Ultrafast lithium-ion batteries (LIBs) are believed to significantly trigger the extensive applications of pure electric buses, taxis, 24 h mobile tools, and so forth[1–3]. The pivotal goal is to exploit practicable anode materials that demand rapid Li⁺ insertion/extraction and high reversibility because this process easily causes severe structural damage and abundant ‘‘dead” Li⁺[4–5]. The rapid charge/discharge performance is chiefly related to the Li⁺ diffusion coefficient [6–7]. The value is about 10^–6 cm² ·s^–1 for Sn[8–10], nearly 1000 times higher than that of Si and graphite[11,12]. Moreover, the Sn–Li alloy can be formed in any proportion with a much lower formation energy (E_f) of –0.573 eV (1 eV = 1.602176 × 10^–19 J) compared to Si and Ge (+0.401 and –0.285 eV, respectively) [13], making it widely considered to date. Among them, SnP shows a high measured capacity of 815 mA·h·g^–1 with the Li⁺ diffusion channel perpendicular to the c-axis [14], and the resultant Li₃P is also a high Li⁺ conductor ( ≈ 3 × 10^–3 cm² ·s^–1 ) [15], which makes it a viable choice for ultrafast LIBs.

The lithium storage mechanism of Sn-based anode materials generally involves the first conversion reaction (SnA_x + Li⁺ + e^– →Li_yA + Sn) and the subsequent alloying (Sn + Li⁺ + e^– →Li_zSn)[16,17]. In the Li⁺ extraction process, it has been found that the atomic level contact of Li_xA and Sn can greatly improve the reversibility of the conversion reaction step, which requires a particle size of less than 10 nm[18,19]. However, high-surface-energy nanocrystals have a strong tendency to coarsen and agglomerate when cycling with fast capacity fading[20,21]. It is therefore critical to stabilize the nanocrystals in the electrochemical process. Furthermore, thermodynamic metastable SnP usually has a quite high formation energy[22,23], making controlled synthesis very difficult. Vacuum quenching and high-temperature molten salt methods are dominantly utilized to synthesize SnP with particle sizes of several hundred nanometers[24,25], resulting in extremely poor cycle stability. For example, 500 nm SnP particles give a capacity attenuation of 10% after only 40 cycles at 0.12 A·g^–1[14]. Recently, nonstoichiometric SnP_x-based hybrids have been reported to have attractive electrochemical performance[26–29]. Typically, tens of nanometers of Sn₄P₃ in the carbon framework can afford more than 100 cycles, which stimulates SnP investigation[30,31]. However, a major challenge is to develop a simple and low-temperature synthesis tactic to controllably synthesize SnP with rapid electrochemical reaction kinetics.

Herein, we first demonstrate an atomic interface catalytic phosphorization strategy to confine SnP/CoP hetero-nanocrystals within reduced graphene oxide (rGO)-coated carbon cubes (denoted as TCP@C/rGO). The density functional theory (DFT) calculations disclose that the relative formation energy of SnP is significantly decreased from +0.12 to –2.05 eV when the phosphate of tin oxide is conducted on the CoP surface. More importantly, the CoP catalyst is also a conversion-type anode material with a lower lithium intercalation potential than SnP. The prior as-formed Co can accelerate the subsequent Li⁺ insertion kinetics of SnP with rapid electron transfer, endowing it with ultrafast charge/discharge capability. A reversible capacity of 260 mA·h·g^–1 is achieved even at 50 A·g^–1 (about 19 s). Furthermore, the confinement of hetero-nanocrystals within carbon frameworks effectively refrains their coarsening, exhibiting superior cycling stability almost without capacity decay (645 mA·h·g^–1 ) after 1500 cycles at 2.0 A·g^–1 .

All reagents and solvents were of analytical grade and were used without any further purification. In the typical process, 0.35 g of SnCl₄·5H₂O (Sinopharm Chemical Reagent Beijing Co., Ltd., China), 0.238 g of CoCl₂·6H₂O (Sinopharm Chemical Reagent Beijing Co., Ltd.), and 1.56 g of sodium citrate (Sinopharm Chemical Reagent Beijing Co., Ltd.) were first dissolved in 15 mL deionized water. Then, 5 mL aqueous solution of NaOH (2 mol·L^–1 ) (Shanghai Aladdin Biochemical Technology Co., Ltd., China) was added under magnetic stirring. After 1 h, 20 mL aqueous solution of NaOH (8 mol·L^–1 ) was added and stirred for another 15 min. The precipitates were collected and washed with water to obtain CoSn(OH)₆–Cit cubes.

