Possible Oxygen Reduction Reactions for Graphene Edges from First

Jul 8, 2014 - Department of Organic and Polymeric Materials, Graduate School of Science and Engineering, Tokyo Institute of Technology,. 2-12-1 I6-31 ...
0 downloads 9 Views 484KB Size
Article pubs.acs.org/JPCC

Possible Oxygen Reduction Reactions for Graphene Edges from First Principles Takashi Ikeda,*,† Zhufeng Hou,‡ Guo-Liang Chai,‡ and Kiyoyuki Terakura‡,§ †

Condensed Matter Science Unit, Quantum Beam Science Center, Japan Atomic Energy Agency (JAEA), Hyogo 679-5148, Japan Department of Organic and Polymeric Materials, Graduate School of Science and Engineering, Tokyo Institute of Technology, 2-12-1 I6-31 Ookayama, Tokyo 152-8552, Japan § Research Center for Simulation Science, Japan Advanced Institute of Science and Technology (JAIST), 1-1 Asahidai, Nomi, Ishikawa 923-1292, Japan ‡

S Supporting Information *

ABSTRACT: N-doped carbon-based nanomaterials are attracting a great interest as promising Pt-free electrode catalysts for polymer electrolyte fuel cells (PEFCs). In this computational study, we demonstrate that N-doped graphene edges can exhibit enhanced catalytic activity toward oxygen reduction reactions by controlling their electron-donating and -withdrawing abilities and basicity, resulting in higher selectivity of 4e− reduction via inner- and outer-sphere electron transfer at edges under acidic conditions, respectively. Our simulations also show that 2e− reduction occurs selectively in the presence of pyridinic N next to carbonyl O at zigzag edges. This study thus rationalizes the roles of doped N in graphenelike materials for oxygen reduction reactions.



INTRODUCTION Polymer electrolyte fuel cells (PEFCs) are one of the most promising power sources due to their prominent characteristics of rather high efficiency, low operating temperature, and low environmental impact. However, their wide commercialization is still prohibited not only by insufficient infrastructure for supplying and storing fuel but also by the cost of expensive Ptbased catalysts currently required to facilitate electrode reactions at operating temperatures of around 80 °C. The oxygen reduction reaction (ORR) at the cathode in H2−O2 fuel cells is known to be significantly slower than the counterpart hydrogen oxidation reaction at the anode. Therefore, many groups have challenged the development of alternative Pt-free cathode catalysts composed of more abundant elements to efficiently reduce the production cost of PEFCs. Recently, several groups1−42 reported that carbon-based nanomaterials including carbon nanotubes, carbon nanofibers, and graphenes doped with a certain number of light heteroatoms such as N, B, S, and P exhibit high ORR activity comparable to that of conventional Pt-based catalysts under alkaline conditions. On the other hand, the ORR activity of those carbon-based materials termed carbon alloys43 (CAs) in this paper is found to be commonly inferior to that of Pt catalysts for currently unknown reasons in concentrated acid media, where cathode catalysts for PEFCs are necessary to work efficiently. Thus, it is of crucial importance to elucidate on the atomistic level what causes the relatively poor activity of CAs in acidic solutions to make it satisfactorily high under the actual operating conditions of PEFCs. For such a purpose, the combined use of various spectroscopic techniques is indispensable to obtaining detailed information on chemical species of dopants and the associated modulation of local electronic structures responsible for © 2014 American Chemical Society

catalytically activating synthesized CAs. Indeed, N 1s X-ray photoemission spectroscopy (XPS) for N-doped CAs (N-CAs), which have been most intensively investigated among CAs, suggests that pyridinic, pyrrolic, and graphitic N are included in N-CAs,44−46 while N K-edge X-ray absorption spectroscopy (XAS) suggests pyridinic, cyanide, and graphitic N as existing moieties in N-CAs.47−49 Although chemical species of N dopants are assigned with some ambiguity from both X-ray techniques, pyridinic and graphitic N atoms are certainly included in N-CAs together with several minor species with varying content depending on N sources and detailed synthesis procedures employed. Recall that actually synthesized CAs are constituted of a substantial number of edges, defects, and pores in addition to a well-developed sp2 C network, thus providing high capabilities of modification. Consequently, the obvious correlation is hardly obtained between the content of each N species in N-CAs and their catalytic activity measured using electrochemical methods. Furthermore, according to recent principal component analysis most of the N atoms in N-CAs are suggested to be irrelevant to the catalytic activity of NCAs.31 Thus, it is extremely difficult to unravel the exact location of active sites in N-CAs for ORR by using experimental techniques alone. In our previous works50−52 systematic analysis was performed for X-ray core-level spectra of N doped along the edges of graphene nanoribbons (GNRs) and graphene nanoflakes by using first-principles-based methods. Our results indicate that chemical shifts of N 1s core-level binding energy range in magnitude over 4 eV, depending on the N Received: April 18, 2014 Revised: June 12, 2014 Published: July 8, 2014 17616

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

energy barriers were estimated at a fixed total Sz, thus allowing an O2 to change its spin state as a result of the interaction with the adsorbents. On the other hand, we performed a series of the usual MD simulations for subsequent ORR by repeatedly introducing additional H+ and e− into the well-equilibrated product of the previous step. A set of trajectories of at least 4 ps for each reduction product was generated in the NVT ensemble at 300 K using the Nosé−Hoover thermostat.64,65 The free-energy difference ΔF for each reduction step was estimated by neglecting the influence of the electric field on the reaction pathways as proposed by Nørskov et al.66 Namely, the total energy Etot(H+ + e−)|U of H+ + e− at finite electrode potential U was computed according to the number of electrons at electrode n as Etot(H+ + e−)|U = 1/2Etot(H2) − neU. The remaining term in ΔF was approximated by the difference in the Kohn−Sham energies averaged over our generated trajectories for the well-equilibrated products of the associated two consecutive steps. Note that our computed ΔF includes rather large errors arising presumably from the insufficient relaxation of the H-bond network, which is necessary to reform depending on the chemical species of edge functional groups including ORR intermediate adducts. The error in the overall ΔF for 4e− ORR is found to be typically 0.8 eV, thus the accuracy of our computed ΔF is at very best on the semiquantitative level.

configuration and its location, i.e., zigzag or armchair edges. Besides the chemical shifts of the N 1s level occupied and unoccupied valence states, reflected respectively in X-ray emission spectra (XES) and XAS, are also altered to varying degrees depending on the details of N dopants and the termination of edges. Thus, N doping in CAs is suggested to be an effective means to enhance the electron-donating and -withdrawing capabilities of delocalized π electrons around dopant sites. In addition, the basicity of edges is modifiable by doping pyridinic N with a lone pair exposed to electrolyte, which tends to attract rather strongly protons of H2O and H2O2 if present nearby. Thus, ORR is potentially facilitated by controlling the basicity of graphene edges as demonstrated in this paper. Although recently developed N-CA-based catalysts show high activity and durability, current progress primarily involves experimental trial and error to combine suitable precursors and pyrolysis conditions to maximize performance. An alternative bottom-up approach instead involves identifying and optimizing various material properties that govern reaction rates, which obviously necessitates a fundamental knowledge of the ORR mechanisms. The present study aims at describing possible reaction paths of ORR for N-CAs by exploiting first-principlesbased computational methods. We find that N-CAs are by their nature capable of catalyzing ORR in multiple reaction paths. Our quantitative estimate of the ORR activity conducted without any assumptions of reaction paths thus enables us to highlight specific material properties of N-CAs to be tuned for the further enhancement of their catalytic activity toward ORR.



