Simona
Achilli
ab,
Gian Franco
Tantardini
ab and
Rocco
Martinazzo
*ab
aDipartimento di Chimica, Università degli Studi di Milano, v. Golgi 19, 20133, Milano, Italy. E-mail: rocco.martinazzo@unimi.it; Fax: +39 02 50314300; Tel: +39 02 50314287
bIstituto di Scienze e Tecnologie Molecolari, Consiglio Nazionale delle Ricerche, Milano, Italy
First published on 27th June 2014
The electronic and transport properties of graphene ribbons sandwiched between hydrogen dimer lines, of the kind recently realized by Nilsson et al., Carbon, 2012, 50, 2052, are investigated with the help of first principles methods. It is found that such lines of hydrogen atoms block conduction between neighboring channels and effectively allow the confinement of graphene charge carriers, thereby opening the possibility of imprinting nano-circuits in graphene by controlled hydrogenation.
Well-known examples of graphene nanostructures are graphene nanoribbons (GNRs)6,7 where, analogously to carbon nanotubes,8,9 electron confinement in the direction perpendicular to the ribbon axis opens a band-gap in their electronic structure and makes them suitable for logic applications, with potentially outstanding performances.10 Current limitations to the practical exploitation of GNRs arise from the lack of control over the edge geometry, though recent progress in controlled hydrogen etching is promising in this respect.11,12 Graphene nanoribbons with desired edges have indeed been produced by a chemical synthetic route on a metallic surface,13,14 and recently proved to be transferable onto an insulating substrate,15 but practical ways to realize bottom-up GNR transistors remain challenging to develop. Likewise if atomic-scale defects (e.g. carbon atom vacancies, voids and chemisorbed atoms) are introduced in the lattice: whether precise control of the defect positions was achieved, regularly arranged superlattice structures would display highly desirable properties – including a band-gap16,17 – thanks to the huge impact that such kinds of defects have on the low energy electronic structure of graphene.18–20
Recently, it has been predicted that parallel lines of chemisorbed hydrogen atoms split a graphene sheet into pseudo-ribbons with semimetal or semiconducting properties, similar to conventional nanoribbons.21–24 Analogues of armchair graphene nanoribbons (aGNRs) were conceived by placing lines of hydrogen atoms along an armchair direction in graphene, and were found to have a band-gap of a size similar to that of an aGNR with the same width.23 Electron waveguides that could act as two or three terminal junctions and serve as nano-diodes, nano-transistors or logic network units were also envisaged with lines of hydrogen atoms playing the role of confining walls.24 Along the same line of thought, Rasmussen et al.25 have recently shown how to stabilize single and multiple kinks in graphene by placing hydrogen lines along an armchair direction, thereby forming pseudo-ribbons with transport properties similar to those of conventional aGNRs. Similarly, Gunlycke et al.26 investigated chemical decoration of a common extended line defect (the 5-5-8 topological defect line), and showed that such nano-structuring creates confinement and resonance bands which trace the bands in zig-zag nanoribbons and display remarkable valley polarization properties. Importantly, in such approaches, defects (chemisorbed atoms) are used to imprint ribbons on graphene rather than cutting them out of it, thereby allowing the use of one and the same substrate as a support for integrated (nano)circuits.
