Membranes play an imperative role in life science and industrial applications, such as disease treatment and water purification [1, 2]. It is of important significance to understand and manipulate the transportation of water through membranes in many physical, chemical, biological, and technical applications. In different artificial and biological membranes, functional groups on the membrane determine the properties of the membrane and have a great influence on the permeability of water [3]. For example, as a two-dimensional (2D) material derived from graphene, GO has almost completely different properties from graphene [4, 5]. Compared to hydrophobic graphene, the polar oxygen-containing functional groups on the membrane make it hydrophilic and have different properties from graphene in many ways [6, 7]. GO membranes, as a feasible artificial membrane for water desalination and wastewater treatment [8–10], was first developed by Nair et al., and it was discovered that stacking GO film can selectively block the motion of non-aqueous solutions and allow water permeation through a unique pathway [11]. From then, many attempts to develop the application of GO membrane and methods to control the interlayer spacing of stacking GO have been discovered [12–15].

The interlayer spacing of the stacking GO membrane determines the ion and water permeation efficiency, a smaller interlayer spacing leads to better ion-sieving performance, while the flow rate of water decreases exponentially with the reduction in interlayer spacing [16–20]. The carboxyl group at the edge of the GO membrane is one of the major factors affecting the permeation of water molecules and other ions through the membrane due to the steric effect [21–24]. Previous attempts made to enhance the permeability of GO were mainly surface modification or reassembling [25, 26]. In contrast, THz electromagnetism (EM) stimulus is a novel and efficient approach. In recent years, the application of THz EM waves in biochemistry and the theoretical discussion on the interaction between light and matter are increasing gradually [27], starting from using an electric field, to enhance the water transport in nanochannels and gas dissolving in aqueous solutions [28, 29], to the application of terahertz-regulated dissolution of methane hydrate [30] or using THz wave to detect different materials and defects [31, 32]. Furthermore, in a biological system, THz stimuli can be applied to enhance the permeability of ion channels by directional interaction with functional groups [33–35] and even can be used to accelerate the DNA unwinding [36]. Biophotons at high-frequency THz range can also participate in the replication of DNA and neurotransmission [37, 38]. Previous research on THz–water interaction proves that THz wave can enhance the water permeability by changing the H-bond structure of one-dimensional (1D) and two-dimensional (2D) water in artificial membrane materials such as graphene and carbon tube [39–44]. As a 2D material sharing analogous behaviors with biological membranes, functional groups on GO membranes can also be stimulated with THz EM with high efficiency and directionally [45]. Inspired by the studies of THz light enhancing water transportation by interacting with matter and functional groups, we propose a light-based modulation method targeting edge functional groups to change the permeability of the GO membrane.

In this study, based on a special GO membrane model and primarily considering carboxyl groups on the edge of the membrane (Figure 1A), we uncovered the interaction between 44.0 THz EM wave and carboxyl, as well as the subsequent influence on the conformation of GO membrane water channels. The underlying mechanism is shown to be that THz EM stimuli can change the oscillation of carboxyl at the entrance and exit of the channel by affecting the carboxyl groups resonantly, thereby reducing the energy barrier for water molecules to enter and exit the GO membrane, thus improving the permeability of GO membranes by nearly five times.

