Ultrafine ferroferric oxide nanoparticles embedded into mesoporous carbon nanotubes for lithium ion batteries

An effective one-pot hydrothermal method for in situ filling of multi-wall carbon nanotubes (CNT, diameter of 20–40 nm, length of 30–100 μm) with ultrafine ferroferric oxide (Fe3O4) nanoparticles (8–10 nm) has been demonstrated. The synthesized Fe3O4@CNT exhibited a mesoporous texture with a specific surface area of 109.4 m2 g−1. The loading of CNT, in terms of the weight ratio of Fe3O4 nanoparticles, can reach as high as 66.5 wt%. Compared to the conventional method of using a Al2O3 membrane as template to fill CNT with iron oxides nanoparticles, our strategy is facile, effective, low cost and easy to scale up to large scale production (~1.42 g per one-pot). When evaluated for lithium storage at 1.0 C (1 C = 928 mA g−1), the mesoporous Fe3O4@CNT can retain at 358.9 mAh g−1 after 60 cycles. Even when cycled at high rate of 20 C, high capacity of 275.2 mAh g−1 could still be achieved. At high rate (10 C) and long life cycling (500 cycles), the cells still exhibit a good capacity of 137.5 mAhg−1.

. Schematic illustration of the formation process of Fe 3 O 4 @CNT: (I) oxidizing multi-wall CNT into mesoporous structure using mixed acids; (II) in situ filling mesoporous CNT with ultra-small Fe 3 O 4 by onepot hydrothermal treatment.
Scientific RepoRts | 5:17553 | DOI: 10.1038/srep17553 cetyltrimethyl ammonium bromide (CTAB)/organic solvent, nor templates, such as Al 2 O 3 membrane; (2) the employed CNT do not need to be cut and opened before filling, which makes it very easy to obtain large scale CNT with uniform morphology and structure, and the prepared CNT filled with Fe 3 O 4 nanoparticles possess the same length-diameter ratio with that of original CNT; (3) the lack of diameter confinement of the CNT (e.g., the method of using Al 2 O 3 membrane often needs larger diameter of CNT (> 50 nm) 39,40 . means the CNT can be filled with more Fe 3 O 4 nanoparticles regardless of the diameter of CNT; (4) the loading level of CNT, in terms of the weight ratio of Fe 3 O 4 nanoparticles filled in CNT, can reach 66.5 wt% which is significantly higher than the highest reported value (51.8 wt%) 38 . This large amount of ultrafine Fe 3 O 4 nanoparticles filled into the CNT backbone enhances the electrochemical reactivity and mechanical integrity of the electrode during the repeated charge/discharge process; (5) the prepared Fe 3 O 4 @CNT exhibited mesoporous properties. In theory, such hybrid structures are expected to greatly improve the electrochemical performance because of their unique structures, relatively high specific surface area and porosity. The highly flexible and conductive CNT backbone provide a three-dimensional network to facilitate the electron transfer, and to provide a large contact area for higher Li + diffusion between the electrode and electrolyte 41 . When evaluated for lithium storage capacity, the capacity of the prepared Fe 3 O 4 @CNT remained at 358.9 mAh g −1 after 60 cycles at a rate of 1.0 C. Even when cycled at high current rate of 20 C, acceptable capacity of 275.2 mAhg −1 was achieved. At high rate (10 C) and long life cycling (500 cycles), the cells still exhibited a similar capacity of 137.5 mAh g −1 , indicating the introduction of mesoporous carbon shell (multi-wall CNT) can greatly enhanced the electrochemical performance of Li storage.

Experimental Section
Synthesis of multi-wall CNT. The muti-wall CNT were mass produced by our previous procedure using a nano-agglomerate fluidized bed reactor method 42 . This procedure involves the design of catalyst, agglomeration control, the fluidization hydrodynamic process, and the large scale fabrication of CNT in an industrial reactor, and routine purification.
Oxidation of muti-wall CNT. In a typical procedure, 1.0 g multi-wall CNT were added in a mixed acid solution containing 10.0 mL HNO 3 and 30.0 mL H 2 SO 4 at room temperature. The mixture was stirred for 10.0 min, and then the mixture was heated to 80 °C for 20.0 min. After that, the mixture was naturally cooled to room temperature. Then, the products were filtered, washed by deionized water three times and finally dried in a freeze-drying apparatus for 24.0 h.

