Transparent conductive oxides (TCOs) and their best-known representative – indium tin oxide (ITO), have been extensively used in various optoelectronic applications in both research and industry over the past few decades. ITO is a n-type wide-bandgap semiconductor and has emerged as one of the most popular TCOs due to its high optical transmittance in the visible and near-infrared regions along with its excellent electrical conductivity [1, 2]. It is a widely used material in everyday life devices ranging from solar cells and touch panels to organic light-emitting diodes [3–5]. In recent years, however, ITO and other TCOs have been rediscovered by the scientific community and are once again brought at the forefront of research. This is because in TCOs, carrier concentration and mobility levels can be adjusted in depth through doping and modifying deposition conditions. As a result, new fascinating and technologically important optical effects associated with free electrons in the material can be brought out. The ability to directly control the properties of the electron gas opened up new promises. Firstly, it is possible to engineer the epsilon-near-zero (ENZ) wavelength in TCOs, i.e., the wavelength at which the real part of the permittivity crosses zero. ENZ materials exhibit various exciting properties and can be used for field localization [6], nonlinear effect enhancement [7], dispersion management [8], phase-matching relaxation [9], ultrafast switching [10], frequency shifting [11, 12], polarization switching [13], high-harmonic generation [14] or metatronic devices [15]. Secondly, the carrier characteristics can be optimized in such a way that the optical losses in the system are reduced. Thus, TCOs have been perceived as alternative materials for plasmonic-based devices, offering an advantage over traditional noble metals [4, 16]. Finally, it has become feasible to electrostatically tune the local charge carrier concentration by applying external voltage. This is viable in configurations where the TCO layer is composed in a metal-oxide-semiconductor (MOS) geometry and has the proper initial carrier density. Such an active photonic structure that utilizes ITO was first proposed by Feigenbaum et al. [17] and exhibited the unique property of a unity-order local refractive index change as an effect of voltage-induced formation of the carrier accumulation layer. This concept has paved the way for various designs of tunable devices, being absorbers [18–20], plasmonic nanoantennas [21], metasurfaces [22] or silicon-based modulators [23, 24] to name a few.
The abovementioned potential of TCOs, and in particular ITO, to serve as key materials for future nanophotonic devices depends heavily on the ability to fabricate semiconductor layers with precisely engineered electrical and optical properties. Over the years, various techniques have been utilized to fabricate ITO films, including e-beam physical vapor deposition (e-PVD) [25, 26], sputtering [27], sol-gel process [28], chemical vapor deposition [29], or pulsed laser deposition [30]. The established position of ITO in modern, everyday optoelectronic devices might suggest that the manufacturing processes for ITO films are well known. However, the fabrication can become complicated when specific details are considered. Two aspects are worth discussing. Firstly, the quality of ITO is dependent on a number of process conditions, with the substrate temperature during the deposition phase being the most critical factor according to reported studies. To obtain high carrier concentration and optical transparency simultaneously, it is necessary to introduce high substrate temperatures, typically between 250 and 500°C, to ensure sufficient recrystallization of the evaporated material [2, 25, 26, 30–33]. As an alternative, it is also suggested to use post-process annealing at 550°C to produce samples of similar quality [15, 16, 34, 35]. Unfortunately, the introduction of elevated temperatures into the manufacturing process can damage other, more temperature-sensitive nanolayers or even entire components of the device in the case of many systems. For example, in the fabrication of flexible electronic devices or next-generation photovoltaics, which designs involve polymers, such conditions are unacceptable, as they cause thermal damage to the materials [36, 37]. The temperature constraints also limit the development of hybrid metal-semiconductor time-varying media, which could leverage the distinct characteristics of free electrons in both materials [38, 39]. The problem of having to use high temperature to fabricate high-quality ITO films has been identified in the literature, and there have been several attempts to resolve it. The research was focused on the optimization of the process conditions in either magnetron sputtering [40, 41], pulsed layer deposition or ion-assisted e-beam deposition [42, 43]. Nevertheless, the fabricated samples still showed reduced electrical conductivity compared to those obtained using substrate heating. Furthermore, the satisfactory result was only achieved for layers thicker than 200 nm, which might exceed the desired thickness for certain nanophotonic applications. This represents the second area of concern in ITO fabrication. Specifically, it relates to the rapid degradation of the electrical properties of ITO with decreasing film thickness [30, 44]. The effect is associated with the kinetics of the initial stages of nucleation and growth of the ITO layer and the resulting general morphology of the thin film, the presence of percolation, and the size of the individual grains [30]. This is a crucial factor because not only does the structure of the layer affect the value of the optical constants, but in more complex multilayer geometries, smooth interfaces between different materials are essential to avoid scattering effects that influence the overall performance of the device.
