Comparison of methods to determine electrocatalysts’ surface area in gas diffusion electrode setups: a case study on Pt/C and PtRu/C

In recent years, gas diffusion electrode (GDE) half-cell setups have attracted increasing attention, bridging the gap between fundamental and applied fuel cell research. They allow quick and reliable evaluation of fuel cell catalyst layers and provide a unique possibility to screen different electrocatalysts at close to real experimental conditions. However, benchmarking electrocatalysts’ intrinsic activity and stability is impossible without knowing their electrochemical active surface area (ECSA). In this work, we compare and contrast three methods for the determination of the ECSA: (a) underpotential deposition of hydrogen (Hupd); (b) CO-stripping; and (c) underpotential deposition of copper (Cuupd) in acidic and alkaline electrolytes, using representative electrocatalysts for fuel cell applications (Pt and PtRu-alloys supported on carbon). We demonstrate that, while all methods can be used in GDE setups, CO-stripping is the most convenient and reliable. Additionally, the application of Cuupd offers the possibility to derive the atomic surface ratio in PtRu-alloy catalysts. By discussing the advantages of each method, we hope to guide future research in accurately determining surface area and, hence, the intrinsic performance of realistic catalyst layers.


Introduction
Electrochemical energy conversion devices such as fuel cells and electrolyzers are essential to support the global transition effort in moving from a fossil-to a renewable energy-based economy [1]. To enable a widespread market penetration of these technologies, however, there is still a need for a significant cost reduction that can be achieved by improving cell performance and minimizing the amount of required noble metal catalyst materials [2]. To reach this goal, massive research efforts have been directed toward the discovery of more advanced electrocatalysts. Over the last several decades, different families of noble metals and noble metals-free electrocatalysts have been identified and tested. Unfortunately, while most of the reported catalysts show improved activity and stability in aqueous model systems such as rotating disk electrodes (RDE), they fail to deliver similar performance in membrane electrode assemblies (MEAs) of real fuel cells [3][4][5].
Recently, gas diffusion electrode (GDE) half-cell setups have been established as a bridging technology to address the observed discrepancy [6][7][8]. GDE half-cell setups have several advantages when compared to both RDE and MEA approaches. In contrast to RDE where reactant gases are dissolved in liquid electrolyte, in GDE half-cell setups similar to MEA systems, a three-phase boundary is established [9]. Therefore, they allow the evaluation of the activity and stability of catalysts and supports in realistic catalyst layers at high mass transport conditions comparable to those in MEA [3,6,10,11]. Moreover, only relatively low amounts of catalysts are required for such measurements, which is essential when testing advanced novel electrocatalysts which are typically synthesized on some milligrams scale [7]. At the same time, the simplicity, rapidity, and comparability of conventional three-electrode setups such as RDE are still maintained with GDE setups [12].
Independent of the method, i.e. RDE, GDE, or MEA, it is inevitable to precisely determine the catalyst's electrochemical active surface area (ECSA) to evaluate and benchmark the intrinsic (specific) activity or stability of new catalyst materials [13,14]. Moreover, the electrochemical determination of the ECSA can be utilized as an in-situ method to track changes in catalyst performance throughout operation. Typical examples of such changes are the removal of surface contaminants during activation procedures or the degradation of catalyst particles during operation (i.e. dissolution, agglomeration, Ostwald ripening, or carbon corrosion) [15,16]. Determining the ECSA in GDE half-cell setups enables activity and stability evaluation of realistic catalyst layers normalized to the corresponding catalyst surface area. Beyond this, in porous electrodes, the ECSA gives insights into catalyst utilization and its dependency on manufacturing parameters (such as the catalyst loading [17,18], the ionomer content [19,20], and the activation protocol [15]) and operating conditions (such as the gas flow [21] and temperature [22]). In general, the ECSA can be determined electrochemically by evaluating the charge that corresponds to a surface-limited electrochemical reaction. Thereby, a suitable electrochemical redox couple has to be selected, depending on the material of interest [23].
In thin-film RDE experiments, mainly three different methods have been developed and applied for Pt-based catalysts on carbon supports, which are the most frequently used electrocatalysts for hydrogen oxidation (HOR) and oxygen reduction reaction (ORR) in fuel cells [13,[24][25][26][27]: (i) The adsorption and desorption of underpotentially deposited hydrogen (H upd ); (ii) The oxidation of adsorbed carbon monoxide (CO-stripping); (iii) The underpotential deposition of copper (Cu upd ).
