Moritz
Hirsbrunner
a,
Anastasiia
Mikheenkova
b,
Pontus
Törnblom
a,
Robert A.
House
c,
Wenliang
Zhang
d,
Teguh C.
Asmara
d,
Yuan
Wei
d,
Thorsten
Schmitt
d,
Håkan
Rensmo
a,
Soham
Mukherjee
a,
Maria
Hahlin
ab and
Laurent C.
Duda
*a
aCondensed Matter Physics of Energy Materials, Division of X-ray Photon Science, Department of Physics and Astronomy, Uppsala University, Box 516, 751 20 Uppsala, Sweden. E-mail: laurent.duda@physics.uu.se
bDepartment of Chemistry – Ångström Laboratory, Uppsala University, Box 523, 751 20 Uppsala, Sweden
cDepartment of Materials, University of Oxford, Oxford, OX1 3PH, UK
dSwiss Light Source, Photon Science Division, Paul Scherrer Institut, CH-5232, Villigen PSI, Switzerland
First published on 2nd July 2024
Vibrationally-resolved resonant inelastic X-ray scattering (VR-RIXS) at the O K-edge is emerging as a powerful tool for identifying embedded molecules in lithium-ion battery cathodes. Here, we investigate two known oxygen redox-active cathode materials: the commercial LixNi0.90Co0.05Al0.05O2 (NCA) used in electric vehicles and the high-capacity cathode material Li1.2Ni0.13Co0.13Mn0.54O2 (LRNMC) for next-generation Li-ion batteries. We report the detection of a novel vibrational RIXS signature for Li-ion battery cathodes appearing in the O K pre-peak above 533 eV that we attribute to OH-groups. We discuss likely locations and pathways for OH-group formation and accumulation throughout the active cathode material. Initial-cycle behaviour for LRNMC shows that OH-signal strength correlates with the cathodes state of charge, though reversibility is incomplete. The OH-group RIXS signal strength in long-term cycled NCA is retained. Thus, VR-RIXS offers a path for gaining new insights to oxygen reactions in battery materials.
Using vibrationally-resolved resonant inelastic X-ray scattering (VR-RIXS), House et al.5 concluded that oxygen redox cathodes effectively convert some of their lattice oxygen into molecular oxygen during the delithiation process. Interestingly, these oxygen molecules are not expelled from the crystal structure; instead, they become trapped within the crystals and appear to be reversibly integrated into the lattice upon cathode re-lithiation.5,6 The discovery is particularly intriguing as it strongly suggests that oxygen redox results in the generation of embedded species that are extrinsic to the oxide sublattice. This was not foreshadowed by Raman spectroscopy studies or photoelectron spectroscopy, instead, identification of these embedded oxygen molecules was enabled by the application of VR-RIXS,5 whereas X-ray absorption spectroscopy (XAS) alone is insufficient to unambiguously detect their presence due to the dominating signal of lattice oxygen. Remarkably, the signatures of these trapped oxygen molecules resemble gas-phase oxygen K-RIXS spectra,5 raising questions about their location, interactions with lattice atoms, and their stability over time.6 Although we will not go into further detail regarding this topic, we note that it has been shown that similar RIXS signatures may arise in certain non-Li-excess cathodes even in the absence of other evidence for oxygen redox.7 However, this observation only adds to the urgency for developing a picture of embedded molecules in Li ion cathodes, in general.
Another direction of inquiry concerns the formation of other oxygen species during synthesis or cycling and their identification using VR-RIXS. For instance, carbonates and intercalated water may typically be formed in battery cathodes, with identifiable signatures at energies above the X-ray absorption pre-peak region.8 Debates also persist about the formation and the impact of hydroxides on battery operation.9 These compounds are generated and evolve during battery cycles, especially in Ni-rich materials exposed to air or moisture. The mechanisms behind long-term cycling degradation and a comprehensive understanding of chemistry within Li-ion battery cathodes remain incomplete but are essential for addressing efficiency bottlenecks.
When several oxygen species are present in a material, disentangling their main band RIXS spectra can become very difficult. On the other hand, their vibrational signatures are often distinguishable and could therefore bring more clarity. High-resolution RIXS stands out as an exceptional tool for gaining detailed insights into the electronic structure changes within materials. Importantly, unlike in other core level spectroscopy techniques, RIXS spectra are not restricted by the core hole's lifetime. As such, it has been successfully employed to discern vibrational contributions in both molecules and condensed materials, which are typically investigated using Raman or IR-spectroscopy.10–13 Thus, when coupled with high-resolution synchrotron radiation, RIXS allows for the resolution of transitions originating from individual molecular vibrations. Though this capability is akin to optical Raman spectroscopy, VR-RIXS has the added advantage of elemental specificity, particularly for distinguishing contributions from oxygen and transition metal sites. Therefore, VR-RIXS is a powerful tool for advancing the understanding of oxygen redox processes in Li-ion battery cathodes.
