Elucidating and controlling phase integration factors in Co-free Li-rich layered cathodes for lithium-ion batteries

Abstract

Li- and Mn-rich layered oxides (LLOs) with a Co-free composition are promising candidates for next-generation cathodes in low-cost and high-energy-density lithium-ion batteries. Despite their potential, the commercialization of Co-free LLOs encounters several electrochemical challenges, such as low activity and initial coulombic efficiency of the first activation cycle and compromised cycle retention, which are primarily attributed to the poor phase integrity between LiTMO₂ and Li₂MnO₃ domains. In this study, we identified that the compromised phase integrity in Co-free LLOs can be driven by the sticking Ni²⁺ compositional design, which induces Li–Ni site-exchange defects in the LiTMO₂ domain, leading to severe TMO₂ slab mismatches between phases and resulting in a penalty in enthalpy mixing energy. To address this, we proposed a rational off-stoichiometric compositional design. By introducing a slight excess of Li, the Ni valence state shifts slightly from +2 to +3, reducing the superexchange interaction and significantly suppressing site exchange formation. The off-stoichiometric Co-free LLO shows highly integrated domains, markedly improving all electrochemical parameters, including coulombic efficiency, cycle stability, and voltage decay. These findings deepen the understanding of designing domain structures to enhance the redox chemistry of the LLO cathode class.

---

New concepts

Co-free Li- and Mn-rich layered oxides (LLOs) are highly promising cathode materials, offering both high energy density and cost efficiency for lithium-ion batteries. However, the poor integration between the LiTMO₂ and Li₂MnO₃ domains significantly hinders their electrochemical performance, limiting the materials' ability to achieve their theoretical potential and creating a major obstacle to commercialization. In this study, we identified the critical factors that undermine phase integrity in Co-free LLOs and, based on these findings, we introduced a rational off-stoichiometric composition design to achieve highly integrated domain structures. The super-exchange interactions of Ni²⁺ trigger Li–Ni site exchange, which increases the misfit between the TMO₂ slabs of the two phases, thermodynamically inhibiting domain mixing. To overcome this, we propose a new off-stoichiometric composition that incorporates a slight lithium excess, which increases the oxidation state of Ni and reduces the adverse magnetic interactions, leading to improved phase integration. This new compositional guidance universally strengthens the phase integrity of Co-free LLOs and significantly enhances key electrochemical performance metrics, including cycle retention, voltage decay, and energy density. These findings provide a cost-effective and strategic framework for designing domain structures that optimize the redox chemistry of LLO cathodes.

Introduction

The rapid increase in the demand for electric vehicles (EVs) has driven significant advancements in lithium-ion batteries (LIBs) technology. Since the energy density and cost of LIBs are heavily influenced by the cathode,^1–3 this has spurred extensive research over the past two decades focused on developing new classes of cathode materials. Among the various materials explored, Li- and Mn-rich layered oxides (LLOs) with the formula Li_1+xMn_yNi_zCo_{1−x−y−z}O₂ (where x + y + z = 1; >280 mAh g⁻¹) have garnered substantial interest due to their potential for low cost^1,4 and the unique electrochemical activity of anionic redox chemistry, which overcomes the capacity limitations of traditional transition metal redox and makes LLOs promising for next-generation EV cathodes.^5–7

The ‘over-lithiated’ oxide cathode possesses a unique structural feature that allows the simultaneous utilization of both cationic and anionic redox chemistries. It consists of two crystallographically distinct phases; the rhombohedral LiTMO₂ phase (TM = Ni, Co, Mn, R phase) and monoclinic Li₂MnO₃ phase (M phase).^8,9 These phases are represented by the formula xLiTMO₂ + (1 − x)Li₂MnO₃ (0 < x < 1) and share an O3 oxygen stacking sequence, enabling coherent integration of the R and M phase domains.⁸ In the R phase domain, the structure is primarily defined by sequential Li–O–TM bonds, where each oxygen atom is coordinated by three transition metals and three lithium ions. The primary redox reactions in the R phase are dominated by the 3d orbital of the transition metal.⁷ In contrast, the M phase features Li–O–Li bonding configurations, with each oxygen atom coordinated by two transition metals and four lithium ions, distinguishing it from the LiTMO₂ phase. This configuration generates a non-bonding O 2p state close to the Fermi level due to an energy gap between the O 2p and Li 2s orbitals, preventing orbital hybridization.⁷ Additionally, recent theories suggest that the t_2g orbitals of transition metals in the M phase form weak π bonds with O 2p orbitals, thereby stabilizing these non-bonding O 2p states.¹⁰ This unique feature allows Li-rich layered oxides to extract additional electrons from these non-bonding states, boosting their electrochemical capacity.

Despite their superior energy density achieved through the integration of cationic and anionic redox chemistries, LLOs face several commercialization hurdles, including voltage hysteresis,⁹ voltage decay,¹¹ surface degradation,¹² low initial coulombic efficiency (ICE),¹³ and lattice oxygen evolution.¹⁴ For example, Li₂MnO₃ domains generate vacancies in the TM slabs at a high state of charge (SoC). During discharge, some of these vacancies are occupied by migrating transition metals rather than being refilled by Li ions. This migration promotes the formation of a Li₆O coordination environment and leads to the localized creation of a highly ionic non-bonding O 2p state.⁹ The elevation of these non-bonding O 2p states leads to progressive voltage decay, contributing to significant charge/discharge hysteresis.⁹ LLO materials also suffer from surface densification caused by Ni migration during charge/discharge cycles or intrinsic site-exchange defects,¹⁵ such as Li–Ni site exchange due to their similar ionic sizes.¹⁶ The Li–Ni anti site defects cause sluggish charge transfer, ultimately leading to degraded electrochemical performance.¹⁷ Additionally, according to a recent report, the low ICE in the first activation cycle has been linked to irreversible phase transitions caused by surface lattice oxygen release and kinetic-related irreversible capacity associated with a metastable phase transition during deep discharge.¹³ Increased lattice mismatch between the R and M phases at high charge has also been suggested to accelerate oxygen gas evolution, further contributing to structural degradation.¹⁴

