Introduction

There are various nuclides in nuclear wastewater, most of which have a short half-life and low radioactivity [1,2,3]. However, cesium (Cs) has drawn much attention owing to its strong γ radiation, long half-life (37Cs: 30.17 years, 134Cs: 2.06 years) and easy mobility (as one high-yield (6.09%) fission product of uranium (U235)) [3, 4]. Sharing similar property with potassium, sodium and calcium, cesium can easily enter human body and cause severe health issues. Therefore, Cs has been deemed as one of the most hazardous nuclides, and its cleanup is an imperative task in the view of environment protection [5].

Tremendous efforts have been devoted to exploring efficient technologies for removal of cesium [6, 7]. These technologies mainly include solvent evaporation, membrane filtration, electrochemical treatment, adsorption and bioremediation, and adsorption has been considered as one of most promising methods for cesium elimination [8, 9]. To date, various materials have been reported for cesium adsorption, including clays, zeolites, crystalline silicotitanates, layer doubled hydroxides, magnetic adsorbents and Prussian blue-based materials [10].

Prussian blue and its analogues (PB and PBAs), also known as transition metal hexacyanoferrates (HCF), are generally built from transition metal center (Fe, Co, Ni, Cu and Zn) coordinating with cyanide ligand [11,12,13]. They mainly display a cubic crystal structure and can serve as an ion-sieve in selectively capturing ions with match size (3.2 Å) [14, 15]. Due to high affinity and selectivity for cesium (3.29 Å), this type of materials has been extensively used for removal of cesium [16, 17]. However, pristine PB or PBAs synthesized by precipitation method were ultrafine powder and almost inseparable from water, and this intrinsic drawback severely limited their practical application [18].

To overcome this difficulty, much work has been conducted to develop a variety of porous supports to immobilize PB/PBAs, like biopolymers (chitosan and alginate), porous silicas, magnetic materials and polyvinyl alcohol sponges [18,19,20,21]. The immobilization strategy has been demonstrated to be a powerful strategy in addressing the separation problem.

It is well known that there are many oxygen-based functional groups in the graphene oxides, and these groups can contribute to cation adsorption. Presently, graphene and its derivatives as excellent support materials with robust porous network, have received particular attention [22, 23]. Graphene-based nanomaterials display a two-dimensional morphology, and they could be easily assembled into three-dimensional porous aerogel which is significant for separation from solution [22, 24].

The assembly of graphene oxide (GO) can be induced by many methods [25, 26]. For instance, Fe ions have been investigated for fabrication of magnetic GO aerogel [27, 28]. Polyvinyl alcohol (PVA), chitin/chitosan and DNA can also be used as cross-linkers [29,30,31]. Besides physical methods, chemical reduction is efficient for GO assembly [32]. These resultant GO aerogels possess a large surface area, a structure of abundant pores and low density, and have been intensively investigated as excellent adsorbent-supports [33, 34].

Thus, it is feasible and efficient to integrate good selectivity of PBAs for cesium and easy separability of aerogels from water. The composites mitigating the drawback of PBAs would show promising in practical applications. To the best of our knowledge, there are few studies available on the combination of PBAs and GO aerogels for cesium elimination.

In this study, Ni ions and PVA were chosen as co-cross-linkers for GO self-assembly, and then Ni ions coordinated with [Fe(CN)6]3− to form precipitate of KNi[Fe(CN)6] in situ. The resultant material (denoted as PVA/GO/KNi[Fe(CN)6]) displayed a hierarchical structure composed of GO sheets decorated with nanoparticles. Moreover, the adsorption performance of cesium by PVA/GO/KNi[Fe(CN)6] was investigated including kinetics, isotherms and effects of solution pH and coexisting ions. The immobilization mechanism was also explored by various characterizations.

