Conjugated microporous polymer membranes for light-gated ion transport.

Inspired by the light-gated ion channels in cell membranes that play important roles in many biological activities, herein, we developed an artificial light-gated ion channel membrane out of conjugated microporous polymers. Through bottom-up design of the monomer molecular structure and by the electropolymerization method, the membrane pore size and thickness were precisely controlled on the molecular level. The obtained membrane exhibited uniform pore size and highly sensitive light-switchable response. The photoisomerization of the polymer chain resulted in a reversible "on and off" light control over the pore size and subsequently led to light-gated ion transport across the membrane for a series of ions including hydrogen, potassium, sodium, lithium, calcium, magnesium, and aluminum ions.

INTRODUCTION

In living cells, light-gated ion channels use light to regulate the ion transport, resulting in control of electrical excitability, calcium influx, and other crucial cellular processes (–). Channelrhodopsins (ChRs) are the first found and thus far the only class of light-gated ion channels identified in biology that have received much attention in recent years (–). In engineering, direct use of the light-gated ChRs has notable limitations due to the generally poor physical and chemical stability of the proteins in external environments. For this reason, extensive efforts have been dedicated to developing artificial light-gated ion channels (–), which are in the spotlight of research for potential applications in neurobiology, bioelectronics, resource recovery, and water purification.

A viable strategy for producing artificial light-gated ion channels is to modify nanopores with light-responsive functional groups (–). However, these methods are not easily processed and typically applicable for single light-gated ion channel preparations, which hinders their large-scale fabrication and practical applications (, –). A better and more practical way is to endow the microporous channels of synthetic membranes with light-responsive properties. This strategy is ambitious and technologically challenging. On one hand, the membranes must have numerous and uniform nanosized channels to ensure highly efficient ion transport and sieving. On the other hand, light-responsive molecules must be precisely incorporated into these channels to act as the photo switches. Although several attempts have been made to fabricate membranes with light-gated ion channels (–), it is still difficult to control the channel structure on molecular level using the existing fabrication techniques. Only with precise design of the channel architecture and chemistry could the membranes achieve the expected light-gated ion transport and sieving.

Conjugated microporous polymers (CMPs) are a unique class of porous organic materials with π-conjugation structures and intrinsic micropores (–). The membranes made from CMPs by the electropolymerization method in our previous works (–) have shown high porosity, uniform pore size, and impressive molecular/ion transport and sieving performance. In addition, the design of CMP monomers on molecular level could yield immense opportunities for the customization of functionalized membranes (, , , ). The structure and thickness of CMPs membranes can be easily and precisely controlled in nanometer scale by the electropolymerization strategy (–). All these advantages make CMPs one of the best candidates as the synthetic materials for the precise construction of smart ion channel membranes. Azobenzene (azo) is one of the most widely studied light-responsive molecules due to its rapid photoisomerization reaction (, –). The reversible trans-cis photoisomerization process of azobenzene can directly convert light energy into mechanical motion, resulting in a corresponding geometric change from the planar trans state with a size of 0.9 nm to the nonplanar cis isomer with a size of 0.6 nm (, ). The incorporation of the azobenzene group into the main chain of CMP molecule is expected to obtain light-responsive CMP monomer for the generation of the smart microporous channels, which has the potential to develop a new avenue to fabricate artificial light-gated ion channel membranes based on electrochemical strategy.

To achieve the desired light-gated response, we synthesized a de novo rigid-flexible azobenzene-containing monomer [azo-CMP, (E)-1,2-bis(4-(3-(9′H-[9,3′:6′,9″-tercarbazol]-9′-yl)propoxy)phenyl)diazene] (figs. S1 to S5). The monomer is homo coupled to each other via a continuous electrochemical oxidation–reduction reaction to form structurally well-defined elementary micropores. These inherent micropores are light-gated and interconnect to serve as smart ion channels in the azo-CMP membrane. The bottom-up design strategy is expected to precisely endow each building block of the respective channels in the membrane with the photo switches mechanism and thus successfully achieving the “on-off-on” photoisomerization response and controllable ion transport.

