Sheet Breakup Yo Lozano
Sheet Breakup Yo Lozano
Sheet Breakup Yo Lozano
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/232282920
CITATIONS READS
44 476
4 authors, including:
Some of the authors of this publication are also working on these related projects:
Manufacture, integration and advanced control methods for a combined heat and power unit based on a
high temperature PEM fuel cell and its application View project
Design and manufacture of a fuel cell-based power plant for a light unmanned aerial vehicle View project
All content following this page was uploaded by Antonio Garcia-Olivares on 04 October 2017.
tent independent of the liquid viscosity and surface tension; where x is a position vector from the origin to any arbitrary
however, these properties do affect the ultimate drop size field point, and the potential function is a harmonic func-
distribution. tion for incompressible flows, chosen to satisfy boundary
ui 共 x,t 兲 ⫽⫺
1
4
冕 s
共 x⫺r兲 ⴛ关 ␥ⴛ共 âⴛb̂ 兲兴
兩 x⫺r兩 3
da•db. 共8兲
n̂ⴛ␥⫽ 冕 /2
⫺ /2
dn. 共5兲
always tangent to the interface surface the following equa-
tion can be derived:
face. If the sheet thickness tends to zero, the vorticity field is where P̃ is the projection tensor defined as
nonzero only at the interface surface and can in accordance
to Eq. 共5兲 be represented by an integral over the three- P i j ⫽ ␦ i j ⫺n i n j . 共11兲
dimensional Dirac delta function
H is the local mean surface curvature and A is the Atwood
共 x,t 兲 ⫽ 冕 S
共 n̂ⴛ␥兲 ␦ 共 x⫺r共 a,b 兲兲 ds, 共7兲
density ratio:
A⫽ 共 1 ⫺ 2 兲 / 共 1 ⫹ 2 兲 . 共12兲
2190 Phys. Fluids, Vol. 10, No. 9, September 1998 Lozano, Garcı́a-Olivares, and Dopazo
Alternative derivations of Eq. 共10兲 are possible. The value produces a damping of the small scale roll-up of the
present one is valid for any point at the mathematical vortical structures 共although maintaining the large scale evo-
vorticity-containing interface and for this reason it does not lution兲. However, the program produced nonphysical results
include the contribution due to material-element surface area due to the growth of spurious perturbations whenever ␦ was
changes ␥(“–ui ) which appears in Wu.17 This contribution assigned a value below the size of a cell.
must be included in any equation for the strength associated Additionally, an artificial viscosity of the form
to a material surface element. On the other hand, Eq. 共10兲 is
correct for our purposes, since the interface is modeled as a
set of Lagrangian mesh cells, each carrying the circulation of
w i ⫽⫺c 共 da 兲 2 冏 冏
␥i ␥i
a a
⫺c 共 db 兲 2
␥i ␥i
b b
冏 冏
, 共14兲
Froude Number: Fr⫽ ␥ 0 / 冑AgL, 共13兲 and, hence, terms involving spatial gradients of the form
( ␥•ⵜ) can be written as
再
due to the presence of surface tension and gravity effects.
Note however that the liquid sheet break up process does not 1
require the presence of gravity, and that similar results are 共 ␥•ⵜ 兲 ⫽ 关共 ␥•â 兲 ⫺ 共 ␥•b̂ 兲共 â•b̂ 兲兴
共 âⴛb̂ 兲 2
a
observed atomizing horizontal7 or vertical10 liquid sheets, be-
cause the perturbation wavelength is usually quite small 共on
the order of 1 mm兲. This considered, gravity effects have
been neglected in the simulations presented in this work.