Experimental section

Materials and reagents

Polyvinyl alcohol (PVA, Mw ~ 145,000) was supplied by Aladdin Chemical Reagent Co., Ltd. Graphene oxide (GO) dispersion (10 mg/mL) was purchased from Ketan New Materials Co., Ltd. Potassium hexacyanoferrate (K3[FeIII(CN)6]), nickel sulfate (NiSO4·6H2O), cesium chloride (CsCl), sodium nitrate (NaNO3), potassium nitrate (KNO3), magnesium nitrate (Mg(NO3)2·2H2O) and calcium nitrate (Ca(NO3)2·6H2O) were purchased from Sinopharm Chemical Reagent Co., Ltd. All reagents were of analytical grade, and used directly when received. The non-radioactive standard solution of cesium (1000 mg/L) was purchased from General Research Institute for Nonferrous Metals (Beijing, China). Cesium stock solution was prepared by dissolving CsCl in deionized water.

Synthesis of PVA/GO/KNi[Fe(CN)6]

The detailed synthesis procedures of PVA/GO/KNi[Fe(CN)6] aerogel were described as follows [35, 36]:

Firstly, 0.2 g of NiSO4·6H2O and 1.6 g of PVA were dissolved in 100 mL deionized water with vigorous stirring at 80 °C for 2 h. A mixed cross-linker solution was thus obtained. Twenty mL GO dispersion and 5 mL deionized water were blended and kept under ultrasonic treatment for 30 min. Then, 2.5 mL diluted dispersion was transferred to a small vessel. And 0.625 mL of cross-linker solution was dropwise added and stirred rapidly with a fine glass bar. Several seconds later, a hydrogel (PVA/GO/Ni2+) finished cross-linking and was shaped. By freeze-drying (24 h), the hydrogel (PVA/GO/Ni2+) would be transformed into its counterpart aerogel.

Secondly, the aerogel (PVA/GO/Ni2+) was totally immersed into 0.1 mol/L K3[Fe(CN)6] solution and kept gently stirring at 50 °C for 6 h, making Ni2+ completely converted into KNi[Fe(CN)6]. After that, the as-synthesized hydrogel (PVA/GO/KNi[Fe(CN)6]) was taken out from solution by the aid of tweezers, rinsed repeatedly with deionized water and dried for 24 h in a lyophilizer. Thus, the PVA/GO/KNi[Fe(CN)6] aerogel was obtained.

Material Characterization

The microstructure and surface elements of as-prepared materials were analyzed on a Hitachi SU-8010 field emission scanning electron microscope (FESEM) equipped with an energy-dispersive X-ray detector (EDX). Transmission electron microscope (TEM) images were observed on a Hitachi-7700 electron microscope. Powder X-ray diffraction (PXRD) patterns were recorded on an X′Pert PRO diffractor with monochromatized Kα radiation (λ = 0.154 nm). Fourier transform infrared spectra (FT-IR) were collected on a Thermo Fisher Scientific Nicolet 6700 instrument. X-ray photoelectron spectroscopy (XPS) was measured on a PHI 5000C ESCA spectrometer with a monochromatized Al Kα source. The peaks were fitted using XPSPEAK41 software and the binding energy was corrected by a reference (284.5 eV of C 1 s). Nitrogen adsorption–desorption isotherm was performed on a Quantachrome Autosorb-1 chemisorption apparatus at 77 K.

Adsorption experiments

Adsorption experiments were conducted in batch mode. The adsorption kinetics were determined by adding 5 mg adsorbent into 25 mL solution with initial cesium concentration of 4.91 mg/L. At predetermined time intervals, aliquots were withdrawn and filtered. Similarly, the adsorption isotherm experiments were performed within the concentration range of 1.7–21.3 mg/L. Adsorption experiments were allowed for proceeding in a shaker (25 °C) for 12 h to assure adsorption equilibrium.

