All-solid-state lithium batteries with NMC₉₅₅ cathodes: PVDF-free formulation with SBR and capacity recovery insights

Materials

The cathode material was meticulously prepared using a mixture of 57% NMC₉₅₅ (Huayou New Energy Technology), 38% LPSCl (NEI corporation), 3% Active Carbon (TIMCAL graphite & carbon super C₆₅), and 2% binder (SBR-1502 KUMHO, cold-polymerized SBR, containing 23.5 at/wt. % styrene, HB Chemical). The solvent used in this solution was toluene (Honeywell, purity ≥ 99.7%).

Cathode preparation

The preparation process involved four main steps [Figure 1]: 1^st the binder SBR was mixed with 5 mL of toluene for 24 h; 2^nd NMC₉₅₅ and C₆₅ were added to the solution and mixed for another 24 h; 3^rd the electrolyte was introduced to the last mixture, and all was stirred for an additional 24 h, and 4^th using a 4-side film applicator with a thickness of 200 μm; the resulting slurry was deposited onto a pre-heated Al foil coated with carbon (at 100 °C for 10 min.) and left to dry for at least 24 h. After drying, the prepared cathode film was cut into discs of a 10 mm diameter using a disk drilling machine for a coin-cell electrode cutter from TMAX-CN. All steps were performed inside the GS Glovebox (O₂ < 1 ppm and H₂O < 1 ppm). A detailed microstructural study of the resulting materials was performed after depositing the cathode material on an Al-coated carbon substrate (MSE Supplies LLC) of 15 μm thickness.

Cell fabrication and electrochemical characterization

A bilayer of the previously prepared cathode material and LPSCl solid electrolyte was formed. To achieve this, 100 mg of LPSCl was pressed onto the cathode disk for one minute at 1 ton, using a mold with a 10 mm diameter and the SPECAC press machine. The CR2032 coin cells were fabricated according to the structure depicted in Figure 1. The anode material used was pure lithium metal, supplied by Goodfellow, with a thickness of 200 µm and 8 mm diameter. The areal capacities of the cathodes vary from 0.167 mAh.cm^-2 to 2.17 mAh.cm^-2, calculated based on the active area of the cathode, which is constrained by the size of the anode and considering the theoretical capacity of 192 mAh.g^-1. The cells were closed using a digital pressure-controlled electric crimper for coin cells MSK-160E. Once the cells were assembled, the electrochemical performance was assessed by a series of electrochemical tests. Three types of measurements were performed. Firstly, potentiostatic electrochemical impedance spectroscopy (PEIS) tests were performed to determine the internal resistance using a Biologic VMP-300 potentiostat. The PEIS experiments were performed with an alternate current (AC) with an amplitude of ±10 mV for an initial differential potential corresponding to the cell’s open circuit voltage (OCV) and the frequency range from 7 MHz to 100 mHz. Subsequently, cyclic voltammetry tests were applied to determine the capacity of the obtained cell using a Biologic VMP-300 potentiostat. The experiments superimposed ±0.1 V to the initial differential potential corresponding to the cell’s OCV. The rate ranged from 0.1 up to 50 mV.s^-1. The cyclic voltammetry measurements were repeated five times, with applied limits for each cycle ranging from 2.0 V to 4.5 V. The galvanostatic charge-discharge (GCD) was performed to evaluate the cyclability and behavior of cells at different C-rates and specific energy. The galvanostatic cycling tests were performed in a Neware-CT-4008Tn-5V-20mA battery tester with 2.6 V and 4.1 V cut-off voltages.

Cathode characterization (Hall effect and sheet resistance)

Hall effect measurements were performed using the Linseis HCS 1 Hall effect measurement system with two magnetic circuits (Neodymium) assembled on a moveable sled. Samples were prepared as positive electrode films, ensuring they were suitable for Hall effect measurements. The prepared sample was then mounted on the sample holder of the Linseis HCS system, with electrical connections established between the sample edges and the input terminals of the measurement system. A perpendicular magnetic field B, B = ±0.7 T, was applied to the sample, and the intensity of the magnetic field was adjustable and accurately measured by the system. A constant current (I) [DC: 1nA up to 125 mA (8 decades/compliance 12 V)] was transported through the sample along its length. The Hall voltage (V_H), generated perpendicular to the applied current and the magnetic field, was measured [DC: low noise/low drift 1 μV up to 2,500 mV, 4 decades amplification, digital resolution: 300 pV]. This voltage arises due to the Lorentz force acting on the charge carriers within the sample. Hall voltage measurements were recorded for various magnetic field intensities and current values to ensure accuracy, with multiple measurements conducted for repeatability. The Hall coefficient (R_H) was calculated using $$ R_{H}=\frac{V_{H} d}{I B} $$, where d is the sample thickness. The charge carrier density ccc_Hall, n, and the mobility µ’ were subsequently determined using $$ n=\frac{1}{e R_{H}} $$ and $$ \mu^{\prime}=\frac{R_{H} \sigma}{e} $$, where e is the elementary charge and σ the conductivity of the sample measured as follows.

Samples were prepared with a uniform thickness suitable for sheet resistance measurements. Following the Van der Pauw method, four collinear electrical contacts were attached to the samples when mounted on the sample holder, and electrical connections were made to the measurement system. A constant current (I) was applied between two adjacent contacts on the sample, and the voltage (V) was measured between the remaining two adjacent contacts. Measurements were taken in different configurations, rotating the contacts to ensure precision and consistency, with multiple readings recorded for each configuration. The sheet resistance (R_S) was calculated using $$ R_{s}=\frac{V}{I} f $$, where f is a geometrical correction factor, typically close to 1, depending on the contact configuration and sample geometry. For the Van der Pauw method, the relationship $$ e^{-\pi R_{s 1} / R_{S}}+e^{-\pi R_{S 2} / R_{S}}=1 $$ is used, where R_S1 and R_S2 are resistances measured in perpendicular directions.

