One-dimensional fossil-like γ-Fe₂O₃@carbon nanostructure: preparation, structural characterization and application as adsorbent for fast and selective recovery of gold ions from aqueous solution

The preparation of 1D α-Fe₂O₃ follows the same procedure as has been reported previously [18]. Typically, 2.0 g of iron(III) nitrate was added into 20 ml of TEAOH (20 wt% in H₂O), and the mixture was stirred for 20 min before being heated in a microwave oven for 5 min (100 W). Afterward, the sample was washed with DI H₂O for five times using a centrifuge, dried at 60 °C for 24 h, and calcined at 300 °C in air for 1 h to obtain 1D α-Fe₂O₃.

Glucose (0.57 g) was dissolved in 40 ml DI H₂O, followed by the addition of 0.20 g 1D α-Fe₂O₃. The mixture was subjected to ultrasonic irradiation to obtain a homogeneous dispersion, after which it was transferred into a Teflon-lined autoclave and then heated up to 180 °C for 18 h. After being cooled to room temperature, the sample was repeatedly washed with water until the pH was ca. 7. The sample was then dried overnight to obtain a dark reddish-brown powder. A second carbon coating was carried out by repeating the above procedures. The carbon-coated samples were calcined at 500 °C in a N₂ atmosphere for 2 h prior to any adsorption experiments to obtain γ-Fe₂O₃@C.

All adsorption experiments were carried out at room temperature. The metal salts used are as follows: (a) Zn(NO₃)₂.4H₂O, (b) Ni(NO₃)₂.6H₂O, (c) Cu(NO₃)₂.3H₂O, (d) Mg(NO₃)₂.6H₂O, (e) Cr(NO₃)₃.6H₂O and (f) HAuCl₄.xH₂O. Certain amounts of the metal salts were added to 100 ml of ultrapure water to yield an initial concentration of 200 mg l⁻¹. After that, 0.10 g γ-Fe₂O₃@C sorbent (Fe@C(2x)-500-N₂) was added into the metal solution and the mixture was stirred for 24 h. During the stirring, 3 ml sample was taken out in an hourly interval for inductively coupled plasma optical emission spectrometry (ICP-OES) analysis. The liquid sample was centrifuged at 13 000 rpm for 5 min and 1 ml of the supernatant was retrieved and diluted for ICP-OES analysis to determine the metal ion concentration, while the remaining liquid sample was returned to the metal solution. The metal removal efficiency (equation (1)) and the equilibrium adsorption capacity (equation (2)) were calculated using the equations below:

$$\begin{eqnarray}&&\mathrm{Metal}\,\,{\rm{removal}}\,\mathrm{efficiency}\,\,( \% )=\displaystyle \frac{{{C}}_{{\rm{i}}}-{{C}}_{{\rm{f}}}}{{{C}}_{{\rm{i}}}}\times 100 \% \end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{{q}}_{{\rm{e}}}=\displaystyle \frac{({{C}}_{{\rm{i}}}-{{C}}_{{\rm{e}}})\times {V}}{{m}}\end{eqnarray} \tag{ 2 }$$

where C_i is the initial metal concentration (mg l⁻¹), C_f is the final metal concentration (mg l⁻¹), q_e is the equilibrium adsorption capacity (mg g⁻¹), C_e is the metal concentration at equilibrium (mg l⁻¹), V is the sample volume (ml) and m is the mass of the adsorbent (g).

