Portal de Periódicos da CAPES

Effect of Substrate Temperature on Splashing of Molten Tin Droplets

2004; ASM International; Volume: 126; Issue: 3 Linguagem: Inglês

10.1115/1.1737778

1528-8943

N. Z. Mehdizadeh, Mehdi Raessi, S. Chandra, J. Mostaghimi,

Additive Manufacturing and 3D Printing Technologies

When a molten metal droplet hits a solid plate and freezes, the shape of the flattened, solidified splat that is formed depends on the temperature of the plate. Figure 1 shows aluminum splats produced by wire arc spraying, a widely used coating process in which an electric arc is struck between the tips of two continuously fed wires. A high velocity air jet directed at the gap between the wires strips off droplets of molten metal and propels them onto the surface being coated. In the study 1 from which Fig. 1 was taken droplet diameters ranged from 16 to 25 μm and velocities from 100 to 125 m/s. Droplets that landed on a polished stainless steel plate at 25°C formed irregular shaped splats (Fig. 1(a)), while those impinging on a plate heated to 350°C produced almost perfectly circular splats (Fig. 1(b)). Why does surface temperature affect splat shape? The answer to this question is of much more than academic interest since coating quality depends on the shape of individual splats 2. Fragmented splats produce porous coatings with low adhesion strength. Splashing—defined as disintegration of a single droplet upon impact to produce satellite droplets such as those visible in Fig. 1(a)—reduces deposition efficiency (the fraction of the coating material which adheres to the surface) because smaller droplets bounce off the surface. The effect of varying substrate temperature is not well understood in the thermal spray industry. Coating applicators cool components during spraying to avoid thermal distortion, but rarely make any further effort to monitor or control substrate temperature. Appropriate thermal management may offer ways of improving coating quality and reducing wastage, but it is not presently clear what optimum surface temperature should be maintained during spraying. Controlling the properties of deposits made by accumulation of molten metal droplets is also of interest in other applications than thermal spray coating. In microcasting 345, complete objects are built up by depositing droplets in a stipulated pattern. The metallurgical properties of parts made with this technique depend on the shape and temperature history of individual droplets 5. The transformation from fragmented deposits to circular “disk splats” as surface temperature rises is well documented 67891011 for a wide range of metals and ceramics. Some researchers 9 have conjectured that freezing around the edges of an impinging droplet makes it splash: liquid flowing out from the center of a drop jets upwards when it hits a solidified rim. Delaying solidification, either by raising surface temperature or increasing thermal contact resistance at the droplet-substrate interface, is expected to suppress splashing. Others 1011 contend that molten droplets superheat volatile compounds on the surface, which evaporate so explosively that they shatter the drops. Preheating the surface removes these contaminants. Since these two mechanisms are not mutually exclusive, it is quite possible that both influence splat shape. Several studies have studied splashing of liquid drops where solidification did not occur. Much of the literature has been reviewed by Rein 12. Photographs have shown that fluid instabilities around the edges of the spreading liquid film lead to the development of fingers that detach to form satellites drops 13. Increased surface roughness is known to enhance fluid instabilities and promote splashing 14. The solidified layer appears to act in much the same way, creating obstacles under the drop that perturb liquid flow. Part of the difficulty in identifying the mechanism responsible for droplet splashing is that no one has directly observed impact of thermal spray particles: we have to infer details of impact dynamics from splat photographs such as those in Fig. 1. Several papers (e.g., 1516) have presented photographs of molten metal droplets impacting on a surface, but these have typically been of large (2–3 mm diameter) droplets falling under their own weight onto a surface with relatively low impact velocities (1–4 m/s). Weber numbers We=ρV02D0/σ, which measure the ratio of droplet kinetic energy to surface energy, were typically less than 103 and Reynolds numbers Re=ρV0D0/μ, which measure the ratio of inertial forces to viscous forces were less than 104. The molten part of impacting drops did not have enough momentum to surmount a solidified rim so that freezing around the edges of droplets suppressed splashing 16. Molten droplets in a wire arc spray typically hit a surface with both Re and We∼104. Whereas Re is the same order of magnitude as in previous single droplet impact experiments 1516, We is much larger. The mechanism triggering splashing under thermal spray conditions may be quite different from that observed at lower We. Computer simulations 17 support the hypothesis that outward flowing liquid is diverted upwards when it hits a solid layer. Fluid instabilities force the jetting liquid sheet to