Composites Science and Technology 113 (2015) 1930

Contents lists available at ScienceDirect

Composites Science and Technology

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

Novel composites with asymmetrical elastic wave properties

Lang-Quan Shui , Zhu-Feng Yue, Yong-Shou Liu, Qing-Chang Liu, Jiao-Jiao Guo, Xin-Dang He
Department of Engineering Mechanics of Northwestern Polytechnical University, Xian 710129, China

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

Article history: The traditional structural composites are known as materials with a spatial order. The temporal order is
Received 26 June 2014 introduced into structural composites in this paper. We theoretically describe and numerically
Received in revised form 7 February 2015 demonstrate the architecture for a structural composite which exhibits tunable and nearly full-banded
Accepted 10 March 2015
asymmetrical elastic wave property. Given its spatiotemporal order and unconventional wave properties,
Available online 19 March 2015
such composite can be named as spatiotemporal metamaterial (STMM). Firstly, the effective material
parameters, which include some new material parameters, are obtained based on improved multi-scale
Keywords:
homogenization techniques. Secondly, the unconventional wave properties are theoretically analyzed. It
A. Structural composites
A. Smart materials
is found that the elastic wave group velocity surfaces do not satisfy the centrosymmetry in STMM, i.e. the
B. Mechanical properties time-reversal symmetry of linear elastic wave is violated. In certain instances, the waves even propagate
Asymmetrical elastic wave unidirectionally. Finally, we present an example of the STMMs, which includes homogenized calculations
and numerical simulations where the shear waves are forbidden in some directions. The unconventional
wave properties of the STMMs proposed in this paper own attractive potential applications, such as
acoustic absorbing, acoustic information processing, energy-saving, and energy-harvesting.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Metamaterials are articially engineered composite structures
that exhibit unconventional wave properties. Originating from
the microwave technology [1], unprecedented applications of
metamaterials have been expanded from optics to mechanics,
acoustics, thermodynamics and other elds [2], which makes this
research topic become one of the fastest growing subdisciplines
of material science [3].
While optics and acoustic cloaking are attractive applications of
metamaterials, the study on metamaterials bears far more
signicance. Leonhardt [4] claimed that perfect geometrical optical
invisibility could be possible when using modern metamaterials,
and also pointed out that the corresponding method could be
applied to other electromagnetic waves or sound. Al and
Engheta [5] discussed the possibility of using plasmonic and meta-
material covers to make spherical and cylindrical objects nearly
invisible. Zhang et al. [6] presented the rst practical realization
of a metamaterial cloak for underwater ultrasound. Popa et al. [7]
proposed a class of two-dimensional acoustic cloakings, which is
easily achievable in practice by using metamaterials. Broadly
speaking, the structure with asymmetrical wave properties
belongs to metamaterials [8]. Achieving asymmetrical wave
Corresponding author. Tel.: +86 15902902159.
E-mail address: shuilq061341@gmail.com (L.-Q. Shui).
properties has become one of the hot topics in the metamaterial
research. The work of Terraneo et al. [9] initiated the possibility
to propose a thermal rectier. Wang and Li [10] established ther-
mal logic gates and discussed the possibility of nanoscale experi-
ments. Fan et al. [11] demonstrated a passive optical diode with
a high transmission ratio based on optical nonlinearity, which
might lead to a revolution in modern information processing.
Liang et al. [12,13] demonstrated a one-dimensional acoustic diode
and achieved acoustic rectication, in which it is emphasized that
such a device could be applied in biomedicine. Nesterenko et al.
[14] and Liang indicated the applications of acoustic diodes in
the areas of energy-saving and energy-harvesting. Boechler et al.
[15] demonstrated a new mechanism for tunable rectication,
which uses bifurcations and chaos, in which the mechanism was
claimed to be applicable to design advanced photonic, thermal
and acoustic materials and devices. Lepri and Casati [16] designed
a diode chain in which the waves propagate with the same
amplitude but different frequencies in opposite directions. In
addition, Silva et al. [17] introduced the concept of metamaterial
analog computing and pointed out that properly designed
metamaterials can perform mathematical operations. Mei et al.
[18] presented a thin-lm acoustic metamaterial which can com-
pletely absorb the selected low-frequency airborne sound. Popa
et al. [19] demonstrated an acoustic metamaterial with effective
tunable material parameters including negative material

http://dx.doi.org/10.1016/j.compscitech.2015.03.007
0266-3538/ 2015 Elsevier Ltd. All rights reserved.
20 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930
parameters. In general, numerous studies about metamaterials
have been published as far.
Despite signicant progress, metamaterials still face challenges,
such as passivity, narrowband and high loss [19,20]. The desired
metamaterials should be fully dynamic, broad bandwidth, loss-free
and truly 3D [3,21,22]. Inspired by the work of Li et al. [23],
Wilczek [24], Shapere and Wilczek [25] and Bergamini et al. [26]
(see Section 2), a temporal order is introduced into metamaterials,
which shows surprisingly promising outcome.
Specically, an architecture is proposed for a structural com-
posite with time-varying unit cells. With some new material
parameters describing the mechanical properties, the effective
material parameters are obtained based on improved multi-scale
homogenization techniques. The result reveals that the coupling
of the spatial and temporal order leads to asymmetric wave
phenomenon. In such material, the elastic waves propagate with
different group velocities in opposite directions and even
propagate unidirectionally, i.e. the time-reversal symmetry is
violated. Obviously, its group velocity surfaces do not satisfy
centrosymmetry. With the time-reversal asymmetry, such a wave
property could be named as spatiotemporal anisotropy (STA).
With these unconventional wave properties, the material might
be proposed for applications in acoustic absorbing, acoustic
information processing, energy-saving, energy-harvesting and so
on [14].
Considering the spatiotemporally ordered mesoscopic struc-
ture, and unconventional wave properties in such a material, we
name it STMM. The temporal variability of the unit cells can be
achieved by using the materials with controllable properties (e.g.
smart materials. See Section 2), which fundamentally enable the
tunability of metamaterials. Thus, the wave properties can be
modied articially. In addition, the homogenization techniques
require much longer wavelength with respect to the scale of the
unit cell, indicating that the wave frequency would fall into any
desirable range if the unit cells are properly small. Overall, the
STMM would be a tunable and nearly full-banded metamaterial.
2. The proposal of STMMs
First, we shall mention the work that inspired us for this pro-
posal. Li et al. [23], Wilczek [24], Shapere and Wilczek [25] pro-
posed and explored the concept of crystals with spatial and
temporal order in 2012, respectively. Coleman [27] highly praised
their work and pointed out that their work might initiate a brand
new eld of research. Bergamini et al. [26] proposed a phononic
crystal with adaptive connectivity in 2013. This phononic crystal
might involve an embryonic idea of STMM. In this phononic crys-
tal, the unit cell has locally controlled connectivity. Bergamini
claimed that a fairly small amount of adaptive material can
produce remarkable effects due to the exploitation of system
periodicity.
In parallel, we propose an architecture with time-varying unit
cells by introducing the temporal order into metamaterials. The
temporal variability of the unit cells can be achieved by
controllable properties of the materials. Such materials can be
the piezoelectric material and electro-rheological (ER) materials
experiencing time-varying electric eld [28,29], the magneto-
strictive material and magneto-rheological uid experiencing
time-varying magnetic eld [30,31], and the shape-memory
material experiencing time-varying temperature eld [32]. Popa
et al. [19] pointed out that the active response of the controllable
properties can be instantaneously achieved because of low group
velocities of the mechanical waves. The so-called time-varying
unit cells may be abstract and contrary to the conventional
cognition. For a better understanding of the architecture, the
implementation of a STMM in Section 5 and some general analyses
of time-varying property in Appendix A can be referred to. Finally,
we want to point out that the controllability of the unit cells
fundamentally enables the tunability of the STMM.
3. The dynamic multi-scale homogenization in STMMs
In the exploration of the material property of structural com-
posites, the elastic wave whose wavelength is much longer than
the unit cell scale is considered. To obtain the macroscopic con-
stitutive relations of an architecture, the effective material
parameters should be calculated. It is well known that
computational homogenization is a particularly suitable technique
to calculate the effective material parameters of structural
composites.
Computational homogenization, as a powerful technique to
solve partial differential equations with rapidly oscillating coef-
cients, has been widely applied in micromechanics [3336]. In
STMMs, the meso-structure exhibits a spatial and temporal order.
Consequently, the corresponding elastic wave equations have
rapidly oscillating coefcients with spatial and temporal periodi-
city. Casado-Diaz et al. [37] have proved the feasibility to apply
the multi-scale homogenization technique to such wave equations.
Dong and Cao [38] have also applied the technique to similar
equations.
A linearized Hookean body is considered in this paper. Let
ui x1 ; x2 ; x3 ; t, i = 1, 2, 3, describe the displacement of a particle
located at x1 ; x2 ; x3 at time t from its position in the natural state.
The innitesimal strain tensor is expressed as
 
