arXiv:1304.6265v2 [cond-mat.soft] 24 Apr 2013
Finite amplitude elastic waves
propagating in compressible solids
Michel Destrade & Giuseppe Saccomandi
2005
Abstract
The paper studies the interaction of a longitudinal wave with transverse waves in general isotropic and unconstrained hyperelastic materials, including the possibility of dissipation. The dissipative term
chosen is similar to the classical stress tensor describing a Stokesian
fluid and is commonly used in nonlinear acoustics. The aim of this
research is to derive the corresponding general equations of motion,
valid for any possible form of the strain energy function and to investigate the possibility of obtaining some general and exact solutions
to these equations by reducing them to a set of ordinary differential
equations. Then the reductions can lead to some exact closed-form
solutions for special classes of materials (here the examples of the
Hadamard, Blatz-Ko, and power-law strain energy densities are considered, as well as fourth-order elasticity). The solutions derived are
in a time/space separable form and may be interpreted as generalized oscillatory shearing motions and generalized sinusoidal standing
waves. By means of standard methods of dynamical systems theory,
some peculiar properties of waves propagating in compressible materials are uncovered, such as for example, the emergence of destabilizing
effects. These latter features exist for highly nonlinear strain energy
functions such as the relatively simple power-law strain energy, but
they cannot exist in the framework of fourth-order elasticity.
1
1
Introduction
Nonlinear elastic wave propagation is a subject of considerable interest for
many natural and industrial applications such as seismology [55], soft tissue
acoustics [50], or the dynamics of elastomers [41]. A detailed study of the
theoretical issues associated with wave phenomena forms the basis of our
understanding of important non-destructive and non-invasive techniques of
investigation such as for instance, the technique of transient elastography for
the analysis of soft solids [20]. The mathematics of wave phenomena is still
an active subject of research where many outstanding problems are waiting
for a definitive systematic treatment. Many papers are devoted to the study
of nonlinear elastic wave propagation and a recent summary on the status
of contemporary research on the subject can be found in the authoritative
review by Norris [47], while an account of the mathematics of hyperbolic
conservation laws can be found in a book by Dafermos [23].
It is important to note that in most studies of nonlinear acoustics, the
investigation is usually restricted to weakly nonlinear waves. Indeed, for
materials such as metals, rocks, or ceramics, the ratio of the dynamic displacement to the wavelength is a small parameter and so, the theories of
third-order elasticity or of fourth-order elasticity seem to be sufficient to account for the nonlinear effects observed in experimental tests. The situation
is however completely different when we consider elastomeric materials such
as those used for vibration isolators or automobile tires, and when we consider biological materials such as arterial walls or glands, under physiological
or pathological conditions. The deformation rates occurring in these materials are so large that their behavior will differ significantly not only from
the behavior predicted by the linear theory of elasticity, but also from the
behavior predicted by the weakly nonlinear theories. It should therefore be
important to be able to derive and to investigate the general equations governing the propagation of finite amplitude longitudinal and transverse waves
in the context of full nonlinear elasticity, which recovers the weakly nonlinear case as a special case. Moreover, recent researches in the constitutive
behavior of rubber-like material indicate that the stiffening effect, a peculiar
but real phenomenon occurring at very large strains, cannot be described
accurately when the full polynomial strain energy function is approximated;
also, high order polynomial theories may give rise to numerical difficulties
in the fitting of the material parameters with the experimental data and to
artificial instability phenomena (see Pucci and Saccomandi [51] and Ogden
2
et al. [49]).
These considerations have led to the present paper which is devoted to
the study of the interaction of a longitudinal wave with transverse waves
in general isotropic and unconstrained hyperelastic materials and this also
when dissipation is taken into account. The aim of our research is first, to
derive the corresponding general equations of motion, valid for any possible
form of the strain energy function; second, to investigate the possibility of
obtaining some general and exact solutions to these equations by reducing
the specific field equations to a set of ordinary differential equations; and
third, to provide, using these reductions, some exact closed-form solutions
for special classes of materials.
To the best of our knowledge, previous studies taking systematically into
account the full nonlinear equations of motion have generally been restricted
to the formal theory of singular surfaces and of acceleration waves. These
studies stem from the fundamental researches initiated by Hadamard [30] in
1903 and were advanced mainly by Ericksen, by Thomas, and by Truesdell in
the 1950s and 1960s (see for example Truesdell [58] and the review by Chen
[22]). They show that the conjunction of the nonlinearity in the material
response and of the hyperbolic nature of the governing equations (in the
purely elastic case) leads inevitably to shock formation. The results obtained
are general and exact but they are formal results: in effect, no solution is
found explicitly but rather, conditions are given that have to be fulfilled by
solutions, if they exist.
In contrast, the present article focuses on the explicit determination of
solutions. Our investigation is directly related to the celebrated finite amplitude elastic motions discovered by Carroll [9, 10, 11, 12, 13, 14, 15]. His
exact solutions are versatile in their fields of application because they are
valid not only for nonlinearly elastic solids, but also for general viscoelastic
solids [14], Reiner-Rivlin fluids [14, 15], Stokesian fluids [14], Rivlin-Ericksen
fluids [15], liquid crystals [16], dielectrics [17], magnetic materials [18], etc.
They also come in a great variety of forms, as circularly-polarized harmonic
progressive waves, as motions with sinusoidal time dependence, as motions
with sinusoidal space dependence, etc. Recently the present authors [25] extended Carroll’s solutions to the case of an incompressible hyperelastic body
in rotation, by considering a new synthetic and very effective method based
on complex variables. In doing so, they discovered a striking analogy between
the equations of motion obtained for a motion general enough to include all
of the above motions, and the equations obtained for the motion of a nonlin3
ear string, as considered by Rosenau and Rubin [53]. The method used by
Rosenau and Rubin shows in a simple and direct way that all (and more) of
the different results obtained by Carroll are a direct consequence of material
isotropy and of the Galilean invariance of the field equations. This is indeed
an explanation for their ubiquity and versatility.
The paper is organized in the following manner. In Section II we lay out
the general field equations of nonlinear elasticity. Because dissipation cannot be neglected for most problems in the dynamics of elastomeric materials
and of soft tissues, we add a simple inelastic tensor of differential type to
the hyperelastic Cauchy stress tensor. This additional term is similar to the
classical stress tensor describing a Stokesian fluid; it is the dissipative term
introduced by Landau and Lifshitz in their monograph on elasticity theory
[43] and it is commonly used in nonlinear acoustics (see Norris [47]). We
point out that this term is different from the one used in continuum mechanics as a first approximation of dissipative effects or as a regularization of the
hyperbolic equations of nonlinear elastodynamics for example in numerical
computations [19]. Then we specialize the equations of motion to the case
of one longitudinal wave and two transverse waves. In Section III we seek
solutions in time/space separable form and derive nonlinear systems of ordinary differential equations for generalized oscillatory shearing motions and
for generalized sinusoidal standing waves. In Section IV, different classes of
materials are considered (Hadamard, Blatz-Ko, power-law, fourth-order elasticity) and some exact solutions are provided. Concluding remarks are made
in Section V, which recaps the main results. In particular, it is emphasized
there that a weak nonlinear elasticity theory (up to the fourth order) cannot
account for some wrinkling phenomenon found in the fully nonlinear theory
(power-law strain energy).
2
2.1
Preliminaries
Equations of motion
Consider a hyperelastic body with strain energy density Σ. Let the initial
and current coordinates of a point of the body, referred to the same fixed
rectangular Cartesian system of axes, be denoted by X and x, respectively.
Hence a motion of the body is defined by
x = x(X, t).
