Computers & Fluids Vol. 24, No. 1, pp.

l-25, 1995
Copyright 0 1995 Elsevier ScienceLtd
Pergamon 00457930(94)ooo19-0 Printed in Great Britain. All rights reserved
Ga45-7930/95 $9.50 + 0.00

NUMERICAL SIMULATION OF NON-CIRCULAR JETS

R. S. MILLER, C. K. MADNIA and P. Grvrt

Department of Mechanical and Aerospace Engineering, State University of New York at Buffalo, Buffalo,
NY 142604400, U.S.A.

(Received 3 January 1994; in revised form 16 May 1994)

Abstract-Results are presented of numerical simulations of spatially developing, three-dimensional jets

issued from circular and non-circular nozzles of identical equivalent diameters. Elliptic, rectangular and
triangular jets are considered with aspect-ratios of 1: 1 and 2: 1. Flow visualization results show that large
scale coherent structures are formed in both cornered and non-cornered jets. The axis-switching
phenomenon is captured in all non-unity aspect-ratio jets and also in the equilateral triangular jet. The
square jet does not show axis-switching; however, the rotation of its axes by 45” is shown to play a
significant role in its entrainment characteristics. All the non-circular configurations are shown to provide
more efficiemt mixers than does the circular jet; the isosceles triangular jet is the most efficient one. It is
demonstrated that the near field entrainment and mixing is characterized by the mean secondary flow
induced by the stream-wise vortices. The transport of a passive Shvab-Zeldovich scalar variable is used
to determine the limiting rate of mean reactant conversion in a chemical reaction of the type
Fuel + Air+Products. The results show that the largest product formation occurs in the isosceles
triangular jet and the lowest occurs in the circular jet. It is also shown that the 2: 1 triangular jet has the
shortest scalar core whereas the rectangular jet has the longest core.

1. INTRODUCTION

The phenomenon of mixing (or lack thereof) is a subject of crucial importance in devices involving
chemically reacting turbulent flows [l]. In these devices, the flow field produced by a “jet”
discharging into a. stagnant or moving (either in the parallel- or the cross-direction) fluid is the most
common configuration in current use. In the majority of previous investigations on turbulent jets,
circular and planar configurations have been considered [2-71; with relatively little effort on the
analyses of jets with other cross-sectional shapes [8]. Results of early investigations of three-
dimensional (3D) rectangular jets have been reported in Refs [g-11], and of elliptic jets in
Refs [12-151. In studies pertaining to elliptic jets, it is now well recognized that the deformation
of large scale vorucal structures is somewhat similar to that of an isolated elliptic ring [16, 171.This
ring is inherently unstable due to the azimuthal variation of the “nozzle” curvature which causes
a non-uniform self-induction mechanism. As a result, the ring deforms in such a way that its two
axes are interchanged. This axis-switching mechanism plays an important role in promoting mixing
by causing an increase of the entrainment as compared to that in circular and planar jets. While
the extent of recent literature on elliptic jets is growing, relatively few experimental investigations
of 3D jets with “corners” have been conducted. Gutmark et al. [8] use a one-component hot-wire
anemometry system to measure the mean and turbulent characteristics and the effect of upstream
forcing on the flow evolution of several non-circular jets. Their main conclusion is that the spatial
growth rates and the amplification of velocity fluctuations vary around the circumference of the
jet and are dependent on the initial local curvature. More recently, Gutmark et al. [ 181have studied
reactive non-circular jets by means of the Planar Laser Induced Fluorescence (PLIF) technique.
They studied the effects of sharp corners on the dynamics of vertical structures as applied to
enhancement of mixing and combustion. These results suggest that a combination of small- and
large-scale mixing in a flow is advantageous in enhancing the product formation in combustion
systems.
Most efforts in analytical treatment of non-circular jets have been based on linear stability
analyses [19-231. The extent of literature on detailed numerical simulations of 3D jet flows is very

tAuthor for correspondence.

2 R. S. MILLER et al.

limited. This is understandable in view of the severe computational resources required for such
simulations. With recent advances in supercomputer technology, however, this situation is
gradually changing [24]. Owing to this technology, it is now possible to perform “model-free”
simulations [25] of jet flows without resorting to “turbulence modeling”. Givi [25] subdivides
model-free simulations into Direct Numerical Simulation (DNS) and Large Eddy Simulation
(LES). Currently the range of physical parameters, such as the Reynolds number, that can be
treated by model free methods is significantly less than that in laboratory experiments. Such
simulations, nevertheless, have proven very effective in elucidating many important features of
turbulent flows; in some cases not easily amenable by other means.
In this work, we make use of model-free simulations to broaden our understanding of some of
the underlying mechanisms involved in the near field of jet flows originating from non-circular
nozzles. Our primary objective is to assess the influence of the nozzle shape on the subsequent
evolution of the jet flows and their mixing characteristics. This is facilitated by analyzing the
processes involved in entrainment. Six nozzle geometries are considered: circular, elliptic, square,
rectangular, and triangular (equilateral and isosceles). Simulations are of a duration sufficient to
determine statistics up to second moments. The flow fields produced by these jets are analyzed to
determine the advantages and/or the drawbacks of non-circular nozzles for mixing enhancement,
as compared with a circular nozzle. Consideration of these nozzles is motivated, at least partly,
by recent experimental findings alluding to their capabilities in facilitating efficient turbulent
combustion systems (e.g. Ref. [18]). The emphasis here is on extracting detailed information from
the numerical simulations to complement the results obtained in laboratory experiments. Details
of the geometrical configurations considered are given in Section 2. Results pertaining to
hydrodynamic transport and those for the analysis of mixing-controlled reacting flows are
presented in Section 3. A summary and conclusions are furnished in Section 4.