Typically, 80 mg of the as-prepared CoSn(OH)₆–Cit cubes and 6.2 mg of graphene oxide (GO) were ultrasonically dispersed in 5 mL H₂O with stirring for 2 h. The brown suspension was frozen in liquid nitrogen and freeze-dried to obtain the CoSn(OH)₆– Cit/GO precursor. Five grams of NaH₂PO₂·H₂O (Sinopharm Chemical Reagent Beijing Co., Ltd.) and 50 mg of the CoSn(OH)₆– Cit/GO precursor were placed at two separate positions in a tube furnace, in which NaH₂PO₂·H₂O was upstream and CoSn(OH)₆– Cit/GO was downstream. The TCP@C/rGO hybrids were obtained by annealing the samples at 350 °C for 30 min with a heating rate of 5 °C·min^–1 under a 100 standard cubic centimeters per minute (sccm) Ar flow.

The typical design concept and fabrication route of the TCP@C/ rGO hybrids is illustrated in Fig. 1. CoSn(OH)₆–Cit cubes are first synthesized by coprecipitating Co²⁺, Sn⁴⁺, and OH^– in the presence of citrate anions. In the subsequent etching process under a high pH environment, a surface stable oxide layer can be readily generated while the as-formed [Co(OH)₄]^2– and [Sn(OH)₆]^2– in the core gradually dissolve, accompanied by the adsorption of citrate anions. To verify this point, a similar method was applied to synthesize pure CoSn(OH)₆ cubes (Fig. S1 in Appendix A) simply without adding citrates. The zeta potential is +35.5 mV, which decreases to +13.2 mV for the CoSn(OH)₆–Cit cubes, as we predicted (Fig. S2 in Appendix A). Positively charged CoSn(OH)₆–Cit hollow cubes are then obtained, which are easily and homogeneously dispersed on the negatively charged graphene oxide surface (–14.0 mV) with the assistance of electrostatic attraction. The traditional direct phosphorization approach easily converts Sn-based compounds into metallic Sn rather than SnP[32,33]. This fact has been further confirmed by phosphating the reported hollow SnO₂ spheres at 350–500 °C [34], indeed forming Sn nanoparticles with high purity (Fig. S3 in Appendix A). In this work, CoSn(OH)₆–Cit/GO hybrids were successfully generated into the corresponding SnP/CoP@C/rGO hybrids only at 350 °C under a PH₃/Ar atmosphere. The ex-situ X-ray diffraction (XRD) patterns and ex-situ Raman spectra are provided in Appendix A Fig. S4 to disclose the specific conversion process. We found that CoP is priority formed when the temperature increases to 350 °C. After retaining it from 5 to 30 min, the SnP component is gradually generated along with the disappearance of SnO_x. These observations indicate that CoP can catalytically phosphate Sn oxides with rapid conversion of SnP. SnP phase tin phosphide is chosen for its unique Li⁺ diffusion channels and layered crystal structure, which make it easier to transport and store Li⁺ than other crystal phases [14].

Fig. 1. Schematic illustration of the design and synthesis of SnP/CoP hetero-nanocrystals within rGO-coated carbon cubes by a catalytic phosphorization strategy.

To further unveil the CoP-catalyzed mechanism, DFT calculations were applied to obtain the relative free energies at different reaction states during the phosphorylation of Sn oxides with and without CoP. When directly phosphating SnO₂ (Fig. 2(a)), we can find that metallic Sn is very easy to generate because it requires a much lower relative free energy of –2.57 eV than SnP (+0.12 eV). The specific formation pathways are illustrated in Appendix A Fig. S5. When introducing the CoP component, totally opposite phenomena are observed at the CoP/SnO₂ heterointerface. As shown in Fig. 2(b), CoP can dramatically reduce the energy barrier and accelerate the formation of SnP with a quite low relative free energy of –2.05 eV compared with metallic Sn (–0.22 eV). The specific formation pathways are also given in Appendix A Fig. S6. It is noted that metallic Sn has difficulty phosphate into SnP even in the presence of CoP. There were no component changes when we phosphate the mixture of CoP and Sn nanoparticles under similar conditions (Fig. S7 in Appendix A). These observations are highly consistent with the aforementioned experimental results.

