Theor. Comput. Fluid Dyn.

DOI 10.1007/s00162-009-0173-y
ORIGINAL ARTICLE
Ali B. Olcay Paul S. Krueger
Momentum evolution of ejected and entrained uid during
laminar vortex ring formation
Received: 2 July 2008 / Accepted: 24 September 2009
Springer-Verlag 2009
Abstract The evolution of total circulation and entrainment of ambient uid during laminar vortex ring
formation has been addressed in a number of previous investigations. Motivated by applications involving
propulsion and uid transport, the present interest is in the momentum evolution of entrained and ejected uid
and momentum exchange among the ejected, entrained uid and added mass during vortex ring formation.
To this end, vortex rings are generated numerically by transient jet ejection for uid slug length-to-diameter
(L/D) ratios of 0.53.0 using three different velocity programs [trapezoidal, triangular negative slope (NS),
and positive slope (PS)] at a jet Reynolds number of 1,000. Lagrangian coherent structures (LCS) were uti-
lized to identify ejected and entrained uid boundaries, and a Runge-Kutta fourth order scheme was used for
advecting these boundaries with the numerical velocity data. By monitoring the center of mass of these uid
boundaries, momentum of each component was calculated and related to the total impulse provided by the
vortex ring generator. The results demonstrate that ejected uid exchanges its momentum mostly with added
mass during jet ejection and that the momentum of the entrained uid at jet termination was < 11%of the total
ring impulse in all cases except for the triangular NS case. Following jet termination, momentum exchange
was observed between ejected and entrained uid yielding significant increase in entrained uids momentum.
A performance metric was dened relating the impulse from over-pressure developed at the nozzle exit plane
during jet ejection to the ow evolution, which increased preferentially with L/D over the range considered.
An additional benet of this study was the identication of the initial (i.e., before jet initiation) location of the
uid to be entrained into the vortex ring.
Keywords Vortex ring formation Entrainment Momentum exchange Impulse
PACS 47.32.cf
1 Introduction
During the formation of a vortex ring from a starting jet, the roll-up of the ejected shear layer in the vicinity
of the nozzle exit plane and subsequent downstream advection of the vortex ring embodies the transport of a
variety of owquantities including circulation, ambient uid entrained into the ring, and uid impulse/momen-
tum. The unsteady advection of these quantities associated with vortex ring formation is relevant for a range of
Communicated by H. Aref
A. B. Olcay (B)
Department of General Engineering, University of WisconsinPlatteville, Platteville, WI 53818, USA
E-mail: olcaya@uwplatt.edu
P. S. Krueger
Department of Mechanical Engineering, Southern Methodist University, Dallas, TX 75275, USA
E-mail: pkrueger@lyle.smu.edu
A. B. Olcay, P. S. Krueger
applications including ow control by synthetic jets [37], uid mixing and heat transfer [26], and propulsion
[3, 6, 16], respectively. Several studies have revealed aspects of the transport processes associated with vortex
ring formation and advection.
The roll-up of the vortex sheet and the connection of this process with the circulation of the resulting
ring has been studied experimentally, numerically, and analytically [10, 13, 24, 30]. It is well recognized that
a dominant factor controlling the circulation of the vortex ring is the magnitude of the jet velocity, owing
to its control over strength of the jet shear layer, giving the so-called slug model for predicting vortex ring
circulation entirely from the jet axial velocity [12, 36]. As apparent from the detailed measurements of Didden
[10], however, transients during jet initiation lead to a non-uniform jet ow and a disproportionately large
contribution to circulation from the initial jet transient, which is not predicted by the slug model. Krueger [15]
proposed a simple model to account for the initial transient based on over-pressure at the nozzle exit plane
during jet initiation for rapidly initiated jets. Approximations made in the model make it inapplicable to slowly
initiated jets, but based on the model premise one can anticipate that slowly initiated jets will not experience
the same disproportionate circulation increase at initiation due to lower nozzle exit over-pressure.
The roll-up of the jet shear layer is also responsible for entrainment of ambient uid into the forming ring
by Biot-Savart induction [1, 26] and as such, the evolution of vortex ring circulation is directly relevant to mass
transport by vortex rings. The dye visualization of Didden [10] allows a qualitative assessment of the evolution
of uid entrainment during shear layer roll-up, showing a tight spiral initially with little entrained uid, which
later opens up and allows a large amount of uid to be entrained into the ring boundary immediately following
jet termination. Olcay and Krueger [26] used dye visualization and Lagrangian coherent structures (LCS)
obtained from digital particle image velocimetry (DPIV) data to identify the vortex ring boundary and quan-
titatively assess ambient uid entrainment during vortex ring formation. Their results show a nearly twofold
increase in the rate of uid entrainment into the ring after jet termination as compared to during jet ejection.
The variation of the jet velocity with time (jet velocity program) significantly affected the observed trend and
the nal fraction of ambient uid present in the vortex ring bubble. Specifically, the fraction of ambient uid
entrained into the ring could be increased by as much as 30% for velocity programs with a rapid initiation and
gradual termination as compared to velocity programs with more gradual initiations and rapid terminations,
even though both the slug model predictions and actual DPIV measurements of nal ring circulation were the
same for both cases. The entrainment benet afforded by the rapidly initiated jets is attributable to the thin-
ner shear layer (tighter vortex spiral) and increased initial vorticity ux characteristic of these jets. Recently,
Shadden et al. [35] also studied transport and stirring features during vortex ring formation using LCS with
experimental and numerical data to monitor vortex ring evolution throughout the formation process. Using
these results, they were able to measure overall entrainment quantitatively and to investigate stirring charac-
teristics of vortex rings and stretching of interfaces between uid regions as the vortex ring forms. Following
ring formation, ambient uid continues to be entrained into the vortex ring by viscous diffusion of vorticity,
which has been quantied and modeled in previous studies [8, 20].
In addition to circulation and mass transport, the total impulse supplied by the jet to the ow (I
T
) is trans-
ported downstream by the vortex ring and can be related to the uid momentum associated with the vortex
ring as
I
T
= (m
EJ
+m
E
+m
A
)U
R
, (1)
where m
EJ
, m
E
, and m
A
are ejected, entrained, and added mass of the vortex ring, respectively, and U
R
is the
velocity of the vortex ring [9, 16]. For starting jets, I
T
can also be expressed as
I
T
= I
U
+ I
P
, (2)
where
I
U
=
t
p
_
0
_
A
u
2
J
(r, t )dAdt (3)
is the impulse supplied by the jet momentum and
I
P
=
t
p
_
0
_
A
( p(r, t ) p

