Scalable Graphene-Based Membranes for Ionic Sieving with Ultrahigh

Letter pubs.acs.org/NanoLett

Scalable Graphene-Based Membranes for Ionic Sieving with Ultrahigh Charge Selectivity Seunghyun Hong,† Charlotte Constans,†,‡ Marcos Vinicius Surmani Martins,†,§ Yong Chin Seow,† Juan Alfredo Guevara Carrió,†,∥ and Slaven Garaj*,†,‡,⊥,# †

Centre for Advanced 2D Materials, National University of Singapore, Singapore 117542 Department of Physics, National University of Singapore, Singapore 117551 § Department of Materials Science and Engineering, National University of Singapore, Singapore 117575 ∥ Engineering School, Presbyterian University Mackenzie, São Paulo 01302-907, Brazil ⊥ NUS Nanoscience & Nanotechnology Institute, National University of Singapore, Singapore 117581 # Department of Biomedical Engineering, National University of Singapore, Singapore 117583 ‡

S Supporting Information *

ABSTRACT: Nanostructured graphene-oxide (GO) laminate membranes, exhibiting ultrahigh water flux, are excellent candidates for next generation nanofiltration and desalination membranes, provided the ionic rejection could be further increased without compromising the water flux. Using microscopic drift−diffusion experiments, we demonstrated the ultrahigh charge selectivity for GO membranes, with more than order of magnitude difference in the permeabilities of cationic and anionic species of equivalent hydration radii. Measuring diffusion of a wide range of ions of different size and charge, we were able to clearly disentangle different physical mechanisms contributing to the ionic sieving in GO membranes: electrostatic repulsion between ions and charged chemical groups; and the compression of the ionic hydration shell within the membrane’s nanochannels, following the activated behavior. The charge-selectivity allows us to rationally design membranes with increased ionic rejection and opens up the field of ion exchange and electrodialysis to the GO membranes. KEYWORDS: Graphene oxide membranes, ionic permeability, surface charges, ion exchange, ionic sieving raphene-based membranes with ultrahigh water flux1,2 and ionic sieving properties2−8 attracted recently significant attention, as a severe strain on the fresh water supply9,10 precipitated a strong research interest in new materials for water purification and desalination. Nanostructured graphene-oxide (GO) membranes,1,3,11−13 scalable, inexpensive, thermally and chemically robust, and integratable with current technologies,14−18 are particularly enticing candidates for the next-generation, high-performance separation membranes.10,19−21 The GO membranes consist of stacked layers of impermeable graphene sheets, L = 1−10 μm in size, spaced by d = 0.9−1.2 nm via functionalized, mostly oxygencarrying groups.3,13 The chemical groups are coalesced into nanoscale domains, delimiting a percolative network of pristine graphene channels, which could accommodate a few layers of water exhibiting frictionless flow. 1,22 Previous experiments,3,15,16,23−25 measuring salt diffusion through centimeterscale membranes over a period of hours, showed no permeation for ions with hydration rates above size cutoff of RH ≈ 4.5 Å and mostly unvarying permeation rate for smaller ions. Those experiments, due to their nature, are ineffective in disentangling all the physical mechanisms contributing to the permeability, are unable to distinguish permeability of different constituting ions in the salt, and could be prone to artifacts due to external defects and tears over larger areas of the membrane. To

G

© 2016 American Chemical Society

understand intrinsic membrane properties, we implemented a highly sensitive drift−diffusion technique, revealing ultrahigh charge-selectivity of the GO membranes. We measured the ionic permeability of a 3 μm thick GO membrane, mounted across an array of 200 × 200 nm2 apertures in a 300 nm thick, free-standing, insulating SiNX layer on a Si substrate chip (Figure 1a). By limiting the exposed membrane area to ∼5 μm2 and keeping it relatively thick, we ensured there are no unintended cracks and defects that would skew the results. The GO membrane and its constituting GO crystallites were extensively characterized using atomic force microscopy, X-ray diffraction, Fourier transform infrared spectroscopy, etc. (see Supporting Information). The membrane chip was inserted in a custom-build fluidic cell, so that it separated two compartments, each subsequently filled with ionic solutions electrically contacted with Ag/AgCl electrodes. The electrodes were connected to a sensitive patch-clamp amplifier (Axopatch 200B), sourcing voltage at a sweep rate of 5 mV/5 s (step function) across the membrane and measuring ionic currents with 10 pA precision (no hysteresis was observed Received: September 12, 2016 Revised: December 13, 2016 Published: December 22, 2016 728

DOI: 10.1021/acs.nanolett.6b03837 Nano Lett. 2017, 17, 728−732

Letter

Nano Letters

Figure 1. Drift−diffusion experiment. (a) Schematics of the experimental setup: the graphene oxide membrane was mounted on a freestanding SiNx membrane with an 12 × 12 array of square-shaped windows, separating two electrolyte-filled reservoirs; Ag/AgCl electrodes in each reservoir are used to apply an electric potential across the GO membrane and to measure the ionic currents flowing through the membrane. (b,c) Depiction of the ionic flow across the membrane driven by the concentration gradient (diffusion) and by the voltage difference (drift), respectively. (d) The ionic current−voltage characteristic of the membrane for different salts, measured under the concentration gradient 0.1 M/0.01 M across the membrane.

