Structural Characterization of MFe₂O₄-montmorillonite

In the XRD pattern of MFe₂O₄-montmorillonite (Fig. 1), the diffraction peaks at 6.5°2θ and 19.8°2θ were assigned to the (001), (101) faces, respectively, consistent with montmorillonite-15A (JCPDS No: 29-1498) (Chang et al. Reference Chang, Ma, Ma, Zhang, Qiao and Hu2015). Besides, the characteristic reflections at 18.2, 30.0, 35.4, 43.0, 53.4, 56.9, and 62.5°2θ, corresponding to the crystal surfaces of (111), (220), (311), (400), (422), (511), and (440), respectively, demonstrate the existence of the MFe₂O₄ pure phase (Fe₃O₄, JCPDS-No: 19-0629; MnFe₂O₄, JCPDS-No: 10-0319; CoFe₂O₄, JCPDS-No: 22-1086; ZnFe₂O₄, JCPDS-No: 89-1010 and NiFe₂O₄, JCPDS-No: 87-2338) (Wang et al. Reference Wang, Li, Wang, Zhao and Jiang2012a; Hu et al. Reference Hu, Gao and Yang2008). Moreover, other characteristic peaks were not observed, suggesting that as-synthesized MFe₂O₄-montmorillonite with good crystallinity was prepared successfully.

Fig. 1 XRD patterns of MFe₂O₄-montmorillonite

To further demonstrate the bond structures of the MFe₂O₄-montmorillonite, the bonding patterns data were analyzed with FTIR and Raman spectroscopies. The double absorption peaks near 594 cm^–1 are assigned to the stretching band of Fe–O in MFe₂O₄ octahedra and tetrahedra (Fig. 2a), and the other typical bands of MFe₂O₄ were also observed at 1380 cm^–1 and 2920 cm^–1. Moreover, the frequency of 1090 cm^–1 represents the stretching bands of Si–O in montmorillonite, and 3630 cm^–1 represents the octahedral O–H stretching bands. The other absorption peaks at 3430 cm^–1 and 1640 cm^–1 are assigned to the O-H stretch vibrations and the H-O-H bending vibration of H₂O, respectively. In the Raman patterns (Fig. 2b), the signature peaks of E_g and A_1g bands in MFe₂O₄ appeared at 303 cm^–1 and 667 cm^–1, respectively. In addition, the Raman shift around 1600 cm^–1 was observed, and the weak peak was predicted to be attributed to the long-range electronic transition bonds of AlO–OFe(M), which suggests the formation of interface oxygen vacancies that could capture the -OH groups from water molecules (Li et al. Reference Li, Yi, Tang, Liu, Wang and Cui2016). From the data above, the as-synthesized samples consisted of MFe₂O₄ and montmorillonite.