RESULTS

Synthesis and characterization of the azo-CMP monomer

The azo-CMP monomer has a butterfly wing–like structure as shown in Fig. 1A and fig. S6. Azobenzene is used as the light-switchable hinge of the wing; the alkyl chain serves as a soft linker to bridge the hinge and the electroactive carbazole scaffold. The geometric changes of the monomer upon photoisomerization are related to the end-to-end distance of the azo-CMP monomer (). Therefore, the length of the soft linker was rationally designed to increase the net distance change. Besides, the deformable alkyl chain is expected to provide sufficient free space for the photoisomerization of the azobenzene moiety, which plays a crucial role in the large geometric changes of the ion channels (, ). The photoisomerization of the monomer was first analyzed by molecular simulation. The results shown in Fig. 1A indicated that the theoretical end-to-end molecular distance of the azo-CMP monomer in the trans-isomeric form and the cis-isomeric form were 3.8 and 1.3 nm, respectively. The geometrical change in end-to-end distance (about 2.5 nm) that occurs upon photoisomerization is much larger than that of the pristine azobenzene unit (from 0.9 to 0.6 nm; fig. S7). Meanwhile, the simulation results also showed that the theoretical size of the elementary pore channel of the membranes could change substantially as shown in Fig. 1B. From the point of view of molecular structure, the soft carbon chains connecting the azobenzene and carbazole, are pulled when the azobenzene unit shrinks from 9 to 6 Å under ultraviolet (UV) light irradiation. The elongated soft chains and the nonplanar cis isomer invade the enclosed three-dimensional (3D) subnanometer space, resulting in the reduced channel size.

Fig. 1.

Schematic representation of the “bottom-up” design strategy for the construction of artificial light-gated ion channel membrane.

(A) Trans-cis-trans reversible isomerization of the synthesized azo-CMP monomer and (B) the elementary pore structure of the azo-CMP membrane.

During the experimental tests, the monomer exhibited rapid and reversible photoisomerization properties. The original transmonomer exhibits a strong absorption maximum peak at 342 nm (fig. S8A), which could be attributed to the π-π* azobenzene transition (, ). After only 10-s UV 365-nm irradiation, the intensity of the peak decreased rapidly from 100 to 69.1%, and the peak at 440 nm attributed to the n-π* transition increased correspondingly. After 90-s UV irradiation, the photoisomerization reached the photostationary state with a high ratio of trans state about 61.7% (fig. S8C). Meanwhile, the reversible cis-to-trans process can be almost fully achieved after 90-s visible-light (vis-light) irradiation with the peak of the π-π* absorption band recovered (fig. S8, B and D). The reversible photoisomerization of the monomer can be easily recycled by changing the wavelength of the irradiation as shown in fig. S9.

