Potential Dependent Ionic Sieving Through Functionalized Laminar MoS2 Membranes

Laminar MoS2 membranes show outstanding potential for practical applications in energy conversion/storage, sensing, and as nanofluidic devices. For water purification technologies, MoS2 membranes can form abundant nanocapillaries from layered stacks of exfoliated MoS2 nanosheets. These MoS2 membranes have previously demonstrated excellent ionic rejection with high water permeation rates, as well as long-term stability with no significant swelling when exposed to aqueous or organic solvents. Chemical modification of these MoS2 membranes has been shown to improve their ionic rejection properties, however the mechanism behind this improvement is not well understood. To elucidate this mechanism we report the potential dependant ion transport through functionalized MoS2 membranes. The ionic permeability of the MoS2 membrane was transformed by chemical functionalization with a simple naphthalene sulfonate dye (sunset yellow) and found to decrease by over a factor of ~10 compared to the pristine MoS2 membranes and those reported for graphene oxide and Ti3C2Tx (MXene) membranes. The effect of pH, solute concentration, and ionic size/charge on the ionic selectivity of the functionalized MoS2 membranes is also reported. The potential dependant study of these dye functionalized MoS2 membranes for ionic sieving with charge selectivity should enable future applications in electro-dialysis and ion exchange for water treatment technologies.


Introduction
Two dimensional (2D) materials, in the form of laminar membranes, have been widely studied for water purification applications such as desalination, ion exchange, and electro-dialysis. [1][2][3] Recently, graphene and graphene oxide (GO), when formed into layers of randomly re-stacked nanosheets, have shown to be promising membrane materials, demonstrating high rejection properties towards small ionic solutes while maintaining high water permeation rates due to a network of nanocapillary channels formed between the individual layered materials. [3][4][5][6] However, GO-based membranes are reported to be unstable when submerged in aqueous solutions as the membranes become swollen. These membranes consequently have an enlarged interlayer spacing resulting in poor rejection properties. 7,8 To overcome this, a number of membrane modification strategies have been developed, including additives to crosslink individual layers 8,9 and physical confinement 3 to restrict the membrane's swelling. Membranes modified by these approaches exhibited improved rejection properties but a decrease in the water flux. This drawback will hinder capacity for large scale exploitation in industry.
Other 2D materials have also been reported as suitable to form laminar membranes, e.g. transition metal carbides/nitrides, MXenes (e.g. Ti 3 C 2 and Ti 3 CN), 10,11 and transition metal dichalcogenides, TMDs (e.g. MoS 2 and WS 2 ). [12][13][14] Deng et al. 12 reported that MoS 2 laminar membranes showed good performance with high stability under extreme pH conditions, without any significant expansion of the interlayer distance. These promising MoS 2 membranes are not only stable in a range of aqueous media but also show significantly higher water permeation rates than similar laminar membranes. Despite this promise there are few studies on the ion transport through MoS 2 membranes; however recent reports have demonstrated improved ionic rejection, surpassing similar GO membranes, by chemical functionalization but the mechanism for this improvement is unclear. TMD laminar 3 membranes also demonstrated molecular sieving/separation performance for both organic vapour and liquid media which lends them to extensive uses in membrane technology applications under extreme conditions. 12,14 Chemical functionalization is crucial in altering the desired surface chemistry of 2D materials, providing control over the membrane's properties for use in various applications such as water treatment, 1, 15 energy storage, 16 and photocatalysis 17 . For water purification, we have previously described the functionalization of MoS 2 membranes with organic dyes, which exhibit an increase in water flux by a factor of 4, with high ion rejection (~99%) when compared to membranes of pristine, exfoliated MoS 2 of comparable thickness. 1 These dye functionalized MoS 2 membranes also showed ca. 5-fold increase in water flux compared to the previously described GO-based membranes. 3,4 Ion transport through GO-based membranes under the influence of an applied potential has been studied previously by Hong et al. 18 , where the membranes were found to exhibit high ionic rejection resulting from the electrostatic repulsion and the size exclusion imparted by the negative surface charge and confinement by the nanochannels. These GO membranes were, however, of extremely limited applicability due to their aforementioned tendency to swell in aqueous solutions, leading to poor ionic rejection. Specifically, the interlayer spacing increases to nearly 70 Å when the membrane was completely soaked in deionized water. 7 In this work, we firstly investigated ion mobility through laminar MoS 2 membranes in the presence of an applied potential and salt concentration gradient. As-prepared functionalized membranes are stable in a range of aqueous solutions including acidic and basic media (no delamination was detected). The sunset yellow (disodium 6-hydroxy-5-[(4sulfophenyl)azo]-2-naphthalenesulfonate; anionic dye) functionalized, laminar MoS 2 membranes (MoS 2 /SY) show significant retardation of ion transport as well as capability for size and charge selectivity compared to the pristine MoS 2 membranes, previously reported laminar membranes (GO and Ti 3 C 2 T x ), and commercial polymeric membranes (Nafion membranes). A range of characterization techniques including optical microscopy, electron microscopy (scanning electron microscopy, SEM, and scanning transmission electron microscopy, STEM), powder X-ray diffraction (PXRD), X-ray photoelectron spectroscopy (XPS), and zeta (ζ) potential measurements were employed to determine the quality of the MoS 2 membrane, along with its thickness, stability, chemical composition, and surface charge. Moreover, MoS 2 /SY membranes were also characterized as a function of permeate concentration and solution pH, thereby demonstrating their excellent cation selectivity at low permeate concentration and high pH condition.

