CONTENTS
5.2 Graphene Substrate Interactions and Band Gap Formation
5.3 Graphene Formation on h-BN(0001)
5.3.1 Graphene Growth on h-BN Monolayers Formed on Metal Substrates
5.3.2 Graphene/Boron Nitride Interfacial Interactions
5.4 Graphene Growth on MgO(111)
5.4.3 Interfacial Charge Transfer
5.5 Graphene Growth on Mica by MBE
5.6 Graphene Growth on Co3O4(111)/Co(111) by MBE
5.6.2 Magnetic Properties and Device Applications
Graphene consists of a single atomic layer of graphite—C atoms arranged in a hexagonal lattice, with sp2 hybridization. Measurements on isolated single- and few-layer graphene sheets have demonstrated a number of properties with great potential for novel device applications, including electron or hole mobilities orders of magnitude greater than in Si [1,2], electron spin diffusion lengths on the order of microns [3], and predicted [4] ferromagnetism in the vicinity of a ferromagnet. The initial demonstration, by Novoselov et al., of some of these properties on isolated single-and few-layer graphene sheets [5,6] has stimulated intense interest in the electronic and spintronic properties of graphene.
A key step toward the industrial development of graphene-based devices is the ability to directly grow graphene on dielectric substrates. To date, however, the great majority of published accounts employ one of two approaches: (1) the physical transfer of graphene sheets grown on metal substrates [7,8,9] or from highly oriented pyrolytic graphite (HOPG) [5,10] onto patterned SiO2/Si substrates, or (2) the graphitization of SiC(0001) by high-temperature evaporation of Si [11,12,13]. The former approach has been successful in terms of elucidating fundamental physical properties of the graphene system, but obviously presents severe problems to device fabrication on an industrial scale, not the least of which is the formation of nanoscale interfacial inhomogeneities [14]. The latter approach has been used successfully to fabricate high-frequency field effect transistors (FETs) [15,16], but the absence of a band gap (at least in zero applied field) leads to low device on/off current ratios (see below). Additionally, this approach is limited to SiC.
Recently, macroscopically large and uniform graphene films (as opposed to nanoflakes) have been grown, without metal catalysts, on h-BN(0001) monolayers on Ru(0001) [17], on MgO(111) [18,19], on mica [20,21], and on Co3O4(111)/Co(111) [22]. Preliminary accounts [23] also suggest that continuous graphene films can be grown on α-Al2O3(0001). These studies have employed scalable methods, including chemical vapor deposition (CVD), physical vapor deposition (PVD), and molecular beam epitaxy (MBE). The results, reviewed below, present exciting possibilities for the development of graphene-based electronic and spintronic devices. These results, however, also demonstrate that the properties of graphene sheets are strongly impacted by interactions with the substrate. This is generally true even for isolated graphene sheets physically transferred to high dielectric constant substrates [15,24], or to boron nitride (BN) nanocrystals [25], where dramatic increases in carrier mobility have been reported, relative to mobilities for graphene on SiO2. In the case of high dielectric substrates, screening of electron scattering from charged impurities may be responsible for the observed mobility enhancement [24], while for graphene on BN, substrate flatness appears to play a role [25]. For graphene grown directly on substrates, such issues as graphene/substrate charge transfer [17,18], interfacial chemical reaction or reconstruction [19,26], and the relative orientations and interactions of graphene sheets [12,27] also present themselves. Such issues not only are important in a fundamental, scientific sense, but directly impact practical device applications.
The primary focus of this review is on direct graphene growth on dielectric substrates, specifically h-BN(0001), and certain oxides—MgO(111), mica, and Co3O4(111). There is a growing literature on the use of metal catalysts to form graphene sheets on insulating substrates [28,29,30], but the chemistry involved appears to be essentially that of graphene growth on transition metal substrates [8,31]. Additionally, the presence of transition metal impurities presents potentially serious difficulties for device applications because even low concentrations of transition metal impurities in Si or other semiconductors can significantly degrade electronic device properties.
Finally, a section of this review is devoted to exploring potential device applications—including electronic applications made possible by the controlled growth of graphene on MgO, which results in a band gap, or spintronics applications for graphene grown on magnetically polarizable oxides, such as Co3O4. Graphene conduction electrons are predicted to become spin-polarized in proximity to a ferromagnet [4]. Preliminary calculations [32] suggest the possibility that such a system may exhibit nonlocal magnetoresistance values orders of magnitude higher than spin valves or spin Direct Graphene Growth on Dielectric Substrates FETs that are based on the injection and diffusion of individual spins through a graphene sheet. Such effects, if demonstrated, could yield manufacturable, low-power, and nonvolatile spintronics devices operating at or above room temperature [32].
5.2 GRAPHENE SUBSTRATE INTERACTIONS AND BAND GAP FORMATION
The lattice of a single graphene sheet (Figure 5.1a) consists of a two-atom unit cell with crystallographically distinct sites A and B that are chemically identical in the isolated graphene lattice. This chemical identity results in the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) being degenerate at the Dirac point (Figure 5.1b) [33]. Thus an isolated graphene sheet is a zero-band gap semiconductor. This band structure—including the singularity at the Dirac point and linear dispersion in its vicinity—is responsible for many of graphene’s unusual properties, including measured DC mobilities for an isolated graphene sheet in excess of 100,000 V cm–2 s–1 (at low carrier densities), several orders of magnitude greater than that of an electron in Si [1]. The band structure shown in Figure 5.1b, however, presents a major problem for applications in such devices as FETs. Specifically, the “off state” in such a device would, at room temperature or above, still exhibit a considerable current, leading to very low on/off ratios. FETs formed with graphene grown on the Si face of SiC(0001), for example, exhibit on/off ratios of ~30 [34]. A band gap of 0.26 eV for single-layer graphene on Si-terminated SiC(0001) has been reported on the basis of angle-resolved photoemission measurements [35]. The magnitude of the gap, however, decreases with additional graphene layer formation [35], and a band gap of this magnitude may in any case be too small for conventional (nontunneling) FET applications [36,37]. Importantly, there is a lack of transport measurements pointing to the formation of such a band gap.
FIGURE 5.1 Substrate-induced band gap formation in graphene. (a) Structure of the isolated graphene lattice. A sites and B sites are crystallographically distinct but chemically identical, leading to (b) HOMO/LUMO degeneracy at the Dirac point. (c) Interaction of graphene with a commensurate substrate (atoms arbitrarily labeled X and Y, as shown) can, depending on the relative orientation of the two layers, destroy the A site/B site chemical equivalence, lifting HOMO/LUMO degeneracy at the Dirac point and (d) resulting in a band gap.
The formation of a ~0.25 eV band gap in bilayer graphene in the presence of an external electric field has been demonstrated [38]. The required presence of an external field, however, as well as the apparent limiting value of only 0.25 eV, presents obvious difficulties for attaining a high on/off ratio in a graphene-based nontunneling [36,37] FET, even assuming the ability to precisely form a bilayer graphene channel. Band gap formation has also been reported [39] for graphene nano-ribbons, but this effect is sensitive to the width of the ribbon on the nanoscale, and perhaps as well to the termination of the nano-ribbon edge sites [40]. Thus the most robust way forward in introducing a band gap to graphene for device applications would appear to be the growth of graphene on a suitable substrate to induce such interactions as shown in Figure 5.1c and d.
