Letter pubs.acs.org/NanoLett
Oxidation Resistance of Reactive Atoms in Graphene Matthew F. Chisholm,*,† Gerd Duscher,†,‡ and Wolfgang Windl§ †
Materials Science and Technology Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831 Department of Materials Science and Engineering, University of Tennessee, Knoxville, TN 37996 § Department of Materials Science and Engineering, The Ohio State University, Columbus, Ohio 43210 ‡
S Supporting Information *
ABSTRACT: We have found that reactive elements that are normally oxidized at room temperature are present as individual atoms or clusters on and in graphene. Oxygen is present in these samples but it is only detected in the thicker amorphous carbon layers present in the graphene specimens we have examined. However, we have seen no evidence that oxygen reacts with the impurity atoms and small clusters of these normally reactive elements when they are incorporated in the graphene layers. First principles calculations suggest that the oxidation resistance is due to kinetic effects such as preferential bonding of oxygen to nonincorporated atoms and H passivation. The observed oxidation resistance of reactive atoms in graphene may allow the use of these incorporated metals in catalytic applications. It also opens the possibility of designing and producing electronic, opto-electronic, and magnetic devices based on these normally reactive atoms. KEYWORDS: Graphene, impurity oxidation, ADF imaging, aberration-correction
A
defects and incorporated impurities, but the same defects and impurities have been seen in all the forms of graphene we have examined. Another common feature of all graphene that we have examined by electron microscopy is the large amount of extra carbon-containing material present on the observed graphene microscopy specimens. We have also found that many graphene samples are in fact graphite or multilayered graphene. These samples are not fully single layer materials and, thus, cannot be used to evaluate properties of “bulk” graphene. In this Letter, we will limit the discussion to Si, O and Fe impurities. Carbon flakes were deposited on holey (or lacey) carbon transmission electron microscopy (TEM) grids and were examined using ADF imaging in an aberration-corrected scanning transmission electron microscope operated at 60 kV. This acceleration voltage is below the knock-on damage threshold for carbon atoms in graphene although carbon atoms at lattice edges and in some lattice defects were seen to be displaced by the 60 kV beam.4 Medium-angle ADF imaging (58 mrad inner angle and ∼200 mrad outer angle) has already been demonstrated to be capable of determining the location and identity of all atoms in single-layer materials.1 It is acknowledged that lattice-resolution transmission electron microscopy studies of graphite-based carbon nanostructures followed soon after the introduction of objective lens aberration correctors and the isolation of free-standing graphene.5−7 The
consequence of doping a two-dimensional material like graphene is that the dopants are fully exposed to the atmosphere and the usual mechanisms for corrosion protection are not applicable. Here we show how elements are incorporated into graphene and explain how these reactive surface atoms remain unoxidized even when stored in air. In the present work, annular dark-field (ADF) imaging in a scanning transmission electron microscope is used to locate and identify the impurity atoms in the graphene lattice.1 Electron energyloss spectroscopy (EELS) is also used to confirm the impurity atom identification.2,3 We further combine the imaging data with first-principles density functional calculations of the reaction of the impurity with graphene and the impurity with oxygen. Impurity atoms on the graphene surface are seen to quickly move across and off open areas of perfect graphene. Interestingly, our density functional theory calculations indicate that even when neglecting reaction barriers the energy required to replace a carbon atom with an impurity is prohibitively large for the impurities we studied. This is true for carbon atoms in the perfect graphene lattice and for carbon atoms in structural defects in graphene. We find that these impurities can only be incorporated into the graphene lattice when filling existing vacancies. Graphene samples can be produced using several techniques. In this investigation, flakes of graphene films grown on Cu and Ni foils (after the metal substrate has been etched away) and from reduced graphene oxide produced by different processing methods have been examined. The reduced graphene oxide samples typically contained a higher density of structural © 2012 American Chemical Society
Received: May 23, 2012 Revised: July 23, 2012 Published: August 20, 2012 4651
dx.doi.org/10.1021/nl301952e | Nano Lett. 2012, 12, 4651−4655
Nano Letters
Letter
Figure 1. ADF images of graphene. The as recorded data (a,b,c) were corrected to remove noise and probe tail effects (d,e,f). The images from an area of perfect graphene (a,d) show individual carbon atoms are resolved. Images b and e show a Si atom in a C vacancy site. Images c and f show an Fe atom in a C divacancy site. The scale mark on each image corresponds to 0.2 nm.
