Lithium Volatilization and Phase Changes during Aluminum-Doped Cubic Li_6.25La₃Zr₂Al_0.25O₁₂ (c-LLZO) Processing

Abstract

Stabilized Li_6.25La₃Al_0.25 Zr₂O₁₂ (cubic LLZO or c-LLZO) is a Li⁺-conducting ceramic with ionic conductivities approaching 1 mS-cm. Processing c-LLZO so that it is suitable for use as a solid state electrolyte in all solid state batteries, however, is challenging due to the formation of secondary phases at elevated temperatures. The work described in this manuscript examines the formation of one such secondary phase La₂Zr₂O₇ (LZO) formed during sintering c-LLZO at 1000 °C. Specifically, spatially resolved Raman spectroscopy and X-ray Diffraction (XRD) measurements have identified gradients in Li distributions in the Li ion (Li⁺)-conducting ceramic Li_6.25La₃Al_0.25 Zr₂O₁₂ (cubic LLZO or c-LLZO) created by thermal processing. Sintering c-LLZO under conditions relevant to solid state Li⁺ electrolyte fabrication conditions lead to Li⁺ loss and the formation of new phases. Specifically, sintering for 1 h at 1000 °C leads to Li⁺ depletion and the formation of the pyrochlore lanthanum zirconate (La₂Zr₂O₇ or LZO), a material known to be both electronically and ionically insulating. Circular c-LLZO samples are covered on the top and bottom surfaces, exposing only the 1.6 mm-thick sample perimeter to the furnace’s ambient air. Sintered samples show a radially symmetric LZO gradient, with more LZO at the center of the pellet and considerably less LZO at the edges. This profile implies that Li⁺ diffusion through the material is faster than Li⁺ loss through volatilization, and that Li⁺ migration from the center of the sample to the edges is not completely reversible. These conditions lead to a net depletion of Li⁺ at the sample center. Findings presented in this work suggest new strategies for LLZO processing that will minimize Li⁺ loss during sintering, leading to a more homogeneous material with more reproducible electrochemical behavior.

1. Introduction

All solid-state batteries (ASSBs) represent the next generation of energy storage devices [1,2,3]. Defined by their solid ion-conducting electrolytes, ASSBs promise improved safety, as well as higher energy densities, relative to traditional batteries that rely on liquid or gel electrolytes to convey ions between electrodes. Such advances are needed if Li⁺ and other alkali ion-based batteries are to assume greater roles in transportation and large-scale grid power applications [4]. Given the intrinsic advantages that ASSBs should have over their liquid electrolyte counterparts, considerable attention has focused on the discovery and development of new solid-state materials having sufficiently high ionic conductivities, as well as material compatibility with electrodes including elemental Li metal electrodes [5].

One such material identified for ASSBs is garnet-structured lithium lanthanum zirconate (Li₇La₃Zr₂O₁₂ or LLZO). Cubic LLZO (or c-LLZO) has a relatively high ionic conductivity (~1 mS/cm), good thermal stability, and wide electrochemical window [6]. Furthermore, c-LLZO is not overly reactive to humidity and is chemically compatible with solid Li electrodes [7]. LLZO is typically in its tetragonal phase (t-LLZO) at room temperature, and t-LLZO has an ion conductivity approximately two orders of magnitude less than c-LLZO [2]. Preserving c-LLZO’s high conductivity at ambient temperatures requires stabilizing the cubic phase with cation dopants such as Al, Ta, Nb, and others [6,8,9]. Doping creates lithium vacancies that allow for greater ionic conductivity. Several reports have discovered that c-LLZO’s conductivity peaks with Li⁺ stoichiometries of 6.25–6.5 [10,11].

