Summary of activated LR-NMC’s charge-compensation mechanism and electrochemical kinetics. Key curves from voltage window opening experiments in Fig. 4 are superimposed on each other, keeping the same colors as before. Electrochemical kinetics is classified as either slow or fast based on impedance analysis in Fig. 5. The small Mn3+/4+ contribution is restricted to low potentials (shaded blue) on charge with some anionic activity also. Further charge to 4.1?V (shaded gray) leads to the peak at 3.8?V mainly from cationic oxidation (Ni2+/3+/4+ and Co3+/4+) along with some anionic contribution. If charging is limited to 4.1?V, the discharge curve (shaded gray) shows two peaks, respectively, due to cationic (3.8?V peak) and anionic (3.2?V peak) reductions. If charging is continued to 4.8?V (shaded red and green), it is mainly charge compensated by anions and then the corresponding discharge capacity is split at high potential (shaded green) and low potential (3.2?V reduction peak, shaded red), thus causing hysteresis. Fast kinetics accompanies cationic redox on either charge or discharge
Rechargeable batteries are enabling the widespread adoption of electrified transportation and renewable energy. System-level predictions reveal that Li-ion batteries powered by Li-rich layered-oxide cathodes, e.g., Li1.2Ni0.13Mn0.54Co0.13O2 (LR-NMC), hold the highest promise regarding practical energy density and cost. LR-NMC can deliver capacities reaching 300?mAh?g-1, a value under-explained if solely the transition metals (TMs) participate in redox processes. Recent advances have steered a consensus that such high capacities arise from the reversible redox of O2– anions. Anionic redox has thereby emerged as a transformational approach for designing new high-energy cathodes, several of which have been discovered lately with diverse crystal chemistries.
As battery researchers are moving into this new direction, it is necessary to revisit the Li-rich systems with a fresh perspective by including anionic redox in order to understand the fundamental origins of some practical roadblocks (i.e., voltage fade, poor kinetics, and voltage hysteresis). These issues jeopardize cycle life, power rate, energy efficiency, and state of charge (SoC) management, and hence have plagued the commercialization of LR-NMCs despite several years of academic and industrial interest.8,9 Recently, we pointed out the poor electrochemical kinetics of anionic redox and its detrimental role in triggering voltage fade and hysteresis using a “model” Li-rich layered-oxide, Li2Ru0.75Sn0.25O3.10 In light of these findings, it is imperative to ask whether the same is true for LR-NMC since it is an archetypical Li-rich system of substantial real-world potential.
This task is not straightforward because LR-NMC is complicated by the redox activity of Ni, Mn, Co, and O, whose redox potentials must first be identified to establish the charge-compensation mechanism. Starting with oxygen, it was believed for quite a few years that the anomalous first charge capacity is compensated by irreversible oxygen loss from the electrode’s surface resulting in densification. However, recent measurements suggest that such oxygen loss is largely insufficient to account for the extra first charge capacity. Thus, the early models of surface oxygen loss were only partially complete and it is now believed that most of the oxidized oxygen remains in the lattice and participates in redox, as directly observed via ex situ X-ray absorption spectroscopy (XAS), X-ray photoelectron spectroscopy (XPS), and hard X-ray photoelectron spectroscopy (HAXPES). The remaining task now is to look beyond the first charge and identify the electrochemical potentials with corresponding states of charge (SoC) where anionic redox occurs.
As for charge compensation from the TMs, operando XAS as well as ex situ XAS at TM K-edges have clearly shown how the Ni2+/3+/4+ redox process proceeds but there are still discrepancies concerning the extent of Mn3+/4+ contribution because Mn K-edge interpretation is not unambiguous. Finally for Co, the XAS data in literature is insufficient to precisely conclude the redox potential of Co3+/4+, which is often wrongly assumed when interpreting the differential capacity (dQ/dV) plots. We note that soft-XAS at Mn and Co L-edges can provide a clearer signature of oxidation states, but these data are scarce in literature.
