Swimming

Swimming
FROM DRAG-BASED TO LIFT-BASED PROPULSION

633
100
5
o
O
a.
CD
in
o
O
100 1000 10000 100000
Mass (kg)
o Human
A Mink
Q Muskrat
Sea otter (surface)
Sea otter (submerged)
Seal
A Sea Lion
Whale
Fish
FIG 3 Relationship between Cost of Transport (CT) and body mass. Open symbols represent CT values for
drag-based propulsion when surface swimming; closed symbols represent undulatory and lift-based propulsion
during submerged swimming. The solid line represents the extrapolated minimum CT for fish (Davis et al.,
1985) Data from Costello and Whitlow (1975); Kruse (1975); 0ritsland and Ronald (1975); P. E. DiPrampero,
personal communication (1979); Fish (1982); Sumich (1983); Williams (1983); Innes (1984); Davis et al. (1985);
Worthy et al. (1987); Feldkamp (1987a); Williams (1989); Williams et al. (1992).
move a unit mass a given distance and CT
is inversely proportional to efficiency
(Tucker, 1970). Drag-based swimmers have
higher values of CT compared to lift-based
swimmers of similar body mass (Fig. 3).
Compared to the extrapolated line for min-
imum CT of fish (Davis et al, 1985), pad-
dling mammals have values of CT 10-25
times greater than fish of similar mass;
whereas lift-based swimmers are only 1.9-
4.6 times greater. A particularly significant
data set was collected by Williams (1989)
on sea otters showing the energetic advan-
tage of shifting from paddling at the water
surface to submerged swimming using un-
dulation. CT of surface paddling was 69%
greater than submerged swimming.
ASSOCIATION BETWEEN TERRESTRIAL AND
AQUATIC MOVEMENTS
Neuromotor patterns for locomotion ap-
pear to be conservative (Jenkins and Gos-
low, 1983; Goslow et al, 1989; Smith,
1994) and are independent of changes in
morphology and functional role (Smith,
1994). Kinematics and electromyographic
analysis of terrestrial locomotion in the rep-
tile Varanus and mammal Didelphis
showed similar activity patterns between
homologous muscles with the same attach-
ment sites (Jenkins and Goslow, 1983). Fur-
thermore, homologous muscles with differ-
ent attachments had similar activity pat-
terns. Neuromotor conservation was dem-
onstrated also between walking and flight
(Goslow et al, 1989; Dial et al, 1991). Al-
though muscles have been functionally re-
organized for flight, several muscles have
similar timing patterns in their respective
locomotor cycle with walking gaits (Dial et
al, 1991).
The basis for conservation of neuromotor
pattern is the existence of central pattern
generators. Central pattern generators are
neural networks in the central nervous sys-
tem that govern specific, repetitive motions
by providing the correct timing for mus-
cular contraction (Grillner, 1975, 1996; Pe-
ters, 1983; Pearson, 1993). Motor patterns
are executed without peripheral feedback,
although afferent feedback can modify the
motor programming and its output (Pear-
son, 1993).
The rhythmic nature of the central pat-

a
t

U
n
i
v
e
r
s
i
d
a
d
e

F
e
d
e
r
a
l

d
e

M
i
n
a
s

G
e
r
a
i
s

o
n

N
o
v
e
m
b
e
r

7
,

2
0
1
0
i
c
b
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

AMER. ZOOL., 36:628-641 (1996)
Transitions from Drag-based to Lift-based Propulsion in
Mammalian Swimming
1
FRANK E. FISH
Department of Biology, West Chester University,
West Chester, Pennsylvania 19383
SYNOPSIS. The evolution of fully aquatic mammals from quadrupedal,
terrestrial mammals was associated with changes in morphology and
swimming mode. Drag is minimized by streamlining body shape and
appendages. Improvement in speed, thrust production and efficiency is
accomplished by a change of swimming mode. Terrestrial and semiaquatic
mammals employ drag-based propulsion with paddling appendages,
whereas fully aquatic mammals use lift-based propulsion with oscillating
hydrofoils. Aerobic efficiencies are low for drag-based swimming, but
reach a maximum of 30% for lift-based propulsion. Propulsive efficiency
is over 80% for lift-based swimming while only 33% for paddling. In
addition to swimming mode, the transition to high performance propul-
sion was associated with a shift from surface to submerged swimming
providing a reduction in transport costs. The evolution of aquatic mam-
mals from terrestrial ancestors required increased swimming performance
with minimal compromise to terrestrial movement. Examination of mod-
ern analogs to transitional swimming stages suggests that only slight mod-
ification to the neuromotor pattern used for terrestrial locomotion is re-
quired to allow for a change to lift-based propulsion.
INTRODUCTION
Numerous paleontological, morphologi-
cal, and molecular studies have focused on
the phylogenetic relationships both within
clades of aquatic mammals and with their
terrestrial ancestors (Barnes et al., 1985;
Wyss, 1988, 1989; Berta, 1991; Milinkov-
itch et al., 1993; Adachi and Hasegawa,
1995). Such studies were helpful in pro-
ducing phylogenies and in elucidating the
historical sequence of structural changes.
However, phylogenetic studies are limited
and lack a functional approach to determine
the causal reasons for evolutionary change
(Lauder, 1990). Functional analysis pro-
vides an understanding at a mechanistic
level through measures of performance
(i.e., exercise metabolism, locomotor kine-
matics, maximal swimming speed). For
1
From the Symposium Aquatic Locomotion: New
Approaches to Invertebrate and Vertebrate Biome-
chanics presented at the Annual Meeting of the Society
for Integrative and Comparative Biology, 27-30 De-
cember 1995, at Washington, D.C.
aquatic mammals and their transitional
forms, functional analysis is therefore nec-
essary to determine the possible environ-
mental factors that produced the various
aquatic adaptations.
The similarity in morphology and swim-
ming mode between fish and cetaceans is
the quintessential example of evolutionary
convergence (Howell, 1930; Fish, 1993a).
This familiar textbook example illustrates
how similar functional requirements are
met by organisms that do not share a single
clade. Such convergence, with its resulting
homoplasy, is associated with similar con-
straints imposed on animals by the physical
environment and with selection for adapta-
tions for effective swimming (Fish, 1993a).
Adaptations associated with increased
aquatic behavior and swimming perfor-
mance evolved independently in several
clades of mammals (i.e., Cetacea, Pinnipe-
dia, Sirenia). Modern groups of aquatic
mammals display swimming adaptations
that optimize use of energy by reduction in
drag and improvement in thrust production
628

a
t

U
n
i
v
e
r
s
i
d
a
d
e

F
e
d
e
r
a
l

d
e

M
i
n
a
s

G
e
r
a
i
s

o
n

N
o
v
e
m
b
e
r

7
,

2
0
1
0
i
c
b
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Terminologia
238 IEEE JOURNAL OF OCEANIC ENGINEERING, VOL. 24, NO. 2, APRIL 1999
Fig. 1. Terminology used in the text to identify the ns and other features
of sh.
ns can also be characterized as either short-based or long-
based, depending on the length of their n base relative to the
overall sh length. The n dimensions normal and parallel to
the water ow are called span and chord, respectively.
Swimming involves the transfer of momentum from the sh
to the surrounding water (and vice versa). The main momen-
tum transfer mechanisms are via drag, lift, and acceleration
reaction forces. Swimming drag consists of the following
components:
1) skin friction between the sh and the boundary layer of
water (viscous or friction drag): Friction drag arises as
a result of the viscosity of water in areas of ow with
large velocity gradients. Friction drag depends on the
wetted area and swimming speed of the sh, as well as
the nature of the boundary layer ow.
2) pressures formed in pushing water aside for the sh to
pass (form drag). Form drag is caused by the distortion
of ow around solid bodies and depends on their shape.
Most of the fast-cruising sh have well streamlined
bodies to signicantly reduce form drag.
3) energy lost in the vortices formed by the caudal and
pectoral ns as they generate lift or thrust (vortex or
induced drag): Induced drag depends largely on the
shape of these ns.
The latter two components are jointly described as pressure
drag. Comprehensive overviews of swimming drag (including
calculations for the relative importance of individual drag
components) and the adaptations that sh have developed to
minimize it can be found in [11] and [12].
Like pressure drag, lift forces originate from water viscosity
and are caused by assymetries in the ow. As uid moves past
an object, the pattern of ow may be such that the pressure
on one lateral side is greater than that on the opposite. Lift
is then exerted on the object in a direction perpendicular to
the ow direction.
Acceleration reaction is an inertial force, generated by the
resistance of the water surrounding a body or an appendage
when the velocity of the latter relative to the water is changing.
Different formulas are used to estimate acceleration reaction
depending on whether the water is accelerating and the object
is stationary, or whether the reverse is true [13]. Acceleration
reaction is more sensitive to size than is lift or drag velocity
and is especially important during periods of unsteady ow
and for time-dependent movements [14], [15].
(a)
(b)
Fig. 2. (a) The forces acting on a swimming sh. (b) Pitch, yaw, and roll
denitions. (Adapted from Magnuson [11].)
The forces acting on a swimming sh are weight, buoyancy,
and hydrodynamic lift in the vertical direction, along with
thrust and resistance in the horizontal direction [Fig. 2(a)].
For negatively buoyant sh, hydrodynamic lift must be
generated to supplement buoyancy and balance the vertical
forces, ensuring that they do not sink. Many sh achieve this
by continually swimming with their pectoral ns extended.
However, since induced drag is generated as a side effect
of this technique, the balance between horizontal forces will
be disturbed, calling for further adjustments for the sh to
maintain a steady swimming speed. For a discussion on this
coupling of the forces acting on a swimming sh, see [11]. The
hydrodynamic stability and direction of movement are often
considered in terms of pitch, roll, and yaw [Fig. 2(b)]. The
swimming speed of sh is often measured in body lengths per
second (BL/s).
For a sh propelling itself at a constant speed, the mo-
mentum conservation principle requires that the forces and
moments acting on it are balanced. Therefore, the total thrust
it exerts against the water has to equal the total resistance it
encounters moving forward. Pressure drag, lift, and accelera-
tion reaction can all contribute to both thrust and resistance.
However, since lift generation is associated with the inten-
tional movement of propulsors by sh, it only contributes to
resistance for actions such as braking and stabilization rather
then for steady swimming. Additionally, viscous drag always
contributes to resistance forces. Finally, body inertia, although
not a momentum transfer mechanism, contributes to the water
resistance as it opposes acceleration from rest and tends to
maintain motion once begun. The main factors determining the
relative contributions of the momentum transfer mechanisms
to thrust and resistance are: 1) Reynolds number; 2) reduced
frequency; and 3) shape [15].
The Reynolds number (Re) is the ratio of inertial over
viscous forces, dened as
resistncia
peso
flutuao + sustentao hidrodinmica
propulso
arfagem (pitch)
guinada (yaw)
rolagem (roll)
Terminologia
Flutuao
Flutuao
Flutuao
Nadar envolve transferncia de movimento para a gua (e vice e versa) na forma de
arrasto, sustentao e foras de reao.
Arrasto
Arrasto viscoso
Arrasto decorrente da forma
Perda de energia devido aos vrtices
Sustentao Foras de reao
SFAKIOTAKIS et al.: REVIEW OF FISH SWIMMING MODES FOR AQUATIC LOCOMOTION 239
Fig. 3. Diagram showing the relative contribution of the momentum transfer
mechanisms for swimming vertebrates, as a function of Re. The shaded area
corresponds to the range of adult sh swimming. (Adapted from Webb [15].)
where is a characteristic length (of either the sh body
or the propulsor), is the swimming velocity, and is the
kinematic viscosity of water. In the realm of Re typical of adult
sh swimming (i.e., ), inertial forces are
dominant and viscous forces are usually neglected. At those
Re, acceleration reaction, pressure drag, and lift mechanisms
can all generate effective forces (Fig. 3).
The reduced frequency indicates the importance of un-
steady (time-dependent) effects in the ow and is dened
as
where is the oscillation frequency, is the characteristic
length, and is the swimming velocity. The reduced fre-
quency essentially compares the time taken for a particle of
water to traverse the length of an object with the time taken
to complete one movement cycle. It is used as a measure of
the relative importance of acceleration reaction to pressure
drag and lift forces. For , the movements considered
are reasonably steady and acceleration reaction forces have
little effect. For , all three mechanisms of
force generation are important, while for larger values of
acceleration reaction dominates. In practice, for the great
majority of swimming propulsors, the reduced frequency rarely
falls below the 0.1 threshold [15].
Finally, the shape of the swimming sh and the specic
propulsor utilized largely affect the magnitude of the force
components. The relationship is well documented for steady-
state lift and drag forces, but relatively little work has been
done on the connection between shape and acceleration reac-
tion.
A common measure of swimming efciency is Froude
efciency , dened as
where is the mean forward velocity of the sh, is the
time-averaged thrust produced, and is the time-averaged
power required.
B. Main Classications
Fish exhibit a large variety of movements that can be char-
acterized as swimming or nonswimming. The latter include
specialized actions such as jumping, burrowing, ying, and
gliding, as well as jet propulsion, the description of which is
beyond the scope of this paper. Swimming locomotion has
been classied into two generic categories on the basis of the
movements temporal features [16]:
1) Periodic (or steady or sustained) swimming, character-
ized by a cyclic repetition of the propulsive movements.
Periodic swimming is employed by sh to cover rela-
tively large distances at a more or less constant speed.
2) Transient (or unsteady) movements that include rapid
starts, escape maneuvers, and turns. Transient move-
ments last milliseconds and are typically used for
catching prey or avoiding predators.
Periodic swimming has traditionally been the center of scien-
tic attention among biologists and mathematicians. This has
mainly been because, compared to sustained swimming, ex-
perimental measurements of transient movements are difcult
to set up, repeat, and verify. Therefore, periodic swimming
will inevitably be the main focus of this paper. However,
given the signicant aspects of locomotion associated with
transient movements, which provide sh with unique abilities
in the aquatic environment and the more recent interest among
scientists in describing them, reference will also be made to
transient propulsion where possible.
The classication of swimming movements presented here
adopts the (expanded) nomenclature originally put forth by
Breder in [17]. Breders nomenclature has recently been criti-
cized as oversimplied and ill-dened (see, for example, [18]
and [19]) in describing sh swimming. Nevertheless, since we
are mainly concerned with descriptions of the sh propulsors,
on which Breders classication is based, it serves as a conve-
nient reference frame, provided its limitations are held in mind.
The interested reader is referred to [19], where a more holistic
classication scheme of swimming is proposed, relating the
swimming propulsors, kinematics, locomotor behavior, and
muscle ber used to the notion of swimming gaits.
Most sh generate thrust by bending their bodies into a
backward-moving propulsive wave that extends to its caudal
n, a type of swimming classied under body and/or caudal
n (BCF) locomotion. Other sh have developed alternative
swimming mechanisms that involve the use of their median
and pectoral ns, termed median and/or paired n (MPF)
locomotion. Although the term paired refers to both the
pectoral and the pelvic ns (Fig. 1), the latter (despite provid-
ing versatility for stabilization and steering purposes) rarely
contribute to forward propulsion and no particular locomotion
mode is associated with them in the classications found in
literature. An estimated 15% of the sh families use non-BCF
modes as their routine propulsive means, while a much greater
number that typically rely on BCF modes for propulsion
employ MPF modes for maneuvering and stabilization [18].
A further distinction, and one that is common in literature,
made for both BCF and MPF propulsion is on the basis of
the movement characteristics: undulatory motions involve the
passage of a wave along the propulsive structure, while in
oscillatory motions the propulsive structure swivels on its base
without exhibiting a wave formation. The two types of motion
should be considered a continuum, since oscillatory move-
ments can eventually be derived from the gradual increase of
the undulation wavelength. Furthermore, both types of motion
Medida usual de velocidade de um peixe:
[comprimento do corpo/segundo]
Fatores importantes para o movimento de um peixe:
1. Nmero de Reynolds, Re;
2. Frequncia do movimento;
3. Forma.
Nmero de Reynolds
Fig. 3. Diagram showing the relative contribution of the momentum transfer
mechanisms for swimming vertebrates, as a function of Re. The shaded area
corresponds to the range of adult sh swimming. (Adapted from Webb [15].)
where is a characteristic length (of either the sh body
or the propulsor), is the swimming velocity, and is the
kinematic viscosity of water. In the realm of Re typical of adult
sh swimming (i.e., ), inertial forces are
dominant and viscous forces are usually neglected. At those
Re, acceleration reaction, pressure drag, and lift mechanisms
can all generate effective forces (Fig. 3).
The reduced frequency indicates the importance of un-
steady (time-dependent) effects in the ow and is dened
as
where is the oscillation frequency, is the characteristic
length, and is the swimming velocity. The reduced fre-
quency essentially compares the time taken for a particle of
water to traverse the length of an object with the time taken
to complete one movement cycle. It is used as a measure of
the relative importance of acceleration reaction to pressure
drag and lift forces. For , the movements considered
are reasonably steady and acceleration reaction forces have
little effect. For , all three mechanisms of
force generation are important, while for larger values of
acceleration reaction dominates. In practice, for the great
majority of swimming propulsors, the reduced frequency rarely
falls below the 0.1 threshold [15].
Finally, the shape of the swimming sh and the specic
propulsor utilized largely affect the magnitude of the force
components. The relationship is well documented for steady-
state lift and drag forces, but relatively little work has been
done on the connection between shape and acceleration reac-
tion.
A common measure of swimming efciency is Froude
efciency , dened as
where is the mean forward velocity of the sh, is the
time-averaged thrust produced, and is the time-averaged
power required.
B. Main Classications
Fish exhibit a large variety of movements that can be char-
acterized as swimming or nonswimming. The latter include
specialized actions such as jumping, burrowing, ying, and
gliding, as well as jet propulsion, the description of which is
beyond the scope of this paper. Swimming locomotion has
been classied into two generic categories on the basis of the
movements temporal features [16]:
1) Periodic (or steady or sustained) swimming, character-
ized by a cyclic repetition of the propulsive movements.
Periodic swimming is employed by sh to cover rela-
tively large distances at a more or less constant speed.
2) Transient (or unsteady) movements that include rapid
starts, escape maneuvers, and turns. Transient move-
ments last milliseconds and are typically used for
catching prey or avoiding predators.
Periodic swimming has traditionally been the center of scien-
tic attention among biologists and mathematicians. This has
mainly been because, compared to sustained swimming, ex-
perimental measurements of transient movements are difcult
to set up, repeat, and verify. Therefore, periodic swimming
will inevitably be the main focus of this paper. However,
given the signicant aspects of locomotion associated with
transient movements, which provide sh with unique abilities
in the aquatic environment and the more recent interest among
scientists in describing them, reference will also be made to
transient propulsion where possible.
The classication of swimming movements presented here
adopts the (expanded) nomenclature originally put forth by
Breder in [17]. Breders nomenclature has recently been criti-
cized as oversimplied and ill-dened (see, for example, [18]
and [19]) in describing sh swimming. Nevertheless, since we
are mainly concerned with descriptions of the sh propulsors,
on which Breders classication is based, it serves as a conve-
nient reference frame, provided its limitations are held in mind.
The interested reader is referred to [19], where a more holistic
classication scheme of swimming is proposed, relating the
swimming propulsors, kinematics, locomotor behavior, and
muscle ber used to the notion of swimming gaits.
Most sh generate thrust by bending their bodies into a
backward-moving propulsive wave that extends to its caudal
n, a type of swimming classied under body and/or caudal
n (BCF) locomotion. Other sh have developed alternative
swimming mechanisms that involve the use of their median
and pectoral ns, termed median and/or paired n (MPF)
locomotion. Although the term paired refers to both the
pectoral and the pelvic ns (Fig. 1), the latter (despite provid-
ing versatility for stabilization and steering purposes) rarely
contribute to forward propulsion and no particular locomotion
mode is associated with them in the classications found in
literature. An estimated 15% of the sh families use non-BCF
modes as their routine propulsive means, while a much greater
number that typically rely on BCF modes for propulsion
employ MPF modes for maneuvering and stabilization [18].
A further distinction, and one that is common in literature,
made for both BCF and MPF propulsion is on the basis of
the movement characteristics: undulatory motions involve the
passage of a wave along the propulsive structure, while in
oscillatory motions the propulsive structure swivels on its base
without exhibiting a wave formation. The two types of motion
should be considered a continuum, since oscillatory move-
ments can eventually be derived from the gradual increase of
the undulation wavelength. Furthermore, both types of motion
Re =
LU
comprimento caracterstico
viscosidade cinemtica
velocidade
Frequncia do movimento
= 2
fL
U
frequncia reduzida
velocidade
frequncia de oscilao
comprimento caracterstico
0,1 < < 0, 4
Frequncia reduzida compara o tempo que uma partcula de gua leva para percorrer o comprimento do objeto com o tempo para o
objeto completar um movimento cclico
Eficincia de nado (eficincia de Froude)
Forma
=
T U
P
propulso mdia produzida
em um intervalo de tempo
velocidade
potncia mdia requerida
em um intervalo de tempo
Swimming Non Swimming
Movimentos
cavar saltar
voar
seguir a correnteza
propulso a jato transiente permanente
Movimentos que incluem disparos,
manobras de fuga e guinadas
sbitas. Estes movimentos duram
milsimos de segundos e acontecem
tipicamente para capturar presas ou
fugir de predadores.
Nado permanente, peridico ou sustentado
caracteriza-se por uma repetio cclica de
movimentos de propulso. O nado
permanente empregado pelo peixe para
cobrir longas distncias a velocidade mais
ou menos constante.
Este movimentos so especializados demais para uma classificao
genrica, pois so particulares de algumas espcies.
Movimento de deslocamento
na gua (Swimming)
Corpo e/ou barbatana caudal
(body and/or caudal fin) BCF
Barbatana medial e/ou barbatanas
duplas (medial and/or paired fin) MPF
85% das espcies de peixes utilizam
esta modalidade de movimento para
se deslocar.
15% das espcies de peixes utilizam esta modalidade de movimento para deslocamento. Este
tipo de movimento utilizado pela maioria das espcies para manobras e estabilizao
Movimento ondulatrio envolve a passagem
de uma onda ao longo da estrutura que est
executando a propulso.
Movimento oscilatrio da estrutura de
propulso envolve o movimento a partir de um
ponto fixo sem exibir movimento de onda.
Movimentos de Propulso
Propulso BCF Propulso MPF
Oscilaes
Ondulaes
Oscilaes das barbatanas Ondulaes das barbatanas
Movimentos transientes Nado permanente
Acelerao Cruzeiro Manovrabilidade
(a)
(b)
Fig. 5. Swimming modes associated with (a) BCF propulsion and (b) MPF propulsion. Shaded areas contribute to thrust generation. (Adapted from
Lindsey [10].)
Fig. 6. Thrust generation by the added-mass method in BCF propulsion.
(Adapted from Webb [20].)
(a) (b) (c) (d)
Fig. 7. Gradation of BCF swimming movements from (a) anguilliform,
through (b) subcarangiform and (c) carangiform to (d) thunniform mode.
(Taken from Lindsey [10].)
locomotion. Similar movements are observed in the sub-
carangiform mode (e.g., trout), but the amplitude of the
undulations is limited anteriorly, and increases only in the
posterior half of the body [Fig. 7(b)]. For carangiform swim-
ming, this is even more pronounced, as the body undulations
are further conned to the last third of the body length
[Fig. 7(c)], and thrust is provided by a rather stiff caudal n.
Carangiform swimmers are generally faster than anguilliform
or subcarangiform swimmers. However, their turning and
accelerating abilities are compromised, due to the relative
rigidity of their bodies. Furthermore, there is an increased
tendency for the body to recoil, because the lateral forces are
concentrated at the posterior. Lighthill [24] identied two main
morphological adaptations that increase anterior resistance in
order to minimize the recoil forces: 1) a reduced depth of
the sh body at the point where the caudal n attaches to
the trunk (referred to as the peduncle, see Fig. 1) and 2) the
concentration of the body depth and mass toward the anterior
part of the sh.
Thunniform mode is the most efcient locomotion mode
evolved in the aquatic environment, where thrust is generated
by the lift-based method, allowing high cruising speeds to be
maintained for long periods. It is considered a culminating
point in the evolution of swimming designs, as it is found
among varied groups of vertebrates (teleost sh, sharks, and
marine mammals) that have each evolved under different
circumstances. In teleost sh, thunniform mode is encountered
in scombrids, such as the tuna and the mackerel. Signicant
lateral movements occur only at the caudal n (that produces
more than 90% of the thrust) and at the area near the narrow
peduncle. The body is well streamlined to signicantly reduce
pressure drag, while the caudal n is stiff and high, with a
crescent-moon shape often referred to as lunate [Fig. 7(d)].
Despite the power of the caudal thrusts, the body shape and
mass distribution ensure that the recoil forces are effectively
minimized and very little sideslipping is induced. The design
of thunniform swimmers is optimized for high-speed swim-
ming in calm waters and is not well-suited to other actions such
as slow swimming, turning maneuvers, and rapid acceleration
from stationary and turbulent water (streams, tidal rips, etc.).
BCF
Ondulatrio
Oscilatrio
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
MPF
Movimentos ondulatrios
das barbatanas
Movimentos oscilatrios
das barbatanas
peitoral dorsal anal anal e dorsal
As reas marcadas em cinza representam aquelas que contribuem para a propulso
BCF Ondulatrio
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
Mtodo da massa adicionada
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
Mtodo da vorticidade
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
BCF Ondulatrio
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
Mtodo da massa
adicionada
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
Fig. 9. The ow eld around the body of a carangiform swimmer, as
obtained from PIV data. The symbols and correspond to pressure and
suction zones that form the basis of the undulating pump mechanism.
(Adapted from M uller et al. [30].)
with the bound vortices created by the tail movements, forming
the discrete vortices shed in the wake. Based on their particle
image velocimetry (PIV) data, the authors concluded that about
a third of the total energy shed to the water is provided by the
anterior body.
A similar mechanism is proposed by Triantafyllou and
Triantafyllou [33] as a means of recapturing energy and
reducing the apparent drag of swimming, by at least 50%
[5]. This could provide an explanation to Grays Paradox
that has long troubled scientists. Gray [34] estimated the
power requirements for a cruising dolphin, assuming that
its drag can be approximated by that of a rigid model and
considering turbulent ow. The calculations indicated that the
power required exceeded the estimates of muscle power output
by a factor of seven, thus the paradox. Despite the numerous
adjustments and corrections of Grays original estimations
and the varied explanations suggested (see [11][13], [21]),
no denite conclusions have been drawn on the matter. The
new hypothesis is supported by efciency measurements of an
articulated robot swimming by body undulations (the Robo-
Tunasee [5] and [6]), ow visualizations of swimming sh
[35], as well as experiments and simulations with oscillating
foils extracting energy from incoming vortices [35][37]. The
implication of this theory is that the apparent swimming drag
is actually lower for an undulating body than that of the
rigid equivalent. This is in complete contrast to the traditional
assumptions that estimate the apparent swimming drag to be
three to ve times that of the rigid-body equivalent, due to
the increased friction drag and inertial recoil energy losses
associated with BCF undulatory motions. There is a need to
reexamine existing data, assumptions, and trends observed in
nature and assess them in the context of these new theoretical
developments.
Vorticity control mechanisms were originally proposed in
the early 1970s in the context of sh schooling behavior often
observed in scombrids. As each vortex is shed by individual
swimming sh, it induces a water motion that is opposite
to the swimming direction immediately behind the sh, but
in the swimming direction at the sides [Fig. 8(c)]. Therefore,
a sh situated laterally midway between the two sh of the
preceding column (Fig. 10), rather than directly behind one of
them, avoids having to overcome increased incoming ow. A
channeling effect has also been suggested, provided the sh
stay close together, to utilize the favorable ow at the sides
of the vortex-street. The advantage is greater when sh in the
same column swim in antiphase with the neighbors. These
Fig. 10. Plan view of a horizontal layer of a sh school, showing its
diamond-shaped building block structure. The conguration is described by
the wake width , the vortex spacing , and the lateral distance amongst
sh of the same column. (Adapted from Weihs and Webb [16].)
requirements point to an elongated diamond-shape pattern as
the basic optimum structure in sh schools (Fig. 10). Evidence
from aerial photographs of schooling scombrids support this
prediction. The hydrodynamic benets of schooling seem to
vary for each column, as partial or complete cancellation of
vortices can occur. Magnuson [11] estimates average energy
savings of 10%20% from schooling. Details can be found in
[11], [16], and [38].
D. Mathematical Analysis
Scientists from varying backgrounds have attempted to
formulate mathematical models to describe the observed kine-
matics of sh. Work has been hindered by the inherent
variability and complexity encountered in natural processes,
limiting the accuracy and repeatability of experiments and
measurements compared to other areas of engineering.
Early resistive hydrodynamic models (see, for example,
[39]) were based on a quasi-static approach that uses steady-
state ow theory to calculate the uid forces for sequential
frames of the shs motion. Their applicability is restricted
to low Reynolds numbers, due to neglecting inertial forces
and the oversimplied assumptions concerning sh motions
and body shapes. Later models dealt with more realistic sh-
type motions, assuming an inviscid (frictionless) uid. Wu
[40] originally developed a two-dimensional (2-D) waving
plate theory, treating the sh as an elastic plate. Along with
the slender body theory that stems from aerodynamics, it
formed the basis for Lighthills elongated-body theory [41],
[42], which is well suited to subcarangiform and carangiform
modes. The ows induced by the undulating body are assumed
to cancel out over a tailbeat cycle and the mean thrust is
estimated from the trailing edge kinematics. The original
theory was extended by Lighthill in [43] to cater for sh
motions of arbitrary amplitude, leading to the large-amplitude
elongated-body theory that is better suited to carangiform
swimming, where the lateral motions of the caudal n are
large. Mechanical thrust power for a sh swimming at an
average speed is calculated [27] as
where
is the added mass per unit length ( is the trailing edge span
and the density of the water), while
is the rms value of the lateral speed of the trailing edge (
is the frequency of the caudal n oscillations and is their
amplitude). The velocity given to the water at the trailing
edge is obtained as
where is the velocity of the propulsive wave. Finally, is the
angle of the trailing edge to the lateral plane of motion. Filmed
sequences of the swimming sh are used to determine these
parameters. The hydromechanical efciency is calculated as
As the above equation shows, is never less than 0.5 (as
, while for ).
For examples of practical application of the large-amplitude
elongated body, see [44] and [45]. Lighthills work has been
further rened to include the effects of body elasticity [46],
recoil movements [47], centerline curvature, and the interac-
tion of the caudal n with the vortex sheets shed from dorsal
ns [48]. The importance of body thickness effects in relation
to thrust and drag have been studied in [49] and [50]. All
these analytical approaches have shed signicant light on the
morphology and swimming mode of sh. Large-amplitude
elongated-body theory has also been used by Weihs to study
the hydrodynamics of BCF turning maneuvers [51] and fast
starts [52]. The outlines of large-amplitude elongated-body
theory found in [44] and [18] are recommended as introductory
texts on the subject. Linear and nonlinear extensions of the
waving plate theory have also appeared in Tong et al. [53]
and Root and Long [54]. The latter allows the analysis of fast
starts as well as steady swimming.
Elongated-body theory cannot be applied to thunniform
mode, because the shape of the caudal and pectoral ns
violates the fundamental assumption of slenderness. The the-
ories that have been developed stem from work on oscillating
aerofoils and consider the caudal n independent from the rest
of the sh body. These are presented separately in Section
III-E. Models that integrate viscous and pressure drag with
acceleration reaction should provide further insights, particu-
larly for anguilliform locomotion, where viscous forces seem
to play a signicant role. This need has been set forth early [24]
and, in principle, viscous and inertial forces can be calculated
separately, the latter estimated using inviscid theory. Charac-
terizing the ow around the sh body is very complicated,
rendering the formulation of such a model problematical and
possibly impractical for application to and validation by actual
data [21].
In another hydrodynamic approach, the energy costs of
swimming are estimated indirectly by calculating the energy
shed into the wake, based on the size and circulation of the
discrete vortices [30]. Application of this method to PIV data
obtained for a swimming mullet yielded a propulsive efciency
greater than 90%.
Concerning ostraciiform locomotion, Blake in [55] consid-
ered the thrust force generated by a rigid tail oscillating, while
the sh body is held straight. He applied both elongated-body
theory and a reactive model of a nite circular oscillating
disc moving in its own plane and in a perfect uid, found in
[56]. Propulsive efciencies were calculated to be around 0.5,
considerably lower than those obtained for undulatory BCF
modes.
As a nal remark, numerical studies involving computa-
tional uid dynamics (CFD) techniques have lately appeared
in literature, exploiting the increased power of computers. The
objective is to calculate the ow patterns and pressure eld
around the undulating sh body and/or caudal n by solving
the NavierStokes equations in order to determine the forces
generated as a result of the momentum changes. The potential
benets in understanding the way the swimming body interacts
with water are immense, as many of the assumptions found
in analytical methods can, in theory at least, be dispensed
with. The increased computational task for such simulations
meant that initial attempts assumed 2-D ows or simplied
movements [57]. Recently. 3-D CFD models have emerged,
utilizing advanced computational techniques and the power of
supercomputers [28], [58], [59].
E. Elements of Lunate Tail Propulsion
The thunniform mode being a highly efcient method of
swimming has attracted much recent interest, due to its poten-
tial for providing articial systems with advanced propulsor
designs. The benets have already been demonstrated in the
form of the RoboTuna robotic sh [5] that was shaped after an
actual tuna and combined oscillating foil tail movements with
carangiform body kinematics (i.e., presenting more extensive
undulations than those encountered among actual scombrids).
Mean propulsive efciencies as high as 91% have been re-
ported for the RoboTuna. Its success spawned further work
in the area of swimming robots [1]. In [4], the use of a dual
apping foil device for propulsion and/or maneuvering of a
rigid cylinder-shaped body is demonstrated, investigating both
a clapping and a waving mode of operation. Work has
also been directed at the prospect of applying oscillating foil
propulsion to traditional sea-surface vessels (see [2] for a list
of references).
Fish swimming in the thunniform mode are characterized
by a stiff caudal n, shaped like a tapered hydrofoil of a
moderate sweepback angle with a curved leading edge and
a sharp trailing edge [Fig. 11(a)]. The caudal n performs
a combination of pitching and heaving motions, tracing an
oscillating path as the sh moves forward, characterized by
a peak-to-peak amplitude , a tail-beat frequency , and a
wavelength [Fig. 11(b)]. There are very small lateral move-
ments of the body, mainly concentrated near the penduncle
BCF
Ondulatrio Anguilliform
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
Modo eficiente Modo veloz
BCF
Ondulatrio Caringiform
anterior body.
developments.
[11], [16], and [38].
1549 Hydrodynamics of carangiform swimming
total drag reduction observed in these experiments is mainly due to
the ability of the undulatory body motion to drastically reduce the
form drag. Note that the simplified inviscid hydrodynamic models
(Lighthill, 1971; Fish et al., 1988; Fish, 1993) could not explain the
drag reduction observed in experiments since they did not considered
form drag and the assumption was that the thrust overcomes the
drag only due to friction.
As we conclude this section it is appropriate to discuss the
physical mechanisms that lead to the observed reduction in form
drag. It has been hypothesized (Triantafyllou et al., 2000) that in
undulatory swimming the travelling body wave contributes to a
decreased drag force by eliminating separation and suppressing
turbulence. This hypothesis was supported by early experiments
(Taneda and Tomonari, 1974), which visualized the flow past a
waving flat plate and showed that when the wave phase velocity V
is smaller than the flow velocity U, the boundary layer separates at
the back of the wave crest, while when V>U the boundary layer
does not separate. A more recent study (Shen et al., 2003) carried
out direct numerical simulation of flow past a waving plate and
confirmed the earlier findings (Taneda and Tomonari, 1974) by
showing that for V>U separation is indeed eliminated and drag is
reduced, relative to the stationary wavy wall, monotonically with
increasing V. Shen et al. also emphasized the relevance of their
waving flat plate results in understanding drag-reduction
mechanisms in fish-like swimming (Shen et al., 2003).
Our computed results also show that the ratio of the undulatory
wave phase velocity V to the swimming speed U is indeed a critical
parameter insofar as drag reduction is concerned. In our simulations
V exceeds U when St>0.22 and it is evident from Figs4 and 5 that
for the viscous flow simulations the drag force is first reduced below
that of the rigid body for St>0.25, i.e. when the condition V>U has
been satisfied. In Fig.6 we show instantaneous streamlines and
0 0.2 0.4 0.6 0.8 1.0 1.2
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Total drag
Form drag
Skin drag
Strouhal number
D
r
a
g