Fig. 2 (a) FTIR patterns and (b) Raman spectra of MFe₂O₄-montmorillonite

SEM and TEM examinations of MFe₂O₄-montmorillonite were performed to explore the morphology and interface structure properties of samples. A typical SEM image of MFe₂O₄-montmorillonite revealed that the magnetic Fe₃O₄ nanoparticles grew on montmorillonite layer surfaces with unavoidable agglomeration (Fig. 3a). Fe₃O₄ (–34690.42 eV) had a lower crystal energy than MFe₂O₄ (–33000.22 eV, –35432.89 eV, –38601.21 eV, –36106.85 eV, M = Mn, Zn, Ni, or Co) (Bian et al. Reference Bian, Li, Li, Nie, Dong, Dong, Bian, Li and Li2017). Hence, the inhibited agglomeration could be observed clearly, whereas the average particle size increased from 60 to ~200 nm. In the TEM images, the agglomeration of nanoparticles was observed more intuitively, and the MFe₂O₄ grew primarily on the edge sites of montmorillonite (Fig. 4a). The relative lattice distances (d) and crystal faces were measured based on HRTEM images (Fig. 4b), as shown in Table 1. Compared with the reported values, the difference of MFe₂O₄-montmorillonite faces ranged from 0.01 to 0.03 nm. For the MFe₂O₄ crystal, the typical sites for Fe²⁺ in Fe₃O₄ were occupied by M ²⁺ ions. Due to the differences in possible atomic radius of M ²⁺ ions (0.153 nm to ~0.167 nm) and Fe²⁺ (0.179 nm) (Sun et al. Reference Sun, Zeng, Robinson, Raoux, Rice, Wang and Li2004; Gao et al. Reference Gao, Liu, Zhu, Wang and Xie2016), Fe-O tetrahedral shrinkage affected the lattice of Fe₂O₃ octahedra nearby, decreasing the lattice distances of (111)/(311) by 0.01 nm and increasing that of (220) by 0.01 nm. The surface energy (1.06 to ~1.18 J·m^-2) of the (111) face in MFe₂O₄ was less than that of a high-index (311) face (Yu et al. Reference Yu, Huo, Li, Wang and Jiao2012), thus the MFe₂O₄(111) combined preferentially with the exposed Al₂O₃(100) face of the montmorillonite edge sites (Hong et al. Reference Hong, Kim, Fang, Hong and Chiang2017). In addition, the data reflected the good matching with MFe₂O₄(111) faces and exposed Al₂O₃(100). Furthermore, HRTEM-EDS analysis was conducted to analyze the element distribution. Fe and O on the montmorillonite surfaces proved the uniform distribution of MFe₂O₄ (Fig. 3b). In addition, the results revealed that the atomic ratio of Si:Al reached 18:1, indicating that the Al³⁺ in the montmorillonite crystal lattice was often replaced by some low-valence cations in solution (e.g. Fe²⁺, Zn²⁺, Mg²⁺) (Brigatti et al. Reference Brigatti, Galán, Theng, Bergaya, Theng and Lagaly2013). However, the exposed Al-O octahedron on the edge of montmorillonite still exhibited high activity, and it could provide active electrons for the adsorption reaction (Hennig et al. Reference Hennig, Reich, Dahn and Scheidegger2002).

Fig. 3 (a) SEM images of MFe₂O₄-montmorillonite and (b) HRTEM-EDS images of MnFe₂O₄-montmorillonite

Fig. 4 (a) HRTEM and (b) TEM images of MFe₂O₄-montmorillonite

Table 1 The lattice distance (d) data and crystal faces of MFe₂O₄-montmorillonite

F represents MFe₂O₄ and M represents montmorillonite.

Experment^a-d: d data and crystal planes reported from HRTEM (Phumying et al. Reference Phumying, Labuayai, Swatsitang, Amornkitbamrung and Maensiri2013; Liu et al. Reference Liu, Fang, Jing, Li, Li and Chen2014a; Gao et al. Reference Gao, Liu, Zhu, Wang and Xie2016; Yan et al. Reference Yan, Li, Yu, Shan, Du and Liu2016).

Present work: HRTEM results.

The surface composition of MFe₂O₄-montmorillonite was further demonstrated using XPS. The characteristic peaks appearing at 72.7 eV, 99.5 eV, and 150.8 eV are assigned to Al 2p₃ (Al³⁺), Si 2p₃, and Si 2s (Si⁴⁺) in the montmorillonite carrier, and the 725.5 eV corresponded to Fe 2p_3/2 (Fe³⁺) ions coexisting in the MFe₂O₄ (Fig. 5). Moreover, an interesting peak of 712 eV, indicating the presence of Fe 2p_1/2 (Fe²⁺), was observed in each MFe₂O₄-montmorillonite. This abnormal phenomenon could occur as the active electrons transferred to MFe₂O₄, as mentioned previously, resulting in some Fe³⁺ ions being reduced to Fe²⁺. The other characteristic peaks of Mn 2p (Mn²⁺), Co 2p (Co²⁺), Ni 2p (Ni²⁺), and Zn 2p (Zn²⁺) were demonstrated clearly with binding energies at 641 eV, 784eV, 855 eV, and 1022 eV, respectively. The detailed peak-fitting data are shown in Supplementary Fig. S2. The presence of M ²⁺ is thus strong evidence for homogeneous bonding in MFe₂O₄, and the samples consisted of montmorillonite with MFe₂O₄.

