Nonlinear Euler buckling

BY ALAIN GORIELY
1,2,
*
, REBECCA VANDIVER
1
AND MICHEL DESTRADE
3
1
Program in Applied Mathematics, and
2
Department of Mathematics,
University of Arizona, Building no. 89, Tucson, AZ 85721, USA
3
UPMC Univ. Paris 6 and CNRS, Institut Jean le Rond d’Alembert
(UMR 7190), 4 Place Jussieu, 75252 Paris, France
The buckling of hyperelastic incompressible cylindrical tubes of arbitrary length and
thickness under compressive axial load is considered within the framework of nonlinear
elasticity. Analytical and numerical methods for bifurcation are developed using the
exact solution of Wilkes for the linearized problem within the Stroh formalism. Using
these methods, the range of validity of the Euler buckling formula and its first nonlinear
corrections are obtained for third-order elasticity. The values of the geometric
parameters (tube thickness and slenderness) where a transition between buckling and
barrelling is observed are also identified.
Keywords: nonlinear elasticity; Euler buckling formula; barrelling
1. Introduction
Under a large enough compressive axial load an elastic beam will buckle. This
phenomenon known as elastic buckling or Euler buckling is one of the most
celebrated instabilities of classical elasticity. The critical load for buckling was
first derived by Euler in 1744 (Euler 1744, 1759; Oldfather et al. 1933) and
further refined for higher modes by Lagrange in 1770 (Lagrange 1770;
Timoshenko 1983). Both authors reached their conclusion on the basis of simple
beam equations first derived by Bernoulli (Todhunter 1893; figure 1). Since then,
Euler buckling has played a central role in the stability and mechanical
properties of slender structures from nano- to macrostructures in physics,
engineering, biochemistry and biology (Timoshenko & Gere 1961; Niklas 1992).
Explicitly, the critical compressive axial load N that will lead to a buckling
instability of a hinged–hinged isotropic homogeneous beam of length L is
N
Euler
Z
p
2
EI
L
2
; ð1:1Þ
where ‘p is the circumference of a circle whose diameter is one’ (Euler 1759); E is
Young’s modulus; and I is the second moment of area, which, in the case of a
cylindrical shell of inner radius A and outer radius B, is I ZpðB
4
KA
4
Þ=4.
There are many different ways to obtain this critical value and infinite
variations on the theme. If the beam is seen as a long slender structure, the one-
dimensional theory of beams, elastica, or Kirchhoff rods, can be used successfully
Proc. R. Soc. A
doi:10.1098/rspa.2008.0184
Published online
* Author for correspondence (goriely@math.arizona.edu).
Received 4 May 2008
Accepted 16 June 2008 1 This journal is q 2008 The Royal Society
to capture the instability, either by bifurcation analysis, energy argument
(Timoshenko & Gere 1961) or directly from the exact solution, which in the case
of rods can be written in terms of elliptic integrals (Nizette & Goriely 1999). The
one-dimensional theory can be used with a variety of boundary conditions, it is
particularly easy to explain and generalize and it can be used for large geometric
deflections of the axis (Antman 1995). However, since material cross sections
initially perpendicular to the axis remain undeformed and perpendicular to the
tangent vector, no information on the elastic deformation around the central
curve can be obtained. In particular, other modes of instability such as barrelling
cannot be obtained. Here, by barrelling, we refer to axisymmetric deformation
modes of a cylinder or a cylindrical shell. These modes will typically occur for
sufficiently stubby structures.
The two-dimensional theory of shells can be used when the thickness of the
cylindrical shell is small enough. Then, the stability analysis of shell equations
such as the Donnell–von Ka´rma ´n equations leads to detailed information on
symmetric instability modes, their localization and selection (Hunt et al. 2003).
However, the theory cannot be directly applied to obtain information on the
buckling instability (asymmetric buckling mode).
The three-dimensional theory of nonlinear elasticity provides, in principle, a
complete and exact description of the motion of each material point of a body
under loads. However, due to the mathematical complexity of the governing
(a)
(b)
Figure 1. Euler problem: (a) illustrations from Euler (1744). (b) Lagrange solutions (1770), modes
1, 2 and 3.
A. Goriely et al. 2
Proc. R. Soc. A
equations, most problems cannot be explicitly solved. In the case of long slender
structures under loads, the buckling instability can be captured by assuming that
the object is either a rectangular beam (Biot 1962; Levinson 1968; Nowinski
1969) or a cylindrical shell under axial load. Using the theory of incremental
deformations around a large deformation-stressed state, the buckling instability
can be recovered by a bifurcation argument, usually referred to, in the nonlinear
elasticity theory, as small-on-large, or incremental, theory. By taking the proper
asymptotic limit for long slender structures, the Euler criterion can then be
recovered. In comparison to the one- and two-dimensional theories, this
computation is rather cumbersome as it is based on non-trivial tensorial
calculations, but it contains much information about the instability and the
unstable modes selected in the bifurcation process.
Here, we are concerned with the case of a cylindrical shell under axial load.
This problem was first addressed in the framework of nonlinear elasticity in a
remarkable 1955 article by Wilkes who showed that the linearized system around
a finite axial strain can be solved exactly in terms of Bessel functions. While
Wilkes only analysed the first axisymmetric mode (nZ0, see below), he noted in
his conclusion that the asymmetric mode (nZ1) corresponds to the Euler strut
and doing so, opened the door to further investigation by Fosdick & Shield
(1963), who recovered Euler’s criterion asymptotically from the solution of
Wilkes. These initial results constitute the basis for much of the modern theory
of elastic stability of cylinders within the framework of three-dimensional
nonlinear elasticity (Haughton & Ogden 1979a,b; Simpson & Spector 1984; Duka
et al. 1993; Pan & Beatty 1997; Bigoni & Gei 2001; Dorfmann & Haughton 2006).
The experimental verification of Euler’s criterion was considered by Southwell
(1932) and by Beatty & Hook (1968).
The purpose of this article is threefold. First, we revisit the problemof the stability
of an incompressible cylindrical shell under axial load using the Stroh formalism
(Stroh 1962) and, based on the solution of Wilkes, we derive a new and compact
formulation of the bifurcation criterion that can be used efficiently for numerical
approximation of the bifurcation curves for all modes. Second, we use this
formulation to obtain nonlinear corrections of Euler’s criterion for arbitrary shell
thickness and third-order elasticity. Third, we consider the problem of determining
the critical aspect ratio where there is a transition between buckling and barrelling.
2. Large deformation
We consider a hyperelastic homogeneous incompressible cylindrical tube with
isotropic cross sections of initial inner radius A, outer radius BandlengthL, subjected
to a uniaxial constant strain l
3
and deformed into a shorter tube with current
dimensions a, b and l. The deformation bringing a point at (R, Q, Z), in cylindrical
coordinates in the initial configuration, to (r, q, z) in the current configuration is
r Zl
1
R; q ZQ; z Zl
3
Z; ð2:1Þ
where l
1
Za=AZb=B and l
3
Zl/L. Since the material is isotropic in the cross
sections, the physical components of the corresponding deformation gradient F are
½FŠ Zdiagðl
1
; l
1
; l
3
Þ; ð2:2Þ
3 Nonlinear Euler buckling
Proc. R. Soc. A
showing that the principal stretches are the constants l
1
, l
2
Zl
1
, l
3
; and that
the pre-strain is homogeneous. Owing to incompressibility, det FZ1, so that
l
1
Zl
K1=2
3
: ð2:3Þ
The principal Cauchy stresses required to maintain the pre-strain are (Ogden 1984)
s
i
ZKp Cl
i
vW
vl
i
ði Z1; 2; 3Þ ð2:4Þ
(no sum), where p is a Lagrange multiplier introduced by the internal constraint of
incompressibility and W is the strain energy density (a symmetric function of the
principal stretches). Inour case, s
2
Zs
1
because l
2
Zl
1
. Also, s
1
Z0 because the inner
and outer faces of the tube are free of traction. It follows that:
p Zl
1
W
1
; s
3
Zl
3
W
3
Kl
1
W
1
; ð2:5Þ
where W
i
hvW/vl
i
, andwe conclude that the principal Cauchystresses are constant.
3. Instability
To perform a bifurcation analysis, we take the view that the existence of small
deformation solutions in the neighbourhood of the large pre-strain signals the
onset of instability (Biot 1965).
(a ) Governing equations
The incremental equations of equilibrium and incompressibility can be written
as (Ogden 1984)
div s Z0; div u Z0; ð3:1Þ
where s is the incremental nominal stress tensor and u is the infinitesimal
mechanical displacement. They are linked by
s ZA
0
gradðuÞ Cp gradðuÞK _ pI; ð3:2Þ
where _ p is the increment in the Lagrange multiplier p and A
0
is the fourth-order
tensor of instantaneous elastic moduli. This tensor is similar to the stiffness
tensor of linear anisotropic elasticity, with the differences that it possesses only
the major symmetries, not the minor ones, and that it reflects strain-induced
anisotropy instead of intrinsic anisotropy. Its explicit non-zero components in a
coordinate system aligned with the principal axes are (Ogden 1984)
A
0iijj
Zl
i
l
j
W
ij
;
A
0ijij
Zðl
i
W
i
Kl
j
W
j
Þl
2
i
=ðl
2
i
Kl
2
j
Þ; if i sj; l
i
sl
j
;
A
0ijij
ZðA
0iiii
KA
0iijj
Cl
i
W
i
Þ=2; if i sj; l
i
Zl
j
;
A
0ijji
ZA
0jiij
A
0ijij
Kl
i
W
i

ð3:3Þ
A. Goriely et al. 4
Proc. R. Soc. A
(no sums), where W
ij
hv
2
W=ðvl
i
vl
j
Þ. Note that some of these components are
not independent because here l
1
Zl
2
. In particular, we have
A
02121
ZA
01212
; A
02323
ZA
01313
; A
02222
ZA
01111
;
A
02233
ZA
01133
; A
02332
ZA
01331
; A
03232
ZA
03131
;
A
01221
CA
01212
A
01111
KA
01122
Z2A
01212
CA
01331
KA
01313
:

ð3:4Þ
(b ) Solutions
We look for solutions that are periodic along the circumferential and axial
directions, and have yet unknown variations through the thickness of the tube, so
that our ansatz is
fu
r
; u
q
; u
z
; _ p; s
rr
; s
rq
; s
rz
g ZfU
r
ðrÞ; U
q
ðrÞ; U
z
ðrÞ; PðrÞ; S
rr
ðrÞ; S
rq
ðrÞ; S
rz
ðrÞg
!expðiðnq CkzÞÞ; ð3:5Þ
where nZ0, 1, 2, .is the circumferential number; k is the axial wavenumber; the
subscripts (r, q, z) for u and s refer to components within the cylindrical
coordinates (r, q, z); and all upper-case functions are functions of r alone.
The specialization of the governing equations (3.1) to this type of solution has
already been conducted in several articles (see Wilkes 1955; Fosdick & Shield 1963;
Mack 1989; Pan & Beatty 1997; Negron-Marrero 1999; and Dorfmann & Haughton
2006 for the compressible counterpart). Here we adapt the work of Shuvalov (2003a)
on waves in anisotropic cylinders to develop a Stroh-like formulation of the problem
(Stroh 1962). The central idea is to introduce a displacement–traction vector,
hh½U
r
; U
q
; U
z
; irS
rr
; irS
rq
; irS
rz
Š
T
; ð3:6Þ
so that the incremental equations can be written in the form
d
dr
hðrÞ Z
i
r
GðrÞhðrÞ; ð3:7Þ
where Gis a 6!6 matrix, with the block structure
GZ
G
1
G
2
G
3
G
C
1
¸ ¸
; G
2
ZG
C
2
; G
3
ZG
C
3
: ð3:8Þ
Here the superscript ‘C’ denotes the Hermitian adjoint (transpose of the complex
conjugate) and G
1
, G
2
and G
3
are the 3!3 matrices
i Kn Kkr
Kn Ki 0
Kkr 0 0