c
o
e
f
f
i
c
i
e
n
t
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
A Re=300
B Re=4000
Fig.5. Variation of the skin (friction) drag, form drag and total drag with
Strouhal number at two Reynolds numbers: Re=300 (A) and Re=4000 (B).
The drag forces are calculated using Eqn17.
Fig.6. Pressure contours and streamlines on the midplane of the fish
relative to a frame moving with the body wave phase speed V (Re=300).
(A) Rigid body (St=0). (B) Flow separates for St=0.1, U/V=2.11. (C) Flow
does not separate for St=0.3, U/V=0.7.
1550
pressure contours at the mid-plane of the virtual swimmer in the
frame of reference moving with the undulatory wave phase velocity
V for Re=300 and St=0 (rigid body), 0.1 (U/V=2.11) and 0.3
(U/V=0.7), respectively. Note that the moving frame of reference
is selected because in the case of a swimming fish flow separation
occurs relative to the undulating body and can only be visualized
clearly in the frame of reference that moves with the body wave
velocity V[see Shen et al. (Shen et al., 2003) for a detailed discussion
of this issue]. As seen from Fig.6A the flow around the rigid body
(St=0) does not separate, due to the streamlined body shape. As the
body begins to undulate, and as long as V is less than U, the flow
separates at the posterior of the body (see results in Fig.6B for
St=0.1) due to the fact that the undulatory body wave is such that
it acts to retard the near-wall flow relative to the free stream. The
onset of separation explains the initial increase of the form drag
force relative to the rigid body drag observed in Figs4 and 5. At
St sufficiently high for the condition V>U to be satisfied (St>0.22
in our case), however, separation is eliminated (see Fig.6C for
St=0.3) and the drag force is reduced below that of the rigid body
drag at the same Re. Under these conditions (i.e. V>U), the motion
of the undulating fish body is piston-like and acts to accelerate the
slower moving ambient fluid, thus creating a positive (stagnation)
pressure region at the posterior portion of the fish body, which
reduces the form drag; this is clearly evident in the pressure contours
shown in Fig.6C. It is of course important to note that the reduction
in the form drag at higher St is not for free. The fish has to beat its
tail faster to achieve drag reduction at higher St and thus needs to
expend more power to accomplish this. As discussed in the previous
section the lateral power cost of such drag reduction was found to
be higher than the inline power gained by the undulatory motion in
our simulations.
Three-dimensional wake structure
The wake of carangiform swimmers has been studied extensively
in the laboratory using particle image velocimetry (PIV), which can
provide the velocity field in several 2D planes (Muller et al., 1997;
Wolfgang et al., 1999; Nauen and Lauder, 2002). These experiments
showed the vortices in the wake of free swimming carangiform
fishes organize in a single row such that a jet flow is formed between
the vortices, which has been dubbed a reverse Karman street (Rosen,
1959).
In our simulations we also find a reverse Karman street wake
consisting of a single row of vortices for the self-propelled inviscid
flow case (St*=0.26), which is the case that corresponds more closely
(both in terms of Re and St) to the available in the literature
experiments with live carangiform swimmers. The simulated near-
wake velocity and vorticity fields in the horizontal and vertical planes
for this case are shown in Fig.7. The flow patterns shown in this
figure are very similar to those obtained experimentally (Nauen and
Lauder, 2002) using PIV on the horizontal and vertical planes near
the caudal fin of a swimming mackerel [see figs3 and 4 in Nauen
and Lauder (Nauen and Lauder, 2002), corresponding to Fig.7A
and B in the present study].
Our simulations have also revealed different wake patterns
depending on the Re and St numbers, as shown in Fig.8, which
depicts three such representative wake patterns. To facilitate the
classification of the various wake patterns observed in our
simulations let us introduce the following wake characterization
convention. (1) Depending on the direction of the common flow
between the wake vortices, the wake can be characterized as being
of thrust (reverse Karman street) or drag (regular Karman street)
type; (2) depending on the layout of the vortical structures the wake
can be characterized as consisting of single or double row vortices.
A single row wake can be either of drag or thrust type, as shown
in Fig.8A (Re=4000 and St=0.2) and Fig.8B (Re= and St=0.26),
respectively. The main characteristic of this wake pattern is that it
remains confined within a relatively narrow parallel strip that is
centered on the axis of the fish body and consists of Karman-street
like vortices. A double row wake that is of thrust type is shown in
Fig.8C (Re=4000, St=0.7). This wake pattern is distinctly different
than the single row wake as it is characterized by the lateral
divergence and spreading of the vortices away from the body in a
wedge-like arrangement.
Our computed results show that for fixed Re both the single and
double row wake structures can emerge, depending on the St number.
Typically at low St the single row wake structure is observed (see
Fig.8A,B), while at high St the wake splits laterally and the double
row pattern emerges (Fig.8C). The dependence of the wake
structure on the St is to be expected since by definition (see Eqn2)
the St can be viewed as the ratio of the average lateral tail velocity
to the axial swimming velocity. Therefore, at high St the vortices
shed by the tail tend to have a larger lateral velocity component,
which advects them away from the centerline causing them to spread
in the lateral direction.
The exact value of St at which the transition from the single to
the double row wake structure occurs depends on the Re. Due to
limitations in the computational resources at our disposal we did
not attempt to precisely calculate the wake transition St as a function
I. Borazjani and F. Sotiropoulos
Fig.7. Calculated out-of-plane vorticity contours with velocity vectors for the
Re=, St=0.26 case (A) on the horizontal (x1x3) mid-plane and (B) the
vertical (x2x3) plane.
BCF
Ondulatrio Caringiform
anterior body.
developments.
[11], [16], and [38].
1551 Hydrodynamics of carangiform swimming
of Re. Our results, however, do suggest that for Re=300 the wake
transition occurs within 0.3<St<0.6 while for Re=4000 and it
occurs at somewhat lower St in the range 0.3<St<0.5.
To illustrate the effect of Re on the wake structure, we plot in
Fig.9 the instantaneous vorticity field and streamlines on the mid
plane for all three Re for St=0.3. As one would anticipate, for lower
Re the thickness of the viscous regions around the body and overall
width of the wake become greater as the diffusive effects of the
viscous forces begin to dominate. For the Re=300 case, the wake
is of drag type with a single row of vortices. At Re=4000 the wake
is still of drag type but it is clearly more disorganized and complex
than the Re=300 case. The wake pattern is of single row type but
its emerging complexity signals its upcoming (for St>0.3) transition
from the single to double row structure. For the inviscid case the
wake also consists of single row vortices but it is now of thrust
type. In comparison with the inviscid wake shown in Fig.8B for
St=0.26, the wake for the St=0.3 case has become more complex.
The vortices have intensified, adjacent layers of positive and
Fig.8. Instantaneous streamlines with vorticity contours showing (A) a single
row regular Karman street (Re=4000, St=0.2); (B) singe row reverse Karman
street (Re=, St=0.26); and (C) double row reverse Karman street (Re=4000,
St=0.7). The red arrows show the general direction of the wake flow.
Fig.9. Instantaneous streamlines and vorticity contours at the horizontal
mid-plane for St=0.3 highlighting the effect of Reynolds number on the
wake structure. (A) Re=300; (B) Re=4000 (middle); (C) Re=.
1552
negative vorticity are observed in the wake, and two cores of high
vorticity emerge within the primary wake vortices especially at some
distance downstream of the tail. Once again, the emergent
complexity of the wake is also suggestive of the transition to the
double row structure that will occur at somewhat higher St.
To visualize the 3D structure of the wake we show in Fig.10,
Fig.11 and Fig.12 instantaneous iso-surfaces of the q-criterion
(Hunt et al., 1988) for Re=300, 4000 and , respectively. The
quantity q is defined as q=G(
2
S
2
), where S and denote the
symmetric and antisymmetic parts of the velocity gradient,
respectively, and
.
is the Euclidean matrix norm. According to
Hunt et al. (Hunt et al., 1988), regions where q>0, i.e. regions where
the rotation rate dominates the strain rate, are occupied by vortical
structures. For each Re we show two St corresponding to the single
and double row vortex patterns. The 3D wake structure of
carangiform swimmers has been approximately reconstructed in the
past from the results of PIV experiments using the velocity data in
conjunction with the Helmhotz theorem for inviscid barotropic
fluids. For carangiform swimmers with a single vortex row wake
structure, a series of connected vortex rings has been suggested
(Lighthill, 1969), while for anguilliform (eel-like) swimmers, which
exhibit a double vortex row, two disconnected vortex rings have
been hypothesized (Muller et al., 2001). As seen in Figs1012, both
types of 3D vortical structures are observed in the simulations
depending on the St. Nevertheless, the rather simple wake structure
that was hypothesized in previous experiments is observed in our
simulations only for the lowest Re case (Fig.10). For this case, the
double row wake consists of laterally dislocated vortex loops while
the single row wake consists of a train of inverted hairpin-like
vortices braided together such that the legs of each vortex are
attached to the head of the preceding vortex. This wake structure
resembles the 3D wake structure observed in laboratory experiments
with vibrating spheres (Govardhan and Williamson, 2005) and
flapping foils (Buchholz and Smits, 2006). For Re=4000 (see
Fig.11), the wake structure becomes significantly more complex.
The single row wake consists of braided hairpins as for the Re=300
case, but the hairpins in this case have longer and more stretched
legs and more slender heads. In addition, smaller-scale vortical
structures attaching to the hairpin vortices are beginning to emerge
in the wake. The double row pattern is no longer made up of two
rows of simple vortex loops but consists of very complex and highly
3D coherent structures connected together through complex
columnar structures. To further appreciate the enormous complexity
of the wake structure at this combination of Re and St, we show in
Fig.13 different views of the q iso-surfaces superimposed to out-
of-plane vorticity contours. Finally, for the inviscid case the single
row wake consists of connected vortex loops, which are more flat
in shape and stretched in the streamwise direction. The double row,
Fig.10. Three-dimensional (3D) vortical structures visualized using the
q-criterion showing 3D wake structures simulated for the Re=300 case.
(A) Double row wake at St=1.2; (B) single row wake at St=0.3.
Fig.11. Three-dimensional (3D) vortical structures visualized using the
q-criterion showing 3D wake structures simulated for the Re=4000 case.
(A) Double row wake at St=0.7; (B) single row wake at St=0.2.
BCF
Ondulatrio SubCaringiform
Obstculo a montante Fluxo livre
Ostraciiform locomotion is the only purely oscillatory BCF
mode. It is characterized by the pendulum-like oscillation of
the (rather stiff) caudal n, while the body remains essentially
rigid. Fish utilizing ostraciiform mode are usually encased in
inexible bodies and forage their (usually complex) habitat
using MPF propulsion [25]. Caudal oscillations are employed
as auxiliary locomotion means to aid in thrust production at
higher speeds, to ensure that the body remains adequately
rigid, or to aid prey stalking [10]. Despite some supercial
similarities with thunniform swimmers, the hydrodynamic
adaptations and renements found in the latter are missing
in ostraciiform locomotion, which is characterized by low
hydrodynamic efciency.
B. Body Undulations and Friction Drag
Swimming viscous drag is calculated using the standard
Newtonian equation
where is the drag coefcient (which depends on the
Reynolds number and the nature of the ow), is the wetted
surface area, and is the water density. Flexing the body to
achieve propulsion is expected to increase viscous drag by a
factor of compared to that for an equivalent rigid body, since
the motion of the propulsive elements increases their velocity
with respect to the surrounding uid. This is known as the
boundary layer thinning effect, as lateral body movements
reduce the boundary layer, resulting in increased velocity
gradients and, hence, shear stress. Exactly how extensive
the increase in viscous drag is has long troubled scientists.
Originally, indirect estimations suggested (see, for example
[26] and [24]) that lies between 4 and 9. Webb in [27]
indicates that this must be a signicant overestimation, placing
a greater importance on the energy losses arising from recoil
forces. A value of for a swimming tadpole has been
calculated in [28] using three-dimensional (3-D) numerical
simulation, at . The rather low Re prohibits safe
application of this value of to adult sh swimming. In
the same study, it is shown that the relative amplitude of
body undulations in tadpoles is signicantly larger than those
observed in sh. When the model was adapted to swim using
the kinematics of a saithe, was reduced to 1.12, stressing
the connection between large lateral motions and increased
friction drag [28].
C. Wake Structure and Generation
The wake left behind the tail of undulatory BCF swimmers
is a staggered array of trailing discrete vortices of alternating
sign, generated as the caudal n moves back and forth. A
jet ow with alternating direction between the vortices is also
visible [Fig. 8(c)]. The structure of the wake is of a thrust-type,
i.e., has a reversed ratational direction compared to the well-
documented drag-producing Karman vortex street. The latter
is typically observed in the wake of bluff (nonstreamlined)
objects [Fig. 8(a)] for a specic range of Reynolds numbers
(roughly ), but also in the wake of stationary
[Fig. 8(b)] or low-frequency-heaving aerfoils (see [29]).
(a) (b)
(c)
Fig. 8. The Karman street generates a drag force for either (a) bluff or (b)
streamlined bodies, placed in a free stream. (c) The wake of a swimming sh
has reverse rotational direction, associated with thrust generation.
The main parameter characterizing the structure of such
wakes is the Strouhal number, dened, for a sh swimming
by BCF movements, as
where is the tail-beat frequency in hertz, is the wake
width (usually approximated as the tail-beat peak-to-peak
amplitude, and is the average forward velocity. The Strouhal
number is essentially the ratio of unsteady to inertial forces.
Triantafyllou et al. [29] concluded that, in oscillating foils,
thrust development is optimal for a specic range of St (namely
). Existing data on a number of sh species
revealed that, for high-speed swimming, their calculated St
values lie within this predicted range. Interestingly, this was
valid for species representing not just thunniform (traditionally
associated with oscillating foils) but also subcarangiform and
carangiform modes, at a range of . These
results have placed increased signicance to vorticity effects
and established the Strouhal number as a prominent factor
when analyzing BCF modes. Detailed data on the morphology
of the wave shed behind a mullet (swimming at )
can be found in [30].
The generation mechanism of this wake structure is still
unclear, as a number of contradicting hypotheses have been
put forth. Lighthill [24] and Videler [18] support that the
reversed Karman street results exclusively from the tail move-
ments. This wake structure has indeed been observed behind
oscillating foils that were not attached to a body (see, for
example, [31]). Rosen [32] was among the rst to conduct ow
visualization experiments of carangiform sh and observed
attached vortices being generated by the anterior half of the
sh body. He proposed a vortex peg mechanism, whereby
sh thrust their body against these vortices, extracting their
rotational energy to move forward. Fully formed attached vor-
tices have not been observed in the more recent visualization
experiments. Rather, suction and pressure zones appear in the
ow pattern (Fig. 9).
M uller et al. supported an undulating pump mechanism
whereby these zones create a circulating ow around the
inection points of the body. The circulating ow propagates
along the body, and upon reaching the caudal n, it interacts
St =
fA
U
0, 25 < St < 0, 4
10
4
< Re <10
6
operate at similar St. Does this reect selection to constrain St, or is
it merely coincidental? To answer this, we rst provide two inde-
pendent conrmations that St is tightly constrained in cruising
ight. The rst conrmation comes from considering how St varies
when an animal is forced to y other than at its preferred speed. St
varies more in four individual zebra nches (Taenopygia guttata)
27
forced to y between 4 and 14 ms
21
than across all 42 species ying
at their preferred speeds (Fig. 2). Even excluding the lowest forced
ight speeds (which are lower than any preferred ight speed), the
standard deviation (s.d.) of St for the zebra nch across speeds is
more than twice that for all 42 species ying at their preferred
speeds. Flight speed affects St so strongly because wingbeat fre-
quency and amplitude are tightly constrained, presumably by
physiology and morphology. This makes it even more notable
that, when only cruise performance is considered, a sample span-
ning ve orders of body mass and three independent evolutionary
origins of ight shows less than half the variation in St found in four
individuals of a single species forced to y at different speeds.
A Monte Carlo analysis provides a second conrmation that St is
tightly constrained in cruising ight (Fig. 3). As frequency, speed
and amplitude all scale with body mass (m), if wing kinematics are
indeed tuned to optimize St, their residual variation should co-vary
appropriately to constrain St. We therefore regressed log(f ), log(U)
and log(A) separately against log(m) and calculated the tted values
for each species. We then randomly allocated 1 of the 42 residuals
from each of the regressions to each species without replacement
and calculated the s.d. of St in the resulting sample (n 42). We
repeated this procedure 50,000 times.
Figure 3 shows the distribution of the s.d. of resampled St.
Regressions and residual plots (Supplementary Fig. S1) showed
sufcient homoscedasticity for the procedure to be valid. Only 53 of
the 50,000 randomized combinations of residuals (,0.11%) had a
lower s.d. of St than the original sample. The actual s.d. of St (0.10)
was therefore signicantly smaller than expected by chance
(P 0.001) from the allometry of f, U and A. This is very strong
evidence that St is tightly constrained during cruising ight. It
would not be unexpected to nd St tightly constrained in similar
species, because similar morphological and physiological con-
straints usually produce dynamic similarity
1
; however, it is surpris-
ing to nd St tightly constrained in this morphologically and
physiologically disparate sample, unless St is constrained by one
uniting factoraerodynamics.
The propulsive efciency of an isolated apping foil usually peaks
at St < 0.3 (refs 37), therefore we used a two-tailed t-test to check
that actual mean St 0.29 did not differ signicantly (P 0.136)
from this (after square-root transforming the data to remove skew
arising because St is a ratio). The data are not normally distributed
even after transformation (Supplementary Fig. S2), but the t-test is
robust to large deviations from normality and, as the 95% con-
dence interval (0.25 , St , 0.31) fell comfortably inside the opti-
mal range 0.2 , St , 0.4, we can be reasonably sure that the lack of
any statistically signicant difference is not merely an artefact of low
statistical power. Median St 0.25 (inter-quartile range 0.200.35)
was less than mean St because of the positive skew, but still did not
differ signicantly from the expected optimum (sign test,
P 0.081). St therefore seems to have converged on the expected
optimum St < 0.3 in cruising ight, with about 75% of species
falling in the range 0.19 , St , 0.41. In addition, the t is remark-
ably tight given the much larger variation in St found in a single
species forced to y at different speeds.
Because the assumptions of parametric analysis of variance are
unlikely to hold for our data set, we used a KruskalWallis test to
compare St among taxonomic classes. The test just failed to attain
statistical signicance (P 0.054), mainly because the n 2
sample size for insects limited its efciency. Using a Wilcoxon
rank sum test to compare directly birds and bats showed instead
that St is signicantly lower in birds than in bats (two-tailed,
P 0.020). We then dropped all families of n 1 and used a
Figure 2 Strouhal number for 42 species of birds, bats and insects in unconned, cruising
ight. Published ranges
3,4
of St in cruising sh and dolphins are included for comparison.
Dotted lines mark the range 0.2 , St , 0.4, in which propulsive efciency usually
peaks; dashed line marks the modal peak at St 0.3. Unbroken lines indicate the range
of variation in St across other non-zero ight speeds, where such data exist.
letters to nature
NATURE | VOL 425 | 16 OCTOBER 2003 | www.nature.com/nature 709 2003 Nature PublishingGroup
concentration using a Varian CP-3800 gas chromatograph with a thermal conductivity
detector. Rates of CO2 production were calculated by dividing concentrations by the time
since jars had been sealed. On the same date, we collected samples of gas from the
headspace of incubation jars for determining the d
13
C value of the accumulated CO2. We
injected 7 ml of headspace gas into He-ushed gas-tight LabCo exetainers, which were
then placed into an autosampler and analysed on a FIN-MAT Gas Bench II connected to
the IRMS. Soil respiration d
13
C values for control and elevated O3 treatments were
225.0 ^ 0.2 across the incubation. For soils from elevated CO2, soil respiration values
initially were 230.9 ^ 0.3 and increased to 228.8 ^ 0.2 by day 281 of incubation.
For the elevated O3 CO2 treatment, values initially were 231.1 ^ 0.4 and
228.2 ^ 0.2 at day 281 of incubation.
Stable isotope tracer, calculations and statistical analysis
To investigate the contribution of new carbon inputs to soil and microbial respiration in
the elevated CO2 and elevated O3 CO2 plots, we used the d
13
C signature derived from
the fossil-fuel fumigation gas mixed with ambient air (218.7 ^ 1.0 compared to
28.6 ^ 0.1 for control and elevated O3 plots). Subsequent fractionation by the plants
produced leaf and root tissue with d
13
C values of approximately 241.6 ^ 0.4 in
elevated CO2 and elevated O3 CO2 plots. Composite soil samples collected from all 12
plots before the experiment began in 1997 and from the three control plots in 2001 had
similar (P 0.47) d
13
C values, averaging 226.7 ^ 0.2. The proportional contribution
of carbon derived from the fumigation gas ( f ) was calculated using the equation f
dt 2do=di 2do; where dt refers to the isotopic composition (d
13
C) of the soil organic
carbon or soil respiration from the fumigated (elevated CO2 or elevated O3 CO2) plots,
do is the d
13
C of soil organic carbon or soil respiration from control plots, and di is the
13
C
signature of the plant leaves and roots collected in 2001 fromthe fumigated plots, weighted
equally and averaged across species as there were no signicant differences found among
the aspen and aspenbirch plots. We do not expect that the isotopic signature of the plants
varied appreciably over the life of this experiment because fumigation was initiated when
the trees were still seedlings. Because there were differences among treatments in the
amount of carbon entering soils derived from fumigation (Fig. 1), we calculated the
relative mass loss as the cumulative amount of carbon respired from each of these pools
divided by the amount of soil carbon in that pool. Data were analysed using analysis of
variance with general linear models for a split-plot randomized complete design using SAS
8.02 (Cary, North Carolina).
Received 25 February; accepted 12 September 2003; doi:10.1038/nature02047.
1. IPCC Climate Change 2001: Technical Summary (Report of the Intergovernmental Panel on Climate
Change, IPCC Secretariat, Geneva, 2001).
2. Gregg, J. W., Jones, C. G. & Dawson, T. E. Urbanization effects on tree growth in the vicinity of New
York City. Nature 424, 183187 (2003).
3. McLaughlin, S. B. & Downing, D. J. Interactive effects of ambient ozone and climate measured on
growth of mature forest trees. Nature 374, 252254 (1995).
4. Chameides, W. L., Kasibhatla, P. S., Yienger, J. & Levy, H. I. Growth of continental-scale metro-agro-
plexes, regional ozone pollution, and world food production. Science 264, 7477 (1994).
5. Percy, K. E. et al. Altered performance of forest pests under CO2- and O3-enriched atmospheres.
Nature 420, 403407 (2002).
6. Latest Findings on National Air Quality: 2002 Status and Trends (US Environmental Protection
Agency).
7. Findlay, S., Carreiro, M., Krischik, V. & Jones, C. G. Effects of damage to living plants on leaf litter
quality. Ecol. Appl. 6, 269275 (1996).
8. Coleman, M. D., Dickson, R. E., Isebrands, J. G. & Karnosky, D. F. Carbon allocation and partitioning
in aspen clones varying in sensitivity to tropospheric ozone. Tree Physiol. 15, 593604 (1995).
9. Andersen, C. P. Source-sink balance and carbon allocation below ground in plants exposed to ozone.
New Phytol. 157, 213228 (2003).
10. King, J. S. et al. Fine-root biomass and uxes of soil carbon in young stands of paper birch and
trembling aspen as affected by elevated atmospheric CO2 and tropospheric O3. Oecologia 128,
237250 (2001).
11. Dickson, R. E. et al. Forest Atmosphere Carbon Transfer and Storage (FACTS-II)The Aspen Free-air
CO2 and O3 Enrichment (FACE) Project: An Overview (Technical Report NC-214, USDA, Washington
DC, 2000).
12. Leavitt, S. W., Follett, R. F. & Paul, E. A. Estimation of slow- and fast-cycling soil organic carbon pools
from 6N HCl hydrolysis. Radiocarbon 38, 231239 (1996).
13. Paul, E. A. et al. Radiocarbon dating for determination of soil organic matter pool sizes and dynamics.
Soil Sci. Soc. Am. J. 61, 10581067 (1997).
14. Phillips, R., Zak, D. R., Holmes, W. E. &White, D. C. Microbial community composition and function
beneath temperate trees exposed to elevated atmospheric carbon dioxide and ozone. Oecologia 131,
236244 (2002).
15. Larson, J., Zak, D. R. & Sinsabaugh, R. L. Extracellular enzyme activity beneath temperate trees
growing under elevated carbon dioxide and ozone. Soil Sci. Soc. Am. J. 66, 18481856 (2002).
Acknowledgements This research was supported by the US Department of Energys Ofce of
Science (BER: Program for Ecosystem Research and National Institute for Global Environmental
Change), the USDA Forest Service (Northern Global Change and North Central Research
Station), the National Science Foundation (DEB, DBI/MRI), and the USDA Natural Research
Initiatives Competitive Grants Program. G. Hendry, K. Lewin, J. Nagey, D. Karnosky and
J. Sober have been instrumental in the successful implementation of this long-term eld
experiment.
Competing interests statement The authors declare that they have no competing nancial
interests.
Correspondence and requests for materials should be addressed to W.M.L. (wmloya@mtu.edu).
..............................................................
Flying and swimming animals
cruise at a Strouhal number
tuned for high power efciency
Graham K. Taylor, Robert L. Nudds* & Adrian L. R. Thomas
Zoology Department, University of Oxford, Tinbergen Building, South Parks
Road, Oxford OX1 3PS, UK
* Present address: School of Biology, University of Leeds, L. C. Miall Building, Clarendon Way,
Leeds LS2 9JT, UK
.............................................................................................................................................................................
Dimensionless numbers are important in biomechanics because
their constancy can imply dynamic similarity between systems,
despite possible differences in mediumor scale
1
. Adimensionless
parameter that describes the tail or wing kinematics of swim-
ming and ying animals is the Strouhal number
1
, St 5 fA/U,
which divides stroke frequency (f ) and amplitude (A) by forward
speed (U)
28
. St is known to govern a well-dened series of vortex
growth and shedding regimes for airfoils undergoing pitching
and heaving motions
6,8
. Propulsive efciency is high over a
narrow range of St and usually peaks within the interval
0.2 < St < 0.4 (refs 38). Because natural selection is likely to
tune animals for high propulsive efciency, we expect it to
constrain the range of St that animals use. This seems to be
true for dolphins
25
, sharks
35
and bony sh
35
, which swim at
0.2 < St < 0.4. Here we show that birds, bats and insects also
converge on the same narrow range of St, but only when cruising.
Tuning cruise kinematics to optimize St therefore seems to be a
general principle of oscillatory lift-based propulsion.
Experiments with isolated pitching or heaving foils have
measured extremely high peak propulsive efciencies within the
interval 0.2 , St , 0.4 (modal peak at St < 0.3)
37
. In this range,
the propulsive efciency (dened as the ratio of aerodynamic power
output to mechanical power input) can be as high as 70% (ref. 7) or
even 80% (ref. 6). Optimal St depends subtly on kinematic par-
ameters including geometric angle of attack, amplitude-to-chord
ratio, airfoil section and phase of motion
68
but, for any given
motion, efciency is usually high (.60%) over a range narrower
than 0.2 , St , 0.4 (refs 37). For example, measured efciency
can plummet from80%at St 0.27 to 10%at St 0.09 (ref. 6) and
also drops off at higher St, albeit more gently
7,8
. Measured propul-
sive efciency usually peaks when the kinematics result in maxi-
mum amplication of the shed vortices in the wake and an average
velocity prole equivalent to a jet
3,4,6
.
Theoretical treatments of apping wings
3,4,6,8
further conrm the
empirical result that St tightly constrains propulsive efciency. In
fact, St is bound to affect aerodynamic force coefcients and
propulsive efciency, because it denes the maximum aerodynamic
angle of attack and the timescales associated with the growth and
shedding of vortices, which are the source of aerodynamic force
production
8,9
. Natural selection is expected to favour wing kin-
ematics that combine high propulsive efciency with a high aero-
dynamic force coefcient. These need not peak at identical St, but
can do for certain motions
7,8
; if not, selection should optimize the
trade-off. Propulsive efciency may be the more important selec-
tion pressure in cruising, whereas high aerodynamic force coef-
cients may be more important in accelerations, slow locomotion or
hovering. In cruising ight or swimming, we therefore predict that
St will be tuned for high propulsive efciency.
This suggestion has already been made for cruising sh and
dolphins
25
, which operate within the range 0.2 , St , 0.4, and the
principle is considered so general for swimming animals that it has
even been used to predict the speeds of extinct ichthyosaurs
10
.
Whereas the uid dynamic results described above
38
refer to
letters to nature
NATURE | VOL 425 | 16 OCTOBER 2003 | www.nature.com/nature 707 2003 Nature PublishingGroup
1637 Dolphin Strouhal numbers
With no statistically signicant correlation, A/L was found
to be relatively insensitive to both U/L (Fig.1A) and f (Fig.1B)
for all species. The mean value of A/L for all odontocetes was
0.210.03 (n=267). Mean A/L ranged from 0.250.02 for S.
frontalis to 0.170.02 for G. melaena with 89% of the data
residing between 0.15 and 0.25. ANOVA showed that there
was a signicant difference for A/L among species (P<0.001;
F=9.76; d.f.=6, 260). Aggregating all the odontocete data
(n=267), f was found to increase linearly with increasing U/L
as f=0.89(U/L)+0.59 (r
2
=0.8; Fig.2). A positive linear
relationship between f and U/L is similar to results reported for
cetaceans, sh and other marine mammals (Bainbridge, 1958;
Hunter and Zweifel, 1971; Webb and Kostecki, 1984;
Feldkamp, 1987; Fish et al., 1988; Scharold et al.,1989).
Regression equation for f and A/L with respect to U/L for each
species is provided in Table2. The negative slope in the
regression equation for f by G. melaena is due to the limited
speed range.
Strouhal data
The computed Strouhal number showed little dependence on
body length or swim speed for the delphinid species (Fig.3).
Aggregating animals for each species (Fig.4), mean St values
generally reside near the lower boundary of the 0.250.35
range (Table3) predicted by Triantafyllou et al. (1991, 1993)
for peak propulsive efficiency. Excluding D. leucas, the mean
St for the delphinids was 0.260.05 (n=248). The predicted
0.250.35 St range captured 55% of the delphinid St data
(Fig.5), whereas the range from 0.2 to 0.3 contained 74%. For
an incremental St range of 0.05, the majority of the data were
found between 0.225 and 0.275 (44%). D. leucas had a mean
St of 0.350.10 (n=19), which was conspicuously higher than
most of the St values for the delphinids.
Propulsive efficiencies, which were previously reported by
Fish (1998), are plotted as a function of St in Fig.6. For P.
crassidens, O. orca and T. truncatus, propulsive efficiencies
were found to broadly peak at about 0.90, 0.87 and 0.85,
respectively, over a relatively narrow range of St (0.230.28).
Outside this St range, where measurements exist, efficiencies
drop off rapidly. The St range favored by P.
crassidens, O. orca and T. truncatus was
within this same range, 0.225<St<0.30
(Fig.7AC). The efficiency data for D. leucas
were lower (0.83) and exhibit a conspicuously
broader peak at St=0.250.40. The distribution
of D. leucas St was relatively at, with a
narrow peak occurring at St=0.4250.45
(Fig.7D).
Discussion
Cetaceans swim by oscillatory heaving and
pitching of the caudal ukes, which act as a
hydrofoil (Lighthill, 1969; Webb, 1975; Fish
and Hui, 1991; Fish, 1993, 1998). The
oscillating movements of a hydrofoil result in
unsteady shedding of vorticity from the trailing
edge (Anderson et al., 1998). The pattern and
spin of the staggered array of vortices generate
a jet ow, which produces thrust to overcome
the drag on the body. Triantafyllou et al. (1991,
1993, 2002) considered the jet to be
convectively unstable, acting as a tuned
amplier with a narrow range of frequencies of
maximum amplication (i.e. maximum thrust
production). The pattern and periodicity of
vortices shed into the wake, therefore,
determine the optimal thrust production for
maximum efficiency. The arrangement of
vortices for maximum efficiency is a reverse

