Lithium-ion batteries (LiBs) have been leading candidates for vehicle applications due to their high energy density and high power capability. ^{[ 1 – 4 ]} However, LiBs’ safety issues have to be soundly addressed before large scale application. ^{[ 5 – 7 ]} When an LiB experiences certain abusive situations, for example, shorting, the temperature of the battery can easily rise to the threshold of so-called “thermal runaway,” in which the temperature rises very rapidly and is out of control. ^{[ 6 ]} A series of chemical reactions are triggered during such a process, and a considerable amount of heat can be released, possibly leading to fire or explosion. Each component in the battery, including anode, ^{[ 8 – 10 ]} separator, ^{[ 11 – 13 ]} electrolyte, ^{[ 14 – 17 ]} and cathode, ^{[ 18 – 23 ]} has its role to play in this process, and the role of the cathode has been proved to be particularly important. Only oxide systems as cathode materials are discussed in the present review. Readers interested in thermal stability of polyanion systems are encouraged to read references in the literature. ^{[ 24 – 26 ]}

Dahn et al. investigated the thermal stability of charged cathodes, including Li _x NiO ₂ , Li _x CoO ₂ , and λ -MnO ₂ (fully charged state of LiMn ₂ O ₄ ), and found that they all release oxygen upon heating. ^{[ 22 ]} Arai et al. conducted DSC measurements for charged Li _x NiO ₂ , ethylene carbonate (EC, a major component of the electrolyte), and the combination of these two, ^{[ 23 ]} finding that heating the mixture of charged Li _x NiO ₂ and EC is much more exothermic than heating either of them individually. This suggests that oxygen-release from a cathode is a key factor contributing to heat generation of the whole reaction. More advanced cathode materials normally contain multiple metal elements. Thermogravimetric and calorimetric studies of charged Li _x Ni _0.8 Co _0.15 Al _0.05 O ₂ and charged Li _x Ni _1/3 Co _1/3 Mn _1/3 O ₂ reveal that the former releases a larger amount of oxygen at a lower temperature than the latter. ^{[ 27 ]} This correlates well with the fact that the former has much poorer safety characteristics in terms of a much greater amount of heat generated than the latter. ^{[ 17 ]} Such a relationship between oxygen-release and safety characteristics is further confirmed in conditions closer to real battery operation by testing full-battery cells which include anode, separator, electrolyte, and cathode. ^{[ 28 ]}

When a charged cathode is heated, it tends to release oxygen. This is considerably detrimental because the released oxygen can react with the electrolyte in a highly exothermic way, significantly accelerating the elevation of temperature, and initiating further disastrous reactions. Therefore, designing cathode materials with suppressed oxygen-release and/or increased onset temperature for oxygen-release would be an effective approach to such safety issues. While thermogravimetric, ^{[ 9 ]} calorimetric, ^{[ 23 ]} and computational studies ^{[ 29 ]} have provided useful information like heat generation rate, chemical reaction kinetics, and so on; it is desirable to understand the roles of crystal and electronic structural changes in the oxygen-release process, especially through in situ techniques. Hence, our group designed in situ x-ray diffraction (XRD) and in situ x-ray absorption (XAS) techniques for studying the thermal stability of charged cathode materials. Here, we start with the experimental setup, then review results from our studies, ^{[ 30 – 35 ]} and finally discuss the implications of these results for designing a stable cathode.

A technique combining in situ x-ray diffraction with mass spectroscopy (MS) during heating was developed at the beamline 7B of the National Synchrotron Light Source (NSLS) facility by our group. ^{[ 32 ]} This technique enables us to monitor crystal structure changes and oxygen-release simultaneously. In an in situ XRD-MS experiment illustrated in Fig.  1 , a charged cathode is placed inside a tailored quartz tube where it can be heated at a controlled rate. The quartz tube is then connected to the MS. The whole system is purged with helium gas to provide an inert atmosphere. Once the charged cathode is heated, its structure is probed and recorded using a high intensity synchrotron beam as the x-ray source in a time-resolved fashion, while the outlet gas information is recorded by the MS. A detailed description of the experimental setup can be obtained from our previous papers. ^{[ 31 – 34 ]}

Fig. 1.