⫹ 关共 ␥•b̂ 兲 ⫺ 共 ␥•â 兲共 â•b̂ 兲兴
b
冎 . 共16兲
Initial conditions FIG. 2. Initial (t⫽0) and final (t⫽3.0) stages of a liquid sheet simulation,
only with a streamwise sinusoidal perturbations. As in the rest of the figures,
To trigger the instability growth, the flow was initially the actual computation domain has been replicated twice in each direction.
perturbed. From the multiple possibilities for this initial per-
turbation, an asymmetric sinusoidal perturbation applied to
the spatial field in the streamwise direction was chosen. The RESULTS AND DISCUSSION
amplitude was taken to be one-half of the sheet thickness, Longitudinal perturbation
corresponding to 10% of the perturbation wavelength. Cases
Figure 2 shows the initial and final (t⫽3.0) stages of a
with an amplitude ten times smaller have also been calcu-
calculation when the flow is initially perturbed only in the
lated. A second sinusoidal spatial perturbation with half the
streamwise direction. As in the rest of the figures that will be
amplitude and the wavelength was introduced in the span-
presented, the image corresponds to the computation domain
wise direction. It has to be noted that if the liquid sheet is
replicated twice in each direction (60⫻60 grid points兲. The
considered to be infinite, the only relevant spatial scales are
value of the Atwood number is A⫽0.99 corresponding to the
the perturbation wavelengths and amplitudes and the sheet
water/air case, and surface tension has been neglected. It can
thickness. Our wavelength to thickness ratio choice would
be seen that the model reproduces correctly the expected
correspond to the wave number of maximum growth rate for
evolution of the Kelvin–Helmholtz instability as described in
a Weber number based on the sheet thickness of 1000, as
previously reported 2D simulations.11,12 When entering the
predicted by linear perturbation theory.12 To ensure that the
nonlinear deformation regime, points far away from the axis
results were not imposed by the initial conditions, other am-
of symmetry, which will be denoted as maxima, accelerate
plitudes were also tested. In particular, cases were calculated
following the surrounding air, while points close to the axis,
starting with one-half and one-tenth of the original amplitude
which will be denoted as minima, move more slowly. At the
of the transverse perturbation, maintaining the initial ampli-
same time, there is vorticity advection from minima to
tude of the longitudinal wave.
maxima, resulting in the generation of rollers in the maxima
In the second geometry described in the previous sec-
points that cause the sheet to convolute. As time evolves, the
tion, starting with an initial mesh of equally sized cells, the
vortex centers assume a sawtooth configuration. As the vor-
limited number of grid points yielded a sheet aspect ratio of
tices rotate, the sheet grows thinner at the initial minima
14, quite thicker than the sheets used in reported experi-
locations. In the final stage, the thinning tends to a limit
ments. To reproduce the aspect ratio of these experiments
where the upper and lower interfaces finally touch. As the
maintaining equally sized cells and a number of nodes at the
sheet thickness tends to zero in these points, any perturbation
sheet edges capable of yielding enough resolution would
in a real case 共e.g., acoustic noise, pressure wave, etc.兲 would
have implied an excessive computation time. The contribu-
cause the sheet to tear. The tear would generate a hole with
tion of edge effects can, however, be observed even for this
regions of high curvature, where the effects of surface ten-
thickness.
sion would be very intense. This mechanism explains the
The initial vortex sheet strength was normalized to 1,
generation of spanwise ligaments, oriented parallel to the
and was defined to have only downstream component, being
nozzle.
always constrained to be tangent to the air/water interface.
Defining an initial geometry and vortex sheet strength is
Transverse perturbation
enough to start the calculation of the interface evolution.