Figure 2. Electron energy-loss spectrum image data from single atoms incorporated in graphene. Simultaneously collected ADF image of a Si atom in graphene (a), Si composition map (b), and spectrum obtained by summing 9 pixels over the Si atom (c). Simultaneously collected ADF image of an Fe atom (d), Fe composition map (e), and the spectrum obtained by summing 9 pixels over the Fe atom (f). No evidence for oxygen (O−K edge at 535 eV) is seen in either spectrum. The scale marks in each image correspond to 0.2 nm.
the etchant used to free graphene films by removing their metal substrates. Figure 1 shows ADF images of perfect graphene, a single Si atom segregated to a carbon vacancy and a single Fe atom segregated to a carbon divacancy. Si has also been found in graphene divacancies, but Fe has not been imaged in graphene single vacancy sites. The observed intensities are consistent with there being just a single impurity atom in Figures 1b,c, that is, no additional oxygen. EELS confirms the identification of these impurity atoms as Si and Fe. The spectra in Figure 2 were obtained using the spectrum-image strategy.12−14 The Si-L2,3 edge summed over 9 pixels centered on the maximum shows an edge onset of about 102 eV, which is less than the oxide value and close to the value of SiC.15,16 No oxidation features are apparent in the Si-L2,3 edge and no O−K edge was detectable in the spectrum. The edge feature in
use of microscope operating voltages below the knock-on damage threshold of sp2 bonded carbon has also been discussed.8−10 While individual C atoms in the boron nitride lattice have been previously imaged1 and high-resolution TEM (HRTEM) has been used to look at charge distribution around N atoms in graphene,11 the current work provides direct evidence that reactive elements such as Fe and Si when incorporated in graphene remain nonoxidized. We have found that silicon, chromium, iron, cobalt, nickel, and copper impurity atoms on graphene remain in elemental form even when the doped graphene is exposed to air for extended times. In this Letter, we will concentrate on the two most commonly seen impurities found on our specimens, Si and Fe. The source of the Si impurities is likely the glassware used to process graphene oxide. The Fe atoms are mainly from 4652
dx.doi.org/10.1021/nl301952e | Nano Lett. 2012, 12, 4651−4655
Nano Letters
Letter
Figure 3. Calculated atomic structures of Si and Fe atoms in (left) single vacancy site, (middle) divacancy site (V2), and (right) a region of 5- and 7atom defects that result from reconstruction of the graphene lattice with four vacancies (V4). Bond lengths are given in the Supporting Information.
Information. The regular Stone-Wales defect in contrast conserves the number of C atoms and just consists of a bond rotation. The formation energy (5.2 eV) of this defect makes it energetically highly unfavorable. Both Si and Fe adatoms are found to bind to perfect graphene. However, these adatoms are extremely mobile. This allows these adatoms to move over the graphene surface and to quickly find and bind to defect sites, where their energy is lowered. This is consistent with the fact that no Si or Fe adatoms were observed in their equilibrium adsorption sites. The formation energies of single, double, and 5−7 reconstructed vacancies per missing carbon atom are 8.0, 3.8, and 4.6 eV, respectively. In thermal equilibrium, one would thus expect the majority of vacancies to be present in the form of divacancies. The next most likely configurations are various 5−7 reconstructed vacancy complexes. Essentially no single vacancies should be present. This is confirmed by our STEM observations in which divacancies and 5−7 reconstructions have been observed, but single vacancies not passivated with impurities have not been seen. While both Si and Fe in single vacancy sites reduce the total energy of the system, single vacancies in graphene are highly unlikely to be present. Experimentally, Si was observed as an incorporated impurity in single vacancies, double vacancies, and reconstructed 5−7 vacancy complexes (Figure 3), whereas Fe was only observed in the latter two, but not in single vacancies. The calculated energies of these configurations are summarized in Table 1. We
Figure 2b not only identifies the atom to be Fe it also indicates that it is not oxidized. The ratio of the Fe L3/L2 edges is lower than any seen for the various forms of Fe oxide.17 These edges are better seen in the spectrum obtained by locating the electron beam on the Fe atom as presented in the Supporting Information (see Figure 1 in Supporting Information). No evidence, structural, visual, or spectroscopic, of oxidized Si and Fe atoms incorporated in graphene was found. To our knowledge, there have been no experimental nor theoretical studies on the oxidation resistance of iron or silicon incorporated in graphene. Bulk iron and silicon carbides are ceramic-like in mechanical behavior and chemically more inert than pure Si or Fe. Thus, it may not be surprising to some that these incorporated elements in graphene are also less reactive. However, it must be remembered that these two-dimensional carbides are not encountered outside of graphene. Additionally, the surface Fe and Si atoms in three-dimensional carbides are readily oxidized at room temperature in air. The corrosion resistance seen in bulk carbides is the result of a protective coating of oxide or graphitic carbon.18−20 Structural defects in graphene are predominately connected 5- and 7-atom rings embedded in the 6-atom rings of the perfect lattice.21−23 Larger rings (8- and 