Synthesis of indium oxi-sulfide films by atomic layer deposition: The essential role of plasma enhancement

Summary This paper describes the atomic layer deposition of In2(S,O)3 films by using In(acac)3 (acac = acetylacetonate), H2S and either H2O or O2 plasma as oxygen sources. First, the growth of pure In2S3 films was studied in order to better understand the influence of the oxygen pulses. X-Ray diffraction measurements, optical analysis and energy dispersive X-ray spectroscopy were performed to characterize the samples. When H2O was used as the oxygen source, the films have structural and optical properties, and the atomic composition of pure In2S3. No pure In2O3 films could be grown by using H2O or O2 plasma. However, In2(S,O)3 films could be successfully grown by using O2 plasma as oxygen source at a deposition temperature of T = 160 °C, because of an exchange reaction between S and O atoms. By adjusting the number of In2O3 growth cycles in relation to the number of In2S3 growth cycles, the optical band gap of the resulting thin films could be tuned.


Introduction
Chalcopyrite-type thin film solar cells that are based on a Cu(In,Ga)Se 2 (CIGS) absorber have reached high efficiencies, up to 20.3% [1] in 2011 and 20.4% [2] on flexible substrates in 2013. The best efficiencies were obtained by using cadmium sulfide (CdS) as buffer layer in solar cells with a glass/Mo/ CIGS/CdS/i-ZnO/ZnO:Al stack. The buffer layer is an n-type semiconductor that forms the p-n junction with the p-type CIGS absorber, and also modifies the CIGS surface chemistry, which is usually too sensitive for a direct deposition of the window layers. However, because of the toxicity of cadmium and the low optical band gap of CdS (2.4 eV [3]) that limits the light conversion of CIGS in the UV range of the solar spectrum,  [24] indium sulfide InCl 3 H 2 S 300 1.4 [25] In(acac) 3 H 2 S 160, 180, 160, 150 0.6, 0.7, 0.44, 0.3 [9,[26][27][28] alternative materials have been developed. Most Cd-free buffer layers are based on zinc and indium-compounds, with current record efficiencies obtained by chemical bath deposition (CBD, 19.7% and 19.1% for Zn(S,O,OH) [4,5], 15.7% for In(S,O,OH) [6]) or atomic layer deposition (ALD, 18.5% for Zn(O,S) [7], 18.1% for (Zn,Mg)O [8], 16.4% for In 2 S 3 [9], and 18.2% for (Zn,Sn)O [10]). Recently, our group has synthesized new mixed films of ZnS/In 2 S 3 by using ALD and applied them as buffer layers in CIGS solar cells [11,12]. ALD is based on sequential self-saturated reactions that allows the conformal and uniform growth of thin films with a high control of their properties [13][14][15]. It is therefore a suitable technique for the deposition of buffer layers. Platzer-Björkman et al. have used ALD to improve the energy-band alignment between the CIGS and the front electrode by controlling the oxygen concentration in Zn(S,O) buffer layers [4,16]. Oxygen-doping of In 2 S 3 films is known to increase their optical band gap value [6,17,18] [18]. Thus, based on our previous results and those studies, we became interested in adjusting the optical properties of In 2 S 3 by incorporating oxygen atoms while using the advantages of ALD. Typical ALD processes for the deposition of In 2 S 3 and In 2 O 3 are referenced in Table 1. As ALD processes of In 2 O 3 report relatively small growth rates, we will consider the case of plasma enhancement. Indeed, plasma-enhanced ALD (PEALD), in which various reactive species are generated, has been the key for the development of fast thin-film deposition processes at low temperature. It is widely used to enhance the thin-film deposition of materials such as Al 2 O 3 , ZnO, Ta 2 O 5 , TiN, TaN and SiN x [19].
In this study, ALD and PEALD have been used to synthesize In 2 (S,O) 3 thin films and carry out optical band-gap engineering. The structural, optical and growth properties of the films will be studied and the role of the plasma will be discussed.