Time has been made nondimensional with a factor T The presence of longitudinal filaments cannot be ex-
⫽/ ␥ 0 . plained if a transverse perturbation is not included. To start
2192 Phys. Fluids, Vol. 10, No. 9, September 1998 Lozano, Garcı́a-Olivares, and Dopazo
FIG. 3. Initial (t⫽0) and final (t⫽3.0) stages of a liquid sheet simulation, FIG. 4. Initial and final stages (t⫽3.0) of a liquid sheet simulation, with
with antisymmetric spanwise and streamwise sinusoidal perturbations. symmetric spanwise and antisymmetric streamwise sinusoidal perturbations.
the growth of these structures, the sheet was initially altered similar characteristics, although the contact points are dis-
by symmetric and antisymmetric sinusoidal perturbations, placed half a wavelength. Tearing at the touching points can
with an amplitude of 25% of the sheet thickness, correspond- be expected causing tubular sections with pointed edges. The
ing to 5% of the wavelength. Figure 3 presents the initial and high curvature of these edges will be quickly smoothed by
final (t⫽3.0) stages of a case with antisymmetric perturba- surface tension forces, originating the cylindrical streamwise
tions both in the longitudinal and transverse directions, filaments observed in the experiments. It has to be pointed
where surface tension has been included. In Fig. 4 the initial out that according to this theory streamwise filaments would
transverse perturbation is symmetric. The evolution of the appear if there is a transverse wave, either if it is symmetrical
longitudinal perturbation is essentially similar to that de- or antisymmetrical. Its wavelength and amplitude would de-
scribed in the previous section. Both air/liquid interfaces end termine the filament spacing and diameter. Note also that the
up touching each other. However, in this case, the first con- surface undulation causes the two interfaces to touch earlier,
tact does not occur simultaneously on a whole line transverse and at points closer to the roller crests.
to the sheet, but in single points of this line. Figure 5 shows Different values for the initial spanwise perturbation am-
the evolution of a plane perpendicular to the liquid sheet at plitude were tested 共1/2 and 1/10 of the nominal case兲. It is
the location where contact between the interfaces first hap- observed that the transverse perturbation grows at a slower
pens for the simulations presented in Fig. 3 共antisymmetric rate than the longitudinal one whose initial growth is expo-
transversal perturbation兲 and Fig. 4 共symmetric perturba- nential. This is due to the fact that both air and water veloci-
tion兲. In both cases, the upper surface has been notably flat- ties are initially oriented in the longitudinal direction. By
tened due to the higher stretching. The cross sections share itself, the transverse perturbation growth rate might be insuf-
FIG. 5. Evolution of a section of the liquid sheets presented in Fig. 3 and Fig. 4 perpendicular to the plane where contact between the two air/water interfaces
first occurs. Cuts correspond to times t⫽0, t⫽1.5, and t⫽3.0.
Phys. Fluids, Vol. 10, No. 9, September 1998 Lozano, Garcı́a-Olivares, and Dopazo 2193
FIG. 6. Initial and final stages (t⫽3.0) of a liquid sheet simulation with FIG. 7. Initial and final stages (t⫽3.0) of a liquid sheet simulation with
antisymmetric perturbations oriented at ⫾45° with the flow direction. Note same initial perturbation of Fig. 6 plus an in-phase antisymmetric longitu-
that in this figure flow direction is towards the viewer. dinal perturbation.
ficient to generate longitudinal filaments starting from an in- phase longitudinal perturbation parallel to the vortex sheet
finitesimal initial perturbation. This could not be a problem if strength, i.e., wave numbers k⫽( , ,0), k⫽( ,⫺ ,0),
there is a mechanism that can generate finite initial trans- and k⫽(2 ,0,0), the end result is similar to that described
verse perturbations, for example inside the nozzle. However, in our nominal case 共Fig. 3兲 but with an enhanced growth
even if this is not the case, higher growth rates have been of the transverse wave. Figure 7 presents a final stage
observed when the initial perturbation is not orthogonal to (t⫽3.0) of this calculation. Figure 8 depicts the temporal
the initial vortex sheet strength. Figure 6 illustrates this evolution of a transverse section (t⫽0, 1.5, 3.0, and 4.0兲. It
point, when the initial vortex sheet strength forms an angle is to be compared with Fig. 5, considering that now the ini-
of ⫾45° with two perturbations, whose initial amplitudes in tial amplitude is five times smaller. This suggests that in a
this case were 1% of the wavelength 共5% of the sheet thick- real situation, with formation of streamwise filaments, the
ness兲. This case results in a final situation where the trans- triggering perturbation could be a superposition of different
verse section shows a sinusoidal undulation of finite ampli- waves with more than just a longitudinal and a transverse
tude. If this same initial condition is combined with an in- components. The presence of this oblique waves will con-
FIG. 8. Evolution of a section of the liquid sheet presented in Fig. 7 perpendicular to the plane where contact between the two air/water interfaces first occurs.
