As illustrated in Fig. 1(a), GO membranes were prepared from 40 mL, 7.5 mg/L GO suspensions on mixed cellulose ester (MCE) substrates using vacuum filtration. Then, 5~500 mg/L different salts (FeCl3, Pb(NO3)2, ZnSO4 and CuSO4) solutions were added to the feed side, respectively. Under a pressure of 1 bar, the salt solutions were filtered through GO membranes. Filtrates were collected after 20 min when the filtration process became steady. Water permeances were calculated by measuring the time required to collect 30 mL filtrate, and the corresponding concentrations of FeCl3, Pb(NO3)2, ZnSO4 and CuSO4 in the filtrate were measured using inductive coupled plasma-optical emission spectrometry (ICP-OES). Rejection rates were calculated from the concentration of feed and permeate solution (see details in Methods). Three parallel permeation tests were performed to obtain averaged values of water permeance and rejection rate.
As shown in Fig. 1(b), the water permeance was 75.2 L m−2 h−1 bar−1 with a high rejection rate of 99.9% for a 5 mg/L FeCl3 1 bar. Increasing the concentration of the FeCl3 solution to 50 mg/L, 250 mg/L and 500 mg/L, the permeances dropped to 63.5 L m−2 h−1 bar−1, 54.9 L m−2 h−1 bar−1 and 48.2 L m−2 h−1 bar−1, with corresponding rejection rates of 98.9%, 90.1% and 89.8%, respectively (see Supplementary Fig. S2 in Supporting Information). These permeances are one to two orders of magnitude higher than those of most GO membranes reported previously18,37,38,41 (see details in Supporting Information section PS 3) . We note that the highest water permeance previously reported very recently for state-of-the-art nanofiltration membranes is about 62 L m−2 h−1 bar−1 with only a 98% rejection rate42, as shown in Fig. 1(c) and Table S1 in Supporting Information section PS 3. Thus, the water permeances we obtained are much superior to the permeances of all other nanofiltration membranes with reasonable rejection rates for 50~1000 mg/L multivalent metal ion solutions. In addition, for other multivalent ions of CuSO4, Pb(NO3)2 and ZnSO4, the water permeances were 56.6 L m−2 h−1 bar−1, 46.6 L m−2 h−1 bar−1, and 48.7 L m−2 h−1 bar−1 with corresponding rejection rates of 97.8%, 86.9%, and 83.0%, respectively. Although these permeances and rejection rates are lower than the permeance for FeCl3 at the same concentration, they are still superior to most other membranes, as shown in Fig. 1(b), (c), and the Table S1 in Supporting Information section PS 3. Thus, the GO membranes in the present work exhibit ultrahigh water permeances with high rejection rates for multivalent ions.
The GO membranes also showed outstanding stability with superior performance. Fig. 1(e) shows that the water permeance decreases from 95.5 L m−2 h−1 bar−1 and reaches a stable value of 75.2 L m−2 h−1 bar−1 within 100 min, while the corresponding rejection rate is maintained at more than 97% for 5mg/L FeCl3. We noted that the feed salt concentration continuously increased in the dead-end filtration set-up. In addition, the increased salt concentration (concentration polarization) could affect the membrane performance. Therefore, a semi-continuous process was employed for the long-term operation experiment of the high-concentration solution filtered by the GO membranes43: the surface and filtration cleaning of the GO membranes by DI water were performed before the next filtration (see details in Supporting Information section PS 7). The water permeances slightly decreased slightly from 49.0 L m−2 h−1 bar−1 to 46.5 L m−2 h−1 bar−1, while the corresponding rejection rates increased from 85.0% to 96.4%, showing an outstanding stability with superior filtration performance of the GO membranes (see Fig. S5 in Supporting Information section PS7). We believe that the increase in the rejection rate and the slight decrease in the water permeance were mainly due to the improved self-assembly and more compacted structures between the GO sheets under the long-term vacuum filtration conditions.
