Promoting the electrochemical properties of yolk-shell-structured CeO₂ composites for lithium-ion batteries

INTRODUCTION

With the development of society, carbon neutralization has been augmented immensely due to the consumption of fossil fuels and climate change. There are significant demands for the development of renewable energy and storage technologies. Among the various energy storage systems, lithium-ion batteries (LIBs) are considered as promising energy storage devices due to their high energy density and specific power and long cycle life^[1-4]. Rapid charging/discharging technology is currently being increasingly investigated worldwide, which makes the development of new LIBs that provide stable cyclic stability and fast rate capability increasingly more urgent^[5,6]. However, it is difficult to satisfy these requirements since traditional graphite is limited by a theoretical specific capacity of only 372 mAh·g^-1. Simultaneously, the capacity of batteries with graphite anodes rapidly decreases to ~66% after 100 cycles^[7]. The demands of high electrochemical performance have engendered the search for alternative anode materials, including carbonaceous materials^[8-11], carbon-silicon composites^[12,13] and transition metal compounds^[14-19]. Rare earth anode materials have also been widely investigated because of their low redox voltage, high cycling rates and other promising properties^[20-22].

Cerium oxide (CeO₂) has been widely explored in the fields of supercapacitors, LIBs, catalytic supports in fuel cells and gas sensors^[23-25] due to its oxygen deficiency and the fast mutation between Ce(III) and Ce(IV)^[26-28]. Bare CeO₂ films^[29], core-shell nanospheres^[27], hollow spheres^[30] and CeO₂/carbon composites^[31] have all been investigated as LIB anodes. The basic working mechanism of cerium oxide in a LIB charge/discharge procedure can be described as^[27]:

2CeO₂ + 2Li⁺ + 2e^- ⇔ Ce₂O₃ + Li₂O

However, the main residual issues of cerium oxide as an anode material are low electronic conductivity and a crystal structure crushing (shatter effect) during cycling. To solve these issues, researchers have dedicated considerable effort to constructing composites with materials such as carbon-coated TiO₂^[32,33], SnO₂^[34-36] and Fe₃O₄^[37,38] to markedly enhance the electrochemical properties of cerium oxide. Structures compounded with carbon can effectively confine the matrix to alleviate volume changes and obstruct the aggregation of active particles. For oxide and sulfide anodes, due to the prominent volume expansion during Li⁺ intercalation/deintercalation, the use of core-shell structures may result in stress fractures and exfoliation^[27]. Inspired by previous studies, an attractive strategy to fabricate yolk-shell structures with tunable void spaces was proposed for CeO₂ with the aid of an outer carbon layer to avoid damage to the architecture^[39].

In this contribution, we report the design and synthesis of yolk-shell-structured CeO₂@void@C using a self-template strategy method. First, uniform CeO₂ nanospheres are synthesized from a solvothermal method. A layer of SiO₂ is then coated on the surface of CeO₂ to form CeO₂@SiO₂, which is further solved in a resorcinol formaldehyde (RF) solution to obtain core-shell-structured CeO₂@SiO₂@RF. By annealing and etching, the target CeO₂@void@C is obtained and fully characterized by structural determination and electrochemical investigations. The electrochemical performance of CeO₂@void@C is demonstrated according to its well-designed nanoarchitecture. The electrochemical reaction mechanisms are probed by X-ray diffraction (XRD) and nuclear magnetic resonance (NMR) spectroscopy.

MATERIALS AND METHODS

Sample preparation and characterization

Synthesis of yolk-shell-structured CeO₂@void@C nanospheres

Initially, uniform CeO₂ nanospheres were synthesized by a hydrothermal method and adopted as the initial materials. Typically, 1 g of Ce₂(CO₃)₃·6H₂O was dissolved in 1 mL of deionized water, with 1 mL of glacial acetic acid, 0.1 g of sodium citrate and 30 mL glycol then added. The mixture was stirred for 30 min and then transferred to a Teflon-lined hydrothermal autoclave, which was heated at 180 °C for 4 h. The product was washed with ethanol and deionized water three times and placed in a blast drying oven at 60 °C for 12 h to obtain the homogenous CeO₂ nanospheres.