0
0.1
0.2
0.3
0.4
0.5
0.6
0 1 2 3 4 5 6 7 8 9
Swimming speed, U (m s
1
)
0
0.1
0.2
0.3
0.4
0.5
0.6
A
B S
t
r
o
u
h
a
l

n
u
m
b
e
r
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
Body length, L (m)
Fig.3 (A) Strouhal number, St, of odontocete cetaceans as a function of body length, L.
(B) Strouhal number, St, of odontocete cetaceans as a function of swimming speed, U.
Symbols indicate particular species: Tursiops truncatus (solid red circles), Pseudorca
crassidens (open green square), Orcinus orca (solid blue triangles), Globicephala
melaena (blue crosses), Lagenorhynchus obliquidens (solid blue diamonds) and
Stenella frontalis (open brown diamonds).
Table 3. Mean Strouhal values
Sample size Species Mean S.D.
n=107 Tursiops truncatus 0.260.05
n=69 Pseudorca crassidens 0.260.05
n=30 Orcinus orca 0.280.05
n=19 Delphinapterus leucas 0.350.1
n=17 Lagenorhynchus obliquidens 0.240.03
n=13 Stenella frontalis 0.330.03
n=12 Globicephala melaena 0.240.02
1638
Karmen street (Triantafyllou et al., 2002). The
wake dynamics are dominated by the non-
dimensional Strouhal number, St, in which the
distance between vortices and their rate of
formation co-vary with speed (Vogel, 1994;
Triantafyllou and Triantafyllou, 1995;
Triantafyllou et al., 2002). Experimental studies
of heaving and pitching foils have found that the
structure of the vortex wake changes with St (von
Ellenrieder et al., 2003; Taylor et al., 2003),
maximum thrust occurred between 0.25 and 0.4
(Triantafyllou et al., 1993; Anderson et al., 1998)
and maximum efficiency was within the range of
0.250.4 (Triantafyllou et al., 1991, 1993;
Anderson et al., 1998). Efficiencies for apping
foil experiments have been reported to peak
below the optimum 0.250.35 St range predicted.
The data for all delphinids (n=248) showed
little dependence on St over a range of swim
speeds from about 28ms
1
(Fig.3B). It has
been hypothesized for cruising ight and
swimming that St would be tuned for high
propulsive efficiency (Triantafyllou et al., 1991,
1993; Taylor et al., 2003). Cruising speeds for
the cetaceans have been reported from
~15ms
1
(see Fish, 1998). For the present St
data (Fig.5), a conspicuous peak was not
apparent at cruising speeds (Fig.3B) or where
maximum propulsive efficiency was predicted
(St=0.250.35). Moreover, the St data were not
most concentrated in the predicted range.
Whereas 55% of the data fell within the
predicted range of St=0.250.35, 74% of all the
St data occurred between 0.2 and 0.3. Some
of the scatter in the St data is a result of
measurement uncertainty. Wolfgang et al.
(1999) have reported St uncertainties of ~30%
for studies with sh (Danio malabaricus).
However, a large part is presumably due to
natural variation of the animals swimming
motion (Rosen, 1959; Wolfgang et al., 1999).
Kayan and Pyatetskiy (1977) reported a
dependence of St on acceleration for captive T.
truncatus, with St increasing with increasing
acceleration. Taylor et al. (2003) similarly found
that, for birds, St was signicantly higher for
intermittent as opposed to direct ight. Although
data from the present video analysis were limited
to steady swimming speeds, effects due to small
accelerations were possible. A dependence of
St on acceleration may partly explain the
difference in St values for L. obliquidens
between the present data (St=0.240.03, n=17)
and those inferred from traces of an accelerating
animal (St=0.30, n=1; Triantafyllou et al., 1991,
1993).
J. J. Rohr and F. E. Fish
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
S
t
r
o
u
h
a
l

n
u
m
b
e
r

r
a
n
g
e
D
e
l
p
h
i
n
a
p
t
e
r
u
s
l
e
u
c
a
s
O
r
c
i
n
u
s

o
r
c
a
S
t
e
n
e
l
l
a

f
r
o
n
t
a
l
i
s
L
a
g
e
n
o
r
h
y
n
c
h
u
s
o
b
l
i
q
u
i
d
e
n
s
G
l
o
b
i
c
e
p
h
a
l
a
m
e
l
a
e
n
a
P
s
e
u
d
o
r
c
a
c
r
a
s
s
i
d
e
n
s
T
u
r
s
i
o
p
s
t
r
u
n
c
a
t
u
s
Fig.4. Strouhal number (St) range ( S.D.) for cetaceans.
0
10
20
30
40
50
60
70
80
P
e
r
c
e
n
t

o
f

t
o
t
a
l
o
d
o
n
t
o
c
e
t
e

c
e
t
a
c
e
a
n

d
a
t
a
Strouhal range
0
.
1
2
5
0
.
2
2
5
0
.
1
5
1
0
.
2
5
0
0
.
1
7
6
0
.
2
7
5
0
.
2
0
1
0
.
3
0
0
0
.
2
2
6
0
.
3
2
5
0
.
2
5
1
0
.
3
5
0
0
.
2
7
6
0
.
3
7
5
0
.
3
0
1
0
.
4
0
0
0
.
3
2
6
0
.
4
2
5
0
.
3
5
1
0
.
4
5
0
0
.
3
7
6
0
.
4
7
5
0
.
4
0
1
0
.
5
0
0
0
.
4
2
6
0
.
5
2
5
Fig.5. Histogram of percentage of delphinid cetacean Strouhal number (St) data
(n=248); incremental St range is 0.1, incremental step is 0.25.
0.75
0.8
0.85
0.9
0.95
1
0.1 0.2 0.3 0.4 0.5 0.6
Strouhal number
P
r
o
p
u
l
s
i
o
n