4
(1)
The response of a homogeneous compressible isotropic elastic solid to deformations from an undistorted reference configuration is described by the
constitutive relation [4],
p ∂Σ
I2 ∂Σ p ∂Σ
2 ∂Σ
TE = 2( √
+ I3
)1 + √
B − 2 I3
B−1 ,
∂I3
∂I2
I3 ∂I2
I3 ∂I1
(2)
B := FFT ,
(3)
where TE is the (elastic) Cauchy stress tensor, 1 is the unit tensor, B is the
left Cauchy-Green strain tensor defined by
F := ∂x/∂X being the deformation gradient tensor, and I1 , I2 , I3 are the
first three invariants of B,
I1 := tr B,
I2 := 12 [I12 − tr(B2 )],
I3 := det B.
(4)
To describe the simultaneous effects of thermal and viscoelastic dissipation, we introduce the following viscous-like stress tensor [43],
TD = 2η[Ė − 31 (tr Ė)1] + (ζ + χ)(tr Ė)1,
(5)
where E is the Green-Lagrange strain tensor,
E := 21 (FT F − 1),
(6)
and the dot indicates the time derivative. In (5), the constant η (> 0) is
the shear viscosity coefficient, the constant ζ (> 0) is the bulk viscosity
coefficient, and the constant χ (> 0) is the coefficient related to the thermal properties of the solid (that is, ambient temperature, thermal expansion
coefficient, and specific heat).
Let ρ and ρ0 denote the mass densities of the body measured in the
current configuration and in the reference configuration, respectively. Then
the equation of motions, in the absence of body forces, are: div (TE + TD ) =
ρẍ in current form, or equivalently,
p
(7)
Div [ I3 (TE + TD )(F−1 )T ] = ρ0 ẍ,
∂ p
∂ 2 xi
−1
] = ρ0 2 ,
[ I3 (TikE + TikD )Fjk
∂Xj
∂t
in referential form.
5
Remark on the dissipative stress tensor
For incompressible solids, det F = 1 at all times (so that I3 = 1 at all times)
and an arbitrary spherical pressure term, to be determined from initial and
boundary conditions, is introduced so that
TE + TD = −p1 + 2
∂Σ
∂Σ −1
B−2
B + 2η Ė,
∂I1
∂I2
(8)
where p = p(x, t) is a Lagrange multiplier. We note that in order to study
the propagation of finite amplitude motions in viscoelastic incompressible
materials, some authors [34, 35, 24, 36] chose to add to the elastic Cauchy
stress tensor a viscous term linear in the stretching tensor D, which is the
symmetric part of the velocity gradient tensor ḞF−1 . In other words, they
chose to write
TE + TD = −p∗ 1 + 2
∂Σ −1
∂Σ
B−2
B + 2νD,
∂I1
∂I2
(9)
for some Lagrange multiplier p∗ and for some constant ν. However it must
be recalled (e.g. Chadwick [21, p.146]) that Ė = FT DF 6= D in general.
Thus we remark, for the time being, that the assumptions (8) and (9) are
not equivalent a priori. We discuss these issues further at the end of the next
Subsection.
2.2
Finite amplitude longitudinal and transverse plane
waves
Now we consider the following class of motions,
x = u(X, t),
y = Y + v(X, t),
z = Z + w(X, t),
(10)
which describes the superposition of a transverse wave, polarized in the (Y Z)
plane and propagating in the X-direction with a longitudinal wave propagating in the X-direction. Then,
1
0
0
uX 0 0
−1
1
−vX uX 0 ,
F ij =
[F]ij = vX 1 0 ,
(11)
uX
−wX 0 uX
wX 0 1
6
and
2
2
2
(uX + vX
+ wX
)t vXt wXt
h i
1
vXt
0
0 .
Ė =
2
ij
wXt
0
0
(12)
Here and hereafter, a subscript letter denotes partial differentiation (thus
uX = ∂u/∂X, vtt = ∂ 2 v/∂t2 , etc.) It follows from the definitions (4) that the
first three invariants of B are given by
2
2
I1 = 2 + u2X + vX
+ wX
,
2
2
I2 = 1 + 2u2X + vX
+ wX
,
I3 = u2X ,
(13)
and consequently that the strain-energy density Σ = Σ(I1 , I2 , I3 ) is here a
2
2
function of u2X and vX
+ wX
alone,
2
2
Σ = Σ(u2X , vX
+ wX
).
(14)
With the expressions (11) and (12) substituted into the Cauchy stress (2)
and into the viscous-like tensor (5), the equation of motions (7) reduce to
2
2
)t ]X ,
+ wX
ρ0 utt = [(Q1 + Q2 )uX + 12 (ζ + χ + 43 η)(u2X + vX
ρ0 vtt = [Q1 vX + ηvXt ]X ,
ρ0 wtt = [Q1 wX + ηwXt ]X ,
(15)
2
2
2
2
where the functions Q1 = Q1 (u2X , vX
+ wX
) and Q2 = Q2 (u2X , vX
+ wX
) are
defined by
∂Σ
∂Σ
∂Σ
∂Σ
, Q2 := 2
.
(16)
+
+
Q1 := 2
∂I1 ∂I2
∂I2 ∂I3
Eqs. (15) form a system of two coupled nonlinear partial differential equations, generalizing the system derived by Carroll in [9] for an elastic compressible material.
The mathematical treatment of initial-boundary-value problems for equations such as (15) is not trivial at all. There is a considerable literature on
the subject and the most updated reference is the recent paper by Antman
and Seidman [2], to which we refer for further information. Here we use a
semi-inverse method and we shall consider the possibility to use our explicit
results to solve initial-boundary-value problems only a posteriori. Obviously,
the possibility of general uniqueness theorems (such as those established in
[2]) is quite valuable to give a clear and definitive status of the solutions
found by a semi-inverse method.
7
Now we differentiate (15) with respect to X, and we recast the resulting
equations in the form
ρ0 Utt = [(Q1 + Q2 )U ]XX
+ 21 (ζ + χ + 34 η)[U 2 + V 2 + W 2 ]XXt ,
ρ0 Vtt = [Q1 V ]XX + ηVXXt ,
ρ0 Wtt = [Q1 W ]XX + ηWXXt ,
(17)
where we have introduced the new functions: U := uX , V := vX , W := wX .
Further, these equations can be rewritten in an even more compact form,
by introducing the complex function Λ, with modulus Ω and argument θ,
defined by
Λ(X, t) = Ω(X, t)eiθ(X,t) := V + iW,
(18)
so that
V = ℜ(Λ) = Ω cos θ,
V 2 + W 2 = Ω2 .
W = ℑ(Λ) = Ω sin θ,
(19)
Then the three equations (17) are
ρ0 Utt = [(Q1 + Q2 )U ]XX + 21 (ζ + χ + 34 η)[U 2 + Ω2 ]XXt ,
ρ0 Λtt = [Q1 Λ]XX + ηΛXXt .
(20)
Now also, the invariants I1 , I2 , I3 found in (13) are written as
I 1 = 2 + U 2 + Ω2 ,
I2 = 1 + 2U 2 + Ω2 ,
I3 = U 2 ,
so that Q1 = Q1 (U 2 , Ω2 ) and Q2 = Q2 (U 2 , Ω2 ) or, more explicitly,
∂Σ
∂Σ
∂Σ
Q1 = 2
, Q2 = 2
−
.
∂(Ω2 )
∂(U 2 ) ∂(Ω2 )
(21)
(22)
These equations have been developed in the purely elastic case of a single
shear wave by Carman and Cramer [8], who determined some numerical and
asymptotic solutions.
8
Remark on finite amplitude motions in incompressible materials
In incompressible solids, det F = 1 at all times and so by (11)1 , uX = 1.
Then we find
2
2
(vX + wX
)t vXt wXt
h i
1
vXt
0
0 ,
(23)
Ė =
2
ij
wXt
0
0
and
0 vXt wXt
1
0 .
[D]ij = vXt 0
2
wXt 0
0
(24)
Div [(−p1 + 2η Ė)(F−1 )T ] = −Grad p̂ + ηvXXt j + ηwXXt k,
(25)
It then follows that
2
2
where p̂ := p − (vX
+ wX
)t , and that
Div [(−p∗ 1 + 2νD)(F−1 )T ] = −Grad p∗ + νvXXt j + νwXXt k.