In this study, we present a novel observation in O K-RIXS from hydroxide compounds forming in the oxygen redox active materials Li1.2Ni0.13Co0.13Mn0.54O2, a Li-rich, Mn-based NMC compound, and LixNi0.90Co0.05Al0.05O2, a Ni-rich NCA active material from commercial electrodes. LRNMC has a high initial capacity but suffers from capacity loss due to structural reformation after the initial cycle while NCA is structurally more stable making it already commercially viable. Considering their differences in composition and performance, interestingly, both display O2- and OH-formation.
Employing VR-RIXS, we unveil, for the first time, the vibrational progressions of these hydroxide groups in battery cathode materials. Our investigation reveals the occurrence of various vibrational modes of OH-groups, whose intensities are sensitive to the type of material and the state of charge. We discuss whether the observed OH-vibrations originates from compounds in the cathode electrolyte interphase (CEI) or from the bulk of the cathode particles. Suggestions on the formation pathways of the OH-groups are also presented. Furthermore, we track the evolution of these OH-groups during the initial cycle and observe changes in OH concentration during the aging process. In a broader context, this study underscores the chemical specificity of RIXS, which is achieved through energy-selective excitation within the O K-edge, resulting in distinct VR-RIXS signatures from various oxygen sites within the examined material.
The present study also investigates LixNi0.90Co0.05Al0.05O2 (NCA) in addition to the LRNMC. Note that the preparation of the NCA samples of the present study has already been described in a recent publication.4 For convenience, we also present relevant details in the following. The sample preparation of the NCA electrodes for the current study consisted of: (i) opening the full cylindrical cell and extracting the electrodes; (ii) reassembling the extracted electrodes into half-cells; (iii) charging the half-cells until reaching the desired state of charge (SOC) for further RIXS investigation. The full cells were extracted from commercial 21700 cylindrical cells originating from a Tesla Model 3 2018 battery pack. Cells extracted from the pack were considered beginning of life. However, such cells have been pre-cycled at the factory. Two additional cells were aged at 45 °C by cycling between 0–50% SOC (resulting samples are called ‘aged 0–50’) and 0–100% SOC (named ‘aged 0–100’). Here, 0% SOC constitutes a voltage of 2.55 V, 50% SOC corresponds to 3.6 V and 100% SOC corresponds to 4.2 V. The charge and discharge curves for each sample are presented in Fig. S4 (ESI†). The cells achieved 950 (aged 0–50) and 1050 (aged 0–100) full equivalent cycles until reaching ∼20% capacity loss, which was considered the end of life for the aged cells. After, the cells were disconnected and discharged for further disassembly in the Ar-filled glove box. The central rectangular region of the cathode material was selected as the region of interest for the electrode extraction. After taking the central regions from all cells, the electrodes were mechanically cleaned from one side of the active material and punched with a diameter of 8 mm. The resulting electrodes were reassembled in pouch cells using pouch material (Skilstuna Flexible, pre-dried at 60 °C in an oven) with Li foil (thickness 450 μm, used as received) as a counter electrode, Celgard 2325 (Celgard, cleaned with ethanol, deionized water and dried in Buchi oven at 60 °C for 12 hours) as a separator, and LP40 (1 mol L−1 lithium hexafluorophosphate) dissolved in 1
:
1 v/v Ethylene carbonate: Diethyl carbonate (LP40, 1 M LiPF6 in 1
:
1 EC:DEC, Solvionic, used as received) was used as electrolyte. Pouch cells were cycled until the required potential using Arbin and Biologic cycling equipment. The assembled half-cells were first put at rest for 6 h, then discharged to 2.55 V and then cycled with a constant current of 50 μA until reaching the required SOC. After that, a constant voltage was applied until the current was lower than C/50. After, the cells were disassembled in an Ar-filled glovebox. A 2 mm strip was cut from every electrode and was placed on a sample holder for further RIXS studies.
Now, we turn to the features highlighted by the orange dashed boxes at incident photon energies above ∼533 eV. Two peaks in this region can easily be identified by visual inspection of the RIXS maps. The lowest energy loss peak appears at an energy loss of ∼0.4 eV and a second, weaker peak at ∼0.8 eV. These features are representative of a vibrational progression which can be assigned to vibrational modes of an OH-group (described in detail further below). Note that this OH-signal is absent in the pristine LRNMC, indicating that the OH is formed during charge.
