Report

Polyamide membranes with nanoscale Turing structures for water purification

See allHide authors and affiliations

Science  04 May 2018:
Vol. 360, Issue 6388, pp. 518-521
DOI: 10.1126/science.aar6308

Turing structures at the nanoscale

Turing structures arise when imbalances in diffusion rates make a stable steady-state system sensitive to small heterogeneous perturbations. For example, Turing patterns occur in chemical reactions when a fast-moving inhibitor controls the motion of a slower-moving activator. Tan et al. grew polyamide membranes by using interfacial polymerization, where the reactions occur at the interface between oil and water layers. The addition of polyvinyl alcohol to the aqueous phase reduced the diffusion of the monomer. This process generates membranes with more bumps, voids, and islands, which prove to be better for water desalination.

Science, this issue p. 518

Abstract

The emergence of Turing structures is of fundamental importance, and designing these structures and developing their applications have practical effects in chemistry and biology. We use a facile route based on interfacial polymerization to generate Turing-type polyamide membranes for water purification. Manipulation of shapes by control of reaction conditions enabled the creation of membranes with bubble or tube structures. These membranes exhibit excellent water-salt separation performance that surpasses the upper-bound line of traditional desalination membranes. Furthermore, we show the existence of high water permeability sites in the Turing structures, where water transport through the membranes is enhanced.

Alan Turing’s 1952 paper (1), “The chemical basis of morphogenesis,” theoretically analyzed how two chemical substances, activator and inhibitor (2) (Fig. 1A), can, under certain conditions, react and diffuse with each other to generate spatiotemporal stationary structures. Turing’s ideas have profoundly influenced theoretical understanding of pattern formation in chemical (3) and biological (4, 5) systems, but it was not until nearly 40 years after his paper was published that experimental evidence was obtained for the chlorite-iodide-malonic acid (CIMA) reaction (6, 7). About 10 years later, stationary Turing states were also observed in the Belousov-Zhabotinsky (BZ) reaction microemulsion consisting of reverse micelles (8). Recently, a variety of two- and three-dimensional stationary structures were studied in chemical (9, 10) and biological (1115) systems.

Fig. 1 Turing-type structures in interfacial polymerization.

(A) Schematic diagram of activator-inhibitor interaction in a reaction-diffusion process. Reactions leading to Turing structures rely on competing activation (red) and inhibition (blue) kinetic pathways. (B) Spatial representation of local activation and lateral inhibition. In two dimensions, Turing structures generally consist of spots or stripes. (C) Schematic illustration of interfacial polymerization Turing system. The inhibitor (TMC) is dissolved in the organic phase (top), and the activator (PZ) and the macromolecule (PVA) are dissolved in the aqueous phases (bottom). The membrane (PA) with nanoscale Turing structures forms on the porous support (PSU). (D and E) AFM topography images of the Turing-type PA membranes. Bright yellow and orange regions correspond to the formed solid-state nanoscale Turing structures. Initial concentrations for nanoscale spots (D) are [TMC] = 6 mM, [PZ] = 28 mM, and [PVA] = 12 mM, and for nanoscale stripes (E), [TMC] = 8 mM, [PZ] = 23 mM, and [PVA] = 32 mM. Scan area is 2 μm by 2 μm.

Turing structures typically emerge in reaction-diffusion processes far from thermodynamic equilibrium (1), in which the diffusion coefficient of the inhibitor must be larger than that of the activator, resulting in the “local activation and lateral inhibition” phenomenon (Fig. 1B) that underlies diffusion-driven instability (2). However, this condition is not easily satisfied in homogeneous solutions, for most chemical reactions involve small molecules with similar or inappropriately differing diffusion coefficients. In the classic Turing systems, two main approaches have been developed to selectively control the effective diffusion coefficients of reactants: (i) Introduce a macromolecule that reversibly binds the activator, like starch or polyvinyl alcohol (PVA) in the CIMA reaction, and (ii) use a heterogeneous finely dispersed multiphase reaction system in which the activator resides in a low-mobility phase, such as when polar BZ reagents are confined within nanosized aqueous droplets (610). On the basis of theoretical analyses and experimental observations, we successfully applied these chemical and physical approaches to aqueous-organic interfacial polymerization and developed a facile route to generate nanoscale Turing structures with high water permeability under ambient conditions.

