Asorption of H 2 O, NH 3, CO, NO 2, an NO on graphene: A first-principles stuy O. Leenaerts,* B. Partoens, an F. M. Peeters Departement Fysica, Universiteit Antwerpen, Groenenborgerlaan 171, B-2020 Antwerpen, Belgium Receive 26 July 2007; revise manuscript receive 20 February 2008; publishe 17 March 2008 Motivate by the recent realization of graphene sensors to etect iniviual gas molecules, we investigate the asorption of H 2 O, NH 3,CO,NO 2, an NO on a graphene substrate using first-principles calculations. The optimal asorption position an orientation of these molecules on the graphene surface is etermine an the asorption energies are calculate. Molecular oping, i.e., charge transfer between the molecules an the graphene surface, is iscusse in light of the ensity of states an the molecular orbitals of the asorbates. The efficiency of oping of the ifferent molecules is etermine an the influence of their magnetic moment is iscusse. DOI: 10.1103/PhysRevB.77.125416 PACS number s : 68.43. h, 73.20.Hb, 68.43.Bc, 81.05.Uw I. INTRODUCTION The synthesis of monolayer graphite i.e., graphene 1 an the experimental observation of Dirac charge carriers in this system 2,3 have awakene an enormous interest in this twoimensional material. The unusual properties of carriers in graphene are a consequence of the gapless an approximately linear electron ispersion at the vicinity of the Fermi level at two inequivalent points of the Brillouin zone. In the low-energy limit, the quasiparticles in these systems are escribe in terms of massless chiral relativistic fermions governe by the Dirac equation. The goo sensor properties of carbon nanotubes are alreay known for some time, 4 but recently, the possibility to use graphene as a highly sensitive gas sensor was also reporte. 5 It was shown that the increase in graphene charge carrier concentration inuce by asorbe gas molecules can be use to make highly sensitive sensors, even with the possibility of etecting iniviual molecules. The sensor property is base on changes in the resistivity ue to molecules asorbe on the graphene sheet that act as onors or acceptors. The sensitivity of NH 3, CO, an H 2 Oupto1ppb parts per 10 9 was emonstrate, an even the ultimate sensitivity of an iniviual molecule was suggeste for NO 2. These excellent sensor properties of graphene are ue to two important facts: i graphene is a two-imensional crystal with only a surface an no volume, which maximizes the effect of surface opants, an ii graphene is highly conuctive an shows metallic conuctance even in the limit of zero carrier ensity. To fully exploit the possibilities of graphene sensors, it is important to unerstan the interaction between the graphene surface an the asorbate molecules. We perform in this letter first-principles calculations for the molecules NH 3,NO 2, NO, CO, an H 2 O asorbe on graphene. We etermine their exact orientation on the surface an their preferential bining site by calculating their bining energy. Their charge transfer to the graphene surface is investigate in orer to etermine the onor or acceptor character of the molecular opant. use for the stuy of molecular asorbates on single-walle carbon nanotubes SWNT. 6 10 All our DFT calculations were carrie out with the ABINIT coe, 11 within the generalize graient approximation GGA of Perew Burke Ernzerhof. 12 The avantage of GGA over the local ensity approximation LDA in this work is that the GGA will not lea to a strong boning of the molecules as in LDA. So, if the molecules bin in GGA, they will efinitely bin in a real system an in LDA too. The istance between asorbate an the graphene surface, however, will be somewhat overestimate an consequently the bining energy will be unerestimate. We use a plane wave basis set with a cutoff energy of 816 ev an pseuopotentials of the Troullier Martins type. 14 For the asorption of the molecules NH 3, CO, an H 2 O, we use non-spin-polarize calculations, while for NO 2 an NO, we use spin-polarize ones. The total system consists of a 4 4 graphene supercell 32 C atoms with a single molecule asorbe to it Fig. 1 an with a istance of 16 Å between ajacent graphene layers. The sampling of the Brillouin zone is one using a 5 5 1 Monkhorst Pack 15 gri, which is teste to give converge results for all the properties we calculate. For the calculation of the ensity of states DOS, weusea15 15 1 Monkhorst Pack gri an a Gaussian smearing of 0.14 ev. Charge transfers are calculate base on the Hirshfel charge analysis. 16 The atomic charge Q A for each atom is obtaine by with r the calculate ensity an A 0 r the electron ensity compute for the isolate atom A an taken from Ref. 11 II. COMPUTATIONAL DETAILS The first-principles calculations are performe using ensity functional theory DFT which has been successfully FIG. 1. Color online H 2 O on graphene. 4 4 supercell of graphene with asorbe H 2 O molecule. 1098-0121/2008/77 12 /125416 6 125416-1 2008 The American Physical Society