Cuts correspond to times t⫽0, t⫽1.5, t⫽3, and t⫽4.
2194 Phys. Fluids, Vol. 10, No. 9, September 1998 Lozano, Garcı́a-Olivares, and Dopazo
FIG. 10. Final stage of a nominal case simulation with a surface tension
coefficient multiplied by a factor of 1000.
Edge effects
The results presented in previous sections have been cal-
culated for a liquid sheet infinite both in the cross-stream and
FIG. 9. Final stages (t⫽3.0) of a liquid sheet simulation with same initial downstream directions. To study the influence of the sheet
conditions as those in Fig. 3 except that the initial separation between the
two air/water interfaces has been multiplied by a factor of 共a兲 0.25 共upper兲,
edges some simulations were performed for an infinite liquid
共b兲 0.5 共middle兲, and 共c兲 2.0 共lower兲, respectively. slab surrounded by an air co-flow. As can be seen in Fig. 11
the edges tend to bend toward the sheet causing the sacklike
structures that are observed in experiments for low air/water
tribute to the evolution of a finite transverse wave starting velocity ratios.7–10 For high ratios sheet breakup occurs too
from infinitesimal perturbations, in agreement with the close to the nozzle for these structures to be observed clearly.
mechanism proposed by Wu and Stewart.25 For an example of these structures, see for example Fig. 14
Figure 9 compares the perturbation evolution for differ- of Mansour and Chigier.7 This is not only a surface tension
ent initial separations between the two air/water interfaces. effect, as it appears even when the inverse of the Weber
Initial separations in Figs. 9共a兲, 9共b兲, and 9共c兲 were 0.25, 0.5, number is zero. Surface tension, however, causes real sheets
and 2 times larger than that depicted in Fig. 2. As the longi- to collapse in the spanwise direction into a jet, an effect that
tudinal perturbation was spanning the same number of grid is more easily observed in the absence of air co-flow.7
points in all cases, a thinner sheet is equivalent to a longer
wavelength. As predicted by linear perturbation theory,
shorter wavelength perturbations grow faster.
CONCLUSIONS ACKNOWLEDGMENTS
A three-dimensional model based on the assumption that This project has been partially supported by the Comi-
the interface behaves as an inviscid vortex sheet has been sión Interministerial de Ciencia y Tecnologı́a of the Spanish
developed to study the instability growth leading to the Government under Contracts Nos. AMB93-0917 and
breakup of a water sheet surrounded by an air co-flow. AMB96-0427-C03-01. The authors are grateful to the refer-
Transverse filaments observed in experiments can be ex- ees especially because one of their criticisms has led us to
plained with a 2D Kelvin–Helmholtz instability mechanism. review our initial considerations of the transverse instability
Longitudinal filaments which are experimentally observed growth mechanisms.
for most air/water velocity ratios leading to efficient atomi-
zation require a 3D analysis. Simulations with initial longi-
APPENDIX: DISCRETIZED EQUATIONS
tudinal and transverse sinusoidal perturbations indicate that
the sheet eventually collapses at discrete points in a trans- Notice the combination of a Mac-Cormack type algo-
verse cross section. Tearing along these points helped by rithm for the hiperbolic part of the ␥ equation and a
surface tension effects would explain the formation of the Predictor-Corrector for its (du/dt) and (dn/dt) sources. a, b
longitudinal filaments. Initial symmetric or antisymmetric are unit vectors along the non-orthogonal coordinates a, b
transverse perturbation result in very similar final configura- defined as the distances on the deformed mesh, initially car-
tions. The presence of in-phase oblique waves may explain tesian, on the surface layer. e 3 is the unit vector in the direc-
the growth of the transverse wave starting from an infinitesi- tion of the gravity field. ⌬ ⫹ a is the distance increment be-
mal perturbation. Edge effects have also been studied, to tween two adjacent mesh points in the increasing a direction,
explain the sacklike structures that can be observed for low and () m⫹1/2⬅ 关 () m ⫹() m⫹1 兴/2. Boldface characters mean 3-
air/water velocity ratios. component variables.
