Lithium Metal-Copper Vanadium Oxide Battery with a Block Copolymer Electrolyte

Electrolyte preparation

Synthesis of the α-CuVO₃ material

Electrode preparation and battery assembly

Battery cycling

After assembly inside the glove box, battery replicates were transferred into a 90°C oven connected to a Maccor cycler where they were heated overnight prior to any cycling. The battery cycling rate is reported in terms of C/n where n is the number of hours used to fully charge or discharge the battery during the first cycle. Five conditioning charge-discharge cycles were performed at an approximate rate of C/20 for all the batteries. Batteries were then cycled at the C-rate of interest. All potentials reported in this article are relative to the Li⁺/Li° couple. Charging was done using a constant current (I) until the battery potential (E) reached 3.6 V because the electrochemical window stability of the SEO electrolyte is about 4 V. The constant current step was followed by a constant voltage step at 3.6 V and this step was stopped when the current was 0.1 × I. After a 1 h rest period, a discharge current identical to the charge current was applied until the potential reached 2 V, followed by a 1 hour rest period. Batteries were examined at C/50, C/20, C/4, and C/3. For the batteries cycled at C/20 a second C/20 discharge step was executed after the first discharge step. This final discharge step was conducted to distinguish between capacity fade due to transport limitations within the α-CuVO₃ electrode and that due to irreversible changes in the battery. The capacity related to this final discharge step was negligible (1.3 ± 0.4% of the total discharge capacity). For batteries cycled at C/50, a constant voltage step at 2 V was added after the constant current discharge step until the current returned to 0.1 × I to see any discharge capacity gain. The capacity related to this constant voltage discharge step was negligible (2.5 ± 1.1% of the total discharge capacity). A typical charge-discharge profile is shown in Figure 1 where battery potential (E) obtained during the charging and discharging steps, and rest steps is plotted versus time (t) for one C/20 cycle. For completeness E as a function t for a C/50 and a C/3 cycle is given in Figure S1.

Zoom In Zoom Out Reset image size

For each cycle, the charge (Q_c) and discharge (Q_d) capacities normalized by the estimated mass (g) of α-CuVO₃ in the positive electrode (using the measured mass of each electrode disks and assuming constant composition) was calculated by integrating the current over time (mA.h). For consistency, only the first constant current discharge step was used to calculate Q_d for the C/50 and C/20 batteries. The Coulombic efficiency (η) in percent is defined by:

Wide-angle X-ray scattering

The crystal structure of the α-CuVO₃ pristine powder was determined by ex situ wide-angle X-ray scattering (WAXS) at the Advanced Light Source beamline 7.3.3 at the Lawrence Berkeley National Laboratory using a 10 keV monochromatic X-rays.⁴⁷ Inside the glove box, the powder was hand-pressed into a 150 μm thick fiberglass Garolite-10 spacer with a 3.175 mm diameter hole at room temperature to form a dense pellet. The pellet was sealed off in a custom-designed airtight sample holder with Kapton windows and heated overnight at 90°C in the glove box antechamber under vacuum. Another sample without the powder was prepared in the same manner and used as a blank. The evolution of the crystal structure of α-CuVO₃ during a battery cycle was studied by in situ WAXS. The samples were similar to the pouched batteries with the precaution that the battery tabs needed to be attached to the edges of the electrodes to ensure electrical contact without blocking the path of the X-rays. After assembly, four battery replicates were preconditioned at 90°C by using five C/20 cycles and taken to the WAXS instrument in the discharged state.

The samples were placed in a custom heating stage and the battery tabs were connected to a multipotentiostat VMP3 (Bio-Logic SAS). A Dectris Pilatus 2 M detector with a pixel size of 0.172 mm × 0.172 mm was used to acquire WAXS data in transmission mode. The distance between the sample and detector, and the scattering vector coordinate were determined by calibrating the diffraction images with a silver behenate reference.⁴⁷ The pristine α-CuVO₃ material was studied at room temperature and at 90°C. The batteries were studied during one charge-discharge cycle at a current density corresponding to a C/4 rate at 90°C. For all the samples, after a 2 seconds exposure time, a 1D scattering profile was obtained by azimuthally averaging the 2D scattering patterns using the Igor Pro software with the Nika macro.⁴⁸ The scattering profiles are presented as scattering intensity (I_s) subtracted from the blank sample using the Nika macro, as a function of the scattering wave vector magnitude (q) as defined by:

where λ is the wavelength of the X-rays (0.124 nm) and θ is the scattering angle.