All measurements were conducted in a controlled environment. The system was evacuated to 10^-2 bar, and subsequently, an N₂ flux of 4-5 L/min, was used to minimize spurious reactions, temperature fluctuations, and external noise.

The results were analyzed to determine the Hall coefficient, carrier concentration, mobility, and sheet resistance. They were then compared with theoretical values and reference materials for validation.

Cathode characterization (scanning electron microscope/energy dispersive spectroscopy and atomic force microscope/scanning kelvin probe microscopy)

A scanning electron microscope (SEM) FEI QUANTA 400 FEG ESEM was used. The chemical composition analysis was performed with an EDAX Genesis X4M energy-dispersive X-ray spectrometer. Microstructural observations and chemical composition analyses were conducted using an accelerating voltage of 15 kV, a spot size between 3 and 5, and a working distance of 10 µm. Images were taken at magnifications ranging from 20× to 20,000×. All images were obtained with a backscattered electron (BSE) detector. To further characterize the cathode material, an Atomic Force Microscope (AFM) capable of conducting Scanning Kelvin Probe Microscopy (SKPM) was employed (Oxford Instruments Atomic Force Microscope MFP-3D Origin⁺ AFM). The analysis was made with a laser-driven tip with a resonant frequency of approximately 75 kHz ASYELEC-01-R2. The tip material comprises silicon with a reflective coating of Ti/Ir and a spring constant (k) of 2.8 N.m^-1. The topography analysis was conducted using tapping mode, which consists of gently oscillating the probe over the sample surface to acquire high-resolution topographic information. Simultaneously, this analysis of the SKPM acquisition was performed. SKPM is a non-invasive technique derived from atomic force microscopy that allows the electrochemical characterization of materials.

Cathode characterization (TOF-SIMS and Raman)

TOF-SIMS analyses were carried out in a M6 TOF-SIMS instrument (IONTOF GmbH, Münster, Germany) using 30 keV Bi₃⁺ primary ions. The sample was mounted on the sample holder (top-mount, without contact with the cathode surface) inside a glove box and transferred into the TOF-SIMS instrument using an air-tight transfer vessel (IONTOF GmbH). Positive and negative ion data were recorded in the all-purpose spectrometry mode and the delayed extraction fast imaging mode, over analysis areas ranging from 100 × 100 µm² to 250 × 250 µm². Only negative ion data is presented in this study. Before analysis, the top layers of the sample surface were removed by sputtering using 2 keV Cs⁺ ions (58 nA, 600 × 600 µm², 37 s).

The cycled cell used in the TOF-SIMS analysis had an areal capacity of 1.90 mAh.cm^-2 and was subjected to PEIS, cyclic voltammetry, and electrochemical cycling [Supplementary Figure 3]. The electrochemical potential range was set between 2.6 V and 4.1 V. The cell underwent ten cycles, with three cycles at C/50, 3 at C/20, 3 at C/10, and 1 at C/5 before TOF-SIMS post-mortem analyses.

Raman spectra were collected at room temperature with a Raman spectrometer (Renishaw inVia Confocal Raman, Reinshaw, Wotton-under-Edge, UK) equipped with a laser of 532 nm excitation wavelength focused through an inverted microscope (Leica, Wetzlar, Germany), via a 50× objective. A composite NMC₉₅₅ electrode was placed in an air-sensitive sample holder and the sample preparation was performed inside the glovebox (H₂O, O₂ ≤ 1 ppm).

The cell used for post-mortem Raman testing had a cathode loading of 1.6 mAh.cm² and was cycled between 2.6 V and 4.1 V at a C-rate of C/20 for up to ten cycles. A CV step was introduced during charge and discharge.

Cell characterization (Calorimetry)

Heat flow analysis was conducted using a TAM IV microcalorimeter (TA Instruments), which provides high sensitivity measurements. The experiment was performed at a constant temperature of 30 °C. The cell under investigation was cell CC I [Supplementary Table 1] and consisted of lithium metal as the anode, LPSCl as the electrolyte-separator, and an NMC₉₅₅-based cathode material. The calorimetric analysis was the final test performed on the cell. For the test, a reference cell was also assembled for comparison, consisting solely of a top cap, bottom cap, separator, and a spring all made of stainless steel. When the heat flow started, the cell displayed an OCV of 3.7 V. To further investigate the internal heat flow behavior, external resistors (217 kΩ and 253 kΩ) were connected to the cell through an external circuit. These resistors allowed for an evaluation of heat flow during the discharge of the battery.

