Covalent triazine framework membranes with tunable pore chemistry
Covalent triazine framework chemistry gives rise to a wide variety of microporous materials and offers enormous diversity in pore chemistry. We thus synthesized a stand-alone triazine framework membrane from 4,4'-biphenyldicarbonitrile and a derivative of 3-hydroxy-[1,1'-biphenyl]-4,4'-dicarbonitrile bearing a quaternary ammonium moiety via a superacid-catalyzed organic sol‒gel procedure (Fig. 1a and Supplementary Figures S1-S4) (26). The process yields a free-standing membrane (namely, QCTF) with a Young’s modulus and tensile strength of 0.91 GPa and 32.0 MPa, respectively (Supplementary Figure S5). The skeletal triazine rings of QCTF were subsequently protonated with HCl or methylated with CH3I, affording P-QCTF and M-QCTF, respectively. Overall, we constructed three covalent triazine framework polymers with similar molecular configurations and pore structures that can be processed into hydrophilic, uniform and robust ion-selective membranes via an organo-sol‒gel procedure (Supplementary Figures S6-S8, Supplementary Table S4), but with slightly different and deliberately tailored pore chemistries.
Carbon dioxide (CO2) adsorption experiments and molecular simulations were conducted to probe the micropore structure of the covalent triazine framework polymers. CO2 sorption isotherms measured at 273 K revealed that powder samples of QCTF, P-QCTF, and M-QCTF had similar CO2 uptake capacities of 16, 15.2, and 14.7 cm3 g− 1 STP, respectively (Fig. 1b). Notably, QCTF, P-QCTF, and M-QCTF exhibit almost identical pore size distributions, ranging from 0.3 nm to 0.9 nm, as derived from CO2 adsorption isotherms based on density functional theory (DFT) calculations (Fig. 1c). These experimental results are further supported by molecular simulations of the 3D framework structure and the computation of CO2 distributions within the framework structures (Supplementary Figures S9 and S10). This again indicates that QCTF, P-QCTF, and M-QCTF have similar framework structures, interconnected micropores and pore size distributions.
The amount of charged functional groups (quaternary ammonium groups) within the pristine QCTF membrane, characterized by the ion exchange capacity (IEC, in mmol g− 1), is 1.20 mmol g− 1 for QCTF (as-designed IEC value is ~ 1.00 mmol g− 1). During protonation, approximately 55% of the triazine rings were protonated and the same amount of triazine rings was methylated after methylation, as revealed by X-ray photoelectron spectroscopy (XPS, Fig. 1d). This suggests that P-QCTF and M-QCTF should have identical IEC values, which was confirmed by titration and zeta potential measurements (Supplementary Figure S11).
Ion Transport and Selectivity
Despite the similar framework structure and almost identical pore size/size distributions, our experimental results reflect that cross-membrane ion transport is significantly affected by pore chemistry. We speculate that the difference is synergistically determined by Coulombic/steric effects and specific ion‒pore wall interactions, as shown in Fig. 2a. The current‒voltage (I‒V) curves across the membranes, as measured in a two-compartment diffusional H-cell under a 10-fold concentration gradient KCl solution (Fig. 2b), reveal a net anion flux, indicating anion selectivity. The anion transference number (t−) calculated for QCTF is 0.940, while the values for protonated QCTF (P-QCTF) and methylated QCTF (M-QCTF) are 0.947 and 0.953, respectively (Supplementary Figure S12). These values suggest the superior anion selectivity of the QCTF membranes compared to that of commercial anion exchange membranes (AEMs). This result is reasonable considering the Coulombic repulsion of the < 1 nm pore channel within the QCTF membranes.
The measured transference numbers align with the cross-membrane permeation/diffusion rates for BTMAP-Vi (a redox-active organic cation) and Cl− (Fig. 2c and 2d, Supplementary Figures S13-S15, Supplementary Tables S5-S6), which are dramatically different in size. Compared with commercial AEMs (Fig. 2c), all the QCTF membranes exhibited superior blocking capabilities toward BTMAP-Vi. The diffusion coefficients of BTMAP-Vi across the QCTF and the P-QCTF were determined to be 4.5×10− 11 cm2 s− 1 and 3.4×10− 11 cm2 s− 1, respectively. These values are at least one order of magnitude smaller than those of commercial AEMs. Note that the value further decreases to 3.1×10− 11 cm2 s− 1 for M-QCTF, a value that is over 20 times smaller than that of Selemion® DSV. The diffusion coefficients of Cl− through the QCTF and P-QCTF are 1.8×10− 7 cm2 s− 1 and 2.6×10− 7 cm2 s− 1, respectively. By contrast, commercial anion-selective membranes demonstrated Cl− diffusion coefficients at least one order of magnitude smaller than those of QCTF membranes. Surprisingly, the Cl− diffusion coefficient measured for M-QCTF reached 3.0×10− 7 cm2 s− 1, which is nearly 2 times that for the QCTF membrane (Fig. 2d). A comparison of the Cl− diffusion coefficients and the Cl−/BTMAP-Vi selectivity for QCTF membranes, commercial AEMs and previously reported membranes implies that these framework membranes can simultaneously deliver fast ion permeation and high selectivity, overcoming the usual tradeoff observed for many ion exchange membranes (Supplementary Figure S16 and Supplementary Table S6).