LEENAERTS, PARTOENS, AND PEETERS Q A = 0 A r A 0 A r r r, from which the charge transfer is euce. From this result we etermine whether or not the asorbate acts as an acceptor or a onor. It shoul be note that the size of the charge transfer is slightly epenent on the metho use to calculate it. The istance from the asorbate to the graphene surface is calculate from the ifference in weighte averages of the ifferent atoms of the molecule an the carbon atoms of the graphene sheet, where we use the atomic number Z of the atoms as the weight factor. III. RESULTS For each asorbate, three asorption sites are consiere, namely, on top of a carbon atom T, the center of a carbon hexagon C, an the center of a carbon-carbon bon B see Fig. 1. For these positions, ifferent orientations of the molecules are examine an the asorption energy is calculate for all of them. The asorption energy is the energy of the isolate graphene sheet an isolate molecule minus the energy of the fully relaxe graphene sheet with the molecule asorbe to it. We i not correct these energies for the ipole-ipole interactions which occur ue to the finite size of the supercells. However, these energies mainly cancel, resulting in corrections of the orer of 2 mev in the worst cases H 2 O an NH 3. This has no influence on any of our conclusions, so we ecie to neglect them. The strength of the molecular oping is iscusse in light of the ensity of states an the highest occupie molecular orbital HOMO an lowest unoccupie molecular orbital LUMO of the asorbate. The position of these orbitals, visible as peaks in the DOS, is practically inepenent of the orientation an asorption site of the molecule, so we only show the total DOS for one geometry per molecule. We can now istinguish two charge transfer mechanisms: i a charge transfer can occur ue to the relative position in the DOS of the HOMO an LUMO of the asorbate. If the HOMO is above the Fermi level of pure graphene the Dirac point, there is a charge transfer to graphene. If the LUMO is below the Dirac point, charge will transfer to the molecule. ii The charge transfer between asorbate an graphene is also partially etermine by the mixing of the HOMO an LUMO with the graphene orbitals hybriization. This mixing scales with the overlap of the interacting orbitals an the inverse of their energy ifference. It is more ifficult to iscuss the asorption energy in this way because of the large amount of possible interacting orbitals present in graphene. Our investigation starts with the nonmagnetic molecules H 2 O, NH 3, an CO, followe by the paramagnetic ones, NO 2 an NO. We iscuss the molecules in the orer of increasing complexity of their charge transfer mechanism. A. H 2 O on graphene We examine the following orientations of the H 2 O molecule with respect to the graphene surface: starting from the TABLE I. H 2 O on graphene: the asorption energy, the istance of H 2 O above the graphene surface, an the charge transfer from the molecule to graphene for ten ifferent B u 18 3.70 0.021 T u 19 3.70 0.021 C u 20 3.69 0.021 B n 24 3.55 0.013 T n 24 3.56 0.015 C n 27 3.55 0.014 B 18 4.05 0.009 T 19 4.05 0.009 C 19 4.02 0.010 C v 47 3.50 0.025 O atom the H-O bons pointing up u, own, or parallel to the graphene surface n. Another orientation v was suggeste in a theoretical stuy, base on an empirical metho, of the asorption of H 2 O on graphite. 13 This orientation has one O-H bon parallel to the surface an the other one pointing to the surface. All properties were foun to be almost invariant with respect to rotations aroun the axis perpenicular to the surface an through the oxygen atom, an therefore, we will not iscuss this orientation. The results of the calculations are given in Table I. From Table I, we learn that all the asorption energies are small, which is partially a consequence of the calculation metho. This is not very important because the asorption energies are only use to compare the ifferent geometries an to fin the best position an orientation of the molecule for which we nee only relative values. Table I also shows that the asorption energy is primarily etermine by the orientation u,, n, an v an to a lesser egree by the position C, B, an T of the molecule. The energy ifferences are 5 6 mev with respect to the orientation, but they vary with about 1 2 mev when changing the position. This ifference in importance of position an orientation is even more pronounce when we look at the charge transfers. If the