Synthesis of mesoporous Fe 3 O 4 @CNT.
Typically, 0.9 g oxidized CNT was mixed with ferric citrate and FeSO 4 ·7H 2 O solution in the mole ratio of 1:2. When the mixture was stirred for 5.0 min, 0.10 g vitamin C (Vc) was added. The effect of Vc was to inhibit the oxidation of Fe 2+ during the reaction process. The suspension was further mixed for 20 min. Then, the pH of the suspension was adjusted to 10.0 using a by NaOH solution (0.4 M). Then the mixed solution was transferred into a Teflon-lined stainless autoclave and heated to 180 °C for 20.0 h. After reaction, the obtained mesoporous Fe 3 O 4 @CNT was separated by a magnet, washed by deionized water five times and dried in a freeze-drying apparatus for 24.0 h.
Characterization. The composition of the synthesized mesoporous Fe 3 O 4 @CNT was determined with an X-ray powder diffractometer (XRD, Rigaku, Japan) using Cu K α radiation at 1.5418 Å at a scanning rate of 5° min −1 . Scanning electron microscopy (SEM, FEI-Sirion 200), transmission electron microscopy (TEM, JEM-2010), high-resolution transmission microscopy (HRTEM) and selected area electron diffraction (SAED) were used to observe the crystal structure and sizes. Atomic force microscopy (AFM, E-Sweep) was used to obtain three-dimensional images of the as-prepared mesoporous Fe 3 O 4 @CNT. The thermal analysis was determined by a thermogravimetric analyzer (Pyris 1 TGA, PerkinElmer, USA) under N 2 atmosphere and in air, respectively, at a heating rate of 10 °C min −1 from 20 °C to 900 °C. The specific surface area and pore distribution were analyzed by Brunauer-Emmett-Teller (BET) tests using ASAP 2020 (Micromeritics Instuments) analyzers. The magnetic property of the synthesized Fe 3 O 4 @ CNT was evaluated by a vibrating sample magnetometer (VSM, Lakeshore 736, USA). The Fourier transform infrared spectroscopy (FTIR) was done by a Perkin Elmer Paragon-1000 spectrometer.
Electrochemical evaluation of mesoporous Fe 3 O 4 @CNT for LIBs. The electrochemical experiments using the synthesized mesoporous Fe 3 O 4 @CNT were performed using CR2025 coin cells. The working electrode was prepared by mixing the active materials, acetylene black and polyvinylidene fluoride (PVDF) with weight ratios of 80:10:10 in N-methyl-2-pyrrolidone solvent. Coin cells were assembled in an argon-filled glovebox in the presence of an oxygen scavenger and sodium drying agent. The loading amount of active material is 0.82 mg cm −2 . The mesoporous Fe 3 O 4 @CNT composites act as the working electrode, metallic lithium as counter/reference electrode, 1.0 M LiPF 6 in ethylene carbonate, diethyl carbonate and ethylmethyl carbonate (EC/DMC/EMC, volume ratio of 1:1:1) electrolyte, and Celgard 2400 polypropylene as separator. The capacitance is calculated based on Fe 3 O 4 .

