PHYSICS OF FLUIDS 21, 051901 共2009兲

Wave propagation and induced steady streaming in viscous fluid

contained in a prestressed viscoelastic tube
Ye Ma and Chiu-On Nga兲
Department of Mechanical Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong
共Received 15 October 2008; accepted 29 April 2009; published online 22 May 2009兲
The oscillatory and time-mean motions induced by a propagating wave of small amplitude through
a viscous incompressible fluid contained in a prestressed and viscoelastic 共modeled as a Voigt
material兲 tube are studied by a perturbation analysis based on equations of motion in the Lagrangian
system. The classical problem of oscillatory viscous flow in a flexible tube is re-examined in the
contexts of blood flow in arteries or pulmonary gas flow in airways. The wave kinematics and
dynamics, including wavenumber, wave attenuation, velocity, and stress fields, are found as
analytical functions of the wall and fluid properties, prestress, and the Womersley number for the
cases of a free or tethered tube. On extending the analysis to the second order in terms of the small
wave steepness, it is shown that the time-mean motion of the viscoelastic tube with sufficient
strength is short lived and dies out quickly as a limit of finite deformation is approached. Once the
tube has attained its steady deformation, the steady streaming in the fluid can be solved analytically.
Results are generated to illustrate the combined effects on the first-order oscillatory flow and the
second-order steady streaming due to elasticity, viscosity, and initial stresses of the wall. The present
model as applied to blood flow in arteries and gas flow in pulmonary airways during high-frequency
ventilation is examined in detail through comparison with models in the literature.
© 2009 American Institute of Physics. 关DOI: 10.1063/1.3139250兴

I. INTRODUCTION cardiac functions.2 HFV, which mimics a panting mode, i.e.,

The problem of oscillatory flow and mass transport
through a viscous fluid contained in an initially stressed flex-
ible tube has long been, and continues to be, of interest to
investigators who are concerned with, inter alia, biofluid me-
chanics. In cardiovascular fluid flows, an important consid-
eration is the shear stress. Blood is a living fluid, and if the
forces applied to the fluid are excessively large, the resulting
shear stress can cause harm such as in vivo hemolysis,1 a
condition corresponding to the destruction or dissolution of
red blood cells with subsequent release of hemoglobin into
the surrounding fluid. Also, there are studies indicating that
fluid shear stress may play a direct role in the initiation and
localization of atherosclerosis and influences other aspects of
the disease process as well. Arterial damage consequent to
atherosclerosis is responsible for the largest number of
deaths in the world. One major aspect of modeling the blood
flow is therefore to determine the shear stress distribution in
both the fluid and the blood vessel, so as to help the basic
understanding or diagnosis of problems in the system,
thereby guiding the treatment of the disease.
Gas transport in pulmonary airways during high-
frequency ventilation 共HFV兲 is another biophysical applica-
tion of oscillatory flow through a flexible tube. In many clini-
cal situations, the breathing of a patient is assisted by a
mechanical ventilator. Mechanical ventilation that mimics a
normal breathing mode would require large air pressures,
which may result in pulmonary barotraumas and impaired
a兲
Author to whom correspondence should be addressed. Electronic mail:
cong@hku.hk.
pumping a smaller tidal volume of air at a high frequency by
a smaller pressure gradient, is safer to use especially for a
vulnerable patient such as an infant. An understanding of the
flow and transport that are associated with the mechanical
ventilation may help guide the development of the technique.
For blood flow in large arteries3,4 or gas flow in the
upper airways,5 a tube of lumen radius of O共1兲 cm and lon-
gitudinal length of O共10兲 cm is normally considered. The
ratio of wall thickness to lumen radius is approximately 15%
in the case of blood vessels and 5% in the case of pulmonary
airways. As commonly found in the literature, a basic model
for pulsatile blood flow in arteries or oscillatory gas flow in
a pulmonary airway is to examine a viscous incompressible
fluid flowing in a straight, flexible, and initially stressed tube
as driven by a periodic pressure loading. Although omitting
some important effects such as pipe curvature and branching,
such a basic model has been extensively applied to the study
of many biofluid phenomena. Our work here is to look into
the propagation of waves and the induced steady streaming
in such a system, where the fluid-tube interaction is con-
trolled by the viscoelastic properties, initial stresses, and the
outer wall condition 共free or constrained兲 of the tube.
Since Womersley,6–8 who pioneered a linearized theory
of oscillatory flow in a straight, uniform, isotropic, thin-
walled, and elastic tube, the problem has been advanced
through numerous developments. In view of the nature of
human tissues, the model has been extended to tubes which
are thick walled,3,4,9,10 viscoelastic,4,11–13 orthotropic,14,15
tethered by the surrounding tissues,14,16,17 or with initial
stresses.18,19 Conventionally, a tube is modeled to be a thin-

1070-6631/2009/21共5兲/051901/25/$25.00 21, 051901-1 © 2009 American Institute of Physics

051901-2 Y. Ma and C.-O. Ng
walled structure 共like a shell or membrane兲 when its wall
thickness is supposed to be infinitesimally small compared
with the tube radius. This has been a classical model based
on some simplifying assumptions, e.g., a constant wall thick-
ness, uniform normal stress, and zero shear stress across the
wall under deformation. In contrast, a tube is so-called thick
walled when its wall thickness is comparable to, or a finite
fraction of, the tube radius 共specifically, when the fraction
ⱖ0.1兲.1 In fact, a thick-wall model does not necessarily im-
ply that the tube under consideration must have a thick wall.
The term is used only to mean that the model is not subject
to those simplifying conditions normally assumed in a thin-
wall model. In this more general model, the radial gradients
of the wall deformations and other effects are not ignored a
priori as in a thin-wall model. It is therefore a more compre-
hensive and exact model that is valid even down to the limit
of a thin wall.20 As was pointed out by Chandran et al.,1 for
many cardiovascular applications of interest, the wall
thickness-to-radius ratio of arteries will generally require one
to use a thick-wall model. Some recent examples are as fol-
lows. A three-dimensional thick-wall model was introduced
by Tang et al.,21 who studied steady flow and wall compres-
sion in stenotic arteries. A thick-walled flexible tube was
considered by Marzo et al.22 for three-dimensional collapse
and steady flow in physiological conduits. Holzapfel and
Gasser23 analyzed stress deformation for a coronary artery by
a two-layer, thick-walled tube model, and more recently Tsa-
mis and Stergiopulos24 proposed a thick-walled tube model
for arterial remodeling.
Based on the assumptions of small-amplitude waves
and/or low Reynolds number, most of the analytical works
mentioned above are essentially leading-order analyses, in
which the nonlinear fluid inertia is ignored. Steady
streaming,25 which is a steady current induced in oscillatory
flows when the velocity components are not ␲ / 2 out of
phase, is an important phenomenon in processes such as the
vascular disease26 and the gas transport during HFV.5 The
streaming owes its existence to the convection and viscous
diffusion of momentum, which gives rise to a net drift of
particles even though the forcing is purely oscillatory. To
find the streaming, one has to extend the analysis to the
second order in the wave steepness.
Ling and Atabek27 numerically solved the fully nonlinear
Navier–Stokes equations for the velocity profiles in arteries
modeled as an elastic tube based on the assumption of zero
longitudinal wall motion owing to the strong perivascular
tethering by the surrounding tissues. The same model was
later applied by Wang and Tarbell26 to finding streaming in
weakly nonlinear flows by the method of perturbation. The
elastic model of Atabek and Lew18 was followed by Dragon
and Grotberg,5 who studied oscillatory flow and mass trans-
port through a thin-walled viscoelastic tube in the context of
HFV. They investigated the effects of steady streaming on
the time-averaged mass transfer rate, from which an optimal
frequency of oscillation depending on the airway flexibility
was determined for HFV. As in any Eulerian formulation,
Dragon and Grotberg had to expand by Taylor’s theorem the
moving wall position about the mean position, where they
imposed the boundary conditions. Such an approach would,
Phys. Fluids 21, 051901 共2009兲
however, limit the validity of their results to waves of ex-
tremely small amplitude when the oscillation frequency, as
indicated by the Womersley number, ␣ = a共␴ / ␯ f 兲1/2 共where a
is the tube radius, ␴ is the angular frequency of oscillation,
and ␯ f is the kinematic viscosity of fluid兲, is high. One way
to interpret the Womersley number is that it is a ratio of two
length scales: the tube radius a and the Stokes layer thick-
ness ␦ = 共2␯ f / ␴兲1/2. The Stokes layer is a viscous boundary
layer formed near the boundary of the fluid domain when
subjected to wavy motion; the layer is thinner for lower fluid
viscosity or higher frequency. Mathematically, one can easily
show that the Taylor expansion of the instantaneous wall
position about the mean position is applicable only when the
wall displacement is much smaller than the Stokes layer
thickness, which is already small when the Womersley num-
ber is large. To circumvent this problem, one may resort to
the use of either the moving curvilinear coordinates or the
Lagrangian coordinates. The latter is adopted in the present
work.
Soft biological tissues are complex in structure, and
hence exhibit complex mechanical behaviors in general. As a
first approximation, they can be modeled as simple as a lin-
ear isotropic elastic material, in which the stress is linearly
proportional to the strain. A more sophisticated model is to
consider the tissues to behave like a viscoelastic material, in
which the stress is related not only to the strain but also to
the strain rate. As the incorporation of both elasticity and
viscosity, through different fashions of combination, can lead
to results that better agree with clinical observations, vis-
coelastic models of various kinds have been put forward to
describe the mechanical properties of physiological tissues.
A viscoelastic material may exhibit some distinct behaviors,
such as relaxation, creep, hysteresis 共a phase lag兲 under cy-
clic loading, wave attenuation, and imperfect rebound on the
termination of loading, which are phenomena often observed
in the mechanical response of many biomaterials. When
studying resonance in the terminal aortic circulation, Burat-
tini et al.28 found that viscosity was responsible for the
damping that inhibited sharp peaks from occurring in the
pressure and flow pulses in their experiments, and hence
concluded that viscoelasticity, rather than pure elasticity, was
influential in the aortic impedance. Morgan and Kiely11 were
among the first to investigate wave propagation in a viscous
fluid contained in a viscoelastic tube with thin walls. With a
focus only on the leading-order simple harmonic motion,
they derived that the viscoelastic stress-strain relations are
the same as those for a purely elastic material provided the
elastic constants are replaced by complex parameters, where
the imaginary parts represent the viscous dissipating effects.
In the context of blood flow, Cox4 studied propagation of
harmonic pressure waves in fluid contained in a viscoelastic
tube of finite wall thickness, where the wall properties would
depend on the frequency through a three-parameter
relaxation-type model. Again, based on the linearized mo-
mentum equations ignoring the convective acceleration, Cox
showed that the wave characteristics are very sensitive to
changes in the degree of wall viscoelasticity. Nonlinear vis-
coelastic models, where the nonlinearity is essentially for the
elastic part, have been adopted in some recent works. A non-
051901-3 Wave propagation and induced steady streaming
linear elastic constitutive relation in terms of the Green strain
tensor was introduced by Kyriacou and Humphrey12 for a
numerical study on the mechanics of a saccular aneurysm,
while a Kelvin–Voigt type of viscoelastic model in terms of
the Finger deformation tensor was used by Demiray13 for
arteries subjected to large initial static deformations and
small dynamic displacements.
Although viscoelasticity is increasingly recognized as an
important property of soft biological tissues and has been
extensively studied for its first-order effects on the oscilla-
tory flow, its effects on the second-order steady streaming in
waves are much less studied. Wang and Tarbell26 studied
steady streaming in blood flow but only for flow in a purely
elastic tube. Dragon and Grotberg5 presented an analytical
study on HFV oscillatory flow and steady streaming in a
thin-walled viscoelastic tube. Their modeling of the wall vis-
coelasticity is, however, rather heuristic: simply by adding a
linear damping term to the inertia in the stress balance equa-
tions for the tube wall. While the use of a constant damping
coefficient may be acceptable as far as the first-order oscil-
latory flow is concerned, it may not properly reflect the vis-
coelastic effects on the second-order streaming. A more sys-
tematic approach should have the viscoelastic stress-strain
relationships be expressed by a tensorial constitutive law that
can account for the finite deformation effects. Apparently, no
work in the literature has examined the viscoelastic effects
on the steady streaming following such an approach. This
has motivated the present study.
On the other hand, the in vivo state of a blood vessel is
inflated with a mean pressure of approximately 100 mm Hg
and is also under longitudinal tension with a stretch ratio
approximately equal to 1.5. These create relatively high ini-
tial circumferential and axial stresses in the tube. The prob-
lem of the influences of the initial stress on the blood flow is
of great interest and has been the focus of many studies in
the literature. Atabek and Lew18 first studied the wave propa-
gation in an initially stressed elastic tube. They pointed out
that the circumferential and longitudinal initial stresses play
different roles in the two possible distinct waves that may
propagate along the tube. Later, Atabek14 extended their iso-
tropic model to an orthotropic tube while making a sign cor-
rection to their original dispersion equation.18 Kuiken19 also
investigated the effects of the initial stresses on the wave
speeds but with a formulation that was different from that of
Atabek and Lew. However, Kuiken’s theory was questioned
by Hart and Shi29 for its inconsistency in the use of stress
terms in the equations of motion. Apparently, although many
works have been done on the problem of wave propagation
along a prestressed tube, some basic relations between the
waves and the initial stress remain unclear and need to be
studied further. Also, the works mentioned above have all
modeled the tube to be a linear thin-walled elastic shell
where the circumferential and longitudinal initial stresses can
be represented by two constants. For a thick-walled tube
model with a nonlinear constitutive law, the initial stresses in
the tube can no longer be constants but functions of the
radial coordinate. Demiray13 carried out an analysis on a
Kelvin–Voigt viscoelastic tube subjected to a large initial
static deformation. However, he did not consider the influ-
Phys. Fluids 21, 051901 共2009兲
ences of the initial stresses on the two distinct waves and the
second-order steady current, which will be analyzed in the
present work in detail.
A specific aim of the present work is to develop a model
for small-amplitude oscillatory flow and steady streaming,
induced by periodic pressure forcing, of a Newtonian viscous
fluid through a prestressed, viscoelastic tube that can be a
thin-walled or thick-walled structure. In order not to exces-
sively complicate the problem, a linear Voigt type of vis-
coelastic model is adopted in this work. By this model, the
stress is a sum of two parts: the viscous part, which is lin-
early proportional to the rate of deformation tensor, and the
elastic part, which is linearly proportional to the Finger de-
formation tensor. When the viscous part vanishes, the model
will reduce to the neo-Hookean model, which extends
Hooke’s law to fairly large deformations 共e.g., Ref. 30兲. For
generality, we place no restriction on the tube wall thickness,
which can be much smaller or comparable to the tube radius;
the condition of a thin wall is a limiting case of the present
problem. In addition, the tube is assumed to be long, straight,
prestressed, and with a uniform cross section initially. The
tube wall material is further assumed to be isotropic, homo-
geneous, and incompressible. Two alternative boundary con-
ditions are considered on the outer surface of the tube: it can
be either stress-free 共i.e., a free tube without constraint by the
surrounding tissues兲 or zero displacement axially 共i.e., a teth-
ered tube constrained by the surrounding tissues兲. Other ef-
fects, such as tapering, curvature, torsion waves, and gravity,
are neglected. Fully developed flow is considered here, so
that the end effects can also be neglected.
The present analysis is limited to small but finite defor-
mations. The wave amplitude, or the wall radial displace-
ment, is assumed to be much smaller than the tube radius so
that their ratio, which is much less than unity, can be used as
an ordering parameter for a perturbation analysis. The small
parameter also reflects a sharp contrast in length scales in the
radial and axial directions. As in Refs. 31 and 32, a pertur-
bation analysis based on Lagrangian coordinates is per-
formed here to the second order in the small parameter: the
first order for the purely oscillatory motions and the second
order for the time-mean motions. For the present problem
involving a moving wall that is initially stressed, the La-
grangian approach, by which boundary conditions are ap-
plied exactly on the moving boundaries, has three major ad-
vantages over the Eulerian approach. First, as remarked
above, this will avoid the problem encountered by Taylor
expanding a wavy surface about its mean position, which is
applicable only when the wave amplitude is much smaller
than the Stokes layer thickness. Second, the initial static de-
formations and stresses can be handled more conveniently
and naturally by a Lagrangian description. Third, the stream-
ing velocity, which depicts the velocity of a particle, is La-
grangian in nature. Therefore, it is more direct to solve for
the steady streaming in the Lagrangian system.
II. MATHEMATICAL FORMULATION
Let us consider the propagation of small-amplitude
waves traveling along a long flexible tube of circular section.
051901-4 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

