In-situ Li₂O-atmosphere assisted solvent-free route to produce highly conductive Li₇La₃Zr₂O₁₂ solid electrolyte

INTRODUCTION

With the development of electric vehicles and mobile vehicles in recent years, the need for lithium-ion batteries has increased significantly. Solid-state lithium metal batteries (SSLMBs) based on solid-state electrolytes are strong candidates for next-generation technologies due to their potential high safety and high energy density^[1-3]. SSLMBs transcend the capacity limitations of conventional liquid lithium-ion batteries and answer the transportation industry's requirement for increased power and energy density^[4,5]. Therefore, the solid-state electrolyte has received extensive research^[6,7]; highly ionic conductive solid electrolytes (SEs), including sulfide electrolytes and oxide electrolytes, have been developed by numerous researchers^[8-12].

Among these SEs, Li₇La₃Zr₂O₁₂ (LLZO) is attractive due to its high room-temperature ion conductivity ~10^-3 S cm^-1[13], wide electrochemical window of 0~4.3 V^[14], strong interfacial stability against lithium metal^[15], and ease of manufacture in ambient atmosphere^[16]. Nevertheless, stoichiometry LLZO is the tetragonal phase at room temperature with two to three orders of magnitude lower Li⁺ conductivity than the cubic phase^[17,18]. Element doping, such as Ta⁵⁺ and Nb⁵⁺ at the Zr-site^[19,20], Al³⁺ and Ga³⁺ at the Li-site^[21,22], is used to successfully stabilize the cubic phase structure. However, the lab-scale production craft of LLZO is less reported^[23], requiring a reliable method for research. Due to low cost and suitability for atmospheric settings, the solid-state reaction method is employed for the mass manufacture of LLZO powders and ceramics. Typically, a wet-chemistry method is used to mix the precursors and tailor the powder, requiring the use of a solvent such as ethanol (EA) or isopropanol (IPA). Huang et al. developed a non-mother-powder method using a MgO crucible for Ta-doped LLZO ceramics^[24]. Although production costs have been lowered, issues such as limited preparation speed and low consistency of the manufactured ceramics remain. Previous works investigated the relationship between lithium loss and densification during the sintering process based on the wet milling^[13,25] and further developed the wet route to produce Ta-doped LLZO as the practical example^[26]. Although ceramics can be successfully prepared delivering relatively good performances, the problem lies in the narrow parameters at each craft step, which requires careful tuning for each batch of production. In another work, we examined the negative effects of IPA solvents on milled particles^[27], where lots of LiOH·H₂O and Li₂CO₃ contaminations formed on the surfaces due to the Li⁺/H⁺ exchange^[28,29]. Other works have investigated the compatibility, reaction properties and proton exchange between the solvent and LLZO^[30-32]. Zheng et al. developed a solvent-free method to prepare Ta-doped LLZO ceramics that are more stable to water^[33,34]. However, there is a lack of research systematically studying the differences between wet and dry routes for the large-scale production.

On the other hand, due to its high sintering temperature (Zr-site- and Li-site-doped LLZO typically exceeds 1,150 and 1,050 °C, respectively), the lithium loss is a serious problem^[35,36]. Traditional mother powder [such as Li_6.55La₃Zr_1.55Nb_0.45O₁₂ (Nb-LLZO) with Li content excess 15%] was usually adopted to cover the green pellet or to supplement Li₂O atmosphere. The lithium atmosphere from the mother powder enters the green pellets from the exterior to the interior, which will result in slow and uneven sintering. It is essential to identify inexpensive lithium supplement materials and more advanced lithium supplement procedures. Zheng et al. have investigated the in-situ lithium supplementation method in previous studies^[37,38]. However, they employed wet procedure, and the consistency of the as-prepared ceramics was not further investigated. For large-scale batch preparation of LLZO ceramics with high performance and high consistency, it is critical to simplify the experimental procedure as much as possible combining with in-situ lithium supplementation strategy.