Results and Discussion
The crystallographic structure and phase composition of the synthesized Fe 3 O 4 @CNT are determined by X-ray powder diffraction (XRD), as shown in Figure S1. All of the diffraction peaks can be assigned to the inverse cubic spinel structure Fe 3 O 4 (JCPDS No. 19-629) and carbon nanotubes. Two peaks at 2θ = 26.5° and 40.6° can be attributed to the (002) and (110) planes of CNT, respectively. The diffraction peak at 40.6° confirms the multi-wall structure of carbon nanotubes. The broad peaks of Fe 3 O 4 indicated that the size of Fe 3 O 4 particles is very small. According to the (311) plane data and applying the Scherrer's diffraction equation (D = Kλ /β cosθ ), the average crystalline size is approximately 10.4 nm. The rather sharp diffraction peaks reveal that the Fe 3 O 4 nanoparticles have a relatively high crystallinity. The Brunauer-Emmett-Teller (BET) surface areas and porous structure of Fe 3 O 4 @CNT were investigated by nitrogen adsorption/desorption isotherms, as shown in Figure S2. The synthesized Fe 3 O 4 @CNT has type IV nitrogen adsorption and desorption isotherm 43 , suggesting the existence of mesoporous which may contribute to a relatively high BET surface area. Our results indicate that the BET surface area of the Fe 3 O 4 @CNT was 109.4 m 2 g −1 , which is much higher than that of the commercial Fe 3 O 4 (~2.0 m 2 g −1 ). Although large surface area is detrimental to Coulombic efficiency 44 , based on previous work, the surface area of our synthesized Fe 3 O 4 @CNT was in a reasonable range. The SnO 2 /CNT composites with a large surface area (180.3 m 2 g −1 ) have been demonstrated to be effective as anodes for lithium ion batteries 45 . The CNT/V 2 O 5 composites with a surface area (80 m 2 g −1 ) also exhibited high-performance for lithium ion electrodes 46 . The pore size distribution obtained through the Barrett-Joyner-Halenda (BJH) method indicates that most pores have the sizes of ~2.5 nm, which is attributed to the surface carbon layers of Fe 3 O 4 @CNT. Figure 2a shows a panoramic image of mesoporous Fe 3 O 4 @CNT by the SEM observation. It is evident that the samples contain numerous dispersed one-dimensional Fe 3 O 4 @CNT, and the length of mesoporous Fe 3 O 4 @CNT is 30-100 μ m. Figure 2b shows that the diameter of mesoporous Fe 3 O 4 @CNT varies from 20 to 40 nm. We also observed that a small number of Fe 3 O 4 nanoparticles decorated with the outer surface of CNT. The formation of mesoporous Fe 3 O 4 @CNT structure was further confirmed by TEM. Figure 2c illustrates the backbones of CNT are uniformly filled with Fe 3 O 4 nanoparticles, and the in situ filling process does not influence the morphology of CNT significantly. This means that the size of the filling Fe 3 O 4 nanoparticles was very small. A higher magnification in Fig. 2d shows that the   (Fig. 3f) were evidently observed. The C element (Fig. 3g) arising from Fe 3 O 4 @CNT was not evident because the samples were dispersed on the carbon grid, and the carbon substrate will weaken the relative contrast intensity of carbon element, which also leads to the EDS mapping area larger than that in the TEM image. From the EDS mapping results, the Fe 3 O 4 nanoparticles (Fe and O elements) were uniformly distributed in the CNT without strong agglomeration. Such uniform distribution of Fe 3 O 4 nanoparticles in the CNT was beneficial for improving the electrical conductivity, enhancing the cyclic stability and rate capability, and tolerating the volume expansion of Fe 3 O 4 nanoparticles during the lithium ion insertion/extraction process. The mesoporous structure of the Fe 3 O 4 @CNT facilitates the electrolyte penetratinge the carbon shell of CNT, promoting rapid lithium ion diffusion and resulting in the high accessibility of the active materials.