Despite c-LLZO’s many attractive properties as a solid-state, the Li⁺-conducting electrolyte material itself is very challenging to fabricate and process [12]. Li⁺ volatilization, together with c-LLZO’s intrinsic phase instability at ambient temperatures, require careful attention to all steps during development of a dense, solid state electrolyte [13,14,15,16]. Small differences in starting material purity, as well differences in sintering procedures and uncharacterized thermal gradients, can all contribute to heterogeneity in the desired c-LLZO product [17,18]. Efforts to improve c-LLZO quality and stability have included adding excess Li (in the form of Li₂CO₃) to accommodate Li volatilization during sintering, as well as fast sintering procedures, to limit Li loss and resulting phase heterogeneity [19,20]. Li₂CO₃, however, can also segregate on the surface, forming a metal carbonate layer that serves as an insulator between the electrode and the electrolyte. Li⁺ depletion will lead to the formation of new undesirable material phases, including a lanthanum zirconate pyrochlore (La₂Zr₂O₇ or LZO) that is known to have both electronic and ion-insulating properties [21,22,23].

The findings presented in this study characterize the LZO that forms as the result of sintering c-LLZO at 1000 °C. LZO formation is associated directly with decreasing Li⁺ content and indirectly with Li volatilization (often as Li₂O) [1]. By making use of spatially resolved vibrational Raman measurements, together with X-ray diffraction, we characterize LZO formation in a pellet that has been sintered at 1000 °C while covered on its top and bottom surfaces with inert substrates, meaning that the only way Li can volatilize from the sample is by leaving from the (round) sample’s perimeter (Figure 1).

Data show that LZO forms with a higher concentration at the center of the sample and lower concentrations closer to the edges. This discovery implies that Li⁺ diffusion in the material is faster than Li volatilization so that Li lost from the sample through volatilization at the sample perimeter is quickly replaced by Li⁺ diffusing from the center of the material in response to a concentration gradient. Furthermore, Li⁺ migration from the center to the edge cannot be completely reversible, otherwise the sample would have a uniform Li⁺ concentration in the material and a more uniform LZO distribution as the Li⁺ evaporates.

2. Materials and Methods

Sample processing. A 15 mm-diameter LLZO pellet was prepared using Aluminium-doped, cubic phase LLZO (c-LLZO) powder (manufactured by Ampcera (Tucson, Arizona, USA, and distributed by MSE Supplies, Tucson, Arizona). Initially, the powder was weighed in a plastic boat and transferred to a glass beaker where it was mixed with a small amount of organic polymer binder. The LLZO-binder mixture was poured into a 15 mm-diameter pellet press die set and subjected to compression using a manual hydraulic press at 495 MPa for 10 s. Subsequently, the compressed pellets were placed in a muffle furnace and sintered at 1000 °C for one hour, with alumina slides covering both the top and bottom surfaces. A schematic illustration of the experimental procedure is shown in Figure 2.

The sintering process consisted of increasing the temperature from ambient to 1000 °C at 5 °C per minute, holding it at 1000 °C for one hour, and reducing it at 5 °C per minute back to ambient temperature. In all the studies reported in this work, LLZO samples measured 15 mm in diameter with a thickness of 1.6 mm.

Characterization—Raman spectroscopy. Raman spectra were acquired using a Renishaw InVia Raman microscope (West Dundee, IL, USA) equipped with a 488 nm single line argon ion laser (488 nm). All spectra were recorded using 0.1 mW of incident laser power. Wire 3.3 software for the microscope enables mapping measurements that scan either part of one field of view or create a montage that can be scanned across to create an image of the material. A mapping measurement using a 15 × 2 grid pattern was used to evaluate the extent of LZO formaton after sintering. Each box in the grid measured 1.25 mm × 1.25 mm. Experiments were repeated at least three times on three independently prepared samples to ensure data were reproducible. All Raman spectra are the result of 3 × 10 s acquisitions. The laser spot size in these experiments had an approximate 5 μm diameter. Mapping data were acquired using Renishaw’s Wire 3.3 software, and spectra were analyzed using Igor Pro 8.04 (Wavemetrics, Portland, OR, USA).

