Designing the next generation of symmetrical organic redox flow batteries using helical carbocations

INTRODUCTION

Meeting the ever-growing global energy demand is an urgent imperative, necessitating the development of energy-storage solutions. These solutions play a crucial role in facilitating the transition to renewable energy grid integration, decarbonizing our economy, and tackling the challenges posed by global warming^[1-3]. In order to achieve the objectives outlined in the International Energy Agency roadmap^[4], the emergence of a variety of metal-based battery solutions has been observed^[5-8], notably marked by the widespread adoption of high-performance lithium-ion batteries^[9-11]. However, as concerns about fire hazards linked to the accumulation of low-valent metal have risen and the current demand rapidly overtaking the availability of this metal^[12,13], the industry is being directed toward safer, cost-effective, and scalable electricity energy storage (EES) solutions^[14,15]. Redox flow batteries (RFBs) have emerged as a compelling choice for large-scale stationary applications within the EES domain, primarily because of their engineering adaptability and scalability enabled by the decoupling of power and capacity^[16-19].

Currently, commercially established RFB technologies are based on the redox characteristics of transition metal ions in aqueous solutions (AqRFBs). Among these, the most extensively studied options are vanadium oxide and chromium and iron variants^[20-24]. Nevertheless, several drawbacks hinder the widespread deployment of AqRFBs, such as the limited electrochemical stability window of water (1.5 V), low energy densities induced, high costs associated with ionic separators, environmental implications related to ore mining, management of substantial quantities of acid, and the inherent toxicity of certain compounds^[25-27].To broaden the electrochemical window and enhance energy density under mild pH conditions, non-aqueous RFBs present a significant advancement^[28-30]. They enable the use of Redox-active Organic Molecules (ROMs), demonstrating enhanced sustainability compared to transition metal-based redox-active materials^[31]. ROMs have a lower global warming potential index, indicating an environmental impact one-fourth that of transition metal-based materials^[32]. Moreover, they will ensure a consistent market price at one-fourth the cost per kilogram and can be sustainably sourced and tailored synthetically^[31-33]. Most of reported Non-Aqueous fully-Organic RFBs (NAORFBs) exhibit an asymmetric configuration, using two distinct ROMs as catholyte and anolyte. This setup imposes selectivity requirements on the exchange membrane, leading to alternative avenues for chemical degradation and irreversible capacity fading^[34,35]. Various effective approaches have been investigated to mitigate redox-active material crossover, including interface improved size selectivity^[36,37], enhanced membrane composition^[38-40], and eliminating the need for a membrane by adopting tandems of non-miscible redox electrolytes^[41].

Concurrently, with the rise of innovation in the field, an elegant solution has emerged: drawing inspiration from vanadium RFBs (VRFBs) and their quasi-identical tank chemistry, symmetrical Organic RFBs (SORFBs) have gained prominence as the most promising approach for eliminating membrane crossover and prolonging EES lifespan^[23,42-44]. The primary characteristic of SORFBs lies in their use of a singular bipolar redox molecule (BRM) on both sides of the cell^[45]. This robust molecule features in its initial state -at least- three stable redox states, owing to entirely reversible reduction and oxidation processes. Consequently, these molecules can function as both anolyte and catholyte within a battery^[46-48]. The resultant symmetrical RFB relies on identical solution components in each half-cell, offering notable advantages^[42]. Firstly, using the same redox-active material reduces the chemical gradient of electroactive species, negating the necessity for highly selective membranes and effectively mitigating crossover. In the event of crossover, rather than enduring permanent contamination, SORFBs will undergo self-discharge^[23]. Furthermore, when in the discharged state, there is no chemical or electrochemical gradient across the membrane, guaranteeing SORFBs to be stored indefinitely without contamination due to leakage or irreversible side reactions. Finally, recent research has demonstrated that capacity loss resulting from compound degradation can be reclaimed through regular polarity reversals, thus extending the battery lifespan^[42,49-51]. Additionally, relying on a single charge carrier species could help achieve significant economies of scale while streamlining the supply chain for broad implementation. The recent surge in SORFB progress has inspired chemists to craft a varied array of BRMs, employing three molecular approaches to fashion novel redox-active compounds [Figure 1]^[45].

