Biomass waste-derived nitrogen and iron co-doped nanoporous carbons as electrocatalysts for the oxygen reduction reaction

Biomass from agricultural by-products is gaining increasing interest as cheap and abundant precursor in the development of active materials for efficient and environmentally friendly devices like fuel cells. Herein, we investigated iron and nitrogen co-doped nanoporous carbons derived from aronia, peach stones and coal tar pitch/furfural as electrocatalysts for the electrochemical oxygen reduction reaction (ORR) in alkaline media. Urea was used as nitrogen precursor and two annealing steps with intermediate acid leaching served to activate the catalysts. Within the series, the peach stone-derived catalyst exhibited a catalytic activity for the ORR close to the benchmark Pt/C, with a 60 mV dec −1 Tafel slope upon the incorporation of 0.57 wt% Fe and proper combination of N-Fe species (20%) with pyridinic/pyridonic moieties (49%). We concluded that the microporosity and a certain content of meso/macro-pores of the activated carbon, together with the creation of graphitic domains result in a high relative amount of FeN 4 and nitrogen functionalities, which determine the electrocatalytic performance. © 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )


Introduction
Polymer electrolyte fuel cells (PEFC) are considered ideal power sources in a sustainable energy scenario due to their high power density and quasi-zero emissions [ 1 , 2 ]. At these electrochemical devices, the oxygen reduction reaction (ORR) represents the most critical process due to its sluggish kinetics. It takes place thanks to an electrocatalyst at the cathode, which usually consists of noblemetal nanoparticles supported on a carbon material. Electrocatalysts based on platinum or platinum group metals (PGM) present a high activity toward the ORR and an adequate long-term stability [1] . However, these metals are expensive and not abundant, making them not suitable for the widespread commercialization of PEFC in a sustainable situation where non critical raw materials are required. Therefore, the development of active catalysts from highly-abundant and cheap sources remains an ongoing challenge.
Different strategies have been proposed in order to reduce the PGM amount in PEFC, such as alloying with cost-effective metals (e.g., Fe, Co, etc.) [1] or the use of advanced support ma-terials with a developed porosity for the deposition of metal nanoparticles [ 1 , 3 , 4 ]. Although the total Pt requirement has been considerably decreased in the last two decades (present status ~0.4 mg Pt cm −2 at cathode) [5] , the sluggish kinetics of the ORR obstacles a sharper reduction.
The development of PGM-free catalysts has attracted special attention in the last years to overcome these problems [6][7][8]. Non-precious transition metals [ 9 , 10 ], metal oxide/carbide/nitride/phosphide materials [11][12][13] or metal-free carbons [14][15][16] have been studied for ORR. Among these formulations, electrocatalysts based on atomically dispersed transition metals coordinated with nitrogen and carbon (M-N-C, M = Fe, Co, Mn, etc) are the most promising candidates for ORR [17][18][19][20][21]. M-N-C catalysts are typically obtained by the pyrolysis of multiple metal, nitrogen and carbon precursors [ 2 , 6 , 19 ]. The metal-ion center has been found to play an important role on the electrochemical properties of M-N-C catalysts, being Fe and Co the metals with a higher activity for ORR and encouraging stability even in acid media [ 2 , 6 , 22 ]. Regarding the active sites for the ORR, although porphyrin-like M-N 4 structures have been found to provide a high activity, there is still a lack of deep understanding and identification of the catalytic species [23][24][25][26][27] On the other hand, the control of the carbon texture is also essential to obtain highly active M-N-C catalysts. Sophisticated routes involving templates, metal-organic frameworks or coordinated polymers have been investigated to tune the porosity of nitrogen and transition metals co-doped carbons [ 6 , 28-31 ]. Although outstanding performances have been obtained, most of these strategies are expensive and time-consuming, making the catalyst production not feasible. Moreover, nitrogen is traditionally introduced by pyrolysis under NH 3 atmosphere with the consequent toxicity and hazard concerns or by addition of petroleumbased ligands (such as pyridine, imidazole, polyaniline) [ 28-30 , 32 ]. On the other hand, the influence of the carbon precursor and hence of the resulting carbon material of the M-N-C catalytic systems on the performance for the ORR is not fully clear. The physicochemical features of the carbon material affect not only the textural properties but also the surface chemistry, electrical conductivity and morphology of electrocatalysts [33] . Different carbon materials have been used for electrocatalytic applications, such as carbon blacks, carbon nanofilaments, graphene, etc. [ 1 , 4 , 12 ]. However, the preparation of these carbon materials often involves the use of fossil fuel-based precursors (such as phenol, pitch and CH 4 ), toxic reagents or harsh synthesis conditions (e.g., chemical vapor deposition and electric-arc discharge), with the consequent cost increase and environmental impact [34] .
