Early View e202301659
Research Article
Open Access

Unveiling the Role of Electrografted Carbon-Based Electrodes for Vanadium Redox Flow Batteries

Matthias Kogler

Matthias Kogler

Institute of Applied Physics, Vienna University of Technology, 1040 Vienna, Austria

Center for Electrochemical Surface Technology GmbH, 2700 Wr. Neustadt, Austria

Contribution: Data curation (lead), ​Investigation (lead), Methodology (supporting), Writing - original draft (lead)

Search for more papers by this author
Nikolai Rauh

Nikolai Rauh

Institute of Applied Physics, Vienna University of Technology, 1040 Vienna, Austria

Contribution: Data curation (supporting), ​Investigation (supporting)

Search for more papers by this author
Dr. Soniya Gahlawat

Dr. Soniya Gahlawat

Institute of Applied Physics, Vienna University of Technology, 1040 Vienna, Austria

Center for Electrochemical Surface Technology GmbH, 2700 Wr. Neustadt, Austria

Contribution: Data curation (supporting), ​Investigation (supporting)

Search for more papers by this author
Muhammad Adeel Ashraf

Muhammad Adeel Ashraf

Center for Electrochemical Surface Technology GmbH, 2700 Wr. Neustadt, Austria

Contribution: Data curation (supporting), ​Investigation (supporting)

Search for more papers by this author
Dr. Markus Ostermann

Dr. Markus Ostermann

Center for Electrochemical Surface Technology GmbH, 2700 Wr. Neustadt, Austria

Contribution: Formal analysis (supporting), ​Investigation (supporting), Writing - review & editing (supporting)

Search for more papers by this author
Prof. Markus Valtiner

Prof. Markus Valtiner

Institute of Applied Physics, Vienna University of Technology, 1040 Vienna, Austria

Center for Electrochemical Surface Technology GmbH, 2700 Wr. Neustadt, Austria

Contribution: Conceptualization (supporting), Funding acquisition (lead), Methodology (equal), Supervision (equal), Validation (supporting), Writing - review & editing (supporting)

Search for more papers by this author
Dr. Christian M. Pichler

Corresponding Author

Dr. Christian M. Pichler

Institute of Applied Physics, Vienna University of Technology, 1040 Vienna, Austria

Center for Electrochemical Surface Technology GmbH, 2700 Wr. Neustadt, Austria

Contribution: Conceptualization (lead), Funding acquisition (supporting), ​Investigation (supporting), Methodology (equal), Supervision (equal), Validation (lead), Writing - review & editing (lead)

Search for more papers by this author
First published: 22 March 2024

Graphical Abstract

Carbon-based materials play a pivotal role for vanadium redox reactions, yet the origin of their active surface remains a contentious topic. This study systematically explores the impact of various structural elements and functional groups on the activity of graphite felt electrodes. Contrasting non-activated and thermally activated felt reveals the striking discrepancy between model and real-world system.

Abstract

Carbon-based electrodes are used in flow batteries to provide active centers for vanadium redox reactions. However, strong controversy exists about the exact origin of these centers. This study systematically explores the influence of structural and functional groups on the vanadium redox reactions at carbon surfaces. Pyridine, phenol and butyl containing groups are attached to carbon felt electrodes. To establish a unique comparison between the model and real-world behavior, both non-activated and commercially used thermally activated felts serve as a substrate. Results reveal enhanced half-cell performance in non-activated felt with introduced hydrophilic functionalities. However, this cannot be transferred to the thermally activated felt. Beyond a decrease in electrochemical activity, a reduced long-term stability can be observed. This work indicates that thermal treatment generates active sites that surpass the effect of functional groups and are even impeded by their introduction.

Introduction

The imminent energy transition and replacement of fossil fuels necessitates continuous growth in renewable energy sources. However, the inherent volatilities of wind, water, and solar power require adequate energy storage systems to balance their fluctuations. Cost-effective, safe, and reliable large-scale energy storage solutions have become essential to facilitate the energy transition. Among various electrochemical storage technologies, vanadium redox flow batteries (VRFBs) have gained widespread attention due to their high energy efficiency, rapid response time, easy scalability and low environmental impact.1-3 The latter is particularly attributed to the ability of disassembling these systems and effectively recovering the electrolyte, a key component of the battery.4-6 For long-term energy storage, durability and consistent power output are crucial. Vanadium redox flow batteries utilize two redox couples, urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0001 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0002 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0003 in the negative and urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0004 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0005 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0006 in the positive electrolyte, with charge storage reactions proceeding at the electrode surface. The electrode provides reaction sites, influencing characteristics like energy conversion efficiency and power density.

The exact nature of the interaction between vanadium species and electrode surface, even though the principle of VRFBs having been first proposed as early as the 1980s,7 is still subject to discussion and fueling ongoing research.8-10 In commercial systems, carbon-based materials, more precisely graphitic carbon felts, are commonly used as electrodes in both half-cells. Despite their advantageous properties such as low cost, electronic conductivity, porosity, ease of handling and good mechanical strength, pristine graphite felt has the distinct disadvantage of low electrochemical activity with respect to vanadium redox reactions.11-13 In order to improve the electrode kinetics and to enhance the affinity of the surface towards the vanadium species, prior activation of the graphite fibers is therefore required. For this purpose, thermal, chemical, electrochemical or plasma treatments can be applied, aiming to modify physical and chemical properties in a way to enhance the redox reaction kinetics and introduce preferential reaction sites.14-19 Many studies suggested that the latter is achieved by introducing abundant oxygen functional groups. Hydroxyl and carboxyl groups are assumed to serve as the main binding sites for vanadium ions and thus facilitate the charge storage reaction.20-22 In the multitude of studies, however, contradictory results are often obtained regarding the exact involvement of the oxygen groups and their influence on the battery performance.