Results & Discussion
The ionic mobility of dye functionalized MoS 2 membranes was measured as schematically shown in Figure 1a; after first using the functionalization procedure described in our previous work (see Figure S1-2). 1 Briefly, 1 mM sunset yellow (SY) was used to functionalize the laminar MoS 2 membranes and they were then cleaned to remove any excess dye molecules until no remaining dye was detected, by UV-visible spectroscopy and electro-spray mass spectroscopy. The characterization of the pristine and dye functionalized MoS 2 membranes are provided as Supporting Information. Figure 1b shows the experimental setup for ion transport in a variety of salt solutions, under a concentration gradient, with 100 mM and 10 mM in the feed and permeate reservoirs, respectively, using a four-electrode system. [19][20][21] The MoS 2 membranes were supported on polyvinylidene fluoride (PVDF) membranes during their pressure filtration driven by self-assembly as shown in Figure 1c. Figure 1d Figure S3-4, as well as energy dispersive X-ray (EDX) analysis mapping of Mo and S elements in Figure S4. Figure  whereas the corresponding electrodes in the permeate chamber were connected to the working and sense terminals. The transmembrane potential was cycled using a triangular potential waveform from −200 mV to 200 mV at a rate of 1 mV s −1 with a reverse cycle to ensure there was no hysteresis in the response. Figure 1f shows the I-V characteristics of different valence ions, measured at a constant concentration ratio (10 mM/100 mM). As a function of the size and charge of the ions, the membrane potential (zero-current potential, the applied potential at which there is no net flow of ions) decreased to a more negative potential with increasing cationic charge. The net current at zero applied voltage is indicative 6 of different diffusion rates between cations and anions resulting in the shift of the I-V response along the voltage axis for both the functionalized ( Figure 1f) and pristine MoS 2 membranes ( Figure S7). A negative current at zero applied voltage corresponds to the higher mobility of cations compared to that of Cl − anions, vice versa for a positive current as shown in Figure 1f. The permeation properties of the MoS 2 membrane are therefore transformed by dye functionalization. The effect of SY functionalization is clearly seen in Figure 1g, which reveals that the membrane conductance decreased significantly, with a related fall in the zerocurrent potential resulting in a 2-fold reduction of mobility ratio as determined by the Goldman-Hodgkin-Katz (GHK) equation (Eq. 1), compared to the pristine MoS 2 . The conductance of salt with three different cationic charges (KCl, BaCl 2 , and AlCl 3 ) has been plotted as a function of MoS 2 /SY thickness as shown in Figure 1h. This analysis shows that the trivalent salt with larger hydrated cation radius (AlCl 3 ) has the highest conductance for a given membrane thickness. Moreover, the conductance values decreased dramatically with increasing membrane thickness, from 1 µm to 3 µm, but were only slightly lower for thicker   To quantify the influence of charge on ion transport through the MoS 2 membranes, I-V measurements known as 'drift-diffusion' experiments were performed, as they are driven by both the diffusion due to concentration gradient and the applied voltage difference. 18,22,23 The GHK current equation, which assumes independence of the ion movements across a membrane, was used to express the cation/anion mobility ratio (µ + /µ − ) as shown in Eq. 1. 18, 24- where E m is the membrane potential (zero-current potential, often called the reversal with increasing hydrated cation radii, which changed over one order of magnitude from K + to Al 3+ ions, after the treatment of the chemical functionalization. This is in agreement with previous work studying ion transport through angstrom-scale slits. 25 Moreover, the individual mobility of cations (µ + ) and anions (µ − ) were calculated from the relation between ionic conductivity (σ) and ion mobility: The conductivity of chloride solutions was measured using a relatively high concentration of 0.1 M in both feed and permeate reservoirs (see Table S1-2 for a comparison of the ionic conductivity between the MoS 2 membranes and bulk chloride solutions), which allows the 9 surface charge contribution to be neglected. 18,25,27 Figure 2b plots the ion mobility (•■ symbols) of the cations, and their corresponding chloride counter ions (○□ symbols) within the MoS 2 membranes, as a function of hydrated cation radii 25,28,29 obtained by the combination of Eq. 2 and the calculated mobility ratio. The cation mobility (µ + ) in laminar MoS 2 membranes decreased by ca. 10 and 100-fold for a pristine MoS 2 membrane and the dye functionalized MoS 2 membrane, respectively, compared to ion mobility in bulk solutions ( symbols). 30 Moreover, our measured ion mobilities are also compared to the literature values reported for the same type of laminar membranes (graphene oxide (see Supporting Information: Figure S8) and Ti 3 C 2 T x ) 11,18 and porous polymeric separation membranes (Nafion-117 and Nafion XL perfluorosulfonic acid (PFSA) membranes) 31, 32 as shown in Figure 2b. This analysis indicated that a pristine MoS 2 membrane exhibited the retardation of ion transport of a similar magnitude to previously reported membranes. The variation of Cl − mobility was found to be within 15% and 20% for the pristine and the dye functionalization, respectively, estimated from average Cl − mobility for several ionic salt solutions. This finding is in good agreement with the variation in Cl − mobility found for transport through Ångström-scale slits (±15%). 