Quasi-solid polymer electrolytes with binary and ternary salt mixtures for high-voltage lithium metal batteries

Materials

PVdF-HFP (experimental copolymer grade) was provided by Arkema and dried under vacuum at 100 °C before use. LiFSI (Foranext®) was provided by Arkema and used without further purification. LiBOB (American Elements, 99.9%) and LiNO₃ (Alfa Aesar, 99%) were dried under vacuum at 60 and 70 °C, overnight, respectively. Methyl ethyl ketone (MEK, anhydrous, 99.5%, Thermo Scientific) and acetone (anhydrous, 99.8%, max., 0.005% H₂O, Thermo Scientific) were used without further purification. PC (99%, < 10 ppm H₂O, Sigma Aldrich) and EC (Battery grade, Fujifilm) were used without further purification. Tetra(ethylene glycol)dimethyl ether (99%, Sigma Aldrich) was dried under vacuum at room temperature and stored on molecular sieves. Spray-dried LiNi_0.8Mn_0.1Co_0.1O₂ (NMC-811), used for the full cell tests in coin cells, was kindly provided by Cerpotech. The NMC-811 used for the cycling tests in pouch cells was purchased from Targray. Li sheets laminated on Cu (Li-Cu) were kindly provided by ABEE.

QSPEs preparation

Unsupported quasi-solid electrolytes (QSPEs) for the preliminary study were processed by solvent casting, in an Ar glovebox, by dissolving the PVdF-HFP in acetone and then by adding a solution of lithium salt in a plasticizer to the PVdF-HFP/acetone solution. The resulting solution was then cast on Mylar® films by a doctor blade and dried at room temperature and reduced pressure (500 mbar) for 16 h. Membranes with thicknesses ranging from 50 to 100 µm and an area of ~200 cm² were produced in this way. With unsupported QSPEs, three plasticizers were tested: tetra(ethylene glycol) dimethyl ether (TEGDME), PC, and EC. LiFSI was used as a lithium salt.

Supported QSPEs were prepared inside a glovebox by solvent casting on a microporous polyolefin separator (Celgard® 2500). MEK was used as a processing solvent. In a typical preparation, the lithium salts (LiFSI, LiBOB, and LiNO₃) were dissolved in EC in the chosen concentration (see Table 1), while 0.4 g of PVdF-HFP was dissolved separately in 6 mL of MEK. Then, 1.6 g of the EC solution was added to the PVdF-HFP/MEK solution, and the resulting mixture was stirred overnight at room temperature. The solution was then cast with a doctor blade on a 10 × 10 cm² Celgard® 2500 sheet, which was placed previously onto a poly(ethylene terephthalate) (Mylar®) foil substrate. The casting was conducted at a speed of 1 mm s^-1 and with a blade gap of 300 µm. The casting solvent was finally removed by keeping the membrane under Ar flow overnight at room temperature. Three QSPE compositions were prepared, with LiFSI only, binary LiFSI/LiBOB mixture (in a 4:1 molar ratio), and ternary LiFSI/LiBOB/LiNO₃ mixture (in a 4:1:1 molar ratio). The composition of the prepared QSPEs is reported in Table 1.

QSPE characterization

Thermogravimetric analysis (TGA) was performed with a TGA 209 F1 Libra analyzer (Netzsch). The temperature scans were performed from room temperature up to 600 °C, at a heating rate of 10 °C min^-1, and under argon (60 mL min^-1). Differential scanning calorimetry (DSC) was performed with a DSC 2500 differential calorimeter (TA Instruments). The scans were conducted in the temperature range between -80 and 100 °C, with a heating rate of 2 °C min^-1. The samples were cycled twice between -80 and 100 °C, and the second heating scan was used for the analysis. For the sample preparation, 5-10 mg of each sample was placed in sealed aluminum pans under argon atmosphere.

Ionic conductivity was determined by electrochemical impedance spectroscopy (EIS). The EIS spectra were collected with a Solartron 1260A Impedance/Gain-Phase Analyzer in the frequency range between 32 MHz and 1 Hz (20 points per decade), with a signal amplitude of 10 mV, and in the temperature range between 20 and 80 °C (with 10 °C step). Additionally, the ionic conductivity was measured at 25 °C. For the conductivity measurements, the membranes were placed in coin cells CR2032, with three stainless-steel (SS) plates of 0.5 mm thickness. The temperature was controlled with a Binder KB23 cooling incubator. The ionic conductivity was calculated using:

where σ is the ionic conductivity, R is the measured resistance, A is the electrode surface area, and L is the membrane thickness. The latter was measured with a digital micrometer after the experiment. The measurement was repeated on three different cells for each composition, and the values of resistance and conductivity used in the analysis are the average values of the three measurements.

The morphology of a QSPE cross-section, obtained by an ion milling technique (Hitachi 4000 Plus), was analyzed by field emission scanning electron microscopy (FESEM, APREO 2 S HiVac FESEM). QSPE samples were transferred between the glovebox, the ion milling and the scanning electron microscope (SEM), under argon atmosphere, with a transfer device. Cross-sectioning was performed at -70 °C with a cryogenic module.

The lithium transference number (T⁺) was measured on a VMP3 potentiostat (BioLogic), in Li||Li coin cells, at 25 °C. The lithium T⁺ was determined by combining a chronoamperometry step with the measurement of the impedance spectra, collected before and after the chronoamperometry step. For the chronoamperometry, a constant voltage of ± 10 mV was applied for a duration of 20 min, and the resulting current was registered with a frequency of 100 Hz during the 30 s, and 1 Hz during the rest of the chronoamperometry. The impedance spectra were collected in the frequency range between 1 MHz and 100 mHz (20 points per decade) and with potential amplitude of 10 mV. For each composition, the experiment was repeated on three cells and six times on each cell, by alternating positive and negative potentiostatic polarization. The cells were allowed to relax for one hour after each chronoamperometric step, and for 10 min after the EIS measurement. The T⁺ was then calculated using^[55]:

where I_SS is the initial and steady-state current, ΔV is the applied potential, R_int,SS is the interfacial resistance after the chronoamperometric step, and R_b,0 is the bulk resistance before the chronoamperometry.

The oxidative stability of the three QSPEs was studied, by linear sweep voltammetry (LSV) and floating test, in coin cells with carbon-coated Al foil discs (cc-Al) as working electrodes, and lithium metal electrodes as counter/reference electrodes. LSV scans were carried out at an ultra-low scan rate of 5 µV s^-1 between the open circuit potential and 5.0 V vs. Li/Li⁺. Floating current tests were conducted in the same cell configuration. First, the cell potential was increased to 3.9 V vs. Li/Li⁺ at 0.1 mV s^-1. Then, the potential was increased stepwise up to 5.0 V vs. Li/Li⁺, with step amplitude of 100 mV and step duration of 10 h.