The magnetic properties of the synthesized Fe 3 O 4 @CNT reflect the filling level of Fe 3 O 4 nanoparticles in the CNT, which is related to electrochemical performance of lithium ion battery. Herein, the magnetic performance of the CNT, the oxidized CNT, and the Fe 3 O 4 @CNT was evaluated by a VSM analyzer at room temperature. Figure 4a presents the magnetic hysteresis curves for the CNT, oxidized CNT and the synthesized Fe 3 O 4 @CNT composites. As for paramagnetic materials, the magnetic moment of the entire crystallite tends to align with the magnetic field. As for superparamagnetic materials, the values of remanent magnetization and coercivity are all zero, and the two magnetic hysteresis curves overlap and go through the zero point, suggesting the synthesized CNT are paramagnetic materials. The saturation magnetization and remanent magnetization are 1.78 and 0.45 emu g −1 , respectively. The coercivity is found to be 251.7 Oe. The magnetic behavior of CNT arises from the use of metal catalysts during the catalytic growth. After the CNT were oxidized by the mixed acid solution, the saturation magnetization (0.48 emu g −1 ), remanent magnetization (0.08 emu g −1 ) and coercivity (197.9 Oe) all decreased. This was attributed to the partial dissolution of residual metal catalyst in the CNT. When the CNT were filled with ultra-small Fe 3 O 4 nanoparticles, the remanence and coercivity are negligible. This evolution shows that the synthesized mesoporous Fe 3 O 4 @CNT exhibit a superparamagnetic behaviour due to the presence of the ultra-small Fe 3 O 4 nanoparticles. The saturation magnetization is found to be 12.1 emu g −1 . After oxidation of CNT by the mixed acid solution, the surface of CNT contains many oxygen-containing active groups (e.g., − COO − , − OH and − C= O). These active groups may serve as the nucleation sites for the Fe 2+ in the reaction solution. The presence of Fe 2+ also may lead to the reduction of the oxygen-containing groups during the hydrothermal process which can be confirmed by FTIR spectroscopy, as shown in Fig. 4b. The FTIR spectra were carried out in the wavenumber range of 400-4000 cm −1 . The intense peak at 3446.2 cm −1 was attributed to the O-H stretching vibration. The peaks at 1633.8 cm −1 was assigned to the C= O stretching vibrations from carbonyl and carboxylic groups. The peak at 1396.1 cm −1 can be assigned to the skeletal vibrations of CNT 47 . The peak at 1112.5 cm −1 can be attributed to the C− O− C stretching vibrations 48 , and the peak at 611.4 cm −1 may be due to the residual metal catalyst in the CNT. When the oxidized CNT were filled with Fe 3 O 4 nanoparticles, the peak of C− O− C group becomes a very weak peak (1113.8 cm −1 ), and has a slight shift. The characteristic absorption peak of oxygen-containing C− O− C group decreased dramatically indicates that the oxidized CNT have been reduced. The peak at 849.2 cm −1 can be attributed to the vibrations of Fe-O 49 .
The Raman spectra of the oxidized CNT and the synthesized mesoporous Fe 3 O 4 @CNT are shown in Fig. 5a. It is obvious that there are two strong peaks at 1344.5 cm −1 and 1577.8 cm −1 , respectively. The 'D' band at 1344.5 cm −1 is related to the defects and disorders in the hexagonal framework of the nanotuble walls 50 . So far, the contribution to the 'D' band from the defects of nanotube walls and other forms of carbon (e.g., rings) is still not completely understood. The 'G' band appeared at 1577.8 cm −1 , which is related to the vibration of sp 2 -bonded carbon atoms in a two-dimensional hexagonal lattice. This peak corresponds to the E 2g graphite-like tangential mode. As for the synthesized mesoporous Fe 3 O 4 @CNT, three peaks characteristic of iron oxides appeared at 214.4, 280.4, and 692.5 cm −1 , which were assigned to the E g , T 2g and A 1g vibration modes, respectively 51,52 . The intensity ratio of the 'D' to 'G' band is about 1.13 and 1.18 for the oxidized CNT and the mesoporous Fe 3 O 4 @CNT, respectively. Generally, the intensity ratio of 'D' peak (I D ) to 'G' peak (I G ) can be used to estimate the defect degree of multi-wall CNT. The increase of I D /I G ratio indicates that after filling the oxidized CNT with Fe 3 O 4 nanoparticles in the hydrothermal system, the surface defects of CNT have a slightly increase. On the other hand, when the oxidized CNT were filled with Fe 3 O 4 nanoparticles, all the intensities of 'D' peak and 'G' peak have a further increased compared with the oxidized CNT, indicating the defect density on theof carbon materials has decreased. In order to evaluate the surface properties and loading level of CNT by the Fe 3 O 4 nanoparticles, the synthesized mesoporous Fe 3 