Bridging Transition Metal and Anion Redox Processes in Li-Rich Sulfide Cathodes Authors: Eshaan S. Patheria, Leah S. Soldner, Nayantara Ramakrishnan, Colin T. Morrell, Michelle D. Qian, Victoria K. Davis, Kimberly A. See Published: 2026 ABSTRACT Li-ion batteries are essential for decarbonizing global transport and energy, but their scalability is constrained by limited supplies of critical cathode elements, such as Ni, Mn, Co, and P. To address this, we previously introduced high-energy-density Li-ion cathodes composed of Al, Fe, and S, which are elements already produced globally at industrial scale and battery-grade purity. These cathodes leverage sulfide anion redox, involving nonbonding S 3p states and localized distortions that form and break S–S bonds, enabling high capacity. Here, we expand this chemical space by incorporating Cu into cathodes Li2.2– z Cu z Al0.2Fe0.6S2 (0 ≤ z ≤ 0.4), where highly covalent Cu–S interactions stabilize holes on Cu as Cu>1+. This Cu redox extends charge compensation that was previously restricted to localized, electronically isolated S–S bonds. Cu also limits capacity, which we attribute to structural destabilization of the delithiated phase, despite the thermodynamic stability of Cu>1+. By describing the effects of Cu on charge compensation and phase stability, we present a sulfide anion redox mechanism for next-generation multielectron redox Li-ion cathodes, where highly covalent transition metal states participate in otherwise electronically isolated redox processes involving anion nonbonding states. Introduction An estimated 100–400 TWh of Li-ion batteries are needed to decarbonize global transportation and energy by 2050. − However, with only 2.4 TWh deployed as of 2023, this ≳100× scale-up is hindered by supply limitations of elements used in cathodes, e.g., Ni, Co, Mn, , and P. − This challenge would be alleviated by cathodes reliant only on “industrial elements”, which we previously defined as elements with annual global production of at least 107 metric tons in primarily elemental form with ≳99.5 wt % purity. While Li, produced at just over 106 tons for the first time in 2024, does not yet meet the criteria, it is immutable in Li-ion cathodes. Accordingly, we selected industrial elements Al, Fe, and S to develop high-energy density Li-rich cathodes Li2+y Al y Fe1–2y S2 (0 ≤ y ≤ 0.5), which represents the solid solution of the end members Li2FeS2 (Figure a) and Li5AlS4 (Figure b). The highest energy density is achieved in Li2.2Al0.2Fe0.6S2 through multielectron redox, in which anions participate in charge compensation during (de)­lithiation, thereby surpassing the capacity limits of conventional intercalation reactions that rely primarily on transition metal redox. , Reversible 2S2–/(S2)2– redox of ≈80% of the S reaches ≳1000 Wh·kg−1. Crystal structures of (a) Li2FeS2, (b) Li5AlS4, and (c) LiCuFeS2 projected down the c-axis (top structures) and down the b-axis (bottom structures). In each panel, the solid black line indicates the unit cell of the structure shown. Early investigations of anion redox in sulfides were pioneered by John B. Goodenough and Jean Rouxel, among others, who observed that for transition metal sulfides containing “high” oxidation states of late transition metals, the transition metal is inactive upon oxidation, with oxidation being facilitated by formal oxidation of the sulfide anions. Around this time, Gard et al. first showed evidence of (S2)2– formation upon electrochemical oxidation of Li2FeS2. More recent work on electrochemical redox processes of Li-rich transition metal sulfides also assign charge compensation of anion redox to S 3p nonbonding states. − Similarly, O redox has been extensively characterized as a route to high energy density Li-ion cathodes. However, oxidized (O2)(2–n)– (n > 0) species are often unstable, leading to a variety of compensatory and frequently irreversible structural rearrangements that remain the subject of intense debate. − To mitigate this instability, recent efforts have suggested that “holes” on oxide ligands generated by electrochemical oxidation can be stabilized in highly covalent transition metal-O bonds, such as Mn4+–O(2–n)–, Fe3+–O(2–n)–, and Ni2+–O(2–n)–. These oxidized compounds are suggested to be “negative charge transfer” compounds in which the electronic state is conceptualized as d (n–1)+ L, where L is a ligand hole, rather than d n+. The term “negative charge transfer” was first introduced by Mizokawa et al. in 1991 and later popularized by Sawatzky and co-workers. This mechanism is still highly debated, as characterization of the electronic structure is complex. Further, “holes” on O are typically unstable and can condense to form O2(g) or related species causing degradation. By contrast, Li2+y Al y Fe1–2y S2 (0 ≤ y ≤ 0.5) shows distinct anion oxidation from nonbonding S 3p states resulting in the formation of (S2)2– with no commensurate transition metal redox activity. , The anion oxidation mechanism shows a characteristic plateau around 2.5 V vs Li/Li+, during which the Fe shows no characterizable redox contributions. , Persulfide formation typically causes less degradation than O2(g) formation, because persulfides motifs are stable and found in many thermodynamic materials. Further, subsequent electrochemical reduction recovers the pristine, crystalline phase, highlighting the reversibility of the 2S2–/S2 2– redox couple. , While the stability of “holes” on O is suggested to depend on highly covalent metal-anion bonds, depopulated nonbonding S 3p states are responsible for (S2)2– formation, and thus the formation of the (S2)2– is largely independent of metal bonding interactions. , However, here we aim to probe the effect of transition metal d states on the formation of (S2)2– by incorporating a metal that we hypothesize will have high degrees of covalency with the ligand field. Accordingly, we now expand the chemical space of the solid solution between Li2FeS2 and Li5AlS4 by incorporating Cu as Li2.2–z Cu z Al0.2Fe0.6S2 (0 ≤ z ≤ 0.4). The substitution of Cu for Li is enabled by the structural similarity of LiCuFeS2 (Figure c) to Li2FeS2 and Li5AlS4. Cu has covalent, unique electronic interactions with S that favor the Cu+ d 10 closed-shell electronic configuration, distinguishing it from any other 3d transition metal-S pairing. Within the pyrite MS2 series (M = Mn, Fe, Co, Ni, Cu, Zn), CuS2 has the shortest S–S bond length of 2.03 Å, , compared to 2.16 Å in FeS2, indicating stronger S–S bonding than in the other analogues. While the MS2 compounds have formal oxidation states M 2+ and (S2)2–, CuS2 is often assigned Cu+ and highly oxidized (S2)−, stabilized by the favorable Cu+ d 10 electronic configuration. ,− Comparing the isoatomic phases FeS and CuS further illustrates the preference for Cu+. FeS consists of Fe2+ octahedrally coordinated by S2– ligands, while CuS contains both tetrahedral and trigonal planar Cu, with unusually short S–S bonds (2.07 Å) that connect the bases of the tetrahedra. This unusual structure has been described as a composite of Cu2S and CuS2, with formal oxidations states Cu+, S2– and Cu+, (S2)−, respectively. The oxidation states in ternary CuFeS2 (chalcopyrite) have also been intensely debated. Some authors propose formal oxidation states Cu+Fe3+S2 2–, consistent with the strict Cu+ d 10 electronic configuration, − while others propose a more delocalized picture: Cu(1+n)+Fe(3–n)+S2 2–, where 0< n ≤ 1. − This contrasts with isostructural CuAlS2, where the oxidation states are well-defined as Cu+Al3+S2 2–. Based on the known phases discussed here, we hypothesize that Cu in Li2.2–z Cu z Al0.2Fe0.6S2 should not act as a spectatorrather, it will alter the redox behavior of Fe and S, and overall phase stability with (de)­lithiation. Although we target the substitution of Cu for Li as “isovalent” substitution, as “Cu+” for Li+, the precise oxidation state of Cu in the pristine materials is difficult to ascertain. We therefore examine relative changes in the Cu oxidation state and electronic structure compared to the pristine state, referring to the oxidation state of Cu in the pristine state as “Cu+” for simplicity. Here, we demonstrate Cu redox activity in Li2.2–z Cu z Al0.2Fe0.6S2, where electrochemical oxidation