In practice, the possibility of realizing the above structures by simply adsorbing hydrogen atoms on graphene is hampered by the unavoidable randomness of the adsorption process. Though well defined dimers (the so-called ortho- and para-dimers) do preferentially form when exposing graphene (graphite) to a hot hydrogen beam,27 the regioselectivity of the adsorption process strongly depends on the number and position of the already adsorbed species, as a consequence of the aromatic character of the substrate.28 As a result, a random distribution of dimers is usually found on the surface, which turns into clustering at high hydrogen doses,27,29 because the substrate softening accompanying hydrogenations causes a (local) enhancement of the reactivity of the surface. Eventually, disordered hydrogenated graphene is achieved, possibly with local graphane-like domains,30 and a metal–insulator transition is already triggered at a modest hydrogen coverage.31
Progress in controlling hydrogenation has been made by exploiting superlattice (Moiré) structures, which arise naturally when graphene grows on a crystalline substrate. Indeed, beside inducing superlattice potentials for graphene charge carriers,32 substrates may act as templates for patterned adsorption even when graphene binds only weakly to them. This is the case, for instance, for graphene on Ir(111) where hydrogenated structures were formed which followed the superlattice periodicity, and were shown to open a gap in the electronic band structure.33
More recently, it has been shown that pairs of hydrogen lines of the type mentioned above do indeed form on graphite upon hydrogen exposure if the surface is pre-covered with a self-assembled monolayer of cyanuric acid.34 Such structures were investigated by combining Scanning Tunneling Microscopy (STM) with Density Functional Theory (DFT) calculations and were found to be highly stable, thanks to the balanced mixture of ortho- and para-dimer configurations which naturally appears along an armchair direction. One specific pair of such hydrogen-dimer lines was identified by its STM image and appeared as bright lines enclosing a dark region, thereby suggesting a reduction of the conductivity between the defect lines.34
In view of the above, in this paper we study in detail the electronic and transport properties of the structures produced by Nilsson et al.34 and similar structures of different sizes. In particular, we investigate whether electronic confinement does indeed occur and how it affects electronic transport. To this end, we consider transport in a direction either parallel or perpendicular to the dimer lines (in the disorder-free ballistic regime where the geometry of the transport channels alone determines the conduction properties), and characterize its spatial features by means of transport eigenchannels. Our results confirm that hydrogen-dimer lines indeed act as impenetrable walls for graphene charge carriers and split the graphene sheet into independent transport channels that can be used for realizing nano-transistors and electron waveguides in graphene.
The paper is organized as follows. Section 2 gives some details concerning the chosen models and the adopted theoretical approach, Sections 3.1–3.4 describe, respectively, the electronic structure and the transport properties of the systems considered, and finally Section 4 summarizes and concludes.
Transport properties were computed with standard non-equilibrium Green's function (NEGF) techniques that use the corresponding Kohn–Sham one-electron Hamiltonian, HKS. In such a set-up three different regions are identified, two semi-infinite leads and a central scattering region. Transport properties require calculation of the retarded Green's function of the scattering region, upon proper inclusion of the electrodes self-energies in the scattering-region Hamiltonian. The latter are obtained separately by standard DFT periodic calculations for the semi-infinite electrodes alone, and the Hamiltonian (Green's function) is made self-consistent with the charge density it describes, for each applied bias, as implemented in TRANSIESTA.41
The zero-temperature, linear-regime conductance G is given (in units of quantum of conductance G0 = 2e2/h = 77.48 μS appropriate for spin-unpolarized channels) by the cumulative transmission probability N(ε)42
N(ε) = Tr[G†(ε)ΓR(ε)G(ε)ΓL(ε)] |
More generally, the current flowing from L to R follows from
Spatial features of conduction modes were analyzed with transmission (left) eigenchannels.43 These are eigenvectors of the product current operator in the device subspace of the left-incoming scattering states, and describe independent transport channels with transmission probabilities tm(ε) given by the corresponding eigenvalues . They were computed with the above G(ε) and ΓL(R), as described in ref. 43 and 44, by means of INELASTICA.45
![]() | ||
Fig. 1 Fully relaxed 6-SL (top) and (6,11)-SL (bottom) structures along with their unit cells projected on the surface plane. The inset is a blow-up of the hydrogen-dimer line. |
All the structures considered were fully relaxed until the force acting on each cartesian coordinate was below 0.01 eV Å−1, using a sufficiently large mesh cutoff for the real-space integration (500 Ry) that minimized egg-box effects. Only the cell parameters were kept fixed to the values they have in pristine graphene, to mimic the behaviour of realistic dimer lines embedded in graphene, disregarding the interesting effect of a tensile or a compressive stress normal to the lines24 whose analysis would be beyond the aims of this work.