9-atoms) and smaller rings (much rarer 4-atom) have also been observed. It is worth noting that no pure carbon Stone-Wales defects21 (a specific form of two 5- and two 7-atom rings) have been observed. It is possible that, in the forms of graphene we have examined, the bond rotation necessary for a Stone-Wales defect does not occur. Instead vacancies are formed, clustered, and reconstructed to produce the defect structures seen. Si and Fe impurities have been seen in these 5-and 7- atom defects (see Figure 3) but they do not appear to be preferentially located on these structural defects. Density functional theory (DFT) calculations have been used to study the energetics of Si, Fe, and O atoms in contact with perfect and defective graphene, using PAW potentials24,25 within PBE26 and Grimme’s parametrization for van der Waals dispersive interaction.27 Strong-correlation effects for the Fe delectrons have been addressed by adding a Hubbard-U to the Hamiltonian with U = 4 eV and J = 0.95 eV as recently proposed.28 The diffusion coefficient of Si and Fe adatoms on graphene was estimated by a series of DFT-MD simulations at varying temperatures. The results are summarized below and presented in more detail in the Supporting Information. Examination of the observed 5−7 atom defect structures, show that they have locally less C atoms than perfect graphene and thus can be considered as reconstructions of vacancy clusters. These reconstructions result in huge energy savings when compared to the sum of the vacancy formation energies. For further discussion of specific examples see the Supporting
Table 1. PBE+U+vdW Results for Reaction Energies of Iron and Silicon with Single Vacancies (V1), Divacancies (V2), and Observed Four-Vacancy Reconstructions (V4, Figure 3c) in Graphene reactant 1
reactant 2
product
ΔE (eV)
Fead Fead Fead Siad Siad Siad
V1 V2 V4 V1 V2 V4
Fes FeV2 FeV4 Sis SiV2 SiV4
−2.6 −4.3 −4.5 −7.4 −6.8 −6.6
find that the reason Si atoms are seen in single vacancy sites is that reactions where a vacancy is split off the divacancy leaving behind a “substitutional” Sis are energetically favorable. This is not the case for Fe. Since the chemical potential (energy contribution of each atom) of the carbon atoms is given by the energy of C atoms in perfect graphene, the chemical potential can be calculated for the impurity atoms. Whereas Si and Fe usually have a chemical potential lower than the energy of atoms in the gas phase, this is not the case for Fe incorporated 4653
dx.doi.org/10.1021/nl301952e | Nano Lett. 2012, 12, 4651−4655
Nano Letters
Letter
reduction mechanism, we have also seen from MD that Hpassivated Si and Fe atoms incorporated in graphene repel approaching O2 molecules (Movies 3 and 4 in Supporting Information). Thermodynamically, H should be replaced by oxygen, but there is a barrier to be overcome before the reaction can occur. As a result, no measurable concentration of oxidized impurities remains in the graphene layer. On the other hand, adsorbed impurities bind to oxygen with an energy similar to the gas phase and have typically a small adsorption energy often dominated by the vdW interaction and thus are desorbed from the surface typically within the MD time frame (Movie 5 in Supporting Information). This means that the catalytic effect of the impurities is only activated by their incorporation into defect sites. Carbon layers down to a thickness of a single layer have been known to form on metal surfaces for more than 50 years.30−34 Graphene on metal surfaces is also known in catalysis, where the deposition or formation of graphitic carbon can lead to deactivation.35,36 So while it is known that graphene layers on metals provide an oxidation barrier, it has not yet been recognized that graphene layers offer a level of oxidation protection to individual atoms or small clusters of atoms in/on the graphene layer. In a very recent paper,37 it is shown that metallic impurities (with Fe being the largest fraction of them) remain even after the oxidation treatment to obtain graphene oxide and after the thermal reduction process to produce graphene. These authors use chemical reactions to determine that the metals remained but did not distinguish between the elemental metal and its oxide. In our study, the individual metal atoms are imaged, their incorporation sites are identified, and it is conclusively shown that the normally reactive elements have remained nonoxidized. With this information, first principles calculations explain these observations are due to preferential bonding of O to nonincorporated atoms and H passivation effects. This is a potentially important discovery; improved resistance to oxidation has important consequences for some catalytic reactions and small devices based on single atoms or small clusters of non-noble metals. Oxidation of most metals precludes their use as catalysts because in their oxidized state they are catalytically inactive or show vastly reduced catalytic activity. Graphene as a substrate appears to protect single atoms and small clusters of atoms from oxidation without completely isolating them. While we have only examined the microstructure and have not explored catalytic potential of impurities incorporated in graphene, it is anticipated that inert graphene can be transformed to a very active catalyst by embedding metal clusters and individual atoms in defects in graphene. This phenomenon can be used to employ the catalytic, electronic, and magnetic properties of more elements than are usually available for unprotected materials in air or other corrosive environments.