R=b
R=a
outer wall surface
deformation gradient tensor given in Eq. 共9兲 and B = F · FT is
inner wall surface
viscoelastic wall the Finger or left Cauchy–Green tensor.30,33 In its two limits,
particle displacements
the constitutive relation 共4兲 represents a Newtonian viscous
viscous fluid subjected to
r - R = r’(R,Z,t) fluid when G = 0 and represents a neo-Hookean elastic solid
pf = p cos(σt) oscillatory pressure forcing
o

when ␮ = 0. In a sense, the neo-Hookean law generalizes

at Z = 0 z - Z = z’(R,Z,t)

R .
(r,z) Hooke’s law. While the latter involves only linear strains and
.
(R,Z) deformation is good for infinitesimally small deformations, the former
Z tube axis
R=0 contains quadratic strains and may describe fairly large de-
formations of an isotropic material. Note that F is not sym-
FIG. 1. Schematic diagram of the problem, where description is based on
Lagrangian axial and radial coordinates, Z and R. Viscous fluid contained in metric, but D, E, and B are symmetric tensors, and for a
a viscoelastic tube is subjected to time-periodic oscillatory pressure forcing basic constitutive equation that is frame indifferent, it is ac-
at Z = 0. As a result, a traveling wave is induced in the fluid and the wall. A ceptable to express the Cauchy stress to be a linear function
fluid element initially centered at 共R , Z兲, on undergoing deformation, moves
of E or B.34 By including a term that is proportional to B−1,
to a new center 共r , z兲 at time t.
the neo-Hookean model can be further generalized into the
Mooney–Rivlin model that can handle even larger deforma-
tions. Such a nonlinear model is, however, not considered in
The core of the tube, which has an initially inflated radius a, this work.
is filled with an incompressible Newtonian viscous fluid. The Based on the constitutive equation 共4兲, the deviatoric
wall of the tube, whose outer radius is initially stressed at b, stress components can be expressed as follows:
is supposed to be made up of a Voigt type of material. By
Lagrangian description, the instantaneous radial and axial
coordinates of a particle of fixed identity, 共r , z兲, and the pres-
sure p are functions of the initial coordinates of the particle,
␶rr = 2␮
⳵ ṙ
⳵r
+G 冋冉 冊 冉 冊 册
⳵r
⳵R
2
+
⳵r
⳵Z
2
−1 , 共5兲

共R , Z兲, and time t. Axisymmetry is assumed, and hence any

dependence on the azimuthal position ␪ is eliminated
共Fig. 1兲. As a result of an oscillatory pressure gradient of
angular frequency ␴, a progressive wave of wavenumber k is
ṙ
␶␪␪ = 2␮ + G
r
冋冉 冊 册
r
R
2
−1 , 共6兲

induced in the system, where k and ␴ are related by a dis-

persion equation, as will be deduced later.

A. Eulerian governing equations

␶zz = 2␮
⳵ ż
⳵Z
+G 冋冉 冊 冉 冊 册
⳵z
⳵R
2
+
⳵z
⳵Z
2
−1 , 共7兲

冋 册 冋冉 冊冉 冊 冉 冊冉 冊册
Let us first recall the conventional continuity and mo-
mentum equations expressed in terms of Eulerian cylindrical ⳵ ṙ ⳵ ż ⳵r ⳵z ⳵r ⳵z
␶rz = ␶zr = ␮ + +G + ,
coordinates: ⳵z ⳵r ⳵R ⳵R ⳵Z ⳵Z
⳵ ṙ ṙ ⳵ ż 共8兲
+ + = 0, 共1兲
⳵r r ⳵z
and other components being identically zero by axisym-
dṙ
dt
=− +
␳ ⳵r ␳ r ⳵ r
+ 冋
1 ⳵ p 1 ⳵ 共r␶rr兲 ⳵ ␶rz ␶␪␪
⳵z
−
r
, 册 共2兲
metry.

B. Lagrangian governing equations

dż
dt
=− +
␳ ⳵z ␳ r ⳵ r
+ 冋
1 ⳵ p 1 ⳵ 共r␶zr兲 ⳵ ␶zz
⳵z
, 册
where an overdot denotes time derivative, d / dt stands for the
total derivative, ␳ represents the density, p is the pressure,
and ␶ij are the deviatoric stress components 共i.e., the isotro-
pic normal stress being subtracted from the total stress com-
ponents兲. Here, we need to introduce a constitutive equation
to relate the stress and the deformation in order to close the
problem. A linear Voigt viscoelastic model is proposed: the
Cauchy stress T is a linear combination of viscous and elas-
tic parts:31–33
T = − pI + 2␮D + GE,
where I is the identity tensor, ␮ is the classical Newtonian
viscosity, D = 21 共L + LT兲 is the rate of deformation tensor,
L = Ḟ · F−1 is the velocity gradient tensor, G is the shear
modulus, and E = B − I is the strain tensor, in which F is the
共3兲
One may refer to Monin and Yaglom35 for a derivation
of the Lagrangian form of the Navier–Stokes equations in
Cartesian coordinates. This Lagrangian approach, which con-
veniently handles the free-surface boundary conditions for
ocean waves, as was first adopted by Pierson,36 has been
followed by Ünlüata and Mei,37 and more recently by Ng38
and Ng and Zhang.32
In the present Lagrangian description, cylindrical coor-
dinates are to be used, and the equations do not seem to have
been explicitly deduced in the literature. The dynamic prob-
lem here is to be expressed in terms of the instantaneous
coordinates 共r , z兲 and pressure p as functions of the initial
共4兲 coordinates 共R , Z兲 and time t. In the following, let us provide
some details outlining how the governing equations of mo-
tion and boundary conditions in terms of Lagrangian cylin-
drical coordinates are developed.
The deformation gradient tensor F in cylindrical coordi-
nates is given by30
051901-5 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

ជ ជ
F = ⳵ x/⳵ X = 0
冢
⳵ r/⳵ R 0 ⳵ r/⳵ Z
r/R 0 ,
冣 共9兲
␶lzz = 2␮l
rl ⳵ 共rl,żl兲
R ⳵ 共R,Z兲
+ Gl 冋冉 冊 冉 冊 册
⳵ zl
⳵R
2
+
⳵ zl
⳵Z
2
−1 , 共19兲

⳵ z/⳵ R 0 ⳵ z/⳵ Z and

where X ជ and xជ are the position vectors of a particle in the
initial and instantaneous states, respectively. The third invari- ␶lrz = ␶lzr = ␮l 冋 册
rl ⳵ 共rl,ṙl兲 ⳵ 共żl,zl兲
+
R ⳵ 共R,Z兲 ⳵ 共R,Z兲

冋冉 冊冉 冊 冉 冊冉 冊册
ant of the deformation gradient tensor, or the determinant of
F, physically represents a ratio of the volume of a material ⳵ rl ⳵ zl ⳵ rl ⳵ zl
+ Gl + . 共20兲
element in its present 共deformed兲 state to that in its past ⳵R ⳵R ⳵Z ⳵Z
共undeformed兲 state.34 By mass conservation, the volume ra-
tio is equal to the inverse of the density ratio. If further the In these equations, G f = 0 for the fluid is purely viscous.
material is assumed to be incompressible so that its density is
invariant with time and the volume ratio is equal to unity, C. Boundary conditions
then the conservation of mass is mathematically given by To complete the formulation, we need the following
det F = 兩F兩 = 1, or boundary conditions. By symmetry about the tube axis at R
rl ⳵ 共rl,zl兲 = 0:
= 1, 共10兲
R ⳵ 共R,Z兲 ⳵zf
r f = 0, =0 on R = 0. 共21兲
where the Jacobian35 ⳵R

⳵ 共rl,zl兲
⳵ 共R,Z兲
= 冏
⳵ r l/ ⳵ R ⳵ r l/ ⳵ Z
⳵ z l/ ⳵ R ⳵ z l/ ⳵ Z
, 冏 共11兲
By the continuity of particle displacements and stress com-
ponents on the fluid-wall interface at R = a and zero stresses
on the outer wall at R = b for a free tube:
and the subindex l is used to distinguish between the fluid r f = r w, z f = z w, T f = T w, N f = Nw on R = a, 共22兲
and wall domains,

l= 再 f for fluid in a ⬎ R ⬎ 0
w for wall in b ⬎ R ⬎ a,
冎 共12兲
Tw = 0, Nw = 0 on R = b,
where T and N are, respectively, the tangential and normal
共23兲

in which a and b are the initial inner and outer radii of the stress components on a material surface as seen in an 共R , Z兲
tube wall, respectively. Using Cramer’s rule,36,39 for any plane. We may derive that to a material curve S
variable f, the Eulerian derivatives can be transformed into = 共r共R , Z , t兲 , z共R , Z , t兲兲, the unit tangential and normal vectors
the Lagrangian ones through the following relations: are t = 共⳵r / ⳵Z , ⳵z / ⳵Z兲 / 兩n兩 and n = 共⳵z / ⳵Z , −⳵r / ⳵Z兲 / 兩n兩, where

冒
兩n兩 = 关共⳵r / ⳵Z兲2 + 共⳵z / ⳵Z兲2兴1/2. Hence, the tangential and nor-
⳵ f ⳵ 共f,z兲 ⳵ 共r,z兲 r ⳵ 共f,z兲 mal stress components, T = 共T · n兲 · t and N = 共T · n兲 · n, can be
= = , 共13兲
⳵ r ⳵ 共R,Z兲 ⳵ 共R,Z兲 R ⳵ 共R,Z兲 expressed by

⳵ f ⳵ 共r, f兲
=
⳵ z ⳵ 共R,Z兲
冒 ⳵ 共r,z兲 r ⳵ 共r, f兲
=
⳵ 共R,Z兲 R ⳵ 共R,Z兲
. 共14兲
Tl兩n兩2 = 共␶lrr − ␶lzz兲
⳵ rl ⳵ zl
⳵Z ⳵Z
+ ␶lzr 冋冉 冊 冉 冊 册
⳵ zl
⳵Z
2
−
⳵ rl
⳵Z
2
, 共24兲

By virtue of these relationships, the Eulerian radial and axial

momentum equations 共2兲 and 共3兲 and the stress components
Nl兩n兩2 = 共− pl + ␶lrr兲 冉 冊
⳵ zl
⳵Z
2
+ 共− pl + ␶lzz兲 冉 冊 ⳵ rl
⳵Z
2
− 2␶lrz
⳵ rl ⳵ zl
⳵Z ⳵Z
.
共5兲–共8兲 can be transformed into the Lagrangian form as fol-
共25兲
lows:
⳵ 2r l
⳵t 2 =− + 冋
rl ⳵ 共pl,zl兲 1 1 ⳵ 共rl␶lrr,zl兲 rl ⳵ 共rl, ␶lrz兲 ␶l␪␪
␳lR ⳵ 共R,Z兲 ␳l R ⳵ 共R,Z兲
+
R ⳵ 共R,Z兲
−
rl
, 册 For a tethered tube, the longitudinal motion of the wall is
strongly constrained, and hence the axial displacement of the
outer surface of the wall can be assumed to be zero. Such a
共15兲 boundary condition will be presented in the following first-
order dynamic problem.
⳵ 2z l
⳵t 2 =− + 冋
rl ⳵ 共rl,pl兲 1 1 ⳵ 共rl␶lzr,zl兲 rl ⳵ 共rl, ␶lzz兲
␳lR ⳵ 共R,Z兲 ␳l R ⳵ 共R,Z兲
+
R ⳵ 共R,Z兲
, 共16兲 册 We suppose that the tube radius and the wall thickness
are comparable to each other, both of them being much
smaller than the wavelength, which is of the order of the tube
where length L. Based on this assumption of sharp contrast in

冋冉 冊 冉 冊 册
length scales, the following parameter:
rl ⳵ 共ṙl,zl兲 ⳵ rl 2
⳵ rl 2
␶lrr = 2␮l + Gl + −1 , 共17兲 ⑀ ⬅ a/L Ⰶ 1, 共26兲
R ⳵ 共R,Z兲 ⳵R ⳵Z

冋冉 冊 册
will be used as an ordering parameter for the perturbation
2
ṙl rl analysis below. This condition is commonly met in biofluid
␶l␪␪ = 2␮l + Gl −1 , 共18兲
rl R flows.3,5
051901-6 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

D. Normalization
Let us now introduce the following normalized vari-
ables, which are distinguished by a caret:
⳵2ẑ f
⳵ t̂ 2
=−
r̂ f ⳵ 共r̂ f ,p̂ f 兲
R̂ ⳵ 共R̂,Ẑ兲
+ 冋
1 1 ⳵ 共r̂ f ␶ˆ fzr,ẑ f 兲
␣ R̂ ⳵ 共R̂,Ẑ兲
2 + ⑀2
r̂ f ⳵ 共r̂ f , ␶ˆ fzz兲
R̂ ⳵ 共R̂,Ẑ兲
, 册
共31兲
共rl,R,b兲 = a共r̂l,R̂,b̂兲, 共zl,Z兲 = L共ẑl,Ẑ兲, k = L−1k̂, t = ␴−1t̂,
⳵2r̂w ␤ r̂w ⳵ 共p̂w,ẑw兲
⑀2 =−
⳵ t̂ 2 ␣4 R̂ ⳵ 共R̂,Ẑ兲
p f = 共␳ f ␴2L2兲p̂ f , pw = Gw p̂w, N f = ⑀−2␮ f ␴N̂ f , Nw = GwN̂w ,

共␶ frz,T f 兲 = ⑀−1␮ f ␴共␶ˆ frz,T̂ f 兲,

+ 冋
␤ 1 ⳵ 共r̂w␶ˆ wrr,ẑw兲 2 r̂w ⳵ 共r̂w, ␶ˆ wrz兲 ␶ˆ w␪␪
␣4 R̂ ⳵ 共R̂,Ẑ兲
+⑀
R̂ ⳵ 共R̂,Ẑ兲
−
r̂w
, 册
共27兲 共32兲
共␶ frr, ␶ f ␪␪, ␶ fzz兲 = ␮ f ␴共␶ˆ frr, ␶ˆ f ␪␪, ␶ˆ fzz兲,
⳵2ẑw ␤ r̂w ⳵ 共r̂w,p̂w兲
=−
共␶wrz,Tw兲 = ⑀Gw共␶ˆ wrz,T̂w兲, ⳵ t̂ 2 ␣4 R̂ ⳵ 共R̂,Ẑ兲

共␶wrr, ␶w␪␪, ␶wzz兲 = Gw共␶ˆ wrr, ␶ˆ w␪␪, ␶ˆ wzz兲.

