Lorenzo
Mazzoli
a,
Angus
Pedersen
*ab,
Simon
Kellner
a,
Robert D.
Hunter
a,
Rongsheng
Cai
c,
Mengnan
Wang
ab,
Kevin
Sivula
d,
Sarah J.
Haigh
c and
Jesús
Barrio
*a
aDepartment of Chemical Engineering, Imperial College London, London SW7 2AZ, UK. E-mail: a.pedersen19@imperial.ac.uk; j.barrio-hermida@imperial.ac.uk
bDepartment of Materials, Royal School of Mines, Imperial College London, London SW7 2AZ, UK
cDepartment of Materials, University of Manchester, Oxford Road, Manchester, M13 9PL, UK
dLaboratory for Molecular Engineering of Optoelectronic Nanomaterials, Institute of Chemical Sciences and Engineering, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
First published on 2nd February 2024
Iron–nitrogen–carbon (FeNC) electrocatalysts are emerging as a low-cost alternative to Pt-based materials for electrochemical oxygen reduction at the cathode of alkaline exchange membrane hydrogen fuel cells. The valorisation of waste biomass is a sustainable pathway that could allow the large-scale production of such catalysts. By means of hydrothermal carbonization (HTC), a biomass-derived carbohydrate can be converted into a carbonaceous framework, however, the electrocatalytic performance of the metal–nitrogen–carbon electrocatalysts prepared through HTC is suboptimal owing to the lack of microporosity in the highly crosslinked carbon frameworks. In this work, we address this issue by adding polystyrene sulfonate (kayexalate) in the HTC of xylose. Kayexalate's negative charges mitigate particle aggregation, resulting in smaller carbon-based particles, with the O2 activation leading to a four-fold increase in specific surface area (127 vs. 478 m2 g−1). Subsequent high-temperature pyrolysis in the presence of an N and Fe source leads to an active FeNC. This produces a corresponding increase in the electrocatalytic activity for the oxygen reduction in alkaline media in a rotating disk electrode (1.45 vs. 14.3 A g−1 at 0.8 V vs. RHE) and in a gas diffusion electrode at high current densities (≥2 A cm−2). The sustainable character of the reported catalyst as well as the high electrocatalytic activity at industrially relevant current densities provides a pathway to catalyst design for low-cost cathodes in alkaline exchange membrane fuel cells.
To align with the growing emphasis on combining sustainability and commercial viability, it becomes imperative to explore synthetic approaches that utilize sustainable sources for the preparation of efficient FeNC oxygen reduction electrocatalysts.17 In that regard, lignocellulosic biomass is considered a potential sustainable carbon precursor in the synthesis of electrocatalyst materials owing to its abundance as well as the high carbon and (often) heteroatom content.18,19 Furthermore, every year a significant portion of biomass waste (agricultural biomass production within the European Union amounts to approximately 956 Mt annually, with residues accounting for 46%)20 is left in fields to decompose or is eventually discarded and used as a low-grade energy source, contributing to greenhouse gas emissions. This abundant bio-waste can be converted into carbonaceous catalysts and supports by means of hydrothermal carbonization (HTC), which entails mild carbonization at temperatures <250 °C and self-generated pressure in the presence of water.21,22 Among the different types of biomass, hemicellulose can be used to obtain xylose. Commercially, hydrolysis of xylan (25–30% of corn cob, for instance) can be employed to produce xylose (C5H10O5).23,24 Xylose is therefore a cheap, abundant and sustainable carbon-based precursor which has been employed for the preparation of heterogeneous catalysts as well as FeNC materials.25 However, the electrochemical performance of xylose-derived FeNC oxygen reduction electrocatalysts is often low. This is due to the low porosity and specific surface area of the materials obtained after hydrothermal carbonization, which does not allow full access to the FeNx active sites. For instance, Feng et al.26 developed a Fe–N–C xylose-derived electrocatalyst, using melamine and Iron(II) chloride as nitrogen and iron precursors. However, the obtained electrocatalysts displayed specific surface areas ranging 40–70 m2 g−1, which led to moderate oxygen reduction activity in alkaline media. The introduction of microporosity is crucial for the development of active