Graphene nanoribbons (GNRs) hold great promise for future electronics because of their edge and width dependent electronic bandgaps and exceptional transport properties. While significant progress toward GNR devices has been made, the field has been limited by difficulties achieving narrow widths, global alignment, and atomically pristine GNR edges on technologically relevant substrates. A recent advance has challenged these limits by using Ge(001) substrates to direct the bottom-up growth of GNRs with nearly pristine armchair edges and widths near ∼10 nm via atmospheric pressure chemical vapor deposition. In this work, the growth of GNRs on Ge(001) is extended to ultra-high vacuum conditions, resulting in the realization of GNRs with widths narrower than 5 nm. Armchair graphene nanoribbons oriented along Ge 〈110〉 surface directions are achieved with excellent width control and relatively large bandgaps. The bandgap magnitude and electronic uniformity of these sub-5 nm GNRs are well-suited for emerging nanoelectronic applications.
The superlative electronic transport properties of graphene are derived from its conical bandstructure at low energies, yielding massless Dirac charge carriers.1,2 This bandstructure makes graphene a zero-bandgap semiconductor with effectively metallic properties at finite temperatures. While many characteristics of the quasiparticle charge carriers in graphene are attractive for electronic devices (e.g., large Fermi velocity and minimized electronic backscattering3), the gapless bandstructure is a significant hindrance since it limits the on/off ratio in transistors to a level far short of the requirements for digital electronics.4,5 However, lateral nanoscale confinement of graphene in the form of a nanoribbon can open an electronic bandgap.6 The largest bandgaps are expected when the nanoribbon long-axis coincides with an armchair direction in graphene (i.e., parallel to C-C bonds), in which case the bandgap will generally increase with decreasing width.6 Experimental measurements of the electronic properties of laterally confined graphene nanostructures have confirmed these theoretical predictions,7–10 motivating efforts to develop scalable methods of producing structurally and electronically homogeneous graphene nanoribbons (GNRs).
Previous attempts to fabricate GNRs have generally followed one of two paths: lithographic definition from a top-down approach or self-assembly from a bottom-up approach. Both strategies have struggled to concurrently achieve uniform placement, orientation, and atomically pristine edges. In the case of top-down lithography, the placement and orientation of GNRs can be defined with respect to a global reference,11,12 but the etching required to fabricate GNRs results in significant edge disorder and roughness that ultimately limits the resulting electronic properties.13,14 On the other hand, bottom-up assembly has achieved atomically pristine edges by using aromatic, graphene-like molecules to build GNRs piece-by-piece,15,16 but at the expense of global alignment.17 Bottom-up assembly approaches have also been limited to metallic substrates, which complicate efforts to measure and utilize the self-assembled GNRs in electronic devices.
Recent work by Jacobberger et al. has made significant strides toward extending the bottom-up synthesis of GNRs to include directional control on non-metallic substrates.18 In this case, an epitaxial relationship between graphene and the Ge(001) growth substrate drives oriented GNR growth.18 Using similar chemical vapor deposition (CVD) protocols as graphene growth on copper foils,19,20 oriented GNRs have been grown on semiconducting Ge(001) wafers with excellent control over the GNR widths (w) down to the 10 nm scale. Despite this significant advance, further reduction in GNR width down to the 3–5 nm level is required for high-performance digital electronic applications.
Here, we realize controlled, sub-5 nm widths by growing GNRs on atomically pristine 2 × 1 reconstructed Ge(001) surfaces in ultra-high vacuum (UHV). These narrow, oriented GNRs exhibit bandgaps approaching 1 eV, which is comparable to the bandgap of crystalline silicon that is the current workhorse for conventional digital electronics. The resulting GNRs also exhibit ordered armchair edges with roughness generally limited to the atomic scale over the entire GNR length. The controlled growth of oriented GNRs with sub-5 nm widths on semiconducting Ge(001) substrates constitutes a significant advance toward the development of graphene-based digital electronics.
