Splat formation and microstructure of solution precursor thermal sprayed Nb-doped titanium oxide coatings

Solution precursor thermal spray can become a breakthrough technology for the deposition of coatings with novel chemistries; however, the understanding of the process that the feedstock material undergoes is still poorly understood when compared to more traditional presentations (i.e. powder and suspension). In this paper, niobium-doped TiO2 coatings were deposited by solution precursor high velocity oxy-fuel spraying, studying its microstructure and phase. It was reported that a lower flame temperature produced a highly porous coating, while the porosity was reduced at higher flame temperature. Investigation of the phase content showed that, contrary to our current understanding, a higher flame power implied an increase of the anatase phase content for solution precursor spray. Three methods were used: Rietveld refinement, peak height and peak area of the x-ray diffraction patterns. Additionally, single splats were analysed, showing that as the precursor travels through the flame, pyrolysis and sintering takes place to form the solid material. These results were used to derive a model of the physico-chemical transformation of the solution precursor. This work proves that solution precursor thermal spray is a promising technique for the deposition of doped ceramic coatings, being the microstructure and phase content controllable through the spraying parameters


Introduction
Titanium oxide coatings are widely used in a number of fields owing to its unique properties, such as a prominent photocatalytic effect (Ref 1), its electrochemical activity (Ref 2), its coherent change in electrical conductivity under gas exposure (Ref 3) or its ability to produce transparent coatings (Ref 4,5). Nevertheless, the properties of titanium oxide (also known as titania) can be tailored to be more beneficial through doping (Ref [6][7][8][9][10][11]. Among those elements, niobium presents some desirable properties that makes it a suitable candidate for the formation of doped titanium oxide coatings. It effectively increases the electrical conductivity which, in addition to the favourable electrochemical properties of TiO2 (such as high capacity and low volume expansion during ion charge/discharge (Ref [12][13][14]), makes it a viable option as an anode material for high-power Li-ion batteries. The ability to produce transparent coatings has increased its interest regarding its use as a transparent conductive oxide (TCO) coating. In the recent years there has been a growing demand for a replacement of the current TCO industrial reference, Sn-In2O3, due to limited resources and increasing costs ( Ref 15). An increase of the coating conductivity, coupled with high transmittance in the visible range, positions Nb-doped TiO2 as a viable candidate for the next generation of TCOs (Ref [16][17][18]. In addition to the improvement in electrical conductivity, Nb-doped TiO2 coatings have also found application as sensors, as the niobium limits the grain growth and inhibits the anatase to rutile phase transformation to temperatures up to 650 °C ( Ref 19), being the rutile phase traditionally considered detrimental for sensing applications ( Ref 11,[20][21][22][23]. Due to the plethora of applications present, production of niobium-doped TiO2 has been reported in the literature using techniques such as sputtering ( Ref 18,24,25), pulsed laser deposition ( Ref 26,27), atomic layer deposition (Ref 28), spin-coating ( Ref 11,29) or aerosol-assisted chemical vapour deposition ( Ref 17). At the same time, thermal spraying deposition of titanium oxide has been thoroughly studied before; using powder, suspensions and solution precursors as feedstock material . In thermal spray, a heat source is used to melt the feedstock material while a jet carries the molten particles towards the piece to be coated ( Ref 34). In the case of high velocity oxy-fuel (HVOF) thermal spraying, the flame is produced via the combustion of mixed oxygen and fuel on a pressurised combustion chamber, which leaves through a nozzle creating a supersonic jet. The traditional presentation of the feedstock material is in powder form; however, that imposes a lower limit to the size of the particles to ensure adequate flowability. Suspension, and more recently, solution precursor thermal spraying have been devised as a route to avoid this limitation. In particular, solution precursor eliminates the need for suspended particles in a liquid medium. Instead, the precursors are mixed in a liquid form and then react in-flight, due to the heat transfer, to form the Although a preference for titanium dioxide coatings with a high anatase content has been the norm, recent studies on the photocatalytic activity of suspension HVOF thermal sprayed titania (Ref 38) suggest that the microstructure of the deposited coating and most importantly, the interaction between rutile and anatase regions, plays an essential role in the presence of an enhanced photoactivity at relatively low anatase content (~20%). Therefore, quantitative phase content determination of XRD diffraction pattern of the deposited Nb-TiO2 has been carried out in this work to better understand the formation of anatase and rutile from solution precursor feedstock.
In this work, a comprehensive study of the deposition process and microstructure of Nb-doped TiO2 coatings produced using solution precursor high-velocity oxy fuel (SP-HVOF) thermal spray is presented. The process from the original precursor solution into individual droplets when exposed to the flame and the formation of the coating upon impact with the substrate, as well as the characteristics of the deposited coatings, were investigated. Scanning electron microscopy (SEM) was used to determine the morphology of the individual splats and the coatings, and x-ray diffraction was applied to evaluate relationship between the phases present, their content and the spraying parameters used. To analyse the evaporation process and high temperature behaviour of the solution precursor, thermogravimetry (TGA) and differential scanning calorimetry (DSC) analysis were performed.

