P)-Doped ... - ACS Publications

Research Article pubs.acs.org/acscatalysis

Fe-Treated Heteroatom (S/N/B/P)-Doped Graphene Electrocatalysts for Water Oxidation Fatemeh Razmjooei,† Kiran Pal Singh,† Dae-Soo Yang, Wei Cui, Yun Hee Jang,* and Jong-Sung Yu* Department of Energy Systems Engineering, DGIST, Daegu 42988, Republic of Korea S Supporting Information *

ABSTRACT: Anodic water splitting is driven by hydroxide (OH−) adsorption on the catalyst surface and consequent O2 desorption. In this work, various heteroatoms (S/N/B/P) with different electronegativities and oxophilicities are introduced to alter the catalytic activity of reduced graphene oxide (RGO) as a catalyst for the oxygen evolution reaction (OER). It is found that, surprisingly, S-doped RGO outperforms the other RGOs doped with more electropositive or electronegative and more oxophilic heteroatoms, and this effect becomes more prominent after Fe treatment of the respective catalysts. Herein, we evaluate the OER activity of a series of Fe-treated mono-heteroatom (S/N/ B/P)-doped RGO (Fe-X-G) catalysts, among which interestingly S-doped RGO catalyst treated with Fe (Fe-S-G) is found to show better OER activity than the well-known active Fe-N-C catalyst, demonstrating the best activity among all of the prepared catalysts, close to that of the state of the art IrO2/C catalyst, along with pronounced long-term stability. Density functional theory (DFT) calculations indicate that the OER activity highly depends on the electroneutrality and oxophilicity of doped heteroatoms and doping-induced charge distribution over RGO, demonstrating that S with mediocre electronegativity and the least oxophilicity exhibits optimal free energy for the adsorption of the OER intermediate and desorption of the final OER product. Furthermore, it is found that Fe treatment greatly helps in enhancing the number of active sites through the regeneration of reduced catalytically active S sites and improving the conductivity and surface area of the S-doped RGO, which are found to be key factors to furnish the Fe-S-G catalyst with the capability to catalyze the OER with high efficiency, even though Fe is found to be absent in the final catalyst. KEYWORDS: heteroatom doping, iron, oxygen evolution reaction, reduced graphene oxide, electrocatalysis

1. INTRODUCTION Water electrolysis has huge implications on future technological advancements. While the cathodic-side hydrogen evolution reaction (HER) is kinetically facile, the anodic-side oxygen evolution reaction (OER), in which hydroxide (OH−) ions developed at the cathode are utilized at the anode to produce oxygen and water (4OH− ↔ O2 + 2H2O + 4e−), is challenging due to large overpotential and hence requires active noblemetal catalysts such as RuO2 and IrO2.1−21 The high price, instability, and scarcity of these noble metals have impeded their large-scale implemenatation.1−21 Various transition-metal oxides and chalcogenides, whose surface metal (M) active sites can adsorb a series of OER intermediates (e.g., M−OH, M−O, M−OOH, and M−OO), have been investigated as alternative OER electrocatalysts.22−27 However, as the electrochemical performance of catalysts is directly related to their electrical conductivity, the inherent low conductivity of these oxide catalysts is a key drawback that seriously impairs their catalytic potency and limits their practical application.28 Recently, nitrogen (N)-doped carbonaceous materials have shown high OER performances,15,28−31 demonstrating that carbon materials with proper physical or chemical modifications can also be efficient OER catalysts. These types of OER catalysts have © 2017 American Chemical Society

evoked strong interest due to their low cost, high electrical conductivity, and unique structural features, but their OER performance still needs to be improved to meet expectations. Adsorption and desorption of intermediates on catalysts directly affect the reaction kinetics of the catalysis,15 and thus a key content would be to develop an OER catalyst that can efficiently adsorb the OER intermediates and easily desorb final products at the same time. The OER on heteroatom-doped carbon material surfaces is generally known to proceed through a sequence of four different electron/proton transfer steps, where an electron (or proton) is transferred in each step (Scheme 1).32 In particular, since OH− adsorption at the first step (I) dictates the OER activity and the optimal C−OH ads bond strength is a requisite for high OER activity,14,15 the OER activity of carbonaceous catalysts can be governed by the reactivity of OH− with the catalyst surface. It is believed that a facile adsorption of water oxidation intermediates (OH−, OOH−) on the positively charged carbon atom (C(δ+)) Received: November 18, 2016 Revised: February 5, 2017 Published: February 16, 2017 2381

DOI: 10.1021/acscatal.6b03291 ACS Catal. 2017, 7, 2381−2391

Research Article

ACS Catalysis

in deciding the OER performance, and these properties of B and P render them OER inactive. Density functional theory (DFT) calculations can also make great contributions to such developments and evaluations.35 Apart from the adsorption and desorption kinetics of reactant and product, respectively, the physicochemical properties of the catalyst are also important parameters to ponder upon. Additional treatment of the N-doped carbon with nonprecious transition metals (M) such as iron (Fe) and cobalt (Co) is found to have a significant effect on the OER performance, which is governed by N content, M content, M−N interaction, and conductivity of the resulting carbon framework.36−38 One particularly intriguing point is that, while such transition metals certainly enhance the conductivity and create extra active sites in the carbon framework,39 their physical presence on the resulting carbon catalyst does not necessarily enhance the OER activity of the N-doped carbons.15 Our interpretation of the role of the transition metals is that they collect the oxygens (O) from the neighboring N sites to form metal oxides, which are washed away by acid washing, and thus convert inactive (oxidized and occupied by O) NO or N−O sites into active N catalytic sites. On the other hand, it has been reported that a decrease in the amount of framework oxygen helps to increase the conductivity of the material,40 consequently improving OER activity. Although OER catalysts based on non-preciousmetal-derived N-doped carbon have been reported, to the best of our knowledge, non-precious-metal-treated carbon materials modified with other heteroatoms with different electronegativities (or oxophilicities) such as S, B, or P remain untouched for the OER. Hence, in this work we have prepared a series of monoheteroatom (S/N/B/P)-doped reduced graphene oxides (RGOs) in the presence of Fe to synthesize Fe-X-G, where X represents each heteroatom. For comparison, catalysts in the absence of Fe have also been prepared and designated as X-G. The resulting series of Fe-X-G electrocatalysts for the OER have been investigated for the first time in this work. Surprisingly, Fe-S-G shows better OER activity than the wellknown active Fe-N-C catalyst, demonstrating the best activity among all the prepared catalysts, where S is less oxophilic and electronegative than N. Fe treatment greatly increases the number of active sites and the conductivity of the Fe-S-G, which are found to be key factors in furnishing the Fe-S-G catalyst with the capability to catalyze the OER with high efficiency, even though Fe is found to be absent in the final catalyst. It is also worth mentioning that Fe-S-G catalyst shows high OER activity, which is very close to that of the state of the art IrO2/C catalyst, along with pronounced long-term stability. To better understand the underlying mechanism of the OER on various Fe-X-G catalysts, DFT calculations are also applied to study catalysts for water oxidation.