25 The cation mobility decreased with increasing hydrated cation radius, R H, from K + to Al 3+ with the trend lines shown across cation radii, especially for MoS 2 /SY. The dye functionalized MoS 2 membrane not only significantly suppressed ion transport through the membranes but also decreased the mobility for the cations with the larger hydrated radii. This is clearly revealed by comparing cation mobility between the functionalized and pristine MoS 2 membranes, which correspond to ca. 9.5, 10.3, and 17.9fold reduction in cation mobility for a range of mono-, di-, and trivalent cations, respectively, for the dye functionalized MoS 2 membrane compared to its pristine counterpart. By comparing the similarly sized ions K + (3.31 Å) and Cl − (3.32 Å), 28 the potassium cations are transported through MoS 2 membranes slightly more quickly than the chloride anion (K + /Cl − > 1), an effect attributed to the negatively charged surface of MoS 2 membranes. This was observed for both pristine and dye functionalized MoS 2 membranes.  in Figure 3a. By decreasing the KCl concentration on the permeate side, the diffusion current at zero applied voltage rapidly dropped due to the suppressed transport of Cl − through the MoS 2 /SY membrane as shown in Figure 3b. To understand the effect of diffusive pressure on the ion transport across the MoS 2 membrane, we measured the zero-current potential, as well as the Cl − mobility, for a variety of KCl permeate molarities as shown in Figure 3c. The Cl − mobility exponentially dropped through the MoS 2 membrane with the decreasing salt concentration at the permeate side in agreement with previous literature using graphene oxide membranes, 18 single-layer MoS 2 /graphene nanopores, 22, 23 carbon/boron nitride nanotube, 33,34 and nanochannel slits. 25,27,35 This is due to the effective surface charge of the membrane, which dominates membrane transport at the low salt concentration in the permeate reservoir. 35,36 This suggestion is also supported by the zeta potential measurements showing that the functionalized MoS 2 nanosheets are negatively charged at pH values around 7. The zero-current potentials increased in proportion to the log of the permeate molarity, which results in the increase of the calculated K + /Cl − mobility ratio (K + >> Cl − ). In addition, the ionic selectivity (%S) at different permeate molarity can be deduced from the ionic mobility (µ i ) defined as; (3) Figure 3d shows ion selectivity as a function of KCl concentration in the permeate reservoir.
The selectivity for both ions (K + and Cl − ) are nearly similar (~50%) for the highest concentration ratio (10 mM/100 mM), and diverges with decreasing KCl concentration at the permeate side. This phenomenon can be explained by the increase of the surface zeta potential, due to the decrease of ionic strength as previously reported in pristine MoS 2 membranes. 37 The increasing zeta potentials result in repulsion between negatively charged MoS 2 /SY surface and Cl − ions.   Table S1 for bulk conductivity at various pH). 18,38 Similar behavior with increasing conductance at high pH has been reported for GO membranes, 18 carbon nanotubes, 38 and hBN nanochannels. 25 Iso-pH conditions were maintained for both feed and permeate reservoirs for this drift-diffusion experiment. Figure   4c shows the K + /Cl − mobility ratio of the functionalized MoS 2 membrane and a pristine MoS 2 membrane as a function of pH. A pristine exfoliated MoS 2 membrane exhibited a small change in mobility ratios for a range of pH solutions with slightly increased at basic pH (K + /Cl − > 1) which indicated to partial adsorption of hydroxide ions on the surface of the pristine MoS 2 channels. Interestingly, the mobility ratio of the dye functionalized MoS 2 membrane significantly decreased at low pH values, which is consistent with some protonated MoS 2 /SY surface corresponding to the decreasing negative surface charge of MoS 2 /SY, as indicated by the zeta potential measurements (see Figure S9). By increasing pH values, the mobility ratio dramatically increased due to the greater extent of OH − adsorption at the functionalized sites of the MoS 2 /SY surface. This significant change in mobility ratio indicated the altered surface chemistry of MoS 2 after functionalization that leads to the effectiveness of the membrane for nanofiltration with a high cation selectivity of ~80% for 14 basic pH. Figure 4d shows the K + mobility increased and Cl − mobility decreased (K + /Cl − >> 1) at high pH due to the influence of polarized water molecules around their hydration shells. This is attributed to preferential co-ordination of H + and OH − groups on the K + and Cl − , respectively. 39 This corresponds to an increase in the OH − adsorption on the nanochannel's surface at higher pH (see Figure 4b); leading to a strong influence on the Cl − transport on the MoS 2 /SY surfaces (higher Cl − friction).  Table   S1).
In addition, the change in surface chemistry after functionalization is reflected in changes in the water contact angle (WCA) in air, by using ultra-pure deionized water droplets placed on the surface of the MoS 2 membranes (see Figure S10). The WCA of the MoS 2 membrane after functionalization (~68°) was much lower than a fresh, pristine MoS 2 (~85°), i.e., the dye functionalized MoS 2 membrane is more hydrophilic. This is attributed to the dye functionalization altering the surface chemistry and roughness of the MoS 2 . This change may result from the increase in dye functionalization on the Mo-edge of MoS 2 (see HAADF-STEM in Figure S5) due to the high number of defects created by the bath sonication process, which is evident by the presence of Mo−N bonding determined by Raman and X-ray photoelectron spectroscopy (XPS) as shown in Figure S11

