Transition Metal Dissolution Mechanisms and Impacts on Electronic Conductivity in Composite LiNi_0.5Mn_1.5O₄ Cathode Films

Abstract

The high-voltage LiNi_0.5Mn_1.5O₄ (LNMO) spinel cathode material offers high energy density storage capabilities without the use of costly Co that is prevalent in other Li-ion battery chemistries (e.g., LiNi_xMn_yCo_zO₂ (NMC)). Unfortunately, LNMO-containing batteries suffer from poor cycling performance because of the intrinsically coupled processes of electrolyte oxidation and transition metal dissolution that occurs at high voltage. In this work, we use operando electron paramagnetic resonance (EPR) and nuclear magnetic resonance (NMR) spectroscopies to demonstrate that transition metal dissolution in LNMO is tightly coupled to HF formation (and thus, electrolyte oxidation reactions as detected with operando and in situ solution NMR), indicative of an acid-driven disproportionation reaction that occurs during delithiation (i.e., battery charging). Leveraging the temporal resolution (s-min) of magnetic resonance, we find that the LNMO particles accelerate the rate of LiPF₆ decomposition and subsequent Mn²⁺ dissolution, possibly due to the acidic nature of terminal Mn-OH groups. X-ray photoemission electron microscopy (XPEEM) provides surface-sensitive and localized X-ray absorption spectroscopy (XAS) measurements, in addition to X-ray photoelectron spectroscopy (XPS), that indicate disproportionation is enabled by surface reconstruction upon charging, which leads to surface Mn³⁺ sites on the LNMO particle surface that can disproportionate into Mn²⁺_(dissolved) and Mn⁴⁺_(s). During discharge of the battery, we observe high quantities of metal fluorides (in particular, MnF₂) in the cathode electrolyte interphase (CEI) on LNMO as well as the conductive carbon additives in the composite. Electronic conductivity measurements indicate that the MnF₂ decreases film conductivity by threefold compared to LiF, suggesting that this CEI component may impede both the ionic and electronic properties of the cathode. Ultimately, to prevent transition metal dissolution and the associated side reactions in spinel-type cathodes (particularly those that operate at high voltages like LNMO), the use of electrolytes that offer improved anodic stability and prevent acid byproducts will likely be necessary.

Introduction

The spinel structure with composition LiNi_0.5Mn_1.5O₄ (LNMO) is a promising cathode for Li-ion batteries (LIBs) because of its high energy density of 690 W h/kg^1,2 and high rate performance.^3,4 In contrast to other commercially available cathode materials currently used in LIBs, LNMO does not contain any Co, making it a potentially cheaper alternative that may also alleviate the serious geopolitical concerns associated with the continued mining of Co in the Democratic Republic of the Congo.⁵ However, LNMO-containing batteries suffer from poor cycling performance, likely due to the oxidation of organic electrolytes at high voltages (the average operating voltage of LNMO is 4.7 V vs Li/Li⁺)^6,7 and transition metal dissolution from the active material.⁸ Transition metal dissolution often occurs in tandem with surface reconstruction into an ionically resistive surface layer that results in a loss of active material.⁹⁻¹¹ The dissolved metals can chemically decompose the bulk electrolyte¹² and/or deposit on the anode,¹³⁻²⁰ where Mn is particularly insidious, and can increase the permeability of the solid electrolyte interphase (SEI) to electrolyte.^20,21 The byproducts formed during electrolyte oxidation (e.g., HF) may also react with dissolved transition metals to form resistive compounds in the cathode electrolyte interphase (CEI) that deposit on the composite surface and increase the barrier to Li intercalation.^11,22−27

Despite the clear correlation between transition metal dissolution and LNMO performance degradation, the link between the chemical and structural mechanisms driving this dissolution is not well understood. In the Mn-only spinel, LiMn₂O₄ (LMO) which contains mixed Mn³⁺/Mn⁴⁺ states, Mn dissolution can occur in the presence of acid (e.g., HF produced during electrolyte oxidation) via the disproportionation of 2Mn³⁺ centers into Mn²⁺ and Mn⁴⁺.²⁸ Because Mn²⁺ is more soluble in standard carbonate electrolytes, it tends to dissolve, while Mn⁴⁺ forms MnO₂-type structures on the active particle surface.^11,29 However, in LNMO, 1/4 of the Mn atoms in LMO are replaced with Ni²⁺, so in principle, all of the Mn is in the 4+ oxidation state and LNMO is not expected to suffer from Mn³⁺ disproportionation and subsequent dissolution of Mn²⁺.^28,30 In practice, trace Mn³⁺ is still found in pristine LNMO because of the formation of oxygen vacancies and impurity phases during synthesis,^15,30−33 which may explain, at least in a small part, why Mn dissolution still occurs in LNMO.^7,14,15,26 However, contrary to expectations, multiple reports claim that Mn dissolution proceeds beyond the redox plateau of Mn^3+/4+ at 4.1 V vs Li/Li⁺³¹ where all of the Mn should be in the 4+ oxidation state.^7,29,34,35

In order to explain this anomalous observation, three possible explanations have been proposed in the literature: (1) Mn³⁺ impurities in the bulk of the active particles are not fully oxidized during charge and allow disproportionation reactions to occur at high voltage, (2) Mn⁴⁺ is reduced during oxidation of the electrolyte salt, forming both HF and Mn³⁺-containing surface phases for the disproportionation reaction, and/or (3) Mn⁴⁺ is reduced all the way to Mn²⁺ during electrolyte solvent oxidation, facilitating direct dissolution. Evidence for mechanism (2) was first reported by Choi et al., who reported that significantly more Mn had dissolved when LiPF₆ was present in carbonate solvents during high temperature aging of LMO, indicating that PF₆^– rather than the electrolyte solvent reduces Mn⁴⁺.²⁹ More recently, mechanism (2) was also supported by the observation of a 2 nm thick, Mn³⁺-containing surface layer on LNMO particles after charge.¹⁰ Mechanism (3) was proposed by both Aoshima et al.³⁶ as well as Pieczonka et al.⁷ who claimed that the correlation between transition metal dissolution and diethyl carbonate (DEC) decomposition into ethanol (detected with gas chromatography–mass spectrometry (GC–MS)) provided evidence of a dissolution reaction by which Mn⁴⁺ is directly reduced to Mn²⁺ by ethanol at the surface of LMO and LNMO cathodes. A recent soft X-ray absorption spectroscopy (sXAS) study by Yang and coworkers³⁴ found higher quantities of Mn²⁺-containing phases on the LNMO surface after the first charge compared to the first discharge, which led them to believe that surface Mn⁴⁺ was directly reduced to Mn²⁺ upon reaction with the electrolyte at high voltage, supporting mechanism (3). While mechanisms (2) and (3) are similar in that they both agree that dissolution is enabled by the reactivity of Mn⁴⁺ with the electrolyte, they differ in whether dissolution occurs by HF-driven Mn³⁺ disproportionation or via direct dissolution of Mn²⁺ in the absence of HF.³⁶ Because these reaction mechanisms have overlap between reactants and products (e.g., dissolved Mn²⁺, HF, and other electrolyte decomposition species), it is difficult to discern experimentally which mechanism is occurring with traditional ex situ techniques. Additionally, while it is generally understood that HF leads to the dissolution of transition metal (e.g., the disproportionation of Mn³⁺), there is some debate on whether HF is present in the electrolyte because of trace water (from hydrolysis of LiPF₆) or if it is generated due to electrolyte oxidation reactions (that also produce water, but lead to increasing quantities of HF at high voltage because hydrolysis is exacerbated).^7,37−39