Scheme 1. OER Mechanism in Alkaline Medium

adjacent to electronegative N decreases the activation energy of OER in N-doped carbonaceous catalysts.33,34 On the other hand, overly strong adsorption of such intermediates on the Ndoped carbon may hinder the desorption and evolution of the final product O2 (last step, IV) and eventually lower the overall OER performance. As a consequence, the oxygen evolution kinetics depends strongly on the surface nature of the catalyst, and to intelligently design a catalyst, it is crucial to extrapolate which immanent material characteristics control OER catalysis. In this regard, oxophilicity (or oxygen affinity), that is, the reactivity and binding strength of an element toward oxygen, is a useful concept in interpreting a wide range of catalytic processes, including the catalytic electrolysis of water. A simple generic scale of oxophilicity has recently been determined over the whole periodic table by Kepp.35 More detailed systemspecific oxophilicity values of element X beyond the Kepp scale35 was determined by density functional theory (DFT) calculations on bond dissociation energies of X−O in this work. We expect that replacing the N dopant by other heteroatoms such as sulfur (S), phosphorus (P), and boron (B) with lower electronegativities (N (3.04) > S (2.58) ≈∼ C (2.55) > P (2.19) > B (2.04)) and/or different oxophilicities (B (1.0) > P (0.7) ≈ N (0.7) > S (0.5))35 would result in a variation in OH− adsorption strength and O2 evolution rate, and this in turn will give differences in the OER performance. In addition, developing novel materials by using other heteroatoms besides N is also of great significance in terms of amplifying the spectrum of new catalytic materials and their various applications. Since N is more electronegative than S, the adjacent carbon atoms become more positively charged in Ndoped carbon in comparison with those of S-doped carbon. Therefore, the adsorption of hydroxide ion (OH−) with a negative charge occurs more easily and strongly on N-doped than on S-doped carbon. This strong adsorption of reaction intermediates can make desorption of the final reaction product, O2, more difficult in N-doped carbon in comparison with the carbon doped with other heteroatoms. On the other hand, B and P, which are much more electropositive than carbon, can provide electrons to the adjacent C, leading to an increase in its nucleophile strength (C(δ−)), which makes carbon-adjacent B and P species inefficient in adsorbing OH− with negative charge, resulting in direct adsorption of OH− on B and P atoms with positive charge. Since B and P with positive charge are highly oxophilic, the adsorption of the negative intermediate species will occur strongly on B or P sites, which in turn will make the desorption of the final product, O2, more difficult in B- or P-doped carbon in comparison with that of Nor S-doped carbon. Hence, it can be surmised that, apart from the electrophilicity, the oxophilicity also plays a prominent role

2. EXPERIMENTAL SECTION 2.1. Synthesis of Graphene Oxide (GO). GO was synthesized from natural flake graphite (325 mesh, 99.8%, Sigma-Aldrich) using an improved Hummers method.41 In brief, a 9/1 (360 mL/40 mL) mixture of concentrated H2SO4/ H3PO4 was added to 3.0 g of graphite flake and 18.0 g of KMnO4. The oxidation was then carried out by heating and stirring the solution mixture at 50 °C for 12 h.41 After the oxidation reaction, the mixture was poured into a mixed solution of ice (∼400 mL) and 30% H2O2 (3.0 mL). The obtained solution was centrifuged for 1 h at 4000 rpm, the 2382


Research Article

ACS Catalysis

two metallic probes placed in the middle of the powder sample. A Keithley Model 6220 and Model 2182A were used as the dc current source and voltmeter, respectively. The current was varied from 0 to 10 mA, and the corresponding voltage was measured. The electrical conductivity of the samples was estimated using the formula

supernatant was decanted away, and the brown powder left behind was collected. The obtained powder was then washed repeatedly with a mixture of ethanol, HCl, and water and finally dried overnight at 60 °C to collect a dried GO sample. A 200 mg portion of dry GO was dispersed in 200 mL of water by sonication and used for further synthesis. 2.2. Preparation of Fe-Treated Heteroatom-Doped RGO Catalysts. For the synthesis of Fe-treated heteroatomdoped RGO samples, different acid family members including sulfuric, nitric, boric, and phosphoric acids were used for the S, N, B, and P precursors, respectively. In a typical synthesis of Fetreated heteroatom-doped RGO, 200 mL of a GO (1.0 mg mL−1) water dispersion was mixed with 2 mL of the respective acid and 0.1 g of FeCl2·4H2O with vigorous stirring at room temperature for 5 h. The mixture was kept at 80 °C in an oven for 10 h to collect a dried sample. The remaining solid was ground and pyrolyzed at 800 °C for 2 h with a ramp rate of 5 °C/min under an N2 atmosphere. To remove excess iron, the obtained catalyst was treated with 0.5 M sulfuric acid at 80 °C for about 8 h, rinsed with deionized water, and finally dried at 60 °C. The resulting catalysts are hereafter denoted as Fe-X-G, where X corresponds to each different heteroatom (S/N/P/B). Heteroatom-doped RGO (X-G) was also prepared by the same procedure without using an iron precursor. Pristine RGO was prepared by direct pyrolysis of dried GO alone using identical pyrolyzing conditions. 2.3. Characterization. The morphology of the samples was observed with a Hitachi S-4700 scanning electron microscope (SEM) operated at an accelerating voltage of 10 kV and with an EM 912 Omega transmission electron microscope (TEM) operated at 120 kV. X-ray diffraction (XRD) data were acquired by a Rigaku Smartlab diffractometer with Cu Kα radiation (1.5406 Å) operated at 40 kV and 40 mA. Raman spectra were recorded with a Renishaw spectrometer using an Ar+ ion laser (λ 514.5 nm). X-ray photoelectron spectroscopy (XPS) was conducted to analyze surface elemental compositions of all samples using an ESCALAB-250 spectrometer with a monochromated Al Kα (150 W) source. The nitrogen adsorption−desorption isotherms were measured at −196 °C using a Micromeritics ASAP 2020 accelerated surface area and porosimetry system. The specific surface area was determined on the basis of the Brunauer−Emmett−Teller (BET) method from nitrogen adsorption data at a relative pressure range between 0.05 and 0.2. The total pore volume was determined from the amount of gas adsorbed at the relative pressure of 0.99. Pore size distribution (PSD) was derived from adsorption branches by the Barrett−Joyner−Halenda (BJH) method. Thermal gravimetric analysis (TGA) was carried out on a Bruker TG-DTA3000SA thermal analyzer at a heating rate of 5 °C/min in flowing air (60 mL/min), increasing from room temperature to 1000 °C to investigate the decomposition course of the materials. Electrical conductivity measurements for all of the samples were carried out using a homemade fourpoint probe apparatus by varying the applied pressure according to previous works.42−44 The cell was made of a nonconducting Teflon block carved into a hollow cylinder, which was covered by two metallic brass pistons, one as a base and another as a lid, to which the pressure was applied. A finely powdered carbon sample was filled in the hollow Teflon chamber, which in turn was sealed using two brass pistons, and resistivity measurements were performed by increasing the applied pressure. Current was applied to the sample through the metallic pistons, and the voltage was measured across the

σ=

l RA

where σ is the electrical conductivity, R is the resistance of the sample, A is the cross-sectional area of the sample (0.126 cm2), and l is the distance between the voltage probes (0.2 cm). While the resistivity was measured, pressure was applied to the metallic pistons by using steel plates of known weights. Theoretical calculations based on DFT were carried out on small model compounds at the level of B3LYP/6-311G** implemented in Jaguar v8.5 (Schrödinger LLC, New York, 2014). 2.4. Electrode Preparation and Electrochemical Characterization. The working electrode was polished with alumina slurry to obtain a mirrorlike surface, then washed with water and acetone, and dried before use. The slurry was prepared by mixing 5.0 mg of catalyst added with a 1.0 mL solvent mixture of Nafion (5 wt %) and water in a v/v ratio of 1/9 for 20 min in an ultrasonicator. For comparison, a commercially available catalyst, 20 wt % IrO2/C, was used, and a 1.0 mg/mL commercial IrO2/C suspension was prepared according to a procedure identical with that described above. The slurry was placed on a precleaned working electrode, and the electrode was allowed to dry at room temperature before the measurement. This led to a catalyst loading of 0.4 μg cm−2 for all of the obtained catalysts or commercial 20 wt % IrO2/C. All electrochemical measurements were carried out with a Biologic VMP3 electrochemical workstation. All obtained catalysts and commercial 20 wt % IrO2/C were used directly as the working electrode without further treatments. OER activities of the electrodes were characterized by OER polarization curves, electrochemical impedance spectroscopy (EIS), and chronoamperometry. Electrochemical studies of the samples were carried out at room temperature using a rotatingdisk electrode (RDE) in a three-electrode geometry. A glassycarbon RDE was used as the working electrode, and Pt wire and Ag/AgCl saturated with KCl were used as counter and reference electrodes, respectively. The Ag/AgCl reference electrode was calibrated with respect to a reversible hydrogen electrode (RHE). The calibration was performed in a highpurity H2 (99.99%) saturated electrolyte using a Pt electrode and Pt wire as the working and counter electrodes, respectively. Linear scanning voltammetry (LSV) was then run at a scan rate of 10 mV/s, and the potential at which the current crossed zero was taken to be the thermodynamic potential (vs Ag/AgCl) for the hydrogen electrode reactions. In 0.1 M KOH, ERHE = EAg/AgCl + 0.973 V. Calibration of the reference electrode against the RHE is shown in Figure S1 of the Supporting Information. The OER activity was evaluated by linear sweep voltammetry (LSV) on a rotating disk electrode with a rotation rate of 1600 rpm and a scan rate of 5 mV s−1. The impedance of each catalyst was measured by electrochemical impedance spectroscopy (EIS) over a frequency range of 1 × 105 to 0.1 Hz with a sinusoidal perturbation amplitude of 5 mV. The turnover frequency is defined as the rate of evolved molecular O2 per surface active site per second, which can be calculated by the equation 2383