The morphology and the particle size were observed by a transmission electron microscope (TEM, JEOL-3010 at 300 kV). The crystal structure of the samples was characterized by powder x-ray diffractometry (XRD; Bruker D8 Advance diffractometer system) with a Cu Kα (1.5418 Å) source, while thermogravimetric analysis (TGA) was performed on a Q500 TA instrument in an air flow. The surface area was measured at 77 K (liquid nitrogen) on a Quantachrome Autosorb-6B surface area and pore size analyzer. Before the measurement, the samples were degassed at 200 °C for several hours. The pore size distribution was calculated with the Barrett−Joyner−Halenda method using the desorption isotherm branch. The specific surface area was determined according to the Brunauer–Emmett–Teller (BET) method in the relative pressure (p/p⁰) range of 0.05−0.3. The infrared spectra for dried samples were recorded on a Bio-Rad Excalibur spectrometer employing a KBr disc method. X-ray photoelectron spectroscopy (XPS) spectra were recorded in a Kratos AXIS Ultra DLD spectrometer and Al Kr radiation was used as the x-ray source. The C1s peak at 284.6 eV was used as a reference for the calibration of the binding energy scale. The saturation magnetization curves of the prepared samples were obtained on a vibrating sample magnetometer (VSM), Lake Shore model 7407, while the magnetic properties were investigated using a superconducting quantum interference device (SQUID), Quantum Design MPMS 3, from 10 to 300 K.

One-dimensional Fe₂O₃ and its corresponding carbon-coated samples were prepared according to the procedures shown in scheme 1, and the samples were characterized with BET, TEM, XRD, etc. Table 1 lists the prepared samples and their preparation conditions. Mesoporous α-Fe₂O₃ nanorods were initially obtained by calcination of 1D FeO(OH) precursors at 300 °C in air, which is denoted as Fe-300-Air. The FeO(OH) precursors were prepared by the microwave irradiation of Fe(III) nitrate and TEAOH. As seen in figure 1(A), the diameters of the nanorods were 10–25 nm and their lengths were about 40–200 nm. After hydrothermal coating with carbon, a core–shell structure with Fe₂O₃ nanorods as the core and carbon as the shell formed (figures 1(B)–(D)). However, some Fe₂O₃ nanorods were bundled together and not self-standing, probably due to the static hydrothermal treatment. The TEM images in figure 1 show a distinct carbon layer on the ferric oxide nanorod surface and its thickness can be readily tuned by changing the glucose/Fe₂O₃ ratio during the hydrothermal coating step or by increasing the coating cycle. This is evident by comparing figure 1(B), which displays iron oxide with single carbon coating Fe@C(1x), with figure 1(C) for double carbon coating (Fe@C(2x)) as the latter shows a thicker carbon layer than the former. Magnification of Fe@C(2x) (figure 1(D)) shows a carbon layer with thickness ranging from 4–8 nm. XRD analysis (the XRD patterns are not shown here) indicates that the iron oxide species in the Fe-300-Air sample was α-Fe₂O_3, which well matches the diffraction data (JCPDS 00-001-1053). It should be noted that the carbon coating step did not alter the α-Fe₂O₃ phase as the XRD pattern of Fe@C(1x) remained unchanged.

