Raman spectroelectrochemistry for operando characterization of redox flow batteries

Raman spectroelectrochemistry for operando characterization of redox flow batteries Lara Lubian a,b , Rub´ en Rubio-Presa a,b , Virginia Ruiz a,b , Alvaro Colina b,* , Edgar Ventosa a,b,** a International Research Center in Critical Raw Materials-ICCRAM, Universidad de Burgos, Plaza Misael Ba˜ nuelos s/n, E-09001, Burgos, Spain b Departamento de Química, Facultad de Ciencias, Universidad de Burgos, Pza. Misael Ba˜ nuelos s/n, E-09001, Burgos, Spain GRAPHICAL ABSTRACT ARTICLE INFO Keywords: Redox flow batteries Raman spectroelectrochemistry Operando characterization Degradation mechanism ABSTRACT Despite the potential of Aqueous Organic Redox Flow Batteries (AORFBs) to address intermittent energy generation from renewable sources, they must improve some key performance indicators to become competitive, in particular cycle stability. Development of advanced in-situ and time-resolved techniques plays a critical role to improve performance of AORFBs by enabling elucidation of the sources for energy storage capacity fading. The development and implementation of operando Raman spectroscopy is herein reported for dihydroxyanthraquinone–ferrocyanide alkaline flow battery. Validation of the technique is carried out using symmetrical cells, confirming that Raman spectroscopy can monitor in-situ the state of charge. In a full battery, time-resolved Raman spectroscopy is used for investigating the Faradaic imbalance process, showing that presence of oxygen in the anolyte leads to the progressive loss of available ferrocyanide. Raman spectroscopy also shows that ferricyanide self-discharges when left at open circuit. Finally, this technique is used to track in-situ the crossover of 2,6-dihydroxyanthraquinone. Surprisingly, the rate is found to increase at full state of charge, supporting recent * Corresponding author. ** Corresponding author. International Research Center in Critical Raw Materials-ICCRAM, Universidad de Burgos, Plaza Misael Ba˜ nuelos s/n, E-09001, Burgos, Spain. E-mail addresses: [email protected] (A. Colina), [email protected] (E. Ventosa). Contents lists available at ScienceDirect Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour https://doi.org/10.1016/j.jpowsour.2025.237272 Received 1 August 2024; Received in revised form 17 March 2025; Accepted 1 May 2025 Journal of Power Sources 646 (2025) 237272 Available online 8 May 2025 0378-7753/© 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC license ( http://creativecommons.org/licenses/bync/4.0/ ). findings using NMR. This operando Raman spectroscopy is anticipated to provide unique insights into critical processes in emerging redox flow battery chemistries. 1. Introduction Energy generation from renewable sources has become crucial to achieve decarbonization and the transition to a more sustainable energy production [1]. It is expected that energy from renewable sources will account for approximately 25 % of the global electricity supply by 2030 [2]. However, the unpredictable and intermittent nature of renewable energy sources such as solar or wind power requires efficient and cost-effective energy storage systems (ESSs) to match energy production and demand [3–5]. Redox flow batteries (RFB) are especially suitable for large-scale stationary energy storage, with competitive advantages over conventional batteries including cost-effectiveness, long cycle life and independent scalability of energy and power [5,6]. The all-vanadium redox flow battery (AVRFB) is currently the state-of-art RFB yet it suffers from drawbacks such as the use of a critical raw material and highly corrosive strongly acidic solutions [5]. This has triggered interest in aqueous organic redox flow batteries (AORFB), where inorganic redox species are replaced by more sustainable and abundant organic redox-active compounds [5,7–10]. Several families of organic molecules have been recently researched as redox actives species in AORFB such as anthraquinone [11–13], phenazine [14–16] or fluorenone [17] derivatives, which have proven to deliver high performance as anolytes in alkaline media. In turn, viologen derivatives exhibit promising performance in terms of energy density and cycle stability at neutral pH [4]. While many of those organic compounds feature suitable redox potentials, fast kinetics and high solubility, their cycle stability should be further improved for large-scale deployment [4,7]. The progressive decay in energy storage capacity in a redox flow battery is usually referred to as capacity fading. There are three main sources of capacity fading in AORFB [18]. The first one includes degradation of the active species due to incompatibility among molecules or intrinsic decomposition [4,19,20]. Another source of capacity loss is related to the cell design, such as membrane degradation electrolyte leakage or crossover [21]. The third source of capacity fading is charge (faradaic) imbalance between both sides of the RFB due to parasitic side reactions such as hydrogen and oxygen evolution or oxygen reduction [18]. Identifying the contribution of each source is crucial to develop strategies to prevent or minimize capacity fading [18,22]. This requires advanced in situ characterization techniques to investigate the performance of redox flow batteries operando conditions, as ex-situ techniques may lead to misinterpretations in case of short-lived redox active species or highly reactive molecules [23]. So far, techniques such as Ultraviolet–Visible (UV–Vis) absorption [24–26], IR [27], electron paramagnetic resonance (EPR) [28] and nuclear magnetic resonance (NMR) [23,28,29] spectroscopies have been used for in situ characterization of RFB to investigate reaction mechanisms and degradation processes of organic redox species in AORFBs. Among the analytical techniques complementary to the operando techniques established so far, time-resolved Raman spectroscopy can