Influence of the aqueous solution composition on the morphology of Zn1xMgxO films deposited by spray pyrolysis

Tailoring the band gap of metal oxides is an attractive property for various device applications. Zinc–magnesium-oxide (Zn1 xMgxO) is a sustainable material that allows control of the energy band gap through the Mg content. To reduce the deposition costs of Zn1 xMgxO, solution-based techniques can be employed. This study investigates spray-pyrolysis as a versatile, environment-friendly, solution-based deposition method to fabricate Zn1 xMgxO films with morphology and properties that strongly depend on the composition of the water-based precursor solution. Films that are either compact or covered with nanorods are obtained, with tunable amounts of incorporated Mg and corresponding band gap values. In order to show the applicability of these films, Zn1 xMgxO was sprayed onto transparent conducting oxide substrates that are commonly used in various devices, such as solar cells with superstrate architecture.


Introduction
Oxide-based wide-band gap materials are attractive for an extensive range of applications, such as photovoltaics, 1-3 functional coatings, 4,5 (opto)electronic devices [6][7][8] or sensors. 9,10 Among them, ZnO is one of the most investigated metal oxides, because of its facile fabrication and doping, especially by low-cost solution-based techniques. [11][12][13] ZnO can be doped with group-13 elements to enhance the intrinsic n-type conductivity or with divalent elements such as Mg or Cd, which increase or decrease the band gap due to differences in the electron affinity. 14,15 The tunability of the band gap with Mg incorporation renders Zn 1Àx Mg x O especially interesting for semiconductor devices, as the energy band alignment at the heterojunction is of utmost importance. An example is given by Hariskos et al. who reported the application of Zn 1Àx Mg x O as n-type buffer layer in Cu(In,Ga)(Se,S) 2 -based thin film solar cells 16 or Minami et al. who applied a Zn 1Àx Mg x O buffer layer in Cu 2 O-type solar cells. 17 Although Zn 1Àx Mg x O can be deposited by a variety of vacuumbased deposition techniques, such as pulsed laser deposition (PLD), 18 sputtering, 19 molecular beam epitaxy (MBE) 20 or atomic layer deposition (ALD), 21 solution-based methods are particularly interesting regarding deposition costs. Solution-based Zn 1Àx Mg x O deposition includes chemical bath deposition (CBD), 22 electrochemical deposition (ECD), 23 and sol-gel techniques. 24 The first two take place close to ambient temperature, but they have the disadvantage that the thermodynamic solubility limit of MgO in ZnO is B4%. 25 Sol-gel techniques can achieve higher Mg incorporation up to 23%, 24 but require a high temperature (4500 1C) annealing step for sufficient film quality.
Spray pyrolysis is a versatile solution-based growth technique, which yields high-quality ZnO films at temperatures below 400 1C. [26][27][28] Further, this method is easy to upscale and, therefore, suitable for large area deposition. In spray pyrolysis a precursor solution, containing volatile metal salts, forms a mist of droplets that are directed onto a heated substrate. For the droplet generation many techniques are available. Among them, the ultrasonic atomizing system has been proven to be precise, controllable and repeatable. 29 Zn 1Àx Mg x O films deposited by spray pyrolysis have been reported frequently, 15,[30][31][32][33][34][35][36][37][38][39][40][41] but in the majority of the reports toxic or flammable solvents, such as methanol [30][31][32] or ethanol, 15,34,35 are used. Studies on water-based solutions are rare. 36,37,40,41 Moreover, due to the unfavorable properties of water as a solvent in spray pyrolysis (such as strong hydrogen bonds and a high boiling point), 27 temperatures above 4400 1C are commonly employed. 36,37 Among the studies using water-based solutions and temperatures below 400 1C, none achieved Mg incorporation larger than 10 mol%. 40,41 For certain device applications, the use of nanostructures instead of compact films can be an advantage. [42][43][44] For the same reasons as for compact films, it is highly desired to deposit nanostructures with a tailored band gap. A common approach to obtain Mg-doped ZnO nanorods from aqueous solutions is hydrothermal deposition, but so far the amount of incorporated Mg by this technique was limited to below 1%. 45,46 Up to now, there is no report on Zn 1Àx Mg x O nanorod deposition by a onestep spray pyrolysis process.
In this work Zn 1Àx Mg x O was deposited by ultrasonic spray pyrolysis (USP) from water-based solutions at 360 1C, in air. Solely low-cost and environment-friendly raw materials were chosen for the precursor solution. This study elucidates the effect of the solution composition on the deposited Zn 1Àx Mg x O films and presents a novel, one-step synthesis approach to obtain Zn 1Àx Mg x O films that are either compact or covered with nanorods on various substrates. The Mg incorporation in the Zn 1Àx Mg x O films was investigated and optical band gaps were determined. In order to show the applicability of these films in applications, Zn 1Àx Mg x O was sprayed onto three different transparent conducting oxide substrates which are commonly used as electrodes in devices.

