With the density functional theory (DFT) (Hohenberg, 1964; Kohn and Sham, 1965), the process of Li-ion de/insertion in α-Fe₂O₃ and TiO₂ is studied. Ultrasoft pseudopotentials were used to calculate the interaction of the ionic core and valence electrons for Fe 3d⁶4s², Ti 3s²3p⁶3d²4s², Li 1s²2s¹, and O 2s²p⁴. The exchange and correlation terms were described with the generalized gradient approximation (GGA) of Perdew–Burke–Ernzerhof (PBE) (Perdew et al., 1996; Perdew et al., 1998). Monkhorst–Pack k-point meshes were conducted to address the Brillouin-zone integrations (Monkhorst and Pack, 1976). Convergence criteria were set as follows: the maximum force, the maximum stress, and the maximum displacement on the atom were below 0.1 eV nm⁻¹, 0.02 GPa, and 5.0 × 10^–5 nm, respectively.

A periodic structure was adopted in the calculation. The unit cells of α-Fe₂O₃ and anatase TiO₂ (space group of R-3C and 141/AMD) contained 12 iron atoms, 18 oxygen atoms, four titanium atoms, and eight oxygen atoms (Supplementary Figure S1). Moreover, 380 eV was set as the cutoff energy for both. The number of k-points was set as 6 × 6×2 and 7 × 7×3. The calculated lattices agreed well with the theoretical and experimental data (Finger and Hazen, 1980; Arlt et al., 2000; Zhang et al., 2017; Jenness et al., 2018) (Supplementary Table S1).

With vigorous stirring in an oil bath at 75°C, 50 ml of 2.0 mol/L FeCl₃·6H₂O solution was added to a round-bottom flask containing 50 ml of 5.4 mol/L NaOH. After stirring for 5 min, a red-brown Fe(OH)₃ colloid formed in the flask, which was then transferred to a high-temperature and high-pressure PTFE reactor. The hydrothermal reaction was conducted at 100°C for 4 h. After cooling to room temperature, the obtained red precipitate was centrifuged, successively rinsed with deionized water and ethanol three times, and then dried overnight to obtain α-Fe₂O₃.

The synthesized α-Fe₂O₃ was treated with oxalic acid, for which 8 ml of deionized water was added to a beaker containing 0.2 g of α-Fe₂O₃, and then 0.1 g of oxalic acid was added. After shaking at room temperature for 6 h, the red precipitate was obtained by centrifugation and then successively rinsed with deionized water and ethanol three times. After drying overnight, oxalic acid-treated α-Fe₂O₃ was obtained.

The core–shell α-Fe₂O₃@TiO₂ synthesis process is illustrated in Figure 1. In a beaker containing 33 ml of absolute ethanol, 0.1 g of oxalic acid-treated α-Fe₂O₃ was added, and then, 0.1 ml of concentrated ammonia water was added with stirring and stirred for 5 min. Next, 0.25 ml of tetrabutyl titanate (TBOT) was added with vigorous stirring, and ultrasonication was performed for 40 min. The solution in the beaker was transferred to the PTFE reactor. After hydrothermal reaction at 45°C for 24 h, the solution was cooled to room temperature and then centrifuged. The red precipitate was rinsed with deionized water and ethanol three times and then dried overnight. In an air environment, the product was calcined at 450°C for 2 h to obtain α-Fe₂O₃@TiO₂, and then, 0.1 g α-Fe₂O₃@TiO₂ was added to 25 ml of 10 mol/L HCl and oscillated for 0.5, 1, 2, 4, and 12 h. Then, they were successively rinsed with deionized water and ethanol three times and dried overnight to obtain α-Fe₂O₃@TiO₂ at different etching degrees. The samples were denoted as FT-0.5h, FT-1h, FT-2h, FT-4h, and FT-12h (pure TiO₂).