(26)
Now we see that the equations of motion (7) are the same whether the choice
(8) or the choice (9) is made. The two approaches in modeling the dissipative effects are reconciled, at least as far as finite amplitude motions in
incompressible materials are concerned.
3
Solutions in separable form
The general solution of (20) may be found only via numerical methods. To
obtain some simple analytical information here, we look for reductions of
the governing equations (20) to a set of ordinary differential equations. The
underlying idea is to search for Λ(X, t), defined in (18), in a separable form
such as
Ω(X, t) = Ω1 (X)Ω2 (t), θ(X, t) = θ1 (X) + θ2 (t),
(27)
where Ω1 , θ1 are functions of space only and Ω2 , θ2 are functions of time only.
Some lengthy computations, not reproduced here, show that in only two
general cases can equations (20) be separated in such way for any material
response.
The results of this Section generalize the results of Carroll [10, 13] by
considering more general types of displacement fields and by including viscoelastic effects. We also acknowledge that the solutions presented here are a
9
generalization of the non-linear three-dimensional motions of an elastic string
found by Rosenau and Rubin [53] (see [25] for a different generalization, to
finite amplitude waves in rotating incompressible elastic bodies.)
3.1
Generalized oscillatory shearing motions
The first general type of separable solution is of the form,
Λ(X, t) = [ψ(t) + iφ(t)]kei(kX−θ(t)) ,
(28)
where k is a constant and ψ, φ, θ are arbitrary real functions of time. Here
Λ is indeed of the separable form given by (18), with
Ω1 (X) = k = const.,
θ1 (X) = kX,
1
Ω2 (t) = [φ(t)2 + ψ(t)2 ] 2 ,
θ2 (t) = θ(t) + tan−1 [φ(t)/ψ(t)].
(29)
Separating real and imaginary parts, we find that the transverse displacement
field (v, w) derived from (28) is
v(X, t) = φ(t) cos(kX − θ(t)) + ψ(t) sin(kX − θ(t)),
w(X, t) = φ(t) sin(kX − θ(t)) − ψ(t) cos(kX + θ(t)).
(30)
In this case,
Ω2 = k 2 [φ2 (t) + ψ 2 (t)],
(31)
is a function of time only.
Now the first equation of motion (20)1 reduces to
ρ0 Utt = [(Q1 + Q2 )U ]XX + 12 (ζ + χ + 43 η)[U 2 ]XXt ,
(32)
where
Q1 = Q1 (U 2 (X, t), Ω2 (t)),
Q2 = Q2 (U 2 (X, t), Ω2 (t)).
(33)
On the other hand, the second equation of motion (20)2 reduces to
ρ0 Λtt = Q1 ΛXX + ηΛXXt .
(34)
Using the separable form (28) for Λ, factoring out the exponential term, and
separating the real and imaginary parts yields
′
ρ0 (φθ′′ − ψθ 2 + 2φ′ θ′ + ψ ′′ ) + k 2 Q1 ψ + k 2 η(φθ′ + ψ ′ ) = 0,
′
ρ0 (φ′′ − 2ψ ′ θ′ − φθ 2 − ψθ′′ ) + k 2 Q1 φ + k 2 η(φ′ − ψθ′ ) = 0.
10
(35)
This is a nonlinear system of two ordinary differential equations in the three
unknowns φ(t), θ(t), and ψ(t) either: when Q1 is a constant independent
of its arguments U 2 (X, t) and Ω2 (t), or: when Q1 = Q1 (t). By (16)1 , this
latter condition is satisfied if and only if U (X, t) = U (t), a function of time
only. It then follows by (33)2 that Q2 = Q2 (t) also. By substitution into the
first equation of motion (20)1 , we obtain ρ0 U ′′ = 0. Then the longitudinal
displacement field is given by
U (t) = C1 t + C2 = uX ,
(36)
u(X, t) = (C1 t + C2 )X + C3 ,
(37)
so that
where C1 , C2 , C3 are constants. This is a homogeneous accelerationless motion.
Hence we showed that for any compressible elastic material with viscoelastic dissipative part defined as in (5), maintained in a state of longitudinal
simple extension, the special transverse waves (30) may always propagate.
Because (35) is a system of only two equations in three unknowns, there
is a certain freedom in considering special classes of solutions. One such class
of special solutions is obtained by considering ψ := 0 in (30), leading to
v(X, t) = φ(t) cos(kX − θ(t)),
w(X, t) = φ(t) sin(kX − θ(t)),
(38)
a direct generalization of the classical damped harmonic circularly-polarized
wave solution,
v(X, t) = Ae−ht cos(kX − ωt),
w(X, t) = Ae−ht sin(kX − ωt),
(39)
where A, h, and ω are suitable constants. For these special motions at ψ := 0,
the system (35) reduces to
ρ0 (φθ′′ + 2φ′ θ′ ) + ηk 2 φθ′ = 0,
ρ0 (φ′′ − φθ′2 ) + k 2 Q1 φ + ηk 2 φ′ = 0.
(40)
The equation (40)1 admits the first integral
2
2 ′
φ θ = Ee
11
t
− ηk
ρ
0
,
(41)
where E is a constant of integration.
If E = 0, then θ(t) = θ0 , a constant. We end up with standing waves,
v(X, t) = φ(t) cos(kX − θ0 ),
w(X, t) = φ(t) sin(kX − θ0 ),
(42)
where, according to (40)2 , φ is a solution to
ρ0 φ′′ + k 2 ηφ′ + k 2 Q1 (k 2 φ2 , t)φ = 0,
(43)
which is the equation of a damped nonlinear oscillator.
If E 6= 0, then solving (41) for θ′ , we end up with the single equation
(40)2 , which reduces to
ρ0 φ′′ + k 2 ηφ′ + k 2 Q1 φ −
2
ρ0 E 2 − 2ηk
t
ρ0
= 0.
e
3
φ
(44)
In the elastic case (η = 0), this equation describes the plane motion of a
particle in a central force field. In general (η 6= 0), it is a nonlinear and
non-autonomous differential equation.
For harmonic wave propagation, θ(t) is taken as θ(t) = ωt. By (41), φ
2
t
− ηk
2ρ
and describes the usual linear damping function.
is proportional to e
However by substitution in (40)2 , we see that this situation is possible only
if the material response is such that Q1 is a constant,
∂Σ
ω2 η2k2
∂Σ
Q1 = 2
= const. = ρ0 2 +
+
.
(45)
∂I1 ∂I2
k
4ρ0
0
In this Subsection we analyzed the solutions (38) which are the analogue
of the Class I solutions presented by Rosenau and Rubin [54] for purely elastic
strings. Qualitatively the solutions describe particles moving in a helical path
which lies on a cylindrical surface of time-varying radius.
3.2
Generalized sinusoidal standing waves
We now consider the second class of solutions in separable form, namely
Λ(X, t) = [(iφ(X) + ψ(X))θ′ (X) + (φ′ (X) − iψ ′ (X))]ei(ωt+θ(X)) ,
(46)
where k is a constant and ψ(X), φ(X), θ(X) are arbitrary functions of space.
Here Λ(X, t) is indeed of the separable form (18), with
1
Ω1 (X) = [(φ′ + ψθ′ )2 + (φθ′ − ψ ′ )2 ] 2 , Ω2 (t) = 1 = const.,
θ1 (X) = θ + tan−1 [(φθ′ − ψ ′ )/(ψθ′ + φ′ )], θ2 (t) = ωt.
12
(47)
The transverse displacement field (v, w) corresponding to (46) is
v(X, t) = φ(X) cos(ωt + θ(X)) + ψ(X) sin(ωt + θ(X)),
w(X, t) = φ(X) sin(ωt + θ(X)) − ψ(X) cos(ωt + θ(X)).