O 4 @CNT were measured by TGA under N 2 atmosphere (Fig. 5b) and in air ( Figure S3), respectively. From Fig. 5b, we can see that there was a tiny weight loss (3.2%) in the temperature range of 20-160 °C. This weight loss can be assigned to the loss of adsorbed water on the surface of mesoporous Fe 3 O 4 @CNT. An approximately 18.1% weight loss occurring in the temperature range from 160-599 °C was attributed to the decomposition of labile oxygen functional groups from CNT layer. In the temperature interval of 599-832 °C, the weight loss (5.3%) was ascribed to the decomposition of stable oxygen functional groups from the CNT layer. When the temperature is over 832 °C, the TGA curve has a sharp decrease. The significant weight loss (12.5%) takes place between 832 °C and 837 °C can be attributed to the removal of very-stable oxygen functional groups from CNT layer. With the further increase of temperature (837-886 °C), a slow mass loss can be observed. The tiny weight loss (1.8%) may be attributed to the much more stable oxygen functional groups from CNT. From Figure S3, we can calculate the loading level of CNT, in terms of the weight ratio of Fe 3 O 4 nanoparticles filled/embedded in CNT and it was found that the loading level can be as high as ~66.5 wt%. Figure 5c shows the Fe2p high-resolution X-ray photoelectron spectra (XPS) of the synthesized mesoporous Fe 3 O 4 @CNT. There is no evident satellite peak at 719.2 eV, which is the characteristic peak for The electrochemical performance of the synthesized mesoporous Fe 3 O 4 @CNT was evaluated as anodes for LIBs in a CR2025 coin-type cells. Figure 6a shows representative discharge/charge voltage profiles of the synthesized mesoporous Fe 3 O 4 @CNT at a current density of 1.0 C within the cut-off voltage window of 0.01-3.0 V. As for the first discharge profile, there is a voltage plateau at 0.8 V and then the voltage plateau followed by a long slope. This voltage plateau reveals the reduction of Fe 3+ to Fe 0 , and the formation of amorphous Li 2 O as well as the irreversible reaction with the electrolyte. The sloping tail (referred to the additional plateau at ~0.5 V) below the conversion regime can be assigned to the formation of gel-like film and/or interfacial lithium storage 55,56 . The initial discharge and charge capacities are found to be 981.4 and 881.9 mAhg −1 , respectively. The irreversible capacity loss of about 10.2% may be ascribed to the conversion reaction of the Li-Fe-O compound to Fe and Li 2 O, and the formation of inorganic solid electrolyte interface (SEI) film, electrolyte decomposition and the reaction of lithium ions with oxygen-containing groups in the CNT layer 57 . At the following cycles, the discharge capacity of mesoporous Fe 3 O 4 @CNT electrode in the 15 th , 30 th , 45 th and 60 th cycle is found to be 490.8, 397.8, 376.3 and 358.9 mAh g −1 , respectively. In the charge process, the charge capacity is found to be 471.3, 389.3, 363.3 and 343.8 mAh g −1 in the 15 th , 30 th , 45 th and 60 th cycle, respectively. After 60 cycles, the discharge and charge profiles nearly overlap, suggesting that the mesoporous Fe 3 O 4 @CNT electrode has a good electrochemical cycling ability. The cyclic stability of the mesoporous Fe 3 O 4 @CNT was evaluated at a rate of 0.5 C, as shown in Fig. 6b. From the 15 th cycle onwards, the discharge capacities of the mesoporous Fe 3 O 4 @CNT gradually decreased from 622.7 mAhg −1 to 571.1 mAhg −1 within 70 cycles with a high Coulombic efficiency of about 95-97%. For pure Fe 3 O 4 nanoparticles, the specific capacity decreased significantly and retained only about 200 mAhg −1 at a current density of 200 mAg −1 after 50 cycles 58 . The results indicate that the synthesized mesoporous Fe 3 O 4 @CNT exhibit good structural stability. Figure 6c shows the rate capability of the mesoporous Fe 3 O 4 @CNT at different current densities. It is obvious that the mesoporous Fe 3 O 4 @CNT delivered very high capacities of 684.1, 541.5, 432.2 and 358.9 mAh g −1 at the current densities of 2.0, 5.0, 10 and 15 C, respectively. Even as a high current density of 20 C, the capacity still remains as high as 275.2 mAh g −1 , suggesting the introduction of mesoporous CNT may significantly enhance the rate