Biomass-derived carbon materials have emerged as a green alternative to produce new electrodes due to their low cost, low toxicity and high abundance [34][35][36][37][38][39][40]. In particular, agriculture byproducts appear to be promising precursors for the production of activated carbons (AC) both from economic and environmental viewpoint. Thermo-chemical processes, such as pyrolysis and gasification, along with fermentation technologies, biodiesel production, etc. have been widely applied for biomass utilization [34] . The chemical characteristics of the lignocellulosic biomass precursors, such as lipids, proteins, hemicellulose, cellulose and lignin content have a significant impact on the formation of carbon porosity. The appropriate choice of the lignocellulosic biomass precursor and the activation procedure results in the formation of activated carbons with tunable surface area and porosity, which is prerequisite for the different applications [ 34 , 35 , 41 ].
In the field of ORR electrocatalysts, recent publications report efficient M-N-C electrocatalysts from different bioresources like soybeans [42] , pomelo peel [43] , or waste reed [32] as examples of these emerging materials. Porous carbons derived from biomass usually present a high content of micropores [34] , which are essential for hosting M-N x species on carbon and thus developing a high density of active sites [ 17 , 44 ]. The valorization of biomass-derived carbons represents thus a promising approach for the synthesis of cost-effective and active M-N-C catalysts for fuel cells. Besides to micropores to anchor the active phase, the presence of mesopores is required to facilitate a fast transport of oxygen in the porous framework [41] . In this regard, a high activity for the ORR has been reported for biomass-derived hierarchical porous carbon materials doped with nitrogen and/or transition metals by different synthesis routes [ 32 , 45-50 ]. Among them, template-assisted procedures (usually using hard-templates, e.g. mesoporous silica) or chemical activation methodologies (often involving harsh chemicals, such as KOH) are the conventional strategies, but the difficult posterior processing limits the scale-up fabrication.
In this work, we provide an effective and additive-free procedure to obtain biomass-derived Fe-N-C electrocatalysts with a remarkable activity for the oxygen reduction reaction. The methodology involves the carbonization and water steam activation of the bioresources, followed by thermal treatment of the resultant nanoporous carbon in the presence of urea and iron nitrate as cheap and abundant nitrogen and metal precursors, avoiding the direct use of ammonia. Two agro-waste precursors are investigated to host iron and nitrogen-based active sites: aronia and peach stones. The approach is also explored for a nanoporous carbon from a mixture of low-rank coals (coal tar pitch) and biomass (furfural), which is of particular interest to address the mining industry activities to greener alternatives in the current energy transition scenario. Up to date, the use of these carbon sources to develop Fe-N-C catalysts for the ORR has not been reported yet. The performance of the catalysts is discussed with regard to their physico-chemical properties, aimed to gather further insights on the desired characteristics of bioresources to promote their valorization into useful catalysts for energy devices.

Synthesis of activated carbons
The activated carbon foam, denoted as CF, was prepared by treatment of a mixture of 75 g coal tar pitch and 75 g furfural with concentrated HNO 3 until the carbon foam formation. Then, the obtained solid product (135 g) was treated at 600 °C (heating rate of 10 °C min −1 ) under nitrogen atmosphere for 30 min. 98 g of the obtained product were activated with water vapour at 850 °C for 1 h. The activated carbons denoted as PS and AR were produced from peach stones and dried aronia residues, respectively, by carbonization of 150 g raw material in nitrogen atmosphere at 550 °C (heating rate of 10 °C min −1 ) for 30 min. 85 g of the obtained product were subjected to activation with water vapour for 1 h at 750 °C and 700 °C, respectively.

Synthesis of Fe-N-C catalysts from activated carbons
The catalysts were synthesized by impregnation of the activated carbons with urea (CH 4 N 2 O, 98%, Sigma Aldrich) as nitrogen source and iron nitrate (Fe(NO 3 ) 3 •9H 2 O, 98%, Alfa Aesar) as metal precursor. Bare activated carbons presented an unsatisfactory hydrophilicity to be well dispersed in water, as required for the introduction of nitrogen and iron species. To improve dispersion in water, each activated carbon was ground to a fine powder in an agate mortar and then it was oxidized with concentrated nitric acid (65 wt.%, Panreac) at 80 °C for 30 min under reflux in a ratio of the carbon mass and the acid volume of 28. The resulting material was thoroughly washed with deionized water and dried overnight at 60 °C. After that, 350 mg of the oxidized carbon material were dispersed by sonication in 50 mL of ultrapure water and added to a second aqueous solution (50 mL) of urea (0.17 M) and Fe(NO 3 ) 3 •9H 2 O (0.05 M). The resultant dispersion was mixed under continuous stirring at 80 °C until evaporation of the solvent. The relative concentration of both precursors was calculated so that the N/Fe atomic ratio was 6, while the total iron amount relative to carbon was 20 wt.%.
The resulting catalyst precursor was subjected to a thermal treatment (T1) at 950 °C for 1 h in N 2 atmosphere (flow rate = 30 mL min −1 ), with a heating rate of 10 °C min −1 . After pyrolysis, an acid leaching procedure (L1) was carried out to remove unstable and inactive iron species using a 0.1 M HClO 4 aqueous solution at 60 °C for 30 min. So-obtained catalysts were washed with deionized water and then dried overnight at 60 °C. A second heat treatment was carried out (T2) under the same conditions (950 °C, 1 h in N 2 atmosphere) for all the catalysts aimed to remove any unstable functional groups generated during acid leaching and to further improve the graphitization of the catalyst [51] . An additional second acid leaching (L2) and third thermal treatment (T3) were performed for the catalyst obtained from dried aronia. The catalysts will be labelled as Fe-N-C followed by the activated carbon precursor (CF, AR or PS) and including the last synthesis step (T1, L1, T2 or T3).