The effect of solution pH was examined within pH of 3.0–1.0. The interfering ion tests were investigated by separately adding ions (K+, Na+, Mg2+ and Ca2+) into cesium solution (4.9 mg/L). The mass concentration ratio of interfering ions and cesium ions was approximately 10:1. After adsorption, a small portion of sample solutions were withdrawn and filtered through 0.22 µm syringe filters and analyzed on a polarized Zeeman Atomic Absorption Spectrophotometer (ZA3000 Series AAS, Hitachi, Japan). In this study, adsorption experiments were performed in duplicate and reported by average values. The removal efficiency (R, %), adsorption amount (qt, mg/g) and distribution coefficient (kd, L/g) were calculated from Eqs. (1), (2) and (3):

$$R = \frac{{\left( {C_{0} - C_{{\text{e}}} } \right)}}{{C_{0} }} \times 100\%$$
(1)
$$q_{{\text{t}}} = \frac{{\left( {C_{0} - C_{{\text{t}}} } \right) \times V}}{m}$$
(2)
$$k_{{\text{d}}} = \frac{{\left( {C_{0} - C_{{\text{e}}} } \right) \times V}}{{C_{{\text{e}}} \times m}}$$
(3)

where C0, Ct and Ce (mg/L) are the initial concentration, the instant concentration at time t and the equilibrium concentration of cesium ions in the solution, respectively. V (mL) represents the volume of the cesium solution, and m (g) the mass of the adsorbent.

Equations of adsorption kinetics:

Pseudo-first-order model

$$\ln \left( {q_{e} - q_{t} } \right) = lnq_{e} - k_{1} t.$$
(4)

Pseudo-second-order model

$$\frac{t}{{q_{{\text{t}}} }} = \frac{1}{{k_{2} q_{{\text{e}}}^{2} }} + \frac{1}{{q_{{\text{e}}} }}t.$$
(5)

Intraparticle diffusion model

$$q_{{\text{t}}} = k_{i} t^{0.5} + C_{i}$$
(6)

where qt and qe (mg/g) in Eqs. (1) and (2) represent the instant adsorption amount of cesium ions (t) and the equilibrium adsorption amount of cesium ions (equilibrium time). Ci (mg/g) in Eq. (3) depicts the effect of boundary layer. k1 (min−1), k2 [g (mg min)−1] and ki [mg (g min−0.5)−1] are the rate constants of pseudo-first-order, pseudo-second-order and intraparticle diffusion in Eqs. (4), (5) and (6), respectively.

Equations of adsorption isotherm.

Langmuir model

$$\frac{{C_{{\text{e}}} }}{{q_{{\text{e}}} }} = \frac{{C_{{\text{e}}} }}{{q_{{\text{m}}} }} + \frac{1}{{K_{{\text{L}}} q_{{\text{m}}} }}.$$
(7)

Freundlich model

$${\text{ln}}q_{{\text{e}}} = {\text{ln}}K_{{\text{F}}} + \frac{1}{n}{\text{ln}}C_{{\text{e}}}$$
(8)

where Ce (mg L−1) and qe (mg g−1) are the equilibrium concentration of Cs+ and its corresponding equilibrium adsorption capacity, respectively. KL (L mg−1) and KF (mg g−1)(L mg−1)1/n represent the constant of Langmuir model (Eq. 7) and Freundlich model (Eq. 8), respectively.

Dubinin–Radushkevich (D–R) isotherm (Eq. 9)

$$q_{e} = q_{{{\text{maxD}}}} {\text{exp}}\left[ { - \beta \left( {{\text{RTln}}\left( {1 + \frac{1}{{C_{{\text{e}}} }}} \right)} \right)^{2} } \right]$$
(9)
$$E = \frac{1}{{\left( {2\beta } \right)^{1/2} }}$$
(10)

where β (mol2/J2) is a constant related to the mean free energy of adsorption per mole of the adsorbate, qmaxD (mmol/g) is the theoretical saturation capacity and ε (J/mol) is the Polanyi potential (= RTln(1 + 1/Ce)), where R (J/mol/K) is the gas constant and T (K) is the absolute temperature. The constant gives an idea about the mean free energy E(J/mol) of adsorption per molecule of the adsorbate when it is transferred to the surface of the solid from infinity in the solution and can be calculated using the relationship in Eq. (10).

