Solar-driven abnormal evaporation of nanoconfined water

Abstract

Intrinsic water evaporation demands a high energy input, which limits the efficacy of conventional interfacial solar evaporators. Here, we propose a nanoconfinement strategy altering inherent properties of water for solar-driven water evaporation using a highly uniform composite of vertically aligned Janus carbon nanotubes (CNTs). The water evaporation from the CNT shows the unexpected diameter-dependent evaporation rate, increasing abnormally with decreasing nanochannel diameter. The evaporation rate of CNT₁₀@AAO evaporator thermodynamically exceeds the theoretical limit (1.47 kg m⁻² hour⁻¹ under one sun). A hybrid experimental, theoretical, and molecular simulation approach provided fundamental evidence of different nanoconfined water properties. The decreased number of H-bonds and lower interaction energy barrier of water molecules within CNT and formed water clusters may be one of the reasons for the less evaporative energy activating rapid nanoconfined water vaporization.

A nanoconfinement strategy alters intrinsic water properties for the ultrafast evaporation.

INTRODUCTION

Global issues on water scarcity, clean energy, and environmental pollution call for advanced technological solutions. Solar-thermal water evaporation is viewed as a promising eco-friendly technique to concurrently harvest clean and abundant solar energy and to alleviate freshwater scarcity. Traditional thermal- and membrane-based evaporation methods consume too much heat or electricity, respectively, and require complex infrastructure. Pioneering explorations have been made on achieving efficient solar-driven water evaporation with increased vapor generation rates and energy efficiencies by improving photothermal-conversion efficiency (1), architectures for water regulation (2, 3), and heat management (4, 5). However, intrinsic water evaporation demands a high energy input of >40 kJ mol⁻¹, which profoundly limits the solar vapor generation (SVG) rate (≤1.47 kg m⁻² hour⁻¹) of conventional designs under natural sunlight (solar flux, ≤1 kW m⁻²) (5, 6). Thus, an improved fundamental understanding of water evaporation mechanisms and kinetics would allow rational manipulation of the evaporation process and further develop evaporation-based technologies for a range of applications such as desalination and brine concentration.

For the past 130 years, the evaporation kinetics at the bulk liquid-vapor interface has been described by the classical Hertz-Knudsen (H-K) equation, where the local evaporation molecular flux j_ev is described as (7, 8)

$$j_{\text{ev}}={\left(\frac{M}{2\mathrm{\pi }R}\right)}^{1/2}({\mathrm{\sigma }}_{\mathrm{e}}\frac{P_{\mathrm{v},\text{eq}}}{\sqrt{T_{\mathrm{i}}}}-{\mathrm{\sigma }}_{\mathrm{c}}\frac{P_{\mathrm{v}}}{\sqrt{T_{\mathrm{v}}}})$$

In Eq. 1, two empirical parameters (the evaporation σ_e and condensation σ_c coefficients) govern the evaporation kinetics (9), and both are dependent on interfacial curvature, temperature, pressure, and in particular the molecule structure and self-interaction energy. The reported values of the water evaporation coefficient had a range of nearly three orders of magnitude due to variable surface structures and coatings (10), and water evaporation mechanism is enabled by the concerted ultrafast hydrogen-bond dynamics of interfacial water (11). Thus, alterations to water structure can substantially affect water-water interaction energies and evaporation kinetics, such as observed with water in hydrogels and spatial confinement (12–15).

Water molecules confined in low-dimensional spaces, e.g., one-dimensional (1D), exhibit novel structural and dynamic properties as compared to bulk water, such as altered viscosity, solvation, wetting, and phase, due to variations in water’s hydrogen-bond network (13, 16–18). For example, confined water has an extremely low dielectric constant of 2 as compared with 80 for bulk water as a result of a highly ordered water structure (16, 19). Similarly, anomalous confined water phase and state were observed due to reduced H-bonding capability and proton conductance (20, 21). Water confinement effects in 1D carbon nanotubes (CNTs) have been reported to be a function of CNT internal diameter in regard to the melting-freezing points (22) likely due to changes in water entropy, enthalpy, and free energy as a function of the degree of confinement (23). In general, confined water molecules exhibit unusual state and structure, in particular the H-bonding network structure that shifts away from the bulk tetrahedral geometry, which has a substantial impact on the water phase transition behaviors such as the enthalpy of vaporization (12, 23–26).