1 @ui @uj
eij 1
2 @xj @xi
In a small deformation framework, the particle velocity can be
expressed as
@ui
vi 2
@t
The corresponding stress and momentum are expressed as
rij kijmn emn 3
pi qv i 4
where kijkl is the adiabatic elastic tensor which satises Voigt
symmetry, and q is the density. The elastic tensor and density are
periodic functions of space and time. For isotropic materials, the
elastic tensor kijmn kdij dmn ldim djn din djm , where k and l are
Lame constants, and dij is the Kronecker symbol.
The wave equation of the linearized Hookean body is written as
@pi @ rij
fi 5
@t @xj
where f i is the volume force. Based on Eqs. (1)(5), the wave
equation can be expressed as
   
@ @u @ @um
q i fi kijmn 6
@t @t @xj @xn
The homogenization of Eq. (6) usually begins with separation of
variables [25,39]. The frequency invariance makes it possible to
separate the temporal variable and spatial variables. However,
for the material we proposed, the time-varying property breaks
the frequency invariance (Appendix A). Thus we adopt new ideas
to homogenize Eq. (6), where the wave equation can be rewritten
as
 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930 21

 
@ @um Substituting Eq. (9) into Eq. (11), we get
kiamb fi 7
@xa @xb 1 0 1 
A um Ai0 u1 0
m Ai1 um
In this paper, some conventions are made about the indices of e2 i0 e 
the tensors. The Einstein summation convention is used. The Ai0 u2 1 0
m Ai1 um Ai2 um
Latin indices take values 1, 2, 3. The Greek indices take values 1, fi 12
2, 3, 4, and the 4 corresponds to the time coordinate. Thus the
tensor xa represents x1 ; x2 ; x3 ; x4 and x4 t. For isotropic materials, where the higher-degree terms with respect to e are neglected. Eq.
the tensor kiamb kdia dmb ldim dab dib dam  l qda4 db4 dim (12) gives us three identical equations according to the power of e:
can be named as elastic-density tensor. When a; b4; kiamb degen- Ai0 u0
m 0 13
erates to the adiabatic elastic tensor. When a; b 4 and
i m; kiamb degenerates to the density. Otherwise, kiamb 0. An Ai0 u1 0
m Ai1 um 0 14
index variable and its underlined form take the same value.
When an index variable and its underlined form appear once in Ai0 u2 1 0
m Ai1 um Ai2 um f i 15
one term respectively, the index can be regarded as a free index.
For example, in Eq. (29), .i Hi4i4 represents .1 H1414 ; By direct integrating Eq. (13), we have
.2 H2424 ; .3 H3434 . 0
um x; y u0
m x 16
In this paper, the material properties are considered to be per-
0
iodic in time and space. A basic duplicable unit of the material Eq. (16) indicates that um does not depend on y, which implies
property framework in time and space is named as spatiotemporal that
0
um should be the homogenized solution of Eq. (7). We may
unit cell. The four-dimensional space that is occupied by a unit cell 1 2 0
eliminate um and um by Eqs. (14) and (15) to obtain um .
is expressed as T4 . The characteristic length of the spatiotemporal
Specically, substituting elimination method is used. After ana-
unit cell is dened as the meso-scale of the material structure. The
lyzing Eqs. (14) and (15), the solving process can be divided into
meso-scale is much less than the macro-scale of the material struc- 1
two steps. In the rst step, um can be expressed as a function of
ture and expressed as e. The elastic-density tensor kiamb kiamb y 0
is a period function of y, where y x=e x1 ; x2 ; x3 ; x4 =e. The um according to Eq. (14). As for the second step, we obtain the
R 0 1 2
integration T4 dy is dened as the spatiotemporal volume of a unit equation of um by substituting um into Eq. (15) and eliminating um
cell, and the dimension without periodicity does not participate in by an implied condition. Detailed derivations are shown as follows.
the integration. The purpose of the dynamic homogenization is to
study the limit of ui when e ! 0 and to nd out an wave equation Step 1: By Eq. (16), Eq. (14) can be simplied as
that is not related to y. " #
1 0
Based on above assumptions, a small parameter e appears in Eq. @ @um x; y @kiamb y @um x
kiamb y  17
(7). Thus we can seek a power series solution of Eq. (7) about e: @ya @yb @ya @xb
0 1 2 By separation of the variables, the solution of Eq. (17) can be
ui ui eui e2 ui . . . 8
expressed as
Assume that the items in Eq. (8) explicitly depend on slow vari- 0
@um x
mb
ables xa and fast variables ya xa =e. As the coefcient of Eq. (7), i.e. u1
r x; y vr y 18
@xb
the elastic-density tensor kiamb , is a periodic function of y, all items
of the power series solution (Eq. (8)) should also be periodic where vmb
1
r y stands for the so-called rst-order corrector. As ur is
functions of y. Thus Eq. (8) can be assumed as periodic with respect to y; vmb should be periodic with respect to y
r
0 1 2 according to Eq. (9). Substituting Eq. (18) into Eq. (17), we obtain an
ui x ui x; y eui x; y e2 ui x; y . . . 9
equation as
0 1 " #
where ui ; ui ; . . . are periodic with respect to y. In the homogeniza- @ @ vmb y @kiamb y
tion, the meso- and macro-scales are separated, hence x and y are kiarc y r  y 2 T4 19
@ya @yc @ya
regarded as independent variables, which indicates that the partial
derivative with respect to xa becomes Considering that vmb is periodic with respect to y, the boundary
r
conditions of Eq. (19) should be periodic. It is obvious that the
@ 1 @ coefcient kiarc is a function of the variable y only. Thus Eq. (19)
10
@xa e @ya
has no oscillating coefcient in T4 . Hence Eq. (19) can be easily
Replacing the differential operator in Eq. (7) by Eq. (10), and solved by numerical software. The problem of solving vmb r y is