In this work, we address the above issues and present comprehensive studies on sub-100°C ion-assisted electron beam deposition of 70 nm thick ITO films with excellent electrical conductivity (resistivity of 5.2x10-4 Ωcm), high carrier concentration reaching 1.2x1021 cm-3, high transparency (85%) and low surface roughness (RMS < 1 nm). Our study is focused on the measurements of electrical and optical properties of the layer, as well as its surface morphology – a crucial factor in nanophotonic design. We provide other researchers with useful guidelines for the selection of e-beam evaporation process parameters such as deposition rate, substrate temperature, presence of oxygen gas and oxygen plasma. In particular, we show that the parameters of the oxygen plasma, namely the gas flow rate, the discharge voltage, and the discharge current, significantly affect the ITO layer properties. Finally, we demonstrate how the carrier concentration and mobility in the ITO film can be tuned to adjust the ENZ wavelength throughout the NIR spectral range.
Electrical and morphological properties of ITO films
ITO electrical properties arise from its complex crystal structure. A unit cell of ITO resembles that of indium oxide (In2O3), containing 80 atoms, and the indium cations are located in two different six-fold-coordinated sites [2]. In this degenerated semiconductor, the majority carriers are electrons that originate from the partial substitution of indium In3+ by tin Sn4+ ions and from the creation of doubly charged oxygen vacancies O2- in the lattice [2, 45]. These two mechanisms are strongly correlated with the substrate temperature used in the manufacturing process and the introduction of post-annealing treatment [35, 46]. Since ITO layers are typically prepared using physical vapor deposition techniques, the deposited films are polycrystalline and have significantly more complex carrier transport mechanisms than those found in single crystals. In such samples, the dominant electron scattering processes are associated with ionized impurities, grain barriers, and crystallographic defects, all of which limit electron mobility in thinner films surfaces and interfaces particularly, and thus reduce the conductivity with accordance to the formula [46, 47]:
,
where \(\rho\) is resistivity, \(N\) electron concentration, \(e\) electron charge and \(\mu\) mobility. When estimating electrical properties by different methods, such as Hall effect measurements and ellipsometry, it is also important to consider the discrepancy between optical and electrical thickness. On the basis of the series of samples with ITO film thicknesses ranging from 30 nm to 100 nm, we have determined this difference to be 13.8 nm, which is similar to the values stated in other works [30, 48].
As reported in the literature, different levels of ITO crystallinity can be achieved by selecting the PVD technique and controlling the deposition conditions. This results in films with different electrical and optical properties. Identifying the most favorable conditions for producing high-quality ITO films requires multi-parameter studies with countless number of combinations. The general concept behind our current work, however, is to develop a process that would allow for substituting indium atoms with tin and creating desired levels of oxygen vacancies without the need for high temperatures during the deposition process or post-annealing. Therefore, we start our research by only roughly determining the boundary conditions for the most important process parameters that will yield ITO layers with initially satisfactory properties, and then fine-tune the process by adjusting the oxygen gas or oxygen plasma parameters. Since thermal activation is a crucial factor in the crystallization process, our optimization procedure starts with identifying the temperature threshold below which the electrical properties of the samples rapidly deteriorate. In Fig. 1a, we present the measured resistivity values of 70 nm thick ITO films deposited at 1 Å/s rate in processes with different substrate temperatures. We observed that the conductivity of the samples improved as the substrate temperature increased. Although this is an expected general trend, the rate of improvement tends to slow down after reaching 80°C. The analysis of carrier concentration and mobility (Fig. 1b) provides some additional information regarding the mechanisms behind this relationship. The decrease in resistivity is mainly due to the increase in carrier mobility associated with the improvement in crystallinity [47]. At the same time, the opposite effect occurs, whereby the electron concentration decreases as the temperature of the substrate increases. This is a rather unexpected result, as the data reported in the literature suggest the opposite trend [26]. However, it should be kept in mind that our preliminary tests were carried out in an oxygen-free vacuum environment, where there is a tendency towards dendritic growth (see Fig. 1c), and the presence of free carriers is mainly due to doubly charged oxygen vacancies rather than Sn doping. Based on the described results, we decided that the temperature threshold should be set at the level of 80°C, corresponding to a resistivity of 1.4·10− 3 Ω cm (235 Ω/□). Such temperature is safe for the polymers and does not damage thin metallic nanolayers [49].