Figure 2. Charge-selective permeability. (a) Permeation rates (p) for different cations (circles) and corresponding chloride counterions (open and filled squares) as a function of hydrated radius (RH) of the cations. The filled square represents the chloride permeability when in RbCl solution, where the hydration radii are very similar for both ions; the two-headed arrow shows the permeation difference resulting purely from the chargerejection effects. The solid blue line is a guide to eye. (b,c) Schematics of the dominant ion rejection mechanisms: size exclusion (b) and electrostatic repulsion (c). (d) The cationic selectivity of GO membranes for different salts, reaching values in excess of 90%. Inset: the permeation rates of chloride ions as a function of the valence of the position counterion in the salt, revealing the effect of the correlated charge inversion in the subnanometer channels.

Figure 1d shows representative current−voltage (I−V) curves, measured at a fixed concentration gradient: the slope of the curve is indicative of Idrift; whereas membrane potential Vm = V(I = 0) is indicative of Idiff. More precisely, we extract the individual permeabilities by modeling the total current density J(Δc,ΔV) across the membrane26 using the equation:

for such low rates). The polydimethylsiloxane (PDMS) gasket seal precluded ionic solution from leaking around the edges of the membrane. To discern the separate permeabilities of cations (P+) and anions (P−) in the salt, we implemented the drift−diffusion technique to measure ionic currents driven by both the voltage and the concentration gradient (Figure 1b). The fluidic compartments were filled with different concentrations of a salt, and we could measure diffusive current across the membrane for zero applied voltage Idiff ≈ (P+ − P−)Δc, driven by the concentration gradient Δc = chigh − clow. As we applied a voltage difference ΔV across the membrane, the added electrophoretic component to the overall current is Idrift ≈ (P+ + P−)ΔV (Figure 1c). Measuring the two current components, we could deduce both P+ and P− permeabilities.

J=

∑ PXzX X

2F

(

z F ΔV

X ΔV [X ]f − [X ]p exp RT z F ΔV RT 1 − exp X

2

(

RT

)

)

For each ionic species X in the solution, PX is membrane permeability, zX is the valence, and [X]f and [X]p are the ionic concentrations in the feed and permeate chambers, respectively. Potential across the membrane ΔV was adjusted for the electrodes’ redox potential; R is the universal gas constant; F is 729

DOI: 10.1021/acs.nanolett.6b03837 Nano Lett. 2017, 17, 728−732

Letter

Nano Letters

Figure 3. Molarity and pH dependence of conductance and ionic permeation. (a) Current−voltage (I−V) curves across the membranes at KCl salt concentration cKCl = 10 mM, measured for different pH values. (b) Conductance vs pH. (c) Permeation rates for potassium and chloride ions for different pH values. (d) The ionic conductance vs molarity (circles) deviates from the Ohmic behavior (full line), even at high salt concentrations, due to subnanometer channel heights. (e,f) Molarity dependence for the permeation rates (e) and for the cation selectivity (f). Dashed curves in all the graphs are fit to the mean-field model, discussed in the text.

predicted by the electric double layer (EDL) model.27 Permeability of Cl− ions in monovalent salts remained independent of counterions (Rb+, K+, Na+, Li+); and the cation selectivity S+ = P+/(P+ + P−) reached values in excess of 90% (Figure 2d). Interestingly, the EDL model breaks down in the case of chloride salts with divalent and trivalent cations, and P(Cl−) reverts to the value predicted for uncharged channels (Figure 2a,d). We attribute this effect to correlation-induced charge inversion,30 where multivalent ions overcompensate monovalent surface groups, leading to a sharp drop, or even an inversion, of the effective surface charge. A similar effect has been observed previously in highly charged protein channels,31−33 such as bacterial porin OmpF, and in narrow silica channels.34 To further investigate the ionic selectivity of GO membranes, we performed a series of drift−diffusion and ionic conductivity measurements using KCl aqueous solutions for a range of pH and molarity values. Figure 3a shows current−voltage (I−V) curves at same salt concentration on both sides, cKCl = 10 mM, measured for different pH values (see also Figure S3a). At this low molarity, nonlinear nature of the I−V curves at larger voltages is likely due to overlimiting currents driven by the concentration polarization at the surface of the membrane.35 The ionic conductance of the membrane was calculated from the slopes of the I−V curves in the Ohmic regime at low voltage (Figure 3b). The increase in pH (reduction in hydronium concentration) leads to increased dissociation of the carboxyl and hydroxyl groups within the GO sheets:

