Investigation of optical properties of aluminum-doped zinc oxide films via flow-limited field-injection electrostatic spraying

Aluminum-doped zinc oxide (AZO) is a transparent conducting oxide (TCO) of interest in plasmonic applications due to its higher mobility and, therefore, lower optical loss compared to other TCOs such as indium-doped tin oxide or gallium-doped ZnO.¹ The low optical loss helps prolong the lifetime of the surface plasmon modes and enhance device performance. In addition, AZO offers tunable optical response that can be controlled by the amount of doping.^2–5 

Previous studies, both theoretical and experimental, have indicated that a minimal resistivity in AZO is achieved by optimizing Al doping to maximize the free carrier concentration without degrading the crystallinity and reducing the mobility.^6–8 The electron scattering rate for DC conduction in AZO was reported to be dominated by grain-boundary scattering, prompting the growth of large grains.^9,10 However, this effect is not as relevant at high frequencies, thus rendering ionized-impurity scattering more important for optical systems.^11,12 Systematic studies on the optical response of AZO under differing Al stoichiometries may further elucidate the scattering mechanisms at optical frequencies.

A common obstacle in investigating the effect of film stoichiometry is the desorption of constituent elements during deposition, causing the stoichiometry to deviate from the expected value. Unlike vapor-phase methods, liquid-phase processing allows for homogeneous mixing of Al and Zn precursors dissolved in solution, enabling facile control of the Al-to-Zn ratio.^13–18 However, this may lead to film inhomogeneity due to the formation of Al clusters during deposition. These clusters will reduce dopant activation and increase optical losses by acting as large scattering sites.¹⁰ Sol–gel processed precursors, via chemical reactions rather than physical mixing, help prevent cluster formation to result in homogeneous distribution of dopants in the deposited material.^19,20

The advantages of liquid-phase precursors are, however, limited by the applicable deposition techniques such as dip-coating or spin-coating.²¹ These methods have poor control of film thickness and uniformity, particularly in conjunction with coating complex nano-structures that are required for photonic and plasmonic systems.^22–24 Conventional spray techniques, relying on micro-sized droplets, may be used as an alternative but are still not suitable for nano-structures. Flow-limited field-injection electrostatic spraying (FFESS) overcomes these limitations via controlled field-injection charging of the precursor solutions, subsequently generating a spray of charged uniform nano-sized droplets.^25,26 These nanodrops evenly cover the substrates due to the electrostatic repulsion,²⁷ and can coat nanostructures conformally,^28,29 producing highly pure and crystalline films at low deposition temperatures.^19,30 In this work, FFESS is employed to deposit AZO films using a sol–gel processed precursor to precisely control the film stoichiometry and to study its effect on the optical properties.

Using zinc acetate dihydrate, aluminum acetylacetonate, and methoxyacetic acid, a sol–gel processed AZO precursor solution in ethanol was prepared using nitric acid as a catalyst. FFESS deposition of AZO films with 0 at. %–5 at. % Al doping was subsequently performed on c-plane sapphire substrates at 350 °C in ambience. To deliver the same amount of solute for each deposition, the precursor solution flow-rate and film growth-time were fixed. Figure 1 shows scanning electron micrographs of a thus-prepared AZO film, exhibiting high density and uniform thickness.

To characterize the crystalline parameters of the films, a Bruker D8 x-ray diffractometer (XRD) was used for coupled 2-θ/ω scans using the Bragg–Brentano configuration. For parameter extraction, the GSAS-II software was used to obtain peak information and perform Rietveld refinements with augmentation from spherical harmonics to compensate for texture effects.³¹ 

Figure 2 shows XRD spectra of AZO films doped with 1 at. % and 4 at. % Al, indicating the formation of highly (002)-oriented ZnO. Previously, we reported FFESS deposition of highly oriented oxide films, including ZnO.^30,32 It is well known that crystallographic orientations of oxide films are affected by various factors, e.g., precursor chemistry and the processing conditions. Considering that the previous FFESS-deposited ZnO utilized a precursor without sol–gel processing, it is intriguing to observe again highly oriented ZnO in the present work. As the doping increases from 1 at. % to 4 at. % Al, the (101), (100), and (202) peak intensities decrease, calling for an investigation into crystalline texture as a function of the Al content.