Formation of hydrogen-dimer lines induces substantial lattice reconstruction, as a consequence of the sp2–sp3 rehybridization which necessarily occurs on the underlying carbon atoms if they have to bind a hydrogen atom each. Structural relaxation extends for quite large distances from the dimer lines and determine a sizable curvature ξ of the graphene sheet which scales linearly with the inverse lattice dimension up to n = 10; for larger values of n the bending of the sheet extends ∼4–5a0 on either side of each hydrogen dimer line and a flat region resembling pristine graphene develops in the middle of the channels. Similar results were obtained when allowing the cell parameters to vary and/or accounting for the presence of a substrate (see ESI†). The above bending of the graphene sheet is essential for the hydrogen-dimer lines to act as effective confinement walls, since transport calculations on model, flat SL structures show sizable tunneling across the defect lines (data not shown).
All these properties translate almost unchanged in more complex (n1,n2,…)-SL structures, where the geometry of each ni channel closely resembles that of the constituent ni-SL structure, see e.g. the case of the (6,11) SL structure shown in Fig. 1.
As for the details of the relaxed structures, the smallest H–H distance in the lines (i.e. the dimer extension) is 2.02 Å, the separation between dimers is 2.24 Å and the carbon–hydrogen bond length is 1.14 Å. Similarly to isolated ortho-dimers,28 the hydrogen atoms are not aligned perpendicularly to the graphene sheet, rather they are slightly tilted by steric repulsion; consequently, the benzene rings accommodating hydrogen atoms in para-positions take a stable boat conformation.
We further analyzed the spatial properties of the electronic states by looking at the Local Density of States (LDOS) ρ(x,ε) = 〈x|δ(ε − HKS)|x〉 which measures the weight of the system eigenstates at energy ε in a given point of space x. For the structure shown in Fig. 2 we evaluated the LDOS at several energies above and below the Fermi level, close to the van Hove singularities which mark the channel opening. An example is given in Fig. 3, where the isosurfaces of the LDOS at the energies of the νh,1 and νe,1 bands at the Γ point are reported. Note that they are clearly electron–hole symmetric, with the phase in one sublattice reverted when passing from h to e and the bond patterning changing accordingly.‡ The zeros of the localized state wave functions are clearly distinguished for m = 1 only because for m > 1 a number of states (with different kinetic energy along the channel) contribute to the LDOS.
![]() | ||
Fig. 3 Isosurfaces of the local density of states evaluated in correspondence of the lowest energy van Hove singularity of Fig. 2, νh1 and νe1, in the left and right panels, respectively. |
Finally, the energy gap in n-SL structures is generally decreasing for increasing n, though the actual decay depends strongly on whether n = 3m, 3m + 1 or 3m + 2 (m integer), analogously to what happens in armchair nanoribbons.6,7 This is shown in Fig. 4 where the energy gap of several SL structures, as computed at the above level of theory, is reported as a function of n (for a comparison, notice that the number of CC dimers belongs to the 3m + 2, 3m + 1 and 3m sequences, respectively). Note in particular, that the sequence n = 3m has a gap which is almost unchanged with the width. The computed values agree reasonably well with those previously reported for these structures,24 once the differences in the DFT functionals adopted are taken into account.
Furthermore, we considered the energy gap of several (n1,n2)-SL structures and found, in agreement with the dispersionless behavior along M–Γ, that it is always given, to a good approximation, by the ni-SL with the smallest gap. This supports the idea that the conduction channels are independent of each other, as we detail in the next section by showing the results of our transport calculations.
Fig. 6 reports the zero-bias cumulative transmission function N(ε) ≡ G(ε)/G0 of 6-SL structures, both along the hydrogen-dimer lines and perpendicular to them. The results refer to a single unit cell along (x, y), and were obtained with a fine k-mesh containing 100 k-points parallel to the slab. It is evident from this figure that electronic transport is strongly anisotropic, with a nearly vanishing transmission probability perpendicular to the lines, despite the presence of a single hydrogen-dimer line to be crossed. In contrast, transport along the dimer lines occurs without scattering and the transmission function shows an almost perfect step-like behavior as expected when transmission modes progressively open. Similar results were obtained for different SL structures considered, and thus in the following we focus only on the case where transport occurs parallel to the hydrogen lines.
