Sei sulla pagina 1di 32

Journal of the Mechanics and Physics of Solids

51 (2003) 1477 – 1508


www.elsevier.com/locate/jmps

Experiments and theory in strain gradient


elasticity
D.C.C. Lama;∗ , F. Yanga; b , A.C.M. Chonga , J. Wanga; b , P. Tonga
a Departmentof Mechanical Engineering, The Hong Kong University of Science and Technology,
Clear Water Bay, Kowloon, Hong Kong
b Institute of Computational Engineering and Science, Southwest Jiaotong University, Chengdu 610031,

Sichuan, People’s Republic of China

Received 13 May 2002; accepted 10 March 2003

Abstract
Conventional strain-based mechanics theory does not account for contributions from strain
gradients. Failure to include strain gradient contributions can lead to underestimates of stresses
and size-dependent behaviors in small-scale structures. In this paper, a new set of higher-order
metrics is developed to characterize strain gradient behaviors. This set enables the application
of the higher-order equilibrium conditions to strain gradient elasticity theory and reduces the
number of independent elastic length scale parameters from 9ve to three. On the basis of this
new strain gradient theory, a strain gradient elastic bending theory for plane-strain beams is
developed. Solutions for cantilever bending with a moment and line force applied at the free
end are constructed based on the new higher-order bending theory. In classical bending theory,
the normalized bending rigidity is independent of the length and thickness of the beam. In the
solutions developed from the higher-order bending theory, the normalized higher-order bending
rigidity has a new dependence on the thickness of the beam and on a higher-order bending
parameter, bh . To determine the signi9cance of the size dependence, we fabricated micron-sized
beams and conducted bending tests using a nanoindenter. We found that the normalized beam
rigidity exhibited an inverse squared dependence on the beam’s thickness as predicted by the
strain gradient elastic bending theory, and that the higher-order bending parameter, bh , is on
the micron-scale. Potential errors from the experiments, model and fabrication were estimated
and determined to be small relative to the observed increase in beam’s bending rigidity. The
present results indicate that the elastic strain gradient e=ect is signi9cant in elastic deformation
of small-scale structures.
? 2003 Elsevier Science Ltd. All rights reserved.

Keywords: A. Strain-gradient e=ects; B. Constitutive behavior; Elastic material; Beams; C. Mechanical testing

∗ Corresponding author. Tel.: +852-2358-8661; fax: +852-2358-1543.


E-mail address: medcclam@ust.hk (D.C.C. Lam).

0022-5096/03/$ - see front matter ? 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0022-5096(03)00053-X
1478 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

1. Introduction

1.1. Elastic deformation in small-scale structures

Structures in engineering are shifting from the micron scale to the nanometer scale
(Albrecht et al., 1990). Micrometer-scale thin 9lms are widely used in engineering
for their electrical, optical, hardness and corrosion-resistance properties. Thin 9lms
are in electronic devices, on the lenses of glasses and in cars. These 9lms are ap-
plied onto structures primarily for non-structural functions and applications. On the
other hand, microelectromechanical systems (MEMS) and nanoelectromechanical sys-
tems (NEMS) are created to probe surfaces, study cells and neurons, move microliters
of Guids and divert photons in optoelectronics (Bashir et al., 2000; Carr and Craighead,
1997; Craighead, 2000; Manias et al., 2001). These structures are microns and nanome-
ters in size (collectively called small-scale structures in this paper). They are in the
form of beams and plates, carry load and deform elastically. Conventional strain-based
elasticity has been used to predict their deformation behaviors, but it is unclear whether
conventional elasticity remains valid in small-scale structures.

1.2. Classical and higher-order elasticity

The linear elastic behavior of materials is founded on Hooke’s Law, which relates
the applied force, F, in a uniaxial tensile test of a solid bar linearly to the change in
the axial displacement, Iu, via a spring constant ks :
F = ks Iu: (1)
This observed linear elastic behavior is the physical foundation of classical elastic-
ity. Instead of force and displacement, stress, ij , and strain, ij , are used as pri-
mary measures in classical elasticity and related linearly via the elastic modulus, E,
giving
 
E

ij = ij + kk ij (2)


1+
1 − 2

for isotropic materials, where


is Poisson’s ratio. (In the equation above and in subse-
quent equations, the index notation will be used with repeated indices denoting summa-
tion from 1 to 3.) In contrast to ks ; E and
are material constants and are independent
of the geometry of the test specimen.
The description of the deformation behavior of solids is not limited to relations
between elastic stresses and strains. Higher-order theories of plasticity, which include
contributions from strain gradients, have been developed by Fleck and Hutchinson
(1997, 2001) and Gao et al. (1999), and experimentally investigated by Nix (1989),
Stelmashenko et al. (1993), Fleck et al. (1994), Ma and Clarke (1995), Poole
et al. (1996), StLolken and Evans (1998), Chong and Lam (1999) and Lam and Chong
(1999). Toupin (1962), Mindlin and Tiersten (1962), Koiter (1964) and Mindlin (1964,
1965) have developed higher-order theories of elasticity. Higher-order theories can be
classi9ed into general strain gradient theories and couple stress theories according to
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1479

the deformation metrics used. In the couple stress theories (Toupin, 1962; Mindlin and
Tiersten, 1962; Koiter, 1964; Mindlin, 1964), the higher-order rotation gradients are
included as the deformation metrics. The rotation gradients are the anti-symmetric part
of the second order deformation gradients and have eight independent components. The
symmetric part of the second-order deformation gradient is generally neglected in the
couple stress theories.
Yang et al. (2002) modi9ed the classical couple stress theories by introducing the
concept of the representative volume element. In addition to the classical equilibrium
equations of forces and moments of forces, they developed a new additional equilibrium
equation to govern the behavior of higher-order stresses, the equilibrium of moments
of couples. They concluded that the couple stress tensor is symmetric and that the
anti-symmetric part of the rotation gradient tensor does not contribute to deformation
energy. Mindlin (1965) developed a general higher-order stress theory, which includes
higher-order strain gradients. In his simpli9ed version, only the second-order defor-
mation gradients (9rst-order strain gradients) are included as additional deformation
metrics. The second-order deformation gradient (18 independent components) has an
anti-symmetric part (8 independent components) and a symmetric part (10 independent
components). The governing equations and boundary conditions are derived from the
principle of virtual work. From his work, Mindlin concluded that there are 9ve linear
elastic parameters corresponding to the second-order deformation gradient for isotropic
center-symmetric materials. Fleck and Hutchinson (1997, 2001) reformulated Mindlin’s
simpli9ed theory and renamed it the strain gradient theory. In the strain gradient theory,
the second-order deformation gradient tensor is decomposed into two independent parts,
the stretch gradient tensor and the rotation gradient tensor. In this formulation as in
prior ones, conventional equilibrium relations are used and higher equilibrium condi-
tions governing the behavior of higher-order stresses are ignored.
In this paper, a stratifying decomposition scheme of the second-order deformation
gradient tensors is developed to enable the application of new higher-order equilib-
rium relation (Yang et al., 2002). The higher-order stress tensor work-conjugate to
the new higher-order deformation metrics and the corresponding constitutive relations
are de9ned. Because of the requirement of the higher-order equilibrium condition, the
anti-symmetric part of the rotation gradient does not contribute to the deformation
energy and the number of material length scale parameters is reduced from 9ve to
three. On the basis of this modi9ed strain gradient elasticity theory, a linear elas-
tic bending theory for plane-strain beams is developed in Section 3 of this paper.
A power series expansion in terms of the beam thickness is used to simplify the
three-dimensional di=erential equations and boundary conditions. Solutions for can-
tilever bending and other bending modes are developed. The results indicate that the
normalized bending sti=ness, which is independent of beam thickness in strain-based
elasticity, has an additional inverse squared dependence on the beam thickness. In Sec-
tion 4 of the paper, deformation experiments on beam bending to determine the presence
and signi9cance of the predicted strain gradient elastic size e=ect are described. The
errors associated with the bending theory and the experiments are analyzed and dis-
cussed. Finally, the major conclusions and developments of strain gradient elasticity are
summarized.
1480 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