¸
¸
¸;
0 0 0
0 K
1
A
01212
0
0 0 K
1
A
01313

¸
¸
¸
¸
¸
¸
¸
¸
;
k
11
ik
12
k
13
Kik
12
k
22
k
23
Kik
13
k
23
k
33

¸
¸
¸; ð3:9Þ
5 Nonlinear Euler buckling
Proc. R. Soc. A
respectively, with
k
11
Z4A
01212
CðA
03131
KA
01313
Þk
2
r
2
; k
23
Znð2A
01212
CA
01313
Þkr;
k
12
Z4nA
01212
; k
13
Z2A
01212
kr; k
22
Z4n
2
A
01212
CA
03131
k
2
r
2
;
k
33
Zn
2
A
01313
Cð4A
01212
C2A
01122
KA
01111
CA
03333
K2A
01133
Þk
2
r
2
:

ð3:10Þ
As it happens, there exists a set of six explicit Bessel-type solutions to these
equations whenns0. This situationis inmarkedcontrast withthecorrespondingset-
up in linear anisotropic elastodynamics, where explicit Bessel-type solutions exist
only for transversely isotropic cylinders with a set of four linearly independent modes
and do not exist for cylinders of lesser symmetry (Martin & Berger 2001; Shuvalov
2003b). As mentioned in §1, the six Bessel solutions are presented in the article by
Wilkes (1955; for a derivation see Bigoni & Gei 2001).
First, denote by q
2
1
, q
2
2
the roots of the following quadratic in q
2
:
A
01313
q
4
KðA
01111
CA
03333
K2A
01331
K2A
01133
Þq
2
CA
03131
Z0: ð3:11Þ
Then the roots of this quartic in q are Gq
1
and Gq
2
, and it can be checked that
the following two vectors are solutions to (3.7):
h
ð1Þ
; h
ð2Þ
Z iI
0
n
ðqkrÞ;K
n
qkr
I
n
ðqkrÞ;KqI
n
ðqkrÞ;
¸
K
kr
q
ðA
01313
q
2
CA
03131
ÞI
n
ðqkrÞ C2A
01212
I
0
n
ðqkrÞK
n
2
qkr
I
n
ðqkrÞ

;
K2inA
01212
I
0
n
ðqkrÞK
1
qkr
I
n
ðqkrÞ

;Kið1 Cq
2
ÞkrA
01313
I
0
n
ðqkrÞ

T
;
ð3:12Þ
where qZq
1
, q
2
in turn and I
n
is the modified Bessel function of order n.
Similarly, we checked that the following vector h
(3)
:
h
ð3Þ
Z i
n
kr
I
n
ðq
3
krÞ;Kq
3
I
0
n
ðq
3
krÞ; 0;K2nq
3
A
01212
I
0
n
ðq
3
krÞK
1
q
3
kr
I
n
ðq
3
krÞ

;
¸
!iq
3
A
01212
2I
0
n
ðq
3
krÞK q
3
kr C2
n
2
q
3
kr

I
n
ðq
3
krÞ

; KinA
01313
I
n
ðq
3
krÞ

T
ð3:13Þ
is also a solution when q
3
is the positive root of the quadratic equation
A
01212
q
2
KA
03131
Z0: ð3:14Þ
Finally, we also checked that the vectors h
(4)
, h
(5)
and h
(6)
, obtained by replacing
I
n
with the modified Bessel function K
n
in the expressions above, are solutions too.
Next, we follow Shuvalov (2003a) and introduce N(r) as a fundamental matrix
solution to (3.7):
N ðrÞ Z

h
ð1Þ
jh
ð2Þ
j.jh
ð6Þ

: ð3:15Þ
A. Goriely et al. 6
Proc. R. Soc. A
It clearly satisfies
d
dr
N ðrÞ Z
i
r
GðrÞN ðrÞ: ð3:16Þ
Let M(r,a) be the matricant solution to (3.7), i.e. the matrix such that
hðrÞ ZMðr; aÞhðaÞ; Mða; aÞ ZI
ð6Þ
: ð3:17Þ
It is obtained from N(r) (or from any other fundamental matrix made of linearly
independent combinations of the h
(i )
) by
Mðr; aÞ ZN ðrÞN
K1
ðaÞ ð3:18Þ
and it has the following block structure:
Mðr; aÞ Z
M
1
ðr; aÞ M
2
ðr; aÞ
M
3
ðr; aÞ M
4
ðr; aÞ
¸ ¸
; ð3:19Þ
say.
(c ) Boundary conditions
Some boundary conditions must be enforced on the top and bottom faces of
the tubes. Considering that they remain plane (U
z
Z0 on zZ0, l ) and free
of incremental shear tractions (S
rz
ZS
rq
Z0 on zZ0, l ) leads to
k Z
mp
l
Z
mp
l
3
L
; ð3:20Þ
where mZ1, 2, 3, .but, since the equations depend only on k, we can take mZ1
without loss of generality.
The other boundary conditions are that the inner and outer faces of the
tube remain free of incremental tractions. We call S h½S
rr
; S
rq
; S
rz
Š
T
the traction
vector and U h½U
r
; U
q
; U
z
Š
T
the displacement vector. We substitute the
condition S(a)Z0 into (3.17) and (3.19) to find the following connection:
rSðrÞ Zzðr; aÞUðrÞ; where z hiM
3
M
K1
1
ð3:21Þ
is the (Hermitian) 3!3 impedance (Shuvalov 2003a). Since S(b)Z0, a non-
trivial solution exists only if the matrix z(b, a) is singular, which implies the
bifurcation condition
det zðb; aÞ ZKi
det M
3
ðb; aÞ
det M
1
ðb; aÞ
Z0: ð3:22Þ
This is a real equation since zZz
C
(Shuvalov 2003a) that applies independently
of the nature (i.e. real or complex (Pan & Beatty 1997), simple or double
(Dorfmann & Haughton 2006)) of the roots q
1
, q
2
and q
3
.
7 Nonlinear Euler buckling
Proc. R. Soc. A
4. The adjugate method
We are now in a position to use the bifurcation condition (3.22) to compute
explicitly bifurcation curves for each mode n. We note that the components of A
0
depend on the strain energy density W and on the pre-strain, which by (2.3)
depends only on l
3
; so do q
1
, q
2
, q
3
, by (3.11) and (3.14). According to (3.12) and
(3.13), the entries of M(b, a) thus depend (for a given W) on l
3
, n, ka and kb
only. For a given material (W specified) with a given thickness (b/aZB/A
specified), the bifurcation equation (3.22) gives a relationship between a measure
of the critical pre-stretch: l
3
Zl
K2
1
, and a measure of the tube slenderness:
kbZ2pmðb=lÞZ2pml
K3=2
3
ðB=LÞ, for a given bifurcation mode (n specified). That
is, for a given tube slenderness, what is the axial strain necessary to excite a
given mode?
While this bifurcation condition is formally clear, it has not been successfully
implemented to compute all bifurcation curves. Indeed, for mode nO1, the root
finding of det(z) becomes numerically unstable and numerical methods become
unreliable (as observed in Dorfmann & Haughton (2006) for a similar problem)
and, in explicit computations, most authors do not use the exact solution by
Wilkes but use a variety of numerical techniques to solve the linear boundary-
value problems directly (such as the compound matrix method (Haughton & Orr
1997), the determinantal method (Ben Amar & Goriely 2005) or the Adams–
Moulton method (Zhu et al. 2008)). Note that from a computational perspective,
the Stroh formalism is particularly well suited and well behaved (Biryukov 1985;
Fu 2005) and if numerical integration was required it would provide an ideal
representation of the governing equation.
Rather than integrating the original linear problem numerically, we now show
how to use an alternative form of (3.22) to compute all possible bifurcation
curves. This method bypasses the need for numerical integration and reduces the
problem to a form that is manageable both numerically and symbolically, to
study analytically particular asymptotic limits. The main idea is to transform
condition (3.22) by factoring non-vanishing factors. We start by realizing that
since the fundamental solutions {h
(i )
, iZ1, ., 6} are linearly independent, the
matrix N(r) is invertible for all r2[a,b], which implies that the elements of
M(r, a) are bounded for r2[a,b]. Therefore, det(M
1
(r, a)) is uniformly bounded
away from zero and det zZ0 implies det(M
3
(b, a))Z0. Instead of expressing
det(M
3
(b, a)) as the determinant of a 3!3 submatrix of a matrix obtained as the
product of two 6!6 matrices, we first decompose N(r) as
N ðrÞ Z
N
1
ðrÞ N
2
ðrÞ
N
3
ðrÞ N
4
ðrÞ
¸ ¸
; ð4:1Þ
say, where each block is a 3!3 matrix. We also rewrite equation (3.18) as
Mðr; aÞN ðaÞ ZN ðrÞ; ð4:2Þ
and write explicitly the two entries N
3
(r) and N
4
(r), which are
M
3
ðr; aÞN
1
ðaÞ CM
4
ðr; aÞN
3
ðaÞ ZN
3
ðrÞ; ð4:3Þ
M
3
ðr; aÞN
2
ðaÞ CM
4
ðr; aÞN
4
ðaÞ ZN
4
ðrÞ; ð4:4Þ
A. Goriely et al. 8
Proc. R. Soc. A
which implies
M
3
ðr; aÞ N
1
ðaÞN
K1
3
ðaÞKN
2
ðaÞN
K1
4
ðaÞ

