Introduction

Proton exchange membrane fuel cells (PEMFCs) are energy converters which can be used for automotive, stationary, or nomad applications. These systems produce electrical energy and heat thanks to the electrochemical conversion of hydrogen and oxygen into water. To ensure the viability of this technology, some drawbacks, such as their insufficient durability or high production costs, need to be overcome [1]. Those are mainly linked to the core materials of the PEMFCs. For instance, their insufficient durability is strongly related to the degradation of the electrocatalyst support, especially at the cathode side [2]. These catalytic supports are mainly based on carbon black materials, which are thermodynamically unstable under the usual working conditions of PEMFC cathodes (T = 80 °C in humid and acidic environment, cell voltage of 0.6 to 1.0 V) [3,4,5] and even more in harsh conditions experienced during start-stop procedures or fuel starvation events [6,7,8,9,10,11]. Theoretically, the carbon corrosion begins with the formation of oxygen-containing groups (carboxyl and carbonyl groups, quinone, phenol, and lactone [4, 5, 12]) in the amorphous part of the carbon materials, at the edge of the carbonaceous lattice or in the defects of the carbonaceous crystallographic planes. These groups are then irreversibly transformed into carbon dioxide. Experimentally, the carbon corrosion is observed at E = 0.9 V vs. the reversible hydrogen electrode (RHE) [13, 14]. The platinum nanoparticles—the most popular electrocatalyst in PEMFCs—catalyze carbon corrosion [15,16,17], the later hence starting at potentials as low as E = 0.6 V vs. RHE [4, 13,14,15]. Evidently, several parameters like the working potential [13], the relative humidity [13], the temperature [13, 15, 18,19,20,21], and the nature of the gas atmosphere [22] play a role on carbon corrosion and therefore on the alteration of the physicochemical properties of the active layer. Indeed, the formation of oxygen-containing groups onto the carbon surface (i) increases the hydrophilicity of the active layer [4, 5, 12], leading to an increase of the diffusion overpotential [23], and (ii) decreases the electrical conductivity of the catalytic support [24, 25]. Carbon corrosion also modifies the carbon morphology and is prone to trigger the destabilization of the catalyst nanoparticles and the decrease of the electrochemical active surface area (ECSA). These modifications are responsible for a drastic loss of performances of the whole PEMFC system [26]. Several methods are studied to improve the durability of the catalyst support. For instance, other carbon materials were used to substitute carbon blacks, e.g., nanotubes [27, 28], boron-doped diamonds [29], or doped-graphene [30, 31]. These carbon supports exhibited a better resistance towards corrosion phenomena as a result of their graphitic nature (i.e. graphitic carbons are more stable upon corrosion at high potential values [4, 5, 16, 17]). However, some chemical treatments (under HNO3, H2SO4, etc.) were necessary, especially on carbon nanotubes [32,33,34], to increase the surface hydrophilicity and achieve an acceptable deposition of active material nanoparticles on the carbon surface, as organized carbon areas provide fewer anchoring sites for the latter [35, 36]. Heat treatments were also performed on carbon blacks to improve their degree of the organization [4] (graphitized carbons [5, 12, 37]). Unfortunately, such heat treatments modified the morphology of the carbon blacks by reducing their specific surface area [36], leading to poor quality of the catalyst nanoparticle deposition.

As a matter of fact, none of the methodologies described above enabled to fully stop the carbon corrosion; at best, a high degree of organization slows down the kinetics of the carbon oxidation reaction (COR), at least when potential values above 1 V vs. RHE are experienced [4, 5]. Hence, other studies aimed at modifying the chemical surface of the carbon supports in the hope to find means to further impede the COR. For example, Berthon-Fabry et al. and Asset et al. used fluorination to reduce the kinetics of the corrosion [38, 39]. Because of its high reactivity, elemental fluorine is a very strong oxidizing agent. Its huge reactivity allows its bonding with almost any element, both metals and non-metals. In this case, dangling groups of the structurally disordered carbon materials will preferentially combine with fluorine; this strategy leads to strong covalent C–F bonds, which have been demonstrated less prone to oxidation than C–O bonds, even though again, the COR is not totally suppressed [39]. Others studied carbon phase-free catalyst supports such as conducting polymers [40, 41] and carbides [42,43,44], or even carbon-free supports like borides [45, 46], metal nitrides [47,48,49], bi-metal aerogels PtMx (with M = Ni, Cu, or Co) [50,51,52,53,54], or metal oxides [55,56,57,58,59,60,61,62,63,64,65,66] have been investigated as catalyst support. Metal oxides have been particularly investigated because of their theoretical stability [3]. For instance, Takabatake et al. studied the initial catalytic activity and the durability of six electrocatalysts Pt/MyOx (M = Mo, Nb, Ta, W, Ti, and Sn) for the oxygen reduction reaction (ORR) by rotating disk electrode analysis (RDE) in a 0.1 M HClO4 electrolyte at 25 °C [67]. It turned out that the electrocatalyst deposited on tin oxide, Pt/SnO2, displays the most encouraging results. Nevertheless, one of the most significant drawbacks of metal oxides for the catalyst support applications is their semi-conductor nature, incompatible with the high electronic conductivity required for electrode support materials [68]. Therefore, other studies focused on doping metal oxides to improve the overall electrical conductivity of the material: many electrocatalysts have been prepared with doped tin oxide (with niobium, antimony, vanadium, and tantalum elements), to probe their initial electrochemical performances and their durability [61, 69,70,71]. These performances have not always been proved satisfactory because the dopant is gradually depleted from the material’s surface leading to pollution and electronic conductivity issues [72, 73]. For this reason, the protection of the carbon surface with a thin (a few nanometers) metal oxides coating, MyOx/C, has also been evaluated [74,75,76,77,78,79,80,81,82,83,84,85,86] in the hope to create a more resistant composite. This methodology conveys multiple advantages: the MyOx/C composites can keep the morphology and the good electrical properties of carbon materials (provided the metal oxide film is indeed thin), thanks to the electrical percolation, and also provide better resistance to the COR owing to the intrinsic thermodynamic stability of the metal oxide. Furthermore, some metal oxides (TiO2, SnO2) are expected to stabilize the Pt nanoparticles thanks to the strong metal support interaction (SMSI) effect [87,88,89]. Encouraging results were generally found for tin oxide–modified carbon supports, SnO2/C [80, 84,85,86]. These good electrochemical properties and durability were explained by (i) the stability of the tin dioxide coating and (ii) the modification of the electronic structure of the Pt nanoparticles induced by tin dioxide [86]. Other examples of materials investigated to be used as catalyst support for low-temperature fuel cells are provided in the following reviews [40, 90,91,92,93,94].