Only a few studies question the genuine importance of these oxygen groups and thus the fundamental reaction mechanism. Within this context, some studies emphasize the non-negligible influence of structural changes (such as introduction of carbon-edge sites) during the activation process,23-25 while others achieve performance enhancement by heteroatom substitution, especially with nitrogen,26-28 thus fundamentally querying the importance of oxygen functional groups. This prevailing lack of consensus on the nature of interfacial processes in redox flow batteries makes it particularly difficult to understand subsequent processes such as performance deterioration or degradation phenomena. Therefore, a unified understanding regarding the influence of individual functional groups and the nature of the surface structure in general, is a prerequisite to advance the development of vanadium redox flow battery systems.

This work provides a systematic investigation on how different structural elements and functional groups on the carbon electrode influence the vanadium redox reactions. Functional group-containing molecules were attached to the graphite felt surface via an electrochemical grafting method. At first, pristine PAN-based graphite felts served as a model system. Subsequently, the procedure was extended to commercial thermally activated electrodes in order to evaluate the effects with materials applied in real-world RFB systems with higher structural complexity. Contrary to most other previous studies, we investigate the real-world electrode material and not only model systems. Structural characterization was performed by Raman spectroscopy and X-ray photoelectron spectroscopy (XPS), while electrochemical methods such as cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) were used to investigate the respective half-cell activity. Additionally, half-cell as well as flow-cell cycling tests were conducted for selected materials.

This approach allowed it to gain new insights regarding the actual impact of oxygen and nitrogen functional groups on the vanadium redox reaction. Furthermore, the decisive role of carbon defects (vacancies or edge sites) could be highlighted. Herein, crucial differences between commonly applied model systems and real-world electrodes were discovered.

Results and Discussion

Contradictory results exist regarding the activity of specific functional groups towards the vanadium redox reactions.11, 23, 27 In order to provide more clarification, a targeted modification of the graphite felt surface with selected functional groups was carried out, followed by a comprehensive structural and electrochemical characterization.

The precursor molecules, depicted in Figure 1, were selected due to the distinct chemical characteristics of their head groups. Hydroxyl as well as nitrogen groups are reported to facilitate the redox reaction in question, while the butyl group is expected to impede the process due to hydrophobic and steric effects. The molecules were electrografted onto the graphite fiber surface exploiting a diazonium ion intermediate, followed by the release of N2. The detailed procedure and reaction mechanism being described elsewhere.29

Details are in the caption following the image

Molecules used for the modification via electrochemical grafting. (a) 4-(aminomethyl)pyridine, (b) 4-(aminomethyl)phenol, (c) 4-butylaniline.

Untreated, non-activated graphite felt (GF) served as an initial model electrode material to study the influence of the functional groups in the absence of structural changes due to pre-activation.

Structural and Morphological Properties

Non-Activated Felt

The three samples modified with molecules containing pyridine (Py) (Figure 1a), phenol (Ph) (Figure 1b) and butyl moieties (BuAn) (Figure 1c) as well as pristine non-activated graphite felt (Prist) were first subjected to physicochemical characterization to understand the structural and chemical changes of the surface.

Morphological changes caused by the modification process were studied using SEM. Figure 2a shows that the general structure of the modified fibers remains unchanged compared to pristine felt. This suggests that no microstructural damage is induced during synthesis and that the chemical alteration is limited to the surface region. This is confirmed by SEM-EDX measurements (Figure S1), where no significant differences in elemental bulk composition were found among the samples.

Details are in the caption following the image

Structural characterization of pristine and electrochemically modified non-activated (GF) and thermally activated (GF-TA) graphite felt. (a, b) SEM images displaying the morphological structure of the pristine and modified carbon fibers. (c) Deconvoluted C 1s high-resolution XPS spectra of GF and GF-TA samples.

To study the surface chemical composition of the graphite fibers XPS was applied. The C 1s region (Figure 2c) was analyzed in more detail, to gain deeper insight into the chemical nature of the carbon species. Seven different peak contributions can be distinguished in all samples. The peaks were attributed to sp2 and sp3 hybridized carbon at ~284.4 eV and ~285 eV, hydroxyl and ether groups at ~286.3 eV, carbonyls at ~287.5 eV, carboxyl at ~289.0 eV and π-π* shake-up satellite at 290.5–293 eV30-32 Another feature lower than sp2 carbon at ~284.0 eV was assigned to disordered carbon (Cd), associated with the altered environment for C in aliphatic groups,33-35 and was included in all further analyses.

When comparing the different samples, a modest increase in the C−O fraction in GF-Ph was found, associated with the introduction of the phenol group. A more pronounced difference was determined for GF-BuAn, where the ratio of sp2 to sp3/Cd was significantly changed. This is attributed to successful modification with the butyl group, which increases the relative sp3 content on the surface, thereby increasing the sp3/Cd signal contribution.

Spectra of the core level regions of C 1s and O 1s were used to determine the O/C ratio on the surface (Figure 3), as well as N 1s to monitor the nitrogen content.

Details are in the caption following the image

Oxygen to carbon ratio on the electrode surface of GF and GF-TA samples obtained by XPS analysis.

As expected, pristine GF shows the lowest oxygen (O) to carbon (C) ratio (0.02), indicating a low degree of initial surface oxidation. After modification with 4-(aminomethyl)phenol, this ratio increases to 0.15, while for modification with pyridine and butyl group the ratio rises to 0.06 and 0.08, respectively. This indicates that the electrochemical grafting process can lead to the introduction of oxygen functional groups, due to the acidic reaction media and oxidative electrochemical potential applied.29, 36 However, the significantly increased O/C ratio for the Ph-modified sample suggests that the immobilization of the O-containing phenol group was successful.

