III. RESULTS AND DISCUSSION

Figure 1(a) depicts the discharge–charge curves of a 23.8 nm-thick Li₄Ti₅O₁₂(111) film, which was synthesized on Nb-doped SrTiO₃(111) during the 30 cycles. The spectroelectrochemical cell used for in situ NR consisted of a Li₄Ti₅O₁₂(111)/Nb:SrTiO₃(111) cathode, 1.0 mol/dm³ LiPF₆-containing PC electrolyte, and a Li metal anode. A plateau region was observed at around 1.5 and 1.6 V in the discharge and charge curves, respectively, which represents a two-phase coexistence of spinel-type Li_4+αTi₅O₁₂ and ordered rock-salt type Li_7−βTi₅O₁₂ domains ^{Reference Borghols, Wagemaker, Lafont, Kelder and Mulder12,Reference Guerfi13,Reference Ohzuku, Ueda and Yamamoto37} that demonstrates the lithium intercalation activity of the Li₄Ti₅O₁₂(111) film electrode. The first discharge and charge curves showed slope regions from 2.3 to 1.7 V and from 2.4 to 3.0 V, respectively, which could be associated with side reactions occurring in the cell. As a Nb:SrTiO₃(111)/1.0 mol/dm³ LiPF₆ PC/Li cell delivered no significant capacities in these voltage regions (see Supporting Information S1), electrochemical decomposition of electrolyte species did not contribute the discharge and charge capacities in the slope regions. Thus, the slope regions observed during the first cycle suggested an irreversible phase change in the Li₄Ti₅O₁₂(111) film. During the following cycles, the Li₄Ti₅O₁₂(111) film exhibited reversible lithium intercalation with high capacity retention and high cycle efficiency, as shown in Fig. 1(b). The discharge capacity was 2.18 μAh (116 mA h/g) at the 3rd cycle and remained stable at 2.16 μAh (115 mA h/g) at the 30th cycle. The discharge capacities were much smaller than the theoretical value of lithium intercalation from Li₄Ti₅O₁₂ to Li₇Ti₅O₁₂ of 3.28 μAh for the 23.8 nm-thick film (corresponding to 175 mA h/g). From these results, we can speculate that lithium intercalation reversibly proceeded into the Li₄Ti₅O₁₂(111) except for a part of the film which irreversibly changed to an inactive phase at the initial electrochemical process.

FIG. 1.

(a) Discharge–charge curves of a 23.8 nm-thick Li₄Ti₅O₁₂(111)/SrTiO₃(111) film with an electrolyte of 1 mol/dm³ LiPF₆ PC and (b) variation of discharge capacity and cycle efficiency during 30 cycles. The discharge–charge measurements were collected using an in situ NR cell. The constant current applied was approximately 6.6 μA, which corresponds to a current density of 350 mA/g (2 C) for the Li₄Ti₅O₁₂ film electrode. The active area of the Li₄Ti₅O₁₂(111) film was 15 × 15 mm.