In a previous study, the authors have synthesized thin tin oxide coatings on the surface of a wide set of carbon materials (carbon nanotubes, with or without pre-treatment, two different carbon blacks and a home-made carbon aerogel) displaying a large panel of textural, chemical, or structural properties [95]. They studied the influence of different experimental parameters (the pH value of the reactive medium, the amount of tin oxide precursor, and the intrinsic carbon material properties) on the quantity and the quality (homogeneity and covering aspect) of SnO2 coatings. The mechanisms of tin oxide formation, depending on the pH of the reactive medium, were also discussed. Some of these samples based on carbon aerogel keep an interesting morphology, which opens the way to their use as catalytic supports for the PEMFC application.

In the present work, the authors pursued the effort and synthesized Pt-based electrocatalysts with tin oxide–coated carbon aerogel materials. To this end, Pt nanoparticles of similar individual crystallite sizes were deposited on these new support materials using a modified polyol method [96]. The electrocatalysts were characterized by physicochemical (transmission electron microscopy, X-ray diffraction) and electrochemical (in rotating disk electrode configuration) techniques. Their initial performances towards the oxygen reduction reaction (ORR) and also their durability were compared to those of a benchmark Pt/C electrocatalyst synthesized with the raw carbon aerogel. For this purpose, two accelerated stress tests (ASTs), mimicking either a base-load procedure of PEMFC or the start-stop procedure of PEMFC, were used. In this process, the influence of the quality and quantity of tin oxide coatings on the initial performances and the durability of the electrocatalysts were evaluated and discussed. Finally, the electrochemical performances of the herein-synthesized samples were compared to those obtained using commercial carbon powders as a substrate and a similar polyol synthesis of Pt nanoparticles, already tested in [39].

Experimental

Syntheses

Synthesis of the Raw Carbon Aerogel

The carbon aerogel was synthesized following the Pekala method [97], as described in earlier contributions [38, 98]. Briefly, resorcinol (R: C6H6O2, 99%, Alfa Aesar) and formaldehyde (F: CH2O, 37 wt% in aqueous solution stabilized with 10–15 wt% methanol, Acros Organics) were mixed with the sodium carbonate catalyst (C: Na2CO3, 99.9999% metal basis, Sigma-Aldrich) in deionized water with molar ratios F/R = 2, R/C = 300, and 5 wt% solid in solution. The mixture was heated at T = 85 °C for 1 week for gelation and mechanical handling. After several exchanges of the water by acetone, the gel was dried in supercritical CO2 conditions (80 bars, 37 °C), and the resulting organic aerogel was pyrolyzed at T = 1050 °C under nitrogen flow (1.5 L min−1) into the carbon aerogel sample (CA). After pyrolysis, the carbon aerogel was ground in a mortar for 15 min to obtain a fine powder. All the samples (raw and composites) originate from the same batch.

Synthesis of the Tin Oxide Coating

The objective of this synthesis is to obtain, for comparison purposes, different qualities of tin oxide coatings on the CA surface. Protocols using two different pH values (acidic or basic values) of the reactive medium and quantities of the tin oxide precursor have been used. For the acidic one, the authors followed a protocol proposed initially by Han and Zettl [99]. Firstly, 100 mg of carbon aerogel was mixed in deionized water (400 mL) with a sonotrode (3 × 5 min, 50 W) to improve the dispersion of the carbon materials and break carbon agglomerates. Hydrochloric acid (HCl, 37% reagent grade, Sigma-Aldrich) (7 mL) was added in the solution during the sonication. After that, a predefined quantity of tin chloride (SnCl2.2H2O, 98%, Alfa Aesar), used as tin oxide precursor was added to the solution, which was mixed at room temperature during 1 h, and then filtered. The filtrate solution was then washed with deionized water until pH ≥ 5. Finally, the obtained powder was dried at T = 90 °C for 6 h under confined atmosphere. For the basic one, the authors used their new protocol, where 5 M sodium hydroxide solution (NaOH, flake, 98%, Alfa Aesar) is progressively added after the sonication, but also after the addition of SnCl2.2H2O to maintain a stable basic pH value (pH = 11.5). In this case, no hydrochloric acid HCl was used. More details of these syntheses are given in ref. [95]. The obtained samples are denoted as follows: SnOx(pH value-quantity of SnCl2.2H2O)/CA.