into a single vacancy, giving clear evidence that this configuration is energetically prohibited. We have examined the interaction between oxygen and Si and Fe in the gas phase as well as in the adsorbed and observed impurity sites on graphene, substitutional (s), divacancy-center (V2), and four-vacancy reconstruction (V4), see Figure 3. O2 in the gas phase and adsorbed on graphene (where we find the adsorption energy of 0.13 eV to be nearly entirely due to van der Waals interactions) was used as the interacting species. Atomic oxygen can in principle be present on graphene; the strong O2 binding energy in the gas phase of 6.7 eV is reduced to 1.8 eV on graphene. However, the high-binding energy of O to graphene (2.5 eV), along with its considerably slower diffusion29 should make it a far less likely initial reactant than molecular O2. Negative reaction enthalpies are calculated between oxygen and all Si and Fe configurations examined (Tables 2 and 3), Table 2. PBE+U+vdW Results for Reaction Enthalpies of Silicon with Oxygen reactant 1
reactant 2
product
ΔE (eV)
Sig Siad Sis SiV4 SiV2 Sig Siad Sis SiV4 SiV2
O2,g O2,g O2,g O2,g O2,g Oad Oad Oad Oad Oad
(SiO2)g (SiO2)ad SisO2 SiV4O2 SiV2O2 (SiO)g (SiO)ad SisO SiV4O SiV2O
−7.3 −7.2 −2.3 −1.6 −0.6 −6.3 −5.9 −3.6 −2.9 −1.8
Table 3. PBE+U+vdW Results for Reaction Enthalpies of Iron with Oxygen reactant 1
reactant 2
product
ΔE (eV)
Feg Fead Fes FeV2 FeV4 Feg Fead Fes FeV2 FeV4
O2,g O2,g O2,g O2,g O2,g Oad Oad Oad Oad Oad
(FeO2)g (FeO2)ad FesO2 FeV2O2 FeV4O2 (FeO)g (FeO)ad FesO FeV2O FeV4O
−2.6 −3.1 −4.7 −1.1 +0.2 −2.5 −2.9 −5.6 −2.1 −0.9
which in principle suggest that oxidation should take place. These incorporated impurities are not impervious to oxidation. Our MD simulations also confirm this (Movie 1 in Supporting Information). However, the calculated numbers show that the reactions between gaseous and adsorbed impurities and oxygen result in significantly higher energy gain than oxygen reactions with incorporated impurities. This has the important effect that oxygen bound to an incorporated impurity is easily reduced from that impurity by other adsorbed or gaseous impurities passing nearby, which happens even on the first-principles MD time scale of a few picoseconds (Movie 2 in the Supporting Information). The only exception to this is the case of Fes, where oxygen is stronger bound to the incorporated than to the adsorbed or gaseous impurity. Fes, however, as we have discussed above, should not be present. In addition to this
■
ASSOCIATED CONTENT
* Supporting Information S
Additional details on imaging conditions, computational methods and results and movies from our molecular dynamics simulations. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Notes
The authors declare no competing financial interest. 4654
dx.doi.org/10.1021/nl301952e | Nano Lett. 2012, 12, 4651−4655
Nano Letters
■
Letter
ACKNOWLEDGMENTS Graphene samples examined in the course of this study were supplied by Gyula Eres of ORNL, Wei He of University of Tennessee and Hongjie Dai of Stanford University. This research was supported by the Materials Sciences and Engineering Division of the Office of Basic Energy Sciences, U.S. Dept. of Energy and by NSF Award Number DMR0925529 and the Center for Emergent Materials at The Ohio State University, a NSF MRSEC (Grant DMR-0820414). W.W. acknowledges support from the Ohio Supercomputer Center under project PAS0072.