ZN
3
ðrÞN
K1
3
ðaÞKN
4
ðrÞN
K1
4
ðaÞ:
ð4:5Þ
Using again the fact that the entries of N are bounded, we have that the
bifurcation condition det(M
3
(b, a))Z0 implies that
det Qðb; aÞ Z0; ð4:6Þ
where
Qðb; aÞ ZdetðN
4
ðaÞÞN
3
ðbÞadjðN
3
ðaÞÞKdetðN
3
ðaÞÞN
4
ðbÞadjðN
4
ðaÞÞ; ð4:7Þ
and adj(A) is the adjugate matrix of A, i.e. the transpose of the cofactor matrix
(which in the case of an invertible matrix is simply adj(A)Zdet(A)A
K1
). This
new bifurcation condition is equivalent to the previous one but has many
advantages. The matrix Q involves only products of 3!3 matrices and is
polynomial in the entries of N, i.e. det Q(b, a) is a polynomial of degree 18 in
Bessel functions and has no denominator (hence no small denominator). Both
numerically and symbolically, this determinant is well behaved, even in the
limits a/0, which corresponds to a solid cylinder, and nZ0, which corresponds
to the first barrelling mode (and usually requires a special treatment). We will
refer to the use of this form of the bifurcation condition as the adjugate method.
15
0.2
5 25
0.7
20
0.9
30
0.5
1.0
0.1
0.6
0
0.8
0.4
0.3
10
kb
Figure 2. Bifurcation curves (stretch as a function of stubbiness) of a homogeneous neo-Hookean
cylindrical tube for modes nZ0–8 with b/aZB/AZ2 and C
1
Z1.
9 Nonlinear Euler buckling
Proc. R. Soc. A
(a ) Numerical results
As a first test of the stability of the numerical procedure, we consider a neo-
Hookean potential WZC
1
ðI
1
K3Þ=2, where we set C
1
Z1 without loss of
generality and consider the typical value B/AZ2. We compute the critical
value of lhl
3
as a function of the current stubbiness kbZpb/l (the initial
stubbiness is nZB=LZkbl
3=2
=p) for the first 9 modes (nZ0–8) as shown
in figure 2. The known classical features of the stability problem for the
cylindrical shell are recovered, namely for slender tubes, the Euler buckling
(nZ1) is dominant and becomes unavoidable as the slenderness increases; there
is a critical slenderness value at which the first barrelling mode nZ0 is the first
unstable mode (in a thought experiment where the axial strain would be incre-
mentally increased until the tube becomes unstable); and for very large kb, the
critical compression ratio tends asymptotically to the value lZ0.444, which
corresponds to surface instability of a compressed half-space (Biot 1962).
For a second test, we consider very thin neo-Hookean tubes with B/AZ1.01. Here
we are interested in the mode selection process. As the stubbiness increases, the
buckling mode rapidly ceases to be the first excited mode and is replaced by different
barrelling modes. Fromfigure 3, it appears clearly that as kb increases, modes nZ1–9
are selected (modes nZ0 and 10 remain unobservable). There is one particularly
interesting feature inthese two sets of bifurcationcurves. Dependingonboththe tube
thickness and the stubbiness, the instability mode of a tube transition occurs from
buckling to barrelling, the material transition from either the one-dimensional
behaviour of slender columntothe two-dimensional behaviour of a thinshort tube, or
the three-dimensional behaviour of a thick short tube. Accordingly, we will refer to
these particular geometric values where transition occurs as dimensional transitions
and obtain analytical estimates for them in §5.
0.94
10 9 6
0.99
4 1
0.96
8 5
0.97
7
0.98
0
0.95
3
1.00
2
1
2
3
4
5 6 7 8 9
kb
Figure 3. Bifurcation curves (stretch as a function of stubbiness) of a homogeneous neo-Hookean
cylindrical tube for modes nZ0–10 with b/aZB/AZ1.01 and C
1
Z1.
A. Goriely et al. 10
Proc. R. Soc. A
5. Asymptotic Euler buckling
We are now in a position to look at the asymmetric buckling mode (nZ1)
corresponding to the Euler buckling in the limit l/1. The asymptotic form of
the Euler criterion cannot be obtained for a general strain-energy density. This is
why we choose the Mooney–Rivlin potential, which, for l close to 1, corresponds
to the most general form of third-order incompressible elasticity (see §6 and
Rivlin & Saunders (1951)),
W ZC
1
ðI
1
K3Þ=2 CC
2
ðI
2
K3Þ=2; ð5:1Þ
where C
1
R0 and C
2
O0 are material constants; I
1
Zl
2
1
Cl
2
2
Cl
2
3
; and I
2
Z
l
2
1
l
2
2
Cl
2
2
l
2
3
Cl
2
3
l
2
1
. Close to lZ1, we introduce a small parameter related to the
stubbiness ratio
e Zkb Zpb=l; ð5:2Þ
and look for the critical buckling stretch l as a function of e of order M,
l ZlðeÞ Z1 C
¸
M
mZ1
l
m
e
m
COðe
MC1
Þ: ð5:3Þ
Similarly, we expand dðlÞZdet Qðb; aÞ in powers of e,
dðlÞ Z
¸
M
d
mZ1
d
m
e
m
COðe
M
d
C1
Þ; ð5:4Þ
and solve each order d
m
Z0 for the coefficients l
m
. This is a rather cumbersome
computation. It can be checked that l
m
vanishes identically for all odd values of
m and that the first non-identically vanishing coefficient appears at order 24.
A computation to order 28 is necessary to compute the correct expression for a,
which is found to be to order 6 in e
l Z1 Cl
ð2Þ
e
2
Cl
ð4Þ
e
4
Cl
ð6Þ
e
6
COðe
8
Þ; ð5:5Þ
with
l
ð2Þ
ZK
r
2
C1
4r
2
; ð5:6Þ
l
ð4Þ
Z
ð19C
2
C28C
1
Þr
4
C2ð53C
2
C62C
1
Þr
2
C19C
2
C28C
1
144ðC
1
CC
2
Þr
4
; ð5:7Þ
l
ð6Þ
ZK
1
4608r
6
ðr
4
K1ÞðC
1
CC
2
Þ

ð973C
1
C341C
2
Þðr
10
K1Þ
Cð7073C
1
C3385C
2
Þr
2
ðr
6
K1Þ C4392 lnðrÞðC
1
CC
2
Þðr
6
Cr
4
Þ
C4ð377C
2
C1141C
1
Þðr
6
Kr
4
Þ

; ð5:8Þ
where rhB/AZb/a. It is of interest to compare the different approximations.
We recover the Euler formula by keeping only the term up to e
2
, which we denote
by Euler
2
. We define similarly Euler
4
and Euler
6
by keeping terms up to orders 4
and 6 in e. We show the different approximations as a function of e
2
for rZ1.01
11 Nonlinear Euler buckling
Proc. R. Soc. A
(figure 4a) and rZ10 (figure 4b). The classical Euler formula is well recovered in
the limit e/0, but the Euler
4
and Euler
6
approximations clearly improve the
classical formula for larger values of e. It also appears from the analysis of Euler
4
that for C
2
R0 the classical Euler formula always underestimates the critical
stretch for instability.
6. Nonlinear Euler buckling for third-order elasticity
The analytical result presented in §5 was formulated in terms of parameters and
quantities natural for the computation and the theory of nonlinear elasticity. In
order to relate this result to the classical form of Euler buckling, we need to
express equation (5.5) in terms of the initial geometric values A, B, L, the axial
load acting on the cylinder and the elastic parameters entering in the theory of
linear elasticity.
We first consider the geometric parameters. We wish to express the critical
load as a function of the initial stubbiness nZB/L and tube relative thickness
rZB/A. Recalling that eZpb/l and lZl/L, bZl
K1/2
B, we have
e
2
l
3
Zp
2
n
2
: ð6:1Þ
To express e as a function of n, we expand e in powers of n to order 6, and solve
(6.1) to obtain
e
2
Zp
2
n
2
K3p
4
l
ð2Þ
n
4
Kð3p
6
l
ð4Þ
K15p
6
l
2
ð2Þ
Þn
6
COðn
8
Þ; ð6:2Þ
where l
(2)
and l
(4)
are defined in (5.6) and (5.7) and come from the expansion of l
in powers of e.
1.00 (a) (b)
0.98
0.25
0.96
0.15
0.92
0.90
0.99
0.30
0.97
0.95
0.20
0.94
0.93
0.91
0.10 0.05 0
Euler
2
exact
Euler
6
Euler
4
Euler
2
exact
Euler
6
Euler
4
0.6
1.000
0.8
0.900
1.0 0.9
0.975
0.875
0.5 0.4 0.2 0.7 0
0.950
0.925
0.3
0.850
0.1
Figure 4. Comparison of the different Euler formulae obtained by expanding the exact solution to
order 2 (the classical Euler buckling formula), Euler
4
and Euler
6
for a neo-Hookean potential
C
1
Z1, C
2
Z0. For comparison purpose, we show the critical stretch for mode nZ1 versus e
2
in which
case the graph becomes linear in the limit e/0. (a) rZb/aZB/AZ1.01, (b) rZb/aZB/AZ10.
A. Goriely et al. 12
Proc. R. Soc. A
Second, we want to relate the axial compression to the actual axial load N. To
do so, we integrate the axial stress over the faces of the tubes, i.e.
N ZK2p

b
a
rs
3
dr: ð6:3Þ
Since s
3
is constant and given by (2.5), we have
N ZKpðb
2
Ka
2
Þs
3
ZK
p
l
ðB
2
KA
2
Þs
3
ZK
p
l
3
ðB
2
KA
2
Þ½ðl
4
KlÞC
1
Cðl
3
K1ÞC
2
Š: ð6:4Þ
Third, we relate the elastic Mooney parameters C
1
and C
2
to the classical
elastic parameters. Here, we follow Hamilton et al. (2004; Destrade &
Saccomandi 2005) and write the strain-energy density to third order for an
incompressible elastic material as
W ZK2mi
2
Cn
3
i
3
; ð6:5Þ
where m is the usual shear modulus, or second Lame ´ parameter, and n
3
is a third-
order elasticity constant; m is related to Young’s modulus by EZ3m; also, in
Murnaghan’s notation, n
3
Zn and in Landau’s notation, n
3
ZA (see Norris (1998)
for other notations). In (6.5), i
1
, i
2
, i
3
are the first three principal invariants of the
Green–Lagrange strain tensor, related to the first three principal invariants I
1
, I
2
,
I
3
of the Cauchy–Green strain tensor by
I
1
Z2i
1
C3; I
2
Z4i
1
C4i
2
C3; I
3
Z2i
1
C4i
2
C8i
3
C1: ð6:6Þ
Since I
3
Z1, we can solve this linear system for i
2
and i
3
and write the strain-
energy density (6.5) as a function of I
1
and I
2
, i.e.
W Z mC
n
3
8

I
1
K
m
2
K
n
3
8

I
2
; ð6:7Þ
which by comparison with (5.1) leads to
C
1
Z2mCn
3
=4; C
2
ZmKn
3
=4: ð6:8Þ
To write the nonlinear buckling formula, we consider (6.4) and first expand l
in e using (5.5), then expand e in n using (6.2) and, finally, substitute the values
of the moduli in terms of the elastic parameters, which yield
N Z
3
4
p
3
B
2
mðr
4
K1Þn
2
r
4
K
1
96
p
5
B
2
ðr
2
K1Þð20r
4
mC9r
4
n
3
C176r
2
mC18r
2
n
3
C20mC9n
3
Þn
4
r
6
C
B
2
p
7
n
6
512r
8
mðr
2
C1Þ
323m
2
r
8
K3n
2
3
K240r
2
mn
3
K9r
2
n
2
3
K9r
10
m
2
C9m
2

C3r
10
n
2
3
C1464 lnðrÞr
6
m
2
C1464 lnðrÞr
4
m
2
C240r
8
mn
3
C240r
6
mn
3
K180r
6
m
2
C180r
4
m
2
C9r
8
n
2
3
C6r
6
n
2
3
K6r
4
n
2
3
K323r
2
m
2
K240r
4
mn
3

