Indium-cadmium-oxide films having exceptional electrical conductivity and optical transparency: Clues for optimizing transparent conductors

Abstract

Materials with high electrical conductivity and optical transparency are needed for future flat panel display, solar energy, and other opto-electronic technologies. In_xCd_1-xO films having a simple cubic microstructure have been grown on amorphous glass substrates by a straightforward chemical vapor deposition process. The x = 0.05 film conductivity of 17,000 S/cm, carrier mobility of 70 cm²/Vs, and visible region optical transparency window considerably exceed the corresponding parameters for commercial indium-tin oxide. Ab initio electronic structure calculations reveal small conduction electron effective masses, a dramatic shift of the CdO band gap with doping, and a conduction band hybridization gap caused by extensive Cd 5s + In 5s mixing.

Substances in thin film form having high electrical conductivity and high optical transparency will be critical for next-generation flat-panel displays, photovoltaic cells, organic light-emitting diodes (LEDs), energy-efficient windows, and other opto-electronic technologies (1, 2). Although transparent conducting oxides (TCOs) such as Sn-doped In₂O₃ (ITO) and F-doped SnO₂ are used extensively (3), their electrical and optical characteristics are inadequate for future applications (1), and fundamental scientific understanding of their properties has been hindered by their exceedingly complex microstructures. Recent advances in the TCO field include new ZnO conductors (refs. 4 and 5; ref. 6 and references therein), high carrier mobility Cd₂SnO₄ (7), new ternary and layered compounds (8, 9), p-type conductors (10), detailed analyses of phase relationships (ref. 11 and references therein), and detailed criteria for materials selection (12). Nevertheless, fundamental understanding of the constraints on maximum mobility and transparency imposed by crystal and electronic structure, film microstructure, and doping level is lacking. We report here a combined film growth, microstructure/charge transport/optical characterization, and first-principles electronic structure analysis of an In_xCd_1-xO (13, 14)‡‡ thin-film TCO family having simple crystal structures, exceptionally high electrical conductivity (≈2–5 × commercial ITO), and excellent optical transparency. The results reported here provide insight into criteria for optimizing TCO film properties.

For growth of metal oxide films, metal-organic chemical vapor deposition (MOCVD) offers the attraction of simple apparatus, conformal coverage, growth under high gaseous partial pressures, and adaptability to large-scale/large-area depositions (15). In the present case, it was necessary to develop the first easily handled Cd MOCVD precursor exhibiting high volatility and thermal stability, while introducing negligible carbon deposits. The synthesis of Cd(hfa)₂(TMEDA) (hfa, hexafluoroacetylacetonate; TMEDA, tetramethylethylenediamine; Fig. 1A) is described in ref. 16 and references therein. In combination with In(dpm)₃ (dpm, dipivaloylmethanate; Fig. 1B; ref. 5) in a low-pressure reactor (ref. 17 and references therein; 2 Torr, 360°C substrate temperature, Ar carrier gas, H₂O-saturated O₂ reactant gas), a series of In_xCd_1-xO films (18††), where x was systematically varied, was grown on smooth 1.25 cm × 0.50 cm glass substrates (float glass; growth rate ≈2.5 nm/min; film composition established by inductively coupled plasma spectrometry). Typical film thickness was 0.15 μm.

Figure 1.

Molecular structures of the metal-organic film growth precursors (A) Cd(hfa)₂(TMEDA) is bis(hexafluoroacetylacetonato)(tetramethylethylenediamine)cadmium(II). (B) In(dpm)₃ is tris(dipivaloylmethanato)indium(III). (C) θ-2θ x-ray diffraction scans of In_xCd_1-xO films grown on glass substrates by metal-organic chemical vapor deposition at 360°C. The unit cell dimensions contract linearly with In doping level until x ≈ 0.11.

X-ray diffraction reveals that for x ≤ 0.11, the as-grown In_xCd_1-xO films are phase-pure, crystalline, and highly [100] textured, with features assignable to a cubic CdO-type (NaCl) crystal structure (Fig. 1C). The lattice parameters contract monotonically with increasing x (indium content), with the reflections becoming broad for x ≥ 0.11, and with In₂O₃ features now evident. The nanoscale structure of the films in cross-sectional and plan-view was investigated by using cold field emission gun transmission electron microscopy. For the most conductive (vide infra) x = 0.05 film, the images clearly reveal pronounced [100] texture of the columnar, submicron (≈300 nm) grains, whereas electron diffraction confirms the cubic structure (Fig. 2). X-ray microchemical sampling evidences uniform presence of In throughout the film, with no evidence for significant grain boundary segregation, indicating a full solid solution of In in CdO. No other elements are detectable, with F⁻ below the EDS (energy dispersive x-ray spectroscopy) and EELS (electron energy loss spectroscopy) detection limits of ≈1.2 wt %. Cross-sectional imaging indicates atomically abrupt substrate–film interfaces with no evidence for chemical reaction or solute segregation. Atomic force microscopy also indicates that the films are continuous and smooth with rms roughness of ≈10 nm over a 4 μm² area.

