Professional Documents
Culture Documents
16456M M.van-Hecke - Cleaned
16456M M.van-Hecke - Cleaned
16456M M.van-Hecke - Cleaned
Muamer Kadic1,2 , Graeme W. Milton3 , Martin van Hecke4,5 , and Martin Wegener2
Metamaterials are composites which are rationally designed, aiming at effective material param-
eters that go beyond (=”meta”, Greek) those of the ingredient materials. For example, negative
metamaterial properties can result from positive-only ingredients. Likewise, large metamaterial pa-
rameter values can arise from all-zero constituents (e.g., magnetic from non-magnetic, chiral from
achiral, anisotropic from isotropic, etc.). Pushing the boundaries of accessible material behavior has
been one of the driving forces of the metamaterial field. While the field emerged from linear elec-
tromagnetism two decades ago, it addresses nearly all conceivable aspects of solids today, ranging
from electromagnetic/optical, mechanical/acoustic, to transport properties – linear and nonlinear,
reciprocal and non-reciprocal, monostable and multistable (”programmable”), active and passive, as
well as static and dynamic. Here, dynamic not only refers to frequency-dependent yet fixed material
properties but to properties that actually change versus time via response to stimuli. In this Review,
we focus on three-dimensional periodic metamaterials. We will outline fundamental bounds of these
composites and summarize the state-of-the-art in theoretical design and experimental realization.
Figure 1. From atoms via 3D materials to designed unit cells and 3D metamaterials.|
(a) and (b) Ordinary crystalline or amorphous materials are built up from atoms. (c) To compress
the underlying complexity, materials are often treated as fictitious continuous media with associated
effective parameters such as the electric conductivity, the optical refractive index, or the mechanical
Young’s modulus. (d) These effective media are used as the ingredients for rationally designed artificial
unit cells. (e) Out of these, periodic or non-periodic 3D metamaterials are assembled. (f) Again,
to compress complexity, metamaterials are mapped onto fictitious continuous media. Notably, the
resulting effective metamaterial properties can go beyond (= ”meta”, Greek) those of the ingredient
materials, qualitatively and/or quantitatively. This includes, e.g., the possibility of sign reversals and
unbounded behavior. The example unit cell shown in (d) leads to auxetic behavior. Further examples
are depicted in Figure 2.
truly fundamental bounds – which may be dif- Homogenization theory [8–11] aims at pro-
ferent for 1D, 2D, and 3D – cannot be over- viding a sound mathematical basis for map-
come. ping periodic structures onto effective media
Effective parameters. Materials are com- or continua. Notably, it is (presently) not pos-
monly described by effective macroscopic ma- sible to homogenize just any periodic struc-
terial parameters referring to fictitious con- ture and map it onto an effective-medium
tinua [5, 7]. Examples are the electric conduc- description, although high-frequency homog-
tivity, the Hall coefficient, the electric permit- enization has been around for some years
tivity, or the Young’s modulus. In this man- [7, 8, 12] and continues making progress in
ner, the complexity of a large system com- this direction [13]. Issues persist in regard
posed of many different materials can be re- to dealing with interfaces in an unambiguous
duced. With the addendum way [11]. Homogenization becomes especially
challenging in the limit that the material con-
trast of the ingredient materials is large or
. . . The properties of the metamaterial can be
even diverges as the cell size shrinks to zero
mapped onto effective-medium parameters.