The yolk-shell-structured CeO₂@void@C was prepared using the following steps. First, 0.1 g of CeO₂ nanospheres were dispersed in a 140 mL ethanol-water (1:6) solvent mixture via ultrasonic treatment for 30 min. Then, 3 mL of NH₄OH (28 wt.%) and 0.5 mL of tetraethyl orthosilicate were added, followed by stirring for an additional 6 h. Afterwards, the CeO₂ particles were coated by a SiO₂ layer, followed by centrifugal action, washing and drying at 60 °C for 12 h. Next, 0.1 g of CeO₂@SiO₂ were dispersed in a 100 mL ethanol-water (1:4) solvent and a homogeneous mixture was generated by ultrasound treatment for 30 min. Then, 0.1 g of resorcinol and 3 mL of NH₄OH reagent were added and subsequently stirred for 30 min. This was followed by the addition of 0.05 mL of formaldehyde into the mixed solution to react under continuous stirring for 6 h. The final product was further dried at 60 °C for 12 h to obtain CeO₂@SiO₂@RF. A condensed layer of SiO₂ and a polymeric layer of RF were coated onto CeO₂ via a successive sol-gel process. The carbon layer was coated on CeO₂@SiO₂ using RF as a precursor by annealing under an inert atmosphere (N₂). The procedure of making yolk-shell-structured CeO₂@void@C is illustrated in Figure 1. Finally, the uniform yolk-shell-structured CeO₂@void@C was obtained by etching off the sacrificial SiO₂ layer with a sodium hydroxide solution.

Figure 1. Schematic of preparation process for yolk-shell-structured CeO₂@void@C nanospheres.

Structure and morphological characterization

Crystal structures were determined using a PANalytical Empyrean X-ray diffractometer. The morphologies were characterized by scanning electron microscopy (SEM) with a JSM-7900F field emission scanning electron microscope. Thermogravimetric analysis (TGA) was performed on a Netzsch STA449 F5/F3 Jupiter thermal analyzer. The Raman spectra were determined using an inVia™ confocal Raman microscope. The specific surface area and aperture of the samples were measured by an ASAP2460 physical adsorption analyzer. The chemical states of the samples were studied with X-ray photoelectron spectroscopy (XPS) using a Thermo ESCALAB 250xi. The microscopic crystal structures were characterized using transmission electron microscopy (TEM) with a JEM-F200 field emission transmission electron microscope. Solid-state ⁷Li NMR spectra were measured on a Bruker spectrometer with a ⁷Li Larmor frequency of 155.52 MHz. All cycled anodes were packed into 4 mm rotors in an Ar glovebox. The samples were spun at a magic angle spinning rate of 8 or 10 kHz to assign the isotropic signal. The recycle delay d₁ was set to 6 s and the single 90° pulse length was 2.6 μs at a power of 300 W. Shifts were externally referenced to the 0-ppm peak of a 1 M LiCl solution. The spin-lattice relaxation times (T₁) were determined using the inverse recovery technique. The spectra were analyzed according to Dmfit^[40].

Electrochemical characterization

Lithium foil was employed as the counter electrode for the evaluation of the electrochemical performance. The active materials (yolk-shell-structured CeO₂@void@C or bare CeO₂), conductivity agent (Super P) and polyvinylidene fluoride in a weight ratio of 60:30:10 were mixed with the N-methyl-2-pyrrolidine solvent to generate a homogenous slurry. It was then painted onto Cu foil and dried at 80 °C overnight under a vacuum. LiPF₆ (1 M) was solved in diethyl carbonate/dimethyl carbonate/ethylene carbonate (1:1:1 vol.%) with the addition of 10 wt.% fluorinated ethylene carbonate to serve as the electrolyte. Celgard 2400 was used as the separator. The batteries were assembled in CR2032 coin cells. The electrochemical performance was studied through a Lanthe CT2001A/B testing system. Electrochemical impedance spectroscopy (EIS) was carried out using an SP-150 workstation (BioLogic, France).

