Asymmetric ion transport through heterogeneous bilayers of covalent organic frameworks

Introduction

Artificial computers utilize electrons and holes for data processing; conversely, biological brains employ a variety of ions and their highly efficient transport across nanometer-scale synaptic ion channels for the transmission, storage and processing of information. In this manner, the energy costs of the brain are ultralow. For example, the human brain operates at approximately 20 W per day^1,2. Owing to their intrinsic fragility, biological nanochannels cannot be directly used in practical manufacturing and daily life. Instead, various bioinspired artificial nanochannel systems have been constructed and intensively studied for mass transport; this field is known as “nanofluidics”³. These systems range from zero-dimensional nanopores^4,5 and one-dimensional nanotubes^6,7,8 to two-dimensional slits^{9,10,11,12,13} and three-dimensional gel networks^{14,15,16,17,18}. Understanding and manipulating ion transport in nanometric confinement is one of the significant goals of nanofluidics. The attainment of this goal is extremely beneficial for the application of artificial nanochannels in energy conversion^{19,20,21,22,23}, sea water desalination^24,25, membrane separation^26,27, neuromorphic computation^28,29,30,31, and wastewater management^32,33.

Very recently, benefiting from the successful fabrication of covalent organic framework (COF) monolayers featuring ultrahigh pore density ( ~ 10¹³ cm^-2), molecule-scale thickness and tailorable chemical ingredients³⁴, several attempts have been made to investigate the ion transport behaviors within these structures. For example, ion transport characteristics driven by a salinity gradient in both aqueous^35,36 and organic³⁷ phases have been reported, where ultra-large ion fluxes and recorded output power densities were obtained. Based on these findings, ion transport through the COF monolayers driven by coupled mechanical forces and electrical fields was also studied. By combining experimental results with theoretical modeling, a distinct mechanical‒electrical interplay mechanism for modulating ion transport has been proposed; this is different from that of low-permeability nanochannel systems³⁸. However, thus far, nanofluidic devices based on COF monolayers have been homogeneous, leading to nearly symmetric ion transport properties. Studies on ionic transport behavior across ultrathin heterogeneous COF membranes are limited and still present significant challenges.

Herein, we describe the fabrication of bi-layered tetraphenylporphyrin-based COF heterojunctions using a two-step successive self-assembly and subsequent interfacial polymerization methodology. These heterojunctions consist of a negatively charged monolayer atop a positively charged monolayer. When an electric field is applied as the driving force, rectified ion transport is observed for the first time (Fig. 1a). Notably, this ionic current rectification property can be tuned by the salt concentration and structural composition of the membranes, and it is universal across various COF heterojunctions with different chemical constituents. These findings are significant for the development of innovative nanofluidic materials and devices.

Asymmetric ion transport through heterogeneous bilayers of covalent organic frameworks — **Fig. 1: Characterization of ultrathin COF membranes.**

Materials and methods

Materials

5,10,15,20-Tetrakis(4-aminophenyl)-porphyrin (TAPP) and 2,5-dihydroxyterephthalaldehyde (DHTA) were purchased from Bide Pharmatech Ltd., China. 5,10,15,20-Tetrakis(4-aminophenyl)-porphyrin zinc(II) chloride (Zn-TAPP), 5,10,15,20-tetrakis(4-aminophenyl)-porphyrin nickel(II) chloride (Ni-TAPP) and 5,10,15,20-tetrakis(4-aminophenyl)-porphyrin copper(II) chloride (Cu-TAPP) were obtained from Yanshen Technology Co., Ltd., Jilin Chinese Academy of Sciences, China. Scandium(III) triflate was obtained from Aladdin, China. A SiN_x/Si chip with a 100 nm-thick SiN_x membrane (0.5 mm × 0.5 mm) supported on a silicon substrate (5 mm × 5 mm, 200 μm-thick) was acquired from YW MEMS (Suzhou) Co., Ltd., China. Ultrapure water (18.2 MΩ•cm, Milli-Q) was used in all experiments.