5.3 GRAPHENE FORMATION ON h-BN(0001)
5.3.1 GRAPHENE GROWTH ON H-BN MONOLAYERS FORMED ON METAL SUBSTRATES
The basal plane of hexagonal boron nitride (h-BN(0001)) is in many respects an ideal dielectric surface for graphene heteroepitaxy. Isoelectronic with graphene, h-BN(0001) has an in-plane lattice constant of 2.51 Å—less than a 2% lattice mismatch to graphene. h-BN(0001) monolayers have been formed on Ni(111), Pt(111), and Pd(111) substrates by thermal decomposition of borazine (B3N3H6) [41,42]. The inertness of the initial BN surface layer to further thermal reaction and precursor decomposition results in a self-limiting monolayer growth process [42]. Such monolayers have also been formed by borazine decomposition on Ru(0001), with observation of a puckered “nanomesh” structure [43]. An R30 (√3 × √3) structure was observed (Figure 5.2a), however, for a monolayer (1 ML) BN film on Ru(0001) formed by atomic layer deposition (ALD) using two cycles of BCl3 and NH3 precursors [17]. This structure is consistent with a flat BN layer rotated 30° with respect to the Ru(0001) lattice (Figure 5.2a, inset). Bulk h-BN(0001) has a band gap of ~6 eV [44]. Scanning tunneling microscopy (STM) dI/dV data, however (Figure 5.2b), proportional to the local density of states, indicate a band gap of ~2 eV, demonstrating substantial Ru/BN orbital hybridization. Such hybridization has also been reported for epitaxial h-BN layers on certain other transition metal surfaces [41,45]. Subsequent exposure of h-BN(0001)/Ru(0001) to C2H4 at 1000 K (60 s, 0.1 Torr) yields an sp2-characteristic C(KVV) Auger feature (not shown), the low-energy electron diffraction (LEED) image shown in Figure 5.2c, and the STM dI/dV data in Figure 5.2d, demonstrating formation of a graphene overlayer.
FIGURE 5.2 Formation of graphene/BN heterojunction on Ru(0001): (a) LEED image for BN(0001) monolayer deposited by ALD. Bifurcation of outer spots due to Ru substrate is shown (arrow). Inset is the model for unit cell orientation relative to the Ru lattice. (b) STM dI/dV data indicating BN/Ru orbital hybridization. (c) LEED after CVD of graphene using C2H4. (Note satellite features attributed to multiple scattering.) (d) Corresponding STM dI/dV data indicating graphene-like density of states near the Fermi level. (From C. Bjelkevig et al., J. Phys. Cond. Matt., 22 (2010) 302002. With permission.)
The well-defined graphene/monolayer h-BN(0001)/Ru(0001) LEED image shown in Figure 5.2b contrasts with the diffuse nature of the LEED image reported [46] for graphene/monolayer h-BN(0001) heterojunction formation on Ni(111). One obvious cause for the LEED spot broadening in the latter case could be very small domain sizes. The authors suggest, however, on the basis of high-resolution electron energy loss data, that graphene interaction with BN results in the BN layer undergoing a slight relaxation, due to a weakening of BN/substrate interactions upon formation of the graphene layer [46]. In either case, the differing LEED data for graphene/mono-layer h-BN(0001)/Ru(0001) (Figure 5.2a) and for graphene/monolayer h-BN(0001)/Ni(111) demonstrate that interlayer interactions between graphene and the BN/metal substrate depend strongly on the nature of the metal, and perhaps the degree of BN/metal lattice matching.
FIGURE 5.3 Raman spectra (a) for few-layer graphene grown on BN nanoflakes by flame pyrolisis (S1–S3 refer to different samples grown by the same method) and (b) for monolayer graphene grown by CVD on monolayer BN/Ru(0001). Graphene/BN data in upper trace. Lower trace is a reference spectrum acquired from HOPG. Note the significant red-shift of the 2D feature for graphene/BN/Ru (arrow) relative to graphene/BN nanoflake. (Data in (a) adopted from T. Lin et al., J. Mater. Chem., 22 (2012) 2859. Data in (b) adopted from C. Bjelkevig et al., J. Phys. Cond. Matt., 22 (2010) 302002. With permission.)
5.3.2 GRAPHENE/BORON NITRIDE INTERFACIAL INTERACTIONS
Significant interactions occur between graphene and BN monolayers on metal substrates. The nature of such interactions can be discerned in part by a comparison of Raman spectra for graphene/monolayer-BN(0001)/Ru(0001) [17] with similar data for multilayer graphene grown on BN nanoflakes (50–250 nm) by flame pyrolysis of a glassy carbon source [47] (Figure 5.3).
The Raman spectra for graphene grown directly on a BN nanoflake (Figure 5.3a) exhibit a significant D feature, indicative of a high concentration of edge or defect sites, as expected for a small domain film grown on a nanoflake, and a 2D feature at a typical energy of 2687 cm–1 [47]. The Raman spectrum for graphene/BN/Ru(0001) (Figure 5.3b, red trace) contains normally IR-active BN modes, due to the broken symmetry of the sample, no discernable D feature, and a G/2D intensity ratio consistent with single-layer graphene [17]. The 2D mode, however (Figure 5.3b, arrow), is red-shifted by over 300 cm–1 from the usual value, as demonstrated by comparison with the 2D mode for HOPG (Figure 5.3b, lower trace). Since the 2D mode in graphene is a mixed vibrational-electronic two-phonon mode [48], such a red-shift is strongly suggestive of charge transfer into the graphene π* band, and mode softening. Thus Raman data indicate negligible substrate → graphene charge transfer for graphene grown on BN nanoflakes, and substantial substrate → graphene charge transfer for graphene/monolayer h-BN(0001)/Ru(0001).
FIGURE 5.4 C(1s) XPS data for graphene grown directly on BN nanoflakes. The peak maximum is at a bindng energy of 284.7 eV, indicative of some graphene charge transfer to the substrates. (Data adapted from T. Lin et al., J. Mater. Chem., 22 (2012) 2859. With permission.)
The absence of substantial substrate → graphene charge transfer for graphene/BN nanoflake heterojunctions is corroborated by x-ray photoelectron spectroscopy (XPS) C(1s) data (Figure 5.4). The data indicate a C(1s) peak maximum at 284.7 eV binding energy. While the figure also indicates some C-N and C=O bonding, which may occur at edge sites, a main peak binding energy of 284.7 eV is somewhat higher than the 284.5 eV commonly reported for graphite. Since shifts in core level binding energy usually correspond to changes in the valence charge density around an atom—the ground state atomic charge population [49]—a slight shift toward higher binding energy indicates some charge transfer from graphene to the BN nanoflake substrate. By comparison, multilayer sheets comprised of mixed BN and graphene domains formed by CVD exhibit a C(1s) binding energy of 284.4 eV [50], close to that of bulk graphite. These data therefore indicate that graphene grown directly on multilayer BN nanoflakes exhibits some charge transfer from the graphene to the BN substrate.
In contrast, there is pronounced charge transfer in the other direction (i.e., to the graphene film from the substrate) in the case of graphene/monolayer h-BN(0001)/Ru(0001), nearly filling the π* band [17]. This is demonstrated by the data displayed in Figure 5.5, which compare angle-integrated valence band photoemission and k-vector-resolved inverse photoemission (conduction band) spectra (Figure 5.5a) with STM dI/dV data for graphene/monolayer h-BN(0001)/Ru(0001) (Figure 5.5b and Figure 5.2d), with monolayer h-BN(0001)/Ru(0001) (Figure 5.5c and Figure 5.2b), and with corresponding inverse photoemission data for multilayer graphene on SiC(0001) (Figure 5.5e) [51]. Specifically, the feature closest to the Fermi level in the inverse photoemission data (Figure 5.5a)—that is, the lowest unoccupied level—is the σ* feature, as shown by the dispersion of this feature in k-space (Figure 5.5d). In contrast, an empty π* feature is the lowest unoccupied level for corresponding measurements on multilayer graphene/SiC(0001) [51], as well as for graphene on Ni or Cu [18]. These comparisons demonstrate that for graphene/monolayer h-BN(0001)/Ru(0001), the graphene π* band is effectively filled, consistent with Raman data (Figure 5.4b) indicating vibrational mode softening for the 2D mode. More detailed analysis [18] indicates that this charge transfer amounts to 0.07 e– per carbon atom. The dispersion of the σ* band, however, yields an estimate for the effective mass of 0.05 me [17], in good agreement with transport measurement-derived values of 0.4–0.6 me [14,52,53] on transferred graphene sheets. These data demonstrate that the charge transfer results from graphene band filling, rather than from interfacial carbon rehybridization. Therefore this charge transfer does not significantly alter the basic graphene band structure.