e
f
f
i
c
i
e
n
c
y
Fig.6. Propulsion efficiency of cetaceans as a function of Strouhal number (St).
Colors indicate particular species: Tursiops truncatus (red), Pseudorca crassidens
(green), Orcinus orca (blue) and Delphinapterus leucas (black).
1633
An important aspect of swimming is the ability to move
efficiently. Paradoxically, early attempts at building sh-
inspired mechanisms achieved disappointingly low
efficiencies (Triantafyllou and Triantafyllou, 1995). It was
only through a deeper understanding of the vorticity produced
along the swimming animal and in its wake that signicant
progress was achieved. Beginning almost 40years ago, Rosen
(1959, 1961, 1963) discerned, through a series of innovative
ow visualization experiments, a system of vortices appearing
along the sides of swimming sh and dolphins. Rosen (1959,
1961, 1963) hypothesized that some of the rotational energy
surrounding the undulating motion of a sh or dolphin could
be regained for propulsion through proper synchronization of
the animals body to the vortex ow. Rosen (1959, 1961)
further deduced an equation for sh and dolphin motion. This
equation predicted swimming speed to be proportional to the
product of the tail beat amplitude and frequency. Rosen (1959,
1961) referred to this proportionality as the sh constant and
hypothesized that it was nearly the same for sh and dolphins.
A similar conclusion, but through more rigorous theoretical
analysis and detailed experimental studies, has been drawn by
Triantafyllou et al. (1991, 1993). Performing stability analysis
of the mean velocity proles of a pitching airfoil, Triantafyllou
et al. (1991, 1993) have shown that maximum spatial
amplication and optimum creation of thrust-producing jet
vortices lies in a narrow range of nondimensional frequencies
referred to as the Strouhal number (St). The predicted St range
for maximum spatial amplication occurs between 0.25 and
0.35, peaking at 0.30 (Triantafyllou et al., 1991, 1993;
Streitlien and Triantafyllou, 1998). Triantafyllou and
Triantafyllou (1995) have argued that for St=0.250.35,
swimming efficiency for sh and cetaceans would also peak.
Experiments with isolated oscillating foils have found highest
propulsive efficiencies for St between 0.20 and 0.40
The Journal of Experimental Biology 207, 1633-1642
Published by The Company of Biologists 2004
doi:10.1242/jeb.00948
Swimming efficiencies of sh and cetaceans have been
related to a certain synchrony between stroke cycle
frequency, peak-to-peak tail/uke amplitude and mean
swimming speed. These kinematic parameters form a non-
dimensional wake parameter, referred to as a Strouhal
number, which for the range between 0.20 and 0.40 has
been associated with enhanced swimming efficiency for
sh and cetaceans. Yet to date there has been no direct
experimental substantiation of what Strouhal numbers are
preferred by swimming cetaceans. To address this lack of
data, a total of 248 Strouhal numbers were calculated for
the captive odontocete cetaceans Tursiops truncatus,
Pseudorca crassidens, Orcinus orca, Globicephala melaena,
Lagenorhynchus obliquidens and Stenella frontalis.
Although the average Strouhal number calculated for
each species is within the accepted range, considerable
scatter is found in the data both within species and among
individuals. A greater proportion of Strouhal values occur
between 0.20 and 0.30 (74%) than the 0.250.35 (55%)
range predicted for maximum swimming efficiency.
Within 0.05 Strouhal increments, the greatest number of
Strouhal values was found between 0.225 and 0.275 (44%).
Where propulsive efficiency data were available (Tursiops
truncatus, Pseudorca crassidens, Orcinus orca), peak
swimming efficiency corresponded to this same Strouhal
range. The odontocete cetacean data show that, besides
being generally limited to a range of Strouhal numbers
between 0.20 and 0.40, the kinematic parameters
comprising the Strouhal number provide additional
constraints. Fluke-beat frequency normalized by the ratio
of swimming speed to body length was generally restricted
from 1 to 2, whereas peak-to-peak uke amplitude
normalized by body length occurred predominantly
between 0.15 and 0.25. The results indicate that the
kinematics of the propulsive ukes of odontocete cetaceans
are not solely dependent on Strouhal number, and the
Strouhal number range for odontocete cetaceans occurs at
slightly (~20%) lower values than previously predicted for
maximum swimming efficiency.
Key words: Strouhal number, swimming, dolphin, cetacean,
odontocete, Tursiops, Pseudorca, Orcinus, Lagenorhynchus,
Globicephala, Stenella, Delphinapterus.
Summary
Introduction
Strouhal numbers and optimization of swimming by odontocete cetaceans
Jim J. Rohr
1,
* and Frank E. Fish
2
1
SSC San Diego, 53560 Hull Street, 211, San Diego, CA 92152, USA and
2
Department of Biology,
West Chester University, West Chester, PA 19383, USA
*Author for correspondence (e-mail: rohr@spawar.navy.mil)
Accepted 17 February 2004
anterior body.
developments.
[11], [16], and [38].
BCF
Ondulatrio Fish School
BCF Ondulatrio
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
(a)
(b)
Lindsey [10].)
(a) (b) (c) (d)
part of the sh.
Fig. 11. (a) Lateral view of caudal n shape for thunniform swimmers, showing span , chord , pitching axis position , sweepback angle and surface area
. (b) Trail of an oscillating caudal n showing amplitude , wavelength , feather angle , and attack angle of the n. (Adapted from Magnuson [11].)
area. As the n moves along this trail, its forward velocity
is the same as that of the sh, while its lateral velocity
changes in time. Other important parameters of its motion
include the angle of attack (with respect to its trail) and
the feathering angle [Fig. 11(b)]. Feathering is the angle
between the n trail and the overall path of the sh. Both
and change as the caudal n sweeps laterally in order to
obtain maximal thrust during the whole of the n-beat cycle.
Detailed data from several references on all these variables for
the scombrids sh family have been gathered in [11], where
tail-beat frequencies as high as 14.5 Hz are documented for
a 40-cm-long Kawakawa (swimming at a speed of 8.2 BL/s,
). Thrust is obtained by the lift force acting
on the oscillating n surface and by leading-edge suction, i.e.,
the action of the reduced pressure in the water moving around
the rounded leading edge of the caudal n. The developed
thrust and the propulsive efciency generally depend on the
following parameters:
1) the aspect ratio (AR) of the caudal n. This is dened
as the n span squared, divided by the projected n
area [Fig. 11(a)]
High aspect ratio ns lead to improved efciency,
because they induce less drag per unit of lift or thrust
produced. In thunniform swimmers, AR values range
from 4.5 to about 7.2.
2) the shape of the caudal n, as it is dened by the
sweepback angle and the curvature of its leading
edge [Fig. 11(a)]. A curved leading edge is benecial,
because it reduces the relative contribution of leading-
edge suction to the total thrust, avoiding boundary layer
separation for high thrust values [11].
3) the n stiffness. The benet of a higher degree of stiff-
ness (achieved by fusing the many n-rays consisting
the caudal n) is increased thrust generation capability,
with only a relatively small drop in efciency [11].
4) the oscillatory motions of the n.
To study the effects of 4) on thrust production, three factors
have traditionally been considered in the models developed
for lunate tail propulsion: the reduced frequency and the
proportional feathering parameters, along with the position of
the pitching axis [dened by in Fig. 11(a)]. For a thunniform
swimmer, the reduced frequency represents the ratio of the
time to swim a distance equal to the caudal n chord (usually
calculated as ) to the tailbeat period
The proportional feathering parameter , originally proposed
by Lighthill in [24], is dened as the ratio of slopes between
and and can be computed [12] as
where is the angle of attack in radians (the slope of ),
the maximum lateral velocity of the n, and is the
swimming speed. Values of between 0.6 and 0.8 have been
calculated by Lighthill [60] to yield optimal combinations of
leading-edge suction and hydromechanical efciency.
Lighthill [42] was the rst to apply a simple linear 2-D
(i.e., for ) wing theory on lunate tail propulsion.
The uid is assumed inviscid and irrotational, and potential
theory is used to calculate the thrust for small-amplitude
oscillations. In his optimization analysis, Wu [61] calcu-
lated that efciencies close to unity are attainable in such
a 2-D model. A large-amplitude 2-D theory based on the
impulse approach was developed by Chopra [62]. Extension
to three dimensions (conned to rectangular wings and small-
amplitude oscillations) based on the vorticity distribution
was made by Chopra in [63]. Chopra and Kambe used a
3-D unsteady lifting-surface theory in [64] to study thrust
production from a variety of different wing shapes. Lan
also considered a 3-D problem in [65], where an unsteady
quasi-vortex lattice method is used. All these models assume
rigid tails. The effects of passive chordwise exibility of the
caudal n performing large-amplitude motions for the 2-D
case were studied by Katz and Weihs in [66]. A linearized
low-frequency unsteady lifting-line theory was applied by
Ahmadi and Widnall in [67]. A strip theory considering small-
amplitude pitching motions was developed by Bose and Lien
in [68] to calculate the hydrodynamic performance of a n
Fig. 11. (a) Lateral view of caudal n shape for thunniform swimmers, showing span , chord , pitching axis position , sweepback angle and surface area
. (b) Trail of an oscillating caudal n showing amplitude , wavelength , feather angle , and attack angle of the n. (Adapted from Magnuson [11].)
area. As the n moves along this trail, its forward velocity
is the same as that of the sh, while its lateral velocity
changes in time. Other important parameters of its motion
include the angle of attack (with respect to its trail) and
the feathering angle [Fig. 11(b)]. Feathering is the angle
between the n trail and the overall path of the sh. Both
and change as the caudal n sweeps laterally in order to
obtain maximal thrust during the whole of the n-beat cycle.
Detailed data from several references on all these variables for
the scombrids sh family have been gathered in [11], where
tail-beat frequencies as high as 14.5 Hz are documented for
a 40-cm-long Kawakawa (swimming at a speed of 8.2 BL/s,
). Thrust is obtained by the lift force acting
on the oscillating n surface and by leading-edge suction, i.e.,
the action of the reduced pressure in the water moving around
the rounded leading edge of the caudal n. The developed
thrust and the propulsive efciency generally depend on the
following parameters:
1) the aspect ratio (AR) of the caudal n. This is dened
as the n span squared, divided by the projected n
area [Fig. 11(a)]
High aspect ratio ns lead to improved efciency,
because they induce less drag per unit of lift or thrust
produced. In thunniform swimmers, AR values range
from 4.5 to about 7.2.
2) the shape of the caudal n, as it is dened by the
sweepback angle and the curvature of its leading
edge [Fig. 11(a)]. A curved leading edge is benecial,
because it reduces the relative contribution of leading-
edge suction to the total thrust, avoiding boundary layer
separation for high thrust values [11].
3) the n stiffness. The benet of a higher degree of stiff-
ness (achieved by fusing the many n-rays consisting
the caudal n) is increased thrust generation capability,
with only a relatively small drop in efciency [11].
4) the oscillatory motions of the n.
To study the effects of 4) on thrust production, three factors
have traditionally been considered in the models developed
for lunate tail propulsion: the reduced frequency and the
proportional feathering parameters, along with the position of
the pitching axis [dened by in Fig. 11(a)]. For a thunniform
swimmer, the reduced frequency represents the ratio of the
time to swim a distance equal to the caudal n chord (usually
calculated as ) to the tailbeat period
The proportional feathering parameter , originally proposed
by Lighthill in [24], is dened as the ratio of slopes between
and and can be computed [12] as
where is the angle of attack in radians (the slope of ),
the maximum lateral velocity of the n, and is the
swimming speed. Values of between 0.6 and 0.8 have been
calculated by Lighthill [60] to yield optimal combinations of
leading-edge suction and hydromechanical efciency.
Lighthill [42] was the rst to apply a simple linear 2-D
(i.e., for ) wing theory on lunate tail propulsion.
The uid is assumed inviscid and irrotational, and potential
theory is used to calculate the thrust for small-amplitude
oscillations. In his optimization analysis, Wu [61] calcu-
lated that efciencies close to unity are attainable in such
a 2-D model. A large-amplitude 2-D theory based on the
impulse approach was developed by Chopra [62]. Extension
to three dimensions (conned to rectangular wings and small-
amplitude oscillations) based on the vorticity distribution
was made by Chopra in [63]. Chopra and Kambe used a
3-D unsteady lifting-surface theory in [64] to study thrust
production from a variety of different wing shapes. Lan
also considered a 3-D problem in [65], where an unsteady
quasi-vortex lattice method is used. All these models assume
rigid tails. The effects of passive chordwise exibility of the
caudal n performing large-amplitude motions for the 2-D
case were studied by Katz and Weihs in [66]. A linearized
low-frequency unsteady lifting-line theory was applied by
Ahmadi and Widnall in [67]. A strip theory considering small-
amplitude pitching motions was developed by Bose and Lien
in [68] to calculate the hydrodynamic performance of a n
St =
fA
U
0, 25 < St < 40
15
o
<
max
< 25
o
BCF Ondulatrio
90
achieve because most particle image velocimetry (PIV)
measurements provide flow fields in 2-D planes (Tytell, 2007).
Therefore, the efficiency in the recent experimental studies has
been obtained using hydromechanical models such as LighthillFs
elongated body theory (EBT) (Lighthill, 1969). Based on EBT
the thrust is produced only at the tail and the rest of the body
undulations create wasted energy (Lighthill, 1969). Therefore,
according to this theory carangiform swimmers are more efficient
than anguilliform swimmers because most of their body
undulations are restricted in the caudal fin region. This theory,
however, incorporates many simplifying assumptions and neglects
viscous forces that can produce thrust along the body of
anguilliform swimmers (Shen et al., 2003; Taneda and Tomonari,
1974). As discussed in Tytell and Lauder, conclusions based on
EBT could lead to results that are contrary to what is observed
in nature (Tytell and Lauder, 2004). For example, many eels
migrate thousands of miles (van Ginneken and van den Thillart,
2000) and many sharks swim steadily in the sub-carangiform
(between carangiform and anguilliform) mode (Gemballa et al.,
2006).
In our previous work (Borazjani and Sotiropoulos, 2009a) we
helped to reconcile a number of such inconsistencies obtained
from simpler theories by performing 3-D numerical simulations
of tethered carangiform (Borazjani and Sotiropoulos, 2008) and
anguilliform (Borazjani and Sotiropoulos, 2009a) virtual
swimmers under similar conditions (Reynolds number). We
found that: (1) the carangiform swimmerFs Froude efficiency (h
f
)
is maximized as the Re* approaches while the anguilliform
swimmerFs efficiency is maximized somewhere in the transitional
regime; (2) the temporal evolution of the net force acting on the
anguilliform swimmer varies more smoothly than for the
carangiform swimmer case; (3) the swimming power required for
a self-propelled anguilliform swimmer is lower than that for the
carangiform swimmer at the same Re*; and (4) for both swimmers
the swimming power is higher than the power required to tow
the respective rigid body at the same Re* (Borazjani and
Sotiropoulos, 2009a).
Our previous work (Borazjani and Sotiropoulos, 2009a) pointed
to a number of similarities as well as to a number of differences
between the two modes of swimming. The similarities are due to
similar BCF undulatory propulsion and similar governing flow
parameters such as Re* and St*. The differences, however, could
be due to either the difference in the respective body shapes (form)
and/or the type of body undulations (kinematics). The purpose of
the present study is to systematically investigate and quantify the
effects of form and kinematics on the hydrodynamic performance
of undulatory swimming. The main difficulty for accomplishing such
an undertaking is finding a rational approach for isolating the effects
of form and kinematics for different swimmers. To isolate such
effects, in this work we construct virtual, self-propelled swimmers
of a given body shape (fixed form) and make them swim with
different kinematics. For example, to quantify the effects of
kinematics on a mackerel body we can compare the performance
of a mackerel swimming like a mackerel (with carangiform
kinematics) with that of a mackerel swimming like an eel
(anguilliform kinematics). Similarly, to isolate the effects of form
we compare the performance of a mackerel body and an eel body
both swimming with the same kinematics (carangiform or
anguilliform). More specifically, we employ the following virtual
swimmers (Fig.1): (1) a mackerel body swimming like a mackerel
(denoted as MM); (2) mackerel body swimming like a lamprey
(ML); (3) a lamprey body swimming like a lamprey (LL); and (4)
a lamprey body swimming like a mackerel (LM). It is of course not
possible in nature to make a live mackerel swim like an eel or an
eel swim like a mackerel. Thus, using virtual (Borazjani and
Sotiropoulos, 2008; Borazjani and Sotiropoulos, 2009a; Kern and
Koumoutsakos, 2006; Liu and Kawachi, 1999; Liu and Wassersug,
1997) or biorobotic (Barrett et al., 1999; Hultmark et al., 2007;
Lauder et al., 2007) swimmers are the only feasible alternatives for
such an undertaking.
To compare the performance of the four virtual swimmers,
we perform self-propelled, fluidbstructure interaction (FSI)
simulations under the same conditions. All virtual swimmers are
placed in the same, initially stagnant, fluid with viscosity n and
start undulating their bodies with the desired kinematics with a
tail-beat frequency f. In the present self-propelled simulations,
there is no tether to absorb the excess force F imparted by the
flow [as was the case in Borazjani and Sotiropoulos (Borazjani
and Sotiropoulos, 2008; Borazjani and Sotiropoulos, 2009a)] and
the fish can accelerate or decelerate depending on the sign of F.
Body undulations impart a thrust-type force F on the body, which
initially accelerates (on average) the virtual swimmer to higher
velocities. The swimming speed continues to increase until the
mean force F during one tail-beat cycle becomes zero (constant
speed, inline swimming limit). Thereafter, the virtual swimmer
has reached a quasi-stationary state characterized by constant
values of the mean swimming speed U, Re* and St*. The
swimming performance of the various swimmers is quantified at
this quasi-stationary, constant-mean-speed state in terms of
various performance metrics, such as power loss, Froude
efficiency h
f
, velocity over power (similar to mile per gallon),
etc. The wake structure of various swimmers is also examined at
the quasi-stationary, constant-mean-speed limit. A small portion
of the results presented in this paper has recently appeared in
Borazjani and Sotiropoulos (Borazjani and Sotiropoulos, 2009b).
The rest of the paper is organized as follows. First, we briefly
describe the numerical method and present the details of the fish
model and prescribed kinematics. Second, we provide the results
of our virtual swimmers and discuss the effects of body shape and
kinematics on their performance. Finally we summarize our findings,
present the conclusions of this work and outline the areas for future
research.
Body shape
K
i
n
e
m
a
t
i
c
s
MM LM
ML LL
Fig.1. Four different virtual swimmers. Each row has the same kinematics
whereas each column has the same body. (A)mackerel swimming like a
mackerel (MM); (B) lamprey swimming like a mackerel (LM); (C) mackerel
swimming like a lamprey (ML); (D) lamprey swimming like lamprey (LL).
BCF Ondulatrio
90
2006).
found that: (1) the carangiform swimmerFs Froude efficiency (hf)
swimmers (Fig. 1): (1) a mackerel body swimming like a mackerel
efficiency hf, velocity over power (similar to mile per gallon),
research.
Body shape
K
i
n
e
m
a
t
i
c
s
MM LM
ML LL
Fig. 1. Four different virtual swimmers. Each row has the same kinematics
whereas each column has the same body. (A) mackerel swimming like a
97 Hydrodynamics of undulatory swimming
the same time during the cycle but the magnitude of the power is
clearly not as small as for the MM swimmer (Fig.6C). This can be
explained by examining the body shape and kinematics of the MM
swimmer. The large surface area and large amplitude of motion in
the posterior region relative to the other swimmers allows the MM
swimmer to use the pressure pockets in the flow most effectively.
Another important observation that emerges collectively from
Figs5 and 6 is that for all three hydrodynamic environments,
swimmers with the mackerel body (MM, ML) not only consistently
reach higher velocities than swimmers with the lamprey body (LM,
LL) but use more power as well. Reaching higher velocities does
not necessarily indicate higher efficiency if more power is consumed,
and more quantitative analysis is required to further probe this
important point. To accomplish this, we report in Table1 various
mean quantities and efficiencies in the quasi-steady-state. In addition
to the computed mean value of U* and the resulting Re* and St*,
we also provide results for: (1) the power coefficient and mean power
equation (Eqn24); (2) the thrust coefficient and mean thrust equation
(Eqn23); (3) the Froude efficiency h
f
equation (Eqn25); (4) mean
efficiency equation (Eqn26); and (5) the root mean squared (r.m.s.)
of the swimming speed fluctuations, which are computed once the
quasi-steady, constant-mean-speed state has been reached. It can be
observed from the table that in the viscous case R1, the ML swimmer
is most efficient, has the smallest power coefficient, produces the
most amount of non-dimensional thrust and reaches the highest
speed. However, the LM swimmer uses the smallest power C
P
o
. In
the moderately viscous case R2, the ML swimmer produces the
highest non-dimensional thrust and reaches the highest swimming
speed but the LL swimmer is most efficient and uses less power.
Finally, in the inviscid case R3, the MM swimmer produces the
highest non-dimensional thrust and reaches the highest speed and
is most efficient but the LL swimmer uses the least power. In all
cases the MM swimmer always exhibits velocity fluctuations with
the highest r.m.s. From this table the effects of body shape and
kinematics cannot be readily deduced. We will discuss these issues
in more detail in the Discussion section below.
Wake structure
Figs8U10 show the wake structures visualized by the q-criterion of
different swimmers for the cases R1, R2 and R3, respectively. It
can be observed from the Figs8 and 9 that a double row structure
is present in the viscous and transitional cases for all the swimmers.
25 26 27 28 29 30 31 32
1
0
1
2
3
MM
ML
LM
LL
16 17 18 19 20 21 22 23
13 14 15 16 17 18 19 20
0
0.5
1
1.5
2
2.5
3
t/T
A
0
0.5
1
1.5
2
2.5
3
B
C
C
o
1
0
1
P
C
o
1
0
2
P
C
o
1
0
3
P
Fig.6. Non-dimensional power time history of all virtual swimmers in case
(A) R1 (Re
o
300, St
o
1.1), (B) R2 (Re
o
4000, St
o
0.6) and (C) R3 [Re
o
(inviscid), St
o
0.3] for several simulated swimming cycles. See Fig.1 for
the definition of various virtual swimmers.
Fig.7. Pressure contours in the mid-plane of the MM swimmer in case R3
[Re
o
(inviscid), St
o
0.3] at time (A) t/T32.67 (C
P
o
3.5610
4
) and (B)
t/T32.75 (C
P
o
3.7310
4
). The arrows indicate the tail motion. C
P
o
is non-
dimensional power.
BCF Ondulatrio
90
2006).
research.
Body shape
K
i
n
e
m
a
t
i
c
s
MM LM
ML LL
98
However, a single row wake structure emerges in the inertial case
R3 for all swimmers (Fig.10).
Apart from large-scale characteristics of the wake (single vs
double row), the body shape can affect more subtle wake features,
such as the shape of the vortex rings. For example, in Fig.9 the
rings shed by the lamprey body (LL and LM swimmers) are more
circular and simpler in shape whereas the rings shed by the mackerel
body (MM and ML swimmers) are more complex and less
organized. Moreover, in Fig.10, where all swimmers have a single
row wake, the swimmers with the mackerel body show a remarkable
vortex-within-a-vortex structure in the vicinity of the tail, which is
not observed in the swimmer with lamprey body. Fig.11 examines
the vortex-within-a-vortex structure more closely for the MM and
ML swimmers, which shows two vortex rings: vortex ring 1 is
generated by the leading edge of the tail; while vortex ring 2 is
generated by the trailing edge of the tail. The swimmers with the
lamprey body do not show such structure probably due to the lack
of a homocercal large aspect-ratio tail. This important wake feature
of the mackerel body swimmers will be discussed further in a
subsequent section. Comparing the wake structure of the swimmers
with similar body shapes but different kinematics in Figs8P10, we
find that the kinematics appear to have a weaker effect on the small
features of the wake than the body shape. In fact, for all cases the
wake of swimmers with the same body shape but different
kinematics looks fairly similar.
DISCUSSION
Self-propelled vs tethered simulations and experiments
As mentioned earlier to achieve swimming speeds close to one,
which are desirable from a numerical standpoint, the tail-beat
frequency of each virtual swimmer at Re
o
is selected close to the
St
o
* (Stouhal number at which net mean force on the tether is zero
for Reynolds number Re
o
) found from tethered simulations of the
MM swimmer (Borazjani and Sotiropoulos, 2008). Inversely, we
can pose the question whether the swimming speed that results from
the self-propelled, FSI simulations yields values for St* and Re*
that are comparable with those obtained from the tethered model.
The Re* and St* values reported in Table1, i.e. St*1.10, 0.61 and
Fig.8. Wake structure visualized by the iso-surfaces of q-
criterion for case R1 (Re
o
300, St
o
1.1) for self-propelled
virtual swimmers: (A) MM; (B) LM; (C) ML; (D) LL. See
Fig.1 for the definition of various virtual swimmers.
Fig.9. Wake structure visualized by the iso-
surfaces of q-criterion for case R2 (Re
o
4000,
St
o
0.6) for self-propelled virtual swimmers: (A)
MM; (B) LM; (C) ML; (D) LL. See Fig.1 for the
definition of various virtual swimmers.
BCF Ondulatrio
90
2006).
research.
Body shape
K
i
n
e
m
a
t
i
c
s
MM LM
ML LL
98
However, a single row wake structure emerges in the inertial case
R3 for all swimmers (Fig.10).
Apart from large-scale characteristics of the wake (single vs
double row), the body shape can affect more subtle wake features,
such as the shape of the vortex rings. For example, in Fig.9 the
rings shed by the lamprey body (LL and LM swimmers) are more
circular and simpler in shape whereas the rings shed by the mackerel
body (MM and ML swimmers) are more complex and less
organized. Moreover, in Fig.10, where all swimmers have a single
row wake, the swimmers with the mackerel body show a remarkable
vortex-within-a-vortex structure in the vicinity of the tail, which is
not observed in the swimmer with lamprey body. Fig.11 examines
the vortex-within-a-vortex structure more closely for the MM and
ML swimmers, which shows two vortex rings: vortex ring 1 is
generated by the leading edge of the tail; while vortex ring 2 is
generated by the trailing edge of the tail. The swimmers with the
lamprey body do not show such structure probably due to the lack
of a homocercal large aspect-ratio tail. This important wake feature
of the mackerel body swimmers will be discussed further in a
subsequent section. Comparing the wake structure of the swimmers
with similar body shapes but different kinematics in Figs8P10, we
find that the kinematics appear to have a weaker effect on the small
features of the wake than the body shape. In fact, for all cases the
wake of swimmers with the same body shape but different
kinematics looks fairly similar.
DISCUSSION
Self-propelled vs tethered simulations and experiments
As mentioned earlier to achieve swimming speeds close to one,
which are desirable from a numerical standpoint, the tail-beat
frequency of each virtual swimmer at Re
o
is selected close to the
St
o
* (Stouhal number at which net mean force on the tether is zero
for Reynolds number Re
o
) found from tethered simulations of the
MM swimmer (Borazjani and Sotiropoulos, 2008). Inversely, we
can pose the question whether the swimming speed that results from
the self-propelled, FSI simulations yields values for St* and Re*
that are comparable with those obtained from the tethered model.
The Re* and St* values reported in Table1, i.e. St*1.10, 0.61 and
Fig.8. Wake structure visualized by the iso-surfaces of q-
criterion for case R1 (Re
o
300, St
o
1.1) for self-propelled
virtual swimmers: (A) MM; (B) LM; (C) ML; (D) LL. See
Fig.1 for the definition of various virtual swimmers.
surfaces of q-criterion for case R2 (Re
o
4000,
St
o
0.6) for self-propelled virtual swimmers: (A)
MM; (B) LM; (C) ML; (D) LL. See Fig.1 for the
BCF Ondulatrio
90
2006).
research.
Body shape
K
i
n
e
m
a
t
i
c
s
MM LM
ML LL
0.25 for Re*299, 3910 and , respectively, are in excellent
agreement with the values found in the tethered MM swimmer
simulations reported in our previous publication (Borazjani and
Sotiropoulos, 2008), i.e. St
o
*1.08, 0.6 and 0.26 for Re
o
300, 4000
and , respectively. Similarly, the values obtained from the LL
swimmer in Table1, i.e. St*1.39, 0.67 and 0.47 for Re*238,
3598.8 and , respectively, are also in agreement with the values
from the tethered LL swimmer with a
max
0.1L (Borazjani and
Sotiropoulos, 2009a), i.e. St
o
*1.3, 0.63 and 0.46 for Re
o
300, 4000
and , respectively. It can be observed that for the LL swimmer,
the difference between St* of self-propelled and tethered simulations
is somewhat higher relative to the MM swimmer because there is
a larger difference between the Re* of self-propelled and tethered
simulations for the LL swimmer. Note that this is because of the
fact that, as already discussed above, the St
o
and tail-beat frequency
were chosen based on the St
o
* of the tethered MM swimmer.
In spite of the differences in kinematics and shape, most fish in
nature swim in a range of Strouhal numbers 0.25S0.35 (Triantafyllou
et al., 1993; Triantafyllou and Triantafyllou, 1995). Our virtual
swimmers with their differences in shape and kinematics get closer
to this range as Re increases (inviscid simulations). Some fish, e.g.
pacific salmon, however, have been observed to swim at high
Strouhal number at low swimming velocities (Lauder and Tytell,
2006). Our results underscore the conclusion of Borazjani and
Sotiropoulos (Borazjani and Sotiropoulos, 2008) that, for a given
body shape and kinematics, for each Re* there is unique St* at which
self-propelled swimming is possible, i.e. the pacific salmon has to
swim at high St* because this is the only St* that can make it swim
steadily at low Re* (swimming speed). According to all of the values
given in Table1, St* is a decreasing function of Re* for all
swimmers, which is also in agreement with our pervious findings
from the tethered simulations (Borazjani and Sotiropoulos, 2008;
Borazjani and Sotiropoulos, 2009a). Furthermore, comparing the
self-propelled and tethered MM and LL swimmers Re* and St*, we
observe that the self-propelled swimmers have lower Re* but higher
St* than the tethered ones. For example, the self-propelled LL
swimmer in the R2 case has Re*3599, which is lower than the
corresponding value for the tethered LL swimmer Re
o
4000 but
has St*0.67, which is higher than the tethered LL swimmer
St
o
*0.63. These results are also in agreement with our finding that
St* is a decreasing function of Re*.
The Froude efficiency and power coefficient values reported in
Table1 for MM and LL are also consistent with the values previously
found for the tethered swimmers (Borazjani and Sotiropoulos, 2008;
Borazjani and Sotiropoulos, 2009a) and show the same trend. The
efficiency of the MM swimmer increases as Re* increases but the
surfaces of q-criterion for case R3 [Re
o
(inviscid), St
o
0.3] for self-propelled virtual
swimmers: (A) MM; (B) LM; (C) ML; (D) LL. See
Fig.1 for the definition of various virtual
swimmers.
Fig.11. Wake structure visualized near the tail by the iso-surfaces of q-
criterion for case R3 [Re
o
(inviscid), St
o
swimmers (A) MM, (B) ML shows the vortex-within-a-vortex structure. See
Fig.1 for the definition of various virtual swimmers. See Fig.1 for the
BCF Ondulatrio
90
2006).
research.
Body shape
K
i
n
e
m
a
t
i
c
s
MM LM
ML LL
0.25 for Re*299, 3910 and , respectively, are in excellent
agreement with the values found in the tethered MM swimmer
simulations reported in our previous publication (Borazjani and
Sotiropoulos, 2008), i.e. St
o
*1.08, 0.6 and 0.26 for Re
o
300, 4000
and , respectively. Similarly, the values obtained from the LL
swimmer in Table1, i.e. St*1.39, 0.67 and 0.47 for Re*238,
3598.8 and , respectively, are also in agreement with the values
from the tethered LL swimmer with a
max
0.1L (Borazjani and
Sotiropoulos, 2009a), i.e. St
o
*1.3, 0.63 and 0.46 for Re
o
300, 4000
and , respectively. It can be observed that for the LL swimmer,
the difference between St* of self-propelled and tethered simulations
is somewhat higher relative to the MM swimmer because there is
a larger difference between the Re* of self-propelled and tethered
simulations for the LL swimmer. Note that this is because of the
fact that, as already discussed above, the St
o
and tail-beat frequency
were chosen based on the St
o
* of the tethered MM swimmer.
In spite of the differences in kinematics and shape, most fish in
nature swim in a range of Strouhal numbers 0.25S0.35 (Triantafyllou
et al., 1993; Triantafyllou and Triantafyllou, 1995). Our virtual
swimmers with their differences in shape and kinematics get closer
to this range as Re increases (inviscid simulations). Some fish, e.g.
pacific salmon, however, have been observed to swim at high
Strouhal number at low swimming velocities (Lauder and Tytell,
2006). Our results underscore the conclusion of Borazjani and
Sotiropoulos (Borazjani and Sotiropoulos, 2008) that, for a given
body shape and kinematics, for each Re* there is unique St* at which
self-propelled swimming is possible, i.e. the pacific salmon has to
swim at high St* because this is the only St* that can make it swim
steadily at low Re* (swimming speed). According to all of the values
given in Table1, St* is a decreasing function of Re* for all
swimmers, which is also in agreement with our pervious findings
from the tethered simulations (Borazjani and Sotiropoulos, 2008;
Borazjani and Sotiropoulos, 2009a). Furthermore, comparing the
self-propelled and tethered MM and LL swimmers Re* and St*, we
observe that the self-propelled swimmers have lower Re* but higher
St* than the tethered ones. For example, the self-propelled LL
swimmer in the R2 case has Re*3599, which is lower than the
corresponding value for the tethered LL swimmer Re
o
4000 but
has St*0.67, which is higher than the tethered LL swimmer
St
o
*0.63. These results are also in agreement with our finding that
St* is a decreasing function of Re*.
The Froude efficiency and power coefficient values reported in
Table1 for MM and LL are also consistent with the values previously
found for the tethered swimmers (Borazjani and Sotiropoulos, 2008;
Borazjani and Sotiropoulos, 2009a) and show the same trend. The
efficiency of the MM swimmer increases as Re* increases but the
surfaces of q-criterion for case R3 [Re
o
(inviscid), St
o
swimmers: (A) MM; (B) LM; (C) ML; (D) LL. See
Fig.1 for the definition of various virtual
swimmers.
Fig.11. Wake structure visualized near the tail by the iso-surfaces of q-
criterion for case R3 [Re
o
(inviscid), St
o
swimmers (A) MM, (B) ML shows the vortex-within-a-vortex structure. See
Fig.1 for the definition of various virtual swimmers. See Fig.1 for the
BCF Ondulatrio
Fish move by transferring momentum to the surrounding
uid; thus, patterns of ow in the vicinity of freely swimming
sh, including the boundary layer and wake, are of interest to
those studying aquatic locomotory mechanisms. Technological
advances over the last decade now allow investigators to
quantify ow characteristics in the vicinity of freely swimming
sh using digital particle image velocimetry (DPIV). Flow is
visualized by the addition of reective, inert, near-neutrally
buoyant particle markers to the uid, and ow-eld
characteristics including velocity and vorticity are calculated
from successive images of ow captured at high frequency
(Raffel et al., 1998). DPIV has become the method of choice
to analyze complex ow elds generated by aquatic animals.
Wakes of larval (Mller et al., 2000) and adult (Stamhuis and
Videler, 1995; Mller et al., 1997, 2000; Drucker and Lauder,
1999, 2000, 2001a; Wilga and Lauder, 1999, 2000; Hanke et
al., 2000; Lauder, 2000) shes, ow elds along the body of
swimming sh (Anderson et al., 2000; Nauen and Lauder,
2001a) and feeding and ventilatory currents of copepods and
shrimp (Stamhuis and Videler, 1995) have been quantied
using DPIV.
In the present study, we use DPIV to describe the wake
morphology and force balance for chub mackerel Scomber
japonicus (Teleostei: Scombridae). S. japonicus is a basal
1709 The Journal of Experimental Biology 205, 17091724 (2002)
Printed in Great Britain The Company of Biologists Limited
JEB4064
As members of the derived teleost sh clade
Scombridae, mackerel exhibit high-performance aquatic
locomotion via oscillation of the homocercal forked caudal
n. We present the rst quantitative ow visualization of
the wake of a scombrid sh, chub mackerel Scomber
japonicus (2026cm fork length, FL), swimming steadily
in a recirculating ow tank at cruising speeds of 1.2
and 2.2FLs
1
. Thrust was calculated from wake
measurements made separately in the horizontal (frontal)
plane and vertical (parasagittal) planes using digital
particle image velocimetry (DPIV) and compared with
drag measurements obtained by towing the same
specimens of S. japonicus post mortem.
Patterns of ow indicated that the wake consisted of a
series of linked elliptical vortex rings, each with central jet
ow. The length of the minor axis (height) of the vortex
rings was approximately equal to caudal n span; the
length of the major ring axis was dependent on swimming
speed and was up to twice the magnitude of ring height.
Proles of wake velocity components were similar to
theoretical proles of vortex rings.
Lift, thrust and lateral forces were calculated from
DPIV measurements. At 1.2FLs
1
, lift forces measured
relative to the X axis were low in magnitude (11mN,
mean S.D., N=20) but oriented at a mean angle of 6 to
the body axis. Reaction forces tend to rotate the sh about
its center of mass, tipping the head down. Thus, the
homocercal caudal n of S. japonicus functions
asymmetrically in the vertical plane. Pitching moments
may be balanced anteriorly via lift generation by the
pectoral ns. Thrust estimates for the two smallest sh
based on DPIV analysis were not signicantly different
from drag measurements made by towing those same
animals. At a speed of 1.2FLs
1
, thrust magnitude was
116mN (mean S.D, N=40). Lateral force magnitudes
were approximately double thrust magnitudes (226mN,
mean S.D., N=20), resulting in a mean mechanical
performance ratio (thrust/total force) of 0.32 at 1.2 FLs
1
.
An increase in speed by a factor of 1.8 resulted in a mean
increase in thrust by a factor of 4.4, a mean increase in
lateral forces by a factor of 3, no change in the magnitude
of lift produced and an increase in mean mechanical
performance to 0.42. The relatively high lateral forces
generated during swimming may be a necessary
consequence of force production via propagated waves of
bending.
Movies available on-line.
Key words: hydrodynamics, locomotion, swimming, scombrid, sh,
force balance, mackerel, Scomber japonicus.
Summary
Introduction
Hydrodynamics of caudal n locomotion by chub mackerel, Scomber japonicus
(Scombridae)
Jennifer C. Nauen* and George V. Lauder
Department of Organismic and Evolutionary Biology, Harvard University, 26 Oxford Street, Cambridge, MA 02138,
USA
*e-mail: jnauen@oeb.harvard.edu
Accepted 13 March 2002
1711 Hydrodynamics of mackerel caudal n locomotion
these images were used to determine the shs position relative
to the light sheet and the angle of the body to the X axis
(dened in Fig. 1). A second camera viewed the movement of
particles illuminated by the light sheet. The cameras elds of
view were calibrated by recording images of a ruler at the
beginning of each experiment. Note that as the caudal n
moved across the eld of view it blocked portions of the light
sheet (which resulted in dark areas in the images; e.g.
Fig. 1A,B).
Three Scomber japonicus (fork length 22cm, 24cm and
26cm) were used for both horizontal and vertical light-sheet
experiments. One other individual (23cm fork length) was
used for horizontal light-sheet experiments only. A nal
individual (20cm fork length) was used for vertical light-sheet
experiments only. Fish swam at speeds of 1.2 FLs
1
(the lowest
speed at which steady swimming was observed) and 2.2FLs
1
.
These speeds are within the range of swimming speeds
Atlantic mackerel can sustain for longer than 200min
(0.43.5total lengths s
1
, where TL is total length; Wardle and
He, 1988), are well below the maximum sustainable swimming
speeds of 45FLs
1
for S. japonicus of this size (Sepulveda
and Dickson, 2000) and match speeds used in a previous study
of S. japonicus caudal n kinematics (Gibb et al., 1999).
To obtain accurate force measurements for steady
swimming from the DPIV data (see below), video sequences
were analyzed only when mackerel were holding position for
several strokes and not drifting either vertically or laterally
because even small deviations from steady swimming
produced markedly different ow patterns in the wake.
Measurements of body angle to the X axis were obtained at the
time the vortex of interest was shed from the caudal n by
analyzing video images of mackerel body position obtained
by the second video camera synchronized with the DPIV
recordings (Fig. 1). Stroke duration was determined as the time
between maximum excursions of the caudal n in the XZ plane.
Digital particle image velocimetry (DPIV)
DPIV methods have been described in detail previously
(Willert and Gharib, 1991; Anderson, 1996; Raffel et al.,
1998); the specic methods used in the present study are based
on previous studies on the wakes of sh conducted in our
laboratory (e.g. Drucker and Lauder, 1999; Lauder, 2000; Liao
and Lauder, 2000; Wilga and Lauder, 2000). Video images
were imported into a PC computer using DT-Acquire software
with a Data Translation video card (Data Translation, Inc.,
USA). Using Insight software (v. 3; TSI, Inc., USA), selected
areas (e.g. Fig. 1A,B) of sequential pairs of video images (4ms
apart in time) were analyzed by subdividing the analysis
area of interest into a series of interrogation windows and
comparing these data subsets using two-frame cross-
Water
flow
Y
X
Z
Mirror
Laser
B
1
2
A
Y
X
Z
X
Fig. 1. Mackerel swam at a constant speed of
1.2 or 2.2FLs
1
, where FL is fork length, in
a ow tank, and a laser light sheet oriented
in the vertical (parasagittal, XY) (shown) or
horizontal (frontal, XZ) plane illuminated
small particles in the ow. Camera 1
recorded images of the caudal n and the
area of the light sheet posterior to it; in the
example shown in A, the tail is beating out
of the plane of the light sheet towards the
reader with the caudal n tilted at an acute
angle to the Y axis (as also shown by Gibb et
al., 1999). The ventral lobe of the caudal n,
illuminated by the light sheet, is trailing the
dorsal lobe, which has previously passed
through the sheet. Camera 2 recorded
synchronous images of the body of the
mackerel that were used to determine the
orientation of the body to the X axis. Images
of the light sheet oriented in the XZ plane (B)
were recorded from below the tank using a
front-surface mirror. Note that the caudal
peduncle and n blocked the light sheet as
they moved through it (the arrow in B
indicates the direction of n movement),
creating a shadow (seen, for example, in B).
Yellow dotted lines in A and B outline the
approximate image area analyzed using
digital particle image velocimetry. The white
scale bars in the lower left of A and B
represent 1cm.
BCF Ondulatrio
1716
vortices were symmetrical at 1.2FLs
1
. Vortex circulation and
momentum were also independent of individual (P=0.087 and
P=0.130, respectively; Table 2) and the interaction effect of
individuallight-sheet orientation (P=0.593 and P=0.559,
respectively, Table 2). Vortex circulation and momentum
values determined from the horizontal light-sheet data were
approximately twice the magnitude of those determined from
the vertical light-sheet data (Table 1) because the strong lateral
ow visible with the horizontal sheet (see, for
example, Figs 2 and 3) was out of the plane of the
vertical light sheet and was thus not viewed in the
vertical light-sheet images. Despite these
differences, the effect of light-sheet orientation was
not statistically signicant for vortex circulation or
momentum (P=0.018 and P=0.019, respectively,
Table 2) because a low P value was used to
determine signicance in order to correct for
multiple comparisons.
The distance between vortex centers was
independent of individual (P=0.043, Table 2) and
the interaction of individuallight-sheet orientation
(P=0.029, Table 2). Although the effect of light-
sheet orientation on the distance between vortex
centers was not statistically signicant (P=0.033),
ring width was on average 1.7 times ring height at
the speed of 1.2FLs
1
(Table 1), suggesting that the
rings were elliptical instead of circular. This ratio
increased to 1.9 at a speed of 2.2FLs
1
(Table 1),
indicating that the long axis of the ellipse increased
in length with increased speed.
Ring axis angle relative to the X axis was
independent of the effects of individual and
individuallight-sheet orientation (P=0.013 and
0.367, respectively; Table 2). Light-sheet
orientation had a signicant effect on ring axis angle
(P=0.006, Table 2). Jet angle to the X axis was also
independent of the effects of individual and
individuallight-sheet orientation (P=0.689 and
P=0.413, respectively, Table 2), but was dependent
on light-sheet orientation (P=0.001, Table 2). The
dependence of these variables on light-sheet
orientation reects the fact that the vertical and
horizontal light sheets showed perpendicular views
of the obliquely oriented vortex rings.
Wake views obtained using the vertical light
sheet (Fig. 4) suggested that the vertical ring axis
was approximately perpendicular to the X axis.
Values of ring axis angle to the X axis were close
to 90 for both speeds (Table 1) although a range
of values were seen, with some rings tilted by up to
30 from vertical (Fig. 6). Vertical light-sheet
views of the wake also suggested that the angle of
the jet to the X axis was nearly zero. Overall, jets
were angled slightly negative to the X axis at both
speeds (Table 1). The mean value of body angle at
the swimming speed of 1.2 FLs
1
(34, mean
S.D., N=20, Table 1) was signicantly different from zero (t-
test, P=0.02). In the vertical plane, jet angle was independent
of ring axis angle at the speed of 1.2 FLs
1
because the slope
of a linear regression model tted to the data was not
signicantly different from zero (ANOVA, P=0.57; Fig. 6). Jet
angles at the speed of 2.2FLs
1
showed a similar range and
relationship to ring axis angle to those at the lower speed
(Fig. 6).
J. C. Nauen and G. V. Lauder
30 40 50 60 70 80 90 100 110 120
10
20
30
40
50
60
70
80
12
4
4
12
20
40 50 60 70 80 90 100 110 120
10
20
30
40
50
60
70
80
40
24
8
8
24
40
X
Y
Y