RESULTS AND DISCUSSION

To make the synthesis procedure clearer, the complete routine is presented in Figure 1. First, CeO₂ nanoparticles were obtained using a solvothermal method, in which the homogenous mixture of Ce₂(CO₃)₃·6H₂O and Na₃C₆H₅O₇ was kept at 180 °C for 4 h. A layer of SiO₂ was then coated onto the as-synthesized CeO₂ via a sol-gel method to form CeO₂@SiO₂, which was further coated with RF to produce the core-shell structure of the CeO₂@SiO₂@RF nanospheres. Finally, the outer RF was annealed to the carbon layer and the interlayer of SiO₂ was etched to form CeO₂@void@C with a yolk-shell structure.

XRD was performed to determine the crystal structures of the as-synthesized materials. Cubic CeO₂ nanospheres were successfully synthesized according to the XRD patterns [Figure 2A], in which the peaks at 28.6°, 33.1°, 47.5° and 56.3° correspond to the (111), (200), (220) and (311) crystal faces, respectively. No position shifts were observed for pure CeO₂ or CeO₂@void@C, indicating that the coating treatment does not influence the lattice parameters of the CeO₂ core. However, all the peaks became sharper for the yolk-shell-structured CeO₂@void@C, which should be attributed to the annealing process at 600 °C, resulting in a better crystalline phase. Additionally, there is no diffraction peak detected for carbon, revealing that the coating layer is amorphous. The Raman spectra of CeO₂ and CeO₂@void@C were also characterized [Figure 2B]. The peak at 460 cm^-1 for the characterization of the F_2g vibration was detected for both CeO₂ and CeO₂@void@C^[27]. For the yolk-shell-structured CeO₂@void@C, two extra intensive broad peaks at 1338 and 1588 cm^-1 were observed, which could be assigned to the D band (disorder induction) and G band (graphite), respectively. The intensity ratio of the D and G bands (I_D/I_G) was referenced to evaluate the graphitization degree^[5]. The I_D/I_G ratio was calculated as ~0.844 for CeO₂@void@C, which reveals good graphitization and electronic conductivity, resulting in faster cycling rate performance of CeO₂@void@C when compared with the untreated CeO₂.

Figure 2. (A) X-ray diffraction patterns of untreated CeO₂ and CeO₂@void@C. (B) Raman spectra of CeO₂ and CeO₂@void@C.

The morphologies of CeO₂ and CeO₂@void@C are displayed in Figure 3 and are homogeneously distributed in size. As shown in Figure 3A and C, the particle diameters of both CeO₂ and CeO₂@void@C are ~100 and ~220 nm, respectively. The tapped density of the anode has a certain impact on the performance and specific capacity and depends on the intrinsic density and particle size distribution. A higher intrinsic density and narrower particle size distribution result in a higher tapped density. Although there is a certain space between the core and shell, the high density core and homogeneous particle size ensure good tapped density. The good tapped capacity could be further improved by optimizing the void volume between the CeO₂ core and the carbon shell. The etching and annealing process does not change the size of the carbon coating layer when compared to Figure 3B and C. The high-resolution TEM images of pure CeO₂ are shown in Figure 3E.

Figure 3. Scanning electron microscopy images of (A) CeO₂, (B) CeO₂@SiO₂@RF and (C) CeO₂@void@C. High-resolution transmission electron microscopy images of (D, E) CeO₂ and (G, H) CeO₂@void@C. EDS elemental maps of (F) CeO₂ and (I) CeO₂@void@C.

The lattice fringe spacing of CeO₂ is in good agreement with the XRD patterns [Figure 2]. The energy dispersive X-ray spectrum elemental mapping images indicate that Ce and O are homogenous for CeO₂ [Figure 3F]. Obviously, there is no signal for carbon in the CeO₂ sample. With regards to the CeO₂@void@C sample, the carbon coating layer is determined as being ~20 nm thick, as shown in Figure 3G. An obvious void space is found between the CeO₂ core and the carbon layer [Figure 3C and G]. Due to the shielding of the carbon layer, the lattice fringe is not clear for the coated CeO₂@void@C [Figure 3H]. The elemental mappings show that both Ce and O are localized within the CeO₂ core [Figure 3I], while carbon is distributed at the outer layer. To check the stability of the structure, cycled samples were also measured. CeO₂@void@C maintains good morphology after cycling [Supplementary Figure 1]. However, the CeO₂ nanoparticles without the carbon shell agglomerated badly after 100 cycling scans [Supplementary Figure 2].