)dAdt (4)
Momentum evolution of ejected and entrained uid
is the impulse due to nozzle exit over-pressure [16]. In Eqs. 3 and 4, u
J
(r, t ) and p(r, t ) are the jet velocity and
pressure at the nozzle exit plane, respectively, and t
p
is the jet duration.
1
The pressure impulse, I
P
, provides
additional impulse over what would be expected from the jet momentum alone, which leads to thrust augmen-
tation in pulsed jets [6, 17] and improved propulsive efciency in pulsed jet propulsion utilizing short jet pulses
that yield isolated vortex rings with no trailing jets [3, 4, 23]. Relating Eqs. 1 and 2, Krueger and Gharib [16]
reasoned that I
P
is related to the acceleration of entrained and added mass during vortex ring formation, but this
sort of before-and-after comparison cannot resolve the relative contributions of the entrained and added mass
or the phase of the ring formation process most responsible for their contribution. Nevertheless, understanding
the evolution of the added mass and entrained uid momentummay help optimize propulsive benets obtained
through I
P
. Moreover, for applications such as mixing one may be interested in knowing the impulse required
to move m
E
downstream. Is a significant increase in total impulse required to entrain m
E
or does m
E
steal
some of the momentum of the ejected uid? For applications where advection of the entrained uid, m
E
, is the
goal (e.g., convective cooling), entraining m
E
with minimum additional input from the vortex ring generator
would be desirable.
Several studies have observed the relationship between the momentum gained by uid entrained after
vortex ring formation is complete and the motion of the vortex ring. During this phase of motion the ring slows
as additional mass is entrained due to conservation of momentum [2, 21, 31]. These studies do not address the
ring formation process or quantify the momentum of the entrained uid. The vortex ring entrainment results
of Olcay and Krueger [26] provide a quantitative perspective of entrainment during the formation process and
suggest that the entrained uid obtains most of its momentum following jet termination because the largest
quantity of ambient uid is entrained during this period. This is certainly not conclusive, however, since the
ring velocity is highest just after jet termination (suggesting the uid already entrained during jet ejection has
high momentum) and the uid entrained after jet termination was already set in motion by the ring forma-
tion. Clearly, Lagrangian tracking of the ejected and entrained uid domains during vortex ring formation is
required to fully assess the evolution of their momenta. A few studies [2729] have considered the momentum
associated with vortices in animal wakes for analysis of locomotion. While effective and reasonably general,
these methods consider the vortex as a whole and do not allow segregation of momentum associated with
ejected and entrained uid as observed in starting jet ows.
While previous studies have addressed circulation and entrainment evolution during vortex ring forma-
tion and subsequent advection, little information is available on the evolution of ejected and entrained uid
momenta during vortex ring formation or the fraction of I
T
related to entrained uid. The objective of this
study is to investigate the ejected and entrained uid momentum during vortex ring formation, determine the
relative momentum contribution of m
EJ
compared to m
E
and m
A
, elucidate the momentum exchange process,
and analyze how this may be leveraged to optimize I
P
and minimize the impulse required to entrain uid by
adjusting parameters controlling vortex ring formation such as jet velocity program. To this end, vortex ring
formation by transient jet ejection from a tube is studied numerically to obtain velocity data inside and outside
the tube. LCS results obtained from the CFD data are used to identify the ejected and entrained uid regions at
jet initiation (t

= 0). Then, by monitoring the center of mass (CM) of these uid regions, momentumevolution
of the ejected and entrained uid is obtained and related to the fully developed vortex rings momentum.
In Sect. 2 the numerical model and methods are briey described. In Sect. 3 the time evolution of ejected
and entrained uid is presented for three velocity programs and jet slug length-to-diameter (L/D) ratios in
the range between 0.5 and 3.0 are discussed. In Sect. 4 ejected and entrained uid momenta are discussed in
comparison to the added mass effect.
2 The numerical model
Transient jet ejection from a tube is simulated using the domain geometry shown in Fig. 1. The geometry is
axisymmetric, therefore, only the x r plane is shown in Fig. 1. The domain dimensions in the x r plane
were chosen as 15D in length and 3.5D in height, where D is the tube diameter, to ensure negligible domain
size inuence on the ow evolution. Tube length was arbitrarily set to 5D.
On the outer boundary, ambient pressure conditions and zero normal derivatives of velocity were specied
for the left, right, and top boundaries in Fig. 1. These outowconditions yielded faster solution convergence for
1
Strictly speaking, the upper limit of time integration in Eqs. 3 and 4 should be taken to t

> t
p
where t

is sufciently large
that the vortex ring has convected away from the nozzle exit plane by this point. For the Reynolds numbers and jet slug length to
diameter (L/D) ratios of interest in this investigation, however, I
P
and I
U
are determined primarily during jet ejection, so Eqs. 3
and 4 are utilized here.
A. B. Olcay, P. S. Krueger
5D
15D
0.5D
3.5D
Wall
Outlet
Axis
Nozzle exit
plane
x
r
U
P
(t)
Fig. 1 Solution domain and boundary conditions
each time step and the results showed negligible difference (< 1% in hydrodynamic impulse for L/D = 2.0)
when compared with outow conditions that specify zero normal pressure gradient. The tube center line was
taken as the axis of symmetry. The no-slip condition was enforced on the tube wall. To simulate the piston
motion, a uniform inlet velocity, U
P
(t ), was specied at the tube inlet.
A critical parameter for vortex ring formation is the stroke length-to-diameter ratio (L/D), dened as the
ratio of the total ejected uid displacement (during jet ejection) to the tube diameter, namely
L
D
=
1
D
t
p
_
0
U
p
(t )dt (5)
Stroke ratios of L/D = 0.5, 1.0, 2.0, and 3.0 were simulated. Only L/D < 4.0 was considered so that isolated
vortex rings are formed and vortex ring pinch off [11] was avoided.
Three different protocols were specied for the time evolution of U
P
(t ) for a given L/D: trapezoidal, trian-
gular negative slope (NS), and positive slope (PS) velocity programs. Plots of the velocity programs are shown
in Fig. 2. Acceleration and deceleration periods were set to 0.1t
p
for a trapezoidal velocity program. Triangular
PS and triangular NS velocity programs use 0.9t
p
and 0.1t
p
in the acceleration phase, 0.1t
p
and 0.9t
p
during
the deceleration phase, respectively. The triangular PS and NS cases address the effects of non-impulsive jet
start up and termination, respectively, on the shear layer roll-up and associated ambient uid entrainment.
Jet Reynolds number (Re
J
) is calculated based on the jets maximum velocity (U
M
), tube diameter (D)
and the uid kinematic viscosity () as
Re
J
=
U
M
D