Filling the above-mentioned charge-compensation knowledge gaps is necessary to explore the role of anionic redox on battery performance issues, i.e., voltage hysteresis, poor kinetics, and voltage fade. So far, the Mn-rich nature of LR-NMC leaning toward Li2MnO3 was blamed for these issues. A phenomenological model involving irreversible and reversible TM migrations was proposed to explain voltage fade and hysteresis, respectively.31 However, it cannot be ruled out that such migrations are in fact a consequence of anionic redox. Concerning kinetics, there is XAS-based evidence for sluggish reaction of the Li2MnO3-type component25 and also evidences of a mysterious impedance rise at low SoCs that encourage us to examine whether this is actually because of anionic redox.
Toward these goals, we report here the complete charge-compensation mechanism in LR-NMC as deduced by bulk-sensitive synchrotron-based spectroscopic techniques, namely HAXPES and soft-XAS. The ability of HAXPES to systematically increase probe depths35 allowed us to distinguish between surface and bulk effects. We consequently correlate these findings with detailed electrochemical measurements to reveal the interplay between anionic/cationic redox and electrochemical kinetics/thermodynamics of these electrodes. We provide direct evidence of bulk anionic redox activity, which compensates for a substantial capacity not just at high, but quite surprisingly at low potentials also. Moreover, we confirm, as previously reported for Li2Ru0.75Sn0.25O3, that the cationic redox from Ni2+/3+/4+ and Co3+/4+ exhibits fast kinetics and diffusion in comparison to the sluggish anions. On the other hand, Mn is found to be essentially redox-inactive with only a small contribution from Mn3+/4+ at the lowest potentials. To effectively convey these findings, the results will be structured in two parts, dealing first with the charge-compensation mechanism followed by a detailed electrochemical investigation.
Via detailed spectroscopic and electrochemical analyses of LR-NMC cathodes, we have revealed (i) their charge-compensation mechanism from anionic/cationic redox, and (ii) how the interplay between these two processes governs application-wise important challenges, such as kinetics, hysteresis, and voltage fade. Next, we connect our results with published knowledge in this area.
The two-stepped first charge voltage profile, typical for Li-rich cathodes, starts with a classical cationic redox step up to ~4.4?V, which is fully reversible26,28,51,59,60, followed by the 4.5?V activation plateau where lattice O2– oxidizes to On– (n?2). We have unambiguously spotted On– in the bulk thanks to the high probe depth of HAXPES (~29?nm, 120 layers). HAXPES characterization of anionic redox is repeatable, quantitative, and tunable from surface to bulk, which makes this technique indispensable for studying charge-compensation mechanisms, especially if operando mode is developed.35
A simple charge balance suggests that n?=?1.24?±?0.07 in the charged sample for 33?±?3% On– to compensate for the capacity of 0.5?e– per formula unit (157?mAh?g-1) that was proposed by Luo et al.4 This value of n is clearly insufficient to make true peroxide (O2)2– and the physio-chemical nature of On– should be investigated on a fundamental basis. On further cycling, HAXPES demonstrated the reversibility of anionic reactivity spread in a wide voltage range from 2.0 to 4.8?V. Although such reactivity is beneficial for capacity enhancement, the preceding anionic activation plateau unavoidably causes (i) some irreversible oxygen loss4,15,16 and (ii) permanently modifies the electrochemical profile, likely due to oxygen network distortion and rearrangement that is challenging to experimentally visualize. Therefore, we recommend designing cathodes showing a complete reversibility and stability of anionic redox from the very beginning—a task that is theoretically difficult with 3d TM-layered oxides.
LR-NMC’s S-shaped-sloped electrochemistry after first cycle activation consists of several dQ/dV peaks, which we have now assigned properly by combining spectroscopic and electrochemical evidences, as summarized in Fig. 7. Cationic redox occurs at nearly the same potentials during either charge or discharge, such that the middle dQ/dV peak around 3.8?V shows Ni2+/3+/4+ and Co3+/4+ activity and the lowest potentials have a minor Mn3+/4+ capacity. In contrast, anionic redox proceeds asymmetrically between charge and discharge. On oxidation, anionic redox is spread from 2.0 to 4.8?V with a large contribution above 4.1?V. While on discharge, anionic reduction at high potential is small and the remaining reduction occurs at much lower potentials (peak below 3.6?V). This resolves several charge-compensation-related discrepancies in literature, some of which were called for by recent investigations61,62 and reviews8,63. For example, the first dQ/dV discharge peak (4.8–4.1?V) is not from Co reduction as many studies assumed, but rather due to anionic redox. Moreover, this peak is also exhibited by Co-free Li-rich compositions34,52,60,62,64, which further confirms our assignment. Another example is the dQ/dV reduction peak below 3.6?V, which is mainly charge compensated by anionic redox and not by Mn3+/4+, whose contribution is very small. Nevertheless, follow-up experiments are underway to quantify the evolution of Mn3+/4+ contribution over long cycling.