The goal of this work is to create a large-scale preparation process platform for LLZO SEs that is dependable and repeatable. Due to the low cost of the Nb element, we chose Nb-doped LLZO as a typical material. The Nb-LLZO was prepared without solvent using the conventional solid-state method, and the high-efficiency sintering of a large number of green pellets was achieved by the “pellet-on-gravel” stacking sintering method^[13]. The lithium-rich additive Li₆Zr₂O₇ [LiZO, decomposes into Li₂O + Li₂ZrO₃ during high temperature sintering (> 1,160 °C)] is selected to compensate for the loss of lithium atmosphere during the sintering process of the green pellets while also enhancing the bonding of grain boundaries and speeding up the ceramic sintering [Scheme 1]. In addition, detailed evaluations on the phase, relative density (RD), conductivity, microstructure, and grain revolution were conducted for ceramics prepared via wet and dry routes under different amounts of LiZO additives. Finally, the electrochemical performance of symmetric and Li-LiFePO₄ cells fabricated with Nb-LLZO ceramics produced by the optimized conditions was evaluated to verify the application of these ceramics.

Scheme 1. Schematic illustration of the roles of Li₆Zr₂O₇ (LiZO).

EXPERIMENTAL

Powder and ceramic preparation

Due to the high conductivity and low cost, a composition of Nb-LLZO (0% excessive Li, Li0) and 0/2/4 wt% LiZO (Li0L0/2/4) (Abbreviations and full names summarized in Supplementary Table 1) additive was utilized in this work. The precursors included LiOH·H₂O (Aladdin, AR 98%), ZrO₂ (NanJingMingShan, AR, 99.99%), La₂O₃ (Sinopharm Chemical Reagent, 99.95%), and Nb₂O₅ (Aladdin, 99.9%). The wet route was prepared as follows. Precursor powders with a stoichiometry ratio of Nb-LLZO were ball milled at 180 rpm (revolution) for 3 h with IPA and yttrium-stabilized zirconium oxide (YSZ) beads as the milling media. After drying at 55 °C for 12 h, dried powders were calcinated at 850 °C for 6 h to obtain cubic garnet-phase Nb-LLZO powders. These powders were sieved through P80 mesh and then further ball milled at 150 rpm (revolution) for 3 h to obtain a tailored powder. These fine powders were dried at 55 °C for 6 h and then sieved through 200-grit mesh. Furthermore, 1 g of sieved powder was uniaxially pressed into 12 mm green pellets at 3 tons. The solvent-free method saves solvents and drying processes compared to the wet route and differs in the material of the ball-milling tank. Scheme 2 depicts the preparation process for the conventional wet approach and the solvent-free method (hereafter known as the dry method). Table 1 displays the specific parameters for the dry and wet routes. LiZO powders were prepared by a solid-state reaction. After weighing the raw powders LiOH·H₂O (Aladdin, AR 98%) and ZrO₂ (NanJingMingShan, AR, 99.99%) at the molar ratio of lithium-to-zirconium equal to 4, they were ball-milled for 6 h at 400 rpm (rotation, revolution 200 rpm)^[37]. The mixture was dried and calcined at 800 °C for 6 h to obtain the LiZO additive.

Scheme 2. Schematic diagrams of (A) the conventional wet-milling route and (B) the modified solvent-free route, respectively.

Table 1

Preparation process comparison

Parameters	Wet	Dry
Tank material and inner dimensions	Nylon, φ110 × h100 mm	Polyurethane, φ110 × h100 mm
Grinding media	YSZ, φ10 mm × 100 g, φ5 mm × 600 g
Planetary milling machine	MITR Co Ltd, YXQM-8L. Main disk, φ420 mm
Mixing weight	~206 g + 400 g IPA	~206 g
Revolution/rotation (mixing)	180/360 r min × 3 h
Drying (mixing)	55 °C × 12 h	No
Loading in φ95 × h95 mm MgO	120 g	200 g
Calcination	Open crucible, 3 °C min-1, 850 °C × 6 h
Tailoring weight	80 g Nb4.5Li0 + 0/1.6/3.2 g LiZO + 160 g IPA	80 g Nb4.5Li0 + 0/1.6/3.2 g LiZO
Revolution/rotation (tailoring)	150/300 r min × 3 h
Drying (tailoring)	55 °C × 6 h	No
Sieving	200 grits, stainless steel screen mesh
Formation	φ12 mm, 1 g, 3 tons (~265 MPa)
Sintering	φ25 × h28 mm MgO Crucible & Lid, 7 pellets stack