Fabrication of the azo-CMP membranes

Next, the azo-CMP monomer was used to prepare the CMP membranes via an electropolymerization process (Fig. 2A) using a three-electrode electrochemical cell (fig. S10). The reaction conditions were optimized to obtain smooth and defect-free azo-CMP membranes (table S1 and figs. S11 to S13). Figure 2B shows the cyclic voltammetry (CV) scan curve of the membranes during the fabrication process. The redox peaks observed in the positive and negative scans indicate the oxidation and reduction of the carbazole/dimeric carbazole units, respectively (, , ). Specifically, the electroactive carbazoles were first oxidized. The generated carbazole radicals in the oxidation were then coupled with each other to form dimeric carbazole cations. The formed dimeric carbazoles were subsequently reduced during the negative scans. The current obtained in each cycle rises gradually with increasing the CV scan, suggesting the growth of the membrane. With prolonged reaction time, the polymer chain was extended and the solubility was decreased. Eventually, a continuous azo-CMP membrane was formed. The chemical structure of the prepared azo-CMP membranes was investigated by Fourier transform infrared (FT-IR) spectroscopy (fig. S14), which confirmed the polymerization of carbazoles and the existence of azobenzene units in the membranes. The thickness of the membranes increases linearly with the number of CV scans, with a growth rate of 1.2 nm per CV cycle (Figs. 2C, fig. S15, and table S2). Thus, the thickness of the resulting membranes can be precisely and easily tailored. To the best of our knowledge, such precise thickness control is difficult to achieve with conventional synthetic routes, apart from the complex molecular layer-by-layer assembly/deposition (–). In addition, the morphology and surface hydrophilicity of the membranes can be easily modified by tuning the synthesis parameters (figs. S11 to S13). The fabrication parameters including scan rate, number of CV cycles, monomer concentration, solvents ratio, and voltage range were investigated and optimized. Briefly, under a relatively slow scan rate, more polymers were formed and deposited in each CV cycle, leading to a relatively rough and nonuniform membrane surface with many micro/nanostructures. The membranes prepared with a CV scan rate of 10 mV/s showed rough surface with large granular structures (fig. S11), exhibiting the well-known “lotus effect,” along with an extremely high water contact angle of 148°. The morphology of the membrane surface became smooth gradually as the scan rate increased. When the scan rate increased to 200 to 300 mV/s, the micro/nanostructures on membrane surface disappeared, and the water contact angle decreased to approximately 74° (fig. S11D). The number of CV cycles and the concentration of the monomer also showed notable effects on membrane surface structures. It tends to obtain more smooth and uniform membranes using relatively less CV cycles and low monomer concentration as shown in figs. S12 and S13.

Fig. 2.

Azo-CMP membranes.

(A) Structure of the synthesized monomer and the mechanism of electropolymerization. (B) CV profiles of the electrochemical oxidation–reduction reaction recorded over 50 CV scan cycles. (C) Membrane thickness as a function of the number of CV cycles. (D) Large-area surface SEM image of the azo-CMP@200-50c membrane on a copper grid. (E) High-magnification SEM image of the surface of the azo-CMP@200-50c membrane. (F) Cross-sectional SEM image of azo-CMP@200-50c membrane on an anodic aluminum oxide (AAO) support. (G) AFM height image of the azo-CMP@200-50c membrane transferred onto a silicon wafer and (H) corresponding height profile of the membrane. (I) AFM image of the azo-CMP@200-50c membrane. RMS, root mean square. (J) AFM image with the peak force quantitative nanomechanical mapping (PFQNM) and (K) the corresponding Young’s modulus profile of the membrane.

Last, the membranes fabricated under a scan rate of 200 mV/s and 50 CV cycles (azo-CMP@200-50c) were selected for further characterization and testing. Figure 2 (D and E) shows the surface morphology of the membrane azo-CMP@200-50c. The membrane is very thin and uniform, therefore, the underlying copper grid is visible. The membrane showed a very smooth surface with an average roughness of 4.1 ± 0.4 nm (Fig. 2I). The membrane structure was further analyzed by transmission electron microscopy (TEM) as shown in fig. S16, which revealed numerous micropores. The membrane azo-CMP@200-50c has a thickness of ~60 nm, which was measured by scanning electron microscopy (SEM) (Fig. 2F and fig. S17) and double-checked by atomic force microscopy (AFM) (Fig. 2, G and H). Figure 2 (J and K) shows that the Young’s modulus of the membrane azo-CMP@200-50c is 2.9 GPa, which is higher than that of polyamide membranes in the range of 0.1 to 2.7 GPa (, ).

Photoisomerization of the membrane

If photoisomerization occurs, then we would expect structural changes in the molecules and thus geometrical changes in the ion channels. These changes would cause a surface potential difference of the semiconductive azo-CMP membranes, which could be detected by the real-time in situ Kelvin probe force microscopy (KPFM) (fig. S18). As shown in Fig. 3 (A and B) and fig. S19A, the surface potential of the transmembrane is in the range of 590 to 610 mV. As expected, when the membrane was irradiated with UV light, the surface potential immediately jumped up to 680 mV (Fig. 3B and fig. S19B). The change in potential before and after UV light irradiation was able to be recorded in a KPFM image during the image capture process (Fig. 3A), indicating the rapid photo response of the membrane. The geometrical deformation that occurs during photoisomerization could allow more efficient interchain transport of charge at close crossing points, which increases the conductivity and surface potential of the membrane (). However, a deeper understanding of the mechanism of the photoelectric effect is required, and we are currently conducting further in-depth studies.