■
REFERENCES
(1) Krivanek, O. L.; et al. Nature 2010, 464, 571. (2) Egerton, R. F. Electron Energy-Loss Spectroscopy in the Electron Microscope, 2nd ed.; Plenum Press: New York, 1996. (3) Suenaga, K.; Koshino, M. Nature 2010, 468, 1088. (4) Zobelli, A.; Gloter, A.; Ewels, C. P.; Seifert, G.; Colliex, C. Phys. Rev. B 2007, 75, 245402. (5) Suenaga, K.; Wakabayashi, H.; Koshino, M.; Sato, Y.; Urita, K.; IIjima, S. Nat. Nanotechnol. 2007, 2, 358−360. (6) Meyer, J. C.; Kisielowski, C.; Erni, R.; Rossell, M. D.; Crommie, M.; Zettl, A. Nano Lett. 2008, 8, 3582−3586. (7) Gass, M. H.; Bangert, U.; Bleloch, A. I.; Wang, P.; Nair, R. R.; Geim, A. K. Nat. Nanotechnol. 2008, 3, 676−681. (8) Sawada, H.; et al. Ultramicroscopy 2010, 110, 958−961. (9) Kaiser, U.; et al. Ultramicroscopy 2011, 111, 1239−1246. (10) Krivanek, O. L.; et al. Ultramicroscopy 2010, 110, 935−945. (11) Meyer, J. C.; et al. Nat. Mater. 2011, 10, 209−215. (12) Jeanguillaume, C.; Colliex, C. Ultramicroscopy 1989, 28, 252. (13) Bosman, M.; et al. Phys. Rev. Lett. 2007, 99, 086102. (14) Kimoto, K.; et al. Nature 2007, 450, 702. (15) Batson, P. E. Nature 1993, 366, 727. (16) Biggerstaff, T. L.; et al. Appl. Phys. Lett. 2009, 95, 032108. (17) Kurata, H.; Tanaka, N. Microsc. Microanal. Microstruct. 1991, 2, 183. (18) Powers, J. M.; Somorjai, G. A. Surf. Sci. 1991, 244, 39. (19) Zhao, X. Q.; et al. J. Appl. Phys. 1996, 80, 5857. (20) Galwey, A. K.; et al. Corros. Sci. 1974, 14, 527. (21) Stone, A. J.; Wales, D. J. Chem. Phys. Lett. 1986, 128 (5−6), 501−503. (22) Hashimoto, A.; et al. Nature 2004, 430, 870. (23) Huang, P.; et al. Nature 2011, 469, 389. (24) Blöchl, P. E. Phys. Rev. B 1994, 50, 17953. (25) Kresse, G.; Joubert, J. Phys. Rev. B 1999, 59, 1758. (26) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865. (27) Grimme, S. J. Comput. Chem. 2006, 27, 1787. (28) Wehling, T. O.; Lichtenstein, A. I.; Katsnelson, M. I. Phys. Rev. B 2011, 84, 235110. (29) Radovic, L. R.; et al. Carbon 2011, DOI: 10.1016/ j.carbon.2011.05.037. (30) Hagstrom, S.; Lyon, H. B.; Somorjai, G. A. Phys. Rev. Lett. 1965, 15, 491. (31) Lyon, H. B.; Somorjai, G. A. J. Chem. Phys. 1967, 46, 2539. (32) Morgan, A. E.; Somorjai, G. A. Surf. Sci. 1968, 12, 405. (33) May, J. W. Surf. Sci. 1969, 17, 267. (34) Grant, J. T.; Haas, T. W. Surf. Sci. 1970, 21, 76. (35) Schlogl, R. In Handbook of Heterogeneous Catalysis; Ertl, G., Knozinger, H., Schuth, F., Weitkamp, J., Eds.; Wiley-VCH: Weinheim, 2008; Vol. 1, p 357. (36) Moulijn, J. A.; van Diepen, A. E.; Kapteijn, F. In Handbook of Heterogeneous Catalysis; Ertl, G., Knozinger, H., Schuth, F., Weitkamp, J., Eds.; Wiley-VCH: Weinheim, 2008; Vol. 4, p 1829. (37) Ambrosi, A.; et al. Angew Chem., Int. Ed. 2012, 51, 500−503.
4655
dx.doi.org/10.1021/nl301952e | Nano Lett. 2012, 12, 4651−4655