Density functional theory and structural, electronic, and potential energy simulations

In the computational analysis, the study employed density functional theory (DFT)^[49], facilitated by the Vienna Ab initio Simulation Package (VASP)^[49] code. Simulations were focused on a [Li₁₂MnCo(Ni₅O₁₂)₂]₂ - LiNi_0.833(3)Mn_0.083(3)Co_0.083(3)O₂ (P1) supercells originally optimized from a R-3m (hP18) rhombohedral disordered structure with a = b = 2.874 Å and c = 14.165 Å and α = β = 90 and γ = 120. To the latter disordered structure, Li was randomly removed until it became Li₇Mn₂Co₂Ni₂₀O₄₈ - Li_0.292Ni_0.833(3)Mn_0.083(3)Co_0.083(3)O₂. The DFT simulation was based on DFT and the generalized gradient approximation (GGA)-Perdew-Burke-Ernzerhof (PBE)^[50] exchange-correlation functional for describing the interactions. Electronic structure calculations were performed with a cut-off energy of 500 eV and a k-spacing of 0.3 Å^-1 corresponding to a 5 × 2 × 2 mesh and electronic convergence of 10^-5 eV. Model-determined non-magnetic and spin-polarized magnetic simulations were performed to compare optimized structures and stabilities (energies of formation). Supercells with (100), (010), and (001) surfaces were built with a > 10 Å vacuum layer to disrupt periodicity for both [Li₁₂MnCo(Ni₅O₁₂)₂]₂ and Li₇Mn₂Co₂Ni₂₀O₄₈ structures [Supplementary Figure 2] and determine the total local potential and its 10 Å integration range macroscopic average potential. The work function (WF), or electrochemical potential $$ \bar{\mu} $$, determines the energy difference between the vacuum surface and the Fermi-level macroscopic average potentials. The electrochemical potential of a species i (element, phase, cathode, current collector,) is $$ \bar{\mu}_{i}=\mu_{i}+e \phi_{i} $$, where µ_i is its chemical potential and ϕ_i indicates its surface potential due to electrical contact with another species or to self-induced surface effects. Hence, if $$ \phi_{i}=0 \Rightarrow \bar{\mu}_{i}=\mu_{i} $$, which was the assumption used herein, except for cell-like electrical contact between species. When two species i and j come in electrical contact, the electrochemical potentials tend to equalize, $$ \bar{\mu}_{i}-\bar{\mu}_{j}=0=\mu_{i}-\mu_{j}+e (\phi_{i} -\phi_{j}) $$ and, if one of the materials is an insulator such as the LPSCl solid electrolyte, an electrical double layer capacitor (EDLC) with potential difference ΔV_ij = (ϕ_i - ϕ_j) will be spontaneously formed at the interface of the two species, e.g., electrode/electrolyte where the energy is stored. The potential vs. Li⁰ simulated and plotted in Figure 2 is $$ \Delta V_{NMCvsLi^{0} }=\frac{\mu_{Li^{0}}-\mu_{M M C 955}}{e}=\left(\phi_{L i^{0}}-\phi_{N M C 955}\right)=-1.39-\phi_{N M C 955} $$ with $$ \phi_{N M C 955} < \phi_{L i^{0}} $$^[51]. Charging corresponds to a chemical potential increase of the negative electrode, and a decrease of the chemical potential of the positive electrode (increasing bias potential). Discharging is a spontaneous process and corresponds to the opposite variation of the chemical potentials (decreasing bias potential). The zero-energy reference corresponds to the maximum energy and the energy of the electrons at rest in a surface in vacuum^[51]. The capacity in Figure 2 is calculated from the number of charge carriers using $$ { capacity }=\frac{{ carrier_{-}densty }\left({ number_{-}of } _{-} { electrons. } c m^{-3}\right) \times\bar{e}(C)}{\rho(g.c m^{-3})} \frac{1}{3.6}\left(\mathrm{mAh} \mathrm{.g^{-1}}\right) $$ where $$ \bar{e} $$ = 1.6 × 10^-19C is the elementary charge and ρ = 4.71 g.cm^-3 the volumetric mass of NMC₉₅₅ simulated herein.

Tools from VASP and Electronics within the MedeA 3.7 software were utilized for calculating chemical potentials µ_i, charge carrier densities n, sheet conductivities σ, mobilities µ’, electronic band structures, and DOS.

Simulations to determine optimal cycling thresholds

To comprehend the behavior of NMC₉₅₅, ab initio simulations using DFT^[49] were performed to determine optimal parameters for testing, such as the potential (V) limits for the electrochemical cycles [Figure 2A] and cycle curve morphologies [Figure 2A]. The cut-off charging voltage was determined to be 4.05 V, coinciding with the (Li_xNMC₉₅₅) with 0.29 < x < 1 solid solution transforming from (LiNMC₉₅₅) to (Li_0.29NMC₉₅₅), and the net number of free electrons reduced to a minimum, corresponding to a capacity of ~10 mAh.g^-1. Up to 4.05 V, the slightly more stable solid solution composition is LiNMC₉₅₅ as its chemical potential versus Li⁰ is higher. Above 3.36 V, NMC₉₅₅ is a solid solution, a homogeneity region of (Li_xNMC₉₅₅) with 0.29 ≤ x ≤ 1.00^[52]. While discharging to the cut-off potential of 2.60 V, the cell initially charged to Li_0.29NMC₉₅₅ shows a theoretical simulated capacity of 196 mAh.g^-1 per comparison with the calculated from Li_0.3NMC₉₅₅ ↔ 0.7Li⁺ + 0.7$$ \bar{e} $$ + (Li_0.3NMC₉₅₅): ~192 mAh.g^-1.

Herein, the importance of discharging at CV of ~2.60 V becomes clear as the number of free electrons should achieve 2.07 × 10²²$$ \bar{e} $$.cm^-3 corresponding to the fully formed (LiNMC₉₅₅). This allows the cell to be subsequently charged to 4.05 V, forming (Li_0.29NMC₉₅₅) and keeping structural stability. The latter leads to a long cycle life corresponding to a primary redox activity involving approximately 0.7Li⁺ + 0.7$$ \bar{e} $$ + (Li_0.29NMC₉₅₅) (LiNMC₉₅₅) and the nickel transitioning from higher to lower oxidation states (Ni⁴⁺ → Ni³⁺) down to 3.00 V. While discharging form the first “shoulder” observed in Figure 2A at 3.12 V to 2.60 V, cobalt may also play a role lowering its oxidation state (Co⁴⁺ → Co³⁺), not substantially affecting the integrity of the (NMC₉₅₅) structure. The only species found on the cathodes’ surface after bombarding the sample with a focused beam of high-energy ions with TOF-SIMS are NiO₂^- and CoO₂^- in agreement with the Ni³⁺ and Co³⁺ discharged species between 4.05 V and 2.6 V [Figure 2E].