Recently, there have been both theoretical simulations and experimental evidence of capillary evaporation during water nanoconfinement, such as between graphene sheets, with reported evaporation fluxes surpassing the H-K limit by an order of magnitude (27, 28). CNTs are an ideal 1D confinement structure for the solar-driven nanoconfined evaporation due to their unique combination of physical and chemical properties related to solar-thermal conversion and 1D fluid flow (29–31). For instance, owing to the unique nanoconfinement effect and frictionless characteristics of CNT, the ultrafast gas/liquid transport behaviors in CNT break classical kinetic theory of gases or hydrodynamic flow Hagen-Poiseuille equation (32, 33). This is also the theoretical basis for choosing CNT as the solar evaporator. However, the underlying mechanisms remain poorly understood and quantitative elucidation of the confined evaporation process remains elusive (34), particularly for the 1D confinement-dependent evaporation.

Here, via a facile template-assisted chemical vapor deposition (CVD) method and selective plasma modification, we produced a highly uniform, Janus hydrophobicity, vertically aligned CNT membrane using an anodic aluminum oxide (AAO) template. The CNT array was composed of uninterrupted uniform diameter nanochannels featuring a selective 1D water transport pathway and a precise solid-liquid-vapor phase exchange interface (Fig. 1). The water evaporation from the CNT nanochannel/membrane was systematically evaluated by experiment, theory, and modeling. The overall nanochannel evaporation mechanism was accurately predicted using theoretical models, and the underlying molecular mechanism of enhanced confinement evaporation was elucidated by experiments and MD simulations. The CNT evaporator was further evaluated for SVG with high-salinity brines and oily wastewaters, and strong evaporative separation performance was observed in all cases due to the Janus hydrophobicity alleviating both salt scaling and organic fouling issues detrimental to existing water evaporator technologies.

Fig. 1. Design and characterizations of CNT array membranes.

(A) 3D schematic of the CNT@AAO array membrane assembled by the CVD process. (B to D) SEM and TEM (top right) images of the top surface and edge of CNT@AAO membranes with different diameters. (E) Water CA of the top and bottom surfaces of CNT membranes. (F) Comparison of the traditional bulk evaporation in hydrophilic and hydrophobic tortuous or straight channels, and nanoconfinement evaporation in a Janus hydrophobicity and charged CNT. (G) Photograph of the pristine AAO and CNT@AAO membranes. (H) Light absorption property of the CNT absorbers and standard air mass 1.5 G solar spectrum.

RESULTS

Design and characterization of Janus CNT array membranes

Here, AAO was selected as the nanoporous template due to two key features: (i) the AAO templates have a prominent highly ordered and straight nanochannel array and (ii) the alumina can catalyze the transformation of hydrocarbons into CNT within the internal AAO nanochannels, avoiding metal-based catalysts that could block the internal CNT pore (note S1) (35). The CNT@AAO membrane was fabricated through a template-assisted CVD process (fig. S2). To explore the pore size effects of confinement on evaporation kinetics, four CNT nanochannels with diameters of ~250 nm (the maximum AAO pore size), ~90 nm, ~20 nm, and <10 nm (the minimum AAO pore size) were prepared and denoted as CNT₂₅₀@AAO, CNT₉₀@AAO, CNT₂₀@AAO, and CNT₁₀@AAO, respectively. The characterization and detailed parameters of AAO substrates are displayed in figs. S3 and S4 and table S1.