Hard X-ray microtomography

Electron microscopy

The ex situ WAXS scattering profiles of the pristine α-CuVO₃ measured at 30°C and 90°C are shown in Figure 2. The scattering profile of the pristine α-CuVO₃ is independent of the temperature between 30°C and 90°C. These two temperatures correspond to the operating temperature of Li battery comprising either a conventional liquid electrolyte (30°C) or a solid polymer electrolyte (90°C). The three peaks between 2.2 and 2.5 Å⁻¹ are the α-CuVO₃ characteristic peaks. These peaks located at q values of 2.22, 2.23 and 2.43 Å⁻¹ can be ascribed to the (131), (122), and (401) planes, respectively.⁹

Zoom In Zoom Out Reset image size

The result of the in situ WAXS experiment is shown in Figure 3 for one representative battery. The battery was charged and discharged at C/4 at 90°C while WAXS patterns were collected at different times. We define the number of Li inserted/de-inserted (x) as the amount of Li in the lithiated phase Li_xCu_6.9V₆O_18.9. Figure 3a shows the battery potential (E) as a function of x. Also shown in Figure 3a are the points at which WAXS spectra were acquired. Those spectra are shown in Figure 3b and the characteristic peaks position of the α-CuVO₃ and those of Cu metal located at q values of 3.02 and 3.49 Å⁻¹ are indicated. The two very intense peaks located at q values of 2.70 and 3.13 Å⁻¹ correspond to the positive electrode Al current collector.

Zoom In Zoom Out Reset image size

In Figure 3a, we see that during the charge step, E increases from the open circuit value of 2.0 V (the battery was initially in discharged state) until a first plateau of about 2.6 V is reached at x = 0.9. E increases slowly until x = 3.6 and then raised rapidly until a second plateau of about 2.8 V is reached. (Both plateaus have measurable slopes.) The second electrochemical process begins at x = 4.5, and reaches x = 8.3. When E reaches the 3.6 V cutoff potential, close to 10 equivalents of Li have been extracted (x = 9.8); in principle electrons may also be released by side-reactions such as oxidation of Cu metal. At the end of the constant voltage step at 3.6 V, the cell reached x = 10.1, corresponding to a charge capacity of 242 mAh/g. During the subsequent discharge E quickly drops down to 2.71 V as x decreases to 9.4, then smoothly decreases along a 2.55 V plateau ending when x is about 2. The remaining capacity is produced while the E drops rapidly to the cutoff potential of 2.0 V, showing the reversibility of the electrochemical cycle. This cycling profile is very similar to the one reported by Kamiya et al. for Cu_6.7V₆O_18.7 in a battery containing a liquid electrolyte.⁹ It is evident that the charge-discharge characteristics obtained with a solid SEO electrolyte are similar to that obtained in a liquid electrolyte.