capability. When returning to the initial current density of 1.0 C, the mesoporous CNT electrode returns to a relatively high capacity (826.4 mAh g −1 ) compared with the original capacity (939.6 mAhg −1 ). This confirms that the mesoporous CNT electrode can keep its integrity during the repeated cycling process. In order to support the superior electrochemical performance of the mesoporous Fe 3 O 4 @CNT composites, the electrochemical data of CNT electrode and pure Fe 3 O 4 nanoparticles electrode are provided to compare with that of the Fe 3 O 4 @CNT composites. Figure 6d shows the cycling performance of CNT electrodes at 0.5 C. It can be seen that there is a large capacity loss between the first and second cycles for the pure CNT electrodes because of the formation of SEI layer on the surface of electrode. After 30 cycles, the capacity is only ~188.1 mAh g −1 and maintains this value until 75 cycles. We also investigated the rate performance of CNT electrodes under different current densities, as shown in Fig. 6e. It can be seen that the CNT electrodes delivered low capacities of 410.3, 305.4, 254.1, 226.1 and 186.7 mAh g −1 at the current densities of 0.2, 0.5, 1.0, 2.0 and 5.0 C, respectively. As for mesoporous Fe 3 O 4 @CNT electrodes, the capacity can reach as high as 684.1 and 541.5 mAhg −1 at the current density of 2.0 and 5.0 C which were about 3 times that of CNT electrode. When the CNT electrode returns to the initial current density of 0.2 C, the capacity can only return to 328.9 mAhg −1 . Figure 6f shows the cycling performance of pure Fe 3 O 4 nanoparticles electrode at 0.5 C. The synthetic procedure of Fe 3 O 4 nanoparticles used for LIBs test was similar to the synthesis of Fe 3 O 4 @CNT, only the oxidized CNT were not added. The morphology of Fe 3 O 4 nanoparticles was shown in Figure S4. From  Fig. 6f, it can be seen that the capacity has a quickly decreased between the first and the third cycles for the pure Fe 3 O 4 nanoparticles electrode. The initial discharge capacity was 862.6 mAhg −1 which is slightly lower than that of the Fe 3 O 4 theoretical capacity value (928 mAh g −1 ). The capacity dramatically decreased to 539.8 mAhg −1 after 3 cycles, and maintaining about 485.4 mAh g −1 after the following cycling. The electrochemical data of CNT electrode and pure Fe 3 O 4 nanoparticles electrode indicate that the synthesized mesoporous Fe 3 O 4 @CNT composites exhibited good electrochemical performance.
In order to investigate the rate performance of the synthesized mesoporous Fe 3 O 4 @CNT, cells were cycled at high current densities (5.0 and 10 C) over 500 cycles. Figure 7a shows the long-term cycling performance under high current density of 5.0 C. The capacity decreased in the first 10 cycles, which was attributed to the formation of SEI film. After 30 cycles, the capacity stabilized at around 231.1 mAhg −1 . The capacity decayed gradually to 188.3 mAhg −1 after 500 cycles. Figure 7b shows the charge-discharge profiles of the mesoporous Fe 3 O 4 @CNT for the 1 st , 100 th , 200 th , 300 th , 400 th and 500 th cycle at a rate of 5.0 C. It is clearly observed that the capacity curves almost overlap after 100 cycles, and the capacity maintain around 180 mAh g −1 at the following cycles. The relatively low capacity of the mesoporous Fe 3 O 4 @CNT electrode may originate from the CNT components, not the ultrafine Fe 3 O 4 nanoparticles. With the further increase of current density, the mesoporous Fe 3 O 4 @CNT still exhibit a good electrochemical performance, as shown in Fig. 7c. At 10 C, while the capacity decreased in the first 10 cycles, while the capacity stabilized at around 215.2 mAh g −1 after 30 cycles, and remained at around 137.5 mAh g −1 after 500 cycles. Figure 7d shows the charge-discharge profiles of this composite at the 1 st , 100 th , 200 th , 300 th , 400 th and 500 th cycle under a rate of 10 C. The capacity was found to be 182.6 mAh g −1 at the 100 th cycle, and stabilized at around 140 mAhg −1 after the following cycles. During the test of charge-discharge, one cell was used for 5.0 C and another cell is used for 10.0 C. The individual difference of cells may result in the difference of charge-discharge profiles, especially for the first cycle profile.