Most of the normalization scales are easily identifiable by the

+ 冋
␤ 1 ⳵ 共r̂w␶ˆ wzr,ẑw兲 r̂w ⳵ 共r̂w, ␶ˆ wzz兲
␣4 R̂ ⳵ 共R̂,Ẑ兲
+
R̂ ⳵ 共R̂,Ẑ兲
. 册 共33兲

balance of dominant terms in the equations above. The scal- The dimensionless stress components are given as below:
ing for the shear and tangential stresses in the wall needs
r̂ f ⳵ 共ṙˆ f ,ẑ f 兲 ṙˆ f r̂ f ⳵ 共r̂ f ,żˆ f 兲
some explanation, however. The normal stress in the fluid is ␶ˆ frr = 2 , ␶ˆ f ␪␪ = 2 , ␶ˆ fzz = 2 , 共34兲
dominated by the pressure, and hence N f ⬃ ␳ f ␴2L2 R̂ ⳵ 共R̂,Ẑ兲 r̂ f R̂ ⳵ 共R̂,Ẑ兲
⬃ ␮ f ␴共L / ␦兲2, where ␦ = 共2␯ f / ␴兲1/2 is the Stokes layer thick-
ness. For the problems under consideration, the oscillation
frequency may vary over a range such that ␦ ⱕ O共a兲, and
therefore a proper scaling for N f is ⑀−2␮ f ␴. On the other
␶ˆ frz = ␶ˆ fzr =
r̂ f
R̂
冋 ⑀2
⳵ 共r̂ f ,ṙˆ f 兲
⳵ 共R̂,Ẑ兲
+
⳵ 共żˆ f ,ẑ f 兲
⳵ 共R̂,Ẑ兲
册 , 共35兲

hand, the normal stress in the wall is scaled by Nw ⬃ pw

⬃ Gw. Assuming that the normal stress in the wall is at least
comparable to that in the fluid, Nw ⱖ O共N f 兲, we get a scaling
relationship for the wall elasticity,
␶ˆ wrr = 2
␣2␥ r̂w ⳵ 共ṙˆw,ẑw兲
␤ R̂ ⳵ 共R̂,Ẑ兲
+ 冋冉 冊 冉 冊 册 ⳵ r̂w
⳵ R̂
2
+ ⑀2
⳵ r̂w
⳵ Ẑ
2
−1 , 共36兲

Gw ⱖ O共⑀−2␮ f ␴兲.

By shallowness, the shear stress in the fluid is an order of

共28兲
␶ˆ w␪␪ = 2
␣2␥ ṙˆw
␤ r̂w
+ 冋冉 冊 册 r̂w
R̂
2
−1 , 共37兲

magnitude larger than the normal stress in the fluid, ␶ frz

⬃ ⑀−1␶ fzz ⬃ ⑀−1␮ f ␴. It follows that, by continuity of shear
stress on the interface and relation 共28兲, the shear stress in
the wall is scaled by ␶wrz ⬃ ⑀Gw, which is an order of mag-
␶ˆ wzz = 2
␣2␥ r̂w ⳵ 共r̂w,żˆw兲
␤ R̂ ⳵ 共R̂,Ẑ兲
+ ⑀−2 冋冉 冊 冉 冊 册
⳵ ẑw
⳵ R̂
2
+
⳵ ẑw
⳵ Ẑ
2
− 1 , 共38兲

nitude smaller than the scaling for the normal stress in the and

冋 册
wall. Despite the difference in the scalings, the wall shear
stress cannot be neglected as its radial gradient is comparable ␣2␥ r̂w ⳵ 共r̂w,ṙˆw兲 −2 ⳵ 共żˆw,ẑw兲
␶ˆ wrz = ␶ˆ wzr = +⑀
to the axial gradient of the normal stress by virtue of the ␤ R̂ ⳵ 共R̂,Ẑ兲 ⳵ 共R̂,Ẑ兲

冋 冉 冊冉 冊 冉 冊冉 冊册
shallowness condition 共26兲, as will be seen when the first-
order problem is formulated in Sec. IV. ⳵ r̂w ⳵ ẑw ⳵ r̂w ⳵ ẑw
Substituting the normalized variables, the problem can + ⑀−2 + . 共39兲
⳵ R̂ ⳵ R̂ ⳵ Ẑ ⳵ Ẑ
be cast into a dimensionless form as follows. The continuity
equation is In these equations, three dimensionless parameters show up:

⳵ 共r̂l,ẑl兲
⳵ 共R̂,Ẑ兲
=
R̂
r̂l
. 共29兲 ␣=a 冉冊 ␴
␯f
1/2
, ␤=
⑀ 2a 2G w
␳w␯2f
, ␥=
⑀ 2␯ w
␯f
, 共40兲

where ␯l = ␮l / ␳l is the kinematic viscosity. Here, ␣ is the

The momentum equations become Womersley number, which is equal to the square root of the
ratio of the wave frequency to the rate of radial diffusion of
⳵2r̂ f r̂ f ⳵ 共p̂ f ,ẑ f 兲
⑀2 =− momentum across the tube. As remarked earlier, the Womer-
⳵ t̂ 2
R̂ ⳵ 共R̂,Ẑ兲 sley number is also related to the ratio of the tube radius to

+ 冋
␣2 R̂ ⳵ 共R̂,Ẑ兲
+
R̂ ⳵ 共R̂,Ẑ兲
− 册
⑀2 1 ⳵ 共r̂ f ␶ˆ frr,ẑ f 兲 r̂ f ⳵ 共r̂ f , ␶ˆ frz兲 ␶ˆ f ␪␪
r̂ f
, 共30兲
the Stokes layer thickness ␣ ⬃ a / ␦, which in biological sys-
tems typically lies in the range O共1兲 – O共10兲. The second
parameter, ␤, represents the significance of the tube elastic-
051901-7 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

ity. Using Eq. 共28兲, one may show that ␤ ⬃ ␳ f a2 / 共␳w␦2兲 共␶ˆ frr, ␶ˆ f ␪␪, ␶ˆ fzz, ␶ˆ lrz兲 = ⑀共␶ˆ frr1, ␶ˆ f ␪␪1, ␶ˆ fzz1, ␶ˆ lrz1兲
ⱖ O共1兲. In the limit of ␤ → ⬁, the wall becomes perfectly
rigid. The third parameter, ␥, is a ratio of the tube viscosity + ⑀2共␶ˆ frr2, ␶ˆ f ␪␪2, ␶ˆ fzz2, ␶ˆ lrz2兲 + ¯ , 共51兲
to the fluid viscosity. Here, the tube viscosity is allowed to and
vary from as small as zero for pure elastic wall to as large as
two orders of magnitude larger than the fluid viscosity. 共␶ˆ wrr, ␶ˆ w␪␪, ␶ˆ wzz兲 = 共␶ˆ wrr0, ␶ˆ w␪␪0, ␶ˆ wzz0兲 + ⑀共␶ˆ wrr1, ␶ˆ w␪␪1, ␶ˆ wzz1兲
Therefore, we may suppose ␥ ⱕ O共1兲. + ⑀2共␶ˆ wrr2, ␶ˆ w␪␪2, ␶ˆ wzz2兲 + ¯ , 共52兲
The normalized boundary conditions become
where the O共⑀兲 terms are the leading-order displacements
⳵ ẑ f consisting of purely first harmonic components, while the
r̂ f = 0, =0 on R̂ = 0, 共41兲
⳵ R̂ higher-order terms consist of higher harmonics as well as
steady components. The O共1兲 normal stress components rep-
␣2 ␣2 resent the initial stresses of the tube as it is subjected to static
r̂ f = r̂w, ẑ f = ẑw, T̂ f = T̂w, N̂ f = N̂w on R̂ = 1, fluid pressure Q = ␳ f ␴2L2 p̂ f0 = ␳ f L2␯2f ␣4 p̂ f0 / a4 and axial elon-
M␤ M␤
gation given by the stretch ratio S. To suppress the ⑀−2 terms
共42兲 in the wall stress components given by Eqs. 共38兲 and 共39兲,
the axial displacement of the wall can only have a very weak
T̂w = 0, N̂w = 0 on R̂ = b̂, 共43兲 dependence on R̂ such that ⳵ẑw1 / ⳵R̂ = O共⑀2兲, ⳵ẑw2 / ⳵R̂ = O共⑀2兲,
where and so on. In view of this, we should further expand ẑw1 and
ẑw2 into
␳w
M= = O共1兲, 共44兲 ẑw1共R̂,Ẑ,t̂兲 = ẑw11共Ẑ,t̂兲 + ⑀2ẑw12共R̂,Ẑ,t̂兲, 共53兲
␳f
which represents the ratio of wall density to fluid density.
ẑw2共R̂,Ẑ,t̂兲 = ẑw21共Ẑ,t̂兲 + ⑀2ẑw22共R̂,Ẑ,t̂兲, 共54兲
The dimensionless expressions for T̂l and N̂l are

T̂ f 兩n兩 = ⑀ 共␶ˆ frr − ␶ˆ fzz兲

2 2 ⳵ r̂ f ⳵ ẑ f
⳵ Ẑ ⳵ Ẑ
+ ␶ˆ fzr 冋冉 冊 冉 冊 册
⳵ ẑ f
⳵ Ẑ
2
−⑀ 2 ⳵ r̂ f
⳵ Ẑ
2
,
where the leading terms ẑw11 and ẑw21 are independent of R̂.
The axial displacement of the wall will have a nearly uni-
form profile across a section or the wall will undergo motion
共45兲 with very little shear deformation: ⳵zw / ⳵R ⬃ ⳵rw / ⳵Z = O共⑀2兲.
Despite its small magnitude, the shear strain of the wall is,

N̂ f 兩n兩2 = 共− ␣2 p̂ f + ⑀2␶ˆ frr兲 冉 冊 ⳵ ẑ f

⳵ Ẑ
2
+ ⑀2共− ␣2 p̂ f + ⑀2␶ˆ fzz兲 冉 冊 ⳵ r̂ f
⳵ Ẑ
2 however, required to be taken into account and will show up
even in the leading-order problem.
Substitution of the above expansions into the governing
⳵ r̂ f ⳵ ẑ f equations 共29兲–共33兲, the stress components 共34兲–共39兲, and
− 2⑀2␶ˆ frz , 共46兲 the boundary conditions 共41兲–共43兲 and collecting terms of
⳵ Ẑ ⳵ Ẑ equal power of ⑀, we may obtain the zeroth- 共initial stress兲,

冋冉 冊 冉 冊 册
first-, and second-order problems as detailed in the following
⳵ r̂w ⳵ ẑw ⳵ ẑw 2
⳵ r̂w 2
sections.
T̂w兩n兩2 = 共␶ˆ wrr − ␶ˆ wzz兲 + ␶ˆ wzr − ⑀2 ,
⳵ Ẑ ⳵ Ẑ ⳵ Ẑ ⳵ Ẑ
共47兲
III. INITIAL STRESS PROBLEM
and

冉 冊 冉 冊
It is assumed that, in the initial static state, the tube is
⳵ ẑw 2
⳵ r̂w 2
under a steady uniaxial extension and also subjected to
N̂w兩n兩 = 共− p̂w + ␶ˆ wrr兲
2
+ ⑀ 共− p̂w + ␶ˆ wzz兲
2

⳵ Ẑ ⳵ Ẑ steady fluid pressure Q on the inner wall in the absence of

any shear stresses. The continuity equation, which relates the
⳵ r̂w ⳵ ẑw stress-free state to the prestressed state of the tube, is
− 2⑀2␶ˆ wrz . 共48兲
⳵ Ẑ ⳵ Ẑ
⳵ ĉ ⳵ d̂ R̂
= , 共55兲
E. Asymptotic expansions ⳵ R̂ ⳵ Ẑ ĉ

On assuming that the induced radial and axial displace- where 共ĉ , d̂兲 are the coordinates of a particle when un-
ments in both the fluid and the wall are one order of magni- stressed. The uniaxial extension amounts to
tude smaller than the corresponding scalings, we may follow
Pierson36 in expanding the variables into powers of ⑀: ⳵ d̂ 1
= , 共56兲
⳵ Ẑ S
共r̂l,ẑl兲 = 共R̂,Ẑ兲 + ⑀共r̂l1,ẑl1兲 + ⑀ 共r̂l2,ẑl2兲 + ¯ ,
2
共49兲
where S ⬎ 1 is the longitudinal stretch ratio. Following
共p̂ f ,p̂w兲 = 共p̂ f0,p̂w0兲 + ⑀共p̂ f1,p̂w1兲 + ⑀2共p̂ f2,p̂w2兲 + ¯ , 共50兲 Demiray,13 the solution to the equations above is simply
051901-8 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

ĉ2 = 共R̂2 − j1兲S,

where j1 is an integration constant. By Eq. 共57兲, the initial
共57兲 2b̂2 − j1
2b̂2S
−
1
2S
ln 冉 冊
b̂2
b̂2 − j1
− j2 = 1. 共66兲

stress components read as follows:

冉 冊
Hence, the initial static stresses of the tube can be found for
2
⳵ R̂ R̂2 − j1 given values of the loading parameters S and Q̂ and the outer
␶ˆ wrr0 = −1= − 1, 共58兲
⳵ ĉ R̂2S wall radius b̂.

␶ˆ w␪␪0 = 冉冊 R̂
ĉ
2

−1=
R̂2
共R̂2 − j1兲S
− 1, 共59兲
IV. FIRST-ORDER PROBLEM
A. First-order solution

and At O共⑀兲, the governing equations for the fluid are

␶ˆ wzz0 = 冉冊 ⳵ Ẑ
⳵ d̂
2

− 1 = S2 − 1. 共60兲
⳵ ẑ f1
⳵ Ẑ
+
⳵ r̂ f1
⳵ R̂
+
r̂ f1
R̂
= 0, 共67兲

The leading-order stress balance equation in the radial direc- ⳵ p̂ f1

= 0, 共68兲
tion is ⳵ R̂

冋 册
⳵ p̂w0 ⳵ ␶ˆ wrr0 ␶ˆ wrr0 − ␶ˆ w␪␪0
− + + = 0, 共61兲 ⳵2ẑ f1 ⳵ p̂ f1 1 ⳵ ␶ˆ frz1 ␶ˆ frz1
⳵ R̂ ⳵ R̂ R̂ =− + + , 共69兲
⳵ t̂2 ⳵ Ẑ ␣2 ⳵ R̂ R̂
which results in

冉 冊
and those for the tube wall are
1 R̂2
j1
p̂w0 = − + ln + j2 , 共62兲 ⳵ ẑw11 ⳵ r̂w1 r̂w1
2R̂2S 2S R̂2 − j1 + + = 0, 共70兲
⳵ Ẑ ⳵ R̂ R̂
where j2 is another integration constant. The leading-order
stress boundary conditions include ⳵ p̂w1 ⳵ p̂w0 r̂w1 ⳵ p̂w0 ⳵ ẑw11 ⳵ ␶ˆ wrr1 ⳵ ␶ˆ wrr0 ⳵ ẑw11
− − − + +
− p̂w0 + ␶ˆ wrr0 = − Q̂ on R̂ = 1, 共63兲 ⳵ R̂ ⳵ R̂ R̂ ⳵ R̂ ⳵ Ẑ ⳵ R̂ ⳵ R̂ ⳵ Ẑ

where Q̂ = Q / Gw = ␣4 p̂ f0 / 共M ␤兲 and ⳵ ␶ˆ wrr0 r̂w1 ␶ˆ wrr1 − ␶ˆ w␪␪1 r̂w1共␶ˆ wrr0 − ␶ˆ w␪␪0兲

+ + − = 0,
⳵ R̂ R̂ R̂ R̂2
− p̂w0 + ␶ˆ wrr0 = 0 on R̂ = b̂. 共64兲
共71兲
Substitution of the initial stress components into the above
boundary conditions, the two integration constants j1 and j2 ␣4 ⳵2ẑw11 ⳵ p̂w1 ⳵ p̂w0 ⳵ r̂w1 ⳵ ␶ˆ wrz1 ␶ˆ wrz1 ⳵ ␶ˆ wzz1
can be determined from the following equations: =− + + + + ,

冋 册
␤ ⳵ t̂ 2
⳵ Ẑ ⳵ R̂ ⳵ Ẑ ⳵ R̂ R̂ ⳵ Ẑ
b̂2共1 − j1兲
ln − 共1 − b̂−2兲j1 = − 2SQ̂ 共65兲 共72兲
2
b̂ − j1
where the stress components ␶ˆ lij1 are given in Appendix A.
and The first-order boundary conditions are

⳵ ẑ f1
r̂ f1 = 0, =0 on R̂ = 0, 共73兲
⳵ R̂

冦 冧
r̂ f1 = r̂w1, ẑ f1 = ẑw11
␣ 2
⳵ r̂w1
␶ˆ frz1 = ␶ˆ wrz1 + 共␶ˆ wrr0 − ␶ˆ wzz0兲
M␤ ⳵ Ẑ on R̂ = 1, 共74兲