RESULTS AND DISCUSSION

Precursors mixed using wet and dry processes exhibit various XRD characteristic peaks after calcination [Figure 1] (The standard cubic and tetragonal LLZO XRD patterns are calculated by Diamond software referring to Refs.^[39,40].). When LiZO was calcined after 800 °C × 6 h, a negligible quantity of new impurity phases could be observed. The ratio of Li to Zr was 4:1 for the synthesis of LiZO, where the Li₂O was excessive. At this temperature, the vapor pressure of Li₂O is low^[41] and there is essentially no extra lithium loss, indicating that the LiZO phase needs additional Li to maintain its main phase. After being calcined at 850 °C × 6 h, the wet-mixed powders showed a pure cubic phase, which was compatible with the findings in prior literature^[42,43]. The dry-mixed powder had a wider FWHM, and the peak corresponding to the La₂O₃ raw material phase (PDF # 05-0602) could be observed. This indicated that some of the tough La₂O₃ particles were not thoroughly pulverized and blended during the dry mixing procedure^[27]. A small amount of La₂O₃ would not influence the final sintering results^[44].

Figure 1. Phase result of sintered powders. XRD patterns of as-calcinated Li₆Zr₂O₇ and Li0 via wet and dry routes.

XRD patterns of Li0L0 and Li0L4 powders tailored by dry and wet routes with corresponding Rietveld refinement are shown in Figure 2. Four structure models, including cubic and tetragonal garnet-type LLZO, monoclinic-type LiZO, and trigonal-type La₂O₃, were applied to refine the diffraction data. The phase proportions were listed in Table 2. Minor La₂O₃ phase occurred in all powder refining data, and the dry routes are more apparent than the wet technique for this phenomenon. Cubic LLZO remained the dominant phase in all refinement XRD data, comprising 98% or more. As a result, the main phase is used as the basis for the evaluation of powder lattice parameters in this work (i.e., cubic phase LLZO). Dry ball milling is more effective at breaking the lattice because it is solvent-free, which causes the LLZO particles to undergo amorphization to a greater extent. There was a significant difference in the lattice parameters between fine powders produced by dry and wet routes. The lattice parameters of wet-tailored powders (Li0L0/4 wet) were all higher than 13.0 Å, whereas the lattice values of dry-tailored powders (Li0L0/4 dry) were all near to 12.9 Å. The enlarged lattice after wet tailoring originated from the solvent's absorption of moisture, which introduced serious Li⁺/H⁺ exchange. Powders in cubic LLZO with less Li⁺/H⁺ exchange after dry tailoring would benefit the sintering results.

Figure 2. Refinement profiles of XRD data of (A) Li0L0 wet, (B) Li0L0 dry, (C) Li0L4 wet, and (D) Li0L4 dry powders. The software applied for Refinement is Fullprof suite. The CIF files adopted for refinement are La₂O₃, Li₆Zr₂O₇, tetragonal LLZO, and cubic LLZO from the COD database.