considering homogeneous boundary conditions, we get the so-called unit cell problem, which is the core of the
homogenization.
( 
1
e2 Ai0 1e Ai1 Ai2 um f i ; x; y 2 X  T4
11 Step 2: Eq. (15) can be rewritten as
um 0; x; y 2 @ X  T4 ; X  R4
Ai0 u2 1 0
m Ai1 um  Ai2 um f i 20
where the operators Ai0 ; Ai1 ; Ai2 are in the forms of
Ai0 @kiamb y@=@yb =@ya ; Ai1 @kiamb y@=@xb =@ya @kiamb y It can be proved that the average value of the right hand side of
@=@yb =@xa ; Ai2 @kiamb y@=@xb =@xa ; X is an open set with Eq. (20) in a unit cell must be zero to maintain the posedness of Eq.
smooth boundaries. It should be noted that arbitrarily chosen (20). Thus
spatiotemporal volume of the unit cell with a constant shape does 1
hAi1 um Ai2 u0
m i fi 21
not affect the homogenized results. For convenience, the R
spatiotemporal volume of the unit cell is set as one (see where the averaging operator hi T4 dy, which involves the
Fig. 3(c) and (d)) in homogenizing. setting that the spatiotemporal volume of the unit cell is one. By
22 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930

substituting Eqs. (16) and (18) into Eq. (21) and tedious The discussion below mainly focuses on a monochromatic
simplications, we nally get the homogenized equation: elastic wave. The solution with frequency x can be expressed as
0
hp i
@ 2 um 0
ui U i exp 1kj xj  xt 30
Hiamb fi 22
@xa @xb
p
where U i characterizes the polarization direction, 1 represents
where the homogenized material parameters are expressed as
* + the imaginary unit, and kj is the component of the wave vector.
@ vmb
r y The relationship between kj and x is determined by Eq. (29).
Hiamb kiamb y kiarc 23
@yc Substituting Eq. (30) into Eq. (29), we obtain

vmb can be solved numerically based on Eq. (19). Substituting Gim  Y im x  Rim x2 U m 0 31
r
vmb
r into Eq. (23), the homogenized material parameters can be where Gim Hijmn kj kn ; Y im cimn kn ; Rim .i dim . It is clear that the
obtained.
coefcient matrix of Eq. (31) is a symmetric matrix. Eq. (31) has
nonzero solution only if the determinant of the coefcient matrix
4. The asymmetrical property analysis of STMMs is equal to zero, i.e.

It is not difcult to prove that the homogenized material jMj det M 0 32

parameters satisfy the following equations where M is a 3  3 matrix, and the component in the i-th row and
Hijmn Hmnij 24 m-th column of the matrix is Gim  Y im x  Rim x2 .
Eq. (32) is the dispersion equation of STMMs, which determines
Hi4m4 0im 25 the relationship between the wave frequency and the wave vector.
Substituting the roots of Eq. (32) into Eq. (31), the displacement
Hi4i4 XHj4j4 ij 26 directions, i.e. the polarization directions, can be solved. It can be
proved that the signs of the roots of Eq. (32) determine the direc-
where X indicates not identical to, while indicates com- tions of wave propagation.
pletely different from. In addition, the homogenized material It is well known that the group velocity is
parameters do not satisfy the Voigt symmetry. That is,
@x
ci 33
Hijmn XHjimn XHjinm ij; nm 27 @ki
If both elastic tensor and density are non-constant functions of It can be proved that x is a rst-degree homogeneous function
both space and time, the parameter of ki (Eq. (32)). Thus, ci is a zero-degree homogeneous function of
ki , i.e. the group velocity depends on the wave propagation
Hi4mn X0 28
direction only. By differentiating Eq. (31) with respect to kp and
and Hi4mn would lead to the STA. multiplying both sides with U i , the group velocity can be solved as
According to Eqs. (24)(28), the material parameter tensor of
2Hkimn kn  ckmi xU k U m
the STMMs is outlined in Fig. 1. ci 34
The next issue to be discussed is how Hi4mn leads to the STA. The cqrs ks U r U q 2x.s U 2s
elastic wave group velocity in STMMs is to be discussed. Disregard
where x is a function of the wave vector direction and can be given
f i , Eq. (22) can be rewritten as
by Eq. (32).
0
@ 2 um
0
@ 2 um @ 2 ui
0 Specially, in the symmetric case, i.e. Hi4mn 0, left hand side of
Hijmn cimn  .i 0 29 Eq. (32) would be an even-degree polynomial. The roots can be
@xj @xn @t@xn @t2
proven to be real values. Hence Eq. (32) has three pairs of roots
where cimn Hi4mn Hinm4 ; .i Hi4i4 . We dene Eq. (29) as the which are opposite numbers mutually, i.e. the roots are sym-
free elastic wave equation of STMMs. metrically distributed to the origin. Thus the two waves have the
same polarization direction and group velocity in opposite direc-
tions according to Eq. (34). These characteristics coincide with
the usual wave theory in crystals.
Generally, in STMMs, roots of Eq. (32) are asymmetrically dis-
tributed to the origin. Thus there can be zero to six polarization
directions corresponding to one wave vector in STMMs. The num-
ber of the waves in one direction may not be equal to 3. It can be
proven that such asymmetric distribution of the roots makes the
elastic waves propagate with different group velocities in opposite
directions and even propagate unidirectionally, i.e. the time-rever-
sal symmetry and wave group velocity surface centrosymmetry are
broken, which might be proposed for applications in acoustic
absorbing, acoustic information processing, energy-saving,
energy-harvesting and so on.
It is noted that an elastic wave has constant group velocity in
STMMs if the wave vector and polarization direction are deter-
mined. The propagation characteristics of the elastic waves along
a determined direction with determined polarization make no
Fig. 1. The schematic diagram of the parameter tensor of STMMs: we use the
difference between STMMs and traditional materials. In addition,
framework of second-order tensor to describe Hiamb . It is noted that the modulus it is likely that Eq. (29) may not be a hyperbolic equation.
and density in STMMs should not be the usual ones any more. Corresponding discussion is reserved for future work.
 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930 23