To explore the influence of THz EM stimulus on the dynamic properties of edge functional groups on a membrane and its subsequent effect on the conformation of the water channel and the permeability of the membrane, we designed a GO membrane model by stacking two sheets of GO in parallel with lengths, widths, and interlayer spacings of 3.42, 4.94, and 0.78 nm, respectively. Conventionally, the molecular structure of GO contains hydroxyl and epoxy functional groups on the basal plane and carboxyl grafted on the edges of the sheet, following the well-known Lerf–Klinowski model [46]. In our previous research, we studied the effect of the interaction between light and hydroxyl groups on the permeability of the GO membrane, while the epoxy shares a similar property with hydroxyl [45]. Relatively, the location of carboxyl groups exerts its effect on water permeation different from that of hydroxyl and epoxy groups, but it is an ideal model for studying the influence of edge functional groups on water permeation. Therefore, we focused on a pure carboxyl-functionalized GO membrane model to simplify the molecular dynamic simulation and the following data analysis. In order to ensure that the water can be affected by the carboxyl on the path of permeating through the GO membrane, we grafted the carboxyl at the edge of the GO sheet with a certain space around 0.4–0.8 nm. The charge model for carboxyl was based on the carboxyl group in ASPH amino acid to ensure that the system was electrically neutral. The GO membrane was centered in the simulation box with side lengths of L _x = 5.75 nm, L _y = 5.11 nm, and L _z = 3.78 nm. Then, we filled the rest of the box with flexible SPCE-modelled water molecules [47–49]. Consequently, the size of the GO membrane model and the box ensures that the water molecules can only permeate through the GO membrane along the x-axis (Figure 1B; Supplementary Figure S1).

EM waves will be applied to the entire system at the beginning of the MD simulation. Since the electric component of EM waves dominates the interaction of the EM wave with small molecules at room temperature [40], an electric field E t = A · u · cos ω t + φ was applied to simulate the EM wave, where A is the maximum amplitude of the electric field that determines the strength of the EM wave. The polarization direction of the wave was set to u = (1, 0, 0), which is parallel to the membrane plane. The frequency f of the EM wave relates to the angular frequency ω via the equation f = ω/2π. Furthermore, to investigate the effect of hydrostatic pressure and THz stimuli together on the permeability of the membrane, we used the accelerated function of GROMACS to generate desired pressure difference ( ∆ P ) [50]. F was applied to define the permeability of the GO membrane (Figure 1C), where F ₁ and F ₂ are the number of water molecules traversing forward and backward from one side of the membrane to the other along the x-axis, respectively. Furthermore, F _flow and F _flux are the sum and difference of F ₁ and F ₂ divided by the product of the simulation time and the cross-section area of the channel, respectively. Therefore, we use F _flow to describe the bidirectional permeability and F _flux to describe the unidirectional permeability of the membrane. To minimize the influence of temperature fluctuation on the simulation, a Nose–Hoover thermostat with a time constant for coupling of 0.2 ps was applied to maintain the average system temperature at 300 K with a fluctuation under 5 K (Supplementary Figure S2). More details of the simulation methods are shown in Supplementary Section S1.

Under the effect of 44.0 THz EM, the GO membrane becomes more permeable to water molecules, but repellency to ions is maintained (Figures 1D–F). Evidently, the F _flow tends to increase non-linearly with increasing EM stimuli strength A under an equilibrium state, while the F _flux only fluctuates within a certain value without a substantial increase (Figure 1D). Under normal conditions without osmotic pressure and THz EM stimuli, the value of the F _flow and F _flux is around 2.42 ± 0.09 ns⁻¹·nm⁻² and 0.25 ± 0.01 ns⁻¹·nm⁻², respectively. With the stimuli of 44.0 THz EM, the F _flow will increase rapidly as A increases from 0.1 to 1.0 V/nm and remain in a super state with F _flow = 11.89 ± 0.44 ns⁻¹·nm⁻² when A ≥ 1.0 V/nm. In contrast, the F _flux fluctuated between 0.25 ns⁻¹·nm⁻² and 0.65 ns⁻¹·nm⁻² and does not increase with the stimulation of 44.0 THz EM. The reason behind this fact is that the values of F ₁ and F ₂ are close to each other under normal conditions, while the THz EM stimuli will boost up both F ₁ and F ₂, resulting in F _flux fluctuating within a certain range. However, if we applied 44.0 THz EM stimuli under a gradient field of ΔP = 500 MPa, both F _flow and F _flux will be excited to a super permeation state (Figure 1E). This is because the gradient field of ΔP = 500 MPa will boost up F ₁ or F ₂ depending on the direction of osmotic pressure, leading to a huge diverse between F ₁ and F ₂; thus, the value of F _flow and F _flux will be close to each other. Consequently, in the absence of THz EM stimuli, the osmotic pressure of ΔP = 500 MPa can enhance the permeability of the membrane in both F _flow and F _flux to 4.81 ± 0.31 ns⁻¹·nm⁻² and 4.78 ± 0.29 ns⁻¹·nm⁻², respectively. Subsequently, both F _flow and F _flux can increase by about four times under the stimulation of 44.0 THz EM with strength A ≥ 1.0 V/nm, reaching the values of F _flow = 20.60 ± 0.99 ns⁻¹·nm⁻² and F _flux = 19.54 ± 0.91 ns⁻¹·nm⁻². This fact indicates that THz EM stimuli can enhance both the bidirectional and unidirectional permeability of the membrane with the presence of a gradient field.