It was previously shown that all three methods can be utilized to determine the ECSA for pure Pt catalysts [24,28,29]. In contrast, for Pt-alloy catalysts, the H upd method leads to a systematic underestimation of ECSA [28]. Therefore, CO-stripping should be used for ECSA determination of Pt-alloy materials. For PtRu, Cu upd has been demonstrated as a suitable method to determine its ECSA [24]. Additionally, unlike H upd and CO-stripping methods, Cu upd stripping voltammetry provides a unique opportunity to extract information on the surface composition of the alloy due to different binding energies of Cu upd on Pt or Ru surface sites [24,25].
In MEA experiments, the H upd method is commonly utilized due to its experimental simplicity (applying cyclic voltammetry (CV) in an inert atmosphere) [16,18,20,30]. In more advanced MEA test stations, the CO-stripping method can also be applied [22,31]. In both cases, the ECSA is determined in a two-electrode mode by purging an inert gas (N 2 or Ar) or CO through the electrode of interest while using the other (Pt-) electrode under H 2 as counter and reference electrode simultaneously. Analogously to RDE experiments, it was found that H upd and CO-stripping lead to similar ECSA results for Pt catalysts, whereas for Pt-alloys, a significantly higher ECSA is measured via CO-stripping [22]. When comparing MEA results to the ones obtained in RDE setups, lower ECSA values are usually obtained in MEA due to incomplete catalyst utilization in porous electrodes [32]. Cu upd has not been applied in the MEA environment due to a lack of a liquid electrolyte.
Following the MEA research, in GDE half-cell setups, H upd and CO-stripping have been applied for the ECSA determination [7,12]. Recently, we showed that Pt or Fe ions dissolved from catalyst could move through the catalyst layer into the aqueous electrolyte [10,11]. Hence, an opposite diffusion of metal ions from the aqueous electrolyte into the catalyst layer should be feasible. Therefore, the Cu upd method could, in contrast to MEA, in principle be applicable in GDE half-cell setups.
In this work, we systematically investigate the determination of ECSA of Pt-and PtRu-containing catalyst layers in a GDE half-cell setup. For this, the ECSAs in alkaline, as well as acidic electrolyte, are evaluated using H upd and CO-stripping methods. Next, the applicability of Cu upd for the determination of the ECSA of Pt-and PtRu-containing catalyst layers is examined and the obtained results are compared with those from the other methods. Moreover, the method is used to determine surface Pt and Ru coverages for different PtRu/C catalysts. Based on this comparison, the strength, limitations, and finally the applicability of each method for the evaluation of the ECSA in GDE half-cell setups is discussed. With this contribution, we hope to guide future research in the choice of a suitable method for the evaluation of the ECSA of GDEs.

Electrode manufacturing
Catalyst layers are prepared using an ultrasonic spray coater (Biofluidix). The applied ink consists of 1 wt% solids in a water/isopropanol (4:1 wt) mixture. The solid fraction comprises of Nafion D520 (5 wt% solution, DuPont) and high surface area carbon supported catalyst (Pt/C: TEC10E40, Tanaka, 40 wt% metal content; Pt 1 Ru 1 /C: TEC62E58, Tanaka, 50 wt% metal content; Pt 1 Ru 1.5 /C: TEC61E54, Tanaka, 53 wt% metal content; or Pt 1 Ru 2 /C: TEC66E50, Tanaka, 47 wt% metal content). A structural characterization of these commercial catalysts including transmission electron microscopy analysis has been previously reported [33][34][35][36]. The ratio between catalyst powder and Nafion solution is adjusted to achieve an ionomer to carbon (I/C) ratio of 0.7. After isopropanol and Nafion solution are added dropwise to a mixture of catalyst powder and water, the ink is homogenized with an ultrasonic horn (Hielscher) at 60 W and 0 • C for 20 min. Subsequently, the ink is spray-coated onto a commercial gas diffusion layer (H23C8, Freudenberg) on a heated stage at 85 • C aiming for a catalyst loading of 0.1 mg metal cm −2 . The latter is determined by the weight difference before and after coating a specific area (Sartorius Cubis®, 0.001 mg).