The surface area and pore size distribution were determined by nitrogen adsorption-desorption isotherms. The specific surface of the untreated CeO₂ [Figure 4A] is only 26 m²·g^-1 and was obtained by the Brunauer-Emmett-Teller method. In comparison, the specific surface area increases profoundly to 166 m²·g^-1[Figure 4B] for CeO₂@void@C, which obviously exhibits typical type IV isotherms for the mesoporous structures of the carbon layer. The higher specific surface of CeO₂@void@C means more active sites for lithium accommodation, thereby offering a larger capacity as a LIB anode. Moreover, the pore size of CeO₂@void@C was calculated via the Barrett-Joyner-Halenda method [Figure 4C], which shows a narrow pore size distribution centered at ~7.6 nm. TGA was also carried out to check the stability of the obtained material, as shown in Figure 4D. CeO₂@void@C shows good thermal stability below 370 °C but experiences a large mass loss between 370 and 500 °C due to the decomposition of the carbon layer. The carbon content is calculated as ~28 wt.% in the total yolk-shell-structured CeO₂@void@C.

Figure 4. N₂ adsorption-desorption isotherms of as-prepared CeO₂ (A) and CeO₂@void@C (B). (C) Pore size distribution of CeO₂@void@C. (D) Thermogravimetric analysis curve of CeO₂@void@C.

XPS was carried out to investigate the surficial elements and valence states of CeO₂@void@C. All signals of Ce, O and C appear in the full spectrum [Figure 5A]. For the CeO₂@void@C [Figure 5B], the XPS spectrum of Ce 3d included five peaks of Ce 3d3/2 and three peaks of Ce 3d5/2^[41]. Based on the signal integration of Ce 3d5/2, the peaks of Ce⁴⁺ are located at 879.8 and 886.5 eV, while one peak of Ce³⁺ is located at 883.0 eV. The corresponding areas were integrated and the ratio of Ce³⁺ and Ce⁴⁺ is 31:69, which could be approximated as the atomic fractions in redox states. The Ce 4d peak in the full survey also belongs to Ce₂O₃, proving the existence of Ce³⁺ in CeO₂. Therefore, the redox state of Ce in the as-prepared CeO₂@void@C nanospheres is both tetravalent and trivalent.

Figure 5. (A) Full X-ray photoelectron spectroscopy (XPS) spectrum of CeO₂@void@C. (B) High-resolution XPS spectrum and simulation of Ce 3d for CeO₂@void@C.

The electrochemical properties were fully investigated for the in-depth study of the potential application and mechanism of CeO₂@void@C. Figure 6A shows the cyclic voltammetry (CV) analysis of CeO₂@void@C at a scan rate of 0.1 mV·s^-1 in the voltage range of 0.01 to 3.00 V. In the first cathodic polarization process, two broad cathodic peaks were observed. The first cycle is partially irreversible because of side reactions, such as electrolyte decomposition and the formation of a solid electrolyte interface (SEI) film. The cathodic peak (1.24 V) was observed only in the first cycle, which can be attributed to the initial insertion of Li ions and the breakdown of electrolytes. After the first cycle, the following cathodic peak located between 0.8 and 0.1 V may be ascribed to the reductive transformation of CeO₂ to Ce₂O₃ and Li₂O according to the conversion reaction. The CV curves quickly become stable and the discharge-charge curves almost overlap with each other, indicating that the electrochemical reactions are highly reversible for CeO₂@void@C.

Figure 6. (A) Cyclic voltammetry of CeO₂@void@C at a scan rate of 0.1 mV·s^-1 at room temperature. (B) Cycling performance of CeO₂@void@C. (C) Rate performance of untreated CeO₂ and CeO₂@void@C under different applied current densities, whose values are marked in columns in mA·g^-1. (D) Nyquist impedance curves for the untreated CeO₂ and CeO₂@void@C materials. (E) Long life cycling performance of CeO₂@void@C under a current density of 0.1 A·g^-1.