When investigating the nitrogen species via the N 1s spectra (Figure S2) presence of N is found for all modified samples, but not for the pristine sample. The predominant peak at ~401.4 eV is related to graphitic N, which can be incorporated into the felt by the diazo mediated modification process. An additional peak can be identified at ~399.7 eV, which is particularly pronounced for BuAn and can be attributed to the presence of a surface bound diazo group.29, 37-39 In this case it must be considered that the modification agent contains an aniline moiety instead of the benzylamine structure of the other compounds. It has been shown that anilines can be also immobilized by formation of diazo-bonds (retaining the nitrogen),40 which is less common for the benzylamine (Figure 1a, b) structures, which are forming a direct C−C bond (via the intermediate benzyl radical).

To study the structural nature of the carbon substrates in greater detail, Raman spectroscopy was applied. The intensity ratio of D and G band (ID/IG) was determined, to assess the degree of structural disorder, representing a measure of graphitic defects in the material. A higher value corresponds to an increased number of defect sites in the graphitic structure. As illustrated for GF-Prist in Figure 4a, the D band appears at ~1350 cm−1 and is associated with the A1g breathing mode. This is accompanied by the G band (~1584 cm−1), corresponding to the first-order allowed Raman mode E2g.41-43 To access the intensity values, the raw first-order spectra had to be deconvoluted, using a total of five distinct peaks. Alongside the D and G bands, contributions from D*, D‘ and D‘‘ were included in the evaluation. Figure 4b shows Raman spectra of all four GF samples. As depicted in Figure 4e, the pristine non-activated felt material (0.77) exhibits a lower degree of disorder, while the number of defects increased after modification. The highest number of defects was observed in GF-Ph (1.14), whereas the effect is less pronounced in GF-Py (0.91) and GF-BuAn (1.00).

Details are in the caption following the image

Raman spectroscopic analysis of the modified GF samples. (a, c) Exemplary Raman spectrum of the first-order modes of GF-Prist and GF-TA-Prist, deconvoluted with five peaks (D*, D, D”, G, D’). (b, d) Raman spectra of all GF and GF-TA samples, given a vertical offset for better visualization. First-order D and G band as well as second-order modes are depicted. (e) A comparison of ID/IG ratio for GF and GF-TA samples is given, obtained from the deconvolution of the first-order spectra.

The overall increase in ID/IG can be explained by the fact that introducing molecules via the radical mechanism necessitates bond rearrangements within the graphitic structure (forming of an additional bond to the modifying molecules). As XPS results show, also some surface oxidation is occurring. Both factors together increase the defect concentration in the graphitic carbon structure. The varying extent of this phenomenon for the different molecules can be attributed to the differing degrees of surface modification and parallel oxidation.

Thermally Activated Felt

For real-world applications, thermally activated carbon felt has established itself as the preferred electrode over non-activated GF, as it exhibits significantly better cell performance. Furthermore, the facile activation process makes it attractive for utilization on larger scale.

To investigate whether the previously obtained findings on the surface functionalization can be transferred to real-world conditions, modification experiments were extended to thermally activated graphite felt samples. The chemical surface alteration was conducted following the same procedure as for non-activated felts, with the molecules shown in Figure 1. Pristine thermally activated graphite felt (GF-TA-Prist) served as a reference sample.

Morphological investigations by SEM (Figure 2b) reveal an increased number of pits compared to the non-activated felt (Figure 2a), associated with thermal treatment. This microstructural change is considered to be responsible for the increase in surface area and the exposure of additional edge sites.44

Moreover, this process is generally performed in an oxygen-rich atmosphere,45 leading to thermal oxidation of the surface, which can be confirmed by an increased initial oxygen content of GF-TA-Prist (0.04), depicted in Figure 3. As SEM-EDX (Figure S3) does not show any noticeable compositional changes, this process only occurs on the surface of the graphitic fiber. Similar to non-activated (GF) samples, an increase in oxygen can be observed in all modified samples, Py (0.05), Ph (0.08) and BuAn (0.07). Once again, this is attributed to the mild oxidation, e. g. introduction of oxygen functional groups during the electrochemical treatment. The increased O/C ratio of GF compared to GF-TA after modification can be explained by GF-TA initially having fewer oxidizable sites than GF. As was shown by Derr et al.,46 oxidizable groups in thermally activated material can be saturated by atmospheric oxygen after a comparable time of storage. Hence, electrochemical grafting can lead to a higher background oxidation in GF compared to GF-TA. However, while this effect is less prominent in GF-TA-Py and GF-TA-BuAn, a more pronounced increase is found for the introduction of phenol groups (GF-TA-Ph).

This observation can be corroborated by detailed analysis of the C 1s region of modified GF-TA samples (Figure 2d). Deconvolution of the high-resolution spectra was performed as described in Section “Non-activated felt”. From this, it can be extracted that BuAn again shows an increased ratio of sp3 and Cd compared to sp2, suggesting a successful incorporation of butyl groups onto the surface.

Raman spectroscopic studies further revealed the distinct microstructural difference between non-activated (Figure 4a, b) and thermally activated graphitic materials. Using again a total number of five peaks to deconvolute the region of interest (Figure 4c), a comparative analysis of all four samples in terms of change in defect concentration was conducted (Figure 4d). Herein, the well-established ID/IG ratio was used to determine the degree of disorder in the modified felts. Figure 4e demonstrates the pronounced effect of thermal treatment on fiber structure, with ID/IG almost doubling from GF-Prist (0.77) to GF-TA-Prist (1.42). The already stated reasons for this can be found in the cleavage or rearrangement of carbon bonds caused by thermal processing, as well as the subsequent decomposition of oxygen functionalities from deeper graphitic layers.44, 47 These processes are now considered to be responsible for the increased electric double layer capacitance (EDLC) (as shown later), the higher number of defects and thus the greater number of active centers. A notable difference can be observed when comparing the effects of the modifications depending on the two different substrates, GF and GF-TA.