Materials and coating deposition
The solution precursor was provided by EpiValence Ltd. (Cleveland, United Kingdom) and contained a mixture of titanium ethoxide and niobium ethoxide with weight percentages of 15.0 % and 1.35 % respectively, dissolved on 2-isopropoxyethanol.
The coatings were deposited using a modified GTV TopGun HVOF system with an injector diameter of 0.3 mm directed towards a 22 mm long combustion chamber. A detailed description of the setup can be found elsewhere ( Ref 39). Two sets of spraying parameters were used, corresponding to a flame power of 25 kW and 75 kW. For the 25 kW flame, the hydrogen flow rate was 78 l/min and the oxygen flow rate was 182 l/min. For the 75 kW flame, the hydrogen flow rate was 229 l/min and the oxygen flow rate was 533 l/min. In both cases the stand-off distance was 85 mm, the carousel rotation speed was 73 rpm (which corresponds to a surface speed of 1 m/s) and the gun traverse speed was 5 mm/s. 10 passes were performed to build up a coating of the desired thickness.
In order to elucidate the transformations that take place once the solution precursor enters the HVOF flame, single splats were collected on stainless steel polished substrates following a swipe test.
To do so, the carousel rotation and gun traverse speed were increased to their maximum values (100 rpm and 30 mm/s respectively), while only allowing one pass of the flame. In addition, the spraying was repeated three times maintaining all spraying parameters fixed with the exception of the standoff distance, which was chosen to be 65, 85 and 105 mm, aiming to provide three different snapshots of the evolution of the droplets as they travel along the flame.
The substrates used, with dimensions 60 x 25 x 2 mm, were AISI 304 stainless steel (SS) with nominal composition of Fe-19.0Cr-9.3Ni-0.05C (in wt. %). For the deposition of coatings, the substrates were grit blasted with a blast cleaner from Guyson (Dudley, England) with fine F100 brown alumina (0.125 -0.149 mm) particles at 3 bar. Following grit blasting, the substrates were cleaned in industrial methylated spirit (IMS) using an ultrasonic bath for up to 10 minutes and blown dry with compressed air. In the case of single splat collection, the surface of the substrate was ground and polished down to 1 µm finish. The process was done starting with Buehler SiC grinding paper (Essligen, Germany) grit 220 (P240) until a uniform ground surface was achieved. The process was continued using grinding papers with grit 320 (P400), 400 (P800) and 600 (P1200). The polishing process was carried out using a 6 µm polishing paper, finally moving into 1 µm polishing paper for the final preparation.