(48)
In this case, Ω = Ω1 (X), a function of X only, and the first equation of
motion (20)1 reduces to
ρ0 Utt = [(Q1 + Q2 )U ]XX + 12 (ζ + χ + 43 η)[U 2 ]XXt ,
(49)
where
Q1 = Q1 (U 2 , Ω21 (X)),
Q2 = Q2 (U 2 , Ω21 (X)).
(50)
Now, introducing (46) into the second equation of motion (20)2 and separating the real part from the imaginary part, we obtain
′
− ρ0 ω 2 (ψθ′ + φ′ ) = (ψθ′ + φ′ )(Q1 )XX + 2(2ψ ′ θ′ + ψθ′′ + φ′′ − φθ 2 )(Q1 )X
′
′
+ (3ψ ′′ θ′ + 3ψ ′ θ′′ + ψθ′′′ − ψθ 3 − 3φ′ θ 2 − 3φθ′ θ′′ + φ′′′ )Q1
′
′
− ηω(3φ′′ θ′ + 3φ′ θ′′ + φθ′′′ − φθ 3 + 3ψ ′ θ 2 + 3ψθ′ θ′′ − ψ ′′′ ), (51)
and
′
− ρ0 ω 2 (φθ′ − ψ ′ ) = (φθ′ − ψ ′ )(Q1 )XX + 2(2φ′ θ′ + φθ′′ − ψ ′′ + ψθ 2 )(Q1 )X
′
′
+ (3φ′′ θ′ + 3φ′ θ′′ + φθ′′′ − φθ 3 + 3ψ ′ θ 2 + 3ψθ′ θ′′ − ψ ′′′ )Q1
′
′
+ ηω(3ψ ′′ θ′ + 3ψ ′ θ′′ + ψθ′′′ − ψθ 3 − 3φ′ θ 2 − 3φθ′ θ′′ + φ′′′ ). (52)
A sufficient condition for the equations (51) and (52) to compose a nonlinear
system of two ordinary differential equations in the three unknowns φ , ψ,
and θ is that the longitudinal field depends only on X: U (X, t) ≡ U (X). In
this case, Q1 = Q1 (X), Q2 = Q2 (X) and (49) reduces to
[(Q1 (X) + Q2 (X))U (X)]′′ = 0.
(53)
Further progress is made by fixing one of the three unknown functions,
θ := 0, say. Then the governing equations (51) and (52) reduce to
(Q1 φ′ )′ + ρ0 ω 2 φ = −ηωψ ′′ ,
(Q1 ψ ′ )′ + ρ0 ω 2 ψ = ηωφ′′ .
(54)
At η = 0, these equations are formally equivalent to the equations derived
by Carroll [11], with the difference that the materials response functions Q1
13
and Q2 depend not only on Ω2 (the amount of shear) but also on U . In
Carroll [15] (see system 5.12 in that reference), the attention is restricted to
incompressible fluids and solids, for which U (X) must be constant.
The solutions considered in this sub-Section contain the Class II solutions
of Rosenau and Rubin [54].
4
Specific materials
Now that we have established the possibility of reducing the general nonlinear partial differential equations to ordinary differential equations, we must
specify constitutive relations in order to make progress.
4.1
Hadamard materials
Finite amplitude motions in elastic Hadamard materials have been thoroughly studied, see John [42] and Boulanger et al. [6]. The analysis conducted in this paper allows for a treatment of a dissipative Hadamard solid.
The Hadamard strain energy function is defined by
2Σ = C(I1 − 3) + D(I2 − 3) + G(I3 ),
(55)
where G(I3 ) is a material function and C, D are material constants such
that [6] C > 0, D ≥ 0, or C ≥ 0, D > 0. The connection with the Lamé
constants λ and µ of the linear theory of isotropic elasticity is made through
the relations:
C = 2µ + G′ (1),
D = −µ − G′ (1),
2G′′ (1) = λ + 2µ.
(56)
For this material, the functions Q1 and Q2 defined in (16) are
Q1 = C + D,
Q2 = D + G′ (U 2 ),
(57)
and the equations of motion (20) reduce to
ρ0 Utt = {[C + 2D + G′ (U 2 )]U }XX
+ 21 (ζ + χ + 34 η)(U 2 + Ω2 )tXX ,
ρ0 Λtt = (C + D)ΛXX + ηΛtXX .
14
(58)
We point out that the equation governing the propagation of the transverse waves (58)2 is always linear, for all Hadamard materials, elastic or
dissipative. Hence we see at once that classical harmonic (damped and attenuated) transverse waves are always possible for these materials.
The material function G(I3 ) accounts for the effects of compressibility.
For example, Levinson and Burgess [44] proposed the following explicit form,
p
G(I3 ) = (λ + µ)(I3 − 1) − 2(λ + 2µ)( I3 − 1).
(59)
This function leads here to a remarkably simple system of equations:
ρ0 Utt = (λ + 2µ)UXX + 21 (ζ + χ + 34 η)(U 2 + Ω2 )tXX ,
ρ0 Λtt = µΛXX + ηΛtXX .
(60)
Now we take a look at some of the solutions investigated in the previous
Section. Classical damped harmonic circularly polarized waves such as (39)
are possible in a dissipative Hadamard material because then Ω = kAe−ht , a
function of time only, and Q1 = C + D = µ, a constant. The corresponding
dispersion equation (45) is here
µ = ρ0
ω2 η2k2
+
.
k2
4ρ0
(61)
The associated longitudinal motion may be either the homogeneous accelerationless motion (36) or any solution to the nonlinear equation (58)1 , here:
ρ0 Utt = {[G′ (U 2 ) − G′ (1)]U }XX + 21 (ζ + χ + 34 η)(U 2 )tXX ,
(62)
which for the Levinson and Burgess choice (59) of G is:
ρ0 Utt = (λ + 2µ)UXX + 21 (ζ + χ + 34 η)(U 2 )tXX .
(63)
Now consider the sinusoidal standing waves of Eq.(48) at θ(X) = 0.
When the longitudinal field U depends on X only, their behavior is governed
by Eqs.(54), here:
µφ′′ = ρ0 ω 2 φ + ηωψ ′′ ,
µψ ′′ = ρ0 ω 2 ψ − ηωφ′′ ,
(64)
or equivalently, by the single complex equation
(µ − iηω)(φ + iψ)′′ = ρω 2 (φ + iψ).
15
(65)
This equation possesses the following class of attenuated solutions,
φ(X) = e−αX [k1 sin(βX) + k2 cos(βX)],
ψ(X) = e−αX [k3 cos(βX) + k4 sin(βX)],
where k1 , k2 are disposable constants and
s
i
hp
ρ0 ω 2
2 + η2ω2 ± µ .
µ
α, β =
2(µ2 + η 2 ω 2 )
(66)
(67)
These solutions may be used to solve some simple boundary value problems.
For example, consider the case of semi-infinite body bounded by a vibrating
rigid plate at X = 0, and take the velocity components of the plate as
ẋ = 0,
ẏ = V ω cos(ωt),
ż = V ω sin(ωt).
(68)
Then φ(0) = 0, ψ(0) = V , and the associated transverse displacements are
v(X, t) = e−αX [sin(βX) cos(ωt) + V cos(βX) sin(ωt)],
w(X, t) = e−αX [sin(βX) sin(ωt) − V cos(βX) cos(ωt)].
(69)
The associated longitudinal displacement independently satisfies (53), here:
{[G′ (U 2 ) − G′ (1)]U }′′ = 0,
(70)
which for the Levinson and Burgess choice (59) of G is simply: U ′′ = 0.
Then the basic displacement (69) may be superimposed upon an axial static
stretch or a homogeneous motion in the axial direction. In a similar way,
it is possible to solve the case of a slab fixed at X = 0 and oscillating at
X = L, or it is possible to use the various integration constants to fix the
stress traction or shear stress at the boundary of a vibrating half-space or of
a slab.