2. DESCRIPTION OF THE PROBLEM

The flow configurations are produced by unsteady, 3D, spatially-developing jets in the presence
of a co-flowing free-stream. The evolution of the flow is considered for several different inflow
conditions. These conditions are produced via six different nozzle configurations: circular, elliptic,
square, rectangular, triangular (equilateral and isosceles); see Table 1 and Fig. 1. The aspect-ratio
for the elliptical, rectangular, and isosceles triangular jets is 2: 1 (for the isosceles triangle the ratio
refers to that of the height to the base). The dimensions in each jet are set in such a way as to
yield the same equivalent diameter, D,. This diameter corresponds to that of a circle with an
equivalent area. The flow field is considered within the domain identified by Cartesian coordinates
x (stream-wise), and y, z (cross-stream); see Fig. 1 for the orientation of the coordinates with respect
to the cross sections of the jets.
The analysis is based on the numerical simulations of the compressible Navier-Stokes equations,
the energy conservation, and a passive scalar conservation equation with Fourier heat conduction
and Fickian diffusion assumptions. These equations are solved numerically without resorting to
any turbulence or imposed subgrid models. The fluid is assumed calorically perfect and the
magnitudes of the kinematic viscosity, thermal conductivity, and scalar diffusion coefficients are
assumed constant. The values of the Prandtl number and the Schmidt number are set to unity and
the ratio of the specific heats is set equal to 1.4. The fluid density and the temperature at the inflow
are uniform and are set to p = 1.Okg/m3 and T = 300 K, respectively. The jet exit velocity is
U, = 86.8 m/s, and the co-flowing free-stream is U, = 17.4 m/s. In all the cases, the flows are
initialized with identical Reynolds number, Re = AUD,lv, where AU = U, - U, and v denotes the

Table I. Flow configurations

RUII Configuration Aspect ratio valax i(~,hn,”
I Circle I:1 I .oo
2 Ellipse 2:1 I.51
3 Square l:I 1.23
4 Rectangle 2:1 1.59
5 Triangle l:I I .28
6 Triangle 2:1 2.00
 Numerical simulation of non-circular jets

(a) lb)
Y Y

-0
z z

if
(d)

Fig. I. Jet profiles as represented by Uh contours.

kinematic viscosi,ty. This implies Re = 800 for the jets with D, = 0.02 m. The Mach numbers at the
two streams are A4 = 0.25 and M = 0.05, respectively. This yields a convective Mach number [26]
of A4, = 0.15 which is sufficiently low to not cause significant compressibility effects.
In order to provide a measure of the extent of mixing, the transport of a conserved scalar variable
J is considered. This scalar is initialized in such a way as to yield the limiting values of 0 and 1
in the jet and in the free stream, respectively. The transport of this scalar determines the limiting
rate of reactant conversion in a binary chemical reaction of the type “Fuel + Oxidizer -+ Products.”
With the usual definition of the ShvabZeldovich variable [27], the limiting values J = 0, 1
correspond to pure fuel and to pure oxidizer, respectively. In this way, effectively, the maximum
rate of product formation of a fuel jet issuing into a co-flowing oxidizer is being simulated. With
the assumption of unity mass fractions at the feeds of each of the two respective reactants, J, = l/2
corresponds to the stoichiometric surface where, by definition, the rate of chemical reaction is
infinitely fast and the product mass fraction is unity.
Several other parameters are calibrated in order to facilitate a direct comparison of the flow field
produced by the jets. The flow is initialized by a “smooth” top-hat stream-wise velocity profile at
the jet exit. The normalized cross-stream gradient of the Shvab-Zeldovich scalar is the same as that
of the stream-wise mean velocity distribution. The velocity and the scalar gradients at the inlet are
adjusted such that all inflows have identical momentum inflow and product thickness
[6,(x) = Ssj,,(PY,,)dy dz; where ( ) indicates the time average]. The ratio of the maximum to
minimum value of momentum thickness (denoted by 0,) at the exit of the jets is listed in Table 1.
These values are in good agreement with those in the experiments [13, 121. In addition to the base
flow, low frequency perturbations are added at the inflow. The forcing is with the same Strouhal
4 R. S. MILLERet al.

number, St, = /?D,/U,= 0.4 in all the jets. The integrated perturbed momentum is held constant
for all jets. The amplitude of the velocity perturbation is set at approx. 15% of the jet exit velocity
to expedite the formation of large scale structures within the domain considered.
The computational scheme is based on an explicit time marching procedure by means of a
monotone Flux Corrected Transport (FCT) finite difference algorithm [28]. The algorithm used
here is second order accurate in time, fourth order phase accurate in space and has been successfully
employed in transitional shear layer studies [24,29]. At the outflow, the first derivatives of the
variables are assumed zero. Free slip conditions are employed at the boundaries in the cross-stream
directions. The grid configuration consists of 120 x 95 x 95 nodes for the unity aspect-ratio jets
and varies slightly for the remaining cases. In all cases, the grid is compressed at the location of
maximum mean gradients to provide a finer resolution. The computational requirements associated
with each simulation vary slightly from one simulation to the other. In total, approx. 500 h of CPU
time on a Cray-YMP supercomputer were required to complete this study. The resolution employed
here is to a large extent dictated by the available computational resources. With the magnitudes
of the physical parameters considered, this resolution is less than that required to resolve the
Kolmogorov length scale. No explicit models are employed for the closure of subgrid fluctuations;
the numerical dissipation inherent in the FCT algorithm provides the only means of modeling such
fluctuations [30]. For “debates” on the usage of an appropriate label (DNS or LES) for these
simulations we refer to Boris [31] and Givi [32].

