Toward Design of Synergistically Active Carbon-Based Catalysts for Electrocatalytic Hydrogen Evolution

Replacement of precious Pt catalyst with cost-effective alternatives would be significantly beneficial for hydrogen production via electrocatalytic hydrogen evolution reaction (HER). All candidates thus far are exclusively metallic catalysts, which suffer inherent corrosion and oxidation susceptibility during acidic proton-exchange membrane electrolysis. Herein, based on theoretical predictions, we designed and synthesized nitrogen (N) and phosphorus (P) dual-doped graphene as a nonmetallic electrocatalyst for sustainable and efficient hydrogen production. The N and P heteroatoms could coactivate the adjacent C atom in the graphene matrix by affecting its valence orbital energy levels to induce a synergistically enhanced reactivity toward HER. As a result, the dual-doped graphene showed higher electrocatalytic HER activity than single-doped ones and comparable performance to some of the traditional metallic catalysts.

lectrocatalytic reduction of water to molecular hydrogen via hydrogen evolution reaction (HER) provides a promising solution to energy supplies in the future. The corresponding devices always require noble metal catalysts, such as Pt, to facilitate low overpotential and fast kinetics for practical applications. 1,2 However, the high cost as well as low abundance of Pt limit their widespread use. Therefore, replacement of noble metals with costeffective but highly efficient alternatives is desirable to achieve large-scale energy production. In recent years, a wide variety of transition metals (Co, Ni, Fe, Mo) and derivative components have been selected as effective candidates; however, the inherent corrosion and oxidation susceptibility largely limits their utilization in acidic proton-exchange membrane-based electrolysis for sustainable hydrogen production. 3À9 On the other hand, various carbon-based materials feature unique advantages for designated catalysis due to their tunable molecular structures, abundance, and strong tolerance to acidic/alkaline environments. Very recent advances in low-dimensional carbon materials, as metal-free catalysts, have shown their promising future in energy-related electrocatalytic oxygen reduction/evolution reactions (ORR/OER).
In the case of ORR, we have found that some carbon-based catalysts showed comparable electrocatalytic activity with that of precious Pt and further possessed enhanced stability and fuel tolerance, which make them as competitive alternatives to Pt. 10,11 Generally, the engineering of pristine graphene by chemical substitution of some carbon atoms with heteroatoms, such as N, B, P, F, and S, is an effective way to tailor its electronic structure and (electro)chemical properties. 12À17 In particular, the codoping of two elements with reverse electronegativity to that of carbon, e.g., B and N, could lead to a unique electron-donor property of carbon by the so-called synergistic coupling effect between two heteroatoms and, consequently, largely boost its ORR activity. 10,11 However, current knowledge about electrocatalytic HER on graphene surface is very inconclusive, and the systemic catalysts' molecular design toward dual-doped graphenes for highly efficient HER is still lacking.
Although graphene-based materials have been extensively studied in electrochemical applications such as fuel cells, supercapacitors, and Li batteries, 18 there are few reports about their application in electrocatalytic hydrogen production, and more importantly, no theoretical calculations have been performed to predict their unexplored electrocatalytic capacities in relation to this key energy conversion process. 19 Herein, we present our attempt to design a dualdoped graphene catalyst with a synergistically enhanced HER activity to extend the electrochemical applications of graphene materials. First, we used density functional theory (DFT) calculations to explore several nonmetallic element-doped graphene models, followed by selecting one couple of heteroatoms (N and P) as codopants with remarkable contrast in their charge population in graphene matrix. By downshifting the valence bands of active carbons in graphene, the codoping of N and P indeed resulted in the highest HER activity among various doped graphene models. On the basis of the theoretical prediction, the proof-ofconcept studies were carried out by simultaneously incorporating N and P heteroatoms into the graphene matrix through a typical chemical-doping procedure. The resultant N,P-graphene catalyst showed much lower HER overpotential and higher exchange current density than those of all investigated pure and doped graphene samples. The newly developed nonmetallic catalyst also showed robust stability and applicability in a wide range of pH values. These findings provide solid evidence, both theoretical and experimental, that similar to precious metals, the well-designed carbon-based counterparts also have a great potential as highly efficient HER catalysts.