The Thermogravimetric analysis- differential scanning calorimetry (TG-DSC) results of as-tailored Li0L0~4 powders were shown in Figure 3. Obviously, dry and wet tailoring gave the powders different characteristics. For wet tailoring powders, there were five stages of weight loss, whereas just two steps for the dry procedure. When cooled in an ambient setting, the sintered powders will absorb H₂O and CO₂ from the ambient atmosphere, forming chemical adsorption materials on the surface of LLZO^[45]. After wet tailoring, the weight loss stages dramatically increased to five in comparison with the dry route, indicating more serious and complex pollution in wet-milled powders. Wet-milling powders lost more weight overall than 10 wt%, indicating a considerable protonation of wet tailored powders^[46]. The weight loss of Li0L0 and Li0L4 powders after dry tailoring at stage VII was just 1.5 and 1.16 wt%, proving that the dry route greatly prevented the protonation degradation. Contrary to our expectations, the weight loss ratios for Li0L0 and Li0L4 powders after both wet or dry tailoring showed no correlation with the lithium-rich additive (LiZO). Li0L4 powder containing LiZO should lose more weight than Li0L0 because the excessive Li should decompose into additional Li₂O in LiZO powder, which would react with H₂O and CO₂ to form LiOH and Li₂CO₃. It is possible that the extra lithium formed a small amount of other lithium-rich Li-Zr-O species, such as Li₈ZrO₆, whose decomposition temperature is higher than 1,150 °C (LiZO is 1,160 °C)^[35]. In addition, the Li⁺/H⁺ exchange in Li-Zr-O compounds seemed not as serious as that in LLZO, which contributed to the less weight loss of Li0L4-wet/dry powders than that of Li0L0-wet/dry. Dry milled Li0L4 powder was expected to deliver the best sintering result due to less contamination and plenty of in-situ Li₂O atmosphere.

Figure 3. TG-DSC of Li0L0 and Li0L4 powders treated with wet-dry routes. The purge gas is 80% N₂ and 20% O₂ monitoring the practical calcination and sintering under ambient air. The I~VII stages corresponded to the following reactions: I: 20~60 °C, CH₃CHOHCH₃ (l) → CH₃CHOHCH₃ (g); II: 80~230 °C, LiOH·H₂O → LiOH·0.5H₂O + H₂O (g), H₂O (Absorbed) → H₂O (g), CO₂(Absorbed) → CO₂ (g); III: 230~330 °C, Li_6.55-xH_xLa₃Zr_1.55Nb_0.45O₁₂ → H₂O (g) + Decompositions; IV: 330~500 °C, Li_6.55-xH_xLa₃Zr_1.55Nb_0.45O₁₂ → 0.5xH₂O (g) + Li_6.55-xLa₃Zr_1.55Nb_0.45O_12-0.5x, LiOH·0.5H₂O → LiOH + H₂O (g); V: 500~1,000 °C, LiOH → 0.5Li₂O + 0.5H₂O (g), Li₂CO₃ → Li₂O + CO₂ (g); VI: 150~280 °C, H₂O (Absorbed) → H₂O (g), CO₂ (Absorbed) → CO₂ (g), Li_6.55-xH_xLa₃Zr_1.55Nb_0.45O₁₂ → H₂O (g) + Decompositions; VII: 300~1,000 °C, Li₂CO₃ → Li₂O + CO₂ (g).

Figure 4 shows the cross-sectional morphology of the as-sintered Li0L0 and Li0L4 pellets, proving the good sintering results for dry milled Li0L4 powder. Both the dry and wet milled Li0L0 cannot be densified at 1,220 °C, where many worm-like grains with rough surfaces and lots of pores occurred [Supplementary Figure 1]. The absence of plenty of Li₂O environment^[23] inhibited the densification of wet/dry milled Li0L0 [Supplementary Figures 2 and 3]. In stark contrast, the RD of the dry milled Li0L4 was as high as 96.45%. The decomposition of LiZO produced an in-situ Li₂O atmosphere, which served the same role as mother powder in promoting the sintering^[47]. When the sintering temperature reached 1,260 °C, the RDs of the dry milled Li0L0 and Li0L4 ceramics were 93.75% and 95.60%, respectively, which were much higher than that of the wet milled Li0L0 and Li0L4 (90.48% and 94.14%). In addition, the wet milled Li0L4 contained huge grains, whereas the dry milled Li0L4 contained uniform grains. The vapor pressure of Li₂O increases more rapidly compared to LLZO, allowing the high-sintering activity LLZO particles prepared by the wet route to grow rapidly, leading to the formation of an abnormal grain growth (AGG) phenomenon. Contaminations in wet milled Li0L4 powders contributed to the poor densification of corresponding ceramics. Nevertheless, LiZO effectively provided the Li₂O atmosphere for densification, and dry milling strategy ensured the dense ceramic after sintering.