Results
Study of In(acac) 3 , H 2 S and H 2 O system First, a controlled growth of pure In 2 S 3 films was established and the film properties were measured in order to clearly identify the influence of oxygen pulse later in the study. For that, In 2 S 3 growth was achieved in the temperature range between 140 and 240 °C. An In 2 S 3 growth cycle consists of the following steps: In(acac) 3 exposure/N 2 purge/H 2 S exposure/N 2 purge = 0.1/5/0.1/5 s, the relative long purge time being chosen to ensure a good homogeneity. Figure 1a shows the growth rate of In 2 S 3 thin films at various temperatures. It globally increases with the temperature. An ALD window can be speculatively observed between 160 °C and 200 °C with a mean growth rate of 0.84 Å/cycle. The variation of the In 2 S 3 growth rate with different In(acac) 3 pulse lengths at a process temperature of 160 °C is illustrated in Figure 1b. This variation only slightly influences the growth rate and a saturation by lengthening the precursor pulse is not observed. The data suggest that the results displayed on Figure 1a may not have been obtained under completely self-limiting conditions. Structural and optical properties of the films were also investigated. In 2 S 3 thin films have an amorphous structure for deposition temperatures below 180 °C and a β-tetragonal crystal structure at higher temperatures. Their indirect optical band gap varies from 2.0 eV to 2.2 eV.
Then, we attempted to synthesize In 2 (S,O) 3 film by inserting an In 2 O 3 growth cycle. For this H 2 O was pulsed, instead of H 2 S in the growth of pure In 2 S 3 , which led to the supercycles n·{In 2 S 3 } + {In 2 O 3 } with n = 1,2,3,5,9,14,19, which corres- pond to ratios of {50%, 33%, 25%, 10%, 6.7%, 5%} of In 2 O 3 cycles at a deposition temperature of 200 °C. All samples were deposited performing a total of 2000 growth cycles, i.e., 100 supercycles for n = 19, 133 supercycles for n = 14, etc. Energy dispersive X-ray spectroscopy analysis was performed on the samples and gave atomic ratios of 0.4 for In/(In+S) and 0.6 for S/(In+S), which correspond to typical In 2 S 3 atomic ratios. The oxygen contents are similar to those of pure In 2 S 3 films, which is assigned to the oxygen contamination of the substrate. Those results were confirmed by GI-XRD measurements. They were performed to investigate the influence of the H 2 O pulse on the microstructure of the films (Figure 2b). Not all samples were crystalline and the crystalline ones can be attributed to β-In 2 S 3 with a random orientation by comparing the diffraction patterns with the reference data and with the literature [27]. Indeed, we should observe a peak shift due to increasing oxygen doping when changing the In 2 O 3 /In 2 S 3 ratio. However, the peaks remain at the same diffraction angles. Comparing the FWHM of the (109) peak, the maximum FWHM measured was 1.2° for the 10%-In 2 O 3 sample, which corresponds to the thickest film.
In general it can be said that the thinner the films, the lower the FWHM. From these observations, it seems that we obtained In 2 S 3 films only. Thin films optical absorption were determined from transmittance (T) and reflectance (R) measurements by using the following formula [29] where α is the absorption coefficient and t is the film thickness. Figure 3 shows absorption spectra of the thin films. They are presented in the form of (α) 0.5 = f(E), which is linear for indirect band gap materials and allows for the determination of the optical transition. The optical band gaps correspond to an indirect transition in the range from 1.9 to 2.2 eV, which is roughly similar to that of pure In 2 S 3 film optical properties. No correlation could be found between either the ratio of In 2 O 3 cycles or the film thickness and the optical measurements. These results are in accordance with the observations of the structural analysis. Consequently, this method is not suitable to synthesize In 2 (S,O) 3 thin films. In parallel, we attempted to synthesize pure In 2 O 3 films from In(acac) 3 and H 2 O at temperatures of 160 and 200 °C. This remained unsuccessful, because no films could be grown under these conditions.   Figure 4a. When increasing the ratio from 4.8% to 9.1%, the growth rate increases up to 1.4 Å/cycle and then decreases again. The variation of the film thickness with the number of ALD cycles for a ratio of 10% of In 2 O 3 cycles is illustrated in Figure 4b. A linear growth is observed up to 1500 ALD cycles. GIXRD measurements revealed an amorphous structure in all the samples.
