The characterization of WSMWe designed a freezing-exfoliation method to afford a uniform WSM (Fig. 2a, left) containing the cuticle, epidermis and part of the hypodermis35, as confirmed by transmission electron microscopy (TEM) (Fig. 2a, right). This study investigated the functions of the three layers composing the WSM (Fig. 2b). The cuticle layer impedes the ion transport of the WSM. The epidermal layer, with high pectin content and closely woven cellulose36, exhibits excellent ion selectivity and mechanical properties. The hypodermis layer showcases remarkable ion transport performance. These specific experimental details are elaborated on later.Fig. 2: The structures and compositions of WSM.a Schematic of watermelon after application of the freezing-exfoliation method (left) and the TEM image (right) of the obtained WSM. b Illustration of the three layer structure of WSM and their functionality descriptions. c WSM obtained using the freezing-exfoliation method (thickness: 75 ± 5 µm). d Optical microscopic image of WSM. e Cross-sectional SEM image of freeze-dried WSM. f FTIR spectra of WSM-outer side, WSM1-outer side and Hypodermis@WSM. (g) Solid-state 13C NMR spectrum of WSM, and typical structures of cellulose/hemicellulose and pectin. h TEM image of the cell wall of WSM cells. i Magnified TEM image of the sample shown in h. j Magnified TEM image of the sample shown in i.The optical microscopic and scanning electron microscopy (SEM) images showcase the overall morphology of the WSM (Fig. 2c-e), which is 75 ± 5 µm thick and composed of well-assembled cell layers. The WSM was characterized using Fourier transform infrared spectroscopy (FTIR)37 and solid-state nuclear magnetic resonance (NMR)38. The characteristic peaks of these two spectra (Fig. 2f and Fig. 2g) indicate that the main components of WSM included cellulose/hemicellulose and pectin39. From Fig. 2f, compared to WSM, the two peaks (2920 cm−1 and 2846 cm−1) belong to the -CH2 antisymmetric and symmetric stretching vibrations respectively, originating from the lipid materials with long alkane chains that constitute the cuticle layer (Fig. S1a, b)40,41. However, the peaks aforementioned are significantly weakened in Hypodermis@WSM, while the other peaks show no apparent changes. This indicates that there’s no cuticle layer in the Hypodermis layer and SEM images substantiated this result (Fig. S1d, e). Atomic force microscope (AFM) topography further illustrates that the morphologies of the outer and inner sides of the watermelon skin are completely different (Fig. S2). The peaks at 1610 cm−1 and 3350 cm−1 have been assigned to carboxylate (COO-) and -OH respectively41 and the higher intensity of these peaks in the spectra of Hypodermis@WSM demonstrated that the cell wall contains more hydrophilic function groups than cuticle layer (Fig. S1c, f). The 13C chemical shifts at 170-180 ppm have been assigned to the carbonyl group (C=O) of COOH (176 ppm), acetyl (OCOCH3), and methyl ester (COOCH3) from hemicellulose and pectin. The shifts at 90-110 ppm are assigned to the ether group (R-O-R) from cellulose/hemicellulose and pectin39. The chemical structures of cellulose, hemicellulose and pectin are shown in the Fig. S3.Both the cell wall and the plasma membrane of plants could be the locations for the transport of water and ions (Fig. S4). The cell viability is identified by using propidium iodide (PI) as a fluorescent identification agent42. Red spots were detected inside the cells once a WSM prepared by the freezing-exfoliation method was treated with the PI solution in the dark for 20 minutes (Fig. S5), which indicates that the cells were dead, and plasma membranes were not integrated. Thus, the main ion-transport site of the WSM was the cell wall rather than the plasma membrane. The cell wall structure of WSM cells was characterized using SEM(Fig. 2h) and TEM(Fig. 2i, j). The cellulose fibers were arranged regularly, forming three-dimensional restricted channels with a diameter of 2-5 nm (Fig. S6), and pectin uniformly filled the regularly arranged three-dimensional fiber channels43. This unique structure limits the volume swelling caused by the high-water-uptake pectin, which allows the WSM to have both good water absorption and