p
o
s
i
t
i
o
n

(
m
m
)
Vorticity (s
1
)
20
0.10 m s
1
Vorticity (s
1
)
0.15 m s
1
Y

p
o
s
i
t
i
o
n

(
m
m
)
X position (mm)
Fig. 4. Views of the wake in the vertical (XY) plane of an individual of Scomber
japonicus 26cm in fork length swimming steadily at speeds of 1.2 FLs
1
(A) and
2.2 FLs
1
, where FL is fork length (B). The orientation of the caudal n to the
light sheet is shown in the diagram. Flow velocity is represented by black vectors
plotted over a color background indicating vorticity. At both swimming speeds,
the caudal n sheds one pair of counter-rotating vortices with a central jet of high-
velocity ow per stroke. The white arrows and labels in B illustrate the axes of
velocity proles presented in Fig. 5. X is the longitudinal axis of the ring; Y is
perpendicular to X.
1719 Hydrodynamics of mackerel caudal n locomotion
Anderson, 1996; Anderson et al., 1998). Reverse von Krmn
wakes are produced during caudal n locomotion by bluegill
sunsh Lepomis macrochirus (Lauder, 2000; Drucker and
Lauder, 2001b), giant danio Danio malabaricus (Anderson,
1996; Wolfgang et al., 1999), mullet Chelon labrosus (Mller
et al., 1997; Videler et al., 1999), trout Oncorhynchus mykiss
(Blickhan et al., 1992) and white sturgeon Acipenser
transmontanus (Liao and Lauder, 2000). The wake
morphology of a double row of pairs of vortices, reported for
caudal n locomotion by eel Anguilla anguilla (Mller et al.,
2001) and danio Brachydanio albolineatus (Rosen, 1959), was
not created by S. japonicus. In addition, the wake of a single
large vortex ring modeled for caudal n locomotion by eel
(Carling et al., 1998) and the model of wakeless caudal n
locomotion due to the sequential creation and destruction of
vortices modeled for a forked caudal n by Ahlborn et al.
(1991) were not supported by the present data. The timing of
vortex shedding during the rst quarter of each stroke in S.
japonicus is similar to that described for the danio Brachydanio
rerio (McCutchen, 1976) and the giant danio (Wolfgang et al.,
1999).
Vortex geometry
Vortex ring height was approximately equal to caudal n
height, indicating that caudal n size determines one parameter
of vortex size (as noted by Blickhan et al., 1992; Mller et al.,
2000). The magnitude of caudal n excursion in the Z dimension
determined the distance between the vortices in the XZ plane
and, thus, the length of the vortex ring. These data suggest that
the vortex rings were elliptical rather than round in shape
(Fig. 8), as was inferred for the vortex rings in the wake of the
mullet Chelon labrosus (Mller et al., 1997). As the length of
the ellipsoid was dependent on speed and was approximately
double the height of the ellipse at the speed of 2.2FLs
1
, the
data indicate that accurate measurements of ring geometry
(which are necessary for force calculations) require
measurements of vortex ring geometry in both the XY and XZ
planes.
The elliptical shape of the vortex rings seen here is a
departure from the axisymmetrical shape of an ideal vortex
ring (according to Helmholtzs theorem; see Milne-Thompson,
1966). A second departure of the present data from the
characteristics of an ideal vortex ring is seen in the geometric
relationship between the vertical ring axis and the central jet.
The central jet ow of an ideal vortex ring (such as that created
by ejection of uid from a circular nozzle by a piston, see
Webster and Longmire, 1997; Raffel et al., 1998) extends
perpendicular to the plane of the vortex ring. In the wake of
Scomber japonicus, however, the angle of the jet to the X axis
was typically close to 0, although the ring axis angle to the X
axis varied from 60 to 110 (with 90 being vertical).
Variability in the angle between the vertical ring axis and the
central jet has been seen in wakes created during pectoral n
locomotion by bluegill sunsh Lepomis macrochirus (Drucker
and Lauder, 1999) and during caudal n locomotion by
sturgeon Acipenser transmontanus (Liao and Lauder, 2000)
and L. macrochirus (Lauder, 2000). The described differences
between the wake structures seen here and those of ideal vortex
rings have been attributed to the fact that these wakes are
produced by reciprocating hydrofoils that are exible, deform
in a complex manner under load and are undergoing unsteady
motions [see, for example, bluegill pectoral n kinematics
(Gibb et al., 1994) and mackerel caudal n kinematics (Gibb
et al., 1999)]. Given the differences in shape and jet orientation
between ideal vortex rings and the wake described here, the
wake of Scomber japonicus may be most appropriately
described as a series of linked vortex loops rather than rings
(Dickinson and Gtz, 1996).
Fig. 8. Summary of the empirically
determined hydrodynamic forces
created by chub mackerel Scomber
japonicus swimming steadily at
1.2FLs
1
. Flow elds and forces are
depicted in the vertical (XY, top) and
horizontal (XZ, bottom) planes.
Vortex rings are shown in blue; ring
dimensions and orientation represent
those determined empirically in this
paper (Table 1). Thick black arrows
represent central jet ow. Red arrows
represent the reaction force on the
caudal n; in the vertical plane,
reaction force is directed behind the
body, so the red arrow is broken.
Thrust, lift and lateral force values are
the mean reaction forces averaged
over a stroke period. S. japonicus swims with its body slightly tilted down (3 on average); the jet is tilted down 3 in the opposite direction.
Thus, the reaction force at the caudal n has a lift component that acts over the lever arm of body length to the center of mass (black-and-white
checkered circle) to rotate the head down (green arrow). Lift generated by abducted pectoral ns (P), such as has been observed in previous
kinematic studies, could counterbalance lift generated at the caudal n.
X
Y
Z
X
Y
Z
67
11 mN
22 mN
1 mN
11 mN
3
P
3
BCF Ondulatrio
2087
INTRODUCTION
Skeleton-reinforced bio-membranes are ubiquitous and play critical
roles in many biological functions. For example, the flexibility of
the wings of many insects is determined primarily by the
architecture of the embedded veins. By experimentally measuring
and numerically characterizing the wing structure of 16 different
species of insects, Combes and Daniel concluded that an
anisotropic flexibility is achieved through the architecture of these
composite structures (Combes and Daniel, 2003a; Combes and
Daniel, 2003b). In all the species they studied, the spanwise
bending stiffness was found to be at least 12 orders of magnitude
higher than the chordwise bending stiffness (see also Newman and
Wootton, 1986; Wootton, 1992). It was attributed mostly to the
existence of clustered or thickened veins near the leading edge.
This suggests that, through specific arrangement of supporting
veins, these insect wings allow certain passive deformations while
discouraging others. Structures with similar architecture are also
found in the fins of aquatic animals. Many fishes possess caudal,
dorsal, pectoral or ventral fins that are stiffened by fin rays
embedded in a thin layer of collagenous membrane (Lauder, 1989;
Lauder and Drucker, 2004; Fish and Lauder, 2006; Lauder and
Madden, 2007). Overall, there are over 25000 species of fish with
these ray-strengthened fins. According to morphological studies
(Harder, 1971; Kardong, 1998), a fin ray contains a central bundle
of collagen surrounded by small segmented bony elements called
hemitrichs. These elements are paired and resemble a bimetallic
strip with two elongated bony elements separated by the central
collagen core. Along the length of a hemitrich, and at the ends,
there exist short ligaments and elastic fibers. In addition, the basal
end of each ray attaches to four separate muscles. This architecture
enables a fish to control the motion of each individual ray and
thus achieve three-dimensional fin deformations. This property is
expected to have fundamental effects on the hydrodynamic
performance of the fins in generating propulsion force during
cruising, bursting and maneuvering.
The exact function of the anisotropic flexibility of skeleton-
strengthened membranes in insect flying and fish swimming remains
to be fully understood. Despite a limited number of investigations
(Cheng et al., 1998; Pedley and Hill, 1999), existing studies in fish
swimming and insect flying are predominantly based upon the
assumption that the wings and fins are rigid (Liu and Kawachi, 1998;
Ramamurti and Sandberg, 2002; Ramamurti and Sandberg, 2007;
Wang, 2000; Pullin and Wang, 2004) or with prescribed
deformations (Liu and Bose, 1997; Ramamurti et al., 2005; Mittal
et al., 2006). It has always been speculated that the structural
flexibility of these composite bio-structures, with their stiffness
determined by the coupling of the embedded skeletons and the
surrounding materials, attributes to the highly efficient force
generation through flapping motions. Evidence of the beneficial
effects of structural flexibility in the generation of lift and propulsion
forces comes from different sources, in particular the analytical,
numerical and experimental investigations of flapping foils, as
discussed below.
Unlike conventional aero- and hydro-foils, which utilize steady
fluid dynamics for force generation, flapping foils are modeled on
fish fins and insect wings so that they produce aerodynamic or
hydrodynamic lift, propulsion or maneuvering forces using unsteady
foil oscillations (Triantafyllou et al., 1991; Tuncer and Platzer, 1996).
Through various investigations, it has been shown that structural
flexibility, per se, increases the propulsion efficiency in many cases.
For example, by experimentally studying the performance of a semi-
flexible flapping foil, Yamamoto et al. reported up to a 27% increase
in propulsion efficiency compared with a rigid foil (Yamamoto et
al., 1995). Similar performance enhancement was observed by
Heathcote et al. in their investigation of a thin steel plate undergoing
periodic heaving motion in still water (Heathcote et al., 2004). By
doi:10.1242/jeb.016279
Propulsion performance of a skeleton-strengthened fin
Qiang Zhu* and Kourosh Shoele
Department of Structural Engineering, University of California, San Diego, La Jolla, CA 92093, USA
*Author for correspondence (e-mail: qizhu@ucsd.edu)
Accepted 8 April 2008
SUMMARY
We examine numerically the performance of a thin foil reinforced by embedded rays resembling the caudal fins of many fishes. In
our study, the supporting rays are depicted as nonlinear EulerBernoulli beams with three-dimensional deformability. This
structural model is then incorporated into a boundary-element hydrodynamic model to achieve coupled fluidstructure interaction
simulation. Kinematically, we incorporate both a homocercal mode with dorso-ventral symmetry and a heterocercal mode with
dorso-ventral asymmetry. Using the homocercal mode, our results demonstrate that the anisotropic deformability of the ray-
reinforced fin significantly increases its capacity of force generation. This performance enhancement manifests as increased
propulsion efficiency, reduced transverse force and reduced sensitivity to kinematic parameters. Further reduction in transverse
force is observed by using the heterocercal mode. In the heterocercal model, the fin also generates a small lifting force, which
may be important in vertical maneuvers. Via three-dimensional flow visualization, a chain of vortex rings is observed in the wake.
Detailed features of the wake, e.g. the orientation of the vortex rings in the heterocercal mode, agree with predictions based upon
particle image velocimetry (PIV) measurements of flow around live fish.
Key words: fish locomotion, flexible fin, skeleton-reinforced membrane, fluidstructure interaction.
2089 Propulsion by skeleton-strengthened fin
By solving the potential flow around the body and determining
the flow potential , the hydrodynamic pressure p on the fin surface
S
b
is readily obtained through Bernoullis equation. We have:
where is the density of water (10
3
kgm
3
). Integrating p over the
fin surface, we obtain the hydrodynamic force as:
The power required to drive the fin is given as:
By definition, the thrust force, F
t
, is the x component of F. The
propulsion efficiency, , is defined as:
= F
t
U / P, (4)
where F
t
and P represent the thrust force and power consumption
averaged over one period T, respectively. The fin oscillation also
creates a transverse force F
r
, which is the y component of F. In the
heterocercal mode, there also exists a non-zero lifting force, F
z
,
leading to a mean lifting coefficient F
z
/GU
2
c
2
.
We thus characterize the performance of the fin by its mean thrust
coefficient, F
t
/GU
2
c
2
, the propulsion efficiency, , the mean lifting
coefficient, and the transverse force coefficient, F
r
(1)
/GU
2
c
2
, where:
As mentioned before, the structural function of the membrane in
which the rays are embedded is simplified as linear spring-dampers
spanning between neighboring rays. Specifically, in the following
calculations, we chose the spring constant to be 20Nm
2
, and the
(5)

. F
r
(1)
=
2
T
Re F
r
(t )e
it
dt
t
0
t
0
+T
{ }
(3)

P = p x , .
.
t
( )
n V
b
x ,t
( )
dx
S
b

(2)

F = p x , . t
( )
ndx
S
b
(1)

