Pouch full cells were prepared by the CATL Company. The cathode was composed of LiCoO₂, super P black (SP), and polyvinylidene difluoride (PVDF) in a weight ratio of 94:3:3. The anode was made of hard carbon and PVDF binder in a weight ratio of 92:8. The electrolyte of LiPF₆ () was mixed in ethylene carbonate/diethyl carbonate (EC/DMC) with volume ratio 1:1. The assembly process was operated in a drying room.

All cells were tested by using the electrochemical workstation at 45 °C. For the 1st cycle, the cells produced in the same batch were firstly charged to 3.9 V at a rate of 0.1 C then to a certain cutoff voltage (i.e., 4.2 V, 4.4 V or 4.6 V) at a rate of 0.7 C and kept for 0.2 h, and then discharged to 3 V at a rate of 0.7 C and kept for 0.2 h. For the subsequent cycling test, the schedule was almost the same as that in the 1st cycle except that the charging rate to 3.9 V is 0.7 C.

All the tested cells were transported into Ar-filled glovebox and disassembled at an Ar atmosphere with O₂ and H₂O content lower than 0.1 ppm. The cathode samples were washed three times by using anhydrous DMC to remove the interference of electrolyte. To avoid any possible interference from the atmosphere, the samples were sealed with Al-plastic bags, taken to the Taiwan Light Source (TLS), and then transferred to the test chamber through an Ar-filled glovebox.

The samples with various cutoff voltage including 3.9 V, 4.2 V, 4.4 V, and 4.6 V at the 1st cycle were prepared for the charge compensation study. Samples at fully charged (4.4 V or 4.6 V) or discharged (3 V) states with the different capacity retention ratios (i.e., 100%, 99%, 51%, 42% at 4.6 V, and 100%, 72%, 50% at 3 V) and samples cycled 10 times between different voltage ranges (3 V–4.4 V versus 3 V–4.6 V) were prepared for studying the correlation between the evolution of the electronic structure and capacity fading.

The sXAS measurements of the LiCoO₂ cathode were carried out at beamline 20A1 of the Taiwan Light Source (TLS) at HsinChu, Taiwan, beamline 08U of Shanghai Synchrotron Facility (SSRF) and beamline 12B-a of the National Synchrotron Radiation Laboratory (NSRL) at Hefei. In particular, the beamline 20A1 was equipped with a 6-m high-energy spherical grating monochromator (6-m HSGM) to supply a photon beam with resolving power up to 8000. The spectra of Co L_2,3 and O K-edge were collected in TEY and TFY mode in an ultrahigh-vacuum chamber with a base pressure of about (1 Torr = 1.33322 × 10² Pa). The TEY and TFY spectra were normalized to the photon flux of incident beam monitored by the Au mesh. The photon energy was calibrated with the spectra of reference samples (CoO for Co L_2,3-edge and SrTiO₃ for O K-edge).

The amount of Li⁺ extraction with various cutoff voltages is shown in Fig. 1(a). With the cutoff voltage increasing in steps of 0.1 V from 4.3 V to 4.6 V, the corresponding increments of the amount of Li⁺ extraction are 0.0811, 0.0607, and 0.1083, which correspond to 8.1%, 6.1%, and 10.8% of the theoretical capacity of LiCoO₂, respectively. It indicates that each increment of voltage by 0.1 V from 4.3 V is crucial to improve the energy density of the LiCoO₂ cell, among which the increase from 4.5 V to 4.6 V is the most significant.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (color online) Electrochemical tests of LiCoO2, showing (a) amount of Li+ increasing with upper cutoff voltage, and (b) cycling performances for the charge cutoff voltages of 4.4, 4.5, and 4.6 V.

