Electrolyte Reactivity at the Charged Ni-Rich Cathode Interface and Degradation in Li-Ion Batteries

NMC111 and NMC811 cathodes and LTO anodes were prepared at large-scale in the Cell Analysis, Modeling, and Prototyping (CAMP) facility at Argonne National Laboratory. The cathodes consisted of 90 wt % NMC (NMC111, Toda; NMC811, Targray), 5 wt % polyvinylidene difluoride binder (PVDF, Solvay 5130), and 5 wt % conductive carbon (Timcal C45) cast onto 20 μm thick aluminum foil using N-methyl-2-pyrrolidone (NMP) as the solvent. The NMC111 and NMC811 cathode sheets had loadings of 10.10 mg_NMC cm^–2 and 8.21 mg_NMC cm^–2, respectively, corresponding to ∼1.48 mAh cm^–2 based on 147 mAh g^–1_NMC for NMC111 and ∼1.52 mAh cm^–2 based on 185 mAh g^–1_NMC for NMC811. The anodes consisted of 87 wt % LTO (Samsung Fine Chemicals), 5 wt % conductive carbon (Timcal C45), and 8 wt % PVDF binder (Kureha 9300) cast onto 20 μm thick aluminum foil using NMP as the solvent. The anode sheets had loadings of 12.27 mg_LTO cm^–2 corresponding to ∼1.96 mAh cm^–2 based on 160 mAh g^–1_LTO. After drying, electrodes were calendered using a heated (80 °C) two-roller hydraulic-driven roll press (A-PRO Co.) to 30% porosity. Circular electrodes with 14 mm (cathode) and 15 mm (anode) diameters were punched and dried at 120 °C for at least 12 h under dynamic vacuum before being transferred to an Ar filled glovebox (<0.5 ppm of O₂ and H₂O, MBraun). The electrode capacity balancing of anode and cathode (N/P ratio) was set to ≈1.3:1.0.

NMC111 powder (Toda) was annealed in an alumina crucible at 750 °C for 4 h with a 5 °C min^–1 heating rate under air atmosphere using a muffle furnace (MTI). After natural cooling, a slurry consisting of 90 wt % annealed NMC111, 5 wt % PVDF binder (PI-KEM), and 5 wt % conductive carbon (C45, PI-KEM) was cast onto 20 μm thick aluminum foil using NMP as the solvent. The annealed NMC111 cathodes had loadings of 8.0 mg_NMC cm^–2 corresponding to ∼1.2 mAh cm^–2 based on 147 mAh g^–1_NMC.

LiMn₂O₄ (LMO) cathodes consisted of 90 wt % LMO (Sigma-Aldrich, < 0.5 μm particle size (BET), > 99%), 5 wt % PVDF binder (PI-KEM), and 5 wt % conductive carbon (C45, PI-KEM) cast onto 20 μm thick aluminum foil using NMP as the solvent. The LMO cathodes had loadings of 3.3 mg_LMO cm^–2 corresponding to ∼0.5 mAh cm^–2 based on 148 mAh g^–1_LMO.

The baseline electrolyte was 1 M LiPF₆ in ethylene carbonate (EC):ethyl methyl carbonate (EMC) 3:7 (by weight, LP57, SoulBrain). Single solvent electrolytes were 1.5 M LiPF₆ (99.9%, Solvionic) in EC (anhydrous 99%, Sigma-Aldrich) and 1.5 M LiPF₆ in EMC (99.9%, Solvionic).

The 2032-type coin cells (Hohsen) were assembled in a full cell setup with a 14 mm diameter cathode, 15 mm diameter anode, and 16 mm diameter Celgard 2325 separator (PI-KEM) or glass fiber separator (grade GF/A, Whatman) soaked in 40 or 80 μL of electrolyte, respectively. Separators were dried for at least 24 h under dynamic vacuum at 60 or 120 °C, respectively. Three-electrode PAT cells (EL-Cell) were assembled with 18 mm diameter cathode and anode, glass fiber separator (260 μm thickness, grade GF/A) soaked in 100 μL of electrolyte, and a lithium metal ring electrode set in an insulation sleeve (EL-Cell).

After assembly, the cells were charged galvanostatically at C/20 (assuming a practical capacity of 147 mAh g^–1_NMC for NMC111 and 185 mAh g^–1_NMC for NMC811) to 3.05 V (vs LTO) and held at that voltage for 60 h while the current was recorded, after which they were discharged at C/20 to 1.45 V (Biologic VMP3 or BCS 805 series). For simplicity, this protocol is referred to as the “60 h voltage hold (VH) protocol.” Subsequently, the potential-dependent impedance of the NMC cathode was measured in the three-electrode PAT cells by charging the NMC cathode at C/20 to various potentials vs. the lithium reference electrode: 3.8, 4.1, 4.3, 4.5, and 4.6 V for NMC111 and NMC811 and also 4.7, 4.8, and 4.9 V for NMC111. The cell was held at each potential for 1 h to reach a steady-state and then allowed to rest at open circuit potential (OCP) for 1 h before the EIS was measured. Potential-controlled EIS was conducted in a frequency range of 500 kHz to 10 mHz with an AC voltage perturbation of 5 mV (Biologic VMP3). The SOC of NMC at each potential was calculated based on the charge passed and a theoretical capacity of 277.9 mAh g^–1 for NMC111 and 275.5 mAh g^–1 for NMC811. All electrochemical protocols were performed in climate chambers set at 25 °C. Two or more cells were evaluated for each condition to ensure reproducibility, which is indicated by error bars in the respective figures.