To address this gap in understanding, we use a combination of operando ¹⁹F NMR and EPR to simultaneously track HF formation and Mn²⁺ dissolution as a function of charging voltage in an LNMO-containing battery. Here, the operando experiment was key to tracking the kinetics and concentrations both HF and Mn²⁺ species as they were generated during the first charge at room temperature, in contrast to ex situ methods,^{7,9,19,29,40−45} many of which rely on high temperature aging procedures and postmortem analyses, in which dissolved Mn species have been shown to react during sample preparation.⁴⁶ With the operando measurements, we found a strong correlation between HF formation and dissolved Mn concentration that increased with voltage, suggesting that Mn dissolution is driven by HF produced by electrolyte oxidation. This reaction begins ≤ 1 h into charging, suggesting that the LNMO particles themselves lower the activation barrier to electrolyte decomposition. At the end of charge, electrolyte oxidation products, such as vinyl compounds, were detected with in situ ¹H NMR spectroscopy. We used a combination of X-ray photoemission electron microscopy (XPEEM) and X-ray photoelectron spectroscopy (XPS) to detect changes in the particle surface structure and oxidation state, including the emergence of Mn^2+/3+-containing phases that form during charge and feed the transition metal disproportionation/dissolution reaction at high voltage. In addition, XPEEM, XPS, and solution NMR spectroscopy were used to characterize the chemical composition of the resulting CEI and show that this layer is rich in MnF₂, which covers the entire composite surface. While it is well known that metal fluorides in the CEI hinder Li ion transport to/from the active particles, we use four-line probe measurements to demonstrate that MnF₂ also disrupts the electronic conductivity of the LNMO cathode film, more so than LiF, possibly also contributing to cell degradation.

Materials and Methods

Materials

Li metal ribbon (0.75 mm thick), 1 M LiPF₆ in ethylene carbonate:dimethyl carbonate (EC:DMC 1:1 v/v, LP30, battery grade), N-methyl-2-pyrrolidinone (NMP, anhydrous, 99.5%), LiF (≥99.9%), and MnF₂ (≥98%) were purchased from Sigma-Aldrich. DMSO-d₆ (≥99.9%) was purchased from Cambridge Isotope Laboratories. Prior to use, DMSO-d₆ was dried over molecular sieves in an Ar-filled glovebox (O₂ < 0.1 ppm, H₂O < 0.5 ppm) for 48 h. The surface oxide layer on Li metal was brushed off prior to assembly in Li/LNMO half cells. All other chemicals were used as received. LNMO cathode powder was purchased from MTI Corporation and stored in an Ar-filled glovebox. Carbon super P C45 and polyvinylidene fluoride (PVDF) binder were purchased from MTI Corporation and used as received.

Electrode Fabrication

LNMO cathode films were prepared from a slurry of 8:1:1 LNMO:C45:PVDF. The black powders, LNMO and C45, were first hand-ground in a mortar and pestle for 10 min. This mixture was then added to a solution of the PVDF binder in NMP to create a viscous slurry. The slurry was then cast onto an Al current collector (25 μm thick, MTI Corporation) using a 150 μm doctor blade and dried at 100 °C under vacuum overnight. The dried film was punched into 12.7 mm diameter disks to use in cell assembly. Once dried, cathode films were stored in an Ar-filled glovebox to minimize exposure to air and moisture. Typical mass loadings of active material per cathode were 4–9 mg cm^–2. These electrodes were used for all electrochemical testing and extracted for postmortem characterization (see XPEEM and XPS sections below).

Electrochemistry

Electrochemical tests for ex situ XPEEM and XPS were conducted using 2032 coin cells assembled in an Ar-filled glovebox with a 12.7 mm Li metal disc as the anode and LNMO films as the cathode. Each cell used a Whatman glass microfiber (GF/A) separator and was cycled in ∼0.2 mL of battery-grade, LP30 electrolyte. Galvanostatic cycling experiments were performed using C-rates of C/10 or C/20 (based on theoretical capacities of 147 mAh/g for LNMO) between 3.5 and 5 V vs Li/Li⁺.

Ex Situ and Operando Solution NMR

All solution NMR experiments were performed on a Bruker Avance III 400 spectrometer equipped with a triple resonance broadband observe (TBO) probe head. One-dimensional (1D) ¹H (30° single pulse, 1 s recycle delay, 16 scans, internally referenced to residual DMSO solvent at 2.5 ppm) and ¹⁹F (30° single pulse, 2 s recycle delay, 32 scans, internally referenced to PF₆^– electrolyte salt at −74.5 ppm) were recorded at room temperature.

For ex situ solution NMR measurements, Li/LNMO half cells were disassembled in the glovebox after cycling. The glass fiber separator was soaked in 1 mL of DMSO-d₆ for 5 min and compressed using clean tweezers to fully extract the electrolyte from the separator. The resulting solution was pipetted into a 5 mm airtight J-Young NMR tube for analysis.

For operando solution NMR, measurements were collected in a previously reported operando electrochemical NMR tube cell³⁷ (shown in Figure S1). In brief, the cell contains a LNMO cathode, a Cu anode, and a LP30 electrolyte (300 μL) assembled in a 3.65 mm o.d. FEP (fluorinated ethylene polypropylene copolymer) NMR tube liner (Wilmad LabGlass, dried at 60 ° C overnight prior to use). The tube cell is submerged into a 5 mm o.d. glass NMR tube containing 0.1 mL of DMSO-d₆ (for locking and shimming) and secured in place with Teflon tape, allowing us to collect high-resolution ¹H and ¹⁹F NMR spectra during electrochemical cycling. Electrodes were placed into the tube approximately 25 mm from the bottom of the FEP liner, completely submerged in the electrolyte. The cathode used in the operando experiments is a 20 × 2.5 mm wide rectangular film cast on an 8 in long strip of Al foil, where only the end section containing the LNMO cathode material is submerged in the electrolyte and the bare Al foil runs to the end of the tube as a current collector and electrical lead. A 20 × 2.5 mm piece of Cu film (6 μm thick) was used as the anode, which was rinsed in acetone and dried prior to use. The Cu anode was connected via heat-shrink chemically resistant electrical insulation tubing to a 22 cm long pure Cu wire, whereby upon heat-treating the tubing, the shrinkage electrically connects the Cu anode to the Cu wire within the tubing. The heat-shrink tubing prevented cell shorting along the length of the NMR tube. The Cu anode was wrapped in a single layer of Celgard 2325 separator (dried at 60 °C for 24 h prior to use). The active material loading was estimated such that the first cycle was consistent with C/10 cycling (∼2–3 mg), which was typically ∼30–40 μA.