mechanical stability.The water uptake, mechanical strength, gas permeability, and ion conductivity are the main factors affecting the performance of ITMs. Hence, these physical parameters of WSM were investigated and compared with commercial benchmark AEMs, Fumasep FAB-PK-130 (Fumasep)44,45, Sustainion X37-50 Grade T (Sustainion)10,46 and quaternary ammonia poly(N-methyl-piperidine-co-p-terphenyl) (QAPPT)47. To accurately quantify the physical properties of the three layers of the WSM, detailed studies were conducted on three samples: WSM, WSM1, and Hypodermis@WSM (Fig. S7). Among them, WSM1 is a sample obtained by soaking the WSM sample in CHCl3 for 72 hours and dissolving a portion of its surface cuticle (Fig. S7d)48. The Hypodermis@WSM is a sample with a thickness of approximately 120 ± 5 µm, exfoliated from the watermelon.As shown in Table 1, the average water uptake of WSM, WSM1, and Hypodermis@WSM at room temperature is 470.9%, 482.1%, and 731.7%, respectively, all higher than the water uptake of Fumasep (11.3%), Sustainion (26.0%) and QAPPT (17.3%) under the same condition. WSM and WSM1 in a fully hydrated state exhibit good mechanical performance, with tensile stresses of 6.1 MPa and 4.7 MPa, and a strain at break of 13.3% and 14.1%, respectively, superior to Sustainion under the same conditions (5.9 MPa, 7.4%). Due to its woven structure, Fumasep is prone to local fracture during tensile deformation (Fig. S8 and S9) with a tensile stress of 13.8 MPa and a strain at break of only 4.5%. WSM and WSM1 maintain a balance between mechanical strength and water uptake, which is mainly attributed to the unique structural features of the cell wall: highly water-absorbent pectin uniformly fills the weak hydrophilic three-dimensional nanochannels composed of cellulose, thereby restricting excessive volume expansion. According to the literature36, it has been reported that the close and interlocking arrangement of cellulose in the cell wall structure of the epidermis layer contributes to its strong mechanical properties, which is consistent with the results of this experiment. Gas permeability is an important physical parameter for the membranes used for gas separation in electrolysis equipment, such as the anode and cathode gases. WSM and WSM1 have gas permeability similar to Fumasep, Sustainion and QAPPT for H2, O2, and CO2, indicating their excellent gas permeation resistance. This study tested the through-plane ionic conductivity of the membranes, with specific testing methods detailed in the supplementary information. To validate the universality of this work, ion conductivity tests were carried out on two types of watermelon skins with thicknesses of 75 ± 5 µm and 35 ± 5 µm respectively. The results demonstrated that both samples exhibited similar ion conductivities (Fig. S10). The ionic conductivity of WSM1 (101.2 mS cm−1) is approximately twice that of WSM (49.1 mS cm−1). The ionic conductivity of Hypodermis@WSM (282.3 mS cm−1) is approximately 5.5 times higher than that of WSM and also surpasses that of Fumasep (6.7 mS cm−1), Sustainion (94.1 mS cm−1) and QAPPT(69.3 mS cm−1) under the same condition. These results demonstrate that the structure of the epidermis and hypodermis plays an important role in the high ion transport characteristics of the WSM. Additionally, because it is not affected by the cuticle of the epidermis, the ionic conductivity of Hypodermis@WSM is closer to the actual ion conductivity of the cell wall, higher than the ionic conductivity of 1 M KOH water solution (195.1 mS cm−1) at room temperature49.Table 1 Characteristic physical parameters of the membranes analyzed in this studyBased on the above experimental data, the following conclusions can be drawn: (i) The ion conductivity of WSM1, with part of the cuticle layer removed, significantly increases, indicating that the presence of the cuticle layer impedes the ion transport of WSM; (ii) The unique structure of the epidermal layer gives it excellent mechanical properties; (iii) Hypodermis@WSM has an ion conductivity much higher than AEMs, indicating that the ion transport mechanism of cell walls is more efficient. Given the relatively similar mechanical properties and gas permeability of WSM and WSM1, but WSM1 