Nevertheless, it should be pointed out that, theoretically speaking, Hence, the properties vary synchronously with the external elds.
the homogenized solution is full-banded because of the In the lamellar structure, we suppose that the external eld in con-
supposed sufciently large wave length relative to the unit cell trolled layer ii 1; 2; 3; . . . is in the square wave form (Fig. 3(a)):
spatiotemporal scale. Thus the wave frequency would fall into
FA FB
any range if the unit cell is sufciently small, i.e. the Fi 1 Xy4 ; ui  1  Xy4 ; ui  35
unconventional wave properties would be nearly full-banded 2 2
in a well-designed practical model or by using nanostructure. where Xy4 ; ui signsin2py4 =T e ui , sign represents the
sign function, and ui is the phase of the external eld in controlled
5. Theoretical understanding layer i. By neglecting the delay of the property response, the
property in layer i should be
The theoretical understanding is divided into three parts. In the PA PB
rst part, a simple description of a STMM architecture is given by Pi 1 Xy4 ; ui  1  Xy4 ; ui  36
2 2
using a lamellar structure. In the second part, the group velocity
of the lamellar STMM is calculated based on Eqs. (29) and (34) to A unit cell of the lamellar STMM based on Eq. (36) is outlined in
gain an intuitionistic understanding of the STA of STMMs. In the Fig. 3(b). Considering the ideal situation that d=D; s=T e ; D ! 0,
third part, the lamellar STMM is numerically demonstrated by the properties of the whole lamellar structure as layer by layer
using the COMSOL Multiphysics software [40] based on Eq. (7). described in Eq. (36) can be simply expressed as
Part 1: A practical lamellar STMM should contain the control
PA PB
and controlled layers which are distributed alternatingly (Fig. 2). P 1 Xy1  Vy4 ; u1  1  Xy1  Vy4 ; u1  37
2 2
The control and controlled layers thicknesses are d and D,
respectively. The materials in the controlled layers include some where V 2p=T e  D=u ^. u
^ ui1  ui is the phase difference
materials with tunable properties. The materials in all controlled between F i1 and F i . It is obvious that V can be an arbitrary real
layers are of the same kind. The control layers can apply proper value, which can be used to describe the property distribution in
elds to change the material properties in controlled layers. the lamellar structure. A unit cell of the STMMs based on Eq. (37)
Fig. 3(a) is a schematic diagram of the control process. In is outlined in Fig. 3(c). The unit cell length along y1 -axis can be
Fig. 3(a), the materials in controlled layers exhibit property P A or derived as
P B when the external eld F A or F B is applied, respectively. Thus
the periodically varying external elds lead to the periodically Le 2pD=ju
^j 38
e
varying properties of the controlled layers. The eld period T Which indicates that the structure periodicity in space usually
and the property response delay s are shown in Fig. 3(a). is not equal to the unit cell scale in space in STMMs. Such a feature
As actuators of the control system, the control layers can be is completely different from that of the microstructure of the tradi-
made sufciently thin (or be integrated to the controlled layers, tional material. It should be pointed out that the unit cell in
see Fig. A2) to satisfy d  D. According to Popas analysis [19], Fig. 3(c) may not be a good choice due to its complex calculation
the delay of the property response can be neglected, i.e. s  T e . in homogenization. The one in Fig. 3(d) can be used to simplify
the calculation. In such a unit cell, the properties are homogenized
along the property interface AB.
Part 2: The issue in this part is to illustrate the asymmetric wave
property of the lamellar STMM (Fig. 2) by actual material parame-
ters. Based on the properties described by Eq. (37), the homoge-
nized material parameters can be solved analytically. As the
property is homogenized along the property interface AB
(Fig. 3(d)), the rst-order corrector should be the function of
y1  Vy4 . Considering a power series solution of the rst-order
corrector with respect to y1  Vy4 , obviously, the solution of Eq.
(19) is in the form of
mb mb
vmb
r y1 ; y4 Ar y1  Vy4 C r 39
Fig. 2. Elastic bulk with a periodic lamination in space: the control and controlled
layers distribute alternatingly with thicknesses of d and D, respectively. The control
layers are actuators of the control system. The control layers can apply proper elds
where C mb
r is useless in Eq. (23), thus it can be set as zero.Let
to change the material properties in controlled layers. It should be noted that the vmb
r vmb
r if P P A , and vr
mb
vmb
r if P P B . The indices with signs
materials in all controlled layers are of the same kind. ;  are to be used with similar meanings in the following. The

(a) (b) (c)

Control layer 1
Property response of controlled layer ( )
Layer i+1
Layer i

( )

External field applied by control layer (d)
2
( )

( )
1

Fig. 3. Unit cell of the lamellar STMM: (a) control of the material properties in a controlled layer (see Section 2); (b) a rectangle unit cell. The property varying characteristics
in a controlled layer refers to (a); (c) the idealized rectangle unit cell with unit volume: d/D, s=T e ; D ! 0; and (d) an idealized oblique unit cell with unit volume.
24 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930

Considering isotropic materials in the controlled layers, the

properties can be expressed as P q; E; m, where q; E; m are den-
sity, Youngs modulus, and Poissons ratio, respectively. The Lame
constants can be calculated by these parameters. Let the mass den-
sity, Youngs modulus and Poissons ratio for P A be 10,000 kg/m3,
346 GPa, and 0.3 respectively. Corresponding parameters for P B
are 6000 kg/m3, 54 GPa, and 0.3. We would like to discuss two
cases with V = 0 and V = 1400 m/s (Eq. (37)). The case V = 0 is
discussed just for comparison of the material parameters with
the case V = 1400 m/s. In the case V = 0, the STMM degenerates to
a transversely isotropic material.
Based on Eqs. (23), (39) and (42), the homogenized material
parameters are obtained (Fig. 4) as listed in Tables 1 and 2. In
Table 1, the material parameters have the same characteristics as
the transversely isotropic materials. This indicates that our theory
is compatible with the traditional homogenization technique. In
Table 2, we list 14 new material parameters in bold that do not
exist in traditional elastic materials (Table 1). Note that the last
row or column contains two pairs of indices, respectively, thus
the 4 new parameters in such row or column actually represent
8 ones (See note b of Table 2). Eqs. (24)(26) can be veried
directly. Eq. (27) can be veried by H1331 H3131 in Table 2. In
addition, the 14 new material parameters in Table 2 indicate t x1
hat Hi4mn X0, i.e. Eq. (28) is tenable. These equations illustrate the
difference between STMMs and traditional materials.
Based on the material parameters in Tables 1 and 2, by imple-
Fig. 4. Procedure of the group velocity calculation and numerical demonstration of
the lamellar STMM. menting the group velocity calculating procedure in Section 4
(Eqs. (29) and (34), Fig. 4), the homogenized group velocity of
the lamellar STMM is obtained as shown in Fig. 5, in which the
symmetric part below the x1 -axis is neglected.
periodic boundaries on the left and right sides of the unit cell can be When V = 0, the materials are transversely isotropic materials.
expressed as For the quasi-longitudinal (qL) wave with a wave vector along
x1 -axis, the velocity is 3964.8 m/s, and it is 5507.9 m/s along
vmb mb
r y1 ; y1 =V vr y1 ; y1  1=V 40
x2 -axis. For the quasi-shear wave polarized in the vertical plane
The continuities of the corrector at the interface AB can be (qSV wave) with a wave vector along x1 -axis, the velocity is
expressed as 2116.1 m/s, and it is 2119.2 m/s along x2 -axis. For the shear
( wave polarized in the horizontal plane (SH wave) with a wave vec-
mb
vmb
r y1 ; yAB =V vr y1 ; yAB =V tor along x1 -axis, the velocity is 2119.2 m/s, and it is 3099.2 m/s
mb
41
@ vmb
r y1 ; yAB =V @ vr y1 ; yAB =V along x2 -axis. It is obvious that the wave velocity surfaces are cen-
trosymmetric about the origin (Fig. 5).
where yAB y1  1=2. By using Gauss Formula, the second When V = 1400 m/s, the materials are STMMs. For qL wave with
continuous variable in Eq. (41) can be obtained as @ vmb r a wave vector along x1 -axis, the velocities are 4239.6 and
ki1r1 @ vmb
r =@y1 ki1mb  Vk i4r4 @ vmb
r =@y4 ki4mb . 2618.3 m/s, and they are (1397.7, 5607.3) m/s along x2 -axis.
According to Eqs. (40) and (41), the coefcient Amb
r in Eq. (39) For qSV wave with a wave vector along x1 -axis, the velocities are
can be solved as 1840.9 and 50.1 m/s, and they are (348.5, 1758.7) m/s along
x2 -axis. For SH wave with a wave vector along x1 -axis, the
kr1mb  kr1mb  Vkr4mb  kr4mb
Amb mb
r Ar 42 velocities are still 1840.9 and 50.1 m/s, and they are (947.5,
kr1r1 kr1r1 V 2 kr4r4 kr4r4 3075.6) m/s along x2 -axis. According to the data that is given,