Cathode active material preparation

NMC-811 powder from Cerpotech was produced by spray pyrolysis. In a typical synthesis, an aqueous solution was prepared by dissolving stoichiometric amounts of metal nitrates in distilled water and mixing individual precursor solutions under stirring. The homogeneous solution was atomized by a two-phase nozzle (pressurized air + solution) into a pre-heated rotary furnace (≈2 rpm) (Entech Energiteknik AB) under constant air flow, yielding an approximate average residence time of 2 s at 700-1,000 °C. This caused instant vaporization of water and onset of nitrate decomposition, as mixed metal oxide began to form. The collected powders were calcined in alumina crucibles in a tube furnace (Entech) at 850 °C for 6 h under a constant oxygen flow (Praxair 5.0) with heating and cooling rates of 200 °C h^-1.

Cathode preparation

The N-Methyl-2-pyrrolidone (Merck)-based slurry, composed of 81.3 wt% of NMC-811 (Cerpotech), 4.5 wt% of C45 carbon black (Imerys) and 14.2 wt% of catholyte, was prepared using a mechanical stirrer (IKA). The catholyte comprised a mixture of PVdF-HFP (Arkema), 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)imide (PYR14TFSI, 99.9%, Solvionic), and LiFSI (Arkema), to promote Li⁺ ionic conductivity within cathode active layer and simultaneously maintain mechanical integrity of the positive electrode. The homogenized slurry was cast over a carbon-coated aluminum current collector with a thickness of 20 µm by an automatic film applicator 1137 (SHEEN Instruments). The electrode was dried in a convection oven at 110 °C for 15 min. The dried electrode with a final loading of 13 mg_NMC-811 cm^-2 was compressed by a laboratory hydraulic roll calender (DPM Solutions, Canada) to 3.0 ± 0.1 g cm^-3 density and finally dried under dynamic vacuum at 120 °C for 16 h in the dry room. NMC-811||Li-Cu coin cells were fabricated with the spray-dried NMC-811 cathode active material.

Owing to the limited availability of the spray-dried active material, NMC-811||Li-Cu single-layer pouch cells were fabricated with commercial NMC-811 (T81RX, Targray). Firstly, a 1 kg slurry batch with the abovementioned formulation was prepared in a dry room (dew point of -40 °C) using a mechanical mixer. Then, the one-sided positive electrode coating of a carbon-coated 20 µm aluminum current collector (Gelon) was carried out using a modified base-coater pilot narrow width machine (up to 300 mm) with a dryer comprising three separated 1-meter-long zones (COATEMA). Finally, the cathode application with confirmed loading level (2.5 mAh cm^-2, 13.8 mg_NMC-811 cm^-2) was calendered at room temperature until reaching an electrode density of 3.0 ± 0.1 g cm^-3 using roll press LDHY400-N45 (Naknor). Before pouch cell assembly, notched cathodes with dimensions of 50 × 60 mm² were dried under dynamic vacuum at 120 °C for 16 h in a dry room (dew point of -50 °C). Lithium metal anodes with a thickness of 50 µm (Albemarle) and dimensions of 50 × 60 mm² were manually cut inside a dry room (dew point of -50 °C) right before pouch cell assembly.

Cell assembly and testing

All coin cells (Hohsen Corp.) were assembled in an Ar.-filled glovebox (MBraun) with H₂O and O₂ concentrations below 1 ppm, and with a manual coin cell crimper (Hohsen Corp.).

The plating/stripping performance of QSPEs was tested at 25 °C, in Li||Li symmetric coin cells with Li electrodes supported on Cu sheets (14 mm-diameter, ABEE). The cells were cycled galvanostatically with a Neware battery tester at 25 °C and at a fixed capacity of 2 mAh cm^-2. Five cycles were performed at 0.2 mA cm^-2 (C/10), followed by 5 cycles at 0.4 mA cm^-2 (C/5), 0.67 mA cm^-2 (C/3), 1 mA cm^-2 (C/2) and 2 mA cm^-2 (C/1). Thereafter, the cells were cycled at C/10 until failure.

Plating/stripping coulombic efficiency was determined by cycling Cu||Li cells, with 500 µm-thick Li discs (14 mm diameter) as counter electrodes and Cu discs (16 mm diameter) as working electrodes. Thick Li foil was used to avoid any effect due to depletion of the negative electrode, as the study of the deposition and stripping process on the Cu foil was the target of this test. Tests were conducted with a Neware battery tester at 25 °C. Firstly, a current of -0.1 mA cm^-2 was applied for 20 h, followed by application of a positive current with the same amplitude, until a cutoff voltage of 0.2 V was reached. This full plating/stripping cycle was performed twice and served as a conditioning cycle, to establish a stable SEI on the Cu foil surface. Then, a plating half cycle was performed, with the same conditions described above, with the aim of creating a Li reservoir on the Cu foil. Thus, this Li reservoir corresponds to an area capacity of 2 mAh cm^-2. Thereafter, fifty short plating/stripping cycles were performed, with a current density of 0.1 mA cm^-2 and a half cycle time of 2 h. In these short cycles, a capacity of 0.2 mAh cm^-2 is cycled back and forth. Finally, a positive current of 0.1 mA cm^-2 was applied, until the cutoff voltage of +0.2 V was reached. The coulombic efficiency of the first 2 cycles was calculated by the ratio of the stripping capacity to the fixed plated capacity. In addition, the average coulombic efficiency during the intermediate short cycles was calculated by^[56]:

where n is the number of short cycles performed, Q_c is the capacity cycled in the short cycles (0.2 mAh cm^-2),Q_S is the capacity of the final stripping half cycle, and Q_T is the total plated capacity in the Li reservoir (2 mAh cm^-2).

NMC-811||Li-Cu coin cells were prepared with NMC-811 electrode discs (12 mm diameter) and Li-Cu discs (14 mm diameter) as positive and negative electrodes, respectively. The top side of the QSPE during casting faced the positive electrode in the full cells. Galvanostatic C-rate and cyclability tests of NMC-811||Li-Cu coin cells were conducted with a Neware battery tester. During the C-rate capability test, the cells were cycled between 3.0 and 4.3 V. Two cycles were performed at C/20 (calculated on a nominal capacity of 170 mAh g_NMC-811^-1), followed by 5 cycles at each of the following C-rates: C/10, C/5, C/3, C/2, and C/1. Thereafter, the cells were further cycled at C/10 until failure. Cyclability tests and EIS were carried out in coin cells with a BT-Lab battery tester. In these tests, the cells were cycled between 3.0 and 4.3 V at C/20 (2 cycles) and at C/10 (100 cycles). EIS spectra were collected at the end of discharge (EoD) every 5 cycles. EIS tests upon cycling in three-electrode cell configuration were carried out with ECC-PAT-core cells (EL-CELL®), with a VMP3 potentiostat (BioLogic). Cells were cycled between 3.0 and 4.3 V at C/50 (2 cycles) and C/20 (30 cycles). EIS spectra were collected both at the end of charge (EoC) and discharge. All full cell tests were performed at 25 °C.