Catalyst thin films for RDE analysis are prepared on a glassy carbon tip (AFE5T050GC, Pine®, 0.1963 cm 2 ). The tip is first polished using diamond paste (Mol R3, 3 µm, Struers®) and a polishing pad applying circular movement for 5 min, then washed with ultrapure water. The ink is prepared by mixing 3 ml water/isopropanol (4/1 v/v%), 2 µl of Nafion (5 wt. %, Sigma-Aldrich®), and the corresponding amounts of Pt or Pt 1 Ru 1.5 catalyst to obtain an electrode with 25 µg cm −2 total metal loading by drop-casting (20 µl) onto the polished glassy-carbon tip.

Electrochemical measurements
Electrochemical measurements in the GDE half-cell setup are performed using a VSP-300 potentiostat (Biologic) and a setup as presented in previous work [6]. All values of potential are reported with respect to a reversible hydrogen electrode (RHE). The spray-coated GDEs are utilized as working electrodes with an active geometric surface area of 2.01 cm 2 . An Ag/AgCl electrode serves as a reference electrode (Metrohm) and a platinized Ti-mesh (METAKEM) in alkaline electrolyte or a glassy carbon rod (HTW Hochtemperatur-Werkstoffe GmbH) in acidic electrolyte as a counter electrode. The electrolyte is prepared by dilution of KOH (>86.5 wt%, Emsure, Merck) or H 2 SO 4 (96%, Suprapure, Merck) with ultrapure water to 1 M or 0.1 M, respectively. Gases (Ar, 99.998%, Air Liquide; H 2 99.999%, Air Liquide; CO, 99.997%, Air Liquide) are purged into the electrolyte and the gas chamber with a flow rate individually controlled via mass-flow controllers (EL-Flow Select, Bronckhorst). In between measurements, the GDE-cell is cleaned by boiling in 1 wt% HNO 3 (diluted from 65%, Suprapure, Merck) and subsequently by boiling in water to remove Cu residues.
Prior to an electrochemical measurement, GDEs are pre-treated by floating on top of 1 M KOH for 15 h to ensure sufficient ionomer wetting and ion exchange. For electrochemical cleaning, 40 CVs are performed. Thereby an upper potential limit (UPL) of 1.2 (Pt) or 0.9 (PtRu) V vs. RHE is chosen as a compromise between the oxidation of residues and avoidance of excessive catalyst dissolution [37]. Afterward, the GDEs are polarized to potentials up to 0.7 V vs. RHE under H 2 atmosphere to evaluate their performance toward the HOR in alkaline media. The HOR performance results will be discussed in an upcoming work. Subsequently, the different methods to determine the ECSA are performed in the sequence H upd , CO-stripping in 1 M KOH and, H upd , CO-stripping, and Cu upd in 0.1 M H 2 SO 4 . H upd charges are derived from CV under Ar atmosphere at a scan rate of 100 mV s −1 and a correction by a slanted baseline as discussed later. CO-stripping voltammetry is performed at a scan rate of 20 mV s −1 after a potential hold at 0.1 V vs. RHE while adsorption of CO at a flow rate of 200 ml min −1 for 1-3 min and subsequent cleaning of the gas lines for 30-45 min with Ar [12,38]. The second cycle of the recorded voltammograms is used for background correction.
After performing CV in Cu-free electrolyte, a potential hold at 0.36 (Pt) or 0.33 (PtRu) V vs. RHE is applied for 4 min to reduce catalyst surface oxides. Subsequently, a 0.5 M solution of CuSO 4 prepared from CuSO 4 ·5 H 2 O (99.0%-100.5%, Emsure, Merck) is added to the electrolyte to obtain a Cu 2+ concentration of 2 mM. If not stated otherwise, Cu is deposited with another potential hold at 0.36 (Pt) or 0.33 (PtRu) V vs. RHE for 10 min and oxidized during a linear sweep at a scan rate of 50 mV s −1 . For sequential Cu upd experiments, a potential hold for 120 s at the UPL is included in between experiments to ensure the complete removal of Cu [24].
Potentials derived from cyclic and linear sweep voltammetry are 90% in-situ iR-compensated and 10% iR post-corrected. The uncompensated resistance is determined via impedance spectroscopy [6]. The surface-specific charge Q spec is calculated by integrating the stripping current following Simpson's rule using Python's library SciPy (Version 1.5.2). From this charge and the metal loading m metal the ECSA is derived assuming a conversion factor of f = 210 µC cm −2 (H upd ) or 420 µC cm −2 (CO-stripping, Cu upd ) as well as a complete monolayer coverage of the catalyst nanoparticles (θ = 1), To determine the Pt:Ru surface ratio, the charge contribution of Ru surface atoms is derived by fitting the first peak of the Cu-stripping voltammograms corrected by the CV in Cu-free solution by Gaussian distribution and subsequential integration of the current as suggested by Green and Kucernak [25].
