- Email: [email protected]
Thin film characterization of zinc tin oxide deposited by thermal atomic layer deposition
Thin Solid Films 556 (2014) 186–194
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Thin film characterization of zinc tin oxide deposited by thermal atomic layer deposition Marja N. Mullings a, Carl Hägglund a, Jukka T. Tanskanen a, Yesheng Yee b, Scott Geyer a, Stacey F. Bent a,⁎ a b
Department of Chemical Engineering, Stanford University, Stanford, CA 94305, USA Department of Electrical Engineering, Stanford University, Stanford, CA 94305, USA
a r t i c l e
i n f o
Article history: Received 1 August 2013 Received in revised form 23 January 2014 Accepted 24 January 2014 Available online 31 January 2014 Keywords: Zinc tin oxide Atomic layer deposition Thin film Transparent conductive oxide
a b s t r a c t Zinc tin oxide (ZTO) thin films are of interest for many applications, including transparent conducting oxides and buffer layers in thin film solar cells. In this work, the ability of atomic layer deposition (ALD) to control both thickness and composition of the ternary material is explored and the resulting film properties are characterized. ZTO was deposited at 150 °C by alternating growth of tin oxide (SnOx) and zinc oxide (ZnO) using tetrakis(dimethylamido)tin and diethyl zinc as metal precursors and water as the oxygen source. The growth behavior of ZTO was systematically examined as a function of the relative fraction of SnOx to ZnO ALD cycles and their bilayer period. Variable angle spectroscopic ellipsometry (VASE) showed that the ZTO growth rate was strongly reduced as compared to the growth rates of the binaries, especially at low bilayer periods and low tin cycle fractions. Inductively-coupled plasma optical emission spectroscopy showed that the composition, together with other properties of ZTO films, was not simply proportional to the binary ALD cycle fractions utilized in the deposition. Structurally, the deposited films were shown by X-ray diffraction to be amorphous for tin content exceeding 10%, for which only smooth features were observed by scanning electron microscopy. Optically, VASE revealed a minimum refractive index in the visible range for intermediate compositions and a maximum for the pure SnOx phase. Conversely, the ZTO bandgap was maximized for intermediate metal compositions near ~50%, and converged to the direct ZnO bandgap of 3.3 eV for compositions with decreasing amounts of Sn in the films. Linear extrapolations gave a range of 2.9 to 4.0 eV for SnOx bandgap depending on the bandgap type (indirect or direct). © 2014 Elsevier B.V. All rights reserved.
1. Introduction Transparent conducting oxides (TCOs) are wide bandgap (≥3 eV), transmissive to visible light (≥ 80%), and highly conductive (≥ 103 S cm−1) metal oxides with many applications in consumer electronics and in photovoltaics [1–4]. The oxides of indium, tin, and zinc are the most commercially-important TCOs [4,5]. These binary TCOs are typically doped with an impurity to improve stability at high temperatures [6]. Within the indium/tin/zinc oxide family, indium tin oxide (ITO) remains the industry-standard for high-end TCOs because of its high conductivity (≥104 S cm−1) [7–9]. However, the scarcity of indium, with an abundance in the earth's upper crust of about 0.05 ppm [10], coupled with the increasing demand in consumer electronic and photovoltaic markets has led to rising costs [11], which motivates the need to find alternative TCO materials. Impurity-doped or alloyed zinc oxides are interesting candidates for this purpose, where mixtures with aluminum (AZO) or gallium (GZO) are well known examples [1,12–14]. While AZO and GZO exhibit attractive properties as inexpensive indium-free alternatives, other options such as zinc tin oxide (ZTO) are also being studied as an inexpensive ⁎ Corresponding author. E-mail address: [email protected] (S.F. Bent). 0040-6090/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.tsf.2014.01.068
indium-free TCO [1,11,15–23]. Notably, the upper crustal abundances of Zn and Sn are around 70 and 5.5 ppm [24], these values being significantly higher than the 0.05 ppm abundance of indium. In addition to its use as a TCO and buffer layer in high efficiency solar cells [25], ZTO has been used in thin-film transistors [26–28] and it has potential in varistor applications as well [29,30]. Despite its promise, a review of the literature suggests little agreement among various structural, optical, and electrical properties of ZTO, motivating a more detailed characterization of this material. In general, there is an agreement that ZTO forms two crystal structures. The metastable ZnSnO3 phase, known as zinc metastannate, has two polymorphs — a face-centered perovskite [31] and an ilmenite [32] structure. The other structure, Zn2SnO4, is known as zinc orthostannate and has a cubic spinel structure that is stable under ambient conditions [33]. However, the correlation between composition and crystallinity is not well-understood. This lack of understanding applies to many other ternary oxides as well, and it originates from the mixing of binaries with potentially different crystal structures, densities, and/or stoichiometries to produce the ternary material. The most extensively used deposition technique for ZTO has been sputtering [6,15,16,18–20,34]. During crystallization, ZTO transforms to ZnSnO3 from 300 to 500 °C and to Zn2SnO4 for temperatures exceeding 600 °C. At these high temperatures, however, obtaining pure Zn2SnO4
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
is difficult, as mixed phases including both forms of ZTO and SnO2 are formed [33]. Other techniques for ZTO deposition have included sol– gel spin coating [27,28,35], chemical vapor deposition [36], and only very recently, atomic layer deposition (ALD) [25,37–40]. A lack of agreement on the ZTO crystal structure as a function of composition may explain the variation in reports on its optical properties. One study (using pulsed layer deposition) reported bandgaps for amorphous ZTO of 2.80 eV for ZnSnO3 and 2.86 eV for Zn2SnO4 [41]. Young et al. found crystalline Zn2SnO4 to have a fundamental bandgap of 3.35 eV that was widened to 3.89 eV by a Burstein–Moss shift due to high carrier concentration [42]. For ZTO synthesized by hydrothermal methods, the optical bandgap for crystalline Zn2SnO4 has been reported with great variation in the range of 3.3 to 3.9 eV [43,44]. The bandgap for crystalline ZnSnO3 has not been rigorously characterized experimentally, but one report presents a value of ~3.9 eV [45], which is significantly higher than the gap of 2.80 eV reported for the amorphous phase (see above). Most previous studies using ZTO focused on device performance instead of film characterization, resulting in few comprehensive reports of material and electrical properties as a function of composition. Because ALD allows one to carefully dope or alloy material in an atomic-level controlled fashion with uniform and conformal growth [46–48], it provides an opportunity to systematically analyze ZTO properties by fine tuning the composition. In addition to ZTO, ALD has been used to deposit several other ternary metal oxides including AZO, [49–53] GZO [54,55], ITO [56], (Zn,Mg)Ox [57], and (Sn,Al)Ox [58]. In this study, we demonstrate a reproducible thermal ZTO ALD process and determine the composition, optical properties, and structural characteristics of the resulting thin films by variable angle spectroscopic ellipsometry (VASE), X-ray photoelectron spectroscopy (XPS), inductively-coupled plasma optical emission spectroscopy (ICP-OES), scanning electron microscopy (SEM), X-ray diffraction (XRD) and ultraviolet–visible spectroscopy (UV–vis). An analytical model based on analysis of the ZTO growth rates is developed for understanding the main features of the ZTO growth characteristics. Relationships between deposition parameters, namely bilayer period and relative fraction of SnOx to ZnO ALD cycles (cycle fraction), and ZTO growth rate, composition, structure, and optical properties are established and discussed in the context of previous studies on ZTO. 2. Experimental methods 2.1. ZTO ALD process The ALD experiments were carried out in a custom-built, viscousflow, hot-wall reactor previously described [59]. Diethyl zinc (DEZn; Sigma-Aldrich) and tetrakis(dimethylamido)tin (TDMASn; Strem, N99% purity) were used as precursors with deionized water as the counter-reactant. Precursors were dosed from 10 cm3 stainless steel sample vials using computer controlled air-actuated valves. The water and DEZn sample vials were kept at room temperature, while the TDMASn vial was heated to 45 °C during depositions. A positive temperature gradient was maintained toward the reaction zone to prevent precursor condensation. During the ALD process, excess precursor was purged with nitrogen for half-reaction separation and reaction completion. For ZnO ALD, we express the pulse time sequences as z1:z2:z3:z4, where z1 and z3 correspond to the pulse lengths of DEZn and water, respectively, and z2 and z4 correspond to N2 purge times following these doses. A typical ZnO ALD pulse time sequence was 1:15:1:30 s. The purge times of z2 = 15 s and z4 = 30 s were chosen based on the return of the precursor concentrations to their baseline values as monitored by in situ mass spectrometry, as well as previously-found saturation in this range for the ZnO system [50,60]. For SnOx ALD, we similarly express the ALD sequences as t1:t2:t3:t4, where t1 and t3 correspond to the pulse lengths of TDMASn and water, respectively, and t2 and t4 correspond to N2 purge times following these doses. A typical timing sequence
187
was 1:30:2:30 s which corresponded to saturated ALD of SnOx as previously described [59]. Designing a reproducible ZTO ALD process first required optimization of the binary ZnO and SnOx ALD processes in our reactor. It is convenient that there is overlap in the ALD temperature windows for these binary metal oxide processes. It was reported previously that within the ZnO ALD temperature window, a maximum growth rate of about ~ 2.0 Å/cycle occurs at 150 °C [61], which is close to the optimized growth rate of 1.85 Å/cycle for our reactor. The ALD of SnOx, reported elsewhere [59], results in SnOx growth rate of about 0.70 Å/cycle at 150 °C. For the optimized SnOx ALD system, the as-deposited film is amorphous, whereas as-deposited ZnO grown by ALD from DEZn and water at 150 °C has a crystalline hexagonal wurtzite structure [62]. The ALD super cycle is defined as the minimum sequence of SnOx and ZnO cycles that is repeated over the course of the ZTO ALD process. A corresponding super cell results from the stack of layers produced by one super cycle. To characterize the ALD system, we define t and z, the total number of SnOx and ZnO ALD cycles in one ZTO ALD super cycle, respectively. Using these parameters, illustrated in Fig. 1, the system can be characterized by the bilayer period Λ, defined as the number of cycles in one super cycle Λ = (t + z), and the relation for the total number of ZnO and SnOx cycles in a ZTO ALD process, referred to as cycle fraction. If a total number of ALD super cycles S are performed, then the total number of ALD cycles in the deposition is given by S × Λ . Substrates for ZnO, SnOx, and ZTO ALD were n-type Si(100) with phosphorous dopant and a resistivity of 1.0–5.0 Ω cm (WRS Materials) and fused silica quartz (ChemGlass). Substrates were pre-cleaned using a UV-ozone (PSD Benchtop UV-Ozone Cleaner) to remove organic contamination, after which the thin films were deposited onto the native Si oxide. 2.2. Characterization A PHI VersaProbe 5000 scanning X-ray photoelectron spectrometer with Al-Kα radiation was used to survey film elemental content. The C1s peak at 285.0 eV was used as a reference. After initial survey scans, adventitious carbon was removed by Ar+ sputtering for 12 s (1 keV, 0.5 μA, 1 mm × 1 mm at 45° incidence from the sample). A Thermo Scientific ICAP 6300 Duo View spectrometer was used to perform inductively-coupled plasma optical emission spectroscopy (ICP-OES) on ZTO deposited on Si(100). For these measurements, films were dissolved in aqua regia (3:1 concentrated HCl:HNO3) and sonicated for at least 6 h before characterization. X-ray diffraction (XRD)
Fig. 1. Schematic of ZTO super cycle formalism visualized as a lamellar stack of ALD cycles of constituent binary oxides. One super cycle includes a certain number of SnOx cycles, t, and ZnO cycles, z, resulting in a bilayer period Λ = (t + z).
