J. Phys. Chem. B 2005, 109, 13175-13179
13175
Vacancy-Induced Chemisorption of NO2 on Carbon Nanotubes: A Combined Theoretical and Experimental Study Francesco Mercuri,*,† Antonio Sgamellotti,† Luca Valentini,‡ Ilaria Armentano,‡ and Jose` M. Kenny‡ ISTM-CNR, Dipartimento di Chimica, UniVersita` di Perugia, INSTM UdR Perugia, Via Elce di Sotto, 8, 06123 Perugia, Italy, and Materials Engineering Centre, UniVersita` di Perugia, INSTM UdR Perugia, Terni, Italy ReceiVed: February 10, 2005; In Final Form: May 4, 2005
The role of structural defects on the adsorption of NO2 on carbon nanotubes (CNTs) is analyzed here by means of both statical density functional theory calculations and Car-Parrinello molecular dynamics and further confirmed by X-ray photoelectron spectroscopy measurements. The interaction of a NO2 molecule with an active site produced by a single vacancy on the sidewall follows two possible reaction routes, leading to the formation of a C-N bond or to dissociation of NO2. Accounting for defective adsorption sites allows a better understanding of microscopic mechanisms involved in technological applications of CNTs, e.g., gassensing devices.
Introduction Since their discovery by Iijima in 1991,1 carbon nanotubes (CNTs) have attracted much interest, as indicated by the large amount of both theoretical and experimental work produced on this topic in the past few years. However, a strong correlation between theoretical simulations and experimental results is often lacking, especially in studies concerning the adsorption of small molecules on the nanotube sidewall. These discrepancies can be mainly ascribed to strong differences between idealized models, utilized in calculations, and the situation arising in real experiments. Indeed, the effect of structural details of real CNTs, though crucial in determining their chemical and physical properties, have been so far poorly investigated. Considerations about the need for a better correlation between theoretical simulations and experiments in the study of CNTs play a major role when the latter are employed in the fabrication of gassensing devices.2 A detailed understanding of the microscopic mechanism taking place upon the absorption of a molecule on the nanotube sidewall might in fact lead to the development of more efficient and better tailored gas sensors. Hence, a step forward in the studies cited above would be given by an adequate description of the gas-nanotube interaction starting from structural models of the nanotube sidewall able to account for the real situation, as indicated by experiments. Namely, topological and structural defects are known to be a common feature in planar and rolled graphene layers, as recently demonstrated by experimental and theoretical works.3-5 In this work we analyze, by means of theoretical calculations, the interaction between an NO2 molecule and CNTs exhibiting defects on their sidewall, as suggested by experiments conducted on CNT-based gas-sensing devices.6,7 In particular, vacancies of carbon atoms on the sidewall change completely the behavior of nanotubes exposed to the gas. Theoretical predictions concerning effects triggered by the presence of defects, namely, * Author to whom correspondence should be addressed. Phone: +39 (0)75 585 5526. Fax: +39 (0)75 585 5606. E-mail:
[email protected]. † ISTM-CNR, Dipartimento di Chimica. ‡ Materials Engineering Centre.
vacancies, are further experimentally confirmed by means of X-ray photoelectron spectroscopy (XPS) measurements. Experimental Details Carbon nanotubes were deposited onto Si3N4/Si substrate patterned with platinum by pulsed plasma enhanced vapor deposition, according to a similar method described previously.8 The sample was inserted in an ultra-high-vacuum chamber (base pressure 10-7 mbar) via an entry lock. Measurements were performed at the gas-phase photoemission beamline of the ELETTRA synchrotron facility, with the sample mounted on a manipulator that allows annealing up to 500 K. An undulator source provides high-intensity synchrotron radiation in the photon energy range of 20-900 eV. Core-level spectra were measured using a photon energy of 339 eV for C1s, 470 eV for N1s, and 598 eV for O1s. High-resolution scans for the C, N, and O core regions were obtained in a vacuum and when NO2 molecules were injected in the XPS measurement system with a pressure of 10-4 mbar. The photoelectrical measurements were performed by fixing the temperature of the film at 430 K. Models and Computational Details Defects, in terms of vacancies and topological imperfections on the nanotube sidewall, represent a common feature in the plasma growing conditions. Transmission electron microscopy (TEM) and scanning tunneling microscopy performed on CNTs grown as described above show that the sidewalls contain a variety of defects,6,9,10 just like topological imperfections, such as pentagon-heptagon pairs11 and vacancies. This was also confirmed by Raman scattering measurements as previously reported.12 The effect of topological defects, like pentagonheptagon pairs, on the adsorption of small molecules on the nanotube sidewall has been already investigated in recent theoretical works.6,13,14 In particular, in the case of absorption of NO2, a weak interaction is found to take place, leading to a physisorbed species that is only slightly more bound than in the case of a defect-free sidewall. However, the presence of a vacancy might induce the formation of a strongly reactive site,
10.1021/jp0507290 CCC: $30.25 © 2005 American Chemical Society Published on Web 06/21/2005
13176 J. Phys. Chem. B, Vol. 109, No. 27, 2005
Mercuri et al.
