International Journal of Solids and Structures 80 (2016) 28–37

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

A reformulation of constitutive relations in the strain gradient elasticity

theory for isotropic materials
Shenjie Zhou a,b,∗, Anqing Li a,b, Binglei Wang c
a
School of Mechanical Engineering, Shandong University, Jinan, Shandong 250061, People’s Republic of China
b
Key Laboratory of High Efficiency and Clean Mechanical Manufacture, Shandong University, Ministry of Education, People’s Republic of China
c
School of Civil Engineering, Shandong University, Jinan, Shandong 250061, People’s Republic of China

a r t i c l e i n f o a b s t r a c t

Article history: The general isotropic strain gradient elasticity theory with five higher-order elastic constants is reformu-
Received 12 April 2015 lated by introducing two different orthogonal decompositions of the strain gradient tensor. Just applying the
Revised 12 September 2015
mathematical reformulations, no extra conditions needed, the constitutive relations, equilibrium equation
Available online 26 October 2015
and boundary conditions are reformulated. In the reformulated theory, the number of independent higher-
Keywords: order elastic constants is proved to be three for isotropic materials, which indicates that the five higher-order
Strain gradient elasticity elastic constants in the general isotropic strain gradient elasticity theory are dependent with each other.
Constitutive behavior Therefore, the general strain gradient elasticity theory contains only three independent material length-scale
Microstructure parameters for isotropic materials in addition to the Lame constants. The new theory is different from the ex-
Size effects isted strain gradient elasticity theory with one or three material length-scale parameters, which introduces
extra conditions during deriving process. Moreover, the reformulated theory can be directly reduced to that
of incompressible materials by assuming the terms associated with hydrostatic strains to be zero. Some ex-
amples, such as torsion of cylindrical bars, shearing of fixed-end layers, and pure bending of thin beams, are
performed to reveal the necessity of including multi-length-scale parameters in the strain gradient elasticity
theory to predict size effects at micron scale.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction involved in the constitutive relations. Different versions of higher-

Many experiments have shown the size-dependent deformation
behaviors in micron scale. In the non-uniform plastic deformation,
the size effects have been observed in the experiments of measur-
ing micro-indentation hardness of metallic materials (Ma and Clarke,
1995), shear strength of copper wires in torsion (Fleck et al., 1994),
and bend moments of ultra-thin beams (Stolken and Evans, 1998).
By contrast, in elastic deformation, the size dependence of the nor-
malized bending rigidity exists in micro-beams of both metals and
polymers (Guo et al., 2005; Tang and Alici, 2011a, 2011b; Lam et al.,
2003). Because of the lack of internal length-scale parameters, the
classical elasticity and plasticity theories fail to describe such a be-
havior in the micron scale. Meanwhile, load or geometrically induced
stress singularities cannot be properly accounted by the standard
continuum mechanical models. However, this is possible with the use
of higher-order continuum mechanics theories, where intrinsic pa-
rameters correlating the microstructure and the macrostructure are
∗
Corresponding author at: School of Mechanical Engineering, Shandong University,
Jinan, Shandong 250061, People’s Republic of China. Tel.: +86 531 88396708; fax: +86
531 88392700.
E-mail address: zhousj@sdu.edu.cn (S. Zhou).
order continuum mechanics have been developed by many authors.
Relationships between those theories have been discussed by Tekoglu
and Onck (2008).
Generally, higher-order theories can be classified into couple
stress theories and general strain gradient theories, according to
the deformation metrics used. In the classical couple stress theory
(Toupin, 1962; Mindlin and Tiersten, 1962; Koiter, 1964), only the
gradient of the rotation vector enters the strain energy density func-
tion, and, hence, two additional material parameters are introduced
besides the Lame constants for isotropic materials. By introducing
a so-called equilibrium condition of moments of couples (the cou-
ple of force couples) to force the couple stress tensor to be symmet-
ric, Yang et al. (2002) modified the classical couple stress theory to
include only one additional material parameter. Recently, however,
Hadjesfandiari and Dargush (2011) came to a contrary conclusion that
the couple stress tensor is of the skew-symmetric character.
The more general strain gradient elastic theory including all com-
ponents of the higher-order deformation is proposed by Mindlin
(1964) to describe the linear elastic behavior of microstructures.
This theory requires 16 additional length constants for isotropic ma-
terials in addition to two Lame constants. The application of this
theory is limited as it requires the formidable task of determining,

http://dx.doi.org/10.1016/j.ijsolstr.2015.10.018
0020-7683/© 2015 Elsevier Ltd. All rights reserved.
 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37 29

theoretically or experimentally, 16 additional constants. For practical number of independent material constants needed for an isotropic
purposes, Mindlin and Eshel (1968) further formulated three simpler material in the general strain gradient elasticity theory. The rest is
versions of the general isotropic theory, utilizing only two material organized as follows. Section 2 reviews the general isotropic strain
and five internal length-scale constants in the final constitutive re- gradient theory. In Section 3, two new sets of independent higher-
lation rather than 18 used in Mindlin’s initial model. One of these order deformation metrics are developed to split the strain gradi-
versions uses the classical strains and the second-order gradient of ent tensor into mutually independent parts and the corresponding
displacement as the deformation metrics. In the second version, the work-conjugated higher-order stress tensors are defined. In Section 4,
second-order gradient of displacement is replaced by the gradient of the constitutive relations are reformulated and the number of inde-
strain, and in the third version, the deformation variables include the pendent higher-order material constants is proved to be three for
classical strains, the gradient of rotation, and the fully symmetric part isotropic linear elastic materials. Then, the general isotropic strain
of the gradient of strain. Although the simpler versions reduce inde- gradient elasticity theory containing three higher-order elastic con-
pendent length-scale parameters from 16 to five for isotropic mate- stants is re-expressed in the form of strain gradient components, and
rials, the application of this theory in engineering is limited as five the equilibrium relations and boundary conditions are derived by ap-
length-scale parameters are still very difficult to be determined ex- plying the variational principle of the strain gradient theory. Section 5
perimentally. Fleck and Hutchinson (1993, 2001) extended the first presents the contribution of each strain gradient component and the
version of Mindlin theory to plasticity and proposed a deformation influence of higher-order length-scale parameters through three ba-
theory of strain gradient plasticity, which involves three length-scale sic problems. Finally, conclusions are summarized in Section 6.
parameters. Lam et al. (2003) reformulated the theory by applying
a set of higher-order metrics to characterize strain gradient behav- 2. Review of general strain gradient elasticity theory
iors, and proposed a isotropic strain gradient elasticity theory. In their
theory, the equilibrium condition of moments of couples is applied In the general strain gradient elasticity theory (Mindlin and Eshel,
to force the symmetric character of the couple stress tensor and re- 1968), the total strain energy density is a function of strain and its
duce the number of elastic length-scale parameters from five to three. first-order gradient, given by
However, in our opinion, the mechanical effect of the moment cou- w = w(εi j , ηi jk ), (1)
ple is unable to be comprehended for the free character of moment
vectors. where ɛij is the symmetric strain tensor and ηijk is the strain gradient
In addition to the simplified model of the general strain gradient tensor with the minor symmetry in the last two indices. The strain
elasticity theory developed by Mindlin, a simple model of isotropic tensor and strain gradient tensor are defined, respectively, as
strain gradient elasticity with only one length-scale parameter has 1
εi j = (ui, j + u j,i ), (2)
been formulated by Aifantis (1992), in which classical stresses are re- 2
lated to classical strains and the Laplacian gradient of strains. Subse-
quently, this theory has been extended with additional terms to ac- ηi jk = εk j,i , (3)
count for surface effects (Vardoulakis and Sulem, 1995; Vardoulakis where ui is the displacement vector and a comma denotes the dif-
et al., 1996; Exadaktylos, 1998). The theory proposed by Aifantis can ferentiation with respect to the coordinates. Then, the corresponding
be formally obtained as a special case of the Mindlin theory (Li et al, stress σ ij and the higher-order stress τi jk ( = τik j ) work-conjugated to
2004; Lazar and Maugin, 2005; Askes and Aifantis, 2011). Although the strain ɛij and the strain gradient ηijk , respectively, can be written
one length-scale parameter can be expediently determined by simple as
bending or torsion test, it has been demonstrated that the strain gra-
∂w ∂w
dient plasticity theory with a single length parameter does not have a σi j = , τi jk = . (4)
scope to include the wide range of small-scale phenomena (Fleck and
∂εi j ∂ηi jk
Hutchinson, 2001). Therefore, the strain gradient theory with multi- For a volume V of a solid with boundary S and sharp edge C, the
ple length parameters is necessary to capture the size effects of me- principle of virtual work for the strain gradient theory is
    
