You are on page 1of 36

Annu. Rev. Fluid Mech. 2000.

32:275308
Copyright 2000 by Annual Review. All rights reserved
00664189/00/01150275$12.00 275
LIQUID JET INSTABILITY AND ATOMIZATION IN
A COAXIAL GAS STREAM
J. C. Lasheras and E. J. Hopnger
Department of Mechanical and Aerospace Engineering, University of California,
San Diego, La Jolla, California 92093-0411; e-mail: lasheras@ames.ucsd.edu
LEGI-CNRS/UJF/INPG, B.P. 53, 38041 Grenoble Cedex, France; e-mail:
emil.hopnger@hmg.inpg.fr
Key Words stability, jets, sprays, combustion, turbulence
Abstract An overview of the near and far-eld breakup and atomization of a
liquid jet by a high speed annular gas jet is presented. The various regimes of liquid
jet breakup are discussed in the parameter space of the liquid Reynolds number, the
aerodynamic Weber number, and the ratio of the momentum uxes between the gas
and the liquid streams. Recent measurements of the gas-liquid interfacial instabilities
are reviewed and used to analyze the underlying physical mechanisms involved in
the primary breakup of the liquid jet. This process is shown to consist of the periodic
stripping of liquid sheets, or ligaments, which subsequently break up into smaller
lumps or drops. Models to predict the liquid shedding frequency, as well as the global
parameters of the spray such as the liquid core length and spray spreading angle are
discussed and compared with the experiments. The role of the secondary liquid
breakup on the far-eld atomization of the liquid jet is also considered, and an attempt
is made to apply the classical turbulent breakup concepts to explain qualitatively the
measurement of the far-eld droplet size distribution and its dependence on the liquid
to gas mass and momentum ux ratios. Models for the droplet breakup frequency in
the far-eld region of the jet, and for the daughter-size probability density function,
which account for the effect of the liquid loading on the local turbulent dissipation
rate in the gas, are discussed in the context of the statistical description of the spray
in the far eld. The striking effect of the addition of swirl in the gas stream is also
examined.
1. INTRODUCTION AND HISTORICAL PERSPECTIVE
When a liquid jet issues from a nozzle and discharges into a stagnant gas, it
becomes unstable and breaks into droplets. Using surface energy arguments, Pla-
teau (1873) proposed that, to achieve a state of minimum surface energy, a round
jet of diameter D
l
must break into equal segments whose length must be about
4.5 D
l
. Rayleigh (1879) demonstrated that this breakup results from a hydrody-
namic instability caused by surface tension and occurs at a relatively low jet
276 LASHERAS HOPFINGER
Reynolds number. He found that the most unstable wavelength is the one sug-
gested by Plateau. Weber (1931) extended Rayleighs analysis and showed that
the liquid viscosity has a stabilizing effect that lowers the breakup rate and
increases the size of the observed droplets. At a larger Reynolds number, the jet
becomes wavy owing to aerodynamic effects, and at an even larger Reynolds
number (of the order of 10
5
) atomization (the formation of droplets with sizes
much smaller than the diameter of the jet) takes place owing to short-wavelength
shear instability (Hoyt & Taylor 1977a, Lefebvre 1989, Lin & Reitz 1998).
In the presence of a high-speed coaxial gas stream, which is the subject of this
review, the breakup and atomization of the liquid jet is fundamentally different.
When surrounded by a gas with a momentum ux greater than that of the liquid,
the breakup of the jet is caused by the transfer of kinetic energy from the high-
speed gas to the liquid, a process known as air-blast atomization (Lefebvre 1989,
Lin & Reitz 1998). This type of breakup has widespread practical applications,
ranging from fuel injectors in gas turbines and jet engines, to two-phase ow
chemical reactors, chemical separators, spray drying, food processing, etc. In
liquid propellant rocket engines, for instance, liquid oxygen is atomized by a
high-speed annular coaxial hydrogen gas jet, and the formation of a stable ame
front is intimately related to the quality of atomization and the mixing of the
liquid oxygen with the hydrogen gas. Generally, the instability of the liquid-gas
interface and subsequent primary atomization are strongly dependent on the initial
conditions. Most of the theoretical treatments of the liquid-gas interfacial insta-
bility consider the gas/liquid interface to be innitely thin, not taking into account
the dominant role of the boundary layers formed in the nozzles.
In practical applications, the drop size is of primary interest and is generally
correlated by a power law dependence on the gas velocity. An explanation for
such a power law has been given only recently (Lasheras et al 1998). Equally
important is the drop size resulting from the primary breakup. Other character-
istics of interest are the spreading rate of the spray and the unbroken length or
liquid intact length L
b
where breakup begins, as well as the liquid core length L,
or length needed for the liquid jet to be completely broken into drops and
ligaments.
The format of this review is as follows. In Section 2, we discuss the different
breakup regimes of a liquid jet by a coaxial gas stream in a Reynolds-Weber
number parameter space. An analysis of the near-eld interfacial instability lead-
ing to the primary droplet formation is given in Section 3. High-speed video
images and optical-probe measurements demonstrate that the primary gas/liquid
instability is controlled by the high momentum gas and that its characteristic
wavelength and frequency are primarily determined by the thickness of the inter-
facial vorticity layer and the liquid/gas density ratio. Possible mechanisms for the
drop formation in the near-eld are also examined in this section. The mechanisms
determining the evolution of the far-eld of the spray are discussed in Section 4.
In this section we review recent experimental measurements of the far-eld drop-
let size probability density function (PDF) and discuss various models for the
AIR ASSISTED LIQUID JET ATOMIZATION 277
secondary liquid breakup. The striking effect of the addition of swirl in the gas
stream is examined in Section 5.
2. REGIMES OF LIQUID JET BREAKUP IN A COAXIAL
GAS STREAM
When a liquid jet of diameter D
l
and velocity U
l
discharges into a stagnant gas,
a Rayleigh instability manifests when the jet diameter is small and the jet Rey-
nolds number Re
l
U
l
D
l
/m
l
is not too large (of the order 10
2
). The upper limit
of Re
l
depends on the Ohnesorge number, Oh l
l
/(q
l
rD
l
)
1

2
, where l
l
and q
l
are
the viscosity and density of the liquid, and r is the interfacial surface tension. At
larger Reynolds numbers, the jet becomes wavy because of aerodynamic effects,
a regime called non-axisymmetric Rayleigh breakup or rst wind-induced regime.
When the Reynolds number is further increased, the wind stress at the gas/liquid
interface strips off droplets, and at the larger Reynolds number (around 10
5
),
atomization due to short-wavelength shear instability takes place (Hoyt & Taylor
1977a,b, Reitz & Bracco 1982).
Concerning the breakup length, in the Rayleigh regime, beyond dripping, the
intact length increases with the Reynolds number until a local maximum of about
L
b
/D
l
10
2
is reached. It then decreases in the non-axisymmetric regime to about
L
b
/D
l
10, before reaching a second maximum in the wind stress regime also
called second wind-induced regime. Beyond this point the intact length reduces
to an asymptotic value in the fully developed spray regime. This asymptotic value
depends strongly on the nozzle ow conditions, in particular on the boundary
layer thickness and the turbulence level (Reitz & Bracco 1982; Lin 1996). The
liquid core length L coincides with L
b
in the Rayleigh regimes but is generally
considerably longer otherwise. In the atomization regimes of interest here
L
b
0.
In this paper, we concentrate mainly on the analysis of the stability and the
atomization (spray formation) of a liquid jet injected into a high velocity annular
coaxial gas stream. In addition to the liquid Reynolds, Re
l
, and Ohnesorge, Oh,
number, the presence of the coowing gas stream introduces four newparameters:
the aerodynamic Weber number, ; the gas Reynolds number,
2
We q U D/r
g g l
Re
g
U
g
(D
g
D
l
)/m
g
, the momentum ux ratio
1
; and the mass
2 2
M q U /q U
g g l l
ux ratio, m q
l
U
l
A
l
/q
g
U
g
A
g
. Here, A
l
and A
g
are the areas of the liquid and
gas nozzle cross sections, respectively. In order to x ideas, the geometry and the
expected instabilities, their wavelengths, and breakup processes are sketched in
Figure 1.
1
M is actually the dynamic pressure ratio. We keep here the terminology (local) momentum
ux ratio used in previous publications.
278 LASHERAS HOPFINGER
Figure 1 Schematic representation of the liquid breakup indicating the geometry and
different lengths used in the analysis.
Depending on the application, a wide range of nozzle geometries is used. In
order to bring out the essential features of the various regimes encountered in
these ows, our discussion focuses primarily on coaxial jets issuing from con-
vergent nozzles where the initial conditions are known. Typical examples are
shown in Figure 2. Jets issuing from long tubes, as are often used in practical
applications, show a qualitatively similar behavior (Lasheras et al 1998). An
important point to notice in the visualizations shown in Figure 2 is the long
wavelength of the instability that develops at the gas/liquid interface. In Figure
2a, the aerodynamic Weber number is small (We 38), and one can see that the
surface tension just prevents the instability from growing and drops from being
formed. In Figure 2b, the Weber number is only slightly larger, We 58, and
one can now observe the formation of ligaments. In this case, the surface tension
acts on a scale about two thirds the size. Owing to its small radius of curvature,
the tip of the ligament recedes and a bulge is formed. The aerodynamic pressure
exerted by the gas can then blow up the sheet behind the bulge into a bag or
membrane (see also Lasheras et al 1998their Figure 2; and Chigier & Reitz
1996their Figure 8). This breakup regime is often referred to as membrane
breakup. In Figure 2c, the Weber number is 118, and one sees that the scale of
the ligaments is further reduced to about half the size of those in Figure 2b. Figure
2d shows a considerably ner spray corresponding to a Weber number of 316.
At these large Weber numbers, the breakup is said to occur in the form of bers
leading to the formation of smaller droplets (Farago & Chigier 1992, Hopnger
& Lasheras 1994, Lasheras et al 1998). In the downstream eld of Figure 2d, one
can see the existence of patches where the liquid is more concentrated. The origin
and the signicance of these patches will be explained in Section 4.
AIR ASSISTED LIQUID JET ATOMIZATION 279
a b
a b
c d
c d
Figure 2 Images of jet breakup by a coaxial gas ow. The nozzle contraction ratio is
7:1, and liquid and gas diameters are D
l
7.6 mm, D
g
11.3 mm respectively. (a) U
l
0.16 m/s and U
g
19 m/s, (b) U
l
0.55 m/s and U
g
21 m/s. (c) U
l
0.86 m/s
and U
g
30 m/s, and (d) U
l
0.26 m/s and U
g
50 m/s.
Due to the high contraction ratio of the nozzle used in the experiments shown
in Figure 2, the velocity proles at the exit are at with very thin laminar boundary
layers (vorticity layers) in both the gas and liquid streams. The wavelength of the
observed instability, which as mentioned above, is of the order of the jet diameter,
is considerably larger than the thickness of these vorticity layers (by a factor of
10
2
). Figure 2d also shows the possible coexistence of a short wavelength insta-
bility whose wavelength k
2
is shown in Figure 1. For jets issuing from long tubes
with turbulent velocity proles, the vorticity thickness is also small, and the
observed wavelength is also larger than the diameter of the liquid jet. In contrast,
the wavelength of the instability developing in the near-eld (up to 10 jet diam-
eters) of a high Reynolds number liquid jets discharging into a stagnant gas is
280 LASHERAS HOPFINGER
Figure 3 Breakup regimes in the parameter space Re
l
We. Lines of constant M are
calculated for water-air and D
l
7 mm. , conditions of Fig. 2a to 2f; , conditions of
Figure 6, | | |, regime of rocket engines. Conditions corresponding to the experiments
analyzed in section 4. Here, , Re
l
U
l
D
l
/m
l
; . (Note
2 2 2
We q U D/r M q U /q U
g g l g g l l
that D
l
3.5 mm in Figure 6 and in the measurements of section 4.2. We have shifted
the values up by a factor of 2 so that these fall in the right range of M.)
always very short compared to the diameter of the liquid jet (Hoyt & Taylor
1977a,b).
In order to have an overview of the different breakup and atomization pro-
cesses taking place over the wide range of liquid and gas conditions, it is useful
to establish a regime diagram similar to that used for pressure atomizers (Reitz
& Bracco 1982, Lin & Reitz 1998). The coaxial jet case of interest here is,
however, considerably more complicated. The composition of such a diagramcan
be very cumbersome since one has to deal with a much larger number of non-
dimensional numbers.
An attempt to establish such a diagram has been made by Farago & Chigier
(1992), which excludes the effect of M. A more complete diagram has been
proposed by Hopnger (1998). Hopngers diagram is reproduced in Figure 3 in
a slightly modied form. The various breakup regimes are presented in the param-
eter space of the liquid Reynolds number and the aerodynamic Weber number,
AIR ASSISTED LIQUID JET ATOMIZATION 281
and they are calculated for water and the geometry corresponding to the cases
shown in Figure 2. Lines of constant M appear as straight lines, Re
l