High-resolution O K-edge RIXS maps of commercial automotive grade NCA battery cathodes have also been obtained and can be seen in Fig. 2. This sample series includes samples in both lithiated (discharged) and delithiated (charged) states from fresh, aged 0–50, and aged 0–100 NCA. Similarly to the LRNMC, the RIXS maps of the delithiated NCA in Fig. 2 display the O2-signal (white dashed boxes) which has been discussed by Mikheenkova et al.4
In the delithiated LRNMC and NCA samples, the intensity is an order of magnitude weaker than the corresponding O2-signal. We find that both the O2 and the OH-group signal are more intense (when at their strongest) in LRNMC compared to NCA when normalizing to the main band features at the respective excitation energies (Fig. S1, ESI†). Interestingly, the RIXS maps of NCA show a consistent absorption energy shift of the OH signal depending on the state of charge and we can infer the same tendency from the LRNMC RIXS maps, see Fig. S3 (ESI†). Both display a shift towards (higher) lower energies when (de)lithiated, indicating that the chemical environment of the OH-group changes with the state of charge.
We analyze the OH-signal from the RIXS maps further by creating 1D energy loss spectra, shown in Fig. 3a. The shown integrated spectra help clarify the observed peak structure whose energy positions can be more accurately determined. The displayed data is derived from the RIXS maps in Fig. 1 and 2 by integrating RIXS spectra along the incident photon energy axis in the range between 533.4 eV and 535.6 eV, combining them into a single spectrum. From this, we achieve a clear picture of the vibrational features for each sample. To enhance clarity, a binomial smoothing of 14 channels (63 meV) was applied to the spectra, which does not significantly reduce the experimentally attained resolution. The corresponding raw data can be found in the ESI† in Fig. S2.
![]() | ||
Fig. 3 O K-edge RIXS spectra in the region of OH vibrations are shown. (a) Initial cycle LRNMC samples: pristine (lithiated), delithiated, beginning of plateau (BoP), end of plateau (EoP) and (re-)lithiated. On the bottom, fresh, aged 0–50 and aged 0–100 NCA samples in both delithiated and lithiated states are shown. The displayed data is derived by integrating over the RIXS spectra from 533.4 eV to 535.6 eV incident photon energy. A binomial smoothing over 14 channels (63 meV) was applied to all spectra. (b) Reference O K-edge RIXS vibrational signals are shown for oxygen gas, liquid water and Kaolinite (Al2Si2O5(OH)4), adapted from Schreck et al., 2016,15 as well as LiOH (powder sample, this work). |
For the LRNMC samples, a significantly smaller (larger) peak intensity of lithiated (delithiated) material is evident from the data presented in Fig. 3a, suggesting partially reversible behaviour. To a lesser degree this is also the case for the cycle of “fresh” NCA cathodes. We recall that “fresh” does not correspond to the very first cycle, so that the presence of a signal in lithiated NCA is consistent with the incomplete reversibility observed for cycled LRNMC. For the aged NCA samples, the OH-peak intensity stays at a high level regardless of lithiation, suggesting some degree of accumulation of OH during aging.
In the integrated spectra, a third peak is evident as well as an intricate substructure. For instance, close inspection of the peak at ∼430 meV reveals that the position noticeably shifts towards a lower energy loss position from the lithiated to the delithiated samples in both LRNMC and NCA. Additionally, for delithiated LRNMC, the 430 meV peak appears to be composed of multiple peaks. Providing an accurate identification of these features is outside of the scope of the present work and future improved resolution will certainly reveal more details. Here, we focus on the peaks belonging to two vibrational progressions. Firstly, the strongest progression is composed of a series of peaks at about 430 meV, 835 meV and 1222 meV (denoted as mode A #1, #2, #3 in the bottom half of Fig. 3a). We observe the ΔE between peaks decreasing for consecutive overtones from 430 meV to 405 meV and then to 387 meV. Furthermore, two peaks at 225 meV and 656 meV can be seen which belong to #1 and #3 of mode B, another vibrational progression which is most apparent in LRNMC. Peak #2 of this progression coincides with peak #1 of mode A, making it imperceptible. Below we discuss the origin of these progressions.