RESULTS AND DISCUSSION Undoped GNR. First, we describe as a reference the ORR activity of undoped graphene suggested from our simulations. Figure 1 shows our computed free-energy diagram for ORR at



COMPUTATIONAL METHODS We performed Car−Parrinello molecular dynamics53,54 (MD) simulations within a plane-wave pseudopotential scheme for density functional theory (DFT). The Becke55−Lee−Yang− Parr56 (BLYP) generalized gradient approximations were employed for the exchange and correlation functionals in an unrestricted-spin scheme combined with Grimme’s semiempirical van der Waals corrections.57 The valence−core interaction was described by Troullier−Martins58 pseudopotentials (PPs) for C, N, O, and S and a von Barth−Car analytical PP for H.59 The sampling of the Brillouin zone was restricted to the Γ point. A systematic set of simulations was performed for various models composed of graphene sheets doped with heteroatoms, one O2 molecule, and a certain number of H2O molecules in a simulation box with periodic boundary conditions. More specifically, our slab models for an edge plane contain about 250 atoms consisting of 4 graphene sheets with a hydrogenated bottom C fixed to the positions of bulk graphite, while for the basal plane our models consist of 2 graphene sheets with a bottom layer restrained to oscillate around the crystal positions by harmonic potentials. For both planes, a water layer of ∼10 Å and one O2 molecule initially in a spin-triplet state are included in our simulation box to mimic the environment around a hypothetical CA-based cathode in PEFCs. The O2 adsorption was examined using the blue moon ensemble60,61 generated at 300 K by velocity rescaling as described in refs 62 and 63. In brief, free-energy barriers for the adsorption of an O2 initially located at a distance of ∼5 Å from graphene edges were computed by decreasing the distance between O2 and a hypothetical adsorption site. For each point, we performed constrained MD simulations for at least 1.0 ps to reequilibrate the system. As remarked in ref 63, all of the free-

Figure 1. Free-energy diagram for ORR at the zigzag-edge plane of undoped GNR at 0, 0.36, and 1.28 V vs SHE constructed from our first-principles-based simulations. Here, the maximum potential of 1.28 V was obtained from our computed ΔF for the overall 4e− ORR. Snapshots for ORR products (d)−(g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. The structure of GNR that we employed is also shown in the upperright panel. Atom colors are white for H, gray for C, and red for O.

the zigzag-edge plane of undoped GNR. As already reported in ref 62, O2 molecules are adsorbed in an end-on configuration at the zigzag edges of pure graphene (Figure 1c). The activation barrier of 0.15 eV for O2 desorption estimated from the blue moon ensemble is found to be much smaller than the corresponding estimate of 0.85 eV for the adsorption, indicating that O2 adsorption at the zigzag edges of undoped graphene is much weaker than that for N-doped graphene discussed below. Thus, one e− reduction of an adsorbed O2 17617

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

represented as *O2 in this study leads to an *OOH adduct (Figure 1d), where * indicates the adsorption site. In the second reduction step the first H2O is produced via the O−O bond cleavage of the adduct, which leaves an O atom at the adsorption site (Figure 1e). The resulting *O can be further reduced to an *OH as displayed in Figure 1f. Finally, the second H2O formed by reducing the *OH is desorbed from the adsorption site (Figure 1g). In these successive ORRs each electron is transferred through the C−O bond of partial covalent character. The ΔF values estimated using the procedures given in the previous section are −0.36, −3.11, −0.64, and −1.70 eV at 0 V vs SHE in the order of four elementary ORR steps as mentioned above. Note that in the ideal situation the free energy of the system decreases evenly by 1.23 eV at 0 V vs SHE for each of four reduction steps because the overall 4e− ORR of O2 + 4H+ + 4e− → 2H2O occurring at the cathode under acidic conditions shows ΔF = −4.92 eV under standard-state conditions. When ΔF for elementary processes deviates from the ideal value, some ORR processes are activated at higher potentials. Indeed, our simulations suggest that for the zigzag-edge plane of undoped GNR the first and third reduction processes become uphill at potentials higher than 0.36 and 0.64 V vs SHE, respectively. Small energy lowering at the first electrochemical step indicates that *OOH is only weakly adsorbed. This aspect together with the large activation barrier of 0.85 eV for O2 adsorption implies that the undoped zigzag edge would not provide efficient ORR active sites. Graphitic N-Doped GNR. Our previous studies showed that graphitic N substituted for the C atom directly bonded with edge carbons of zigzag GNR can activate its neighboring edge carbons toward ORR.62,63,67 Such a specific site for Ndoping is called an edge-1 site, and the resulting graphitic N is denoted as Nedge−1 in this paper. Figure 2 shows our computed

the density of states (DOS) at these C sites just below the Fermi level (EF) accompanied by the diminishingly small DOS just above EF as explained in ref 67. Therefore, the electron transfer (ET) from the edge C to an incoming O2 will be significantly enhanced upon the adsorption. Indeed, with decreasing distance dO2−* between an O2 and the adsorption site * the O−O bond length averaged over our trajectories starts to increase from dO2−* = 2.4 Å as shown in Figure 3 and

Figure 3. Averaged O−O bond length of an O2 molecule as a function of the distance between an O2 and the adsorption site * obtained from our constrained MD simulations for the zigzag-edge plane of undoped (black) and Nedge−1-doped GNR (red).

finally reaches ∼1.5 Å, which is about 0.1 Å larger than that for undoped GNR. This final O−O bond length is also found to be much larger than that of superoxide anion O2− in our aqueous system (1.34 Å) obtained from a separate MD run. Although the charge state of molecules adsorbed at solid surface is hardly identified unambiguously, O2 molecules are certainly negatively charged and the terminal O has three lone pairs. Thus, O−O bond cleavage easily occurs in the first ORR step, resulting in the formation of an OH− released from the surface along with an O atom left at the adsorption site of positively charged graphene (Figure 2d) instead of the *OOH adduct observed for undoped GNR (Figure 1d). The remaining three steps lead formally to the same products as those for undoped GNR. However, it is worthwhile to point out that the ratedetermining step of ORR is switched to the third process of *O + H+ + e− → *OH from the first one. We also observed that the last ORR process has a kinetic barrier of about 0.2 eV even at 0 V vs SHE as already pointed out in ref 62. Although this barrier height is much lower than that for O2 adsorption, the presence of the kinetic barrier in the last process together with the obvious activation barrier in the third one at potentials of U > 0.32 V vs SHE suggests that adsorption sites for the zigzag-edge plane of Nedge−1-doped graphene tend to be poisoned by *O and *OH at higher potentials in accordance with Okamoto.68 Note that in the presence of Nedge−1 in GNRs, ΔF for the third step decreases in magnitude by a nonnegligible amount of 0.32 eV with respect to that for undoped ones possibly due to the resonance effect of π-excessive aromatic systems, preferring to increase the number of double bonds participating in aromatic regions. Therefore, our present simulations indicate that the Nedge−1 dopant alone cannot improve very remarkably the catalytic activity of N-CAs toward 4e− ORR for an O2 adsorbed in an end-on configuration, whereas it efficiently promotes the O−O bond cleavage due to its strong electron-donating capability.

