Periodic Nanoarray of Graphene pn-Junctions on Silicon Carbide Obtained by Hydrogen Intercalation

24/05/2022 1 Introduction

The growth of epitaxial graphene on silicon carbide (SiC) surfaces is considered to be one of the most promising methods for graphene production due to its scalability and compatibility with standard complementary metal-oxide–semiconductor (CMOS) fabrication processes.[1-6] The choice of the particular SiC surface plays a key role in determining the transport properties of graphene. As it was shown previously, on-axis epitaxial graphene layers exhibit a much lower resistance anisotropy[7] compared to graphene layers grown on vicinal SiC surfaces.[89] In fact, the quality of epitaxial graphene layers produced over the past years has improved to an extent that nowadays graphene on semi-insulating SiC(0001) is being used for quantum metrology.[1011] Yet, in spite of its high-quality production achieved thus far and its innate remarkable attributes,[12-14] tailoring the electronic properties of graphene through various means is a crucial requirement for the actual development of graphene-based technologies. For instance, 1D confined armchair graphene nanoribbons (AGNRs) are produced when graphene is epitaxially grown on the sidewalls of 6H-SiC mesa structures oriented along the [112⎯⎯0][112¯0]-direction of SiC, which coincides with the [11⎯⎯00][11¯00]-direction of graphene/graphite.[15-17] Contrary to a pristine graphene layer, the as-grown ribbons display a width-dependent semiconducting gap in their electronic band structure[1819]—a key ingredient regarding logic electronics. The electronic behavior is drastically modified for graphene grown on mesas structured parallel to the [11⎯⎯00][11¯00]-direction of SiC (i.e., [112⎯⎯0][112¯0] of graphene/graphite), resulting in the formation of zigzag graphene nanoribbons (zzGNRs).[17] The latter reveal topologically protected spin-polarized ballistic transport channels at room temperature, which are essential for the realization of spintronic devices.[20-24]

Another well-established technique of functionalizing epitaxial graphene is through the intercalation of foreign atomic elements at the graphene/SiC interface.[25-31] Typically, the intercalated atoms saturate the dangling bonds of the topmost Si layer of the substrate, thereby decoupling the electronically inactive (63⎯⎯√×63⎯⎯√)𝑅30∘(63×63)R30∘ reconstructed carbon buffer layer and transforming it into a quasi-free-standing monolayer graphene (QFMLG). The intercalation of Ge and Au are of particular interest for the current study, since they lead to the formation of graphene pn-junctions on mesoscopic scales depending on the amount of intercalated element.[32-36] A plethora of applications have been hypothesized, and in some cases experimentally demonstrated, that are based on graphene pn-junctions. In the ballistic transport regime, Klein tunneling devices and graphene-made metamaterials with extreme electron focusing capabilities could be engineered.[37-40] These include Veselago lenses,[4142] Fabry–Pérot interferometers,[43] electron waveguides analogous to optical fibres,[44] and high-frequency rectifiers.[45] A different class of optoelectronic devices, that do not necessarily require ballistic carriers, consists of photodetectors covering a broad range of the electromagnetic spectrum, where the underlying physical mechanisms governing the conversion of absorbed photons into electric signals are the photovoltaic and photo-thermoelectric effects.[46-49]

In the majority of the aforementioned studies, graphene pn-junction devices were fabricated using appropriately designed top gate structures. While this approach ensures the flexibility to tune the carrier concentration in graphene, it suffers from the presence of electrostatic stray fields at the gate edges and rough junction interfaces. Such imperfections result in spatially extended pn-junctions of finite widths, causing the devices to underperform.[5051] In this respect, chemically gating graphene via intercalation offers a compelling alternative, as it was shown to produce atomically sharp pn-junctions.[33] However, the drawback of the current intercalation procedures (i.e., the intercalation of epitaxial graphene on SiC(0001) surfaces), is the inability to predetermine the layout of the differently doped graphene regions. The p- and n-phases of graphene will be randomly distributed over the sample due to the statistical nature of the intercalation and deintercalation processes.[32] Even with the help of a shadow mask, that may provide a better control over the amount of material being intercalated on different parts of the sample,[36] a process allowing the precise structuring of graphene pn-junctions at the nanoscale is yet to be accomplished.