provide real-time information about the chemical composition of the redox-active molecules in the electrolytes, as Raman spectra are like fingerprints of the compounds. Raman spectroscopy that has shown to be a powerful in situ technique for non-flow battery research [30–32] did not attract much attention in the field of RFBs. Indeed, there are only few examples employing Raman spectroscopy for redox flow battery research. For example, Raman spectroscopy was used in a static cell to study vanadyl-ion oxidation mechanisms on carbon paper electrodes [33]. This technique was also used to evaluate the evolution of vanadium electrolyte by taking samples at different state of charge [34,35]. Also, Raman spectroscopy has been applied to monitor the evolution of ZnBr concentration upon cycling in a conventional zinc-bromine RFB [36]. Despite the interesting results, this powerful characterization technique has not been applied for time-resolved investigation under operando conditions in a flow battery for the emerging field of AORFBs. In this report, we use for the first time dynamic and time-resolved Raman spectroscopy for characterization of an AORFB operando conditions. A tailored optical flow cell for in situ Raman confocal microscopy measurements was designed, 3D printed and its performance validated with the widespread used potassium ferrocyanide/ferricyanide redox couple. We illustrate the power of time-resolved Raman spectroelectrochemistry for in situ evaluation of AORFB with an anthraquinone derivative//ferrocyanide alkaline redox flow battery. The technique allowed detecting loss of ferrocyanide due to the accumulation of charged species (ferricyanide) in the catholyte caused by oxygen-induced parasitic reactions at the negative compartment, corroborating that the capacity fading observed in the AORFB was due to Faradaic imbalance between both half-cells. Furthermore, in situ Raman spectroscopy corroborated the mechanism of an electrochemical charge balancing strategy implemented in an imbalanced AORFB for reverting the effects of the side reaction and mitigating capacity fading. We believe that the proposed examples illustrate the promise of timeresolved Raman spectroscopy for operando evaluation of AORFB performance, a technique that can contribute to a faster large-scale deployment of AORFB by providing better understanding of degradation processes that limit their cycling stability. 2. Experimental section 2.1. Materials Potassium ferrocyanide (II) trihydrate, potassium ferrocyanide (III) (>98 %, Thermoscientific) and KOH (VWR) were used as received. The synthetic procedure of 2,6-dihydroxantraquinone (2,6-DHAQ) was previously reported (Section S1). 2.2. Preparation of electrolytes In the symmetric K 3 Fe(CN) 6 //K 4 Fe(CN) 6 flow cell, the catholyte was prepared by dissolving potassium ferrocyanide in 1 M KOH to afford 13 mL of 0.3 M ferrocyanide electrolyte. The anolyte was prepared by dissolving potassium ferrocyanide and potassium ferricyanide in 1 M KOH to afford 45 mL of 0.1 M ferrocyanide and 0.2 M ferricyanide electrolyte (oversized counter-compartment). In the 2,6-DHAQ//K 4 Fe(CN) 6 flow batteries, the catholytes were prepared by dissolving potassium ferrocyanide in 1 M KOH to afford 14 mL of 0.3 M ferrocyanide electrolyte in all full batteries. The 2,6-DHAQ sample systematically delivered 1.75 exchanged electrons as shown in Fig. S1, Section S1 (large excess of catholyte). This is attributed to 12.5 wt% water remaining after the synthesis. As for the anolyte, experiments using two different volumes were employed: i) 10 mL of 0.2 M 2,6DHAQ in 1.4 M KOH electrolyte leading to a 20 % excess of potassium ferrocyanide (for 1.75 exchanged electrons for the 2,6-DHAQ), and ii) 12 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH electrolyte leading to a system with a balanced amount of charge for catholyte and anolyte. In fact, the latter leads to 5 % excess of as it is very difficult to obtain a Coulombic efficiency above 95 % in the first cycles since oxygen adsorbed on the carbon felt and traces of oxygen in the electrolyte are extremely difficult to be removed without an Ar-filled glovebox. Anolyte was purged with argon prior to use. All electrolytes were prepared with deionized water. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 2 2.3. In situ Raman spectroelectrochemistry setup 2.3.1. Flow battery A filter-pressed flow cell using expanded graphite (SGL Carbon), graphite felt (SGL Carbon) and Nafion 212 (Ion Power) as current collector, electrode and ion selective membrane, respectively was used. The projected area of the cell was 10 cm 2 . A peristaltic pump (MasterFlex L/ S) was used to provide a flow rate of 24 mL min −1 . 2.3.2. Raman flow cell The Raman flow cell (Fig. 1) was designed using the SketchUp software and manufactured using an Ultraviolet (UV) Liquid Cristal Display (LCD) - based stereolithography (SLA) 3D printer (Photon Mono SE, Anycubic) and a commercial clear resin (Anycubic). After the cleaning procedure (with 70 % isopropanol solution), the printed pieces were assembled with a PET window sandwiched between two Viton® gaskets (1 mm thick) and all the pieces were held together with screws (section 3.1, Fig. 1). Dimensions of the Raman flow cell were 10 x 8 ×3 cm 2 . 2.4. Spectroelectrochemical characterization 2.4.1. Electrochemical characterization Galvanostatic charge-discharge measurements were conducted using a potentiostat/galvanostat PGSTAT302 (AUTOLAB). The batteries were galvanostatically cycled at ±20 mA cm −2 with voltage limits of ±0.3 V for the K 3 Fe(CN) 6 //K 4 Fe(CN) 6 symmetric flow cell and +1.6 V/+0.5 V for the 2,6-DHAQ//K 4 Fe(CN) 6 flow battery. For the electrochemical rebalancing protocol, the upper voltage limit was increased to +2 V. 2.4.2. Raman spectroscopy characterization Raman spectra were measured using a confocal Raman microscope Voyage (BWTEK). A 20 ×objective was used, with an excitation line at 532 nm and a power of 5 mW. Raman spectra were collected by a CCD array with a spectral resolution of 3.8 cm −1 . Sampling interval was 30 s for cycling the K 3 Fe(CN) 6 //K 4 Fe(CN) 6 symmetric flow cell and for the 2,6-DHAQ//K 4 Fe(CN) 6 flow battery. The integration time for each spectrum was 3000 ms. Matlab was used to analyze, plot and correct the baseline of measured Raman spectra. Raman spectrometer is synchronized with the potentiostat using a pulse trigger in order to obtain concomitantly the optical and electrical response. 