exhibiting approximately double the ionic conductivity of WSM, our subsequent characterization and testing efforts within this study will primarily concentrate on WSM1.Ion-transport channels in the cell wallTo investigate the intrinsic reason for WSM1 which exhibits excellent ion conductivity, in-depth experimental verification was conducted. Figure 3a shows a schematic diagram of the two possible pathways for OH- transport in WSM1. Pathway 1 represents the transport of OH- along the cell wall, while pathway 2 describes how OH- can pass through the cell wall and the cell cavity filled with electrolytes. The ion conductivity of Hypodermis@WSM in 1 M KOH electrolyte (282.3 mS cm-1) is significantly higher than the intrinsic ion conductivity of 1 M KOH (195.1 mS cm-1), suggesting that the ion transport channels within the cell wall can accelerate the transport of OH-. To figure out the structure-activity relationship, both positron annihilation lifetime spectroscopy (PALS) and Brunauer-Emmett-Teller (BET) techniques were initially employed. The results (Fig. 3b, Fig. S11, S12) indicated that WSM1 possessed a microporous structure. According to previous reports, microporous polymers had a certain interchain gap size, which provided a low-hindrance transfer path for hydroxide7,9,50.Fig. 3: Ion-transport channels in the cell wall.a The schematic diagram of the transport pathway of OH- in WSM1. b Pore size distribution of WSM1 obtained from PALS. c Schematic of the Grotthuss mechanism: proton transfer mechanism (I); hydroxides transfer mechanism (II); Schematic of Newton’s cradle (III). d FTIR spectra of H2O, WSM1-H2O (WSM1 fully absorbed H2O, and the surface H2O was wiped off), and WSM1-dry (the dried WSM1). e FTIR spectra of the WSM1-dry, WSM1−1M-2h-dry (the WSM1 soaked in 1 M H2SO4 solutions for 2 h and then dried), and WSM1−1M-4h-dry (WSM1 soaked in 1 M H2SO4 solutions for 4 h and then dried). f Pore size distributions of the WSM1 and WSM1-1M-2h-dry obtained from the BET method. g FTIR spectra of H2O, WSM1-H2O, and WSM1-1M-2h-H2O (After soaking WSM1 in a 1 M H2SO4 solution for 2 hours, rinse it repeatedly with deionized water, dry it, and finally fully absorb H2O). h Ion conductivity of WSM1 after acid treatment. The error bars represent the standard deviation of ion conductivity.The transport of protons and hydroxides in the water system is controlled by the Grotthus mechanism, i.e., hydrogen-bonding network transport51,52. The construction of continuous hydrogen-bonding networks is beneficial for accelerating the transport of protons and hydroxides (Fig. 3cI, Fig. 3cII)53,54. For example, when protons are transmitted through a hydrogen-bonding network (Fig. 3cI), the proton bridging two hydrogen-bonded water molecules switch from one molecule to the other, while kicking an existing proton out of its adopted molecule and triggering a series of resembles displacements through the continuous hydrogen-bonding networks. This motion is similar to Newton’s cradle (Fig. 3cIII), where the associated local displacement leads to long-range proton transport.Since the hydroxyl and carboxyl groups in cellulose and pectin are hydrophilic, water molecules are readily confined within the micropores constructed by cellulose and pectin. Additionally, hydroxyl and carboxyl groups are proton donors. Concentrated hydrated hydrogen ions can be inferred to induce the rearrangement of water molecules of enclosed water near hydroxyl and carboxyl groups, thereby promoting the formation of a continuous hydrogen-bonding network55 and accelerating the proton/ hydroxides transport rate.FTIR was used to investigate whether the WSM1 contained continuous hydrogen-bonding networks. As shown in Fig. S13, in the water system, H2O molecules have two distinct characteristic vibration peaks at 1638 cm-1 and 3300 cm-1, which are attributed to the bending vibration of the O-H bond and the stretching vibration of the O-H bond, respectively55. Due to the large amount of -OH in the WSM1-dry, the characteristic vibration peak at 3300 cm-1 coincides with the 3300 cm-1 peak in the H2O system and has a wider peak width (Fig. S14. Thus, to explore whether continuous hydrogen-bonding networks were formed, the offset of the bending vibration peaks