Understanding the Role of Proton and Hydroxide Transport in Forward‐Bias Bipolar Membrane for Electrochemical Applications

A forward‐bias bipolar membrane (BPM) provides an alkaline cathode condition, which can be beneficial to some electrochemical reactions, such as the CO2 reduction reaction (CO2RR), but the water association (WA) in forward‐bias BPM is not well understood at all. In this study, BPMs are designed with different interfacial polymeric catalysts to investigate the WA reaction under forward‐bias for electrochemical applications. An enhanced current density is observed with added polymeric catalysts (−OH, −O−, −N−, and graphene oxide) compared to the blank control. Temperature‐dependent measurements indicated that the WA in BPM is not kinetically controlled. The in‐plane and through‐plane ions diffusion is investigated, which showed that the WA in BPM is limited mostly by the transport of OH− and, to a lesser degree, H+ at the interface. Molecular dynamic studies presented that the migration rate of OH− at the interface is approximately one order of magnitude lower than that of H+, indicating that the WA is mainly governed by the transport of OH−. Finally, a forward‐bias CO2RR electrolyzer is demonstrated with an Faradaic effeiciency CO (FECO) of 92.2 ± 2.7%. This work provides important fundamental insights into the WA reaction that would enable the use of forward‐bias BPM electrolyzers in future electrochemical applications.


Introduction
Excessive utilization of fossil fuels has caused a dramatic increase in atmospheric CO 2 concentration, rising from 280 ppm in the 1800s to ca. 400 ppm in 2020.This surge has accelerated DOI: 10.1002/admi.202400034climate change and led to multiple adverse effects. [1]Capturing and converting anthropogenic CO 2 into chemical fuels using renewable energy represents a promising pathway toward establishing a circular carbon economy. [2]The majority of CO 2 electrolyzer rely on an anion exchange membrane (AEM) to separate the anode/cathode and manage ions transport. [3]In such systems, a high FE for CO 2 is achieved as the cathode resides in an alkaline environment, which suppresses the hydrogen evolution reaction (HER).However, ions such as OH − , CO 3 2− , and HCO 3 − can crossover from the cathode to the anode, reducing the utilization of CO 2 and overall energy efficiency. [4]4b,5] However, the acidic environment in CEM can promote the competing HER, thereby lowering the FE of CO 2 reduction.One possible approach to solving these obstacles is to apply a BPM in CO 2 electrolyzers. [6]Typically, a BPM consists of an AEM and a CEM laminated together, with a catalyst layer at the interface.6d,7] Most BPM CO 2 electrolyzers operate under reverse-bias conditions, where the CEM and AEM face the cathode and anode, respectively.At the interface of the BPM, water dissociation (WD), H 2 O → H + + OH − , provides H + and OH − for charge balance. [8]ithout further engineering design, CO 2 electrolyzer in reverse bias would suffer from low CO 2 FE similar to the CEM due to the acidic local environment.The BPM can also be operated in forward bias where the WA, H + + OH − → H 2 O, provides charge balance by moving H + and OH − from the anode and cathode respectively. [9]Forward bias could also lead to ion accumulation at the interface due to the imperfect ion selectivity of the AEM and CEM, such as CO 3 2− .A large current in forward bias can also generate water at the BPM interface faster than it can dissipate leading to delamination.Early demonstrations of electrochemical devices in forward bias have also been reported, but they are much less common than those in reverse bias. [9,10]Since the majority of studies on BPM devices have been done in reverse bias, [11] there have been a number of studies reporting on the investigation of water dissociation catalysts to enhance the reverse bias current in BPM. [8,12]However, to improve the CO 2 RR FE in BPM devices, forward bias provides an alkaline cathode environment, which can inhibit the competitive HER and promote CO production.This contrasts with a reverse bias acidic cathode environment.In this context, understanding the water association in BPM under forward-bias is critical for lowering the overpotential and enhancing the catalytic activity.
In general, water association (H + + OH − → H 2 O) is thermodynamically spontaneous.10a] This study is the first to highlight the importance of understanding the H + /OH − recombination mechanism in BPM setting.The recombination of H + + OH − is perhaps the most fundamental proton transfer reaction, and a rate constant of up to 1.4 × 1011 L mol −1 s −1 has been reported. [13]To put this number into perspective, a 1 M H + and 1 M OH − solution would neutralize each other in ≈10 −11 s -the recombination kinetics of the water association reaction is very fast to begin with in aqueous conditions.However, there are additional considerations for this reaction when it is happening inside a BPM.First, H + and OH − are confined and separated in CEM and AEM respectively before recombination at the interface of the BPM.Since the BPM is under an electrical bias, the polarization and dielectric properties of H + /OH − /H 2 O are different from those in a free solvent, which would affect the reaction kinetics.Also, H + and OH − must travel through a polymeric membrane and also need to find each other at the interface before the recombination reaction can occur.Hence, the enhancement in catalytic current for the water association in a BPM could be due to one or more of the transports and kinetic processes, which are not well understood.
In this study, we explored the origin of the enhanced catalytic current in forward-bias BPM using both metal oxide nanoparticles and polymeric material as the water association catalyst.We focused on understanding whether the origin of the enhanced current was caused by either reaction kinetics or mass transport.Notably, the forward-bias BPM with polyethylene oxide (PEO) with −O− functional groups exhibited the highest current density among the materials tested, up to 247.3±10 mA cm −2 .We discovered that the functional groups (−OH, −O−, and −NH−) of the polymeric materials affected the transport of the H + and OH − at the interface of the BPM.The in-plane mass transport mechanisms for OH − and H + were the major contributions to the enhanced catalytic current, more than kinetics.Molecular dynamic calculations further ascertained that OH − transport was affected by the interfacial catalyst and played a significant role in the spontaneous water association reaction.With our homemade BPM, we demonstrated a gas-fed membrane electrode assembly (MEA) CO 2 electrolyzer in forward-bias conditions.A maximum FE CO of 92.2 ± 2.7% was obtained at a current density of 150 mA cm −2 .These results highlight the importance of the mass transport mechanism inside the BPM, and that the forward-bias configuration MEA design could enable better performance for different electrochemical systems.

