Can humidity assisted tuning of “permeation channels” in graphene oxide membranes be useful for sea water desalination?

Many layered materials (clays, Mxenes) exhibit gradual change in d(001) spacing us a function of humidity with increments smaller than size of water molecule. This effect is not related to size of “permeation channels” in GO membranes according to common structural model of GO hydration.

The paper by J.Abraham at al1 was presented to mass media in 2017  with titles like:

  “Graphene sieve turns seawater into drinking water”

However, no experiments were performed with sea water and no pure water was produced in any experiments.

 All sea salts were found to diffuse through pores of graphene oxide membranes “precisely controlled” by humidity . Moreover filtration experiments in this study were performed only with pure water. Pure  water indeed penetrated across the membranes. 

Citing the study: “Due to the difficulties of fabricating samples with such large areas for pressure filtration, systematic filtration experiments with salt water were not performed”.  

So, what was actually studied?  First part is about diffusion from  compartment with salted water into compartment with pure water along the planes of glue encapsulated membranes.    Authors assumed that the size of  “capillaries”  in GO membranes  is smaller than size of hydrated ions and hoped that membrane will stop diffusion of these ions. However, all salts were found to diffuse through the “precisely tuned” pores of GO membranes. New desalination mechanism proposed as an idea did not function in experiments.

The second part of study is demonstration of forward osmosis in partly reduced GO membranes.  That is, to make very concentrated sugar solution  less concentrated. There is no doubt that desalination of water using osmosis-related methods is possible using graphene oxide membranes.  In fact, high pressure reverse osmosis experiments with large area graphene oxide membranes were demonstrated already in 1960-s reporting NaCl  rejection rates up to 94%.  The study was performed using standard industrial type of testing and published by the US Department of Interior in 1970.2  Note that in these old studies experiments were performed across the planes of membranes with relatively large area, new experiments in 2017 only along the planes with extremely small cross section area.

The industrial research with graphene oxide membranes was abandoned in 1970-s when much better reverse osmosis membranes were developed and introduced into industrial desalination plants. The first generation membranes were rather cheap and based on very common acetylated cellulose .  The reverse osmosis membranes  were further developed over past 50 years and now third generation is implemented in industrial desalination plants.

 It fact the study J.Abraham1   is yet to achieve the results reported back in 1970. The main problem of graphene oxide membranes is very poor mechanical stability. The 50 years old study solved this problem by encapsulating GO membranes into a clay envelop. The encapsulation enabled high pressure reverse osmosis tests on relatively large area membranes.  The 2017 study still did not overcome the issue of mechanical stability and that is why forward osmosis is used in order to avoid using high pressure. Reverse osmosis allows to obtain pure water while forward osmosis only dilutes concentrated draw solutions and requires additional methods to extract water. Nevertheless, forward osmosis membranes are also very common.   

   The forward osmosis membranes based e.g. on cellulose are cheap and superior to graphene oxide membranes in many ways.  

      But what about those precisely tuned “permeation channels” in graphene oxide membranes described in the study by J.Abraham et al and desalination?  

Let’s see how this discovery is described in a motivation for Prince Sultan Bin Abdulaziz International Prize for Water received by two of the authors.

  “.....membranes based on graphene oxide laminates ....act as atomic-scale sieves allowing water to pass through while blocking salts and other molecules, a mechanism completely different from that of polymer-based membranes.” 

But that is not what readers find in the publication. The study by J.Abraham1 demonstrated that  all sea salts  diffuse along the membranes, even those  which were glue encapsulated   at  low humidity.   The main result is that salts diffuse along the membrane if there is any swelling  present, even if rather small.

    It is possibly not surprising result for researchers who are familiar with most common model of multilayered graphene oxide hydration and effect of interstratification.  The “inter-layer distance” found by X-ray diffraction is not equal to size of permeation channels in graphene oxide according to well-known mechanism of GO hydration. The shift of lattice spacing for XRD reflection  is not directly equal the change in size of true “permeation channels”. It is  a feature relevant to specifics of averaging in  XRD method in a system with randomly packed hydrated and not hydrated layers. 

Authors do not propose any alternative ideas to explain how the 1Å increments  can be achieved by intercalation of ~3Å size water molecules and do not mention interstratification.  Nevertheless, J.Abraham et al 1 insist in their reply to my comment that 1Å scale changes in d(001) spacing found by XRD correspond to true and uniform inter-layer distance for “permeation channels” in graphene oxide (GO) membranes.  

