In the past few decades, porous materials have been used in many fields, such as filters, catalysts, cells, supports, optical materials, and so on 123. In general, porous materials can be classified into three types depending on their pore diameters, namely, microporous (<2 nm), meso- or transitional porous (2-50 nm), and macroporous (>50 nm) materials, respectively 4. Currently, the mesoporous materials have attracted growing research interests and have great impact in the applications of catalysis, separation, adsorption and sensing due to their special structural features such as special surface area and interior void 25678. On the other hand, iron oxide nanomaterials have been extensively studied by material researchers in recent years, due to their novel physicochemical properties and advantages (high saturation magnetization, easy synthesis, low cost, etc.) and wide applications in many fields (magnetic recording, pigment, magnetic separation, and magnetic resonance imaging, MRI) 910111213141516.

However, it is crucial to realize the magnetic iron oxide materials with mesoporous structure which can further adjust the physical and chemical properties of iron oxides for expanding application. According to the previous studies, the porous iron oxide nanomaterials have remarkable magnetic properties, special structures and greatly potential applications in targetable or recyclable carriers, catalyst and biotechnology 1718. For example, Yu et al. 19 fabricated novel cage-like Fe₂O₃ hollow spheres on a large scale by hydrothermal method. In the report carbonaceous polysaccharide spheres were used as templates, and the prepared Fe₂O₃ hollow spheres exhibit excellent photocatalytic activity for the degradation of rhodamine B aqueous solution under visible-light illumination. Wu et al. 20 successfully developed porous iron oxide-based nanorods used as nanocapsules for drug delivery, and this porous magnetic nanomaterial exhibited excellent biocompatibility and controllability for drug release.

It is well known that the intrinsic properties of an iron oxide nanomaterial are mainly determined by its size, shape, and structure. A key problem of synthetically controlling the shape and structure of iron oxide nanomaterials has been intensively concerned by many researchers. In previous studies, there have been various porous iron oxide nanomaterials, such as porous α-Fe₂O₃ nanorods, Fe₃O₄ nanocages, and so on 92122232425. However, to our best knowledge, there are few reports for fabricating the mesoporous structure of monodisperse spindle-like Fe₃O₄ NPs. Thus, we employ forced hydrolysis method to prepare spindle-like α-Fe₂O₃ NPs first. Then as-prepared α-Fe₂O₃ NPs were reduced by hydrogen gas at different temperatures. The structure, morphology, and magnetic properties of samples were investigated by multiple analytical technologies. The results reveal that spindle-like Fe₃O₄ mesoporous NPs could be obtained after annealing at 350°C.

Forced hydrolysis method has been widely used for preparing α-Fe₂O₃ NPs since the first study by Matijevic et al. 4 and Cornell and Schwertmann 27. In general, in the presence of water, the Fe³⁺ salt dissociates to form the purple, hexa-aquo ion, the electropositive cations induce the H₂O ligands to act as acids (except at very low PH) and hydrolysis by heating. In addition, the Fe salt was added to preheated water in order to avoid nucleation of geothite during the initial heating stage 428. The synthesis of Fe₃O₄ NPs can be reached by reduction of α-Fe₂O₃ NPs in hydrogen atmosphere. In brief, the whole experimental process can be described as follows 4:

FeCl 3   +   6H 2 O   →   Fe ( H 2 O ) 6 3 + +   3Cl −

2Fe ( H 2 O ) 6 3 +   →   Fe 2 O 3   +   6H +   +   9H 2 O

3Fe 2 O 3   +   H 2   →   2Fe 3 O 4   +   H 2 O

In the hydrolysis process, the features that affect the products of the experiment generally include additive, reaction temperature, aging time, PH value. On the basis of previous reports, the addition anions have great effect on the shape of α-Fe₂O₃ NPs. The used PO₄^3- anions will adsorb onto the crystal planes parallel to the c-axis of α-Fe₂O₃, which causes the growing of the α-Fe₂O₃ NPs along the c-axis direction and promotes the formation of spindle-like α-Fe₂O₃ NPs 222930. More detailed formation mechanisms in this study are currently under way.

