Long‐Life Aqueous Organic Redox Flow Batteries Enabled by Amidoxime‐Functionalized Ion‐Selective Polymer Membranes

Abstract Redox flow batteries (RFBs) based on aqueous organic electrolytes are a promising technology for safe and cost‐effective large‐scale electrical energy storage. Membrane separators are a key component in RFBs, allowing fast conduction of charge‐carrier ions but minimizing the cross‐over of redox‐active species. Here, we report the molecular engineering of amidoxime‐functionalized Polymers of Intrinsic Microporosity (AO‐PIMs) by tuning their polymer chain topology and pore architecture to optimize membrane ion transport functions. AO‐PIM membranes are integrated with three emerging aqueous organic flow battery chemistries, and the synergetic integration of ion‐selective membranes with molecular engineered organic molecules in neutral‐pH electrolytes leads to significantly enhanced cycling stability.


1
) and operated at a flow rate of 1 mL min -1 and 60 °C. Molecular weight and polydispersity index (PDI) was calculated based on poly(methyl methacrylate) standards. Thermogravimetric analysis (TGA). TGA traces of polymer powder were obtained by a NETZSCH STA 449 F5 Jupiter thermogravimetric analyser at a heating rate of 10 °C min -1 from room temperature to 800 °C under a nitrogen atmosphere. Skeletal density test. Skeletal density test was performed by a Micromeritics Accupyc II 1340 helium pycnometer with a 3.5 cm 3 sample chamber at 25 °C. AO-PIM powders were degassed overnight at 110 °C under vacuum prior to tests. The mean values and standard deviation were calculated based on 10 measurements. Tensile test. Tensile test was conducted on a Lloyd-Ametek EZ50 Material Testing Machine at ~50% relative humidity and room temperature. Membrane specimens with the width of ~5 mm and effective length of ~30 mm were strained at a rate of 3 mm min -1 . Average values and standard deviation were calculated on 3-5 specimens. Nitrogen/carbon dioxide physisorption. Low-pressure gas physisorption were measured by a Micromeritics 3Flex surface characterization analyser with N 2 isotherms at 77 K and CO 2 physisorption at 273 K. Powder samples (~100 mg) were degassed at 110 °C under high vacuum for 12 h followed by loading into the instrument and in situ degassed at 110 °C for another 12 h prior to measurements. Dynamic water vapour sorption (DVS). Dynamic water vapour sorption was measured on a dynamic vapour sorption Endeavour gravimetric sorption analyser (Surface Measurement Systems Ltd.) with the relative humidity in the range of 0 -90%. Membrane samples (20-30 mg) were dried at 110 °C under vacuum for 12 h followed by loading into the instrument and in situ dried at 25 °C under flowing dry air over 24 h until the mass keeping constant. Water vapour uptake isotherms were obtained by measuring sample mass change in dry to fully hydrated state at a specified relative humidity. Small-angle X-ray scattering (SAXS)/wide-angle X-ray scattering (WAXS). SAXS and WAXS experiments were carried out on a Ganesha 300XL (SAXSLAB) instrument at the EPSRC CNIE research facility at University College London by employing a high brilliance microfocus Cu source (λ = 1.54 Å). SAXS and WAXS patterns were recorded using a Pilatus 300 K solid-state photon-counting detector with a 2 mm beam stop with a sample-to-detector distance of 1041 mm (SAXS) and of 101 mm (WAXS), respectively. The beam center and the sample-to-detector distance were calibrated using the position of diffraction peaks from a standard silver behenate powder. SAXS and WAXS patterns were radially averaged around the direct beam position using SAXSGUI software.

Experimental Measurements
Water/electrolyte uptake measurement Water/electrolyte uptake was measured by mass changes of AO-PIM membrane samples in dry and fully hydrated conditions. Membrane samples were dried at 110 °C under vacuum overnight to obtain the dry mass (W dry ). These samples were immersed in deionized water, 1M aqueous KOH and 1M aqueous KCl in sequence by using the same membranes for 24 h at room temperature. Before changing the liquid, the wet mass of membranes was measured immediately after wiping off excess liquid from the surface (W wet ). The water or aqueous electrolyte uptake (WU/EU) was calculated by: The mean values and error bars are based on 3 measurements using 3 different membrane samples.