GNRs were grown on atomically pristine Ge(001) surfaces following exposure to ethylene at 800–900 °C in UHV. The Ge(001) crystals were first cleaned by Ar+ ion bombardment and annealing to achieve atomically pristine 2 × 1 reconstructed surfaces (unit cell highlighted in green in Fig. 1(a)). Figure 1(b) shows a large-scale scanning tunneling microscope (STM) image of multiple GNRs aligned with the 〈110〉 directions on the 2 × 1 reconstructed Ge(001) surface. The fast-Fourier transform (FFT) shown in the inset highlights the 〈110〉 directions, coincident with the dimer row directions, and is translated to real-space directions in the top corner of Fig. 1(b). In large-scale STM images, only the directions of the nanoribbon edges can be confirmed to align with the 〈110〉 Ge directions, schematically shown in Fig. 1(a). The GNR edge directionality is more precisely shown to be within 0.6° of the 〈110〉 Ge directions through the atomically resolved STM image in Fig. 1(c), where the [110] direction of the Ge(001) surface is highlighted in light gray and the edges of the graphene nanoribbon are accented by the darker gray lines. This directional alignment deviates slightly from the CVD growth method, where a 3° offset was observed.18 This discrepancy likely results from differences in the growth conditions such as the absence of H2, the lower pressure at the growth surface, the difference in precursor, and the lower growth temperature in UHV compared to CVD.
(a) Schematic diagram of a 2 × 1 reconstructed Ge(001) surface with a GNR (gray) atomic lattice overlaid with the armchair direction parallel to the [110] direction. (b) Large-scale (160 nm × 160 nm) STM image of aligned graphene nanoribbons. Inset: FFT highlighting dimer-row periodicity and providing reference for 〈110〉 Ge directions (V = 1 V, I = 400 pA, scale bar = 20 nm). (c) Atomically resolved GNR (edges shown in dark gray) aligned with the [110] direction of Ge (highlighted with light gray lines) (V = 0.2 V, I = 200 pA, scale bar = 4 nm). (d) Line profile from (c) taken along the dashed blue line showing atomic steps in the underlying Ge(001) substrate (gray horizontal lines in (d)) and GNR cross-section.
(a) Schematic diagram of a 2 × 1 reconstructed Ge(001) surface with a GNR (gray) atomic lattice overlaid with the armchair direction parallel to the [110] direction. (b) Large-scale (160 nm × 160 nm) STM image of aligned graphene nanoribbons. Inset: FFT highlighting dimer-row periodicity and providing reference for 〈110〉 Ge directions (V = 1 V, I = 400 pA, scale bar = 20 nm). (c) Atomically resolved GNR (edges shown in dark gray) aligned with the [110] direction of Ge (highlighted with light gray lines) (V = 0.2 V, I = 200 pA, scale bar = 4 nm). (d) Line profile from (c) taken along the dashed blue line showing atomic steps in the underlying Ge(001) substrate (gray horizontal lines in (d)) and GNR cross-section.
The GNRs appear as elongated elevated protrusions because of the significant Ge evaporation at the growth temperature and the fact that the GNR actively suppresses evaporation while the surrounding material is removed. The atomic steps surrounding the GNR are clearly visible at the theoretically predicted a0/4 step height indicated by the horizontal gray lines in Fig. 1(d). A close examination of the STM image reveals that the graphene rests on one Ge terrace, making the observed spacing between the Ge and GNR equal to ∼3.6 Å, in reasonable agreement with the out-of-plane spacing of 3.3 Å in graphite. Furthermore, the underlying Ge terrace has a dimer row reconstruction aligned with the long axis direction of the GNR. The precise, global alignment of the GNR long axis direction with the dimer row reconstructions of the Ge(001) surface points to a surface-assisted anisotropic growth motif. The approximate reproduction of this feature from CVD growths18 reveals that the anisotropic crystal growth phenomena that drive GNR formation on Ge(001) occur at various pressures, regardless of the presence of hydrogen.
Figure 1(b) also demonstrates the uniformity of the resulting GNR widths, confined here to w < 5 nm. To further quantify the distribution of GNR widths, 300 distinct GNRs were measured to statistically characterize the width distribution. The histogram of the resulting widths is shown in Fig. 2. Fit to a log-normal distribution due to the asymmetric positive tail, the distribution is centered at w = 3.40 nm with a variance of 1.65 nm. This corresponds to 82.3% of the studied nanoribbons having w < 5 nm, while 97.8% of the observed nanoribbons fall under w = 8 nm. These GNR widths are significantly narrower than the results previously achieved with CVD on Ge(001).