in particular toward spin-polarized species, as shown by our previous theoretical calculations performed within a hybrid quantum mechanical/molecular mechanical (QM/MM) formalism.6 The increase in reactivity due to vacancies on the hexagonal network of nanotube sidewalls should present strong analogies with the case of graphene sheets, recently investigated by Causa` et al.15 Larger reactive propensities, especially those leading to strong chemical bonds on the defective sites, are expected to play a major role in the gas-sensing properties of nanotube-based devices. Hence, in our calculations we employed a model constituted by a (6,0) single wall CNT exhibiting a vacancy of a carbon atom on the sidewall. The high degree of curvature increases the reactivity on the sidewall surface, thus allowing us to emphasize differences with respect to the case of graphene. The interaction with a single molecule of NO2 was considered, since the poor reactivity of its dimer, N2O4, toward CNTs has been already clarified.16,17 Calculations were performed by applying the density functional theory as implemented in the CPMD program package.18 Gradient corrections to the local exchange and correlations were applied according to the schemes proposed by Becke19 and Lee, Yang, and Parr,20 respectively, within a local-spin-density (LSD) formalism. The electronic density was expanded in a basis set of plane waves within an energy cutoff of 70 Ry. Martins-Troullier pseudopotentials21 were used to describe the atom cores in the calculations. Periodic boundary conditions in one dimension were employed, thus considering a nanotube infinitely extended along its axis and choosing a cell length (12.79 Å) large enough to neglect interactions between the reactive site on the sidewall and its images. In directions orthogonal to the nanotube axis, a supercell approach was used, with a choice of cell parameters (14 and 10 Å, respectively) that allowed us to keep the simulating systems and their periodic images well separated. For all structures, geometry optimizations have been performed by minimizing the energies of structures obtained after a simulated annealing procedure, consisting of thermally annealing the system up to 300 K, equilibrating for around 1 ps, and cooling to almost negligible kinetic energies. This procedure should ensure that the optimized structure corresponds to the global energy minimum. Molecular dynamics simulations were carried out according to the Car-Parrinello formalism22 with a time step of 0.09676 fs and an electronic mass of 600 au Results and Discussion The oxygen, nitrogen, and carbon core-level photoemission spectra of CNT samples prior to and after exposure to NO2 are shown in Figure 1. It is worth nothing that the sample has been exposed to air throughout the transitional stages of the experiment (e.g., moving samples from the deposition apparatus to the spectrometer vacuum system). Before the acquisition time, the as-grown sample was maintained at 430 K and pumped from the atmospheric pressure down to 10-7 mbar for 12 h. Figure 1a shows the O1s core level of the CNT film while exposing at 430 K in a vacuum, 10-4 mbar of NO2, and a vacuum after NO2 exposure. From this figure, it is clear that no oxidation effects were revealed before exposure to NO2. However, by exposing the sample to 10-4 mbar of NO2 at 430 K, we observed a strong change in the O1s core-level signal, thus indicating the formation of a carbon-oxygen bond.13 This change in the O1s core level was not reversible; by annealing the sample in a vacuum for more than 10 h, no desorption of oxygen was detected. Figure 1b shows the N1s photoemission spectrum taken while exposing the sample to the gas. On the basis of the work of Jirsak et al.,26 we assign the N1s binding energy at
Figure 1. (a) O1s spectra of the CNT film while exposing the sample at 430 K in a vacuum (squares), 10-4 mbar of NO2 (diamonds), and a vacuum after NO2 exposure (circles). (b) N1s spectra of the CNT film while exposing the sample at 430 K to 10-4 mbar of NO2. (c) C1s spectra of the CNT film while exposing the sample at 430 K to 10-4 mbar of NO2. The inset of Figure 1c shows the C1s spectrum of the CNT film while exposing the sample at 430 K in a vacuum.