Single-layer pouch cells were manually assembled by stacking alternate layers of the positive electrode, QSPE film and negative electrode. When the cell stack was assembled, nickel and aluminum tabs (MTI Corp.) were ultrasonically welded to negative and positive electrode tabs, respectively. Finally, the solid-state battery was sealed under vacuum in an aluminum laminated pouch bag. The entire assembly was conducted in a dry room with a dew point of -50 °C. Before cycling, the freshly fabricated solid-state battery was placed between two metal plates, clamped with a torque of 0.3 N m^-1, and kept for three hours at 25 °C to ensure the formation of intimate contacts between both electrodes and the solid electrolyte. Finally, the assembled solid-state pouch cells were cycled at 25 ± 1 °C by BaSyTec cell test system, in accordance with the protocol described in Table 2.

X-ray photoelectron spectroscopy analysis

The surface chemistry of Li deposits on Cu under striping/platting experiments was analyzed by X-ray photoelectron spectroscopy (XPS) using a Phoibos 150 XPS spectrometer (SPECS GmbH) installed in an ultrahigh vacuum (UHV) chamber with a base pressure below 5 × 10^-10 mbar. Samples were transferred in an argon-filled air-tight container from the glove box to the instrument after cell disassembly. Spectra were collected in Fixed Analyzer Transmission mode with ~2 mm field of view on the sample using a non-monochromatic Mg anode source (hν = 1253.6 eV). E_step = 0.5 eV and E_pass = 90 eV were employed for the survey spectra and E_step = 0.1 eV and E_pass = 30 eV for the high-resolution regions. The atomic surface species were quantified using tabulated Scofield cross sections, applying corrections to the integrated intensity considering the energy-dependent analyzer transmission and the variations in the effective attenuation lengths (EAL) of the collected photoelectrons depending on their kinetic energies using the analytical expression proposed by Seah^[57]. Data analysis was carried out using CASA XPS software^[58]. The binding energy scale was calibrated by referencing the aliphatic carbon at 285 eV. The inelastically scattered photoelectron background was simulated by a Shirley function, and Voigt profiles (30% Gaussian, 70% Lorentzian) were employed as line shapes for the photoelectron peaks.

QSPE characterization

The chosen composition of the QSPEs results from previous optimization work, aimed at achieving the best compromise between mechanical and transport properties, oxidative stability, and plating stripping performance, to allow room-temperature operation in HV-LMBs. The PVdF-HFP content was optimized to give the highest ionic conductivity while retaining mechanical integrity and was finally set to 20 wt%. At lower PVdF-HFP contents, no self-standing membranes could be obtained. Initially, unsupported membranes were prepared and tested. For these initial trials, acetone was used as a processing solvent. Several lithium salt solvents (plasticizers) were tested, including PC, tetra(ethylene glycol)dimethyl ether, and EC, among which EC gave the best results in terms of ionic conductivity [Supplementary Figure 1] and plating/stripping performance in Li||Li symmetric cells [Supplementary Figure 2]. When tested with thin Li-Cu electrodes, the membranes allowed plating/stripping up to a capacity of 1 mAh cm^-2[Supplementary Figure 3] but underwent rapid short circuit after a few cycles at a plated/stripped capacity of 2 mAh cm^-2 and at a C-rate of C/10 (0.2 mA cm^-2, results not shown). Consequently, a microporous separator was added to increase the resistance against dendrites penetration and to allow cycling in full cells at high cathode active material loading (2-3 mAh cm^-2), necessary to achieve a high energy density at the cell level. The focus of this study is on further optimizing the QSPE composition and examining the effect of the salt mixture composition, particularly LiNO₃, on cell performance. Three supported QSPEs were prepared and characterized, containing LiFSI as a unique salt (QSPE-1), a binary mixture of LiFSI and LiBOB in a 4:1 molar ratio (QSPE-2) and a ternary salt mixture of LiFSI, LiBOB, and LiNO₃, in a 4:1:1 molar ratio (QSPE-3). The exact composition of the three QSPEs is reported in Table 1. The QSPEs were prepared by solvent casting on a microporous polyolefin separator (Celgard 2500), resulting in homogeneous, translucent membranes with a thickness of ca. 60-80 µm. Having a high thickness was necessary to ensure complete penetration of the QSPE into the separator and, additionally, to ensure good contact with both the anode and the cathode. The excess QSPE on the cathode side may help compensate for possible residual porosity in the catholyte phase and thus reduce the interfacial resistance between the cathode layer and the QSPE. The relatively high thickness of the supported QSPEs might obviously affect the cell energy density. However, the QSPE thickness might be reduced by optimizing the casting conditions or using polyolefin separators with a reduced thickness. For instance, by using a (commercially available) 15 µm thick separator, final thicknesses below 50 µm may be easily achieved.

The preparation process of the supported QSPEs was slightly modified to make it more compatible with a possible scale-up to a continuous roll-to-roll process. First, the processing solvent was changed, from acetone to MEK, which has a lower evaporation rate. Additionally, the drying method was altered from drying under static at 500 mbar at room temperature to drying under Ar flow. As in the previous case, this change was necessary to replicate the processing conditions in a roll-to-roll coating line, where vacuum cannot be applied.

Cross-section SEM images of a QSPE membrane [Supplementary Figure 4] show the good adhesion of the overlaying QSPE layer on the infiltrated microporous separator, while porosity formation was observed during SEM acquisition, owing to EC evaporation under high vacuum. The presence of residual MEK was controlled by Fourier transform infrared spectroscopy (FTIR) and nuclear magnetic resonance (NMR). No residual MEK was detected by either NMR or FTIR. In the proton NMR spectra [Figure 1A], MEK gives an intense signal at 2.1 ppm, which was not detected in the spectra of the QSPEs. The most intense peak in the QSPE spectra is due to EC at ca. 5 ppm. FTIR spectra of QSPE-2, processed either with acetone or MEK, are reported in Figure 1B. The spectra are similar, and most importantly, the intense C=O stretching peak of MEK at 1,718 cm^-1 was not observed, confirming absence of residual process solvents. In the carbonyl stretching region, only the doublet related to EC, at 1,774 and 1,801 cm^-1, is visible in both spectra.

Figure 1. Chemical, thermal and transport characterization of supported QSPEs: (A) ¹H NMR spectra of the three QSPEs in DMSO, together with the ¹H NMR spectra of ethylene carbonate (EC), and methylethylketone (MEK); (B) FTIR spectra of QSPE-2, processed in acetone and MEK, between 400 and 2,000 cm^-1; (C) thermogravimetric profiles, collected under Ar atmosphere, at a heating rate of 10 °C min^-1, between room temperature and 1,000 °C; (D) Derivative TGA profiles, up to 600 °C; (E) DSC profiles, collected at a heating rate of 2 °C min^-1, between -80 and 100 °C (cycle 2); (F) Ionic conductivity vs. temperature, between 20 and 80 °C (average of three cells).