at 1638 cm-1 of the O-H bond in three samples of H2O, WSM1-dry and WSM1-H2O were chosen for comparison. As shown in Fig. 3d, a characteristic vibration peak was found at 1612 cm-1 in the WSM1-H2O system, with a redshift relative to the bending vibration peak of the O-H bond in the H2O system. This redshift was due to the increased stiffness of the O-H bond bending mode caused by the electrostatic force between H and the acceptor atoms after the formation of continuous hydrogen-bonding networks56. Therefore, continuous hydrogen-bonding networks formed in the WSM1, which is beneficial for accelerating H+/OH- transport.Since the main components of WSM1 are cellulose and pectin, we prepared cellulose and pectin membranes using commercially available cellulose and pectin, respectively. After water absorption, the bending vibration frequency of the O-H bond in these two types of membranes did not show a redshift compared to the pure water system (Fig. S15a, b), which indicated that continuous hydrogen-bonding networks were not formed in these two membrane structures57. Therefore, disordered hydroxyl-rich polymer structures have difficulty forming continuous hydrogen-bonding networks.To investigate the role of pectin in the formation of micropores and continuous hydrogen-bonding networks in the WSM1, we performed pectin removal treatment on the WSM1. According to the literature41, the dissolution of pectin can be accelerated in an acidic medium at 60 °C. Therefore, we soaked WSM1s in 1 M H2SO4 solutions and placed them in a 60 °C oven for 2 (Figs. S16) or 4 hours. Figure 3e shows the FTIR spectra of the WSM1s before and after H2SO4 treatment. According to the spectrum, the intensity of the peak at 1610 cm-1, which is the characteristic vibration peak of -COO- in pectin41, significantly decreased. Solid-state 13C NMR spectra of WSM1 and WSM1-1 M-2 h also demonstrated that the signal of pectin has decreased after H2SO4 treatment (Fig. S17). This suggests that the pectin content of the WSM1 decreased after acid treatment. The microporous volume of the WSM1 after pectin dissolution increased (Fig. 3f), and their adsorption capacity for CO2 gas significantly decreased (Fig. S18). Through FTIR characterization, it was observed that WSM1-1 M-2 h exhibited a redshift decrease in the peak at 1638 cm-1 relative to untreated WSM1, as illustrated in Fig. 3g. This indicated that after pectin dissolved, the hydrogen-bonding strength weakened. At the same time, the dissolution of pectin also caused a significant decrease in the ion conductivity of the WSM1 (Fig. 3h). The above experiments indicate that (i) pectin plays an important role in the formation of micropores and continuous hydrogen-bonding networks in cell walls and (ii) micropores and continuous hydrogen-bonding networks constructed by the synergistic effect of pectin and cellulose are crucial for improving ion-transport efficiency.Ion selective permeation of WSM1A prominent issue arises from the crossover of anionic CO2 reduction products, such as formic acid, acetic acid, and propionic acid, occurring through the electromigration process within the AEMs10,25. An ideal ITM used for CO2RR should have excellent ion selectivity. Here, we selected five common liquid products from the CO2RR for ion selectivity testing of WSM1. The electrochemical measurement system for the CO2RR liquid product crossover test of WSM1 is shown in Fig. 4a. Five common liquid products of the CO2RR58,59, i.e., formic acid, acetic acid, propionic acid, ethanol, and propanol, were added to the catholyte, and the system was operated for 5 hours with no CO2 reactant at a current density of 200 mA cm-2. The catholyte was collected at different intervals and analyzed using 1H NMR.Fig. 4: Ion selective permeation of WSM1.a Schematic diagram of the liquid product crossover test device. b–f Remaining ratios of formate, acetate, propionate, ethanol, and propanol in the catholyte when WSM, WSM1, Hypodermis@WSM, WSM1-1 M-2 h (the WSM1 soaked in 1 M H2SO4 solutions for 2 h), Fumasep, Sustainion or QAPPT was used as the ITM. The error bars represent the standard deviation of the remaining ratios. g Pectin-containing model diagram for molecular dynamic simulation. h Position of formate in