Kinetics Investigations
Two families of materials, metal oxide nanoparticles and polymers, were chosen as catalytic materials for the water association reaction under forward bias in BPM.A custom BPM was hot-pressed with the respective catalytic materials at the interface and tested in a four-electrode electrochemical cell.The details of the synthesis and testing conditions are available in the Supporting Information.Cross-sectional scanning electron microscopy (SEM) images (Figure S8, Supporting Information) of our in-house BPM show uniform binding between the AEM and CEM layer.To obtain the jV curve of the BPM, the CEM side of the BPM faced the anode with 1 M H 2 SO 4 , and the AEM side faced the cathode with 1 M KOH (illustrated in Figure S1, Supporting Information).In this operation mode, H + and OH − transport through the CEM and AEM, respectively, and recombine at the interface of the BPM.In open circuit voltage condition (OCV), when the BPM is neither in forward bias nor reverse bias, the OCV should be ca.0.8 V to reflect the H + /OH − neutralization reaction thermodynamically.Figures 1 and S2 (Supporting Information) show the jV curves for BPMs with different catalysts in the potential range of −0.4-0.4V. Applying a positive potential (0-0.4V) drives the H + and OH − toward the BPM junction, inducing a forward bias current.Conversely, applying a negative potential (−0.4-0 V) resulted in no current flow since the 0.8 V of thermodynamic voltage for water dissociation was not reached yet.Under forward bias conditions, the current densities for all the catalysts were higher than for a blank BPM without an interfacial catalyst.10a] All the metal oxide nanoparticles were controlled for similar size and loading.Among the metal oxide nanoparticle catalysts, TiO 2 exhibited the best performance at 188.4 mA cm −2 .When subjected to the same loading, molecular catalysts, particularly PEO and polyvinyl alcohol (PVA), exhibited higher catalytic activity than metal oxide nanoparticles.PEO (40.0 μg cm −2 ) exhibited the highest current density at 247.2 mA cm −2 , which is 2.6 times higher than that of a blank BPM.We also tested the BPMs under reverse bias (Figures S3,S4, Supporting Information).The corresponding LSV curves for BPMs with different catalysts under both forward and reverse biases are presented in Figure S5 (Supporting Information).In line with the jV curves, the BPM-PEO achieved the highest current density under forward bias, but a lower current density was observed under reverse bias.While graphene oxide (GO) was only a moderate catalyst for the water association reaction, it exhibited the highest promotion for the water dissociation reaction.12b,14] It also revealed that the reaction mechanism for the water dissociation and water association are different.We focused this study on the forward bias water association reaction.The difference in reaction mechanism for WA and WD will be the subject of future investigation.
We used SEM to evaluate the distribution of the catalysts.Uniformity was a critical factor for activity calculation, as electrochemical reactions are surface-related.During hot-pressing, catalyst materials were deposited on the AEM surface before bonding with the CEM.Consequently, Figures 2 and S6, S7 (Supporting Information) display the surface of a blank AEM and AEM deposited with different catalysts.The blank AEM (Figure 2a) had a relatively smooth surface, whereas aggregated SiO 2 , TiO 2 , and CeO 2 nanoparticles could be observed distributed on the AEM's surface (Figure 2b; Figure S6a,b, Supporting Information).While typical surface roughness was expected for solid nanoparticles, extra care was necessary to normalize the current and determine the intrinsic activity, given that the interfacial area had become three-dimensional.Conversely, minimal to no surface roughness was seen in the polymeric coated materials (Figure 2c; Figure S6d, Supporting Information) since all the polymers were dissolved in the appropriate solvent before being air-sprayed onto the AEM.The GO layers were also evenly distributed on the AEM's surface (Figure 2d), though wrinkling was observed due to drying in the SEM chamber.To systematically investigate the interfacial reaction at the BPM, maintaining the smoothness of the interfacial layer was crucial.While roughness might enhance catalytic activity, it made it challenging to normalize the current to compute the intrinsic current density.The interfacial distance between AEM and CEM was also affected by the thickness of the metal oxide nanoparticles, potentially influencing the local electric field.Taken together, we focused our subsequent investigations on polymeric materials only.As a control experiment, we also screened two additional polymers with ether functional group, polyethylene glycol (−(CH 2 −O−CH 2 −) n −, n≈1500) (PEG) and ethylene glycol diethyl ether (CH 3 (CH 2 OCH 2 ) 2 CH 3 ) (EGDE) to compare against PEO.As shown in Figure S2d (Supporting Information), both PEG and EGDE achieved similar current densities as PEO as the interfacial catalyst layer in BPM, suggesting that the chain length for polymers was less important than the type of functional group present.Therefore, we developed the mechanistic investigation focusing on polymers with different characteristic functional groups (−OH in PVA, −O− in PEO and −NH− in polyethyleneimine (PEI)).In a catalytic reaction, the enhanced current density can arise from either improved intrinsic kinetic activity or enhanced mass transport.Hence, we first conducted a series of temperaturedependent electrochemical impedance spectroscopy (EIS) in a four-electrode configuration to probe the kinetic effects.If the water association was kinetically limited, the temperature dependence of the current density would be expected to follow an Arrhenius-type relationship.Figures 3a,b and S9 (Supporting Information) display the EIS spectra of blank BPM, BPM−PEO, BPM−PEI, and BPM−PVA.All of these showed a progressively decreasing charge transfer resistance (R ct ) as the temperature increased, indicating improvements in both kinetics and/or transport.A singular linear relationship was expected for ln (R ct −1 ) versus 1/T if kinetics were the dominant factor in the jV curves.Table S1 (Supporting Information) provided the fitted data and the equivalent circuit for EIS curves.Using the temperaturedependent R ct data for the different polymer materials, we calculated the activation energy for each BPM, and the results are presented in Table S4 (Supporting Information).Notably, only the blank BPM (inset in Figure 3a) displayed a "two-region" linearity in the Arrhenius plot, which suggests a change in mechanism as the temperature varied.The deviation from a linear Arrhenius plot implies that the OH − and H + recombination could operate under different mechanisms at varying temperature conditions.Given that the OH − and H + recombination reaction is already one of the fastest reactions known, we believe that the introduction of the polymeric interface layer affected the transport of reactants rather than the kinetics of the reaction.In the low-temperature region, the blank BPM was transport-limited with a high activation energy of 70.6 kJ mol −1 .As the temperature rises, ion mobility increases and the activation energy decreases.BPM−PEO (inset in Figure 3b), BPM−PEI (inset in Figure S9a, Supporting Information), and BPM−PVA (inset in Figure S9b, Supporting Information) all display a single linear slope with activation energies of 46.1, 46.7, and 58.1 kJ mol −1 , respectively.The lowest activation energy for BPM−PEO was consistent with our jV curves, where BPM−PEO as a catalyst exhibited the highest forward bias current.Based on these findings, we proposed that mass transport was the initial current limiting step in the blank BPM at room temperature.However, as a catalyst was incorporated, mass transport improved, and water association became more kinetically limited.The corresponding temperature-dependent LSV curves (Figure S10, Supporting Information) showed that both onset potential and current density increased with rising temperature, aligning with the temperature-dependent EIS results.

