A FLOW VISUALIZATION STUDY OF THE GAS DYNAMICS OF LIQUID METAL

ATOMIZATION NOZZLES

S.P. Mates and G.S. Settles

Gas Dynamics Laboratory
Department of Mechanical Engineering
The Pennsylvania State University
University Park, Pennsylvania

ABSTRACT

An assessment of the how well close-coupled liquid metal atomization nozzles expand high-
pressure gas to high velocities, and of how effectively they bring the high-velocity gas to bear on the
liquid metal, may point the way towards enhancing powder production capability through improved
nozzle design. The first step towards this is to study the gas-only flowfields produced by these nozzles.
The gas dynamic behavior of three different nozzles is examined through flow visualization. Two of
these nozzles are representative of the current state of technology: a converging (HPGA) and a
converging-diverging (Ünal) design. A truncated plug nozzle from the propulsion literature is included
for comparison. All nozzles are examined using flow visualization over a range of operating pressures
between 2 and 5.5 MPa. Base (aspiration) pressures are measured, and pitot surveys of the flowfields are
taken to complement the visualizations. The HPGA nozzle produces a highly underexpanded flowfield
which undergoes an abrupt change, called wake closure, at a high pressure ratio. Shock waves resulting
from underexpansion cause high kinetic energy losses in the flow, especially at wake closure, which may
inhibit atomization downstream. The Ünal nozzle produces a large wake at all pressure ratios, keeping
high-velocity gas away from centerline, while its converging-diverging design limits underexpansion
losses. The truncated plug nozzle flowfield combines some features of the other nozzles, and suggests
nozzle design alternatives.

INTRODUCTION

Gas atomization of liquid metal via close-coupled nozzle technology is used to produce metal
powders with characteristics that cannot be achieved with other methods. It is often used to fabricate fine,
highly spherical powders typically used in applications requiring dense packing and good flow
characteristics. Alloy powders that cannot be made with chemical techniques are readily made using gas
atomization. In addition, the rapid cooling rates inherent in the process show the potential to produce
amorphous powders with unique properties. These capabilities ensure that gas atomization will continue
to be an important process in powder metallurgy.
 However, despite its importance, the physics of gas atomization is not well understood. Efforts to
understand it have been thwarted by the inherently complex nature of the atomization process, which
involves turbulence, liquid stability, two- and three-phase flows, compressibility, and shock waves. Heat
transfer, solidification, and melt chemistry also play important roles. In addition to this, progress has
been slowed by a general lack of continuity in research efforts. For example, although many current
close-coupled nozzles are of similar design, novel nozzle concepts such as that of Ünal1 may involve
different atomization mechanisms2 which make direct comparisons of atomizer performance difficult. As
a further complication, some atomization nozzle concepts are believed to be held as trade secrets, and are
thus not available for evaluation.
An better understanding of the physics of gas atomization is essential if improvements in
atomization performance are to be made. Gas atomization is regarded as a relatively inefficient
production method in comparison to other techniques3. Control over particle size and size distribution
during production has the potential to cut production costs by increasing the yield of sellable powder per
batch produced. Also, demands continue to be made for finer particle sizes. These issues have motivated
research efforts geared towards improving the performance of close-coupled atomizers.
This research has primarily involved empirical investigation of various laboratory and
production-scale atomizers4. Experiments have shown that process parameters such as gas pressure, melt
superheat, and gas-to-liquid-metal flow rate ratio are important factors determining powder size and
production efficiency. Liquid metal delivery tube shape, size, and protrusion distance from the gas nozzle
exit have been varied in attempts to enhance atomization performance. Some nozzles use discrete jets,
while others use annular arrangements, both with various impingement angles. While this work is
important, its scope may be too limited. For example, specific results obtained from these experiments
are usually applicable only to a single atomizer, producing powder of a particular chemistry, and are thus
not widely applicable. More importantly, they do not provide much insight as to how high velocity gas
actually atomizes liquid metal in these nozzles.
Atomization researchers have agreed that high-velocity gas is important to effective atomization.
Studies have shown that maximizing the Weber number leads to quicker, more complete breakup in both
primary5 and secondary6 atomization. Since the Weber number is directly proportional to the relative
velocity squared, high gas velocity is desirable. To take full advantage of this requires understanding
how high-pressure gas is expanded to high velocities in close-coupled nozzles, and how it is brought to
bear upon the liquid metal. Efforts to study the flowfields of these nozzles indicate the presence of shock
waves near the nozzle exit, but these are not well understood to date. Their presence indicates that the
atomizing gas has not been perfectly expanded, which means that some of the gas energy, potentially
available to break up liquid metal, is lost. Alternative nozzle concepts, such as those used in air-breathing
propulsion, are specifically designed to expand the gas to high velocity over a range of operating
conditions. An investigation of the potential of such nozzle types to atomize liquid metal may prove
valuable, and has not yet been attempted to any significant extent.
An important first step towards understanding the behavior of current close-coupled nozzles is a
more complete evaluation of the flowfields they produce. This knowledge will provide a basis for
understanding how well high velocity gas is brought to bear on the liquid metal. Comparisons of
different nozzle concepts on this basis will provide a means to measure their relative ability to expand
high pressure gas to high velocity, something never before tried. The potential of new nozzle concepts
can also be evaluated. The goal of the present study is thus threefold: first, to provide a more complete
description of the gas dynamic phenomena present during the operation of representative close-coupled
nozzles, second, based on these results, to make comparisons of the relative effectiveness of these nozzles
in producing a high-velocity flowfield near the nozzle exit, and third, to set the stage for further
experiments involving atomization of liquid metal.

EXPERIMENTAL PROCEDURE

2
Nozzles. Schematics of the three nozzles investigated in this study are shown in Figure 1. Two are taken
from the liquid metal atomization literature, while the other is of a type used in air-breathing propulsion.
All three are fabricated to be similar in size. Critical dimensions including throat areas, exit-to-throat area
ratios, and design pressure ratios of the three nozzles are listed in Table 1.

Table 1. Nozzle design parameters.

Nozzle Minimum Area Area Ratio Design Pressure

(mm2) (Ae/Amin) Ratio (Po/Pa)
HPGA 9 1 -
Ünal 13 4 35
TPN 30 4 35

The High Pressure Gas Atomizing (HPGA) nozzle is shown in Figure 1a. The liquid delivery
tube is tapered at the end, and high pressure gas is expanded through an annular channel around it. This
channel amounts to a converging nozzle, since its flow area decreases to a minimum at the nozzle exit.
This design is patterned after a nozzle studied by NIST7, and is similar to a nozzle patented by Anderson8.
It is believed that this design represents a large class of nozzles currently used to produce metal powders,
thus it is hoped that results gained in this study will be of general use. Actual dimensions have been
selected to closely match the NIST nozzle. The principal difference is that an annulus is used here rather
than a symmetric array of discrete jets. The annular design was chosen for ease of implementation, and
because differences in gas dynamic behavior were not expected to be dramatic.
Our second nozzle is called the Ünal nozzle, after the investigator who first used it to atomize
liquid metal. An annular converging-diverging nozzle is used to expand the gas to supersonic velocity.
Since it is not properly contoured, the exit flow will never be perfectly expanded and shock waves will
form at all pressure ratios. Dimensions of this nozzle (Figure 1b) are similar to Ünal's original design1.
The third nozzle is the Truncated Plug Nozzle (TPN) shown in Figure 1c. It is based on a nozzle
studied by Sule and Mueller9 for use in air-breathing propulsion. It is termed an "altitude-compensating"
nozzle due to its ability to adjust its effective area ratio in response to changing overall pressure ratios,
which allows it to minimize losses due to imperfect expansion. The TPN design amounts to an annular
converging-diverging nozzle, similar to the Ünal nozzle in that the diverging section is conical, but
different in that the slope is on the inner surface rather than on the outer surface of the nozzle. This
results in a tapered liquid delivery tube tip similar to the HPGA design. It is included in the present study
for comparative purposes, and to examine a possible alternate design for liquid metal atomization nozzles.

Experimental Apparatus. The plenum chamber of the experimental atomizer is made up of a short
section of 6.35 cm diameter pipe. Flanges are welded to the pipe section on either end, and high-strength
round aluminum plates are bolted to these flanges. A thin cylindrical centerbody serves as the liquid
delivery tube and runs through the center of the plenum chamber. Each delivery tube has a 4 mm
diameter bore along its centerline through which liquid may flow. The top end of the delivery tube
protrudes through an O-ring seal in the center of the top plate to allow access to a pressure tap or to
connect to a liquid reservoir. The bottom tip of the tube protrudes through an orifice in the bottom plate,
forming the annular channel through which gas is expanded. Nozzles are interchanged by installing the
appropriate bottom plate/centerbody combination.

3
 (a)

(b) (c)

Figure 1. Nozzle dimensions in millimeters of (a) HPGA nozzle, (b) Ünal nozzle,
(c) Truncated Plug Nozzle.

Figure 2 is a schematic of the gas delivery system. A large reservoir is charged with air by a 7
MPa compressor. High pressure air is supplied to the plenum chamber through a feed pipe connected to
the chamber through the top plate. Plenum pressure is controlled by a regulator in between two ball
valves, and can be monitored during operation by a pressure tap in the plenum chamber. Nozzle base
pressure (aspiration pressure) is measured by a pressure line connected to the delivery tube. Both plenum
and base pressures are measured by diaphragm-type pressure transducers whose outputs are recorded on a
digital oscilloscope.
The flowfields produced by the atomizing nozzles are visualized using a Töpler schlieren
system10 with 15 cm diameter field lenses. The schlieren effect visualizes regions of density gradient in
the flowfield, such as shock waves and shear layers. A halogen lamp provides continuous illumination.
The schlieren cutoff is a graded filter positioned perpendicular to the flow direction. Images are captured
by a CCD video camera and recorded on SVHS video tape. The effective exposure time of 1/30 second
visualizes only the steady density gradients in the flowfields. Final images are digitized from the
videotape and processed electronically.
Pitot pressure surveys are made by traversing a small pitot tube aligned with the flow direction
along planes perpendicular to the flow11. The probe tip is 0.9 mm in diameter and is made of stainless
steel tubing. A pressure transducer is mounted directly below the probe, and the entire assembly is
attached to a support arm driven by stepper motors in two directions. A linear position transducer is used
to monitor probe location along the survey path. Pitot pressure is recorded at a data rate of 1 kHz as the

4
probe is moved radially out from the centerline of the flow. Position and plenum pressure data are
recorded along with pitot pressure using a LeCroy data acquisition system. The probe traversing speed
was selected to minimize the run length while allowing enough time for pressure equilibration. Run times
for each survey were about 7 seconds, during which the plenum pressure typically remained constant
within five percent. This was deemed an acceptable variation, since no changes in the flowfield were
noticeable during the surveys. The variations in flow direction relative to the axis of the probe were not
large enough to cause significant errors, since this probe design is relatively insensitive to such changes.

Figure 2. Schematic of high pressure air delivery system.

Since the flowfields produced by the present nozzles contain both subsonic and supersonic
regions, the interpretation of pitot pressures is not always straightforward and bears explanation. In
subsonic portions of the flow, the probe measures local stagnation pressure directly. When the probe
moves into a supersonic region, a normal shock will form at the tip, and the pressure measured is then the
stagnation pressure downstream of the shock. With higher Mach numbers the shock strength increases,
resulting in a greater stagnation pressure loss. Thus, in supersonic regions, a low pitot pressure generally
indicates a high Mach number. The local flow Mach number can be determined from the pitot pressure
only in regions of the flowfield where the stagnation pressure upstream of the probe tip, or the local static
pressure, is known.

RESULTS AND DISCUSSION

HPGA Nozzle. Visualization of the HPGA nozzle flowfield at pressure ratios (Po/Pa) ranging from 24 to
57 is shown in Figure 3. The images have been rotated 90 degrees for presentation, so the flow direction
is left-to-right. The HPGA flowfields are underexpanded in all cases shown. An underexpanded
supersonic flow occurs when the static pressure of the flow leaving a nozzle is higher than the back
pressure. This condition causes the flow to expand outside the nozzle to adjust itself to the surrounding
pressure. The resulting expansion waves reflect from the jet boundary or from solid surfaces and
eventually form shock waves. In the HPGA flowfield this results in the formation of an internal shock,
which originates near the nozzle exit and converges on centerline. The shock becomes stronger as the
pressure ratio is increased, and eventually forms an irregular crossing with a Mach disk12 (Figure 3d). A
wake is created beyond the delivery tube base due to flow separation. Boundaries of the wake near the

5
base are marked by the internal shear layer, which is visible as dark lines attached to the edge of the
delivery tube.
At a pressure ratio (Po/Pa) of 24 (Figure 3a), the internal shock is small and extends just beyond
the edge of the delivery tube. The shear layer at outer edge of the jet is visible near the nozzle exit as
dark lines, which disappear after a short distance as the mixing layer thickens due to viscous effects. A
small recompression shock originates near the corner of the delivery tube and turns the flow away from
centerline. After two diameters downstream, the wake region has disappeared, and the flow narrows
considerably. As the pressure ratio increases (Figures 3b,c), these features become more pronounced.
The external shear layer is darker and thicker, owing to higher velocities just past the nozzle exit, and the
internal shock grows larger, reaching farther downstream. The recompression shock moves off the
delivery tube and locates near the end of the internal shock. Reflections of both shocks are visible farther
downstream. The internal shear layer, which marks the wake boundary, becomes visible, growing larger
and more inclined toward centerline at higher pressures.

(a) (b)

(c) (d)

Figure 3. HPGA Nozzle at operating pressure ratios (Po/Pa) of (a) 24, (b) 43, (c) 53.5, (d) 57.

A dramatic change in the flowfield occurs at a slightly higher pressure ratio, as shown in Figure
3d. The internal shock suddenly crosses downstream as Mach disk, forming a continuous shock cell that
completely encloses the wake region, which shrinks to a third of its previous size. The flow on centerline
becomes supersonic after little more than one delivery tube tip diameter downstream, and the
recompression shock elongates in response to the greater flow expansion past the delivery tube tip. One
of the most striking things about this transition is that it occurs almost instantaneously. Since the wake
becomes entirely enclosed by the shock cell in this configuration, it is called "closed-wake" operation.
Conversely, lower pressure operation will be called "open-wake".
The transition from open- to closed-wake may be understood on a phenomenological basis by
examining the behavior of ordinary underexpanded axisymmetric jets. A schematic of the underexpanded
flow generated by operating a circular converging nozzle at a high pressure ratio is shown in Figure 4a.
The shock cell in this flowfield is formed by the same gas dynamic phenomena which produce the

6
internal shock the HPGA flowfield. Placement of a large cylindrical blunt body in the center of this
nozzle (Figure 4b) will produce a large separated wake. If the wake size is large compared to the jet
diameter, the shock cell will be forced open by the reflection of the internal shock off the wake boundary.
If the pressure ratio is increased to a critical value, the internal shock will elongate and cross once again
(Figure 4c), which is "wake closure". In the HPGA nozzle, the relative wake size is directly related to the
annular exit area (the "minimum area" in Table 1), with a larger exit area corresponding to a smaller
relative wake size. Thus, the critical pressure ratio required for wake closure is a strong function of exit
area. Dependence of this pressure ratio with exit area for the HPGA nozzle is shown in Figure 5.

(a) (b)

(c)

Figure 4. Schematic of flowfields produced by (a) simple underexpanded converging nozzle,

(b) "open wake" operation, and (c) "closed wake" operation.

The base pressure (aspiration pressure) behavior of the HPGA nozzle over this same pressure
ratio range is shown in Figure 6. During open-wake operation, the pressure is relatively uniform and sub-
atmospheric. A sudden drop in base pressure occurs at wake closure (around Po/Pa=55). This indicates
that the overall wake pressure has dropped, which is evident from the schlieren pictures before and after
wake closure (Figures 3c,d). The lower wake pressure allows the supersonic flow to expand more sharply
around the delivery tube tip, resulting in a steeper inclination of the internal shear layer towards centerline
in closed-wake operation.

7
 Figure 5. Dependence of wake closure pressure ratio on exit area in HPGA nozzle

Figure 6. Comparison of HPGA base pressure variation with operating pressure

data taken by NIST7 for a similar discrete jet nozzle.

The base pressure data for the HPGA nozzle are in general agreement with similar data taken by
NIST7 with their discrete-jet nozzle of otherwise similar design. Their data show a similar falloff at high
pressure ratio, indicating wake closure has occurred. Moreover, the general agreement in base pressure
data indicates that the flowfields produced by these nozzles are similar whether an annular channel or
discrete jets are used.
To complement the qualitative results provided by flow visualization, pitot surveys of the HPGA
flowfield have been performed. Results of pitot surveys of closed-wake operation are shown in Figure 7.
Pitot profiles are plotted at the axial location on the schlieren image corresponding to where the surveys
were taken, with the exception of the last two surveys which fall farther downstream. Pitot pressures are
normalized to plenum pressure in each case.
Rapid rises in pitot pressure occur near shocks and shear layers, across which there are abrupt
changes in flow velocity and stagnation pressure. Valleys in the pressure profiles near the nozzle exit
indicate high Mach number supersonic flow. The wake is marked by very low pitot pressures close to the
flow centerline. At x/Do=1.5, the pitot pressure on centerline has risen substantially, indicating the

8
centerline flow has accelerated to a high velocity at this point. By four diameters downstream, both
flowfields have coalesced, and exhibit smooth pitot pressure profiles typical of fully developed
supersonic freejets.
In the initial expansion region near the nozzle exit (region 1, Figure 7), the supersonic flow has
not yet passed through any shocks, and therefore the local stagnation pressure is equal to the plenum
pressure Po. For closed-wake operation this region is large, and Mach numbers can be estimated from the
pitot pressures at three axial locations. Maximum Mach numbers of 4.2, 4.8 and 5.0 are found at x/D=
0.1, 1.0 and 1.5, respectively. Flowfield calculations of the NIST nozzle at a similar operating pressure
ratio by Kuntz12 show similarly high Mach numbers, indicating general agreement with these
measurements.
Using oblique shock theory, flow property changes across the recompression shock can be
estimated. In region 2 of Figure 7, the Mach number remains high at about 4, and there is only a 5
percent loss in stagnation pressure, corresponding to a relatively weak recompression shock. Across the
Mach disk, however, the stagnation pressure loss is much more severe. For a normal shock at Mach 4,
gas dynamic theory predicts a 90 percent loss in stagnation pressure. This is confirmed by the survey at
x/D=2 (region 3 in Figure 7), which indicates the pitot pressure just behind the Mach disk is less than 10
percent of the plenum pressure. It is important to note that these Mach number estimates are based on the
assumption of inviscid flow, which is not applicable in regions where viscous effects are important.
Therefore, such estimates are not attempted in regions near the freejet boundary, where viscous mixing
occurs between the jet and the ambient air.

2 3 0 4 mm

Figure 7. Pitot surveys of HPGA nozzle in "closed wake" operation.

The investigation of the HPGA nozzle flowfield has revealed two features that may have a strong
effect on atomization. First, the internal shock is seen to increase in size and strength as the pressure ratio
is increased. The drop in gas velocity across this shock will therefore be greater at higher operating
pressures. The formation of a Mach disk at wake closure worsens the situation, as it decelerates the
centerline flow to subsonic velocity just two diameters downstream of the delivery tube tip. Clearly this

9
nozzle is not very effective in sustaining high-velocity gas flow near the nozzle exit, especially at high
operating pressures. Based solely on this argument, one might speculate that the atomization efficiency of
this nozzle will decrease as plenum pressures are increased. In fact, this has been suggested as a general
limitation of gas atomization3. If such a limitation exists, it may be due to shock losses, which can be
reduced by better expansion of the gas flow with an improved nozzle design. The second observation is
that the abruptness with which wake closure occurs may result in confusing variations in the resulting
powder properties near the critical pressure ratio if wake closure has not been anticipated.

Ünal Nozzle. Schlieren images of the Ünal nozzle flowfield at two pressure ratios are shown in Figure 8.
At Po/Pa=20 (Figure 8a), the nozzle is slightly overexpanded, meaning the static pressure of the gas at the
nozzle exit is lower than ambient. In response to the pressure mismatch, a shock is formed at the nozzle
exit and reflects off the wake boundary just past the delivery tube tip as an expansion fan. Another shock
is generated at the tip and turns the flow away from centerline. In highly underexpanded operation
(Figure 8b) the recompression shock has moved farther downstream and become stronger. In addition, it
appears to be somewhat smeared by flow unsteadiness. An internal shock can now be identified as a thin
dark line located just inside the visible shear layer. It is much weaker than the internal shock in the
HPGA flowfield because of the expansion provided by the diverging section of the nozzle. Large losses
in stagnation pressure across this internal shock are therefore avoided. Also, the wake generated by the
delivery tube remains almost unchanged over the pressure range examined, with no wake closure taking
place. As shown in Figure 9, the base pressure decreases monotonically as pressure ratio increases. Thus
the Ünal nozzle appears to operate in open-wake mode over the entire pressure range tested.

(a) (b)

Figure 8. Ünal Nozzle at operating pressure ratios (Po/Pa) of (a) 20, (b) 54

Pitot surveys of the Ünal nozzle operated at Po/Pa=54 are shown in Figure 10. It is evident that
the high-velocity gas is confined to an annular region far from the centerline for at least two tip diameters
downstream. A "dip" in the pitot pressure distribution is present just past the delivery tube tip at
x/D=0.16, indicating Mach number variation in this region. Based on the same assumption used in the
HPGA analysis, the maximum Mach number, indicated by the lowest pitot pressure, is estimated to be 3.9
at this location. Further estimations of Mach number are not possible due to viscous effects. By four tip
diameters downstream, the flow begins to coalesce into a single supersonic jet.

10
 Figure 9. Ünal nozzle base pressure variation with operating pressure.

0 4 mm

Figure 10. Pitot surveys of Ünal nozzle at a pressure ratio (Po/Pa) of 54.

The Ünal flowfield behavior is very uniform over the pressure range examined, with no abrupt
changes taking place such as wake-closure. The converging-diverging design avoids the formation of
strong shocks near the nozzle exit, such as in the HPGA flowfield, even when the nozzle is operated well
above its design pressure. Losses in gas velocity close to the nozzle are therefore less severe than in the

11
HPGA nozzle. However, the wake is large at all pressures, and high-velocity gas is kept away from the
centerline for two tip diameters downstream.

TPN nozzle. Schlieren images of the Truncated Plug Nozzle (TPN) at four pressure ratios are shown in
Figure 11. At the lowest pressure (Figure 11a) the flow is overexpanded. A shock forms at the nozzle
exit to adjust the jet pressure to ambient, then reflects off the wake as an expansion. A recompression
shock is attached to the delivery tube tip and extends away from centerline to the freejet boundary. The
wake region, though not well defined in the schlieren image, is large at this point.
Close to the design pressure ratio (Figure 11b), the TPN flowfield is still open-wake. At a
slightly higher pressure ratio (Figure 11c), wake closure has occurred, and the flowfield is now much like
that of the HPGA nozzle at wake closure. The wake has shrunk abruptly to a fraction of its former size,
allowing greater expansion around the delivery tube tip. The main difference is that the TPN internal
shock forms a regular crossing12 at wake closure rather than a Mach disk as in the HPGA case. This
means that centerline flow remains supersonic downstream of the crossing, and retains more of its kinetic
energy than in the case of the HPGA closed-wake flowfield. At the highest pressure (Figure 11d), the
shock crossing moves farther downstream.

(a) (b)

(c) (d)

Figure 11. TPN Nozzle at operating pressure ratios (Po/Pa) of (a) 21, (b) 31, (c) 39, (d) 45.

Base pressure behavior of the TPN nozzle is shown in Figure 12. Initially, the nozzle operates in
open-wake mode and the base pressure decreases monotonically, similar to the Ünal nozzle. At wake
closure (Po/Pa ~ 35), there is a sharp drop in base pressure, similar to the HPGA nozzle. For higher
operating pressures, the base pressure rises steadily. Thus the minimum base pressure occurs just at wake
closure.
Pitot surveys of the closed-wake TPN flowfield are shown in Figure 13. The results are similar to
surveys of the HPGA nozzle in closed-wake operation. Using the same approximation as before to find
Mach numbers in the initial expansion zone (region 1 in Figure 13), maximum Mach numbers are found
to be 3.6 at x/D=0.3, and 4.2 at x/D=1.2.

12
 Figure 12. TPN nozzle base pressure variation with operating pressure.

2
0 6mm

Figure 13. Pitot surveys of the TPN at a pressure ratio (Po/Pa) of 39.

In summary, it has been shown that TPN flowfield behaves much like the Ünal nozzle at low
pressures, but undergoes wake closure like the HPGA nozzle. This design takes advantage of expansion
within the nozzle to reduce shock loss, and is capable of closed-wake operation, which brings high-
velocity gas to the centerline much closer to the delivery tube tip. Also, the avoidance of the Mach disk
at wake closure is an added advantage because losses in flow energy are less severe. It must be noted that

13
Mach disk formation is a possibility with this design9, and whether or not it occurs is, in general, a
function of the particular nozzle dimensions and operating pressure ratio. However, elimination of the
Mach disk is shown to be possible in such a close-coupled nozzle with proper design considerations.

CONCLUSIONS

The gas dynamic behavior of current-generation close-coupled nozzles used to atomize liquid
metal to produce metal powder has been examined. Visualization of the flowfields produced by these
nozzles, along with base pressure data and pitot pressure surveys, have provided a comparative look at the
high-velocity gas flow in the vicinity of the delivery tube tip.
The underexpanded nature of the HPGA flowfield results in strong shock waves near the nozzle
exit, which seriously reduce gas energy available for atomization further downstream. An abrupt change
in the flowfield, called wake closure, occurs at a high plenum pressure and is likely to produce confusing
changes in atomization performance if not anticipated. At wake closure, high-velocity gas is brought
much closer to the centerline near the nozzle exit. The Mach disk formed at wake closure was shown to
incur a tremendous loss in gas kinetic energy, and is probably not desirable. The plenum pressure
required to produce wake closure is dependent on the nozzle exit area, and in general is likely to be a
strong function of geometry in other close-coupled nozzles of similar design. Also, the difference in gas
dynamic behavior between discrete-jet nozzles and annular nozzles of this type does not appear to be
significant.
The Ünal nozzle produces an internal supersonic expansion with its converging-diverging design,
thus limiting losses in gas kinetic energy due to a strong internal shock. The large wake region
downstream of the delivery tube tip keeps the high-velocity gas away from centerline at all pressure ratios
tested.
The Truncated Plug Nozzle incorporates features of both the HPGA and Ünal nozzles. It
provides expansion within the nozzle, thus minimizing flow loss due to strong shocks as in the Ünal
nozzle. It can also be operated in closed-wake mode, bringing high-velocity gas close to centerline near
the nozzle exit, and it avoids the formation of a Mach disk, which allows high gas velocities to be
maintained on centerline.
Further experiments are planned to visualize liquid metal breakup in these nozzles so that a better
assessment can be made of the specific impacts these flowfield features have on atomization. Results of
these experiments, together with the results presented here, will direct new nozzle design efforts using gas
dynamic principles, which may improve upon current gas atomization technology.

ACKNOWLEDGMENTS

This work is supported by the National Science Foundation under grant No. CTS-9221863. The authors
wish to thank J. D. Miller and Lori J. Dodson for their assistance, and Sanjay Garg, and Charles and
Elizabeth Hackett for their technical insight.

REFERENCES
1 A. Ünal, "Influence of Gas Flow Performance of 'Confined' Atomization Nozzles",
Metallurgical Transactions B, vol. 20B, 1989 pp. 833-843
2 A. J. Yule and J. J. Dunkley, Atomization of Melts for Powder Production and Spray
Deposition, Oxford Univ. Press, New York, 1994, pp. 165-174
3 R. M. German, Powder Metallurgy Science, Metal Powder Industries Federation,
Princeton, NJ, 1994, p. 116

14
4 S.A. Miller, "Global and Midfield Imaging of Liquid Metal Atomization", Advances in Powder
Metallurgy and Particulate Materials - 1993, Vol. 1, compiled by Alan Lawley and Armour
Swanson, Metal Powder Industries Federation, Princeton, NJ, 1993
5 W. O. H. Meyer, "Coaxial atomization of a round liquid jet in a high speed gas stream: A
phenomenological study", Experiments in Fluids, 16, 1994, pp. 401-410
6 L.-P. Hsiang and G. M. Faeth, " Near-Limit Drop Deformation and Secondary Breakup", Int. J.
Multiphase Flow, Vol. 18, No. 5, 1992, pp. 635-652
7 S. D. Ridder, "Optimization of Inert Gas Atomization", Physical Chemistry of Powder Metals,
Production and Processing, The Minerals, Metals, and Materials Society, 1989, pp. 163-173
8 I. E. Anderson, "Fluid Flow Effects in Gas Atomization Processing", Physical Chemistry of
Powder Metals, Production and Processing, The Minerals, Metals, and Materials Society,
1989, pp. 229-249
9 W. P. Sule and T. J. Mueller, "Annular Truncated Plug Nozzle Flowfield and Base
Pressure Characteristics", J. of Spacecraft and Rockets, Vol. 10, No. 11, Nov, 1973, pp. 689-
695
10 W. Merzkirch, Flow Visualization, Academic Press, Orlando, 1987.
11 D. W. Bryer and R. C. Pankhurst, "Pressure Probe Methods for Determining Wind Speed and
Flow Direction", National Physical Laboratory, Her Majesty's Stationary Office, London, 1971,
pp. 7-15
12 M.A. Saad, Compressible Fluid Flow, pp. 190-291, Prentice-Hall, New Jersey, 1985.
13 D. Kuntz, 1995, Sandia National Laboratories, private communication

15