The OEMS system consists of a stainless-steel tube carrying gas through the electrochemical cell (Swagelok type), connected through self-sealing quick-connects (Beswick Engineering) so the system is never exposed to air, to a mass spectrometer. Mixtures of several gases can be selected via mass flow controllers (Bronkhorst), with a pressure controller (Bronkhorst) maintaining the gas line at a constant 1.1 bar(a). For this work, Ar (BOC N6.0), connected to a purifier (Bronkhorst) before flow control, was used as the carrier gas. The gas line is connected to a quadrupole mass spectrometer (Pfeiffer) through a heated capillary (120 °C) to prevent condensation. A potentiostat (Ivium) controls the electrochemical operations.

Calibration was performed by flowing through the system a mixture of H₂, CO, C₂H₄, O₂, and CO₂ (1000 ppm each) in Ar (BOC N6.0) to establish a correlation between channel ion currents and gas molar flow. The molar flow of CO was established by subtracting the proportional amount of CO₂ measured in channel m/z = 44 to channel m/z = 28 according to the relative intensities of both channels in the spectral pattern of CO₂.

Cells for OEMS were assembled in half cell setup with an 18 mm diameter cathode, 15.6 mm diameter metallic Li chip counter/reference electrode (0.25 mm thickness, LTS Research Laboratories), and 25 mm diameter glass fiber separator (Whatman, GF/B) soaked in 300 μL of electrolyte. In some experiments, a 25 mm diameter lithium ion conducting glass-ceramic (LICGC) separator (0.15 mm thickness, Ohara AG-01) was placed between two 22 mm diameter glass fiber separators to prevent the migration of decomposition species in the electrolyte to the opposite electrode. After assembly, the cells were connected to the OEMS system and the potentiostat and held at the OCP for 6 h before starting the electrochemical protocol. The cells were cycled using an analogous protocol to that described for the coin cells above. Specifically, the cells were charged galvanostatically at C/20 to 4.6 V vs Li/Li⁺ and held at that potential for 40 h, after which they were discharged at C/20 to 2.5 V vs Li/Li⁺ and held at OCP for at least 12 h.

The surface area of the NMC powders was determined by nitrogen gas physisorption at 77 K, measuring isothermally at 10 points between 0.07 ≤ p/p₀ ≤ 0.30 (3Flex, Micromeritics). The water content of the electrolytes was measured by Karl Fischer titration (899 Coulometer, Metrohm). NMR experiments were conducted on a Bruker Avance III HD spectrometer equipped with a 11.7 T magnet (ν_1H = 500 MHz) using a BBO probe. ¹H, ¹⁹F, and ³¹P NMR spectra were acquired using a one pulse pulse sequence; ¹⁹F and ³¹P experiments were conducted without the use of ¹H decoupling. ¹H chemical shifts were referenced to the DMSO-d₆ solvent peak at 2.50 ppm. ¹⁹F and ³¹P chemical shifts were referenced to LiPF₆ at −74.5 and −145.0 ppm, respectively. Pristine electrolyte was measured by pipetting 40 μL of the electrolyte solution into 0.7 mL of DMSO-d₆ (99.9 atom % D, 99% CP, Sigma-Aldrich), which was transferred to an airtight NMR tube fitted with a Young’s tap.

After the 60 h VH protocol, the NMC/LTO coin cells were disassembled in an Ar filled glovebox. The separator was extracted and soaked in 0.7 mL of DMSO-d₆ for 5 min. The solution was transferred to an airtight NMR tube fitted with a Young’s tap for measurement. The NMC and LTO electrodes were extracted, rinsed with 1 mL of dimethyl carbonate (DMC, anhydrous, ≥99%, Sigma-Aldrich), and vacuum-dried at ambient temperature for 1 h prior to measurement by XPS and HRTEM.

XPS measurements were performed using a Thermo Scientific Nexsa X-ray photoelectron spectrometer system utilizing Al Kα X-rays. The electrodes were transferred inertly into the system without air exposure. Energy calibration was performed by setting the carbon black feature in the C 1s core level region to 284.4 eV for the NMC electrodes and the Ti 2p_3/2 peak to 459.3 eV for the LTO electrodes. A Shirley type background was subtracted from all spectra besides the region containing the transition metal 3p core levels. The probing depth corresponding to the intensity of 95% of the emitted photoelectrons was calculated according to²⁷

where d is the probing depth, λ is the inelastic mean free path, and θ is the electron take off angle, i.e., the angle between the sample surface and the analyzer. The inelastic mean free path was calculated to 3.38 nm based on the photoelectrons emitted from the Ti 2p orbital traveling through a SEI consisting of polyethylene and 3.19 nm for photoelectrons emitted from the O 1s orbital traveling through a CEI consisting of polyethylene.²⁸ The equation assumes the surface is flat and the chemical composition is known and homogeneous. This is not the case for the electrodes in this study, although it should still serve as a reasonable approximation.