The thermal properties of the three QSPEs were analyzed by TGA and DSC. The TGA profiles [Figure 1C and D] show a slow mass loss up to at least 100 °C, attributed to EC evaporation, followed by a rapid mass loss at ca.150 °C, which is caused by faster EC evaporation and possibly salt degradation. A final mass loss is observed at 500 °C, corresponding to the thermal degradation of the polymer matrix. The initial mass loss is slightly faster for QSPE-1, for which the 95 wt% threshold is reached at 97 °C. For comparison, the same threshold is reached at 137 and 122 °C in QSPE-2 and QSPE-3, respectively. The mass loss of the three QSPEs at 300 °C corresponds approximately to the content of EC, namely 60 wt%. The DSC profiles [Figure 1E] show some differences between the three QSPEs. The QSPE-1 has a glass transition temperature (T_g) of ca. -72 °C and two melting transitions, at -25 and 22 °C, preceded by a weak crystallization at -50 °C, with enthalpy of ca. 1 J g^-1. The SPEs with binary and ternary salt mixtures, QSPE-2 and QSPE-3, exhibit T_g of -67 and -62 °C, respectively. Both show a melting transition at 20 °C, which is attributed to the melting of the EC. The different crystallization/melting behaviors with the binary and ternary salt mixtures possibly arise from the increased disorder, caused by the presence of various anions, whereas the increase of the T_g might indicate a lower mobility of BOB^- and NO₃^-, with respect to FSI^-. Nonetheless, the ionic conductivity of the three QSPEs is remarkably similar, slightly higher than 1 mS cm^-1 at room temperature and ~3 mS cm^-1 at 80 °C [Figure 1F, Table 3]. Interestingly, the ionic conductivity is close to that of unsupported QSPEs [Supplementary Figure 1B], although the addition of the separator, with a 55% porosity, was expected to lower the ionic conductivity. This is probably due to the large thickness of the QSPEs, dampening the effect of the microporous separator, which has a thickness of 25 µm. Altogether, this conductivity is expected to be high enough to allow cell cycling at room temperature. The lithium-ion T⁺ at room temperature was determined by the potentiostatic polarization method and is also reported in Table 3. T⁺ is close to 0.1 and increases slightly from QSPE-1 (0.09 ± 0.01) to QSPE-3 (0.14 ± 0.01). The low value of T⁺ indicates that most of the ionic conductivity is due to the motion of the anions and is attributed to the strong coordination of Li⁺ by the EC molecules. The slight increase observed from QSPE-1 to QSPE-3 is attributed to the partial substitution of FSI^- with less mobile BOB^- and NO₃^- anions. Some representative chronoamperometric profiles and correspondent EIS spectra are reported in Supplementary Figure 5. In the chronoamperometry profiles, the steady state is reached within a few seconds (less than one minute in most cases), owing to the high ionic conductivity of the QSPEs. Additionally, the EIS spectra were used to estimate the Li|QSPE area-normalized interface resistance (R_int^*, Table 3). This was calculated by multiplying R_int by the electrode area and dividing the resulting value by two (thus accounting for the presence of two Li electrodes). In this case, the effect of the salt mixture composition is significant, with R_int^* decreasing from 240 ± 50 to 107 ± 5 Ω cm² for QSPE-1 and QSPE-3, respectively. The less resistive interface and higher T⁺ of QSPE-3 result in a correspondingly higher steady-state current [Table 3] and may result in better Li plating/stripping performance for this electrolyte.

The oxidative stability of QSPEs was analyzed by LSV and floating tests, as described in the experimental section. The use of cc-Al working electrodes allowed studying the oxidative stability of the SPEs in conditions resembling the operational cell conditions, as cc-Al is used as a current collector for the NMC-811 cathodes. A higher oxidative current and lower oxidative stability were expected due to larger electrochemical surface area and possible catalytic activity of carbon coating. Furthermore, possible corrosion phenomena involving the Al foil could contribute to the parasitic current at high voltages. The LSV profiles are shown in Figure 2. LSV was conducted at an ultra-low scan rate of 5 µV s^-1 to mimic the residence time in the floating current test and avoid any capacitive contribution. The oxidative stability limit for QSPE-1 is quite low, 4.2 V vs. Li/Li⁺. LSV conducted with SS working electrodes on unsupported QSPEs showed a slightly higher oxidation potential of 4.3 V vs. Li/Li⁺ [Supplementary Figure 6]. Addition of LiBOB (QSPE-2) results in an increase of the oxidative stability, with a current density of 0.5 µA cm^-2 at 4.4 V vs. Li/Li⁺ and a marked increase from 4.5 V vs. Li/Li⁺. A similar result was also observed in SS||Li cells, in which the electrolyte containing a mixture of LiFSI and LiBOB showed much higher stability than the electrolyte containing LiFSI only. QSPE-3 shows intermediate stability, with the current increasing slowly from 4.3 V vs. Li/Li⁺ and a steep increase from 4.4 V vs. Li/Li⁺. After the oxidation onset, the current increases much more steeply with the LiNO₃-containing QSPE-3, indicating that this salt may be deleterious for the oxidative stability of the electrolyte. LiBOB, on the other hand, is confirmed as a fundamental CEI-forming additive for stabilizing the electrolyte at high voltages^[1,38]. Analogue results were obtained also in the floating current tests [Figure 2B]. QSPE-1 showed a steep current increase already at 4.3 V vs. Li/Li⁺, while QSPE-2 and QSPE-3 were stable up to at least 4.4 V vs. Li/Li⁺. Additionally, in this case, the current increase was much steeper for QSPE-3 than for QSPE-2, indicating faster oxidation or current collector corrosion for the LiNO₃-containing electrolyte.

Figure 2. (A) LSV profiles of cc-Al||Li cells with different QSPEs, at 25 °C and 5 µV s^-1; (B) floating current profiles of cc-Al||Li cells, with different QSPEs, collected at 25 °C and with step time of 10 h every 100 mV, between 3.8 and 5.0 V vs. Li/Li⁺.

Plating-stripping performance

The plating/stripping performance of the QSPEs was evaluated by galvanostatic cycling in Li||Li cells at room temperature, with a constant plated/stripped capacity of 2 mAh cm^-2 and current density ranging from 0.2 to 2 mA cm^-2. Constant capacity mode allows a performance comparison among electrolytes at an area capacity comparable to the one of practical full cells. The study focused on QSPE-2 and QSPE-3 due to their higher oxidative stability than QSPE-1. Initially, Li|QSPE-2|Li cells with 500 µm-thick Li electrodes were cycled to evaluate the performance of this QSPE with a large excess of Li.