. (6)
The Re
J
for all simulations was 1,000.
2.1 Flow eld solution
The ow evolution was simulated using the axisymmetric, unsteady, incompressible NavierStokes equations
with zero swirl, namely,
1
r

r
(ru
r
) +
u
x
x
= 0
u
r
t
+u
r
u
r
r
+u
x
u
r
x
=
1

p
r
+
_
1
r

r
_
r
u
r
r
_
+

2
u
r
x
2

u
r
r
2
_
(7)
u
x
t
+u
r
u
x
r
+u
x
u
x
x
=
1

p
x
+
_
1
r

r
_
r
u
x
r
_
+

2
u
x
x
2
_
Momentum evolution of ejected and entrained uid
t / t
P
U
P
(
t
)
/
U
M
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
1.2
(a)
t / t
P
U
P
(
t
)
/
U
M
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
1.2
(b)
t / t
P
U
P
(
t
)
/
U
M
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
1.2
(c)
Fig. 2 Normalized velocity programs for L/D = 2.0: a trapezoidal, b triangular NS, and c triangular PS velocity programs
The governing equations were solved using a nite volume method with a xed time step. The solver scheme
was second order accurate for both time and space. The SIMPLE algorithm was used for pressurevelocity
coupling and the QUICK scheme was used for discretization of the convective terms.
The domain was discretized into 388120 nodes (see Fig. 3) in the axial and radial directions, respectively.
A total of 46,560 quadrilateral cells were employed with increased mesh density near the nozzle exit plane
and the tube wall region to properly resolve wall gradients during the formation process.
A. B. Olcay, P. S. Krueger
Fig. 3 Sample mesh used during the study. 388 120 nodes were used in the axial and radial directions, respectively. Only every
fth mesh point is shown
Table 1 Space and time convergence for the trapezoidal velocity program, Re
J
= 1,000, and L/D = 2.0 at t

= 2.5
I
T
U
M
D
3
% Difference
Mesh resolution
258 80 1.6878 1.834
388 120 1.6591 0.103
582 180 1.6574 0
Time step
0.04 1.6310 1.759
0.02 1.6557 0.271
0.01 1.6602 0
Table 2 Summary of the tested cases
L/D Velocity program
2.0 Triangular NS
2.0 Triangular PS
0.5, 1.0, 2.0, 3.0 Trapezoidal
Re
J
= 1,000 for all cases
Because of the present focus on impulse, the hydrodynamic impulse [18, 22] was used to monitor space
and time convergence since it represents the total impulse required to generate the ow. The hydrodynamic
impulse was computed as
I
T
=
_ _

r
2
dxdr, (8)
where

is the azimuthal component of vorticity. The integration domain was chosen as 0 x/D 10 and
0 r/D 3.5.
Mesh convergence was tested for the trapezoidal velocity program, L/D = 2.0, and t = tU
M
/D = 0.02
on meshes with resolutions 258 80, 388 120, and 582 180. I
T
was computed from the vorticity eld
at t

= 2.5, where t

is non-dimensional time dened as t /t

P
. The results given in Table 1 show that the
388 120 mesh (given in Fig. 3) is sufcient for this study since this mesh provided 0.103% difference in the
total impulse compared to the total impulse for the 582 180 mesh.
For time-step convergence, simulations with t = 0.01, 0.02, and 0.04 on the 388 120 mesh were
performed for the trapezoidal velocity program with L/D = 2.0. The results for I
T
are shown in Table 1. The
0.02 time step was chosen since this time step provided 0.271% difference in the total impulse compared to
the total impulse for the 0.01 time step.
As L/D and the velocity program are key controlling parameters in the vortex ring formation and uid
entrainment process, a wide range of L/D and velocity programs were studied. A list of cases simulated is
shown in Table 2.
Momentum evolution of ejected and entrained uid
2.2 Lagrangian coherent structures (LCSs) implementation
Olcay and Krueger [26] and Shadden et al. [33, 34] briey discuss the use of LCS. In the present study, the
Eulerian CFD data were used to obtain Lagrangian information about the ow as expressed using the nite
time Lyapunov exponents (FTLE). Details of this approach may be found in Shadden et al. [33]; however, a
brief overview following Olcay and Krueger [26] will be given here for completeness.
The equation describing the trajectory of a uid particle at position x
0
at time t
0
may be expressed as
x(t ; t
0
, x
0
) = V(x(t ; t
0
, x
0
), t ), (9)
where x(t
0
; t
0
, x
0
) = x
0
. The right side of Eq. 9 can be obtained fromthe CFDvelocity eld data. The solution
to (9) is a ow map (
t
0
+T
t
0
(x
0
)) describing the position at time t = t
0
+ T of the uid particle initially at x
0
at time t
0
, namely,

t
0
+T
t
0
(x
0
) = x(t
0
+ T; t
0
, x
0
) (10)
Then the FTLE is dened as

T
t
0
(x)
1
|T|
ln
_

max
, (11)
where
max
is the maximum eigenvalue of
_

t
0
+T
t
0
(x)
_

t
0
+T
t
0
(x)
_
(12)
and ()

denotes the adjoint operation. It can be shown [33] that the separation of particles advected by the ow
is proportional to e

T
t
0
(x)|T|
to leading order. Hence, the FTLE is roughly a measure of the maximum expansion
rate of particle pairs advected by the ow. For example, strong and convex peaks (ridges) in the FTLE eld
correspond to the stable (for T >0) and unstable (for T <0) manifolds of hyperbolic xed points in steady
ows (separatrices).
Generalizing to aperiodic and unsteady ows, LCS are dened as the ridges in the FTLE eld. Shadden
et al. [33] show that the ux across a LCS scales like
1
|T|
and thus, for large |T| can be treated as a material
line, transport barrier, or separatrix in the ow. LCS can be obtained for both forward and backward time inte-
gration of the velocity eld data (e.g., right side of Eq. 9). The resulting LCSs can be regarded as the unsteady
generalization of the stable (i.e., T > 0, also called repelling LCS) and unstable manifolds (i.e., T < 0, also
called attracting LCS) of the ow eld. Since LCSs behave like material lines, they can identify boundaries
separating uid particles in different domains. This was illustrated by Shadden et al. [34], who combined the
attracting and repelling LCSs to study entrainment of a formed ring. Since the formulation applies equally
well to unsteady ows, it can be used to identify the uid that will comprise the nal vortex ring before it is
formed.
LCSs in this study were calculated by using a software package called ManGen developed by Francois
Lekien and Chad Coulliette in 2001 (http://www.lekien.com/francois/software/mangen/). ManGen provided
the FTLE eld by computing
T
t
0
(x) from Eq. 11 for a grid of massless particles placed in the domain and
advected using the computed velocity eld. For the present investigation, computation of was performed
with a uniform grid of massless particles with resolution 0.01D to produce sharp ridges for the repelling LCS,
and both repelling and attracting LCSs are used to determine the initial (t

