1. Introduction

Transparent conductive oxide (TCOs) films have been extensively investigated for applications in solar cells, transistors, and flexible electronics [1]. TCO based photonic structures and devices have also been investigated over the past few years. Electrically gated structures utilizing TCOs as an active element were first proposed and demonstrated by Feigenbaum et al. [2]. The work demonstrated refractive index tunability via an accumulation of free charge carriers in a metal-oxide-conductor heterostructure incorporated with an indium tin oxide (ITO) layer. The most important points of the work are that the accumulation occurs within ~5 nm of the TCO – dielectric interface and that the dielectric spacer thickness must be minimized to maximize the tunability. The following few years saw the concept used in multiple theoretical investigations of electro-optic waveguide modulators but in the important telecommunication wavelength regime [3–10]. More comprehensive reviews of modulated TCO based structures can be found in Refs [11-12]. Other thin TCO-based applications investigated have included beam steering [11,13], non-linear optics [14-15], metamaterials [16], and perfect light absorbers [17–19] which we note may also have applications for light harvesting and thermal emitters. We note that Ref [18]. specifically has pushed TCO applications more into the mid-infrared, which is a focus also of this work. The above works primarily depended on Drude model, thick film based permittivity data for active TCO layers, or have other unknowns. Our aim is to comprehensively investigate ITO, as an example TCO, to provide the knowledge base for accurate photonic device designs.

The knowledge of the thickness-dependent optical constants of ultra-thin ITO films is important for designing new photonics devices. Previously, Losurdo et al. [20] performed spectroscopic ellipsometry measurement of ITO films of thicknesses ~90 nm. The novel part of that was the implementation of an index grading layer that accounted for changes observed near the substrate/film interface. This work however only extended up to wavelengths of ~800 nm leaving much of the infrared unexplored. Naik et al. investigated optical properties of conductive oxides that extended to wavelengths ~2 μm [21-22]. In terms of thickness dependence, they investigated gallium doped zinc oxide with film thicknesses that varied from near 300 nm down to 46 nm. In that case, 46 nm films showed a clear shift in the epsilon-near-zero (ENZ) wavelength near the important telecommunication range.

In this work, we investigate conductive ITO for deposited films with thicknesses between 10 nm and 100 nm using complimentary optical and electrical techniques. This includes determination of complex permittivity over the near to long-wave infrared wavelength range. We will show that a non-electrical or low mobility interface layer with thickness on the order of ~14 nm is present when ITO is deposited on silicon. On a similar thickness scale, the ITO films will also be shown to be thinner than the critical percolation threshold where instead of continuous films, isolated islands and structures are observed on the surface rending electrical measurements impossible [23–25]. The measured complex optical constants and electrical properties for the films can be used to model photonic structures throughout the infrared for the applications mentioned above. Perhaps more importantly, the analysis process used here can be repeated for any conductive oxides, or other plasmonic materials, and on any substrate, for the goal of achieving accurate properties that enables efficient design and modeling of photonic structures. Keys to such design are optical understanding, accounting for a non-electrical substrate/film interface layer, and ensuring avoidance of the percolation regime if the application requires continuous films.

2. Deposition of ultra-thin ITO films

The ITO films in this work were deposited via pulsed laser deposition (PLD) on single side polished 3 inch diameter (100) Si substrates which were p-type with specified resistivity > 5000 ohm-cm. Such high resistivities and correspondingly low free carrier concentrations of < 10¹⁴ cm⁻³ [26], were selected to avoid possible substrate influence on characterization. The Si substrates were cleaned and striped of native oxide using an HF solution and transferred to the PLD chamber. PLD was performed in a Neocera Pioneer 180 pulsed laser deposition system with a KrF excimer laser (Coherent COMPex Pro 110, λ = 248 nm, 10 ns pulse duration). The chamber base pressure was 2 x 10⁻⁵ Pa with a 5% O₂ /95% Ar background gas mixture introduced during deposition to a pressure of 1.3 Pa. Substrates were heated by a backside heater to 300°C and rotated during deposition. The laser operated at a pulse rate of 30 Hz and an energy density of 2.6 J/cm² was measured at the target. The target-to-substrate distance was 50 mm with a 45° laser angle of incidence to the target. The 90 wt. % In₂O₃ and 10 wt. % SnO₂ target was a 50 mm diameter by 6 mm thick sintered oxide ceramic disk (99.99% purity). The ITO films, which are inherently n-type, were deposited with target thicknesses of 5, 10, 15, 20, 50 and 100 nm. We refer to these targeted values as the nominal thicknesses.

