A novel high-performance all-liquid formic acid redox fuel cell: simultaneously generating electricity and restoring the capacity of flow batteries

Abstract

Conventional formic acid fuel cells rely on the oxygen reduction reaction (ORR) to generate the cathode potential. However, this approach is plagued by mixed-potential issues caused by formic acid crossover and poor cathodic electrochemical kinetics. To address these limitations, we propose an innovative design that replaces the oxygen with a liquid redox couple. By implementing the Bi-modified Pt/C electrocatalyst that can facilitate the formic acid oxidation reaction with robust CO tolerance, this novel redox fuel cell achieves an open circuit voltage and a peak power density of 1.23 V and 281.5 mW cm⁻², respectively, representing a 55.7% and 235.1% improvement over the cell with the ORR cathode. Such performance metrics are among the highest recorded for formic acid fuel cells. Density functional theory calculations and a mathematical model are utilized to describe the increased activity of the produced catalysts and the cell working principles. Moreover, the redox fuel cell can be used to restore the capacity of flow batteries by using the degraded electrolyte as a cathode fuel. For example, the capacity of vanadium redox flow batteries can be recovered to 97.6% of the initial highest level after 400 cycle tests. Exhibiting high safety and convenience, this innovative cell offers a feasible, economically viable avenue for significantly extending the flow batteries’ cycle life.

---

Broader context

Improving the power density and expanding applications are key to the broad commercialization of formic acid fuel cells. We introduce an innovative all-liquid formic acid redox fuel cell (LFARFC), using the degraded electrolyte of a redox flow battery (RFB) as a cathode fuel alternative to traditional oxygen. It addresses water flooding and mixed potential issues in the cathode of traditional formic acid fuel cells, achieving an open-circuit voltage of 1.23 V. By implementing the Bi-modified Pt/C electrocatalyst that can facilitate the formic acid oxidation reaction with robust CO tolerance, the LFARFC achieves a peak power density of 281.5 mW cm⁻². In addition, this novel redox fuel cell not only facilitates efficient power generation but also regenerates the electrolyte of the RFB, marking a significant stride in extending the cycle life of flow batteries.

Introduction

Liquid fuel cells offer highly advantageous solutions for sustainable energy sources due to their considerable energy density, storage convenience, and safety features.^1,2 Among diverse variants, the formic acid fuel cell (FAFC) is distinguished by its benefits.^3,4 Formic acid, for example, boasts a higher ignition point and reduced toxicity compared to fuels like methanol. Moreover, as a carbon-neutral fuel, liquid formic acid can be readily produced by carbon dioxide reduction through artificial photosynthesis, offering more accessible transportation and storage than hydrogen gas, which poses a more significant explosion hazard. It can store substantially more protons per volume than compressed hydrogen and achieve an outstanding energy density of 2086 W h L⁻¹.^5,6 These alluring qualities have generated a great deal of interest in FAFC development both inside and outside of academic research.

Despite these merits, certain challenges still require being addressed before this technology can be widely commercialized.⁷ Some major technical issues are related to the oxygen reduction reaction (ORR) cathode and the formic acid oxidation reaction (FAOR) anode.^8,9 For the cathode, the real electrode potential is substantially lower than the theoretical value because the crossover of formic acid from the anode to the cathode chamber oxidizes and decreases the potential.¹⁰ Besides, when liquid water accumulates in the cathode, a phenomenon known as “water flooding” significantly raises the mass transfer resistance of oxygen and polarization loss.^11,12 In addition, it is challenging to synthesize active and durable catalysts for the FAOR. The indirect oxidation of formic acid on Pt-based catalysts with the dehydration pathway poisons the Pt by strongly adsorbing CO* (served as intermediates in formic acid oxidation processes),¹³ thus decreasing the durability and activity of the catalysts. Therefore, constructing durable anode catalysts and a suitable cathode is essential to enhance the performance and stability of FAFCs.

Conventional FAFC supplies the anode fuel (anolyte) with formic acid solutions. The FAOR occurs to generate electrons, protons, and CO₂.^14,15 Meanwhile, oxygen/air is supplied to the cathode, where it reacts with protons to form water (Fig. S1A, ESI†).^16,17 The anode half-cell reaction is:

HCOOH − 2e⁻ → 2H⁺ + CO₂ E⁰_FAOR = −0.25 V (1)

The cathode half-cell reaction is:

1/2O₂ + 2H⁺ + 2e⁻ → H₂O E⁰_OOR = −1.23 V (2)

The overall reaction can be summarized as follows:

1/2O₂ + HCOOH → H₂O + CO₂ E⁰ = 1.48 V (3)

In the cathode, Pt-based nanomaterials are typically used as catalysts for the ORR, resulting in an undesired potential loss, as formic acid that crosses over the membrane to the cathodic Pt surface will directly oxidize and form a mixed potential. To address this issue, we proposed a novel all-liquid formic acid redox fuel cell (LFARFC). Unlike conventional wisdom, the redox fuel cell supplies the liquid redox species as the cathode fuel (catholyte). The redox cathode's half-cell reactions can be summarized as follows:

Aⁿ⁺ + ze⁻ → A^(n−z)+ (4)

The high-valence redox ions (Aⁿ⁺) are reduced (electrochemically discharged) by accepting electrons in the LFARFC (Fig. S1B, ESI†), thereby forming low-valence redox ions (A^(n−z)+).^18,19 Therefore, the overall cell reaction can be expressed as:

z/2HCOOH + Aⁿ⁺ → A^(n−z)+ + z/2CO₂ + zH⁺ (5)

In particular, the LFARFC does not use a platinum-group-metal catalyst on the cathode side since redox couples do not require any noble metal for acceptable reaction kinetics. The absence of noble metals at the cathode extenuates the mixed potential issues by fuel crossover since the formic acid can only be oxidized on the platinum-group metal surface. On top of this inventive design, a facile chemical reduction method is proposed to introduce the Bi element into Pt/C nanoparticles. The obtained catalyst can optimize the FAOR reaction pathway by inhibiting CO* generation and improving CO tolerance, as evidenced by density functional theory (DFT) calculations and in situ Fourier transform infrared (FTIR) experiments. Specifically, the Bi-modified Pt/C achieves a specific activity of 80.67 mA cm⁻², which is 3.1 times higher than that of Pt/C. The LFARFC with the prepared catalyst exhibits an open circuit voltage and a peak power density of 1.23 V and 281.5 mW cm⁻², respectively, 55.7% and 235.1% higher than those of the conventional FAFC with ORR cathodes. It is among the highest performance of formic acid fuel cells ever reported. Aside from more efficiently converting the chemical energy of formic acid into electrical energy, another critical feature of the redox fuel cell is that it can restore the capacity of redox flow batteries (RFBs).