−
␣
M␤
4

冉
2p̂ f0
⳵ ẑ f1
⳵ Ẑ
冊
+ p̂ f1 = − p̂w1 + ␶ˆ wrr1 + 2共− p̂w0 + ␶ˆ wrr0兲
⳵ ẑw11
⳵ Ẑ

and
051901-9 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

冦 冧
␶ˆ wrz1 + 共␶ˆ wrr0 − ␶ˆ wzz0兲
⳵ r̂w1
⳵ Ẑ
=0 共case 1兲 冋
zowv共R̂兲 = k̂2共B4 + B5兲 −
␣
␨
B4 册
R̂2
4
+ ln共R̂兲B3

ẑw11 = 0 共case 2兲

− p̂w1 + ␶ˆ wrr1 + 2共− p̂w0 + ␶ˆ wrr0兲

⳵ ẑw11
=0
on R̂ = b̂.
+
␨S 冋
ik̂B2␤ 3j1 1
−
8R̂2 2
冕 ln共R̂2 − j1兲
R̂
dR̂

冉冑 冊 册
⳵ Ẑ
共75兲 +
2
j1
冕冕
1
R̂
R̂ ln
R̂
R̂2 − j1
dR̂dR̂

冉冑 冊
In the outer boundary condition above, case 1 represents a
free tube, where the stresses on the outer wall are absent,
while in case 2 the tube is tethered by the perivascular tissues +
k̂2 j1B4␤
2␨S
冕 1
R̂
ln
R̂2 − j1
R̂
dR̂, 共83兲
so that the longitudinal motion of the outer surface is
strongly constrained and the leading order of the axial dis- where ␭ = ␣i3/2, ␨ = ␣2␥i + ␤, S is the longitudinal stretch ra-
placement can be taken as zero.27 tio, j1 is a constant of initial stress as defined in Eq. 共65兲, Jn
One can now see that the wall shear stress ␶ˆ wrz1, which is the Bessel function of the first kind of order n, and the
shows up in the equation of motion 共72兲 and the boundary complex integration constants B1 – B5 as well as the complex
conditions 共74兲 and 共75兲, is indeed a quantity that plays a wavenumber k̂ are to be determined by the boundary condi-
role in the first-order problem. This quantity will not lose its tions.
importance even when the wall thickness becomes very In case 1 for a free tube, substituting the general solu-
small. tions into the boundary conditions 共74兲 and 共75兲 results in a
We suppose that the pressure forcing that is applied to homogeneous linear algebraic system of equations for the six
the fluid at the origin p̂ f1共Ẑ = 0兲 is a simple harmonic function constants po and B1 – B5:
of t̂, and as a result, a traveling wave is induced in the fluid:
AX = 0, 共84兲
o i共t̂−kˆZˆ兲
p̂ f1 = Re关p e 兴, 共76兲 where the matrix A is given in Appendix B, and

where Re stands for the real part, po is a known constant X = 共po B1 B2 B3 B4 B5 兲T . 共85兲
equal to the amplitude of the pressure forcing at Ẑ = 0, and The system will have a nontrivial solution if and only if the
i = 冑−1 is the complex unit. To seek the solution, let us now determinant of the coefficient matrix A vanishes. This condi-
express the first-order quantities in the same harmonic form tion yields the characteristic equation governing the eigen-
as p̂ f1: value, which in this case is the wavenumber k̂, as a function
ˆˆ
of the frequency ␣ and material properties ␤ 共wall elasticity兲
共r̂ f1,ẑ f1,r̂w1,ẑw11,ẑw12,p̂w1兲 = Re关共rof ,zof ,rwo,zwc
o
,zowv,pwo兲ei共t̂−kZ兲兴, and ␥ 共wall viscosity兲, which are defined in Eq. 共40兲. Since it
共77兲 is a relation between wavenumber and frequency, the char-
acteristic equation is also called the dispersion equation,
where rof , zof , zof , rwo, zowv, and pwo are complex functions of R̂, which can be obtained in the following algebraic form after
o
and zwc is an unknown constant to be determined by the some manipulation of the matrix:
boundary conditions. On substituting these expressions into
A1k̂4 + A2k̂2 + A3 = 0, 共86兲
the governing equations, the following general solutions sat-
isfying Eq. 共73兲 can be deduced: where A1 – A3 are given in Appendix C. Equation 共86兲 is a
quadratic equation that can be readily solved for k̂2. It admits
zof 共R̂兲 = − ik̂po + ␭J0共␭R̂兲B1 , 共78兲 four complex solutions for k̂ = k̂r + ik̂i, where the real part k̂r
denotes the wavenumber, and the imaginary part k̂i is the
k̂2R̂ o attenuation constant. Our focus is on the forward traveling
rof 共R̂兲 = p + ik̂J1共␭R̂兲B1 , 共79兲 waves, so we only consider the two solutions with k̂r ⬎ 0 and
2
k̂i ⬍ 0. Each of these two solutions corresponds to a distinct
wave mode. The wave with a slower phase velocity C1
o
zwc = B4 , 共80兲 = ␴ / kr1 is called Young wave representing a pressure wave
propagating in the fluid, while the one with a higher phase
velocity C2 = ␴ / kr2 is called Lamb wave representing a wave
ik̂R̂ B2
rwo共R̂兲 = B4 + , 共81兲 traveling largely along the wall. A Young wave is generated
2 R̂ by an oscillatory pressure forcing in the fluid, while a Lamb
wave can be expected when the tube wall is excited harmoni-

pwo =
␨
ik̂B +
␤ 5 冋 1
R̂ 2
−
j1
2R̂ 4
−
2
j1 冉 冑 冊册
ln
R̂
2
R̂ − j1
B2
S
, 共82兲
cally in the direction of the axis.3 Most of our discussions
below will be on the Young wave mode, which is more im-
portant in the present study.
051901-10 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

1
The constants B1 – B5 are related to the pressure ampli-
14
tude po by the following relations:
po B1 B2 B3 B4 B5 12 0.8
= = = = = , 共87兲

Transmission per wavelength

⌬11 ⌬12 ⌬13 ⌬14 ⌬15 ⌬16

Phase velocity C 1 (m/s)

10
where ⌬kl is the cofactor obtained by eliminating the kth row 0.6

and lth column of 兩A兩. 8

In case 2 for a tethered tube, B4 = 0 and the dispersion

equation becomes much simpler, which can be expressed as 6 0.4

follows:
4
␣4␭J0共␭兲 Womersley model
6
0.2
k̂2 = − , 共88兲 Cox model 40

M共␣ ␥i + ␤兲关J1共␭兲 − ␭J0共␭兲/2兴关2共1 − 1/b̂ 兲 + ␹兴

2 2 2 Present free-tube model
Present tethered-tube model

where 0 0

冉 冊
2 4 6 8 10
α
␤ 1 j1 j1 2 b̂冑1 − j1
␹= 共89兲
b̂2 2 2b̂4 j1 冑b̂2 − j1
1− − + + ln . FIG. 2. Phase velocity C1 共upper four lines corresponding to the left y axis兲,
共␣ ␥i + ␤兲S
2
and transmission per wavelength 共lower four lines corresponding to the right
y axis兲 of the Young wave, as computed from Womersley’s elastic tube
The Lamb wave does not exist this time and only the Young model 共Ref. 6兲, Cox’s model 共Ref. 40兲, the present free-tube model, and the
wave can survive in the tethering case. present tethered-tube model, where b̂ = 1.15, M = 1.0417, ␤ = 37 237, and
Let us first study the problem of wave propagation ␥ = 0 for a pure elastic tube. The present model agrees with Cox’s model
very well.
within an initially stress-free tube 共i.e., the initially circum-
ferential pressure Q̂ = 0 and the longitudinal stretch ratio S
= 1兲. The dispersion equation 共86兲 for a free tube without The wave speed and the transmission per wavelength from
tethering 共case 1兲 can be simplified as follows: Womersley’s thin-walled tube model6 are given by
C0关Re共␰兲兴−1 and exp关2␲ Im共␰兲 / Re共␰兲兴, respectively, where
A1⬘k̂4 + A2⬘k̂2 + A3 = 0,
冑
共90兲
␺1 + 冑␺21 − 3␺2/4
where ␰= , 共93兲
2
A1⬘ = 3M共␣2␥i + ␤兲共1 − b̂2兲, 共91兲

3␣4␭J0共␭兲b̂2
␺1 =
3
4共1 − F10兲
+
␺3 1
+ ,
2 4
冉 冊 ␺2 =
1 + 2␺3
1 − F10
− 1, 共94兲
A2⬘ = ␣4关1 − M共1 − b̂2兲兴 − . 共92兲
2关J1共␭兲 − ␭J0共␭兲/2兴
共b − a兲␳w 2J1共␭兲
␺3 = , F10 = , 共95兲
For a constrained tube in case 2, ␹ in Eq. 共88兲 is zero in the a␳ f ␭J0共␭兲
absence of initial stresses.
and
B. Results and discussions on the first-order solution
To check our theory, we first compare results with those
C0 = 冋 E共b − a兲
2a␳ f
册 1/2
共96兲

obtained by Womersley6 and Cox40 for wave propagation in
blood flow through a purely elastic tube without initial
stresses, as shown in Fig. 2. For the case shown in this fig-
ure, the physical inputs are4 a = 2 ⫻ 10−3 m, b̂ = b / a = 1.15,
␳ f = 1.056⫻ 103 kg/ m3, ␳w = 1.1⫻ 103 kg/ m3 共or M
= 1.0417兲, ␮ f = 3.3⫻ 10−3 Pa s, ␮w = 0, and Gw = 106 Pa. Al-
though the physical problem itself does not depend on the
tube length 共or the parameter ⑀ = a / L, as long as it is much
smaller than unity兲, we nevertheless need to specify this
quantity as required by normalization. From here onward in
this paper, a typical small value of ⑀ = 0.01 will be used. We
reiterate that any value used for ⑀ will affect only the dimen-
sionless, but not the physical, form of the solution. Substi-
tuting the values into Eq. 共40兲, we get ␤ = 37 237.
Following Cox,40 we show in Fig. 2 the phase velocity
of the Young wave C1 and the transmission per wavelength
given by exp共2␲k̂i1 / k̂r1兲 as functions of the Womersley num-
ber ␣, which represents the frequency as given in Eq. 共40兲.
is the Moens–Korteweg formula,18 in which E is Young’s
modulus, a and b are the inner and outer radii of the tube,
and ␭ = ␣i3/2. For incompressible material, E = 3Gw. As
clearly seen in the figure, the present free-tube model gener-
ates results for the wave propagation speeds that are essen-
tially the same as those by Cox’s model in this case of a pure
elastic tube. Both these models are on the basis of a thick-
walled tube, yielding the phase velocity and the transmission
per wavelength being smaller than those of Womersley’s
thin-walled tube model. The agreement here suggests that, as
far as the first-order oscillatory motion is concerned, our
model is as good as Cox’s model when applied to small-
amplitude wave propagation through a Newtonian fluid con-
tained in an elastic tube of finite wall thickness. Cox did not
apply the long-wave approximation in his model, and his
dispersion equation was so lengthy that he had to solve it by
an iterative approach. With the long-wave assumption, our
model has a much simplified dispersion equation that can be
051901-11 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

1 12 1
14

Transmission per wavelength

Transmission per wavelength
10

Phase velocity C1 (m/s)

C1 (m/s)
12 0.8 0.8

10 8

Phase velocity C
0.6 0.6
8
6

6 0.4 0.4
4
4
0.2 0.2
2
2
(a) ^b = 1.15 (b) ^b = 1.10
0 0 0 0
2 4 6 8 10 2 4 6 8 10
α α
1 4 1

Transmission per wavelength

Transmission per wavelength

C1 (m/s)
C1 (m/s)

0.8 0.8
3
6

Phase velocity C
Phase velocity C

0.6 0.6

2
4
0.4 0.4

1
2 0.2 0.2

(c) ^b = 1.05 (d) ^b = 1.01

0 0 0 0
2 4 6 8 10 2 4 6 8 10
α α
FIG. 3. Phase velocity C1 共upper two lines corresponding to the left y axis兲, and transmission per wavelength 共lower two lines corresponding to the right y
axis兲 of the Young wave, as computed from Womersley’s elastic tube model 共Ref. 6兲 共dashed lines兲 and the present free-tube model 共solid lines兲 for tube
thickness b̂ = 1.15, 1.1, 1.05, 1.01, where M = 1.0417, ␤ = 37 237, and ␥ = 0. The present thick-wall model works equally well in the limit of a thin wall.

solved directly and analytically. The present comparison be- r̂1共R̂ = 1,Ẑ = 0兲 = D cos共t̂ − ␾兲, 共97兲
tween the two models reveals that over the frequency range
␣ ⱕ O共10兲 in blood flow, the long-wave condition is indeed
where D is the amplitude and ␾ is the phase lag between the
well satisfied. When the tube is constrained, the Young wave
forcing and the wall response. Figures 4 and 5 illustrate how
speed will become faster and the wave damping will be
D and ␾ vary with the frequency parameter ␣ for several
stronger as shown in Fig. 2.
values of the wall elasticity and viscosity in the absence of
Although a thick-walled tube model is being considered
initial stresses and surrounding tissues. The results are based
here, the present model works equally well in the limit of a
on a typical value of fluid pressure amplitude equal to
thin wall, i.e., it should approach Womersley’s thin-walled
20 mm Hg,41 which corresponds to po = 20⫻ 133.322
tube model when the free-tube wall becomes very thin. This
⫻ 共␳ f ␴2L2兲−1 / ⑀ = 10 343␣−4, where po is the first-order pres-
tendency is illustrated by Fig. 3, where we show that as the
sure amplitude as defined in Eq. 共76兲. As shown in Fig. 4共a兲,
outer radius b̂ = b / a decreases toward unity 共i.e., the inner the wall displacement amplitude D can become very large
radius兲, the results of the two models are increasingly in when the elasticity parameter ␤ is small for a soft wall, as is
agreement with each other. As in Fig. 3共d兲, the two models expected. In contrast, as shown in Fig. 5共a兲, the wall dis-
give practically the same results when b̂ = 1.01. By and large, placement D tends to be not affected by the wall viscosity ␥
Womersley’s model turns out to be of good accuracy when until ␣ is large, by which D will only be modestly lowered
共b − a兲 / a ⬍ 0.05, or the wall is less than 5% as thick as the by higher ␥. By and large, Figs. 4共a兲 and 5共a兲 show that for
inner radius. Also, Fig. 3 shows that the wave velocity de- oscillation under a fixed amplitude of the pressure forcing,
creases 共or the wavenumber k̂r increases兲 dramatically in re- the wall radial displacement D is rather insensitive to change
sponse to a small decrease in the tube thickness. In the lim- in the frequency ␣. Unlike the amplitude D, the phase lag ␾
iting case when the thickness approaches zero, the wave is strongly affected by the wall viscoelasticity as well as the
speed tends to zero as well or the wavenumber goes to in- frequency; see Figs. 4共b兲 and 5共b兲. We remark that the de-
finity. The transmission per wavelength is on the contrary pendence on ␣ will qualitatively change if the oscillation is
much insensitive to change in the wall thickness. volume cycled 共i.e., under a fixed tidal volume兲.
The wall deformation, particularly in the radial direction, We further show in Fig. 6 the cross-sectional profiles of
is of practical interest. The radial displacement r̂1 of the in- the axial and radial displacements at time instants within half
terface R̂ = 1 at the origin Ẑ = 0 can be expressed as an oscillation period: t̂ = 0 , ␲ / 6 , . . . , ␲, at the origin Ẑ = 0 for
051901-12 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

(a) 15 (b) 0.5

0.4

10 0.3

D φ 0.2
5 0.1

0 -0.1
2 4 6 8 10 2 4 6 8 10
α α
FIG. 4. 共a兲 The amplitude of radial displacement and 共b兲 the phase lag of the Young wave on the fluid-wall interface at Ẑ = 0 as functions of the Womersley
number ␣, where b̂ = 1.15, M = 1.0417, ␥ = 10, and ␤ = 2000 共solid line兲, 5000 共dashed line兲, and 20 000 共dot-dashed line兲. D is rather insensitive to change in
the frequency ␣ but is sensitive to change in the elasticity ␤.