In Figure 3b, the scattering profile of the battery in the discharged state (x = 0) shows the presence of two Cu peaks, at q values of 3.02 and 3.49 Å⁻¹. During the charge step, the Cu peak intensities decreased overall, then increased again during the discharge process. At the same time, the three α-CuVO₃ characteristic peaks, located in the q interval between 2.2 and 2.5 Å⁻¹, showed the opposite trend. To get a better understanding of the electrochemical processes occurring during this charge-discharge cycle and its reversibility, the area of the reflections corresponding to Cu metal (q = 3.02 and 3.49 Å⁻¹) and α-CuVO₃ (2.22 ≤ q ≤ 2.43 Å⁻¹) were integrated. At a given x value, the Cu peak area was divided by that at x = 0 while the total area under the three α-CuVO₃ characteristic peaks was divided by the total area of those peaks at x = 10.1. Figure 4 represents the area ratio of the Cu peak (A_Cu) at q = 3.02 Å⁻¹ and the total area ratio of the α-CuVO₃ (A_CuVO3) characteristic peaks as a function of x. Similar plots were obtained when the q = 3.49 Å⁻¹ peak area was used. One might expect A_CuVO3 to be zero at x = 0 if the battery was fully discharged. In contrast the data in Figure 4 shows that A_CuVO3 is 0.2 when x = 0. Similarly, one might except A_Cu to be zero at x = 10.1 if the battery was fully charged. The data in Figure 4 show that A_Cu is 0.2 when x = 10.1. We conclude that the five conditioning cycles resulted in small irreversible changes in the crystal structure. During the charge step, A_CuVO3 follows a sigmoidal curve, increasing slowly until x = 4.6 with an average slope of 0.02. This is followed by a rapid increase in A_CuVO3 until the final plateau at x = 10. The average slope in the second region is 0.12. The break between the two regimes at x = 4.6 corresponds to the transition between the two voltage plateaus identified in Figure 3a. The decrease in A_Cu during charge also occurs in two regimes demarcated by x = 4.6. The WAXS data provide evidence for two simultaneous processes during charge: (i) de-insertion of Li from the VO₃ framework, and (ii) insertion of metallic Cu into the vanadate framework. These processes are similar to the observation in liquid electrolytes.⁹ After an initial plateau until x = 8.3, the discharge data in Figure 4 are also differentiated into two additional regimes demarcated by x = 2. Three regimes are also seen in the discharge curve in Figure 3a. The changes in slope in A_CuVO3 and A_Cu (Figure 4) between x = 10.1 and 8.3 corresponds to the step decrease in E (Figure 3a). The data in Figures 3 and 4 indicate that the charge and discharge pathways are not identical. Similar conclusions are contained in reports on other Cu-based active materials.^51,52 The other demarcation point seen in the discharged state in Figure 4 (x = 2) corresponds to the end of the long plateau in Figure 3a. At the end of the charge-discharge cycle, A_Cu is 0.951 at x = 0.6 (Figure 4). This value is less than 1 indicating that some Cu is lost during charge, probably by dissolution in the polymer electrolyte instead of reinsertion into the oxide, but the majority is reversibly returned into the host material. In other words, Cu reversibility in this cycle is estimated to be 95.1%. Based on the area ratio evolution (Figure 4) and literature data on CuVO,^9,18 we propose that the discharge process occurs in two steps: (i) insertion of Li into the vacancy of the Cu site until x = 1.8 (= 10.1 −8.3), and at higher values of x, (ii) simultaneous insertion of Li in to CuVO accompanied by Cu extrusion.

Zoom In Zoom Out Reset image size

Battery cycle life was studied by charging and discharging battery replicates at different fixed current densities. Figure 5 represents a typical potential profile as a function of the charge and discharge capacity (Q) every 10 cycles for a battery cycled at C/20. For C/20 batteries, the first cycle charge capacity was 242.7 ± 1.9 mAh/g, while the discharge capacity was 239.9 ± 1.9 mAh/g. Q_c and Q_d decrease continuously with increasing cycle number (N). For the battery in Figure 5, after 150 cycles, Q_d decreased by 19.7%. In addition, at a given capacity value the difference between the charge and discharge potential plateaus increases with cycle number, indicating increased polarization. For completeness, the potential profiles at C/3 and C/50 are presented as a function of Q in Figure S2. These potential profiles are qualitatively similar to data presented in Figures 5 and 3a.

Zoom In Zoom Out Reset image size

For the different C-rates studied, Q_c and Q_d are plotted as a function N in Figure 6 and the inset shows the average Coulombic efficiency (η) as a function of N. For comparison, data obtained on a Li metal coin cell battery comprising a conventional liquid electrolyte, i.e. 1 M LiPF₆ in ethylene carbonate and dimethyl carbonate (50:50 vol.%), is also shown in Figure 6. This battery was cycled at room temperature at C/35 using the same cutoff potentials as those used for the polymer batteries, i.e. in between 2.0 V and 3.6 V. For the liquid electrolyte based battery, the initial Q_d was 342 mAh/g. After 20 cycles, 13% of the capacity was lost, and after 100 cycles, 52% of the capacity was lost. This capacity fading is mainly attributed to the dissolution of Cu from the positive electrode at each cycle. In contrast at a higher C-rate, the C/20 polymer batteries, after 20 cycles, 1.8% of the capacity was lost, and after 100 cycles, 14% of the capacity was lost. This observation suggests that Cu dissolution is much slower in polymer batteries. A deeper understanding of Cu dissolution in liquid and solid electrolyte is necessary to confirm this tentative conclusion. It is not clear if the difference in positive electrode formulation between the liquid and solid battery affects capacity fade. The dependence of Q_c and Q_d on N for the polymer batteries is more-or-less independent of the C-rate. At the lowest C-rate, C/50, Q_d is initially 248.6 mAh/g, and decreases to 243 mAh/g after 30 cycles, which corresponds to a 97.8% capacity retention. At C/20, Q_d stabilizes after 22 cycles at about 237 mAh/g, and after 50 cycles the capacity decays on average by 0.33 ± 0.14% per cycle. At C/4 and C/3, Q_d drops during the first 10 cycles before stabilizing at around 217 and 159 mAh/g, respectively. The C/3 batteries were cycled up to N = 105 times with a steady discharge capacity, whose value is close to the battery based on a liquid electrolyte after 100 cycles. The two C/4 batteries failed after 17 and 31 cycles. The main difference between C/4 and C/3 batteries was the electrolyte thickness, 14 ± 2 and 33 ± 1 um, respectively.

