LLZO fillers with cubic phase were obtained by a solid-state synthesis employing double- doping strategy for high ionic conductivity and short reaction time.27 Resistive Li2CO3 layers were formed at the LLZO surface upon unavoidable exposure to ambient air. Transmission electron microscopy (TEM) visualizes the Li2CO3 layer of ~15 nm thickness (Fig. 1a).39 The poorly crystallized Li2CO3 layers are hardly detectable by X-ray diffraction (XRD) (Fig. S1). Fig. 1b schematically describes our well-refined dry etching process for the resistive layers. Reactive ion etching (RIE) is a reliable dry etching by means of ion bombardment, widely used in the microfabrication.40 Herein, RIE strips off the thin Li2CO3 layers at the powder state LLZO fillers, which is difficult with other methods, such as mechanical polishing.39,41,42 In a RIE process, the degree of anisotropy for etching rate, generally expressed as A = 1-RL/RV (RL : lateral etch rate, RV : vertical etch rate), is expected to be 1 for an ideal vertical etching.43 Practical RIE cannot feature such an ideal case and sidewall passivation is commonly used to prevent the undercut beneath the etching mask in microelectromechanical system (MEMS) fabrication.44 In addition, the typical etching rate with fluorocarbon gases such as CF4 is significantly higher for the amorphous surface layer rather than crystalline core.45 Such a non-ideal directionality of RIE along with high etching selectivity could lead to a conformal etching of Li2CO3 layers. Previously, simple thermal treatment has been exploited to remove the Li2CO3 layers.46 Nonetheless, our etching approach can be generally extended to the unwanted surface layers of other materials even with high thermal vulnerability. We used CF4 RIE for the selective etching of amorphous surface layers. The crystalline structure of LLZO core was well-maintained after surface etching (Fig. S1).
X-ray photoelectron spectroscopy (XPS) was carried out to confirm the surface chemistry of LLZO fillers. Two distinctive peaks were detected in C 1s scan around 285 and 290 eV for adventitious carbon and carbonate, respectively (Fig. 1c).32,41 The carbonate peak diminished with the removal of Li2CO3 layers. In O 1s spectra, the carbonate peak at ~532.1 eV decreased with etching, while the peak for crystalline LLZO increased at ~528.5 eV.34 Similarly, Li 1s spectra consist of the two principal peaks at ~55.4 and 54.8 eV for Li2CO3 and Li-O, respectively (Fig. 1e). Along with etching time, the intensity of Li2CO3 peak decreases and the overall Li 1s peak shifts from ~55.1 to ~54.8 eV, supporting the removal of Li2CO3. There is no significant difference in the 15 and 20 min etching, implying that 15 min is sufficient to remove the majority of Li2CO3 layers (Fig. S2).
Fig. 1f displays the prominent change of Zr 3d signal at LLZO upon surface etching. A weaker Zr 3d signal was observed at non-etched LLZO, which is quite plausible considering the presence of Li2CO3 layer and the detection limit of ~10 nm for XPS. After etching, the intensity of Zr 3d signal strengthened. Quantitative analysis for the etching profile is implemented by plotting the elemental ratios of C/La and C/Zr as a function of etching time (Fig. 1g). Both ratios reduce rapidly within 10 min and saturates around 15 min (Those ratios are higher than zero throughout the etching process due to adventitious carbon).39,41 No significant difference in the carbon content was detected between 15 min etched sample and others treated with different conditions (Fig. S3). It is also critical to keep the etched samples in an inert atmosphere to avoid the re-growth of surface layer (Fig. S4). Energy dispersive X-ray spectroscopy (EDS) mapping could directly visualize the surface selective etching of Li2CO3 layers (Fig. 1f and Fig S5). While the C signal became blurred after 15 min etching, no significant compositional change was detected in the crystalline LLZO core, as also supported by Raman spectroscopy (Fig. S6).
The typical properties of HSEs strongly depend on the size, morphology and content of inorganic fillers.47 Here, planetary ball milling was used for the size reduction of LLZO fillers, considering its mechanical simplicity and potential large-scale production.48 Fig. 2a present SEM images of the as-synthesized (referred to as asLLZO) and ball-milled (referred to as bmLLZO) LLZO, respectively. Compared with asLLZO (12.97 ± 4 um), bmLLZO shows a considerably decreased size (1.2 ± 0.5 um) (Fig. 2b). XRD patterns of both samples showed identical profiles, implying the well-maintained crystalline structures during the size reduction process (Fig. S7). Next, free-standing films of PVDF-LLZO HSE and the PVDF based solid polymer electrolyte (SPE) were prepared by solution-casting (Fig. S8). XRD patterns of LLZO, pure PVDF, PVDF-SPE and PVDF-LLZO HSEs taken at room temperature are displayed in Fig. 2c and Fig. S9. The diffraction peaks for LiClO4 salt is absent in the PVDF-SPE, indicating a successful ionic complexation.37 Upon hybridization, the crystalline phase of LLZO does not change. Interestingly, the weak crystallinity of PVDF-SPE becomes slightly less apparent in the HSEs loaded with bmLLZO or etched bmLLZO compared to those with asLLZO. The larger PVDF/LLZO interface area with the smaller-size ball-milled LLZO fillers should contribute to the further broadening of PVDF-SPE peaks, also evidenced by the planar elemental mapping (Fig. S10). The surface etching of Li2CO3 layers caused no noticeable difference again in the XRD patterns of HSE-bmLLZO30 (30 wt% bmLLZO in HSE) and HSE-etched bmLLZO30 (Fig. 2c).