RFBs are upcoming technologies for large-scale, long-duration energy storage and have recently attracted much attention.^20–25 However, after cycling operation, they are susceptible to unavoidable capacity decay issues caused by unbalanced concentration, volume, and valence of the positive and negative electrolytes.^26–30 While the concentration and volume can be rebalanced by remixing the positive and negative electrolytes,^31–33 the irreversible valence increase induced by side reactions and air oxidation requires external intervention.³⁴ The issue of increased valence can lead to irreversible capacity decay in flow batteries, becoming an urgent problem that needs to be resolved. Currently, the commonly used way to regulate the electrolyte valence is to use an electrolytic cell,^35–37 or add an agent to the electrolyte for a chemical reduction process.^38–40 Nevertheless, both methods have certain drawbacks, such as the need for additional electrical or thermal energy to drive the electrolytic and chemical reaction processes. LFARFCs can overcome these shortcomings perfectly and effectively. For example, by reforming the degraded electrolyte of vanadium redox flow batteries (VRFBs) with LFARFCs, the capacity of VRFBs can be recovered to 97.6% of the initial highest level after 400 cycles, significantly extending the cycle life. In the restoration process, the cell operates at room temperature without any thermal energy input. It offers a high degree of safety and convenience, making it very promising for the operation and maintenance of large-scale RFBs.

Results

Working principle of all-liquid formic acid redox fuel cells

The ORR exhibits a standard electrode potential of 1.23 V,⁹ significantly higher than that of the V⁵⁺/V⁴⁺ redox reaction (1.00 V).⁴¹ In combination with the FAOR, thermodynamically theoretical open circuit voltages (OCVs, taken at zero current) of FAFCs and LFARFCs are expected to be 1.48 and 1.25 V (Fig. 1A and Fig. S2, ESI†), respectively. However, the measured OCV (0.76 V) of FAFC is only about half of the theoretical value (Fig. 1A and Table S1, ESI†). Since no current is applied at this point, the only significant factor that can reduce the cell potential is the direct reaction of the fuel and O₂ at the cathode, resulting from unavoidable formic acid crossover.⁴² When crossed formic acid reaches the cathode Pt/C surface, the FAOR (electron donor) and the ORR (electron acceptor) can spontaneously occur, resulting in an undesired potential loss (Fig. 1B).⁴³ Besides, the liquid water in formic acid solution on the anode can permeate the membrane and arrive at the cathode, exaggerating the water-flooding phenomenon.¹⁰ Both factors result in a catastrophic decrease in the cell performance and inhibit robust cell operation. When replacing the ORR with the V⁵⁺/V⁴⁺ redox reaction, the measured OCV of the LFARFC can reach 1.23 V (Fig. 1A), close to the theoretical value (1.25 V), indicating that the mixed-potential issue can be effectively suppressed. This enhancement can be attributed to two factors. Firstly, the V⁵⁺/V⁴⁺ redox cathode uses a carbon material (graphite felt) as the electrode and does not contain any noble metal catalyst, which is not catalytically active for the FAOR (Fig. 1C). Secondly, formic acid fuel diffusing across the membrane would be dissolved and diluted to the catholyte before reaching the electrode surface. The simultaneous absence of catalytically active sites and reactants significantly inhibits the undesired parasitic FAOR, mitigating the OCV drop caused by the mixed potential at the cathode side. Moreover, the catholyte is an aqueous solution, and water transport across the membrane does not affect liquid redox reactions, fundamentally eliminating the water-flooding issue in conventional FAFCs. When normalized to the electrode surface area, the kinetic current density of the ORR on commercial Pt/C was calculated to be 4.6 mA cm⁻² at 0.8 V vs. reversible hydrogen electrode (RHE) (Fig. 1D and Supplementary Note 1, ESI†). In contrast, the kinetic current density of V⁵⁺/V⁴⁺ on graphite felt reaches as high as 45.1 mA cm⁻² at 0.8 V vs. RHE, indicating that the cathodic electrochemical kinetics of V⁵⁺ to V⁴⁺ is much superior to those of the ORR, even without any metal electrocatalysts. Due to significantly higher OCV and better redox kinetics, the peak power density of the LFARFC reaches 65.8 mW cm⁻² (Fig. 1E), which is about 1.5 times higher than that of the FAFC of 26.5 mW cm⁻², demonstrating that pairing a suitable redox cathode can substantially improve the performance of the formic acid fuel cell.

Fig. 1 The comparison between an FAFC and LFARFC. (A) The theoretical and measured OCVs of the FAFC and LFARFC. (B) and (C) Schematic diagram and mechanism of the FAFC and LFARFC. (D) Linear sweep voltammetry (LSV) curves of the ORR on Pt/C and V⁵⁺ reduction reaction on graphite felt. (E) Polarization and power density curves of the FAFC and LFARFC.

Although promising, the power density still needs to be significantly enhanced, compared to state-of-the-art FAFCs.^44–46 A higher power density is beneficial for decreasing the cell size for a given power requirement, thereby reducing the capital cost of cell component materials such as bipolar plates, membranes, electrodes, etc.

Catalyst design for the FAOR

For the anodic reaction, Pt is renowned for its effectively catalyzing FAOR.⁴⁵ However, the active sites of Pt are prone to CO poisoning due to indirect oxidation processes. In one step of the FAOR, CO is produced as the intermediate CO*, which absorbs strongly on the Pt surface, reducing the active sites for reactions. To tackle this issue, a surface modification method is developed by synthesizing the Bi-modified Pt/C catalyst (details in the Experimental section and Fig. 2A). This process involves diffusing Bi atoms into the Pt interstices to form irregular layers through a chemical reduction using NaBH₄ and Bi₂O₃. Although previous studies have demonstrated that introducing Bi elements to create intermetallic PtBi or PtBi₂ could significantly enhance the activity and CO poisoning tolerance for the FAOR, a relatively high temperature is usually required for the intermetallic transformation of PtBi₂, such as 271 °C for cubic β-PtBi₂, 420 °C for trigonal γ-PtBi₂, and even higher for σ-PtBi₂.^47,48 In our method, the synthesis routine is simple and kept under mild conditions, which would be easier for practical applications.

