rXXXX American Chemical Society A dx.doi.org/10.1021/jp201991j | J. Phys. Chem. C XXXX, XXX, 000–000

ARTICLE

pubs.acs.org/JPCC

Mechanisms of Oxygen Reduction Reaction onNitrogen-Doped Graphene for Fuel CellsLipeng Zhang‡ and Zhenhai Xia*,†

†Department of Materials Science and Engineering and Department of Chemistry, University of North Texas, Denton, Texas 76203,United States‡Department of Mechanical Engineering, University of Akron, Akron, Ohio 44325, United States

1. INTRODUCTION

Fuel cells can directly convert chemical energy into electricenergy with high conversion efficiency, high power density,quiet operation, and no pollution. Among many factors affect-ing the chemical-electrical energy conversion, oxygen reductionreaction (ORR) on cathode is the pivot in fuel cell. Thisreaction is a kinetically slow process,1 which dominates theoverall performance of a fuel cell. The ORR can proceedthrough two ways. One is a direct four-electron pathway, inwhich O2 is reduced directly to water without involvement ofhydrogen peroxide, O2 þ 4Hþ þ 4e� f 2H2O. The other is aless efficient two-step two-electron pathway in which hydrogenperoxide is formed as an intermediate, O2 þ 2Hþ þ 2e� fH2O2. To achieve a high efficiency fuel cell, the four-electronpathway is expected to occur. Because the ORR process is veryslow in nature, catalysts must be used to facilitate the four-electron pathway to boost the efficiency of fuel cells. Tradi-tionally, such electrocatalysts are platinum and its alloys,2�4 butthey are expensive and susceptible to time-dependent drift5 andCO poisoning,6 which limits large-scale application of the fuelcell. There have been intensive research efforts,7�11 to reduceor replace Pt and Pt based alloys electrodes in fuel cell.Recently, it has been demonstrated that vertically alignednitrogen-containing carbon nanotubes (VA-NCNTs)12 andnitrogen-containing graphene sheets (N-graphene),13 reducedgraphene oxide/platinum supported electrocatalysts (Pt/RGO),14

and stabilizing metal catalysts at metal�metal oxide�graphenetriple junctions (Pt-ITO-graphene)15 show a much better electro-catalytic activity, with long-term operation stability, and tolerance to

crossover and poison effects, than platinum electrodes forORR. Butthe detailed electrocatalytical mechanisms of these nitrogen dopedcarbon nanomaterials remains unclear. A fundamental understand-ing of the catalytic mechanism will provide guideline for furtherincreasing the efficiency of these catalysts and discovering newcatalysts.

There have been some reports on mechanisms of ORR onelectrode of fuel cells. A suitable mechanism table was formulatedfor the prominent pathway of ORR in proton exchange mem-brane (PEM) fuel cell and the kinetics of the proposed none-lectrochemical reactions on platinum were studied using B3LYPdensity functional theory (DFT).16 The DFTmethods were alsoemployed to elucidate the mechanisms of ORR on carbonsupported Fe-phthalocyanine (FePc/C) and Co-ptthalocyanine(CoPc/C) catalysts in 0.1 M NaOH solution.17 Anderson et al.studied the oxygen reduction on graphene, nitrogen-dopedgraphene and cobalt�graphene�nitride systems18�20 usingB3LYP hybrid DFT method. Electroreduction of oxygen tohydrogen peroxide was presented in their study, which showstwo-electron pathway. Another simulation method was alsoapplied to study the ORR mechanism on electrodes for fuel cell.Car�Parrinello molecular dynamics simulations had been per-formed to investigate the ORR on a Pt (111) surface.21 However,to our knowledge, there are no reports about the ORR mechan-isms of the four-electron pathway on the catalytic electrode of

Received: March 1, 2011Revised: May 3, 2011

ABSTRACT: Graphene and its derivatives are attractive forelectrocatalytical application in fuel cells because of their uniquestructures and electronic properties. The electrocatalyticalmechanism of nitrogen doped graphene in acidic environmentwas studied by using density functional theory (DFT). Thesimulations demonstrate that the oxygen reduction reaction(ORR) on N-doped graphene is a direct four-electron pathway,which is consistent with the experimental observations. Theenergy calculated for each ORR step shows that the ORR canspontaneously occur on the N-graphene. The active catalyticalsites on single nitrogen doped graphene are identified, whichhave either high positive spin density or high positive atomiccharge density. The nitrogen doping introduces asymmetry spin density and atomic charge density, making it possible forN-graphene to show high electroncatalytic activities for the ORR.