(We/M)
1/2
(D
l
r/q
l
v
l
2
). Unfortunately, there is not enough experimental data avail-
able to give the precise location of the different boundaries to any degree of
accuracy, except when We r 0 (in this limit the boundaries are shown for Oh of
about 10
3
). Recent experiments performed in the range of Weber numbers 20
We 10
3
and 800 Re
g
8.10
3
(Farago & Chigier 1992, Lasheras et al
1998) give a good indication of the upper and lower boundaries of the liquid
membrane breakup. Good atomization, in the sense of a ne spray with uniformly
small droplets, is achieved beyond the upper boundary of the membrane breakup.
The farther to the right or above, the ner is the spray. For a given combination
of M(1) and We, the mass loading m affects the downstream spray formation
by draining kinetic energy from the gas stream. We discuss this effect in more
detail in Section 4.
2.1 Liquid Core Length, L
The momentum ux ratio M is the primary parameter determining the length of
the unbroken liquid core. When M K 1, this length is determined by the liquid
jet (Reitz & Bracco 1982), whereas for M 1 the gas jet is responsible for the
breakup (Hopnger & Lasheras 1994, Lasheras et al 1998). When M is large, the
liquid core is very short, and above a certain critical value M
c
(of about 50), a
gas cavity forms downstream of the liquid core, which is truncated by a recir-
culating motion.
The core length L can be estimated from conservation of mass uxes m
s
A
s

q
l
U
l
A
1
, where A
l
is the cross sectional area of the liquid nozzle, A
s
is the surface
area of the liquid core, and m
s
is the ux of liquid mass shed per unit area (see
for example Przelawas 1996). The problem is to determine this mass ux across
the surface A
s
. The semiempirical expression m
s
C
1
/r/q
l
]
1/3 2 2
[l (q U )
l g g
obtained using the capillary wave theory by Mayer (1961) is often used.
This gives a core (cone) length
1/3
L 1 r
. (1)
2/3
D 2C M lU
l 1 l l
This formula, which involves an adjustable constant C
1
, is in reasonably good
agreement with experiments when the physical parameters are not varied over a
wide range of conditions. However, it does not give the correct limit when the
surface tension goes to zero, as is the case in applications such as liquid propellant
rocket engines operating under supercritical conditions. Other empirical correla-
tions such as (Arai et al 1985, Eroglu et al 1991, Lin & Reitz
b
a
L/D C We Re
l 2 l
1998) with 0.3 a 0.7 and 0.5 b 0.5, have the same shortcomings.
A different expression for the core length has been proposed by Lasheras et
al (1998). Since the most unstable wavelength measured in the experiments is
very large (of the order of the jet diameter; see Figure 2), they argued that surface
282 LASHERAS HOPFINGER
tension does not have any effect on the interfacial instability when the Weber
number is large. Using this fact, they were able to generalize the simple entrain-
ment model proposed by Rehab et al (1997) to estimate L. This model is based
on dynamic pressure continuity at the interface in the form
2 2
q u C q u , (2)
l e e g g
where is the r.m.s. velocity in the mixing layer (which can be approximated u
g
as 0.17 (U
g
U
l
)), u
e
is the velocity at which liquid is entrained, and C
e
is a
proportionality constant obtained experimentally (C
e
0.25) (Rehab et al 1997).
Substituting m
s
q
l
u
c
in the mass ux conservation, one gets
1
L 6 U
l
1 . (3)

