Comparison of Li₃In_xY_(1−x)Cl₆ Solid Electrolytes Synthesized by Mechanochemical and Water-Based Methods for All-Solid-State Batteries

Abstract

Halide solid electrolytes (HSE) have shown remarkable stability against high-voltage cathodes. Some HSE, such as Li₃InCl₆ (LIC), can be readily synthesized via aqueous routes. Here, we expand the aqueous synthesis of LIC to include Y substitution, which has different hydration coordination strengths, to form Li₃In_xY_1−xCl₆ (LIYC, 0 ≤ x ≤1). This composition is intended to combine the high ionic conductivity of LIC with the superior stability of Li₃YCl₆ (LYC). We compared solution-synthesized products with those derived mechanochemically. We found that adding ammonium chloride in a 3:1 ratio to YCl₃ + InCl₃ produces a phase-pure product, with X-ray diffraction (XRD) revealing structure similarity for both routes. Through nuclear magnetic resonance (NMR) and impedance measurements, we evaluate how the synthesis method affects ionic transport, particularly regarding correlated motion. Despite lower initial grain boundary impedance in mechanochemical samples, full cells made from solution-synthesized samples show superior cycling performance. This work establishes a scalable aqueous synthesis route for LIYC that achieves properties comparable to traditional mechanochemical methods.

1. Introduction

Solid electrolytes have the potential to enable safer solid-state batteries (SSBs) that are less flammable than conventional lithium-ion batteries (LIBs), which use organic liquid electrolytes. Some SSBs offer improved oxidative stability with high voltage cathodes, and limit Li metal dendrite formation thanks to their better mechanical properties and either a larger electrochemical voltage window or passivation reactions. Some halide solid electrolytes, such as LIC and LYC, can be synthesized in solution, enabling large-scale processing. Furthermore, halide solid electrolytes tend to be more stable with high voltage cathodes than sulfides, and more flexible than oxides. Electrolytes’ thermal and electrochemical stability are improved by close-packing structures. Whereas theoretical calculations indicate that close-packed sulfide and oxide electrolytes are not favorable for Li⁺ migration, most halide electrolytes are close-packed while benefiting from mobile Li⁺ ions (Figure 1). This low diffusion barrier is enabled by the monovalent halide anions bonding only weakly with Li⁺. Similarly, the low binding strength of halide anions with core metal ions and their high polarizability allow these materials to have superior deformability [1].

Since Asano et al.’s 2018 breakthrough [2] in the first reported halide solid electrolyte with ionic conductivity greater than 1 mS cm⁻¹, there has been significant research focused on understanding their conduction mechanisms and other relevant properties. Stainer et al. have recently investigated the ion dynamics in LIC synthesized via aqueous dissolution-precipitation and ball-milling [2]. While LIC can be easily synthesized in water and shows promise for large-scale processing, it suffers from poor stability with cathodes. LYC is more stable and has demonstrated good cycling performance using LiIn alloy anode [3,4,5]; however, LYC achieves an ionic conductivity of ~0.1 mS·cm⁻¹ when synthesized using high-energy ball milling. Here we report on the similarities and differences between synthesizing Li₃In_xY_1−xCl₆ (LIYC) through milling and water routes, with different Y and In ratios, and how it impacts the structure–property correlations, i.e., ionic conductivity and cathode cycling stability. Aliovalent doping with elements such as Zr⁴⁺, Hf⁴⁺, and Nb⁵⁺ can maintain the parent structure while increasing both the charge carrier concentration and the disorder, which effectively boosts the ionic conductivity [6,7,8,9]. However, the aqueous synthesis of these phases, even in the presence of ammonium chloride, is prevented by overly stable metal–oxygen bonds. Similarly, other recent oxyhalide solid electrolytes (SEs) [10,11,12,13,14,15] show very promising performance but rely on high-energy ball-milling to reach a highly disordered phase, challenging their mass production.

We hypothesized that combining Y and In in LIYC could yield a solid electrolyte that balances stability with high-voltage cathodes and benefits from the high conductivity of the LIC monoclinic structure. This report examines the aqueous synthesis of mixed metal ion LIYC. In our previous report investigating LYC synthesis [16], we showed that in order to synthesize LYC, it is necessary to form an ammonium chloride complex intermediate species and that grain boundary differences arise from the synthesis routes.

$$\mathrm{Y}{\mathrm{C}\mathrm{l}}_3·n{\mathrm{H}}_2\mathrm{O}+3\mathrm{N}{\mathrm{H}}_4\mathrm{C}\mathrm{l}\stackrel{-n{\mathrm{H}}_2\mathrm{O}}{\to }{\left(\mathrm{N}{\mathrm{H}}_4\mathrm{C}\mathrm{l}\right)}_3\left[\mathrm{Y}{\mathrm{C}\mathrm{l}}_6\right]$$

(1)

$$\mathrm{Y}{\mathrm{C}\mathrm{l}}_3·n{\mathrm{H}}_2\mathrm{O}\stackrel{\Delta }{\to }\mathrm{Y}\mathrm{O}\mathrm{C}\mathrm{l}+\left(n-1\right){\mathrm{H}}_2\mathrm{O}+2 \mathrm{H}\mathrm{C}\mathrm{l}$$

(2)

In a water solution of YCl₃ without ammonium chloride, Y³⁺ is coordinated by eight H₂O at the vertices of a cube [17]. These local clusters arrange themselves in a cubic face-centered close-packed fashion, with non-coordinated Cl⁻ anions accommodated in the free spaces of the lattice [18]. This coordination is drastically different from InCl₃ solutions, which form octahedral InCl₆³⁻ complexes. The addition of ammonium chloride increases the Cl⁻ activity, shifting Y³⁺ speciation to the hexachloride, and stabilizes the (NH₄)₃[YCl₆] monoclinic C2/c intermediate at low hydration level [16,19].