	Figure Option View Download New Window
	Fig. 1. (a) In situ XRD-MS and (b) in situ XAS experimental setup. The figure is adapted from Ref. [ 32 ].

Different contents and concentration ratios of metal elements in the cathode materials result in different thermal stability of the cathodes. This suggests that each element can play a unique role in stabilizing or destabilizing the structure. Commonly used metal elements for lithium-ion battery cathodes are mainly 3d transition metal elements, which are very close to each other in the periodic table, making it difficult to distinguish them by XRD alone. Therefore, x-ray absorption is a very valuable tool in differentiating them. An XAS spectrum includes both the x-ray absorption near edge structure (XANES) and the extended x-ray absorption fine structure (EXAFS). These parts can provide information about the oxidation state, electronic structure, and local environment in an elemental selective way. An in situ XAS technique (shown in Fig.  2 ) has been employed by our group at beamline 18A of NSLS. In such experiments, a charged cathode is positioned in the center of a helium-purged chamber and exposed to the incoming x-ray with variable wavelength. As the sample is heated up, the absorption coefficient of each element as a function of incoming x-ray wavelength is recorded in a spectrum, which is to be analyzed later for electronic structural and local structural information. Combining the techniques of in situ XRD-MS and in situ XAS not only reveals the fundamental structural reason for oxygen-release, but also specifies the positive or negative role of each individual element in the stability of the cathode. This information provides valuable guidelines for the development of safer cathodes. In the following discussions, a review of our work on layer-structured Li _x Ni _0.8 Co _0.15 Al _0.05 O ₂ , Li _x Ni _1/3 Co _1/3 Mn _1/3 O ₂ , and high voltage spinel Li _x Ni _0.5 Mn _1.5 O ₄ is given, followed by discussions of the importance of these studies for cathode development. Note that all of the work discussed here was carried out on charged cathode samples in the absence of the electrolyte. Although adding the electrolyte to the sample could make the experimental conditions closer to real cell operation, it would reduce the comparability and reproducibility of the experiments. Since our main interest was in the fundamental understanding of the cathode materials, we excluded electrolyte from the setup of these experiments.

From the above examples, the oxygen-release at high temperatures is an inevitable event for charged oxide cathode materials. Such inevitability can be understood from a thermodynamic point of view. ^{[ 29 ]} During the temperature elevation, the total amount of metal cations, including lithium and transition metals, is fixed but the amount of oxygen can vary. This implies that the phase diagram of multi-metal oxides with a fixed metal-to-metal ratio can provide the thermodynamic roadmap for the phase transition that occurs during temperature increase. (For instance, phase transitions of fully delithiated LiNi _0.5 Mn _1.5 O ₄ at high temperature can be well understood by referring to the phase diagram of Ni/Mn oxide with a nickel-to-manganese ratio of 1:3.) Since, in these phase diagrams, the high temperature phase is normally low in oxygen compared to the low temperature phase, it is not surprising that oxygen-release can hardly be avoided for charged oxide cathode materials. In addition, the high oxygen partial pressures of oxides with highly oxidized cations such as Ni ⁴⁺ and Co ⁴⁺ imply that the cations all have a strong tendency to reduce. ^{[ 53 ]} Therefore, both the phase diagram and the high oxidation state of transition metals in the charged cathode samples can explain the thermodynamic origins of the oxygen-release.