Fig. 2. The relative free energy versus reaction states of (a) direct phosphorization and (b) CoP-catalyzed phosphorization of SnO₂; (c) high-magnification (the inset showing the corresponding selected area electron diffraction (SAED) pattern); (d–f) high-resolution transmission electron microscope (TEM) images of the TCP@C/rGO hybrids; (g) VBXPS spectra; (h) XPS Co 2p spectra of the samples.

The detailed structural characterizations of the TCP@C/rGO hybrids are shown in Fig. 2(c). Hollow TCP@C cubes with lengths of about 200 nm are well covered by rGO nanosheets. The same observation is also obtained in the low-magnification scanning electron microscope (SEM; S-4800, Hitachi Ltd., Japan) image (Fig. S8 in Appendix A). We can only see the SnP and CoP electron diffraction patterns in the corresponding inset. Further magnifying the wall of the TCP@C cubes, we can observe the SnP/CoP heteronanocrystals in Fig. 2(d) due to the CoP-catalyzed phosphorization strategy, which are embedded into citrate-derived carbon. The typical staggered arrangement of the SnP and CoP nanocrystals indicates that CoP can effectively prevent the migration and agglomeration of Sn during the electrochemical process. Both of them are marked with white (SnP) and yellow (CoP) dashed circles, respectively. Lattice distances of 0.30 and 0.32 nm are assigned to the (002) and (011) planes of SnP in Fig. 2(e), while 0.38 and 0.28 nm are the (101) and (002) planes of CoP in Fig. 2(f). The SnP/CoP heterostructure features are also verified by valanceband X-ray photoelectron spectroscopy (VB-XPS; AXIS Ultra DLD, Kratos Empire Ltd., China), as shown in Fig. 2(g). The as-obtained TCP@C/rGO hybrids give a valence-band maximum (VBM) value of 1.48 eV, between –0.22 eV for pure CoP and 2.25 eV for pure SnP because of the valence band curvature. The SnP–CoP@C/rGO hybrids show VBM values of –0.22 eV because heteronanocrystals are not formed. A similar phenomenon can also be observed in their ultraviolet–visible (UV–vis; UV-2600, Shimadzu Co., Japan) spectra (Fig. S9 in Appendix A), indicating a strong interaction between them. An in-depth analysis of the electronic structure was implemented by XPS (AXIS Ultra DLD). The Co 2p spectra in Fig. 2(h) show that an obvious blueshift of about 0.8 eV for Co–P peaks can be observed in the TCP@C/rGO hybrids compared to the SnP/CoP@C/rGO hybrids, indicating the formation of Co–P–Sn bonds at the hetero-nanocrystal interfaces [35]. The corresponding P 2p spectra further verify this point (Fig. S10 in Appendix A). A detailed discussion is provided in the Appendix A.

The atomic ratio of Sn, Co, and P is 1:1:2 for the TCP@C/rGO hybrids with a weight ratio of 79.7% based on the inductively coupled plasma mass spectrometry (ICP–MS ELAN DRC-e, PerkinElmer, Inc., USA) results. We then completely removed the SnP/CoP hetero-nanocrystals by acid etching. As shown in Appendix A Fig. S11, many mesopores are left in the resultant carbon frameworks, and a strong peak at 4.0 nm appears in the pore size distribution curve, further clarifying the embedded nanostructures and indicating that the grain coarsening of the nanocrystals during charge/discharge cycles can be effectively inhibited. Based on the above discussion, Fig. S12 in Appendix A shows the formation schematic illustration of SnP for direct phosphorization and CoPcatalyzed phosphorization of SnO₂. The interface-catalyzed phosphorization at the atomic level determines the total conversion of SnO₂ into SnP. If phosphorization occurs at the particle level, the main component is metallic Sn with few SnP in their interface, which is confirmed in the experiment (Fig. S13 in Appendix A).

