Effect of fructose-containing feedstocks on the microstructure of multicomponent coatings deposited by suspension plasma spraying

Abstract This work addressed to investigate the use of fructose as an additive in the water-based suspension feedstock of a Y-TZP/Al2O3/SiC multicomponent coating manufactured by suspension plasma spraying. The effect of fructose on suspension rheology and surface tension and on the microstructure and thermal conductivity of the resulting coatings was assessed. It was observed that addition of fructose slightly affected the rheological behaviour of the suspensions while a strong decrease in the surface tension of water occurred. The fructose addition led to the development of columnar-like structures, probably associated with its effect on surface tension. X-ray diffraction patterns in the final coating displayed that crystallinity of tetragonal zirconia formed when fructose was added whereas silicon carbide crystalline phase was practically preserved. The determination of thermal conductivity showed that the formation of a controlled columnar structure along with inter-columnar porosity can be beneficial for thermal insulation.


Introduction
Thermal barrier coatings (TBCs) include a class of materials designed to protect gas turbine components from high temperatures and aggressive environments. TBCs are composed of refractory ceramic oxides, such as alumina, titania, magnesia, and their mixtures; nevertheless, yttrium-doped zirconia (YSZ) represents the most commonly used material for TBCs owing to its moderately low thermal expansion, low thermal conductivity and high phase stability and corrosion resistance [1].
Advanced TBCs providing gas turbines with higher efficiency and lower polluting emissions constitute one of the greatest challenges for the current and future gas turbine industry [2]. Thus, the development of new coating architectures, novel multicomponent materials or a combination of both approaches are currently under intense research [3][4][5][6].
One of the simplest way to improve the efficiency of the turbine deals with increasing its thermal insulation, which means that TBCs can operate at higher temperatures [7].
Numerous works in the field of thermal barrier coatings of YSZ have demonstrated that developing columnar microstructure throughout the coating can enhance TBC lifetime owing to the decrease of residual stresses and the increase of thermal shock resistance [8][9][10][11][12][13][14].
Otherwise, suspension plasma spraying (SPS) technique has been proven to be very effective in generating columnar microstructures in YSZ coatings leading to high thermal cycling resistance [10][11][12]. Thus, Bernard et al [12] found thermal conductivity of the YSZ coating processed by SPS lower than 1 W m -1 K -1 from room temperature up to 1100 ºC irrespective of the column morphology developed. This result is in agreement with previous works where YSZ porous columnar structures display thermal conductivity values in the range of 0.6-1 W m -1 K -1 compared to typical values of YSZ in EB-PVD coatings (1.5 W m -1 K -1 ) [8,13,14]. 4 With regard to the design of the columnar microstructure by SPS, literature also recognises the impact of feedstock characteristics and the subsequent suspension spraying into the plasma torch on the genesis and development of such microstructure. Thus, Curry et al. [15] and VanEvery et al [16] reported that the decrease in the viscosity and the surface tension of nanoparticle suspension feedstocks reduce the effective size of the droplet radius assisting the growth of columnar coatings in SPS processes. Nevertheless, although the feedstock effect was reported no research has been addressed on the modification of suspension characteristics to propitiate columnar microstructure appearance.
In the last years, several works have been reported where saccharides and their derivatives were found to be effective in reducing the viscosity of aqueous nanoparticle and microparticle suspensions due to the adsorption of the polymer on the particle surface displacing the adsorbed water molecules, thereby decreasing the viscosity of the suspension [17][18][19]. Regarding the effect of fructose or other saccharides on surface tension in ceramic suspension have not reported To the best of our knowledge. On the other hand, recent research has shown the use of different saccharides such as sucrose, glucose or fructose as pore formers in ceramic materials, demonstrating that it is possible to obtain ceramic materials with tailored porous microstructure by adding some saccharides [20][21][22][23].
A fine porous microstructure can improve the thermal insulation properties of TBCs due to the high difficulty of heat transfer provided by well distributed multiscale porosity [14,24]. It should also be emphasized that SPS process can produce TBCs with very fine porous microstructure as well as high deposition rates that enable to produce large amounts of different coatings at low cost. 5 Finally, the use of fructose has been revealed for stabilizing the tetragonal phase of yttriastabilized zirconia (Y-TZP). Heshmatpour et al. [25] studied the synthesis of zirconia nanoparticles by the sol-gel method in presence of glucose and fructose as organic additives. The use of these organic compounds stabilised the tetragonal zirconia reducing the formation rate of monoclinic zirconia and played an important role in the morphology and crystallite size of the synthesised nanoparticles. On the other hand, saccharides display interesting characteristics such as low cost, non-toxicity, water-solubility and ease of storing. In addition, saccharides can be easily removed from samples by burnout resulting in a fine, porous microstructure [26,27].
As a consequence, fructose represents a serious candidate to be studied as pore modifier in SPS suspension feedstocks for YSZ coatings. Thus, the objective of this work was to investigate the use of D-fructose as an additive in the feedstock suspension and its relationship with the microstructure and thermal properties of a Y-TZP/Al2O3/SiC multicomponent coating manufactured by SPS. The use of this multicomponent system is based on previous research by the authors who are exploring advanced TBCs coatings [3,6]. For this purpose, suspension feedstocks containing different concentrations of fructose and solids were prepared. Rheological characterization of feedstock suspensions was carried out, as well as microstructural examination and thermal conductivity determination. To the best of our knowledge, there is no previous, available research in the thermal spray community on the use of some organic-based compound in a suspension feedstock to promote a tailored porous and/or columnar microstructure in the final SPS coating as targeted in this work. 6