X-ray photoelectron spectroscopy (XPS) with intermediate sputter-etch steps was used to confirm sample uniformity with respect to depth on the thickest and thinnest film. We confirmed that indium, tin, and oxygen were uniform throughout the film. XPS also confirmed that a ~1 nm layer of oxygen and carbon exists on the surface of both films that likely formed post deposition. No other measureable impurities were identified. The films also exhibit a concurrent decrease of oxygen, indium and tin as we neared the silicon substrates. This confirms that effects observed in this work are not caused by an abundance or lack of constituent materials near the substrate interface.

Deposited samples were etched with a 1:1 solution of HCl:H₂O for height measurements. Electron dispersive scattering (EDS) was used to confirm complete removal of the deposited film. Step heights were characterized via atomic force microscopy (AFM) and surface profilometry.

AFM was also utilized to characterize the deposited ITO film surfaces. Figure 1 shows AFM images from films with nominal thicknesses of 5, 10, 15 and 20 nm. The literature has cases where metal films close to the critical percolation thickness show a metal to insulator transition [23–25]. In the case of the nominal 5 nm film, the ITO particles are discontinuous and sparse, with lateral and height dimensions of ~30 nm, clearly being below the percolation thickness. The nominal 10 nm film gives indication that this thickness is still near the percolation regime, although it is noted that the particles are mostly connected with some observed gaps. The nominal 15 nm sample shows a definite coalescing, indicating that the thickness is now above the percolation thickness. RMS roughness for the films A-B in Fig. 1 are 4.2 nm, while for the C and D this decreases to 3.1 and 2 nm respectively. This decrease in RMS roughness is another indicator of the coalescing. These roughness values are negligible when compared with the optical wavelengths measured.

 

Fig. 1 AFM images for pulse laser deposited ITO films with nominal thicknesses of 5 nm (A), 10 nm (B), 15 nm (C), and 20 nm (D).

Download Full Size | PDF

X-ray diffraction (XRD) was completed in the θ-2θ configuration on the ITO films. The ITO diffraction peaks observed match those found in literature and indicate a polycrystalline film [27]. Crystalline ITO is present in even the thinnest film but it is likely that some of that film is amorphous. Analysis of the full-width half maximum and relative intensity of the (222) peak at 30.3° indicates that crystallite size and texture increase with thickness up to roughly ~20 nm, after which little change is observed.

3. Optical constants of ultra-thin ITO films

Spectroscopic ellipsometry was completed using both a J.A. Woollam IR-VASE and V-VASE which covers the visible to infrared wavelength regimes. Analysis of the data used the corresponding WVASE software to extract complex optical constants, as well as thickness and parameters for later comparison with electrical characterization. A separate highly-resistive single side polished Si substrate was also measured and used in the analysis software when building models for the ITO films.

Each respective metallurgical ITO film was treated as a single layer in the ellipsometry model. The nominal film thicknesses were used as a baseline, but were allowed to vary for best fit. The complex permittivity of the ITO included a summation of Drude and Gaussian oscillators as well as a Cauchy term. The Drude oscillator term is used to model the free carrier response in the ITO. The fitted Drude parameters included the resistivity, ρ, and the carrier relaxation time, τ. The Drude portion of the complex permittivity from the model was

Figure 2 presents the determined thicknesses for the films from the spectroscopic ellipsometry. Comparative data points are added from films that were etched to the substrate as confirmed via AFM and profilometry. For both the thick and thin films these measurements matched fairly well to the ellipsometry based thicknesses. In all cases, the films were thicker than the nominal thickness by 5-10 nm. It is noted that the data extracted from ellipsometry is a spatial average of length scales ~3 cm due to the large spot size of the systems. Thus local variations minimally impact the characterization. These fitted thickness values in Fig. 2 will be referred to for the remainder of this work simply as the thickness.

 

Fig. 2 Determined ITO film thicknesses compared with the nominal values.