Notably, we used higher osmotic pressure to make the phenomena more obvious. However, similar phenomena will occur in a lower gradient field, such as ΔP = 50 MPa (see details in Supplementary Figure S3; Supplementary Section S2). Neither gradient field nor THz EM can change the ion rejection of the membrane (Figure 1F). The GO membrane with an interlayer spacing of 0.78 nm has a strong steric effect on ions, which can refuse ions to enter the channel, as F _Ion remained 0 ns⁻¹·nm⁻² at any stimuli strength A. In other words, 44.0 THz EM stimuli can enhance the water permeability of the membrane while keeping the ion rejection of the membrane unchanged.

In order to reveal the mechanism behind the phenomenon of the specific frequency of EM stimuli that can enhance the permeability of functionalized membranes, we focused on the vibration spectrum of carboxyl groups, which are essential for the permeation of water through the membranes. In the classical approximation, the absorption spectrum intensity (I) in terms of frequency can be calculated by using Fourier transform for the autocorrelation function of total charge current J t = ∑ i q i v i t [44], where q _i and v _i represent the charge and velocity of the ith atom, respectively. The first principle of frequency selection is that the stimulation at a particular frequency can resonate with the vibration of functional groups on the membrane (see the frequency selection of other membrane models in Supplementary Figure S4). Furthermore, the stimulation at the selected frequency is weakly absorbed by the bulk water and confined water. As shown in Figure 2A, the carboxyl groups have characteristic peaks at 35.0 THz, 44.0 THz, and 112.0 THz, which correspond to the bending vibration of O-H and the stretching vibrations of C-O and O-H, respectively. They are far away from the prominent absorption peaks of bulk water and confined water, denoting that carboxyl can resonantly absorb the photon energy from the 35.0, 44.0, and 112.0 THz EM stimuli. The vibration modes of C-O and C=O show certain similarities.

To clarify the belonging of the frequency of 44.0 THz, we calculated the vibration of C-O and C=O separately and found out that the vibration of 44.0 THz is mainly contributed to the C-O stretching (Figure 2B). It is essentially in agreement with the result from the FT-IR measurement, indicating that the simulations are well calibrated. Among these characteristic frequencies, only 35.0 and 44.0 THz EM stimuli are able to enhance the permeability of the GO membrane, and the mechanism behind this enhancement is similar (the effect of 35.0 THz EM stimuli on the permeability of the membrane is shown in Supplementary Figure S5; Supplementary Section S4). Therefore, we will focus on 44.0 THz EM stimuli to uncover the mechanism behind this enhancement. From Figure 2C, we can clearly see that the vibration mode of the carboxyl group changes greatly after the application of 44.0 THz EM stimuli. Notably, the vibration mode with an intrinsic frequency of 44.0 THz is greatly enhanced by the corresponding stimuli, while other vibration modes had been reduced to a relatively weak state. In other words, due to the over-damping phenomenon caused by EM stimuli, all other vibration modes of carboxyl are suppressed, except for the vibration mode with an intrinsic frequency of 44.0 THz. As a result, the motion of the carboxyl group at the entrance of the water channel has changed dramatically under the influence of EM wave stimulation, which undoubtedly changes the permeability of the GO membrane.