In this work, we propose a scalable method for producing periodic nanoarrays of graphene pn-junctions on semiconducting SiC substrates. Via H-intercalation, 1D confined AGNRs are transformed into a single 2D graphene carpet rolling over the SiC mesa structures. Due to the different surface terminations of the vicinal and basal SiC planes, different carrier types are locally induced into the graphene sheet, resulting in spatially well-defined pn-junctions. Angle-resolved photoelectron spectroscopy (ARPES) measurements clearly demonstrate the conversion of the electronic system under consideration from 1D to 2D, and reveal two symmetrically doped graphene regions, with p-type being located on the basal planes and n-type on the facets. Scanning tunneling microscopy (STM) indicates a successful delamination of the buffer layers from the substrate through H-intercalation—an observation that is further corroborated by means of low-energy electron diffraction (LEED) and soft-X-ray photoelectron spectroscopy (XPS).

2 Results and Discussion

2.1 Hydrogen Intercalation of Epitaxial Armchair Graphene Nanoribbons

The epitaxial growth of graphene on 6H-SiC mesa structures is performed by the thermal decomposition process detailed in refs. [17] and [18] (the parameters used for growth are also provided in the Sample Preparation paragraph of the Experimental Section below). The 3D representation of the atomic force microscopy (AFM) image in Figure 1a, shows the periodic arrangement of the mesa structures oriented parallel to the [112⎯⎯0][112¯0]-direction of SiC. In this particular orientation, the mesa sidewall decomposes into an array of mini-facets as a result of graphene growth, hosting 1D confined AGNRs separated from each other by electronically inactive nanobuffer regions.[18] A characteristic STM topography image taken on a mesa sidewall is displayed in Figure 1b, showing a mini-facet periodicity of D = (3.7 ± 0.2) nm. After H-intercalation, which was carried out following the procedure introduced by Riedl et al.[25] (detailed in Section 4 of the present manuscript as well), the apparent distinction between AGNR and nanobuffer layers fades away, and a continuous hexagonal lattice emerges on the mesa sidewall (Figure 1c). This implies that through H-intercalation, the buffer layers decouple from the SiC substrate, and the otherwise segregated 1D confined AGNRs transform into a single 2D QFMLG rolling over the mesa structures as shown in the schematic structural model of Figure 1d.

Details are in the caption following the image
Figure 1
Structural properties of epitaxial graphene grown on 6H-SiC mesa structures. a) Perspective AFM view of the mesa structures oriented along the [112⎯⎯2¯0]-direction of SiC, having a periodicity of 200 nm and a trench depth of 20 nm. b,c) STM topography image taken on a mesa sidewall b) before and c) after H-intercalation. The tunneling parameters were (1 V, 0.5 nA) and (0.2 V, 0.4 nA) respectively. The [11⎯⎯1¯00]- and [112⎯⎯2¯0]-directions of graphene (Gr) are shown at the bottom left corners of each panel. d) Schematic structural model of the sample demonstrating the delamination of the buffer and nanobuffer (NB) layers from the substrate. The array of AGNRs transforms into a single QFMLG rolling over the facet and basal SiC regions. The mini-facet periodicity is D = (3.7 ± 0.2) nm and the step height H is the height of a 6H-SiC unit cell.

In order to support the structural model of the intercalation process sketched in Figure 1d, LEED and XPS measurements are presented in Figure 2. Before intercalation, diamond-like spot arrangements (marked by the gray diamond) accompanied by weak (1 × 1) graphene spots (green circles) are observed in the LEED patterns. They clearly indicate the presence of electronically inactive (63⎯⎯√×63⎯⎯√)𝑅30∘(63×63)R30∘ reconstructed carbon buffer layers on the plateaus and trenches of the mesas, i.e., the basal planes of SiC.[3] The spot-chains oriented perpendicular to the [112⎯⎯0][112¯0]-direction of SiC originate from the well-ordered mini-facet structures on the mesa sidewalls.[18] After H-treatment, the diffraction spots belonging to the (63⎯⎯√×63⎯⎯√)𝑅30∘(63×63)R30∘ superstructure grid get significantly suppressed, and intense graphene (1 × 1) spots emerge from both basal (green circles) and facet (red circles) regions, demonstrating a successful decoupling of the buffer layers from the substrate.[25] The facet graphene spots are replicated with a period of D′ = (0.20 ± 0.03) Å−1, which agrees with the mini-facet periodicity D observed in STM (Figure 1b). The real-space and reciprocal-space periodicities are related to each other by D′ = 2π/D.