The cycling performances with the upper cutoff voltage of 4.4, 4.5, and 4.6 V are shown in Fig. 1(b). The LiCoO₂/carbon battery cycled between 3 V and 4.4 V only has a capacity loss of 4.22% after 120 cycles. With the upper cutoff voltage increased to 4.5 V, the cells can only keep 19.40% of the initial capacity after 110 cycles. The capacity decreases sharply to 0 within 20 cycles with the upper cutoff voltage of 4.6 V. It demonstrates that the cycling performance is severely deteriorated by each increase of 0.1 V from 4.4 V even though the energy density can be improved notably.

To avoid any possible interference from the side reaction at the electrode/electrolyte interface, the TFY spectra of Co L_2,3-edge and O K-edge, which reflect the bulk evolution, are firstly studied. As shown in Fig. 2(a), Co L-edge is divided into L₂ (792 eV to 802 eV) and L₃-edge (778 eV to 788 eV), which originate from the transition from the 2p_1/2 and 2p_3/2 states due to the core-hole spin-orbital coupling. Co L₃-edge of the initial LiCoO₂ has 3 peak features, marked as B, A, and C, respectively.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (color online) (a) Normalized Co L2,3-edge x-ray absorption spectra and (b) normalized oxygen K-edge x-ray absorption spectra of Li1−xCoO2 with respect to the cut-off voltage in the TFY mode, and ((c) and (d)) magnified parts of Co L3 and O K pre-edge spectra above 4.4 V, respectively, which are convenient for comparison.

When charged to 3.9 V, the intensity of peak B increases notably while the peak A shifts to higher energy and becomes broader. During charging from 3.9 V to 4.4 V, the intensity of peak B continues to increase while peak A continues to shift to higher energy and become broader. However, there is almost no change when charging from 4.4 V to 4.6 V, as confirmed by the careful comparison in Fig. 2(c). In other words, all the features of the Co L₃-edge, including the increasing intensity of peak B and blue-shift of peak A, evolve below 4.4 V. In addition, it should be noticed that peak C disappears gradually when the battery is charged to 4.6 V. We cannot determine, however, whether it is an intrinsic evolution or just because of the broadening of peak A. Moreover, it has been suggested that peak C results from Co³⁺ in high spin state and is less sensitive to the distortion of Co–O octahedron.^[42,43] Thus, only the evolution of features B and A are discussed. Peak B is attributed to the generation of Co²⁺ in some research.^[44,45] If this assignment were true, we would conclude that a fair amount of Co²⁺ ions generate at least at 3.9 V during the 1st charging process. It contradicts with the experimental facts that major redox should be oxidized with Li⁺ extracted. In addition, CoO is thought to be electrochemical inactive in the voltage range from 3 to 4.6 V.^[46,47] Thus the assignment also contradicts with the good cyclability of LiCoO₂ with a cutoff voltage less than 4.2 V. A more reasonable assignment is the contribution from Co⁴⁺.^[41] The evolution of peak B indicates that the oxidation of Co³⁺ to Co⁴⁺ provides electrons with the battery being charged below 4.4 V. Peak A, which takes the major part in Co L-edge, is mainly related to Co³⁺. It has been pointed out that the gravity of the L₃ shifts to higher energy mainly due to the oxidation of Co³⁺.^[45,48] Someone has suggested that the evolution of peak A is also closely associated with the re-hybridization between Co and O molecular orbital.^[28,49] Nevertheless, the evolution of peak A is consistent with peak B, indicating that the changes of Co occur below 4.4 V.