Characterization—X-ray diffraction. All XRD data were acquired using a 1 kW Bruker D8 Advance X-ray (Madison, WI, USA) diffractometer rotational stage and processed using instrument installed Diffrac Eva and MDI Jade (International Centre for Diffraction Data, Newtown Square, PA, USA) peak matching software. We placed a sintered LLZO pellet onto a microscope slide that had clay underneath, pressed it down in a deep sample holder until flat, and then placed the sample holder in the diffractometer. Data were acquired from 2θ = 8° to 80°, with 0.2 s per step in 0.05 degree steps. All peak PDFs were taken from the ICDD full database. All Rietveld refinements and elemental analyses were performed with Jade peak match report functionality.

3. Results

This work was motivated by the need to understand the phase stability and phase changes that occur in c-LLZO during the processing required to create viable Li⁺ ion-conducting electrolytes for ASSB applications. The most likely products to form as c-LLZO loses lithium through volatilization are t-LLZO and LZO. Figure 3A shows the XRD patterns for these three materials. As expected, c-LLZO (red, space group $Ia\overline{3}d$) had a simpler diffraction pattern, with 21 observed diffraction peaks compared to t-LLZO’s 32 diffraction peaks (blue, space group I4₁/acd, PDF 04-018-9023). The c-LLZO pattern resulted in a lattice constant of 12.972 Å. Figure 3 also shows the 2θ locations of LZO’s (black) diffraction peaks (space group $F\overline{d}3m$, PDF 04-008-6353, lattice constant 10.800 Å). These data were used to calculate material composition in the processed samples. Figure 3B shows Raman spectra for c-LLZO, t-LLZO, and LZO. In general, the LLZO spectra are characterized by strong vibrational bands in three regions. A collection of oxygen-bending modes were centered near 250 cm⁻¹, vibrations between 300 and 600 cm⁻¹ were assigned to Li-oxygen-stretching vibrations, and the symmetric stretching motion of ZrO₆ octahedra was assigned to the strong feature between 635 and 650 cm⁻¹ [7,9]. Despite these similarities, several differences are important to note. First, features are more sharply resolved in the t-LLZO relative to the c-LLZO. Second, the t-LLZO spectrum does contain several smaller features not observed in the c-LLZO spectrum. Two such vibrational bands in the t-LLZO spectrum appear at 284 cm⁻¹ and 593 cm⁻¹.

LZO’s XRD pattern and vibrational Raman spectrum are simpler than those of t-LLZO and c-LLZO, with one dominant feature centered at 300 cm⁻¹ assigned to F_2g and/or A_1g lattice vibrations [24]. Notably, this vibrational feature falls at a frequency where both t-LLZO and c-LLZO were spectroscopically quiet, although the broad, unresolved phonon structure of c-LLZO retains some intensity at this frequency. In the analyses below, we will compare the spectral intensity at 300 cm⁻¹ to the spectral intensity of the ZrO₆ stretching vibrations near 645 cm⁻¹ to assess how much LZO formed during processing. Based on the data above, we note that the limits of this intensity ratio range from 0.55 (pure c-LLZO) to ≥10.0 (pure LZO). (If comparing LZO to t-LLZO, the ratio ranges from 0.40 (pure t-LLZO) to ≥10.0 (pure LZO), although in the data shown below we never observe conclusive evidence that t-LLZO has formed).

Given that the top and bottom of the c-LLZO pellet are covered with alumina blocks (Figure 2), the only way Li can be lost through volatilization is from the pellet edge. Figure 4A shows a software mapping image of the LLZO pellet after sintering at 1000 °C for 1 h. This image is a collection of still image viewer snapshots that are stitched together using the Wire 3.3 software provided with the InVia microscope. Figure 4A contains two separate images. The image on the right illustrates the intensity at 300 cm⁻¹ in the Raman spectra acquired at each spot using heat map imagery. In this case, the brighter red (at center) corresponds to higher intensities (and more LZO), and the darker red at the edges corresponds to lower intensities (and less LZO).