      Unfortunately, the reply of authors to my letter reveals misunderstanding of interstratification. Moreover, the incorrect assumptions about interstratification are presented by author’s  as a part of some “Talyzin’s model”.

  Citing the reply:  “Indeed, the X‐ray peak position corresponds to the periodicity (interlayer spacing) within our GO membranes and is not an average of two distinct XRD peaks coming from dry and hydrated regions.”   

    That is, authors incorrectly assume that TWO diffraction peaks will be observed by XRD independently on the type of layering.    However,  my letter stated  something completely different: “In case of random stacking of layers only ONE diffraction peak is observed for d(001) of GO while position of this reflection shifts depending on proportion between the numbers differently hydrated layers.” 

   It is not in any way “Talyzin’s model” but the most common explanation for continuous shifts of 00ℓ reflections in humidity dependent swelling of several common layered materials.

      It is well known that continuous changes in d(001)   (or changes in swelling stated upon change of temperature) can be observed by X-ray diffraction us a function of humidity in several kinds of layered hydrophilic materials.3-5 Small increments in d(001) are commonly observed in a process of hydration of clays6 and graphite/graphene oxides7 and  mxenes.8 The d(001) value is often cited simply as  “inter-layer distance” of graphite/graphene oxide structure and that  resulted in the controversy.

     The difference between randomly layered (one XRD peak) and segregated layering (two XRD peaks) is shown schematically  in the figure below.

Effects of interstratification were described first in 1930-s, cited as well known in 1960-s and broadly accepted in modern times.9

For example, fundamental study of graphite oxide hydration by A.Lerf et al  published in 200610 describes XRD data as following:  

“..continuous increase in the basal spacing follows as more water is intercalated. Such a behavior is indicative for a nearly random interstratification of differently filled interlayer spaces as suggested earlier [14]. With increasing humidity the relative number of both layer distances changes and this shifts the layer distance towards higher values.”   The ref.14 in this citation is to study by Ruess from 1946.

Another citation from this paper:

"Hydration follows a random interstratification of non-hydrated interlayers and fully occupied water monolayers at layer distances below 8.5Å” 

One more example is paper by De Boer, J. and Van Doorn, A. (1958) "Graphite Oxide. V. The Sorption of Water."  

".. the röntgenographically  determined c-spacing is wrong: that is to say: the c-spacing determined in this manner is only an average value and does not  represent the actual situation."

Citation from the reply by authors: Talyzin suggests that the resulting GO membranes had both hydrated and dry layers, and the interlayer spacing observed using XRD was their average. We disagree.”

 “.. we agree that the observed strong changes in ion permeation, which are caused by rather small changes in the interlayer spacing (smaller than the size of water molecules), are puzzling and remain to be fully understood. Recent molecular dynamics simulations already provide some insight into this [3,4].”

Authors disagree that interlayer distance in GO membranes  is averaged over hydrated and not hydrated layers in XRD. As alternative explanation they propose  that ….interlayer distance in GO membranes  is averaged over hydrated and not hydrated layers, that is according to cited paper from 2019.   

The theoretical model  presented by C.D.Williams et al   (2019)  includes many layers of differently hydrated GO. The study   comes to the same conclusions as the conclusions which are confronted in the “Reply”.  

 Citation: “For a water content of 15% (dGO = 0.80 nm), the membrane water is predominantly found in monolayers, but dehydrated regions are also observed.”    

 That is, you can not make the size of “permeation channels”  smaller than the size of water monolayer.   The 0.8 nm value calculated in this model  is averaged over many regions and layers.     It is cited as “average” all throughout the study. Similar kind of averaging occurs in analysis XRD data. The d(001) is averaged value of hydrated and not hydrated layers over many layers.   

   I note that the model  do not include hydrophobic  “graphene capillaries” postulated in earlier studies by A.Geim and R.Nair  (2012, 214) to explain permeation across GO membranes.  

The study C.D.Williams et al   (2019) also  correctly cites our paper from 2014: “The coexistence of two distinct interlayer distances upon hydration can be attributed to interstratification (i.e., randomly packed hydration states throughout the interlayer space of the membrane).”   

