Adsorbate-Induced Defect Formation and Annihilation on Graphene

Jan 8, 2009 - Department of Physics, Aristotle UniVersity of Thessaloniki, GR-54124 Thessaloniki, Greece, Department of. Physics and Astronomy, Vander...
0 downloads 0 Views 2MB Size
J. Phys. Chem. B 2009, 113, 941–944

941

Adsorbate-Induced Defect Formation and Annihilation on Graphene and Single-Walled Carbon Nanotubes Leonidas Tsetseris*,†,‡ and Sokrates T. Pantelides‡,§ Department of Physics, Aristotle UniVersity of Thessaloniki, GR-54124 Thessaloniki, Greece, Department of Physics and Astronomy, Vanderbilt UniVersity, NashVille, Tennessee 37235, and Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831 ReceiVed: October 18, 2008

We used density functional theory calculations to probe the chemical reactivity of graphene and single-wall carbon nanotubes (CNTs) toward the small molecules O2, H2, N2, C2H2, CO, and CO2. We found that there is a threshold CNT size below which C2H2 and CO, typical feedstock precursors for CNT growth, become trapped in decorated hillock-like defects on the side walls of CNTs. We also found that O2, H2, and CO2 can etch isolated C adatoms and C adatom pairs. These processes play a role not only in the growth of CNTs, but also in the postgrowth evolution of defects on CNTs through exposure to typical ambient gases. Introduction The adsorption of chemical species on graphene-based systems can have either detrimental or beneficial effects, because it can lead to functionalization or defect formation, affecting a plethora of electronic, chemical, mechanical, and transport properties. In particular, the interaction of carbon nanotubes (CNTs) with a number of small molecules is especially important because these species are often used as feedstock material for the growth of CNTs, for example, acetylene and carbon monoxide, or are present during growth as components of carrier gas, such as H2. In addition, molecules such as O2 and N2 are typical ambient gases, and for this reason, their interactions with CNTs might modify the population of CNT defects in the postgrowth stage and during the operation of CNTbased devices. Several studies have considered reactions of small molecules with carbon nanotubes, but to our knowledge, the majority of the investigations pertain to adsorption on the edges of openended CNTs.1-3 In the case of oxygen, first-principles calculations4 found that O2 molecules adsorb on the side walls of singlewalled CNTs (SWCNTs) and then diffuse to their edges. Other studies5 probed the reactions of a number of molecules with vacancies on the graphite surface, and recent investigations6-8 have explored the possibility of hydrogen chemisorption on CNTs as a tool of effective H storage. Acetylene molecules are known9 to play a key role in the growth of polycyclic aromatic hydrocarbons, and interactions of C2H2 molecules with CNTs have been suggested10 as a rate-limiting process for SWCNT growth. Clearly, the identification of atomic-scale mechanisms that relate to strong interactions between feedstock precursors of CNT growth, such as C2H2 and CO, or carrier gas and ambient molecules, such as H2 and O2, is very important for the understanding of the evolution of CNT defects and related properties during and after growth. Among other significant effects, reactions of CNTs with O and H species are important for the seamless embedment of CNTs in dielectric media.11 * Corresponding author. Tel.: [email protected]. † Aristotle University of Thessaloniki. ‡ Vanderbilt University. § Oak Ridge National Laboratory.

+30-2310-998039.

E-mail:

In this article, we identify reactions of various molecules either with pristine graphene and single-wall CNTs or with the important defect class of C adatoms. We show that C2H2 and CO molecules can become trapped as decorated 7-5-5-7 hillock-like species12-15 on the side walls of small CNTs, and we describe the kinetics of key processes, such as transformation, diffusion, and desorption, that ultimately control the population of these defects. In large CNTs, isolated C adatoms and adatom pairs are unstable against etching to gaseous products, which reveals that O2 and H2 molecules, as well as N2 and CO2 species, can, in certain cases, induce healing of residual defects from CNT surfaces. Method The results were obtained using density functional theory (DFT) calculations with a local density approximation (LDA)16 exchange-correlation functional, plane waves as a basis set, and ultrasoft pseudopotentials,17 as implemented in the code VASP.18 The energy cutoff was set at 400 eV, and the Γ point was used for sampling in reciprocal space. Large periodic supercells with 160 C atoms in the case of graphene and 21.3-Å- (19.7-Å-) long zigzag (armchair) CNTs were employed for the study of isolated defects and defect complexes. The energy barriers of transformations were obtained with the elastic band method,19 based on experience with various systems.20,21 Results Reactions with H2 and C2H2. We start with the results on the adsorbate-induced healing of graphene and CNT defects. In particular, we focus on the so-called 7-5-5-7 pair of C adatoms,12-15 which is known to affect the electronic,22 transport,23 mechanical,24 and chemical25,26 properties of CNTs. A 7-5-5-7 complex stays idle unless the sample is heated to very high temperatures.27 This defect has very interesting behavior because it self-heals out to mobile protrusions after capture of a third C atom, but it is favored to reappear in pairs or in larger clusters upon further adatom agglomeration.27 For these reasons, the 7-5-5-7 hillock is one of the best candidates for residual defects on graphene-based systems. Graphene-based systems are, in general, chemically inert because of the high bonding stability of their sp2 networks.

10.1021/jp809228p CCC: $40.75  2009 American Chemical Society Published on Web 01/08/2009

942 J. Phys. Chem. B, Vol. 113, No. 4, 2009

Tsetseris and Pantelides

Figure 2. Adsorption of an O2 molecule on a 7-5-5-7 hillock of graphene: (a) intermediate structure, (b) final configuration. The transformation from a to b has a barrier of 0.9 eV and decreases the energy by 2.2 eV. (Carbon, gray spheres; oxygen, red spheres.)

Figure 1. (a) HsH pair on pristine graphene; its formation from H2 molecules in vacuum is endothermic with an energy penalty of 1.5 eV. (b-f) Hydrogen-mediated healing of defects on graphene: (b-d) exothermic adsorption of H2 on 7-5-5-7 defects (reaction energy ∆E ) -2.25 eV, barrier Ea ) 1.25 eV), (d-f) exothermic desorption of C2H2 species (∆E ) -1.26 eV, Ea ) 1.24 eV). Parts c and e show the transition states. (C, gray spheres; H, white spheres.)

Defects such as the 7-5-5-7 pair, however, lead25,26 to chemical activation and can thus enable functionalization. In this work, we used hydrogen as a prototype adsorbate to study both the energetics and the kinetics of reactions at 7-5-5-7 sites. Hydrogen is an ubiquitous species in many electronic devices,28,29 and it can be expected to be present to varying degrees in CNT-based systems. We found that, whereas H2 chemisorption on a pristine graphene layer carries an energy penalty of 1.5 eV, adsorption on a graphene hillock is an exothermic reaction that releases 2.25 eV of energy. Figure 1b,d depicts the initial and final configurations of adsorption on a graphene 7-5-5-7 complex, starting from a H2 molecule above the hillock. The reaction barrier is 1.25 eV, and the transition state (TS) is shown in Figure 1c. The reverse barrier of 4.5 eV is in excellent agreement with a measured activation energy30 for hydrogen desorption from graphite. The chemisorption of H2 demonstrates the chemically active character of graphene hillocks, in agreement with previous studies on other adsorbates.25,26 However, we further found that the adsorbed structure of Figure 1c is, in fact, a metastable configuration that can precede the removal of the 7-5-5-7 defects themselves. Specifically, the adsorption of H2 can be followed by the exothermic desorption of the hydrogenated 7-5-5-7 defect in the form of an acetylene molecule C2H2. This process is depicted in Figure 1d-f and its reaction energy and barrier are 1.26 and 1.24 eV, respectively. The net conclusion is that adsorbate-mediated healing can, in some cases, compete with functionalization of graphene-based systems. For hydrogen on graphene, the desorption barrier is comparable to the chemisorption activation energy, and healing wins over functionalization. In contrast, the H complex of Figure 1c is more stable than the desorbed C2H2 configuration of Figure 1e for SWCNTs in the ranges from (8,0) to (15,0) and from

