A 2D Bismuth-Induced Honeycomb Surface Structure on GaAs(111)

Abstract

Two-dimensional (2D) topological insulators have fascinating physical properties which are promising for applications within spintronics. In order to realize spintronic devices working at room temperature, materials with a large nontrivial gap are needed. Bismuthene, a 2D layer of Bi atoms in a honeycomb structure, has recently attracted strong attention because of its record-large nontrivial gap, which is due to the strong spin-orbit coupling of Bi and the unusually strong interaction of the Bi atoms with the surface atoms of the substrate underneath. It would be a significant step forward to be able to form 2D materials with properties such as bismuthene on semiconductors such as GaAs, which has a band gap size relevant for electronics and a direct band gap for optical applications. Here, we present the successful formation of a 2D Bi honeycomb structure on GaAs, which fulfills these conditions. Bi atoms have been incorporated into a clean GaAs(111) surface, with As termination, based on Bi deposition under optimized growth conditions. Low-temperature scanning tunneling microscopy and spectroscopy (LT-STM/S) demonstrates a well-ordered large-scale honeycomb structure, consisting of Bi atoms in a √3 × √3 30° reconstruction on GaAs(111). X-ray photoelectron spectroscopy shows that the Bi atoms of the honeycomb structure only bond to the underlying As atoms. This is supported by calculations based on density functional theory that confirm the honeycomb structure with a large Bi-As binding energy and predict Bi-induced electronic bands within the GaAs band gap that open up a gap of nontrivial topological nature. STS results support the existence of Bi-induced states within the GaAs band gap. The GaAs:Bi honeycomb layer found here has a similar structure as previously published bismuthene on SiC or on Ag, though with a significantly larger lattice constant and only weak Bi-Bi bonding. It can therefore be considered as an extreme case of bismuthene, which is fundamentally interesting. Furthermore, it has the same exciting electronic properties, opening a large nontrivial gap, which is the requirement for room-temperature spintronic applications, and it is directly integrated in GaAs, a direct band gap semiconductor with a large range of (opto)electronic devices.

Keywords: 2D layer; DFT; GaAs; STM; bismuth; bismuthene; honeycomb structure.

Figure 1

Honeycomb structure upon Bi deposition on GaAs(111)B at 250 °C. (a) Overview STM image of the honeycomb structure. Green circles indicate some hollow areas and green arrows some Bi clusters on the surface. (b) Atomically resolved STM image of the honeycomb structure. Each bright dot presents an individual Bi atom. The STM color scale extends over 44 pm. (c) Fast Fourier transform (FFT) image of image (b), showing a clear single periodicity of sixfold symmetry. The additional weak spots close to the center are probably related to surface step edges of the underlying GaAs(111)B substrate. (d) Height profile of a line scan across the honeycomb structure as indicated in (b) by a white line. (e) Overview STM image of the clean GaAs(111)B surface, for comparison. Pink arrows indicate step edges of the surface terraces along ⟨110⟩ directions. (f) Height profile of the clean GaAs surface, as indicated by the white line in (e), showing the surface step height. STM imaging parameters are V_T = −3 V, I_T = 50 pA for (a), V_T = −5 V, I_T = 100 pA for (b) and V_T = −3 V, I_T = 80 pA for (e).

Figure 2

Bi-induced honeycomb structure on GaAs(111)B after quick annealing at 400 °C. (a) Overview STM image. Yellow (blue) zone presents the top (lower) surface terrace. Some step edges along ⟨110⟩ orientations are marked with pink arrows; yellow arrows point out areas with an incomplete honeycomb structure. (b,c,d) Close-view STM images of the selected areas marked in (a), presenting different topographies. (e) Line scan along the green dashed line in (a), showing the height profile of the lower terrace (background in green) and the top terrace (background in blue). STM color scale extends over 421, 522, and 316 pm in (b), (c), and (d), respectively. V_T = −3.5 V, I_T = 50 pA for all STM images.