Since both structural change and transition metal reduction (and therefore oxygen-release) are thermodynamically driven, strategies for suppressing the oxygen-release leverage kinetic factors. In the charged oxide cathode materials that consist of Ni, Co, and Mn, the ions of Ni, Co, and Mn are oxidized close to Ni ⁴⁺ , Co ^{x +} (3.5 < x < 4), and Mn ⁴⁺ . ^{[ 54 ]} Among these, Ni ⁴⁺ is the easiest ion to reduce (to Ni ²⁺ ), while Mn ⁴⁺ the most difficult to reduced (to Mn ²⁺ ). Therefore, the reduction of the Ni ⁴⁺ ions takes place at the early stage of the heating process and generates a relatively large amount of oxygen gas. This is a critical step that governs the overall thermal instability of the material. To complete the structural transformation (such as layered to spinel and spinel to rock-salt) associated with the reduction of Ni ⁴⁺ , rearrangement of the Ni ions and other transition metal ions is necessary. This process requires migration of the transition metal ions from their original octahedral sites into either octahedral sites in neighboring layers or tetrahedral sites. ^{[ 55 ]} Direct migration between octahedral sites is energetically unfavorable because the transition metal ions face strong columbic repulsion on the travel path. They prefer to travel through the nearest tetrahedral sites as shown in Fig.  4 . Therefore, how easily the structure can transform into another form depends on how easily the transition metal ions can migrate from octahedral to tetrahedral sites, depending on the site preference of individual transition metal as well as the availability of the tetrahedral vacancy sites. The transition metal ions’ preference for tetrahedral or octahedral site is determined by the 3d electron configuration, which can be simply explained by crystal field theory (CFT). ^{[ 56 , 57 ]} In the transition metal oxides, the 3d orbitals of the transition metal ions split into different energies due to interaction with the surrounding oxygen atoms, and this process depends on the coordination environment, as shown in Fig.  8 . The energy gain, called crystal field stabilization energy (CFSE), can be calculated on the basis of the 3d electron configuration of the transition metal ions. By comparing the CFSE between octahedral and tetrahedral environments, the site preference of the transition metal ions can be estimated. ^{[ 58 ]} Ni ⁴⁺ (3d ⁶ ) and Ni ³⁺ (3d ⁷ ) have strong preference for octahedral site, while Ni ²⁺ (3d ⁸ ) has a less strong preference for octahedral site. Therefore, migration of the Ni ions between octahedral and tetrahedral sites requires their reduction into Ni ²⁺ . Therefore, in order to make the Ni migration-associated structural transformation occur, a relatively large amount of Ni ⁴⁺ needs to be reduced to Ni ²⁺ first, resulting in a large amount of oxygen released at an early stage of the heating process. For NCA, it can be considered that some of the Ni in LiNiO ₂ is replaced by Co and Al. The reduction of the Co ions follows that of the Ni ions during the heating process, but Co ^3.67+ (charged state) needs less reduction (Co ^2.67+ in Co ₃ O ₄ ) to be able to migrate, so less oxygen gas is generated during this process. Meanwhile, the occupation of Co ions in tetrahedral sites will impede the migration of Ni ions, pushing the formation of the rock-salt phase to a higher temperature and spreading the oxygen-release in a wider temperature range compared with pure LiNiO ₂ . Due to the high nickel content and low cobalt and alumina contents in NCA material, the rock-salt phase still forms at a relatively low temperature and a considerable amount of oxygen is released, because there are too few Co ions to occupy the tetrahedral sites and stabilize the structure. For NCM series materials, the onset oxygen-release temperature increases with increasing cobalt content, but decreases with increasing nickel content. The amount of Co ions in LiNi _0.5 Co _0.2 Mn _0.3 O ₂ is sufficient to stabilize the overall structure at the spinel phase, so no further reduction of manganese is necessary during the entire heating process. Therefore, the onset temperature of oxygen-release is delayed and less oxygen gas is released in comparison with NCA. In contrast, the Mn ⁴⁺ ions need to be reduced for migration in high voltage spinel LiNi _0.5 Mn _1.5 O ₄ during the heating process. Although Mn ⁴⁺ needs only to be reduced to Mn ³⁺ for migration (through a charge disproportionation reaction Mn ²⁺ prefers tetrahedral sites), ^{[ 55 ]} the relatively large amount of Mn ion reduction still generates a large amount of oxygen, causing thermal instability.

Fig. 8.

Figure Option View Download New Window

Fig. 8. Crystal field splitting of the transition metal 3d orbital in tetrahedral and octahedral environments. The tetrahedral crystal field splitting energy Δ t is smaller than the octahedral crystal field splitting energy Δ 0 and yields the relationship Δ t = 4/9 Δ 0 . The site preference of the transition metal ion can be predicted by crystal field stabilization energy (CFSE) calculated based on its electron configuration.