Research Article

ACS Catalysis

facile OH− adsorption and O2 desorption kinetics suggest more favorable OER kinetics on the S-doped RGO. It is quite surprising to see that S-G shows better OER activity than the well-known active N-G. From the above results it can be concluded that the OER activity decreased in the order S-G > N-G > P-G > B-G. It will be interesting to find a relationship between OER activity and the surface properties of X-G samples. Interestingly, the OER activity matches well with the oxophilicity of doped heteroatoms. To investigate the effect of Fe treatment on the OER performance of the catalysts, polarization curves for Fe-S-G, FeN-G, Fe-B-G, and Fe-P-G catalysts in 0.1 M KOH solution with a scan rate of 5 mV s−1 were also obtained along with that of 20 wt % IrO2/C, as shown in Figure 1a. The potential required for

TOF (s−1) =

(number of oxygen turnovers)/(cm 2 geometric area) (number of active sites)/(cm 2 geometric area)

TOF (s−1) = (number of oxygen turnovers)/(number of active sites) = (J/4F)/n, where J is the OER current density at a given overpotential, n is the number of active sites, and F is the Faraday constant (96584.3 s A mol−1). J/4F represents the total oxygen turnover. The overpotential used for the calculation of the TOF was set at 480 mV, in which a current density of 10 mA cm−2 resulted for Fe-S-G.

3. RESULTS AND DISCUSSION The schematic illustration of the synthesis process is depicted in Scheme 2. While the doping conditions and dopant Scheme 2. Schematic Illustration of Various Fe-Treated Heteroatom-Doped RGOs

Figure 1. (a) LSV profiles and (b) Tafel plots of Fe-X-G catalysts and 20 wt % IrO2/C. (c) EIS Nyquist plots measured for all the prepared Fe-X-G catalysts. (d) Chronoamperometric response at a constant potential of 10 mA cm−2 for Fe-S-G and 20 wt % IrO2/C. The inset in (d) gives the LSV plots for the 1st and 800th potential cycles for Fe-SG.

concentrations were kept identical, different acid families such as sulfuric, nitric, boric, and phosphoric acids were used as S, N, B, and P precursors, respectively. Briefly, all of the Fe-treated heteroatom-doped RGOs (Fe-S-G, Fe-N-G, Fe-B-G, and Fe-PG) were prepared by pyrolyzing a mixture of FeCl2 as an iron precursor, each respective acid as a heteroatom precursor, and GO as a substrate and carbon precursor at 800 °C (see the Experimental Section) followed by acid washing to remove residual unstable Fe species. In this case, it is found that acid functionalities also help in enhancing the surface area of the resulting Fe-X-G samples. The OER activity of all the resulting Fe-X-G catalysts was assessed by LSV in 0.1 M KOH electrolyte at room temperature (25 °C). OER polarization curves for the resulting Fe-free samples S-G, N-G, B-G, and P-G at 1600 rpm are shown in Figure S2a of the Supporting Information. Since the Tafel slope provides insight into the reaction mechanism, we compare the Tafel plots of the Fe-free heteroatom-doped RGO catalysts. As shown in Figure S2b of the Supporting Information, the S-G catalyst exhibits a much smaller Tafel slope (151 mV per dec) in comparison with N-G (158 mV/ dec), P-G (171 mV/dec), and B-G (183 mV/dec) catalysts. These results suggest more favorable OER kinetics for S-G in comparison with other counterparts. It can be seen that, among all the of catalysts, S-G outperforms in terms of onset and the water oxidation current and also shows much smaller Tafel slope in comparison with the other X-Gs, which indicates the

water oxidation to obtain the current density of 10 mA cm−2 is commonly used to judge the OER activity.14,24 As can be concluded from Figure 1a, interestingly, Fe-S-G generates an OER current density of 10 mA cm−2 at a much lower potential of 1.70 V vs RHE, in comparison with the other counterparts Fe-N-G (1.75 V), Fe-B-G (1.89 V), and Fe-P-G (1.95 V), which is considerably closer to that of IrO2/C (1.60 V). The impact of S doping and iron treatment on the OER can be elucidated from Figure S3a of the Supporting Information. S doping helps to significantly improve the OER activity of RGO in the S-G sample along with greater catalytic activity than its counterparts. Particularly, this is highly significant since Ndoped carbons are among the most studied metal-free substitutes for noble-metal catalysts for the OER. We also find that a simple Fe treatment further improves the activity of the S-G in Fe-S-G. This could expedite the larger scale use of Sdoped carbon-based OER catalysts. Furthermore, as shown in Figure 1b, the Fe-S-G catalyst exhibits a much smaller Tafel slope (88 mV per dec) in comparison with the other Fe-X-G catalysts. These results suggest more favorable OER kinetics (OER intermediate adsorption and desorption) for Fe-S-G in comparison with 2384


Research Article

ACS Catalysis

collective data confirm that the Fe-S-G is highly efficient toward the OER. As physical parameters have a huge effect on the catalytic properties of OER catalysts, various analyses have been carried out on Fe-X-G materials. The XRD patterns of all the asprepared Fe-X-G samples before acid washing are shown in Figure S6a,b of the Supporting Information. It can be seen that, before acid washing, Fe-S-G and Fe-N-G possess typical Fe3O4 peaks, whereas Fe-B-G and Fe-P-G show peaks for Fe3BO5 (ICDD card no. 01-078-2285) and Fe2P4O12 (ICDD card no. 01-073-1945), respectively. Interestingly, as shown in Figure 2a,

other counterparts. A similar trend was observed in Figure S3b of the Supporting Information, where Fe-S-G exhibits a much smaller Tafel slope in comparison with that of its Fe-free counterpart (S-G), suggesting the synergistic effect of Fe treatment on the S-G catalyst. Figure 1c and Figure S3c of the Supporting Information show Nyquist plots based on EIS analysis. The smaller charge transfer resistance of Fe-S-G in comparison to the other Fe-X-G catalysts and its Fe-free counterpart (S-G) clearly indicates its better electronic transport capability and better reactant diffusivity toward the electrode surface, which clearly justify the superior OER performance of Fe-S-G. Furthermore, Fe-S-G demonstrates a strong durability in alkaline electrolytes with only 25% anodic current loss during ∼800 continuous potential cycles (inset of Figure 1d). As shown in Figure 1d, the high stability of Fe-S-G for the OER is also proved by chronoamperometric methods, exhibiting significant current retention of ∼67% after continuous operation at a constant potential of 1.70 V at 10 mA cm−2 over 30000 s, in contrast to the rapid activity decrease of IrO2/C (52% retention) at a constant potential of 1.60 V. The stability of the catalysts was measured at the constant potential required for water oxidation to obtain a current density of 10 mA cm−2 for each sample, this potential being different for each sample. The potential required for a current density of 10 mA cm−2 is 1.70 V for Fe-S-G and 1.60 V for IrO2/C as mentioned. The EIS measurements were done at the potential near to the onset of the OER to understand the mechanism of intermediate adsorption on the Fe-X-G catalysts and their Fe-free counterparts, as shown in Figure S4a,b of the Supporting Information. It can be seen in Figure S4a of the Supporting Information that the charge transfer resistance for the Fe-N-G catalyst is much lower than those of the other counterparts in the order Fe-N-G < Fe-S-G < Fe-B-G < Fe-P-G. This clearly indicates the facile adsorption of OER intermediates (−OH−) at the Fe-N-G catalyst and proves the point that the highly electropositive surface of Fe-N-G due to N doping facilitates the −OH− adsorption. Even though the adsorption of the intermediate is quite facile on Fe-N-G, its final OER activity is less than that of the Fe-S-G, which can be justified from the oxophilicity of the catalyst. As the oxophilicity of S is lower than that of N, final oxygen desorption is much facile on Fe-S-G in comparison with Fe-N-G catalyst. Therefore, Fe-S-G was assumed to have the most optimal M−O bond strength among the other X-G species, suggesting facile adsorption of OH− and O2 desorption on the active sites of the catalyst as the rate-limiting step. The same trend is observed for their Fe-free counterparts, as shown in Figure S4b of the Supporting Information. In order to make meaningful comparisons of activity trends, it is critical to compare the intrinsic activity of all of the catalysts on the basis of specific activity and TOF. Specific activity versus TOF is plotted in Figure S5 of the Supporting Information for all of the samples. As shown in Figure S5 of the Supporting Information, Fe treatment and sulfur doping have led to reduced graphene oxide achieving turnover frequencies as high as 0.34 s−1 at an overpotential of 480 mV for O2 production at high specific activity (0.71 mA cm−2 BET). A comparison of turnover frequencies and activities for all of the catalysts shows that the TOF and specific activity associated with Fe-S-G are higher than those of Fe-N-G (0.30 s−1 and 0.12 mA cm−2 BET) ,Fe-B-G (0.28 s−1 and 0.02 mA cm−2 BET), and Fe-P-G (0.26 s−1 and 0.03 mA cm−2 BET). These