Recent studies have shown that these electrochemical challenges are closely linked to the degree of ‘homogeneity’ between the R and M domains. Compromised domain integrity increases the possibility of Li coordinating with MnO₆ units, which reduces Mn–O covalency.¹⁸ As a result, the electron density of lattice oxygen depletes further during the charging process in M phase domains, weakening the lattice structure.¹⁸ Moreover, the highly active TM 3d–O 2p electronic structure, typically found at the R–M phase interface, diminishes when the two phases are not coherently mixed.¹⁹ To address these issues, several strategies have been proposed to enhance phase integrity and create more dispersed domains. These include adjustment of synthesis protocols,^18,20 compositional variations,^19,21 and phase-targeted substitution approaches.²² LLOs with highly integrated R–M phases synthesized through these methods exhibit significantly increased anionic redox activity during the initial cycle.¹⁸ Furthermore, the well-integrated phases demonstrate superior cycling performance, attributed to reduced internal strain¹⁴ and the simultaneous suppression of lattice oxygen release.²¹

The challenge of integrating the R and M phases in LLOs is further heightened in Co-free compositional designs. Liu et al. reported that Co³⁺ is more likely to be incorporated into the Li₂MnO₃ phase, while Ni²⁺ shows a higher affinity for the LiTMO₂ phase, indicating the potential for phase separation in Ni²⁺-containing Co-free LLOs.²³ Indeed, several research groups have reported the substantial spatial segregation between Ni²⁺ and Mn⁴⁺ in Co-free LLOs particles, illustrating less entanglement between the Ni-rich R phase and the Mn-dominant M phase.^24–26 Building on these observations, our group conducted a systematic comparison between Co-free and Co-containing LLOs, following the conventional valence state designs for Ni²⁺, Co³⁺, and Mn⁴⁺. This comparative study uses an identical phase ratio (R : M = 50 : 50), comparing Co-free Li_1.2Ni_0.2Mn_0.6O₂ (Ni : Mn = 25 : 75, hereafter NM2575) and Co-included Li_1.2Ni_0.13Co_0.13Mn_0.54O₂ (Ni : Co : Mn = 1 : 1 : 4, hereafter NCM114). The secondary particles, approximately 3 μm in size, were identified through scanning electron microscopy (SEM, Fig. S1, ESI†). Fig. S1 (ESI†) shows synchrotron high-resolution X-ray diffraction (sXRD) patterns of the (003) peak and the superstructure peak induced by the honeycomb structure in the M phase for NM2575 and NCM114, respectively, obtained using monochromatic incident X-rays. Notably, the FWHM (Full Width at Half Maximum), asymmetry, and integrated intensity of these peaks—key indicators of the degree of homogeneity between the R and M phases—are slightly larger and stronger in the Co-free composition compared to the Co-containing composition, even under identical synthetic conditions. This indicates a reduction in phase integrity in the Co-free composition, as observed in previous reports^24–26 (for details, see the ESI†).

In this study, we identified one of the factors for phase integrity in Co-free LLOs and introduced a rational compositional design to achieve highly integrated structures. Our findings reveal that strict adherence to the Ni²⁺ composition plays a critical role in hindering the integration of the R and M phases. When Ni remains in the +2 state in the LiTMO₂ phase, its similar ionic size to Li and strong superexchange interactions increase the concentration of Ni_Li anti-site defects (Ni at Li site). These defects cause significant oxygen framework mismatch between the R and M phases, driving the LLO from a dispersed state to a segregated state. To address this issue in a straightforward manner, we examine an off-stoichiometric compositional strategy: introducing a small excess of Li. This adjustment slightly raises the valence state of Ni from +2 to +3, thereby reducing the strong magnetic interactions of Ni²⁺. Consequently, Ni_Li anti-site defects are significantly suppressed, improving coherence between the R and M phases and promoting better phase integration. This method enables the easy and cost-effective formation of homogeneous domains in phase-segregated LLOs without the need for additional expensive doping strategies or further processing. The resulting highly integrated structure stabilizes anionic redox and surface oxygen while greatly enhancing electrochemical performance, as demonstrated by improved coulombic efficiency and cycle stability and reduced voltage decay. This work sheds light on the mechanisms of domain segregation and offers a new approach for designing homogeneous domains in Co-free LLOs.

Results and discussion

Phase integrity issue in Co-free LLOs

In general, the transition metal design of Co-free LLOs follows the line of Ni²⁺ and Mn⁴⁺, as depicted by the navy line connecting LiNi_1/2Mn_1/2O₂ and Li₂MnO₃ in Fig. 1(a). Increasing the Ni³⁺ content by raising the Ni/Mn ratio (shifting left from the Ni²⁺ line), while keeping the amount of Li and the Mn valence state at +4 constant, leads to a reduction in the Li₂MnO₃ phase fraction and a decrease in the exploitable capacity from anionic redox.^27–29 Alternatively, a compositional design that decreases the Ni/Mn ratio while maintaining Ni in the +2 valence state and increasing Mn³⁺ content (shifting right from the Ni²⁺ line) induces Jahn–Teller distortion, which compromises the structural stability of LLOs and reduces their long-term cycle retention.³⁰ Therefore, to maximize energy density and ensure electrochemical stability, the transition metal redox state design for LLOs typically adheres to the navy line in Fig. 1(a).

Fig. 1 Phase segregation phenomenon in Co-free LLOs. (a) Ternary phase diagram of Li, Ni, and Mn for designing Co-free LLOs. (b) Whole XRD patterns of Co-free LLOs. The yellow-shaded region highlights the (104) peak, which clearly illustrates the trend in peak separation. The inset shows the asymmetry profile of the (104) peak of NM2575. (c) First activation galvanostatic profiles of Co-free LLOs. (d) dQ dV⁻¹ curves of the charging profile during the first activation cycle. (e) Theoretical capacity and first cycle charge capacity of Co-free LLOs.

