Wetting behavior of painted LM with naturally oxidized skin
Herein, we choose Ta2O5 as doping agent for cubic garnet electrolyte in view of the structural stabilization and high ionic conductivity of Li6.5La3Zr1.5Ta0.5O12 (LLZT) electrolyte (Supplementary Figs. 1–3)11. The electrochemical impedance spectra (EIS) of Ag/garnet/Ag symmetric configuration were measured to obtain a RT ionic conductivity of LLZT of 4 × 10−4 S cm−1 with an activation energy of 0.36 eV (Supplementary Fig. 4). Pristine garnet electrolyte with a lithiophobic surface from insulating Li2CO3 causes poor wettability for lithium anode and therefore point contact between anode and electrolyte (Fig. 1a). There is a big gap that separates molten lithium from pristine garnet electrolyte surface (Supplementary Fig. 5). In contrast, with simple painting of gallium LM on garnet surface (marked as LM@LLZT) under air, a highly lithiophilic and conformal interlayer is expected to be constructed due to the elastic Ga2O3 skin coverage on LM inclusions (Fig. 1b). Under oxygen-lean condition (e.g., in vacuum or argon), LM has low viscosity and high surface tension, and exists in the form of nearly spherical shape, which is hard to wet garnet substrate and shows a large contact angle more than 90° (Fig. 1c). After LM is exposed to air to generate an oxide layer, significantly reduced surface tension and improved viscosity endow LM with excellent wettability and paint behavior37. After removing excess gallium metal, a dark film composed of residual elastic gallium oxide is well formed on garnet surface. The realistic photos at different stages are correspondingly provided as insets of Fig. 1c. To further certify the versatility of the unique painting behavior of LM, we dropped them on different substrates of Al2O3 plate, Cu foil and A4 paper (Fig. 1d, e). The excellent spreadability is also observed for all the substrates when LM is operated under air with the formation of self-passivating Ga2O3 film. This stable pasting behavior is caused by surface-dispersed Van der Waals force of gallium oxide, which is favorable for the gap or void healing at anode–electrolyte interface38. The scanning electron microscopy (SEM) images of LM@LLZT further certify the wetting behavior of surface-oxidized LM. Note that the Ga2O3 film enables a smooth and continuous coating on garnet grains (Fig. 1f and Supplementary Fig. 6), and the rough GBs are not discernable. From the view of cross-section images (Fig. 1g and Supplementary Fig. 7), LM with oxide skin totally wets the bumpiness zone and GBs at electrolyte surface without any gap exposure. Energy dispersive X-ray spectroscopy (EDS) mapping of LLZT proves the homogenous distribution of Ga element as consequence of favorable painting of LM (Supplementary Fig. 8).
When dipping LM@LLZT into molten lithium, fast lithiation can be achieved in merely 2 min (Fig. 2a), while molten lithium cannot wet the surface of pristine LLZT covered with an intrinsic lithiophobic skin (Supplementary Fig. 9)17,18,19,20. The garnet surface after lithiation shows a compact coverage of lithium with desired metallic luster. The improved wettability of Li metal toward LM-painted garnet is further confirmed in Supplementary Fig. 10. When a Li disc is heated at 200 °C, the molten Li spreads quickly and covers the entire surface of LM@LLZT within 5 min with the color turning from pristine dark (the color of gallium-oxide skin) to yellow (the color of Li). After lithiation, the smooth and continuous morphology of interlayer is well preserved (Fig. 2b, c). The excess Li and lithiated LM are elaborately blended without texture segregation phenomenon, indicating an excellent spatial compatiblity of LM interlayer with both anode and electrolyte. Even at a large capacity up to 1 mAh cm−2, the Li/LM@LLZT/Li symmetric cell can run stably (Supplementary Fig. 11). The newly formed interfaces after Li plating and stripping still show tight contact between Li and LLZT (Fig. 2d, e), benefiting from the high mobility of LM, which enables a prompt filling or wetting of pits or gaps left after Li stripping or deposition.