Transport Investigations
To determine the transport properties of H + and OH − , we measured the through-plane and in-plane EIS on the AEM and CEM surface with different catalysts tested in previous section. [15]Illustration of the measurement setup are provided in Figure S11a,b (Supporting Information).The through-plane EIS represented the H + /OH − transport through the bulk of the polymeric materials and the in-plane EIS gave information on the diffusion of H + and OH − at the interface of the BPM, which was required for the H + and OH − to find each other for recombination.As shown in Figure 3c,d, no obvious difference, within experimental fluctuation, was observed for the through-plane EIS curves between blank CEM/AEM versus CEM/AEM coated with catalyst, suggesting that the through plane H + /OH − transport was minimally affected by the added catalyst layer.This result was reasonable since the CEM/AEM itself was ca. 100 μm thick and the catalyst layer was likely in the 10s of nm.We noted that in the high-frequency region (Figure S12, Supporting Information), a small semicircle was observed for all the EIS spectra with AEM, which is attributed to the ion equilibrium between the membrane-electrolyte interface due to the Donnan potential effect. [16]The same semicircle features for CEM in similar frequency regions are less discernable likely to the faster ion equilibrium for H + .The in-plane EIS curves for CEM/AEM with catalysts all showed a smaller R ct than the blank CEM/AEM (Figure 3e,f).The EIS spectra were fitted with an equivalent circuit model and the fitted results were shown in Tables S2 and S3 (Supporting Information).Elongated semi−circles were observed for both the CEM and AEM with catalyst layer, which suggest multiple electrochemical processes.In the equivalent circuit, R 0 represented the ohmic resistance of the system.Based on similar diffusion measurement in literature, [17] R 1 represented the impedance of the interface of membrane and Pt electrodes, while R 2 can be ascribed to the diffusion of H + and OH − at the interface of the membrane.While catalyst coated CEM had slightly smaller R 1 and R 2 than the blank CEM (Figure 3e; Table S2, Supporting Information) in general, the difference was small and within experimental fluctuation, suggesting that the diffusion of H + was not a rate limiting factor in the water association reaction.On the other hand, R 1 and R 2 between blank AEM and AEM with catalysts (Figure 3f and Table S3, Supporting Information) showed a significantly larger difference, suggesting that the OH − transport was significantly enhanced with the added polymeric materials, compared to control.Furthermore, the AEM−PEO displayed the smallest R 1 (6.5 Ω) and R 2 (7.2 Ω), compared to all the other catalyst materials.This was consistent with the jV data that PEO delivered the highest enhancement in current density.Based on these experimental observations, we posited that the functional group of the polymeric materials can improve the water association process via promoting the OH − transport at the interface of CEM/AEM along the in-plane direction.