Table 1
The material parameters Hiamb (SI) of the STMMs with V = 0.a

ia mb
11 22 33 23, 32b 12, 21(13, 31)c 14 24(34)
11 1.26E+11 5.39E+10 5.39E+10 0 0 0 0
22 5.39E+10 2.43E+11 8.90E+10 0 0 0 0
33 5.39E+10 8.90E+10 2.43E+11 0 0 0 0
23, 32 0 0 0 7.69E+10 0 0 0
12, 21(13, 31) 0 0 0 0 3.59E+10 0 0
14 0 0 0 0 0 8.00E+03d 0
24(34) 0 0 0 0 0 0 8.00E+03e
a
In fact, when V = 0, the lamellar STMM degenerates to transversely isotropic material.
b
The comma in this table means or, which indicates that many material parameters with the same value are merged. For example, according to this table,
H2323 H2332 H3223 H3232 7.69E+10.
c
The material parameters with the same value are merged, too. However, the indices in and out of the brackets are not allowed to be combined. For example, according to
this table, when ia 12 or 21, then mb 12 or 21; when ia 13 or 31, then mb 13 or 31; thus H1212 H1221 H2112 H2121 H1313 H1331 H3113 H3131 3.59E+10.
d,e
These values represent the opposite number of density.
 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930 25

Table 2
The material parameters Hiamb (SI) of the STMMs with V = 1400 m/s.

ia mb
11 22 33 23 or 32 12(13)a 21(31) 14 24(34)
11 9.15E+10 5.01E+10 5.01E+10 0 0 0 2.17E+06b 0
22 5.01E+10 2.41E+11 8.74E+10 0 0 0 9.30E+05 0
33 5.01E+10 8.74E+10 2.41E+11 0 0 0 9.30E+05 0
23 or 32 0 0 0 7.69E+10 0 0 0 0
12(13) 0 0 0 0 2.54E+10 2.54E+10 0 2.57E+06
21(31) 0 0 0 0 2.54E+10 7.63E+08 0 2.57E+06
14 2.17E+06 9.30E+05 9.30E+05 0 0 0 8.03E+03 0
24(34) 0 0 0 0 2.57E+06 2.57E+06 0 8.13E+03
a
See note c of Table 1.
b
The bold parameters are the new material parameters that do not exist in traditional elastic materials (Table 1).

the shear waves (S waves) are unidirectional near x1 -axis. The for-
bidden directions of qSV and SH waves are shown in Fig. 5 as aFqSV
and aFSH , respectively. It can be seen that the wave property of
STMMs exhibits signicant asymmetry (Fig. 5).
Part 3: The issue in this part is to numerically demonstrate the
lamellar STMM. Meanwhile, the numerical results are used to
Fig. 5. The group velocity surfaces of the lamellar STMM: the symmetric part below
the c1-axis is omitted. It can be seen that S waves propagate unidirectionally near
x1-axis. aFqSV and aFSH are the forbidden directions of qSV and SH waves
respectively.
verify the improved homogenization theory for STMMs. To simu-
late the wave propagation behavior in the lamellar STMM, actual
^ and T e , or V and Le . The used
values should be assigned to D; u
wave should have wavelength that is much greater than Le and
wave period that is much greater than T e . In addition, it is obvious
that the wave propagation in a lamellar STMM satises Eq. (7).
Thus, the simulation is based on Eq. (7).
Finite-element simulations with V = 1400 m/s have been per-
formed to solve Eq. (7) with point source by using the COMSOL
Multiphysics software. The numerical model is based on a cube
E0 F0 G0 H0 E00 F00 G00 H00 with an edge length of 600 m as shown in
Fig. 6(a). A slender circular through-hole along x3 -axis with a
radius of 6 m is located at the center of the model. A cross section
EFGH of the model is shown in Fig. 6(b). The boundary conditions
on E0 F0 G0 H0 and E00 F00 G00 H00 are innite boundaries (equivalent to per-
iodic boundaries in this simulation). The boundary condition on
the inner edge of the hole is Dirichlet boundary, which is set as
x  x x x  2
1 2 1 2
u ; ; 1  e4000t0:05 m 43
6 6
where x21 x22 36. The through-hole plays the role of wave source.
This wave source is a Gauss pulse whose peak appears at t = 0.05 s.
Considering that the through-hole is extremely small compared
with the model scale, the wave source can be regarded as a point
source. The boundary conditions on the other faces of the model
are free boundary. Based on Eq. (37), the properties of the model
can be expressed as
P P A  0:5 f x1  Vt P B  0:5  f x1  Vt 44

(a) H' (b) H

E'
Through-hole

G'

F' O
Wave source (c)
600m

E
1.5m

E" G O

G"
F

F"

Fig. 6. The numerical model for the lamellar STMM: (a) the simulation model: a cube with slender circular through-hole along x3-axis; (b) a cross section EFGH of the model;
and (c) part of the meshed elements in the cross section EFGH.
26 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930

(a) (b) (c)

qL wave qSV wave SH wave

t=0.075s t=0.1s t=0.1s

0 0.13 0 0.26 -0.12 0.61

(d) qL wave (e) (f) SH wave

qSV wave
15
10 10
Forbidden Forbidden

Errors (%)
Errors (%)

Errors (%)

Directions Directions
5
5
0
0

-5 -10 -5
0 90 180 270 360 0 90 180 270 360 0 90 180 270 360
() () ()
q q
Fig. 7. The wave propagation in the lamellar STMM: (a) u21 u22 induced by qL waves; (b) u21 u22 induced by qSV waves; (c) u3 induced by SH waves; (d) the group
velocity errors of qL waves; (e) the group velocity errors of qSV waves; and (f) the group velocity errors of SH waves. The boundary dimensions of this gure (a)(c) are
600  600 m2 (Fig. 6(b)). The error is calculated by error = {simulated group velocity  homogenized group velocity}/{homogenized group velocity}, where the homogenized group
velocity refers to Fig. 5, the simulated group velocity is calculated based on the shapes of wave peaks in this gure (a)(c). a in this gure (d)(f) is introduced in Fig. 5. It can
be seen that S waves propagate unidirectionally near x1-axis and the forbidden directions do exist (Figs. 5 and 7(b), (c), (e) and (f)).