Physical-chemical characterization
Transmission electron microscope (TEM) micrographs were obtained using a Tecnai F30 microscope (300 kV) equipped with a scanning transmission electron microscopy (STEM) module and a high-angle annular dark-field detector. The samples were dispersed in ethanol and a drop of solution was then deposited on a copper grid covered with Lacey carbon.
The iron concentration in the catalysts was determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) using a Jobin Ybon 20 0 0 spectrometer. Elemental analysis (EA) of as-synthesized and oxidized carbon materials were performed in a CHNS-O Analyzer Thermo FlashEA 1112. Thermogravimetric analyses (TGA) were performed on a thermogravimetric SETARAM Setsys Evolution under air atmosphere, varying the temperature from 30 to 950 °C with a heating rate of 5 °C min −1 .
BET specific surface area, total pore volume and pore size distribution of the samples were investigated by low-temperature nitrogen physisorption using a Quantachrome Instruments NOVA 1200e (USA) apparatus. Total pore volume was registered at p/p 0 = 0.99 and V-t-method was applied for micropores volume calculation. Non Localized Density Functional Theory (NLDFT) and Dubinin-Ast akhov methods were used for the elucidation of the meso-and micropores size distribution, respectively.
X-ray diffraction (XRD) patterns were acquired in a Bruker D8 Advance Series 2 diffractometer (USA). The Bragg angle (2 θ ) range scanned was 15-70 °using a counting step of 0.01 °and acquisition time of 4 s. XRD data were fitted using the software TOPAS (Bruker, USA). The interlayer spacings (d 002 ) and the mean crystallite sizes (L c ) were calculated from the position of the peak applying Bragg's law, and using the Scherrer formula with a value of K = 0.89, respectively [52] . The average number of graphene layers (NL) was calculated as (L c /d 002 ) + 1 [53] . X-ray photoelectron spectroscopies (XPS) were acquired on an ESCAPlus OMICRON system equipped with a hemispherical electron energy analyser operating with an Al anode (1486.7 eV) at 225 W (15 mA, 15 kV. Analysis area = 1.75 × 2.75 mm). Survey scans were recorded from 0 to 10 0 0 eV at 0.5 eV step, 0.2 s dwell and 50 eV pass energy. High-resolution spectra were obtained at 0.1 eV step, 0.5 s dwell and 20 eV pass energy. The C1s binding energy (BE) of the graphitic peak was referenced at 284.5 eV for calibration purposes. Shirley type background for all peaks, peak fitting and quantification was performed using CasaXPS software. Gaussian (70%)-Lorentzian (30%) profiles were used for each component.
The transmission Mössbauer spectra were obtained at room temperature with an electromechanical spectrometer (Wis-senschaftlicheElektronik GMBN, Germany) working in a constant acceleration mode. A 57 Co/Rh (activity ∼ = 50 mCi) source and α-Fe standard were used. Samples Fe-N-C-CF T2 and Fe-N-C-PS T2 were measured in a velocity range of ±4.5 mm s −1 , while Fe-N-C-AR T2 was recorded in a velocity range of ±10 mm s −1 . The experimentally obtained spectra were fitted according to the least squares method with the WinNORMOS for Igor 6.37 package. The parameters of hyperfine interaction such as isomer shift ( δ iso ), quadrupole splitting ( E Q ), effective internal magnetic field (hyper-fine field), the linewidths at half maximum (FWHM) and the relative spectral area (A) of the partial components of the spectra were determined.

Electrochemical measurements
Electrochemical measurements were carried out in a three electrode electrochemical cell at room temperature. The working electrode consisted of a thin layer of a catalyst on a rotating disk electrode (RDE) made of glassy carbon (5 mm of diameter). To do this, catalyst inks were prepared by dispersing the catalyst (2 mg mL −1 ) in water and Nafion® dispersion (5 wt.%, Sigma-Aldrich) under sonication. The catalyst loading on RDE was 500 μg cm −2 (100 μg of catalyst) and 15 wt% Nafion®, close to optimum ionomer-tocatalyst ratio for this type of Fe-N-C catalyst according to previous works [54] . A reversible hydrogen electrode (RHE) was used as reference, whereas a high surface area graphite rod was employed as counter electrode. A Metrohm Autolab potentiostat/galvanostat was used.
The electrolyte was a 0.1 M NaOH aqueous solution (80 mL), prepared with 99.99% NaOH (Alfa Aesar) and ultrapure water (milli-Q, 18.2 M cm). First, cyclic voltammograms (CV) were recorded in deaerated electrolyte from 0.05 to 1 V vs. RHE at 0.2 V s −1 until a stable voltammogram was obtained. Afterwards, the electrolyte solution was saturated with oxygen to study the activity of the catalysts for the ORR. Linear sweep voltammetry (LSV) curves were carried out in the potentiostatic mode with a scan rate of 5 mV s −1 and at different rotation rates from 200 to 2500 rpm.