FIGURE 5.5 Density of states data for graphene/monolayer h-BN(0001)/Ru(0001). All data plotted relative to the Fermi level (EF): (a) Valence band-integrated photoemission (left) and k-vector-resolved inverse photoemission data (right), (b) STM dI/dV data for graphene/BN/Ru (as in Figure 5.2d), (c) corresponding data for BN/Ru (as in Figure 5.2b), (d) dispersion data for inverse photoemission feature closest to Fermi level, and (e) inverse photoemission data (i) compared to those of (ii) multilayer graphene/SiC(0001). (Data in (e)(ii) from I. Forbeaux et al., Phys. Rev. B, 58 (1998) 16396. Figure adopted from C. Bjelkevig et al., J. Phys. Cond. Matt., 22 (2010) 302002. With permission.)
The data presented in Figures 5.3,5.4,5.5 thus demonstrate a significant difference in interfacial interactions for graphene/BN(nanoflakes) vs. graphene/monolayer h-BN(0001)/Ru(0001). For graphene grown directly on a BN nanoflake, there is some graphene → substrate charge transfer. In contrast, for graphene/monolayer h-BN(0001)/Ru(0001), there is strong substrate → graphene charge transfer, almost completely filling the graphene π* band. The ability to grow multilayer epitaxial h-BN(0001) with layer-by-layer precision on a transition metal substrate should therefore present a route toward “tuning” graphene/transition metal interactions. Such precision growth is possible using an atomic layer deposition (ALD) process incorporating cycles of BCl3 and NH3, at ~550 K, which has been used to grow conformal polycrystalline BN films on a variety of substrates [54]. This process has been adapted to the layer-by-layer growth of epitaxial h-BN layers on Ru(0001), as shown by the LEED and STM data in Figure 5.6. The ability to form a h-BN(0001) trilayer on Ru(0001) (Figure 5.6; six BCl3/NH3 cycles at 550 K, followed by annealing to ~1000 K in ultrahigh vacuum (UHV)) suggests that this ALD process is a viable route toward forming varying thicknesses of BN on appropriate metal substrates for systematic variation of graphene/metal interactions. The data shown in Figures 5.3,5.4,5.5 also suggest that the chemical reactivity of a monolayer of BN on a transition metal substrate with strong BN/metal hybridization (e.g., Figure 5.2b) could differ substantially from that of a multilayer film that approaches the electronic structure of bulk h-BN. Thus a graphene film grown directly on the surface of mono-layer h-BN(0001)/Ru(0001) may well display different electronic properties than a graphene layer grown on a multilayer h-BN(0001) film. An obvious issue is whether a multilayer film will be as reactive toward C-H bond scission during CVD as the monolayer BN film on Ru(00001). Recent studies, however, suggest that processes such as flame pyrolysis [47], or perhaps MBE, will permit direct graphene growth on multilayer h-BN(0001) films.
FIGURE 5.6 Trilayer h-BN(0001) on Ru(0001) by ALD: (a) LEED of a 3 ML film produced by 6 BCl3/NH3 cycles at 550 K, followed by annealing to 1000 K in UHV (beam energy = 75 eV), and (b) STM constant current image for a BN(0001) film grown epitaxially on a Ru(0001) substrate. STM line trace (c) shows two steps with an individual step height of 3.2 Å, within experimental error of the nominal BN interlayer distance of 3.3 Å.
Finally, the data in Figure 5.2d and Figure 5.5 demonstrate that graphene/mono-layer h-BN(0001)/Ru(0001) does not exhibit an observable band gap at 300 K. This is somewhat in contrast to recent density functional theory (DFT) calculations predicting the formation of a small band gap of ~0.05 eV in graphene/BN bilayers, due to adoption of a relative orientation similar to that in Figure 5.1c [55]. Aside from well-known difficulties in the use of DFT methods to predict band gap size [56], there are several factors that could contribute to this apparent absence of an experimental graphene band gap, the most obvious being that a band gap so small would be difficult to observe by room temperature STM dI/dV or photoemission/inverse photoemission (PES/IPES) measurements. Other factors include the possible orientation of the graphene and BN layers being such as to lead to equivalent average bonding environments for graphene lattice A sites and B sites. Indeed, an analysis of graphene-related LEED intensity spots on h-BN(0001)/Ru(0001) (Figure 5.2c) revealed true C6v symmetry in the LEED pattern [19], thus indicating graphene A site/B site chemical equivalence, and implying either that the relative orientation of the graphene and BN layers differed from that predicted by the DFT calculations [55], or that the charge transfer from the Ru substrate was sufficiently great as to overwhelm and obscure effects resulting from graphene/BN interfacial interactions. In any event, the LEED, STM dI/dV, and PES/IPES data all indicate the absence of a band gap at room temperature for graphene grown directly on monolayer h-BN(0001)/Ru(0001).
5.4 GRAPHENE GROWTH ON MgO(111)
Graphene has been grown on the surface of bulk MgO(111) single crystals, using either chemical vapor deposition (thermally dissociated C2H4) [18], annealing of adventitious carbon layers in the presence of C2H4 (Figure 5.7) [19], or magnetron sputter deposition from a graphite source at room temperature, followed by annealing in UHV [26]. In all these cases, annealing in UHV to 1000 K appears to be sufficient to induce (111)-ordered sp2-hybridized C overlayers—graphene. That these overlayers are macroscopically continuous is evident from the fact that the samples can be repeatedly exposed to ambient without significant change in, e.g., LEED or other surface spectra [18]. XPS and LEED data (Figure 5.7) [19] demonstrate that formation of an ordered overlayer is accompanied by an interfacial chemical reaction that results in at least some of the carbon atoms in a higher oxidation state, as evidenced by a C(1s) feature at >288 eV binding energy (Figure 5.7d), and as observed for graphene deposited by other deposition methods [18,26]. Thus at least part of the interfacial carbon layer is strongly oxidized.
The existence of a highly oxidized component in the C(1s) XPS spectrum (Figure 5.7d) is strong evidence that the graphene layer (perhaps partially oxidized) is in contact with and reacting with an oxygen-terminated or hydroxyl-terminated MgO substrate surface. The bulk-terminated (111) layers of MgO and other oxides with the rock salt structure are composed entirely of either oxygen anions or metal cations, as shown schematically in Figure 5.8. This composition—either 100% cation (Figure 5.8 left) or 100% anion (Figure 5.8 right)—results in a net surface charge and instability in the Madelung potential, which is often stabilized by hydroxyl termination of the surface layer [57,58,59,60]. For this reason, the (111) surfaces of rock salt oxides such as MgO are prone to reconstruction upon, e.g., the adsorption of transition metals [61].
FIGURE 5.7 Graphene formation on a MgO(111) single crystal. (a) C(1s) spectrum of an adventitious C overlayer upon sample insertion into UHV. (b) After annealing at 700 K in the presence of 10–6 Torr O2. The XPS-derived average C thickness is 1 ML. (c) Corresponding LEED pattern exhibits C3v symmetry. (d) C(1s) XPS after additional annealing to 1000 K in the presence of 5 × 10–7 Torr. (e) Corresponding complex LEED pattern. (f) Close-up of the pattern in (e) with integrated, background-subtracted intensities (arbitrary units) for A and B spots (circled). The A spots have an average intensity of 18.7 ± 3, while the B spots have an average intensity of 12.9 ± 1. The uncertainties are the standard deviations. Other spots in the image are weaker and are attributed to multiple diffraction. The LEED patterns were acquired at 80 eV beam energy. The XPS spectra binding energies are referenced to a MgO lattice oxygen O(1s) binding energy of 530.0 eV. (Adopted from S. Gaddam et al., J. Phys. Cond. Matt., 23 (2011) 072204.)