Figure 1. Three strategies to develop innovative bipolar redox molecules for the rise of Symmetrical Organic Redox Flow Battery. n corresponds to n-type molecules, p for p-type molecules, while Ab means for ambiphilic molecules. Examples of SORFBs based on strategy (A) could be found in the references^[52-58], strategy (B) at^{[49,50,59,60]} and strategy (C) at^[51,61-64].

Redox-active organic materials are usually categorized as n-type or p-type based on their ability to either store or provide electrons in their neutral state during electrochemical reactions. n-type redox-active organic materials typically experience reduction from their neutral state, resulting in a negatively charged molecular state. In contrast, p-type organic materials undergo oxidation from their neutral state, yielding positively charged. Both ROMs must exhibit reversible redox events to be used. One straightforward method involves pairing an n-type moiety with a p-type moiety using a simple linker, most of the time insulating [Figure 1A]. The resulting molecule can function as a bipolar entity (^-●n-p$$\rightleftharpoons$$n-p$$\rightleftharpoons$$n-p^●+), provided that none of its three redox states are compromised or interact with uncharged molecules in the medium. This “combi-molecule” approach^[65] is both fundamental and effective, as it leverages the extensive research conducted on catholyte and anolyte structures for fully organic RFBs, each with a known potential, leading to a BRM with a predicted potential gap (E_gap = E_1/2^Ox - E_1/2^Red) between both sides of the SORFB^[43,52-58]. However, this approach underutilizes half the molecule on each battery side, reducing atom efficiency and significantly increasing the molecular weight of the charge carrier, limiting potential energy density per kilogram of ROM^[66].

An improved strategy for developing innovative BRMs enhances this design concept by electronically merging established n-type and p-type conjugated scaffolds within a single molecule [Figure 1B]. By closely integrating the conjugation of two aromatic materials, the electronic characteristics of each component are anticipated to synergistically influence the resulting electrochemical potentials. Rising the E_gap while maintaining a reasonable molecular weight ensures a favorable capacity per unit weight. This route has proved successful, leading to the formation of closed-shell^{[49,58,60,67]} and open-shell molecules^[59,68] that exhibit improved properties compared to their parent units. Nonetheless, the synthetic access to such compounds can be a gradual and intricate process, involving multiple steps and introducing the risk of antagonism effects due to significant structural alteration of the initial n- and p- moieties. And, in contrast to method a, only a comprehensive electrochemical assessment of the final molecule can guarantee stability across all oxidation states (^-●n|p$$\rightleftharpoons$$n|p$$\rightleftharpoons$$n|p^●+).

The latest approach shifts away from merging the electronic characteristics of n- or p-type molecules. Instead, it revolves around exploring existing literature for compounds that exhibit bipolar activity (Figure 1, ^-●Ab $$\rightleftharpoons$$ Ab $$\rightleftharpoons$$ Ab^●+). Although such compounds are relatively scarce, the concept of relying on already well-defined and robust bipolar molecules proves refined and straightforward. Consequently, this facilitates maintaining a low molecular weight while customizing the scaffold to enhance E_gap, stability, solubility, or cost-effectiveness [Figure 1C]. Along this path, open-shell compounds, acting as ambiphilic radicals with three stable states, have achieved significant success when appropriately managing their inherent reactivity^{[51,61,62,69]}. The search for additional ambipolar compounds can be rooted in redox-active ligands featuring multiple stable states commonly found in homogeneous catalysis^[63,64,70] or in organophotocatalysis characterized by three well-defined redox states^[71].