disintegrate. The objective of our study was to photograph impact of molten metal droplets at high Weber and Reynolds numbers. We used a combination of experiments and numerical simulation to determine the effect of varying both impact velocity and substrate temperature on droplet dynamics. It is difficult to accelerate a droplet to high enough impact velocities to obtain Weber numbers typical of thermal spray processes. We chose to accelerate the test surface by mounting it on the rim of a flywheel rotating in the horizontal plane. Droplets produced by a generator mounted above the edge of the flywheel fell vertically under their own weight and were hit by the surface moving horizontally at high speed. In this study we photographed splashing of molten tin droplets on a stainless steel surface. Droplet diameter (0.6 mm) and surface roughness (0.03 μm) were kept constant while impact velocity (10–40 m/s) and substrate temperature (80–260°C) were varied. Weber numbers ranged from 8.0×102 to 1.3×104 and Reynolds number from 2.3×104 to 9.2×104. A three-dimensional model that simulates fluid flow and heat transfer, developed earlier by Pasandideh-Fard et al. 18 and Bussmann et al. 19, was used to simulate droplet impact and solidification and obtain insight into the cause of splashing. A molten metal droplet generator 20 was used to produce uniform sized molten tin droplets. Figure 2 shows a schematic of the system. The main body of the generator was made of stainless steel. A band heater (Model HBA-202040, Omega Company, Stamford, Connecticut) was used to maintain the chamber temperature above the melting point of tin. A temperature controller (Model CN9000A, Omega Company, Stamford, Connecticut) and a thermocouple were connected to the heater to monitor the conditions. The chamber was filled with tin shot (99 percent pure, Aldrich Chemical Company, Milwaukee, Wisconsin). A commercially available synthetic sapphire nozzle with a 178 μm orifice was inserted into a hole drilled through the other wall of the chamber and sealed in place with Teflon tape. Teflon O-rings sealed both bottom and top of the chamber. A T-fitting was connected to the top plate, so that one outlet acted as a vent while the other was connected to a nitrogen tank whose outlet pressure was varied from 125–200 KPa. A solenoid valve (Model 8262G202, Convalve Company, Toronto, Ontario) was placed between the nitrogen tank and chamber. When the solenoid valve opened briefly (8–15 ms) a pressure pulse was sent to the chamber forcing a molten tin droplet out through the nozzle. The pressure in the chamber was then relieved by gas escaping through the vent hole, preventing more droplets from being ejected. By this method single droplets could be produced on demand by sending a signal to the circuit controlling the solenoid valve. To prevent oxidation of tin droplets emerging from the droplet generator an aluminum pipe with an inner diameter of 10 mm was attached to the bottom of the chamber. Nitrogen was injected into this pipe to shield droplets from the atmospheric oxygen. The volume flow rate of nitrogen was adjusted until it was just enough to prevent oxidation. The effect of oxidation on droplet formation was immediately obvious since it produced non-spherical, teardrop shaped droplets. The test surfaces on which molten metal droplets impinged were stainless steel coupons (38.1 mm long, 25.4 mm wide and 0.51 mm thick) polished on a metallurgical wheel to a mirror finish with average surface roughness 0.05 μm. Test coupons were mounted on a 9.5 mm thick aluminum plate with the same length and width as the coupons, which was bolted to the outer rim of a 406.4 mm diameter aluminum flywheel. Another identical plate was attached to the opposite side of the flywheel to act as a counterweight. A vertical rod inserted through the hub of the flywheel was connected to the shaft of a variable speed DC motor (model MS 3130-04/T, Dynetic Systems, Elk River, MN) through a flexible coupling (see Fig. 3). By varying the voltage applied to the motor, rotational speeds of up to 3500 rpm were obtained, giving the test surface linear velocities of up to 80 m/s. The whole system was mounted on a vibration isolation table. A slip ring (Model S4, Michigan Scientific, Charlevoix, MI) was mounted on the upper end of the flywheel shaft and used to carry electric power to the test surface and thermocouple signals from it. The stationary part of the slip ring was fixed to a support frame. To heat the substrate two 120 W cartridge heaters (Omega Engineering Co., Stamford, CT) were inserted into holes in the aluminum plate backing the test coupons. A Chromel-Alumel thermocouple was inserted into the center of the plate with its tip touching the stainless steel coupon. The rear surface and sides of the aluminum plate were insulated to minimize heat losses due to convection to the air and conduction to the flywheel. By varying the voltage applied to the heaters the temperature of the test surface could be controlled. The temperature of the surface was allowed to reach steady state while rotating before depositing drops on it. Spatial temperature variations from one side of the test surface to the other when it was moving were less than 5°C. A CCD video camera (Sensicam, Optikon Corporation Ltd., Kitchener, Ontario) was used to photograph droplet dynamics during impact. It had an intensified CCD chip capable of recording 30 frames per second with a resolution of 1280×1024 pixels. The camera could also superimpose up to ten images in every frame, each with an exposure time as short as 0.1 μs, separated by delays that varied from 0 to 1 ms (selectable in 0.1 μs time steps). A 0.1 μs exposure time was short enough to capture the deformation of the droplet during the impact, without any blurring caused by the extremely fast motion of the substrate. To hit a falling droplet with the moving substrate, and to photograph its impact, three events had to be synchronized with the position of the arm: ejection of a droplet, triggering of the camera, and triggering of a flash to provide illumination. An optical sensor was used to pick up the signal caused by the flywheel rotation. This signal was then used to trigger the camera, flash, and droplet generator. Since the frequency of this signal was too high to directly drive the droplet generator, it first passed through a frequency divider, which reduced the frequency by a factor that varied from 2 to 32, depending on the rotational speed of the arm. The low frequency signal formed one input of an AND gate (see Fig. 2). When we were ready to take a photograph we pressed a switch which activated the second input of the AND gate, so that the pulses at the other input were transmitted to a time delay unit. The rising edge of each pulse provided a reference we used to time all other events. The digital time delay generator (Model DG 535, Stanford Research Systems, Sunnyvale, California) controlled the timing of three subsequent actions with pico-second resolution. We made droplets collide with the substrate by varying the delay between the reference pulse and triggering of the droplet generator. Each droplet was ejected from the generator and fell to a position coincident with the center of the test surface just as the arm approached the droplet. The droplet velocity as it exited the generator was less than 1 m/s, small enough to reasonably assume impact was normal to the surface even for the lowest velocities in our experiments, 10 m/s. Tin droplets did not adhere to the stainless steel surface after impact but were thrown off by centrifugal forces. The flash (Model MVS 7000, EG&G Corp., Salem, Massachusetts) timing was adjusted so that droplet impact was illuminated by a 10 μs long burst of light. While the flash was on, the camera was activated to take a single 0.1 μs exposure of an impacting droplet. By varying the time at which the camera was triggered, different stages of droplet impact were recorded, and the entire process of droplet impact reconstructed from a sequence of such pictures. This single-shot method was used instead of a high-speed camera because it provided high-resolution photographs in which satellite droplets were clearly visible. The entire process of droplet impact and spreading takes between 100 to 200 μs. The repeatability of photographs was only about ±20 μs, which was approximately the interval between successive frames in a sequence of photographs. Therefore, it was not possible to assign an exact time to each frame. This was not a serious shortcoming, as we did not make any time resolved measurements from photographs. They were used only to observe the shape of droplets as they splashed. In our experiments we accelerated the substrate to achieve high velocity impact. This technique did not exactly replicate a spray coating application, where droplets are accelerated by a high velocity gas flow before they impact on the surface. Nevertheless, we do not expect the gas velocity to have any significant effect on droplet shape or impact dynamics, because the viscous shear force exerted by the surrounding gas on a droplet is much less than the surface tension force that keeps it spherical. The capillary number Ca=μV0/σ gives the relative magnitudes of viscous and surface tension forces. For a tin droplet traveling with a velocity V0=100 m/s we estimated Ca∼10−3, showing that the shear exerted by the surrounding gas is relatively small even at high velocities. A droplet colliding with a rotating surface is acted upon by both centrifugal and Coriolis forces. An order-of-magnitude analysis 21 of the Navier-Stokes equations in which they appear as additional body forces, showed that for D0/Rf≪1 (where Rf is the distance between the center of the droplet and center of rotation) both centrifugal and Coriolis forces may be neglected, which was the case in our experiments. Figure 4 illustrates the spreading of a 0.6 mm molten tin droplet impacting with a velocity of 10 m/s on a stainless steel surface at a temperature of 20°C, photographed with a camera inclined at an angle of 45° to the substrate. The frames are labeled a to e, representing successive stages of impact. Triangular protrusions formed around the periphery of the droplet (Fig. 4(b)) as it spread. These protrusions then detached from the rim forming tin fragments that flew outwards (Fig. 4(c)). A roughly circular, solid splat was left on the surface (Fig. 4(e)). The pattern of splashing was quite different from that observed at lower impact