Comparison between micrometer and millimetre sized metal and ceramic lamellas, for a better understanding of the spray process.

Transcription

1 Comparison between micrometer and millimetre sized metal and ceramic lamellas, for a better understanding of the spray process. S. Goutier, M. Vardelle, and P. Fauchais SPCTS Laboratory, University of Limoges, 123 avenue Albert Thomas, Limoges Cedex Abstract: An experimental set-up has been developed, at the SPCTS Laboratory of the University of Limoges, to produce fully melted, millimetre-sized, ceramic or metal drops with impact velocities up to 10 m/s. Such impact velocities allow reaching impact Weber numbers, close to those of the plasma spray process (We=2300) with droplets in the micrometer sized range. A fast camera (4000 image/s) combined to a fast pyrometer (4000 Hz), allows following the drop flattening. The flattening of droplets (at the micrometer scale), has been studied with a direct current (DC) plasma torch spraying particles about 45 µm in diameter. The corresponding experimental set-up comprised a very fast (50ns) two-color pyrometer and two fast exposure (at least 1 µs) CCD cameras (one orthogonal and the other tangential to the substrate). The flattening of millimetre and micrometer sized particles are compared. First are studied impacts of alumina and nickel-aluminium drops (millimetre sized) with impact velocities up to 10 m/s. Then are considered the same particles but micrometer sized (about 45 µm in diameter) sprayed with the DC plasma torch. A correlation has been found between both flattening scales. This work shows that when comparing phenomena at the two different scales (three orders of magnitude difference both in size and flattening time), the Weber number is by far more important than the Reynolds one and it is mandatory to have Weber numbers as close as possible to make pertinent comparisons. Keywords: plasma spray, Weber, Reynolds, particle flattening, surface treatments, flattening velocity 1. Introduction The flattening of plasma sprayed droplets (few tens of micrometer in diameter) is very difficult to follow experimentally according to the very short flattening time (a few µs) 1-4. That is why many works have been devoted to the flattening of millimeter-sized drops where characteristic flattening times are in the millisecond range Indeed, the flattening studies for millimeter sized and micrometer sized melted particles, called successively in this study drop and droplet, showed similarities for the liquid material flow onto the substrate and the different types of liquid droplet ejections 11. To predict the maximum flattening diameter, most studies were focused on the energy balance at impact, and/or series of experiments with different impacting materials, melted particle diameters and impact velocities. These researches allowed deducing evolutions of the maximum flattening diameter linked to the dimensionless numbers (mainly the Reynolds, Re o, and the Weber, We o, numbers at impact). These equations, established according to the particle characteristics at impact, however do not take into account substrate properties and its surface chemical and physical characteristics. In the frame of this study, the parameters of the incident particles were kept constant as much as it possible. Only the substrate surface parameters were modified and to take them into account the only expressions available are those related to the thermal contact resistance (Dhiman 12 ), or the contact angle (Pasandihed 13 ), or the flattening velocity (Fukumoto 14 ). However experiments do not allow measuring

2 the time evolution of the thermal contact resistance or the contact angle of the liquid with the substrate. The only measurable parameter is the flattening velocity. In a first part of this work, the evolution the flattening velocity is presented at both scales according to the impacting particles and the substrate temperature. Then in a second part, in order to take into account the viscosity and the surface tension strength of the particle, the Reynolds and Weber numbers are presented according to the particle flattening conditions. 2. Experimental setup Experimentations are carried out with two techniques for studying impact and flattening phenomena. The first one is a modified free-falling set-up to study millimeter-sized drops with high Weber numbers (up to 2000) and the second one uses a plasma spraying set-up with micrometer-sized drops. Running in parallel these two studies enable to compare particle flattening and cooling on smooth stainless steel (304L) or titanium alloy (Ti-6242) substrates, preheated or not, at spatial and time scales differing by almost three orders of magnitude Millimeter sized particles study To produce liquid ceramic or metal drops, a rod is introduced and melted in an electrical arc furnace. A suspended drop is formed. When the gravity force overcomes that of the surface tension, the drop falls. This set-up is disposed into an argon controlled atmosphere chamber. The substrate, fixed onto a pneumatic jack, can be moved up during the drop fall and the relative impact velocity can reach up to 10 m/s 15. A detector located at the chamber output, generates a TTL pulse when one single drop crosses its measuring volume. This pulse is then directed to the measuring system composed of a fast camera (Photron 4000 image/s) targeting the substrate Micrometer sized particles study Particles are plasma sprayed using a direct current (D.C.) plasma torch (PTF4 type) with a 6 mm internal diameter nozzle and running with a mixture of argon-hydrogen. The arc current is 650A, the argon flow rate 33 L/min and the hydrogen volume percentage 25%. An alumina powder and nickel aluminium powder with particle sizes between 40 and 50 µm are used in this study. These spray conditions result in fully melted particles with velocities at impact around 200 m/s. Lamellas are collected on a smooth (Ra=0.06 µm) 304L substrate at a distance of 110 mm downstream of the nozzle exit. The measuring system comprises: - In-flight measurement of droplet velocity by a two points measuring optical detector and a fast (response time = 50ns) dichromatic pyrometer to follow the particle temperature during its flattening and cooling. The two wavelengths of the bichromatic pyrometer are 690 µm and 710 µm respectively. -Imaging techniques which allow following the flattening of the particle. On each camera are fixed a macro lens with a focal distance of 200 mm and two focal doublers. One camera is disposed orthogonal to the substrate and the other one tangential to it. With different exposures times and a variable time delay compared to the impact time, it is possible to observe the flattening at different times but for different particles (assuming they have the same parameters at impact) Substrates Substrates are made of stainless steel (304L) and titanium alloy (Ti-6242), are mirror polished by using SiC paper 4000 and disposed at 110 mm from the nozzle exit. They are fixed on copper supports heated up to 200 C with a heating rate of 0.5 C/s by two small resistances (each one with a power of 150 W). A monochromatic pyrometer (Ircon 5 µm, 10 ms response time) controls the substrate temperature during the preheating stage Flattening velocity The measurement of the flattening velocity is complex. Indeed for droplets, it is not possible to follow the particle flattening in real time. It is thus necessary to make some hypotheses to determine the