RESULTS AND DISCUSSION
To explore the effects of various dopants in graphene toward HER activity, we carried out systematic studies on the electronic properties of differently doped graphene models by DFT calculations. We selected six nonmetallic heteroatoms (N, B, O, S, P, F) with various electron negativities to investigate their effects on the adjacent C atoms, which act as the active sites for HER (Supporting Information, Figure S1). All the proposed doping configurations were constructed on the basis of the reported experimental X-ray photoelectron spectra (XPS) data by taking into account the chemical environment of each heteroatom. 20 The natural bond orbital (NBO) population analyses (Figure 1a) show that six dopants possess various charge populations in graphene matrix: N and O are negatively charged (act as electron acceptors for the adjacent C), while F, S, B, and P are positively charged (act as electron donors). Therefore, we selected one couple of the investigated atoms with the most noticeable differences in the charge population (i.e., N and P) as codopants for graphene to maximally activate the adjacent C atom by tailoring its electron donorÀacceptor property and consequently enhance its HER activity.
The HER activities of a series of the proposed N-and/or P-doped graphene clusters were then theoretically ARTICLE predicted by the HER free energy diagram ( Figure 1b). Generally, the overall HER pathway, simply written as 2H þ þ 2e À f H 2 , can be described by a three-state diagram comprising of an initial state H þ þ e À , an intermediate adsorbed H*, and the final product 1 / 2 H 2 . The Gibbs free-energy of the hydrogen adsorption, |ΔG H* |, has been considered as the major descriptor of the HER activity for a wide variety of metal surfaces. 21,22 The optimal value of |ΔG H* | should be zero; if ΔG H* is positive, the formation of intermediate H* would be the rate-determining step (RDS) of the whole HER; while if ΔG H* is negative, desorption of H* to form H 2 is the RDS. 22 Among graphene-based models, pure graphene shows a highly positive ΔG H* value (1.85 eV), representing a negligible HER activity. Both N and/or P incorporation could reduce the ΔG H* values to enhance the initial H* adsorption, in which pyridinic N and P dual-doped model (pN,P-G) demonstrates the lowest |ΔG H* | value of 0.08 eV (Figure 1b and Supporting Information, Figure S2), indicating its highest HER activity with the most favorable H* adsorptionÀdesorption property. In contrast to most of the single-doped graphenes which have a positive ΔG H* (Supporting Information, Figure S3), the negative ΔG H* of the dual-doped pN,P-G indicates a relatively strong H* adsorption on its surface; for such a model, the C atom between meta-type pN and P in the heteroring acts as the most active site for H* adsorption (Supporting Information, Figure S4).
We further explored the origin of the different H* adsorption behavior on these doped graphenes using the molecular orbital theory. 23 As shown in Figure 1c, the ΔG H* values of all investigated models show a linear trend in relation to E diff (E diff is defined as the difference between the lowest valence orbital energy of C active center and the highest valence orbital energy of the graphene cluster). The principle underlying this relationship is that the valence band (v) of the active C atom hybridizes with the bonding (σ) orbital of the adsorbed H* to form bonding (v-σ) and antibonding (v-σ)* states, as illustrated in the Supporting Information (Scheme S1). An increase in the filling of the (vÀσ) state, as induced by lower C valence orbitals, could lead to a stronger bonding between H* and active C to reduce the ΔG H* . Among all investigated models, the pN,P-G cluster shows the highest value of E diff , which corresponds to the lowest value of ΔG H* and apparently the highest predicted HER activity (Figure 1c).