In addition, we analysed the the ions adsorption by the GO membranes in our filtration experiments44. As shown in Fig. 2, the adsorption efficiency of the GO membranes for ZnSO4, CuSO4, FeCl3, and Pb(NO3)2 were 13%, 19%, 26%, and 42%, respectively. As mentioned above, the rejection rates of the GO membranes were 83% ~ 99.9% (see Fig. 1(b)), which are much higher than 13% ~ 42% removed by adsorption. These results indicate that the significant effect on the removal multivalent ions at a low concentration of 5 mg/L is mainly due to rejection by the GO membranes, even though the adsorption effect cannot be negligible. Further, we performed adsorption experiments using FeCl3 solution with a high concentration of 500 mg/L. As shown in Fig. S6 in Supporting Information section PS8, the FeCl3 solution with 500 mg/L showed stable concentration for over 120 min. The removal rate by adsorption was < 4%. Considering the rejection rate for 500 mg/L FeCl3 solution was 89.8 %. Thus, for the high salts concentrations, the salts removal are almost entirely attributed to the rejection rather than adsorption by the GO membranes, which was further confirmed by the mass balance experiment (see details in Supporting Information section PS 9).
Our previous experimental studies showed that the order in which the GO membrane is exposed to ions is important for efficient ion sieving41. One type of ions enter the membrane first, controlling the interlayer spacing and potentially excluding other cations that require a larger interlayer spacing, or allowing other hydrated cations with smaller sizes to pass through. Thus, we performed additional permeation experiments with GO membranes using ZnSO4 or FeCl3 as controlled ions, which are denoted as ZnSO4-controlled GO membranes or FeCl3-controlled GO membranes. Explicitly, 20 mL, 5 mg/L ZnSO4 or FeCl3 solutions were first added to the feed side and then filtered at a pressure of 1 bar. Subsequently, salt solutions including 5 mg/L ZnSO4 or FeCl3 with 5 mg/L target salts (Pb(NO3)2, CuSO4, FeCl3, or ZnSO4) were added to the feed side, and the mixed salt solutions were filtered through GO membranes at the same pressure. The filtrates of the mixed solutions were collected after 20 min when the filtration process became steady, and the concentrations of target testing ions in the filtrates were measured. Rejection rates were calculated from the concentrations of feed and permeate solutions (see details in Methods).
We found that ZnSO4-controlled GO membranes and FeCl3-controlled GO membranes had different rejection performance. The ZnSO4-controlled GO membranes showed high rejection rates of 99.6%, 95.1%, and 92.6% for FeCl3, Pb(NO3)2, and CuSO4, with corresponding water permeances of 44.2 L m−2 h−1 bar−1, 42.9 L m−2 h−1 bar−1, and 24.2 L m−2 h−1 bar−1, respectively (see Fig. 3(a)). In contrast, FeCl3-controlled GO membranes have rejection rates for Pb(NO3)2, CuSO4, and ZnSO4 of only 47.6%, 39.8%, and 24.8% with corresponding water permeances of 61.9 L m−2 h−1 bar−1, 42.9 L m−2 h−1 bar−1, and 44.7 L m−2 h−1 bar−1, respectively (see Fig. 3(c)). The rejection rates for the FeCl3-controlled GO membranes are much lower than the rejection rates for the ZnSO4-controlled GO membranes.
To illustrate the underlying physical mechanism, the interlayer spacings of the GO membranes fabricated in the present work were analyzed by X-ray diffraction. As shown in Fig. 1(d), the interlayer spacings indicated by the Bragg peaks of X-ray diffraction) were 15.8 ± 0.1 Å, 15.7 ± 0.1 Å, 15.3 ± 0.1 Å, and 15.2 ± 0.1 Å for GO membranes immersed in only FeCl3, Pb(NO3)2, CuSO4, or ZnSO4 solutions, respectively. These values are all about 3 Å larger than the interlayer spacings of the GO membranes prepared by the conventional drop-casting method41,45,46 (see details in Methods), which were 13.1 Å, 13.0 Å, 13.0 Å, and 12.8 Å for the drop-casting GO membranes immersed in FeCl3, Pb(NO3)2, CuSO4, and ZnSO4 solutions. The large interlayer spacings of the GO membranes prepared by the present method can be further demonstrated by the interlayer spacing of the GO membrane immersed in pure water, which was 16.2 Å (Fig. 1(d)), this is also ~3 Å larger than the interlayer spacing of 13.0 Å for the GO membrane fabricated by the drop-casting method immersed in pure water, as reported earlier41