X-ray photoelectron spectroscopy (XPS) analysis was performed to characterize the surface components of LM@LLZT before and after lithiation (Fig. 2f, g). For pristine LM decoration, the Ga2O3 peak is found at 20.9 eV for Ga 3d as expected, and it is more intensive than that of Ga metal at 19.5 eV, which is conformally coated by the former39,40. For O 1s spectrum, the corresponding Ga2O3 peak is located at 531.9 eV41. The other should peak (with lower intensity) at 530.5 eV likely stems from the generation of Ga2O3−x as a consequence of insufficient surface oxidation. After lithiation, Ga2O3 and Ga components are expected to convert into LixGa and Li2O products42. Therefore the LixGa (19.1 eV) and Li2O (528.5 eV) peaks become dominant in the Ga 3d and O 1s spectra, respectively4. Note that the undecomposed LixGaOy still remains from the appearance of peaks at 20.7 eV for Ga 3d and 531.1 eV for O 1s40. We also compare the corresponding XPS results before and after etching the surface for 10 s (Supplementary Figs 12 and 13). Note that the fraction of metallic Ga (from Ga 3d spectra) does not remarkably increase with the etching of Ga2O3 skin at outer surface, indicating that the oxidation depth is not very shallow. The increased fraction of Ga2O3−x (compared with Ga2O3, from O 1s spectra) on etched surface agrees with the attenuation or reduction of Ga2O3 during etching process. After lithiation (Supplementary Fig. 13), the LixGaOy fraction increases (compared with Li2O, from O 1s spectra) in the etched sample, indicating the lithiated conversion reaction progresses more sufficiently at the outer region than at the inner region. This phenomenon agrees with the increased fraction of LixGaOy component (compared with LixGa) from Ga 3d spectra. The slight inhomogeneity of lithiation process does not influence the Li-ion transport across LM coating and its oxidation skin. The phase assignment of lithiated LM is further confirmed by X-ray diffraction (XRD) pattern (Supplementary Fig. 14a). The LixGa alloy contains the phases of Li2Ga and Li3Ga2, and LixGaOy is composed of LiGaO2, Li5GaO4, and LiGa5O8 phases. The existence of these phases is consistent with the Li-Ga-O ternary phase diagram (Supplementary Fig. 14b).
Li2CO3-affiliative mechanism via LM painting
Note that the good contact between LLZT and LM can be obtained even without intentional removal of the so-called passivated Li2CO3. There is a speculation that alloyable element (e.g., Zn) may react with Li2CO3 to create more Li vacancies in passivation layer for faster Li-ion transport at the interface zone43. However there is no strong evidence to support this prediction. Herein, we propose a Li2CO3-affiliative mechanism in which Li2CO3 on garnet surface is wetted and downsized by LM drops, and it is torn into Li2CO3 nanodomains by surrounding LM nanoparticles with similar nanoscale. The superior wettability can guarantee a homogeneous mixing of Li2CO3 and LM nanodomains. Therefore the separated Li2CO3 grains cannot form continuous passivation layer to retard Li-ion transport. Instead, the well-dispersed LM grains can serve as ion wires after lithiation to construct the alternative ion channels especially when they penetrate into continuous conductive network. To prove our hypothesis, we ground LM with Li2CO3 or garnet powder (Fig. 3a). Under the mechanical force during the mixing process in air, the shiny LM is pulverized into much smaller particles that blend with white Li2CO3 or light-yellow garnet powder. The resultant darkening stems from the higher-fraction oxidized regions of tearing LM grains by air or oxide powder. In contrast, the pristine LM still maintains its metal luster even after grinding in air in view of the self-limiting surface oxidation. These phenomena indicate the facile dispersion capability of LM when contacting with carbonate or ceramic powder. The strong Van der Waals’ force of self-formed gallium-oxide skin is responsible for its coverage on powder grains38. The XRD pattern of the mixture of LM and Li2CO3 (LM@Li2CO3) does not display the evident peaks ascribed to LM (Supplementary Fig. 15), further indicating the attenuation of LM after dispersing and downsizing. The enrichment of gallium oxide does not cause the appearance of excess diffraction peaks due to its relatively poor crystallinity37,44. The pronounced peaks of Li2CO3 are still preserved after mixing with LM, considering the mechanical robustness of well-crystallized Li2CO3. Similarly, the mixing of LM with LLZT (LM@LLZT) also causes the weakening and elimination of LM diffraction peaks (Supplementary Fig. 16). The cubic phase structure of LLZT is not destroyed after mixing with LM.