Synthesis from water would drastically reduce costs in terms of synthesis and processing, as well as recycling. These cost-saving factors are thought to aid in offsetting the raw material cost of In and Y. While the sourcing cost of metal salts varies according to the purity and hydration, we can estimate it from a bulk anhydrous source (upper limit) and the main, most unrefined ore source (lower limit). The average price of Y₂O₃ has fluctuated between $3 and $12/kg over the past 5 years [20]. Indium is only found as a byproduct of Zn ore processing. It is coextracted with Zn and S from sphalerite and the 2024 average cost of 99.99% In metal was $340/kg [21], about 10 times more expensive than Y ($33/kg in 2024). An estimate provides a cost of $596/kg and $320/kg for bulk InCl₃ and YCl₃, respectively, and of $1030/kg and $75/kg for their hydrated counterparts [22,23]. Thus, YCl₃∙6H₂O may be used to further reduce synthesis cost from solution, but based on the cost of reactants, anhydrous InCl₃ is more indicated than InCl₃·4H₂O to synthesize the LIYC compounds. This is because the anhydrous compound is made by direct chlorination from In metal, while YCl₃∙6H₂O is synthesized from Y₂O₃ in a solution of hydrochloric acid, further refined to YCl₃ if required through ammonium coordination and dehydration. Given that Lithium Nickel Manganese Cobalt Oxide (NMC) is insoluble in water, recycling of the spent trivalent metal center, halide-based solid electrolytes introduced here may be designed based on selective dissolution in water, followed by appropriate post-treatment such as mild annealing or freeze drying to recover the optimal grain size and ionic conductivity [24,25]. Such recycling would be less energy intensive than the current techniques designed for liquid electrolyte batteries using pyro- or hydro-metallurgical methods.

2. Materials and Methods

2.1. Synthesis

The LIYC solid electrolytes were synthesized from anhydrous LiCl (Sigma-Aldrich 99%, 793620, St. Louis, MO, USA), InCl₃ (Thermo Scientific 99.99%, 041977.18, Watham, MA, USA), and YCl₃ (Sigma Aldrich 99.99%, 451363) that were stored in an Ar atmosphere. Based on the stoichiometric amounts, 10 wt% excess InCl₃ and YCl₃ were added to offset the losses during grinding of the mixed precursors because the rare earth halide salts stick to the agate mortar [26]. To compensate for partial hydration based on thermogravimetry analysis (TGA) determination (Figure S1), an additional 4.4 wt% of YCl₃ has been added, making the total YCl₃ excess 14.4 wt%. For instance, in the case of an equimolar amount of Y and In and a 5 g target sample mass, 1.81 g (8.20 mmol) and 1.66 g (8.52 mmol) of InCl₃ and YCl₃, respectively, were added instead of 1.65 g and 1.45 g (7.45 mmol of both) without excess.

In the case of the mechanochemical-annealed (MCA) route, the precursors and Y-stabilized ZrO₂ balls (5 mm diameter) were loaded into a 40 mL Retsch milling jar in a media-to-powder mass ratio of 20:1. We pre-mixed the precursors at 100 rpm for two times 5 min with direction reversal, then milled the mixture at 500 rpm for two times 24 h cycles and scraped the wall of the jar in between cycles with a planetary mill (Retsch PM 200, Retsch GmbH, Haan, Germany). The samples were then pelletized at 632 MPa, placed in alumina vessels covered with a lid, and annealed at 500 °C for 0 ≤ x ≤ 0.5 or 350 °C for 0.5 < x ≤ 1 for 5 h in a box furnace within an Ar-filled glovebox.

In the case of the water-based (W) route, the precursors were added to a flask and dissolved in 18 MΩ water (40 mL for 5 g of precursors) and sonicated at 60 °C for 30 min. The ammonium coordinated samples (AC) were prepared in the same stoichiometric amounts as MCA ones, but NH₄Cl (Sigma Aldrich 99.5%, A4514) was added to the AC samples in a 1:3 ratio to YCl₃. The unbound water was removed through Rotovap (BÜCHI Labortechnik AG, Flawil, Switzerland) at 60 °C before drying overnight at 150 °C under a dynamic vacuum. In an Ar atmosphere, the product was ground with a mortar and pestle, pelletized, placed in an alumina combustion boat, and annealed at 500 °C for 5 h under an active vacuum in a quartz tube furnace. The pure In sample was annealed at 350 °C.

2.2. Characterization

2.2.1. X-Ray Diffraction

The XRD samples were placed in an airtight knife-edge Rigaku (Akishima, Japan) sample holder under an argon atmosphere. Spectra were measured using a Smartlab (Rigaku Corporation, Akishima, Japan) (38 kV, 30 mA) and a Cu Kα source, collecting data at 1.5°/min. The Rietveld refinement was performed using Highscore Plus v3.0 (Malvern Panalytical, Almelo, The Netherlands).