FeNC electrocatalysts, as it allows the formation of a higher number of active sites.27,28 Templating methods with either silica or molten salts (ionothermal carbonization29) can lead to high surface areas. Our group for instance employed magnesium chloride hexahydrate in the presence of 2,4,6-triaminopyrimidine as reactants for the synthesis of highly porous nitrogen-doped carbons that efficiently exposed the FeNx active sites, resulting in a much higher surface area of 3295 m2 g−1.30 Sodium salts have been widely employed as porogens as well.31–33 Waterhouse and co-workers for instance employed a mixture of zeolitic-imidazole framework 8 and NaCl to synthesize FeN4 electrocatalysts with a surface area >1900 m2 g−1 and 26.3 × 1019 sites g−1.34 Silica templating has been thoroughly investigated by Atanassov and co-workers, who often observe electrocatalysts with specific surface areas up to 600 m2 g−1.14,35 Yet such templating methods require a high salt/organic precursor ratio as well as an acid wash, which requires HF in the case of SiO2, which leads to potential significant impacts on human health.6 Mazzucato and coworkers, on the other hand, employed CO2 and steam activation to regulate the formation of micro and mesopores in carbon black, which increased the site density and turnover frequency of the final catalyst.36
Therefore, in this work, we have addressed this issue by inducing microporosity in xylose-based FeNC electrocatalyst through the addition of small amounts of poly(sodium 4-styrene sulfonate) (kayexalate) during the HTC of xylose. The charged sulfonate groups of kayexalate avoid the aggregation of the carbon particles during the HTC leading to a much smaller sphere size and increased surface area of the final material.26,37 The particle size reduction in the hydrothermal spheres was proven through scanning electron microscopy (SEM) images. The enhancement in surface areas of the final catalysts (prepared by pyrolysis of the hydrothermal spheres in the presence of melamine and FeCl2) was evaluated with N2 sorption and double-layer capacitance (Cdl) measurements (proportional to the electrochemical surface area (ECSA)). The homogeneous distribution of nitrogen and iron was confirmed by means of scanning transmission electron microscopy – energy dispersive X-ray spectroscopy (STEM-EDX) – and the chemical composition was further evaluated by X-ray photoelectron spectroscopy (XPS). The electrochemical oxygen reduction activity was evaluated in alkaline electrolyte (0.1 M KOH) in a rotating disk electrode (RDE), where the insertion of kayexalate led to an almost 10-fold enhancement in the mass activity. Furthermore, owing to the increased specific surface area (478 m2 g−1) and microporosity, the material was tested in a gas diffusion electrode (GDE) configuration and compared to the state-of-the-art FeNC catalyst (PMF-D14401, Pajarito Powder).
The kinetic current densities (jkin) were calculated at 0.80 and/or 0.75 VRHE, using the geometric disk current density (jd) and the limiting current density (jlim) found at 0.3 VRHE following eqn. (1):
(1) |
(2) |
(3) |
The selectivity was evaluated through the %H2O2 produced employing eqn (4):
(4) |
Subsequently, the double-layer capacitance Cdl was calculated by plotting the difference between the anodic and cathodic current density at a certain potential (in a non-faradaic region, at 0.50 VRHE) versus different scan rates (in this work 20, 40, 60 and 80 mV s−1).40 The tests were pursued in N2-saturated KOH 0.1 M, at 0 rpm. The slope of this linear plot is Cdl in F cm−2; by dividing this number by the total amount of loaded catalyst, a normalized value of Cdl in F g−1 is obtained. The double-layer capacitance provides valuable insights into the interfacial electrochemical behaviour and the extent to which an electrocatalyst participates in ORR reactions. By being directly proportional to the ECSA (electrochemical surface area), a higher Cdl indicates a larger electrochemically active surface area, often correlated with enhanced catalytic performance.41 The stability of the catalysts was evaluated through accelerated stress tests through cyclic voltammetry cycles at potentials between 0.8 and 0.4 VRHE in O2, at 100 mV s−1. The catalysts were then tested after 1000 and 8000 cycles by recording a cyclic voltammetry in the potential range 1.00–0.20 VRHE.