Details are in the caption following the image
Figure 2
LEED and XPS experiments performed before (AGNR) and after (QFMLG) H-intercalation. In the LEED patterns, the spots originating from the (63⎯⎯√×63⎯⎯√)𝑅30∘(63×63)R30∘ superstructure are represented by the gray diamond, while the basal and facet (1 × 1) graphene spots are indicated by the green and red circles respectively. The corresponding (0,0) spot positions of the basal and facet graphene layers are also shown. The facet graphene spots are replicated by a period of D′ = (0.20 ± 0.03) Å−1 as indicated by the inset of the bottom left LEED picture. The core level spectra are measured at different photon energies (hν = 140 eV for Si 2p, hν = 400 eV for C 1s before H-intercalation, and hν = 427 eV for C 1s after H-intercalation). Characteristic 2p3/2 and 2p1/2 components are presented by the green dotted lines for the bulk Si 2p peak before H-intercalation. In addition to the major Si 2p and C 1s components, minor defects were also observed, which are indicated by the gray lines. A detailed discussion about the fitting procedures can be found in Note S1 of the Supporting Information.

The XPS measurements are also in accordance with the LEED and STM results. Before intercalation, the Si 2p signal consists of a bulk component (marked by green) located at 101.17 eV, and a basal surface component (orange) that is 0.40 eV higher in binding energy relative to its bulk counterpart, as expected for epitaxial graphene buffer layer on SiC(0001).[52] The third peak situated at a binding energy of 100.48 eV (purple line)—not commonly observed on standard epitaxial graphene samples—is attributed to facet SiC. As shown by the schematic ball-and-stick model of Figure 1d, the bulk-truncated vicinal planes on the mesa sidewalls exhibit a different surface structure as compared to that of the basal planes. The vicinal surfaces are composed of both Si and C atoms—showing no strong interaction with their overhead free-standing AGNRs[18]—while the basal surfaces, composed of only Si atoms, are partially bonded to the graphitic layer, giving rise to the buffer and nanobuffer regions. This implies that the bonding configurations associated with the topmost substrate atoms comprising the basal and vicinal SiC planes are nonidentical. Furthermore, in order to minimize their surface energies, it is likely for the vicinal planes to undergo a reconstruction (not shown in Figure 1d) rather than maintaining their bulk-terminated forms. These differences in chemical environments effectively lead to the facet related SiC peaks revealing different binding energies as compared to those of the basal surface and bulk SiC components. Due to spin-orbit coupling, the Si 2p components are represented by doublets with an energy splitting of 0.60 eV between the 2p3/2 and 2p1/2 peaks (individually shown by green dotted lines in the bulk Si 2p peak before intercalation), and a relative peak area ratio of 2:1. The binding energy values of the different components are defined with respect to their corresponding 2p3/2 peaks. The C 1s core level signal is comprised of a bulk peak located at a binding energy of 283.43 eV (shown in green), and two surface components S1 (284.55 eV) and S2 (285.30 eV) representing the partial bonding of the buffer layer to the substrate (orange lines).[53] The additional contributions arising from the presence of mesa structures are the facet SiC peak (282.78 eV—purple line), which was similarly observed in the Si 2p spectrum, and the AGNR peak situated at 284.13 eV (black line). A minor portion of the AGNR signal originates from parasitic graphene layers (p-Gr), that in small patches have overgrown on the mesa sidewalls. Upon intercalation, the bulk components of both Si 2p and C 1s core levels, along with the surface peak of Si 2p, are shifted toward lower binding energies by about 0.9 eV. The surface components S1 and S2 of the C 1s core level on the other hand are transformed into a single graphene peak (orange line—basal graphene) located at 284.17 eV. These observations are characteristic signatures of a successful delamination of the buffer layers from the substrate through saturation of surface Si dangling bonds by H, and a subsequent development of QFMLG.[25] In fact, the complete disappearance of the intense buffer layer peaks S1 (284.55 eV) and S2 (285.30 eV), is evidence for a homogeneous intercalation of the sample on a large scale, ruling out the possibility of a partial intercalation scenario. In addition to the basal graphene peak, a facet graphene peak (black line) is discerned at 284.62 eV. The apparent difference in binding energy between the facet and basal graphene peaks is not due to a difference in bonding configuration—as was the case for the facet substrate peaks—but a difference in doping level between the two graphene regions, whose occurrence and plausible causes giving rise to it are explored in Section 2.3 below. Note that after H-intercalation, basal QFMLG lies flat on the SiC(0001) plane, while facet QFMLG is subjected to slight corrugation as it rolls over the mini-facets. These structural variations may also play a minor role in the determination of the absolute binding energy values of the respective core levels in XPS. Interestingly, the binding energies of the facet SiC peaks in both Si 2p and C 1s spectra remain almost unchanged. We tentatively ascribe this behavior to a difference in interlayer band bending present on the facet as compared to that on the basal plane, while emphasizing the need and importance of future studies that would properly characterize the complex structural and chemical properties comprising this surface.