generates holes on Cu (i.e., Cu>1+). We hypothesize that the Cu redox activity is enabled by uniquely high Cu–S covalency. The redox activity of Cu broadens charge compensation during the anion oxidation plateau beyond just S 3p nonbonding states to Cu 3d states in Li2.2–z Cu z Al0.2Fe0.6S2. At the same time, the total capacity (in mole e– per formula unit) decreases linearly with increasing Cu content. We compare multielectron redox in Li2.2Al0.2Fe0.6S2 (z = 0) and Li1.8Cu0.4Al0.2Fe0.6S2 (z = 0.4) to evaluate how Cu influences the capacity limit. We find that Cu>1+ destabilizes the delithiated phase, introducing a competing phase transition to CuFeS2, which suppresses deep delithiation. This Cu-enabled redox behavior expands the scope of multielectron mechanisms by showing how transition metal redox can couple to otherwise isolated S 3p bond-forming/breaking processes. Results and Discussion Structural Characterization of Li2.2–z Cu z Al0.2Fe0.6S2 Substitution of Cu+ into Li2+y Al y Fe1–2y S2 (0 ≤ y ≤ 0.5) is motivated by structural similarities between Li2FeS2 (Figure a), Li5AlS4 (Figure b), and LiCuFeS2 (Figure c). Fong et al. first reported LiCuFeS2 and its complete solid solution with Li2FeS2 (Li2–x Cu x FeS2, where 0 ≤ x ≤ 1) in 1989, motivated by previous reports of partial Cu+ substitution for Li+ in Li2FeS2. , The crystal structure of LiCuFeS2, shown in Figure c along the c and b axes, is isostructural to Li2FeS2 and adopts the P3̅m1 trigonal space group, with a hexagonal close-packed (HCP) sulfide anion framework and cations alternating between octahedral and tetrahedral sites in layers. The tetrahedral sites in LiCuFeS2 have Cu and Fe in a 1:1 ratio while the tetrahedral sites in Li2FeS2 have Li and Fe in a 1:1 ratio. We previously reported Li2+y Al y Fe1–2y S2 (0 ≤ y ≤ 0.5), the solid solution of Li2FeS2 and Li5AlS4, whose structural relationship is described in our previous work. Motivated by the complete substitution of tetrahedral Li+ by Cu+ in Li2FeS2, we hypothesize that the tetrahedral Li+ in Li2+y Al y Fe1–2y S2 can also be replaced by Cu+. We therefore target substituting Cu+ for Li+ in Li2.2Al0.2Fe0.6S2 as Li2.2–z Cu z Al0.2Fe0.6S2 (0 ≤ z ≤ 0.4)a systematic perturbation of the Li2.2Al0.2Fe0.6S2 (z = 0) parent phase. We prepare Li2.2Al0.2Fe0.6S2 (z = 0), Li2Cu0.2Al0.2Fe0.6S2 (z = 0.2), and Li1.8Cu0.4Al0.2Fe0.6S2 (z = 0.4) by solid-state synthesis from Li2S, Cu, Al2S3, FeS, and FeS2 at 900 °C. To confirm the incorporation of Cu, we analyze synchrotron X-ray diffraction (sXRD) patterns for Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 shown in Figure , with the corresponding Rietveld refinements (Table S1), reflections associated with each phase in the fit, and the difference between the fit and data. We also show the sXRD data on a log intensity scale to make the data and fit at high Q more visible (Figure S1). The P21/m unit cell of Li2.2Al0.2Fe0.6S2 is used to fit all compositions. The unit cell contains three unique tetrahedral sites: two at 2e Wyckoff positions and one at a 4f position. Al occupies only one of the 2e sites, while Li and Fe are disordered across all three. Given the similar effective ionic radii of monopositive Li+ and Cu+ in tetrahedral coordination, Cu substitution is modeled by replacing Li with Cu across all sites. For example, Li2Cu0.2Al0.2Fe0.6S2 is generated by reducing Li occupancy by 10% and adding 10% Cu occupancy at each 2e and 4f site. In this way, Li2Cu0.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 are generated as Cu-substituted derivatives of Li2.2Al0.2Fe0.6S2. Synchrotron powder X-ray diffraction of (a) Li2.2Al0.2Fe0.6S2, (b) Li2Cu0.2Al0.2Fe0.6S2, and (c) Li1.8Cu0.4Al0.2Fe0.6S2 along with the corresponding Rietveld refinement, reflection locations in Q (Å–1) of each phase in the fit, and difference between fit and data. Superstructure reflections that are not fit by the Rietveld refinement are indicated by asterisks (*). The resulting lattice parameters (d) a, (e) b, and (f) c from each Rietveld refinement with respect to z in Li2.2–z Cu z Al0.2Fe0.6S2 with linear fits indicated by dashed lines. The lattice parameters largely follow a linear Vegard’s trend, indicating that Cu+ successfully substitutes for Li+ in Li2Cu0.2Al0.2Fe0.6S2 (z = 0.2) and Li1.8Cu0.4Al0.2Fe0.6S2 (z = 0.4). The lattice parameters a, b, and c from the fits are plotted vs Cu content in Figure d–f, respectively. The lattice parameters largely follow linear Vegard trends, supporting the substitution of Cu+ for Li+. Cu substitution causes a decrease in the unit cell volume, suggesting a more covalent Cu–S interaction compared to Li–S. Weak reflections between 1 and 2 Å–1 for Li2Cu0.2Al0.2Fe0.6S2 are not described by the fit and are marked with asterisks (*). The reflections are not associated with any reasonable impurity phases and their position in Å–1 suggest they are likely superstructure reflectionswe have described this hypothesis in more detail in our previous work. Electrochemical Characterization of Li2.2–z Cu z Al0.2Fe0.6S2 The electrochemical performance is evaluated with galvanostatic cycling. The first cycles of Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 at C/10 based on 1 e– per formula unit are shown in Figure a–c, respectively. All materials exhibit characteristics associated with sequential transition metal (Fe2+/>2+) and anion (2S2–/(S2)2–) oxidation during charge, corresponding to the initial sloping region and subsequent plateau, respectively. , We note the open-circuit voltage (OCV) increases linearly with Cu content (Figure S2), which suggests that Cu substitution alters the pristine state monotonically. Capacity changes the most in the plateau region, which decreases as Cu is added from 1.52 ± 0.09 e– in Li2.2Al0.2Fe0.6S2, to 1.36 ± 0.04 e– in Li2Cu0.2Al0.2Fe0.6S2, and to 1.16 ± 0.01 e– in Li1.8Cu0.4Al0.2Fe0.6S2. In contrast, the capacity in the sloping region changes relatively little: it decreases from 0.46 ± 0.02 e– in Li2.2Al0.2Fe0.6S2 to 0.34 ± 0.01 e– in Li2Cu0.2Al0.2Fe0.6S2, and remains constant at 0.34 ± 0.01 e– in Li1.8Cu0.4Al0.2Fe0.6S2. To better visualize changes in the sloping region, we plot the differential capacity (dQdV) versus voltage in Figure S3. While the sloping region in Li2.2Al0.2Fe0.6S2 consists of two distinct low and high voltage features, the low voltage feature diminishes significantly in Li2Cu0.2Al0.2Fe0.6S2 and disappears altogether in Li1.8Cu0.4Al0.2Fe0.6S2. By contrast, the high voltage feature is consistent across all three materials. We note that the loss of the low voltage feature with Cu content coincides with the trend of higher OCV with Cu content. Three replicate cells from for each material are shown in Figure S4, with tabulated capacities provided in Table S2. We plot the total average charge capacity of these replicates versus Cu content (Figure S5) and find that the capacity limit, in mol e– per f.u., decreases linearly with Cu content z (R 2 = 0.987). Despite this trend, Li2Cu0.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 still deliver substantial charge/discharge capacities of 352 ± 12 mAh·g–1/366 ± 15 mAh·g–1 and 287 ± 1 mAh·g–1/288 ± 1 mAh·g–1, respectively, and high energy densities 873 ± 29/841 ± 35 and 710 ± 3/664 ± 2 Wh·kg–1, respectively. The gravimetric capacities, average voltages, and energy densities of all materials are tabulated in Table S3. First cycle galvanostatic charge and discharge curves of (a) Li2.2Al0.2Fe0.6S2, (b) Li2Cu0.2Al0.2Fe0.6S2, and (c) Li1.8Cu0.4Al0.2Fe0.6S2 cycled at C/10 based on 1 e– per formula unit. (d) Cycling data of Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 at C/10. (e) Rate capability tests of Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 for 5 cycles each at C/10, C/5, C/2, 1C, and again at C/10. All C rates are based on 1 e– per formula unit. The data points in (d) and (e) are the average of three replicate cells and error bars indicate the standard deviations. The data for Li2.2Al0.2Fe0.6S2 are adapted with permission from Patheria et al. Copyright 2025 American Chemical Society. Preliminary long-term cycling and rate capability performance are shown in Figure d,e, respectively. We use free-standing electrodes (see Freestanding cathode preparation), which are poorly suited for extended