To verify the theoretical predictions, we incorporated N and P in graphene simultaneously by annealing chemically exfoliated graphene oxide (GO) powder with a melamine (N source) and triphenylphosphine (P source) mixture at 950°C in Ar atmosphere. During the reduction process of GO, N and P were chemically doped into the defective sites of graphene by substituting C or O atoms, yielding a doping concentration of 4.60 atom % of N and 1.63 atom % of P, respectively (based on the XPS results). As control samples, the single-doped N-or P-graphenes were respectively synthesized by the same procedure with addition of a single respective precursor (Supporting Information, Figure S5a,b). The nanosheet morphology of the pristine GO was well preserved in the resultant N,Pgraphene-1 without noticeable residues of the solid precursors, as shown in the scanning electron microscope (SEM) and transmission electron microscopy (TEM) images (Figure 2a,b). The electron energy loss spectrum (EELS) elemental mapping shows a homogeneous distribution of the various dopants in the graphene matrix (Supporting Information, Figure S6). The nature of doping was first identified by near-edge X-ray adsorption fine structure spectroscopy (NEXAFS) as shown in Figure 2c,d. In the C K-edge (Figure 2c), all three N-and/or P-doped samples exhibited typical graphitic-sp 2 carbon structures with strong π* and σ* resonances at ∼285.5 and ∼292.0 eV, representing the core level electron transitions to the unoccupied antibonding π* and σ* orbitals, respectively. The successful N incorporation was revealed by the N K-edge NEXAFS spectra (Figure 2d), showing three π* resonances representing NÀC bonds in the form of pyridine (398.5 eV), pyrrole (399.7 eV), and graphite (401.2 eV) structures, respectively, also consistent with the XPS  (Figure 2e). 14 Because the π* resonances of CÀC and CÀP species have an overlapping position in the C K-edge NEXAFS (only a weak shoulder at ∼283.2 eV that can be attributed to the π* CÀP resonance in P-containing N,P-graphene-1 and P-graphene samples but not N-graphene, as shown by the arrow in Figure 2c), the doping of P in N,P-graphene-1 was confirmed by Synchrotron-based high-resolution XPS (Figure 2f), indicating that P heteroatom is mainly bonded to C or O atoms, along with some Ph 3 P precursor residue. 15 Note that both NEXAFS and XPS spectra indicate that N and P heteroatoms were not bonded to each other to form any PÀN species (i.e., phosphorus nitride, P 3 N 5 ) in or out of the graphene basal plane which was predicted has worst HER activity than meta and para site PÀCÀN moieties (Supporting Information, Figure S4).
The electrocatalytic activities of the as-synthesized N-and/or P-doped graphenes toward HER were examined in both acidic (0.5 M H 2 SO 4 ) and alkaline (0.1 M KOH) aqueous solutions. Figure 3a shows the HER polarization curves of various electrocatalysts in 0.5 M H 2 SO 4 solution. Graphite (as the example of pure graphene without any heteroatom doping) had a negligible electrocatalytic activity, while either single N or P doping could significantly enhance the HER activity of graphene by generating additional active sites, which is consistent with theoretical prediction (Figure 1b). To achieve a 10 mA/cm 2 HER current density, the dual-doped N,P-graphene-1 required an overpotential of ∼420 mV, lower than those of both single-doped N and P-graphenes (Supporting Information, Table S1). It should be noted that the HER activity of the N,P-graphene-1 is comparable to that of some traditional metallic catalysts, such as bulk Au, Mo and Mo/Ni alloy, etc., 24À26 while it is not as good as that of the state-of-the-art nanostructured MoS 2 /WS 2 and NiÀMo alloy catalysts. This may be due to the bulk nature and the lack of a large population of active sites because of the stacking graphene layers during the N,P-graphene-1 electrode preparation (drying) process. Both of these factors result in a relatively lower electrochemically active surface area on N,P-graphene-1 than other nanostructured catalysts. However, the comparable normalized HER exchange current densities of N,P-graphene-1 and the well-studied metallic catalysts (nanostructured MoS 2 for example, 7,27À30 see Figure S7 and Table S2 in the Supporting Information for detailed normalization procedure) indicate that the graphenebased materials have the potential to achieve comparable activity to the metallic catalysts via nanostructure engineering/optimizing.
The electrocatalytic activity enhancement in the case of N,P-graphene-1 as compared to single-doped samples could be fundamentally revealed by exploring their HER mechanisms. As shown in Figure 3b, all N-and/or P-doped graphenes showed Tafel slopes in a range of ∼90 to ∼130 mV/decade (N,P-graphene-1 possessed the lowest value of 91 mV/decade), indicating that an initial proton adsorption (i.e., Volmer step: H þ þ e À f H*) was the rate-determining step of the whole HER process. 3 These behaviors could be theoretically explained by thermodynamic calculations (Figure 1b and Supporting Information, Figure S3), in which almost all investigated graphene-based catalysts demonstrate positive ΔG H* values related to a weak H* adsorption. 22 On the contrary, the N,Pgraphene model shows a slightly negative ΔG H* with an enhanced capability to adsorb hydrogen which therefore can accelerate the most sluggish step (H* adsorption) of the whole HER process. As the result, a higher HER activity was experimentally observed.