Preparation of the heterogeneous HCOF@ZnCOF membranes

The heterogeneous HCOF@ZnCOF membranes are prepared utilizing a custom-fabricated Teflon reactor that is tightly mounted to minimize any possible shaking during the process of membrane growth and transfer, with the goal of ensuring membrane integrity. The procedure involves two successive steps of laminar assembly and interfacial polymerization (IP); details on these steps can be found in our work or others (Figure S1)^34,35,38. Briefly, during each step, the reactor is initially filled with an aqueous solution of the DHTA linker (0.05 mM) with scandium(III) triflate (0.5 mg/mL) as the catalyst, followed by immersion of the target substrate in the aqueous phase at an inclination angle of 15°. Then, an immiscible organic pentane phase is slowly decanted onto the aqueous solution to form a stable and sharp interface. For the first IP process, a silicon wafer, quartz, TEM grid or SiN_x/Si chip is used as the substrate, and 14 μL of the monomer Zn-TAPP solution (0.1 mM, pyridine/methanol = 3:1 v/v) is injected into the pentane phase at a flow rate of 10 μL/min via a syringe pump (Longer Pump System Inc., LSP02). Afterward, the amphiphilic Zn-TAPP molecules assemble into a continuous laminar monolayer at the interface and gradually polymerize with DHTA over a period of 30 min. Next, the bottom phase is pumped out at a constant rate of 1 mL min^-1, while the ZnCOF monolayer settles atop the tilted substrate. Finally, the as-formed COF membrane is flushed sequentially with n-hexane, isopropyl alcohol, and methanol to remove surface residues and blown dry with a mild nitrogen flow. For the second IP course, the ZnCOF monolayer prepared above serves as the new substrate, and 12 μL of the monomer TAPP solution (0.1 mM, pyridine/methanol = 3:1 v/v) is injected into the pentane at an injection rate of 10 μL/min. After a duration of 30 min for preassembly and subsequent polymerization and a period of 30 min for expulsion of the bottom phase, the newly formed HCOF monolayer settles onto the ZnCOF monolayer. Similarly, the as-fabricated HCOF monolayer is also washed sequentially with n-hexane, isopropyl alcohol, and methanol and dried under a nitrogen flow. For the preparation of the HCOF@NiCOF and HCOF@CuCOF heterojunction membranes, the Zn-TAPP solution is replaced with a Ni-TAPP (0.1 mM) solution and a Cu-TAPP (0.1 mM) solution, respectively.

Characterization

A morphological study of the COF membranes is carried out using scanning electron microscopy (SU8010, Hitachi, Japan) and atomic force microscopy (OLS 4500, Olympus, Japan). Porous carbon TEM grids with 2 μm diameter holes and silicon wafers are used as substrates for these characterizations. The chemical composition of the COF membranes is characterized using UV‒vis spectroscopy via a Lambda-950 (Perkin Elmer Instruments) UV‒vis‒near-infrared spectrophotometer, where quartz is used as the substrate. Zeta potential measurements of the ZnCOF and HCOF monolayers in KCl solutions of various concentrations were conducted on a solid surface zeta potential analyzer (SurPass, Anton Paar, Austria). For the ZnCOF monolayer, a commercial hydrophilic polytetrafluoroethylene (PTFE) film was used as the substrate, whereas for the HCOF monolayer, a plasma-treated hydrophobic PTFE membrane was chosen as the substrate.

Electrical measurement

First, an opening aperture with a diameter of 2 μm is manufactured within the free-standing 100 nm thick SiN_x membrane via a focused ion beam system (Nova200 NanoLab, FEI Inc.). Then, the single-layered and bi-layered COF membranes, which are grown via the abovementioned method, are transferred onto the 2 μm wide aperture on the SiN_x/Si chip. The COF membrane covering the SiN_x/Si chip is then mounted on a customized Teflon disk with a 3 mm wide pore and sandwiched between two reservoirs filled with different electrolyte solutions of various concentrations (Fig. 1a).