X-ray diffraction (XRD, Bruker D8, Germany) was used to characterize the synthesized powder samples. The morphology of each sample was observed by field emission scanning electron microscopy (FESEM) and high-resolution transmission electron microscopy (HRTEM, JEOL, JSM-2100, Japan). The component and valence analysis of the synthesized samples were carried out by X-ray photoelectron spectroscopy (XPS, ESCALAB 250, an Al K excitation source, United States). Additionally, HRTEM (TF20, JOEL, 2100F, Japan) was used to map the elemental distribution of the samples. The specific surface area and pore size analyzer (BET, ASiQ Mini QUANTACHROME) was used to determine the porosity of the samples.

The prepared samples were mixed with acetylene black and polyvinylidene fluoride (PVDF) at a weight ratio of 7:2:1 in N-methyl-2 pyrrolidone (NMP). The obtained slurry was uniformly smeared on a 9-μm-thick copper foil and dried in a vacuum oven at 80°C for 10 h, and then a 12-mm-diameter negative plate was cut, and 1 M LiPF₆, in a mixture of 1:1:1 vol.% ethylene carbonate (EC), diethyl carbonate (DEC), and dimethyl carbonate (DC) was used as the electrolyte (ternary electrolyte LBC305-1, Shenzhen Kejinzida Technology Co., Ltd.). With a 15.6-mm × 0.45-mm lithium sheet as the positive electrode and a Celgard 2325 diaphragm as the battery separator, the prepared negative electrode was assembled into a CR2025 battery in a glove box (Super, Germany Mikrouna) that was filled with argon gas (water concentration <0.1 ppm and oxygen concentration <0.1 ppm). The cycling and rate performance were tested with a Blue Power instrument (Wuhan Blue Electronics Co., Ltd.). An electrochemical workstation (Shanghai Chenhua Instrument Co., Ltd.) was used to obtain cyclic voltammetry (CV) curves, and a different electrochemical workstation (ZAHNER Company, Germany) was used to conduct electrochemical impedance spectroscopy (EIS).

According to the symmetry of α-Fe₂O₃ and anatase TiO₂, there are one and two symmetrical positions for Li ions to be inserted into α-Fe₂O₃ (6b) and TiO₂ (16 h, 8 days), respectively (Figure 2). One unit cell of α-Fe₂O₃ can accommodate six Li ions, but one unit cell of TiO₂ can only accommodate four Li ions. Thus, the reaction equation of Li ions and Fe₂O₃ is as follows: Fe 2 O 3 + Li + + e − → Li ( Fe 2 O 3 ) . (1)

Eq (1) is consistent with Yang et al. (2019). The reaction process of Li ions and TiO₂ is given in Eq. 2: TiO 2 + Li + + e − ↔ LiTiO 2 . (2)

From Eq. 2, the theoretical capacity of TiO₂ can be calculated (335 mA h g⁻¹), which is consistent with Liang et al. (2020). After six Li ions were inserted into the α-Fe₂O₃ unit cell, the volume expanded by 9.64%. The volume of TiO₂ expanded by 1.15% after four Li ions were inserted. Moreover, the one, two, three, four, and five and one, two, and three Li ions inserted in the cells at the different sites of α-Fe₂O₃ and TiO₂ were also calculated (Supplementary Table S2). With an increasing number of Li ions inserted in the α-Fe₂O₃ cell, the volume continued to expand. However, it was amazing to find that the expansion rate of TiO₂ always remained below 2.2%. With the same number of Li ions inserted in the cell, the volume expansion of α-Fe₂O₃ was larger than that of TiO₂. There are two reasons for this result: one is that the interstitial vacancy (8 days and 16 h) of TiO₂ is larger than that of α-Fe₂O₃ (6b) and the other is that with Li ions insertion, the angle of Ti-O-Ti (Figures 2C,D) changes from 155.148° to 179.022°, which acts as a buffer. Hence, TiO₂ can effectively maintain its cell structure. Furthermore, it is believed that with an increasing number of Li ions inserted in α-Fe₂O₃ and the combination of O ions, the cell will continue to expand. This mechanism is the reason to coat α-Fe₂O₃ with TiO₂. Moreover, it is not found that the bonds of Ti-O and Fe-O are broken, which can be verified from Figure 2, the population analysis (Supplementary Table S3), and the density of states (DOS) (Figure 3).