Model compositions following this general guideline (navy line) were prepared with Ni to Mn ratios of 25 : 75, 35 : 65, and 40 : 60 (Li_1.2Ni_0.2Mn_0.6O₂-NM2575; yellow spot, Li_1.13Ni_0.304Mn_0.565O₂-NM3565; light green spot, and Li_1.091Ni_0.364Mn_0.545O₂-NM4060; green spot). All Co-free LLOs were synthesized under identical conditions, with the Li and TM sources carefully adjusted to achieve the correct stoichiometric ratio (see the synthesis part in the Experimental Methods section). Here, the compositions were designed to gradually increase the fraction of the R phase in order to systematically study its effects (see Fig. S2 for the detailed phase fraction of Co-free LLOs derived from Rietveld refinement, ESI†). The secondary particles, approximately 3 μm in size, were identified through scanning electron microscopy (SEM, Fig. S3, ESI†). Additionally, X-ray absorption spectroscopy (XAS) analysis revealed that the valence states of Ni and Mn were +2 and +4, respectively, across all samples (Fig. S4, ESI†). To assess the degree of R and M domain integration, peak bifurcation was analyzed using mono-wavelength sXRD measurements (Fig. 1(b) and Fig. S5, ESI†). Significant (104) peak separation was observed for the two prepared samples, NM3565 and NM4060, directly demonstrating pronounced domain separation between the R phase and M phase. Although NM2575 displayed a relatively well-merged (104) peak, a broad and asymmetric pattern was still evident in the inset of Fig. 1(b), suggesting compromised integrity within the R–M phase domains. Further analysis of the (003) peak also showed that all Co-free LLO powders exhibited broad peaks (see Fig. S5, ESI†), consistent with previous studies highlighting the limitations of phase mixing in Co-free compositions.

The electrochemical activity test on the prepared Co-free LLO electrodes further suggests compromised phase integrity, as similarly observed in previous reports. The galvanostatic profile for the initial electrochemical activation cycle and the corresponding dQ dV⁻¹ curve are shown in Fig. 1(c) and (d). Both cationic and oxygen redox processes occur in all samples, but they fall short of their theoretical capacities. Additionally, the oxygen oxidation peak shifts to higher voltages as indicated in the dQ dV⁻¹ curve, suggesting slower oxygen redox kinetics resulting from the domain segregation.³¹ This trend was observed in the order of NM2575, NM4060, and NM3565. This order aligns with the tendency for phase segregation, indicating that domain segregation acts as a kinetic barrier to oxygen redox. Fig. 1(e) summarizes the theoretical and actual capacities of each electrode, clearly illustrating the strong link between redox activity and phase integrity. The NM3565 electrode, which exhibits the most severe phase segregation, achieved much lower capacities than its theoretical limits, whereas NM2575, with the best XRD peak merge (i.e., phase integrity), approached its theoretical capacity more closely. Despite exhibiting the highest charge–discharge capacity, NM2575 shows relatively low initial coulombic efficiency during the activation cycle. This is because NM2575 contains a relatively higher M phase fraction compared to other Co-free LLOs, which inherently leads to substantial irreversible capacity due to extensive oxygen reactions. Further electrochemical cycling analysis, as shown in Fig. S6 (ESI†), revealed an increasing capacity trend with continued cycling. This behavior, common in LLO electrodes with poor phase integrity, suggests delayed anionic redox activation.³² The capacity increase observed in NM3565 after 15 cycles will be discussed in the subsequent section on electrochemical performance.

Key factor affecting phase segregation in Co-free LLOs

Combined local and global structural analyses were conducted to elucidate the factors determining phase segregation in Co-free LLOs. High-angle annular dark-field STEM (HAADF-STEM) analysis of the NM3565 sample, which exhibited the most compromised phase integrity, confirmed that the R and M phases formed segregated domains, consistent with previous XRD observations (Fig. S7, and see Fig. S8 for additional confirmation of the domain distribution in other Co-free LLO compositions, ESI†). Notably, another key observation was the significant presence of Ni_Li site exchange defects exclusively within the R phase domain (Fig. 2(a) and Fig. S9, ESI†). The black arrows in the right inset profile of Fig. 2(a) indicate the detection of Ni in the Li slab, whereas the normal layered region without anti-site defects shows no such signal (Fig. S10, ESI†). Global structural analysis through extended X-ray absorption fine structure (EXAFS) confirmed the high concentrations of Ni_Li anti-site defects in the R phase across all prepared samples, extending the local observations to the entire particle level. It is important to note that EXAFS has limited sensitivity to Li element; thus, the ∼3.9 Å peak in the Ni K-edge EXAFS serves as a qualitative marker for assessing the concentration of Ni_Li defects.^18,33 Additionally, since Ni is exclusively located within the LiTMO₂ phase, this peak provides clear information about anti-site defects specifically within the R domain.⁸ For better comparison, LiNiO₂ (LNO) with approximately 2.2% defect concentration was used as a reference. All samples exhibited higher peak intensities at ∼3.9 Å compared to LNO (Fig. 2(b)), indicating higher concentrations of Ni_Li anti-site defects. Notably, the concentration of these defects followed the same order as phase integration quality: NM2575 < NM4060 < NM3565. Further quantitative analysis using XRD Rietveld refinement corroborated the Ni K-edge EXAFS findings, showing that all sample groups had higher Ni_Li anti-site concentrations than LNO (2.2%). The measured defect concentrations—NM2575 (6.2%), NM4060 (9.4%), and NM3565 (11.5%)—emphasize the strong relationship between structural integrity, electrochemical performance, and Ni_Li anti-site defects (see Tables S2–S4 for more details, ESI†). Although NM4060 has the highest Ni content, NM3565 shows a slightly higher Li–Ni anti-site defect concentration. This can be attributed to the fact that NM3565, with its higher surface area, experiences more severe Li/O loss at the surface during high-temperature synthesis, resulting in increased defect formation.^34,35

Fig. 2 Origin of phase segregation behavior. (a) HAADF-STEM image along the [110]_R direction of NM3565 displaying a Ni_Li anti-site in the R phase. (b) Ni K-edge EXAFS region of Co-free LLOs. (c) Average distance between 10 slabs in the layered and anti-site regions, as measured from the HAADF-STEM image. (d) The c-lattice parameters and TM slab distance for the R and M phases obtained through Rietveld refinement. (e) Scheme of achieving coherent lattice alignment between phases with different TM slab distances. (f) DFT calculation of the energy penalty associated with varying slab distances of R and M phases to form a coherence lattice (where Δd/d_I represents (d_M − d_I)/d_I, with d_M as the modified slab distance for phase matching and d_I as the initial slab distance of the relaxed R and M phases).