Molecular Dynamic Simulations
To further explore the role of ions transport, molecular dynamics (MD) calculations were carried out to determine the effect of different polymer groups on the diffusion coefficient of OH − and H + .In the simulation, OH − and H + were allowed to diffuse within the polymer network over 15 ns, and the diffusion coefficient was calculated using the mean square displacement.Figures 4a,b and S13 (Supporting Information) display the diffusion paths of OH − and H + along different polymer backbones.The OH − and H + can be directly observed in the enlarged versions of the diffusion models.Video S1 (Supporting Information) illustrates that hydroxide ions move freely within the PEO network, rather than hopping between active sites, suggesting that vehicular diffusion was the predominant conduction method in our study.As illustrated in Figure 4c, the calculated diffusion coefficients of OH − in PEI, PEO, and PVA were 0.53, 0.57, and 0.54 Å 2 ps −1 , respectively.These values align with the impedance measurements from the experiment in Table S3 (Supporting Information), where AEM−PEO exhibited the lowest impedance, followed by AEM−PVA, while AEM−PEI had the highest impedance.The H + diffusion rate from the MD simulation (Figure 4d) also corresponded with the experimental EIS results in Table S2 (Supporting Information), in which the PEO network most significantly enhanced proton diffusion.The consistency between the experimental and computational results further supports our hypothesis that the recombination of OH − and H + is influenced by the addition of the polymeric interface through a transport mechanism.