3. Results and discussion 3.1. Design and validation of the in situ Raman spectroelectrochemical setup As indicated in the introduction, to the best of our knowledge, timeresolved Raman spectroscopy had not been previously conducted in situ during electrochemical cycling of AORFB for real-time monitoring of active species or degradation products operando. Aware of the great promise of this technique for exploring degradation processes causing capacity fading in AORFB, we have developed a Raman flow cell and integrated it in the flow system of a K 3 Fe(CN) 6 //K 4 Fe(CN) 6 symmetric flow cell and a 2,6-DHAQ//K 4 Fe(CN) 6 flow battery to follow the evolution of active species during cell/battery operation. The experimental setup for in situ Raman spectroelectrochemical measurements with AORFB is shown schematically in Fig. 1A. The Raman flow cell was mounted on the confocal Raman microscope holder at the inlet of the electrochemical reactor in the catholyte side to investigate the state of charge of the electrolyte in the tank. The detailed configuration of the 3D printed Raman flow cell is shown in Fig. 1B. It essentially consists of two printed parts made of resin (one piece having a flow-through and the other having an opening), a rectangular piece of PET as optical window and two Viton® gaskets (1 mm thick) to seal the cell. Dimensions of each part are given in Fig. S2 (plans). The printed components were carefully designed to ensure proper electrolyte flow, avoid solution leakage and maximize measured Raman intensity. On the one hand, the tailored shape of the flow channel (2.5 mL) in the bottom flowthrough piece, widening at the inlet and narrowing at the outlet (see cross-section, Fig. S2) allows total filling and emptying of the chamber, thus preventing retention of active species inside it that would result in capacity fading of the battery. On the other hand, the top optical piece was printed with a conical cavity of the right size (see cross-section, Fig. S2) to insert the Raman microscope objective and allow moving it in the z-axis to focus the laser beam on different planes of the flow channel for optimal signal monitoring. The cavity depth in the optical piece is Fig. 1. (A) Schematic illustration of the experimental setup for in situ Raman spectroelectrochemical measurements at the catholyte side of an AORFB, and (B) a scheme showing a breakdown of the Raman cell. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 3 large enough to prevent collision of the objective with the PET window. The PET optical window is sandwiched between two rectangular Viton® gaskets with the same size as the printed top and bottom cell parts. An opening in the bottom gasket (Viton A) with the same shape and size as the uppermost plane of the bottom printed flow piece allows the solution reaching the optical window. The laser beam passes through a circular central hole in the top gasket (Viton B). All cell components except the PET window have holes at their edges for the screws to keep all the pieces joined and the cell tightly sealed. A previously calibrated torque wrench was used to ensure applying a homogeneous pressure to all the screws. An initial validation of the Raman flow cell and the whole experimental setup for in situ Raman spectroelectrochemical characterization of RFB was done during galvanostatic cycling of a symmetric potassium ferrocyanide/ferricyanide alkaline redox flow cell. This model redox system was selected because it is one of the most widely used aqueous catholytes in AORFB and also to explore the stability of the developed flow cell under alkaline conditions. First, the Raman flow cell was calibrated using aqueous solutions with different concentration ratios of potassium ferrocyanide/ferricyanide in 1 M KOH (Section S3). Both species of the redox couple have very distinct Raman features. While the Raman spectrum of potassium ferrocyanide (Fig. S4) exhibits two bands centered at 2052 and 2090 cm −1 corresponding to the stretching vibration of cyanide bonds ν (C ≡N), the spectrum of potassium ferricyanide (Fig. S4) shows only one cyanide vibrational band at 2124 cm −1 [37]. Calibration curves obtained for the intensity of characteristic Raman bands of both potassium ferrocyanide and ferricyanide (Fig. S5) reveal high linearity of the optical signal with analyte concentration for all bands. The volume of the anolyte compartment (45 mL, 0.2 M potassium ferricyanide) of the ferrocyanide/ferricyanide symmetric flow cell and, thus, the amount of active species and capacity was significantly increased so that the catholyte compartment (13 mL, 0.2 M potassium ferrocyanide), where Raman spectra were monitored, became the limiting side. Thus, the electrochemical data can be attributed to the catholyte limiting compartment, which becomes the “working compartment”, while the large anolyte compartment acts as “countercompartment”. Moreover, using a large excess of active species in the Fig. 2. In situ Raman spectroelectrochemical characterization of a symmetric ferrocyanide/ferricyanide alkaline redox flow cell operando during galvanostatic cycling at ±20 mA cm −2 with voltage limits of ±0.3 V. (A) Time evolution of cell voltage and Raman intensity at 2090 cm −1 during the first charge/discharge cycle. (B) Raman spectra of the catholyte monitored at the cycle times indicated in (A). (C) Charge/discharge capacity and Raman intensity at 2090 cm −1 measured in situ during cycling. (D) Evolution of the normalized capacity retention and the normalized Raman intensity at 2090 cm −1 . Catholyte: 13 mL of 0.3 M potassium ferrocyanide in 1 M KOH, anolyte: 45 mL of 0.1 M ferrocyanide and 0.2 M ferricyanide in 1 M KOH. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 4 counter-compartment reduces capacity fading due to occurrence of the OER, which cannot be neglected in alkaline media [22]. Time evolution of cell voltage and Raman intensity at 2090 cm −1 (characteristic of potassium ferrocyanide) during the first galvanostatic charge/discharge cycle are plotted together in Fig. 2A. As can be seen, electrochemical and spectroscopic signals are totally correlated. At the start of the cycle, the Raman intensity at this wavenumber (Fig. 2A, point 1) confirmed the presence of potassium ferrocyanide in the catholyte (which contained 0.2 M potassium ferrocyanide). The full spectrum measured (Fig. 2B–1) corresponds to the starting uncharged potassium ferrocyanide. Upon charging, charged potassium ferricyanide starts to be produced at the catholyte side, giving rise to a decrease of the Raman intensity (Fig. 2A, point 2) that reaches a minimum value at the end of the charging step (Fig. 2A, point 3) when all potassium ferrocyanide has been oxidized to ferricyanide. Full Raman spectrum measured at half state of charge (Fig. 2B-2) exhibits three bands, revealing the presence of both ferrocyanide (bands at 2052 and 2090 cm −1 ) and ferricyanide (band at 2124 cm −1 ). Consumption of initial ferrocyanide upon charging is evidenced by the lower intensity of its characteristic bands while emerging of the new band at 2124 cm −1 proves concomitant ferricyanide generation. The spectrum monitored at full state of charge (Fig. 2B–3) corresponds to ferricyanide alone, as the characteristic bands of ferrocyanide have completely disappeared. Upon discharging, ferricyanide is reduced back to ferrocyanide and hence, the intensity of its characteristic Raman band increases until reaching the initial value at the beginning of the cycle (Fig. 2A, point 4). The full spectrum at the end of the cycle (Fig. 2B–4) overlaps with the initial one (Fig. 2B–1), corroborating that the flow cell catholyte is fully discharged. As expected, Raman intensity of the ferricyanide band follow exactly the opposite trends during the charge/discharge cycle (Fig. S6C), increasing in the charge step and returning to the initial value at the end of the discharge step, confirming full interconversion between both active species of the redox pair. After validating the Raman spectroelectrochemistry setup to follow in situ the state of charge in the catholyte of a symmetric ferrocyanide/ ferricyanide alkaline redox flow cell in only one cycle, the technique was applied to assess the cycling stability of this flow cell. Evolution of charge/discharge capacity and Raman intensity at 2090 cm −1 (ferrocyanide band) measured in situ during cycling (Fig. 2C) reveals a remarkable correlation between electrochemical and optical responses. Thus, the 14 % drop in capacity noted after the first cycle was accompanied by a comparable decrease (16 %) of Raman intensity amplitude of the charge. Both capacity and Raman intensity remained roughly stable for the rest of the cycles due to the high stability of the ferrocyanide/ferricyanide redox pair in alkaline media under the operation conditions. This correlation between electrochemical and optical responses can be easily observed in Fig. 2D by comparing the normalized capacity retention (normalized by the capacity in the first cycle) with the Raman intensity retention (normalized by the intensity in the first cycle). The excellent correlation between optical and electrochemical signals for ferrocyanide in a simple symmetrical cell confirms that this signal (2090 cm −1 ) can be used in more complex systems to monitor the state of charge of the catholyte. It should be noted that similar information can be inferred from the additional Raman signals of ferrocyanide (2052 cm −1 ) and ferricyanide (2124 cm −1 ) that are shown in Figs. S6 and S7. The following sections will focus on the Raman signal at 2090 cm −1 (potassium ferrocyanide) to simplify the explanations. 3.2. Probing capacity imbalance of a 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB by in situ Raman spectroscopy After demonstrating the capability of the in-situ Raman spectroelectrochemistry setup for operando characterization of a symmetric redox flow cell, a very relevant use case was selected to explore applicability of the technique for AORFB testing. Specifically, the setup was used to probe capacity fading due to Faradaic imbalance during cycling stability assessment of a state-of-art alkaline flow battery chemistry, namely 2,6-DHAQ//K 4 Fe(CN) 6 . Faradaic imbalance refers to the uneven Fig. 3. Schematic illustration of: (A) a redox flow battery suffering of parasitic side reaction related to the presence of oxygen in the negative compartment, (B) the concept of battery imbalance during the first charge and (C) discharge cycle in a flow battery by oxygen-induced parasitic reactions or hydrogen evolution in the negative electrode. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 5 state of charge of catholyte and anolyte triggered by parasitic side reactions that irreversibly consume charges under the operating conditions such as the reaction of oxygen with reduced active species in the negative electrode [18]. It should be noted that crossover of active species through the membrane also leads to the imbalance of available charge between anolyte and catholyte, but this process does not lead to an accumulation of charged species in one of the sides which is the main process to be monitored in this case study. Thus, crossover is not further discussed in this explanation that is focused on Faradaic imbalance leading to accumulation of charged species. In order to prevent these oxygen-induced parasitic reactions and the associated irreversible capacity losses in long-term cycling, AORFB are usually operated in argon-filled gloveboxes for academic research. However, this approach is not practical for an eventual large-scale implementation of AORFB, which urges to seek operando analytical tools for providing insights into side reactions or degradation processes leading to battery charge capacity fading [18]. Inevitable hydrogen evolution reaction and oxygen evolution reaction for active species having redox potentials beyond the stability window of the electrolyte have the same consequences on the Faradaic imbalance. To explain the process, a hypothetical case is discussed in which the values are arbitrarily taken. Fig. 3 illustrates schematically the concept of battery imbalance during the first charge/discharge cycle in a flow battery due to oxygen-induced parasitic reactions in the negative electrode, The anolyte (N, 2,6-DHAQ in our case) suffering oxygen-induced parasitic reactions will not be fully charged (e.g. 90 %) as some charges (e.g. 10 %) are irreversibly consumed by the parasitic reaction (HER or ORR), while the catholyte (P, K 4 Fe(CN) 6 in our case) will be able to be fully charged (e.g. 100 %). The occurring of parasitic reactions (HER and ORR) in the anolyte irreversibly consumed charge leading to a Faradaic imbalance in the state of charge of anolyte and catholyte at the end of the charge step (Fig. 3B). During the discharge process, the discharge capacity will be limited by the state of charge achieved in the anolyte (e. g. 90 %), hence full discharging will not be achieved in the catholyte, which will remain partially charged at the end of the discharge step (e.g. 10 %), as illustrated in Fig. 3C. This will reduce the capacity of the battery in the following charge step as now only 90 % of the catholyte species can be charged. Moreover, the anolyte will not be fully charged again due to oxygen-induced parasitic reactions, thus resulting in an increasingly larger mismatch between the states of charge of anolyte and catholyte with the number of cycles as long as oxygen is present, as illustrated schematically for the first three charge/discharge cycles in Fig. S8. It should be noted that the occurrence of hydrogen evolution reaction would have the same effect. In our study case of the 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB, an oversized tank for the catholyte with 20 % excess of potassium ferrocyanide was used, for 1.75 exchanged electrons (as explained in the experimental section) for the 2,6-DHAQ (94 mAh). According to the results of the first 2 cycles (Fig. 4A), the capacity (94.7 mAh) was indeed limited by the anolyte (theoretical capacity of 94 mAh and 112.5 mAh for anolyte and catholyte, respectively), while the excess of ferrocyanide enabled maintaining a stable capacity for the first cycles. However, the excess of ferrocyanide is quickly consumed by the remaining amount of oxygen in the electrolyte, adsorbed oxygen on the graphite felts and/or leakage of small amount of oxygen in the negative compartment leading to a poorer Coulombic efficiency in the first cycles. In other words, ferrocyanide oxidation is accompanied by parasitic oxygen reduction rather than 2,6-DHAQ reduction, leading to a decrease in Coulombic efficiency. After 3 cycles, both charge and discharge capacity started to decrease steadily with the number of cycles (Fig. 4A) leading to only 60.1 % capacity retention after 25 cycles. Based on the Coulombic efficiency of ca. 97 %, a leakage of small amount of oxygen in the negative compartment is likely the origin of the capacity fading. Although this fading mechanism is plausible, an in-situ technique is needed be provide evidence and support such a hypothesis. Therefore, accumulation of charged active species and the consequently increasing deficit of ferrocyanide at the beginning of each charge in the catholyte during battery operation was explored operando by Raman spectroscopy. Evolution of Raman intensity at 2090 cm −1 (characteristic of potassium ferrocyanide) measured in situ in the catholyte during the first galvanostatic charge/discharge cycle is plotted in Fig. 4B together with the voltage profile. Full Raman spectra of the catholyte measured at the beginning and end of the charge step (points 1 and 2, respectively) and at the end of the discharge step (point 3) are shown in Fig. 4C. At the start of the cycle, the measured spectrum (Fig. 4C–1) corresponded to uncharged potassium ferrocyanide, bands at 2052 and 2090 cm −1 (Fig. 4B, point 1), hence the null Raman intensity at 2124 cm −1 (Fig. 4C, point 1). Upon battery charging, potassium ferricyanide started to be produced at the catholyte side and Raman intensity at the characteristic wavenumber of ferrocyanide decreased until reaching a minimum value at the end of the charging step (Fig. 4B, point 2). Spectrum measured at this point (Fig. 4C–2) displayed not only the band of charged potassium ferricyanide at 2124 cm −1 but also small-intensity bands at 2052 and 2090 cm −1 , revealing the presence of a small amount of uncharged potassium ferrocyanide as we used an oversized tank for the catholyte with 20 % excess of ferrocyanide to compensate charge consumption by oxygen-induced parasitic reactions at the anolyte. Note the difference with the spectrum monitored at the end of the charge step in the case of the symmetric ferrocyanide/ferricyanide flow cell with oversized counter-compartment (Fig. 2B–3), where the characteristic bands of ferrocyanide disappeared completely as potassium ferrocyanide was fully oxidized to Fig. 4. In situ Raman spectroelectrochemical measurement for a 2,6-DHAQ//K 4 Fe(CN) 6 alkaline redox flow battery during galvanostatic cycling at ±20 mA cm −2 with voltage limits of +1.6/+0.5 V. (A) Discharge capacity and Coulombic efficiency versus number of cycles. (B) Time evolution of cell voltage and Raman intensity at 2090 cm −1 measured in