Faraday’s constant; and T is the temperature. For details on the model and the method, see Supporting Information. To elucidate the ionic selectivity of the GO membranes, we investigated the permeability of a wide selection of aqueous salt ions, with varying ionic charges and spanning a wide range of effective hydrated ionic volumes. Figure 2a depicts the permeation rates (p = P·Δc) of different cations (circle) and their corresponding Cl− counterions (squares) as a function of the cation’s hydration radii.27,28 Two general trends are revealed: (a) cation permeability decreases exponentially with increased hydration radius, followed by the sharp cutoff at RH ≈ 4.6 Å; and (b) permeability of the negatively charged Cl− ion is suppressed by an order of magnitude compared to the positive K+ and Rb+ ions, despite all those ions having very similar hydration volumes. We conclude that the two dominant mechanisms for the ion rejection in GO membranes are size exclusion due to compression of the ionic hydration shell in narrow channels8,29 and electrostatic repulsion due to membrane surface charge (Figure 2b,c). The earlier diffusion experiments3,15,16,23−25 measured the combined permeability of all the salt ions, which is determined by the value for the least permeable species in a salt; for monovalent salts they were actually measuring permeability of the chlorine counterion, not cations. This led to apparent size-independent permeability for ions with hydration radii below the cutoff size defined by the channel height (implying rigid hydration shells around ions). Instead, by properly separating cations and anions, we observe the exponential dependence of the permeability on an ion’s hydration radius, consistent with the compressible hydration shell model, where coordinated water molecules could rearrange or detach themselves to allow passage of the hydrated ion through a narrow channel.8,29 We postulate that the high charge selectivity of the GO membranes is a result of the negatively charged nanochannels in a GO membrane, due to the protonable oxygen groups. This leads to the expulsion of the negatively charged Cl− ions from nanochannels and suppression of the anionic permeability, as

grapheneOH ⇄ grapheneO− + H+

This leads to an increase in negative surface charge density in the graphene nanochannels and is reflected in an increased conductance and current rectification. The ionic currents associated with the excess hydronium (H3O+) or hydroxide (OH−) ions are subtracted as shown in Figure S4a. At higher pH, we also observe an increase in p(K+), a decrease in p(Cl−), 730

DOI: 10.1021/acs.nanolett.6b03837 Nano Lett. 2017, 17, 728−732

Letter

Nano Letters and an increase in cation selectivity S+ (Figure 3c), all consistent with the increase in the nanochannels’ charge. The strong surface charge effects were revealed in the membrane’s conductance G0 variation with the electrolyte concentration c (Figure 3d). Starting from c = 1 M, the observed G0 immediately deviates from the expected linear regime for a charge-neutral membrane (black solid line), indicating the compression of the EDL in the nanochannels even at high ionic strengths. In contrast, the charge effects were previously observed to dominate the conductance in solid-state constrictions only at much lower salt concentrations.36−38 We note that the cation selectivity, as deduced from ionic permeabilities, could reach as high a value as S+ = 96% at low salt concentration (Figure 3e,f). To gain insight into the surface charge-driven ionic transport, we applied mean-field theoretical model based on the Poisson− Boltzmann and Navier−Stokes equations (see the Supporting Information for more details). The model fits the observed pH and molarity dependence of both the conductivity and the charge selectivity well (Figure 3), assuming the ions flow in pristine graphene nanochannels with an effective height of hG = 0.9 nm, an effective width in the range of wG ≈ 50 nm, an effective channel length of Leff = 0.4 mm, and a constant density of active site on the sidewalls corresponding to one protonable charged site per 2 nm (Figure S6). A crucial assumption of the model is the infinite-slip boundary condition for the water flow at the top and bottom graphene surfaces, and no-slip condition at the oxidized sidewalls. The large slip-length is consistent with the effect of frictionless water flow, as reported in GO membranes.1 The other possible geometries could not replicate the observed pH and molarity dependence of the conductance (see Supporting Information). We employed the same set of parameters to concurrently simulate all the independent experiments. The parameters deduced from the model are within the range expected for the GO membrane, despite the crude approximations. Although this continuous-media model has a limited scope at nanometer length scales, it has been shown to capture the relevant physics and to give sufficient semiquantitative insight, while the intermolecular and steric interactions are renormalized into the effective hydrodynamic dimensions.36,39 In conclusion, we have shown that the ion-rejection in graphene-oxide membranes is driven as much by the electrostatic repulsion (defined by the nanochannel surface charge) as it is by the activated size-exclusion (defined by the nanochannel height). Hence, the engineering of the surface charge of the membrane offers a new venue for increasing the overall salt rejection, without constraining the water flux. We have demonstrated that the GO membranes exhibit ultrahigh charge selectivity, reaching up to 96%, driven by the negative surface charge of the oxygen-carrying functional groups in the membrane’s nanochannels. Coupled with their high-durability and scalability, the GO membranes are well positioned for applications in high-performance ion exchange and electrodialysis technologies.

