Report

In situ visualization of Li/Ag2VP2O8 batteries revealing rate-dependent discharge mechanism

See allHide authors and affiliations

Science  09 Jan 2015:
Vol. 347, Issue 6218, pp. 149-154
DOI: 10.1126/science.1257289

Watching the silver lining inside

Some types of batteries contain both a transition metal reducible metal, such as the cathode material Ag2VP2O8. During operation, both Ag and V ions are reduced, and the Ag atoms can form wires to enhance the internal conductivity. Kirshenbaum et al. probe the discharge of a battery at different rates and track the formation of Ag atoms using in situ energy-dispersive x-ray diffraction (see the Perspective by Dudney and Li). They show how the discharge rate affects whether the Ag or V is preferentially reduced and also the distribution of the Ag atoms, and then correlate this to the loss of battery capacity at higher discharge rates.

Science, this issue p. 149; see also p. 131

Abstract

The functional capacity of a battery is observed to decrease, often quite dramatically, as discharge rate demands increase. These capacity losses have been attributed to limited ion access and low electrical conductivity, resulting in incomplete electrode use. A strategy to improve electronic conductivity is the design of bimetallic materials that generate a silver matrix in situ during cathode reduction. Ex situ x-ray absorption spectroscopy coupled with in situ energy-dispersive x-ray diffraction measurements on intact lithium/silver vanadium diphosphate (Li/Ag2VP2O8) electrochemical cells demonstrate that the metal center preferentially reduced and its location in the bimetallic cathode are rate-dependent, affecting cell impedance. This work illustrates that spatial imaging as a function of discharge rate can provide needed insights toward improving realizable capacity of bimetallic cathode systems.

Following the introduction of lithium iron phosphate (1), polyanion framework materials have emerged as a cathode class of interest due to increased thermal stability and higher operating voltage relative to oxides. Materials with multiple metal centers—such as LiCo1/3Mn1/3Ni1/3O2 (2), LiMn0.15Fe0.85PO4 (3), and LiNi0.5Mn1.5O4 (4), as well as numerous bimetallic sulfates (5, 6)—are being explored for use in high-voltage electrodes; however, high electronic resistance and low volumetric capacity are two inherent limitations that have been observed in these materials. Various strategies have been implemented to facilitate ion access through the use of nanosized materials and for mitigation of resistance on a practical level, including intimate mixing or coating with conductive carbon (7). Substantial gains in power output were achieved, albeit at the expense of gravimetric and volumetric energy density.

An alternate conceptual approach is the rational design of multifunctional bimetallic polyanion framework materials in which a redox active center offering the opportunity for multiple electron reductions per formula unit (i.e., vanadium) is used in conjunction with a redox active center that reduces to form a conductive metal (i.e., silver) (8). Design of active cathode materials that form conductive networks in situ can reduce or potentially eliminate the need for conductive additives that do not add to the capacity of the cell and require additional processing, thereby providing a greater overall energy density. One such polyanionic material, silver vanadium diphosphate (Ag2VP2O8), displays a reduction displacement mechanism (9) in which both vanadium and silver are reduced during the discharge process. The overall reduction processes can be expressed as Ag2V4+P2O8 + (x+y)Li0 → Lix+yAg2–yV(4–x)+P2O8 + yAg0.

Here we seek to determine the effect that discharge rate has on the discharge mechanism, including homogeneity and spatial distribution of discharged material in the cathode. Determination of the discharge conditions yielding an optimal conductive matrix (i.e., uniformly distributed throughout the cathode) is expected to improve the realizable capacity by using the entire cathode. The first task in this regard is identification of an appropriate marker to track the discharge process. The Ag0 metal product formed upon discharge is a strong x-ray scatterer, and as such its presence can be detected in small amounts (10, 11). In previous studies, it was shown that pure Ag2VP2O8 cathodes free from binders and conductive additives can be discharged, facilitating direct interrogation of the progress of the active material reduction process via tracking of silver metal as a reduction product by using thick cathode pellets (0.5 mm) to enable spatial visualization of the discharge progress within an active electrode (12). This study explores the effect of discharge rate and combines spatially resolved in situ energy-dispersive x-ray diffraction (EDXRD) data with ex situ x-ray absorption spectroscopy (XAS) measurements to elucidate the conditions under which a more optimal conductive network can be formed. We will show that these techniques offer an approach for the exploration of active materials by providing information about the redox mechanism as well as the location of the reactivity within the electrode. These techniques could be applied to other bimetallic electrode materials for which the rate of discharge may influence the reduction mechanism and affect its performance.