2.2 Dimensionality Change of the Electronic System

Figure 3 presents a collection of ARPES measurements showing the dimensionality change of the electronic system from 1D to 2D. Before intercalation, the constant energy cuts (CECs) taken at the Fermi level and 1 eV below, consist of straight lines oriented across the facets (i.e., across the AGNRs) as expected for 1D confined Dirac electrons (Figure 3a). The intensity modulations observed in the CECs are due to Dirac cones located at the K⎯⎯⎯K¯-points of the p-Gr layers overgrown in small patches on the oppositely inclined facets.[18] After H-intercalation, new Dirac cones arranged in a sixfold symmetric pattern emerge, indicating the decoupling of the buffer layers from the substrate and the development of QFMLGs on the basal planes.[25] We note that these six Dirac cones are not the primary cones situated at the K⎯⎯⎯K¯-points of the basal graphene Brillouin zone, but rather satellite cones replicated by the basal SiC reciprocal lattice vectors 𝑆→1S→1 and 𝑆→2S→2.[54] In addition, the straight lines in the CECs that are characteristic features of 1D confined electronic systems vanish upon intercalation, and are replaced by a series of π-band replicas whose origins will be discussed in detail below (Section 2.4). The 2D reciprocal space maps for the CECs measured at hν = 60 eV are displayed in Figure 3b. They are constructed by drawing the hexagonal Brillouin zones of facet and basal graphene layers on top of the LEED patterns of corresponding kinetic energy.[18] Figure 3c shows high-resolution ARPES energy-momentum cuts taken at the K⎯⎯⎯K¯-points of the oppositely inclined facet graphene layers, whose positions in k-space, at hν = 50 eV, coincide with the Γ⎯⎯⎯Γ¯-point of basal graphene (the measurement direction is indicated by the orange line in Figure 3e). Both spectra acquired before (AGNR) and after (QFMLG) H-intercalation are differentiated twice (second derivative) along their energy axes. The electronic structure of AGNRs is comprised of a number of distinct subbands (the most prominent ones are marked by the red arrows), which are striking signatures of 1D quantum confinement.[18] The subbands completely disappear and a standard graphene Dirac cone develops subsequent to H-intercalation, demonstrating a clear transition from 1D confined electrons to a 2D gas of massless Dirac fermions. A similar behavior is observed in the ARPES cuts measured across the facets as shown in Figure 3d (the measurement orientation is denoted by the purple line in Figure 3e). The nondispersive subbands marked by the red arrows transform into strongly dispersive π-bands located at the K⎯⎯⎯K¯-points of the oppositely inclined facet QFMLGs, indicating once more that 1D confinement is lifted as a result of H-intercalation. In this specific orientation of the ARPES measurement, the two branches of each facet Dirac cone (located at the red and blue K⎯⎯⎯K¯-points) are not equally intense due to the well-known dark corridor effect.[55]