While PVDF-SPE film is transparent, HSE films show a dramatic variation in the color, depending on the dimension, composition and surface chemistry of loaded fillers (Fig. 2d and Fig. S11). The interaction between N,N-Dimethylformamide (DMF) solvent and LLZO is known to provide an alkaline-like environment.37,49 This can induce the dehydrofluorination of PVDF and thus HSE-bmLLZO30 turns dark brown. By contrast, HSE-etched bmLLZO30 exhibits light brown since a less alkaline-like condition with etched LLZO might alleviate the degradation of PVDF. Furthermore, the good mechanical flexibility is illustrated by the easy deformation of HSE-etched bmLLZO30 film.
For the optimal choice of LLZO fillers, HSE films were prepared with various filler contents of 0–50 wt%. Electrochemical impedance spectroscopy (EIS) of all films presents a similar trend that the ionic conductivity maximizes at 30 wt% content and declines above (Fig. 2e).37,47 The HSE-etched bmLLZO30 displays the maximum ionic conductivity of 4.05 x 10–4 S cm–1 approximately two times higher than HSE-bmLLZO30 (2.12 x10–4 S cm–1) at room temperature. Temperature-dependent conductivities and activation energies are compared where HSE- etched bmLLZO30 shows the lowest activation energy of 0.31 eV (Fig. 2f and Table S1).
Specific effect from eliminating Li2CO3 layers on the electrochemical performance has been investigated in a Li symmetric cell structure and SSBs with the HSEs including 30 wt% LLZO fillers. All the cells were tested at room temperature. Voltage-time profiles of the cells with PVDF-SPE and HSE-asLLZO30 present stable cycling at a low current density of 0.05 mA cm- 2 (Fig. 3a). However, together with the increase of current density, the overpotential abruptly increased and eventually the cell failure occurred by exceeding a pre-set voltage limit (after 168 h for PVDF-SPE, after 244 h for HSE-asLLZO30), which can be ascribed to the large internal resistance. In HSE-asLLZO30, the relatively small PVDF/asLLZO interfacial area causes a high overall cell resistance, possibly by the dendritic grow of Li metal during cycling. Fig. 3b and c presents the cycling results for HSE-bmLLZO30 and HSE-etched bmLLZO30 cells from 0.05 to 0.15 mA cm–2. Both cells displayed similar trends up to 0.10 mA cm–2 in terms of the shape of voltage plateau and cycling stability (Fig. S12). HSE-etched bmLLZO30 maintained the stable cycling with a low overpotential (0.045 mV) and flat voltage plateau at 0.15 mA cm–2, whereas HSE-bmLLZO30 suffered from the increase of overpotential from 0.072 to 0.135 mV along with the irregular shape of the voltage hysteresis. The interfacial resistance of HSE-bmLLZO30 and HSE-etched bmLLZO30 was characterized by EIS upon Li stripping/plating test. (Fig. S13). As summarized in Fig. 3d, the interfacial areal specific resistance (ASR) shows much milder increase in HSE-etched bmLLZO30 (from 72.6 to 110 Ω cm2) compared to HSE-bmLLZO30 (78 to 232.5 Ω cm2). This further supports that removal of the Li2CO3 layers can boost up Li-ion transport and ultimately suppress the increase of cell resistance.
For the demonstration of practical cell performance, SSBs based on the configuration of LiNi0.6Mn0.2Co0.2/LLZO-PVDF HSEs/Li were assembled and cycled at room temperature. The rate performances of LiNi0.6Mn0.2Co0.2/HSE-bmLLZO30/Li and LiNi0.6Mn0.2Co0.2/HSE- etched bmLLZO30/Li are compared in Fig. 3e. There is a similar level of discharge capacity for both samples at low current density (≤ 0.5 C, 1 C = 170 mA g–1). However, a higher current density reveals the superior cycling profile of the SSB with HSE-etched bmLLZO30. Noticeably, SSB with HSE-etched bmLLZO30 exhibits a high discharge capacity of ~ 79 mA h g–1 at 4 C, which is three times higher than the one with HSE-bmLLZO30 (~26 mA h g–1) as well as comparable to or even higher than the previous conventional batteries based on LiNi0.6Mn0.2Co0.2 cathode with liquid electrolyte50 or sulfide-based electrolyte.51 This outstanding rate performance is the highest value among HSE-based SSBs employing the high capacity cathode materials such as LiNi0.6Mn0.2Co0.2 or LiFePO4 reported to date (Fig. 3f and Table S2). The selected charge/discharge curves for each rate step are compared for clarification (Fig. S14). The voltage-capacity profiles during the long cycling are displayed in the inset in Fig. 3e. SSB with HSE-etched bmLLZO30 exhibits a more stable cycling stability than HSE-bmLLZO30. In addition, cycling test was also conducted at a low temperature of 10 ℃ (Fig. S15). The SSB with HSE-etched bmLLZO30 still exhibited the high capacity of 167 and 163 mA h g–1 at 0.1 and 0.2 C, respectively, while the SSB with non-etched LLZO delivered relatively inferior capacity (155 and 133 mA h g–1 at 0.1 and 0.2 C).