4.2
Blatz-Ko materials
Experiments on compressible polyurethane rubber lead Blatz and Ko [3]
to propose some strain energy functions which have since received much
attention, see for instance, Beatty [4]. In particular, two reduced forms of
the Blatz-Ko general constitutive equation were deemed appropriate to model
16
certain polyurethane rubber samples. One is the Blatz-Ko strain energy
function for a solid, polyurethane rubber,
Σ=
µ
1/β
[I1 − 3 + β(I3 − 1)],
2
(71)
where µ and β are constants. Direct comparison with (55) shows that this
material is a special Hadamard material, with the identifications: C = µ,
1/β
D = 0, G(I3 ) = −µβI3 . The other is the Blatz-Ko strain energy function
for a foamed, polyurethane elastomer,
µ I2
1/β
(72)
Σ=
− 3 + β(I3 − 1) ,
2 I3
where µ and β are constants. The continuity with linear elasticity requires
that µ is the infinitesimal shear modulus and that β = (1 − 2ν)/ν, where ν
is the infinitesimal Poisson ratio. Blatz and Ko’s experiments showed that
typically, ν = 1/4 (and so, β = 2).
For this latter material, the functions Q1 and Q2 defined in (16) are
Q1 = µU −2 ,
Q2 = −µU −2 − µ(1 + Ω2 )U −4 + µU
and the equations of motion (20) reduce to
i
h
3β−2
ρ0 Utt = −µ (1 + Ω2 )U −3 − U β
1
(ζ
2
4
η)(U 2
3
2(β−1)
β
,
(73)
XX
2
+
+χ+
+ Ω )tXX ,
−2
ρ0 Λtt = µ(ΛU )XX + ηΛtXX .
(74)
Here, and in contrast with the case of Hadamard materials, the transverse
wave is coupled to the longitudinal wave.
First, consider the generalized oscillatory shearing motions of Section
III.A. As seen there, they are governed by nonlinear ordinary differential
equations when the longitudinal displacement u is the homogeneous accelerationless motion of (36)2 . Then the standing waves of (42) are governed
by
µk 2
ρ0 φ′′ + ηk 2 φ′ +
φ = 0.
(75)
(C1 t + C2 )2
At C1 = 0, this is the classical equation of a damped oscillator, whose
type of motion (exponential and/or sinusoidal) is decided by the sign of the
17
quantity: (ηkC2 )2 − 4ρ0 µ; hence the nature of the transverse motion depends
on the longitudinal extension via the parameter C2 .
At C1 6= 0 (linear stretch rate), we perform the change of variables ζ =
C1 t + C2 , so that (75) reduces to
ζ2
d2 φ
dφ Γ
+ 2Υζ 2
+ φ = 0.
2
dζ
dζ
4
(76)
where the quantities Υ, Γ are defined by
Υ :=
ηk 2
,
2ρ0 C1
Γ :=
4µk 2
.
ρ0 C12
(77)
In the purely elastic case (η = 0), Υ vanishes and the solution to this equation
is quite simple: we find that for
√
Γ < 1,
φ(ζ) = k1 ζ λ+ + k2 ζ λ− , where λ± = (1 ± 1 − Γ)/2;
p
p
Γ = 1,
φ(ζ) = k1 ζ + k2 ζ ln ζ;
p
p
√
√
(78)
Γ > 1,
φ(ζ) = k1 ζ cos( Γ − 1 ln ζ) + k2 ζ sin( Γ − 1 ln ζ);
where k1 and k2 are integration constants. We note that all these solutions
blow up as t → ∞ (equivalent to ζ → ∞); that for Γ ≤ 1, the solutions are
monotonic increasing; and that for Γ > 1, the solutions have a slight oscillatory character. In the viscoelastic case (η 6= 0), equation (76) is solvable in
terms of special functions and a richer variety of solutions emerges, because
now the solutions are not necessarily unbounded. Moreover, although they
are damped solutions, they do not necessarily vanish with time. In order to
solve equation (76) explicitly, we introduce the function f defined by
φ(ζ) = e−Υζ f (2Υζ),
(79)
and we find that it satisfies a Whittaker equation:
Γ
1
′′
f (2Υζ) = 0.
f (2Υζ) + − +
4 4(2Υζ)2
(80)
It follows that the solution to (76) is expressed in terms of Whittaker’s functions as:
h
i
φ(ζ) = e−Υζ k1 W0, 1 √1−Γ (2Υζ) + k2 M0, 1 √1−Γ (2Υζ) ,
(81)
2
2
18
φ(t)
C1 = 1.5
3
2
1
C1 = 1
0
20
40
60
80
100
120
140
t
–1
C1 = 0.2
Figure 1: Influence of the stretch rate C1 in the X-direction on the shearing
motion solutions for a Blatz-Ko material (foamed, polyurethane elastomer).
where k1 and k2 are integration constants. To illustrate the influence of the
stretch rate in the X-direction (through the constant C1 ) upon the behavior
of the solutions, we consider that the material is unstretched at t = 0 (so
that C2 = 1 and φ(t = 0) = 0); we take φ′ (t = 0) = 1, µk 2 /ρ0 = 1, η = µ/10
(so that Υ = 1/(20C1 ) and Γ = 4/C12 ); and we let C1 = 0.2, 1, 1.5. Figure
1 shows that for a “small” rate, the solution undergoes a few oscillations
before it vanishes completely; as C1 increases, the oscillations disappear but
the solution tends to an increasing finite asymptotic value. We argue that
this behavior can be related to the problem of a tensile impact: when the
material is plucked at “low” speed, then a few oscillations take place before
the material returns to its original state; when the material is plucked at
“high” speed, then the material can accomodate an asymptotic transverse
wave.
Now, consider the generalized sinusoidal standing waves of Section III.B.
As seen there, they are governed by nonlinear ordinary differential equations
when the longitudinal displacement u is such that uX = U = U (X), a
19
function of X only. This condition leads to (53), here integrated twice as
′
′
(φ 2 + ψ 2 − 1)U −3 + U
(β−2)
β
= C3 X + C4 ,
(82)
where C3 , C4 are disposable constants. Taking the typical value β = 2
corresponding to Poisson ratio ν = 1/4, we find
′
′
φ2+ψ2−1
U=
C3 X + C4 − 1
31
.
(83)
Hence for this material, the governing equations (54) for the transverse solutions of type (48) at θ(X) = 0 are, using (83) and (73)1 ,
′
′2
− 32
′2
µ φ + ψ − 1
φ′ + ρ0 ω 2 φ = −ηωψ ′′ ,
C3 X + C4 − 1
′
′2
− 32
′2
µ φ + ψ − 1
(84)
ψ ′ + ρ0 ω 2 ψ = ηωφ′′ .
C3 X + C4 − 1
The final governing equations are highly nonlinear, even with the choice
β = 2. The possibility of solving them in closed form seems quite remote. Of
course, numerical and qualitative analyses may be performed with the usual
methods of dynamical systems theory.
4.3
A separable strain energy density
Recall that the strain energy function of an incompressible material, Σinc say,
e
corresponds to the restriction to the subspace (I1 , I2 , 1) of a strain-energy Σ
say, defined in the full space (I1 , I2 , I3 ):
e 1 , I2 , 1).
Σinc (I1 , I2 ) := Σ(I
(85)
Σinc = µ(I1 − 3)/2,
(86)
Hence the classical incompressible neo-Hookean form,
may be associated for example with the full-space strain energy
p
e = µ (I1 − 3 − 2 ln J), J := I3 ,
Σ
2
20
(87)
often used in the classical molecular theory of rubber [26].