Figure 4. Cross-section SEM images of the Li0L0 and Li0L4 pellets sintered at (A) 1,220 °C × 20 min, (B) 1,260 °C × 20 min.

Nyquist plots of as-sintered L0/4- wet/dry ceramics further proved the effect of Li₂O atmosphere in-situ provided by Li₆Zr₂O₆. Li0L0 ceramic sintered at 1,220 °C contained two semicircles [Figure 5A], which corresponded to the grains and grain boundaries, respectively. The inflection point between the first and second semicircle for Li0L0 was 215 kHz, where this point was defined as the terminal frequency (TF). Dry-processed ceramics had a much higher TF than that of wet-processed ceramics in the second flat semicircle in the mid-low frequency range, where all ceramics sintered at 1,220~1,280 °C exhibited this feature [Figure 5A and B, Supplementary Figure 4A and C]. The TF was positively related to the rate of Li⁺ migration at the grain boundaries^[48]. The low TF in all Li0-wet ceramics indicated the sluggish Li⁺ migration across the grain boundaries, which originated from the lack of Li₂O and the deterioration of contaminations introduced in wet milling. Comparatively, the second semicircle disappeared in Li0L4-wet/dry ceramics [Figure 5A and B, Supplementary Figure 4B and D]. The characteristic relaxation times corresponding to LLZO grain and grain boundaries were so close to each other that Li0L4 ceramics with tight grain boundaries exhibited no distinguishable two semicircles in Nyquist plots^[49]. Herein, the Nyquist plots of Li0L0 and Li0L4 ceramics sintered at 1,220~1,280 °C were fitted using the distribution of relaxation times (DRT). It is a mathematical modification to display the frequency domain-based Nyquist plots into time domain-based DRT spectra^[50,51] as follows: $$ Z(\omega)=R_{\infty}+\int_{0}^{\infty} \frac{\gamma(\tau)}{1+j \omega \tau} d \tau $$, where R_∞ denotes the ohmic resistance, γ(τ) is the distribution function of relaxation time, and τ is the relaxation time. As shown in Figure 5C and D, the Li0L0 clearly contained two characteristic peaks belonging to grains and grain boundaries, whereas Li0L4 contained only one peak, which was similar to the one peak observed for Ta-doped LLZO^[27]. As illustrated in Figure 5E and F, distinctive resistances contributed by grains and grain boundaries were calculated from DRT peaks. The Li0L0-wet sintered at 1,220 °C delivered the maximum total, Rg and Rgb resistances. The Rg of Li0L0-dry was 693 Ω, while the Rgb of 208 Ω was only half of that of Li0L0-wet. Rg and Rgb of Li0L0 tended to decrease with elevated sintering temperature, whereas Rgb could not be diminished. Compared to Li0L0 samples, the Rg and Rgb of Li0L4 were much lower than that of Li0L0, where the grain boundary impedance of Li0L4 samples could be decreased to 0.

Figure 5. Normalized Nyquist plots of Li0L0/4 ceramics sintered at (A) 1,220 °C × 20 min and (B) 1,260 °C × 20 min. (C and D) DRT transition of Nyquist plots. g: grain, bulk. gb: grain boundary. (E and F) Resistance data calculated from the DRT results. (G) Li⁺ conductivities and (H) Relative densities of Li0L0/4 sintered at 1,220~1,280 °C for 20 min.

Figure 5G and H depicts the conductivities and RD of Li0L0/4 ceramics at room temperature (23.7 °C). The RD and conductivity of all Li0L0 samples improved with increasing sintering temperature, whereas the Li0L4 exhibited a decreasing trend above 1,240 °C. The conductivities of Li0L4 ceramics maintained above 6 × 10^-4 S cm^-1, where no distinction occurred between samples processed via dry or wet milling. The RD of ceramics prepared by a dry route was approximately 2% higher than that prepared by a wet route within all sintering temperatures, proving the advantages of fewer contaminations of the dry route. An RD of no more than 90% represents a loose microstructure, where loose grain boundaries make the migration of Li⁺ difficult, resulting in low ionic conductivity. Therefore, the conductivity of Li0L0-dry samples is higher than Li0L0-wet samples at 1,220~1,280 °C sintering. Due to the absence of Li₂O atmosphere, Li0L0 pellets could only be densified at a high temperature of 1,280 °C, whereas high impedance contributed by grain boundaries could not be diminished, as shown in Supplementary Figures 4 and 5. Li0L4 ceramics could well be densified at a relatively low temperature. The vapor pressure of Li₂O generated from LiZO increased much higher than that of Nb-LLZO when the sintering temperature was raised to over 1,200 °C^[35], which promoted sintering.