COðn
8
Þ: ð6:9Þ
13 Nonlinear Euler buckling
Proc. R. Soc. A
While it is not surprising, it is comforting to recover to order n
2
the classical
Euler buckling formula (1.1) (using rZB/A, nZB/L and mZE/3).
7. Dimensional transition
Finally, we use the buckling formula to compute the transition between modes as
parameters are varied. That is, to identify both the geometric values and the
axial strain for which there is a transition between buckling and barrelling
modes. Here we restrict again our attention to the neo-Hookean case (with C
1
Z1).
Fromfigures 2 and 3, it appears clearly that for e small enough there is a transition
(depending on the value of r) from either mode nZ1 to mode nZ0 (large r), or
from mode nZ1 to mode nZ2 (r close to 1) as e increases. We refer to this
transition as a dimensional transition, in the sense that the material mostly
behaves as a slender one-dimensional structure when it buckles according to
mode nZ0 and mostly as a two-dimensional structure when it barrels with mode
nZ2. Indeed both modes of instability can be captured by, respectively, a one- or
a two-dimensional theory. For r close to unity, the transition nZ0/nZ1 occurs
for small values of e. Therefore, in this regime, we can use the approximation
(5.5) for the barrelling curve and substitute it in the bifurcation condition of
mode nZ2. Expanding again this bifurcation condition in e as well as r, one
identifies the values r
t
of r and l
t
of l at which the transition occurs, as
r
t
Z1 C
3
4
e
2
K
53
32
e
4
C
2393
384
e
6
COðe
8
Þ;
l
t
Z1K
1
4
e
2
C
13
8
e
4
K
665
96
e
6
COðe
8
Þ:

ð7:1Þ
In terms of the initial stubbiness nZB/L, we have
r
t
Z1 C
3
4
p
2
n
2
K
17
32
p
4
n
4
C
161
384
p
6
n
6
COðn
8
Þ: ð7:2Þ
This relationship also provides a domain of validity for the Euler buckling formula.
For sufficiently slender tube (n small), the buckling mode disappears when rOr
t
at
the expense of the nZ2 barrelling mode. For stubbier and fuller tubes, this
approximation cannot be used. To understand the dimensional transition, we solve
numerically the bifurcation condition, using the adjugate method, for the
intersection of two different modes. That is, for a given value of r
Ã
, we find the
value of e
Ã
such that both the bifurcations for either modes nZ1 and 2, or modes
nZ1 and 0 are satisfied. If the corresponding value l
Ã
is the largest value for which
a bifurcation takes place, the pair (e
Ã
, r
Ã
) is a transition point. The corresponding
transition point in terms of the initial parameters is n
Ã
Zðe
Ã
=pÞðl
Ã
Þ
3=2
; r
Ã
. In
figure 5, we show a diagram of all such pairs for both transitions.
8. Conclusion
This article establishes a reliable and effective method to study the stability of
tubes based on the exact solution of the incremental equations proposed by
Wilkes (1955) within the Stroh formalism. It then puts the method to use, to
A. Goriely et al. 14
Proc. R. Soc. A
obtain the first geometric and material corrections to the Euler buckling. The
method can be also used to obtain the transition between buckling and barrelling
modes when a tube becomes unstable.
The method presented here can be easily generalized to different materials and
different boundary conditions. For instance, using the exact solution of the
incremental equations proposed in Dorfmann & Haughton (2006) for compres-
sible materials and the adjugate method, an explicit form of the bifurcation
condition in terms of Bessel functions can be obtained by following the steps
presented here and various asymptotic behaviours can be obtained. Similarly, a
variety of boundary-value problems can be analysed by the adjugate method,
such as the stability problem of a tube under pressure and tension (Han
2007; Zhu et al. 2008), the problem of a tube embedded in an infinite domain
(Bigoni & Gei 2001) and the problem of a tube with coating (Ogden et al. 1997).
In all these cases, useful asymptotic formulae for the buckling behaviour could be
obtained by perturbation expansions.
It is also enticing to consider the possibility of performing an analytical
post-buckling analysis of the solutions. Since the solutions of the linearized
problem can be solved exactly, a weakly nonlinear analysis of the solution
should be possible to third order. This would yield, in principle, an equation
for the amplitude of the unstable modes containing much information not
0.9
2.6
2.2
0.7
1.8
0.5
1.0
1.0
3.0
2.8
2.4
0.8
2.0
1.6
0.6
1.4
1.2
0.4 0.3 0.2 0.1 0
approximation
L
Figure 5. Dimensional transition for a neo-Hookean cylindrical tube of initial length L and initial
radii A and B. All tubes in the nZ1 regions will become unstable by buckling. As the tubes get
stubbier or thinner (arrows), it will not buckle but instead will be subjected to a barrelling
instability. Note that only the transition curves from mode nZ0 are shown. Tubes in the barrelling
regions may be subjected to other unstable modes.
15 Nonlinear Euler buckling
Proc. R. Soc. A
only about the actual amplitude of the unstable modes but also on the
localization of unstable modes after bifurcation. We leave this daunting task for
another day.
This study is based in part upon work supported by the National Science Foundation under grant
no. DMS-0604704 (A.G.) and made possible by a CNRS/USA Collaborative Grant from the
French Centre National de la Recherche Scientifique (M.D.).
References
Antman, S. S. 1995 Nonlinear problems of elasticity. New York, NY: Springer.
Beatty, M. F. & Hook, D. E. 1968 Some experiments on the stability of circular, rubber bars under
end thrust. Int. J. Solids Struct. 4, 623–635. (doi:10.1016/0020-7683(68)90078-4)
Ben Amar, M. & Goriely, A. 2005 Growth and instability in soft tissues. J. Mech. Phys. Solids 53,
2284–2319. (doi:10.1016/j.jmps.2005.04.008)
Bigoni, D. & Gei, M. 2001 Bifurcations of a coated, elastic cylinder. Int. J. Solids Struct. 38,
5117–5148. (doi:10.1016/S0020-7683(00)00322-X)
Biot, M. A. 1962 Exact theory of buckling of a thick slab. Appl. Sci. Res. 12, 183–198.
Biot, M. A. 1965 Mechanics of incremental deformations. New York, NY: Wiley.
Biryukov, S. V. 1985 Impedance method in the theory of elastic surface waves. Sov. Phys. Acoust.
31, 350–354.
Destrade, M. & Saccomandi, G. 2005 Finite amplitude elastic waves propagating in compressible
solids. Phys. Rev. E 72, 016 620. (doi:10.1103/PhysRevE.72.016620)
Dorfmann, A. & Haughton, D. M. 2006 Stability and bifurcation of compressed elastic cylindrical
tubes. Int. J. Eng. Sci. 44, 1353–1365. (doi:10.1016/j.ijengsci.2006.06.014)
Duka, E. D., England, A. H. & Spencer, A. J. M. 1993 Bifurcation of a solid circular elastic cylinder
under finite extension and torsion. Acta Mech. 98, 107–121. (doi:10.1007/BF01174297)
Euler, L. 1744 Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes sive
solutio problematis isoperimetrici latissimo sensu accepti. Lausanne, Geneva: Marc-Michel
Bousquet & Co.
Euler, L. 1759 Sur la force des colonnes, vol. 13. Berlin, Germany: Me `moires de l’ Acade ´mie de Berlin.
Fosdick, R. L. & Shield, R. T. 1963 Small bending of a circular bar superposed on finite extension
or compression. Arch. Ration. Mech. Anal. 12, 223–248. (doi:10.1007/BF00281227)
Fu, Y. B. 2005 An explicit expression for the surface-impedance matrix of a generally anisotropic
incompressible elastic material in a state of plane strain. Int. J. Nonlin. Mech. 40, 229–239.
(doi:10.1016/j.ijnonlinmec.2004.05.002)
Hamilton, M. F., Ilinskii, Y. A. & Zabolotskaya, E. A. 2004 Separation of compressibility and shear
deformation in the elastic energy density. J. Acoust. Soc. Am. 116, 41–44. (doi:10.1121/
1.1736652)
Han, H. C. 2007 A biomechanical model of artery buckling. J. Biomech. 40, 3672–3678. (doi:10.
1016/j.jbiomech.2007.06.018)
Haughton, D. M. & Ogden, R. W. 1979a Bifurcation of inflated circular cylinders of elastic material
under axial loading-II. Exact theory for thick-walled tubes. J. Mech. Phys. Solids 27, 489–512.
(doi:10.1016/0022-5096(79)90027-9)
Haughton, D. M. & Ogden, R. W. 1979b Bifurcation of inflated circular cylinders of elastic material
under axial loading-I. Membrane theory for thin-walled tubes. J. Mech. Phys. Solids 27,
179–212. (doi:10.1016/0022-5096(79)90001-2)
Haughton, D. M. & Orr, A. 1997 On the eversion of compressible elastic cylinders. Int. J. Solids
Struct. 34, 1893–1914. (doi:10.1016/S0020-7683(96)00122-9)
Hunt, G. W., Lord, G. J. & Peletier, M. A. 2003 Cylindrical shell buckling: a characterization of
localization and periodicity. Discrete Contin. Dyn. Syst. B 3, 505–518.
Lagrange, J. L. 1770 Sur la figure des colonnes. Miscellanea Taurinensia 5, 123–166.
A. Goriely et al. 16
Proc. R. Soc. A
Levinson, M. 1968 Stability of a compressed neo-Hookean rectangular parallelepiped. J. Mech.
Phys. Solids 16, 403–415. (doi:10.1016/0022-5096(68)90004-5)
Mack, W. 1989 On the deformation of circular cylindrical rubber bars under axial thrust. Acta
Mech. 77, 91–99. (doi:10.1007/BF01379744)
Martin, P. A. & Berger, J. R. 2001 Waves in wood: free vibrations of a wooden pole. J. Mech. Phys.
Solids 49, 1155–1178. (doi:10.1016/S0022-5096(00)00068-5)
Negron-Marrero, P. V. 1999 An analysis of the linearized equations for axisymmetric deformations
of hyperelastic cylinders. Math. Mech. Solids 4, 109–133. (doi:10.1177/108128659900400107)
Niklas, K. J. 1992 Plant biomechanics: an engineering approach to plant form and function.
Chicago, IL: The University of Chicago Press.
Nizette, M. & Goriely, A. 1999 Towards a classification of Euler–Kirchhoff filaments. J. Math.
Phys. 40, 2830–2866. (doi:10.1063/1.532731)
Norris, A. 1998 Finite amplitude waves in solids. In Nonlinear acoustics (eds M. F. Hamilton &
D. T. Blackstock), pp. 263–277. San Diego, CA: Academic Press.
Nowinski, J. L. 1969 On the elastic stability of thick columns. Acta Mech. 7, 279–286. (doi:10.
1007/BF01377530)
Ogden, R. W. 1984 Non-linear elastic deformation. New York, NY: Dover.
Ogden, R. W., Steigmann, D. J. & Haughton, D. M. 1997 The effect of elastic surface coating on
the finite deformation and bifurcation of a pressurized circular annulus. J. Elast. 47, 121–145.
(doi:10.1023/A:1007448209058)
Oldfather, W. A., Ellis, C. A. & Brown, D. M. 1933 Leonhard Euler’s elastic curves. Isis 20,
72–160. (doi:10.1086/346767)
Pan, F. & Beatty, M. F. 1997 Instability of a Bell constrained cylindrical tube under end thrust-
part 1: theoretical development. Math. Mech. Solids 2, 243–273. (doi:10.1177/10812865970
0200301)
Rivlin, R. S. & Saunders, D. W. 1951 Large elastic deformations of isotropic materials. VII.
Experiments on the deformation of rubber. Phil. Trans. R. Soc. Lond. A 243, 251–288. (doi:10.
1098/rsta.1951.0004)
Shuvalov, A. L. 2003a A sextic formalism for three-dimensional elastodynamics of cylindrically
anisotropic radially inhomogeneous materials. Proc. R. Soc. A 459, 1611–1639. (doi:10.1098/
rspa.2002.1075)
Shuvalov, A. L. 2003b The Frobenius power series solution for cylindrically anisotropic radially
inhomogeneous elastic materials. Quart. J. Mech. Appl. Math. 56, 327–345. (doi:10.1093/
qjmam/56.3.327)
Simpson, H. C. & Spector, S. J. 1984 On barrelling instabilities in finite elasticity. J. Elast. 14,
103–125. (doi:10.1007/BF00041660)
Southwell, R. V. 1932 On the analysis of experimental observations in problems of elastic stability.
Proc. R. Soc. A 135, 601–616. (doi:10.1098/rspa.1932.0055)
Stroh, A. N. 1962 Steady state problems in anisotropic elasticity. J. Math. Phys. 41, 103.
Timoshenko, S. 1983 History of strength of materials. New York, NY: Dover.
Timoshenko, S. P. & Gere, J. M. 1961 Theory of elastic stability. New York, NY: McGraw-Hill.
Todhunter, I. 1893 A history of the theory of elasticity and of the strength of materials: from Galilei
to the present time. Cambridge, UK: Cambridge University Press.
Wilkes, E. W. 1955 On the stability of a circular tube under end thrust. Quart. J. Mech. Appl.
Math. 8, 88–100. (doi:10.1093/qjmam/8.1.88)
Zhu, Y., Luo, X. Y. & Ogden, R. W. 2008 Asymmetric bifurcations of thick-walled circular
cylindrical elastic tubes under axial loading and external pressure. Int. J. Solids Struct. 45,
3410–3429. (doi:10.1016/j.ijsolstr.2008.02.005)
17 Nonlinear Euler buckling
Proc. R. Soc. A

