CanCNSM
Vancouver
June 19-23, 2002
SURFACE STABILITY ANALYSIS OF A PREDEFORMED
BELL-CONSTRAINED HALF-SPACE
Michel Destrade
Texas A&M University
Department of Mathematics
College station, TX 77843-3368, USA.
e-mail: destrade@math.tamu.edu
Key Words: Predeformed semi-infinite medium, Bell constraint, surface waves, exact
secular equation, near-the-surface instability, bifurcation criterion.
Abstract.
Near-the-surface instability is examined for a finitely and homogeneously deformed hyperelastic semi-infinite body subject to the Bell constraint, by using surface (Rayleigh) waves.
These inhomogeneous plane waves propagate in the direction of a principal axis of the
finite homogeneous static deformation, and decay exponentially away from the free flat
surface of the half-space, in the direction of another principal axis. The exact secular
equation, giving the speed of propagation v, is found. Then by letting v tend to zero, the
‘bifurcation criterion’ or ‘stability equation’ is established; this equation delimits a surface
in the stretch ratios space which separates a region where the deformed half-space is always
stable with respect to perturbations from a region where the half-space might be unstable.
Finally two specific examples of materials are treated: ‘Bell’s empirical model’ and ‘Bell
simple hyperelastic material’, and it is seen that in each case, the stability equation is
universal to the whole subclass considered.
CanCNSM 2002, Vancouver
1
M. Destrade
INTRODUCTION
According to the Introduction of a recent textbook by Guz,1 the three-dimensional theory
of deformable bodies stability can be traced back to the early works of Biot2, 3 in the 1930’s,
at least as far as incremental deformations of finitely prestressed bodies are concerned.
In a series of articles, most of which are summarized in his book,4 Biot analyzed both the
internal and the near-the-surface instabilities of homogeneous as well as heterogeneous
prestressed elastic and viscoelastic materials. He pointed out that the study of surface
instability is formally analogous to the study of surface (Rayleigh) waves.
Surface waves are indeed a valuable tool when the stability of an elastic half-space x 2 ≥ 0
(say) comes under study. This is so because they may be decomposed into a combination
of inhomogeneous plane waves, propagating with speed v on the flat surface x2 = 0, in a
direction x1 (say), and with attenuation in the x2 direction, so that they are perturbations
of the form
ǫℜ{Aeik(x1 +ξx2 −vt) }.
(1)
Here, ǫ is a real parameter, ‘small’ in the sense that terms of order higher than ǫ 1 may
be neglected; A is a complex vector, describing the polarization of the wave; k is the real
wave number; and ξ is a complex scalar such that ℑ(ξ) > 0, in order for the amplitude
to vanish away from the surface x2 = 0. Now consider a semi-infinite body, made of
some hyperelastic material, maintained in an equilibrium state of finite pure homogeneous
deformation under uniform compressive (or tensile) loads P1 and P3 . Let x1 , x2 , x3 , be
the three principal axes of the deformation, x2 = 0 be the flat surface delimiting the
upper end of the body, and let a wave such as (1) propagate near this surface. Once the
equations of motion are solved and the boundary conditions are satisfied, an equation
– the secular equation – is obtained for v 2 . The parameters in this equation depend
on the material constants characteristic of the elastic body and on the principal stretch
ratios of the primary static deformation. For a given configuration, a unique positive
root of the secular equation is usually found for v 2 , and the configuration is said to be
stable with respect to a perturbation of the form (1), or using Fourier analysis, with
respect to any dynamical perturbation. However, it may happen that for certain stretch
ratios, the secular equation admits v 2 = 0 as a root. This phenomenon corresponds to the
vanishing of the apparent surface rigidity4 and is generally associated with the appearance
of ‘ripples’ or ‘wrinkles’ on the free surface of the half-space. Furthermore, there might
exist some configurations where the secular equation admits only negative roots v 2 < 0,
leading to a perturbation (1) increasing exponentially with time and rendering the linear
approximation of small perturbations superimposed on large invalid. The secular equation
written at v 2 = 0 is usually called the stability equation or the bifurcation criterion. It
defines a surface in the stretch ratios space which, in the linearized theory, separates stable
configurations of the half-space from unstable configurations.