Figure 1 shows the XRD patterns of the samples. Curve a is the pattern of S1. The diffraction peaks (2θ = 24.1°, 33.2°, 35.6°, 40.9°, 49.5°, 54.1°, 62.4°, and 64.1°) are coincided well with the value of JCPDS card 33-0664 (shown as green lines in the bottom), which could be well indexed to the pure hexagonal phase of hematite ((012), (104), (110), (113), (024), (116), (214), and (300)). Curve b displays the diffraction peaks of S2 (250°C). In this curve all the peak positions do not change, which reveals that the sample is still in α-Fe₂O₃ phase after annealing at this temperature. However, when the annealing temperature elevates to 300°C (S3), some new peaks (2θ = 30.2°, 43.3°, 57.3°, and 62.8°) are appeared in curve c. These peaks can be indexed to cubic spinel magnetite (JCPDS card 19-0629, indexed with red lines in the bottom). Moreover, the peaks of α-Fe₂O₃ become weak, which implies that the α-Fe₂O₃ NPs partially transform to Fe₃O₄ NPs after annealing at 300°C. Subsequently, all the peaks in the pattern of S4 (350°C) could be attributed to Fe₃O₄, their intensity become much stronger. The peaks attribute to α-Fe₂O₃ are almost disappeared, which demonstrates that the NPs is mainly Fe₃O₄ NPs. When the temperature was increased to 400°C (S5, shown in curve e), the peaks (2θ = 44.7°, and 65.0°) can be attributed to α-Fe (JCPDS card 06-0696, shown as blue lines in the bottom). Finally, the sample of S6 mainly transforms to α-Fe phase after annealing at 450°C (curve f).

The morphologies of the samples were studied by SEM analysis. The SEM image of S1 in Figure 2a clearly shows the formation of uniform spindle-like α-Fe₂O₃ NPs with the length and outer diameter approximately 250 and 60 nm, respectively. It is obvious that each of the spindle-like particles possesses a rough surface composed of many small particles. Figure 2b,c,d,e,f shows the SEM images of S2, S3, S4, S5, and S6, respectively. In the Figure 2b,c,d, their particle shape and size are preserved well. However, as shown in Figure 2e, when the annealing temperature increases to 400°C, the shape of the particles is damaged and many particles are melted. For the sample annealed at 450°C (shown in Figure 2f), the spindle-shape of precursor α-Fe₂O₃ NPs is disappeared completely. Instead, the obtained particles have irregular morphology. All the XRD and SEM results clearly indicate that α-Fe₂O₃ NPs can be transformed to Fe₃O₄ NPs after annealing in the reducing atmosphere with temperature up to 350°C, meanwhile the shape and size of the NPs are kept.

For further discussing the morphologies and structures of the samples, TEM images of S1, S2, S4, and S5 are presented, as shown in Figure 3. It can be found in Figure 3a that the as-prepared α-Fe₂O₃ NPs are consisted of smaller closely packed particles, which causes rough surfaces. The inserted SAED pattern is in agreement with the structure plane of α-Fe₂O₃, which also reveals that the α-Fe₂O₃ NPs are in polycrystal. The TEM image of S2 in Figure 2b clearly illustrates that the NPs are mesoporous structure. The SAED pattern demonstrates that the sample is also in polycrystal feature with α-Fe₂O₃ phase. The results reveal that the porous structure has been formed after annealing at 250°C. Figure 3c shows the TEM image of S3 annealed at 300°C. It can be clearly seen that the shape and size of the particles are well preserved. Moreover, the size of the pores in the sample becomes larger than that of the pores in S2. This is because more vacancies are produced after reducing by H₂. These vacancies aggregate to form larger pores. The inserted SAED pattern implied that the sample S3 is a compound of Fe₃O₄ and α-Fe₂O₃, which coincides with the XRD result. Figure 3d displays the TEM images of S4 (350°C). Although the sample S3 and S4 have similar porous structure, the SAED patterns of the samples are changed and the ring patterns of S4 can be indexed as a cubic spinel phase of magnetite, which demonstrates that the sample S4 are in Fe₃O₄ phase. Figure 3e shows the TEM images of S5. Clearly, some particles are also spindle-like and porous in structure. However, most of the particles are irregularly shaped, meaning that the shape of the sample has been partly damaged after annealing temperature at 400°C. This may be due to the collapse of NP structure, which is because too many large pores are produced inside the NP. The inserted SAED patterns reveal that the sample is a compound of Fe₃O₄ and α-Fe. The TEM result is in good agreement with the XRD and SEM results. Moreover, it proves that the annealing treatment can cause the mesoporous structure.