Ex situ XRD and XRR analyses were conducted to investigate the structural change of the Li₄Ti₅O₁₂(111) induced by the electrochemical processes. Figure 2(a) shows out-of-plane XRD patterns of the 23.8 nm-thick Li₄Ti₅O₁₂(111) in pristine condition and after 30 cycles. The intensities were normalized using the value of the SrTiO₃(111) diffraction peak, which was obtained by fitting a Gaussian curve. The pristine film generated diffraction peaks at 18.4°, 37.2°, and 79.6° along the out-of-plane [111] direction, which are attributed to the hhh diffraction peaks of Li₄Ti₅O₁₂ with the $Fd\bar 3m$ space group, and these observations confirmed the (111) orientation of the Li₄Ti₅O₁₂ film. The out-of-plane lattice parameter was calculated to be 8.365(3) Å, and no considerable change in lattice parameter was observed for the cycled sample, which supported the reversible intercalation of the Li₄Ti₅O₁₂ film during the 30 cycles. By contrast, the diffraction peaks showed a decrease in the intensities after the electrochemical cycling, which indicates a partial phase change from Li₄Ti₅O₁₂ to another phase, as proposed from transmission electron microscopy and scanning probe microscopy studies. ^{Reference Kitta, Akita and Kohyama26,Reference Kitta, Akita, Maeda and Kohyama27} Figure 2(b) illustrates the XRR analysis results and the x-ray SLD profiles of the pristine and cycled Li₄Ti₅O₁₂ films. The reflectivity spectra were plotted as functions of the scattering vector, Q _z = 4π sin θ/λ, where λ is the x-ray wave length (0.1541 nm) and θ is the glancing angle. A two-layer model consisting of a surface impurity and the Li₄Ti₅O₁₂ film closely fitted the reflectivity curve of the pristine film. The thickness, surface roughness, and x-ray SLD value of the Li₄Ti₅O₁₂ film were refined to 23.6 ± 0.4 nm, 1.5 ± 0.6 nm, and 2.83 ± 0.02 × 10⁻³ nm⁻², respectively [Fig. 2(c)]. The SLD value was consistent with that expected for stoichiometric Li₄Ti₅O₁₂ (2.82 × 10⁻³ nm⁻²). The pristine film was covered with a surface impurity layer with a thickness of 1.0 ± 0.3 nm, and the surface impurity had a SLD value of 1.69 ± 0.04 × 10⁻³ nm⁻². Considering the SLD values of typical lithium containing oxides (1.74 × 10⁻² nm⁻² for Li₂CO₃, 1.61 × 10⁻³ nm⁻² for Li₂O, and 1.25 × 10⁻³ nm⁻² for LiOH), the surface layer appears to mainly consist of Li₂CO₃. The fringes of the XRR spectrum of the cycled sample shifted to lower Q _z values, which indicates an increase in the total thickness of the film [Fig. 2(b)]. Thus, we added a SEI layer to the structural model since a SEI layer has been reported to form on the Li₄Ti₅O₁₂ surface by electrolyte decomposition. ^{Reference Liu, Bian, Li, Ji, Hao and Yu21–Reference He, Liu, Huang, Zhang, Yu, Li, Kang and Kim24} Furthermore, we divided the Li₄Ti₅O₁₂ layer into two parts: a bulk region and a surface region for investigating the irreversible phase change indicated by the discharge–charge and XRD results. The three-layer model composed of SEI layer/Li₄Ti₅O₁₂-surface/Li₄Ti₅O₁₂-bulk/SrTiO₃ fitted more closely (χ² = 5.22 × 10⁻³) than the three-layer model composed of SEI layer/Li₄Ti₅O₁₂/SrTiO₃ (χ² = 6.41 × 10⁻³). The fitting curve and the SLD profile as calculated using the four-layer model are shown in Figs. 2(b) and 2(c). In the SLD profile, the bulk region of the cycled Li₄Ti₅O₁₂ had a similar SLD value to that of the pristine sample, whilst the 10-nm surface region had a higher SLD value of 3.00 ± 0.03 × 10⁻³ nm⁻². The XRR analyses highlighted the phase change in a high-density phase in the surface region. The SEI layer on the Li₄Ti₅O₁₂ surface had a thickness of approximately 4 nm after 30 cycles, and the slope of the SLD profile implied a gradation of chemical composition in the SEI layer.

FIG. 2.

(a) Out-of-plane XRD patterns, (b) XRR analysis results of Li₄Ti₅O₁₂(111)/SrTiO₃(111) films, and (c) x-ray SLD profiles in pristine condition and following electrochemical cycling. The XRD patterns and XRR spectra were collected in air.

To clarify the phase change process occurring at the Li₄Ti₅O₁₂ surface, we conducted in situ NR analyses at the Li₄Ti₅O₁₂/LiPF₆-containing PC interface upon battery operation. Figure 3 shows the NR spectrum fitting results and neutron SLD profiles of the Li₄Ti₅O₁₂(111) film during the first and fifth discharge–charge processes. The reflectivity spectra were plotted as functions of the scattering vector, Q _z = 4π sin θ/λ, where λ is the neutron wave length and θ is the glancing angle. Similar to the XRR analysis, a two-layer model consisting of surface impurity/Li₄Ti₅O₁₂/SrTiO₃ fitted closely with the reflectivity curve for the pristine film [Fig. 3(a)]. The SLD value of the pristine Li₄Ti₅O₁₂ films was 2.03 ± 0.04 × 10⁻⁴ nm⁻², which is consistent with the theoretical value of 2.04 × 10⁻⁴ nm⁻² as calculated with the ICSD data (#15787). The pristine film was covered with a surface impurity layer with a thickness of 2.0 ± 0.3 nm, and the surface impurity had a higher neutron SLD value than the Li₄Ti₅O₁₂ layer [Fig. 3(b)]. Considering the SLD values of typical lithium containing oxides (3.46 × 10⁻⁴ nm⁻² for Li₂CO₃, 8.15 × 10⁻⁵ nm⁻² for Li₂O, and 6.02 × 10⁻⁶ nm⁻² for LiOH), the surface layer consisted of Li₂CO₃, which is consistent with the XRR result. The formation of the carbonate and hydroxide components was indicated by x-ray photoelectron spectroscopy (see Supporting Information S2). These observations demonstrate that NR can detect the structural changes that occur within several nanometers of the electrode surface.