A representative 3D schematic of a CNT@AAO array is presented in Fig. 1A consisting of a perforated carbon film, CNT, and a nanoporous AAO template. A typical high-resolution scanning electron microscopy (SEM) image of the top surface and edge of a CNT@AAO is presented in Fig. 1 (B to D), where the CNT can be clearly seen to protrude from pores at the top of the membrane edge corresponding to pores of the AAO substrates. The cross section of the CNT₁₀@AAO looks similar to a typical CNT array, and the individual CNT has a nanotube structure with a uniform diameter and open ends (fig. S5). Transmission electron microscopy (TEM) images reveal that CNT₁₀ has a uniform and smooth nanochannel (<10 nm) and wall (about 5 nm thickness), further indicating successful CNT formation. Size exclusion experiments supported TEM measurements that the pore size of the CNT₁₀ nanochannels is less than 10 nm by filtering gold nanoparticles (~95% rejection for 10-nm Au particles; note S10), and the transport in CNT₁₀ occurs exclusively through the inner pores of the CNT spanning the membrane. Similarly, CNT₉₀ and CNT₂₅₀ have a uniform diameter (about 90 and 250 nm, respectively), open ends, and smooth nanochannel walls (fig. S6). These consistent features are critical for water transport through the nanochannels and eventual evaporation at the nanopore-air interface.

The pristine CNT array is hydrophobic in nature with a water contact angle (CA) of 110^o due to the nonpolar conjugated π-orbital structure, which can hinder water uptake into the CNT nanochannels. To enable water uptake, asymmetric oxygen plasma functionalization was applied to selectively modify one side of the CNT membrane to a greater extent than the other with hydrophilic oxy-functional groups as supported by x-ray photoelectron spectroscopy (XPS) analysis (fig. S7), which enables the hydrophilic-hydrophobic transition within the length CNT (fig. S8). The plasma oxidized side of the CNT membrane was now charged and hydrophilic (CA = 20^o) (fig. S9 and Fig. 1E). Thus, unlike traditional evaporation channels designed for capillary transport or vapor diffusion, the CNT nanochannels here were designed with Janus hydrophobicity/hydrophilicity for a directional water transport pathway and a three-phase hydrophobic interface, ensuring rapid water into, through, and evaporation from the nanochannel, respectively (Fig. 1F). Additionally, the curvature of formed droplets in nanochannels can also affect the state of water molecules for enhanced evaporation.

In addition, the black CNT arrays displayed excellent light absorption, with a relatively constant optical absorptivity of 0.98 to 0.99 across a wide spectral range (250 to 2500 nm), similar to a blackbody (Fig. 1, G and H) (36). The overall solar energy absorption approaches 99% (calculation details in note S3), which results in a high solar-thermal conversion efficiency (fig. S10). Overall, the engineered CNT array membranes were rationally designed and fabricated to have straight nanochannel geometries, well-defined and narrow pore size distributions, efficient broadband light absorption, and asymmetric surface functionalities across the length of the nanochannel to yield a high-performance solar-thermal water evaporation membrane.

Exceptional evaporation in Janus nanochannel evaporators

The biomimetic solar evaporator architecture inspired by natural transpiration systems was designed with a sufficient water supply (Fig. 2A and fig. S11). Infrared imaging was used to monitor the temperature of the evaporator over time under one-sun irradiation (Fig. 2B), where the center temperature of CNT membrane increased from 19.7°C to 32.8°C in 5 min and stabilized around 41.2°C after 30 min. There is no obvious heat loss in other parts of the system. COMSOL Multiphysics was used to simulate the solar evaporation process (Fig. 2C and note S14), and results indicated that both capillary water wicking and suppressed parasitic heat dissipation are occurring simultaneously for the CNT solar evaporator. During the SVG process, the top surface of the CNT evaporator remained dry due to Janus hydrophobicity, while that of conventional CNT filter was wetted, further verifying that the evaporation occurs from within the nanochannel rather than from the bulk surface.

Fig. 2. Solar-driven water evaporation in and through CNT nanochannels.