O atom points to the graphene surface, there is a small charge transfer to graphene, but if the H atoms point to the surface, there is a small charge transfer to the water molecule. This is a consequence of the form of the HOMO an LUMO of H 2 O an their relative position with respect to the Dirac point see Fig. 2. The HOMO 1b 1 is completely locate on the O atom, but the LUMO 4a 1 is mostly locate on the H atoms. In the u orientation, the HOMO plays the ominant role an onates, through a small mixing with graphene orbitals above the Fermi level, some charge to graphene. There is also a stronger mixing with the orbitals below the Dirac point because they are closer in energy, but this oes not inuce any charge transfer because all these orbitals are fille. In the an v orientations, the LUMO of H 2 O interacts stronger with the surface an is able, through a small mixing with the 125416-2
ADSORPTION OF H 2 O, NH 3,CO,NO 2, AND FIG. 2. Color online H 2 O on graphene. Inset: a the HOMO an b the LUMO of H 2 O the H atoms are white an the oxygen atom is re; green an yellow inicate ifferent signs of the orbital wave function. Main panel: DOS of H 2 O on graphene. The blue otte lines show the positions of the molecular orbitals of H 2 O. graphene orbitals below the Dirac point, to accept some charge from graphene. There is also a stronger mixing with orbitals above the Dirac point now, but this oes not inuce any charge transfer because all these orbitals are empty. In the n orientation, it is again the HOMO that will interact stronger, but now there is also some interaction with the LUMO. There will be a charge transfer from the molecule to graphene, but because of the interaction with the LUMO, it will be smaller. Experimentally, 5 one fins that H 2 O acts as an acceptor on graphene which is in accorance with our theoretical results where we fin that the acceptor character C, v is energetically favore on perfect graphene. B. NH 3 on graphene Two orientations of the ammonia molecule were investigate, one with the H atoms pointing away from the surface u an the other with the H atoms pointing to the surface. All properties were again foun to be almost invariant to rotations aroun the axis perpenicular to the surface an through the nitrogen atom. The results of the calculations are given in Table II. The asorption site an the orientation are now seen to be of the same importance for the asorption energy. The charge transfer, however, is solely etermine by the orientation of the NH 3 molecule. TABLE II. NH 3 on graphene: the asorption energy, the istance of NH 3 above the graphene surface, an the charge transfer from the molecule to graphene for six ifferent B u 21 3.86 0.026 T u 20 3.86 0.026 C u 31 3.81 0.027 B 15 4.08 0.001 T 16 3.97 0.000 C 25 3.92 0.001 FIG. 3. Color online NH 3 on graphene. Inset: a the HOMO an b the LUMO of NH 3 the N atom is blue an the H atoms are white. Main panel: DOS of NH 3 on graphene. There is a small charge transfer from the molecule to the graphene surface of 0.03e in the u orientation an there is almost no charge transfer in the orientation. We can see how this comes about by looking at the HOMO 3a 1 an LUMO 4a 1 of the NH 3 molecule Figs. 3 a an 3 b. In the u orientation, the HOMO is the only orbital that can have a significant overlap with the graphene orbitals an thus can cause charge transfer. As a consequence, the NH 3 molecule will act as a onor. In the orientation, both HOMO an LUMO can cause charge transfers which are similar in magnitue but in opposite irections. The net charge transfer is therefore close to 0. The u orientation is energetically favore which explains the onor character as observe experimentally. 5 We also performe LDA calculations for the asorption of NH 3 on graphene. The results of these are similar for the charge transfer, but the asorption energy is much larger 100 mev. The real asorption energy is known to lie between the two approximate values obtaine through GGA an LDA. This is consistent with the experimental observation 5 that the asorbates can be remove from the surface by annealing at 150 C. C. CO on graphene Three ifferent orientations were use for the CO molecule. Two with the molecule perpenicular to the surface, with the O atom above the C atom u an the other way aroun, an one parallel to the surface n. From Table III, we notice that the CO molecule always acts as a onor. The size of the charge transfer only epens on the orientation of the molecule with respect to the surface Fig. 4. The ifferences in charge transfer are ue to ifferences in orbital overlap between the HOMO 5 of the CO molecule an graphene. The LUMO 2 seems to play no important role in the oping process although it is closer to the Dirac point than the HOMO. To unerstan this, we have to take into account the symmetry of this orbital an the graphene orbitals. The DOS below an close 3 ev to the Dirac point originates from mostly boning combinations of the p z atomic orbitals of the C atoms of graphene. The DOS above the Dirac point is mostly ue to antiboning combinations. The completely antisymmetric LUMO will 125416-3