Synthesis of Platinum-Based Electrocatalysts

The Pt nanoparticle deposition onto the catalytic support (raw carbon aerogel (CA), or tin oxide–coated carbon aerogel (SnO2/CA)) used a modified polyol method [96]. Firstly, a colloidal suspension was obtained by reducing a given amount of Pt salt (H2PtCl6-xH2O, 99.9% metal basis, Alfa Aesar) dissolved in a 1:1 volume ethylene glycol (Rotipuran > 99.9%, Roth):ultrapure water solution. The solution was basified (pH = 12.5) and mixed at T = 160 °C at least during 3 h under argon atmosphere. Meanwhile, the chosen catalytic support was mixed in a 1:1 volume ethylene glycol: ultrapure water solution, the quantity of catalytic support being calculated to obtain a final Pt content of 40 wt%. Finally, the two solutions were mixed and the pH adjusted at 3.0 to allow the Pt nanoparticle deposition. The final solution was mixed at least during 12 h, filtered, and washed with ultrapure water. The obtained powder was dried at T = 110 °C for 1 h. The two Pt/C benchmarks deposited on commercial carbons using the same polyol synthesis for Pt nanoparticle deposition were synthesized and thoroughly characterized in [39]. Briefly, one of the materials was prepared using a high surface area, mildly graphitized, carbon black (ca. 800 m2 g−1); the corresponding catalyst is labeled Pt/EN; it is loaded at 33 wt% Pt. The other was prepared using a lower surface area strongly graphitized carbon black (ca. 130 m2 g−1); the corresponding electrocatalyst, labeled Pt/YS, is loaded at 31 wt% Pt.

Material Characterizations

Physicochemical Characterizations

The morphology of different catalyst supports was investigated by scanning electron microscopy (SEM) imaging, with a Zeiss® Supra 40 apparatus equipped with a Gemini column operated at 3.00 kV.

Nitrogen sorption analysis was used at 77 K with an ASAP 2020 apparatus (MICROMERITICS). Samples (CA or SnO2/CA) were first degassed under secondary vacuum at 90 °C for 12 h. The specific surface areas were determined following the Brunauer-Emmett-Teller (BET) model. The Barret-Joyner-Halenda (BJH) method applying to the adsorption branch of the isotherms was used to calculate the pore size distributions [100]. The density functional theory (DFT) model was used to assess the micro-porosity of the different samples. The values are reported by mass of composite.

Transmission electron microscopy (TEM) imaging was used to observe the deposition of the Pt nanoparticles on the different support materials and to evaluate the nanoparticle size dispersion. The apparatus used is a JEOL 2010 microscope with a LaB6 filament, operated at 200 kV, equipped with a retractable SDD Centurio detector. Samples were observed in bright field mode with a point-to-point resolution of 0.19 nm. The obtained micrographs were used to determine the average diameter dTEM and the volume-weighted average diameter dv (Eq. (1)) of the Pt particles deposited on the different support.

$$ {d}_{\mathrm{v}}=\frac{\sum {n}_i{d}_i^4}{\sum {n}_i{d}_i^3} $$
(1)

where ni is the number of particles of diameter di.

Only isolated and well-defined particles were measured, and a minimum of 100 Pt nanoparticles were counted for each sample.

X-ray diffraction (XRD) measurements were performed using a PANalytical X’Pert pro Philips diffractometer operating at 45 kV and 40 mA using Cu Kα radiation (λ = 0.15418 nm) and a diffracted beam monochromator. The Pt crystallite size was determined from the (111) diffraction peak at 2θ about 39.8° using the Debye-Scherrer equation. As XRD is sensitive to the volume of particles, the crystallite size corresponds to an average volume diameter and must be compared to the volume-weighted average diameter dv.

To evaluate the quantity of the tin oxide coating in the composites, thermogravimetric analysis (TGA) experiments were performed on a TGA-50 apparatus (Shimadzu) with an accuracy of 0.1 mg. The measurements were carried out in the air with a heating rate of 2 °C min−1 from room temperature to 600 °C with a 30 min step at T = 600 °C.

Raman spectroscopy was used to probe the carbon structure of the initial bare CA and composites and their aged counterparts after the two ASTs. The spectra were recorded ex situ using a Jobin Yvon T64000 spectrometer with a wavelength LASER excitation of λ = 514 nm. Each sample was analyzed in five different areas. For the sake of comparison, the Raman spectra were normalized to the intensity of the graphitic lattice band (ν ≈ 1585 cm−1). For the carbon materials, five main bands could be detected [4], so spectra were recorded in intervals including these Raman shifts. The obtained curves are used to determine the lateral extension of the graphite planes (hereafter denoted by La) following the Knight and White formula [101] (Eq. (2)):

$$ {L}_{\mathrm{a}}=4.4\times \left({A}_{\mathrm{Gband}}/{A}_{\mathrm{D}1\mathrm{band}}\right) $$
(2)

where AGband and AD1band are, respectively, the area of the G and D1 band.

It turns out that the G and D2 bands are close (ν ≈ 1585 cm−1 and ν ≈ 1610 cm−1, respectively), so curve fitting, using the LabSpec software, was performed to calculate the precise area of the G band. Even if AGband/AD1band is not proportional to the La value for the amorphous carbon materials (it is rather proportional to La2 but no quantitative description of this relationship is provided in the literature) [102, 103], it remains possible to compare the La value of the fresh and the aged CAs to evaluate any structural evolution after the two ASTs.