cycling. We are currently optimizing cast electrodes to better assess the capacity fade and rate capability. The cycling stability at C/10 of each material is shown in Figure d. We tabulate the rates of capacity fade in Table S4, obtained from linear fits of discharge capacity versus cycle number in Figure d. Each data point and error bar represent the average and standard deviation of three replicate cells, with the individual cell data and Coulombic efficiencies shown in Figure S6. The discharge capacity fade rate decreases with increasing Cu content. This trend may partly reflect the reduced extent of (de)­lithiationespecially of structurally burdensome 2S2–/(S2)2– redoxin the lower-capacity Cu-substituted materials. We also compare the rate capabilities of Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 in Figure e (Figure S8 shows individual cell data). We compare the rate capabilities by taking the 1C discharge capacity (average of five cycles) as a percentage of the initial C/10 capacity (average of the first five cycles) and find that it does not vary monotonically with z. The Li2Cu0.2Al0.2Fe0.6S2 stoichiometry shows the best rate capability, suggesting that Cu incorporation has some benefit to kinetics although too much Cu hinders kinetics. The decrease in rate capability from Li2Cu0.2Al0.2Fe0.6S2 to Li1.8Cu0.4Al0.2Fe0.6S2 may reflect disruption of Li+ percolation networks by Cu+a phenomenon previously observed in Li-rich cation-disordered rocksalt cathodes, where cation site occupancy influences Li+ transport pathways and thus rate capability. To probe the role of Cu on the kinetics further, we use galvanostatic intermittent titration technique (GITT) to assess overpotentials at C/10. The equilibrium voltage (V eq) and overpotential (η) during charge for Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2, extracted from GITT, are shown in Figure S10. Full GITT charge/discharge curves are shown in Figure S11. Both V eq and η are unchanged during charge across all Cu contents despite the capacity fade and rate capability differences with z. Thus, despite possible benefits to kinetics at high rates based on the rate capability data, at C/10, Cu does not affect overpotentials. This indicates that Cu alters total capacity, capacity fade, and rate capability through structural or electronic effects that do not significantly impact the kinetics of the multielectron redox reactions. We note, though, that comparisons of rate capability and cycling retention are difficult to make since the initial capacities are different. Additionally, given the limitations of freestanding electrodes for long-term performance analysis, we now focus on evaluating differences in the first cycle between the materials via mechanistic understanding of the first cycle. Spectroscopic Characterization of the Multielectron Redox Mechanism The galvanostatic data, interpreted based on our previous work, suggest that the decreasing capacity with increasing Cu content primarily reflects reduced 2S2–/(S2)2– redox, while Fe2+/>2+ redox remains relatively unchanged. This inference is based on changes in the two distinct features of the charge curves: the initial sloping region, associated with Fe2+/>2+ oxidation, and the subsequent plateau, associated with 2S2–/(S2)2– oxidation. No new features appear during charge in the Cu-substituted materials, and thus the role of Cu in the redox mechanism is not clear from the electrochemistry alone. To probe how Cu influences the multielectron redox mechanism, we perform ex situ X-ray absorption spectroscopy (XAS). We measure Fe and Cu K-edge XAS at four states of charge (SOCs): (1) pristine, (2) transition (at the transition point between the sloping and plateau regions), (3) charged, and (4) discharged. We measure S K-edge XAS at these same SOCs and at two additional intermediate points: midslope and midplateau. To highlight the mechanistic role of Cu, we focus on comparing Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2, while data for Li2Cu0.2Al0.2Fe0.6S2which shows systematic trends consistent with this comparisonis presented in the Supporting Information. First, we discuss the Fe K-edge XAS in Figure (see Figure S12 for Fe K-edge XAS of Li2Cu0.2Al0.2Fe0.6S2). The Fe K-edge near-edge regions for Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 at the various SOCs are shown in Figure a,b. The spectra exhibit common features: a pre-edge at 7113.0 eV, labeled w, and a rising edge that shifts depending on the SOC. For both materials, the pre-edge intensity increases at the transition SOC and remains nearly constant at the charged state, suggesting that electronic states with significant Fe 3d character change predominantly during the sloping region. Although the pre-edge intensities of the transition and charged states of Li2.2Al0.2Fe0.6S2 are nearly equivalent, we note the pre- and rising edge change shape slightly in the charged state. The changes correspond to slightly different electronic configurations , that we probed via higher-sensitivity Mössbauer spectrometry in our previous work. Ex situ Fe K-edge XAS spectra of (a) Li2.2Al0.2Fe0.6S2 and (b) Li1.8Cu0.4Al0.2Fe0.6S2. The first derivative of the rising edge regions for (c) Li2.2Al0.2Fe0.6S2 and (d) Li1.8Cu0.4Al0.2Fe0.6S2. The energies of the maxima of the first derivatives at each of the SOCs overlaid with the corresponding galvanostatic cycling data for (e) Li2.2Al0.2Fe0.6S2 and (f) Li1.8Cu0.4Al0.2Fe0.6S2. The dashed lines in all panels indicate the approximate positions of the pre-edge w, at 7113.0 eV, the two rising edges observed at different SOCs, x and y, at 7117.2 and 7118.2 eV, respectively, and the additional high energy feature, z, at 7119.7 eV. The data for Li2.2Al0.2Fe0.6S2 are adapted with permission from Patheria et al. Copyright 2025 American Chemical Society. To more clearly observe features in the rising edge, we plot the first derivative of the XAS data in Figure c,d for Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2. Several factors govern the shape of the Fe K-edge rising edge, including ligand-field effects and metal–ligand covalency, which can produce multiple rising-edge features. In Li2.2Al0.2Fe0.6S2, the pristine state is described by a single feature at x, ≈7117.2 eV. At the transition state, this feature shifts to ≈7118.2 eV, labeled y, and remains at the ≈7118.2 eV at the charged state. A weak shoulder appears at ≈7119.7 eV, labeled z, at the transition state and grows in intensity at the charged state. By contrast, in Li1.8Cu0.4Al0.2Fe0.6S2, the pristine state exhibits two features, where the lower energy feature is at the same energy x as in Li2.2Al0.2Fe0.6S2. While this feature also shifts to higher energy in Li1.8Cu0.4Al0.2Fe0.6S2 upon oxidation, the magnitude of the shift is smaller than in Li2.2Al0.2Fe0.6S2, shifting only to ≈7117.9 eV, labeled y′, as opposed to ≈7118.2 eV (y). The smaller shift supports the inference from the electrochemistry that Fe2+/>2+ contributes less capacity in the Cu-containing phases than in Li2.2Al0.2Fe0.6S2, based on the shorter sloping region in the charge curves of the Cu-containing phases (Figure a–c). The higher energy feature in the pristine state of Li1.8Cu0.4Al0.2Fe0.6S2 shifts to ≈7119.7 eV (z) at the transition state, the same energy at which it appears in Li2.2Al0.2Fe0.6S2. As in Li2.2Al0.2Fe0.6S2, this feature stays at the same energy and grows in intensity at the charged state. Overall, the rising edge features of Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 show similar trends during charge, and revert to a spectrum resembling the pristine material at the discharged state. The Fe K-edge XAS spectra and first derivatives of pristine Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 are overlaid in Figure S13, which shows the low-energy rising edge maximum is at the same energy, x, across all the materials. To track a single, comparable ‘edge position’ across all states of charge (SOCs), we reference the low-energy rising edge maximum which behaves similarly to higher-energy feature. This confirms that Fe2+/>2+ oxidation ceases after the transition SOC despite the majority of electron removal occurring during the plateau in both materials. To be thorough, we note that when the low-energy feature