A stability test of the most active Fe-N-C catalyst was carried out. The working electrode was cycled from 0.6 to 1.0 V vs. RHE at 50 mV s −1 for 40 0 0 cycles in the O 2 -saturated 0.1 M NaOH aqueous electrolyte in idle mode [55] . A LSV was recorded every 10 0 0 cycles at 5 mV s −1 and a rotation speed of 1600 rpm to investigate the variation of ORR activity upon stress test.

Physicochemical characterization of activated carbons and Fe-N-C catalysts
Three different activated carbons are investigated in this work. A nanoporous carbon was prepared using treatment products from low rank coals (coal tar pitch) and biomass (furfural). This activated carbon will be labelled as CF. Waste products from canning industry, such as dried fruit residue of aronia and peach stones, were also used as activated carbons precursor. The AC will be denoted as AR and PS, respectively.
Due to the importance of microporosity and surface chemistry of the carbon matrix on the electroactivity of Fe-N-C catalysts [44] , here we present the most remarkable properties of the AC precursors. Table 1 summarizes the most important textural properties of the bare AC, whereas N 2 adsorption-desorption isotherms are depicted in Fig. S1 (supporting information).
The activated carbon obtained from peach stones (PS) presented the highest specific surface area (S BET ) and pore volume (V pore ), followed by the one prepared from coal tar pitch and furfural (CF). All carbon materials possess mixed micro-mesoporous texture with a considerably high contribution of micropores. However, the micropore to meso-macropore volume ratios are of 1.6, 2.3 and 4.0 increasing in the order AR < PS < CF. The latter indicates significant differences in the porosity distribution: the material derived from coal tar pitch and furfural exhibits a more developed microporosity (ca. 80% in terms of pore volume), while the carbon from aronia presents the highest non-microporous contribution (38%). In terms of surface area, predominantly microporous texture is exhibited with more than 95% of area coming from micropores. Fig. 1 shows the pore size distributions (PSD) obtained by NLDFT and Dubinin-Astakhov methods, respectively, from N 2 adsorption data. In the mesopore region (PSD determined by NLDFT), the three carbons present a main maximum at ca. 4 nm, whereas a wide contribution is observed at larger pores (from 4 to 9 nm), specially for the materials derived from aronia and peach stones ( Fig. 1 a ). Pore sizes larger than 10 nm were not obtained in PSD ( Fig. 1 a is reported only up to 16 nm for the sake of clarity). Additionally, PS and AR exhibit a continuous pore distribution over the entire range from 2.5 to 9 nm, which could promote the anchor- Table 1 Textural properties of activated carbons as determined by N 2 physisorption.  ing of iron nanoparticles as will be discussed later. On the other hand, the PSD in the micropore range (Dubinin-Astakhov, Fig. 1 b ) presents differences in the average pore size, increasing in the order CF < PS < AR, with narrow distributions centred at 1.2, 1.6 and 1.9 nm, respectively. To prepare ORR active catalysts, Fe and N species were introduced in the AC structures by impregnation with iron nitrate and urea followed by consecutive stages of thermal treatment and acid leaching, as described in the experimental section. The thermal treatment is aimed to create active sites while acid leaching is carried out to remove inactive and unstable iron species from the catalyst. Catalysts will be labeled as Fe-N-C followed by the biomass precursor (CF, AR or PS) and the last treatment stage (T for thermal treatment, L for acid leaching, and a number standing for the cycle).
The chemical composition of the catalysts was determined by X-ray photoelectron spectroscopy (XPS) and inductively coupled plasma (ICP, Fe content), as described in the experimental section. No other elements than Fe, N, C, and O were detected. The complete set of results is included in Table S1 of the supporting information. Upon the first thermal treatment (T1), the three materials exhibited a high bulk iron content ranging from 26 to 39 wt.%, following the same trend than mesopore volume of activated carbons. This might indicate that the anchoring of iron nanoparticles can be correlated with the amount of the largest mesopores. Metallic iron nanoparticles or iron oxide in the outermost layers of their surface (as evidenced in the high resolution Fe 2p XPS spectra, Fig. S2 ) are not active (or less active than FeN 4 moieties) for ORR as described in the literature [ 56 , 57 ] and will be also shown in the following section. Therefore, acid leaching (L1) resulted in an effective iron removal (inactive and unstable species) with bulk Fe contents from 0.33 to 1.10 wt.% (ICP , Table S1 ). Finally, a second thermal treatment (T2) led to a decrease of oxygen, which may be attributed to the complete removal of the remaining iron oxide after the lixiviation process as revealed in Fe 2p XPS spectra ( Fig. S2 ), with the subsequent creation of new active sites. Consequently, a slight increase of the iron concentration (0.45-1.20 wt.%) was found for Fe-N-C T2 catalysts compared to L1-series. These Fe contents are in the optimal interval for ORR according to the published literature [58] . Such low Fe content is expected in this kind of catalysts, where it is found typically below 1.5 wt.% considering that the amount of atomically dispersed iron coordinated with nitrogen and carbon in micropores relies on the micropore surface area and the geometry of micropores [44] . It is also worth to mention that the atomic N/Fe ratio determined by XPS is found above 2, thus Fe-N 2 and Fe-N 2 + 2 coordinates can be formed together with nitrogen functional groups bonded to carbon and/or oxygen, as will be discussed later.