Suspension preparation and characterisation
Multicomponent suspensions were prepared to solids contents of 20 and 30 vol.% (57 and 63 wt-%, respectively) in deionised water at several sonication times and with two fructose concentrations in order to address the effect produced by this compound on the final coating microstructure. Thus, these multicomponent suspensions were obtained by adding the different reagents in water: tetragonal zirconia polycrystals doped with 3 mol% Y2O3 (TZ-3YS, Tosoh Co., Japan) and α-alumina (CT3000SG, Almatis, Germany) according to the eutectic mass ratio of 51/34 and α-silicon carbide (UF-15, Hermann C. Starck, Germany), as disperse phase in a concentration of 15 wt%. More details of raw materials are described in previous works [6,28]. Finally, two amounts of D-fructose Al2O3, and 1.5 wt% synthetic polyelectrolyte (PKV, Produkt KV5088, Zschimmer-Schwarz, Germany with polycarboxylic nature) for SiC), which has been studied in a previous work [28]. Rheological behaviour was analysed using a rheometer (Haake RS50; Thermo, Karlsruhe, Germany) which operated in controlled shear rate mode. The rheometer was designed with an element of measurement of double-cone and plate and equipped with a system to avoid the evaporation of solvent. The test cycle consisted in uploading the shear rate from 0 to 1000 s -1 in 5 minutes, maintaining at 1000 s -1 for 1 minute and downloading from 1000 to 0 s -1 in 5 minutes. In addition, the stability of the 7 multicomponent suspensions containing fructose was evaluated by means of turbiscan stability index (TSI), which is a parameter that allows to quickly and easily compare the sedimentation trend of suspensions [7]. The sedimentation tests were carried out by means of a multiple light scattering equipment (Turbiscan TM LAB stability Analyzer, Formulaction, France). The test consists of measuring the sedimentation every 1 hour for three days. Also, the density and surface tension of the suspensions were determined using a pycnometer and tensiometer (KRÜSS K12, KRÜSS GmbH, Germany) respectively.
The tensiometer operated using a Krüss standard plate and a constant room temperature in all samples. In addition, three repeated tests were performed for each sample and the mean values are averaged.

Coatings preparation
Suspensions were sprayed onto metallic substrates by a plasma torch (F4-MB, Oerlikon Metco, Switzerland) and controlled with a robot (IRB 400, ABB, Switzerland). Before spraying, the metallic substrates with dimensions of Ø 25 mm x 10 mm were grit blasted with black corundum at a constant pressure of 4.2 bars and cleaned in ethanol and ultrasonic bath to remove surface residues. Bond coats (Amdry 997, Sulzer-Oerlikon, Germany) were sprayed by atmospheric plasma spraying under standard conditions set out in a previous research [28] in order to enhance the adhesion between layers. The suspensions feeding was carried out using an injection system developed by the Institute for Ceramic Technology (ITC, Castellón, Spain), which consists in a peristaltic pump to inject the suspensions from the vessels to the plasma gun through a pressure nozzle with a hole size of 150 µm diameter. Before injection, the suspensions were filtered to remove agglomerates larger than 75 µm. Suspension plasma spraying conditions were as follows:

Coatings microstructure characterization 2.3.1 Morphological characterization
The microstructure of multicomponent coatings was examined in surface and crosssection with a field-emission scanning electron microscope (Quanta 200FEG, FEI Company, USA). To improve the quality of the analysis, the cross-sectional coatings were prepared by standardized metallographic procedure (cutting, mounting and polishing) using a semiautomatic polishing machine (Tegramin-25, Struers, Denmark).
Statistical analysis of column (cauliflower-like agglomerate) size features in coatings microstructure was performed using an image analysis software (Microimage analysis) from surface SEM images. Each histogram and column diameter were estimated from more than 100 measurements of the cauliflower-like agglomerates on the coating surface.
Therefore, it allows representative and repeatable statistical results to be presented.
However, the accurate determination of the mean column (cauliflower) diameters must be performed with some care, as the results may be affected by the magnification and low contrast observed between the columns.
The evaluation of thickness, porosity and resolidifed particles distribution in the coatings was also performed by analysing (Microimage analysis) more than 15 micrographs of the cross-sectional images of coatings and averaging the results.

