Unveiling and controlling the electronic structure of oxidized semiconductor surfaces: Crystalline oxidized InSb(100)(1 × 2)-O

The exothermic nature of oxidation causes nearly all semiconductor applications in various fields like electronics, medicine, photonics, and sensor technology to acquire an oxidized semiconductor surface part during the application manufacturing. The significance of understanding and controlling the atomic scale properties of oxidized semiconductor surfaces is expected to increase even further with the development of nanoscale semiconductor crystals. The nature of oxidized semiconductor layers is, however, hard to predict and characterize as they are usually buried and amorphous. To shed light on these issues, we pursue a different approach based on oxidized III-V semiconductor layers that are crystalline. We present a comprehensive characterization of oxidized crystalline InSb(100)$\left(1\times 2\right)$-O layers by ab initio calculations, photoelectron spectroscopy, scanning tunneling microscopy, and spectroscopy, and demonstrate the electronic band structures of different oxidized phases of the semiconductor, which elucidate the previous contradictory semiconductor-oxidation effects. At 0.5 monolayer (ML) oxidation, oxygen atoms tend to occupy subsurface Sb sites, leading to metallic states in the semiconductor band gap, which arise from top dimers. When the oxidation is increased to the 1.0–2.0 ML concentration, oxygen occupies also interstitial sites, and the insulating band structure without gap states is stabilized with unusual occupied In dangling bonds. In contrast, the 2.5–3.0 ML oxide phases undergo significant changes toward a less ordered structure. The findings suggest a methodology for manipulating the electronic structure of oxidized semiconductor layers.

Figure 1

(a) Spheres mark the oxidation conditions found for InSb(100)$\left(1\times 2\right)$-O. (b) LEED pattern from InSb(100)$\left(1\times 2\right)$-O: white square marks the lattice of the $\left(1\times 1\right)$ bulk plane, and white rectangle is the lattice of surface superstructure. (c) Large-scale empty-state STM image from InSb(100)$\left(1\times 2\right)$-O taken with 1.67 V and 0.05 nA.

Figure 2

(a) Stable $\left(1\times 2\right)$ model with 0.5 ML of oxygen. The oxygen sites are numbered. (b)–(d) Stable 1.0, 1.5, and 2.0 ML atomic models, respectively.

Figure 3

Two different views (a) and (b) for the ordered and (c) and (d) for less ordered models of the InSb(100)$\left(1\times 2\right)$-O surface with 2.5 ML of oxygen. Two different views (e) and (f) for the ordered and (g) and (h) less ordered models of the InSb(100)$\left(1\times 2\right)$-O surface with 3.0 ML of oxygen.

Figure 4

(a) Filled-state zoomed STM image measured from InSb(100)$\left(1\times 2\right)$-O; the tunneling current is 0.1 nA and the voltage is −2.9 V. (b) Simulated filled-state (−2.9 V) STM images for the energetically favored models with different oxygen concentrations. (c) Empty-state zoomed STM image measured from InSb(100)$\left(1\times 2\right)$-O; the tunneling current is 0.05 nA and the voltage is +2.1 V. (d) Simulated empty-state (+2.1 V) STM images for the energetically favored models with different oxygen concentrations.

Figure 5

(a) Differentiated STS curve of InSb(100)$\left(1\times 2\right)$-O. (b) Valence-band photoemission from InSb(100)$\left(1\times 2\right)$-O alongside the Fermi edge measured on a Ta plate in electrical contact with the InSb.

Figure 6

(a) The calculated band structure of InSb(100)$\left(1\times 2\right)$-O at 0.5 ML of oxygen. (b) The corresponding density of states spectrum.

Figure 7

(a) Calculated band structure (LDA) along high symmetry Brillouin directions of the InSb(100)$\left(1\times 2\right)$-O structure with 1.0 ML of oxygen [i.e., the model in Fig. 2]. The Fermi level is at 0 eV. (b) The same as (a) but calculated using HSE06 exchange-correlation functionals. (c) Hybrid exchange-correlation band structure for InSb(100)$\left(1\times 2\right)$-O with 2.0 ML of oxygen [i.e., the model in Fig. 2].

Figure 8

(a) Electron localization function images for the energetically favored 1.0 ML model of InSb(100)$\left(1\times 2\right)$-O. (b) The same as (a) but with the subtraction of the one-electron data.

Figure 9

(a) Measured core-level spectra from InSb(100)$\left(1\times 2\right)$-O and their fittings. The surface sensitivity increases toward the top spectra. For the Sb and In components, the solid lines show the 4$d_{5/2}$ peaks of the doublet emission, and the dot lines mark the corresponding 4$d_{3/2}$ peaks. The kinetic energies of photoelectrons $\left(E_{\mathrm{KIN}}\right)$ have been estimated with the BEs of the bulk 4$d_{5/2}$ peaks and lower BE component of O 1$s$.

Figure 10

(a) Large-scale STM image from a well-ordered InSb(100)$\left(1\times 2\right)$-O surface. (b) Large-scale STM image from the InSb surface after a prolonged oxidation, still showing a weak $\left(1\times 2\right)$ LEED. (c) Zoomed STM image from the same InSb surface after a prolonged oxidation showing a bumpy structure.

Figure 11

STM image of the InSb(100)$\left(1\times 2\right)$-O surface containing defects induced by overoxidation. (b) STS curves from different defective areas. (c) Schematic MOSCAP structure. (d) 10 kHz CV curve from MOSCAP with InSb(100)$\left(1\times 2\right)$-O interface. (e) 10 kHz CV curve from the MOSCAP with an overoxidized InSb surface where the well-ordered $\left(1\times 2\right)$ reconstruction has disappeared. The CV curve from the MOSCAP without the preoxidation is similar to (e).