Figure 6B shows the first three electrochemical cycles of a Li//CeO₂@void@C battery at a current density of 100 mA·g^-1. The initial discharge and charge capacities were characterized as 615 and 321.13 mAh·g^-1, respectively, with a Coulombic efficiency of 51.38%. The initial irreversible capacity loss is mainly due to the electrolyte decomposition, the formation of an SEI film and the irreversible formation of Li₂O. During the second and third circles, the discharge and charge contours are almost identical, reflecting the excellent structural stability of the electrode. Figure 6C compares the rate performances of the untreated CeO₂ and the CeO₂@void@C electrodes. With increasing current density from 100 to 200, 500 and 1000 mA·g^-1, the Li//CeO₂@void@C battery delivered reversible discharge capacities of 313, 265, 227 and 196 mAh·g^-1_, respectively. For the Li//CeO₂ battery, the capacities are determined only as 104.8, 84.1, 61.2 and 50 mAh·g^-1 under the corresponding current densities. The capacity of the Li//CeO₂@void@C battery was maintained at ~300 mAh·g^-1 and 100 mAh·g^-1 was recovered for the Li//CeO₂ battery when the current density was restored to 100 mAh·g^-1. Along with all different applied currents, the capacities of CeO₂@void@C are almost three times those obtained for the untreated CeO₂.

EIS of CeO₂@void@C and untreated CeO₂ was performed to determine their resistance and diffusion behavior [Figure 6D]. The impedance data were fitted as an equivalent electrical circuit, which is composed of the solution resistance (Rs), the charge transfer resistance (Rct), the constant phase element and the Warburg impedance (Rw)^[42]. The analyzed impedance results are listed in Table 1. The values of Rct for the CeO₂@void@C anode and the untreated CeO₂ anode were calculated to be 232.5 and 464 Ohm, respectively. Obviously, faster kinetics in the electrochemical reactions exist for the CeO₂@void@C anode compared to the untreated CeO₂. The total resistance of CeO₂ is larger than that of the CeO₂@void@C as an electrode. The long-term cycling performance was also tested [Figure 6E] and the capacity gradually decreases from ~300 to ~200 mAh·g^-1 at a current density of 100 mAh·g^-1 after 1000 cycles. The initial 50 cycles are shown in Supplementary Figure 3 for better comparison. Surprisingly, CeO₂@void@C shows an excellent Coulombic efficiency of 100% due to the structural stability endowed by the buffer space.

To understand the electrochemical mechanism of the CeO₂@void@C anode, CV was carried out under different rates, as shown in Figure 7A. The storage mechanism is manipulated by two types of process, namely, diffusion-controlled behavior and pseudocapacitive storage, which can be estimated by the ratio of the square root of current and the scan rate as the factor of b values^[43,44]. The process is completely controlled by ion diffusion when b is 0.5, while the capacitance effect dominates when b is 1. As shown in Figure 7B, b values of 0.93 and 0.84 are determined for peaks 1 and 2, respectively, revealing a pseudocapacitive behavior for CeO₂@void@C. A pseudocapacitive contribution of 77.3% at a 0.8 mV·s^-1 scanning rate is determined for the CeO₂@void@C anode, as calculated in Figure 7C. By varying the scan rate [Figure 7D], the pseudocapacitive contributions are 54.3%, 60.1%, 68.6%, 70.5%, 73.4%, 75.9% and 77.3% under scanning rates of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7 and 0.8 mV·s^-1, respectively. Therefore, the pseudocapacitive process shows a significant role in the total capacity and its contribution grows with increasing scanning rate.

Figure 7. (A) Cyclic voltammetry curves of CeO₂@void@C electrode under different scan rates. (B) Fitting of the square root of the peak current and the scan rate (v^1/2). (C) Capacitive contribution of CeO₂@void@C to the total storage. (D) Contribution ratios of the capacitive and diffusion-controlled process under various scan rates.

Ex-situ XRD at the first cycle was carried out to investigate the component and crystal structure variation of CeO₂@void@C. Certain cycling states of the batteries were selected, as marked to the right of Figure 8. The XRD pattern of the pristine anode shows the pure phase of CeO₂. Upon discharge, the diffraction peak slightly shifted to a lower value of 2θ, which is consistent with the peak position trend of Ce₂O₃. Meanwhile, lithium ions are embedded in the lattice, making the lattice parameters larger. Thus, the partial formation of the Ce₂O₃ phase structure and cell volume expansion are confirmed during lithiation. In the subsequent charge process, the peaks inversely move back to larger 2θ values, which could be due to the extraction of Li ions, causing oxidation of Ce³⁺ to Ce⁴⁺ and cell shrinking. The peak positions of the fully charged sample are still slightly lower than those of the pristine material, which is attributed to the irreversible reaction and residual lithium ions. No new phase was detected in the whole charge/discharge process. Therefore, the electrochemical cycling of the CeO₂@void@C electrode could be considered as an intercalation/extraction reaction process.