While with the former an increase in defect concentration for GF-Py and GF-Ph (Figure 4e) is visible, the very opposite is found for both groups with GF-TA as the substrate (ID/IG 1.2). The trend towards a decrease in disorder could be attributed to the partial healing of the defect structures by the introduction of the aromatic molecules. However, this behavior is not apparent for BuAn, which may caused by the molecular structure itself.

Electrochemical Properties

Non-Activated Felt

To assess the electrode performance in vanadium redox flow batteries, electrochemical half-cell experiments were performed, including cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS).

Figure 5a and 5b illustrate the notable discrepancy in activity of non-activated felt (GF) towards the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0007 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0008 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0009 (negative) and urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0010 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0011 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0012 (positive) redox reaction, respectively.

Details are in the caption following the image

CV curves of GF (a, b) and GF-TA (c, d) samples in the negative (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0013 ) and positive (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0014 ) half-cell. All spectra were recorded with a scan rate of 3 mV/s in a 10 mM urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0015 electrolyte. urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0016 is displayed for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0017 redox peaks of GF (b).

The pristine electrode shows nearly no activity for the negative half-cell reaction. While the oxidation peak (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0018 ) is only weakly pronounced, the reverse reaction (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0019 ) is superimposed by the H2 production due to the large peak-to-peak separation and the low reduction current.

Different aspects of the cyclic voltammogram can be utilized to assess the electrochemical performance of the electrodes. Apart from the maximum peak current Ip, the ratio of anodic and cathodic peak current Ip,a/Ip,c as well as the separation between the oxidation and reduction reaction peak urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0020 can be understood as a measure of the reversibility of the electron transfer process.

Although a quantitative evaluation of the negative half-cell performance was not feasible due to the overlapping H2 evolution in the low potential region, Figure 5a shows a significant increase in performance for GF-Py and GF-Ph, while GF-BuAn, on the other hand, leads to complete deactivation.

This deterioration can also be observed for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0021 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0022 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0023 reaction. The GF-BuAn sample shows non-reversible behavior with a peak current ratio (Ip,a/Ip,c) of 3.4 (Figure 6a), which however improves to some extent with higher scan rate. Furthermore, an almost twofold higher urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0024 (Figure 5b) is observed for GF-BuAn (686 mV) compared to GF-Prist (389 mV). It has to be mentioned that due to overall peak shape and additional broad features at ~0.2 V and ~0.9 V, attributed to azo functionalities,48, 49 the quantitative evaluation of GF-BuAn remained challenging. The deteriorating behavior of GF-BuAn can be explained by the enhanced hydrophobic characteristics resulting from the incorporation of an alkyl chain. Furthermore, the presence of butyl groups causes steric hindrance at existing active centers at the graphitic lattice.

Details are in the caption following the image

Electrochemical characterization of non-activated GF samples. (a) Evaluation of the reversibility via the peak current ratio for all measured GF samples, depicted for the positive half-cell at scan rates 3, 5, 7, 10 and 20 mV/s. Due to the shape of CV curves, the values for the negative half-cell could not be evaluated. (b) Electrical double layer capacitance evaluated as a measure for the electrochemically active surface area. (c) Electrochemical impedance spectra of GF-Prist for the positive half-cell, recorded at 0.4 V vs. Hg/Hg2SO4 in the range of 1–50 MHz. The illustrated equivalent circuit was used for fitting the raw data. (d) Evaluation of the charge transfer resistance (RCT) for the negative and positive half-cell, obtained from impedance measurements at −1.0 V and 0.4 V vs. Hg/Hg2SO4, respectively.

Examining the maximum peak current Ip both GF-Ph and GF-Py exhibit an improvement, with no significant difference between the two groups. This can be attributed to the general increase in number of active sites as well as in hydrophilicity due to the introduction of the functionalities. The type of functional group does not appear to be decisive for the activity towards the positive half-cell reaction. Also in terms of reversibility (Ip,a/Ip,c, Figure 6a) similar behavior can be observed in comparison with non-activated felt. It is worth noting that overall, there is an imbalance favoring the oxidation reaction of (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0025 ) over the reduction (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0026 ). Regarding urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0027 the behavior remains nearly unchanged with determined values of 404 mV for pyridine and 392 mV for phenol (Figure 5b).

The results obtained from CV studies highlight the importance of adequate functionalization of the graphite felt, with the observed effect having a greater impact on the negative half-cell. This can be attributed to different reaction mechanisms, as an inner-sphere mechanism is assumed for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0028 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0029 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0030 couple and and outer sphere mechanism for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0031 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0032 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0033 couple.50, 51

The inner sphere mechanisms for the negative half-cell, includes an adsorption step of the vanadium species on the electrode surface. The pyridine or phenol functionalities can serve as adsorption sites for this purpose, explaining the improved electrochemical characteristics in Figure 5a. For the positive half-cell general surface characteristics, such as conductivity or hydrophilicity, more significant than the specific nature of the reaction sites. Hence, the relative activity enhancement is more pronounced for the negative half reaction.

The next parameter determined was the electrical double layer capacitance (EDLC) (Figures 6b and S4).52-54

Pristine GF exhibits an expectedly low EDLC (0.57 mF) due to the absence of defects and primary exposure of basal planes.47 In contrast, GF-Py, despite having a higher ID/IG ratio (Figure 4e), does not show a significant increase (0.61 mF). However, this changes for GF-Ph (2.60 mF), which is consistent with Raman investigations and the elevated oxygen concentration on the electrode surface (Figure 3). Despite its presumed detrimental properties, GF-BuAn shows a high EDLC, potentially attributed to possible overlap of hidden Faradaic reactions, affecting the evaluation.

EIS experiments were performed for both half-cells, to assess the charge transfer resistance urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0034 (Figure 6c, d) for the respective redox reaction. This parameter can be associated with the electron transfer properties at the electrode/electrolyte interface and therefore enables a comparative analysis of the electrode kinetics. Figure 6c depicts the representative impedance spectra of pristine GF for the positive and negative half-cell and the applied equivalent circuit (for the full spectra see also Figure S5). Herein, Rel (Table S1) represents the resistance originating from the electrolyte, RCC/GF and CPECC/GF the interface between graphite current collector and graphite felt, respectively, and W the diffusion dependent Warburg impedance.