(A) Scheme of the 2D solar evaporator used in the water evaporation test, consisting of the polystyrene foam for thermal insulator, cellulose wick and layer for water supply (200-μm water layer), and Janus CNT membrane (50 μm) for solar absorption and evaporation. (B) Infrared images of the CNT₁₀ evaporator showing the temperature distribution with an irradiation time of 0, 5, 15, and 30 min under one-sun (1 kW m⁻²) radiation. Inset is a photograph of the CNT₁₀ evaporator. (C) COMSOL stimulation of the model (left) and temperature (right) distribution of the evaporator, including the CNT@AAO membrane (AAO, gray square block with 50-μm thickness; CNT channels, black narrow bars), water supply layer (blue quadrate, 200 μm), and polystyrene insulation (milk-white background). (D) Water mass loss of CNT evaporators and pure water as the control under one-sun illumination (1 kW m⁻²). (E) Illustration of overall evaporation processes in nanochannel, including water capillary supply, evaporation, and vapor diffusion steps.

Figure 2D shows the measured evaporation rate of CNT evaporators under one sun, where the abnormal experimental phenomenon is presented (fig. S12A). On the one hand, there is the unexpected diameter-dependent water evaporation rate. Namely, the evaporation rate of CNT membranes with smaller pore size and porosity is actually higher than that of CNT membranes with larger pore size and porosity (table S2). The phenomenon that evaporation rate is strongly negatively correlated with the pore diameter is the opposite of that expected for traditional gas/fluid transport, and the underlying mechanism for this exception to the rule is worth further exploration (8, 37). On the other hand, the evaporation performance of the CNT₁₀@AAO evaporator not only outperforms all reported 2D evaporators but also thermodynamically exceeds theoretical upper limit of the evaporation rate (1.47 kg m⁻² hour⁻¹ under 1 kW m⁻² solar illumination using the enthalpy of vaporization at 20°C; note S9 and fig. S12) (28, 38, 39).

To improve our understanding of the enhanced evaporation rate through the Janus CNT nanochannel, the overall evaporation process was divided into three fundamental three mass transport steps: (1) liquid water transport through the hydrophilic portion of the nanochannels to the hydrophobic solid-liquid-vapor interface, (2) water evaporation at the three-phase interface, and (3) water vapor diffusion through the hydrophobic section of the nanochannels (Fig. 2E). These three steps were investigated in detail to evaluate the relative contribution to the overall evaporation rate enhancement.

In regard to step 1, liquid water transport in the nanochannel can be modeled using capillary limit (Q_c), which is determined by the capillary pressure (∆P) and the total viscous resistance (R) associated with the structure (detailed theory can be found in note S4) (8, 27):

$$Q_{\mathrm{c}}=\frac{\mathrm{\epsilon }\mathrm{\tau }}{A}\frac{∆P}{R}$$ (2)

The capillary limits of the hydrophilic region of the CNT nanochannels are all greater than the corresponding measured evaporation rates. For example, the calculated capillary limit of liquid water transport for CNT₁₀ is 79.1 kg m⁻² hour⁻¹. This result indicates that the liquid water transport in the nanochannel does not limit the overall evaporation rate, in agreement with previous reports (27, 37, 40).

Water vapor diffusion out of the nanochannel was modeled using the Knudsen diffusion equation, the classical kinetic theory of gases (Knudsen diffusion and molecular diffusion were both used for the 90- and 250-nm nanochannels), in which the mean free path (λ) of water molecule is larger than the nanochannel dimensions (d) and transport enters the molecular flow regime (note S5) (33). The vapor transport rate through CNT nanochannels will not cause the resultant evaporation rate to exceed the theoretical limit. That is to say, the water evaporation enhancement at the nanochannel solid-liquid-vapor interface is the main factor.