chanic behaviors at the micron scale.
Recently, strain gradient theory has new development. Polizzotto (σi j δεi j + τi jk δηi jk )dV = b̄k δ uk dV + t̄k δ uk + r̄k Dδ uk dS
V V S
(2012) proposed a gradient elasticity theory for continua featured 
by not only a strain energy depending on the strain and the first- + f¯k δ uk dC, (5)
C
order strain gradient, but also a kinetic energy depending on the ve-
locity and the first-order velocity gradient, in which the effects of where b̄k is the body force per unit volume, t̄k is the surface traction,
both strain gradient and higher-order inertia are combined. Further, r̄k is the surface double-force traction, f¯k is the line load along the
this theory has been even extended to the second-order strain gradi- sharp edge, and D = ni ∂i denotes the normal gradient operator. The
ent elasticity with second-order velocity gradient inertia (Polizzotto, equilibrium equation in the body V can be derived using the virtual
2013). Moreover, Auffray et al. (2013) derived and provided the ex- work principle as
plicit matrix representations of the sixth-order elastic tensor for all σik,i − τi jk,i j + b̄k = 0, (6)
the three-dimensional (3D) anisotropic cases in a compact and well-
structured manner. In addition, Mühlich et al. (2012) developed an al- and the boundary conditions on S and along C are, respectively,
   
ternative method for the approximation of the material properties in t̄k = ni σik − τi jk, j + (D p n p )ni n j τi jk − Di n j τi jk or ūk = uk , (7)
linear elastic strain gradient effective media. Bacca et al. (2013) pro-
vided an analytical approach to the determination of the parameters
r̄k = ni n j τi jk or Duk = Duk , (8)
defining an elastic higher-order (Mindlin) material as the homoge-
nization of a heterogeneous Cauchy elastic material. Although many and
new achievements have been made, the strain gradient theory should f¯k = [ni k j τi jk ] or ūk = uk , (9)
be contributed more as the basis to form a unified and effective the-
ory for application. where ni is a unit vector normal to the boundary surface S, Di =
The purpose of this article is to propose a general strain gradient (δik − ni nk )∂k is the surface gradient operator, and kj is the outer co-
elasticity theory by reformulating the constitutive relations in terms normal vector satisfying the following relation
of two sets of independent higher-order metrics and determine the k j = eik j si nk , (10)
30 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37

with the alternating tensor eikj and the unit vector si tangent to the 1 
edge C. The square brackets in Eq. (9) represent the difference be-
τi(jk1) = τijks − s + δ τ s + δ τ s
δi j τmmk jk mmi ki mm j , (21)
5
tween the values of the enclosed quantity on the two sides of the
1  
edge. τijkas =  − τ + e e τ + e e τ ,
2τi jk − τ jik k ji i jp kts tsp ikp jts tsp (22)
6
3. Development of independent higher-order metrics
1 
τi(jk2) = s + δ τ s + δ τ s
δi j τmmk jk mmi ki mm j
In this section, two sets of independent higher-order deformation 5
metrics will be developed by introducing two different orthogonal 1  
decompositions of the strain gradient tensor, of which one is based
+ 2τ i jk − τ  jik − τ  k ji − ei jp ekts τ  tsp − eikp e jts τ  tsp , (23)
6
on a hydrostatic/deviatoric splitting, the other on a symmetric/anti- with
symmetric splitting.
1  
For the hydrostatic/deviatoric splitting, it is known that the strain τijks =  + τ ,
τi jk + τ jki ki j τijk = τi jk − τihjk , (24)
tensor can be expressed as its spherical and deviatoric parts 3
where τijk
s is the symmetric component of the deviatoric higher-order
1
εi j = δi j εnn + εi j , (11) stress tensor, and the deviatoric part τijk of higher-order stress is
3
work-conjugate to the deviatoric strain gradient ηi jk .
where δ ij is the Kronecker delta, 13 εnn is the mean or hydrostatic
For the symmetric/anti-symmetric splitting, the strain gradient
strain associated with a change in volume, and εi j is the deviatoric
tensor ηijk can be directly decomposed into its symmetric and anti-
strain tensor associated with a change in shape. Following Eq. (11), symmetric parts, ηisjk and ηiajk , (Fleck and Hutchinson, 1997), respec-
the strain gradient tensor ηijk can also be decomposed into a hydro-
tively, as follows:
static part ηihjk and a deviatoric part ηi jk as
1 
ηisjk = ηi jk + η jki + ηki j , ηiajk = ηi jk − ηisjk
ηi jk = η h
i jk + ηi jk , (12) 3
1 1
where = ei jp χ pk + eikp χ p j , (25)
3 3
1 1
η h
i jk = δ jk εnn,i = δ jk ηinn , ηi jk = ηi jk − ηihjk = εk j,i . (13) where χi j = eipq η pq j is the curvature tensor. Then, the symmetric part
3 3
ηisjk can be further split into a trace part, ηi(jk0) , and a traceless part,
Here, the deviatoric part ηi jk is the deviatoric strain gradient εk j,i
associated with the shape change in strain gradient, and, in fact, ηi(jk1) , (Lam et al., 2003) and, moreover, new independent strain gradi-
equals the plastic strain gradient introduced by Fleck and Hutchinson ent metrics can be obtained by splitting the anti-symmetric part ηiajk
(2001). In accordance with the decomposition of plastic strain gradi- into two independent parts, ηiasjk and ηiaa jk
, according to the decompo-
ents into three parts proposed by Fleck and Hutchinson (2001), the sition of the curvature tensor χ ij into its symmetric part χisj and anti-
strain gradient ηijk can be decomposed into four independent com- symmetric part χiaj . Thus, the new set of independent components is
ponents as given by
ηi jk = ηihjk + ηi(jk1) + ηias (2)
jk + ηi jk , (14) ηi jk = ηi(jk0) + ηi(jk1) + ηiasjk + ηiaajk , (26)
with where
(1) 2   1 
ηi jk = ηisjk − δi j ekpq + δ jk eipq + δki e jpq χ a ,
pq (15) ηi(jk0) = δi j ηmmk
s
+ δ jk ηmmi
s
+ δki ηmm
s
j , (27)
15 5

ηias
jk =
1 s + e χ s ,
ei jp χ pk ikp p j
1
(16) ηi(jk1) = ηisjk − ηi(jk0) , (28)
3 3
1 1
  a ηiasjk = ei jp χ pk
s
+ eikp χ ps j , (29)
(2) 1 a + 1 e χ a + 2 δ e 3 3
ηi jk = ei jp χ pk ikp p j i j kpq + δ jk eipq + δki e jpq χ pq ,
3 3 15 and
(17) 1 1
ηiaajk = ei jp χ pk
a
+ eikp χ pa j , (30)
where 3 3
1   with
ηisjk = η i jk + η ki j + η jki , χ  i j = eipq η pq j = eipq ε  jq,p 1  1 
3
χisj = χi j + χ ji , χiaj = χi j − χ ji . (31)
1  1  2 2
χijs = χ  i j + χ  ji , χija = χ  i j − χ  ji . (18)
2 2 By defining the corresponding higher-order stress components
Accordingly, the decomposition of the higher-order stress tensor τi(jk0) , τi(jk1) , τiasjk , and τiaa
jk
, work-conjugate to the set of higher-order de-
τ ijk , work-conjugate to the strain gradient tensor ηijk , can be written (0)
formation metrics ηi jk , ηi jk , ηiaa
(1)
, and ηiasjk , respectively, the higher-
jk
as
order stress tensor can also be expressed as
τi jk = τihjk + τi(jk1) + τijkas + τi(jk2) , (19)
τi jk = τi(jk0) + τi(jk1) + τiasjk + τiaajk , (32)
(1) (2)
where the components τihjk , τi jk , τijk
as , and τ , work-conjugate to (0) (1)
i jk where τi jk and τi jk are the trace and traceless parts, respectively,
(1) (2)
the strain gradient components, ηihjk , ηi jk , ηias
jk
, and ηi jk , respectively, split from the symmetric part of the higher-order stress tensor τisjk =
are defined as
3 (τi jk
+ τ jki + τki j ), which is work-conjugate to the symmetric part
1
1 of strain gradient tensor ηisjk . While τias and τiaa are other two inde-
τ h
i jk = δ jk τinn , (20) jk jk
3 pendent parts split from the anti-symmetric part of the higher-order
 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37 31

stress tensor τiajk = τi jk − τisjk , which is work-conjugate to the anti- depends on δ jk , and the property of the deviatoric tensor ηi jk requires
symmetric part of strain gradient tensor ηiajk . the tensor Fidjkl pq has Fissl
d = 0.
pq
In the above-presented discussions, the two sets of higher-order On the other hand, according to the symmetric and anti-
deformation metrics have been given in the form of different orthog- symmetric parts of the strain gradient tensor in Eq. (25), the higher-
onal components in Eqs. (14) and (26). The difference between them order part of strain energy density can also be written as
is that the latter three terms in the first set are independent of the di- 1 s 1
latational deformation, but the first term depends on dilatational de- wh = F ηs ηs + F a ηa ηa , (36)
2 i jkl pq l pq i jk 2 i jkl pq l pq i jk
formation; nevertheless, the second and third terms in the second set
in which the sixth-order isotropic tensors Fisjkl pq and Fiajkl pq
have
are independent of the dilatational deformation, but the first and last
terms are related to the dilatational deformation. With the help of the the major symmetry, Fisjkl pq = Fkis jl pq = Fjkil s
pq
, and the major anti-
relations εk j,i = εk j,i + 13 δ jk εnn,i , εni,n = εni,n
 + 1ε s
3 nn,i , and χi j = χi j ,
symmetry, Fiajkl pq + Fkia jl pq + Fjkil
s a = 0, in addition to obeying the sym-
pq
(1) metry of Fijklpq .
it is found that the components ηi jk and ηiasjk of the strain gradient
(1)
The sixth-order isotropic tensors having their own properties on
tensor are equal to the components ηi jk and ηias
jk
of the deviatoric the symmetries or anti-symmetries can be read as a linear combi-
(1) (1)
strain gradient tensor, respectively, that is ηi jk = ηi jk and ηias
jk
= ηiasjk . nation of their basic tensors provided in Appendix. From Eqs. (A.5)–
(1) (A.9), the sixth-order isotropic tensors, Fijklpq , Fihjkl pq , Fidjkl pq , Fisjkl pq , and
Correspondingly, the components τi jk and τias of the higher-order
jk Fiajkl pq can be expressed, respectively, using their own basic tensors,
(1)
stress tensor are equal to the components τi jk and τijk
as of the de-
as
(1) (1)
viatoric higher-order stress tensor, respectively, that is τi jk = τi jk Fi jkl pq = a1 (S1 )i jkl pq + a2 (S2 )i jkl pq + a3 (S3 )i jkl pq + a4 (S4 )i jkl pq
and τijk
as = τ as . And, hence, no differentiation will be made between
i jk + a5 (S5 )i jkl pq , (37)
ηi(jk1) and ηi(jk1) , ηias
jk
(1) (1)
and ηiasjk , τi jk and τi jk , τijk
as and τ as in the rest of
i jk
the current paper. Moreover, it is confirmed that the traceless part, Fihjkl pq = d(S3 )i jkl pq , (38)
ηi(jk1) of the strain gradient tensor is the same as provided by Lam et
al. (2003), who refer to it as the deviatoric stretch gradient. In addi- Fidjkl pq = d1 (D1 )i jkl pq + d2 (D2 )i jkl pq + d3 (D3 )i jkl pq , (39)
tion, it is also shown that all independent components of both strain
gradient tensor and the higher-order stress tensor carry the minor
symmetry of the strain gradient tensor. Fisjkl pq = b1 (K1 )i jkl pq + b2 (K2 )i jkl pq , (40)

4. Reformulation of constitutive relations Fiajkl pq = c1 (A1 )i jkl pq + c2 (A2 )i jkl pq , (41)

where an (n = 1, . . . , 5), d, dn (n = 1, 2, 3), bn , and cn (n = 1, 2) are the

4.1. Independent higher-order material constants
higher-order material constants, and the tensor Sn (n = 1, . . . , 5), S3 ,
Dn (n = 1, 2, 3), K1 and K2 , A1 and A2 are the basic tensors provided
For linear elastic isotropic materials, the total strain energy den-
in Appendix for the elasticity tensors Fijklpq , Fihjkl pq , Fidjkl pq , Fisjkl pq , and
sity in the strain gradient theory (Eq. (1)) consists of the conventional
part wc , depending on strains, and the higher-order part wh , depend- Fiajkl pq , respectively.
ing on strain gradients. Thus, the total strain energy density can be Substituting Eq. (37) into Eq. (34), the constitutive relations for the
expressed as higher-order metrics are written as
1  
w = wc + wh =
1 1
C εi j εkl + Fi jkl pq ηi jk ηl pq , (33) τi jk = Fi jkl pq ηl pq = a1 δi j ηknn + δik η jnn + 2δ jk ηnni
2 i jkl 2 4
1  
where Ci jkl = λδi j δkl + μ(δil δ jk + δik δ jl ) is the conventional elastic + a2 δi j ηnnk + δik ηnn j + a3 δ jk ηinn + a4 ηi jk
2
tensor with the Lame constants λ and μ, and Fijklpq is the sixth-
1  
order isotropic elastic tensor obeying the symmetry Fi jkl pq = Fl pqi jk + a5 ηki j + η jki , (42)
2
and, moreover, requires the symmetries Fi jkl pq = Fik jl pq = Fi jklqp be-
cause of the minor symmetry of the strain gradient tensor ηijk . The which is similar to the form presented by Mindlin (1964). Considering
resulting constitutive relations can be derived using Eq. (4), as fol- the relations,
lows: 1  ,η = δ η  1
ηnni = ηinn + εni,n i jk jk inn + ε jk,i , (43)
3 3
σi j = Ci jkl εkl = λδi j εnn + 2μεi j
it can be recognized that the terms in Eq. (42) are coupled, that is,
τi jk = Fi jkl pq ηl pq . (34) the five higher-order constants are not independent. The relationship
Because of the development of the new independent higher-order between these five constants will be found by expressing the consti-
metrics, the higher-order part wh in the total strain energy density tutive relations in terms of the components of higher-order metrics.
can be expressed in terms of the differential independent compo- From the expression of higher-order strain energy density in Eq. (35),
nents of the strain gradient tensor. On the one hand, for the hydro- the higher-order constitutive relations can be obtained as
static and deviatoric parts of the strain gradient tensor in Eq. (12), the τihjk = Fihjkl pq ηlhpq , τijk = Fidjkl pq ηlpq . (44)
higher-order part has the following form:
Another form of higher-order constitutive relations can be ob-
1 1 tained from Eq. (36), as follows
wh = Fihjkl pq ηlhpq ηihjk + Fidjkl pq ηlpq ηi jk , (35)
2 2 τisjk = Fisjkl pq ηlspq , τiajk = Fiajkl pq ηlapq . (45)
where the sixth-order isotropic tensors and Fihjkl pq
have a sym- Fidjkl pq
Applying the decomposing relations and the orthogonal proper-
metry similar to that of Fijklpq ; however, in addition, the minor sym- ties of the higher-order components, the higher-order stress can also
metric property of the hydrostatic strain gradient tensor ηihjk enforces be expressed in terms of the above-described elasticity tensors as
that the symmetry of the tensor Fihjkl pq with respect to its indices (j, k)
 
τi jk = τisjk + τiajk = Fisjkl pq + Fiajkl pq ηl pq , (46)
32 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37

  5 
τi jk = τihjk + τijk = Fihjkl pq + Fidjkl pq ηl pq . (47) τi(jk2) = a2 + a4 +
1
a5 ηi(jk2) . (57)
3 6
Thus, had the hydrostaic/deviatoric model in Eq. (44) and the
symmetric/antisymmetric model in Eq. (45), together with the gen- Therefore, the strain energy density can be rewritten as
eral five-constants one in Eq. (42), to represent a same isotropic strain 1 1 h h 1 (1) (1) 1 as as 1 (2) (2)
gradient material, that is Eq. (42) equals Eqs. (46) and (47), the rela-
w= σij εij + τijk ηijk + τijk ηijk + τijk ηijk + τijk ηijk
2 2 2 2 2
tionship between these sixth-order isotropic tensors is obtained as
1 1 5

1

follows: = λεii εjj + μεij εij + a2 + a4 + a5 ηijk
h h
ηijk
2 2 3 6
Fi jkl pq = Fisjkl pq + Fiajkl pq = Fihjkl pq + Fidjkl pq . (48) 1 1 1
 
(1) (1)
+ (a4 + a5 )ηijk ηijk + a4 − a5 ηijk ηijk
as as

By suing the relationships between these basic tensors provided in 2 2 2

Appendix, we can obtain the relationships between the higher-order 
1 5 1

(2) (2)
material constants for the strain gradient tensor as + a2 + a4 + a5 ηijk ηijk . (58)
2 3 6
2 2 1 The last two invariants of Eq. (58) are
a1 = − (a2 + a5 ), a3 = a2 + a5 , (49)
1 
3 3 6
2 s s 2
for the hydrostatic and deviatoric components as ηijk ηijk =
as as
χij χij = χijs χijs = χij χij + χij χji , and
3 3 3

d=
5 1
d1 + d2 +
1
d3 , (2) (2) 6 a a 3 
9 3 18
(50) ηijk ηijk = χ χ = χij χij − χij χji . (59)
5 ij ij 5
and for the symmetric and anti-symmetric parts as In order to evaluate the contribution of each strain gradient com-
4 6 ponent intuitively, the constants are defined as,
b1 = c1 − (b2 − c2 ). (51)  
5 5 1 5 1 1
a2 + a4 + a5 = 3μl02 , (a4 + a5 )
These relations show that any expressed form of the constitutive re- 2 3 6 2
lations using either the strain gradient tensor or its components con- 1
 1

= μl12 , and a4 − a5 = 3μl22 , (60)
tains only three independent higher-order constants. We can con- 2 2
clude that there are only three independent higher-order material
where ln (n = 0, 1, 2) are three length-scale parameters with the di-
constants involved in the general isotropic strain gradient elastic-
mension of length. Therefore, the strain energy density is rewritten
ity theory. This conclusion is different from that drawn by both Lam
as
et al. (2003) and Fleck and Hutchinson (2001). In the theory by Lam
1
et al., the symmetric character is enforced to the couple stress ten- w= kεii ε j j + με  i j ε  i j + μl02 εnn,i εmm,i + μl12 ηi(jk1) ηi(jk1)
2
sor, whereas the theory by Fleck and Hutchinson is plasticity theory  9
  9

only for incompressible materials. The demonstration procedure of + μ l22 + l02 χ  i j χ  i j + μ l22 − l02 χ  i j χ  ji ,
independent higher-order constants as presented in this section can 5 5
be similarly applied to the reformulation of the nonlinear second- (61)
gradient model proposed by Dell’lsola et al. (2009).
where k and μ are the bulk and shear modulus, respectively. The con-
stitutive relations of the general strain gradient elasticity are obtained
4.2. Constitutive relations
from Eq. (61), as follows:
According to Eqs. (34), (42) and (49), the constitutive equations of ∂w
σi j = = kδi j εnn + 2μεi j , (62)
isotropic strain gradient elasticity theories are rewritten as ∂εi j
σi j = λδi j εnn + 2μεi j , (52) ∂w
pi = = 2μl02 εnn,i , (63)
and ∂εnn,i
1   ∂w
τi jk = a2 δi j ηnnk + δik ηnn j τi(jk1) = = 2μl12 ηi(jk
1)
, (64)
2
1   ∂ηi(jk1)
− (a2 + a5 ) δi j ηkmm + δik η jmm + 2δ jk ηnni    
6 ∂w 9 9
1 1   mi j = = 2μ l22 + l02 χij + 2μ l22 − l02 χ ji , (65)
+ (4a2 + a5 )δ jk ηinn + a4 ηi jk + a5 ηki j + η jki , (53) ∂χ i j
 5 5
6 2
(1) (1)
with three independent higher-order elastic constants in addition where pi , τi jk , and mi j are work-conjugates to ɛnn, i , ηi jk , and χij , re-
to the traditional Lame constants. Further, in order to identify the spectively.
specific corresponding relationship between solids deformation and For the incompressible material in which the hydrostatic defor-
its strain gradient components, constitutive relations in the form of mation vanishes, Eq. (61) simplifies to
higher-order strain and stress components are obtained from Eq. (53),  9

according to the first set of higher-order metrics in Eqs. (14)–(24), as w= μεi j εi j + μl12 ηi(jk1) ηi(jk1) + μ l22 + l02 χij χij
5
follows:  9

5 1
 + μ l22 − l02 χij χ ji . (66)
τihjk = a2 + a4 + a5 ηihjk , (54) 5
3 6
The number of length-scale parameters is equal to that of the strain
(1) (1) gradient plasticity by Fleck and Hutchinson (1997). In the strain gra-
τijk = (a4 + a5 )ηijk , (55)
dient plasticity, the effect of the term χij χ ji in the effective plas-
 1
 tic strain is usually excluded, because no example has been identi-
τijk
as
= a4 − a5 ηijk
as
, (56) fied yet for which this invariant plays a particularly important role
2
 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37 33

 1

(Begley and Hutchinson, 1998; Fleck and Hutchinson, 2001). Here,
−Dk ni pi − ei js m is n j or ūk = uk , (73)
if we exclude any dependence on this invariant in the deformation 3
energy
√ density function for the elasticity theory by assuming l2 =
3/ 5 l0 , the number of length parameters will further reduce to two μ̄k = nl mkl − 2eslk ni τi(js1) n j nl or θ̄k = (δki − nk ni )θi , (74)
from three, resulting in the modification of the strain energy density
(42) as
1
r̄ = ni pi + ni τi(jk1) n j nk − eis j ns mi j or ε̄N = ni n j εi j , (75)
1 18 3
w = kεii ε j j + μεi j εi j + μl02 εnn,i εkk,i + μl12 ηi(jk
1) (1)
ηi jk + μl02 χij χij .
2 5 and along the edge C of the piece-wise smooth surface is
(67)  1  
(1)
f¯k = kk (ni pi ) + k j ni τ jik + e jsk ni mil nl ks n j + ni n j kl τi(jl1) nk
Thus, experimental data from two different types of micro-scale test 2
are able to independently determine the length parameters l0 and l1 , 1  

such as torsion and bending tests. − kk ei jl n j mil or ūk = uk , (76)
3

4.3. Governing equations and boundary conditions where θ i is the rotation vector. θ̄k and ε̄N are the given rotation tan-
gential to the surface and the given normal strain, respectively. New
According to Eqs. (13)–(18), the strain energy density in Eq. (61) equilibrium and boundary conditions in terms of the independent
can also be rewritten as metrics are now established for the present theory. These equations

1 9 2

4 2
 can directly reduce to the case of incompressible materials when the
w= kεii ε j j + με  i j ε  i j + μ l − l − l22 ηiik η j jk terms associated with the hydrostatic strain vanish.
2 5 0 15 1
6 4 2
 6 1 2

−μ l02 + l1 − 2l22 ηk j j ηiik + μ l02 − l1 − l22 ηkii ηk j j 5. Solutions to simple problems
5 15 5 15
1  2 
The general isotropic strain gradient elasticity theory in the form
+μ l 2 + 2l22 ηi jk ηi jk + μ l 2 − 2l22 ηki j ηi jk , (68)
3 1 3 1 of strain gradient components offers a good opportunity to identify
the contribution of each strain gradient component. In this section,
which is similar to that of Mindlin (1965), but the independent
three basic problems are considered within the framework of the
higher-order constants are three in the current theory rather than
present theory with three length-scale parameters and the simple
five. Then, the constitutive relations are derived as
theory with single parameter. The results of present three-parameter
σi j = kδi j εnn + 2μεi j , (69) theory will reveal which strain gradient component enters the cor-
responding problem and controls its size effect. Comparison of solu-
9 4 2
 
tions from the three-parameter and one-parameter versions allows
τijk = μ l02 − l − l22 δij ηnnk + δik ηnnj one to draw conclusions about the necessary of including multiple
5 15 1 material length parameters in strain gradient theory. For the simple
3 2 2
  theory without surface energy by Vardoulakis et al. (1996), the strain
− μ l02 + l − l22 δij ηkmm + δik ηjmm + 2δjk ηnni energy density has the form as
5 15 1
 
 12 2 2
 2  ws =
1 1
λεii ε j j + μεi j εi j + l 2 ληinn ηikk + μηi jk ηi jk , (77)
+μ l02 − l − 2l22 δjk ηinn + μ l12 + 4l22 ηijk 2 2
5 15 1 3
2  
which is a special case of the Mindlin theory. Obviously, the strain
+μ l12 − 2l22 ηkij + ηjki . (70) energy density (77) contains only one length-scale parameter l with
3 the dimension of length.
In addition, the equilibrium equation and boundary conditions have
been provided in Eqs. (6)–(10). Although the equilibrium equation 5.1. Torsion of bar
and boundary conditions listed in Eqs. (6)–(10) are simple and easy to
apply, the equilibrium equation and boundary conditions in the form Consider a cylindrical bar with constant radius R, whose axis is as-
of higher-order stress components are also essential for the sake of sumed to be the z-axis. When the bar twists, the displacement com-
identifying the contribution of each component. Applying the princi- ponents, same as those in classical theory, are expressed as
ple of the virtual work,
u1 = −θ yz, u2 = θ xz, u3 = 0, (78)
   
(1) (1)
σi j δεi j + pi δε,i + τi jk δηi jk + m ij δχ  ij dV = bk δ uk dV where θ is the constant angle of twist per unit length. The derived
V
   V strain is identical to that in the classical theory. The hydrostatic strain
+ t̄k δ uk + μ̄k (δkl − nk nl )δθl + r̄δεN dS + f¯k δ uk dC, (71) equals zero, and the non-zero deviatoric strains are
S C
1
 = ε  = − θ y, ε  = ε  = θ x. 1
the equilibrium equation in the volume V is given as ε13 31 23 32 (79)
2 2
1 1
σik,i − pi,ik − τi(jk,
1)
ji
− eilk mi j,l j + eil j mi j,lk + bk = 0, (72) For the three-parameter version, the dilatation gradient and the
2 3 deviatoric stretch gradient vanish, and the deviatoric rotation gradi-
as well as, the boundary conditions on the surface S are ent components, the only non-vanishing strain gradient quantities,
are
1 1
t̄k = ni [σik − pn,n δik − τi(jk,
1)
− e m + e m δ ] 1 1
 = − θ , χ  = − θ , and χ  = θ .
j 2 lik l j, j 3 l js ls, j ik χ11 (80)
 1
 222 33
2
+(D p n p ) nk ni pi + ni n j τi(jk1) + ni τi(js1) n j nk ns − ei js nk n j m is From Eq. (61), the strain energy density can be obtained as
3
1   
 (1) (1) 1
− Ds e ni n m n j + ni τsik + ni τi js n j nk w= μθ 2 R2 + 6l22 . (81)
2 jsk l il 2
34 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37

Fig. 2. Simple shear problem.

larger than h. The Cartesian coordinate system shown in the fig-

ure is adopted in the formulation below. The only non-vanishing x-
component u(y) of the displacement vector is induced by a shear
Fig. 1. Size effect of the normalized torsion rigidity.
force P acting on the top surface y = h. The hydrostatic strain equals
zero, and the non-zero deviatoric strains are

 = ε = 1 du
On the basis of the work-energy principle, εxy yx . (87)
  2 dy
L
1
wdAdz = T θ L, (82) In the three-parameter version, the dilatation gradient vanishes.
0 A 2
The non-zero strain gradient quantities are the deviatoric stretch gra-
the twist angle is derived as dient and the deviatoric rotation gradient, which are given as
T 1 d2 u (1) 4 d2 u (1)
θ=  , (83) (1)
η111 =− (1)
, η221 = η212 (1)
= η122 = (1)
, η331 = η313 (1)
= η133
μ Ip + 6Al22 5 dy2 15 dy2
where T denotes the torque, L denotes the length of the bar, A denotes 1 d2 u
=− , (88)
the area of cross section, and Ip denotes the polar moment of inertia 15 dy2
of the cross section. Therefore, the torsion rigidity is obtained from
the above relation as  =− 1 d2 u
  χ32 . (89)
2 dy2
S = μ Ip + 6Al2 . 2
(84)

For the one-parameter version, the torsion rigidity is given by The work-conjugated stress and higher-order stress quantities ob-
  tained from Eqs. (62)–(65) are listed as
S = μ Ip + 2Al . 2
(85)
du
Considering the torsion rigidity for the classical theory, Sc = μI p , σxy = σyx = μ , (90)
dy
the torsion rigidities for both versions can be normalized as
 
9 2 d2 u 9
 d2 u

S l2 m32 = −μ l22 + l0 , m23 = −μ l22 − l02 ,
= 1 + 12 ∗2 , (86) 5 dy 2 5 dy2
(91)
Sc R
where l∗ depends on the formulation according to l∗2 = l22 for the
three-parameter version and l∗2 = 1/3 l 2 for the one-parameter ver- (1) 2 d2 u (1) (1) (1) 8 d2 u (1) (1)
τ111 = − μl12 2
, τ221 = τ122 = τ212 = μl12 2 , τ331 = τ313
sion. 5 dy 15 dy
From the formulation of the three-parameter version, one knows 2 d2 u
(1)
that only the length parameter l2 determines the size effect in bar = τ133 =− μl12 2 . (92)
15 dy
torsion, and the other length parameters l0 and l1 do not enter this
problem. The contribution of the three-parameter version to the size Substituting Eqs. (90)–(92) into Eq. (72) leads to the governing
effect of the torsion rigidity is 6μAl22 , while the simple version has equation,
2μAl2 as its contribution to the torsion rigidity. The results from the
d4 u d2 u
two theories in the form of normalized torsion rigidities as a func- g2 4
− = 0, (93)
tion of the ratio of radius to length parameter are compared in Fig. 1. dy dy2
It is seen that the size-dependencies of normalized torsion rigidities in which g2 = 15 8 2
l1 + 12 l22 + 10
9 2
l0 . According to the boundary condi-
from gradient effects predicted in the two versions are obviously dif- tions of the current problem
ferent. Identical gradient effects from the two versions would be ex-
pected when l 2 = 3l22 . Nevertheless, the factor is strongly problem- du(h) d3 u(h) P d2 u(h)
u(0) = 0, εxy (0) = 0, − g2 = , = 0,
dependent, which will be seen in the following shear and pure bend- dy dy3 μbL dy2
ing problems.
(94)

5.2. Shearing of fixed-end layers the solution of Eq. (93) can be readily obtained as


Consider a block of width b, length L and height h undergoing P sinh[(1 − ξ ) g ] h
h
u(y) = [y + g   − g tanh ], (95)
a shear deformation, as shown in Fig. 2. Assume b and L are much μbL cosh h g
g
 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37 35

Fig. 3. Distribution of normalized shear strain with different characteristic size.

Fig. 4. Size effect of normalized bending rigidity.

where the normalized coordinate ξ = y/h. The non-vanishing strain,

hence, is found to be where R denotes the radius of curvature of the central axis of the
  beam after bending in the xz-plane. Then, the strains can be written
1 P cosh (1 − ξ ) g h
as
εxy = 1−   , (96)
2 μbL cosh hg z v v
ε11 = , ε22 = − z, ε33 = − z. (100)
R R R
and the corresponding strain for the one-parameter version is For the three-parameter version, the high-order deformation
  quantities are given as
1 P cosh (1 − ξ ) h
εxy = 1−   l
. (97) 1
2 μbL cosh hl εnn,3 = (1 − 2v), (101)
R
Considering the non-vanishing strain in the classical theory, εxy
c =
11+v 21+v
1 P  =
χ12 ,  =
χ21 , (102)
2 μbL , the non-vanishing strains in the two versions can be normal- 3 R 3 R
ized as
(1) 11
εxy cosh[(1 − ξ ) lh∗ ] η333 =− (1 + v), (103)
= 1 −   , (98) 5R
εxy
c
cosh lh ∗

where l∗ depends on the formulation according to l∗ = g for the three- (1) (1) (1) 4 1 (1) (1) (1)
η113 = η131 = η311 = (1 + v), η223 = η232 = η322
parameter version and l∗ = l for the one-parameter version. 15 R
The distribution of the normalized shear strain along the height of 1 1
=− (1 + v). (104)
the block in Eq. (98) is shown in Fig. 3 for various values of h/l∗ . The 15 R
size-dependent normalized shear strain is seen in the shear problem. From Eq. (61), the strain energy density can be obtained as
Size effects of shear strain predicted in the two formulations are ex-   1  
pected to be identical when l 2 = 15 8 2
l1 + 12 l22 + 10
9 2
l0 . Thus, the corre- 1 1 4 2
w= Ez2 + 2μ l02 (1 − 2v)2 + l02 + l + l22 (1 + v)2 .
sponding factor is 1/2 if l0 = l1 = 0, which is different from the fac- 2 R2 5 15 1
tor in the torsion problem. One knows that the equivalent relations (105)
in the torsion and shear problems are strongly problem-dependent
On the basis of the work-energy principle,
rather than only the deformation of materials. If each of the two for-
 L 
mulations were to be calibrated by the torsion problem such that 1 L
l 2 = 3l22 , then at least one of the versions must clearly be signifi-
wdAdx = M , (106)
0 A 2 R
cantly erroneous in predicting shear problem. This strong problem
the curvature of the central axis of the beam after bending is obtained
dependence of the size effect is the essence underlying the necessity
as
of the multiple-parameter theory to predict the size effect in micro-
structures. Here, we arrive at the same conclusion as in the plastic 1 M
= , (107)
deformation theory (Fleck and Hutchinson, 2001). R D
where M denotes the bending moment and D denotes the bending
5.3. Pure bending of thin beams rigidity expressed as
 1 4 2

Consider a rectangular beam with height h, width b and length D = EI + 2μA l02 (1 − 2v)2 + l02 + l + l22 (1 + v)2 , (108)
5 15 1
L. The Cartesian coordinate system is used. We assume the x-axis to
where E is the Young’s modulus, v is the Poisson ratio, and I is the
coincide with the centerline of the beam and the other axes parallel
moment of inertia. The corresponding bending rigidity for the one-
to the sides of the cross section. The displacement components are
parameter version is
1 v v  1 2
u1 = xz, u2 = − yz, u3 = y2 − z 2 − x , (99) D = EI + EAl 2 . (109)
R R 2R 2R
36 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37

The bending rigidities, including gradient effects, in the two ver- where Aijklpq is the sixth-order elasticity tensor. The property required
sions are normalized using the bending rigidity of the classical the- by isotropic materials implies the elasticity tensor to conform
ory, EI, as follows
Ai jkl pq = Ahmnrst Qhi Qm j Qnk Qrl Q ps Qtq , (A.2)
D l2
= 1 + 12 ∗2 , (110) for every orthogonal transformation Qij . The tensor Aijklpq can be read
EI h
as a linear combination of sixth-order isotropic components Tn (n =
2
where l∗2 depends on the formulation according to l∗2 = l02 (1−2
1+ν
v)
+ 1...15), which are called as basic tensors, given by (Monchiet and Bon-
net, 2011; Suiker and Chang, 2000)
( 15 l02 +
15 l1 + l2 )(1 + v) for the three-parameter version and l∗ = l
4 2 2 2 2