Download Full Size | PDF

Figure 3 presents the determined complex permittivity for the ITO from 500 nm to 12 μm in wavelength. The near-infrared (NIR), mid-wave infrared (MWIR) and long-wave infrared (LWIR) bands are highlighted for reference. The data sets with the same targeted nominal thickness are colored the same. This figure presents a clear trend of the thinner films becoming less metallic, i.e. the large negative ${\epsilon }^{\prime }$ approaches zero or even becomes positive. The imaginary part of the electric permittivity ${{\epsilon }^{\prime }}^{\prime }$, trends towards zero for decreasing film thickness, although this is more prominent in the longer wavelengths in Fig. 3.

 

Fig. 3 Infrared complex permittivity of ITO films as determined by spectroscopic ellipsometry. The similar colored data sets are for similar targeted thicknesses.

Download Full Size | PDF

The thinnest films are near or below the percolation thickness, which results in islands rather than a uniform film as seen in Fig. 1. For these films, the determined thickness is really treating the structures as an effective medium which gave the best fit. Percolated films can be characterized by a decreasingly metal-like permittivity, not un-like what is seen in the Drude regime of MWIR to LWIR in Fig. 3. However, as discussed in Ref [24], as the films enter the percolation regime while being thinned, the permittivity goes from negative to positive maxima and then decreases to a smaller positive value, with the maxima occurring for films near the critical thickness. The films here do not show the same trend in ${\epsilon }^{\prime }$ as it is thinned beyond the critical percolation thickness even though. While the metal to insulator percolation based transition may be present, it is not the dominating factor in the changes observed here in the permittivity. The mechanism for the metal to insulator change will be discussed more in the following section.

We now investigate the meaning of these results with respect to plasmonic applications using the quality factors proposed by West et al. [28]. The quality factor for localized surface plasmon resonance, and surface-plasmon polariton excitation both take the form of ${\epsilon }^{\prime }/ {{\epsilon }^{\prime }}^{\prime }$ with the later application having the real part in the numerator squared. This analysis is one of many figures of merit to analyze potential for plasmonics using permittivity, although it may be the most generalized and direct. In the MWIR where ${{\epsilon }^{\prime }}^{\prime }$ of all but the thinnest film are similar, these quality factors are driven by ${\epsilon }^{\prime }$ in the numerator; i.e., the thicker films give a better quality plasmonic material. For λ > ~5 μm however, the best quality factors will not be driven by real part only due to the imaginary parts now diverging for films of different thicknesses. In that regime, quality factors will be a tradeoff between ${\epsilon }^{\prime }$ and ${{\epsilon }^{\prime }}^{\prime }$ that is application specific. For surface plasmon-polariton waveguide structures, specific wavelength ranges will give a tradeoff between mode confinement and propagation length like those in Ref [29]. Aside from the thickness dependent optical properties shown here, thickness also plays a role in properties of plasmonic waveguides such as mode confinement [30]. In design of such structures, one needs to carefully take into consideration the shape of the structure in addition to the thickness-dependent optical properties demonstrated here.

Figure 4 presents the complex permittivity in the NIR range from Fig. 3 for clarity. Figure 4 more clearly illustrates the wavelengths where ${\epsilon }^{\prime }$ = 0 for these films, which we define as the epsilon-near-zero (ENZ) wavelength, λ_ENZ. Recent works have utilized ENZ ITO in the NIR as means for tunable perfect absorbers [17–19] as noted earlier making this investigation useful to that community. In Fig. 4, λ_ENZ clearly red-shifts with decreasing thickness with the 10.8 nm film remaining entirely positive now. We will revisit λ_ENZ in a later section.

 

Fig. 4 Near-infrared complex permittivity of ITO films as determined by spectroscopic ellipsometry. The similar colored data sets are for similar targeted thicknesses.

Download Full Size | PDF

4. Electrical properties of ultra-thin ITO films

Electrical properties of the ITO films were characterized by two methods. The film sheet resistances were mapped using a Lehighton Electronics, Inc contactless rf-eddy current system. The contactless measurements determine sheet resistance R_s, which we use mainly for comparison purposes. This data set is referred to as “contactless resistivity” from here forward. Hall Effect measurements were completed in Van der Pauw configuration using an Accent Optical Technologies system. The Hall measurements for most of the films were completed on cleaved pieces with indium soldered contacts. For the 16.1 nm film (nominal 10 nm), gold contacts were patterned onto the corners of square-cut samples to avoid potential thermal annealing from the soldering process, which has been shown to impact specifically λ_ENZ [31-32]. The Hall and combined sheet resistance measurements determine R_s, carrier mobility μ, and sheet carrier concentration N_s.