Now, constitutive equations for compressible hyperelastic materials come
in many formulations. One of them consists in adding a purely volumetric
e whose restriction
term Σvol (J) say, to a basic strain energy density function Σ,
(85) is the strain energy function of an incompressible material Σinc . Then
the final strain energy density for a compressible hyperelastic material can
be written as
e 1 , I2 , I3 ) + Σvol (J).
Σ = Σ(I
(88)
Several models have have been proposed in the literature for this pure
volumetric part (or bulk term) of the strain energy function. Ogden [48]
proposed the form:
ΣIvol (J) = λβ −2 (β ln J + J −β − 1),
(89)
where λ is the first Lamé modulus and β (> 0) is an empirical parameter.
Flory [27] proposed the form
2
ΣII
vol (J) = (c/2)(ln J) ,
(90)
and Simo and Pister [56], the form
2
ΣIII
vol (J) = (c/2)[(J − 1) − ln J],
(91)
which is (89) at β = −2. Recently Bischoff et al. [5] proposed the form
ΣIV
vol (J) =
c
[cosh β(J − 1) − 1].
β2
(92)
In this manner, several strain energy functions may be constructed. As
an example, consider the power-law function of Knowles [40],
n
C
b
Σinc =
1 + (I1 − 3) − 1 ,
(93)
b
n
where C, b, and n are constitutive parameters, all three assumed positive.
This simple strain energy captures several important features of rubber-like
materials. At n = 1, it recovers the neo-Hookean form (86); at n > 1 it
describes strain-stiffening materials; at n < 1, strain-softening materials.
These hyperelastic properties prove crucial to the understanding of complex
phenomena such as dynamic fracture, as shown recently by Buehler et al [7].
21
Then, using the formulation exposed above, we may construct the following
compressible strain energy function
n
C
b
Σ=
(94)
1 + (I1 − 3) − 1 − C ln J + Σvol (J).
b
n
Turning
back to the finite amplitude motions of Section II, we find that
√
J = I3 = uX = U , so that Σvol (J) = Σvol (U ), and the functions Q1 , Q2
defined in (16) are here,
b 2
(U + Ω2 − 1)]n−1 ,
n
Q2 = U −1 Σ′vol (U ) − CU −2 .
Q1 = 2C[1 +
(95)
In general, the corresponding equations of motion are coupled and rather
involved. Consider the special case of the standing waves (42), with the
choice U (t) = C2 (C1 = 0 in (36)). Then φ(t) satisfies (43), which is here
φ′′ +
k2C
b
k2η ′
φ +2
[1 + (C22 − 1 + k 2 φ2 )]n−1 φ = 0.
ρ0
ρ0
n
(96)
We notice that this equation may admit not only the static solution:
φ := 0, but also the nontrivial static solutions:
q
φ = ±(1/k) 1 − C22 − n/b.
(97)
These solutions are real if and only if b(1 − C22 ) − n > 0: when the body is
compressed in the X-direction (u = C2 X, C2 < 1), then the solution is real
if b > n/(1 − C22 ) > 0; when the body is stretched (u = C2 X, C2 > 1), the
solution is not real (recall that Knowles [40] assumed that b > 0 in (93)).
Another important general property of equation (96) is that it can be
derived [57] from Lagrangian
k2 η
t
L = 21 e ρ0 [φ′2 − V(φ)],
(98)
where
Z
k2C φ
b
V(φ) = 2
[1 + (C22 − 1 + k 2 ζ 2 )]n−1 ζdζ
ρ0 0
n
C
b
=
[1 + (C22 − 1 + k 2 φ2 )]n .
ρ0 b
n
22
(99)
As an example we consider a strain-stiffening material, by taking n = 2
in (94). Then (96) is a Duffing equation with damping,
φ′′ + δφ′ − βφ + αφ3 = 0,
where
δ :=
k2η
,
ρ0
β :=
k2C
[b(1 − C22 ) − 2],
ρ0
(100)
α := b
k4C
.
ρ0
(101)
Clearly, α > 0, δ > 0. To make the connection with results by Holmes [37], we
impose β > 0 also, which happens only when the body is compressed in the
X-direction (u = C2 X, C2 < 1) and when b > 2/(1 − C22 ). As seen for (100),
this equation possesses three
p fixed points in the phase plane, namely: (0, 0),
which is a saddle, and (± β/α, 0), which are two sinks. Holmes [37] showed
that equation (101) is locally and globally stable. It follows that as t → ∞,
our standing waves must approach one of the four wrinkled configurations
p
lim v(X, t) = ± β/α cos(kX − θ0 ),
t→∞
p
lim w(X, t) = ± β/α sin(kX − θ0 ),
(102)
t→∞
for almost any initial conditions. As any one of the four final configurations
is equivalent to another, we conclude that the material is destabilized by
the waves (only some special choices of the initial conditions will lead to the
unstressed deformation corresponding to the saddle point (0,0); these special
initial conditions are found by computing the corresponding stable manifold
in the phase plane, and we refer to Holmes [37] for the details.)
The situation described here can be extended to all positive integers n,
because the leading of nonlinearity of the potential V(φ) in (99) is φ2n so
that the leading order of nonlinearity in the resulting differential equation
is always to an odd integer power. We also point out that for a given b, an
increasing n requires a decreasing C2 for the appearance of the fixed points
(97) that is, a greater compressive stretch in the X-direction. This is in
agreement with the “physical” expectation that the stiffer a material is, the
harder it is to destabilize it.
For non-integer powers n, the situation is more complex because destabilizing configurations may or may not appear in the case of strain-hardening
materials (n > 1) and in the case of strain-softening materials (n < 1).
Obviously the dynamical systems uncovered here are susceptible to a more
complete and more careful analysis than the one provided, which may reveal
23
even richer behaviors than those already described, such as for example the
appearance of parametric resonance; these aspects are however beyond the
scope of the present paper.
Finally, we note that the instability evoked in this Subsection is somewhat
reminiscent of the buckling of an elasticita under a compressive load, but we
point out several differences. The buckling of an elastica is a phenomenon
related to geometrical non-linearities, whereas our instability is due to constitutive non-linearities; in an elastica, more than one inflexion may occur
whereas here, we could multiply instabilities only by considering a multiple
well strain energy function (and such a strain energy is not usual for the
modeling of soft tissues and elastomers); for an elastica, the stability of the
various fixed points must be examined thoroughly, whereas here, the stability
of such points comes out directly from our dynamical approach.
4.4
Fourth-order elasticity
In the linearized theory of “small-but-finite” amplitude waves, the strain
energy function must be expanded up to the fourth order in the strain, in
order to reveal nonlinear shear waves [59]. In that framework, Murnaghan’s
expansion [46] is often used (see for instance Porubov [52]):
λ + 2µ 2
l + 2m 3
i1 − 2µi2 +
i1 − 2mi1 i2 + ni3 + ν1 i41 + ν2 i21 i2 + ν3 i1 i3 + ν4 i22 ,
2
3
(103)
where λ, µ are the Lamé moduli, l, m, n are the third-order moduli, and
ν1 , ν2 , ν3 , ν4 are the fourth-order moduli (other notations exist: see Norris
[47] for the connections). In this expansion, we used the first three principal
invariants i1 , i2 , i3 of E, the Green-Lagrange strain tensor; they are related
to the first three principal invariants I1 , I2 , I3 of B defined in (4) by the
relations [4],
Σ=
I1 = 2i1 + 3,
I2 = 4i1 + 4i2 + 3,
I3 = 2i1 + 4i2 + 8i3 + 1.
(104)
By chain rule differentiation, we find that the functions Q1 and Q2 defined
24
in (16) are here,
ν2 2
)i
2 1
− (2m + ν4 )i2 + 4ν1 i31 + 2ν2 i1 i2 + ν3 i3 ,
ν2
Q2 = −µ − mi1 + i21 + ν4 i2 ,
2
Q1 = µ + (λ + 2µ + m)i1 + (l + 2m −
(105)
where now i1 , i2 , i3 are found from (104) and (21) as
i1 = (U 2 + Ω2 − 1)/2,
i2 = −Ω2 /4,
i3 = 0.