Swelling ratio measurement
The swelling ratio (SR) was calculated from the difference in linear dimensions between the wet (x wet ) and dry (x dry ) AO-PIM membranes according to: Error bars are standard deviations derived from 3 measurements based on 3 different membrane specimens.

Ionic conductivity measurement
Ionic conductivity of AO-PIM membranes was performed on electrochemical impedance spectroscopy (EIS) using the potentiostat mode at an AC bias of 10 mV and a frequency range from 0.2 MHz to 10 Hz. For apparent ionic conductivity tests, membrane samples were pretreated in 1M aqueous KOH overnight to fully deprotonate hydroxyl groups (i.e., to convert -OH form into -Oform), followed by soaking in deionized H 2 O for 24 h to remove residual KOH, then equilibrated in deionized H 2 O, 1M KOH or 1M KCl aqueous electrolyte (termed as a KCl conductivity) for 24 h. Membrane samples in non-deprotonated form (i.e., -OH form) were directly soaked in 1M aqueous KCl for 24 h for the intrinsic ionic conductivity tests (termed as b KCl conductivity). Membrane samples were sandwiched between two stainless steel electrodes and sealed with coin cells (Type 2032). The assembly procedure was carried out in deionized H 2 O or KCl/KOH electrolytes to avoid air bubbles being trapped in cells. The ionic conductivity in the range of 30 to 80 °C was calculated from the resistance [5] by: where σ is the membrane ionic conductivity in S cm -1 , L is the membrane thickness in cm, A is the membrane active area of 2.00 cm 2 , and R m is the membrane resistance in Ω. The resistance of blank coin cell (0.04 Ω) was subtracted from R m before applying Equation S3.

Diffusion and Crossover measurement
Ion diffusion and redox molecule crossover tests were performed using concentration-driven dialysis diffusion H-shaped cells. AO-PIM membranes were sandwiched between two chemically resistant O-rings and secured in the middle of H-cells by clips. Continuous stirring was used in both feed and permeate sides to alleviate the concentration polarization near membranes. In ion diffusion dialysis tests, 50-µm-thick AO-PIM membranes in non-deprotonation, neutral-charge form (i.e., -OH) were first applied to evaluate the permeation rates of common salts to investigate their ion-sieving performance. Aqueous electrolytes (i.e., 1 M KCl, NaCl, LiCl, CaCl 2 and MgCl 2 , 50 mL) was used as a feed solution and deionized H 2 O (50 mL) was used in the permeate side. Besides, AO-PIM membranes in deprotonation, negative-charge form (i.e., -O -) were then used to measure the permeation rates of KOH (1M, 50 mL, pH = 14.0) and KCl (1M, 50 mL, pH = 9.0) across the membranes. The ionic conductivity in the permeate side was continuously recorded by a conductivity meter (Orion Star A210, Thermo Scientific). The concentration change of aqueous electrolytes over time was obtained from the corresponding electrolyte calibration curves according to the linear relationship between the ionic conductivity and electrolyte concentration. The mean values and error bars of ion permeation rate were derived from 3 individual measurements based on 3 different samples.
In redox molecule crossover tests, AO-PIM membranes were pretreated in 1M KOH to obtain their deprotonated form. In alkaline system, K 4 Fe(CN) 6 (0.1M) or 2,6-DHAQ (0.1M) in aqueous KOH solution (1M, 50 mL, pH = 14.0) was used as a feed solution and blank aqueous KOH solution (1M, 50 mL, pH = 14.0) was used in the permeate side. In near neutral-pH system, K 4 Fe(CN) 6 (0.1M) or 2,6-DPPAQ (0.1M) in KCl aqueous solution (1M, 50 mL, pH = 9.0) was used as a feed solution and blank KCl aqueous solution (1M, 50 mL, pH = 9.0) was used in the permeate side. The concentration change of K 4 Fe(CN) 6 in the permeate side was quantitatively detected by inductively coupled plasma-optical emission spectrometry (ICP-OES). Permeate aliquots were collected three times per day and diluted with 2.0 wt.% HNO 3 (100 or 50 times) prior to analysis. The concentration changes of 2,6-DHAQ and 2,6-DPPAQ in the permeate side was quantitatively detected by the calibrated UV-Vis spectrometer. Permeate aliquots were analysed without dilution and recycled to permeate side. The permeation rate of ions and redox molecules across the membrane over a short-time period without volume change can be determined by Fick's first law: where J is the permeation rate in mol cm -2 s -1 , V is the solution volume of 50 mL, A is the membrane active area of 1.54 cm 2 , C is permeate concentration in mol cm -3 , and t is diffusion time in s. Over the process, the difference between permeate and feed concentration can be ignored (i.e., C 1 -C 2 = C 1 ), and the flux can be assumed to be constant. Consequently, Fick's first law can be simplified as: where D is the diffusivity in cm 2 s -1 , C 1 is feed concentration in mol cm -3 , and l is the membrane thickness in cm.