B dx.doi.org/10.1021/jp201991j |J. Phys. Chem. C XXXX, XXX, 000–000

The Journal of Physical Chemistry C ARTICLE

nitrogen doped graphene. In this paper, the DFT method wasused to study the mechanism process of ORR on nitrogen dopedgraphene. The electrocatalytic active sites and the processes ofelectron transformation were examined. Energy variations werecalculated during each reaction steps. The electrocatalytic activityof nitrogen doped graphene is related to electron spin densityand atomic charge density distribution on it. Finally, the mechan-isms of a four-electron pathway on the nitrogen-doped graphenewere analyzed.

2. SIMULATION METHOD AND MODELS

B3LYP hybrid density functional theory of Gaussian 0322 wasemployed with a basis set of 6-31G(d,p).23�26 Considering thebreaking and forming of the chemical bond, unstrict polarizationsetting27 was used in the calculation. Two different nitrogen-containing graphene sheets (C45NH20 and C45NH18) were built,containing pyridine and pyrrole species, respectively, as shown inFigure 1. For comparison, graphene sheets with the same config-uration but no N-doping (C46H20, and C46H18) were also con-structed. Carbon or nitrogen atoms on the edge of the grapheneare terminated by hydrogen atoms. The ORR processes weresimulated beginning with the first electron transmission, in whichprocess the intermediate molecule OOHwas already formed. Thisis possible because in acidic environment, O2 can adsorb an H

þ toform Hþ�O�O,21 because the whole system is charge neutral.OOHþ could be simplified toOOH, and subsequent adsorbedHþ

could be taken as H by considering the ionization potentials. Andthere are no net charges on the N-graphene.

After the first electron transfer, the product OOH was placednear the N-graphene with the OOH molecular plane parallel tothe N-graphene plane, with a distance of 3.0 Å away from thegraphene. Four-electron transformation reactions were simu-lated by keeping introducing H atoms into the system. At eachstep, the optimization structure was obtained, and adsorptionenergy for these molecules on the N-graphene was calculated.The adsorption energy is defined as the energy difference bet-ween the adsorption and the isolated systems. Here, the energyof the isolated system refers to sum of energies of fore-stepadsorbed N-graphene and the individual isolated adsorbatemolecules. Thus, negative adsorption energy indicates that theadsorbate molecules would be energetically favorable to beadducted to the surface of the N-garaphene.

3. RESULTS AND DISCUSSION

We first consider ORR behavior of the N-graphene withpyridine structure (Figure 1a). Figure 2 shows the structural

change of N-graphene C45NH20 and adsorbedmolecules for eachreaction step, at which an H atom was sequentially introducedinto the system. The variation of distance between different atomsor molecules is listed in Table 1.

In the first step of the reaction, OOH moves from initialposition (Figure 2a) to the graphene and adsorbs to a carbonatom (C37) close to the nitrogen atom (Figure 2b). The carbonatom (C37) moves out of the N-graphene plane to form atetrahedral structure, suggesting that chemical bond is formedbetween the carbon and oxygen atoms. The distance between thecarbon (C37) and oxygen (O67) atoms reduces to 1.50 Å frommore than 3.0 Å, further confirming the formation of a chemicalbond between OOH and graphene. This is an important step forN-graphene to have catalytical activities because adsorption andformation of chemical bond is necessary for the followingreactions.

A similar procedure was used to examine the adsorption ofOOH on the graphene sheet with pyrrole species and those withno doping. OOH molecule can also adsorb to a carbon atom(C16) close to the nitrogen atom, but it cannot adsorb on thepure graphene sheets. Because the OOH adsorption is a muststep for promoting ORR, pure graphene sheets do not possesscatalytical capability for fuel cell.