pectin-containing systems as a function of the simulation time. i Position of formate in a pectin-free system as a function of the simulation time. j Mechanism diagram of selective permeation of WSM1.As shown in Fig. 4b–f, when Fumasep/Sustainion/QAPPT was used as the ITM, the crossover rates of the formate, acetate, propionate, ethanol, and propanol were 64.0%/47.4%/58.75%, 42.3%/31.8%/42.99%, 17.7%/30.4%/42.43%, 6.8%/26.4%/21.28%, and 7.2%/24.8%/24.38%, respectively. In contrast, when WSM1 was used as the ITM, the crossover rates of the same liquid products were 5.0%, 5.6%, 8.9%, 6.1%, and 1.6%, respectively. The crossover rate of formate for WSM1 was approximately 12.8/9.6/11.8 times lower than that for Fumasep/Sustainion/QAPPT when used under the same conditions. Thus, the WSM1 exhibited excellent ion selectivity compared to Fumasep, Sustainion, and QAPPT.A detailed comparison was conducted among WSM, WSM1, and Hypodermis@WSM regarding ion selectivity. As shown in Fig. 4b-f, WSM, WSM1, and Hypodermis@WSM exhibited excellent ion selectivity for common ion products in CO2RR. However, Hypodermis@WSM displayed a slightly lower ion selectivity compared to the other two. The epidermis layer exhibits the most exceptional ion selectivity. We conducted a detailed analysis of the cellulose and pectin content in WSM1 and Hypodermis@WSM. The mass ratio of pectin to cellulose in WSM1 was 0.067, while in Hypodermis@WSM, it was 0.024. This indicates that the mass ratio of pectin to cellulose in the epidermal layer is greater than 0.067. Therefore, among the three layers of the watermelon skin, the epidermal layer has the highest pectin content. Because pectin contains an abundance of negatively charged groups, it has the ability to repel acid anions, resulting in excellent ion selectivity within the epidermal layer.With WSM1-1 M-2 h (with partial pectin removed) employed as the ITM, the crossover rates of formate, acetate, propionate, ethanol, and propanol were higher than those of the untreated WSM1 (Fig. 4b–f). This indicates that the partial removal of pectin reduces the ion selectivity of WSM1. To verify the inhibitory effect of negative charge groups on acid anions, we chose Nafion115 PEM, which has a large number of sulfonic acid groups, for ion-selective permeation experiments. The experimental results are shown in Fig. S19. Compared with Fumasep, Nafion115 has a lower permeation rate of acid anions, indicating that negative charge groups can inhibit the permeation of acid anions. In addition, we also found that WSM1 has better ion selectivity than Nafion115. Accordingly, there are two main reasons why the WSM1 exhibits good ion selectivity: (i) the carboxyl groups in pectin are negatively charged and thus repel negatively charged ions such as HCOO- and CH3COO- and (ii) although the many hydroxyl groups in the WSM system bind with OH- and acid ions through hydrogen bonding, OH- could be transferred through the continuous hydrogen-bonding networks, whereas acid ions could not.To further theoretically demonstrate the above conclusion, we conducted molecular dynamics simulations on the WSM1 system. Two models were constructed: one with cellulose channels filled with pectin (Fig. 4g) and another with cellulose channels without pectin (Fig. S20). Both models included a formate molecule within the channel. Molecular dynamics simulations were employed to investigate the migration pathway of formate60,61,62. In the system containing pectin, formate was initially located inside the simulated channel, while the center of mass of the entire simulated unit was located at the origin. The channel spanned approximately 95 Å in total length along the x-axis. Therefore, if the x-coordinate of the formate molecule exceeded the range of −47 to 47 Å, the formate molecule entered the adjacent unit. As shown in Fig. 4h, in the pectin-containing model, the formate molecule oscillated back and forth at varying speeds throughout the entire 400 ns simulation time. The start and end positions of the formate molecule in the simulation were both within the original simulation unit. In the pectin-free model system (Fig. 4i), the formate molecule experienced a significant shift in the negative field strength direction. The formate