Figure 3 compares the peak area contributing to the specific orientation to the total amount of the crystalline peak area,³³ 

The substrate peak is excluded in this calculation. This p_hkl value gives an approximate volume percent (vol. %) of the crystallites oriented in the (hkl) direction against the total crystalline material. The top plot of Fig. 3 demonstrates that the vol. % of the (002) peak tends to increase linearly with the amount of doping, whereas that of the (101) peak decreases. The bottom plot of Fig. 3 presents the vol. % of the lower-intensity peaks. The (100) vol. % reaches a minimum at 3 at. % Al, which is in line with the reported solubility limit of Al in bulk ZnO, i.e., 2.5 at. %–3 at. %.^34,35 The (110) vol. % also decreases with Al doping but, unlike the (100) peak, continues to decrease past 3 at. %. Notably, no zinc aluminate or aluminum phases appear at any doping concentrations. This confirms the ability of sol–gel processing to uniformly distribute the dopant throughout the lattice without causing clustering.

Figure 4 shows the crystalline strain in AZO, calculated from the textured Rietveld refinement, presented as the percent change in the lattice constant relative to the undoped measurement, (a(N_d) − a(0))/a(0), where N_d denotes the Al concentration. Error bars are a reflection of the fit accuracy and as such are dependent on peak intensities of the planes orthogonal to the said constant. In general, both the lattice constants a and c decrease from their undoped values, 3.251 Å and 5.205 Å, respectively, indicating compressive strain that is likely the result of the smaller ionic radius of Al over Zn. It was previously reported that below the solid solubility limit, Al occupies Zn substitutional sites, which would cause a reduction in the lattice constant.^6–8,36 Above the solid solubility limit, Al would begin to occupy interstitial sites, which would take up a higher volume and relax the lattice constants toward the undoped value. Here, the lattice constant a reaches a minimum at 2 at. %, which may indicate the onset of Al occupying interstitial sites. Meanwhile, the lattice constant c decreases further, reaching a minimum at 4 at. % Al. With c being the larger of the two constants, its maximal deformation is likely to occur at a higher Al concentration, which is observed. The added strain to the lattice may help prevent the formation of native defects in the lattice.

Figure 5 shows the rms crystallite size in AZO calculated from the Rietveld refinement, plotted against Al doping. The plot demonstrates a peak in the crystallite size at 4 at. % Al, which corresponds to the maximum rms lattice strain (⁠$a2+c2$⁠). This indicates that the higher compressive strain may discourage the misorientation of the crystallites, reducing the number of grain boundaries to increase the crystallite size.

For optical characterization, a Woollam variable-angle spectroscopic ellipsometer was used to measure the changes in polarization of the reflected light, Ψ and Δ, between 400 nm and 1700 nm using back-side roughened substrates. The WVASE software was then used to fit the measurements to film thickness and Drude relative dielectric permittivity,

where ε(∞) is the relative background permittivity, Γ is the Drude damping constant, and ω_p is the plasma frequency. This model with an intermix layer to represent the roughness led to mean-squared fit errors below 10° in Ψ and Δ. Error bars are included as 90% confidence intervals for parameters extracted from fits.

The measured plasma frequency, ω_p, was converted to an observed activated carrier density, n_e, using the relationship $ωp=e2ne/ε0m*$⁠, where e is the electron charge, ε₀ is the free-space permittivity, and m^* is the effective mass that is reported to be 0.36 times the electron rest mass.³⁷ This calculation ignores changes in the effective mass resulting from the non-parabolic band structure, which will lead to an underestimation of n_e for high values, but should not change the general trends.