After the OOH adsorbs on the N-graphene, we introduceanother H to the system. Because of random nature, this Hatom may first move to the position near the oxygen atom(O67) that is bonded to the carbon atom, or the other one. Inthe former case, the subsequent reaction process is noted asReaction Path I while the latter case is Reaction Path II. We firstconsider the former case. In the simulation, the H atom isplaced near the oxygen atom O67 within a distance of O�Hbonding length, as shown in Figure 2c. In such a step, theHOO*H molecule is assumed to form and adsorb to theN-graphene. After optimization, we found that one of theoxygen atoms still bonds to the graphene at C37 in the formof OH while the other one with a hydrogen drifts away andadsorbs to another carbon atom (C14) adjacent to the nitrogen(Figure 2d). The distance between two O atoms now increasesto 2.89 Å, indicating that the bond of O�O is broken. Thus theHOO*H is not a stable product. When an H is introduced intothe system near the OOH, it leads to the decomposition ofHOO*H into two OH molecules. Similar reaction is alsoobserved on the graphene sheet with pyrrole species. This isanother important step because the break of O�Obonds repre-sents four-electron transformation pathway in ORR. Other-wise, it is a two-electron transformation pathway. Obviously,

Figure 1. Nitrogen-containing graphene sheets of (a) C45NH20 and (b) C45NH18. The larger gray circles are carbon atoms, the larger blue circle is anitrogen atom, and the smaller light white circles are hydrogen atoms; all the atoms are numbered in the circles.

C dx.doi.org/10.1021/jp201991j |J. Phys. Chem. C XXXX, XXX, 000–000

The Journal of Physical Chemistry C ARTICLE

the ORR on N-graphene is a four-electron transformationpathway, which is consistent with the experimental results.13

We further add the second H to the system near the oxygen(O67) that first bonds to the graphene (Figure 2e). Theintroduction of the H atom causes the break of C�O bond(C37�O67 bond) and the formation of the first water molecule.The distance between O67 and C37 now increases from 1.47 to3.30 Å. At the same time, another adsorbed OH molecule isstretched by the newly formed water molecule and a hydrogen

bond is formed between the hydrogen in adsorbed OH (H69)and the oxygen in the water molecule (O67). As the third H isintroduced into the system, the second water molecule is formedas shown in Figure 2h. After the two water molecules drift awayfrom the N-graphene, the N-graphene recovers to its initial state.These reactions also occur on the graphene sheet with pyrrolespecies. Here, the four-electron transformation process finishesand the N-graphene is ready for the next catalytic reaction cycle.

For the case where the first introducedH is close to the oxygenatom (O68) that is bonded to another H atom, four-electrontransformation also occurs but the subreaction paths are different,as shown in Figure 2c0�h0. Instead of producing two OH mol-ecules after the first H is introduced, one water molecule isgenerated while the oxygen atom (O67) alone adsorbs on thegraphene. Consequently, when the next twoH atoms were addednear the oxygen (O67), another water molecule is generated.With the analysis above, the reactions of the electron transforma-tion of ORR on the N-graphene are as follows:

Reaction Path I

OOH f �OOH ð1Þ

�OOHþ e� þHþ f �OHþ �OH ð2Þ

Figure 2. Initial states (a), (c), (e), (g), (c0), (e0), and (g0) and finial optimization structures (b), (d), (f), (h), (d0), (f0), and (h0) of the system for eachreaction step, showing only part of the graphene, with the large gray circles for carbon atoms, the large blue circles for nitrogen atoms, the large red circlesfor oxygen atoms, and the small light gray circles for hydrogen atoms.

Table 1. Variation of Distance between Different Atomsduring Each Reaction Step for Two Different Reaction Paths(Unit/Å)a

step 1 step 2 step 3 step 4

reaction path I D(O�O) 1.45 2.89 2.83 2.81

D(1�G) 1.50 1.47 >3.30 >3.30

D(2�G) >2.40 1.46 1.42 >3.20

reaction path II D(O�O) 1.45 2.73 2.91 2.76

D(1�G) 1.50 1.41 1.49 >3.20

D(2�G) >2.40 >3.30 >3.30 >2.80a D(O�O), the distance between two oxygen atoms; D(1�G), thenearest distance between O67 and the graphene; D(2�G), the nearestdistance between O68 and the graphene.