Li/Ag2VP2O8 cells were discharged at C/168, C/608, and C/1440 rates (fig. S1). The fastest discharge rate (C/168, 7-day) was used to provide a condition with notable evidence of polarization, whereas the slower rates (C/608, 25-day; C/1440, 60-day) were used to provide conditions with reduced polarization. It is noted that higher battery volumetric energy density can be achieved through the use of thicker cathodes and minimizing inert components. Thus, appropriate discharge rates were selected for this high-capacity cell design containing pure Ag2VP2O8 cathodes free from binders and conductive additives. The ac impedance results further highlight the effect of discharge rate (table S1), with details concerning the equivalent circuit fits provided in the supplementary materials and methods. The ac impedance data was fit to the equivalent circuit pictured in fig. S2. Upon partial discharge [0.5 electron equivalents (elec. equiv.)], the value of the charge transfer resistance (Rct) decreases by more than three orders of magnitude, from ~1 megohm in the nondischarged cell to <10 kilohm for the cells discharged at C/168 and C/1440. However, when comparing the fast-discharged (C/168) and slow-discharged (C/1440) cells, Rct is three times higher in the cell discharged at the faster rate at the same discharge level, indicating considerably higher impedance resulting from the faster discharge rate.

Energy-dispersive x-ray diffraction can be used to achieve spatial resolution of the electrochemical reduction process within the electrode inside an intact electrochemical cell. White-beam radiation coupled with a synchrotron “wiggler” insertion device emits high-energy radiation, which can penetrate bulk engineering materials such as steel, enabling in situ measurement of conventionally designed prototype and production-level batteries (13) to obtain a tomographic profile as a function of both depth of discharge and spatial location within the electrode (13, 14). In situ EDXRD was measured at several x-direction locations spaced 3 mm apart in several stainless steel coin cells with Ag2VP2O8 cathodes discharged to 0.5 elec. equiv. (Fig. 1). At each location, spectra were collected in 20-μm increments through the thickness of the electrode. For comparison, an as-prepared (not discharged) cell was also analyzed by EDXRD.

Fig. 1 Experimental setup and EDXRD spectra resulting from in situ measurements.

EDXRD spectra obtained within three coin cells: (A) the nondischarged cell, (B and C) two different locations within a cell discharged to 0.5 elec. equiv. at C/1440, and (D to F) three different locations within a cell discharged to 0.5 elec. equiv. at C/168. Spectra were obtained every 20 μm through the cathode; for clarity, only half of the scans are presented in the figure. Spectra toward the top of the figure (red hues) were obtained within the side of the cathode closer to the stainless steel coil cell top, and spectra toward the bottom of the figure (black hues) were obtained closer to the Li anode.

The nondischarged cell shows a uniform diffraction pattern throughout the cathode thickness (Fig. 1A), with differences apparent only in the first (black) scan [Ag(111) peak apparent at 1/d = 0.4239 Å (where d is the interplane spacing in the crystal lattice)], indicating formation of some Ag0 near the cathode surface at the side opposing the anode (12). The source of this diffraction peak was further investigated. Previous studies of bimetallic silver vanadium oxide materials have shown that the silver ion can dissolve into the electrolyte due to ion exchange and dissolution (15). Once the silver ion is in solution, it migrates to the anode surface and is reduced at the anode, forming a silver metal layer. Thus, it is likely that the silver metal that appears near the cathode surface is actually located at the anode surface and appears near the cathode surface due to a small amount of cell tilt during the EDXRD experiment. Ex situ analysis of the cathode of a nondischarged cell showed small numbers of dark spots on the surface where the inhomogeneous distribution of reduction products at the cathode surface is identified with optical images of cathode pellets (fig. S3). The presence of silver metal could not be identified by ex situ x-ray diffraction (XRD), providing further support that the silver metal observed near the cathode-anode interface by EDXRD may be due to silver metal deposited on the anode.

The spectra for the two positions in the slower-discharged cell (Fig. 1, B and C) are similar. The Ag(111) peak intensity at 1/d = 0.4239 Å relative to the intensity of the Ag2VP2O8 peaks is also similar between the two positions. In contrast, at the faster discharge rate (C/168), the intensity of the Ag(111) peak varies greatly between the three positions. At position 1 (Fig. 1D), the intensity of the Ag peak is larger than that observed at positions 2 (Fig. 1E) and 3 (Fig. 1F). To further clarify this point, the intensities of several characteristic peaks as a function of beam position along the z direction for the slow- and fast-discharged cells were determined (Fig. 2). Details are provided in the supplementary materials and methods section.