Results and discussion

Structure characterization

Figure 1 shows SEM images of PVA/GO and PVA/GO/KNi[Fe(CN)6]. It can be easily observed that an interwoven structure consisted of GO layers and these layers were full of wrinkles as marked in images (Fig. 1a, b). This unique structure may be generated from the strong hydrogen bonds between GO and PVA. After reacting with [Fe(CN)6]3−, the interwoven morphology retained and the skeleton seemed to be strengthened. It is interesting that many nanoparticles can be even identified in the wrinkle space (Fig. 1c, d). In TEM images (Fig. 2), a striking structure of layers decorated with particles can be clearly observed. These particles were dispersedly embedded on a GO layer plane (yellow marks in image). By comparison, we can speculate that these particles were probably KNi[Fe(CN)6] species and the resultant aerogel displayed a hierarchical structure.

Figure 1
figure 1

SEM images of PVA/GO (a, b) and PVA/GO/KNi[Fe(CN)6] (c, d)

Full size image
Figure 2
figure 2

TEM images of GO (a, b) and PVA/GO/KNi[Fe(CN)6] (c, d)

Full size image

As depicted in Fig. 3a, there were several diffraction peaks in the PXRD pattern of PVA/GO/KNi[Fe(CN)6]. The broad peak located around 2θ = 8° can be ascribed to the characteristic peak of GO and the shift (comparing with pristine GO) was probably due to an enlarged d-spacing when the layer PVA molecules were intercalated into GO layers. Other diffraction peaks located at 2θ = 17.3°, 24.6°, 35.0° and 39.5° can be well indexed to (200), (220), (400) and (422) spaces of cubic KNi[Fe(CN)6] phase (JCPDS No. 00-051-1897), respectively [37]. This agreed well with the observation of SEM and TEM.

Figure 3
figure 3

PXRD patterns (a) and FT-IR spectra (b) of PVA/GO/KNi[Fe(CN)6]

Full size image

In comparison to PVA/GO, two typical peaks can be found in FT-IR spectrum of PVA/GO/KNi[Fe(CN)6] (Fig. 3b). One weak absorption peak around 592 cm−1 could be ascribed to the stretching vibration of Fe–N and the other strong absorption peak in the vicinity of 2116 cm−1 presented the characteristic vibration of C≡N [18]. These information can confirm the successful formation of KNi[Fe(CN)6] in the substrate of PVA/GO/Ni2+.

To further obtain insight into the aerogel structure (PVA/GO/KNi[Fe(CN)6]), XPS was used to analyze its surface element composition. As depicted in Fig. 4a, the characteristic peaks of C 1 s, N 1 s, O 1 s, K 2p, Fe 2p and Ni 2p can be identified in XPS survey. Figure 4b–d shows the high-resolution XPS spectra of N 1 s, Fe 2p and Ni 2p. The peak located in 398.1 eV represented N1s spin orbit. The three peaks in Fe 2p spectra corresponded to Fe 2p3/2 (709.8 eV), Fe 2p1/2 (723.0 eV) and satellite peak (719.0 eV), respectively [38]. The three peaks in Ni 2p spectra corresponded to Ni 2p3/2 (856.5 eV), Ni 2p1/2 (874.0 eV) and satellite peak (7863.3 eV), respectively [39]. The KNi[Fe(CN)6 phase was formed on the PVA/GO surface.

Figure 4
figure 4

XPS survey of PVA/GO/KNi[Fe(CN)6] (a); high-resolution XPS spectra of N 1 s (b), Fe 2p (c) and Ni 2p (d)

Full size image

Nitrogen adsorption–desorption isotherm and pore size distribution of PVA/GO and PVA/GO/KNi[Fe(CN)6] are shown in Fig. 5. The two isotherms both displayed a notable hysteresis loop in the relative pressure of 0.4 to 0.9 and belonged to type IV isotherm, which suggested their mesoporous property (Fig. 5a and c) [40]. The BET specific surface areas (SAA) for PVA/GO and PVA/GO/KNi[Fe(CN)6] were 4.1 and 26.9 m2/g, while the average pore sizes were 15.9 and 24.8 nm (Fig. 5b and d), respectively. Notably, the phenomenon of increased SAA and pore size was likely associated with the generation of KNi[Fe(CN)6. Large SAA and broad pore size would be conducive to exposure of active sites and mass transfer of adsorbates [41].