Zoom In Zoom Out Reset image size

The Coulombic efficiencies (η) at C/50, C/4 and C/3 were 99.2 ± 0.3, 98.6 ± 0.7 and 99.5 ± 0.6%, respectively (inset of Figure 6). At C/20, η was 98.6 ± 0.7% during the first 50 cycles then dropped to 94.6% in the last cycle (N = 155). For the liquid electrolyte based battery cycled at C/35, η was 98.6 ± 1.1%. The use of a solid polymer electrolyte extends the cycle life of the α-CuVO₃ battery.

To get a better understanding on the capacity fade reported in Figure 6, selected batteries were imaged by hard X-ray microtomography. Figure 7 shows microtomography cross-section slices of batteries obtained at room temperature. For clarity, each cycled battery reported is labelled from A to G, and Table I summarizes their cycling condition and the point at which cycling was stopped for the tomography experiment. Also given in Table I is the total number of cycle (N_total) and the amount of charge passed during cycling (C_p), including the conditioning cycles. Figure 7a is an image of an uncycled battery that was heated at 90°C overnight prior being imaged at the beamline in order to observe the initial state of the battery interfaces. Four distinct layers are seen in intimate contact with each other: Li, SEO, positive electrode, and Al current collector. Note, only a small portion of the 150 μm thick Li metal is shown. Typical tomography slices of batteries after cycling at C/50 (battery B), C/20 (battery E) and C/4 (battery H and I) are shown in Figures 7b through 7e, respectively. All of the cycled batteries shown in Figure 7 were stopped in discharged state. The composite positive electrodes of all batteries were punched out of one sheet, while the SEO electrolyte used were punched out of different films which explains the qualitative difference in electrolyte thicknesses (l_SEO), also listed in Table I. For comparison, batteries were also imaged in the charged state after cycling at C/50 (battery A) and C/20 (battery C); their typical cross-section slices are reported in Figure S3. There were no qualitative differences in X-ray microtomography images obtained from charged and discharged batteries.

Zoom In Zoom Out Reset image size

The grayscale pixel values in Figure 7 correspond to the relative X-ray linear absorption coefficients of the material at that specific position. Brighter pixels correspond to higher X-ray absorption at that position. Thus, the positive electrodes appear brighter as they are less X-ray transmissive. The Li/SEO interface in Figure 7a is embellished by a thin dark band on the electrode side and a thin bright band on the electrolyte side. This is a result of Fresnel phase contrast arising during the imaging of samples containing interfaces.^53,54 The main difference between uncycled and all the cycled batteries is that in addition to the Fresnel contrast band, darker and brighter features appear at the Li/SEO interface. The measured linear X-ray absorption coefficient in the darker features (dashed yellow circles in Figure 7b) is significantly lower than that obtained in the uncycled battery while those on the brighter features are significantly higher.

In a previous study on Li metal batteries comprising a similar SEO electrolyte and a LiFePO₄ positive electrode this darker band was also observed and was attributed to a delamination of Li/SEO interface due to battery cycling.⁴⁹ In addition, similar capacity fading was also observed for those batteries (0.35 ± 0.18% per cycle) which is comparable to that obtained from our α-CuVO₃ batteries (0.33 ± 0.14%). We conclude that the dark bands in Figures 7b through 7e and Figure S3 indicate the emergence of voids, i.e. delamination of the Li/SEO interface. This delamination effect is seen in all batteries, independent of C-rate. We note in passing that these voids are not the result of the surface roughness of the Li metal electrodes. In references 37 and 49 we cycled symmetric Li-polymer-Li cells using Li metal electrodes with similar roughness and observed no evidence of void formation.