Figure 2. (a) XRD patterns and (b) Raman spectra for all of the Fe-XG catalysts.

after acid washing, with the exception of Fe-P-G, the other FeX-G catalysts show no metal complex impurities. Generally, the diffraction profiles of reduced graphene oxide based materials exhibit two prominent broad bands centered around 2θ = 25.6 and 43°, associated with diffraction of the (002) and (100) planes, respectively, which are characteristic of turbostratic carbon. However, as shown in Figure S7 of the Supporting Information, in the early region within the range of 2θ = 21− 30°, XRD patterns for Fe-S-G, Fe-N-G, and Fe-B-G show the presence of more developed and separated peaks with maxima at 2θ = 21.9 and 25.5° for Fe-S-G, 21.5 and 25.3° for Fe-N-G, and 21.0 and 25.2° for Fe-B-G, respectively. In this present study, different acid family members including sulfuric, nitric, boric, and phosphoric acids are used as S, N, B, and P precursors for the synthesis of Fe-S-G, Fe-N-G, Fe-B-G, and Fe-P-G, respectively, which at the same time can act as chemical activation agents to increase the surface area of graphene-based materials. It has been proposed that, in chemical activation using acid, carbon materials show richer diffraction profiles at lower angles with diffraction peaks at ∼25.6 and ∼21.0°, which correspond to mean d002 spacings of 3.39 and 3.85 Å, respectively.45 This observation can be attributed to fragmented crystallites due to the use of acid families as activation agents, which leads to the formation of disorders and defects in the graphene-based materials. As can be seen in Figure S7 of the Supporting Information, both peaks, which correspond to the (002) peak for Fe-N-G and Fe-B-G, shift toward the lower angle in comparison to that of Fe-S-G, which implies that Fe-N-G and Fe-B-G possess larger interlayer spacing between graphene sheets.33 This means that Fe-S-G possesses a more intact graphitic structure in comparison to other Fe-X-Gs. In the case of Fe-P-G, since Fe2P4O12 is a chemically stable and insoluble complex, it remains in the Fe-PG carbon lattice even after acid washing.46−48 To further justify our claim regarding the complete removal of Fe after acid washing, we have opted for the carbon 2385


Research Article

ACS Catalysis

temperature, volatilization of acidic groups can lead to the gasification of carbon and distortion of the graphitic structure, which can generate pores and wrinkles, resulting in increased surface area. A further surface area increase seen after incorporation of Fe in Fe-S-G is mainly due to the removal of iron oxides, formed during the high-temperature pyrolysis process, by acid washing. As shown in Figure S11c,d of the Supporting Information, PSD curves based on the BJH model exhibit a mesoporous nature for all of the samples. Surface characteristics for all of the samples are summarized in Table S1 of the Supporting Information. To investigate the topological features and properties of all of the prepared catalysts, SEM and TEM measurements were carried out as shown in Figure 3a−h and Figures S12a−h and

oxidation method. On the basis of the concept that carbon is oxidized at a higher temperature under an oxygen atmosphere, whereas Fe oxidizes to its oxide forms (the only product left over), we have oxidized the Fe-S-G sample before and after acid washing (Fe-S-G-BW-OX and Fe-S-G-OX). For comparison we also prepared a sample washed for a shorter period of time (2 h) (Fe-S-G-2 h W), which was further oxidized to give Fe-S-G2 h W-OX. Figure S8a of the Supporting Information shows the TGA analysis of washed and unwashed Fe-S-G samples as well as Fe-S-G-2 h W in atmospheric air. It can be seen that oxidation of carbon starts at ∼450 °C and finishes at around 700 °C, indicating complete removal of carbon up to 700 °C. In the washed sample (Fe-S-G), 100% removal of samples was observed, implying that the washed sample is devoid of any Fe impurities. On the other hand, unwashed Fe-S-G-BW and Fe-SG-2 h W samples show only 90% and 93% degradation, respectively, which implies the presence of Fe impurities in these samples. To identify the nature of the species obtained after carbon oxidation, XRD analysis was carried out. Figure S8b of the Supporting Information shows the XRD patterns for Fe-S-G-BW and sample washed for a shorter period (2 h) after oxidation (Fe-S-G-BW-OX and Fe-S-G-2 h W-OX), which show typical peaks for Fe2O3 (ICCD card no. 01-073-3825). As can be seen in Figure S9 of the Supporting Information, S-G shows a more crystalline nature and has a sharper (002) diffraction peak in comparison with the other counterparts. It is clear that S doping significantly keeps the graphitic structure of graphene carbon more intact in comparison with the other counterparts. The comparative XRD plots of graphite, GO, RGO, and S-G are shown in Figure S10a of the Supporting Information. Figure 2b compares the Raman spectra of all of the Fe-X-G samples. The ratios (ID/IG) of the D (1352 cm−1) to the G band (1580 cm−1), which is commonly cited as a measure of structural disorder in a graphitic structure, are found to be 1.14, 1.20, 1.26, and 1.18 for Fe-S-G, Fe-N-G, Fe-B-G, and Fe-P-G, respectively. This clearly indicates that Fe-S-G possesses a more crystalline and ordered structure with high graphitic nature in comparison to other Fe-X-G samples, which is in agreement with XRD results. A new band at ∼1050 cm−1 is seen in the Fe-P-G sample, which are due to the symmetric PO43− stretching mode associated with the PO43− tetrahedron and can be ascribed to the presence of Fe2P4O12.47 The comparative Raman spectra of RGO (ID/IG = 0.97) and S-G (ID/IG = 1.12) are also shown in Figure S10b of the Supporting Information. Surface properties such as surface area and pore distribution are critical parameters for electrochemical catalysis. Nitrogen isotherms display type IV adsorption/desorption behaviors with a pronounced hysteresis loop typical of mesoporosity for the prepared materials, as shown in Figure S11a of the Supporting Information. The BET surface areas were determined to be 212.24, 300.81, 307.58, and 354.36 m2 g−1 for Fe-P-G, Fe-S-G, Fe-N-G, and Fe-B-G, respectively. The higher surface areas of Fe-S-G, Fe-N-G, and Fe-B-G in comparison with that of Fe-P-G could be due to the removal of iron residues in the process of acid washing. As shown in Figure S11b and Table S1 of the Supporting Information, with the introduction of S into the pristine RGO and incorporation of Fe into the S-doped RGO, the BET surface area starts increasing from 185.12 m2 g−1 for pristine RGO to 240.10 m2 g−1 for S-G and to 300.81 m2 g−1 for Fe-S-G. Such an increase in surface area in S-G can be ascribed to the pore-generating ability of the acidic groups of sulfuric acid. At high pyrolysis

Figure 3. SEM images of (a) Fe-P-G, (c) Fe-S-G, (e) Fe-N-G, and (g) Fe-B-G and TEM images of (b) Fe-P-G, (d) Fe-S-G, (f) Fe-N-G, and (h) Fe-B-G.