The morphology of activated carbons and Fe-N-C catalysts was studied by transmission electron microscopy (TEM). Fig. 2 shows a comparison of images from the materials before and upon iron and nitrogen incorporation after two thermal treatments and one acid leaching stage. The three ACs (pictures at the left side in Fig. 2 ) present the typical amorphous structure of activated carbon, as a result of the random aggregation of a porous carbon matrix hosting the micropores.
The pyrolysis at 950 °C in the presence of the iron precursor led to the formation of graphitized ribbons forming a porous structure, which was evident for the catalysts derived from AR and PS in line also with XRD results (see later). The inset of Fig. 2 d and 2 f contain a detail of such ribbons, with an interlayer distance of 0.34 nm, characteristic of ordered graphitic carbon. We did not find evidences of graphitic ribbons in the Fe-N-C catalyst from CF ( Fig.  2 b ) but rather an amorphous carbon structure. There were neither evidences of iron particles in Fe-N-C catalysts, which indicates the effective removal of inactive Fe upon acid leaching. This is also in agreement with the relative low concentration of iron in the catalysts (0.45-1.20 wt.%) and may indicate that Fe is present in atomic dispersion, or at least, small Fe clusters not advisable in TEM. The crystallinity of the three activated carbons, and the Fe-N-C catalysts upon the lixiviation step (L1) and the second heat treatment (T2) steps were investigated by XRD. All the patterns are depicted in Fig. S3 . The diffractograms presented mainly the reflections of the carbon domains at Bragg angles (2 θ ) of about 25-26 °, together with low intensity reflections of different iron species in the case of L1 and T2 catalysts (Fe 3 C, α-Fe and γ -Fe). Neither iron species nor any other crystalline phase were evident in the activated carbons before Fe-N co-doping. Fig. 3 and Table 2 show the peak deconvolution and the resulting structural parameters of the (002) diffraction peak of graphite for the Fe-N-C T2 catalysts, including peak position (2 θ ), integrated intensity (I), graphene interlayer distance (d 002 ), crystallite size in c-axis of carbon (L c ) and the average number of graphenic layers (NL = L c /d 002 + 1). Three different crystalline structures were identified: amorphous carbon (peak I), disordered or turbostratic carbon (peak II) and hexagonal graphite (peak III). Fe-N-C-CF T2 only showed the presence of small-size amorphous carbon domains (L c = 1.1 nm and NL = 4 layers). Conversely, the contribution of the peaks related to turbostratic and hexagonal graphite at 2 θ = 26.2 °and 26.5 °, respectively, was observed for Fe-N-C-AR T2 and Fe-N-C-PS T2. Some ordered carbon domains were also observed in TEM images (see Fig. 2 d and 2 f ). According to the literature, turbostratic and graphitic structures present different in- terlayer spacings of 0.340-0.346 and 0.334 nm, respectively [59] . These results evidence the graphitization of AR and PS during the thermal treatment in the presence of iron nitrate, considering that the activated carbons before treatment present an amorphous structure, see Fig. 2 a , 2 c and 2 e (images at the left) and Fig. S3 (black patterns). To better understand the extent of the formation of turbostratic or hexagonal phases, the percentages of their integrated intensities are included in Table 2 . The relative content of hexagonal graphite is known as graphitization degree by some authors [ 53 , 60 ]. In the present case, the graphitization degree was higher in Fe-N-C-PS T2.
Raman analyses were also carried out on the Fe-N-C catalysts to get more information on the structure of carbon. Raman spectra, depicted in Fig. S4 , show a similar trend in carbon ordering degree like that observed in XRD patterns. The D-band to G-band intensity ratio (I D /I G ) is a parameter typically used to evaluate the ordering degree of carbon materials. I D /I G was 1.04 for Fe-N-C-CF T2 and decreased to 0.94 for Fe-N-C-AR T2 and to 0.79 for Fe-N-C-PS T2 catalyst, indicating an increase of the ordering degree of carbon substrate in the order CF < AR < PS.

Deeper insights on nitrogen and iron speciation of Fe-N-C T2 electrocatalysts
Attending to ORR activity results, the Fe-N-C T2 catalysts presented the most interesting electrochemical behaviour. As a consequence, these will be the focus of discussion in the following paragraphs.