Phase composition characterization
Crystalline phases composition of the coatings was evaluated by X-ray diffraction analysis using an advance diffractometer (Bruker Theta-Theta, Germany) which operated under previously mentioned conditions [6]. Rietveld refinement analysis to quantify the 9 different crystalline phases in the coating was carried out with DIFFRACplus TOPAS software supplied by BRUKER [30].

Thermal diffusivity analysis
Thermal conductivity of ceramic layers was calculated by means of a xenon flash lamp equipment (LFA467 HT Hyperflash, Netzsch-Gerätebau GmbH, Selb, Germany) which determines the thermal diffusivity of coatings (10 mm x 10 mm square samples cut from the coatings). The thermal conductivity was measured at 1000 ºC and it was calculated using Proteus analysis software (Netzsch-Gerätebau GmbH, Selb, Germany) from the measured diffusivity of the complete coating system and theoretical data (Cp) and experimental data (dimensions and thickness) from the individual materials that compose each layer [8]. Moreover, the three-layer model was used to correct the emitted pulse by flash lamp and heat loss between the different layers of the coating [31]. In this method, the light beam, which is generated by xenon lamp, heats the lower sample surface and an infrared detector measures the temperature increase on the upper sample surface.
Moreover, to improve the signal of the radiation emitted by the coating, a thin layer of graphite was deposited on both sides of the coating. Argon gas is used to prevent oxidation of the sample at high temperatures. The equation for correlating the thermal diffusivity to the thermal conductivity is = · · where k is the thermal conductivity (W/(m·K)), Cp is the specific heat capacity (J·kg −1 ·K −1 ) at a certain temperature and constant pressure, is the density of the coatings (kg/m³) and α is the thermal diffusivity (m²/s). The density of the coatings was obtained from the following equation: where ℎ is the theoretical density for the coatings and P is the porosity in percentage that was measured by image analysis technique.

Results and discussion
3.1 Establishing an optimum range for fructose addition As set out in the introduction, multifunctionality of fructose in ceramic processing together with its high compatibility in aqueous suspensions allow to use this saccharide in large amounts in ceramic suspensions. However, the viscosity of the suspension must be kept below certain limits in order not to compromise the feeding of the suspension into the plasma torch as reported elsewhere [29,32]. For this reason, preliminary experimentation was carried out to determine the effect of fructose addition on water viscosity. Not less important, surface tension of water after the addition of different amounts of fructose was also determined due to the great impact of this property on the suspension injection process and the subsequent droplet formation. interior is usually negative because pure sugar displays poor affinity to the surface resulting in an increase in surface tension [33]. With regard to viscosity (determined at high shear rate value of 1000 s -1 ) it can be observed that the addition of fructose produces a plateau from 20 to 50 wt% of fructose contents where viscosity is very stable around 8 mPa·s. As observed, for contents higher than 50 wt%, the effect is pretty similar to that observed in surface tension, i.e. a continuous increase of the property as a consequence, 11 in this case, of the progressive growth of the effective volume of polymer present in the suspension [34]. Due to the aforementioned limitation of SPS system in terms of viscosity as well as the effect observed on water surface tension two amounts of fructose addition (20 and 50 wt%) were proposed for feedstock formulation in this research. Table 1 shows all the suspension feedstock prepared and the final density of the suspensions.  The rheological behaviour of suspensions prepared to 30 vol.% solids content adding the same two fructose contents were very similar to that observed in Figure 2 with the only difference that slightly higher viscosities were obtained due to the higher solids content.
Hence, for the sake of simplicity they have not been included in this paper.
13 14 Viscosity curves of one minute-sonicated suspensions with 20 vol. % solids content are plotted in Figure 3. As it can be seen, rheological behaviour close to a Newtonian profile is confirmed. From these curves, viscosities at high shear rate (1000 s -1 ) were obtained, the values being 4.8, 6.2, and 8.5 mPa·s for 0, 20, and 50 wt % of fructose respectively.
In the same way, for the 30 vol. % solids content suspensions viscosity values were 8.5, 11.4, and 17.4 for 0, 20, and 50 wt % of fructose respectively. It can be seen that the addition of fructose duplicates the viscosity of the suspensions, but it maintains always below the critical viscosity of the injection system as set out elsewhere [29]. previously reported [32]. According to equipment specification, an index value lower than 10 TSI indicates low sedimentation tendency, for this reason an even lower TSI value of 5 has been established in order to estimate an optimum stability time for the SPS process. As observed in Fig.4 the two suspensions containing fructose show stability times longer than 6 hours, which is sufficient for the whole SPS process to occur.
Moreover, the prepared suspensions are easily redispersable, recovering their starting characteristics without persistent agglomeration.  surfaces where clusters of these structures are completely developed. As previously reported, these surface topographies characterized by cauliflower-like structures crown 16 columnar structures inside the coating, as we will see later on [15,16]. As it can be seen, the cauliflower-like topography becomes more pronounced and presents a larger agglomerate average diameter for samples with 20 wt % of added fructose while the coatings obtained from 50 wt % of fructose display smaller size cauliflowers but higher concentration of these structures. These observations can be confirmed by looking at Table 3  replacing ethanol by water as a solvent dramatically increases atomized droplet size [15].