= 0) positions of both ejected and

entrained uid.
A contour plot of the FTLE eld at t

= 0 for the trapezoidal velocity program with L/D = 2.0 is shown

in Fig. 4. It is noted that high FTLE values (i.e., ridges) illustrate the LCS as described by Shadden et al. [33].
During post-processing, a threshold is applied to FTLE eld to identify the LCS and to locate the entrained
uid domain as shown in Fig. 5.
The LCS results for the trapezoidal velocity program and L/D = 2.0 at t

= 0 are shown in Fig. 5. The

repelling LCS was calculated from ManGen taking |T| = 10 (shown as the red solid line), and the nozzle exit
plane is identied as the attracting LCS at t

= 0 (shown as blue dashed line). Since repelling and attracting

LCSs give the transport barriers between ejected and entrained uid at t

= 0, these data were used for

identifying uid particles that will comprise the ejected and entrained uid in the formed ring and are labeled
as such from t

= 0. For t

> 0, the boundaries of these regions were advected using a Runge-Kutta fourth
order scheme [5] to determine the time evolution of the ejected and entrained uid.
A. B. Olcay, P. S. Krueger
Fig. 4 Color contour plot of the FTLE eld with T = 10s for an L/D = 2.0 trapezoidal velocity program at t

= 0
x / D
r
/
D
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
Nozzle wall
Attracting LCS
Repelling LCS
Fig. 5 Ejected and entrained uid boundaries at t

=0 for the trapezoidal velocity program with L/D=2.0

As the domain boundaries were stretched during advection, the distance between points forming the ejected
or entrained uid boundaries were monitored and additional points were added where excessive stretching
occurred. The maximum allowable distance between points, z
MAX
, was adjusted to provide sufcient res-
olution without excessive points. The appropriate z
MAX
was chosen based on a convergence study. To test
the convergence, the entrained uid was advected using z
MAX
= 0.05D, 0.025D, and 0.01D. Total volume
conservation and time evolution of the axial location of the center of mass (CM
x
) were monitored during this
process. The results are plotted in Fig. 6. The center of mass axial location was calculated using
CM
x
=
_
x f (x)dx
_
f (x)dx
, (13)
where f (x) represents the dimensionless area obtained by revolving a slice of the uid domain of interest about
the tube centerline. The integration was taken over the uid domain of interest (e.g., ejected uid or entrained
uid). Based on the results shown in Fig. 6, z
MAX
= 0.025D was chosen for further calculations since it
provided constant volume with < 1% deviation from the 0.01D case and CM
x
proved nearly insensitive to
z
MAX
over the tested range.
Once CM
x
was obtained, the axial velocity of CM
x
was calculated using a second order central differencing
method [5] for a constant-time step. The computed velocity of CM
x
had second order truncation error.
Momentum evolution of ejected and entrained uid
t / t
P
V
E
/
V
M
E
0 0.5 1 1.5 2 2.5 3
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
0.01D
0.025D
0.05D
(a)
t / t
P
C
M
X
/
D
0 0.5 1 1.5 2 2.5 3
0.5
1
1.5
2
2.5
0.01D
0.025D
0.05D
(b)
Fig. 6 Fluid domain boundary convergence test for z
MAX
= 0.05D, 0.025D, and 0.01D. a Volume change of entrained uid
with the time; btime evolution of CM
x
for the entrained uid. The results are for the trapezoidal velocity programwith L/D = 2.0.
V
ME
is the entrained uid volume at t

= 2.5
3 Results
3.1 Time evolution of ejected and entrained uid
Figure 7 illustrates the evolution of ejected and entrained uid for L/D = 2.0 and the trapezoidal velocity
program. The ejected uids shape initially looks like a cylinder as one would expect; however, the back edge
is not straight like the front edge. This actually indicates that uid near r = D/2 is not pulled into the ring
toward the end of the jet pulse. As the ow evolves, the entrained uid domain is swept over the developing
vortex ring spiral by Biot-Savart induction and enters the spiral through the rear such that the union of the two
uid domains gives the expected ellipsoidal shape to the vortex ring boundary in Figs. 7e, 8. The initial location
and time evolution of the ejected and entrained uid domains agrees qualitatively with Peng and Dabiri [27]
although Peng and Dabiri do not investigate the two domains separately.
The precise overlapping of the domain boundaries illustrated in Fig. 8 conrms the accuracy of the advec-
tion scheme. As further conrmation of the results, comparing the nal volume of the vortex ring with the
experimental results of Olcay and Krueger [26] (for Re
J
=1,000, L/D=2.0 and trapezoidal velocity program)
gives < 5% difference, which is within experimental uncertainty.
3.2 The effect of velocity program and L/D on the initial location of the entrained uid boundary
As illustrated in the preceding section, LCS at t

= 0 identies the location of ambient uid that will eventually

be entrained into the vortex ring. Such information is especially useful for the applications involving mixing
since it indicates where uid components should reside at t

= 0 to mix most completely (i.e., minimal waste).