The electrical conductivity of the catalyst supports was investigated by a direct resistance measurement at room temperature with a home-made cell. Briefly, a quantity of 50 mg of sample was introduced between two copper electrodes (Sgeo = 0.7854 cm2) surrounded by a Teflon ring. Different currents were applied (I = − 105 mA, 105 mA, and 400 mA) by a potentiostat (Biologic HPC-803) at a pressure of 6.37.107 Pa. The voltage value was recorded for each current, and the electrical conductivity (σ) was calculated as an average of the three values obtained with the three currents following Eq. (3):

$$ \sigma =e.I/U.{S}_{\mathrm{geo}} $$
(3)

where e is the thickness of the sample (cm), I the current applied (A), U the voltage recorded (V), and Sgeo the surface of the electrode (cm2).

The Pt content in the electrocatalyst was determined by atomic absorption spectroscopy (AAS) (PinAACle 900F, PerkinElmer). To that goal, each electrocatalyst (m = 5 ± 1 mg) was dissolved in 1 mL of a HCl:HNO3 solution (3:1 volume) at least overnight at T = 60 °C. The solution was then diluted with ultrapure water to reach the AAS range for Pt. The Pt content was determined with Syngistixen software, basing on calibration curves obtained by Pt standard solutions. The wavelength considered for Pt was λ = 266 nm.

Electrochemical Characterizations

Prior to any electrochemical measurement, the glassware was soaked with the concentrated Caro’s acid (96% H2SO4-30% H2O2 mixture) at least overnight, thoroughly rinsed with ultrapure water.

The electrochemical measurements were performed in a four-electrode electrochemical cell using a rotating disk electrode setup. The temperature was controlled using a double envelope. The electrolyte used (0.1 M HClO4) was prepared with ultrapure water and HClO4 (70% Merck Suprapur). The electrochemical characterizations and the ASTs were carried out in two different cells. The working electrode was a rotating disk electrode (RDE) made of glassy carbon (Sgeo = 0.196 cm2), the counter electrode was a glassy carbon plate, and the reference electrode was a reversible hydrogen electrode (RHE) connected to the cell via a Luggin capillary to limit the Ohmic drop. A Pt wire was connected to the reference electrode with a capacitive bridge to filter the high-frequency electrical noise [104].

The catalytic ink, containing the studied electrocatalysts, was composed of Nafion® (5 wt% perfluorinated-sulfonic acid solution with 15–20 wt% water, Aldrich), ultrapure water and isopropanol (99.5%, Acros Organics). The quantity of Nafion® was normalized by the catalytic support mass to have a ratio equal to 0.3 gNafion® gsupport−1. The solution was ultrasonicated during 15 min to obtain a homogeneous ink, 10 μL of which being deposited on the working electrode: the Pt loading on the electrode was then 20 μg cm−2geo.

The electrochemical characterizations were performed at T = 25 °C prior to and after the two ASTs. Firstly, the working electrode was immersed in the argon-purged electrolyte at controlled potential E = 0.40 V vs. RHE. Then, 50 cyclic voltammetries (CVs) at ν = 50 mV s−1, between 0.05 V vs. RHE and 1.23 V vs. RHE, were performed, following by 3 CVs at ν = 20 mV s−1 between 0.05 V vs. RHE and 1.23 V vs. RHE. The Pt-specific surface area (SPt) was determined by coulometry of the under-potential desorption of H atoms (Hupd). For the calculations, the authors assumed that the electrooxidation of a Hads monolayer requires 210 μC cm−2. Finally, the electrolyte was saturated of oxygen, and two CVs were performed at ν = 5 mV s−1, between 1.05 V vs. RHE and 0.20 V vs. RHE at ω = 1600 rpm to evaluate the activity of oxygen reduction reaction (ORR). The latter was expressed in terms of surface (SA0.9) and mass (MA0.9) activities, determined at E = 0.90 V vs. RHE after correction for the Ohmic drop and oxygen mass-transport in a solution.

The two ASTs were performed at T = 80 °C to mimic the working conditions of the PEMFCs [105]. They are either a base-load cycle procedure to study the aging of the Pt catalyst (succession of 15,000 cyclic voltammetries under inert atmosphere between E = 0.6 V vs. RHE and E = 1.0 V vs. RHE at v = 50 mV s−1, hereafter denoted by “AST P1”) or a start-stop procedure to study the aging of the catalyst support (succession of 1000 3-s potential steps under inert atmosphere between E = 1.0 V vs. RHE and E = 1.5 V vs. RHE, hereafter denoted by “AST P2”).

Results and Discussion

Physicochemical Properties of the Catalyst Supports and Platinum-Based Electrocatalysts

Physicochemical Properties of the Catalyst Supports

XRD, X-ray energy-dispersive spectrometry (XEDS), SEM (Fig. 1), N2 sorption (Fig. 2), and TGA have been carried out on the raw carbon aerogel and on the tin oxide/carbon aerogel composites to probe the influence of the experimental synthesis parameters (pH value of the reactive medium, quantity of tin precursor) on the quality and the quantity of the coating. Their electrical conductivity was also measured. The results of these characterizations are summarized in Table 1.