Figure 2. Free-energy diagram for ORR at the zigzag-edge plane of Nedge−1-doped GNR at 0, 0.32, and 1.03 V vs SHE obtained from our simulations. Snapshots for ORR products (d)−(g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. Atom colors are white for H, gray for C, red for O, and blue for N.

free-energy diagram for ORR at the zigzag-edge plane of Nedge−1-doped GNR. Although the activation barrier of O2 adsorption is nearly the same as that for undoped GNR, the barrier of O2 desorption at the edge carbon next to Nedge−1 increases significantly by 0.59 eV compared to that for pure GNR. This indicates that O2 molecules are more strongly bound in an end-on configuration (Figure 2c) at the neighboring edge C atoms of Nedge−1 due to the increase in 17618

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

serious consequence is that O2 chemisorption at edge carbons is also significantly strengthened, thus leading eventually to the poisoning of catalytic sites at higher electrode potentials. With these considerations we naturally expect that edge carbons will be able to be catalytically more activated by embedding a couple of complementary moieties of π-deficient and πexcessive aromatics within the π-conjugated region so that the density of π states around dopant sites is effectively compensated for. This modulation of electronic structures will provide edge C atoms of graphenelike materials with more balanced electron-donating and -withdrawing abilities of their π electrons. Hence, those materials having such dual abilities are expected to work in principle as catalysts not only for reduction but also for oxidation. To prove our fundamental idea we inspect in this study three kinds of binary-doped GNRs containing either pyridiniumlike N, pyranlike O, or thiopyranlike S at edges as an electron acceptor and Nedge−1 as a common electron donor. Recall that the above three kinds of electron acceptors have formal isoelectronic structures, showing a tendency to be positively charged to varying degrees. The local DOS shown in Figure S1 indicates that the electronic structures around dopant sites are indeed modified by incorporating the binary dopants in GNRs as described above. Therefore, we can argue how the balance of electron-donating and -withdrawing abilities generally affects the ORR activity by comparing the expected catalytic activity for these three cases even if the doping of pyranlike O and thiopyranlike S is difficult to realize in practice. Nedge−1 and Pyridiniumlike N. Figure 5 shows our computed free-energy diagram for ORR at the zigzag-edge step of GNR

Contrary to the zigzag-edge plane, O2 molecules are adsorbed in a side-on configuration at edge carbons of the zigzag-edge step of Nedge−1-doped GNR as shown in Figure 4c.

Figure 4. Free-energy diagram for ORR at zigzag-edge step of Nedge−1doped GNR at 0, 0.62, and 1.07 V vs SHE obtained from our simulations. Snapshots of ORR products (d)−(g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. Atom colors are white for H, gray for C, red for O, and blue for N.

Thus, the reduction of 2*O can proceed at both the adsorption sites via the intermediates of *OH + *O (Figure 4d), 2*OH (Figure 4e), and *OH (Figure 4f). For these first three processes ΔF was estimated in order as −2.19, −1.23, and −0.62 eV at 0 V vs SHE, and thus the deviations of ΔF from the ideal value of −1.23 eV are significantly reduced compared to those for an O2 adsorbed in an end-on configuration. On the other hand, the last process is commonly expressed as *OH + H+ + e− → H2O among the three cases examined so far. Therefore, the zigzag-edge step of Nedge−1-doped GNR also tends to suffer from the poisoning of active sites by *OH due to the presence of a kinetic barrier of ∼0.2 eV irrespective of the adsorption configuration of *O2, as indicated in ref 62. Codoped GNR. To enhance the ORR activity of graphene edges the deviation of ΔF at 0 V vs SHE from the ideal value of −1.23 eV must be minimized for each reduction step. Moreover, the activation barrier for O2 adsorption is required to decrease to a yet unknown suitable value from the kinetic viewpoint. In a previous paper we proposed the codoping of B and N at the edges of graphene to reconcile its reactivity and stability.63 However, the ORR activity of a B−N pair doped at zigzag edges turned out to be rather poor by using more a simplified estimation of catalytic activity. Our computational estimate of the ORR activity based on the binding energy of adducts is presented elsewhere. Thus, we adopt in this work a different approach to enhancing the ORR activity based on our fundamental understanding of the origins of large overpotentials required in the third step of *O + H+ + e− → *OH for the zigzag-edge plane of Nedge−1-doped GNR and the kinetic barrier in the last step of *OH + H+ + e− → H2O observed in our simulations irrespective of the surface orientation. Graphenes doped with graphitic N including Nedge−1 are regarded as typical π-excessive heteroaromatics, which exhibit enhanced electron-donating ability to an adsorbed O2, thus facilitating the O−O bond cleavage as demonstrated in Figure 2. Furthermore, the increase in the density of reactive π states at adsorption sites tends to increase the barrier height for O2 adsorption itself due to enhanced electrostatic repulsion between an approaching O2 and the adsorption site. A more

Figure 5. Free-energy diagram for ORR at the zigzag-edge step of GNR doped with Nedge−1 and pyridiniumlike N at 0, 0.33, and 1.01 V vs SHE obtained from our simulations. Snapshots for ORR products (d)−(g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. The structure of GNR that we employed is also shown in the upper-right panel. Atom colors are white for H, gray for C, red for O, and blue for N.

doped with Nedge−1 and pyridiniumlike N. In this N-doped GNR a secondary N dopant is substituted for one of the second-nearest-edge C atoms of Nedge−1. According to our previous analysis of interactions between two doped N atoms in graphene, this N configuration corresponds to a metastable one with an interaction energy that is as small as 0.1 eV.67 Interestingly, the activation barrier for O2 adsorption in the end-on configuration at one of the neighboring edge C atoms of Nedge−1 (see Figure 5c) becomes 0.45 eV lower than that for the zigzag-edge step of Nedge−1-doped GNR (Figure 4). The O2 desorption barrier is also found to decrease by 0.29 eV compared to that for an edge plane doped with Nedge−1 only, whose edge C atoms next to Nedge−1 also adsorb O2 molecules 17619

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

in an end-on configuration (Figure 2c). The resulting *O2 undergoes 4e− ORR for apparently the same reaction path taken for the zigzag-edge plane of Nedge−1-doped GNR (Figure 2). From our MD simulations, ΔF for four successive elementary processes was estimated in order as −1.53, −0.91, −1.01, and −0.33 eV at 0 V vs SHE, thus indicating that the deviations of ΔF are efficiently suppressed from the ideal situation except for the last process. Note that our blue moon simulations suggest that free energy of the system including *O2 is 0.27 eV lower than that for the initial state. This means that the further stabilization of graphene systems will contribute to increasing the reaction rate of *OH + H+ + e− → H2O. Therefore, we can expect that further improvement of the overall ORR activity is achievable by modifying the edge termination as discussed later. Nedge−1 and Pyranlike O. By substituting pyranlike O for pyridiniumlike N, the electron-withdrawing ability of edge carbons around the dopant will be enhanced because of the large electronegativity of O (3.5) compared to that of N (3.0) and C (2.5). In fact, our computed free-energy diagram shown in Figure 6 for the zigzag-edge step doped with pyranlike O

Figure 7. Free-energy diagram for ORR at the zigzag-edge step of GNR doped with Nedge−1 and thiopyranlike S at 0, 0.43, and 1.09 V vs SHE obtained from our simulations. Snapshots of ORR products (d)− (g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. Atom colors are white for H, gray for C, red for O, blue for N, and yellow for S.