The electronic population is usually used to assess the covalent or ionic nature of a bond. A high value of the bond population expresses a strong covalent bond (Aouadi, 2006). Both α-Fe₂O₃ and TiO₂ have two types of bonds, which are marked I and II in Supplementary Figure S1. From Supplementary Table S3, it can be found that the electronic populations of Fe-O I were little larger than those of Fe-O II. Meanwhile, the bond length of Fe-O I was shorter than that of Fe-O II. Nevertheless, the electronic population of Ti-O I was larger than that of Ti-O II, which means Ti-O I had stronger covalent bonds. When Li ions were inserted in α-Fe₂O₃ and TiO₂, the electronic population and bond length increased. Bond length increase can be attributed to volume expansion. Among them, the electronic population of Ti-O I increased the most, which means Ti-O I was enhanced. Instead of breaking, Ti-O I was strengthened.

The DOS of α-Fe₂O₃ and TiO₂ express two main regions near the Fermi level (Figure 3). The more important part is the first region, which is predominantly formed by non-metal O 2p states and Ti 3d (Fe 3d) states located between −10 and 0 eV. This part forms a strong p-d covalent bonding. With the Ti (Fe) and O atoms getting squeezed, it is found that the DOS below the Fermi level move to the low-energy part (the DOS of TiO₂+4 Li and Fe₂O₃+6 Li), which shows that Ti-O and Fe-O are enhanced. Furthermore, the TiO₂+4 Li DOS move more than the Fe₂O₃+6 Li DOS. This result indicates that Ti-O is more enhanced, which agrees well with the electronic population analysis.

The SEM images of α-Fe₂O₃ without coating TiO₂ are shown in Figures 4A,B. α-Fe₂O₃ is successfully synthesized during the experiment. The XRD result is consistent with the diffraction peak of α-Fe₂O₃ (Haematite, syn) in PDF#33-0664 (Supplementary Figure S2). α-Fe₂O₃ has a cubic morphology with relatively smooth surfaces and a size of approximately 400–500 nm. In order to compare the effect of oxalic acid-treated and untreated α-Fe₂O₃ coating with TiO₂, untreated α-Fe₂O₃ is used for TiO₂ layer cladding. According to Figures 4C,D, when untreated α-Fe₂O₃ is coated with TiO₂, it is more prone to heterogeneous nucleation, and agglomeration is easy to occur. After cladding, the homogeneous Fe₂O₃ particles become inhomogeneous α-Fe₂O₃@TiO₂ particles.

Figure 5 shows the samples of oxalic acid-treated α-Fe₂O₃ coated with TiO₂ and etched for different periods of time. Clearly, with an increasing etching time, α-Fe₂O₃@TiO₂ gradually changes from a core–shell structure to a completely hollowed out TiO₂ structure. The average core–shell sizes of different samples are listed in Supplementary Table S4. With hydrochloric acid etching, the thickness of shell changes a little, reducing from 58 to 45 nm, and maintains at 53 nm during the etching time of 1–4 h, while the core size changes greatly. The vertical diagonal (diagonal 1) changes from 560 to 100 nm. The horizontal diagonal (diagonal 2) changes from 568 to 419 nm. Thus, hydrochloric acid etching mainly aims to Fe₂O₃.