FIGURE 5.8 Schematic of the bulk-terminated (111) layers of an oxide with the rock salt structure: (left) metal cation terminated; (right) oxygen anion terminated.
The data in Figure 5.7 also strongly suggest that, in addition to reaction with carbon, the oxide surface undergoes a reconstruction to yield an interface commensurate with the (partially oxidized) graphene overlayer. Specifically, the LEED data for the ordered graphene overlayer (Figure 5.7e) exhibit C3v symmetry, with the LEED spots thus exhibiting a strong (A site) and weak (B site) configuration. The O-O nearest-neighbor distance for bulk-terminated MgO is ~2.8 Å, and should therefore form an incommensurate interface with graphene (lattice constant 2.5 Å). In such an incommensurate structure, both graphene A sites and B sites will be in an ensemble of chemical environments, and one would expect the average A site and B site electron density to be the same. This would then give rise to a C6v LEED pattern and (Figure 5.1b) the absence of a band gap at the Dirac point. To the contrary, a C3v LEED pattern, as in Figure 5.7f, indicates that graphene lattice A sites have a higher electron density than the B sites—implying the formation of a commensurate interface [26] and rehybridization of the HOMO and LUMO molecular orbitals, lifting their degeneracy at the Dirac point [33], as in Figure 5.1c and d. Thus the data in Figure 5.7 indicate that graphene on MgO(111) should exhibit a band gap.
The formation of a 0.5–1 eV band gap for a 2.5 ML (average thickness) graphene film formed on MgO(111) by thermally dissociated C2H4 has been reported [18]. Valence band photoemission (angle-integrated PES) and conduction band inverse photoemission (k-vector-resolved IPES) data for this film are shown in Figure 5.9. The PES data (He I source) are in excellent agreement with those reported for high-quality graphene films grown on transition metal surfaces [62], while the IPES data exhibit the expected π* and σ* features in good agreement with IPES data for multilayer graphene/SiC(0001) [51]. Between the PES/valence band and IPES/conduction band, there is a feature-free region—a band gap. The magnitude of this band gap is ~0.5–1 eV, with the uncertainty being due to possible charging effects and the inherent limiting resolution (400 meV) in the IPES measurements [18].
FIGURE 5.9 Band gap formation for 2.5 ML graphene on MgO(111). (Left) Angle-integrated valence band photoemission (He I source): PES. (Right) K-vector-resolved inverse photoelectron data: IPES. Energies are plotted relative to the Fermi level (0 eV). (Adopted from L. Kong et al., J. Phys. Chem. C, 114 (2010) 21618. With permission.)
Recent charge transport measurements [63] for a single-layer graphene film on MgO(111) confirm semiconducting behavior, with an estimated activation energy for carrier transport of 0.64(± 0.05) eV derived from the temperature-dependent conductivity data. Although the activation energy can differ from the band gap energy by the final state exciton binding energy and other factors, an activation energy of 0.6 eV for the monolayer film is in very good agreement with the magnitude of the band gap indicated by the photoemission/inverse photoemission data (Figure 5.9) for a few-layer film. Importantly, LEED data for both films exhibit C3v symmetry [19].
5.4.3 INTERFACIAL CHARGE TRANSFER
Although substrate → graphene charge transfer occurs readily for graphene on transition metals and on monolayer h-BN(0001)/Ru(0001) [18] graphene/MgO exhibits charge transfer in the opposite direction: the graphene becomes p-type due to charge transfer to the substrate [18]. This is demonstrated by the data shown in Figure 5.10, comprising inverse photoemission spectra for graphene grown on a variety of substrates, compared to multilayer graphene/SiC(0001) [51]. Since inverse photoemission spectra reveal only the unoccupied states, the energy of a specific feature, such as the σ*, relative to the Fermi level, gives a quantitative measure of the charge transfer into the graphene conduction band from the substrate, or out of the graphene valence band to the substrate [18]. Taking the IPES data for multilayer graphene/SiC(0001) [51] as indicative of negligible graphene/substrate charge transfer [64], the relative energies of the π* and σ* features in the IPES spectrum for graphene/MgO(111) indicate charge transfer from graphene to the oxide. This finding is also consistent with XPS spectra [26,65].
In summary, the formation of single-layer graphene on MgO(111) yields an ordered graphene layer with C3v symmetry, and a band gap of ~0.5–1 eV, as determined by DC transport measurements [63]. A 2.5 ML film on MgO(111) also exhibits a band gap of about the same magnitude, as determined by photoemission/inverse photoemission [18]. Both films exhibit C3v symmetry in the LEED pattern, indicating that (1) interfacial reconstruction is probably occurring to yield a commensurate interface as well as (Figure 5.7d) a partially oxidized initial carbon layer, and (2) that such interactions are not completely screened in the second layer. A band gap of 0.5–1 eV is highly suitable for conventional FET-type logic devices [36,37]. Further, (111)-textured MgO films have been deposited on Si(100) by pulsed laser deposition [66], indicating a straightforward route toward the production of simple graphene-based FET structures integratable with Si CMOS.
FIGURE 5.10 Multilayer graphene structure on mica. (a) Optical micrograph. (b) I-V behavior. (From G. Lippert et al., Phys. Stat. Sol. B, 248 (2011) 2619. With permission.)
5.5 GRAPHENE GROWTH ON MICA BY MBE
The growth of graphene by MBE on mica has recently been reported [20,21]. The particular sample, “black mica” or biotite, has a stoichiometry of K(Mg, Fe)3AlSi3O10(OH)2 (Fe content < 5% of Mg) [21] and a structure consisting of sheets of silica/alumina layers separated by potassium cations—permitting facile cleavage exposing ordered silica layers.
Raman mapping [21] of the G/2D intensities for graphene MBE growth at 1073 K indicates that graphene thickness is nonuniform, and that the quality of the graphene layer—as judged by the relative intensity of the D peak—decreases with increasing number of graphene layers [21].
The temperature sensitivity of the growth process, and the macroscopic variability in graphene thickness, strongly suggest that the growth process involves an initial carbon interaction with the substrate to form a buffer layer, and that succeeding sp2 carbon-graphene layers grow in an island-like mode and coalesce. Ab initio growth simulations [20] indicate that growth may start at defect, or seed sites on the flat mica surface, with graphene flakes becoming immobilized when they coalesce. The studies further suggest that carbon oxidation occurs at the edge sites of the graphene flakes, which could also immobilize the separate domains. This last suggestion is certainly consistent with the finding of enhanced graphene quality with increased number of graphene layers, i.e., that graphene MBE growth on graphene occurs more readily than the growth of the initial layer on the mica.
Graphene/mica sheets grown in this way were fabricated into FETs, as shown in Figure 5.10a [20], and yielded I-V behavior characteristic of multilayer graphene [20] (Figure 5.13b). The ability to grow graphene directly on a SiO2-like substrate suggests a path toward direct graphene integration with Si CMOS. However, the apparent increase in graphene quality with increasing graphene layers indicates that growth on such substrates involves formation of an interfacial layer, with perhaps some oxidation at edge sites, followed by a more regular growth process on the initial graphene layer. In other words, the growth of graphene on graphene apparently yields a higher-quality graphene sheet than graphene on silica. This simple fact indicates that the ability to grow few-layer or multilayer graphene, instead of just a single layer, may be of significant practical importance for device fabrication. In this respect, MBE would appear to have significant advantages over CVD, as the formation of the initial graphene (or graphene-like) layer results in a surface relatively inert toward further C-H bond scission [17].
