Synthesis and characterization of ultrafine ZnO/rGO nanocomposites. Figure 1a illustrates a general heterogeneous nucleation and diffusion-controlled growth process of ultrafine metal oxide/rGO nanocomposites using ultrafine ZnO/rGO nanocomposites as an example. The GO is rich in oxygen functional groups, and it is supposed that these oxygen functional groups are responsible for the heterogeneous nucleation. More typically, in a mixed ethylene glycol (EG) solution of GO and zinc acetate dihydrates (Zn(Ac)2·2H2O), the presence of oxygen groups which contributes to an overall negatively charged surface of -40 mV will capture the positively charged [Zn(EG)2]2+ complexes through electrostatics interactions (Fig. S1). Under solvothermal conditions, the adsorbed [Zn(EG)2]2+ complexes could be hydrolyzed to adsorbed Zn(OH)2, which is equivalent to the formation of preferential sites. At such sites, the effective surface energy is lower, thus diminishes the free energy barrier and facilitating heterogeneous nucleation. The as-formed nucleus will grow slowly via a diffusion-controlled growth process, leading to the formation of ultrafine and isolated island nanoparticles on the surface of rGO.
The TEM images reveal that the as-prepared ZnO/rGO nanocomposites has a very transparent laminated structure with less wrinkles and folding, which is similar to the pristine GO nanosheets (Fig. 1b). The high-resolution TEM images (Fig. 1c-d) clearly demonstrate that the rGO surface is consisted of densely distributed ZnO nanoparticles with the size of ca. 2.5 nm. Besides, energy dispersive X-ray spectroscopy (EDS) elemental mappings (Fig. 1e) also confirm that ZnO nanoparticles are distributed homogeneously within the rGO nanosheets. The weight ratio of ZnO in the composite is determined to be close to the designed value 50 wt% (Fig. 2f), suggesting that all Zn precursors participate fully the hydrolysis reaction to ZnO.
Mechanism study of the formation of ultrafine ZnO nanoparticles on GO surface. To understand the nucleation and growth of ZnO nanoparticles on rGO surface, the time-dependent experiments were conducted. At the initial solvothermal time of 0.5 h, the ZnO nanoparticles are already formed, but their size is only 1 nm corresponding to ~ 20 ZnO molecular clusters (Fig. 2a and Fig. S2). With the time prolonging, the size gradually increases to ca. 2.5 nm at 3 h while remains unchanged until reaching to their final stage (Fig. 2b-e). It is interesting to note that the particle number density is almost the same of 9 per 10 nm⋅10 nm for all samples (Fig. 2e and Fig. S3), which clearly indicates that the nucleation occurs simultaneously at GO surface and no new nuclei form during growth process. This result implies that the particle number density should be depended on the surface characteristics of GO, more especially the number of oxygen functional groups.
To verify the structure and chemical state of the nanocomposites produced at the first few hours of solvothermal synthesis, the thermogravimetric (TG), X-ray photoelectron (XPS), and X-ray diffraction (XRD) characterizations were conducted. The weight ratio of ZnO in composites from TG results (Fig. 2f) has similar increasing tendency with the size of ZnO nanoparticles. At the initial 0.5 h, the weight ratio of ZnO nanoclusters is only 10 wt%, and then gradually increases to the designed value of 50 wt% at 3 h. It is noted that the ZnO/rGO-0.5 h has a significant weight loss at temperature of 160–220 ℃ in TG curve, which should be due to deoxygenation of incompletely reduced GO. This could be well confirmed by XRD pattern (Fig. 2g), in which the ZnO/rGO-0.5 h exhibit two shoulder peaks centered at 11o and 23o, responding to the characteristic of GO and rGO[33, 34], respectively. While the time increase to 2 h, the characteristic peak at 10o disappears, suggesting the successful reduction of GO. And no obvious characteristic peaks could be discerned for the ZnO crystals, further confirming their ultrafine size. In addition, XPS patterns (Fig. 2h-i and Fig. S4) suggest that the GO is incompletely reduced at the initial 0.5 h, as indicated by a high ratio of C = O. It should be noted that the Zn 2p spectra (Fig. 2j) for all samples are very similar with two obvious bands located at 1045.3 eV and 1022.3 eV, corresponding to the Zn2+ 2p1/2 and Zn2+ 2p3/2, respectively[35]. All these results indicates that the reduction of GO and the growth of ZnO on rGO nanosheets could be accomplished at the initial 3 h under solvothermal conditions.