3 flattening velocity. The only solution, to determine the flattening time, is to measure the time necessary to reach the maximum flattening diameter. So it must be to consider that during flattening the cooling is negligible. To confirm this assumption, it can be noticed in images presented in the figure 1 that the maximum of the pyrometer signal is reached at the maximum flattening. the maximum flattening diameter, to calculate this mean velocity. To compare drop and droplet flattening, it is necessary to have a dimensionless approach considering important diameter and velocity variations. Thus, results are normalized by using the factors ξ max and a, defined by the following expressions: - Factor of maximum flattening: ξ max = D max /D o where: D max : diameter of maximum flattening. D o : diameter of the incident particle. - Normalized flattening velocity: a=v e /v o Figure 1: Pictures of droplet flattening, on a cold stainless steel substrate for alumina and nickel aluminium droplets. The shaded area on the pyrometer signal represents the opening time of the camera shutter. Immediately afterwards the ejections and the film rupture occurs 11. By measuring the maximum flattening diameter (D max ), thanks to the image of the camera 1 and the time corresponding to the maximum of the pyrometer signal at a given wavelength, it is possible to calculate a mean flattening velocity. For drops, as shown in figure 2, two flattening velocities can be observed. Figure 2: Evolution of flattening diameter for an alumina drop impacting on a stainless steel at room temperature In order to simplify the comparisons, only the mean flattening velocity is considered. By using images of the fast camera, it is possible to deduce the time of where: v e : mean flattening velocity 3. Results v o : particle impact velocity 3.1. Case of Al 2 O 3 on 304L In figure 3 are presented the maximum flattening factor (ξ max =D max /D o ) and the normalized flattening velocity velocity (a=v e /v o ) for substrates heated to various temperatures respectively for alumina drops (millimeter sized) and droplets (micrometer sized) impacting on a stainless steel substrates. For droplets, when the factor a increases, the droplet flattens much more. In the drop case, the experimental points are rather dispersed (see figure 3) when the impact velocity is modified (from 3 to 10 m/s). The ratio is similar but when the kinetic energy at the impact is larger, the particle flattens more. Normalized flattening velocity velocity, noted a, seems to depend on the substrate temperature. That is confirmed in figure 3 for drops and droplets. For a surface at the room temperature, factors a are close for the two scales. But when the substrate surface is heated, the factor a falls down much faster for droplets than for drops. That means that on a hot surface, the droplet flattening is slowed down faster than in the case of a drop. The particles cooling are not identical at the two scales. The