The evaporation kinetics at the liquid-vapor interface can be described by the semi-empirical H-K, as shown in Eq. 1, which can be further simplified into Eq. 3 (7, 8):

$$j_{\text{ev}}=\mathrm{\sigma }∆P\sqrt{\frac{M}{2\mathrm{\pi } \mathit{RT}}}$$ (3)

where M and R represent the molecule molar weight and gas constant. T is temperature (the temperatures of liquid T_i and vapor T_v are considered equal), and ∆P is the difference between the equilibrium vapor pressure of liquid at interface (P_v,eq) and partial pressure of the vapor in gas phase (P_v). The experimental conditions for all CNT@AAO evaporators were the same. If we set $A = \mathrm{∆}P\sqrt{\frac{M}{2\mathrm{\pi } \mathit{RT}}}$ (constant), then j_ev = σA. Hence, the coefficient σ directly determines the evaporation flux, j_ev. At thermodynamic equilibrium, the evaporation coefficient σ_e and the condensation coefficient σ_c are similar to emissivity and absorptivity, respectively, i.e., σ_e = σ_c = σ. Assuming A is constant and simplifying Eq. 3, the H-K equation indicates that a large increase in evaporation velocity when the nanochannel is decreased to 10 nm would be attributed to the nanochannel evaporation coefficient being much larger as compared to much larger diameter nanochannels. However, since the evaporation coefficient is typically determined empirically, minimal insight is gained on the underlying mechanisms for this observed increase; thus, evaporation fluxes still cannot be quantitatively determined from first principles.

Mechanism of the nanoconfined water evaporation

Fundamentally, water’s strong hydrogen-bond network (H-bonding) determines most of its physical, chemical, and biological properties such as viscosity, solvation, wetting, and phase transition thermodynamics (13, 16–18). Furthermore, water molecules exhibit abnormal physicochemical properties and behaviors under nanoconfinement, where the H-bonding network structure can be substantially altered, which would likely affect the evaporation coefficient and kinetics (11, 12, 23–26).

The structure of water molecules confined in a single CNT was examined by high-angle angular dark field–scanning transmission electron microscopy (HAADF-STEM) and electron energy loss spectroscopy (EELS) (notes S6, S7, and S10) (41). EELS detected the water oxygen peak (537 eV) of water confined by CNT₁₀ as shown in fig. S15. The H-bonding structure of confined water within CNT₁₀ was radially examined in three sections (sections are overlaid on a TEM in Fig. 3A). In the center of the CNT (3), the EELS displays a symmetric saturated bulk water H-bonding peak with a post-edge peak at 543 eV. In contrast, the water nearest the CNT inner wall (1 and 5) displays a dominant pre-edge at 535 eV and no post-edge peak (543 eV), indicating asymmetric H-bonding with a metastable structure. The EELS data indicate that water near the CNT inner wall has a lesser degree of H-bonding than that at the center forming water cluster. The strong nanoconfinement effects are mainly observed within the first ~2 nm of the CNT inner wall, which would be ~64% of the volume of a 10-nm CNT, but only ~36% of a 20-nm CNT, <9% of the volume of a 90-nm CNT, and <4% of the volume of a 250-nm CNT. This is in agreement with the results of Fig. 2D where CNT₁₀ had a greater evaporation rate than CNT₂₅₀ and CNT₉₀, with CNT₂₀ having an intermediate rate to the two extrema. The less energy would be required to break H-bonds of confined water for evaporation (11), especially near the internal CNT surface, and the interfacial water may envelop the central water in the form of water-cluster evaporation from CNT, resulting in reducing vaporization energy (12, 42). The measured Li⁺ concentration in the evaporated water verified that water evaporation in CNT with smaller pore size is more likely to be in water clusters (note S11; water-cluster evaporation from CNT) (12, 39).

Fig. 3. Mechanism of the confined water evaporation in nanochannels.

(A) Nanoconfined water molecule structure along with the CNT radial divided into five parts (HAADF-STEM image) and EELS analysis corresponding image area. (B and C) Fitting curves of O-H stretching modes for bulk water and confined water in CNT₁₀. (D) IW/FW ratios of bulk water and confined water in CNT₂₅₀ and CNT₁₀ membranes detected by in situ Raman. (E to G) MD simulations. (E) Schematics showing the water evaporation models at the bulk interface or within CNT nanochannel (d = 90.0 Å) at T = 313 K. (F) Average number of H-bond per water molecule and radial liquid density. (G) Total interaction energy profile of bulk water and confined water in CNT. Distance is modulated from the CNT center (zero point) to the CNT wall.