The electrochemical measurements were then conducted using a Keithley 2634B source meter (Keithley Instruments). A pair of Ag/AgCl electrodes are used to apply a transmembrane voltage and record the ion current through the COF membranes.

Results and discussion

The bi-layered heterogeneous COF membranes are prepared by sequentially depositing positively and negatively charged tetraphenylporphyrin-based COF monolayers, with square nanopore arrays; these are denoted as ZnCOF and HCOF based on the desired substrates (Figure S1). This process follows the recently developed preassembly of 2D monomers and the subsequent interfacial polymerization methodology^34,35. After the silicon substrate is covered with the first ZnCOF layer, atomic force microscopy (AFM) is used to examine its surface topography and thickness. The micrograph reveals a flat membrane with a thickness of approximately 1 nm, extending across a micrometer scale (Fig. 1b). The as-prepared ZnCOF membrane subsequently serves as a new substrate for the deposition of the second HCOF layer. The AFM image of the newly formed membrane indicates a further increase in the thickness of the membrane to approximately 2.1 nm (Fig. 1c). Notably, the HCOF@ZnCOF membrane is also very smooth at the micrometer scale. Since the theoretical van der Waals thickness of both ZnCOF and HCOF is 0.92 nm according to density functional theory (DFT) calculations^35,38, the prepared membrane is a bilayer of ZnCOF and HCOF. This deduction is further supported by the scanning electron microscopy (SEM) results, UV‒vis characterization and X-ray photoelectron spectroscopy (XPS) analysis of the heterogeneous COF membranes (Fig. 1d, e and Figure S2). The SEM images of the membrane surface on a porous carbon grid revealed the integrity and free-standing nature of the COF membrane across a large area. Notably, the presence of areas with varied contrasts indicates a height difference between the first and second COF layers (Fig. 1d). As demonstrated in Fig. 1e, UV‒vis characterization of the single-layered ZnCOF and HCOF membranes revealed adsorption peaks at approximately 438 and 431 nm, respectively; these results indicate the formation of large conjugated frameworks^34,35. However, for the bi-layered HCOF@ZnCOF membrane, an overlapping and intensified adsorption peak at 435 nm is observed. This result indicates the presence of two independent ZnCOF and HCOF components in the bi-layered heterogeneous HCOF@COF membrane. As shown in Figure S2a, the successful fabrication of the ZnCOF monolayer on the Si substrate is verified by the presence of both the Zn2p1 peak at 1045 eV and the Zn2p3 peak at 1022 eV. Notably, after the deposition of the second HCOF layer without Zn, the intensities of both the Zn2p3 peak and the Zn2p1 peak are reduced by more than 2 times (Figure S2b); this indicates the formation of the bi-layered heterogeneous HCOF@ZnCOF membrane.

To investigate the ion transport properties through single-layer and bi-layer COF membranes, a SiN_x chip covered with membranes is mounted on a Teflon disk and sandwiched between two reservoirs filled with 0.1 mM KCl solution for measurement of the current–voltage (I–V) curve (Fig. 1a). A voltage bias is applied on the electrolyte bath in contact with the ZnCOF layer through a pair of Ag/AgCl electrodes, while the HCOF side is grounded. The ratio of the current at −2 V and +2 V, I(V = -2 V)/I(V = 2 V), which is also known as the ionic current rectification (ICR) ratio, is employed as a parameter to quantify the ICR performance of the nanofluidic membranes. Under a scanning voltage between −2 V and +2 V, the I-V curves for the single-layer membrane of the ZnCOF and the HCOF supported on the 2 μm-wide SiN_x aperture show very weak rectification, with an ICR ratio of no more than 1.2 (Fig. 2a, b). These phenomena likely originate from the heterostructures between the porous COF monolayers, which feature 2.6 nm-wide nanopore arrays, and the underlying SiN_x aperture³⁹. In sharp contrast, when the positively charged ZnCOF monolayer is further covered by a negatively charged HCOF monolayer, the I-V curve of HCOF@ZnCOF becomes remarkably nonlinear and rectified with an ICR ratio of 3.0 (Fig. 2c, d). Notably, this ICR performance is very stable over repeated transmembrane voltage scans (Figure S3). The preferential direction of the ionic current is from the negatively charged HCOF layer to the positively charged ZnCOF layer; this result is in accordance with the predictions stemming from the nanofluidic resistance model described below (Figure S4)^40,41.