4 kinetic energy variation of the flattening droplet is more important according to the viscosity energy, which takes part in the flow deceleration. This phenomenon is less important for drops, where a bigger quantity of matter is present. Thus at micrometer scale, flattening is controlled by the boundary layer between the flattening particle and the substrate, whereas this boundary layer is negligible for drop flattening. For drops, the flattening velocity rises according to the increase of the maximum flattening factor and decreases with the increase in the substrate temperature. However, in the case of a droplet impacting on a titanium alloy substrate, no real evolution of the factor a is detected, both with the maximum flattening factor and the substrate temperature. It can be due to the weak variation of the diameter when the substrate is heated. Figure 3: Evolutions of normalized flattening velocity, noted a, (a=v e /v o ), for an alumina melted particle on a stainless steel substrate, in function of the maximum flattening factor ξmax (ξ max =D max /D o ), the substrate temperature Tsubstrat 3.2. Case of NiAl on Ti-6242 In figure 4 are presented the maximum flattening factor (ξ max =D max /D o ) and the normalized flattening velocity velocity (a=v e /v o ) for substrates heated to various temperatures respectively for nickel aluminum drops (millimeter sized) and droplets (micrometer sized) impacting on titanium alloy (Ti- 6242) substrates. Figure 4: Evolutions of normalized flattening velocity, noted a, (a=v e /v 0 ), for an alumina melted particle on a titanium (Ti- 6242) substrate, in function of the maximum flattening factor ξ max (ξ max =D max /D o ), the substrate temperature Tsubstrat An important point must also be underlined: in all impacts of NiAl droplets on titanium alloy substrates, it was not possible to obtain perfect disk shaped lamellas. All lamellas presented either a film or a peripheral phenomenon of ejection, indicating the presence of kinetic energy excess, which this translated here by the important (around 0.8) value of the factor a. It is probable that if it had been possible to heat the substrate at higher temperatures, the factor a would have more decreased.

5 3.3. Conclusions of experiments The comparison of the flattening velocity evolutions with the maximum flattening diameter and substrate temperature highlights, the differences, between a metal particle and a ceramic one. For a ceramic droplet, the range of normalized flattening velocities is important (from 0.2 to 1.1) whereas for a NiAl droplet, it is weak (from 0.8 to 1.1). The important variation observed in the case of ceramic droplets can be related to the evolution of the liquid viscosity according to the cooling velocity. Indeed, for alumina drops, cooling on cold and hot substrates are overall close together and the variation of normalized flattening velocities a is weak (from 1 to 1.3), whereas for droplet, cooling velocities are very different and the variation of a is important (0.2 to 1.1). As expected, the faster is the cooling, the more the liquid flow on the substrate is slowed down. For drops, on cold substrates, alumina and NiAl have the same flattening velocity, but it is no more the case on hot substrates. The NiAl lamellas are adherent on the substrate, whereas it is not the case of Al 2 O 3 ones: the latters have not a sufficient contact with the substrate, to drastically influence the flattening liquid flow, in spite of desorption of adsorbates and condensates at hot substrate surfaces. Indeed, in spite of the substrate temperature beyond the transition temperature and circular lamellas, ejections are always present in their periphery for alumina (see Table 1, 15 ). necessary to carry out a dimensionless approach of our results. 4. Discussion 4.1. Variation of flattening velocity according to the surface tension, and the viscosity of the particle. In order to find a link between drops and droplets, the flattening diameter evolution is presented according to the physical parameters of the particle. Thus, Reynolds and Weber numbers are calculated according to flattening parameters and not with the impacting parameters. For these dimensionless numbers, it is necessary to know characteristic dimensions of the flow. Table 1: Alumina drop flattening on polished stainless steel for an impact velocity of 10 m/s and an subtrate temperature of 200 C. t=0ms t=3.2ms t=6.4ms In conclusion, for drops and droplets, to avoid the peripheral ejection phenomenon, it is necessary to achieve normalized flattening velocities lower than 0.8. But it exist some differences between the alumina and nickel aluminum case. It is now Figure 5: Evolutions of the flattening particles (metal and ceramic) at two scales, with the Reynolds (Re=ρv e R max / µ), the Weber (We=ρv e 2 R max /σ), the velocity taken into account being that of the liquid on the substrate and characteristic dimension being equal to the radius of maximum flattening. The mean flattening velocity is considered, and for the characteristic dimension it is the maximum flattening radius, in order to obtain curves presented in figure 5.