Figure 2.

Nanoscopic images of the In_xCd_1-xO film microstructure for x = 0.05. (A) Plan-view transmission electron micrograph showing sub-micron columnar grains. (B) Electron diffraction pattern showing cubic crystal structure.

Charge transport measurements used instrumentation and techniques described elsewhere (19); reproducibility in conductivity measurements was ±1% on the same sample, ±5% from sample-to-sample. Data are compiled in Table 1. For x = 0.0, the four-probe electrical conductivity (3,000 S/cm; n-type) is approximately comparable to that of commercial ITO, whereas the charge carrier mobility of ≈150 cm²/Vs rivals or exceeds that of known TCO films, including that of CdO films grown by other deposition methods (ref. 20 and references therein). We speculate that the present rather “soft” growth technique provides higher O₂ partial pressures as well as conditions for greater film microstructural orientation and crystallinity. As the In content is increased, the conductivity increases dramatically (Fig. 3A) as does n-type carrier concentration (Fig. 3B), whereas the mobility declines moderately (Fig. 3C). The highest conductivity is achieved at x = 0.05, where σ = 16,800 S/cm with an n-type carrier concentration (n) of 1.5 × 10²¹ cm⁻³ and a Hall mobility (μ) of 70 cm²/Vs. These parameters can be compared with those for films of the recently reported, high-mobility spinel TCO, Cd₂SnO₄, where σ = 8,300 S/cm, n = 9.0 × 10²⁰ cm⁻³, and μ = 60 cm²/Vs, and to those of the high In-content spinel TCO, CdIn₂O₄, where σ = 4,300 S/cm, n = 6.1 × 10²¹ cm⁻³, and μ = 44 cm²/Vs (7). In the present case, increasing x beyond 0.05 results in declining conductivity and mobility, whereas the carrier concentration plateaus around x = 0.07 and then gradually declines (Fig. 3B). Variable-temperature data (Fig. 3D) reveal “metal-like” (∂σ/∂T < 0) behavior at low temperatures. Although carrier scattering mechanisms are not completely understood in polycrystalline TCO films (1), decreasing mobility with increasing carrier concentration combined with the modest conductivity increase with falling temperature argues for a predominant contribution from ionized impurity scattering, with a subordinate electron-phonon scattering role (21). Optical transmission spectra (Fig. 4A) reveal high transmittance and a widening visible transparency window with increasing In doping. Estimating the band gap from the absorption edge yields data (Fig. 4B) consistent with a classical Burstein-Moss (B-M) shift on conduction band filling (3), likely offset partially by gap shrinkage effects (22). From this analysis, the band gap of the x = 0.05 phase is ≈3.1 eV—rivaling or exceeding reported values for commercial ITO (3.0–3.7 eV).

Table 1.

Charge transport characteristics of In_xCd_1−xO films

Comp, x	Conductivity, S/cm	Carrier concentration, 1021 cm−3	Mobility, cm2/V⋅s
0.00	3,560	0.152	146
0.02	9,370	0.640	91.5
0.03	15,600	1.26	76.9
0.04	12,300	1.27	60.5
0.05	16,800	1.51	69.2
0.09	9,010	1.65	34.0
0.11	6,740	1.06	39.6

All data at 25°C. 

Figure 3.

Charge transport properties as a function of In doping level (x) for a series of In_xCd_1-xO films having 0.15 μm thickness. (A) Electrical conductivity at 300 K. (B) Charge carrier concentration at 300 K. (C) Charge carrier Hall mobility at 300 K. (D) Electrical conductivity as a function of temperature.

Figure 4.

Optical properties as a function of In doping content for a series of In_xCd_1-xO films. (A) Transmittance as a function of incident wavelength. (B) Derivation of the apparent optical band gaps.