Predictor
␥ equation
␥*
i j ⫽ ␥ i j ⫺0.5A⌬t 共 f i, j 共 ␥ i⫹1 j ⫺ ␥ i j 兲 ⫹g i, j 共 ␥ i j⫹1 ⫺ ␥ i j 兲兲
m m m m m m m
⫺c0.5⌬t⌬a 2⫹ 冉冏 ⌬ ⫹a
冏 ⌬ ⫹a
⫹冏 ⌬ ⫺a
冏
␥ mi⫹1 j ⫺ ␥ mi j ␥ mi⫹1 j ⫺ ␥ mi j ␥ mi j ⫺ ␥ mi⫺1 j ␥ mi j ⫺ ␥ mi⫺1 j
⌬ ⫺a
冊
⫺c0.5⌬t⌬b 2⫹ 冉冏 ⌬ ⫹b
冏 ⌬ ⫹b
⫹冏 ⌬ ⫺b
冏
␥ mi j⫹1 ⫺ ␥ mi j ␥ mi j⫹1 ⫺ ␥ mi j ␥ mi j ⫺ ␥ mi j⫺1 ␥ mi j ⫺ ␥ mi j⫺1
⌬ ⫺b
冊
⫺2A⌬t
⌬u
⌬t 冉 冉
⫺ n•
⌬u
⌬t
n 冊冊 m⫺1/2
ij
⫺⌬t 共 ␥ m
i j •ⵜui j ⫺ni j • 共 ␥ i j •ⵜui j 兲 ni j 兲 ⫺2B⌬t 共 ⵜH 兲 i j ⫹2A⌬tg 共 e3⫺ 共 e3–n 兲 n 兲 i j ⫺⌬t n ␥ •
m m m m m m m
冉冉 ⌬n
⌬t 冊冊m⫺1/2
ij
,
and:
␥ **
i j ⫽ ␥ i j ⫺0.5A⌬t 共 f *
m
i, j 共 ␥ *
ij⫺␥* i, j 共 ␥ *
i⫺1 j 兲 ⫹g * ij⫺␥*
i j⫺1 兲兲
⫺c0.5⌬t⌬a 2⫹ 冉冏 ␥*
i⫹1 j ⫺ ␥ *
⌬ ⫹a
冏
i j ␥ i⫹1
冏 冏
* j⫺ ␥ *i j ␥ *i j ⫺ ␥ *i⫺1 j ␥ *i j ⫺ ␥ *i⫺1 j
⌬ ⫹a
⫹
⌬ ⫺a ⌬ ⫺a
冊
⫺c0.5⌬t⌬b 2⫹ 冉冏 ␥*
i j⫹1 ⫺ ␥ *
⌬ ⫹b
ij ␥*
冏
i j⫹1 ⫺ ␥ *
⌬ ⫹b
ij
⫹
␥*
ij⫺␥*
⌬ ⫺b
冏
i j⫺1 ␥ *
ij⫺␥*i j⫺1
⌬ ⫺b
冏⫺2A⌬t
⌬u
⌬t
⫺ n•
⌬u
⌬t
n 冊 冉 冉 冊冊 m⫺1/2
ij
⫺⌬t 共 ␥ *
i j •ⵜui j ⫺ni j • 共 ␥ *
m m
i j •ⵜu 兲 i j ni j 兲 ⫺2B⌬t 共 ⵜH 兲 i j ⫹2A⌬tg 共 e3⫺ 共 e3–n 兲 n 兲 i j ⫺⌬t n ␥ •
m m m m
冉冉 ⌬n
⌬t 冊冊 m⫺1/2
ij
,
and: ␥ m⫹1
ij ⫽( ␥ *
i j ⫹ ␥ **
i j )/2, where c is the artificial viscosity coefficient, and:
i, j ⫽
Hm
1
2 共 a∧b兲 3 det 0.5 再 冉
共 ai⫹1,j ⫺ai j ai, j ⫺ai⫺1,j
⌬ ⫹a
⫹
⌬ ⫺a
⫺0.5
⌬b ⫹
⫹ 冊 冉
ai j⫹1 ⫺ai j ai j ⫺ai j⫺1
⌬b ⫺ 冊