Characterisation
Cross section of the coatings were prepared cutting a section of the substrate using a SiC cutting wheel (MetPrep Ltd, Coventry, United Kingdom) at a speed of 0.010 mm/s on an Brilliant 220 (ATM GmbH, Mammelzen, Germany) cut-off machine. The cut section was then hot-mounted using Conducto-Mount resin from MetPrep following the recommended standard procedure. The mounted cross section was then grounded and polished down to 1 µm using the same procedure as described above. Briefly, the process was done using SiC grinding papers with grits P240, P400, P800 and P1200 and 6 and 1 µm polishing papers.
A FEI Quanta 600 (FEI Europe, Eindhoven, Netherlands) scanning electron microscope (SEM) was used to image the cross section, surface and single splats of the deposited Nb-TiO2, using secondary electron (SE) and backscattered electron (BSE) modes. A spot size of 2.5 nm and an acceleration voltage of 20 kV were used as the imaging parameters. The coatings were also analysed using a Siemens D500 powder X-ray diffractometer in Bragg-Brentano θ -2θ geometry equipped with copper anode X-ray tube and a scintillation point detector. The 2θ range scanned by CuKα radiation (with a wavelength of 1.54 Å) was from 20° to 120° with 0.04° step size and 22 s of counting time in each step.
Peak identification was performed using the diffracsuite EVA (Bruker Software) and Rietveld refinement procedure was applied to the obtained results using TOPAS V5 software. A split pseudo-Voigt function was used to account for the base broadening of the two anatase and rutile reflections, believed to be caused by stacking faults on the crystallographic structure and some degree of amorphous content. To account for instrumental broadening effects, the specifics of the XRD instrument, such as source emission profile, detectors and slits, were defined during the refinement process. Structural values for the anatase and rutile were obtained from the inorganic Crystal Structure Database (ICSD). Since both coatings have a thickness bellow 20 µm, x-ray penetration caused the appearance of peaks from the stainless steel substrate. These peaks were taken into account during the refinement as well, although they were not considered for the total phase quantification.
In addition to Rietveld refinement, two other methods were applied to calculate the phase content of the coatings. The    For the peak fitting of the XRD diffraction patterns the star quality PDF entries with reference codes 01-072-7376 (Nb-TiO2, rutile phase), 00-021-1272 (TiO2, anatase phase) and 00-033-0397 (304 austenite, stainless steel) were used. Unfortunately, no niobium doped anatase XRD diffraction pattern could be found in the database used (PDF-4+ 2018). The XRD spectra already indicates that the 75 kW coating presents a higher content of anatase phase.
Due to the importance of the content of both phases for the physico-chemical properties of the coating, Rietveld refinement was performed, as shown in Figure 3, to obtain the normalised phase content percentage of anatase and rutile. It should be noted that the Rietveld refinement process was carried out including all phases detected in the XRD diffraction patterns: the two titanium phases (anatase and rutile), those associated with the stainless steel substrate (iron α and iron γ) as well as a weak signal corresponding to corundum and associated with alumina particles from the grit blasting of the substrate. Since the thickness of the two coatings is not the same, and x-ray penetration on the sample at 25 kW was deemed to be excessive, the values presented in Figure 3

In-flight transformation
In order to better understand the in-flight transformation of the solution precursor into solid particles, and consecutively a coating, DSC-TGA analysis was performed to gain insight on the thermodynamic behaviour of the solution. Figure 5 shows the results obtained from the analysis, where both the weight change and the heat flow from room temperature to 1500 °C are presented. To study in detail the pyrolysis and crystallisation of the amorphous Nb-TiO2, TGA analysis of the dried solution was performed with a heating rate of 10 °C/min up to 800 °C. The results can be seen in Figure 6. Since the temperature range has been reduced and the thermodynamic contribution of the solvent has been greatly removed, the data in Figure 6 provides a more clear insight into the formation of Nb-doped TiO2, its crystallisation and the phase transformation. The first peak, at 277.65 °C, corresponds to the previously identified peak for the pyrolysis of the precursor. The next peak, at 393.08 °C presents a broader aspect, usually associated with a crystallisation process. Nevertheless, the complete crystallisation from amorphous phase takes place at 498.38 °C (Ref 33). The last peak, at 563.34 °C, would correspond to the transformation from anatase to rutile.