In summary, reduction of the transition metal ions, accompanied by oxygen-release and structural transformation, is unavoidable during the heating process in most cases. Since the structural transformation requires migration of the transition metal ions through tetrahedral sites, cations that need no or less reduction to become mobile and prefer to occupy tetrahedral sites favor overall thermal stability of the cathode materials. Such cations stabilize the structure at a certain phase at an early stage of the heating process, delaying further structural transformation to higher temperatures. Less oxygen gas will be generated, and oxygen-release will be spread across a wider temperature range. For example, some fixed valence cations (Mg ²⁺ , Al ³⁺ , Zn ²⁺ , etc.), migrate easily to tetrahedral sites during the heating process and have been proven to improve thermal stability of transition metal oxide cathode materials. However, these cations are electrochemically inactive, so a considerable amount of substitution will come at the expense of the capacity of the materials. Substitution of specific electrochemically active elements, such as Co and Fe, could be a better choice to enhance the thermal stability of oxide materials. This has been demonstrated in the above discussion of the NCA and NCM series materials and will be further studied in Fe-substituted high voltage spinel in our future work. ^{[ 59 ]}

The time-resolved XRD and XAS studies reveal that the structure and chemical composition of the materials play important roles in determining their structural stability during heating. These XRD and XAS data were collected by averaging a sample area on mm scale, therefore only reflecting the average structural changes. In fact, the structural transformation during the heating process involves nucleation and propagation of new structure, which takes place at the atomic scale. Exploring the structural changes during heating with high spatial resolution and precise awareness of location would provide valuable insights for understanding the overall thermal stability of the materials. High resolution transmission electron microscopy (HRTEM) is a suitable tool for probing the local details of the phase transformation during heating, because it offers both local structure and chemical information. ^{[ 60 , 61 ]}

In our work, in situ TEM has been employed to study the structural origin of the overcharge-induced thermal instability of two cathode oxide materials that exhibit significant differences in thermal stability, NCA and NCM. ^{[ 30 ]} Detailed TEM analysis reveals that overcharged NCA and NCM particles both have complex core-shell-surface structures, which cannot be detected by XRD. For overcharged NCA, HRTEM imaging reveals three structures in the scale of tens of nanometers, as shown in Fig.  9 : rhombohedral (located in the core of the particle), spinel (the shell near the surface), and rock-salt (at the surface layer of the particle). In contrast, the overcharged NCM particles have a core-shell-surface structure with O1 (CdI ₂ -type) on the surface, the spinel phase in the shell, and the rhombohedral phase in the core. In situ TEM experiments for NCA during heating reveal rapid growth of the rock-salt phase along with oxygen-release, while for NCM, slow structural transformation into the spinel phase was observed. These observations are in good agreement with the time-resolved XRD results. More importantly, the TEM results indicate that the difference in surface structures of overcharged NCA and NCM particles before heating is responsible for their different thermal decomposition behaviors. The rock-salt phase on the surface of overcharged NCA acts as a seed to accelerate the phase transformation to spinel and rock-salt phases during heating. The rapid growth of the rock-salt structure, accompanied by a release of a large amount of oxygen gas, causes thermal instability in NCA. In contrast, the CdI ₂ -type surface structure on overcharged NCM particles protects the material from losing oxygen during the heating process. Therefore, it postpones the oxygen-release reaction to a higher temperature, resulting in better thermal stability. This research reveals that the surface structure of the materials will strongly influence the phase transformation kinetics during heating. Therefore, surface modification is another effective way to improve the thermal stability of the materials. For example, a concentration-gradient layered oxide material, ^{[ 62 ]} with a surface rich in manganese and bulk rich in nickel, shows better thermal stability than pure high-nickel-content layered oxide materials (such as NCA), because the decomposition of the manganese rich surface structure forms a CdI ₂ -type MnO ₂ structure, which suppresses the oxygen-release associated with the phase transformation for the entire material.

Fig. 9.