Histogram of GNR widths resulting from UHV growth conditions. The orange line shows the log-normal fit used to extract mean and variance parameters.
Histogram of GNR widths resulting from UHV growth conditions. The orange line shows the log-normal fit used to extract mean and variance parameters.
To confirm the semiconducting properties of the GNRs, scanning tunneling spectroscopy (STS) measurements were performed to probe the GNR density of states (DOS). A representative STS curve for the w = 2.7 nm (N = 11 hexagonal unit cells) GNR in Fig. 3(b) is shown in Fig. 3(a). The curve shows semiconducting characteristics including zero differential tunneling conductance at the Fermi energy (Ef) and sharp valence/conduction band edges on either side of Ef. This picture contrasts starkly with both the metallic Ge(001) reference (gray line in Fig. 3(a))21,22 and full graphene monolayers on Ge grown with a similar method.23 In that case, the graphene DOS remains non-zero throughout a range of energies near Ef and exhibits no flat band behavior in that region.23 As previously discussed,18 quantitatively calculating the DOS bandgap on the Ge(001) growth substrate remains difficult due to the presence of ionized vacancies within the Ge(001) bulk and the metallic Ge(001) surface state that presents electronic states within the GNR and bulk Ge bandgaps. Due to the negligible vertical thickness of the GNRs, these available states can contribute to the dI/dV signal within the bandgap (Eg) of the GNR.24,25 Consequently, we conservatively estimate Eg > 0.8 eV for the GNR in Fig. 3(b). Allowing for small width variation, the average prediction for N = 10, 11, and 12 using GW corrected DFT is 0.95 eV—in good agreement with the measurement.6
(a) STS point spectra of (blue) GNR in (b) and (gray) reconstructed Ge(001) surface. (b) STM image of w = 2.7 nm GNR (V = −2 V, I = 400 pA, scale bar = 2 nm). (c) STS line spectra taken along length of GNR in (b) with colorbar indicating dI/dV intensity in arbitrary units.
(a) STS point spectra of (blue) GNR in (b) and (gray) reconstructed Ge(001) surface. (b) STM image of w = 2.7 nm GNR (V = −2 V, I = 400 pA, scale bar = 2 nm). (c) STS line spectra taken along length of GNR in (b) with colorbar indicating dI/dV intensity in arbitrary units.
To further probe the GNR electronic characteristics, we collected dI/dV point spectra along the length of the GNR near the center (Fig. 3(c)) to probe for spatial variations. Interestingly, the data indicate that while spatial variations exist within the valence and conduction bands, the resulting bandgap is nearly unaffected. Despite atomic-scale line-edge roughness, the band edges show relatively small energetic fluctuations along the GNR length. These results contrast observations of atomically precise bandstructure transitions in bottom-up architectures.26,27 The primary localized bandstructure modifications are the atomic scale point modification near the middle of the GNR (indicated by the gray arrow), which might be associated with the topographic protrusion in the GNR, either related to a particularly strong scattering center or an underlying surface defect in the Ge surface. The other bandstructure deviation in Fig. 3(c) is the strong DOS increase in the conduction band near the end of the GNR (gray oval), which could be attributed to a localized end state.28 Overall, these results show that despite minor atomic scale deviations from pristine armchair edges, the w < 5 nm GNRs demonstrate spatially invariant bandgaps along their entire length.
Finally, the observed GNRs exhibit rather unique atomic-scale structure that is attributed to quasiparticle interference patterns from quantum confinement. The six-fold symmetry of the graphene Fermi surface confines the available scattering vectors allowed for these carriers and thus modifies the observed patterns in STM. An atomically resolved image of a w = 2.9 nm GNR is shown in Fig. 4(a) exhibiting both six- and two-fold scattering patterns attributed to point and line defects, respectively. Fig. 4(b) shows a smaller portion of the GNR near the end exhibiting characteristic √3 and λf (dashed gray lines) scattering patterns. Due to the unique confinement of charge density to C-C bonds,29 the scattering patterns can be directly used to identify the location of the graphene lattice, shown in gray at the bottom of Fig. 4(b). This analysis confirms the GNR edges have a pristine armchair configuration. The quasiparticle scattering is confirmed in the FFT (Fig. 4(c)) where periodicity is observed at all 6 K/K′ points in reciprocal space. Similar to alternative synthetically derived GNRs, the scattering patterns in the graphene nanoribbons do not exhibit one-dimensional uninterrupted charge density along their entire length,30 possibly accounting for their overall semiconducting behavior. At w < 5 nm, the nominal hexagonal graphene lattice is not observed, and all atomic-scale features are entirely derived from the confluence of these scattering behaviors, thus indicating the transition to confinement-derived electronic properties.