So, he authors first disagree with interstratification, then agree and again disagree in following text.

Citation from reply. “This is inconsistent with the other observations provided in our article, which show that water permeation through differently hydrated GO membranes changes only marginally, whereas ion permeation rates exhibit exponential changes.”

How to compare permeation of water and ions if these are measured in different setups, completely different conditions and even expressed in different units?

 The original study from 2017 shows in one figure permeation data for ions and pure water. Permeation rates for ions were obtained in experiments with diffusion of salts  from compartment with solution to compartment with pure water driven by difference in concentration. The units are “Moles per Hour per Square Meter”.   The concentration difference is 1M at the starting point of experiment and decreases over time due to diffusion of salt. 

Water can permeate in these experiments only  in the direction opposite  to diffusion of ions  due to osmotic flow. This permeation could possibly be studied.  But that is not how water permeation was measured for publication.

 Instead, the water permeation was measured in completely different permeation setup where permeation is driven by evaporation  from the membrane surface.  In this case one side of membrane is exposed to water vapour and second side is open for evaporation of water into humidity free atmosphere.

The specified units are: “Liter per Hour per Square Meter per Bar”   What was used here as a pressure is not specified.  It is not pressure driven filtration and possibly partial water pressure is cited as a driving force .  

 One can only wonder how amount in liter per bar can be compared to mole per (1 (?) mole concentration gradient) in experiments with completely different setups and driving forces behind permeation.  What is the meaning of this for desalination?   

Once again, interstratification is cited in many studies of graphite oxides and graphene oxides published over past decade to explain gradual shifts of XRD peaks related to inter-layer distance. However, hydration of graphene oxide appeared to be even more complex due to inherently inhomogeneous oxidation of its surface.

          In 2014 we published microscopic study of hydration for  two-layered graphene oxide.2  The distance between  two layers of GO was measured as a function of humidity using SFM in low and high resolution mode.  Low resolution scans showed that “inter-layer distance” between two GO sheets gradually changes upon increase in humidity with visible increments smaller than 1Å. The two-layered sample  excludes possibility of interstratification.  However, high resolution scans revealed that hydration of GO interlayer is inhomogeneous on the scale of few nm.  Hills were found in hydrated areas and valleys in less hydrated.  The visibly gradual change of “inter-layer distance” in low resolution scans was then explained by effects of averaging over many differently hydrated nm-sized areas. This effect we proposed to name intrastratification to reflect inhomogeneous hydration within one inter-layer rather than in many stacked layers.

   I note that these results also demonstrate that GO flakes are perforated by many holes. It is evidenced by formation of “hills” due to local hydration. Without holes the hydration would start on the edges of flakes and propagate into the inner parts as it is observed for penetration and escape of water under high quality hole-free graphene on mica.11  I long argued  that  “superfast flow” rate of water in GO membranes  is result of incorrectly applied idealized geometrical model which does not take into account existence of holes in GO sheets .12 The SFM study also shows that effect of “pillaring” by water molecules is not observed on few nm scale, thus not allowing formation of interconnected “graphene capillaries”  speculated in earlier studies.13 

 Surprisingly, our study14 was cited in the paper by J.Abraham et al1  (2017) only as evidence of true change in the size of GO membrane permeation channels with increments as small as 1-1.5Å.  The most important part of study about high resolution scans and inhomogeneous hydration was ignored.

    It is not the size of “permeation channels” which decreases at low humidity, it is the relative number of hydrated channels and hydrated areas which become smaller.  It is also likely that hydrated layers are not anymore well interconnected at low humidity levels  which results in exponential drop of diffusion rates for salt ions. Smaller humidity results in less hydration and smaller number of hydrated inter-layers. 

      The most common high pressure reverse osmosis method of sea water desalination is not yet  challenged by graphene oxide membranes.  

 I note that my letter was submitted in August 2018 and with all my best efforts could not include discussion of papers published in 2021.  

   Finally, the question about calculation of rejection rates in forward osmosis was not included in my original submission. It was cited independently by two reviewers. I barely mentioned this question in my letter following request of editor not to expand this part.   

Authors admitted in their "Reply" that  the calculation described in their original study was erroneous but insist that it was done in a different way.  I note that several later studies copied incorrect method of calculation citing paper by J.Abraham et al. 1


This post was updated on 27 of January. Two comments to the "reply" of authors were added.  