(106)
Several comments are in order at this point. First of all, we recall that
the equations corresponding to (15) for third-order elasticity (without dissipation) were first derived by Gold’berg [29] in 1960.
Then we note that recently there has been a renewed interested in fourthorder elasticity following experiments on the nonlinear acoustic properties of
soft tissue-like solids [20]. Hence Hamilton et al. [32] proposed the following
reduced version of the expansion (103), suitable when the relative portion of
energy stored in compression is negligible, as is expected for shear deformations and motions of soft tissues,
Σ = −2µi2 + ni3 + ν4 i22 .
(107)
In this case, the functions Q1 and Q2 defined in (16) simplify to
Q1 = −Q2 = µ − ν4 i2 = µ + ν4 Ω2 /4,
(108)
and we check at once that the longitudinal equation of motion (15)1 becomes
a trivial identity in the purely elastic case (ζ = χ = η = 0), because here
Q1 + Q2 = 0.
Finally, we also notice that the theories of third- and fourth-order elasticity have been used to study bulk solitons propagation in elastic materials.
For instance, Hao and Marris [33] discuss the issue of acoustic solitons from
both experimental and theoretical points of view; they consider the possibility to produce KdV solitons by adding a fourth-order spatial derivative to the
longitudinal wave equation and by performing some ad hoc approximations.
The general expressions (103), (105), and (106) yield quite complicated
equations of motion. If for example we restrict our attention to the standing
waves (42), we find that the polynomial nonlinearity of the resulting ordinary
25
differential equation is of the fifth order for third-order elasticity, and of
seventh order for fourth-order elasticity. If we further restrict our attention to
the reduced expansion (107) we find that equation (43) reduces to a damped
Duffing equation,
φ′′ +
ηk 2 ′
k2
φ +
(4µ + k 2 ν4 φ2 )φ = 0.
ρ0
4ρ0
(109)
This equation is characterized by a positive linear stiffness and therefore
the peculiar behavior found in the previous sub-section for some power-law
materials is ruled out. For completeness, we point out that Holmes and Rand
[38] computed an approximate solution of equation (109), using the method
of averaging.
5
Concluding Remarks
We studied the propagation of finite amplitude waves in a general nonlinear
elastic material, with dissipation taken into account by means of a simple
mechanism of differential type. We showed that the general equations of motion admit some beautiful and most interesting reductions to ordinary differential equations, using a direct method based on complex functions. Clearly,
these reductions are a consequence of the inherent symmetrical structure of
the balance and constitutive equations and may therefore be recovered also
by the standard methods of group analysis; we argue however that our direct
method is more simple and more revealing from a mechanical point of view.
Moreover, our results complement and generalize the well-known results of
Carroll in various directions, as already pointed out in the Introduction;
they also emphasize the analogy between Carroll’s solutions and some exact
solutions proposed for nonlinear strings by Rosenau and Rubin [53].
The solutions provided here can be used not only as benchmarks for the
more complicated numerical analysis required in “real life” applications, but
also for a better understanding of the mechanical properties and of the mathematical structure of various usual models. These advantages were highlighted
in the discussion on standing waves solutions conducted in Section III. There,
we focused much of the discussion on the special case of standing waves for
the sake of simplicity and brevity, but it is clear that our general methods
of investigation can be applied to all the ordinary differential equations that
were derived.
26
One of the important findings uncovered by the analysis of our solutions
is the possibility to predict the appearance of highly symmetric wrinkles in
an elastic medium compressed longitudinally. This phenomenon has been
observed experimentally by several researchers when, for example, a hard
film is deposited on a soft material and put under compressive strain [39].
Our analysis showed that these wrinkles may indeed appear when the compression parameter satisfies some simple inequalities; it also showed that the
detailed viscoelastic behavior of the material is unimportant because of the
highly attracting nature of the wrinkled states of deformations. We note of
course that our solutions have a higher symmetry than the patterns usually
observed in experiments, probably because our solutions are characterized
by strong invariance and because they are bulk solutions, in contrast to the
half-space geometry generally used in experiments. Despite these limitations,
we found that these solutions are compatible with the relatively simple nonlinear power-law strain energy density whereas weak non-linear theories (up
to the fourth-order) are not adequate to predict these features.
Our approach allowed a direct and complete comparison of the models
used in nonlinear acoustics and in continuum mechanics. It was seen that
although the popular weak nonlinear theories are often used to study some
special features of wave propagation, a consequence of their polynomial nature is that they cannot contain all the necessary information of the general
constitutive theory.
Last but not least, we gave some explicit solutions to some models of
compressible hyperelasticity. New exact solutions are always welcomed and
valuable additions to the short list of exact solutions found in the literature.
We conclude by pointing out that several generalizations of our results
are possible in principle. For instance, the consideration in polar coordinates
of a wave propagating in the radial direction coupled to two shear waves in
the axial and azimuthal direction is one of such possible generalizations. In
this case the following class of motions is considered
r = r(R, t),
θ = Θ + g(R, t),
z = Z + u(R, t),
(110)
where (R, Θ, Z) are cylindrical polar coordinates associated with the undeformed configuration and (r, θ, z) are cylindrical polar coordinates associated
with the deformed configuration. It is possible to show that from (110) and
the corresponding kinematical quantities of interest, the equations of motion
may be reduced to a set of three partial differential equations, similar to
27
what we did for the motions (10) in (15). Unfortunately, it seems that in
this case the equations of motion do not allow general solutions in separable
form, as was the case in Section III for the equations (15)). Of course, we do
not exclude that for special classes of materials, it is possible to find some
classes of wave solutions (see for example [31] for compressible materials and
[28] for incompressible materials).
References
[1] S.S. Antman, Nonlinear Problems of Elasticity (Springer, New York,
2004).
[2] S.S. Antman and T.I. Seidman, The parabolic-hyperbolic system governing the spatial motion of nonlinearly viscoelastic rods, Arch. Ration.
Mech. Analysis 175 (2005) 85–150.
[3] P.J. Blatz and W.L. Ko, Application of finite elasticity to the deformation of rubbery materials, Trans. Soc. Rheol. 6 (1962) 223–251.
[4] M.F. Beatty, Topics in finite elasticity: Hyperelasticity of rubber, elastomers, and biological tissues - with examples, Appl. Mech. Rev. 40
(1987) 1699–1733.
[5] J.E. Bischoff, E.M. Arruda, and K. Grosh, A new constitutive model for
the compressibility of elastomers at finite deformations, Rubber Chem.
Tech. 74 (2001) 541–559.
[6] Ph. Boulanger, M. Hayes, and C. Trimarco, Finite-amplitude waves in
deformed Hadamard elastic materials, Geophys. J. Int. 118 (1994) 447–
458.
[7] M.J. Buehler, F.F. Abraham, and H. Gao, Hyperelasticity governs dynamic fracture at a critical length scale, Nature 426 (2003) 141–146.
[8] G.P. Carman and M.S. Cramer, Nonlinearity parameters for pulse propagation in isotropic elastomers, J. Acoust. Soc. Am. 91 (1992) 39–51.
[9] M.M. Carroll, Some results on finite amplitude elastic waves, Acta Mech.
3 (1967) 167–181 .
28
[10] M.M. Carroll, Oscillatory shearing of nonlinearly elastic solids, Z. angew.
Math. Phys. 25 (1974) 83–88 .
[11] M.M. Carroll, Plane elastic standing waves of finite amplitude, J. Elast.
7 (1977) 411–424.
[12] M.M. Carroll, Finite amplitude standing waves in compressible elastic
solids, J. Elast. 8 (1978) 323–328 .
[13] M.M. Carroll, Reflection and transmission of circularly polarized elastic
waves of finite amplitude, J. Appl. Mech. 46 (1979) 867–872.