Flow battery measurement
The flow cell was assembled with a cell fixture (Scribner Associates) and POCO single serpentine pattern graphite plates. A piece of ~50-µm-thick membrane (active area = 7 cm 2 ) was sandwiched between electrodes comprising a stack of carbon paper (three sheets, SGL) in each side. The rest space between graphite plates was sealed by Viton gasket. Electrolytes were fed into the cell at a flow rate of 100 mL min -1 by a Cole-Parmer peristaltic pump. Carbon paper was pretreated by baking at 400 °C in air for 24 h. Nafion ® 212 and 115 membranes were pretreated by soaking in deionized water at 80 °C for 20 min, followed in hydrogen peroxide solution (6%) at room temperature for 35 min, and stored in 0.1M aqueous KCl solution at room temperature for further tests. AO-PIM membranes were pretreated in 1M KOH for 24 h to obtain their deprotonated form. Prior to full cell tests, Nafion ® and AO-PIM membranes were equilibrated in 1M aqueous KOH (for alkaline flow cell tests) or 1M aqueous KCl with pH = 9 (for near neutral-pH flow cell tests) overnight. For alkaline 2,6-DHAQ|K 4 Fe(CN) 6 flow cell tests, the electrolytes consisted of 0.1M K 4 Fe(CN) 6 , or 0.1M 2,6-DHAQ combined with 0.2M KOH in 10 mL 1M aqueous KOH solution. For near neutral-pH 2,6-DPPAQ|K 4 Fe(CN) 6 flow cell tests, the electrolytes consisted of 0.1M K 4 Fe(CN) 6 , or 0.1M 2,6-DPPAQ combined with 0.4M KOH in 10 mL 1M aqueous KCl solution. Trace amount of KOH was added to catholyte and analyte solutions to adjust the pH to 9.0. For neutral-pH BTMAP-Vi|BTMAP-Fc flow cell tests, the electrolytes consisted of 0.1M BTMAP-Vi or 0.1M BTMAP-Fc in 10 mL 1 M aqueous NaCl solution.
EIS spectra at 0 and ~100% state of charge (SOC) was measured using the potentiostat mode at an AC bias of 10 mV and a frequency range from 0.5 MHz to 10 Hz in the assembled flow cells. Charging-discharging curves were obtained using an electrochemical station (Biologic SP-150 potentiostat) with a constant current density at room temperature. To obtain an electrochemical polarization curve, the cell was charged to a desired SOC followed by polarized using linear galvanic sweep method at a rate of 200 mA s -1 from -6000 to 6000 mA. The corresponding power density at specific SOC (10, 50, ~100 %) was derived from the current -voltage curve. Long-term cycling tests were performed at 80 mA cm −2 for the 2,6-DHAQ|K 4 Fe(CN) 6 and 2,6-DPPAQ|K 4 Fe(CN) 6 cells and at 25 mA cm −2 for BTMAP-Vi|BTMAP-Fc cells in an argon-filled glove box. The coulombic efficiency (CE) was calculated by the ratio of discharge capacity to charge capacity. The voltage efficiency (VE) was calculated by the ratio of average discharge voltage to average charge voltage. The energy efficiency (EE) was calculated by the ratio of average discharge energy to average charge energy. The cyclic capacity decay rates of different membranes were evaluated by the overall percentage loss of capacity divided by the total operation time (cycle numbers or days).