FIG. 3.

(a) In situ NR analysis results of epitaxial Li₄Ti₅O₁₂(111) films during cycles 1 and 5 in 1 mol/dm³ LiPF₆ PC electrolyte. Refined neutron SLD profiles (b) at the first discharge and (c) during the following processes. The insets in (b, c) depict magnified profiles near the Li₄Ti₅O₁₂ regions.

NR spectra collected under the electrochemical conditions were analyzed using a three-layer model (SrTiO₃/Li₄Ti₅O₁₂-bulk/Li₄Ti₅O₁₂-surface/surface impurity or SEI/electrolyte), which was a simple extension of the model used for the cycled sample in the ex situ XRR analysis. Following the cell construction (OCV: open circuit voltage), the SLD value of the Li₄Ti₅O₁₂ layer increased in the 10–15 nm region from the Li₄Ti₅O₁₂/electrolyte interface [Fig. 3(b)], whereas no significant change was observed in the bulk region. This result indicates that the phase change from the Li₄Ti₅O₁₂ was initiated following soaking in the electrolyte prior to electrochemical cycling. Kitta et al. have reported that a phase transition to a lithium-deficient rocksalt (LiTi₅O_10.5) occurs at the Li₄Ti₅O₁₂ surface in 1.0 mol/dm³ LiPF₆-containing PC electrolyte, characterized from TEM/electron energy-loss spectroscopy. ^{Reference Kitta, Akita and Kohyama26} The SLD increase in the surface region is consistent with the previous report, because the lithium-deficient phase should have a higher SLD value than Li₄Ti₅O₁₂ because natural lithium has a negative coherent scattering length for neutrons (−1.9 fm for ⁷Li:⁶Li = 92.5:7.5). Further, the decrease in the SLD value of the surface impurity region under the OCV condition was observed, which indicates the chemical or electrochemical reactions of the surface impurity phases with electrolyte species at the interface. This could be attributed to the initiation process of the SEI formation. During the first discharge from OCV to 1.2 V, the Li₄Ti₅O₁₂ surface region showed a further increase in the SLD, and the phase change proceeded in the surface region during the first discharge. In contrast to the surface region, the bulk region showed no significant change in the SLD from the OCV to 1.6 V and from 1.4 V to 1.2 V. Thus, the discharge capacity observed from 3.0 V to 1.6 and 1.4 V–1.0 V in the first discharge curve may primarily correspond to the phase change of Li₄Ti₅O₁₂ in the surface region. The SLD value in the Li₄Ti₅O₁₂ bulk region decreased to 1.76 ± 0.08 × 10⁻⁴ nm⁻² from 1.6 V to 1.4 V. The decrease in the SLD corresponds to lithium intercalation in the Li₄Ti₅O₁₂ bulk region, which was consistent with the plateau region observed at around 1.5 V in the first discharge curve. The SLD value of the bulk region at the discharged state corresponded to the theoretical value for Li₇Ti₅O₁₂ (1.71 × 10⁻⁴ nm⁻²). The lithium intercalation proceeded in the Li₄Ti₅O₁₂ bulk region through the surface region. The neutron SLD profiles during the subsequent conditions are shown in Fig. 3(c). After the first charging, the SLD value in the Li₄Ti₅O₁₂ bulk reversibly increased to 2.02 × 10⁻⁴ nm⁻², and the Li₄Ti₅O₁₂ bulk showed similar SLD values under the discharged and charged conditions at cycle 5 to those observed at the first cycle. Furthermore, the SLD value and the thickness of the Li₄Ti₅O₁₂ surface region remained stable at around 2.3 × 10⁻⁴ nm⁻² and 15 nm after the first cycle, respectively. The low electrochemical activity of the surface-reconstructed phase may cause a decrease in the reversible capacity, which is consistent with the small discharge–charge capacities of the Li₄Ti₅O₁₂(111) film (approximately 115 mA h/g) compared to the theoretical value (175 mA h/g). In contrast, highly-reversible lithium intercalation proceeds in the bulk region by lithium diffusion through the surface region during the electrochemical cycling. The SEI layer at the interface was grown in the initial cycles, and the thickness after the fifth cycle was 6.1 ± 0.2 nm. The SEI formation could suppress the further phase change in the surface region of Li₄Ti₅O₁₂. The depth profiling of the Li₄Ti₅O₁₂/electrolyte interface using in situ NR clarified the intermediate process of the surface-reconstruction and the SEI formation during the initial electrochemical cycling.