■

■

Physiochemical characterizations of graphene oxide nanosheets; interplanar spacing expansion of graphene oxide membranes in water; quantitative analysis of ion selectivity across the membranes; calculation of ionic conductance and surface charge density; pH-dependent ionic conductances and surface charge densities; effects of excess hydronium or hydroxide ions; pH-dependent drift−diffusion measurements; mean field model for ion transport in the nanochannels; validation of the analytical continuum models; ion strength-dependent ionic conductance and cation permselectivity; voltage drop across bare SiNx scaffold (PDF)

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Slaven Garaj: 0000-0001-5529-4040 Author Contributions

H.S. and S.G. conceived the concept of the study. H.S. performed the experiments and analyzed the data together with C.C. and J.A.G.C. H.S., M.V.S.M., Y.C.S., and J.A.G.C. fabricated and characterized the membranes. The continuum model analysis was performed by C.C. The manuscript was written by H.S. and S.G. with comments and input from all authors. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We acknowledge support from the National Research Foundation, Prime Minister’s Office, Singapore, under the NRF Fellowship Program (Award No. NRF-NRFF2012-09) and Competitive Research Program (Award No. NRF-CRP132014-03). J.A.G.C. also cknowledges support by Brazilian agency FAPESP (2013/21190-2). The authors are grateful to Dr. Eugene Choo of ZEISS Advanced Imaging Centre (Singapore) for assisting the FIB preparation of nanopores.

■

REFERENCES

(1) Nair, R. R.; Wu, H. A.; Jayaram, P. N.; Grigorieva, I. V.; Geim, A. K. Science 2012, 335 (6067), 442−444. (2) Cohen-Tanugi, D.; Grossman, J. C. Nano Lett. 2012, 12 (7), 3602−3608. (3) Joshi, R. K.; Carbone, P.; Wang, F. C.; Kravets, V. G.; Su, Y.; Grigorieva, I. V.; Wu, H. A.; Geim, A. K.; Nair, R. R. Science 2014, 343 (6172), 752−754. (4) Garaj, S.; Hubbard, W.; Reina, A.; Kong, J.; Branton, D.; Golovchenko, J. A. Nature 2010, 467 (7312), 190−3. (5) O’Hern, S. C.; Stewart, C. A.; Boutilier, M. S. H.; Idrobo, J.-C.; Bhaviripudi, S.; Das, S. K.; Kong, J.; Laoui, T.; Atieh, M.; Karnik, R. ACS Nano 2012, 6 (11), 10130−10138. (6) Celebi, K.; Buchheim, J.; Wyss, R. M.; Droudian, A.; Gasser, P.; Shorubalko, I.; Kye, J. I.; Lee, C.; Park, H. G. Science 2014, 344 (6181), 289−92. (7) Surwade, S. P.; Smirnov, S. N.; Vlassiouk, I. V.; Unocic, R. R.; Veith, G. M.; Dai, S.; Mahurin, S. M. Nat. Nanotechnol. 2015, 10 (5), 459−464. (8) Sint, K.; Wang, B.; Kral, P. J. Am. Chem. Soc. 2008, 130 (49), 16448−16449. (9) Service, R. F. Science 2006, 313 (5790), 1088−1090. (10) Elimelech, M.; Phillip, W. A. Science 2011, 333 (6043), 712− 717.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.6b03837. 731