The CO-stripping experiments on RDEs are performed using a Gamry (Reference 600) potentiostat and a PINE MSR Electrode Rotator in a custom-made Teflon cell. A glassy carbon rod (HTW Hochtemperatur-Werkstoffe GmbH) and an Ag/AgCl electrode (Metrohm) are used as counter and reference electrodes, respectively. After adsorption of CO for 15 min at 0.1 V vs. RHE and removal of remaining CO with Ar for 35 min, a CO-stripping experiment is performed at a scan rate of 20 mV s −1 with a UPL of 1.2 V vs. RHE (Pt) or 0.9 V vs. RHE (PtRu). A subsequent CV cycle is used for background correction.

Qualitative analysis of the CV curves
To qualitatively assess the applicability of the chosen methods to determine the ECSA in GDE half-cell setups, the electrochemical profiles of carbon-supported electrocatalysts in different electrolytes were evaluated first. Figure 1 presents the CVs of Pt and Pt 1 Ru 1.5 containing electrodes in an inert environment as well as the stripping voltammetry of a monolayer adsorbed CO or Cu upd . 0.1 M H 2 SO 4 and 1 M KOH are used as acidic and alkaline electrolytes, respectively. To prevent significant Ru dissolution that takes place at higher anodic potentials [37], a UPL of 0.9 V vs. RHE is selected for PtRu electrodes. As significant Pt dissolution starts at ca. 1.1-1.2 V vs. RHE, higher UPLs are selected for Pt electrodes to guarantee complete oxidation of the respective surface species (i.e. adsorbed CO or Cu upd ).
The CVs for Pt and PtRu electrodes in Ar-saturated H 2 SO 4 or KOH electrolytes shown in figures 1(a) and (b), exhibit typical profiles, as already reported in literature [39,40]. Specifically for Pt in alkaline compared to acidic electrolyte, a shift of the H upd region to higher potentials is observed, which has been explained by an increase in Pt-H binding energy [39]. On the other hand, the H upd region of PtRu compared to Pt is shifted towards more negative potentials, which can be attributed to a decrease in metal-H binding energy [40]. This shift in the H upd potential region is usually used as a descriptor to explain the HOR activity of different catalysts. According to this hypothesis: (i) due to strong metal-H binding, Pt has a significantly lower HOR activity in alkaline compared to acidic media, and (ii) due to lower metal-H binding, PtRu is more active towards HOR in alkaline media compared to Pt catalysts [39][40][41].
Figures 1(c) and (d) display changes in CO-stripping voltammetry for both catalysts in the two electrolytes. On Pt surfaces in acidic electrolyte, a single peak (0.73-0.84 V vs. RHE) is observed, whereas, in alkaline media, multiple peaks appear, resulting in a broad voltammetric response (0.35-0.73 V vs. RHE). The lower peak onset potential in 1.0 M KOH indicates a higher activity of Pt towards CO-oxidation in alkaline media [42,43]. In the case of PtRu, the peak potential is further shifted cathodically, illustrating its even higher CO-oxidation activity. As previously reported for PtRu in literature, no significant change in the shape of the oxidation curves for the two electrolytes is observed [40].
Cu upd stripping voltammetry can only be applied in acidic electrolytes [44]. Figure 1(e) exhibits two separate peaks for both, Pt and PtRu. This can be related to surface sites with different Cu upd binding energies [24,45]. Specifically, in the case of PtRu, the two peaks can be related to Cu upd stripping from Ru and Pt surface atoms, as shown by Green and Kucernak [24]. As the deposition of Cu does not occur on Ru oxide sites, they could measure Cu upd stripping voltammograms selectively from Pt atoms in the alloy by irreversibly oxidizing Ru atoms prior to the Cu upd experiment. With this, they were able to assign the peak at lower and higher potential to Cu upd stripping from Ru and Pt sites, respectively [24,25]. Based on this assignment, even the Pt:Ru surface ratio can be calculated as presented later.