LEENAERTS, PARTOENS, AND PEETERS TABLE III. CO on graphene: the asorption energy, the istance of CO above the graphene surface, an the charge transfer from the molecule to graphene for seven ifferent B u 10 3.75 0.019 T u 10 3.75 0.019 C u 13 3.73 0.019 T 8 3.72 0.009 C 10 3.70 0.010 B n 14 3.74 0.013 C n 14 3.74 0.012 therefore mostly interact with the DOS above the Dirac point which oes not cause any oping. The HOMO is thus the more important orbital an the charge transfer is consequently always to graphene. Because the HOMO is mainly locate on the C atom, the charge transfer is largest when the C atom is closest to the surface u orientation, smallest when the O atom is closer to the surface orientation, an intermeiate when both atoms are at an almost equal istance from the surface n orientation. FIG. 5. Color online NO 2 on graphene. Inset: HOMO an LUMO of NO 2 the N atom is blue an the O atom is re. Main panel: spin-polarize DOS of NO 2 on graphene. smaller than the first one but it is still noticeable in the strength of the magnetic moment of the system. The total magnetic moment of graphene an asorbate in, e.g., the B, orientation is 0.862 B. The charge transfer from graphene M =0 B to NO 2 M =1 B is 0.099e, so the orbital mixing causes a charge transfer of 0.039e to graphene. D. NO 2 on graphene In Ref. 17, it was state that asorbates with a magnetic moment in general result in a larger oping. Therefore, we will now turn our attention to paramagnetic molecules. The first one is NO 2 which has, in a spin-polarize calculation, an energy that is 0.4 ev smaller as compare to an unpolarize one an therefore is paramagnetic. We examine three ifferent orientations of the NO 2 molecule: starting from the N atom, the N-O bons pointing up u, own, or parallel to the graphene surface n. The LUMO 6a 1, of NO 2 is locate 0.3 ev below the Dirac point Fig. 5. This inuces a large charge transfer to the molecule. However, there are also some NO 2 orbitals close enough to the Dirac point to cause some charge transfer in the opposite irection through orbital mixing especially the HOMO 6a 1, Table IV. The latter charge transfer is E. NO on graphene To test whether or not there is always strong oping by paramagnetic molecules, we will investigate another one. For NO, a spin-polarize calculation gives an energy that is 0.3 ev lower than a non-spin-polarize one, so NO is inee a paramagnetic molecule. We investigate the same orientations an use the same notations as for the CO molecule replace C with N. Contrary to the claim mae in Ref. 17, we i not fin that NO inuces any strong oping. The charge transfers are an orer of magnitue smaller than in the case of the NO 2 molecule see Table V which is comparable to the nonmagnetic molecules. Physically, we can unerstan this if we compare the DOS of the asorbates NO Fig. 6 an NO 2 on graphene. For NO 2 asorbe on graphene, the LUMO is situate 0.3 ev below the Dirac point of graphene. TABLE IV. NO 2 on graphene: the asorption energy, the istance of NO 2 above the graphene surface, an the charge transfer from the molecule to graphene for six ifferent FIG. 4. Color online CO on graphene. Inset: a the HOMO an b the LUMO of CO the C atom is black an the O atom is re. Main panel: DOS of CO on graphene. B 67 3.61 0.099 T 65 3.61 0.099 C 63 3.64 0.098 B u 55 3.83 0.089 T u 55 3.93 0.090 C n 67 3.83 0.102 125416-4
ADSORPTION OF H 2 O, NH 3,CO,NO 2, AND TABLE V. NO on graphene: the asorption energy, the istance of NO above the graphene surface, an the charge transfer from the molecule to graphene for six ifferent TABLE VI. Summary of results. Asorbate Theory Expt. a C u 16 4.35 0.006 T u 14 4.35 0.006 C 13 4.11 0.007 T 11 4.27 0.005 C n 28 3.71 0.018 B n 29 3.76 0.017 H 2 O Acceptor Acceptor 47 0.025 NH 3 Donor Donor 31 0.027 CO Donor Donor 14 0.012 NO 2 Acceptor Acceptor 67 0.099 NO Donor 29 0.018 a Reference 5. This inuces a strong oping. In the case of NO Fig. 6, the HOMO is egenerate 2 x,2 y an is half fille so it is also the LUMO an lies only 0.1 ev below the Dirac point. This inuces a very small charge transfer from graphene to NO, but ue to its small strength, it can be over compensate by orbital mixing. The HOMO an/or LUMO, of NO can, because it is half fille, cause charge transfer in both irections by mixing