2. Strain gradient elasticity

2.1. A summary of classical strain gradient theory for linear elasticity

In this section, the classical strain gradient theory is summarized. Mindlin devel-
oped a higher-order elastic theory in 1960s. Fleck and Hutchinson modi9ed it to a
second-order version in 1997. The major di=erence between the strain gradient theory
and the conventional elastic theory is that the strain energy density, w, depends on both
the conventional strain (the symmetric part of the 9rst-order deformation gradient) and
on the second-order deformation gradient,
w = w(ij ; ijk ); (3)
where ij and ijk are the strain tensor and second-order deformation gradient tensor,
respectively. In Eq. (3), the strain tensor and second-order deformation gradient tensor
are de9ned, respectively, as
ij = 12 (9i uj + 9j ui ); ijk = 9ij uk ; (4)
where 9i is the forward gradient operator and ui is the displacement vector. The strain
tensor has six independent symmetric components and the second-order deformation
gradient tensor has 18 independent components that are symmetric in the 9rst two
indices.
In the derivations of Mindlin (1965) and Fleck and Hutchinson (1997), the Cauchy
stress tensor, ij , and double stress tensor, ijk , are conjugated with the strain tensor
and the second-order deformation gradient tensor via
9w 9w
ij = ; ijk = : (5)
9ij 9 ijk
The variation of the total strain energy in the volume V of the body can be written as
 
w dV = (ij ij + ijk ijk ) dV: (6)
V V
Applying Gauss’s divergence theorem to Eq. (6) yields the 9rst variation of the total
strain energy as
 
w dV = − (9i ik − 9ij ijk ) uk dV
V V

+ [nj (jk − 9i ijk ) uk + ni ijk 9j uk ] dS; (7)
9V
where ni is a unit vector normal to the boundary surface, 9V , of the volume, V .
The gradient of the displacement variation on the boundary surface in Eq. (7) can be
decomposed into a surface gradient and a normal gradient,
9j uk = Dj uk + nj D uk ; (8)
where the surface gradient operator, Dj , and normal gradient operator, D, are de9ned,
respectively, as
Dj = ( jk − nj nk )9k ; D = n k 9k : (9)
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1481

The boundary surface is assumed to be divisible into a 9nite number of smooth parts,
Sn , each bounded by an edge, Cn . Using Stokes’ surface divergence theorem on the
smooth surface Sn , we have
  
Dj (ni ijk uk ) dS = ni kj ijk uk ds + (Dp np )ni nj ijk uk dS: (10)
Sn Cn Sn

The surface integral in Eq. (7) then becomes



{[nj (jk − 9i ijk ) + ni nj ijk (Dl nl ) − Dj (ni ijk )] uk + ni nj ijk D uk } dS
9V

+ (ni kj ijk ) uk ds; (11)
m Cm

where kj is the surface unit vector normal to the edge Cm ; m is the number of sharp
edges, and  represents the di=erence between the bracketed terms on the two sides
of the edge. The 9nal form of the principle of virtual work is thus
  
(ij ij + ijk ijk ) dV = fk uk dV + (tk uk + rk D uk ) dS
V V 9V

+ pk uk ds; (12)
m Cm

where the body force per unit volume is fk , the surface traction on surface S is tk ,
the surface double-force traction is rk , and the line load along any sharp edges Cm , is
pk . Thus, the equilibrium equation in the body is
9i ik − 9ij ijk + fk = 0 in V (13)
and the boundary conditions on S and along Cm are, respectively,
nj (jk − 9i ijk ) + ni nj ijk (Dl nl ) − Di (nj ijk ) = tQk or uk = uQ k

ni nj ijk = rQk or Duk = Duk (14)


and
(ni kj ijk ) = pQ k or uk = uQ k : (15)
For linear elastic isotropic materials, Mindlin (1965) de9ned the density of strain
energy as
w = 12 ii jj + ij ij + a1 ijj ikk + a2 iik kjj + a3 iik jjk
+ a4 ijk ijk + a5 ijk kji ; (16)
where  and  are the conventional LamRe constants, corresponding to the two invariants
of strain, and an (n = 1; 2; : : : ; 5) are the 9ve additional second-order elastic constants
corresponding to the invariants of the second deformation gradients. The energy den-
sity relation contains cross terms for both stretch and rotation gradients in the strain
energy density. These cross terms are also present in the strain gradient theories by
1482 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

Fleck and Hutchinson (1997). Their presence disallows the direct application of the
higher-order equilibrium conditions.

2.2. Development of independent metrics, and equilibrium and boundary conditions

In the section, we will establish a new set of strati9ed metrics for use in strain gradi-
ent elasticity that will enable the application of the higher-order equilibrium conditions
to the higher-order theory. Fleck and Hutchinson (1997) decomposed the second-
order deformation gradient, ijk , into symmetric and anti-symmetric parts, sijk , and aijk ,
giving,
sijk = 13 ( ijk + jki + kij ); aijk = 23 (eikl "lj + ejkl "li ); (17)
where eijk is the alternating tensor and "ij = 1=2eipq jpq is the curvature tensor. New
independent second strain metrics are obtained by splitting the symmetric second-order
deformation gradient, sijk into a trace part, (0) (1)
ijk , and a traceless part, ijk , as

sijk = (0) (1)


ijk + ijk ; (18)
where
(0) 1 s s s
ijk = 5 ( ij mmk + jk mmi + ki mmj );

(1) s (0)
ijk = ijk − ijk ;

smmk = 13 ( mmk + 2 kmm ): (19)


The curvature tensor is decomposed into symmetric and anti-symmetric parts as
"ij = "ijs + "ija ; (20)
where
"ijs = 12 ("ij + "ji ); "ija = 12 ("ij − "ji ): (21)
The number of independent components in the trace part of the symmetric second-order
deformation gradient (0)
ijk is three; in the traceless part of the symmetric second-order
deformation gradient (1)
ijk , the number is seven; and in the symmetric and anti-symmetric
curvature tensors, "ijs and "ija , the numbers are 9ve and three, respectively. The trace
part of the symmetric second-order deformation gradient is a function of the dilatation
gradient and the anti-symmetric part of the curvature,
sipp = ; i + 23 eimn "mn
a
= ; i + 23 eimn "mn ; (22)
where  is the dilatation strain,
 = mm : (23)
For easy reference, we name ; i the dilatation gradient, (1)
the deviatoric stretch
ijk
gradient and "ij the rotational gradient. The second-order virtual work density written
in terms of the new strain metrics, (0) (1) a
ijk ; ijk ; ijk , is

ŵ = (0) (0) (1) (1) a a


ijk ijk + ijk ijk + ijk ijk ; (24)
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1483

where

(0) 1 s s s
ijk = 5 ( ij mmk + jk mmi + ki mmj );

(1) s (0)
ijk = ijk − ijk ; (25)

are the trace and traceless parts of the symmetric part of the double stress tensor,
respectively. The tensors (0) (1)
ijk and ijk are orthogonal to each other. The set of stress
metrics, (0) (1) a (0) (1) a
ijk ; ijk ; ijk , is work-conjugate to the set of strain metrics, ijk ; ijk ; ijk .
Using ;i ; (1)
ijk ; "ij as the second-order strain metrics, we obtain

ŵ = pi ;i +(1) (1) 


ijk ijk + mij "ij ; (26)

where

pi = 35 smmi ;

mij = 43 aipq ejpq − 25 eijk smmk : (27)

Thus, pi , (1)  (1)


ijk and mij are work-conjugates to ;i ; ijk ; "ij , respectively. Correspond-
ingly, the inverse forms of Eq. (27) are

aijk = 14 (eikp mjp + ejkp mip ) − 16 (2 ij pk − ik pj − jk pi );

(0) 1
ijk = 3 ( ij pk + jk pi + ki pj ): (28)

Finally, application of the principle of virtual work gives the equilibrium equation in
V as

9i ik − 12 ejlk 9il mij − 9ik pi − 9ij (1)


ijk + fk = 0 (29)

and the boundary conditions on S as

nj (jk − 12 ejkl 9i mil − jk 9i pi − 9i (1)


ijk )

+ (Dl nl )(np pp nk + ni nj (1) (1)


ijk + np nq nr pqr nk )

− 12 ejlk Dl (np nq mpq nj ) − Dk (pp np ) − Dj (ni (1) (1)


Q
ijk ) − Dl (ni nj ijl nk ) = t k

or uk = uQ k ; (301 )

ni mij − (np mpq nq )nj + 2ejlk nl np nq (1)


pqk = qQj

or ( ij − ni nj )%i = %Qj ; (302 )

ni pi + ni nj nk (1)
ijk = r;
Q

or n = ni nj ij = Qn (303 )


1484 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

and along Cm as
 
I 12 (np mpq nq )ejlk nj kl + kk (pi ni ) + kj ni (1) (1)
ijk + ni nj kl ijl nk = p
Qk

or uk = uQ k : (31)

New equilibrium and boundary conditions in terms of the new independent strati9ed
metrics are now established.