Spray pyrolysis set-up
USP was carried out in a Sono-Tek Exacta Coat s system, using a Sono-Tek Impact s 120 kHz nozzle in the horizontal geometry with an applied power of 3.5 W. The solution flow rate was kept at 0.8 ml min À1 and compressed air (0.5 bar) was used as the carrier gas. The substrate-to-nozzle distance was kept at 200 mm and the scanning speed of the nozzle over the substrate was 25 mm s À1 . The number of spraying cycles over the substrate was for all experiments fixed to 80 repetitions, unless otherwise stated, resulting in a film thickness between 200 and 300 nm. The substrates were heated to 360 1C using a hot plate covered with an aluminium substrate holder.

Sample preparation and precursor solutions
Borosilicate glass substrates (Schott Nexterion s D, 7.5 Â 2.5 cm 2 ) were ultrasonically cleaned at 50 1C for 30 minutes in Hellmanex s III washing solution, rinsed with deionized water (DI, 18 MO cm À1 ), then rinsed with isopropanol and finally dried in an air stream. Commercial fluorine-doped tin oxide (FTO, 2.2 mm thick, 7 O Sq À1 , Sigma-Aldrich 735140-5EA) and indium tin oxide glass substrates (ITO, 1 mm thick, Sigma-Aldrich 703192, 8-12 O Sq À1 ) were rinsed with deionized water, rinsed with isopropanol and dried in an air stream. For the preparation of 1.8 mm-thick indium-doped ZnO (IZO) films on borosilicate glass substrates the same recipe as previously reported by our group 47 was used.
In the first set of experiments Zn 1Àx Mg x O films were deposited on clean glass substrates using precursor solutions with various concentrations of magnesium dopant and a constant concentration of zinc salt. Metal acetates were chosen as precursor salts due to their low-cost, non-toxicity and suitable properties for spray deposition. The sprayed aqueous solutions A-F contained 0.2 M of zinc acetate dihydrate (Zn(CH 3 COO) 2 Â 2H 2 O, Sigma-Aldrich 96459) and 0, 5, 10, 15, 20 and 30 mol% magnesium acetate tetrahydrate (Mg(CH 3 COO) 2 Â 4H 2 O, Sigma-Aldrich M5661) dissolved in deionized water, as summarized in Table 1. Additionally, the solutions contained 1.4 M acetic acid (CH 3 COOH, Sigma-Aldrich A6283) to achieve a lower surface tension, resulting in a smaller droplet size, as well as a pH value suitable for the deposition.
To investigate the influence of the zinc acetate molarity and the solution pH value, solutions G-L were prepared with a constant Mg concentration (15 mol% Mg) and varied zinc acetate molarities or acetic acid concentrations as displayed in Table 1.
For the deposition of nanorods and compact films on different TCO substrates, solutions D and I with 15 mol% Mg and different zinc acetate molarities were sprayed onto IZO, FTO and ITO.