The image in Figure 4A implies more LZO formed in the center of the pellet compared to at the edges. Figure 4B shows the spectra taken radially from the pellet center. While each spectrum has the general appearance of c-LLZO, the spectra closest to the center show clear evidence of a new band appearing at 300 cm⁻¹, indicating the growth of LZO. Figure 4C plots the ratio of the LZO intensity divided by the c-LLZO ZrO₆ vibration intensity (at 648 cm⁻¹). The dashed horizontal line designates a ratio of 0.55 or the limit of pure c-LLZO with no detectable LZO. From these data, we conclude that, after sintering, the c-LLZO sample has a measurable amount of LZO at its center, while c-LLZO content at the edges appears unchanged from its pre-sintering composition. Also, we note that, while some of the spectra may show a hint of a shoulder near 284 cm⁻¹, implying possible t-LLZO formation, we do not observe a corresponding feature at 593 cm⁻¹ that would confirm t-LLZO growth. This latter observation, coupled with a I₃₀₀/I₆₄₈ intensity ratio that never drops below 0.55, shows that t-LLZO formation was negligible.

A remaining question about c-LLZO’s phase evolution during thermal processing is whether the changes observed in Figure 4 persist through the bulk of the material. Raman measurements shown in Figure 4 will probe ~100 nm into the material, but only a small fraction of the sample’s 1.6 mm thickness. To assess if these radially dependent material changes measured in the near surface region extend throughout the entire sample, we broke a sample pellet in half and then measured Raman spectra at approximately 1 mm intervals from the center of the pellet out to the edge, as indicated schematically in Figure 5A. Raman spectra from the different points are shown in Figure 5B.

Again, using a ratio of the Raman intensities assigned to LZO (at 300 cm⁻¹) and LLZO (648 cm⁻¹), we evaluated the extent of LZO formation in the bulk material as a function of position relative to the pellet’s center. The spatially dependent ratios (Figure 5C) show clearly that the pellet is defined by a compositional gradient with more LZO in the center and less towards the perimeter. This LLZO–LZO distribution pattern is consistent with what is observed on the pellet’s surface. The radial distribution observed in the pellet’s cross-section, along with the surface LZO composition, show a consistent LZO concentration normal to the surface throughout the sample’s interior.

XRD measurements support findings from the vibrational Raman studies. Figure 6 shows XRD patterns from LLZO samples arranged in two different geometries, and the LLZO pattern from c-LLZO (red, bottom). The top pattern (labeled “Top surface”) was acquired from a surface of a single sample, similar to the sampling geometry used for the Raman mapping shown in Figure 4. Following Rietveld refinement, the total calculated material percentages were 97% LLZO and 3% LZO (Rwp = 17.4%). Efforts to include t-LLZO in the fit did not lead to any improvement in the refinement. From this finding, we conclude that any t-LLZO that may have formed during sintering was below detection limits, consistent with conclusions from the the Raman spatial mapping experiments. Based on the Raman mapping experiments shown in Figure 4, the area evincing clear evidence of LZO formation had a radius of ~3 mm or an area of 28 mm². This Li⁺-depleted region constituted ~16% of the total surface area. Given that the Raman data in Figure 4 show that the center region contains a mixture of both c-LLZO and LZO, the 3% value reported by the XRD pattern is not unreasonable.

The middle pattern (“cross-section”) shown in Figure 6 resulted from four samples that had been broken in half and then stacked together so that the “surface” being sampled in the XRD experiment consisted of eight cross-sections similar to the sampling geometry used for the Raman cross-sectional mapping data shown in Figure 5. At first glance, the two patterns (top and middle) look similar, but close inspection shows small differences. Rietveld refinement (Rwp = 27.2%) leads to a calculated composition of 93% LLZO and 7% LZO. Again, assuming that the Li⁺-depleted region extends ~3 mm away from the sample center, geometric considerations show that approximately 40% of the total area should contain measurable amounts of LZO. With these differences in LZO area being sampled in the two different geometries, we expect to observe 2.5 times more LZO in the cross-section experiment (40%/16%) compared to the top surface experiment. This estimate agrees with the 2.3-fold excess of LZO measured in the cross-section XRD pattern, especially given the rough estimates used to assess the Li⁺-depleted regions based on the Raman mapping measurements and uncertainties in area irradiated by the incident X-rays.