**Fig. 2: Rectified ion transport behaviors in bi-layered heterogeneous COF membranes.**

Briefly, the transmembrane ion current is determined by the concentration of total charge carriers (K⁺ and Cl⁻) within the heterogeneous HCOF@ZnCOF membrane. An equivalent electric circuit of the total nanofluidic ion resistance can be considered as a tandem connected resistance from the positively charged ZnCOF layer (R_p(ZnCOF)) and the negatively charged HCOF layer (R_n(HCOF)) (Figure S4a). Higher ion concentrations within the ZnCOF (c_p(K⁺), c_p(Cl⁻)) and the HCOF (c_n(K⁺), c_n(Cl⁻)) layers correlate to their lower corresponding ion resistances (Figure S4b). The concentrations of K⁺ and Cl⁻ in the bulk solution are denoted as c_bulk(K⁺) and c_bulk(Cl⁻), respectively. When V = 0, owing to electrostatic equilibrium, the K⁺ concentration follows the sequence c_n(K⁺) > c_bulk(K⁺) > c_p(K⁺), whereas the Cl⁻ concentration follows the sequence c_n(Cl⁻) < c_bulk(Cl⁻) < c_p(Cl⁻). When V > 0, driven by the electric field, both K⁺ and Cl⁻ are expelled from the membrane into the bulk solution; however, their transportation from the bulk solution to the oppositely charged ZnCOF and HCOF parts is electrostatically impeded. This results in a low concentration of the charge carriers inside the COF membrane, yielding higher ionic resistance and lower ion current. Conversely, when V < 0, K⁺ and Cl⁻ are carried from the bulk solution into the HCOF and ZnCOF layers of the membrane. This increases the concentration of charge carriers inside the membrane, leading to lower ionic resistance and higher ion current.

Notably, the observed ICR phenomenon can also be effectively reproduced in other HCOF@ZnCOF devices (Fig. 3a). By varying the KCl concentration from 0.1 mM to 1 M, the I-V curve of the HCOF@ZnCOF membrane transitions from a nonlinear asymmetric state in 0.1 mM, 1 mM, 10 mM and 0.1 M KCl (Fig. 3a‒d) to a linear symmetric state in 1 M KCl (Fig. 3e). Correspondingly, the ICR ratio progressively decreases from 2.7 to 2.1, 1.6, 1.4 and 1.0, respectively (Fig. 3f). For comparison, single-layer homogeneous COF membranes were also assessed. As shown in Figure S5 and Figure S6, both the ZnCOF and HCOF membranes exhibit nearly symmetric and non-rectified I-V curves across the entire tested concentration range. These results confirm that the asymmetric ion transport originates from the heterogeneous structure of the HCOF@ZnCOF membranes. Specifically, the opposite surface charge polarities of the ZnCOF and HCOF layers (Figure S7) are responsible for the rectified ion transport behavior. This is further shown by the surface charge density-dependent ICR performance of the COF heterojunction in electrolytes with various pH values (Figure S8).