The NMC cathode materials (including NMC111 or NMC811 particles, conductive carbon, and PVDF) were scratched from the NMC electrode on an aluminum current collector and ground with an agate pestle and mortar. The powder material was then transferred onto the holey carbon copper grid in an Ar filled glovebox. A single tilt holder was used for TEM (JEM-2100Plus, JEOL) characterization.

Elemental analysis was performed using inductively coupled plasma-optical emission spectroscopy (ICP-OES, Thermoscientific) calibrated with standards prepared from an ICP multielement solution (VWR, Aristar). For ICP-OES analysis, NMC/LTO coin cells were constructed with 120 μL of electrolyte and either three separators (Celgard, glass fiber, Celgard; for LP57 and EMC electrolyte) or one separator (glass fiber; for EC electrolyte). The higher volume of electrolyte (transferred within the glass fiber separator) enables the extraction of a sufficient volume for analysis, while the Celgard separator facilitates easier separation of the electrodes from the glass fiber. After cycling with the 60 h VH protocol, the cells were disassembled in an Ar filled glovebox. The LTO anode and separators were extracted, placed in a 15 mL polypropylene tube and centrifuged at 5000 rpm for 10 min. The LTO anode and the separators were removed and 4.0 mL of ∼2% nitric acid (diluted from concentrated nitric acid; 67–69%, trace metal grade, Fischer Chemical) was added to the extracted electrolyte (34–70 μL) before analysis. The LTO coating was scraped from the current collector (19–26 mg_LTO) and soaked in 248 μL of 5.1 M nitric acid for 5 days before being diluted to 4.0 mL with 3.75 mL of water for analysis. The LTO coating scraped from uncycled electrodes was measured as a baseline.

Figurea,b shows the voltage profile and current trace for an NMC811/LTO cell with LP57 electrolyte (1 M LiPF₆ in EC/EMC 3/7 by volume) during the first charge–discharge cycle with a 60 h potentiostatic hold at 3.05 V, referred to henceforth as the “60 h voltage hold (VH) protocol”. The three-electrode cell measurement in Figurec illustrates that LTO exhibits a potential plateau at 1.55 V vs Li/Li⁺ during charging (lithiation), and therefore at a full cell voltage of 3.05 V, the NMC potential is 4.6 V vs Li/Li⁺.

Figured shows the oxidation current (normalized to the NMC BET surface area, Table S1↗) during the potentiostatic hold with NMC111 and 811 and for three electrolyte solutions. A conventional LP57 electrolyte is tested alongside two model, single solvent electrolytes, 1.5 M LiPF₆ in EMC and 1.5 M LiPF₆ in EC. Potential profiles for all the conditions tested are shown in Figure S1↗. Owing to the lower delithiation potential of the Ni-rich NMC811 compared to NMC111, the amount of capacity extracted in the charge to 3.05 V, and hence the state-of-charge (SOC), is greater for NMC811 (90% SOC) than for NMC111 (76% SOC), as shown in Figure S2↗. A 1.5 M salt concentration was used for the EMC electrolyte to improve the ionic conductivity² and for the EC electrolyte to lower the viscosity,²⁹ and glass microfiber separators (grade GF/A) were used in cells with the EC electrolyte since it does not wet Celgard 2325 polymer separator. Control experiments for the influence of the salt concentration (1.0 and 1.5 M) and separator (Celgard and GF/A) on the current in the potentiostatic hold are provided in Figure S3↗ and show equivalent electrochemical behavior. The water contents in the three electrolytes as measured by Karl Fischer titration are given in Table S2↗.

In the first 20–30 h of the potentiostatic hold, the current decays rapidly as the electrolyte polarization relaxes and as the concentration of lithium in the bulk of the NMC particles reach the equilibrium value set by the applied potential. At later times, the current is dominated by oxidative decomposition reactions at the electrolyte–NMC interface. To quantitatively compare the stability of each electrolyte at the NMC interface, the average current in the final 20 h of the potentiostatic hold is plotted in Figuree. For NMC111, the average current is largely independent of the electrolyte, at a value of 0.40(3) mA m^–2_NMC. With LP57, NMC811 results in a 21% higher average current (0.50(3) mA m^–2_NMC), in line with recent literature reporting poorer cathode surface/oxygen stability for Ni-rich NMCs at an equivalent cathode potential.^4,30 Further, for NMC811, the current is dependent on the electrolyte solvent(s) and is 18% lower for EMC electrolyte and 23% higher for EC electrolyte compared to LP57. As mentioned earlier, the current measured can have contributions from lattice oxygen release and electrolyte oxidation, which releases gases and produces soluble and insoluble electrolyte degradation products. The gas, liquid, and solid phases resulting from these processes are next characterized in situ by OEMS and ex situ by solution-state NMR on extracted electrolyte and XPS on the extracted electrodes. TM dissolution from NMC is also investigated by ICP-OES on the extracted electrolyte and anodes.