The fast Cl− transport across the triazine framework membranes is further supported by the membrane conductivity measurements. Compared with commercial AEMs, triazine framework membranes show high Cl− conductivity at relatively low hydration numbers (Fig. 2e, Supplementary Figure S17 and Supplementary Tables S7-S8). The Cl− conductivity of QCTF, as measured by four-point electrochemical impedance spectroscopy (EIS), is 13.2 mS cm− 1 at 30.0°C and approaches 42.0 mS cm− 1 at 80°C at low hydration numbers (3.5 at 30°C, 4.4 at 80°C). In comparison, the Cl− conductivity of P-QCTF is 20.0 mS cm− 1 at 30°C and increases to 48.4 mS cm− 1 at 80°C. We find that the Cl− conductivity of M-QCTF is 26.0 at 30.0°C, which is nearly twice that of QCTF, and reaches 53.0 mS cm− 1 at 80°C. The activation energy (Ea) for Cl− conduction across the QCTF membrane is 20.6 kJ mol− 1, as derived from the conductivities at various temperatures (Fig. 2f and Supplementary Figure S18), contrasting an Ea of 12.9 kJ mol− 1 for K+ transport across an otherwise identical membrane with sulfonate functional groups (ref 14). Surprisingly, the Ea value for M-QCTF is as low as 13.1 kJ mol− 1, which is nearly half that of QCTF and lower than any value reported in the literature (Fig. 2g and Supplementary Table S1). Considering the similar framework structure and almost identical pore size/size distributions, this significant result indicates that the methylation of triazine rings alters the transport energy barrier for Cl− ions.
Due to the aforementioned results, we conclude that electrostatic interactions alone cannot explain the differences in Cl− diffusion coefficients, Cl− conductivity or activation energy for cross-membrane Cl− transport. To unravel why methylation of the triazine ring promotes fast Cl− conduction, compared to the protonated triazine ring in P-QCTF and the charge-neutral triazine ring in QCTF, the charge distribution and the Cl− transport routes within the matrix of the triazine framework membranes were portrayed based on molecular simulations, and the two-dimensional free-energy landscapes were computed according to current methodology (13, 14). Our calculations show that the charge distributions of triazine framework membranes vary dramatically after protonation and methylation (Fig. 3a, Supplementary Figure S19). The most even charge distribution is observed for M-QCTF. We speculate that the variation in charge distribution alters the interactions between anions and the membrane frameworks and helps establish low-energy-barrier pathways for anion transport. This is supported by free energy calculations for Cl− conduction (Fig. 3b). The simulation results showed that Cl− can interact with quaternary ammonium (QA) groups (Fig. 3c, Supplementary Figures S20 and S21) and lower the free energy, but an energy barrier must be overcome for Cl− ions to approach adjacent QA groups. The energy barrier for Cl− conduction is the highest for QCTF (Fig. 3b, left panel) and decreases when the triazine ring is protonated (Fig. 3b, middle panel), while methylation of the triazine ring in M-QCTF improves the diffusivity of Cl− within the framework and creates a Cl− diffusion pathway with the lowest energy barrier (Fig. 3b, right panel). We suspect that the synergy of electrostatic interactions between Cl− and the methylated triazine ring and the change in electron density along the Cl− diffusion path after methylation may account for the emergence of the low-energy-barrier diffusion pathway.
Molecular simulation results are further supported by measurements of transmembrane F− diffusion coefficients via 19F pulsed-field gradient-stimulated-echo nuclear magnetic resonance (19F PFG-NMR; 19F was selected owing to its higher sensitivity compared with 35Cl).19F PFG-NMR revealed two separate F− signals for Selemion® DSV and Selemion® AMV membranes (Fig. 3d and Supplementary Figure S22), with the upfield signal corresponding to free F− in water (located at the same position as that in 0.1 M KF aqueous solution) and the downfield signal corresponding to associated F− within the membrane. In contrast, only the upfield signal was observed for all three triazine framework membranes (Fig. 3d), which is an indication of freely exchangeable F− within the membrane, with slight variations in the 19F chemical shifts. By fitting the echo profiles with the Stejskal‒Tanner equation (Supplementary Figure S23), the derived F− diffusion coefficients within the P-QCTF and QCTF are 0.93×10− 9 m2 s− 1 and 0.63×10− 9 m2 s− 1, respectively (Fig. 3e). The value reaches 1.1×10− 9 m2 s− 1 for M-QCTF, almost a twofold increase compared to that for QCTF. Notably, this value is 12.8 times that of Selemion® AMV and 10.8 times that of Selemion® DSV (Fig. 3e and Supplementary Figure S23) and approaches the measured diffusion coefficient of F− in water (1.2×10− 9 m2 s− 1; Supplementary Figure S23). In summary, by tailoring the pore chemistry of framework membranes, intimate ion‒pore wall interactions provide a low-energy-barrier diffusion pathway for anions. Taken together with the Coulombic/steric exclusion by the charged framework micropores, the triazine framework membranes, particularly M-QCTF, will be of interest in applications demanding extremely fast and highly selective transport of anions.