Figure 5
figure 5

N2 adsorption–desorption isotherm and pore size distribution of PVA/GO (a, b) and PVA/GO/KNi[Fe(CN)6] (c, d)

Full size image

Adsorption performance of PVA/GO/KNi[Fe(CN)6]

The adsorption amount of cesium by PVA/GO and PVA/GO/KNi[Fe(CN)6] increased as the contact time (t) increased (Fig. 6a). For both systems, a fast adsorption stage occurred in the first 60 min, and a slow stage followed between 60 and 300 min. Till 720 min, an adsorption equilibrium would occur. While a similar adsorption trend could be shared, the adsorption amount of PVA/GO/KNi[Fe(CN)6] (23.65 mg/g) was as 3.5-fold as that of PVA/GO (6.70 mg/g). The increased adsorption capacity was strongly related to the introduction of KNi[Fe(CN)6] into PVA/GO. For exploring insightful adsorption behavior, three models were used to fit the adsorption data.

Figure 6
figure 6

Effect of contact time on adsorption amount of Cs(I) by PVA/GO and PVA/GO/KNi[Fe(CN)6] (a); Pseudo-first-order model (b); Pseudo-second-order model (c) and Intraparticle diffusion model (d). Experimental conditions: [Cs(I)]0 = 4.9 mg/L, [adsorbent] = 0.2 g/L, pH = 7.0, T = 298 K

Full size image

Adsorption kinetics

The pseudo-first-order model (PFO), the pseudo-second-order model (PSO) and intraparticle diffusion model (IPD) are shown in Fig. 6b–6d, respectively. The fitting parameters are listed in Table 1. The correlation coefficients (R2) of PSO (0.9971) and IPD (1.000 and 0.9959 of the first two stages) were higher than that of PFO (0.9744). Moreover, the qe of PSO (24.41 mg/g) and IPD (20.15 mg/g) was much closer to the experimental value (23.6 mg/g) than that of PFO (13.84 mg/g). It can easily be found that PSO and IPD fitted better for kinetic data than PFO. This demonstrated that the adsorption process was jointly controlled by chemisorption and diffusion mechanism, which may be determined by the unique structure of PVA/GO/KNi[Fe(CN)6] [42].

Table 1 Kinetics parameters of three models (Fig. 6)
Full size table

Adsorption isotherm

The adsorption isotherm was analyzed by the Langmuir model, Freundlich model and Dubinin–Radushkevich model (Fig. 7). The fitting parameters are shown in Table 2. As for isotherm, the adsorption amount gradually increased when the initial concentration increased. When the active sites were completely occupied by cesium ions, a saturation adsorption amount can be achieved (Fig. 7a). According to the R2 (0.9970), it can be understood that the Langmuir model can fit best for adsorption data, indicating this adsorption was mostly a monolayer homogeneous adsorption [43]. The unsatisfactory R2 suggested the adsorption did not conform to the Freundlich model (R2 = 0.7554). Calculated from the Langmuir model, the theoretically maximum adsorption capacity (qm) was 47.66 mg/g. A comparison was made among the cesium adsorbents, listed in Table 3. There are three categories of adsorbents, i.e., chitosan-based hybrids, magnetic composites and Prussian blue-based adsorbents [5, 18, 44,45,46,47]. Their adsorption capacity toward cesium ions was among 4.08–58.73 mg/g. The as-prepared aerogel (PVA/GO/KNi[Fe(CN)6]) displayed a comparable adsorption capacity (47.67 mg/g) to most of the adsorbents listed. Considering simple preparation and convenient separation, PVA/GO/KNi[Fe(CN)6 may be a potential adsorbent for cesium.