Due to the unique mesoporous structure of the CNT and the presence of large amount monodisperse Fe 3 O 4 nanoparticles filled in CNT, our materials exhibit good cycling capability, stability and rate performance. It is widely accepted that the specific capacity and cycling performance of iron oxides (e.g., α -Fe 2 O 3 and Fe 3 O 4 ) are closely related with their shapes and microstructures. Our previous study shows that α -Fe 2 O 3 microdisks exhibited good cyclic stability and rate performance for LIBs 59 . The result demonstrated that the disk-like structure was facilitating the transfer of Li + ions and electrons. Among the promising metal oxides for LIBs anode materials, magnetite (Fe 3 O 4 ) has high theoretical capacity, which is about three times than that of conventional graphite. However, the specific capacities of pure Fe 3 O 4 decay readily because: (1) the chemical reaction of Fe 3 O 4 in lithium storage (Fe 3 O 4 + 8 Li + + 8e − ↔ 3Fe + 4Li 2 O) induces structure pulverization after the lithium inclusion; (2) the generated Fe nanoparticles induce some irreversible reactions and result in poor capacity retention; (3) the SEI film on Fe 3 O 4 are destroyed due to the large volume change and the repeated formation/decomposition of SEI film 60 . One of the effective methods to overcome these drawbacks is to coat the iron oxides with a carbon shell. Our mesoporous Fe 3 O 4 @CNT possess all of these properties, which facilitate the transmission of lithium ions and electrons, shorten the diffusion time of lithium ions and improve the electron conductivity, and enhance the cycling stability and rate performance. The Fe 3 O 4 @CNT electrode shows excellent electrochemical properties, which can be attributed to the nanoconfinement effect 61 . and the excellent conductivity of CNT as well as the good stability of such special nanostructure. On the other hand, the Fe 3 O 4 @CNT can form a three-dimensional network which not only can accommodate the volume variation on insertion/extraction of lithium ions, but also protect the active materials from severe aggregation 33   nanohorns on the CNT backbone for LIBs 61 . When cycled at high rate of 1-3 C, comparable capacities of 420-500 mAhg −1 to those observed here can be maintained. Yu et al. reported a hybrid material of CNT-encapsulated Fe 2 O 3 nanoparticles for LIBs 38 . Electrochemical testing shows that a stable capacity of 335 mAhg −1 can be achieved when the current density is 1.2 C. Very recently, Chen and co-workers reported a unique composite with tiny Fe 3 O 4 dispersed into ~54.6% carbon via a high pressure and temperature based solvothermal route with a capacity of 610 mAh g −1 after 100 cycles 62 . Herein, a high capacity of 684.1 mAhg −1 can be obtained at 2.0 C on the current mesoporous Fe 3 O 4 @CNT anode. When cycled at high rate 5 C, a capacity of 541.5 mAhg −1 can be maintained. Even at 10 C, a similar capacity of 432.2 mAhg −1 can still be obtained. Compared with the reported data, the reported mesoporous Fe 3 O 4 @ CNT is not the anode that offers the highest discharge capacity, but is comparable with the state-of-the-art of Fe 3 O 4 based anode. Such mesoporous Fe 3 O 4 @CNT anodes have excellent rate performance ascribed to its interconnected porous nanostructures, which offers a new route for high rate anode.
In order to evaluate the electrochemical dynamical behavior, an electrochemical impedance spectroscopy (EIS) analysis was performed, as shown in Fig. 7e. The EIS spectra of mesoporous Fe 3 O 4 @CNT before and after 50 cycles have a quasi-semicircle in the high frequency region and an inclined linear part at low frequency. The depressed capacitive arc was related to the charge transfer resistance, and the sloped line was attributed to the Warburg impedance that is derived from lithium ion diffusion 27 . It is clearly seen that the diameter of capacitive arc after 50 cycles is smaller than that of the samples before cycles, indicating that the electrical conductivity of mesoporous Fe 3 O 4 @CNT improved after cycling. The small diameter of capacitive arc after cycling indicates the low charge transfer resistance, which will improve the electron kinetics in the electrode materials 18 . Figure 7f shows the HRTEM image of the active materials (Fe 3 O 4 @CNT) after 50 cycles at a rate of 2 C. It is obvious that the Fe 3 O 4 nanoparticles are still embedded in the CNT. These results indicated the excellent electrochemical properties of the synthesized mesoporous Fe 3 O 4 @CNT. Multi-wall CNT were oxidized into mesoporous structure by a mixed acid solution (HNO 3 and H 2 SO 4 ) at 80 °C. As for CNT, they generally exhibit low dispersibility in water and organic solvents because of their carbonic nature and the presence of van der Waals attraction between the nanotubes due to their hydrophobic properties. After the CNT were oxidized at 80 °C, the CNT are easily dispersed in water and the surface of CNT is porous. The mixed acid solution is strongly oxidising which can cause C-C bond cleavage