S13a−j of the Supporting Information. SEM and TEM images of all Fe-treated samples before acid washing are shown in Figure S12a−h of the Supporting Information. Images show the presence of some particles with different sizes formed on the RGO surface during the annealing process. These particles are attributed to different iron oxide compositions generated under different conditions. The obtained samples were treated with 0.5 M sulfuric acid to remove excess iron complexes before OER catalysis. Figure 3a,b for Fe-P-G shows that particles with different sizes still remain on the graphene surface, which can be attributed to Fe2P4O12 complexes, as shown in the XRD pattern in Figure 2a. However, no such iron complexes are observed on the graphene surface for the other Fe-X-G catalysts after acid washing. In addition, the SEM and TEM images in Figure S13c−j of the Supporting Information and Figure 3c−h reveal the evolution of wrinkled patterns in all of the X-G and Fe-X-G catalysts in comparison with RGO, which can be ascribed to heteroatom doping in the RGO framework as well as removal of iron complexes after mild acid washing. It is worth mentioning that the graphene sheets in Fe-S-G show a less disordered structure with fewer wrinkles, which implies a much more preserved graphitic structure in Fe-S-G in comparison with that of Fe-N-G and Fe-B-G, which is in line with the XRD and Raman data in Figure 2. From the above discussions, it can be concluded that Fe-S-G possesses a less disordered structure in comparison with other prepared catalysts. Therefore, it can be expected that the conductivity of Fe-S-G should be better than that of the other counterparts.49 To evaluate this, the electrical conductivities of the Fe-X-G, S-G, and RGO samples were examined using a 2386


Research Article

ACS Catalysis

nitrogen, NOx (∼402.4 eV) in Figure 4b.33 The pyridinic N, which is known to be the OER active site, is found to be the dominant N state in the Fe-N-G sample.15 Even though N is more electronegative than C, its effect largely depends on its doping configuration. As reported, the quaternary and pyrrolic N atoms in graphene can provide electrons to the p-conjugated system, which can increase nucleophilic strength for the adjacent carbon rings (C(δ−)), making the carbon atoms near the quaternary and pyrrolic-N energetically unfavorable for adsorption of water oxidation intermediates (OH− and OOH−) in alkaline solution and consequently unfavorable for the OER. However, pyridinic N, which is an electron-withdrawing group with lone pair electrons, inductively removes electron density from the adjacent carbon (C(δ+)), which facilitates the adsorption of water oxidation intermediates on the adjacent carbon (C(δ+)), hence accelerating the OER.26 It is worth mentioning that, in the case of sulfur, the S3 peak is found to be absent in Fe-S-G, showing the increased C−S bond formation (Figure 4a), and in the case of N the intensity of the N1 and N4 species increases and decreases, respectively, in comparison with that of N-G (Figure 4b). As can be seen in Tables S2 and S3 in the Supporting Information, it can be noted that, after Fe treatment, the oxygen content significantly drops down from 3.27 for S-G to 1.33 for Fe-S-G and from 4.03 for N-G to 2.69 for Fe-N-G, which can be attributed to the removal of formed Fe oxide species by acid washing. These results clearly demonstrate that introduction of Fe in S-G and N-G to form Fe-S-G and Fe-N-G can collect O from the neighboring sites (S3 and N4), to establish Fe3O4 oxides (Figures S6 and S12 of the Supporting Information), which can be etched out by acid washing. Thus, this Fe treatment helps to convert inactive oxidized sites to active reduced sites, decreases O content in the Fe-S-G, increases the conductivity (Figure S15 of the Supporting Information), and in turn improves the OER performance. However, Fe collects oxygen from the SOx more easily than NOx, which might be due to the lower oxophilicity of S in comparison with that of N. The B 1s XPS spectra for B-G and Fe-B-G are shown in Figure 4c. B is presented in the form of (B1) BC3 (∼191.2 eV), (B2) partially oxidized B (BC2O) (∼191.9 eV), (B3) BCO2 (∼193.2 eV),57 and (B4) B2O3 (193.7 eV).58 However, even though the B4 peak is found to be absent in Fe-B-G due to an iron reaction to form the Fe3BO4 complex, which can be etched out by acid washing, it is still found that B2 and B3 are the dominant phases of the doped B in Fe-B-G.59,60 This may be evidence for the high amount of oxygen present in Fe-B-G (Table S2 of the Supporting Information), which is not removed from the B-doped carbon due to the high oxophilicity of B. Thus, Fe-B-G, due to the low amount of active species such as B1 in the carbon framework and low conductivity due to the high amount of oxygen (see Figure S14b of the Supporting Information), may lead to lower OER activity in comparison with Fe-S-G and Fe-N-G. Figure S17 of the Supporting Information shows the relative distribution of prominent heteroatom species for Fe-S-G, Fe-NG, and Fe-B-G catalysts and their Fe-free counterparts. It can be seen that, due to the lower oxophilicity of the sulfur, Fe treatment can effectively etch out the oxygens present in the vicinity of sulfur, which is in agreement with the complete disappearance of the S3 peak in Fe-S-G sample in comparison to S-G sample. In the case of nitrogen, due to higher oxophilicity of N, Fe treatment is found to only partially decrease the oxygen content, and hence only a partial decrease

homemade four-probe apparatus (Figures S14a,b and S15 of the Supporting Information).42,50 It can be seen that, as predicted, Fe-S-G presents higher conductivity in comparison with the other catalysts and S-G.51 Fe-P-G shows the lowest conductivity among all the Fe-X-G catalysts, which could be due to formation of nonconductive Fe2P4O12 species on a graphene surface.46 The peak survey and atomic contents of all the samples evaluated using XPS are shown in Figure S16a−d and Tables S2 and S3 of the Supporting Information. Interestingly, with the exception of Fe-P-G, the XPS survey scans indicate the absence of Fe species in Fe-S-G, Fe-N-G, and Fe-B-G frameworks, showing the complete removal of Fe complexes after the acid washing. The S contents for Fe-S-G (1.83 wt %) and S-G (1.71 wt %) are found to be almost identical, indicating that Fe has almost no effect on S doping concentration in the carbon framework, as shown in Tables S2 and S3 of the Supporting Information. In addition, interestingly, the results show that the O content is reduced with a corresponding increase in the carbon content for Fe-S-G in comparison with S-G. The S 2p XPS spectra for S-G and FeS-G are shown in Figure 4a. As can be seen in Figure 4a, the S

Figure 4. Deconvoluted XPS spectra revealing (a) relative distribution of prominent S species for S-G and Fe-S-G, (b) N 1s for Fe-N-G, (c) B 1s for Fe-B-G, and (d) P 2p for Fe-P-G.

2p narrow scan spectra for S-G and Fe-S-G catalysts are fitted by three peaks at 163.61, 164.82, and 168.21 eV corresponding to S p3/2 (S1) and S p1/2 (S2) for C−S−C and SOx (S3). The S1 and S2 peaks can be attributed to sulfur bonded directly to the carbon atoms in a heterocyclic configuration, whereas the S3 peak at 168.21 eV is assigned to −C−SOx− species.40,26,52−54 It is reported that a sulfur dopant in the form of thiophene with a five-membered-ring structure is active for the OER.55 Since Fe-S-G exhibits prominent OER activity, we also excogitate that these C−S−C species exist in the thiophene form, which is consistent with previously reported observations.56 In the present Fe-S-G, the intensity of the S1 and S2 species, which are considered to be active for the OER, increases in comparison with that of S-G.55 The N 1s XPS spectra for N-G and Fe-N-G are shown in Figure 4b. As can be seen in Figure 4b, deconvolution of the N 1s spectra of N-G and Fe-N-G results in four major peaks, assigned to (N1) pyridinic (∼398.3 eV), (N2) pyrrolic (∼399.4 eV), (N3) quaternary (∼401.1 eV), and (N4) oxides of 2387