[14] M.M. Carroll, Unsteady homothermal motions of fluids and isotropic
solids, Arch. Ration. Mech. Analysis 53 (1974) 218–228.
[15] M.M. Carroll, Plane circular shearing of incompressible fluids and solids,
Q. J. Mech. Appl. Math. 30 (1977) 223–234.
[16] J.H. Curran and M.M. Carroll, Sinusoidal shearing of liquid crystals,
Int. J. Eng. Sci. 18 (1980) 971–977.
[17] M.M. Carroll, On circularly-polarized nonlinear electromagnetic waves,
Q. Appl. Math. 25 (1967) 319–323.
[18] M.M. Carroll and M.F. McCarthy, Finite amplitude wave propagation in
magnetized perfectly electrically conducting elastic materials, In: Elastic Wave Propagation, M.F. McCarthy and M.A. Hayes, eds. (Elsevier,
North-Holland, 1989) pp. 615–621.
[19] C. Carstensen and G. Dolzmann, Time-space discretization of the nonlinear hyperbolic system utt =div(σ (Du) + Dut ), SIAM J. Numer.
Analysis 42 (2004) 75–89.
[20] S. Catheline, J.L. Gennisson, and M. Fink, Measurement of elastic nonlinearity of soft solid with transient elastography, J. Acoust. Soc. Am.
114 (2003) 3087–3091.
[21] P. Chadwick, Continuum Mechanics (Dover, New York, 1999).
[22] P.J. Chen, Growth and decay of waves in solids, in: C.A. Truesdell (ed.)
Handbuch der Physik, Vol. VIa/3 (Springer, New York, 1973).
29
[23] C.M. Dafermos, Hyperbolic Conservation Laws in Continuum Physics,
(Springer, New York, 2000).
[24] M. Destrade and G. Saccomandi, Finite amplitude inhomogeneous waves
in Mooney-Rivlin viscoelastic solids, Wave Motion 40 (2004) 251–262.
[25] M. Destrade and G. Saccomandi, Some results on finite amplitude elastic
waves propagating in rotating media, Acta Mech. 173 (2004) 19–31.
[26] B. Erman, Molecular aspects of rubber elasticity, in: G. Saccomandi
and R.W. Ogden (eds.) Mechanics and Thermomechanics of Rubberlike
Solids, CISM Lecture Notes 452 (Springer, New York, 2004).
[27] P.J. Flory, Thermodynamic relations for high elastic materials, Trans.
Faraday Soc. 57 (1961) 829–838.
[28] R.L. Fosdick and G.P. Macsithigh, Shearing motions and the formation
of shocks in an elastic circular tube, Q. Appl. Math. 38 (1980) 191–207.
[29] Z.A. Gold’berg, Interaction of plane longitudinal and transverse wave,
Akusticheskii Zhurnal 6 (1960) 307–310.
[30] J. Hadamard, Leçons sur la Propagation des Ondes et les Equations de
l’Hydrodynamique (Hermann, Paris, 1903).
[31] J.B. Haddow and R.J. Tait, A note on the propagation of finite amplitude shear waves on a hyperelastic solid, ASME J. Appl. Mech. 53
(1986) 470–471.
[32] M.F. Hamilton, Y.A. Ilinski, and A.A. Zabolotskaya, Separation of
compressibility and shear deformation in the elastic energy density, J.
Acoust. Soc. Am. 116 (2004) 41–44.
[33] H.Y. Hao and H..J. Marris, Experiments with acoustic solitons in crystalline solids, Phys. Rev. B 64 (2001) 064302.
[34] M.A. Hayes and G. Saccomandi, Finite amplitude transverse waves in
special incompressible viscoelastic solids, J. Elast. 59 (2000) 213–225.
[35] M.A. Hayes and G. Saccomandi, Finite amplitude waves superimposed
on pseudoplanar motions for Mooney-Rivlin viscoelastic solids, Int. J.
Nonlinear Mech. 37 (2002) 1139–1146.
30
[36] M.A. Hayes and G. Saccomandi, Antiplane shear motions for viscoelastic
Mooney-Rivlin materials, Q. J. Mech. Appl. Math. 57 (2004) 379–392.
[37] P. Holmes, A nonlinear oscillator with a strange attractor, Phil. Trans.
Roy. Soc. London A 292 (1979) 419–448.
[38] P. Holmes and R. Rand, The bifurcations of Duffing’s equation: an
application of catastrophe theory, J. Sound Vibr. 44 (1976) 237–253.
[39] Z. Huang, W. Hong, and Z. Suo, Evolution of wrinkles in hard films on
soft substrates, Phys. Rev. E 70 (2004) 030601.
[40] J. Knowles, The finite anti-plane shear field near the tip of a crack for a
class of incompressible elastic solids, Int. J. Fract. 13 (1977) 611–639.
[41] O. Kramer, S. Hvidt, and J.D. Ferry, Dynamical mechanical properties
in: J.M. Mark, B.Erman, and F. Eirich (eds.) Science and Technology
of Rubber (Academic Press, San Diego, 1994).
[42] F. John, Plane elastic waves of finite amplitude. Hadamard materials
and Harmonic materials, Comm. Pure Appl. Math. 19 (1966) 309–341.
[43] L.D. Landau and E.M. Lifshitz, Theory of Elasticity (Pergamon, New
York, 1986).
[44] M. Levinson and I.W. Burgess, A comparison of some simple constitutive
relations for slightly compressible rubber-like materials, Int. J. Mech.
Sci. 13 (1971) 563–572.
[45] G.A. Maugin, Nonlinear Waves in Elastic Crystals (University Press,
Oxford, 1999).
[46] F.D. Murnaghan, Finite Deformation of an Elastic Solid (John Wiley:
New York, 1951).
[47] A. Norris, Finite amplitude waves in solids, In: M.F. Hamilton and
D.T. Blackstock (eds.) Nonlinear Acoustics (Academic Press, San Diego,
1999) pp. 263–277.
[48] R.W. Ogden, Large deformation isotropic elasticity: on the correlation
of theory and experimental for compressible rubberlike solids, Proc. R.
Soc. Lond. A 328 (1972) 567–583.
31
[49] R.W. Ogden, G. Saccomandi, and I. Sgura, Fitting hyperelastic models
to experimental data, Comput. Mech. 34 (2004) 484–502.
[50] K.J. Parker and R.M. Lerner, Sonoelasticity of organs: shear waves ring
a bell, J. Ultras. Med. 11 (1992) 387–392.
[51] E. Pucci and G. Saccomandi, A note on the Gent model for rubber-like
materials, Rubber Chem. Tech. 75 (2002) 839–851.
[52] A.V. Porubov, Amplification of Nonlinear Strain Waves in Solids
(World Scientific, Singapore, 2003).
[53] P. Rosenau and M.B. Rubin, Motion of a nonlinear string: some exact
solutions to an old problem, Phys. Rev. A, 31 (1985) 3480–3482.
[54] P. Rosenau and M.B. Rubin, Some nonlinear three-dimensional motions
of an elastic string, Physica D 19 (1986) 433–439.
[55] B.K. Sinha and K.W. Winkler, Formation nonlinear constants from sonic
measurements at two borehole pressures, Geophys. 64 (1990) 1890–1900.
[56] J.C. Simo and K.S. Pister, Remarks on rate constitutive equations for
finite deformations, Computer Meth. Appl. Mech. Eng. 46 (1984) 201–
215.
[57] W.H. Steeb and A. Kunick, Lagrange functions of a class of dynamical
systems with limit-cycles and chaotic behavior, Phys. Rev. A 25 (1982)
2889–2892.
[58] C. Truesdell, General and exact theory of waves in finite elastic strain,
Arch. Rat. Mech. Analysis 8 (1961) 263–296.
[59] E.A. Zabolotskaya, Sound beams in a nonlinear isotropic solid, Sov.
Phys. Acoust. 32 (1986) 296–299.
32