The values from those independent measurements are compared also with electrical parameters determined from the ellipsometry fittings. Besides the extracted complex permittivity from the ellipsometry measurements, the individual fitted parameters of the Drude term are of interest and will be compared with the above independent measurements. The Drude term from the ellipsometry included the resistivity, ρ, and the carrier relaxation time, τ according to Eq. (1). From the ellipsometric ρ and τ, we can calculate the electron mobility μ, and free carrier concentration N via

We note that for the 10.8 nm film both Hall and contactless resistivity measurements resulted in values comparable only to those of silicon substrate so that data was removed. This is due to the discontinuous islands seen in the AFM (Fig. 1 A). The 16.1 nm film however contained semi-continuous densely packed islands (see Fig. 1 B), seemingly allowing electrical Hall measurements which will be mentioned again later.

Figure 5 presents R_s as a function of thickness from the three measurement techniques. Since ellipsometry treated each film as a single layer, the corresponding R_s, values were directly obtained by Eq. (4) using the fitted thicknesses from Fig. 2. We present R_s first because little is assumed and all three techniques are in good qualitative agreement. For thick films, R_s approaches a low limit of ~20 Ω/□. As the films become thinner, R_s rapidly increases.

 

Fig. 5 R_s, and N_s as a function of film thickness. The inset shows a magnified portion of N_s.

Download Full Size | PDF

Now we look at N_s as a function of thickness. The Hall measurements directly give this value, but ellipsometry requires calculation using Eqs. (3-4). The difficulty in calculating N (and also later μ) from ρ and τ lies in m* which has some uncertainty for ITO. Studies of ITO have shown effective masses in the range of 0.15 – 0.4 times the electron mass [33]. More recently, a study quantified the nonlinear dependence of m* on N [34] with the range being 0.35 – 0.53 times the electron mass. Looking into the literature, Refs [35-36]. explain how N is in most cases independent of thickness, beyond an initial interface region, even though it may seem that it is not true. You must first however renormalize the sheet values according to the “electrical thickness”, as opposed to the metallurgical thickness. At this point, we must assume constant N in our films similar to Ref [35], which is an assumption that will be revisited later. For now, constant N allows us to likewise use a constant m*, in agreement with Ref [36]. Since the Hall measurements do not use knowledge of m*, we fit the ellipsometry determined N_s (Eqs. (3-4) to those values via m*. We observed a clear trend of if m* was too large or too small, the ellipsometry N_s curve shape deviates significantly from that of the Hall curve. Our m* value was 0.38 ( ± 0.01) times the electron mass. We will also revisit this value for m* at a later point with regards to the non-parabolic formalism of Ref [36]. Due to this fitting of m* to match ellipsometry parameters to Hall measurements, the following analysis will rely primarily on the later.

Figure 5 shows qualitative agreement in N_s for thicknesses down to ~20 nm, albeit the ellipsometry used a fitted m* as discussed above. We completed Hall measurements on numerous pieces of the 16.1 nm film with results for N_s varying drastically. To rule out annealing effects via soldering of indium bonds, we photolithographically patterned ohmic contacts. The error bar in Fig. 5 still shows a wide range from different pieces of that 16.1 nm film. We cannot rule out annealing effects especially when taking into accounting the interconnect islands (Fig. 1(B)), but we are encouraged by the lower end of the error bar still overlapping with the ellipsometry data point. Due to this uncertainty, we will exclude this Hall data point for the forthcoming normalization process.

In addition to the error seen from the Hall for the 16.1 nm film, the ellipsometry gives a peculiar trend for this thicknesses at and below this value. N_s clearly has upward shift observed below the 16 nm range (see Fig. 5 inset). We speculate that percolation effects [24] may be responsible for this shift. However more comprehensive investigations are required to determine the properties near and below the percolation threshold which are beyond the scope of this paper.

We now will determine the electrical thickness d_e, which we use in determination of ρ and τ. The electrical thickness d_e, is less than the actual thickness d by a value δd, which is effectively a dielectric “dead” layer [36]. Linear fitting of N_s(d) and determining the x-intercept gives δd = 14.2 nm (indicated in Fig. 5). In our case, this layer is likely a mixture of ITO and In₂O₃ with the later becoming dominant closer to the substrate. This layer is also likely to include about 2 nm of interface SiO₂.