Vibrational excitations of stretching modes of OH groups have previously been observed between 3100 cm−1 (ΔE = 385 meV) and 3800 cm−1 (ΔE = 471 meV) using both experimental and theoretical techniques.11,12,16–22 The lower of these energy excitations have been related to more strongly-bound OH groups, while the largest ΔE can be assigned to free OH− vibrations.16,20,23 Comparing this to the observed ΔE of mode A of 430 meV (3470 cm−1) suggests an assignment of these features to stretching mode vibrations of a bound OH-group. Strong vibrational progressions in O K-RIXS of liquid15,19 and gas phase water24 as well as from other compounds15 have been reported earlier. In Fig. 3b we show examples of compounds with similar signatures of previously reported O K-RIXS spectra from liquid water, kaolinite,15 and LiOH powder (this work). Additionally, the dashed lines indicate the energies taken from the vibrational progression mode A in Fig. 3a. We find that the observed main vibrational progressions of the references are similar with a somewhat larger spacing between the energy levels, as well as an intense low-energy structure that is markedly distinct from the OH-groups that we detect. The slight differences of the observed OH-signal in the cathode materials with the above reference compounds could be the result of a mixture of such OH-compounds, or it could mean that the vibrational frequency of the present OH-compound is altered because of its chemical environment. Note that, for free water molecules, both symmetric and asymmetric stretching vibrations would lie above 446 meV (3600 cm−1), which would suggest that the OH-groups are either bound to the structure directly in the form of a TM-OH, or there are water molecules which are restricted inside of the material structure. For reference, water can exhibit lower energy stretching vibrations in the range of 430 meV when hydrogen-bonded in complexes of four hydrogen molecules or more.25
In LRNMC, additional peaks at 225 meV and 660 meV belonging to a separate vibrational progression are apparent, while the NCA cathodes show weak hints of peaks at these energy losses (mode B in Fig. 3a). Tentatively, this could be attributed to bending modes (see gas phase water24) or other stretching modes26 (see also LiOH in Fig. 3b). On the other hand, bending mode energies of OH groups bound to transition metals are on the order of ∼50–80 meV (∼400–650 cm−1).22 Thus, the observation of different OH-peak mode intensities in LRNMC and NCA suggests that the OH-group surroundings are at least distinct in the two compounds, if not signifying completely different chemistries, which is corroborated by the richer substructure around the 430 meV peak of LRNMC.
In the studied NCA samples, OH-groups were found in both fresh and aged materials in both the lithiated and the delithiated states (see Fig. 3a). While the presence of OH-vibrations in the fresh NCA samples could be interpreted as originating from OH-groups that are produced during the material synthesis, we explain in the following why we deem this to be unlikely. The presence of OH-groups in Ni-rich materials has been established in previous studies using other techniques.27–34 Hydroxide compounds can be found in transition metal oxide materials, such as NCA, at different stages of the materials production (the detailed description of the formed and consumed species is provided in the ESI†). However, as mentioned in the ESI,† there should, at most, only be trace amounts of OH-containing compounds concentrated at the surface of the NCA from the material production. Detection of these trace amounts with the used experimental parameters is unlikely. Therefore, we believe that the observed OH signal is related to OH-groups that formed by cycling, which is supported by the fact that an OH signal is absent in the pristine LRNMC sample. Thus, we conclude that the observed signal is not a result of remnant precursors inside of the samples.
One possible OH-formation pathway could stem from the proton exchange mechanism which has been reported for Ni-rich materials.31 This effect can be observed both at the surface and in the bulk because protons can migrate into the active material structure,41 making the OH-groups formed through this mechanism the most relevant for RIXS measurements due to its bulk sensitivity. There are two ways of forming H+ in the cell. Both include degradation of the electrolyte at higher and lower potentials. First, the dehydrogenation of EC to VC (vinylene carbonate) can take place at lower potentials according to reaction (1).42 The reaction occurs at the surface of the cathode material. Since H+ has high mobility, there are significant chances of protonating the bulk of the active material.
![]() | (1) |
Another pathway of protonation includes degradation of the EC electrolyte as a result of oxidation at higher potentials.33,43 The mechanism is a two-step process. First, oxygen, previously shown to be formed within the LRNMC and NCA structure, can move to the interface where it oxidizes EC electrolyte according to reaction (2):43
![]() | (2) |
Support for this reaction is the detection of CO2 when cycling at higher potentials which was measured on the same NCA electrodes.44 The generated water further reacts with Li+, which is present on the surface of the secondary particles29,31 and the following reaction occurs:
Li+ + H2O → H+ + LiOH | (3) |
Reaction (3) takes place on the surface of the active material. Subsequently, the formed protons can diffuse into the active material and once H+ is inside of the cathode structure, it is likely to form OH groups (4):
LiTMO2 + H+ → LiTMO-OH | (4) |
These oxyhydroxides have been observed in Ni-rich NMC before30 and previous studies of oxyhydroxides have shown that they are stable up to 4.2 V vs. Li/Li+,45 which supports the hypothesis that the formation of OH-groups is partially irreversible in the working range of the material. However, note that, assigning the exact mechanism of OH formation inside the cathode is difficult and outside the scope of the current study. Fig. 4 shows a schematic depiction of the proton exchange mechanism in connection to the charging curve of NCA. Low potential electrolyte decomposition is shown in light blue while the high potential proton exchange mechanism is shown in orange. We point out that all the formation pathways mentioned above may be substantially assisted by internal lattice superstructures,5 nanoporosity4 or interstitial pores or voids from microstructural defects within the cathode particles.46
Footnote |
† Electronic supplementary information (ESI) available: Surface OH-formation, raw RIXS data, electrochemical data. See DOI: https://doi.org/10.1039/d4cp01766h |
This journal is © the Owner Societies 2024 |