Epitaxial K0.5Na0.5NbO3 thin films by aqueous chemical solution deposition

We report on an environmentally friendly and versatile aqueous chemical solution deposition route to epitaxial K0.5Na0.5NbO3 (KNN) thin films. The route is based on the spin coating of an aqueous solution of soluble precursors on SrTiO3 single crystal substrates followed by pyrolysis at 400°C and annealing at 800°C using rapid thermal processing. Strongly textured films with homogeneous thickness were obtained on three different crystallographic orientations of SrTiO3. Epitaxial films were obtained on (111) SrTiO3 substrates, while films consisting of an epitaxial layer close to the substrate followed by an oriented polycrystalline layer were obtained on (100) and (110) SrTiO3 substrates. A K2Nb4O11 secondary phase was observed on the surface of the thin films due to the evaporation of alkali species, while the use of an NaCl/KCl flux reduced the amount of the secondary phase. Ferroelectric behaviour of the films was investigated by PFM, and almost no dependence on the film crystallographic orientation was observed. The permittivity and loss tangent of the films with the NaCl/KCl flux were 870 and 0.04 (100-orientation) and 2250 and 0.025 (110-orientation), respectively, at 1 kHz.


Introduction
Lead-free piezo-and ferroelectric materials have received considerable attention owing to the environmental issues with the state-of-the-art lead zirconate titanate (PZT)-based materials.
& 2019 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. and the concentration was determined by thermogravimetric analysis. The solution was further mixed with 25 mol% excess of KNO 3 and NaNO 3 to compensate for alkali loss during processing. The solution was stirred at 708C for 5 h to obtain approximately 0.2 M Nb-K -Na aqueous precursor solution. In addition, a 1 : 1 molar ratio mixture of NaCl/KCl salts (Sigma-Aldrich, St. Louis, US) (total 25 mol% relative to KNN) was added to selected precursor solutions by dissolving the chlorides in the solution. The NaCl/KCl mixture was proposed to act as a flux during the crystallization of the films. In order to increase the wettability of the precursor solutions on the substrate, 1 wt% of a solution of 15 wt% polyvinyl alcohol (MW 40 -88, Merck, Darmstadt, Germany) in distilled water was added as a surfactant.
The stable precursor solutions were deposited onto (100), (110) and (111) oriented SrTiO 3 /Nb:SrTiO 3 (STO/Nb:STO) substrates (Crystal GmbH, Berlin, Germany) by spin coating at 2500 -3000 r.p.m. for 20 -40 s. The films were dried at 2008C for 2 min using a hotplate followed by pyrolysis at 4008C for 5 min by rapid thermal processing (RTP) with a heating rate of 40 K s 21 (Jipelec Jetfirst 200 mm, Semco Technologies, Montpellier, France). In order to increase the film thickness, the deposition, drying and pyrolysis steps were repeated 15 times. The final films were heat treated at 8008C for 5 min in flowing oxygen by RTP with a heating rate of 40 K s 21 . The films with excess alkali are named KNN-E-xxx and the ones with the addition of the salt flux in addition to excess alkali are named KNN-ES-xxx where xxx describes the orientation of the substrate (100, 110 or 111).
Phase purity of the thin films was determined by X-ray diffraction (XRD) (Siemens D5005, Germany) analysis using CuKa. Regular u 2 2u scans were performed in continuous mode in the 2u range from 20 to 808 with a step size of 0.018 and dwell time of 2 s at each step. Grazing incidence XRD was performed both on Siemens D5005 and Bruker D8 Discover diffractometers by using a grazing incidence angle of 38 in the 2u range from 10 to 808. A scanning step of 0.088 and a counting time of 16 s per step were used.
Cross-section transmission electron microscopy (TEM) samples of selected films were prepared by focused ion-beam milling (FIB) with a dual-beam Helios Nanolab 600 from FEI. The final polishing with the Ga ion-beam was done at 5 kV. The entire sample was coated with 30 nm of Au prior to the FIB preparation to avoid sample charging. Electron beam assisted Pt deposition was further used to coat the region chosen for TEM to avoid ion-beam damage of the KNN film. Secondary electron SEM images of the film cross-section were recorded in the FIB. A Cs probe-and image-corrected cold-FEG JEOL ARM 200F, operated at 200 kV, was used to characterize the sample. The TEM was equipped with a Centurio silicon drift detector (0.98 sr solid angle) for EDX and a Quantum GIF with dual electron energy loss spectroscopy (EELS).
The electronic properties were studied by piezoresponse force microscopy (PFM), a scanning probe microscopy technique, and impedance analysis. The local piezoelectric properties were studied by PFM applying an AC bias between the tip and the sample bottom electrode (Nb:STO), resulting in a local piezoresponse (PR). PR imaging was carried out using a MultiMode series AFM with a Digital Instruments Nanoscope IIIA controller equipped with conductive tips (Pt/Ir-coated silicon) to serve as a top electrode. The PR signal is composed of a phase and an amplitude (PR amplitude and PR phase). Local piezoelectric loops were obtained by subsequently adding a slowly varying DC bias between the tip and the bottom electrode. Before the measurements, the films were carefully cleaned with ethanol to remove traces of the NaCl/KCl flux.
The relative dielectric permittivity (1 r ) and the dissipation factor (tan d) were measured in the 10 2 2 -10 5 Hz frequency range using an Alpha-A impedance analyser (Novocontrol Technology). The measurements were performed at a zero bias and 1 V oscillation level. The top electrode (diameter 0.4 mm) was made by sputtering Au through a Kevlar fibre mask on the top of films.