  1. Abraham, J.; Vasu, K. S.;  Williams, C. D.;  Gopinadhan, K.;  Su, Y.;  Cherian, C. T.;  Dix, J.;  Prestat, E.;  Haigh, S. J.;  Grigorieva, I. V.;  Carbone, P.;  Geim, A. K.; Nair, R. R., Tunable sieving of ions using graphene oxide membranes. Nat Nanotechnol 2017, 12 (6), 546-+.
  2. Bober, E. S.; Flowers, L. C.;  Lee, P. K.;  Sestrich, D. E.;  Wong, C.-M.;  Gillam Sherman, W.;  Johnson, S.; Horowitz, R. H., Final report on Reverse osmosis membranes Containing Graphitic Oxide. Research and development progress report No.544, US Department of the Interior, reprints from the colelction of the University of Michigan Library 1970, 1-113.
  3. Hofmann, U.; Frenzel, A.; Csalán, E., Die Konstitution der Graphitsäure und ihre Reaktionen. Justus Liebigs Annalen der Chemie 1934, 510 (1), 1-41.
  4. You, S. J.; Sundqvist, B.; Talyzin, A. V., Enormous Lattice Expansion of Hummers Graphite Oxide in Alcohols at Low Temperatures. Acs Nano 2013, 7 (2), 1395-1399.
  5. Talyzin, A. V.; Luzan, S. M.;  Szabo, T.;  Chernyshev, D.; Dmitriev, V., Temperature dependent structural breathing of hydrated graphite oxide in H2O. Carbon 2011, 49 (6), 1894-1899.
  6. Mcatee, J. L., Determination of Random Interstratification in Montmorillonite. Am Mineral 1956, 41 (7-8), 627-631.
  7. Iakunkov, A.; Talyzin, A. V., Swelling properties of graphite oxides and graphene oxide multilayered materials. Nanoscale 2020, 12 (41), 21060-21093.
  8. Celerier, S.; Hurand, S.;  Garnero, C.;  Morisset, S.;  Benchakar, M.;  Habrioux, A.;  Chartier, P.;  Mauchamp, V.;  Findling, N.;  Lanson, B.; Ferrage, E., Hydration of Ti3C2Tx MXene: An Interstratification Process with Major Implications on Physical Properties. Chem Mater 2019, 31 (2), 454-461.
  9. Ban, S.; Xie, J.;  Wang, Y. J.;  Jing, B.;  Liu, B.; Zhou, H. J., Insight into the Nanoscale Mechanism of Rapid H2O Transport within a Graphene Oxide Membrane: Impact of Oxygen Functional Group Clustering. Acs Appl Mater Inter 2016, 8 (1), 321-332.
  10. Lerf, A.; Buchsteiner, A.;  Pieper, J.;  Schottl, S.;  Dekany, I.;  Szabo, T.; Boehm, H. P., Hydration behavior and dynamics of water molecules in graphite oxide. Journal of Physics and Chemistry of Solids 2006, 67 (5-6), 1106-1110.
  11. Severin, N.; Lange, P.;  Sokolov, I. M.; Rabe, J. P., Reversible Dewetting of a Molecularly Thin Fluid Water Film in a Soft Graphene-Mica Slit Pore. Nano Lett 2012, 12 (2), 774-779.
  12. A.V.Talyzin, Graphene oxide membranes: on the absence of "graphene capillaries", "ultrafast flow" rate and "ultraprecise sieving". arxiv 2018,
  13. Nair, R. R.; Wu, H. A.;  Jayaram, P. N.;  Grigorieva, I. V.; Geim, A. K., Unimpeded Permeation of Water Through Helium-Leak-Tight Graphene-Based Membranes. Science 2012, 335 (6067), 442-444.
  14. Rezania, B.; Severin, N.;  Talyzin, A. V.; Rabe, J. P., Hydration of Bilayered Graphene Oxide. Nano Lett 2014, 14 (7), 3993-3998.
  15. Kim, S.; Choi, S.;  Lee, H. G.;  Jin, D.;  Kim, G.;  Kim, T.;  Lee, J. S.; Shim, W., Neuromorphic van der Waals crystals for substantial energy generation. Nat Commun 2021, 12 (1).


Please sign in or register for FREE

If you are a registered user on Springer Nature Materials Community, please sign in