Zoom In Zoom Out Reset image size

The subsequent calcination of Fe@C(1x) and Fe@C(2x) at 500 °C in a N₂ atmosphere and exposure to air transformed the initial α-Fe₂O₃ phase to γ-Fe₂O₃. The samples are denoted as Fe@C(1x)-500-N₂ and Fe@C(2x)-500-N₂, respectively. Structural analysis was conducted by XRD and high-resolution TEM (HRTEM). Figure 2 compares the XRD patterns of Fe@C(1x)-500-N₂ (red line), Fe@C(2x)-500-N₂ (blue line), and Fe-300-Air (pink line). From the XRD patterns we can see the formation of γ-Fe₂O₃ for the calcined samples, which is in agreement with the standard γ-Fe₂O₃ diffraction data (JCPDS 00-004-0755) both in the peak positions and the relative intensities. It is observed there is no peak at 2θ = 18.2°, which corresponds to the Fe₃O₄ (111) plane (JCPDS 82-1533), suggesting Fe@C(1x)-500-N₂ and Fe@C(2x)-500-N₂ contain negligible Fe₃O₄ phases. The TEM and HRTEM images of Fe@C(1x)-500-N₂ after storage for 6 months are shown in figures 3(A) and (B). It is found that even after long-time storage at the ambient atmosphere, the morphology and the pore structure are still clearly discerned. From the HRTEM image shown in figure 3(B), the lattice spacing is measured to be 0.29 nm, which corresponds to the γ-Fe₂O₃ (220) plane. The selected area electron diffraction (SAED) pattern shown in the inset (between figures 3(A) and (B)) confirms the crystal structure of γ-Fe₂O₃, in good agreement with the above XRD results. The HRTEM image of Fe-300-Air (figure 3(D)) shows a lattice spacing of 0.25 nm, which is assigned to the α-Fe₂O₃ (110) plane, hence confirming the α-Fe₂O₃ phase of Fe-300-Air.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The surface area analysis is presented in table 2 and it shows that the initial α-Fe₂O₃ nanorods (Fe-300-Air) have a specific surface area of 132 m² g⁻¹, a pore volume of 0.398 cm³ g⁻¹, and a pore diameter of 11.7 nm, much larger than those of commercial non-porous α-Fe₂O₃, of which the surface area is usually between 10 and 20 m² g⁻¹. After carbon coating on the Fe₂O₃ nanorods (Fe@C(2x)), the surface area becomes 24.2 m² g⁻¹. However, after calcination in N₂ at 500 °C (Fe@C(2x)-500-N₂), the specific surface area increases to 162 m² g⁻¹, mainly due to the generation of micropores in the carbon layer, as indicated by the measured pore diameter (<2 nm).

XPS analysis was used to further determine the oxidation state of the Fe element in the composite (figure 4). It is observed that the Fe 2p_1/2 and Fe 2p_3/2 peaks of both Fe@C(1x)-500-N₂ (figure 4(A)) and Fe-300-Air (figure 4(B)) are accompanied by satellite peaks. The position of the satellite peak for Fe 2p_3/2 is determined to be ∼719.0 eV for both samples. The satellite peak is attributed to Fe3d–O2p hybridization and suggests that both samples have a composition of Fe₂O₃ [31]. No satellite peak is observed around the binding energy of 715.5 eV, which rules out the existence of Fe²⁺. Also, the absence of the characteristic Fe2p peak for metallic Fe at 706.6 eV excludes the presence of Fe⁰ clusters [32].

Zoom In Zoom Out Reset image size

Figure 5 shows the TGA results of Fe-300-Air, Fe@C(2x), and Fe@C(2x)-500-N₂ in air from room temperature up to 800 °C. Without carbon coating, Fe-300-Air hardly exhibits any weight loss in the examined temperature range, except for moisture loss from room temperature up to 200 °C. After double coating with carbon, the weight loss is very significant, mainly due to the combustion of carbon, which starts at 250 °C and ends at 500 °C. The carbon composition can be deduced from the weight loss of about 61.5 wt% before calcination in N₂ (Fe@C(2x)) and 53.2 wt% after calcination in N₂ (Fe@C(2x)-500-N₂). The decrease of carbon content after calcination is due to the removal of functional groups, such as –COOH, –OH, and –C=O, during calcination. This carbon amount is higher than that of the sample with single carbon coating (Fe@C(1x)-500-N₂), which is ca. 37%.