To investigate the effect of the electrolyte composition on the surface-reconstruction behavior, we conducted NR analyses of the Li₄Ti₅O₁₂(111) film in contact with different electrolytes. Figure 4 shows NR analysis results of Li₄Ti₅O₁₂(111) films in pristine condition and following soaking in different electrolytes of 2.4 mol/dm³ LiPF₆ PC and 1.0 mol/dm³ LiPF₆ EC/DEC (3:7 vol.). Both pristine Li₄Ti₅O₁₂(111) films were covered with the surface impurity layer. NR spectra collected under the OCV condition were analyzed using the three-layer model (SrTiO₃/Li₄Ti₅O₁₂-bulk/Li₄Ti₅O₁₂-surface/SEI/electrolyte) [Figs. 4(a) and 4(b)], which was the same model used for the Li₄Ti₅O₁₂(111) film in the 1.0 mol/dm³ LiPF₆ PC electrolyte. When the film electrode was soaked in the 2.4 mol/dm³ LiPF₆ PC electrolyte, the SLD increased in the entire region of the 16 nm-thick Li₄Ti₅O₁₂ film [Fig. 4(c)] and the SLD value of approximately 2.4 × 10⁻⁴ nm⁻² was similar to that observed for the Li₄Ti₅O₁₂ surface following the first cycle in the 1.0 mol/dm³ LiPF₆ PC (2.5 × 10⁻⁴ nm⁻²). The surface-reconstruction could be promoted via contact with the PC electrolyte of higher LiPF₆ content. Conversely, the Li₄Ti₅O₁₂ layer showed no significant changes in the SLD using 1.0 mol/dm³ LiPF₆ EC/DEC electrolyte [Fig. 4(d)]. Furthermore, no surface-reconstruction proceeded at the Li₄Ti₅O₁₂ surface following soaking in the EC/DEC-based electrolyte, and the surface-reconstruction of Li₄Ti₅O₁₂ was influenced by the solvent and the solute species in the organic electrolytes.

FIG. 4.

(a, b) In situ reflectometry analysis results and (c, d) neutron SLD profiles of epitaxial Li₄Ti₅O₁₂(111) films in pristine conditions and after soaking in different electrolytes of 2.4 mol/dm³ LiPF₆ PC (a, c) and 1 mol/dm³ LiPF₆ EC:DEC (3:7 vol.) (b, d).

Figure 5 shows the discharge–charge curves of the coin-type cell, which consisted of a 23.5 nm-thick Li₄Ti₅O₁₂(111) cathode, 1 mol/dm³ LiPF₆ PC electrolyte, and a Li anode. The discharge–charge curves had no clear plateau region at around 1.5 V, which corresponds to the lithium intercalation even though the same combination of cathode, electrolyte, and anode materials were used. It is generally considered that water and hydrogen fluoride in electrolytes lead to electrode deterioration, but it could not explain the difference in the deterioration because the coin-type cell contained a smaller amount of the electrolyte (∼0.5 mL) than the in situ spectroelectrochemical cell (∼5 mL). In contrast to the in situ spectroelectrochemical cell, which was made from electronically-insulating poly-chloro-trifluoro-ethylene, the coin-type cell was made from electron-conductive stainless steel. Thus, side reactions of the cell body with the electrolyte species could not have been negligible to identify the intercalation properties of the Li₄Ti₅O₁₂(111) film. The side reactions were investigated via the discharge–charge curves of the Nb:SrTiO₃ substrate in the 1 mol/dm³ LiPF₆ PC electrolyte which were collected using the coin-type cell [Fig. S1(b)]. The Nb:SrTiO₃ substrate without the deposited Li₄Ti₅O₁₂ active material exhibited the first discharge and discharge capacities of 1.6 and 1.2 μAh, respectively. These values were much larger than those observed for the in situ spectroelectrochemical cell [Fig. S1(a)], which confirmed the significance of the side reactions in the coin-type cell.

FIG. 5.

Discharge–charge curves of a 23.5 nm-thick Li₄Ti₅O₁₂(111)/SrTiO₃(111) film with an electrolyte of 1 mol/dm³ LiPF₆ PC. The discharge–charge measurements were collected using a 2032-type coin-cell. The constant current applied was approximately 1.3 μA, which corresponds to a current density of 175 mA/g (1 C) for Li₄Ti₅O₁₂ film electrode. The active area of the Li₄Ti₅O₁₂(111) film was 10 × 8 mm.