velocities, where finger shaped perturbations formed and detached forming satellite droplets around the edge 16. Suppressing solidification—by increasing substrate temperature above the melting point of tin (232°C)—prevented droplet splashing at an impact velocity of 10 m/s. Figure 5 shows a 0.6 mm molten tin droplet landing with a velocity of 10 m/s on a stainless steel surface. The experimental conditions were exactly the same as those in Fig. 4, except that the substrate temperature was raised to 240°C. The droplet smoothly spread out into a thin disk (Fig. 5(d)) with small fingers forming at regular intervals around its edge. Fingers are created due to Rayleigh-Taylor instability, which occurs when the surface of a spreading droplet undergoes large deceleration 2223. Surface tension pulled back the molten tin (Fig. 5(e)), so that there was almost no splashing, with only the tips of a few fingers detaching. The final position of the liquid was asymmetrical due to centrifugal forces that drew the liquid to one side. Increasing impact velocity enhanced splashing. Figure 6 shows the impact of a 0.6 mm molten tin droplet with a velocity of 40 m/s on a surface at 20°C. At this high impact velocity splashing was much more pronounced than it was at 10 m/s (compare with Fig. 4), producing a cloud of debris ahead of the spreading rim. Violent fragmentation of droplets occurred immediately after impact (Fig. 6(b)). Large sections of the droplet flew off the surface (Fig. 6(d)) instead of the small fragments seen in Fig. 4. A small circular splat remained on the surface (Fig. 6(e)). At high impact velocities droplets splashed even on a hot substrate. When a droplet hit a 240°C surface with a velocity of 40 m/s (Fig. 7) there was some break-up around its periphery (Fig. 7(b–d)), though less than that observed during impact on a cold surface (compare with Fig. 6). There was no discernible formation of fingers and no recoil of the droplet, unlike that seen at lower impact velocity, and a roughly circular splat with irregular edges (Fig. 7(e)) remained on the substrate. To model solidification and splashing of molten metal droplets we used a three-dimensional model of droplet impact and solidification developed by Pasandideh-Fard et al. and Bussmann et al. 1819. The model has been discussed in detail in previous publications 17181924, so it is described only very briefly here. The model solves the mass, momentum and energy conservation equations, discretized using a finite volume technique on a three-dimensional, Eulerian structured grid. Fluid flow was assumed to be Newtonian, laminar and incompressible. The Volume-of-Fluid (VOF) algorithm was used to track the free surface of the droplet. Surface tension was incorporated as a component of the body force acting on the fluid free surface using the Continuum Surface Force (CSF) model 1925. An adiabatic boundary condition was applied at the droplet free surface. The model requires a value for thermal contact resistance Rc at the droplet-substrate interface, that we had no way of measuring in these experiments. Previous measurements have shown that contact resistance under impinging droplets decreases as impact speed is raised, varying from 5×10−6 °C/W for tin droplets impacting with a velocity of 1 m/s to 10−6 °C/W at 4 m/s 16. Lacking experimental data at higher velocities we tried using Rc=5×10−7 °C/W and found that predicted droplet shapes agreed with photographs; all simulations shown in this paper were done using this value. Advancing and receding contact angles at the liquid-solid contact lines were set to 140 deg and 40 deg, respectively, based on measurements by Aziz and Chandra 16, who found that contact angles did not depend on impact velocity. Only a quarter segment of each droplet was simulated and the entire droplet reconstructed by reflecting the results about planes of symmetry. The computational domain was a cube with sides 6 times droplet radius and height 3 times radius. Based on a mesh refinement study 1824, the mesh size was chosen to be 1/22 of the droplet radius. Previous experience 15171824, confirmed here, showed that predicted droplet shapes do not change significantly when the computational grid is made smaller. However, once a droplet splashes and small satellite droplets detach, some of them are of the same size as the mesh spacing. Our simulations, therefore, do not give reliable information about the shapes of these secondary droplets and are meant only to model the deformation of the main droplet. Numerical computations were performed on an AMD Athlon 1.4 GHz PC and the maximum CPU time was 24 hours. Properties of tin used in calculations are given in Table 126. Figure 8 shows simulated images of 0.6 mm tin droplets, initially at 233°C, impinging with an impact velocity of 40 m/s onto stainless steel surfaces with initial temperatures (a) 80°, (b) 180°, and (c) 260°C, respectively. Each column shows a droplet during successive stages of impact on a surface at a given initial temperature. To show clearly the extent of solidification in the computer generated images the liquid was made transparent, with a light gray color. Solid layers were assigned a darker shade of gray. Substrate temperature distributions are given in Fig. 9 for substrate temperatures of (a) 80°C and (b) 180°C, at the same times as those in Fig. 8. Temperature distributions were not given for a substrate at 260°C, where impact was almost isothermal. In Fig. 9 liquid portions of droplets are shown in white and solid in black. During impact on a substrate at 80°C (Fig. 8(a)), solidification started first along the edges of the spreading drop t=22 μs, where it first contacted the colder substrate. The temperature distribution in the substrate was essentially one-dimensional, with the substrate a very short distance beyond the edges of the drop at its original temperature of 80°C (see Fig. 9(a) t=22 μs). The highest temperature gradient, and consequently the highest heat flux from the droplet, was along its edge. The band of solid material grew wider (Fig. 8(a), t=34 μs) and freezing also started at the center of the drop, which had been longest in contact with the substrate. Liquid jetted over the solidified layer along the droplet rim and broke up into droplets (Fig. 9(a), t=34 μs). By t=46 μs the bottom of the droplet was entirely solidified, but a film of liquid remained on top of it and continued to flow outwards, forming fingers of liquid around the periphery of the drop that detached and broke-up to form satellite droplets. At a higher substrate temperature, 180°C, solidification was delayed until t=34 μs and was visible only around the fringes of the droplet. In this case too the lowest surface temperature was just beyond the edge of the drop (Fig. 9(b)). The droplet spread to a greater extent than it did on the surface at 80°C. There was clearly less splashing in this case because there was no solidified outer rim to perturb the liquid flow. By the time there was significant solidification the droplet has already spread to its maximum extent t=62 μs and liquid velocities were very low. When substrate temperature was raised above the melting point of tin (Fig. 8(c)) no solidification occurred and the droplet spread to form a smooth disk. Rings of molten tin detached along the rim of the droplet, leaving a disk of molten metal in the center. How well do our numerical simulations capture the mechanism of droplet splashing? Figure 10 shows comparisons between photographs of splats at their maximum spread and corresponding simulations at temperatures of temperatures (a) 80°, (b) 150°, and (c) 260°C. At the lowest temperature, 80°C, thin radial fingers projecting from the splat are visible in both the photograph and simulated image. Bussmann et al. 24 simulated fingering around the edges of liquid droplets—without solidification—by artificially perturbing the velocity field after impact to initiate the growth of fingers. We did not do that in this case, since solidification around the droplet edges was enough to trigger an instability that produced fingering. Raising the surface temperature to 150°C reduced the number of fingers, leaving an irregular shaped splat (Fig. 10(b)). Above the melting point, at a substrate temperature of 260°C, the splat was almost perfectly circular, though material that had detached was visible around the splat. Explosive vaporization of contaminants on the surface, principally adsorbed water, has been proposed as one possible mechanism that causes splashing of impacting drops 1011. Though this mechanism is possible in thermal spray applications where droplets of very high melting point materials impact the substrate, it is unlikely to be significant in our experiments with tin droplets. Simulations of droplet impact on a surface initially at 20°C showed that substrate temperature always remained below 100°C, precluding rapid vaporization of water. Previous researchers 6789 examining thermal spray splats have identified a “transition temperature,” at which droplets stop splashing and form circular splats. There is a certain subjectivity in identifying this temperature, since the transition to circular splats is gradual: we identified a transition temperature range of approximately 160–180°C, above which splashing greatly diminished. We designed and built an experimental apparatus to photograph high-speed impact of small molten tin droplets on surfaces of varying temperature. We held droplet diameter and surface roughness (0.6 mm, 0.05 μm) constant while varying surface temperature and impact velocity (120–260°C, 10–40 m/s). On a cold surface (20°C) fragments from the periphery of the droplet broke off and flew off the surface during impact. At a high surface temperature (240°C) there was much less splashing and the droplet formed a roughly circular splat. The transition from splashing to circular splats took place over a range of temperature between 160–180°C. A three-dimensional model of droplet impact and solidification was used to simulate droplet splashing. Numerical results agreed qualitatively with photographs of droplet splashing. Simulations showed that freezing around the edges of droplets creates instability and triggers splashing. D0= droplet diameter Rf= radius of flywheel V0= droplet impact velocity Greek Symbolsσ = droplet surface tension ρ = droplet density μ = droplet viscosity ν = droplet kinematic viscosity Dimensionless NumbersRe = Reynolds number =ρV0D0/μWe = Weber number =ρV02D0/σCa = Capillary number =μV0/σ