of local CNT backbone and formation of numerous oxygen-containing groups. The oxidization temperature is very important for the formation of mesoporous CNT. When the temperature was elevated to 120 °C for 20 min, the CNT backbone will be destroyed ( Figure S5). Second, when the oxidized CNT were dispersed in the solution containing iron ions (M z+ , z = 2 and 3, M= Fe), the inside of CNT will be filled completely by the solution based on the capillarity effect, and the oxygen-containing groups on the surface of CNT will act as the active site for the complexation of iron ions by the charge interaction. At this time, the distribution of M z+ is in fact composed of two parts: one part of M z+ is located in the solution, and another part of M z+ is located in inside the CNT. When the NaOH solution was added, the M z+ located in the solution will precipitate first. The citrate ions present help control the size of the dispersed M z+ precipitates leading to a narrow diameter distribution of particles. When the M z+ in solution was gradually consumed, the M z+ located at the ends of CNT will be precipitated. Then, the ends of CNT will be blocked with M z+ precipitates. However, there will still be some spaces or channels around the blocking M z+ precipitates, which facilitate the transmission of the OH − in solution to the M z+ in CNT. Driven by the concentration gradient of OH − in solution, the loading level of M z+ precipitates located at the ends of CNT will become more and more compact leading to the formation of bigger-size agglomerates. Our experimental results already confirm this fact (see Fig. 2b and Fig. 3a). If the CNT are not oxidized in the mixed acid solution, further filling CNT with iron oxides nanoparticles will be inhibited. Aiming at this challenging problem, we propose this pre-oxidation strategy for the CNT. First, after oxidizing CNT in mixed acid solution, the CNT will become water soluble, which is key factor to fill the CNT with iron oxides nanoparticles in situ. In order to avoid the destruction of the CNT backbone, fine control the oxidation parameter is very important. Second, the oxidation of CNT will make the surface of CNT possess numerous defects or pores. These porous structures of CNT will improve the transmission of lithium ions and electrons. Because of the mesoporous structure of CNT, OH − can permeate the carbon shell into the inner of CNT and react with the M z+ ions. After that, the inner and outer space of CNT are full of M z+ precipitates. When the temperature was elevated to 180 °C for 20 h, M z+ precipitates are completely converted into the Fe 3 O 4 nanoparticles. When the hydrothermal reaction is completed, the cooled solution mixture is composed of Fe 3 O 4 nanoparticles and mesoporous Fe 3 O 4 @CNT. As for the Fe 3 O 4 nanoparticles, they are water soluble, small size ( Figure S6) and weak magnetic intensity. When a magnet is applied, the Fe 3 O 4 nanoparticles in solution have no response to the magnet. As for the mesoporous Fe 3 O 4 @CNT, they were full of Fe 3 O 4 nanoparticles (about 66.5 wt%). This endows the mesoporous Fe 3 O 4 @CNT with a relatively strong magnetic intensity (12.1 emu/g). Therefore, the mesoporous Fe 3 O 4 @CNT are easily isolated from the solution uisng magnet within a few seconds. After washing the mesoporous Fe 3 O 4 @ CNT with deionized water five times, the Fe 3 O 4 nanoparticles in solution, which are not inside the CNT, are completely removed. The efficacy in generating large scale mesoporous Fe 3 O 4 @CNT arises from the capillarity effect, and the mesoporous Fe 3 O 4 @CNT can be obtained on a high scale (about 1.42 g per one-pot, 50 mL hydrothermal reactor). Compared with the conventional method using Al 2 O 3 membrane Scientific RepoRts | 5:17553 | DOI: 10.1038/srep17553 as template to fill CNT with iron oxides, our current strategy is very effective, low cost, high filling level and easy large scale production.

Conclusions
An effective method for in situ filling of multi-wall CNT (diameter of 20-40 nm, length of 30-100 μ m) with ultrafine Fe 3 O 4 nanoparticles (8-10 nm) has been demonstrated. The loading level of CNT, in terms of the weight ratio of Fe 3 O 4 nanoparticles, can reach 66.5 wt% which is significantly higher than the highest reported value (51.8 wt%). The prepared Fe 3 O 4 @CNT exhibited mesoporous properties, and can be obtained on a large scale (~1.42 g per one-pot). The BET surface area of the mesoporous Fe 3 O 4 @CNT was 109.4m 2 g −1 , which is much higher than that of the commercial Fe 3 O 4 (~2 m 2 g −1 ). Even when cycled at a high current rate of 20 C, a high capacity of 275.2 mAhg −1 could still be achieved. The advantages for the present Fe 3 O 4 @CNT composite compared with that of most previous Fe 3 O 4 /carbon hybrid electrodes are that they facilitate charge transport, maintain the electrode integrity, and endow the electrodes with high capacity, high rate performance and excellent cycling stability showing they are promising anode material for LIBs.