Details are in the caption following the image
Figure 3
1D to 2D transition of the electronic system. a) Fermi surface and a CEC 1 eV below the Fermi level acquired at a photon energy of 60 eV before (AGNR) and after (QFMLG) H-intercalation. b) 2D reciprocal space maps: LEED patterns at 55.5 eV corresponding to a photon energy of 60 eV, superimposed by the hexagonal Brillouin zones of facet and basal graphene layers. 𝑆→1S→1 and 𝑆→2S→2 represent the reciprocal lattice vectors of the basal SiC surface. c,d) Second derivative of high-resolution ARPES energy-momentum cuts measured c) along and d) across the facets. The red arrows indicate the positions of the most prominent AGNR subands, which disappear after H-intercalation. The k-space conversion in (d) is made relative to the normal emission (Γ⎯⎯⎯Γ¯-point) of the blue facet QFMLG. e) 2D reciprocal space maps corresponding to photon energies of 50 and 35 eV. The orange and purple lines indicate the orientations of the ARPES spectra taken in panels (c) and (d) respectively. The LEED patterns used in these 2D reciprocal space maps are those measured subsequent to H-intercalation.

2.3 Periodic Nanoarray of Graphene pn-Junctions

The doping difference between the facet and basal QFMLGs is assessed on the basis of ARPES measurements. Figure 4a presents a CEC taken 1 eV below the Fermi level, where the entire Brillouin zones of basal (green hexagon) and facet (red and blue hexagons) QFMLGs are observed. The measurement was performed by means of a momentum microscope (nanoESCA), which can cover nearly the full photoemission horizon in a single shot when using a photon energy of 40.8 eV (He II α radiation). A side-view of the mesa structure is also displayed (Figure 4b), demonstrating how a facet inclination in real-space results in shifting the corresponding Brillouin zones (represented by the red and blue Γ⎯⎯⎯Γ¯-points) away from the normal emission of the basal plane (green Γ⎯⎯⎯Γ¯-point) in reciprocal space. This angular separation allows to investigate the electronic properties of the different graphene regions separately. ARPES energy-momentum cuts acquired at the K⎯⎯⎯K¯-points of facet and basal QFMLGs are shown in Figures 4c and 4d, respectively. Their spatial distribution over the SiC mesa structures is represented by the red/blue and green colored planes overlaid on top of the 3D view of the AFM image in Figure 4e. Note that the latter is the same AFM image used in Figure 1a with a different perspective applied to it. Since both red and blue facet QFMLGs are electronically similar, a single ARPES cut is used to represent their Dirac cones. The green and red dotted lines superimposed on the ARPES data track the spectral maxima of the MDCs (and linearly extrapolate the curves in the case of basal QFMLG), revealing Dirac point crossings located 0.22 eV above the Fermi level for basal QFMLG and 0.31 eV below the Fermi level for facet QFMLG. The resulting pn-junction potential barrier height is 0.53 eV, with hole concentrations of p = 3.86 × 1012 cm−2 on the basal planes, and electron concentrations of n = 3.12 × 1012 cm−2 on the facets. The doping levels are evaluated using the equation 𝜌=𝑘2F/𝜋ρ=kF2/π, where ρ is the carrier density and kF is the Fermi wavevector.[41]

Details are in the caption following the image
Figure 4
Periodic nanoarray of graphene pn-junctions across the facets. a) ARPES CEC taken 1.0 eV below the Fermi level using a photon energy of 40.8 eV (He II α radiation). The k-space field of view is set to FOV = 6.2 Å−1, covering nearly the full photoemission horizon. The entire Brillouin zones of facet (red/blue hexagons) and basal (green hexagon) QFMLGs are captured in a single shot. The facet QFMLG Brillouin zones are centered around their own Γ⎯⎯⎯Γ¯-points (red/blue), which are shifted away from the normal emission of the basal plane (green Γ⎯⎯⎯Γ¯-point) as shown in b) the schematic side-view of the mesa structure. c,d) ARPES energy-momentum cuts acquired at the K⎯⎯⎯K¯-points of c) basal and d) facet QFMLGs, respectively. e) The colored regions superimposed on top of the 3D AFM image (which is the same image used in Figure 1a with a different perspective applied to it) show the spatial distribution of basal (green) and facet (red/blue) QFMLGs over the SiC mesa structures. Basal QFMLGs are p-doped, while facet QFMLGs are n-doped, forming periodically repeating graphene pn-junctions across the facets, with a doping difference between the two graphene layers amounting to 0.53 eV.