3. RESULTS

Simulations are performed within a domain large enough to accommodate for the growth of the
jets and to minimize the effects of boundaries in the cross-stream direction. With available
computational resources, it is possible to consider a domain with L, = 9D,, and L, x L, z 4D,,
where Li denotes the length in the i-th direction. The simulated results are analyzed both
instantaneously and statistically. The instantaneous results provide an effective means of flow
visualization whereas the statistical data are useful for comparative assessments with laboratory

(4

Fig. 2. Surfaces of constant vorticity magnitude (Iwl/lolmax= 0.55). (a) Run 1, (b) Run 2.
 Numerical simulation of non-circular jets

(a)

Fig. 3. Surfaces of constant vorticity magnitude (IwJ/IwI,,, = 0.55). (a) Run 3, (b) Run 4.

(b)

Fig. 4. Surfaces of constant vorticity magnitude (IwI/Jw(,,, = 0.60). (a) Run 5, (b) Run 6.
6 R. S. MILLER et al.

data. Of course with the low value of the Reynolds number considered, it is not possible to make
quantitative comparisons with such data. Qualitative comparisons, nevertheless, are possible and
are made.

3.1. Flow visualization

A qualitative assessment of the formation and dynamics of large scale flow vertical structures
formed at the near field of the jets is possible by examining the instantaneous surfaces of constant
vorticity magnitude. Figure 2 represents an iso-surface of vorticity magnitude for the non-cornered
jets of Run 1 and Run 2. Part (a) of this figure shows that for the circular jet the growth of

(4
2.0

1.0

9% 0.0

-1.0

-2.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

2.0

1.0

“N 0.0

-1.0

-2.0 L
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

(b)

‘N j~+~I:):)._
0.0 1.0 2.0 3.0 4.0 5.0
X*
6.0 7.0 8.0 9.0 10.0

Fig.5-caption on p. 8.
 Numerical simulation of non-circular jets 7

perturbations introduced at the inflow results in circular vortex-ring-like structures [33]. The shape
and the dynamics of the structures formed in the elliptic jet are markedly different as shown in
Fig. 2(b). The structures observed in this figure show the azimuthal variation of the vorticity
magnitude. This variation is due to the initial shape of the ellipse which causes a non-uniform
self-induction velocity leading to a three-dimensional deformation. The iso-surfaces of vorticity
magnitude for Ru:ns 3 and 4 are shown in Fig. 3. This figure suggests that as the aspect ratio is
increased, the initial structure of the jet is less preserved. The triangular jets (Run 5 and Run 6)
exhibit different characteristics as observed in Fig. 4. In these jets, larger scale structures are
produced near the flat surfaces and are masked by small scale structures produced near the corners.

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 6.0 9.0 10.0
X*

2.0

Fig. 5-caption on p. 8.
8 R. S. MILLERet al.

As a result, the flows appear to be characterized by a much more “turbulent” vorticity distribution
than those produced by the other jets.

3.2. Statistical consideration

The statistical analysis of the generated data is based on an ensemble of 1200 realizations. This
is conducted within a time period equal to 2.5 times the residence time of the flow within the domain
considered. The downstream evolution of the mean stream-wise velocity profiles in both cross-
stream directions is shown in Fig. 5. All distances are normalized by the equivalent diameter

(4
2.0

1.0

‘r. 0.0

-1.0

-2.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 1
X*

J
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10
X*

2.0

1.0

‘>. 0.0

-1.o

-2.0
t 1 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 1
X*

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 1
X*
Fig. 5. Stream-wise evolution of (U ) in y - x and z - x planes. Scale: One stream-wise equivalent
diameter corresponds to 100 m/s. (a) Circular jet, (b) elliptic jet, (c) square jet, (d) rectangular jet,
(e) equilateral triangular jet, (f) isosceles triangular jet.
 Numerical simulation of non-circular jets 9

0.00
0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 6.0 8.0 10.0
X* X*
Fig. 6. Downstream evolution of the normalized mean Fig. 7. Downstream evolution of fluctuating centerline
centerline velocity for Runs l-6. velocity for Runs I-6. Legends are the same as those in
Fig. 6.

(e.g. x* = x/D,). In the circular jet [Fig. 5(a)], it is apparent that the velocity profiles in the vertical
and the horizontal center-planes are very similar. This suggests the radial symmetry of the flow
and indicates adequacy of the numerical resolution. This is not the case for the elliptic jet [Fig. 5(b)].
In this case, the width of the profile in the major plane is observed to contract for the first four
equivalent diameters, and then to expand slowly. In contrast, the width in the minor plane spreads
rapidly throughout the evolution. This trend is in good agreement with that observed in previous
investigations (e.g. Ref. [12]). An interesting feature observed in this figure is the formation of
shoulders on the minor plane velocity profiles downstream of x* = 6. This is associated with the
cross-stream mean secondary flow, as will be discussed in Section 3.3. The results for the square

()[Qooocr;o
a z

Fig. 8. Axis switching as depicted by (Us) contours. Contours are in increments of x* = 1. (a) Run 2,
(b) Run 3, (c) Run 4, (d) Run 5, (e) Run 6.
10 R. S. MILLER et al.

jet [Fig. 5(c)] show that the profiles in both planes are very similar and grow uniformly in
the stream-wise direction. Downstream of x* = 5, two additional peaks appear on the mean
velocity profiles. They are formed as a result of the induced flow field of the stream-wise vortices
(Section 3.3). No contraction of the width of the profiles is observed in either axes of the rectangular
jet [Fig. 5(d)]. For x* > 5 the width along the minor axis becomes larger than that along the major
axis. This behavior is different from that observed in the elliptic jet. The peaks in the velocity
profiles are also observed in the rectangular jet and are more pronounced in the major axis plane.
The spread of the two triangular jets, shown in Figs 5(e) and 5(f) is the most complex. The
contraction and the subsequent expansion of the profiles in the transverse direction y is observed
in both jets. The initial non-symmetry with respect to the minor axis prevails throughout the entire
stream-wise evolution in both jets.
The downstream evolution of the normalized mean centerline velocity (Uc,/U,) is shown in
Fig. 6. This figure shows that the magnitude of this velocity initially rises slightly above unity due
to forcing and then decays. The rate of decay is highest for the isosceles triangular jet. The variation
of the longitudinal fluctuating velocity along the jet centerline is presented in Fig. 7. Except for
the two triangular jets, a peak intensity occurs at x* x 4.5 for all cases. As discussed in Ref. [13]
the amplitude and location of this peak are dependent on the value of the Strouhal number.