into OH− and *O in the first step. The subsequent reductions basically follow the same reaction path as observed for the zigzag-edge plane of Nedge−1-doped GNR (Figure 2). Our estimated ΔF values of −2.09, −1.10, −0.43, and −0.69 eV at 0 V vs SHE for successive reduction processes are also found to be very close to the corresponding ones for ORR at the zigzagedge plane doped with Nedge−1 only. However, we noticed that the kinetic barrier in the last process of *OH + H+ + e− → H2O observed commonly for Nedge−1 only disappears by the codoping of Nedge−1 and thiopyranlike S at zigzag edges. Thus, we expect the ORR activity exhibited by graphene edges to be possibly remarkably improved by incorporating Nedge−1 and thiopyranlike S within the same π-conjugated region included in graphenelike materials. Further improvement of the catalytic activity of the N−S codoped zigzag step is suggested to be achievable by modifying the edge termination. Figure 8 shows our computed free-energy

Figure 6. Free-energy diagram of ORR at the zigzag-edge step of GNR doped with Nedge−1 and pyranlike O at 0 and 1.05 V vs SHE obtained from our simulations. Snapshots for ORR products (d)−(g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. Atom colors are white for H, gray for C, red for O, and blue for N.

along with Nedge−1 clearly shows that embedding pyranlike O in graphene lowers the ORR activity of edge carbons. More precisely, O2 molecules adsorbed by overcoming the small activation barrier of 0.32 eV hardly start to be electrochemically reduced because ΔF for the first process was estimated to be as small as −0.1 eV at 0 V vs SHE. Once ORR is initiated, subsequent steps of *OOH + H+ + e− → *O + H2O, *O + H+ + e− → *OH, and *OH + H+ + e− → H2O are expected to proceed smoothly in the range of U < 0.75 V vs SHE. Therefore, a huge overpotential is required only to initiate ORR at the edges codoped with pyranlike O and Nedge−1. Nevertheless, considering that one of the serious problems of N-CAs is a strong tendency to poison catalytic sites according to our present study and Okamoto,68 the doping of pyranlike O would be useful in suppressing such undesired side effects of N doping. Nedge−1 and Thiopyranlike S. Since the electronegativity of S is nearly the same as that of C, the substitution of pyranlike O with thiopyranlike S is expected to shift the too strong electronwithdrawing capability of edge carbons in the desired direction. In fact, our computational results summarized in Figure 7 show that O2 molecules adsorbed at one of the edge C atoms next to Nedge−1 by overcoming a small barrier of 0.31 eV are dissociated

Figure 8. Free-energy diagram of ORR at the zigzag-edge step of GNR doped with Nedge−1 and thiopyranlike S at 0, 0.75, and 1.18 V vs SHE obtained from our simulations. Snapshots of ORR products (d)−(g) are shown together with those for initial (a), transition (TS, b), and final (c) states for O2 adsorption. Here dihydrogenated C atoms at the edges of both sides except around the O2 adsorption site are shown by arrows in the upper-right panel. Atom colors are white for H, gray for C, red for O, blue for N, and yellow for S.

diagram for ORR at the zigzag-edge step of modified N−S codoped GNR, where mono- and dihydrogenated C atoms coexist at the edges of both sides except around the O2 adsorption site. Recent theoretical studies including ours on the edge-carbon termination suggest that except for very low hydrogen partial pressure (lower than 10−20 bar at 300 K), edge 17620

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

H2O2 is immediately dissociated into a hydroxyl radical and a hydroxyl anion for which an extra electron is supplied from the graphene sheet, including NH that participates in the H-bond complex. The resulting OH− easily deprotonates the nearest NH group, reverting to pyridinic N. Then the remaining ·OH radical was also observed to react quickly with another NH group via reactions ·OH + NH → OH− + NH+ → H2O + N. Thus, two H2O molecules are produced (Figure 9d). Finally pyridiniumlike N is recovered in the electrochemical reaction of N + H+ + e− → NH (Figure 9e,f). Although the estimated ΔF of about −0.25 eV at 0 V vs SHE for the third and fourth steps is relatively small in magnitude, the actual reactions are expected to occur more easily in acid media of pH ∼0, which is not considered in the present computational treatment. Note that our ΔF for the overall ORR is largely underestimated partially because of the lack of an accurate estimate of ΔF for the O2 fixation process (Figure 9a,b), which should correspond to the binding energy of a newly formed H bond (on the order of −0.1 eV). Another computational error is expected to arise from the inadequate description of electronic structures of ·OH radicals involved in the second reduction process. It is shown that an unpaired electron of hydroxyl radicals is much too delocalized and their neighboring H2O molecules are artificially spin-polarized at the level of the current standard DFT that we employed,71 thereby our computed ΔF for the second ORR is considered to include rather large errors. In our additional simulations for pyridinic N contained in armchair-edge planes ORR was observed to occur again selectively via 2e− ORR and hydroxyl radicals were also produced as a result of the molecular dissociation of a temporarily produced H2O2. As mentioned in our previous paper,52 two pyridiniumlike N atoms along armchair edges in the amiddle pydm configuration in the notation of ref 52 are found to catalyze again two successive 2e− ORRs to fully reduce an O2 via a Fenton-like reaction, revealed using charge-constrained DFT (CDFT)72 expressed as H2O2 + Cδ− + H+ → OH− + ·OH + C(1−δ)+ + H+ → H2O + ·OH + C(1−δ)+, where Cδ− is the edge C in between two pyridiniumlike N atoms on which an excess electron is located. The detailed results obtained using CDFT will be given elsewhere. Then an ·OH was again observed to react quickly with an NH group at the edge. Therefore, ORR suggested from our simulations for pyridinic N includes the formation of reactive hydroxyl radicals regardless of the type of edges. Implications for the synthesis of N-CAs as cathode catalysts for PEFCs are discussed in the last section. Pyridinic-Carbonyl/Pyridiniumlike-Hydroxyl Redox Couple. Actually synthesized N-CAs are known to contain a nonnegligible number of oxygens depending on precursors and detailed synthesis procedures. Prior to a discussion of the influence of N dopants on 2e− ORR we describe a likely reaction path for 2e− ORR associated with O-containing groups at the edges of undoped GNR. Figure 10 shows the free-energy diagram obtained from our simulations for ORR at the zigzagedge step of undoped GNR with carbonyl and hydroxyl groups at edges. When these O-containing groups expected to be included abundantly are next to each other at graphene edges, an intra-H bond of OH···O formed between the two groups substantially stabilizes the graphene edges. Thus, for the electrochemical reduction of the CO group ΔF is nearly zero or small positive numbers in most cases even at 0 V vs SHE. When the two groups are separated, e.g., by dihydrogenated C atoms as shown in Figure 10a, the one electron reduction of CO groups turns slightly exergonic, thus showing ΔF of

carbons of zigzag-GNRs are terminated not only as monohydrogenated but also as dihydrogenated.51,69,70 This means that graphenes terminated fully with monohydrogens can be stabilized by mixing dihydrogenated C atoms to some extent. Compared to our computed ΔF for the previous GNR without edge modification, our estimated ΔF values of −2.06, −1.13, −0.75, and −0.93 eV at 0 V vs SHE for four successive reductions show clearly that the edge modification makes ΔF greater even among the elementary processes, although *O is converted to *OH (Figure 8e) in the second step instead of the third one. Therefore, the ORR activity of the N−S codoped zigzag step is potentially further improved as evidenced by our present (semi)quantitative estimate of catalytic activity. Pyridinic N-Doped GNR. Pyridinic/Pyridiniumlike Redox Couple. Apart from the electron-donating and -withdrawing capabilities, the basicity of edges in N-CAs is modifiable by doping pyridinic N with a lone pair exposed to electrolyte. A lone pair located at base sites rather strongly attracts protons of H2O and H2O2 if present nearby, enabling the formation of the H-bond network necessary for facile proton transfer to reactants. Thus, these sites are potentially associated with ORR. Moreover, pyridiniumlike N produced more or less by protonating pyridinic N under acidic conditions is able to attract an O2 by forming H bonds in the form of N···HO2 (or NH···OH2···O2). When supplied O2 molecules participate in the H-bond network developed from NH at edges, 2e− ORR is found to selectively occur via outer-sphere ET through H bonds. If the resulting product is still kept near pyridinic N, an unstable H2O2 has a chance of further reduction, eventually ending up as more preferable 4e− ORR. An example of such cases we encountered is shown in Figure 9. In this simulation