Next, we determined the optimum deposition rate, assuming the substrate is heated to 80°C. From Fig. 1d it can be seen that the films with the lowest resistivity were obtained at a deposition rate of 10 Å/s. We attribute this to two reasons. Firstly, increasing the deposition rate leads to higher kinetic energy of the atoms or molecules, which could improve the crystallization process [43]. Secondly, raising the temperature of the material in the crucible activates more Sn ions, which effectively increases the carrier concentration and improves conductivity [26]. A deposition rate of 10 Å/s was chosen for further investigation as deposition rates above this value would significantly reduce the vacuum in the chamber, deteriorating the sample quality.
In the subsequent studies, we decided to focus on the role of oxygen gas in the deposition process and its influence on the electrical properties of ITO. There are contradictory results published in the literature on this matter, with some papers suggesting a decrease in conductivity while others indicate the opposite [25, 42, 46, 50]. Therefore, we decided to make our series of customized processes with varying gas flows. In Fig. 2a and 2b, we present measured resistivity, carrier concentration and mobility of fabricated samples. The results indicate that as the oxygen flow rate increases, the electrical properties of the deposited films are adversely affected, and all three electrical parameters deteriorate. Several factors account for this. By incorporating oxygen atoms into ITO films, the number of oxygen vacancies, and thus, free carriers is reduced, leading to higher layer resistivity. Additional oxygen could also introduce defects and structural imperfections in the film [51] or alter the orientation of ITO nanocrystal growth, both of which may negatively impact carrier mobility. However, we believe that in the discussed case, the most important factor influencing the electrical properties is the general surface morphology. Judging from the SEM images (Fig. 2c and d), it occurs that for the given set of parameters (deposition rate of 10 Å/s and substrate temperature of 80 oC) the neutral oxygen atoms do not participate actively in the formation of ITO structure, for any oxygen flow rate. This lack of building material forces the growing crystal to maximize its surface-to-volume ratio, leading to the dendritic structure already mentioned in the previous paragraphs [52].
To address the insufficient quantity of oxygen atoms in the ITO structure, we decided to introduce a reactive form of oxygen into the process – oxygen plasma. We carried out two sets of depositions: (i) varying the discharge voltage, which affects the ion energy, and (ii) changing the discharge current, which influences the amount of gas converted to plasma, and thus, the number of ions reaching the sample. The oxygen flow was kept constant at 5 sccm in both series. The selected flow rate enabled us to cover a wide range of discharge currents and voltages. For the samples deposited at different discharge voltages, the discharge current was fixed at 0.4 A, whereas for the series with varying discharge current, the system was operated at a discharge voltage of 150 V. The ion bombardment was switched off immediately after the desired sample thickness was reached. In addition to the previously described results, the electrical parameters of the layers were also extracted from the ellipsometric measurements. This was possible due to a significant improvement in the optical and morphological properties in this set of samples.
The electrical measurement results show that the oxygen plasma has a significant impact on the conductivity of ITO layers, resulting in a reduction of the film resistance, from 1.8x10-3 Ωcm, when only pure oxygen gas was present, to less than 6.0x10-4 Ωcm for most of the discharge voltage values tested (Fig. 3a). The qualitative and quantitative agreement between the Hall and ellipsometric measurements proves the validity of the constructed ellipsometric model (see section 4). We found that increasing the discharge voltage leads to a decrease in carrier concentration and a simultaneous increase of their mobility (Fig. 3b). We attribute these changes to two main mechanisms. The first is related to the number of ionized scattering centers. The higher voltage allows for more efficient insertion of oxygen atoms into the ITO structure that reduces the number of oxygen vacancies, but also, as indicated by Yamaguchi, it leads to a drop in Sn4+ donor concentration [25]. In the end, there are fewer free electrons, but also fewer scattering centers. However, in layers deposited with a plasma discharge equal to 180 V, this trend breaks down because different growth conditions are favored at this voltage value. This results in visibly bigger ITO crystallites, as seen in Fig. 4e, but with some kind of structural disorder degrading the mobility. The second mechanism is related to the modification of surface morphology and the degree of crystallinity in the ITO layer. In contrast to the process with neutral oxygen, the usage of plasma has led to the formation of well-developed crystallites. As can be seen from the SEM images presented in Fig. 4, their size structure and separation between the grains depend strongly on the oxygen plasma discharge voltage, and the gradual increase in the voltage results in the increase of the grain size. This also boosts the carrier mobility in the case of some of the samples as lower density of grain boundaries reduces electron scattering. Compared to the samples deposited in the presence of neutral oxygen gas the overall mobility increased from 1–2 cm2/Vs to 10–15 cm2/Vs in the series with oxygen plasma.