situ in the catholyte during the first charge/discharge cycle. (C) Raman spectra of the catholyte corresponding to the cycle times indicated in (B). Catholyte: 14 mL of 0.3 M ferrocyanide in 1 M KOH; anolyte: 10 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 6 ferricyanide. In the discharge process of the 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB, the Raman fingerprint of ferrocyanide increased during the whole step without reaching the initial value at the start of the cycle (Fig. 4B, point 3). The full spectrum (Fig. 4C–3) confirmed the presence of accumulated charged ferricyanide in the catholyte at the end of the discharge step (and hence less ferrocyanide than initially), providing the spectroscopic evidence of battery Faradaic imbalance from the first cycle due to irreversible charge consumption at the anolyte that prevented this “capacity limiting side” reaching its full state of charge. Indeed, the Raman signal indicates a 12 % accumulation of charge in the first cycles, which is in good agreement with the Coulombic efficiency. Importantly, the Raman signal confirms that Faradaic imbalance is triggered by side reactions in the negative electrode. Overall, operando Raman spectroelectrochemical characterization of catholyte active species during cycling stability testing of the 2,6DHAQ//K 4 Fe(CN) 6 alkaline RFB showed remarkable correlation between electrochemical and optical responses (Fig. S9). Voltage profiles are plotted together with the Raman intensity of the ferrocyanide band (2090 cm −1 ) measured versus the number of charge/discharge cycles in Fig. 5. While capacity gradually faded and cycles became shorter, the spectroscopic signal revealed progressive loss of potassium ferrocyanide after each discharge cycle, as Raman intensity did not return to its initial value (green circles in Fig. 5). Furthermore, the excess of ferrocyanide is consumed by cycle 3 (purple square in Fig. 5)., as the band of ferrocyanide disappears at the end of the 3rd charge step (Fig. S10), which is consistent with the beginning of the capacity fading (Fig. 4A). Similar information can be inferred from the additional Raman signals of ferrocyanide (2052 cm −1 ) and ferricyanide (2124 cm −1 ) that are shown in Fig. S11. 3.3. Probing electrochemical charge balancing of a 2,6-DHAQ//K 4 Fe (CN) 6 alkaline RFB by operando Raman spectroscopy Once demonstrated by operando Raman spectroscopy that Faradaic imbalance in a 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB leads to the loss of potassium ferrocyanide in the catholyte by progressive accumulation of ferricyanide, the technique was used to shed light on the fundaments of reverting the battery Faradaic imbalance by applying a simple electrochemical charge balancing strategy recently proposed [18]. In a previous work, our group proved that the effects of the side reaction could be reverted by an electrochemical rebalancing procedure, thus mitigating the capacity fading and prolonging the cycling performance of AORFB (remarkable 20-fold reduction of capacity fading). While detailed discussion on the implication of the method (capacity fading over long-term, corrosion, energy efficiency, etc) can be found elsewhere [18], the fundaments of the processes involved were not unraveled due to the lack of suitable operando techniques. Therefore, the implementation of operando Raman spectroscopy on the previously reported electrochemical balancing methods is considered a suitable case study to showcase the potential of the technique to provide unambiguous results to unravel key processes to fully understand mechanism. Regarding the methods, simply explained, electrochemical “rebalance” of an unbalanced battery consists of intentionally promoting an irreversible side reaction in the “unbalanced compartment” (where charged species are accumulated) to drive also an accumulation of “charged species” at the opposite side, thus “rebalancing” the state of charge of both sides. In the case of our 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB, where oxygen-induced parasitic reactions at the anolyte (O 2 +2H 2 O +4e − → 4OH − ) is responsible for the Faradaic imbalance, promotion of the oxygen evolution reaction (OER) by water splitting (4OH − → O 2 +2H 2 O +4e − ) at the catholyte will compensate the charges irreversibly consumed by the ORR, leading to a more balanced state of charge of both electrolytes. Since the redox potential of the OER decreases with increasing pH, the OER is facilitated by the battery alkaline media and hence can be promoted by increasing the upper cut-off voltage in the charging step to a higher value (+2.0 V) than that previously applied in the standard cycling protocol reported in Section 3.2 (+1.6 V). To simplify the explanation, discussion is focused on the first cycle that usually delivers a lower Coulombic efficiency due to the presence of small amount of oxygen (even after purging with Ar) and oxygen absorbed on the graphite felts. The excess of ferrocyanide was removed (for both Fig. 6A and B), so that there is not buffer of ferrocyanide to compensate the ORR (and avoid conducting multiple cycles to consume this buffer). For a cut-off voltage of +1.6 V, Fig. 6A shows that the ferrocyanide signal is not restored to its initial value since some ferrocyanide was irreversible oxidized to compensate the charge consumed by oxygen-related reactions in the anolyte. Raman signal form Fig. 6A confirms that an accumulation of ferricyanide took place, implying that the charge step was limited by the catholyte (there are no enough charged species in the anolyte to fully discharge the catholyte). When the cut-off voltage is lifted to +2.0 V, a second plateau appeared at higher voltage (Fig. 6B) for a balanced battery (113 mAh for positive and negative sides). This could be attributed to the occurrence of oxygen evolution reaction in the positive electrode, which compensates the oxygen reaction in the anolyte and