Fig. 5 XPS spectra of MFe₂O₄-montmorillonite

Electron-Transfer Mechanism of MFe₂O₄-montmorillonite

In general, adsorption behavior was affected by the surface potential of the adsorbent. In the montmorillonite, the interlayer potential showed a permanent negative charge by isomorphic substitution (Zuo et al. Reference Zuo, Gao, Yang, Zhang, Chen, Li, Shi and Wu2017); the potentials of MFe₂O₄ were reported as 17.5 to ~20 mV (M = Mn, Fe, Zn) and –15 to ~–30 mV (M = Co, Ni) at pH=5–6 (Chen et al. Reference Chen, Wen, Kong, Tian, Ding and Xiong2014; Jiang et al. Reference Jiang, Sun, Xu, Lu and Shi2016; Liu et al. Reference Liu, Yang and Zhang2016; Tu et al. Reference Tu, Chan, Tu, Wang, You and Chang2016; Deng et al. Reference Deng, Chen, Lu, Ma, Feng, Gao and Li2017). Moreover, the zeta potential of MFe₂O₄-montmorillonite was measured as –6.5 to ~26 mV at pH 5.5 (Fig. 6a), and the potential decreased with the rise in pH because increasing OH^- was adsorbed to the edge surfaces of the montmorillonite (Min and Lee Reference Min and Lee2010). Thus, the MFe₂O₄ on the surface of montmorillonite was vital for the surface potential of the adsorbent (Fig. 6b). At the interfaces, M ²⁺ in MFe₂O₄, an electron acceptor, could provide an oxygen vacancy and lead to active electrons at the montmorillonite edge sites being transferred to compensate the electron holes (Bian et al. Reference Bian, Song, Dong, Duan, Xu and Li2015). Thus, MFe₂O₄-montmorillonite exhibited greater positive surface potential than Fe₃O₄-montmorillonite. In addition, the zeta potential was used to describe the dispersion stability of the adsorbent in aqueous solution. The potential of MFe₂O₄-montmorillonite (M = Co, Ni, Zn) was ascertained to be 8.9 to ~41 mV, which indicates that the adsorbent systems were stable. The pH of zero potential for Fe₃O₄-montmorillonite was ~5.5, suggesting the interactions between metal ions and adsorbent were easier to enhance at pH < 5.5. Thus, the adsorption experiments were carried out at pH 5.5.

Fig. 6 (a) Zeta potential curves and (b) surfaces/interfaces/interlayer charge distribution of MFe₂O₄-montmorillonite

The UV-Vis diffuse reflectance spectra were obtained to investigate further the mechanism of electron transfer at the MFe₂O₄-montmorillonite surface and interface. Also, the band gaps were estimated by the photon energy equation:

(6) $\mathrm{Eg}=\frac{h}{k}\times \frac{C}{\lambda }$

where h denotes Planck’s constant; k is a constant (1.6×10^–19 J·eV^–1); C/λ is wave frequency based on the tangent line; Eg is band gap energy (eV).

The band gap of montmorillonite was calculated as 2.5 to ~3.06 eV and MFe₂O₄ was ~1.62 eV (Fig. 7a). Due to the particle confinement effect as reported by Fatimah et al. (Reference Fatimah, Wang and Wulandari2011), the band gap energy of MFe₂O₄-montmorillonite decreased, and the peak shifted to a higher wavelength compared with Fe₃O₄-montmorillonite. This result implied that the addition of MFe₂O₄ led to an increase in particle size. It is noteworthy that a small band gap (0.78 to ~0.82 eV) appeared near 1900 nm. Electron holes could be formed due to the presence of M ²⁺ mediators to receive and transmit electrons which were provided by the electron donor (Fig. 7b) (Rekhila et al. Reference Rekhila, Trari and Bessekhouad2017), this suggested that the interfaces of MFe₂O₄-montmorillonite were the main region for electron transfer.