Swipe test
As mentioned in the experimental methods section, swipe tests were performed to study the  As it can be seen in Figure 7, a liquefied splat has deformed upon impact, while preserving a brighter exterior layer and a darker interior. This difference in brightness is also reflected on the EDS measurements. Along the four different points measured, the titanium and oxygen contents remains fairly constant, being the only difference the niobium content. As expected due to the heavier nature of niobium, the brighter areas, associated with the exterior layer, present a higher content of niobium.
The top surface of the substrates, with the produced splats, is shown in Figure 8. Some features can be identified from the images, such as the reduction in deposition efficiency as the stand-off distance is increased. Nevertheless, the surface of the samples at 65 mm is mostly formed by solid particulates with a reduced presence of round, fully molten splats. Such features are more evident at 85 mm, where a mixture of solid particulates and molten splats is present. In the case of stand-off distance of 105 mm, it is considered to be excessive, as the deposition efficiency dramatically drops, without an improvement on the morphological aspect of the splats present. Some fully molten splats are still seen, accompanied by smaller, round particles. From the SEM images in Figure 8, it can be extracted that the amount of bright, nanosized particles is reduced as the solution precursor travels along the flame. On the other hand, the amount of molten material with relation to the solid particles increases with increased stand-off distance. This effect is partially due to the reduced deposition efficiency at larger stand-off distances; however the proportion of solid particles to molten material provides information of the physical transformation of the precursor.
A proposed model for the physical transformation that takes place once the solution precursor is injected into the system is here presented, being schematically depicted in Figure 9.  The specific dynamic behaviour is influenced by the dimensionless values of the Weber number (We) and the Reynolds number (Re), being defined as: Where ρ is the density of the solution, νrel is the initial relative velocity between ambient and drop, d is the initial diameter, σ is the surface tension and µ is the dynamic viscosity. These two quantities provide a quantification of the ratio of the fluid inertia to the surface tension (We number) and the ratio of the fluid inertia to viscosity (Re number).
Once the mixture of solution precursor and gases comes out through the nozzle, entering the HVOF flame, the droplets experience a secondary fragmentation into smaller structures due to shear deformation from the drag forces (Ref 51). This process stablishes the final size of the droplets, being therefore crucial for the differentiation between the various morphologies, as illustrated in Figure 10.
In this case, the We number is the one mostly responsible for the changes in morphology, as an increase in its value causes the droplets to experience a different secondary fragmentation from Figure   10, being the first morphology associated with lower We number and the last with higher values. This solid phase, present as bright dots within deposited splats visible only at high magnification (visible in Figure 9a, marked with arrows), was confirmed to be niobium-rich phases through energydispersive x-ray spectroscopy. The increased niobium content was caused by the segregation of the substitutional niobium present in the lattice of the bulk material (Ref 46,54) when exposed to high temperatures.
From the first appearance of the solid phase, two phenomena can occur. If the rate of solvent vaporisation is greater than the diffusion of the solute, a supersaturated external layer will appear surrounding the molten material, as seen in Figure 9b, and confirmed by EDS measurements on the swipe test section. In the opposite case, the condensed phases diffuse to form precipitates. This route starts with the formation of precipitates, which is followed by pyrolysis and sintering, where three morphologies can arise depending on the conditions experienced in-flight by the droplet. In the first case, if a small droplet experiences a low heat rate, a dense solid particle will form. The presence of such solid particles has been observed indirectly, as the surface of some samples from the swipe test at stand-off distance of 65 mm showed multitude of craters such as the one see in Figure 9c, indicating the impact of solid material of micrometric size. The lack of evidence for the presence of solid features on samples with larger stand-off distances indicates that these particles either lose most of its kinetic energy after traveling 65 mm, not creating craters upon impact, or they melt when traveling longer distances. Another possibility arises if the size of the initial droplets is larger or they experience higher heating rates. In this case, the formation of agglomerates of round particles will be favoured, being an example shown in Figure 9e. These agglomerated structures were more common than the craters from solid material, which provides information on the conditions of the flame and the primary and secondary fragmentation, seemingly favouring large droplets and/or higher heating rates. A third option, with characteristics in between the two previous structures was also identified and can be seen in Figure 9d. In this case, a large agglomerate possesses a solid core, marked with a circle on the BSE-SEM image. The final possible stage of the deposition process takes place when the agglomerates break up into smaller fragments, or if they never reach the critical size to endure the trajectory within the flame without re-melting, as exemplified by the images on Figure 9f and Figure 9g.

Conclusions
The SP-HVOF deposition technique was used to produce niobium-doped titanium oxide coatings for the first time. The results show that the flame power, chosen in this study to be 25 kW and 75 kW, has critical implications on the microstructural features and phase content of the produced coatings.
The effects observed were:  Flame power was proven an effective way to modify the microstructure of the deposited coatings. Low flame powers induced the sintering of solid material but failed to melt them, producing a coating with porous structure. Higher flame power allows the melting of the solid particles, leading to a lower porosity.
 Three calculation methods confirmed that an increase in flame power equated to an increase in the anatase phase content. This behaviour is believed to be caused by the presence of doped niobium, which hinders the anatase to rutile phase transformation, and the complete melting of the solid content formed in-flight at 75 kW, promoting the anatase phase content.
 In addition to the flame power, three stand-off distances (65, 85 and 105 mm) were used to collect single splats following a swipe test. The features present, as well as the morphologies discovered, were the base for the development of a model of the physicochemical transformations that the solution precursor experiences. A detailed mechanism for the sintering of solid materials, along with the changes as it progresses through the flame, is presented.

Acknowledgments
This work was supported by the Engineering and Physical Sciences Research Council (grant number EP/L016206/1). The authors would like to thank Rory Screaton and John Kirk for their assistance during the SP-HVOF spray, Sunil Chadha for the computational simulations of the flame temperature and his input on droplet fragmentation and EpiValence Ltd. for providing the solution precursor.