where V = 1400 m/s, f x flc2hssinpx=6; 0:05  0:5. c2hs(x,
scale) is the built-in smooth Heaviside function in COMSOL
Multiphysics software. Values of P A and P B are equal to the ones
in calculation of group velocity above. The model is meshed by
the hexahedral elements. The length and width of the elements
are 1.5 m except the rened ones near the hole (Fig. 6(c)). The ele-
ment height can be any value considering the innite boundary.
With a time step of 0.0001 s in the solution process, the successive
over relaxation (SOR) iterative solver is used. Eq. (44) directly pro-
vides the parameters as Eq. (7) requires. The above scheme equiva-
lently establishes a lamellar STMM model with D 0:14 m,
d 0; u^ 0:0733; Le 12 m and T e 0:00857 s.
As the elastic waves in STMMs have no frequency dispersion
(Eq. (34)), the Gauss pulse keeps its waveform during propagating.
By numerically solving Eq. (7), the wave propagation behavior with
a point source in a lamellar STMM is obtained. The numerical
results are shown in Fig. 7(a)(c). The boundary dimensions of
Fig. 7(a)(c) are 600  600 m2 . The waves in Fig. 7(a) correspond
to the qL wave at t = 0.075 s. The waves in Fig. 7(b) and (c) corre-
spond to the qSV and SH waves at t = 0.1 s, respectively. It can be
seen that the elastic waves in STMMs propagate asymmetrically
with respect to the wave source. It is amazing that S waves propa-
gate unidirectionally near x1 -axis and some forbidden directions
do exist.
The above simulation results can be used to verify the homoge-
nized results. As the point source is used in the simulation, the
homogenized results should also be based on the point source. In
fact, Fig. 5 indirectly describes the propagation distances of waves
from a point source in a unit time in the homogenized lamellar
STMM. Thus, it is unnecessary to give additional homogenized
numerical examples, and the group velocity surfaces can be
regarded as homogenized results of wave propagation with a point
source. The homogenized results can be veried by comparison
with the simulation results. The errors of the numerical group
velocity are shown in Fig. 7(d)(f). The error is calculated
by error fsimulated group v elocity  homogenized group v elocityg=
fhomogenized group v elocityg, where the homogenized group
velocity refers to Fig. 5, the simulated group velocity is calculated
based on the shapes of wave peaks in Fig. 7(a)(c). The horizontal-
axis in Fig. 7(d)(f) is the angle between the group velocity direc-
tion and the x1 -axis (Fig. 5). The numerical and homogenized
results of qL wave agree with each other with an error percentage
less than 16%. The numerical error percentage of S waves is less
than 13%. In Fig. 7(e) and (f), the forbidden directions are marked
by brick-shaped patterns.
It is noticed that the errors in some directions (Fig. 7(d)(f)) are
greater than those reported in literature when using the multi-
scale homogenization technique [41]. Analyses show that such
large error is due to the difculties in the numerical simulation
based on Eq. (7). We found that it is extremely difcult to reduce
the error by any conventional method. Dong and Cao [38] pointed
out that the symplectic scheme, which is stable for a long time in
many cases, is not good for solving equations with time-varying
oscillating coefcients. It can be proved that solution of Eq. (7)
satises continuity if the coefcients vary discontinuously as used
in Eq. (44). In other words, the discontinuous coefcients should
not be the root causes of the large errors. Shui et al. [42] proved
that elastic waves in time-varying media do not satisfy frequency
invariance and mechanical energy conservation. Thus, for a unit
cell, STMMs do not satisfy frequency invariance and mechanical
energy conservation. However, for the whole architecture,
STMMs do satisfy frequency invariance and mechanical energy
conservation because the homogenized wave equation (Eq. (29))
has constant coefcients. Therefore, it can be concluded that the
 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930
numerical difculties result from the irreversible energy variance
in the numerical simulation. The meshed element parameters are
varying because of time-varying properties. Such element parame-
ter variance exacerbates the numerical dispersion, which nally
break the total energy balance. Weekes [43] proposed a moving
meshing method to deal with such simulations and gained excel-
lent numerical results. Rousseaus method in fact forms meshed
elements with constant parameters which avoid irreversible
energy variance. However, Rousseaus method is difcult to be
improved for a high dimensional numerical application. In our sim-
ulation, the static meshing method is used, which nally leads to
large errors.
Though the error is large, it is believed that the homogenization
theory for STMMs is still reliable. Firstly, the above analyses indi-
cate that the error occurs in the simulation of equations with
time-varying oscillating coefcients and there is no numerical dif-
culty in the homogenization (see analysis after Eq. (19)).
Secondly, according to Figs. 5 and 7(a)(c), it can be seen that
the overall trends of wave propagation, i.e. the shapes of the
homogenized group velocity surfaces and the shapes of wave peaks
in the simulation results, coincide with each other. The average
percentage errors of group velocities with respect to qL, qSV and
SH waves in all directions are less than 5.46%, 5.40% and 3.65%,
respectively. These indicate that the homogenization theory is reli-
able. Furthermore, the theory of homogenization technique is
widely applied in micromechanics and has been thought effective
[3339,41]. As a result, it is reasonable to prove the reliability of
the homogenization theory of STMMs.
6. Conclusions and perspectives
A structural composite named spatiotemporal metamaterial
(STMM) is theoretically proposed and numerically demonstrated.
An improved dynamic multi-scale homogenization technique is
proposed to obtain the effective material parameters of STMMs.
Some new material parameters of STMMs can lead to asymmetrical
elastic wave property. The wave equation of STMMs is established.
The elastic wave group velocity formula is derived. The unconven-
tional wave property is analyzed theoretically and veried by sim-
ulations. Specically, the STMMs exhibit asymmetrical wave
properties, including group velocity surfaces with broken cen-
trosymmetric i.e. the wave property with broken time-reversal
symmetry. In certain instances, the waves even propagate
unidirectionally. With these unconventionally wave properties,
the STMMs might be proposed for attractive applications in acous-
tic absorbing, acoustic information processing, energy-saving,
energy-harvesting and so on. Compared with existing materials,
the STMMs are tunable and nearly full-banded, which have sig-
nicant advantages in practice.
Although we propose a reasonable mathematical model of
STMMs, it should be pointed out that a lot of difculties and
challenges still exist in STMMs research, such as the analysis of
the wave equation type, the energy analysis, the experimental veri-
cation, reveal of the mechanism and related application research.
Thus many improvements are required. As analyzed in Section 2,
the STMMs can be articially constructed based on smart materials
or structures. However, it remains a challenge to achieve a small-
scaled or nano-scaled, loss-free, rapid responsive, and 3D unit cell
to perform further experimental verications.
Acknowledgements
This work is supported by the National Natural Science
Foundation of China (Grant No. 51205312), and the NPU
Foundation for Fundamental Research (Grant No. JC20110255), to
27
whom the authors express their deep gratitude. We thank J.C.
Qiao for critical reading of the manuscript.
Appendix A. Some issues of time-varying materials
This section involves several basic issues of time-varying mate-
rials, such as the implementation of the time-varying properties
and the balance of energy, which are the basis of our theory. The
Appendix is divided into four parts. In the rst part, a simple intro-
duction of time-varying materials is given. In the second part, two
examples of time-varying materials are given. In the third part, the
energy balance of time-varying materials is analyzed. In the fourth
part, the applicable conditions of the theory of STMMs are given.
Part 1: The time-varying materials specically refer to the mate-
rials which can exhibit rapid time-varying properties (e.g. modu-
lus), and the property variance is not in company with signicant