Aligned with strategy (c) and contributing to the SORFB research field, our group chose to employ a specific category of BRMs within the carbenium ion class. Among the members of this carbocation family, certain compounds have demonstrated exceptional stability^[72-74] and hold significant utility in coordination chemistry^[75-77], small molecules activation^[78,79], and, notably, organophotocatalysis^[80-86]. And while a planar triangulenium motif has recently enabled the deployment of a robust SORFB model^[87], it is the particularly rich and versatile chemistry of [4]helicenium ions that will be discussed in what follows^[88]. Dimethoxyquinacridiniums (DMQA⁺) are sturdy helical carbenium ions, well-documented for their biological applications, photophysical features and electrochemical properties^[89-96]. The reversible electrochemical oxidation and reduction processes involving these carbenium ions have been notably disclosed by Herse and Sørensen et al.^[89,97]. In line with this, our group has recently published a study on the chemical synthesis, isolation, and characterization of neutral helicene radicals produced through chemical reduction of the corresponding carbocation precursors^[98]. All of these preliminary efforts have enabled us to introduce a robust SORFB model based on a relatively straightforward N,N′-di-n-propyl-1,13-dimethoxyquinacridinium (^nPrDMQA⁺) with the capability to operate in a poleless mode^[99]. This research led to the publication of a second model featuring an enhanced DMQA⁺ framework through nitro group incorporation (^nPrDMQA^NO2+), resulting in a significant boost in energy density at the expense of stability^[100].

In this context, this work aims to demonstrate how the utilization of pre-established bipolar redox scaffolds facilitates effective screening. We propose a methodology for rapidly evaluating the properties and compatibility of new BRMs for SORFB deployment. This approach is illustrated through the examination of 11 differently substituted helicenium examples, leveraging our expertise on the DMQA⁺ core. The results will be showcased through electrochemical characterization using a three-electrode cell. A thorough investigation of electrochemical kinetic parameters will help qualify the effects induced by various substitutions and structure tethering. Lastly, the robustness of ROMs, which have demonstrated suitable parameters for use in the RFB field, will be evaluated through cycling in H-cells, also known as “static RFB”.

EXPERIMENTAL

RESULTS AND DISCUSSION

The choice of the [4]helicenium motif as the core for structural modifications is related to its high durability and the numerous applications we have attributed to it within our research group. Indeed, the DMQA⁺ core, a stable carbocation in both air and water in its synthesized form, exhibits three stable redox states. The DMQA⁺ can thus be electrochemically reduced by one electron to form a neutral DMQA^● radical at E_1/2^Red and conversely oxidized by one electron to form a DMQA^●++ dication radical at E_1/2^Ox [Figure 2].

Figure 2. Stables redox states of DMQA⁺ core and structural modifications explored in this study.

The access and stability of these two states are crucial for obtaining a suitable BRM for symmetrical RFBs. Hence, all structural modifications explored must be compatible with the various oxidation states. Two variation types that showcase the extensive tunability of the DMQA⁺ motif can be investigated. Alterations of the aromatic core of the DMQA⁺ motif through positions X or Y with electrophilic or nucleophilic groups are feasible and detailed at various synthesis stages, whether initially on the tris-aryl carbenium synthon for Y or during the late functionalization stage for X [Supplementary Figure 1]^[94,100]. Changes in the nature of the R¹ and R² “arms” are almost invariably introduced during a double nucleophilic aromatic substitution [Supplementary Figure 1]^[103].

Considering the results obtained with the SORFB model based on ^nPrDMQA^+[99], initially, we examined the consequences of modifying the X group through late-stage functionalization. The synthesis of these compounds had been well described and was achievable by adding only a few steps to the initial synthesis path^[94]. The groups NO₂, OMe, NH₂, and NMe₂, along with their influence on the electrochemical properties of the ^nPrDMQA⁺ scaffold as a reference, were thus studied in acetonitrile, our model solvent (Figure 3, ^nPrDMQA⁺ as black trace)^[104,105]. The impact of the X-position substitution on the core has proved to be significant for the E_gap and, thus, for attainable energy density^[31]. As previously reported, the introduction of an electron-attracting nitro group contributes to shifting the potentials of the E_1/2^Red and E_1/2^Ox processes towards the cathodic region of the spectrum, while introducing a second reversible reductive event (Figure 3 red trace). In addition, the gap between these fully reversible processes was increasing, transitioning from E_gap = 2.12 V to E_gap = 2.24 V^[100]. Conversely, when an OMe ether group is added at the X position, a slight decrease in the E_1/2^Red potential (-60 mV) is observed, while the E_1/2^Ox value drops by 220 mV, resulting in a lower E_gap of 1.84 V (Figure 3 orange trace). Following this trend, the introduction of a primary amine NH₂ and a tertiary NMe₂ at the X position only minimally affects the value of the DMQA⁺ core reduction process, with a maximum difference of -20 mV for E_1/2^Red compared to ^nPrDMQA⁺. However, a drastic decrease in the value of E_1/2^Ox is noted in both cases, with a value of -510 mV for ^nPrDMQA^NH2+ and -560 mV for ^nPrDMQA^NMe2+, leading to E_gap values of 1.60 and 1.54 V, respectively (Figure 3 yellow and salmon traces).