188
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
Table 1 Sn composition fractions (cSn) and growth rates for ZTO films for different bilayer periods with a constant Sn/(Sn + Zn) cycle fraction of 0.75 and a constant total number of ZTO ALD cycles of 800. Total number of super cycles (S), bilayer period (Λ), the number of SnOx ALD cycles in a super cycle (t), and the number of ZnO ALD cycles in a super cycle (z) used to deposit the films are also given. The growth rates and compositions were determined by VASE and ICP-OES, respectively. The films were deposited at 150 °C on Si(100). S
Λ
t
z
cSn
Growth rate (Å/cycle)
200 100 20 5 2 1
4 8 40 160 400 800
3 6 30 120 300 600
1 2 10 40 100 200
0.35 0.46 0.36 0.37 0.38 0.42
0.52 0.55 0.79 0.94 0.97 1.06
measurements of ZTO on fused-silica quartz were performed as a function of ALD cycle fractions using the PANalytical X'Pert PRO system in parallel beam mode with Cu Kα radiation at 45 kV and 40 mA. Grazing incidence X-ray diffraction (GIXRD) was also performed on ZTO films deposited on Si(100) using 12.7 keV X-rays that were incident on the sample at 0.48°. Diffracted X-rays were collected with an MAR34 twodimensional imagining detector calibrated using a LaB6 reference. Film thickness and optical properties of ZTO deposited on Si(100) were determined using a J.A. Woollam M2000 Variable Angle Spectroscopic Ellipsometer (VASE) at 65, 70 and 75° angles of incidence and wavelengths ranging from 210 to 1688 nm. In the analysis, the optical properties of the films were represented by a sum of oscillator terms, described in more detail below. A model with a single homogeneous layer (to be referred to as a single layer model) was found accurate for finely layered ZTO, but for the most coarsely layered stacks, the root mean squared error was lower if taking the periodic structure into account with a two layer unit cell corresponding to the super cycle of ZnO and SnOx used in the ALD sequence. We will refer to the latter as a super cycle model. Plan view images of ZTO film morphology on Si(100) were examined using a FEI Magellan 400 XHR Scanning Electron Microscope with FEG source (5 kV and 25 pA). Optical extinction spectra of films deposited on quartz were recorded using a Varian Cary 6000i UV–vis-NIR spectrometer.
Fig. 3. Tin composition fraction in ZTO films from ICP-OES (markers) and from an analytical model (Eq. (1)) as a function of bilayer period, Λ . The black dashed line gives the composition linearly interpolated from the growth rates and densities of the binaries. The films were deposited at 150 °C on Si(100) by using a Sn/(Sn + Zn) cycle fraction of 0.75.
Atomic layer deposition allows the sequential dosing of an arbitrary number of SnOx and ZnO cycles in any desired ratio to deposit ZTO. We hypothesize that the more finely spaced these dosing sequences are, the more laminated the film will be and the less the film will resemble its constituent binaries in growth rate and material properties. The fineness of the cycling can be described by the bilayer period, Λ .
The ZTO ALD super cycle formalism provides a systematic approach to determining the effects of the bilayer period on the growth rate and other properties (see the Experimental methods section). To study the effect of Λ without changing the cycle fraction, we fixed the total number of ZnO and SnOx cycles to 200 and 600, respectively. Fixing these õparameters yielded a tin cycle fraction fSn = Sn/(Sn + Zn) = 0.75, which based on the binary growth rates and densities would result in an intermediate (cSn ~ 43%) overall composition. We then varied the super cell size (t + z) in order to sample different Λ . Table 1 summarizes the effect of varying bilayer period on ZTO growth rates and compositions, and these results are also illustrated in Figs. 2 and 3, respectively. In the figures, theoretical results produced by an analytical model of ZTO growth are also illustrated (see Section 3.2). The data in Table 1 and Fig. 2 show that the growth rate of the intermixed system increases with increasing Λ . When a weighted average of the binary growth rates is calculated for fSn = 0.75, a value of 0.99 Å/cycle results, which is close to the growth rates of the least lamellar films with large Λ values. In contrast, for the highly lamellar films, the growth behavior deviates significantly from this weighted average (labeled as “linearly interpolated” in Fig. 2). This reduced growth rate suggests that at least one of the materials does not grow as well on its counterpart as on itself. In general, as shown in Table 1 and Fig. 3, the overall tin content by ICP-OES varies slightly from the value of cSn = 0.43 that results from calculating the composition for fSn = 0.75 using the binary growth rates and densities (see discussion above). We will discuss the behavior of the ZTO growth rate and composition later in the context of the analytical model developed to understand the ZTO ALD growth characteristics. X-ray photoelectron spectroscopy (XPS) was performed to verify the elemental composition of the ZTO films deposited by varying Λ . Fig. 4 shows an XPS survey scan for a film with a large bilayer period of
Fig. 2. ZTO ALD growth rates from VASE and from an analytical model (Eq. (1)) as a function of the bilayer period, Λ . The black dashed line gives the growth rate linearly interpolated from the growth rates of the binaries. The films were deposited at 150 °C on Si(100) by using a Sn/(Sn + Zn) cycle fraction of 0.75.
Fig. 4. X-ray photoelectron spectroscopy survey scan of a ZTO film deposited on Si(100) at 150 °C after Ar+ sputtering for 12 s. The ZTO film was deposited using a bilayer period, Λ , of 400 and a Sn/(Sn + Zn) cycle fraction of 0.75.
3. Results 3.1. ZTO ALD as a function of bilayer period
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
189
Λ = 400. No contaminants are observed after removal of adventitious surface carbon by Ar+ sputtering, which is representative for the other films investigated here. As this also holds for the binary films [59], we conclude that this thermal ZTO ALD process deposits films of high purity. We also studied film morphology as a function of Λ using SEM (Fig. 5). The films appeared to be smooth and without large and distinct crystalline domains for all bilayer periods studied, with a morphology resembling that of pure and X-ray amorphous SnOx [59]. 3.2. ZTO properties as a function of cycle fraction We now turn to the influence of the cycle fraction on ZTO growth and properties. Because film characteristics vary significantly as a function of bilayer period, Λ , we fixed this value to 6 to examine the role of cycle fraction – which then determines the film composition – on ZTO properties. With a fixed total number of super cycles S = 100, which corresponds to a total number of ALD cycles A = 600, the tin cycle fraction, fSn, was sampled in multiples of 1/6. Note that internal cycling, which would increase the film lamination, was not varied here. For illustration, the film corresponding to fSn = 0.50 was deposited using a cycling of zzzttt in a super cycle instead for instance a cycling of ztztzt. The ZTO growth rates are plotted against the fSn in Fig. 6. The two endpoints on the plot, fSn = 0 and fSn = 1.0, correspond to the two binary systems, ZnO and SnOx, respectively. Based on a linear interpolation of the two binary growth rates (black dashed curve), the observed growth rate for the intermixed system is lower than expected. The ZTO ALD growth rate behavior as a function of fSn resembles the growth behavior of the mixed ZnO/Al2O3 (AZO) ALD process, an analog to ZTO ALD, reported by George and Elam [50]. Notably, similar ALD growth characteristics have been observed also for other ternary materials, including Zn1 − xMgxO [63], and (Sn,Al)Ox [58]. George and Elam deposited AZO films from DEZn and trimethylaluminum (TMA) with H2O as the oxygen source. Supported by detailed quartz crystal microbalance (QCM) experiments, the authors discuss potential causes for the reduced AZO growth, and these include etching of surface Zn by TMA, deficiency of hydroxyl groups on the binary surfaces and surface hydroxyls that are less reactive to the TMA and DEZn reactants.
Fig. 6. Film growth characteristics determined by VASE and an analytical model described by Eq. (1) as a function of Sn/(Sn + Zn) cycle fraction, fSn. The black dashed line gives the growth rates linearly interpolated from the growth rates of the binaries. Films were deposited on Si(100) at 150 °C using a bilayer period of Λ = 6 for a total of 100 super cycles.
The same arguments, which have been discussed in the context of Zn1 − xMgxO ALD as well, apply for ZTO ALD, except for the etching. In the ZnO/Al2O3 system, etching was attributed to two processes: (1) the transfer of CH3 ligands from TMA to surface Zn and concomitant release of DMZn, and (2) the formation of thermodynamically stable ZnAl2O4 spinel. While etching in ZTO ALD could take place via cleavage of a C\N bond in a N(CH3)2 ligand of TDMASn followed by formation of DMZn and the associated C\Zn bond, this is likely to be hindered by a stronger C\N bond, with a bond strength of about 82 kcal/mol in HN(CH3)2 [64], as compared to the C\Zn bond having a bond strength of about 64 kcal/mol [65]. Notably, Al\C bond strength is nearly equal with the bond strength of C\Zn in DMZn [64]. Accordingly, the observed reduced ZTO ALD growth is likely to originate from reduced surface hydroxyl density and/or less reactive surface hydroxyls. Analysis of the ZTO growth rate behavior (Fig. 6) provides insight into the origin of the growth rate reduction as compared to the growth rates of the binaries. The growth rate drops particularly strongly, about 30%, with a cycle of SnOx followed by five cycles of ZnO within a super cycle (fSn = 0.17), and then at higher fSn converges toward the growth rate of pure SnOx ALD. This could be understood by a disruption of the ZnO ALD process by SnOx cycle(s), causing less deposition of ZnO, a hypothesis that is supported by compositional analysis of the films (vide infra). Using this approach, we calculated growth rates for ZTO films deposited at fSn values of 0.0, 0.17, 0.33, 0.50, 0.67, 0.83, and 1.0 by assuming that ZnO growth rate after the SnOx cycle(s) is first reduced and then linearly restored during the subsequent ZnO cycles. ZnO growth rates reduced by factors of 1/6, 2/6, 3/6, 4/6, and 5/6, and restoration of the ZnO growth rate in 1, 2, 3, 4, and 5 ZnO cycles were considered. Thus, the growth rates (rZTO) were calculated from
r ZTO ¼
Fig. 5. Plan-view, high-resolution SEM images for ZTO morphology deposited by varying the bilayer period from 800 to 4, as indicated. The films were deposited at 150 °C on Si(100) using a Sn/(Sn + Zn) cycle fraction of 0.75.
" # nZnO X 1 i−1 n þ rZnO nZnO −nZnO þ r ZnO R þ ð1−RÞ r ðnZnO þ nSnOx Þ SnOx SnOx n ZnO i¼1
ð1Þ
where r, n, n⁎ZnO, and R refer to the growth rate, the number of ZnO or SnOx ALD cycles in a super cycle, the number of ZnO ALD cycles with reduced growth rates in a super cycle, and a factor by which the ZnO growth rate is reduced, respectively. The best least squares fit with experimental growth rates was obtained for a ZnO growth rate reduced down to R = 3/6 of the pure ZnO growth rate and with restoration of the pure ZnO growth rate within 3 ZnO cycles (see Fig. 6), providing an estimate for the nucleation period for restoring the pure ZnO ALD growth. Note that these parameters that were derived from the data in Fig. 6 were also utilized to produce the results shown in Figs. 2 and 3, where in general correct trends for growth rates and ZTO composition as a function of bilayer period were produced. Notably, a nucleation period of 4–6 ALD cycles, of a similar magnitude to the value of n⁎ZnO determined here, was reported for ZnO in the ZnO/Al2O3 ALD on the basis of in situ QCM experiments [50]. More insight into the mechanistic
190
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
where r, n, ρ, and M refer to the growth rate, the number of ALD cycles in a super cycle, the density, and the molecular mass, respectively. n⁎ZnO refers to the number of ZnO ALD cycles with reduced growth rates in a super cycle. Note that Eq. (2) also generates the linearly interpolated compositions by setting n⁎ZnO = 0. Densities and molecular masses of both SnO and SnO2 were used for the SnOx values, but the results were practically equal and hence only results for SnO2 were used to calculate the compositions shown in Fig. 7. These values for the densities of the constituent binaries are as follows: ρZnO = 5.61 g/cm3, ρSnO = 6.45 g/cm3 and ρSnO2 = 6.95 g/cm3 [66]. Fig. 7 shows that excess tin with respect to the linearly interpolated tin composition fraction (dashed black curve) is observed for all fSn below 0.83. On the other hand, below this fSn the compositions from analytical modeling (dashed red curve) are in good agreement with the experimental values, providing additional evidence for the reduced ZnO growth after SnOx cycle(s) that forms the basis of this model. The analytically determined compositions deviate from the observed composition significantly only at fSn = 0.83, which suggests that mixing small numbers of ZnO cycles into a SnOx ALD process results in deposition of less Sn as compared to pure
SnOx ALD, and future surface chemistry investigations will provide insight into this behavior. Previous reports of as-deposited ZTO grown by ALD show that the films are largely amorphous [25,37–39]. To better understand the ZTO composition–crystallinity relationship, we characterized the structural properties of as-deposited films as a function of fSn. Fig. 8 shows diffraction patterns for the constituent binary oxides and intermediate ZTO compositions performed using grazing incidence X-ray diffraction (GIXRD). The diffraction pattern of the pure ZnO sample is characteristic of the hexagonal wurtzite phase, showing reflection of the (100), (002), (101), (110), and (201) planes at 31.86°, 34.4°, 36.8°, 56.6°, and 69.1° referenced to Cu-Kα, respectively. The peak broadening for these patterns is due to the limited resolution of the two dimensional GIXRD experiment. For fSn = 0.17 (~10% Sn), the film still shows patterns, albeit significantly broader and weaker, for reflection of the (101), (002), (101), and (201) planes of hexagonal wurtzite ZnO. The result suggests that mixing ZnO ALD with a small fraction of SnOx cycles begins to decrease the long-range order of deposited ZnO, and the origin of this finding represents an interesting target for a future detailed study. At higher cycle fractions, the films are X-ray amorphous, as is pure SnOx deposited from TDMASn and H2O [59], showing no reflections for ZnO, SnO, SnO2, Zn2SnO4 or ZnSnO3. In general, the consequences of the mixing of crystalline and amorphous materials on the atomic scale, as is the case in mixing ZnO and SnOx, are poorly-understood [67], and future mechanistic studies are necessary for elucidating the main factors causing the transition from crystalline to amorphous ZTO. For the pure ZnO sample and the ZTO sample with fSn = 0.17 that exhibit a hexagonal wurtzite structure, the growth orientation can be determined from GIXRD. The 2d GIXRD data give information on the relative intensities of the crystal planes that are oriented out of the plane of the substrate (Qz) and in the plane of the substrate (Qxy), as depicted in Fig. 9a). Fig. 9b–c) shows the 2d GIXRD data for pure ZnO and a fSn = 0.17 ZTO sample. For both samples, a strong 002 component is observed in-plane, indicating that the c-axis of the wurtzite crystal is growing parallel to the substrate. This is consistent with the SEM images below, in which elongated crystals are shown growing in the substrate plane. For pure ZnO, the 002 plane is also observed in the out-of-plane direction, indicating that a fraction of the crystals grow with a 002 fiber texture. The absence of the out-of-plane 002 feature for the fSn = 0.17 ZTO sample demonstrates the sensitivity of this growth direction toward incorporation of tin. In sum, the XRD data demonstrates a crystalline-to-amorphous transition when mixing ZnO ALD with SnOx cycles, with the deposited films being amorphous above tin cycle fractions of 0.17 (~10% Sn). The XRD results are further supported by film morphologies observed by SEM, with most of the films resembling amorphous SnOx, as illustrated in Fig. 10. This finding is in agreement with previous work on ZTO ALD [37].
Fig. 7. ZTO compositions determined by ICP-OES and from an analytical model described by Eq. (2) as a function of Sn/(Sn + Zn) cycle fraction, fSn. The black dashed line gives the compositions linearly interpolated from the growth rates and densities of the binaries. Films were deposited on Si(100) at 150 °C using a bilayer period of Λ = 6 for a total of 100 super cycles.
Fig. 8. Grazing incidence X-ray diffraction (GIXRD) patterns of ZTO films deposited using Sn/(Sn + Zn) cycle fractions from 0 to 1. Films were deposited on Si(100) at 150 °C using a bilayer period of Λ = 6 for a total of 100 super cycles. The Sn/(Sn + Zn) cycle fractions are indicated on the right.
details of the ZTO ALD growth mechanism will be provided by future theoretical and experimental investigations. Reduced ZnO growth rates after SnOx ALD cycle(s) should be reflected in the ZTO composition as an excess of Sn with respect to the composition estimated from the growth rates and densities of the binaries, and hence compositional analysis of the films provided the means to verify this. The fractional tin compositions determined experimentally by ICP-OES for the ZTO films deposited using a constant number of ALD cycles in a super cycle are shown in Fig. 7 as a function of fSn. Also shown in Fig. 7 for comparison are two predicted composition curves. The first (referred to as “analytical” in Fig. 7) is based on the analytical model introduced earlier, in which a reduction in ZnO growth rate on SnOx is considered. These compositions are calculated using Eq. (2), which is derived from Eq. (1) by taking into account the densities of the binaries. A simpler predicted composition curve calculated solely from the linearly interpolated growth rates (see Fig. 6) and densities of the binaries is also included for comparison (referred to as “linearly interpolated”). The analytical model compositions were calculated using equation
nSnOx
cSn ¼ nSnOx
ρSnOx r SnOx MSnOx
ρSnOx r SnOx
MSnOx " # nZnO X ρZnO r ZnO 1 1 i−1 þ þ nZnO −nZnO þ 2 2 nZnO M ZnO i¼1 ð2Þ
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
191
Fig. 9. a) Cartoon depicting the geometry of the in-plane (Qxy) and out-of-plane (Qz) components of the reciprocal lattice vector. b) 2d GIXRD pattern for pure ZnO exhibiting 002 growth both in-plane and out-of-plane. c) 2d GIXRD pattern for ZTO film with fSn = 0.17.
Fig. 10. High resolution SEM images of ZTO films deposited on Si(100) at 150 °C using Sn/(Sn + Zn) cycle fractions from 0 to 1. The cycle fractions, together with ICP-OES determined tin compositions (in parentheses) are shown in the images. The scale bar is 500 nm.
192
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
Spectroscopic ellipsometry was used to characterize the optical properties of the ZTO films grown on Si(100) as a function of fSn, and thus composition. The results are shown in Fig. 11. We believe that this is the first comprehensive report of optical constants over a full range of compositions for ZTO by thermal ALD. An oscillator model was used to represent the optical properties of the ZTO films. This included a polynomial spline function (the psemi-m0 oscillator [68]), a Lorentz oscillator and poles on each side of the spectrum to represent the optical constants of the films in a Kramers–Kronig consistent manner. A multi-sample analysis was used to fit a common set of oscillator parameters to samples sharing the same Sn cycle fraction. However, film thicknesses were allowed to vary individually for each sample in the analysis. As a control for the ellipsometry results, UV–vis extinction measurements were performed at normal incidence for films deposited on fused silica quartz in the same ALD runs, as shown in Fig. 11c). The general trends and features including absorption onset show good agreement with the extinction coefficients obtained from VASE. For pure ZnO, the extinction coefficient displays a characteristic, sharp and direct type bandgap (Fig. 11b). The introduction of Sn into ZnO appears to increase the absorption over the entire visible range. This could, for instance, be due to defects associated with a more disordered phase, and/or free carrier-like absorption as the Sn act as an n-type dopant in the ZnO [69]. For intermediate ZTO compositions, the refractive index shown in Fig. 11a) drops with increasing Sn content. When completely eliminating the ZnO content, however, a rather
dramatic increase of the refractive index is observed for the pure SnOx-phase. The higher reflectance expected from this change is also consistent with the increased extinction (Fig. 11c). To analyze the bandgap dependence on composition, extrapolations of linear segments in the absorption onset region were performed as shown in Fig. 12. As it is not clear what bandgap type is most appropriate for the intermediate compositions, fits to both direct and indirect/ amorphous type bandgaps are presented in Fig. 12a) and b), respectively. The resulting bandgap values are presented as a function of cycle fraction and composition in Fig. 12c) and d), respectively. The direct bandgap of ZnO was found to be 3.3 eV, in line with previous literature. The Tauc plot analysis for the amorphous type bandgap suggests a nontrivial dependence on cycle fraction and composition (Fig. 12c and d), where the introduction of a low Sn concentration (10%) results in a bandgap minimum of 2.8 eV. The bandgap then widens to a maximum of ~ 3.4 eV for a ZTO film with 50% Sn, and then again a narrows to ~2.9 eV for the pure SnOx-phase. Interestingly, Kapilashrami et al. observed a similar dependence for amorphous ZTO ALD films by soft Xray spectroscopy [39]. The increased refractive index observed for the pure SnOx film is concomitant with a reduction of both type of bandgaps, but it is not clear what the primary cause for these changes is. It was recently suggested that the optical properties of SnOx depend critically on the microstructure of the film, which may entail grain sizes below the detection limit of XRD [59]. If this is causing the strong dependence of the optical properties on composition here, it implies that infrequent interruptions of the SnOx growth by single ALD cycles of ZnO (fSn = 0.83) has a significant effect on the resulting microstructure. Overall, the non-trivial dependence of optical properties on ZTO composition is likely affected by the interplay between crystal structure, dopant concentrations, and microstructural changes of the material. 4. Conclusions We report the structural and optical properties of zinc tin oxide (ZTO) deposited by thermal ALD by systematically varying the relative fractions of SnOx and ZnO ALD cycles and the bilayer period in ZTO ALD. Non-trivial dependences on these parameters were observed for the ALD growth rates, structures, and properties. Namely, growth rates decreased with increasing amounts of tin in ZTO, and were lower than what would be expected from a linear interpolation of the growth rates of the binary oxides. An analytical model accounting for the decreased growth was developed to provide insight into the ZTO ALD growth rate process. X-ray diffraction showed that the crystallinity of the pure ZnO is largely lost when ZTO is deposited using a tin cycle fraction higher than 0.17 (~10% Sn). The optical properties, especially the amorphous type bandgap size, shows a non-trivial dependence on ZTO composition which may be understood as an interplay between dopant concentrations and microstructural changes of the material. The mild ZTO ALD process reported herein may be easily adapted for use in thin film solar cells and in thin film transistors. Because the ALD processes for these binary metal oxides can be performed at nearroom temperature [59], this ZTO ALD process may also be used in systems where low-temperature processing is required, such as for biological and organic electronic applications. Acknowledgments
Fig. 11. Spectroscopic ellipsometry results for a) refractive index and b) extinction coefficient vs. wavelength for ZTO films deposited on Si(100) at 150 °C using Sn/(Sn + Zn) cycle fractions from 0 to 1. For comparison, optical extinction vs. wavelength from UV–vis transmission measurements of the ZTO on fused silica is shown in c). The cycle fractions are indicated in the legends.
This work was supported by the Center for Nanostructuring for Efficient Energy Conversion, an Energy Frontier Research Center (EFRC) funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences (BES) under Award No. DE-SC0001060. The X-ray diffraction studies were supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award #DE-SC0004782. MNM acknowledges the Bell Labs Graduate Fellowship Research Program for graduate fellowship support. CH is grateful for financial support from the Marcus and Amalia Wallenberg Foundation. JTT gratefully
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
193
Fig. 12. VASE multi-sample analyses of bandgaps for ZTO films deposited on Si(100) at 150 °C using Sn/(Sn + Zn) cycle fractions from 0 to 1. The bandgaps are found by extrapolating linear portions of the function (αhv)x to zero. In this expression, hv is the photon energy and α the absorption coefficient (α = 4πk / λ with k the extinction coefficient of the material, as determined by VASE). For the direct type bandgap fits in a), the exponent x = 2, while for the indirect or amorphous type bandgap in b), x = 0.5. The results are summarized versus c) the Sn cycle fraction and d) composition. The cycle fractions are indicated in the legends.
acknowledges the Academy of Finland (Grant 256800/2012) and the Finnish Cultural Foundation for financial support. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]
T. Minami, Semicond. Sci. Technol. 20 (2005) S35. I. Hamberg, C.G. Granqvist, J. Appl. Phys. 60 (1986) R123. R.G. Gordon, MRS Bull. 25 (2000) 52. C.A. Hoel, T.O. Mason, J. Gaillard, K.R. Poeppelmeier, Chem. Mater. 22 (2010) 3569. E. Fortunato, D. Ginley, H. Hosono, D.C. Paine, MRS Bull. 32 (2007) 242. T. Minami, H. Nanto, S. Shooji, S. Takata, Thin Solid Films 111 (1984) 167. G. Frank, H. Kostlin, Appl. Phys. A: Mater. Sci. Process. 27 (1982) 197. P. Nath, R.F. Bunshah, B. Basol, O.M. Staffsud, Thin Solid Films 72 (1980) 463. I.A. Rauf, Mater. Lett. 18 (1993) 123. A.C. Tolcin, U.S. Geol. Surv. (2012) 74. T. Minami, Thin Solid Films 516 (2008) 5822. J. Montero, C. Guillen, J. Herrero, Thin Solid Films 519 (2011) 7564. J. Perrenoud, L. Kranz, S. Buecheler, F. Pianezzi, A.N. Tiwari, Thin Solid Films 519 (2011) 7444. H. Liu, V. Avrutin, N. Izyumskaya, U. Ozgur, H. Morkoc, Superlattice. Microst. 48 (2010) 458. T. Minami, S. Takata, H. Sato, H. Sonohara, J. Vac. Sci. Technol. A 13 (1995) 1095. Y. Hayashi, K. Kondo, K. Murai, A. Moriga, I. Nakabayashi, H. Fukumoto, K. Tominaga, Vacuum 74 (2004) 607. T. Moriga, Y. Hayashi, K. Kondo, Y. Nishimura, K. Murai, I. Nakabayashi, H. Fukumoto, K. Tominaga, J. Vac. Sci. Technol. A 22 (2004) 1705. J. Perkins, J. del Cueto, J. Alleman, C. Warmsingh, B. Keyes, L. Gedvilas, P. Parilla, B. To, D. Readey, D. Ginley, Thin Solid Films 411 (2002) 152. D. Young, D. Williamson, T. Coutts, J. Appl. Phys. 91 (2002) 1464. H. Enoki, T. Nakayama, J. Echigoya, Phys. Status Solidi A 129 (1992) 181. E. Cetinorgu, S. Goldsmith, Z. Barkay, R.L. Boxman, J. Phys. D Appl. Phys. 39 (2006) 5245. S. Vijayalakshmi, S. Venkataraj, M. Subramanian, R. Jayavel, J. Phys. D Appl. Phys. 41 (2008) 035505. A. Kurz, M.A. Aegerter, J. Sol-Gel Sci. Technol. 31 (2004) 267. Z. Hu, S. Gao, Chem. Geol. 253 (2008) 205. A. Hultqvist, C. Platzer-Bjorkman, U. Zimmermann, M. Edoff, T. Torndahl, Prog. Photovoltaics 20 (2012) 883. H.Q. Chiang, J.F. Wager, R.L. Hoffman, J. Jeong, D.A. Keszler, Appl. Phys. Lett. 86 (2005) 013503. Y.-. Chang, D.-. Lee, G.S. Herman, C.-. Chang, Electrochem. Solid-State 10 (2007) H135. S.K. Park, Y. Kim, H. Kim, J. Han, Electrochem. Solid-State 12 (2009) H256. G.Z. Zang, J.F. Wang, H.C. Chen, W.B. Su, W.X. Wang, P. Qi, C.M. Wang, J. Mater. Sci. 39 (2004) 3537.
[30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59]
P.R. Bueno, J.A. Varela, E. Longo, J. Eur. Ceram. Soc. 28 (2008) 505. S. Yu-Sheng, Z. Tian-Shu, Sensors Actuators B Chem. 12 (1993) 5. D. Kovacheva, K. Petrov, Solid State Ionics 109 (1998) 327. S. Baruah, J. Dutta, Sci. Technol. Adv. Mater. 12 (2011) 013004. W.M. Kim, J.H. Ko, I.H. Kim, D. Kim, K.S. Lee, T.S. Lee, B. Cheong, Appl. Surf. Sci. 253 (2007) 7398. Y. Jeong, K. Song, D. Kim, C.Y. Koo, J. Moon, J. Electrochem. Soc. 156 (2009) H808. R.G. Gordon, Final Technical Report, Harvard University, National Renewable Energy Laboratory, 2001. W.S. Choi, J. Korean Phys. Soc. 57 (2010) 1472. A. Hultqvist, M. Edoff, T. Törndahl, Prog. Photovoltaics Res. Appl. 19 (2011) 478. M. Kapilashrami, C.X. Kronawitter, T. Törndahl, J. Lindahl, A. Hultqvist, W. Wang, C. Chang, S.S. Mao, J. Guo, PCCP 14 (2012) 10154. J. Heo, S.B. Kim, R.G. Gordon, Appl. Phys. Lett. 101 (2012) 113507. M.K. Jayaraj, K.J. Saji, K. Nomura, T. Kamiya, H. Hosono, J. Vac. Sci. Technol. B 26 (2008) 495. D.L. Young, H. Moutinho, Y. Yan, T.J. Coutts, J. Appl. Phys. 92 (2002) 310. X. Fu, X. Wang, J. Long, Z. Ding, T. Yan, G. Zhang, Z. Zhang, H. Lin, X. Fu, J. Solid State Chem. 182 (2009) 517. J. Zeng, M. Xin, K. Li, H. Wang, H. Yan, W. Zhang, J. Phys. Chem. C 112 (2008) 4159. M. Miyauchi, Z. Liu, Z. Zhao, S. Anandan, K. Hara, Chem. Commun. 46 (2010) 1529. S.M. George, Chem. Rev. 110 (2010) 111. M. Leskela, M. Ritala, Thin Solid Films 409 (2002) 138. M. Leskela, M. Ritala, Angew. Chem. Int. Ed. 42 (2003) 5548. H. Yuan, B. Luo, D. Yu, A. Cheng, S.A. Campbell, W.L. Gladfelter, J. Vac. Sci. Technol. A 30 (2012) 01A138. J.W. Elam, S.M. George, Chem. Mater. 15 (2003) 1020. T. Dhakal, D. Vanhart, R. Christian, A. Nandur, A. Sharma, C.R. Westgate, J. Vac. Sci. Technol. A 30 (2012) 021202. C. Gong, Y. Choi, H. Kim, C. Park, H. Park, J.G. Jang, H.J. Chang, G.Y. Yeom, J. Vac. Sci. Technol. A 31 (2013) 01A101. J.Y. Kim, Y. Choi, H. Park, S. Golledge, D.C. Johnson, J. Vac. Sci. Technol. A 28 (2010) 1111. P.R. Chalker, P.A. Marshall, S. Romani, M.J. Rosseinsky, S. Rushworth, P.A. Williams, J. Buckett, N. McSporran, J. Ridealgh, MRS Proc. 1315 (2011) 39. K. Saito, Y. Hiratsuka, A. Omata, H. Makino, S. Kishimoto, T. Yamamoto, N. Horiuchi, H. Hirayama, Superlattice. Microst. 42 (2007) 172. J.W. Elam, D.A. Baker, A.B.F. Martinson, M.J. Pellin, J.T. Hupp, J. Phys. Chem. C 112 (2008) 1938. C. Platzer-Bjorkman, T. Torndahl, A. Hultqvist, J. Kessler, M. Edoff, Thin Solid Films 515 (2007) 6024. J. Heo, Y. Liu, P. Sinsermsuksakul, Z. Li, L. Sun, W. Noh, R.G. Gordon, J. Phys. Chem. C 115 (2011) 10277. M.N. Mullings, C. Hägglund, S.F. Bent, J. Vac. Sci. Technol. A 31 (2013) 061503.
194
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
[60] J.R. Bakke, J.T. Tanskanen, C. Hägglund, T.A. Pakkanen, S.F. Bent, J. Vac. Sci. Technol. A 30 (2012) 01A135. [61] J.T. Tanskanen, J.R. Bakke, T.A. Pakkanen, S.F. Bent, J. Vac. Sci. Technol. A 29 (2011) 031507. [62] L. Wachnicki, M. Lukasiewicz, B. Witkowski, T. Krajewski, G. Luka, K. Kopalko, R. Minikayev, E. Przezdziecka, J.Z. Domagala, M. Godlewski, E. Guziewicz, Phys. Status Solidi B 247 (2010) 1699. [63] T. Törndahl, C. Platzer-Björkman, J. Kessler, M. Edoff, Prog. Photovoltaics Res. Appl. 15 (2007) 225.
[64] Y. Luo, Comprehensive Handbook of Chemical Bond Energies, CRC Press, University of Science and Technology of China, Hefei, 2007. [65] R.L. Jackson, Chem. Phys. Lett. 163 (1989) 315. [66] In: W.M. Haynes (Ed.), CRC Handbook of Chemistry and Physics, 94th edition, CRC Press, 2013. [67] D. Wang, A. Fernandez-Martinez, Science 337 (2012) 812. [68] Guide to using WVASE32, J. A. Woollam Co., Inc, Lincoln, NE., 2008 [69] S. Venkataraj, S. Hishita, Y. Adachi, I. Sakaguchi, K. Matsumoto, N. Saito, H. Haneda, N. Ohashi, J. Electrochem. Soc. 156 (2009) H424.