(6,6) to (9,9) for zigzag SWCNTs and arm-chair SWCNTs, respectively, and H-induced healing is not operative in these cases. There is, however, a monotonic decrease in the energy penalty, Ep, for C2H2 as the SWCNT size increases, and for the (15,0) SWCNT, Ep is only 0.4 eV. Based on the asymptotic limit of graphene, we can conclude that H healing wins over for SWCNTs that are not much larger than (15,0). Reactions with O2 and CO. Similarly to H2, O2 chemisorption on 7-5-5-7 sites is a strongly exothermic reaction. The attachment of two O atoms in the configuration of Figure 2b releases 3.88 eV in the case of graphene and 3.7-3.9 eV for the above-mentioned SWCNTs. Another local minimum-energy configuration is depicted in Figure 2a, with the two O atoms and the underlying C 7-5-5-7 atoms in a square-bonding arrangement, similar to that obtained in previous studies.25,26 This structure, however, is significantly less stable than that of Figure 2a, by 2.2, 2.2-2.3, and 2.0-2.1 eV in the case of graphene, zigzag SWCNTs, and arm-chair SWCNTs, respectively. O2 interactions with SWCNTs differ from the H2 case not only in the configuration of chemisorbed species, but also in the ability of molecular oxygen to remove 7-5-5-7 defects from small SWCNTs. In particular, emission of a pair of CO molecules is marginally favored energetically over the defect of Figure 2b for (10,0) and (8,8) SWCNTs. The energy gain for CO emission becomes 0.57 and 0.24 eV for (13,0) and (9,9) SWCNTs, respectively, and it increases even further, to 0.81 eV, for a (15,0) carbon nanotube. In the case of graphene, the simultaneous desorption of two vicinal CO molecules, which is depicted in Figure 3, is a strongly exothermic process, releasing 2.6 eV, and its barrier is equal to about 1.5 eV. Another possible desorption pathway comprises two steps, whence one of the two CO species desorbs in the first stage and the emission of the second CO molecule ensues. The barrier of the first (second) step is about 1.5 eV (0.3 eV), and it results in a decrease in energy of 0.5 eV (2.1 eV). On the other hand, the uptake of a pair of CO molecules from the gaseous phase to a defect of the type in Figure 2b is energetically favorable in the case of the smaller (6,6), (7,7), (8,0), and (9,0) CNTs. The corresponding energy gains are 0.75, 0.33, 0.73, and 0.48 eV, respectively. As in the case of acetylene, carbon monoxide can adsorb on the side walls of small CNTs, but in a significant difference from C2H2, the CNT threshold for this process is much lower for CO precursors. The formation of CO vicinal pairs on small CNT systems can occur as a second-order reaction with the participation of two CO molecules. In principle, another possibility pertains to adsorption of individual CO molecules and subsequent mutual trapping of migrating CO adsorbates into the stable pair of Figure 2b. As we now explain, however, this latter scenario is unlikely. First, the adsorption of individual CO molecules on

Adsorbate-Induced Nanotube Defects

J. Phys. Chem. B, Vol. 113, No. 4, 2009 943

Figure 3. Desorption of two CO molecules off graphene: (a) initial configuration, (b) transition state, (c) final configuration. The barrier is 1.5 eV. The desorption is exothermic for graphene and CNTs equal to or larger than (10,0) and (8,8), whereas it is endothermic for smaller CNTs. (Carbon, gray spheres; oxygen, red spheres.)