5.6 GRAPHENE GROWTH ON Co3O4(111)/Co(111) BY MBE
Direct graphene growth by MBE on Co3O4/Co(111) at 1000 K has been reported [22], and the ability to grow graphene directly on a magnetically polarizable oxide suggests new possibilities for spintronics applications. Further, graphene growth on Co3O4(111) exhibits important differences with graphene growth on MgO(111) or on mica with respect to both the nature of interfacial chemical interactions and the mode of growth. Co3O4(111) film formation occurred by the deposition, at 750 K, of Co(111) films ~40 Å thick, on Al2O3(0001), followed by annealing in UHV at 1000 K. The resulting oxygen segregation resulted in formation of a Co3O4(111) film ~3 ML thick [22]. Subsequent MBE from a solid carbon source yielded sp2-hybridized carbon at average carbon coverages from 0.4 ML up to the thickest coverage examined, 3 ML (Figure 5.11a).
Importantly, the evolution of the Auger-derived average carbon thickness (Figure 5.11b) is well fit by a series of linear segments, with changes in slope corresponding to the completion of one carbon layer and the beginning of another [22] indicative of layer-by-layer growth [67]. LEED data acquired at 0.4 ML and 3.0 ML coverages (Figure 5.12) indicate the formation of a hexagonal carbon lattice with a 2.5 Å lattice constant, thus forming an incommensurate interface with the unreconstructed oxide surface, which has an O-O distance of 2.8 Å [22].
Additional analysis of the LEED data indicates that the graphene LEED pattern at both 0.4 and 3 ML graphene thickness displays C6v symmetry [22], consistent with the formation of an incommensurate graphene/oxide interface [26], and indicating that the graphene lattice A sites and B sites are chemically equivalent at all coverages examined. Additional analysis of the full width at half maxima of LEED diffraction spots indicates an average graphene domain size of ~1800 Å, comparable to HOPG [22]. C(1s), Co(2p), and O(1s) XPS spectra acquired for the 3 ML film are displayed in Figure 5.13. The data indicate a characteristically asymmetric C(1s) feature (Figure 5.13a) with the expected π → π* satellite feature (Figure 5.13a, inset). The Co(2p) and O(1s) spectra are as expected for a 3 ML Co3O4(111) film on Co(111) [22,68]. In contrast to XPS spectra for graphene/mica (Figure 5.12a), or for graphene/MgO(111) (Figure 7d in [26]), there is no evidence of C chemical reaction with the oxide substrate. The observed C(1s) binding energy, 284.9(± 0.1) eV, however, is significantly higher than that of bulk graphite—284.5 eV. This observed binding energy is also quite close to that observed, 284.75 eV, for the first C(1s) layer on Si-terminated SiC(0001) [64]. This increase in binding energy is indicative of charge donation from the graphene layer to the oxide substrate [64,69]. Such graphene → Co3O4(111) charge transfer was also inferred from the blue-shift in the π → π* absorption feature for spectroscopic ellipsometry data, relative to that of multilayer graphene/SiC or graphene physically transferred to SiO2 [22]. Thus the XPS data in Figure 5.13a indicate significant graphene → Co3O4(111) charge transfer, as to also inferred from IPES spectra for graphene on MgO(111).
FIGURE 5.11 Auger spectra for MBE graphene growth at 1000 K on Co3O4(111)/Co(111). (a) Auger spectra at 0.4 ML (dashed trace) and 3 ML (solid trace) carbon coverage—insert shows expanded view of C(KVV) spectra; (b) evolution of average carbon overlayer thickness as a function of MBE deposition time. (From M. Zhou et al., J. Phys. Cond. Matt., 24 (2012) 072201. With permission.)
FIGURE 5.12 LEED and corresponding line scan data for (a, b) 0.4 ML graphene on Co3O4(111), and (c, d) 3 ML graphene on Co3O4(111). Arrows (a, c) mark diffraction spots associated with Co3O4(111), as do inner spots in the outer ring of bifurcated features (e.g., O1, O2—b, d). Outer spots in the outer ring of bifurcated features (e.g., G1, G2—b, d) are graphene related. LEED beam energy is 65 eV. (From M. Zhou et al., J. Phys. Cond. Matt., 24 (2012) 072201. With permission.)
Macroscopic continuity and uniformity of the graphene sheets deposited by MBE on Co3O4(111) is indicated by two pieces of evidence [22]:
FIGURE 5.13 Core level XPS spectra acquired for a 3 ML graphene film grown on Co3O4(111)/Co(111). (a) C(1s) (inset: π à π* satellite feature), (b) Co(2p), and (c) O(1s) spectra. (From M. Zhou et al., J. Phys. Cond. Matt., 24 (2012) 072201. With permission.)
1. The sample with 3 ML graphene coverage, upon exposure to ambient, showed no change in Auger or LEED spectra. A 3 ML Co3O4(111)/Co(111) film, unmodified, would be extremely sensitive to air exposure. This passivating effect of continuous graphene films has been observed for graphene on other air-sensitive substrates, including Ru(0001) [18] and Co(0001) [70].
2. Micro-Raman spectra (Figure 5.14). Micro-Raman spectra were acquired at spots separated by ~3 mm on a 1 × 1 cm sample with a 3 ML graphene film. The spectra acquired at full laser power (Figure 5.14a—upper trace, Figure 5.14b) display identical spectra, indicating macroscopic film uniformity. The observed G/2D intensities are consistent with multilayer graphene [48,71]. Importantly, however, both spectra display an intense D peak, generally associated with grain boundaries and edge sites [71]. This is inconsistent with the determination, from LEED data, of an average domain size—1800 Å—comparable to HOPG [22], which typically yields a negligible D peak (e.g., Figure 5.3b, blue trace). The explanation for this inconsistency can be found in a comparison of two spectra acquired from the same spot at full and half laser power (Figure 5.14a, red and black traces, respectively). The data indicate that the intensities of the G and 2D features increase proportionally to the laser power, but that the increase in the D feature is much greater, roughly as (laser power)2. This is indicative of an interaction between the D vibrational mode—the only mode in the spectrum with an out-of-plane vibrational component—and the dipole from the substrate oxide layer [72]. Thus the Raman data (Figure 5.14) indicate that the 3 ML graphene film is macroscopically uniform and continuous. However, the data also demonstrate that Raman spectroscopy is not a reliable metric for film quality (as determined by D peak relative intensity) on oriented polar oxide substrates.
FIGURE 5.14 Micro-Raman spectra of a 3 ML graphene film on Co3O4(111)/Co(111) acquired at two regions ~3 mm apart on a 1 cm × 1 cm sample. (a) Spectra acquired from area 1 at (top) full and (bottom) 50% laser power; (b) spectrum acquired at full laser power from area 2. (From M. Zhou et al., Phys. Cond. Matt., 24 (2012) 072201 (supplementary data online). With permission.)
In summary, the data in Figures 5.11,5.12,5.13,5.14 demonstrate that macroscopically continuous and uniform graphene sheets have been grown by MBE at 1000 K on Co3O4(111)/Co(0001). Furthermore, the growth proceeds in a layer-by-layer manner, with no evidence of carbon reaction/oxidation at edge sites, or other evidence of interfacial chemical reaction. This is consistent with growth under MBE conditions on this surface being much more rapid at edge sites than at terrace sites. In this respect, growth of graphene by MBE on Co3O4(111)/Co(0001) differs significantly from the reported results of graphene growth by the same method on silica-terminated mica [20,21], or of graphene growth on MgO(111) [18,19,26]. It would appear that the choice of oxide substrate rather than growth method determines the extent of interfacial interactions. However, the ability to grow graphene multilayers in a uniform and controlled manner, which is apparently feasible by MBE on Co3O4(111) and (albeit at >1273 K) on SiC(0001) [73,74], may have important implications for controlling and limiting such effects as graphene/substrate charge transfer [64].