The p-type doping observed in basal QFMLG is explained by the polarization doping model introduced by Ristein et al.[56] 6H-SiC, being a pyroelectric material, possesses an intrinsic spontaneous polarization in its bulk oriented along the 𝑐→c→ axis. At the surface of the substrate, the polarization density discontinuously drops to zero due to the termination of the bulk. This sudden change in polarization is equivalent to bound surface charges (pseudo charges) at the graphene/SiC interface, acting as a reservoir of acceptor-like states pulling electrons away from the basal graphene layer, namely p-doping it.[5657] The situation on the facet is more complex since the surface of the substrate consists of both Si and C atoms, contrary to the basal plane of SiC (i.e., the Si-face of SiC) composed of only Si atoms. During H-treatment, the apparent energetically favorable state for H is to form Si-H covalent bonds while keeping the C dangling bonds intact. Otherwise, graphene buffer layers would have been hydrogenated—something that has never occurred in a H-intercalation experiment performed in similar conditions as the ones used in the present study. Although there have been reports of H-trapping in multilayer stacks of graphene on SiC,[58] such hydrogenation or C-H bond formation has not been observed on H-intercalated buffer layer graphene samples.[5960] For this reason, on the sidewalls of SiC mesa structures, H presumably binds to Si subsequent to H-intercalation, leaving behind an interfacial layer of substrate C atoms with unsaturated dangling bonds, which essentially plays a similar role for facet QFMLG as a buffer layer does for MLG on SiC(0001).[56] The donor-like states associated with the unsaturated substrate C dangling bonds overcompensate the pseudo charges on the mesa sidewalls (whose density is proportional to the spontaneous polarization component normal to this surface[61]) and transfer electrons to facet QFMLG, which would explain the n-type doping observed in ARPES. We note that the 6H-SiC substrates used in this study are n-doped with an estimated bulk carrier concentration of n ≈ 5 × 1017 cm−3.[57] When performing a similar experiment on semi-insulating 6H-SiC substrates (n < 1014 cm−3), the Fermi levels of both facet and basal QFMLGs are expected to shift rigidly downward by few tens of meV—a behavior previously reported for H-intercalated mono- and bilayer graphene grown on n-doped and semi-insulating SiC substrates.[57] This rigid shift in Fermi level would change the overall doping level within the two graphene layers, while keeping the doping difference between basal and facet QFMLGs, i.e., the pn-junction potential barrier height, constant.