all studied CNTs and on graphene is endothermic. The corresponding energy penalty varies monotonically from the small values 0.10 eV for an (8,0) CNT and 0.12 eV for a (6,6) CNT to 1.00 and 0.73 eV for (15,0) and (9,9) CNTs, respectively. The initial, final, and transition states for this adsorption on the smallest CNTs are shown in Figure 4. In addition, the figure also depicts the corresponding states for adsorption on graphene, in which case the energy penalty extends to 2.15 eV. The respective reaction barriers were calculated as 2.5 eV on graphene, 1.4 eV on an (8,0) CNT, and 1.7 eV on a (6,6) CNT. A second factor against the scenario of individual adsorption, migration, and mutual trapping relates to a peculiar feature of the behavior of CO molecules. In particular, we studied the diffusion of CO molecules using various initial sequences for the hopping of CO species between neighboring bridge sites of (6,6) and (8,0) CNTs. In all cases, the results indicate that there is no true migration path; instead, as the CO species attempt to hop, they are favored to first desorb off the CNT and then readsorb at another CNT site. The significance of this finding is the following: the stable decorated hillock of Figure 2b can form only as the result of adsorption of two CO molecules on the same site; in other words, it does, indeed, correspond to a second-order process the likelihood of which depends strongly on the concentration of CO in the gas phase. Oxygen molecules can also interact with isolated C adatoms, a reaction that can etch away the adatom in the form of a CO2 molecule. Figure 5 depicts a number of configurations with O2 molecules attached to a C adatom on graphene and CNTs. The attachment is exothermic, leading to energy gains of 3.11 eV for the graphene structure of Figure 5a and 3.58, 3.67, and 4.67 eV for the CNT configurations of Figure 5b-d, respectively. These geometries are, in fact, intermediate products before the emission of CO2 molecules. Overall, the O2 etching of isolated C adatoms into CO2 species is strongly exothermic for graphene and for all the studied CNTs, leading to the release of 9.2-10.6 eV of energy. Reactions with CO2 and N2. CO2 is a product of reactions between oxygen and C adatoms, but carbon dioxide can, in turn, react with these defects and etch them off CNT surfaces. In

Figure 4. Endothermic adsorption of a carbon monoxide molecule on (a) a graphene layer (barrier Ea ) 2.5 eV, reaction energy ∆E ) 2.15 eV), (c) an (8,0) carbon nanotube (Ea ) 1.4 eV, ∆E ) 0.12 eV), and (e) a (6,6) carbon nanotube (Ea ) 1.7 eV, ∆E ) 0.26 eV). (b,d,f) Corresponding transition states of adsorption. (Carbon, gray spheres; oxygen, red spheres.)

Figure 5. Reaction of an oxygen molecule with a C adatom on (a) graphene, (b) an (8,0) carbon nanotube, and (c-d) a (6,6) carbon nanotube. Emission of a CO2 molecule decreases the energy by (a) 7.56, (b) 5.7, (c) 5.54, and (d) 4.54 eV. (Carbon, gray spheres; oxygen, red spheres.)

particular, a CO2 molecule can pair with a C adatom to form a decorated hillock of the type in Figure 2b, a process that is energetically favorable by 0.44 eV on graphene and by 1.59-2.33 eV on the studied CNTs. As discussed above, the complex of Figure 2b is unstable against emission of two CO molecules for CNTs equal to or larger than (10,0) and (8,8).

944 J. Phys. Chem. B, Vol. 113, No. 4, 2009

Tsetseris and Pantelides In summary, we showed that hydrogen, oxygen, and nitrogen molecules can remove adatom defects from graphene and large carbon nanotubes, but we also demonstrated that reactions with feedstock precursors for CNT growth, namely, carbon monoxide and acetylene, can lead to defect formation on sufficiently small carbon nanotubes. Thus, there is a competition between defect formation and healing of defects, and the balance is tilted in favor of one of the two depending on the size of the carbon nanotubes, the defects, and the chemical species involved.

Figure 6. (a) Pair of vicinal CN radicals on a graphene sheet. (b) Cyanogen molecule emitted from the graphene defect in part a. The emission is exothermic by 2.2 eV in energy. (Carbon, gray spheres; nitrogen: cyan spheres.)