Fig. 3.

Trans-cis-trans reversible isomerization of the azo-CMP@200-50 membrane.

(A) Real-time in situ KPFM image of the membrane and (B) corresponding potential profile. (C) UV-vis absorption spectra of trans-to-cis isomerization under UV light and (D) ratio of trans/cis state with UV light irradiation time. (E) UV-vis spectra of cis-to-trans isomerization under vis-light and (F) ratio of trans/cis state with vis-light irradiation time.

We found that the free-standing azo-CMP membranes exhibited fast and reversible light-triggered motions (movie 1 and fig. S20). The light response is real-time, which is even faster than some reported well-designed polymeric light actuators and artificial muscles (, ). We attributed the light-triggered motions to the efficient geometric changes in the membrane channel structure, and these macroscopic motions provide direct evidence for the rapid light response of the membrane.

UV–visible (UV-vis) spectroscopy further confirmed the isomerization of the membranes. The results are shown in Fig. 3 (C to F). After 1 min of UV 365-nm irradiation, the intensity of the peak that arises from the π-π* azobenzene transition (, ) decreased rapidly by 14.8%. Meanwhile, the peak corresponding to the n-π* transition increased, simultaneously. After 15 min of UV irradiation, the photoisomerization reached the photostationary state with a final percentage of trans state about 65.6%. Note that the photoisomerization efficiency of the azo-CMP membranes is very high, as compared to the reported azobenzene-dangling metal-organic frameworks with a maximum cis yield rate of ~63% after over 30 min of UV irradiation (). The reversible cis-to-trans process was almost fully achieved after 15 min vis-light irradiation. However, without vis-light irradiation, the cis-to-trans relaxation is extremely slow. After 1 week in the dark at room temperature, the relaxation recovers only 5.2% calculated on the basis of the increased π-π* absorption band, as shown in fig. S22. This result indicates that the cis state of the azo-CMP membranes can be considered stable in the dark for the duration of our experiment. Figure S23 shows five consecutive cycles of the membrane under alternating irradiation with UV light and vis-light, indicating that the on-off-on light-switchable channels in the azo-CMP membrane can be repeatedly controlled by lights. In summary, both the results from the in situ KPFM and UV-vis spectroscopy indicate that the polymerized azo-CMP membranes have rapid and stable photo responsive trans-cis-trans isomerization.

The changes in channel size of the membrane in trans and cis states were determined from the nitrogen adsorption isotherm measured at 77 K (fig. S24). The pore-size distributions were estimated on the basis of the nonlocal density functional model. The isotherms of azo-CMP@200-50c membrane showed a steep rise at very low-pressure region in both trans and cis states (figs. S25C and S26C), indicating the microporous nature of the membrane. The transmembrane showed four peaks at 10.5, 11.5, 13.3, and 18.8 Å in micropore range (Fig. 4A and fig. S25B). The fraction of mesoporous (20 to 100 Å) in the pore-size distribution curves is negligible compared with that of the micropore. The former could result from the defects of the membranes formed in the sample collection, drying, or evacuation. After the in situ UV light irritation, the sample showed only one peak at 9.5-Å in micropore range (Fig. 4B and fig. S26B). This means the cis-membranes have narrower pore-size distribution and smaller pores than that of the transmembrane. In other words, the initial large pore channels with size over 10 Å in transmembrane were successfully “switched off” in the cis state. In addition, the nitrogen-sorption analysis further revealed that these smaller pores in cis-membrane can switch back to the original “on-state” after in situ vis-light irradiation (fig. S27). These results match well with the reversible isomerization of the membrane checked by UV-vis absorption spectra and KPFM method.