p x,t
( )
=
x,t
( )
t
+ ,
1
2
x,t
( )
2
damping constant to be 2Nsm

2
(both are measured per unit length).
With such a constraint, in the following simulations the maximum
variation of the surface area of the fin is around 510%.
We employ an iteration method to solve the fully coupled
fluidstructure interaction problem. Below we list the primary steps
of this algorithm.
(1) At time t, we start from an initial guess, S
b
(0)
, of the
configuration of the fin.
(2) The hydrodynamic problem is solved based on S
b
(0)
.
(3) After the hydrodynamic pressure, p, within panel j is evaluated
using Eqn1, we assume that the hydrodynamic force acting on this
panel, pnS
bi
(S
bi
is the area of the panel), is transferred equally
to four grid points that are closest to the panel on the two neighboring
rays (two points on each ray). Summing up the forces on all the
panels, the hydrodynamic loads, F
h
, on the nine rays are thus
determined.
(4) The BernoulliEuler beam equations (EqnA9) are numerically
integrated to obtain the updated positions of the rays. Based upon
these, the updated configuration of the fin, S
b
, is determined via
linear interpolation.
(5) If S
b
is not sufficiently close to S
b
(0)
, we set S
b
(0)
=S
b
and repeat
Steps 24.
To eliminate the well-known added mass induced instability, we
apply an implicit added mass scheme. In this approach, the added
mass effects are subtracted from both sides of EqnA9 during time
integration. As long as convergence of the iteration is achieved, the
exact values of the added mass will not affect the results. A detailed
description of this method is provided in Connell and Yue (Connell
and Yue, 2007).
Convergence tests
The validity and accuracy of both the EulerBernoulli beam solver
and the boundary integral solver have been extensively tested and
documented through numerous convergence studies and
comparisons with experiments or other numerical/theoretical
Base of a ray
(unsegmented)
Ray (segmented)
Membrane
Hemitrichs
Tendons
Tendons
Ligament
Cartilage pad
Tapering and
branching
Upper edge
Lower edge
Leading edge
Trailing edge
d
c
x
x
y
z
z
s
0
1
2
3
4
5
6
7
8
9
A C
B
Fig. 1. (A) Schematic showing a typical ray-strengthened caudal fin [modified from Alben et al. (Alben et al., 2007)]. We note that the rays are segmented and
are tapered and branched near the trailing edge. (B) The dorsal view of the internal structure of a ray [courtesy of S. Alben and the Royal Society (Alben et
al., 2007)]. (C) A geometrically and structurally simplified fin employed in the present modeling study. See List of symbols and abbreviations for definitions.
BCF Ondulatrio
2087
INTRODUCTION
discussed below.
doi:10.1242/jeb.016279
SUMMARY
2096
(3) Structural strength and lightness: the biomimetic composite
membrane provides a light structure with high strength.
(4) Deployability: the skeleton-reinforced membrane resembles
deployable structures such as cable roof buildings. These structures
can be easily folded and unfolded, making them an ideal design as
a portable device.
It is necessary to point out the limitations of the current model.
First, our work concerns a geometrically and structurally
simplified caudal fin. Effects of the fish body and attachments
such as dorsal, pectoral or anal fins are not considered. As
illustrated by Wilga and Lauder (Wilga and Lauder, 2002), in
leopard sharks (Triakis semifasciata) and bamboo sharks
(Chiloscyllium punctatum), the inclination angle of the body plays
a pivotal role in balancing the torque created by the tail. In
addition, as shown by Tytell (Tytell, 2006), vortices shed from
the dorsal and the anal fins of a bluegill sunfish may contribute
significantly to the near-body flow. Recent experiments with a
mechanical fish have shown that the fish body itself might also
generate a sequence of weaker vortex rings in the wake (Brucker
and Bleckmann, 2007). The dynamic effects of the vortices
generated upstream of the caudal fin depend not only on their
strength but also on their phase with respect to the caudal fin
(Zhu et al., 2002). This, in turn, is determined by the exact
morphology of the fish as well as its kinematics. Another effect
not considered in the present study is vorticity generation from
the fin surface in locations other than the trailing edge, in
particular the upper and lower edges. This vorticity
generation will not only change the local dynamics (e.g.
the pressure distribution) around the upper/lower edges
but will also affect the overall performance of the fin
in propulsion. To fully account for this effect, a
NavierStokes solver capable of studying three-
dimensional fluidstructure interactions is required
[for a possible method to achieve this, see Connell
(Connell, 2006)]. A fully viscous study will not only
illustrate effects of vortex generation from places other
than the trailing edge but will also provide more
accurate predictions of the flow field and the
hydrodynamic forces, especially in cases when the
Reynolds number is low. These NavierStokes
simulations, however, are computationally expensive.
In that respect, a potential flow method like ours
provides an alternative way to conduct systematic tests
over a wide range of parameters. To improve the
accuracy of our current method, investigations are
underway to develop methods to study vortex
generation from the leading edge area or the
upper/lower edges of the fin. The knowledge gained
from these studies will pave the way for future
simulations with more comprehensive numerical tools
and experiments using mechanical devices. These
studies may eventually contribute to the creation of
biomimetic apparatus that will be installed on AUVs
(autonomous underwater vehicles) or other vehicles (see Tangorra
et al., 2007).
Finally, we would like to mention that, in addition to insect wings
and fish fins, there exist a large number of bio-structures in nature
with similar structural characteristics, which can thus be categorized
as skeleton-reinforced membranes. A typical example is found in
the membrane of the erythrocyte (red blood cell), which features a
composite structure including a lipid bilayer and a protein skeleton
consisting primarily of actin and spectrin (Mohandas and Evans,
1994). The combined visco-elasticity of the bilayer and the skeleton,
as well as their dynamic inter-connectivity, impart remarkable
stability and deformability essential for the cell functionality when
it circulates around the body and squeezes through capillaries half
its own diameter. Similar structures are found in the biopolymer
membranes within the shells of mollusks such as the red abalone
(Haliotis rufescens) and the nautilus (Nautilus pompilius).
Remarkable mechanical properties of these structures (e.g. structural
strength, stability, durability and deployability) suggest biomimetic
applications and justify concentrated research efforts to understand
the detailed structural mechanics and fluidstructure interactions of
these bio-structures.
APPENDIX A. MODEL OF RAYS VIA NONLINEAR BEAM
DYNAMICS
Our ray model is based upon a three-dimensional nonlinear
EulerBernoulli beam formulation reported by Tjavaras et al.
Q. Zhu and K. Shoele
Fig. 13. Iso-surfaces of vorticity in the wakes behind (A) a rigid
fin in the homocercal mode, (B) a flexible fin in the homocercal
mode and (C) a flexible fin in the heterocercal mode. The iso-
surfaces are plotted at the level of 0.9 (after normalization by
U/c). S
t
=0.3. For the homocercal mode,
0
=10; for the
heterocercal mode,
0
=35. U, forward speed; c, chord length
of the foil.
2096
(3) Structural strength and lightness: the biomimetic composite
membrane provides a light structure with high strength.
(4) Deployability: the skeleton-reinforced membrane resembles
deployable structures such as cable roof buildings. These structures
can be easily folded and unfolded, making them an ideal design as
a portable device.
It is necessary to point out the limitations of the current model.
First, our work concerns a geometrically and structurally
simplified caudal fin. Effects of the fish body and attachments
such as dorsal, pectoral or anal fins are not considered. As
illustrated by Wilga and Lauder (Wilga and Lauder, 2002), in
leopard sharks (Triakis semifasciata) and bamboo sharks
(Chiloscyllium punctatum), the inclination angle of the body plays
a pivotal role in balancing the torque created by the tail. In
addition, as shown by Tytell (Tytell, 2006), vortices shed from
the dorsal and the anal fins of a bluegill sunfish may contribute
significantly to the near-body flow. Recent experiments with a
mechanical fish have shown that the fish body itself might also
generate a sequence of weaker vortex rings in the wake (Brucker
and Bleckmann, 2007). The dynamic effects of the vortices
generated upstream of the caudal fin depend not only on their
strength but also on their phase with respect to the caudal fin
(Zhu et al., 2002). This, in turn, is determined by the exact
morphology of the fish as well as its kinematics. Another effect
not considered in the present study is vorticity generation from
the fin surface in locations other than the trailing edge, in
particular the upper and lower edges. This vorticity
generation will not only change the local dynamics (e.g.
the pressure distribution) around the upper/lower edges
but will also affect the overall performance of the fin
in propulsion. To fully account for this effect, a
NavierStokes solver capable of studying three-
dimensional fluidstructure interactions is required
[for a possible method to achieve this, see Connell
(Connell, 2006)]. A fully viscous study will not only
illustrate effects of vortex generation from places other
than the trailing edge but will also provide more
accurate predictions of the flow field and the
hydrodynamic forces, especially in cases when the
Reynolds number is low. These NavierStokes
simulations, however, are computationally expensive.
In that respect, a potential flow method like ours
provides an alternative way to conduct systematic tests
over a wide range of parameters. To improve the
accuracy of our current method, investigations are
underway to develop methods to study vortex
generation from the leading edge area or the
upper/lower edges of the fin. The knowledge gained
from these studies will pave the way for future
simulations with more comprehensive numerical tools
and experiments using mechanical devices. These
studies may eventually contribute to the creation of
biomimetic apparatus that will be installed on AUVs
(autonomous underwater vehicles) or other vehicles (see Tangorra
et al., 2007).
Finally, we would like to mention that, in addition to insect wings
and fish fins, there exist a large number of bio-structures in nature
with similar structural characteristics, which can thus be categorized
as skeleton-reinforced membranes. A typical example is found in
the membrane of the erythrocyte (red blood cell), which features a
composite structure including a lipid bilayer and a protein skeleton
consisting primarily of actin and spectrin (Mohandas and Evans,
1994). The combined visco-elasticity of the bilayer and the skeleton,
as well as their dynamic inter-connectivity, impart remarkable
stability and deformability essential for the cell functionality when
it circulates around the body and squeezes through capillaries half
its own diameter. Similar structures are found in the biopolymer
membranes within the shells of mollusks such as the red abalone
(Haliotis rufescens) and the nautilus (Nautilus pompilius).
Remarkable mechanical properties of these structures (e.g. structural
strength, stability, durability and deployability) suggest biomimetic
applications and justify concentrated research efforts to understand
the detailed structural mechanics and fluidstructure interactions of
these bio-structures.
APPENDIX A. MODEL OF RAYS VIA NONLINEAR BEAM
DYNAMICS
Our ray model is based upon a three-dimensional nonlinear
EulerBernoulli beam formulation reported by Tjavaras et al.
Q. Zhu and K. Shoele
Fig. 13. Iso-surfaces of vorticity in the wakes behind (A) a rigid
fin in the homocercal mode, (B) a flexible fin in the homocercal
mode and (C) a flexible fin in the heterocercal mode. The iso-
surfaces are plotted at the level of 0.9 (after normalization by
U/c). S
t
=0.3. For the homocercal mode,
0
=10; for the
heterocercal mode,
0
=35. U, forward speed; c, chord length
of the foil.
BCF Ondulatrio
2087
INTRODUCTION
discussed below.
doi:10.1242/jeb.016279
SUMMARY
2097 Propulsion by skeleton-strengthened fin
(Tjavaras et al., 1998). This paradigm employs an EulerLagrangian
dual coordinate system, including the above-defined global
coordinate system (x, y, z) and a local reference system with unit
vectors , n and b in the tangential, normal and bi-normal directions
of the ray, respectively. The two coordinate systems are related by:
(A1)

.
,

n
b
1
]
1
1
1
= C
i
j
k
1
]
1
1
1
where i, j and k are unit vectors in the x, y and z directions,

respectively. C is the orthogonal tensor representing the change
of orientation from the global to the local reference frames. C is
constructed by using the singularity-free method of Euler
parameters based upon Eulers principal rotation theorem, which
depicts an arbitrary reorientation as a single rotation by a principal
angle about the principal unit vector of C, i.e. l.
Correspondingly, four Euler parameters
0
,
1
,
2
and
3
are
defined as:
The rotation matrix, C, thus becomes:
Dynamic equations of the ray movement are derived by
considering the conservation of translational and angular momenta
of an infinitesimal segment of the ray. We have:
and
where m is the mass per unit length, and s is the distance from this
point to the basal end along the unstretched ray.
V(s,t)=V
+V
n
n+V
b
b and (s,t)=
+
n
n+
b
b are the translational
and angular velocities, respectively. T(s,t)=T
+T
n
n+T
b
b is the
internal force inside the ray. M(s,t)=M
+M
n
n+M
b
b is the internal
moment. (s,t) is the strain. (s,t)=
+
n
n+
b
b is the Darboux
vector measuring the material torsion and curvatures of the ray. F
c
is the constraining force from the membrane that controls the
distance between neighboring rays. F
h
is the hydrodynamic force
to be defined later.
The compatibility relations, which state that the rays
configuration must be continuous in both space and time, should
also be satisfied as:
The internal forces and moments are related to the strain and
the Darboux vector through the constitutive relations T
=EA,
M
=GJ
, M
n
=EI
n
, M
b
=EI
b
. By definition, A is the cross-
sectional area. EI and GJ are the bending and torsional stiffnesses,
respectively. Our current simulations do not consider cases with ray
twisting; therefore, the value of GJ is irrelevant. At this point, a
hysteretic (material) damping can be incorporated by replacing the
Youngs modulus E with E+D/t, where D is the damping
(A2)

.

1
]
1
1
1
1
1
=
cos / 2
( )
l
x
sin / 2
( )
l
y
sin / 2
( )
l
z
sin / 2
( )
1
]
1
1
1
1
1
1
(A3)
.

C =
0
2
+
1
2

2
2

3
2
2
1
2
+
0
3
( )
2
1
3

0
2
( )
2
1
2

0
3
( )

0
2

1
2
+
2
2

3
2
2
2
3
+
0
1
( )
2
1
3
+
0
2
( )
2
2
3

0
1
( )

0
2

1
2

2
2
+
3
2
1
]
1
1
1
1
(A6)

t
+ 1+
( )
=
V
s
+ V .
(A4)

,

m
V
t
+ V
_
,
=
T
s
+ T + 1+
( )
F
h
+ F
c
( )
(A5)

M
s
+ M + 1+
( )
3
T = 0 ,
Fig. 14. Flow field within the y=0 plane behind a flexible fin in (A) the
homocercal mode and (B) the heterocercal mode. xl is the x-location of the
leading edge. The contour displays the y-component of the vorticity
(normalized by U/c). St=0.3. For the homocercal mode, 0=10; for the
heterocercal mode, 0=35.
MPF Ondulatrio
(a) (b)
(a)
Fig. 12. Vector analysis of an undulating n. (a) A single n-ray oscillating
exerts an upward thrust. (b) When many n-rays are connected via a exible
membrane (plan view), additional forces are exerted as indicated by the black
arrows. Their resultant is parallel to the n base. (c) Perspective view of
an undulating n, showing both force vectors. (Adapted from Breder and
Edgerton [77].)
(a)
(b) (c)
Fig. 13. Diagrams relating morphology to how the vector components of
undulating ns could be combined to yield a net forward thrust for (a) an
amiiform, (b) a gymnotiform, and (c) a balistiform swimmer.
propulsive waves combine to produce a net forward thrust
[Fig. 13(c)]. A signicant advantage of this arrangement is
that elaborate maneuvering can be achieved by varying the
individual force components of the median ns and direct
the resulting force vector with precision. Breder and Edgerton
observed this high degree of maneuverability for the seahorse
[77] which swims exclusively by undulations of its dorsal and
anal ns. They identied a number of physical and behavioral
factors that can alter the relative magnitude of the parallel and
normal force components. Physical factors include variations
in the interdistance, length, and exibility of the individual
n-rays. Behavioral factors affect the amplitude, wavelength,
and phase differences along the n and in time. The n-rays
also perform small longitudinal as well as lateral movements,
and they tend to be held like an open fan. The musculature
supporting the seahorse ns is exible and strong enough to
provide additional functionality; Blake [78] observed that the
ns can change their long axes relative to the body axis,
as well as move parts of the n relative to others. These
abilities are utilized during turn maneuvers to compensate for
the inexible body of the seahorse. Also, the entire n may be
held at various angles to the body, allowing it to be deected
far to one side and undulated in that position. Finally, most
undulatory median n swimmers are able to swim backward
just as effectively as forward by simply reversing the direction
of propagation of the propulsive wave [79], [80].
In undulating pectoral ns, the vertical force components
are lateral to the sh body and create yawing couples that
are generally cancelled out for symmetrical movements of
the ns. Powered maneuvers can be obtained by asymmetric
movements and different phase relationships in the undulating
paired ns [81].
Apart from swimming, sh utilize n undulations exten-
sively for hovering in mid-water. Small corrective forces are
generated by the ns to compensate for disturbances due to
pressure variations, minor sudden currents, or even the jet-
effect of the respiratory ow [26].
C. Mathematical Analysis
A simple method used to calculate the hydromechanical
efciencies for undulatory n swimmers is the actuator-disc
theory, a special application of the momentum principle in
uid dynamics. The mechanism operating on the uid (in this
case, an undulating n) is reduced to an idealized device
(actuator disc) that generates a pressure rise in the uid
passing through it. The thrust force can be calculated by
integrating the pressure rise over the whole disc. The main
advantage of this approach is that the n is regarded as a
black box, requiring no detailed knowledge of its kinematics.
However, the assumptions involved can be quite restrictive.
For applications of the actuator-disc theory to sh propulsion
and hovering, along with discussions on its limitations, see
[82][84].
The similarity of the waveforms observed in median ns
and those found in the undulating bodies of BCF swimmers
has encouraged the application of large-amplitude elongated-
body theory to the undulatory median n propulsion modes.
The initial work reported in [80], [84], and [85] was extended
in a series of papers [86][89] by Lighthill and Blake. It is
there shown that, for rigid deep-bodied sh, the momentum
shed into the water can be increased by a factor of about
three, compared to the momentum expected by the movement
of the ns on their own. This increment does not apply
to the shedding of unproductive energy into the wake.
Furthermore, the minimization of lateral forces, due to the
fact that they largely cancel out over the n length, means that
the sh body can remain rigid, avoiding increases in viscous
drag. These factors all combine to signicantly increase the
overall efciency of undulating median n propulsion. For
a speed range from 0.2 to 5 BL/s (corresponding to a
from 10 to 10 ), Blake [80] calculated a propulsive efciency
between 0.7 and 0.9 for electric eels and knifeshes. The ap-
plication of the latest wake theories developed for undulatory
MPF Ondulatrio
(a) (b)
(a)
Fig. 12. Vector analysis of an undulating n. (a) A single n-ray oscillating
exerts an upward thrust. (b) When many n-rays are connected via a exible
membrane (plan view), additional forces are exerted as indicated by the black
arrows. Their resultant is parallel to the n base. (c) Perspective view of
an undulating n, showing both force vectors. (Adapted from Breder and
Edgerton [77].)
(a)
(b) (c)
Fig. 13. Diagrams relating morphology to how the vector components of
undulating ns could be combined to yield a net forward thrust for (a) an
amiiform, (b) a gymnotiform, and (c) a balistiform swimmer.
propulsive waves combine to produce a net forward thrust
[Fig. 13(c)]. A signicant advantage of this arrangement is
that elaborate maneuvering can be achieved by varying the
individual force components of the median ns and direct
the resulting force vector with precision. Breder and Edgerton
observed this high degree of maneuverability for the seahorse
[77] which swims exclusively by undulations of its dorsal and
anal ns. They identied a number of physical and behavioral
factors that can alter the relative magnitude of the parallel and
normal force components. Physical factors include variations
in the interdistance, length, and exibility of the individual
n-rays. Behavioral factors affect the amplitude, wavelength,
and phase differences along the n and in time. The n-rays
also perform small longitudinal as well as lateral movements,
and they tend to be held like an open fan. The musculature
supporting the seahorse ns is exible and strong enough to
provide additional functionality; Blake [78] observed that the
ns can change their long axes relative to the body axis,
as well as move parts of the n relative to others. These
abilities are utilized during turn maneuvers to compensate for
the inexible body of the seahorse. Also, the entire n may be
held at various angles to the body, allowing it to be deected
far to one side and undulated in that position. Finally, most
undulatory median n swimmers are able to swim backward
just as effectively as forward by simply reversing the direction
of propagation of the propulsive wave [79], [80].
In undulating pectoral ns, the vertical force components
are lateral to the sh body and create yawing couples that
are generally cancelled out for symmetrical movements of
the ns. Powered maneuvers can be obtained by asymmetric
movements and different phase relationships in the undulating
paired ns [81].
Apart from swimming, sh utilize n undulations exten-
sively for hovering in mid-water. Small corrective forces are
generated by the ns to compensate for disturbances due to
pressure variations, minor sudden currents, or even the jet-
effect of the respiratory ow [26].
C. Mathematical Analysis
A simple method used to calculate the hydromechanical
efciencies for undulatory n swimmers is the actuator-disc
theory, a special application of the momentum principle in
uid dynamics. The mechanism operating on the uid (in this
case, an undulating n) is reduced to an idealized device
(actuator disc) that generates a pressure rise in the uid
passing through it. The thrust force can be calculated by
integrating the pressure rise over the whole disc. The main
advantage of this approach is that the n is regarded as a
black box, requiring no detailed knowledge of its kinematics.
However, the assumptions involved can be quite restrictive.
For applications of the actuator-disc theory to sh propulsion
and hovering, along with discussions on its limitations, see
[82][84].
The similarity of the waveforms observed in median ns
and those found in the undulating bodies of BCF swimmers
has encouraged the application of large-amplitude elongated-
body theory to the undulatory median n propulsion modes.
The initial work reported in [80], [84], and [85] was extended
in a series of papers [86][89] by Lighthill and Blake. It is
there shown that, for rigid deep-bodied sh, the momentum
shed into the water can be increased by a factor of about
three, compared to the momentum expected by the movement
of the ns on their own. This increment does not apply
to the shedding of unproductive energy into the wake.
Furthermore, the minimization of lateral forces, due to the
fact that they largely cancel out over the n length, means that
the sh body can remain rigid, avoiding increases in viscous
drag. These factors all combine to signicantly increase the
overall efciency of undulating median n propulsion. For
a speed range from 0.2 to 5 BL/s (corresponding to a
from 10 to 10 ), Blake [80] calculated a propulsive efciency
between 0.7 and 0.9 for electric eels and knifeshes. The ap-
plication of the latest wake theories developed for undulatory
MPF Ondulatrio
3491 Ribbon-fin propulsion
hydrodynamics of ribbon-fin propulsion. Second, artificial ribbon
fins may provide a superior actuator for use in highly maneuverable
underwater vehicles for applications such as environmental
monitoring (Epstein et al., 2006; MacIver et al., 2004).
We use computational fluid dynamics to examine the flow
structures and forces arising from a sinusoidally actuated ribbon
fin. We compare the computed flow structures with those measured
from a robotic ribbon fin using digital particle image velocimetry
(DPIV) and compare the computed surge force with the drag force
measured from towing a cast of the fish. Whereas tow drag can
provide a useful estimate of the thrust needed during steady
swimming, for the impulsive motions modeled in this study, the
thrust needed to undergo typical accelerations is more directly
relevant. Thus, we also compare computed forces with the thrust
that we estimate is needed for two different types of swimming
direction reversals: reversals that occur during prey capture strikes
from kinematic data collected in a previous study (MacIver et al.,
2001) and reversals that occur during refuge tracking behavior,
where fish placed in a sinusoidally oscillating refuge will move to
maintain constant position with respect to the refuge.
For the present study, we idealize the fin kinematics as a
traveling sinusoid on an otherwise stationary (i.e. non-translating,
non-rotating) membrane (Fig. 1C). As a consequence, the top edge
of the fin remains fixed at all times, and all points on the fin below
this edge move in a sinusoidal manner. The fin deformation is
specified in Eqn1 below. As indicated in Fig. 1C, positive surge is
defined as the force on the fin from the fluid in the direction from
the tail to the head. If the traveling wave passes from the tail to the
head, then the force on the fin from the fluid would be from the
head to tail, corresponding to negative surge. Positive heave is
vertically upward.
We chose to characterize the hydrodynamics of a fixed fin in a
stationary flow because this is most relevant for understanding forces
arising from maneuvering movements that occur when the body is
at near-zero velocity with respect to the fluid far away from the
body. In future work, using this approach will also allow us to
compare our simulated force estimates with those obtained
empirically from a robotic ribbon fin placed on a linear track pushing
against a force sensor (Epstein et al., 2006). In subsequent studies,
we will be examining the hydrodynamics of a stationary fin under
imposed flow conditions and when the fin and an attached body are
allowed to self-propel through the fluid.
Flow visualizations from computational simulations and DPIV
indicate that the mechanism of thrust generation is a streamwise central
jet and associated attached vortex rings. We show that, despite the
lack of cylindrical symmetry in the morphology of the fin, its peculiar
deformation pattern the traveling wave produces a jet flow often
found in highly symmetric animal forms, such as jellyfish and squid.
Whereas previous research focused exclusively on the surge force
(Blake, 1983; Lighthill and Blake, 1990), we find that the ribbon fin
is also able to generate a heave force, which pushes the body up. This
arises from the generation of counter-rotating axial vortex pairs that
are shed downward and laterally from the bottom edge of the fin. We
hypothesize that the slanted angle of the fin base with respect to the
spine observed in many gymnotids (e.g. Fig. 1A) leverages this heave
force for forward translation. We also find that, in certain cases, as
the number of waves on the fin decreases to below approximately
two-thirds, the heave force surpasses the surge force. This switchover
from an undulatory parallel thrust mode to an oscillatory normal thrust
mode may provide insight into how the position and orientation of
median fins varies with the length of the fin and the number of
wavelengths that can be placed on it.
We show how the surge force from the fin scales as a function
of a few key parameters. We found that for a stationary fin without
imposed flow: (1) the surge force is proportional to (frequency)
2
(angular amplitude)
3.5
(fin height/fin length)
3.9
(wavelength/fin
length), where is a function that approximates the variation of
surge force as a function of wavelength normalized with fin length;
(2) for angular deflections above =10deg., where is defined in
Fig. 1C, previous analytical work (Lighthill and Blake, 1990)
underestimated the magnitude of surge force; (3) surge force shows
a peak when the wavelength is approximately half the fin length,
similar to what is observed biologically (Blake, 1983) and contrary
B
C
A
Flow: 15 cm s
1
Surge
Roll
Head
Wave motion
Tail
Sway
Pitch
Heave
Yaw
L
fin
h
fin
x
y
z
m
r
Fig. 1. Apteronotus albifrons, the black ghost knifefish of South America.
(A) Photograph courtesy of Neil Hepworth, Practical Fishkeeping magazine.
Inset shows side view of the fish to illustrate the angle of the fin with
respect to the body. (B) Frame of video of the fish swimming in a water
tunnel, from ventral side. (C) Model of the ribbon fin, shown together with
symbols for fin length (L
fin
), fin height (h
fin
) and two of the kinematic state
variables, angular deflection () and the vector from the rotational axis to a
point on the fin (r
m
). The Eulerian (fluid) reference frame (x, y, z; positive
direction indicated by inset) is shown together with names for velocities in
the body frame fixed to the rigid fin rotation axis indicated by the red line
(surge, heave, sway, roll, pitch, yaw; positive direction indicated by arrows).
In this work, simulations were performed with the body-fixed frame oriented
to the Eulerian frame as illustrated, and forces are with respect to a
traveling wave moving in the head-to-tail direction indicated.
MPF Oscilatrio
570
Outgroup
CP
G. V. LAUDER AND B. C. JAYNE
Centrarchidae
Morone
cyanellus microlophus macrochirus
Micropterus Pomoxis
Lepomis
FIG. 2. Phylogenetic relationships of representative taxa within the North American teleost fish family Cen-
trarchidae (following Mabee [1993, 1995] and Wainwright and Lauder [1992]) and an outgroup clade (Perci-
chthyidae: Morone) to show evolutionary patterns to pectoral fin shape within this clade. Images of the left
pectoral fin above each clade have been scaled proportionally to the size of the fin in that clade, and all taxa
are drawn to the same total length. Note the trend in the Centrarchidae from a primitive short and blunt fin
(suggested to be characteristic of drag-based propulsion) to the longer wing-like fin used in lift-based propulsion.
Images of fishes modified from McGillis (1984), Freshwater Fishes of California, 1984 by the Regents of the
University of California.
that this condition is primitive for sunfishes
as a clade.
Lift-based propulsion is associated with
more "wing-like" fin shapes where the dis-
tal tip is tapered and the fin as a whole is
more "diamond-shaped." Species within
the sunfish genus Lepomis possess more
wing-like pectoral fins (Fig. 2) with greater
relative area than the blunt fins character-
istic of basal groups in this clade.
METHODS FOR THREE-DIMENSIONAL
ANALYSIS
The most basic data needed for an anal-
ysis of pectoral fin locomotion in fishes are
kinematic. Without an understanding of
how the fin moves in three-dimensional
space we lack the ability to conduct accu-
rate modeling (either empirical or theoreti-
cal), estimate thrust production, or under-
stand the extent to which fin movements are
under active muscular control. Fin motion
is inherently three-dimensional as the pec-
toral fin typically moves in both dorsoven-
tral and anteroposterior directions during
swimming due to the oblique orientation of
the base of the fin with the body. Therefore,
it is essential that kinematic data be mea-
sured in three-dimensional space and that
specific marked points on the fin be fol-
lowed through time. Tracking the move-
ment of individual markers allows both the
visualization of parts of the fin that can not
normally be seen due to the diaphanous na-
ture of the fin membrane, and also permits
division of the fin surface into smaller units
that can be analyzed individually. A full
three-dimensional analysis also allows cal-
culation of 3D angles of attack of different
portions of the fin surface, rather than re-
stricting angular calculations to their pro-
jection onto one plane. As we will show
below, two-dimensional analyses can result
in gross errors that severely compromise
understanding the mechanics and hydrody-
namics of pectoral propulsion.
In order to analyze movement of the
pectoral fin, it is useful to visualize the fin