Figure 4 shows the ATR-FTIR spectra of the samples S1 (a) and S4 (b). The absorption band at 558.86 cm^-1 in the curve a is attributed to the bending vibrations of the Fe-O in α-Fe₂O₃ 31, while the fingerprint bands at 1037.89, 1004.85, 967.99, and 928.40 cm^-1 could be related to PO₄^3- anions 32. In the curve b, there is an absorption band at 971.16 cm^-1. This band is attributed to NaFePO₄ 33, which indicates that a new component (NaFePO₄) might be generated on the surface of the particles after annealing. The absorbtion band at 585.97 cm^-1 is associated with the Fe-O stretching mode of the Fe₃O₄ NPs 343536. In addition, the absorption band at about 685 cm^-1 is observed in both of the curves, which is assigned to the bending modes of Fe-O-H 31. The ATR-FTIR results further prove the phase transformation of NPs from α-Fe₂O₃ to Fe₃O₄. Moreover, the detection of the phosphate reveals that the phosphate possibly plays an important role in the formation of the spindle and porous structures.

Nitrogen adsorption-desorption isotherms were performed to determine the surface area and pore size of S4, which is shown in Figure 5. The BET surface area is measured using multipoint BET method within the relative pressure (P/P₀) range from 0.05 to 0.3. The pore size distribution was determined by the Barret-Joyner-Halender (BJH) method using desorption isotherm. The pore volume and average pore size for the sample were determined according to the nitrogen adsorption volume at the relative pressure (P/P₀) of 0.9956. As shown, the sample exhibits a type H3 hysteresis loop according to Brunauer-Deming-Deming-Teller (BDDT) classification, which indicated the presence of mesopores (2-50 nm) with a cylindrical pore mode 37. According to the BET method, the specific surface area of the samples is determined to be 7.876 m² g^-1. The BJH adsorption cumulative volume of pores between 17 and 300 nm is 0.15 cm³ g^-1. However, the BJH adsorption average pore of the sample is 78.1 nm, which is probably because the pores in the particles are hermetic, nitrogen could not be contact with the internal wall of the pores 37. On the other hand, the aggregation of the Fe₃O₄ NPs will cause many spaces among them, which can also lead to the larger result of the pore size 3839. The density of the sample based on the current BET result is calculated to be 2.16 g cm^-3 (Assuming that each Fe₃O₄ NPs is an ellipsoid, thus ρ = M V , and M = A_s · S, where ρ is the density of the sample; M, S and V are the mass, surface area and volume of one Fe₃O₄ particle, respectively; A_s is the BET surface area of the sample. As V = 4 3 π r a r b 2 and S = 2 π r b ( 7 3 r a 2 + 2 3 r a r b + r b 2 ) , where r_a and r_b are the length and outer diameter of the Fe₃O₄ NPs, the density of the sample based on the BET result is estimated to be 2.16 g cm^-3), it is smaller than 5.18 g cm^-3 for corresponding bulk Fe₃O₄, which indirectly proves that the Fe₃O₄ NPs are in porous.

