Bioinspired graphene membrane with temperature tunable channels for water gating and molecular separation

Smart regulation of substance permeability through porous membranes is highly desirable for membrane applications. Inspired by the stomatal closure feature of plant leaves at relatively high temperature, here we report a nano-gating membrane with a negative temperature-response coefficient that is capable of tunable water gating and precise small molecule separation. The membrane is composed of poly(N-isopropylacrylamide) covalently bound to graphene oxide via free-radical polymerization. By virtue of the temperature tunable lamellar spaces of the graphene oxide nanosheets, the water permeance of the membrane could be reversibly regulated with a high gating ratio. Moreover, the space tunability endows the membrane with the capability of gradually separating multiple molecules of different sizes. This nano-gating membrane expands the scope of temperature-responsive membranes and has great potential applications in smart gating systems and molecular separation.

This is an interesting paper, authors reports the modulation of water flux in GO based membrane by polymerization of a temperature-responsive polymer with GO sheets. I think it opens a new route on controlling the laminar structure and modulating the mass transport through GO membrane. The idea and experimental data are fairly solid. However, on the mechanism understanding, I have several comments that need to be addressed. 1. Authors explained the gating of water transport by shrinkage of channel inside the GO membrane, however, from the Fig. 4c, when the membrane was heated to 500C, one can only see <0.5 degree shifting in XRD spectrum, this means there is only ~0.2 nm shrinking on the channel, the little change should not be enough to explain either the 7 times decrease of water flux(if calculated by Hagen-Poiseuille equation, that is equation 1 in manuscript), or the sieving properties(RB show 60% rejection at 25 0C, which increase to 90% at 50 0C). I noticed the XRD was performed with relative humidity of 40%, I think, in fact, this could not be used as an effective evidence to show the channel width dependence on temperature. XRD measurements of samples immersed in liquid water, or at least 100% humidity ,which reflects the real filtration conditions, to see if there is a large enough shrinkage would be a appropriate evidence. If after the XRD measurement, the change of channels is still not enough to explain the experiment results, then a proper explanation should be made to gain the insights on the mechanism.
2. The thickness of individual P-GOM sheets is ~5 nm, the thickness of PNIPAM layer is then estimated to be 4 nm(page 7, line 122-125). This estimation is based on there is only one layer of GO encapsulated by PNIPAM, however, AFM could not prove there is only one layer GO sandwiched/wrapped by PNIPAM. In particular, GO is prone to agglomerate during the proposed process, as the precursor-NIPAM is positively charged, then, the GO sheets probably are multilayer stack. Authors need a justification on this statement. 3. One inconsistency data. the rejection of [Fe(CN)6] is written as 38.8% in Fig. 5e, however, from This manuscript reports the preparation of a PNIPAM-GO membrane where thermal properties of PNIPAM allow the spacing between the GO sheets to respond with respect to temperature. The work combines experimental and simulation evidence to support the proposed mechanism. When the temperature exceeds a certain value, the GO sheets collapse and shutdown transport of water. Intermediate temperatures can be used to do selective separations in the several nm to order 10 nm molecular size range. The results seem reasonable and are accompanied by a fair amount of hype. It would be helpful to back the hype up with comparisons to existing membranes that are used to separate the types of molecules that are considered in this work. Instead, the manuscript largely stays away from any quantitative comparison, preferring to rely on language to convince the reader that the materials are particularly special (from an application perspective) -no doubt, though, that the physical mechanism is interesting.
For example, it would be useful to report permeance and/or permeability data for transport rates as opposed to just flux. That would help the reader to compare these materials to other membranes where, particularly permeance units, are commonly reported.
Presumably the materials aren't useful for desalination, or else data with sodium would have been used. One is left to wonder what the actual separation resolution really is for these materials. The 'hype' would be more justified by doing separations between ions, for example, as opposed to the Cu2+ and RB cutoff that is reported (i.e., a hydrated 'big' ion versus an organic dye). They authors should at least report the resolution of separations that can reasonably be achieved for dye molecules versus ions using other membrane approaches.
Additionally, no rational is given for the low concentrations that were used in the experiments. One is left to wonder whether those conditions are really relevant / how the material would perform if challenged with a more concentrated feed solution.
As an aside, it seems unlikely that highly confined water molecules in the nano-channels would behave like a bulk viscous flow (such that the Hagen-Poiseuille equation would be valid). It seems as though a more molecular approach to analyzing transport would be appropriate compared to a continuum approach.
We are greatly encouraged that all three referees affirm our work "interesting". The comments are invaluable and very helpful for improving our paper, as well as the important guiding significance to our research. We have evaluated comments carefully and made revisions according to these suggestions. Revised portion are marked in yellow in the paper. The main revisions in the paper and the responses to the reviewers' comments are as following:

Responses to Reviewers' comments:
Reviewer #1: The authors reported a negative temperature-response gating membrane composed of GO and PNIPAM. They not only show the tunable water gating and precise small molecules separation of the membrane but also explains the mechanism of the temperature-responsive behaviour. The results are interesting.

Response:
We appreciate reviewer's very encouraging comment.

Response 2:
Thanks for referee's kind advice. We have added the detailed explanations about the Raman spectra

Comment 4:
In line 260, page 14, authors XRD peak positions of P-GOMs at 25°C and 50°C. By calculation, 2θ = 5.43° corresponds to a layer distance of 1.627 nm, 2θ = 5.59° corresponds to a layer distance of 1.581 nm, the difference between them is 0.046 nm. I think this layer distance is so small that cannot explain the large water flux difference. This may be not real layer distance because samples were dried for XRD. I would strongly recommend the authors do the XRD of the membranes in wet state.

Response 4:
We appreciate this very helpful suggestion. In the initial manuscript, we performed the XRD characterizations of P-GOMs in dry state, the diffraction peak of P-GO shifted from 2θ = 5.43° to 2θ = 5.59°. However, this could not fully reflect the real lamella distance change in aqueous environment as the referee indicated. Following the reviewer's good suggestion, we have performed the XRD characterization of the membranes in wet state. The diffraction peak shifts from 2θ = 4.71° to 2θ = 5.91° when the temperature increased from 25°C to 50°C, demonstrating the d-spacing of P-GOMs changes from 1.87 nm to 1.49 nm. The smaller d-spacing of P-GOMs at 50°C, undoubtedly, will decrease the water flux through the P-GOMs. Besides, the entanglement of PNIPAM chains between two adjacent GO sheets also leads to the decrease of water flux, because the entangled PNIPAM In the revised manuscript, we replaced the original dry state XRD spectra in Fig. 4b (page 16) with the wet state XRD spectra below (Fig. R2). And we added the explanation of the large water flux difference in the revised manuscript (page 15, line 262). Reviewer #2: This is an interesting paper, authors report the modulation of water flux in GO based membrane by polymerization of a temperature-responsive polymer with GO sheets. I think it opens a new route on controlling the laminar structure and modulating the mass transport through GO membrane. The idea and experimental data are fairly solid. However, on the mechanism understanding, I have several comments that need to be addressed.

Response:
We appreciate very much for the referee's highly affirmative comment.

Comment 1:
Authors explained the gating of water transport by shrinkage of channel inside the GO membrane, however, from the If after the XRD measurement, the change of channels is still not enough to explain the experiment results, then a proper explanation should be made to gain the insights on the mechanism.

Response 1:
Thanks for referee's helpful comment. In the initial manuscript, we performed the XRD characterizations of P-GOMs in dry state, the diffraction peak of P-GO shifted from 2θ = 5.43° to 2θ = 5.59°. As the referee indicated, this could not fully reflect the real lamella distance change in aqueous environment. Following the reviewer's suggestion, we have performed the XRD characterization of the membranes in wet state. The diffraction peak shifts from 2θ = 4.71° to 2θ = 5.91° when the temperature increased from 25°C to 50°C, demonstrating the d-spacing of P-GOMs  In the revised manuscript, we replaced the XRD spectra in Fig. 4b (page 16) with the XRD spectra below (Fig. R3). And we added the explanation of the large water flux difference in the revised manuscript (page 15, line 262).