(a) Atomically resolved GNR (V = −0.1 V, I = 200 pA, scale bar = 1 nm). (b) Closer look at top boxed region in (a) showing √3 and λf scattering similar to defects or boundaries in graphene (scale bar = 0.5 nm). The graphene lattice is overlaid using the quasiparticle scattering patterns to determine the location of the lattice. (c) FFT of (b) with K/K′ points highlighted indicating strong presence of intervalley scattering (scale bar = 2 nm−1). (d) Atomic scale image from bottom boxed region in (a) showing curved symmetry (scale bar = 0.5 nm). (e) FFT of (d) illustrating no periodicity at K/K′ points (scale bar = 2 nm−1).
(a) Atomically resolved GNR (V = −0.1 V, I = 200 pA, scale bar = 1 nm). (b) Closer look at top boxed region in (a) showing √3 and λf scattering similar to defects or boundaries in graphene (scale bar = 0.5 nm). The graphene lattice is overlaid using the quasiparticle scattering patterns to determine the location of the lattice. (c) FFT of (b) with K/K′ points highlighted indicating strong presence of intervalley scattering (scale bar = 2 nm−1). (d) Atomic scale image from bottom boxed region in (a) showing curved symmetry (scale bar = 0.5 nm). (e) FFT of (d) illustrating no periodicity at K/K′ points (scale bar = 2 nm−1).
Examining the termini of the GNRs, we find often one end exhibits a distinct geometry, as shown in Fig. 4(d). Rather than bearing a relation to the overall GNR edges, this alternative geometry is nearly circular and exhibits a starkly contrasted scattering profile. Despite similar periodicity, the scattering in Fig. 4(d) is isotropic (see FFT in Fig. 4(e)), rather than directionally defined. This lack of directionality eliminates the possibility that the patterns are directly related to Dirac quasiparticle intervalley scattering in graphene, as no periodicity is observed at the K/K′ points in Fig. 4(e). Rather, this pattern is reminiscent of C60 and likely related to physical curvature at the GNR edge. This spherical symmetry is a common characteristic to many of the GNRs grown here and could be related to the nucleation and growth of the GNRs. This curvature might also arise from bonding between the short edges of the GNR and the underlying Ge substrate.
In conclusion, we have utilized the catalytic properties of the anisotropic Ge(001) surface to grow armchair oriented GNRs with w < 5 nm in UHV. These GNRs are globally aligned with the 〈110〉 directions of the Ge(001)–2 × 1 surface and have long axis directions parallel to the in-plane dimer row directions. Nearly 83% of the GNRs possessed widths below the 5 nm threshold and greater than 97% were less than 8 nm wide. The GNRs have technologically relevant bandgaps consistent with theoretical predictions, which vary minimally along their length. These advances in GNR physical and electronic structure homogeneity represent an important step toward the long-standing goal of graphene-based digital electronics.
This work was performed, in part, at the Center for Nanoscale Materials, a U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences User Facility under Contract No. DE-AC02-06CH11357. This work was supported by the U.S. Department of Energy SISGR Contract No. DE-FG02-09ER16109, the Office of Naval Research (Grant No. N00014-14-1-0669), and the National Science Foundation Graduate Fellowship (DGE-1324585 and DGE-0824162). R.M.J. and M.S.A. acknowledge support from the Department of Energy (DOE) Office of Science Early Career Research Program through the Office of Basic Energy Sciences (No. DE-SC0006414) for graphene synthesis, and R.M.J. also acknowledges support from the Department of Defense (DOD) Air Force Office of Scientific Research through the National Defense Science and Engineering Graduate Fellowship (No. 32 CFR 168a).