398.8 eV to NO physisorbed on the nanotube surface that, as will be addressed below, could be attributed to the dissociation of the NO2 molecule on the defect. Moreover, this peak cannot be attributed to to the NO2 signal that is located at about 407
Vacancy-Induced Chemisorption of NO2 on CNTs
Figure 2. Spin-density isosurfaces for the (6,0) CNT model with a vacancy in the singlet (top) and triplet (bottom) spin states. Green and red surfaces correspond to positive and negative spin densities, respectively. Surfaces corresponding to an electron density of 0.01e are plotted.
eV.28 In Figure 1c, the C1s core level of the sample before and after exposure to 10-4 mbar of NO2 is shown. A typical XPS C1s spectrum of as-deposited CNT film (inset of Figure 1c) shows a C1s peak located at 284.4 eV, which can be associated to an aromatic (pure graphitic) structure.29,30 Upon exposure to NO2, the C1s peak is deconvoluted (Figure 1c) into two additional components attributed to a carbon-nitrogen bond at 286 eV and a small contribution due to a carbon-oxygen bond at 288 eV, respectively.31 These experimental findings have been rationalized in terms of theoretical calculations. First of all, we analyzed the electronic structure of a (6,0) CNT exhibiting a vacancy of one carbon atom on the sidewall. Geometry optimizations, performed by means of the simulated annealing procedure described above, were carried out on both spin singlet and triplet states. Calculations provide almost the same total energy for both spin states. Nevertheless, the negligible energy splitting (less than 0.1 kcal mol-1) indicates small differences in the values of the electronic exchange integral in the two spin states. The analysis of the spin density, defined as the difference between the electronic density of R-electrons and that of β-electrons, plotted in Figure 2, gives a clear explanation of the small singlet-triplet energy gap. In both the triplet and the singlet states, two electrons occupy a different region of space; one electron is strongly localized on a carbon atom beside the vacancy hole, possessing σ-character, while another electron belongs to a strongly delocalized band, on the nanotube surface, with π-character. The lack of a strong
J. Phys. Chem. B, Vol. 109, No. 27, 2005 13177 correlation between these molecular orbitals causes a minor dependence of the total electronic energy on the spin coupling of the two electrons involved. Hence, according to our calculations, we hereafter focus on the singlet state, which is the most stable one. Upon geometry relaxation, the system undergoes distortion of the structure around the vacancy, with the reconstruction of a carbon-carbon bond, leading to the formation of a pentagonal ring (Figure 2). Moreover, an out-of-plane displacement of the dangling carbon atom opposite the pentagon is observed. These findings agree with tight-binding calculations performed by Charlier et al.23 The formation of a pentagonal ring and the out-of-plane displacement of the opposite carbon atom has already been observed in the case of a vacancy on graphene15,24 where, however, the 3-fold symmetry of the system induces relatively small differences in the bond configuration around the hole. However, in the case of CNTs the symmetry lowering allows degeneracy breaking and, thus, a stronger localization of the electronic pair involved in the bond reconstruction, as suggested by Heggie, Briddon, and co-workers in the case of a curved graphene sheet.24 The occurrence of pentagonal rings in defected CNTs is confirmed by a direct observation of STM images, as we already discussed in a previous paper.9 Besides structural properties, the highly localized σ-electron on the carbon atom tilted out of the hexagonal network will of course induce strong changes on the nanotube propensity to reaction, especially toward spin-polarized species. Therefore, we considered the interaction of the defective site with a NO2 molecule. A detailed exploration of possible reactive sites indicated that a strong chemisorption spontaneously occurs upon an “on-top” coordination with the dangling carbon atom, out of the sidewall surface. This is not surprising since the highly localized σ-electron should be the most active site toward a spin doublet species, such as the NO2 molecule, as discussed above. The interaction between an undefected nanotube and a NO2 molecule with the nitrogen atom coordinating with a carbon atom on the sidewall, hereafter referred to as the “nitro” configuration, has been extensively studied by means of theoretical calculations.16,25 However, the dangling carbon atom arising upon formation of a vacancy is also likely to react with NO2 in the same coordination, leading to the formation of a carbon-nitrogen bond. Indeed, calculations indicate a fair chemisorption of NO2 on the vacancy in a “nitro” configuration, as shown in Figure 3, the computed binding energy being 17.5 kcal mol-1. Moreover, the C-N bond length is 1.46 Å, only 0.03 Å longer than in nitrobenzene. Concerning the electronic properties, calculations suggest a spin doublet state for the chemisorbed adduct. The unpaired electron is mostly delocalized on the nanotube sidewall, as shown by the spin-density isosurface plotted in Figure 3, and possesses a strong π-character. Thus, the electronic structure of the chemisorbed species can be interpreted as a simple pairing of the σ-electron, localized on the dangling carbon atom, and the unpaired electron belonging to the NO2 molecule. The π-electron, delocalized on the nanotube surface, is almost unaffected by this process, being responsible for the spin polarization. This picture is confirmed by a practically complete overlap of spin densities on the CNT surface before and after reaction with NO2, as shown by a comparison of Figures 2 and 3. The formation of a C-N chemical bond can be directly related to the gas-sensing properties of defected CNTs and to the strong, nonreversible changes in their electronic properties upon reaction, as discussed in our previous work.6 However, although thermodynamically reasonable, the nitro attack of NO2 on the defect, shown by our
13178 J. Phys. Chem. B, Vol. 109, No. 27, 2005
Figure 3. Spin-density isosurfaces for the (6,0) CNT model with an NO2 molecule chemisorbed on a vacancy site in the nitro configuration. Surfaces corresponding to an electron density of 0.01e are plotted.
XPS measurements, is ruled by a small binding energy. Thus, more favorable processes might compete with chemisorption in the nitro configuration, and theoretical calculations have been focused on alternative reaction routes. To this end, we also considered the interaction between a vacancy on the (6,0) nanotube sidewall and a NO2 molecule in the “oxo” coordination, i.e., with the oxygen atom coordinating with a carbon atom on the sidewall. In this case, calculations do not show the formation of a stable adduct in which the NO2 molecule is directly bound to the nanotube sidewall. Rather, the NO2 molecule dissociates on the vacancy site, leading to the formation of a strong carbon-oxygen bond and releasing nitrogen monoxide, according to the reaction shown in Scheme 1. The computed reaction energy, defined as the difference between the electronic energies of products and reagents, respectively, is -62.3 kcal mol-1, suggesting a high propensity to dissociation of NO2 on the vacancy. The coordination between the CNT and the NO2 molecule in the oxo configuration is realized by means of the σ-electron localized on the dangling carbon atom opposite the pentagonal ring. Upon dissociation, a strong carbon-oxygen bond is formed, with a computed bond length of 1.245 Å, close to that of a typical carbonylic C-O double bond. Moreover, the dissociation process induces a modification of the electronic structure on the CNT, leading to a spin singlet stable state for the oxidized species and a negligible spin density on the sidewall surface, indicating a strong electron pairing, thus in contrast with the case of the unreacted nanotube. The dissociation of NO2 on the dangling carbon atom formed upon the creation of a vacancy on a SCHEME 1
Mercuri et al.