Zoom In Zoom Out Reset image size

Magnetic measurements are carried out to demonstrate the magnetic properties of the freshly prepared and 6-month aged samples (figures 6 and 7) by subjecting them to a VSM. It can be seen from the magnetic hysteresis loops (figure 6) that the 6-month old Fe@C(1x)-500-N₂ (black solid line) and the fresh sample (red solid line) have a saturation magnetization (M_S) of ∼47 emu g⁻¹, while Fe@C(2x)-500-N₂ (green solid line and blue solid line) has a M_S of ∼32 emu g⁻¹. It should be noted that the saturation magnetization here is calculated by using the total weight of each sample. The results show that there is no difference in the magnetization values for the freshly prepared samples and the aged ones, indicating excellent magnetic stability under ambient storage. By considering the weight of the carbon amorphous layer obtained from TGA, the corrected saturation magnetizations of sample Fe@C(2x)-500-N₂, as shown by the green and blue dashed lines, are ∼72 emu g⁻¹, quite close to the M_S value of bulk γ-Fe₂O₃ (76.3 emu g⁻¹), giving further evidence that the core phase is γ-Fe₂O₃ [33]. On the contrary, Fe-300-Air has a very small M_S of 0.4 emu g⁻¹, affirming that it is non-ferromagnetic α-Fe₂O₃ (the bulk M_S is 0.4–1.9 emu g⁻¹) [34]. In figure 7, Fe@C(1x)-500-N₂ and Fe@C(2x)-500-N₂ are further characterized by the SQUID from 10 to 300 K. Magnetic hysteresis measurement by the SQUID agrees well with the VSM measurement at room temperature. With the increase of temperature from 10 to 300 K, the measured saturation magnetization and coercivity of both samples are decreased due to thermal fluctuations of the magnetic moments and a reduced magnetocrystalline anisotropy field, respectively. The temperature dependence of magnetic moments between 10 and 300 K is also measured under field-cool (FC) and zero-field-cool (ZFC) conditions. In the FC measurement, the sample was first cooled down from 300 to 10 K under an applied magnetic field of 500 Oe. The magnetic moment was then measured as the temperature was raised from 10 to 300 K. In the ZFC measurement, the sample was cooled to 10 K without applying a magnetic field and magnetic measurement was then performed when heating the sample to 300 K. It can be seen in figures 7(C) and (D) that both samples do not show a Verwey transition at ∼120 K, which is a characteristic behavior of Fe₃O₄. The measurements exclude the presence of Fe₃O₄ and again suggest that the sample core is γ-Fe₂O₃.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The above results and our previous work [17] clearly demonstrate that the initially formed 1D α-Fe(O)(OH) was converted to 1D and porous α-Fe₂O₃ after calcination in air at 300 °C. This phase structure was maintained after the hydrothermal carbon coating. An evaluation of the FTIR spectra (figure 8) shows the characteristic Fe-O stretching bands in α-Fe₂O₃ at 445 and 529 cm⁻¹. After fresh calcination in N₂, because of the partial reduction, an Fe₃O₄ phase may be initially formed. This is evident from the FTIR spectra (figure 8) of the fresh Fe@C(2x)-500-N₂, which shows a single Fe-O vibration at 566 cm⁻¹ that is attributed to the Fe₃O₄ phase, as has been reported elsewhere [35, 36]. However, after exposure to air, this Fe₃O₄ phase was re-oxidized to γ-Fe₂O_3, suggesting oxygen molecules could effectively penetrate the carbon layer due to its porous nature and oxidize the Fe₃O₄ core. Dar et al [37] reported that the conversion of Fe₃O₄ to γ-Fe₂O₃ took place at ca. 200 °C, while that of γ-Fe₂O₃ to α-Fe₂O₃ occurred at ca. 500 °C. Jia et al [38] prepared single-crystal α-Fe₂O₃ nanorings and converted them to Fe₃O₄ after a reduction in H₂ gas at 360 °C, and the latter was converted to γ-Fe₂O₃ after exposure to air at 240 °C. However, this conversion is also dependent on the interaction between iron oxide and the support. As previously reported, the complete transformation of γ-Fe₂O₃ to α-Fe₂O₃ could be retarded to as high as 700 °C in the presence of amorphous Al₂O₃ support [39, 40]. Guo et al [23] also prepared magnetic composite spheres consisting of a magnetic core and a mesoporous carbon shell by direct replication from hollow mesoporous aluminosilicate spheres, followed by casting of iron nitrate solution, impregnation of the organic carbon source and carbonization, reduction of the α-Fe₂O₃ core to Fe₃O₄ and removal of the silica template. Compared to these literature methods, the route shown in this work is quite straightforward. The 1D α-Fe₂O₃ phase was reduced to Fe₃O₄ first, and the latter was re-oxidized into γ-Fe₂O₃ under ambient conditions within almost two weeks.