Analogous hybrids with SnP/CoP hetero-nanocrystals sized about 20 and 50 nm were also synthesized as controls just by reannealing the TCP@C/rGO hybrids for the prescribed time, denoted TCP-1@C/rGO and TCP-2@C/rGO, respectively (Fig. S14 in Appendix A). The three samples are evaluated by charge–discharge measurements. The initial coulombic efficiency of the TCP@C/rGO hybrids was 74.1%. The average reversible capacities for five cycles at 0.5 A·g^–1 are illustrated in Fig. 3(a). The capacity shows rapid attenuation as the size of the SnP nanoparticles increases. The targeted TCP@C/rGO hybrids show the highest capacity of 907 mA·h·g^–1 at 0.5 A·g^–1 compared to the two controls. More impressively, it possesses ultrafast charge–discharge capability and ultrastable cycle performance. Even at 50 A·g^–1 (charging for 19 s), 260 mA·h·g^–1 is still retained. After 1500 continuous cycles at 2.0 A·g^–1 , the capacity is always 645 mA·h·g^–1 without degradation, as shown in Fig. 3(b). Nevertheless, the TCP-1@C/rGO hybrids show a rapid capacity fading from 522 to 247 mA·h·g^–1 only after 300 cycles. The TCP-2@C/rGO hybrids exhibit a 21.1% capacitance retention only after 200 cycles with poor reversibility, which is unnecessary in the latter discussion. The rate and cycling performances imply that the SnP particle size has significant effects on the lithium storage ability. Our TCP@C/rGO hybrids can also achieve 82% capacity retention after 500 cycles, even at 10.0 A·g^–1 (Fig. S15 in Appendix A). Such superior lithium storage performances have been compared with some typical Sn-based anode materials by choosing several key indices, including reversible specific capacity, charging/discharging rates and cycle life at various current densities[36–41], as shown in Fig. 3(c). The targeted TCP@C/rGO hybrids exhibit obviously enhanced LIB performance.

Fig. 3. (a) The reversible specific capacity at 0.5–50.0 A·g^–1 and (b) cycling stability at 2.0 A·g^–1 of the targeted TCP@C/rGO hybrids with different SnP/CoP sizes; (c) comprehensive comparison of the reported Sn-based anode materials with the targeted TCP@C/rGO hybrids; (d) cyclic voltammetry (CV) curves at 3.0 mV·s^–1 from the 1st to 100th cycle; (e) galvanostatic intermittent titration technique (GITT) curves of the targeted TCP@C/rGO hybrids before and after 100 cycles (inset showing the TEM image after 100 cycles). CE: coulombic efficiency; C₁₀: capacity at 10 A·g^–1 ; C₁: capacity at 1 A·g^–1 .

Fig. 3(d) shows the cyclic voltammetry (CV) curves of the TCP@C/rGO hybrids during 100 cycles at 3.0 mV·s^–1 . The welloverlapped CV curves and the slight shift of the redox peak to low potential imply a gradually decreased electrochemical polarization and high charging/discharging reversibility. In contrast, the redox peak decreases obviously for the TCP-1@C/rGO hybrids during 100 cycles (Fig. S16 in Appendix A), indicating relatively poor reversibility when increasing the SnP/CoP nanoparticles to 20 nm. The electrochemical impedance spectroscopies (EISs) of the above two samples at the 3rd, 50th, and 100th cycles are also provided in Fig. S17 and Table S1 in Appendix A. The charge transfer resistance (R_ct) of the TCP@C/rGO hybrids decreases from 21.7 to 5.2 Ω, while the TCP-1@C/rGO hybrids show an increase from 95.7 to 205.6 Ω after 100 cycles. Furthermore, the Li⁺ diffusion coefficient () of the TCP@C/rGO hybrids reaches 5.0 × 10⁹ cm² ·s^–1 , which is almost constant before and after 100 cycles, as shown in Fig. 3(e). The corresponding inset gives the transmission electron microscope (TEM) image after cycling. The about 4 nm SnP/CoP particles are maintained and well distributed in the carbon framework. In the TCP-1@C/rGO hybrids with about 20 nm SnP/CoP particles, the value of is only 1.0 × 10⁹ cm² ·s^–1 at the 3rd cycle and 10¹¹ cm² ·s^–1 at the 100th cycle (Fig. S18 in Appendix A). These results suggest that the particle size has significant effects on the electrochemical performance, which are further verified by studying the lithium storage behaviors of the two samples after the 3rd and 100th cycles. As shown in Appendix A Fig. S19, the capacitive dominated contribution of the TCP@C/rGO hybrids gives a slight decrease from 76.8% at the 3rd cycle to 69.7% at the 100th cycle. However, the TCP-1@C/ rGO hybrids exhibit a sharp drop of 40% after 100 cycles. This analysis indicates that the ultrasmall SnP/CoP nanocrystals contribute much more capacitive capacity, therefore endowing a high specific capacity even at ultrafast charge/discharge rates and superior cycling stability. 