D U
l g M
Liquid propellant rocket engines typically operate at M10 (Herding et al 1998),
which results in a liquid core length of about 2D
l
(note that for liquid-gas jets
M 1 also implies U
l
/U
g
K 1). On the other hand, for high-speed water jets
discharging into a stagnant or nearly stagnant gas, we have M r 0 (U
g
r 0), and
Equation 3 reduces to L/D
l
6 (q
l
/q
g
)
1/2
. In this limit the core length is indepen-
dent of the velocity, and the expression is of the same form as the one given by
Taylor (1963). For the case of U
g
U
l
, the core length is theoretically innite
(see Figure 3, color insert). The actual core length of the water jet depends on
the ratio of the thickness of the vorticity layer at the nozzle exit to the jet diameter.
When this ratio is small, the core length of a water jet in air can reach values as
large as 200D
l
(Hoyt & Taylor 1977). For larger ratios, the jet breakup is com-
pleted before the initial shear created in the nozzle has relaxed, and the core length
is considerably shorter. It is known that the core length also depends on the nozzle
geometry (Reitz & Bracco 1982, Hiroyasu et al 1996, Lin & Reitz 1998).
Recent experiments performed in a water-air shear layer (Raynal 1997) indi-
cate that Equation 3 is in good agreement with the measured core lengths. Raynal
also measured a weak dependency on U
l
, with L decreasing with increasing U
l
.
One explanation for this weak dependency on U
l
is that although surface tension
does not affect the instability, it can limit the elongation of the liquid sheet or
ligaments (3-D effect) by the effect of the surface tension force. In the case shown
in Figure 2a, for instance, we can see that this force allows hardly any breakup,
whereas in the case of Figure 2b the ligaments reach a considerably longer length
and subsequently break. To account for this effect, the dynamic pressure balance
at the interface could be modied as:
2 2
qu Br/d C q u , (4)
l e e g g
where d is the thickness of the sheet or ligament, and B is a coefcient to be
found from experiments. The problem is to formulate an expression for d. Assum-
AIR ASSISTED LIQUID JET ATOMIZATION 283
ing that the minimum thickness is given by the dynamics of the shear layer, one
can use the strain rate-diffusion balance U
g
/k d
2
/m, and U
l
/U
g
K1, to arrive at
the expression
L 6 1
. (5)
1/2
D (1 B r/l U )
l 1 g g M
The coefcient B
1
, which depends weakly on the gas Reynolds number and
weakly on Oh, is evaluated from Raynals (1997) experiments to be about 10
3
.
This expression gives the correct limit when r goes to zero (Raynal 1997). Cor-
relations of the type L/D
l
C/M
n
with n 0.3 have also been proposed by
Engelbert et al (1995).
2.2 Spray Angle
The spray angle is also a global quantity of considerable practical interest. The
growth rate of the liquid-gas shear layer is related to the liquid jet core length,
which for large gas velocities gives an angle tanc M
1/2
/6. This relation remains
valid as long as the liquid core remains cone-like, that is as long as M is less than
about 30, giving a maximum angle c of about 40. The spray angle is generally
larger than the liquid-gas shear layer angle because the inertia of the drops allows
them to escape the shear layer boundary. Raynal (1997) measured spray angles
of about 50 with respect to the liquid cone surface and this angle was found
to depend only weakly on M. The stability analysis in section 3.1 suggests a spray
angle of 45 when M
1/2
k1. Because the liquid cone angle increases with M,
the total angle h of the spray cone decreases with M from h 2( c/2)
90 to 60.
3. THE NEAR-FIELD REGION
3.1 Linear Stability Analysis of a Liquid/Gas Shear Interface
The classical stability analysis of two parallel ows with different densities has
been performed on numerous occasions to study the primary instability devel-
oping at a gas/liquid interface (Taylor 1963, Reitz & Bracco 1982, Lin & Lian
1989, Lin 1996, and many others). An extension of this analysis to the particular
case of the coaxial jets with high momentum ux ratios of interest here has been
done by Raynal (1997), and Raynal et al (1999). Raynal showed that assuming
an innitely thin shear interface gives results that are not consistent with the
frequencies and wavelengths of the instabilities measured in both two-
dimensional gas/liquid shear layers and in axisymmetric coaxial gas/liquid jets.
His stability analysis indicates that in order to properly simulate the conditions
encountered in these ows, the vorticity layers resulting from the nozzles walls
284 LASHERAS HOPFINGER
must be included. Furthermore, since the experimental observations suggested
that the instability has a long wavelength compared to the thickness of the vor-
ticity layers (see Figure 2), he assumed that the gas vorticity layer is responsible
for the instability and that the surface tension can be neglected. In other words,
he postulated that the gas stream xes both the velocity and length scales in the
problem, and that the liquid damps high frequency perturbations and only allows
large wavelengths to be amplied. Performing a classical linear stability analysis
following Chandrasekhar (1961) of a piecewise velocity prole composed of the
two free streams and a constant vorticity layer in the gas of thickness d
2
, he
calculated the amplication rate as a function of wave number for values of the
density ratio S q
g
/q
l
ranging from1 to 10
3
, and for large values of the velocity
ratio U
g
/U
l
. He showed that the preferentially amplied wavelength, as well as
its amplication rate, decreases as the density ratio is decreased. When S 1/10,
the amplication rate is proportional to the density ratio q
g
/q
l
, while the group
velocity and the selected wave number are proportional to (q
g
/q
l
)
1/2
. For the case
of very small density ratios (S 1.2 10
3
), he found that the Strouhal number
of the most amplied wave is f d
x
/U
c
7.6 10
3
, where U
c
is the group
velocity.
3.2 Experimental Results
In an attempt to clarify this interfacial instability under the conditions of interest
here (large gas momentum ux and large gas Reynolds number), Raynal (1997)
performed a two-dimensional liquid-gas shear layer experiment where he mea-
sured both the frequency and the convection velocity of the interfacial waves
under a wide range of gas and liquid velocities. He measured the frequency of
the liquid interfacial waves with a laser beam and a photo-diode and determined
the convection velocity U
c
by correlating the photo-diode signal with the signal
of a hot-lm placed at a known distance downstream of the laser beam.
The measured Strouhal number, based on the vorticity thickness, d
x
, of the
gas layer, f d
x
/U
c
8.7 10
3
, (Figure 4) was in very good agreement with
his linear stability analysis, which, as mentioned above, gives a Strouhal number
of f d
x
/U
c
7.6 10
3
. This value is much lower than the Strouhal number
f d
x
/U
c
0.13 measured in a homogeneous shear layer. The difference is due to
the much larger wavelength developing in the liquid-gas shear layer. This result
suggests that the liquid damps out the high frequency excitations imposed by the
gas stream. The measured convection velocity was found to agree with the one
proposed by Bernal & Roshko (1986) and Dimotakis (1986) for the large eddies
in a shear layer forming between two streams of unequal density U
c
. Note that if the instability under consideration ( q U q U )/( q q ) l l g g l g
here were driven by the liquid vorticity layer, the wavelength would be very short
and consequently the Strouhal number would be large. For large gas velocities,
this short wavelength instability k
2
can also develop, coexisting with the long
wavelength one k
1
(see sketch in Figure 1).
AIR ASSISTED LIQUID JET ATOMIZATION 285
Figure 4 Measured frequency of the interfacial instability, f vs U
c
/d
x
. From Raynal et al
(1999).
3.3. Primary Drop Formation Mechanisms
3.3.1 Phenomenological Description of the Near-Field Breakup When the
interfacial waves are amplied, the liquid is drawn out into sheets of thickness d
by the strain imposed by the gas stream. These sheets subsequently break into
droplets whose sizes are proportional to and of the order of d. The problem then
is to predict the thickness, d, of this sheet. Figure 5 shows a visualization (span-
wise view) of the amplication of the interfacial perturbation and breakup of the
sheet in the liquid-gas shear layer experiments (Raynal 1997). As the perturba-
tions grow and a sheet begins to form, spanwise perturbations at the rim of the
sheet also grow (compare, for instance, the left-hand side of Figure 5b with Figure
5c and 5d). In time, the sheets thickness and rim radius decrease until surface
tension forces become of the order of the aerodynamic forces. The rim is nally
disrupted by aerodynamic forces and Rayleigh instability (see Figure 5e and Fig-
ure 5f ). Behind the rim, the liquid sheet is blown out into a bag, forming a
membrane, which breaks into small drops. The mechanism shown in Figure 5 is
characteristic of the relatively small Weber numbers (the Weber number of Figure
5 is about 80 based on the liquid layer thickness at the nozzle exit), where mem-
brane breakup occurs (see Figure 3). This breakup mechanism is similar to the
scenario observed in numerical simulations by Keller et al (1984), Li (1996), and
Zaleski et al (1998).
286 LASHERAS HOPFINGER
Figure 5 Visualization of the temporal growth of the interfacial waves leading to sheet
and subsequent ligament formation, followed by the breakup into drops. U
l
0.42 ms
1
,
U
g
22 ms
1
. From Raynal (1997).
From a simple balance between the aerodynamic forces (drag) acting on the
rim and the opposing surface tension force, one learns that the drop size, d
i
d,
scales with the Weber number as , where d is the ligament
1 2
d /D We U
i l g
thickness.
When the Weber number is large and ber-type breakup occurs, it is likely
that surface tension is unimportant in determining the nal thickness of the sheet
or ligament and hence the primary drop size. The size of the bers d
f
may scale
with the streamwise vortices of the interfacial gas shear layer given by the balance
between the strain rate u
g
/x U
g
/k and viscous diffusion rate . Since k
2
m /d
g f
d
x
(q
l
/q
g
)
1/2
the size of the bers would scale as d
f
/D
l
(m
g
/U
g
D
l
)
1/2
(d
x
/
D
l
)
1/2
(q
l
/q
g
)
1/4
. With d
x
D
l
[m
g
/(D
g
D
l
)U
g
]
1/2
, for laminar conditions at the
nozzle exit, the drop size varies as . The crossover from d
i
d to
3/4
d /D U
i l g
d
i
d
f
occurs at a certain gas velocity, or Weber number, which corresponds to
the upper boundary of membrane breakup in Figure 3 (We 10
2
).
Villermaux (1998) proposed an expression for the initial drop size, valid over
the whole Weber number range, in the form where We
x
1/5 2/5
d /d We (q /q )
i x x l g
is dened with . In obtaining this expression it is supposed that the
1/2
d U
x g
rate of straining of the liquid sheet is U
i
/k, and it continues until surface-tension
forces balance the straining. This expression gives a dependency of d
i
on U
g
in
AIR ASSISTED LIQUID JET ATOMIZATION 287
the form . Villermaux further suggested that the vorticity layer grows
4/5
d U
i g
with x and replaces d
x
by the vorticity layer evaluated at the end of the potential
cone. In this case he nds . Other expressions for the initial or primary
1
d U
i g
drop size have also been proposed by Przekwas 1996.
Measurements made at the edge of the spray, where the drops are thought to
originate from the primary instability, seem to indicate a variation when
n
d U
i g
We 200 and 0.5 n 0.7 (Hopnger 1998). On the other hand, for small
We, the size of the ligaments seem to decrease inversely proportional to We (see
Figure 2). This would give support to the existence of two regimes. For large We
the drops resulting from the possible short wavelength instability k
2
(see Figure
1) are likely to scale on , with 0.5 n 2. Further measurements of
n
k U
2 g
the drop size in the near-eld are required to conrm this conjectures or guide
new models.
4. THE FAR-FIELD OF THE SPRAY
As described above, the primary instability of the gas/liquid interface leads to the