Using this background, the present article establishes the secular equation for surface
CanCNSM 2002, Vancouver
M. Destrade
waves on a deformed Bell-constrained half-space in order to study its stability. The
Bell constraint was obtained experimentally by James F. Bell5 for the behavior of certain
annealed metals and has since been extensively studied in a theoretical manner within
the context of hyperelasticity (see Beatty6 and the references therein). Note that the
secular equation for this problem was recently found7 using the method of first integrals
proposed by Mozhaev.8 This method is rapid and elegant but presents a minor inconvenient because the secular equation is obtained in a polynomial form (a cubic in v 2 ) which
corresponds to the rationalization of the exact secular equation and the introduction of
spurious roots.9, 10 Consequently, there is a need to obtain the exact secular equation,
from which the exact bifurcation criterion may be deduced. This procedure is carried on
in Section 3, using previously established results recalled in Section 2. Then in Section 4,
the analysis is specialized to two specific classes of Bell-constrained materials. First, for
‘Bell’s empirical model’, whose strain energy function depends upon only one material parameter, the bifurcation criterion found by Beatty and Pan11 is quickly recovered; second,
for a ‘simple hyperelastic Bell material’, whose strain energy function depends upon two
material parameters, the bifurcation criterion is found to be a very simple relationship between the stretch ratios. In both cases, the bifurcation criterion is universal to the whole
subclass considered, as it does not depend upon any material constant. In the latter case,
the influence of the prestrain on the speed of the Rayleigh waves and on the stability of
the half-space is discussed. Finally, comparisons are made with the stability of half-spaces
subject to the constraint of incompressibility (which coincides with the Bell constraint in
the isotropic limit12 ), specifically with neo-Hookean and Mooney-Rivlin materials, which
model the behavior of rubber.
2
2.1
STABILITY EQUATIONS
Basic deformation of a Bell half-space
Let (O, x1 , x2 , x3 ) ≡ (O, i, j, k) be a Cartesian rectangular coordinate system. Let the halfspace x2 ≥ 0 be occupied by a hyperelastic Bell-constrained material, with strain energy
density Σ. This material is subject to the internal constraint that for any deformation, 12, 13
i1 ≡ tr V = 3,
(2)
at all times, where V is the left stretch tensor. Hence, for isotropic Bell materials,
Σ depends only upon i2 and i3 , the respective second and third invariants of V. So,
Σ = Σ(i2 , i3 ), where
i2 = [(tr V)2 − tr (V2 )]/2, i3 = det V,
(3)
and the constitutive equation giving the Cauchy stress tensor T is12
T = pV + ω0 1 + ω2 V2 ,
(4)
CanCNSM 2002, Vancouver
M. Destrade
where p is an arbitrary scalar, to be found from the equations of motion and the boundary
conditions, and the material response functions ω0 and ω2 are defined by
ω0 =
∂Σ
,
∂i3
ω2 = −
1 ∂Σ
,
i3 ∂i2
(5)
and should verify the Beatty–Hayes A-inequalities12
ω0 (i2 , i3 ) ≤ 0,
ω2 (i2 , i3 ) > 0.
(6)
In the case where the material is maintained in a state of finite pure homogeneous static
deformation, with principal stretch ratios λ1 , λ2 , λ3 along the x1 , x2 , x3 axes, the Cauchy
stress tensor is the constant tensor To given by
To = (po λ1 + ω0 + λ21 ω2 )i ⊗ i + (po λ2 + ω0 + λ22 ω2 )j ⊗ j + (po λ3 + ω0 + λ23 ω2 )k ⊗ k. (7)
Here ω0 and ω2 are evaluated at i2 , i3 given by
i2 = λ 1 λ 2 + λ 2 λ 3 + λ 3 λ 1 ,
i 3 = λ 1 λ2 λ3 .
(8)
Of course,
λ1 + λ2 + λ3 = 3,
(9)
in order to satisfy (2). It is assumed that the boundary x2 = 0 is free of tractions so that
To22 = 0; and that the compressive loads P1 and P3 are applied at x1 = ∞ and x3 = ∞
to maintain the deformation so that P1 = −To11 and P3 = −To33 . Hence,
po = −(ω0 + λ22 ω2 )/λ2 ,
2.2
Pγ = (λ2 − λγ )(−ω0 + λγ λ2 ω2 )/λ2 ,
(γ = 1, 3).