After the first cycle, a large quasi-static hysteresis of ~500?mV is seen even at C/30053, which negatively affects energy efficiency and complicates SoC management. It is now clear, as stated above, that the anionic/cationic redox sequence differs on charge vs. discharge, resulting in path dependence and hysteresis. Such hysteresis mechanism contrasts from LiFePO4 (non-monotonic equilibrium potential65) or conversion electrodes66. Consistent with the anionic/cationic sequence hypothesis, our Ni 2p3/2 HAXPES results show a hysteresis loop in Ni oxidation state vs. capacity (Fig. 3b). However, we could not observe a complementary hysteresis loop for % On– (Fig. 2f) in a sense opposite to that of Ni, as one would expect for charge balancing. Besides, it also remains unanswered why anionic redox occurs in a wide potential range but differently on charge vs. discharge. One possible explanation could be nested in changes of n with SoC, implying different types of On– that can only be unambiguously distinguished with very sharp energy resolutions, thus calling for technical improvements on HAXPES. Previous works suggested a structural origin for hysteresis, i.e., a back-and-forth cationic migration resulting in different thermodynamic potentials during Li removal vs. insertion.31,53,67,68 Although this hypothesis is valid, it must be recalled that structural distortions and rearrangements are in fact a consequence of instability induced by anionic oxidation. Nevertheless, the hysteresis mechanism poses a fundamental challenge which could greatly benefit from operando investigations as well as phenomenological models.
Apart from hysteresis, poor rate-capability has been another barrier in the success of LR-NMC. Our results highlight the contrasting electrochemical kinetics of cationic and anionic redox, wherein the latter displays sluggish charge-transfer and diffusion; a situation alike the one previously reported for the “model” Li2Ru0.75Sn0.25O3 system, where anionic/cationic activities were unambiguously decoupled.10,46 Moreover and worth mentioning is LR-NMC’s large impedance growth at high and low SoCs that is disadvantageous, respectively, for fast charging and discharging.32,33,34,69 To counter this, one can hypothesize, besides electrode blends, a new high-capacity cathode in which cationic and anionic redox occur at the same potential, in order to benefit from fast kinetics of the former as well as extra capacity from the latter. One such known model material is currently being investigated.
Lastly for voltage fade, we have systematically revealed the detrimental role of time spent with highly oxidized oxygen. Thus, the stability of On– is very important for preventing permanent reorganizations, such as irreversible cationic migrations, oxygen loss by recombination or nucleophilic attack, and oxygen lattice modification. Low potential anionic redox is exciting by being more robust against voltage fade70, and hence could be a new focus for materials design.
In summary, our detailed spectroscopic and electrochemical investigations of LR-NMC have attempted to answer long-standing questions about this practically valuable system. First, bulk reactivity of reversible anionic redox is now confirmed thanks to HAXPES’s deep probe. Moreover, quantitative tracking of anionic/cationic charge-compensation mechanism in LR-NMC has allowed us to fully understand its dQ/dV curve. Application-wise, anionic redox boosts capacity but unfortunately, it is also associated with hysteresis, poor kinetics, and voltage fade. Quite remarkably, electrochemical investigation of a model Li-rich system, Li2Ru0.75Sn0.25O3, had warned about these side-effects10, thus emphasizing the added value of model systems in not just revealing fundamental insights, but also inspiring mitigation strategies. Our future direction would therefore be to master holistically the underlying thermodynamics and kinetics of anionic redox by bridging the learning between model and practical systems. Such an approach is essential for taking high-capacity anionic redox cathodes beyond the labs and into the market.