As shown in Figure 6, the evolution of grain and grain boundaries was studied to reveal the advantages of LiZO additive during the dry route. The densification procedure for Nb-LLZO was divided into six stages from green pellet to dense ceramic. Numerous amorphous, glassy substances with dark hues were found among grains in wet ball-milling powders [Figure 6B and D]. Deep protonated LLZO after wet milling would first decompose at 300-700 °C and then react with Li-rich contaminations at 700-1,000 °C during the sintering process^[45], which hindered densification. Compared to the particles contaminated with solvents, dry milled powders were clean, indicating the few contaminations. When sintered at 1,080 °C for one minute (Stage 2), where the lithium-rich additive LiZO did not release Li₂O, all particles had grown with small necks developing between grains. Wet processed samples delivered a lower density than dry processed samples. The green pellets pressed from dry milled powder usually delivered high densities due to the few contaminations in powders. As the temperature increased to over 1,160 °C (Stage 3), cross-sectional microstructures of Li0L0 and Li0L4 ceramics differed substantially. The grains in Li0L4 grew significantly large with clear polyhedron surfaces and obvious linking necks. In contrast, the grains of the Li0L0 samples almost retained the same shape as that in stage 2. At stage 4, Li0L4 densified more with well-developed edges and distinct grain boundaries. In contrast, the Li0L0 grains failed to form a unique polyhedral shape, and they began to join with each other. At stage 5, the grains of Li0L4 further developed. The Li0L4-dry ceramic was almost densified while the Li0L4-wet ceramic contained lots of pores. In contrast, the Li0L0 ceramics exhibited trans-granular fractures, indicating the weak bulk strength, which might originate from the lack of Li in the LLZO lattice. At stage 6, grains in Li0L4-dry were tightly connected with clear grain boundaries and few isolated pores. In Li0L4-wet, the pores could not be completely evacuated. Li0L0 dry reached a relatively tight cross-sectional microstructure in trans-granular fracture at a high temperature of 1,300 °C, whereas Li0L0-wet exhibited abnormally grown grains with a great number of pores.

Figure 6. Evolution of the grain morphology. Cross-sectional SEM images of Nb-LLZO pellets: (A) Li0L0 wet, (B) Li0L0 dry, (C) Li0L4 wet, and (D) Li0L4 dry.

The evolution of ceramic microstructures explicated the role of LiZO in ceramic sintering: promoting the grain growth and enhancing grain boundary adhesion. The influence of dry and wet routes on particles was also clearly demonstrated: ceramics manufactured by the dry route could achieve a higher density than that by the wet route. Li0L4-dry required just 1,220 °C × 1 min to reach an RD higher than 90%. As shown in Figure 7, the changes in the RD and ion conductivity of the samples correlated to their microstructures.

Figure 7. Conductivity and relative density of Nb-LLZO ceramics shown in Figure 6.

The role of LiZO at grain boundaries was further revealed by EDS, as shown in Figure 8. A typical triangular grain boundary with glue-like segregations was selected to show the distributions of La, Zr, Nb, and O. Figure 8B depicted the element ratios at points 1, 2, and 3. Zr, La, and O ratios at bulk grains (point 3, 11.4%, 16.2%, and 67.8%, respectively) were similar to stoichiometric Nb-LLZO. The Zr ratio at the grain boundary was doubled (19.4%) to the ratio at the bulk grains. Additionally, the Zr-O ratio of the two kinds of sites ranged between 1:3 (point 3) and 2:7 (points 1 and 2); the second phase at the grain boundary was predominantly composed of Li₂ZrO₃ or LiZO species. A small excess of Nb may originate from the formation of Li_xNbO_y from Li₂O and Nb during sintering^[27].