Results
The deposited thin films were homogeneous with constant thickness, and no cracks or defects were observed by eye. Hence a successful aqueous route to KNN thin films was developed where the amount of organic additives was kept at a minimum.
The u 2 2u XRD patterns of the KNN films deposited using excess alkali on the three different orientations of STO heat treated at 8008C are shown in figure 1a-c. Only (100), (110) and (111) diffraction reflections are observed, demonstrating that the films have conformed the orientation of the substrates. The diffractograms for the films prepared using a salt flux in addition to the excess alkali show similar features (figure 1d-f) to the films prepared without the salt flux. However, grazing incidence diffractograms included in figure 1g,h reveal that the surface of the films also contains a K 2 Nb 4 O 11 secondary phase. The use of the salt flux significantly reduced the amount of K 2 Nb 4 O 11 formed. The diffraction lines from the secondary phase are not observed in the u 2 2u patterns of the films except for some weak reflections for the film on (111) STO synthetized without salt flux (data not shown), which demonstrates that the secondary phase is mainly present at the surface of the films.
The film thickness and microstructure were studied by TEM, which confirmed that the orientation of the films followed the orientation of the substrates as shown by XRD. Cross-section TEM image of the KNN-ES-100 film is provided in figure 2a, revealing a film thickness in the range 310-345 nm. Close