The two-dimensional theory of shells can be used when the thickness of the cylindrical shell is small enough. in principle. other modes of instability such as barrelling cannot be obtained.2 (a) A. 2 and 3. However. the theory cannot be directly applied to obtain information on the buckling instability (asymmetric buckling mode). These modes will typically occur for sufficiently stubby structures. A . In particular. (b) Lagrange solutions (1770). modes 1. their localization and selection (Hunt et al. R. energy argument (Timoshenko & Gere 1961) or directly from the exact solution. However. Here. Then. Goriely et al. to capture the instability. since material cross sections initially perpendicular to the axis remain undeformed and perpendicular to the tangent vector. Soc. The one-dimensional theory can be used with a variety of boundary conditions. Euler problem: (a) illustrations from Euler (1744). the stability analysis of shell equations ´ ´ such as the Donnell–von Karman equations leads to detailed information on symmetric instability modes. no information on the elastic deformation around the central curve can be obtained. The three-dimensional theory of nonlinear elasticity provides. (b) Figure 1. a complete and exact description of the motion of each material point of a body under loads. However. due to the mathematical complexity of the governing Proc. we refer to axisymmetric deformation modes of a cylinder or a cylindrical shell. it is particularly easy to explain and generalize and it can be used for large geometric deflections of the axis (Antman 1995). 2003). either by bifurcation analysis. by barrelling. which in the case of rods can be written in terms of elliptic integrals (Nizette & Goriely 1999).

subjected to a uniaxial constant strain l3 and deformed into a shorter tube with current dimensions a. Second. we derive a new and compact formulation of the bifurcation criterion that can be used efficiently for numerical approximation of the bifurcation curves for all modes. usually referred to. outer radius B and length L. l3 Þ. this computation is rather cumbersome as it is based on non-trivial tensorial calculations. we are concerned with the case of a cylindrical shell under axial load. Since the material is isotropic in the cross sections. most problems cannot be explicitly solved. we consider the problem of determining the critical aspect ratio where there is a transition between buckling and barrelling. R. This problem was first addressed in the framework of nonlinear elasticity in a remarkable 1955 article by Wilkes who showed that the linearized system around a finite axial strain can be solved exactly in terms of Bessel functions. ð2:1Þ where l1 Z a=AZ b=B and l3Zl/L. see below). the Euler criterion can then be recovered. The deformation bringing a point at (R. opened the door to further investigation by Fosdick & Shield (1963). In comparison to the one. q Z Q. First. 1993. the physical components of the corresponding deformation gradient F are ½F Š Z diagðl1 . theory. Levinson 1968. z Z l3 Z. Pan & Beatty 1997. the buckling instability can be recovered by a bifurcation argument. Simpson & Spector 1984.and two-dimensional theories. Duka et al. but it contains much information about the instability and the unstable modes selected in the bifurcation process. as small-on-large. These initial results constitute the basis for much of the modern theory of elastic stability of cylinders within the framework of three-dimensional nonlinear elasticity (Haughton & Ogden 1979a. in the nonlinear elasticity theory. Dorfmann & Haughton 2006). 2. Bigoni & Gei 2001. Q. the buckling instability can be captured by assuming that the object is either a rectangular beam (Biot 1962. A ð2:2Þ . or incremental.b. who recovered Euler’s criterion asymptotically from the solution of Wilkes. z) in the current configuration is r Z l1 R. Third. in cylindrical coordinates in the initial configuration. Here.Nonlinear Euler buckling 3 equations. based on the solution of Wilkes. The experimental verification of Euler’s criterion was considered by Southwell (1932) and by Beatty & Hook (1968). we revisit the problem of the stability of an incompressible cylindrical shell under axial load using the Stroh formalism (Stroh 1962) and. q. The purpose of this article is threefold. l1 . he noted in his conclusion that the asymmetric mode (nZ1) corresponds to the Euler strut and doing so. b and l. Proc. While Wilkes only analysed the first axisymmetric mode (nZ0. Using the theory of incremental deformations around a large deformation-stressed state. we use this formulation to obtain nonlinear corrections of Euler’s criterion for arbitrary shell thickness and third-order elasticity. Large deformation We consider a hyperelastic homogeneous incompressible cylindrical tube with isotropic cross sections of initial inner radius A. to (r. Soc. In the case of long slender structures under loads. Z ). Nowinski 1969) or a cylindrical shell under axial load. By taking the proper asymptotic limit for long slender structures.

not the minor ones. showing that the principal stretches are the constants l1. 9 > > > > > = if i sj. and we conclude that the principal Cauchy stresses are constant. (a ) Governing equations The incremental equations of equilibrium and incompressibility can be written as (Ogden 1984) div s Z 0. Goriely et al. ð3:2Þ _ where p is the increment in the Lagrange multiplier p and A0 is the fourth-order tensor of instantaneous elastic moduli. ð2:5Þ where WihvW/vli . we take the view that the existence of small deformation solutions in the neighbourhood of the large pre-strain signals the onset of instability (Biot 1965). with the differences that it possesses only the major symmetries. ð3:1Þ where s is the incremental nominal stress tensor and u is the infinitesimal mechanical displacement. i i j A 0ijij Z ðA 0iiii KA 0iijj C li Wi Þ=2. li Z lj . l3. A if i sj. R. In our case. and that it reflects strain-induced anisotropy instead of intrinsic anisotropy. This tensor is similar to the stiffness tensor of linear anisotropic elasticity. ð3:3Þ . and that the pre-strain is homogeneous. s2Zs1 because l2Zl1. det FZ1. where p is a Lagrange multiplier introduced by the internal constraint of incompressibility and W is the strain energy density (a symmetric function of the principal stretches). 2. s1Z0 because the inner and outer faces of the tube are free of traction. A 0ijji Z A 0jiij A 0ijij K li Wi Proc. l2Zl1. s3 Z l3 W3 K l1 W1 . so that l1 Z l3 vW vli K1=2 : ð2:3Þ The principal Cauchy stresses required to maintain the pre-strain are (Ogden 1984) si ZKp C li ði Z 1. It follows that: p Z l1 W1 .4 A. Instability To perform a bifurcation analysis. div u Z 0. Also. li slj . 3. 3Þ ð2:4Þ (no sum). Owing to incompressibility. A 0ijij Z ðli Wi Klj Wj Þl2 =ðl2 Kl2 Þ. Soc. > > > > > . They are linked by _ s Z A 0 gradðuÞ C p gradðuÞK pI . Its explicit non-zero components in a coordinate system aligned with the principal axes are (Ogden 1984) A 0iijj Z li lj Wij .

1. Srz ðrÞg !expðiðnq C kzÞÞ. Uz ðrÞ. G2 Z GC. The central idea is to introduce a displacement–traction vector. Srr ðrÞ. and Dorfmann & Haughton 2006 for the compressible counterpart). irSrz ŠT . where Wij h v2 W =ðvli vlj Þ. uz . srq . A 01221 C A 01212 A 01111 KA 01122 Z 2A 01212 C A 01331 KA 01313 : A 02323 Z A 01313 . Srq ðrÞ. and have yet unknown variations through the thickness of the tube. z). ð3:5Þ ð3:4Þ where nZ0. Negron-Marrero 1999. 2 C G3 G1 ð3:7Þ ð3:6Þ G3 Z GC: 3 ð3:8Þ Here the superscript ‘C’ denotes the Hermitian adjoint (transpose of the complex conjugate) and G1.Nonlinear Euler buckling 5 (no sums). with the block structure " # G1 G2 GZ . . In particular. Note that some of these components are not independent because here l1Zl2. Pan & Beatty 1997. A . the subscripts (r. Uz . G2 and G3 are the 3!3 matrices 3 2 0 0 0 3 2 3 6 7 2 k i Kn Kkr ik12 k13 11 7 6 7 6 60 K 1 0 6 7 7. Uq . Uq ðrÞ. The specialization of the governing equations (3. 6 5 4 Kn Ki A 01212 7 6 6 1 7 5 40 Kik13 k23 k33 Kkr 0 0 0 K A 01313 Proc. 9 A 02222 Z A 01111 . we have A 02121 Z A 01212 . srr .1) to this type of solution has already been conducted in several articles (see Wilkes 1955. uq . Fosdick & Shield 1963. so that the incremental equations can be written in the form d i hðrÞ Z GðrÞhðrÞ. 2. h h ½Ur . A 03232 Z A 03131 . srz g ZfUr ðrÞ. R. irSrq . z) for u and s refer to components within the cylindrical coordinates (r. > = A 02233 Z A 01133 . A 02332 Z A 01331 . (b ) Solutions We look for solutions that are periodic along the circumferential and axial directions. Soc. PðrÞ. p. irSrr . 4Kik12 k22 k23 7. and all upper-case functions are functions of r alone. > . is the circumferential number. ð3:9Þ 0 5. q. so that our ansatz is _ fur . k is the axial wavenumber. q. Mack 1989. dr r where G is a 6!6 matrix. Here we adapt the work of Shuvalov (2003a) on waves in anisotropic cylinders to develop a Stroh-like formulation of the problem (Stroh 1962).