[14–16].
cussed in this review, the mapping onto effec- mous textbook [27] on electromagnetic con-
tive parameters is possible because some char- tinua, Landau and Lifshitz argue, based on an
acteristic length scale, e.g., the wavelength or inequality containing the atomic lattice con-
a transport length, is sufficiently large com- stant a < 1 nm, that the relative magnetic
pared to the period or lattice constant. Still, permeability at optical frequencies is close to
one should be aware that effective continuum unity, µ = 1. In other words, the magnitude
descriptions are usually not perfect [17–19], of the magnetic dipole density excited by the
not even for ordinary crystals. Furthermore, magnetic field of light is negligible. As a re-
the mapping onto effective parameters is gen- sult, most optics textbooks essentially only
erally not unique. As a simple example, we deal with the relative electric permittivity r ,
can describe the optical properties of silicon which describes an electric dipole density ex-
either by a frequency-dependent conductivity cited by the electric field of light.
or by a frequency-dependent complex refrac-
One should be surprised: It is no prob-
tive index. For other architectures discussed
lem to induce a current by a time-varying
below, it is not yet clear whether homogeniza-
magnetic field via Faraday’s law into a small
tion is possible. However, researchers have
coil with inductance L, made out of metalli-
improved and conceptually expanded homog-
cally conducting wire. The induced circulat-
enization theories over the years and are con-
ing current leads to a local magnetic dipole
tinuing to do so. Hence, a structure that can-
moment via Ampère’s circuital law. This mo-
not be mapped onto an effective medium to-
ment can be made large by building a res-
day, may be tomorrow. For yet other periodic
onant LC circuit out of the coil and a ca-
architectures, homogenization does not make
pacitance C. Densely packing many such cir-
sense. For example, it is generally problem-
cuits into a 3D metamaterial leads to a large
atic to map the complex band structure of a
positive magnetic permeability µr below the
photonic [20] or phononic [21] crystal, which
resonance frequency and to µr < 0 slightly
may even contain topological band gaps [22–
above the resonance. Many of the metamate-
24], onto effective material parameters – with
rial unit cells shown in Figure 2 include vari-
the notable exception of the lowest bands in
ations of this motif of a coil with one or more
the long-wavelength limit.
of slits, i.e., a ”split-ring resonator” [28, 29]
Several other notions used in the literature (see Fig. 2(b)), also discussed in the literature
are more or less synonymous to the notion under the names ”slotted-tube resonator” [30],
of metamaterials, notably “architected mate- ”loop-gap resonator” [31], or ”cut-wire pairs”
rials” (mainly used in the context of mechan- [32].
ics), ”designer matter”, ”artificial materials”,
or ”properties on demand”. Metamaterials What prevents us from reaching optical fre-
can be periodic in 1D, in 2D (”metasurfaces” quencies? As the Maxwell equations are scal-
[25]) or in 3D. This Review focusses on 3D able, reducing the size by a factor of ten also
metamaterials. reduces the resonance wavelength by the same
We start historically with negative-index factor, increasing the resonance frequency by
metamaterials and end with the exciting per- a factor of ten. The answer lies in the finite
spective of stimuli-responsive and space-time electron density hence finite constituent met-
metamaterials. In between, however, this als plasma frequency, which is usually in the
review is not organized chronologically, but ultraviolet spectral region. When remotely
rather aims at emphasizing analogies and dis- approaching the plasma frequency, the metal
similarities between electromagnetic/optical, properties deviate from those of an ideal con-
mechanical/acoustic, and transport metama- ductor. The same physics can alternatively
terials. We spare out designed inhomoge- be expressed by the kinetic inductance [38],
neous metamaterial distributions, which, for which adds to the Faraday inductance [39]
example, enable invisibility cloaks and coun- and results in an upper limit of the LC fre-
terparts thereof, as this extension would de- quency. In a circuit picture, further miniatur-
serve a review article on its own [26]. ization towards the atomic scale is counter-
productive, because the resistance R scales
Electromagnetic/optical metamaterials inversely with size, hence damping increases
Sometimes, bounds believed to be fundamen- and the resonance gets washed out. This situ-
tal are actually not. The field of optical ation creates opportunities for metamaterials
metamaterials started like that: In their fa- with lattice constants a much larger than the
4
Figure 2. Gallery of designed 3D optical metamaterial unit cells and corresponding exper-
imental realisations.| Classic unit cells used in all areas: (a) ABAB... laminate [33] Unit cell used in
optics: (b) arrangement of metallic split-ring resonators leading to artificial magnetism [34], (c) fishnet
arrangement for uniaxial negative refractive indices [35], (d) helices providing chiral behavior [36], and
(e) multiple intertwined helices for recovering three-fold rotational symmetry [37].