FeXK Gas Diffusion Electrodes (GDEs) were crafted in-house using an airbrush set and an Iwata Smart Jet Pro Airbrush compressor. The Pajarito PMFD14401 GDEs were created with an Exactacoat ultrasonic spray coater on a 5 cm2 gas diffusion layer (Freudenberg H23C8). The catalyst loading was 1.29mgFeNC cm−2, determined by pre- and post-weighing the gas diffusion layer after spraying. Electrochemical measurements were performed using a GDE setup filled with a 1 M KOH aqueous solution. The potentiostat PGSTAT204 with FRA32M Module (Metrohm) was employed, along with galvanostatic steps coupled with Electrochemical Impedance Spectroscopy (EIS). The detailed methodology, including ink formulation and spray-coating specifics, is provided in the ESI.† Procedures for all the material characterizations are also included in the ESI.†
Finally, from the N2 isotherm (Fig. 2a), the increased quantity of N2 adsorption at high relative pressure (P/P0 > 0.9) in FeXK indicates an increased macropore volume arising between aggregates of FeXK sphere agglomerates.42 This possibly evolves from the smaller kayeaxalate-derived spheres (Fig. S1†). X-ray diffraction (XRD) analysis, conducted to study the crystalline structure of the catalysts, suggests that the change in the morphology of the hydrothermal precursor also affects the chemical nature of the final catalysts (Fig. 3). FeXK presents more accentuated peaks with higher intensity; the sharp peak at 43.8° confirms the presence of Fe (111), although this diffraction peak could point to the presence of Fe3C, nevertheless the diffraction peak at 45°, arising from Fe3C (110) unequivocally proves the formation of carbide species. Additionally, at 51.2°, FeXK displays a Fe (0,2,0) peak. The high intensity of these peaks might hide the presence of the overlapping Iron-Nitrides-related signals, especially in the region 43–44°, as shown in previous work.26 XRD data suggests that the decreased size of HTC-XK versus HTC-X may favor the formation of crystalline Fe-based species that stabilize ORR reaction intermediates and improve the catalytic performance. Further characterization of the materials was carried out using Raman spectroscopy (Fig. S3 and S4†), which showed the presence of two dominant peaks in the first-order spectra. These peaks are centered at approximately 1340 and 1585 cm−1, corresponding to the D and G bands respectively. The G band indicates the presence of graphitic carbon and is present in all graphitic materials, while the D band signifies disorder within the carbon structure.
Fig. 3 XRD patterns of FeX (a) and FeXK (b) (Iron (96-901-4057) and Iron carbide (00-044-1292) and Carbon (01-075-1621)). |
The intensity ratio of the D and G bands (ID/IG) is often used as an indicator of the graphitization degree in carbon materials. The ID/IG values of 1.03 and 1.15 (Table S2†) displayed by FeX and FeXK suggest the presence of nanocrystalline graphitic domains, which is consistent with the broad carbon peaks observed at approximately 26.2 and 44.4°2θ in XRD. For small graphitic crystallite sizes, the ID/IG ratio is correlated to the crystallite size using the Ferrari–Robertson model, which suggests that the ID/IG ratio increases with increasing crystallite size up to a size of 2 nm.43 Therefore, the higher value of 1.15 displayed by FeXK suggests greater graphitic ordering. Also, the variation of the ID/IG value measured at 25 different locations across the sample is greater for FeX (standard deviation of 0.17 compared to 0.05 for FeXK, Table S2†) indicating the non-uniform nature of the carbon structure, with some locations displaying higher graphitic ordering, while other locations have a more disordered structure. The less variable ID/IG values observed in the Raman spectra of FeXK are likely related to the more uniform dispersion of small Fe-based particles throughout the carbon structure. Scanning transmission electron microscopy imaging was then employed to obtain further insights into the morphology, dimensionality, and chemical composition of the Fe-based particles. Fig. 4 and Fig. S5† display the HAADF-STEM images and EDX elemental maps for FeXK and FeX, respectively. Images of FeXK show spherical nanoparticles, approximately 100 nm in diameter, with EDS mapping revealing they contain homogeneous distributions of N, C, O, and Fe. At high magnification, Z-contrast HAADF-STEM imaging reveals bright single atoms and atomic clusters (Fig. 4b), which can be assigned to Fe as the only high atomic number element present. Denser Fe-rich crystalline nanoparticles are also observed with diameters of approximately 10 nm (Fig. 4c). Selected area electron diffraction and Fast Fourier transform (FFT) analysis of high-resolution images from these particles reveals lattice spacings that can be ascribed to the Fe3C phase and/or pure Fe phase (Fig. S6†). FFT analysis along with XRD characterization confirms the presence of both Fe and Fe3C particles as observed in the diffraction peaks at 43.8 and 45 °C (Fig. 3), as well as the Miller indices shown on Fig. S6.†