stops shifting, the high-energy feature also does not shift. The position of the low-energy feature is plotted in Figure e,f, with the galvanostatic data. For both Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2, the rising edge position increases during the sloping region indicating oxidation, and stays constant during the plateau. Next, we discuss the S K-edge XAS as a function of SOC, which is shown along with the associated electrochemistry in Figure . The same data for Li2Cu0.2Al0.2Fe0.6S2 are shown in Figure S14. The S K-edge XAS spectra of pristine Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2 are overlaid in Figure S15. The pre-edge feature u at 2469.2 eV is unchanged across the materials and increases in intensity with Cu content, confirming that Cu and S are highly covalent. For Li2.2Al0.2Fe0.6S2, u increases in intensity during the sloping region, indicating greater S 3p character in the Fermi-level states as a result of oxidation, i.e., greater transition metal-S covalency (Figure a). , During the plateau, u remains constant with no further increase in intensity, while a new feature v emerges at 2471.8 eV. Feature v, which indicates (S2)2– formation, , peaks in intensity at the end of charge and disappears completely at the discharged state. The growth of v during the plateau, alongside negligible change in u, marks a switch from increasing transition metal–S covalency via Fe2+/>2+ oxidation to 2S2–/(S2)2– oxidation. The spectra and corresponding galvanostatic data for Li1.8Cu0.4Al0.2Fe0.6S2, shown in Figure c,d, follow the same trends as Li2.2Al0.2Fe0.6S2, with the same pre-edge features u and v. The lower intensity of v in the charged state of Li1.8Cu0.4Al0.2Fe0.6S2 confirms less 2S2–/(S2)2– oxidation capacity relative to Li2.2Al0.2Fe0.6S2. 2S2–/(S2)2– redox is structurally reversible with respect to long-range order in both materials after the first cycle, as confirmed by ex situ XRD (Figures S16 and S17) and our previous work. , (a) Ex situ S K-edge XAS and (b) representative first cycle curve indicating the SOCs at which the XAS data was collected for Li2.2Al0.2Fe0.6S2. The corresponding data for Li1.8Cu0.4Al0.2Fe0.6S2 are in (c) and (d), respectively. The dashed lines in (a) and (c) indicate the two pre-edge features u and v, at 2469.2 and 2471.8 eV, respectively. The data for Li2.2Al0.2Fe0.6S2 are adapted with permission from Patheria et al. Copyright 2025 American Chemical Society. The Fe and S K-edge XAS exhibit consistent trends across Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 as a function of SOC. We now discuss the Cu K-edge XAS of Li1.8Cu0.4Al0.2Fe0.6S2, shown in Figure (see Figure S18 for Cu K-edge XAS of Li2Cu0.2Al0.2Fe0.6S2). The near-edge spectra at different SOCs are shown in Figure a. A pre-edge feature at ≈8979.0 eV (labeled r), magnified in Figure b, appears at the transition state, intensifies at the charged state, and disappears at the discharged state. The pre-edge, formally assigned to the Cu 1s to 3d transition, is absent in the pristine state, where Cu is introduced via isovalent substitution for Li+ and expected to adopt a closed-shell d 10 configuration (Cu+). Thus, with no empty Cu 3d states in the pristine and discharged materials, pre-edge excitation is not possible, explaining the absence of pre-edge intensity. The emergence and growth of r during charge thus supports the progressive emptying of Cu 3d states, consistent with hole accumulation on Cu as it is oxidized to Cu>1+. As ligand-field effects can induce weak pre-edge intensity even in Cu+ d 10 systems by mixing empty Cu 4p character into the 3d states, we examine the rising edge to distinguish between changes in coordination environment and genuine Cu+/>1+ oxidation. In Figure a, the rising edge shifts from ≈8981.9 eV (labeled s) in the pristine state to ≈8982.4 eV (labeled t) at the charged state. This shift is visualized more clearly in the first derivative shown in Figure c, with the derivative maximum shifting progressively from s to t throughout the entire charge curve, and returning to its original position at the discharged state. The rising edge position in the pristine state matches previously reported edge positions near ≈8982 eV for monopositive Cu+ coordinated by S. ,, The simultaneous shift in the rising edge and increase in pre-edge intensity supports Cu+/>1+ oxidation as the primary driver of these changes, rather than structural or ligand-field effects. We summarize this result in Figure d, overlaying the rising edge positions with the galvanostatic data. Unlike the Fe K-edge, which remains mostly unchanged during the plateau, the Cu rising edge position increases continuously across both the sloping region and the plateau. This continuous shift indicates that Cu+/>1+ oxidation occurs throughout the full charge process, bridging the otherwise electronically isolated Fe2+/>2+ and 2S2–/(S2)2– redox reactions. (a) Ex situ Cu K-edge XAS spectra, (b) magnified pre-edge region, and (c) first derivative of the rising edge regions for Li1.8Cu0.4Al0.2Fe0.6S2. (d) Energies of the maxima of the first derivatives at each of the SOCs overlaid with representative galvanostatic cycling data. The dashed lines in all panels indicate the approximate positions of the pre-edge r, at 8979.0 eV, and the lowest and highest energy rising edge positions observed at different SOCs, s and t, at 8981.9 and 8982.4 eV, respectively. Effect of Cu>1+ on Local Structure and Delithiated Phase Stability To complement the electronic structure changes observed by XAS that indicate Cu+/>1+ redox in Li2Cu0.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2, we now use extended X-ray absorption fine structure (EXAFS) analysis to investigate the accompanying local structural changes that correlate to the multielectron redox mechanism. We conduct Fe and Cu K-edges EXAFS analyses at the pristine, transition, charged, and discharged SOCs of Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2, with the parameters associated with the first-shell shown in Figure . Parameters associated with the second-shell are in Figure S19, and fitted χ­(k) and |χ­(R)| in Figures S20–S22 and S23–S25, respectively. All fit parameters are tabulated in Table S5. (a) Bond length, r eff., and (b) coordination number, N, associated with the first shell from Fe K-edge EXAFS analysis of Li2.2Al0.2Fe0.6S2, with (c) accompanying galvanostatic data indicating the SOCs at which EXAFS analysis is conducted. Panels (d), (e), and (f), respectively, indicate the same corresponding data for Li1.8Cu0.4Al0.2Fe0.6S2. Panels (d) and (e) additionally include r eff. and N, respectively, from Cu K-edge EXAFS analysis of Li1.8Cu0.4Al0.2Fe0.6S2. All r eff. and N values are shown with statistical error bars from fitting. The data for Li2.2Al0.2FeS2 are adapted with permission from Patheria et al. Copyright 2025 American Chemical Society. The first shell is associated with the Fe–S bonds. First, we discuss trends in the first-shell bond length (r eff.) and coordination number (N) of Li2.2Al0.2Fe0.6S2, shown in Figure a,b, respectively, with corresponding galvanostatic data plotted in (Figure c). r eff. decreases during the sloping region, consistent with Fe2+/>2+ oxidation, remains constant during the plateau, and increases back to its original value at the discharged state. r eff. thus mirrors the Fe K-edge rising edge position summary in Figure e. N remains fairly constant at all SOCs, between ≈3 and 3.3 with its maximum at the transition SOC (Table S5), which indicates Fe is persistently distorted toward the tetrahedral base. , Next, we discuss the EXAFS analysis for both the Fe and Cu K-edge data obtained from Li1.8Cu0.4Al0.2Fe0.6S2. The first-shell r eff. and N are shown in Figure d,e, respectively, with corresponding galvanostatic data (Figure f). For clarity we use subscripts to distinguish r eff.,Fe and N Fe for Fe, and r eff.,Cu and N Cu for Cu. For Fe, r eff.,Fe again mirrors the Fe K-edge rising edge positions (Figure f) as it did in Li2.2Al0.2Fe0.6S2. Similarly, N Fe changes minimally across SOCs as in Li2.2Al0.2Fe0.6S2. However, N Fe differs in that it is now closer to the ideal tetrahedral value of 4, between ≈3.6 and 3.9 with its minimum at the transition SOCopposite to the trend in Li2.2Al0.2Fe0.6S2. The similar Fe–S bond contraction in Li2.