For reasons explained in Supplementary note S1↗, an NMC/Li half-cell is used for the OEMS experiments. In brief, decoupling the cathode and anode gas evolution from NMC/Li cells is more straightforward than from NMC/LTO cells; a lithium ion conducting glass-ceramic separator (Ohara, LICGC) is used for some experiments to decouple the anode from cathode processes. The current rate and potential applied to the NMC cathode (shown in Figure S4↗ for all the conditions tested) were kept the same as in Figure, although the potentiostatic hold time was reduced to 40 h since most of the gas evolution took place at times <40 h. Gas evolution profiles are shown in Figure as a function of potential for charge and discharge and time for the potentiostatic hold. For NMC111 with LP57 electrolyte, the onset for CO₂ evolution is ∼3.8 V vs Li/Li⁺. This is characteristic of the oxidation of carbonate impurities^{31 −33} that form on the surface of NMC particles during long-term storage (even in a dry room environment) via reactions with carbon dioxide and water.^34,35 These surface contaminants can (largely) be removed via thermal treatment,³⁶ and additional OEMS experiments conducted on electrodes prepared with annealed NMC111 powder (750 °C for 4 h in air) yield a CO₂ onset potential of ∼4.4–4.5 V vs Li/Li⁺ (Figure S5↗). While the NMC111 cathodes appear to have “aged”, NMC811 does not evolve CO₂ at potentials below 4.4 V vs Li/Li⁺ indicating that the NMC811 sample is largely free of surface impurities. It is well understood that the high surface reactivity of Ni-rich cathodes make them more prone to the formation of surface impurities during storage,^37,38 and hence it is likely that the storage conditions and/or time are different for the NMC111 and 811 electrodes used in this work.

At ∼4.4 V vs Li/Li⁺, and independent of the electrolyte, the signals from CO, O₂, and CO₂ begin to rise simultaneously for NMC811 (Figure). Gasteiger and co-workers also observed this phenomenon^6,10 and have shown that reactive lattice oxygen (e.g., singlet oxygen) released from NMC reacts with the electrolyte solvent producing CO and CO₂.^11,12 O₂ detected in the OEMS experiment may arise from the ground state (i.e., triplet) O₂ release from the lattice and/or from deactivation of ROS via nonradiative (electronic-to-vibrational coupling to solvent molecules) and radiative transition to the ground state.¹² The signals from CO, O₂, and CO₂ continue to rise before reaching a maximum at, or shortly after (<3 h), the start of the potentiostatic hold. For LP57, the total quantity of each gas released, shown in Figurea, is dependent on the NMC composition: the amount of CO and CO₂ evolved are 8.5 and 6.4 times higher for NMC811 (29.7 and 204 μmol m^–2_NMC, respectively) compared to NMC111 (3.5 and 31.7 μmol m^–2_NMC, respectively), with the amount of CO₂ for NMC111 over-represented due to CO₂ evolution from the surface impurities present. Note that differences in the slurry preparation and electrode manufacture limit direct comparison between the annealed NMC111 data in Figure S5↗ and the other conditions in Figure. Further, O₂ evolution is not detected from NMC111 (Figurec and Figure S5↗) but is clearly evident for NMC811 (0.7 μmol m^–2_NMC), which is likely related to the different NMC SOC at 4.6 V vs Li/Li⁺ (i.e., 90% for NMC811 vs 76% for NMC111, Figure S2↗). However, the onset of CO₂ release at ∼4.4–4.5 V vs Li/Li⁺ for annealed NMC111 is likely indirect evidence of a small amount of lattice oxygen release. This is consistent with a ∼ 4.6 V vs Li/Li⁺ onset potential for O₂ evolution from NMC111 reported previously.¹⁰ It should be noted that purely electrochemical oxidation of electrolyte solvents also produces CO₂ and CO gas^10,39,40 (oxidation onset potentials in the range 4.5–6.0 V vs Li/Li⁺ have been reported^{41 −46}) and therefore may contribute to the gas evolution in Figure and Figure S5↗. Excluding any catalytic effect of Ni (ruled out in experiments reported by Jung et al.^6,10), the gas evolution with NMC111 provides an upper limit for the contribution from direct electrochemical oxidation. Therefore, the majority of the CO₂ and CO evolution with NMC811 stems from chemical oxidation pathways or a coupled chemical and electrochemical process. Even if we assume no lattice oxygen release from NMC111 at 4.6 V vs Li/Li⁺, we estimate that electrochemical oxidation can account for at most ∼12% of the evolved CO₂ for NMC811, in line with calculations by Jung et al.⁶