molecule tended to move vertically toward the cellulose channel wall and interact with the hydroxyl group of the sugar ring through hydrogen bonding, which restricted its movement. However, the interaction between the formate molecule and the hydroxyl group of the sugar ring was not very strong, so the formate molecule was easily detached and re-incorporated into the solvent. The formate continued to move under the influence of the electric field. The above results indicate that pectin rich in -COO- functional groups repelled HCOO- and reduced its transmission. Thus, the calculation simulation yielded consistent results with the experimental findings.From the above experimental and modeling results, we formulated the ion transport mechanism in the WSM1: OH- is accelerated through the continuous hydrogen-bonding networks and microporous channels, while acid ions are repelled by negatively charged groups in pectin and form certain hydrogen bonds with hydroxyl groups, making penetration through the WSM1 difficult (Fig. 4j). This is the fundamental reason why WSM1 has both high ion conductivity and excellent ion selectivity.Performance of the CO2RR system with WSM1The high ion conductivity and low CO2RR liquid product crossover behavior of WSM1 make it particularly suitable for CO2RR systems. Thus, we evaluated the performance of the WSM1 for the CO2RR utilizing a currently widely used electrochemical flow cell system (Fig. 5a, Fig. S21, S22) in which the cathode catalyst was previously reported to be CoPc63, while NiFeOOH64 was used as the anode catalyst. The cell voltages for CO production applied on the two electrodes at different current densities were measured without iR compensation. Online gas chromatography was used to detect and quantify the evolution of the electrolytic products CO over a current density range of 100-500 mA cm-2. When 1 M KOH was used as the electrolyte (Fig. 5b), the average cell voltages of WSM1 at 100 mA cm-2, 200 mA cm-2, and 500 mA cm-2 are 2.30 V, 2.56 V, and 2.98 V, respectively, significantly lower than Fumasep but comparable to Sustainion / QAPPT. The average cell voltages of Hypodermis@WSM at 100 mA cm-2, 200 mA cm-2, and 500 mA cm-2 are 2.12 V, 2.25 V, and 2.55 V, respectively, significantly lower than Sustainion, QAPPT and Fumasep. Furthermore, at current densities ranging from 100-500 mA cm-2, both WSM1 and Hypodermis@WSM maintain a CO Faradaic efficiency (FECO) of approximately 90%. The device that employs WSM1 as the ISM can run stably for approximately 115 hours (RT, Fig. 5c) at a current density of 100 mA cm-2 and with 1 M KOH as the electrolyte. These data indicate that the ion transport channels in the cell wall of watermelon skin have remarkable ion transport capabilities and can serve as a benchmark for artificial biomimetic membranes.Fig. 5: The performance of the CO2RR system with WSM1 as ISM.a Schematic diagram of the CO2 reduction device. b Applied cell voltage and CO2RR selectivity for CO formation as a function of the current density with different membranes using 1 M KOH as electrolyte. Cathode catalyst: 2.0 mg/cm² CoPc loading on PTFE/Cu; anode catalyst: NiFeOOH. The error bars represent the standard deviation of Ecell and FECO. c Temporal stability of the CO2RR system with WSM1 as the ISM at 100 mA cm-2 using 1 M KOH as electrolyte. Cathode catalyst: 2.0 mg/cm² CoPc loading on PTFE/Cu; anode catalyst: NiFeOOH. d The cell voltage curve recorded for the CO2RR system with WSM1 as ISM operating at a current density of 200 mA cm-2. e FE of the CO2RR system with WSM1 as the ISM at a current density of 200 mA cm-2 using 1 M KOH as electrolyte. Cathode catalyst: PTFE/Cu; anode catalyst: NiFeOOH. The error bars represent the standard deviation of FE.Furthermore, the FE was evaluated at a current density of 200 mA cm-2 (Fig. 5d), and was monitored at the first, third, and fifth hours during the reaction with copper serving as a cathode catalyst (Fig. S24). The system comprised both gaseous and liquid products, and the total FE was maintained at ~95% for over 5 hours (Fig. 5e), which indicates that the application of WSM1 in the CO2RR system is beneficial for collecting cathode products and objectively evaluating the performance of cathode catalysts.