volume change. It is recognized that certain materials can be in
the state as a function of temperature, electric eld or other exter-
nal elds. This paper mainly focuses on the mechanical behavior of
composites constructed by time-varying materials. In this paper,
the time-varying materials belong to a time-varying mechanical
system. Such a system consists of two subsystems (Fig. A1): the
controlled mechanical system and the control system. The control
system should be non-mechanical (such as electrical system). The
external eld F (such as electric eld) that is manipulated by the
control system can change the material properties P (such as
modulus and density) of the controlled system (P has a direct
effect on its mechanical behavior). Thus a time-varying mechanical
system can be regarded as a multi-eld coupling system. The prop-
erty variance of the material is based on certain special non-me-
chanical mechanism (such as second-order phase transition) or
different types of mechanical mechanism (such as translational
system in certain special rotational system, Fig. A2).
Part 2: In this part, two examples are given to demonstrate the
time-varying properties. The rst example is the giant electrorheo-
logical (GER) material [44]. Such a material has a shear modulus l
that is proportional to the electric eld strength E. Thus the electric
eld becomes the external eld that controls a mechanical
property. A time-varying electric eld signal can produce a
time-varying shear modulus. This example shows the type that
implements the property variance based on a non-mechanical
mechanism.
The second example is a special structure that can implements
the time-varying density. The time-varying density sounds
strange as it is rarely seen. A lot of work about new concepts of
density have made breakthrough in recent years, such as negative
density [4547], anisotropy density [48,49], which were thought
unimaginable in the past. Thus, it is reasonable to believe that
Non-mechanical control system
The time-varying mechanical system
Meso-/Micro-
structure
Mechanical Mechanical
excitation response
The controlled mechanical system
Fig. A1. The time-varying mechanical system: P is the mechanical properties (such
as modulus and density). F is the external elds (such as electric eld) that is
controlling the properties.
28 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930
One layer
Mass
Fig. A2. Time-varying density realization: a period structure with time-varying
moment of inertia (a generalized density) based on controlling of the radial
distribution of the masses (considering torsional wave propagation with large
wavelength). Each layer is controlled independently, i.e. the radial movement of the
mass in one layer does not have an effect on the radial movement of the mass in
adjacent layers.
the time-varying density can be carried out by newly developed
mechanism. The example is shown in Fig. A2. The structure in
Fig. A2 is arranged by a series of disk-shaped layers along an axis.
A layer contains a pair of masses that distributes symmetrically to
the center and can move along radial paths. The movement of the
masses along radial paths can be realized by a rationally designed
non-mechanical system with certain smart materials such as
magnetostrictive material and shape memory materials. Each layer
is controlled independently, i.e. the radial movement of the mass in
one layer does not have an effect on the radial movement of the
mass in adjacent layers. The key issue may be how to obtain the
time-varying density by using such a structure. It should
be pointed out that the propagation of a torsional wave is the con-
cerned mechanical behavior in this example. In this situation, the
moment of inertia J plays the role of generalized density. It can
be proved that the change of moment of inertia is proportional
to r 2 , where r is the distance between the mass and the center of
corresponding layer. A time-varying signal can control the position
of the mass and nally produce time-varying moment of inertia. In
this example, the control mechanism is non-mechanical, however
the property variance mechanism is mechanical. The above two
examples are outlined in Table A1.
Part 3: The issue in this part is the energy balance analysis. The
total time-varying system satises the energy conservation.
However, the control or controlled system itself is energy open.
First, let us review the energy balance of materials with invariable
properties. Considering the isothermal process, the mechanical
energy E can be expressed as
E Ek F A1 F kx
where E k is the kinetic energy, F is the deformation free energy
(strain energy), and dF rij deij . Considering the reversible process,
the change in internal energy DU should be
DU Q  W A2
Table A1
The characteristics of the two examples.
Example GER
F Electric eld strength E
P Shear modulus l
The relationship between F and P Dl / jEtj
The affected mechanical quantities rij ; eij ; v i
Type of the control system Electrical
Property variance mechanism Phase transformation
Type of property variance mechanism Electrical and thermodynamic
where Q is the quantity of heat added to the materials, W is the
work done by the materials, and
W DF A3
In the following, we consider the energy balance of materials
that experienced a transient property variance from the above
invariable situation. Let the work done by F on the materials be
W F . The control system changes the properties and motion of the
controlled system, simultaneously [42]. The property variance
results from the materials meso-/micro-structure change which
would change the internal energy with DU F . The mechanical
state of the materials performs as the impedance of changing the
meso-/micro-structure of the materials. This means that, to change
the material properties, the materials in different mechanical
states may require different quantities of work done by the control
system. The property variance directly lets E k and F have
increments DE kF and DF F , respectively. To sum up, the total
energy balance can be expressed as following:
E V E DE kF DF F A4
DU V DU DU F DF F A5
W V DF  DF F A6
W F DE kF DF F DU F A7
where the quantities with a subscript of V corresponds to the
ones that experienced a transient property variance from the origi-
nal invariable situation. In addition, it should be noted that the
external eld may do negative work to the system (for example,
in Fig. A2, the torsional wave makes the controlled layer get a
centrifugal eld. In such a eld, the radial moving mass
obviously gets the chance to do negative work if the torsion speed
is high enough), and it can be proved that high frequency property
variance in STMMs does not need innite energy under certain
conditions.
Part 4: This part focuses on the applicable conditions of the
theory of STMMs. To analyze the applicable conditions, a basic
carrier is proposed, rst. A general time-varying system is an
energy and momentum open system. For simplicity and
generality, a simple harmonic oscillator formed by mass con-
nected to spring (Fig. A3(a)) is used to conduct the preliminary
analysis. It should be noted that such an oscillator can suf-
ciently reect the basic features of an elastic-inertia system.
In Fig. A3(a), the mass m, when displaced from its equilibrium
position, experiences a restoring force F, proportional to the dis-
placement x:
A8
where k > 0 is the spring constant. The mass has a moving
velocity v .
With consideration of time-varying properties, the spring
constant k and mass m should be a function of time. The oscillator
transient should be a basic process of the time variance. Based on a
Fig. A2
Certain non-mechanical eld
Moment of inertia (a general density) J
DJ / r2 t
Angular velocity, angle strain, torque
Certain non-mechanical
Mechanical structure change
Mechanical
 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930
(a) (b) (c)
m Fig. A2, PV1 should be Dh 0; X0 0, where h is the angle strain
Connection
m + m
F m m'= m
x
x'
O
O'
k+k
k'=k
k k
External energy and momentum
Fig. A3. Simple harmonic oscillator formed by mass connected to spring: (a)
passive case: the spring constant k and mass m are constants; (b) transient case: the
spring constant k and mass m get increments of Dk and Dm, respectively; and (c) an
equivalent model of the transient harmonic oscillator: parallel connection of the
original oscillator system and additional imaginary pre-deformed spring and
moving mass.
logical view, the (non-mechanical mechanism based) transient