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Thin film characterization of zinc tin oxide deposited by thermal atomic layer deposition Marja N. Mullings a, Carl Hägglund a, Jukka T. Tanskanen a, Yesheng Yee b, Scott Geyer a, Stacey F. Bent a,⁎ a b
Department of Chemical Engineering, Stanford University, Stanford, CA 94305, USA Department of Electrical Engineering, Stanford University, Stanford, CA 94305, USA
a r t i c l e
i n f o
Article history: Received 1 August 2013 Received in revised form 23 January 2014 Accepted 24 January 2014 Available online 31 January 2014 Keywords: Zinc tin oxide Atomic layer deposition Thin film Transparent conductive oxide
a b s t r a c t Zinc tin oxide (ZTO) thin films are of interest for many applications, including transparent conducting oxides and buffer layers in thin film solar cells. In this work, the ability of atomic layer deposition (ALD) to control both thickness and composition of the ternary material is explored and the resulting film properties are characterized. ZTO was deposited at 150 °C by alternating growth of tin oxide (SnOx) and zinc oxide (ZnO) using tetrakis(dimethylamido)tin and diethyl zinc as metal precursors and water as the oxygen source. The growth behavior of ZTO was systematically examined as a function of the relative fraction of SnOx to ZnO ALD cycles and their bilayer period. Variable angle spectroscopic ellipsometry (VASE) showed that the ZTO growth rate was strongly reduced as compared to the growth rates of the binaries, especially at low bilayer periods and low tin cycle fractions. Inductively-coupled plasma optical emission spectroscopy showed that the composition, together with other properties of ZTO films, was not simply proportional to the binary ALD cycle fractions utilized in the deposition. Structurally, the deposited films were shown by X-ray diffraction to be amorphous for tin content exceeding 10%, for which only smooth features were observed by scanning electron microscopy. Optically, VASE revealed a minimum refractive index in the visible range for intermediate compositions and a maximum for the pure SnOx phase. Conversely, the ZTO bandgap was maximized for intermediate metal compositions near ~50%, and converged to the direct ZnO bandgap of 3.3 eV for compositions with decreasing amounts of Sn in the films. Linear extrapolations gave a range of 2.9 to 4.0 eV for SnOx bandgap depending on the bandgap type (indirect or direct). © 2014 Elsevier B.V. All rights reserved.
1. Introduction Transparent conducting oxides (TCOs) are wide bandgap (≥3 eV), transmissive to visible light (≥ 80%), and highly conductive (≥ 103 S cm−1) metal oxides with many applications in consumer electronics and in photovoltaics [1–4]. The oxides of indium, tin, and zinc are the most commercially-important TCOs [4,5]. These binary TCOs are typically doped with an impurity to improve stability at high temperatures [6]. Within the indium/tin/zinc oxide family, indium tin oxide (ITO) remains the industry-standard for high-end TCOs because of its high conductivity (≥104 S cm−1) [7–9]. However, the scarcity of indium, with an abundance in the earth's upper crust of about 0.05 ppm [10], coupled with the increasing demand in consumer electronic and photovoltaic markets has led to rising costs [11], which motivates the need to find alternative TCO materials. Impurity-doped or alloyed zinc oxides are interesting candidates for this purpose, where mixtures with aluminum (AZO) or gallium (GZO) are well known examples [1,12–14]. While AZO and GZO exhibit attractive properties as inexpensive indium-free alternatives, other options such as zinc tin oxide (ZTO) are also being studied as an inexpensive ⁎ Corresponding author. E-mail address: [email protected] (S.F. Bent). 0040-6090/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.tsf.2014.01.068
indium-free TCO [1,11,15–23]. Notably, the upper crustal abundances of Zn and Sn are around 70 and 5.5 ppm [24], these values being significantly higher than the 0.05 ppm abundance of indium. In addition to its use as a TCO and buffer layer in high efficiency solar cells [25], ZTO has been used in thin-film transistors [26–28] and it has potential in varistor applications as well [29,30]. Despite its promise, a review of the literature suggests little agreement among various structural, optical, and electrical properties of ZTO, motivating a more detailed characterization of this material. In general, there is an agreement that ZTO forms two crystal structures. The metastable ZnSnO3 phase, known as zinc metastannate, has two polymorphs — a face-centered perovskite [31] and an ilmenite [32] structure. The other structure, Zn2SnO4, is known as zinc orthostannate and has a cubic spinel structure that is stable under ambient conditions [33]. However, the correlation between composition and crystallinity is not well-understood. This lack of understanding applies to many other ternary oxides as well, and it originates from the mixing of binaries with potentially different crystal structures, densities, and/or stoichiometries to produce the ternary material. The most extensively used deposition technique for ZTO has been sputtering [6,15,16,18–20,34]. During crystallization, ZTO transforms to ZnSnO3 from 300 to 500 °C and to Zn2SnO4 for temperatures exceeding 600 °C. At these high temperatures, however, obtaining pure Zn2SnO4
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
is difficult, as mixed phases including both forms of ZTO and SnO2 are formed [33]. Other techniques for ZTO deposition have included sol– gel spin coating [27,28,35], chemical vapor deposition [36], and only very recently, atomic layer deposition (ALD) [25,37–40]. A lack of agreement on the ZTO crystal structure as a function of composition may explain the variation in reports on its optical properties. One study (using pulsed layer deposition) reported bandgaps for amorphous ZTO of 2.80 eV for ZnSnO3 and 2.86 eV for Zn2SnO4 [41]. Young et al. found crystalline Zn2SnO4 to have a fundamental bandgap of 3.35 eV that was widened to 3.89 eV by a Burstein–Moss shift due to high carrier concentration [42]. For ZTO synthesized by hydrothermal methods, the optical bandgap for crystalline Zn2SnO4 has been reported with great variation in the range of 3.3 to 3.9 eV [43,44]. The bandgap for crystalline ZnSnO3 has not been rigorously characterized experimentally, but one report presents a value of ~3.9 eV [45], which is significantly higher than the gap of 2.80 eV reported for the amorphous phase (see above). Most previous studies using ZTO focused on device performance instead of film characterization, resulting in few comprehensive reports of material and electrical properties as a function of composition. Because ALD allows one to carefully dope or alloy material in an atomic-level controlled fashion with uniform and conformal growth [46–48], it provides an opportunity to systematically analyze ZTO properties by fine tuning the composition. In addition to ZTO, ALD has been used to deposit several other ternary metal oxides including AZO, [49–53] GZO [54,55], ITO [56], (Zn,Mg)Ox [57], and (Sn,Al)Ox [58]. In this study, we demonstrate a reproducible thermal ZTO ALD process and determine the composition, optical properties, and structural characteristics of the resulting thin films by variable angle spectroscopic ellipsometry (VASE), X-ray photoelectron spectroscopy (XPS), inductively-coupled plasma optical emission spectroscopy (ICP-OES), scanning electron microscopy (SEM), X-ray diffraction (XRD) and ultraviolet–visible spectroscopy (UV–vis). An analytical model based on analysis of the ZTO growth rates is developed for understanding the main features of the ZTO growth characteristics. Relationships between deposition parameters, namely bilayer period and relative fraction of SnOx to ZnO ALD cycles (cycle fraction), and ZTO growth rate, composition, structure, and optical properties are established and discussed in the context of previous studies on ZTO. 2. Experimental methods 2.1. ZTO ALD process The ALD experiments were carried out in a custom-built, viscousflow, hot-wall reactor previously described [59]. Diethyl zinc (DEZn; Sigma-Aldrich) and tetrakis(dimethylamido)tin (TDMASn; Strem, N99% purity) were used as precursors with deionized water as the counter-reactant. Precursors were dosed from 10 cm3 stainless steel sample vials using computer controlled air-actuated valves. The water and DEZn sample vials were kept at room temperature, while the TDMASn vial was heated to 45 °C during depositions. A positive temperature gradient was maintained toward the reaction zone to prevent precursor condensation. During the ALD process, excess precursor was purged with nitrogen for half-reaction separation and reaction completion. For ZnO ALD, we express the pulse time sequences as z1:z2:z3:z4, where z1 and z3 correspond to the pulse lengths of DEZn and water, respectively, and z2 and z4 correspond to N2 purge times following these doses. A typical ZnO ALD pulse time sequence was 1:15:1:30 s. The purge times of z2 = 15 s and z4 = 30 s were chosen based on the return of the precursor concentrations to their baseline values as monitored by in situ mass spectrometry, as well as previously-found saturation in this range for the ZnO system [50,60]. For SnOx ALD, we similarly express the ALD sequences as t1:t2:t3:t4, where t1 and t3 correspond to the pulse lengths of TDMASn and water, respectively, and t2 and t4 correspond to N2 purge times following these doses. A typical timing sequence
187
was 1:30:2:30 s which corresponded to saturated ALD of SnOx as previously described [59]. Designing a reproducible ZTO ALD process first required optimization of the binary ZnO and SnOx ALD processes in our reactor. It is convenient that there is overlap in the ALD temperature windows for these binary metal oxide processes. It was reported previously that within the ZnO ALD temperature window, a maximum growth rate of about ~ 2.0 Å/cycle occurs at 150 °C [61], which is close to the optimized growth rate of 1.85 Å/cycle for our reactor. The ALD of SnOx, reported elsewhere [59], results in SnOx growth rate of about 0.70 Å/cycle at 150 °C. For the optimized SnOx ALD system, the as-deposited film is amorphous, whereas as-deposited ZnO grown by ALD from DEZn and water at 150 °C has a crystalline hexagonal wurtzite structure [62]. The ALD super cycle is defined as the minimum sequence of SnOx and ZnO cycles that is repeated over the course of the ZTO ALD process. A corresponding super cell results from the stack of layers produced by one super cycle. To characterize the ALD system, we define t and z, the total number of SnOx and ZnO ALD cycles in one ZTO ALD super cycle, respectively. Using these parameters, illustrated in Fig. 1, the system can be characterized by the bilayer period Λ, defined as the number of cycles in one super cycle Λ = (t + z), and the relation for the total number of ZnO and SnOx cycles in a ZTO ALD process, referred to as cycle fraction. If a total number of ALD super cycles S are performed, then the total number of ALD cycles in the deposition is given by S × Λ . Substrates for ZnO, SnOx, and ZTO ALD were n-type Si(100) with phosphorous dopant and a resistivity of 1.0–5.0 Ω cm (WRS Materials) and fused silica quartz (ChemGlass). Substrates were pre-cleaned using a UV-ozone (PSD Benchtop UV-Ozone Cleaner) to remove organic contamination, after which the thin films were deposited onto the native Si oxide. 2.2. Characterization A PHI VersaProbe 5000 scanning X-ray photoelectron spectrometer with Al-Kα radiation was used to survey film elemental content. The C1s peak at 285.0 eV was used as a reference. After initial survey scans, adventitious carbon was removed by Ar+ sputtering for 12 s (1 keV, 0.5 μA, 1 mm × 1 mm at 45° incidence from the sample). A Thermo Scientific ICAP 6300 Duo View spectrometer was used to perform inductively-coupled plasma optical emission spectroscopy (ICP-OES) on ZTO deposited on Si(100). For these measurements, films were dissolved in aqua regia (3:1 concentrated HCl:HNO3) and sonicated for at least 6 h before characterization. X-ray diffraction (XRD)
Fig. 1. Schematic of ZTO super cycle formalism visualized as a lamellar stack of ALD cycles of constituent binary oxides. One super cycle includes a certain number of SnOx cycles, t, and ZnO cycles, z, resulting in a bilayer period Λ = (t + z).
188
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
Table 1 Sn composition fractions (cSn) and growth rates for ZTO films for different bilayer periods with a constant Sn/(Sn + Zn) cycle fraction of 0.75 and a constant total number of ZTO ALD cycles of 800. Total number of super cycles (S), bilayer period (Λ), the number of SnOx ALD cycles in a super cycle (t), and the number of ZnO ALD cycles in a super cycle (z) used to deposit the films are also given. The growth rates and compositions were determined by VASE and ICP-OES, respectively. The films were deposited at 150 °C on Si(100). S
Λ
t
z
cSn
Growth rate (Å/cycle)
200 100 20 5 2 1
4 8 40 160 400 800
3 6 30 120 300 600
1 2 10 40 100 200
0.35 0.46 0.36 0.37 0.38 0.42
0.52 0.55 0.79 0.94 0.97 1.06
measurements of ZTO on fused-silica quartz were performed as a function of ALD cycle fractions using the PANalytical X'Pert PRO system in parallel beam mode with Cu Kα radiation at 45 kV and 40 mA. Grazing incidence X-ray diffraction (GIXRD) was also performed on ZTO films deposited on Si(100) using 12.7 keV X-rays that were incident on the sample at 0.48°. Diffracted X-rays were collected with an MAR34 twodimensional imagining detector calibrated using a LaB6 reference. Film thickness and optical properties of ZTO deposited on Si(100) were determined using a J.A. Woollam M2000 Variable Angle Spectroscopic Ellipsometer (VASE) at 65, 70 and 75° angles of incidence and wavelengths ranging from 210 to 1688 nm. In the analysis, the optical properties of the films were represented by a sum of oscillator terms, described in more detail below. A model with a single homogeneous layer (to be referred to as a single layer model) was found accurate for finely layered ZTO, but for the most coarsely layered stacks, the root mean squared error was lower if taking the periodic structure into account with a two layer unit cell corresponding to the super cycle of ZnO and SnOx used in the ALD sequence. We will refer to the latter as a super cycle model. Plan view images of ZTO film morphology on Si(100) were examined using a FEI Magellan 400 XHR Scanning Electron Microscope with FEG source (5 kV and 25 pA). Optical extinction spectra of films deposited on quartz were recorded using a Varian Cary 6000i UV–vis-NIR spectrometer.
Fig. 3. Tin composition fraction in ZTO films from ICP-OES (markers) and from an analytical model (Eq. (1)) as a function of bilayer period, Λ . The black dashed line gives the composition linearly interpolated from the growth rates and densities of the binaries. The films were deposited at 150 °C on Si(100) by using a Sn/(Sn + Zn) cycle fraction of 0.75.
Atomic layer deposition allows the sequential dosing of an arbitrary number of SnOx and ZnO cycles in any desired ratio to deposit ZTO. We hypothesize that the more finely spaced these dosing sequences are, the more laminated the film will be and the less the film will resemble its constituent binaries in growth rate and material properties. The fineness of the cycling can be described by the bilayer period, Λ .
The ZTO ALD super cycle formalism provides a systematic approach to determining the effects of the bilayer period on the growth rate and other properties (see the Experimental methods section). To study the effect of Λ without changing the cycle fraction, we fixed the total number of ZnO and SnOx cycles to 200 and 600, respectively. Fixing these õparameters yielded a tin cycle fraction fSn = Sn/(Sn + Zn) = 0.75, which based on the binary growth rates and densities would result in an intermediate (cSn ~ 43%) overall composition. We then varied the super cell size (t + z) in order to sample different Λ . Table 1 summarizes the effect of varying bilayer period on ZTO growth rates and compositions, and these results are also illustrated in Figs. 2 and 3, respectively. In the figures, theoretical results produced by an analytical model of ZTO growth are also illustrated (see Section 3.2). The data in Table 1 and Fig. 2 show that the growth rate of the intermixed system increases with increasing Λ . When a weighted average of the binary growth rates is calculated for fSn = 0.75, a value of 0.99 Å/cycle results, which is close to the growth rates of the least lamellar films with large Λ values. In contrast, for the highly lamellar films, the growth behavior deviates significantly from this weighted average (labeled as “linearly interpolated” in Fig. 2). This reduced growth rate suggests that at least one of the materials does not grow as well on its counterpart as on itself. In general, as shown in Table 1 and Fig. 3, the overall tin content by ICP-OES varies slightly from the value of cSn = 0.43 that results from calculating the composition for fSn = 0.75 using the binary growth rates and densities (see discussion above). We will discuss the behavior of the ZTO growth rate and composition later in the context of the analytical model developed to understand the ZTO ALD growth characteristics. X-ray photoelectron spectroscopy (XPS) was performed to verify the elemental composition of the ZTO films deposited by varying Λ . Fig. 4 shows an XPS survey scan for a film with a large bilayer period of
Fig. 2. ZTO ALD growth rates from VASE and from an analytical model (Eq. (1)) as a function of the bilayer period, Λ . The black dashed line gives the growth rate linearly interpolated from the growth rates of the binaries. The films were deposited at 150 °C on Si(100) by using a Sn/(Sn + Zn) cycle fraction of 0.75.
Fig. 4. X-ray photoelectron spectroscopy survey scan of a ZTO film deposited on Si(100) at 150 °C after Ar+ sputtering for 12 s. The ZTO film was deposited using a bilayer period, Λ , of 400 and a Sn/(Sn + Zn) cycle fraction of 0.75.
3. Results 3.1. ZTO ALD as a function of bilayer period
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
189
Λ = 400. No contaminants are observed after removal of adventitious surface carbon by Ar+ sputtering, which is representative for the other films investigated here. As this also holds for the binary films [59], we conclude that this thermal ZTO ALD process deposits films of high purity. We also studied film morphology as a function of Λ using SEM (Fig. 5). The films appeared to be smooth and without large and distinct crystalline domains for all bilayer periods studied, with a morphology resembling that of pure and X-ray amorphous SnOx [59]. 3.2. ZTO properties as a function of cycle fraction We now turn to the influence of the cycle fraction on ZTO growth and properties. Because film characteristics vary significantly as a function of bilayer period, Λ , we fixed this value to 6 to examine the role of cycle fraction – which then determines the film composition – on ZTO properties. With a fixed total number of super cycles S = 100, which corresponds to a total number of ALD cycles A = 600, the tin cycle fraction, fSn, was sampled in multiples of 1/6. Note that internal cycling, which would increase the film lamination, was not varied here. For illustration, the film corresponding to fSn = 0.50 was deposited using a cycling of zzzttt in a super cycle instead for instance a cycling of ztztzt. The ZTO growth rates are plotted against the fSn in Fig. 6. The two endpoints on the plot, fSn = 0 and fSn = 1.0, correspond to the two binary systems, ZnO and SnOx, respectively. Based on a linear interpolation of the two binary growth rates (black dashed curve), the observed growth rate for the intermixed system is lower than expected. The ZTO ALD growth rate behavior as a function of fSn resembles the growth behavior of the mixed ZnO/Al2O3 (AZO) ALD process, an analog to ZTO ALD, reported by George and Elam [50]. Notably, similar ALD growth characteristics have been observed also for other ternary materials, including Zn1 − xMgxO [63], and (Sn,Al)Ox [58]. George and Elam deposited AZO films from DEZn and trimethylaluminum (TMA) with H2O as the oxygen source. Supported by detailed quartz crystal microbalance (QCM) experiments, the authors discuss potential causes for the reduced AZO growth, and these include etching of surface Zn by TMA, deficiency of hydroxyl groups on the binary surfaces and surface hydroxyls that are less reactive to the TMA and DEZn reactants.
Fig. 6. Film growth characteristics determined by VASE and an analytical model described by Eq. (1) as a function of Sn/(Sn + Zn) cycle fraction, fSn. The black dashed line gives the growth rates linearly interpolated from the growth rates of the binaries. Films were deposited on Si(100) at 150 °C using a bilayer period of Λ = 6 for a total of 100 super cycles.
The same arguments, which have been discussed in the context of Zn1 − xMgxO ALD as well, apply for ZTO ALD, except for the etching. In the ZnO/Al2O3 system, etching was attributed to two processes: (1) the transfer of CH3 ligands from TMA to surface Zn and concomitant release of DMZn, and (2) the formation of thermodynamically stable ZnAl2O4 spinel. While etching in ZTO ALD could take place via cleavage of a C\N bond in a N(CH3)2 ligand of TDMASn followed by formation of DMZn and the associated C\Zn bond, this is likely to be hindered by a stronger C\N bond, with a bond strength of about 82 kcal/mol in HN(CH3)2 [64], as compared to the C\Zn bond having a bond strength of about 64 kcal/mol [65]. Notably, Al\C bond strength is nearly equal with the bond strength of C\Zn in DMZn [64]. Accordingly, the observed reduced ZTO ALD growth is likely to originate from reduced surface hydroxyl density and/or less reactive surface hydroxyls. Analysis of the ZTO growth rate behavior (Fig. 6) provides insight into the origin of the growth rate reduction as compared to the growth rates of the binaries. The growth rate drops particularly strongly, about 30%, with a cycle of SnOx followed by five cycles of ZnO within a super cycle (fSn = 0.17), and then at higher fSn converges toward the growth rate of pure SnOx ALD. This could be understood by a disruption of the ZnO ALD process by SnOx cycle(s), causing less deposition of ZnO, a hypothesis that is supported by compositional analysis of the films (vide infra). Using this approach, we calculated growth rates for ZTO films deposited at fSn values of 0.0, 0.17, 0.33, 0.50, 0.67, 0.83, and 1.0 by assuming that ZnO growth rate after the SnOx cycle(s) is first reduced and then linearly restored during the subsequent ZnO cycles. ZnO growth rates reduced by factors of 1/6, 2/6, 3/6, 4/6, and 5/6, and restoration of the ZnO growth rate in 1, 2, 3, 4, and 5 ZnO cycles were considered. Thus, the growth rates (rZTO) were calculated from
r ZTO ¼
Fig. 5. Plan-view, high-resolution SEM images for ZTO morphology deposited by varying the bilayer period from 800 to 4, as indicated. The films were deposited at 150 °C on Si(100) using a Sn/(Sn + Zn) cycle fraction of 0.75.