Fig. 2 The synthesis and analysis of Bi-modified Pt/C. (A) Schematic diagram of the preparation process. (B) Aberration-corrected TEM image of Bi-modified Pt/C (top) and intensity analysis along the lines L₁ and L₂ (bottom). (C) HAADF-STEM image and corresponding images with elemental mapping. (D) TGA analysis of Pt/C and Bi-modified Pt/C. XPS spectra of (E) Pt 4f and (F) Bi 4f of different catalysts.

Transmission electron microscopy (TEM) images (Fig. S3, ESI†) of Pt/C and Bi-modified Pt/C catalysts reveal that introducing Bi atoms does not significantly alter the overall morphology and distribution of the Pt/C nanoparticles. However, a difference in the average particle size is evident, with sizes measured at 2.92 nm for Pt/C and 3.63 nm for Bi-modified Pt/C (Fig. S4, ESI†). This increase in size, based on the analysis of 100 nanoparticles from each sample, is ascribed to successful Bi adsorption on the Pt surface, leading to an expansion of the particle dimensions. Aberration-corrected TEM (Fig. 2B) is employed to investigate the structural change further. Through the intensity analysis of the atomic column in the aberration-corrected TEM image of the unevenness (L₁ and L₂), it can be found that these peaks are distributed at fixed distances. The lattice spacings of 0.33 nm (L₁) and 0.23 nm (L₂) correspond to the (012) crystal plane of Bi and the (111) crystal plane of Pt,^49,50 respectively, further confirming the presence of Bi on the Pt/C surface. Bi and Pt's locations are examined using a high-angle annular dark-field scanning transmission electron microscope (HAADF-STEM) and the corresponding energy-dispersive X-ray spectroscopy (EDS) mappings (Fig. 2C) in the modified catalyst. The results indicate that Bi atoms are disorderly distributed on the Pt surface, forming irregular layers. Thermogravimetric analysis (TGA) is performed to quantify the Bi adsorption in weight (Fig. 2D). The analysis shows residual values of 47.01% for commercial Pt/C and 50.43% for the Bi-modified catalyst after carbon support oxidation at around 400 °C indicates that the Bi-modified Pt/C catalyst contains 47.01% Pt, with 3.42% Bi adsorbed on the Pt. Due to the low Bi/Pt ratio (Bi/Pt ratio < 0.07), Bi is not observed in the X-ray diffraction (XRD) tests of the modified catalyst (Fig. S5, ESI†).⁵¹ X-ray photoelectron spectroscopy (XPS) analysis is utilized to gain more detailed surface information on the synthesized catalysts, with the findings in Fig. 2E and F. A doublet of Pt 4f_7/2 and 4f_5/2 is analyzed in the Pt spectrum (Pt 4f). These peaks can be deconvoluted into three distinct components: metallic Pt (green), Pt²⁺ in PtO (blue), and Pt⁴⁺ in PtO₂ (yellow). For Bi-modified Pt/C, the metallic Pt content is 54.4% of Pt, PtO, and PtO₂, similar to that of Pt/C (54.1%). The content of Pt metal is almost unaffected after modification within the measurement error.

A certain amount of PtO and PtO₂ is detected in both samples due to partial oxidation in the ambient air during tests, as seen in Fig. S6 (ESI†). The XPS result of Bi-modified Pt/C significantly reveals new peaks at 157.0 and 162.3 eV, corresponding to the metallic Bi, which are absent in Pt/C. This finding unequivocally confirms the successful incorporation and presence of Bi on the modified Pt nanoparticles. The electrochemical half-cell tests are further conducted to compare the reaction activity of commercial Pt/C and Bi-modified Pt/C. Firstly, the electrochemically active surface areas (ECSAs) of the two catalysts were determined through cyclic voltammetry (CV) tests in a three-electrode setup in N₂-saturated 0.5 M H₂SO₄. As shown in Fig. 3A, the ECSAs calculated using the hydrogen adsorption/desorption isotherms (0.05 to ∼0.45 V vs. RHE) are 47.66 and 11.61 m² g⁻¹ for Pt/C and Bi-modified Pt/C, respectively (Supplementary Note 2, ESI†).⁵² The observed decrease in the ECSA following Bi modification is due to the shielding effect of Bi atoms on Pt surfaces. Theoretically, each Bi atom can prevent three Pt atoms from coming into contact with hydrogen.⁷ Therefore, the Bi coverage for modified Pt/C was calculated to be about 0.25 (Fig. S7 and Supplementary Note 3, ESI†), which is within the optimum range (0.18–0.33) for the FAOR.⁴⁴ To evaluate the FAOR activity, further tests are performed in an N₂-saturated 0.5 M H₂SO₄ solution containing 0.5 M HCOOH (Fig. 3B–E). The forward scan of the LSV test displays that the Bi-modified Pt/C shows a much lower potential for the FAOR at 10 mA cm⁻² (0.527 V) than that of Pt/C (0.622 V) (Fig. 3B).

Fig. 3 FAOR performance evaluation of Bi-modified Pt/C and commercial Pt/C. (A) CV curves of catalysts in N₂-saturated 0.5 M H₂SO₄ solution. (B) LSV curves, (C) mass and specific activities, (D) Tafel plots, and (E) chronoamperometric curves of catalysts in N₂-saturated 0.5 M H₂SO₄ solution containing 0.5 M HCOOH. (F) Polarization and power density curves of the LFARFC with different catalysts. (G) and (H) In situ FTIR spectra of Pt/C and Bi-modified Pt/C.