a
t

U
n
i
v
e
r
s
i
d
a
d
e

F
e
d
e
r
a
l

d
e

M
i
n
a
s

G
e
r
a
i
s

o
n

N
o
v
e
m
b
e
r

7
,

2
0
1
0
i
c
b
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

MPF Oscilatrio
BCF propulsion to gymnotiform and amiiform locomotion
presents an interesting eld of research that, along with ow
visualization experiments, could determine whether vorticity
control mechanisms are employed by sh swimming in these
modes.
Finally, rajiform locomotion has been analyzed by Daniel
using a combination of unsteady aerofoil theory and blade-
element theory in [90], where the signicance of unsteady
effects and wing shape in thrust generation is demonstrated.
V. MPF OSCILLATIONS
A. General
Fin oscillations usually involve short-based median or
paired ns. In tetraodontiform mode, the dorsal and anal ns
are apped as a unit, either in phase or alternating to achieve
propulsion. The ocean sunsh is an extreme example of
tetraodontiform swimmer: it has virtually no caudal n or body
musculature and propels itself by synchronized oscillations of
its very high dorsal and anal ns. Tetraodontiform mode
can be viewed as a continuation of balistiform mode, where
the wavelength of the propulsive wave is very large, and,
consequently, the individual n-rays oscillate more or less
in phase.
In labriform mode, propulsion is achieved by oscillatory
movements of the pectoral ns. Due to the large variability
of these movements, as well as the signicance of pectoral
n swimming amongst sh and the potential for building
stabilization/maneuvering devices based on them (see, for
example, [7] and [91]), the following section is dedicated to a
more detailed discussion of labriform locomotion.
B. Labriform Swimming
Swimming using the pectoral ns is widespread among
teleost sh, but only recently has it received scientic atten-
tion. This is largely because of the difculty in observing and
analyzing the n kinematics due to the speed, variability, and
complexity of the movements performed (apping, rotations
and undulations), as well as the transparent nature of the
n membrane. Recently, a number of sophisticated lming
techniques have evolved, enabling the acquisition of detailed
kinematic data, that can help us gain a better understanding of
the hydrodynamic forces involved.
Blake [82] identied two main oscillatory movement types
for the pectoral ns: 1) a rowing action (drag-based labri-
form mode) and 2) a apping action, similar to that of
bird wings (lift-based labriform mode). According to Vogel
[13], drag-based methods are more efcient at slow speeds,
when the chordwise ow over the n is small, while lift-based
methods are more efcient at higher speeds. Later observations
(see [91] and [92]) emphasized the importance of acceleration
reaction in thrust generation. They also indicate that pectoral
n movements are usually very complicated owing to the
highly exible character of the membrane and the n-rays,
as well as to the hydrodynamic interactions of the ns with
the moving water and the sh body. Thus, sh rarely exhibit
a clearly rowing or apping movement. Instead, they use a
(a) (b)
Fig. 14. Diagram showing the n positions and attack angles during (a) the
power stroke and (b) the recovery stroke for a sh swimming in drag-based
labriform mode. (Adapted from Blake [84].)
combination of them that generally varies with speed. Undula-
tions are also often passed along the ns (diodontiform mode),
and the great diversity of movements attainable can generate
thrust in almost any direction, achieving high maneuverability.
The complexity of the pectoral n motions is illustrated in
the detailed 3-D kinematic data recently available [91], [92],
[93]. Comprehensive reviews of pectoral n swimming can be
found in [18] and [91]; the latter discusses a number of issues
pertinent to the design of articial ns for use in underwater
vehicles. To understand the basics of thrust generation in
pectoral n movements, it is helpful to go back into the original
studies of the purely drag- and lift-based labriform locomotion,
for which mathematical models have been easier to develop.
1) Drag-Based Mode: Blake presents kinematic data and
a mechanical analysis of drag-based labriform locomotion in
[94] and [95], as it is utilized by angelsh for an extensive
range of swimming speeds. The ns usually have a short base
that forms a high angle with the main axis. Rowing action con-
sists of two phases [94]: the power stroke, when the ns move
posteriorly perpendicular to the body at a high attack angle
and with a velocity greater than the overall swimming speed
[Fig. 14(a)], and a recovery stroke, when the ns are feath-
ered to reduce resistance and brought forward [Fig. 14(b)].
Thrust is generated due to the drag encountered as the n is
moved posteriorly, as well as due to the acceleration reaction
of the water being rapidly hauled at the initial part of the power
stroke. Since thrust is only produced during the power stroke, it
is discontinuous. This is in contrast to BCF propulsion, where
a usefully directed thrust force is generated over most of the
tail-beat cycle.
Blade-element theory has been applied to drag-based labri-
form propulsion, whereby the pectoral ns are divided into a
number of rigid sections, each inclined at an angle to the inci-
dent ow. According to the results obtained for an 8-cm-long
angelsh specimen swimming at about 0.5 BL/s, the outermost
40% of the n area produces over 80% of the total hydrody-
namic force. A propulsive efciency of 16% for the complete
rowing stroke is derived using the same calculations [95].
A simple hydromechanical model developed in [96] predicts
PECTORAL FIN PROPULSION IN FISHES 577
FIG. 8. Three-dimensional reconstruction of the position and movement of the bass pectoral fin during adduction
and retraction (times T, and T
2
see Fig. 7). Note that the fin has been enlarged relative to the body to illustrate
triangle and vector orientations. This reconstruction is based on the X, Y, and Z coordinates of the three fin
markers and a point at the base of the pectoral fin; small spheres indicate the location of fin markers and the
base of the fin used for three-dimensional calculations of triangle position. Vectors originating at the centroid
of each of the two fin triangles show the orientation of the triangular surface: each vector is oriented perpen-
dicular to the fin triangle plane. The length of each vector is proportional to the squared velocity of the triangle
centroid in the direction normal to the triangle surface (see text for further discussion). Note that at time T
2
the
surface of the upper triangle (A) is oriented posteriorly. Triangle B is not visible in dorsal view at time T,
because it is hidden beneath triangle A. The direction of triangle A centroid movement in shown by the arrow
below each fish image. At time T
2
the centroid of triangle A is moving anteriorly and ventrally as seen in lateral
view.
of triangle A is moving laterally, ventrally
and anteriorly, folding the fin along the bor-
der between triangles A and B.
One likely cause of the complexity of fin
movements in bass is the oblique articula-
tion of the fin with the body. The axis of
the fin base is inclined from anterodorsal to
posteroventral (Fig. 2) and the path of fin
motion may thus be largely determined by
anatomy. The orientation of the fin ray bas-
es, their attachment to the underlying radial
elements, and the lines of action of abduc-
tor and adductor muscles may render purely
translational motion along any one axis im-
possible.
COMPARISONS TO OTHER TAXA
While three-dimensional kinematic data
of this kind are not yet available for any
other fish taxon, a basic three-dimensional
description of fin movement in bluegill, Le-
pontis macrochirus, with marked pectoral
fins was presented by Gibb et ai, (1994).
A comparison between bluegill and bass is
instructive because although both species
are members of the same sunfish clade, they
possess patterns of fin structure that would
be expected a priori to conform to lift-
based and drag-based kinematic patterns re-
spectively (Fig. 2). In addition, both species
were studied under the same experimental
conditions and are size-matched by total
length: bluegill averaged nearly 18 cm in
total length (Gibb et al., 1994) while the
bass used for the experiments reported in
this paper had a mean length of about 21
cm.
At a similar total length, bluegill possess
pectoral fins of greater maximum length
and area than bass: the distance from the
most dorsal marker to the fin base is 4.3 cm
in bluegill compared to 2.8 cm in bass,
while the respective total fin areas are 6.5
cm
2
and 4.0 cm
2
.
Bass show little change in fin beat fre-
quency with speed, and over a speed range
of 0.3 to 0.75 Lsec"
1
use significantly high-
er frequencies than bluegill. For example,
at 0.5 Lsec"
1
, bluegill pectoral fin beat fre-
quencies average about 1.35 Hz, while bass
beat frequencies reach almost 2 Hz. Blue-
gill beat frequencies are strongly speed de-

a
t

U
n
i
v
e
r
s
i
d
a
d
e

F
e
d
e
r
a
l

d
e

M
i
n
a
s

G
e
r
a
i
s

o
n

N
o
v
e
m
b
e
r

7
,

2
0
1
0
i
c
b
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

MPF Oscilatrio
212
symmetric biplane during station keeping in a uniform stream
during the outstroke half of the fin beat. The flexibility is a means
of making two fins out of one in order to balance unsteady forces
in the vertical direction. The formation of the two attached edge
vortices shows that the dynamic stall mechanism is present in
sunfish pectoral fins although the projected fin area is much
reduced from the maximum possible.
To achieve station keeping, it is necessary to be able to promptly
produce forces and moments along stream and also normal and cross-
stream in amplitudes that are just enough to cancel the perturbations.
The flexible fin has the ability to control the projected surface area
in all planes as indeed we see in the movie frames in Lauder et al.
(Lauder et al., 2007). The fin folding also changes the local angles
of attack. The twist of the fin can also be controlled at the root,
thereby controlling the vector of the co-flowing jet (Fig.14). These
three traits, with the aid of sensitive lateral line sensors and an
instantaneously deployable controller, could dynamically cancel
force and moment perturbations in all directions and axes. The rigid
fin measurements of force vectors at each instant were examined in
three-dimensional fields. Rigid fins produce large transverse periodic
forces, which may be undesirable when holding position (Beal and
Bandyopadhyay, 2007). To achieve station keeping in the vertical
plane, two symmetric fins are needed to balance the instantaneous
vertical (Y-) forces. Such data show quantitatively that station
keeping can be achieved at all times by symmetric biplanes,
Research article
conjoined or not. The symmetric fins would not have to be mirror
images; their angles of attack to the flow and flapping speed could
be different as long as they cancel undesired unsteady forces. A
sample of the sunfish pitch angle time history from Lauder et al.
(Lauder et al., 2007) is compared with a similar sinusoidal history of
the rigid fin in Fig.15, where a qualitative similarity is seen to exist.
The sum of the pitch angles between the dorsal and ventral rays is
within 0 to 20 during abduction, while it is closer to zero during
the adduction phase. The pitch angles in the biorobotic rigid fins, of
course, sum exactly to zero at all times. In the absence of detailed
kinematics of the sunfish pectoral fin, this zero sum is taken as an
indication of real-time station keeping.
Discussion
What can we learn by comparing the force production in sunfish
due to their flexible pectoral fins and the present relatively stiff
penguin wing-like biorobotic fins? Lauder et al. (Lauder et al.,
2007) have given the time sequences of the flexible fin contortions
and the distributions of local velocity variations about the
freestream velocity for a sunfish holding station in a stream. We
propose the following mechanism to be in play in the flexible fin.
During the outstroke, while the spanwise edges of the fin are
folding inward, due to separation and the ensuing pressure
difference, two symmetric leading-edge vortices are formed
(Fig.14). The projected area of the fin is much reduced from its
maximum value during the cupping process, which suggests that
its thrust production role is smaller. The two leading-edge vortices
coalesce to form a downstream-pointing jet. The spanwise bone
structure and the chordwise fin corrugations would assist this
spanwise jet flow. The jet is inclined rearward to the body and is a
source of thrust. The outstroke takes the fin to a position just short
of being normal to the body. The fin subsequently expands, which
causes the twin leading-edge vortice jet to expand into a diffuser,
allowing pressure recovery, at the end of the outstroke. During the
return stroke, the fin concaves with the fin tip facing upstream. Due
to stiffness, the spanwise edges of the fin this time do not cup
inward and no significant leading-edge vortex is probably
produced, although a stall vortex could still form at the tip.
Conceivably, this tip vortex inclined cross-stream and parallel to
the body could interact with the caudal fin downstream, providing
a means for precision streamwise control of the fish body to station
itself in the face of perturbations, because it is sinusoidal and small
Fig.14. Proposed dynamic stall vortex pairs shown in color in the sunfish
pectoral fin formed during outstroke when the fin is undergoing cupping
motion. The fin picture is from Lauder et al. (Lauder et al., 2007). The total
fish length is 17cm and the stream speed is 8.5cms
1
. The fish is
maintaining its station in a uniform stream, shown by the vertical arrow,
while the fin is turning upstream and the spanwise edges are curling
inward. The stall vortices locally augment the pressure difference across
the two leading edges and could help cancel perturbations in the vertical
directions. The co-flowing jets on both sides of the fish could be vectored
appropriately to hold station laterally and provide some thrust. The vertical
arrow shows the stream direction.
Time (s)
P
i
t
c
h

(
d
e
g
r
e
e
s
)
Ventral
(ray)
Dorsal
(ray)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
100
80
60
40
20
20
40
60
80
100
Fig.15. Comparison of measurements of the variation in fin pitch angle
with time during one cycle (red and blue diamonds) in bluegill sunfish
pectoral fins (Lauder et al., 2007) with our rigid fin data. Sunfish pectoral
fin: roll amplitude, 40.8; frequency, 1.0Hz; pitch amplitude, 44.8.
212
symmetric biplane during station keeping in a uniform stream
during the outstroke half of the fin beat. The flexibility is a means
of making two fins out of one in order to balance unsteady forces
in the vertical direction. The formation of the two attached edge
vortices shows that the dynamic stall mechanism is present in
sunfish pectoral fins although the projected fin area is much
reduced from the maximum possible.
To achieve station keeping, it is necessary to be able to promptly
produce forces and moments along stream and also normal and cross-
stream in amplitudes that are just enough to cancel the perturbations.
The flexible fin has the ability to control the projected surface area
in all planes as indeed we see in the movie frames in Lauder et al.
(Lauder et al., 2007). The fin folding also changes the local angles
of attack. The twist of the fin can also be controlled at the root,
thereby controlling the vector of the co-flowing jet (Fig.14). These
three traits, with the aid of sensitive lateral line sensors and an
instantaneously deployable controller, could dynamically cancel
force and moment perturbations in all directions and axes. The rigid
fin measurements of force vectors at each instant were examined in
three-dimensional fields. Rigid fins produce large transverse periodic
forces, which may be undesirable when holding position (Beal and
Bandyopadhyay, 2007). To achieve station keeping in the vertical
plane, two symmetric fins are needed to balance the instantaneous
vertical (Y-) forces. Such data show quantitatively that station
keeping can be achieved at all times by symmetric biplanes,
Research article
conjoined or not. The symmetric fins would not have to be mirror
images; their angles of attack to the flow and flapping speed could
be different as long as they cancel undesired unsteady forces. A
sample of the sunfish pitch angle time history from Lauder et al.
(Lauder et al., 2007) is compared with a similar sinusoidal history of
the rigid fin in Fig.15, where a qualitative similarity is seen to exist.
The sum of the pitch angles between the dorsal and ventral rays is
within 0 to 20 during abduction, while it is closer to zero during
the adduction phase. The pitch angles in the biorobotic rigid fins, of
course, sum exactly to zero at all times. In the absence of detailed
kinematics of the sunfish pectoral fin, this zero sum is taken as an
indication of real-time station keeping.
Discussion
What can we learn by comparing the force production in sunfish
due to their flexible pectoral fins and the present relatively stiff
penguin wing-like biorobotic fins? Lauder et al. (Lauder et al.,
2007) have given the time sequences of the flexible fin contortions
and the distributions of local velocity variations about the
freestream velocity for a sunfish holding station in a stream. We
propose the following mechanism to be in play in the flexible fin.
During the outstroke, while the spanwise edges of the fin are
folding inward, due to separation and the ensuing pressure
difference, two symmetric leading-edge vortices are formed
(Fig.14). The projected area of the fin is much reduced from its
maximum value during the cupping process, which suggests that
its thrust production role is smaller. The two leading-edge vortices
coalesce to form a downstream-pointing jet. The spanwise bone
structure and the chordwise fin corrugations would assist this
spanwise jet flow. The jet is inclined rearward to the body and is a
source of thrust. The outstroke takes the fin to a position just short
of being normal to the body. The fin subsequently expands, which
causes the twin leading-edge vortice jet to expand into a diffuser,
allowing pressure recovery, at the end of the outstroke. During the
return stroke, the fin concaves with the fin tip facing upstream. Due
to stiffness, the spanwise edges of the fin this time do not cup
inward and no significant leading-edge vortex is probably
produced, although a stall vortex could still form at the tip.
Conceivably, this tip vortex inclined cross-stream and parallel to
the body could interact with the caudal fin downstream, providing
a means for precision streamwise control of the fish body to station
itself in the face of perturbations, because it is sinusoidal and small
Fig.14. Proposed dynamic stall vortex pairs shown in color in the sunfish
pectoral fin formed during outstroke when the fin is undergoing cupping
motion. The fin picture is from Lauder et al. (Lauder et al., 2007). The total
fish length is 17cm and the stream speed is 8.5cms
1
. The fish is
maintaining its station in a uniform stream, shown by the vertical arrow,
while the fin is turning upstream and the spanwise edges are curling
inward. The stall vortices locally augment the pressure difference across
the two leading edges and could help cancel perturbations in the vertical
directions. The co-flowing jets on both sides of the fish could be vectored
appropriately to hold station laterally and provide some thrust. The vertical
arrow shows the stream direction.
Time (s)
P
i
t
c
h

(
d
e
g
r
e
e
s
)
Ventral
(ray)
Dorsal
(ray)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
100
80
60
40
20
20
40
60
80
100
Fig.15. Comparison of measurements of the variation in fin pitch angle
with time during one cycle (red and blue diamonds) in bluegill sunfish
pectoral fins (Lauder et al., 2007) with our rigid fin data. Sunfish pectoral
fin: roll amplitude, 40.8; frequency, 1.0Hz; pitch amplitude, 44.8.
MPF Oscilatrio
that, for a given planform area, triangular ns will create less
interference drag over the sh body than square or rectangular
ones. This is in accordance with the actual n shape observed
in drag-based labriform swimmers. More recently, Kato and
Inaba used the unsteady vortex lattice method to calculate
the hydrodynamic forces on a rigid pectoral n model [3]
in drag-based labriform mode. The propulsive efciencies
calculated do not exceed 10%, a result in accordance with
their experimental measurements and the predictions of blade-
element theory. Despite these low values, Blake suggested [94]
that rowing propulsion is more efcient for slow swimming
than BCF modes, the efciency of which falls off rapidly
for decreasing speed. There is evidence for that in nature, as
many sh use labriform locomotion for slow-speed swimming,
switching to BCF propulsion at higher speeds. In [97], the
velocity at which this transition occurs for a certain species
(Notothenia Neglecta, average adult length 28 cm) is quoted
as 0.8 BL/s. However, the use of labriform locomotion at
low speed could be attributed to nonenergetic factors, such as
higher maneuverability or being less conspicuous to predators.
2) Lift-Based Mode: Lift forces are generated in the plane
perpendicular to the direction of the n motion, whereas drag
forces appear in the plane of the n motion. As a result, in
lift-based labriform mode, for the pectoral ns to propel the
sh forward, they have to move up and down in a plane that
is roughly perpendicular to the main axis of the shs body.
This implies that no recovery stroke is necessary and lift can
be generated during both the upstroke and the downstroke.
Additionally, lift forces can be an order of magnitude greater
than the drag forces generated by a n of the same area.
Thus, lift-based ns can generate larger, more continuous,
and more efcient thrust than ns performing rowing motions
(see [91] for relevant data). The n shapes for lift-based
labriform swimmers tend to differ from those using drag-based
mechanics. One reason is the need to minimize the crossow
around the n tip that decreases lift and increases drag. As
a result, lifting ns tend to be diamond-shaped, with a high
aspect ratio and tapered at both ends, while their base usually
forms a small angle relative to the main axis [20].
Kinematics obtained by Webb [98] for the pectoral n
propulsion of the seaperch have since been used to outline
lift-based labriform mode, although the movements performed
cannot be considered a pure apping; they are, however,
simpler and more tractable than those found in more recent
data (see [99] and [97]). Along with oscillating in a dorso-
ventral motion, the pectoral ns of the seaperch pass a wave
back over their length as a result of phase lags in the movement
of the individual n rays. The wavelength of this wave varies
with swimming speed, resulting in phase lags from about (for
velocities below 2 BL/s) to about (at higher velocities)
between the leading and trailing edges of the n.
Webb divided the n-beat cycle in the seaperch into ab-
duction [Fig. 15(a)], adduction [Fig. 15(b)], and refractory
[Fig. 15(c)] phases. The terminology has been since adopted,
although inconsistencies concerning the movements charac-
terizing each phase do appear in the literature. Generally,
during abduction, the n is moved away from the body
and downwards. It is then brought back to the body surface
(a) (b) (c)
Fig. 15. Dorsal view of the n movements in the lift-based labriform mode
for the seaperch. The diagrams show how the n trailing edge (thin line) lags
behind during (a) the abduction, (b) the adduction, and (c) the refraction of
the leading edge (thick line). The sh used had an average length of 14.3 cm
and swam at 1.2 BL/s. (Adapted from Webb [98].)
(adduction phase) and, during refraction, the n is orientated
to its original position by rotation of the leading edge. Due to
the subtlety of these movements, the angle of attack for the n
changes during each phase. As a result, the lift forces generated
have an elevation as well as a thrust component that causes the
body to move up and down during normal forward swimming.
Additionally, thrust forces will be generated discontinuously
because of the pattern of n-beat and refractory phases.
Between abduction and adduction, and during the refractory
phase, no lift-based thrust is generated. Webb estimated a
propulsive efciency between 0.6 and 0.65, for it has been
suggested in [99] that some small thrust could result from
a jet propulsion effect during refraction as water is being
displaced out of the decreasing space between the ns and
the body. If present, its effect should be minimal, and the sh
generally tends to accelerate during abduction and adduction
and to decelerate in between. The net result of these motions
is that the sh body moves relative to the ow in a gure-
eight motion, whose parameters change with speed, reecting
variations in the elevation and thrust components [98]. Recent
observations show that sh can smooth out their movements
by complimentary actions of the other ns.
A blade-element analysis of apping pectoral ns is given
by Blake in [84]. Again, a purely lift-based labriform mode
is considered and the n is assumed to consist of a series
of straight elements. The generalized applicability of blade-
element theory to labriform locomotion is questioned (see [92]
for a discussion), due to the curvatures and shape changes
observed for the pectoral ns when a combination of lift- and
drag-based methods is used, as is generally the case.
VI. SUMMARY
Having looked at some of the biomechanical aspects of
certain swimming modes employed by sh, one can only
marvel at the developed mechanisms and their signicance in
relation to the aquatic environment. It seems highly desirable
to successfully replicate them in articial devices. However,
Locomotion of Articulated Bodies in a Perfect Fluid
Eva Kanso,
1
Jerrold E. Marsden,
1
Clarence W. Rowley
2
and Juan Melli-Huber
2
1
Control and Dynamical Systems
2
Mechanical and Aerospace Engineering
California Institute of Technology Princeton University
Mail stop: CDS107-81 D232 Engineering-quad
Pasadena, CA 91125 Princeton, NJ 08544
Revision: March 7, 2005
Abstract
This paper is concerned with modeling the dynamics of N articulated solid
bodies submerged in an ideal uid. The model is used to analyze the locomotion
of aquatic animals due to the coupling between their shape changes and the uid
dynamics in their environment.
The equations of motion are obtained by making use of a two-stage reduction
process which leads to signicant mathematical and computational simplici-
ations. The rst reduction exploits particle relabelling symmetry: that is, the
symmetry associated with the conservation of circulation for ideal, incompress-
ible uids. As a result, the equations of motion for the submerged solid bodies
can be formulated without explicitly incorporating the uid variables. This
reduction by the uid variables is a key dierence with earlier methods and it
is appropriate since one is mainly interested in the location of the bodies, not
the uid particles.
The second reduction is associated with the invariance of the dynamics
under superimposed rigid motions. This invariance corresponds to the conser-
vation of total momentum of the solid-uid system. Due to this symmetry,
the net locomotion of the solid system is realized as the sum of geometric and
dynamic phases over the shape space consisting of allowable relative motions,
or deformations, of the solids. In particular, reconstruction equations that gov-
ern the net locomotion at zero momentum, that is, the geometric phases, are
obtained.
As an illustrative example, a planar three-link mechanism is shown to propel
and steer itself at zero momentum by periodically changing its shape. Two so-
lutions are presented: one corresponds to a hydrodynamically decoupled mech-
anism and one is based on accurately computing the added inertias using a
boundary element method. The hydrodynamically decoupled model produces
smaller net motion than the more accurate model, indicating that it is impor-
tant to consider the hydrodynamic interaction of the links.
1
1 Introduction 4
Beal, Lauder and Triantafyllou [2003a],and Liao, Beal, Lauder and Triantafyllou
[2003b]. Therefore, understanding how sh behave in the presence of vortices is
essential in studying aquatic propulsion and stability. For this reason, we view the
present study, that is, swimming in a potential ow with zero circulation, as the
rst step in the generation of a family of rigorous models that will eventually treat
the interaction of the sh with self-generated vortices as well as with vortices shed
by other objects, see Figure 1.1. We expect that recent models of bodies interacting
dynamically with vortices, such as Shashikanth, Marsden, Burdick, and Kelly [2002]
and Borisov and Mamaev [2003] will be useful in our future studies. Such theoretical
and computational models will compliment the recent experimental and empirical
studies on the wake dynamics and its interaction with the shape kinematics of bio-
logical sh, as done in Tytell and Lauder [2004] and Tytell [2004] in the case of an
eel swimming.
(a)
(b)
(c)
Figure 1.1: (a) Fish swimming in an innite potential ow. The body of the sh is modeled as
a system of articulated solid bodies; (b) Vortices shed by a swimming sh, the brighter vortices
being the most recent. The vortices form a thrust wake, that is, a wake that resembles drag wakes
generated behind stationary obstacles but with opposite direction of rotation; (c) Fish swimming
in a drag wake generated behind a solid obstacle. The undulating sh weaves through the vortex
street as if it were generating a thrust wake.
Particle Relabeling Symmetry of the Fluid. A reduced formulation of the
dynamics of the solid bodies that does not explicitly incorporate the ambient uid
is derived. The underlying assumption is that the motion of the solids does not
generate circulation. In this case, Kelvins circulation theorem states that the circu-
lation remains constant around any closed material contour moving with the uid;
see, for example, Chorin and Marsden [1979] for proofs and discussion. Physically,
Figure 8: Snapshot of swimmer and vortex wake. A portion of the wake of
the swimmer is shown for the controlled case at t/t
= 44.5. The red and