⫹0.5 冉 bi⫹1,j ⫺bi j bi j ⫺bi⫺1 j
⌬ ⫹a
⫹
⌬ ⫺a 冊冊 共 a–b兲 ⫹0.5 冉
bi j⫹1 ⫺bi j bi j ⫺bi j⫺1
⌬ ⫹b
⫹
⌬ ⫺b
,a,b 冊 冎m
and:
共 ⵜH 兲 m
i j⫽ 冉 a
1⫺ 共 a–b兲 2 冊再 冉
m
ij
0.5
H i⫹1,j ⫺H i j H i, j ⫺H i⫺1 j
⌬ ⫹a
⫹
⌬ ⫺a 冊
⫺ 共 a–b兲 0.5
⌬ ⫹b 冉
H i, j⫹1 ⫺H i j H i, j ⫺H i j⫺1
⫹
⌬ ⫺b 冊冎 m
⫹ 冉 b
1⫺ 共 a–b兲 2 冊再 冉 m
ij
0.5
H i, j⫹1 ⫺H i j H i j ⫺H i j⫺1
⌬ ⫹b
⫹
⌬ ⫺b 冊
⫺ 共 a–b兲 0.5
⌬ ⫹a 冉
H i⫹1,j ⫺H i j H i j ⫺H i⫺1,j
⫹
⌬ ⫺a 冊冎 m
.
u equation
i j⫽
um
1
4 兺k 兺l 再 共 xi j ⫺xkl 兲 –nkl ␥ kl⫺ 共 xi j ⫺xkl 兲 • ␥ klnkl
共共 xi j ⫺xkl 兲 • 共 xi j ⫺xkl 兲 ⫹ 2 兲 3/2 冎 m
.
x equation
xm⫹1
ij ⫽xm
i j ⫹ui j ⌬t.
m
Corrector
␥ equation
␥*
i j ⫽ ␥ i j ⫺0.5A⌬t 共 f i, j 共 ␥ i⫹1 j ⫺ ␥ i j 兲 ⫹g i, j 共 ␥ i j⫹1 ⫺ ␥ i j 兲兲
m m m m m m m
⫺c0.5⌬t⌬a 2⫹ 冉冏 ⌬ ⫹a
冏 ⌬ ⫹a
⫹ 冏 ⌬ ⫺a
冏
␥ mi⫹1 j ⫺ ␥ mi j ␥ mi⫹1 j ⫺ ␥ mi j ␥ mi j ⫺ ␥ mi⫺1 j ␥ mi j ⫺ ␥ mi⫺1 j
⌬ ⫺a
冊
⫺c0.5⌬t⌬b 2⫹ 冉冏 ⌬ ⫹b
冏 ⌬ ⫹b
⫹ 冏 ⌬ ⫺b
冏
␥ mi j⫹1 ⫺ ␥ mi j ␥ mi j⫹1 ⫺ ␥ i j ␥ mi j ⫺ ␥ mi j⫺1 ␥ mi j ⫺ ␥ mi j⫺1
⌬ ⫺b
冊
⫺2A⌬t
⌬u
⌬t
⫺ n• 冉 冉 冊冊
⌬u
⌬t
n
m⫹1/2
ij
⫺⌬t 共 ␥ m
i j •ⵜui j ⫺ni j • 共 ␥ i j •ⵜui j 兲 ni j 兲 ⫺2B⌬t 共 ⵜH 兲 i j ⫹2A⌬tg 共 j⫺ 共 j–n 兲 n 兲 i j ⫺⌬t n ␥ •
m m m m m m m
冉 冉 冊冊 ⌬n
⌬t
m⫹1/2
ij
,
and:
␥ **
i j ⫽ ␥ i j ⫺0.5A⌬t 共 f *
m
i, j 共 ␥ *
ij⫺␥* i, j 共 ␥ *
i⫺1 j 兲 ⫹g * ij⫺␥*
i j⫺1 兲兲
⫺c0.5⌬t⌬a 2⫹ 冉冏 ␥*
i⫹1 j ⫺ ␥ *
⌬ ⫹a
ij ␥*
冏
i⫹1 j ⫺ ␥ *
⌬ ⫹a
ij
⫹
␥*
ij⫺␥*
冏
i⫺1 j ␥ *