in
1.0

0.5
,--.
_--_/
-,-
___I /’

0.0 _1__i_- 0.0

0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 6.0 8.0 10.0

;I /.
X* X’

(e)
1.5 /_1vw7--

1.0

in ,/---- \>A
:
0.5
/’
/’

0.0
0.0 2.0 4.0 6.0 8.0
1
1IO.O
X*

Fig. 9. Evolution of the jet half width vs stream-wise direction. (a) Run 2, (b) Run 3, (c) Run 4, (d) Run 5,
(e) Run 6.
 Numerical simulation of non-circular jets 11

The stream-wise evolution of the mean jet half velocity (U,) = (( Uc, + U,)/Z) contours is
presented in Fig. 8 for Runs 2-6. Each contour represents conditions within the range 0 < x* Q 5,
with an increment of x* = 1. Figure 8(a) indicates that at the first two stream-wise locations, the
jet maintains its initial elliptic shape with the major axis in the y direction. Further downstream,
there is a continuous reduction of the local aspect-ratio due to a larger spreading of the jet in the
minor-axis plane. By x* = 5, the mean profile is rotated by 90” and the axes are switched. In
contrast, the results in Fig. 8(b) indicate that the flow field of the square jet does not experience
an axis-switching; however, a 45” rotation is observed. Figure S(c) portrays the axis-switching of
the rectangular jet. A 90” rotation of the major axis is observed. The evolved shape is nearly
elliptical except for small stretching along the y-axis. In triangular configurations, a very different
behavior is experienced [Figs 8(d) and 8(e)]. In both jets, a 180” “ flip-flop” of the original profile
is observed, climaxed with a triangular shaped profile which is nearly equilateral.
The location of axis-switching is determined by monitoring the stream-wise variations of the
major and the minor axis half velocity widths for all the jet configurations. These widths are
non-dimensionaliz,ed by D, and are denoted by B,: and B T for the major- and the minor-axis,
respectively. The stream-wise variations of these widths are shown in Fig. 9 and the axis-switching
location for each run is listed in Table 2. Figure 9(a) indicates that the elliptic jet switches its axes
at X*~Z 3. The experimental data of the switch-over location for a 2: 1 elliptic jet is x* z 2.4 as
reported in Ref. [ 131, and x* z 2.8 as suggested in Ref. [12]. In the experiments of Ref. [13], the
jet is forced at St, = 0.4 and the momentum thickness at the jet exit is uniform. In the experiments
of Ref. [12], an unforced jet is considered with a non-uniform momentum thickness. The square
jet does not switch its axes and both widths grow monotonically in x* as shown in Fig. 9(b). The
switching location for the rectangular jet is further downstream than that for the elliptic jet
[Fig. 9(c)]. This can be attributed to the influence of corners. As indicated in Fig. 8(c), the
rectangular jet evolves into a rotated elliptic configuration. It must, therefore, smooth its corners
while simultaneously switching axes. Thus, its spreading rate in the minor plane is slower than that
in the elliptic jet. F?gures 9(d) and 9(e) show two axis-switchings for the two triangular jets. In both
jets, B f initially decreases to a minimum corresponding to the maximum of B ,*, and then increases.
The first axis-switching in the isosceles triangular jet occurs approximately twice as far downstream
as that in the equilateral triangular jet (Table 2). This can be attributed to the larger aspect-ratio
of the isosceles triangular nozzle which results in an initial gap between the major and the minor
axis half widths.

3.3. Entrainment of free-stream fluid

A measure of entrainment of the free-stream fluid provides an effective means of estimating the
mixing efficiency of the jets. Here, the entrainment is quantified by measuring the difference between
the average mass flow rate (conditioned on (J) < 0.99) at a downstream location Q (x*) and that
at the nozzle exist, Q, = Q (x* = 0). Figure 10 shows the normalized value of this parameter as a
function of the stream-wise coordinate. This figure shows that the isosceles triangular jet entrains

1
1.51 r (

1.0
0”
r,
sp
u
- 0.5
Table 2. Streamwise location(s)
of axis switching
RUll Axis switch (x*)
I no switch
2 3.1 0.0 2.0 4.0 6.0 8.0 10.0
3 no switch
4 6.3 X*
5 I .5,4.6 Fig. IO. Downstream variation of the entrainment ratio.
6 2.9,6.5
Legends are the same as those in Fig. 6.
12 R. S. MILLER et al.

.......
9’ .......
.......................

..... \,,I,,,,,, ........

...... ..... ,,,,,,,,,,,
. . . . ,,,,,,,,
........

*a.*
1.0 . . . _ . . .
. . . . . .
. . . . . . .
. . . _ _. .
. _ . . .
. . .
*. .
I

. . ,

. I .
I . .
. , I I *

\ I ,

\\
. . , ,

._ ._.

5
.... ._A
0.0 ._
.\
........
,_.

,,c_ .,

Y
.....

..........
......
. ,,-
. .

.,
........_. . ,
. . .

. , , .

I\\ . . . . . . I .

...... .___I I\\ . . . . . .;I:;;:;) , .

...... _ _ __.,I I . . . .- , . .

...... -_--c ,I/ . . .__ _-.,1//f , , .

. . _ _-- _.
_ . .

...... _ _ _-.///I I , I I I . . .