Fig. 4 FeXK HAADF-STEM images showing Fe-based particles and atomic Fe species (a–c) and EDX elemental maps for C (green), Fe (red), and N (blue) (d). |
FeX is composed of much bigger spherical particles with diameters around 1 μm, which also contain N, C, O, and Fe. The HAADF STEM and EDX elemental mapping reveals that these particles contain Fe-rich nanoparticles of approximately 10 nm diameter, distributed evenly throughout the N, O, C and Fe-containing matrix. High-resolution HAADF-STEM also reveals atomically dispersed Fe species, although these are more difficult to resolve due to the matrix particles’ greater thickness. The atomically dispersed species could be FeNx sites or clusters of FexOy. Due to the small size of the particles and the invisibility of both N and O species relative to the C, O and N support, it is difficult to ascertain the true atomic Fe coordination. These can only be distinguished with certainty via cryo 57Fe Mössbauer spectroscopy.44 The local environment of Fe within FeX was studied by Feng et al.26 by means of X-ray Absorption Spectroscopy where Fe3C was observed, which is in agreement with nanoparticles seen in the HAADF-STEM and XRD data here. The HAADF STEM images of FeX (Fig. S5†), along with its low specific surface area, suggest most of the observed Fe3C nanoparticles are internal to the C–N–O matrix, indicating very low accessibility for the Fe-based species. In contrast, the substantially smaller size of the C–N–O–Fe matrix particles and higher BET surface area of FeXK indicates more efficient exposure of the Fe-based active sites. X-ray photoelectron spectroscopy was employed to provide further insights on the elemental composition of the surface of the prepared catalysts. Carbon, nitrogen, oxygen, and iron were observed in both FeX and FeXK. In terms of chemical composition, substantial differences can be observed between the catalysts (Table S3†). FeX displays a much higher N1s, O1s, and Fe2p surface content than FeXK suggesting a lower electrical conductivity and confirming the presence of Fe-based aggregates. We hypothesize that the higher porosity of FeXK may be responsible for the relatively easier release of nitrogen and oxygen functional groups particularly during the pyrolysis. C1s spectra of both materials display binding energies corresponding to C–C, C–N, and C–O species, while FeX displays an additional chemical species that suggests a higher degree of oxidation in the framework despite the lack of activation step in the O2 atmosphere (Fig. S7†). N1s spectra display four peaks for both materials at 402.2, 400.8, 399.5, and 398.1 eV, indicating the presence of graphitic, pyrrolic, iron–nitrogen, and pyridinic bonds respectively (Fig. S8†). The higher atomic percentage of the pyridinic contribution indicates a pyridinic coordination of the Fe sites within the material, in agreement Fellinger and coworkers who observed that pyridinic active sites are more thermodynamically stable than pyrrolic, which are kinetically preferred above 900 °C.45 Additionally, the amount of Fe in FeXK is 1.46 in wt% (0.32 at%, Fig. S6 and S9†), which is very similar to the Fe content obtained by ICP-MS (1.29 wt%), suggesting that the high porosity of the material exposes most of the Fe with the material resulting in a very similar content in the surface and in the bulk. In contrast, FeX exhibits a higher iron content (3.11 wt%, 0.71 at% according to XPS analysis). As for the nitrogen content, FeX and FeXK yield values of 10.39 wt% and 3.13 wt%, respectively. Such elevated nitrogen content in FeX could be attributed to the retention of C–N based fragments (arising from melamine decomposition) upon pyrolysis.46 We hypothesize that owing to the larger particle diameter and lower porosity of FeX, C–N based fragments are retained in the structure. Meanwhile in the case of FeXK, the smaller particle diameter and higher porosity results