Research Article

ACS Catalysis

To better understand the underlying mechanism of the OER on various doped graphene sheets, a DFT investigation was applied to study Fe-X-G as catalysts for water oxidation. It appears that the O contents (Table S2 of the Supporting Information) in the Fe-X-G catalysts are well correlated with their OER performance. The O contents should be related to the oxophilicity of the heteroatom X in X-G.35 For a more specific understanding of the oxophilicity of the X-G catalysts, we used DFT to calculate the O affinity (ΔE1) and the O2 desorption energy (ΔE2) of X-G, which are defined by eqs 1 and 2:

in N4 species is observed. Similar observation stands true for BG and Fe-B-G as well. After studying the high resolution XPS of the prepared catalysts following order of oxophilicity was observed (B > N > S), which is also similar to our theoretical predicted order. As shown in Figure 4d, the P 2p spectrum for P-G and Fe-PG is deconvoluted into two peaks at 132.1 and 133.3 eV assigned to (P1) P−C and (P2) P−O (oxidized P), respectively. It has been also reported that the binding energies in the range of 132.9−133.9 eV are attributed to Fe phosphates such as the Fe2P4O12 complex.61,62 In the present case with FeP-G, the formation of chemically stable and insoluble Fephosphate complexes is also highly possible, which is in line with XRD, SEM, and TEM analysis and can be gleaned from the decreased P1 and enhanced P2 signal intensity for Fe-P-G in comparison to that of P-G. This indicates that the major P2 species, with high O content, can decrease the conductivity of Fe-P-G. The role of oxygen on the carbon surface is important in catalysis. It is reported that a high amount of oxygen (Table S2 of the Supporting Information) can reduce the conductivity, consequently reducing the OER activity. The O content is related directly to oxophilicity of the heteroatom (B (1.0) > P (0.7) ≈ N (0.7) > S (0.5)).35 Due to the high oxophilicity of B, the oxygen content is rather high in Fe-B-G in comparison with that of Fe-N-G and Fe-S-G. Therefore, the high oxygen content in Fe-B-G agrees with its low conductivity observed in Figure S14b of the Supporting Information, eventually leading to lower OER activity in comparison with Fe-S-G and Fe-N-G. In the case of Fe-P-G, the formation of the Fe2P4O12 complex, which is a chemically stable and insoluble complex, can decrease the conductivity, consequently reducing the OER performance. As shown in Scheme 1, the proposed mechanism of the OER for heteroatom-doped carbon in alkaline media consists of the adsorption of OH− in the first step (I) and desorption of O2 in the last step (IV). The high OER activity of Fe-S-G in comparison with active Fe-N-G is intriguing and worth investigating. Due to the higher electronegativity of N in comparison to S, neighboring carbon atoms become more positive in Fe-N-G in comparison to those of Fe-S-G, resulting in stronger adsorption of OH− and more difficult desorption of the final product (O2), which eventually may lead to lower OER activity in Fe-N-G in comparison with that of Fe-S-G. In the case of Fe-B-G and Fe-P-G, in which B and P are much more electropositive than carbon, a negative charge will be induced on a carbon atom, which makes carbon adjacent to B and P species inefficient in adsorbing OH− with negative charge, resulting in adsorption of OH− directly on B and P with positive charge. Thus, B or P may act as active sites to adsorb OER intermediates, but as is often the case, the strong adsorption due to their high oxophilicity leads to much more difficult O2 desorption; thus the OER performance is not high. Along this line, as B is more electropositive than P, B-G shows lower OER performance in comparison with P-G, as observed in Figure S2 of the Supporting Information. However, interestingly, with the introduction of Fe, Fe-P-G shows lower OER activity in comparison with Fe-B-G, which can be attributed to the formation of chemically stable and nonconductive Fe2P4O12 species. In addition, since the major B and P species in Fe-B-G and Fe-P-G are mainly present as inactive oxide species (see Figure 4c,d), the OER activity should be lower than that expected.

ΔE1 = E[O/X‐G] − E[X‐G] − 0.5E[3O2 ] 3

ΔE2 = E[ O2 ] + E[X‐G] − E[O2 /X‐G]

(1) (2)

A more negative value of ΔE1 represents a higher O affinity (that is, greater stabilization after O adsorption), while a more positive value of ΔE2 represents a higher O2 affinity (that is, higher cost of O2 desorption). The X-G was modeled by the smallest graphene fragments with a heteroatom sitting on the edges in Figure 5a−c. A thiophene-like five-membered-ring

Figure 5. (a) Model DFT calculations to select the most stable O (red) adsorption site among several plausible ones (1−4) on graphene edges doped with S (yellow), N (blue), and B (green) on the basis of the relative O affinity (in kcal/mol; in parentheses). (b) O affinity ΔE1 (in kcal/mol) of the selected O adsorption sites with the atomic charges (in electron units) of the dopants shown together. (c) O2 desorption energy ΔE2 (in kcal/mol) from the same sites with the O− O distance (in Å) shown together.

model (S-G in a box; Figure S18 of the Supporting Information, upper) was used for the S-doped edge of graphene (S-G) and a pyridine-like six-membered-ring model (N-G and B-G in a box; Figure S18 of the Supporting Information, lower) was used for the N- and B-doped edges of graphene (N-G and B-G) in order to satisfy stable closed-shell electron configurations. Indeed, a previous experimental work63 has assigned similar types of models for S-G, that is, exclusively thiophene-like S doping, from their high-resolution XPS spectra, which are essentially identical with ours (Figure 4a).The total energy E of each species was estimated after 2388


Research Article

ACS Catalysis

and more difficult desorption of final product (O2), which eventually may lead to lower OER activity of Fe-N-G in comparison with that of Fe-S-G. As acid washing is considered to be a necessary step to remove inactive species for improving the OER activity, to estimate the variation of OER activity before and after acid washing of the Fe-S-G sample, we characterized the Fe-S-G sample before acid washing as well. As shown in Figure S19 and Table S4 of the Supporting Information, XPS characterization of the Fe-S-G sample before acid washing (Fe-S-G-BW) shows the presence of higher amounts of Fe and O, which might be evidence for the presence of metal oxide (Fe3O4) impurities (as shown in the XRD pattern in Figure S6a of the Supporting Information), which are etched out after acid washing (Figure 2a). As shown in Table S5 and Figure S20a,b of the Supporting Information, the BET surface area and the conductivity of the catalysts are found to improve after acid washing. The higher surface area of Fe-S-G in comparison with that of Fe-S-G-BW can be due to the removal of iron residues during the process of acid washing. Since it has been reported that a decrease in the amount of framework oxygens helps to increase the conductivity of the material, the lower conductivity of Fe-SG-BW in comparison with Fe-S-G could be due to the presence of nonconductive Fe3O4 on the catalyst surface and its higher oxygen content. Figure S21 of the Supporting Information shows a comparison of the OER activity of Fe-S-G samples before and after acid washing (Fe-S-G-BW and Fe-S-G). The OER activity of Fe-S-G is improved after acid washing in comparison with Fe-S-G-BW due to the removal of iron oxide impurities. Therefore, it is interesting to note that, even though Fe is not detected after acid washing as shown in Figure S16a−c and Table S2 of the Supporting Information, the sample treated with Fe significantly improves OER activity in comparison with the Fe-free counterpart due to an increase in the amount of active sites as well as surface area and improvement in graphiticity and electrical conductivity of catalysts. This implies the eminent role played by Fe in preparing an efficient OER catalyst, which is in direct agreement with earlier work by Zhao et al.15