. _ I , I
-1.0
I _ . I , ,

.e-* .,,,,,I

(b) -1.0 0.0 1.0 **

CCC) -1.0 0.0 1.0 z*

Fig. Il. Time averaged cross-stream velocity vectors for Run 2. (a) x* = 2, (b) x* =4, (c) x* = 6.

nearly 125% of its initial mass flow rate as compared to 50% obtained for the circular jet. The
mass entrained by the other jets fall somewhere in between these two extremes. Conceptual
understanding of the entrainment process is aided by examining the time-averaged cross-stream
velocity vectors. Figures 1l-15 show these vectors for Runs 2-6, at stream-wise locations x* = 2,
4, and 6 (except for Run 3 in which vectors at x * = 1 4, and 6 are shown). The solid lines in these
figures denote the (U,,) iso-line. Stream-wise voitical structures are observed for all the
non-circular jets. The elliptic jet (Fig. 11) is shown to entrain fluid into the mixing zone along the
y-axis and to eject it along the z-axis. The effect is to contract the major axis of the jet while
simultaneously stretching the minor axis, resulting in axis-switching. Notice the recirculation
pattern caused by the four vortices. The flow field associated with this recirculation pattern is
responsible for the formation of the shoulders observed on the minor plane velocity profiles
presented in Fig. 5(b). A more complex flow pattern is observed for the square jet (Fig. 12). The
velocity field induced by the four counter-rotating pairs of vortices [Fig. 12(a)] causes an outflow
of the fluid on the flat sides and an inflow at the corners. By x * = 4 the original configuration is
rotated by 45” about the y-axis [Fig. 12(b)]. At this location, four additional pairs of vortices are
formed inside the original ones. Farther downstream, the flow pattern induced by these vortices
results in further stretching of the new corners [Fig. 12(c)]. Also, the inner set of vortex pairs are
no longer distinct and tend to lose their identity. As the flow evolves downstream, the induced
velocity field due to the outer set of stream-wise vortex pairs results in the formation of two
additional peaks on the velocity profiles in both the major plane and the minor plane as shown
in Fig. 5(c). The rectangular jet displays characteristics similar to those of the square and the elliptic
 Numerical simulation of non-circular jets 13

......
I_...
1.0 ::::
. ......
. ......
. . .._._ ... . .
....___..
.. .. .. .. .. .. ._...
..-..
. ...... ... . .
. ...... -....
, . . . . _, ... . .
... . .
4: : : : : : : ... . .

-...
0.0 0.0 : : : :
... . .

...-..
......
t . . . .
t::::::: .. . . . L. - . _
I::::::.. ._...
.. .. .. .. .. .. .. .. .. .. .. .. 1: : ::
.._..

I...,
. ...... ... . .
... . .
... . .
..d....

. . . . . . .

. . . ,,,, ... .
-1.0 t: : : :

(4 -1.0 0.0 1.0 z*

Fig. 12. Time averaged cross-stream velocity vectors for Run 3. (a) x* = I, (b) x* = 4, (c) x* = 6.
14 R. S. MILLERet al.

T
1
,<,,1,,.., . .

,,, . . Y* I,,,,, t,,., . .

,,. .. ,,I ,,,,,,,. , . .

,,. .
~..... l,l,,,,#>?~+..~

+. . . . . . . , ~..... 1.0
11,,,,,~*‘.~~~~
. . . . .
.... . .
... ......____
..___
I: : : ::::: -_....
...... I -_..
-... .

. ...__--
__.. .
......
II:::::::: -_
..........
_....

........__
.
__...
..-e-0

......
...... .. .. ..
.
.
.
.
.
.._
. .

..._--
.
. . . .
...... .,

_
* I
......
........

I
.

........
...... .. I .

.,.

..---
......
........

-): : : : : . .
0.0 ...... 0.0 .. .
...... ,\.
...... ,

1
.._
...... ,,
......
...... . .. ,,...
...... .. .. .. ,,..
_....
__ .... .. _...

~
__
.......... . . . _...

,
_....
.-es
...... _. . . .
......
. . . . I v-e
. . . . . , ,

-1.0 . . . . , , , I ...... -1.0

1.0 :::

d
I

. *

0.0 : : :
.
*
. I

. .

. . .

. . .

. . .

. . _

. . .

. . .

-1.0 : : :

. , ,

. , a

, . ,

. . .

(4 -1.0 0.0 1.0 z*

Fig. 13. Time averaged cross-stream velocity vectors for Run 4. (a) x* = 2, (b) x* = 4, (c) x* = 6.
 Numerical simulation of non-circular jets IS

Y’

1.0

. .
.. .. ..
,, ,, .* . ,.
,I,..

,, <, .
0.0
,,*, ,

,..-.

.A_-.

.-_--

___-_

____.

___ _.

..___

. .._.

. . .._

. . . . .

-1.0

I -1.0 0.0 1.0 2’ (b: -1.0 0.0 1.0 I*

Y* ::::::::::::::::::::::::::::::::::
. . . .
...........................
.......*
I . . .

*.o+
: : : :...........................
...........................
...........................
. .
.. .
...
I
II ... ....,.. ,\
3 . .

t :::: ,--. -
0.0
se..

I____
fiI
.__
__.
___
-__
-
I
_e--
___s
.__
-..
.. .. .. 1
.

tL....
_
~
,
-
~
.
*
. ..I 1
...
I :::: ...
,,,
..,.
..
...
...
*. . I
(4 -1.0 0.0 1.0 I *

Fig. 14. Time averaged cross-stream velocity vectors for Run 5. (a) x* = 2, (b) x* =4, (4 x*=6.
16 R. S. MILLER et al.