with the graphene orbitals below an above the Dirac point. However, as in the case of the LUMO of CO, it interacts mostly with the latter ue to symmetry reasons. So, the orbital mixing leas to charge transfer to graphene. We see from Table V that the charge transfer ue to orbital mixing always overcompensates the small transfer ue to the position of the HOMO an/or LUMO, so NO always act as a onor. We notice that there are large ifferences in an in the istance. They are obviously correlate because a smaller istance between asorbate an graphene leas to a larger orbital overlap an consequently to more orbital mixing i.e., a larger charge transfer. The ifferences in the istance can be explaine by the overlap of the 5 orbital. The position of this orbital is very close in energy to that part of the graphene DOS that originates from the boning combinations of carbon p z orbitals aroun the point. Mixing of these orbitals inuces a net energy shift upwar so they repel each other strongly. The geometry of the 5 orbital gives a large overlap in the u orientation, a smaller overlap in the orientation, an the smallest overlap in the n orientation. This gives a simple explanation for all the ifferences foun from our calculations. IV. SUMMARY AND CONCLUSIONS The charge transfer between the consiere asorbates an graphene is foun to be almost inepenent on the asorption site but it oes epen strongly on the orientation of the asorbate with respect to the graphene surface. We compare two paramagnetic molecules, NO 2 an NO, an foun that NO 2 inuces a relatively strong oping 0.1e, but NO oes not 0.02e. This is in contrast to Ref. 17 where it was claime that paramagnetic molecules are strong opants which we foun inee to be the case for NO 2 but not so for NO. For the consiere asorbates, the sign of the charge transfer agrees with what was foun experimentally see Table VI in Ref. 5. We showe that these charge transfers can be unerstoo from two charge transfer mechanisms. In NO 2 on graphene, it is mainly ue to the position of the LUMO below the Dirac point for all the other stuie asorbates it is cause by the mixing of the HOMO or LUMO orbitals with the orbitals of graphene. The strength of this hybriization can be euce from the geometrical orientation of the HOMO an LUMO orbitals with respect to the graphene surface. The strength of the charge transfer an the bining energies must be seen in light of the use approximations GGA an the metho of calculation Hirshfel metho, which probably lea to an unerestimation. The trens an relative values, on the other han, are much more trustworthy. Our results are also in goo agreement with theoretical stuies of the asorption of molecules on large SWNTs in, e.g., Ref. 6. This suggests that some of the knowlege of asorption on nanotubes shoul be transferable to graphene. ACKNOWLEDGMENTS FIG. 6. Color online NO on graphene. Inset: a 5 orbital an b HOMO/LUMO of NO the N atom is blue an the O atom is re. Main panel: spin-polarize DOS of NO on graphene. This work was supporte by the Flemish Science Founation FWO-Vl, by the NOI-BOF of the University of Antwerp, an by the Belgian Science Policy IAP. 125416-5
LEENAERTS, PARTOENS, AND PEETERS *ortwin.leenaerts@ua.ac.be bart.partoens@ua.ac.be francois.peeters@ua.ac.be 1 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, an A. A. Firsov, Science 306, 666 2004. 2 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, an A. A. Firsov, Nature Lonon 438, 197 2005. 3 Y. Zhang, Y.-W. Tan, H. L. Stormer, an P. Kim, Nature Lonon 438, 201 2005. 4 J. Kong, N. R. Franklin, C. Zhou, M. G. Chapline, S. Peng, K. Cho, an H. Dai, Science 287, 622 2000. 5 F. Schein, A. K. Geim, S. V. Morozov, E. W. Hill, P. Blake, M. I. Katsnelson, an K. S. Novoselov, Nat. Mater. 6, 652 2007. 6 J. Zhao, A. Bulum, J. Han, an J. P. Lu, Nanotechnology 13, 195 2002. 7 S. Santucci, S. Picozzi, F. Di Gregorio, L. Lozzi, C. Cantalini, L. Valentini, J. M. Kenny, an B. Delley, J. Chem. Phys. 119, 10904 2003. 8 J. A. Robinson, E. S. Snow, S. C. Baescu, T. L. Reinecke, an F. K. Perkins, Nano Lett. 6, 1747 2006. 9 S. Peng an K. Cho, Nanotechnology 11, 57 2000. 10 S. Peng an K. Cho, Nano Lett. 3, 513 2003. 11 http://www.abinit.org/ 12 J. P. Perew, K. Burke, an M. Ernzerhof, Phys. Rev. Lett. 77, 3865 1996. 13 B. S. González, J. Hernánez-Rojas, J. Bretón, an J. M. Gomez Llorente, J. Phys. Chem. C 111, 18862 2007. 14 N. Troullier an J. L. Martins, Phys. Rev. B 43, 1993 1991. 15 H. J. Monkhorst an J. D. Pack, Phys. Rev. B 13, 5188 1971. 16 F. L. Hirshfel, Theor. Chim. Acta 44, 129 1977. 17 T. O. Wehling, K. S. Novoselov, S. V. Morozov, E. E. Vovin, M. I. Katsnelson, A. K. Geim, an A. I. Lichtenstein, Nano Lett. 8, 173 2008. 125416-6