Electronic band structure analysis used the precise FLAPW (full-potential linearized augmented plane wave) formalism (ref. 23 and references therein) with exchange-correlation energies treated via the local density approximation (LDA). Furthermore, the screened exchange LDA (sX-LDA) method was used for the first time in a TCO calculation to more accurately evaluate (24) band structure and optical properties. In addition to CdO, computations were performed for x = 0.031, 0.063, and 0.125 using 64-, 32-, and 16-atom supercells, respectively, and CdO atomic coordinates. The sX-LDA-derived CdO band structure (Fig. 5A) features a broad Cd 5s conduction band and O 2p-based valence band. The calculated direct gap at Γ of 2.36 eV (Table 2) compares well with experiment (2.28 eV; ref. 25). The carrier effective masses obtained from the conduction band curvature around Γ are 0.240 m_e, 0.247 m_e, and 0.259 m_e in the [100] [111], and [110] directions, respectively, also in excellent agreement with experiment (0.27 m_e; ref. 26). These rather small values contrast with 0.30–0.42 m_e calculated for In₂O₃ by LDA techniques (27) and suggest that CdO can function as a high mobility TCO host. Indium doping dramatically alters the CdO band structure (Fig. 5B) with a large Fermi energy displacement and B-M shift due to added In dopant electrons, which broadens the optical transparency window (similar effects are observed in LDA calculations on ITO; ref. 28). The present results also reveal extensive mixing of In 5s and Cd 5s states, yielding a hybridization gap (E) in the conduction bands, which further lowers optical absorption by weakening intraband transitions, and may reduce electron mobility at E_F in heavily doped samples. Fig. 5C shows the wave vector k dependence of the conduction band energies as a function of doping. To obtain the optical band gap E_g (k_F), which determines transparency in the visible, we calculated the optical transition matrices at k_F and evaluated the edges of the interband transition energies. Next, E_g (k = 0) for x = 0.031 and 0.063 was evaluated by linearly interpolating the correction to the LDA band gap with respect to x, using sX-LDA data (1.33 eV for x = 0; 1.11 eV for x = 0.125). Table 2 summarizes results for E_g (k = 0) and E_g(k_F) as a function of x. The minimum computed band gaps (in the [110] direction) are in excellent agreement with experiment (Fig. 5C), as is the band gap shrinkage of E_g at k = 0, which arises from many-body effects (22, 29) and counterbalances B-M shifting at high doping levels. Fig. 5C also shows a significant discrepancy from the rigid-band model neglecting any band gap shrinkage but fixing E_g (k = 0) and the effective mass at the experimental values.

Figure 5.

First-principles electronic structure calculations at the screened exchange-local density approximation level (sX-LDA) for In_xCd_1-xO. (A) Band structure of CdO with the direct band gap indicated. The origin of the energy is taken at the bottom of the conduction band. (B) Band structure of heavily doped In_xCd_1-xO, where x = 0.125. The origin of the energy is taken at the Fermi level. Dotted lines indicate states where mixing of Cd 5s with In 5s states is extensive; the direct gaps at k = 0 and k = k_F and the hybridization gap are indicted. (C) Calculated band gap energies of In_xCd_1-xO as a function of k_F (open symbols). The experimental band gaps using k_F = (3π²n)^1/3 (filled circles) and the rigid-band model, E_g(k_F) = E_g(0) + (h²/2m*)k_F², where E_g(0) = 2.28 eV and m* = 0.27 m_e from experiment, are also shown for comparison.

Table 2.

Computed Fermi wave vectors, k_F in atomic units, and band gaps, E_g in eV, for In_xCd_1−xO in the [100](Δ), [111](Λ), and [110](Σ) directions

	x = 0.0	x = 0.031	x = 0.063	x = 0.125
k	0.00	0.158	0.189	0.251
k	0.00	0.173	0.216	0.284
k	0.00	0.168	0.200	0.278
kF(n)	0.00	0.174	0.220	0.277
Eg (k = 0)	2.36	2.27	2.13	1.93
E(kF)	2.36	3.73	3.97	4.46
E(kF)	2.36	3.31	3.36	3.50
E(kF)	2.36	3.12	3.17	3.19

Fermi wave vectors in the free electron model k_F(n) = (3π²n)^1/3, where n = electron carrier density, are also listed. 

For achieving high oxide conductivity and transparency, the present results are instructive, and in building on earlier qualitative models (30), point the way for design and discovery of new TCOs, both with and without Cd. The key to high In_xCd_1-xO carrier mobility is the broadly dispersed, free electron-like s-like conduction band, consistent with a large hopping integral between neighboring sites (28). This broad conduction band and the low carrier effective masses doubtless reflect the close metal–metal contacts possible in the cubic crystal structure where Cd-Cd = 3.320 Å in CdO (six-coordinate Cd⁺² ionic radius = 1.09 Å), vs. closest In-In = 3.345 and 3.360 Å in In₂O₃ (six-coordinate In⁺³ ionic radius = 0.94 Å). Not only should an s-like conduction band promote uniform electronic charge density distribution, hence relatively low scattering, but the simple, isotropic cubic crystal structure should, a priori, afford fewer ionized and neutral scattering centers than more complex ITO-type structures (4, 21, 26, 31) having nonequivalent metal and O sites plus several types of vacancies and disorder. For enhancing optical transparency in the visible, the large doping-induced B-M shift more than compensates for gap shrinkage effects. In addition, doping splits the highly dispersed conduction band, and the resulting hybridization gap contributes to lower intraband optical absorption. Finally, to the extent that grain boundaries make some finite contribution to scattering (21), an isotropic cubic microstructure should help to minimize high-angle effects.

Abbreviations

LDA

local density approximation

TCO

transparent conducting oxide

ITO

Sn-doped In₂O₃