D dx.doi.org/10.1021/jp201991j |J. Phys. Chem. C XXXX, XXX, 000–000

The Journal of Physical Chemistry C ARTICLE

�OHðadsorbed1Þ þ e� þHþ f H2O ð3Þ

�OHðadsorbed2Þ þ e� þHþ f H2O ð4Þor

Reaction Path II

OOH f �OOH ð1aÞ

�OOHþ e� þHþ f �OþH2O ð2aÞ

�Oþ e� þHþ f OH ð3aÞ

�OHþ e� þHþ f H2O ð4aÞwhere asterisk * represents the graphene. Our simulation beganwith the reaction 1. Reactions 3 and 4 are the same reaction, but*OH adsorbed to C37 on the N-graphene in reaction 3, and inreaction 4 *OH adsorbed to C14. Adsorption energies of theeach step are calculated and the relative energy of reactionpathway is shown in Figure 3. In this figure, for the first step,the reference energy state is the total energy of optimizedN-graphene and OOH molecules. For the other reductionsteps, the reference energy states is the total energy of the pro-ducts of previous reaction step and Hþ þ e�, based on whichthe relative energy of each reaction step was calculated. Thesolid lines show the energy variation of Reaction Path I, and thedotted lines refer to that of Reaction Path II. In the first step of

the reaction in this simulation, the energy decreases by 0.85 eVwhen OOH adsorbs to the N-graphene. When the H atom wassubsequently introduced into the system, the adsorption energiesfor the following reaction steps of Reaction Path I are �4.28,�5.01, and �2.48 eV, respectively. For Reaction Path II, theadsorption energies for the following reaction steps are �3.53,�4.66, and�3.77 eV. For the second reaction step, the decreasingenergy of Reaction Path I is more than that of Reaction Path II,suggesting that the reaction favorites the production of two OHmolecules when the bond of O�O is broken. In each step ofelectron transformation, the energy becomes more negative, driv-ing the system to a more stable state. Therefore, the four-electronreaction can spontaneously take place on the nitrogen-dopedgraphene.

Why does the doped graphene have catalytic capability butpure graphene does not? This could be explained from the levelof their chemical reactivity. The highest occupied molecularorbital (HOMO)�lowest unoccupied molecular orbital(LUMO) energy separation has been used as a simple indicatorof kinetic stability. A small HOMO�LUMO gap implies lowkinetic stability and high chemical reactivity, because it isenergetically favorable to add electrons to a high-lying LUMO,to extract electrons from a low-lyingHOMO, and so to form theactivated complex of any potential reaction.28 We have calcu-lated the HOMO�LUMO gap for all those graphene sheetswith or without nitrogen doping (Table 2). For pure graphene,C46H20, the gap is 2.7 eV, which agrees with the results forsimilar size graphene, calculated by Shemella.29 The HOMO�LUMO gap reduces by two times after a nitrogen atom issubstituted into C46H20 to form C45NH20, which is alsoconsistent with results from Zheng et al.30 Hence, the chemical

Figure 3. Relative energy of two different reaction pathways of the ORR on N-graphene (C45NH20). For the first step, the reference energy state is thetotal energy of optimized N-graphene (C45NH20) and OOHmolecules, and for the other reaction steps, the reference energy states are the total energyof the product of previous reaction and Hþ þ e�.

E dx.doi.org/10.1021/jp201991j |J. Phys. Chem. C XXXX, XXX, 000–000

The Journal of Physical Chemistry C ARTICLE

reactivity of the nitrogen doped graphene is significantlyimproved because the electrons are more easily excited fromvalence band to conduction band. For pure graphene C46H18,the HOMO�LUMO gap is already low due to the presentof defects (two pentagon carbon rings). It has been shownthat defects can significantly alter the electronic propertiesof the graphene. Surprisingly, although C46H18 has a low

HOMO�LUMO gap, our further calculations show thatOOH molecule cannot adsorb to the pure graphene, indicatingthat it does not have catalytic capability. Obviously, there areother factors that determine the catalytic capability of thegraphene, e.g., spin density and charge density of individualatoms, which will be discussed later. Although the nitrogendoping makes the HOMO�LUMO gap of C45NH18 slightly