According to the LSV curves, the calculated specific and mass activities of Bi-modified Pt/C (371.75 mA mg_Pt⁻¹ and 80.67 mA cm⁻²) are significantly higher than those of Pt/C (109.53 mA mg_Pt⁻¹ and 25.74 mA cm⁻²), as illustrated in Fig. 3C. In addition, the Tafel slope of the Bi-modified Pt/C is 318.46 mV dec⁻¹ (Fig. 3D), which is significantly lower than that of Pt/C (752.36 mV dec⁻¹), indicating the enhanced FAOR kinetics of Bi-modified Pt/C. Apart from the catalytic activity, the stabilities of the catalysts are evaluated by chronoamperometry (Fig. 3E). It is noteworthy that Bi-modified Pt/C exhibits enhanced stability for the FAOR, with the highest current density of 30.02 mA cm⁻² and retention rate of 18.0% after 3600 s in contrast to 28.82 mA cm⁻² and 7.67% for Pt/C and later shows a quick decrease in current within 60 s. The enhanced electrochemical activities of Bi-modified Pt/C can elevate the cell performance significantly. As shown in Fig. 3F, the peak power density of the LFARFC with Bi-modified Pt/C reaches 281.5 mW cm⁻², which is 327.8% higher than that of the cell with Pt/C (65.8 mW cm⁻²), representing the highest value for FAFCs among related reports (Table S1, ESI†). Notably, the OCV (1.23 V) and peak power density (281.5 mW cm⁻²) of the Bi-modified Pt/C-based LFARFC are 55.7% and 235.1% higher than those of the FAFC (0.79 V and 84.0 mW cm⁻²) with the same catalysts (Fig. S8, ESI†). These advances further demonstrate the superiority of the proposed catalyst design for the FAOR in the LFARFC. To gain a deeper insight into the mechanism by which Bi-modified Pt/C generates high activity, in situ FTIR is introduced to clarify the reaction pathway of the FAOR (Fig. 3G and H and Supplementary Note 4, ESI†). It is widely believed that there are two reaction pathways for the FAOR, including dehydrogenation (HCOOH → HCOO*/COOH* → CO₂) and dehydration (HCOOH → COOH* → CO*).⁵² Dehydrogenation is generally considered to be the optimal pathway because formic acid can be directly oxidized. In the dehydration steps, the linear absorbed CO (CO_L) tends to induce CO poisoning, which can be identified by detecting CO_L during testing. In the FTIR spectrum of Pt/C (Fig. 3G), two characteristic peaks were assigned to the asymmetric stretching vibration of CO₂ (∼2341 cm⁻¹) and CO_L (∼2010 cm⁻¹), suggesting that both pathways are involved. In contrast, the apparent peaks of CO₂ can also be obtained for Bi-modified Pt/C (Fig. 3H), while no CO_L peaks are detected. These results indicate that dehydrogenation steps are preferred for Bi-modified Pt/C, which brings about high resistance to CO poisoning.

To illustrate the mechanism of improved FAOR kinetics for Bi-modified Pt/C, DFT calculations are conducted to estimate the free energy of the involved species (Supplementary Note 5, ESI†). The difference in free energy between the reactant and product is termed the formation energy. Thermodynamically, a higher formation energy indicates a more incredible difficulty in forming the corresponding product. The formation energy of CO* on Pt (0.27 eV, Fig. 4A) is lower than that on Bi-modified Pt (1.45 eV, Fig. 4B), which means that the dehydration steps and CO poisoning are more likely to occur on Pt catalysts. The introduction of Bi elevates the formation energy of CO*, which inhibits the dehydration steps and facilitates the overall FAOR kinetics. Structural analysis (Fig. 4C and D) reveals that Bi-modification leads to the lengthening of C–H bonds in COOH* (from 0.98 to 0.99 Å) and HCOO* (from 1.10 to 1.11 Å), which boosts the bond-breaking reaction of C–H, thereby promoting dehydrogenation and CO₂ generation. The structural analysis reveals that the Pt surface exhibits a propensity to adsorb CO* (Fig. S9, ESI†), which can easily induce CO poisoning. In contrast, Bi-modification obstructs the adsorption and formation of CO* on the catalyst surface (Fig. S10, ESI†) and then enhances CO tolerance. To better understand the impact of catalysts on the polarization losses in LFARFC, a three-dimensional multi-physics model (Supplementary Note 6 and Table S2, ESI†) was established to deconvolute the cell loss, which is well validated and exhibits grid independence (Fig. S11, ESI†). The cathode, ohmic, and anode losses can be separately determined from the simulation, as shown in Fig. 4E and F. For the cell with Pt/C catalysts, anode polarization constitutes the primary loss (Fig. 4E), indicating that Pt/C exhibits inadequate catalytic activity, which fails to reach the required reaction kinetics of the FAOR. With the introduction of Bi-modified Pt/C, anode polarization loss is reduced to a level equivalent to the cathode side (Fig. 4F). Even though ohmic and cathodic polarizations remain the same, the overall losses have been substantially mitigated. The simulation results indicate that the Bi-modified catalysts enhance the performance of LFARFCs by notably reducing the anode polarization loss.

Fig. 4 Mechanistic analysis of the catalyst's performance improvement. (A) and (B) Free energy diagrams for the dehydrogenation and dehydration steps on the surfaces of Pt and Bi-modified Pt. (C) and (D) Optimized surface structures with HCOO* and COOH* on the surfaces of Pt and Bi-modified Pt. (E) and (F) The polarization loss analysis for LFARFCs with Pt/C and Bi-modified Pt/C.