(Fig. A3(b)) can be mechanically equivalent to parallel connection
of the original oscillator system and additional imaginary pre-de-
formed spring and moving mass (Fig. A3(c)). It should be noted
that the control system does not appear in Fig. A3. The additional
pre-deformed spring and moving mass in Fig. A3(c) are indepen-
dent of each other and do not belong to any oscillation system
before the transient. Once parallel connected, the position of the
two masses will be bounded together. Considering that the
variances of spring constant and mass of the original oscillator,
respectively, are Dk and Dm (Fig. A3(b)), the spring constant and
pre-deformation of the imaginary spring, respectively, should be
0 Rev B 2013;88(2):024303.
k Dk and x0 ; and the imaginary mass and its moving velocity,
respectively, should be m0 Dm and v 0 (Fig. A3(c)). The new restor-
ing force RF and momentum Rp should be
RF kx  k0 x0 k Dkx  DkDx A9
Rp mv Dmv 0 A10
where Dx x0  x. The energy and frequency variances of the har-
02
monic oscillator should be Rp2 =m Dm Dkx  mv 2 =2, and [26] Bergamini A, Delpero T, Simoni LD, Lillo LD, Ruzzene M, Ermanni P. Phononic
p p crystal with adaptive connectivity. Adv Mater 2013. http://dx.doi.org/10.1002/
k Dk=m Dm  k=m, respectively. It is obvious that the
frequency variance is an intrinsic property of the time-varying sys-
tem. However the energy variance is associated with mechanical
state. It cannot be ignored that the connection process also requires
energy from the control system.
Using the time-varying simple harmonic oscillator as a basic
carrier, the property variance is classied into two kinds:
Property variance of the first kind PV1 : Dx 0; v 0 0
Property variance of the second kind PV2 : Otherwise:
A11
It is noted that a time-varying system with PV1 is an energy
open but momentum closed system. Such classication can be
easily generalized to an Hookean elastomer, considering the
correspondences between k and kijkl , m and q, x and eij , and v
and v i . It can be proved that the elastodynamics for a material with
PV1 (Eq. (A11)) can be described by Eq. (7), and the one with PV2
(Eq. (A11)) cannot be directly described by Eq. (7). Thus the theory
of STMMs is only applicable to the material with PV1 (Eq. (A11)). It
29
should be noted that the property variances in the two examples
above as listed in Table A1 satisfy PV1, and, in the example in
and X is the angular velocity.
References
[1] Zheludev NI. The road ahead for metamaterials. Science 2010;328(5978):
5823.
[2] Wegener M. Metamaterials beyond optics. Science 2013;342(6161):93940.
[3] Keiser GR, Fan K, Zhang X, Averitt RD. Towards dynamic, tunable, and
nonlinear metamaterials via near eld interactions: a review. J Infrared
Millimeter Terahertz Waves 2013;34(11):70923.
[4] Leonhardt U. Optical conformal mapping. Science 2006;312(5781):177780.
[5] Al A, Engheta N. Achieving transparency with plasmonic and metamaterial
coatings. Phys Rev E 2005;72(1):016623.
[6] Zhang S, Xia CG, Fang N. Broadband acoustic cloak for ultrasound waves. Phys
Rev Lett 2011;106(2):024301.
[7] Popa BI, Zigoneanu L, Cummer SA. Experimental acoustic ground cloak in air.
Phys Rev Lett 2011;106(25):253901.
[8] Deymier PA. Acoustic metamaterials and phononic crystals. Berlin Heidelberg:
Springer-Verlag; 2013. p. 159.
[9] Terraneo M, Peyrard M, Casati G. Controlling the energy ow in nonlinear
lattices: a model for a thermal rectier. Phys Rev Lett 2002;88(9):094302.
[10] Wang L, Li B. Thermal logic gates: computation with phonons. Phys Rev Lett
2007;99(17):177208.
[11] Fan L, Wang J, Varghese LT, Shen H, Niu B, Xuan Y, et al. An all-silicon passive
optical diode. Science 2012;335(6067):44750.
[12] Liang B, Guo XS, Tu J, Zhang D, Cheng JC. An acoustic rectier. Nat Mater
2010;9(12):98992.
[13] Liang B, Yuan B, Cheng JC. Acoustic diode: rectication of acoustic energy ux
in one-dimensional systems. Phys Rev Lett 2009;103(10):104301.
[14] Nesterenko VF, Daraio C, Herbold EB, Jin S. Anomalous wave reection at the
interface of two strongly nonlinear granular media. Phys Rev Lett 2005;95(15):
158702.
[15] Boechler N, Theocharis G, Daraio C. Bifurcation-based acoustic switching and
rectication. Nat Mater 2011;10(9):6658.
[16] Lepri S, Casati G. Asymmetric wave propagation in nonlinear systems. Phys
Rev Lett 2011;106(16):164101.
[17] Silva A, Monticone F, Castaldi G, Galdi V, Alu A, Engheta N. Performing
mathematical operations with metamaterials. Science 2014;343(6167):1603.
[18] Mei J, Ma GC, Yang M, Yang ZY, Wen WJ, Sheng P. Dark acoustic metamaterials
as super absorbers for low-frequency sound. Nat Commun 2012;3:756.
[19] Popa BI, Zigoneanu L, Cummer SA. Tunable active acoustic metamaterials. Phys
[20] Hess O, Pendry JB, Maier SA, Oulton RF, Hamm JM, Tsakmakidis KL. Active
nanoplasmonic metamaterials. Nat Mater 2012;11(7):57384.
[21] Soukoulis CM, Wegener M. Optical metamaterials-more bulky and less lossy.
Science 2010;330(6011):16334.
[22] Zheludev NI, Kivshar YS. From metamaterials to metadevices. Nat Mater
2012;11(11):91724.
[23] Li TC, Gong ZX, Yin ZQ, Quan HT, Yin XB, Zhang P, et al. Space-time crystals of
trapped ions. Phys Rev Lett 2012;109(15):163001.
[24] Wilczek F. Quantum time crystals. Phys Rev Lett 2012;109(15):160401.
[25] Shapere A, Wilczek F. Classical time crystals. Phys Rev Lett 2012;109(15):
160402.
adma.05280.
[27] Coleman P. Quantum physics time crystals. Nature 2013;493(7431):1667.
[28] Crawley EF, Deluis J. Use of piezoelectric actuators as elements of intelligent
structures. AIAA J. 1987;25(10):137385.
[29] Halsey TC. Electrorheological uids. Science 1992;258(5083):7616.
[30] Jolly MR, Bender JW, Carlson JD. Properties and applications of commercial
magnetorheological uids. J Intel Mater Syst Str 1999;10(1):513.
[31] Lee EW. Magnetostriction and magnetomechanical effects. Rep Prog Phys
1955;18:184229.
[32] Liu C, Qin H, Mather PT. Review of progress in shape-memory polymers. J
Mater Chem 2007;17(16):154358.
[33] Geers MGD, Kouznetsova VG, Brekelmans WAM. Multi-scale computational
homogenization: trends and challenges. J Comput Appl Math 2010;234(7):
217582.
[34] Parnell WJ, Abrahams ID. Dynamic homogenization in periodic bre reinforced
media. Quasi-static limit for SH waves. Wave Motion 2006;43(6):47498.
[35] Fritzen F, Bhlke T. Reduced basis homogenization of viscoelastic composites.
Compos Sci Technol 2013;76:8491.
[36] Jacques S, De Baere I, Van Paepegem W. Application of periodic boundary
conditions on multiple part nite element meshes for the meso-scale
homogenization of textile fabric composites. Compos Sci Technol 2014;92:
4154.
[37] Casado-Diaz J, Couce-Calvo J, Maestre F, Martin-Gomez JD. Homogenization
and corrector for the wave equation with discontinuous coefcients in time. J
Math Anal Appl 2011;379(2):66481.
30 L.-Q. Shui et al. / Composites Science and Technology 113 (2015) 1930
[38] Dong QL, Cao LQ. Multiscale asymptotic expansions and numerical algorithms
for the wave equations of second order with rapidly oscillating coefcients.
Appl Numer Math 2009;59(12):300832.
[39] Willoughby N, Parnell WJ, Hazel AL, Abrahams ID. Homogenization methods to
approximate the effective response of random bre-reinforced composites. Int
J Solids Struct 2012;49(13):142133.
[40] COMSOL. Equation-Based Modeling. Version 4.2.1.110; 2008.
[41] Kalamkarov AL, Georgiades AV. Modeling of smart composites on account of
actuation, thermal conductivity and hygroscopic absorption. Compos Part B-
Eng 2002;33(2):14152.
[42] Shui LQ, Yue ZF, Liu YS, Liu QC, Guo JJ. One-dimensional linear elastic waves at
moving property interface. Wave Motion 2014;51(7):117992.
[43] Weekes SL. Numerical computation of wave propagation in dynamic
materials. Appl Numer Math 2001;37(4):41740.
[44] Wen WJ, Huang XX, Yang SH, Lu KQ, Sheng P. The giant electrorheological
effect in suspensions of nanoparticles. Nat Mater 2003;2(11):72730.
[45] Lee SH, Park CM, Seo YM, Wang ZG, Kim CK. Acoustic metamaterial with
negative density. Phys Lett A 2009;373(A5):44649.
[46] Huang HH, Sun CT. Wave attenuation mechanism in an acoustic metamaterial
with negative effective mass density. New J Phys 2009;11:013003.
[47] Lee SH, Park CM, Seo YM, Wang ZG, Kim CK. Composite acoustic medium with
simultaneously negative density and modulus. Phys Rev Lett 2010;104(5):
054301.
[48] Flores-Mendez J, Perez-Rodriguez F. Metasolid with anisotropic mass density.
Epl-Europhys Lett 2013;103(5):54001.
[49] Torrent D, Sanchez-Dehesa J. Anisotropic mass density by two-dimensional
acoustic metamaterials. New J Phys 2008;10:023004.