Figure 6c highlights the substantial difference in RCT between positive and negative half-cell performance, which is in agreement with the findings of the CV studies (Figure 5a, b).

EIS reveals a greater resistance contribution oft the negative half-cell, with an 8-fold higher RCT observed for GF-Prist material in the positive half-cell (Figure 6d). The modification with Py and Ph again has no major influence on the RCT values, while BuAn leads to a visible deterioration. The higher RCT values for the negative half reaction confirm previous literature results and indicate that that the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0035 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0036 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0037 couple is responsible for a larger share of the overall resistance in the VRFB system.

Except for GF-BuAn, little to no change in RCT can be observed for the positive half-cell. This indicates that the type of functional group does not play a fundamental role in the reaction mechanism urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0038 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0039 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0040 as long as the macroscopic properties (conductivity, wettability) are not negatively affected. Both pyridine and phenol show comparable activity in this regard. For GF-BuAn, however, electron transfer is significantly hindered at the electrode/electrolyte interface, as reflected in an RCT approximately 13 times higher than that of GF-Prist. This increase is attributed to a reduced electrical conductivity and increased hydrophobicity caused by the modification.

Concludingly, it could be shown that the introduction of Ph and Py functionalities increases the Ip values for both half-cell reactions. This seems to support previous claims, that the functionalities can indeed serve as adsorption centers for the vanadium redox reaction. This is supported by the fact that the relative increase of Ip is higher for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0041 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0042 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0043 reaction, an inner sphere reaction that actually contains an adsorption step. While the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0044 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0045 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0046 couple, as outer sphere reaction, only shows a comparable slight improvement. As expected, the introduction of butyl groups is highly detrimental for the electrochemical performance by increasing hydrophobicity and blocking access of the vanadium species to the surface. However, it must be proved that this behavior is also valid for real-world electrode materials.

Thermally Activated Felt

As demonstrated by the results above, non-activated GF is, despite modification-related advancements, not suitable for use as electrode material in commercial VRFB applications, due to its overall unfavorable electrochemical performance. Hence, it has to be examined whether the optimized behavior of modified GF can be extended to thermally activated graphite felt.

The electrochemical response of the GF-TA samples reinforces the previously made assumptions regarding the half-cell specific effect of the activation process. As Figure 5a visualizes, the electrochemical behavior in the negative half-cell reaction urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0047 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0048 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0049 is profoundly affected by the thermally induced activation, exhibiting a dramatic increase in performance compared to non-activated GF. This is reflected not only by a substantial increase in Ip, but also by a significantly improved reversibility, represented by Ip,a/Ip,c (Figure 7a), and a reduced peak-to-peak separation urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0050 (Figure 7b).

Details are in the caption following the image

Electrochemical characterization of thermally activated GF-TA samples. (a) Evaluation of the reversibility via the peak current ratio for all measured GF-TA samples, depicted for both half-cells at scan rates 3, 5, 7, 10 and 20 mV/s. (b) Electrical double layer capacitance of all GF-TA samples, evaluated as a measure for the electrochemically active surface area. (c) Peak-to-peak separation urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0051 between the redox peaks as a measure for electrocatalytic activity. (d) Evaluation of the charge transfer resistance (RCT) for the negative and positive half-cell, obtained from impedance measurements at −1.0 V and 0.4 V vs. Hg/Hg2SO4, respectively.

While the introduction of pyridine and phenol groups leads to an enhanced activity for non-activated GF, thermally activated GF shows a different behavior. Results indicate that the chemical modification leads to a slight deactivation regarding the electrochemical performance of the thermally activated graphite felt. This is indicated by the lower Ip values for Ph and Py modified samples in Figure 5c, d. Additionally, the onset of the H2 evolution is shifted to more positive potentials, even for the BuAn modified sample, that showed opposite behavior on the non-activated GF. For BuAn the decrease in Ip is also clearly more pronounced for both half-cells compared to Py and Ph. (Figure 5c, d).

In terms of reversibility, GF-TA modified samples exhibit an overall improved behavior, with now the reduction being slightly favored for both sides (Figure 7a). At 3 mV/s, the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0052 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0053 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0054 reaction features a reversibility of about 0.9 (Prist, Py and Ph) and 0.87 (BuAn), respectively, stabilizing for all samples at about 0.95 with increasing scan rate. Also urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0055 was significantly improved for the thermally activated GF, giving a value of 94 mV (Figure 7b).

For the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0056 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0057 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0058 reaction, on the other hand, the Ip,a/Ip,c ratios improved markedly and show for GF-TA-Prist a ratio of 0.99, Py and Ph 0.96 and BuAn 0.95. Surprisingly, herein all samples experience a decrease at 7 mV/s, shifting the ratio more towards favoring reduction, with the initial behavior being restored at 20 mV/s.

Also for the positive half-cell urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0059 has improved compared to non-activated GF, but GF-TA-Prist illustrates the substantial discrepancy between the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0060 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0061 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0062 (202 mV) and urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0063 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0064 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0065 (94 mV) reaction. (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0066 could not be determined for the negative half reaction with the non-activated GF). This indicates that a major effect of utilizing thermally activated GF is actually the improvement of the electrochemical mechanism in the negative half reaction.