Defects in cathode materials have a profound impact on their physical and electrochemical properties. For example, Ni migration on the surface of high-Ni layered cathodes induces a fatigue phase, leading to a pinning effect within the c-lattice. This effect restricts lattice flexibility and diminishes electrochemical activity.¹⁷ Similarly, Ni_Li site exchange defects cause lattice changes in the R phase. A comparison of the average slab distance with varying levels of Ni_Li anti-site defects (Fig. 2(c)) shows a modest 0.8% increase in the c-axis lattice parameter, from 4.82 nm to 4.86 nm, which aligns with previous observations.^17,36,37 XRD refinement of the NM3565 sample further confirms this, revealing a lattice mismatch of about 0.012 Å between the expanded c-lattice of the R phase and the M phase (Fig. 2(d)).

It is well known that even a small amount of Ni_Li defects in the R phase can significantly increase the distance of the TM slab while reducing the distance of Li slab.^38,39 Therefore, it is essential to prioritize the energy penalties associated with phase mixing caused by the expansion of the TM slab and contraction of the Li slab in the R phase domain. Since the TM–O bond is significantly stronger than the Li–O bond^7,40,41 (see Fig. S11 for a more detailed and precise analysis of our system, ESI†), the changes in the TM slab distance are expected to play a more critical role in limiting phase mixing. Our refined analysis revealed that the TM slab distance in the R phase, increased by anti-site defects, was 8.3% larger than the M phase slab distance in the most segregated composition, NM3565. Additionally, the TM slab distance in the R phase of NM4060 and NM2575 expanded by 6.8% and 3.6%, respectively, compared to the M phase (Tables S2 and S4, ESI†). On the other hand, the M phase generally exhibits negligible site exchange defect concentration even at a high calcination temperature of 900 °C.⁴² Given that most materials struggle to accommodate lattice mismatches greater than 5%, this degree of lattice incoherency is significant.⁴³

A key requirement for homogeneous domain formation is minimizing lattice mismatch between phases. Failing to do so increases interface strain energy, raises Gibbs free energy, and reduces thermodynamic solubility.^43,44 In LLO materials, maintaining oxygen stacking in an O3 sequence is essential, making TM slab alignment essential for maximizing thermodynamic mixing feasibility. However, our findings suggest that the significant TM slab mismatch caused by anti-site defects drastically increases the enthalpy mixing energy penalty. Density functional theory (DFT) calculations were conducted to further investigate the effects of TM slab distance variation and Ni_Li anti-site defects in the R phase on the material. Since NM3565, the most phase-segregated sample, exhibits an anti-site defect concentration of approximately 11.5%, we used an optimized model with a slightly higher concentration of 16% for further analysis. Table S5 (ESI†) presents the lattice parameters at the ground state for the R phase (LiNi_0.5Mn_0.5O₂), the R phase with anti-site defects ((Li_0.84Ni_0.16)(Li_0.16Ni_0.34Mn_0.5)O₂), and the M phase (Li₂MnO₃). The relaxed R phase with anti-site defects shows a 0.02 Å expansion of the TM slab compared to the defect-free R phase, consistent with experimental results.

To ensure coherent coexistence between the R and M phase domains, the slab mismatch between these phases must be addressed. This can be achieved by either: (i) contraction of the TM slab in the R phase or (ii) expansion of the TM slab in the M phase (Fig. 2(e)). We evaluated the energy penalty required to adjust the slab distances for phase integration (Fig. 2(f)). The zero-energy point indicates the relaxed state of each phase, representing its unstrained original state when existing independently (i.e., in an unmixed state). The star-shaped markers identify the points at which each phase achieves slab matching with the other, corresponding to the maximum energy penalty during the adjustment process. As slab distance variation increases toward the mixed state, the energy required for phase integration rises accordingly.

When anti-site defects are incorporated into the R phase, the TM slab distance deviates more significantly from that of the M phase compared to the defect-free R phase. Consequently, aligning the R phase containing Ni_Li defects with the M phase requires greater contraction of the TM slab than is needed for the defect-free R phase. On the other hand, the steeper slope of the strain curve in the absence of anti-site defects arises because the TM slab consists entirely of TM–O bonds. Consequently, contracting or shrinking the TM slab demands relatively higher strain energy. Similarly, the M phase experiences greater TM slab expansion and encounters a higher energy penalty to align with the defect-containing R phase. According to the calculation results, the defect-free R and M phases must overcome maximum energy barriers of 4.6 meV f.u.⁻¹ and 3.4 meV f.u.⁻¹, respectively. In contrast, the defect-rich R and M phases encounter penalties that are four times higher, at 12.6 meV f.u.⁻¹ and 12.2 meV f.u.⁻¹, respectively, for phase alignment. Moreover, since the formation energy of anti-site defects is positive (Table S5, ESI†), the LLO particle is inherently in a more unstable state, making phase mixing less thermodynamically favorable. The energy penalty was further confirmed upon the formation of a fully coherent interface (Fig. S12, ESI†). In this model, the defect-rich R phase requires an additional 72.2 meV f.u.⁻¹ in energy compensation compared to the defect-free R phase to achieve phase alignment with the M phase. As phase formation naturally favors energy reduction, the R and M phases tend to form in ways that minimize their interfacial area. This increase in energy penalty promotes domain segregation into distinct R and M phases, hindering the formation of a coherent interface and complicating dispersed domain formation.

Rational composition design for reduction in defect concentration

Our findings suggest that Ni_Li anti-site defects in the R phase are some of the critical factors in disrupting oxygen framework coherency between the R and M phase domains (Fig. 3(a)). Therefore, it is crucial to develop synthesis strategies aimed at substantially reducing these anti-site defects within the material. The formation of the Ni_Li anti-site is well-established to result from the close ionic radii size between Ni²⁺ (69 pm) and Li⁺ (76 pm).¹⁶ Furthermore, recent studies have shown that the antiferromagnetic interaction facilitates this site-exchange due to the robust 180° super-exchange interactions within the Ni²⁺–O–Ni²⁺ configuration^45,46 (Fig. 3(b)); the Ni²⁺ ion possesses two unpaired spins in the e_g orbital and forms a linear alignment with the O 2p orbital, resulting in a stronger superexchange interaction compared to other ions such as Ni³⁺, Co³⁺ and Mn⁴⁺. In short, the Ni²⁺ oxidation state inevitably triggers undesirable Li–Ni site exchange in the layered structure.

Fig. 3 Off-stoichiometric composition design for homogeneous domain. (a) Scheme of slab structure alterations in layered oxide due to Ni_Li anti-site defects. (b) Scheme of the magnetic interaction in a layered oxide cathode with and without anti-site defect induced magnetic interaction. (c) Guideline for highly structural integrated Co-free LLOs.