Capacity recovery of RFBs by LFARFCs

Capacity degradation is an issue that has long plagued the operation of RFBs. In the case of VRFBs, for example, (Supplementary Note 7, ESI†), mixing the initial negative electrolyte (V³⁺) and positive electrolyte (V⁴⁺) will lead to an electrolyte with an average valence of 3.5. However, unexpected side reactions (e.g., hydrogen evolution, air oxidation, etc.) increase the valence of the vanadium electrolyte (usually above 3.5). The relation between the available capacity and the mixed electrolyte containing V^X+ is depicted in Fig. 5A. It is found that the more significant the valence increase, the greater the capacity degradation. In particular, when the valence in the mixed electrolyte reaches 4, the available capacity diminishes to 0% of its theoretical value. In detail, when charging with V⁴⁺ as the initial electrolyte, the negative and positive electrolytes are charged to V³⁺ and V⁵⁺, respectively. This leads to an imbalanced valence that cannot release capacity, as the negative electrolyte is in a fully discharged state. The electrolyte valence adjustment can be realized by LFARFCs employing formic acid and high-valence V⁵⁺ as the anolyte and catholyte. As shown in Fig. 5B, the discharge curve of LFARFCs at 100 mA cm⁻² is accompanied by a change in the valence of vanadium ions, reflected by the color change of the catholyte from yellow (V⁵⁺) to blue (V⁴⁺). The vanadium valence can be measured quantitatively through the ultraviolet-visible (UV-vis) test. By analyzing the V⁴⁺ adsorption peak at 765 nm, the vanadium valence of the discharged electrolyte was calculated to be 4.02 (Fig. 5C), which means 98% of V⁵⁺ is reduced to V⁴⁺. In addition, the simulation result shows that when discharged from 1 V to approximately 0 V, the catholyte valence drops from 5.00 to below 4.02 (Fig. 5D and Table S3, ESI†), which fits well with the experiments. At different stages during discharge, the theoretical faradaic capacity of the electrolyte matches well with the actual discharge capacity (Fig. 5E and Fig. S12, ESI†), meaning that the vanadium valence in the catholyte can be effectively controlled by adjusting the discharge capacity of the cell. Moreover, The LFARFC setup functions with decent stability, as the newly obtained discharge curve highly overlaps with the previous one after replacing the catholyte and anolyte (Fig. 5F). The CV and EIS (Fig. 5G and H and Supplementary Note 8, ESI†) curves reflect equal peak potentials, currents, and impedances, which means the properties of the discharged catholyte are similar to those of the pure V⁴⁺ electrolyte.

Fig. 5 The capacity decay and recovery of VRFBs. (A) Schematic of the capacity decay of VRFBs with the different valences of the mixed electrolytes. (B) Electrolyte regeneration in LFARFCs. (C) UV-vis curves of the variation of vanadium electrolytes. (D) Simulation result of the relationship between the discharge voltage and electrolyte valence of LFARFCs. (E) Comparison of the experimental discharge capacity and theoretical faradaic capacity. (F) Stability testing of LFARFCs. (G) CV and (H) EIS curves for the pure V⁴⁺ electrolyte and the discharged catholyte.

Then, we applied the setup to recover the capacity of a VRFB, which underwent 500 cycles at a constant current density of 200 mA cm⁻², as shown in Fig. 6A and Table S4 (ESI†). Although the efficiency remained stable, the battery undergoes inevitable capacity decay induced by the imbalance of positive and negative electrolytes. During the first 100 cycles, the discharge capacity dropped from 481.0 to 397.8 mA h. Without intervention, the battery capacity would continue to decrease (as shown by the dotted line in Fig. 6B). To restore the capacity, the battery operation stopped, and a recovery process was implemented at the end of the 100th cycle (Supplementary Note 9, ESI†). The process can be described as follows: the negative and positive electrolytes are first mixed and then divided into two equal parts, successfully balancing the volume and concentration. After mixing, the average electrolyte valence was measured to increase from V^3.50+ to V^3.61+ (Fig. 6C). Next, the battery was fully charged to get V⁵⁺ and V^2.22+ on the positive and negative sides, respectively. Then, the LFARFC involved the use of the obtained V⁵⁺ as the initial catholyte and discharged it to V^∼4.78+ (a symmetrical state of V^2.22+) by controlling the discharge capacity to 117.9 mA h. Finally, the V^∼4.78+ catholyte was returned to the positive tank of the flow battery, and the cycle tests continued. At this point, the entire capacity recovery step ends. It was encouraging to observe that the battery capacity attained 474.8 mA h after the recovery process (Fig. 6B), reaching 98.7% of the previous highest level (481.0 mA h). The average valence of the mixed electrolyte was successfully reduced from 3.61 to 3.48 through the UV-vis tests (Fig. 6C). Charge–discharge curves of points A, B, and C are depicted in Fig. 6D. The charge–discharge curve of C almost returns to its initial location and covers the first cycle. Four restorations were performed throughout the 500 cycles (each restoration was performed at 100 cycle intervals). After the fourth recovery, the battery capacity can still maintain at 469.6 mA h, equivalent to 97.6% of the initial capacity (481.0 mA h), demonstrating the crucial role of the LFARFC in regulating electrolyte valence and recovering VRFB capacity. The recovery process took place at room temperature without additional thermal or electrical energy input, outweighing the traditional recovery method (electrolysis^35–37 or chemical reduction^38–40) and making it highly safe and easy to operate. In practical applications, the RFBs and LFARFCs can be easily integrated through the transport of liquid electrolytes (vanadium electrolyte) and reactants (formic acid), allowing for a continuous and convenient mode switch between energy storage, capacity detection and recovery (Table S5, ESI†).

Fig. 6 The application of LFARFCs for recovering the capacity of a VRFB. (A) The capacity recovery test. CE is the coulombic efficiency; VE is the voltage efficiency; and EE is the energy efficiency. (B) The discharge capacity curves in the recovery process. (C) The valence changes of the VRFB electrolyte during operation and after recovery. (D) The charge–discharge curves of the VRFB at points A, B, and C.

Other redox cathodes such as Fe³⁺/Fe²⁺

Except for VRFBs, other types of RFBs, such as iron-based ones, also face the problem of capacity degradation due to elevated electrolyte valence.^22,53 There are several types of iron-based RFBs, such as Fe–S RFBs,⁵⁴ the Fe–Cr RFBs,⁵⁵ and all iron RFBs,⁵⁶ each designed to meet specific operational demands. Given the complexity of different Fe-based RFBs, the capacity recovery of the Fe^2+/3+ containing electrolyte is demonstrated without the evaluation of a specific iron-based RFB. To verify the universality of our proposed cell in restoring the capacity of iron-based RFBs, Fe³⁺ is used as the catholyte in this section. By equipping Bi-modified electrocatalysts, the peak power density of the cell can reach 76.45 mW cm⁻² (Fig. 7A), which is 2.75 times higher than that of the cell with a conventional Pt/C electrocatalyst, ascribed to the robust electrocatalytic activity of the Bi-modified catalyst. Fig. 7B shows the discharge performance of the cell operating at a current density of 40 mA cm⁻². The theoretical capacity of the ferric solution is 536 mA h, while the practical discharge capacity reaches 513 mA h, indicating that a majority of high-valence Fe³⁺ can be effectively reduced to Fe²⁺. The change in the solution color from brown (Fe³⁺) to yellow–green (Fe²⁺) further reflects the successful reduction of the ferric valence. The experimental results show that the LFARFC can also effectively adjust the valence of iron electrolytes and is expected to be applied to the capacity recovery of some iron-based RFBs. Practically, it is demonstrated that the cell can power an electronic device with a voltage booster, illustrating the effectiveness of the LFARFC in electricity generation (Fig. S13, ESI†).