XPS analyses were further performed to examine the nitrogen speciation of the catalysts. The high resolution N1s spectra for the catalysts upon the second heat treatment are included in Fig. 4 . The signals have been deconvoluted into 6 types of N species, centred at 398.5 , 399.7 , 400.9 , 401.9 , 403 and 405.4 eV, corresponding to pyridinic N, N-Fe, pyrrolic/pyridonic N, quaternary N, graphitic N and N oxides (NOx), respectively [ 61 , 62 ]. There is an important contribution of ORR active species (pyridinic N, N coordinated to iron and pyrrolic/pyridonic N) to the N1s signal area in all catalysts. Some differences were found, such as quaternary N representing the most important species for Fe-N-C-CF T2, pyrrolic/pyridonic one in the case of Fe-N-C-AR T2 and pyridinic N for Fe-N-C-PS T2, despite differences in relative content were below 10%.
To better differentiate Fe species with similar structure in different spin and/or oxidation states, Fe-N-C T2 catalysts were characterized by Mössbauer spectroscopy [23] . A combination of sin- Table 2 XRD structural parameters of the Fe-N-C T2 catalysts obtained by deconvolution of the (002) diffraction peak of graphite.

Fe-N-C-CF-T2
Fe-N-C-AR T2 Fe-N-C-PS T2   Table 3 . All Fe-N-C T2 catalysts presented one doublet (D1) with similar isomer shift ( δ iso ~0.39-0.45) and quadrupole splitting ( E Q in the range 0.9-1.2). This doublet is assigned to square-planar Fe II N 4 /C coordination with Fe II in a low-spin state [ 23 , 24 , 26 ]. It is a center consisting of a Fe-ion coordinated by four pyrrolic N-groups hosted to the carbon surface (Fe II N 4 /C). The catalyst obtained from coal tar pitch and furfural presented a second doublet (D2), attributed to Fe II N 4 sites with the iron ion in the Fe II medium-spin state (Fe II N 2 + 2 ) [ 23 , 24 , 26 ].
In addition to FeN x species, Mössbauer spectra evidence the presence of a singlet for all the Fe-N-C T2 catalysts, with an isomer shift of -0.10 mm s −1 , which is assigned to metallic paramagnetic iron ( γ -Fe) [ 25 , 26 ]. Additionally, the catalyst obtained from aronia Table 3 Mössbauer parameters for the three Fe-N-C T2 catalysts from the deconvoluted spectra. Colours indicate the assignments following the pattern of Fig. 5 . presented also two sextet components, Sx1 (Hyper-fine field = 32.6 T) and Sx2 (Hyper-fine field = 20.7 T), which are attributed to α-Fe and iron carbide, respectively [25] .
According to the literature [ 23 , 24 , 26 , 65 ], D1 corresponds to a FeN x C y moiety that is more electrocatalytically active towards ORR than the species corresponding to D2. Additionally, D2 exhibits an anomalous large linewidth at half maximum (FWHM , Table 3 ), indicating a distribution of local environments around the iron ions [ 23 , 65 ]. The latter causes an overlap of several doublets having nearly the same parameters of isomer shift and quadrupole splitting. As a result, the second doublet in the Fe-N-C-CF T2 can be attributed to a distorted ferrous FeN 4 center: (i) out-of-plane position of iron, (ii) the bonding to nitrogen belonging to two adjacent graphene layers, and/or (iii) an inhomogeneous chemical environment in the periphery of the FeN 4 center. Table 3 includes also the relative content (%) of iron species assuming the same Lamb-Mössbauer factor (A) and the Fe mass concentration from ICP analyses. The relative concentration of FeN 4 (low spin) increases from 0.17 to 0.25 wt.% in the order AR < PS < CF. On the other hand, despite the acid leaching procedure, there is still a significant amount of Fe not coordinated with nitrogen that has resisted the etching step.

Oxygen reduction electrochemical activity of Fe-N-C catalysts
The electrocatalytic activity of the Fe-N-C catalysts for the oxygen reduction reaction (ORR) was evaluated by means of linear sweep voltammetry (LSV) in an oxygen saturated 0.1M NaOH solution. The LSV curves for the catalysts at 1600 rpm are included in Fig. 6 . The three curves for each activated carbon-based catalyst correspond with the formulation upon the different synthesis stages (T1, L1 and T2). All the curves present an initial increase of negative current corresponding to ORR, from a certain value of potential (onset potential) and following an exponential variation of current with overpotential, according to the Butler-Volmer equation. The current density then reaches a more or less defined  plateau, associated to the limiting diffusion current, resulting in the sigmoidal form of the LSV curve for ORR.
The introduction of iron and nitrogen species within the activated carbon structure resulted in an effective strategy for the preparation of active catalysts toward the ORR. The acid leaching (L1) and the second thermal treatment (T2) led to a significant increase of activity, indicating both the removal of inactive species and the creation of more active sites. This is particularly evident in terms of the ORR onset potential for the three catalyst formulations, with an improvement of over 200 mV from T1 to T2, regardless the activated carbon nature. In terms of limiting current density, the positive effect of acid leaching and thermal treatments is more evident in the case of activated carbons from aronia and peach stones.