CO 2 RR Performance
To demonstrate the potential application of a forward bias BPM, we showcased a BPM CO 2 electrolyzer in which the CO 2 cathode faced the AEM under forward bias conditions.The alkaline interface from the AEM created a favorable environment for CO 2 RR.The design of the BPM effectively prevented product crossover, a significant issue with monopolar AEM.Therefore, the forward bias BPM electrolyzer offered optimal conditions for CO 2 RR development.A schematic illustration of the MEA setup is shown in Figure 5a.Carbon foam with Ag nanoparticles was used as the cathode, while a Pt mesh served as the anode.During operation, humidified CO 2 and sulfuric acid were introduced to the cathode and anode, respectively.Under the applied voltage, the reactions occurring at each component of the cell can be summarized in equations 1 through 4 below: BPM : Due to the alkaline conditions for CO 2 RR, CO 2 can react with OH − to produce CO 3 2− .This compound can then be transported via the AEM to the interface of the BPM.Once there, CO 3 2− can influence the recombination kinetics of OH − and H + .To investigate this mechanism, we designed a three-chamber experiment to measure the transference number of OH − and CO 3 2− across the AEM (refer to Figure S14, Supporting Information).Further experimental details can be found in the Supporting Information section.After maintaining electrolysis at 10 mA cm −2 for 1 hour, the ratio of OH − to CO 3 2− transported across the AEM was ≈ 8:1.This indicates that OH − is the predominant anion at the BPM interface.
The jV curves for the forward bias CO 2 RR cell are displayed in Figure 5b, an overall cell current density of −413.9 mA cm −2 was achieved for BPM-PEO at −4.5 V without any iR correction, a 42.0% improvement over blank BPM (−291.6 mA cm −2 ).11b,18] The theoretical value for the cell voltage (V cell ) can be expressed as V cell = V cathode −V anode +V membrane +V ohmic , [19] where each term denotes the voltage drop across the cathode, anode, membrane, and ohmic components, respectively.The cathode voltage, V cathode , is further defined as V cathode = E 1/2,cat + activation + transport , while V anode possesses analogous terms for the anodic reaction.The ohmic voltage, V ohmic , encompasses the membrane resistance, membrane−catalyst interface resistance, and contact resistance.Considering the CO 2 -to-CO reaction at the cathode and the oxygen evolution reaction (OER) at the anode, the theoretical cell voltage should be ca.1.34 V. [11a,20] Given a typical overpotential ( activation ) of 400 mV each for OER and CO 2 RR, the onset potential for a CO 2 electrolyzer should span between 2 -2.3 V, excluding ohmic resistance.This correlated with our data in Figure 5b closely, wherein our onset potential measures roughly V.
V ohmic significant at higher potential, which was what contributed to the high cell voltage observed at Figure 5b at large current density.The Faradic efficiencies for the CO 2 conversion to CO at different current densities are shown in Figure 5c, A 92.2% FE CO was achieved at 150 mA cm −2 , further increasing the current density would lower the FE, similar to Ag catalyst behaviors reported in literature.Such high FE in CO 2 RR with BPM is typically not feasible in the reverse bias conditions or can only be done with a buffer layer between the catalyst and membrane.We also conducted stability tests on our home−made membrane at 150 mA cm −2 (Figure S16, Supporting Information) over a duration of 6 hours.The current density and FE were stable for the duration of the testing and there was no discernible degradation or delamination to the BPM post reaction.