Density Functional Theory (DFT) Calculations
DFT calculations were carried out to probe the charged properties of the redox active species and AO-PIM polymer segments. Geometry optimisations of the redox active species as well as single hydrogen capped cluster models of the AO-PIM polymers were carried out using Gaussian16 [6] with the Becke's 3 parameter exchange function combined with Lee-Yang-Parr correlation functional (B3LYP) [7] . The Los Alamos National Laboratory 2 Double-Zeta (LANL2DZ) [8] basis set was used for iron (Fe) atoms and the 6-31+G(d,p) basis set was used for all other atoms (H, C, N, O, P). Molecular electrostatic potential (MESP) surfaces were calculated for each optimized geometry and plotted on an isosurface of 0.002 a.u.  Table S1. ESP of three redox couples for aqueous organic RFBs (The unit of ESP map is a.u.)

Catholytes Anolytes
Fe ( The characteristic peak appearing at 2239 cm -1 is assigned to the stretching band of CN groups in parent PIMs, however, no CN group peak is visible in AO-PIMs after performing AO functionalisation. Other obvious vibrations for AO groups appear including the bands at 3480 and 3332 cm -1 from antisymmetric and symmetric stretching of NH 2 groups, 3170 cm -1 from stretching vibration of OH groups, 1645 and 915 cm -1 from C=N and N-O stretching of oxime groups [4] , respectively. These distinct peaks attributed to AO functionality indicate the complete conversion of nitrile groups to amidoxime groups in all four PIM polymers.  [a] T d1 is the peak degradation temperature for the mass loss of AO groups.
[b] T d2 is the peak degradation temperature for the backbone degradation.  Figure S7. Stress-strain curves of AO-PIM membranes. The mechanical behaviours of AO-PIMs follow a general feature of glassy polymers, i.e., exhibiting relatively high toughness and stiffness [9] , which indicate their high rigidity and low mobility of PIM polymeric chains.   [a] SA BET was obtained by N 2 adsorption in the range of P/P 0 = 0.001-0.1.
[b] Total pore volume was obtained from N 2 adsorption based on the NLDFT model.
[c] Water vapour uptake was measured using initial non-deprotonated AO-PIM membranes by DVS.
[d] AO-PIM membranes were soaked in 1M NaOH solution overnight to yield deprotonated forms then their water vapour uptake was obtained by DVS.
[e] The values of SA BET, total pore volume and water vapour uptake of PIM-1 and AO-PIM-1 were from our previously reported work [5] . Power-law scattering patterns relevant to a fractural surface structure at low q (< 0.05 Å -1 ) were observed for all samples. A broad scattering feature at q ~0.15 Å -1 (~4.2 nm in real space) was observed for wet Nafion 212 membrane, corresponding to the spacing of hydrophilic water domains in Nafion. This typical ionomer peak and nanoscale pore channel within Nafion membranes has been widely studied [10] . In contrast, all AO-PIM membranes did not present distinct scattering features over a wide low-q range from 0.05 to 0.3 Å -1 in both dry and wet states which supports the assumption of amorphous homopolymers without large intersegmental volume and local phase separation. were observed for dry AO-PIM membranes, which agree well with the pore information obtained from gas adsorption experiments. For wet AO-PIM membranes, the broad peaks at q ~ 0.9-2.0 Å -1 shifted to higher q positions, indicating that the polymer network is more densely packed in wet AO-PIM membranes. This is also supported by the lack of low-q (<0.5 Å -1 ) intensity. A similar phenomenon was observed and reported in our previous work [5] . In comparison, Nafion membranes presented a sharp scattering peak at q ~1.17 Å -1 derived from the scattering among crystallites in a dry Nafion amorphous matrix.    