Figures and 3a also show that the gas evolution with NMC811 is dependent on the electrolyte. Compared to NMC111, enhanced CO and CO₂ gas evolution is observed for NMC811 with LP57, EC, and EMC only electrolytes, indicating that reactive lattice oxygen reacts with both cyclic and linear carbonates, although the quantity and ratio of gases vary. The ratio of each gas evolved with EC relative to EMC electrolyte (EC/EMC, Figureb) is particularly insightful as it probes the relative gassing behavior of the individual solvents. With EC electrolyte, the amount of CO₂ evolved (225 μmol m^–2_NMC) is 1.9 times higher than with EMC electrolyte (117 μmol m^–2_NMC), while LP57 (204 μmol m^–2_NMC) shows only a slight reduction compared to EC electrolyte. A similar trend is observed for O₂, with 1.6 times more O₂ with EC electrolyte (0.886 μmol m^–2_NMC) relative to EMC electrolyte (0.560 μmol m^–2_NMC), although the quantities of O₂ are more than 10² times lower than CO₂ and therefore more susceptible to errors. Conversely, the amount of CO evolved is marginally higher for EMC electrolyte (38.7 μmol m^–2_NMC) compared to that for EC electrolyte (32.5 μmol m^–2_NMC) and LP57 (29.7 μmol m^–2_NMC), yielding an EC/EMC fraction of 0.8. The similar quantity of each gas evolved with LP57 and EC electrolyte, along with the equivalent CO₂/CO fraction of 6.9 (compared to 3.0 for EMC electrolyte, Figurec) strongly suggests that the NMC-induced gassing in LP57 is dominated by EC.

Taking an O₂/CO₂ mole ratio of 1:1 for the reaction of EC and EMC with reactive oxygen (as proposed by Jung et al. for EC,⁶ and as proposed for EMC in the Discussion section (see Scheme), the higher fraction of CO₂ evolved in the EC electrolyte (1.9 times) indicates significantly more lattice oxygen release from NMC811 with EC-containing electrolytes. This finding is also supported by the proposed O₂/CO mole ratios and the relative quantities of CO produced, which will be discussed in further detail in the Discussion section.

Turning now to the H₂ evolution in Figurea, for the EC-containing electrolytes, there is a clear onset potential between 4.4 and 4.6 V vs Li/Li⁺, similar to that seen for CO, O₂, and CO₂. H₂ is then evolved throughout the potentiostatic hold and only stops once the NMC potential drops below ∼4.2 V vs Li/Li⁺ on discharge. Reduction of trace water in the electrolyte at the anode (Table S3↗) cannot account for this trend since (i) with a lithium metal anode this process would be potential independent, and (ii) the total quantity of H₂ evolved in each electrolyte is >30 times that expected from the measured trace water content alone; see the calculations in Supplementary note S2↗. Instead, the observed H₂ evolution is likely the result of electrode crosstalk, in which protic oxidation species formed at high SOCs at the cathode diffuse to the anode where they are reduced.⁴⁰ The identity of these protic species could include water, protons, and protic electrolyte solvent oxidation fragments.^6,47,48 Mechanisms for the formation of these species are discussed further in the Discussion section. The sustained evolution of H₂ throughout the potentiostatic hold indicates that protic species are continuously produced when the NMC is above the threshold potential.

Unfortunately, the relative quantity of H₂ evolved with different electrolytes is not necessarily an accurate measure of the amount of protic species formed at the cathode. A more effective SEI on the anode, such as that formed with EC-containing electrolytes or other specialized additives, hinders the reduction of protic species⁴⁰ which may then react via alternate pathways. This likely explains the much larger H₂ evolution with EMC electrolyte (Figurea, 3.8 times that with LP57), since EMC is a poor SEI former on lithium metal and graphite anodes, and in the latter case forming an SEI that is nonuniform and thinner compared to an EC-based SEI.⁴⁹ To prove this hypothesis, a NMC811/Li cell was built with a lithium ion conducting glass-ceramic separator to block the liquid-state diffusion of protic species to the anode. EMC electrolyte was used as the catholyte and LP57 as the anolyte. Schematics of the cell stack with and without the Ohara glass separator are shown in Figure S6↗, and the potential profiles are shown in Figure S4↗ b. As expected, the total amount of CO₂ evolved is very similar between the two runs with EMC electrolyte (117 μmol m^–2_NMC and 108 μmol m^–2_NMC with and without the Ohara glass separator, respectively), which validates the comparison. As shown in Figurea, with the glass-ceramic separator and LP57 anolyte, the quantity of H₂ detected is significantly lower, although the detection of a small amount of H₂ suggests an imperfect seal between the two compartments of the cell. Finally, we note that variability in the treatment of the lithium metal anode (e.g., scraping) prior to the experiment may also introduce variability in the active surface area and hence the rate of H₂ evolution between experiments with the same electrolyte.