2.2.2. Impedance Spectroscopy

The LIYC samples were pelletized at 8 T for at least 1 min in a 12.7 mm die set with 18 µm thick carbon-coated aluminum symmetric blocking electrodes. The conductivity of the material was determined by electrochemical impedance spectroscopy (EIS) measurements, which were carried out with a 10 mV excitation amplitude between 7 MHz and 1 Hz using a VSP potentiostat (Biologic, SP-240, Bio-logic Science Instruments, Seyssinet-Pariset, France). The EIS spectra of LIYC were fitted using a ZARC equivalent circuit using a loop made of a constant phase element (CPE) Q_gb describing space charging and a resistor R_gb for the grain boundaries; in series with a resistor R_ion englobing lattice ionic resistance and ohmic losses, and another Q_SS for the contact capacitance in the absence of diffusion (stainless steel blocking electrodes). The associated capacitance C to R/CPE loops can be calculated from:

$$C = {\left(R^{1-\alpha }Q\right)}^{1/\alpha }$$

(3)

where Q and α are the CPE parameters.

The ionic conductivity in a crystalline material can be modelled by the Arrhenius equation:

$${\sigma }_{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{i}\mathrm{c}} T={\sigma }_0 exp\left(-{\displaystyle \frac{E_{\mathrm{a}}}{k_{\mathrm{B}}T}}\right)$$

(4)

where ${\sigma }_0$, $k_{\mathrm{B}}$, and T are respectively the temperature-independent Arrhenius pre-factor, the Boltzmann constant, and the absolute temperature. ${\sigma }_0$ represents the frequency of lithium hopping events at a standard carrier concentration within the solid ionic conductor. We used Equation (4) to fit the temperature-dependent impedance measurements.

2.2.3. Thermogravimetry Analysis

A TA instruments TGA Q500 (New Castle, DE, USA) was used to measure the TGA of YCl₃. The sample was loaded into aluminium pans and heated to 350 °C at a heating rate of 5 °C/min under a nitrogen flow of 50 mL/min.

2.2.4. Nuclear Magnetic Resonance

⁷Li static NMR spectra and pulsed field gradient-NMR were recorded using a Bruker Avance Neo 400 MHz narrow bore spectrometer (Bruker Corporation, Billerica, MA, USA) equipped with a Diff50 high gradient ¹H/X diffusion probe (Bruker Corporation, Billerica, MA, USA). All NMR measurements were conducted with temperature control using the built-in variable temperature module and samples were equilibrated for at least 5 min at specified temperatures prior to the collection of data.

Self-diffusion coefficients for the Li⁺ cation were obtained using the pulse-gradient stimulated echo (PGSTE) experiment on the ⁷Li nucleus. The Stejskal-Tanner equation (Equation (3)) was used to fit the integrated peak area as a function of the gradient strength (g) [27].

$$A(g)=A_0 exp\left(-D{\gamma }^2{g}^2{\delta }^2\left(∆-{\displaystyle \frac{\delta }{3}}\right)\right)$$

(5)

In Equation (3), A₀ is the peak area at zero gradient (g = 0), A(g) is the peak area at gradient g, γ is the gyromagnetic ratio of the nucleus, δ is the gradient pulse duration, Δ is the diffusion time between gradient pulses, and D is the self-diffusion coefficient. The values of the adjustable parameters (g, δ, and Δ) were selected to achieve ≥85% signal attenuation in at least 16 gradient steps. The fits of the signal data were performed by exporting the data from Bruker’s Topspin v. 3.6.4 software and fitting a monoexponential decay in Origin 2024.

The Nernst-Einstein Equation (6) relates the microscopic property of diffusion to the macroscopic ionic conductivity $\sigma$. We use it to estimate an equivalent ionic conductivity ${\sigma }_{\mathrm{d}\mathrm{i}\mathrm{f}\mathrm{f}}$ from NMR. The inverse Haven ratio in (7) is a useful metric to compare $\sigma$ to ${\sigma }_{\mathrm{d}\mathrm{i}\mathrm{f}\mathrm{f}}$.

$${\sigma }_{\mathrm{d}\mathrm{i}\mathrm{f}\mathrm{f}}={\displaystyle \frac{n{e}^2}{k_{\mathrm{B}}T}}{D}^{\ast }$$

(6)

$$H_{\mathrm{R}}^{-1}={\displaystyle \frac{\sigma }{{\sigma }_{\mathrm{d}\mathrm{i}\mathrm{f}\mathrm{f}}}}$$

(7)

where n is the concentration of Li⁺ in the unit cell and $D^{\ast }$ the diffusion measured through NMR. Given that the NMR diffusion measurements were conducted at a higher temperature, we extrapolated the values at lower temperatures using the Arrhenius relationship.

2.2.5. Scanning Electron Microscopy (SEM)

SEM pictures of the powder samples were collected using a Hitachi TM3030 Plus tabletop microscope (Hitashi High-Tech Corporation, Tokyo, Japan) with a 15 kV beam and mixed backscattered electrons (BSE)/secondary electrons (SE) mode. In a glovebox, the samples were deposited on carbon-coated tape prepared on a holder made of aluminum, then brought to the SEM in a Mason jar and quickly loaded into the microscope for observation.

2.3. Cycling

2.3.1. LIYC Downsizing Treatment

300 mg of W LIYC (x = 0.1, 0.3, 0.5) powders were loaded and ball-milled into ZrO₂ pots containing ZrO₂ balls (12 mm and 3 mm dia.) with a planetary ball mill (Retsch GmbH, Haan, Germany) at 100 rpm for 60 min. The LIYC synthesized by ball-milling was used in the cells without further downsizing treatment.