Figure 3. Cyclic voltammograms of 1 mM ^nPrDMQA⁺, ^nPrDMQA^NO2+, ^nPrDMQA^OMe+,^nPrDMQA^NH2+,^nPrDMQA^NMe2+,^nPrDMQA^(pMe)3+,^nPr/PhDMQA⁺, ^nPr/CH2CF3DMQA⁺, ^(CyNHnPr)DMQA⁺, ^PEGDMQA⁺ and [6]helicene⁺ in 0.1 M TBAPF₆ CH₃CN solution at 100 mV·s^-1. All values of potential are reported in Table 1. Grey box corresponding to E_gap of ^nPrDMQA⁺.

When it comes to modifying the Y position of the aromatic core, options for late-stage functionalization are practically absent. The alteration of this position in the para of the carbocation center of tris-aryl carbenium synthons must take place during a preliminary step [Supplementary Figure 1]. Access to an electronically enriched scaffold through the introduction of OMe and NMe₂ groups was unsuccessful^[72,106]. Despite numerous attempts, the electron richness of acridinium moieties containing three σ- and π-donating groups, such as OMe and NMe₂, in the para position inhibits the second aromatic nucleophilic substitution required to form the helicenium scaffold^[83]. Fortunately, σ-donating groups such as methyl groups which were introduced in the para position to form the synthon tris(2,6-dimethoxy-4-methylphenyl)carbenium tetrafluoroborate did not hamper the synthesis of the helicenium analog DMQA^(pMe3)+ (see Supplementary Materials). This new compound sees its E_1/2^Red value decrease by -130 mV, while its E_1/2^Ox value only decreases by -60 mV compared to ^nPrDMQA⁺. This leads to an E_gap = 2.19 V, which is significantly comparable to the reference DMQA⁺ (Figure 3 green trace).

We then focus our interest on the nature of the arms carried by the DMQA⁺ scaffold and their impact on the electrochemical properties of the BRM. Through the introduction of a Ph group in R² through a preliminary synthesis step, followed by adding an nPr group in R¹, a dissymmetrical DMQA⁺ was obtained^[103]. It was quickly observed that the presence of a Ph aromatic group, while not significantly altering the potential of the electrochemical events, severely disrupts their reversibility (Figure 3 deep blue trace). The E_1/2^Red was recorded at -0.99 V vs. AgNO₃, a difference of +150 mV compared to the ^nPrDMQA⁺, while the E^Ox became completely irreversible. This initial drawback renders this molecule unsuitable for the deployment in SORFB. The substitution of the R² arm with an electrophilic trifluoroalkyl group CH₂CF₃ was then initiated to assess its influence on the electrochemical processes (Figure 3 sky blue trace). It was observed that the E_1/2^Red potential shifted towards more positive values by +340 mV, with the appearance of a second distinct process E_1/2^Red2. The E_1/2^Ox process was measured higher by +110 mV as compared to the reference DMQA⁺ but also became bielectronic, making use of ^nPr/CF3CH2DMQA⁺ as a BRM very complicated as the electron exchanges would be unbalanced. From a more structural perspective rather than electronic, the introduction of two bulky amino-alkyl R¹/R² arms (3-(cyclohexylamino)propyl) in order to increase the molecule's steric hindrance does not seem to induce any noticeable effect on the E^Red and E^Ox processes, which are modified by +160 and +50 mV, respectively (Figure 3 light blue trace). However, this ^(CyNHnPr)DMQA⁺ molecule experiences an alteration of both processes as they became pseudo-reversible. The pursuit of increased solubility of the different oxidation states of the DMQA⁺ core has also guided the synthesis of a variation where R¹ and R² were introduced in the form of an alkyl ether moiety, so-called pegyl chain (3-(2-methoxyethoxy)propyl = PEG)^[45,107]. Remarkably, this ^PEGDMQA⁺ molecule presents values of E_1/2^Red and E_1/2^Ox very similar to those of the reference ^nPrDMQA⁺, with potential variations less than -50 mV, allowing the preservation of an E_gap = 2.07 V (Figure 3 blue trace)^[108].