(a) 2 (b) 0.5

0.4
1.5
0.3

D 1 φ 0.2
0.1
0.5
0

0 -0.1
2 4 6 8 10 2 4 6 8 10
α α
FIG. 5. 共a兲 The amplitude of radial displacement and 共b兲 the phase lag of the Young wave on the fluid-wall interface at Ẑ = 0 as functions of the Womersley
number ␣, where b̂ = 1.15, M = 1.0417, ␤ = 20 000 共normal blood vessel兲, and ␥ = 0 共pure elastic兲 共solid line兲, 10 共dashed line兲, and 100 共dot-dashed line兲. D
tends to be not affected by the wall viscosity ␥ until ␣ is large.

(a) (b)
1 1
^t = π 5π/6 2π/3 π/2 π/3 π/6 0
0.8 0.8

0.6
R^
0.6
R^
0.4 0.4

^t = 0 π π/6 5π/6 π/3 2π/3 π/2

0.2 0.2

0 0
-2 0 2 4 6 8 -1 0 1
^z ^r
1 1
FIG. 6. Cross-sectional profiles of 共a兲 the axial and 共b兲 the radial displacements of the Young wave at time instants in half a period when the tube is free of
tethering at Ẑ = 0, where b̂ = 1.15, M = 1.0417, ␣ = 5, ␥ = 10, and ␤ = 20 000. The axial displacement amplitude is the largest at the tube center and the smallest
in the tube wall; the radial displacement is the smallest at the tube center and the largest near the tube wall.
051901-13 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

(a) 10 (b) 0.8

Transmission per wavelength

8
0.6

Phase velocity C1 (m/s)

6

0.4

4
q =0 q =0
q = 1000 q = 1000
q = 2000 0.2 q
2 = 2000
q = 3000 q = 3000
q = 4000 q = 4000
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
α α
FIG. 7. 共a兲 Phase velocity C1 and 共b兲 transmission per wavelength of the Young wave as functions of the Womersley number ␣ for a free elastic tube, where
the initial pressure q varies from 0 to 4000, and the longitudinal stretch ratio S = 1. The phase velocity and the wave transmission of Young wave decrease with
increasing q.

moderate frequency ␣ = 5 and large elasticity ␤ = 20 000. The stresses on the wave properties according to the dispersion
axial displacement amplitude is the largest at the tube center equation 共86兲. Figures 7–10 show the influences of the initial
and the smallest in the tube wall. As discussed earlier, the mean pressure q = ␣4 p̂ f0 = 共a4 / ␳ f L2␯2f 兲Q and the longitudinal
axial displacement of the wall has only a weak dependence stretch ratio S on the wave speeds and the transmission per
on R̂, and is hence nearly uniform across the wall section. wavelength for the Young and Lamb waves along a purely
For this high value of rigidity 共␤ = 20 000兲, the axial wall elastic tube 共␥ = 0兲. The static pressure q is so defined that the
displacement is many times smaller than the maximum axial dependence on frequency is removed from the normaliza-
fluid displacement that happens along the tube axis. In sharp tion. For a normal blood vessel, Q ⯝ 150 mm Hg and
contrast, the radial displacement amplitude is the smallest S ⯝ 1.44.18 Therefore, the following ranges of values have
共equal to zero兲 at the tube center and the largest near the tube been considered in our calculations: q = 0 – 4000 and
wall. As a result, the radial wall displacement is comparable S = 1 – 1.6. While a detailed comparison of results with the
to the maximum radial fluid displacement that happens near literature is impossible owing to basic differences in the
the wall. When the tube becomes tethered, the velocity pro- modeling, a qualitative comparison of the effects is neverthe-
files are largely the same as those when the tube is free. The less possible. The qualitative effects of the two parameters q
major difference is in the axial wall displacement, which is and S on the wave properties are summarized in Table I, as
much closer to zero because of the large longitudinal con- predicted by the present model, and the models by Atabek,14
straint by the surrounding tissues outside the tube. Some pre- and Kuiken.19 One can find that the present predictions are
vious authors26,27 have assumed a priori in their problem largely in accord with those by Atabek. One difference is in
formulations that the wall would undergo deformation purely the effect of the mean pressure on the Lamb wave speed. As
in the radial direction with identically zero axial displace- in Fig. 8共a兲, the present model gives that the wave speed C2
ment across the wall. Our approach to the tethering case is, may increase or decrease with increasing q depending on the
however, different from these previous works. Here the axial frequency ␣ being small or large. Atabek’s model predicts
displacement of the wall, although much smaller than that of that the wave speed always increases with increasing q irre-
the fluid, is not exactly zero across the wall layer because of spective of the frequency. This inconsistency in prediction,
ẑw12共R̂兲 ⫽ 0 as defined in Eq. 共53兲, which in fact is important which is relatively minor, can be ascribed to the different
to the stress distribution in the wall. constitutive equations being applied in the present and
All the discussions so far have centered on an initially Atabek’s models. Kuiken’s model predicts the same effects
unstressed tube. Here, let us examine the effects of the initial of the mean pressure q but completely opposite effects due to

TABLE I. The effects of initial stresses on the wave propagation by 共a兲 the present model, 共b兲 Atabek 共Ref. 14兲, and 共c兲 Kuiken 共Ref. 19兲.

Transmission Transmission
C1 of Young wave C2 of Lamb wave

As q increases 共a兲 Decreases 共a兲 Decreases 共a兲 Increases for small ␣, decreases for large ␣ 共a兲 Increases
共b兲 Decreases 共b兲 Decreases 共b兲 Increases 共b兲 Increases
共c兲 Decreases 共c兲 Decreases 共c兲 Increases 共c兲 Increases

As S increases 共a兲 Decreases 共a兲 Increases 共a兲 Decreases 共a兲 Decreases

共b兲 Decreases 共b兲 Increases 共b兲 Decreases 共b兲 Decreases
共c兲 increases 共c兲 Decreases 共c兲 Increases 共c兲 Increases
051901-14 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

(a) 35 (b)0.8
q =0
q = 1000

Transmission per wavelength

q = 2000

Phase velocity C 22 (m/s)

0.6
q = 3000
30 q = 4000

0.4

25 q =0
q = 1000
q = 2000 0.2
q = 3000
q = 4000
20
0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
α α
FIG. 8. 共a兲 Phase velocity C2 and 共b兲 transmission per wavelength of the Lamb wave as functions of the Womersley number ␣ for a free elastic tube, where
the initial pressure q varies from 0 to 4000, and the longitudinal stretch ratio S = 1. The wave phase velocity of Lamb wave increases for small ␣ but decreases
for large ␣, and the wave transmission increases with increasing q.

(a) 10 (b)
0.8
Transmission per wavelength

8
Phase velocity C11 (m/s)

0.6
6

0.4
4
S =1 S =1
S = 1.2 S = 1.2
S = 1.4 0.2 S = 1.4
2
S = 1.6 S = 1.6

0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
α α
FIG. 9. 共a兲 Phase velocity C1 and 共b兲 transmission per wavelength of the Young wave as functions of the Womersley number ␣ for a free elastic tube, where
the longitudinal stretch ratio S varies from 1 to 1.6, and the initial pressure q = 0. The phase velocity of Young wave decreases and the wave transmission
increases with increasing S.

(a) 35 (b) 0.8

S =1
S = 1.2
Transmission per wavelength

30 S = 1.4
0.6
Phase velocity C2 (m/s)

S = 1.6

25
0.4

20

S =1 0.2
15 S = 1.2
S = 1.4
S = 1.6
10 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
α α
FIG. 10. 共a兲 Phase velocity C2 and 共b兲 transmission per wavelength of the Lamb wave as functions of the Womersley number ␣ for a free elastic tube, where
the longitudinal stretch ratio S varies from 1 to 1.6, and the initial pressure q = 0. The phase velocity and the wave transmission of Lamb wave decrease with
increasing S.
051901-15 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

the longitudinal stretch ratio S. As mentioned earlier, Kuik-
en’s theory has been questioned by others for the inconsistent
use of stress terms.
Table I also presents a contrast in the roles played by q
and S. In particular, the initial pressure tends to diminish the
Young wave but to enhance the Lamb wave, as can be in-
ferred from the wave transmission. Remarkably, opposite ef-
fects will result from an initial extension of the tube. On
comparing Figs. 7共a兲 and 9共a兲, one finds that the Young wave
velocity changes more appreciably with respect to the initial
pressure than with respect to the stretch ratio within the
ranges of values specified above. In contrast, on comparing
evance to the HFV, in which gases fill the tube. To examine
the velocity and wall displacement fields in HFV, the follow-
ing physical values are chosen, following Dragon and
Grotberg:5 a = 9 ⫻ 10−3 m, b̂ = 1.056, ⑀ = 0.1, ␳f
= 1.14 kg/ m , ␳w = 9.97⫻ 102 kg/ m3 共M = 874.56兲, ␯ f = 1.5
3
⫻ 10−5 m2 / s, ␯w = 4.5⫻ 10−4 m2 / s, and Gw = 138.5 Pa for
normal tracheae 共␤ = 500, ␥ = 0.3兲. Different from the blood
circulation, HFV requires a fixed leading-order tidal volume
VT at a given axial position, say, Z = 0, in one half-cycle, and
d = VT / ␲a2 is the stroke length. The dimensionless flow rate
can be expressed by5

Figs. 8共a兲 and 10共a兲, one finds that the Lamb wave velocity is
more affected by the stretch ratio than by the initial pressure.
As noted above, the Lamb wave is essentially associated
with longitudinal motion only, and is hence more responsive
to the initial axial deformation. The effects shown in these
figures remain qualitatively the same for a viscoelastic tube,
when ␥ ⬎ 0. For a constrained elastic tube, the initial stresses
will have no effect on the wave damping since ␹ is real and
the initial stress terms are canceled out on evaluating k̂i / k̂r by
Eq. 共88兲. The influence of the initial pressure on the Young
wave speed is quite similar to the free-tube case as shown in
Fig. 7共a兲. However, in the absence of the mean pressure,
q = 0, it follows from the dispersion equation, Eq. 共88兲, that
the Young wave velocity in a constrained tube is independent
of the initial stretch S, which is different from what is shown
in Fig. 9共a兲 for a free tube. Even when the constrained tube is
viscoelastic 共␹ becomes a complex number兲, the wave at-
tenuation and the phase velocity can only be weakly affected
by the initial stresses and by the initial extension of the tube,
respectively.
As remarked earlier, the present study is also of rel-
冕
1
A
żof R̂dR̂ = , 共98兲
0 4
where A = d / a is a nondimensional amplitude parameter, re-
lating the stroke length to the undisturbed inner tube radius.
In this study, A is fixed at 5. Dragon and Grotberg assumed
comparable length scales for the radial and axial wall dis-
placements, which is a condition that holds only when the
tube is tethered. Therefore, the present constrained-tube
model 共case 2兲 will be used when we do the comparison
below. However, different from blood flow in arteries, the
elasticity of tracheae can reach a small value, say, ␤ = 50,
while the oscillation frequency can be as high as ␣ = 15. Un-
der this rather extreme condition, the radial acceleration term
⑀2⳵2r̂w / ⳵t̂2 in Eq. 共32兲 becomes comparable to the elastic
stress terms, or ⑀2␣4 / ␤ = O共1兲, and the term should be taken
into account in the leading-order analysis. Therefore, when
the present tethered tube model 共case 2兲 is applied to the
pulmonary gas flow during HFV, the dispersion equation
共88兲 should be changed to

␣4␭J0共␭兲
k̂2 = − . 共99兲
M共␣2␥i + ␤兲关J1共␭兲 − ␭J0共␭兲/2兴关2共1 − 1/b̂2兲 − ⑀2␣4 ln共b̂兲/共␣2␥i + ␤兲 + ␹兴

Figure 11 shows a comparison of the amplitude D and

the phase ␾, as defined in Eq. 共97兲, of the radial displace-
ment of the wall-fluid interface R̂ = 1 at Ẑ = 0 as functions of
k̂o = 冋 2J0共␭兲 R20␣4
J2共␭兲 M̃共R2␣4 − iR2共2˜␥兲␣2 − ␬/0.75兲
0 0
册 1/2
共101兲

the Womersley number and the elasticity, where ␥ = 0.3, be-

and
tween the present model and Dragon and Grotberg’s model.
We recall that the latter model is based on the thin-wall as-
b−a ␳w共b − a兲 ␯w
sumption, by which the wall is merely represented by a solid R0 = 1 + , M̃ = , ˜␥ = ,
surface. By their model,5 D and ␾ are given by 兩␵兩 and 2a ␳fa 2␯ f
tan−1关−Im共␵兲 / Re共␵兲兴, respectively, where 共102兲
Ea2
␭ = ␣i3/2, ␬= ,
␳w␯2f
A
␵= k̂o , 共100兲
4 where E is Young’s modulus. As was assumed by Dragon
and Grotberg,5 the initial stresses are ignored in this particu-
in which lar example.
051901-16 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

6 1.6

4.5 1.2

D 3
φ 0.8
β = 50 β = 50
1.5 0.4

0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14
2 0.8

1.5 0.6

D 1 φ 0.4

0.5
β = 500 0.2 β = 500

0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14

0.8 0.8

0.6 0.6

D 0.4 φ 0.4
β = 5000
0.2 0.2 β = 5000

0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14
α α
FIG. 11. Radial wall displacement on the interface at Ẑ = 0 as computed from the present model 共solid line兲 and from Dragon and Grotberg’s model 共Ref. 5兲
共dashed line兲 as a function of the Womersley number ␣ for elasticities ␤ = 50, 500, 5000, where b̂ = 1.056, M = 874.56, and ␥ = 0.3. The two models agree very
well with each other in the case of strong elasticity at low frequencies but give rather different values for weak elasticity at high frequency.