Prior to the experiment, the tube cell was connected to double-shielded wires which were connected to an electromagnetic shield (NMR Service, GmbH). Once the cell was inserted into the NMR magnet, the shield was placed to lie on top of the upper magnet bore. The double-shielded wires extended out from the other side of the shield to connect to a Biologic SP-150 potentiostat for electrochemical cycling. Prior to acquisition, the sample was locked to DMSO-d₆ and shimmed, which was key for detecting the dilute species reported in this work. NMR acquisition parameters are the same as described above for ex situ solution NMR, except 128 scans were used for ¹⁹F acquisition, making each acquisition 7 min and 2 s long, defining the time resolution of the operando experiment. All operando measurements were repeated at least twice. The resulting 2D NMR experiments were analyzed in Python by importing the data with the Python package Nmrglue.⁴⁷ The concentration of HF in the ¹⁹F operando measurement was determined by comparing the integrated area of the HF doublet at each time point to that of the LiPF₆ doublet in the ¹⁹F NMR taken just before the electrochemical measurement started, which corresponds to 1 M LiPF₆ (pristine LP30). The complete charge profile for the electrochemistry shown in Figure 1 is shown in Figure S2 to confirm that the operando cell exhibits the expected phase transitions in the LNMO charge curve, where there is a slight plateau beginning at 4.1 V for a small amount of Mn^3+/4+ redox and another set of plateaus beginning at approximately 4.7 and 4.75 V for the Ni^2+/3+ and Ni^3+/4+ redox reactions^48,49 (Figure S3).

Figure 1.

(a) Charge profile of the operando LNMO/Cu cell at a rate of C/10 used in the operando solution NMR experiment shown in (b). This charge profile is also representative of the tube cell used for the operando EPR experiment shown in (c), where the tube cells were charged to 5 V vs Li/Li⁺ at a rate of C/10. Circles are placed along the voltage profile at the values that correspond to select voltages of each color-coded NMR/EPR spectrum shown in panel (b) and (c) as follows: OCV, 4.68, 4.73, 4.76, 4.80, 4.90, and 5 V vs Li/Li⁺. (b) Representative ¹⁹F solution NMR spectra extracted from the operando solution NMR experiment showing HF production during charge. (c) Representative EPR spectra from the operando EPR experiment showing Mn²⁺ dissolution into the electrolyte during charge. (d) Concentration of HF and Mn²⁺ detected during charge of the operando cell, which were calculated from area integrals of all spectra recorded for the experiments shown in (b,c). For the EPR, the integrals of the absolute value of the spectra were used. The cycling profile from (a)/Figure S2 is reprinted for visual aid.

EPR Spectroscopy

X-band EPR measurements were performed on an ESR-5000 spectrometer operating at a microwave frequency of 9.47 GHz manufactured by Freiberg Instruments for Bruker, sold by Rotunda Scientific Technologies. All measurements used a B-field sweep from 300 to 370 mT, a microwave power of 5 mW, a 0.5 mT modulation amplitude with 100 kHz frequency, and a 60 s scan time for a single scan. All EPR spectra were processed using background subtraction to remove the slanted baseline.

For operando EPR measurements, EPR spectra were acquired every 15 min, capturing a 60 s snapshot of the chemistry in the cell at a given voltage. The operando EPR measurements used the same electrochemical tube cell as the operando NMR measurements described in the previous sections with the exception that no glass NMR tube containing DMSO-d₆ for locking and shimming was required (i.e., the FEP tube cell was inserted into the EPR cavity as is). The cell was directly connected to the Biologic SP-150 potentiostat via alligator clips. The concentration of Mn²⁺ in the EPR operando measurement was quantified by comparing the integral of the absolute value of each spectrum in the operando experiment to that of the integrated area of the absolute value of a spectrum of the Mn²⁺ signal from a standard sample of known Mn²⁺ concentration (300 μL of 200 ppm Mn(II) triflate dissolved in LP30 in an FEP tube liner).

X-Ray Photoemission Electron Microscopy

Prior to electrochemical cycling, LNMO cathode films fabricated as described in the “Electrode Fabrication” section were pressed at 2 tons (40 MPa) in a hydraulic press (Specac) to minimize sample roughness for analysis with XPEEM. After cycling, cathode films were removed from the coin cells in the glovebox and dipped twice into 1 mL of DMC for 3 s each to remove residual salts and minimize surface layer outgassing in the XPEEM vacuum chamber. Cathode films were then dried under vacuum overnight to remove residual solvent. Samples were double-sealed in airtight pouches under Ar and transferred to the XPEEM facility at the Center for Functional Nanomaterials at Brookhaven National Laboratory (BNL) for measurement. Samples were mounted on XPEEM holders in an Ar-filled glovebox and quickly transferred to the XPEEM vacuum chamber, where air exposure during the transfer was less than 20 s.

Pixel-wise X-ray absorption spectra (XAS) were obtained by recording a series of XPEEM images at each energy in each absorption edge range at sequential increments of 0.2 eV. Mn L-edge spectra were referenced to the low energy peak characteristic of Mn²⁺, which has a reported value of 640 eV.³⁴ Other edges were translated accordingly, except the O K-edge which was referenced to the sharp low energy peak occurring from LNMO particles at 529.5 eV⁵⁰ and the C K-edge which referenced the lowest energy peak, assigned to conductive carbon, at 285.4 eV.^51,52 Local XAS spectra can be extracted from the XPEEM images by plotting the intensity of a given spatial region (set of pixels) at each energy. For example, for Mn L-edge, 112 images were acquired from 635 to 657 eV. All local XAS spectra extracted were normalized individually to their highest intensity value, unless otherwise noted. XPEEM elemental contrast images were obtained by subtracting the pre-edge image from the desired X-ray absorption energy. The C K-edge image at 293.0 eV was extracted to display the spatial distribution of conductive carbon. The pre-edge image at 280.0 eV was then subtracted, yielding the final elemental map of conductive carbon. The same procedure was conducted for other elemental maps, as specified in Figures 2 and 3. Representative C K-edge spectra are shown in Figure S4, where we followed the same normalization procedure as outlined by El-Kazzi and co-workers⁵¹ and our previous study.³⁷ In brief, we normalized our measured C-K edge spectra by that of the C K-edge spectrum of an Au microparticle deposited on a pristine LNMO sample that was imaged in XPEEM. The C K-edge spectra in Figure S4 on conductive carbon regions exhibit characteristic absorption peaks at 285.4 and 291.9 eV that correspond to C=C π* and σ* transitions.⁵¹

Figure 2.