Figure 7
figure 7

Adsorption isotherm of Cs(I) by PVA/GO/KNi[Fe(CN)6] (a), Langmuir model (b), Freundlich model (c) and Dubinin–Radushkevich model (d). Experimental conditions: [Cs(I)]0 = 1.7–21.3 mg/L, [adsorbent] = 0.2 g/L, pH = 7.0, t = 12 h, T = 298 K

Full size image
Table 2 Isotherm parameters of three models (Fig. 7)
Full size table
Table 3 Comparison of Cs(I) adsorption capacity for newly reported adsorbents
Full size table

The Dubinin–Radushkevich model can be used to differentiate the physical adsorption and chemisorption. According to E = (2β)−1/2, the value of E is 18.5 kJ/mol. Generally, when E < 8 kJ/mol, the adsorption behavior belongs to physical sorption; when E > 8 kJ/mol, the adsorption behavior belongs to chemisorption [48]. Obviously, the adsorption behavior of cesium by PVA/GO/KNi[Fe(CN)6 was a chemisorption process, and this agreed well with the result of adsorption kinetics.

Effect of solution pH and coexisting ions

The effect of solution initial pH is shown in Fig. 8a. When pH ranged from 5.0 to 9.0, R (removal efficiency, %) kept 95.5%. When pH dropped to 3.0 or increased to 11.0, R underwent a moderate decline to 81.2% or 74.5%, respectively. In a weak acid system, the decline of R may be resulted from protons’ competitive influence. However, in an alkaline system, the reason was probably due to the reduction of active sites. As previously reported, a plenty of OH− may induce a fraction of degradation of KNi[Fe(CN)6. In a word, PVA/GO/KNi[Fe(CN)6 exhibited excellent removal performance toward cesium ions in neutral solution and was not favored in an acidic or alkaline environment. Evidence can also be found in Fig. 8b. The distribution coefficient (kd) suffered from a dramatic drop when the solution pH was 3.0 (kd = 21.5 L/g) or 11.0 (kd = 15.0 L/g), whereas kd kept above 107.6 L/g when the solution pH was 5.0 to 9.0. These results suggested an acidic or alkaline condition was not conducive to the adsorption [49].

Figure 8
figure 8

Effect of solution pH (a) and the corresponding distribution coefficient (kd) (b); effect of coexisting ions (c) and the corresponding distribution coefficient (kd) (d). Experimental conditions: [Cs(I)]0 = 4.9 mg/L, [adsorbent] = 0.2 g/L, pH = 7.0, t = 12 h, T = 298 K

Full size image

Likewise, the effect of coexisting ions was also investigated (Fig. 8c). In a mixed system, K+ and Na+ imposed a severe impact on the R of cesium ions, leading to R down to 71.1% and 55.6% from 98.3% (control test), respectively. In contrast, Mg2+ and Ca2+ showed a very slight effect, and R still retained 97.4% and 97.5%, respectively. A similar phenomenon can be found in Fig. 8d. K+ and Na+ interfered strongly with the adsorption of cesium ions, making kd fall down to 12.3 and 6.3 L/g from 287.4 L/g. Mg2+ and Ca2+ competed slightly with cesium ions, and kd was 193.7 and 199.3 L/g. As known, K+ and Na+ are located in the same group with the cesium ion, and they shared many similar physicochemical properties. In particular, the hydrated radius of K+ (3.3 Å) and Na+ (3.6 Å) is much closer to Cs+ (3.29 Å) [50]. In an ion-exchange process, K+ and Na+ would tend to consume many active sites, thus decreasing the adsorption of cesium ions [5].

Adsorption mechanism

To explore the adsorption mechanism, the PXRD patterns of PVA/GO/KNi[Fe(CN)6 before and after adsorption were performed (Fig. 9). After adsorption of cesium, the intensity of characteristic peaks became much weaker, and the positions of peaks displayed a slight shift to the positive direction. This suggested that a replacement of K+ by Cs+ may induce a shrinkage of KNi[Fe(CN)6 lattice. In other words, ion-exchange was one of the adsorption mechanisms. Moreover, the SEM–EDX spectra can provide another evidence (Fig. 10). Before adsorption, there was approximately 1.4wt% potassium on the surface of the aerogel. After adsorption, there was no signal of potassium on the surface, whereas 3.8wt% cesium was obviously detected. This further confirmed the ion-exchange mechanism, as previously reported [47]. In addition, the unique cubic lattice of PB or PBAs may trap a fraction of Cs+ ions to some extent.