In addition to the higher Ip values, also larger electrochemical double layer capacitance is found for the GF-TA samples, as shown in Figure 7c. The EDLC of GF-TA-Prist (26.02 mF) shows a more than 40-fold increase compared to GF-Prist (0.57 mF, Figure 6b), which highlights the structural effect of the thermal processing. Although the samples show a higher EDLC after modification with Py (32.82 mF) and Ph (30.39 mF), no distinct positive correlation with the electrochemical activity can be established. However, analyzing the lower EDLC of GF-TA-BuAn (17.27 mF), it appears that a stronger influence on the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0067 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0068 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0069 reaction occurs, whereas the evident effect on the opposite side turns out to be minor. This suggests that although the chemical modification leads to an increase in EDLC due incorporation of the functional groups mentioned (Py and Ph), these do not provide relevant active sites for the vanadium redox reactions to the same extent. As demonstrated, it must be emphasized that the use of EDLC as a descriptor for electrochemical behavior is not without limitations, as detailed performance analysis can be overshadowed by other factors such as the effects of carbon defects, as shown later.

Finally, EIS measurements were conducted. The data depicted in Figure 7d, provide further explanation for the improved performance of GF-TA. The RCT (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0070 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0071 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0072 ) exhibits drastic reduction from 577 Ω (non-activated GF-Prist) to 1.4 Ω (GF-TA-Prist) after activation, whereas this step is less pronounced for urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0073 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0074 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0075 (70 Ω to 9.5 Ω). Considering the similar electrochemical response of GF-TA-Prist on both sides, this significant reduction corresponds to an increased activity, especially facilitating the negative half-cell reaction, highlighting the different reaction mechanisms regarding the two vanadium redox couples. Modification with Py or Ph lead to a minor additional decrease in RCT (urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0076 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0077 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0078 ), while for the negative half reaction the values remain nearly unchanged. Only for BuAn a substantial RCT increase is found for both half reactions, indicating deterioration in electrochemical performance, as expected.

It must be highlighted that for the GF-TA the additional modification of the felts with Py or Ph only led to minor improvements among the measured electrochemical parameters, e. g. RCT (for the positive half-cell reaction) and urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0079 . This is contrasting the results from the non-activated GF, where Py and Ph modification led to higher Ip values for both half-cell reactions, which is a key indicator for the electrochemical performance. This raises doubts about the actual influence of functional groups on the electrochemical reactions. As it would be expected, that if the functional groups (Py, Ph) would act as active sites for the vanadium redox reactions, this effect should also be reproducible for the GF-TA materials to a more distinct extent. It is often argued that thermal activation introduces oxygen functionalities as well (which was confirmed in this study), which are responsible for the improved performance of those materials. However, the modification of the GF-TA with Ph introduces even more oxygen functionalities without improving the electrochemical performance metrics considerably. Furthermore, the O/C ratios are in a very similar range for the pristine non-activated GF and GF-TA.

Hence, it seems that oxygen functionalities appear not to be a key factor for determining the electrode performance in VRFBs.

A significant difference is nevertheless found when comparing the ID/IG values, which are significantly higher for GF-TA compared to non-activated GF. Furthermore, an exact opposite trend to the defect concentration can be observed when modifying the carbon felts with Py and Ph. For the non-activated GF samples, the ID/IG ratio increases upon Py or Ph modification, which is reflected in increased Ip values for both half-cell reactions. For modified GF-TA samples, however, the ID/IG ratio decreases upon Py or Ph modification and Ip values are reduced. This is a strong indication that the defect concentration in the carbon material is of greater relevance for the electrochemical performance.23

Half-Cell Cycling Behavior

To investigate the long term stability of the modified samples, half-cell aging experiments were performed. Since information about the real-world behavior is acquired, only thermally activated electrodes were used for these measurements. GF-TA-BuAn was excluded due to its poor electrochemical performance, as demonstrated earlier. Therefore, GF-TA-Prist, GF-TA-Py and GF-TA-Ph were cycled 100 times in the negative and positive half-cell, respectively. urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0080 and Ip,a/Ip,c were determined for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0081 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0082 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0083 reaction. However, due to the overlapping H2 evolution for the negative half-cell reaction, a quantitative assessment herein was not feasible.

Figure 8 shows the aging performance of the sample electrodes for the negative half-cell. GF-TA-Py exhibits an enhanced initial activity, compared to GF-TA-Prist, which cannot be sustained as cycling continues. This pronounced decrease in stability cannot be found for GF-TA-Ph to the same extent. However, the shifted onset of H2 evolution leads to a distinct overlap of the reduction peak.

Details are in the caption following the image

Long-term cycling stability performance of pristine thermally activated GF-TA as well as modified GF-TA-Py and GF-TA-Ph for the negative half-cell reaction. The color bar indicates the number of cycles for the CV experiments. All spectra were recorded with a scan rate of 3 mV/s in a 1 mM urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0084 electrolyte.

For the positive half-cell reaction similar behavior is found for the cycling performance (Figure 9a). Again, GF-TA-Py displays the highest initial activity, although only for the oxidation peak. It appears that throughout all the samples, pristine and modified, there is a transition towards favoring reduction with increased cycle life. This trend can also be observed in Figure 9b, where the ratio Ip,a/Ip,c of all samples continuously declines and leveling of at around 90–100 cycles. Herein, it is also visualized, that the transition is more pronounced for the modified samples, with the pristine GF-TA exhibiting more stable behavior.

Details are in the caption following the image

Long-term stability performance of pristine thermally activated GF-TA as well as modified GF-TA-Py and GF-TA-Ph for the positive half-cell reaction. (a) Cyclic voltammetry curves for the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0085 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0086 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0087 reaction, with the color bar indicating the number of cycles. Evolution of (b) urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0088 and (c) Ip,a/Ip,c with increasing number of cycles. All spectra were recorded with a scan rate of 3 mV/s in a 1 mM urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0089 electrolyte.

Figure 9c illustrates on the basis of urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0090 , that the initial improvement of reversibility happens within the first 20 cycles, after which the values only experience minor variation. Comparison of the samples shows that again GF-TA-Py exhibits improved behavior in the beginning, although all 3 samples approach a common value with increasing cycling.