The voltage vs. time profile of one of the cells cycled is shown in Figure 3A. The cell showed very low polarization even at a C-rate of 1C (2 mA cm^-2) and excellent cyclability at C/10, with stable performance for up to 1,000 h. After 1,000 h, the polarization gradually increased, and the cell underwent a short circuit after 1,500 h cycling. The excellent Li plating/tripping performance, compared to that of unsupported QSPEs at the same capacity [Supplementary Figure 2], confirms the importance of adding a mechanical barrier to the growth of lithium dendrites. Subsequently, cells were also cycled with thin Li-Cu electrodes to assess the effect of the lithium amount. The voltage vs. time profiles of Li-Cu|QSPE-2|Li-Cu and Li-Cu|QSPE-3|Li-Cu are shown in Figure 3B and C, respectively. These cells showed higher overvoltage compared to those with thick Li electrodes, approaching 1 V at 1C. Nonetheless, cycling remained stable, and the two cells shown passed the rate test without experiencing short circuits. However, the occurrence of short circuits was observed on other replicas (results not shown), reflecting the harsh cycling conditions at high C-rates. The low reproducibility of the short circuit occurrence at high C-rates hinders the use of this parameter for the comparison of the different QSPEs. The polarization, on the contrary, is quite reproducible among different cells and is thus a convenient parameter for evaluating the plating/stripping performance. Between the two QSPEs, QSPE-3 showed lower polarization, suggesting a beneficial effect of LiNO₃. A closer examination of the voltage vs. capacity profiles shows that the higher polarization with thin Li-Cu electrodes, compared to the cells with thick Li electrodes, was caused by a voltage spike towards the end of each plating/stripping semi-cycle, which was not observed in cells with thick Li electrodes (compare Figure 3D and E). This voltage spike is attributed to Li depletion in the thin Li-Cu electrodes. On the contrary, the plateau voltage in the middle of each semi-cycle is relatively comparable in all cells. Thus, the average value of the plateau voltage was used to compare the performance of the different QSPEs. The average values of plateau voltage at different C-rates were plotted vs. the current density and interpolated [Figure 3F], and the resulting slope was used to estimate the internal cell resistance in steady-state conditions, that is, before the onset of the additional polarization due to Li depletion. The values for Li-Cu|QSPE-1|Li-Cu cells are also provided for comparison (voltage profiles are shown in Supplementary Figure 7). The internal resistance for Li|QSPE-2|Li cells is 75 Ω cm^-2, lower than the one of Li-Cu|QSPE-2|Li-Cu cells (92 Ω cm^-2). The difference is possibly due to variations in internal cell pressure (despite reducing the number of spacers in cells with thick Li electrodes compared to those with Li-Cu electrodes), or it may result from differences in the wettability of the QSPEs with the two types of Li electrodes. This variation reflects the better cycling performance obtained with thick Li electrodes. Li-Cu|QSPE-1|Li-Cu cells showed a similar resistance of 94 Ω cm^-2, whereas Li-Cu|QSPE-3|Li-Cu cells showed the lowest internal resistance of 53 Ω cm^-2, thus confirming the beneficial effect of LiNO₃ for the plating/stripping process.

Figure 3. Plating/stripping performance of Li||Li cells with supported QSPEs, at room temperature and with plating/stripping capacity of 2 mAh cm^-2. (A) Galvanostatic cycling of a Li|QSPE-2|Li cell, with 500 µm-thick Li electrodes; (B) Galvanostatic cycling of a Li-Cu|QSPE-2|Li-Cu cell, with thin Li-Cu electrodes; (C) Galvanostatic cycling of a Li-Cu|QSPE-3|Li-Cu cell, with thin Li-Cu electrodes; (D) voltage vs. capacity profile at different C-rates of a Li|QSPE-2|Li cell with 500 µm-thick Li electrodes (fifth cycle at each C-rate); (E) voltage vs. capacity profile at different C-rates of a Li-Cu|QSPE-2|Li-Cu cell (fifth cycle at each C-rate); (F) average plateau voltage vs. current density, with different cell configurations. The dashed lines represent the interpolation of the experimental points, and the slope was used to estimate the internal cell resistance in steady-state conditions.

The effect of LiNO₃ on the plating/stripping process was further examined by cycling in Cu||Li cells. The protocol described in the experimental section (Equation 3) was used to calculate the average coulombic efficiency of 50 plating/stripping cycles in Cu||Li cells with QSPE-2 and QSPE-3 as electrolytes. Again, QSPE-3 showed far better performance than QSPE-2, with an average coulombic efficiency of 99.3% ± 0.1%, compared to 95.4% ± 0.7% of QSPE-2 [Supplementary Figure 8 and Supplementary Figure 9]. The results further confirm the positive effect of LiNO₃ on the Li plating/stripping process and suggest that the LiNO₃ salt additive protects the electrolyte from decomposition at the Li interface. To confirm this hypothesis, XPS analysis was performed on Cu electrodes recovered from Cu||Li cells, cycled either with QSPE-2 or QSPE-3. The cells were cycled with a similar profile to the previous cells but were left in the plated state; that is, the last stripping semi-cycle was not performed. For comparison, XPS spectra were also collected on Cu electrodes cycled only for three cycles at a capacity of 2 mAh cm^-2 and at a current density of 0.1 mA cm^-2 (fully stripped Cu electrodes), and on Cu electrodes cycled for two and a half cycles in the same conditions (fully plated Cu electrodes). The three cycling profiles are shown in Supplementary Figure 10. The aim was to study the evolution of the SEI throughout the plating/stripping process and its long cycling stability. Although the SEI formed on Cu might differ from the one formed on Li^[59], this comparative study was useful to assess the effect of LiNO₃ on the electrolyte stability at the anode interface. The XPS spectra collected on the cycled Cu electrodes are shown in Figure 4, whereas those of stripped and plated Cu electrodes are shown in Supplementary Figure 11 and Supplementary Figure 12, respectively.

Figure 4. X-ray photoelectron spectra of Cu electrodes recovered from Cu||Li cells, cycled either with QSPE-2 (upper panels) or QSPE-3 (lower panels) as electrolyte. Cycling was performed at room temperature, with two and a half cycles at 2 mAh cm^-2 and 0.1 mA cm^-2, followed by 50 cycles at 0.2 mAh cm^-2 and 0.1 mA cm^-2. (A) F 1s, (B) O 1s, (C) N 1s, (D) C 1s, (E) B 1s and( F) S 2p core levels with corresponding fittings based on internal references as well as^[38,60,61] and refs therein.