The potential/SOC-dependent impedance of NMC after the 60 h VH protocol was measured in three-electrode format, controlling the potential of the NMC versus the lithium metal reference electrode. The electrochemical protocols applied for NMC111 and 811 are shown in Figure S7↗. Figurea–d shows the Nyquist plots for NMC111 with LP57 (NMC111 with EC and EMC electrolytes are shown in Figure S8↗) and for NMC811 with LP57, EC, and EMC electrolytes—the NMC potential for each spectrum is indicated in Figurea–d and the corresponding SOC is provided in Table S4↗. Three features are evident (labeled in the inset in Figureb), a high frequency semicircle (hf), a midfrequency semicircle (mf), and a Warburg impedance tail at low frequencies (lf). The diameter of the midfrequency semicircle, which can be attributed to the electrolyte-oxide interfacial impedance,^8,50 was extracted by fitting a simplified equivalent circuit to the data (see Supplementary note S3↗) and is plotted as a function of potential and SOC in Figuree,f. The interfacial resistance shows a huge growth at potentials >4.1 V vs Li/Li⁺ for NMC811 and >4.5 V vs Li/Li⁺ for NMC111 (Figuree). This apparent difference is largely due to NMC811 approaching the delithiated state at lower potentials than NMC111, which is accounted for in Figuref, where the resistance is plotted as a function of Li content. For fully delithiated NMC a quasi-infinite charge transfer resistance is expected, referred to as a blocking condition,⁵¹ justifying the general trend seen in the data. Nevertheless, at high SOCs (x > 0.7 in Li_1–xTMO₂), the impedance is dependent on both the NMC composition and the electrolyte. With LP57, the impedance at x > 0.7 for NMC811 is higher than NMC111, e.g., at x = 0.84(1) the impedance of NMC811 (68.8 Ω.cm²) is 4.5 times that of NMC111 (15.2 Ω cm²), see the inset in Figuref. In terms of the electrolyte dependence, for x > 0.7, NMC811 with LP57 and EC electrolyte exhibit similar impedance values, while with EMC electrolyte the impedance is 70–80% lower, e.g., at x = 0.85(1) the impedance with EC electrolyte (67.8 Ω.cm²) is 3.6 times that of EMC electrolyte (18.9 Ω cm²) (inset in Figuref). This indicates that the impedance growth in LP57 is dominated by the EC solvent contribution. The EC in LP57 was also found to dominate the gassing behavior with NMC811, which stems from lattice O₂ release (see above). This suggests a correlation between impedance and gas evolution, which will be explored in the Discussion section.

TEM was used to study the interfacial structure of NMC nanoparticles in the pristine state and after the 60 h VH protocol. The HRTEM images in Figurea,e show the layered structure of pristine NMC111 and NMC811, respectively, with the layered structure being confirmed by the corresponding fast Fourier transformation (FFT) images. Pristine NMC811 also has some layers with cation mixing between the Li and TM layers, as shown in Figuree, likely stemming from the higher fraction of Ni²⁺ and the propensity for Li/Ni site-disorder; dislocations/grain boundaries are also seen. Formation of a surface reconstruction layer (SRL) was observed on the surface of NMC111 (Figureb–d) and NMC811 (Figuref–h) after the 60 h VH protocol with LP57, EC electrolyte, and EMC electrolyte. The EC electrolyte shows a thicker SRL on the surface of NMC111 than EMC electrolyte or LP57, including rock-salt structure and cation mixing layer, as shown in Figurec and the corresponding FFTs. In comparison, there are only fine SRL structures, mainly a cation mixing layer of 2–4 nm in thickness, on the surface of the NMC111 for LP57 and EMC electrolyte, as shown in Figureb,d and their corresponding FFTs, respectively. However, some regions with cation mixing can be seen in the bulk area of NMC111 with LP57, as shown in Figureb.

Compared with NMC111, more and thicker SRLs form on the surface of NMC811 after the 60 h VH protocol with LP57, EC electrolyte, and EMC electrolyte, as shown in Figuref–h, respectively. LP57 electrolyte leads to the formation of a thick rock-salt structure on the surface of NMC811, as shown in Figuref and the corresponding FFT. Figuref also shows the probable formation of cathode electrolyte interface (CEI) on the surface of NMC811 in LP57 solution. A thick rock-salt structure can also be seen on the surface of NMC811 with the EC electrolyte, as shown in Figureg and the corresponding FFTs, and some parts of the particle have a cation mixing layer. In contrast, EMC electrolyte leads to a uniform rock-salt structure and cation mixing layer with a thickness of 3–5 nm on the surface of NMC811, as shown in Figureh and the corresponding FFTs. While we recognize that only a minute fraction of the electrode material is sampled by HRTEM, the results shown are representative of many particles sampled at random from the electrode. Moreover, they are in accord with the electrochemistry and gas analysis studies described above.