2.3. Higher-order equilibrium constraint

The couple stress theory is a special case of the higher-order stress theory in which
the e=ects of the dilatation gradient and the deviatoric stretch gradient are assumed to
be negligible. The density of the total internal virtual work simply becomes

w = ij ij + mji "ij : (32)

Thus, pi and (1)


ijk become

9w(ij ; ; i ; (1)
ijk ; "ij ) 9w(ij ; ; i ; (1)
ijk ; "ij )
pi = = 0; (1)
ijk = = 0: (33)
9; i 9 (1)
ijk

The equilibrium equations in Eq. (29) and the boundary conditions in Eqs. (30) and
(31) then become,

9i ik − 12 ejlk 9il mij + fk = 0 in V (34)

nj (jk − 12 ejkl 9i mil ) − 12 ejlk Dl (np nq mpq nj ) = tQk


or uk = uQ k on S

ni mij − (np mpq nq )nj = qQk

or ( ij − ni nj )%i = %Qj (35)

and
1
I 2 (np mpq nq )ejlk nj kl = pQ k

or uk = uQ k along Cm : (36)

In addition to the classical equilibrium equations of forces and moments of forces,


 
fQi dv + tij nj ds = 0;
v 9v
 
(∈ijk xj fQk + lQi ) dv + (∈ijk xj tkl + mil )nl ds = 0; (37)
v 9v
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1485

the equilibrium of moments of couples


 
Q
∈ijk xj (lk − ∈klm tlm ) dv + ∈ijk xj mkl nl ds = 0 (38)
v 9v

must also be satis9ed (Yang et al., 2002). Satisfaction of the higher-order equilibrium
equation requires that the couple stress tensor,
mij = mji ; (39)
be symmetric. As a result, the anti-symmetric part of the rotation gradient tensor does
not contribute to the deformation energy. In Eqs. (37) and (38), xi is a position vector,
ni is a normal vector, fQi and lQi are the body force and the body couple, respectively,
tij and mij are the Cauchy stress and the couple stress, respectively, and ∈ijk is the
alternating tensor. Since the deformation energy is independent of the anti-symmetric
part of the rotation gradient tensor,
9w(ij ; ; i ; (1) s a
ijk ; "ij ; "ij )
maij = =0 (40)
9"ija
maij must be zero. Note that the constraint in Eq. (40) cannot be obtained naturally from
Fleck and Hutchinson’s de9nition of the strain gradient (1997) because the symmetric
part of the strain gradient, sijk , involves the anti-symmetric part of the rotational gra-
dient, "ija , as shown in Eq. (21). Therefore, one cannot simply set (0) (1)
ijk and ijk to zero
to reduce the strain gradient theory to the couple stress theory. In contrast, the new
strati9ed metrics de9ned in this paper allow a natural reduction of the present strain
gradient theory into the couple stress theory (Yang et al., 2002).

2.4. Linear elastic constitutive relations

Application of the higher-order equilibrium constraint Eq. (38) to the strain gradient
theory leads to the constraint in Eq. (40). The total deformation energy density becomes
a function of the symmetric strain tensor, the dilatation gradient vector, the deviatoric
stretch gradient tensor and the symmetric rotation gradient tensor, but it is independent
of the anti-symmetric rotation gradient tensor,
w = w(ij ; ; i ; (1) s
ijk ; "ij ): (41)
For linear elastic center-symmetric isotropic materials, the constitutive function is a
quadratic function of the invariant strain metrics,
w = 12 kii jj + ij ij + a0 mm; i nn; i + a1 (1) (1)  s s
ijk ijk + a2 "ij "ij ; (42)
where ij is deviatoric strain,
ij = ij − 13 mm ij ; (43)
where k and  are the bulk and the shear moduli, respectively, and (n = 0; 1; 2)an
are the additional independent material parameters associated with dilatation gradients,
deviatoric stretch gradients and rotation gradients, respectively.
1486 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

In Eq. (42), the di=erent second-order strain metrics are orthogonal to each other
and there is no coupling amongst them. The second and the last two terms in Eq. (42)
are independent of the dilatational deformation. By rewriting the constants as

a0 = l20 ; a1 = l21 ; a2 = l22 ; (44)

where ln (n=0; 1; 2) are three material length scale parameters, we 9nd the constitutive
relations as

ij = k ij mm + 2ij ;

pi = 2l20 mm; i ; (1) 2 (1)


ijk = 2l1 ijk ; msij = 2l22 "ijs : (45)

Instead of 9ve independent higher-order length scale parameters for the second strains
in Mindlin’s strain gradient theory, there are only three in the present theory. The
reduction is a signi9cant step in enabling experimental delineation of strain gradient
behavior since only bending and torsion are required to completely characterize the
elastic strain gradient of an approximately incompressible material.

3. Bending theory for plane-strain beams

3.1. Approach

In this derivation, we set the mid-plane of the thin plane-strain beam (an in9nitely
wide plate) as the x–y (x1 –x2 ) plane of a Cartesian coordinate system with the width
direction along the y-axis. By de9nition, the thickness of the thin beam is much smaller
than its length and width. The displacement along the width direction is ignored and
the beam deformation is independent from the width coordinate, y. Also, the body
force in the beam is ignored, while only the normal tractions are applied to the top or
bottom surfaces of the beam. To facilitate the derivation, the bending stress resultants
are 9rst de9ned in terms of the stress measures in the strain gradient theory. Then, the
bending equilibrium equations and force-prescribed boundary conditions are derived.
Afterwards, a new active set of bending deformation measures is proposed by using
Taylor expansions of displacements with respect to the beam thickness to establish the
bending constitutive relations.

3.2. Equilibrium relations

In this section, the equilibrium relations and force-prescribed boundary conditions for
a plane-strain beam are developed. The focus is on the displacement along the length
and thickness directions of the beam. The displacement along the width is ignored and
the beam deformation is independent of the coordinate, y. In the strain gradient theory,
the deformation measures, i.e., the strain tensor, ij , the dilatation gradient vector, )i ,
the deviatoric stretch gradient tensor, (1)
ijk , and the symmetric rotation gradient tensor,
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1487

"ij from the previous section are summarized in Eq. (46).

ij = 12 (9i uj + 9j ui );

)i = 9i mm ;

(1) 1
ijk = 3 (9i jk + 9j ki + 9k ij )

1
− 15 [ ij (9k mm + 29m mk ) + jk (9i mm + 29m mi )
+ ki (9j mm + 29m mj )];

"ij = 12 (∈ipq 9p qj + ∈jpq 9p qi ): (46)

The stress measure is the classical stress tensor, ij , which is the work-conjugate to
ij . The higher-order stresses, pi , (1) (1)
ijk and mij , are the work-conjugates to )i , ijk and
"ij , respectively. The higher-order deformation and stress measures are categorically
ignored in conventional elasticity. The equilibrium equations and boundary conditions
for the beam are derived from the general equations. The equilibrium equations along
the directions of length and thickness of the beam are

91 (11 − 91p1 − 91 (1)