The most significant difference between the electrodes cycled with QSPE-2 and QSPE-3 lies in the relative concentration of FSI^- anion decomposition products compared to the FSI^- residuals. As seen in Figure 4, FSI^- residuals are more intense in the Cu electrode cycled with QSPE-3, whereas the signals of the FSI^- decomposition products (e.g., LiF in the F 1s spectrum, SO_x signal in the S 2p spectrum) are much weaker. Therefore, the addition of LiNO₃ limits the FSI^- reduction by Li. In addition, the fitting of the B 1s spectrum of the electrode cycled with QSPE-2 evidences the lack of BOB^- anion residuals, while other boron oxides are present, indicating that in the vicinity of the Cu electrode, this anion is completely decomposed during plating. This finding agrees with the reported reduction potential of BOB^- at ~1.8 V vs. Li/Li⁺. On the contrary, in the case of QSPE-3, a weak signal of BOB^- is still present, indicating that LiNO₃ partially protects this anion from decomposition. The lower reduction of FSI^- and BOB^- anions when LiNO₃ is present in the electrolyte seems to indicate a reduced electrolyte consumption likely by forming a more effective SEI rich in LiNO₃ decomposition products. Interestingly, no signal from Li₃N was clearly observed, suggesting that this SEI component, if present, might be buried more deeply in the SEI to be visible by XPS.

Similar results were also obtained with Cu electrodes analyzed after three full Li plating/stripping cycles (stripped Cu electrodes, Supplementary Figure 11) and after two and a half cycles (plated Cu electrodes, Supplementary Figure 12), except that, in the latter case, the BOB^- signal was observed as well when QSPE-2 was employed, although in minor concentration with respect to the case of QSPE-3. The relative concentration of the different components is reported in Supplementary Figure 13, also showing similar SEI composition in the different cycling stages, indicating quite stable chemistry throughout cycling. Larger differences were observed in the C 1s spectrum, between the electrodes cycled with QSPE-2 and QSPE-3, which may be caused by the presence of adventitious carbon and the rest of the electrolyte polymer matrix. Nevertheless, the systematically higher content of carbon bonds in the electrodes cycled with QSPE-2 vs. those cycled with QSPE-3, especially in form of hydrocarbon, C-O and COO, may denote a more organic SEI due to higher solvent decomposition. Altogether, the XPS analysis evidenced the effectiveness of LiNO₃ in protecting the electrolyte from intense reduction by metallic lithium, preventing the other anions and the solvent from strong decomposition, which has a beneficial effect on the plating/stripping coulombic efficiency.

Full cell performance

The full cell performance of QSPE-2 and QSPE-3 was first evaluated in NMC-811||Li-Cu cells, in coin cell configuration, with NMC-811 synthesized by spray drying and with cathode active material loading of ~13 mg cm^-2. The cells were cycled at room temperature, with C-rates ranging from C/20 to 1C, and then at a constant C-rate of C/10. The specific discharge capacity of two cells with QSPE-2 and QSPE-3, during the first 150 cycles, is depicted in Figure 5A, whereas the corresponding coulombic efficiency and capacity retention during the long cycling at C/10 are shown in Figure 5B and C, respectively. The complete discharge capacity profile of three cells with QSPE-3, up to 220 cycles, is shown in Supplementary Figure 14. The cell with QSPE-3 delivered slightly higher specific capacity during the initial C-rate test, with initial specific discharge capacity of 165 mAh g^-1 at C/20 and ~60 mAh g^-1 at C/1. During the rate test, the capacity decay is faster at low C-rates, which might be due to the longer residence time at high voltages and thus to a more severe electrolyte degradation. In the subsequent long cycling at C/10, the capacity fading rate was initially lower for the cell with QSPE-2, so that after 80 cycles, the capacity was approximately equal for both cells. In this initial range, the capacity retention was similar or slightly higher for QSPE-2 than for QSPE-3 (region I in Figure 5C). Nonetheless, after cycle 80, the cell with QSPE-2 started to lose capacity more rapidly and underwent sudden failure after approximately 130 cycles. On the contrary, the cell with QSPE-3 cycled for over 200 cycles without showing sudden failure, despite with a continuous capacity fade [Supplementary Figure 14]. The trends in the capacity fade correlate with some slight differences observed in the coulombic efficiency during long cycling. Initially, the cell with QSPE-2 has higher coulombic efficiency, concordant with the lower capacity fade rate. However, during the subsequent cycling, the coulombic efficiency decreased continuously for QSPE-2, while increasing for cells with QSPE-3. The coulombic efficiency dropping below 99% for QSPE-2 coincides with the sudden capacity drop observed in this cell. This is probably caused by the depletion of the cyclable Li in the anode^[62], exacerbated by the low amount of Li reservoir on the anode (~20 µm). This effect was not observed with QSPE-3 for over 200 cycles, which is attributed to the higher coulombic efficiency for the lithium plating/stripping process in presence of LiNO₃. To summarize, the aging behavior of the analyzed cells is characterized by two distinct regions. In the first region, the cells experience linear capacity fade, slightly more accelerated for cells with QSPE-3 than cells with QSPE-2. In the second region, which is dominated by anode depletion, the cells undergo sudden capacity fade. The onset of this second aging regime is delayed in cells with QSPE-3 as an electrolyte.

Figure 5. Galvanostatic cycling of NMC-811||Li-Cu coin cells with QSPE-2 and QSPE-3, performed at room temperature, in the voltage range between 3.0 and 4.3 V, and with cathode active material loading of ~13 mg cm^-2. (A) Specific discharge capacity during the first 150 cycles; (B) coulombic efficiency during the long cycling at C/10; (C) Capacity retention during the long cycling at C/10, normalized on the first cycle at C/10 after C-rate test.

EIS study in two- and three-electrode cell configurations

To understand the origin of the different behavior with the two electrolytes, two other cells were cycled at a constant C-rate of C/10 (after a formation cycle at C/20). In addition, EIS spectra were collected at the EoD, every 5 cycles, in the frequency range between 10 kHz and 10 mHz. The specific discharge capacity and coulombic efficiency of the two cells are represented in Figure 6A. The discharge capacity in the initial cycles was similar, namely 164 and 161 mAh g^-1 at C/20 and 142 and 138 mAh g^-1 at C/10 for QSPE-2 and QSPE-3, respectively. In this experiment, no sudden failure was observed, presumably because cycling was stopped before this mode of failure occurred. However, as in the previous experiment, the two cells showed a constant capacity decrease, with slightly better cyclability for the cell with QSPE-2. The capacity fade rate was similar in the two cases [Supplementary Figure 15], but the final capacity retention at cycle 100, with respect to the first cycle at C/10, was slightly higher for QSPE-2, namely 42% vs. 35% for QSPE-2 and QSPE-3, respectively. Concordantly, the coulombic efficiency was higher for QSPE-2, with average values (excluding the formation cycle) of 99.2% and 98.7%, respectively. The EIS spectra upon cycling were analyzed to shed light on the origin of the capacity fade. The full EIS spectra are shown in Supplementary Figure 16, whereas Figure 6B and C shows the zoom on high-frequency region of the spectra. The spectra are characterized by a small high-frequency semicircle, in the range from 10 kHz-100 Hz, followed by a large semicircle at lower frequencies. Thus, the spectra were fitted with a ladder equivalent circuit model comprising three resistances, labeled high-frequency resistance (R_HF), middle-frequency resistance (R_MF), and low-frequency resistance (R_LF), and two constant phase elements [Supplementary Figure 17]. As customary, the R_HF was attributed to the sum of the cell ohmic resistances, among which the main contribution is probably the bulk electrolyte resistance, whereas the middle-frequency semicircle (R_MF) was assigned to the sum of the different planar interface resistances, such as the cathode/QSPE and anode/QSPE interface resistances, and the cathode-current collector contact resistance. Since the area of these interfaces is similar, corresponding to the geometrical area of the cell, their EIS contributions are expected to be highly overlapped and appear in the middle-high frequency region^[63,64]. Finally, the lowest resistive contribution (R_LF) was assigned to the cathode charge transfer resistance^[65,66]. R_LF is several orders of magnitude larger than the other two resistances, which agrees with the expected dependence of the charge transfer resistance on the state of charge, particularly the expected increase towards the EoC/EoD^[67].