Fig. 7 The application of the LFARFC in regulating the valence of iron-based electrolytes. (A) Steady-state polarization and power density curves of the LFARFC with an Fe³⁺/Fe²⁺ redox catholyte and different catalysts. (B) Variation of the catholyte during different discharge stages in the LFARFC.

Conclusion

In summary, we report a novel LFARFC capable of simultaneously generating electricity and recovering the capacity of RFBs. The cell engenders the cathode potential by introducing redox species in liquid electrolytes. Introducing a redox cathode eliminates mixed-potential and water-flooding issues in conventional ORR cathodes and significantly improves the electrochemical kinetics. To effectively demonstrate the cell performance, we designed a Bi-modified Pt/C electrocatalyst to facilitate formic acid oxidation with robust CO* tolerance. The novel cell with an electrolyte containing V⁵⁺ as the catholyte results in a high open circuit voltage of 1.23 V and a power density of 281.5 mW cm⁻², 55.7% and 235.1% higher than those of the cell with a conventional ORR cathode, which is among the highest performance of formic acid fuel cells. Lastly, we investigated the effectiveness of the cell in restoring the capacity of RFBs. Our findings reveal that the proposed LFARFC demonstrates notable advantages in RFBs’ capacity recovery. By adjusting active species’ valence in aqueous electrolytes (e.g., vanadium, iron, etc.), the cell can recover the capacity of the VRFB to 97.6% of the initial value after 400 cycle tests, thereby substantially extending the cycle life of RFBs.

Experimental

Preparation of Bi-modified Pt/C

Bi-modified Pt/C nanoparticles were synthesized by a chemical reduction method with NaBH₄ as the reducing agent. Firstly, 20 mL of 0.5 M sulfuric acid (95–98%, National Medicine, China) was mixed with 46.6 mg of Bi₂O₃ (99.99%, Sigma-Aldrich, USA) to form a solution in which the Bi³⁺ concentration was 5 mM. Subsequently, 60 mg of Pt/C (47 wt%) was mixed with the solution and stirred continuously for 48 hours to enable the irreversible binding of Bi³⁺ ions onto the surface of Pt/C nanoparticles. After that, distilled water washes were performed to eliminate residual Bi³⁺ ions from the solution. Then, 90 mg of NaBH₄ powder and treated Pt/C were mixed in 100 mL of dichloromethane solution and magnetically agitated for 1 hour to reduce the Bi³⁺ to Bi metal on the Pt surface. The resulting mixture was filtered using filter paper. The as-prepared catalyst was dried in an oven at 60 °C for 6 hours.

Characterization of the catalysts

Aberration-corrected TEM was performed on an FEI Titan Cubed Themis G2300. TEM was performed on an FEI Talos F200x. HAADF-STEM and EDS (Super X) were used to detect the distribution of Bi ions on the surface of Pt nanoparticles. In addition, XPS (Thermo Fisher ESCALAB Xi+) was used to analyze the electronic state of the catalysts with monochromate Al Kα radiation. The crystalline structure of the synthesized Bi-modified Pt/C was characterized using XRD (Rigaku SmartLab 3 kW diffractometer), employing a Cu Kα radiation source operating at 40 kV and 30 mA, with a 2θ range from 10° to 80° and a scan rate of 10° min⁻¹. TGA (Netzsch STA 449 F3) was performed to determine the quantity of Bi adsorbed.

Half-cell tests

The half-cell tests were performed on an electrochemical workstation (Shanghai CH Instruments Inc.). The working, counter, and reference electrodes were as-prepared Pt/C or Bi-modified Pt/C deposited glassy carbon, Pt mesh, and saturated calomel electrodes, respectively. The catalyst ink was prepared by dispersing the catalyst (5 mg) in a solution containing 24.75 μL of Nafion D520 ionomer (Sigma-Aldrich) and 500 μL of isopropyl alcohol (IPA, Sigma-Aldrich). After sonication for 15 min, 10 μL of ink was drop-cast onto the glassy carbon electrode (0.196 cm²) to prepare the rotating disk electrode (RDE). The catalyst loading on the electrode was 0.024 mg_Pt cm⁻², and the rotation speed of RDE was 1600 rpm in half-cell tests. CV curves were conducted at a scan rate of 50 mV s⁻¹ and the range of 0.05 to 1.2 V vs. RHE in 0.5 M H₂SO₄ solution. LSV was performed in Ar-saturated 0.5 M H₂SO₄ with 0.5 M HCOOH solution at a scan rate of 50 mV s⁻¹ and in the range of 0.05 to 1.2 V vs. RHE. The chronoamperometry curves were tested in 0.5 M H₂SO₄ and 0.5 M HCOOH for 3600 s at 0.46 V vs. RHE.

Fuel cell tests

A fuel cell was constructed with an interdigital flow field (detailed in Fig. S14, ESI†), and its components included bipolar plates, membranes, and electrodes. Polytetrafluoroethylene (PTFE) gaskets were placed between the two half fixtures. The graphite felt (Liaoning Jingu Carbon Materials Co., Ltd, China) was thermally treated and used as the cathode electrode and the substrate of the anode catalysts. The active area was 2 × 2 cm² with a thickness of 1.5 mm. The cathode and anode were separated using a membrane (Nafion® 212). The fuel solution (50 mL of 5 M formic acid) was fed into the anode side at 6.0 mL min⁻¹ in the closed loop mode. In the following text, V^X+ represents the solution containing vanadium ions with a valence of X, where X = 2, 3, 4, and 5 correspond to V²⁺, V³⁺, V⁴⁺, and V⁵⁺, respectively. If the value of X is within the range of 3–4, it represents a combination of V³⁺ and V⁴⁺ in different proportions. The catholyte (20 mL of 1 M V⁵⁺ + 3 M H₂SO₄ or 1 M Fe³⁺ + 3 M H₂SO₄) was fed at a rate of 40 mL min⁻¹ using the closed loop mode. The anode catalyst ink was made by combining 50 mg of catalyst (Pt/C, Bi-modified Pt/C) with 10 mL of deionized water, 120 μL of 5% Nafion solution, and 15 mL of IPA to generate the catalyst layer for the anode. The prepared ink was then sprayed onto a graphite-felt surface with a Pt loading of 1.5 mg cm⁻². An electrochemical workstation (Arbin BT2000, USA), controlled using MITS Pro software, was used to record the polarization and power density curves. The details of the measurement of the polarization curves can be seen in Supplementary Note 10 (ESI†).