Underlying mechanism for the promoted performance of the HSEs with etched LLZO fillers could be characterized with solid-state magic angle spinning (MAS) 7Li nuclear magnetic resonance (NMR). The plausible route for Li-ion migration within HSE volume consists of PVDF polymer matrix, LLZO fillers and/or the interface between them. From 7Li NMR profiles, the resonance peaks for LLZO fillers and PVDF-SPE were observed at 0.36 and –0.63 ppm, respectively (Fig. 4a). The bmLLZO shows a slightly broad and weak profile compared to the etched counterpart, indicating the influence from surface barrier layer, resistant to mobile Li-ion.52 Narrow and intense peak for PVDF-SPE originates from LiClO4 salt, homogeneously dissolved and complexed within the polymeric matrix. Non-blocking Li symmetric cells of 6Li/HSE-bmLLZO30 or HSE-etched bmLLZO30/6Li were assembled and cycled repeatedly at 10 μA cm–2 for every 5 min. 6Li from an electrode passes through HSE and reaches the opposite electrode while partly replacing 7Li in the preferable pathway. Monitoring of the residual 7Li in HSE can trace the possible Li-ion pathways. Noticeably, dissimilar NMR signals were detected in those samples after cycling, suggesting distinct Li-ion migration behaviors. (Fig. S16) Further, the NMR spectra can be deconvoluted to reveal a new peak at –0.28 ppm in addition to the two main signals, which is attributed to locally modified environment at LLZO and PVDF-SPE interface (Fig. 4b and Fig. S17).53,54
Preferable Li-ion pathway was evaluated from the integral area of deconvoluted NMR spectra, as plotted in Fig. 4c. Significant reductions in 7Li were observed in the PVDF-SPE (from 30.6 to ~ 16.5 %) and interface (from 24.5 to 7.9 %) particularly for HSE-etched bmLLZO30. These results suggest the preferable Li-ion pathways through the polymer matrix and interface. Obviously, the interface seems more dominant as follows. First, LLZO fillers are known to modify the local surrounding environment by increasing mobile Li-ions.55–57 Second, uniform dispersion of LLZO fillers increases the total interfacial area in HSE, establishing abundant routes for ion transport, hardly achieved with a lower (< 20 wt%) or higher (> 40 wt%) filler loading (see Fig. 2e). Noteworthy that severe agglomeration of LLZO fillers could occur at the higher loading and result in a less generation of active interfaces.37,47
Fig. 4d displays the imaginary part of impedance plotted against frequency, known as a Debye plot together with a fitting of a Lorentzian function, where the relaxation time is related to ion hopping within a conductive medium. The characteristic frequency at the Debye peak maxima can be determined by the reciprocal of conductivity relaxation time (τ) or conductivity (σ) according to the following equation.58,59
2𝜋𝑓𝑚𝑎𝑥 = 𝜔 = (𝜏)−1 = 𝜎(𝑒0𝜀′)−1 (1)
Where 𝑒0 is the permittivity of free space (8.854 x 10–14 F cm–1) and 𝜀′ (real component in complex permittivity) is the permittivity independent of frequency. Interestingly, no significant difference in the peak maxima signifies the similar level of ionic mobility for both HSEs.
Consequently, the promoted Li-ion conductivity in HSE-etched bmLLZO30 should be attributed to the higher population of mobile Li-ions particularly around the modified interface. The dielectric constant (relative permittivity) of doped cubic phase LLZO is known to be 40- 60, sufficiently high enough to promote Li-ion dissociation from salt.55 In addition, surface defects or vacancies at etched LLZO surface may complex with Li-ions through Lewis acid- base reaction to further increase the concentration of free Li-ions in HSE.56 Here, the surface Li2CO3 layers with much lower dielectric constant of 4.9 may weaken the dissociation of ions around LLZO.57 Taken together, a proposed scheme is depicted for the ionic conductivity enhancement in Fig. 4e. While Li-ion transport can be facilitated with the enriched Li-ions around LLZO fillers, effective removal of the resistive surface layers can boost up the population of mobile ions and ultimately lead to the improved ionic conductivity and rate performance.