The number of charge carriers is measured using the Hall effect with two magnetic circuits with a magnetic field (B_max = 0.7 T). A current of 6.7 mA.cm^-2 is applied to the cathode sheet, and the Hall potential V_H and constant R_H are determined, leading to the charge carrier density or concentration (ccc) calculation (Experimental Section). The ccc_Hall resulting from applying the Hall effect to a 1.6 × 2.0 cm² NMC₉₅₅ cathode with 57% LiNMC₉₅₅+38% LPSCl+3% C₆₅+2% SBR is directly comparable to the simulated ccc_simulated of (LiNMC₉₅₅) for different charged states [Figure 2B]. It should be noted that the cathode in the Hall effect and sheet resistance experiments is charged only with electrons, not with Li⁺-ions, as the cathode is not inserted into a battery cell. The simulations match the experiments; for the chemical potential µ = 3.32 V and 4.01 V vs. Li⁰, the concentration of electrons $$ \bar{e} $$ surpasses the number of cations: electron vacancies and holes, represented by (h⁺). For µ = 4.17 V and 5.58 V vs. Li⁰, h⁺ surpasses $$ \bar{e} $$, in agreement with the Hall and electrochemical experiments, indicating that the cathode was over-charged with oxygen (O₂) release possibly occurring [Figure 2E]; it is noteworthy to highlight yet again that the cathode is not charged with Li⁺ as it is not charged as a cell component but as a single sheet.

The absolute chemical potential of the solid solution (LiNMC₉₅₅) (coinciding with its Fermi level when electrically insulated with null surface potential) is µ₀ = 4.15 V vs. Li⁰, corresponding to 2.74 × 10²¹ h⁺.cm^-3[Figure 2A] obtained by simulating the total local potential for a surface of Li₁₂Ni₁₀MnCoO₂₄ [Figure 2, right, Supplementary Figure 2]. It agrees with the chemical potential µ₀ = 4.09 V vs. Li⁰ of the average of charge carriers concentration ccc_Hall = 1.13 × 10²¹ h⁺.cm^-3 in Figure 2B.

Note that the analyzed cathode contained 38% LPSCl electrolyte, a semiconductor with an electronic structure band gap energy of 2.5 eV^[28], which may have been under-evaluated by DFT-GGA as traditionally occurs with semiconductors; the band gap may ascend to 4.2 eV. The cathode also contained 3% C₆₅+2% SBR, but other materials besides the active cathode did not considerably affect the experimental results for 0 ≤ T(°C) ≤ 80, as shown for the sheet conductivity in Figure 2C. Below 0 °C, the experimental data marks the behavior of a conductor with increased conductivity as the temperature drops; the presence of carbon black in the cathode may have influenced the conductivity. The simulated data, however, shows an increase in conductivity as the temperature decreases for µ = 2.89 V vs. Li⁰; for all other values, the material behaves as a semiconductor, as expected for LiNMC₉₅₅ [Figure 2D and E]. All experimental reversible results (common to I, II, and III) fall within the values of the simulated data. The electrical conductivity for the entire range of temperatures increases as the chemical potential rises to µ = 3.65 V vs. Li⁰ and decreases for µ = 4.17 Vvs. Li⁰, which agrees with the general drop in electrical conductivity as the cathode is charged. The simulated mobility of the charged species is shown in Supplementary Figure 2.

The cathode’s surface potential ϕ_cathode, measured by AFM with its scanning kelvin probe (SKP) mode, is 2.71 Vvs. Li⁰, which coincides with the expected OCV from the analysis of Figure 2A and D. The 2.63 V marks the higher voltage of a small band gap with null DOS and the lowest voltage of a region with a high DOS, where the electrochemical cycles should be confined. The DOS steadily decreases above 4.04 V, as expected. The cut-offs of 2.63 V and 4.04 V and the inflection voltage at 3.36 V vs. Li⁰ represent pinpoints in the DOS vs. chemical potential plot of (LiNMC₉₅₅) and (Li_0.29NMC₉₅₅) [Figure 2D and E]. The lithiation between 4.04 V and 3.00 V predominantly corresponds to the redox reaction (Ni⁴⁺ → Ni³⁺) and between 3.00 V and 2.63 V to (Co⁴⁺ → Co³⁺). Below 2.24 V, the second “shoulder” gain in capacity is due to the (Mn⁴⁺→Mn³⁺) transformation [Figure 2A]. However, this extra capacity may be paid for with structural instability as the latter pairs are highly paramagnetic, conversely to (Co^4+/3+), likely causing magnetic frustration. It is worth noting that the post-mortem analyses of cells that were cycled between 2.6 V and 4.1 V show evidence of small oxidation of the sulfates in the electrolyte on the LiNMC₉₅₅ surface. This oxidation is due to sulfur diffusion and cannot be completely avoided. However, keeping the cell within the cycling range of 2.6 V to 4.1 V prolongs the cell cycle life by reducing to the barely minimum the oxidation and the formation of rock salts, Li_xNiO_2-δ resulting from oxygen deficiency due to lattice oxygen release, and Ni_1-δO following from reduced nickel oxide phase due to oxygen loss.