	Figure Option View Download New Window
	Fig. 9. HRTEM images, selected area electron-diffraction pattern (SAEDP) and simulated structure of overcharged Li x Ni 0.8 Co 0.15 O 2 . Three phases, rhombohedral (in the core), spinel (in the shell), and rock-salt (on the surface), have been identified on the charged particle in the same image. [ 30 ]

Note that structural decomposition has been found on the surface of the overcharged particle at room temperature, as mentioned above. For Li _x Ni _0.8 Co _0.15 O ₂ ( x < 0.15), the overcharge-induced new phase propagates from the surface to the bulk of particles, following the process of rhombohedral to spinel to rock-salt. This resembles the phase transformation sequence observed in heating. For most transition metal oxide cathode materials, the structural decomposition, accompanied by oxygen-release, often occurs in a deeply delithiated state (high voltage charging). This kind of structural transformation normally proceeds mildly and starts from the surface of the material, making it difficult to track by bulk characterization tools such as XRD. However, understanding the origin of the structural decomposition is vital to the development of stable cathode materials, because these subtle irreversible structural changes will accumulate during electrochemical cycling and result in capacity fading during long-term cycling. As revealed, the surface structure evolution during electrochemical cycling is similar to the bulk structure evolution during heating. The knowledge gained from thermal studies might provide valuable information for predicting the structural evolution of the material over long-term electrochemical cycling. This is understandable, because the metastable highly delithiated (or overcharged) structure tends to transform into a stable structure during electrochemical cycling, and heating accelerates this transformation. Therefore, approaches that are effective to improve the thermal stability of the material could also be applied to enhance its structural stability during electrochemical cycling. Recent research on lithium rich manganese based layered oxide cathode materials demonstrates this point. The lithium rich materials attract lots of interest nowadays, because they can deliver exceptionally high reversible capacity, exceeding 250 mA·h·g ⁻¹ between 4.8 V and 2 V. ^{[ 54 , 63 , 64 ]} Despite delivering high capacity, lithium-rich materials exhibit several practical shortcomings, such as continuous fading of both capacity and voltage during electrochemical cycling. ^{[ 65 – 67 ]} The scientific community has made substantial efforts to understand these phenomena. The layered-to-spinel structural transformation, accompanied by oxygen release, is found to occur on the surface of the material during charging at high voltage. This phase transformation behavior, which is considered to be one of the primary factors responsible for the voltage and capacity fading of lithium-rich layered material, is analogous to its thermal decomposition behavior at an early stage of heating (< 250 °C, results will be reported in our future work). ^{[ 68 , 69 ]} Surface coating, ^{[ 70 ]} an effective approach to improve thermal stability for most oxide materials, can also be applied to suppress the oxygen-release of lithium-rich materials during high voltage charging and to retard the structural transformation into the spinel phase. Strategies of substituting specific cations that are able to inhibit the thermal decomposition of the material during heating ^{[ 67 ]} can also be applied to alleviate the structural transformation of lithium-rich layered material during electrochemical cycling. Bear in mind that the structural evolution revealed by thermal studies is more significant and easier to identify than the subtle structural changes occurring during each electrochemical cycle. Thermal studies may provide an alternative way to understand the structural origin of the cyclic instability of the materials and offer useful guidance in developing more structurally stable electrode materials for lithium ion batteries.

The structure changes of several typical charged oxide cathode materials (NCA, NCM, and high voltage spinel) during a heating process, at both bulk and atomic levels, are reviewed in a comparative way, based on results obtained from in situ time-resolved XRD and MS, in situ XAS, and in situ TEM experiments. It has been found that the structural transformation (or decomposition) together with oxygen-release is inevitable for charged oxide cathode materials during the heating process. Several approaches are proposed to improve the thermal stability of oxide cathode materials: (i) substituting specific cations that require only slight reduction to migrate into tetrahedral sites of the oxygen framework at early stages of heating is an effective way to improve the intrinsic thermal stability of the material. On one hand, the slight reduction of these cations will generate only a small amount of oxygen. On the other hand, the occupation of these ions in the tetrahedral sites will impede the migration of other transition metal ions that are required for further structural transformation, so the onset temperature of structural transformation will be pushed higher. (ii) Surface modification that can prevent the structure decomposition from the surface is another effective approach to improve the thermal stability of the cathode materials. In addition, it has been revealed that the structural transformation of the oxide cathode materials, due to high voltage charging during electrochemical cycling, is similar to their structural transformation observed during heating. Therefore, the information obtained from thermal studies may also provide valuable insights for developing electrode materials with better cycle stability.