we checked that the following vector h(3):   n 1 ð3Þ 0 0 I ðq krÞ . the six Bessel solutions are presented in the article by Wilkes (1955. Goriely et al. 9 k23 Z nð2A 01212 CA 01313 Þkr. we also checked that the vectors h(4). k33 Z n2 A01313 Cð4A 01212 C2A01122 KA 01111 CA 03333 K2A 01133 Þk 2 r 2 : As it happens.7): h i N ðrÞ Z hð1Þ jhð2Þ j. there exists a set of six explicit Bessel-type solutions to these equations when ns0. denote by q1 . > > = 2 2 2 ð3:10Þ k12 Z 4nA 01212 . hð2Þ Z iIn ðqkrÞ. Similarly.7): 0 hð1Þ . where explicit Bessel-type solutions exist only for transversely isotropic cylinders with a set of four linearly independent modes and do not exist for cylinders of lesser symmetry (Martin & Berger 2001. k22 Z 4n A 01212 CA 03131 k r .Kq3 In ðq3 krÞ. obtained by replacing In with the modified Bessel function Kn in the expressions above.Kið1 C q ÞkrA 01313 In ðqkrÞ . KinA01313 In ðq3 krÞ q3 kr n 3 ð3:13Þ is also a solution when q3 is the positive root of the quadratic equation A 01212 q 2 KA 03131 Z 0: ð3:14Þ n I ðqkrÞ.jhð6Þ : ð3:15Þ Proc. are solutions too. for a derivation see Bigoni & Gei 2001). 2nq h Z i In ðq3 krÞ. h(5) and h(6). As mentioned in §1.6 respectively. K ðA 01313 q C A 03131 ÞIn ðqkrÞ C 2A 01212 In ðqkrÞK q qkr n   !T 1 0 2 0 I ðqkrÞ . Next. k11 Z 4A 01212 CðA 03131 KA 01313 Þk 2 r 2 . K2inA 01212 In ðqkrÞK qkr n ð3:12Þ where qZq1.K   kr n2 2 0 I ðqkrÞ .K n ðqkrÞ. Soc. with A. > > . we follow Shuvalov (2003a) and introduce N (r) as a fundamental matrix solution to (3. and it can be checked that the following two vectors are solutions to (3. Shuvalov 2003b). A .K 3 A01212 In ðq3 krÞK kr q3 kr n 3   !T   n2 0 !iq3 A01212 2In ðq3 krÞK q3 kr C 2 I ðq krÞ . q2 in turn and In is the modified Bessel function of order n. 0. qI qkr n Finally. 2 2 First. q2 the roots of the following quadratic in q2: A 01313 q 4 KðA 01111 C A 03333 K2A 01331 K2A 01133 Þq 2 C A 03131 Z 0: ð3:11Þ Then the roots of this quartic in q are Gq1 and Gq2. R. k13 Z 2A 01212 kr. This situation is in marked contrast with the corresponding setup in linear anisotropic elastodynamics.

but. aÞhðaÞ. Srq . real or complex (Pan & Beatty 1997). aÞ say. which implies the bifurcation condition det zðb. Since S(b)Z0. l ) leads to kZ mp mp Z . aÞU ðrÞ.Nonlinear Euler buckling 7 It clearly satisfies d i N ðrÞ Z GðrÞN ðrÞ: dr r Let M(r. aÞ Z M 3 ðr. K1 where z h iM 3 M 1 ð3:21Þ is the (Hermitian) 3!3 impedance (Shuvalov 2003a). The other boundary conditions are that the inner and outer faces of the tube remain free of incremental tractions. aÞ ZKi det M 3 ðb. R. ð3:19Þ ð3:18Þ M 2 ðr. Uz ŠT the displacement vector. aÞ where mZ1. . simple or double (Dorfmann & Haughton 2006)) of the roots q1. 2.e. aÞ ð3:22Þ This is a real equation since zZzC (Shuvalov 2003a) that applies independently of the nature (i. A . M ða. 3. aÞ M ðr. Considering that they remain plane (UzZ0 on zZ0. aÞ Z I ð6Þ : ð3:17Þ ð3:16Þ It is obtained from N (r) (or from any other fundamental matrix made of linearly independent combinations of the h(i )) by M ðr.19) to find the following connection: rSðrÞ Z zðr. aÞ Z 0: det M 1 ðb.e. since the equations depend only on k. we can take mZ1 without loss of generality. (c ) Boundary conditions Some boundary conditions must be enforced on the top and bottom faces of the tubes. a) is singular. l ) and free of incremental shear tractions (Srz Z Srq Z 0 on zZ0.a) be the matricant solution to (3. the matrix such that hðrÞ Z M ðr. i. Uq .17) and (3. Proc. Srz ŠT the traction vector and U h ½Ur .7). We substitute the condition S(a)Z0 into (3. a nontrivial solution exists only if the matrix z(b. Soc. aÞ M 4 ðr. l l3 L ð3:20Þ # . q2 and q3. aÞ Z N ðrÞN K1 ðaÞ and it has the following block structure: " M 1 ðr. We call S h ½Srr .

M 3 ðr. Fu 2005) and if numerical integration was required it would provide an ideal representation of the governing equation. the determinantal method (Ben Amar & Goriely 2005) or the Adams– Moulton method (Zhu et al. Rather than integrating the original linear problem numerically. The adjugate method We are now in a position to use the bifurcation condition (3. This method bypasses the need for numerical integration and reduces the problem to a form that is manageable both numerically and symbolically.13). n.22) to compute explicitly bifurcation curves for each mode n.22) by factoring non-vanishing factors. the entries of M(b. the root finding of det(z) becomes numerically unstable and numerical methods become unreliable (as observed in Dorfmann & Haughton (2006) for a similar problem) and. to study analytically particular asymptotic limits. iZ1. We start by realizing that since the fundamental solutions {h(i ). the Stroh formalism is particularly well suited and well behaved (Biryukov 1985. Note that from a computational perspective. aÞN 3 ðaÞ Z N 3 ðrÞ. where each block is a 3!3 matrix. the bifurcation equation (3. 4. Proc. for mode nO1. R. for a given bifurcation mode (n specified). in explicit computations. . aÞN ðaÞ Z N ðrÞ. 6} are linearly independent. a) are bounded for r2[a. which by (2. Soc. which are M 3 ðr. which implies that the elements of M(r. Instead of expressing det(M3(b. a) thus depend (for a given W ) on l3.22) to compute all possible bifurcation curves.3) depends only on l3. and write explicitly the two entries N 3(r) and N 4(r)..22) gives a relationship between a measure K2 of the critical pre-stretch: l3 Z l1 . ð4:1Þ N 3 ðrÞ N 4 ðrÞ say. det(M1(r. aÞN 4 ðaÞ Z N 4 ðrÞ.b]. Therefore.11) and (3. We note that the components of A 0 depend on the strain energy density W and on the pre-strain. it has not been successfully implemented to compute all bifurcation curves. Indeed. For a given material (W specified) with a given thickness (b/aZB/A specified).8 A.14). a))Z0. for a given tube slenderness. q2. The main idea is to transform condition (3. we first decompose N (r) as " # N 1 ðrÞ N 2 ðrÞ N ðrÞ Z . ka and kb only. most authors do not use the exact solution by Wilkes but use a variety of numerical techniques to solve the linear boundaryvalue problems directly (such as the compound matrix method (Haughton & Orr 1997). the matrix N (r) is invertible for all r2[a. by (3. q3. aÞN 1 ðaÞ C M 4 ðr. and a measure of the tube slenderness: K3=2 kbZ 2pmðb=lÞZ 2pml3 ðB=LÞ. We also rewrite equation (3. That is. what is the axial strain necessary to excite a given mode? While this bifurcation condition is formally clear. so do q1. a)) is uniformly bounded away from zero and det zZ0 implies det(M3(b. aÞN 2 ðaÞ C M 4 ðr. According to (3. 2008)). a)) as the determinant of a 3!3 submatrix of a matrix obtained as the product of two 6!6 matrices.b]. we now show how to use an alternative form of (3.18) as M ðr. Goriely et al. A ð4:2Þ ð4:3Þ ð4:4Þ .12) and (3.

where Qðb. ð4:7Þ ð4:6Þ and adj(A) is the adjugate matrix of A.4 0. Proc. a))Z0 implies that det Qðb. det Q(b.9 0. Both numerically and symbolically. aÞ Z detðN 4 ðaÞÞN 3 ðbÞadjðN 3 ðaÞÞKdetðN 3 ðaÞÞN 4 ðbÞadjðN 4 ðaÞÞ. aÞ N 1 ðaÞN 3 ðaÞKN 2 ðaÞN 4 ðaÞ Z N 3 ðrÞN 3 ðaÞKN 4 ðrÞN 4 ðaÞ: ð4:5Þ Using again the fact that the entries of N are bounded.3 0. even in the limits a/0. i.5 0. Soc. and nZ0. we have that the bifurcation condition det(M3(b. a) is a polynomial of degree 18 in Bessel functions and has no denominator (hence no small denominator). i.2 0. R.1 9 0 5 10 15 kb 20 25 30 Figure 2. A .8 0. the transpose of the cofactor matrix (which in the case of an invertible matrix is simply adj(A)Zdet(A)AK1).0 0.Nonlinear Euler buckling 1.e.7 0. This new bifurcation condition is equivalent to the previous one but has many advantages. which implies À Á K1 K1 K1 K1 M3 ðr. Bifurcation curves (stretch as a function of stubbiness) of a homogeneous neo-Hookean cylindrical tube for modes nZ0–8 with b/aZB/AZ2 and C1Z1. this determinant is well behaved.e. We will refer to the use of this form of the bifurcation condition as the adjugate method. which corresponds to the first barrelling mode (and usually requires a special treatment). The matrix Q involves only products of 3!3 matrices and is polynomial in the entries of N . which corresponds to a solid cylinder.6 0. aÞ Z 0.