Materials characterization with SEM/energy-dispersive X-ray spectroscopy and atomic force microscopy/SKPM

The microstructure of the active material LiNMC₉₅₅ in powder was analyzed by Scanning Electron Microscopy/Energy-Dispersive X-ray Spectroscopy (SEM/EDX) and presented in Figure 3A. The nano-sized particles of NMC show a relatively irregular size and shape and a tendency to form agglomerations. Additionally, the EDX analysis demonstrates that the atomic Ni:Mn:Co ratio is approximately 90:5:5, consistent with the anticipated composition [Supplementary Table 2]. The cathode discs prepared in this study were also analyzed using SEM/EDX [Figure 3B and C]. A uniform cathode deposition is demonstrated in Figure 3B. By increasing the magnification in Figure 3C, the homogeneous distribution of the active material nanoparticles is observed in detail: light grains, point 2 [Supplementary Table 2]. The dark gray particles correspond to the LPSCl electrolyte, point 1 in Figure 3C. The SBR binder is also apparent, denoted by the black spots uniformly dispersed throughout Figure 3B and C, effectively binding the materials together within the sample. Additional proof of the homogeneous consistency of the cathode is given in Supplementary Figures 4 and 5, with the element distribution maps (C, O, Co, Mn, Ni, Cl, P, and S).

Figure 3. Characterization of cathode materials. SEM/EDX images acquired with the BSE detector of (A) pure NMC₉₅₅ powder at a magnification of 20,000; (B and C) cathode sample of 57% NMC₉₅₅+38% Li₆PS₅Cl+3% C₆₅+2% SBR at different magnifications 500 and 2,000×. Surface topography of cathode material using atomic force microscopy and SKPM; (D) 3D view; 2D electrochemical profile in (E) and 3D in (F). Note: the electrochemical potentials are plotted as analyzed vs. SHE (standard hydrogen electrode); to refer to Li⁰: Vνs Li⁰ = Vνs SHE + 3.05. The areal capacity of the cathodes in SEM/EDX (B and C) and atomic force microscopy/SKPM (D-F) analyses is ~0.17-0.20 mAh.cm^-2. SEM: Scanning electron microscope; EDX: energy-dispersive X-ray spectroscopy; SKPM: scanning Kelvin Probe Microscopy.

To characterize the prepared cathode foil’s topography and surface chemical potential, an AFM with the ability to perform SKPM was used. The topography profile of the cathode is shown in Figure 3D, demonstrating that the cathode is uniform despite nanotopographic “peaks”. The corresponding electrochemical profile is shown in Figure 3E and F. As expected, there is no relationship between the topography [Figure 3D] and the electrochemical profile [Figure 3F] and Supplementary Figure 6. In the electrochemical profile [Figure 3F], charge accumulation regions with minimal amplitude can be observed between 0.155 V and -0.845 V vs. standard hydrogen electrode (SHE) [Figure 3E], corresponding to 3.205 V and 2.205 V vs. Li⁰. These results demonstrate that this cathode material exhibits quasi conductor-like behavior despite containing a significant proportion of the semiconductor LiPSCl electrolyte and the insulator SBR. In fact, the potential variation of the stripes in Figure 3E and F may indicate semiconductor behavior with a band gap approximately < 1 V [Figure 2A, D and E] for LiNMC₉₅₅ between 2.6 V and 2.2 V. This observation aligns with the results obtained experimentally and simulated for the conductivity and DOS of the cathode [Figure 2C and D]. Another significant observation is that the average potential of this cathode sample is -0.35 V vs. SHE, 2.71 V vs. Li⁰, and -4.09 eV in the physical scale (absolute scale)^[51] where the maximum chemical potential is zero for the electrons at rest in a surface in vacuum. The latter value is within the limit of the anode-cathode typical barrier^[13,51].

Electrochemical characterization

The obtained cathode foils were used to assemble all-solid-state batteries with LPSCl as the electrolyte separator and lithium metal as an anode. All cells analyzed in this article are summarized in Supplementary Table 1. After assembly, the cells underwent electrical impedance spectroscopy (EIS) and cyclic voltammetry tests [Figure 4A and B]. The equivalent circuit of the EIS was determined, along with the values of its components. R₁ is an ohmic resistor and corresponds to the resistance to the conduction of most agile species at higher frequencies. It represents asymmetric phenomena mostly occurring at the interfaces. Ideally, R₁ should approach zero. On the other hand, R₂ is the diameter of the first semicircle and corresponds to ionic hoping in a capacitor with no charge accumulation at the interfaces. R₂ refers to the bulk impedance of the dielectric (electrolyte) in a capacitor with $$ C=\varepsilon_{0} \varepsilon_{r} \frac{A}{d} $$, where A is the cross-section area, d is the thickness of the electrolyte, and ε = ε₀ε_r represents the dielectric constant of the electrolyte, where ε₀ is the permittivity of vacuum and ε_r stands for the real relative permittivity of the electrolyte. The resistance of the electrolyte is $$ R_{2}=\frac{1}{\sigma} \frac{d}{A} $$ where σ is the ionic conductivity of the electrolyte. The ionic hoping in a solid inorganic electrolyte affects both electrodes simultaneously as Li⁺ displacing in a direction implies a Li⁺ vacancy charged negatively V_a^- moving in the opposite direction; therefore, the best element to represent the opposition to Li⁺ hoping is a resistor in parallel with the correspondent capacitor. The phenomena occurring at the interfaces in Electrical Double Layer Capacitors (EDLCs) are related to Q₃, R₃ and Rd₄ [Figure 4A]. Non-ideal Li⁺ diffusion occurs from the EDLCs and the diffusion pathway should only correspond to a small path/thickness in the positive electrode.