2.3.2. Cell Preparation

First, 50 mg of Li₆PS₅Cl (LPSCl) was used as a separate layer to construct a stable anode–electrolyte interface, by adding the powder into a PEEK cell (0.5-inch dia.) and pressing at 79 MPa for 1 min. Then, 100 mg of LIYC was added and pressed the same way. The composite cathode powder was prepared by mixing sc-NMC811, LIYC, and vapor-grown carbon nano-fibers (VGCF) at a mass ratio of 50:47:3. Approximately 12.7 mg of the composite cathode powder was evenly dispersed on the LIYC side of the pellet and then pressed at 395 MPa for 15 min. A piece of Cu foil coated with 40 μm Li was then added at the surface of the Li₆PS₅Cl and pressed at 79 MPa for 30s. Galvanostatic cycling of the cells was carried out at the same pressure, in the voltage range of 2.6–4.2 V (vs. Li/Li⁺), starting with two formation cycles at C/10, followed by cycling at C/3 (1C = 200 mAh·g⁻¹) using a Neware battery cycler (Shenzhen, China) at 30 °C.

3. Results and Discussion

3.1. Synthesis and Phase Evolution

The structure of LIC is monoclinic; however, LYC can exist as two different structures, trigonal or orthorhombic, as shown in Figure 1. The XRD patterns of the sample with x = 0.5 in Figure 2 mostly reveal a product with single monoclinic C 1 2/m 1 close to the base LIC structure, with trace amounts of LiCl. The broad peaks of the as-prepared sample show a more disordered material, whereas the annealed sample at 260 °C has sharper features due to it becoming more ordered.

To synthesize LIYC from water (W LIYC), the stoichiometric amount of NH₄Cl needed to be added in a 3:1 ratio to YCl₃ + InCl₃, suggesting that an InCl₃(NH₃)₃ complex is formed [28]. It is likely that the higher ionization potential of In compared to Y causes the negatively charged Cl⁻ to form more stable bonds with In, causing a lack of NH₄Cl around Y and hydrolysis of YCl₃ if the ratio of NH₄Cl to metal center is lower than 3. Having excess NH₄Cl on a per YCl₃ + InCl₃ basis does not seem to change the product structure; rather, it lowers its final total ionic conductivity, and Sheng et al. also found lower phase purity [28]; see Figures S2 and S3 for more details on the effect of the NH₄Cl amount on W LIYC. The W LIYC was also synthesized in two steps, with the first being the removal of the bulk solvents, and the second being the annealing step that locks in the product structure and removes the ammonium chloride. Figure 2b shows that the removal of the bulk solvent from Li₃(InY)_0.5Cl₆ results in an intermediate hydrated monoclinic structure we previously investigated [29] and revisited by Wissel et al., with characteristic reflections at 14, 16, 17, 24, and 28° 2θ [30]. We are not observing the ammonium yttrium chloride hydrated intermediate from our previous report. The material prepared via the ammonium coordinated (AC) route has more pronounced LiCl features around 30 and 35°. The trend of the relative intensities of the peaks around 14 and 34° for the annealed is inverted compared to the milled samples; the monoclinic peak around 14 is much taller in the AC sample. The final product has no reflections of the intermediate phase and is similar to the MCA sample.

The XRD reflections show a similar structural trend for MCA and W samples in Figure 3. Overall, the size of the unit cell of the monoclinic cell increases with increasing x from 427 ų for MCA LIC to 448 ų for Li₃In_0.3Y_0.7Cl₆, and from 425 for W LIC to 444 ų for Li₃In_0.5Y_0.5Cl_6, reflecting the larger ionic radius size of Y replaced into the structure. W Li₃In_0.3Y_0.7Cl₆ has four different phases; therefore, the volume of the monoclinic phase being close to x = 1 may hint at nearly pure LIC phase with Y being entirely absorbed by the hcp phases. The orthorhombic phase also sees a slight volume increase from 875(.5) ų for MCA Li₃In_0.3Y_0.7Cl₆ to 876 ų for LYC, and from 869 to 878 ų for W LIYC, due to the decrease of In content in the structure. In the MCA sample, the Y substitution in the monoclinic phase in Li₃In_0.3Y_0.7Cl₆ is more important, and the In substitution in the orthorhombic phase is less pronounced. For more details, the reader is referred to Figures S4 and S5 and Tables S1–S12. The impurity amount and unit cell volume are detailed in Figure S6. The samples from x = 1 to x = 0.5 exhibit mostly monoclinic phase with characteristic LIC peaks around 16, 16.5, 28, 29.5, and 34.5° 2θ. These peaks shift to the left with increasing Y content, pointing to an increasing size of the unit cell, consistent with a substitution of In by Y, which has a greater atomic radius. The W Li₃In_0.7Y_0.3Cl₆ sample has LiCl peaks around 30 and 35°, maybe due to the trade-off between the lower melting point of LIYC with increasing In content and the necessity to anneal at 500 °C to remove the ammonium chloride. As x decreases, there are smaller and fewer secondary features indicating increasing disorder. For x = 0.3 and x = 0.1, the reflections correspond mostly to an orthorhombic LYC phase, whereas Li₃In_0.1Y_0.9Cl₆ has a trigonal LYC phase. Both phases have some similar reflections but can be unambiguously distinguished and matched to the appropriate samples because the low-angle region shows some key differences: two peaks around 16 and 17.4° for the trigonal phase, whereas the orthorhombic phase has two twin peaks around 16° and one peak around 18°, along with a multitude of different smaller features between 20 and 32°.