All the cycled batteries also show the emergence of bright features from highly absorptive elements located in between the Li metal and the electrolyte. We attribute these features of copper containing compounds. This was confirmed by imaging the Li electrode surface by SEM. The results are shown in Figure 8. The SEM image (Figure 8a) reveals the presence of disperse agglomerates all over the Li surface similar to those observed by X-ray microtomography (Figures 7b through 7e). Some of these agglomerates were characterized by EDS (inset of Figure 8a) and the corresponding spectra shown in Figure 8b, confirm the presence of Cu. In addition, Cu was not detected by EDS on the portion of Li metal free of agglomerates (Figure 8c). However, all over the surface, with or without Cu-rich agglomerates, carbon, oxygen, fluorine, and sulfur were identified. This indicates that a layer of SEO/LiTFSI polymer electrolyte remained on top of the Li metal. To estimate the chemical composition of the Cu-rich agglomerates, the linear adsorption coefficient of the bright features at the Li/SEO interface seen by hard X-ray microtomography (Figure 7) were determined. The linear adsorption coefficient of the bright features is 70.8 ± 15.1 cm⁻¹ whereas based on the hard X-ray energy (25 keV) employed the theoretical linear adsorption coefficient of Cu metal is 157 cm⁻¹.⁵⁵ Despite a 30% uncertainty on the experimental linear adsorption coefficient due to beamline calibration errors, we conclude that Cu metal is not the sole element in the bright features at the Li/SEO interface. Based on the experimental linear adsorption coefficient and the EDS results, we propose that the agglomerates contained light elements such as lithium and oxygen in addition to copper. For example, the linear adsorption coefficient of LiCu₂O₂ at 25 keV is 73.2 cm⁻¹, consistent with our measurements.^56,57 Further work is needed to determine the exact composition of the agglomerates, which could in principle contain other elements such as fluorine, carbon, and nitrogen.

Zoom In Zoom Out Reset image size

Battery H, whose microtomography cross-section is shown in Figure 7d, failed after 31 cycles at C/4. In this image the Cu-rich agglomerates plated onto the Li metal electrode, punctured the thin SEO electrolyte layer, and contacted the positive electrode leading to a short. The point of contact in between the two electrodes is highlighted by a white dashed square in Figure 7d. The microtomography cross-section of battery I (Figure 7e), which failed after 17 cycles, shows a globular dendritic structure, highlighted by a white dashed square, at the Li/SEO interface that expanded through the electrolyte leading to a short. The dendrite shape is identical to that obtained in failed Li symmetric cells comprising an SEO electrolyte.^37,54,58 It is evident that the use of an SEO electrolyte does not completely eliminate dendrite growth. In our previous study on Li metal polymer batteries with an LiFePO₄ based positive electrode, no evidence of dendrites was found in hard X-ray microtomography images.⁴⁹ Failure modes of the Li metal negative electrode in polymer batteries are clearly affected by the chemical composition of the positive electrodes (FePO₄ versus α-CuVO₃).

To quantify the delaminated interface and the Cu-rich agglomerate volumes, the Avizo software package was used to segment specific voxels located at the Li/SEO interface, using the procedure described in Reference 47. Voxels with brightness below a certain threshold, approximately the lowest value seen in the uncycled battery, are considered voids. Voxels with brightness higher than those seen in the uncycled batteries are assumed to be part of the Cu-rich agglomerates. The segmentation procedure was repeated on each battery imaged by hard X-ray tomography. A typical example is shown in Figure 9 for battery A. In Figure 9a we show the two orthogonal slices through the battery along the xz and yz planes before the segmentation process. The coordinate system is defined in Figure 9. The volume reconstructions of the void (V_void) and the Cu-rich agglomerate (V_Cu-rich) domains are shown in Figures 9b and 9c, respectively. The void voxels are colored blue while the Cu-rich agglomerate voxels are colored yellow. Using the same color scheme the complete volume reconstruction is shown in Figure 9d where the white region corresponds to contact area between the Li metal and the SEO electrolyte that are defect-free.