It is clear from Fig. 11 that, for sufficiently strong elas-
ticity such that ␤ ⱖ 500, the results by the two models agree
very well with each other. Some differences, however, ap-
pear in the case of weak elasticity, ␤ = 50, when the Womer-
sley number is so high that ␣ ⱖ 6. Essentially, Dragon and
Grotberg’s model predicts that D will reach a local maximum
at a certain frequency, corresponding to the occurrence of
resonance. The resonance frequency is lower for smaller
elasticity ␤. Therefore, for elasticity as weak as ␤ = 50, the
resonance may occur at a frequency that is within the range
of ␣ being considered. However, the resonance frequency
and radial displacement predicted by the present model are
slightly different from those by Dragon and Grotberg.5 The
differences are possibly due to one basic difference between
the two models. The radial displacement r̂wo in Eq. 共81兲 of the
present model is a function of the radial coordinate R̂ whose
radial gradient is responsible for the normal stress, while the
thin-wall model by Dragon and Grotberg5 can only admit
constant radial wall deformation. The disparities will be en-
hanced as the wall thickness increases and further magnified
under the occurrence of resonance. On the other hand, the
resonance could happen only when the radial wall accelera-
tion is comparable to the elastic stresses, which can be sat-
isfied for pulmonary gas flow during HFV. For blood flow in
arteries, the elasticity is so high that the resonance frequency
is far beyond the range of ␣ being considered. Hence, for
blood flow, it is reasonable to neglect the radial acceleration
term as in previous works.6,18
Figure 12 presents another comparison between the
present and Dragon and Grotberg’s models, in which D and
␾ are shown as functions of the Womersley number ␣ and
the tube wall viscosity ␥, and the elasticity ␤ = 500 for nor-
mal tracheae. For a pure elastic tube 共␥ = 0兲, the two models
give results very close to each other. As ␥ increases, differ-
ences in the results at high frequencies are growingly exhib-
ited, especially for the phase lag ␾. The difference in the
effects of the wall viscosity on the wall motion can be related
to the fact that the wall damping has been accounted for in
different manners by the two models. This comparison shows
that the difference is intensified as the wall viscosity in-
creases. The effects of the viscous damping will be further
discussed in the second-order analysis below.
Owing to the small displacements, the Lagrangian prob-
lem gives the same solution as the Eulerian problem at the
first order. This is no longer true at higher orders. At the
second order, the time-averaged Lagrangian problem can be
solved directly for the Lagrangian streaming velocity, which
differs from the Eulerian streaming velocity by the Stokes
drift. In biofluid problems, substances are transported in the
fluid mainly by the Young wave mode. Hence, the Lamb
wave will be ignored in the following second-order analysis.
V. SECOND-ORDER PROBLEM
A. Second-order solution
At O共⑀2兲, the governing equations for the fluid flow are
⳵ ẑ f2 ⳵ r̂ f2 r̂ f2 r̂2f1 ⳵ r̂ f1 ⳵ ẑ f1 ⳵ ẑ f1 ⳵ r̂ f1
+ + = + − , 共103兲
⳵ Ẑ ⳵ R̂ R̂ R̂2 ⳵ Ẑ ⳵ R̂ ⳵ Ẑ ⳵ R̂
051901-17 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

2 0.8

1.5 0.6

D 1 φ 0.4
γ=0 γ=0
0.5 0.2

0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14
2 0.8

1.5 0.6

D 1 φ 0.4
γ=1 γ=1
0.5 0.2

0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14
2 0.8

1.5 0.6

D 1 φ 0.4
γ = 10
0.5 0.2
γ = 10
0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14
α α
FIG. 12. Radial wall displacement on the interface at Ẑ = 0 as computed from the present model 共solid line兲 and from Dragon and Grotberg’s model 共Ref. 5兲
共dashed line兲 as a function of the Womersley number ␣ for tube viscosity ␥ = 0 , 1 , 10, where b̂ = 1.056, M = 874.56, and ␤ = 500. The two models give results
very close to each other for weak wall viscosity, but the differences at high frequencies are growingly exhibited as ␥ increases.

⳵ p̂ f2 ⳵ ẑ f1 ⳵ p̂ f1 ␣4 ⳵ ẑw2 ⳵ p̂w2 ⳵ ␶ˆ wzr2 ␶ˆ wrz2 ⳵ ␶ˆ wzz2

= , 共104兲 + − − −
⳵ R̂ ⳵ R̂ ⳵ Ẑ ␤ ⳵ t̂2 ⳵ Ẑ ⳵ R̂ R̂ ⳵ Ẑ
⳵ ẑw11 ⳵ p̂w1 ⳵ r̂w1 ⳵ p̂w1 ⳵ r̂w1 ⳵ ␶ˆ wzr1 r̂w1
= + − − ␶ˆ wrz1

冉 冊
⳵ Ẑ ⳵ Ẑ ⳵ Ẑ ⳵ R̂ ⳵ R̂ ⳵ R̂ R̂2
⳵ ẑ f2 ⳵ p̂ f2 ⳵ ␶ˆ fzr2 ␶ˆ fzr2
␣2 + − − ⳵ ẑw11 ⳵ ␶ˆ wzz1 ⳵ r̂w1 ␶ˆ wzz1
⳵ t̂ 2
⳵ Ẑ ⳵ R̂ R̂ − − , 共108兲
⳵ Ẑ ⳵ Ẑ ⳵ Ẑ ⳵ R̂
⳵ ẑ f1 ⳵ p̂ f1 ⳵ r̂ f1 ⳵ ␶ˆ fzr1 ⳵ ẑ f1 ⳵ ␶ˆ fzr1 r̂ f1
= ␣2 − − − ␶ˆ frz1 . 共105兲
⳵ Ẑ ⳵ Ẑ ⳵ R̂ ⳵ R̂ ⳵ R̂ ⳵ Ẑ R̂2
where the stress components ␶ˆ lij2 are given in Appendix D.
Our interest here is to find the steady streaming in the
To simplify the problem, let us for the time being consider a fluid, which can be obtained by taking the time average of
tube without initial stresses, for which the governing equa- the above equations. On the substitution of Eq. 共77兲, the time
tions at O共⑀2兲 can be written as averaging of Eqs. 共103兲 and 共106兲 yields

⳵ ẑw21 ⳵ r̂w2 r̂w2 2

r̂w1 ⳵ ẑw11 ⳵ r̂w1 ⳵¯ẑl2 ⳵¯r̂l2 ¯r̂l2
+ + = − , 共106兲 ˆ ˆ
+ + = Fle2kiZ , 共109兲
⳵ Ẑ ⳵ R̂ R̂ R̂ 2
⳵ Ẑ ⳵ R̂ ⳵ Ẑ ⳵ R̂ R̂

⳵ p̂w2 ⳵ ␶ˆ wrr2 ␶ˆ wrr2 − ␶ˆ w␪␪2 where

− −
⳵ R̂ ⳵ R̂

再 冎
R̂

= 冉 ⳵ r̂w1
⳵ R̂
−
r̂w1
R̂
冊冉 ␶ˆ wrr1 − ␶ˆ w␪␪1
R̂
冊 , 共107兲 Ff =
Re 兩rof 兩2
2 R̂2
+ ik̂ⴱroⴱ
f
dzof
dR̂
− ik̂ ⴱ oⴱ
z f
drof
dR̂
, 共110兲
051901-18 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

Fw =
2 R̂2
再
Re 兩rwo兩2
− ik̂ⴱzwc
o
oⴱ drw

dR̂
. 冎 共111兲
s
are functions of R̂ and t̂, and zwc is a function of t̂ only, where
time dependence here is nonperiodic. Before solving for the
steady streaming in the fluid, we first determine the steady
The overbar denotes time average over one wave period and motion of the tube.
the asterisk denotes the complex conjugate. The time- The continuity equation for the time-mean velocity is
averaged momentum equations are obtainable by first differentiating Eqs. 共103兲 and 共106兲 with
respect to time followed by time averaging over a period:
⳵ ¯p̂ f2 ˆ ˆ
= ⌫e2kiZ , 共112兲
⳵ R̂ ⳵¯ẑ˙l2 ⳵¯r̂˙l2 ¯r̂˙
l2
+ + = 0. 共119兲
⳵ Ẑ ⳵ R̂ R̂
⳵ ¯p̂ f2 ⳵2¯ẑ˙ f2 1 ⳵¯ẑ˙ f2 ˆ ˆ Substituting Eq. 共118兲 into Eqs. 共109兲, 共114兲, 共115兲, and
␣2 − − = He2kiZ , 共113兲
⳵ Ẑ ⳵ R̂ 2
R̂ ⳵ R̂ 共119兲 yields the following equations for the axial time-mean

冋 冉 冊册
displacement at Ẑ = 0:

⳵ ¯p̂w2
⳵ R̂
−2
␣2␥ ⳵2¯r̂˙w2 1 ⳵¯r̂˙w2 ¯r̂˙w2
␤
+
⳵ R̂2 R̂ ⳵ R̂
−
R̂
␤ 冉
␣2␥ d2żswv 1 dżswv
dR̂ 2
+
R̂ dR̂
+ 4k̂2i żwc
s
+
d2zswv 1 dzswv
dR̂ 2
+
R̂ dR̂
冊
+ 4k̂2i zwc
s

−2 冋 ⳵2¯r̂w2
⳵ R̂2
+
R̂
冉
1 ⳵¯r̂w2
⳵ R̂
−
¯r̂w2

R̂
冊册 = Ie 2kˆiZˆ
, 共114兲
= K,

where K = K共R̂兲 is given by

共120兲

and 冉 冕 冊
冋
共121兲

冉 冊
K = 2k̂i Fw + IdR̂ + M 0 − P.

⳵ ¯p̂w2 ␣2␥ ⳵2¯r̂˙w2 ⳵2¯ẑ˙w22 1 ⳵¯r̂˙w2 ⳵¯ẑ˙w22 Equation 共120兲 can be readily solved to give the general
− + + +
⳵ Ẑ ␤ ⳵ R̂ ⳵ Ẑ ⳵ R̂2 R̂ ⳵ Ẑ ⳵ R̂ solutions of the steady axial velocity profile:

册冋 冉 冊
2
s
zwc = M 1关1 − e−␤t̂/␣ ␥兴 共122兲
⳵2¯ẑ˙w21 ⳵2¯r̂w2 ⳵2¯ẑw22 1 ⳵¯r̂w2 ⳵¯ẑw22
+2 − + + + and
⳵ Ẑ2 ⳵ R̂ ⳵ Ẑ ⳵ R̂2 R̂ ⳵ Ẑ ⳵ R̂

册
2
zswv = N关1 − e−␤t̂/␣ ␥兴, 共123兲
⳵ ẑw21
2¯
2kˆiZˆ
+2 = Pe , 共115兲
⳵ Ẑ2 where N = N共R̂兲 is given by

R̂2

再 冎
where N = 共K − 4k̂2i M 1兲 + M 2 ln共R̂兲 + M 3 , 共124兲
4
Re dzof
⌫= ik̂ⴱ Poⴱ
f , 共116兲 where M 0 – M 3 are the constants of integration. Equations
2 dR̂
共122兲 and 共123兲 show that the time-mean axial velocity ¯ẑw21

冉 冊
冦
dzof and ¯ẑw22 will die out exponentially and tend to the limits M 1

冏 冏
d R̂zoⴱ
f and N, respectively, over a timescale ␣2␥ / 共␤␴兲, or ␮w / Gw.
Re 2 2 oⴱ o 2k̂ⴱ dR̂ dzof 2
The time-mean radial velocity rws has similar exponential de-
H= ␣ 兩k̂兩 p f z f − − k̂
2 R̂ dR̂ dR̂ cay with time. The timescale ␮w / Gw is the so-called relax-
ation time, which characterizes the rate of creep of a Voigt

冊冧
body in response to a unit-step force. For a viscoelastic ma-

+i 冉 droⴱ
f

dR̂
−
roⴱ
f

R̂
冊冉 1 dzof
R̂ dR̂
−
d2zof
dR̂2
, 共117兲
terial with sufficient strength such that Gw / ␮w ⱖ O共␴兲, which
implies that the time for the transient motion is much shorter
than the wave period, the time-mean motion of the tube is
short lived, as the ultimate mean deformation can be readily
and I and P are two complicated functions of R̂ containing
attained within several wave periods or so.31 Under this con-
terms of products of first-order quantities.
dition, the transient motion is of no practical significance,
According to the forcing terms on the right-hand side of
and we may consider an instant of time that is long enough
the time-averaged equations 共109兲–共115兲, the second-order
for the steady motion of the wall to have already stopped. In
solutions of the steady motion in both the fluid and wall can
biofluid systems, the elasticity of the tissue is typically
be expressed as follows:
strong enough so that the assumption Gw / ␮w ⱖ O共␴兲 can be
¯˙ ,ẑ
¯˙ ¯ ¯ ¯ ¯ ¯ satisfied in general. One may check that this condition is
共r̂ f2 f2,p̂ f2,r̂w2,ẑw21,ẑw22,p̂w2兲
indeed satisfied using the physical input values that have
ˆ ˆ
= 共URs,ULs,psf ,rws,zwc
s
,zswv,pws兲e2kiZ , 共118兲 been earlier introduced for the blood flow and HFV systems.
For an initially stressed tube, the above argument remains
where URs, ULs, and psf are functions of R̂ only, rws, zswv, and pws valid. Based on this assumption, the boundary conditions for
051901-19 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

the steady streaming of the fluid can be written as Substitution of Eq. 共118兲 into Eqs. 共109兲–共113兲 and the
boundary conditions 共125兲 and 共126兲, the time-mean pressure
¯˙
dẑ and axial steady streaming velocity at Ẑ = 0 can be readily
¯r̂˙ = 0, f2
=0 on R̂ = 0 共125兲
f2 obtained as

冕
dR̂

and psf = ⌫dR̂ + M 4 , 共128兲

¯r̂˙ = 0,
f2
¯ẑ˙ = 0
f2 on R̂ = 1. 共126兲
ULs = 冕 冕Rˆ

1
1
R̂ 0
Rˆ
共2k̂i␣2 psf − H兲R̂dR̂dR̂, 共129兲
Integrating Eq. 共119兲 over the cross section while making use
of the boundary condition 共126兲 gives where the integration constant M 4 can be determined from
URs by substituting Eq. 共129兲 into Eq. 共119兲 which is then
¯˙ 共1兲 integrated with respect to R̂ with the boundary conditions
¯˙ 典 = − r̂ f2 = 0,
具ẑ 共127兲
f2 共125兲 and 共126兲:
2k̂i

where the angular brackets denote the cross-sectional aver- URs兩Rˆ=1 =

1
R̂
冕 0
1
共− 2k̂iULs兲R̂dR̂ = 0. 共130兲
age. This result shows that the net Lagrangian time-mean
flux across a section will be identically equal to zero when Substituting F, ⌫, and H into Eq. 共129兲 results in an expres-
the time-mean motion of the wall dies out. sion for ULs as follows:

再
ULs = Re ␭B1 poⴱ 冉 1 2
2
冊
兩k̂兩 i + k̂ⴱ2i + k̂ik̂ⴱ 关J0共␭R̂兲 − J0共␭兲兴 + ␣2兩B1兩2k̂ⴱ 冕1
Rˆ
Jⴱ0共␭R̂兲
⳵ J0共␭R̂兲
⳵ R̂
dR̂ 冎 1
+ k̂r␣4兩B1兩2
2

冕 冕 冕 冕冏 冏 2
Rˆ 1 Rˆ 1 Rˆ 1 Rˆ 2J1共␭R̂兲 1
⫻ 兩J1共␭R̂兲兩2R̂dR̂dR̂ + k̂i␣2兩B1兩2 ␭J0共␭R̂兲 − R̂dR̂dR̂ + k̂i␣2共R̂2 − 1兲共4M 4 − 兩k̂兩2兩po兩2兲, 共131兲
1 R̂ 0 2 1 R̂ 0 R̂ 8

where ␭ = ␣i3/2. The steady Lagrangian drift, which is domi- and Grotberg’s model.5 The dimensionless Eulerian steady
nantly in the axial direction, is now determined. This steady axial velocity w1 in Dragon and Grotberg’s model is given by
current is affected through the parameter B1 by the viscoelas- ˆ ˆ
w1 = 关W共0兲 共0兲ⴱ
1 共R̂兲 + W1 共R̂兲兴e
2koiZ
, 共132兲
ticity and initial stresses of the tube, as given in Eq. 共87兲. Our
second-order analysis and the resulting expression for the where the expression for W共0兲1 共R̂兲 is given in Appendix E and
Lagrangian drift appear not to have been reported in the ex- k̂oi is the imaginary part of k̂o as defined in Eq. 共101兲. The
isting literature. By virtue of the Lagrangian approach, the dimensional Eulerian steady axial velocity ue in Dragon and
boundary conditions are prescribed exactly on the interfaces, Grotberg’s model can be obtained by multiplying w1 with the
and hence the expressions deduced above are good irrespec- velocity scale:
tive of the displacement amplitude of the boundaries. This
⑀␯ f
renders the present model to be superior to an Eulerian ue = w1 . 共133兲
a
model, which is limited by the condition that the amplitude
of the interfacial displacement must be much smaller than To enable the comparison, the Stokes drift is added to the
the Stokes layer thickness; otherwise the Taylor expansion of Eulerian drift42,43 to get the Lagrangian drift uL:
the boundary conditions about the mean position will not uL = ue + us , 共134兲
work.
where the dimensional Stokes drift us can be calculated by

B. Results and discussions

on the second-order solution us =
⑀␯ f
␣ 2a 冋冉冕 t̂
w0dt 冊 冉冕 冊
dw0
dẐ
+
t̂
u0dt
dw0
dr̂
册
Numerical results are presented below to help us look
into the effects of viscoelasticity on the steady streaming.
First, let us compare in the context of HFV the steady axial
= 2 Re 再 冋⑀␯ f
␣a
2 k̂o兩W0共r̂兲兩2 + iUo共r̂兲ⴱ
dW0共r̂兲
dr̂
册冎 ˆ ˆ
e2koiZ , 共135兲

velocity profiles as given by the present model and Dragon in which u0 and w0 are, respectively, the dimensionless first-
051901-20 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