Table 2. HOMO, LUMO, and HOMO�LUMO Energy Gap of r Electrons and β Electrons for C45NH20, C45NH18, C46H18 andC46H20 (Unit/eV)

a

C45NH20 C45NH18 C46H20 C46H18

R electron β electron R electron β electron R electron β electron R electron β electron

HOMO �3.29 �4.87 �4.57 �4.58 �4.87 �4.87 �4.61 �4.61

LUMO �1.89 �2.14 �1.98 �3.11 �2.10 �2.10 �3.43 �3.43

HOMO�LUMO gap 1.40 2.73 2.59 1.47 2.77 2.77 1.18 1.18aHOMO�LUMO gap is the difference between LUMO and HOMO energy levels.

Table 3. HOMO and LUMO Spatial Distributions of r Electron and β Electron for C45NH20, C45NH18, C46H20 and C46H18

F dx.doi.org/10.1021/jp201991j |J. Phys. Chem. C XXXX, XXX, 000–000

The Journal of Physical Chemistry C ARTICLE

increase, the value of the gap (1.47 eV) is still comparable tothat of the graphene containing pyridine species. Thus, nitrogendoping is a key for graphene to possess high catalytic reactivity.

We also obtained the HOMO and LUMO spatial distributionsof R and β electrons for these graphene structures (Table 3). Forthe pure graphene C46H20, both HOMO and LUMO spatialdistribution are delocalized, with the same spatial distribution forR electron and β electrons. After the graphene is doped to formC45NH20, the HOMO and LUMO of R electron are localizeddistribution. For the pure graphene with two pentagon carbonrings, C46H18, the spatial distribution of molecule orbital ischanged due to the presence of the pentagon rings, resulting inslight localization of the HOMO and LUMO although all ofthe electrons are paired. With doped nitrogen, the LUMO ofβ electron is localized in the graphene C45NH18. Obviously,nitrogen doping introduces an unpaired electron, which causesthe localized distribution of the molecule orbitals. As a result, thechemical reactivity of the graphene is significantly enhanced bythe nitrogen doping.

It is of interest to determine the active sites for catalyticreaction on N-graphene. Parts a and b of Figure 4 show atomiccharge and spin density distribution on the N-graphene(C45NH20), the atomic charge and spin density distributenonuniformly around the nitrogen atom. The carbon atom(C6), the second neighbor of the nitrogen, has the largest atomiccharge value 0.169 while the carbon atom C9 bonding to thenitrogen has the second largest value 0.150. C37 in the oppositeposition of the same hexagon ring as the nitrogen has the largestspin density 0.235 while C9 has the second largest spin density0.194. For pure graphene, the spin density is zero while the

charge density is small and distributes relatively uniformlycompared to those with N doping.

Putting an OOHmolecule over the carbon atoms at a distanceof 3.0 Å, we have examined most carbon atoms that could possiblyact as active sites for catalysis. There is no active site identified onpure graphene but we found that OOH can adsorb to C37 and C9,the carbon atoms with high spin density. Thus, there are two activesites near single nitrogen dopant on the N-graphene (C45NH20).The adsorption energy of OOH bonding to C37 andC9 is equal to�0.85 and �1.04 eV, respectively. Although C6 has the largestatomic charge value, the OOH molecule could not adsorb to thiscarbon atom as its spin density is�0.045. On the other hand, C37has the small charge density of 0.084, but its spin density is thelargest. OOH molecule can adsorb to this carbon atom.

