Persistent metallic Sn-doped In2O3 epitaxial ultrathin films with enhanced infrared transmittance

Infrared transparent electrodes (IR-TEs) have recently attracted much attention for industrial and military applications. The simplest method to obtain high IR transmittance is to reduce the electrode film thickness. However, for films several tens of nanometres thick, this approach unintentionally suppresses conduction due to surface electron scattering. Here, we demonstrate low sheet resistance (<400 Ω □−1 at room temperature) and high IR transmittance (>65% at the 2.5-μm wavelength) in Sn-doped In2O3 (ITO) epitaxial films for the thickness range of 17−80 nm. A combination of X-ray spectroscopy and ellipsometry measurements reveals a persistent electronic bandstructure in the 8-nm-thick film compared to much thicker films. This indicates that the metallicity of the film is preserved, despite the ultrathin film configuration. The high carrier mobility in the ITO epitaxial films further confirms the film’s metallicity as a result of the improved crystallinity of the film and the resulting reduction in the scattering defect concentration. Thus, ITO shows great potential for IR-TE applications of transparent photovoltaic and optoelectronic devices.


Existence of defects near ITO/YSZ interfaces
Introducing a minute number of defects can degrade the transmittance and conductivity of transparent electrodes (TEs), as they tend to act as scattering centres. The misfit strain in 3 epitaxial films has been widely accepted to easily generate more defects near the film/substrate interfaces 1,2 . As mentioned in the paper, our ITO epitaxial films were under in-plane tensile strain with a +1.4% mismatch. Therefore, we investigated the levels of strain and defects with film thickness variation. Figure S2a shows a reciprocal space map around the (113) diffraction peak of YSZ. For the 80-nm-thick film, the bright spot originating from the (226)ITO diffraction has Qx-and Qz-values close to the bulk values, indicating that the tensile strain was fully released.
With a reduction in thickness, the Qx-value of the (226)ITO diffraction of the 17-nm-thick film approached the Qx-value of the (113)YSZ diffraction, indicating that these ITO films were partially under in-plane tensile strain. The 8-nm-thick film showed the same Qx-values; thus, the ITO films thinner than 8 nm were fully strained. Figure S2b shows X-ray photoemission spectroscopy (XPS) spectra of O-1s, Sn-3d, and In-3d in the ranges of 534-526, 498-484, and 455-441 eV binding energies, respectively 3 . Irrespective of the thickness (8-80 nm), peaks were present at nearly the same binding energy, without a distinguishable change in the spectral shape, indicating homogeneous stoichiometry throughout the thickness. The 3-nm-thick film showed a slight deviation towards a lower binding energy. In this case, a higher density of defects was likely generated near the interface to release the misfit strain, as also evidenced by the transmission electron microscopy (TEM) results ( Fig. 1c in the paper). Figure S2. (a) Reciprocal space mapping around (113)YSZ. With a decrease in the film thickness, the Qx-value of (226)ITO is the same as that of (113)YSZ, indicating fully in-plane tensile-strained films below 8 nm. (b) X-ray photoemission spectroscopy (XPS) spectra of O-1s, Sn-3d, and In-3d. Regardless of the thickness, the peak position and shape are nearly the same over the range of 8-80 nm, indicating that the oxidation states do not significantly change. The 3-nm-thick film shows a peak shift towards lower binding energy, indicating the existence of defects near the interface. The numbers in the XPS spectrum for Sn-3d and In-3d indicate the oxidation states of Sn and In, respectively 3 .

Well-ordered atomic arrangement over a wide area and flat film surfaces
We observed a well-ordered atomic arrangement over an ~20-nm-narrow area in the cross-sectional TEM image shown in Fig. 1c. To support this excellent crystallinity, we took cross-sectional TEM images over an ~110-nm-wide area. Figure S3a shows the image for the 80nm-thick ITO epitaxial film. The atomic arrangement is well ordered, and there are few grain boundaries and dislocations in the region far from the ITO/YSZ interface. It should also be noted that the film surface is very flat, as confirmed by X-ray reflectivity (Fig. 1b). The root-meansquare surface roughness of the 80-nm-thick ITO film is 2.3 nm, as shown in

Metal-to-insulator transition of the 8-, 17-, and 30-nm-thick films
The 8-, 17-, and 30-nm-thick films showed a metal-to-insulator transition (MIT) upon cooling, as shown in Fig. S4. Similar thickness-dependent MIT behaviour has been observed in other materials, including nickelate ultrathin films of 2-3 unit cells 4 . To understand the MIT in ITO epitaxial films, we refer to three scenarios proposed in an earlier study 4 . First, the spatially directional d orbitals in Ni 3+ ions with a 3d 7 electron configuration are sensitive to strain, which may modify the bond length and oxygen octahedral rotation patterns and amplitudes. However, the overlap of spherically symmetrical In-5s orbitals, which are responsible for conduction in ITO, is minimally affected by lattice distortions. Second, quantum confinement is rarely a consideration for our sample, as this MIT behaviour occurs even in 30-nm-thick ITO films. A similar MIT was also observed in 300-nm-thick ITO polycrystalline films 5 . Therefore, another mechanism that is not relevant to strain or quantum effects plays an important role in the MIT of ITO films. Third, defects populate the ITO/YSZ interfaces to release the misfit strain, as evidenced by the TEM (Fig. 1c) and XPS (Fig. S2b) results, and the thermally activated carriers are frozen out at interfacial defects at low temperature. The transition temperature increases with decreasing film thickness (e.g., 100 K for the 30-and 17-nm film thicknesses and 200 K for the 8-nm film thickness), supporting that the MIT of ITO films may be caused by interfacial impurity scattering. ρ(400 K) denotes the resistivity at 400 K.

Thickness dependence of the transmittance and sheet resistance
We compared the transmittances at 2.5 μm of various thickness ITO, SrVO3, and CaVO3 films, as shown in Fig. S5a. The transmittance of sub-nanometre-thick ITO was 30% higher than those of SrVO3 and CaVO3 and approached 60% for tens-of-nanometre-thick films. The room temperature sheet resistance comparisons, as shown in Fig. S5b, indicated that ITO was sufficiently conductive, but more resistive than SrVO3 and CaVO3. Transmittance as a function of thickness. The transmittance of ITO (measured at the 2.5-μm wavelength) is higher than those of CaVO3 and SrVO3. (b) Sheet resistance as a function of thickness. The sheet resistance (measured at room temperature) of ITO is sufficiently low, except for that of the 3-nm-thick film. We digitized the data of SrVO3 and CaVO3 from reference [6].

Calculation of indirect and direct bandgaps
With the assumption of parabolic-like electronic bands, we determined the indirect bandgap and direct bandgap by extrapolation to zero of the linear portion of the sharp rise in the 1 2 ∝ (ℏ − ) and 2 ∝ (ℏ − ) curves 7 , respectively, where α, ℏ , and Eg denote the absorption coefficient, photon energy, and bandgap, respectively (Fig. S6).