Figure 4. Reaction energy for the formation of a CNT-NO2 adduct in the oxo configuration. Computed values correspond to full geometry optimizations with the constraint of a fixed C-O distance. The sum of the energies for the separate species (infinite distance) is taken as reference.
nanotube sidewall should then proceed in a similar way to that observed for the vacancy on a graphene sheet, as discussed in ref 15. The microscopic details of the dissociation process have been investigated by means of Car-Parrinello molecular dynamics calculations. Simulations have been performed by applying constrained molecular dynamics in which a slowly decreasing oxygen-carbon distance, along the trajectory, was imposed. Our calculations indicate a spontaneous dissociation of NO2 at 300 K when oxo-coordinated to the vacancy. An animation showing the reaction trajectory is available as Supporting Information. This picture can be related to the more general case of the reaction of spin-polarized species with radical organic substrates, which are known to proceed with small barriers. Indeed, we also performed a linear transit search on intermediate configurations extracted from the molecular dynamics, and in all attempts, it was not possible to find a suitable saddle point, as expected. Geometry optimizations performed with the constraint of a fixed C-O distance (Figure 4) show a monotonic increase in the interaction energy until the dissociative domain (around 1.7 Å) is reached, thus suggesting the lack of high energy barriers. From our calculations, we can then conclude that the reaction of NO2 with a vacancy in the oxo coordination is an energetically favored process. Rather, the reaction barrier is more likely due to the decrease in entropy associated with rearrangements leading to suitable configurations for the reactive process. It is worth noting that the formation of NO upon dissociation has also been detected by XPS measurements, as already discussed. Hence, dissociative chemisorption of NO2 on a vacancy is very likely to be observed at operating temperatures, and it can considerably affect the electronic
Vacancy-Induced Chemisorption of NO2 on CNTs properties of a defective CNT, due to the strong electronic effects of the oxygen atom. As stated above in the case of nitro coordination, chemical processes on the nanotube sidewall are expected to play a major role in the observed nonreversible, strong gas-sensing behavior of defected CNTs, and this picture is further confirmed in the case of a dissociative oxo coordination. Moreover, the formation of a stable carbon-oxygen double bond can be directly observed by means of standard chemical analysis techniques, thus bearing out the picture given by our calculations. Conclusions In summary, ab initio calculations on the interaction between NO2 and defected CNTs give clear evidence of strong chemical interactions. Such interactions correspond to both associative and dissociative adsorption of the NO2 molecule on the nanotube sidewall, giving rise to new chemical bonds. Indeed, XPS measurements carried out on CNT samples obtained in growing conditions suitable for the construction of gas-sensing devices further confirm the nonreversible formation of stable C-O and C-N bonds, as a result of NO2 absorption. In particular, experimental measurements give a stronger evidence of the formation of a carbonyl group than a C-N bond, thus agreeing with the different computed reaction energies for the dissociative and associative processes, respectively. These results indicate that many supposedly intrinsic properties measured on asprepared CNTs may be severely reinterpreted by the presence of defects, namely, vacancies, coming from the deposition procedure. As expected, the presence of vacancies on the nanotube sidewall dramatically changes the reactive propensity of CNTs; the role of defects must then be accounted for in theoretical modeling of nanotubes grown under particular conditions. Such an approach allows a stronger correlation between experimental measurements and theoretical simulations, particularly relevant in application fields such as the realization of CNT-based gas-sensing devices. Acknowledgment. L. Lozzi and S. Santucci are gratefully acknowledged for their help in XPS measurements as well as M. Coreno and M. De Simone (INFM-TASC Gas Phase Beamline, Sincrotrone Trieste) for giving access to their equipment and for technical support. F.M. thanks F. Tarantelli and F. De Angelis for helpful discussions. Supporting Information Available: Three-dimensional rotable images of the optimized structures and an animation of
J. Phys. Chem. B, Vol. 109, No. 27, 2005 13179 the dissociative reaction in Scheme 1. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Iijima, S. Nature 1991, 354, 56. (2) Kong, J.; Franklin, N. R.; Zhou, C. W.; Chapline, M. G.; Peng, S.; Cho, K. J.; Dai, H. J. Science 2000, 287, 622. (3) Hashimoto, A.; Suenaga, K.; Gloter, A.; Urita, K.; Iijima, S. Nature 2004, 430, 870. (4) Lu A.; Pan, B. C. Phys. ReV. Lett. 2004, 92, 105504. (5) Rossato, J.; Baierle, R. J.; Fazzio, A.; Mota, R. Nano Lett. 2005, 5, 197. (6) Valentini, L.; Mercuri, F.; Armentano, I.; Cantalini, C.; Picozzi, S.; Lozzi, L.; Santucci, S.; Sgamellotti, A.; Kenny J. M. Chem. Phys. Lett. 2004, 387, 356. (7) Valentini, L.; Armentano, I.; Lozzi, L.; Cantalini, C.; Santucci, S.; Kenny, J. M. J. Vac. Sci. Technol., A 2004, 22, 1450. (8) Valentini, L.; Armentano, I.; Kenny, J. M.; Lozzi, L.; Santucci, S. Mater. Lett. 2004, 58, 470. (9) Valentini, L.; Armentano, I.; Puglia, D.; Lozzi, L.; Santucci, S.; Kenny, J. M. Thin Solid Films 2004, 449, 105. (10) Charlier, J.-C. Acc. Chem. Res. 2002, 35, 1063. (11) Stone, A. J.; Wales, D. J. Chem. Phys. Lett. 1986, 128, 501. (12) Chhowalla, M.; Teo, K. B. K.; Ducati, C.; Rupesinghe, N. L.; Amaratunga, G. A. J.; Ferrari, A. C.; Roy, D.; Robertson, J.; Mine, W. I. J. Appl. Phys. 2001, 90, 5308. (13) Chakrapani, N.; Zhang, Y. M.; Nayak, S. K.; Moore, J. A.; Carroll, D. L.; Choi, Y. Y.; Ajayan, P. M. J. Phys. Chem. B 2003, 107, 9308. (14) Grujicic, M.; Cao, G.; Singh, R. Appl. Surf. Sci. 2003, 211, 166. (15) Ghigo, G.; Maranzana, A.; Tonachini, G.; Zicovich-Wilson, C. M., Causa`, M. J. Phys. Chem. B 2004, 108, 3215. (16) Yim W. L.; Gong X. G.; Liu Z. F. J. Phys. Chem. B 2003, 107, 9363. (17) Ellison, M. D.; Crotty, M. J.; Koh, D.; Spray, R. L.; Tate, K. E. J. Phys. Chem. B 2004, 108, 7938. (18) CPMD, version 3.7; IBM Corp., MPI fu¨r Festko¨rperforschung: Stuttgart, Germany, 1990-2003, 1997-2001 (http://www.cpmd.org). (19) Becke, A. D. Phys. ReV. A 1988, 38, 3098. (20) Lee, C.; Yang, W.; Parr, R. C. Phys. ReV. B 1988, 37, 785. (21) Troullier, N.; Martins, J. Phys. ReV. B 1991, 43, 1993. (22) Car, R.; Parrinello, M. Phys. ReV. Lett. 1985, 55, 2471. (23) Ajayan P. M.; Ravikumar, V.; Charlier, J.-C. Phys. ReV. Lett. 1998, 81, 1437. (24) El-Barbary A. A.; Telling R. H.; Ewels C. P.; Heggie M. I.; Briddon P. R. Phys. ReV. B 2003, 68, 144107. (25) Zhao, J.; Buldum, A.; Han, J.; Lu, J. P. Nanotechnology 2002, 13, 195. (26) Jirsak, T.; Kuhn, M.; Rodriguez, J. A. Surf. Sci. 2000 457, 254. (27) Zheng, W. T.; Sjostrom, H.; Ivanov, I.; Xing, K. Z.; Broitnen, E.; Salameck, W. R.; Greene, J. E.; Sundgren, J. E. J. Vac. Sci. Technol., A 1996, 14, 2696. (28) Goldoni, A.; Larciprete, R.; Petaccia, L.; Lizzit, S. J. Am. Chem. Soc. 2003, 125, 11329. (29) Baker, M. A.; Hammer, P. Surf. Interface Anal. 1997, 25, 629. (30) Mansour, Z. A.; Ugolini, D. Phys. ReV. B 1993, 47, 10201. (31) Marton, D.; Boyd, K. J.; Al-Bayati A. H.; Todorov S. S.; Rabalais, J. W. Z. Phys. ReV. Lett. 1994, 73, 118.