formation of liquid sheets and ligaments that break into drops. Due to the large-
scale vortical motion of the gas/gas and gas/liquid shear layers shown in Figure
1, portions of these sheets and ligaments also may detach from the liquid jet.
These liquid droplets and lumps of complex shapes are subsequently convected
downstream (see sketch in Figure 1) and undergo breakup and coalescence as
they are subjected to the stresses of the surrounding turbulent gas. For the case
of very large aerodynamic Weber numbers, this secondary breakup taking place
downstream of the liquid core may account for a considerable fraction of the
liquid atomization, thus determining the nal droplet size distribution of the spray
(Faeth 1990, Hsiang & Faeth 1992, Lasheras et al 1998).
The secondary breakup is particularly apparent in cases where the length of
the core is small (as in the condition shown in Figures 6a and 6b), and large
portions of the liquid sheets are pinched off from the liquid by the large eddies
that develop at the gas/liquid and gas/gas shear layers. This regime, shown in
Figure 6, is dominated by the dynamics of the large-scale eddies and manifests
itself as an unsteadiness in the liquid void fraction, thus the name superpulsating
mode as used by Farago & Chigier (1992) and Chigier & Reitz (1996).
4.1. Secondary Breakup Mechanisms
Following the classical decomposition of the turbulent motion into a mean plus
a uctuating component, one can express the forces s acting on the liquid particles
as the sum of a force resulting from the mean relative velocity between the droplet
and the gas (mean slip velocity), and a force due to the turbulence uctuations
of the surrounding gas. To differentiate between the breakup processes resulting
288 LASHERAS HOPFINGER
a
b
Figure 6 Instantaneous ow visualization of the break up of a liquid jet by an annular
air jet U
l
0.5m/s. a) U
g
85m/s, We 489, M 31; b) U
g
56 m/s, We 210,
M 13. From Lasheras et al (1998).
from these two effects, we refer to them as shear break up and turbulent break
up, respectively.
4.1.1. Shear Break Up When a liquid drop of size d is suddenly exposed to an
airow of constant relative speed, break up occurs if the shear Weber number
exceeds a critical value We
g
q
g
(u
g
u
d
)
2
d/r (Lane 1951). Here u
d
and u
g
refer
to the mean values of the drop and gas velocities. Depending on the value of the
shear Weber number, several types of breakup regimes exist: a bag or balloon-
type breakup (Merrington & Richardson 1947, Kennedy & Roberts 1990); a strip-
AIR ASSISTED LIQUID JET ATOMIZATION 289
ping or shear breakup (Ranger & Nicholls 1969, Borisov et al 1981); and an
explosive or catastrophic breakup (Hanson et al 1963, Reinecke & Waldman
1970). At low relative velocities, the pressure forces from the gas deform the
droplet into a disk shape, which is oriented normal to the relative velocity. At a
critical shear Weber number of about 12, the disk deforms into a balloon shape,
which is then stretched until it ruptures, producing a bimodal droplet size distri-
bution composed of very small droplets resulting from the breakup of the balloon
and very large ones from the breakup of the rim. According to Borisov et al
(1981), the bag-like breakup regime occurs in the range 12 We
s
80 and
0.2 We
s
Re
0.5
1.6. At shear Weber numbers greater than 80, one observes a
second breakup mode consisting of the stripping of ligaments that are shed off
from the surface of the liquid lump, leading to the production of very small
droplets. This regime appears in the range of 80 We
g
210
4
and 1 We
g
Re
0.5
20 (Borisov et al 1981). For Weber numbers We
g
210
4
, a catastrophic
or explosive breakup is observed in which the droplet appears to disintegrate very
quickly into a very large number of droplets with sizes much smaller than the
parent one. This third mode has been observed exclusively in the interaction of
a droplet with strong shock waves, and thus it is of no interest here. A summary
of these breakup regimes is also given by Faeth (1990). Unfortunately, beyond
the phenomenological descriptions given above, there is virtually no information
on either the breakup time or the droplet size distribution that results from each
one of these regimes. In addition, one has to apply caution in extending the above-
mentioned experiments to the situations encountered downstream of the liquid
core in coaxial jets. The drop/shock interaction experiments in which some of
these modes have been observed involve the sudden acceleration of a droplet and
its exposure to a high-temperature dense gas. This causes a Rayleigh/Taylor type
of instability at the accelerating interface and, in some cases, nucleated boiling
within the droplet, followed by rapid evaporation.
In the situation of interest here, the liquid jet is surrounded coaxially by a high-
speed air jet, and the detached sheet and ligaments (such as the ones shown in
Figure 6) are not suddenly exposed to a high-speed gas, but rather to a gas speed
much smaller than the initial gas speed. It is therefore unlikely that an explosive
breakup can occur in the coaxial airassisted atomization under consideration
where the initial gas velocity is limited to Mach number unity and the droplet
Weber number never reaches values above the initial aerodynamic Weber number,
based on the diameter of the liquid jet (1000). Even in applications such as the
atomizers in liquid propellant rocket engines, where the initial aerodynamic
Weber number can be very large (low surface tension and large densities), the
drops or liquid patches are not suddenly exposed to these conditions. Recent
measurements by Lasheras et al (1998) show that although the shear breakup is
very important in the region of several jet diameters downstream of the liquid
core length (in particular the stripping ligament mode; see Figure 6), the resulting
droplets are quickly accelerated up to the gas speed, and their Weber numbers
become very small at downstream distances greater than or equal to ten jet diam-
290 LASHERAS HOPFINGER
eters. Beyond this point, the liquid lumps traveling at approximately the same
velocity as the gas are still surrounded by a high intensity turbulent gas, and the
turbulent stresses acting on the droplets may lead to their breakup. This breakup
process continues as the liquid droplets are convected downstream to a location
where the turbulent stresses acting on the surface of the droplet become equili-
brated with the connement forces due to viscosity and surface tension. At this
point, the break up nishes and a frozen PDF is achieved. Thus, it should be
expected that, regardless of the initial break up, the maximum droplet size of the
equilibrium (or frozen) PDF achieved in the far-eld of coaxial jet sprays should
be determined by the nal turbulent breakup process.
4.1.2. Turbulent Break up When a liquid lump of size d
0
is immersed in a
turbulent gas, the turbulent stresses acting on the droplet may be larger than the
connement stresses due to both surface tension and internal viscous forces, and
the droplet breaks up into smaller ones in a characteristic time t
b
(Kolmogorov
1949, Hinze 1955). When the Reynolds number of the air ow is very large, as
in the case considered here (i.e., of the order of 10
4
to 10
5
), Kolmogorov found
that if the droplet size compared to the Kolmogorovs microscale, g, satised k
g (m
l
/m
g
)
3/4
, the effect of the viscosity inside the drop is unimportant, and the
determining factor in the turbulent break up is only the dynamic pressure caused
by the velocity change over distance of the order of the droplet diameter. The
dynamic pressure force from the turbulent motion per unit surface is ,
2
q u(d)
g
where is the mean square of the relative velocity uctuations between two
2
u(d)
points diametrically opposed on the surface of the droplet. The surface tension
force per unit area is r/d. The turbulent Weber number is then dened as
. Thus, when We
t
is greater than a critical value, (We
t
)
c
, of order
2
We q u(d) d/r
t g
1 (Clay 1944), atomization of the liquid occurs because the dynamic pressure
forces from the turbulent motion are sufciently large to overcome the conne-
ment of the surface tension (Hinze 1955).
Since the air jet is at a very high Reynolds number, at its centerline an inertial
subrange exists in which the energy spectrum of the turbulence conforms to Kol-
mogorovs hypothesis of local isotropy. In other words, the spectrum of the tur-
bulent velocity uctuations includes a range of high wave numbers (the universal
equilibrium range), which is uniquely determined by the turbulent dissipation rate
in the air. For this local isotropy to exist, the linear scale of the energy-containing
eddies must be large compared to the scale of the small energy-dissipating eddies.
For very small values of d(d g), the form of the universal function can be
obtained by dimensional analysis C(ed)
2/3
. When the residence time of
2
u(d)
the liquid in the turbulent region is longer than the breakup time, the maximum
stable droplet size, d
max
, can be obtained from the relation d
max
.
3/5 2/5
[r(We ) /q ] e
t c g
In the following we discuss in some detail the statistical description of the
droplet size distribution in the far-eld of the spray.
AIR ASSISTED LIQUID JET ATOMIZATION 291
4.2. Statistical Description of the Spray Far-Field
To analyze the process in the far-eld, one can use a statistical description of the
liquid droplets by introducing a droplet size distribution function (or density func-
tion), f (d, x, v, t), dened as the probable number of droplets per unit volume
with diameters in the range about d, located in the spatial range about the vector
position x, with a velocity in the range about v, at a time t. A Boltzmann-type
equation describing the time evolution of f (d, x, v, t) may be derived, using
arguments similar to those used in the kinetic theory of gases (Williams 1985,
Coulaloglou & Tavlarides 1977, Konno et al 1980 1983),
f
(vf ) (Ff ) (Rf ) Q Q G, (6)
x v b c
t d
where F(d, x, v, t) is the force per unit mass acting on a liquid particle of size d,
and R(d, x, v, t) is the time rate of change of its size due to evaporation or
condensation (which in general depends on the thermodynamic properties of the
liquid and on the local temperature and vapor concentration in the surrounding
gas). This statistical formulation is widely used in the chemical engineering and
combustion communities to describe the evolution of two-phase ows such as
gas/liquid systems or liquid/liquid dispersions (i.e., emulsions) and is referred to
as the population balance equation (pbe) or the spray equation. is the time Q
b
rate of increase of f due to particle breakup; is the rate of change of f due to Q
c
droplet coalescence; and G is the rate of change of f due to collisions that do not
result in coalescence (i.e., the changes in f resulting from the variation in the
velocity of the particle caused by inter-particle collisions, which did not result in
coalescence). Eliminating the velocity dependence by integrating over the whole
velocity space for the steady-state, nonvaporizing ows of interest here, the popu-
lation balance equation simply reduces to:
(vn) Q dv Q dv 0, (7)
x b c
where n(d, x) fdv is the mean number density of droplets of size d at a
location x, and v(d, x) ( vfdv)/ fdv is the mean velocity of the liquid droplet
of size d at location x. Expanding the rst term in Equation 7 yields
v n n v Q Q , (8)
x x b c
with dv, and dv. To close the problem, the use of models Q Q Q Q
b b c c
is required for Q
b
and Q
c
. In addition, the mean velocity of the liquid particles
v(d, x) must be calculated from the conservation of mass and linear momentum
between the phases, which requires information on F.
In what follows we discuss the various theories proposed to calculate Q
b
and
Q
c
and compare them with experiments. The role of the convective term
x
(vn)
in the variation of the droplet size pdf with the downstream distance is not covered
here. A discussion of this effect can be found in Lasheras et al (1998).
292 LASHERAS HOPFINGER
The time rate of change of the droplet number density n due to the turbulent
breakup can be calculated as:

Q (d, x) m(d )b(d, d )g(d )n(d )d(d ) g(d)n(d), (9)


b o o o o o
d
where g(d
0
) is the break frequency of a droplet of size d
0
(or the inverse of the
breakup time t
b
), m(d
0
) is the mean number of daughter droplets resulting from
the breakup of a parent droplet of size d
0
, and b(d, d
0
) is the size probability
density function of the daughter droplets formed from the breakage of a mother
droplet of size d
0
(Konno et al 1983). The rst term on the right-hand side of the
equation represents the rate of birth of droplets of size d due to the turbulent break
of larger droplets of size d
0
, while the second term represents the rate of death
(or depletion) of droplets of size d due to their turbulent breakup. Use of Equation
9 requires closure models for g(d
0
), m(d
0
), and b(d, d
0
).
4.2.1. Droplet Breakup Frequency, g(d
o
) In the past, a large number of models
have been proposed for the drop breakup rate, namely the drop elongation in a
shear ow (Taylor 1934); turbulent pressure uctuations (Hinze 1955); relative
velocity uctuations; and drop/eddy collision, (Coulaloglou & Tavlarides 1977
and Prince & Blanch 1990). Despite the above efforts, inconsistencies still exist
between the models and the experiments, and none of them can describe the
breakup frequency over the large range of conditions encountered in the coaxial
jet ows of interest here. Using arguments similar to kinetic theory of gases,
Prince & Blanch (1990) and Coulaloglou & Tavlarides (1977) have developed a
closure model widely used in the chemical engineering community, which
assumes that the breakup frequency g(d
0
) is the product of an eddy-drop collision
frequency and a breakup efciency, g(d
0
) h
be
kdn
e
, where the collision fre-
quency is given by , and the breakup efciency is calculated
2 2 1/2
h S (u u )
be be b e
as . The drawback of this model is that it is based on an k exp( E /Ee)
r
idealized view of the turbulence composed of an array of eddies of various sizes
and energies. In addition, it requires the use of submodels for the droplet/eddy
collision cross section h
be
, the breakup energy E
r
, and the number of eddies of
each size n
e
. This model predicts a monotonic increase of the breakup frequency
with the droplet diameter, a fact contrary to experimental evidence (Mart nez-
Bazan (1999 a,b)).
Recently, Mart nez-Bazan et al (1999a,b) have proposed a phenomenological
model for g(d
0
) based on pure kinematic arguments and applicable to the range
of conditions encountered in coaxial jets. This model gives excellent agreement
with experimental results of the turbulent breakup of bubbles and droplets
immersed in homogeneous isotropic turbulent ows such as the fully developed
region of turbulent coaxial jets. The model is applicable when the droplets are of
diameters in the inertial subrange of the turbulence (between the Kolmogorov
microscale of viscous dissipation and the integral length scale, g d
0
L
x
), and
AIR ASSISTED LIQUID JET ATOMIZATION 293
when the internal viscous deformation forces can be neglected in comparison to
the surface tension forces. Under these conditions, the only two effects acting on
the surface of the droplet are the turbulent stresses and the surface tension. When
the average deformation energy per unit volume acting on the surface of the
droplet is greater than the connement energy
2 2/3 2/3
e 1/2 qu(d ) 1/2 qde d
t o o
per unit volume due to surface tension , the droplet breaks in a time e 6r/d
s o
t
b
. In estimating the deformation energy an homogeneous turbulence ow in equi-
librium was assumed and d is an integration constant (see Batchelor 1956). This
criterion, rst proposed by Kolmogorov, denes a critical diameter d
c
(12r/
dq)
3/5
e
2/5
, below which the droplets do not break (Kolmogorov 1949, Baranaev
et al 1949, Hinze 1955). Accordingly, the average time t
b
needed to break a droplet
of this size d
c
is innite, and the breakup frequency is zero.
Mart nez-Bazan et al (1999a) postulated that the surface of any other droplet
with a diameter greater than d
c
will deform with an acceleration a
b
which, on the
average, is proportional to the difference between the deformation and the con-
nement forces acting on it:
2
d u(d) 6r
a . (10)
b
2 2
t 2d qd
b
This simple argument suggests that the break frequency depends not only on the
droplet size but also on e as
r 2
u(d) 6r 2/3
8.2(ed) 12


qd
l 2d qd
g(e, d) K , (11)
g
t d d
b
where K
g
is the only proportionality constant involved in the model (Figure 7).
The model shows that in the limit of very large droplets, d/d
c
k 1, surface
tension effects are very small and the breakup frequency decreases with the drop-
let size as g(e, d) e
1

3
d

3
, while for small droplets, d
c
d d
gmax
, it increases
rapidly with the drop diameter as g(e, d) e
3/5
(r/q)
2/3
. It also d/d 1 c
predicts a diameter d
gmax
1.63d
c
for which the breakup frequency is maximum.
Experiments conducted by Mart nez-Bazan et al (1999 a,b), and Eastwood et
al (1999) (Figure 8) show that in the range d/d
c
k 1, the breakup frequency is
well predicted by the model and increases as e
1

3
. However, no experimental con-
rmation exists of the value of d
gmax
.
4.2.2. Number and Size PDF of the Daughter Droplets Resulting from the
Turbulent Break Up Three main approaches have been used to model b(d,d
0
):
phenomenological models based on surface energy considerations (Tsouris &
Tavlarides 1994); statistical models (Coulaloglou & Tavlarides 1977, Prince &
Blanch 1990, Longuet-Higgins 1992, Novikov & Dommermuth 1997); and
hybrid models based on a combination of both (Konno et al 1983, Cohen 1991).
294 LASHERAS HOPFINGER
Figure 7 Breakup frequency g(e, d) and breakup velocity model proposed by Mart nez-
Bazan et al (1999a).
Figure 8 Comparison of the breakup frequency measured experimentally with the model
given by Equation 11. From Mart nez-Bazan et al (1999a).
AIR ASSISTED LIQUID JET ATOMIZATION 295
Among the most widely used phenomenological models based on surface energy
considerations is the one proposed by Tsouris & Tavlarides (1994). They assumed
that, upon breakup, only two daughter droplets of size d
1
and d
2
are formed whose
most probable sizes are inversely proportional to the amount of surface energy
created in the breakup process. This gives a minimum probability for the for-
mation of two particles of the same size (since their surface energy is maximum),
and a maximum probability for the formation of a pair made up of a very large
particle and its complementary very small one
e [e e(d )]
min max 1
b(d , d ) , (12)
1 0
d
0
(e [e e(d )])dd
min max 1 1
0
where the energy to form a particle of size d
1
is given by:
2
e(d ) prd
1 1
. The maximum energy e
max
corresponds to the formation of two
2
prd 2 prd
2 0
particles of equal volume d
1
d
2
. This PDF has a U-shape with a
1/3
d /2
0
minimum value when a particle of minimum diameter and a complementary one
of maximum size are formed. This model has been shown to lead to results that
are not consistent with the experimental distributions measured in stirred tanks
and other turbulent ows (Hesketh et al 1991, Sathyagal & Ramkrishna 1996,
Kostoglou & Karabelas 1998, Mart nez-Bazan et al 1999a,b, and others). In addi-
tion, an important yet unresolved issue in the model is the need to dene a cri-
terion for the value of d
min
.
Statistical distributions have also been proposed by Coulaloglou & Tavlarides
(1977), Prince & Blanch (1990), and others. Coulaloglou & Tavlarides also
assumed a binary breakup and that the probability distribution function b(d, d
0
)
is given by a normal distribution (as proposed by Valentas & Amundson 1966).
Purely statistical models have also been proposed by Longuet-Higgins (1992),
Novikov & Dommermuth (1997), and others. Longuet-Higgins proposed a simple
mechanism for the breakup process consisting of a series of random divisions of
a cubical block by a number of planes parallel to the faces of the block. In studying
the PDF resulting from the cuts performed in one, two, and three dimensions, he
obtained an innite number of possible distributions, depending on the number
of random partitions performed in the initial dimension. Adjusting the combina-
tion of the number of cuts performed in each dimensions, he showed that the
model could t various experimental results (Longuet-Higgins 1999). Novikov
& Dommermuth (1997) extended Kolmogorovs similarity arguments and pro-
posed a statistical description of the droplets in turbulent sprays using ideas of
similarity, cascade process, and innitely divisible distributions. These statistical
models can describe the droplet size PDF measured in the far-eld of the coaxial
jets.
Konno et al (1980, 1983) proposed a hybrid model based on the assumption
that three daughter droplets are produced from the breakup of one mother, m
3, and the probability of obtaining a certain combination of droplets is weighted
296 LASHERAS HOPFINGER
by a factor proportional to the energy contained in the turbulent scales of sizes
equal to the size of the daughter droplets
P E(K )E(K )E(K ). . . . E(K ), (13)
1 2 3 m
where E(E
i
) is the spectrum function of the turbulent kinetic energy. Konnos
distribution gives a low probability for combinations of very large and very small
particles and gives a maximum probability for combinations of particles of similar
sizes, a result totally opposite to Coulaloglou & Tavlarides model. Konnos
model is also decient in that the distribution of the daughter droplets is dependent
on neither the size of the mother droplets, nor the turbulent kinetic energy of the
underlying turbulence, two facts contrary to experimental observations. In fact,
Konnos distribution is practically a universal one and can be approximated by a
beta function (Konno et al 1983).
A hybrid model based on energy and entropy considerations has been proposed
by Cohen (1991). Using entropy arguments, he found that the most probable
distribution resulting from the shattering of a mother droplet is similar to the beta
function proposed by Konno (1983). However, his PDF model incorporated a
dependency with e. Although Cohens model is an elegant one, it has the draw-
back of producing an explosive breakup in which a very large number of daughter
droplets (several thousands) are formed, a fact not yet observed in any
experiments.
Mart nez-Bazan et al (1999b) have recently proposed a model for b(d, d
0
) that
assumes a cascade of binary droplet divisions (m 2) and is based on an exten-
sion of the kinematic arguments used in their breakup frequency model (Figure
9).
When a droplet of size d
0
breaks into two droplets of sizes d
1
and d
2
, the
daughter size PDF, b(d,d
0
), is assumed to be proportional to the product of the
difference in the stresses associated with the formation of a droplet of size d
1
,
Ds
1
1/2 qc(ed
1
)
2/3
6r/d
0
, and the formation of the complementary droplet
of size d
2
, Ds
2
1/2 qc(ed
1
)
2/3
6r/d
0
1 6r 1 6r
2/3 2/3
b(d, d ) K qc(ed ) qc(ed ) . (14)
0 b 1 2
2 d 2 d
o o
This model predicts a dependency of the PDF on both d
0
and e is in very good
agreement with experiments performed on the breakup of liquid droplets or bub-
bles injected into the fully developed region of high Reynolds number turbulent
jets.
Figure 10 shows a comparison of the various droplet size distribution models
discussed above. At very high values of the dissipation rate, Mart nez-Bazans
model closely resembles Konnos model, while it differs considerably for small
e. On the other hand, Tavlarides model is diametrically opposed to the other two
models.
AIR ASSISTED LIQUID JET ATOMIZATION 297
Figure 9 Dependency of the daughter size PDF, b(d, d
0
) on (a) the dissipation rate of
turbulent kinetic energy and (b) the droplet size. From Mart nez-Bazan et al (1999b).
298 LASHERAS HOPFINGER
Figure 10 Comparison of the various models proposed for the size PDF of the daughter
droplets formed from the turbulent breakup of a mother one of size d
0
.
4.2.3 Models for the Droplet Coalescence The turbulence of the surrounding
gas results in drop/drop collisions and thereby in the possibility of coalescence.
However, only a fraction of those collisions involving a sufcient amount of
kinetic energy result in a coalescence (Brazier-Smith et al 1972, Ashgriz & Poo
1990, Jiang et al 1992, and others). When two droplets collide, a thin lm of air
forms between the two colliding droplets and acts as a cushion, which may cause
them to rebound. If the time involved in a collision is sufciently large, the air
lm separating the droplets gradually thins down below a certain threshold, the
boundary between the two droplets will adhere and coalesce. Inversely, if the
turbulence of the surrounding gas is sufciently high (large e), the time involved
in the drop/drop collision is shorter than that needed for the lms drainage, and
the droplets will rebound without undergoing coalescence.
The rate of change of the droplet number density n due to coalescence can be
calculated as
AIR ASSISTED LIQUID JET ATOMIZATION 299
m/2
Q h(m m, m)k(m m, m)n(m m, m)n(m)d(m)
C
0