(10)
Incremental near-the-surface motions
Beatty and Hayes14 wrote the general equations for small-amplitude motions in a Bellconstrained material maintained in a state of finite pure homogeneous deformation (as described in the previous subsection). These equations were then specialized by this author
to surface (Rayleigh) waves. For a wave of the form (1), written as ǫℜ{U(x 2 )eik(x1 −vt) },
where U is an unknown function of x2 , it was proved that the equations of motion and
the boundary conditions relative to surface waves could be written in a quite simple man∗
∗
ner in terms of the incremental tractions σ12
, σ22
, acting upon the planes x2 = const.
Explicitly, and introducing the scalars functions t1 (x2 ) and t2 (x2 ), defined by
∗
σ21
(x1 , x2 , t) = t1 (x2 )eik(x1 −vt) ,
∗
σ22
(x1 , x2 , t) = t2 (x2 )eik(x1 −vt) ,
(11)
it was found that the equations of motion are
1
2
(λ1 λ−1
2 C − X)(µλ1 − X)t1 = 0,
µλ22
2
′′
′
2 −2
2
(λ1 λ−1
2 C − X)t2 + iβ12 t1 − λ1 λ2 [µ(λ1 − λ2 ) − X]t2 = 0,
[µ(λ21 − λ22 ) − X]t′′1 + iβ12 t′2 −
(12)
CanCNSM 2002, Vancouver
M. Destrade
and the boundary conditions are simply
t1 (0) = t2 (0) = t1 (∞) = t2 (∞) = 0.
(13)
Here, X = ρv 2 where ρ is the mass density of the material, and
µ=
Cαβ
−ω0 + λ1 λ2 ω2
,
λ2 (λ1 + λ2 )
= 2λ2α δαβ ω2 − λ2β (ω02 + λ2α ω22 ) + λ1 λ2 λ3 (ω03 + λ2α ω23 ),
−1
2
2
C = λ−1
1 λ2 C11 + λ1 λ2 C22 − C12 − C21 − 2ω0 − (λ1 + λ2 )ω2 ,
2
2
−1
β12 = λ1 λ−1
2 [µ(λ1 − λ2 ) + C] − (1 + λ1 λ2 )X,
(14)
where the derivatives ω0Γ , ω2Γ (Γ = 2, 3) of the material response functions ω0 , ω2 are
taken with respect to iΓ and evaluated at i2 , i3 given by (8).
3
EXACT SECULAR EQUATION AND BIFURCATION CRITERION
Now a law of exponential decay is chosen for t(x2 ) = [t1 (x2 ), t2 (x2 )]T ,
t(x2 ) = eikξx2 T,
with ℑ(ξ) > 0,
where T is a constant vector. Then the incremental equations of motion (12) are
¸· ¸ · ¸
·
T1
0
α11 ξ 2 + γ11
β12 ξ
,
=
0
β12 ξ
α22 ξ 2 + γ22 T2
(15)
(16)
where β12 is defined in (14)4 and
α11 = [µ(λ21 − λ22 ) − X], α22 = (λ1 λ−1
2 C − X),
γ11 =
1
2
2 −2
2
2
(λ1 λ−1
2 C − X)(µλ1 − X), γ22 = λ1 λ2 [µ(λ1 − λ2 ) − X].
µλ22
(17)
Hence, nontrivial solutions exist when ξ is root of the biquadratic
ξ 4 − Sξ 2 + P = 0,
where
S=
2
β12
− α11 γ22 − α22 γ11
,
α11 α22
P =
γ11 γ22
(µλ21 − X)λ21
=
> 0.
α11 α22
µλ42
(18)
(19)
The quantity P is positive because, owing to the A-inequalities (6), µ defined in (14) 1
is positive, and because in the subsonic range, a surface wave propagates a speed which
is lower than that of any body wave, and in particular, lower than the pure shear wave
propagating along the x1 direction (X < µλ21 ).