Figure 4. Electrochemical characterization of a cell containing the 57% NMC₉₅₅+ 38% Li₆PS₅Cl+3% C₆₅+2% SBR cathode with active material Li⁰/Li₆PS₅Cl/NMC₉₅₅. (A and B) PEIS and cyclic voltammetry analysis of the cell before the electrochemical cycling; (C) best electrochemical cycles for each C-rate; (D) cycles at C/10, C/5 and C; (E and F) Specific capacity vs. cycle number for charge and discharge, respectively, for the following C-rates: C/50, C/20, C/10, C/5, C/3, C/2, C, C/10, 2C, 3C, C/10, C/5, C/3, C/2, and C; (G and H) PEIS and cyclic voltammetry analysis of the cell after the electrochemical cycling. Cells tested at 25 C with no applied pressure; areal capacity of the cathode: 0.167 mAh.cm^-2. Note: In E the lowest charging C-rate at CV varies from C/15 to C/76 for an average of C/46 and in F the lowest discharging average C-rate at CV is C/24. The maximum discharge capacity at C C-rate is > 120 mAh.g^-1. SBR: Styrene-butadiene rubber; PEIS: potentiostatic electrochemical impedance spectroscopy; CV: constant voltage.

The cyclic voltammetry [Figure 4B] clearly show a pronounced reduction peak at 3.3 V while the cell is discharged. Notably, the current’s response sharply increases at ≥ 4.0 V, suggesting a shift in charge carrier origin from $$ \bar{e} $$ to h⁺ (with possible loss of O^δ-), emphasizing the importance of charging at CV, avoiding charging beyond this cut-off voltage. As discussed previously, this limit was also verified in theoretical simulations and experiments in Figure 2. These findings allow the establishment of electrochemical cycling limits ranging from 2.6 V to 4.1 V based on the cyclic voltammetry [Figure 4B], which agree with the simulations in Figure 2.

The cell was subjected to electrochemical cycles at different C-rates according to the estimated electrochemical limits. The theoretical specific capacity of the NMC cathode was considered 192 mAh.g^-1 since only ~70% of the Li ions were allowed to be delithiated from the cathode material while charging^[13]. The best cell performance for each of the implemented C-rates: C/50, C/20, C/10, C/5, C/3, C/2, C, 2C, and 3C are shown in Figure 4C. Based on the electrochemical cycles [Figure 4C], it is marked the significant stability in cell performance.

NMC₉₅₅ cathode materials are prone to safety issues, such as thermal runaway, structural instability, and oxygen release due to the eminent presence of Ni^2+[9,22]. Employing appropriate fabrication methods may help mitigate these risks by ensuring uniformity and minimizing the presence of impurities or defects that could trigger unwanted reactions under stressful conditions^[22]. From the electrochemical performance in this study, it is evident that this cathode combination enables the creation of stable, long-lasting cells with unique performance, reaching the theoretical capacity multiple times for different C-rates, Figure 4D and Supplementary Figure 7A and B, showcasing a coulombic efficiency of ~100% considering the CV step in the electrochemical cycling [Supplementary Figure 7C]. Incorporating a CV step during both the charge and discharge stages is crucial for ensuring the cell’s long-term performance and little to no Ni²⁺ formed. The redox reactions for nickel are restricted to (Ni⁴⁺ ↔ Ni³⁺). This distinction becomes evident while analyzing the maximum discharge capacity obtained in the experiments shown in Figure 4E and F, Figure 5, and Supplementary Figure 7. At the charging process at higher C-rates, the cell’s capacity is achieved at CV, as shown in Figure 4E. The discharge at a constant current attains a higher capacity than the corresponding charging step [Figure 4F]. By including a final CV step, the cell repeatedly reaches its theoretical capacity at C for 130 cycles, for a total of 200 cycles. At C above cycle seventy, the lowest charge C-rate range obtained for the CV step was C/15 → C/76, for an average of C/46 [Figure 4E]. In fact, charging at C was found to be impossible above cycle 70, contrary to expectations, as charging is usually conducted faster than discharging in all-solid-state batteries. The average lowest C-rate for the discharge correspondent to the CV step was C/24. At C discharge C-rate, the capacity varied from 100 mAh.g^-1 to 120 mAh.g^-1 for a total theoretical capacity of 192 mAh.g^-1 obtained for almost 200 cycles while accounting for the CV step [Figure 4F].

Figure 5. Electrochemical cycling of two cells, Li⁰/Li₆PS₅Cl/NMC_955, with a 57% NMC₉₅₅+38% Li₆PS₅Cl+3% C₆₅+2% SBR cathode simulated as (LiNMC₉₅₅) = Li₁₂Ni₁₀MnCoO₂₄ = LiNi_0.833Mn_0.083Co_0.083O₂ and (Li_0.29NMC₉₅₅) = Li₇Ni₂₀Mn₂Co₂O₄₈ = Li_0.291Ni_0.833Mn_0.083Co_0.083O₂. (A) Simulation of (LiNMC₉₅₅) and (Li_0.29NMC₉₅₅) versus experiments showing pinpoints for the electrochemical cycling, such as cut-offs and capacities; the discrepancies between simulated and experimental data are hypothesized to be due to the difference between simulated and experimental NMC₉₅₅ stoichiometries, impedances and finite discharge rates; (B) two cells showing evidence for the necessity of including a constant voltage step that allows for (Ni⁴⁺ ↔ Ni³⁺ and Co⁴⁺ ↔ Co³⁺) completion. Cell 1 areal capacity is 0.167 mAh.cm^-2, and cell 2 0.434 mAh.cm^-2; the differences in capacities are mostly due to the lack of cycling at constant voltage; (C) Two cells with areal capacity of ~2 mAh.cm^-2 showing evidence for the necessity of including a constant voltage step. The areal capacity is also a determinant of the electrochemical performance and accounts for the differences in capacities and discharge voltages of cells in B and C. SBR: Styrene-butadiene rubber.