1 1 1
β = 500 β = 500 β = 500
0.8 α=2 0.8 α=6 0.8 α = 10
0.6 0.6 0.6
R^ R^ R^
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
-2 -1 0s 1 2 -4 -2 0 2 4 -10 -5 0 5 10
uL usL usL
1 1 1
β = 50 β = 50 β = 50
0.8 α=2 0.8 α=6 0.8 α = 10
0.6 0.6 0.6
^
R^ R R^
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
-2 -1 0 1 2 -4 -2 0 2 4 -10 -5 0 5 10
usL usL usL
1 1 1
β=5 β=5 β=5
0.8 α=2 0.8 α=6 0.8 α = 10
0.6 0.6 0.6
^ ^ ^
R R R
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
-2 -1 0 1 2 -4 -2 0s 2 4 -10 -5 0 5 10
usL uL usL
FIG. 13. Cross-sectional profiles of steady axial velocity at Ẑ = 0, resulting from the Young wave in the context of HFV, for elasticities ␤ = 5 , 50, 500 and
Womersley numbers ␣ = 2 , 6 , 10, as computed from the present model 共solid lines兲 and from Dragon and Grotberg’s model 共Ref. 5兲 共dashed lines兲, where
b̂ = 1.056, M = 874.56, and ␥ = 0.3. Differences between the two models are similar to those in Fig. 11. The streaming velocity increases in magnitude as ␤
decreases. The velocity profile changes qualitatively when the frequency is varied.

order radial and axial velocities in Dragon and Grotberg’s
model:
ˆ ˆ
共u0,w0兲 = 2 Re兵关U0共R̂兲,W0共R̂兲兴ei共t̂−koZ兲其,
where the expressions for U0共R̂兲 and W0共R̂兲 are also given in
Appendix E. The final step is to normalize uL by the velocity
scale:
ˆ ˆ
uL = 共⑀2L␴兲UL⬘se2koiZ .
Now, UL⬘s given above by Dragon and Grotberg’s model can
be compared with ULs given by Eq. 共131兲 of the present
model. As in Sec. IV, the case of volume-cycled oscillation
共i.e., the tidal volume is fixed when the frequency is varied兲
in a tethered tube without initial stresses is considered for
this comparison in the context of HFV.
Figure 13 shows the steady axial velocity profiles at Ẑ = 0
for different values of the elasticity ␤ and the Womersley
number ␣, which represents frequency as given in Eq. 共40兲.
Again, it can be seen that the present results agree well with
those by Dragon and Grotberg’s model except when ␤ is
small and ␣ is large. As reasoned earlier when discussing
Fig. 11, the difference in the results can be ascribed to the
basic difference in handling the wall shear deformation and
stress between the two models. From this figure, one may see
that in general the streaming velocity will increase in mag-
nitude as ␤ 共or the stiffness of the tube兲 decreases. This is
associated with the fact that a softer tube will move with a
larger amplitude under the same forcing. More remarkably,
the velocity profile will change qualitatively when the fre-
quency is varied. At a low frequency 共␣ ⬇ 1兲, the steady
共136兲 streaming can be in the forward direction near the center of
the tube and is in the backward direction near the wall. The
pattern will be reversed, however, when the frequency be-
comes large enough 共␣ ⬎ 5兲.
The effect of the wall viscosity ␥ on the steady stream-
共137兲 ing velocity is shown in Fig. 14. Again, some differences in
the results of the two models are exhibited when both the
wall viscosity ␥ and the frequency ␣ are large. When this is
examined in conjunction with Fig. 12, one can see that the
wall viscosity has a stronger diminishing effect on the veloc-
ity profile in the present model than in Dragon and Grot-
berg’s model. This difference can be ascribed to the fact that
different constitutive equations have been used in the two
models. In the present work, a Voigt type of constitutive law
for the viscoelastic tube as given in Eq. 共4兲 is part of the
problem formulation. In this constitutive equation, the wall
viscosity is a term contributing to the deviatoric stress com-
ponents. On examining the steady streaming velocity given
by Eq. 共131兲, one may find that the time-mean velocity is
controlled by several parameters 共e.g., B1 , po , k̂兲 which are
all functions of the term ␣2␥i + ␤ as a result of the Voigt
model. The wall viscosity is hence introduced into the
present model through the formal stress-strain rate relations.
Of course, the wall shear stress is one of the components
051901-21 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

1 1 1
γ=0 γ=0 γ=0
0.8 α=2 0.8 α=6 0.8 α = 10
0.6 0.6 0.6
R^ R^ R^
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
-0.2 -0.1 0s 0.1 0.2 -0.4 -0.2 0s 0.2 0.4 -3 -2 -1 0s 1 2 3
uL uL uL
1 1 1
γ=1 γ=1 γ=1
0.8 α=2 0.8 α=6 0.8 α = 10
0.6 0.6 0.6
^
R^ R R^
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
-0.2 -0.1 0 0.1 0.2 -0.4 -0.2 0s 0.2 0.4 -3 -2 -1 0s 1 2 3
usL uL uL
1 1 1
γ = 10 γ = 10 γ = 10
0.8 α=2 0.8 α=6 0.8 α = 10
0.6 0.6 0.6
^ ^ ^
R R R
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
-0.2 -0.1 0 0.1 0.2 -0.4 -0.2 0 0.2 0.4 -3 -2 -1 0 1 2 3
usL usL usL
FIG. 14. Cross-sectional profiles of steady axial velocity at Ẑ = 0, resulting from the Young wave in the context of HFV, for tube viscosities ␥ = 0 , 1 , 10 and
Womersley numbers ␣ = 2 , 6 , 10, as computed from the present model 共solid lines兲 and from Dragon and Grotberg’s model 共Ref. 5兲 共dashed lines兲, where
b̂ = 1.056, M = 874.56, and ␤ = 500. The wall viscosity has a stronger diminishing effect on the velocity profile in the present model than in Dragon and
Grotberg’s model.

affected by the viscosity. The approach adopted by Dragon
and Grotberg is dramatically different, where the viscous
damping is accounted for by terms added directly to the in-
ertia alongside the acceleration terms in the momentum
equations deduced previously by Atabek and Lew.18 For ex-
ample, the viscous damping in their radial momentum equa-
tion is a term that is proportional to the radial velocity of the
wall: 2˜␥ ⳵ rw / ⳵t, where the frictional coefficient ˜␥ is given in
Eq. 共102兲. In other words, the wall viscosity is introduced
into their model through a damping effect acting directly on
the wall inertia, which is essentially not the same as through
the stress-strain rate relations. With wall shear stresses ig-
nored, Dragon and Grotberg’s model falls short in fully de-
scribing the wall viscosity effects on the first-order as well as
the second-order fluid flows. This should explain the differ-
ence in the results we have seen thus far between the two
models when the wall is soft and viscous and the oscillation
frequency is high.
The results being not shown here, we have also evalu-
ated the steady axial streaming velocity using input values
that are relevant to blood flow for a free tube. In this case,
the oscillation is based on a fixed pressure gradient when the
frequency is varied. Under this condition, increasing the fre-
quency will actually decrease the stroke length or the tidal
volume. As a result, the streaming velocity will decrease in
magnitude as the frequency increases. Other than this aspect,
the streaming velocity profile is essentially affected by the
elasticity and viscosity of the wall in a manner similar to that
found in Figs. 13 and 14. The streaming velocity will also
change in direction as the frequency is varied. The effects
remain qualitatively the same even when the tube is con-
strained.
We show in Figs. 15 and 16 the influences of the initial
stresses on the steady streaming for a free tube in the context
of blood flow. The initial mean pressure q will in general
enhance the magnitude of the velocity profiles as shown in
Fig. 15. As has been shown in Fig. 7共a兲, the initial mean
pressure q is to decrease the Young wave speed or to increase
the wavenumber k̂. A larger wavenumber corresponds to a
larger amplitude of the first-order axial oscillatory velocity,
thereby leading to a larger value of the second-order time-
mean velocity, as these two velocities can be found from
Eqs. 共78兲 and 共131兲, respectively. The initial longitudinal
stretch ratio S also enhances the steady streaming when the
frequency is large but tends to have little effect on the
streaming at a low frequency, as shown in Fig. 16. By and
large, the streaming is affected to a greater extent by the
initial mean pressure than by the initial stretch. When the
tube becomes constrained 共results not shown here兲, we have
found that the effects of the mean pressure on the steady
streaming are similar to those shown in Fig. 15 for a free
tube. However, when q = 0, the initial stretch ratio S will have
no effect on the steady streaming.
051901-22 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

1 1
(a) α=2 (a) α=2
0.8 0.8
q =0 S =1
q = 1000 S = 1.2
0.6 q 0.6
= 2000 S = 1.4
^ q = 3000 S = 1.6
^
R 0.4 q = 4000 R 0.4

0.2 0.2

0 0
-100 0 100 200 300 -20 -10 0 10 20 30 40 50
s
uL usL
1 1
(b) α=6 (b)
S
α=6
0.8 q =0 =1
q 0.8 S
= 1000 = 1.2
q = 2000 S = 1.4
0.6 0.6 S
q = 3000 = 1.6
^ q = 4000
R 0.4 ^
R 0.4

0.2 0.2

0 0
-15 -10 -5 0 5 10 -8 -6 -4 -2 0 2
usL usL
1
(c) α = 10 (c)
1
α = 10
0.8 q =0 S =1
q 0.8 S = 1.2
= 1000
q = 2000 S = 1.4
0.6 0.6
q = 3000 S = 1.6
^ q = 4000
R 0.4 ^
R 0.4

0.2
0.2

0
-15 -10 -5 0 5 10 0
s -4 -2 0 2
uL s
uL
FIG. 15. Cross-sectional profiles of steady axial velocity at Ẑ = 0, resulting FIG. 16. Cross-sectional profiles of steady axial velocity at Ẑ = 0, resulting
from the Young wave in the context of blood circulation for a free tube for from the Young wave in the context of blood circulation for a free tube for
various values of the initial mean pressure q, where ␣ = 2 , 6 , 10, ␤ = 20 000, various values of the longitudinal stretch ratio S, where ␣ = 2 , 6 , 10, ␤
␥ = 10, and S = 1. The initial mean pressure can considerably enhance the = 20 000, ␥ = 10, and q = 0. The initial stretch enhances the steady streaming
magnitude of the steady velocity profiles. when the frequency is large but tends to have little effect at low frequency.

two models in handling the radial wall displacement and in

VI. CONCLUDING REMARKS
We have performed a Lagrangian perturbation analysis
to the second order for the oscillatory flow and the steady
streaming of a fluid as induced by a wave propagating in a
prestressed viscoelastic tube modeled as a Voigt material.
The cases of a free tube, which is free of stresses on the outer
wall surface, and a constrained tube, which is strongly teth-
ered by the surrounding tissues longitudinally, have been ex-
amined. The classical oscillatory viscous flow in a flexible
tube has been revisited. The present model has been checked
to be capable of yielding results that are in agreement with
those by the thin-walled tube model of Womersley6 and by
the thick-walled tube model of Cox.40 The influences of the
initial stresses on the propagation of Young and Lamb waves
have been studied and compared with other prestressed tube
models. The first-order results derived from the present
tethered-tube model in the context of HFV have also been
compared with those from Dragon and Grotberg’s model.5
Differences in the results of the two models are appreciable
for small elasticity and large viscosity of the tube at high
frequencies. We have also advanced to the second-order
problem to find the time-mean streaming velocity of the fluid
for an instant of time that is long enough for the mean mo-
tion of the wall to have already died out. Again, a compari-
son with Dragon and Grotberg’s model for the streaming
velocity shows similar differences in the results when the
tube is soft and the frequency and the damping are high. On
reviewing the two models, we have pointed out that the dis-
parities can stem from fundamental differences between the
the basic constitutive equation for the viscoelastic tube.
The present work is limited by the condition of small
displacements of the solid and fluid particles in order to en-
able the problem to be solved by an asymptotic analytical
approach. For finite deformations, where the displacements
of the wall are comparable to the tube radius, the problem
will still be formulated in terms of a Lagrangian description,
but the strongly nonlinear problem has then to be solved
numerically. This may happen when the forcing is very
strong, or when the tube is very soft, or when the frequency
is very high, e.g., the Womersley number ␣ ⱖ 15. Our prob-
lem presented here can then be used as a benchmark to test
the accuracy of a computational scheme that is to be devel-
oped to solve the Lagrangian problem numerically.
Our next problem is to look into the associated mass
transport. An important application is the dispersion due to
HFV of a vaporous substance in the respiratory system. With
the oscillatory flow field given by the present model, we may
proceed to find the concentration field using an effective
transport equation. Dragon and Grotberg5 examined the
steady rate of axial mass transport and determined an optimal
frequency of the steady mass transport as a function of the
Womersley number. It is worth finding out how the vis-
coelastic properties of an airway may influence this optimal
frequency based on the present flow model.
ACKNOWLEDGMENTS
The work was supported by the Research Grants Council
of the Hong Kong Special Administrative Region, China
through Project No. HKU 7192/04E and also by the Univer-
051901-23 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

sity of Hong Kong through the Seed Funding Programme for ␣2␥ ⳵ ṙˆw1 ⳵ r̂w1
Basic Research under Project Code 200611159078. This ␶ˆ wrr1 = 2 +2 , 共A3兲
␤ ⳵ R̂ ⳵ R̂
study constitutes part of the work performed by the first au-
thor for his Ph.D. thesis under the supervision of the second
author at the University of Hong Kong. ␣2␥ ṙˆw1 r̂w1
␶ˆ w␪␪1 = 2 +2 , 共A4兲
␤ R̂ R̂
APPENDIX A: FIRST-ORDER STRESS COMPONENTS,
AS IN EQS. „69…, „71…, and „72…
␣2␥ ⳵ żˆw11 ⳵ ẑw11
⳵ ṙˆ f1 ṙˆ f1
␶ˆ wzz1 = 2 +2 , 共A5兲
␶ˆ frr1 = 2 , ␶ˆ f ␪␪1 = 2 , 共A1兲
␤ ⳵ Ẑ ⳵ Ẑ
⳵ R̂ R̂
and

␶ˆ fzz1 = 2
⳵ żˆ f1
⳵ Ẑ
, ␶ˆ frz1 =
⳵ żˆ f1
⳵ R̂
, 共A2兲 ␶ˆ wrz1 =
␤
冉
␣2␥ ⳵ ṙˆw1 ⳵ żˆw12
⳵ Ẑ
+
⳵ R̂
+ 冊
⳵ r̂w1 ⳵ ẑw12
⳵ Ẑ
+
⳵ R̂
. 共A6兲

APPENDIX B: MATRIX A, AS IN EQ. „84…

冢
k̂2/2 ik̂J1共␭兲
− ik̂ ␭J0共␭兲
0 ␣4J1共␭兲/共M ␨兲
A=
␣4共1 − 2p̂ f0k̂2兲/共␨ M兲 − 2p̂ f0i5/2k̂␣5J0共␭兲/共␨ M兲
0 0
0 0

−1 0

冋 册
0 0

− ik̂ +
ik̂␤
␨S
−
ln共1 − j1兲 2
2
+
j1
冕
1
R̂ ln 冉冑 冊 R̂
R̂2 − j1
dR̂ +
j1
4
+ S3 − 1 1

−2−
␤
␨S
冉
j1 2
1 − − ln
1
2 j1 冑1 − j1
冊 0

− ik̂ +
ik̂␤
␨S
− 冋
ln共b̂2 − j1兲 2
2
+
j1
冕 bˆ
R̂ ln 冉冑 冊 R̂
2
R̂ − j1
dR̂ +
j1
4b̂2
+ S3 − 1 册 1