The key to the synthesis of ultrafine ZnO nanoparticles should be the oxygen functional groups on GO surface, which have binding ability toward Zn2+ through electrostatic attractions, and consequently trigger the heterogenous nucleation. According to the elemental analysis, GO used in this work has a C/O ratio of ca.1.8. And XPS spectrum in Fig. 2h reveals that it is consisted of majorly epoxide and carbonyl functional groups on the basal plane, and hydroxyl and carboxyl groups on its edges[36]. To get insights into the binding abilities of these oxygen functional groups with Zn 2+ in EG, we perform density-functional theory (DFT) calculation and disclose that the order of adsorption stability towards Zn2+ is as follows: epoxy(-COC) > carbonyl (-CO) > carboxyl (-COOH) > > hydroxyl (-OH) (Fig. 3a). Furthermore, the hydrolysis of adsorbed Zn2+ and then its transformation to ZnO on GO surface and without GO are analyzed. The optimized structures of intermediates and their Gibbs free profiles are displayed in Fig. 3b-c and Table S1. The overall Gibbs free energy changes on oxygen functional groups (-OH, -COC, -COOH and -CO) in EG reaction system are − 6.433 eV, -6.999 eV, -6.608 eV, and − 7.759 eV, respectively, which are all more negative than that of without GO (-4.783 eV). The last step of Zn(OH)2/rGO*→ZnO/rGO* exhibits positive Gibbs energy change (ΔG), which infers this step is the thermodynamic-limiting step in the whole process. From the results of Gibbs energy changes, it can be concluded that the total transformation process of Zn2+ to ZnO is a spontaneous process and thermodynamically more favorable on GO surface. The carbonyl is suggested as the most active site for nucleation and followed by epoxy and carboxyl.
The controlled nucleation is triggered by the oxygen functional groups on GO surface due to the lower free energy barrier, while the consequent growth of ZnO nanoparticles should be a diffusion-controlled process that is limited by the rate of incorporation of adatoms into ZnO growth centers. Herein, the growth of ZnO nanoparticles on the rGO surface was studied through a series of control experiment by changing the reaction temperature. As shown in Fig. S5a, the ZnO/rGO nanocomposites prepared at temperature of 90℃ show ultrasmall ZnO nanoparticles of ca. 1.3 nm. Such low nucleation temperature should be due to the lowering heterogenous nucleation barrier, while the small size should be caused by the limited diffusion at low temperature. With increasing the reaction temperature to 180℃, the average size of ZnO nanoparticle increases gradually, but the particle number density keeps almost unchanged on the rGO surface (Fig. S5-7). In addition, the concentration of precursor Zn (Ac)2·2H2O was also changed to adjust the loading contents of ZnO in ZnO/rGO nanocomposites. At a low precursor concentration, the TG curve reveals that the content of ZnO is 30 wt%, being the same with the designed value (Fig. S7b). However, the resulting nanocomposites (Fig. S5d) show an almost transparent feature without the visible nanoparticles on the rGO surface, which should be due to the formation of ultrasmall ZnO clusters (< 1 nm) that are hardly displayed by the TEM images. While the ZnO loading is larger than 50%, the nanocomposites show almost similar morphologies of ZnO nanoparticles (Fig. S5e-f and Fig. S7b).
Furthermore, the synthesis of pure ZnO particles without GO was conducted under the same solvothermal conditions. The obtained ZnO product shows spherical aggregates with size of ca. 600 nm (Fig. S8), which should be due to the homogenous nucleation continued by the growth of primary nanoparticles and thereafter the spherical aggregation. In addition, the solvent is also critical for the formation uniform ZnO nanoparticles on the rGO surface. EG is selected as the solvent due to its strong chelating and reduction ability[37]. While changing solvent from EG to deionized water and ethanol (Fig. S9-10), large and aggregated particles are formed on the wrinkled rGO surface.
General synthesis of ultrafine metal oxide/rGO nanocomposites. Based on the above discussion, the oxygen functional groups-induced heterogenous nucleation should work as a general strategy for the synthesis of other metal oxide/rGO nanocomposites. Herein, a family of nanocomposites with ultrafine metal oxide nanoparticles grown on rGO surface are successfully prepared, including CdO/rGO, CoO/rGO, CuO/rGO, Fe2O3/rGO, MgO/rGO, La2O3/rGO, MoO3/rGO and Nb2O5/rGO. As shown in Fig.4a-h and Fig. S11-12, TEM images verify that all nanocomposites consist of monodispersed ultrafine metal oxide nanoparticles with the size less than 3 nm. Moreover, metal sulfide nanoparticles/rGO nanocomposites, such as ZnS/rGO and MoS2/rGO could also be prepared by the present strategy through adding the thiourea in the precursor solution, as shown in Fig. 4i-jand Fig. S13.
The synthetic procedures reported in this work is highly reproducible and readily applicable for the large-scale synthesis of ultrafine metal oxide/rGO nanocomposites. And more importantly, due to the effective inhibition of wrinkling of the rGO sheets, the as-prepared nanocomposites are highly dispersed into EG, forming a homogenous colloidal suspension of ca. 4 mg/ml without obvious sedimentation even after one month (Fig. 4k). Furthermore, the as-prepared nanocomposites could be also dispersed in the organic solvent such as N, N-dimethylformamide (DMF), isopropanol or N-methylpyrrolidone (NMP), which are desirable for wet processing to enable a homogenous arrangement of nanocomposites as filler, coatings, or thin films.