(a) XPEEM elemental images of a pristine LNMO cathode film (top) and an LNMO cathode film that underwent 50 cycles (cycled at C/10, disassembled at 5 V vs Li/Li⁺). LNMO particles are mapped in magenta via the image corresponding to 643.0 eV and conductive carbon is mapped in cyan using the image at 293.0 eV. Dark regions indicate surface roughness caused by indirect scattering. Upon probing the F K-edge energies, no fluorine absorption signal was detected within the detection depth of the technique (∼2–10 nm), suggesting that PVDF was not present within the first few nanometers of sample imaged here. (b) Local O K-edge and (c) local Mn L-edge XAS spectra extracted for LNMO particles on the pristine sample or the cycled sample in (a), where the color of the spectrum corresponds to the location of the marker of the same color in (a). All images scale to an area of 30 × 30 μm. In the XAS spectra, dashed lines are marked along distinctive features for visual aid.

Figure 3.

XPEEM elemental images of pristine LNMO (a) before cycling, reproduced from Figure 2a and (b) 50 cycles (cycled at C/10, disassembled at 3.5 V), where LNMO particles are mapped in magenta via the image corresponding to 643 eV and conductive carbon is mapped in cyan using the image at 293.0 eV. (c) Reproduced image from (b) overlaid with an elemental map of Mn²⁺ (yellow, image corresponding to 640 eV). (d) Local Mn L-edge XAS spectra extracted from the pristine sample, regions as marked in (a). (e) Local Mn L-edge XAS spectra extracted from the 50 cycles sample, regions as marked in (b). (f) Local F K-edge XAS extracted from the 50 cycles sample, regions as marked in (b). All images scale to an area of 30 × 30 μm. In the XAS spectra, dashed lines are marked along distinctive features for visual aid.

X-Ray Photoelectron Spectroscopy

After cycling, the cathodes were removed from coin cells in the glovebox and dipped twice into 1 mL of DMC for 3 s each to remove residual salts and minimize sample outgassing. Cathode films were then dried under vacuum overnight to remove residual solvent. The samples were brought in airtight containers to the XPS instrument, then quickly mounted on the XPS chamber (air exposure <30 s) prior to pumping down the XPS antechamber. Spectra were collected using a PHI 5600 XPS system with a hemispherical analyzer and an Al X-ray source with XPS base chamber pressure <1.0 × 10^–8 Torr. XPSPEAK software was used to fit spectra. The adventitious carbon peak in the C 1s spectrum of each sample was referenced to 284.8 eV. All peaks were fit using a Shirley baseline correction with two constraints: (i) the Gaussian–Lorentzian ratio was the same for all peaks in each orbital and (ii) the full width at half maximum (FWHM) was the same for all peaks in each orbital and constrained to < 1.7 eV.

Electronic Conductivity Measurements

We followed the procedure reported by Peterson and Wheeler⁵³ to construct a four-line probe to measure the electronic conductivity of LNMO composite cathode films to determine the impact of different surface coatings. LNMO cathode slurries were combined with various quantities of inorganic CEI species, such as MnF₂ and LiF, to see how these components altered electronic conductivity. In these experiments, LNMO powder and C45 were preground for 10 min in a mortar and pestle before combining with PVDF in NMP as described in the Electrode Fabrication section above. At this stage, various quantities of metal fluorides (5.0, 9.7, 23 and 6.4 mg of MnF₂ and LiF, respectively) were added to the slurry and mixed in a mortar and pestle for 2 min prior to casting as a 22 × 22 mm square-shaped film using a 150 μm doctor blade onto a 22 × 75 × 1 mm glass microscope slide or PTFE sheet (for MnF₂ samples, to avoid reaction with glass). The final film thickness ranged from 100 to 150 μm and typical mass loading ranged from 45 to 65 mg.

The film-containing slide was placed face down onto the four-line probe. A 500 g weight was placed on top of the sample to ensure good electrical contact between the probe and the film. A current of 0.3 mA was run through the outer leads using an Arbin LBT20084 battery cycler. The voltage between the inner two probe lines was measured using a Keysight Technologies 34461A 6 1/2 Digital Multimeter. The electronic conductivity was then measured using eq 1, where L is the length of the sample along the direction of the current, I is the current being passed through the sample, and S is the cross-sectional area of the sample perpendicular to the direction of the current.

1

Four measurements were made on a sample to compute the displayed error bars. The sample was rotated clockwise 90° between each of the four measurements.

Scanning Electron Microscopy

Scanning electron microscopy (SEM) images of pristine LNMO composite cathode films used in this study (pressed at 2 tons for XPEEM measurements) were taken using a Zeiss Sigma VP Schottky Thermal Field emission SEM with a Gemini objective lens.

Results and Discussion

Operando Monitoring of HF Formation and Mn²⁺ Dissolution

We performed operando ¹⁹F solution NMR and operando EPR experiments to determine how HF formation is coupled to Mn dissolution during the first charge of LNMO in a working battery (Figure 1). At the open circuit voltage (OCV), there is little to no visible ¹⁹F NMR signal in the region corresponding to HF, nor Mn in the EPR, indicating that neither HF nor Mn are present in the electrolyte (or that they are below the detection limit of the measurement) upon battery assembly (Figure 1, bottom black spectra). In ¹⁹F NMR, a doublet centered at −191.3 ppm assigned to HF^54,55 is observed, where the doublet indicates ¹H–¹⁹F coupling (J = 474 Hz) (Figure 1b). HF prominently appears upon charging to 4.68 V vs Li/Li⁺, at approximately 61 min into the 10 h long charge experiment and persists at higher voltages (1st spectrum from bottom, Figure 1b), although fluctuations in HF concentration can be observed at earlier time points (< 1 h, Figure 1d). This coincides with the onset of electrochemical oxidation of the electrolyte and is intriguing because the operando measurement indicates that the kinetics of this process is much faster than that previously reported (on the order of 50 h,⁵⁶ this point will be discussed later). Meanwhile in the EPR, a sextet from the hyperfine interactions characteristic of Mn²⁺,^18,57,58 is first detected at 4.73 V (2nd spectrum from the bottom, Figure 1c), suggesting that HF formation closely coincides with transition metal dissolution in LNMO. As charging continues, the signal intensity associated with the spectroscopic signatures of both HF and Mn²⁺ continue to increase (Figure 1d), indicating that their concentrations rise in tandem with one another. The HF and Mn²⁺ concentrations shown in Figure 1d were determined by relating the area integrals of the NMR and EPR spectra from the experiment to the spectra of species of known concentration (see the corresponding Experimental sections). When first detected, HF and Mn²⁺ are around 15 ppm and by the end of the charge, the concentration of HF is approximately 120 ppm while the concentration of Mn²⁺ is about 90 ppm.

The operando NMR and EPR measurements shown in Figure 1 indicate that HF formation is strongly correlated with Mn²⁺ dissolution from the LNMO cathode. Although the increase in Mn dissolution with voltage is in line with previous reports for LNMO,^7,38 it is surprising that it is so strongly correlated with HF in the electrolyte because this finding suggests that Mn dissolution occurs via acid-driven disproportionation reaction at high voltage when Mn should be in the 4+ oxidation state.²⁸ As stated previously, for transition metal dissolution to occur via this route, the Mn at the particle surface must be in the 3+ oxidation state. In addition, the HF concentration continues to increase at a constant rate, even though HF is likely quickly consumed at the particle surface during disproportionation to produce soluble Mn²⁺ (which is observed simultaneously). Continuous HF production is consistent with oxide surface etching reactions proposed by Benedek and Thackeray (that form dissolved transition metal cations or transition metal fluorides) that result in H₂O formation as a byproduct²⁸ and go on to participate in subsequent hydrolysis reactions with LiPF₆ to replenish the HF in a cyclic manner.^59,60 It is also possible that the local HF concentration closest to the electrodes (which is outside of the measurement area) differs from the average, bulk concentration that we measure in the experiment.