In Supplementary Figure 7C, the coulombic efficiency versus cycle number with and without the CV step was calculated considering the previous full charge. Otherwise, while cycling at C, where the charge obtained at constant current is almost null and, therefore, the charge is almost fully obtained at CV, the efficiency would always be > 100% for all cycles between 70 and 200 [Supplementary Figure 7A-C]. After completing the electrochemical cycles, PEIS and cyclic voltammetry tests were conducted [Figure 4G and H]. The PEIS profile changed significantly compared to the initial analysis [Figure 4A]. As shown in Figure 4G, the profile now features three semicircles, resembling the PEIS results obtained before the 2C C-rate electrochemical cycles [Supplementary Figure 7D]. An increase in the R₁ value was observed throughout the impedance tests, indicating intensified asymmetric phenomena at the different interfaces. Similarly, the bulk impedance of the electrolyte increased over successive tests. The third semicircle, visible in Figure 4G and Supplementary Figure 7D, lacks a diffusion component that was observed in the initial impedance test [Figure 4A] suggesting a reduced diffusion, contrasting with the initial test where diffusion was evident. Comparing Figure 4G with Supplementary Figure 7D demonstrates a substantial increase in resistance, which may indicate early signs of cell degradation. This is further supported by the cyclic voltammetry curves [Figure 4B and H, and Supplementary Figure 7E], which, while maintaining similar overall behavior, exhibit a measurable decrease in currents at equivalent voltages.

The direct comparison between experiments and stable solid solutions simulations is shown in Figure 5A, where all pinpoints and morphology of the electrochemical curves are observed and compared with the experiments for C/50 and C/10. The simulated capacity is calculated as the difference between the capacity at 2.60 V (lower voltage cut-off) and the capacity obtained when Li_0.29NMC₉₅₅ becomes the stable solid solution, at 4.05 V, which happens for a capacity of 10 mAh.g^-1. Hence, the simulated capacity is 196 mAh.g^-1 (2%) when compared with the calculated 192 mAh.g^-1 for a final composition of Li_0.3NMC₉₅₅. The resolve to perform CV steps on charge and discharge is evidenced by two Li⁰/LPSCl/NMC₉₅₅ cells in Figure 5B and C, Supplementary Figures 8-10. The difference between cells in Figure 5B and C lies in the amount of active material, which directly affects the areal capacity of the cell. Figure 5B shows electrochemical results for cells with areal capacities ranging from 0.2 mAh.cm^-2 to 0.4 mAh.cm^-2, while Figure 5C presents data for cells with an areal capacity of approximately 2 mAh.cm^-2 [Supplementary Figure 10]. In Figure 5B, cell 1 has a similar areal capacity to cell 2; however, when accounting for the CV step for charge and discharge, a capacity gain of > 3× is obtained in cell 1. A capacity gain of > 2× is obtained for cell 2 when comparing the same number of cycles with and without CV step, where the first cycles were conditioning cycles with CV steps, followed by the same number of cycles without. Although preserving a coulombic efficiency of approximately 100%, the capacity is substantially different, as preconized. Even more substantial evidence is obtained for the cell cycling at a C-rate of C [Supplementary Figure 7B], where all the charge capacity is obtained at a CV. In contrast, the discharge capacity is partially obtained at a constant current, > 120 mAh.g^-1, and part at CV (< 70 mAh.g^-1) [Figure 4F]. Figure 5C shows the behavior of two cells containing 5 and 6 mg of active material, 10× the cells in Figure 5B. Both cell 1 and cell 2 achieve their highest capacity at C/10 after the first electrochemical cycle. After this, the cells experience a decrease in performance across subsequent cycles. In both cases, the capacity exhibits similar trends.

Pre/Post mortem TOF-SIMS analysis

TOF-SIMS is an advanced technique that provides detailed chemical and spatial information about the surfaces^[53]. It works by bombarding the sample with a focused beam of high-energy ions, which release secondary ions that are analyzed to create a mass spectrum of the surface layers. By scanning the ion beam over the sample, spectra are obtained with the distribution of specific molecular species, or masses, from chosen areas of interest. In this study, TOF-SIMS was used to obtain molecular and spatial information about the degradation process of LPSCl/NMC₉₅₅ composite cathodes^[53]. Negative ion mass spectra of the cathode surface (acquired after sputter removal of the top surface, see Experimental section) displayed a multitude of peaks corresponding to secondary ions that can be associated with the different sample components, including Ni_xO_y^- and Co_xO_y^- ions representing the LiNMC₉₅₅ discharged cathode material, Cl^-, S_x^- and PS_x^- ions representing the LPSCl electrolyte and C_xH^- ions representing the SBR binder.