Figure 6 presents ρ, τ, N, and μ as a function of thickness. We calculate ρ and N by Eq. (4) using the values presented in Fig. 5 and the electrical thickness d_e = d-δd. On the ρ and N plots, a vertical dashed line was included to illustrate the thickness of the non-electrical interface layer δd. The ρ and N data points for films thinner than 20 nm showed trends that indicated the approach of the mathematical limit of dividing by zero (see Eq. (4). These data points were likely an unphysical artifact of this normalization method and removed for clarity. τ and μ however used no thickness in any calculation and thus all data points appear to have a physical meaning.

 

Fig. 6 ρ, t, N_s and μ as a function of film thickness. ρ, and N are determined using the electrical thickness of the films. The magenta dashed line on those curves indicates the thickness of the non-electrical dead layer.

Download Full Size | PDF

In Fig. 6, ρ follows the same trend of increasing with decreasing thickness as observed in R_s (Fig. 5). Importantly, all three techniques still qualitatively agree. The observed lower limit for thick films is ρ > 2 x 10⁻⁴ cm²/Vs. To understand the lower thickness limit of resistivity, we must look at τ and N as these are related via Eq. (4).

Before that we observe that N is indeed now constant when taking into account the electrical thickness as was observed in [35] for d > δd. Averaging over the data points for both the Hall and ellipsometry measurements, we arrive at nearly the same values. With Hall being the most direct measurement, we take the value of 7.9 ( ± 0.4) x 10²⁰ cm⁻³ to be the most accurate representation of the deposited ITO films. With this near constant value, N cannot be responsible for the low-thickness increase in ρ.

Looking at τ in Fig. 5, an upper limit appears to be nearly reached of ~8 x 10⁻¹⁵ s for thick films. As the films become thin, τ decreases. The reasoning for this is most likely scattering due to defects at the ITO/Si interface. Correspondingly, μ follows the same trend as τ, as they should via Eq. (3), reaching a maximum value of ~37 cm²/Vs for thick films. Typical values for μ in polycrystalline ITO films are 1-20 cm²/Vs [33]. It is noted that our grain sizes are on the nanoscale (~20-50 nm as seen in AFM); this indicates a mixture of amorphous and polycrystalline material in agreement with the XRD analysis and literature [37]. Annealing techniques can likely improve this mobility via reduction of amorphous content and correspondingly increasing grain size [37-38]. Our values show marginal improvement in mobility over the literature for thick films. We note that both the relaxation time and electron mobility for the thin films can likely be improved by the addition of a buffer layer [36]. The decreasing τ (and μ) with < 20 nm film thickness is directly correlated with the increasing resistivity.

We investigated the mobility dependence on thickness which directly is impacted by potential interface defects. Like that of Ref [36], we use

Lastly, we return to the assumption that we made earlier about m* being a constant = 0.38. Ref [34]. gives function of m* on N. However since we have now demonstrated a near constant N value for the electrically active portion of the film, the constant m* value is valid. For our value of N = 7.9 x 10²⁰ cm⁻³, m* = 0.45 according to their non-parabolic formulation. This is 18% larger than ours but with the wide values in the literature [33], this difference is minimal especially considering that these values are strongly deposition or growth dependent.

5. Epsilon-near-zero property of the ultra-thin ITO films

This section highlights the optical λ_ENZ (the wavelength where ${\epsilon }^{\prime }$ = 0) and compares the electrical and Drude based parameters as a function of film thickness. λ_ENZ is specifically relevant for modulators, perfect light absorbers, and other applications [13–19]. First we calculate the Drude parameters ω_p and Γ which can help to give an intuitive sense of the observed material changes with many articles in the literature reporting these values. These are calculated via

Figure 7 (upper) presents ω_p and Γ using the N determined from the electrical thickness (average from Hall in Fig. 6) and the $\mu$ function (Eq. (5) that fitted the Hall data respectively. Also plotted in Fig. 7 with these Drude parameters are $\mu$ values that were used in the Eq. (5) calculation. The constant N correspondingly gives ω_p that is independent of thickness. For our films, ω_p = 1.2 x 10¹⁵ rad/s. For a similar thick ITO film, Ref [30]. gave a value of 2.9 x 10¹⁵ rad/s. For a true comparison, we must scale their value by $\sqrt{{\epsilon }_{\infty }}$ since their fitting function did not include this factor. This brings their value to 1.5 x 10¹⁵ rad/s which is in agreement with ours with their slightly higher frequency agreeing with their slightly lower λ_ENZ (discussed next). For our thickest film, Γ = 0.13 x 10¹⁵ rad/s which is comparable also to the value of 0.18 x 10¹⁵ rad/s from Ref [39].