Let us revisit the ternary phase diagram in Fig. 1(a). When designing Co-free LLO, we maintain Ni in the +2 oxidation state to maximize energy density. However, this approach increases the possibility of Ni_Li anti-site defect formation, which hinders phase integration. Strategies such as doping²² or post-annealing^18,20 are effective in improving phase integrity; however, they inevitably reduce the available capacity or require additional processing costs. To mitigate these defects while maintaining a Co-free composition with only Ni and Mn, a straightforward design is to create a moderate superexchange interaction, such as forming a Ni²⁺–O–Ni³⁺ configuration.⁴⁵ Here, two feasible compositional approaches to introduce Ni³⁺ involve (i) adjusting the Ni/Mn ratio and (ii) increasing the Li content as shown by the red and navy arrows in the phase diagram (Fig. 3(c)). Slightly increasing the Ni content shifts the transition metal ratio in the R phase towards higher Ni levels, facilitating the formation of Ni³⁺, as indicated by the red arrow in Fig. 3(c). However, this also increases the cationic redox fraction, reducing the available capacity for anionic redox chemistry. Alternatively, increasing the Li content promotes Li₂MnO₃ formation while integrating Ni³⁺ into the R phase. Thus, an effective way to reduce anti-site defects while preserving anionic redox capacity is through off-stoichiometric composition design by adding Li.

Off-stoichiometric design for highly integrated phases

Following the new guideline, we introduced an additional 0.05 mol of Li as a model study to adjust the redox state of Ni in the Co-free NM3565 LLO (hereafter, referred to as H-NM3565, Li/TM = 1.35, Li_1.149Ni_0.298Mn_0.553O₂; ‘H’ indicates homogeneous domain distribution). H-NM3565 was synthesized under identical conditions to NM3565 to ensure a fair evaluation of phase integrity and to form secondary particles of comparable size to NM3565, approximately 3 μm (Fig. S13, ESI†). The introduction of excess Li results in an elevation of the Ni valence state, as evidenced by a higher energy shift in the half-edge of the Ni K-edge compared to the stoichiometric composition (Fig. 4(a)). Additionally, the signal originating from the Ni–O peak at ∼1.6 Å, the 1st neighbor of Ni, showed slight peak broadening due to the Ni³⁺ soft Jahn–Teller effect, as illustrated in Fig. S14(a) (green-shaded area, ESI†). Meanwhile, the half-edge and pre-edge of the Mn K-edge remained unchanged, indicating that the Mn valence state was maintained at +4 (Fig. 4(b)).

Fig. 4 H-NM3565, a highly integrated structure. (a) X-ray absorption near edge spectra (XANES) of the Ni K-edge. (b) XANES spectra of the Mn K-edge. (c) Magnetic susceptibilities of NM3565 and H-NM3565. (d) c-lattice parameter (left side) and TM slab distance (right side) of the R phase and the M phase in NM3565, H-NM3565. (e) HAADF-STEM image of H-NM3565 along the [110]_M direction. (f) Intensity profile and FFT diffraction points of (e). (g) XRD pattern showing the superstructure peak and (104)_R peak of NM3565 and H-NM3565.

Magnetic analysis using superconducting quantum interference device (SQUID) measurements directly illustrates the alleviation of strong antiferromagnetic (AFM) interaction through the introduction of Ni³⁺ (Fig. 4(c)). The SQUID allows for the assessment of the AFM momentum by analyzing the bifurcation between the zero-field-cooled (ZFC) and field-cooled (FC) curves. In general, LLO materials exhibit strong AFM properties attributed to the 90° superexchange of Mn⁴⁺–O–Mn⁴⁺ from the LiMn₆ honeycomb ordering in the M phase.^47,48 Additionally, the anti-site induced Ni²⁺–O–Ni²⁺ configuration in the R phase contributes to the AFM counts, resulting in significant bifurcation between ZFC and FC in NM3565, as shown in Fig. 4(c). However, despite possessing an identical amount of M phase fraction and a Mn valence state of +4, H-NM3565 exhibits notably reduced bifurcation, demonstrating the suppression of the 180° AFM interaction induced by the anti-site-originated Ni²⁺–O–Ni²⁺.

A qualitative and quantitative reduction in anti-site defects within the R phase was observed, as shown by the decreasing intensity around ∼3.9 Å in the Ni K-edge EXAFS (Fig. S14(a), ESI†) and a reduction in anti-site defect concentration from 11.5% to 7.8%, as indicated by Rietveld refinement (Table S6, ESI†). Consequently, the c-lattice parameter difference between the R and M phases decreased from 0.012 Å to 0.002 Å, as depicted on the left side of Fig. 4(d). Similarly, the TM slab distance between the R and M phases decreased from 0.183 Å to 0.0611 Å, as shown on the right side of Fig. 4(d).

Further local analysis using HADDF-STEM with the [110]M zone axis for H-NM3565 (Fig. 4(e)) identifies a blue-shaded M phase domain and an R phase domain. In the intensity profile of the blue box, the left side represents the R phase characterized by TM–TM–TM ordering, while the right side corresponds to the M phase characterized by TM–TM–Li ordering⁴⁹ (Fig. 4(f)). These two phases are randomly mixed into small domains, only a few nanometers in size, within a single particle. The FFT image in Fig. 4(f) shows a streak, indicated by the white arrow, resulting from the high homogeneity of the two domains. This homogeneity leads to short-range ordering in each domain, which in turn stretches the diffraction spots.

Global crystal analysis using sXRD clearly demonstrates the high phase integration between the R and M phases, reflecting new compositional guidance (Fig. S15, ESI†). Upon examining the superstructure peak at the top of Fig. 4(g), it is observed that although H-NM3565 has approximately a 5% larger mole fraction of the M phase due to additional Li content, the superstructure peak appears blurred. This indicates a loss of long-range order in Li₂MnO₃.¹⁸ Additionally, the (104) peak at the bottom merges perfectly, suggesting a well-shared Bragg position between the R and M phases. Furthermore, the Ni K-edge and Mn K-edge EXAFS regions confirm domain homogeneity, as detailed in Fig. S14 (ESI†). This design is not only applicable to NM3565 but also universally relevant to NM2575 and NM4060, showcasing the strategic versatility of structural integrity. Upon adding extra Li to both NM2575 and NM4060, an increase in the oxidation state of Ni was observed, accompanied by a decrease in anti-site defect concentration. Consequently, enhanced phase integrity was confirmed for both compositions (Fig. S16 and S17, ESI†).