As shown in Figure 6, the sample etched for 1 h (FT-1h) possesses a well-defined core–shell structure with a buffer chamber between the α-Fe₂O₃ core and the TiO₂ shell. From Figure 6A, it can be clearly observed that TiO₂ is successfully coated on α-Fe₂O₃. Furthermore, the TiO₂ coating layers are very uniform with few ruptures (Figure 6B). In addition, Figure 6C shows that there is a chamber between the α-Fe₂O₃ core and the TiO₂ shell whose thickness is about 80 nm. According to the EDS mapping of a single α-Fe₂O₃@TiO₂ core–shell structure (Figure 6D), Fe, Ti, and O are evenly distributed. The red section in Figure 6E shows that the TiO₂ layer covers the surface of α-Fe₂O₃@TiO₂, and the yellow part in Figure 6F represents the α-Fe₂O₃ core structure. The green section illustrated in Figure 6G refers to O that is jointly contained by α-Fe₂O₃ and the TiO₂ covering layer in a single α-Fe₂O₃@TiO₂ core–shell structure. The composition of composite is further demonstrated by energy-dispersive spectroscopy (EDS), shown in Figure 6H. From the illustration, it shows that the compound is mainly composed of Fe₂O₃ and TiO₂, and Fe₂O₃ is the main part. Figure 7 displays the HRTEM image of FT-1h α-Fe₂O₃@TiO₂. The stripes and grids on dark and light are different, which demonstrated that they were assembled by small crystals with measured lattice d-spacings of 0.35 and 0.37 nm, corresponding to the (101) plane of anatase TiO₂ and the (012) plane of Fe₂O₃, respectively. So, it can be proven that TiO₂ is uniformly dispersed on the surface of Fe₂O₃.

The chemical composition and crystal phase of the samples were investigated by XRD analysis. Figure 8 shows that the diffraction peak 2θ = 25.28°, 37.8°, 48.05°, 55.06°, and 62.69° corresponds to the (101), (004), (200), (211), and (204) planes of TiO₂, which is consistent with PDF#21-1272 (the diffraction peak of anatase). Samples FT-0.5h, FT-1h, FT-2h, and FT-4h present the characteristic peaks of Fe₂O₃ and TiO₂, which expressed that TiO₂ coating is successfully combined with α-Fe₂O₃. Furthermore, it can be observed that the intensity of the characteristic peak at 24.1° of α-Fe₂O₃ decreases, and the characteristic peak at 25.3° of TiO₂ increases with an increasing etching time. In addition, no additional peaks were observed, suggesting high purity of the samples, which shows good agreement with the results of XPS spectra (Figure 9).

The XPS characterization was conducted to examine the chemical composition of samples (FT-0.5h, FT-1h, FT-2h, FT-4h, and FT-12h (pure TiO₂)). The complete survey spectrum (Figure 9A) of α-Fe₂O₃@TiO₂ samples (FT-0.5h, FT-1h, FT-2h, and FT-4h) reveal the existence of Ti, Fe, and O elements. For the C 1s XPS spectrum, the peak at 285.00 eV is attributed to adventitious carbon (Zhong et al., 2017; Yang et al., 2019). The high-resolution XPS spectrum of Fe 2p was composed of two distinct peaks at 710.90 and 724.40 eV (Figure 9B), which correspond to Fe 2p^2/3 and 2p^1/2 with satellite lines. The spectrum is consistent with the characteristic of Fe³⁺ α-Fe₂O₃ (Zhong et al., 2017; Yang et al., 2019). In the spectrum of Ti 2p (Figure 9C), the Ti 2p^3/2 and Ti 2p^1/2 peaks were located at binding energies of 458.70 and 464.65 eV, respectively, which is in agreement with the value of Ti⁴⁺ in the TiO₂ lattice (Zhong et al., 2017; Zhao et al., 2018). The spectrum of the O1s core level is shown in Figure 9D, where binding energy peaks at 531.80 eV originate from bonded hydroxyl groups (Yang et al., 2019). For the broad peak centered at 530.00 eV, it is attributed to metal-bonding in both oxides (Zhao et al., 2018; Yang et al., 2019).