Fig. 4.

Pore-size distribution of azo-CMP@200-50c membrane.

(A) The membrane in trans and (B) cis states. (C) Simulated pore-size distribution of the membrane in trans and cis states. A 3D view of the membrane in (D) trans and (E) cis states (free volume in gray and Connolly surface in blue).

Molecular dynamics (MD) simulations were performed to help better reveal the changes in channel size. The results are shown in Fig. 4 (C to E) . The pore size of the transmembrane in the simulations distribute in the range of 3 to 15 Å. After fully photoisomerization, the cis-membrane showed a narrower pore-size distribution of 3 to 10 Å. The relatively large pores with a size of 10 to 15 Å were disappeared after the UV light irritation. In addition, the main peak left shifted by ~1.5 Å from transmembrane to cis-membrane. The trend of the changes in channel size presented in the MD simulations agrees well with the nitrogen-sorption results. The simulation study indicates that the trans-to-cis photoisomerization shrinks the pore size of the membrane and narrows their distribution and vice versa. These angstrom-scale changes in channel size are expected to contribute to obvious distinctions in ion permeability and selectivity.

Light-gated ion transport of the membranes

To investigate the performance of light-gated ion channel membranes for controllable ion transport, electrically driven ion permeation tests were performed in a laboratory quartz cell with two chambers (Fig. 5A). The two chambers were filled with the same concentrated salt solution. Ion transport was measured by analyzing the current-voltage (I-V) characteristics of the azo-CMP membranes in the trans and cis form. First, Al3+ with a hydrated size of 9.5 Å was used to check the membrane performance (Fig. 5B). Because of the existence of the large pores with a size over 10 Å inside the transmembrane, these large channels allow Al3+ ions to cross the membrane at a relatively fast permeation rate. As a result, the corresponding current/ion conductance obtained in this on-state (i.e., trans state) is larger than that in the cis state of the channels. The average Al3+ conductance reached 349.2 nS when the light-gated channels of the membrane are in the on-state (trans state). In contrast, the conductance greatly decreases to 25.0 nS after giving the membrane UV light irradiation. The decreased conductance indicates a reduced ion transport rate due to the changes in channel size of the cis-membrane. Specifically, the large-pore channels with size over 10 Å were switched off, meanwhile the size of the channels in cis-membrane became smaller, which hinders the transport of Al3+ ions. After irradiation with vis-light, it was interesting to find that the ion-gated channels were restored to the basic on-state. The initial current was almost fully recovered with a conductance of 313.1 nS.

Fig. 5.

Light-gated ion transport of the membranes azo-CMP@200-50c.

(A) Schematic diagram of the setup for the tests under electric field. (B) Al3+ conductance changes under alternating UV light and vis-light irradiation calculated on the basis of the data in fig. S28. The insets show the illustration of the controllable ion transport in the ion channels with on and off states. (C) I-V curves of the membranes recorded in 10 mM KCl solution during the trans-to-cis isomerization under UV light. (D) K+-relative conductance changes in successive cycles under alternating UV light and vis-light irradiation. The relative conductance is derived from comparing the conductance of K+ to that of deionized water (fig. S29). (E) Current of common ions recorded in on-state and off-state of the membrane under a voltage of 0.5 V. Note: The current in (E) was normalized by the number of the ion charges based on the data in figs. S28 and S30. (F) K+ and Al3+ permeation rate tested under concentration-driven ion permeation process. The inset shows the details of Al3+ permeation rate.