Ionic conductivity measurements were designed to decouple two key factors, i.e., free volume voids and AO groups that enable the ion transport through AO-PIM membranes. We assume that the free volume voids for ion transport would be disabled by using deionized H 2 O rather KCl electrolyte, while AO groups can be activated from neutral-charged form (i.e., -OH) to negative-charged form (i.e., -O -) by deprotonation at pH 14. Firstly, we activated AO groups in AO-PIM membranes and measured the ionic conductivity in deionized H 2 O (termed as H 2 O conductivity) to study the relative contribution of AO groups for ion conduction. All AO-PIM membranes show low H 2 O conductivities in the order of 10 -3 to 10 -4 mS cm -1 with AO-PIM-1 showing the highest value and AO-PIM-DBMP showing the lowest value ( Figure S15a). This trend agrees with the relative AO content of AO-PIM polymers. Lacking vehicular transport of charge carriers, the overlap of ionisable AO hydration shell is the only way to allow the charge conduction. Similarly, we applied nonactivated AO-PIM membranes (AO groups in OH forms) in 1M aqueous KCl for ionic conductivity tests (termed as b KCl conductivity) to study the relative contribution of free volume voids for ion conduction. The values for b KCl conductivity jump orders of magnitude to 10 -1 to 10 0 mS cm -1 with AO-PIM-DBMP showing the highest value while AO-PIM-1 shows the lowest value ( Figure S15b). This observed change is consistent with the relative micropore architecture that AO-PIM-DBMP exhibits the highest total pore volume hence the most sufficient pathways allowing fast transport of charge carriers. Importantly, by activating AO groups and using KCl/KOH electrolytes together, KCl ionic conductivity (termed as a KCl conductivity) or KOH ionic conductivity (termed as KOH conductivity) of AO-PIM membranes are significantly improved to 10 1 mS cm -1 that are comparable or even superior to commercial benchmark Nafion 212 membrane ( Figure 2 h&i and Figure S15c). We further normalised the H 2 O conductivity and b KCl conductivity of four AO-PIM membranes to quantify the relative contribution for ion conduction from free volume voids and AO groups. AO groups contribute the largest portion to the ion conduction for AO-PIM-1 membranes, while high free volume voids contribute the largest portion to the ion conduction for AO-PIM-DBMP membranes, and AO groups and free volume voids contribute a balanced portion to the ion conduction for both AO-PIM-SBF and AO-PIM-BTrip membranes ( Figure S15d). However, all of four fully activated/deprotonated AO-PIM membranes exhibit high conductivity under aqueous electrolytes, which suggests a collective property owing to the combination of high micropore volume and interaction between charge carriers and ionisable AO groups.    [a] Permeation rates and calculated ideal selectivity are tested in 1M neutral pH aqueous electrolytes by non-deprotonated AO-PIM membranes.
[b] Permeation rates and calculated ideal selectivity are tested in 1M neutral pH aqueous electrolytes by deprotonated AO-PIM membranes (i.e., pre-treated in 1M aqueous KOH overnight).       Figure S21. Electrochemical reaction mechanism of (a) 2,6-DHAQ|K 4 Fe(CN) 6 redox couple in 1 M aquerous KOH and (b) 2,6-DPPAQ| K 4 Fe(CN) 6 redox couple in 1M aqueous KCl at pH = 9. Typical C-V curves of (c) 2,6-DHAQ (red trace) and K 4 Fe(CN) 6 (blue trace) in 1M aqueous KOH [5] and (d) 2,6-DPPAQ (red trace) and K 4 Fe(CN) 6 (blue trace) in 1M aqueous KCl at pH = 9 using three-electrode electrochemical systems.  [c] SA BET value of AO-PIM-1 was from our previously reported work [5] . . Similar EIS spectra of each membrane in fresh and aged stage suggests the retention of microporous structure of swollen hydrophilic AO-PIM membranes. This is because typical relaxation of glassy polymers including PIMs from nonequilibrium to equilibrium packing state over time generally leads to significant reduction of free volume (i.e. aging behaviour) hence the deterioration of electrochemical performance including ionic conductivity [11] .     Figure S2 and Table S1) might have been expected to encourage severe absorption of this redox couple into the AO-PIM-1 membrane. However, the relatively stable battery performance suggests that the strong size-sieving effect of AO-PIM membranes plays a key role in achieving high molecular selectivity.