Finally, the rich chemistry of heliceniums^[109] prompted us to take an interest in dioxo[6]helicenium (labeled [6]helicene⁺ in Figure 3 and Table 1). [6]helicene⁺ is a parent helicenium of diaza[4]helicenium, the DMQAs presented thus far, devoid of R¹/R² arms where the bridging amino groups (NR) are replaced by a bridging oxygen atom. Expansion from [4] to [6]helicenes is provided by replacing the methoxy groups of the core with phenyl rings^[110]. This [6]helicene⁺ presents two completely reversible reduction phenomena with a first reduction at E_1/2^Red = -0.48 V vs. AgNO₃ at a much higher potential than the reduction of the DMQA⁺ model consistent with the less electron donating effect of the O vs. NR bridge groups (Figure 3 purple trace). However, no reversible electrochemical process is accessible in oxidation, consequently eliminating this molecule, as well as ^nPr/PhDMQA⁺, from use as a BRM, and these heliceniums will not be further discussed hereafter.

While the development of new BRMs for the deployment of SORFB is experiencing significant growth, it should be noted that a certain lack of coherence in the reporting and evaluation methods of their physicochemical specificities makes the comparison of new ROMs difficult^[111]. However, certain kinetic parameters need to be highlighted. Thus, parameters such as diffusion (D) and electron-transfer rate (k⁰) will greatly determine the expected performances for using a ROM in SORFB. Through effective mass transport of the different oxidation states of the BRM from the bulk solution to the electrode surface, higher current density can be achieved while limiting overpotential. The importance of the parameter k⁰ lies in its description of the efficiency of electron transfer from the electrodes to the ROM. Its value range also indicates the reversibility of the electronic process, with complete reversibility above 10^-1 cm·s^-1, quasi-reversibility between 10^-1 and 10^-5 cm·s^-1, and irreversibility below this range^[112]. While having the highest possible values of D and k⁰ is indeed preferable, it should be noted that in the context of BRMs and due to the symmetrical aspect of SORFBs and the use of a single molecule, it is also important that the values corresponding to the reduction and oxidation processes are in the same order of magnitude. This is to avoid any overpotential or loss of energy efficiency due to significant differences in kinetics during full flow deployment.

During the study of the reversibility of the electronic processes of each helicene, in-depth studies of the electrokinetic parameters D and k⁰ of each of them were conducted for practical reasons under the conditions of cyclic voltammetry. For this purpose, CVs of isolated electronic events were performed at seven different scan rates (10, 25, 75, 100, 250, 400, and 500 mV·s^-1). D determination was relying on the resolution of the Randles-Sevcik equations [Supplementary Equation 1] and k⁰ on the numerical application of the method developed by Nicholson et al. and Lavagnini et al. [Supplementary Equation 2]^[113,114].