5.6.2 MAGNETIC PROPERTIES AND DEVICE APPLICATIONS
A principal motivation for direct graphene growth on dielectric substrates is, of course, the development of graphene-based electronic or spintronic devices. With respect to charge-based devices, the ability to induce a significant band gap for graphene/MgO(111) [18,19] suggests direct applications to graphene FETs, particularly as (111)-textured MgO films have been deposited on Si(100) by pulsed laser deposition [66]. While there is thus a direct potential route toward formation of graphene/MgO(111) FET structures, the formation of a band gap at the Dirac point leads to a change in the conduction and valence band dispersions in the vicinity, and therefore to an increase in the effective mass and decrease in the electron mobility. For graphene/MgO(111)-based FETs to be of real value, there must be a “sweet spot” in the trade-off between increased band gap (higher on/off ratio) and decreased mobility. Semiconducting behavior has been determined by transport measurements on single-sheet graphene on MgO(111) with a band gap of ~0.5–1 eV [63], consistent with spectroscopic measurements on few-layer graphene [18]. Band gaps of this magnitude are generally suitable for conventional (e.g., nontunneling) FETs [36,37].
The ability to deposit graphene on magnetically polarizable substrates raises the possibility of developing spintronic devices based on the proximity effect [4], in which the spin polarization of graphene electrons is predicted to be induced by proximity to a ferromagnetic layer. Such applications, because of the nature of the spin transport, would be distinctly different from conventional graphene spin valves or spin FETs based on injection and diffusion of discrete spins in graphene. Originally proposed by Semenov et al. [75], the spin FET involves injection of spin-polarized electrons at a source (Figure 5.15, S) into a graphene layer, with the diffusion of such spins through a graphene layer, and transport into a drain (Figure 5.15, D). A ferromagnetic dielectric (Figure 5.15, FMD) controls spin precession. This causes a difference in conductance (C) and magnetoresistance (MR) when the source and drain are in parallel (P) or antiparallel (AP) configurations:
(5.1) |
The long spin diffusion lengths and small spin-orbit coupling in graphene make such devices attractive [3], and variants have been constructed and tested by several groups [3,76,77].
The actual performance of such devices has been unexciting, however, at least with respect toward practical device applications, as typical MR values (Equation 5.1) are ~10%, and then only at cryogenic temperatures [3,76,77]. There are several reasons for this rather disappointing result, one of which is the necessity of using spin-polarized tunneling, typically through an oxide barrier, to maximize the low efficiency of spin injection/detection [78]. A second reason is that such devices involve the diffusion of individual spins through the graphene layer, with the result that the net spin polarization, and therefore MR, decreases significantly with source/drain distance [3] and possibly also with spin scattering at impurities and grain boundaries.
A second, somewhat different type of spin FET has been proposed [32], based on graphene in intimate contact with a FM substrate, as in an intercalated graphene/FM/graphene structure (Figure 5.16, left). In such a structure, exchange interactions with the unpaired electrons on the ferromagnetic centers (Figure 5.16, left—middle spheres) induce spin polarization of the graphene conduction electrons, as previously predicted for graphene in proximity to EuO [4]. In the case of graphene/FM/graphene structures, where FM = Co, Ni, or Fe, DFT calculations predict AB stacking (graphene lattice A sites in one layer corresponding to lattice B sites in the next layer), with Fe atoms located over C-C bonds [22]. Substantial graphene conduction band polarizations (>60%) are also predicted for Co and Fe (but not for Ni) with quantum transport calculations suggesting MR values of ~200% at low source/drain bias and at room temperature (Figure 5.16, right) Notably, even at bias voltages of >0.1 V, the calculated MR at 300 K is several times better than reported values [3,76,77] for spin valves based on the tunneling injection and diffusion of discrete spins.
FIGURE 5.15 Proposed spin FET, consisting of a ferromagnetic source and drain (S, D) separated by a ferromagnetic dielectric (FMD) gate on top of a graphene channel, lying in turn on top of a graphene substrate. The external applied field from the FMD affects spin of the electron diffusing through the graphene channel, and therefore the conductance. (From Y.G. Semenov et al., Appl. Phys. Lett., 91 (2007) 153105. With permission.)
An important issue is whether the proximity effect, if it occurs, relies simply on the presence of a local magnetic mean field, as predicted [4], or whether induced magnetization is a sensitive function of C/FM orbital hybridization [32]. The graphene/Co3O4(111) interface is incommensurate [22] (Figure 5.15), indicating minimal graphene/Co orbital hybridization. The formation and testing of spin FETs on graphene/Co3O4(111)/Co(111) substrates would provide a rigorous test of these competing models, as well as perhaps open the way toward spin FETs with much enhanced magnetoresistance values at room temperatures.
The results presented in this chapter demonstrate that graphene can be deposited directly on h-BN(0001), MgO(111), mica, and Co3O4(111) by a variety of scalable, industrially practical methods, including CVD, PVD, and MBE. These data also demonstrate that the type of substrate can significantly alter the electronic structure of graphene:
FIGURE 5.16 Proposed spin FET based on (left) graphene sheets (top and bottom layers of spheres) intercalated with an atomic layer of a ferromagnetic material (middle layers of spheres). (Right) Predicted MR values vs. source/drain voltage difference. (From J. Zhou et al., J. Phys. Chem. C, 115 (2011) 25280. With permission.)
• Graphene/BN. For graphene/monolayer h-BN(0001)/Ru(0001), deposited by CVD using C2H4, there is substantial charge transfer (0.07 e–/carbon atom) from the substrate to the graphene, filling the graphene π* band, without impacting the dispersion of the σ* band. This coincides with red-shifting (by more than 300 cm–1) of the graphene 2D Raman band, indicative of mode softening. In contrast, no such red-shift is observed for graphene/BN(nanoflakes). This and C(1s) spectra indicate that in fact there is some charge transfer from graphene to the substrate nanoflakes, and suggest that the formation of epitaxial BN multilayers by ALD can systematically decrease substrate → graphene charge transfer. Although a commensurate interface is observed for graphene/monolayer h-BN(0001)/Ru(0001), LEED images display C6v symmetry, indicating the chemical equivalence of A and B sites, and consistent with the lack of a discernible band gap in STM dI/dV or PES/IPES measurements.
• Graphene/MgO(111). Single- or few-layer films of graphene have been deposited on MgO(111) by CVD (thermally predissociated C2H4), by PVD at room temperature followed by annealing to 1000 K, and by annealing a monolayer of adventitious carbon in the presence of C2H4. Both single-and few-layer films exhibit C3v LEED symmetry, indicative of interfacial reconstruction resulting in a commensurate interface, and of graphene lattice A site/B site chemical inequivalence. XPS also indicates partial oxidation of the carbon overlayer, possibly at edge sites. Consistent with the LEED data, conductivity measurements indicate a band gap of ~0.6 eV for a single-layer film, and photoemission/inverse photoemission indicates a band gap of 0.5–1 eV for a few-layer film. Inverse photoemission data also indicate some graphene → oxide charge transfer.
• Graphene/mica. Graphene multilayers have been deposited on mica at 1000 K by MBE from a solid carbon source. XPS data and growth simulations suggest initial graphene nucleation at oxide defect sites, with oxidation of carbon edge sites. The growth process is highly temperature-sensitive, with graphene growth occurring at 1073 K, but not at 473 K or 1273 K. Raman data indicate macroscopic non-uniformity of the graphene films, with higher quality graphene growth (as determined by a decrease in the Raman D peak). XPS data indicate that, in addition to graphitic carbon, smaller amounts of carbidic carbon and carbon in a partially oxidized state are present.