Modelling the ions constitutive of LIYC with charged spheres, their stacking can assume only three categories with the following sites occupation: trigonal and orthorhombic hexagonal close-packed (respectively hcp-T and hcp-O), and monoclinic cubic close-packed (ccp-M). Sun et al. infer a correlation between the transition metal radii, charges, and the final structure of the halide compounds of the form Li_aMX_b, with M being one or more metal centers and X a halide [31]. The interaction between the cation and anion frameworks determines the final symmetry of the material.

The ionic potential I indicates how strong the electrostatic attraction is between ions carrying opposite charges. I is described as the quotient of a cation charge (z⁺) divided by its radius (r⁺). In a molecule, φ equals the ionic potential of a species times its stoichiometric number. Sun et al. explain how, for Li_aMX_b, the cationic polarization factor τ would predict the crystal structure [31].

$$\tau ={\displaystyle \frac{\sum {\Phi }_{\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}}}{\sum {\Phi }_X}}={\displaystyle \frac{{\Phi }_{Li}+\sum {\Phi }_M}{\sum {\Phi }_X}}$$

(8)

where ${\Phi }_{Li}$ is the stoichiometric ionic potential of Li⁺, $\sum {\Phi }_M$ is the sum of the stoichiometric ionic potential of the metal centers, and $\sum {\Phi }_X$ that of the halide anions.

When τ is large, the cations have a weaker impact on the halide anions, and the high symmetry arrangement is favored (ccp-M). As τ decreases, the relative polarizing power of cations increases and distorts the symmetry, forcing the anions to shift from a ccp arrangement into a less symmetric hcp arrangement. In the case of Li₃In_xY_1−xCl₆, τ ≲ 1.74 would correspond to trigonal compounds, 1.74 ≲ τ ≲ 1.76 to orthorhombic, and 1.76 ≲ τ to monoclinic. However, our findings nuance this ionic potential model, as can be seen in Figure 4. In this model, the ionic potential for LYC is ≈1.73, making it belong to the trigonal region, whereas we observe the orthorhombic structure, as in their initial report on LYC [32]. Figure S6 shows that errors due to moisture determination cannot account for the discrepancy between the predicted hcp-T LYC phase and the observed hcp-O. Moreover, the ionic radius considered for Y in the model seems appropriate since x = 0.1 has a trigonal structure [32]. Thus, there are other parameters regulating the structure of Li₃MCl₆. That said, we obtained the same product trend via different synthesis routes, suggesting that the synthesis temperature, dwell time, and cooling rate are the main drivers. For example, Ref. [32] reports an orthorhombic structure for MCA LYC with 260 °C annealing temperature for 5 h, as does Meyer et al. [33] when melting at 650 °C and slowly cooling at 1 °C/min, and Lutz et al. when melting between 500 and 600 °C and using the Tamman-Stöber cooling technique to produce large crystals. Banik et al. report the trigonal phase for LYC when synthesizing at 550 °C, maintaining the temperature for a week, which likely leaves time to reach the thermodynamically favored structure, and cooling down without controlling the rate [34].

NH₄Cl needs to be added in a 3:1 ratio to the metal center, either Y or In in the case of the AC route. Annealing is necessary to remove the flux, and in the case of MCA, it promotes a more ordered phase with sharper XRD reflections. The resulting phases are very similar, except in the case of x = −0.7 with more LiCl impurities for the AC sample, which is likely caused by the increased amount of In, lowering the melting temperature and interfering with the elevated temperature required to remove NH₄Cl. The monoclinic phase is maintained from x = 0.3 to 1 in this study, transitioning to hcp for x ≤ 0.1. Interestingly, the thermodynamically favored hcp-T phase for τ ≲ 1.74 and x ≲ 0.1 is only observed for x = 0.1; LYC has the hcp-O phase. The In substitution may cause the Li⁺ to be less tightly bound, allowing it to reach the thermodynamically favored hcp-T phase faster than for LYC. It may be possible to avoid the thermodynamically favored hcp-T phase for LYC by frustrating the structure with a relatively short dwelling time at high temperature or relatively fast quenching, thus locking a slightly more ionically conductive hcp-O phase as discussed in the next section.

3.2. Li⁺ Transport

We chose to focus mostly on the samples with x = 0.1, 0.3, and 0.5 in Li₃In_xY_1−xCl₆ because they are on each side of the observed phase transition from trigonal to monoclinic at x = 0.2, and because x = 0.5 seems to exhibit high ionic conductivity and potentially improved cathode stability [32]. There is an optimal annealing temperature for the LIYC samples. For x = 1 (pure LIC), the ionic conductivity increases with the annealing temperature from unannealed up to 350 °C, which is the best ionic conductivity measured; the impedance relationship with annealing temperature is available in Figure S8. Decreasing x (increasing Y content) increases the optimal annealing temperature. In coherence with previous studies, the annealing temperature for x ≤ 0.5 has been chosen at 500 °C, which is close to the optimal annealing temperature in terms of ionic conductivity [34].

