Comparing metal oxide thin films as transparent p-type conductive electrodes

The development of transparent and p-type conductive layers remains a challenge to achieve more efficient hole collection and to combine with the most common n-type counterparts into transparent p–n junctions. Here, several candidates based on abundant materials: Cu2O, NiO and SnO have been prepared, characterized and comparatively evaluated. Thin-film deposition methods (evaporation and sputtering) have been used along with thermal treatments (oxidation and sulfurization) to maximize the transmittance and conductivity for each material. The highest quality is achieved by Cux(S, O) layers prepared by sulfurization of Cu2O at 250 °C. Besides, the NiO films obtained by reactive sputtering at room temperature have a good quality to be applied on heat-sensitive substrates.


Introduction
Most conducting materials are opaque to visible light, while most wide band gap materials are electrically insulating. However, a class of oxide semiconductors named as transparent conductive oxides (TCOs) constitute an exception to this rule, because they are rather good conductors of electricity and simultaneously transparent in the visible region [1]. TCO materials in thin films are widely used in a variety of optoelectronic devices ranging from flat panel displays to smart windows or photovoltaic solar cells [2,3], and other advanced photo-sensors [4]. For these applications, transparent metal oxides with high n-type conductivity are commonly used, such as Sn-doped In 2 O 3 , Al-doped ZnO or Sb-doped SnO 2 [2]. However, the development of high performance p-type TCOs remains a challenge to achieve more efficient hole collection and also to combine with the common n-type TCOs into transparent p-n junctions with active functionalities [5][6][7].
For the best design of p-type TCOs, both experimental and theoretical data about the electronic structure of metal oxides should be taken into account [7][8][9]. In most metal oxides, the minimum of the conduction band (CBM) is formed by metallic s orbitals that are spatially extended. Therefore, the electrons in the metal oxides have a small effective mass and high electron conduction can be achieved by an easy doping with donor impurities, resulting in n-type TCOs. On the other hand, the maximum of the valence band (VBM) is formed by oxygen 2p orbitals, which are quite localized, giving rise to a large effective mass for holes that results in very low mobilities for these charge carriers. Furthermore, the dispersion of the valence band tends to be small, and the VBM level is so deep that p-type doping is difficult [9].
Metals with d orbitals close to the O 2p can be used to favor the formation of highly hybridized orbitals that raise the VBM level and facilitate doping by holes. In this sense, Cu 2 O is a good candidate because the 3d 10 configuration of Cu + is very close to the O 2p 6 . Besides, it is intrinsically copper-deficient due to formation of copper vacancies V Cu that act as a shallow and efficient hole producer [10]. The 3d 8 configuration of Ni 2+ is also suitable for obtaining p-type NiO, with an easy hole dopability by cation deficiency [11]. An interesting alternative is tin monoxide, because the 4d 10 5s 2 configuration of Sn 2+ allows hybridization of the cation s states with the O 2p orbitals. Since the s orbitals are more delocalized than the d orbitals, this hybridization with s orbitals can lead to a greater dispersion of the valence band and therefore lower effective mass for holes than in the case of the Cu 2 O, where the hybridization is performed with d orbitals [12,13]. The calculations made on a large set of compounds indicate that the hybridization of cation s orbital with O 2p is more pronounced for the Sn 2+ , followed by Pb 2+ and finally Tl 1+ , which has the less dispersive valence band [8].
Otherwise, it has been proposed to add an anion of the type S 2− or P 3− because these have 3p orbitals more delocalized than the O 2p and can modify the valence band so that the holes effective mass decreases [7,8]. Since the pure metal sulfides or selenides have band-gap energies lower than the corresponding oxides, a better option is the formation of oxychalcogenides, in order to decrease the effective mass of holes and maintain high values for the energy gap that allow achieve a high transparency [14]. Although both oxygen and chalcogens (S or Se) are group VI elements, the sizes of their ions are considerably different. Hence, oxygen and chalcogens do not form a solid solution in crystals and are generally spatially separated, forming a layered structure [12].
For large-scale device applications, it is now of the utmost importance to reduce the cost of the materials and preparation processes. Hence, the research is directed towards abundant materials and with low environmental impact to reduce costs in handling and security. In addition, the substitution of conventional glasses by plastic substrates, cheaper and less fragile or heavy, can significantly lower final prices and enhance the industrial viability of the preparation processes. Especially interesting is the development of thin film materials at room temperature by continuous or monolithic processes [15]. For this purpose, both evaporation and sputtering systems were built and installed in our lab, which allow the deposition of thin films of metals and metal oxides on substrates up to 30×30 cm 2 [16,17].
In the present work, evaporation and sputtering processes have been developed to obtain different p-type TCO films based on abundant materials. Specifically, Cu 2 O, NiO and SnO oxides have been synthesized, characterized and evaluated comparatively. The deposition processes have been adjusted to achieve optimum transmittance and conductivity for each material, but also minimizing the working temperature because this means greater economy and larger application on heat-sensitive substrates. Besides, the effect of adding S 2− anions has been investigated by the formation of Cu x (S, O) layers by three different ways: (a) oxidation of CuS, (b) sulfurization of Cu 2 O and c) direct sulfoxidation of metallic copper.