Fig. 2 Intensities of Ag and Ag2VP2O8 and crystallite size of Ag as a function of beam position along the z direction.

(A) Cathode discharged to 0.5 elec. equiv. at the faster rate (C/168). Three x-direction locations were measured (see schematic inset). arb., arbitrary units. (B) Cathode discharged to 0.5 elec. equiv. at the slower rate (C/1440). Two locations along the x direction were measured in this cell (see schematic inset). Error bars represent the uncertainty in the fit (see supplementary materials).

The three x-direction locations in the cell discharged at the faster rate show a large spatial distribution of silver (Fig. 2A): At location 1, the intensity of the Ag(111) peak is higher than that of the Ag2VP2O8 peak. At location 3, the intensities are comparable, whereas at location 2, the Ag(111) peak is absent through most of the bulk of the cathode. The intensity of the Ag2VP2O8 peaks is also lower at location 1 than at locations 2 and 3, indicative of a heterogeneous (nonuniform) discharge process under the higher (C/168) discharge rate. In contrast, the results from the cell discharged at the slower rate (Fig. 2B) show a much more uniform discharge across the cathode, with similar spatial distributions of Ag0 corresponding to comparable local depths of discharge. This uniformity allows for a more complete discharge of the active material by providing electron access to more of the active material simultaneously, thereby increasing the functional capacity of the cell.

The thickness of the cathode is uneven in the cell discharged at the faster rate. From the onset and offset of the Ag0 and Ag2VP2O8 intensities, the cathode is ~60 μm (~10%) thicker at location 1 than it is at location 2, where location 1 shows the lowest intensity of Ag0 through the thickness of the cathode (Fig. 2A). Complete discharge from Ag2VP2O8 to Li2VP2O8 (corresponding to total replacement of Ag by Li) would result in >8% decrease in the interlayer spacing, where microscale particle fracture may result from the crystallographic stress (9). Thus, we conclude that the uneven swelling of the cathode is related to the local depth of discharge (DOD), potentially due to fracturing of the particles of Ag2VP2O8 as they discharge. In the cell discharged at the slower rate, there is no difference in thickness between the two positions, suggesting uniform interlayer spacing through the active material (Fig. 2B).

In every discharged cathode, the Ag(111) peak is much broader than the nearby Ag2VP2O8 peaks, indicative of small Ag particles on the order of several nanometers (Fig. 1). The size of the Ag0 crystallites within the cathodes was determined from the broadening of the Ag(111) peak by applying a derivation of the Scherrer equation in reciprocal space (see supplementary materials and methods for details). This derivation was included as the Scherrer equation is typically used for diffraction data collected at a fixed wavelength, unlike the data reported here. In the cell discharged at the slower rate, we determined that the crystallite size was ~10 to 14 nm at both horizontal locations and did not vary appreciably through the thickness of the cathode (Fig. 2B). In the cell discharged at the faster rate, however, the crystallite size varied greatly between the three locations: At location 1, the crystallites were ~5 nm, whereas at location 2, there were crystallites larger than 500 nm (Fig. 2A).

To complement these data, ex situ x-ray powder diffraction spectra were acquired (Fig. 3). During this process, the complete cathodes from an as-prepared cell and cells tested to various depths of discharge under slower (C/1440) and faster (C/168) discharge rates were removed and ground before measurement (see supplementary materials and methods section for details).

Fig. 3 Ex situ XRD patterns of Ag2VP2O8 cathodes.

(A) Spectra were obtained from cells discharged at the slower rate (C/1440) to 0.1 and 0.5 elec. equiv.. (B) Spectra were obtained from cells discharged at the faster rate (C/168). (C) Area of the Ag(111) peak and crystallite size, as determined by the peak width. Error bars represent the uncertainty in the fit (see supplementary materials).