Precise determination of the ECSA-specific charges for the different methods
Before comparing the electrocatalysts' surface areas obtained from the different techniques, the methodologies for deducting ECSA from the experimental CV profiles are discussed. For H upd , the lower potential limit (LPL) and the method for correction of capacitive charge contribution need to be considered to derive the surface-specific charge [12]. The LPL of the CVs must not be chosen too cathodic to avoid significant hydrogen evolution reaction (HER) but at the same time not too anodic to prevent incomplete coverage and, therefore an underestimated ECSA. To avoid overestimation due to HER contribution, a high scan rate of 100 mV s −1 is selected (see figure S1). In addition, only desorption charges are evaluated [46]. As previously proposed, the anodic current is corrected for capacitive charge contribution by subtracting a slanted baseline, whereby the slope of the baseline is derived from the CO-stripping experiment assuming that the capacitive charge of the CO-covered surface is the same as that of the H upd -covered surface [12,13]. The same approach is used for PtRu-containing electrodes. However, it has to be stated here, that the choice of a correct baseline is challenging and can lead to significant uncertainties in the resulting ECSA.
The CO-stripping voltammograms are corrected for capacitive current as well as for surface oxidation at high potential by subtraction of the CV after complete oxidation of adsorbed CO [12,13]. On Pt surfaces in acidic electrolyte, the CO-stripping peak up to 1.0 V vs. RHE is integrated [13]. In alkaline electrolyte, an overlap of the H upd and CO-stripping region can be observed, leading to an underestimation of the surface-specific current (see figures 1(d) and S2, red solid and dashed lines). Fortunately, this region is clearly distinguishable from its CO-stripping peak in acidic electrolyte (see figure 1(c), red solid line). Therefore, up to 0.5 V vs. RHE, the CO-stripping voltammogram of Pt in alkaline electrolyte is corrected for capacitive current contribution with the corresponding CO-stripping voltammogram in acidic electrolyte (see figure  S2, red dash-dotted line) as previously reported by Rheinländer et al [47].
Similar to CO-stripping, contributions of capacitive current and surface oxidation have to be taken into account to derive the surface-specific Cu upd stripping charge. This goal is achieved by subtracting a CV recorded beforehand in a Cu-free solution. Before Cu upd can be utilized to determine the ECSA, experimental parameters of the method (such as deposition time and potential, gas flow rate, and scan rate) have to be optimized for both catalyst systems (PtRu, see figure 2 and Pt, see figures S3-S5).
To determine the optimal deposition potential, at which a complete monolayer coverage is established while no bulk deposition occurs, the deposition potential is varied between 0.29 and 0.37 V vs. RHE. As depicted in figure 2(a), the deposition of Cu on PtRu below 0.33 V vs. RHE leads to an additional peak in the stripping voltammogram, which can be assigned to the bulk deposition of Cu. On the other hand, when choosing a higher deposition potential, no complete monolayer of Cu upd can be established. Hence, the optimal deposition potential was found to be 0.33 V vs. RHE for PtRu and 0.36 V vs. RHE for Pt electrodes (following the same logic, see figure S3). The slightly higher deposition potential for Pt compared to PtRu can be explained by the shift in peak potential as illustrated in figure 1(e), corresponding to the lower binding energy of Cu upd on Ru surface sites [25]. Moreover, this cathodic shift at PtRu electrodes leads to a less clear separation between underpotential and bulk deposition (see figures 2(a) and S3). Thus, the deposition potential for Cu upd experiments must be chosen very carefully.
By variation of the deposition time, it is observed that a minimal potential hold for 600 s is needed to ensure complete coverage of Cu upd (see figures 2(b) and S4). This time is significantly longer than the 100-120 s determined in experiments on thin film electrodes [24,48,49]. When the charge that needs to be transported per geometric area in the different systems utilized in literature is compared, no clear trend towards the necessary deposition time can be observed (see table S1). This finding suggests that the deposition of Cu itself is not the major limiting factor. When comparing unsupported catalyst nanoparticles used by Green and Kucernak to the high-surface-area carbon-supported catalysts in a realistic catalyst layer with higher ionomer content used in this work, the increase in deposition time can be explained by the hindered Cu 2+ mass transport through the thicker catalyst layer, as was assumed for Pt ions [10,24]. Thus, applying Cu upd to new systems (and catalyst layers) will require an optimization of the deposition time. For this, also the current response during the deposition is a valuable indicator when monolayer coverage is completed (see figure S6).