111 )

+93 (13 − 12 91 m12 − 91p3 − 291 (1) 1 (1)


113 − 2 93 m32 − 93 133 ) = 0;

93 (33 + 12 91 m32 − 91p1 − 291 (1) 1 (1)


133 − 2 93 p3 − 93 333 )

+ 91 (13 + 12 91 m12 − 91 (1)


113 ) = 0: (47)

The traction-prescribed boundary conditions (see Eq. (30)) on the top and bottom
surfaces of the beam are:

33 − 91 p1 − 93 p3 − 391 (1) (1)


133 − 93 333 = qQ±h=2 ;

1 1 h
13 − 91 m12 − 291 (1)
113 − 91 p3 − 93 m32 − 93 (1)
133 = 0 ∀z = ± ;
2 2 2
m32 + 2(1)
133 = 0; p3 + (1)
333 = 0; (48)

On the surfaces normal to the x-axis, the traction conditions are

tQ1 = 11 − 91 (p1 + (1) (1)


111 ) − 93 (p3 + 3113 );

tQ3 = 13 + 12 91 (m12 − 2(1) 1 (1) (1)


113 ) + 2 93 (m23 + 2133 ) − 93 (p1 + 3133 );

gQ2 = m12 − 2(1)


113 ;

rQ = p1 + (1)
111 ; (49)
1488 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

where qQ±h=2 is the prescribed normal force-traction on the top or bottom surfaces in
units of N=m2 , tQ1 and tQ3 are the prescribed force-tractions in units of N=m2 , gQ2 and
rQ are the prescribed tractions (N m=m2 ) of the couple and double force, respectively.
The boundary conditions at the edges of the top and bottom surfaces of the beam (Eq.
(31)) are
h
PQ 1 = ±(p3 + 3(1)113 ); PQ 3 = ±(p1 + 3(1)
133 ); z=± ; (50)
2
where PQ 1 and PQ 3 are the prescribed line forces per unit width (N/m).
With the presence of higher-order stress measures, the stress resultants on a beam
are de9ned as
 h=2  h=2

1
N= 11 d z; Q = 13 + 91 m12 − 91 (1)
113 d z;
h=2 h=2 2
 h=2
M= (z11 + m12 + p3 + (1)
113 ) d z;
h=2
 h=2  h=2
Nh = (p1 + (1)
111 ) d z; Mh = z(p1 + (1)
111 ) d z; (51)
h=2 h=2

where N is the axial stress resultant per unit width, Q is the shear stress resultant per
unit width, and M is the moment per unit width in the beam. New higher-order stress
resultants per unit width, N h , and higher-order moment resultants per unit width, M h ,
are introduced to account for the higher-order e=ect. In addition, the resultant shear
stress, Q, and the moment, M , now contain higher-order stress terms. Combining the
equilibrium equations in Eq. (47), the boundary conditions in Eq. (48) and the stress
resultant de9nitions in Eq. (51), we 9nd
dN d2 N h
− = 0; (52)
dx d x2
where
dQ dM d2 M h
+ qQ = 0; − − Q = 0; (53)
dx dx d x2
where qQ = qQh=2 + qQ−h=2 Eq. (53) can be re-written as
d2 M d3 M h
− + qQ = 0: (54)
d x2 d x3
A solution satisfying these two equations will satisfy the equilibrium and top and
bottom boundary conditions automatically. Then, from Eqs. (49) and (50), the stress
resultants at the boundary where the tractions prescribed are:
 h=2
dN h
N− = NQ = tQ1 d z + PQ 1 |z=−h=2 + PQ 1 |z=h=2 ; (551 )
dx −h=2
 h=2
Q=Q= Q tQ3 d z + PQ 3 |z=−h=2 + PQ 3 |z=h=2 ; (552 )
−h=2
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1489

 h=2
dM h Q h h
M− =M = (z tQ1 + gQ2 ) d z − PQ 1 |x=−h=2 + PQ 1 |x=h=2 ; (553 )
dx −h=2 2 2
 h=2  h=2
N h = NQ h = rQ d z; M h = MQ h = rz
Q d z; (554 )
−h=2 −h=2

where NQ is the prescribed total axial force in units of N/m; QQ is the prescribed shear
force in units of N/m; MQ is the prescribed moment per unit width in units of N m/m; NQ h
is the prescribed higher-order axial force in units of N m/m; and MQ h is the prescribed
higher-order moment per unit width in units of N m2 =m. In bending, the prescribed axial
stress resultant, NQ , and the higher-order axial stress resultant, NQ h , which correspond to
pure tension, are both zero.
In brief, the independent stress resultant measures for bending are the moment, M ,
and the higher-order, moment M h . The shear force, Q, is associated with M and M h
(Eq. (53)). The equations detailed in this section are used in the following section for
development of the bending equations.

3.3. Deformation and constitutive relations

In this section, we shall develop a new set of deformation measures for bend-
ing. The expressions of deformation and constitutive relations for thin beams can be
simpli9ed by expanding the displacements as a power series of the beam thickness
(O’Malley, 1991)

 ∞

u= ui (x; s)hi ; v = 0; w= wi (x; s)hi ; (56)
i=0 i=0

where s=z=h; h is the beam thickness and z is the coordinate in the thickness direction.
After expansion, the non-vanishing strains and strain gradients are

 9ui (x; s)
11 = hi ;
9x
i=0


 9wi+1 (x; s)
33 = hi ;
9s
i=0

∞  
1  9wi (x; s) 9ui+1 (x; s) i
13 = + h; (571 )
2 9x 9s
i=0

  
9 9ui (x; s) 9wi+1 (x; s) i
)1 = + h;
9x 9x 9s
i=0


  
9 9ui+1 (x; s) 9wi+2 (x; s) i
)3 = + h; (572 )
9s 9x 9s
i=0
1490 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

∞  
1  9 9wi (x; s) 9ui+1 (x; s) i
"21 = − − h;
4 9x 9x 9s
i=0

∞  
1  9 9wi+1 (x; s) 9ui+2 (x; s) i
"21 = − − h; (573 )
4 9s 9x 9s
i=0
∞ 


9 2 9ui (x; s) 2 9wi+1 (x; s) 1 92 ui+2 (x; s) i
(1)
111 = − − h;
9x 5 9x 5 9s 5 9s2
i=0

∞ 


9 4 9wi (x; s) 8 9ui+1 (x; s) 1 92 wi+2 (x; s) i
(1)
113 = + − h;
9x 5 9x 15 9s 5 9s2
i=0

∞ 

9 1 9ui (x; s) 8 9wi+1 (x; s) 4 92 ui+2 (x; s) i
(1)
133 = − − − h;
9x 5 9x 15 9s 15 9s2
i=0

∞ 

9 1 9wi (x; s) 2 9ui+1 (x; s) 2 92 wi+2 (x; s) i
(1)
333 =− + − h;
9x 5 9x 5 9s 5 9s2
i=0

(1)
122 = −( (1) (1)
111 + 133 ); (1) (1) (1)
223 = −( 133 + 333 ); (574 )

where
9u0 (x; s) 9w0 (x; s) 92 u1 (x; s) 92 w1 (x; s)
= = = =0 (58)
9s 9s 9s2 9s2
for the strains and strain gradients to be 9nite. In general, the constitutive relations for
linear elastic isotropic materials are


ij = 2 ij + mm ij ; pi = 2l20 )i ; (1) 2 (1) 2


ijk = 2l1 ijk ; mij = 2l2 "ij :
1 − 2

(59)

Given the non-vanishing strains and strain gradients for bending detailed above, the
non-vanishing stresses and higher-order stresses are
∞ 
 
1 −
9ui
9wi+1 i
11 = 2 + h;
1 − 2
9x 1 − 2
9s
i=0

∞ 
 
1 −
9wi+1
9ui i
33 = 2 + h;
1 − 2
9s 1 − 2
9x
i=0

∞ 
 
9wi 9ui+1 i
13 =  + h;
9x 9s
i=0

22 =
(11 + 33 ); (601 )
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1491


  
9 9ui (x; s) 9wi+1 (x; s) i
p1 = l20 + h;
9x 9x 9s
i=0

 ∞  
9 9ui+1 (x; s) 9wi+2 (x; s) i
p3 = l20 + h; (602 )
9s 9x 9s
i=0
∞  
l22  9 9wi (x; s) 9ui+1 (x; s) i
m21 = − − h;
4 9x 9x 9s
i=0

∞  
l22  9 9wi+1 (x; s) 9ui+2 (x; s) i
m23 = − − h (603 )
4 9s 9x 9s
i=0

and
∞ 


9 2 9ui (x; s) 2 9wi+1 (x; s) 1 92 ui+2 (x; s) i
(1)
111 = l21 − − h;
9x 5 9x 5 9s 5 9s2
i=0

∞ 


9 4 9wi (x; s) 8 9ui+1 (x; s) 1 92 wi+2 (x; s) i
(1) 2
113 = l1 + − h;
9x 15 9x 15 9s 5 9s2
i=0

∞ 


9 1 9ui (x; s) 8 9wi+1 (x; s) 4 92 ui+2 (x; s) i
(1) 2
133 = −l1 − − h;
9x 5 9x 15 9s 15 9s2
i=0

∞ 


9 1 9wi (x; s) 2 9ui+1 (x; s) 2 92 wi+2 (x; s) i
(1)
333 = −l21 + − h;
9x 5 9x 5 9s 5 9s2
i=0

(1)
122 = −((1)
111 + (1)
133 ); (1) (1) (1)
223 = −(133 + 333 ): (604 )
Rewriting the second and the third equations in boundary conditions (48) to include
only the non-vanishing stresses and higher-order stresses yields
13 − 12 91 m12 − 291 (1) 1 (1) 2
113 − 91p3 − 2 93 m32 − 93 133 = O(h );

m32 + 2(1) 2
133 = O(h ): (61)
Further combining the second equation in Eq. (47) with Eqs. (60) and (61) gives
9w2 (x; s)

u1 (x; s) = −w0 (x)s; = w (x)s;


9s 1−
0

2 −1  2
2 
92 u2 (x; s) 8l1 l22 2l1  16l1 l22 9w1 (x; s)
= + u (x) − − ;
9s2 15 4 5 0 15 4 9x9s

2 −1  2
2 
92 u3 (x; s) 1 8l1 l22 2l1 2l1 l22
= − + + −
w0 (x)s: (62)
9s2 1 −
15 4 5 3 4
The equilibrium relation for the axial stress resultant (Eq. (52)) combined with the
vanishing requirement of the prescribed axial and higher-order axial stress resultants
1492 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

give
9w1 (x; s) 92 u2 (x; s)
u0 (x) = = = 0: (63)
9s 9s2
For beam deformation, the curvature, 2, and the curvature gradient, 2h , are, respectively
2 = −w0 (x); 2h = −w0 (x); (64)
where w0 is the deGection of the beam at mid-surface. By combining Eqs. (60), (62)
and (64), the constitutive equations for the stresses and higher-order stresses in terms
of the curvature and the curvature gradients are
2
11 = z2 + O(z 3 ); (651 )
1−

l2 (1 − 2
) h l2 (1 − 2
)
p1 = 0 z2 + O(z 3 ); p3 = 0 2 + O(z 2 ); (652 )
1−
1−


2 −1 
2
2 
l21 8l1 l22 2l1 l22 2l1 l22
(1)
111 = + + − −
z2h + O(z 3 );
5(1 −
) 15 4 3 2 3 4

l21 4 −

(1)
113 = 2 + O(z 2 );
15 1 −


2 −1
2
l21 8l1 l22 l1 3l22
(1)
133 = − + +
z2h + O(z 3 );
5(1 −
) 15 4 4 4

l21 (1 +
)
(1)
333 = − 2 + O(z 2 ); (653 )
5(1 −
)
l2
m21 = 2 2 + O(z 2 ); m23 = −2(1) 133 : (654 )
2
The constitutive relations for beam bending can now be obtained by substituting
Eqs. (651–4 ) into Eq. (51). This gives the constitutive relations as
M = D2; M h = Dh 2h ; (66)
where the bending rigidities are

2
bh
D = D0 1 + ; D h = D 0 2 (67)
h
in which, D0 is the conventional bending rigidity,
h3
D0 = ;
6(1 −
)
2
b2h = 6(1 − 2
)l20 +(4 −
)l21 + 3(1 −
)l22 ;
5

−1 
2
2 
l2 l2 8l21 l2 2l1 l2 2l1 l2
2 = 0 (1 − 2
) + 1 + 2 + 2 − − 2
(68)
2 10 15 4 3 2 3 4
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1493

and bh and are the higher-order bending parameters, which characterize the thickness
dependence of beam bending.
From Eqs. (53), (552 ), and (553 ), the boundary conditions are
dM d2 M h
Q= − = QQ or w0 = wQ 0 ;
dx d x2
dM h
M− = MQ or w0 = wQ 0 ;
dx
Mh = 0 or w0 = wQ 0 ; (69)
h
where M is zero for a zero applied higher-order moment. The displacement boundary
conditions are
w0 = wQ 0 ; w0 = wQ 0 ; w0 = wQ 0 (70)
at a 9xed end,
9M h
w0 = wQ 0 ; M− = MQ ; M h = 0 (71)
9x
at a simply supported end, and
9M 92 M h Q 9M h
Q= − 2
= Q; M− = MQ ; M h = MQ h (72)
9x 9x 9x
at a free end.
After solving the bending problems, one can obtain the stresses and higher-order
stresses from Eqs. (651–4 ), except for 13 and 33 . The quantities 13 and 33 is
determined by integrating Eq. (47) from the bottom surface to z to give
1 1
13 = 91 m12 + 91 p3 + 291 (1)113 + 93 m32 + 93 (1)
133
2 2
 z
− 91 (11 − 91 p1 − 91 (1)
111 ) d z;
−h=2
1
33 = − 91 m32 + 91 p1 + 291 (1) (1)
133 + 93 p3 + 93 333 − qQ−h=2
2
 z
− 91 (13 + 12 91 m12 − 91 (1)
113 ) d z: (73)
−h=2

3.4. Cantilever beam with applied moment at the free end

In Fig. 1, a moment, MQ , is applied at the free end of a cantilever beam. The length
of the beam is a. For a vanished q, Q the equilibrium equation (54) becomes
2 3 h
d M d M
− = 0: (74)
d x2 d x3
The boundary conditions at the free end in Eq. (72) give
dM d2 M h dM h
M h = 0; − 2
= 0; M − = MQ ∀x = a: (75)
dx dx dx
There are two possible boundary conditions at the 9xed end, x = 0.
1494 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

z
Q
M
x

Fig. 1. A cantilever beam with an applied moment or force at the free end.

Case (a): The boundary conditions at x = 0 are


M h = 0; w0 = w0 = 0: (76)
From the constitutive relation in Eq. (66), we have,
Dh dM
Mh = ; (77)
D dx
where D and Dh are de9ned in Eq. (67). Solving Eqs. (64), (66) and (74) with the
boundary conditions in Eqs. (76) and (75) gives the moment and deGection as
1 MQ a 2
M = Mc = MQ a ; wa = wc = − x ; (78)
2 D
where MQ a is the applied moment. The moment and deGection have the same forms as
the classical solutions, except that the bending rigidity, D, has additional dependence
on the thickness and the higher-order material length scale parameters (Eq. (67)).
Case (b): The boundary conditions at x = 0 are
M
w0 = = 0; w0 = w0 = 0: (79)
D
Solving Eqs. (64), (66) and (74) with the boundary conditions in (79) yields the
moment as
 (a−x)= 
Q e + e(a−x)=
M = Mc − M (80)
ea= + e−a=
and the deGection as
 a=

MQ e − e−a= 2 e(a−x)= + e−(a−x)=
w0 = wc + x a= −  1 − ; (81)
D e + e−a= ea= + e−a=
where  is the ratio of Dh to D,
Dh ( =h)2
2 = = h2 (82)
D 1 + (bh =h)2
The quantities Mc and wc de9ned in Eq. (78) are the same as in the classical solution,
except that the bending rigidity, D, is the rigidity with the thickness size e=ect as
de9ned in Eq. (67).
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1495

3.5. Cantilever beam with applied force at the free end

Consider the case where a transverse force, Q,Q (Fig. 1) is applied at the free end of
a cantilever beam. The equilibrium equation is the same as before (Eq. (74)) and the
boundary conditions at the free end, x = a, are
dM d2 M h Q dM h
M h = 0; − = −Q; M− = 0: (83)
dx d x2 dx
The conditions at x = 0 are the same as in the previous example. For case (a), the
moment, M , and the deGection, w0 , are
 (a−x)= 
Q e + e−(a−x)= − ex= − e−x=
M = Mc − Q  ; (84)
ea= + e−a=

(a−x)= 
QQ 2 e + e−(a−x)= − ex= + e−x= 1 − ea=
w0 = wc +  x + 3 + (85)
D ea= − e−a= 1 + ea=
and for case (b) the solutions for M and w0 are
 (a−x)= 
(e + e−(a−x)= ) − (ex= − e−x= )
M = M − QQ  ; (86)
ea= + e−a=

(a−x)= 
QQ 2 + a(ea= − ea= ) 2 a(e + e−(a−x)= ) − e(e−x= − ex= )
w0 = wc + x + −a ;
D ea= + ea= ea= + e−a=
(87)

where


Q − x); QQ 1 3 1 2
Mc = Q(a wc = x − ax ; (88)
D 6 2
are the same as the classical solution except that D is the thickness-dependent bending
rigidity. The solutions for both cases (a) and (b) are compared and discussed in next
section.

3.6. Bending rigidity and higher-order bending rigidity

In the conventional bending theory, the bending rigidity, D0 , is proportional to the


cube of the beam thickness. In the present higher-order bending theory, the bending
rigidity, D, has additional dependence on the material length scale parameters and the
beam thickness and length. The bending behavior is also dependent on a higher-order
bending rigidity Dh , which is a function of the material length scale parameters only.
For discussion, we assume that all the material length scale parameters are the same,
i.e., l0 = l1 = l2 = l. From Eq. (68), we have
54 − 99
2
b2h = (10:6 − 15:4
)l2 ; 2 = l ≈ (0:57 − 1:05
)l2 : (89)
94
1496 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

15

13 ν =0.0
11 ν =0.38

9 ν =0.5
D/D 0
7

1
0 1 2 3 4 5
h/l

Fig. 2. Variation of D=D0 on h=l. It is assumed that l0 = l1 = l2 = l.

0.35

0.3 ν = 0.0
ν = 0.38
0.25 ν = 0.5

0.2
ε/ h

0.15

0.1

0.05

0
0 1 2 3 4 5
h/l

Fig. 3. Dependence of =h on h=l. It is assumed that l0 = l1 = l2 = l.

The normalized bending rigidity, D=D0 , for di=erent Poisson ratios is plotted in
Fig. 2. The 9gure shows that the bending rigidity, D, increases signi9cantly in com-
parison with the classical one when the beam thickness  is less than or close to the
material length scale parameter. The coeUcient,  = Dh =D, de9ned in Eq. (82) for
several Poisson ratios, is plotted in Fig. 3. This 9gure shows that the coeUcient, , is
less than a quarter of the beam thickness, h, which means that Dh is 7% or less of
Dh2 . The signi9cance of Dh will be explored in more detail in the next section.
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1497

1
ε =0.05h, a=5h, 10h, 50h
0.8 ε =0.1h, a=5h, 10h, 50h
ε =0.2h, a=5h, 10h, 50h

M − Mc a
M c x =0 ε
0.6

0.4

0.2

0
0 0.5 1 1.5 2
(a) r/h

0.006
a/ε =25
0.005
a/ε =50
a/ε =100
0.004
w − wc
wc x = a

0.003

0.002

0.001

0
0 0.2 0.4 0.6 0.8 1
(b) x/a

Fig. 4. The normalized di=erences between the higher-order solutions (excluding D’s size dependence) and
the classical solutions. The higher-order moment is set to zero at the 9xed end.

3.7. The e<ect of boundary conditions

Bending solutions in conventional elasticity are a=ected by the assumed boundary


condition at the 9xed end of the beam. The e=ects of di=erent boundary conditions on
bending behavior in higher-order solutions are explored in this section.
The higher-order moment can be set to zero. In this case, the solution closely ap-
proximates the conventional beam solution in form except that D0 is replaced with the
size-dependent bending rigidity, D. In addition the di=erence between the higher-order
solution and the classical one involves the higher-order bending rigidity, Dh . This dif-
ference is absent in the example of a cantilever beam with a moment applied at the
free end. The di=erences in the moment and deGection in the example of a cantilever
beam with a force applied at the free end are plotted in Fig. 4. Fig. 4(a) shows that
a boundary layer exists near the 9xed end and another near the free end of the can-
tilever beam. The sizes of the boundary layers, are dependent on the coeUcient, . The
di=erence in the moment is about 4% for  ≈ 0:2 and a = 5h (a short beam) near the
beam end, where the di=erence is at the maximum. Since the di=erence in the moment
1498 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

1
ε =0.05h, a=5h, 10h, 50h
0.8 ε =0.1h, a=5h, 10h, 50h
ε =0.2h, a=5h, 10h, 50h

M − Mc
M c x =0
0.6

0.4

0.2

0
0 0.5 1 1.5 2
(a) r/h

a/ε =25, M
0.16 a/ε =50, M
a/ε =100, M
a/ε =25, F
0.12 a/ε =50, F
w − wc
wc x = a

a/ε =100, F
0.08

0.04

0
0 0.2 0.4 0.6 0.8 1
(b) x/a

Fig. 5. The di=erences between the higher-order solutions (excluding the D’s size dependence) and the
classical solutions for cases with a moment or a force applied at the free end of the beam. The second-order
derivative for the deGection is set to zero at the 9xed end.

in the boundary layers is proportional to the coeUcient, , and inversely proportional


to the beam length, a, this di=erence further diminishes when a=h is large. Fig. 4(b)
shows that the maximum di=erence in deGection is about 0.5% at  ≈ 0:2h (near the
maximum) and a = 5h (a short beam) is at the free end. Given the small di=erence
and the fact that the di=erence diminishes with beam length, the contribution from Dh
relative to that from D can be ignored.
An alternate boundary condition is to have zero second derivatives for the deGection.
This case is similar to the 9rst case, except that a boundary layer exists near the 9xed
end. Setting aside once again the size dependence in D, the di=erences between the
moment and deGection solutions and the classical solutions are plotted in Fig. 5. The
size of the boundary layer is again dependent only on the small  shown in Eq. (82) for
both cases. The maximum di=erence in deGection at the free end is about 11% for the
moment load and 6% for the force load. In comparison to the dominant inverse squared
dependence of bending rigidity, D, on the thickness, these contributions stemming from
Dh can be ignored.
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1499

10
Eq. (88)
8 Eq. (85), a=5h
Eq. (85), a=10h
Eq. (85), a=20h

3D0 w x =a
6 Eq. (87), a=5h
Qa3 Eq. (87), a=10h
4 Eq. (87), a=20h

0
0 1 2 3 4 5
h/l

Fig. 6. The size dependence of a cantilever beam bending using several solution forms and beam lengths.

Finally, the normalized load parameter group is plotted as a function of h=l in Fig.
6 for several beam lengths according to the solutions in Eqs. (85) and (87) and their
lead term in Eq. (88)2 . The 9gure shows that all curves collapse onto a single curve.
This indicates that the size dependence of cantilever beam is dominated by the bending
rigidity, D, in the beam, and the e=ect of Dh is limited in the boundary layer near the
9xed end.

4. Deformation experiments with and without strain gradients

4.1. Bending with strain gradient e<ects

The behavioral di=erence between classical and strain gradient elasticity can be ex-
perimentally investigated using simple cantilever beam bending. The bending theory
developed in the last section predicts that the bending sti=ness has additional size de-
pendence. In this section, experiments and analyses on the structural scale in which size
sti=ening becomes signi9cant are detailed. Consider the bending of a thin cantilever
beam. The classical relation between the load and deGection is
Q 3
1 Qa
w0 = ; (90)
3 D0 h3
where
E
D0 = (91)
12’
is a bending sti=ness parameter independent of the beam geometry with ’ = 1 −
2
for plane strain and ’ = 1 for plane stress, a is the distance of the loading point to
the cantilever mount and h is the thickness of the beam. When strain gradients are
1500 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

6 h>>b
h

5 h=2b
h
h= b
Q (N/m)

4 h
D'
h=0.5bh
3

1 D0'

0
0 0.5 1 1.5 2 2.5 3 3.5
3w0(h/a)3 (nm)

Fig. 7. The load versus normalized deGection for cantilever bending for di=erent beam thicknesses showing
size e=ect.

included, the load-deGection solution is


Q 3
1 Qa
w0 ≈ ; (92)
3 D  h3
where (from Eq. (67))

2
  bh
D = D0 1 + : (93)
h

In contrast to D0 ; D is dependent on the beam thickness, h, and the higher-order


bending parameter, bh ; is de9ned in Eq. (68)2 and is dependent on the higher material
length scale parameters and Poisson’s ratio. For beams with h much larger than bh , nor-
malized load-deGection are predicted to collapse onto the thick line, but departure from
the classical solution is predicted for thin beams (thin lines in Fig. 7) with thickness h
near bh . In this section, bending experiments to determine the size dependence of D
are described. During bending, the load–displacement behavior is potentially a=ected
by the compliance of the supporting mount. In the following section, the deformation
contribution of the mount is analyzed to facilitate the specimen design.

4.2. Experimental design

At the joint of the beam and the support, the local deformations w0 in the parentheses
are functions of the local force and moment (Q0 and M0 ),
   
w0 SPP SPM Q0
= ; (94)
w0 SPM SMM M0
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1501

Normalized bending rigidity D'/D'0


0.95

0.9

a/h=10
0.85 a/h=20
a/h=50

0.8
0 0.1 0.2 0.3 0.4 0.5
Beam thickness (mm)

Fig. 8. The normalized bending rigidity of an epoxy beam as a function of beam thickness showing the
e=ect of the underlying glass substrate on the bending rigidity for di=erent a=h.

where S’s in the square bracket are the compliances, which are dependent on the shape,
size and properties of the support substrate, and the adhesion of the beam and support
substrate. It can be easily shown that the normalized bending rigidity of a cantilever
beam on an elastic support is



2 −1
D  h 2 h h
 = 1 + 3D0 SMM h + 2SPM h + SPP
D0 a a a

−1
h
≈ 1 + 3D0 SMM h2 (95)
a
and is dependent on the S’s of the joint. To determine the inGuence of the support on
the bending rigidities, we conducted a three-dimensional 9nite element analysis using
a commercial FEM software ANSYS with about 14,000 elements. Epoxy beams with
widths of 0:235 mm on glass support substrates with dimensions of 150 × 100 × 5 mm3
are modeled. The normalized rigidities for beams with varied thicknesses are plotted
in Fig. 8. For 9xed ratio a=h, the 9gure shows that the normalized bending rigidity
approaches the result of a perfectly 9xed support when the beam thickness is decreased.
The e=ect of the glass support on the normalized bending rigidity is less than 10% for
a=h = 10 and less than 5% for a=h = 20. These results indicate that the e=ect of the
glass support of epoxy beams on the cantilever bending rigidities is minor for beams
with large a=h.

4.3. Specimen preparations

Epoxy polymeric beams were fabricated by casting. Bisphenol-A epichlorohydrin


resin was mixed with 20 phr of diethylenetriamine hardener. The mixed liquid was
degassed, spin cast and sandwiched between two low-adhesion glass plates. The beam
1502 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

Bending Test
2.5
Tension Test

2
E (GPa)
1.5

0.5

0
0 20 40 60 80 100 120
Thickness (µm)

Fig. 9. Elastic moduli of beams as a function of beam thickness from tension and bending tests.

thicknesses were controlled using spacers. The cast sandwich assembly was then cured
at 100◦ C for 3 h and cooled slowly to room temperature. The glass plates were then
removed from the cured epoxy plates and parallel cuts were made on the epoxy plate
to form rows of beams. One end of the beams was adhesively mounted onto a glass
substrate for cantilever bending tests. Four sets of beams were fabricated with thickness
of 20, 38, 75 and 115 m.

4.4. Elastic behaviors in the absence of strain gradients

The elastic behaviors of the microbeams in the absence of strain gradients were
characterized in uniaxial tension tests using a mechanical tester (Dynamical Mechanical
Analyzer, Perkin Elmer), according to the testing methodology described by Menard
(1999). The microbeams were individually center aligned with a tension jig under a
microscope and adhesively 9xed onto one end of the jig. The jig was then suspended
on the tester, clamped and pulled at room temperature. The elastic moduli obtained
from the load–displacement curves are plotted in Fig. 9. The results indicate that the
elastic behaviors of the beams, in the absence of strain gradient, are independent of
the beam thickness.

4.5. Elastic behaviors in the presence of strain gradients

Unlike simple tension, bending generates a strain gradient in the beam. Bending tests
were conducted using a nano-loading system (Hysitron Triboindenter) with a 20 m
long wedge head with a 90◦ tip angle and a rounded 1 m tip radius. The machine
had an electrostatic actuator with the load range of 10 mN and the load resolution of
0:1 N. The maximum displacement was 5 m and the displacement resolution was
0:1 nm. The loading and unloading behaviors of a typical beam is shown in Fig. 10.
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1503

400

350
h=38 µm
300

Load (µN)
250

200

150

100

50

0
0 500 1000 1500 2000 2500
Deflection (nm)

Fig. 10. Loading and unloading behaviors of a 38 m epoxy cantilever beam showing elastically reversible
behavior.

500
h = 20 (µm)
400 h = 38 (µm)
h = 75 (µm)
h =115 (µm)
300
Q (µ)

200

100

0
0 500 1000 1500 2000
w0 (nm)

Fig. 11. Cantilever beam bending at a=h = 10 for beams with di=erent thicknesses.

The loading and unloading curves are essentially collinear, indicating linear elastic and
reversible behavior.
The classical solution for cantilever bending indicates that the deGection is dependent
on the ratio of the beam length to the beam thickness. The bending load-deGections of
beams with di=erent thickness should collapse onto a single curve if the ratio of the
distance from the mount to the load point, a, to the beam thickness, h, is 9xed (Eq.
(90)). The load-deGection curves for beams with four di=erent thicknesses at 9xed a=h
are plotted in Fig. 11. The data do not collapse onto a single curve as predicted by the
classical elasticity solution. The corresponding D determined from the load-deGection
curves is plotted in Fig. 12, which shows that D increased by about 2.4 times when
the beam thickness reduced from 115 to 20 m.
1504 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

0.35
Higher order

Bending rigidity D' (GPa)


0.3 Conventional

0.25

0.2

0.15

0.1
0 20 40 60 80 100 120
Thickness (µm)

Fig. 12. D and D0 as a function of beam thickness (D0 is computed using E from tension tests).

4.6. Data analyses and models comparisons

The experimentally measured D increased by more than two times when the thick-
ness was decreased. In this section, analyses on possible factors contributing to the
increase are detailed. The normalized bending rigidities, D , may be a=ected by the
mounting support and by deviation from plain strain conditions. The e=ect of the
mounting support has been analyzed earlier and the increase in D (Fig. 12) is far
larger than the e=ect from the support mount (Fig. 8). The error from the deviation
from the model-assumed plain strain condition is also minor. The change in the nor-
malized bending rigidity from the plain stress condition (thickest beam) to the plain
strain condition (thin beam) is
2 =(1 −
2 ), which is approximately 10% for
= 0:3
and about 1=3 for
= 0:5 when the beams are incompressible. These changes are too
small to explain the doubling of D .
Other than errors associated with the experimental set-up and model assumptions,
variations in microstructures from fabrication may also lead to variation in D . The
formation of a sti= surface layer may potentially lead to variation in the elastic behavior.
The variation in the tensile elastic modulus has been determined to be less than 10%
(Fig. 9). Assuming that the variation is related to the presence of a surface layer on
both surfaces, the contribution of the surface layer can be modeled with the sandwich
beam shown in Fig. 13(a). The error on the tensile elastic modulus is then


E − E1 2t E2
et = = −1 ; (96)
E1 h E1

where E2 is the elastic modulus of the top and bottom layers and E1 is the elastic
modulus of the bulk. The corresponding experimental error in bending rigidity in the
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1505

E2
x

th-2t y

E1
t

(a) E2

0.7
Relative error of bending rigidity eb

0.6 e =5%
t
e =10%
t
0.5
e =15%
t
0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5
(b) 2t/h

Fig. 13. The e=ect of variation of elastic modulus on bending rigidity analyzed using a sandwiched beam
geometry.

bending test is


3
D − D1 E2 2t
eb = = −1 1− 1− : (97)
D1 E1 h

Combining these two equations and writing in terms of et gives




2
2t 2t
eb = et 3 − 3 + 6 3et : (98)
h h

The error in the bending rigidity is dependent on the tensile elastic modulus as well
as the ratio of the layer thickness to the specimen thickness. From the equation, the
relative error in the tension test must be larger than 1=3 if the bending rigidity, D ,
in the bending test is to double D0 . For et = 10%, the 9gure shows that eb is less
than 30% (Fig. 13(b)). Thus, the doubling of D is not associated with the presence of
1506 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

surface layers in the beam, and the size dependence is outside the range of variation
in material inhomogeneities.
Since the size e=ect is not observed in the tension with no strain gradients and is
massively observed in the bending when strain gradients are present, the doubling of the
normalized rigidity may be attributed to strain gradient e=ects. The higher-order strain
gradient model predicted D to vary as the inverse square of the beam thickness (Eq.
(93)). Fitting the data in Fig. 12 to the bending model in Eq. (93) using bh = 24 m
indicates that the higher-order strain gradient bending model (the curve in Fig. 12)
agrees well with the model. Secondly, the elastic modulus deduced from the bending
data is also consistent with the elastic modulus from the tension tests. This means that
the conventional elastic dependence on strain remains valid and that strain gradients
are contributing signi9cantly to the elastic deformation of thin beams. The agreement
between experiments and the strain gradient model shows that strain gradients are
potentially important in small-scale structures in MEMS and NEMS.

5. Discussion and conclusions

In this paper, we developed a modi9ed set of second-order deformation metrics,


the dilatation gradient vector, the deviatoric stretch gradient tensor and the rota-
tion gradient tensor, and derived the corresponding work-conjugate stress metrics as
the basic strain and stress measures for a strain gradient theory for elasticity. The
new second-order deformation metrics are mutually independent and only the dilata-
tion scalar and the dilatation gradient vector depend on volumetric deformation. With
the new set of metrics, application of the higher-order equilibrium to the strain gradient
theory leads to the conclusion that the deformation energy density is independent of
the anti-symmetric part of the rotation gradient tensor. The total deformation energy
density is a function of the symmetric strain tensor, the dilatation gradient vector, the
deviatoric stretch gradient tensor and the symmetric rotation gradient tensor. It is inde-
pendent of the anti-symmetric rotation gradient tensor. As a result, the modi9ed strain
gradient elasticity theory has only three higher-order material length scale parameters
for isotropic linear elastic materials.
From the modi9ed strain gradient elasticity theory, an elastic bending theory for thin
plane-strain beams is developed. In the bending theory, the deformation measures are
the conventional second-order derivative (curvature) and the third-order derivative (cur-
vature gradient) of the mid-plane normal deGection. The corresponding work-conjugate
stress measures are the conventional moment and a higher-order moment. The equi-
librium equation, constitutive relations and boundary conditions have been established
accordingly. The equilibrium equation is third-order in terms of the moment and the
higher-order moment and is sixth-order in terms of deGection. In the constitutive re-
lations, both the normalized regular bending rigidity (D=D0 ) and higher-order bending
rigidity (Dh =D0 ) depend on the material length scale parameters, but only D=D0 de-
pends on the beam thickness. The bending theory predicts that the bending rigidity
is dependent on the inverse square of the beam thickness. We conducted experiments
to determine the signi9cance of the thickness dependence and found that micron-sized
D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508 1507

isotropic elastic cantilever epoxy beams clearly demonstrated such behavior. The nor-
malized bending rigidity increases with decreases in beam thickness. This size ef-
fect, which cannot be explained by classical elasticity, is in good agreement with the
higher-order bending solutions. This shows that strain gradients signi9cantly sti=en the
deformation of micron-scale structures resulting in elastic strain gradient-based size
dependence.

Acknowledgements

This work was supported by the Research Grants Council of the Hong Kong Special
Administrative Region, the People’s Republic of China. Fan Yang and Jun Wang ac-
knowledge the support from the Fund of Southwest Jiaotong University in the People’s
Republic of China.

References

Albrecht, T.R., Akamine, S., Carver, T.E., Quate, C.F., 1990. Microfabrication of cantilever for the atomic
force microscope. J. Vacuum Sci. Technol. A-Vacuum Surf. Films 8, 3386–3396.
Bashir, R., Gupta, A., Neudeck, G.W., McElfresh, M., Gomez, R., 2000. On the design of piezoresistive
silicon cantilevers with stress concentration regions for scanning probe microscopy applications.
J. Micromech. Microeng. 10, 483–491.
Carr, D.W., Craighead, H.G., 1997. Fabrication of nanoelectromechanical systems in single crystal silicon
using silicon on insulator substrates and electron beam lithography. J. Vacuum Sci. Technol. B 15,
2760–2763.
Chong, A.C.M., Lam, D.C.C., 1999. Strain gradient plasticity e=ect in indentation hardness of polymers.
J. Mater. Res. 14, 4103–4110.
Craighead, H.G., 2000. Nanoelectromechanical systems. Science 290, 1532–1535.
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., Hutchinson, J.W., 2001. A reformulation of strain gradient plasticity. J. Mech. Phys. Solids. 49,
2245–2271.
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.
Gao, H.J., Huang, Y., Nix, W.D., Hutchinson, J.W., 1999. Mechanism-based strain gradient plasticity-I.
Theory. J. Mech. Phys. Solids. 47, 1239–1263.
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.
Manias, E., Chen, J., Fang, N., Zhang, X., 2001. Polymeric micromechanical components with tunable
sti=ness. Appl. Phys. Lett. 79, 1700–1702.
Menard, K.P., 1999. Dynamic Mechanical Analysis: a Practical Introduction. CRC Press, Boca Raton, FL.
Mindlin, R.D., 1964. Micro-structure in linear elasticity. Arch. Rational Mech. Anal. 16, 51–78.
Mindlin, R.D., 1965. Second gradient of strain and surface tension in linear elasticity. Int. J. Solids Struct.
1, 417–438.
Mindlin, R.D., Tiersten, H.F., 1962. E=ects of couple-stresses in linear elasticity. Arch. Rational Mech. Anal.
11, 415–448.
Nix, W.D., 1989. Mechanical properties of thin 9lms. Metall. Trans. A 20, 2217–2245.
1508 D.C.C. Lam et al. / J. Mech. Phys. Solids 51 (2003) 1477 – 1508

O’Malley, R.E., 1991. Singular Perturbation Methods for Ordinary Di=erential Equations. Springer, New
York.
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 (2), 2855–2865.
StLolken, 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. Rational Mech. Anal. 11, 385–414.
Yang, F., Chong, A.C.M., Lam, D.C.C., Tong, P., 2002. Couple stress Based Strain gradient theory for
elasticity. Int. J. Solids Struct. 39, 2731–2743.

Potrebbero piacerti anche