(4 -110 0:o l:o 2’ (h) --110 0.0 1.0 2‘

Fig. 15. Time averaged cross-stream velocity vectors for Run 6. (a) X* = 2, (b) X* = 4, (c) X* = 6.

jets (Fig. 13). Initially, two vortex pairs are formed with their axes coincident with the z-axis. This
enables the jet to adopt an elliptic configuration and to display a recirculation pattern similar to
that in Fig. 1l(a). However, the influence of the jet origin (nozzle geometry) is not completely lost
and the subsequent evolution is considerably different from that of the elliptic jet. At x* = 4, the
cross-section becomes diamond shaped, similar to that in the square jet at the same location. By
x* = 6, four vortex pairs are formed and the shape of the jet is elliptic with major axis on the z-axis
and with a slightly stretched minor axis. Due to the delay in its axis-switching, the rectangular jet
is a relatively inefficient mixer in the near field. The evolution of the cross-section of the equilateral
triangular jet is depicted in Fig. 14. At x* = 2 the profile for this jet adopts an approximate square
shape. However, the flow pattern and the subsequent evolution are clearly different from that in
the square jet. By x* = 4 [Fig. 14(b)] the profile is heart-shaped with a strong outflow induced by
two vortex pairs located at the corners. The interaction of these vortices continues to distort the
profile downstream [Fig. 14(c)]. For the isosceles triangular jet, Fig. 15(a) shows the entrainment
from the top corner and the base, and ejection of the fluid from the two long flat sides. Further
downstream, this causes a “flip-flop” of the profile. The fluid is entrained from the bottom corner
and is ejected through the top corners.
The results presented here indicate the influence of large scale structures on the global mixing
process in both circular and non-circular jets. These results also show that near-field mixing and
entrainment is characterized by the induced secondary flow field of the stream-wise vortices.

3.4. Influence on reactant conversion

The consequences of the flow evolution on the rate of reactant conversion in reacting jets are
portrayed by considering the transport of the ShvabZeldovich scalar variable, J. Figure 16 depicts
the instantaneous stoichiometric surface J = J,, corresponding to the flame sheet, for each jet. This
figure shows a severe distortion of the flame surface in the non-circular jets. In particular, the two
 Numerical simulation of non-circular jets

(4

(b)

(cl

I
LX

Fig. 16-caption on p. 19.

triangular jets show a highly stretched and convoluted topology with the formation of small scale
structures. The extent of distortion of the stoichiometric surface provides a measure of the
combustion efficiency as measured by the magnitude of the product formation. The downstream
evolution of the mean product mass fraction in the y - x and z - x planes is shown in Fig. 17.
Examination of Fig. 17(a) indicates that for the circular jet, the initial reaction occurs along the
18 R. S. MILLER et al.

(d)

Fig. 16-caption opposite.

jet boundaries and proceeds to spread both outward and inward. The two spikes at x * = 0 are due
to finite gradients of the Shvab-Zeldovich profile at the inflow. For the circular jet it is about eight
diameters downstream before the product reaches the centerline. The profile for the elliptic jet is
initially similar to that of the circular jet. The difference is,in the widths of the profiles in the major
and the minor planes. As the jet evolves downstream, the profiles in the two planes start to differ
and some of the profiles in the minor axis plane develop blunt topped humps [Fig. 17(b)]. Similar
flat-topped peaks are observed in the profiles in the square and the rectangular jets [Figs 17(c) and
17(d)]. The stream-wise vortex pairs [shown in Figs 12(c) and 13(c)] are responsible for increasing
mixing in these regions and the reaction takes place over a larger volume. The square jet profiles
portray approximately the same shape in the y - x and z - x planes. The rectangular jet for which
the initial axis-switching occurs at the stream-wise location of x* = 6.3 maintains its maximum
value of the mean product mass fraction at its outer edges. Both the equilateral and the isosceles
triangular jets [Figs 17(e) and 17(f)] are characterized by a large increase in the product formation
in the center of the jet as compared to those in the other jets.
 Numerical simulation of non-circular jets 19

Fig. 16. Instantaneous surface of the flame sheet. (a) Run 1, (b) Run 2, (c) Run 3, (d) Run 4, (e) Run 5,
(f) Run 6.

(4
2.0 -

1 .(I

‘h 0.0

-1 .o

-2.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

2.0

1 .o

‘N 0.0

-1.0

-2.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*
Fig. 17-caption on p. 22.
20 R. S. MILLER et al.

(b)

-1.0

-2.0 7
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

‘N

-1.0

-2.00.0
- 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

Fig. II-caption on p. 22.

 Numerical simulation of non-circular jets 21

Cd)

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

2.0

1.0

=_T-----
‘N 0.0
a-++_
-1.0

-2.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

I I
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 1
X*

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

Fig. II-cuptim on p. 22
22 R. S. MILLER et al.

2.0 ; I I I I I I I I I I

-1.0 =-

-2.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

2.0 I I

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
X*

Fig. 17. Profiles of the mean product mass fraction in the y - x and z - x planes. (a) Run 1, (b) Run 2,
(c) Run 3, (d) Run 4, (e) Run 5, (f) Run 6. Scale: 0.9 stream-wise equivalent diameter corresponds to
unity mass fraction.

The stream-wise variations of the integrated product thickness are shown in Fig. 18. This figure
shows that for x* < 3 the extent of products formed in the circular jet is lower than those in the
square and the triangular jets, but is higher than those in the elliptic and the rectangular jets. In
the region 3 < x* < 5 the products formed in all the jets are lower than that in the circular jet with
the exception of the isosceles triangular jet. Further downstream (x* > 6) all the non-circular jets
yield higher values of product thickness as compared to the circular jet. The results presented in
this figure are consistent with those in Fig. 10 which indicate that the product formation is directly
related to the entrainment. Therefore, the ratio of the integrated product thickness to the physical
area of the jet is considered. This area refers to regions in the stream-wise planes where (J ) < 0.99.
This ratio is denoted as y and is referred to as the cross-stream product density [34,29]. The
stream-wise variation of this ratio is shown in Fig. 19 for all jets. This figure suggests that after
a period of rapid growth, a plateau at the value of y w 0.55 is reached in all cases. This provides
an explanation for the correlation between the growth rate of the area and that of the product
thickness. Experimental confirmation of the existence of a plateau value of y for higher Reynolds
number turbulent jets is desirable.