The cathode that had not been discharged was found to be pure Ag2VP2O8, with no Ag0 detected. As the cells are discharged, the intensity of the Ag2VP2O8 peaks decreases, indicating that the active material becomes amorphous on discharge, as determined by XRD (Fig. 3, A and B). Similar results have been found in the related compounds Ag2VO2PO4 (16), AgV2O5.5 (17), and Ag4V2O6F2 (18). Under both discharge rates, an increase in the Ag(111) peak area was observed, consistent with the formation of additional silver on discharge (Fig. 3C). In the discharged materials, the Ag0 crystallite size ranged from 5 to 20 nm, consistent with the 7- to 8-nm crystallite size of Ag nanoparticles found in pure, graphite-free cathodes composed of the related material Ag2VO2PO4 (8). Under the slower discharge rate (C/1440), the silver crystallite size remained constant in the cells discharged to 0.1 and 0.5 elec. equiv., whereas under the faster discharge rate, a measurable decrease in average Ag crystallite size was observed in the cells discharged to 0.1, 0.5, and 1.0 elec. equiv. Under the faster rate, we propose that additional heterogeneous nucleation sites are generated as a result of particle fracture upon local discharge of Ag2VP2O8 to Li2VP2O8. These nucleation sites promote formation of additional small silver crystallites, leading to a smaller average crystallite size for the overall population as a function of discharge. In contrast, under the slower rate, the material discharges more uniformly, with lower local stresses. Thus, a consistent Ag0 crystallite size is observed throughout.

The ex situ XRD results also suggested a difference in discharge mechanism of the bimetallic material due to a slight increase in the amount of Ag0 present at the slower discharge rate than at the faster discharge rate at the same depth of discharge (Fig. 3C). To confirm these results the relative amount of Ag0 and Ag+ and the ratio of V3+/V4+ were measured in discharged cathodes from cells discharged at three rates (C/168, C/608, and C/1440) and were compared with as-formed Ag2VP2O8 powder that was never placed in a cell and Ag0 powder as references, using the oxidation state–sensitive measurements of XAS at the Ag K-edge and x-ray fluorescence spectroscopy (XRF) at the V K-edge (Fig. 4).

Fig. 4 X-ray absorption (XAS) and fluorescence (XRF) measurements of partially discharged Ag2VP2O8.

Data in this figure are from cells discharged to 0.1 and 0.5 elec. equiv. at C/1440 (slow rate, S) and 0.1, 0.5, and 1.0 elec. equiv. at C/168 (fast rate, F), as well as Ag2VP2O8 and Ag standards. (A) Normalized absorption at the Ag K-edge. (B) Ag K-edge near-edge region. Arrows point to isosbestic points supporting the existence of a two-phase system (Ag+ →Ag0). (C) Normalized fluorescence of V K-edge. (Inset) Pre-edge peak showing no change in peak height or position until DOD reached 1.0 elec. equiv. (D) Post-edge white line for normalized fluorescence data highlighting the dependence of the shoulder feature on DOD and discharge rate. (Inset) Normalized fluorescence of the V K-edge. The dashed box shows the region that is emphasized in the figure.

As Ag2VP2O8 is discharged and Ag+ is reduced to Ag0, more than eight discrete isosbestic points can be observed in these spectra (Fig. 4, A and B). These points are an indicator that the silver in these pellets exists as a linear combination of two states: octahedrally coordinated Ag+ found in Ag2VP2O8 and metallic (BCC) Ag0 found in silver metal (19, 20). As asymmetry increased as a function of oxidation state in vanadium oxides, the intensity of the pre-edge peak for the V K-edge (Fig. 4C) can be used to determine oxidation state changes (21, 22). Because the V-O bond lengths in Ag2VP2O8 are more anisotropic than those in either V2O3 or VO2, the pre-edge peak intensity is greater, and these materials cannot be used as standards. Therefore, we compare only this peak intensity among Ag2VP2O8 samples discharged at different rates to look for small changes to the oxidation state. We observe no measurable change in the pre-edge peak intensity until a DOD of 1.0 elec. equiv., providing further evidence that at low DOD, Ag+ reduces preferentially over V4+.

Another feature showing subtle changes with discharge is the post-edge shoulder highlighted in Fig. 4D. The post-edge peak (white line) has been assigned as the 1s → 4p transition and the shoulder as the 1s → 4p shakedown transition (21, 22). In our data, we see the height of this shoulder increasing as a function of DOD. Although this is not an indicator of oxidation state, shakedown transitions are associated with metal-to-ligand charge transfer; thus, these changes to the shoulder may be related to slight distortions in the density of states near the vanadium atoms. Previous studies have shown similar changes to the white line/shoulder feature: One study of the lithiation of LixV6O13 showed a similar feature and assigned it to a possible ligand field splitting effect (23). The height of the shoulder is greater in cells discharged at the faster rate, indicating a greater effect on the local vanadium environment.