The texture characteristic of α-Fe₂O₃@TiO₂ was further confirmed by N₂ adsorption/desorption isotherm, as shown in Figure 10. It can be observed from Figure 10A that the specific surface area of pure α-Fe₂O₃ was very small, only 2.07 m³ g⁻¹. After coating with TiO₂, the specific surface area increased to approximately 30 m³ g⁻¹, and that is because pure α-Fe₂O₃ is smoother and denser than TiO₂, which can be seen from Figure 4. The surface bulge of TiO₂ was more obvious than that of pure α-Fe₂O₃. After etching for 0.5, 1, 2, 4, and 12 h (pure TiO₂), more pores formed. The specific surface area increased to 32.1, 34.39, 36.96, 43.845, and 131.91 m³ g⁻¹. The isotherms were identified as IUPAC type IV, which is characteristic of mesoporous materials. The pore size distribution obtained from the isotherm indicates a number of pores 4–8 nm in the samples (Figure 10B). With increasing etching time, the pore volume increases. A structure with abundant mesopores is likely to buffer the volume expansion and allow the penetration of electrolyte for complete contact with the active material, thus playing an important role in improving the electrochemical properties of α-Fe₂O₃@TiO₂.

Electrochemical performances of α-Fe₂O₃@TiO₂ (FT-1h) were investigated as anode materials for LIBs to demonstrate the effectiveness in improving lithium storage. Figure 11A shows the first four CV curves of the FT-1h sample between 0.01 and 3 V at a scan rate of 0.1 mV⁻¹. It can be observed that the sample has a reduction peak and an oxidation peak at a potential of 1.73/2.13 V, which may be attributed to the de/insertion of lithium ions from TiO₂, as shown in Eq. 2. There are two oxidation peaks at potentials of 0.92 and 1.12 V, which correspond to the lithium reaction of Fe₂O₃, respectively, as shown in Eq. 3 (Fu et al., 2015). Fe 2 O 3 + 6 Li + 6 e − → 2 Fe 0 + 3 Li 2 O . (3)

In addition, the oxidation peak between 1.5 and 2.0 V may correspond to Fe⁰ being reoxidized to Fe₂O₃, as shown in Eq. 4 (Fu et al., 2015). Fe 2 O 3 + 6 Li ↔ 2 Fe 0 + 3 Li 2 O . (4)

After the second charge–discharge cycle, the peak position slightly shifts, and the peak strength slightly weakens, which may be attributed to the polarization of the anode material and the formation of an SEI film during the charge–discharge process (Fu et al., 2015). It can be clearly observed that the curves of the fourth charge–discharge cycle and the third charge–discharge cycle coincide well, indicating that the FT-1h sample has good cycling stability.

Charge and discharge curves of the sample FT-1h with a voltage of 0.01–3 V are shown in Figure 11B. According to the previous analysis, a wide slope from 1.75 to 0.8 V is attributed to the Li⁺ intercalation into the crystal structure of α-Fe₂O₃ and the phase transformation from hexagonal to cubic Li₂Fe₂O₃ (Qin et al., 2018). Last, an obvious plateau at 0.8 V is ascribed to the reduction from Fe²⁺ into Fe⁰, and the formation of amorphous Li₂O and an SEI layer (Zhu et al., 2020). After the first charge and discharge cycle, the discharge/charge specific capacities of the FT-1h sample are 1,609.3 mA h g⁻¹ and 1,021.6 mA h g⁻¹, respectively. In the second process, a plateau at 1.0 V is observed, which agrees well with Eq. 3. Furthermore, the discharge specific capacity is reduced to 1,228.6 mA h g⁻¹, and the Coulombic efficiency is 99.1%. The fifth circle is similar to the second, except the length of the plateau, which is shorter than the one in the first circle, and this is due to the irreversible capacity loss derived from the formation amorphous Li₂O (Yang et al., 2019). The discharge specific capacity is 999.4 mA h g⁻¹ after the fifth cycle. The 55th, 100th, and 150th curves express the similar circle curves, whose plateau is at 0.9 V, which is consistent with Eq. 3. After 150 charge–discharge cycles at a current density of 100 mA g⁻¹, the reversible capacity remains at 893.7 mA h g⁻¹, and the Coulombic efficiency is 98.4%.