Agarose salt bridge preparation
Firstly, silver wire (99.99 % purity, 0.35 mm diameter, Goodfellow Cambridge Limited: AG005145/1) was used to form a spring electrode to obtain a high surface area for chloride deposition. A silver wire spring was cleaned by immersing in 0.1 M HNO3 solutions for a few seconds to remove any contamination from the surface and then rinsed with deionized water. Secondly, silver/silver chloride (Ag/AgCl) electrodes were prepared in 0.1 M HCl solution by applying the potential of 0.5, 1.5, and 2.5 V for an hour with a Pt flag used as the counter electrode. Then, the prepared Ag/AgCl electrodes were cleaned by rinsing with deionized water to remove any residue. 1 Finally, the stability of the as-prepared Ag/AgCl reference electrodes were tested as working and reference electrodes in 0.1 M KCl solution by applying current of 0 A for 1 hour. The stable Ag/AgCl reference electrodes showed a potential difference around −0.8 mV (ideally should be less than ±1.0 mV). 3 To maintain a stable electrode potential during the experiment, Ag/AgCl reference electrodes were prepared with agarose salt bridges inside glass luggin capillaries. Briefly, agarose (3%) was dissolved in 3 M KCl solution by heating in a water bath at 75-80 °C under constant stirring. The tip of luggin capillary was immersed into the hot agarose gel which allowed the gel to fill the luggin tube by capillary force and then filled the rest by glass pipette. The Ag/AgCl electrode was immediately immersed in the gel leaving 3-5 mm between the luggin tip and the end of the electrode. The agarose salt bridge was left for an hour to allow the gel to set completely. 2 Then, the top of the luggin tubes was sealed with epoxy glue. The resultant Ag/AgCl agarose salt bridges were kept in 3M KCl solution until use to avoid gel getting dried. The Ag/AgCl salt bridges were used in a custom-made Hbeaker cell for the drift-diffusion experiment to eliminate the potential occurring from the redox reactions on the electrodes under different salt concentration gradients.