As the physicochemical properties of samples are related to their morphologies and structures, the magnetic hysteresis loops of the samples (S1 and S4) were measured by VSM at room temperature, and the results are shown in Figure 6a. From the curve 1, we can see that the sample exhibits weak ferromagnetic behavior before annealing, and its saturation magnetization and coercivity are 0.64 emu g^-1 and 37.6 Oe, respectively. It has been proved that the structure of α-Fe₂O₃ can be described as consisting hcp arrays of oxygen ions stacked along the [001] direction. Two-thirds of the sites are filled with Fe³⁺ ions, which are arranged regularly with two filled sites being followed by one vacant site in the (001) plane thereby forming sixfold rings. In this case, the arrangement of cations produces pairs of Fe(O)₆ octahedra, and Fe³⁺ ions are antiferromagnetically coupled across the shared octahedral faces along the c-axis. In the basal plane, there are two interpenetrating antiferromagnetic sublattices. As the electron spins of these sublattices are not exactly antiparallel (with a canting angle of <0.1°), a weak ferromagnetic interaction is resulted, and this effect dominates the magnetic behavior at room temperature 4. As shown in curve 2 (Figure 6a), the S4 possessed a saturation magnetization of 85.18 emu g^-1 and a coercivity of 86.01 Oe, the saturation magnetization is close to 92 emu g^-1 for corresponding bulk Fe₃O₄ 40, which is because the α-Fe₂O₃ phase of the NPs has transformed to Fe₃O₄ phase after annealing. The structure of magnetite is inverse spinel, where there is a face-centered cubic unit cell based on 32 O^2- ions which are regularly cubic close packed along the [111]. Two different cation sites occupied by Fe²⁺ and Fe³⁺ form two interpenetrating magnetic sublattices. At room temperature the spins on the two sites are antiparallel and the magnitudes of types of spins are unequal, which causes the ferromagnetism of magnetite. In addition, the particle size and crystal morphology affect the coercivity in the order: spheres < cubes < octahedral in line with the increase in the number of magnetic axes along this series of shapes 4. In addition, anisotropy shape of the particles may also affect the magnetism 41. Figure 6b shows the photographs of the samples dispersing in ethanol with and without an external magnetic field. It can be clearly seen that the Fe₃O₄ NPs are well dispersed in ethanol before magnetic separation. However, after magnetic separation all Fe₃O₄ NPs are attracted together by magnet. And the separating time only needs 35 s. For comparison, the α-Fe₂O₃ NPs dispersing in ethanol almost do not change before and after magnetic separation. The results demonstrate that the Fe₃O₄ NPs present excellent magnetic separation property and have good potential application for recyclable nanomaterials.

In conclusion, spindle-like α-Fe₂O₃ NPs were fabricated by forced hydrolysis of FeCl₃ in the presence of PO₄^3- anions. The as-prepared α-Fe₂O₃ NPs were then reduced in hydrogen at 350°C and transformed into spindle-like Fe₃O₄ NPs with mesoporous structure. The as-obtained mesoporous Fe₃O₄ NPs possess a high BET surface area of 7.876 m² g^-1. In addition, the obtained Fe₃O₄ NPs possessed a high saturation magnetization of 85.18 emu g^-1 and a coercivity of 86.01 Oe. Owing to its excellent magnetic separation property and special mesoporous structure, the as-obtained Fe₃O₄ NPs may have a great potential application in the future.

AP: analytically pure; ATR-FTIR: attenuated total reflection fourier transform infrared spectroscopy; BDDT: Brunauer-Deming-Deming-Teller; BET: Brunauer-Emmett-Teller; BJP: Barret-Joyner-Halender; FSEM: field emission scanning electron microscopy; MRI: magnetic resonance imaging; NPs: nanoparticles; SAED: selected area electron diffraction; TEM: transmission electron microscopy; VSM: vibrating sample magnetometer; XRD: X-ray diffraction.

The authors declare that they have no competing interests.

SZ participated in the materials preparation, data analysis and drafted the manuscript. WW, XX and JZ participated in the sample characterization. FR conceived and co-wrote the paper. CZ participated in its design and coordination. All authors read and approved the final manuscript.