To further verify the abovementioned data, Raman spectra were completed to qualitatively examine the content of the free water (FW) with four H-bonds and intermediate water (IW) with weak or no H-bonds in CNT nanochannels. The evaporation of IW is much faster than that of FW owing to the weaker interaction of IW with surrounding water molecules (12, 43). As shown in Fig. 3 (B to D) (fig. S17), the IW/FW ratio increases from 0.29 of bulk water to 0.86 of CNT₁₀, indicating that nanochannel confinement increases the relative IW content, which has been observed to have a lesser latent heat than FW (i.e., reduced IW energy requirement for water vaporization) (12). A control experiment was designed to evaluate the vaporization enthalpy of nanoconfined water in CNT (fig. S18; theoretical analysis is in note S11) (44). The Gibbs free energy analysis indicates that water molecules evaporating from a nanochannel require larger free energy gradients than those evaporating from the bulk. Thus, water evaporation under nanoconfinement is more thermodynamically favorable than from the bulk (see note S12 for details).

To further elaborate the molecular mechanism behind the abnormal confinement evaporation, the average number of H-bond per water molecule radially across the CNT nanochannel was examined with molecular dynamics (MD) simulations using two evaporation models (Fig. 3E) (14, 45). Bulk water typically has ~3.7 H-bonds per water molecule. In contrast, the whole confined water molecule near the center of a CNT only had ~3.1 H-bonds and rapidly declines to <1 H-bond per water molecule within 1 nm of the CNT inner wall as shown in Fig. 3F. The MD simulations confirm previous experimental observations (TEM, EELS, and Raman) that the bulk water structure is lost within 1 to 2 nm of the CNT inner wall, resulting in weaker intermolecular water interactions and elevated evaporation velocity (46). The corresponding radial density profile suggests a coaxial structure of water molecule tubes nearest the CNT wall, where the density within the layers increases to 2.7 g cm⁻³ (bulk water = 1 g cm⁻³) (47). The corresponding radial total interaction energy of water molecules reveals that the confined water (−4.1 kcal/mol), except the first bound water layer near the wall driving water into CNT, is less thermodynamically stable than bulk water (−5.4 kcal/mol) (Fig. 3G) (48). Water evaporation velocity in CNT is faster than the bulk water (fig. S19), which corresponds to the reduction of number of H-bonds and interaction energy of water molecules from experiment and simulation. Additionally, the first-principles calculation further proves that the hydrophilic groups of the CNT wall strongly affect the first layer of water molecules (bound water), which drives the water into CNT. While starting from the second layer of water molecules, the absolute adsorption energy of water molecules rapidly decreases, especially for the second water layer, and gradually approaches the value of bulk water (fig. S20). This is because the attraction of hydrophilic functional groups greatly weakens, and this part water molecules are co-attracted by water-water bonding. This result further proves that hydrophilicity just drives water into CNT, and it is the nanoconfinement effect that reduces the interaction energy of whole confined water molecules (Fig. 3G), resulting in lower energy and easier evaporation than bulk water. The nanoconfinement dominated water molecules exceed 37.97% of the total water within a CNT₁₀ (see note S13 for detailed calculations). Overall, both the experimental results and theoretical simulations/calculations demonstrate the fundamental mechanism of superior water evaporative rates under nanoscale confinement. The reduced H-bonds, especially near the CNT internal surface, and lower interaction energy barrier of confined waters may be one of the reasons for facilitating nanoconfined water vaporization. Note that the mechanism of nanoconfined evaporation is much more in-depth than what we have discovered, and deeper and fundamental research is required (for instance, how to break H-bonds under nanoconfinement). The evaporation rate is expected to further increase in smaller nanochannels where the percentage of water molecules under nanoconfinement is increased (27, 49).