Enhanced electrochemical activity with highly integrated domains

As phase integrity between the two domains increases, a significant enhancement in electrochemical activity is observed, particularly during the first activation cycle, where the capacity approaches theoretical limits (Fig. 5(a)). H-NM3565 achieves over 90% of its theoretical charge capacity, compared to 77.3% for NM3565 (Fig. S18, ESI†). This improvement is attributed to the enhanced redox activity of both cationic and anionic redox processes. Notably, H-NM3565 exhibits lower polarization above 4.45 V, a characteristic typically seen in well-dispersed domains. Furthermore, Li-ion kinetics are enhanced across the entire voltage range, with diffusivity increasing by over three orders of magnitude, particularly in the oxygen oxidation region (Fig. S19, ESI†). In addition to the influence of domain structure modification, the improved electrochemical performance of H-NM3565 can be attributed to the enhanced oxygen redox reactivity at the R–M phase interface. Further DFT calculations (Fig. S20, ESI†) confirmed that the R–M phase without anti-site defects exhibits higher oxygen redox activity in the M phase, which contributes to the superior electrochemical activity of H-NM3565.

Fig. 5 Enhanced electrochemical performance of H-NM3565. (a) Galvanostatic profile for the activation cycle of NM3565 and H-NM3565. (b) O K-edge spectra obtained in the TEY mode for pristine and after the activation cycle of NM3565 and H-NM3565. (c) Mn L-edge spectra obtained in the TEY mode for pristine and after the activation cycle of NM3565 and H-NM3565. (d) Cycle life graph of NM3565 and H-NM3565. (e) Normalized discharge profile of NM3565 and H-NM3565. (f) dQ dV⁻¹ profile of NM3565 and H-NM3565 during cycling.

The proposed off-stoichiometric composition further suppresses lattice oxygen evolution. Fig. 5(b) and (c) show the pre-edge of the normalized O K-edge and Mn L-edge spectra in the total electron yield (TEY) mode, respectively, after the first galvanostatic activation cycle. In NM3565, severe phase segregation accelerates oxygen loss during charging due to inter-domain strain, consistent with previous reports.¹⁴ This leads to a substantial decrease in O K pre-edge intensity, reflecting significant surface oxygen loss. As a result, NM3565 shows greater Mn reduction, with increased peak intensities at 639.6 eV and 641.0 eV, corresponding to Mn²⁺ and Mn³⁺, respectively. In contrast, H-NM3565 exhibits only minor reduction in both the O K-edge and Mn L-edge spectra after activation, indicating improved lattice oxygen stability. Suppression of highly soluble Mn²⁺ dissolution in the electrolyte can prevent dendritic growth on the anode in full cells.^50–52 Additionally, the presence of Mn³⁺ in LLOs can lead to reduced structural stability due to Jahn–Teller distortion, negatively impacting retention.³⁰ As previously discussed, Mn migration is associated with oxygen release, contributing to voltage decay.⁹ Thus, careful management of Mn content is essential for preventing these issues.

Fig. 5(d) and (e) demonstrate the cycle stability and normalized discharge voltage of NM3565 and H-NM3565, respectively, after the activation cycle (see Fig. S21 for the long-term cyclability performance under high-loading conditions designed, ESI†). Due to enhanced redox activity, H-NM3565 consistently delivers a significantly higher specific capacity, exceeding 50 mA h g⁻¹, and maintains an average discharge voltage more than 0.1 V higher than NM3565 across all cycles (Fig. S22, ESI†). Although NM3565 may appear to have a better numerical cycle retention than H-NM3565, it suffers from significant oxygen vacancy formation and migration of active transition metals into the Li slab, leading to severe voltage decay and irreversible electrochemical reactions. In contrast, H-NM3565 demonstrates superior electrochemical performance due to its higher reaction activity and more reversible electrochemical behavior. Similarly, off-stoichiometric compositions of Co-free NM2575 and NM4060 show substantial improvements in energy density compared to Co-free NM LLOs (Fig. S23, ESI†).

The dQ dV⁻¹ plots of NM3565 and H-NM3565 during cycling, shown in Fig. 5(f), reveal that H-NM3565 undergoes only a minor voltage shift, suggesting more reversible reactions, while NM3565 shows a significant voltage shift, indicating a higher degree of irreversibility. Additionally, NM3565 exhibits a distinct Mn redox signal in the low-voltage range (Fig. 5(f)). After the activation cycle, substantial surface oxygen release in NM3565 leads to Mn reduction for charge compensation,⁵³ which triggers the Mn redox process (Fig. 5(b) and (c)). Although this process increases capacity with continued cycling, it compromises structural stability. In contrast, off-stoichiometric H-NM3565 maintains better structural integrity and achieves higher energy density due to its more stable redox reactions compared to NM3565.

Conclusions

The unique coexistence of two phases in LLO materials enables the simultaneous utilization of both cationic and anionic redox chemistries, providing an attractive atomic-scale design for achieving unprecedented energy density. However, since the properties of these materials arise from the interplay between these phases, maximizing their degree of mixing is a crucial strategy. Therefore, in materials like LLOs, where subtle domain segregation occurs between phases, fine-tuning the mixing enthalpy becomes an essential engineering consideration. This study reveals that in Co-free LLOs, the conventional composition design, which adheres to Ni²⁺, fundamentally hinders the desired R–M phase domain mixing. This limitation is due to the high concentration of Ni_Li anti-site defects in the R phase, which causes significant expansion of the TM slab distance and increases the slab misfit with Li₂MnO₃. Our calculations show that this structural mismatch increases the mixing enthalpy by at least 12.2 meV f.u.⁻¹, deviating the material from the ideal slab-matched phase. These insights urge a reconsideration of traditional compositional strategies that rigidly adhere to the Ni²⁺ state, highlighting the need to explore methods that lower Gibbs free energy and enhance phase integrity.