Zoom In Zoom Out Reset image size

Table I lists V_void and V_Cu-rich for each battery obtained by X-ray microtomography and the segmentation procedure. The data corresponds to portions of batteries imaged by tomography. The volume of α-CuVO₃ (V_CuVO3) in the same portion of each battery was estimated using the known loading. We also estimated the amount of Li (l_Li) deposited on the Li electrode during the initial charging step. This value is also reported in Table I. The difference in electrode volumes (ΔV) in the charged and discharged batteries in the observed portion of each battery can then be calculated, given the density of lithium (0.534 g/cm³) and the differences in unit cell volumes of α-CuVO₃ and lithiated α-CuVO₃. The value of ΔV for each battery is given in Table I. It is important to note that V_void, V_Cu-rich, V_CuVO3, and ΔV are all extensive quantities determined in independent experiments. To compare data obtained from different batteries it is essential to normalize intensive quantities. The two parameters related to battery failure are V_void and V_Cu-rich. In Table I we report two intensive quantities related to these parameter: V_void / ΔV and V_Cu-rich / V_CuVO3. In Figure 10 we plot V_void / ΔV and V_Cu-rich / V_CuVO3 as a function of the total charge passed (C_p). It is important to note that the different batteries represented in Figure 10 were cycled using vastly different C-rates and were stopped in charged, discharged, and shorted states. In spite of these differences, V_void / ΔV is 0.96 ± 0.41, independent of C_p and other battery cycling parameters (see Table I). V_void and ΔV are, thus, comparable, a correlation which was also found in LiFePO₄ polymer batteries.⁴⁹ Consequently, the void formation is attributed to the lack of adhesion of the SEO electrolyte onto the Li electrode as the electrodes shrink during discharging. The plot of V_Cu-rich / V_CuVO3 as a function of C_p shows that this ratio is also independent of C_p with an average value of 0.51 ± 0.10 (see Figure 10). In this case, we have one outlier, battery F, with V_Cu-rich / V_CuVO3 = 1.35. We have no explanation for this outlier and the average value was calculated without this datum. We posit that the Cu-rich agglomerates are formed by dissolution of Cu into the electrolyte followed by migration and electrodeposition onto the Li metal electrode. The battery imaged before cycling contained neither voids nor Cu-rich agglomerates, i.e. V_void / ΔV = 0 and V_Cu-rich / V_CuVO3 = 0. We thus conclude that the formation of voids and agglomerates happens primarily during the early cycles, 0 < C_p < 100 C. Both features appeared to be related to intrinsic characteristics of the charge-discharge cycle rather than continued cycling.

Zoom In Zoom Out Reset image size

We report on the charge-discharge characteristics of batteries with an α-Cu_6.9V₆O_18.9 (α-CuVO₃) positive electrode and a Li metal negative electrode. A nanostructured polystyrene-b-poly(ethylene oxide) (SEO) block copolymer served as a solid electrolyte in the battery. The capacity fade of our batteries is less rapid than that of α-CuVO₃–Li metal batteries with a conventional liquid electrolyte. This is attributed due to reduced mobility of Cu²⁺ in the PEO domains and resistance to dendrite growth due to the rigidity of the polymer electrolyte. In situ X-ray scattering revealed that the multistep electrochemical reactions involved are similar in the presence of liquid or solid electrolyte. The presence of polymer electrolyte does not entirely prevent copper dissolution from the positive active material. Hard X-ray microtomography revealed that upon cycling, the interface between the Li metal and the SEO electrolyte became delaminated and Cu-rich agglomerates were found on the Li metal electrode. The void volume due to delamination and volume occupied by the Cu-rich agglomerates were independent of C-rate and cycle number.

List of Symbols

This work was supported by the Joint Center for Energy Storage Research, an Energy Innovation Hub funded by the U. S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES). Didier Devaux was supported by a postdoctoral fellowship from the Philomathia Center at UC Berkeley. The WAXS and microtomography works were carried out at the Advanced Light Source at Lawrence Berkeley National Laboratory at beamlines 7.3.3 and 8.3.2, respectively. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U. S. Department of Energy under Contract No. DE-AC02-05CH11231. We thank Polite Stewart, Chenhui Zhu, and Eric Schaible for their assistance in coordinating and performing the WAXS experiments at the Advanced Light Source Beamline 7.3.3. We thank Jean-Marie Tarascon for insightful discussions.