NMR is very sensitive to the Li-ion local environment of battery materials^[45,46]. The solid-state ⁷Li NMR spectra of Li_xCeO₂@void@C electrodes at different cycling states are shown in Figure 9. As shown in Figure 9A, all the electrodes show wide spinning sidebands ranging from -600 to 600 ppm because of the anisotropy of Li, together with the possible hyperfine interaction between the unpaired electrons of Ce^3+/4+. The electrode discharged to 1.1 V (bottom spectrum in Figure 9A) presents a sharp symmetric peak at -0.4 ppm for isotropic resonance. With further lithiation to 0.4 V, the isotropic peak becomes broader with a shift to a low field. At the end of discharge to 0.01 V, an even broader signal is observed. Reversible evolution is observed when delithiation takes place.

Figure 8. Ex-situ XRD of CeO₂@void@C at various cycling states. The corresponding electrochemical curve is displayed to the right. For better comparison, the indexes of both CeO₂ and Ce₂O₃ are also plotted for reference.

Figure 9. Solid-state ⁷Li NMR spectra of cycled CeO₂@void@C electrodes at different states. (A) Full spectra with the spinning sidebands marked with asterisks. The corresponding longitudinal relaxation time T₁ is noted to each spectrum. (B) The simulation region of the isotropic signal is marked by the rectangular dotted frame, as shown in (A). Blue signals are obtained from experiment and red-dashed curves are the sum of deconvolution. The chemical shifts and occupancies are marked for each simulated peak.

For a clearer discussion, the spectra are simulated and the deconvoluted spectra of the isotropic resonances are presented in Figure 9B. For the electrode discharged to 1.1 V, only one peak at -0.4 ppm is simulated, which is possibly originated from the SEI components, such as LiF and Li₂CO₃^[47]. When the anode is further discharged to 0.4 V, an additional peak at 2 ppm with 41% occupancy appears in addition to the first peak at -0.4 ppm, which accounts for 59%. The signal at 2 ppm is mainly assigned to the product Li₂O. Deeper lithiation to 0.01 V produces an extra signal at even downfield to 8 ppm, with an occupancy of 32%. The change of Fermi contact along the Ce-O-Li bond is the main reason for the spectral shifts, as the electron cloud varies due to the reduction of Ce⁴⁺ to Ce³⁺ upon discharge, or different sites/vacancies are occupied by the intercalated Li⁺. Upon subsequent charge, a reverse trend is observed. When the electrode is charged to 1 V, the signals at 8 and 2 ppm decrease to 7% and 37% from 32% and 47%, respectively. When the battery is further charged to 3 V, the signal at 8 ppm disappeared completely and 10% is still residual for the signal at 2 ppm caused by the irreversible reaction. Furthermore, the relaxation time T₁ decreases upon discharge due to the reduction of Ce⁴⁺ to Ce³⁺, with the latter displaying stronger paramagnetization. This is regained upon the following charge because of reverse evolution. The NMR results are in good agreement with the above-mentioned XRD study.

CONCLUSIONS

In summary, yolk-shell-nanostructured CeO₂@void@C was designed and its electrochemical properties were fully investigated as an anode for LIBs. Both good electronic conductivity and enhanced capacity were achieved according to the hollow carbon shell composite structure of CeO₂@void@C. Furthermore, the excellent cycling stability was further confirmed by the fast charge transfer and long-term cycling test. The structure-property correlation proposes a promising strategy for fabricating nanosized CeO₂ cores and carbon shells with adjustable voids, which show enhanced lithium storage properties. XRD and NMR analysis reveals that the intercalation process dominates the lithiation/delithiation reaction for the CeO₂@void@C anode, which is consistent with the pseudocapacitance storage mechanism for CeO₂@void@C. This work may help to design and fabricate hybrid composites for energy storage applications.