in shorter pathways for the release of N-based species during pyrolysis. The lower N and O content within FeXK may entail a higher electrical conductivity, which contributes to an enhanced electrocatalytic activity. While relatively small amounts of nitrogen within a carbon material can increase the electrical conductivity,47 elevated nitrogen contents result in the formation of semiconducting materials, which are not suitable for electrocatalysis. Carbon nitrides for instance, with a predicted stoichiometry of C3N4, display a band gap of 2.7 eV which translates in a low electrical conductivity (between 10−9–10−7 S cm−1);48 C2N covalent organic materials display a band gap of 1.7–1.9 eV,49 and overall nitrogen contents over 20 at% results in negligible electrocatalytic performances even with the presence of Fe single sites.50 Despite the higher levels of N and Fe in FeX, the presence of visible inhomogeneity in the HAADF-STEM images (Fig. S5†) suggests that a significantly higher portion of the iron formed crystalline nanoparticles instead of single atoms sites. O1s spectra suggests that the activation in air employed in the synthesis of FeXK slightly modifies the nature of the functional groups in the surface of the materials (Fig. S10 and Table S4†). Both materials display three different chemical binding energies corresponding to CO bonds, C–O–C/COOH bonds and C–OH groups.51 However, the atomic contribution of C–O–C/COOH bonds in FeXK is substantially higher than in FeX, suggesting that, despite the lower oxygen content within FeXK, the oxygen species change upon air activation. From this characterization, we conclude that the smaller sphere size obtained in the kayexalate-mediated hydrothermal synthesis of xylose-based carbon frameworks strongly determines the physical and chemical properties of the FeNC materials. The higher BET surface area, homogeneous distribution of Fe-based species and suitable N and O content suggest a higher electrocatalytic performance of FeXK compared to FeX. To elucidate whether the increase in surface area translates into more exposed active sites during electrochemical measurements, we determined the double-layer capacitance, denoted as Cdl, from the slopes observed in cyclic voltammograms obtained at different scan rates. This parameter is in fact directly proportional to the electrochemically active surface area. As illustrated in Fig. S11† the Cdl of FeXK surpasses that of FeX by more than twofold (181.1 mF cm−2vs. 84.5 mF cm−2), underscoring the efficacy of kayexalate as a porogen for tailoring the surface area of carbon-based electrocatalysts.
The electrocatalytic activity for the ORR was then screened in alkaline electrolyte (0.1 M KOH) and compared to that of a commercial FeNC material (Pajarito Powder PMFD14401). The coexistence of Fe–Nx and Fe3C has been found to significantly enhance the catalytic performance of the ORR, as shown in several studies.52–54 The theoretical research (DFT) conducted by Reda et al.55 showed that the presence of Fe3C not only stabilizes the intermediates involved in the ORR but also enhances the catalytic performance in alkaline media owing to its electron-donating nature. As illustrated in Fig. 5a, FeXK outperformed FeX across all parameters, including onset potential Eonset (potential at 0.1 mA cm−2, 0.84 and 0.90 V for FeX and FeXK, respectively), half-wave potential E1/2 (0.70 V and 0.80 V) and limiting current density Jlim (−3.68 and −5.36 mA cm−2). The relative performance improvement was substantial across the entire potential range. The addition of kayexalate led to a higher catalytic activity, with FeXK reaching a kinetic current density Jkin of −3.71 mA cm−2 at 0.8 V vs. RHE and a correspondent kinetic mass activity mkin of 14.30 A g−1. On the other side, FeX only showed values of −0.36 mA cm−2 and 1.45 A g−1. Subsequently, a comparison was made between FeXK and PMFD14401, a commercial FeNC material prepared by means of SiO2 templating and acid etching. PMF D14401 displayed a higher electrocatalytic activity reaching Jkin of −15.7 mA cm−2 and mkin of 60.4 A g−1 at 0.8 ViR-freevs. RHE (Table S5†). However, FeXK displayed a higher current density at more cathodic potentials suggesting improved mass transport to the active site, possibly arising from the higher specific area compared to FeX. While FeXK possesses a lower ORR mkin than PMFD14401, it ranks well in terms of biomass-derived catalysts in the literature (Fig. S12, S13 and Table S6†).