6 No link exists between Re and the maximum flattening when the material (metal or ceramic) and diameters (40 µm => 5 mm) of the impacting particle are modified. But, the evolution of the lamella diameter seems to be strongly related to the evolution of We. Indeed, the flattening curves corresponding to various conditions of the substrate (cold, hot) and of particles (metal, ceramic) are close to each other. This means that the maximum flattening is controlled by the competition between the surface tension and inertia energies. That also means that the effects of the surface tension are preserved during the modification of the particle scale, due to the weak variation of the surface tension when the liquid is cooling. It is not the case of the viscous effects which are strongly related to liquid cooling Comparison with literature As introduced previously, the only relationship previously developed is the factor of Fukumoto noted K f 14. This factor corresponds to the following expression: In the case of NiAl droplets impacting on a titanium alloy substrates, the transition temperature could not be reached (due to the system limits) and the value of K f is higher than 7. However, the shape of the lamellas is close to that of perfect disks (see figure 6, case at 350 C,). For drops, however it was possible to obtain lamellas with very few ejections (T substrate =300 C), and K f parameter lower than 7. K f = 0.5 a 1.25 Re K where K is the Sommerfeld parameter defined by K= We o 0.5 Re o 0.25 In this case, the dimensionless numbers are calculated by using the parameters of the impacting particle. Fukumoto uses drops to find the critical value of K f. This critical value, noted K, is defined as the value below which only circular lamellas are c obtained. He found that K f =7, in the case of metal drops on various materials. However, Fukumoto could not confirm this value for droplets, being unable to measure the flattening velocity at this scale. However here, it is possible to calculate K f parameter at both scales and figure 6 are obtained. For droplets, the alumina lamellas presenting disk shapes and very few ejections (Tsubstrate=200 C) obtained on a stainless steel substrate have a Kf value lower than 7. For alumina drops, in spite of a substrate temperature over T t and disk shaped lamellas, the values of K f are higher than 7, and many ejections are present. c f Figure 6: Evolutions of the factor K f according to the substrate temperature for two types of particles ceramic (Al 2 O 3 ) metal (NiAl). According to 14, below Kfc=7, only circular lamellas are obtained. In conclusion, it seems that, when K f is lower than 7, some disk shaped lamellas are obtained. But below a value of 7, lamellas present a jagged form or ejections. This is in good accordance with results found by Fukumoto. 5. Conclusions The comparison of the two flattening scales, to understand the impact phenomena, shows many similarities (in the flattening process and the ejection phenomena), but also some important differences.

7 Indeed, the thermal thickness of boundary layer of the flattening melted particles plays a more important role with droplets than with drop. Such finding involves modifications in the flattening process, which can be classified according to the substrate temperature: - for a cold substrate, the desorption of the adsorbates and condensates generates a gas cloud at the interface flattening particle-substrate. This gas cloud has not the effect at both scales. For micrometer-sized droplets, the flattening particle cannot solidify because it is lifted over the substrate. Because of low thickness of the liquid film, gases generated at the interface can cross it, destabilize it and cause de-wetting. It is not the case for drops where the cloud effect is negligible due to the important thickness of liquid flow, which is solidified without de-wetting. This difference is observed for the two types of materials. - for a hot substrate, the phenomenon of solidification at the interface allows dissipating the flattening energy and avoid ejections. For drops, due to the more important flowing liquid thickness, the upper part of the flattening drop can be always liquid whereas the bottom part in contact with the substrate is solid. This phenomenon induces the presence of ejections at the time of the maximum flattening and an important recoil phenomenon in the direction of the flattening drop center, while for droplets, because of thickness of the flattening droplet and the cooling velocity, most of the lamella is probably solidified at the time of maximum flattening. References 1. K. Shinoda, Y. Kojima and T. Yoshida, Journal of Thermal Spray Technology 14 (4), (2005). 2. K. Shinoda, H. Murakami, S. Kuroda, K. Takehara and S. Oki, Journal of Thermal Spray Technology 17 (5-6), (2008). 3. N. Z. Mehdizadeh, M. Lamontagne, C. Moreau, S. Chandra and J. Mostaghimi, Journal of Thermal Spray Technology 14 (3), (2005). 4. A. McDonald, S. Chandra and C. Moreau, Journal of Materials Science 43 (13), (2008). 5. M. Fukumoto, E. Nishioka and T. Matsubara, Journal of Thermal Spray Technology 11 (1), (2002). 6. R. Dhiman and S. Chandra, Int. J. Heat Mass Transf. 48 (25-26), (2005). 7. R. Dhiman and S. Chandra, presented at the Proceedings of the ASME Heat Transfer/Fluids Engineering Summer Conference 2004, HT/FED 2004, Charlotte, NC, 2004 (unpublished). 8. S. Amada, M. Haruyama, T. Ohyagi and K. Tomoyasu, Surf. Coat. Technol. 138 (2-3), (2001). 9. W. M. Healy, J. G. Hartley and S. I. Abdel- Khalik, Int. J. Heat Mass Transf. 39 (14), (1996). 10. S. Chandra and P. Fauchais, J Therm Spray Technol 18 (2), (2009). 11. S. Goutier, M. Vardelle, J. C. Labbe and P. Fauchais, Journal of Thermal Spray Technology 20 (1-2), (2011). 12. R. Dhiman, A. G. McDonald and S. Chandra, Surface and Coatings Technology 201 (18), (2007). 13. M. Pasandideh-Fard, Y. M. Qiao, S. Chandra and J. Mostaghimi, Phys. Fluids 8 (3), (1996). 14. M. Fukumoto, E. Nishioka and T. Nishiyama, Surface and Coatings Technology 161 (2-3), (2002). 15. S. Goutier, M. Vardelle, J. C. Labbe and P. Fauchais, J Therm Spray Technol 19 (1-2), (2010).