Fig. 7 (a) DRS curves of MFe₂O₄-montmorillonite and (b) the relative electron transfer reaction

The interface properties of MFe₂O₄-montmorillonite were supported by THz-TDS spectra. The peak widths of equipment-emitted pulses were <0.65 ps, and the amplitudes were nearly 100 to ~150 V (Fig. 8a). With the strong absorption of terahertz rays, light could pass easily through the montmorillonite edge sites without changing velocity (Li et al., 2017). In the refraction index information, the two absorption peaks were near 1.75 THz, which are assigned to the photon adsorption of the O-Fe (2p_1/2) octahedron and O-Fe (2p_3/2) tetrahedron in the MFe₂O₄ (Fig. 8b). The refraction peaks at ~2.15 THz reflected the MFe₂O₄(111)-montmorillonite(100) interfaces, because poorly active electrons at the interfaces were triggered and detected easily. In addition, note that MFe₂O₄-montmorillonite showed a smaller refractive index peak and strong absorption at 1.95 to ~2.15 THz in the absorption spectrum (Fig. 8c). This abnormal phenomenon was predicted as the electronic transition within the interfacial bonding of AlO-OFe(M) and SiO-OFe(M) between the montmorillonite and MFe₂O₄, associated with the synergistic hybridized orbitals effect of O-O sp², etc. The tangent data were calculated to explain better the electronic mechanism of the interfaces, according to the frequency difference (Fig. 8d). The tangent data revealed that the concentration of surface oxygen vacancies of MFe₂O₄(111) increased by the paired electrons of M ²⁺. Therefore, the active electrons of MFe₂O₄(111)-montmorillonite(100) interfaces were of significance to the adsorption of UO₂ ²⁺-X ⁿ⁺ from solution.

Fig. 8 (a) Time dependence, (b) refraction index, (c) absorption, and (d) frequency difference (FD) power of THz signal spectra of the MFe₂O₄-montmorillonite interfaces, where time-resolved spectra of semicoke were obtained via THz-TDS. The time resolution was on a picosecond scale

Competitive Adsorption of Toxic Trace Metal Ions Using MFe₂O₄-montmorillonite

To evaluate the adsorption capacity of the adsorbent, data were obtained through ICP measurements. The results revealed that the equilibrium adsorption capacity increased with the increase in initial metal ion concentration (Fig. 9). The total adsorption capacities of UO₂ ²⁺ and metal ions were 322 to ~396 mg·g^–1 in the initial ionic concentration of 60 mg·L^–1 (Supplementary Table S1). Subsequently, Langmuir, Freundlich, and Dubinin-Radushkevich models were adopted to describe the adsorption behavior (Table 2). The Langmuir linear correlation coefficients (R²) were calculated as 0.913 to ~0.999, larger than those of Freundlich 0.774 to ~0.999. The Langmuir isotherm fitted the experimental data better. The results revealed that Rb⁺, Sr²⁺, Mn²⁺, Ni²⁺, Zn²⁺, Cd²⁺, and Cr³⁺ ions were primarily at the adsorbent surfaces as a result of monolayer physical adsorption. However, the values of R² for UO₂ ²⁺ were approximately 0.897 to ~0.989 (Freundlich) and 0.739 to ~0.977 (Langmuir). The exception showed that the adsorption of UO₂ ²⁺ was more consistent with the Freundlich isotherm, without that of Fe₃O₄-montmorillonite. The Freundlich constants (1/n) were associated with the adsorption intensity, and the constants of 1/n of X ⁿ⁺ were calculated as 0.5 to ~3.2 (0.5<1/n<1), which indicated that the adsorption of Rb⁺, Sr²⁺, Mn²⁺, Ni²⁺, Zn²⁺, and Cd²⁺ ions would be favorable. The constants of 1/n of X ⁿ⁺ UO₂ ²⁺ adsorption were 0.09 to ~0.37, within the range of 0.1 < 1/n < 0.5, which suggests that the UO₂ ²⁺ was more significantly adsorbed on the adsorbent. In addition, the linear correlation coefficients (R²) of the D–R isotherm for X ⁿ⁺ and UO₂ ²⁺ were determined as 0.913 to ~0.997. This indicated that X ⁿ⁺ and UO₂ ²⁺ ions were also adsorbed well in the interlayers of montmorillonite. On the whole, the difference in adsorption-fitting results was associated with the competitive adsorption between UO₂ ²⁺ and X ⁿ⁺.