Fenwick et al recently observed a decrease in Cu upd stripping charge on nanoporous gold GDEs with an increasing flow rate of inert gas from the back [21]. They correlate this to a partial drying out of the catalyst layer close to the gas compartment. In contrast, here, no change in the Cu upd stripping voltammogram with flow rate is observed (see figure 2(c)). This indicates a sufficient wetting of the utilized catalyst layers. To test this assumption, different methods have been developed to derive the wetting state of the catalyst layer of GDEs in-situ [50]. In contrast to the Cu upd stripping voltammogram, the CV in Cu-free solution exhibits an increase in additional ORR current below 0.7 V vs. RHE when decreasing the flow rate of inert gas (see figure 2(c)). This can be attributed to the undesired leakage of ambient air to the GDE setup at low gas flow rates. In the Cu upd stripping voltammogram, this undesired additional ORR current is absent. This indicates a complete coverage of the catalyst surface with Cu and thus, a deactivation towards ORR [51,52]. Still, undesired ORR contributions need to be minimized to avoid an underestimation of the background charge and thus an overestimation of the resulting ECSA. Therefore, in the current work, a gas flow rate of 300 ml min −1 was chosen for Cu upd experiments.
The influence of traces of undesired ambient O 2 can be further reduced by increasing the scan rate of the stripping sweep. As the mass-transport-limited ORR current is independent of the scan rate, its relative contribution to the absolute current decreases with increasing scan rate (see the background as dashed lines in figure S5). However, an anodic shift of the Cu upd stripping peak potential with increasing scan rate is observed due to the rather slow kinetics of the stripping of Cu upd (see figures 2(d) and S5) [53]. Hence, to ensure a complete stripping of Cu upd below 0.9 V vs. RHE, a medium scan rate of 50 mV s −1 is chosen for the ECSA analysis by Cu upd [37]. Figure 3 summarizes the obtained ECSAs from the three different methods for Pt and PtRu catalysts in alkaline and acidic electrolyte. Comparing the ECSA results calculated by H upd , two trends can be observed: (i) a decreased ECSA Hupd in acidic compared to alkaline environment for all catalysts; (ii) a decreased ECSA Hupd for the PtRu alloy compared to Pt-containing electrodes in both electrolytes. These trends can be explained by the architecture of the GDEs and the gas flowing from the back of the electrode leading to an improved mass transport of H 2 gas. As known from MEA research, the improved mass transport leads to an anodic shift of the HER onset potential compared to RDE setups following the Nernst equation [30,54]. This shift partially masks the H upd adsorption/desorption current, which leads to an incomplete coverage with  H upd at the HER onset potential, and thus, erroneously decreases the calculated ECSA [17,30,47,54]. The degree of this effect depends on the H upd binding energy. Following the discussion in section 3.1, the underestimation of the ECSA is more pronounced in acidic than in alkaline electrolyte and for PtRu compared to Pt. Thus, due to the high binding energy of H upd on Pt in alkaline electrolytes, the effect of the HER onset is minimal, leading to ECSA values comparable to CO-stripping and Cu upd . However, due to the comparably broad potential window of the H upd desorption, the correction of the capacitive current can lead to larger uncertainties. In the case of Pt-alloy catalysts, also in conventional three-electrode setups, a reduced H upd charge can be observed [55]. This observation points towards an incomplete surface coverage at the HER onset potential due to lower H upd binding energy compared to bare Pt surfaces. The same trend has been previously reported for Pt-and PtCo-catalyst layers in acidic electrolyte [12]. Thus, the applicability of the H upd method for Pt-alloy catalysts has to be questioned generally [12].

Comparison of methods to determine the ECSA
In contrast to H upd , CO-stripping shows comparable results in both electrolytes and the two catalysts. These results are in line with experiments on polycrystalline Pt electrodes showing the independence of the CO-stripping charge on electrolyte [47]. Also, the results from Cu upd indicate the same ECSA for Pt and PtRu catalysts. In contrast, compared to CO-stripping in the same electrolyte, the values are 9% and 14% higher for Pt and Pt 1 Ru 1.5 , respectively. This trend is in good agreement with findings from Green and Kucernak, who reported increased ECSA values from Cu upd (+6% Pt, +7% for Pt 1 Ru 1.2 ) compared to those from CO-stripping. Larger deviations here can be explained by the contribution of ORR to the CV in Cu-free media leading to an underestimation of background charge and thus an overestimation of the ECSA as discussed before. Additionally, despite the higher deposition potential chosen in this work, the longer required time to deposit Cu could lead to more pronounced bulk deposition and thus overestimation of the Cu upd stripping charge. Overall, it can be concluded that the results of both techniques can be considered adequate to determine the ECSA of Pt and PtRu catalyst layers. This statement holds for at least the three investigated compositions of PtRu-alloy catalysts (Pt 1 Ru 1 , Pt 1 Ru 1.5 , and Pt 1 Ru 2 ).