While the present results are limited to laminar ow (Re
J
= 1,000) with a highly organized mixing process
(see Fig. 7), the entrained uid domains identied herein may also provide insight into applications requiring
ne-scale mixing (such as chemical reactions) since the vortex rings can break down and become turbulent
downstreamfor slightly higher Reynolds number. Also, the initial shear layer roll-up for rings that are turbulent
immediately following formation is qualitatively similar to laminar vortex ring roll-up (see Fig. 8 in Glezer
[12]).
To analyze the effect of jet parameters on the mixing process, LCSs are used to determine the size and
shape of the entrained uid region at t

= 0 for different L/D and velocity programs. Figure 9 shows the

initial shape of the entrained uid regions for the trapezoidal, triangular NS, and PS velocity programs with
L/D = 2.0. As expected from the results of Olcay and Krueger [26], the domain sizes are larger for the trap-
ezoidal and triangular NS velocity programs. Likewise, the trapezoidal and triangular NS velocity programs
entrain a larger fraction of ambient uid from in front of the nozzle (x > 0), which is a consequence of the
A. B. Olcay, P. S. Krueger
x / D
r
/
D
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
Nozzle wall
Entrained fluid
domain
Ejected fluid
domain
Nozzle exit
plane
(a)
x / D
r
/
D
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
Nozzle wall
Entrained fluid
domain
Ejected fluid
domain
(b)
x / D
r
/
D
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
(c)
x / D
r
/
D
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
(d)
x / D
r
/
D
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
0
0.5
1
1.5
(e)
Fig. 7 Evolution of ejected and entrained uid. Panels ae show the ejected and entrained uid at t

= 0.0, 0.5, 1.0, 1.5, and

2.0, respectively, for the trapezoidal velocity program with L/D=2.0. The red solid line and blue dash-dot line represent ejected
uid and entrained uid domains, respectively
Momentum evolution of ejected and entrained uid
x / D
r
/
D
0.5 1 1.5 2 2.5
0
0.2
0.4
0.6
0.8
1
Fig. 8 Enlarged view of ejected and entrained uid regions at t

= 2.0 for the trapezoidal velocity program with L/D = 2.0

x / D
r
/
D
-0.2 0 0.2 0.4 0.6
0
0.2
0.4
0.6
0.8
Trapezoidal
Triangular NS
Triangular PS
Fig. 9 Shape of the entrained uid regions at t

= 0 for the trapezoidal, triangular NS, and PS velocity programs with L/D = 2.0
stronger overall Biot-Savart induction provided by the strong initial shear layer associated with the rapid jet
initiation for these cases.
Figure 10 illustrates the initial shape of the entrained uid regions for the trapezoidal velocity program
with L/D of 0.5, 1.0, 2.0, and 3.0. As in Fig. 9, most of the entrained uid initially resides in front of the
nozzle exit plane. In fact, as the volume of entrained uid increases with L/D, the fraction of it residing in
front of the nozzle increases until for L/D = 3 the largest portion of entrained uid comes from very near to
the nozzle centerline. This trend may seem at odds with the fact that entrainment into the ring occurs from the
rear of the forming spiral and closer to the nozzle lip, but it results from the increased effect of Biot-Savart
induction sweeping the entrained uid over the vortex spiral for larger L/D due to the larger ring circulation
and longer formation period as L/D is increased.
3.3 Impulse evolution of ejected and entrained uid
Figure 11 shows the ejected and entrained uid momentum evolution of a vortex ring from jet initiation to the
steadily translating state for the trapezoidal velocity program with L/D = 2.0. Axial ejected and entrained
momenta (I
EJ
and I
E
) were calculated using
A. B. Olcay, P. S. Krueger
x / D
r
/
D
-0.2 0 0.2 0.4 0.6
0
0.2
0.4
0.6
0.8
L / D = 0.5
L / D = 1.0
L / D = 2.0
L / D = 3.0
Fig. 10 Shape of the entrained uid regions at t

= 0 for the trapezoidal velocity program with L/D = 0.5, 1.0, 2.0, and 3.0
t / t
P
I
(
t
)
/
I
T
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
0
0.2
0.4
0.6
0.8
1
1.2
I
E
I
EJ
I
R
= I
E
+I
EJ
Added mass momentum
Fig. 11 Time variation of ejected and entrained uid momenta for the trapezoidal velocity program with L/D = 2.0
I
EJ
= m
EJ
U
EJ
(14)
I
E
= m
E
U
E
,
where U
EJ
and U
E
represent the axial velocities of the center of masses of the ejected and entrained uids,
respectively. The results are normalized using I
T
computed at t

= 2.5, which is the point where I

T
reaches
its nal, steady value (i.e., the ring is fully developed and boundary effects are minimal). Figure 11 indicates
that the bulk of the ejected uid impulse is obtained during the initial acceleration similar to an isolated rigid
particle. Subsequently, I
EJ
decreases as its momentum is shared with the entrained mass and external uid
(i.e., uid not in either the ejected or entrained uid domain). The latter represents what eventually is the added
mass component, and is clearly more significant than the entrained mass momentum because I
EJ
decreases
faster than I
E
increases for t

< 1.0. The sharp decrease in I

EJ
during 0.9 < t

< 1.0 is because of negative

over-pressure at jet termination, which removes momentum from the ow.
After the jet stops (i.e., t

> 1.0), ejected uid continues to transfer its momentum to the entrained uid.
The exchange between the ejected and entrained uid components appears as oscillations in the respective
components for t

=1.05.0 as a result of the spiraling of the two uid components around each other during
the roll-up and entrainment process as illustrated in Fig. 7.
A key observation from Fig. 11 is that the entrained uid gains most of its momentum after jet termination.
Indeed, at jet termination, the momentum of the entrained uid only accounts for about 10% of I
T
. Following
jet termination, the uid motion evolves according to conservation of total momentum, so one may conclude
Momentum evolution of ejected and entrained uid
t / t
P
I
E
/
I
T
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
0
0.05
0.1
0.15
0.2
0.25
Trapezoidal
Triangular NS
Triangular PS
Fig. 12 Time variation of entrained uid momentum for trapezoidal, triangular PS, and NS velocity programs with L/D = 2.0
t / t
P
I
E
/
I
T
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
-0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
L / D = 0.5
L / D = 1.0
L / D = 2.0
L / D = 3.0
Fig. 13 Time variation of entrained uid momentum for L/D = 0.5, 1.0, 2.0, and 3.0 for the trapezoidal velocity program
that entrainment contributes very little to any impulse benet provided by pulsation. On the other hand, this
also indicates that the momentum cost to transport entrained uid with the ring is small.
Figure 12 illustrates the time variation of I
E
for the three studied velocity programs. Note that I
E
for the
triangular PS case is negligible until t