atomic scale (compare Figure 1), yet smaller tive real part to imaginary part, the so-called
than the operation wavelength λ, such that figure of merit, has not exceeded values on
the structure can approximately be described the order of ten at optical frequencies [35].
as an effective medium. Landau and Lifshitz This means that the light intensity decreases
[27] did not see that coming (Figure 2). by more than 70% over just one medium
Negative refraction. The direction of the wavelength – i.e., a bulk metamaterial with
flow of light can be changed by refraction at an extent of many wavelengths is essentially
an interface between two materials. The re- opaque. In addition, the few experimen-
fractive index, n, in Snell’s law determines tal studies actually considering 3D structures
this change. Unusual behavior can arise from were mostly uniaxial, leading to anisotropic
negative n, from anisotropies, and from chi- refractive index tensors, of which only compo-
rality. nents were negative. Related restrictions ap-
ply to zero-index and epsilon-near-zero meta-
A positive sign of n means that the vec-
materials [52–54].
tors of phase velocity ~vp and energy velocity
~ve or Poynting vector S ~ (which are generally An alternative route to negative refrac-
not parallel to the group velocity vector [40]) tion are anisotropic materials [55]. Hyper-
point in the same direction (~vp · S ~ > 0), while bolic or indefinite [56–60] metamaterials (see
they point in opposite directions if the refrac- Fig. 2(a)) are a special class of anisotropic
~ < 0). This
tive index is negative [41–46] (~vp · S media for which the electric permittivity is
situation, which can occur if the real parts of positive (“dielectric”) along one direction and
r and µr are both negative [41, 42], is highly negative (“metallic”) for the orthogonal one,
unusual (see Fig. 2(c)). It created much of the leading to hyperbolic iso-frequency contours
early excitement in the field of metamaterials, in momentum space. The hyperbola (in con-
in large part due to the idea of ”perfect lenses” trast to a circle or ellipse) gives access to large
[42] overcoming the Abbe diffraction barrier. momenta of light, hence to unusually small ef-
fective wavelengths and ”superresolution” at
Causality imposes finite imaginary parts
optical frequencies.
onto n, reflecting propagation losses. Math-
ematically, the Kramers-Kronig relations [47] Optical magnetism. However, magnetism
derived from causality allow to obtain zero at elevated frequencies creates additional op-
imaginary part of n at a single finite fre- portunities: µr 6= 1 allows for adjusting
quency of light by introducing gain media [48– p electromagnetic wave impedance Z =
the
51], but, unavoidably, losses quickly increase µ0 µ∗r /(0 ∗r )) and make itpequal to the vac-
with increasing detuning from this singular uum impedance of Z0 = µ0 /0 = 376 Ω,
point. In passive structures based on metal- independent of n. By impedance matching,
lic constituents, the inferred ratio of nega- reflections from interfaces between a mate-
5
rial and vacuum/air can thus be eliminated edness of light impinging from vacuum/air at
completely, regardless of the materials refrac- all, whereas they do interact (strongly) with
tive index. Here losses, which have been a the opposite one. This property is intimately
nuisance above, can be turned into an ad- connected to duality [73] (∗r = µ∗r . 3D meta-
vantage in perfect absorbers: Materials that materials with extremal chirality have poten-
neither reflect nor transmit any light [61–64]. tial applications in angle-insensitive helicity-
The idea of achieving impedance matching via filtering glasses in stereoscopic 3D projection
balanced electric and magnetic responses is systems and in the optical sensing of chiral
also key for the field of 2D Huygens meta- molecules [73, 74]. Such metamaterials can
surfaces [65]. For the case of µ∗r < 0 and potentially be built exclusively from dielectric
∗r > 0, magnetic mirrors result [66] that ide- constituents. In essence, the Ohmic current in
ally completely reflect the light, yet with a metals is replaced by the displacement current
different phase shift than ordinary metal mir- in dielectrics, leading to low losses.