Linear combination fitting (LCF) was applied to the Ag and V K-edge data to gain additional information regarding changes in oxidation state as a function of discharge (Fig. 5). Because the silver transitions directly between two states, the ratio of Ag+ to Ag0 present in each sample could be determined using nondischarged Ag2VP2O8 and an Ag foil as the end members in a linear combination fit (Fig. 5A). To estimate the ratio of V3+ to V4+, we used as-prepared Ag2VP2O8 as the V4+ standard and Ag2VP2O8 discharged fully to 3 elec. equiv. at C/608 as the V3+ standard when performing LCF on the V K-edge (Fig. 5B).

Fig. 5 Linear combination fits of data from Fig. 4.

(A) Ag K-edge. (B) V K-edge. Error bars represent the uncertainty in the fit (see supplementary materials). (Insets) Enhancement of the linear combination fit results up to 1 elec. equiv.

Linear combination fitting indicates that Ag+ begins to reduce even at the lowest DOD measured in this study and increases monotonically as DOD increases (Fig. 5A). For cells discharged at the slower rate, the linear combination fits show a greater amount of Ag0 at the same DOD compared with the cells discharged at the faster rate, consistent with the observation of greater Ag0 intensity by EDXRD in the slow-discharged cell, particularly at the anode interface (Fig. 3). As DOD increases, this difference becomes more apparent; at 1 electron equivalent, there is twice as much Ag0 in the cells discharged at the slower rate. The Ag0 content of the cells discharged at the two slower rates (C/608 and C/1440) is consistent, as expected based on the overlapping discharge curves (fig. S1). For the cells discharged at C/608, the Ag0 content remains consistent between 2 and 3 elec. equiv., indicating that most of the silver has reduced by 2 elec. equiv. of discharge.

As with Ag reduction, the amount of V3+ present in the samples is comparable for the two slower rates (C/608, C/1440) and is, as expected, lower than the amount of V3+ in the cells discharged at the faster rate (C/168) (Fig. 5B). There is a sharp upturn in the amount of V3+ above 2 elec. equiv., consistent with our previous conclusion that by 2 electrons equivalents, most of the Ag+ has reduced and the remaining electrons must come from V4+. Tabular LCF results are provided in the supplementary materials (table S2).

Thus, two different reduction processes for the multifunctional bimetallic Ag2VP2O8 cathode material become apparent: Ag+ ions exit the structure and are reduced to Ag0, and V4+ is reduced to V3+. Although at all discharge rates both reduction processes occur, the ratio of the reduction of silver to that of vanadium changes with discharge rate where the reduction of Ag+ is favored by slower discharge rates. The spatial distribution of Ag0 is more uniform in the cell discharged at the slower rate, indicating a comparatively even discharge throughout the cathode with consistent Ag0 crystallite size. In the cell discharged at the faster rate, nonuniform reduction is observed with regions of higher and lower local Ag0 content, leading to more and less favorable electron conduction pathways through the thickness of the cathode. Upon further discharge, reduction will continue to occur preferentially at these favorable locations with enhanced electron access, with incomplete use under high rate discharge as a consequence.

By combining in situ EDXRD with ex situ XRD and XAS measurements, we are able to visualize the formation of the conductive silver matrix within the Ag2VP2O8 electrode and elucidate a rate-dependent discharge mechanism. The results of this study show that by using lower current densities early in the discharge of a multifunctional bimetallic cathode–containing cell, it is possible to preferentially form metallic silver that is more evenly distributed, resulting in the opportunity for more complete cathode use and higher functional capacity. Thus, this approach can be extended to other bimetallic and polyanionic electrode materials to probe their discharge mechanisms and spatially resolve changes as a function of usage profile, providing the insight needed to optimize these materials for their use in the next generation of batteries.

Supplementary Materials

www.sciencemag.org/content/347/6218/149/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S3

Tables S1 and S2

References (2430)

References and Notes

  1. Acknowledgments: E.S.T., K.J.T., A.C.M., C.-Y.L., and D.C.B. acknowledge funding from the U.S. Department of Energy (DOE), Office of Basic Energy Sciences, under grant DE-SC0008512. Use of the National Synchrotron Light Source beamline X17B1 was supported by DOE contract DE-AC02-98CH10886. K.K. acknowledges postdoctoral support from Brookhaven National Laboratory and the Gertrude and Maurice Goldhaber Distinguished Fellowship Program. We thank M. C. Croft for helpful discussions and Y. Belyavina for assistance with the conceptual schematics shown in Fig. 1.
View Abstract

Subjects

Navigate This Article