You are on page 1of 22

LIQUID METAL ATOMIZATION USING 15, pp. xxyy, 2005 NOZZLES, 1 Atomization and Sprays, vol.

CLOSE-COUPLED

[[AU: NOTE THAT EQUATION NUMBERS HAVE BEEN CORRECTED. PLEASE CHECK ALL REFS TO EQ. NOS.]]

A STUDY OF LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, PART 1: GAS DYNAMIC BEHAVIOR S. P. Mates
Metallurgy Division, National Institute of Standards and Technology, Gaithersburg, Maryland, USA

G. S. Settles
Gas Dynamics Lab, Department of Mechanical and Nuclear Engineering, Penn State University[[AU: ADD City, State, USA]]
Original Manuscript Submitted: 2/26/03; Final Draft Received: 8/8/03 Liquid metal atomization using close-coupled nozzles is an established technique for fabricating fine (< 100-m) metal powders for a variety of industrial uses. Despite its widespread use, however, the interrelationships among gas dynamics, nozzle geometry, processing parameters, and particle size remain ill-defined. As a result, efforts to reduce powder costs by improving particle size control and energy efficiency remain hindered. This study examines and compares examples of a convergent and a converging-diverging (c-d) close-coupled nozzle on the basis of their gas dynamic behavior (Part 1) and their liquid metal atomization performance (Part 2). In Part 1, Schlieren photography and Mach number and Pitot pressure measurements are used to characterize the gas dynamic behavior of the nozzles (without liquid metal present) operating at stagnation pressures between 2 and 5 MPa. In Part 2, their liquid metal atomization behavior is examined by high-speed Schlieren photography, and particle size distributions are measured to compare their atomization performance. Results showed that the two nozzles performed similarly in gas flow and atomization tests over most of the range of Po examined, despite their significantly different geometries. The active atomization zone appeared to extend far downstream, indicating that gas velocity decay by turbulent diffusion may play a limiting role in atomization. This also suggests that the importance of the gas-to-liquid mass flux ratio has a physical basis associated with a ratio of velocity decay length to breakup length scales. These observations have potentially important implications for designing efficient liquid metal atomization processes for producing low-cost metal powders.

INTRODUCTION Liquid metal atomization using close-coupled nozzle technology is an established method of fabricating fine metal powders for use in powder metallurgy (P/M) manufacturing as well as a variety of other industrial uses [15]. A typical gas atomizer configuration is sketched in Fig. 1a. In a close-coupled nozzle arrangement, molten metal is fed through a central tube that is surrounded by a coaxial gas nozzle. The nozzle generates a high-velocity gas stream that disintegrates the slower melt stream into fine droplets, which then freeze in-flight into solid powder particles. Liquid metal atomization is usually conducted inside
The authors wish to thank J. D. Miller and Lori Dodson-Dreibelbis of the Penn State Gas Dynamics Lab, and Dr. Ali nal and Dr. Dennis McLaughlin for their valuable input. This work was supported by the National Science Foundation (CTS-9221863).

Copyright 2005 by Begell House, Inc.

S. P. MATES AND G. S. SETTLES

NOMENCLATURE
A* d L m M P q T critical throat area, m2 diameter (nozzle, droplet), m length, m mass flow, kg/s Mach number pressure, kPa dynamic pressure, kPa temperature, K specific heat ratio viscosity, N s/m2 density, kg/m3 surface tension, n/m; std. dev.

Subscripts
a e o s ambient exit stagnation sonic point

Fig. 1 Typical arrangement of a liquid metal atomizer (a) employing a close-coupled nozzle to produce fine metal powder (based on [3]). (b) Typical atomized metal powder.

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

an enclosed chamber to both confine the powder particles and, in some cases, to limit particle oxidation. The particles are typically spherical in shape (Fig. 1b) and range from about 1 to 250 m. Metals atomized commercially range from tin and tinlead alloys to aluminum, steel, and nickel-based alloys. Gases used usually include air, nitrogen, or argon. Industrial-scale atomizers can produce tens of kilograms of powder per minute, while research-scale units generally produce significantly less. Gas-atomized powders are used as rocket propellant additives, chemical catalysts, feedstock for P/M consolidation and thermal spray coatings, and in many other capacities. Gas atomization is a relatively costly method for producing metal powders. Contributing to its high cost are inefficiencies arising from poor energy usage and lack of particle size control. Improvements in these areas would have a significant impact on a wide range of metal powder industries. Unfortunately, efforts to optimize the liquid metal atomization process have been hampered by an incomplete understanding of the underlying physics, particularly that involving supersonic gas flow. For example, the role of shock waves in the atomizing gas flow has often been debated but remains unclear. Further, no reliable guidelines exist for optimizing nozzle geometry to achieve efficient atomization. For example, it remains unclear whether the improved supersonic expansion provided by converging-diverging (c-d) nozzles necessarily leads to better atomization performance (smaller particle sizes) compared to convergent nozzles. Finally, although it is known that increasing gas pressure generally reduces particle size, the physical role of gas pressure remains poorly understood. For example, increasing gas pressure increases both the fully expanded gas velocity as well as the gas mass flux, either of which may contribute to reducing particle size. Is one or the other the dominant effect, or do they play equal roles? These lingering questions limit efforts to optimize the liquid metal atomization process. In many respects this process is similar to the more often studied twin-fluid fuel atomization processes [6]. Although liquid metal solidifies readily in the cold atomizing gas stream, the solidification step generally follows the atomization step [2], so that atomization behavior is determined primarily by liquid instability rather than phase change. However, the unique properties of liquid metals give rise to some important differences. Because the high surface tension of liquid metals makes them difficult to atomize, supersonic gas velocities are usually employed. Further, because liquid metal can easily freeze inside the nozzle and interrupt atomization, heat transfer is a critical issue. The nozzle must not only thermally insulate the melt, it must be chemically inert toward the melt to prevent powder contamination and/or its own erosion. The use of supersonic gas velocities and the importance of heat transfer and chemical reactivity in the nozzle make this a unique twin-fluid atomization process. Optimizing a liquid metal atomization process for increased powder affordability generally involves maximizing the yield of powder particles within some desired size range (usually narrow) while consuming the least possible amount of energy in terms of compressed gas. Typical uses for metal powders call for narrow or specific particle size ranges that are usually some subset of the raw powder size distribution obtained from the atomizer. For example, metal injection molding (MIM) requires feedstock powders below 20 m, while in thermal spray applications the powder must typically be larger than 20 m and less than about 50 m. Thus only a fraction of the powder produced meets the size requirements of any one application. Alternative uses must be found for the remaining powder, or else it must be remelted and reatomized, adding cost. The ability to tune an atomizer for particular sizes would therefore reduce the overall powder cost by minimizing

S. P. MATES AND G. S. SETTLES

off-size material. Energy usage in liquid metal atomization is often put in terms of the amount of gas consumed per quantity of powder produced. Thus the gas-to-metal mass flow ratio (G:M), in addition to being an important physical parameter, is a key indicator of process economy. Typical values range between 2 and 10 for industrial atomizers. Lower values are associated with less costly production, while higher values usually result in finer powder, creating a trade-off between particle size and powder cost. Another desirable goal is to increase the yield of the finest particles, typically below 10 mm. Fine powder is more expensive because it represents a relatively small fraction of the overall yield. Improving particle size control and increasing fine powder yield on industrial scales are thus the most critical research goals. Research on liquid metal atomization involving close-coupled nozzles goes back decades. The earliest scientific work known to the authors in this field was that of Thompson [7], who concentrated on aluminum. Thompson investigated the effect of the major process variables, such as gas pressure, metal flow rate, and metal temperature, on particle size using various nozzle arrangements. More recently, interest in rapid solidification processing in the 1980s prompted an expanded interest in liquid metal atomization. Rapidly solidified metal droplets produced by atomization were found to possess glassy (noncrystalline) microstructures with unique and desirable properties [8]. Under the rapid solidification research initiative, an array of sophisticated experimental tools was applied to examine the atomization of a wide range of metals and alloys. Schlieren optical flow visualization was applied to study the supersonic gas flow patterns produced by closecoupled nozzles when operating gas-only [912] and when atomizing water as a surrogate liquid [13]. High-speed cinematography [9, 10] and holography [14] were used to visualize melt breakup phenomena. Particle size analysis by laser-diffraction, including insitu measurements [15], have been employed to measure fine powder yield at various processing conditions and with different nozzle designs. Computational fluid dynamics has been applied to model gas-only flow [16, 17] as well as liquid metal atomization itself [18]. Both convergent [911] and converging-diverging (c-d) [12] close-coupled nozzle concepts, employing either annular or discrete-jet arrangements, have been evaluated in an attempt to find optimum performers. The effects of gas species [12] and gas temperature [19] on particle size have also been studied. While these efforts have led to improvements in atomizer performance and enhanced process understanding, substantial gaps in knowledge, particularly regarding gas dynamics topics such as the role of gas pressure, shock waves, and nozzle expansion conditions, remain. In addition to these gas dynamics issues, a further uncertainty involves the primary breakup mechanism that has traditionally been associated with close-coupled liquid metal atomization. Primary breakup is believed to involve the formation of a thin melt sheet at the nozzle tip [1, 912], shown schematically in Fig. 2a. Upward-flowing gas along the central axis, along with radial pressure gradients acting across the base of the liquid delivery tube, are believed to create the melt sheet by forcing the liquid metal out to the nozzle edge and into the supersonic cross flow. It is generally believed that fine powder is created by the subsequent disintegration of this melt sheet by a wave instability mechanism akin to the case studied by Dombrowski and Johns [20]. Despite widespread adherence to this theory, however, many observations suggest that the melt sheet does not always form. While sheet formation is observed when the liquid metal flow rate is low [21], as the flow increases, it becomes far less certain [21, 22]. In Thompsons original experiments, sheet formation did not occur at all. Instead, the liquid metal formed a miniature fountain at the nozzle tip,

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

Fig. 2 Melt sheet [1, 912] and fountain [7] primary breakup models.

with a short intact liquid core, shown schematically in Fig. 2b. Others have described a similar jetting effect under some conditions [2]. As will be shown in Part 2 of this article, the conditions examined in the present study do not result in a melt sheet. Sheet formation is therefore not the only possible atomization mechanism at work in this process. In fact, industrial-scale atomizers, which operate with large melt flow rates, are significantly less likely to produce melt sheets than the research-scale units in which sheeting has been observed. Despite this possibility, few attempts have been made to analyze atomization under non-sheet-forming conditions. Investigating this regime further will expand the overall understanding of this atomization process and may suggest new ways to approach improving particle size control and improving efficiency. For example, our initial observations of non-sheet-forming behavior indicated that fine powder appears to be created mostly by the secondary breakup of large globules that are produced by a relatively coarse primary breakup stage. Further, it was not uncommon for secondary breakup to persist well downstream of the region where sheet breakup was previously thought to occur [23]. Thus our observations indicate that, under non-sheet-forming conditions, the important powder-forming breakup mechanisms appear to be quite different. In fact, even under conditions favoring melt sheet formation, one investigator suggested that secondary breakup is important and that atomization is not necessarily limited to the nozzle tip region. nal, who investigated aluminum atomization, estimated that only about one-quarter of the powder was created directly by the initial (primary) disintegration of the sheet [12]. The rest, he hypothesized, was created by the secondary breakup of coarse daughter droplets resulting from primary sheet breakup. Further, nal found that nozzles producing longer supersonic jets generally also produced finer powder [13]. With a longer supersonic jet, he argued, elevated gas velocities are maintained farther downstream, leading to more effective secondary breakup far from the nozzle tip. Despite these obser-

S. P. MATES AND G. S. SETTLES

vations, most previous efforts to improve atomization have focused on enhancing sheet breakup, while little attention has been paid to influencing secondary breakup. nals observations not only show that secondary breakup may be important, they also suggest a novel connection between the gas dynamic behavior of supersonic nozzles and atomization performance. That is, a relationship appears to exist between the supersonic length of the gas jet produced by the nozzle and the outcome of secondary breakup events occurring far downstream from the nozzle tip. Understanding this connection may shed significant light on the fundamental behavior of the liquid metal atomization process, from which important insights on optimization may be gained. For example, among rocket nozzles it is known that supersonic jet length increases significantly as the nozzle stagnation pressure increases [24]. However, no prior study has examined the effect of stagnation pressure on supersonic jet length among liquid metal atomization nozzles, or on how this effect might influence atomization performance. Another implication is that convergingdiverging nozzles ought to produce finer powder than convergent nozzles, since the former produce longer supersonic jets, at least for round (rather than annular) nozzle geometries [2527]. nal reported that c-d nozzles were generally superior to convergent nozzles among basic configurations he studied [28]. However, others claim excellent performance using convergent nozzles, raising doubts about the universal benefit of c-d nozzles [2]. An examination of the effect of stagnation pressure on supersonic jet length behavior among close-coupled nozzles, including comparisons between successful examples of convergent and c-d designs, would therefore prove revealing. In the present study, the gas dynamic behavior and atomization characteristics of a convergent and a converging-diverging close-coupled nozzle are examined and compared. A laboratory-scale atomizer is constructed with the ability to atomize small quantities of molten tin. The two nozzles chosen for this study are based on successful designs described in prior literature as having an ability to atomize liquid metal effectively without suffering from excessive melt freeze-off problems. In Part 1 of this article we examine the gas dynamic behavior of these nozzles over a range of stagnation pressures. Schlieren optical flow visualization and Mach number and Pitot pressure measurements are used to characterize the mean flow behavior of the two nozzles both in the neighborhood of the nozzle tip and farther downstream. In Part 2 we examine the nozzles liquid metal atomization behavior under conditions that do not favor melt sheet formation. Breakup phenomena taking place in the compressible jet flow are observed using microsecond-exposure Schlieren photography. Particle size analysis is then performed to quantify and compare the two nozzles under equivalent atomizing conditions and to examine how nozzle stagnation pressure affects particle size. The observations and data are synthesized into an overall physical picture of the atomization process, including both near-field and far-field effects, to supplement the current understanding. EXPERIMENTAL DESIGN The two close-coupled nozzles chosen for this study are shown in Fig. 3. They include an annular convergent nozzle, based on a discrete-jet design described in [9], and a converging-diverging nozzle based on [12]. The nozzle dimensions were chosen to best emulate the optimum designs identified in the cited references while maintaining some dimensional uniformity (e.g., nozzle tip diameter) to facilitate comparisons. It must be noted that,

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

Fig. 3 Close-coupled metal atomization nozzles investigated. Linear dimensions are in mm.

in addition to differing in gas channel design (e.g., convergent versus converging-diverging), these nozzles differ in other characteristics such as nozzle tip protrusion and taper angle (Fig. 2), owing to their separate heritages. Previous studies of such nozzles have indicated that atomization performance can be sensitive to small changes about these optimum configurations [10, 28]. However, we did not elect to investigate these variants here and instead focus on the difference in their degree of exit flow expansion (convergent versus c-d design) when comparing their behavior. The nominal sonic throat area of the c-d nozzle was somewhat larger than that of the convergent nozzle (13 mm2 to 9 mm2), causing a difference in mass flux at any given stagnation pressure. This difference was accounted for by normalizing the results by the measured mass fluxes when appropriate. The isentropic design pressure ratio of the cd nozzle is 35 based on its exit-to-throat area ratio of 4:1. In this study, the c-d nozzle is often operated at off-design conditions in order to study the effect of stagnation pressure on its gas dynamic and atomization behavior. The term design is misleading in the present context, however, since it pertains specifically to the use of c-d nozzles in thrustproducing devices and not in liquid atomizers. It is well known that operating a c-d nozzle at its design pressure ratio produces maximum thrust, while operating under off-design conditions diminishes thrust [29]. However, the thrust performance of a nozzle depends only on the net pressure force acting on its walls. Thrust is unaffected by the gas velocity field existing beyond the nozzle exit. However, the atomization performance of a c-d nozzle depends only on the characteristics of the external gas velocity field that drives the atomization event. While the external velocity field is certainly influenced by off-design operation (shock waves are introduced, for example), it is not clear that atomization performance should necessarily suffer as a result. Thus, operating at design conditions does not have the same implications for atomizers as it does for thrust-producing devices. Further, the present c-d nozzle will never behave like an ideal converging-diverging nozzle even when operating at its design pressure ratio, since shock waves will be formed because of the conical diverging section and the presence of the central liquid delivery tube.

S. P. MATES AND G. S. SETTLES

The atomizer facility is shown in Fig. 4. Compressed air is supplied to the atomizer assembly from a 454-L reservoir that is maintained at 7 MPa. The stagnation pressure in the atomizer assembly is controlled via a two-stage regulator, and onoff flow control is provided by a ball valve located upstream of the regulator. The system allowed a maximum stagnation pressure of approximately 5.5 MPa. Exhaust air from the nozzle is contained by directing it into a 200-L enclosed plastic container through a 15-cm inlet duct protruding through the containers cover. Excess air exits the container through a side-mounted exhaust duct. When metal is atomized, the container is partially filled with water to scrub the atomized particles from the exhaust air. Further details on powder handling are discussed in Part 2. Air mass flow is measured with a Venturi flow meter selected to minimize stagnation pressure losses [30]. The expanded uncertainty of the air mass flow measurements was determined to be 7%. This uncertainty is larger than the nominal 1% usually assigned to flow meters of this type because of the small differential pressures achieved across the meter during operation, which resulted in a relatively low signal-tonoise ratio. Pressure measurements for the air flow meter and in the nozzle stagnation chamber were made using diaphragm-type pressure transducers. Gas and liquid metal temperatures were measured using K-type thermocouples. Voltage data from the pressure transducers and thermocouples were acquired at 10 Hz and averaged over several seconds to obtain steady-state measurements. The atomizer assembly, also shown in Fig. 4, is made from a short section of schedule 80 pipe (76.2 mm i.d.) fitted with extra-heavy pipe flanges. The plenum is sealed by capping the flanges with 9.25-mm-thick aluminum plates. Holes cut in each cap plate allow the nozzle centerbody to protrude through the plenum at each end to allow liquid metal delivery to the nozzle tip. The liquid reservoir sits atop the centerbody and connects to a quartz capillary tube that is inserted into the centerbody to insulate the molten metal and protect the centerbody from being slowly dissolved. The nozzle shape is formed by the shape of the centerbody tip and the shape of the orifice in the bottom cap plate through which the centerbody tip protrudes (see Fig. 3). Three set screws located upstream of the

Fig. 4 Laboratory atomizer facility. Linear dimensions are in mm.

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

nozzle throat were used to carefully align the centerbody tip inside the bottom plate orifice in order to obtain a symmetric gas flow pattern. During gas-only experiments, the liquid delivery tube is sealed to prevent excess air from being aspirated through the orifice. A single-pass lens-type Schlieren system [31] (Fig. 5) was used for visualizing the compressible flow patterns generated by the two nozzles. Images were captured with a CCD video camera and recorded on S-VHS videotape, after which still frames were digitized with a PC-based frame grabber and subsequently processed for clarity. Long exposures were obtained using a halogen light source, yielding an effective exposure time of 33 ms. Microsecond-exposure images were obtained using a strobed xenon-arc lamp light source that was synchronized to the CCD camera field acquisition rate of 60 Hz. Radial Pitot pressure surveys of the gas flow pattern in the neighborhood of the nozzle tip were conducted using a 0.9-mm-diameter steel Pitot tube mounted on a threeaxis stepper-motor-driven traverse. To minimize gas consumption, the probe was swept through the flow at constant speed while recording pressure data. Probe position was monitored using a variable-resistance position transducer. Axial Pitot pressure surveys conducted along the gas flow centerline were performed with a 1.5-mm-diameter steel Pitot tube mounted on a manually driven traverse. In the axial surveys, the probe was positioned at each measurement location before pressurizing the nozzle. Centerline Mach number measurements were made optically using the Schlieren system and a 10 halfwedge. Mach numbers were computed from the angle of the oblique shock wave formed at the wedge tip using two-dimensional oblique shock theory [32]. The expanded uncertainty in the Mach number data was estimated to be 0.2 Mach. Beyond the optical field of view, Mach numbers were estimated directly from the measured Pitot pressure levels using the Rayleigh-Pitot formula [33] and by assuming ambient static pressure (nominally 100 kPa). The sonic point was taken to be the axial location where the pitot pressure equaled 1.89 times the ambient pressure. RESULTS AND DISCUSSION Nozzle Tip Region Flow Structure Our initial investigation focused on the gas flow patterns produced by the convergent and c-d nozzles in the nozzle tip region, where primary breakup is thought to occur.

Fig. 5 Schlieren optical system.

10

S. P. MATES AND G. S. SETTLES

Differences in the nozzles gas dynamic behavior are anticipated primarily because of differences in their exit pressure ratios, although nozzle tip protrusion and taper angle also contribute. The exit pressure ratio is the ratio of the static pressure of the flow at the nozzle exit (Pe) to the surrounding ambient pressure (Pa). It determines the amount of expansion (or compression) the flow undergoes upon leaving the nozzle to adjust to the pressure of the surroundings. Table 1 compares the exit pressure ratios (Pe /Pa) of the two nozzles at low, medium, and high values of Po /Pa. These values, being computed assuming ideal isentropic one-dimensional flow, are only approximate. Table 1 shows that convergent nozzles exit pressure ratio is always well above unity, meaning its flow will expand significantly in response to the surrounding pressure in the exhaust environment (e.g., the laboratory). Because of this, the convergent nozzle is said to be highly underexpanded. By contrast, the c-d nozzle goes from being slightly overexpanded to being slightly underexpanded over the range tested, meaning the flow it generates undergoes only mild pressure adjustments in response to the surrounding pressure. Differences in the amount of expansion give rise to differences in the strength and pattern of the shock waves that develop in these flows. The higher exit pressure ratios achieved by the convergent nozzle mean that stronger expansion waves will form, which, by wave reflection, will subsequently give rise to stronger shock waves compared to the c-d nozzles flow. One objective of the present study is to determine whether the stronger shock waves expected in the convergent nozzles flow will affect its atomization performance relative to the c-d nozzle. Figure 6 describes the typical flow pattern produced by the convergent nozzle in the neighborhood of the liquid delivery tube. Because this flow pattern is very similar to that produced by plug nozzles studied in connection with rocket propulsion applications [34], many of the terms used here to label flow features are borrowed from this earlier work. Dominating the flow is the strong internal shock wave, which is created by reflections of the strong expansion waves formed at the nozzle exit. The internal shock wave originates close to the nozzle exit, then curves inward toward the flow centerline, finally terminating on an unseen boundary. This boundary is the embedded sonic surface surrounding an internal subsonic flow region called the wake. The wake is created as the supersonic flow exiting the nozzle separates from the outer rim of the liquid delivery tube, leaving a subsonic region of circulating flow, sometimes called a separation bubble. Because the wake is surrounded by a sonic surface, shock waves cannot cut across it. Instead, as in the case of the internal shock, they reflect off the wake boundary away from centerline as an expansion fan. Upon reaching the outer subsonic boundary surrounding the jet, the expansion fan reflects again as a compression fan, creating a new shock wave. Successive reflections of these waves off sonic surfaces give rise to the classic repeating shock cell structure known as shock diamonds.
Table 1 Convergent and c-d Nozzle Exit Pressure Ratios Calculated Using a One-Dimensional Isentropic Approximation Pe /Pa Po /Pa 14 28 48 Convergent nozzle 7.29 14.57 25.50 c-d nozzle 0.39 0.79 1.38

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

11

Fig. 6 Schlieren image of the convergent nozzles near-region flow field (left) in the open-wake configuration (Po /Pa = 45), a flow field sketch (right), and Pitot pressure (Pt2) data (top).

Inside the wake, the mean flow consists of a circulating pattern, as depicted qualitatively in Fig. 6. This circulation pattern is typical of supersonic base flows studied in rocket propulsion [35]. The end of the circulation zone is marked by a single stagnation point on the flow axis. Beyond this point the flow accelerates once again in the streamwise direction, eventually reaching supersonic velocities. The faint horizontal stripes visible in the Schlieren picture are caused by a streamwise vortex sheet present within in the annular supersonic flow region surrounding the wake. Streamwise vortices are commonly observed in underexpanded supersonic jets emanating from round nozzles [36]. Pitot pressure data plotted alongside the Schlieren image in Fig. 6 are used to further define the shape and extent of the wake region and to reveal the structure of the supersonic jet surrounding it. The subsonic region extends approximately two nozzle diameters from the tip of the liquid delivery tube. After six nozzle diameters, the initial doughnut Pitot profile begins to give way to a single-peaked profile. Natural turbulent diffusion processes and entrainment of surrounding air act to smooth and broaden the jet velocity profile as one proceeds downstream. Although velocity information cannot be extracted from the Pitot pressure measurements because the static pressure distribution is unknown, the Pitot data nevertheless show that velocities are generally much lower in the wake region compared to the supersonic jet surrounding it. Therefore liquid breakup will be much more intense outside of the wake region than inside it. In Fig. 7, a series of Schlieren pictures reveals how the convergent nozzle flow pattern varies with stagnation pressure. Of particular interest is that, when the overall pressure ratio (Po /Pa) exceeds about 52, the flow pattern changes dramatically and a Mach disk appears as the internal shock crosses itself. In round nozzles, a Mach disk is generally present when Pe /Pa > 2 [37]. Since the exit pressure ratio of the convergent nozzle always exceeds 2 (Table 1), one might have expected to see a Mach disk at all the pressures examined here.

12

S. P. MATES AND G. S. SETTLES

However, because of the presence of the embedded sonic boundary surrounding the wake, the internal shock is prevented from crossing to form the Mach disk at pressure ratios below 52. The transition occurs because, as Po increases, the point where the internal shock is intercepted by the wake boundary moves progressively farther downstream, as Fig. 7 shows. Eventually this causes the wake to suddenly collapse to a smaller size, allowing the internal shock to propagate freely through the jet and cross to form a Mach disk. The collapse of the wake and the subsequent Mach disk formation occurs very suddenly, at a well-defined pressure ratio. We label the flow pattern containing the Mach disk as being closed wake, while the previous pattern is referred to as open wake. The terms closed wake and open wake are borrowed from [34], but here they carry a slightly different meaning. In the present context they refer specifically to the presence or absence of the Mach disk structure. In [34], wake closure refers to the point at which the wake region loses pressure communication with the surroundings, which happens when the wake first becomes completely enclosed by supersonic flow. In our case, however, the wake is never open in this sense, since it is always completely enclosed by supersonic flow. When the wake closes, the pressure acting at the delivery tube orifice drops dramatically. This is shown in Fig. 8, which plots the pressure measured in the liquid delivery tube against overall pressure ratio. The pressure acting at the tip of the liquid delivery tube is commonly referred to as the aspiration pressure, since it is usually below the ambient pressure and thereby helps aspirate liquid through the tube. At wake closure, the aspiration pressure of the convergent nozzle drops sharply. Similar aspiration pressure behavior observed by other investigators [9, 10] indicates that wake closure is common among convergent close-coupled nozzles used for atomizing liquid metal. Understandably, there has been considerable discussion of the possible effects of wake closure on atomization performance. Some have argued that the presence of a Mach disk probably enhances atomization due to the rapid pressure rise across the wave [10]. Others have suggested that wake closure may affect melt flow stability, which can contribute to poor particle size control [38]. In the present experiments, Schlieren photography revealed that the introduction of molten metal into the flow causes the wake to reopen and the Mach disk to

Fig. 7 Schlieren images of the convergent nozzles near-region flow field at increasing values of Po/Pa (indicated below each image). Wake closure occurs in the region 52 < Po /Pa < 55.

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

13

Fig. 8 Variation in delivery tube pressure (Ptip ) with Po /Pa.

disappear [23]. It was therefore concluded that wake closure did not influence atomization performance in our experiments. Wake closure may have an impact, perhaps, if the conditions are such that transitions between closed-wake and open-wake structures occur randomly while the atomizer is in operation. In such a case, the resulting abrupt changes in aspiration pressure might then affect the stability of the metal flow rate and thereby influence atomization results. Such an extremely unstable situation requires careful balance between melt flow rate and stagnation pressure, however, making it an unlikely operating mode for most atomizers. For the present study, wake closure is not considered to have an important effect on the atomization behavior of the convergent nozzle. The flow generated by the c-d nozzle in the near region is described in Fig. 9. Because its exit pressure ratio is always near unity, the c-d nozzle produces a much less prominent internal shock wave compared to the convergent nozzle. As a result, wake closure does not occur. As Fig. 8 shows, the aspiration pressure declines smoothly as Po /Pa increases and there are no sudden changes. Further, the c-d nozzles flow lacks the prominent streamwise vortices that were present in the convergent nozzles highly underexpanded flow. The wake flow pattern is assumed to be similar to that of the convergent nozzle, one of low-velocity circulation. However, c-d nozzles wake is somewhat broader, and it lacks the concave shape of the convergent nozzles wake. These differences result from its straight-walled liquid delivery tube design and the lack of a strong internal shock wave. As a result, the velocity and pressure distributions inside the c-d nozzles wake are different from the convergent nozzles wake, possibly enough to alter primary breakup behavior. This will be investigated further in Part 2 of this article. The foregoing results pointed out a number of similarities, and a number of differences, in the gas dynamic behavior of the two nozzles in the near region. Both produce a large wake consisting of a low-speed circulating flow surrounded by a high-velocity, supersonic annular jet. Differences in the exit pressure ratio as well as the delivery tube geometry of the two nozzles lead to different shock wave patterns and wake zone characteristics. Because of its highly underexpanded exit flow, the convergent nozzle undergoes

14

S. P. MATES AND G. S. SETTLES

wake closure, in which the wake suddenly collapses, the aspiration pressure drops dramatically, and a Mach disk appears. By contrast, the c-d nozzle undergoes no such wake closure event, most likely because of its near-proper supersonic expansion, resulting in a weaker internal shock wave. However, since the Mach disk is eliminated from the convergent nozzles flow when atomization begins, wake closure is not a factor in the present experiments. In most other respects, the experimental results indicated little reason to expect there to be gross differences in the atomization behavior of the two nozzles in the near region. This will be examined in Part 2. Far-Field Flow Behavior and Comparison Schlieren images of the gas flow patterns produced by the two nozzles at up to 12 diameters downstream of the nozzle tip are shown in Fig. 10. Images are taken when operating the nozzles at nominally low, medium, and high stagnation pressures. In each case, the annular jet structure near the nozzle exit gives way to a predominantly round jet structure a few diameters downstream of the nozzle tip. The round jets exhibit the classic repeating shock diamond patterns typical of imperfectly expanded jets produced by round nozzles [39]. Figure 10 also reveals the dramatic effect of Po on the length of the supersonic region extending from the nozzle exit. This effect, which has generally been overlooked in previous work, is anticipated to have an important impact on secondary atomization events that take place well downstream of the nozzle tip region. Comparing the two nozzles, Fig. 10 indicates that, at Po /Pa 14, the c-d nozzle produces a supersonic jet that is roughly twice the length of the jet produced by the convergent nozzle. Because of the limited field of view of the schlieren optics, however, comparisons at higher stagnation pressures cannot be made. Instead, jet lengths are compared using a combination of centerline Pitot pressure measurements and wedge Mach number measurements. These data, shown in Fig. 11, confirm that the c-d nozzle pro-

Fig. 9 Schlieren images of the c-d nozzles near-region flow field at increasing values of Po /Pa (left), flow field sketch (right, bottom), and Pitot pressure (Pt2) data (right, top) acquired at Po /Pa = 53.

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

15

Fig. 10 Far-field flow patterns produced by the convergent and c-d nozzles at various values of Po /Pa (as indicated).

duces a significantly longer supersonic jet than the convergent nozzle at Po/Pa 14. However, at the higher two stagnation pressures (Po /Pa 28, 48), the supersonic lengths are comparable, although the c-d nozzle enjoys a slight advantage at Po /Pa 48. Some of its advantage derives from its greater mass flux, as noted in the experimental section. Overall, the data indicate that, despite large differences in exit pressure ratio, the two nozzles produce roughly similar supersonic jet lengths. This result appears to contradict literature data that suggest c-d nozzles produce significantly longer supersonic jets than convergent nozzles under equivalent conditions [2527]. One might argue that the present c-d nozzle fails to produce a longer supersonic jet because it never achieves shock-free flow. However, the oblique shocks formed in mildly off-design supersonic jets have been shown to have little influence on velocity decay or supersonic length [39]. The surprise, therefore, may not be that the c-d nozzle produces a shorter supersonic jet than it should. Rather, it seems that the convergent nozzle performs better than expected. The literature suggests a possible explanation for this apparently unexpected result. The shortening of supersonic length due to underexpansion effects appears to be related specifically to the presence of a Mach disk in the jet. As already mentioned, oblique shock cells are known to have little effect on jet velocity decay. Further, experimental data indicate that supersonic lengths produced by convergent and c-d nozzles begin to diverge at jet Mach numbers above about 1.5 [40], which is close to the point where the Mach disk forms in convergent nozzles (e.g., when Pe /Pa > 2) [37]. The present convergent nozzle does not produce a Mach disk until Po /Pa > 52, just beyond the pressure

16

S. P. MATES AND G. S. SETTLES

range under investigation. Thus its surprisingly good performance may be due to the fact that it avoids producing a Mach disk despite being highly underexpanded because of the presence of the embedded wake zone. In fact, subsequent supersonic jet length measurements made on either side of wake closure show that supersonic length is indeed reduced as soon as the Mach disk appears [41]. Figure 12 shows that the present close-coupled nozzles produce significantly shorter supersonic jets compared to ideal round supersonic nozzles [24] under equivalent conditions. The points in Fig. 12 are computed by normalizing the supersonic length measure-

Fig. 11 Centerline Mach numbers from wedge measurements (circles) and Pitot pressure measurements (squares and triangles). Squares are calculated from the Rayleigh-Pitot formula [32], and triangles are calculated assuming isentropic subsonic flow (Po /Pa values in parentheses).

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

17

Fig. 12 Comparison of normalized supersonic length measurements with an empirical correlation for perfectly expanded round supersonic jets [24].

ments from Fig. 10 by the exit diameter of a hypothetical perfectly expanded c-d nozzle operating with the same overall pressure ratio (Po /Pa) and critical throat area (A *). This normalization procedure (details of which appear in [42]) eliminates mass flow from the comparison, highlighting the effect of nozzle geometry on supersonic length. Gas mass flow data needed to normalize the supersonic length data are shown in Fig. 13. The data show a power-law dependence on Po rather than the expected linear dependence. We believe the nonlinear behavior of the mass flow data is caused by small expansion of the sonic throats of the nozzles due to pressure-induced stresses, which cause the bottom plate to flex slightly with increasing Po. Discharge coefficients computed from the measured mass flow rates using the nominal critical throat diameters are also plotted in Fig. 13. The wide variation in Cd with pressure ratio is uncharacteristic of critical nozzles and is also thought to be due to throat expansion effects. Mass flow curves obtained after repeated nozzle installations agreed to within the measurement uncertainties, however, indicating that the installation procedure produced consistent unstressed sonic throat dimensions and that the nozzle did not deform plastically when stressed. The mass-flow normalized supersonic length data in Fig. 12 are fitted reasonably well by the following empirical equation (following [24]):

18

S. P. MATES AND G. S. SETTLES

An attenuation coefficient (A ) has been included to account for nozzle geometry effects. Equation (1) is plotted to fit the present data in Fig. 12 using A = 0.65. Thus the supersonic jets produced by these close-coupled nozzles are approximately 35% shorter than equivalent perfectly expanded supersonic jets produced by round (rather than annular) nozzles. The attenuation is most likely caused by the annular geometry of the closecoupled nozzles, which leads to a larger jet surface area compared to round nozzles, leading to more rapid turbulent diffusion of the jets momentum. The Schlieren photographs of Fig. 10 confirm that by the time the present jets acquire a full velocity profile (a few

Fig. 13 Air mass flow rates measured by Venturi flow meter and power-law regressions (top) and discharge coefficient estimates (bottom) for convergent and c-d nozzles.

L d

(5 M 2

0.8) A

(1)

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

19

diameters downstream of the wake region), they are 2550% thicker than their perfectly expanded round equivalents. The jet produced by the c-d nozzle at Po /Pa 14 fails to follow this pattern, however, as it apparently does not suffer from significantly accelerated diffusion (Fig. 12). At present, this behavior is not well understood. However, it indicates that the attenuation of these jets is more complex than the simple model presented here. In addition to supersonic length, an additional flow quantity of importance to atomization behavior is the dynamic pressure, q. Dynamic pressure in part determines the Weber number applicable to both primary and secondary breakup. It can be calculated from the centerline Mach number data if the local static pressure (P ) of the flow is known using the following [43]:

Thus, for a given static pressure, higher Mach numbers mean higher dynamic pressures and larger corresponding Weber numbers. However, in the present case the static pressure distribution is unknown and varies due to the presence of shock cells. An approximation of the average dynamic pressure along the centerline of these jets can be made by assuming the static pressure is everywhere equal to the ambient pressure. Essentially, this averages out the effect of the shock cells, across which the static pressure rises and falls about the ambient pressure. Smooth polynomial fits are drawn through the centerline Mach number data obtained with the convergent nozzle in Fig. 14 to approximate how the centerline dynamic pressure distributions change with increases in Po. For the convergent nozzle, there is a significant increase in the peak centerline Mach number when Po /Pa increases

Fig. 14 Effect of pressure ratio on the average centerline Mach number distributions produced by the convergent nozzle.

1 PM 2 2

(2)

20

S. P. MATES AND G. S. SETTLES

from 14 to 28, but there is little increase beyond this pressure. Thus, while the maximum dynamic pressure rises initially with Po, it begins to level off with further increases. However, due to the large increase in supersonic length aided by the increasing mass flux, the dynamic pressure remains at a high level over a significantly longer distance. Therefore, increasing Po /Pa from 28 to 48 on the convergent nozzle primarily extends the high region of high dynamic pressures rather than increases the maximum dynamic pressure available. Thus Po has a more dramatic effect on the extent of the aerodynamic forces available to atomize liquid metal than on their peak magnitude. Thus atomization enhancement achieved by increasing Po may have much to do with its effect on secondary breakup occurring downstream of the nozzle tip, rather than on primary breakup.

SUMMARY This study examined and compared the gas dynamic characteristics of convergent and converging-diverging close-coupled nozzles used for atomizing liquid metals to produce fine metal powders. Schlieren photography and Pitot pressure surveys were used to characterize the mean gas flow fields generated at different nozzle stagnation pressures in the nozzle tip region as well as farther downstream. Each flow field contained an embedded region of separated flow (wake) at the base of the liquid metal delivery tube surrounded by an annular region of supersonic flow. This low-velocity circulating flow zone typically extended two to four nozzle tip diameters downstream. Beyond the wake, the annular supersonic flow region gradually gives way to a supersonic jet with a full (single-peaked) velocity profile, similar to an imperfectly expanded supersonic jet generated by a round nozzle. The convergent nozzle flow is highly underexpanded over the entire test range. Its near-region flow pattern is dominated by a strong internal shock wave. The interaction between this strong shock wave and the wake region causes an abrupt change in flow structure, called wake closure, at a pressure ratio between 52 and 55. Wake closure involves the sudden appearance of a Mach disk structure and a precipitous drop in the pressure acting at the liquid delivery tube orifice. The c-d nozzle, by contrast, because it operates with a near-unity exit pressure, does not produce a strong internal shock wave and suffers no wake-closure event. Increasing the nozzle pressure ration from 14 to 55 produced a two- to fourfold increase in supersonic jet length and more moderate increases in the maximum dynamic pressure generated along the flow centerline. Both effects are believed to play a role in enhancing atomization (reducing mean powder particle size) as the stagnation pressure increases, particularly through their influence on secondary breakup. The convergent and c-d nozzles produced similar supersonic jet lengths and dynamic pressures at equivalent values of Po, indicating that nozzle geometry had less influence on the external velocity fields generated by these nozzles than did Po. The c-d nozzle did produce a significantly longer supersonic jet than the convergent nozzle at the low end of the stagnation pressure range (Po /Pa 14), however. The overall similarity in the gas dynamic behavior of these nozzle suggests that they also ought to perform similarly in atomizing liquid metal at medium and high pressure ratios, although not at Po /Pa 14, where the c-d nozzle generates a longer supersonic jet. In Part 2 of this article, the atomization behavior of these nozzles is examined and compared at low, medium, and high stagnation pressures to test this assumption and to examine the effect of increasing supersonic jet length with Po on particle size.

LIQUID METAL ATOMIZATION USING CLOSE-COUPLED NOZZLES, 1

21

REFERENCES
1. E. Klar and J. W. Fesko, Atomization, Metals Handbook, 9th ed., vol. 7, pp. 2551, American Society for Metals, Metals Park, OH, 1984. 2. A. J. Yule and J. J. Dunkley, Atomization of Melts for Powder Production and Spray Deposition, Oxford University Press, Oxford, UK, 1994. 3. A. Lawley, Atomization: The Production of Metal Powders, Metal Powder Industries Federation, Princeton, NJ, 1992. 4. R. M. German, Powder Metallurgy Science, Metal Powder Industries Federation, Princeton, NJ, 1995. 5. R. I. Howells, G. R. Dunstan, and C. Moore, Production of Gas Atomised Metal Powders and Their Major Industrial Uses, Powder Metall., vol. 31, no. 4, pp. 259266, 1988. 6. A. H. Lefebvre, Atomization and Sprays, Hemisphere, New York, NY, 1989. 7. J. S. Thompson, A Study of Process Variables in the Production of Aluminium Powder by Atomization, J. Inst. Metal., vol. 74, pp. 101132, 1948. 8. R. Busch, The Thermophysical Properties of Bulk Metallic Glass-Forming Liquids, J. Metals, pp. 3942, July 2000. 9. S. D. Ridder, S. A. Osella, P. I. Espina, and F. S. Biancaniello, Intelligent Control of Particle Size Distribution during Gas Atomization, Int. J. Powder Metal., vol. 28, no. 2, pp. 133147, 1992. 10. I. E. Anderson, H. Morton, and R.,S. Figliola, Fluid Flow Effects in Gas Atomization Processing, in W. M. Small (ed.), Physical Chemistry of Powder Metals, Production and Processing, pp. 229249, The Minerals, Metals and Materials Society, [[AU: PLACE OF PUBLICATION?]]1989. 11. S. A. Miller, S. Savkar, and R. S. Miller, Global and Midfield Imaging of Liquid Metal Atomization, in A. Lawley and A. Swanson (compilers), Advances in Powder Metallurgy and Particulate Materials1993, pp. 2534, Metal Powder Industries Federation, Princeton, NJ, 1993. 12. A. nal, Effect of Processing Variables on Particle Size in Gas Atomization of Rapidly Solidified Aluminium Powders, Mater. Sci. Technol., vol. 3, pp. 10291039, 1987. 13. A. nal, Influence of Gas Flow on Performance of Confined Atomization Nozzles, Metall. Trans. B, vol. 20B, pp. 833843, 1989. 14. S. D. Ridder, F. S. Biancaniello, P. I. Espina, and S. A. Miller, Laser Holographic Imaging of the Two Fluid Atomization Process, in E. J. Lavernia, H. Henein, and I. Anderson (eds.), Synthesis and Analysis in Materials Processing: Advances in Characterization and Diagnostics of Ceramics and Metal Particulate Processing, pp. 7386, TMS,[[AU: SPELL OUT]] Warrendale, PA, 1989. 15. F. Biancaniello, C. Presser, and S. D. Ridder, Real-Time Particle Size Analysis during Inert Gas Atomization, Mater. Sci. Eng., vol. A124 pp. 2129, 1990. 16. P. I. Espina and U. Piomelli, Numerical Simulation of the Gas Flow in Gas-Metal Atomizers, in Proc. 1998 ASME Fluids Engineering Division Summer Meeting, FEDSM98-4901, ASME, Washington, DC, 1998. 17. J. Mi, R. S. Figliola, and I. E. Anderson, Numerical Simulation of the Effects of Gas Atomizing Pressure on Gas Flow Field in High Pressure Gas Atomization, in M. Phillips and J. Porter (compilers), Advances in Powder Metallurgy and Particulate Materials1995, Part 1: Powder Production and Characterization, pp. 1-411-51, Metal Powder Industries Federation, Princeton, NJ, 1995. 18. D. W. Kuntz and J. L. Payne, Simulation of Powder Metal Fabrication with High Pressure Gas Atomization, in M. Phillips and J. Porter (compilers), Advances in Powder Metallurgy and Particulate Materials1995, Part 1: Powder Production and Characterization, pp. 1-631-77, Metal Powder Industries Federation, Princeton, NJ, 1995.

22

S. P. MATES AND G. S. SETTLES

19. J. T. Strauss, Close-Coupled Gas Atomization Using Elevated Temperature Gas, Advances in Powder Metallurgy and Particulate Materials1999, Part 1, Metal Powder Industries Federation, Princeton, NJ, 2000. 20. N. Dombrowski and W. R. Johns, The Aerodynamic Instability and Disintegration of Viscous Liquid Sheets, Chem. Eng. Sci., vol. 18, pp. 203214, 1963. 21. A. nal, Liquid Break-up in Gas Atomization of Fine Aluminum Powders, Metall. Trans. B, vol. 20B, pp. 6169, 1989. 22. R. S. Miller, S. A. Miller, S. D. Savkar, and D. M. Mourer, Two Phase Flow Model for the Close-Coupled Atomization of Metals, Int. J. Powder Metal., vol. 32, no. 4, pp. 341352, 1996. 23. S. P. Mates and G. S. Settles, High-Speed Imaging of Liquid Metal Atomization Using Two Different Close-Coupled Nozzles, in Advances in Powder Metallurgy and Particulate Materials1996, Part 1: Powder Production and Characterization, pp. 1528, Metal Powder Industries Federation, Princeton, NJ, 1996. 24. H. T. Nagamatsu, R. E. Sheer, Jr., and G. Horvay, Supersonic Jet Noise Theory and Experiments, NASA SP-207, 1969. 25. G. L. Morrison and D. K. McLaughlin, Instability Process in Low Reynolds Number Supersonic Jets, AIAA J., vol. 18, no. 7, pp. 793800, 1980. 26. T.-F. Hu and D. K. McLaughlin, Flow and Acoustic Properties of Low Reynolds Number Underexpanded Supersonic Jets, J. Sound Vibration, vol. 131, no. 3, pp. 485505, 1990. 27. U. Heck, U. Fritsching, and K. Baukhage, Gas Flow Effects on Twin Fluid Atomization of Liquid Metals, Atomization and Sprays, vol. 10, no. 1, pp. 2546, 2000. 28. A. nal, Influence of Nozzle Geometry in Gas Atomization of Rapidly Solidified Aluminium Alloys, Mater. Sci. Technol., vol. 4, 1988, pp. 909915. 29. P. Hill and C. Peterson, Mechanics and Thermodynamics of Propulsion, 2d ed., p. 519, AddisonWesley, Reading, MA, 1992. 30. R. W. Miller, Flow Measurement Engineering Handbook, 2d ed., McGraw-Hill, New York, NY, 1989. 31. G. S. Settles, Schlieren and Shadowgraph Techniques, Springer-Verlag, Berlin, Germany, 2001. 32. F. M. White, Fluid Mechanics, 2d ed., McGraw-Hill, New York, NY, 1986. 33. NACA Report 1135: Equations, Tables and Charts for Compressible Flow, National Advisory Committee for Aeronautics, Ames Aeronautical Laboratory, Moffett Field, CA, 1953. 34. W. P. Sule and T. J. Mueller, Annular Truncated Plug Nozzle Flowfield and Base Pressure Characteristics, J. Spacecraft Rockets, vol. 10, no. 11, pp. 689695, 1973. 35. J. L. Herrin and J. C. Dutton, Supersonic Base Flow Experiments in the Near Wake of a Cylindrical Afterbody, AIAA J., vol. 32, no. 1, pp. 7783, 1994. 36. A. Krothapalli, G. Buzyna, and L. Lourenco, Streamwise Vortices in an Underexpanded Axisymmetric Jet, Phys. Fluids A, vol. 3, no. 8, pp. 18481851, 1991. 37. C. Dup. Donaldson and R .S. Snedeker, A Study of Free Jet Impingement. Part 1. Mean Properties of Free and Impinging Jets, J. Fluid. Mech., vol. 45, part 2, pp. 281319, 1971. 38. J. Ting, M. W. Peretti, and W. B. Eisen, The Effect of Wake-Closure Phenomenon on Gas Atomization Performance, Mater. Sci. Eng. A, vol. 326, no. 1, pp. 110121. 2002. 39. J. M. Seiner and T. D. Norum, Aerodynamic Aspect of Shock Containing Jet Plumes, AIAA Paper 80-0965, AIAA 6th Aeroacoustics Conf., Hartford, CT, June 46, 1980. 40. F. I. Lukhtura, One-Dimensional Theory of Off-Design Supersonic Gas Jets, Izvestiya Rossiiskoi Akademii Nauk, Mekhanika Zhidkosti i Gaza, no. 1, pp. 4856, Jan.Feb. 1993. 41. S. P. Mates, F. S. Biancaniello, and S. D Ridder, Comparison of the Supersonic Length and Dynamic Pressure Characteristics of Discrete-Jet and Annular Close-Coupled Nozzles Used to Produce Fine Metal Powders, in Liquid Metal Atomization: Fundamentals and Practice, The Minerals, Metals and Materials Society, Warrendale, PA, 2000. 42. C. K. W. Tam and H. K. Tanna, Shock Associated Noise of Supersonic Jets from ConvergentDivergent Nozzles, J. Sound Vibration, vol. 81, no. 3, pp. 337358, 1982. 43. A. B. Cambel and B. H. Jennings, Gas Dynamics, McGraw-Hill, New York, NY, 1958.

You might also like