Triazine framework membrane powers fast-charging AORFBs
The extremely fast and highly selective anion (particularly chloride ions) conduction through chemically tuned triazine framework membranes is desirable in electrochemical devices, such as aqueous organic redox flow batteries. As a proof of concept, we configured pH-neutral AORFBs with BTMAP-Vi/FcNCl as the redox-active organic electrolyte couple and triazine framework membranes as the ion-conducting membranes, while Cl− ions were transported back and forth as charge carriers (Fig. 4a). At an electrolyte concentration of 0.1 M, EIS of the BTMAP-Vi/FcNCl cells assembled with QCTF or P-QCTF showed area-specific membrane resistances (ASRs) of 0.63 Ω cm2 and 0.53 Ω cm2, respectively (Supplementary Figures S24-S25). An otherwise identical cell assembled with M-QCTF showed an ASR of 0.37 Ω cm2 (Supplementary Figure S26), which is almost twofold lower than that of the QCTF membrane. This finding aligns with the high conductivity of M-QCTF (Fig. 2e, 3b), which enables charging of the BTMAP-Vi/FcNCl cells at extreme current densities. For example, at 200 mA cm− 2, BTMAP-Vi/FcNCl with M-QCTF exhibited an energy efficiency (EE) of over 60% (Supplementary Figure S26). In contrast, the control BTMAP-Vi/FcNCl cells assembled with Selemion® DSV or Selemion® AMV could not operate at this current density due to the immediate voltage cutoff. At lower current densities ranging from 20 to 80 mA cm− 2, the reported energy efficiency for the control cells drops from 89.4–65.9% for Selemion® DSV or from 80.0–26.6% for Selemion® AMV (27).
At a higher electrolyte concentration of 0.5 M, BTMAP-Vi/FcNCl with M-QCTF demonstrated an even lower ASR of 0.23 Ω cm2 (Fig. 4b), a much lower value than that for Selemion® DSV or Selemion® AMV. The rate performance of the cell reveals an EE of 49.7% and a capacity utilization of 58.8% at an extreme current density of 500 mA cm− 2 (Fig. 4c). Compared with the most recent report of an AEM (MTCP-50 membrane, with the optimal ratio 1:1 of m-terphenyl to p-terphenyl) for pH-neutral AORFBs at 0.5 M (21), M-QCTF achieved a much greater energy efficiency (76.9% vs. 60.1%) and capacity utilization (94.3% vs. 63.7%) at the same current density of 200 mA cm− 2. Notably, alkaline AORFBs that utilize K+ as charge-carrying ions assembled with a cation exchange membrane (SCTF-BP), which allows cation diffusion close to the value in bulk electrolyte, exhibit an EE of 50.4% and a capacity utilization of 62% at 500 mA cm− 2. The current results demonstrate a similar efficiency for Cl− transport and therefore suggest a breakthrough in the charge asymmetry effect.
Robust and exceptional cell performance was observed during long-term galvanostatic cycling of over 2000 cycles at 200 mA cm− 2 (0.1 M electrolyte concentration, Supplementary Figure S26) and over 1000 cycles at 400 mA cm− 2 (0.5 M electrolyte concentration, Fig. 4d). Comparisons of the EE and capacity utilization against the current density shows consistently superior battery performance over multiple cell cycling experiments for the BTMAP-Vi/FcNCl cells with M-QCTF, compared to the pH-neutral AORFB with different membranes (Fig. 4e, 4f and Supplementary Table S10).
This work demonstrates that chloride and fluoride anions traverse the M-QCTF membrane with a very low energy barrier, leading to exceptional flow battery performance. This significant development can be applied more broadly to designing anion exchange membranes for other technologies such as CO2 electrolysers (28) and ion-capture electrodialysis (29). Although the anion diffusion constants within the developed membranes are approaching the theoretical limit of the bulk electrolyte solution, we expect further improvements in overall conductivity to be achievable by eliminating micropore tortuosity and creating perfectly aligned micropore channels with monodispersed pore size distributions.