for the one-parameter version. The size-dependent bending rigid- (T1 )i jkl pq = δi j δkl δ pq , (T2 )i jkl pq = δi j δkp δlq , (T3 )i jkl pq = δi j δkq δl p ,
ity of pure bending of a thin beam with rectangular cross-section is
(T4 )i jkl pq = δik δ jl δ pq , (T5 )i jkl pq = δik δ jp δlq , (T6 )i jkl pq = δik δ jq δl p ,
shown in Fig. 4. It is seen that only if the parameters in the three- and
2 (T7 )i jkl pq = δil δ jk δ pq , (T8 )i jkl pq = δil δ jp δkq , (T9 )i jkl pq = δil δ jq δkp ,
one-parameter versions hold the equivalent relation l 2 = l02 (1−2
1+ν
v)
+
( 15 l02 + 15
4 2
l1 + l22 )(1 + v), the two versions will predict the same size
(T10 )i jkl pq = δip δ jk δlq , (T11 )i jkl pq = δip δ jl δkq , (T12 )i jkl pq = δip δ jq δkl ,
effects for the pure bending problem. The equivalent relation is differ- (T13 )i jkl pq = δiq δ jk δl p , (T14 )i jkl pq = δiq δ jl δkp , (T15 )i jkl pq = δiq δ jp δkl .
ent from the bar torsion and shear problem. The different equivalent (A.3)
relation between the three- and one-parameter versions in the pure
bending problem further shows the necessity of using the multiple- If the sixth-order tensor Aijklpq satisfies some special symmetry,
parameter theory to predict size effects. then its basic tensors can be simplified. For a tensor having the
major symmetries, Ai jkl pq = Al pqi jk , only 11 basic tensors denoted
Mn (n = 1...11) are needed. They are given by
6. Conclusion
1 1
M1 = (T1 + T13 ), M2 = (T2 + T6 ), M3 = T3 , M4
By applying two sets of orthogonal decompositions of the 2 2
strain gradient tensor, the general isotropic strain gradient elastic- 1
= (T4 + T10 ), M5 = T5 M6 = T7 , M7 = T8 , M8 = T9 ,
ity theory with only three independent length-scale parameters is 2
reformulated. In the reformulated frame, the total deformation en- 1
ergy density is a function of four parts: the symmetric strain tensor, M9 = T11 , M10 = (T12 + T14 ), M11 = T15 . (A.4)
2
the dilatation gradient vector, the deviatoric stretch gradient tensor,
Furthermore, if a tensor has the additional minor symmetries,
and the deviatoric curvature tensor. The independent strain gradient
Ai jkl pq = Aik jl pq = Ai jklqp , in addition to the major symmetries, the
parts and the corresponding work-conjugated stress tensors are de-
number of basic tensors reduces from 11 to five. These basic tensors
fined. After strict derivations, the constitutive equations, equilibrium
are
equation and boundary conditions are obtained subsequently. In ad-
dition, the deformation energy density, equilibrium equations, and 1 1
S1 = (T1 + T10 + T13 + T4 ), S2 = (T2 + T3 + T5 + T6 )
boundary conditions can directly reduce to the case of incompress- 4 4
ible materials by assuming the dilatation gradient to be zero. 1 1
S3 = T7 , S4 = (T8 + T9 ), S5 = (T11 + T12 + T14 + T15 ). (A.5)
Three simple examples are studied based on the present strain 2 4
gradient elasticity theory. The results reveal that the dilatation gra-
However, a case of the minor symmetries depending on a Kro-
dient, the deviatoric stretch gradient, and the deviatoric rotation gra-
necker symbol needs only the basic tensor S3 ; whereas, for a case
dient control different higher-order deformation in isotropic solids.
that the components contracting two minor symmetric indices van-
The nonvanishing higher-order deformation in torsion is the devia-
ish, only three basic tensors are needed, given as
toric rotation gradient. The dilatation gradient vanishes in the shear
problem. All the higher-order deformations enter the pure bending 2 1 1 1 2
problem. By comparing the present solutions of three examples with D1 = S2 − S1 + S3 , D2 = S4 − S3 , D3 = S3 + S5 − S1 .
3 9 3 9 3
those of the simple strain gradient theory containing only one higher- (A.6)
order length-scale parameter, the necessity of including more than
one length-scale parameters in the strain gradient theory to consider When a tensor has either the additional symmetric properties
size effects in the micron scale has been explained. The present strain Ai jkl pq = Aki jl pq = A jkil pq , or the additional anti-symmetric properties
gradient theory can provide effective descriptions of size-dependent Ai jkl pq + Aki jl pq + A jkil pq = 0, in addition to the major and minor sym-
behaviors in wide micron-scale isotropic elastic problems. metries, only two basic tensors are needed. They are
1
K1 = (4S1 + S3 + 4S2 )
Acknowledgments 9
1
This study is funded by the National Natural Science Foundation
= (T1 + T10 + T13 + T4 + T7 + T2 + T3 + T5 + T6 ), (A.7)
9
of China (11272186), Specialized Research Fund for the Doctoral Pro-
and
gram of Higher Education of China (20120131110045), and the Natu-
ral Science Fund of Shandong Province of China (ZR2012AM014). 1
K2 = (2S4 + 4S5 )
3
1
Appendix. The basic tensors of sixth-order isotropic tensors = (T8 + T9 + T11 + T12 + T14 + T15 ), (A.8)
3
In the higher-order elasticity theory, both the theories by Mindlin for the additional symmetric conditions, and
and Eshel (1968) and Dell’lsola et al. (2009), the constitutive equa-
4
tions for linear isotropic materials are a linear relation between two A1 = (S3 + S2 − 2S1 )
9
third-order tensors, denoted by aijk and bijk , which is expressed as
1
= (4T7 +T2 + T5 + T3 + T6 − 2T1 − 2T4 − 2T10 − 2T13 ), (A.9)
ai jk = Ai jkl pq bl pq , (A.1) 9
 S. Zhou et al. / International Journal of Solids and Structures 80 (2016) 28–37
and
4
A2 = (S4 − S5 )
3 disclinations in first strain gradient elasticity. Int. J. Eng. Sci. 43, 1157–1184.
1 Li, S., Miskioglu, I., Altan, B.S., 2004. Solution to line loading a semi-infinite solid in
= (2T8 + 2T9 − T11 − T12 − T14 − T15 ) (A.10)
3 Ma, Q., Clarke, D.R., 1995. Size dependent hardness of solver single crystals. J. Mater.
for the additional anti-symmetric conditions. Finally, these sixth-
order tensors with special symmetry can be read as a linear combi-
nation of their basic tensors.
References
Aifantis, E., 1992. On the role of gradients in the localization of deformation and frac-
ture. Int. J. Eng. Sci. 30, 1279–1299.
Askes, H, Aifantis, E, 2011. Gradient elasticity in statics and dynamics: An overview of
formulations, length scale identification procedures, finite element implementa-
tions and new results. Int. J. Solids Struct 48, 1962–1990.
Auffray, N., Quang, H.L., He, Q.C., 2013. Matrix representations for 3D strain-gradient
elasticity. J. Mech. Phys. Solids 61, 1203–1223.
Bacca, M., Bigoni, D., Corso, F.D., Veber, D., 2013. Mindlin second-gradient elastic prop-
erties from dilute two-phase Cauchy-elastic composites. Part I: Closed form ex-
pression for the effective higher-order constitutive tensor. Int. J. Solids Struct. 50,
4010–4019.
Begley, M.R., Hutchinson, J.W., 1998. The mechanics of size-dependent indentation. J.
Mech. Phys. Solids 46, 2049–2068.
Dell’lsola, F., Sciarra, G., Vidoli, S., 2009. Generalized Hooke’s law for isotropic second
gradient materials. R. Soc. Lond. Proc. Ser. A 465 (2107), 2177–2196.
Exadaktylos, G., 1998. Gradient Elasticity with surface energy: Mode-I Crack Problem.
Int. J. Solids Struct. 35, 421–456.
Fleck, N.A., Hutchinson, J.W., 1993. A phenomenological theory for strain gradient ef-
fects in plasticity. J. Mech. Phys. Solids 41, 1825–1897.
Fleck, N.A., Muller, G.M., Ashby, M.F., Hutchinson, J.W., 1994. Strain gradient plasticity:
theory and experiment. Acta Metall. Mater. 42, 475.
Fleck, N.A., Hutchinson, J.W., 1997. Strain gradient plasticity. In: Hutchenson, J.W.,
Wu, T.Y. (Eds.), Advances in Applied Mechanics, vol. 33. Academic Press, New York,
p. p. 295-361.
Fleck, N.A., Hutchinson, J.W., 2001. A reformulation of strain gradient plasticity. J. Mech.
Phys. Solids 49, 2245–2271.
Guo, X.H., Fang, D.N., Li, X.D., 2005. Measurement of deformation of pure Ni foils by
speckles pattern interferometry. Mech. Eng. 27, 22–25 (in Chinese).
Hadjesfandiari, A.R., Dargush, G.F., 2011. Couple stress theory for solids. Int. J. Solids
Struct. 48, 2496–2510.
Koiter, W.T., 1964. Couple stresses in the theory of elasticity I and II. Proc. K. Ned. Akad.
Wet. (B) 67, 17–44.
37
Lam, D.C.C., Yang, F., Chong, A.C.M., 2003. Experiments and theory in strain gradient
elasticity. J.Mech. Phys. Solids 51, 1477–1508.
Lazar, M., Maugin, G.A., 2005. Nonsingular stress and strain fields of dislocations and
gradient elasticity. Int. J. Solids Struct. 41, 3395–3410.
Res. 10, 853–863.
Mindlin, R.D., Tiersten, H.F., 1962. Effects of couple stresses in linear elasticity. Arch.
Rat. Mech. Anal. 11, 415–448.
Mindlin, R.D., 1964. Micro-structure in linear elasticity. Arch. Rat. 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., Eshel, N., 1968. On first strain-gradient theories in linear elasticity. Int. J.
Solids Struct. 4, 109–124.
Monchiet, B.V., Bonnet, G., 2011. Inversion of higher order isotropic tensors with minor
symmetries and solution of higher order heterogeneity problems. Proc. R. Soc. A.
467 (2126), 314–332.
Mühlich, U., Zybell, L., Kuna, M., 2012. Estimation of material properties for linear elas-
tic strain gradient effective media. Eur. J. Mech. A-Solid. 31, 117–130.
Polizzotto, C., 2012. A gradient elasticity theory for second-grade materials and higher
order inertia. Int. J. Solids Struct. 49, 2121–2137.
Polizzotto, C., 2013. A second strain gradient elasticity theory with second velocity gra-
dient inertia - Part I: Constitutive equations and quasi-static behavior. Int. J. Solids
Struct. 50, 3749–3765.
Stolken, J.S., Evans, A.G., 1998. A microbend test method for measuring the plasticity
length scale. Acta Mater. 46, 5109.
Suiker, A.S.J., Chang, C.S., 2000. Application of higher order tensor theory for formulat-
ing enhanced continuum models. Acta Mech. 142, 223–234.
Tang, C, Alici, G, 2011. Evaluation of length-scale effects for mechanical behaviour of
micro- and nanocantilever: I. Experimental determination of length-scale factors.
J. Phys. D: Appl. Phys. 44, 335501.
Tang, C., Alici, G., 2011. Evaluation of length-scale effects for mechanical behaviour of
micro- and nanocantilevers: II. Experimental verification of deflection models us-
ing atomic force microscopy. J. Phys. D: Appl. Phys. 44, 335502.
Toupin, R.A., 1962. Elastic materials with couple-stresses. Arch. Rat. Mech. Anal. 11,
385–414.
Tekoglu, C., Onck, P.R., 2008. Size effects in two-dimensional Voronoi foams: a compar-
ison between generalized continua and discrete models. J. Mech. Phys. Solids 56,
3541–3564.
Vardoulakis, I., Sulem, J., 1995. Bifurcation Analysis in Geomechanics. Blackie/Chapman
and Hall, London.
Vardoulakis, I., Exadaktylos, G., Aifantis, E., 1996. Gradient elasticity with surface en-
ergy : mode-III crack problem. Int. J. Solids Struct. 33, 4531–4559.
Yang, F., Chong, A.C.M., Lam, D.C.C., Tong, P., 2002. Couple stress based strain gradient
theory of elasticity. Int. J. Solids Struct. 39, 2731–2743.