Concomitantly, CO2 can etch isolated C adatoms for the same range of CNT sizes. At this point, we recap the trends of reactions involving O2, CO, and CO2 molecules and graphene-based systems. We note that, even though the DFT values might not be in exact agreement with experimental values, the calculated values for the reaction energies and barriers render the following trends unambiguous. O2 and CO2 molecules etch 7-5-5-7 defects and isolated C adatoms, respectively, on sufficiently large CNTs, releasing two CO species, but they become trapped as decorated hillocks on small CNTs. O2 molecules etch isolated C adatoms on graphene-based systems, releasing CO2, whereas CO molecules become trapped in decorated hillocks on small CNTs through a second-order reaction the efficiency of which depends on the concentration of CO in the gas phase. Finally, we investigated the energetics of the reaction of a N2 molecule with a 7-5-5-7 defect on graphene. The attachment of the N atoms on each of the C atoms of the hillock leads to an energy decrease of 1.0 eV. Unlike oxygen and hydrogen, the addition of N modifies the structure considerably. As shown in Figure 6a, the new complex has two vicinal CN radicals that are located directly above graphene carbon atoms, not on bridge positions over graphene C-C bonds. In fact, this complex is metastable against emission of cyanogen (dimethyldinitrile), and the respective desorbed configuration of Figure 6b is 2.2 eV lower in energy than the structure of Figure 6a. Conclusions The results reported herein clearly indicate a dual role for small molecules to either remove or create defects on CNTs. Acetylene and carbon monoxide are typical carbon precursors in the chemical vapor deposition (CVD) growth of CNTs, and theoretical studies33,34 have investigated their decomposition on appropriate metal nanoparticle catalysts. As noted above, reactions of small molecules have also been investigated theoretically1-3 for adsorption on the open ends of CNTs, a regime that relates mostly to nanotubes grown with arc discharge or laser ablation techniques. To our knowledge, the direct uptake of C in the form of C2H2 or a pair of vicinal CO species on the side walls of CNTs has not been highlighted before. Impurities and defects can affect the properties of CNTs in many different ways,13,26,31,32 but the reactions discussed herein are also significant because of the hazardous nature of certain products. For example, carbon monoxide and cyanogen35 are well-known poisons of very high toxicity, and their release by CNTs is clearly an undesired effect for biological-related CNT applications. Of course, the actual level of concern about the release of toxic substances depends in the first place on the numbers of 7-5-5-7 defects on the CNTs of interest. The evaluation of this population will be the subject of future studies.

Acknowledgment. This work was supported by the William A. and Nancy F. McMinn Endowment at Vanderbilt University and by U.S. DOE Grant DEFG0203ER46096. The calculations were performed at ORNL’s Center for Computational Sciences. References and Notes (1) Mann, D. J.; Halls, M. D.; Hase, W. L. J. Phys. Chem. B 2002, 106, 12418. (2) Espinal, J. F.; Montoya, A.; Mondrago´n, F.; Truong, T. N. J. Phys. Chem. B 2004, 108, 1003. (3) Seo, K.; Kim, C.; Kim, B.; Lee, Y. H.; Song, K. J. Phys. Chem. B 2004, 108, 4308. (4) Zhu, X. Y.; Lee, S. M.; Lee, Y. H.; Frauenheim, Y. H. Phys. ReV. Lett. 2000, 85, 2757. (5) Allouche, A.; Ferro, Y. Carbon 2006, 44, 3320. (6) Bilic, A.; Gale, J. D. J. Phys. Chem. C 2008, 112, 12568. (7) Miller, G. P.; Kintigh, J.; Kim, E.; Weck, P. W.; Berber, S.; Toma´nek, D. J. Am. Chem. Soc. 2008, 130, 2296. (8) Nikitin, A.; Li, X.; Zhang, Z.; Ogasawara, H.; Dai, H.; Nilsson, A. Nano Lett. 2008, 8, 162. (9) Frenklach, M. Phys. Chem. Chem. Phys. 2002, 4, 2028. (10) Louchev, O. A.; Laude, T.; Sato, Y.; Kanda, H. J. Chem. Phys. 2003, 118, 7622. (11) Tsetseris, L.; Pantelides, S. T. Phys. ReV. Lett. 2006, 97, 266805. (12) Orlikowski, D.; Buongiorno Nardelli, M.; Bernholc, J.; Roland, C. Phys. ReV. Lett. 1999, 83, 4132. (13) Orlikowski, D.; Buongiorno Nardelli, M.; Bernholc, J.; Roland, C. Phys. ReV. B 2000, 61, 14194. (14) Xia, Y.; Ma, Y.; Xing, Y.; Mu, Y.; Tan, C.; Mei, L. Phys. ReV. B 2000, 61, 11088. (15) Sternberg, M.; Curtiss, L. A.; Gruen, D. M.; Kedziora, G.; Horner, D. A.; Redfern, P. C.; Zapol, P. Phys. ReV. Lett. 2006, 96, 075506. (16) Perdew, J. P.; Zunger, A. Phys. ReV. B 1981, 23, 5048. (17) Vanderbilt, D. Phys. ReV. B 1990, 41, 7892. (18) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169. (19) Mills, G.; Jo´nsson, H.; Schenter, G. K. Surf. Sci. 1995, 324, 305. (20) Tsetseris, L.; Wang, S. W.; Pantelides, S. T. Appl. Phys. Lett. 2006, 88, 051916. (21) Tsetseris, L.; Kalfagiannis, N.; Logothetidis, S.; Pantelides, S. T. Phys. ReV. Lett. 2007, 99, 125503. (22) Habenicht, B. F.; Kamisaka, H.; Yamashita, K.; Prezhdo, O. V. Nano Lett. 2007, 7, 3260. (23) Mehrez, H.; Taylor, J.; Guo, H.; Wang, J.; Roland, C. Phys. ReV. Lett. 2000, 84, 2682. (24) Xia, Y.; Zhao, M.; Ma, Y.; Ying, M.; Liu, X.; Liu, P.; Mei, L. Phys. ReV. B 2002, 65, 155415. (25) Grujicic, M.; Cao, G.; Singh, R. Appl. Surf. Sci. 2003, 211, 166. (26) Horner, D. A.; Redfern, P. C.; Sternberg, M.; Zapol, P.; Curtiss, L. A. Chem. Phys. Lett. 2007, 450, 71. (27) Tsetseris, L.; Pantelides, S. T. Carbon (2008), doi: 1016/j.carbon.2008.12.002. (28) Tsetseris, L.; Fleetwood, D. M.; Schrimpf, R. D.; Zhou, X. J.; Batyrev, I. G.; Pantelides, S. T. Microelectron. Eng. 2007, 84, 2344. (29) Pantelides, S. T.; Tsetseris, L.; Rashkeev, S. N.; Zhou, X. J.; Fleetwood, D. M.; Schrimpf, R. D. Microelectron. Reliab. 2007, 47, 903. (30) Atsumi, H.; Tauchi, K. J. Alloys Compd. 2003, 705-709, 356. (31) Fan, X.; Dickey, E. C.; Eklund, P. C.; Williams, K. A.; Grigorian, L.; Buczko, R.; Pantelides, S. T.; Pennycook, S. J. Phys. ReV. Lett. 2000, 84, 4621. (32) Fan, Y.; Burghard, M.; Kern, K. AdV. Mater. 2002, 14, 130. (33) Bengaard, H. S.; Norskov, J. K.; Sehested, J.; Clausen, B. S.; Nielsen, L. P.; Molenbroek, A. M.; Rostrup-Nielsen, J. R. J. Catal. 2002, 209, 365. (34) Hofman, S.; Csa´nyi, G.; Ferrari, A. C.; Payne, M. C.; Robertson, J. Phys. ReV. Lett. 2005, 95, 036101. (35) Brotherton, T. K.; Lynn, J. W. Chem. ReV. 1959, 59, 841.

JP809228P