Zoom In Zoom Out Reset image size

Interestingly, the 1D morphology and porous structure are well maintained even after the two-stage phase change, which should be attributed to the good protective characteristic of the carbon layer. This carbon layer is inert, amorphous and porous, in close contact with the iron oxide core but not rigidly enough, so that it can accommodate structural and volume changes. The formation of γ-Fe₂O₃ is important as it is a well-known magnetic material with extensive applications. More importantly, there is no degradation in the magnetic property after long storage in air (>6 months) (figure 6). In a recent analysis by TEM after 1.5 years of storage, the structure and morphology are still well maintained. Therefore, this work demonstrates a feasible and scalable route to prepare robust and stable 1D magnetic materials, overcoming the long-standing technical problem that 1D structures, particularly magnetic ones, are usually not very stable upon aging/heating. Similarly, the successful preparation and surface carbon coating of 1D TiO₂ are shown in figures S1 and S2 in the supplementary material.

The adsorption study was performed by stirring Fe@C(2x)-500-N₂ as a sorbent in Au solution with different concentrations (200 to 5000 mg l⁻¹) for 20 h. The equilibrium adsorption curve in figure 9 was measured and used to calculate the maximum adsorption capacity of Au which is about 600 mg Au/g sorbent. By deducting the γ-Fe₂O₃ core, the maximum adsorption capacity is 1132 mg g⁻¹ carbon, which is almost the highest Au adsorption capacity by carbon reported so far. Similar adsorption behavior with high capacity was observed for a TiO₂@C system (figure S3) as compared to γ-Fe₂O₃@C, while both bare Fe₂O₃ and TiO₂ have poor adsorption capacity towards Au (below 50 mg Au g⁻¹ metal oxide). This indicates that the adsorption capacity of the composite nanostructures is mainly due to the surface carbon. It should be pointed out that, due to the presence of the magnetic core, the sorbent can be readily separated from the solution by simply applying a magnetic field. Figure 10 shows the adsorption properties of Fe@C(2x)-500-N₂ towards Au³⁺, Cr³⁺, Cu²⁺, Mg²⁺, Ni²⁺ and Zn²⁺ ions. It was observed that the sorbent exhibits high adsorption selectivity to Au³⁺ but very low selectivity to other metal cations. The metal removal efficiencies for these metal ions are given in table 3, where Au³⁺ has the highest removal efficiency of 98%, while those of the other metals are below 4%. Mpinga et al [41] tested the adsorption behaviors of activated carbon for the cyanide salts of Au, Pt, Pd, Ni and Cu, and found the affinity of activated carbon is in the following sequence: Au > Pt > Pd > Ni > Cu. Zheng et al [42] utilized carbon material derived from banana peel for the adsorption of metal ions, which exhibits strong selectivity towards Au(III) while negligible affinity to other base metals, such as Cu(II), Pb(II), Ni(II), and Fe(III). The carbon material has the maximum Au adsorption capacity of 801.7 mg g⁻¹ carbon, lower than what we obtained. In this case, one possible explanation is that the carbon is coated and dispersed on the surface of metal oxides, which exposes more adsorption active sites to Au as compared with bulk carbon or carbon not coated on a 1D nanostructure. In addition, glucose-derived carbon probably generates more defects acting as active sites for Au adsorption.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The mechanism for Au ion adsorption is still not very clear. Traditionally, the adsorption of Au cyanide by activated carbon is believed to be due to the formation of Mⁿ⁺-[Au(CN)₂^-], where Mⁿ⁺ is a metal ion, such as Na⁺, K⁺ and Ca²⁺. The adsorbed Au species is finally reduced to a Au cluster [43]. Zheng et al [42] believed that –OH, –NH₂ functional groups on the carbon surface are active for the selective adsorption of Au(III) ions. In the templated synthesis of Au nanoparticles on glucose-derived carbon species, it was confirmed that the –OH, C-O-C functional groups played an important role in anchoring Au species [30]. Wang et al [28] also prepared carbon spherules by hydrothermal treatment of glucose and observed the highly selective adsorption of Au(III) against Pd(II), Pt(VI), Rh(III), Fe(III), Cu(II) and Ni(II) ions. High selectivity