Figure 9
figure 9

PXRD patterns of PVA/GO/KNi[Fe(CN)6] before and after adsorption of Cs(I)

Full size image
Figure 10
figure 10

SEM–EDX of PVA/GO/KNi[Fe(CN)6] before (a) and after adsorption of Cs(I) (b)

Full size image

XPS spectra were used to identify adsorption sites and interfacial interaction. As shown in Fig. 11a and b, the peak of Cs 3d can be found in XPS survey spectra and high-resolution spectra. This observation agreed with the results of EDX spectra, strongly suggesting the adsorption of cesium on the surface of the as-prepared aerogel. The peak of K 2p was also analyzed before (Fig. 11c) and after adsorption (Fig. 11d). The characteristic peak of K 2p disappeared after adsorption of cesium, and this indicated the involvement of K+ in the adsorption process. Moreover, the peak of O 1 s was also measured to investigate the contribution of oxygen-containing functional groups to the adsorption. O 1 s can be deconvoluted into two types of oxygen species (C–O–C/C–O=C and C–OH). The relative content of C–O–C/C–O=C species was 19.83% and 16.22%, while that of C–OH species was 80.47% and 83.78% before and after adsorption. There was a slight shift to high binding energy for C–O–C/C–O=C from 533.48 eV to 534.20 eV, and for C–OH from 532.30 eV to 532.66 eV, respectively. The variation of relative content and binding energy of oxygen species suggested oxygen-containing functional groups may participate in the adsorption process by the way of replacement of protons [46, 51].

Figure 11
figure 11

XPS survey of PVA/GO/KNi[Fe(CN)6] after adsorption of Cs(I) (a); high-resolution XPS spectra of Cs 3d (b); K 2p before adsorption (c) and K 2p after adsorption (d); O 1 s before adsorption (e) and O 1 s after adsorption (f)

Full size image

Based on the above analysis, the adsorption mechanism of cesium ions by PVA/GO/KNi[Fe(CN)6 was tentatively proposed as shown in Fig. 12. There were three immobilization mechanisms involved in the cesium adsorption: (i) the capture of Cs+ by cubic lattice, (ii) the ion-exchange of K+ by Cs+ and (iii) the interaction between electrons of oxygen and cesium orbits (proton substitution) [52, 53].

Figure 12
figure 12

Schematic illustration of adsorption mechanism of Cs(I) by PVA/GO/KNi[Fe(CN)6]

Full size image

Conclusions

In summary, a novel aerogel (PVA/GO/KNi[Fe(CN)6]) was successfully fabricated by combining a synergistic cross-linking reaction with in situ precipitation strategy. Ni2+ and PVA synergistically played roles in the assembly of GO units, and then Ni2+ in situ coordinated with [Fe(CN)6]3− to produce KNi[Fe(CN)6] in the following stage. The resultant aerogel exhibited a hierarchical structure composed of GO sheets decorated with nanoparticles. Evaluation of adsorption performance indicated pseudo-second-order model and intraparticle diffusion model can better fit the kinetic data, and the Langmuir model can depict the isotherm data appropriately. This indicated the adsorption of Cs+ by the as-prepared aerogel was probably a chemisorption process. Characterization analysis further confirmed that the substitution of Cs+ for K+ (KNi[Fe(CN)6]), the physical capture of Cs+ by the cubic lattice (KNi[Fe(CN)6]) and the proton exchange (carboxyl/hydroxyl) were the possible adsorption mechanism. This aerogel adsorbent showed good adsorption performance (with a capacity of 47.66 mg/g) as well as a merit of easy separation from solution and may be a promising alternative for cesium cleanup.