Structural Rearrangements at the Particle Surface Feed the Mn Disproportionation Reaction

To determine if the LNMO active particles exhibit unexpected changes in the Mn oxidation state at a high potential, we performed a series of XPEEM and XPS measurements. XPEEM allows us to extract soft X-ray absorption (sXAS) spectra from different positions on the composite cathode surface before and after cycling with a lateral resolution of approximately 100 nm (Figure 2). XPEEM images acquired for a pristine LNMO film and a film cycled for 50 cycles that was disassembled at 5 V vs Li/Li⁺ are shown in Figure 2a. Representative 1st and 50th cycle electrochemical charge curves of the cells used in both XPEEM and XPS are included in Figure S5. LNMO particles are shown in magenta and conductive carbon regions are mapped in cyan, while dark regions indicate surface roughness that caused indirect scattering. SEM images of pristine pressed cathode films are shown in Figure S6, which support the spatial arrangement of LNMO and conductive carbon seen in the XPEEM maps. Although Mn²⁺, Mn³⁺, and Mn⁴⁺ share absorption peaks at several energies in Mn L-edge and O K-edge XAS, one can identify the oxidation states present via qualitative inspection of the relative intensity of peaks within each spectrum and comparison to reference spectra of standard Mn oxides, which are provided in Figures S7 and S8.^34,61 Extraction of O K-edge and Mn L-edge sXAS spectra from regions that contain LNMO particles in the pristine sample (Figure 2a, top) yields absorption spectra (Figure 2b,c, black) that are consistent with a mixed Mn^3+/4+ oxidation state expected for Mn⁴⁺ in uncycled LNMO (e.g., MnO₆ coordination environments) and associated Mn³⁺ impurities from synthesis.^34,50,62 The assignments from XPEEM are consistent with findings from XPS, where we see mixed Mn^3+/4+ oxidation states upon discharging the cells to 3.5 V vs Li/Li⁺ (Figure S9).

In contrast, the XPEEM images of the cycled sample that was charged to 5 V (Figure 2a, bottom) yield multiple unique O K-edge and Mn L-edge absorption spectra for individual LNMO particles, suggesting that the Mn oxidation state at the particle surface is heterogeneous after cycling (Figure 2b,c). As expected for the charged state of LNMO, most of the Mn on the surface of the LNMO particles is present in the 4+ oxidation state (Figure 2, gray). The Mn⁴⁺ detected here corresponds to fully delithiated LNMO at the particle surface (Ni_0.5Mn_1.5O₄) and/or a surface reconstruction phase that contains Mn⁴⁺ (e.g., λ-MnO₂ as proposed by Aurbach and co-workers^11,63).^10,61,62,64 Other particles that show similar O K-edge and Mn L-edge spectra are marked in Figure S10 of the SI. However, we find that some LNMO particles in the charged state exhibit O K-edge and Mn L-edge spectra that are consistent with mixed Mn²⁺/Mn³⁺ or Mn³⁺ phases, such as Mn₃O₄-like and Mn₂O₃-like compounds, respectively (Figure 2, light and dark green).⁶¹ In particular, the O K-edge spectrum corresponding to the light green region has a broad feature centered at 537 eV that is larger in intensity than the low energy transitions at 529.5 and 531 eV, consistent with that observed for Mn₂O₃-like phases.⁶¹ In the Mn L-edge spectrum for this region (Figure 2c, light green), the ratio of the peak at ∼641.5 eV has slightly increased relative to the peak at ∼643 eV in the pristine sample, which is also consistent with a Mn₂O₃-like (Mn³⁺) phase.⁶¹ In comparison, O K-edge and Mn L-edge spectra extracted from the dark green region are indicative of a Mn₃O₄-like (Mn^2+/3+) phase (Figure 2, dark green). In the O K-edge spectrum extracted for the dark green region, this Mn₃O₄-like phase is supported by the increased intensity of the peak at 531 eV with respect to the peak at 529.5 eV (Figure 2b, dark green).⁶¹ This observation is complemented by the corresponding Mn L-edge spectrum which shows that the sharp peak at 640 eV has increased in relative intensity to those at ∼641.5 and ∼643 eV, which is associated with the absorption of Mn²⁺ (Figure 2c, dark green and Figure S8).³⁴ We also note that the presence of reduced Mn species on the LNMO particle surface is supported by changes in the spectral region ranging from approximately 650–655 eV (Figures 2c and S8). With the ∼2–10 nm detection depth of this experiment, we may also be capturing some of the Mn⁴⁺ from the underlying particle in these particular regions, but the changes in the relative peak intensities indicate that there is a lower oxidation state on the particle surface. Likewise, XPS measurements from cells disassembled at 5 V after electrochemical cycling also show substantial quantities of both Mn³⁺ and Mn⁴⁺ (Figure S9), supporting that LNMO particles undergo surface structural rearrangement at high voltage. The presence of Mn³⁺-containing phases on the surface of LNMO particles in the charged state is theoretically unexpected, as all Mn³⁺ impurities should oxidize to Mn⁴⁺ during the plateau in the LNMO charge curve at ∼4.1 V.³¹ However, this observation is in agreement with more recent scanning transmission electron microscopy (STEM) experiments that confirm the existence of a thin (∼2 nm) Mn₃O₄ phase on the surface of LNMO upon charging¹⁰ as well as sXAS results that detected Mn³⁺ species on the surface of LNMO after charging.³⁴ Taken together, our results combined with others provide evidence that structural rearrangements at the LNMO particle surface produce a source of Mn³⁺ that can participate in the acid-driven disproportionation reaction, producing Mn²⁺ species in the electrolyte.

Spatially Resolved Characterization of the Cathode Electrolyte Interphase on LNMO

In addition to structural changes at the active particle surface, we also used a combination of XPEEM, XPS, and solution NMR to characterize the composition and arrangement of the CEI produced after electrochemical cycling. For analysis of the CEI, XPEEM images were acquired for LNMO films cycled for 50 cycles and disassembled at 3.5 V and compared to the pristine LNMO sample (Figure 3a,b). We note that for CEI characterization, it is critical to analyze the cathode at discharge (i.e., the lithiated state) because of the instability of the CEI at high voltage,^65,66 which has been observed for this material in particular.^23,38,67 The discharged state is also where we expect the CEI to be the thickest because of chemical cross-talk between the anode and cathode.^23,68,69 Observing the CEI when it is thickest will minimize spectral overlap with the LNMO particles for surface-sensitive techniques like XPEEM and XPS, both of which have a maximum penetration depth of ∼10 nm. For completeness, additional samples analyzed on the 50th charge (disassembled at 5 V), as well as the 1st charge (5 V) and discharge (3.5 V), are included in Figure S11.