⌬ ⫺a
ij⫺␥*i⫺1 j
⌬ ⫺a
冏 冊
⫺c0.5⌬t⌬b 2⫹ 冉冏 ␥*
i j⫹1 ⫺ ␥ *
⌬ ⫹b
ij ␥*
冏
i j⫹1 ⫺ ␥ *
⌬ ⫹b
ij
⫹
␥*
ij⫺␥*
⌬ ⫺b
冏
i j⫺1 ␥ *
ij⫺␥*i j⫺1
⌬ ⫺b
冏
⫺2A⌬t
⌬u
⌬t
⫺ n•
⌬u
⌬t
n 冊 冉 冉 冊冊 m⫹1/2
ij
⫺⌬t 共 ␥ *
i j •ⵜui j ⫺ni j • 共 ␥ *
m m
i j •ⵜui j 兲 ni j 兲 ⫺2B⌬t 共 ⵜH 兲 i j ⫹2A⌬tg 共 j⫺ 共 j–n 兲 n 兲 i j ⫺⌬tn ␥ •
m m m m
冉 冊
⌬n
⌬t
m⫹1/2
ij
,
and: ␥ m⫹1
ij ⫽( ␥ *
i j ⫹ ␥ **
i j )/2.
u equation
um⫹1
ij ⫽
1
4 兺k 兺l 再 共 xi j ⫺xkl 兲 –nkl ␥ kl⫺ 共 xi j ⫺xkl 兲 • ␥ klnkl
共共 xi j ⫺xkl 兲 • 共 xi j ⫺xkl 兲 ⫹ 2 兲 3/2 冎 m⫹1
.
x equation
xm⫹1
ij ⫽xm
i j ⫹ui j
m⫹1/2
⌬t.
And finally ␥ calculated in predictor and corrector are averaged:
␥ m⫹1
ij ⫽0.5兵 ␥ m⫹1
ij 共 predictor兲 ⫹ ␥ m⫹1
ij 共 corrector兲 其 .
Phys. Fluids, Vol. 10, No. 9, September 1998 Lozano, Garcı́a-Olivares, and Dopazo 2197
1 14
W. W. Hagerty and J. F. Shea, ‘‘A study of the stability of plane fluid A. Leonard, ‘‘Computing three-dimensional incompressible flows with
sheets,’’ J. Appl. Mech. 22, 509 共1955兲. vortex elements,’’ Annu. Rev. Fluid Mech. 17, 523 共1985兲.
2
G. I. Taylor, ‘‘The dynamics of thin sheets of fluid. II. Waves on fluid 15
E. Meiburg and J. C. Lasheras, ‘‘Experimental and numerical investiga-
sheets,’’ Proc. R. Soc. London, Ser. A 253, 296 共1959兲. tion of the three-dimensional transition in plane wakes,’’ J. Fluid Mech.