Comment 2:
The thickness of individual P-GOM sheets is ~5 nm, the thickness of PNIPAM layer is then estimated to be 4 nm (page 7, line 122-125). This estimation is based on there is only one layer of GO encapsulated by PNIPAM, however, AFM could not prove there is only one layer GO sandwiched/wrapped by PNIPAM. In particular, GO is prone to agglomerate during the proposed process, as the precursor-NIPAM is positively charged, then, the GO sheets probably are multi-layer stack. Authors need a justification on this statement.   one Typo: Fig. 5a-d, Absortance.

Response 3:
Thanks for your helpful comment. We mistook the rejection rate at 35°C (38.8% in Fig. 5i) as 25°C.
We have changed "38.8%" to "12.5%" in the revised manuscript (Fig. 5e, page 21) and Supplementary Information (Supplementary Table S2, page 21). Accordingly revisions have been made in Fig. 3c (page 13) and Fig. 5i (page 21). We have amended "Absortance" to "Absorbance" in the revised manuscript and Supplementary Information. Many thanks for the conscientious referee.

Comment 4:
Citations or a justification is needed on the source of size of ions/molecules (Page 18). This manuscript reports the preparation of a PNIPAM-GO membrane where thermal properties of PNIPAM allow the spacing between the GO sheets to respond with respect to temperature. The work combines experimental and simulation evidence to support the proposed mechanism. When the temperature exceeds a certain value, the GO sheets collapse and shutdown transport of water.
Intermediate temperatures can be used to do selective separations in the several nm to order 10 nm molecular size range. The results seem reasonable and are accompanied by a fair amount of hype.
It would be helpful to back the hype up with comparisons to existing membranes that are used to separate the types of molecules that are considered in this work. Instead, the manuscript largely stays away from any quantitative comparison, preferring to rely on language to convince the reader that the materials are particularly special (from an application perspective) -no doubt, though, that the physical mechanism is interesting.
For example, it would be useful to report permeance and/or permeability data for transport rates as opposed to just flux. That would help the reader to compare these materials to other membranes where, particularly permeance units, are commonly reported.

Response 1:
We appreciate the referee's high evaluation of our work "no doubt, though, that the physical mechanism is interesting". In our work, a pressure of 1 bar was applied to the pressure-driven filtration (described in the Methods section of manuscript, page 25, line 446). Thus the flux data and permeance data are equal in number. We have converted the flux data into permeance data in the revised manuscript and Supplementary Information. And we also added the permeance data of P-GOMs for molecules at 25°C and 50°C in Supplementary Table S2 (revised Supplementary   Information, page 21). In this work, the thickness of P-GOMs is fixed at 1.1 μm in order to achieve a larger temperature-response gating ratio (Fig. 3d in manuscript) and realize the rejection rate of almost 100% for RB at 50°C, thus improving the precision of mixed molecules gradient separation.  (2017)]. Therefore, we can improve the permeance by sacrificing a little selectivity of P-GOMs by using a thinner membrane.
In Fig. 3d of the manuscript, we studied the effect of membrane thickness on temperature-response water gating ratio. We selected 1.1 μm membrane for water gating because it showed a good balance between the permeance and gating ratio. Therefore, the subsequent experiments also used such 1.1 μm membrane for molecular separation. If we decreased the membrane thickness to 0.8 μm, the RB rejection rate slightly decreased from near 100% (1.1 μm) to 98.4% as shown in Fig. R5. But its permeance significantly increased from 2.43 to 20.10 L m -2 h -1 bar -1 . According to the rejection rates In response to these comments, we have made revisions in the revised manuscript (page 19, line 336) and Supplementary Information (page 14, line 143). Fig. R5 The separation performance of 0.8 μm P-GOMs. a) The UV-vis absorption spectra before and after filtering RB from 25°C to 50°C using the 0.8 μm P-GOMs. b) Temperature dependent permeance and rejection rate of the 0.8 μm P-GOMs for RB.