Cross-sectional STEM characterization methods
The cross-sectional scanning transmission electron microscope (STEM) images were carried out using an FEI Titan 80-200 ChemiSTEM equipped with probe-side aberration correction and an X-FEG electron source, which was used for the aberration-corrected scanning transmission electron microscope. STEM images were recorded using an acceleration voltage of 200 kV, a convergence angle of 21 mrad, and beam currents of 90 pA for imaging.

Dye functionalization
To control the exposed membrane area during the functionalization process, the MoS2 membrane was assembled between polyethylene terephthalate (PET) sheets with epoxy resin to give an exposed area of 0.26 cm 2 . Importantly, the MoS2 sandwich membranes were first checked by optical microscopy to ensure there was no obvious physical damage (e.g. cracking, defects, and delamination) as shown in Figure 2a

SEM and EDX characterizations
Top-down and cross-sectional SEM images of the MoS2/SY membranes, obtained using a FEI Quanta 650 FEG ESEM are shown in Figure S3

STEM characterization
High angle annular dark field scanning transmission electron microscope (HAADF-STEM) imaging was performed in near ultra-high vacuum (<1 × 10 −9 Torr) using an aberration corrected Nion UltraSTEM 100 operating at 60 kV and fitted with a cold field emission gun. The probe convergence angle was 30 mrad with a beam current of ~20 pA. The

X-ray diffraction analysis
Powder X-ray Diffraction (PXRD) patterns of the MoS2 membranes were obtained using a PANalytical X'pert X-ray diffractometer. The patterns were recorded in the range 2θ = 5-70°, with a step size of 0.017° with a scan step time of 66 s, which used a Cu-Kα radiation source (0.154 nm wavelength) operating at 40 kV and 30 mA. The (002) peak positions of MoS2 were corrected using the PVDF peak at 2θ of 20.17° as an internal reference peak. Figure S6 shows PXRD patterns of pristine and dye functionalized MoS2 (MoS2/SY) at the (002) peak supported on PVDF membranes. It is clear that the diffraction peak due to the (002) peak of MoS2 after exfoliation is notably broader, compared to the bulk material, indicating the distribution in the number of layers and flake size from the ultrasonication process. 9,10 The PXRD pattern of MoS2/SY exhibited small change in the (002) peak position indicating that there was no significant swelling as the membranes have been exposed in aqueous solution for several months. 6 Figure S7 shows the I-V characteristics of different valence ions (KCl, BaCl2, and AlCl3) of pristine exfoliated MoS2 membrane (~3 µm thick), which were measured at a constant concentration ratio (10 mM/100 mM). It was shown that the zero-current potential significantly decreased to a more negative potential with increasing hydrated cation radii from K + to Ba 2+ , but slightly decreased for Al 3+ . This is because less charge and size selective ion sieving for trivalent cations through MoS2/pristine when compared to MoS2/SY as shown in Figure 2 in the main text.