The membrane also showed light-gated ion transport to the smaller K+ although the channel size is larger than the hydrated size of the ion. As shown in Fig. 5C, the average current obtained in cis state was much lower than that in trans state. The decreased K+ current observed using cis-membrane indicates a reduced ion transport rate due to the shrunken channel size. The current depends directly on the electric charge of ions transported through the channels in a given time. Therefore, the reduced channel size increases the ion transfer resistance and substantially decreases the ion transport rate, which decreases the current intensity. The reversible photoisomerization of the ion channels could be easily recycled as shown in Fig. 5D. This means that the light-gated ion channels can be easily photoswitched between on-off-on (trans-cis-trans) states, resulting in considerable control over the ion transport across the membrane. Moreover, the on-demand ion transport was observed not only in the tests using Al3+ and K+ but in all the tests using seven different species of ions/protons (Figs. 5E and figs. S28, S30, and S31). To be specific, under the “off-state,” the current obtained using H+ decreased by 49.6% compared with that observed in the on-state. In addition, for the larger divalent and trivalent cations, the current decreased over 60 and 90%, respectively, after the UV irradiation. The light-gated ion transport property is advantageous for the remote regulation of the ion flux. In addition, it can be also used for tuning the ion selectivity. The selectivity of K+/Al3+ calculated from the current in Fig. 5E increased from 3.1 (on-state) to 17.2 (off-state). Note that the ion selectivity tested under the electrically driven process appears less impressive. It can be explained by the forced migration of large ions due to partial dehydration when an electric field is applied (–). Although electrical measurement is a highly sensitive and qualitative method for testing the controllable ion transport, the ion selectivity is limited by forced migration that occurs. With this limitation in mind, the followed concentration-driven ion permeation tests were further used to investigate the effects of the smart ion channels on selectivity.

The concentration-driven ion permeation tests were performed using a U-shaped glass diffusion cell (fig. S32). The left-side and right-side chambers were filled with deionized water and salt solution, respectively. As shown in Fig. 5F and fig. S33, the permeation rate across the membrane was almost constant for both K+ and Al3+ ions during the experimental period of 10 hours. Al3+ exhibited a permeation rate of 0.48 mmol/m2 per hour when the membrane was in the on-state. In contrast, K+ showed a much higher permeation rate of 8.92 mmol/m2 per hour due to its smaller hydrated size. After UV irradiation, the permeation rate of Al3+ decreased markedly by 71% and reached an extremely low value of 0.14 mmol/m2 per hour. However, the permeation rate of K+ decreased by only 20%, from 8.92 to 7.12 mmol/m2 per hour. Consequently, the selectivity of K+/Al3+ increased from the 18.6 originally to 50.9, indicating that the light-gated channels can selectively block Al3+ in the off-state. Note that the membrane showed light-gated regulation to all the common ions used in the concentration-driven process (fig. S34). In summary, all the results obtained in both the electrically driven and concentration-driven ion permeation tests indicate that the azo-CMP membrane has smart channels that can be switched between the on-state (large size) and the off-state (small size) by lights, thereby regulating ion transport across the membrane.

The transport properties of the azo-CMP membrane can be easily tuned by lights. It will be possible to extract two or more different target species without changing membranes in a continuous process. To demonstrate this, graded sieving experiments were performed to separate three different solutes using an azo-CMP@200-50c membrane (fig. S35). First, a simulated ternary mixture containing tetracycline, Al3+, and K+ was used to evaluate the sieving performance of the azo-CMP membrane in trans state. The channel size of the transmembrane is large enough for Al3+ and K+, which allows them to diffuse through the membrane. However, tetracycline was successfully retained in the feed solution because of its relatively large size, and no tetracycline was detected in the permeate solution as shown in fig. S35C. Correspondingly, the ternary mixture turned into a binary salt solution with the successful removal of tetracycline in the permeate side. Subsequently, the azo-CMP membrane was in situ irradiated by an UV light for 15 min, which then was used to separate the simulated binary salt solution. The cis-membrane has narrower pore-size distribution and smaller pores compared with the initial transmembrane. K+ could easily diffuse through the membrane alone, while Al3+ was rejected by the cis-membrane, leading to a high selectivity of K+/Al3+ up to ~60 in the binary mixture separation. As a result, the concentration of K+ in the permeate side increased continuously to 0.61 mM as shown in fig. S35D, but the concentration of Al3+ remained lower than 0.01 mM during the tests lasted for 10 hours. The results above demonstrate the good potentials of the light-gated ion channel membrane in graded sieving, which could be beneficial to pharmaceutical industry and smart dialysis.