Figure shows the XPS spectra of the NMC electrodes after the 60 h VH protocol using different electrolytes. Comparing NMC111 and 811 with LP57, the O 1s spectra in Figurea show that the lattice oxygen peak (∼529.4 eV) is much higher in intensity for the NMC111 electrode compared to the NMC811 electrode, indicating a thinner CEI is formed for NMC111. In the corresponding P 2p spectra shown in Figurec, a peak is seen at 135 eV, likely coming from degradation of the LiPF₆ salt into Li_xPO_yF_z compounds, which is more intense for NMC111 than 811 in LP57. Given the thinner CEI of the NMC111, it may be that the phosphorus species are less buried beneath the organic CEI components and/or that the average phosphorus content of the CEI is increased. In either case, the larger phosphorus signal close to the CEI surface is consistent with the hypothesis that the degraded salt stabilizes the electrode preventing further degradation from organic compounds. In the region containing the transition metal 3p core levels, peaks corresponding to cobalt (∼61 eV) and manganese (∼49 eV) are clearly apparent for NMC111 in contrast to 811 where these are barely detectable above the background, consistent with the lower fractions of Mn and Co in NMC811. The same trends observed in Figure for LP57 (i.e., a thinner CEI yet more phosphorus species for NMC111) are also observed for EC electrolyte in Figure S10↗.

Comparing the role of the electrolyte on the Ni-rich NMC surface, the lattice oxygen peaks for all NMC811 electrodes have a similar intensity suggesting no major difference in the CEI layer thickness, although the LP57 electrolyte may result in a slightly thicker CEI since the oxygen peak intensity is slightly lower. As the lattice oxygen is clearly visible for all samples, the CEI thickness should be thinner than the probing depth of ∼10 nm. NMC811 cycled with EC electrolyte shows much higher peak intensities associated with organic oxygen species, phosphorus, and lithium compared to any of the other electrodes. Furthermore, the Ni 3p region in Figured shows a high binding energy shoulder at ∼70 eV for NMC811 with EC electrolyte that is not apparent for the other solvents. This indicates a chemical change in the Ni close to the NMC surface and has previously been observed to coincide with ReSL formation during long-term cycling of NMC811,⁵² potentially corresponding to formation of a Ni–F environment.⁵³ These observations indicate that the EC electrolyte is more reactive toward the NMC811 electrode than the other electrolytes, which is in agreement with the electrochemical data shown in Figured,e.

The C 1s spectra in Figureb are rather similar for all electrolytes, although the intensity from the carbon black is lower for NMC811 cycled with EC electrolyte. The combination of showing more degradation products, similar CEI thickness on NMC811, and a lower intensity for the carbon black peak suggests that the EC is more prone to react and cover the carbon black.

Pristine electrolyte and the electrolyte extracted from NMC/LTO cells after the 60 h VH protocol were characterized using ¹H, ¹⁹F, and ³¹P NMR spectroscopy. Assignments of the observed NMR signals, listed in Table, are made based on results reported in the literature.^{19,48,54 −59} The ¹H NMR spectra for deuterated DMSO (the solvent used to extract the electrolyte and degradation products) and the pristine electrolytes are shown in Figures S11 and S12↗, respectively. Figures and 8 show the ¹H, ¹⁹F, and ³¹P NMR spectra of the cycled LP57 (top), EMC electrolyte (middle), and EC electrolyte (bottom) with NMC111 (left) and 811 (right). In Figure, signals present in the cycled electrolyte, but absent in the pristine, are labeled in green or red depending on whether they arise from EMC or EC degradation, respectively. The color of the labels for LP57 are based on whether the chemical shift of the particular signal matches with that observed for EMC or EC electrolyte.

Starting with the ¹H NMR of EMC electrolyte (Figurec,d), the intense signals at 4.12 ppm (q) and 3.69 ppm (s) are from EMC, while those at 3.34 ppm (s, water) and 3.32 ppm (s), which are present in all pristine and cycled electrolytes, appear to be introduced from impurities in DMSO (see Figure S11↗). The degradation products identified in EMC electrolyte for NMC111 and 811 are methanol (3.18 ppm, d),⁵⁴ polyethylene oxide (EO) based oligomers likely containing carbonate groups (ROCOOCH₂CH₂OR′; multiple peaks in the region 3.38–3.62 ppm; 3.79 ppm, s),⁵⁵ and simple acetal species (RCH(OR₂); possibilities include methanediol, 1,1-ethanediol, methoxymethanol, and 1-methoxyethanol; 5.80 ppm, s; 5.70 ppm, s; 5.68 ppm, s).¹⁹ A doublet at −82.8 ppm in the ¹⁹F NMR spectra with NMC111 and 811 (Figuree,g) indicates the formation of PO₂F₂^–,^48,56,57 but the expected triplet at −16.6 ppm in the ³¹P NMR spectra⁵⁷ (Figuref,h) is not observed, presumably because the quantity present is below the detection limit. Similar electrolyte degradation products for NMC111 and 811 suggest that the reaction pathways are largely independent of Ni content. However, NMC111 has fewer signals from poly-EO based oligomers and a lower PO₂F₂^–/PF₆^– peak area fraction (Table S5↗), indicating less EMC solvent decomposition and less LiPF₆ salt breakdown with the lower Ni cathode.