CanCNSM 2002, Vancouver
M. Destrade
Now we examine in turn different situations that may arise in the resolution of the biquadratic (18): (a) S 2 − 4P > 0 and S > 0; (b) S 2 − 4P > 0 and S < 0; (c) S 2 − 4P < 0
and any real value for S (we exclude the particular case of a repeated root, because
it does not correspond to a surface wave15 ). Case (a) must be ruled out because it
leads to real values for all the roots and no exponential decay
wave as x2 → ∞.
qfor the √
In Case (b), the roots are all purely imaginary, and ξ1 = i (−S + S 2 − 4P )/2 and
q
√
ξ2 = i (−S − S 2 − 4P )/2 have positive imaginary parts. In Case (c), the roots with
q
√
positive imaginary parts are ξ1 = a + ib, ξ2 = −a + ib, where a = (S + 2 P )/4 and
q
√
b = (−S + 2 P )/4. We conclude that in both acceptable cases (b) and (c), we have
√
ξ1 ξ2 = − P .
(20)
Note that the different cases (a), (b), (c) written at X = ρv 2 = 0 are related to the
examination of the internal (or material) stability 4 of the material. These points were
treated by Pan and Beatty16 for Bell materials and are not considered here, as this paper
is concerned only with near-the-surface stability.
Now let T(r) be a vector satisfying (16) when ξ = ξr , (r = 1, 2). It may be written as
· ¸
α11 ξr2 + γ11
1
(r)
, where qr = −
, (r = 1, 2).
(21)
T =
qr
β12 ξr
Then the tractions t defined in (11) are a combination of T(1) and T(2) for some constants A and B, t(x2 ) = Aeikξ1 x2 T(1) + Beikξ2 x2 T(2) , and they must satisfy the boundary
conditions (13)1,2 , that is t(0) = 0, or
A + B = 0,
q1 A + q2 B = 0.
(22)
This linear homogeneous system of two equations for the two unknowns A and B has non
trivial solutions when q2 − q1 = 0, that is when the following exact secular equation is
verified, α11 ξ1 ξ2 − γ11 = 0, or using (17)1,3 , (20), and (19)2 ,
q
q
−1
2
2
2
(23)
[µ(λ1 − λ2 ) − X] µλ1 + (λ1 λ2 C − X) µλ21 − X = 0.
Note that in the process of obtaining the secular equation, the spurious roots corresponding to ξ2 − ξ1 = 0 were dropped. Note also that by a squaring process, the cubic secular
equation,7 which has spurious roots, may be deduced from (23).
Now we may deduce the bifurcation criterion directly from (23), by letting X tend to
zero, as
µ(λ21 − λ22 ) + λ1 λ−1
(24)
2 C = 0.
CanCNSM 2002, Vancouver
M. Destrade
When for certain stretch ratios λ1 , λ2 , λ3 , this criterion is satisfied, the predeformed
semi-infinite Bell medium looses its near-the-surface stability. This equation defines a
surface in the space of the stretch ratios which separates a region where the homogeneous
deformations of the Bell half-space are always stable (where ρv 2 > 0) from a region
where they might be unstable (where ρv 2 < 0, indicating temporal growth, at least in the
linearized theory). Of course, it must be kept in mind that the stretch ratios must always
satisfy the Bell constraint (9).
4
4.1
TWO ILLUSTRATIVE EXAMPLES
Bell’s empirical model
For Bell’s empirical model materials,5 the strain energy function Σ is given by
Σ = 23 β0 [2(3 − i2 )]3/4 ,
(25)
where β0 is a positive constant, and the material response functions ω0 and ω2 provided
by (5) are
1
ω0 = 0, ω2 = β0 [2(3 − i2 )]−1/4 .
(26)
i3
In that context,
λ1 ω2
(λ1 − λ2 )2
µ=
]λ1 λ2 ω2 .
(27)
, C = [2 −
λ1 + λ 2
4(3 − i2 )
Then the bifurcation criterion (24) may be arranged as
3−
(λ1 − λ2 )2
− λ−1
1 λ2 = 0,
4(3 − i2 )
(28)
which is precisely the stability equation (7.10) established by Beatty and Pan,11 using the
Euler stability criterion. It is independent of the material constant β0 . These authors
also conducted a detailed analysis of the regions delimited by the bifurcation criterion.