Figure 8. Second phase species analysis at grain boundary. (A) cross-sectional SEM images of Li0L4 dry ceramics. (B) element analysis results of points at the triple-point grain boundary and grain.

Cross-sectional triangular grain boundaries were further characterized by X-ray photoelectron spectroscopy (XPS) [Figure 9]. The ceramic sample was sintered at 1,220 °C × 1 min and then was prepared inside an Ar-filled glovebox avoiding the contamination from trace-level H₂O and CO₂. Ar⁺ sputtering was performed on the surface of the stripped LLZO to remove contamination before test. And due to the unevenness of the ceramic cross-section, the binding energy of elements might be partially shifted. The O 1s peak at 54.9 eV in Li0L0 dry represents the Nb-LLZO^[52]. The O 1s spectra of Li0L4 dry were deconvoluted into two peaks at 53.7 and 55.9 eV corresponding to Nb-LLZO and Li₂O, respectively^[53]. In Figure 9B, all Zr 3d spectra exhibited a doublet with a fixed difference of 2.43 eV due to the spin-split coupling between 3d_5/2 and 3d_3/2. For dry Li0L0, just one chemical environment peak of Zr could be identified, which corresponded to the Zr in cubic LLZO^[53,54]. The Zr 3d spectrum of the Li0L4 sample exhibited evident distortion and widening, revealing two different chemical environments of Zr. The main peak of Zr 3d_5/2 located at the 182.8 eV corresponds to the Zr in the cubic LLZO while the side peak of the Zr 3d_5/2 at the 181.8 eV was ascribed to the Li-Zr-O (e.g., Li₂ZrO₃)^[55,56] at the cross-section of the sample. The Li-Zr-O could not be detected in the Raman test [Supplementary Figure 6], which originated from the small amount of Li-Zr-O existing at the narrow grain boundaries. Nevertheless, these results further proved the decomposition of LiZO to Li₂O and Li-Zr-O at the grain boundaries.

Figure 9. Second phase species analysis at grain boundary. XPS spectra of (A and B) Li 1s, Zr 3d of the cross-section pellets for Li0L0 dry and Li0L4 dry.

The production of LLZO ceramic pellets was conducted by applying Li0L2 as the composition via a solvent-free method. Li0L4 was not adopted due to the relatively low conductive decomposed Li-Zr-O compounds. Plenty of Li-Zr-O compounds in-situ formed at grain boundaries could improve the robustness of controlling the grain growth; however, these materials would hinder the Li⁺ transportation across the grain boundaries. Figure 10A depicted the RD of seven parallel samples of Li0L2 sintered by four sintering conditions. All these samples had an RD higher than 96%. The consistency was examined using standard deviation (S.D.). Clearly, the average RD of ceramics manufactured by dry method was high with low S.D. This consistency was also proved in EIS. Figure 10B depicted the Nyquist plots (25 °C) of Li0L2-dry sintered at 1,300 °C for 1 min, and Supplementary Figure 7 depicted the data for other samples. All the plots delivered only one semicircle corresponding to dense microstructure. All parallel pellets in each group sintered under four conditions showed high consistency. The Li0L2-dry sintered at 1,300 °C for 1 min had the highest RD and conductivity [Figure 10C] of 96.8% and 8.39 × 10^-4 S cm^-1, respectively. The RD and conductivities of these parallel pellets sintered at different positions inside the MgO crucible were shown in Supplementary Figure 8.

Figure 10. Inspection concerning consistency of as-produced ceramics. Consistency and mean/standard deviation in (A) relative densities, (B) Nyquist plots of Li0L2 dry sintered at 1,300 °C × 1 min, and (C) Li⁺ conductivities of Li0L2 dry pellets.

Figure 11A and B showed the Nyquist plots of the above Li0L2-dry ceramics measured at -20 to 60 °C. By linearly fitting the Arrhenius plots [Figure 11C], the activation energy calculated from the total conductivity is 0.441 eV^[57]. Highly conductive LLZO ceramics in good consistency could be produced with the aid of LiZO and a dry route. As a product, Nb-LLZO SE Li0L2 delivered high consistency in high RD (~97%) and conductivity (> 8 × 10^-4 S cm^-1).