The results of the long-term aging experiments show that especially for the modification with pyridine initial positive effects on Ip,a/Ip,c and urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0091 can be observed for both half-cell reactions. This enhancement, however, cannot be sustained, which leads to a loss of activity. The introduced functionalities do not appear to provide sufficient long-term stability, as indicated by the performance of the modified graphite felts approaching that of the pristine one with increasing number of cycles.

Flow-Cell Cycling Behavior

To evaluate the efficiencies as well as the stability of the electrodes in real-world VRFB systems, flow-cell tests were performed. GF-TA-Py, GF-TA-Ph as well as GF-TA-Prist were cycled at a current density of 100 mA cm−2 for 15 cycles. Determined Coulombic (CE), voltage (VE) and energy (EE) efficiencies are displayed in Figure 10. The corresponding charge/discharge curves can be found in Figure S6. For GF-TA-Prist and GF-TA-Py Coulombic efficiencies of 99.4 % were achieved for the beginning cycles, with the latter showing less stability with increasing number of cycles (Figure 10a). This behavior can also be observed for GF-TA-Ph, where the CE drops already at the start to 99 %. A lower VE of GF-TA-Prist (67 %) compared to GF-TA-Py (71-72 %) and GF-TA-Ph (75 %) can be ascribed to an increased overpotential, as illustrated in Figure 9 by urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0092 .

Details are in the caption following the image

Flow-cell performance of GF-TA-Prist, GF-TA-Py and GF-TA-Ph. (a) Coulombic, (b) voltage and (c) energy effiency of each cycle for the tested electrodes.

These results obtained by flow-cell testing are in good agreement with the half-cell cycling behavior. As shown, an initial positive effect on performance can be achieved through modification with selected groups, e. g. phenol and pyridine. However, the main issue identified in this work is the long-term stability of said functionalities. Active centers (e. g. carbon defects), however, introduced by thermal activation, do not seem to exhibit this behavior to the same extent.

These results support the assumption that the actual effect of nitrogen and oxygen functional groups on vanadium redox reactions is often overstated and other material parameters, such as defect concentration, seem to have a more determining effect.

A question that remains to be answered is why such a great number of studies report a positive influence of oxygen or nitrogen functional groups on the performance towards the vanadium redox reactions.17, 20-22 We suspect that for the vast majority of these studies, model electrode systems (glassy carbon, plain graphite, non-activated GF, etc.) were used. For this case, we found as well that the introduction of those functionalities has indeed a beneficial effect. However, most studies did not deeply investigate real-world materials, as done in this study. By utilizing actual real-world materials, we could demonstrate that the additional functionalization of thermally activated carbon felts does not further improve the electrochemical performance, especially with regard to long-term stability. For real-world systems, the massively increased electrochemical double-layer capacitance (which is directly connected to the electrochemical surface area) and the increased defect concentration seem to be of far greater importance.23, 24, 55

Conclusions

In this work, a comparative study on the effect of functional groups on the electrocatalytic activity of graphite felt electrodes for vanadium redox flow batteries was conducted. Functionalities comprising nitrogen-, hydroxyl- and alkyl-groups were immobilized on the surface by targeted electrografting, whereas both non-activated and prior thermally activated graphite felt served as substrate material. Thus, the striking discrepancy between model and real-world system could be elucidated.

Morphological and structural investigations with SEM and XPS were successfully applied to validate the modification method. The determination of the defect concentration by Raman spectroscopy revealed the distinct influence of thermal treatment on the graphitic structure by introducing a greater number of defects, compared to the non-activated GF. This is also reflected in a substantially increased electrochemical double-layer capacitance.

We found that, while the immobilization of apolar groups led to a drastic deterioration of the electrode characteristics, introducing hydrophilic groups, whether nitrogen or hydroxyl, resulted in an increase in defects and electrochemical capacitance on the non-activated felt and even improved its performance for both half-cell reactions. However, this enhancing effect could not be translated to the real-world, thermally activated felt. Here the overall electrochemical performance was not significantly affected. On the contrary, the introduction of functional groups deteriorated the long-term stability of the electrodes. We observed that in particular the negative half-cell was influenced to a greater extent, highlighting the necessity of targeted activation with respect to the redox kinetics of the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0093 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0094 urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0095 couple.

Contrary to the popular opinion in the literature, functional groups do not appear to be the only reason for improving the electrochemical performance. Especially for the thermally activated, real-world samples their influence seems to be marginal. The influence of parameters such as number as well as type of graphitic defect concentrations and electrochemical surface area/electrochemical double-layer capacitance seems to be paramount for those systems. Our results reveal that thermal treatment generates a class of active sites that surpass the effect of functional groups and are even impeded by their introduction. Unraveling the exact nature of these active centers has yet to be accomplished in future studies, as it holds the potential to unify the current understanding of graphitic surface processes. Nevertheless, this study underscores the disparity between model and real-world electrode materials.

Furthermore, we could provide an explanation for the fact, that oxygen functionalities are still mentioned in literature as crucial elements for VRFB electrodes. For model systems, the effect of oxygen functionalities can be clearly demonstrated, however, in real-world systems it is overshadowed by other material parameters.

Experimental

All chemicals were used as received without further purification.

Sample Preparation

Pristine non-activated (GF) and previously thermally activated (GF-TA) GFD 4.6 graphite felt (SGL Carbon, SIGRACELL battery felts) served as electrode material. Prior to modification, the sheets were cut and thoroughly washed with Milli-Q water (18.2 MW cm−1).