Ion transport through a pristine MoS2 membrane
By using this equation, we can calculate the conductance contributing from solely MoS2 membrane as following: To gain more information about individual ion mobility, the ion conductivity (σ MoS 2 ) can be calculated from the measured conductance (G MoS 2 ) as following: where A MoS 2 is the exposed MoS2 membrane area that perpendicular to the direction of the electrical current and l total is the total length of each nanochannel within the laminar MoS2 membranes.
To estimate the total length of each effective nanochannel (l total ), we used the possible permeation path for ions through laminar stacked MoS2 membranes as suggested by Nair and Geim. 15 Schematic S1 shows the laminar stacking structure of a MoS2 membrane which was separated by the distance between individual MoS2 sheets (nanochannel height, d). The nanochannel height was estimated in the range of ~6-13 Å (mostly 10 Å) measured by the cross-sectional STEM images, as shown in Figure 1e in the main text. This estimated channel 15 height is slightly larger than the free spacing between interlayer MoS2 distances (~6 Å) after deducting the thickness of a single MoS2 sheet as reported by Wang and Mi. 16 The possible path of ion transport through a MoS2 membrane of thickness (h) can be estimated by a number of ion turns (N) as following: where each turn (N) is equivalent to the length of the nanocapillary channel (L) which is measured around 200-300 nm by SEM and TEM images (see Figure S3-5). According to this estimation, each MoS2 crystallites were assumed to be the same dimensions (length, width, and thickness). Therefore, the total length (ltotal) of each effective nanochannel can be estimated by the following equation: For our typical 3 µm thick MoS2 membranes, ltotal can be estimated approximately 0.6 mm for permeation path of ions through laminar MoS2 membranes at which channel height and nanosheet length are 10 Å and 200 nm, respectively.

Ionic conductivity measurements
where µi is the ion mobility of ion i as reported in Figure 2b and F is Faraday's constant.
b The molar Cl − conductivities were calculated using the Cl − mobility of KCl solutions.

Ion transport through GO-based membranes
To compare ion sieving performance with other laminar 2D material-based membranes, we also prepared GO membranes on a PVDF support with a comparable thickness to the MoS2 membranes (~3 µm thick). The GO dispersion used herein was provided with the preparation method as reported in Wang and Dryfe. 18 Figure S8a shows the I-V characteristics of KCl measured at a concentration gradient (10 mM/100 mM) of a GO membrane under the same testing conditions used to study the MoS2 membranes. It was evident that the zero current potential significantly increased to a more positive potential of +31.3 mV which exhibited a high K + /Cl − mobility ratio ca. 5, calculated by the GHK equation as explained in the main text. The mobility ratio from our prepared GO membrane was closed to the literature value reported by Hong et al. 19 with the percentage difference of <10 % as shown in Figure S8b.
Moreover, the individual mobility between K + and Cl − can be also estimated using the same method for the MoS2 membranes. The permeation path for ions through the GO membrane was estimated to be ca. 1.5 mm at which the length of nanocapillary GO channel (L) and d-spacing (d) are 1 µm and 20 Å, respectively. The d-spacing used here was taken from the literature value when the membrane is immersed in water about an hour. 16 Therefore, the individual K + and Cl − mobility (•○ symbols) can be calculated as shown in Figure S8b.
To compare the estimated ion mobility of our prepared GO membrane with the literature value, the ion permeability (Pi) reported by Hong et al. 19 typically related to the diffusion coefficient (Di) and the solubility of the ion transport through the membranes as defined by molar flux density (permeation rate) and Fick's first law, as following equation: According to the Nernst-Einstein relation, the ion mobility is related to the diffusion coefficient as expressed following 17, 20 By combining Eq. S7 and S8, the ion mobility can be related to permeability as the following equation: where βi is the partition coefficient of an ion within the membrane. According to our results for ion mobility within the prepared GO membranes, βi can be assumed to be ca. 10 to evaluate K + and Cl − mobility (▲∆ symbols) within the previously reported GO membrane. 19 In the case of the Ti3C2Tx (MXene) membrane, as shown in Figure 2b, the K + mobility was also estimated from the permeation rate using equation S9 with the ltotal and βi were ca. 1.5 mm and 10, respectively. The βi of MXene membrane was assumed to be ca. 10 due to the equivalent ion permeability within Ti3C2Tx and GO membranes as reported by Ren et al. 21 To confirm the stability of GO membranes in aqueous media, Figure S8c shows a fresh dried GO membrane with a PVDF support prepared by pressure filtration at around 3 µm thick. It was clearly seen that the membrane exhibited significant instability from swelling after being immersed in deionized water for around 1 hour. This is in agreement with the previous literature using GO membranes for water purification. 16 Figure S10, recorded one second after placement of the water droplet.
The water droplets for all membranes were controlled to a volume of ~6 µL, inside a high humidity vessel to control water droplet evaporation. The WCAs were recorded for 10 minutes using a camera to capture the droplet images at 1.9 FPS. The WCAs were calculated using the Young-Laplace equation. 26