Interfacial polymerization is a reaction-diffusion process far from thermodynamic equilibrium (16). It is based on the Schotten-Baumann reaction, in which the irreversible polymerization of two fast-reacting multifunctional monomers occurs near the interface of two immiscible phases of a heterogeneous liquid system (17, 18). This technique has been used to prepare reverse osmosis and nanofiltration membranes for large-scale and low-cost water purification applications (19, 20). In a typical membrane synthesis (fig. S1), organic amines are dissolved in water while acyl chlorides are dissolved in an organic solvent, and a very thin insoluble polyamide (PA) membrane forms on top of a porous support (figs. S2 and S3). In our experiment, piperazine (PZ) is the activator, and trimesoyl chloride (TMC) is the inhibitor (Fig. 1C). The reaction is initiated when the top surface of a porous polysulfone (PSU) support containing an aqueous solution of the activator comes in contact with an organic solution of the inhibitor. Because the acyl chloride has very little solubility in water, the polymerization occurs predominantly on the organic side of the interface. Initially, the activator reacts with the locally available inhibitor in the reaction zone, later it diffuses to penetrate more deeply into the reaction zone, and finally, a cross-linked PA membrane forms across the region of pore openings of the PSU support (figs. S4 to S11 and table S1). This PA membrane formed by a conventional interfacial polymerization reaction is not of Turing type, for there are not appropriate differences between the diffusion coefficients of the activator and the inhibitor. During the reaction, the aqueous solution of the activator is confined within surface nanometer-sized pores of the PSU support, where physical obstruction blocks dispersed aqueous-phase movement and slows the activator transport. The diffusion coefficient of the organic molecules is around 10−5 cm2 s–1, whereas the diffusion of the dispersed aqueous phases in the organic phase can be as low as 10−6 cm2 s–1 (fig. S12). When a certain amount of macromolecule, PVA, was added to the aqueous solution, it interacted with the activator via hydrogen bonding and increased solution viscosity, further reducing the diffusion rate of the activator (fig. S13). Through the synergetic effects of the physical obstruction and chemical interaction, the systems meet appropriate differences in the diffusion coefficients of the activator and inhibitor (21, 22), leading to a diffusion-driven instability and generating nanoscale spotted (Fig. 1D) and striped (Fig. 1E) Turing structures.

Atomic force microscopy (AFM) measurements (Table 1) show that the surfaces of membranes with the nanoscale spotted (TS-I) and striped (TS-II) Turing structures are relatively rough and heterogeneous. The measured average root mean square roughnesses were 22 and 32 nm, respectively, which is quite different from that of traditional semiaromatic PA membrane (figs. S14 and S15 and table S2) with a relatively smooth and homogeneous surface (23). The spotted and striped structures have virtually the same height, whereas the surface area increase of TS-II is approximately two times greater than that of TS-I, suggesting that the continuous striped structures have a larger surface area relative to the discrete spotted structures in the scan area. To further investigate the nanoscale Turing structures, the membranes were characterized by scanning electron microscopy (SEM) and transmission electron microscopy (TEM) analyses. The SEM images show that both structures are uniformly distributed throughout the membranes (Fig. 2A), which is consistent with the corresponding AFM measurements. A closer look at the membrane surfaces (Fig. 2B) reveals that the nanoscale Turing structures generally consist of close-packed hexagonal arrays or interconnected labyrinthine networks (figs. S16 to S18), with diameters ranging between 60 and 80 nm (fig. S19). The TEM analyses not only present the external features on the surfaces of the membranes but also provide morphology information on the internal characteristics of the structures. Projected area (Fig. 2C) and cross-sectional TEM (Fig. 2D) micrographs show that there are two types of voids in the Turing structures, with diameters ranging from 30 to 40 nm (fig. S20). The thickness of the Turing-type PA membranes is about 20 nm or less, two times thinner than that of traditional semiaromatic PA membranes (24). In three dimensions, the Turing structures are bubble or tube shaped, like Turing patterns in the BZ microemulsion system (25).

Table 1 Surface properties of the Turing-type PA membranes.

Comparison of the surface properties of the spotted and striped Turing structures. These results were acquired from AFM measurements over a scanning area of 5 μm by 5 μm. Reported are the averages and standard deviations.

View this table:
Fig. 2 Electron micrographs of the Turing-type PA membranes.

(A) Low-magnification SEM images of the two membrane surfaces. (B) High-magnification SEM images of the two different structures. (C and D) Projected area TEM images (C) and cross-sectional TEM images (D), showing the internal characteristics and three-dimensional morphologies of the two structures.