= 0.75. On the other hand, for the triangular NS case I

E
increases
throughout the pulse duration. Thus, a rapid initial acceleration (e.g., trapezoidal and triangular NS velocity
program) immediately imparts momentumto the entrained uid, and a gradual acceleration delays this process.
Somewhat counter intuitively, Fig. 12 also shows that I
E
rises rapidly when U
P
(t ) begins decelerating. This
observation is related to the more rapid uid entrainment into the ring following jet termination [26].
The entrained uid momentumvariation for different stroke ratios is given in Fig. 13. Clearly the maximum
I
E
is a larger fraction of the total impulse as L/D decreases, except for L/D = 0.5. Figure 13 also shows
that entrained uid momentum for all of the L/Ds except for L/D = 3.0 makes a dip during jet ejection.
This indicates that the entrained uid for these cases is slowed (or even moves backward as for L/D = 0.5,
i.e., the entrained uid gets pushed back during jet start up) by its interaction with the forming ring. During
0.9 t

1.0, a kink appears in entrained uid momentum plot (Fig. 13). This is due to the jet termination
which pulls the uid around the nozzle lip. As in Fig. 12, I
E
increases significantly for all L/Ds following jet
deceleration.
Figure 14 gives summary of I
E
/I
T
values versus I
U
/(U
M
D
3
) for all the studied cases when the jet is
terminated (i.e., t

=1.0). The value of I

E
at this instant is most relevant for assessing the input required
from the jet to entrain m
E
since the ow evolves according to conservation of momentum for t

> 1. The jet

A. B. Olcay, P. S. Krueger
I
U
/(U
M
D
3
)
I
E
/
I
T
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Trapezoidal and L/D = 2.0
Triangular PS and L/D = 2.0
Triangular NS and L/D= 2.0
Trapezoidal L/D = 0.5
Trapezoidal and L/D = 1.0
Trapezoidal and L/D = 3.0
Fig. 14 I
E
/I
T
versus I
U
/(U
M
D
3
) for the trapezoidal velocity program with L/D = 0.5, 1.0, 2.0, 3.0 and triangular NS and PS
with L/D = 2.0 at t

= 1.0
Fig. 15 FTLE elds illustrating entrained uid in front of the nozzle for the trapezoidal velocity program with L/D = 2.0 at
t

= 0.0
momentum I
U
is chosen as the abscissa in Fig. 14 since it is an indicator of the bulk effect of the jet for a given
stroke ratio and velocity program combination. Despite variations in the maximum I
E
obtained (see Fig. 13),
Fig. 14 shows that I
E
/I
T
at t

= 1.0 is nearly insensitive to the generating conditions for nearly all cases tested
except for the triangular NS case. Triangular NS case produces an almost two times larger I
E
/I
T
value than
other cases due to the increased rate of entrainment occurring near the beginning of jet initiation for this case.
In all cases, however, I
E
/I
T
is small, indicating the momentum cost of entraining uid is generally small.
Returning to the overall ow evolution, Fig. 11 also shows the momentum of the uid inside the ellipsoidal
vortex ring bubble (see Fig. 7e), namely, I
R
I
EJ
+ I
E
. (The evolution of I
R
and I
EJ
for the other cases
simulated can be found in Olcay [25].) This ring momentum excludes the momentum associated with added
mass. Figure 11 shows that I
R
decreases for t

2.5. This is due to additional ambient uid entrainment by

vorticity diffusion as documented by Maxworthy [20, 21], Dabiri and Gharib [8] and Olcay and Krueger [26].
The uid entrained for t

2.5 initially resides in front of the nozzle near the jet centerline as illustrated by
the FTLE eld in Fig. 15 (note the ridge near r/D

= 0.2 for 0.0 x/D 4.0). As the ring translates forward,
this uid is swept over the top of the ring by Biot-Savart induction and entrained into the back of the ring. As
this entrainment follows the mechanism described by Maxworthly [20] and Dabiri and Gharib [8] and occurs
primarily after the ring is completely formed, we refer to this uid as uid entrained by viscous diffusion.
Although this additional uid steals some of the ejected uids momentum, the transfer occurs primarily after
Momentum evolution of ejected and entrained uid
Fig. 16 Streamline identifying the vortex ring boundary in a frame of reference moving with the vortex ring. The streamline is
shown with the ejected/entrained uid domains at t

= 2.5
the ring is formed since this additional uids velocity is essentially zero until it is encountered by the ring.
Hence, it does not significantly affect our momentum calculations during ring formation (i.e., t

< 2.5).
A key component contributing to the difference between I
R
and I
T
in Fig. 11 for t

> 1.0 is added mass.

As described by Saffman [32], Krueger [14], Krueger and Gharib [16], and Dabiri [7] added mass is set in
motion by the ring but external to the ring boundary. Added mass in this context is well dened only after
the ring is formed, at which point it may be computed using standard potential ow analysis for the closed
streamline encircling the ring in the frame of reference moving with the ring [18, 22]. In this case, the ring
boundary is reasonably well approximated as an ellipsoid. For an ellipsoid with major axis a and minor axes
b and c, the added mass coefcient is given by
=

0
2
0
, (15)
where
0

_
0
(abc)dz
(a
2
+z)k
z
, and k
z

_
(a
2
+ z)(b
2
+ z)(c
2
+ z) [15].
Then, is related to the momentum associated with added mass through
I
A
= m
R
U
R
, (16)
where m
R
m
EJ
+m
E
is the uid mass inside the ellipsoidal boundary of the ring (see Fig. 7e). For the case
shown in Fig. 11, =0.39 was computed using Eq. 15 with a and b = c determined as 0.90D and 0.62D,
respectively, from a streamline identifying the vortex ring boundary at t

= 2.5 (illustrated in Fig. 16). The

calculation was done at t

= 2.5 because the vortex ring is not formed for t

< 2.5 and additional ambient uid

is entrained for t

> 2.5, so t

= 2.5 gives best accuracy. The computed result for assuming an ellipsoidal
ring gives I
A
/I
T
= 0.23.
Based on Eq. 1 I
EJ
+ I
E
+ I
A
= I
T
. From Fig. 11, I
EJ
/I
T
= 0.38 and I
E
/I
T
= 0.19 at t

= 2.5, which
together with the calculated value for I
A
/I
T
only sums to 0.80 rather than 1.0. The discrepancy is in I
A
/I
T
,
which Fig. 11 shows should be 0.43 at t

= 2.5, not 0.23 as computed above, indicating that the use of

the ellipsoid assumption can have as much as 46% error even for a formed vortex ring. Recently, Dabiri [7]
computed the added mass coefcient of an empirically generated vortex ring (at Re
J
= 1,400 and L/D = 2.0
with an impulsive piston velocity program), and documented to be 0.72 for a fully formed vortex ring. The
use of 0.72 for gives I
A
/I
T
= 0.41 at t