n(m) h(v, m)k(v, m)n(m)d(m), (15a)

0
where h(v, m) is the collision frequency of drops of volumes v and v, which
depends on the dissipation rate of turbulent kinetic energy (relative velocity
between two droplets) and the droplet number density. k(v, v) is the coalescence
efciency of collision between drops of volumes v and v. The coalescence ef-
ciency depends on the ratio between the droplets contact time and a critical
drainage time, T

. The drainage time is


l
r
2/3 2/3
T e d
q
1
(Tsouris & Tavlarides 1994, Jeffreys & Davis 1971). The contact time can be
estimated as the reciprocal of the uid velocity uctuations between two points
separated by a distance equal to the droplets diameter, or turnover time t
e

3
d
2

3
. The coalescence efciency is given by

T
k exp . (15b)

t
The maximum drop diameter for which separation is possible (coagulation will
be prevented) is given by the cut-off condition T

, which gives

t
l
8
1/4
d e . (16)
max
q
1
4.2 Experimental Results of the Far-Field Droplet Size
Distributions in Coaxial Jets
Lasheras et al (1998) have shown that the nonmonotonic variation of the droplet
size measured along the central axis of the spray in the far-eld x/D
1
10 (Figure
11) can be explained by the competition between turbulent breakup and coales-
cence effects. Assuming local isotropy and fully developed turbulence at the
central axis of the jet, the argument made is that both turbulent breakup and
coalescence depend only on the value of the dissipation rate of turbulent kinetic
energy e.
The maximum droplet size that can withstand turbulent breakup is d
c
, while the maximum droplet size above which coalescence is
3/5 2/3
(12r/dq) e
negligible is . Thus, in regions of high e (after the length of
1/4
d l /q e
max g 1
the liquid cone) turbulent breakup dominates, while further downstream, as e
quickly decays, turbulent breakup ends, and coalescence effects determine the
drop size (Shinnar 1961). Lasheras et al (1998) showed that the critical distance
300 LASHERAS HOPFINGER
Figure 11 Downstream variation of the Sauter mean diameter(d
32
) measured at the jets
centerline, U
g
140m/s and Ul 0.13, 0.20, 0.31, 0.43, and 0.55 m/s. From Lasheras
et al (1998).
downstream from the nozzle needed for the turbulent breakup to be completed
can be estimated by equating the liquid residence time t
r

x
n/U(n)
to the droplet breakup time obtained from Equation 11 as
2
x /2U
crit g
d
t . (17)
b
r
2/3
K 8.2(ed) 12
g

qd
The remaining problem is obtaining an expression for e that takes into account
the presence of the liquid mass loading. From energy conservation considerations,
Lasheras et al (1998) proposed:
3
U
g
e . (18)
D (1 m)
g
This estimate of the dissipation rate gives a value for the distance downstream
up to which turbulent breakup takes place
AIR ASSISTED LIQUID JET ATOMIZATION 301
3/5 2/5
1 r
4/5
x U , (19)
crit g
D (1 m) q
g g
and the value of d
c
at the end of the breakup region as:
2/5 3/5 3
U r
g
d . (20)
c
D (1 m) q
g g
Recent measurements of d
90%
and d
32
performed over a wide range of gas veloc-
ities and liquid mass loadings were found to compare well with the d
max
and x
crit
given by the above equations (Lasheras et al 1998). Assuming similarity distri-
butions so that the maximum critical diameter is proportional to the Sauter mean
diameter (SMD), d
c
d
32
, the above equation also shows that , with
b
d U
32 g
the exponent b varying between 6/5 to 8/5 depending on liquid mass loading.
Most of the drop size measurements reported in the literature on coaxial jet sprays
are made downstream of the liquid intact length at x/D
l
10, and they are empir-
ically correlated with a power law of the form , with the exponent b
b
d U
32 g
typically ranging from about 0.8 to 1.4 (Ingebo 1992) and occasionally even 2
(Nukiyama & Takasawas 1989, Gomi 1985). The good agreement between the
experimental measurements and the prediction of Equations 19 and 20 gives
strong support to the model. It is important to remark that the secondary atomi-
zation described above does not play a role in all regimes shown in Figure 3; it
only becomes relevant in the cases where the dissipation rate of turbulent kinetic
energy e given by Equation 18 is very large. The droplet size distribution mea-
surements of Lasheras et al were performed only for very large gas speeds and
very small liquid loadings (M 100), corresponding to very large values of e.
Downstream of x
crit
the turbulent kinetic energy of the air can no longer provide
sufcient pressure deformation forces to overcome the connement due to the
surface tension. From this location onward, droplet coalescence, droplet decel-
eration, and evaporation determine the drop size distribution. The maximum drop
diameter in this region is given by Equation 16 with the local dissipation rate
given by (x/D
g
)
4
. This gives a nearly linear dependency of d
max
on the e q
downstream distance x, which is in agreement with the measurements shown in
Figure 10. The turbulent model of Borghi (Vallet & Borghi 1998) also predict
such a nearly linear increase in this region.
5. THE EFFECT OF THE ADDITION OF SWIRL TO THE
GAS STREAM
It is well known that the addition of swirl can not only enhance the jet breakup,
but also increase the lateral spreading of the liquid droplets. In the liquid-gas
coaxial jet congurations, attempts to enhance jet breakup have generally been
made by swirling the liquid jet. At rst sight this seems a logical approach con-
sidering the large density of the liquid compared with the gas. The importance of
302 LASHERAS HOPFINGER
the momentum ux ratio in liquid jet breakup discussed in Section 2 suggests,
however, that when M 1 the addition of swirl to the gas stream is the optimal
method to produce an enhancement of the liquid breakup. Hopnger & Lasheras
(1996) and Hardalupas & Whitelaw (1998) showed that a swirl in the gas jet is
indeed very effective in breaking up the liquid jet when a critical swirl number
is exceeded. The swirl number is usually dened by the ratio of axial ux of
angular momentum to axial linear momentum ux, which in this case is closely
approximated by S U
t
/U
g
, where U
t
is the tangential gas velocity at the nozzle
exit and U
g
is the axial velocity as above. The effect of a gas swirl is clearly
apparent by comparing the ow visualizations shown in Figures 12a and 12b.
These conditions fall within the membrane breakup regime shown in Figure 3.
Thus, large membrane surfaces are clearly visible in Figure 12b. The value of S
1.27 is somewhat above the critical value S
cr
required for the onset of the
formation of a central recirculating ow (similar to vortex breakdown). For the
conditions of M 2.2, S
cr
1 considered, a central recirculating ow is estab-
lished, and a stagnation point on the central axis of the jet appears. In the exper-
iments conducted with the geometry indicated above, Hopnger & Lasheras
(1996) found that the S
cr
decreases when M increases and reaches an asymptotic
value of about 0.4 at large M.
When the depression on the jets centerline caused by the azimuthal motion
of the gas becomes equal to the dynamic pressure of the uid mixture existing
on the axis, a ow reversal occurs, and a phenomenon similar to a vortex break-
down takes place. Using this criterion, Hopnger & Lasheras obtained an expres-
sion for the critical swirl number in the form
1/2 1/2
(1 ) M
2
S 1 r , (21)
cr 1 2 2
r 2M ln(D /D ) r
1 2 1 g
where r
g
and g denote respectively the ratios of axial air and liquid velocities at
the jets nozzles and at the center line at the stagnation point location, r
g