FeXK exhibited high selectivity towards the four electron ORR, as evidenced by the average electron transfer number n = 3.95 (Fig. 5b and Fig. S14†), which remains very similar to that of PMFD14401 in a wide potential range. The selectivity of FeX to four electron ORR was lower, with an average value of the electron transfer number of 3.76. The observed high selectivity of FeXK towards the four electron ORR pathway can be attributed to its greater porosity and BET specific surface area, which led to a greater density of four electron selective active sites and (or) tortuosity changes. As discussed earlier, micropores have previously been proposed to host active FeNx sites;27,28 therefore, a higher micropore volume should lend itself to a higher density of four electron selective sites. Additionally, a greater porosity within FeXK could lead to a higher tortuosity in the catalyst layer, providing a longer residence time for H2O2 generated in the catalyst to be further reduced, resulting in an observed increase in four electron selectivity.
The stability of the catalysts was then tested by recording the electrochemical performance after 1000 and 8000 cyclic voltammetry recorded between 0.8 and 0.4 VRHE in O2, at 100 mV s−1 and with an applied potential of 1.27 VRHE for the Pt ring (Fig. S15a and b†). The kinetic mass activity of FeXK showed a 50% and 84% decrease respectively after 1000 and 8000 cycles, while PMFD14401 showed a decrease of 25% and 46%, respectively (Fig. S15c†). However, the rate of loss of kinetic mass activity at 0.8 V vs. RHE was comparable between FeXK and PMFD14401 from 0–1000 cycles (8.7 and 9.3 mA g−1FeNC cycle−1) and from 1000–8000 cycles (0.5 and 1.1 mA g−1FeNC cycle−1). This suggests a similar degradation mechanism for both FeNC. FeNC can degrade through several pathways, including carbon corrosion,56 reactive oxygen species,57 and active Fe species agglomeration58 or dissolution.59 Due to the low upper potential limit (0.8 VRHE) and room temperature conditions for the accelerated stress test, carbon corrosion should not play a factor56 on the FeNC degradation here. Additionally, reactive oxygen species generated via Fenton reactions have been found to possess short lifetimes in alkaline environments,57 indicating this should not be the main degradation pathway here. Fe agglomeration has previously been mainly observed at T > 60 °C.58 The observed degradation rate in FeXK and PMFD14401 may therefore arise from Fe dissolution, which are known to be dependent on the FeNC preparation method60 and chemical environment and location of the FeNx sites (e.g. in micropores).61
To test the material under the practical high current density conditions of a fuel cell, we employed a GDE setup.62 The previously benchmarked GDE measurement protocol,63 which coupled galvanostatic steps and electrochemical impedance spectroscopy is used in a small area gas-diffusion-electrode cell.64Fig. 6 shows the GDE results obtained for FeXK and PMFD14401. As can be seen, the performance of the FeXK operates at 47 ± 35 and 57 ± 12 mV below state-of-the-art commercial PMFD14401 at 0.25 and 2 A cm−2, respectively. While the results obtained with iR-correction and normalized per mass are very similar, the uncorrected performance of FeXK decreases compared with that of PMFD14401 which may arise from the unoptimized N and O content within the catalyst, leading to low electronic conductivity and high charge transfer resistance. Still, FeXK demonstrates the ability to operate at practical current densities (−2 A cm−2), holding promise as a suitable xylose-derived catalyst in AEMFCs.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3gc04645a |
This journal is © The Royal Society of Chemistry 2024 |