For modification of the graphite felt surface, 30 mM of the respective molecule, 4-(aminomethyl)pyridine (98 %, Sigma–Aldrich), 4-(aminomethyl)phenol (90 %, Sigma–Aldrich) and 4-butylaniline (97 %, Sigma–Aldrich) was dissolved in 0.5 M HCl (37 %, VWR). Subsequently, 30 mM NaNO2 (98 %, Sigma–Aldrich) was added under vigorous stirring to allow formation of the urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0096 ion. After 1 min, the felt electrode with a modified area of 3×4 cm was immersed in the reaction solution and the diazonium ion was reduced onto the surface using cyclic voltammetry (VSP-300, Biologic). For this, the electrode was cycled 10 times between −0.8 and 0.6 V (vs. Ag/AgCl 3 M NaCl) at 50 mV/s, with a Pt sheet acting as counter electrode (CE). The progress of the reaction was monitored by means of an irreversible reduction peak, which indicated the reduction of the diazonium ion to form the radical intermediate, followed by subsequent bonding onto the surface and the release of N2 gas. After the modification was completed, the felt was washed thoroughly with Milli-Q water and dried under vacuum. An approximation of the amount of each introduced functional group on the electrode surface, as well as the wt %, can be found in Table S2.

Characterization of Modified Electrodes

Raman spectroscopy was used to determine the defect concentration in the graphitic fibers. The spectra were recorded using a Horiba LabRam Aramis, equipped with a 532 nm laser, a 600 gr mm−1 grating and a 100x magnification objective. Measurements were recorded in at least 3 spots. The spectra were corrected in CasaXPS with a spline linear-type background and deconvoluted with Lorentzian and Gaussian-Lorentzian peak shapes, respectively. For visualization, each spectrum was normalized to the highest signal.

X-ray photoelectron spectroscopy (XPS) measurements were performed using a Versa Probe III spectrometer (Physical electronics GMBH) at the ELSA cluster TU Vienna. Monochromated Al Kα (1486.6 eV) was used as the radiation source, with the beam diameter set to 100 μm and the beam voltage to 15 kV. The samples were mounted on a polymer tape and measurements were performed using an E−I neutralization gun to compensate for the charging effect. Survey scans of all samples were recorded at a pass energy of 140 eV and a step size of 0.125 eV. High-resolution core level spectra were recorded at a pass energy of 27 eV and a step size of 0.05 eV.

CasaXPS was used to process the spectra. The C 1s spectra were corrected with a U 3 Tougaard-type background. sp2 hybridized carbon was fitted with an asymmetric peak shape, while the remaining species were deconvoluted with Gaussian-Lorentzian peak shapes. Their position was determined with respect to sp2 carbon, with their FWHM restricted to sp3 carbon. All spectra were binding energy corrected to the sp2 carbon signal at 284.4 eV.

Scanning electron microscopy (SEM) was used to study the morphologies of the modified felts. Measurements were conducted using a Zeiss Sigma EDVP scanning electron microscope, equipped with an Ametek EDAX analyzer for energy dispersive X-ray spectroscopy (EDX) analysis.

Electrochemical Measurements

Cyclic voltammetry measurements were performed with a Squidstat Prime potentiostat (Admiral Instruments) using a custom-built 3-electrode setup. A cut-out piece of the modified graphite felt (1.2×1.2 cm) in contact with a graphitic current collector served as working electrode, with thermally activated felt as counter electrode and Hg/Hg2SO4 (sat. K2SO4) as reference electrode. The felts were thoroughly washed with Milli-Q water before measurements. 0.01 M VOSO4 (99.99 % metal basis, Alfa Aesar) in 2 M H2SO4 was used as electrolyte. A conditioning step for 4 cycles at 20 mV/s was performed in advance.

Half-cell aging experiments were conducted using a commercially available vanadium urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0097 electrolyte (Bolong New Materials) diluted with 2 M H2SO4 to a final concentration of 1 mM V. A conditioning step for 4 cycles at 20 mV/s was performed, followed by 100 cycles at 3 mV/s.

Impedance measurements were performed with a VSP-300 potentiostat (Biologic), with an amplitude of 10 mV and 6 points per decade, in a frequency range of 1–50 MHz. Two spectra were recorded at applied potentials of −0.75 to −1.1 V and 0.2 to 0.5 V, whereas the second spectrum at −1.0 V and 0.4 V was used for the analysis. The impedance spectra were fitted with the software ZView, using the equivalent circuit shown in Figure 6c. It is mentioned that not every element was required for every spectrum.

The electrochemical double layer capacitance was determined in a non-faradaic region between 0 and −0.6 V Hg/Hg2SO4, with 4 cycles recorded at scan rates of 5–200 mV s−1. The 4th cycle was then used to determine the capacitive current at a potential of −0.3 V vs. Hg/Hg2SO4. By plotting the current against the scan rate, the double layer capacitance in F could be determined from the slope of the straight line.

Flow-cell tests were performed in an electrochemical flow-cell (Electrocell Microflowcell) with a commercial anion exchange membrane provided by Fumatech and electrode area of 10 cm2. Continuous charging/discharging for 15 cycles at a current density of 100 mA cm−2 and cutoff voltages of 0.8 V and 1.6 V was performed with a battery tester (Neware, China). Commercially available, urn:x-wiley:18645631:media:cssc202301659:cssc202301659-math-0098 electrolyte (Bolong New Materials) was pumped with a flow rate of 75 rpm to ensure steady-state conditions. The Coulombic efficiency was obtained by dividing the discharge time by the charge time. Voltage efficiency was calculated as the ratio of average discharge voltage to average charge voltage and the energy efficiency as the ratio of average discharge energy to average charge energy.

Acknowledgments

Financial support of “Gesellschaft fuer Forschungsfoerderung NOE” for M. K. and M. M. A. (FTI21 – Dissertationen) and the FFG COMET funding scheme (Competence Centers for Excellent Technologies by BMVIT, BMDW as well as the Province of Lower Austria and Upper Austria) is gratefully acknowledged. The authors acknowledge TU Wien Bibliothek for financial support through its Open Access Funding Programme. We thank Dr. Pavel Mardilovich for helpful discussions. Enerox GmbH is gratefully acknowledged for supporting this work.

    Conflict of interests

    The authors declare no conflict of interest.

    Data Availability Statement

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