Since the carboxyl group is located at the edge of the GO membrane and is hydrophilic, it tends to oscillate up and down in the direction perpendicular to the membrane plane under the attraction of water molecules around and the restriction of the carbon plate (Figure 3A). Figure 3B shows the trajectory of carboxyl in the absence and presence of THz EM stimuli. Evidently, the tendency of carboxyl to oscillate up and down has been suppressed after the application of 44 THz EM, confirming the conclusion of Figure 2C, and all vibration modes except C-O stretching have become weaker. Moreover, we can find that the oscillation period of carboxyl increases with the growth of the EM stimuli strength A (Figure 3C). Originally, it only takes 2.8 ns on average for carboxyl to complete a period of oscillating up and down, but as the stimuli strength A increases to 1.5 V/nm, the oscillation period increased to 8.6 ns on average. The dynamical change in carboxyl also changes the energy barrier for water molecules to permeate through the GO membrane. From Figure 3D, we can see the intermolecular interaction energy between the water molecule and the GO membrane. As the gateway for water molecules permeating through the membrane, carboxyl locates around 1.3 nm and 4.5 nm along the x-axis. Under normal circumstances, hydrophilic carboxyl will attract water molecules by forming H-bond with the interaction energy of E _w-c = -24.0 kJ/mol, making it difficult for water molecules to enter or exit from the inside of the membrane. However, with the application of 44.0 THz EM stimuli, the dynamic conformation of carboxyl changes and the interaction energy E _w-c reduces to -19.0 kJ/mol, making it less attractive to water and easier for water molecules to pass through. This fact leads to the conclusion that the movement change of carboxyl has a great influence on the membrane permeability due to its gateway location of water molecule transportation.

At last, it is worth noting that at the macroscopic level, the electric field required in an experiment to produce the dielectric breakdown of water is 0.07 V/nm. However, in classical molecular dynamics simulations, the electric field needed to affect the dynamic property of water significantly is almost 10 times greater than 0.07 V/nm. We did not observe the dielectric breakdown phenomenon in the simulation because the classical force field fixed the charge of atoms. This treatment just considered the orientation polarization and ignored the electron polarization. At the molecular level, dielectric breakdown is a bond-breaking process [51], and the electric field threshold to dissociate water molecules and sustain an ionic current needs to be up to 3.5 V/nm [52]. Therefore, under this threshold, we can reveal the classical image rather than the quantum image of the influence of the external field on molecular orientation and vibration dynamics. Furthermore, in the simulation, thermal dissipation is based on the thermostat algorithm, while the water bath is used to control the system’s temperature in the experiment. Although 44.0 THz is not the strongest absorption frequency band that water absorbs the EM stimulus, the local temperature will still rise if we irradiate the light on the water. Based on the continuous light source we applied in the simulation, the injected power from 44.0 THz EM stimulus to the unit volume of water is less than 0.6 × 10⁻⁹ W/nm³ when A < 1.5 V/nm (see details in Supplementary Section S5). Therefore, to investigate the non-thermal effect of the enhancement of membrane permeability caused by light, the water bath needs to be designed to take away the injected energy in the process of the experiment as shown in the previous work [35].

In this work, we found that EM stimuli at a frequency of 44 THz can be resonantly absorbed by carboxyl groups at the edge position of the GO membrane, thereby changing the vibration mode and dynamic conformation of carboxyl groups. As a result, it reduced the energy barrier of water molecule transport and ultimately increased the permeability of the membrane by five times. It is worth noting that in various biological and artificial particle transport channels and membranes, the edge functional groups play an important role in screening molecules or ions. Therefore, our findings bring new insights into the study of biological channels and artificial membranes and provide a new approach to manipulating the transmembrane transport of water molecules and other particles in life science and industrial applications.