" # nZnO X 1 i−1 n þ rZnO nZnO −nZnO þ r ZnO R þ ð1−RÞ r ðnZnO þ nSnOx Þ SnOx SnOx n ZnO i¼1
ð1Þ
where r, n, n⁎ZnO, and R refer to the growth rate, the number of ZnO or SnOx ALD cycles in a super cycle, the number of ZnO ALD cycles with reduced growth rates in a super cycle, and a factor by which the ZnO growth rate is reduced, respectively. The best least squares fit with experimental growth rates was obtained for a ZnO growth rate reduced down to R = 3/6 of the pure ZnO growth rate and with restoration of the pure ZnO growth rate within 3 ZnO cycles (see Fig. 6), providing an estimate for the nucleation period for restoring the pure ZnO ALD growth. Note that these parameters that were derived from the data in Fig. 6 were also utilized to produce the results shown in Figs. 2 and 3, where in general correct trends for growth rates and ZTO composition as a function of bilayer period were produced. Notably, a nucleation period of 4–6 ALD cycles, of a similar magnitude to the value of n⁎ZnO determined here, was reported for ZnO in the ZnO/Al2O3 ALD on the basis of in situ QCM experiments [50]. More insight into the mechanistic
190
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
where r, n, ρ, and M refer to the growth rate, the number of ALD cycles in a super cycle, the density, and the molecular mass, respectively. n⁎ZnO refers to the number of ZnO ALD cycles with reduced growth rates in a super cycle. Note that Eq. (2) also generates the linearly interpolated compositions by setting n⁎ZnO = 0. Densities and molecular masses of both SnO and SnO2 were used for the SnOx values, but the results were practically equal and hence only results for SnO2 were used to calculate the compositions shown in Fig. 7. These values for the densities of the constituent binaries are as follows: ρZnO = 5.61 g/cm3, ρSnO = 6.45 g/cm3 and ρSnO2 = 6.95 g/cm3 [66]. Fig. 7 shows that excess tin with respect to the linearly interpolated tin composition fraction (dashed black curve) is observed for all fSn below 0.83. On the other hand, below this fSn the compositions from analytical modeling (dashed red curve) are in good agreement with the experimental values, providing additional evidence for the reduced ZnO growth after SnOx cycle(s) that forms the basis of this model. The analytically determined compositions deviate from the observed composition significantly only at fSn = 0.83, which suggests that mixing small numbers of ZnO cycles into a SnOx ALD process results in deposition of less Sn as compared to pure
SnOx ALD, and future surface chemistry investigations will provide insight into this behavior. Previous reports of as-deposited ZTO grown by ALD show that the films are largely amorphous [25,37–39]. To better understand the ZTO composition–crystallinity relationship, we characterized the structural properties of as-deposited films as a function of fSn. Fig. 8 shows diffraction patterns for the constituent binary oxides and intermediate ZTO compositions performed using grazing incidence X-ray diffraction (GIXRD). The diffraction pattern of the pure ZnO sample is characteristic of the hexagonal wurtzite phase, showing reflection of the (100), (002), (101), (110), and (201) planes at 31.86°, 34.4°, 36.8°, 56.6°, and 69.1° referenced to Cu-Kα, respectively. The peak broadening for these patterns is due to the limited resolution of the two dimensional GIXRD experiment. For fSn = 0.17 (~10% Sn), the film still shows patterns, albeit significantly broader and weaker, for reflection of the (101), (002), (101), and (201) planes of hexagonal wurtzite ZnO. The result suggests that mixing ZnO ALD with a small fraction of SnOx cycles begins to decrease the long-range order of deposited ZnO, and the origin of this finding represents an interesting target for a future detailed study. At higher cycle fractions, the films are X-ray amorphous, as is pure SnOx deposited from TDMASn and H2O [59], showing no reflections for ZnO, SnO, SnO2, Zn2SnO4 or ZnSnO3. In general, the consequences of the mixing of crystalline and amorphous materials on the atomic scale, as is the case in mixing ZnO and SnOx, are poorly-understood [67], and future mechanistic studies are necessary for elucidating the main factors causing the transition from crystalline to amorphous ZTO. For the pure ZnO sample and the ZTO sample with fSn = 0.17 that exhibit a hexagonal wurtzite structure, the growth orientation can be determined from GIXRD. The 2d GIXRD data give information on the relative intensities of the crystal planes that are oriented out of the plane of the substrate (Qz) and in the plane of the substrate (Qxy), as depicted in Fig. 9a). Fig. 9b–c) shows the 2d GIXRD data for pure ZnO and a fSn = 0.17 ZTO sample. For both samples, a strong 002 component is observed in-plane, indicating that the c-axis of the wurtzite crystal is growing parallel to the substrate. This is consistent with the SEM images below, in which elongated crystals are shown growing in the substrate plane. For pure ZnO, the 002 plane is also observed in the out-of-plane direction, indicating that a fraction of the crystals grow with a 002 fiber texture. The absence of the out-of-plane 002 feature for the fSn = 0.17 ZTO sample demonstrates the sensitivity of this growth direction toward incorporation of tin. In sum, the XRD data demonstrates a crystalline-to-amorphous transition when mixing ZnO ALD with SnOx cycles, with the deposited films being amorphous above tin cycle fractions of 0.17 (~10% Sn). The XRD results are further supported by film morphologies observed by SEM, with most of the films resembling amorphous SnOx, as illustrated in Fig. 10. This finding is in agreement with previous work on ZTO ALD [37].
Fig. 7. ZTO compositions determined by ICP-OES and from an analytical model described by Eq. (2) as a function of Sn/(Sn + Zn) cycle fraction, fSn. The black dashed line gives the compositions linearly interpolated from the growth rates and densities of the binaries. Films were deposited on Si(100) at 150 °C using a bilayer period of Λ = 6 for a total of 100 super cycles.
Fig. 8. Grazing incidence X-ray diffraction (GIXRD) patterns of ZTO films deposited using Sn/(Sn + Zn) cycle fractions from 0 to 1. Films were deposited on Si(100) at 150 °C using a bilayer period of Λ = 6 for a total of 100 super cycles. The Sn/(Sn + Zn) cycle fractions are indicated on the right.
details of the ZTO ALD growth mechanism will be provided by future theoretical and experimental investigations. Reduced ZnO growth rates after SnOx ALD cycle(s) should be reflected in the ZTO composition as an excess of Sn with respect to the composition estimated from the growth rates and densities of the binaries, and hence compositional analysis of the films provided the means to verify this. The fractional tin compositions determined experimentally by ICP-OES for the ZTO films deposited using a constant number of ALD cycles in a super cycle are shown in Fig. 7 as a function of fSn. Also shown in Fig. 7 for comparison are two predicted composition curves. The first (referred to as “analytical” in Fig. 7) is based on the analytical model introduced earlier, in which a reduction in ZnO growth rate on SnOx is considered. These compositions are calculated using Eq. (2), which is derived from Eq. (1) by taking into account the densities of the binaries. A simpler predicted composition curve calculated solely from the linearly interpolated growth rates (see Fig. 6) and densities of the binaries is also included for comparison (referred to as “linearly interpolated”). The analytical model compositions were calculated using equation
nSnOx
cSn ¼ nSnOx
ρSnOx r SnOx MSnOx
ρSnOx r SnOx
MSnOx " # nZnO X ρZnO r ZnO 1 1 i−1 þ þ nZnO −nZnO þ 2 2 nZnO M ZnO i¼1 ð2Þ
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
191
Fig. 9. a) Cartoon depicting the geometry of the in-plane (Qxy) and out-of-plane (Qz) components of the reciprocal lattice vector. b) 2d GIXRD pattern for pure ZnO exhibiting 002 growth both in-plane and out-of-plane. c) 2d GIXRD pattern for ZTO film with fSn = 0.17.
Fig. 10. High resolution SEM images of ZTO films deposited on Si(100) at 150 °C using Sn/(Sn + Zn) cycle fractions from 0 to 1. The cycle fractions, together with ICP-OES determined tin compositions (in parentheses) are shown in the images. The scale bar is 500 nm.
192
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
Spectroscopic ellipsometry was used to characterize the optical properties of the ZTO films grown on Si(100) as a function of fSn, and thus composition. The results are shown in Fig. 11. We believe that this is the first comprehensive report of optical constants over a full range of compositions for ZTO by thermal ALD. An oscillator model was used to represent the optical properties of the ZTO films. This included a polynomial spline function (the psemi-m0 oscillator [68]), a Lorentz oscillator and poles on each side of the spectrum to represent the optical constants of the films in a Kramers–Kronig consistent manner. A multi-sample analysis was used to fit a common set of oscillator parameters to samples sharing the same Sn cycle fraction. However, film thicknesses were allowed to vary individually for each sample in the analysis. As a control for the ellipsometry results, UV–vis extinction measurements were performed at normal incidence for films deposited on fused silica quartz in the same ALD runs, as shown in Fig. 11c). The general trends and features including absorption onset show good agreement with the extinction coefficients obtained from VASE. For pure ZnO, the extinction coefficient displays a characteristic, sharp and direct type bandgap (Fig. 11b). The introduction of Sn into ZnO appears to increase the absorption over the entire visible range. This could, for instance, be due to defects associated with a more disordered phase, and/or free carrier-like absorption as the Sn act as an n-type dopant in the ZnO [69]. For intermediate ZTO compositions, the refractive index shown in Fig. 11a) drops with increasing Sn content. When completely eliminating the ZnO content, however, a rather
dramatic increase of the refractive index is observed for the pure SnOx-phase. The higher reflectance expected from this change is also consistent with the increased extinction (Fig. 11c). To analyze the bandgap dependence on composition, extrapolations of linear segments in the absorption onset region were performed as shown in Fig. 12. As it is not clear what bandgap type is most appropriate for the intermediate compositions, fits to both direct and indirect/ amorphous type bandgaps are presented in Fig. 12a) and b), respectively. The resulting bandgap values are presented as a function of cycle fraction and composition in Fig. 12c) and d), respectively. The direct bandgap of ZnO was found to be 3.3 eV, in line with previous literature. The Tauc plot analysis for the amorphous type bandgap suggests a nontrivial dependence on cycle fraction and composition (Fig. 12c and d), where the introduction of a low Sn concentration (10%) results in a bandgap minimum of 2.8 eV. The bandgap then widens to a maximum of ~ 3.4 eV for a ZTO film with 50% Sn, and then again a narrows to ~2.9 eV for the pure SnOx-phase. Interestingly, Kapilashrami et al. observed a similar dependence for amorphous ZTO ALD films by soft Xray spectroscopy [39]. The increased refractive index observed for the pure SnOx film is concomitant with a reduction of both type of bandgaps, but it is not clear what the primary cause for these changes is. It was recently suggested that the optical properties of SnOx depend critically on the microstructure of the film, which may entail grain sizes below the detection limit of XRD [59]. If this is causing the strong dependence of the optical properties on composition here, it implies that infrequent interruptions of the SnOx growth by single ALD cycles of ZnO (fSn = 0.83) has a significant effect on the resulting microstructure. Overall, the non-trivial dependence of optical properties on ZTO composition is likely affected by the interplay between crystal structure, dopant concentrations, and microstructural changes of the material. 4. Conclusions We report the structural and optical properties of zinc tin oxide (ZTO) deposited by thermal ALD by systematically varying the relative fractions of SnOx and ZnO ALD cycles and the bilayer period in ZTO ALD. Non-trivial dependences on these parameters were observed for the ALD growth rates, structures, and properties. Namely, growth rates decreased with increasing amounts of tin in ZTO, and were lower than what would be expected from a linear interpolation of the growth rates of the binary oxides. An analytical model accounting for the decreased growth was developed to provide insight into the ZTO ALD growth rate process. X-ray diffraction showed that the crystallinity of the pure ZnO is largely lost when ZTO is deposited using a tin cycle fraction higher than 0.17 (~10% Sn). The optical properties, especially the amorphous type bandgap size, shows a non-trivial dependence on ZTO composition which may be understood as an interplay between dopant concentrations and microstructural changes of the material. The mild ZTO ALD process reported herein may be easily adapted for use in thin film solar cells and in thin film transistors. Because the ALD processes for these binary metal oxides can be performed at nearroom temperature [59], this ZTO ALD process may also be used in systems where low-temperature processing is required, such as for biological and organic electronic applications. Acknowledgments
Fig. 11. Spectroscopic ellipsometry results for a) refractive index and b) extinction coefficient vs. wavelength for ZTO films deposited on Si(100) at 150 °C using Sn/(Sn + Zn) cycle fractions from 0 to 1. For comparison, optical extinction vs. wavelength from UV–vis transmission measurements of the ZTO on fused silica is shown in c). The cycle fractions are indicated in the legends.