blue circles represent clockwise and counterclockwise vorticity, respectively.
The larger, darker circles correspond to stronger strength vortices while the
smaller, lighter circles are weaker strength vortices.
[4] de Araujo, G., and Koiller, J. Self-Propulsion of N-Hinged An-
imats at Low Reynolds Number. Qualitative Theory of Dynamical
Systems 1 (2003), 128.
[5] Drucker, E. G., and Lauder, G. V. Locomotor function of the dor-
sal n in teleost shes: experimental analysis of wake forces in sunsh.
J. Exp. Biol. 204 (2001), 29432958.
[6] Fish, F. Performance constraints on the maneuverability of exible
and rigid biological systems. Proceedings of the Eleventh International
Symposium on Unmanned Untethered Submersible Technology (1999),
394406.
[7] Kanso, E., Marsden, J. E., Rowley, C. W., and Melli-Huber,
J. B. Locomotion of articulated bodies in a perfect uid. J. Nonlin.
Sci. (2005).
[8] Kelly, S. D. The Mechanics and Control of Robotic Locomotion with
Applications to Aquatic Vehicles. PhD thesis, California Institute of
Technology, Pasadena, California, May 1998.
[9] Maslov, N. Maneuverability and controllability of dolphins. Bionika
4 (1970), 4650.
[10] Mason, R., and Burdick, J. W. Propulsion and control of de-
formable bodies in an ideal uid. In IEEE International Conference on
Robotics and Automation (1999).
11
Models and control of sh-like locomotion
Juan Melli and Clarence W. Rowley
Department of Mechanical and Aerospace Engineering
Princeton University, Princeton, New Jersey
March 14, 2010
Abstract
Inspired by the advanced capabilities of sh-like swimmers, we seek
a greater understanding of the hydrodynamic mechanisms they em-
ploy. Borrowing tools from geometric mechanics, we develop a large-
amplitude gait design technique for self-propelling deformable swim-
mers in potential ow and Stokes ow. Additionally, a heuristically-
developed control algorithm is implemented to nearly optimize thrust-
generation for a swimmer moving through its own nearly-periodic vor-
tex wake.
1 Motivation
The advanced capabilities of biological swimmers, including eciency, ma-
neuverability, and stealth, have attracted increased attention to the study of
underwater swimming in recent years. For example, whereas sea lions have
minimum turning radii of 9% of their body length [6], a traditional sub-
marines turning radius may be two to three body lengths or more [9]. An
improved understanding of the hydrodynamic mechanisms and propulsion
techniques employed by sh and other swimmers may lead to the design of
more advanced man-made underwater vehicles.
Here we review the results from three simplied swimmer models used
to study and better understand the locomotion of deformable bodies in a
uid. For the rst two cases we consider articulated swimmers in both
inviscid and highly viscous uids. Although these two scenarios (modeled
by potential ow and Stokes ow, respectively) are opposites in terms of
the relative importance of inertial to frictional forces, they exhibit similar
dynamical properties (they are both perfectly time-reversible systems) and
1
2436
INTRODUCTION
Two distinct types of hydromedusan propulsion are well known
(Colin and Costello, 2002). Prolate species such as Sarsia tubulosa
primarily use a jetting type of propulsion with large jet velocities
immediately behind the velar aperture. Their swimming is
characterized by quick accelerations during the contraction phase
of swimming followed by periods of gliding with relatively small
accelerations. However, the once widely accepted jetting model fails
to explain the swimming patterns seen in prolate species such as
Aequorea victoria. These hydromedusae use a paddling or rowing
motion to swim and produce more diffuse vortices shed from the
bell margins during the contraction phase.
Prolate, jetting hydromedusae retract their tentacles during
swimming and feed by extending their tentacles while drifting
(Madin, 1988). Swimming is used to escape predators or to ambush
prey. Swimming and feeding are disparate activities since extending
the tentacles during swimming could greatly decrease swimming
performance. Conversely, oblate, paddling hydromedusae leave their
tentacles extended during swimming and the vortices produced
during swimming travel through the extended tentacles (Colin and
Costello, 2002; Colin et al., 2003; Costello, 1992; Costello and Colin,
1994; Ford and Costello, 1997). Prey in the fluid near the bell region
of paddling hydromedusae have even been observed to be carried
into contact with the tentacles by the vortices formed during
swimming (Costello and Colin, 1995). For these reasons, swimming
complements feeding in paddling hydromedusae by helping to draw
prey into the tentacles.
It has been shown that rowing propulsion is a necessary adaptation
for larger hydromedusae due to morphological constraints and
energy efficiency. For jetting propulsion, the force necessary for
propulsion increases faster with size than the available muscle force
to provide the jetting motion (Dabiri et al., 2007). Oblate
hydromedusae make up for this by employing paddling type
propulsion. Models show that the production of stopping vortices
during the relaxation phase of paddling type propulsion allows large
hydromedusae to swim effectively despite their morphological
constraints (Dabiri et al., 2007). Specifically, the stopping vortex
partially cancels the starting vortex, reducing the induced drag on
the oblate hydromedusa and increasing swimming efficiency. The
smaller, prolate species have lower drag due to their shape and
further decrease drag by retracting their tentacles while swimming
(Colin et al., 2003). These factors, combined with more rapid bell
contractions, make jetting hydromedusae much more proficient
swimmers than their oblate relatives (Daniel, 1983).
We are interested in hydromedusan propulsion as a basis for the
design of new propulsion technologies for underwater vehicles.
Recently, jet and vortex propulsion have become a focus in the areas
of underwater maneuvering and locomotion of bio-engineered
vehicles. A vortex thruster loosely mimicking hydromedusa
propulsion was proposed by Mohseni (Mohseni, 2004; Mohseni,
2006). The current generation of these thrusters and their
implementation on an underwater vehicle are discussed by Krieg
and Mohseni (Krieg and Mohseni, 2008) and are capable of
doi:10.1242/jeb.026740
Flow structures and fluid transport for the hydromedusae Sarsia tubulosa and
Aequorea victoria
Doug Lipinski and Kamran Mohseni*
College of Engineering and Applied Science, University of Colorado at Boulder, Boulder, CO 80309, USA
*Author for correspondence (e-mail: mohseni@colorado.edu)
Accepted 8 May 2009 (submitted 27 October 2008)
SUMMARY
The flow structures produced by the hydromedusae Sarsia tubulosa and Aequorea victoria are examined using direct numerical
simulation and Lagrangian coherent structures (LCS). Body motion of each hydromedusa is digitized and input to a CFD program.
Sarsia tubulosa uses a jetting type of propulsion, emitting a single, strong, fast-moving vortex ring during each swimming cycle
while a secondary vortex of opposite rotation remains trapped within the subumbrellar region. The ejected vortex is highly
energetic and moves away from the hydromedusa very rapidly. Conversely, A. victoria, a paddling type hydromedusa, is found to
draw fluid from the upper bell surface and eject this fluid in pairs of counter-rotating, slow-moving vortices near the bell margins.
Unlike S. tubulosa, both vortices are ejected during the swimming cycle of A. victoria and linger in the tentacle region. In fact, we
find that A. victoria and S. tubulosa swim with Strouhal numbers of 1.1 and 0.1, respectively. This means that vortices produced
by A. victoria remain in the tentacle region roughly 10 times as long as those produced by S. tubulosa, which presents an
excellent feeding opportunity during swimming for A. victoria. Finally, we examine the pressure on the interior bell surface of both
hydromedusae and the velocity profile in the wake. We find that S. tubulosa produces very uniform pressure on the interior of the
bell as well as a very uniform jet velocity across the velar opening. This type of swimming can be well approximated by a slug
model, but A. victoria creates more complicated pressure and velocity profiles. We are also able to estimate the power output of
S. tubulosa and find good agreement with other hydromedusan power outputs. All results are based on numerical simulations of
the swimming jellyfish.
Supplementary material available online at http://jeb.biologists.org/cgi/content/full/212/15/2436/DC1
Key words: hydromedusae, feeding, jet propulsion, rowing propulsion, Lagrangian coherent structures, LCS.
2438
Shadden et al. (Shadden et al., 2005) and we provide here a brief
overview for those unfamiliar with this concept.
It is simplest to think of LCS as a post-processing technique
to reveal coherent structures in a given flow. In our case, the flow
is determined by numerical simulations of the swimming
hydromedusae. LCS are based on the finite time Lyapunov
exponent (FTLE), which is analogous to the standard Lyapunov
exponent of classical dynamical systems theory. The FTLE is
defined as:
where t
0
is the time being considered, T is the integration time,
which will be further explained later, x is the position vector, and
max
() is the maximum eigenvalue of the finite time deformation
tensor, . In practice, the domain of interest is seeded with passive
tracer particles, which are then advected from time t
0
to time t
0
+T
using the known velocity field. The resulting particle positions are
then used to compute and then the FTLE field. The flow
considered here is a two-dimensional, axisymmetric flow, and
was calculated appropriately to take this into account. In swirl-free
axisymmetric coordinates, (r, , x), becomes:
where r
i
, r
f
, x
i
and x
f
are the initial and final radial and axial
coordinates of a particle in the flow, respectively.
The resulting FTLE field depends on the integration time, T, in
that larger values of T reveal more structures than smaller values
of T. Therefore, T may be chosen to reveal the desired level of detail
without worrying about influencing the major structures that are
revealed. Additionally, T may be positive or negative, representing
forward and backward particle advection, respectively. Therefore,
there are two types of FTLE fields: forward time and backward
time.
Once the FTLE field has been calculated, LCS are defined as
ridges in the FTLE field, following Shadden et al. (Shadden et
al., 2005). In practice, LCS are usually visualized by looking at
contour plots of the FTLE field. Conceptually, ridges in the
forward FTLE field, called forward LCS, represent lines where
particles diverge most quickly, and backward LCS represent lines
where particles converge. For this region, dye visualization
experiments reveal structures very similar to backward LCS.
Additionally, for LCS which are sufficiently strong, the flux
across the LCS is negligible, a property that makes LCS extremely
useful for analyzing transport in flows. Finally, forward LCS are
analogous to the stable manifolds of a dynamical system and act
as repelling material lines while backward LCS are analogous to
the unstable manifolds of a dynamical system and act as attracting
material lines. The interaction of these LCS largely govern
transport in a flow, and their intersections can be used to exactly
define a vortex without the use of arbitrary thresholds of vorticity
(Shadden et al., 2006).
d
dx
=
r
f
r
i
0
r
f
x
i
0
rf
r
i
0
x
f
r
i
0
x
f
x
i
, (3)
=
d
dx
*
d
dx
, (2)

t
0
T
x
( )
=
1
T
ln
max

( )
, (1)
D. Lipinski and K. Mohseni
RESULTS
Sarsia tubulosa
Sarsia tubulosa employs a jetting type of propulsion. The swimming
cycle consists of three phases: (1) a rapid contraction, (2) relaxation
and (3) a coasting phase. Each cycle results in the expulsion of a
strong vortex along the axis of symmetry, which rapidly propels
the jellyfish forward as seen in the LCS shown in Fig. 1 (also see
Movie1 in supplementary material).
The jetting nature of S. tubulosas propulsion, forming a single
vortex ring with each swimming pulse, is clearly reflected in the
LCS seen in Fig. 1. During the contraction phase, a vortex which
draws fluid in the positive x-direction along the axis of symmetry
is ejected from the hydromedusa. This will be referred to as the
starting vortex. One of the most striking details of this figure is the
presence of very complex flow structures within the subumbrellar
5 4 3 2 1 0 1
Time = 2.76 s
D
1
0
0.5
0.5
1
4 3 2 1 0 1 2
Time = 2.51 s
C
1
0
0.5
0.5
1
x
y
3 3 2 1 0 1 2
Time = 2.26 s
B
1
0
0.5
0.5
1
3 3 2 1 0 1 2
Time = 2.01 s
A
1
0
0.5
0.5
1
Fig. 1. Lagrangian coherent structures (LCS) for one swimming cycle of
Sarsia tubulosa. The forward LCS are shown in blue and the backward
LCS are red. The video version is available in the supplementary material
as Movie1.
2436
INTRODUCTION
Two distinct types of hydromedusan propulsion are well known
(Colin and Costello, 2002). Prolate species such as Sarsia tubulosa
primarily use a jetting type of propulsion with large jet velocities
immediately behind the velar aperture. Their swimming is
characterized by quick accelerations during the contraction phase
of swimming followed by periods of gliding with relatively small
accelerations. However, the once widely accepted jetting model fails
to explain the swimming patterns seen in prolate species such as
Aequorea victoria. These hydromedusae use a paddling or rowing
motion to swim and produce more diffuse vortices shed from the
bell margins during the contraction phase.
Prolate, jetting hydromedusae retract their tentacles during
swimming and feed by extending their tentacles while drifting
(Madin, 1988). Swimming is used to escape predators or to ambush
prey. Swimming and feeding are disparate activities since extending
the tentacles during swimming could greatly decrease swimming
performance. Conversely, oblate, paddling hydromedusae leave their
tentacles extended during swimming and the vortices produced
during swimming travel through the extended tentacles (Colin and
Costello, 2002; Colin et al., 2003; Costello, 1992; Costello and Colin,
1994; Ford and Costello, 1997). Prey in the fluid near the bell region
of paddling hydromedusae have even been observed to be carried
into contact with the tentacles by the vortices formed during
swimming (Costello and Colin, 1995). For these reasons, swimming
complements feeding in paddling hydromedusae by helping to draw
prey into the tentacles.
It has been shown that rowing propulsion is a necessary adaptation
for larger hydromedusae due to morphological constraints and
energy efficiency. For jetting propulsion, the force necessary for
propulsion increases faster with size than the available muscle force
to provide the jetting motion (Dabiri et al., 2007). Oblate
hydromedusae make up for this by employing paddling type
propulsion. Models show that the production of stopping vortices
during the relaxation phase of paddling type propulsion allows large
hydromedusae to swim effectively despite their morphological
constraints (Dabiri et al., 2007). Specifically, the stopping vortex
partially cancels the starting vortex, reducing the induced drag on
the oblate hydromedusa and increasing swimming efficiency. The
smaller, prolate species have lower drag due to their shape and
further decrease drag by retracting their tentacles while swimming
(Colin et al., 2003). These factors, combined with more rapid bell
contractions, make jetting hydromedusae much more proficient
swimmers than their oblate relatives (Daniel, 1983).
We are interested in hydromedusan propulsion as a basis for the
design of new propulsion technologies for underwater vehicles.
Recently, jet and vortex propulsion have become a focus in the areas
of underwater maneuvering and locomotion of bio-engineered
vehicles. A vortex thruster loosely mimicking hydromedusa
propulsion was proposed by Mohseni (Mohseni, 2004; Mohseni,
2006). The current generation of these thrusters and their
implementation on an underwater vehicle are discussed by Krieg
and Mohseni (Krieg and Mohseni, 2008) and are capable of
doi:10.1242/jeb.026740
Flow structures and fluid transport for the hydromedusae Sarsia tubulosa and
Aequorea victoria
Doug Lipinski and Kamran Mohseni*
College of Engineering and Applied Science, University of Colorado at Boulder, Boulder, CO 80309, USA
*Author for correspondence (e-mail: mohseni@colorado.edu)
Accepted 8 May 2009 (submitted 27 October 2008)
SUMMARY
The flow structures produced by the hydromedusae Sarsia tubulosa and Aequorea victoria are examined using direct numerical
simulation and Lagrangian coherent structures (LCS). Body motion of each hydromedusa is digitized and input to a CFD program.
Sarsia tubulosa uses a jetting type of propulsion, emitting a single, strong, fast-moving vortex ring during each swimming cycle
while a secondary vortex of opposite rotation remains trapped within the subumbrellar region. The ejected vortex is highly
energetic and moves away from the hydromedusa very rapidly. Conversely, A. victoria, a paddling type hydromedusa, is found to
draw fluid from the upper bell surface and eject this fluid in pairs of counter-rotating, slow-moving vortices near the bell margins.
Unlike S. tubulosa, both vortices are ejected during the swimming cycle of A. victoria and linger in the tentacle region. In fact, we
find that A. victoria and S. tubulosa swim with Strouhal numbers of 1.1 and 0.1, respectively. This means that vortices produced
by A. victoria remain in the tentacle region roughly 10 times as long as those produced by S. tubulosa, which presents an
excellent feeding opportunity during swimming for A. victoria. Finally, we examine the pressure on the interior bell surface of both
hydromedusae and the velocity profile in the wake. We find that S. tubulosa produces very uniform pressure on the interior of the
bell as well as a very uniform jet velocity across the velar opening. This type of swimming can be well approximated by a slug
model, but A. victoria creates more complicated pressure and velocity profiles. We are also able to estimate the power output of
S. tubulosa and find good agreement with other hydromedusan power outputs. All results are based on numerical simulations of
the swimming jellyfish.
Supplementary material available online at http://jeb.biologists.org/cgi/content/full/212/15/2436/DC1
Key words: hydromedusae, feeding, jet propulsion, rowing propulsion, Lagrangian coherent structures, LCS.
Human Reproduction vol.12 no.5 pp.10061012, 1997
Quantitative observations of agellar motility of
capacitating human spermatozoa
Sharon T.Mortimer
1,4
, Damien Schoevaert
3
, procedures (Schrader et al., 1991; ESHRE Andrology Special
Interest Group, 1996). Sperm kinematic analyses are carried M.Anne Swan
1
and David Mortimer
2
out routinely using computer-aided sperm analysis (CASA)
1
Department of Anatomy and Histology and Institute for
instruments which monitor the movement of the sperm head
Biomedical Research, University of Sydney, Sydney, NSW 2006,
and then analyse different aspects of its reconstructed trajectory.
2
Sydney IVF, Sydney, NSW, Australia and
3
Labo dAnalyse
dImage en Pathologie Cellulaire, Centre Hayem, Hopital St Louis, These analyses are typically performed at image sampling
Paris, France
frequencies of 2560 Hz, depending on the video system used
(Mortimer, 1994).
4
To whom correspondence should be addressed at: Mortimer
Scientic Consulting, 227 Quarter Sessions Road, Westleigh,
The aspects of sperm movement analysed by CASA include
NSW 2120, Australia
the velocity of motion of the sperm head, such as curvilinear
velocity (VCL), straight-line velocity (VSL) and average path
For technical reasons sperm head movement is assessed
velocity (VAP), as well as the amplitude of lateral head
in kinematic analysis, while agellar movement is the
displacement about the direction of travel (ALHmean and/or
determining factor of head movement, not vice versa. It
ALHmax) (Boyers et al., 1989). These kinematic parameters
follows then that the development of new kinematic values
were derived from the study of sperm movement in seminal
to describe the movement of capacitating human spermato-
plasma and resulted from detailed analyses of both head and
zoa should include the analysis of their agellar movement.
agellar movement (David et al., 1981; Serres et al., 1984).
The aim of this study was to establish quantitative differ-
The centroid-based values were found to relate closely to
ences between agellar movement patterns of hyperactiv-
the agellar movement patterns. So while centroid-derived
ated and non-hyperactivated spermatozoa which could then
kinematic values are used in the denition of different sperm
be used in the evaluation of new centroid-based kinematic
movement patterns, it is understood that these values are
values. Spermatozoa were prepared by swim-up from
secondary indicators of aspects of agellar movement when
semen into culture medium supplemented with 30 mg/ml
the spermatozoa are analysed in seminal plasma.
human serum albumin. Sperm movement was recorded in
The relationship between agellar movement and head
50 m-deep chambers using a 200 Hz video system. Sperm
movement may be more complex for human spermatozoa in
movement was classied based on agellar movement, with
capacitating medium, although this has not been studied in
24 non-hyperactivated and 26 hyperactivated spermatozoa
detail. For example, it has been shown that human spermatozoa,
included in the study. Flagellar analysis was performed
when incubated under capacitating conditions, exhibit different
using both a semi-automated analysis system(SIAMFLAG;
motility patterns such as hyperactivated [both progressive
30 images at 200 Hz) and manual methods (100 Hz).
(transitional) and non-progressive (star-spin or thrashing)
Hyperactivated spermatozoa had signicantly larger
head movement patterns], freeze-ex and forward progress-
agellar beat angles (87) and signicantly lower agellar
ive (Burkman, 1984; Robertson et al., 1988), and that they
beat frequencies (29.4 Hz) than non-hyperactivated
are able to switch between these motility patterns (Mortimer
human spermatozoa. In addition, the agellar wave ampli-
and Swan, 1995a). Despite obvious differences in head and
tude was signicantly greater and the bend diameter
agellar movement patterns between the fairly regular seminal
signicantly smaller for hyperactivated spermatozoa in the
spermatozoa and the more erratic capacitating spermatozoa,
proximal region of the agellum (up to 20 m from the
use of the original kinematic values has persisted in the
headmidpiece junction). The velocity of the hyperactivated
description of hyperactivated motility. This has contributed to
wave was low in this region, although it was signicantly
some confusion over track classication, with the result that
slower than the non-hyperactivated wave in all regions of
investigators may not always be certain that a trajectory
the sperm tail.
classied in one laboratory as hyperactivated would have the
Key words: agellar/human/hyperactivation/kinematics/sperm-
same classication in a different laboratory.
atozoa
One way to overcome this confusion would be to derive a
new set of kinematic values for use with capacitating human
spermatozoa. Because sperm head movement is dictated by
Introduction
agellar movement patterns, this would require an understand-
ing of the relationship between the new centroid kinematic The analysis of sperm movement is an integral part of
sperm function testing in infertility diagnosis, as well as in values and agellar movement parameters. The rst step in
this process is to investigate quantitative differences in agellar reproductive toxicology investigations and in quality control
1006 European Society for Human Reproduction and Embryology
S.T.Mortimer et al.
Figure 1. Calculation of agellar beat angle. Consecutive 200 Hz
images were combined using the longitudinal axis of the sperm
head as the point of reference. Print-outs of the agellar envelopes
were made and lines drawn from the neck to the extremes of the
agellar envelope. The angle made by these two lines about the
neck was measured using a protractor.
interest was identied, the tape was advanced to the same starting
point as for the semi-automated analysis. A piece of tape with a cross
Figure 2. Manual agellar reconstruction. Flagellar traces of the drawn on it was attached to the monitors screen close to the
same spermatozoa used in the agellar beat angle analysis were
spermatozoon of interest, but not so close that it covered any of the
made manually (100 Hz). The distance of the peak along the
cell during the analysis sequence. The head outline and tail were
agellum, the bend diameter and the amplitude were determined as
traced onto overhead projector lm sheets using ne-tipped marker
shown.
pens. The cross was traced with every image, giving the position and
orientation of the spermatozoon with respect to it.
Because of the limitations of resolution using video tape, it was
transparency over the agellar trace and nding the circle that best
not always possible to identify the midpiece of a spermatozoon. So
tted the agellar bend. The diameter was then converted to m by
the tracing was only of the head and tail, and analyses were started
dividing it by the magnication correction factor (nal magnication/
at the apparent junction between the head and tail of the spermatozoon
100) for the particular agellar trace (Figure 2).
(i.e. the neck). In addition, it was not possible to differentiate principal
A large-diameter agellar wave was one with a low degree of
and reverse bends using these traces because the image was not
agellar deformation, while a small-diameter agellar wave indicated
always clear enough to differentiate head rotation and there were no
a high degree of agellar deformation, i.e. the agellum had a more
obvious cellular points of reference.
acute bend.
The agellar traces were photocopied using a machine known not
Calculation of the amplitude of agellar waves to distort the copied image, and the copies were compared with the
The amplitude of a wave is the perpendicular distance between the originals and found to be accurate. Manual agellar analysis was
point of inexion and the peak. The point of inexion of the agellar made using the photocopied images.
waves could not be determined accurately, so the total perpendicular
Calculation of the peak of the agellar wave
distance between a consecutive peak and trough was found, and the
The peak of a wave was determined by the geometric bisector method.
distance halved to give the amplitude (in mm). The amplitude was
First, the tangents to the wave were drawn. Then, using a drawing
converted to m by dividing it by the magnication correction factor
compass with the point placed on the point of intersection of the
(nal magnication/100) for the particular agellar trace (Figure 2).
tangents, arcs were drawn onto the tangents. With the drawing
compass set to the same distance, its point was placed on each Calculation of agellar beat frequency (FBF)
intersection of the tangent and arc and another arc drawn below the The FBF was estimated for each spermatozoon by determining the
wave peak. The intersection of these arcs and the intersection of the number of waves initiated per second. The agellar traces were
tangents were joined to give the bisector of the wave (David et al., examined and the initiation of a wave was dened as the rst image
1981). The distance from the neck of the spermatozoon to the bisector showing a agellar bend which was subsequently propagated. The
was measured in millimetres by tracing the agellumwith a curvimeter number of initiations was counted, and the frequency converted to
(Model No. 54M; Burnat, Paris, France). The millimetre distance Hz (cycles/s).
was converted to microns by dividing by the magnication correction
Statistics
factor (nal magnication/100) (Figure 2).
Receiver Operating Characteristic (ROC) curve analyses were used
Calculation of the velocity of propagation of the agellar wave
to determine hyperactivation thresholds. When the values were
The peak of the wave was followed along the agellum in consecutive
normally distributed, comparisons were made using unpaired t-tests,
images and the velocity determined by the change in distance (
and when they were not normally distributed, comparisons were made
distance) over time (since 0.01 s elapsed between successive images).
using unpaired Wilcoxon analyses. All statistical analyses were
performed using MedCalc (MedCalc Software, Mariakerke, Belgium).
Calculation of the diameter of the agellar wave
A series of circles of diameters from 10 to 50 mm (increasing in
1 mm increments) was drawn using CorelDRAW! (Version 3.0;
Results
Corel Corporation, Ottawa, Ontario, Canada) and printed onto a
single sheet of paper. The circles were photocopied onto overhead
Trajectory classication
transparency lm, and the accuracy of the copy checked by placing
Centroid analysis of the tracks conrmed the subjective classi-
the transparency over the original. The accuracy of the diameters was
cation of the spermatozoa as hyperactivated or non-hyperacti-
checked using a mm rule.
The diameter of the agellar waves was estimated by placing the vated. The objective classication of these tracks meant that
1008
S.T.Mortimer et al.
Figure 1. Calculation of agellar beat angle. Consecutive 200 Hz
images were combined using the longitudinal axis of the sperm
head as the point of reference. Print-outs of the agellar envelopes
were made and lines drawn from the neck to the extremes of the
agellar envelope. The angle made by these two lines about the
neck was measured using a protractor.
interest was identied, the tape was advanced to the same starting
point as for the semi-automated analysis. A piece of tape with a cross
Figure 2. Manual agellar reconstruction. Flagellar traces of the drawn on it was attached to the monitors screen close to the
same spermatozoa used in the agellar beat angle analysis were
spermatozoon of interest, but not so close that it covered any of the
made manually (100 Hz). The distance of the peak along the
cell during the analysis sequence. The head outline and tail were
agellum, the bend diameter and the amplitude were determined as
traced onto overhead projector lm sheets using ne-tipped marker
shown.
pens. The cross was traced with every image, giving the position and
orientation of the spermatozoon with respect to it.
Because of the limitations of resolution using video tape, it was
transparency over the agellar trace and nding the circle that best
not always possible to identify the midpiece of a spermatozoon. So
tted the agellar bend. The diameter was then converted to m by
the tracing was only of the head and tail, and analyses were started
dividing it by the magnication correction factor (nal magnication/
at the apparent junction between the head and tail of the spermatozoon
100) for the particular agellar trace (Figure 2).
(i.e. the neck). In addition, it was not possible to differentiate principal
A large-diameter agellar wave was one with a low degree of
and reverse bends using these traces because the image was not
agellar deformation, while a small-diameter agellar wave indicated
always clear enough to differentiate head rotation and there were no
a high degree of agellar deformation, i.e. the agellum had a more
obvious cellular points of reference.
acute bend.
The agellar traces were photocopied using a machine known not
Calculation of the amplitude of agellar waves to distort the copied image, and the copies were compared with the
The amplitude of a wave is the perpendicular distance between the originals and found to be accurate. Manual agellar analysis was
point of inexion and the peak. The point of inexion of the agellar made using the photocopied images.
waves could not be determined accurately, so the total perpendicular
Calculation of the peak of the agellar wave
distance between a consecutive peak and trough was found, and the
The peak of a wave was determined by the geometric bisector method.
distance halved to give the amplitude (in mm). The amplitude was
First, the tangents to the wave were drawn. Then, using a drawing
converted to m by dividing it by the magnication correction factor
compass with the point placed on the point of intersection of the
(nal magnication/100) for the particular agellar trace (Figure 2).
tangents, arcs were drawn onto the tangents. With the drawing
compass set to the same distance, its point was placed on each Calculation of agellar beat frequency (FBF)
intersection of the tangent and arc and another arc drawn below the The FBF was estimated for each spermatozoon by determining the
wave peak. The intersection of these arcs and the intersection of the number of waves initiated per second. The agellar traces were
tangents were joined to give the bisector of the wave (David et al., examined and the initiation of a wave was dened as the rst image
1981). The distance from the neck of the spermatozoon to the bisector showing a agellar bend which was subsequently propagated. The
was measured in millimetres by tracing the agellumwith a curvimeter number of initiations was counted, and the frequency converted to
(Model No. 54M; Burnat, Paris, France). The millimetre distance Hz (cycles/s).
was converted to microns by dividing by the magnication correction
Statistics
factor (nal magnication/100) (Figure 2).
Receiver Operating Characteristic (ROC) curve analyses were used
Calculation of the velocity of propagation of the agellar wave to determine hyperactivation thresholds. When the values were
The peak of the wave was followed along the agellum in consecutive normally distributed, comparisons were made using unpaired t-tests,
images and the velocity determined by the change in distance ( and when they were not normally distributed, comparisons were made
distance) over time (since 0.01 s elapsed between successive images). using unpaired Wilcoxon analyses. All statistical analyses were
performed using MedCalc (MedCalc Software, Mariakerke, Belgium).
Calculation of the diameter of the agellar wave
A series of circles of diameters from 10 to 50 mm (increasing in
1 mm increments) was drawn using CorelDRAW! (Version 3.0;
Results Corel Corporation, Ottawa, Ontario, Canada) and printed onto a
single sheet of paper. The circles were photocopied onto overhead
Trajectory classication
transparency lm, and the accuracy of the copy checked by placing
Centroid analysis of the tracks conrmed the subjective classi-
the transparency over the original. The accuracy of the diameters was
cation of the spermatozoa as hyperactivated or non-hyperacti- checked using a mm rule.
The diameter of the agellar waves was estimated by placing the vated. The objective classication of these tracks meant that
1008
Robot Fish
http://www.robaid.com/bionics/sh-robots-search-for-pollution-in-the-waters.htm
Robot Fish
Respirao
Respirao
Respirao
Respirao
Respirao
Respirao
Comparando peixes de gua doce e de gua
salgada, qual dos dois deve possuir maior rea
de troca gasosa?
Compare a rea de troca gasosa/kg de um
humano e de um peixe. Explique a diferena.

Swimming

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Swimming

Hochgeladen von

Copyright:

Verfügbare Formate

Swimming

FROM DRAG-BASED TO LIFT-BASED PROPULSION

damping constant to be 2Nsm

where i, j and k are unit vectors in the x, y and z directions,

= 44.5. The red and

Das könnte Ihnen auch gefallen