Calculated n_e are plotted against the Al content in Fig. 6. The high n_e for the undoped ZnO is likely the result of substrate-to-film interface states due to the low film thicknesses, in the range of 100–150 nm, or native oxygen vacancies.^1,13 As the Al content increases to 1 at. %, n_e reaches a maximum due to the increased ionized impurities. Afterward, however, n_e begins to decrease, eventually settling below that of the undoped films. Therefore, it is likely that the number of native defects accordingly decreases as a result of the increased lattice strain as speculated in reference to Fig. 4. As previously discussed, Al begins to occupy interstitial sites above 2 at. %–3 at. %, which would decrease the number of ionized substitutional Al and further reduce carrier activation. The number of oxygen vacancies may have also been reduced due to the oxygen-rich deposition environment, decreasing the amount of activated carriers for all films. This warrants further investigation into FFESS deposition under inert ambience.

The Drude damping constant, Γ, is determined by the electron scattering rate, which controls the optical loss, Im(ε). This is also plotted against Al doping in Fig. 6. Notably, there is a significant drop in Γ reaching a minimum at 3 at. % Al. Since grain-boundary scattering is reduced at optical frequencies, the drop should indicate a reduction of ionized-impurity scattering. As discussed, the increased lattice strain may reduce the native defects, which contribute significantly to the ionized-impurity scattering. The previous work using magnetron sputtering reported a similar trend by increasing the oxygen supply to reduce the number of oxygen vacancies.¹¹ The reduction of Γ in the present work may indicate an avenue for strain-induced defect control using Al stoichiometry. Additionally, films grown without sol–gel processing have previously shown an increase in the electron scattering toward the solid-solubility limit due to the formation of Al clusters.¹⁰ However, in the present results, Γ is reduced, which may be a confirmation that the sol–gel processing promotes uniform distribution of the dopant in the material. The control of Γ demonstrated in this work is especially important for plasmonic applications where optical loss plays a key role in improving the device performance.

The relative background permittivity, ε(∞), is plotted against Al doping in Fig. 7, showing a steep decrease with doping and reaching a minimum at 4 at. % Al. This demonstrates a reduction in bound charge density with Al doping, which commonly indicates reduced film density. However, the growth rate in Fig. 7 shows a minimum at 3 at. % Al and the crystallite size shows a maximum at 4 at. %, reducing the density of grain boundaries. Therefore, it seems unlikely that the minimum in ε(∞) results entirely from atomic density. This may be attributed to an increase in the amount of interstitial Al as discussed in reference to Fig. 4. The interstitial Al would tend to remain unionized in the lattice, screening the dipole moment of optical phonons that contribute to ε(∞). No matter the explanation, this behavior is of interest in plasmonic applications due to the direct influence of ε(∞) on the crossover wavelength, i.e., the zero point in ε(ω).

The behavior of optical and crystallographical properties of AZO with respect to Al doping was investigated by ellipsometry and XRD 2-θ/ω scans with Rietveld refinement. The trend in Drude damping may indicate that the decrease in electron scattering is due to a reduction in the number of ionized impurities. The concentration of activated carriers showed a sharp decrease after reaching a maximum at 1 at. % Al, which may have resulted from a reduction in the number of defects. Finally, the background permittivity showed a minimum at 4 at. % Al, which may indicate the role of Al in reducing phonon polarization.

The present results may prove pertinent to plasmonic applications where a significantly reduced optical loss is required. Device tunability can be further enhanced by controlling the dielectric permittivity parameters and thereby the surface plasmon resonance conditions.³ FFESS deposition, due to its unique capabilities associated with liquid-phase processing, may help advance AZO as a plasmonic material.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

This work was financially supported by the U.S. Army CERL, Grant No. W9132T18C0010, and the Kim-Fund of the University of Illinois, Grant No. UIUC-933008-633134. R.V. gratefully acknowledges the Bardeen Fellowship and the ECE Distinguished Fellowship of the university. The materials characterization work was carried out in part in the Materials Research Laboratory Central Research Facilities, University of Illinois. The authors thank Dr. Mauro Sardela, Dr. Julio Soares, and Dr. Mohammed A. Ali for their assistance.