Based on the data shown in Figure 4, Figure 5 and Figure 6, we propose that that Li⁺ diffusion through c-LLZO is faster than lithium volatilization at 1000 °C. A consequence of this behavior is that, as Li is lost from the material at the sample edges, it is replaced rapidly by Li⁺ diffusing from the sample’s center. In principle, fast Li⁺ diffusion should maintain a constant concentration gradient across the sample, but the data show clearly that LZO forms at the center of the sample during sintering, but not at the edges. We propose that, as Li⁺ diffuses from the center to the edge, Li⁺ depletion in the center drives the formation of LZO, preventing Li⁺ “back-migration”, leading Li⁺ to accumulate at the sample edges. We note that our findings complement those reported recently by Zheng and coworkers [25]. In their work, Zheng and co-workers compared different methods for preparing (Ta-stabilized) c-LLZO. The authors found that LLZO precursors first formed an LZO intermediate phase that then converted into LLZO. Samples prepared for conductivity measurements were sintered at 1200 °C—higher than the 1000 °C in our work—but the samples were “covered with the same mother powder” during sintering. Based on our findings presented above, we expect covering the samples in this way limited the amount of Li⁺ volatilization and allowed the samples to retain a homogeneous c-LLZO composition throughout.

Our proposed mechanism is illustrated schematically in Figure 7. In this figure, the relative Li⁺ content along the y-axis corresponds to the Li_6.25 in the original, Al-stabilized material. A Li content of 0 would force the material to revert to LZO, an intermediate identified by Zheng and co-workers during LLZO synthesis. The position along the x-axis corresponds to the relative position along the diameter, with the locations of 0 and 1.0 corresponding to the perimeter edges (at 0 mm and 15 mm) and the center of the sample, 0.5, corresponding to the midpoint (7.5 mm) between the two edges.

This mechanism leads to irreversible loss of Li⁺ in the middle of the material and the irreversible formation of LZO. Furthermore, the data in Figure 6 show clearly that this effect extends throughout the entire sample’s interior. The effects of Li⁺ migration, coupled with LZO formation, are illustrated schematically in Figure 7, where at time t = 0 before sintering the sample consists of c-LLZO with a uniform Li⁺ distribution throughout the sample. As the sample is heated, Li⁺ diffusion to the sample edges to replace volatilized Li₂O coupled with LZO formation at the sample center leads to the heterogeneous distribution of materials observed in the final product. Given more time at 1000 °C, we expect that most, if not all, Li⁺ would be lost and the sample would revert to a homogeneous LZO pyrochlore material.

4. Summary

The experiments described in this work characterize material heterogeneity that developed in a c-LLZO sample during sintering. A combination of spatially resolved Raman vibrational spectra, together with complementary XRD measurements, shows that a LZO concentration gradient developed after the sample was heated at 1000 °C for one hour, with more LZO in the sample’s center and non-observable amounts of LZO at the sample edges. LZO formation in c-LLZO is associated with Li⁺ loss through volatilization. These experiments were designed so that Li⁺ loss opportunities were limited to the 1.6 mm-thick perimeter of the 15 mm-diameter sample. This experimental geometry, coupled with our findings, suggests a mechanism describing LZO formation and the relative stability of c-LLZO during sintering. Specifically, these results require that Li⁺ diffusion within the sample be fast, relative to Li⁺ volatilization (presumably as Li₂O), and that Li⁺ depletion in the sample center promotes LZO formation, making the loss of Li⁺ at the sample center irreversible. The data did not show any evidence of a second LLZO polymorph, t-LLZO, forming during the sintering process. Our conclusions have consequences for strategies intended to use LLZO as an electrolyte in ASSBs. Specifically, once LZO begins to form anywhere within a sample, that region is likely to become irreversibly electrochemically inactive. Furthermore, LZO will form relatively quickly in regions that become Li⁺-depleted. This proposed model now allows material scientists and engineers to design sintering methods and processes that minimize LZO formation and preserve c-LLZO homogeneity throughout samples.