full geometry optimization performed at the level of B3LYP/6311G** of DFT implemented in Jaguar v8.5 (Schrödinger, LLC, New York, 2014). Each optimized geometry was confirmed to be the minimum-energy structure using normalmode analysis. The spin-unrestricted DFT was used for the triplet ground state of O2, and the spin-restricted DFT was used for the singlet ground states of all other species. The most stable O-adsorbed and O2-adsorbed X-G species (used for the estimation of ΔE1 and ΔE2 in Figure 5b,c) were selected among various final structures optimized from different initial structures (Figure 5a). The S and N dopants (position 3), which are (slightly for S) more electronegative than C (N (3.04) > S (2.58) > C (2.55)), are (slightly for S) negatively charged (−0.08 and −0.67 electron unit (|e|) from the electrostatic-potential-fitted atomic charge estimation, as shown in Figure 5b) and therefore unfavorable (ΔE1 = 5.0 kcal/mol on S) or only slightly favorable (ΔE1 = −3.4 kcal/mol on N) for O adsorption (Figure 5a). Instead, the positively charged adjacent C site (position 1) serves as the most favorable O adsorption site (ΔE1 = −38.5 and −51.5 kcal/mol for Fe-S-G and Fe-N-G, respectively, in Figure 5a,b). On the other hand, the B dopant (position 1, Figure 5a), which is more electropositive than C (C (2.55) > B (2.04)), would be positively charged (0.78 |e| in Figure 5b) and serves as the most favorable O adsorption site (ΔE1 = −59.9 kcal/mol as shown in Figure 5a,b). The O affinities to the most stable site in this series of X-G are therefore estimated as −59.9, −51.5, and −38.5 kcal/mol for B-G > N-G > S-G. This is indeed consistent with the reported oxophilicity scale of these heteroatoms (B > N > S).35 Therefore, Fe in Fe-X-G after collection of O from neighboring X sites to form different iron oxide compositions, which are washed away by acid washing, will convert the oxidized (occupied and inactive) sites into reduced (unoccupied and active) X sites, decrease the O content in X-G, enhance the conductivity of X-G, and eventually improve the OER performance of X. This effect is more prominent for S-G than for N-G. On the other hand, doping with oxophilic B or P would have an opposite effect. The higher O affinity of the catalytic sites in B-G or P-G than in N-G and S-G would result in higher resistance to restoration of the reduced (unoccupied and active) sites even after the Fe treatment, as evidenced by the observation of a high amount of oxygen in the respective samples, and thus lower the conductivity and consequently lower the OER performance.35 The oxophilicity is quite general in the sense that the same trend of affinity of X-G (B-G > N-G > S-G) holds for O2 as well as for O on the same site (Figure 5c). The energy costs ΔE2 of O2 desorption from S-G, N-G, and B-G, which is the last step of the OER, were calculated to be 15.9, 28.3, and 53.5 kcal/mol, respectively. In the case of B-G, the initial step of O(H) binding may be very favorable (ΔE1 = −59.9 kcal/mol), but the final step of O2 desorption would be extremely difficult (ΔE2 = 53.5 kcal/mol). On the other hand, in the case of N-G and especially S-G, while the initial step of O(H) binding is still favorable on a carbon adjacent to heteroatom sites (ΔE1 = −51.1 and −38.5 kcal/mol), the final step of O2 desorption is much easier from N-G and S-G (ΔE2 = 28.3 and 15.9 kcal/mol) than from B-G. Such a balance in the adsorption−desorption energetics could also be responsible for the superior OER performance of Fe-SG over Fe-(N/B/P)-G. Summarizing, due to the higher electronegativity of N in comparison to S, neighboring C atoms become more positive in Fe-N-G in comparison to that of Fe-S-G, resulting in stronger adsorption of OH− on Fe-N-G

4. CONCLUSIONS A series of Fe-treated mono-heteroatom (S/N/B/P)-doped RGOs was prepared and investigated for the first time for their OER electrocatalytic activity. It is found that the new Fe-S-G catalyst is very effective in catalyzing oxygen evolution with catalytic activity better than not only that of its metal-free counterpart but also that of other Fe-treated heteroatom (N/ B/P)-doped RGOs prepared in this work. Furthermore, Fe-S-G showed pronounced long-term operational stability. This superior activity can be attributed to the high graphiticity, high surface area, low oxophilicity of S, high conductivity, high amount of active sites, and optimal C−Ox adsorption− desorption energetics in the Fe-S-G framework. Interestingly, it is also found that the physical presence of Fe may not be necessary to enhance the OER activity of the investigated Fe-XG. Fe can efficiently collect the oxygen from the neighboring sites to form iron oxides, which are washed away by acid washing and, hence, can thus render the catalyst to have many more active sites, increased surface area, and improved conductivity, and therefore is required for the preparation of an efficient OER catalyst. Furthermore, theoretical calculations indicate that the OER performance of various Fe-X-G catalysts 2389


Research Article

ACS Catalysis

(12) Gao, Y.; Zhao, H.; Chen, D.; Chen, C.; Ciucci, F. Carbon 2015, 94, 1028−1036. (13) Liu, Q.; Jin, J.; Zhang, J. ACS Appl. Mater. Interfaces 2013, 5, 5002−5008. (14) Subbaraman, R.; Tripkovic, D.; Chang, K.-C.; Strmcnik, D.; Paulikas, A. R.; Hirunsit, P.; Chan, M.; Greeley, J.; Stamenkovic, V.; Markovic, N. M. Nat. Mater. 2012, 11, 550−557. (15) Zhao, Y.; Nakamura, R.; Kamiya, K.; Nakanishi, S.; Hashimoto, K. Nat. Commun. 2013, 4, 2390. (16) Lu, X.; Zhao, C. Nat. Commun. 2015, 6, 6616. (17) Zhao, Z.; Li, M.; Zhang, L.; Dai, L.; Xia, Z. Adv. Mater. 2015, 27, 6834−6840. (18) Surendranath, Y.; Kanan, M. W.; Nocera, D. G. J. Am. Chem. Soc. 2010, 132, 16501−16509. (19) Burke, M. S.; Kast, M. G.; Trotochaud, L.; Smith, A. M.; Boettcher, S. W. J. Am. Chem. Soc. 2015, 137, 3638−3648. (20) McCrory, C. C. L.; Jung, S.; Peters, J. C.; Jaramillo, T. F. J. Am. Chem. Soc. 2013, 135, 16977−16987. (21) Danilovic, N.; Subbaraman, R.; Chang, K.; Chang, S. H.; Kang, Y. J.; Snyder, J.; Paulikas, A. P.; Strmcnik, D.; Kim, Y.; Myers, D.; Stamenkovic, V. R.; Markovic, N. M. J. Phys. Chem. Lett. 2014, 5, 2474−2478. (22) Ganesan, P.; Prabu, M.; Sanetuntikul, J.; Shanmugam, S. ACS Catal. 2015, 5, 3625−3637. (23) Wang, J.; Gao, D.; Wang, G.; Miao, S.; Wu, H.; Li, J.; Bao, X. J. Mater. Chem. A 2014, 2, 20067−20074. (24) Zhao, Y.; Chen, S.; Sun, B.; Su, D.; Huang, X.; Liu, H.; Yan, Y.; Sun, K.; Wang, G. Sci. Rep. 2015, 5, 7629. (25) Trotochaud, L.; Young, S. L.; Ranney, J. K.; Boettcher, S. W. J. Am. Chem. Soc. 2014, 136, 6744−6753. (26) Yang, H. B.; Miao, J.; Hung, S. F.; Chen, J.; Tao, H. B.; Wang, X.; Zhang, L.; Chen, R.; Gao, J.; Chen, H. M.; Dai, L.; Liu, B. Sci. Adv. 2016, 2, e1501122. (27) Cheng, Y.; Jiang, S. P. Prog. Nat. Sci. 2015, 25, 545−553. (28) Li, R.; Wei, Z.; Gou, X. ACS Catal. 2015, 5, 4133−4142. (29) Chen, A.; Duan, J.; Jaroniec, M.; Qiao, S. Z. Adv. Mater. 2014, 26, 2925−2930. (30) Zhang, J.; Zhao, Z.; Xia, Z.; Dai, L. Nat. Nanotechnol. 2015, 10, 444−452. (31) Vezzù, K.; Delpeuch, A. B.; Negro, E.; Polizzi, S.; Nawn, G.; Bertasi, F.; Pagot, G.; Artyushkova, K.; Atanassov, P.; Noto, V. D. Electrochim. Acta 2016, 222, 1778−1791. (32) Handoko, A. D.; Deng, S.; Deng, Y.; Cheng, A. W. F.; Chan, K. W.; Tan, H. R.; Pan, Y.; Tok, E. S.; Sow, C. H.; Yeo, B. S. Catal. Sci. Technol. 2016, 6, 269−274. (33) Razmjooei, F.; Singh, K. P.; Song, M. Y.; Yu, J.-S. Carbon 2014, 78, 257−267. (34) Yu, X.; Zhang, M.; Chen, J.; Li, Y.; Shi, G. Adv. Energy Mater. 2016, 6, 1501492. (35) Kepp, K. P. Inorg. Chem. 2016, 55, 9461−9470. (36) Shui, J. L.; Karan, N. K.; Balasubramanian, M.; Li, S. Y.; Liu, D. J. J. Am. Chem. Soc. 2012, 134, 16654−16661. (37) Bayatsarmadi, B.; Qiao, S. Z. Proceedings of the World Congress on New Technologies 2015, 348, 1−8. (38) Kauffman, D. R.; Alfonso, D.; Tafen, D. N.; Lekse, J.; Wang, X. D.; Lee, J.; Jang, H.; Lee, J. S.; Kumar, S.; Matranga, C. ACS Catal. 2016, 6, 1225−1234. (39) Yang, D.-S.; Song, M. Y.; Singh, K. P.; Yu, J.-S. Chem. Commun. 2015, 51, 2450−2453. (40) Singh, K. P.; Song, M. Y.; Yu, J.-S. J. Mater. Chem. A 2014, 2, 18115−18124. (41) Marcano, D. C.; Kosynkin, D. V.; Berlin, J. M.; Sinitskii, A.; Sun, Z.; Slesarev, A.; Alemany, L. B.; Lu, W.; Tour, J. M. ACS Nano 2010, 4, 4806−4814. (42) Razmjooei, F.; Singh, K. P.; Yu, J.-S. Catal. Today 2016, 260, 148−157. (43) Chaudhari, N. K.; Song, M. Y.; Yu, J.-S. Sci. Rep. 2014, 4, 5221. (44) Chaudhari, K. N.; Song, M. Y.; Yu, J.-S. Small 2014, 10, 2625− 2636.