towards Au(III) is also proposed due to the presence of surface –OH, C-O-C and –COOH groups, which are able to reduce Au(III) to Au⁰, as evident in TEM and XPS analysis. On the other hand, carbon nanotubes are also reported to be active in adsorbing some organic molecules and their interaction forces with the adsorbates include hydrophobic interaction, electrostatic interaction, etc [44]. In our case, sample calcination at 500 °C in N₂ largely removes surface –OH and –COOH functional groups and generates amorphous carbon with a lot of defects. In contrast to conventional carbon materials, during the calcination process, glucose-derived carbon materials release a lot of gas molecules, such as CO₂ and H₂O, which will damage the integrity of the carbon structure and generate defects. It is reflected in the dramatic increase in surface area after calcination (table 2). As shown in figure 10, both the uncalcined and the calcined samples adsorb Au species with high capacity, but the former have a much lower adsorption rate than the latter. In the former, adsorption is mainly via surface groups, while in the latter, the formed structural defects should also exhibit significant contribution to Au adsorption besides the surface groups. The carbon defects are electron-rich centers which can adsorb molecular oxygen to form ${{{\rm{O}}}_{2}}^{-},$ ${{{\rm{O}}}_{2}}^{2-}$ species [45]. The latter can transfer electrons and further reduce Au(III) ions. In addition, these defects can interact with water molecules and generate new surface functional groups, which may attract Au(III) cations and promote Au reduction after several hours of stirring. The other metals such as Ni(II) and Cr(III) could not be reduced into their metal state, hence leading to weak interaction between the metal ions and the carbon surface, which renders low adsorption selectivity. Among the noble metals, Au(III) is highly reducible (metallic Au is the most stable state), thus leading to the highest adsorption selectivity. The reduction of Au(III) to metallic Au is confirmed by TEM analysis. Figure 11 shows the TEM images of the collected Fe@C(2x)-500-N₂ after the adsorption study in 1000 and 2500 mg l⁻¹ Au solution. It can be observed that there are a lot of metallic Au particles distributed on the carbon surface. In the case of the 1000 mg l⁻¹ Au solution, the Au nanoparticles are quite uniform and usually below 5 nm in size, while in the latter case, the Au particles are larger, about 10–60 nm in size, which not only demonstrates that this composite is a very good sorbent for the selective recovery of Au from aqueous solution, but also confirms that the high selectivity towards Au is related to the high reducibility of Au cations. It is known that for the supported Au catalysts, the catalytic performance is usually dependent on the Au particle size, so this method provides a facile way to prepare supported Au catalysts on carbon with tunable Au particle size, which may have important application in catalysis [45]. In addition, the carbon structure on oxide materials may have changed electronic properties [46, 47]; hence it is possible that the high selectivity of 1D γ-Fe₂O₃@carbon is associated with the presence of a γ-Fe₂O₃ core, which should be further investigated in the future.

Zoom In Zoom Out Reset image size

Although the adsorbed Au particles can be recovered by washing the adsorbent with thiourea solution [48], we combusted the Au-adsorbed Fe@C(2x)-500-N₂ in air at 600 °C to remove the carbon layer followed by dissolution of the γ-Fe₂O₃ core in HCl solution for a direct TEM observation of the Au particles. As a result, pure metallic Au particles can be retrieved, as shown by the EDX analysis in figure 12. Before the removal of carbon and iron oxide (figure 12(A)), strong C and Fe spectra are detected together with the Au spectra. After combustion and acid dissolution, the EDX signal for C drops dramatically while the signals for Fe and O completely disappear as seen in figure 12(B), leaving only strong Au signals together with Cu spectra that come from the copper grids. The TEM image shows that the recovered Au nanoparticles have sizes of 20–40 nm. Therefore, this confirms that a straightforward Au recovery method can be effectively performed after the adsorption procedure.

Zoom In Zoom Out Reset image size