there is a critical slenderness value at which the first barrelling mode nZ0 is the first unstable mode (in a thought experiment where the axial strain would be incrementally increased until the tube becomes unstable). the buckling mode rapidly ceases to be the first excited mode and is replaced by different barrelling modes. we consider very thin neo-Hookean tubes with B/AZ1. A . where we set C1Z1 without loss of generality and consider the typical value B/AZ2.98 0.97 0. For a second test. which corresponds to surface instability of a compressed half-space (Biot 1962).01. Soc.10 1. the material transition from either the one-dimensional behaviour of slender column to the two-dimensional behaviour of a thin short tube.99 0.96 0. 6 7 8 9 0. modes nZ1–9 are selected (modes nZ0 and 10 remain unobservable).444.01 and C1Z1. the critical compression ratio tends asymptotically to the value lZ0.94 0 1 2 3 4 5 kb 6 7 8 9 10 Figure 3. There is one particularly interesting feature in these two sets of bifurcation curves. we consider a neoHookean potential W Z C1 ðI1 K3Þ=2. the Euler buckling (nZ1) is dominant and becomes unavoidable as the slenderness increases. Accordingly. Goriely et al. Depending on both the tube thickness and the stubbiness. we will refer to these particular geometric values where transition occurs as dimensional transitions and obtain analytical estimates for them in §5. (a ) Numerical results As a first test of the stability of the numerical procedure. Here we are interested in the mode selection process. As the stubbiness increases. From figure 3. Bifurcation curves (stretch as a function of stubbiness) of a homogeneous neo-Hookean cylindrical tube for modes nZ0–10 with b/aZB/AZ1. or the three-dimensional behaviour of a thick short tube. the instability mode of a tube transition occurs from buckling to barrelling. R. We compute the critical value of lhl3 as a function of the current stubbiness kbZpb/l (the initial stubbiness is nZ B=LZ kbl3=2 =p) for the first 9 modes (nZ0–8) as shown in figure 2.95 0.00 1 2 3 4 5 A. The known classical features of the stability problem for the cylindrical shell are recovered. and for very large kb. Proc. it appears clearly that as kb increases. namely for slender tubes.

which we denote by Euler2. 144ðC1 C C2 Þr4 h 1 ð973C1 C 341C2 Þðr10 K 1Þ lð6Þ ZK 4608r6 ðr4 K1ÞðC1 C C2 Þ C ð7073C1 C 3385C2 Þr2 ðr6 K1Þ C 4392 lnðrÞðC1 C C2 Þðr6 C r4 Þ i C 4ð377C2 C 1141C1 Þðr6 Kr4 Þ . we expand dðlÞZ det Qðb. It is of interest to compare the different approximations. We show the different approximations as a function of e2 for rZ1. We recover the Euler formula by keeping only the term up to e2. R. Md X d m em C OðeMd C1 Þ. The asymptotic form of the Euler criterion cannot be obtained for a general strain-energy density.01 Proc. and I2Z 1 2 3 l2 l2 C l2 l2 C l2 l2 . dðlÞ Z mZ1 ð5:4Þ and solve each order dmZ0 for the coefficients lm. corresponds to the most general form of third-order incompressible elasticity (see §6 and Rivlin & Saunders (1951)). for l close to 1. This is why we choose the Mooney–Rivlin potential. M X lm em C OðeMC1 Þ: ð5:3Þ l Z lðeÞ Z 1 C mZ1 Similarly. Close to lZ1. ð5:8Þ where rhB/AZb/a. with lð2Þ ZK lð4Þ Z r2 C 1 . ð5:1Þ where C1R0 and C2O0 are material constants.Nonlinear Euler buckling 11 5. we introduce a small parameter related to the 1 2 2 3 3 1 stubbiness ratio e Z kb Z pb=l. 4r2 ð5:6Þ ð5:7Þ ð5:5Þ ð19C2 C 28C1 Þr4 C 2ð53C2 C 62C1 Þr2 C 19C2 C 28C1 . We define similarly Euler4 and Euler6 by keeping terms up to orders 4 and 6 in e. Soc. which. This is a rather cumbersome computation. which is found to be to order 6 in e l Z 1 C lð2Þ e2 C lð4Þ e4 C lð6Þ e6 C Oðe8 Þ. I1 Z l2 C l2 C l2 . aÞ in powers of e. A computation to order 28 is necessary to compute the correct expression for a. A . W Z C1 ðI1 K3Þ=2 C C2 ðI2 K3Þ=2. It can be checked that lm vanishes identically for all odd values of m and that the first non-identically vanishing coefficient appears at order 24. Asymptotic Euler buckling We are now in a position to look at the asymmetric buckling mode (nZ1) corresponding to the Euler buckling in the limit l/1. ð5:2Þ and look for the critical buckling stretch l as a function of e of order M.

Nonlinear Euler buckling for third-order elasticity The analytical result presented in §5 was formulated in terms of parameters and quantities natural for the computation and the theory of nonlinear elasticity.12 (a) 1.6 0.6) and (5. C2Z0.91 0.25 0. Soc.20 0. We first consider the geometric parameters. We wish to express the critical load as a function of the initial stubbiness nZB/L and tube relative thickness rZB/A. 6.00 0.10 0.01.90 0 Euler2 exact Euler4 A.1) to obtain e2 Z p2 n2 K3p4 lð2Þ n4 Kð3p6 lð4Þ K15p6 l2 Þn6 C Oðn8 Þ. the axial load acting on the cylinder and the elastic parameters entering in the theory of linear elasticity.950 0.8 0.0 Euler4 Euler2 exact Euler6 Figure 4.15 0.99 0. Euler4 and Euler6 for a neo-Hookean potential C1Z1. Goriely et al.3 0. but the Euler4 and Euler6 approximations clearly improve the classical formula for larger values of e.94 0.900 0.93 0.1 0.95 0.7) and come from the expansion of l in powers of e. we show the critical stretch for mode nZ1 versus e2 in which case the graph becomes linear in the limit e/0. For comparison purpose.2 0.925 0.7 0. we need to express equation (5. (b) rZb/aZB/AZ10. we have e 2 l 3 Z p2 n 2 : ð6:1Þ To express e as a function of n. B.5 0.875 Euler6 0. Recalling that eZpb/l and lZl/L.4 0. Comparison of the different Euler formulae obtained by expanding the exact solution to order 2 (the classical Euler buckling formula). Proc. The classical Euler formula is well recovered in the limit e/0.850 0.98 0.96 0. ð2Þ ð6:2Þ where l(2) and l(4) are defined in (5. we expand e in powers of n to order 6. A . In order to relate this result to the classical form of Euler buckling. bZlK1/2B. L. (figure 4a) and rZ10 (figure 4b).5) in terms of the initial geometric values A. and solve (6.30 0 0. (b) 1.975 0.9 1. R.92 0. It also appears from the analysis of Euler4 that for C2R0 the classical Euler formula always underestimates the critical stretch for instability.05 0. (a) rZb/aZB/AZ1.000 0.97 0.

Soc. I3 of the Cauchy–Green strain tensor by I1 Z 2i1 C 3. I3 Z 2i1 C 4i2 C 8i3 C 1: ð6:6Þ Since I3Z1.2) and. or second Lame parameter.  m n  n  W Z m C 3 I1 K K 3 I2 . A ð6:9Þ . i2. ðb N ZK 2p rs3 dr: ð6:3Þ a Since s3 is constant and given by (2.1) leads to C1 Z 2m C n 3 =4. also. I2. related to the first three principal invariants I1. we want to relate the axial compression to the actual axial load N. we follow Hamilton et al.e. we can solve this linear system for i2 and i3 and write the strainenergy density (6. 2mi ð6:5Þ ´ where m is the usual shear modulus. R. ð6:7Þ 2 8 8 which by comparison with (5. and n 3 is a thirdorder elasticity constant.5) as a function of I1 and I2. C2 Z mK n 3 =4: ð6:8Þ To write the nonlinear buckling formula. in Murnaghan’s notation. i3 are the first three principal invariants of the Green–Lagrange strain tensor. (2004. n 3Zn and in Landau’s notation.4) and first expand l in e using (5. we relate the elastic Mooney parameters C1 and C2 to the classical elastic parameters. I2 Z 4i1 C 4i2 C 3. To do so. In (6. i1. Here. i.5).5). which yield NZ 3 p3 B 2 mðr4 K1Þn2 4 r4 1 p5 B 2 ðr2 K 1Þð20r4 m C 9r4 n 3 C 176r2 m C 18r2 n 3 C 20m C 9n 3 Þn4 K 96 r6 2 7 6  B pn 323m2 r8 K3n 2 K240r2 mn 3 K9r2 n 2 K9r10 m2 C 9m2 C 3 3 8 mðr2 C 1Þ 512r C3r10 n 2 C 1464 lnðrÞr6 m2 C 1464 lnðrÞr4 m2 C 240r8 mn 3 3 C240r6 mn 3 K180r6 m2 C 180r4 m2 C 9r8 n 2 3 à 6 2 4 2 2 2 4 C6r n 3 K6r n 3 K323r m K240r mn 3 C Oðn8 Þ: Proc. Destrade & Saccomandi 2005) and write the strain-energy density to third order for an incompressible elastic material as W ZK 2 C n 3 i3 . i. we have p N ZKpðb2 Ka 2 Þs3 ZK ðB 2 KA2 Þs3 l p ZK 3 ðB 2 KA2 Þ½ðl4 KlÞC1 C ðl3 K1ÞC2 Š: ð6:4Þ l Third. we consider (6. then expand e in n using (6. we integrate the axial stress over the faces of the tubes. finally.e.5).Nonlinear Euler buckling 13 Second. m is related to Young’s modulus by EZ3m. substitute the values of the moduli in terms of the elastic parameters. n 3ZA (see Norris (1998) for other notations).

or from mode nZ1 to mode nZ2 (r close to 1) as e increases. > rt Z 1 C e2 K e4 C = 4 32 384 ð7:1Þ 1 13 665 6 > e C Oðe8 Þ: . For sufficiently slender tube (n small). Indeed both modes of instability can be captured by. rà . in this regime. respectively. we use the buckling formula to compute the transition between modes as parameters are varied.1) (using rZB/A. It then puts the method to use. the buckling mode disappears when rOrt at the expense of the nZ2 barrelling mode. Goriely et al. From figures 2 and 3. Dimensional transition Finally.14 A. Conclusion This article establishes a reliable and effective method to study the stability of tubes based on the exact solution of the incremental equations proposed by Wilkes (1955) within the Stroh formalism. as 9 3 53 2393 6 e C Oðe8 Þ. it is comforting to recover to order n2 the classical Euler buckling formula (1. 7. Expanding again this bifurcation condition in e as well as r. While it is not surprising. one identifies the values rt of r and lt of l at which the transition occurs. That is. for the intersection of two different modes. To understand the dimensional transition. a one. We refer to this transition as a dimensional transition. The corresponding transition point in terms of the initial parameters is nà Z ðeà =pÞðlà Þ3=2 . we find the value of eà such that both the bifurcations for either modes nZ1 and 2. it appears clearly that for e small enough there is a transition (depending on the value of r) from either mode nZ1 to mode nZ0 (large r). In figure 5. we solve numerically the bifurcation condition. For stubbier and fuller tubes. nZB/L and mZE/3). R. to identify both the geometric values and the axial strain for which there is a transition between buckling and barrelling modes. or modes nZ1 and 0 are satisfied. this approximation cannot be used. in the sense that the material mostly behaves as a slender one-dimensional structure when it buckles according to mode nZ0 and mostly as a two-dimensional structure when it barrels with mode nZ2. Soc. Therefore. the transition nZ0/nZ1 occurs for small values of e. rÃ) is a transition point. If the corresponding value là is the largest value for which a bifurcation takes place.5) for the barrelling curve and substitute it in the bifurcation condition of mode nZ2. A . the pair (eÃ. That is.or a two-dimensional theory. to Proc. lt Z 1K e2 C e4 K 4 8 96 In terms of the initial stubbiness nZB/L. we have 3 17 161 6 6 ð7:2Þ p n C Oðn8 Þ: rt Z 1 C p2 n 2 K p4 n 4 C 4 32 384 This relationship also provides a domain of validity for the Euler buckling formula. For r close to unity. we show a diagram of all such pairs for both transitions. using the adjugate method. for a given value of rÃ. we can use the approximation (5. Here we restrict again our attention to the neo-Hookean case (with C1Z1). 8.