Fig. 1
figure 1

SEM micrographs and schematic representations (gray: CA and blue: SnOx) of samples CA, SnOx(0.7–10 g)/CA, SnOx(11.5–2 g)/CA, and SnOx(11.5–1 g)/CA

Full size image
Fig. 2
figure 2

Nitrogen adsorption isotherms of samples CA, SnOx(0.7–10 g)/CA, SnOx(11.5–2 g)/CA, and SnOx(11.5–1 g)/CA, measured at 77 K after an outgassing under secondary vacuum at 90 °C

Full size image
Table 1 Physicochemical characteristics of the samples synthesized. The SBET, Vtot,Vmeso, and Vmicro are reported by mass of composite
Full size table

The raw carbon aerogel (CA) is composed of carbon nanoparticles, with a diameter of 15 ± 5 nm, bound by covalent links leading to the formation of a 3D micro-/meso-/macro-porous material. Whatever the operational parameters, the tin oxide coating (confirmed by XRD and XEDS measurements, not shown here) is visible on the different composites (the surfaces appear rough, which contrasts with the smooth texture of the carbon aerogel). It is formed of nanoparticles with a diameter ca. 5 nm. The values of the textural characteristics (SBET, Vtot, and Vmicro) logically decreased after the coating step (Table 1). The specific surface area evolves from 967 to values around 230 m2 g−1 for the composites, but with a different variation of the porosity/texture. The drop of these specific surface area values is due to the partial filling of the largest pores and the closing of almost all the micropores of the raw carbon aerogel support. Furthermore, it is emphasized owing to the much larger density of tin dioxide versus carbon (6.95 g cm−3 vs. 2.0 g cm−3). Nevertheless, one can observe some differences between the samples after the coating step. The SnOx(0.7–10 g)/CA exhibits a heterogeneous and not-covering coating (some parts of carbon surface remain unprotected). This phenomenon logically leads to a low percentage of SnOx measured by TGA in air (52 wt%), which is assumed as the wt% of the coating. Moreover, few independent aggregates/blocks (up to several micrometers), composed of tin oxide nanoparticles, are observed off the carbon surface on this sample (as previously observed in these synthesis conditions [95]). These blocks are useless and detrimental for the application, because they do not protect the carbon aerogel surface and they would electrically insulate the catalyst nanoparticles from the CA. The two samples, SnOx(11.5–2 g)/CA and SnOx(11.5–1 g)/CA, synthesized in basic solution (pH = 11.5), exhibit a good dispersion of tin oxide nanoparticles, in spite of a lower quantity of tin precursor (2 g and 1 g, respectively). The percentages of SnOx measured by TGA are equal to 81.5 wt% and 64.4 wt%, respectively, logically explained by the difference of the quantity of tin precursor. Considering a totally covering coating of the tin dioxide nanoparticles (with a diameter equal to 5 nm), the theoretical percentage of the coating value is about 95 wt%. So, the sample SnOx(11.5–2 g)/CA exhibits an interesting and significant coating (81.5 wt%) with a visible porosity and interesting SBET and Vtot values (225 m2 g−1 and 0.56 cm3 g−1) for the application. The coating of the sample SnOx(11.5–1 g)/CA is correct but limited by the lower quantity of tin precursor. No detrimental SnOx independent blocks are observed for these two samples. The differences observed between sample synthesized in acidic (pH = 0.7) and basic (pH = 11.5) solution are the consequences of electrostatic interactions (repulsion for samples synthesized in acidic condition or attraction for samples synthesized in basic condition) between the reactive species and the CA surface [95]. That is why the percentages of SnOx measured by TGA are higher for the samples SnOx(11.5–2 g)/CA and SnOx(11.5–1 g)/CA than the percentage of the sample SnOx(0.7–10 g)/CA, even though lower tin precursor contents had been used.

Not only the quantity, but also the quality of the tin oxide coatings, has a major influence on the electrical conductivity values σ of the samples (composites dried at 90 °C in air without any further calcination). For example, the higher the quantity of coating, the lower the electrical conductivity value σ. The more or less pronounced presence of tin dioxide (a semi-conductor material: band gap about 3.5 eV [68]) explains this phenomenon. Furthermore, the sample SnOx(0.7–10 g)/CA exhibits a heterogeneous and non-covering coating with plenty of uncovered areas of carbon aerogel surface, which are enough to allow fast transfer of electrons (small contact resistance): this sample has the highest electrical conductivity value of the composites, but exposes unprotected areas of the CA.

Physicochemical Properties of Platinum-Based Electrocatalysts

The different electrocatalysts synthesized from the raw carbon aerogel and the tin oxide/carbon aerogel composites are analyzed by TEM. The pictures and the Pt nanoparticle size distribution are given in Fig. 3. The XRD diffractograms, plotted for the four electrocatalysts, are shown in Fig. 4. Moreover, the Pt content (calculated with AAS), the average diameter of the Pt dTEM, and the volume-weighted average diameter dv (calculated from TEM pictures) and the average size of Pt nanocrystallites (dXDR, calculated with the Debye-Scherrer equation from the (111) diffraction peak at 2θ about 39.8°) are given in the Table 2.