Scaling resistance and anti-fouling performance of Janus CNT evaporator

The highly uniform vertically aligned Janus CNT₁₀ evaporator was evaluated for solar water vapor generation with high concentration waste brine (10 wt % NaCl), and a randomly ordered CNT filter was used as a control (fig. S21). There is no visible salt accumulation on the CNT₁₀ evaporator after 6.0 hours under one-sun irradiation (Fig. 4A), and the water evaporation rate remained relatively constant. In contrast, the randomly oriented CNT filter surface gradually accumulated salt that eventually crystallized on the absorber’s surface. SEM-EDS (Energy Dispersive Spectrometer) mapping of Cl⁻ and Na⁺ reveals large salt grains on top of the absorber (Fig. 4B), which will increase albedo decreasing solar light absorption and block pathways for water evaporation, consequently degrading the evaporator performance. COMSOL simulations were used to elucidate the potential CNT₁₀ evaporator anti-fouling mechanisms. In the simulations, the high negative charge of the plasma-modified CNT membrane, in particular for the CNT wall, electrostatically repels the dissolved salt decreasing the adjacent salt concentration, especially avoiding salt crystallization on the CNT wall. This effect is further enhanced in the nano-sized channels as displayed in Fig. 4C. Obviously, a strongly bound pure water layer forms between the brine-membrane interface, which further indicates the obstruction of salt ions into the nanochannel. A double cell diffusion experiment further verified that the CNT₁₀@AAO membrane can hinder the diffusion of salt ions (fig. S22). Thus, the hydrophobic prevention and crystallization resistance on the CNT₁₀ wall by the electrostatic interaction contribute to the anti-scaling of the CNT₁₀ membrane.

Fig. 4. Salt resistance and anti-fouling performance of the multifunctional CNT evaporator.

(A) Macroscopic photographs and (B) microcosmic surface morphologies and EDS of the traditional CNT filter and Janus CNT₁₀ membranes under different irradiation time with 10% NaCl solution. The dashed lines refer to the salts that accumulate on membranes. (C) Negative potential distribution and ion concentration of the CNT₁₀ bottom layer simulated by COMSOL. (D) Schematic diagram of the anti-fouling performance and salt resistance of the CNT evaporator based on Janus hydrophobicity and Donnan effect.

The plasma-modified bottom surface of the Janus CNT membranes has an aqueous submersed oleophobicity with a 147^o oil-water CA (fig. S23). The strong aqueous oleophobicity suggests the Janus CNT membrane may also have anti-fouling performance in the presence of hydrophobic and oily chemicals and thus utility for water-oil separations. For example, the CNT evaporator–exposed surface remained dry and clean during the oily wastewater treatment, while the surface of a randomly oriented CNT filter was wetted and fouled as shown in fig. S24. The oily wastewater anti-fouling mechanism is similar to that for brine where the plasma-modified bottom surface strongly binds a layer of water that prevents nonaqueous species from approaching and entering the nanochannel (Fig. 4D). The Janus evaporator can maintain the long-term stability in real seawater (Huanghai Sea; fig. S25).

DISCUSSION

Here, a highly uniform vertically aligned Janus CNT nanochannel membrane was designed, constructed, and evaluated for solar-driven interfacial confinement evaporation. We experimentally determined that the evaporation increased with decreasing nanochannel diameter, and the evaporation rate of Janus CNT₁₀@AAO evaporator thermodynamically exceeds the theoretical limit. Experimental analysis and MD simulations of the molecular water structures indicate that, under nanoconfinement, the number of H-bonds and interaction energy of water molecules are decreased and water clusters are formed, which may result in water requiring a lower evaporation energy (enthalpy) to activate rapid solar-driven water vaporization. Along with ultrafast evaporation, the CNT solar evaporator was demonstrated to be effective and efficient for the treatment of high-salinity brines, oily wastewaters, and natural seawaters due to the Janus functionality, giving the membrane strong anti-scaling and anti-fouling properties. On the basis of the highly uniform Janus CNT membrane, this work provides fundamental molecular insights into the underlying mechanism of nanoconfined evaporation, which may only be the tip of the iceberg. Note that the mechanism of nanoconfined water evaporation is much more in-depth than what we have discovered, and deeper and fundamental research is needed. We hope that this work can spark researchers to have more inspirations and deeper researches, and may make a meaningful contribution to the related field.