We proposed, in a straightforward manner, the compositional design as an atomic-scale engineering solution to reduce anti-site defects, which is linked to strong magnetic interactions, while maximizing oxygen redox utilization by incorporating Ni³⁺. The off-stoichiometric composition was achieved by adding extra lithium to facilitate Ni³⁺ incorporation, effectively reducing Ni_Li site exchange and suppressing the oxygen framework misfit between the R and M phase domains. This design strategy offers a simple approach to address the origins of phase segregation. Consequently, we observed unprecedented phase integrity in Co-free LLOs, with the electrodes exhibiting enhanced oxygen redox activity, reduced lattice oxygen loss, improved cycle retention, and decreased voltage decay across all composition models. Specifically, NM3565, which initially had the most compromised phase integrity, showed significant improvements with a slight addition of Li. The ICE increased from 75.9% to 88.7%, voltage decay reduced from 5.4 mV to 4.3 mV per cycle, and it demonstrated 92.5% capacity retention at C/3 over 50 cycles under harsh conditions (45 °C). Building on this new understanding, this study establishes new criteria for designing homogeneous domains in Co-free LLOs, paving the way for more stable and efficient energy storage solutions. This work not only challenges existing paradigms but also inspires a new direction in cathode material design, emphasizing phase integrity and defect management as essential components of the next generation of high-performance batteries.

Author contributions

Y. L. conceived the idea, designed the overall experiments, conducted all experiments and analyses, and wrote the manuscript. J. K. provided advice on synchrotron-based experiments and the interpretation of the results. J.-H. L. assisted with the electrochemical analysis. E. C. conducted the DFT calculations and analyzed the results. K. E. L. and D. A. provided advice on Rietveld refinement. Y.-T. K. offered guidance on hard XAS analysis. S. P. and C. J. advised on synthesis. Y.-U. H. performed the STEM analysis. I. P. conceived the idea, conducted the DFT calculations, analyzed the results, co-wrote the manuscript, and supervised the research. K.-Y. P. conceived the idea, designed the overall experiments, co-wrote the manuscript, and supervised the entire research.

Data availability

All data supporting the findings of this study are provided in the main article and the ESI.† Further experimental and computational details are available from the corresponding author upon reasonable request.

Conflicts of interest

The authors declare that they have no competing interests.

Acknowledgements

This work was supported by POSCO N.EX.T Hub, as well as by grants from the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (MSIT): RS-2024-00408156 (50%) and 00261543 (25%). Additional support (25%) was provided by the Korea Institute for Advancement of Technology (KIAT) through a grant funded by the Ministry of Trade, Industry and Energy (MOTIE) under the HRD Program for Industrial Innovation (RS-2024-00419413).