DISCUSSION

Inspired by the light-gated ion channels ChRs, artificial light-gated ion channel membranes are developed. First, the azobenzene-containing CMP monomer was exquisitely designed on molecular level. The monomer consists of a light-switchable azobenzene core unit, a soft alkyl chain with finely tuned length and relatively rigid electroactive carbazole building units. Because each building block of the membrane channels has an azobenzene photo switch, a highly effective “trans-to-cis” photoisomerization response from the channels is ensured. Moreover, an electrochemical approach was carried out with a precisely tailored nanometer-scale thickness, in contrast to conventional membrane preparation methods. The azobenzene-containing monomer molecules form a well-defined conjugated 3D network with uniform intrinsic micropores via the continuous oxidation-reduction reaction. These numerous and interconnected micropores serve as smart ion channels. The geometrical changes of the channels induced by the photoisomerization allow on-off-on light control in the channels. As a result, the ion transport across the membrane can be regulated remotely and dynamically.

Light-gated ion transport across membrane has been demonstrated by both electrically driven and concentration-driven ion permeation tests. The artificial light-gated ion channel membranes showed rapid ion transport when the smart channels were in the on-state. The off-state resulted in a notable decrease in ion permeation and a remarkable increase in selectivity. Moreover, reversible photoisomerization of the ion channels could be recycled, and the permeability and selectivity of the membrane can be continuously and repeatedly tuned by light irradiation, which would be of great benefit to separation industry. In addition, the concept of light-gated smart CMPs can be extended beyond membrane applications because of the unique π-conjugation structure and intrinsic microporosity of CMPs, for example, nanoscale molecule memory, smart drug release, light-controlled supercapacitors, and photoresponsive chemosensors.

MATERIALS AND METHODS

Synthesis of the azo-CMP monomer

The design route, synthesis details and characterization of the azo-CMP monomer were presented in the Supplementary Materials (figs. S1 to S5).

The homogeneous electrolyte solution was obtained by dissolving 7.75 g of tetrabutylammonium hexafluorophosphate and 38.7 mg of azo-CMP monomer [(E)-1,2-bis(4-(3-(9′H-[9,3′:6′,9″-tercarbazol]-9′-yl)propoxy)phenyl)diazene] in 200 ml of a mixture of anhydrous dichloromethane and acetonitrile (4:1, v/v). A nonaqueous Ag/Ag+ electrode and a 6 cm–by–6 cm indium tin oxide (ITO)–coated glass was used as reference electrode and working electrode, respectively. A thin titanium metal plate was used as the counter electrode. The three-electrode electrochemical cell attached to an electrochemical workstation (model 660C, CH Instruments Inc.) was used for the fabrication of the azo-CMP membranes. CV was set in the range of −0.50 to 1.08 V with a scanning rate of 200 mV/s. The obtained membranes were immersed in a mixture solution of dichloromethane and acetonitrile overnight to remove electrolytes and unreacted monomers. Then, freestanding azo-CMP membranes were obtained from ITO glass using the reported surface tension–assisted transfer method (). These membranes can be easily transferred on potential energy surface (PES) porous support (0.22 μm in pore size) for further testing and applications.