The Li₃In_0.1Y_0.9Cl₆ samples in Figure 5a,d have a hexagonal close-packed structure and lower total ionic conductivity. MCA and W–Li₃In_0.1Y_0.9Cl₆ have an equivalent capacitance for Q_gb of, respectively, 6.20 · 10⁻¹⁰ and 1.05 · 10⁻⁹ F m⁻¹, mostly attributable to grain boundaries [35]. As can be seen in Figure 5b,e, the samples have increasing conductivity, increasing x between 0 and 1, both for MCA and W samples. Transitioning from the trigonal structure (x = 0.1) with a conductivity of 0.1 mS cm⁻¹ to the mix of orthorhombic and monoclinic phases (x = 0.3) leads to an increase in the effective conductivity. For the MCA samples, this phase transition leads to an order of magnitude increase in total conductivity as the grain boundary impedance drastically shrinks. For the W sample, the overall conductivity increase is more limited, by a factor of 3. Nevertheless, this confirms the conclusion of better conductivity of the hexagonal closed pack orthorhombic (hcp-O) over the trigonal (hcp-T) for Li₃MCl₆ solid electrolytes [31]. The MCA and W samples with x = 0.5 have, respectively, a conductivity of 1.07 mS cm⁻¹ and 1.06 mS cm⁻¹, and negligible grain boundaries. MCA and W samples at x = 0.7 have just above 0.6 and 0.4 mS cm⁻¹ conductivities, respectively, also lower than the x = 0.5 and pure LIC samples above 1 mS cm⁻¹. The W Li₃In_0.7Y_0.3Cl₆ exhibits a larger LiCl reflection on the XRD scans, due to a potential phase separation resulting from the lower melting point of InCl₃, which would occur when annealing to remove the NH₄Cl. The above observations corroborate literature results and confirm the better ionic conductivity of monoclinic cubic close packed (ccp) Li₃MCl₆ over both types of hcp phases [32,34]. However, the LYC sample having the orthorhombic phase has expectedly better conductivity than the x = 0.1 sample.

Figure 5c,f shows Arrhenius behaviour of the total conductivity based on the temperature. The y-axis intersect is −ln(${\sigma }_0$) which are similar for both monoclinic phases x = 0.5 and x = 0.3 but significantly lower for x = 0.1, confirming that the ccp-M phase ionic conductivity is inherently superior to hcp for LIYC. Activation energies (E_a) are similar among the samples, ~0.35 eV, and comparable to Banik et al., reporting activation energies ranging from 0.49 eV for LYC, to 0.38 eV for x = 0.7, and ~0.32 eV for LIC [34]. The E_a is inferior to 0.40 eV for all monoclinic (x ≥ 0.3) samples and just above it at 0.42 eV for the x = 0.1 W sample. The x = 0.1 MCA sample has a similar E_a at 0.38 eV.

Solid-state ⁷Li NMR spectra were collected on powders of LIC and LYC of both W and MCA, and Li₃(InY)_0.5Cl₆ W. Across these samples, as can be seen in Figure 6, those containing Y exhibit broad non-Gaussian peaks which are suggestive of quadrupolar interactions with an asymmetric electric field imposed by the crystal lattice. Upon heating, the central transition in the NMR spectra of Li₃(InY)_0.5Cl₆ W sharpens significantly, and the quadrupolar satellite peaks become easier to resolve (Figure S10). In fact, at 140 °C, two sets of quadrupolar satellite peaks become partially resolved from each other, which would be consistent with a temperature induced ordering in the sample resulting in two non-exchanging inequivalent lithium sites. We extracted a diffusion coefficient (Figure 6b) from these spectra through the Stejskal-Tanner Equation (3).

NMR shows slightly faster diffusivity in the AC than in the MCA sample, and a lower activation energy compared to the impedance measurement. LIC diffusion values are reported at 60 °C 9 × 10^—12 m²s^—1 in [2], between 6 and 7 × 10⁻¹² m²s⁻¹ in [36] and for LYC at 80 °C between 5 and 6 × 10⁻¹² m²s⁻¹ in [16]. Li₃(InY)_0.5Cl₆ is of the same order of magnitude between 1 and 3 × 10⁻¹¹ m²s⁻¹ at 85 °C. $H_{\mathrm{R}}^{-1}$ > 1 shows correlated ionic hopping. Increasing values of $H_{\mathrm{R}}^{-1}$ with temperature indicates that the ionic conduction relies more on this correlated mechanism as the temperature increases. Stainer and Wilkening [37] provided definitive proof through temperature-dependent NMR and appropriate modelling that the Li⁺ diffusion relies mostly on 2D motion in the ab plane of Figure 1A. The Li₃(InY)_0.5Cl₆ NMR activation energies of 0.27 ± 0.02 (W) and 0.26 ± 0.05 eV (MCA) are close to Stainer and Wilkening’s LIC (0.25 eV). The activation energies in Figure 5c,f measured through impedance spectroscopy are higher due to probing not only the local diffusion, but also the slower transitions at grain boundaries.

Because Li₃(InY)_0.5Cl₆ and LIC have similar diffusion, activation energies, and monoclinic structure with Li atoms organized in parallel planes in the lattice, the barrier for Li⁺ diffusion for Li₃(InY)_0.5Cl₆ should also be lowest within the 2D planes [2,34,37]. Doping is commonly used to improve the ionic conductivity in SE by introducing disorder and vacancies. In LIYC, the partial occupation of In and the stacking defects induced by Y substitution may facilitate Li⁺ transport along the c-axis. However, for the 2D layered structure of monoclinic LIYC, such disorder seems to hinder long-range 2D transport. The long-range Li⁺ transport disruption results in the lower ionic conductivities relative to LIC observed in Figure 5b,e [2,6].