= 2.5 providing reasonable agreement with 0.43 obtained from

Fig. 11.
The magnitude of I
A
in Fig. 11 should not go unnoticed. For the completely formed vortex ring (t

= 2.5)
more than 40% of the total impulse is associated with added mass, more than twice the value of the entrained
uid impulse at the same instant. Indeed, at t

= 2.5 almost half of the ejected uids momentum is spent to

set the added mass into motion. Figure 17 conrms the significance of I
A
for the cases studied. While all the
cases show about half of the total impulse is expended to set the added mass in motion, lowering stroke ratio or
gradual deceleration in the velocity program (triangular NS case) can increase the momentum of added mass
more than 20%. In the context of the equivalence between Eqs. 1 and 2, added mass clearly makes a much
larger contribution to nozzle exit over-pressure (I
P
) and the associated impulse benet than entrained mass.
A. B. Olcay, P. S. Krueger
I
U
/(U
M
D
3
)
I
A
/
I
T
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Trapezoidal and L/D = 2.0
Triangular PS and L/D= 2.0
Triangular NS and L/D= 2.0
Trapezoidal L/D = 0.5
Trapezoidal and L/D = 1.0
Trapezoidal and L/D = 3.0
Fig. 17 I
A
/I
T
versus I
U
/(U
M
D
3
) for the trapezoidal velocity program with L/D = 0.5, 1.0, 2.0, 3.0 and triangular NS and
PS with L/D = 2.0 at t

= 2.5
I
U
/(U
M
D
3
)

0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25

0
0.25
0.5
0.75
1
1.25
1.5
1.75
Trapezoidal and L/D = 2.0
Triangular PS and L/D = 2.0
Triangular NS and L/D= 2.0
Trapezoidal L/D = 0.5
Trapezoidal and L/D = 1.0
Trapezoidal and L/D = 3.0
Fig. 18 at t

= 2.5 versus I
U
/(U
M
D
3
) for all cases considered
In order to quantitatively assess the inuence of the entrained and added mass components on I
P
and the
associated inuence of the ring generating conditions (L/D and velocity program), it is observed that
I
P
= I
E
+ I
A
+ I
EJ
I
U
= (I
E
+ I
A
) (I
U
I
EJ
) (17)
(see Eqs. 1 and 2). Physically, I
E
+ I
A
is the momentum gained by the ambient uid and is the resulting
output of the jet action related to I
P
. Conversely, I
U
I
EJ
is the jet momentum contribution, or input, to
the ambient uid momentum gain. Hence, a useful measure of output to input for the ambient uid momentum
contribution to I
P
is
=
I
E
+ I
A
I
U
I
EJ
. (18)
Based on Eq. 17, it is desirable to maximize the performance metric . In particular, maximizing maximizes
the pressure impulse per unit of jet momentumcontributed to the acceleration of ambient uid (I
P
/(I
U
I
EJ
)).
Figure 18 plots evaluated at t

= 2.5 for all cases considered. From Fig. 18 it is observed that increases
as L/D is increased, indicating that higher a stroke ratio provides a stronger relative contribution to pressure
impulse and ambient uid momentum. The rate of increase slows, however, as L/D increases and the forma-
tion number [11] is approached, consistent with the results of Krueger and Gharib [16]. Comparing the open
symbols shows that the trapezoidal and triangular PS velocity programs exhibit higher values compared to
the triangular NS velocity program for the same L/D, indicating that rapid deceleration velocity programs
Momentum evolution of ejected and entrained uid
provide higher ambient uid impulse per amount of momentum transferred from ejected uid. This result is
somewhat surprising as a rapid initial acceleration might seem to have the stronger effect as it promotes rapid
acceleration of the ambient uid. The rapid deceleration of the trapezoidal and triangular PS velocity programs,
however, prevents as much I
U
from being used to accelerate this uid and is higher for these cases. In any
case, the effect of velocity program on is small for the cases considered.
4 Conclusions
Ejected and entrained impulse calculations were performed to investigate each terms contribution to the
formed ring impulse. CFD combined with LCS were utilized to identify initial uid boundaries. Then, these
uid boundaries were advected via fourth order Runge-Kutta integration scheme. The effect of velocity pro-
grams and L/D on the evolution and momentum of the ejected and entrained uid domains were investigated.
The velocity program used to generate a single jet pulse played a key role in the I
E
and I
A
results. Particu-
larly, I
E
was low while the jet was on, indicating that most of the ejected uids momentum was transferred to
the added mass during this phase and that the momentum input required from a starting jet to entrain ambient
uid is small. Most of the momentum gained by the entrained uid occurred after the jet began to decelerate.
This constrained I
E
at jet termination (t