Figure 3

XPS Bi 5d core level spectra at after subsequent processing steps. (a) Overview of Bi 5d spectra along the sample processing timeline (from top to bottom), as described in the main text, after background subtraction. (b–f) Component decomposition and curve fitting of the spectra shown in (a). Corresponding components are marked in (c). The evolution of the Bi intensity upon subsequent processing steps can be followed in (a), where the spectra are shifted along the intensity axis for better visualization, while the intensities are normalized in (b)–(f). Dashed lines in all figures indicate the binding energy (B.E.) position of Bi–As (24.5 and 27.6 eV) and Bi–Bi (23.8 and 26.9 eV) for both 5d_3/2 and 5d_5/2 core levels.

Figure 4

LT-STM images of the Bi-induced honeycomb structure under different tunneling bias, representing different LDOS distribution. (a–j) The same area of a Bi-induced honeycomb structure with tunneling bias ranging from −2.0 V to +3.0 V. The red cross in each image marks the same physical position. All STM images are taken at around 10K. The STM color scale extends over 1.0, 0.9, 2.0, 1.2, 0.9, 0.9, 1.1, 1.0, 0.9, and 0.7 pm in (a)–(j), respectively.

Figure 5

First-principles atomistic model of the Bi honeycomb structure. (a)Top view of the Bi-induced honeycomb structure on top of the GaAs(111)B substrate. The positions of the Bi atoms are emphasized with shaded red circles to highlight its honeycomb lattice that is made up of two sublattices, A and B. The hollow site is set as the origin of the unit cell of the honeycomb lattice structure and a and b define the lattice vectors of the absorbate unit cell. The substrate unit cell is also indicated, shaded in gray and defined using dashed lattice vectors. We see that the absorbate unit is rotated 30° anticlockwise with respect to the substrate and that each absorbate lattice vector is √3 times the length of the corresponding substrate lattice vector, i.e., Bi forms a√3 × √3 30° overlayer in the Wood’s notation. (b) Side view of the GaAs(111)B substrate with (right) and without (left) the honeycomb lattice. The first two Ga–As planes of the substrates are fully relaxed by minimizing the total energy of the system while the third plane is fixed in position to model the bulk. The faint atoms at the bottom of the slab are the pseudohydrogen atoms used to saturate the bonds. On the left part of (b), one sees that surface reconstruction of the pristine GaAs(111)B surface results in crests and troughs at the surface. When deposited, Bi preferentially occupies the troughs, forming Bi–As bonds with the substrate (as shown), and the crests become the hollow sites of the honeycomb lattice (right part of (b)). (c) Calculated charge transfer plot in units of electron charge per supercell volume, along a Bi–As bond. The plot shows a strong localization of charge transfer at the Bi–As bond center.

Figure 6

Electronic properties of the honeycomb structure as-deposited on the GaAs(111)B surface. (a) Calculated STM image using DFT based on the Tersoff–Hamann model. (b) Experimental STM image, acquired at V_T = −2.0 V, I_T = 350 pA. (c) LT-STS () point spectra, obtained at a Bi atom (blue) and a hollow site (orange) of figure (b) and at a clean GaAs(111)B surface (black). (d–f) DFT band structures of the GaAs(111)B surface with and without Bi. The band states are projected onto the atomic pseudowave function of Ga 4p and 5s orbitals (orange), As 4p and 5s orbitals (black), and Bi 6p_x and 6p_y orbitals (teal), with the thickness of the line representing the degree of projection. The Fermi level is at 0 eV. (d) DFT band structure of the pristine (111)B surface of a GaAs substrate with no Bi. (e) DFT band structure of the honeycomb structure as-deposited on the GaAs(111)B surface, calculated without spin–orbit coupling (SOC). There are two overlapping Dirac cones at the K point. (f) DFT band structure of (e) calculated with SOC. SOC leads to the opening of a nontrivial gap at the Dirac cones, and the splitting of the double degenerate bands in the low-energy region near the Fermi level.