Figure 9. Free-energy diagram of ORR at the zigzag-edge plane of GNR doped with pyridinic N at 0, 0.40, and 0.88 V vs SHE obtained from our simulations. Snapshots of ORR products (c)−(f) are shown together with those for initial (a) and final (b) states for an O2 bound to pyridiniumlike N by forming a H bond. Atom colors are white for H, gray for C, red for O, and blue for N.

two pyridinic N atoms contained in zigzag-edge planes of different graphene sheets are assumed to be converted to pyridiniumlike N in the initial configuration (Figure 9a). An O2 initially located a distance of ∼5 Å from the graphene edges spontaneously approaches one of the NH sites, leading to the formation of a H bond in the form of N···HO2 by shifting H originally attached to N upon H-bond formation (Figure 9b). In the first reduction step H of HO2 is shifted back to the original N and an H-bond complex of NH···O2H is formed as shown in Figure 9c. When another proton is supplied to the complex through the H-bond network a temporarily produced 17621

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

kinetic barrier of 0.33 eV estimated from our blue moon ensemble for the process in Figure 11c−e. Thus, such hydrogenated groups are able to work as an efficient hydrogen source for converting an O2 to an H2O2 in ordinary chemical reactions without O2 chemisorption. Therefore, our simulations suggest that 2e− ORR arising from the NCCO/NHC COH redox couple, which can be regarded as an analog of the well-known quinone/hydroquinone couple utilized for the industrial production of H2O2, shows a positive shift of the onset potential by about 0.2 V due to the obviously missing energy loss in nonelectrochemical processes. In contrast to pyridinic N, the presence of Nedge−1 at neighboring sites of carbonyl oxygens is found to change the reduction of the CO group to an activation process with ΔF being estimated as 0.29 eV at 0 V vs SHE. Thus, 2e− ORR associated with the carbonyl/hydroxyl redox couple at graphene edges significantly depends on the chemical species of their neighboring N dopants.

Figure 10. Free-energy diagram for ORR associated with the carbonyl/hydroxyl redox couple at the zigzag-edge step of undoped GNR at 0 and 0.47 V vs SHE obtained from our simulations. Snapshots of the hydrogenation of edge functional groups (a) and (b) are shown together with those for H2O2 production (d) from an O2 via an *OOH (c). Atom colors are white for H, gray for C, and red for O.



−0.12 eV at 0 V vs SHE as estimated from our simulations. The edge functional groups fully saturated with H were observed to react easily with an incoming O2 in line with our previous observation,51 resulting in the conversion to ·OOH. This activated ·OOH released upon its formation is eventually attached to the C edge of the opposite side of GNR (or the C edge of a different piece of GNR) even though N dopants are absent (Figure 10c). Since O2 adsorption at zigzag edges of undoped GNR is much weaker as already stated, the reduction of an *OOH produces a final product H2O2 of 2e− ORR. It should be noticed that the intermediate step in Figure 10b,c is the usual nonelectrochemical reaction. Thus, the corresponding ΔF estimated as −0.57 eV cannot be converted into electricity. Hence, the maximum potential obtained in this reaction path decreases to 0.47 V vs SHE though 2e− ORR is completed. The chemical species of O-containing groups at edges are also influenced by the neighboring N dopant if it exists. As shown in Figure 11, pyridinic N and carbonyl O next to each

CONCLUDING REMARKS In this study possible ORRs for graphene edges doped with N (and O or S) have been investigated by performing firstprinciples-based MD simulations. Our (semi)quantitative estimate of catalytic activity shows that both graphitic and pyridinic N atoms activate graphene edges toward ORR under acidic conditions. However, these two types of N dopants lead to different paths of 4e− ORR via different ET processes. Graphitic N atoms at zigzag edges are found to lead to direct 4e− ORR via inner-sphere ET. However, the catalytic sites are suggested to be rather easily poisoned by carbonyl and hydroxyl groups. Our simulations indicate that the poisoning of the catalytic sites is effectively suppressed when graphene edges are adequately sulfurized (or oxidized) so that excess π electrons introduced by graphitic N are compensated for. Moreover, adequate charge compensation renders electron-donating and -withdrawing abilities of π electrons highly balanced, which contributes to improving the overall performance for ORR exhibited by graphene edges. Therefore, the balance of these two abilities of π electrons is considered to be one of key properties of CAs to be tuned to enhance their catalytic activity toward ORR. The activation barriers for O2 adsorption and desorption are employable as measurable quantities for estimating the degree of charge compensation. According to our attempts to optimize the ORR activity, the barrier height of ∼0.3 eV for both processes is suggested to be an optimum condition for enhanced 4e− ORR. On the other hand, pyridinic N atoms at edges lead to successive 2e− + 2e− ORR via the outer-sphere ET mediated by H bonds under acidic conditions. The 4e− ORR via H2O2 formation is suggested to occur at electrode potentials lower than ∼0.4 V vs SHE. A lone pair directed to water solutions rather strongly attracts protons of H2O and H2O2, and thus the basicity of graphene edges is another key property to be tuned to enhance the ORR activity. However, we found that the reaction path and the corresponding free-energy difference for successive elementary processes of 2e− + 2e− ORR are essentially insensitive to the type of graphene edge. This suggests that the morphology of synthesized N-CAs is of primary importance rather than the detailed electronic structures as long as a rather large electric conductivity is maintained. In contrast, the activity of 2e− ORR arising from redox couples of O-containing groups significantly depends on the chemical species of their neighboring N dopant. Our

Figure 11. Free-energy diagram for ORR at the zigzag-edge step of oxidized GNR doped with pyridinic N at 0 and 0.76 V vs SHE obtained from our simulations. Snapshots for the hydrogenation of edge functional groups (a)−(c) are shown together with those for H2O2 production (d) and (e) from O2 without the adsorption at edge carbons. Atom colors are white for H, gray for C, red for O, and blue for N.

other at zigzag edges are electrochemically reducible due to the absence of an intra-H bond between NH and CO groups contrary to an OH and CO pair. The corresponding ΔF was estimated to be −0.81 and −0.70 eV at 0 V vs SHE, respectively. Since an intra-H bond between resulting NH and OH groups is also absent in this situation, H2O2 molecules are rather easily produced, whenever the fully hydrogenated groups are accessible to supplied O2 molecules, by overcoming a small 17622