Characterizations

The surface morphology and cross-sectional structure of the membranes were analyzed using a SEM (FEI Magellan 400, USA). The surface roughness and thickness of the membranes were analyzed using an atomic force microscope (Bruker Dimension Icon, Germany). AFM with a peak force quantitative nanomechanical mapping (PFQNM) function was also used to measure the surface Young’s modulus of the membranes according to our previously reported method (). The AFM surface potential images were collected via KPFM using a SCM-PIC-V2 probe (Dimension Icon SPM, Bruker). A high-resolution TEM (300 kV; FEI Titan CT, USA) was used to analyze membrane structure. The chemical structures of the azo-CMP membranes and monomer were analyzed by FT-IR (Nicolet iS10, Thermo Fisher Scientific, USA) and nuclear magnetic resonance (500 MHz; Bruker, Germany). The UV-vis spectra of the membranes and the monomer under UV 365-nm irradiation/vis-light irradiation with a light intensity of 0.2 mW/cm2 were analyzed by a UV-vis spectrometer (Evolution 220, Thermo Fisher Scientific, USA). Nitrogen-physisorption experiments performed at 77 K were used for the analysis of membrane porosity via surface characterization analyzer (3Flex, Micromeritics, USA). The samples were evacuated at 120°C for 30 hours before the measurement. More details about characterizations can be found in the Supplementary Materials. Ion concentration of the binary mixture solution was checked by inductively coupled plasma optical emission spectrometry (ICP-OES) (Optima 8300, PerkinElmer, USA). UV-vis spectrometer (Evolution 220, Thermo Fisher Scientific, USA) was used to analyze the concentration change of tetracycline in the solution.

Quantum chemical calculation

The pore size and geometry of the elementary pore unit of the azo-CMP membrane were optimized and calculated by the first-principles density functional theory simulation using B3LYP functional by Gaussian 16 code (–). The 6-311 g++ basis functions were applied to the systems ().

Construction of the atomic model

The Forcite module of the Materials Studio (MS) program was used for MD simulations for both the trans azo-CMP membrane and the cis azo-CMP membrane. First, 64 molecules were placed in a simulation box with the COMPASSII force field, and then the MD simulation for the simulation box compression with the adjustment of pressure and temperature was performed in the NPT ensemble with more than 1500,000 steps. The Berendsen bath coupling scheme was used (). Last, the stable structure was achieved at 298.15 K. The free volume figures of the above constructed membrane system were simulated by the Atom Volume and Surface of Materials Studio (MS) program. Zeo++ () was used to analyze the pore-size distributions.

Tests of the light-controlled ion transport

A laboratory-made two-compartment quartz cell attached to an electrochemical workstation (model 660C, CH Instruments Inc.) was used to test the light-controlled ion transport driven under electric field. An azo-CMP membrane deposited on a PES support was fixed between the two chambers of the cell. Both chambers contained salt solution (10 mM) with the same volume. A pair of Ag/AgCl electrodes was placed in each of the chambers. At least three independent membrane samples were tested. The voltage range was swept from −0.5 to 0.5 V with a step size of 0.05 V/s. The current was recorded as a function of applied voltage after the membranes were alternately irradiated with UV light and vis-light. The ion conductance, G (S), was calculated using Eq. 1where, I (in amperes) and U (in volts) represent current and potential, respectively.

A U-shaped glass diffusion cell was used for the concentration driven ion permeation tests. An azo-CMP membrane loaded on a PES support was mounted between the two filter chambers. Unlike electrically driven ion permeation, the permeate chamber was filled with 30 ml of deionized water, and the feed chamber was filled with 30 ml of salt solution (10 mM). Both chambers were shaken to avoid concentration polarization. A conductivity meter (CON2700, Eutech, USA) was inserted into the permeate chamber, which was connected to an online computer system to record the data and calculate the ion permeation rate. At least three independent membrane samples were tested. The ion permeation rate, J (in millimoles per square meter per hour), was calculated using Eq. 2where C (in millimolars) and V (in liters) is the concentration and volume of the solution on permeate chamber, respectively; A is membrane area (in square meters), and ∆t (in hours) is the test time.

Graded sieving experiments

Ternary mixture separation tests and binary mixture separation tests were performed under the concentration-driven process. The simulated ternary solution contains three solutes, tetracycline (200 parts per million), Al3+ (10 mM), and K+ (10 mM). The simulated binary solution contains two ions, Al3+ (10 mM) and K+ (10 mM). The ternary mixture separation tests were first performed. Subsequently, the membrane was in situ irradiated by a UV light for 15 min used for the binary mixture separation tests. The ion concentration was checked by ICP-OES, and a UV-vis spectrometer was used to quantify the concentration change of tetracycline.