Fig. 9 The equilibrium adsorption capacity data of MFe₂O₄-montmorillonite from heavy metal mixed solution (ppm, parts per million, indicates the concentration of the solute)

Table 2 Isotherm parameters for the uranyl-toxic metal MFe₂O₄-montmorillonite

The adsorption mechanism was considered to be by ligand exchange of hydroxyl groups and an adsorption reaction caused by the transition of electrons in the long-range AlO–OFe(M) bonds, creating surface hydroxyl radical layers in aqueous solution and changing the adsorption behavior from monolayer to multilayer. In the adsorption behaviors, involving successive mono electronic transition, UO₂ ²⁺-X ⁿ⁺ adsorption could be assumed to be the direct electron transfer for increasing •OH and the complexation of •OH with UO₂ ²⁺ or X ⁿ⁺ (Karamanis Reference Karamanis1997). According to previous reports (Xu et al. Reference Xu, Zhang, Wang, Xu, Ding and Li2013), the UO₂ ²⁺ and X ⁿ⁺ ions underwent competitive adsorption on the surfaces of montmorillonite, where the adsorption reaction of surface •OH with positively charged X ⁿ⁺ could form an ion monolayer to block UO₂ ²⁺ from being adsorbed. The adsorption results implied that the adsorption positions of UO₂ ²⁺ and X ⁿ⁺ were not the same as those on the surfaces of MFe₂O₄-montmorillonite, and the adsorption of UO₂ ²⁺ occurred primarily at interfaces. Take ZnFe₂O₄-montmorillonite as an example. The adsorption of X ⁿ⁺ overall fitted the Langmuir and Freundlich models (Fig. 10a), whereas the maximum adsorption capacity of Freundlich was much lower than the actual adsorption capacity, which revealed that the X ⁿ⁺ were adsorbed primarily by monolayer physical adsorption on the adsorbent surfaces. However, the Freundlich model described the adsorption of Cr³⁺ and UO₂ ²⁺ better than the Langmuir model (Fig. 10b), and the saturated UO₂ ²⁺ adsorption capacity of different models differed significantly. Thus, the adsorption positions of UO₂ ²⁺ were primarily at ZnFe₂O₄-montmorillonite interfaces.

Fig. 10 (a) The fitting curves of Sr²⁺ and (b) UO₂ ²⁺, Cr³⁺ to Freundlich, Langmuir, and Dubinin-Radushkevich models at ZnFe₂O₄-montmorillonite surfaces

The maximum adsorption capacity was calculated from the adsorption models. The results revealed that MFe₂O₄-montmorillonite had a better adsorption capacity of metal ions using Langmuir and D-R equation fitting (Fig. 11a-b). However, the adsorption capacities of UO₂ ²⁺ and Cr³⁺ were much greater than that of other X ⁿ⁺ ions using the Freundlich model (Figure 11c). The data demonstrated the highly selective adsorption behavior of UO₂ ²⁺ and Cr³⁺ ions in MFe₂O₄-montmorillonite interfaces (Fig. 11d); the illustration of the adsorption mechanism has been mentioned. The interface adsorption capacity of UO₂ ²⁺ was 25.1 mg·g^–1 and that of Cr³⁺, 60.2 mg·g^–1. Additionally, the maximum adsorption capacity of Cr³⁺ from the Langmuir isotherm was 163.8 mg·g^–1. Compared with previous reports, the adsorption capacities of UO₂ ²⁺ and Cr³⁺ were just 5 to ~33.7 mg·g^–1 using montmorillonite and other modified Fe₃O₄, and the MFe₂O₄-montmorillonite exhibited a large adsorption capacity of UO₂ ²⁺, as listed in Table 3. Thus, the medium (M ²⁺) and material modifications of Fe₃O₄ were proven as one of the effective ways to enhance adsorption capacity. Though the adsorption reaction was known to be a spontaneous and endothermic process, the maximum adsorption capacities of X ⁿ⁺ ions using Langmuir (39.7 to ~449.4 mg·g^–1) were greater than the capacities using the Dubinin–Radushkevich equation (15.1 to ~95.4 mg·g^–1). Thus, the weak electrostatic attraction and the hydroxyl adsorption reaction between MFe₂O₄-montmorillonite and X ⁿ⁺ ions remained the major process for metal-ion adsorption, compared with adsorption by interlayer ion exchange.