enables reaching full state of charge for the anolyte. The Raman Fig. 5. (A) Time evolution of cell voltage and Raman intensity at 2090 cm −1 measured in situ in the catholyte during galvanostatic cycling of a 2,6-DHAQ//K 4 Fe (CN) 6 alkaline RFB at ±20 mA cm −2 with voltage limits of +1.6/+0.5 V. Circles denote potassium ferricyanide accumulation at the end of each cycle). Catholyte: 14 mL of 0.3 M ferrocyanide in 1 M KOH; anolyte: 10 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 7 signal is essential to assess this hypothesis. Fig. 6B shows that the initial concentration of ferrocyanide is recovered at the end of the first discharge process, in contrast to the measurement carried out using a cut-off voltage of +1.6 V (Fig. 6A). According to the Coulombic efficiency in the first cycle (92 %), the OER should account for 8 % of the reversible capacity in the first cycle to enable recovery of initial concentration of ferrocyanide. In addition, the synchronization of the electrochemical and spectroscopy signals enables us to conclude that the Raman intensity of ferrocyanide decreased constantly during the first voltage plateau, but it stopped decreasing during the second plateau (Fig. S12). Thus, the oxidation of ferrocyanide is not the main reaction occurring during the second plateau, but the OER is. Thus, Fig. 6 clearly illustrate the effect of irreversible side reactions in the anolyte on the accumulation of charged species in the catholyte., and how the OER can provide the necessary charges to complete the charge step in the anolyte. 3.4. Probing balanced parasitic reactions in the positive and negative electrodes Interestingly, the relatively low Coulombic efficiency in this particular measurement (88 % for 1st cycle and 97 % for subsequent cycles) made us review the Ar-lines and a small leakage was found. After fixing the leaking, the Coulombic efficiencies (>99.5 %) were comparable to those obtained in an Ar-filled glovebox. The formation of 2,6-DHAQ dimers (evidenced by the appearance of a lower voltage plateau as discussed below, Fig. S13) confirmed that the rate of oxygen leaking was very small for the new measurements (bearing in mind that complete prevention is not possible without an Ar-filled glovebox). Fig. 7A shows the evolution of the capacity and Coulombic efficiency with cycles. The capacity fading rate remained constant for the first 25 cycles. According to the Coulombic efficiency, the small excess of catholyte should have been consumed by cycle 7. However, a change in the capacity fading rate was not observed, which indicates that the parasitic side reactions that are responsible for the Coulombic inefficiency do not result in a shortterm Faradaic imbalance. Indeed, the Raman signal confirmed that the catholyte is not limiting the capacity since the intensity of ferrocyanide signal never approaches values close to 0, as the bands of ferrocyanide are observed at the end of the charge step even for the last cycle (Fig. S14). That is, ferrocyanide was never depleted at the end of the charge step, as it would have been expected when the catholyte was the limiting side as it occurred in the previous experiment (Fig. 5). Therefore, Raman spectroscopy confirms that the chemical degradation of anolyte is responsible for the capacity fading, which is supported by the appearance of a lower voltage plateau in the discharge process (Fig. S13) that is attributed to the formation of 2,6-DHAQ dimers [38]. Thus, a measurement was designed to shed light into the occurrence of irreversible side reactions without the apparent impact on the capacity retention. Specifically, the 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB was operated for 10 cycles. Then, the battery was fully charged and left at open circuit voltage for 12 h. Continuous flow of electrolyte was maintained during the open circuit. After that, the battery was discharged and continued cycling. The capacity-fading rate was constant before and after the application of the open circuit voltage: 4.2 %⋅day −1 . During the open circuit voltage, the capacity fading rate increased (5.4 Fig. 6. In situ Raman spectroelectrochemical measurements for 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFBs. Time evolution of cell voltage and Raman intensity at 2090 cm −1 measured in situ at the catholyte during the first charge/discharge cycle at ±20 mA cm −2 (A) battery with voltage limits of +1.6/+0.5 V (standard conditions) using an oversized tank for the catholyte (catholyte: 14 mL of 0.3 M ferrocyanide in 1 M KOH; anolyte: 10 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH), and (B) battery with voltage limits of +2.0/+0.5 V (electrochemical rebalancing conditions) using a balanced amount of charge for catholyte and anolyte (catholyte: 14 mL of 0.3 M ferrocyanide in 1 M KOH; anolyte: 12 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH). Fig. 7. In situ Raman spectroelectrochemical measurement for a 2,6-DHAQ//K 4 Fe(CN) 6 alkaline redox flow battery during galvanostatic cycling at ±20 mA cm −2 with voltage limits of +1.6/+0.5 V. (A) Evolution of the discharge capacity and Coulombic efficiency upon cycling. (B) Time evolution of cell voltage and Raman intensity at 2090 cm −1 measured in situ in the catholyte during the first 25 charge/discharge cycles. Catholyte: 14 mL of 0.3 M ferrocyanide in 1 M KOH; anolyte: 10 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 8 %⋅day −1 ) because the 2,6-DHAQ is totally reduced, which favors the disproportion leading to dimerization [38]. The appearance of a second voltage plateau at a lower voltage (ca. 0.7–0.6 V) indicates that dimers were still present after leaving the cell at open voltage for 12 h (Fig. S15). Indeed, the length of the lower voltage plateau (dimers) increased after the open circuit as shown in Fig. S15, which is attributed to the accelerated dimerization occurring at fully charged state. The discharge capacity after the application of the