Methods
Evaporation and magnetron sputtering systems were used to obtain thin films of Cu 2 O, Cu x (S, O), NiO and SnO on glass substrates, as grown at room temperature or after heat treatment performed under a controlled atmosphere.
Copper layers were prepared by evaporation and annealed at different temperatures (between 200°C and 500°C) in a tubular quartz furnace, always with flowing N 2 (1.5 l min −1 ) and adding O 2 flow (0.3 l min −1 ) for the oxidation process or elemental S (placed next to the sample in a quartz holder) for the sulfurization. Both flowing O 2 and elemental S were introduced concurrently in the process of sulfoxidation.
Besides, nickel oxide and tin oxide layers were prepared by reactive DC magnetron sputtering from the respective metallic target. The magnetism of pure Ni target involves technical problems that were avoided by using a typical non-magnetic alloy of 92%Ni:8%V [15]. After evacuation of the sputtering chamber to a base pressure of 4×10 −4 Pa, oxygen and argon gases were introduced by independent mass flow controllers to adjust the oxygen partial pressure which determines the film stoichiometry. In order to improve the crystallinity of the sputtered samples, post-deposition heat treatments were performed in air at temperatures ranging from 200°C to 500°C for 30 min.
The structural properties of the layers were examined by x-ray diffraction (XRD) using Cu K α1 (λ=1.54056 Å) radiation in a Philips X'pert instrument with Bragg-Brentano θ−2θ configuration. Crystalline phases were identified by comparison of the measured diffraction peaks with the Joint Committee of Powder Diffraction Standards (JCPDS) files. The mean crystallite size was estimated using the Scherrer formula based on the full width at half-maximum of the main diffraction peak. The optical properties were determined in the 300-1500 nm wavelength range with a double beam spectrophotometer Perkin-Elmer Lambda 9, which has an integrating sphere to measure the total transmittance (T t ) and reflectance (R t ). In order to eliminate the interference peaks, the transmittance is corrected for reflection losses: Then, the optical absorption coefficient is determined as: % , including the film thickness value (t) measured by profilometry. Here, a same thickness of about 0.1 μm has been used to compare the different samples. The band-gap energy (E g ) is calculated according to direct allowed transitions: The electrical conductivity, carrier concentration and mobility were determined with an ECOPIA Hall Measurement System. Before electrical measurements, the samples were cut into square pieces of 1×1 cm 2 and placed on a sample holder with four gold-plated spring loaded pins. Conductive Ag paste is added to improve pin contact to the layer. The conductivity (σ) depends on the sheet resistance (R S ) and the film thickness as: s Electrical measurements were made one or two weeks after film deposition, and no significant changes were detected by environmental exposure in the interval.