10.0
r 7 0.80, 1 _

8.0
0.60

6.0
00” r 0.40
4.0

0.20

0.01 1 I 0.00 ’ * 3 h 1
0.0 2.0 4.0 6.0 8.0 10.0 0.0 2.0 4.0 8.0 8.0 1 .O
X* X’

Fig. 18. Equilibrium product thickness vs stream-wise direc- Fig. 19. Cross-stream product density vs stream-wise direc-
tion. Legends are the same as those in Fig. 6. tion. Legends are the same as those in Fig. 6.
 Numerical simulation of non-circular jets 23

Table 3. Streamwise location of

the scalar cme

RUll Scalar core (I*)

3 0.8
I 7.5
2 7.0 i!
3 8.0
3 0.6
4 8.5
5 6.5 ;;
6 6.0
5 0.4
g
p 0.2

Fig. 20. Normalized mass of unmixed fuel vs stream-wise

direction. Legends are the same as those in Fig. 6.

The traditional definition of the jet potential core is not very applicable to the quasi-transitional
jets studied here. Pertaining to mixing, a scalar core is defined as the stream-wise length of the jet
containing pure, unmixed fuel within the non-reacting jets. This corresponds to J = 0. Figure 20
presents the normalized integrated mass of fuel as a function of stream-wise location. A near linear
decrease of fuel mass fraction is observed in all the cases. The lengths of the scalar cores in all the
jets are listed in Table 3. The square and the rectangular jets, with no axis-switching and
axis-switching far downstream respectively, have the longest cores. The shortest core occurs for the
isosceles triangular jet.

4. SUMMARY AND CONCLUDING REMARKS

Detailed numerical experiments are conducted to study the entrainment and mixing character-
istics of the flow fields generated by non-circular turbulent jets. Simulations are conducted of jet
flows originating From elliptic, rectangular, and triangular nozzles with aspect-ratios of 1: 1 and 2: 1.
The results are compared with those of a circular jet of the same equivalent diameter to determine
the relative efficiency of non-circular nozzles in mixing enhancement. Flow visualization results
show that for both cornered and non-cornered jets, large scale coherent structures are formed. The
shape and dynamics of these structures depend on the azimuthal variation of the curvature of the
profiles at the jet exit. The triangular jets exhibit characteristics markedly different from the other
jets. Coherent large scale structures in these jets are quickly masked by the small scale structures
formed at the corners. In the elliptic and the rectangular jets, the orientations of the cross-sections
are modified by axis-switching. The rectangular jet switches its axes at a stream-wise distance
approximately twice that of the elliptic jet. This can be attributed to the effects of the corners.
Although the square jet does not show axis-switching, it is shown that a 45” rotation of its initial
profile results in entrainment of the free-stream fluid. The triangular jets switch their axes twice.
In the isosceles triangular jet, the first axis cross-over occurs approximately twice as far downstream
as that in the equilateral triangle. This is attributed to the larger aspect-ratio of the isosceles
triangular jet.
The entrainment and mixing in the near field of these jets are shown to be characterized by the
induced mean secondary flow field of the stream-wise vortices. Non-unity aspect-ratios, sharp
corners, and long flat surfaces are important factors in facilitating an efficient mixing configuration.
In the case of the rectangular jet, although it contains many of these features, its axes switch too
far downstream to cause significant near-field mixing. Although a non-unity aspect-ratio is
important for mixing enhancement, it is not sufficient for large entrainment in the near-field. The
isosceles triangular jet is shown to be the most efficient mixer. This jet produces the most intricate
network of stream-wise vortices which are responsible for enhanced mixing. The square jet ranks
as the second most efficient mixer, and the circular jet is the least efficient one. A comparison of
the flow fields produced by the two triangular jets reveals that the formation of small scale
structures at the corners does not have a significant influence in entraining the free-stream fluid.
24 R. S. MILLER et al.

The aspect-ratio is the primary difference between these two jets. The effect of the larger aspect-ratio
of the isosceles triangular jet is to alter the vorticity dynamics in this jet as compared to the
equilateral triangular jet. This results in a different stream-wise vorticity pattern which enhances
entrainment.
The limiting rate of the mean reactant conversion in reacting jets in which the fuel is discharged
to ambient oxidizer is evaluated by considering the transport of a Shvab-Zeldovich scalar variable.
It is shown that the isosceles triangular jet yields the highest amount of chemical products, whereas
the circular jet yields the lowest. However, the magnitudes of the cross-stream product density
approaches a plateau in all the jets. The magnitudes at this plateau are approximately the same
for all the cases. With the transport of the ShvabZeldovich variable, a scalar core is also defined.
It is shown that the 2: 1 aspect-ratio triangular jet has the shortest, and the rectangular jet has the
longest core.
The examination of the effects of harmonic forcing and the role of the initial momentum
thickness on the subsequent development of jet flows under the influence of non-equilibrium
chemical reactions are the subject of our current investigations.

Acknowledgemenrs-This work is part of an effort sponsored by the Office of Naval Research under Grant NOOOl4-90-J-
4013 and by the National Science Foundation under Grant CTS-9253488. Computational resources are provided by the
NSF through the National Center for Supercomputing Applications at the University of Illinois.