tightly depends on the oxophilicity of heteroatoms and dopinginduced charge distribution in the doped carbon framework. In all, the present findings are highly important and innovative with respect to developing new OER catalysts and understanding the underlying principle. This study establishes fundamentals to understand the factors which control the OER activity in heteroatom-doped carbon catalysts and, therefore, demonstrates a new direction toward the development of a new series of cost-effective and efficient catalysts.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.6b03291. Experimental procedures and characterization data (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail for Y.H.J.: [email protected]. *E-mail for J.-S.Y.: [email protected]. ORCID

Jong-Sung Yu: 0000-0002-8805-012X Author Contributions †

These authors contributed equally.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the Global Frontier R&D Program on Centre for Multiscale Energy System (NRF-2011-0031571) and an NRF grant (NRF 2014K2A3A1000240) funded by the Ministry of Education, Science and Technology. The authors also thank KBSIs at Daegu and Busan for TEM and XPS measurements and CCRF in DGIST for SEM measurements.

■

REFERENCES

(1) Oh, H. S.; Nong, H. N.; Reier, T.; Bergmann, A.; Gliech, M.; Araujo, J. F. D.; Willinger, E.; Schlögl, R.; Teschner, D.; Strasser, P. J. Am. Chem. Soc. 2016, 138, 12552−12563. (2) Song, M. Y.; Yang, D.-S.; Singh, K. P.; Yuan, J.; Yu, J.-S. Appl. Catal., B 2016, 191, 202−208. (3) Görlin, M.; Chernev, P.; Araujo, J. F. D.; Reier, T.; Dresp, S.; Paul, B.; Krahnert, R.; Dau, H.; Strasser, P. J. Am. Chem. Soc. 2016, 138, 5603−5614. (4) Liu, G.; Li, P.; Zhao, G.; Wang, X.; Kong, J.; Liu, H.; Zhang, H.; Chang, K.; Meng, X.; Kako, T.; Ye, J. J. Am. Chem. Soc. 2016, 138, 9128−9136. (5) Tao, H. B.; Fang, L.; Chen, J.; Yang, H. B.; Gao, J.; Miao, J.; Chen, S.; Liu, B. J. Am. Chem. Soc. 2016, 138, 9978−9985. (6) Zou, S.; Burke, M. S.; Kast, M. G.; Fan, J.; Danilovic, N.; Boettcher, S. W. Chem. Mater. 2015, 27, 8011−8020. (7) Ping, J. F.; Wang, Y. X.; Lu, Q. P.; Chen, B.; Chen, J. Z.; Huang, Y.; Ma, Q. L.; Tan, C. L.; Yang, J.; Cao, X. H.; Wang, Z. J.; Wu, J.; Ying, Y. B. Adv. Mater. 2016, 28, 7640−7645. (8) Fan, Z. X.; Luo, Z. M.; Chen, Y.; Wang, J.; Li, B.; Zong, Y.; Zhang, H. Small 2016, 12, 3908−3913. (9) Wu, j.; Xue, Y.; Yan, X.; Yan, W.; Cheng, Q.; Xie, Y. Nano Res. 2012, 5, 521−530. (10) Cheng, Y.; Tian, Y.; Fan, X.; Liu, J.; Yan, C. Electrochim. Acta 2014, 143, 291−296. (11) Xiao, Z.; Huang, X.; Xu, L.; Yan, D.; Huo, J.; Wang, S. Chem. Commun. 2016, 52, 13008−13011. 2390


Research Article

ACS Catalysis (45) Girgis, B. S.; Temerk, Y. M.; Gadelra, M. M.; Abdullah, I. D. Carbon Science 2007, 8, 95−100. (46) Razmjooei, F.; Singh, K. P.; Bae, E. J.; Yu, J.-S. J. Mater. Chem. A 2015, 3, 11031−11039. (47) Singh, K. P.; Bae, E. J.; Yu, J.-S. J. Am. Chem. Soc. 2015, 137, 3165−3168. (48) Su, J.; Wu, X. L.; Yang, C. P.; Lee, J. S.; Kim, J.; Guo, Y. G. J. Phys. Chem. C 2012, 116, 5019−5024. (49) Venkatkarthick, R.; Davidson, D. J.; Ravichandran, S.; Vengatesan, S.; Sozhan, G.; Vasudevan, S. Catal. Sci. Technol. 2015, 5, 5016−5022. (50) Song, M. Y.; Park, H. Y.; Yang, D.-S.; Yu, J.-S. ChemSusChem 2014, 7, 1755−1763. (51) Wang, Z.; Li, P.; Chen, Y.; He, J.; Zhang, W.; Schmidt, O. G.; Li, Y. Nanoscale 2014, 6, 7281−7287. (52) Hoque, M. A.; Hassan, F. M.; Seo, M. H.; Choi, J. Y.; Pritzker, M.; Knights, S.; Ye, S.; Chen, Z. Nano Energy 2016, 19, 27−38. (53) Qie, L.; Chen, W.; Xiong, X.; Hu, C.; Zou, F.; Hu, P.; Huang, Y. Adv. Sci. 2015, 2, 1500195. (54) Liu, H.; Sun, P.; Feng, M.; Liu, H.; Yang, S.; Wang, L.; Wang, Z. Appl. Catal., B 2016, 187, 1−10. (55) El-Sawy, A. M.; Mosa, I. M.; Su, D.; Guild, C. J.; Khalid, S.; Joesten, R.; Rusling, J. F.; Suib, S. L. Adv. Energy Mater. 2016, 6, 1501966. (56) Higgins, D.; Hoque, M. A.; Seo, M. H.; Wang, R.; Hassan, F.; Choi, J. Y.; Pritzker, M.; Yu, A.; Zhang, J.; Chen, Z. Adv. Funct. Mater. 2014, 24, 4325−4336. (57) Wang, C.; Guo, Z.; Shen, W.; Xu, Q.; Liu, H.; Wang, Y. Adv. Funct. Mater. 2014, 24, 5511−5521. (58) Rodríguez, E.; García, R. Fuel 2012, 93, 288−297. (59) Choi, C. H.; Park, S. H.; Woo, S. I. ACS Nano 2012, 6, 7084− 7091. (60) Sheng, Z. H.; Gao, H. L.; Bao, W. J.; Wang, F. B.; Xia, X. H. J. Mater. Chem. 2012, 22, 390−395. (61) Wang, Y.; Wang, Y.; Jiang, R.; Xu, R. Ind. Eng. Chem. Res. 2012, 51, 9945−9951. (62) El Haskouri, J.; Moragues, A.; Beltran, A.; Murcia-Mascarós, S.; Plazaola, F.; Legarra, E.; Mauri-Aucejo, A.; Brotons-Gisbert, M.; Sánchez-Royo, J. F.; Beltran, D.; Amoros, P. Microporous Mesoporous Mater. 2014, 187, 14−22. (63) Yang, S.; Zhi, L.; Tang, K.; Feng, X.; Maier, J.; Müllen, K. Adv. Funct. Mater. 2012, 22, 3634−3640.

2391