• Graphene/Co3O4(111). Graphene layer-by-layer growth on Co3O4(111) occurs at 1000 K, by MBE from a solid carbon source. The oxide/graphene interface is incommensurate, and consistent with this, the graphene LEED images exhibit C6v symmetry at low and high graphene coverage, indicative of the chemical equivalency of the graphene A sites and B sites. The LEED data also indicate an average domain size of ~1800 Å, comparable to HOPG. Both inertness upon exposure to ambient and micro-Raman data demonstrate film continuity and uniformity over macroscopic distances. The intensity of the graphene D peak varies in a nonlinear manner with laser power, indicating strong interaction between the out-of-plane D vibrational mode and the oxide substrate dipole, thereby suggesting that Raman spectroscopy is not a reliable fingerprint for graphene domain/edge site density on polar substrates. XPS and spectroscopic ellipsometry data indicate significant graphene → oxide charge transfer.
These results point the way toward direct graphene growth on a variety of oxides and nitride substrates, for device applications ranging from graphene-based FETs to nonlocal spin valves and magnetic tunnel junctions. Further, some of these substrates (e.g., MgO(111), Co3O4(111)) can be formed directly on Si(100), indicating routes toward direct integration with Si CMOS. What is clear, however, is that practical development of graphene-based devices must consider not only the issues of direct graphene growth, but how interfacial chemistry modifies graphene electronic and magnetic properties, and how such interactions can be systematically varied by, e.g., the use of few-layer, rather than single-layer, graphene. Such basic considerations strongly suggest that studies of graphene growth will expand to focus not only on substrate interactions, but on control of the relative orientations and interactions of graphene sheets, in the transition from single- to few-layer graphene.
This work was supported in part by the Semiconductor Research Corporation/Global Research Consortium under Task ID 2123.001 and Task ID 1770.001. Günther Lippert, Peter Dowben, Jincheng Du, and Paul Bagus are acknowledged for many stimulating conversations. Mi Zhou, Cao Yuan, Sneha Gaddam, and Frank Pasquale are gratefully acknowledged for their assistance in preparation of the manuscript.
1. A.H.C. Neto, F. Guineau, N.M.R. Peres, K.S. Noveoselov, A.K. Geim. Rev. Mod. Phys. 81 (2009) 109.
2. E.V. Castro, K.S. Novoselov, S.V. Morozov, N.M.R. Peres, J.M.B. Lopes dos Santos, J. Nilsson, F. Gineau, A.K. Geim, A.H. Castro Neto. Phys. Rev. Lett. 99 (2007) 216802.
3. N. Tombros, C. Jozsa, M. Popinciuc, H.T. Jonkman, B.J. van Wees. Nature 448 (2007) 571.
4. H. Haugen, D. Huertas-Hernando, A. Brataas. Phys. Rev. B 77 (2008) 115406.
5. K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V. Grigorieva, A.A. Firsov. Science 306 (2004) 666.
6. K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jian, M.I. Katsnelson, I.V. Grigorieva, S.V. Dubonos, A.A. Firsov. Nature 438 (2005) 197.
7. X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Velamakanni, I. Jung, E. Tutuc, S.K. Banerjee, L. Colobmo, R.S. Ruoff. Science 324 (2009) 1312.
8. A. Reina, S. Thiele, X. Jia, S. Bhaviripudi, M.S. Dresselhaus, J.A. Schaefer, J. Kong. Nano Res. 2 (2009) 509.
9. P.W. Sutter, J. Flege, E.A. Sutter. Nature Mater. 7 (2008) 406.
10. X. Liang, Z. Fu, S.Y. Chou. Nano Lett. 7 (2007) 3840.
11. C. Berger, Z. Song, T. Li, Z. Li, A.Y. Ogbazghi, R. Feng, Z. Dai, A.N. Marchenkov, E.H. Conrad, P.N. First, W.A. de Heer. J. Phys. Chem. B 108 (2004) 19912.
12. C. Berger, Z. Spong, X. Li, X. Wu, N. Brown, C. Naud, D. Mayou, T. Li, J. Hass, A. Marchenkov A.N., E.H. Conrad, P.N. First, W.A. de Heer. Science 312 (2006) 1191.
13. W.A. de Heer, C. Berger, X. Wu, P.N. First, E.H. Conrad, X. Li, T. Li, M. Sprinkle, J. Hass, M.L. Sadowski, M. Potemski, G. Martinez. Sol. St. Commun. 143 (2007) 92.
14. Y. Zhang, V.W. Brar, C. Girit, A. Zettl, M.F. Crommie. Nature Physics 5 (2009) 722.
15. F. Chen, J. Xia, D.K. Ferry, N. Tao. Nano Lett. 9 (2009) 2571.
16. Y.-M. Lin, C. Dimitrakopoulos, K.A. Jenkins, D.B. Farmer, H.-Y. Chiu, A. Grill, P. Avouris. Science 327 (2010) 662.
17. C. Bjelkevig, Z. Mi, J. Xiao, P.A. Dowben, L. Wang, W. Mei, J.A. Kelber. J. Phys. Cond. Matt. 22 (2010) 302002.
18. L. Kong, C. Bjelkevig, S. Gaddam, M. Zhou, Y.H. Lee, G.H. Han, H.K. Jeong, N. Wu, Z. Zhang, J. Xiao, P.A. Dowben, J.A. Kelber. J. Phys. Chem. C 114 (2010) 21618.
19. S. Gaddam, C. Bjelkevig, S. Ge, K. Fukutani, P.A. Dowben, J.A. Kelber. J. Phys. Cond. Matt. 23 (2011) 072204.
20. G. Lippert, J. Dabrowski, M. Lemme, C. Marcus, O. Seifarth, G. Lupina. Phys. Stat. Sol. B 248 (2011) 2619.
21. O. Seifarth, G. Lippert, J. Dabrowski, G. Lupina, W. Mehr. Proceedings of IEEE Conference on Semiconductors (Dresden, September 27–28, 2011), p. 1.
22. M. Zhou, F.L. Pasquale, P.A. Dowben, A. Boosalis, M. Schubert, V. Darakchieva, R. Yakimova, J.A. Kelber. J. Phys. Cond. Matt. 24 (2012) 072201.
23. M.A. Fanton, J.A. Robinson, B.E. Weiland, M. LaBella, K. Trumbell, R. Kasarda, C. Howsare, M. Hollander, D.W. Snyder. Abstract for the Graphene 2011 Conference (Bilbao, Spain, 2011).
24. R.S. Shishir, D.K. Ferry. J. Phys. Cond. Matt. 21 (2009) 232204.
25. C.R. Dean, A.F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, K.L. Shepard, J. Hone. Nature Nanotechnol. 5 (2010) 722.
26. J.A. Kelber, S. Gaddam, C. Vamala, S. Eswaran, P.A. Dowben. Proc. SPIE 8100 (2011) 81000Y-1
27. J. Hass, F. Varchon, J.E. Millan-Otoya, M. Sprinkle, N. Sharma, W.A. de Heer, C. Berger, P.N. First, L. Magaud, E.H. Conrad. Phys. Rev. Lett. 100 (2008) 125504.
28. Z. Peng, Z. Yan, Z. Sun, J.M. Tour. ACS Nano 5 (2011) 8241.
29. E. Kim, H. An, H. Jang, W. Cho, N. Lee, W. Lee, J. Jung. Chem. Vap. Dep. 17 (2011) 9.
30. W. Gannett, W. Regan, K. Watanabe, T. Taniguchi, M.F. Crommie, A. Zettl. Appl. Phys. Lett. 98 (2011) 242105.
31. A. Reina, X. Jia, J. Ho, D. Nezich, H. Son, V. Bulovic, M.S. Dresselhaus, J. Kong. Nano Lett. 9 (2009) 30.
32. J. Zhou, L. Wang, R. Qin, J. Zheng, W.N. Mei, P.A. Dowben, S. Nagase, Z. Gao, J. Lu. J. Phys. Chem. C 115 (2011) 25280.