After 50 electrochemical charge cycles and disassembly at 3.5 V vs Li/Li⁺, we find that LNMO particles (Figure 3b, green box) are in the Mn³⁺/Mn⁴⁺ mixed oxidation state at 3.5 V. In addition, we find that the LNMO particles also exhibit a sharp absorption at 640 eV, indicating that a layer of Mn²⁺ has formed on the particles. When we extract local Mn L-edge spectra from areas of the surface containing conductive carbon (Figure 3b, gray and black boxes), we detect absorption spectra consistent with Mn²⁺ in both regions. Local F K-edge XAS spectra from the conductive carbon indicates that the Mn²⁺ is likely deposited as MnF₂ (shoulder at 689.6 eV,⁷⁰Figure 3f) because transition metal fluoride species are distributed throughout the composite surface as well, although most concentrated on active particles (Figure S12). The deposition of MnF₂ on the cathode surface is consistent with previous reports on LiMn₂O₄^22,64,71,72 and LNMO.^7,34,50Figure 3c reproduces the image shown in Figure 3b, but with the addition of an elemental map that corresponds to Mn²⁺ (yellow), where we can see that Mn²⁺ is detected on almost the entire composite surface. In addition to finding MnF₂ in the CEI, F K-edge measurements also detect LiF (peak at 702 eV)⁷³ distributed broadly on the surface after one and 50 cycles (Figure S13). The peak at 692 eV may originate from the PVDF binder, LiF, or residual LiPF₆.^73,74 We note that while it is likely that the CEI also contains NiF₂ in the metal fluoride region, this cannot be distinguished from the Ni in the LNMO at the Ni L-edge to confirm the assignment. O K-edge measurements detect Li₂CO₃⁷⁴ distributed broadly on the surface after one and 50 cycles (Figure S14). In contrast to the discharged state, XPEEM measurements in the charged state did not detect signals from CEI species such as metal fluorides in the F K-edge or Li₂CO₃ in the O K-edge (Figure S15). Mn L-edge measurements, with high sensitivity compared to lighter elements, detected Mn²⁺ on the surface composite on charge, but in very dilute concentrations compared to on discharge (Figures S16 and S17). Less CEI on the composite in the charged state indicates that some CEI compounds may not be stable at 5 V.

XPS measurements of cycled LNMO cathodes disassembled at both charge and discharge confirm our XPEEM observations, which show that the quantities of metal fluorides in the CEI fluctuate with voltage (Figure S18). In the F 1s orbital spectra, a peak at 685.9 eV assigned to metal fluorides (e.g., LiF, MeF_x)^16,64,72 is detected on discharge but not on charge. This apparent dissipation of metal fluorides upon charging is consistent with previous reports that find that the CEI thickens upon discharge,^23,73 possibly due to voltage instability.^23,38,67 To explain this phenomenon, Oh and co-workers⁶⁷ hypothesized that these components may be soluble in HF. Conveniently, the absence of metal fluorides at 5 V allowed for the detection of changes in oxidation state due to crystallographic changes at the particle surface with XPEEM (Figure 2) and CEI compositional changes at discharge (Figure 3).

In addition to the inorganic CEI, we can also assign changes in the organic CEI with C 1s and O 1s spectra from XPS (Figure S18). At both charge and discharge, we see peaks that correspond to adventitious/conductive carbon (284.8 eV) and the PVDF binder (286.5, 290.4, and 292.2 eV¹⁶) in the C 1s spectra that can also be identified in the pristine cathode film. The peak at 288.2 eV in all of the samples corresponds to C=O environments that may arise from residual Li₂CO₃ (as seen in the pristine sample), Li alkyl carbonates, other carbonate decomposition compounds, and/or residual EC/DMC solvent.^75,76 The O 1s spectra for the charged sample show further evidence of solvent decomposition in the cycled samples, with a peak at 531.6 eV for C=O (e.g., from Li₂CO₃),²³ 532.5 eV for O–H or O–Li species (e.g., ROCO₂Li, LiOH, or MnOH),^76,77 533.8 eV for C–O-containing compounds,²³ and 535.5 eV that may arise from salt decomposition products⁷⁸ (e.g., Li_xPF_yO_z⁷⁹ or O-CF₂-containing ethers⁸⁰). For the discharged sample, the C 1s spectrum shows the same peaks and relative compositions as the charged spectrum, except there is no 292.2 eV peak for the binder, indicating a thicker CEI on discharge. The O 1s spectrum on discharge shows that ROCO₂Li (C–O) species, hydroxyl species, and polyethers (C–O) are also present as they were on charge.

The electrolyte oxidation reactions that occur at high voltage were also examined via additional analysis of the operando NMR measurements as well as further in and ex situ NMR experiments because these eliminate the need for sample washing prior to analysis when compared to X-ray techniques. One hour after the operando ¹⁹F solution NMR experiment shown in Figure 1 concluded (i.e., once the cell finished charging to 5 V), in situ ¹H NMR was acquired to detect the organic decomposition products produced during charge (Figure S19). In this measurement, we found soluble species indicative of EC decomposition, such as vinyl compounds and Li formate. Unfortunately, NMR signals from 3.5 to 4.5 ppm overlap with EC/DMC peaks from the bulk electrolyte,³⁷ so these measurements must be complemented with ex situ methods. Ex situ solution NMR of the spent electrolyte extracted from an LNMO/Li coin cell after 50 cycles exhibited ¹H resonances at 3.71 and 4.29 ppm corresponding to DMDOHD,³⁷ an oligomer formed from solvent oxidation (Figure S20).

Further inspection of the full spectra acquired during the operando ¹⁹F solution NMR experiment presented in Figure 1 reveals several fluorine species characteristic of LiPF₆ degradation (see Figure 4). The pristine electrolyte shows a doublet centered at −85.2 ppm (d, J_P–F = 938 Hz) that is assigned to PF₂O₂^–, a product from PF₆^– breakdown. This doublet grows in intensity with voltage, indicating that PF₆^– decomposes at high voltage. A doublet at −90.1 ppm (d, J_P–F = 1065 Hz) emerges at 4.68 V and then another at −88.5 ppm (d, J_P–F = 1004 Hz) appears at 4.76 V that correspond to POF₃⁸¹ and OPF₂(OCH₃), respectively.⁷⁸ Both doublets increase in intensity with voltage after initial detection, suggesting that the formation of these compounds coincides directly with the decomposition of LiPF₆ at high voltage. POF₃ and HF are simultaneously detected at 4.68 V, which raises questions regarding the kinetics of the proposed equations (eqs 2 and 3) below for PF₆^– hydrolysis reactions that produce these products:^29,39,82,83

2

3

Figure 4.

Selected spectra reproduced from the operando ¹⁹F NMR experiment in Figure 1b showing the chemical shift range of −82 to −93 ppm. The baseline appears slanted because of the overlap with the wide FWHM of the much larger LiPF₆ peak nearby at −74.5 ppm.