Preparation and characterization of ZnO/rGO membranes. The free-standing ZnO/rGO nanocomposite membranes were fabricated by a simple vacuum filtration of homogenous colloidal suspensions. The continuous suction force from vacuum pump can move the solvent rapidly, which could overcome the agglomeration of nanocomposites and form the laminated membranes. After being washed with ethanol and dried in 60℃, the ZnO/rGO membrane with 45 mm diameter is obtained (Fig. 5a). This membrane can randomly bend 360°, suggesting its excellent flexibility and durability. SEM images (Fig. 5b) indicates that the resulting membrane is continuous and free of macropores or surface defects. And the cross-section image (Fig. 5c) demonstrates a layered structure resembling that of nacre, therefore exhibiting good mechanical strength. The distance between ZnO/rGO bilayer is ~ 4.6 nm determined by AFM images (Fig. 5d-e). By subtracting the rGO monolayer (ca. 0.34–0.4 nm)[23], the 2D nanochannel spacing of ZnO/rGO membrane is estimated at ~ 4.2 nm. Furthermore, the interlayer spacings between nanosheets as pores are evaluated by nitrogen adsorption-desorption measurement. As shown in Fig. 5f, ZnO/rGO membrane displays a Type IV sorption isotherm with Type H3 hysteresis[38], responding to the slit-shaped pores between nanosheets. The resulting BJH pore size distribution reveals that the average pore size is at ~ 4.0 nm, which is consistent with AFM results.
Nanofiltration performance of ZnO/rGO membranes. GO-based membranes are great promising for advanced nanofiltration in water treatments, however their narrow interlayer channels generally limit the water flux and the separation of larger organic molecules[39]. Generally, the separation mechanism of GO-membranes is governed primarily by the interlayer spacing between the nanosheets and the length or tortuosity of the transport pathways[40]. By functioning as pillars, the metal oxide nanoparticles remarkably increase both vertical interlayer spacing and lateral tortuous paths of the rGO membranes, offering the great possibility for the membranes concurrently with high water permeance and high rejections (as illustrated in Fig. 6a).
To evaluate the separation performance, we fabricated the ZnO/rGO membranes supported by nylon substrates and used them for dye permeation tests. As shown in Fig. S16, ZnO/rGO membranes are uniformly coated on a porous nylon 66-0.22 µm substrate. The surface wettability of ZnO/rGO membranes is determined to be ca. 46o by the water contact angle (Fig. S17), indicating excellent hydrophilic properties. The loading mass of ZnO/rGO on nylon membranes can be changed from 0.14 to 0.58 mg cm− 2 by varying the volume of colloidal dispersion during the vacuum filtration. The average thicknesses of the ZnO/rGO membranes are in the range of nanometers (220 nm) to micrometers (1.1 µm) estimated by the SEM observation (Fig. S18).
Due to the variations in transport distance, the loading mass of nanosheets has a critical effect on the water flux and the rejection rate of methyl blue (MB), as shown in Fig. 6b. With the increase of the ZnO/rGO loading mass, the rejection rate increase from 69–98.1%, while water flux decreases from 390 to 190 L m− 2 h− 1 bar − 1. Therefore, we select the optimized loading mass of ca. 0.54 mg cm− 2 for further performance evaluation, which achieve simultaneously an extraordinary high-water flux of 225 L m− 2 h− 1 bar − 1 and MB rejection of 98.1 %. To highlight the advanced performance of ZnO/rGO membranes, the pure GO membranes at the same mass loadings are prepared, which exhibits only 20 L m− 2 h− 1 bar − 1 water permeance as given in Fig. 6c. Moreover, the MB rejections of other reported GO-based membranes are compared in Fig. 6d and Table S2. Except the exciting rejection rate, the ZnO/rGO membranes exhibit much higher water permeance than those of the GO-based membranes reported so far (generally less than 80 L m− 2 h− 1 bar − 1) [41–51].
Molecular selectivity is one of crucial parameters for the nanofiltration. Here we select a series of dye molecules with different molecular weights (361 ~ 960 Da, including evans blue (EB), congo red (CR) and chrome black T (CBT), methyl orange (MO)) for the comparison, and their chemical structures and molecular sizes are given in Table S3. Although the interspacing of ZnO/rGO layers (~ 4.2 nm) is much larger than the molecular sizes of these dyes, the ZnO/rGO membranes still exhibit remarkably high rejections of > 95% for EB, CR and CBT at high water permeances of ~ 220 L m− 2 h− 1 bar − 1, as shown in Fig. 6e. Such excellent rejections should be due to the additional blocking and sieving effects by these monodispersed ZnO nanoparticles, which create narrow and tortuous paths in the 2D interlayer spaces. However, in the case of MO with smallest molecular size of 1.13 nm × 0.42 nm, the ZnO/rGO membranes show only 27% rejection as a result of ineffective sieving effects for smaller molecules.
The long-term stability of ZnO/rGO and GO membranes were also conducted over a duration of 30 h. As shown in Fig. 6f, the GO membranes deliver an initial water flux of 20 L m− 2 h− 1 bar − 1 and then decline rapidly to 3 L m− 2 h− 1 bar − 1, due to the compression of loosely overlapped GO nanosheets under pressure. On the contrast, the rigid ZnO nanoparticles could serve as rigid "pillars” to create permanent high-speed waterways along the surface of rGO nanosheets, achieving a high steady-state flux value of 160 L m− 2 h− 1 bar − 1 after steadily 30 h working.