33. P.A. Cox. The Electronic Structure and Chemistry of Solids. Oxford University Press, Oxford, 1987.
34. X. Li, X. Wu, M. Sprinkle, F. Ming, M. Ruan, Y. Hu, C. Berger, W.A. de Heer. Phys. Stat. Sol.i A 207 (2010) 286.
35. S.Y. Zhou, G.-H. Gweon, A.V. Federov, P.N. First, W.A. de Heer, D.-H. Lee, F. Guinea, A.H. Castro Neto, A. Lanzara. Nature Materials 6 (2007) 770.
36. M. Cheli, P. Michetti, G. Iannaccone. IEEE Trans. Electron Dev. 57 (2010) 1936.
37. P. Michetti, M. Cheli, G. Iannaccone. Appl. Phys. Lett. 96 (2010) 133508.
38. Y. Zhang, T. Tang, C. Cirit, Z. Hao, M.C. Martin, A. Zettl, M.F. Crommie, R.Y. Shen, F. Wang. Nature 459 (2009) 820.
39. M.Y. Han, B. Ozyilmaz, Y. Zhang, P. Kim. Phys. Rev. Lett. 98 (2007) 206805.
40. Y. Son, M.L. Cohen, S.G. Louie. Phys. Rev. Lett. 97 (1997) 216803.
41. A. Nagashima, N. Tejima, Y. Gamou, T. Kawai, C. Oshima. Phys. Rev. Lett. 75 (1995) 3918.
42. C. Oshima, A. Nagashima. J. Phys. Cond. Matt. 9 (1997) 1.
43. A. Goriachko, Y. He, M. Knapp, H. Over, M. Corso, T. Brugger, S. Berner, J. Osterwalder, T. Greber. Langmuir 23 (2007) 2928.
44. K. Watanabe, T. Taniguchi, H. Kanda. Nature Mater. 3 (2004) 404.
45. A.B. Preobrajenski, A.S. Vinogradov, N. Martensson. Surf. Sci. 582 (2005) 21.
46. C. Oshima, A. Itoh, E. Rokuta, T. Tanaka, K. Yamashita, T. Sakuri. Sol. St. Commun. 116 (2000) 37.
47. T. Lin, Y. Wang, H. Bi, D. Wan, F. Huang, X. Xie, M. Jiang. J. Mater. Chem. 22 (2012) 2859.
48. A.C. Ferrari, J.C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K.S. Novoselov, S. Roth, A.K. Geim. Phys. Rev. Lett. 97 (2006) 187401.
49. U. Gelius. Phys. Scripta 9 (1974) 133.
50. L. Ci, L. Song, C. Jin, D. Jariwala, D. Wu, Y. Li, A. Srivastava, Z.F. Wang, K. Storr, L. Balicas, F. Liu, P.M. Ajayan. Nature Mater. 9 (2010) 430.
51. I. Forbeaux, J.-M. Themlin, J.-M. Debever. Phys. Rev. B 58 (1998) 16396.
52. X. Hong, A. Posadas, K. Zou, C.H. Ahn, J. Zhu. Phys. Rev. Lett. 102 (2009) 136808.
53. Y. Zhang, Y. Tan, H.L. Stormer, P. Kim. Nature 438 (2005) 201.
54. J.D. Ferguson, A.W. Weimar, S.M. George. Thin Sol. Films 413 (2002) 16.
55. G. Giovannetti, P.A. Khomyakov, G. Brocks, P.J. Kelly, J. van den Brink. Phys. Rev. B 76 (2007) 073103.
56. B. Lee, L. Wang, C.D. Spartary, S.G. Louie. Phys. Rev. B 76 (2007) 245114.
57. F. Rohr, K. Wirth, J. Libuda, D. Cappus, M. Baumer, H.-J. Freund. Surf. Sci. 315 (1994) L977.
58. D. Cappus, C. Xu, D. Ehrlich, B. Dillmann, C.A. Ventrice Jr., K. Al Shamery, H. Kuhlenbeck, H.-J. Freund. Chem. Phys. 177 (1993) 533.
59. M. Hassel, H.-J. Freund. Surf. Sci. 325 (1995) 163.
60. V.K. Lazarov, R. Plass, H.-C. Poon, D.K. Saldin, M. Weinert, S.A. Chambers, M. Gajdardziska-Josifovska. Phys. Rev. B 71 (2005) 115434.
61. J. Goniakowski, C. Noguera. Phys. Rev. B 66 (2002) 085417.
62. Y.S. Dedkov, M. Fonin, U. Rudiger, C. Laubschat. Phys. Rev. Lett. 100 (2008) 107602.
63. J.A. Kelber, M. Zhou, S. Gaddam, F. Pasquale, L. Kong, P.A. Dowben. Proc. ECS (2012) in press.
64. K.V. Emtsev, F. Speck, T. Seyller, L. Ley. Phys. Rev. B. 77 (2008) 155303.
65. K.V. Emtsev, A. Bostwick, K. Horn, J. Jobst, G.L. Kellogg, L. Ley, J.L. McChesney, T. Ohta, S.A. Reshanov, J. Rohrl, E. Rotenberg, A.K. Schmid, D. Waldmann, H.B. Weber, T. Seyller. Nature Mater. 8 (2009) 203.
66. X.Y. Chen, K.H. Wong, C.L. Mak, X.B. Yin, M. Wang, J.M. Liu, Z.G. Liu. J. Appl. Phys. 91 (2002) 5728.
67. C. Argile, G.E. Rhead. Surf. Sci. Rep. 10 (1989) 277.
68. S.C. Petito, M.A. Langell. J. Vac. Sci. Technol. A 22 (2004) 1690.
69. Briggs and M. P. Seah (eds.), Practical Surface Analysis 2nd Edition, Vol. I: Auger and Photoelectron Spectroscopy, Wiley Interscience, NY (1990) 657 pages.
70. H. Ago, Y. Ito, N. Mizuta, K. Yoshida, B. Hu, C.M. Orofeo, M. Tsuji, K. Ikeda, S. Mizuno. ACS Nano 4 (2010) 7407.
71. Z. Ni, Y. Wang, T. Yu, Z. Shen. Nano Res. 1 (2008) 273.
72. S. Franzen. J. Phys. Chem. C 113 (2009) 5912.
73. J. Park, W.C. Mitchel, L. Grazulis, H.E. Smith, K.G. Eyink, J.J. Boeckl, D.H. Tomich, S.D. Pacley, J.E. Hoelscher. Adv. Mater. 22 (2010) 4140.
74. E. Moreau, F.J. Ferrer, D. Vignaud, S. Godey, X. Wallart. Phys. Stat. Sol. A 207 (2010) 300.
75. Y.G. Semenov, K.W. Kim, J.M. Zavada. Appl. Phys. Lett. 91 (2007) 153105.
76. S. Cho, Y. Chen, M.S. Fuhrer. Appl. Phys. Lett. 91 (2007) 123105.
77. W. Han, K. Pi, H. Wang, M. McCreary, Y. Li, W. Bao, P. Wei, J. Shi, C.N. Laun, R.K. Kawakami. Proc. SPIE 7398 (2009) 739819-1.
78. B. Dlubak, P. Seneor, A. Anane, C. Barraud, C. Deranlot, D. Deneuve, B. Servet, R. Mattana, F. Petroff, A. Fert. Appl. Phys. Lett. 97 (2010) 092502.