![]() | ||
Fig. 6 Conductance function for the 6-SL structure along the hydrogen-dimer lines (red) and perpendicularly to them (black). The latter values have been multiplied by a factor of 10 for clarity. |
In general, transport along the dimer lines remains rather simple in all the n-SL structures: the step-like behavior of N(ε) marks the opening of transport channels, with a step occurring at any energy where a transverse mode appears. The height of each step reaches its maximum (usually corresponding to one perfectly conducting channel) in the very small energy window where dispersion along the transverse direction occurs.
Less trivial results are obtained upon combining superlattices with different widths, and exploiting their different electronic properties (energy gap). For instance Fig. 7 reports the transmission results for the (6,7)-SL structure, along with its band structure. Notice that in this case, the energy gap of the constituent 7-SL structure is about 0.5 eV larger than that of the 6-SL one, despite its larger width (see Fig. 4). As a consequence, the gap in the transmission function closely resembles that in the constituent 6-SL structure (Fig. 6, and black lines in Fig. 7), while the steps mainly follow the denser energy levels of the 7-SL structure. In fact, on closer inspection, the transmission function of the compound (6,7)-SL structure is given, to a very good approximation, by the sum of the transmission in the two constituent structures, the 6- and 7-SL structures. This was similar for all the compound structures we considered (not reported).
Stated differently, the channels confined by the hydrogen-dimer lines essentially act as independent conduction channels, i.e. as resistors in parallel, as a consequence of the weakly dispersing character of the energy bands along the direction transverse to the dimer line. Thus, hydrogen-dimer lines work as hard-wall confining potentials for the electrons in graphene, much like the presence of edges in nanoribbon geometries.
Fig. 8 shows how the eigen-transmissions in the (6,7)-SL structure sum up to give the cumulative transmission probability: they are perfectly conducting channels with different threshold energies (i.e. t(ε) ≡ 1 when the channel is open). Correspondingly, the eigenchannels localize in different superlattice regions, according to the band structure of the constituent elements. For instance, the first and fourth steps in the conductance function are due to the energy levels in the 6-SL structure (see also Fig. 7) and the corresponding eigenchannels localize in the six-unit-wide canal.
These results confirm that confinement is indeed at work in electronic transport and that each section defined by a pair of hydrogen-dimer lines acts as an independent transport channel, as previously suggested by the spatial distribution of the HOMO–LUMO in multiple period structures22 (compare also Fig. 3 with the m = 1 eigenchannel reported in Fig. 8).
As expected, the effect of the bias is rather simple: the potential drop inside the channel occurs smoothly from one electrode to the other, and determines a band-mismatch between the left and right ends. As a consequence, the shape of the conductance function is simply determined by an overlap requirement and can be calculated to a good approximation as G(ε,VSD) = min{G(ε − |e|VSD/2,0),G(ε + |e|VSD/2,0)}.
An exception is for energies within the gap region, where transport can occur, depending on the channel length, from the valence band of one end to the conduction band of the other end, upon tunneling through the gap region. Thus, the peak at the center of the conductance spectrum reflects the increase of tunneling probability that occurs when the barrier width decreases as a consequence of a bias increase. In this regime, the transistor simply breaks down, and the total current has an approximately cubic increase, which fits well to Simmons' tunneling in the intermediate-voltage range,42I ≈ JL(V + γV3). This is shown in the inset of Fig. 9 which reports the I–V characteristic obtained with the above bias values.
All this suggests that hydrogen-dimer lines may represent a valid alternative to graphene nano-ribbons to be used in one and the same support for fabricating (chemically imprinting) integrated nano-circuits. Such lines may be used to define insulating layers and create electron waveguides (as exemplified by a (n1,n2)-SL structure with n1 small – the insulating sections – and n2 ≫ n1 – the electron waveguides), or to draw semiconducting ribbons (when n is sufficiently small), provided transverse lines are introduced in the remaining armchair directions to block conduction outside the channel.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4cp01025f |
‡ Electron–hole symmetry implies that the phase of the wavefunction on one sublattice is reverted when moving from electrons to holes, i.e. if ψ = ϕA + ϕB is the wavefunction at a given energy ε and ϕA, ϕB its amplitudes on the A, B sublattices, then ϕ = ϕA − ϕB is its e–h counterpart at energy −ε. |
This journal is © the Owner Societies 2014 |