Flow battery tests

To assess the capacity restoration capabilities of the LFARFC, a flow battery system featuring interdigitated flow fields was constructed for a long-term cycle test. Graphite felts with an active area of 4 cm² serve as both positive and negative electrodes, separated using a Nafion® 212 membrane (Dupont, USA). 20 mL of 1 M V^3.5+ (10 mL of 1 M V³⁺ and 1 M V⁴⁺) dissolved in 3 M H₂SO₄ were used for both the negative electrolyte and positive electrolyte. The liquid electrolytes were circulated using a dual-channel peristaltic pump (WT600-2J, Baoding Longer Pump Co., Ltd) at a steady flow rate of 30 mL min⁻¹. Before tests, N₂ gas was used to purge residual air from the electrolyte tanks for 10 minutes. Charge–discharge cycles were conducted using a NEWARE battery test system (CT-4008Tn-5V6A-S1, Shenzhen, China) at a fixed current density of 200 mA cm⁻² within a cut-off voltage window of 0.9–1.65 V. The details of the aforementioned electrolytes used in those tests are comprehensively described in Table S6 (ESI†).

Author contributions

D. B. W. conceived the design and performed the experiments. L. M. P. carried out the simulations and data processes. J. S. conducted the methodology and result analysis. D. B. W., L. M. P., and J. S. prepared the initial manuscript. M. S. H., S. B. W., Y. B. L., L. Z., L. W. and T. S. Z. revised the manuscript. M. R. S. and J. C. G. participated in the flow battery experiments. Q. Z., C. L. X., and Z. L. participated in the fuel cell experiments. L. Z., L. W. and T. S. Z. conceived the idea and provided supervision. All authors discussed the results and contributed to the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of interest

The authors declare no competing interests.

Acknowledgements

This work was supported by the National Key R&D Program of China (No. 2022YFB2404902), the National Natural Science Foundation of China (No. 52206089), the Guangdong Major Project of Basic and Applied Basic Research (2023B0303000002), the Guangdong Basic and Applied Basic Research Foundation (2024A1515010288 and 2023B1515120005), the National Natural Science Foundation of Shenzhen City (JCYJ20230807093315033), the Shenzhen Science and Technology Fund (No. ZDSYS20220401141000001), and the High Level of Special Funds (G03034K001). The computation in this work was supported by the Center for Computational Science and Engineering at the Southern University of Science and Technology.