The preliminary studies and results conducted on compounds ^nPrDMQA⁺ and ^nPrDMQA^NO2+ have already demonstrated good performances in static RFBs. Indeed, the reference carbenium with diffusion coefficients close to 1 × 10^-5 cm^-2·s^-1 and k⁰ at 2 × 10^-2 cm·s^-1, both very balanced, enabled the creation of the first SORFB model based on a [4]helicenium^[99]. In the case of ^nPrDMQA^NO2+, it was observed that the introduction of a nitro group on the DMQA⁺ scaffold had consequences on the potential of the electronic processes and on the value of D, and especially of k⁰. The electronic transfer in oxidation is five times less efficient in this case (k⁰_Red = 3.7 × 10^-2 cm·s^-1, k⁰_Ox = 0.8 × 10^-2 cm·s^-1), resulting in less cycling stability and robustness of the compound as published^[100]. It was observed that the introduction of electron-donating groups OMe, NH₂, and NMe₂ at the X position of the DMQA⁺ scaffold had little impact on the potential of the reduction process but contributed to a shift towards less oxidizing values of the E_1/2^Ox phenomenon. This is reflected in balanced D parameters with values ranging between 4-6 × 10^-5 cm^-2·s^-1 for ^nPrDMQA^NMe2+ and ^nPrDMQA^OMe+, while ^nPrDMQA^NH2+ exhibits values similar to that of the DMQA⁺ model in correlation with low structural volume change. Similarly, for these three compounds, the k⁰ values in reduction are comparable to those of ^nPrDMQA⁺ (~ 2 × 10^-2 cm·s^-1) but two to three times higher than those in oxidation processes (≤ 1 × 10^-2 cm·s^-1), an imbalance less pronounced than that of ^nPrDMQA^NO2+. The triple Me functionalization at the Y position leading to ^nPrDMQA^(pMe)3+ does not involve any significant changes in terms of the D parameter, with consistent values similar to those exhibited by the compounds substituted at X. However, it is noted that the k⁰ in reduction appears slightly higher than that of the [4]helicenium model, whereas the k⁰ of the oxidation process is more than an order of magnitude slower (k⁰_Ox = 0.3 × 10^-2 cm·s^-1), which could raise concerns about potential issues during high-rate cycling.

When the time comes to modify the R¹/R² arms without substituting the core, the results become significantly more pronounced. The asymmetric species, ^nPr/PhDMQA⁺ and ^nPr/CH2CF3DMQA⁺, are strongly affected, with R² = CH₂CF₃ leading to a decrease of one order of magnitude in the value of D for the reduction process, while the oxidation process maintains a value close to that observed for the previous compounds. This disparity is also evident in k⁰, with a reduction process sixteen times faster than in oxidation (k⁰_Red = 1.6 × 10^-2 cm·s^-1, k⁰_Ox = 0.1 × 10^-2 cm·s^-1); however, the fact that this phenomenon is bielectronic already eliminates this compound as a potential BRM. When R¹ = R², the D parameters behave more balanced, potentially related to the conservation of the molecule's symmetry. Thus, for ^(CyNHnPr)DMQA⁺, the diffusion parameters are in the range of values observed for the compounds substituted in X and Y, with an average of ~5 × 10^-6 cm²·s^-1. As anticipated during the CV of the compound, the E_1/2^Red process is less reversible than that of its counterparts, with a value of k⁰_Red = 0.5 × 10^-2 cm·s^-1, while the oxidation maintains its reversibility. In the case of ^PEGDMQA⁺, the diffusion parameters are strictly identical (6.2 × 10^-6 cm²·s^-1) for both electron exchanges. A noteworthy value of k⁰_Red = 2.5 × 10^-1 cm·s^-1 is noted for the reduction process, which unfortunately turns out to be twenty-two times faster than the oxidative process (k⁰_Red = 1.1 × 10^-2 cm·s^-1).

As most of the ROM's electrokinetic parameters presented here fall within the range of BRMs used in reported SORFBs, this prompts us to evaluate their robustness in cycling conditions. However, deployment in a complete in-flow RFB system is a task requiring numerous specific hardware adjustments, making it unsuitable for rapid screening of new compounds. That is why we aim to emphasize the practicality and relevance of evaluating new compounds in an H-cell. This model, described as “static RFB” in the literature^[37,102], has the advantage of requiring only a small volume and low working concentration while subjecting the electroactive materials to more stressful cycling constraints. This allows rapid and efficient selection and facilitates the elimination of certain candidates within an accessible timeframe.