Fig. 3
figure 3

TEM pictures and Pt nanoparticle size distribution of Pt/CA, Pt/SnOx(0.7–10 g)/CA, Pt/SnOx(11.5–2 g)/ CA, and Pt/SnOx(11.5–1 g)/CA

Full size image
Fig. 4
figure 4

XRD diffractograms of Pt/CA, Pt/SnOx(0.7–10 g)/CA, Pt/SnOx(11.5–2 g)/CA, and Pt/SnOx(11.5–1 g)/CA electrocatalysts. The number sign and currency sign correspond to the visible peaks of the crystallographic plans of SnO2 and Pt, respectively

Full size image
Table 2 Physicochemical characteristics of the four electrocatalysts
Full size table

Whatever the catalyst support, the deposition of Pt Nanoparticles is obvious (Figs. 3 and 4). Slight differences in dXRD are observed for the various supports values. This agrees with the colloidal method used for their synthesis (samples coming from the same colloidal suspension): apart from different degrees of agglomeration for different supports, the crystallite sizes should remain essentially identical. Differences between dv and dXRD could be observed. The high values of dv could be explained by the agglomerations of platinum nanoparticles, which are relatively abundant in the electrocatalysts Pt/SnOx(0.7–10 g)/CA and Pt/SnOx(11.5–2 g)/CA. Moreover, it is possible to have some small nanoparticles (with a diameter lower than or equal to 1 nm), which are not distinguishable from the background by image analysis, leading to dv values oversized. For the electrocatalyst based on the raw carbon aerogel (Pt/CA), the deposition of the Pt Nanoparticles is homogeneous with a narrow size distribution, centered on 2.1 nm. The carbon aerogel, with its low organized structure and its high specific surface area (967 m2 g−1), displays a lot of anchoring sites for the nanoparticles. For the electrocatalysts based on the tin oxide–coated carbon aerogel (Pt/SnOx/CA), the Pt nanoparticle distributions are wider, with a tendency to gather Pt nanoparticles, as a result of larger degree of agglomeration of individual crystallites. This phenomenon is accompanied by non-negligible Pt nanoparticle agglomeration [106, 107] and can be explained by two main factors. The less pronounced porosity of the SnOx/CA supports leads to lower specific surface area values (between 225 and 236 m2 g−1) and to less anchoring site for the Pt nanoparticles. Moreover, the quantity and the quality of the coating have also a major influence on the Pt nanoparticle deposition. Although the SnOx(11.5–2 g)/CA support has the more homogeneous and the more covering coating, the Pt/SnOx(11.5–2 g)/CA electrocatalyst displays a mediocre Pt dispersion with a significant number of Pt agglomerates. This could be explained by the different interactions between, on the one hand, Pt and carbon and, on the other hand, Pt and tin dioxide [87, 108]. In the second case, Pt tends to agglomerate, as already observed by Cognard et al., on the tin dioxide supports. This phenomenon was observed on non-doped tin dioxide supports, as well as on doped (with niobium or antimony elements) supports [71,72,73]. In other words, not only the textural characteristics, but also the chemical nature of the catalyst support, have a direct influence on the quality of the Pt nanoparticles deposited. Finally, the morphologies, in terms of distribution, of the Pt/EN and Pt/YS samples are very close to those monitored for Pt/CA (see [39] for details).

Electrochemical and Physicochemical Characterizations after the Load Cycle Procedure of Accelerated Stress Tests (AST P1)

The four Pt-based electrocatalysts were tested electrochemically, firstly in argon-saturated electrolyte and then in oxygen-saturated electrolyte to measure their ORR activity. These characterizations were performed prior to and after the AST P1, mimicking a base-load cycle procedure. The CVs curves plotted are given in Figs. 5 and 6. Physicochemical characterizations (TEM (Fig. 7) and Raman spectroscopy) were also performed at the end of AST P1, to probe the changes of the Pt nanoparticles and of the catalytic support, as summarized in Table 3. The determination of the specific surface after the AST P1 and P2 was not performed for all electrocatalysts (as well as the specific activity at 0.9 V vs. RHE). Indeed, the specific surface was determined from the Hupd region, using the capacitive current in the 0.4–0.6 V vs. RHE as a baseline to precisely determine the charge under the Hupd peak. During the aging, the formation of a quinone-hydroquinone peak (due to the partial corrosion of the carbon) was often observed, hence modifying the baseline (possibly without any change of the Hupd) and resulting in an underestimation of the Hupd (this is likely very much the case for Pt/CA, and in a lesser extent for Pt/SnOx11.5–2 g)/CA, for which the capacitive current strongly evolves after aging). To avoid any misinterpretation of the results, we thus decided to comment the evolution of the specific surface areas in a qualitative manner.

Fig. 5
figure 5

a–d Representative cyclic voltammograms in argon-purged electrolyte (0.1 M HClO4, scan rate v = 20 mV s−1) at T = 25 °C of the Pt-based electrocatalysts prior to (fresh, full lines) and after (dotted lines) AST P1 (succession of 15,000 cyclic voltammetries under inert atmosphere between E = 0.6 V vs. RHE and E = 1.0 V vs. RHE at v = 50 mV s−1, in 0.1 M HClO4 at T = 80 °C)

Full size image
Fig. 6
figure 6

a–d Representative ORR cyclic voltammograms at T = 25 °C of the Pt-based electrocatalysts prior to (fresh, full lines) and after (dotted lines) AST P1 (succession of 15,000 cycles)

Full size image
Fig. 7
figure 7

Representative TEM micrographs obtained after AST P1 and corresponding Pt nanoparticle size distribution of the Pt-based electrocatalysts prior to and after AST P1 (succession of 15,000 cycles)

Full size image
Table 3 Summary of the kinetic parameters extracted from the ORR cyclic voltammograms and physicochemical properties of the Pt-based electrocatalysts prior to and after AST P1
Full size table