Wu and coworkers observed an improvement of current density upon a second heat treatment for polyaniline-based Fe-N-C catalysts, but interestingly, the authors did not observe a change in the onset potential between the acid leaching and the thermal treatment [66] . They attributed the activity improvement to an increase of the density of active sites, but maintaining their nature (same intrinsic activity or turnover frequency). The positive effect of a second thermal treatment has also been reported for catalysts based on mesoporous carbons [57] , on polyaniline-derived carbon black [67] and multiwalled carbon nanotubes [68] . The removal of unstable and inactive iron species together with the formation of active ones are pointed out as the main cause for the observed increase of ORR performance. In our results, the significant enhancement of potential between L1 and T2 clearly indicates the formation of new active sites upon the second heat treatment.
An additional acid leaching and thermal treatment (T3) was carried out in one of our Fe-N-C catalysts in order to assess a possible activity enhancement. Fig. S5 of the supporting information includes the ORR activity for Fe-N-C-AR T3 compared to the catalyst T1-, L1-and T2-derived formulations. Although there is an improvement of ORR activity at low overpotential, the curves superimpose at larger current densities, suggesting that a third treatment stage is not as beneficial as the previous ones.
Comparing the LSV curves of the catalysts after two thermal treatments and one acid lixiviation stage (T2), the nature and properties of the activated carbon precursor appears to have a very relevant impact on the ORR activity ( Fig. 7 ). Interestingly, ACs obtained from biomass materials (AR and PS) presented a better ORR activity in terms of half-wave potential (i.e. the potential at half the limiting current density) than that derived from coal tar pitch and furfural.
Indeed, Fe-N-C-PS T2 presents an onset potential very close to that of a commercial Pt/C catalyst (20wt.% Pt), only 20 mV far in terms of half-wave potential (E 1/2 ) and with almost the same limiting current density. ORR activity here discussed is in line with previous publications in which biomass has been considered as precursor for Fe-N-C catalysts, such as soybeans [42] , pomelo peel [43] or waste reed [32] . These results highlight the suitability of using biomass, here in particular from peach stones, as an excellent precursor for low-cost and highly active catalysts for electrochemical applications.
The analysis of Tafel plot is shown in Fig. 8 for the AC-based Fe-N-C catalysts. Significant variations of Tafel slope ( b in graph) and exchange current density ( j 0 in graph) were identified within the three catalyst formulations (upon second annealing). The Tafel slope gives information about the reaction mechanism, and decreases from 154 to 60 mV dec −1 going from CF > AR > PS. A decrease in Tafel slope denotes a change of the rate determining step (rds) in ORR towards a more effective pathway. According to theoretical calculations [69] and considering an associative mechanism occurring in alkaline environment (see Supplementary material for the whole mechanism description), a Tafel slope b = 60 mV dec −1 is associated with the following reaction being the rds: where * denotes an active site; while b = 120 mV dec −1 can be attributed to either of the following reactions being the rds: * OO + e − * OO − (120 mV dec −1 ) Therefore, in the catalyst Fe-N-C-PS T2, the reaction is governed by Eq. (1 ) , with 60 mV dec −1 ; this is also the case of Pt/C catalysts [15] as well as active PGM-free catalysts in alkaline medium [10] . Whereas, a mixed situation with sites reacting in agreement with Eqs. (1 ) and (2 ) is envisaged for Fe-N-C-AR T2 catalyst, with 75 mV dec −1 . Eq. (3 ) is discarded as the rds because in such case a Tafel slope of 40 mV dec −1 should be obtained at low overpotential according to Shinagawa et al. [69] , who describe a large coverage of * OO species governing the ORR, which is not our case. The largest Tafel slope was observed for Fe-N-C-CF T2, which is reflected in the lower current density obtained in the polarization curve ( Fig. 7 ). Another relevant difference from Tafel plot arising from AR and PS catalysts is the exchange current density (j 0 ), with almost two orders of magnitude higher value for the most active catalyst. This parameter indicates the readiness of the catalyst to proceed with the electrochemical reaction, being set as the equilibrium current at zero overpotential (1.23 V vs. RHE in this case). Figure 9 shows the Koutecky-Levich plot for the different Fe-N-C catalysts. The exchanged number of electrons has been determined by applying the Koutecky-Levich, Eq. (4) : where j is the measured current density, j k is the kinetic current density, j d is the diffusion limiting current density (all current density are expressed in mA cm −2 ), B is a constant parameter depending on the measurement conditions, n is the number of electrons and ω is the RDE rotation rate (rad s −1 ). From Levich equation, the B factor is known to depend on Eq. (5) : where F is the Faraday's constant (96,485 C mol −1 ), C O 2 is the concentration of oxygen in the electrolyte, D O 2 is the diffusion coefficient of oxygen in the solution and ν is the solution kinetic viscosity. Considering the Pt/C commercial catalyst as a reference for a 4e − pathway, the value of B was calculated to be 0.0813 mA s 0.5 cm −2 . The number of electrons of Fe-N-C catalysts was thus calculated from Koutecky-Levich plots and is included in Fig. 9 . The catalysts prepared from biomass waste like aronia and peach stones present values of number of electrons closer to 4 (about 3.5 e − ) than the one from coal tar pitch and furfural (3 e − ). These numbers indicate that the reduction of oxygen is proceeding by a mix pathway between 2e − and 4e − route. In this sense, Artyushkova and coworkers suggested different reaction pathways depending on the nature of the active sites through a systematic approach [61] . Nitrogen species like pyridinic or pyrrolic were identified to favour 2e − route while Fe-N x centers catalyse either 4e − path or 2 × 2e − , depending on the exact nature of iron coordination. Varnell and collaborators also found that Fe particles encapsulated by N-doped carbon do also contribute to ORR activity [70] .