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

Notes

simulations show that the presence of pyridinic N next to carbonyl O at zigzag edges improves the onset potential by ∼0.2 V with respect to that for the corresponding undoped graphene oxide. We also observed that ·OH radicals are produced upon molecular dissociation of H2O2, which occurs when protons are supplied to an HO2 bound to pyridinic N by forming an H bond or alternatively when a produced H2O2 approaches the graphene edges containing an excess number of π electrons. Although ·OH radicals were observed to be quenched by reacting with NH groups in our simulations, it is not guaranteed that NH groups are sufficiently included in working N-CA-based catalysts by considering that several redox couples of edge functional groups will also be able to produce H2O2 molecules. Thus, the present study strongly suggests that a sufficient number of additional quenchers of ·OH radicals are necessary to incorporate in actually synthesized N-CAs and/or ionomers otherwise electrolyte membranes will be damaged sooner or later. In addition, OH− anions were observed to be formed as an intermediate of 4e− ORR in our simulations for several Ndoped and N−S codoped CAs. Furthermore, we observed that produced hydroxyl ions, which correspond to the final product of ORR at the cathode in alkaline media, tend to be released from the graphene edges also under acidic conditions. These observations indicate that the energy loss during ORR in concentrated acid media due to nonelectrochemical reactions such as OH− + H+ → H2O will be generally larger than that in alkaline media as suggested in several experimental measurements of the ORR activity for actually synthesized N-CA-based catalysts. Thus, our present simulations explain partially why NCAs exhibit relatively poor ORR activity in acidic solutions. Besides the energy loss, the loss of catalytic sites resulting from the protonation of graphene edges is partially expected to cause the relatively poor activity of N-CAs, which should be examined more carefully with respect to enhancing their catalytic activity toward ORR under the actual operating conditions of PEFCs. Finally, the present comprehensive computational study reveals that the redox of N-CAs is closely associated with the intermediate steps of ORR not only to cleave the O−O bond of O2 and H2O2 molecules through the oxidation of N-CAs but also to hydrogenate O-containing groups at graphene edges (i.e., the reduction of N-CAs). Consequently, N-CAs are by their nature capable of catalyzing ORR in multiple reaction pathways, analogously to the roles of Co species discussed by Olson et al.20 Besides N dopants, metal species such as iron and cobalt ions incorporated into synthesized N-CAs will shift the redox potential of graphenelike materials. Thus, these metal species are also likely to contribute at least indirectly to the enhanced ORR activity of N-CAs. However, a close examination for the elucidation of the exact roles of metal ions included in N-CAs is beyond the scope of this work.



The authors declare no competing financial interest.



ACKNOWLEDGMENTS This computational work was performed under Projects 10000829-0 and 10000832-0 at the New Energy and Industrial Technology Development Organization (NEDO) using the supercomputing facilities at JAEA, the Center for Information Science of JAIST, and the TSUBAME grid cluster at the Global Scientific Information and Computing Center of the Tokyo Institute of Technology.



(1) Maldonado, S.; Stevenson, K. J. Influence of Nitrogen Doping on Oxygen Reduction Electrocatalysis at Carbon Nanofiber Electrodes. J. Phys. Chem. B 2005, 109, 4707−4716. (2) Ozaki, J.; Nozawa, K.; Yamada, K.; Uchiyama, Y.; Yoshimoto, Y.; Furuichi, A.; Yokoyama, T.; Oya, A.; Brown, L. J.; Cashion, J. D. Structures, physicochemical properties and oxygen reduction activities of carbons derived from ferrocene-poly(furfuryl alcohol) mixtures. J. Appl. Electrochem. 2005, 36, 239−247. (3) Sidik, R. A.; Anderson, A. B.; Subramanian, N. P.; Kumaraguru, S. P.; Popov, B. N. O2 Reduction on Graphite and Nitrogen-Doped Graphite: Experiment and Theory. J. Phys. Chem. B 2006, 110, 1787− 1793. (4) Bashyam, R.; Zelenay, P. A class of non-precious metal composite catalysts for fuel cells. Nature 2006, 443, 63−66. (5) Ozaki, J.; Anahara, T.; Kimura, N.; Oya, A. Simultaneous doping of boron and nitrogen into a carbon to enhance its oxygen reduction activity in proton exchange membrane fuel cells. Carbon 2006, 44, 3358−3361. (6) Ozaki, J.; Tanifuji, S.; Kimura, N.; Furuichi, A.; Oya, A. Enhancement of oxygen reduction activity by carbonization of furan resin in the presence of phthalocyanines. Carbon 2006, 44, 1324− 1326. (7) Ozaki, J.; Kimura, N.; Anahara, T.; Oya, A. Preparation and oxygen reduction activity of BN-doped carbons. Carbon 2007, 45, 1847−1853. (8) Matter, P. H.; Wang, E.; Arias, M.; Biddinger, E. J.; Ozkan, U. S. Oxygen reduction reaction activity and surface properties of nanostructured nitrogen-containing carbon. J. Mol. Catal. A: Chem. 2007, 264, 73−81. (9) Shao, Y.; Sui, J.; Yin, G.; Gao, Y. Nitrogen-doped carbon nanostructures and their composites as catalytic materials for proton exchange membrane fuel cell. Appl. Catal. B: Environ. 2008, 79, 89−99. (10) Subramanian, N. P.; Li, X.; Nallathambi, V.; Kumaraguru, S. P.; Colon-Mercado, H.; Wu, G.; Lee, J.-W.; Popov, B. N. Nitrogenmodified carbon-based catalysts for oxygen reduction reaction in polymer electrolyte membrane fuel cells. J. Power Sources 2009, 188, 38−44. (11) Gong, K.; Du, F.; Xia, Z.; Durstock, M.; Dai, L. Nitrogen-Doped Carbon Nanotube Arrays with High Electrocatalytic Activity for Oxygen Reduction. Science 2009, 323, 760−764. (12) Lefèvre, M.; Proietti, E.; Jaouen, F.; Dodelet, J.-P. Iron-Based Catalysts with Improved Oxygen Reduction Activity in Polymer Electrolyte Fuel Cells. Science 2009, 324, 71−74. (13) Liu, G.; Li, X.; Ganesan, P.; Popov, B. N. Development of nonprecious metal oxygen-reduction catalysts for PEM fuel cells based on N-doped ordered porous carbon. Appl. Catal. B: Environ. 2009, 93, 156−165. (14) Wang, X.; Li, X.; Zhang, L.; Yoon, Y.; Weber, P. K.; Wang, H.; Guo, J.; Dai, H. N-Doping of Graphene Through Electrothermal Reactions with Ammonia. Science 2009, 324, 768−771. (15) Biddinger, E. J.; Ozkan, U. S. Role of Graphitic Edge Plane Exposure in Carbon Nanostructures for Oxygen Reduction Reaction. J. Phys. Chem. C 2010, 114, 15306−15314. (16) Liu, G.; Li, X.; Ganesan, P.; Popov, B. N. Studies of oxygen reduction reaction active sites and stability of nitrogen-modified

ASSOCIATED CONTENT

S Supporting Information *

The local density of states for the three kinds of binary-doped GNRs that we considered. This material is available free of charge via the Internet at http://pubs.acs.org.



REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: +81-791-58-2622. Fax: +81-791-58-0311. 17623

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

carbon composite catalysts for PEM fuel cells. Electrochim. Acta 2010, 55, 2853−2858. (17) Qu, L.; Liu, Y.; Baek, J.-B.; Dai, L. Nitrogen-Doped Graphene as Efficient Metal-Free Electrocatalyst for Oxygen Reduction in Fuel Cells. ACS Nano 2010, 4, 1321−1326. (18) Rao, C. V.; Cabrera, C. R.; Ishikawa, Y. In Search of the Active Site in Nitrogen-Doped Carbon Nanotube Electrodes for the Oxygen Reduction Reaction. J. Phys. Chem. Lett. 2010, 1, 2622−2627. (19) Yu, D.; Nagelli, E.; Du, F.; Dai, L. Metal-Free Carbon Nanomaterials Become More Active than Metal Catalysts and Last Longer. J. Phys. Chem. Lett. 2010, 1, 2165−2173. (20) Olson, T. S.; Pylypenko, S.; Fulghum, J. E.; Atanassov, P. Bifunctional Oxygen Reduction Reaction Mechanism on NonPlatinum Catalysts Derived from Pyrolyzed Porphyrins. J. Electrochem. Soc. 2010, 157, B54−B63. (21) Xiong, W.; Du, F.; Liu, Y.; Perez, A.; Supp, M.; Ramakrishnan, T. S.; Dai, L.; Jiang, L. 3-D Carbon Nanotube Structures Used as High Performance Catalyst for Oxygen Reduction Reaction. J. Am. Chem. Soc. 2010, 132, 15839−15841. (22) Jaouen, F.; Proietti, E.; Lefèvre, M.; Chenitz, R.; Dodelet, J.-P.; Wu, G.; Chung, H. T.; Johnston, C. M.; Zelenay, P. Recent advances in non-precious metal catalysis for oxygen-reduction reaction in polymer electrolyte fuel cells. Energy Environ. Sci. 2011, 4, 114−130. (23) Wu, G.; More, K. L.; Johnston, C. M.; Zelenay, P. HighPerformance Electrocatalysts for Oxygen Reduction Derived from Polyaniline, Iron, and Cobalt. Science 2011, 332, 443−447. (24) Sheng, Z.-H.; Shao, L.; Chen, J.-J.; Bao, W.-J.; Wang, F.-B.; Xia, X.-H. Catalyst-Free Synthesis of Nitrogen-Doped Graphene via Thermal Annealing Graphite Oxide with Melamine and Its Excellent Electrocatalysis. ACS Nano 2011, 5, 4350−4358. (25) Wang, S.; Yu, D.; Dai, L.; Chang, D. W.; Baek, J.-B. Polyelectrolyte-Functionalized Graphene as Metal-Free Electrocatalysts for Oxygen Reduction. ACS Nano 2011, 5, 6202−6209. (26) Luo, Z.; Lim, S.; Tian, Z.; Shang, J.; Lai, L.; MacDonald, B.; Fu, C.; Shen, Z.; Yu, T.; Lin, J. Pyridinic N doped graphene: synthesis, electronic structure, and electrocatalytic property. J. Mater. Chem. 2011, 21, 8038−8044. (27) Yang, Z.; Yao, Z.; Li, G.; Fang, G.; Nie, H.; Liu, Z.; Zhou, X.; Chen, X.; Huang, S. Sulfur-Doped Graphene as an Efficient Metal-free Cathode Catalyst for Oxygen Reduction. ACS Nano 2011, 6, 205−211. (28) Wiggins-Camacho, J. D.; Stevenson, K. J. Mechanistic Discussion of the Oxygen Reduction Reaction at Nitrogen-Doped Carbon Nanotubes. J. Phys. Chem. C 2011, 115, 20002−20010. (29) Proietti, E.; Jaouen, F.; Lefèvre, M.; Larouche, N.; Tian, J.; Herranz, J.; Dodelet, J.-P. Iron-based cathode catalyst with enhanced power density in polymer electrolyte membrane fuel cells. Nat. Commun. 2011, 2, 416. (30) Choi, C. H.; Park, S. H.; Woo, S. I. Binary and Ternary Doping of Nitrogen, Boron, and Phosphorus into Carbon for Enhancing Electrochemical Oxygen Reduction Activity. ACS Nano 2012, 6, 7084−7091. (31) Kuroki, S.; Nabae, Y.; Chokai, M.; Kakimoto, M.; Miyata, S. Oxygen reduction activity of pyrolyzed polypyrroles studied by 15N solid-state NMR and XPS with principal component analysis. Carbon 2012, 50, 153−162. (32) Lee, Y.-H.; Li, F.; Chang, K.-H.; Hu, C.-C.; Ohsaka, T. Novel synthesis of N-doped porous carbons from collagen for electrocatalytic production of H2O2. Appl. Catal. B: Environ. 2012, 126, 208−214. (33) Liang, J.; Jiao, Y.; Jaroniec, M.; Qiao, S. Z. Sulfur and Nitrogen Dual-Doped Mesoporous Graphene Electrocatalyst for Oxygen Reduction with Synergistically Enhanced Performance. Angew. Chem., Int. Ed. 2012, 51, 11496−11500. (34) Oh, H.-S.; Kim, H. The role of transition metals in non-precious nitrogen-modified carbon-based electrocatalysts for oxygen reduction reaction. J. Power Sources 2012, 212, 220−225. (35) Wu, G.; Mack, N. H.; Gao, W.; Ma, S.; Zhong, R.; Han, J.; Baldwin, J. K.; Zelenay, P. Nitrogen-Doped Graphene-Rich Catalysts Derived from Heteroatom Polymers for Oxygen Reduction in