2.4 Origin of the π-Band replicas
In order to understand the origin of the π-band replicas observed in the CECs of Figures 3a and 4a, a constant initial state map (CISM) is presented in Figure 5a. The CISM is composed of a series of momentum distribution curves (MDCs), extracted from the Fermi level of ARPES energy-momentum cuts measured across the facets as a function of photon energy. A characteristic energy-momentum cut is displayed in Figure 5b. The red dashed line indicates the MDC extracted at the Fermi level, which is used for constructing the CISM of Figure 5a. The orientation of the ARPES cut across the facets is shown by the purple line in the 2D reciprocal space map of Figure 5c. By increasing (decreasing) the photon energy, the sizes of the Brillouin zones in Figure 5c shrink (expand), and the angular distances between the K⎯⎯⎯K¯-points of the oppositely inclined facet QFMLGs and their corresponding replicas, change accordingly.[18] This trend can nicely be observed in the 2D reciprocal space maps of Figures 3b and 3e. The motion of the distinct Dirac cones as a function of photon energy can be tracked by fitting individual bands in the CISM using the basic photoemission equation
𝜃𝑦=±  [𝑠𝑖𝑛−1(ℏ𝑘𝑦2𝑚𝑒(ℎ𝜈−𝜙)⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯√)−𝛿]θy=±  sin−1ℏky2mehν−ϕ−δ
(1)
where θy is the emission angle across the facets, ky denotes the crystal momentum across the facets, me stands for the mass of an electron, ϕ = 4.39 eV is the analyzer workfunction, and δ represents the inclination of facet QFMLGs relative to the basal plane. In Equation (1), when fixing |ky| to 1.7 Å−1 (which corresponds to the K⎯⎯⎯K¯-point of graphene), two families of facet QFMLG inclinations are found. The first is δ = ±29.4° indicated by the solid red (+29.4°) and blue (−29.4°) curves, which perfectly fit the most intense bands in the CISM (Figure 5d). The second is δ = ±21.6° marked by the solid orange lines, fitting the bands with a slightly steeper dispersion as compared to the bands belonging to the δ = ±29.4° family. It is important to clarify the meaning of the word “dispersion” in this context. In the CISM, the states located at the Fermi level “disperse” or “move” at different rates as a function of photon energy, depending on the facet inclination they originate from. This should not be confused with the common definition of the word “dispersion” (borrowed from band structure theory), which refers to the relationship between the binding energy of an electron and its crystal momentum, i.e., the dispersion relation. The bands having the steepest dispersion in the CISM (green lines, Figure 5d) correspond to ky = ± 0.68 Å−1 and δ = 0°. They represent the Dirac cones of basal QFMLG that are replicated by the basal SiC reciprocal lattice vectors 𝑆→1S→1 and 𝑆→2S→2.[54] The remaining features observed in the CISM are the facet QFMLG π-band replicas, resulting from their underlying periodic arrays of mini-facets constituting the mesa sidewalls. The replica bands of the δ = ±29.4° family appear with a period of 0.20 Å−1 (red and blue dotted lines), in excellent agreement with LEED and STM, whereas those of the δ = ±21.6° family are replicated with a period of 0.15 Å−1 (orange dotted lines). The presence of two different facet inclinations on the sample with different mini-facet periodicities, is supported by spot profile analysis LEED (SPA-LEED) measurements shown in Figure S2 of the Supporting Information.

Details are in the caption following the image
Figure 5
Origin of π-band replicas. a) Constant initial state map (CISM). b) ARPES energy-momentum cut taken at normal emission, across the facets, using a photon energy of 60 eV. The red dashed line represents the MDC at the Fermi level, which is the same MDC marked by the red dashed line in the CISM of panel (a). c) 2D reciprocal space map corresponding to hν = 60 eV. The purple line shows the orientation of the ARPES cut of panel (b). d) Fit to the CISM—determining the origin of the different primary and replicated π-bands.

3 Conclusion
To conclude, we have demonstrated a scalable approach for producing periodic nanoarrays of graphene pn-junctions on a technologically viable substrate. Through H-intercalation, 1D confined AGNRs separated from each other by electronically inactive nanobuffer regions, are transformed into a single 2D graphene sheet rolling over the SiC mesa structures. Contrary to other elements used in a variety of intercalation studies, the case of H is particularly interesting since the doping of the resulting QFMLG is solely determined by the SiC substrate and not the intercalated element. Namely, p-type carriers are induced into QFMLG on the basal planes of SiC, while n-type carriers are induced on the facets. Our results clearly show that via careful structuring of the substrate, the natural tendency of 6H-SiC to dope its overhead graphene differently depending on its surface termination could be exploited, allowing the precise control over the layout and arrangement of the graphene pn-junctions on the sample—something that has not been achieved by standard intercalation methods, i.e., the intercalation of epitaxial graphene on flat SiC(0001) surfaces. Therefore, the epitaxial growth of graphene on structured SiC surfaces combined with H-intercalation, is a promising approach for engineering integrated networks of graphene pn-junctions at the nanoscale, opening the avenue for the development of novel graphene-based optoelectronic devices.