Film characterization
The film morphology and film thickness were analysed by scanning electron microscopy (SEM, Zeiss Ultra 40), using an in-lens detector and a 5 kV acceleration voltage. The transparency of the films was obtained by Fourier transform infrared spectroscopy (FTIR, Bruker Vertex 70) in the UV/vis/IR range between 300 and 1000 nm. X-ray diffraction (XRD, ThermoFisher Scientific ARL Equinox 100) with Cu-K a (l = 1.5419 Å) radiation was used to determine the crystal structure of the deposited films. The elemental composition was determined by inductive coupled plasma spectroscopy (ICP-OES, PerkinElmer Optima 5300 DV). For the ICP-OES measurements liquid samples were prepared by dissolving the Zn 1Àx Mg x O films in 1 M ultrapure HNO 3 in an ultrasonic bath for 20 minutes at room temperature. The sheet resistance was measured using a 4-point probe set-up connected to a semiconductor parameter analyzer (Süss MicroTec probes, Agilent 4156 C parameter analyzer). EDX analysis was done on a scanning electron microscope (Zeiss Supra 55 VP) with an accelerating voltage of 3 kV.    Table 2 and shown in Fig. 1. Independent of the Mg concentration in the solution, X-ray diffractograms demonstrate that all Zn 1Àx Mg x O films exhibit a    Our results suggest a solubility limit of B20 mol% Mg in the solution (resulting in B16 mol% Mg in the film) for the experimental conditions of this study. Samples deposited from 30 mol% Mg solutions showed an especially high amount of Mg (23.6 mol%) in the ICP OES, which is attributed to MgO coprecipitation. A variation in spraying cycles for solution D retains the preferred {0001} ZnO crystal orientation ( Fig. 3(b)) even for the thinnest film, suggesting that there is no competitive growth in the beginning of the film formation.

Results and discussion
One important aspect of this work was the incorporation of Mg into the ZnO lattice. To verify that Mg is indeed substituting Zn atoms in the lattice and not solely occupying interstitial sites, XRD peaks were studied. Since the ionic radius of the tetrahedral Mg is slightly smaller (0.57 Å) 48 compared to Zn (0.60 Å), 48 a substitution by this ion results in a minor peak shift to higher 2y angles. Fig. 3(c) and Table 2    The formation of nanorods by spray pyrolysis is not a common phenomenon. Nanorod growth is either observed if a seed layer is deposited prior to the spray deposition 49 or at higher temperatures around 500-600 1C. 50,51 The growth evolution of nanorods by addition of Mg can be explained by considering the aspects that influence the film growth in spray pyrolysis. According to Arca et al., 27 the main factors are: (i) the thermal properties of the solvent, (ii) the volatility/decomposition properties of the precursors and (iii) the substrate temperature.
The solvent evaporation is mainly defined by the water : acetic acid ratio, which was fixed for this experimental series. Therefore, this influence can be neglected.
As far as the thermal decomposition of the metal salt precursors is concerned, it is often studied by thermal analysis methods, such as thermogravimetric (TGA) or differential thermal analysis (DTA). These methods usually apply much lower heating rates compared to those that the precursors undergo in the spray pyrolysis process. Although the heating rate has a significant influence on the thermal decomposition, some conclusions could be drawn. Reports on TGA/DTA analysis of ZnAc 2 and MgAc 2 show that ZnAc 2 decomposes completely at a much lower temperature (B300 1C) 52 compared to MgAc 2 (B400 1C), 53 at an applied heating rate of 20 1C min À1 . Further, for the ZnAc 2 decomposition it was proposed that a volatile intermediate tetrahedral cluster of ZnO 4 (CH 3 COO) 6 (I) (basic zinc acetate complex BZA) is formed, which can adsorb on the substrate surface and stepwise decompose to ZnO. 54,55 Such a complex was never reported for MgAc 2 decomposition and since Mg 2+ prefers the octahedral over the tetrahedral complex coordination, 56,57 it may not be favorable. However, there is the possibility that some zinc atoms in the BZA are substituted by magnesium atoms. It is not clear to what extent the decomposition of ZnAc 2 is influenced by MgAc 2 addition.