This work was supported by the Center for Nanostructuring for Efficient Energy Conversion, an Energy Frontier Research Center (EFRC) funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences (BES) under Award No. DE-SC0001060. The X-ray diffraction studies were supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award #DE-SC0004782. MNM acknowledges the Bell Labs Graduate Fellowship Research Program for graduate fellowship support. CH is grateful for financial support from the Marcus and Amalia Wallenberg Foundation. JTT gratefully
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
193
Fig. 12. VASE multi-sample analyses of bandgaps for ZTO films deposited on Si(100) at 150 °C using Sn/(Sn + Zn) cycle fractions from 0 to 1. The bandgaps are found by extrapolating linear portions of the function (αhv)x to zero. In this expression, hv is the photon energy and α the absorption coefficient (α = 4πk / λ with k the extinction coefficient of the material, as determined by VASE). For the direct type bandgap fits in a), the exponent x = 2, while for the indirect or amorphous type bandgap in b), x = 0.5. The results are summarized versus c) the Sn cycle fraction and d) composition. The cycle fractions are indicated in the legends.
acknowledges the Academy of Finland (Grant 256800/2012) and the Finnish Cultural Foundation for financial support. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]
T. Minami, Semicond. Sci. Technol. 20 (2005) S35. I. Hamberg, C.G. Granqvist, J. Appl. Phys. 60 (1986) R123. R.G. Gordon, MRS Bull. 25 (2000) 52. C.A. Hoel, T.O. Mason, J. Gaillard, K.R. Poeppelmeier, Chem. Mater. 22 (2010) 3569. E. Fortunato, D. Ginley, H. Hosono, D.C. Paine, MRS Bull. 32 (2007) 242. T. Minami, H. Nanto, S. Shooji, S. Takata, Thin Solid Films 111 (1984) 167. G. Frank, H. Kostlin, Appl. Phys. A: Mater. Sci. Process. 27 (1982) 197. P. Nath, R.F. Bunshah, B. Basol, O.M. Staffsud, Thin Solid Films 72 (1980) 463. I.A. Rauf, Mater. Lett. 18 (1993) 123. A.C. Tolcin, U.S. Geol. Surv. (2012) 74. T. Minami, Thin Solid Films 516 (2008) 5822. J. Montero, C. Guillen, J. Herrero, Thin Solid Films 519 (2011) 7564. J. Perrenoud, L. Kranz, S. Buecheler, F. Pianezzi, A.N. Tiwari, Thin Solid Films 519 (2011) 7444. H. Liu, V. Avrutin, N. Izyumskaya, U. Ozgur, H. Morkoc, Superlattice. Microst. 48 (2010) 458. T. Minami, S. Takata, H. Sato, H. Sonohara, J. Vac. Sci. Technol. A 13 (1995) 1095. Y. Hayashi, K. Kondo, K. Murai, A. Moriga, I. Nakabayashi, H. Fukumoto, K. Tominaga, Vacuum 74 (2004) 607. T. Moriga, Y. Hayashi, K. Kondo, Y. Nishimura, K. Murai, I. Nakabayashi, H. Fukumoto, K. Tominaga, J. Vac. Sci. Technol. A 22 (2004) 1705. J. Perkins, J. del Cueto, J. Alleman, C. Warmsingh, B. Keyes, L. Gedvilas, P. Parilla, B. To, D. Readey, D. Ginley, Thin Solid Films 411 (2002) 152. D. Young, D. Williamson, T. Coutts, J. Appl. Phys. 91 (2002) 1464. H. Enoki, T. Nakayama, J. Echigoya, Phys. Status Solidi A 129 (1992) 181. E. Cetinorgu, S. Goldsmith, Z. Barkay, R.L. Boxman, J. Phys. D Appl. Phys. 39 (2006) 5245. S. Vijayalakshmi, S. Venkataraj, M. Subramanian, R. Jayavel, J. Phys. D Appl. Phys. 41 (2008) 035505. A. Kurz, M.A. Aegerter, J. Sol-Gel Sci. Technol. 31 (2004) 267. Z. Hu, S. Gao, Chem. Geol. 253 (2008) 205. A. Hultqvist, C. Platzer-Bjorkman, U. Zimmermann, M. Edoff, T. Torndahl, Prog. Photovoltaics 20 (2012) 883. H.Q. Chiang, J.F. Wager, R.L. Hoffman, J. Jeong, D.A. Keszler, Appl. Phys. Lett. 86 (2005) 013503. Y.-. Chang, D.-. Lee, G.S. Herman, C.-. Chang, Electrochem. Solid-State 10 (2007) H135. S.K. Park, Y. Kim, H. Kim, J. Han, Electrochem. Solid-State 12 (2009) H256. G.Z. Zang, J.F. Wang, H.C. Chen, W.B. Su, W.X. Wang, P. Qi, C.M. Wang, J. Mater. Sci. 39 (2004) 3537.
[30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59]
P.R. Bueno, J.A. Varela, E. Longo, J. Eur. Ceram. Soc. 28 (2008) 505. S. Yu-Sheng, Z. Tian-Shu, Sensors Actuators B Chem. 12 (1993) 5. D. Kovacheva, K. Petrov, Solid State Ionics 109 (1998) 327. S. Baruah, J. Dutta, Sci. Technol. Adv. Mater. 12 (2011) 013004. W.M. Kim, J.H. Ko, I.H. Kim, D. Kim, K.S. Lee, T.S. Lee, B. Cheong, Appl. Surf. Sci. 253 (2007) 7398. Y. Jeong, K. Song, D. Kim, C.Y. Koo, J. Moon, J. Electrochem. Soc. 156 (2009) H808. R.G. Gordon, Final Technical Report, Harvard University, National Renewable Energy Laboratory, 2001. W.S. Choi, J. Korean Phys. Soc. 57 (2010) 1472. A. Hultqvist, M. Edoff, T. Törndahl, Prog. Photovoltaics Res. Appl. 19 (2011) 478. M. Kapilashrami, C.X. Kronawitter, T. Törndahl, J. Lindahl, A. Hultqvist, W. Wang, C. Chang, S.S. Mao, J. Guo, PCCP 14 (2012) 10154. J. Heo, S.B. Kim, R.G. Gordon, Appl. Phys. Lett. 101 (2012) 113507. M.K. Jayaraj, K.J. Saji, K. Nomura, T. Kamiya, H. Hosono, J. Vac. Sci. Technol. B 26 (2008) 495. D.L. Young, H. Moutinho, Y. Yan, T.J. Coutts, J. Appl. Phys. 92 (2002) 310. X. Fu, X. Wang, J. Long, Z. Ding, T. Yan, G. Zhang, Z. Zhang, H. Lin, X. Fu, J. Solid State Chem. 182 (2009) 517. J. Zeng, M. Xin, K. Li, H. Wang, H. Yan, W. Zhang, J. Phys. Chem. C 112 (2008) 4159. M. Miyauchi, Z. Liu, Z. Zhao, S. Anandan, K. Hara, Chem. Commun. 46 (2010) 1529. S.M. George, Chem. Rev. 110 (2010) 111. M. Leskela, M. Ritala, Thin Solid Films 409 (2002) 138. M. Leskela, M. Ritala, Angew. Chem. Int. Ed. 42 (2003) 5548. H. Yuan, B. Luo, D. Yu, A. Cheng, S.A. Campbell, W.L. Gladfelter, J. Vac. Sci. Technol. A 30 (2012) 01A138. J.W. Elam, S.M. George, Chem. Mater. 15 (2003) 1020. T. Dhakal, D. Vanhart, R. Christian, A. Nandur, A. Sharma, C.R. Westgate, J. Vac. Sci. Technol. A 30 (2012) 021202. C. Gong, Y. Choi, H. Kim, C. Park, H. Park, J.G. Jang, H.J. Chang, G.Y. Yeom, J. Vac. Sci. Technol. A 31 (2013) 01A101. J.Y. Kim, Y. Choi, H. Park, S. Golledge, D.C. Johnson, J. Vac. Sci. Technol. A 28 (2010) 1111. P.R. Chalker, P.A. Marshall, S. Romani, M.J. Rosseinsky, S. Rushworth, P.A. Williams, J. Buckett, N. McSporran, J. Ridealgh, MRS Proc. 1315 (2011) 39. K. Saito, Y. Hiratsuka, A. Omata, H. Makino, S. Kishimoto, T. Yamamoto, N. Horiuchi, H. Hirayama, Superlattice. Microst. 42 (2007) 172. J.W. Elam, D.A. Baker, A.B.F. Martinson, M.J. Pellin, J.T. Hupp, J. Phys. Chem. C 112 (2008) 1938. C. Platzer-Bjorkman, T. Torndahl, A. Hultqvist, J. Kessler, M. Edoff, Thin Solid Films 515 (2007) 6024. J. Heo, Y. Liu, P. Sinsermsuksakul, Z. Li, L. Sun, W. Noh, R.G. Gordon, J. Phys. Chem. C 115 (2011) 10277. M.N. Mullings, C. Hägglund, S.F. Bent, J. Vac. Sci. Technol. A 31 (2013) 061503.
194
M.N. Mullings et al. / Thin Solid Films 556 (2014) 186–194
[60] J.R. Bakke, J.T. Tanskanen, C. Hägglund, T.A. Pakkanen, S.F. Bent, J. Vac. Sci. Technol. A 30 (2012) 01A135. [61] J.T. Tanskanen, J.R. Bakke, T.A. Pakkanen, S.F. Bent, J. Vac. Sci. Technol. A 29 (2011) 031507. [62] L. Wachnicki, M. Lukasiewicz, B. Witkowski, T. Krajewski, G. Luka, K. Kopalko, R. Minikayev, E. Przezdziecka, J.Z. Domagala, M. Godlewski, E. Guziewicz, Phys. Status Solidi B 247 (2010) 1699. [63] T. Törndahl, C. Platzer-Björkman, J. Kessler, M. Edoff, Prog. Photovoltaics Res. Appl. 15 (2007) 225.
[64] Y. Luo, Comprehensive Handbook of Chemical Bond Energies, CRC Press, University of Science and Technology of China, Hefei, 2007. [65] R.L. Jackson, Chem. Phys. Lett. 163 (1989) 315. [66] In: W.M. Haynes (Ed.), CRC Handbook of Chemistry and Physics, 94th edition, CRC Press, 2013. [67] D. Wang, A. Fernandez-Martinez, Science 337 (2012) 812. [68] Guide to using WVASE32, J. A. Woollam Co., Inc, Lincoln, NE., 2008 [69] S. Venkataraj, S. Hishita, Y. Adachi, I. Sakaguchi, K. Matsumoto, N. Saito, H. Haneda, N. Ohashi, J. Electrochem. Soc. 156 (2009) H424.