MATERIALS AND METHODS

The materials used in the experiments, detailed methods, and theoretical analysis are described in the Supplementary Materials.

Fabrication of the 2D solar evaporation device

The CNT array membrane was synthesized by a template-assisted CVD process and selective plasma modification (see note S2 for details), as an absorber. A 2D solar evaporator was designed, in which a Janus CNT absorber is loaded on a mimetic transpiration system inspired by nature tree, where expanded thermal insulator (a polystyrene foam, thermal conductivity of ∼0.04 W/mK) was used as the soil to ensure much suppressed parasitic heat loss of the environment and bulk water, and placed on the water vessel. A hydrophilic cellulose thin layer (200 μm) and wick (diameter: 5 mm) acted as the root of the tree, which provided a 2D water layer and 1D water transport pathway, respectively, based on the capillary force and prevented conduction heat loss simultaneously. The microstructure of the cellulose channel has been characterized in fig. S11.

Steam generation and solar desalination experiments

The 2D Janus CNT evaporator was irradiated by a solar simulator (class AAA, Newport 94043A) with an optical filter for the standard air mass 1.5 G spectrum (6). The power density of the illumination intensity was measured at the level of the sample with the power meter (CEL-NP2000, CEAULIGHT). The temperature of the CNT surface was captured by the infrared camera (InfiSense, P2). The weight change from evaporation was monitored by electronic analytical scale (ME104/02, 0.1-mg accuracy), real-time recorded by a desktop computer (with RS 232 serial ports) once the temperature reached steady state (pre-heating for 30 min), and then used to determine the evaporation rate and efficiency of solar steam generation (fig. S26). All the experimental data of solar vapor generation under the irradiation are calibrated to dark-condition evaporation data.

A home-made double-slope solar still setup was used for steam condensation and water collection. A clean quartz plate was used as the roof to allow light to pass through. Under solar illumination, the steady steam will condense into water when it arrives at the cold chamber wall, and the condensed water automatically flows along the sidewall of the container into the condensation chamber. The concentration of the feed and collected solution was measured by an inductively coupled plasma spectrometer (iCAP7400, Thermo Fisher Scientific). For cycle performance of CNT evaporator, as the evaporation process reached steady state after 30 min, the solar desalination tests were sustained for 1 hour each day. The experiments were typically conducted at an ambient temperature of 26 ± 2°C, humidity of ~41 ± 5%, and minimal convective flow in air-conditioned and dehumidified room.

Characterization of H-bonding network structure of water molecule confined in CNT

In situ ultrahigh vacuum TEM would be the best method to observe and analyze the confined water molecule–level structure. The confined water molecule encapsulated in CNT by the graphene nanovessel strategy was observed by HAADF-STEM (FEI TF20 and Titan3 G2 60-300) and then analyzed by electron energy-loss spectroscopy (EELS) methods to map the water-molecule arrangement with nanoscale resolution (notes S6 and S7).

MD simulation methodology of water evaporation

Evaporation velocity profile, average number of H-bond per water molecule, radial liquid density, and interaction energy profile of bulk water and confined water in CNT were studied by MD simulation. Detailed methodology was depicted in note S13.

COMSOL Multiphysics simulation for the CNT evaporator

Temperature, potential, and ion concentration distributions of the CNT evaporator during the evaporation process were simulated by COMSOL Multiphysics. Detailed calculation method was shown in note S14.

Supplementary Materials

This PDF file includes:

Supplementary Text

Notes S1 to S15

Figs. S1 to S26

Tables S1 to S3

References