The main influence on the growth behaviour most likely arises from the substrate temperature. A possible explanation for the nanorod formation is the modified temperature at the substrate/air interface due to the Zn 1Àx Mg x O deposition. Mg doping lowers the thermal conductivity of ZnO, 58,59 resulting in a temperature gradient that favors nanorod growth (see also the schematic growth behaviour in Fig. 5(a)-(c)). At the same time, the underlying compact film acts as a seed layer for the growth of nanorods.
To further investigate this behaviour, films were deposited with an increasing number of spraying cycles, namely 20, 80 and 160 cycles, from solution D (15 mol% Mg). Fig. 1(d) and 6(a) and (b) show plain view SEM images of these films and Fig. 2(d) and 7(a) and (b) cross sections. From examination of these images it is concluded that a low number of spraying cycles (fast deposition process, less substrate cooling) results in thin and compact films, while for an increasing number of spraying cycles the nanorod length remains constant, while the thickness of the compact part of the film underneath the nanorods increases. This means that nanorods evolve into a compact film as the spraying proceeds.
To verify that the nanorods indeed consist of Zn 1Àx Mg x O, EDX analysis of the compact part of the film and the part with the nanorods was realized. For this, separate scans were acquired on the tilted cross section of a film deposited from solution D (with 15 mol% Mg) (see also Fig. S1, ESI †). In both scans, corresponding to the compact and nanorod part of the film, EDX showed signals from Zn and Mg (Fig. S2, ESI †).
3.1.2. Effect of zinc acetate molarity. The zinc acetate (ZnAc 2 ) molarity was adjusted, while maintaining 15 mol% Mg acetate in the solution. A solution containing 15 mol% Mg and 0.2 M zinc acetate resulted previously in the formation of nanorods (see also Fig. 1(d) and 2(d)). It was observed that with increasing ZnAc 2 molarity denser films can be obtained, while a lower overall metal salt concentration results in thinner compact films with enlarged nanorods as displayed in Fig. 8(a)-(c) and 9(a)-(c). Independent from the Zn 1Àx Mg x O structure, the amount of Mg determined by ICP OES remained unchanged as listed in Table 3 and Fig. 8(a)-(c). This is another indication that the nanorods and the compact film have similar compositions. Fig. 5(d) shows schematically that a high ZnAc 2 molarity results in a large amount of incoming volatile zinc and magnesium species, which hinder the formation of Zn 1Àx Mg x O nanorods and   pack closely forming a compact film. Besides that, as the deposition process is accelerated, there is less surface cooling. At the same time, a higher ZnAc 2 molarity increases the pH value from 3.6 to 4.1, resulting in an altered solution speciation of volatile Zn and Mg complexes, which might also affect the film growth. Upon comparison of the compact ZnO film in Fig. 2(a) with the compact Zn 1Àx Mg x O film in Fig. 9(c), one can observe that they possess different surface morphologies (see also plain view SEM images in Fig. 1(a) and 8(c)). Additionally, the XRD diffractograms shown in Fig. 3(a) Fig. 10(b) and Table 3, is not altered by a variation in the ZnAc 2 molarity. This confirms, together with ICP OES, that Mg incorporation is only dependent on the Mg content in the solution.
In summary, adjustment of the ZnAc 2 molarity offers a simple tool to control the film morphology and deposit films with or without nanorods, with various amounts of incorporated Mg.