References

1. M. K. Debe, Nature, 2012, 486, 43–51.
2. X. Ji, K. T. Lee, R. Holden, L. Zhang, J. Zhang, G. A. Botton, M. Couillard and L. F. Nazar, Nat. Chem., 2010, 2, 286–293.
3. X. Lu, Y. Wu, X. Yuan and H. Wang, Angew. Chem., Int. Ed., 2019, 58, 4031–4035.
4. X. Jiang, X. Yan, W. Ren, Y. Jia, J. Chen, D. Sun, L. Xu and Y. Tang, ACS Appl. Mater. Interfaces, 2016, 8, 31076–31082.
5. X. Fu, H. Li, A. Xu, F. Xia, L. Zhang, J. Zhang, D. Ma, J. Wu, Q. Yue, X. Yang and Y. Kang, Nano Lett., 2023, 23, 5467–5474.
6. S. Mondal and C. R. Raj, ACS Appl. Mater. Interfaces, 2019, 11, 14110–14119.
7. Y. Zhou, Y. Yang, X. Zhu, T. Zhang, D. d Ye, R. Chen and Q. Liao, Adv. Funct. Mater., 2022, 32, 2201872.
8. Y. H. Wang, J. B. Le, W. Q. Li, J. Wei, P. M. Radjenovic, H. Zhang, X. S. Zhou, J. Cheng, Z. Q. Tian and J. F. Li, Angew. Chem., Int. Ed., 2019, 58, 16062–16066.
9. H. Cruz-Martínez, H. Rojas-Chávez, P. T. Matadamas-Ortiz, J. C. Ortiz-Herrera, E. López-Chávez, O. Solorza-Feria and D. I. Medina, Mater. Today Phys., 2021, 19, 100406.
10. P. G. P. Xingwen Yu, J. Power Sources, 2008, 182, 124–132.
11. S. H. Takuya Tsujiguchi, T. Iwakami and N. Nakagawa, J. Power Sources, 2013, 223, 42–49.
12. T. T. Fahimah Abd Lah Halim, Yugo Osaka and Akio Kodama, Int. J. Energy Res., 2019, 43, 8070–8084.
13. T. Shen, S. Chen, C. Zhang, Y. Hu, E. Ma, Y. Yang, J. Hu and D. Wang, Adv. Funct. Mater., 2021, 32, 2107672.
14. S. H. C. Ricea, R. I. Masel, P. Waszczukb, A. Wieckowski and T. Barnard, J. Power Sources, 2002, 111, 83–89.
15. J. D. Lović, A. V. Tripković, S. L. Gojković, K. D. Popović, D. V. Tripković, P. Olszewski and A. Kowal, J. Electroanal. Chem., 2005, 581, 294–302.
16. S. Uhm, H. J. Lee and J. Lee, Phys. Chem. Chem. Phys., 2009, 11, 9326–9336.
17. S. Yang, J. Kim, Y. J. Tak, A. Soon and H. Lee, Angew. Chem., Int. Ed., 2016, 55, 2058–2062.
18. L. An, T. S. Zhao, X. L. Zhou, L. Wei and X. H. Yan, RSC Adv., 2014, 4, 65031–65034.
19. Y. Yao, J. Lei, Y. Shi, F. Ai and Y.-C. Lu, Nat. Energy, 2021, 6, 582–588.
20. Q. Dai, F. Xing, X. Liu, D. Shi, C. Deng, Z. Zhao and X. Li, Energy Environ. Sci., 2022, 15, 1594–1600.
21. L. Zhi, C. Liao, P. Xu, F. Sun, C. Yuan, F. Fan, G. Li, Z. Yuan and X. Li, Energy Environ. Sci., 2024, 17, 717–726.
22. Y. Shi, Z. Wang, Y. Yao, W. Wang and Y.-C. Lu, Energy Environ. Sci., 2021, 14, 6329–6337.
23. G.-M. Weng, B. Yang, C.-Y. Liu, G.-Y. Du, E. Y. Li and Y.-C. Lu, Energy Environ. Sci., 2019, 12, 2244–2252.
24. S. Sun, L. Fang, H. Guo, L. Sun, Y. Liu and Y. Cheng, Adv. Sci., 2023, 10, e2207728.
25. S. Wang, L. Ma, S. Niu, S. Sun, Y. Liu and Y. Cheng, Angew. Chem., Int. Ed., 2024, 63, e202316593.
26. C. Sun and H. Zhang, ChemSusChem, 2021, 15.
27. Z. Wang, Z. Guo, J. Ren, Y. Li, B. Liu, X. Fan and T. Zhao, ACS Cent. Sci., 2022, 9, 56–63.
28. K. Ma, Y. Zhang, L. Liu, J. Xi, X. Qiu, T. Guan and Y. He, Nat. Commun., 2019, 10, 5286.
29. Q. Luo, L. Li, W. Wang, Z. Nie, X. Wei, B. Li, B. Chen, Z. Yang and V. Sprenkle, ChemSusChem, 2012, 6, 268–274.
30. C. Sun, J. Chen, H. Zhang, X. Han and Q. Luo, J. Power Sources, 2010, 195, 890–897.
31. T. Puleston, M. Serra and R. Costa-Castelló, Appl. Energy, 2024, 355.
32. Y. Zhang, L. Liu, J. Xi, Z. Wu and X. Qiu, Appl. Energy, 2017, 204, 373–381.
33. K. Wang, L. Liu, J. Xi, Z. Wu and X. Qiu, J. Power Sources, 2017, 338, 17–25.
34. L. Wei, T. S. Zhao, Q. Xu, X. L. Zhou and Z. H. Zhang, Appl. Energy, 2017, 190, 1112–1118.
35. A. H. Whitehead and M. Harrer, J. Power Sources, 2013, 230, 271–276.
36. J. Piwek, C. R. Dennison, E. Frackowiak, H. Girault and A. Battistel, J. Power Sources, 2019, 439, 227075.
37. Z. Li, L. Liu, Y. Zhao, J. Xi, Z. Wu and X. Qiu, J. Power Sources, 2019, 438, 226990.
38. G. Park, Y. Lim, K. Hyun and Y. Kwon, J. Power Sources, 2024, 589, 233770.
39. L. Wei, X. Z. Fan, H. R. Jiang, K. Liu, M. C. Wu and T. S. Zhao, J. Power Sources, 2020, 478, 228725.
40. J. Heo, J. Y. Han, S. Kim, S. Yuk, C. Choi, R. Kim, J. H. Lee, A. Klassen, S. K. Ryi and H. T. Kim, Nat. Commun., 2019, 10, 4412.
41. H. R. Jiang, J. Sun, L. Wei, M. C. Wu, W. Shyy and T. S. Zhao, Energy Storage Mater., 2020, 24, 529–540.
42. U. L. Zhenni Ma, E. Pahija, J. R. Tavares and D. C. Boffito, Ind. Eng. Chem. Res., 2021, 60, 803–815.
43. C. Rice, S. Ha, R. I. Masela and A. Wieckowski, J. Power Sources, 2003, 115, 229–235.
44. M. Choi, C.-Y. Ahn, H. Lee, J. K. Kim, S.-H. Oh, W. Hwang, S. Yang, J. Kim, O.-H. Kim, I. Choi, Y.-E. Sung, Y.-H. Cho, C. K. Rhee and W. Shin, Appl. Catal., B, 2019, 253, 187–195.
45. X. Hu, Z. Xiao, W. Wang, L. Bu, Z. An, S. Liu, C.-W. Pao, C. Zhan, Z. Hu, Z. Yang, Y. Wang and X. Huang, J. Am. Chem. Soc., 2023, 145, 15109–15117.
46. C. Venkateswara Rao, C. R. Cabrera and Y. Ishikawa, J. Phys. Chem. C, 2011, 115, 21963–21970.
47. S. Luo, W. Chen, Y. Cheng, X. Song, Q. Wu, L. Li, X. Wu, T. Wu, M. Li, Q. Yang, K. Deng and Z. Quan, Adv. Mater., 2019, 31, 1903683.
48. Y. Sun, W. Chen, W. Zhang, Y. Nie, Q. Zhang, L. Gu, M. Luo and S. Guo, Adv. Funct. Mater., 2023, 33, 2303299.
49. Y. Wan, H. Zhou, M. Zheng, Z. H. Huang, F. Kang, J. Li and R. Lv, Adv. Funct. Mater., 2021, 31, 2100300.
50. F. Xiao, Y. Wang, G. L. Xu, F. Yang, S. Zhu, C. J. Sun, Y. Cui, Z. Xu, Q. Zhao, J. Jang, X. Qiu, E. Liu, W. S. Drisdell, Z. Wei, M. Gu, K. Amine and M. Shao, J. Am. Chem. Soc., 2022, 144, 20372–20384.
51. X. Yu and P. G. Pickup, Electrochim. Acta, 2011, 56, 4037–4043.
52. W. Zhu, X. Zhang, F. Yao, R. Huang, Y. Chen, C. Chen, J. Fei, Y. Chen, Z. Wang and H. Liang, Angew. Chem., Int. Ed., 2023, 62, 202300390.
53. Y. K. Zeng, T. S. Zhao, X. L. Zhou, J. Zou and Y. X. Ren, J. Power Sources, 2017, 352, 77–82.
54. J. Lei, Y. Zhang, Y. Yao, Y. Shi, K. L. Leung, J. Fan and Y.-C. Lu, Nat. Energy, 2023, 8, 1355–1364.
55. Y. Ahn, J. Moon, S. E. Park, J. Shin, J. Wook Choi and K. J. Kim, Chem. Eng. J., 2021, 421, 127855.
56. S. Yu, X. Yue, J. Holoubek, X. Xing, E. Pan, T. Pascal and P. Liu, J. Power Sources, 2021, 513, 230457.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ee02450h

‡ These authors contributed equally.

---