Discussion
The XRD patterns of the KNN-based films on STO substrates (figure 1) demonstrated a strong preferential orientation following the crystallographic orientation of the substrate as only the (h00), (0k0) and (00l) diffraction lines are present respectively on the three different orientations of the substrate (except for some weak lines of (100) and (110) on the (111) substrate). The lower 2u values observed for the diffraction lines of the films, as compared with the substrates, reveal that the films have a larger out-of-plane lattice parameter than the substrate (figure 1a-f). An important feature of thin films deposited on a substrate is the presence of two-dimensional in-plane strain [19]. The origin of the misfit strain includes the difference in lattice parameters and the thermal expansion due to  different thermal expansion coefficients of the film and the substrate [20]. The thermal expansion of KNN is lower than for STO [21,22]. After heat treatment at 8008C, the KNN film will therefore be in compression.
Epitaxial growth is most usually obtained by physical deposition methods [6][7][8], but epitaxial growth was clearly obtained by the simple aqueous chemical solution deposition method developed in this work (figure 4). Similar films to the ones obtained here have also been prepared by sol-gel deposition of KNN films using alkoxides and an organic solvent [23]. Yu et al. [23] proposed that the development of the epitaxial layer was dependent on the pyrolysis temperature and observed a higher probability for epitaxy when the pyrolysis temperature was in the range 400-4508C. This is in accordance with our work where a pyrolysis temperature of 4008C was used. A low pyrolysis temperature will induce heterogeneous nucleation to occur at the surface of the substrates due to the lower activation energy for heterogeneous compared with homogeneous nucleation [24]. Even though it is promising to fabricate epitaxial KNN films by this cheap and simple method, more studies with respect to processing parameters are necessary to elucidate the mechanism for the epitaxial growth. For the films on (100) and (110) oriented STO, which contained an epitaxial layer and a polycrystalline part, both heterogeneous nucleation took place promoting the epitaxy and homogeneous nucleation resulting in the polycrystalline part of the film. In the epitaxial films on (111) STO substrate, a few pores were observed close to the film substrate interface. Yu et al. [23] claimed that the large porosity in their epitaxial films was caused by decomposition of alkali carbonates formed by the combustion of organics, forming CO 2 . The uniform KNN films obtained by the method presented here might be explained by the advantage of using inorganic compared to organic precursors resulting in less effective volatility. During the drying of the films, the use of water might be important to facilitate shrinkage due to the high surface tension of water compared to solvents like 2-methoxyethanol as long as the film thickness is small.
The GIXRD patterns and TEM of the KNN thin films confirmed the presence of the K 2 Nb 4 O 11 secondary phase at the surface of the films. The formation of the niobium-rich phase can be explained by the high volatility of K and Na species during the heat treatment. The amount of secondary phase caused by the loss of alkali species was independent of heat treatment temperature in the range 700 -9008C (data not shown). Hence, direct evaporation of K(g) or Na(g) from KNN according to equation (4.1) (simplified for KNbO 3 ) cannot explain the formation of the secondary phases observed from these films. The vapour pressure of K(g) over KNbO 3 starts to be significant for loss of alkali at rather high temperatures [25] and should not be significant at the lowest heat treatment temperatures used here. Another mechanism for the loss of alkali should therefore be considered and the formation of alkali hydroxide/carbonate phases at a lower temperature according to equations (4.2) and (4.3) (simplified reactions with only K) has previously been proposed [25]. Even if the films in this work were heat treated in pure oxygen, the films were exposed to ambient air and the decomposition of the oxalate and small amount of PVA added as a wetting agent could form CO 2 .  4) and (4.5) is therefore the most plausible reason for the alkali loss of the films and the formation of the niobium-rich K 2 Nb 4 O 11 secondary phase. The presence of the NaCl/KCl flux during the processing reduces the amount of K 2 Nb 4 O 11 significantly (figure 1). According to the NaCl -KCl phase diagram [26], the KCl -NaCl eutectic has a melting point of 6508C. During the heat treatment, we propose that the alkali hydroxides/carbonates formed are dissolved in the salt flux due to the ionic character of the species. In order to explain the effect of the salt on reducing the K 2 Nb 4 O 11 content, we deduce that the ionic liquid reduces the activity of the alkalis, hence reducing the volatilization. The PR phase data reported in figure 5 clearly show that the films are ferroelectric. As no significant effect of the substrate orientation on the piezo properties was observed, we propose that due to the compressive stress in the films, the polarization direction is aligned vertically in all the films giving a similar response. The dielectric properties of selected films are illustrated in figure 6. The measured dielectric constant of the films was shown to be higher and the loss tangent lower than most literature data reported for films prepared by CSD and sol-gel methods [16 -18,27 -29].

Conclusion
Homogeneous KNN thin films with a thickness of a few hundred nm were prepared by aqueous chemical solution deposition followed by heat treatment at 8008C. The high degree of orientation, high local piezoelectric properties and high permittivity were evidenced by XRD, PFM and dielectric spectroscopy, respectively. The amount of secondary phase was reduced by the use of alkali excess and a salt flux during the processing. The high degree of orientation of the films demonstrates the advantages of our environmentally friendly synthesis method developed, which can be applicable for the fabrication of oriented and epitaxial lead-free ferroelectric thin films.
Ethics. Our study does not use humans or human tissue for data collection and does not need a research ethics approval. Data accessibility. The XRD data, thermogravimetry data, piezoresponse phase and amplitude data, permittivity data and transmission electron image are available: http://dx.doi.org/10.5061/dryad.g05350g [30].