REFERENCES

1. J. P. Drummond and P. Givi, Suppression and enhancement of mixing in high-speed reacting flow fields, in M. Y.
Hussaini, J. D. Buckmaster, T. L. Jackson and A. Kumar, editors, Combustion in High-Speed Flows, pp. 191-229.
Kluwer Academic Publishers, The Netherlands, in press (1994).
2. 1. Wygnanski and H. Fiedler, Some measurements in the self-preserving jet. J. Fluid Mech. 38, 577 (1969).
3. S. C. Crow and F. H. Champagne, Orderly structure in jet turbulence. J. Fluid Mech. 48, 547 (1971).
4. A. J. Yule, Large-scale structure in the mixing layer of a round jet. J. Fluid Mech. 89, 413 (1978).
5. P. E. Dimotakis, R. C. Maike-Lye and D. A. Papantoniou, Structure and dynamics of round turbulent jets. Phys. Fluidr
26(11), 3185 (1983).
6. E. Gutmark and I. Wygnanski, The planar turbulent jet. J. Fluid Mech. 73, 465 (1976).
7. W. K. Everett and G. A. Robins, The development and structure of plane jets. J. Fluid Mech. 88, 563 (1978).
8. E. Gutmark, K. C. Schadow, D. M. Parr, C. K. Harris and K. J. Wilson, The mean and turbulent structure of
non-circular jets. AIAA Paper 85-0543 (1985).
9. P. M. Sforza, M. H. Steiger and N. Trentacoste, Studies on three dimensional viscous jets. AIAA J. 4, 800 (1966).
10. N. Trentacoste and M. P. Sforza, Further experimental results for three-dimensional free jets. AIAA J. 5, 885 (1967).
Il. A. Krothapalli, D. Baganoff and K. Karamcheti, On the mixing of a rectangular jet. J. F&d Mech. 107, 201 (1981).
12. C.-M. Ho and E. Gutmark, Vortex induction and mass entrainment in a small-aspect-ratio elliptic jet. J. Fluid Mech.
179, 383 (1987).
13. H. S. Husain and F. Hussain, Elliptic jets. Part 1. Characteristics of unexcited and excited jets. J. Fluid Mech. 208,
257 (1989).
14. H. S. Husain and F. Hussain, Elliptic jets. Part 2. Dynamics of coherent structures: pairing. J. Fluid Mech. 233, 439
(1991).
15. H. S. Husain and F. Hussain, Elliptic jets. Part 3. Dynamics of preferred mode coherent structure. J. Fluid Mech. 248,
315 (1993).
16. M. R. Dhanak and B. de. Bernardinis., The evolution of an elliptic vortex ring. J. Fluid Mech. 109, 189 (1981).
17. H. Viets and P. M. Sforza. Dvnamics of bilaterallv svmmetric vortex rines. Phvs. Fluids 15. 230 (1972).
18. E. Gutmark, K. C. Schadow, T. P. Parr, D. M. Hanson-Parr and K. J. Wils&, N&circular jets in combustion systems.
Experiments in Fluids 7, 248 (1989).
19. S. Koshigoe and A. Tubis, Wave structures in jets of arbitrary shape: I. Linear inviscid spatial stability analysis. Phys.
Fluids 29( 12), 3982 (1986).
20. S. Koshigoe and A. Tubis, Wave structures in jets of arbitrary shape: Il. Applications of a generalized shooting method
to linear stability analysis. Phys. Fluids 30(6), 1715 (1987).
21. S. Koshigoe, C.-M. Ho and A. Tubis, Application of a generalized shooting method to the linear stability analysis of
elliptic core jets. AIAA Paper 87-2722 (1987).
22. S. Koshigoe, E. Gutmark, K. C. Schadow and A. Tubis, Instability analysis on non-circular free jets, AIAA Paper
88-0037 (1988).
23. C. K. W. Tam and A. T. Thies, Instability of rectangular jets. J. Fluid Mech. 248, 425 (1993).
24. F. F. Grinstein and C. R. DeVore, Coherent structure dynamics in spatially-developing square jets. AIAA Paper
92-3441 (1992).
25. P. Givi, Model free simulations of turbulent reactive flows. Prog. Energy Combust. Sci. 15, 1 (1989).
26. D. Papamoschou and A. Roshko, The compressible turbulent shear layer: an experimental study. J. Fluid Mech. 197,
453 (1988).
27. F. A. Williams, Combustion Theory, 2nd edition. The Benjamin/Cummings Publishing Company, Menlo Park, CA
(1985).
28. J. P. Boris and D. L. Book, Solution of the continuity equations by the method of flux corrected transport. In Methodr
in Computational Physics. pp. 85-129, Academic Press, New York, NY (1976).
 Numerical simulation of non-circular jets 25

29. R. S. Miller, C. K. Madnia and P. Givi, Structure of a turbulent reacting mixing layer. Cornbust. Sci. Tech. 99, 1 (1994).
30. J. P. Boris, F. F. Grinstein, E. S. Oran and R. L. Kolbe, New insights into large eddy simulations. NRL Report
NRL/MR/4400-921-6979, Naval Research Laboratory, Washington, D.C. (1992).
31. J. P. Boris, On large eddy simulation using subgrid turbulence models, in J. L. Lumley, editor, Whither Turbulence?
Turbulence at the Crossroads, Lecture Notes in Physics, Vol. 357, pp. 344353. Springer, New York, NY (1990).
32. P. Givi, Spectral and random vortex methods in turbulent reacting flows, in P. A. Libby and F. A. Williams, editors,
Turbulent Reacriq Flows, Chap. 8, pp. 475-572. Academic Press, London, UK (1994).
33. S. James, Direct numerical simulation of laminar vortex rings. M.S. Thesis, Department of Mechanical and Aerospace
Engineering, State Univ. of New York at Buffalo, Buffalo, NY (1994).
34. M. M. Koochesfahani and P. E. Dimotakis, Mixing and chemical reactions in a turbulent liquid mixing layer. J. Fluid
Mech. 170, 83 (1986).