Nonaqueous Lithium-O2 Battery Cathodes. ACS Nano 2012, 6, 9764−9776. (36) Yang, S.; Zhi, L.; Tang, K.; Feng, X.; Maier, J.; Mullen, K. Efficient Synthesis of Heteroatom (N or S)-Doped Graphene Based on Ultrathin Graphene Oxide-Porous Silica Sheets for Oxygen Reduction Reactions. Adv. Funct. Mater. 2012, 22, 3634−3640. (37) Zhang, H.-J.; Yuan, X.; Wang, Z.; Yang, J.; Ma, Z.-F. Pyrolyzed iron-triethylenetetramine on carbon as catalyst for oxygen reduction reaction. Electrochim. Acta 2013, 87, 599−605. (38) Xu, J. B.; Zhao, T. S. Mesoporous carbon with uniquely combined electrochemical and mass transport characteristics for polymer electrolyte membrane fuel cells. RSC Adv. 2013, 3, 16−24. (39) Peng, H.; Mo, Z.; Liao, S.; Liang, H.; Yang, L.; Luo, F.; Song, H.; Zhong, Y.; Zhang, B. High Performance Fe− and N- Doped Carbon Catalyst with Graphene Structure for Oxygen Reduction. Sci. Rep. 2013, 3, 1765. (40) Chung, H. T.; Won, J. H.; Zelenay, P. Active and stable carbon nanotube/nanoparticle composite electrocatalyst for oxygen reduction. Nat. Commun. 2013, 4, 1922. (41) Cheon, J. Y.; Kim, T.; Choi, Y.; Jeong, H. Y.; Kim, M. G.; Sa, Y. J.; Kim, J.; Lee, Z.; Yang, T.-H.; Kwon, K.; et al. Ordered mesoporous porphyrinic carbons with very high electrocatalytic activity for the oxygen reduction reaction. Sci. Rep. 2013, 3, 2715. (42) Feng, L.; Yang, L.; Huang, Z.; Luo, J.; Li, M.; Wang, D.; Chen, Y. Enhancing Electrocatalytic Oxygen Reduction on Nitrogen-Doped Graphene by Active Sites Implantation. Sci. Rep. 2013, 3, 3306. (43) Tanabe, Y.; Yasuda, E. Carbon alloys. Carbon 2000, 38, 329− 334. (44) Reddy, A. L. M.; Srivastava, A.; Gowda, S. R.; Gullapalli, H.; Dubey, M.; Ajayan, P. M. Synthesis of Nitrogen-Doped Graphene Films for Lithium Battery Application. ACS Nano 2010, 4, 6337−6342. (45) Wei, D.; Liu, Y.; Wang, Y.; Zhang, H.; Huang, L.; Yu, G. Synthesis of N-Doped Graphene by Chemical Vapor Deposition and Its Electrical Properties. Nano Lett. 2009, 9, 1752−1758. (46) Niwa, H.; Kobayashi, M.; Horiba, K.; Harada, Y.; Oshima, M.; Terakura, K.; Ikeda, T.; Koshigoe, Y.; Ozaki, J.; Miyata, S. X-ray photoemission spectroscopy analysis of N-containing carbon-based cathode catalysts for polymer electrolyte fuel cells. J. Power Sources 2011, 196, 1006−1011. (47) Niwa, H.; Horiba, K.; Harada, Y.; Oshima, M.; Ikeda, T.; Terakura, K.; Ozaki, J.; Miyata, S. X-ray absorption analysis of nitrogen contribution to oxygen reduction reaction in carbon alloy cathode catalysts for polymer electrolyte fuel cells. J. Power Sources 2009, 187, 93−97. (48) Zhang, L.-S.; Liang, X.-Q.; Song, W.-G.; Wu, Z.-Y. Identification of the nitrogen species on N-doped graphene layers and Pt/NG composite catalyst for direct methanol fuel cell. Phys. Chem. Chem. Phys. 2010, 12, 12055−12059. (49) Zhong, J.; Deng, J.-J.; Mao, B.-H.; Xie, T.; Sun, X.-H.; Mou, Z.G.; Hong, C.-H.; Yang, P.; Wang, S.-D. Probing solid state N-doping in graphene by X-ray absorption near-edge structure spectroscopy. Carbon 2012, 50, 335−338. (50) Hou, Z.; Wang, X.; Ikeda, T.; Huang, S.-F.; Terakura, K.; Boero, M.; Oshima, M.; Kakimoto, M.; Miyata, S. Effect of Hydrogen Termination on Carbon K-Edge X-ray Absorption Spectra of Nanographene. J. Phys. Chem. C 2011, 115, 5392−5403. (51) Wang, X.; Hou, Z.; Ikeda, T.; Huang, S.-F.; Terakura, K.; Boero, M.; Oshima, M.; Kakimoto, M.; Miyata, S. Selective nitrogen doping in graphene: Enhanced catalytic activity for the oxygen reduction reaction. Phys. Rev. B 2011, 84, 245434. (52) Wang, X.; Hou, Z.; Ikeda, T.; Oshima, M.; Kakimoto, M.; Terakura, K. Theoretical Characterization of X-ray Absorption, Emission, and Photoelectron Spectra of Nitrogen Doped along Graphene Edges. J. Phys. Chem. A 2012, 117, 579−589. (53) Car, R.; Parrinello, M. Unified Approach for Molecular Dynamics and Density-Functional Theory. Phys. Rev. Lett. 1985, 55, 2471−2474. (54) CPMD; IBM Corp, 1990−2008; MPI für Festkörperforshung Stuttgart, 1997−2001. http://www.cpmd.org. 17624

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625

The Journal of Physical Chemistry C

Article

(55) Becke, A. D. Density-functional exchange-energy approximation with correct asymptotic behavior. Phys. Rev. A 1988, 38, 3098−3100. (56) Lee, C.; Yang, W.; Parr, R. G. Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Phys. Rev. B 1988, 37, 785−789. (57) Grimme, S. Semiempirical GGA-type density functional constructed with a long-range dispersion correction. J. Comput. Chem. 2006, 27, 1787−1799. (58) Troullier, N.; Martins, J. L. Efficient pseudopotentials for planewave calculations. Phys. Rev. B 1991, 43, 1993−2006. (59) Laasonen, K.; Sprik, M.; Parrinello, M.; Car, R. Ab initio” liquid water. J. Chem. Phys. 1993, 99, 9080−9089. (60) Sprik, M.; Ciccotti, G. Free energy from constrained molecular dynamics. J. Chem. Phys. 1998, 109, 7737−7744. (61) Sprik, M. Computation of the pK of liquid water using coordination constraints. Chem. Phys. 2000, 258, 139−150. (62) Ikeda, T.; Boero, M.; Huang, S.-F.; Terakura, K.; Oshima, M.; Ozaki, J. Carbon Alloy Catalysts: Active Sites for Oxygen Reduction Reaction. J. Phys. Chem. C 2008, 112, 14706−14709. (63) Ikeda, T.; Boero, M.; Huang, S.-F.; Terakura, K.; Oshima, M.; Ozaki, J.; Miyata, S. Enhanced Catalytic Activity of Carbon Alloy Catalysts Codoped with Boron and Nitrogen for Oxygen Reduction Reaction. J. Phys. Chem. C 2010, 114, 8933−8937. (64) Nosé, S. A unified formulation of the constant temperature molecular dynamics methods. J. Chem. Phys. 1984, 81, 511−519. (65) Hoover, W. G. Canonical dynamics: Equilibrium phase-space distributions. Phys. Rev. A 1985, 31, 1695−1697. (66) Nørskov, J. K.; Rossmeisl, J.; Logadottir, A.; Lindqvist, L.; Kitchin, J. R.; Bligaard, T.; Jonsson, H. Origin of the Overpotential for Oxygen Reduction at a Fuel-Cell Cathode. J. Phys. Chem. B 2004, 108, 17886−17892. (67) Huang, S.-F.; Terakura, K.; Ozaki, T.; Ikeda, T.; Boero, M.; Oshima, M.; Ozaki, J.; Miyata, S. First-principles calculation of the electronic properties of graphene clusters doped with nitrogen and boron: Analysis of catalytic activity for the oxygen reduction reaction. Phys. Rev. B 2009, 80, 235410. (68) Okamoto, Y. First-principles molecular dynamics simulation of O2 reduction on nitrogen-doped carbon. Appl. Surf. Sci. 2009, 256, 335−341. (69) Wassmann, T.; Seitsonen, A. P.; Saitta, A. M.; Lazzeri, M.; Mauri, F. Structure, Stability, Edge States, and Aromaticity of Graphene Ribbons. Phys. Rev. Lett. 2008, 101, 096402. (70) Wassmann, T.; Seitsonen, A. P.; Saitta, A. M.; Lazzeri, M.; Mauri, F. Clar’s Theory, π-Electron Distribution, and Geometry of Graphene Nanoribbons. J. Am. Chem. Soc. 2010, 132, 3440−3451. (71) VandeVondele, J.; Sprik, M. A molecular dynamics study of the hydroxyl radical in solution applying self-interaction-corrected density functional methods. Phys. Chem. Chem. Phys. 2005, 7, 1363−1367. (72) Oberhofer, H.; Blumberger, J. Charge constrained density functional molecular dynamics for simulation of condensed phase electron transfer reactions. J. Chem. Phys. 2009, 131, 064101.

17625

dx.doi.org/10.1021/jp5038365 | J. Phys. Chem. C 2014, 118, 17616−17625