To provide additional support for the ECSA determination in GDE half-cell setups, CO-stripping experiments are performed on thin film RDEs using the same Pt and Pt 1 Ru 1.5 catalyst in acidic media. The results for the ECSA of RDEs are in acceptable agreement with values reported in literature for the same catalyst (see table S2) [34,56]. Comparing results from GDE and RDE experiments, a reduction in ECSA for realistic catalyst layers compared to catalyst thin films is observed (see table 1). This reduction has been previously observed for GDE half-cells as well as for MEA systems and is attributed to an incomplete utilization of the catalyst within a thicker catalyst layer due to a lack of contact with solid electrolyte for part of the catalyst particles [32,57]. Specifically, a catalyst utilization in the GDE of 77% and 85% for Pt and Pt 1 Ru 1.5 catalysts, respectively, is derived. These values for the catalyst utilization are in good agreement with findings from Nösberger et al in a comparative RDE-GDE study [57]. Following this, comparing the catalyst utilization in different GDEs enables the optimization of the manufacturing process of realistic catalyst layers toward complete utilization.

Determination of Ru content in PtRu alloys
Besides the determination of the ECSA, Cu upd stripping voltammetry provides unique information on the surface composition of PtRu alloys due to the lower binding energy of Cu upd on Ru compared to Pt [24,25]. By fitting the peak at lower potential and relating the corresponding charge to the total Cu upd stripping charge, the Ru content at the surface of the PtRu-alloy can be calculated (see figure 4(a)). However, the exact determination of the Ru content by Cu upd is limited by the arbitrarily chosen Gaussian function to fit the peak at lower potentials. Analysis of Cu upd stripping voltammograms at varied scan rates demonstrates the reliability of the method. Less than 2% deviation in the derived Ru-content is found despite the shift in potential and change in shape (see figure S7).
To underline the applicability of this method, Cu upd stripping voltammetry on different PtRu alloys is performed and the derived Ru content is compared to the one obtained by energy-dispersive x-ray spectroscopy (EDX) and x-ray photoelectron spectroscopy (XPS) as well as the nominal bulk composition (see figure 4(b)) [33]. Thereby, the results from electrochemical and physical characterization follow the same trend as the nominal bulk composition for the different alloy compositions. The absolute value derived from Cu upd are well in line with the ones obtained by XPS, both showing a ∼25% lower Ru content compared to the nominal bulk composition, which was previously verified via EDX [33]. As Cu upd and XPS are surface-sensitive techniques, the results demonstrate the discrepancy between surface and bulk composition. This so-called surface segregation is well-known for PtRu-alloys and an accumulation of Pt atoms at the surface has been reported previously [58]. Thus, the Cu upd method offers the unique possibility to derive the Ru surface content electrochemically without being dependent on physicochemical ex-situ techniques.

Which method should be chosen for ECSA determination?
Choosing a method to determine the ECSA of the material of interest is a delicate task. Besides fundamental considerations, also experimental requirements must be taken into account. Table 2 summarizes the advantages and limitations of the three examined techniques. In general, when using the obtained ECSA to normalize the activity or stability of electrocatalysts, it is important to bear in mind that the determined ECSA does not necessarily coincide with the number of sites active during the reaction of interest. Discrepancies can occur due to different chemical environments or possible blocking of surface sites during the reaction. In addition, it should be noted that underestimation of the ECSA can erroneously lead to an overestimation of the specific activity or stability of the considered material [59].