Figure 6. Galvanostatic cycling of two NMC-811||Li-Cu cells with QSPE-2 and QSPE-3 as electrolyte, coupled with EIS analysis. Galvanostatic cycling was performed at room temperature, in the voltage range between 3.0 and 4.3 V, and at constant C-rate of C/10 (with one initial cycle at C/20). The cathode active material loading was ~13 mg cm^-2. EIS spectra were collected at the end of discharge, every 5 cycles. (A) Specific discharge capacity (left axis) and coulombic efficiency (right axis); (B) selected EIS spectra of the cell with QSPE-2. Dots represent experimental curves and dashed lines of the fitting curves; (C) selected EIS spectra with QSPE-3. Dots represent the experimental curves and dashed lines of the fitting curves; (D-F) evolution of the (D) high-frequency, (E) middle-frequency, and (F) low-frequency resistance components upon cycling.

The evolution of the three resistances upon cycling is shown in Figure 6D-F. R_HF was approximately constant upon cycling and similar for the two electrolytes, concordant with the similar ionic conductivity of the two QSPEs. R_MF was initially 28 and 15 Ω for QSPE-2 and QSPE-3, respectively. R_MF increased steadily upon cycling, with a more pronounced rise observed for QSPE-2. This trend agrees with the less resistive and more stable anode interface obtained using LiNO₃, as observed in Li||Li cells [Supplementary Figure 5], and suggests that R_MF is strongly influenced by the resistance of the anode/QSPE interface. Although XPS analysis indicated stable SEI chemistry for QSPE-2, the SEI thickness likely increases during cycling due to the higher propensity of QSPE-2 for reduction by the lithium anode. Finally, R_LF increased steadily upon cycling, similarly for the two QSPEs, indicating progressive deterioration on the cathode side. This might be due to various processes, such as progressive oxidation of the electrolyte, degradation of the active material (e.g., cracking), or loss of contact between the electrolyte and the cathode active material particles upon repeated lithiation/delithiation cycles. Among the three resistive components, the increase of R_LF was the most evident, from ca. 1,000 Ω to ca., 5,000 Ω at the end of cycling, thus suggesting that the progressive capacity decay may be due to the increase of the cathode charge transfer resistance.

In conclusion, the EIS analysis in two-electrode cell configuration suggests that the progressive capacity fade is mainly associated with the degradation of the cathode interface. In this regard, LiNO₃ has, apparently, a negligible effect. Besides, a progressive increase of R_MF, associated with a progressive deterioration of the anode interface, was observed in the cell with QSPE-2, while this effect was less obvious in the cell with QSPE-3. This agrees with the previous results from Cu||Li cells characterizations, indicating a higher propensity of QSPE-2 to reductive decomposition on the lithium anode.

The effect of LiNO₃ on the cyclability of NMC-811||Li cells was further studied by cycling three-electrode cells at C/20 for 30 cycles, while collecting EIS spectra both at the EoD and at the EoC. The two cycled cells, with QSPE-2 and QSPE-3 as electrolytes, showed decreasing capacity, as observed in the previous experiments, with slightly faster fading for the cell with QSPE-3 [Figure 7A]. The coulombic efficiency was initially higher for the cell with QSPE-2, although it became approximately equal after ca. 15 cycles. So far, this confirmed the conclusion from previous experiments that the addition of LiNO₃ initially worsens the cyclability, although during long cycling, it has a positive effect due to reduced consumption of Li at the anode. We then focused our attention on the positive electrode, trying to clarify whether the addition of LiNO₃ affects cathode aging. Figure 7B shows the open circuit potential of the positive electrode (E_+,OCP), against the lithium metal reference electrode upon cycling time. While the potential at the EoD remained approximately constant, it decreased progressively at the EoC, indicating a decreased ability to extract Li⁺ from the cathode active material upon cycling. This could be attributed to a partial disconnection of the cathode active material particles upon cycling and is obviously related to the linear capacity decay of the cycled NMC-811||Li cells. Between the two cells, the decrease of the E_+,OCP is faster with QSPE-3, in agreement with the faster capacity decay, thus suggesting that the cathode aging is accelerated in presence of LiNO₃. As mentioned above, EIS spectra were collected both at the EoC and at the EOD. Selected EIS spectra of the positive electrodes are shown in Supplementary Figure 18 (QSPE-2) and Supplementary Figure 19 (QSPE-3). The spectra show two semicircles in the high/middle frequency range, clearly separated in the case of QSPE-2 and more overlapped in the case of QSPE-3, followed a large semicircle at low frequencies. The spectra can be modeled through a ladder circuit encompassing four resistances and three constant phase elements [Supplementary Figure 20]. The high-frequency series resistance (R_HF) is related to the electrolyte ionic resistance, whereas the R_MF (R_MF,1 and R_MF,2) is tentatively attributed to the planar cathodic interface resistances, namely the interface resistance between the cathode and the QSPE, and to the contact resistance between the cathode and the current collector. Indeed, planar interfaces should give rise to impedance features in the middle-/high-frequency range^[63,68], and in the system under consideration, once the anode contribution is removed, the only planar interfaces remaining are the QSPE/cathode solid interface and the interface between the current collector and the cathode layer. Finally, as mentioned above, the low-frequency component (R_LF) is attributed to the cathode charge transfer resistance. Each resistance is associated with a parallel capacitance that can be calculated from the CPEs, as described by Brug et al. The product of the resistances by their associated capacitances gives the characteristic time constants for the related polarization phenomena^[69]. The time constants for the middle- and low-frequency polarizations are shown in Supplementary Figure 21. The middle-frequency polarizations have time constants comprised between 10^-6 and 10^-4 s (τ_MF,1) and between 10^-3 and 10^-2 s (τ_MF,2). The low-frequency polarization has a very high characteristic time constant τ_LF > 10 s. The time constants are approximately constant upon cycling. That is, if a resistance increases upon cycling, the associated capacitance decreases. τ_MF,2 and τ_LF have similar values in QSPE-2 and QSPE-3-containing cells. However, τ_MF,1 is an order of magnitude higher for QSPE-3, with respect to QSPE-2, which explains the higher degree of overlap between the two middle-frequency semicircles with the former electrolyte.