The system used here is based on a homemade H-cell consisting of two welded glass tubes for each compartment (Figure 4A, details in Supplementary Materials) separated by a porous frit that serves as a membrane. This design is particularly suitable for BRMs as it allows us to determine whether the molecule can tolerate the use of a porous separator without ion crossover consideration. Each “pole” of the battery is equipped with a highly conductive RVC electrode that boasts a significant specific surface area (~33 cm² within the dimensions used in this study) and a magnetic stir bar, aiding in the diffusion of species in solution. The side where the working electrode (E_w) will be considered is also fitted with an AgNO₃ reference electrode, similar to the one previously employed in the characterization of a three-electrode cell.

Figure 4. (A) Picture of homemade H-cell “static RFB”, full details available in Supplementary Figure 9. (B) Scheme of the H-cell in initial and in charged state (reduction occurring at E_w) loaded with 2 × 5 mL of 1 mM ^R1/R2DMQA^X/Y+ in 0.1 M TBAPF₆ CH₃CN. Display of 134 μAh H-cell cycling monitoring via (C) Capacity in discharge and (D) Coulombic efficiency of ^nPrDMQA⁺, ^nPrDMQA^NO2+, ^nPrDMQA^OMe+,^nPrDMQA^NH2+, ^nPrDMQA^NMe2+, ^nPrDMQA^(pMe)3+, ^(CyNHnPr)DMQA⁺ and ^PEGDMQA⁺. Supplementary Figures 10-17 provide detailed monitoring of the cycling data for each helicenium individually.

Upon the selection of [4]helicenium exhibiting electrokinetic parameters compatible with deployment for RFB, each of them is tested at a concentration of 1 mM in acetonitrile in the presence of 0.1 M TBAPF₆(0.5 M TBABF₄ in the case of ^PEGDMQA⁺). Each side of the H-cell is filled with 5 mL of the same 1mM solution to conduct a test battery with a capacity of 134 μAh. Thus, in the initial state, each side of the static RFB contains the same BRM in the DMQA⁺ oxidation state. During charging, an electron is transferred from the counter electrode side (E_c) to the E_w side, allowing the formation of DMQA^●++ on the “posolyte” side, while the carbocation on the “negolyte” side is reduced to DMQA^● [Figure 4B]. The repetition of these charge-discharge cycles in our RFB model constitutes cycles.

The cycling protocol for each of the compounds involves charging and discharging at 90% of the theoretical capacity at |5| mA. Due to its static design, this type of cell primarily relies on the diffusion of species in solution to access the entirety of the electron exchange. Hence, a constant current followed by a constant voltage (CCCV) protocol was employed. This setup was developed with a reference electrode to control the voltage at Ew, and therefore, the potential limits reached will be E_1/2^Red ± 300 mV. In the static RFB, the key metrics of interest are the discharge capacity (Q_dis) achieved and the coulombic efficiency (CE) attained during the cycles. Thus, it is observed that ^nPrDMQA⁺ maintains a consistent Q_dis = 100% until the 461st cycle, followed by a slow degradation while maintaining a near-perfect CE (Figure 4C and D, black). Similarly, ^nPrDMQA^NO2+ exhibits excellent CE, with a constant 100% Q_dis until the 186th cycle, after which a decrease in discharge capacity is noted until the end of cycling at the 400th cycle (Figure 4C and D, red). These findings demonstrated the relevance of this evaluation system, prompting our subsequent interest in ^nPrDMQA^OMe+. While CE is decreasing at 0.23%/cycle from an acceptable 97% value, it is noticeable that from the first cycle, the Q_dis is below 75% and drops rapidly at an average rate of 0.49% per cycle, making this ROM, although electrokinetically adapted, unsuitable for a SORFB system (Figure 4C and D, orange). A more surprising behavior is observed with the electron-enriched ^nPrDMQA^NH2+, with an initial Q_dis value of 96%, which remains stable for around forty cycles before initiating a decay at a rate of 0.28%/cycle. Meanwhile, its CE values appear slightly erratic, suggesting reactivity in the charged state, with an initial value of 96% and an average decay of 0.25%/cycle (Figure 4C and D, yellow). The transition from a primary amine to a tertiary amine for ^nPrDMQA^NMe2+ has a dramatic effect on its ability to be used as a BRM. It has thus been observed that the Q_dis, in its case, is initially at 15%, dropping to 0% by the tenth cycle, illustrating its complete incompatibility for use in SORFB (Figure 4C and D, salmon). The introduction of three Me groups in the Y position appeared to hold promise in terms of the electrokinetic characteristics of ^nPrDMQA^(pMe)3+. However, when the cyclability of this new DMQA⁺ is considered, the expected results are not met as only 46% of Q_dis is accessible, with an average decrease in capacity value of 0.3% per cycle. This implies that more than half of the charge has been lost between charging and discharging. This is also reflected in a CE value below 50%, which drops to 35% after ten cycles, stabilizing with a loss rate of 0.008% per cycle (Figure 4C and D, green). Although no specific sub-species could be clearly identified, the regrettable outcome in bulk electrolysis is attributed to the intrinsic radical characteristics of the DMQA^●++ and DMQA^● species. The existence of three possible benzylic radical positions, featuring heightened reactivity stemming from resonance stabilization, results in a multitude of decomposition pathways and undesired possible reactions when generated^[115].