Coatings microstructure characterization
As postulated by VanEvery et al smaller droplet size in the plasma plume favours the growth of coatings with columnar structure [16]. For the same reason, further amounts of fructose in the starting suspensions (from 20 to 50 wt%) will result in lower surface tension liquid and consequently larger amount of fed droplets of smaller sizes. As observed in the micrographs (Fig.5) and Table 2    The cross-sectional micrographs displayed in Fig. 6 show the columnar structure features developed in many SPS coatings. This is a typical microstructure in which lamellae form columnar structures topped with cauliflower-like build-up as seen above [15,16]. As reported by VanEvery et al, the mechanism for generation of columnar coatings relates to the appearance of small in-flight particles once the solvent has evaporated [16]. Thus, these authors proposed a microstructure transition from planar (APS-like) to columnar coatings as average impacting particle size decreases. As observed in Fig.6, non-fructose coatings show a continuous layer of Y-ZTP/Al2O3/SiC where neither cracks nor columns are visible whereas the coatings where fructose is added clearly display a columnar structure growth. In addition, as it can be also observed in Fig.6 for coatings with 20 wt% fructose added, the structure appears more compact, and the columns are wider with intercolumnar voids little pronounced. Whereas, in the coatings with 50 wt% added fructose there is a higher number of columns, although they are narrower with deeper intercolumnar voids which achieve in some extent the same bond coat. These findings agree with those reported by Ganvir et al in a very interesting, recent paper where these authors highlighted the effect of suspension characteristics, in particular surface tension, on the final coating microstructure. Thus, in this paper the authors evinced that the increase of suspension surface tension when using water or even a 50/50 ethanol-water mixture as feedstock solvents instead of pure ethanol was directly responsible for not obtaining a columnar microstructure [36].
Then, when comparing Figs. 5 and 6 a connection between the surface cauliflower-like structures with the columnar structure growth inside the coatings can be established as reported elsewhere [15,16,37]. Hence, as the amount of fructose added to the suspension feedstock increases coatings display greater number of cauliflower-like microstructures of smaller size which lead to a more developed columnar structure in the coatings. As set out above, the effect becomes more pronounced as the solids content rises. These results show, for the first time, that the use of a biocompound such as fructose can play an important role when designing the microstructure of SPS coatings. And much more important, incorporating fructose in the suspension feedstock allowed to develop columnar microstructure coating in water-based highly concentrated systems.  Fig.7b shows that, overall, coating porosity increases by adding fructose. This fact is in accordance with the porosity former functionality of fructose as described in literature [20][21][22]. Nevertheless, coatings prepared from suspensions containing 50 wt% fructose do not follow the increasing tendency, particularly the coating deposited from the higher solids content suspension (30 vol.%). This behaviour may be mainly due to the highly developed columnar microstructure where a greater number of narrow columns and intercolumnar voids are formed replacing closed voids inside the coating. Finally, Fig. 7c shows how the addition of fructose in the feedstock suspension greatly decreases the amount of resolidified/unmelted zones. The diminution is especially significant for 30 vol.% coatings, from 24.9% (without fructose) to 9.9% (20 wt% fructose) and 6.7% (50 wt% fructose). This tendency may be probably related to the effect of fructose on surface tension of the feedstock liquid as set out in section 3.1. Thus, the addition of fructose drastically reduces the surface tension of water resulting in a more homogeneous, smaller average droplet size distribution feeding into the plasma torch. This more homogeneous distribution favours a better build-up of molten droplets inside the core zone of the plasma plume. As a consequence, there is a decrease in the amount of suspension droplets that reach the substrate surface in a poor melting state from the plasma fringe, therefore resolidified/unmelted zones formation is diminished [38]. Additionally, an enthalpy increase of plasma plume associated with fructose burnout should not be underestimated [39].