Next, the corresponding spectra of O K-edge are studied. As shown in Fig. 2(b), O K-edge spectra can be divided into two parts. The pre-edge peaks with the energy lower than 535 eV are assigned to the transitions from oxygen 1 s orbital to the states hybridized between O 2p orbital and Co 3d orbital. The hump above 535 eV is assigned to the transition to the states hybridized between O 2p and Co 4sp orbitals. The O K pre-edge of the pristine LiCoO₂ has only one notable feature, marked as peak a. Because most of Co³⁺ ions stay at the low-spin state,^[24,45] where t_2g orbital is fully filled while e_g is completely unfilled (i.e. ), peak a is thought to originate from the hybridization between Co³⁺ e_g states with O 2p states. When the battery is charged to 3.9 V, two new notable features b₁ and b₂ at lower energy appear. When they are charged from 3.9 V to 4.4 V, the intensity of peaks b₁ and b₂ increase a bit. Peaks b₁ and b₂ were reported to result from the hybridization of O 2p states with the t_2g and e_g states of Co⁴⁺.^[41] Thus the increasing intensity is consistent with the oxidation from Co³⁺ to Co⁴⁺, where both the amount of holes in 3d states and the covalency of Co–O increase. So far, the spectral changes of the O K-edge XAS, though relatively more dramatic, are consistent with those of Co L_2,3-edge XAS with respect to the Li extraction. It should be noted that because of the convolution of features at O–K pre-edge resulting from both O 2p hole and the hybridization between Co and O, it is difficult to distinguish the contribution of O 2p holes to the electron compensation below 4.4 V.

What is more attractive is the evolution of O K-edge above 4.4 V where there is no obvious variation at the Co L_2,3-edge. By careful comparison, as shown in Fig. 2(d), peak intensity in the range of about 528 eV–528.8 eV increases slightly while that in the range of about 529 eV–532 eV decreases from 4.4 V to 4.6 V. This evolution corresponds to the charge compensation from O, indicating the formation of O 2p hole with extra Li⁺ extraction. Moreover, we can conclude that the O 2p holes are very local or even not related to the Co-O bond at all, otherwise the Co L-edge will change more or less. The feature in the range of about 529 eV–532 eV is different from that reported previously where O 2p holes were thought to lie between 530 eV–535 eV and led the intensity of peaks to increase with the Li⁺ extracted.^[50–52] The different position might result from a different coordination environment around O in the Li_xCoO₂ compared with those in the Li-excess cathodes. The reducing intensity of this feature might be related to the increasing intensity of the feature in the range of about 528 eV–528.8 eV. This feature is consistent with the feature of reported on KO₂ previously and has also been found in the study on Li₂MnO₃.^[53,54] Thus the evolution of these two features might indicate that the formation of may result from the combination of the O anions with a local hole, which gives an explanation to the origin of or even the evolution of O₂ and needs more confirmation in future. Nevertheless, our results indicate that the evolution of O 2p holes can be found in LiCoO₂ conventional cathodes.

The TEY spectra of Co L_2,3 are shown in Fig. 3(a). The discussion on the Co TEY will be simplified as the evolution trend is very similar to that in the bulk in spite of a better signal-to-noise ratio leading to unambiguous evolution. We thus conclude that Co³⁺ turns into Co⁴⁺ below 4.4 V on the surface, which is confirmed by the careful comparison between the magnified parts, as shown in Fig. 3(c).

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) (a) Normalized Co L2,3-edge x-ray absorption spectra, (b) normalized oxygen K-edge x-ray absorption spectra of Li1−xCoO2 with respect to the cut-off voltage in the TEY mode, and ((c) and (d)) magnified parts of Co L3 and O K pre-edge spectra above 4.4 V, respectively, which are convenient for comparison.

The evolution of surficial O is revealed by the TEY spectra of O K-edge in Fig. 3(b). While the battery is charged from 3.9 V to 4.4 V, relative intensities among b₁, b₂ and a evolve more notably than the changes of bulk O, indicating the interface undergoes more severe evolution than the bulk accompanied with side reactions. Besides, new peak features marked as c and d evolve. According to previous reports,^[55,56] peaks c and d are assigned to the lithium salts from the reaction between LiCoO₂ and electrolyte. With the LiCoO₂ cathode immersed into the electrolyte in open-circuit voltage (OCV) state, the faint peaks c and d indicate that the interface changes a bit but is stable overall. When the battery is charged to 3.9 V, peaks c and d become more obvious, indicating that interfacial reactions already take place at 3.9 V. However, the invisible change of peaks c and d from 3.9 V to 4.6 V indicates that the side reactions turn very weak.