The ¹H NMR spectra of EC electrolyte (Figurese,f) have an intense signal at 4.48 ppm (s) from EC and a trace EMC impurity (marked with asterisks and also present in the pristine electrolyte, Figure S12↗). Signals at 4.08, 3.57, 3.51, and 3.39 ppm (labeled in black) are also present in the pristine electrolyte and are attributed to hydrolysis products of EC, specifically LEMC¹⁹ and poly-EO based oligomers⁵⁵ and/or ethylene glycol.⁶⁰ The degradation signals with NMC111 and 811 are assigned to poly-EO based oligomers likely containing carbonate groups (3.62 ppm, m; 3.81 ppm, m; 4.20 ppm, m),⁵⁵ aldehyde species (RCHO; possibilities include formaldehyde, acetaldehyde, and glyoxal; 9.61 and 2.13 ppm, weak),^19,58 and oxyfluorophosphate salts (3.96 ppm, d; 3.98 ppm, s; e.g., OPF_x(OR)_y).^57,59 A singlet at 7.77 ppm is observed with NMC811 but absent with NMC111, which has been assigned to vinylene carbonate (VC).⁶¹ Signals for PO₂F₂^– are evident in the ¹⁹F and ³¹P NMR spectra in Figurei–l.^48,56,57 The detection of signal from PO₂F₂^– in the ³¹P spectra of the EC electrolyte but not the EMC electrolyte, suggests more LiPF₆ salt decomposition in EC electrolyte. As was found for EMC electrolyte, the reaction pathways for EC electrolyte decomposition appear to be independent of Ni content. Signs of less EC solvent decomposition and less LiPF₆ salt breakdown with the lower Ni cathode are again seen via fewer signals from poly-EO based oligomers and a lower PO₂F₂^–/PF₆^– peak area fraction (Table S5↗).

The ¹H NMR spectra of LP57 (Figurea,b) show intense signals at 4.12 ppm (q) and 3.69 ppm (s) from EMC and at 4.48 ppm (s) from EC. The signatures of degradation detected are a subset of those found in the cycled EC and EMC electrolytes and can be assigned to poly-EO based oligomers.⁵⁵ Very weak signals for PO₂F₂^– are present in the ¹⁹F spectra with both NMC111 and 811 (Figurea,c).

To decouple the species formed, or whose formation is initiated, by chemical oxidation (via reaction with reactive lattice oxygen) and direct electrochemical oxidation pathways, the same electrochemical protocol was applied to a LiMn₂O₄ (LMO) cathode which, unlike NMC, does not evolve oxygen at high SOC. The ¹H NMR spectra of the cycled electrolytes extracted from LMO/LTO cells is shown in Figure S13↗. Aside from signals also present in the pristine electrolyte, there are no signals observed that are in common with those seen in the cycled electrolytes from NMC/LTO cells. This strongly indicates that the degradation species identified for NMC111 and 811 in Figure are initiated by chemical oxidation, or a coupled chemical and electrochemical pathway; these pathways are discussed further in the Discussion section.

To examine the extent of TM dissolution after the 60 h VH protocol, cycled electrolyte and LTO electrodes were extracted from NMC/LTO cells for characterization by ICP-OES. Figure shows the concentration of Ni, Mn, and Co dissolved in the electrolyte and deposited on the LTO anode for NMC111 and 811 and for LP57, EC, and EMC electrolytes; tabulated values are given in Table S6↗. The values are normalized per mass of electrolyte or LTO extracted from the cell. The concentration of TMs in the electrolyte was significantly lower than the concentration on the LTO anode in most cases. The effect of the NMC composition can be determined by comparison of NMC111 and 811 with LP57, and the higher Ni content of NMC811 must be accounted for. For the pristine NMCs, the TM₈₁₁/TM₁₁₁ fraction for Ni, Mn, and Co in the electrode are (0.8/0.33 =) 2.42, 0.30, and 0.30, respectively. However, with LP57 electrolyte the TM₈₁₁/TM₁₁₁ fraction of dissolved/deposited Ni, Mn, and Co are 3.54, 1.20, and 0.48, respectively. This indicates higher relative quantities of TM dissolution from NMC811 for all three TMs and is 1.5 times higher for Ni, 3.9 times higher for Mn, and 1.6 times higher for Co (Figure S14↗). TM dissolution is also strongly influenced by the electrolyte. With NMC111, the concentration of Mn and Co are lower in EMC electrolyte than LP57 and EC electrolyte (by ∼0.24 and ∼0.38 times, respectively), while the Ni concentrations are within the error of the measurement. With NMC811, the electrolyte-dependence is more striking. The concentration of Ni, Mn, and Co are 4.39, 1.78, and 3.42 times higher in EC electrolyte than in LP57, while in EMC the concentration of Ni, Mn, and Co are lower, 0.41, 0.25, and 0.32 times that in LP57.

The XPS spectra of the LTO electrodes (paired with NMC811 cathodes) after cycling in different electrolytes (Figure S15↗) are in agreement with the ICP-OES results. Specifically, the intensities of the Ni 3p, Co 3p, and the Mn 3p are highest for the LTO electrode cycled in EC electrolyte while barely any TMs are detected with EMC electrolyte. In the literature, TM dissolution has been associated with corrosion of NMC by dissolved H⁺ and HF,^62,63 the solvating power of the solvent,^64,65 and lattice oxygen release,⁶⁶ as explored further in the Discussion section.