To further explore this fascinating wetting phenomenon of LM with Li2CO3 powder, transmission electron microscopy (TEM) was resorted to. Fig. 3b discloses the excellent mutual miscibility between LM and Li2CO3 nanodomains, and there is no serious phase segregation observed. Under high resolution (Fig. 3c, d), we clearly observe the homogenous distribution of LM and Li2CO3 nanoparticles with comparable sizes as small as 3–5 nm. The LM particles are discerned from the darker spots, while the Li2CO3 ones are the crystallized domains with typical lattice fringes corresponding to (311), (020), (−112) planes with d-spacings of 0.187, 0.249, 0.262 nm, respectively (Supplementary Fig. 17). In view of the actual interface with excess LM modification compared with the amount of naturally formed Li2CO3, we also intentionally blended higher-fraction LM with Li2CO3 to see the microscopic distribution of mixture. As shown in Fig. 3e–g, some big drop-like spheres of LM are still residual without undergoing the fragmentation due to the insufficiency of carbonate. However their oxidation surface still has good affiliative ability toward surrounding Li2CO3, leading to the appearance of Li2CO3 moieties anchored on LM sphere surface. A part of Ga2O3 skins are likely peeled off and enter into the Li2CO3 network to form the mutual mixture. This would result in the interaction reinforcement at Li2CO3-LM interface as indicated from the firm attachment of LM to the Li2CO3-covered LLZT surface even after Li plating and stripping. The scanning transmission electron microscopy (STEM) image and corresponding EDS mapping of C, O, Ga (Fig. 3h) disclose the similar spatial distributions of carbonate and LM components, further confirming the uniform blending of both the phases after grinding. This Li2CO3-affiliative mechanism mitigates the passivation effect and reinforces the interconnection of conductive interface (Fig. 3i). The penetration of LM conductive network along with the breaking up of carbonate passivation layer provides an alternative pathway for facile Li-ion transport across Li-LLZT interface without the requirement on removing the passivation layer by harsh conditions17,18.
Raman spectra of garnet pellet, gallium mixed Li2CO3, and gallium mixed garnet powders are represented in Supplementary Fig. 18. The garnet pellet shows the typical characteristic peaks roughly located at 243, 376, 645, and 728 cm−1 (corresponding to instinct signals of cubic LLZT, the former two peaks being ascribed to the Li–O bonding), apart from the strong CO32− vibration peak at 1090 cm−1(ref.17). After mixing with LM, the characteristic peaks of garnet are covered, while the Li2CO3 peak is still maintained. This phenomenon indicates a potential interaction between LM (or its oxide skin) and LLTO surface. Compared with the Raman spectrum of LM mixed Li2CO3, we deduce that the broad peak located at 700 cm−1 should stem from the vibration of Ga–O bonding in both the mixture samples45. Note that the blending of LM cannot reduce or decompose the Li2CO3 component, agreeing with the XRD results mentioned above.
Solid-state cells benefitting from superior LM wettability
To characterize the electrochemical benefit from this strategy, lithium symmetric cells were assembled with the LM-painted LLZT (marked as Li|LM@LLZT|Li) and unmodified garnet (marked as Li|LLZT|Li) as solid electrolytes, respectively. The EIS of Li/Li symmetric cells was performed to evaluate the interfacial situation. The interfacial ASR values of symmetric cells are significantly decreased from 1.75 × 104 to 19.5 Ω cm2 at RT, and from 272 to 5 Ω cm2 at 60 °C after LM painting on both the sides of garnet (Fig. 4a, b)43. The ASR values are estimated based on the equivalent circuit in the insets of Fig. 4a, b, which shows good fitting effect46. Therein Rg, Rint, and Rsurf denote the ASRs for garnet, interface transfer, and surface reaction, respectively. CPEint and CPEsurf denote the constant phase elements paralleled with Rint and Rsurf, respectively. The corresponding characteristic time constants (τ) and capacitance values (C) are listed in Supplementary Table 1. C and τ can be expressed by (R1−nCPE)1/n and RC, where n is CPE exponent. The capacitance values referring to interfacial transport are in the range of 10−9–10−7 F cm−2, while those referring to surface reaction are in the range of 10−7–10−5 F cm−2. The discrepancy of capacitance range depending on different process is in accordance with the precious report by Irvine et al.47 After LM painting, the relaxation time is shortened no matter for the interfacial or surface process, also indicating an improved electrochemical kinetics. Both the ASR values of Li|LM@LLZT|Li cell at RT and 60 °C are smaller than most of reported values even if the garnet pellets we used have been intentionally exposed to air for several days (Supplementary Table 2). These comparisons demonstrate the superiority of this facile painting strategy.