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 indicates that Fe is oxidized to a similar extent in both materials. For Cu, r eff.,Cu demonstrates continuous Cu–S bond contraction throughout charge (Figure d), supporting Cu+/>1+ oxidation observed in the X-ray absorption near-edge structure (Figure ). Unlike Fe, for which r eff.,Fe only decreases during the sloping region, Cu+/>1+ oxidation pervades the entire multielectron oxidation, even during the 2S2–/(S2)2–-oxidation plateau, reflected structurally by continuous Cu–S bond contraction. The Cu–S bond length at the charged state, ≈2.28 ± 0.01 Å, matches the Fe–S bond length at the transition and charged states. N Cu, in sharp contrast to N Fe, approaches ideal trigonal planar coordination, between ≈2.5 and 2.9, with its maximum at the charged state when Cu is most oxidized. Recent work by Maji et al. helps explain this trend, where it was demonstrated that Cu>1+-hole delocalization is highly sensitive to and enabled by higher-coordinate Cu environments in sulfides. Thus, the EXAFS analyses reveal local structural changes that clarify the Cu+/>1+ oxidation mechanism, in which electron removal via distinct Fe2+/>2+ and 2S2–/(S2)2– oxidations occur simultaneously with Cu>1+ hole formation, stabilized by gradual Cu–S bond contraction and slight increases in coordination number throughout charge. This additional Cu>1+ charge compensation may contribute to the greater reversibility (Figure d) and higher rate capability (Figure e) of Li2Cu0.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2 by extending charge compensation during 2S2–/(S2)2– redox beyond local structural distortions required to form and break S–S dimers. Although the Cu+/>1+ oxidation mechanism is now well-characterized, it remains unclear why the multielectron redox capacity limit decreases with increasing Cu content z in Li2.2–z Cu z Al0.2Fe0.6S2 (Figure S5). We previously showed that the capacity limit of Li2.2Al0.2Fe0.6S2 is predominantly determined by the stability of the electrochemically delithiated phase relative to competing phase transitions to both pyrite FeS2 and Al2FeS4. These competing phase transitions were inferred by examining the thermodynamically stable configuration of the hypothetically fully delithiated material, ‘Al0.2Fe0.6S2’. To determine this configuration, we carried out a 900 °C solid-state reaction of 0.2Al + 0.6Fe + 2S, the elemental ratio corresponding to the stoichiometry of the hypothetically fully delithiated material, and found that the majority phases in the product are FeS2 (51.1 wt %) and Al2FeS4 (38.3 wt %) (Figure S26). The structural similarity between Al2FeS4 and pristine Li2.2Al0.2Fe0.6S2 thermodynamically stabilizes the delithiated state by suppressing the competing phase transition to electronically and structurally distinct FeS2. Here, to assess whether Cu alters the competing phase transitions, and thus the capacity limits, in Li2Cu0.2Al0.2Fe0.6S2 and Li1.8Cu0.4Al0.2Fe0.6S2, we carry out 900 °C solid-state reactions of Cu, Al, Fe, and S in stoichiometric ratios corresponding to the hypothetically fully delithiated compositions: 0.2Cu + 0.2Al + 0.6Fe + 2S for ‘Cu0.2Al0.2Fe0.6S2’ (Figure S27) and 0.4Cu + 0.2Al + 0.6Fe + 2S for ‘Cu0.4Al0.2Fe0.6S2’ (Figure S28). Neither reaction yields Al2FeS4. Instead, the majority phases of the first reaction are FeS2 (46.5 wt %) and CuFeS2 (40.0 wt %), and of the second reaction are FeS2 (31.6 wt %) and CuFeS2 (63.8 wt %). We note that the 0.2 equiv of Al in both reactions are likely incorporated into CuFeS2, because CuAlS2 is isostructural with CuFeS2, and thus likely forms a solid solution CuFe1–x Al x S2. In fact, Al substitution in CuFeS2 has been shown previously. , We would expect the lattice parameters of CuFeS2 to increase with Al incorporation, and in fact the Rietveld refinement confirms expanded lattice parameters for CuFeS2, which are tabulated in Table S6, supporting Al incorporation. In all cases, including ‘Al0.2Fe0.6S2’, the only other phase is a minor amount (≲10 wt %) of Fe7S8. Thus, although FeS2 remains significant across all reactions, even the small addition of 0.2 equiv of Cu in ‘Cu0.2Al0.2Fe0.6S2’ relative to ‘Al0.2Fe0.6S2’ significantly alters the thermodynamically stable configuration by eliminating Al2FeS4 in favor of Al-incorporated CuFeS2. Notably, at 0.4 equiv of Cu, Al-incorporated CuFeS2 becomes the dominant phase, exceeding FeS2 by ≈2× (by wt %). While CuFeS2 shares the common feature of tetrahedrally coordinated cations with Li2.2–z Cu z Al0.2Fe0.6S2, CuFeS2 adopts the I4̅2d space group with a face-centered cubic anion framework, structurally dissimilar to the hexagonal close-packed anion framework of Li2.2–z Cu z Al0.2Fe0.6S2. Cu eliminates the stabilizing competing phase transition to Al2FeS4 and replaces it with the destabilizing transition to structurally dissimilar CuFeS2, which ultimately dominates the instability of the delithiated phase. We therefore hypothesize that the capacity decreases with increasing Cu content z because Cu destabilizes the delithiated state. To assess the relative thermodynamic and kinetic stabilities of delithiated Li2.2–x Al0.2Fe0.6S2 and Li1.8–x Cu0.4Al0.2Fe0.6S2, we conduct annealing and resting experiments. For both experiments, we charge the cathode to the transition point, stop the cell, remove and rinse the cathode, then either anneal it in an evacuated ampule (≈50 mTorr) at 200 °C for 2 h or rest it at room temperature under the same static vacuum for ≈36 h, and finally reassemble a new cell with the annealed or rested cathode. We evaluate the stability of the materials at the transition SOC. Structurally, the onset of 2S2–/(S2)2– oxidation after the transition SOC causes gradual loss of long-range order. Therefore, choosing the transition SOC enables us to evaluate the inherent structural stability of the crystalline delithiated phase versus competing phase transitions to other crystalline phases. Electronically, at the transition SOC, we observe Cu>1+ (if Cu is present), Fe>2+, and S2–. This closely matches chalcopyrite CuFeS2, which has oxidation states Cu(1+n)+, Fe(3–n)+, and S2–, where 0 ≤ n ≤ 1, but differs from pyrite FeS2, which has lower-valent oxidation states: Fe2+ and (S2)2–. Thus, while the transition SOC is electronically compatible with CuFeS2, it is unstable with respect to FeS2, which would require ligand-to-metal charge transfer to reduce Fe and dimerize S. Accordingly, by comparing the oxidation states before and after annealing, we can assess the stability of the electrochemically oxidized state relative to reorganization into the oxidation states that define the thermodynamically stable, fully delithiated phases. The annealing and resting galvanostatic data, ex situ XRD, and Fe K-edge XAS are in Figure . Additional S and Cu K-edge XAS are in Figures S29 and S30, respectively. First we briefly summarize the Li2.2Al0.2Fe0.6S2 data here, which is reproduced from Patheria et al. The charge curves after annealing and resting retain the shape and full capacity of the uninterrupted 2S2–/(S2)2– plateau, with a small initial Fe2+/>2+ oxidation-like feature (Figure a). The XRD (Figure b) shows that the crystalline delithiated state is stable against FeS2 formation. In the XRD pattern of the annealed state, this is evidenced by the increased intensity of the (0 0 1) reflection and the absence of any FeS2 reflections (Figure S31a), particularly the most intense (2 0 0) peak expected at ≈33.0° 2θ. Although we do not observe significant changes in the S K-edge (Figure S29a), the Fe K-edge shows clear evidence of Fe reduction. This is indicated by a shift of the rising edge to lower energy (Figure c) to an intermediate value between x (7117.2 eV) and y (7118.2 eV), emphasized by the first derivatives (Figure d), and a decrease in intensity of the pre-edge, marked by w, at 7113.0 eV (Figure d). While the source of electrons for this reduction is unclear, we previously argued that the charge compensation is likely complex, resembling unresolved oxidation state assignments in ‘Fe-deficient FeS’, Fe7S8. Nonetheless, despite electronic reorganization that reduces Fe, the delithiated phase is structurally stable against conversion to