Effect of acetic acid concentration.
The amount of acetic acid (HAc) was varied to investigate the influence of the solution pH value and of the solvent evaporation behaviour on the Zn 1Àx Mg x O morphology. The resulting film morphologies are displayed in Fig. 11(a)-(c). With increasing pH value (lower acetic acid concentration) fewer nanorods are observed. The H 2 O : HAc ratio not only has an influence on the temperature during the deposition, it also alters slightly the mean droplet size due to a change in the surface tension. 60 Further, the amounts of volatile Zn and Mg species in the solution depend on the pH value and can also influence the film growth. Compared to the influence of the Mg concentration or the zinc acetate molarity on the Zn 1Àx Mg x O morphology, the influence of the acetic acid concentration is less pronounced.
XRD confirmed, as shown in Fig. 12(a), that all films are polycrystalline, with a preferred {0001} ZnO orientation. Films deposited from a solution containing a high acetic acid concentration have a reduced thickness due to the lower heat required for solvent vaporization and the consequential higher effective deposition temperature. Further, a minor {0001} peak shift to higher 2y angles (see also Fig. 12(b)) can be observed.

Optical and electrical Zn 1Àx Mg x O properties
FTIR spectra (in the UV/vis/IR wavelength region) of the films obtained from all solution compositions are displayed in Fig. 13(a)-(c). For varying Mg concentration (Fig. 13(a)) higher transparencies are observed till 12.1 mol% Mg incorporation, compared to undoped ZnO. Films deposited from solutions E and F with high Mg concentrations exhibit a lower transparency, due to the presence of white MgO precipitates on the film surface. For variations in the ZnAc 2 molarity (Fig. 13(b)) and   the HAc concentration (Fig. 13(c)) the resulting transparency is constant, with values 480%.
Optical band gaps were linearly extrapolated as shown in Fig. 13(d)-(f) from transmittance data using Tauc's relation for   direct band gap semiconductors, 61 where a is the absorption coefficient, hn the photon energy, A a constant and E g is the band gap (in eV): Fig. 13(d) shows a gradual band edge shift to higher values with increasing amounts of incorporated Mg. It was seen that there is a linear dependence of the band gap with Mg incorporation between 0 and 16.3 mol%. For the Zn 1Àx Mg x O films realized by variations in the zinc acetate molarity, the optical band gap remained B3.45 eV, as seen in Fig. 13(e). Alterations in the acetic acid concentration did not affect the band gap as well, as displayed in Fig. 13(f). Fig. 13(g) shows the dependence of the band gap on the amount of incorporated Mg for the different films.
The sheet resistance was measured for Zn 1Àx Mg x O films (B250 nm thick) deposited on glass (as shown in Fig. 2(a)-(d)) with different Mg content (Fig. 13(h)). While for undoped ZnO the sheet resistance is 0.52 MO Sq À1 , it increases with an increasing amount of incorporated Mg to B60 MO Sq À1 for films with 12 mol% incorporated Mg. This behaviour is commonly observed for Zn 1Àx Mg x O films. 62

Zn 1Àx Mg x O deposited on TCO substrates
To show the applicability of the deposition process, Zn 1Àx Mg x O films were sprayed onto three conducting oxide substrates, namely commercial FTO, ITO and sprayed IZO. This is an important step towards device applications in which Zn 1Àx Mg x O is applied on a transparent electrode, such as solar cells with a superstrate architecture. It is important to note that the proposed deposition process might not be applicable to devices in which the substrate properties are impaired by the temperature and/or incoming vapor of precursor solution species.
It is well known that the nature of the substrate influences the surface texture of spray deposited films. [63][64][65] This behaviour can be attributed to the difference in surface energy and thermal properties for each substrate. As stated before, the heat conductivity of the substrate and the deposited material influences the morphology of the sprayed Zn 1Àx Mg x O. Therefore, a lower amount of nanorods in the deposited film is expected due to the higher heat conductivity of metal oxide coated glass substrates, compared to pure borosilicate glass used in the previous experiments. To show that nanorods and dense Zn 0.88 Mg 0.12 O films can be deposited, solution D (for nanorod formation) and solution I (for dense films) were used for the deposition on the various TCO substrates.
The film morphologies are displayed in Fig. 14(a)-(f). On all three TCO substrates it was observed that low concentration solutions (solution D) still lead to the formation of nanorods, while if the ZnAc 2 molarity is increased to 0.8 M compact films can be deposited. Nevertheless, deposition on ITO and FTO results in a lower amount of nanorods. This suggests that the main influence on the morphology of the deposited material stems from the solution composition and the underlying substrate has a minor effect on the film structure. The Zn 0.88 Mg 0.12 O film morphology on FTO substrates (Fig. 14(f)) appears to be rougher than on the other two substrates, which arises from the rougher morphology of the underlying FTO substrate.
The film structure was further analysed by XRD diffractograms of Zn 0.88 Mg 0.12 O deposited on ITO and FTO as shown in Fig. 15(a) and (b). Because of the high film thickness of the IZO electrode (1.8 mm) and the same crystal structure of IZO and Zn 0.88 Mg 0.12 O, it is not possible to assign specific peaks to Zn 0.88 Mg 0.12 O, therefore these diffractograms are not shown here.
In summary, results from the deposition on glass substrates can be transferred to other substrates such as glass coated with TCOs, without altering the material properties. This indicates that the proposed deposition process of Zn 1Àx Mg x O can further be applied in various device applications. An example is given by the application of a compact Zn 0.88 Mg 0.12 O film as a buffer layer in all-solution processed Cu 2 O-type all-oxide solar cells. 47

Conclusion
This article investigates the influence of the precursor solution composition on the structure of Zn 1Àx Mg x O obtained by ultrasonic spray pyrolysis from water-based solutions. Depending on the precursor concentrations, either dense films or nanorods can be obtained. XRD in combination with ICP OES showed that Mg is substitutionally incorporated into the ZnO lattice. It was found that solely the precursor solution defines the amount of incorporated Mg, independent from the Zn 1Àx Mg x O morphology, which offers a valuable tool to deposit nanostructures with defined amounts of Mg and band gaps between 3.28 and 3.50 eV. The applicability of the spray deposited Zn 1Àx Mg x O nanostructures was demonstrated by deposition on conducting metal oxide substrates.

Conflicts of interest
There are no conflicts of interest to declare.