Fig. 4(a) provides the initial three CV curves of the targeted TCP@C/rGO hybrids at 0.2 mV·s^–1 within 0.01–3.00 V. In the first lithiation process, the peak at 1.5 V is assigned to the conversion of CoP and SnP into the corresponding metal as well as the formation of the solid electrolyte interface (SEI) layer, and the peak at 0.01 V belongs to the alloying of Sn into Li_xSn. The following charge and discharge curves highly overlap. The redox reactions of CoP occur in the red area, where the peaks at 1.8 and 2.5 V correspond to the reversible conversion reactions of CoP + xLi Co + Li_xP. The next yellow area is the SnP lithiation and delithiation process. It also shows high reversibility due to the conversion into metallic Co of CoP, which further accelerates the electron transfer during lithiation, as shown in the schematic of the inset. In contrast, the CV curves of the TCP-1@C/rGO hybrids decrease obviously without the redox peaks of CoP reversible reactions, as shown in Fig. 4(b). These phenomena imply that the crystal size has significant effects on the lithiation reactions. To further clarify their reaction mechanism, the discharge/charge profiles and the corresponding ex-situ XPS results of the TCP@C/rGO hybrids are provided in Appendix A Fig. S20. The electrochemical process of the TCP@C/rGO hybrids is summarized as CoP + xLi⁺ + xe^– Co + Li_xP, SnP + yLi⁺ + ye^– Sn + Li_yP, and Sn + zLi⁺ + ze^– Li_zSn. Specifically, in the first discharge process, the Co–P and Sn–P peaks totally disappear at 1.0 V along with the appearance of the corresponding Co⁰ and Sn⁰ peaks. After further discharging to 0.01 V, the peak of Sn⁰ shifts to a higher binding energy of Sn–Li. In the following charge process, the peak of Sn–Li disappears with the appearance of Sn⁰ at 0.8 V, which then fully turns into the Sn–P bond at 1.7 V. When charging to 3.0 V, the Co–P bonds reappear without the Co⁰ peak. In the second discharge process, the chemical bond change is almost the same as that in the first discharge process, indicating high electrochemical reversibility. The coupling effects among them are further verified by the reaction resistance calculated by the galvanostatic intermittent titration technique (GITT) results during the discharge process (Fig. S21 in Appendix A)[42,43]. The TCP@C/rGO hybrids exhibit two obvious decreased reaction resistances at about 2.0 and 0.6 V, assigned to the generation of Co and Sn, respectively. By comparison, the TCP-1@C/rGO hybrids show only one reduced reaction resistance at about 0.6 V, agreeing with the aforementioned observations. Fig. S22 and Table S2 in Appendix A give the EIS curves when discharging from 3.0 to 1.0 V. The R_ct of the TCP@C/rGO electrode is significantly reduced from 21.7 to 8.3 Ω, also indicating the greatly enhanced charge transfer enabled by the generated metallic Co.