When observed, the grain boundary impedance is larger for the W samples owing to more LiCl impurities. The activation energy and grain boundary impedance decrease as the ionic conductivity increases and with In addition, except for the x = 0.1 sample with hcp-T phase, which has the lowest ionic conductivity. The activation energy found through NMR compared to EIS is lower because the environment probed is more local. The grain boundaries have less impact on the measurement, which reveals that the AC samples have slightly higher bulk diffusion than the MCA. NMR shows correlated ionic motions, with the ionic transport being favored through the 2D planes as the temperature increases.

3.3. Cycling

We cycled the MCA and W samples with x = 0.1, 0.3, and 0.5 in full cells with an LPSCl interlayer and a Li metal anode. The anodic sweep of the linear sweep voltammetry (LSV) in Figure S11 shows an onset of 4 V for the oxidative peak, in line with previous reports for LIC [28,38] and LYC [39]. During the conditioning cycle at C/10, Figure 7 shows that samples W Li₃In_0.5Y_0.5Cl₆ and Li₃In_0.3Y_0.7Cl₆ have an initial capacity of about 180 mAh g⁻¹ and initial coulombic efficiencies (ICE) of, respectively, 91% and 88%, whereas Li₃In_0.1Y_0.9Cl₆ showed only 105 mAh g⁻¹ capacity and 70% ICE initially. W Li₃In_0.5Y_0.5Cl₆ has the highest capacity retention at 100 cycles, around 120 mAh, which is also 80% of its initial capacity at C/3. W Li₃In_0.3Y_0.7Cl₆ reaches 80% of its initial capacity at C/3 at about 27 cycles and 110 mAh. W Li₃In₀₁Y_0.9Cl₆ reaches the 80% mark before 10 cycles. We observe initial C/10 capacity and coulombic efficiency for the MCA Li₃In_0.5Y_0.5Cl₆, Li₃In_0.3Y_0.7Cl₆ and Li₃In_0.1Y_0.9Cl₆ of respectively 190 mAh and 89%, 170 mAh and 72%, and 149 mAh and 87%. The capacity fade at 80% of initial performance for these same samples is attained, respectively, at 60, 45, and 30 cycles.

Overall, higher In content leads to improved ICE and cycling stability despite showing greater oxidative currents above 4.2 V (Figure S11). W-LIYC cells only showed similar cycling upon the addition of a quick milling post-processing step. The W-LIYC particles agglomerated during drying and therefore needed to be ground to smaller sizes. When particle size was appropriately reduced, the W LIYC exhibits slightly better stability than the MCA LIYC at all stoichiometries but Li₃In_0.1Y_0.9Cl₆., which is not due to the amount of LiCl impurity, as the MCA and W samples have equivalent LiCl amounts (Figure S6). As such, the amount of LiCl in our samples is not the main predictor of the battery performance, as it is in the same range for x = 0.1, 0.3, and 0.5, and is slightly higher for the W samples.

The scanning electron microscope (SEM) images of the electrolyte in Figure 8 reveal agglomerated particles for W LIYC. The larger, more defined NMC 811 particles are visible for both W and MCA LIYC. The annealing treatment on dense MCA pellets sinters the particles, causing the SE particles in the cathode not to blend as well with the active material, and explaining their faster capacity fade.

The impedance data informs us about the degradation mechanisms in the cell. All the semi-circles on the Nyquist plots in Figure 9a,b have increasing diameter after cycling, showing degradation in the cells. The impedances after cycling are largely convoluted; only one or two features are clearly visible, whereas the distribution of relaxation times (DRT) shows four peaks in Figure 9c,d, and therefore helps pinpoint the sources of the degradations. Only representative Nyquist and DRT plots for cycled cells with W LIYC are shown in Figure 9; the other plots can be found in Figure S12. Long-range, mass-transport-limited diffusion is excluded as it causes divergence in the standard DRT integral [40], and the intrinsic SE ionic conductivity is not represented as it is a purely resistive phenomenon. An increase in the peak area indicates an increase in the impedance of that process, such as material degradation. Faster processes can be assigned to higher frequencies/smaller time constants in the DRT spectra: 10⁻⁶ s characterizes grain boundary ionic transport of both LIYC and LPSCl (peak A, R_GB), 10⁻⁵ s the SE/current collector (CC) and LIC/LPSCl interfaces (peak B, R_{LIC/LPSCl+LIC/CC}), 10⁻⁴ s and 10⁻³ s the SEI and CEI, respectively (peak C, R_SEI and peak D, R_CEI) [41,42,43,44,45]. The evolution of all the resistances associated with the DRT peaks is summarized in Figure 9e,f.

For Li₃In_0.5Y_0.5Cl₆, R_GB is smaller in the W than in the MCA cells, with respectively 1.7 and 3.0 Ω in the pristine cells; increasing 11 and 15 times in the cycled cells. For W Li₃In_0.3Y_0.7Cl₆, R_GB is 7.2 Ω in the pristine W cells and 3.6 Ω in the pristine MCA cells, but it increases 4 times for the cycled W cells and 41 times for the cycled MCA cells. In Li₃In_0.1Y_0.9Cl₆, A is almost identical in the pristine and cycled W cells at 450 Ω. The MCA Li₃In_0.1Y_0.9Cl₆ cell has a feature at 150 Ω in the pristine state, increasing 4 times after cycling. This indicates greater bulk degradation of the MCA samples in this region, especially with a lower In:Y ratio—apart from W Li₃In_0.1Y_0.9Cl₆ [41].