4 Experimental Section
Sample Preparation
Nominally on-axis, single crystalline, n-doped 6H-SiC(0001) wafer pieces (purchased from SiCrystal GmbH) were used as substrates for the preparation of graphene pn-junctions. H-etching was carried out in an inductively heated reactor equipped with a graphite susceptor, in order to remove residual polishing scratches from the surfaces.[62] The latter was achieved by annealing the samples at 1500 °C, in 800 mbar of purified H2 atmosphere (flow rate set to 1 slm), for 20 min. Using a combination of e-beam lithography and reactive ion etching (gas mixture 20/7 SF6/O2, power = 30 W, pressure = 0.05 mbar),[18] mesa structures were produced along the [112⎯⎯2¯0]-direction of SiC, with sidewalls standing at steep angles relative to the basal plane. The lateral periodicity of the mesa structures was chosen to be 200 nm, and a reactive ion etching time of 2 min was used to obtain a mesa height of about 20 nm separating the plateaus from the trenches. The epitaxial growth of graphene on the SiC mesa structures was performed in an inductively heated furnace, by annealing the sample at 1800 °C under 850 mbar of Ar pressure for 1 min.[17] At this stage, the sidewalls relax into facets angled at ≈26° and 18° relative to the basal plane (as shown by the SPA-LEED measurements in the Supporting Information), and decompose into an array of mini-steps, hosting a multitude of ultra-narrow AGNRs. H-intercalation was achieved in the same reactor used for H-etching, by heating the sample at 730 °C in 800 mbar of purified H2 atmosphere (flow rate set to 1 slm) for 1 h.[25]

Angle-Resolved Photoelectron Spectroscopy and Soft X-Ray Photoelectron Spectroscopy
The photoemission experiments were performed at the Max Planck Institute for Solid State Research (MPI-FKF) in Stuttgart and at different beamlines such as Bloch and MAXPEEM at the MAX IV synchrotron facility in Lund, as well as the 12 endstation of the UE112 beamline at Bessy II, Helmholtz–Zentrum Berlin.

MAX IV—The ARPES measurements on the AGNR samples prior to H-intercalation were carried out at the Bloch beamline of the MAX IV synchrotron facility.[18] These include the AGNR CECs of Figure 3a, and the AGNR energy–momentum cuts taken along and across the facets in Figures 3c and 3d. The ARPES spectra were measured using a high performance deflector-based DA30 hemispherical analyzer from Scienta Omicron. The energy and angular resolutions were set to 15 meV and 0.1° respectively. The spot-size of the beam was 10 × 24 μm2, simultaneously probing around 100 mesas, i.e., 2000 AGNRs. These experiments were performed in ultra-high vacuum (UHV) at a sample temperature of 80 K. The core level spectra (before and after H-intercalation) shown in Figure 2 and the photon energy dependent ARPES scans (i.e., the CISM of Figure 5) were also measured at the Bloch beamline, in UHV, at room temperature. Preliminary microscopy and spectroscopy experiments (not shown in the manuscript) were conducted at the MAXPEEM beamline of the MAX IV synchrotron facility.

Bessy II—After H-intercalation, the QFMLG CECs of Figure 3a, and the QFMLG energy-momentum cuts taken along and across the facets in Figures 3c and 3d, were measured at the 12 endstation of the UE112 beamline at Bessy II.[63] The energy-momentum cuts were acquired using a hemispherical analyzer (Scienta R8000), with an energy resolution set to 20 meV and an angular resolution of 0.1°. All data were taken at a sample temperature of 20 K in UHV.
MPI-FKF—The CEC shown in Figure 4a was measured using a momentum microscope (Scienta Omicron NanoESCA) operated at an energy resolution of 100 meV. The instrument is capable of mapping the photoelectron distribution with a 6.2 Å−1 field of view, which at a photon enery of 40.8 eV covers nearly the entire photoemission horizon. The experiment was performed at room temperature, using non-monochromatized He II α radiation (40.8 eV) as its excitation source. The individual ARPES energy-momentum cuts shown in Figures 4b and 4c were acquired with a Specs Phoibos 150 hemispherical analyzer. Monochromatized He II α photons were used to excite the photoelectrons. The energy and angular resolutions were set to 60 meV and 0.3° respectively. Both spectra were collected at room temperature in UHV.

Scanning Tunneling Microscopy
The STM measurements were performed using a commercial Omicron LT-STM. Before transferring the samples to the low temperature chamber, they were degassed at elevated temperatures (≈400 °C) for several hours. Fourier filtering was applied to the STM image taken after H-intercalation (Figure 1c) in order to improve the visibility of the graphene lattice. All data, prior and post intercalation, were acquired at a cryogenic temperature of 80 K in UHV.

Source: https://bit.ly/3yTUqYi, via Wiley Online Library
Share: