You are on page 1of 13

International Journal of Solids and Structures 39 (2002) 2731–2743

www.elsevier.com/locate/ijsolstr

Couple stress based strain gradient theory for elasticity


a,b
F. Yang , A.C.M. Chong a, D.C.C. Lam a,*
, P. Tong a

a
Department of Mechanical Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon,
Hong Kong, PR China
b
Institute of Computational Engineering and Science, Southwest Jiaotong University, Chengdu 610031, Sichuan, PR China
Received 13 January 2001; received in revised form 29 January 2002

Abstract
The deformation behavior of materials in the micron scale has been experimentally shown to be size dependent. In
the absence of stretch and dilatation gradients, the size dependence can be explained using classical couple stress theory
in which the full curvature tensor is used as deformation measures in addition to the conventional strain measures. In
the couple stress theory formulation, only conventional equilibrium relations of forces and moments of forces are used.
The couple’s association with position is arbitrary. In this paper, an additional equilibrium relation is developed to
govern the behavior of the couples. The relation constrained the couple stress tensor to be symmetric, and the sym-
metric curvature tensor became the only properly conjugated high order strain measures in the theory to have a real
contribution to the total strain energy of the system. On the basis of this modification, a linear elastic model for iso-
tropic materials is developed. The torsion of a cylindrical bar and the pure bending of a flat plate of infinite width are
analyzed to illustrate the effect of the modification. Ó 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Couple stress; Strain gradient; Elasticity

1. Introduction

The size dependence of deformation behavior in micron scale had been experimentally observed in metals
(Fleck et al., 1994; St€
olken and Evans, 1998; Nix, 1989; Stelmashenko et al., 1993; Ma and Clarke, 1995;
Poole et al., 1996) and polymers (Lam and Chong, 1999; Chong and Lam, 1999). The behavior cannot be
explained by the conventional theories of mechanics and the couple stress theory (Fleck and Hutchinson,
1997) has been used to explain the size dependence of the deformation behavior. The difference between the
conventional theories of mechanics and the couple stress theories of mechanics are discussed below.
In the mechanics of particles (Martin and Thornton, 1995) , forces applied onto the material particles are
unobservable. The forces can only be determined through observations on the motions of the material
particles. They can also be determined by investigating the change of kinetic energy of the material particles.
In conventional Newtonian mechanics, a material particle in a deformable continuum can only undergo
translations. The applied force on the material particle is the only power to alter its motion and the force is

*
Corresponding author. Tel.: +852-2358-8661; fax: +852-2358-1543.
E-mail address: medcclam@ust.hk (D.C.C. Lam).

0020-7683/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 0 - 7 6 8 3 ( 0 2 ) 0 0 1 5 2 - X
2732 F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743

related to the acceleration of the material particle through Newton’s second law. (Alternately, the force can
be defined from the perspective of the kinetic energy of the material particle using d’Alembert’s principle.)
In conventional mechanics, a force drives a material particle to translate. In the classical couple stress
theories (Toupin, 1962; Mindlin and Tiersten, 1962; Koiter, 1964; Mindlin, 1964) for linear elastic mate-
rials, the applied loads on the material particle include not only a force to drive the material particle to
translate but also a couple to drive it to rotate. In this classical conception, only the conventional equi-
librium relationships of forces and moments (of forces) are enforced and the couple is unconstrained in the
absence of higher order equilibrium requirements.
More recently, Eringen and coauthors have studied deformation behavior in the microscale continua.
Their works are summarized in Eringen’s published works in 1971 and in 1998 (Eringen, 1971, 1998). In
their micropolar theory, they suggested that the location vector and a rigid vector representing its inner
rotation define the motion of a particle. The particle’s motion is characterized by the changes of the vectors.
The equilibrium relations in the micropolar theory are established on the basis of conservational laws of
momentum and moment of momentum. Since there is no higher order equilibrium relation in the micro-
polar, microstretch and micromorphic continua works, the attachment of the rigid vector for rotation to a
particle is arbitrary.
In this paper, the concept of representative volume element is introduced, and the force and couple
applied to a single material particle are defined. A new set of equilibrium relations for a system of material
particles to account for the rotations of these material particles is developed. The results are then gener-
alized to the couple stress theory of continuum. By the introduction of a higher order equilibrium con-
dition, the arbitrary nature of couples in the classical couple stress theory is resolved without the use of rigid
vector attachment condition, as was used in the micropolar theory. The paper will also show that the new
equilibrium relations dictated that the couple stress tensor must be a symmetric tensor. Since the symmetric
part of curvature tensor is the additional measure of deformation that conjugates to the couple stress, this
means that the antisymmetric part of curvature tensor does not conjugate to the couple stress and it does
not appear explicitly in the deformation energy density function. On the basis of this result, a theoretical
framework for a modified couple stress theory is developed, and a linear elastic constitutive model is de-
veloped within this framework. We will show that the number of required material length scale parameters
is only one in the new modified couple stress framework, instead of two in the classical couple stress theory.
Two simple applications of the modified couple stress theory, torsion of a thin cylinder and pure bending of
a flat plate of infinite width, are presented.

2. Equilibrium relationship in conventional mechanics of particles

In conventional theories of continuum mechanics, a material body is modeled as a continuum consisting


of an infinite number of material particles. We regard each material particle as a geometric point with
specific geometric and physical characteristics, including location, mass and motion. When the material
particle is modeled as a geometric point, it can only undergo translation motion, and is changed only by the
force applied onto the particle. Denoting the displacement vector of the particle as u and the force vector
applied to it as F, we can write Newton’s second law of motion as
F ¼ m€
u; ð1Þ
where m is the mass and € u is the acceleration vector of the material particle. According to d’Alembert’s
principle (Martin and Thornton, 1995), the force can be defined from the perspective of the kinetic energy
of the material particle as
D ok
F¼ ; ð2Þ
Dt ou_
F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743 2733

where D=Dt is the total time derivative operator. The kinetic energy of the particle k is
k ¼ 12mu_  u_ : ð3Þ
If a set of forces is applied to the material particle, the resultant force F, the vector sum of the forces, is
related to the acceleration or kinetic energy of the material particle through
X X D ok
F¼ Fi ¼ m€ u or F ¼ Fi ¼ : ð4Þ
Dt ou_
For static equilibrium, the forces on the material particle are balanced. The equilibrium relation for a single
material particle in conventional mechanics becomes
X
F¼ Fi ¼ 0: ð5Þ

In order to obtain the equilibrium relations for a system of material particles, an equivalent system of
forces is investigated. As shown in Fig. 1, a force FA applied to a material particle at point A in a system in
Fig. 1(a) is equivalent to an equal force F0A at another point B of the system and a couple of forces L0A
applied to the system as shown in Fig. 1(c). The equivalence is established through an intermediate step
shown in Fig. 1(b), in which two equal and opposite forces F0A and F00A with F0A ¼ FA are applied to point B
and the force FA and the force F00A form the couple L0A . The equivalent force and couple are, respectively
F0A ¼ FA ; L0A ¼ ðxB  xA Þ  FA ; ð6Þ
where xA and xB are the position vectors of points A and B.
In conventional mechanics, a couple of forces is a free vector in the space of the material particle system.
The couple can be translated and applied to any point in the system, which means that the motive effect of a
couple on the system of material particles is independent of the location where the couple is applied. Thus,
the forces Fi and the couples of forces Li applied to a set of material particles within the system is equivalent
to a resultant force and a resultant couple of forces, and the couple can be applied to an arbitrary point
within the system. The resultant force F0 and the couple of forces L0 are, respectively,
X X
F0 ¼ Fi ; L0 ¼ ðxi  Fi þ Li Þ; ð7Þ

where the origin is specified as the point onto which the equivalent resultant force is applied. If the system
of the material particles is in equilibrium, the resultant force and couple vanish, giving the conventional
equilibrium relations
X X
Fi ¼ 0; ðxi  Fi þ Li Þ ¼ 0; ð8Þ

for a system of material particles.

Fig. 1. Equivalence of force: a force at point A is equivalent to a force at point B and a couple of the force.
2734 F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743

3. Generalized motions and forces of a particle

In general higher order theories, the material particle should possess the conventional characteristics of
translation, but should also possess the characteristics of rotations and deformations. The conventional
conception of material particle as a geometric point is inadequate for use in higher order theories of me-
chanics. Rotations of material particles in higher order theory can be realized by encasing the material
particle with a representative volume element. The representative volume element is hypothesized as an
infinitesimal neighborhood of the material particle consisting of an infinite number of smaller material
particles. Each of the smaller material particles is also encased with an even smaller representative volume
element. Thus, the behavior of a material particle at a specific scale can be characterized by the behavior of
the smaller material particles in its representative volume element. For example, consider an arbitrary
material particle located in the representative volume element of the material particle characterized by a
location vector x and an infinitesimal volume dv. The displacement u0 of the conventional material particle
at x0 can be represented by
u0 ¼ uðx0 ; tÞ ¼ uðx; tÞ þ hðx; tÞ Dx; ð9Þ
where Dx ¼ x0  x and h is the rotation vector. h is defined as
h ¼ 12 2: u r; ð10Þ
in which 2 is the alternating tensor and r is the Hamiltonian differential operator. The velocity of the
conventional material particle at x0 is
u_ 0 ¼ u_ ðx0 ; tÞ ¼ u_ ðx; tÞ þ h_ ðx; tÞ  Dx; ð11Þ
where h_ is the angular velocity. The kinetic energy of the representative volume element at x, is then
Z
1 0 0 0 0
kðx; tÞ ¼ q u_  u_ dv
dv 2
Z Z Z
1 1
¼ q0 dv0 u_  u_ þ q0 Dx dv0  u_  h_ þ h_  q0 ½ðDx  DxÞg  Dx Dx dv0  h_ ; ð12Þ
2 dv dv 2 dv

where q0 is the mass density of the particles and g is the unit tensor. If x is the mass center of the repre-
sentative volume element, the above equation becomes
kðx; tÞ ¼ 12mu_  u_ þ 12h_  I  h_ ; ð13Þ
where m and I,
Z
m¼ q0 dv0 ;
dv
Z ð14Þ
I¼ q0 ½ðDx  DxÞg  Dx Dx dv0
dv

are the mass and the moment of inertia tensor of the particle, respectively. The generalized forces applied to
the particle encased by the representative volume element are defined as
D ok D ok
F¼ ¼ m€
u; L¼ ¼I€
h; ð15Þ
Dt ou_ Dt oh_
where F is a force and L is a couple. F and L drive the translation and the rotation of the particle, re-
spectively. If a set of forces Fi and a set of couples Li are applied to the particle, the resultant force F and
couple L are related to the kinetic energy of material particle via
F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743 2735

X D ok X D ok
F¼ Fi ¼ ¼ m€
u; L¼ Li ¼ ¼ I  €h: ð16Þ
Dt ou_ Dt oh_
In static equilibrium, the forces and couples on the material particle encased by the representative volume
element are
X X
F¼ Fi ¼ 0; L ¼ Li ¼ 0: ð17Þ

Thus, the equilibrium relations for a single material particle consist of the conventional equilibrium relation
of forces as in Eq. (5) and the additional equilibrium relation of couples. This concept of representative
volume element containing couples is now suited for use as a foundation for couple stress theories.

4. Equilibrium of a material particle system

Using the equilibrium relations for a single material particle encased with a representative volume ele-
ment as a basis, we can now establish the equilibrium relations for a system of material particles to account
for the equilibrium of moments of couples. The equilibrium relations of a conventional system of forces
applied to a system of multiple material particles are derived from a resultant force and a resultant couple
of forces applied to an arbitrary point. The couple of forces is a free vector in the conventional mechanics,
which means that the effect of the couple applied on an arbitrary point in the space of the system of material
particles is independent of the position of the point. In other words, the couple can translate to any point in
space freely and the resulting motive effects are unchanged. As a result, only the conventional force
equilibrium and moment equilibrium are involved in the equilibrium relations as given in Eq. (8) (Toupin,
1962; Mindlin and Tiersten, 1962; Koiter, 1964; Mindlin, 1964). The equivalence of a couple that is not a
free vector but a driving force that rotates the material particles is shown in Fig. 2. The couple vector LA at
A in a system of material particles in Fig. 2(a) is equivalent to a couple L0A and a couple of couples M0A
applied to the point B in Fig. 2(c). This equivalence is derived through an intermediate step as shown in Fig.
2(b). In the intermediate step, two opposite couple vectors L0A and L00A of equal magnitude,
L0A ¼ LA ð18Þ
are placed at point B. The couple of couples M0A defined as
M0A ¼ ðxB  xA Þ  LA ð19Þ
is equivalent to the couple LA at point A and the couple L00A at point B. For a system of material particles,
the forces Fi and couples Li applied to the system are equivalent to a resultant force F0 , a resultant couple L0
and a resultant couple of couples M0 applied to the origin where F0 , L0 and M0 are given as

Fig. 2. Equivalence of couple: a couple at point A is equivalent to a couple at point B and a moment of the couple.
2736 F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743
X X X
F0 ¼ Fi ; L0 ¼ ðxi  Fi þ Li Þ; M0 ¼ xi  Li : ð20Þ

In static equilibrium, these equations are equal to zero giving the relations,
X X X
Fi ¼ 0; ðxi  Fi þ Li Þ ¼ 0; xi  Li ¼ 0; ð21Þ

which are the equilibrium relations for a system of discrete particles encased by representative volume el-
ement.

5. Governing equations in couple stress theory of deformable body

5.1. Equilibrium equations

The equilibrium relations in Eq. (21) are for a system of discrete particles. In this section, we shall
develop the equilibrium relations for a continuum. Consider an arbitrary volume v0 of a deformable body
bounded by piecewise smooth surfaces denoted by ov0 . We denote tn and ln respectively as the force and
couple vectors per unit area transmitted through the surface ov0 . The subscript n represents the direction of
the external normal n to the surface. Additionally, we use f and l to denote the body force and the body
couple per unit volume of the material particles respectively. For the volume of the continuum, the first two
equilibrium equations of Eq. (21) can be rewritten as
Z Z
f dv þ tn ds ¼ 0;
0 ov0
Zv Z ð22Þ
ðx  f þ lÞdv þ ðx  tn þ ln Þds ¼ 0;
v0 ov0

where x is the position vector of a material particle in the continuum. Generalizing the conventional
Cauchy’s principle, Koiter (1964) proposed
tn ¼ t  n; ð23Þ
and
ln ¼ l  n; ð24Þ
where t is the stress tensor and l is the couple stress tensor. Using the divergence theorem to transform the
surface integrals in Eq. (22) to volume integrals, we obtain
Z
ðt  r þ fÞdv ¼ 0;
0
Zv ð25Þ
½x  ðt  r þ fÞ 2: t þ l  r þ l dv ¼ 0:
v0

Since the volume v0 is arbitrary, the volume dependence can be eliminated which then leads to
t  r þ f ¼ 0;
ð26Þ
l  r þ l 2: t ¼ 0;
which are the equilibrium relations of deformable body and are equivalent to the first two equations of Eq.
(21). The second equilibrium relation in Eq. (26) indicates that the stress tensor t generates an equivalent
body couple  2: t acting together with l to maintain the equilibrium of the continuum. Since the couple
moment must vanishes, we have
F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743 2737
Z Z
x  ðl 2: tÞdv þ x  ln ds ¼ 0; ð27Þ
v0 ov0

where l 2: t is the residual body couple and ln is the surface couple traction. This is equivalent to the third
equation in Eq. (21). Using the divergence theorem, we can rewrite Eq. (27) as
Z
½x  ðl 2: t þ l  rÞ 2: l dv ¼ 0; ð28Þ
v0

which leads to the conclusion that


2: l ¼ 0; ð29Þ

i.e., the couple stress tensor is symmetric.


The expressions in Eq. (26) can be combined through decomposition of the stress tensor and the cou-
ple stress tensor. Let us denote the symmetric and antisymmetric parts of the stress tensor, respectively,
as
r ¼ 12ðt þ tT Þ; s ¼ 12ðt  tT Þ; ð30Þ
where tT is the transpose of t. Decomposition of the couple stress tensor l into the spherical part lg and the
deviatoric part m gives,
l ¼ lg þ m; l ¼ 13trðlÞ; trðmÞ ¼ 0: ð31Þ
The equilibrium relations in Eq. (26) can be rewritten as
ðr þ sÞ  r þ f ¼ 0;
ð32Þ
lr þ m  r þ l 2: s ¼ 0:
Eliminating the antisymmetric stress tensor s in Eq. (32) gives the final equilibrium equation
r  r þ 12 2: ðm  r r þ l rÞ þ f ¼ 0: ð33Þ
This is in the same form as that of the classical couple stress theory (Koiter, 1964). However, unlike the
classical couple stress theory, Eq. (33) requires that the deviatoric couple stress tensor m to be symmetric.

5.2. Generalized strains and principle of virtual work

In this section, the deformation measures, e and v, conjugate to r and m, are derived via the application
of the principle of virtual work. Consider a convex volume v0 of the deformable body bounded by a surface
ov0 that consists of a finite number of sub-surfaces whose outer normals form a continuous vector field. The
deformation energy density is assumed to be a function of the gradients of generalized motions, which are
the gradients of translation and rotation in the present revised couple stress model. The principle of virtual
work states that the change in the deformation energy equals the work done by the external generalized
loads through the virtual displacement du and the virtual rotation dh, i.e.,
Z Z Z
d wðu r; h rÞdv ¼ ðdu  f þ dh  lÞdv þ ðdu  tn þ dh  ln Þds; ð34Þ
v0 v0 ov0

where w is the deformation energy density per unit volume. Substituting Eqs. (23) and (24) into Eq. (34) and
applying the divergence theorem, we can rewrite the principle of virtual work as
Z Z
dw dv ¼ ½du  ðt  r þ fÞ þ du r : t þ dh  ðl  r þ lÞ þ dh r : l dv: ð35Þ
v0 v0
2738 F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743

Let us denote the symmetric parts of the displacement gradient and the rotation gradient, respectively, as
e ¼ 12ðu r þ r uÞ;
ð36Þ
v ¼ 12ðh r þ r hÞ;
where e is the strain tensor and v is the symmetric curvature tensor. We also define
c ¼ 12ðh r  r hÞ ð37Þ
as the antisymmetric curvature tensor, and define as
h ¼  2: 12u r: ð38Þ
Using Eqs. (35)–(38) and the equilibrium equations of Eq. (26), we derive the following relation,
Z Z
dw dv ¼ ðde : r þ dv : mÞdv; ð39Þ
v0 v0

which shows that e and v are conjugated to r and m, respectively. Note that m and v are both deviatoric
and symmetric. Since the volume v0 is arbitrary, Eq. (39) implies
dwðu r; h rÞ ¼ de : r þ dv : m: ð40Þ
The variation of w is
ow ow ow ow
dwðu r; h rÞ ¼ dwðe; h; v; cÞ ¼ de : þ dh  þ dv : þ dc : : ð41Þ
oe oh ov oc
Thus, we have
ow ow ow ow
r¼ ; m¼ ; ¼ 0; ¼ 0: ð42Þ
oe ov oh oc
The above equations show that the deformation energy density w does not depend explicitly on the rotation
h (the antisymmetric part of displacement gradient) and the antisymmetric curvature tensor c (the anti-
symmetric part of rotation gradient). That is to say, only the symmetric part of displacement gradient
(conventional strain tensor) and the symmetric part of rotation gradient (symmetric curvature tensor) con-
tribute to the deformation energy. The rotation vector and the antisymmetric curvature tensor do not con-
tribute to the deformation energy. The strain tensor and symmetric curvature tensor are the deformation
measures conjugate to the symmetric stress tensor r and the deviatoric couple stress tensor m, respectively.
This conclusion will have important implications for the number of independent higher order material
parameters.

6. Linear isotropic elasticity

A linear elastic constitutive law within the present modified couple stress framework is developed in this
section. A linear constitutive law for isotropic elastic couple stress materials can be derived from the de-
formation energy density as given in Eq. (42). For the linear isotropic materials, w is a quadratic function of
the invariants of generalized strains. It can be written as
2
w ¼ 12kðtr eÞ þ lðe : e þ l2 v : vÞ; ð43Þ
where k and l are the Lame’s constants, and l is a length scale parameter. Substituting Eq. (43) into Eq.
(42), we obtain the linear constitutive relations,
F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743 2739

r ¼ kg trðeÞ þ 2le;
ð44Þ
l1 m ¼ 2llv:
In this model, only one length scale parameter is needed in addition to the conventional Lame constants.
The single material length scale parameter dependence is the direct consequence of the conclusion of Section
5.2 that w is a function of the strain and the symmetric curvature tensors only, and does not depend explicitly
on the rotation (the antisymmetric part of the deformation gradient) and the antisymmetric part of the cur-
vature tensor. This is a particularly useful contribution of the study as this conclusion reduced the exper-
imental problem of finding two independent higher order material length scale parameters to just one.

7. Examples

In this section, two simple examples, torsion of a thin cylindrical bar and pure bending of a plate of
infinite width, are derived to illustrate the effects of strain gradients. The governing equation is Eq. (33) and
the general boundary conditions are detailed in Appendix A.

7.1. Torsion of a thin cylindrical bar

The x-axis of the Cartesian coordinate system is taken along the axis of the cylindrical bar. The dis-
placement field in the cylindrical bar is assumed to be in the same form as in the conventional mechanics,
ux ¼ 0; uy ¼ jxz; uz ¼ jxy; ð45Þ
where j is the twist per unit length of the cylindrical bar. From the displacement field, the components of
the strain tensor, the rotation vector and the symmetric curvature tensor are
exx ¼ eyy ¼ ezz ¼ eyz ¼ 0; exy ¼ 12jz; exz ¼ 12jy;
hx ¼ jx; hy ¼ 12jy; hz ¼ 12jz; ð46Þ
1
vxx ¼ j; vyy ¼ vzz ¼ 2j; vxy ¼ vyz ¼ vzx ¼ 0:
From the constitutive relation in Eq. (44), the components of the stress tensor and the couple stress tensor are
rxx ¼ ryy ¼ rzz ¼ ryz ¼ 0; rxy ¼ ljz; rxz ¼ ljy;
2 2
ð47Þ
mxx ¼ 2ll j; myy ¼ mzz ¼ ll j:
One can easily verify that the stresses and couple stresses satisfy the equilibrium equation (33) and the
traction free boundary conditions in Eq. (A.3) on the outer cylindrical surface. The resultant torque at the
ends of the cylindrical bar can be determined through the virtual work principle,
"  2 #
l
Q ¼ Q0 1 þ 6 ; ð48Þ
a

where a is the radius of the cylindrical bar and Q0 is the torque per unit length of the cylindrical bar
calculated from conventional theory,
Q0 ¼ 12plja4 : ð49Þ
The last term in the bracket of Eq. (48) shows the effect of rotation gradient and cylinder radius on the
torsional rigidity of the cylindrical bar. The significant size effect on the torsional rigidity is shown in Fig. 3,
where the horizontal axis is the ratio of the cylindrical radius to the material length scale parameter, a=l,
2740 F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743

Fig. 3. Effect of cylindrical radius on the torsional rigidity of the cylindrical bar. Q=Q0 is the normalized torsional rigidity, a is radius of
the cylindrical bar and l is the length scale parameter.

and the vertical axis is the normalized torque, Q=Q0 . When the cylindrical radius is close to the material
length scale parameter, the rigidity is about seven times of the magnitude of conventional result. For
torsion of the copper wires, l was determined to be 3 lm (Chong et al., 2001). Elastic deformation in copper
structures in the micron-scale is strongly affected by strain gradients.

7.2. Bending of a plate of infinite width

We take the x–y plane of the Cartesian coordinate system as the midplane of the plate. The y-axis is
parallel to the direction of the width. The displacement field is assumed in the same form as in conventional
mechanics, i.e.,
1 m 1
ux ¼ jxz; uy ¼ 0; uz ¼  jz2  jx2 ; ð50Þ
2 1m 2
where j is the curvature of the bended midplane of the plate, and m is Poisson’s ratio. This displacement
field satisfies the boundary conditions
uy ¼ 0; hx ¼ hz ¼ 0 at y ¼ 1: ð51Þ
The corresponding strain tensor, rotation vector and symmetric curvature tensor are
m
exx ¼ jz; eyy ¼ 0; ezz ¼  jz; exy ¼ eyz ¼ ezx ¼ 0;
1m
hx ¼ hz ¼ 0; hy ¼ jx; ð52Þ
vxx ¼ vyy ¼ vzz ¼ vyz ¼ vzx ¼ 0; vxy ¼ 12j:
From Eq. (44), the components of the stress tensor and couple stress tensor are obtained as follows:
2l 2ml
rxx ¼ jz; ryy ¼ jz; rzz ¼ rxy ¼ ryz ¼ rzx ¼ 0;
1m 1m ð53Þ
mxx ¼ myy ¼ mzz ¼ myz ¼ mzx ¼ 0; mxy ¼ ll2 j:
It is easy to verify that the stresses and couple stresses satisfy the equilibrium equation in Eq. (33) and the
traction-free boundary conditions in Eq. (A.3) at z ¼ h=2 surfaces with h being the plate thickness. The
tractions and the couple tractions on the y ¼ 1 surfaces are respectively,
F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743 2741

Fig. 4. Effect of plate thickness on the bending rigidity of the plate for m ¼ 0:3. M=M0 is the normalized bending rigidity, h is the
thickness of the plate and l is the length scale parameter.

2ml
px ¼ pz ¼ 0; py ¼  jz;
1m ð54Þ
qx ¼ ll2 j; qz ¼ 0;
and the tractions and the couple tractions on the x ¼ a=2 surfaces are
2l
px ¼  jz; py ¼ pz ¼ 0;
1m ð55Þ
qy ¼ ll2 j; qz ¼ 0:
The equivalent resultant moment per unit width is thus
"  2 #
l
M ¼ M0 1 þ 6ð1  mÞ ; ð56Þ
h

where M0 is the corresponding resultant moment per unit width calculated from the classical theory
lh3
M0 ¼ j: ð57Þ
6ð1  mÞ
The last term in the bracket of Eq. (56) gives the contribution of rotation gradients to the bending ri-
gidity and shows the effect of plate thickness on the bending rigidity. The thickness effect on the bending
rigidity is shown in Fig. 4, where the horizontal axis is the ratio of plate thickness to material length scale
parameter, h=l, and the vertical axis is the normalized moment, M=M0 . When the plate thickness is close to
the material length scale parameter for m ¼ 0:3, this bending rigidity is about five times the conventional
rigidity. The plate bending behavior is similarly affected by strain gradients as that for torsion of copper
wires above-mentioned.

8. Discussion

In classical couple stress theories (Toupin, 1962; Mindlin and Tiersten, 1962; Koiter, 1964; Mindlin,
1964), the equilibrium of forces and moments are the same as those in conventional mechanics. The de-
formation measures conjugate to the symmetric stress tensor r and the couple stress tensor l are the strain
2742 F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743

tensor e and the curvature tensor h r, which includes the symmetric part as well as the antisymmetric
part of the curvature tensor. For linear elastic materials, there are two independent length scale parameters
associated with the symmetric and antisymmetric parts of the curvature tensor, respectively. The two pa-
rameters cannot be determined from a single test such as the twisting of a thin cylindrical bar or pure
bending of a thin film. A combination of twisting and bending tests is necessary to determine the two
parameters.
In the present theory, an equilibrium relation for moments of couples is introduced to remedy the free-
floating nature of the couple vector. The additional equilibrium relation restricts the couple stress tensor
to be symmetric. As a result, the deformation energy becomes independent of the antisymmetric part of
the curvature tensor. Consequently for linear isotropic elastic materials, there is only one independent
length scale parameter in the couple stress theory, and it is associated with the symmetric curvature ten-
sor. The single length scale parameter can be determined by twisting of slim cylinders of different dia-
meter. The implications of the additional equilibrium relation for couples in the higher order strain
gradient elasticity have been investigated and are reported in another paper (Yang et al., submitted for
publication).

9. Conclusion

In this paper, equilibrium of the moment of couples is introduced as an additional equation for the
couple stresses. The additional equilibrium relation requires the couple stress tensor to be symmetric. As a
result, the only deformation measures, which contribute to the deformation energy, are the symmetric parts
of the displacement gradient (strain tensor) and the rotation gradient (symmetric curvature tensor). A linear
elastic constitutive law for isotropic couple stress materials is developed on this basis. The number of
material length scale parameters is reduced from the two in the classical couple stress theories to only one in
the present theory. The torsion of a thin cylindrical bar and the bending of a thin plate of infinite width are
presented to illustrate the effect of the strain gradients in the modified couple stress theory.

Acknowledgements

This work has been supported by the Research Grants Council of the Hong Kong Special Adminis-
trative Region, the People’s Republic of China. Fan Yang also acknowledges the support of the fund of
Southwest Jiaotong University of PR-China.