Electric field finite element simulations are applied to extensively analyze the current density distributions on different crystal sizes of the heterostructured nanocrystals within rGO-coated carbon cube hybrids. A circuit is first built, and the two dimensional (2D) section model of the SnP/CoP hetero-nanocrystals distribution is also given in Fig. 4(c). The relative current density distributions of the 4 and 20 nm SnP/CoP heterocrystals in the carbon frameworks are shown in Fig. 4(d). The current density of the conductive carbon is approximately 1000 times higher than that of the SnP/ CoP heterocrystals, hence showing blue for SnP/CoP in lowmagnification mapping. After magnification, almost the same current distribution can be observed for SnP and CoP, chiefly due to their similar resistance values. Obviously, the current density of the crystals with sizes of 4 nm is much higher than that of the corresponding 20 nm crystals. We also simulated the current density distributions of crystals of other sizes (Fig. S23 in Appendix A), and the average current density values at the crystal centers (J_c) are shown in Fig. 4(e). A linear relationship is built when the crystal sizes are in the range of 4–20 nm. However, there is no large current change (< 4%) even for 2 nm crystals. The dynamic relative current density distributions were further investigated under 1.0 V based on the conversion degree (D_c) of CoP into Co during the discharging process (Fig. S24 in Appendix A). Fig. 4(f) gives the current density distributions of the initial (D_c = 0) and final states (D_c = 100%) for SnP crystals, indicating that the full conversion of CoP can remarkably increase the current density. The coefficient of variation (δ) is then introduced to evaluate the uniformity of the current density distribution. The relationship of the average current density in SnP crystals (J_a) and δ with D_c is provided in Fig. 4(g). As D_c increases from 0 to 100%, the value of J_a increases by 4.4 times from 52 to 229 A·cm^–2 . In the process of dynamic conversion of CoP into Co with D_c = 25%–87%, the J_a value of SnP adjacent to Co always reaches 200 A·cm^–2 , while there is almost no change for SnP adjacent to CoP, which results in a relatively high δ value. This value then drops to 47.6% with D_c = 100%, indicating that the high reversibility of CoP conversion is good at improving the current distribution in SnP. These results are highly consistent with the above experimental results and related analysis.

Fig. 4. The initial three CV curves at 0.2 mV·s^–1 of (a) the TCP@C/rGO hybrids and (b) the TCP-1@C/rGO hybrids; (c) schematic of a TCP@C cube in circuit and its 2D section for finite element simulation; (d) the relative current density distribution and (e) the average current density at crystal centers of the simulated models with different crystal sizes; (f) the relative current density distribution and (g) the average current density and the coefficient of variation in SnP crystals of the simulated models with different CoP conversion degrees.

In summary, this work demonstrates a low-temperature CoPcatalytic phosphorization strategy for confining SnP/CoP heteronanocrystals within rGO-coated carbon frameworks as anode materials for ultrafast LIBs with superior cycling stability (645 mA·h·g^–1 after 1500 cycles at 2 A·g^–1 ) and ultrahigh rate capability (260 mA·h·g^–1 at 50 A·g^–1 ). The DFT calculations disclose that the relative formation energy of SnP is significantly decreased when tin oxide is phosphorylated on the CoP surface. The ex-situ XPS analysis and CV curves clarify the lithiation/delithiation mechanism, discovering that the electrochemical reversibility of SnP/ CoP hetero-nanocrystals is closely related to their sizes (< 4 nm) and that the previously formed metallic Co of conversion-type CoP at relatively high potential can further accelerate the subsequent Li⁺ insertion kinetics of SnP, hence endowing it with ultrafast charge/discharge performances. The dynamic relative current density distributions of SnP have been further investigated by finite element analysis, indicating that the current density remarkably increases with the D_c of CoP into metallic Co. This finding offers deep insights to stimulate the rapid developments in Sn-based anode materials for ultrafast LIBs.

This work was supported by the National Natural Science Foundation of China (21975074, 21838003, and 91834301), the Social Development Program of Shanghai (17DZ1200900), the Shanghai Scientific and Technological Innovation Project (18JC1410500), and the Fundamental Research Funds for the Central Universities (222201718002).

Chen Hu, Yanjie Hu, Aiping Chen, Xuezhi Duan, Hao Jiang, and Chunzhong Li declare that they have no conflict of interest or financial conflicts to disclose.

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2021.11.026.