4.2
Simple hyperelastic Bell materials
For simple hyperelastic Bell materials,12 the strain energy function Σ is given by
Σ = α(3 − i2 ) + β(1 − i3 ),
(29)
where α and β are positive constants, and the material response functions ω0 and ω2
provided by (5) are now
ω0 = −β, ω2 = α/i3 .
(30)
In that context,
µ=
β + αλ−1
3
,
λ2 (λ1 + λ2 )
C = 2(β + αλ−1
3 ).
(31)
CanCNSM 2002, Vancouver
M. Destrade
3
1
0.8
scaled squared speed
2
λ3
1
0.6
0.4
0.2
0
0
1
3
0
0.2
0.4
2
λ1 2
1
3
0.6
0.8
1
1.2
1.4
stretch ratio
λ2
–0.2
(b) Surface wave speed
(a) Region of stability
Figure 1: Near-the-surface stability for simple hyperelastic Bell materials.
Then the bifurcation criterion (24) simplifies considerably to
3λ1 − λ2 = 0,
(32)
which is a particularly simple linear relationship between the stretch ratios λ 1 and λ2 ,
universal to the whole class of simple hyperelastic Bell materials. This equation delimits
a plane in the stretch ratios space (λ1 , λ2 , λ3 ), which cuts the constraint plane (9) along
the straight segment going from the point (0,0,3) to the point ( 34 , 49 , 0). Moreover, it will
become apparent in the foregoing analysis that the region which is stable with respect to
perturbations (where X = ρv 2 > 0) is: 3λ1 − λ2 > 0. In Figure 1(a), the plane (32) cuts
the triangle of the possible values for the stretch ratios (9) into two parts, of which the
visible one is the region of stability of any simple hyperelastic Bell material.
Regarding surface waves, a change of variable suggested by Dowaikh and Ogden9 may be
applied and the secular equation (23) may be written as a polynomial in η, defined by
1
X ¤1 h
(λ1 + λ2 )λ2
ρv 2 i 2
η = 1− 2 2 = 1−
,
µλ1
λ21
β + αλ−1
3
(33)
−2 2
f (η) ≡ η 3 + η 2 + (1 + 2λ−1
1 λ2 )η − λ1 λ2 = 0.
(34)
£
as
Clearly, at η = 0 (corresponding to a transverse bulk wave),
while at η = 1 (corresponding to v = 0), we have f (1) =
2
we have f (0) = −λ −2
1 λ2 < 0,
(3λ1 − λ2 )(λ1 + λ2 )λ−2
1 . So,
CanCNSM 2002, Vancouver
M. Destrade
because f is a monotone increasing function of η, the existence of a positive root for ρv 2 in
the subsonic interval [0, µλ21 ] is equivalent to the condition: 3λ1 − λ2 > 0 being satisfied.
In Figure 1(b), the influence of the prestrain upon the speed of the surface wave is made
apparent in the case of plane strain (λ3 = 1). On the horizontal axis, λ1 is increased
from a compressive value (λ1 < 1) to a tensile value (λ1 > 1). The coordinate on the
vertical axis is the squared surface wave speed, scaled with respect to the transverse bulk
wave speed, that is ρv 2 /(µλ21 ). At λ1 = 1, the half-space is isotropic (λ1 = λ2 = λ3 = 1)
and the scaled squared speed is equal to 0.9126, the value found by Lord Rayleigh17 in
the incompressible isotropic case (Beatty and Hayes12 have proved that the constraints
of incompressibility and of Bell are equivalent for infinitesimal motions). A compressive
load P1 > 0, λ1 < 1, will decrease the speed of propagation for the surface wave, until
the critical stretch of (λ1 )cr = 0.5 (see next subsection) where the surface looses stability.
Conversely, a tensile load P1 < 0, λ1 > 1, will increase the speed of propagation for the
surface wave, with the speed of the transverse bulk wave as an upper bound.
4.3
Comparisons with incompressible rubber
The stability of a deformed half-space made of incompressible rubber was first studied by
Biot. He used the neo-Hookean model but noted4 that his results were also valid for the
Mooney-Rivlin model. He obtained the bifurcation criterion, showed that it was universal
to both classes of materials, and computed the value of the critical stretch (λ 1 )cr at which
the rubber half-space looses stability under compressive loads, first in the case of plane
strain, then in the case of biaxial strain. Both cases involved the numerical resolution of
a cubic. Now we show that for Bell’s empirical models and for simple hyperelastic Bell
materials, the critical stretch can be found explicitly.