Raman analysis of MoS2 membranes
Raman spectra were measured using a Renishaw inVia microscope with a 532 nm (2.33 eV) laser, incident perpendicular to the membranes. The laser power was 1 mW with a 50× objective lens, and a grating of 1800 l/mm. Spectra were obtained between 200-1700 cm −1 and averaged over 3 accumulations. Figure S11 Figure S11. Raman spectra of a pristine exfoliated MoS2 and the MoS2/SY membranes compared to SY (raw powder). Inset shows two possible tautomeric SY structures.

XPS analysis of MoS2 membranes
X-ray photoelectron spectroscopy (XPS) analysis was performed with a Kratos Axis Ultra spectrometer, with excitation from a focused monochromated Al Kα source (1486.6 eV) and using an electron flood gun for charge neutralization. All XPS spectra were calibrated using adventitious carbon (C 1s) at 284.8 eV. Peak fitting used a nonlinear Shirley-type background (70% Gaussian and 30% Lorentzian line shapes).
The XPS analysis is shown in Figure S12. The Mo 3d spectrum of the pristine exfoliated MoS2 membrane consists of two main component peaks at around 230 and 233 eV assigned to Mo 4+ 3d5/2 and Mo 4+ 3d3/2 for 2H-phase MoS2, respectively ( Figure S12a). The Mo 3d peaks are shifted to lower binding energies by ~0.6 eV after functionalization which is similar to the S 2p regions of the spectra. This is due to charge transfer of new bound systems for Mo atoms. The peaks assigned at ~395 and ~413 eV correspond to Mo 3p3/2 and Mo 3p1/2, respectively. The N 1s peak at ~400 eV shown in Figure S12b for MoS2/SY can also be confirmed as strongly chemisorbed nitrogen from azo/hydrazone species on the metallic molybdenum atoms. [32][33][34] This was also supported by Becue et al. 33

Comparison between inorganic and organic cations
Furthermore, mobility measurements with the functionalized MoS2 membrane (MoS2/SY) were also carried out with organic cations, specifically tetramethyl-, tetraethyl-, and tetrapropylammonium ions (present as their chloride salts TMA-Cl, TEA-Cl, and TPrA-Cl) to make a comparison with the inorganic cations (NaCl, BaCl2, and CeCl3), which have hydrated radii very close to the radii of the non-hydrated organic ions. [35][36][37] The I-V characteristics for both types of salt were measured with a constant concentration ratio (10 mM/100 mM) as shown in Figure S13a. Figure S13b shows the ion mobility for the inorganic (• symbol) and organic cations (▲ symbol) with their chloride counter ions (○∆ symbols).
With increasing cation radius, the mobility of the hydrated and non-hydrated cations decreased in the similar values. The mobility of a small organic cation (TMA + ) with minor water solvation shell was similar to Na + , whereas others (TEA + /TPrA + ), which are believed to have no hydration shells, 36 exhibited ion mobility closed to that of inorganic cations of comparable sizes. This indicated that that size exclusion predominantly affected on transport across the membrane rather than the influence of charge density for di-and trivalent ions (Ba 2+ /Ce 3+ ).