The cyclic voltammograms obtained in an argon-saturated electrolyte (Fig. 5) display a classical behavior with the absorption/desorption of the proton H+ at low working potentials (E < 0.4 V vs. RHE). The evolution of the electrochemical behavior of the electrocatalysts depends on the catalytic support. The double-layer capacitance (0.4 < E < 0.6 vs. RHE) evolves, as a consequence of the formation of new oxygen-containing groups, which initiate the carbon corrosion mechanism into CO2, thereby destroying the carbon support and leading to increased diffusion overpotential in systems [22]. The degradation of the carbon aerogel is confirmed by the evolution of the La values (from 0.8 to 3.3 nm), signing the irreversible corrosion of the amorphous regions of the CA, leaving only (at least at that stage of the process) the more graphitic regions of the material (hence, the increase of La). Such carbon degradation favors the loss of some Pt nanoparticles. The Pt nanoparticles supported on the raw carbon aerogel are also subjected to the Ostwald ripening mechanism during AST P1 (broadening of the particle size dispersion, an increase of the average diameter, and keeping the spherical shapes of the nanoparticles [109], Fig. 7). The decrease of MA0.9 (ca. 54%) is directly linked to these two phenomena. Overall, the carbon-related degradation of Pt/CA under P1 AST is more important than what witnessed for more graphitic carbons (Pt/EN and more specifically Pt/YS), as detailed in [39] and recalled in supporting information (Table SI-1), whereas its initial performances (SA0.90, M0.90) are intermediate (this phenomenon being induced by the agglomerated nanoparticles observed on Pt/CA and, to a larger extent, Pt/YS (see [39]).

The Pt/SnOx(0.7–10 g)/CA and Pt/SnOx(11.5–1 g)/CA exhibit a slight decrease in SPt during the AST P1 (ca. 20%, see Table 3). The presence of the tin oxide seems to limit the degradation of the carbon aerogel (no evolution of the double-layer capacitance and the La ini and La fin values are similar, Table 3), resulting in more stable catalytic supports. The loss of SPt for Pt/SnOx(0.7–10 g)/CA could be explained by the drastic increase of the average Pt nanoparticle size, leading to an increase of the surface activity (SA0.9 increases from 204.8 to 286.7 μA cmPt−2). On the contrary, the particle size distribution of Pt/SnOx(11.5–1 g)/CA displays a little offset to bigger sizes (Fig. 7). This sample also exhibits similar initial and final catalytic activities (Table 3). In this case, the coating seems to improve the behavior of the electrocatalyst (stability of the Pt nanoparticles and protection of the support). Finally, the Pt/SnOx(11.5–2 g)/CA electrocatalyst exhibits a low SPt ini value compared to the other samples, which could be explained by the mediocre disposition of the Pt (Fig. 3) and most importantly by the low electrical conductivity of the support (0.2 ± 0.1 S cm−1). In that case, the homogeneous and covering tin oxide coating (Table 1) seems to insulate a significant number of Pt particles from the CA support, this phenomenon explaining the low initial value of mass activity MA0.9 ini (55.7 A gPt−1) and why there is no or very little change of the Pt particle size distribution (Fig. 7).

To conclude, the presence of the tin oxide coating can be either beneficial or detrimental to the catalytic performances of the Pt-based electrocatalysts in AST P1, depending on the original “morphology” of the Pt/SnOx/CA sample. While the sample Pt/SnOx(11.5–1 g)/CA exhibits promising performances and durability (its CA surface is only partially covered by the SnO2 deposit), samples where the CA surface is totally covered by SnO2 exhibit unsatisfactory initial catalytic performances, because of the low electrical conductivity values of the support material.

Electrochemical and Physicochemical Characterizations after the Start-Stop Procedure of Accelerated Stress Tests (AST P2)

Finally, the Pt-based electrocatalysts were tested electrochemically under the AST P2, which mimics the frequent start-stops that can be experienced in automotive applications. As previously, the samples were tested in argon-saturated electrolyte and then in oxygen-saturated electrolyte, both prior to and after the AST P2. The corresponding CVs are given in the Figs. 8 and 9, respectively. Physicochemical characterizations (TEM (Fig. 10) and Raman spectroscopy) were also performed at the end of AST P2 (Table 4).

Fig. 8
figure 8

a–d Representative cyclic voltammograms in argon-purged electrolyte (0.1 M HClO4) at T = 25 °C of the Pt-based electrocatalysts prior to (fresh, full lines) and after (dotted lines) AST P2 (succession of 1000 3-s potential steps under inert atmosphere between E = 1.0 V vs. RHE and E = 1.5 V vs. RHE, in 0.1 M HClO4 at T = 80 °C)

Full size image
Fig. 9
figure 9

a–d Representative ORR cyclic voltammograms at T = 25 °C of the Pt-based electrocatalysts prior to (fresh, full lines) and after (dotted lines) AST P2 (succession of 1000 cycles)

Full size image
Fig. 10
figure 10

Representative TEM micrographs obtained after AST P2 and corresponding Pt nanoparticle size distribution histograms of the Pt-based electrocatalysts prior to and after the 1000 cycles of AST P2

Full size image
Table 4 Summary of the kinetic parameters extracted from the ORR cyclic voltammetries and physicochemical properties of the Pt-based electrocatalysts prior to and after the 1000 cycles of AST P2
Full size table