Comment 2:
Presumably the materials aren't useful for desalination, or else data with sodium would have been used. One is left to wonder what the actual separation resolution really is for these materials. The 'hype' would be more justified by doing separations between ions, for example, as opposed to the Cu 2+ and RB cutoff that is reported (i.e., a hydrated 'big' ion versus an organic dye). They authors should at least report the resolution of separations that can reasonably be achieved for dye molecules versus ions using other membrane approaches.

Response 2:
Thanks for referee's helpful comment. In the manuscript, the reason why we used Cu 2+ and RB, two ion/molecule with big size difference, was to better describe the concept of temperature-tuned gradient separation which does not need to change membrane when separating mixed molecules. P-GOMs demonstrated large rejection rate difference for RB at 25°C and 50°C (RB could pass through the P-GOMs at 25°C, while could not at 50°C), and the rejection rate for Cu 2+ was only 16.7% at 50°C. Therefore, we could effectively separate the small Cu 2+ from the mixed ions/molecules solution at 50°C. Then the middle size RB could be separated at 25°C. Besides, we further chose three kinds of saccharides, raffinose (C 18  The concentrations of raffinose, maltopentaose and maltoheptaose were measured by ion chromatography (IC). As shown in Fig. R6, the P-GOMs showed rejection rates of 13.6% for raffinose, 34.3% for maltopentaose and 48.6% for maltoheptaose at 25°C. When the temperature was 50°C, the rejection rates increased to 35.3% for raffinose, 68.4% for maltopentaose and 94.3% for maltoheptaose, respectively. At 50°C, since the P-GOMs demonstrated a relatively high rejection rate for maltoheptaose and a low rejection rate for raffinose, we supposed that we could separate raffinose and maltoheptaose effectively. Subsequently, a mixture of raffinose and maltoheptaose (50:50% by weight) was filtered at 50°C. As shown in Figs. R6d and R6f, the mass content of raffinose in the permeate increased to 84%, indicating that the P-GOMs had a high selectivity towards raffinose and maltoheptaose at 50°C. Therefore, the P-GOMs have the performance of reliability and relatively high separation resolution.
In response to these comments, we have made revisions in the revised manuscript (page 20, line 352) and Supplementary Information (page 15, line 168). maltoheptaose. e) The rejection rates of P-GOMs for raffinose, maltopentaose and maltoheptaose at 25°C and 50°C. f) Source and permeate content of a mixture of raffinose and maltoheptaose.

Comment 3:
Additionally, no rational is given for the low concentrations that were used in the experiments. One is left to wonder whether those conditions are really relevant / how the material would perform if challenged with a more concentrated feed solution.

Response 3:
Thanks for this important comment. Following the reviewer's suggestion, we have performed the separation experiments using more concentrated feed solutions. We chose RB and CBB as examples to demonstrate the question. As shown in Fig. R7, the permeance decreased with the RB concentration increasing from 4 ppm to 200 ppm. The rejection rates for these four concentrations of RB were 99.6%, 93.5%, 93.1% and 91.1%, respectively. It can be seen that there was only a slight rejection rate decrease for RB with the increased concentration. For CBB, the anionic dye, the rejection rates were ~100%, 99.7% and 99.1% using 40 ppm, 100 ppm and 200 ppm feed solutions, and the permeance variation was small (Fig. R8). Therefore, the P-GOMs exhibited a good separation performance for a wide concentration range.   Temperature dependent rejection rates of P-GOMs for 40 ppm, 100ppm and 200 ppm CBB.

Comment 4:
As an aside, it seems unlikely that highly confined water molecules in the nano-channels would behave like a bulk viscous flow (such that the Hagen-Poiseuille equation would be valid). It seems as though a more molecular approach to analyzing transport would be appropriate compared to a continuum approach.

Response 4:
We appreciate this important comment. The understanding of water flow behavior in the nanochannels of graphene oxide membrane is significant to explain the unusual high water permeance. Up to now, many researchers revealed that the water flow in the GO channel is viscous combined with experiments and classical molecular dynamics simulations [Nair, R., Wu, H., Jayaram, P., Grigorieva, . Therefore, we adopted this mechanism in our work. It is indeed that the behavior of confined water molecules in nanochannels plays an important role in fluid transportation. We would like to perform the molecular simulation in our follow-up work.