Figure 11. (A and B) Nyquist plots and (C) Arrhenius plot of the mass-produced Li0L2 dry ceramics at -20 to 60 °C.

Because the Nb-LLZO is unstable against lithium metal^[58-60], the Li|Nb-LLZO|Li cells soon (no more than 50 h) short-circuited at 0.05 mA cm^-2 [Supplementary Figure 9]. Therefore, as-produced Li0L2-dry pellets were fabricated into Li|LE (liquid electrolyte) |Nb-LLZO|LE (liquid electrolyte) |Li symmetric cells. The Nyquist plot of a symmetric cell was shown in Supplementary Figure 10. The interface impedance was calculated to be 252 Ω cm^-2. The Li|LE|Nb-LLZO|LE|Li symmetric cells achieved good cycling stability of 2,000 h [Figure 12A and B]. The critical current density (CCD) performance [Figure 12C] of the Li0L2-dry was as high as 0.9 mA cm^-2 under a time-fixed testing method. Using capacity-constant mode and keeping the capacities at 0.1 mAh, the cell operated well at 0.1 to 2 mA cm^-2 [Supplementary Figure 11]. Figure 12D and E shows the rate performance of the Li|LFP cell, which exhibits reversible discharge capacities of 147, 140, 120, and 92 mAh g^-1, in line with current densities of 0.1, 0.2, 0.5, and 1 C, respectively. The discharge capacity increased to 136 mAh g^-1 and cycled continuously for more than 200 cycles when the rate dropped to 0.2 C. The galvanostatic charge/discharge performance of the Li|LFP cell at 0.1 C, with weak attenuation over 150 cycles [Supplementary Figure 12]. In addition, the Li|LFP full cells guaranteed a high Coulombic efficiency of over 99% during all cycles. The use of the laboratory-scale produced Li0L2-dry ceramics is demonstrated by the results of the Li|LFP battery and Li|Li symmetric cell.

Figure 12. Electrochemical performances (30 °C) of symmetric Li-Li and full Li-LFP cells fabricated with scalable prepared Li0L2-dry pellets sintered at 1,300 °C for 1 min. (A) Galvanostatic cycling performance of Li-Li symmetric cell, (B) segmental voltage profile of (A), (C) rate performance of Li-Li symmetric cell, (D) voltage profiles and (E) galvanostatic cycling performance of Li-LFP batteries at different rates.

CONCLUSIONS

The role of lithium-rich additive LiZO was demonstrated in densifying garnet Nb-LLZO. The LiZO decomposed at grain boundaries during sintering, generating a Li₂O environment that enhanced garnet sintering and offset Li-loss. After LiZO decomposition, Li₂ZrO₃ in amorphous and crystalline mixed phases spread along grain boundaries and acted as a bridge to fill the adjacent grain boundaries to improve conductivity. This work developed an economical, efficient route based on solid-state reaction to synthesize LLZO SE. The solvent-free technique avoided contamination caused by solvent-adsorbed contaminants. Increased milling energy, slightly decreased sintering activity, and increased compaction density could be achieved in the dry route for LLZO powders and green pellets. These features enabled LLZO ceramics produced via a solvent-free route to be denser than those produced by a typical wet route. Combining the above in-situ Li₂O compensating strategy and solvent-free dry method, grain boundaries in as-produced ceramics were strongly bonded in uniform sizes without AGG. The as-synthesized Nb-LLZO with 2 wt% LiZO additional exhibited a high room-temperature ionic conductivity of 8.39 × 10^-4 S cm^-1 at 25 °C, an RD of 96.8%, and high critical current densities of 0.9 mA cm^-2 at 30 °C as prepared. Moreover, the ceramics produced in such a batch exhibit an exceptionally high degree of consistency. This work provided an effective production strategy for LLZO ceramics with high performance and high consistency, which promotes the development of solid-state batteries based on garnet-type ceramics.