The symmetric Li|LM@LLZT|Li cells can achieve an ultralong Li plating/stripping cycling for at least 9930 h (4965 cycles) at 60 °C with an ultra-small overpotential from −10 to 10 mV, as well as from −33 to 33 mV at RT at a current density of 0.2 mA cm−2 based on an areal capacity of 0.2 mAh cm−2 (Fig. 4c). The symmetric cell at RT can be steadily cycled for at least 2000 h. Note that the Li plating and stripping curves of symmetric cells are quite plat and smooth during the early cycling and after long-term cycling (insets of Fig. 4c), indicating the elimination of nucleation overpotential as a consequence of facile charge transport across lithiated LM. In all-solid-state architecture, there is no extra accumulation of solid electrolyte interface from side reaction to cause larger nucleation resistance and overpotential. The modification of impedance and cycling performance also benefits from the prior Li melting step (230 °C for 5 min) for better interface contact. Note that the symmetric cell even without Li melting pretreatment can also cycle for a long time over 750 h, but at a cost of unstable polarization activation process during early cycling (Supplementary Fig. 19). After activation, the overpotential is quite stable and is still small (from −20 to 20 mV) at 60 °C, benefiting from the construction of mixed conductive network induced by electrochemical lithiation of LM interlayer. In contrast, the symmetric cell with unmodified garnet reaches to the short circuit stage quickly after merely several hours (Supplementary Fig. 20). Before short circuit, the overpotential is also large and asymmetric between plating and stripping (with a gap of ~1.3 V) even at a smaller current density of 0.1 mA cm−2. The appearance of tip phenomenon in voltage profiles is caused by the poor interface contact1,2. For the LM-modified symmetric cell, stable rate performance is recorded with the increase of current density from 0.05 to 1.2 mA cm−2 as shown in Fig. 4d. The corresponding voltage profiles do not undergo serious degradation, and they are roughly flat even at much higher current density exceeding 1 mA cm−2. Even when reaching to 1 mA cm−2, the overpotential is still controlled between −28 and 28 mV (Fig. 4e). This low plateau overpotential does not increase remarkably and can last for at least 150 h during the following cycling. A high critical current density (CCD, defined as the highest applied current density that the solid-state electrolyte can endure the lithium dendrite penetration) value up to 1.7 mA cm−2 can be reached for the LM improved symmetric cell (Fig. 4g), while the CCD value for unmodified cell is much smaller (0.3 mA cm−2) even with the assistance of LE wetting (Fig. 4f). This high CCD value in this work is superior to most of the literature reports (e.g., 0.5 mA cm−2 for Sn film alloying, 0.8 mA cm−2 for H3PO4 modification, 1 mA cm−2 for Li-graphite anode), indicating an excellent defense capability of LM interlayer against Li dendrite growth20,29,48. As suggested by Flatscher et al.46, CCD is highly related to the conditions of interface wetting, pressure, and temperature. The obtained high CCD value also confirms the perfect Li wetting and agrees with the significantly reduced ASR values at the fixed conditions without externally applied pressure and high temperature.
Since the symmetric cell architecture employs excess Li, it is not suitable to evaluate the accurate utilization ratio of anode as well as accumulation degree of dendritic Li or dead Li in GBs. To further explore the dendrite suppression effect modulated by LM painting, we also performed the asymmetric cells, which were not often used to estimate the Li plating behavior. Herein the carbon-coated Al foil (denoted as C@Al) is used as the nonlithium electrode. The asymmetric cell with solid electrolyte painted by LM is denoted as Li|LM@LLZT|C@Al, while the control cell with extra LE dropped on the anode side of garnet is denoted as Li|LE@LLZT|C@Al. Benefiting from the excellent wettability of LM and its dissipation effect on passive Li2CO3, the lithiated interphase layer allows a uniform Li+ flux from bulk electrolyte to anode, thus resulting in a high utilization ratio of Li (Fig. 5a). In contrast, for control cell, the dendrites are expected to initiate at interface and grow inside garnet due to the uneven Li+ flux frustrated by continuous Li2CO3, causing a low utilization ratio of Li even with LE addition (Supplementary Fig. 21). The coulombic efficiencies (CEs) for Li|LM@LLZT|C@Al cell are well stabilized at the high values close to 100% for at least 100 cycles after early activation (Fig. 5b), while the Li|LE@LLZT|C@Al cell can only run for <20 cycles with a smaller CE value of ~90% before failure. The former cell has a much smaller voltage hysteresis (e.g., 70 mV in the 10th cycle) than that (325 mV) for the latter cell in the same cycling stages (Fig. 5c). The voltage hysteresis after LM painting is preserved at a low value without serious fluctuation during the whole cycling process (Fig. 5d). For unmodified cell, the voltage hysteresis increases rapidly after ten cycles. These results further confirm the improved reversibility and kinetics of Li plating through the stable LM interlayer, leading to the alleviations of active Li roughening and dead Li formation.
Interface reactions between lithiated LM components and LLZT were also studied (based on first-principle computations)49,50. The interface is considered as a pseudo-binary structure of lithiated interphase and garnet and the most stable phase equilibria based on thermodynamic driving forces is determined. Lithiated interphase (e.g., main components Li2Ga, Li3Ga2, and LiGaO2) shows better chemical stability with garnet (than Li metal) with a minor reaction energy more than −15 meV/atom (Fig. 5e), which is higher than that of Li-LLZT and much higher than previously reported values (e.g., −40 to −90 meV/atom for Si-garnet, −24.78 to −62 meV/atom for Al-garnet, −20 to −100 meV/atom for LixC-garnet)21,29,51