First we consider that the half-space made of Bell-constrained material is deformed in
such a way that there is no extension in the x3 -direction (λ3 = 1). This is possible when
the compressive load P3 = (λ2 − 1)(−ω0 + λ2 λ3 ω2 )/λ2 is applied at infinity. Then we have
λ2 = 2 − λ1 by (9). For Bell’s empirical model, the bifurcation criterion (28) reduces to
3λ1 − 2 = 0,
so that (λ1 )cr = 32 .
(35)
For simple hyperelastic Bell materials, the bifurcation criterion (32) reduces to
4λ1 − 2 = 0,
so that (λ1 )cr = 12 .
(36)
Then we consider that the half-space made of Bell-constrained material is allowed to
expand freely in the x3 -direction, so that P3 = 0. Then we have λ2 = λ3 = (3 − λ1 )/2 by
(10)3 and (9). For Bell’s empirical model, the bifurcation criterion (28) reduces to
11λ1 − 6 = 0,
so that (λ1 )cr =
6
.
11
(37)
CanCNSM 2002, Vancouver
M. Destrade
For simple hyperelastic Bell materials, the bifurcation criterion (32) reduces to
7λ1 − 3 = 0,
so that (λ1 )cr = 37 .
(38)
In Table 1, the numerical values for the critical stretches are given for the classes of
Bell’s empirical model (2nd column), of neo-Hookean and Mooney-Rivlin incompressible
materials4 (3rd column), and of simple hyperelastic Bell materials (4th column), in the
cases of plane strain (2nd row) and of biaxial strain (3rd row). It appears that rubber can
be compressed more than Bell’s empirical model but less than simple hyperelastic Bell
materials, before it looses its near-the-surface stability.
Table 1: Critical stretch ratios for surface instability
λ3 = 1
λ2 = λ 3
Bell empirical
rubber
simple Bell
0.667
0.545
0.544
0.444
0.500
0.429
REFERENCES
[1] A.N. Guz, Fundamentals of the three-dimensional theory of stability of deformable
bodies, Springer-Verlag, Berlin (1999).
[2] M.A. Biot, Ann. Soc. Sci. B54, 91–109 (1934).
[3] M.A. Biot, Phil. Mag. Ser.7 27, 89–115 (1939).
[4] M.A. Biot, Mechanics of incremental deformations, John Wiley, New York (1965).
[5] J.F. Bell, Int. J. Plasticity 1, 3–27 (1985).
[6] M.F. Beatty, In: Nonlinear Elasticity: Theory and Applications. Eds: Y.B. Fu and
R.W. Ogden, 58–134 (Cambridge University Press, London 2001).
[7] M. Destrade, Int. J. Nonlinear Mech. (to appear 2002).
[8] V.G. Mozhaev, In: IUTAM Symposium on anisotropy, inhomogeneity and nonlinearity in solids. Eds: D.F. Parker and A.H. England, 455–462 (Kluwer, Holland 1995).
[9] M.A. Dowaikh and R.W. Ogden, IMA J. Appl. Math. 44, 261–284 (1990).
[10] M. Romeo, J. Acoust. Soc. Am. 110, 59–67 (2001).
[11] M.F. Beatty and F.X. Pan, Int. J. Non-Linear Mech. 33, 867–906 (1998).
[12] M.F. Beatty and M.A. Hayes, J. Elasticity 29, 1–84 (1992).
[13] M.F. Beatty and M.A. Hayes, Q. Jl. Mech. Appl. Math. 45, 663–709 (1992).
[14] M.F. Beatty and M.A. Hayes, In: Nonlinear Waves in Solids. Eds: J.L. Wegner and
F.R. Norwood, 67–72 (ASME 1995).
[15] M. Hayes and R.S. Rivlin, Arch. Rational Mech. Anal. 8, 358–380 (1961).
[16] F.X. Pan and M.F. Beatty, Int. J. Non-Linear Mech. 34, 169–177 (1999).
[17] Lord Rayleigh, Proc. R. Soc. London A17, 4–11 (1885).