of chemical composition.

X-ray attenuation coefficient histograms during electrochemical (A) reduction and (B) oxidation. Color indicates temporal evolution. Q, capacity. (C) Attenuation coefficient distributions during electrochemical reduction and oxidation. Horizontal dotted white lines indicate theoretical attenuation coefficients for end members of the phase evolution.

During subsequent oxidation (Fig. 2B), the 14 cm−1 peak shifts back toward 30 cm−1 and decreases in magnitude, consistent with the dealloying of LixSn and the contraction of the particles. The variation of the attenuation coefficient between the theoretical values for Sn in Li2O (μ = 30 cm−1) and Li4.4Sn in Li2O (μ = 13 cm−1) highlights the permanent existence of LixSn clusters within a lithia matrix. The temporal aspect of the phase evolution can be clearly seen by plotting the histograms from Fig. 2, A and B, as a density plot as a function of capacity, Q, in Fig. 2C. The disappearance of the SnO phase and emergence of the Li2O + Sn phase are clearly visible. We find a linear relation between the attenuation coefficient of the mixed phase and capacity, indicating a steady-state reaction. This bulk analysis sets the visual inspection of individual particles into perspective. For example, we conclude that the particle in the upper left of Fig. 1F lags behind the conversion of the ensemble, as it exhibits an unconverted core (red) at 729 mA·hour g−1. The particle in the lower right is more representative of the ensemble behavior.

As observed in Fig. 1F, the conversion and subsequent alloying reaction and the associated volume change are accompanied by crack formation and propagation in the active particles. Our choice of SnO enables us to identify the crystallographic orientation of particles in the tomography data and relate this to the observed cracking. XRD data and SEM analyses (provided in fig. S1) reveal pure SnO with tetragonal crystal structure and tetragonal particle shape. This coincidence of particle shape and crystal structure permits identification of the crystallographic orientation of tetragonal particles in the tomograms (33): A square face is associated with the (001) plane, whereas a rectangular face corresponds to the (100) or (010) plane. Figure 3, A and B, and movie S2, showing transverse and coronal cross sections through a particle during electrochemical reduction, reveal preferential cracking in the (001) plane originating at multiple sites along the [001] axis. We link these crack sites to grain boundaries that occur preferentially in (001) planes and are observable in the SEM images (fig. S1, A and B).

Fig. 3 Evolution of particle fracture.

(A) Coronal and (B) transverse cross sections through a particle during electrochemical reduction. Horizontal dotted white lines in the cross sections at 0 min indicate cutting planes. Crystallographic directions are identified and indicated. White arrows point to crack locations. (C) Schematic of particle phase evolution and crack growth leading to zig-zag morphology. (D) 3D rendering of subvolume visualizing zig-zag morphology in multiple particles. Black arrows indicate fracture.

As highlighted by the white arrows in Fig. 3A, cracks appear sequentially at opposite edges of the particle. As observed by the absorption coefficient contrast in Fig. 3A and as indicated schematically in Fig. 3C, these cracks propagate inward, exposing fresh SnO surfaces to the electrolyte; these surfaces immediately undergo the conversion reaction. Volume expansion of this freshly exposed SnO contributes to the opening of the crack. This is highlighted by Fig. 3B, which shows that the SnO conversion in the crack plane occurs as a front diagonally traversing the particle. We speculate that as one crack is initiated and propagates, the stress in the vicinity of this crack is relaxed. Continuing reduction of the particle causes a buildup of stress at the opposite side of the particle, where propagation of the first crack does not alleviate the stress. Because cracks occur preferentially at grain boundaries, a new crack therefore initializes at the opposite side of the particle along a parallel grain boundary. These collective effects lead to the sequential delamination of (001) planes in the observed zig-zag geometry. The 3D renderings in Fig. 3D and movies S3 and S4 highlight the fact that this zig-zag delamination occurs in multiple particles. This observation that fracture is linked to crystallographic defects is consistent with the report of no fracture in defect-free SnO2 nanowires (7). Studies of amorphous materials (which, by definition, do not have grains and grain boundaries) or nanostructured alloys would offer the opportunity to complete the picture of fracture as a function of material structure.

Synchrotron radiation x-ray tomographic microscopy further enables quantification of the volume change of the active particles in the entire electrode, a radiograph of which is shown in Fig. 4A. Details of the data processing are described in the supplementary text. Figure 4B shows the volume expansion calculated from our tomography data (gray circles) and the cell potential (purple line) as a function of capacity. During reduction, the calculated volume expansion coincides with that expected from the equilibrium LixSn phases in lithia (blue squares). In particular, at the end of electrochemical reduction, a volume expansion of 258% is found, in good agreement with theoretical predictions of 252% for Li2O + Li4.4Sn with respect to SnO.

Fig. 4 Volume expansion.

(A) X-ray radiograph of the full electrode highlights the accessible volume, which includes the entire electrode as well as the separator and current collector. (B) Quantitative volume expansion of the active materials (gray circles) in the entire electrode compared with equilibrium phases (blue squares) and cell voltage (purple line) during electrochemical reduction and oxidation. (C) Snapshots of radiographs from the electrode center show qualitative electrode expansion. White lines denote the top edge of the electrode. (D) Volume fraction of active particles in the electrode as a function of the distance from the current collector (curr. coll.). Coloring indicates temporal evolution. Dotted lines indicate the average volume fraction and associated electrode thickness before reduction (circles), after reduction (triangles), and after oxidation (squares). Red symbols indicate thickness; blue symbols represent volume fraction. Following the distance from the current collector at a fixed volume fraction (vertical dashed line at 0.1 volume fraction) quantifies (E) the change in thickness of the electrode. (F) Average volume fraction of active particles in the electrode as a function of capacity.