Cu 2 O thin films obtained by oxidation of evaporated Cu layers
The crystalline structure of copper oxide films is found dependent on the heating temperature, as illustrated by the XRD patterns in figure 1. Cubic Cu 2 O with main (111) diffraction (JCPDS no. 05-0667) appears in the form of small crystallites (14 nm diameter) after heating the samples at 250°C, below the temperature expected from the equilibrium diagram [18]. The increment of the heating temperature to 350°C produces transition to monoclinic CuO (JCPDS no. 89-5899) with a higher crystallite size (16 nm diameter), while a mixture of Cu 2 O and CuO phases is obtained at 300°C. The formation of Cu 2 O at low temperatures is attributed to the surface energy of small particles, which promotes the coalescence of nanocrystallites with high symmetry (cubic). Transition to larger CuO crystallites with a lower symmetry (monoclinic) occurs at superior temperatures, involving a sharp decrease in the normalized cell volume [19].
The electrical conductivity decreases from 1.7×10 −2 S cm −1 for the pure Cu 2 O sample to 2.0×10 −4 S cm −1 for the CuO film, being p-type in both compounds and dominated by the carrier concentration, which decreases from 10 16 cm −3 to 10 14 cm −3 ranges. First-principles calculations have stated that copper oxides are p-type semiconductors due to Cu vacancies (V Cu ), because they are energetically favored and induce shallow acceptor states at the top of the valence band [20,21]. Otherwise, copper interstitials (Cu i ) and oxygen vacancies (V O ) create donor levels within the band gap, acting as hole-killer defects [10,21]. The activation energy for such recombination levels is found lower in CuO than Cu 2 O [19], which explains the decrease in free hole densities when CuO crystallizes with increasing temperature. Both Cu 2 O and CuO samples show transparency above 85% in the near infrared region (figure 2). In the visible range, the optical absorption is lower for Cu 2 O (with E g =2.55 eV) than for CuO (E g =2.20 eV), according to other works [22,23]. Thus, high visible transmittance and acceptable electrical conductance are concurrently achieved with single-phase Cu 2 O layers prepared at 250°C.  figure 3. Mean crystallite sizes of 33 nm are obtained for these films grown at room temperature, with a lattice constant a=0.421 nm that is higher than expected from the standard powder file (0.418 nm) [24]. Such enlargement in the lattice can be attributed to the presence of nickel vacancies (V Ni ) and concomitant formation of Ni 3+ species. This is because the coexistence of Ni 2+ and Ni 3+ produces a local spinel arrangement dispersed in the NiO matrix, which alters the cubic rock-salt structure and increases the lattice parameter [25,26]. After heating in air at 300°C or above, the crystallite size enlarges to 42 nm and the lattice constant approaches the standard powder value ( figure 3). This indicates annihilation of V Ni defects by the annealing.

NiO thin films obtained by reactive sputtering from metallic target
The as-grown NiO film exhibits a high p-type conductivity of 0.78 S cm −1 , with a carrier concentration of 6.2×10 18 cm −3 . The conductivity decreases by several orders of magnitude after heating, to 4.0×10 −4 S cm −1 at 300°C and 1.5×10 −4 S cm −1 at 500°C, owing to the decrease in the free holes concentration to the 10 14 -10 13 cm −3 range. Figure 4 shows the corresponding optical data, which indicate that the overall transmittance increases as the electrical conductivity decreases. Such relationship is in accordance with the literature data [27][28][29], where high conductivity and darkening are attributed to the presence of V Ni defects and subsequent transition of adjacent Ni 2+ lattice ions into Ni 3+ , which refers to a structure that contains Ni +2 with additional holes localized on Ni sites [29]. The increment in the holes concentration goes along with an increase in the Coulomb attraction between Ni 3+ and O 2− , decreasing the threshold energy of the p-d charge transfer involved in the interbands transition [30]. This results in a transparency decrease by gap narrowing. The band-gap energy, calculated in the inset of figure 4, is found increasing from E g =2.91 eV for the NiO film asgrown at room temperature to E g =3.66-3.71 eV after heating at 300°C-500°C. In a similar way, the values reported for other nickel oxide films change in the range E g =3.0-4.0 eV depending on the deposition conditions [27][28][29]. Despite their higher visible transparency, the heated layers have a very low electrical conductivity due to the annihilation of nickel vacancies by lattice rearrangement when the temperature increased. Therefore, NiO films as-grown at room temperature are more suitable for acting as p-type conductive electrodes.