DOI: 10.1021/acs.nanolett.6b03837 Nano Lett. 2017, 17, 728−732

Letter

Nano Letters (11) Li, H.; Song, Z.; Zhang, X.; Huang, Y.; Li, S.; Mao, Y.; Ploehn, H. J.; Bao, Y.; Yu, M. Science 2013, 342 (6154), 95−98. (12) Kim, H. W.; Yoon, H. W.; Yoon, S.-M.; Yoo, B. M.; Ahn, B. K.; Cho, Y. H.; Shin, H. J.; Yang, H.; Paik, U.; Kwon, S.; Choi, J.-Y.; Park, H. B. Science 2013, 342 (6154), 91−95. (13) Raidongia, K.; Huang, J. J. Am. Chem. Soc. 2012, 134 (40), 16528−16531. (14) Qiu, L.; Zhang, X.; Yang, W.; Wang, Y.; Simon, G. P.; Li, D. Chem. Commun. 2011, 47 (20), 5810−5812. (15) Han, Y.; Xu, Z.; Gao, C. Adv. Funct. Mater. 2013, 23 (29), 3693−3700. (16) Hu, M.; Mi, B. Environ. Sci. Technol. 2013, 47 (8), 3715−3723. (17) Dikin, D. A.; Stankovich, S.; Zimney, E. J.; Piner, R. D.; Dommett, G. H. B.; Evmenenko, G.; Nguyen, S. T.; Ruoff, R. S. Nature 2007, 448 (7152), 457−460. (18) Eda, G.; Chhowalla, M. Adv. Mater. 2010, 22 (22), 2392−2415. (19) Pendergast, M. M.; Hoek, E. M. V. Energy Environ. Sci. 2011, 4 (6), 1946−1971. (20) Liu, G.; Jin, W.; Xu, N. Chem. Soc. Rev. 2015, 44 (15), 5016− 5030. (21) Lee, A.; Elam, J. W.; Darling, S. B. Environ. Sci.: Water Res. Technol. 2016, 2 (1), 17−42. (22) Boukhvalov, D. W.; Katsnelson, M. I.; Son, Y.-W. Nano Lett. 2013, 13 (8), 3930−3935. (23) Sun, P.; Liu, H.; Wang, K.; Zhong, M.; Wu, D.; Zhu, H. J. Phys. Chem. C 2014, 118 (33), 19396−19401. (24) Sun, P.; Zhu, M.; Wang, K.; Zhong, M.; Wei, J.; Wu, D.; Xu, Z.; Zhu, H. ACS Nano 2013, 7 (1), 428−437. (25) Sun, P.; Zheng, F.; Zhu, M.; Song, Z.; Wang, K.; Zhong, M.; Wu, D.; Little, R. B.; Xu, Z.; Zhu, H. ACS Nano 2014, 8 (1), 850−859. (26) Hille, B. Ionic Channels of Excitable Membranes; Sinauer: Sunderland, MA, 1984; Chapter 10. (27) Israelachvili, J. N. Intermolecular and Surface Forces: with Applications to Colloidal and Biological Systems; Academic: London, 1985. (28) Marcus, Y. Biophys. Chem. 1994, 51 (2−3), 111−127. (29) Zwolak, M.; Lagerqvist, J.; Di Ventra, M. Phys. Rev. Lett. 2009, 103 (12), 128102. (30) Shklovskii, B. I. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 1999, 60 (5), 5802−5811. (31) Aguilella, V. M.; Queralt-Martin, M.; Aguilella-Arzo, M.; Alcaraz, A. Integr. Biol. 2011, 3 (3), 159−172. (32) García-Giménez, E.; Alcaraz, A.; Aguilella, V. M. Phys. Rev. E 2010, 81 (2), 021912. (33) Alcaraz, A.; Nestorovich, E. M.; López, M. L.; García-Giménez, E.; Bezrukov, S. M.; Aguilella, V. M. Biophys. J. 2009, 96 (1), 56−66. (34) van der Heyden, F. H. J.; Stein, D.; Besteman, K.; Lemay, S. G.; Dekker, C. Phys. Rev. Lett. 2006, 96 (22), 224502. (35) Yossifon, G.; Mushenheim, P.; Chang, Y.-C.; Chang, H.-C. Phys. Rev. E 2009, 79 (4), 046305. (36) Hoogerheide, D. P.; Garaj, S.; Golovchenko, J. A. Phys. Rev. Lett. 2009, 102 (25), 256804. (37) Stein, D.; Kruithof, M.; Dekker, C. Phys. Rev. Lett. 2004, 93 (3), 035901. (38) Smeets, R. M. M.; Keyser, U. F.; Krapf, D.; Wu, M.-Y.; Dekker, N. H.; Dekker, C. Nano Lett. 2006, 6 (1), 89−95. (39) Garaj, S.; Liu, S.; Golovchenko, J. A.; Branton, D. Proc. Natl. Acad. Sci. U. S. A. 2013, 110 (30), 12192−6.

732

DOI: 10.1021/acs.nanolett.6b03837 Nano Lett. 2017, 17, 728−732