Like the case in the bulk, the O evolution above 4.4 V draws more attention. Through the careful comparison between 4.4 V and 4.6 V shown in Fig. 3(d), the evolution is the same as that in TFY spectra except that the amplitude in TEY spectra is larger, suggesting that more O 2p holes evolve on the surface.

To shed light on the mechanism of capacity fading from the viewpoint of electronic structure, the sXAS spectra of LiCoO₂ cathodes with different capacity retention ratios are studied, where the samples cycled between 3 V and 4.6 V are taken for example because of the most severe capacity degradation. Since the O K-edge is more sensitive to the evolution than the Co L-edge, only the TEY and TFY spectra of the O K-edge are shown.

The TFY data of the LiCoO₂ cathodes in the 1st cycle are shown in Fig. 4(a). The fully discharged state (i.e., discharged to 3 V) looks like the pristine state except remaining weak features of b₁ and b₂. It indicates that most of the LiCoO₂ in the bulk evolves reversibly in spite of a bit residual of Co⁴⁺. In comparison, the surficial evolution is much more severe as shown in Fig. 4(b). For the fully discharged state, the stronger features b₁ and b₂ while the weaker peak a indicate more residuals of Co⁴⁺ on the surface.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (color online) (a) TFY and (b) TEY spectra of O K-edge of LiCoO2 cathodes in the 1st cycle, (c) TFY and (d) TEY spectra of O K-edge of LiCoO2 cathodes at fully charged state or fully discharged state with various capacity retention ratios.

The capacity degradation during cycling is further studied. As shown in Fig. 4(c), the TFY spectra of fully charged states regardless of capacity retention ratios have little changes compared with the fully charged state in the 1st cycle. It indicates the Co–O bonds at fully charged states are very similar regardless of capacity degradation. However, in the fully discharged states, the spectra with capacity retention of 72% and 50% look very similar to the fully charged states except a small hump corresponding to peak a, which is very different from the evolution in the 1st cycle. It indicates that more Co⁴⁺ ions cannot return back to Co³⁺ but stay at the fully charged state, leading to the capacity fading. The invisible evolution with the ratio from 72% to 50% might indicate that the structural change plays a less important role in the capacity fading after many cycles.

The TEY data are shown in Fig. 4(d) to study the surficial evolution. The intrinsic signal from LiCoO₂ is submerged by peaks c and d with capacity fading regardless of charged states or discharged states. It indicates that a large quantity of lithium salts are generated on the surface of the LiCoO₂ cathode after many cycles. The intensity of peaks b₁, b₂, and a become weaker with the capacity retention ratio decreasing from 99% to 42% at the fully charged state and from 72% to 50% at the fully discharged state. It indicates more and more lithium salts aggregate on the surface and they might be the decisive factor to capacity degradation after many cycles. What is more, it should be noted that the difference in spectrum between the fully discharged state and fully charged state is much larger than the difference in spectrum between the different retention ratios at the same state. This indicates that the surface layer experiences reactions involving decomposition and formation with a certain reversibility during cycling.

The spectra of O K-edge of LiCoO₂ in fully charged state cycled 10 times with cutoff voltages of 4.4 V and 4.6 V are compared. As shown in Fig. 5(a), the TFY data are much similar, indicating that the Co–O bonds are almost the same at fully charged states. When we observe the TEY spectra shown in Fig. 5(b), peaks c and d become dominant in both cases, indicating that a large quantity of lithium salts cover the surface. The weaker intrinsic feature of the LiCoO₂ cycled at 4.6 V demonstrates the more aggregations of lithium salts, which contribute to a larger capacity fading.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. (color online) (a) TFY and (b) TEY spectra of O K-edge of LiCoO2 cathodes after 10 cycles with cutoff voltages of 4.4 V and 4.6 V.