FeS2, supporting deep delithiation via 2S2–/(S2)2– oxidation. (a) Charge curves of Li2.2Al0.2Fe0.6S2 up to the transition point, after which the cell is stopped, disassembled, the cathode is rinsed in DMC, dried in vacuum, annealed at 200 °C under static vacuum (≈50 mTorr) for 2 h, or rested at room temperature under the same vacuum conditions for ≈36 h, and then assembled in a new cell for galvanostatic cycling. The standard galvanostatic cycling charge curve for uninterrupted cycling is indicated by the black, dashed line in panel (a). (b) Ex situ XRD of Li2.2Al0.2Fe0.6S2 focused on the 2θ ranges of the (0 0 1) reflection, and the (2 0 0) reflection of pyrite FeS2 in the pristine, transition, and annealed states. (c) Fe K-edge XAS spectra, (d) magnified pre-edge region, and (e) first derivative of the rising edge regions for Li2.2Al0.2Fe0.6S2 in the pristine, transition, and annealed states. As before in Figure , the dashed lines in panels (c), (d), and (e) indicate the approximate positions of the pre-edge w, at 7113.0 eV, and the two rising edges observed at different SOCs, x and y, at 7117.2 and 7118.2 eV, respectively. Panels (f), (g), (h), (i), and (j) contain the same corresponding data for Li1.8Cu0.4Al0.2Fe0.6S2. Panel (g) focuses on the 2θ ranges of the (0 0 1) reflection of Li1.8Cu0.4Al0.2Fe0.6S2 and the (2 0 4) reflection of chalcopyrite CuFeS2. The dashed lines in panels (h), (i), and (j) indicate the approximate positions of the pre-edge w, at 7113.0 eV, and the two rising edges observed at different SOCs, x′ and y′, at 7118.5 and 7119.7 eV, respectively. The data for Li2.2Al0.2Fe0.6S2 is adapted with permission from Patheria et al. Copyright 2025 American Chemical Society. We now discuss the same experiments for Li1.8Cu0.4Al0.2Fe0.6S2. The charge curves after annealing and resting (Figure f) partially resemble the uninterrupted 2S2–/(S2)2– plateau, with even smaller initial Fe2+/>2+ oxidation-like features than for Li2.2Al0.2Fe0.6S2. However, both have markedly lower capacities than the uninterrupted curve. XRD (Figure g) reveals that the crystalline delithiated phase is unstable against CuFeS2 formation. The (0 0 1) reflection of the annealed phase is significantly diminished, indicating a loss of crystallinity of the original structure. New reflections corresponding to ‘CuFeS2’ also appear, most prominently the intense (2 0 4) peak at ≈49.2° 2θ. Rietveld refinement of this pattern (Figure S31b) estimates the ‘CuFeS2’ content at ≈11 wt %. We hypothesize that this ‘CuFeS2’ likely contains Al3+ substituted at the Fe site based on its expanded lattice parameters relative to pristine CuFeS2, which are tabulated in Table S6. We observe no FeS2 reflections, indicating that the competing phase transition to CuFeS2 indeed dominates. Despite this major structural change, the S (Figure S29b), Cu (Figure S30), and Fe (Figure h–j) K-edges show no detectable changes after annealing. This is consistent with the oxidation states Cu(1+n)+, Fe(3–n)+, and S2–, in CuFeS2, which approximately match those at the transition SOC in Li1.8–x Cu0.4Al0.2Fe0.6S2: Cu>1+ (Figure ), Fe>2+ (Figure b,d,f), and S2– (Figure c). Compared to Li2.2–x Al0.2Fe0.6S2, the lack of edge shift in the Fe K-edge spectra of Li1.8–x Cu0.4Al0.2Fe0.6S2 before and after annealing indicate higher-valent Fe>2+ is more stable, with spectra of the transition and annealed states overlaying exactly (Figure h). The rising edge shifts from x″ (7118.5 eV) in the pristine state to y″ (7119.7 eV) in both the transition and annealed states, highlighted by the first derivatives (Figure i). The pre-edge w (7113.0 eV) also remains unchanged between the transition and annealed states (Figure j). This demonstrates that although Cu>1+ and Fe>2+ are electronically stabilized in Li1.8–x Cu0.4Al0.2Fe0.6S2, the delithiated phase is destabilized by the strong structural driving force of the competing phase transition to form ‘CuFeS2’. The lower capacity after annealing is likely due to structural degradation associated with this phase transition. However, the material also appears to degrade as a result of resting, as evidenced by the decreased charge capacity, even though this degradation is not captured by the formation of a new crystalline phase. The ability of the material to change as a result of simply resting at room temperature highlights the instability imposed by incorporation of Cu. Temperature-controlled operando XRD in future work could help verify this hypothesis. Thus, although incorporation of Cu appears to prevent Fe reduction during annealing of the oxidized material, it instead drives a structural rearrangement upon annealing that reveals a thermodynamic preference for forming CuFeS2, thereby limiting the charge capacity. Conclusion and Outlook The materials presented in this work expand the compositional space of industrial element Li-ion cathodes introduced in our previous work. , We demonstrate a Cu>1+-enabled multielectron redox mechanism that involves continuous charge compensation across otherwise separate Fe2+/>2+ and 2S2–/(S2)2– redox processes. We previously showed that 2S2–/(S2)2– redox invokes electronically isolated nonbonding S 3p states and localized structural distortions to form and break S–S bonds. ,, In contrast, we reveal a complementary mechanism that coexists with anion–anion bonding: the highly covalent Cu–S interaction, described in the Introduction, enables charge compensation to extend beyond electronically isolated S–S bonds by stabilizing holes on Cu as Cu>1+. This leverages the natural tendency of anion redox processes to form and break anion–anion bonds, while promoting greater reversibility possibly due to the redox activity of both Cu>1+ and the anion in the plateau region of the charge. However, improvements in reversibility enabled by Cu are convoluted by a concurrent decrease in capacity with increasing Cu content, which we attribute to Cu-induced structural destabilization of the delithiated phase. Ultimately, the unique Cu–S interaction offers a strategy to address a challenge in developing multielectron redox cathodes: bridging electronically isolated transition metal and anion redox processes by extending charge compensation beyond the sites of bond formation and cleavage. Experimental Section Material Preparation All materials and precursors were handled inside an Ar-filled glovebox (H2O and O2 ≲1 ppm). All Li2.2–z Cu z Al0.2Fe0.6S2 (0 ≤ z ≤ 0.4) materials were prepared by solid-state synthesis. Powders of Li2S (Thermo Fisher Scientific, 99.9%), Cu (Thermo Fisher Scientific, 99%), Al2S3 (Thermo Fisher Scientific, 99+%), FeS (Sigma-Aldrich, 99.9%), and FeS2 (Thermo Fisher Scientific, 99.9%) were weighed to an accuracy of ± 0.1 mg to give total 250 mg of a desired stoichiometry (i.e., value of z) and then hand-mixed in an agate mortar and pestle. The mixed precursor powders were pressed into 14 inch diameter cylindrical pellets with a hand-operated arbor press. The mixed precursor pellets were light gray in color. Pellets were placed inside carbon-coated vitreous silica ampules (10 mm i.d., 12 mm o.d.), evacuated to ≤50 mTorr, and sealed with a methane-oxygen torch without exposure to air. The ampules were coated by first rinsing the inside of the empty ampule with acetone, and then pyrolyzing residual acetone with a methane-oxygen torch. This was repeated at least twice for conformal, continuous coating. The evacuated and sealed ampule was then placed inside a box furnace and heated at 1 °C/min to 900 °C with a dwell time of 12 h. After ambient cooling to room temperature (approximately 1 °C/min), the ampules were opened inside the glovebox and the pellets were ground into fine powders in agate mortar and pestles for further characterization. The materials mostly retained the shape of the original pressed pellet after being heated to 900 °C. All products were black in both pellet and powder forms. Synchrotron X-ray Diffraction (sXRD) Sample Preparation As-prepared materials were packed into individual 0.5 mm (outer diameter) quartz capillaries (Hampton Research) in an Ar-filled glovebox. Each quartz capillary was manually broken at ≈1.5 cm and sealed with epoxy (Torr Seal). After curing, the capillaries were removed from the glovebox and nested inside 0.711 mm Kapton