In contrast to traditional metallic and the state-ofthe-art nanostructured MoS 2 electrocatalysts, which are active under either acidic or basic conditions, 4,31,32 the newly developed graphene-based catalysts showed favorable HER activity in a wide range of pH values. As expected, in alkaline solution, the HER activities of all four graphene-based electrocatalysts were lower than those in acidic solutions but the activity followed the same trend (Figure 3c). The difference may arise from the different HER mechanism under the two conditions, 33 e.g., there is an additional water dissociation step in the alkaline solution and graphene-based catalysts is unfavorable to ARTICLE facilitate this step therefore resulting in a relative high energy barrier in the whole HER process.
The synergistic coupling effect in N,P-graphene-1 was quantitatively evaluated in terms of the exchange current density (i 0 ) derived from the Tafel plots. As summarized in Figure 3e, the dual-doped N,P-graphene-1 showed i 0 = 2.4 Â 10 À7 A/cm 2 and 4.0 Â 10 À10 A/cm 2 in acidic and alkaline solutions, which surpassed the summed value of single doped N-graphene and P-graphene in respective solutions (Supporting Information, Table S1). Such a difference indicates that the enhanced electrocatalytic activity of N,P-graphene-1 is not caused by a simple increase in the number of active sites by codoping N and P heteroatoms but represents the synergistic effect that boost the capacity of proton adsorption and reduction, as revealed by the aforementioned DFT calculations. To gain more insight into the relationship between the apparent HER performance and the theoretical activity descriptor for newly developed carbon-based catalysts, the experimentally measured i 0 was correlated with the DFT-derived ΔG H* , which is commonly used in the study of metal surfaces. 21 As shown in Figure 3f, a good linear trend between the theoretical and experimental values is clearly visible (excluding P-graphene), which also validates the predictive capability of the employed DFT model beyond metals. 21 Regarding P-graphene, different from the cases of other three catalysts, the sample was highly hydrophobic hence the mass transport limitation might be the major contributor for its poor HER activity.
One of significant features of carbon materials is their strong tolerance to both acidic and alkaline solution. As expected, the N,P-graphene-1 catalyst with strong covalent CÀN and CÀP bonds featured a robust stability in both 0.5 M H 2 SO 4 and 0.1 M KOH solutions without corrosion and oxidation, which is required for a sustainable hydrogen production ( Figure 4a). As a control sample, two-step synthesized N,P-graphene-2 (see the Experimental Section for a detailed synthesizing procedure) showed lower HER activity than that of one-step synthesized N,P-graphene-1 (Figure 4b), due to the different N and P doping concentrations (Supporting Information, Figure S5c,d). Also note that not each graphene codoped by a random couple of heteroatoms could achieve an enhanced HER activity. As shown in Figure 4b, the boron-and nitrogencodoped graphene (B,N-graphene), which has demonstrated a synergistic catalysis toward electrocatalytic ORR activity enhancement, 10 unexpectedly showed lower HER activity than that of single-doped N-graphene and N, P-graphene-1. These experimental findings indicate the N/P dual doping is crucial for the favorable HER activity in the case of N,P-graphene-1. Additionally, the dual-doped graphene electrocatalysts showed higher HER activity than all of investigated single-doped ones (N-, B-, O-, S-, P-doped graphenes) in both acidic and alkaline solutions (Figure 4c,d), unambiguously indicating that the synergistic effect of dual doping in N,P-graphene-1 led to its enhanced electrocatalytic activity.

CONCLUSIONS
In summary, based on the DFT-calculations, we designed and synthesized N and P dual-doped graphene as a metal-free catalyst for electrocatalytic HER, which exhibited a significantly improved electrochemical performance as compared to single doped counterparts due to the synergistic effect confirmed by both electrochemical measurements and theoretical calculations. By applying HER as a test reaction, which is also considered as a cornerstone to explore the mechanism of other multielectron transfer processes, 6 we expect that this study will largely expand the spectrum of catalysts for other energy-related electrocatalytic reactions.

EXPERIMENTAL SECTION
Computational Screening Procedure. The explored graphene models (as shown in Supporting Information, Figure S1) are based on C 42 H 16 clusters which have been previously adopted for investigation of ORR on boron and nitrogen doped graphene. 10 The geometry optimizations of these molecular models were carried out using density functional theory (DFT) calculations provided by Gaussian 09. 34 All the calculations were performed at UB3LYP/6-31G(d,p) level of chemistry with all atoms fully relaxed. The solvent (water) effect was considered by the Polarizable