We evaluated separation performance of the two membranes by saltwater desalination tests and explored structure-property relationships in these membranes for water purification. The water and salts transport data show that both membranes exhibit excellent separation performance, surpassing the water-salt separation upper-bound line (Fig. 3A) of traditional nanofiltration membranes (26). Counterintuitively, water permeability and water-salt selectivity are both high, in contrast to the trade-off behavior of traditional polymer membranes (tables S3 and S4), where higher water permeability invariably leads to lower water-salt selectivity (27). Additionally, tube-structured membrane TS-II exhibits higher water flux and similar salt rejections compared to that of bubble-structured membrane TS-I under the same test conditions (Table 2). The water flux of TS-II is as high as 125 liters m–2 hour–1, which is approximately two times higher than that of TS-I. This result correlates well with the excess surface area ratio of the membranes, indicating that the Turing structures have a large effect on the water flux. On the basis of these observations, we hypothesized that there must be some specific sites with relatively higher water permeability in the Turing structures and that these high-permeability sites lead to membranes with an enhanced water transport property.

Fig. 3 Spatial distribution of water permeability sites in the Turing-type PA membranes.

(A) Correlation between water permeability and water-salt selectivity for the Turing-type PA membranes [TS-I (blue diamond) and TS-II (red star)] and other nanofiltration membranes (open circles). These data were obtained from water-salt separation tests (2000 ppm MgSO4, 25°C, 4.8 bar). The dashed red line is the permeability-selectivity trade-off for traditional semiaromatic PA membranes, and the solid black line is the empirical upper-bound relationship (26). Ps, salt permeability. (B) Schematic diagrams of the dynamic filtration experiments with GNPs and the transport of water across the Turing-type PA membranes. (C to F) Projected area TEM images showing nanoparticle deposition on the surfaces of the Turing-type PA membranes after 10-min filtration tests (1.0 × 1012 particles ml–1, 25°C, 4.8 bar). In (C) and (D), GNP percent surface area coverage is given in the upper right corner of each image. In (E) and (F), high-resolution TEM images of outlined areas from (C) and (D), respectively, show spatial distribution of nanoparticle deposition patterns and Turing structures.

Table 2 Separation performance of the Turing-type PA membranes.

The operating pressure was controlled at 4.8 bar, and the temperature was maintained at 25°C by a heat exchanger. The feed flow rate was 6 liters min–1, and the concentrations of salts in the feed solutions were 2000 parts per million. All measurements were made 1 hour after starting the filtration to stabilize the membrane performance. The rejections were calculated on the basis of the electrical conductivities of feed and permeate solutions.

View this table:

To verify this hypothesis, we used gold nanoparticles (GNPs) as probes in combination with microscopy methods to visually examine the spatial distribution of water permeability sites in the Turing-type membranes (Fig. 3B). GNPs are negatively charged under neutral conditions, and both membranes showed essentially the same surface charge behavior as GNPs (table S5). Consequently, for deposition to occur (figs. S21 and S22), drag forces had to overcome repulsive forces originating from nanoparticle-membrane electrostatic interactions (28, 29). Projected area TEM micrographs revealed that the deposition of GNPs was not uniformly distributed across the membrane surfaces. Nanoparticle surface area coverage for TS-I and TS-II were 6.0 (Fig. 3C) and 12.8% (Fig. 3D), respectively. GNPs deposited in specific areas of the membrane surfaces and formed clusters, leaving other areas of the surfaces uncovered or with considerably fewer sparsely distributed GNPs (figs. S23 and S24 and tables S6 and S7). Most of GNPs were deposited around bubble (Fig. 3E) or tube structures (Fig. 3F), which provides visual evidence supporting the existence of relatively higher water permeability sites in the nanoscale Turing structures (figs. S25 and S26).

Our work demonstrates that Turing structures can be produced by interfacial polymerization when appropriate initial conditions are created. Microscopic characterization of the Turing-type membranes reveals that the spatial distribution of relatively higher water permeability sites agrees well with the corresponding Turing structures at the nanoscale. These unusual nanostructures, which are generated by diffusion-driven instability, enable outstanding transport properties in both water permeability and water-salt selectivity.

Supplementary Materials

www.sciencemag.org/content/360/6388/518/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S26

Tables S1 to S7

References (3039)

References and Notes

Acknowledgments: We thank X. A. Zhao for helpful discussions and L. He, J. Hong, N. H. Rong, S. D. Shen, H. Wang, Y. Xu, M. J. Yu, and H. J. Zhang for technical assistance. Funding: This work was supported by the National Natural Science Foundation of China (nos. 51578485 and 21671171) and the National Basic Research Program of China (no. 2015CB655303). Author contributions: Z.T. performed the experiments. Z.T., S.F.C., X.S.P., and L.Z. designed the experiments and analyzed the data. All authors discussed the results and wrote the manuscript. Competing interests: L.Z., Z.T., and S.F.C. are inventors on patent application 201810120316.x submitted by Zhejiang University, which covers Turing-type polyamide membranes. Data and materials availability: All data are available in the manuscript or in the supplementary materials.
View Abstract

Subjects

Navigate This Article