obtain the first geometric and material corrections to the Euler buckling.5 0.6 1.1 0. In all these cases.4 2. Dimensional transition for a neo-Hookean cylindrical tube of initial length L and initial radii A and B.4 0. it will not buckle but instead will be subjected to a barrelling instability. The method presented here can be easily generalized to different materials and different boundary conditions.0 2.2 0. Zhu et al.6 0. a variety of boundary-value problems can be analysed by the adjugate method. Soc. such as the stability problem of a tube under pressure and tension (Han 2007.4 1.8 1. All tubes in the nZ1 regions will become unstable by buckling. Tubes in the barrelling regions may be subjected to other unstable modes. an explicit form of the bifurcation condition in terms of Bessel functions can be obtained by following the steps presented here and various asymptotic behaviours can be obtained. using the exact solution of the incremental equations proposed in Dorfmann & Haughton (2006) for compressible materials and the adjugate method.6 2.3 0. a weakly nonlinear analysis of the solution should be possible to third order. 2008). A . Since the solutions of the linearized problem can be solved exactly. It is also enticing to consider the possibility of performing an analytical post-buckling analysis of the solutions.Nonlinear Euler buckling 15 3. As the tubes get stubbier or thinner (arrows).8 2. For instance. The method can be also used to obtain the transition between buckling and barrelling modes when a tube becomes unstable.0 0 0.2 1. useful asymptotic formulae for the buckling behaviour could be obtained by perturbation expansions. This would yield.7 0.0 approximation L Figure 5. 1997). the problem of a tube embedded in an infinite domain (Bigoni & Gei 2001) and the problem of a tube with coating (Ogden et al.2 2. Note that only the transition curves from mode nZ0 are shown.0 1. R. Similarly. in principle.9 1. an equation for the amplitude of the unstable modes containing much information not Proc.8 0.

A . 98. L. A.2005. A. 13. Phys.. (doi:10. elastic cylinder. E 72. M.16 A. M. 179 –212. Beatty.018) Haughton. rubber bars under end thrust. L. Solids 53. Mech. & Shield.1016/j. & Spencer. Ration. R. J. Acta Mech. G.1007/BF00281227) Fu. 505–518. 12. 34. Phys.) and made possible by a CNRS/USA Collaborative Grant from the French Centre National de la Recherche Scientifique (M. 1997 On the eversion of compressible elastic cylinders. S. M. 116. D.06. R. vol. Biryukov. only about the actual amplitude of the unstable modes but also on the localization of unstable modes after bifurcation. 1016/j.ijnonlinmec. 1979a Bifurcation of inflated circular cylinders of elastic material under axial loading-II. 1979b Bifurcation of inflated circular cylinders of elastic material under axial loading-I. S. M. H.1736652) Han. 1968 Some experiments on the stability of circular. A. Germany: Memoires de l’ Academie de Berlin.06. M. J. 123 –166. Berlin. DMS-0604704 (A. Mech.016620) Dorfmann. 1893 –1914. (doi:10. & Haughton. 107–121. 2006 Stability and bifurcation of compressed elastic cylindrical tubes. Phys. T. 489 – 512. S. E.014) Duka.). & Hook. E. 183–198. 1744 Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes sive solutio problematis isoperimetrici latissimo sensu accepti. 31. W. Rev.. A. R. Phys. H. Mech. Mech. 1963 Small bending of a circular bar superposed on finite extension or compression. 1770 Sur la figure des colonnes. & Ogden. (doi:10. England. V. G. Int. J. Int. Destrade. Lagrange. D. M. Solids Struct. J.jmps. A. J. C. 4. NY: Wiley.72. Sci. Lausanne. M. F. Phys. References Antman. 2003 Cylindrical shell buckling: a characterization of localization and periodicity. Solids 27. & Saccomandi. D. 229–239. Int.2007. 623 – 635. (doi:10. & Orr. Acoust. Arch. 016 620.1016/S0020-7683(00)00322-X) Biot. 1985 Impedance method in the theory of elastic surface waves.04. Lord. Ilinskii. A. & Gei. J.05. 2004 Separation of compressibility and shear deformation in the elastic energy density. R.2004. Int. (doi:10. Geneva: Marc-Michel Bousquet & Co. R. Goriely et al. Int. Miscellanea Taurinensia 5. J. 5117– 5148.1103/PhysRevE. Soc. 12. (doi:10.1016/0022-5096(79)90001-2) Haughton. A. Biomech. Solids 27. E. 1962 Exact theory of buckling of a thick slab. We leave this daunting task for another day. A.ijengsci. B 3. W. 2005 Growth and instability in soft tissues. 1993 Bifurcation of a solid circular elastic cylinder under finite extension and torsion.1016/j.002) Hamilton. 2005 Finite amplitude elastic waves propagating in compressible solids. 38. J. Exact theory for thick-walled tubes. 44.1007/BF01174297) Euler. M. D. F. & Peletier. Syst. New York. ` ´ Euler. J. B.G. Discrete Contin. Anal.1016/0022-5096(79)90027-9) Haughton. & Ogden. A. M. Res. A.1016/j. Solids Struct.. (doi:10. Solids Struct.1016/S0020-7683(96)00122-9) Hunt. Eng. M. Appl. J. Biot. 2001 Bifurcations of a coated. L. Soc. Sci. 2284–2319. & Goriely. 1995 Nonlinear problems of elasticity. 223–248. Nonlin. (doi:10. Fosdick. J. 1965 Mechanics of incremental deformations. J. New York. L. W.1016/0020-7683(68)90078-4) Ben Amar.2006. 40. D.008) Bigoni. J. NY: Springer.D. This study is based in part upon work supported by the National Science Foundation under grant no. & Zabolotskaya. Proc. Sov. 2007 A biomechanical model of artery buckling. Y. 3672–3678. Y. G. D.1121/ 1. 1353 –1365. M. (doi:10. (doi:10. Mech. 350– 354. Am. (doi:10. (doi:10. M. 2005 An explicit expression for the surface-impedance matrix of a generally anisotropic incompressible elastic material in a state of plane strain. M. 41–44. 1759 Sur la force des colonnes. Dyn.jbiomech. D. 40. (doi:10. Acoust. Membrane theory for thin-walled tubes.

Math. & Berger. 2008 Asymmetric bifurcations of thick-walled circular cylindrical elastic tubes under axial loading and external pressure. J. 1007/BF01377530) Ogden. (doi:10. (doi:10. CA: Academic Press. & Gere. Solids 16. 14. Cambridge. 103–125. Chicago. M.2002. Mech. UK: Cambridge University Press. L. 2830 –2866.1093/qjmam/8. 1932 On the analysis of experimental observations in problems of elastic stability. C. Phys. R. (doi:10. D.1007/BF00041660) Southwell. 40. Phys.. Timoshenko. R. J. L. Int. W. 1893 A history of the theory of elasticity and of the strength of materials: from Galilei to the present time. Mech. Solids 4. Trans. Math. 2001 Waves in wood: free vibrations of a wooden pole. D. Mech. F.ijsolstr. 2003a A sextic formalism for three-dimensional elastodynamics of cylindrically anisotropic radially inhomogeneous materials. S. R. J. 1983 History of strength of materials. & Goriely. N. 91–99. Acta Mech.Nonlinear Euler buckling 17 Levinson. J. 1984 On barrelling instabilities in finite elasticity. & Beatty.1098/rspa. 1997 The effect of elastic surface coating on the finite deformation and bifurcation of a pressurized circular annulus. A. Solids 2. Ellis. E. S. (doi:10. Ogden. NY: Dover.532731) Norris. 1969 On the elastic stability of thick columns. 1961 Theory of elastic stability. Blackstock). Elast. W. Luo. C. Quart. A 459.005) Proc. 327–345. (doi:10. J. (doi:10. W. Solids 49. (doi:10. Soc. & Haughton. V.02. 1997 Instability of a Bell constrained cylindrical tube under end thrustpart 1: theoretical development. J. (doi:10. 7. J. 45.1093/ qjmam/56. 279 –286. 1999 Towards a classification of Euler–Kirchhoff filaments. P. F. Quart. Math. 56. R. Math. Math. W. A. J.88) Zhu.. 1998 Finite amplitude waves in solids. 1999 An analysis of the linearized equations for axisymmetric deformations of hyperelastic cylinders. J. Nizette. (doi:10. (doi:10. 47.1016/0022-5096(68)90004-5) Mack. A. R. 72–160. Phil. Experiments on the deformation of rubber.0004) Shuvalov. New York. (doi:10. pp. Soc. W. R. R. Timoshenko. Phys. Acta Mech. (doi:10.2008. Proc. Phys. J. San Diego. 243–273. & Brown. R. A 135. H. (doi:10. 403 – 415. 1098/rsta. NY: McGraw-Hill. M. A . Lond. R. X. D.3. Steigmann. Mech. 1951 Large elastic deformations of isotropic materials. Mech. 121–145. A.1177/108128659900400107) Niklas. Appl. 3410–3429. 77. P. 251–288. 1968 Stability of a compressed neo-Hookean rectangular parallelepiped. New York. D.1951. A. M. W. 263–277. Y. Wilkes. Todhunter. W. J. K. 1155–1178. & Ogden. & Spector. Soc. New York. A.1086/346767) Pan. Isis 20. I. Soc. J. L.1016/S0022-5096(00)00068-5) Negron-Marrero. 88–100. Solids Struct. (doi:10. T. S. A 243. Nowinski. IL: The University of Chicago Press.1023/A:1007448209058) Oldfather. J. (doi:10. M. 8. 2003b The Frobenius power series solution for cylindrically anisotropic radially inhomogeneous elastic materials. 109 –133. In Nonlinear acoustics (eds M. 1989 On the deformation of circular cylindrical rubber bars under axial thrust. F.. 1933 Leonhard Euler’s elastic curves. J. Y.1932. NY: Dover. 103.1177/10812865970 0200301) Rivlin. S. Elast. V. Mech.1016/j. A. R. 1984 Non-linear elastic deformation. & Saunders.1063/1.1. M.1098/ rspa.1007/BF01379744) Martin. Proc. Math. A. Appl. (doi:10. 1955 On the stability of a circular tube under end thrust. 1611–1639. 41. Hamilton & D. M. P. VII.327) Simpson. 1992 Plant biomechanics: an engineering approach to plant form and function. J. 601–616. 1962 Steady state problems in anisotropic elasticity.1075) Shuvalov.0055) Stroh.

Sign up to vote on this title
UsefulNot useful