The cyclic voltammograms obtained in argon-saturated electrolyte (Fig. 8) evolve a lot upon AST P2, which is the consequence of a severe degradation of the catalytic support. The Hupd features (hence, the specific surface area of Pt) partially disappear for Pt/CA and Pt/SnOx(0.7–10 g)/CA and are completely suppressed for Pt/SnOx(11.5–2 g)/CA and Pt/SnOx(11.5–1 g)/CA. Furthermore, samples synthesized with raw carbon aerogel (Pt/CA) or with the heterogeneous and non-covering tin oxide coating (Pt/SnOx(0.7–10 g)/CA) display a significant enlargement of their double-layer capacitance, characteristic of the formation of oxygen-containing groups at the available carbon surface. The appearance of the oxidation peak relative to the quinone into hydroquinone transformation at E ≈ 0.6 V vs. RHE [4] is another indicator of the formations of these new functional groups. The harmful effect of these groups for the catalytic support for the PEMFC application was previously discussed in this study. As a result, at the end of AST P2, the raw CA support is heavily degraded, in line with the increase of the La value (from 0.8 to 1.8 nm): the amorphous carbon firstly corrodes (leading to an apparent increase of La), but then the organized parts of the CA also degrade in these conditions (later), leading to a further decrease of La. Overall, these degradations lead to the detachment of a significant number of Pt nanoparticles from the CA, further explaining the considerable drop of the mass activity (about 71%), this latter value being calculated with respect to the initial mass of Pt contained in the sample.

The Pt/SnOx(0.7–10 g)/CA electrocatalyst exhibits a similar percentage of loss of MA0.9 than Pt/CA. The plentiful uncovered surface areas of the carbon seem to undergo the same degradations as for the bare Pt/CA. In these conditions, the tin oxide coating has no beneficial effect on the durability of the catalytic support. The Pt/SnOx(11.5–2 g)/CA and Pt/SnOx(11.5–1 g)/CA samples also display visible changes of the double-layer capacitance after AST P2, especially for Pt/SnOx(11.5–2 g)/CA. In this case, one can wonder if the tin oxide coatings were also modified under the very oxidizing conditions of AST P2. The similar values of the La ini and La fin obtained for the two samples (Table 4) seem to indicate a relative stability of the carbon aerogel, though, probably because no CA is directly exposed to the electrolyte. This apparent stability must, however, not be understood as a positive effect. Indeed, as stated above, the tin oxide deposit seems strongly modified. This assumption is corroborated by the drastic loss of the mass activity (and of specific surface, as evidenced by the disappearance of the Hupd) when no change of Pt particle size is observed. The alteration of tin oxide catalytic supports was already observed by Cognard et al. [73]: some dissolution of tin in argon-purged 0.1 M electrolyte at T = 57 °C after a holding of the disk potential at E = 1.2 V vs. RHE for 30 min using rotating disk and rotating ring-disk electrode was noticed. According to the authors, such tin oxide dissolution leads to the formation of an amorphous layer with a low electrical conductivity at the surface of the samples. In the present case, this transformation could lead to an electrical disconnection of the Pt nanoparticles, this likely causing the drastic drop of MA0.9 for Pt/SnOx(11.5–2 g)/CA and Pt/SnOx(11.51 g)/CA noted in Table 4.

Finally, after AST P2, the Pt/CA sample exhibits a broader Pt nanoparticle size distribution histogram, with a shift towards bigger sizes (Fig. 10). Thus, for this sample, the degradation of the raw CA support is likely coupled with some Ostwald ripening. This is not observed for the tin oxide–coated CAs (Fig. 10), the difference being tentatively explained by the electrical disconnection of the Pt nanoparticles discussed previously for these samples. In comparison, it seems that having a more graphitized carbon support like YS (see Table SI-1) results in a slightly higher (if not similar) durability in start-stop conditions, the HSAC support (Pt/EN) suffering from higher degradation.

Conclusion

Several tin oxide coatings were synthesized by impregnation and dried at 90 °C on the surface of a carbon aerogel to improve the durability of this catalytic support under the working conditions of a PEMFC. The influence of the quantity and quality (homogeneity and covering aspect) of this coating on the Pt nanoparticle depositions, their catalytic performances towards the oxygen reduction reaction, and the durability of the electrocatalysts were studied for two accelerated stress tests procedures, one mimicking a base-load cycling (AST P1) and the other a start-stop procedure (AST P2). It turns out that the not only textural characteristics of the composite (especially the porosity and the specific surface area), but also the chemical nature (the Pt nanoparticles, tend to agglomerate on the tin oxide surface) of the catalyst support have a direct influence on the Pt nanoparticle deposition. At the end of AST P1, some coatings lead to an improvement of the durability of the electrocatalysts: for example, the Pt/SnOx(11.5–1 g)/CA sample exhibits interesting initial performances, a good durability of the catalytic support, and stable Pt nanoparticles. Nevertheless, if the carbon surface is totally covered, the initial catalytic performances are unsatisfactory, because of the low electrical conductivity value of the catalytic support. Finally, all electrocatalysts exhibit significant degradations after AST P2, leading to a severe drop of the catalytic activity. It could be explained either by the degradation, under the harsh experimental conditions of AST P2, of the uncovered carbon surface areas or by the alteration of the tin oxide coating, resulting in the formation of an amorphous layer with a low electrical conductivity, thereby disconnecting a significant number of Pt nanoparticles. As a result, the real beneficial interest of tin oxide coating to protect carbon supports is still open to question, hence requesting investigation with different carbon supports, as YS or EN.