The operando measurements performed here offer the time resolution necessary to clearly show that HF and POF₃ production occur within 1 h of charging (Figures 1 and 4), rather than multiple hours as previously reported.^{56,59,84−86} Using both experimental methods and molecular dynamics (MD) simulations, Liu et al.⁸⁵ found that under high voltage conditions, the oxidation barrier for PF₆^–-H₂O coordination complexes can be lowered to ∼1 eV for a one electron oxidation reaction shown in eq 4, yet this still corresponds to a reaction time > 1 h.

4

The faster kinetics observed in our experiments can be rationalized in two possible ways: (i) eqs 2–4 are accurate, but in a real system are catalyzed by reactive metal centers on the LNMO particle surface during battery operation or (ii) in addition to eqs 2–4, an entirely different (and faster) reaction occurs at the cathode that feeds the production of HF. In scenario (i), Mn and/or Ni sites on LNMO may lower the activation energy for LiPF₆ hydrolysis or oxidation, leading to faster kinetics during charging. Several reports have hypothesized that the partially or fully delithiated structure, Ni_0.5Mn_1.5O₄ (isostructural with λ-MnO₂), is catalytic,^{6,7,29,34,36,43,87} but this property is still not well understood. In addition to water-driven hydrolysis, the LNMO particles themselves may oxidize or hydrolyze PF₆^– in the absence of H₂O that leads to the production of HF (scenario (ii)). Choi et al. suggested that direct PF₆^– oxidation at the particle surface causes concomitant surface transition metal reduction and proton abstraction from a nearby solvent molecule or water via eq 5,²⁹ but the kinetics of this process at the LNMO surface are also not well understood.

5

This general mechanism was supported by Solchenbach et al.,⁸⁶ who recently consolidated several density functional theory (DFT) predictions⁸⁸⁻⁹⁰ with their experimental findings to propose that oxidized EC fragments act as proton donors (Brønsted acids) that allow for the formation of HF independent to H₂O, following a direct PF₆^– oxidation process similar to eq 5. Furthermore, previous DFT studies predict that Mn-containing cathodes may be terminated with acidic Mn-OH functional groups upon solvent oxidation,^6,91,92 supporting that PF₆^– hydrolysis (eq 6) or direct oxidation (eq 7) proceeds via an alternative proton source other than water that can lower the activation energy of the reaction because of the more acidic nature of the Mn-OH group. The −OH environments seen in O 1s XPS (Figure S18) could come from Mn-OH surface hydroxyls, supporting this mechanism. Equations 6–7 are supported by Figure 4, which are either associated with the increase in HPO₂F₂ or PF₅ with voltage (which may be detected as POF₃ or HPO₂F₂ via eq 3 or eqs 8 and 9, see below).

6

7

Both mechanism (i) and (ii) are possible and may be occurring in tandem, as several electrolyte oxidation pathways generate water^39,93 in addition to the HF etching reactions of the LNMO surface²⁸ and explain the cyclic production of the reaction products (e.g., increasing HF concentration seen in Figure 1b, while Karl Fischer titration of the pristine electrolyte yielded < 10 ppm H₂O and HF is below the NMR detection limit in the pristine electrolyte). Similarly, HPO₂F₂ is associated with the formation of HF, as shown in eqs 8 and 9.^59,93

8

9

In any case, we strongly believe that the active material surface plays a role in accelerating the kinetics of electrolyte decomposition and that this finding warrants further investigation.

Influence of Metal Fluoride Deposition on Cathode Film Properties

Four-line probe measurements⁵³ were conducted to measure the sheet conductivity of the porous cathode films to investigate the impact of metal fluorides on the electronic conductivity of the LNMO composite to understand the functionality of CEI (Figure 5). We found that the pristine LNMO film had a sheet conductivity of 523 ± 26.1 mS/cm. Upon adding dilute amounts of MnF₂ (5 mg, 4.7% of the total sample composition), the sheet conductivity dropped more than twofold to 200 ± 73 mS/cm. The conductivity dropped to 90.5 ± 20.0 mS/cm and then to 49.6 ± 1.40 mS/cm when 9.7 and 23 mg of MnF₂ were added, respectively. LiF was also measured as a reference compound, where we matched the volume of LiF in an LNMO film to the volume of MnF₂ in a separate LNMO film for direct comparison (i.e., 6.4 mg of LiF corresponds to the same volume as 9.7 mg of MnF₂ based on the densities of LiF (2.64 g/cm³) and MnF₂ (3.98 g/cm³)). From this measurement, we find that MnF₂ is more electronically insulating than LiF for a given volume (290 ± 22.9 mS/cm for 6.4 mg of LiF and 90.5 ± 20.0 mS/cm for 9.7 mg of MnF₂). These measurements show that MnF₂ in the CEI is more electronically disruptive to the bulk properties of the electrode than other components, such as LiF, providing insight that is only gained from a quantitative understanding of the compositional heterogeneity of the CEI.

Figure 5.

Four-line probe measurements of different LNMO cathode films (8:1:1 LNMO:C:PVDF), each with dilute amounts (as specified on the x-axis) of LiF or MnF₂ added to the slurry prior to casting and drying.

Mechanistic Understanding of Transition Metal Dissolution, CEI Formation, and Impact on Film Properties in LNMO

The combination of time resolved and spatially resolved characterizations used in this study lend insight into the complex degradation reactions at play in LNMO batteries. With time resolved NMR (i.e., the operando solution NMR), we discovered that electrolyte oxidation (e.g., transformation of PF₆^– into POF₃) occurs at the same potential as HF formation (4.68 V). These coupled processes occur on timescales that are much faster than previously thought (min vs h), underscoring the critical role of the active cathode itself in accelerating electrolyte decomposition reactions. XPEEM analyses showed that this electrolyte oxidation resulted in the formation of Mn³⁺-rich surface layers during charge, which are susceptible to acid-driven disproportionation reactions at high voltage. These findings agree with our operando EPR measurements that detect Mn²⁺ dissolution as the voltage is raised above 4.7 V during battery charging. The fact that LiPF₆ decomposes into HF and POF₃ at the same time that Mn dissolution occurs provides new evidence that LiPF₆ oxidation plays a major role in transition metal dissolution in LNMO batteries. These observations align with studies showing that dissolution in LNMO is greatly mitigated in aging experiments using LiPF₆-free vs LiPF₆-containing electrolytes.^29,36 Together, these findings indicate that Mn dissolution should, in theory, be mitigated by using Li salts that have higher anodic stability and/or are less prone to generating HF, such as LiFSI.^8,94