capillaries (provided by the Stanford Synchrotron Radiation Lightsource at SLAC National Accelerator Laboratory), which were sealed on one end with modeling clay. High-resolution sXRD patterns were collected on the sample-loaded capillaries at beamline 2–1 (λ = 0.729371 Å) at the SSRL at SLAC National Accelerator Laboratory. The diffraction patterns were fit using the Rietveld method with the General Structure Analysis System II, and crystal structures were visualized with VESTA. Freestanding Cathode Preparation Freestanding cathodes were prepared by mixing 60/20/20 (wt %) active material, carbon (Super P, Alfa Aesar, >99%), and PTFE binder (Sigma, 1 μm powder), respectively, in agate mortar and pestles. The active material and carbon were mixed first, then binder was added to evenly distribute the active material and carbon in the binder framework. The hand grinding with binder creates small flakes (≈1 mm2) that were broken into smaller pieces/a powder by hand with a stainless steel spatula. Roughly 6–10 mg of the composite fragmented mix was weighed and pressed into a 6 mm diameter electrode under ca. 2 tons of force using a manual hydraulic press (Vivtek), resulting in a freestanding cathode with a thickness of approximately 100 μm and an areal loading of roughly 12.7–21.2 mg·cm–2 of active material. Electrochemical Characterization All electrochemical cells were assembled in an Ar-filled glovebox. Li-foil anodes with a diameter of 12 inch were punched from either Li ribbon (Sigma, 99.9%, 0.75 mm) or Li chips (AOT Battery, 99.9%), both first mechanically cleaned with an Xacto blade immediately prior to cell assembly. Each Li anode, ≈0.75 mm thick, weighed ≈45 mg before cleaning, with minimal loss in both mass and thickness afterward, as the cleaning primarily removed the surface passivation layer. The electrolyte, 1 M LP30, was a 1 M solution of LiPF6 (Oakwood chemical, battery grade) in a 1/1 (by volume) mixture of ethylene carbonate (EC) and dimethyl carbonate (DMC), both Sigma, ≥99%. The electrolyte was made (and stored) in an HDPE bottle by combining the carbonates and the salt, and was first stored at least overnight before use to ensure all components dissolved/mixed well. All electrochemistry was performed in 2032 coin cells (MTI) assembled with a ≈0.88 g stainless steel top cap, ≈0.18 g stainless steel spring (MTI), ≈0.73 g stainless steel spacer (MTI), Li anode on the spacer, ≈17 mg 12 inch diameter glass-fiber separator (Whatman, GF/D), 100 μL of LP30 electrolyte (30 μL on the anode, 40 μL on the separator, 30 μL on the cathode), freestanding cathode, and ≈0.90 g stainless steel bottom can. The typical total mass of an as-assembled coin cell was ≈2.9 g. All stainless steel coin cell components were sonicated in roughly 1/1 acetone/isopropyl alcohol for 30 min and then dried overnight in a vacuum oven at 60 °C prior to use in the glovebox. The glass fiber separators were dried overnight in a vacuum oven at 60 °C prior to use in the glovebox. The coin cells were crimped shut with a manual crimper (Pred Materials). All electrochemical experiments were performed at room temperature (≈25 °C) with a BCS 805 battery cycler (Bio-Logic). For continuous galvanostatic cycling experiments, all materials were charged at a C/10 rate based on 1 e– per f.u. up to 3 V, and discharged at the same rate to 1.7 V. For GITT experiments, currents were applied at the same C/10 rate for 20 min at a time separated by 4 h rest periods. The equilibrium potential V eq and overpotential η was extracted from the GITT using Python. The discharge capacity fade rate (in % per cycle) for cycling at C/10 over 25 cycles was determined by calculating linear fits of the average discharge capacities versus cycle number of three replicate cells, and then dividing the slope over the value of the fit function at x = 1 (cycle 1). For the rate capability tests, the same charge/discharge voltage cut-offs of 3/1.7 V were used, and the cells were cycled for 5 cycles each, sequentially, at C/10, C/5, C/2, 1C, and back to C/10 (25 cycles total). All Ex Situ Samples All ex situ samples were prepared in 2032 coin cells (MTI) with freestanding cathodes as previously described. Samples are first cycled to one of the following SOCs: midslope, transition, midplateau, charged, and discharged. For midslope samples, voltage cutoffs of ≈2.38 V for Li2.2Al0.2Fe0.6S2, ≈2.33 V for Li2Cu0.2Al0.2Fe0.6S2, and ≈2.30 V for Li1.8Cu0.4Al0.2Fe0.6S2 were used. For transition samples, a voltage cutoff of ≈2.54 V was used for Li2.2Al0.2Fe0.6S2, Li2Cu0.2Al0.2Fe0.6S2, and Li1.8Cu0.4Al0.2Fe0.6S2. For midplateau samples, time cutoffs of ≈9.8 h for Li2.2Al0.2Fe0.6S2, ≈9.36 h for Li2Cu0.2Al0.2Fe0.6S2, and ≈8.61 h for Li1.8Cu0.4Al0.2Fe0.6S2 were used. Due to slight cell-to-cell variation, cutoffs for these intermediate SOCs varied slightly and so for clarity are always shown with the full corresponding galvanostatic charge and discharge curves. For charged and discharged samples, voltage cutoffs of 3 and 1.7 V, respectively, were used. After cycling to one of the above-defined cutoffs, the cells were decrimped and opened with a manual disassembling tool (Pred Materials) in an Ar-filled glovebox. The ex situ cathodes were gently scraped off the current collector by hand using a stainless steel spatula, keeping the cathode intact. Any visible glass fiber separator stuck on the cathode was manually scraped off the cathode surface with a stainless steel spatula. The cathodes were then rinsed with ≈300 μL of DMC for 2–3 min to wash away residual electrolyte. Any residual DMC was dabbed with a dry Kim wipe, and then the cathode was dried under vacuum for roughly 30 min. The dry intact cathodes were then either kept intact or broken into smaller pieces or a powder by hand with a stainless steel spatula, depending on the requirements of the subsequent characterization. X-ray Absorption Spectroscopy (XAS) Samples for ex situ XAS were prepared in 2032 coin cells as previously described for ex situ samples. The intact ex situ cathodes, treated as previously described, were broken up into loose powders with a stainless steel spatula. All sample preparation described below was conducted in an Ar-filled glovebox. Prepared sample holders were sealed in Ar in aluminized mylar pouches (2–3 layers) using an impulse heat sealer (Uline) for transport to the respective synchrotrons. Calibration, background correction, and data processing of X-ray absorption near-edge structure was done in Athena from the IFEFIT suite. For Fe and Cu K-edges XAS, the loose powders were loaded into aluminum sample holders provided by the SSRL at SLAC National Accelerator Laboratory, encapsulated between two pieces of Kapton tape (1 mil film thickness, 2.5 mil total thickness, Uline). All Fe and Cu K-edges XAS (except for data of Li1.8Cu0.4Al0.2Fe0.6S2 in Figures h–j and S30 – see below) was measured in transmission mode at beamline 4–3 at the SSRL at SLAC National Accelerator Laboratory. In all cases, during measurement, the sample holder was placed in a continuous He-flushed chamber to minimize air exposure, and O2 levels were measured to be ≈500 ppm with an O2 sensor. Fe and Cu K-edges data were calibrated to collinear Fe and Cu foils, respectively, present for each sample. The data shown are three averaged sweeps of each sample, with each sweep taking roughly 20 min. For Fe K-edge data of Li1.8Cu0.4Al0.2Fe0.6S2 in Figure h–j, the loose powders were mixed by hand with agate mortar and pestles with boron nitride (BN) (Alfa Aesar, 99.5%) so that the total sample concentration gave a calculated edge jump of ≈1 – this was ≈14 mg of loose pellet powder for every ≈13 mg of BN. Roughly 27 mg of each composite BN-sample mix was pressed into 38 inch diameter pellets using a hand-operated arbor press. For Cu K-edge data of Li1.8Cu0.4Al0.2Fe0.6S2 in Figure S30, the undiluted loose powders were directly pressed into into 38 inch diameter pellets using a hand-operated arbor press. X-ray absorption near edge spectroscopy (XANES) measurements were performed using a EasyXAFS300+ XAS/XES spectrometer at ambient temperature and pressure in the XAFS transmission mode set up. Transmission mode spectra were collected in a Rowland circle geometry. Samples were irradiated with a Mo X-ray source