Appendix A. Boundary conditions

The boundary conditions associated with the above modified couple stress theory are examined in this
section. Koiter (1964) derived the boundary conditions of the classical couple stress theory using the
principle of virtual work for the whole body. He concluded that the normal component of the rotation
vector on the boundary surface could not be prescribed independently. In other words, only the three
components of the displacement vector and the two tangential components of the rotation vector can be
specified as the generalized displacement conditions on a boundary surface. Thus, the total number of
generalized displacement conditions is five. The resulting boundary conditions are summarized below:

u¼
u;
ðA:1Þ
h  ðh  nÞn ¼ 
ht
F. Yang et al. / International Journal of Solids and Structures 39 (2002) 2731–2743 2743

on the smooth part of ovi ; and

uk ;
uk¼ ðA:2Þ

along the intersection Ci of two smooth parts of the boundary surface, where the overbar denotes a pre-
scribed quantity, 
ht is the prescribed rotation vector tangent to the boundary surface, and uk is the dis-
placement component along the direction k of the intersection Ci of two smooth parts of the boundary
surface. The corresponding generalized traction conditions are

r  n þ 12n  ½m  r  ðn  m  nÞr þ l ¼ r
n  12n  ½ð
ln  nÞr ;
ðA:3Þ
m  n  ðn  m  nÞn ¼ ln  ð
ln  nÞn

on the smooth part of the boundary surface ovi and


1
2
½ðn  m  nÞþ  ðn  m  nÞ ¼ 12½ð
ln  nÞþ  ð
ln  nÞ ; ðA:4Þ

along the intersection Ci of two smooth parts of the boundary surface, where r n and l n are the force-
traction and couple-traction on the boundary surface, respectively, and the subscripts ‘‘þ’’ and ‘‘’’ in-
dicate the two opposite sides of the intersection.

References

Chong, A.C.M., Lam, D.C.C., 1999. Strain gradient plasticity effect in indentation hardness of polymers. J. Mater. Res. 14, 4103–4110.
Chong, A.C.M., Yang, F., Lam, D.C.C., Tong, P., 2001. Torsion and bending of micron-scaled structures. J. Mater. Res. 14, 1052–
1058.
Eringen, A.C., 1971. Continuum Physics. Academic Press, New York.
Eringen, A.C., 1998. Microcontinuum Field Theories. Springer-Verlag, New York.
Fleck, N.A., Hutchinson, J.W., 1997. Strain gradient plasticity. In: Hutchinson, J.W., Wu, T.Y. (Eds.), Advances in Applied
Mechanics, vol. 33. Academic Press, New York, pp. 295–361.
Fleck, N.A., Muller, G.M., Ashby, M.F., Hutchinson, J.W., 1994. Strain gradient plasticity: theory and experiments. Acta Metall.
Mater. 42, 475–487.
Koiter, W.T., 1964. Couple stresses in the theory of elasticity, I and II. Proc. K. Ned. Akad. Wet. (B) 67, 17–44.
Lam, D.C.C., Chong, A.C.M., 1999. Indentation model and strain gradient plasticity law for glassy polymers. J. Mater. Res. 14, 3784–
3788.
Ma, Q., Clarke, D.R., 1995. Size dependent hardness of silver single crystals. J. Mater. Res. 10, 853–863.
Martin, J.B., Thornton, S.T., 1995. Classic dynamics of particles and systems, fourth edition. Saunders College Publishing,
Philadelphia.
Mindlin, R.D., 1964. Micro-structure in linear elasticity. Arch. Ration. Mech. Anal. 16, 51–78.
Mindlin, R.D., Tiersten, H.F., 1962. Effects of couple-stresses in linear elasticity. Arch. Ration. Mech. Anal. 11, 415–448.
Nix, W.D., 1989. Mechanical properties of thin films. Metall. Trans. A 20, 2217–2245.
Poole, W.J., Ashby, M.F., Fleck, N.A., 1996. Micro-hardness of annealed and work-hardened copper polycrystals. Scripta Metall.
Mater. 34 (4), 559–564.
Stelmashenko, N.A., Walls, M.G., Brown, L.M., Milman, Y.V., 1993. Microindentations on W and Mo oriented single crystals: an
STM study. Acta Metall. Mater. 41 (10), 2855–2865.
St€
olken, J.S., Evans, A.G., 1998. A microbend test method for measuring the plasticity length scale. Acta Metall. Mater. 46, 5109–
5115.
Toupin, R.A., 1962. Elastic materials with couple stresses. Arch. Ration. Mech. Anal. 11, 385–414.
Yang, F., Chong, A.C.M., Lam, D.C.C., Tong, P., submitted for publication. Strain gradient theory for linear elasticity.

You might also like