The modification of the DMQA⁺ core has a significant impact on the electrophysical characteristics of ROMs, as demonstrated above. However, it seems that the variation in the nature of the R¹/R² arms - when allowing the conservation of two entirely reversible electrochemical events - has little effect on the electrokinetic potentials and parameters of this BRM class. However, when the cycling of ^(CyNHnPr)DMQA⁺ is assessed, the result is disastrous. The Q_dis is < 0.3%, and the CE value is null, illustrating how an evaluation in a three-electrode cell is insufficient for characterizing the suitability of a new BRM for SORFB. Thus, bulk electrolysis, such as static RFB cycling, is far too damaging for ^(CyNHnPr)DMQA⁺ (Figure 4C and D, cyan). Finally, the proposal to integrate pegyls arms seemed promising, as the CE value is > 98% and remains nearly constant up to cycle 200. Subsequently, a decay of 0.24% per cycle appears and can be correlated with what is observed in terms of Q_dis values. Indeed, it is observed that ^PEGDMQA⁺ exhibits two decay regimes: the first, from cycle 5 to 200, shows its Q_dis drop from 98% to 78% at a rate of 0.10% per cycle, and then, at cycle 201, this phenomenon accelerates with a decline of 0.74% per cycle (Figure 4C and D, deep blue). This behavior is not easily rationalized, but it is worth remembering that for effective discrimination, the H-cell system and the current intensities used here are extremely stressful for the electroactive material. A CE value and Q_dis access rates exceeding 95% still make it a convincing BRM that deserves to be evaluated in a flow system.

CONCLUSIONS

The significance of this research work resides in the practicality of an empirical methodology, selecting an ambipolar scaffold and simple electrochemistry tools, to effectively identify and screen a variety of new BRMs. The electrochemical richness of the [4]helicenium class has illustrated how the choice of a tunable core is a rapid pathway to numerous bipolar compounds with a wide range of possible substitutions. The qualification through measurements of essential electrochemical parameters, such as diffusion coefficients and electron transfer rate constants, could be carried out using a simple three-electrode cell, making this process easily accessible. Finally, the assessment of the robustness of these BRMs in a static RFB cell constitutes a quick tool for discriminating the most promising and robust ROMs in bulk electrolysis, with the future aim of deployment in a flow RFB cell. Through the importance of this research effort, we hope to assist research groups working on the development of new BRMs in accessing promising systems more swiftly.

A potential future enhancement of this methodology could involve the incorporation of Machine Learning tools capable of predicting the characteristics of potential BRMs through computational design, based on the collection of extensive data^[116,117]. This would require active participation from the community in building a large database, but it has the potential to predict the entirety of electrokinetic parameters, decomposition pathways, and potential improvements for existing bipolar molecules.