Figure 7. EIS analysis of three-electrode NMC-811||Li-Cu cells with QSPE-2 and QSPE-3 as electrolytes. Cells were cycled between 3.0 and 4.3 V, at C/50 (2 cycles) and C/20 (30 cycles). EIS spectra were collected during cycling at C/20, at the end of charge (EoC) and at the end of discharge (EoD), after one hour of resting. (A) Specific capacity (left axis) and coulombic efficiency (right axis); (B) open circuit potential of the positive electrode (E_+,OCP) vs. Li/Li⁺, at the EoC and EoD, upon cycling time (each point corresponds to one cycle at C/20); (C) evolution of the main resistances of the positive electrode impedance spectra, at the EoC and EoD, with QSPE-2 as electrolyte; (D) evolution of the main resistances contributing to the positive electrode impedance spectra, at the EoC and EoD, with QSPE-3 as electrolyte.

The evolution of the resistance components upon cycling is shown in Figure 7C (QSPE-2) and Figure 7D (QSPE-3). The overlapping between the middle-frequency impedance components in QSPE-3 spectra results in a high degree of uncertainty on the related resistances. For a better comparison between the two electrolytes, the sum of the R_MF was considered in the analysis. The R_HF is insensitive to the state of charge and was stable upon cycling, indicating that the bulk properties of the electrolyte are not affected during these few cycles. R_MF increases moderately upon cycling for both electrolytes. The value is initially higher for QSPE-3 (27 vs. 53 Ω at the EoC), but the increase is more pronounced for QSPE-2, namely 90% vs. 47%. A slight divergence between EoC and EoD is observed, especially in the case of QSPE-2. The increase upon cycling suggests a progressive degradation of the cathode/QSPE interface, or a progressive loss of contact between the cathode-coated layer and the current collector. The higher resistance of the middle-frequency component and the higher value of the time constant τ_MF,1, with QSPE-3, suggest a more sluggish charge transfer across the cathode/QSPE interface. However, the most striking differences are observed in the R_LF. In the case of QSPE-2, R_LF increases from ~1,500 to ~2,500 Ω after 30 cycles, with similar values at the EoC and EoD. As for the two-electrode cells, the high values of cathode charge transfer resistance are caused by the blocking behavior of the cathode in the fully lithiated/delithiated states. For QSPE-3, however, the increase at the EoC is much more pronounced. R_LF increases from 1,200 to 2,000 Ω at the EoD, and from a similar initial value to 5,500 Ω at the EoC. This increase suggests a faster degradation of the cathode interface at high voltages, possibly involving oxidation of the catholyte, and may explain the faster capacity decay with the LiNO₃-containing QSPE-3. The EIS analysis in a three-electrode cell configuration confirms the results from LSV and floating current experiments, showing faster electrolyte oxidation for QSPE-3 with respect to QSPE-2. It is interesting that this effect was not observed in the two-electrode cell, as in that case, the EIS spectra were collected at the EoD. The study remarks the importance of collecting EIS spectra at the EoC during cycling, rather than at the EoD, as degradation effects due to high voltage may be visible only in the charged state.

To sum up, the addition of LiNO₃ improved the coulombic efficiency for the lithium plating/stripping process, delaying sudden cell failure due to lithium depletion at the anode. However, the primary cause for capacity fade in this type of cell appears to be related to the degradation of the positive electrode or the related electrolyte interface. In this regard, the addition of LiNO₃ did not show any significant benefit. On the contrary, cells containing LiNO₃ showed slightly lower coulombic efficiency and higher capacity fade rate, possibly due to accelerated electrolyte oxidation at high voltages. In addition, EIS analysis in a three-electrode cell configuration showed that this accelerated capacity fade is related to a steeper increase of the cathode charge transfer resistance at the EoC, and thus to a faster degradation of the cathode active material in the delithiated state, or to a faster oxidation of the electrolyte/catholyte at high voltages. This effect, which is further confirmed by the previous results from LSV and floating experiments, might be due to several factors, such as crossover of the LiNO₃ decomposition products from the anode to the cathode, a catalytic activity towards electrolyte oxidation, or LiNO₃ interference with the formation of the LiBOB-borne CEI.

Pouch cell performance

To complete the study, a monolayer pouch cell with QSPE-2 as an electrolyte was assembled and cycled, first at different C-rates, and then at a constant C-rate of C/10. Figure 8 shows the specific capacity and coulombic efficiency of this cell. The initial specific capacity of this cell (at C/20) was close to 200 mAh g^-1, close to the theoretical specific capacity of NMC-811. The higher capacity obtained with this cell, compared to the previous results in coin cells, is attributed to the use of commercial, optimized NMC-811, while non-optimized research-grade NMC-811 was used for the study of coin cells. The cell showed also excellent C-rate capability, with a specific capacity of 150 mAh g^-1 at 1C, despite having a cathode with high area capacity of 2.5 mAh cm^-2. The cyclability was also good, with a capacity retention of 80% reached after ca.90 cycles. No sudden failure was observed over 160 cycles, possibly thanks to a higher and more homogeneous pressure distribution in the pouch cell format. The average coulombic efficiency, during the long cycling at C/10, was over 99.5%.

Figure 8. Full cell performance of a NMC-811||Li-Cu monolayer pouch cell with QSPE-2 as electrolyte. Galvanostatic cycling was performed at room temperature, between 3.0 and 4.3 V (see Table 2 for detailed cycling protocol). The nominal area capacity for this cell was 2.5 mAh cm^-2.

Altogether, the results confirm the suitability of this QSPE for application in high-voltage quasi-solid-state lithium metal batteries, although its cyclability must be further improved to meet the requirements for practical use. The limited cyclability arises from the degradation of the electrolyte/catholyte at high voltages, the chemical reactivity of delithiated NMC-811, and lithium depletion at the anode during cycling. These issues are further exacerbated when cycling at high area capacities, which are necessary for achieving high energy densities^[70]. Tackling these challenges will require comprehensive optimization of all cell components and their interfaces^[71]. For instance, electrolyte degradation at the cathode interface could be mitigated by applying coatings to the cathode active material particles^[72-74] or further refining the electrolyte formulation. Improving cyclability will also necessitate enhancing the plating-stripping coulombic efficiency. Lithium reactivity at the anode could be reduced by applying a passivation layer on its surface^[75-77]. Additionally, the use of SEI-forming additives, such as LiNO₃, has proven highly effective, as demonstrated in our work.