Considering the experimental requirements, H upd is the easiest available method. It can be concluded directly from CVs in an inert atmosphere, whereas for CO-stripping, additional safety measurements to handle CO gas as well as extra time to remove all traces of CO from the gas lines by purging inert gas, are required. The latter can add up to 1 h per experiment with the current setup and procedure, which is comparable to MEA setups [62]. The time required to remove residual CO can be reduced by decreasing gas line volume and by using diluted CO [22,61]. A short time is preferred as traces of O 2 can oxidize surface-bound CO and, thus, decrease the CO-stripping charge [22]. In the case of Cu upd , only a CuSO 4 solution must be added to the electrolyte, which is beneficial in terms of safety and time consumption. However, the removal of Cu 2+ ions is crucial to avoid influences on activity measurements by residual impurities [51,52]. Therefore, an additional cell-cleaning step in diluted HNO 3 in between different samples has to be added to remove all traces of Cu 2+ ions. Thus, without a proper cleaning protocol, the repetitive determination of ECSA for one sample is not possible. Therefore, changes in ECSA due to certain effects such  [34]. Recently, Chattot et al investigated the effect of potential cycling to low potential after the application of an accelerated stress test protocol. Their results demonstrate that the observed agglomeration of catalyst particles is induced by cycling to low potential rather than a consequence of the performed stress protocol [63]. Therefore, special care has to be taken, when ECSA methods are performed after or in between stress tests. When comparing the validity of the determined ECSA values for each technique, different phenomena become apparent. As discussed above, ECSA derived from H upd is likely to be underestimated due to insufficient coverage at the LPL because of the anodic shift in HER onset potential and low H upd binding energy for Pt-alloy catalysts. Additionally, small errors in the capacitive current correction can significantly vary the result. Therefore, the systematic error of this method to determine a catalyst's ECSA is large. Contrastingly, for CO-covered surfaces, a disappearance of the H upd indicates a close to complete surface coverage (>80%) [38]. In addition, if no CO oxidation features can be observed in the CV following the CO-stripping voltammetry, it can be validated that adsorbed CO as well as any excess CO gas, have sufficiently been removed. When these control measures are considered, reliable ECSA determination by CO-stripping is possible in alkaline and acidic media. However, for measurements at elevated temperatures, the decrease in CO surface coverage should be borne in mind [22,61]. When using Cu upd , it is crucial to optimize the deposition time and potential for each catalyst system to achieve a complete monolayer coverage and avoid bulk deposition. Thus, time and potential dependent experiments are required for each new system to identify the best compromise as discussed in section 3.2. Additionally, the applicability of the Cu upd method in other electrolytes than H 2 SO 4 remains uncertain.
In conclusion (see figure 5), we recommend using CO-stripping to determine the ECSA of Pt and Pt-alloy catalyst systems in alkaline or acidic media. Cu upd is a viable alternative when experimental parameters are optimized for each new catalyst system. It should be preferred if the Pt:Ru surface composition should be assessed additionally. Quantitative ECSA determination via H upd should only be applied for pure Pt catalysts and with cautious optimization of experimental and integration parameters. However, it remains the most facile method to follow qualitative ECSA trends in-situ to assess catalyst activation or degradation.

Conclusion and outlook
The ECSAs of Pt and PtRu containing catalyst layers in KOH and H 2 SO 4 electrolytes are determined using H upd , CO-stripping, and Cu upd methods in a GDE half-cell setup. The obtained results suggest that both CO-stripping and Cu upd are adequate methods to determine the ECSA. H upd , however, underestimates the ECSA for Pt in acidic and PtRu in both electrolytes. This underestimation can be attributed to an increased overlap with HER because of the increased mobility of H 2 in GDEs in general and the lower H upd binding energy in PtRu in specific. For future research, we suggest the use of CO-stripping as the standard method to determine the ECSA considering (i) its potential window excluding an overlap with other electrochemical reactions, (ii) the possibility of using the method multiple times on the same sample in between activity protocols and (iii) its broad application in literature. When evaluating PtRu material, Cu upd represents a viable alternative especially if the Pt:Ru surface ratio can give further insights into the question of research. With this contribution, future research on the intrinsic activity and stability of Pt and PtRu catalysts is guided and its possible dependence on PtRu-alloy composition as well as parameters of manufacturing and operation can be explored. Further research is necessary to validate or develop ECSA methods in GDE half-cell setups for other catalyst materials i.e. different Pt-based alloys or different noble and non-noble metals with potential use in electrochemical devices.

Data availability statement
The data cannot be made publicly available upon publication because no suitable repository exists for hosting data in this field of study. The data that support the findings of this study are available upon reasonable request from the authors.

Acknowledgments
N C R acknowledges Pascal Lauf and Valentín Briega Martos for scientific discussions. S C acknowledges funding by the German Federal Ministry for Economic Affairs and Energy (BMWi) within the project 03ETB027A. K E acknowledges Heinrich Böll Foundation for financial support.

Authorship contributions
N C R, Y-P K, KE S C. contributed to the conceptualization of this work. N C R performed GDE experiments, data analysis, and wrote the original draft of the manuscript. M M performed RDE experiments. Y-P K, K E, S C provided critical feedback to the interpretation, reviewed, and edited the manuscript.