Transmittance and reflectance measurements were carried out on the In 2 (S,O) 3 samples. Figure 5 shows the transmittance of   The atomic ratios of oxygen, sulfur and indium determined by using EDX are presented in Table 2 and correlated to the optical band gap values. The dependence of the atomic ratio of oxygen and the optical band gap on the number of In 2 O 3 cycles is not clear. In general, high oxygen concentrations of more than 66 atom % were measured in the films.   We also tried to synthesize In 2 (S,O) 3 by using a single O 2 plasma pulse instead of In 2 O 3 pulse cycles. The following cycle program was used: 20·{In 2 S 3 } + 2·{O 2 +plasma exposure}/N 2 purge with the same process parameters. This corresponds to 9.1% of indium cycles, which can be compared to the previous deposition with a pulse of In(acac) 3 before the O 2 plasma exposure. Table 3 shows the properties of these two samples, along with those of pure In 2 S 3 . Even without the In(acac) 3 pulse, O and S atomic ratios indicate that the synthesized film corresponds to a In 2 (S,O) 3 film and no significant differences were observed between the samples.
The growth of In 2 (S,O) 3 growth could be achieved when using O 2 plasma as oxygen precursor. The maximum growth rate was 1.4 Å/cycle, which is higher than the growth rates of In 2 S 3 shown in Figure 1 and those reported in the literature for this deposition temperature [9,[26][27][28]. Optical measurements revealed an onset absorption moving to higher energies when increasing the number of In 2 O 3 cycles. At the same time, the optical band gap increased from 2.2 eV to 3.3 eV for In 2 O 3 cycle ratios in the range from 0 to 11.8%. EDX analysis showed that those films have a high oxygen content. Finally, all attempts to synthesize pure In 2 O 3 films from In(acac) 3 and O 2 plasma remained unsuccessful .

Discussion
It has been observed that inserting an In 2 O 3 cycle during the deposition of In 2 S 3 when using H 2 O as oxygen precursor has no influence on the oxygen content and on the film properties. It only affects the growth rate as the thickness varies. Attempts to synthesize pure In 2 O 3 thin films were also unsuccessful, which suggests a low reactivity of H 2 O towards In(acac) 3 . Several authors reported the difficulty to synthesize In 2 O 3 by ALD using β-diketonates (In(acac) 3 , In(hfac = hexafluoropenta-dionate) 3 , In(thd = 2,2,6,6-tetramethyl-3,5-heptanedioneate)) and water [20,21,25]. In most cases, they assigned the low growth rates or the absence of grown films to the low reactivity of water toward β-diketonates.
As no pure In 2 O 3 films could be grown in our case, the synthesis of mixed films by a simple addition of two layers, i.e., In 2 O 3 + In 2 S 3 , is not possible. However, the deposition of ternary materials can also occur via exchange reactions. For instance, when synthesizing zinc indium sulfide (ZIS) thin films, substitution mechanisms between diethylzinc (DEZ) and In 2 S 3 could be demonstrated [11]. Similar mechanisms also occur when inserting H 2 O in pure ZnS during the growth of Zn(S,O) by using ALD [4,16]. Such processes do not seem to occur in our case, because the In 2 (S,O) 3 deposition method that uses H 2 O remained unsuccessful. A possible thermodynamic explanation for the unfavorable deposition of In 2 (S,O) 3 using H 2 O as oxygen precursor is that the following exchange reaction is endothermic and thus unlikely to occur [31]. (1) Due to the high reactivity of radicals, PEALD generally allows the achievement of many chemical reactions that cannot occur with thermal ALD [13,19]. Here In 2 (S,O) 3 films could be grown while using O 2 plasma as oxygen source. But the growth of pure In 2 O 3 films remained unsuccessful. This suggests that the oxygen contained in In 2 (S,O) 3 films is not generated from single layers of In 2 O 3 but rather by exchange reactions as described in the previous section. Indeed, the O 2 plasma can directly react with the film surface and induce an exchange reaction with surface sulfur atoms. Figure 7 presents a scheme of the assumed substitution mechanism at the surface. The following exchange reactions can explain the substitution of S atoms by reactive oxygen species generated in the plasma. Indeed, their free standard enthalpies all have negative values: Thus, when comparing these reactions with the reaction between In 2 S 3 and H 2 O, it seems that the doping is only favorable when using O 2 plasma as oxygen precursor, because these reactions are all exothermic. This thermochemical analysis and the observation that In 2 (O,S) 3 films obtained from the two different ALD pulse programs 20·{In 2 S 3 } + 2·{In 2 O 3 } and 20·{In 2 S 3 } + 2·O 2 plasma have similar properties, show the critical role of activated oxygen during the deposition of In 2 (S,O) 3 .