The ORR activity differences encountered with the choice of carbon precursor were quite large and significant, as shown in Table 4 in terms of half-wave potential (E 1/2 ), number of electrons or Tafel slope. Interestingly, the ORR activity of our catalysts did not rely on the concentration of N and/or Fe active species (i.e. Fe-N 4 , pyridinic or pyrrolic N), thus there must be additional reasons to explain the different behaviour. The most noteworthy differences were found to be the occurrence of graphitic planes in the case of AR-and PS-derived catalysts (see Figs. 2 and 3 ) which were not detected in CF-related ones, as discussed before. CF presented also a pure type I nitrogen adsorption/desorption isotherm ( Fig.  S1 ), indicating absence of mesopores, while AR-and PS-activated Table 4 Summary of ORR electrokinetic parameters and relative concentration of FeN 4 and nitrogen ORR active species from Mössbauer and XPS deconvolution, respectively.

Parameter
Fe-N-C-CF T2 Fe-N-C-AR T2 Fe-N-C-PS T2  carbons exhibited a certain amount of adsorbed nitrogen at partial pressure above 0.4, with a wide mesopore size distribution. The lowest performance of Fe-N-C-CF T2 might be mainly correlated to the absence of such graphitic domains upon thermal treatment. Mesopores in the activated carbon are most probably favouring the anchorage of iron particles in the first synthesis step, facilitating the graphitizing effect of iron on carbon substrate. Indeed, iron concentration from ICP analyses ( Table S1 , supporting information) was higher in AR (33 wt.%) and PS (39 wt.%) than in CF (26 wt.%), confirming this hypothesis. Therefore, active sites are most probably forming on the edges of graphenic planes where the charge transfer is largely favoured compared to amorphous carbon matrix [ 71 , 72 ]. In accordance with these results, the most active catalyst (Fe-N-C-PS T2) exhibited the higher contribution of graphite crystalline structures ( Fig. 3 ). Concerning the two best formulations, the catalyst from peach stones exhibited higher activity than the one from aronia, with a potential over 60 mV better ( Fig. 8 , 110 mV if E 1/2 is considered, Table 4 ), which is a very significant difference in ORR activity. Even though the concentrations of Fe and N are higher in Fe-N-C-AR T2 catalyst, it appears that it is not the total but the relative content of active species that determines the ORR behaviour. Assuming that active sites are in the form of Fe-N coordination on one hand, and pyridinic/pyrrolic/pyridonic species on the other hand [61] , Fe-N-C-PS T2 catalyst presents the largest relative amount for all of them ( Table 4 ). This aspect points out the high importance of the occurrence of neighbouring active sites for a proper electrocatalytic activity in case of 2 × 2 e − pathway, as it has been concluded to take place in this investigation.
To assess the stability of the most active catalyst studied in this work (Fe-N-C-PS T2), an accelerated stress test (AST) was performed based on continuous cycling from 0.6 to 1.0 V vs. RHE. Dominant degradation mechanisms of Fe-N-C catalysts which may take place at this potential window are: i) the demetallation of FeN 4 moieties by H 2 O 2 radicals (E < 0.8 V vs. RHE) and; ii) the loss of active species by carbon corrosion, mainly occurring at E > 0.9 V vs. RHE [ 55 , 73 , 74 ]. Fig. 10 shows the ORR polarization curves before and upon gradual potential cycling. A negative shift in the half wave potential is observed up to 30 0 0 cycles, with the most significant performance fading upon the first 10 0 0 scans. Upon 40 0 0 cycles, the potential loss in the E 1/2 reaches values of 53 mV and the diffusion current density decreases 17%. A similar performance loss has been reported for the state-of-the-art Pt/C catalyst subjected to comparable degradation tests [75] .

Conclusions
An effective strategy for the preparation of active ORR catalysts from biomass, urea and iron nitrate is presented as non-critical raw alternative material catalyst for PEFC cathodes. The methodology takes advantage of the use of biomass-derived materials, such as peach stones and aronia, as carbon precursor for the preparation of nanoporous carbons. N and Fe species are introduced in the carbon structure by a route comprising thermal treatment of the porous carbons in the presence of available and cost-effective sources (urea and iron nitrate) and subsequent acid etching. The as-synthesized catalysts are active for the ORR with a better performance than their analogue obtained from coal tar pitch and furfural precursors. Interestingly, the catalyst obtained from peach stones, presented a performance close to that obtained with a commercial Pt/C catalyst (state-of-the-art catalyst). The graphitic structure created during annealing and the porosity of the activated carbon play a key role in determining the activity of the derived Fe-N-C catalyst, with iron coordinated to nitrogen and nitrogen functionalities as the active species.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.