**Fig. 3: Concentration-dependent rectification properties of the bi-layered heterogeneous COF membrane.**

Notably, previous studies on the concentration dependence of ICR phenomena in 1D and 2D nanofluidic membranes have shown that ICR ratios often peak at intermediate electrolyte concentrations and decrease at both low and high electrolyte concentrations^41,42. However, for the bi-layered COF heterojunctions, the ICR ratio reaches its maximum value at the lowest concentration (0.1 mM) and decreases as the KCl concentration increases. This distinct trend is likely caused by the limited surface charges of the tetraphenylporphyrin-based COF monolayers, which require fewer counterions to establish electrostatic equilibrium at V = 0 (Figure S4b). At high electrolyte concentrations, the electrostatic screening effect from the bulk ions narrows the gaps among c_n(K⁺), c_bulk(K⁺) and c_p(K⁺) and among c_n(Cl⁻), c_bulk(Cl⁻) and c_p(Cl⁻); this leads to a gradual loss of the ICR function. Specifically, our ultrathin COF heterojunctions work better in more dilute electrolytes. Additionally, the ionic conductances of ZnCOF, HCOF and the composite HCOF@ZnCOF at different KCl concentrations were also calculated from their corresponding I-V curves (Figure S9). Compared with the single-layered COF membranes, the bi-layered membrane shows reduced but still comparable transmembrane conductance in higher-concentration KCl solutions (10 mM, 0.1 M and 1 M). As the concentration further decreases, all three membranes display saturated conductance; thus, the ion transport is governed by the surface charge of the membrane at low concentrations.

Interestingly, the introduction of an additional ZnCOF or HCOF monolayer to the heterogeneous HCOF@ZnCOF membrane results in diode-like I-V characteristic for the tri-layered HCOF@ZnCOF@ZnCOF and HCOF@HCOF@ZnCOF heterojunctions (Fig. 4a, b). However, compared with those of the HCOF@ZnCOF membrane with equivalent relative proportions, the ICR functions of these tri-layered heterogeneous membranes are weakened (Fig. 4c) because of their tendency toward a more homogeneous membrane structure. This observation aligns with previous empirical findings in 1D and 2D nanofluidic nanochannels, where the optimal ICR ratios are achieved in relatively balanced asymmetric pore structures^43,44.

**Fig. 4: Structure–composition–dependent rectification behaviors of the heterogeneous COF membranes.**

Next, we replace the ZnCOF monolayer with other positively charged COF monolayers, such as NiCOF and CuCOF, to prepare different bi-layered heterogeneous COF membranes with various chemical compositions. The HCOF@NiCOF and HCOF@CuCOF membranes are characterized using the I-V curves in KCl solutions of different concentrations (Fig. 5, Figure S10 and Figure S11). Remarkable ICR phenomena were also observed at both the HCOF@NiCOF and HCOF@CuCOF heterojunctions, as shown by the smooth, nonlinear, and asymmetric I-V curves recorded in 0.1 mM KCl (Fig. 5a, c). The ICR ratios are calculated to be 3.2 and 4.8 for the HCOF@NiCOF and the HCOF@CuCOF membranes, respectively (Fig. 5b, d). The preferential direction for ion transport is from the negatively charged HCOF monolayer to the positively charged NiCOF or CuCOF monolayers; this result is identical to that of the HCOF@ZnCOF heterojunctions. By replacing 0.1 mM KCl with more concentrated solutions ranging from 1 mM to 1 M, the I-V curves of the two types of COF heterojunctions gradually transition from nonlinear and rectified states to the linear and non-rectified states. Accordingly, the ICR ratio progressively decreases with increasing salt concentration to 1.0 for HCOF@NiCOF and 1.1 for HCOF@CuCOF in 1 M KCl; this is the same trend as that of the HCOF@ZnCOF membrane. These results clearly demonstrate the wide universality of asymmetric ion transport in ultrathin heterogeneous COF membranes.

**Fig. 5: Universality of the rectification properties of the different COF heterojunctions with various chemical ingredients.**

Conclusion

In summary, for the first time, we prepare ultrathin bi-layered heterogeneous nanofluidic membranes composed of two oppositely charged COF monolayers and demonstrate their asymmetric ion transport behavior. The preferred direction for ion transport is from the negatively charged layer to the positively charged layer. Moreover, the ionic current rectification performance of the COF heterojunctions can be modulated by varying the electrolyte concentration and structural composition, and this behavior is widely observed in heterogeneous membranes with different chemical compositions. These ultrathin COF heterojunctions can be applied in various fields, such as nanofluidics, smart ionic devices, and energy conversion.