We also checked all of the possible active sites on theN-graphene with a pyrrole structure (C45NH18). Its atomiccharge density and spin density distribution are shown inFigure 4c,d, respectively. We found that OOH can adsorb toC63, C11, C24, C9,and C16 atoms. C63 has the largest spindensity 0.681. C11 and C24 have the second and third largestspin densities 0.200 and 0.198, respectively. Different fromC45NH20, the carbon atoms C9 and C16 both have the largestatomic charge value of 0.237 but with negative spin densities of�0.009 and �0.007, respectively. OOH can be adsorbed tothese two atoms, C9 and C16. This may be attributed to bothfacts: large charge density and relatively small absolute value ofnegative spin density of these two atoms. Thus, compared toatomic charge density, spin density is much more important indetermining the catalytic active sites. If negative spin density onan atom is small, atomic charge density will play a key role in

Figure 4. (a), (c) Charge distribution and (b), (d) spin density distribution on theN-graphene with pyridine structure (C45NH20) and pyrrole structure(C45NH18), respectively. The number on the circle is the number of atom. The fractions on the side of these atoms in (a), (c) are atomic charge value andspin density value on the atoms. The denominator is the charge value while the numerator is the spin density number. In (b), (d) spin density distributeson the electron density isovalue plane; the most negative value is red while the most positive value is blue.

G dx.doi.org/10.1021/jp201991j |J. Phys. Chem. C XXXX, XXX, 000–000

The Journal of Physical Chemistry C ARTICLE

determining whether it is an active site or not. For N-graphenemodels, C45NH20 and C45NH18, containing pyridine andpyrrole structures, both have electrocatalytic property forORR of four-electron transformation process, consistent withthe experimental results.13,31 Here, the substituting N atoms inN-graphene leads to the asymmetry spin density and atomiccharge density, thus making it possible for N-graphene to showhigh electroncatalytic activities for the ORR.

More generally, any chemical species in the form of eithersubstitution or attachment on graphene, which can lead to ahigh asymmetric spin density and atomic charge density ongraphene, could promote high electroncatalytic activities forthe ORR. Recently, Wang et al. demonstrated that poly-(diallyldimethylammonium chloride) (PDDA) functiona-lized/adsorbed carbon nanotubes can act as effective catalystsfor ORR in fuel cells with similar performance as Pt catalysts.32

With a strong electron-withdrawing ability, PDDA could causehigh positive spin density and atomic charge density onnanotubes, creating active sites for facilitating ORR. Thus,our results here may provide a general rule for searching fornew catalysts for ORR in fuel cells.

4. CONCLUSION

The DFT method was used to study the mechanism of ORRon the N-graphene cathode of fuel cells in acidic environment.The simulation results on the electron transformation processshow that the ORR is a four-electron pathway on the N-graphenebut pure graphene does not have such catalytic activities. WhenH is introduced into the system, the sequential reactions canoccur, including the formation of O�C chemical bond betweenoxygen and graphene, O�O bond break, and creation of watermolecules. For each reaction step, the system energy decreasesaccordingly, indicating that the four-electron transformationreaction takes place spontaneously. The catalytic active sites onthe N-graphene depend on spin density distribution and atomiccharge distribution. The substituting nitrogen atom introducesno-pair electrons to the graphene and changes the atomic chargedistribution on it. Generally, the carbon atoms that possess highestspin density are the electrocatalytic active catalytic sites. If thenegative value of spin density is small, the carbon atoms with largepositive atomic charge density may act as the active sites.

’AUTHOR INFORMATION

Corresponding Author*E-mail: Zhenhai.xia@unt.edu. Tel: 940-369-7673. Fax:940-565-4824.

’ACKNOWLEDGMENT

We thank the National Science Fundation (NSF) for thesupport of this research under the contract no. 1000768.

’REFERENCES

(1) Steele, B. C. H.; Heinzel, A. Nature 2001, 414, 345–352.(2) Neergat, M.; Shukla, A. K.; Gandhi, K. S. J. Appl. Electrochem.

2001, 31, 373–378.(3) Greeley, J.; Stenphens, I. E. L.; Bondarenko, A. S.; Johansson,

T. P.; Hansen, H. A.; Jaramillo, T. F.; Rossmeisl, J.; Chorkendorff, I.;Norskov, J. K. Nat. Chem. 2009, 1, 552–556.(4) Ji, X. L.; Lee, K. T.; Holden, R.; Zhang, L.; Zhang, J. J.; Botton,

G. A.; Couillard, M.; F.Nazar, L. Nat. Chem. 2010, 2, 286–293.

(5) Yu, X.; Ye, S. J. Power Sources 2007, 172, 145–154.(6) Winter, M.; Brodd, R. J. Chem. Rev. 2004, 104, 4245–4269.(7) Che, G.; Lakshmi, B. B.; Fisher, E. R.; Martin, C. R.Nature 1998,

393, 346–349.(8) Gong, K. P.; Yu, P.; Su, L.; Xiong, S. X.; Mao, L. Q. J. Phys. Chem.