U
g
/u
g
(z), r
l
U
l
/u
l
(z), and is the local value of the volumetric void fraction of
the liquid.
There are two limiting regimes given by either very small or very large values
of the momentum ux ratio M. For M of order 1, we have m
. Under these conditions, and consistent with the experimental q AU/q A U k 1
l l l g 2 g
observations, one can assume that the ow reversal occurs at a downstream loca-
tion where, without swirl, the liquid core would remain unbroken ( 1 at the
center), and the liquid velocity is close to U
l
(r
l
1). The critical swirl number
(Equation 21) is then given by
1/2
S [2M ln(D /D)] . (22)
cr g l
In the other limiting case of M k 1 (corresponding also to small liquid to air
mass ow rates m K 1), the experimental observations show that prior to the
ow reversal, the liquid core is quickly atomized by the high-speed coaxial gas
AIR ASSISTED LIQUID JET ATOMIZATION 303
Figure 12 Images of water jet breakup with and without swirl of the coaxial gas jet for
U
l
0.55 m s
1
and U
g
22 m s
1
, M 2.2. (a), S U
l
/U
g
0; (b), S 1.27. D
l
7.6 mm, D
g
11.3 mm.
jet. Thus, in this case one can assume m K 1, and the leading term in
Equation 21 is now
2 1/2
S [2r ln(D /D)] , (23)
cr g g l
indicating that S
cr
is independent of M. Estimating r
g
2.5, the above equation
gives S
cr
0.4 for the conditions studied by Hopnger & Lasheras.
304 LASHERAS HOPFINGER
Figure 14 Normalized radial distribution of the droplet SMD (d
32
) measured at x/D
20 for various swirl numbers S. From Preaux et al (1998).
When the swirl number is above the critical value, the spray is composed of
three regions: an external region I made up mainly of large droplets following
ballistic trajectories; an inner region II dominated by a recirculating ow where
the SMD (d
32
) is minimum; and an intermediate hollow conical region where the
droplet size distribution is bimodal (Figure 12). The SMD in the outer region is
found to be only slightly increased with the swirl number, while the sizes in the
droplets in the inner region are strongly affected by the degree of swirl.
Recent measurements of the radial variation of droplet size distribution by
Preaux et al (1998) show that the addition of swirl has practically no effect on
the drop size measured at the periphery of the spray, but reduces the mean drop
size at the center (Figure 13, see color insert). The main advantages of a swirl is
then to produce (a) a central recirculation zone populated by small size droplets,
a feature with very important implications in combustion, (b) a more rapid
decrease of axial velocity associated with a large spreading angle of the spray
(increased by more than a factor of two), and (c) a more complete atomization
(absence of liquid blobs).
In the case of high air-to-water momentum ux ratio (M 50), Preaux has
also shown that the addition of relatively small amounts of swirl to the outer air
ow results in a considerable enhancement of the lateral spreading of the spray.
However, when the swirl number is above a critical value, he also found that the
AIR ASSISTED LIQUID JET ATOMIZATION 305
droplet size distribution and the liquid void fraction exhibit stronger radial inhom-
ogeneities throughout the spray (Figure 14). The degree of non-uniformity
depends on the swirl number.
The above results indicate that the primary atomization is the mechanismmost
likely to be responsible for the atomization of the droplets in the outer region,
while turbulent breakup appears to play a major role in the intermediate region.
Although this intermediate region is pushed outward with increasing swirl, the
mean size of the droplets existing there was found to be independent of the swirl
number, a result consistent with the model shown in Section 4.
ACKNOWLEDGMENTS
Many ideas and results presented in this paper must be shared with our colleagues
E. Villermaux, L. Raynal, C. Mart nez-Bazan, and J. Montanes. The work was
nancially supported by the SEP (Societe Europeenne de Propulsion) under con-
tract n 910023 (EJH), and by the US Ofce of Naval Research under contract
ONR#N00014-96-1-0213 (JCL).
Visit the Annual Reviews home page at www.AnnualReviews.org.
LITERATURE CITED
Arai M, Shimizu M, Hiroyasu H. 1985.
Breakup length and spray angle of high
speed jets. Proc. 3rd Int. Conf. Spray Sys-
tems, pp. IB/4/1-IB/4/10
Ashgriz N, Poo JY. 1990. Coalescence and sep-
aration in binary collisions of liquid drops.
J. Fluid Mech. Vol. 221, pp. 183204.
Baranaev Y, Tevenovskiy N, Treguboval EL.
1949. On the measurements of minimal
uctuations in turbulent ows. Dokl. Akad.
Novk. SSSR, 66(5) 8212424
Batchelor GK. 1956. The Theory of Homoge-
neous Turbulence. Cambridge, UK: Cam-
bridge Univ. Press
Bernal LP, Roshko A. 1986. Streamwise vortex
structure in a plane mixing layer. J. Fluid
Mech. 170:499525
Borisov A, Gelfand B, Natanzon M, Kossov
O. 1981. Droplet break regimes and criteria
for their existence. J. Engin. Phys. 40(1)
Brazier-Smith PR, Jennings SG, Latham J.
1972. The interaction of falling drops: coa-
lescence. Proc. R. Soc. London, A326:393
408
Chandrasekhar S. 1961. Hydrodynamic and
Hydromagnetic Stability. New York: Dover.
Chigier N, Reitz RD. 1996. Regimes of jet
break up and break up mechanisms. Recent
advances in spray combustion: spray atom-
ization and droplet burning phenomena.
Progress in Astronautics and Aeronautics.
Volume 166. pp 109136
Clay PH. 1940. Proc. R. Acad. Sci. Amster-
dam, The Netherlands 43:852
Cohen RD. 1991. Shattering of a liquid drop
due to impact. Proc. R. Soc. London, A
435:483503
Coulaloglou CA, Tavlarides LL. 1977.
Description of interaction processes in agi-
tated liquid-liquid dispersions. Chem.
Engin. Sci. 32:128997
Dimotakis PE. 1986, Two-dimensional shear
layer entrainement. Am. Inst. Aeronaut.
Astronaut. J. 24(11):179196
Eastwood E, Lasheras JC. 1999. Effect of the
density and viscosity ratio on the breakup
frequency of a droplet injected into a tur-
bulent ow. Phys. Fluids. To be submitted
306 LASHERAS HOPFINGER
Engelbert C, Hardalupas Y, Whitelaw JH.
1995. Breakup phenomena in coaxial air-
blast atomizers. Proc. R. Soc. London A
451:189229
Eroglu H, Chigier N, Farago Z. 1991. Coaxial
atomizer liquid intact lengths. Phys. Fluids
A3(2):3038
Faeth GM. 1990. Structure and atomization
properties of dense sprays. Int. Symp. Com-
bustion. pp. 134552. Pittsburgh, PA: Com-
bustion Institute
Farago Z, Chigier N. 1992. Morphological
classication of disintegration of round jets
in a coaxial airstream. Atomization Sprays,
2:13753
Gomi H. 1985. Pneumatic atomization with
coaxial injectors. NAL-TR-888T, N 86-
27595
Hanson AR, Domich EG, Adam HS. 1963.
Shock tube investigation of the break-up of
drops by air blast. Phys. Fluids 6:107080
Hardalupas Y, Whitelaw JH. 1998. Coaxial air-
blast atomizers with swirling air stream.
Recent advances in spray combustion: mea-
surements and model simulation Vol. II.
Progress in Astronaut. Aeronaut. Am. Inst.
Aeronaut. Astronaut. 171:20132
Herding G, Snyder R, Rolon C, Candel S.
1998. Investigation of cryogenic propellant
ames using computerized tomography of
emission images. J. Propulsion Power
13(2):14651
Hesketh RP, Etcgells AW, Russell TWF. 1991.
Bubble break up in a pipe ow. Chem.
Engin. Sci. 46:(1):19
Hinze JO. 1955. Fundamentals of the hydro-
dynamic mechanism of splitting in disper-
sion processes. Am. Inst. Chem. Engin. J.
28995
Hiroyasu H, Arai M, Shimizu M. 1996.
Breakup length of a liquid jet and internal
ow in a nozzle. Recent advances in spray
combustion: spray atomization and droplet
burning phenomena. Progress in Astronaut.
Aeronaut. Am. Inst. Aeronaut. Astronaut.
166:17384
Hopnger EJ. 1998. Liquid jet instability and
atomization in a coaxial gas stream. Proc.
Eur. Turbulence Conf. 7, ed. U. Frisch,
Advances in Turbulence VII. Kluwer Aca-
demic. In press
Hopnger E, Lasheras JC. 1994. Break-up of
a water jet in a high velocity co-owing air.
Proc. Sixth Int. Conf. Liquid Atomization
and Spray Systems, ed. Yule & Dumouche,
pp. 110117. New York: Begell House
Hopnger EJ, Lasheras JC. 1996. Explosive
breakup of a liquid jet by a swirling coaxial
gas jet. Phys. Fluids. 8:169698
Hoyt JW, Taylor JJ. 1977a. Waves on water
jets. J. Fluid Mech. 83:
Hoyt JW, Taylor JJ. 1977b. Turbulence struc-
ture in a water jet discharging in air. Phys.
Fluids 20(10):120
Hsiang L-P, Faeth GM. 1992. Near-limit defor-
mation and secondary break up. Int. J. Mul-
tiphase Flows. 18(5):63552
Ingebo RD. 1992. Effect of gas mass ux on
cryogenic liquid break-up. Cryogenics
32(2):19193
Jeffreys GV, Davies GA. 1971. Coalescence of
liquid droplets and liquid dispersion.
Recent Advances in Liquid-Liquid Extrac-
tion, ed. C. Hanson, p. 495. New York:
Pergamon
Jiang YJ, Umemura A, Law CK. 1992. An
experimental investigation on the collision
behavior of hydrocarbon droplets. J. Fluid
Mech. 234:17190
Keller FX, Li J, Vallet A, Vandromme D,
Zaleski S. 1984. Direct numerical simula-
tion of interface breakup and atomization.
Proc. Sixth Int. Conf. Liquid Atomization
and Spray Systems, ed. Yule & Dumouchel.
New York: Begell House
Kennedy JB, Roberts J. 1990. Rain ingestion
in gas turbine engines. Proc. 4th Annu.
Conf. Instabilities of Liquid Atomization
Spray Systems, pp 154162. Hartford, CT.
Kolmogorov AN. 1949. On the disintegration
of drops by turbulent ows, Dokl. Akad.
Nank. SSSR, 66:82528
Konno M, Aoki M, Saito S. 1983. Scale effect
on breakup process in liquid-liquid agitated
tanks. J. Chem. Engin. Japan 16(4):31319
AIR ASSISTED LIQUID JET ATOMIZATION 307
Konno M, Matsunaga Y, Arai K, Saito S. 1980.
Simulation model for break-up process in
an agitated tank. J. Chem. Engin. Japan
13(1):6773
Kostoglou M, Karabelas AJ. 1998. On the
attainment of steady state in turbulent pipe
ows of dilute suspensions. Chem. Engin.
Sci. 53(3):50513
Lane WR. 1951. Shatter of drops in stream of
air. Ind. Eng. Chem. 43(6):131217
Lasheras JC, Villermaux E, Hopnger EJ.
1998. Break-up and atomization of a round
water jet by a high speed annular air jet. J.
Fluid Mech. 357:35179
Lefebvre AH. 1989. Atomization and Sprays.
New York: Hemisphere
Li J. 1996. Resolution numerique de lequation
de Navier-Stokes avec reconnection
dinterfaces: Methode de suivi de volume et
application a` latomization. The`se de Doc-
torat, Univeriste Pierre et Marie Curie, Paris
Liang PY, Ungerwitter RJ. 1996. Modeling and
atomization of secondary breakup fromrst
principles. Recent advances in spray com-
bustion: spray atomization and droplet
burning phenomena. Vol. I. Prog. Astro-
naut. Aeronaut. 166:481504
Lin SP. 1996. Regimes of jet break up and
break up mechanisms (Mathematical
aspects). Recent advances in spray com-
bustion: spray atomization and droplet
burning phenomena. Vol. I, Prog. Astro-
naut. Aeronaut. 166:109136. Am. Inst.
Aeronaut. Astronaut.
Lin SP, Lian ZW. 1989. Absolute instability of
a liquid jet in a gas. Phys. Fluids 32:120
26
Lin SP, Reitz RD. 1998. Drop and spray for-
mation from a liquid jet. Annu. Rev. Fluid
Mech. 30:85105
Longuet-Higgins M. 1992. The crushing of air
cavities in a liquid. Proc. R. Soc. London.
A. 439:61126
Mart nez-Bazan C, Montanes JL, Lasheras JC.
1999a. On the break up frequency of an air
bubble injected into a fully developed tur-
bulent ow. J. Fluid Mech. 401:157182
Mart nez-Bazan C, Montanes JL, Lasheras JC.
1999b. On the size pdf resulting from the
break up of an air bubble immersed into a
turbulent liquid ow. J. Fluid Mech.
401:183207
Mayer E. 1961. Theory of liquid atomization
in high velocity gas streams. ARS
31(12):178385
Merrington A, Richardson EG. 1947. The
break up of liquid jets. Proc. Phys. Soc.
59:1
Novikov EA, Dommermuth DG. 1997. Distri-
bution of droplets in a turbulent spray.
Phys. Rev. E. 56(5):547982
Nukiyama GE, Takasawa Y. 1989. Experi-
ments on the atomization of liquids by
means of air streams, parts III and IV.
Transact. Soc. Mech. Eng. Japan 5(18):63
75
Plateau J. 1873. Stattique Experimentale et
Theoretique des Liquids Soumie Aux Seules
Forces Moleicularies. Paris: Chanthier
Vallars
Przekwas AJ. 1996. Theoretical modeling of
liquid jet and sheet breakup processes. In
Recent Advances in Spray Combustion:
Spray Atomization and Droplet Burning
Phenomena. Prog. Astronaut. Aeronaut.
166:21139.
Preaux G, Lasheras JC, Hopnger EJ. 1998.
Atomization of a liquid jet by a high
momentum coaxial swirling gas jet. Proc.
3rd Int. Conf. Multiphase Flow, Lyon.
France.
Prince WJ, Blanch HW. 1990. Bubble coales-
cence and breakup in air-sparged bubble
columns. Am. Inst. Chem. Eng. J.
36(10):148999
Ranger R. and Nicholls JA. 1969. The aero-
dynamics of shattering of liquid drops. Am.
Inst. Aeronaut. Astronaut. J. 7:28590
Rayleigh L. 1879. On the instability of jets.
London Math. Soc. 10:36171
Raynal L. 1997. Instabilite et entrainement a`
linterface dune couche de melange
liquide-gaz. The`se de Doctorat, Universite
Joseph Fourier, Grenoble.
308 LASHERAS HOPFINGER
Raynal L, Villermaux E, Hopnger EJ. 1999.
Primary instability of a plane liquid-gas
shear layer. J. Fluid Mech. Submitted
Rehab H, Villermaux E, Hopnger EJ. 1997.
Flow regimes of large velocity ratio coaxial
jets. J. Fluid Mech. 345:35781
Reinecke WG, Waldman GD. 1970. A study of
drop break up behind strong shocks and
applications to ight. AVCO Report A VSD-
0110-70-77
Reitz RD, Bracco FV. 1982. Mechanism of
atomization of a liquid jet. Phys. Fluids,
25:173042
Sathyagal AN, Ramkrishna D. 1996. Droplet
breakup in stirred dispersions. Breakage
functions from experimental drop-size dis-
tributions. chemical engineering sciences.
51(9):137791
Scardovelli R, Zaleski S. 1999. Direct numer-
ical simulation of free surface and interfa-
cial ows. Annu. Rev. Fluid Mech. 33:567
603
Shinnar R. 1961. On the behavior of liquid dis-
persions in mixing vessels. J. Fluid Mech.
10:25975
Taylor GI. 1934. The formation of emulsions
in denable eld of ow. Proc. R. Soc. Lon-
don, Ser. A 146:501
Taylor GI. 1963. Generation of ripples by wind
blowing over a viscous liquid. In The Sci-
entic Papers of Sir G. I. Taylor, ed. G. K.
Batchelor, 3:24454 Cambridge, UK: Cam-
bridge Univ. Press
Tsouris C, Tavlarides LL. 1994. Breakage and
coalescence models for drops in turbulent
dispersions. Am. Inst. Chem. Eng. J.
40(3):395406
Valentas KJ, Amudson NR. 1966. Breakage
and coalescence in dispersed phase sys-
tems. Eng. Chem. Fundam. 5:53342
Vallet A, Borghi R. 1998. An Eulerian model
of the atomization of a liquid jet. Proc. 3rd
Int. Conf. on Multiphase Flows. ICMF98.
Lyon, France, June 812
Villermaux E. 1998. Mixing and spray forma-
tion in coaxial jets. J. Prop. Power 14(5):
Weber C. 1931. Zum zerfall eines ussigkeits-
trahles (on the disruption of liquid jets). Z.
Angew. Math. Mech. 2(2)
Wierzba A, Takayama K. 1987. Experimental
investigations of liquid droplet breakup in
a gas stream. Rep. Inst. High Speed Mech.
53(382):293
Williams FA. 1985. Combustion Theory.
Menlo Park, CA: Benjamin/Cummings.
2nd ed.
Zaleski S, Li J, Succi S. 1998. Two-dimen-
sional Navier-Stokes simulation of defor-
mation and breakup of liquid patches. Phys.
Rev. Lett. 75:24447
LASHERAS
I
HOPFINGER C-1
F
i
g
u
r
e

1
3
L
i
q
u
i
d

v
o
i
d

f
r
a
c
t
i
o
n

a
n
d

d
r
o
p
l
e
t

s
i
z
e

d
i
s
t
r
i
b
u
t
i
o
n
s

i
n

a

c
o
a
x
i
a
l

s
p
r
a
y

w
i
t
h

a

s
w
i
r
l

n
u
m
b
e
r

a
b
o
v
e

t
h
e

c
r
i
t
i
c
a
l
.
F
r
o
m

P
r
e
a
u
x

e
t

a
l

1
9
9
8
.
Copyright of Annual Review of Fluid Mechanics is the property of Annual Reviews Inc. and its content may
not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.

You might also like