SnO thin films obtained by reactive sputtering from Sn target
The tin oxide layers prepared by reactive sputtering at room temperature are found amorphous for the x-ray diffraction [31]. Nevertheless, they show crystallization in the tetragonal SnO structure (JCPDS card No 06-0395) after heating at T200°C, as it is illustrated in figure 5for thin films prepared at 12% O pp . The SnO crystallite size increased from 24 to 30 nm when the temperature was raised from 200 to 450°C, although some   2+ and Sn 4+ states), which is typically formed during the oxidation from SnO to the most stable SnO 2 phase [32][33][34].
For these SnO samples, a significant improvement in the optical and electrical properties is obtained by increasing the annealing temperature from 200 to 400°C. The evolution of the optical transmittance is represented in figure 6, where the highest transparency is observed after heating at 350°C-400°C. This corresponds to p-type conductivities of 0.8 S cm −1 with free hole densities around 1.5 ×10 18 cm −3 , attributed to the presence of acceptor defects such as Sn vacancies (V Sn ). The band-gap energy calculated in the figure 6 inset achieves a maximum E g =2.93 eV for the layers heated at 350ºC-400°C, with no appreciable differences in that temperature range. The subsequent optical and electrical worsening at 450°C is related to the crystallization of SnO 2 and intermediate oxides in competition with the main SnO, as evidenced by the respective XRD pattern in figure 5. It is known that SnO 2 is intrinsically n-type due to a low formation energy for donor oxygen vacancies [35,36]. The best optical and electrical performance is then achieved for single-phase SnO layers after annealing at a moderate temperature of about 350°C.   The various processes were performed at a same temperature of 250°C, changing the heating atmosphere in one or two steps. The crystalline structure of these samples is analyzed in figure 7. The XRD graph for oxidized CuS shows diffraction peaks corresponding to cubic Cu 2 S (JCPDS no. 84-1770) and cubic Cu 2 O (JCPDS no. 05-0667), instead of the hexagonal CuS phase (JCPDS no. 06-0464) that was the only one present before the oxidation treatment [37]. The initial CuS layer consisted of crystallites with 34 nm sizes, which have evolved into slightly larger Cu 2 S crystallites (36 nm in diameter) along with small Cu 2 O crystallites (15 nm average size) during the oxidation process. Analogous evolution from hexagonal CuS to cubic Cu 2 S nanocrystals, when heated at 250°C in flowing nitrogen, has been attributed to a loss of sulfur atoms and subsequent lattice rearrangement [38]. In other way, the Cu 2 O sample represented in figure 1 (showing cubic crystallites of 14 nm) has changed to hexagonal CuS with larger crystallites (34 nm in size) after sulfurization. For the direct sulfoxidized copper film, additional peaks have been observed and identified in figure 7. They correspond to orthorhombic CuSO 4 (JCPDS no. 15-0775) and monoclinic CuSO 4 ·3H 2 O (JCPDS no. 22-0249), with average crystallite sizes of 48 nm and 40 nm, respectively. The coexistence of anhydrous and hydrated CuSO 4 phases is commonly found [39,40]. This is due to the spontaneous formation of hydrated species at ambient pressure and temperature, the reaction of anhydrous to trihydrate being specially favored [40]. The close connections between anhydrous and hydrated phases make possible the aggregation of molecules in crystals with different polymorphisms [41].
The optical data obtained for the different Cu x (S, O) layers are shown in figure 8. Direct sulfoxidized Cu and oxidized CuS display similar transmittance in the visible range, but a lower transmittance is obtained for the  oxidized CuS in the infrared region. A further transmittance decrease (in both visible and infrared wavelengths) is observed for the sulfurized Cu 2 O. Nevertheless, the band-gap energy moves in a small range E g =2.44-2.48 eV for the various samples, as it is calculated in the inset of figure 8. Then, the transmittance variations should be related to the respective holes concentration and subsequent free carrier absorption [42]. For these layers, the transmittance decreases as the carrier concentration increases in the 10 18 -10 22 cm −3 interval, which is well above the carrier concentration obtained for pure Cu 2 O (10 16 cm −3 in the section 3.1). The incorporation of sulfur anions is proven effective to increase the carrier concentration and conductivity, as expected from previous investigations [43]. More specifically, the Cu x (S, O) films obtained by sulfurization of Cu 2 O have achieve the best electrical performance in this work, with a holes density of 1.1×10 22 cm −3 corresponding to p-type conductivity of 2.2x10 2 S cm −1 .