 

Fig. 7 (Upper) ω_p, Γ, and μ as a function of film thickness using the electrically and optically determined parameters. (Lower) λ_ENZ determined from ellipsometry as a function of thickness.

Download Full Size | PDF

Figure 7 (lower) presents λ_ENZ as a function of thickness for the films that have negative permittivities which were extracted directly from Fig. 4. The 104.4 nm thick film has λ_ENZ in the middle of the important telecommunication wavelength range. Upon decreasing thickness however, λ_ENZ red-shifts to approximately 2.1 μm, after which the thinner film has permittivity that remains positive. It is immediately apparent that this red-shift is correlated with increasing Γ and decreasing $\mu$. Re-visiting the Drude formulation of Eq. (8), one can determine that the ${\epsilon }^{\prime }$ = 0 condition is given by ${\omega }_{Drude-ENZ}=\sqrt{{\omega }_p^2-{\text{Γ}}^2}$. While the λ_ENZ data is not a direct relation to this condition due to the Gaussian and Cauchy terms used in the ellipsometry modeling, that correlation is corroborated. For this condition, as long as $\text{Γ}\ll {\omega }_p$, λ_ENZ is relatively unaffected by changes in $\text{Γ}$, or by $\mu$. As $\text{Γ}$ becomes comparable to ${\omega }_p$ however, λ_ENZ begins red-shifting which is observable in Fig. 7. This is specifically the case in our thin films where $\mu$ decreases with decreasingly film thickness, and those corresponding changes in $\text{Γ}$ and λ_ENZ are observed. Notably in Fig. 7, λ_ENZ reaches a maximum for our thin films which is defined more by a Gaussian absorption after the Drude term in the ellipsometry fitting gives comparatively small contributions. This is the case where thicknesses are on the order of or less than the percolation thickness of the film (See Fig. 1(A)-1(B)). In that thickness regime the optical constants are driven by both interface defects and dipole interaction in the observed islands [24].

Figure 7 shows the negative electric permittivity regime (“plasmonic”) and the positive electric permittivity regime. These regimes fall on the right- and left-side of our λ_ENZ curve respectively. As an example, if one were designing a telecom wavelength-range ITO structure that was only 50 nm thick, but using optical parameters determined from a much thicker film, discrepancies or failure could result if the non-electrical interface layer was not taken into account. It’s easy to see from Fig. 7 how the “plasmonic” region (or λ_ENZ depending on application) could be pushed out of the telecom wavelength region of interest in that case. Accounting for the non-electrical interface layer will be most crucial for applications in this NIR range where nanometer scale films are more heavily used in applications.

6. Summary

We have investigated optical and electrical properties of ultra-thin conductive ITO films on Si as a function thickness. The films were deposited via PLD on Si substrates held on a heated stage. XRD was used to determine that the films were predominantly crystalline. Optical characterization included visible to infrared spectroscopic ellipsometry where film thicknesses and complex permittivies were extracted. The complex electric permittivity data clearly shows decreasingly “metal-like” behavior in the infrared range of interest and shows a λ_ENZ that red-shifts from 1.5 to 2.1 μm with a film thickness decrease from ~100 to 10 nm. At first glance, one assumes this is due to a decreasing free carrier concentration, but ours was determined to be ~8 x 10²⁰ cm⁻³ for the films. Hall measurements showed a mobility that decreases from 35 cm²/Vs to single digits corresponding with the decreasing film thickness. This decreasing mobility is the dominant factor in the observed infrared optical property changes and is a product of a non-electrical dead layer that is ~14 nm thick. This layer is present for all the films investigated here regardless of thickness. For our depositions, films smaller than ~15 nm also entered the so-called “percolation regime” where the films became discontinuous islands. Both of these features are deposition and substrate dependent. The complex permittivity data presented here will be useful in designing and modeling ITO based structures for photonic applications, such as plasmonic modulators and perfect light absorbers for operation from the NIR to LWIR wavelength ranges. On a broader scale, the methods demonstrated here provide a blueprint for determination of thickness dependent optical and electrical properties that are useful for design and modeling of many photonic structures and devices.