The understanding that Mn dissolution at high voltages in LNMO is driven by electrolyte-facilitated reduction of surface transition metal centers can help us understand the similarities and differences in transition metal dissolution between spinels and layered oxide cathode materials (e.g., NMC) that also have highly reactive surfaces to facilitate electrolyte decomposition and undergo structural rearrangement during electrochemical cycling.^95,96 During NMC charging, electrolyte oxidation at the particle surface results in a concomitant release of oxygen that results in the production of HF^37,97,98 as well as a reduction of surface transition metals that forms an oxygen-depleted rock salt layer composed of Ni²⁺, Mn^2+/3+, and Co²⁺.⁹⁹⁻¹⁰¹ Thus, in both cathode structures, the surface reconstruction process provides both a source of reduced transition metals and HF needed for metal dissolution, but the extent of dissolution will be highly dependent on the exact composition of the surface layer, which is typically determined by the original stoichiometry of the active material.^{99,100,102,103}

However, in Ni-rich NMCs like NMC811, where only 10% of the transition metal layer contains Mn, it is not entirely clear how Mn dissolution, or transition metal dissolution in general, impacts performance decline. Yet, deposition of transition metals on graphite anodes in NMC cells is correlated with capacity fade in in LIBs.^99,104,105 Preferential Mn dissolution over Ni and Co has been observed in NMC cells and often deviates from what might be expected based on the stoichiometry of the material.^{97−99,105,106} This preferential Mn dissolution has been attributed to the higher solubility of Mn²⁺ in nonaqueous electrolytes compared to Ni and Co^10,98,99 as well as the tendency of surface Ni in Ni-rich NMCs to form a stable NiO rocksalt phase.^105,107 One study on the comparative aging between spinels and layered cathodes showed that LMO/LNMO exhibit at least 3× more transition metal dissolution compared to NMC-type cathodes during aging,^40,44 but how transition metal dissolution proceeds with different cathodes during electrochemical cycling is still not well understood. In addition, if surface hydroxyl groups are responsible for accelerating the production of HF that ultimately etches surface Mn species as proposed in eqs 6 and 7, the crystallographic structure and its termination will be critical for determining transition metal dissolution rates from the cathode. For this reason, more studies focused on the direct comparison of cycling-induced dissolution behavior between spinel-type cathodes (LMO, LNMO) and layered transition metal oxides (e.g., NMC) as well as a more robust understanding of their surface chemistries are needed to better understand the relative magnitude of dissolution in these systems.

Because LNMO and NMC811 both undergo transition metal dissolution driven by HF formation (albeit to different extents), metal fluoride species resulting from these reactions like MnF₂, NiF₂, and LiF have been observed in the CEI of both materials.^{7,8,22,23,34,50,67,108−110} However, we note that LNMO produces measurably higher quantities of MnF₂ compared to NMC811, even when cycled to high voltage (we have used 4.6 V vs Li/Li⁺ in previous work on NMC811 to study the CEI).³⁷ In recent work, we found that the CEI layer on NMC811 (50 cycles, C/10, 3–4.6 V) is rich in LiF but contained no detectable quantities of MnF₂ compared to LiF and PVDF.³⁷ These observations are in stark contrast to the XPEEM results presented here for LNMO (50 cycles, C/10, 3.5–5 V, disassembled at 3.5 V) which show a CEI that contains high quantities of MnF₂ in addition to LiF, possibly due to differences in HF formation and Mn dissolution during electrochemical cycling at voltages ≥ 4.7 V.^38,39 The reaction energy associated with MnF₂ formation is favored over the dissolved transition metal and corresponding fluoride anion in highly acidic environments, which may explain the differences in CEI composition between the layered oxides and the spinels.²⁸

The result of large quantities of MnF₂ deposition on the cathode film surface is a decrease in both the electronic conductivity (assessed with four-line probe measurements) as well as the ionic conductivity (which is widely reported in the literature^{64,71,72,111,112}). However, the scale of the MnF₂ particles (micron-sized) used in this experiment differs from those expected to form on the CEI in a working system (nanometer-sized), but given the high band gap of this material (between 9 and 10 eV¹¹³), we expect that the native CEI will still be electronically insulating. While these conductivity measurements may initially suggest that Mn-rich composites may require higher quantities of conductive carbon, adding much more will decrease the energy density of the cell and possibly lead to other adverse interactions at the cathode/electrolyte interface. As MnF₂ has been previously observed before in the CEI for a wide variety of cathode chemistries that perform well in existing Li-ion cells, what these results instead indicate is that the carbon content in composite electrodes must overcompensate for these losses during electrochemical cycling. If we can prevent the formation of MnF₂ (and other insulators) in the first place, we may have the opportunity to reduce the carbon content in the cell and improve the overall energy density.

Conclusions

We used operando EPR and NMR to correlate transition metal dissolution with HF generation in the electrolyte during the first charge of an LNMO battery. The operando experiments showed that HF is generated because of electrochemical oxidation of the electrolyte salt and that the amount of HF and dissolved Mn are strongly coupled during the first charge, within the first hour. The rapid HF generation reaction during charging indicates that the active particles play a role catalyzing electrolyte decomposition. The detection of HF and Mn dissolution occurred simultaneously at 4.7 V when the cathode is expected to have oxidized fully to Mn⁴⁺, indicating that Mn³⁺ must be produced during charge to enable the Mn disproportionation reaction to occur. This was confirmed via XPEEM and XPS analyses that detected Mn³⁺-containing surface phases, supporting that surface Mn⁴⁺ is likely reduced during electrolyte oxidation, allowing for transition metal dissolution to occur at high voltages. The simultaneous detection of POF₃ and HF at the same voltage with operando solution NMR indicates that PF₆^– oxidation largely drives transition metal dissolution and indicates that this dissolution can be mitigated by avoiding the formation of HF during cycling. Chemical and spatial characterization of the CEI indicates that these reactions lead to large amounts of MnF₂ deposits not only on LNMO particles but also on the carbon additives that decrease the electronic conductivity of the LNMO cathode film and likely alter Li transport to the active particles. However, because MnF₂ desorbs from the surface at high voltage, high voltage pulses or holds may be able to disrupt these phases on the cathode surface. Together, these results highlight the importance of developing advanced electrolytes with improved anodic stability to prevent Mn⁴⁺ reduction and enable low- or no-Co cathode active materials for Li batteries.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmaterialsau.2c00060.

• Operando NMR/tube cell design diagram; electrochemical cycling data; solution NMR spectra; SEM images; XPS spectra; XPEEM results and associated XAS spectra; and NMR spectral data sets presented in this study can be accessed freely on the Open Science Framework (OSF) (PDF)

Author Contributions

CRediT: Julia C. Hestenes conceptualization (equal), data curation (lead), formal analysis (lead), investigation (lead), methodology (lead), visualization (equal), writing-original draft (lead), writing-review & editing (equal); Jerzy T. Sadowski formal analysis (supporting), investigation (equal), methodology (lead), resources (lead), writing-review & editing (supporting); Richard May formal analysis (supporting), writing-review & editing (supporting); Lauren E Marbella conceptualization (equal), data curation (supporting), formal analysis (supporting), funding acquisition (lead), investigation (supporting), methodology (supporting), project administration (lead), resources (lead), software (lead), supervision (lead), validation (equal), visualization (equal), writing-original draft (supporting), writing-review & editing (equal).

The authors declare no competing financial interest.