R_{LIC/LPSCl+LIC/CC} in the W Li₃In_0.5Y_0.5Cl₆ cells increases 4 times after cycling from 1.6 Ω in the pristine state. In the MCA Li₃In_0.5Y_0.5Cl₆ cells, it increases 19 times from 2.6 Ω. In the W Li₃In_0.3Y_0.7Cl₆ cells, it increases 3 times after cycling from 5.2 Ω in the pristine state. In the MCA Li₃In_0.3Y_0.7Cl₆ cells, it surges 57 times from 2.3 Ω. In both W and MCA Li₃In_0.1Y_0.9Cl₆, it increases greatly after cycling, 13 times from 140 Ω in the W sample and up to 520 Ω from a non-visible level in the pristine MCA sample. The increase in R_{LIC/LPSCl+LIC/CC} after cycling suggests either particle degradation in the cathode mix or delamination between LIC and LPSCl is one of the main causes of the capacity loss, and this failure cause is more severe with increasing amount of Y and overall in the MCA samples [42].

Consistent with being associated with the SEI formation, R_SEI is not perceivable in the pristine cells, except for W Li₃In_0.1Y_0.9Cl₆ at 29 Ω. It generally does not contribute significantly to the impedance of cycled cells, remaining at 5.4 Ω for W Li₃In_0.5Y_0.5Cl₆ and 9.3 for Li₃In_0.3Y_0.7Cl₆, and at 37, 85, and 61 Ω for the MCA samples with x = 0.5, 0.3, and 0.1, respectively, and demonstrating the formation of a stable SEI. W Li₃In_0.1Y_0.9Cl₆ stands out with a 20 times increase to 590 Ω after cycling, possibly linked to the great R_{LIC/LPSCl+LIC/CC} increase and showing growing dendrites or isolated Li causing a larger LIC/LPSCl interface as well as SEI [42].

R_CEI does not contribute as much to the performance decrease. It remains stable from 8.8 to 9.6 Ω in the W Li₃In_0.5Y_0.5Cl₆ before and after cycling, increases 2 times from 20 Ω for W Li₃In_0.3Y_0.7Cl₆, and 7 times from 67 Ω in Li₃In_0.1Y_0.9Cl₆. In the MCA samples, R_CEI grows 7 times from 8.6 Ω in Li₃In_0.5Y_0.5Cl₆, 12 times from 10 Ω in Li₃In_0.3Y_0.7Cl₆, and 2 times from 48 Ω in Li₃In_0.1Y_0.9Cl₆.

R_GB and R_{LIC/LPSCl+LIC/CC} are the resistances contributing most to the impedance increase after cycling, suggesting the degradations occur mostly through mechanical degradation of LIYC in the bulk and causing delamination of the LIC/LPSCl interface. These two mechanisms are also the ones most worsened by decreasing In:Y ratio, which is particularly noticeable for the W Li₃In_0.1Y_0.9Cl₆ cell with R_{LIC/LPSCl+LIC/CC} growing from 140 Ω to 1900 Ω, and associated with growing isolated Li or dendrites since there is also a large R_SEI growth.

The total resistance triples after cycling for W samples of x = 0.5 and 0.3, whereas it is an order of magnitude higher for the corresponding MCA samples. Comparatively, higher In content results in lower impedance before and after cycling and achieves better stability. Importantly, the slightly greater ionic conductivity of the MCA samples does not help prevent subsequent degradation during cycling, with a larger increase in the grain boundaries of MCA versus W samples.

4. Conclusions

Our work compares the structure, electrochemical properties, and cycling performance of LIYC synthesized in aqueous solution and mechanochemically. Ammonium chloride is used as a sacrificial additive for Y to be incorporated into the LIC structure. Its proportion needs to be 3:1 relative to In + Y, leading to LiCl impurities if this proportion is not respected, and oxychloride as well if below. The substitution of Y for In shifts the low conductive hcp phase to the highly conductive ccp-M phase. The transition from hcp to ccp-M with increasing In:Y ratio follows a similar trend for both synthesis routes. The LYC samples have hcp-O phase, whereas sphere packing modelling of the ions in the unit cell predicts hcp-T [31]. Both the thermodynamically favored hcp-T LYC phase [34] and hcp-O [32,33,46] are encountered in the literature; longer annealing time may favor the hcp-T phase. Expectedly, the volume of the unit cell decreases with increasing In:Y. Due to lower intrinsic ionic conductivity in hcp samples, the grain boundaries bottlenecks are more easily resolved in Li₃In_0.1Y_0.9Cl₆ and LYC. The grain boundary impedance is also magnified by the aqueous synthesis for all LIYC stoichiometries because it tends to leave more LiCl impurities. As the In:Y ratio and the ionic conductivity increase, both the activation energy and grain boundary impedance decrease, apart from the x = 0.1 sample with hcp-T phase exhibiting the lowest ionic conductivity. The activation energy determined by NMR is lower than that obtained from EIS because NMR probes more localized environments, making grain boundaries less influential in the measurement. This indicates that AC samples possess slightly higher bulk diffusion compared to MCA samples. Additionally, NMR results demonstrate correlated ionic motions, with ionic transport becoming increasingly favored within the 2D planes as temperature rises. In-rich samples with ccp-M phase experience less capacity fade thanks to their order of magnitude higher ionic conductivity. Although the grain boundary impedance is more important for samples synthesized through the aqueous route, the corresponding cells show better stability, owing to smaller grains that tend to fragment less and keep better contact during cycling.