As highlighted in movie S4, the volume expansion of the active particles drives expansion of the entire electrode, which can lead to mechanical breakdown of the polymeric binder and carbon black matrix (34). In Fig. 4C, snapshots of radiographs of a central section of the electrode are depicted during reduction and oxidation. The top edge of the electrode is indicated by a white line, and tracing its temporal evolution shows increasing electrode thickness during electrochemical reduction and decreasing thickness upon oxidation. To quantify the relation between the volume change of the active particles and thickness change in the electrodes, we calculate the volume fraction of active particles in the direction normal to the current collector (Fig. 4D) from slices 1 voxel high and 512 by 512 voxels in area. Plotting the maximum distance from the current collector at a specific volume fraction allows us to quantify the thickness change (Fig. 4E). We compare this to the temporal evolution of the average volume fraction of the active particles (Fig. 4F). The freshly prepared electrode measures 50 μm, and the average active particle volume fraction is 0.24. During electrochemical reduction, the electrode expands to more than 120 μm. At first, this electrode expansion occurs concomitantly with particle volume expansion, as evidenced by the roughly constant volume fraction. However, after the electrode thickness has doubled, even though the total electrode thickness continues to grow, the particles begin to occupy an increasingly large volume fraction of the electrode, as demonstrated by the rise in active material volume fraction in the electrode. This behavior is indicative of particle volume expansion beyond what can be accommodated by the polymer binder and carbon black matrix. After oxidation, the electrode contracts to a thickness of 80 μm, but the particle volume fraction decreases to 0.21, a value below the starting volume fraction, which implies that the polymer binder and carbon black matrix are permanently distorted in the first reduction step. This distortion of the conductive matrix, together with particle fracture, is known to electrically disconnect particles from the rest of the electrode, leading to capacity loss (1, 2).

To test whether our tomography data show evidence of electrically disconnected particles, we carefully inspect the attenuation coefficient histograms in fig. S3A. We observe broadening of the Li2O + LixSn peak during oxidation, which we quantify in fig. S3B by plotting the full width at half maximum. Because the attenuation coefficient is related to the lithium content (x in Li2O + LixSn), we attribute the spread in attenuation coefficient to a spread in lithium content across different particles and interpret this as a sign for electrical disconnection of individual particles from the rest of the electrode.

Our electrochemical data are consistent with this interpretation. For SnO, it is known that below 1.0 V, dealloying of LixSn is the only source of capacity. Above 1.5 V (shaded region in Fig. 4B), decomposition of Li2O and reformation of Sn-O contacts, leading to unknown and possibly amorphous or disordered phases, have been reported (35). In our cell, the extracted capacity during oxidation below 1.5 V is ~550 mA·hour g−1, ~37% below the theoretical value of 873 mA·hour g−1.

Finally, we test if the identified spread in lithium content identified in the tomography data can account for the electrochemically determined capacity loss. We estimate the state of charge, parameterized by the average lithium content Embedded Image, by numerically integrating over the Li2O + LixSn attenuation coefficient peak, as discussed in the supplementary text. We find that the estimated lithium content is Embedded Image at a cell voltage of 1.5 V, which corresponds to a lost capacity of ~43%, in good agreement with the 37% determined electrochemically. This finding demonstrates that SRXTM can track and quantify cause and effect of electrochemical and mechanical processes and enables a self-consistent description of battery degradation mechanisms.

The experiment presented here can be repeated for a number of anode and cathode materials to develop a comprehensive framework comprising all electrochemical and mechanical aspects of conversion and alloying materials. For example, this framework could resolve open questions associated with the lithiation of silicon, a potential anode material. Highly anisotropic lithiation has been observed in crystalline silicon micro- and nanostructures (18, 19), although lithium diffusion in silicon is expected to be isotropic (13). Current explanations include anisotropic volume expansion and unequal reaction kinetics for different crystal planes (13). Electrochemical amorphization and the sudden appearance of crystalline Li15Si4 from these amorphous phases (36), stress-dependent potential of lithiated silicon (17), and a rate-independent polarization hysteresis (16) add to a complicated picture. However, the type of quantitative, 3D, and time-resolved images of particle lithiation introduced in this work will provide the experimental data necessary to comprehend the complex electrochemical and mechanical interactions in silicon and related materials. The development of chemistries and particle morphologies that hinder crack formation or growth, as well as expansion-tolerating composite electrode architectures, may enable better batteries.

Supplementary Materials

www.sciencemag.org/content/342/6159/716/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S3

Table S1

References (37, 38)

Movies S1 to S4

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: The tomography experiments were performed on the Tomographic Microscopy and Coherent Radiology Experiments (TOMCAT) beamline at the Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. We thank P. Modregger, L. Nowack, and M.-F. Lagadec for their support during the beamtime; K. Kunze and W. Woodford for insightful discussion; D. Norris for access to SEM; and O. Waser and S. Pratsinis for access to XRD. We gratefully acknowledge material donations from TIMCAL and Arkema.
View Abstract

Navigate This Article