Rate measurements of pure α-Fe₂O₃, oxalic acid-treated α-Fe₂O₃ coated with TiO₂ and etched for different periods of time (FT-0.5h, FT-1h, FT-2h, FT-4h, and FT-12h (pure TiO₂), and the charge–discharge tests at a current density of 100 mA g⁻¹ for each cell after the rate performance test are shown in Figure 11C. The reversible capacities of FT-1h at current densities of 100 mA g⁻¹, 200 mA g⁻¹, 500 mA g⁻¹, 1 A g⁻¹, and 2 A g⁻¹ reach 1,489.3, 586.8, 438.7, 332.7, and 224.5 mA h g⁻¹, respectively. When the current density is returned to 100 mA g⁻¹, the reversible capacity is still maintained at 555.5 mA h g⁻¹. After 100 charge–discharge cycles at a current density of 100 mA g⁻¹, the reversible capacity remains at 682.0 mA h g⁻¹, and the Coulombic efficiency is 99.7%. During the long cycle, the Coulombic efficiency is maintained above 98%. The capacity decreases at the beginning and continuously increases during the latter long term cycle test. It is supposed to be the grain boundaries of iron metal and Li₂O formed in the electrochemical process (Eq. 4), which contributes to the extra energy storage of the composite (Li S. et al., 2019). This phenomenon is common for most MO _x (M = Fe, Co, and Ni) attributed to the activation of the materials (Kaprans et al., 2018). Comparing the different etching times, it can be found that the battery performances of the etched samples are higher than that of pure α-Fe₂O₃ and TiO₂, indicating that the different degrees of buffer chambers can relieve, to a certain extent, the volume expansion of α-Fe₂O₃ during the discharge/charge process. Furthermore, the SEM images of pure Fe₂O₃ and FT-1h after 100 charge–discharge cycles are shown in Supplementary Figure S3. Fe₂O₃ morphology gets damaged severely if TiO₂ is not protected. On the contrary, sample FT-1h maintained the original morphology, which proved again that TiO₂ can effectively inhibit the volume expansion. Among the etched samples, FT-1h possesses the best performance. Figure 11C shows that FT-1h still has good cycling performance after the large rate performance test, while the reversible capacities of pure α-Fe₂O₃ and pure TiO₂ are low. The as-prepared α-Fe₂O₃@TiO₂ in this work shows superior cyclability and capacity compared with others summarized in Table 1.

EIS resistance tests are conducted for pure α-Fe₂O₃, pure TiO₂, and FT-1h, and the Nyquist plots of FT-1h, pure α-Fe₂O₃, and pure TiO₂ are shown in Figure 11D. All of them exhibit a similar profile. There is a depressed semicircle in the high-frequency region and an inclined line in the low-frequency region. The diameter of the semicircle in the high-frequency region of each cell is related to the tenable resistance of the electrolyte (R _e ), SEI layer (R _s ), and charge transfer resistance (R _ct ) (Wang et al., 2020). R _ct is associated with the transfer of electrons and Li⁺. The inclined line in the low-frequency region represents the Warburg impedance (Z _w ), which is derived from lithium-ion diffusion in electrode materials. A semicircle with a larger diameter in the high-frequency region represents a larger R _ct (Yang et al., 2019). From Figure 11D, it can be found that the impedance of α-Fe₂O₃ alone is relatively large, and the impedance decreases after coating with TiO₂, indicating that the electrical conductivity of the electrode is improved by forming a composite material. In addition, the higher slope of FT-1h compared with pure α-Fe₂O₃ in the low-frequency region is evidence of a stronger interaction between Li ions and FT-1h.