Commonly existing species in oxygen plasmas are atomic oxygen that is created from molecular oxygen dissociation, excited oxygen species at different electronic levels, ionized oxygen or recombined species like O 3 [32,33]. Consequently, exchange reactions between adsorbed oxygen and oxygen species from the gas phase or recombination reactions have to be considered. Marinov et al. studied the interactions between a radiofrequency O 2 plasma and oxide surfaces like TiO 2 , SiO 2 and Pyrex [32]. They demonstrated that these materials surfaces are continuously re-structured under O 2 plasma exposure because of the exchange reactions that occur between O atoms in the films and oxygen species of the gas phase. They also reported that reaction products undergo oxidation at the surface and assumed that two surface mechanisms could occur; O + O → O 2 and O + O 2 → O 3 . Such mechanisms might also occur in our case considering the exchange reactions described in Figure 7 during the first O 2 plasma pulse. Indeed, no match was found between the ratios of In 2 O 3 cycles during the deposition of In 2 (S,O) 3 films and the oxygen content of the films determined by EDX. When the number of In 2 O 3 cycles varied from 4.8% to 11.8%, in the same time the oxygen content of the films varied from 68 atom % to 75 atom %. Oxidation mechanisms during deposition process can explain these high differences between the expected values and those measured. On-going studies focus on a better understanding on the nature of the oxygen species generated by the plasma, their role in oxidizing mechanisms and the reason of the relatively low indium content. One of them could be an excessive adsorption of oxygen in the film and the formation of sulfates. Further studies, in particular by using X-ray photoelectron spectroscopy, are in progress to assess the presence or not of such groups. Experiments will also be performed to study the influence of other oxygen sources such as O 2 alone and O 3 .

Conclusion
In this study we reported the atomic layer deposition of In 2 (S,O) 3 films by using In(acac) 3  introduced in a remote RF plasma generator with Argon (99.9997%, Messer) as carrier gas, and the plasma power was kept at 2600 W. All sources were kept at room temperature while In(acac) 3 was heated to 200 °C. The carrying and purge gas was nitrogen with a purity of 99.9999% (Messer). The pressure in the reaction chamber was kept in the range from 1 to 4 mbar.
The thickness of the films was measured using a VEECO DEKTAK 6M profilometer on glass substrates. Thicknesses were determined after creating steps in the films, by masking film parts with chemically resistant tape and dipping the film in nitric acid (45% in water) at room temperature for 60 s. The uncertainty given for the thickness is the standard deviation of six measurements taking into account the uncertainty of the profilometer, the sharpness of steps, the film roughness, and the film inhomogeneity. Transmittance and reflectance spectra were obtained by using a PerkinElmer lambda 900 Spectrophotometer with a PELA-1000 integrating sphere. All optical measurements were performed on borosilicate glass substrates. X-Ray diffraction (XRD) studies were performed under grazing incidence X-ray diffraction conditions with a PANalytical Empyrean diffractometer while using Cu Kα radiation. X-Ray reflectometry analyses were also performed to confirm thickness measurements. Thin film compositions were obtained by using a Magellan 400L scanning electron microscope provided by FEI. It is equipped with an energy dispersive X-ray spectroscopy detector INCASynergy 350. All EDX measurements were carried out on Si(100) substrates and the values reported are atomic percentages (atom %).