Conclusion
In summary, we have elucidated that the limitation for WA under forward-bias BPMs is predominantly due to mass transport limitation, specifically of OH − , in the in-plane direction at the junctions.Firstly, both the chosen metal oxide nanoparticles and polymers exhibited increased current densities for water association.The temperature-dependent EIS indicated that reaction kinetics might not be the primary factor in the spontaneous WA process.Notably, the stark differences observed in the in-plane EIS curves between blank AEM and AEM with catalysts showed that the in-plane transport of OH − is the limiting factor, as opposed to reaction kinetics.Molecular dynamic simulations show similar trend in the OH − and H + diffusion coincidence, further corroborated the transport mechanism.When applied to a MEA for CO 2 , a maximum FE CO of 92.2 ± 2.7% was achieved at a current density of 150 mA cm −2 .With the inclusion of PEO under a forward-bias BPM, the current densities reached a peak of 247.3 ± 10 mA cm −2 .Our insights into mass transport limitation can inform future catalyst design, emphasizing that achieving a swift reaction rate demands not only rapid reaction kinetics but also enhanced mass transport to facilitate an improved recombination rate.

Figure 1 .
Figure 1.a) The jV curves of forward-bias BPMs with different loading of GO. b) The jV curves of forward-bias BPMs with different loading of PEO.c) The jV curves of forward-bias BPMs with different loading of PEI.d) The jV curves of forward-bias BPMs with different loading of PVA.Error bars represent one standard deviation from three independent measurements.

Figure 2 .
Figure 2. a) SEM images for blank AEM.b) SEM images for SiO 2 −AEM.c) SEM images for PVA−AEM.d) SEM images for GO−AEM.

Figure 3 .
Figure 3. a) EIS curves in forward-bias for blank BPMs under different temperature.The insert image is the relationship between R ct and 1/T.b) EIS curves in forward-bias for BPM-PEO under different temperature.The insert image is the relationship between R ct and 1/T.Error bars represent one standard deviation from three independent measurements.c) Through-plane EIS curves for CEM with different catalysts.d) Through-plane EIS curves for AEM with different catalysts.e) In-plane EIS curves for CEM with different catalysts.f) In-plane EIS curves for AEM with different catalysts.

Figure 4 .
Figure 4. a) The sketch of the diffusion model when OH − migrates in the polymer network composed of PEO.b) The sketch of the diffusion model when H + migrates in the polymer network composed of PEO.c) The calculated diffusion coefficient of OH − in PEI, PEO, PVA polymer network.d) The calculated diffusion coefficient of H + in PEI, PEO, PVA polymer network.

Figure 5 .
Figure 5. a) Schematic of the BPMs under forward bias in a MEA cell.b) jV curves for MEA cell.c) Faradaic efficiency for BPM-PEO at different current densities under forward bias.Error bars represent one standard deviation from three independent measurements.
funding supporting.Y.C.L. would also like to thank the Office of International Education at the University at Buffalo, SUNY for seed grant funding supporting.The computational study is supported by the Marsden Fund Council from Government funding (21−533 UOA−237) and Catalyst: Seeding General Grant (22−UOA−031−CGS), managed by Royal Society Te Apārangi.Z.W. and C.Z. would like to acknowledge the use of New Zealand eScience Infrastructure (NeSI) high-performance computing facilities, consulting support and/or training services as part of this research.X.L. would like to thank Dr. Zheng Li for assistance in SEM imaging.All authors discuss and contribute to the writing of the manuscript.