operating at 30 kV/15 mA for Fe K-edge measurements, and 35 kV/20 mA for Cu K-edge measurements. A beam mask size of 4.5 × 2.0 mm was used for all measurements. For Fe K-edge XANES measurements, a Si­(1,1,0) crystal analyzer (r = 495) at the fourth harmonic was used. A crystal monochromator offset of 0.099 was measured and applied from a crystal theta alignment scan on the Si­(1,1,0) crystal performed during instrument calibration. The collected Fe XANES spectra were averaged from 12 scans of each sample and 12 scans of I0. For Cu K-edge XANES measurements, a Si­(5,5,1) crystal analyzer (r = 495) at the first harmonic was used. A crystal monochromator offset of −0.007 was measured and applied from a crystal theta alignment scan performed on the Si­(5,5,1) crystal during instrument calibration. The collected Cu XANES spectra were averaged from 10 scans of each sample and 10 scans of I0. XANES spectra were calibrated to appropriate metallic reference foils (Fe: 7.5 μm; Cu: 7.5 μm; EXAFS Materials, Danville, CA) measured during the same experimental runs. For S K-edge XAS, the loose powders were mixed by hand with agate mortar and pestles with boron nitride (BN) (Alfa Aesar, 99.5%) so that the total sample concentration was ≤5% by mass. We first describe the sample preparation for S K-edge XAS data in Figure a of Li2.2Al0.2Fe0.6S2, and in Figure S29 of Li1.8Cu0.4Al0.2Fe0.6S2. Roughly 10–15 mg of each composite BN-sample mix was pressed into 14 inch diameter pellets using a hand-operated arbor press. The pellets were then loaded into plastic sample holders provided by the National Synchrotron Light Source II (NSLS-II) at Brookhaven National Laboratory (BNL), sandwiched between a polypropylene layer and Kapton tape and adhered to the sample holder using Kapton tape. S K-edge XAS was measured in fluorescence mode at beamline 8-BM at NSLS-II at BNL. During measurement, the sample holder was placed in a continuous He-flushed chamber to minimize air exposure. S K-edge data were calibrated to a gypsum, i.e., sulfate S6+, standard (1 wt % CaSO4·2H2O in polyethylene glycol). The data shown are three averaged sweeps of each sample, with each sweep taking roughly 15 min. We now describe the sample preparation for S K-edge XAS data in Figure c of Li1.8Cu0.4Al0.2Fe0.6S2, and in Figure S14 of Li2Cu0.2Al0.2Fe0.6S2. Each composite BN-sample mix was loaded into aluminum sample holders provided by the SSRL at SLAC National Accelerator Laboratory, encapsulated between one piece of polyester acrylic tape (1 mil thickness, CHR M60, Saint-Gobain) and polypropylene film (5 μm thickness, SPEX 3520). S K-edge XAS was measured in fluorescence mode at beamline 4–3 at the SSRL at SLAC National Accelerator Laboratory. During measurement, the sample holder was placed in a continuous He-flushed chamber to minimize air exposure, and O2 levels were measured to be ≈500 ppm with an O2 sensor. The data shown are two averaged sweeps of each sample, with each sweep taking roughly 15 min. CuKα XRD CuKα XRD patterns were collected on a Rigaku SmartLab diffractometer. To prevent air exposure during measurement, samples were loaded inside an Ar-filled glovebox into a Rigaku-built air-free sample holder with a low background silicon sample platform. For synthesized materials, roughly 10 mg of sample powder was placed and compressed (by hand using a stainless steel spatula) onto the sample platform. For ex situ cathodes, the cathode was kept intact and gently placed onto the sample platform. The diffraction patterns were fit using the Rietveld method with the General Structure Analysis System II (GSAS-II) and crystal structures were visualized with VESTA. Fe and Cu K-Edges EXAFS Fitting We use the Athena and Artemis software from the IFEFIT suite for Fe and Cu K-edges EXAFS fitting. In Fe and Cu K-edges XAS, oscillations at energies beyond the primary electronic transition arise from the interference of the excited photoelectron with itself after scattering off neighboring atoms. The oscillations are converted to a function of the wavenumber, χ­(k), which is then Fourier transformed to real space (|χ­(R)|). The χ­(k) are fit within a k window of roughly ≈3.0 ± 0.1 to ≈10 ± 1 Å–1 (sample dependent), with dk = 2 Å–1. The |χ­(R)| are fit within an R window of roughly ≈1.1 ± 0.1 to ≈10 Å­(sample dependent), with dR = 0.2 Å. The amplitude reduction factor S 0 2 for both Fe and Cu K-edges EXAFS is held fixed for all ex situ samples and is determined by fitting the first shell of the χ­(k) and |χ­(R)| of the Fe and Cu calibration foils, using a k window of 5–11 Å–1 (dk = 2 Å–1) and R window of 1–3 Å (dR = 2 Å). The intensity of the Fourier transform, |χ­(R)|, represents the oscillation intensity in real space, corresponding to correlation shells, with either Fe or Cu located at 0 Å. We fit the first and second shell correlations with defined scattering paths and determine the correlation distance, r eff. (Å), and the coordination number N. The first shell describes the immediate coordination of Fe or Cu by S, and the second shell describes the nearest neighbor cations at the edge-sharing tetrahedral sites closest to a given Fe or Cu. We model the second shell using only Fe atoms for both Fe and Cu K-edge EXAFS, since Fe is the majority cation compared to Cu and Al, and Li contributes negligibly to the scattering intensity. The EXAFS analysis is conducted at the pristine, transition, charged, and discharged SOCs. The resolution of our EXAFS data (the oscillations ≳50 eV above the rising edge), limits us to fitting both the first and second shells using only single scattering paths. This captures overall changes but not specific atomic positions or heterogeneity thereof. For example, the first shell actually includes 4 separate S atoms, each with its own scattering path, but we fit the first shell with a single scattering path, capturing the average effect of the multiple S atoms through the coordination number N. The EXAFS second shell r eff. and N are shown in Figure S19, with the fitted χ­(k) and |χ­(R)| in Figures S20–S22 and S23–S25, respectively, and all fit parameters in Table S5. Syntheses of Hypothetically Fully Delithiated Phases The material and precursors were handled in an Ar-filled glovebox. Powders of Cu (Thermo Fisher Scientific, 99%), Al (Alfa Aesar, 99.5%), Fe (Acros Organics, 99.0%), and S8 (Acros Organics, >99.5%) were weighed to an accuracy of ±0.1 mg to give total 250 mg of each fully delithiated composition and then hand-mixed in an agate mortar and pestle. The mixed precursor powders were pressed into 14 inch diameter cylindrical pellets with a hand-operated arbor press. The mixed precursor pellets were gray in color. The pellets were each placed inside separate vitreous silica ampules (10 mm i.d., 12 mm o.d.), evacuated to ≤50 mTorr, and each sealed with a methane-oxygen torch without exposure to air. The evacuated and sealed ampules were then placed inside a box furnace and heated at 1 °C/min to 900 °C with a dwell time of 12 h. After ambient cooling to room temperature (roughly at a rate of 1 °C/min), the ampules were opened inside the glovebox. All pellets after heating were a glittery dark gray color and fully melted, conforming to the shape of the ampule with slight bright yellow S residues stuck to the sides of the ampules. Ex Situ Annealing and Resting For annealing and resting experiments, the cathode was first charged to the transition SOC. The cell was then stopped and disassembled in an Ar-filled glovebox, and the cathode was removed and rinsed as previously described. The intact cathode was then placed in vitreous silica ampules (10 mm i.d., 12 mm o.d.), evacuated to ≤50 mTorr, and sealed with a methane-oxygen torch without exposure to air. For annealing experiments, the cathode in the evacuated ampule was placed inside a box furnace and then heated at 1 °C/min to 200 °C, then annealed at 200 °C for 2 h, and then ambiently cooled to room temperature by shutting off the furnace. For resting experiments, the cathode in the evacuated ampule was kept at room temperature for ≈1 week. The ampule containing the annealed or rested cathode was then opened inside an Ar-filled glovebox, and a new cell was assembled as previously described with the annealed or rested cathode. Supplementary Material