References

1. G. Assat and J.-M. Tarascon, Nat. Energy, 2018, 3, 373–386.
2. W. Zuo, M. Luo, X. Liu, J. Wu, H. Liu, J. Li, M. Winter, R. Fu, W. Yang and Y. Yang, Energy Environ. Sci., 2020, 13, 4450–4497.
3. P. Rozier and J. M. Tarascon, J. Electrochem. Soc., 2015, 162, A2490.
4. H. Zhao, W. Y. A. Lam, L. Sheng, L. Wang, P. Bai, Y. Yang, D. Ren, H. Xu and X. He, Adv. Energy Mater., 2022, 12, 2103894.
5. K. Luo, M. R. Roberts, R. Hao, N. Guerrini, D. M. Pickup, Y.-S. Liu, K. Edström, J. Guo, A. V. Chadwick and L. C. Duda, Nat. Chem., 2016, 8, 684–691.
6. M. Sathiya, G. Rousse, K. Ramesha, C. Laisa, H. Vezin, M. T. Sougrati, M.-L. Doublet, D. Foix, D. Gonbeau and W. Walker, Nat. Mater., 2013, 12, 827–835.
7. D.-H. Seo, J. Lee, A. Urban, R. Malik, S. Kang and G. Ceder, Nat. Chem., 2016, 8, 692–697.
8. Y. Song, X. Zhao, C. Wang, H. Bi, J. Zhang, S. Li, M. Wang and R. Che, J. Mater. Chem. A, 2017, 5, 11214–11223.
9. R. A. House, G. J. Rees, M. A. Pérez-Osorio, J.-J. Marie, E. Boivin, A. W. Robertson, A. Nag, M. Garcia-Fernandez, K.-J. Zhou and P. G. Bruce, Nat. Energy, 2020, 5, 777–785.
10. M. Okubo and A. Yamada, ACS Appl. Mater. Interfaces, 2017, 9, 36463–36472.
11. J.-J. Marie, R. A. House, G. J. Rees, A. W. Robertson, M. Jenkins, J. Chen, S. Agrestini, M. Garcia-Fernandez, K.-J. Zhou and P. G. Bruce, Nat. Mater., 2024, 1–8.
12. A. Boulineau, L. Simonin, J.-F. Colin, C. Bourbon and S. Patoux, Nano Lett., 2013, 13, 3857–3863.
13. L. Fang, D. Han, S. Kang, U.-S. Heo, K.-W. Nam and Y.-M. Kang, Energy Environ. Sci., 2023, 16, 3053–3062.
14. T. Liu, J. Liu, L. Li, L. Yu, J. Diao, T. Zhou, S. Li, A. Dai, W. Zhao and S. Xu, Nature, 2022, 606, 305–312.
15. H. Dixit, W. Zhou, J.-C. Idrobo, J. Nanda and V. R. Cooper, ACS Nano, 2014, 8, 12710–12716.
16. H.-X. Wei, L.-B. Tang, Z.-Y. Wang, Y.-H. Luo, Z.-J. He, C. Yan, J. Mao, K.-H. Dai and J.-C. Zheng, Mater. Today, 2021, 51, 365–392.
17. C. Xu, K. Märker, J. Lee, A. Mahadevegowda, P. J. Reeves, S. J. Day, M. F. Groh, S. P. Emge, C. Ducati and B. Layla Mehdi, Nat. Mater., 2021, 20, 84–92.
18. J. Hwang, S. Myeong, W. Jin, H. Jang, G. Nam, M. Yoon, S. H. Kim, S. H. Joo, S. K. Kwak and M. G. Kim, Adv. Mater., 2020, 32, 2001944.
19. H. Zhuo, H. Peng, B. Xiao, Z. Wang, X. Liu, Z. Li, G. Li, X. Bai, L. Wang and X. Huang, Adv. Energy Mater., 2023, 13, 2203354.
20. J. Hwang, S. Myeong, E. Lee, H. Jang, M. Yoon, H. Cha, J. Sung, M. G. Kim, D. H. Seo and J. Cho, Adv. Mater., 2021, 33, 2100352.
21. Z. Wei, Z. Shi, X. Wen, X. Li, B. Qiu, Q. Gu, J. Sun, Y. Han, H. Luo and H. Guo, Mater. Today Energy, 2022, 27, 101039.
22. Z. Li, Y. Li, M. Zhang, Z. W. Yin, L. Yin, S. Xu, C. Zuo, R. Qi, H. Xue and J. Hu, Adv. Energy Mater., 2021, 11, 2101962.
23. C. Yin, Z. Wei, M. Zhang, B. Qiu, Y. Zhou, Y. Xiao, D. Zhou, L. Yun, C. Li and Q. Gu, Mater. Today, 2021, 51, 15–26.
24. M. Gu, I. Belharouak, A. Genc, Z. Wang, D. Wang, K. Amine, F. Gao, G. Zhou, S. Thevuthasan and D. R. Baer, Nano Lett., 2012, 12, 5186–5191.
25. A. Devaraj, M. Gu, R. Colby, P. Yan, C. M. Wang, J. Zheng, J. Xiao, A. Genc, J. Zhang and I. Belharouak, Nat. Commun., 2015, 6, 8014.
26. M. Gu, A. Genc, I. Belharouak, D. Wang, K. Amine, S. Thevuthasan, D. R. Baer, J.-G. Zhang, N. D. Browning and J. Liu, Chem. Mater., 2013, 25, 2319–2326.
27. J. C. Knight and A. Manthiram, J. Mater. Chem. A, 2015, 3, 22199–22207.
28. C. Zhang, P. Hou, X. Shi, D. Song, J. Song and L. Zhang, RSC Adv., 2015, 5, 36015–36021.
29. S. Chong, Y. Liu, W. Yan and Y. Chen, RSC Adv., 2016, 6, 53662–53668.
30. H. Kim, G. Yoon, I. Park, K.-Y. Park, B. Lee, J. Kim, Y.-U. Park, S.-K. Jung, H.-D. Lim and D. Ahn, Energy Environ. Sci., 2015, 8, 3325–3335.
31. J. Zhang, Q. Zhang, D. Wong, N. Zhang, G. Ren, L. Gu, C. Schulz, L. He, Y. Yu and X. Liu, Nat. Commun., 2021, 12, 3071.
32. D. Ye, K. Ozawa, B. Wang, D. Hulicova-Jurcakova, J. Zou, C. Sun and L. Wang, Nano Energy, 2014, 6, 92–102.
33. M. G. Kim and C. H. Yo, J. Phys. Chem. B, 1999, 103, 6457–6465.
34. J. Zhao, W. Zhang, A. Huq, S. T. Misture, B. Zhang, S. Guo, L. Wu, Y. Zhu, Z. Chen and K. Amine, Adv. Energy Mater., 2017, 7, 1601266.
35. Y. Duan, L. Yang, M.-J. Zhang, Z. Chen, J. Bai, K. Amine, F. Pan and F. Wang, J. Mater. Chem. A, 2019, 7, 513–519.
36. D. Rathore, C. Liang, E. Zsoldos, M. Ball, S. Yu, C. Yang, Q. Wang and J. Dahn, J. Electrochem. Soc., 2023, 170, 020520.
37. X. Hou, K. Ohta, Y. Kimura, Y. Tamenori, K. Tsuruta, K. Amezawa and T. Nakamura, Adv. Energy Mater., 2021, 11, 2101005.
38. K. Kang and G. Ceder, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 74, 094105.
39. W. Lee, S. Muhammad, T. Kim, H. Kim, E. Lee, M. Jeong, S. Son, J. H. Ryou and W. S. Yoon, Adv. Energy Mater., 2018, 8, 1701788.
40. X. Yan, X. Zhou, C. Zhu, W. Huang and Y.-J. Zhao, J. Mater. Chem. A, 2024, 12, 3722–3733.
41. J. Jiao, G. Lai, S. Qin, C. Fang, X. Xu, Y. Jiang, C. Ouyang and J. Zheng, Acta Mater., 2022, 238, 118229.
42. A. Boulineau, L. Croguennec, C. Delmas and F. Weill, Solid State Ionics, 2010, 180, 1652–1659.
43. Z. Wang, Y. Huang, C. Liu, J. Li and J. Wang, Intermetallics, 2019, 109, 139–144.
44. R. Dingreville, A. Hallil and S. Berbenni, J. Mech. Phys. Solids, 2014, 72, 40–60.
45. J. Zheng, G. Teng, C. Xin, Z. Zhuo, J. Liu, Q. Li, Z. Hu, M. Xu, S. Yan and W. Yang, J. Phys. Chem. Lett., 2017, 8, 5537–5542.
46. Y. Xiao, T. Liu, J. Liu, L. He, J. Chen, J. Zhang, P. Luo, H. Lu, R. Wang and W. Zhu, Nano Energy, 2018, 49, 77–85.
47. D. Mohanty, A. Huq, E. A. Payzant, A. S. Sefat, J. Li, D. P. Abraham, D. L. Wood III and C. Daniel, Chem. Mater., 2013, 25, 4064–4070.
48. J. Lee, N. Dupre, M. Jeong, S. Kang, M. Avdeev, Y. Gong, L. Gu, W. S. Yoon and B. Kang, Adv. Sci., 2020, 7, 2001658.
49. B. Qiu, M. Zhang, L. Wu, J. Wang, Y. Xia, D. Qian, H. Liu, S. Hy, Y. Chen and K. An, Nat. Commun., 2016, 7, 12108.
50. H. Y. Asl and A. Manthiram, Science, 2020, 369, 140–141.
51. Z. Rao, Z. Yang, W. Gong, S. Su, Q. Fu and Y. Huang, J. Mater. Chem. A, 2020, 8, 3859–3864.
52. W. Li, J. Electrochem. Soc., 2020, 167, 090514.
53. M. Oishi, K. Yamanaka, I. Watanabe, K. Shimoda, T. Matsunaga, H. Arai, Y. Ukyo, Y. Uchimoto, Z. Ogumi and T. Ohta, J. Mater. Chem. A, 2016, 4, 9293–9302.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5mh00072f

---