To further understand the effects of the side reactions on the intercalation inactivity of the Li₄Ti₅O₁₂(111) film, the discharge–charge measurements were conducted using coin-type cells with different electrolytes. Figure 6 shows the discharge–charge curves of 10-nm-thick Li₄Ti₅O₁₂(111) films in PC and EC/DEC based electrolytes with different LiPF₆ concentrations. The insets in each figure depict the dQ/dV plots at the first and 50th cycles. The discharge–charge curves of all coin-type cells mainly consisted of the slope regions, which were associated with the side reactions, and the capacities of the slope regions increased as the LiPF₆ concentration increased, which indicated the participation of the Li⁺ and PF₆ ^– ions in the side reactions. When the 0.5 mol/dm³ LiPF₆-containing PC was used as the electrolyte, the plateau regions were observed at around 1.4 and 1.7 V in the discharge and charge curves, respectively [Fig. 6(a)]. The sharp dQ/dV peaks observed at the 50th cycle confirmed that the reversible lithium intercalation proceeded in the Li₄Ti₅O₁₂(111) film. By contrast, the dQ/dV peaks became smaller as the LiPF₆ concentration increased to 1.0 and 2.4 mol/dm³ [Figs. 6(b) and 6(c)]. As described in the above sections, the Li₄Ti₅O₁₂(111) film delivered the reversible lithium intercalation using the spectroelectrochemical cell, even when the 1.0 mol/dm³ LiPF₆-containing PC electrolyte was used. Thus, we can speculate that the side reactions between coin-cell components and the electrolyte species generate some products that enhance the deterioration of the Li₄Ti₅O₁₂ electrode. It must be noted that the Li₄Ti₅O₁₂(111) films in LiPF₆-containing EC/DEC electrolytes showed clear plateau regions in the discharge–charge curves as the LiPF₆ concentration increased from 0.5 to 2.4 mol/dm³ [Figs. 6(d)–6(e)]. The discharge capacities in the plateau regions at the 50th cycle were 0.17, 0.19, and 0.18 μAh when the LiPF₆ concentrations were 0.5, 1.0, and 2.4 mol/dm³, respectively, which indicates no dependence of the capacity on the LiPF₆ concentration. Furthermore, the dQ/dV peaks were observed at around 1.55 V for both discharge and charge processes, showing a considerably small polarization compared to those observed for the PC-based electrolytes. These results confirmed that the highly-reversible lithium intercalation of Li₄Ti₅O₁₂ proceeded in the LiPF₆-containing EC/DEC electrolytes. Figure 7 shows XRD patterns of the 10-nm-thick Li₄Ti₅O₁₂(111) films after 50 cycles in different electrolytes. The Li₄Ti₅O₁₂(111) film cycled in the 1.0 mol/dm³ LiPF₆ EC/DEC showed a 111 diffraction peak at 18.5°, which supports the reversible lithium intercalation. By contrast, the Li₄Ti₅O₁₂(111) films cycled in the LiPF₆ PC electrolytes showed a diffraction peak at 22.5°, which could not be attributed to the Li₄Ti₅O₁₂ and Li₇Ti₅O₁₂ spinels. An irreversible phase change from the spinel-type phase was clearly detected after the discharge–charge reactions in the coin-type cells. The intensity ratios of the Li₄Ti₅O₁₂(111) to the new peaks decreased with LiPF₆ concentration, and the Li₄Ti₅O₁₂(111) peak was not observed for the sample cycled in the 2.4 mol/dm³ LiPF₆ PC. The decrease in the intensity ratio can be reasonably related to the decrease in the electrochemical capacity in the plateau regions shown as Figs. 6(a)–6(c). Our results demonstrated that the electrochemical stability of the surface region of Li₄Ti₅O₁₂ varied with electrolyte composition and cell components, and a phase change from the Li₄Ti₅O₁₂ spinel to an electrochemically-inactive phase occurs during the initial electrochemical process, which results in the small capacity and large polarization during the subsequent cycles.

FIG. 6.

Discharge–charge curves of 10-nm-thick Li₄Ti₅O₁₂(111)/SrTiO₃(111) films with (a–c) PC and (d–f) EC/DEC(3:7 vol.) electrolytes containing (a, d) 0.5, (b, e) 1.0 and (c, f) 2.4 mol/dm³ LiPF₆. The constant current applied was approximately 0.5 μA, which corresponds to a current density of 175 mA/g (1 C) for Li₄Ti₅O₁₂ film electrode.

FIG. 7.

Ex situ XRD patterns of the 10-nm-thick Li₄Ti₅O₁₂(111) films following 50 cycles in (a) 1 mol/dm³ LiPF₆ EC:DEC (3:7 vol.) and (b) 0.5, (c) 1.0, and (d) 2.4 mol/dm³ LiPF₆ PC electrolytes.