3
N. Dombrowski, D. Hasson, and D. E. Ward, ‘‘Some aspects of liquid 190, 1 共1988兲.
flow through fan spray nozzles,’’ Chem. Eng. Sci. 12, 35 共1960兲. 16
4 W. T. Ashurst and E. Meiburg, ‘‘Three-dimensional shear layers via vor-
N. Dombrowski and W. R. Johns, ‘‘The aerodynamic instability and dis-
integration of viscous liquid sheets,’’ Chem. Eng. Sci. 18, 203 共1963兲. tex dynamics,’’ J. Fluid Mech. 109, 87 共1988兲.
17
5
N. K. Rizk and A. H. Lefebvre, ‘‘The influence of liquid film thickness on J. Z. Wu, ‘‘A theory of three-dimensional interfacial vorticity dynamics,’’
airblast atomization,’’ J. Eng. Power 102, 706 共1980兲. Phys. Fluids 7, 2375 共1995兲.
6
T. Arai and H. Hashimoto, ‘‘Disintegration of a thin liquid sheet in a 18
G. K. Batchelor, Fluid Dynamics 共Cambridge University Press, London,
concurrent gas stream,’’ Proc. ICLASS-85 London 共1985兲. 1967兲.
7 19
A. Mansour and N. Chigier, ‘‘Disintegration of liquid sheets,’’ Phys. Flu- G. Tryggvason, ‘‘Numerical simulations of the Rayleigh-Taylor instabil-
ids A 2, 706 共1990兲. ity,’’ J. Comput. Phys. 75, 253 共1988兲.
8
A. Mansour and N. Chigier, ‘‘Dynamic behavior of liquid sheets,’’ Phys. 20
L. Rosenhead, ‘‘The formation of vortices from a surface of discontinu-
Fluids A 3, 2971 共1991兲.
9 ity,’’ Proc. R. Soc. London, Ser. A 134, 170 共1931兲.
B. E. Stapper and G. S. Samuelsen, ‘‘An experimental study of the 21
G. Tryggvason, W. J. A. Dahm, and K. Sbeih, ‘‘Fine structure of vortex
breakup of a two-dimensional liquid sheet in the presence of co-flow air
shear,’’ AIAA paper #90-22730 共1990兲. sheet rollup by viscous and inviscid simulation,’’ J. Fluids Eng. 113, 31
10
A. Lozano, C. J. Call, C. Dopazo, and A. Garcı́a-Olivares, ‘‘Experimental 共1991兲.
and numerical study of the atomization of a planar liquid sheet,’’ Atomi-
22
J. W. Rottman and P. K. Stansby, ‘‘On the ‘␦-equations’ for vortex sheet
zation Sprays 6, 77 共1996兲. evolution,’’ J. Fluid Mech. 247, 527 共1993兲.
11
R. H. Rangel and W. A. Sirignano, ‘‘The linear and non-linear shear 23
D. Potter, Computational Physics 共Wiley, London, 1973兲.
instability of a fluid sheet,’’ Phys. Fluids A 3, 2392 共1991兲. 24
R. Aris, Vectors, Tensors and the Basic Equations of Fluid Mechanics
12
H. Q. Yang, ‘‘Interfacial instability between a liquid film and the sur- 共Prentice-Hall, Englewood Cliffs, NJ, 1962兲.
rounding compressible gas,’’ 30th Aerospace Sciences Meeting, AIAA, 25
X. Wu and P. A. Stewart, ‘‘Interaction of phase-locked modes: a new
1992.
13 mechanism for the rapid growth of three-dimensional disturbances,’’ J.
A. Leonard, ‘‘Vortex methods for flow simulation,’’ J. Comput. Phys. 37,
289 共1980兲. Fluid Mech. 316, 335 共1996兲.