open circuit voltage decreased by 14.5 %, and the most of it (11.8 %) was recovered in the subsequent cycle (Fig. 8A). While electrochemical data alone is not enough to elaborate further on fundamentals of the key processes, the addition of the Raman signal provided in situ information that allowed us to propose a reasonable explanation. One important point for this discussion is that Raman spectroscopy confirmed that ferrocyanide was not limiting the capacity after the open circuit voltage (neither before), as shown in Fig. S16 and discussed later. After ruling out Faradaic imbalance, the main sources for irreversible capacity fading can be attributed to crossover and degradation of 2,6-DHAQ. Raman spectroscopy enabled an estimation for the crossover of 2,6-DHAQ. The 2, 6-DHAQ has a distinct signal at 1773 cm −1 . Indeed, the intensity at 1773 cm −1 increased lineally with 2,6-DHAQ concentration for a calibration carried out in the presence of ferricyanide (Fig. S17). During the open circuit, the signal from 2,6-DHAQ measured in the catholyte slightly increased (Fig. S18). However, the amount was determined to be below 0.5 mM (<0.25 % of initial concentration). Thus, the crossover accounted only for a capacity fading of 0.5 %⋅day −1 (0.25 % in 12 h) out of the 5.4 %⋅day −1 (irreversible capacity fading) observed during the open circuit. Interestingly, in situ monitoring of crossover during the open circuit voltage step suggests that crossover of 2,6-DHAQ increased at higher state of charge. This is rather surprising since 2,6-DHAQ becomes more negatively charged at full state of charge, which should disfavor crossover when using a cation-exchange membrane. Nevertheless, recently, in-situ NMR measurements also reported this surprising increase in crossover at higher state of charge [21], This work concluded that the crossover is mainly driven by the migration of K + from the negative compartment to the positive compartment. This hypothesis was proposed based on the observation that the crossover rate increased with increasing current density. However, our findings were obtained at open circuit potential where migration of K + through the membrane does not occur. Consequently, our finding indicates that migration of K + through the membrane is not the only factor. Since our results were obtained at equilibrium, the origin is not likely to be electrochemically driven. Considering the large differences in electrolyte viscosity for fully charged and fully discharged state, this property may be related to the enhanced crossover. In particular, the anolyte increases significantly its viscosity while the catholyte undergoes minor changes. Since the system was operated at constant flow rate, pressure in the negative half-cell is expected to increase with increasing state of charge, while pressure in the positive half-cell almost does not change. The pressure differences in the two half-cells should reach a maximum at full state of charge, which is speculated to promote the crossover. Nevertheless, a complete study should be carried out to obtain unambiguous results to clarify this unexpected observation in the crossover rate. In any case, our in-situ Raman measurements are highly relevant to support such an unexpected behavior. Returning to the capacity fading, compared with the value reported in literature (4.6 %⋅day −1 , [39]), the slightly higher capacity fading of 5.4 %⋅day −1 observed during the open circuit at full state of charge is now attributed not only to the accelerated degradation of 2,6-DHAQ but also to the slightly higher crossover of charged 2,6-DHAQ. The most intriguing aspect of this measurement occurred during the open circuit voltage. The discharge capacity decreases with respect to the previous charge capacity. We refer it to as total capacity loss (14.5 %). The capacity of the subsequent cycle increased almost to the value sof the previous cycles. However, a loss of 2.7 % with respect to the capacity of the cycle before the open circuit was observed, which is referred to as irreversible capacity loss. Thus, ta large loss of capacity was recovered, which is referred to as the reversible capacity loss and accounted for 11.8 % (14.5 % total – 2.7 % irreversible) during the open circuit voltage (Fig. 8A). The reversible capacity loss is attributed to the self-discharge processes, namely oxygen evolution in the positive electrolyte which converts ferricyanide to ferrocyanide, and hydrogen evolution or oxygen reduction in the negative electrolyte to oxidize reduced 2,6-DHAQ. If this capacity fading was due to oxygen-induced parasitic reactions, a Faradaic imbalance would occur leading to partial irreversible losses. In addition, it should be considered that 2,6DHAQ can operate outside the thermodynamic stability window of the electrolyte (0.142 V vs RHE) [13] at high state of charge (Nernstian shift of the potentials with the state of charge), so that hydrogen evolution reaction may be expected as well, which would have the same impact as oxygen-induced parasitic reactions in the Faradaic imbalance. Although the kinetics of the hydrogen evolution reaction (HER) are expected to be sluggish, a comprehensive assessment of its long-term contribution to Faradaic imbalance necessitates further analysis, which is beyond the Fig. 8. In situ Raman spectroscopy probing balanced parasitic reactions in the positive and negative electrodes of a 2,6-DHAQ//K 4 Fe(CN) 6 alkaline RFB. (A) Discharge capacity and Coulombic efficiency versus number of cycles. (B) Voltage profiles and Raman intensity measured in situ during galvanostatic cycling of a 2,6DHAQ//K 4 Fe(CN) 6 alkaline redox flow battery at the catholyte side at wavenumbers characteristic of ferrocyanide: 2090 cm −1 . Current density: ±20 mA cm −2 , voltage limits of +2.0/+0.5 V. Catholyte: 14 mL of 0.3 M ferrocyanide in 1 M KOH; anolyte: 10 mL of 0.2 M 2,6-DHAQ in 1.4 M KOH. L. Lubian et al. Journal of Power Sources 646 (2025) 237272 9