Comparison of the obtained thin film materials as p-type TCOs
In order to evaluate and compare the quality of different layers as transparent and conductive electrodes, a figure of merit is defined by Haacke as [44]: which considers the optical characteristics at 550 nm wavelength (transmittance T V or absorption coefficient a V ) as well as the electrical properties (sheet resistance R s or conductivity s). Besides, n is a number that can take different values setting the film thickness and transmittance limits that maximize the function: Thus, values of n equal to 2, 4, 8 or 10 place the maximum at visible transmittances of 61%, 78%, 88% or 90%, respectively. Recent studies have pointed out that the resolution of Φ H decreases when the transmittance power n increases [45], owing to the high predominance of optical characteristics that vary in a relatively small range (0T V 1). However, the resolution greatly improves when the power is charged in another way: giving a new figure of merit named H-HR (Haacke high resolution) [45]. It should be noted that a same film thickness maximizes both Φ H and Φ H-HR functions, since for the last expression it is also The selection of the n or m value depends on the transparent conductor application. In most cases, a visible transmittance of about 80% is considered appropriate, and therefore a value m=4 has been selected to compare the present samples. Table 1 summarizes the main optical and electrical characteristics for the different metal oxides analyzed in the previous sections. For each material, the respective α V and σ data have been used to calculate the evolution of the high-resolution figure of merit in function of the film thickness, as it is illustrated in figure 9. This figure shows experimental points corresponding to samples with different thicknesses and also extrapolation lines in a wider range. The highest quality is achieved by the Cu 2 (S, O) layers crystallized at 250°C, which give a maximum Φ H-HR =0.126 Ω −1/4 at t∼30 nm, being Φ H-HR 0.044 Ω −1/4 for t230 nm. For superior thicknesses, the SnO films crystallized at 350°C achieve a better quality. Moreover, the NiO layers obtained at room temperature have a good quality to be applied on heat-sensitive substrates, with Φ H-HR 0.020 Ω −1/4 for t70 nm.

Conclusions
Several metal oxides (Cu 2 O, NiO and SnO) have been prepared, characterized and evaluated as transparent and p-type conductive films, according to the Haacke-high-resolution figure of merit. NiO is the only one that exhibits a crystalline structure as-grown at room temperature, with cubic crystallites of 33 nm in size, and some lattice distortion related to the presence of nickel vacancies. These acceptor defects allow achieving a hole density of 6.2x10 18 cm −3 and p-type conductivity of 0.78 S cm −1 . The optical absorption coefficient is relatively high (1.4x10 5 cm −1 at 550 nm wavelength) and, therefore, the film thickness should be kept below 70 nm in order to have a figure of merit within the 0.02-0.03 Ω −1/4 range.
The lowest visible absorption (2.0×10 4 cm −1 at λ=550 nm) corresponds to tetragonal SnO crystallites (26 nm mean size) obtained at 350°C, with p-type conductivity of 0.80 S cm −1 . These SnO layers maintain good quality for a wide range of thicknesses, being the figure of merit in the interval 0.03-0.04 Ω −1/4 for films with t=10-600 nm.
In comparison, copper oxides show a poor conductivity of 1.7×10 −2 Scm -1 for Cu 2 O crystallites (14 nm in size) obtained at 250°C and lower for CuO crystallized at 350°C, due to low free carrier concentrations. However, the incorporation of sulfur anions improves the electrical characteristics. In fact, Cu x (S, O) layers obtained by sulfurization of Cu 2 O have reached the highest hole density (1.1×10 22 cm −3 ) and p-type conductivity of 2.2×10 2 S cm −1 . This corresponds to a figure of merit above 0.10 Ω −1/4 for Cu x (S, O) film thickness below 100 nm, with a maximum of 0.126 Ω −1/4 for t∼30 nm.

Declarations of interest
None. Table 1. Summary of the main characteristics obtained for the various compounds. They are included the band-gap energy, absorption coefficient at 550 nm wavelength, p-type conductivity, carrier concentration, mobility and maximum figure of merit achieved for each material.