= 1.0) to nearly the same value for all but the triangular NS case,
which initiated deceleration very early in the jet pulse. Once the jet was terminated, both I
E
and I
EJ
oscillated
as the ejected and entrained uid rotated about the center of vorticity in the ring. On the other hand, it was
observed that I
A
captured about 50% of total impulse of all the cases studied when the ring was formed. It
was also determined that lower L/Ds and/or gradual deceleration in the velocity program could increase the
momentum of added mass more than 20% indicating that the effect of velocity program and associated stroke
ratio can be significant in the determination of added mass momentum. Generally, changes in the velocity
program had a much larger effect than might otherwise be anticipated from a simple analysis based on the slug
model since the jet impulse and circulation predicted by the slug model were the same for the triangular NS
and PS programs.
Since I
P
= I
E
+ I
A
(I
U
I
EJ
), = (I
E
+ I
A
)/(I
U
I
EJ
) was used as a performance metric to gage
the effectiveness of different cases in contributing to I
P
as a function of the vortex formation process. It was
observed that the rate of increase of slows as L/D increases and the formation number [11] is approached.
It was also observed that the triangular PS velocity program and higher L/D provided a larger , indicating a
larger contribution to I
P
for a given input I
U
I
EJ
in these cases.
An additional benet of this study is the identication of the initial shape and location of entrained uid
around the nozzle exit plane. This information can be useful for promoting efcient mixing of two reactive sub-
stances by indicating where a regent should be concentrated or what formation parameters (velocity programs
or L/D) should be adjusted to enhance the process.
While this study only investigates the impulse variation of starting jets, the many of the general trends
in I
E
and I
A
can be applicable to pulsed jets where the time duration between the pulses is long enough to
avoid strong vortex interaction between pulses [17]. Often applications involving pulsed jets such as pulsed jet
propulsion [3, 4, 23], the jet pulses are widely separated so that the ow between pulses is nearly undisturbed.
The present results can be expected to be representative of the trends observed in such cases.
Acknowledgements This material is based upon work supported by the National Science Foundation under Grant No. 0347958.
References
1. Auerbach, D.: Stirring properties of vortex rings. Phys. Fluids A 3, 13511355 (1991)
2. Ball, R.S.: Account of experiments upon the resistance of air to the motion of vortex rings. Philos. Mag. 342, 208210 (1871)
3. Bartol, I.K., Krueger, P.S., Thompson, J.T., Stewart, W.J.: Swimming dynamics and propulsive efciency of squids through-
out ontogeny. Integr. Comp. Biol. 48, 720733 (2008)
4. Bartol, I.K., Krueger, P.S., Stewart, W.J., Thompson, J.T.: Hydrodynamics of pulsed jetting in juvenile and adult brief squid
Lolliguncula brevis: evidence of multiple jet modes and their implications for propulsive efciency. J. Exp. Biol. 212,
18891903 (2009)
5. Burden, R.L., Faires, J.D.: Numerical Analysis. Brooks/Cole, Belmont (2001)
6. Choutapalli, I.M.: An Experimental Study of a Pulsed Jet Ejector. Ph.D. Dissertation, Florida State University, Tallahassee,
FL (2006)
7. Dabiri, J.O.: Note on the induced Lagrangian drift and added-mass of a vortex. J. Fluid Mech. 547, 105113 (2006)
A. B. Olcay, P. S. Krueger
8. Dabiri, J.O., Gharib, M.: Fluid entrainment by isolated vortex rings. J. Fluid Mech. 511, 311331 (2004)
9. Dabiri, J.O., Colin, S.P., Costello, J.H.: Fast-swimming hydromedusae exploit velar kinematics to form an optimal vortex
wake. J. Exp. Biol. 209, 20252033 (2006)
10. Didden, N.: On the formation of vortex rings: rolling-up and production of circulation. Z. Angew. Math. Phys. 30,
101116 (1979)
11. Gharib, M., Rambod, E., Shariff, K.: A universal time scale for vortex ring formation. J. Fluid Mech. 360, 121140 (1998)
12. Glezer, A.: The formation of vortex rings. Phys. Fluids 31, 35323542 (1988)
13. Heeg, R.S., Riley, N.: Simulations of the formation of an axisymmetric vortex ring. J. Fluid Mech. 339, 199211 (1997)
14. Krueger, P.S.: The Significance of Vortex Ring Formation and Nozzle Exit Over-pressure to Pulsatile Jet Propulsion. Ph.D.
Thesis, California Institute of Technology (2001)
15. Krueger, P.S.: An over-pressure correction to the slug model for vortex ring circulation. J. Fluid Mech. 545, 427443 (2005)
16. Krueger, P.S., Gharib, M.: The significance of vortex ring formation to the impulse and thrust of a starting jet. Phys.
Fluids 15, 12711281 (2003)
17. Krueger, P.S., Gharib, M.: Thrust augmentation and vortex ring evolution in a fully pulsed jet. AIAA 43, 792801 (2005)
18. Lamb, H.: Hydrodynamics. Dover Publications, New York (1932)
19. Lim, T.T., Nickels, T.B. : Vortex rings. In: Green, S.I. (ed.) Fluid Vortices, Kluwer, New York (1995)
20. Maxworthy, T.: The structure and stability of vortex rings. J. Fluid Mech. 51(part 1), 1532 (1972)
21. Maxworthy, T.: Some experimental studies of vortex rings. J. Fluid Mech. 81(part 3), 465495 (1977)
22. Milne-Thomson, L.M.: Theoretical Hydrodynamics. Dover, New York (1996)
23. Nichols, J.T., Moslemi, A.A., Krueger, P.S.: Performance of a self-propelled pulsed-jet vehicle.In: AIAA Fluid Dynamics
Conference, Seattle, WA, June, AIAA 2008-3720 (2008)
24. Nitsche, M., Krasny, R.: A numerical study of vortex ring formation at the edge of a circular tube. J. Fluid Mech. 276,
139161 (1994)
25. Olcay, A.B.: Measurement of Ambient Fluid Entrainment and Calculation of Vortex Ring Impulse During Vortex Ring
Formation. Ph.D. Dissertation, Southern Methodist University, Dallas, TX (2007)
26. Olcay, A.B., Krueger, P.S.: Measurement of ambient uid entrainment during laminar vortex ring formation. Exp. Fluids
44, 235247 (2008)
27. Peng, J., Dabiri, J.O.: A potential-ow, deformable-body model for uid-structure interactions with compact vorticity:
application to animal swimming measurement. Exp. Fluids 43, 655664 (2007)
28. Peng, J., Dabiri, J.O.: An overview of a Lagrangian method for analysis of animal wake dynamics. J. Exp. Biol. 211,
280287 (2008)
29. Peng, J., Dabiri, J.O., Madden, P.G., Lauder, G.V.: Non-invasive measurement of instantaneous forces during aquatic loco-
motion: a case study of the bluegill sunsh pectoral n. J. Exp. Biol. 210, 685698 (2007)
30. Pullin, D.I.: Vortex ring formation at tube and orice openings. Phys. Fluids 22, 401403 (1979)
31. Reynolds, O.: On the resistance encountered by vortex rings, and the relation between the vortex rings and the streamlines
of a disk. Nature 14, 477 (1876)
32. Saffman, P.G.: Vortex Dynamics. Cambridge University Press, Cambridge (1992)
33. Shadden, S.C., Lekien, F., Marsden, J.E.: Definition and properties of Lagrangian coherent structures from nite-time
Lyapunov exponents in two-dimensional aperiodic ows. Physica D 212, 271304 (2005)
34. Shadden, S.C., Dabiri, J.O., Marsden, J.E.: Lagrangian analysis of uid transport in empirical vortex ring ows. Phys.
Fluids 18, 047105/11 (2006)
35. Shadden, S.C., Katija, K., Rosenfeld, M., Marsden, J.E., Dabiri, J.O.: Transport and stirring induced by vortex formation.
J. Fluid Mech. 593, 315331 (2007)
36. Shariff, K., Leonard, A.: Vortex rings. Annu. Rev. Fluid Mech. 24, 235279 (1992)
37. Shuster, J.M., Smith, D.R.: Experimental study of the formation and scaling of a round synthetic jet. Phys. Fluids
19, 045109 (2007)