−
2
b̂ 2
−
␤ 1
− 冉
j1 2
− ln
␨S b̂2 2b̂4 j1
b̂
2
b̂ − j1 冑 冊 0

冣
− ik̂/2 0
−1 0

k̂2 −
␣ ik̂␤
+
4

2␨ ␨S
−
ik̂j1
2
冋
ln冑1 − j1 +
ik̂共j1 + S − 1兲
2
3
册 k̂2/2

2ik̂␣4 p̂ f0 . 共B1兲
ik̂ + − ik̂
M␨

k̂2b̂2 −
␣4b̂2 ik̂␤ ik̂j1
2␨
+
␨S 2
ln 冋 冉冑 冊 冉 b̂
b̂2 − j1
+
ik̂b̂2 j1
2 b̂2
+ S3 − 1 冊册 k̂2b̂2/2

ik̂ − ik̂
051901-24 Y. Ma and C.-O. Ng Phys. Fluids 21, 051901 共2009兲

再 册冎
APPENDIX C: THE TERMS A1 – A3, AS IN EQ. „86…

A1 = M共␣2␥i + ␤兲 3共1 − b̂2兲 +

␤
共␣ ␥i + ␤兲S
2 j1ln
b̂冑1 − j1
冋
冑b̂2 − j + 共S − 1兲共b̂ − 1兲
3 2

冋 冉 冊册
1

␤b̂2 1 j1 j1 2 b̂冑1 − j1
⫻ 1+ 共C1兲
2␨共b̂2 − 1兲S
1−
b̂2
− + + ln
2 2b̂4 j1 b̂2 − j1 冑 ,

A2 = ␣4 1 + 冋 冉 ␤
2S␨
1−
j1
2b̂ 2
−
2b̂2
j 1
ln
b̂
冑b̂ − j
2
冊册 −
3␣4␭J0共␭兲b̂2
2关J1共␭兲 − ␭J0共␭兲/2兴

冋 冉 冊册
1

␤b̂2 1 j1 j1 2 b̂冑1 − j1
− ␣4M共1 − b̂2兲 1 + 1− − + + ln
2␨共b̂2 − 1兲S b̂2 2 2b̂4 j1 b̂2 − j
1

冋 册−
冋 冉 冊册
bˆ冑1−j 1
冑b̂2 − j1
+
␤␣4␭J0共␭兲b̂2 j1 ln b̂2−j1

2␨共b̂ − 1兲S关J1共␭兲 − ␭J0共␭兲/2兴

2
+ 共S3 − 1兲共b̂2 − 1兲 ␣4b̂2␤
␨共b̂ − 1兲S
2
−
2
j1
冕 1
bˆ
R̂ ln
冑R̂2 − j1
R̂
dR̂ + ln
1 − j1
+
j1
4
1−
1
b̂2
,

共C2兲

A3 = 冋 共␣2␥i + ␤兲共b̂2 − 1兲关J1共␭兲 − ␭J0共␭兲/2兴

␣8b̂2
册冋 J1共␭兲 ␭J0共␭兲共1 − b̂2兲
M
−
2
. 册 共C3兲

APPENDIX D: SECOND-ORDER STRESS

COMPONENTS, AS IN EQS. „105…, „107…,
AND „108…
␶ˆ wzz2 = 2
␤
冉
␣2␥ ⳵ ẑ˙w2 ⳵ r̂w1 ⳵ ẑ˙w11 r̂w1 ⳵ ẑ˙w11
⳵ Ẑ
+
⳵ R̂ ⳵ Ẑ
+
R̂ ⳵ Ẑ
+2
⳵ ẑw21
⳵ Ẑ
冊
␶ˆ frr2 = 2 冉 ⳵ r̂˙ f2
+
⳵ ẑ f1 ⳵ r̂˙ f1
−
⳵ ẑ f1 ⳵ r̂˙ f1
+
r̂ f1 ⳵ r̂˙ f1
冊 , 共D1兲
+ 冉 冊 ⳵ ẑw11
⳵ Ẑ
2
, 共D7兲
⳵ R̂ ⳵ Ẑ ⳵ R̂ ⳵ R̂ ⳵ Ẑ R̂ ⳵ R̂

␶ˆ f ␪␪2 = 2 冉 r̂˙ f2
R̂
−
r̂˙ f1r̂ f1
R̂2
冊 , 共D2兲
␶ˆ wrz2 = ␶ˆ wzr2

= 冉
␣2␥ ⳵ ẑ˙w22 ⳵ r̂˙w2 ⳵ r̂w1 ⳵ r̂˙w1 r̂w1 ⳵ r̂˙w1
+ + +
␤ ⳵ R̂ ⳵ Ẑ ⳵ R̂ ⳵ Ẑ R̂ ⳵ Ẑ

␶ˆ fzz2 = 2 冉 ⳵ ẑ˙ f2
⳵ Ẑ
+
⳵ r̂ f1 ⳵ ẑ˙ f1
⳵ R̂ ⳵ Ẑ
−
⳵ r̂ f1 ⳵ ẑ˙ f1
⳵ Ẑ ⳵ R̂
+
r̂ f1 ⳵ ẑ˙ f1
R̂ ⳵ Ẑ
冊 , 共D3兲 −
⳵ r̂w1 ⳵ r̂˙w1
⳵ Ẑ ⳵ R̂
+
⳵ ẑw11 ⳵ ẑ˙w12
⳵ Ẑ ⳵ R̂
+
r̂w1 ⳵ ẑ˙w12
R̂ ⳵ R̂
−
⳵ ẑw12 ⳵ ẑ˙w11
⳵ R̂ ⳵ Ẑ
冊
⳵ ẑw22 ⳵ r̂w2 ⳵ r̂w1 ⳵ ẑw12 ⳵ r̂w1 ⳵ ẑw11
+ + + + . 共D8兲
⳵ ẑ˙ f2 ⳵ ẑ f1 ⳵ ẑ˙ f1 ⳵ ẑ f1 ⳵ ẑ˙ f1 r̂ f1 ⳵ ẑ˙ f1 ⳵ R̂ ⳵ Ẑ ⳵ R̂ ⳵ R̂ ⳵ Ẑ ⳵ Ẑ
␶ˆ frz2 = ␶ˆ fzr2 = + − + , 共D4兲
⳵ R̂ ⳵ Ẑ ⳵ R̂ ⳵ R̂ ⳵ Ẑ R̂ ⳵ R̂

␶ˆ wrr2 = 2
␤ ⳵ R̂
冉
␣2␥ ⳵ r̂˙w2 ⳵ ẑw11 ⳵ r̂˙w1 r̂w1 ⳵ r̂˙w1
+
⳵ Ẑ ⳵ R̂
+
R̂ ⳵ R̂
+2
⳵ r̂w2
⳵ R̂
冊 APPENDIX E: EULERIAN STEADY AXIAL VELOCITY
AT Ẑ = 0, AS IN DRAGON AND GROTBERG’S
MODEL „REF. 5…

+ 冉 冊 ⳵ r̂w1
⳵ R̂
2
, 共D5兲
W共0兲 1 共0兲 2
1 共R̂兲 = 2 k̂oi P1 R̂ + ␸共R̂兲 + l1 , 共E1兲

␶ˆ w␪␪2 = 2
␤ R̂
− 冉
␣2␥ r̂˙w2 r̂˙w1r̂w1
R̂2
+2
r̂w2
R̂
+
r̂w1
R̂
冊 冉 冊 2
, 共D6兲 ␸共R̂兲 = 冕 冕冋0
Rˆ 1
q̂ 0
q̂
Uⴱ0
dW0
dŝ
册
− ik̂oWⴱ0W0 ŝdŝdq̂, 共E2兲
051901-25 Wave propagation and induced steady streaming Phys. Fluids 21, 051901 共2009兲

P共0兲
1 =
−4
k̂oi
冋冉Hⴱ1
dW0
dR̂
共1兲 +
1 dU0
k̂oi dR̂
冊
共1兲 + ␸共1兲 16
superposed on steady flow in an orthotropic initially stressed elastic tube,”
Biorheology 13, 7 共1976兲.
U. Dinnar, “The role of the surrounding tissue in the propagation of waves

册
through the arterial system,” TIT J. Life Sci. 5, 49 共1975兲.

冕 1 17
K. Jagielska, D. Trzupek, M. Lepers, A. Pelc, and P. Zielinski, “Effect of
−2 ␸共R̂兲R̂dR̂ , 共E3兲 surrounding tissue on propagation of axisymmetric waves in arteries,”
0 Phys. Rev. E 76, 066304 共2007兲.
18
H. B. Atabek and H. S. P. Lew, “Wave propagation through a viscous
and incompressible fluid contained in an initially stressed elastic tube,” Bio-

冉 冊 冕 1 phys. J. 6, 481 共1966兲.

dW0 2 dU0 19
l1 = Hⴱ1 共1兲 + 共1兲 + ␸共1兲 − 4 ␸共R̂兲R̂dR̂. C. A. Kuiken, “Wave propagation in a thin-walled liquid-filled initially
stressed tube,” J. Fluid Mech. 141, 289 共1984兲.
dR̂ k̂oi dR̂ 0 20
J. D. Humphrey and S. L. Delange, An Introduction to Biomechanics:
共E4兲 21
Solids and Fluids, Analysis and Design 共Springer, New York, 2004兲.
D. Tang, C. Yang, S. Kobayashi, and D. N. Ku, “Steady flow and wall
Here, H1, U0共R̂兲, and W0共R̂兲 are the first-order quantities compression in stenotic arteries: A three-dimensional thick-wall model
with fluid-wall interactions,” J. Biomech. Eng. 123, 548 共2001兲.
given by 22
A. Marzo, X. Y. Luo, and C. D. Bertram, “Three-dimensional collapse and
steady flow in thick-walled flexible tubes,” J. Fluids Struct. 20, 817
H1 = 81 Ak̂o , 共E5兲 共2005兲.

冋 册
23
G. A. Holzapfel and T. C. Gasser, “Computational stress-deformation
analysis of arterial walls including high-pressure response,” Int. J. Cardiol.
iA␣2 J0共␭兲 J1共␭R̂兲 R̂
U0共R̂兲 = k̂o − , 共E6兲 116, 78 共2007兲.
4 J2共␭兲 ␭J0共␭兲 2 24
A. Tsamis and N. Stergiopulos, “Arterial remodeling in response to in-
creased blood flow using a constituent-based model,” J. Biomech. 42, 531
and 共2009兲.

冋 册
25
N. Riley, “Steady streaming,” Annu. Rev. Fluid Mech. 33, 43 共2001兲.
A␣2 J0共␭兲 J0共␭R̂兲 26
D. M. Wang and J. M. Tarbell, “Nonlinear analysis of flow in an elastic
W0共R̂兲 = −1 . 共E7兲 tube 共artery兲: Steady streaming effects,” J. Fluid Mech. 239, 341 共1992兲.
4 J2共␭兲 J0共␭兲 27
S. C. Ling and H. B. Atabek, “A nonlinear analysis of pulsatile flow in
arteries,” J. Fluid Mech. 55, 493 共1972兲.
28
1
K. B. Chandran, A. P. Yoganathan, and S. E. Rittgers, Biofluid Mechanics: R. Burattini, S. Natalucci, and K. B. Campbell, “Viscoelasticity modulates
The Human Circulation 共CRC, Boca Raton/Taylor & Francis, London, resonance in the terminal aortic circulation,” Med. Eng. Phys. 21, 175
2007兲. 共1999兲.
29
2
J. B. Grotberg, “Pulmonary flow and transport phenomena,” Annu. Rev. V. G. Hart and J. Shi, “Effects of initial stretches on wave speed in thin
Fluid Mech. 26, 529 共1994兲. orthotropic hyperelastic tubes containing fluid flow,” Appl. Math. Model.
3
W. Klip, P. Loon, and D. A. Klip, “Formulas for phase velocity and damp- 18, 198 共1994兲.
30
ing of longitudinal waves in thick-walled viscoelastic tubes,” J. Appl. C. W. Macosko, Rheology: Principles, Measurements and Applications
Phys. 38, 3745 共1967兲. 共Wiley, New York, 1994兲.
31
4
R. H. Cox, “Wave propagation through a Newtonian fluid contained within X. Zhang and C. O. Ng, “On the oscillatory and mean motions due to
thick-walled, viscoelastic tube,” Biophys. J. 8, 691 共1968兲. waves in a thin viscoelastic layer,” Wave Motion 43, 387 共2006兲.
32
5
C. A. Dragon and J. B. Grotberg, “Oscillatory flow and mass transport in C. O. Ng and X. Zhang, “Mass transport in water waves over a thin layer
a flexible tube,” J. Fluid Mech. 231, 135 共1991兲. of soft viscoelastic mud,” J. Fluid Mech. 573, 105 共2007兲.
33
6
J. R. Womersley, “Oscillatory motion of a viscous liquid in a thin-walled R. C. Batra and J. H. Yu, “Effect of inertia forces on the damping of a
elastic tube—I: The linear approximation for long waves,” Philos. Mag. constrained layer finitely deformed in shearing,” J. Sound Vib. 241, 913
46, 199 共1955兲. 共2001兲.
7 34
J. R. Womersley, “Method for the calculation of velocity, rate of flow and W. M. Lai, D. Rubin, and E. Krempl, Introduction to Continuum Mechan-
viscous drag in arteries when the pressure gradient is known,” J. Physiol. ics 共Butterworth-Heinemann, Oxford, 1993兲.
35
127, 553 共1955兲. A. S. Monin and A. M. Yaglom, Statistical Fluid Mechanics 共MIT Press,
8 Cambridge, 1971兲.
J. R. Womersley, “Oscillatory flow in arteries: The constrained elastic tube
36
as a model of arterial flow and pulse transmission,” Phys. Med. Biol. 2, W. J. Pierson, “Perturbation analysis of the Navier-Stokes equations in
178 共1957兲. Lagrangian form with selected linear solution,” J. Geophys. Res. 67,
9
D. K. Whirlow and W. T. Rouleau, “Periodic flow of a viscous liquid in a 3151, DOI: 10.1029/JZ067i008p03151 共1962兲.
37
thick-walled elastic tube,” Bull. Math. Biophys. 27, 355 共1965兲. Ü. Ünlüata and C. C. Mei, “Mass transport in water waves,” J. Geophys.
10
H. Demiray and S. Dost, “Solitary waves in a thick walled elastic tube,” Res. 75, 7611, DOI: 10.1029/JC075i036p07611 共1970兲.
38
Appl. Math. Model. 22, 583 共1998兲. C. O. Ng, “Mass transport and set-ups due to partial standing surface
11
G. W. Morgan and J. P. Kiely, “Wave propagation in a viscous liquid waves in a two-layer viscous system,” J. Fluid Mech. 520, 297 共2004兲.
39
contained in a flexible tube,” J. Acoust. Soc. Am. 26, 323 共1954兲. H. Lamb, Hydrodynamics 共Dover, New York, 1945兲.
12 40
S. K. Kyriacou and J. D. Humphrey, “Influence of size, shape and prop- R. H. Cox, “Comparison of linearized wave propagation models for arte-
erties on the mechanics of axisymmetric saccular aneurysms,” J. Biomech. rial blood flow analysis,” J. Biomech. 2, 251 共1969兲.
41
29, 1015 共1996兲. W. N. Wilmer and F. O. Michael, McDonald’s Blood Flow in Arteries:
13
H. Demiray, “A viscoelastic model for arterial wall materials,” Int. J. Eng. Theoretical, Experimental, and Clinical Principles 共Arnold, London,
Sci. 32, 1567 共1994兲. 1998兲.
14 42
H. B. Atabek, “Wave propagation through a viscous fluid contained in a M. S. Longuet-Higgins, “Mass transport in water waves,” Philos. Trans. R.
tethered, initially stressed, orthotropic elastic tube,” Biophys. J. 8, 626 Soc. London, Ser. A 245, 535 共1953兲.
43
共1968兲. C. C. Mei, The Applied Dynamics of Ocean Surface Waves 共World Scien-
15
H. Schwerdt and A. Constantinesco, “Periodic flow of a viscous fluid tific, Singapore, 1989兲.