C 2007, 111, 1882–1887.(9) Kongkanand, A.; Kuwabata, S.; Girishkumar, G.; Kamat, P.

Langmuir 2006, 22, 2392–2396.(10) Collman, J. P.; Devaraj, N. K.; Decreau, R. A.; Yang, Y.; Yan,

Y. L. Science 2007, 315, 1565–1568.(11) Winther-Jensen, B.;Winther-Jensen, O.; Forsyth,M.;MacFarlane,

D. R. Science 2008, 321, 671–674.(12) Gong, K.; Du, F.; Xia, Z.; Dustock, M.; Dai, L. Science 2009,

323, 760–764.(13) Qu, L. T.; Liu, Y.; Baek, J. B.; Dai, L. M. ACS Nano 2010,

4, 1321–1326.(14) Sharma, S.; Ganguly, A.; Papakonstantinou, P.; Miao, X. P.; Li,

M. X.; Hutchison, J .L.; Delichatsios, M.; Ukleja, S. J. Phys. Chem. C 2010,114, 19459–19466.

(15) Kou, R.; Shao, Y. Y.;Mei, D. H.; Nie, Z.M.;Wang, D. H.;Wang,C. M. J. Am. Chem. Soc. 2011, 133, 2541–2547.

(16) Stephen, W.; Abhishek, D.; Masoud, A.; Heinz, P. J. Phys. Chem.C 2008, 112, 8464–8475.

(17) Chen, R. R.; Li, H. X.; Chu, D.; Wang, G. F. J. Phys. Chem. C2009, 113, 20689–20697.

(18) Reyimjan, A.; Sidik; Alfred, B. A. J. Phys. Chem. B 2006,110, 1787–1793.

(19) Ellen, V.; Alfred, B. A. J. Phys. Chem. C 2007, 111, 9330–9336.(20) Kiera, A.; Kurak; Alfred, B. A. J. Phys. Chem. C 2009,

113, 6730–6734.(21) Wang, Y. X.; Balbuena, P. B. J. Phys. Chem. B 2005,

109, 14896–14907.(22) Frishch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;

Robb,M. A.; Cheeseman, J. R.;Montgomery, J. A.; Vreven, T., Jr.; Kudin,K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.;Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.;Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hadegawa,J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.;Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.;Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A.J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.;Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich,S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.;Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.;Baboul; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.;Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham,M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.;Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Gonzalez; Pople,J. A. Gaussian 03, revision E.01; Gaussian, Inc.: Wallingford, CT, 2004.

(23) Ditchfield, R.; Hehre, W. J.; Pople, J. A. J. Chem. Phys. 1971, 54,724–728.

(24) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56,2257–2261.

(25) Hariharan, P. C.; Pople, J. A. Theor. Chem. Acc. 1973, 28,213–222.

(26) Francl, M.M.; Pietro,W. J.; Hehre,W. J.; Binkley, J. S.; DeFrees,D. J.; Pople, J. A.; Gordon, M. S. J. Chem. Phys. 1982, 77, 3654–3665.

(27) Frisch, M. J.; Pople, J. A.; Binkley, J. S. J. Chem. Phys. 1984,80, 3265–3269.

(28) Aihara, J.-I. J. Phys. Chem. A 1999, 103, 7487–7495.(29) Shemella, P.; Zhang, Y. M.; Mailman, M. Appl. Phys. Lett. 2007,

91, 042101.(30) Zheng, H. X.; Duley, W. Phys. Rev. B 2008, 78, 155118.(31) Yu, D. S.; Zhang, Q.; Dai, L. M. J. Am. Chem. Soc. 2010,

132, 15127–15129.(32) Wang, S. Y.; Yu, D. S.; Dai, L. M. J. Am. Chem. Soc. 2011,

133, 5182–5185.