Fig. 11 Maximum adsorption capacities of uranyl and toxic metal ions at MFe₂O₄-montmorillonite surfaces/interfaces/interlayer. (a) Langmuir, (b) Dubinin-Radushkevich, (c) Freundlich models, and (d) Illustration of the selective adsorption mechanism

Table 3 Comparison of UO₂ ²⁺ and Cr³⁺ ion-adsorption performance among various adsorbents

HRTEM-EDS images were used to verify the presence of X ⁿ⁺ and UO₂ ²⁺ ions after adsorption. The elemental distribution of Fe₃O₄-montmorillonite was observed, and the UO₂ ²⁺/X ⁿ⁺ ions were found to have been adsorbed uniformly in the sample (Fig. 12a). Ignoring dissociation of the X ⁿ⁺ ions, the elemental distribution of MFe₂O₄-montmorillonite after desorption showed a strong response to the existence of UO₂ ²⁺ and Cr³⁺ (Fig. 12b). Thus, the collaborative adsorptions of UO₂ ²⁺ and Cr³⁺ should be noted. The UO₂ ²⁺ preferentially compensated electron holes and adsorbed at the interfaces compared with Cr³⁺ metal ions, reported as Cr³⁺4s₁ electron orbital, which is easier to protonate than the U 5f₃6d₁ orbital (De et al. Reference De Decker, Folens, De Clercq, Meledina, Van Tendeloo, Du Laing and Van Der Voort2017). Consequently, the selective adsorption of UO₂ ²⁺ was enhanced through the active electron/hole regulation by the M ²⁺ ionic medium modification, and the adsorption of UO₂ ²⁺ was treated as favorable.

Fig. 12 (a) HRTEM-EDS of uranyl-toxic metal ions in Fe₃O₄-montmorillonite and (b) HRTEM-EDS of uranyl-toxic metal ions in MFe₂O₄-montmorillonite after desorption, where “at.%” means the atom ratio

The charge transition of interaction between UO₂ ²⁺ and X ⁿ⁺ was characterized using cyclic voltammetry. The appearance of the reduction peak revealed that UO₂ ²⁺ ions could be reduced to UO₂ ⁺ by gaining electrons when the potential was scanned from positive to negative (Fig. 13a). However, only a portion of UO₂ ⁺ oxidized back to UO₂ ²⁺ compared with the integral area of the redox reaction, suggesting that UO₂ ²⁺ was oxidized heterogeneously to UO₂ ⁺ (Singer et al. Reference Singer, Chatman, Ilton, Rosso, Banfield and Waychunas2012) and the adsorption of UO₂ ²⁺ at the interfaces was stable. The X ⁿ⁺ ions did not undergo redox reaction throughout the cycle, because the exchange of active electrons took place primarily in the interface layer. For instance, the oxidation-reduction peaks of Sr²⁺ ions did not appear (Fig. 13b). However, the integral reduction area of Cr³⁺ ions was expanded, suggesting that some of the Cr³⁺ was reduced to Cr²⁺. In addition, the coupling of electrons with Cr²⁺-Cr³⁺ co-orbitals on the interfaces led to inhibition of the oxidation. The total peaks were obtained through electric signals covering (Fig. 13c), and the total oxidation signal was weakened. This situation predicted that active electrons transferred from Cr³⁺ to UO₂ ⁺ on the interfaces and formed UO₂ ²⁺-UO₂ ⁺-Cr³⁺-Cr²⁺ redox reactions (Fig. 13d). Hence, the selective adsorption of UO₂ ²⁺ in the MFe₂O₄-montmorillonite interfaces was not only associated with electrostatic interaction but was also affected by complexation of Cr³⁺ ions.

Fig. 13 Cyclic voltammograms of FTO glass electrode in 60 ppm solute concentrations. (a) UO₂ ²⁺, (b) Cr³⁺-Sr²⁺, (c) the mixed uranyl-toxic metal ion solution (pH 5.5), and (d) the charge transition mechanism at MFe₂O₄(311)/(111)-montmorillonite(100) interfaces. Therein, the scan rate was 25 mV∙s^–1, “O” and “R” represent the oxidation and reduction peaks, respectively (in parts a–c, the horizontal axis, E (V vs. SCE) represents the scanning potential and the vertical axis, (J (mA∙cm^–3)), the current density (i.e. the ratio of the current to the interface area of the electrode at one potential)