Two cells were analyzed, a pristine and a cell that was cycled between 4.1 V and 2.6 V. The electrochemical profile of the cycled cell is shown in Supplementary Figure 3 and described in detail in Experimental Section. A comparison between the spectra from the uncycled pristine (“Fresh”) and cycled cathode samples revealed a strong intensity increase of the SO₃^- and SO₄^- peaks upon cycling, particularly when compared with the (less oxidized) S₂O^- and S₃^- ions, respectively [Figure 6A]. In contrast, the PS^- and PO₂^- ions remained essentially unchanged upon cycling [Figure 6A]. These observations indicate that some of the most important chemical changes upon cycling are associated with oxidation of sulfur. Additionally, throughout the analysis, the oxidation of Mn³⁺ was not observed, even in the cycled samples. As indicated in Figure 2E, simulations indicate that manganese oxidation occurs above 4.1 V, where irreversible reactions occur. The absence of this oxidation in the analyzed samples highlights the importance of carefully defining charging limits and demonstrates how simulations are crucial in determining them.

Figure 6. TOF-SIMS analysis of “Fresh” (pristine) and cycled electrode samples. (A) Negative ion spectra of pristine and cycled electrode samples focusing on peaks corresponding to PS^-/PO₂^- (left), S₂O^-/SO₃^- (center), and S₃^-/SO₄^- (right); (B) Ion images of secondary ions representing sulfates (top left), LiNMC₉₅₅ (top right), LPSCl (bottom left) and SBR binder (bottom right); (C) Three-colour overlay image (from B) showing LPSCl in red, sulfates in green and LiNMC₉₅₅ in blue. Field of view 100 × 100 µm². TOF-SIMS: Time-of-flight secondary ion mass spectrometry; LPSCl: Li₆PS₅Cl.

The spatial distribution of the different components on the surface of the cycled cathode sample is visualized in the ion images [Figure 6B and C]. These images show a clear spatial separation between ions representing LPSCl and LiNMC₉₅₅, respectively, indicating that the particles of these components are well separated on the sample surface and that their respective localization is confidently visualized in the ion images. Furthermore, the sulfate ion image reveals a close correlation with the LiNMC₉₅₅ ion image, showing that the sulfates are mainly localized to the LiNMC₉₅₅ particles. This is a strong indication that the sulfur oxidation, found to be associated with cycling, mainly occurs by sulfur diffusion to the LiNMC₉₅₅ particles, rather than oxygen diffusion to the LPSCl particles. The latter assertion reinforces the necessity of not charging above 4.1 V to prevent the oxidation of Li_0.3NMC₉₅₅ avoiding reactions with sulfur, among other spurious reactions.

Pre/Post mortem Raman and calorimeter analysis

To evaluate the chemical and structural information of LiNMC₉₅₅ cathodes fabricated by wet processing, Raman measurements were conducted on as prepared and cycled LiNMC₉₅₅ electrodes [Figure 7A]. The Raman peak intensities were normalized on the highest peak intensity from LPSCl.

Figure 7. Analyses on pristine and post-mortem cycled cells. (A) Raman spectroscopy analysis as-prepared (pristine) and cycled cathode material; (B) Heat flow analysis on cell from Figure 4 at 30 °C considering different cycling states: (1) OCV (without external resistor), (2) R_ext = 217 kΩ, (3) OCV, and (4) R_ext = 553 kΩ. Schematics of the battery calorimeter set up on the right. OCV: Open circuit voltage.

The main vibration modes for LPSCl show up as a prominent signal at 423 cm^-¹, due to the symmetric stretching of (PS₄^3-), along with additional peaks at 197 cm^-¹ and 265 cm^-¹ and a band between ~500 cm^-1 and 600 cm^-1, which come from asymmetric stretching of (PS₄^3-), as explained by Taklu et al.^[54]. For LiNMC₉₅₅, most Raman-active vibrations are found in the 400-600 cm^-1 range, with a strong signal approximately ~570 cm^-1 from the A_1g mode of Ni, as reported by Li et al.^[55]. There are also smaller signals between ~470 cm^-1 and 610 cm^-1. A broader signal appears between ~1,200 cm^-1 and 1,700 cm^-1 where the presence of D and G bands arising from the carbon materials are observed^[56-58], which could come from the carbon-coated aluminum foil supporting the cathode or from the conductive filler. This variety makes it difficult to pinpoint the exact source of these signals.

By analyzing Figure 7A, it becomes clear that most signals from the cathode’s main components overlap in the 500-600 cm^-1 range, making it challenging to assign these peaks specifically to the active material or the argyrodite phase, a common issue in the literature^[56,59,60]. Even so, the main signal from the argyrodite phase remains clearly visible in both the as-prepared and cycled Raman spectra, confirming its presence in the cathode composite, already demonstrated with TOF-SIMS. Interestingly, the as-prepared cathode shows a stronger signal for the LiNMC₉₅₅ phase at ~570 cm^-1 compared to the cycled version, which shows a lower overall intensity in the 500-600 cm^-1 band. This difference may be due to degradation of the cathode material during cycling, reacting with the argyrodite LiPSCl, or to the potential damage while removing the cycled cathode from the pellet. In fact, the difference may be due to the formation of SO₃^- and SO₄^- on the surface of LiNMC₉₅₅, as shown by TOF-SIMS in Figure 6. The calorimeter analysis on the cell from Figure 4 obtained while discharging from 3.7 V shows that the higher the discharge current the greater the amount of heat released [Figure 7B]. This may indicate that oxygen is mostly released immediately after the beginning of discharge at higher C-rates (exothermic reaction), as shown in the inset of Figure 4B. More than 2.2 h are necessary to start stabilizing the heat flow at ~C/5, with the first hour corresponding to a pronounced exothermic variation of 14.6 μW.