In addition to improving conductivity, plasma treatment also results in a smoother surface (Fig. 4f) compared to the oxygen-only deposition process. In particular, we show that 70 nm thick ITO films deposited with oxygen ion assistance can achieve surface roughness RMS as low as 0.9 nm at 165 V discharge voltage. This is a comparable result to those presented in Refs. [35, 44, 46], given the significantly lower temperature of our process, preserved good electrical properties of ITO films and despite the smaller grain sizes in our samples. Lastly, it is worth noting that the use of oxygen plasma in the range of discharge voltages between 120 V and 165 V allows for low resistivity to be maintained, while providing a very wide range of carrier concentration and mobility at the same time. This is crucial for many nanophotonic applications, as it enables low power consumption and offers a choice between high concentration or high mobility depending on the device requirements [53].
In the second series of samples deposited with the assistance of oxygen plasma we varied the discharge current between 0.2 A to 0.8 A, whereas the discharge voltage was held constant at 150 V. The voltage value was chosen to ensure stable plasma conditions over the widest possible discharge current range. This time the measured resistivity values exhibited a well-like shape trend (Fig. 5a). Although, for most of the samples the resistivity value oscillated around 6x10-4 Ωcm again, the samples deposited with the extreme plasma discharge currents presented different electrical characteristics, with much higher resistivity values. The measured carrier concentration and mobility dependencies (Fig. 5b) and acquired SEM images (Fig. 6) indicate that the resultant resistivity depends on the subtle interplay between the carrier characteristic parameters and morphology of the sample. The general trend is that with the increase of discharge current value more amount of oxygen gas is converted to plasma, which leads to a reduction of the number of oxygen vacancies (thus decreased concentration) and larger crystallites. Except for two extreme scenarios, the increase in carrier mobility seems to balance the effect of reduced electron concentration. However, at a discharge current of 0.2 A, the high density of grain boundaries additionally reduces the mobility of the carriers, while the additional drop in carrier concentration at high ion intensity of 0.8 A may indicate that the strong ion flux also affects the amount of Sn sites or causes some other structural changes.
The lowest sheet resistance of 90 Ω/□ was obtained for layers deposited with the discharge current of 0.4 A, the same as used in the previous series of plasma-assisted samples. The influence of the discharge voltage value on the surface morphology can also be detected in the RMS value measured with the AFM microscope (Fig. 6f). Initially, a decrease in surface roughness is observed as the crystallite size increases. However, above a certain ion flux threshold (discharge current of 0.5 A), the separation between the ITO clusters starts to become larger, and consequently, the RMS value starts to rise again.
To summarize the results presented so far, it can generally be said that the use of oxygen plasma seems to be obligatory to obtain high-quality ITO layers in the e-beam process with a substrate temperature of 80°C and a deposition rate of 10 Å/s. The samples produced with the assistance of ions have the lowest resistivity, the smoothest surface, and at the same time, maintain a reasonable level of carrier mobility and concentration (Fig. 7). Both the adjustment of the plasma discharge voltage and the discharge current allow the properties of the ITO electron gas to be tuned, although the second approach permits this to be done over a wider range of values. Moreover, the use of oxygen plasma allows to omit the process of annealing, as our samples offer comparable or even better electrical properties than those subject to post-annealing treatment [2, 15, 35].
Optical properties of ITO films
Samples with the most promising electrical and morphological properties, i.e., those which were deposited in the presence of oxygen plasma, were selected for further characterization of the optical properties. Variable angle spectroscopic ellipsometry measurements performed over a wide spectral range have allowed us to reconstruct the permittivity dispersion curves of ITO films. The parameters of the oxygen ions flux have a significant effect on the optical constants, both for samples obtained at variable discharge voltage and at different discharge currents (Fig. 8). This is expected because the optical response of ITO layers in the NIR spectral range originates mainly from the properties of the ITO electron gas and, as we have shown in the previous section, both carrier concentration and mobility are affected by the oxygen plasma. The general observation is that higher discharge voltages and currents yield lower values of the imaginary part of the permittivity (Figs. 8a and 8b, respectively), which can be linked with the reduced carrier concentrations in these samples. At the same time, the values of the real part of the permittivity increase, which is consistent with the Kramers-Kronig relations, and indicate that the samples have become less metallic (see detailed discussion in the next section).
The transmittance measurements, presented in Fig. 9, further confirm the correctness of the retrieved permittivity curves. The samples with the lowest values of the imaginary part of the dielectric constant are also the most transparent ones. In particular, the transmittance can be as high as 85% in the VIS/NIR range, depending on the choice of plasma parameters. The transmittance of the layer deposited at a plasma discharge current of 0.2 A is less than 60%, fundamentally different from the others in this series. The uniqueness of this sample was already recognized in the previous paragraph and is caused by insufficient ionization of the oxygen gas.