rors (where r < 0 and µr = 1). If the excitation of magnetic dipoles by the
Cross couplings. So far, we have tacitly ne- electric field of light is not ”symmetric” with
glected chiral effects. Macroscopically, chiral respect to the excitation of electric dipoles by
effects such as, e.g., optical activity, can be the magnetic field, a medium breaks time-
described by the dimensionless chirality pa- reversal symmetry, becomes Faraday active,
rameter ξ(ω) [67, 68]. Microscopically, ξ sum- and behaves non-reciprocally. For a linear
marizes that magnetic dipole moments are ex- lossless stationary passive medium, this re-
cited by the electric component of the elec- quires an internal or external static magnetic
tromagnetic light wave and vice versa. In the field. As one consequence, the transmission
general anisotropic case, these “cross terms” of light in one direction is no longer equal
are refered to as bianisotropy [67, 68]. We to the transmission in the opposite direction.
will encounter analogous couplings in Erin- For example, the bulky optical isolators in
gen continuum mechanics [69] below. These telecommunication systems are cost drivers
cross terms can only be nonzero if space in- and, hence, improved Faraday metamaterials
version symmetry is broken. For light imping- would be highly welcome. Interesting steps
ing under normal incidence onto a planar 2D in this direction have been taken [75]. This
system and emerging from it along the same effect must not be confused with asymmetric
axis, chiral effects are zero by symmetry [70]. polarization conversion, which is misleadingly
Therefore, 3D metamaterials are of particular sometimes referred to as ”asymmetric trans-
interest here. mission” [76]. We will come back to other op-
As mentioned at the beginning of this sec- tions of breaking time-reversal symmetry in
tion, chirality enables negative refractive in- time-dependent and nonlinear media below.
√
dices via the relation n∗± = ∗r µ∗r ± ξ ∗ . Here, Finally, to obtain any linear material re-
± refers to left- and right-handed circular po- sponse, one needs more than just electric and
larization, respectively. More interestingly, magnetic dipole moments. Toroidal moments
chiral 3D metamaterials can exhibit optical have been discussed as a new multipole family
activity many orders of magnitude larger than for 3D optical metamaterials [77, 78]. How-
what is found in natural substances. Further- ever, reference [79] recently proved mathe-
more, the wave impedance and the absorp- matically that electric and magnetic multipole
tion coefficient also depend on the handed- moments suffice to capture all degrees of free-
ness of the circularly polarized light eigen- dom for this purpose.
states. This allows for realizing circular po- Nonlinearities. The combination of res-
larizers [36], i.e., materials that transmit one onances, local-field enhancements, and local
circular polarization of light and that reflect symmetry breaking has raised hopes for novel
and/or absorb the opposite handedness nearly highly efficient nonlinear optical [80–82] and
completely. 3D helical metallic metamaterials electro-optic 3D metamaterials [83, 84]. In-
(see Fig. 2(d) and (e)) approach this ideal and deed, novel geometries and record-high non-
even work over one to two octaves of band- linear optical susceptibilities have been re-
width [37, 71, 72]. ported [82], especially for the nonlinear refrac-
Recently, the concept of 3D metamaterials tive index, n2 , of metamaterials with r ≈ 0,
with extremal chirality was introduced [73]. hence n ≈ 0 [82]. However, it has also been
Such media do not interact with one hand- pointed out [85] that 3D metamaterials and
6
theoretical blueprints for them have not im- fractive indices under meaningful conditions
proved compared to established nonlinear op- is not trivial [101]. Most solid materials
tical crystalline materials in regard to the have an acoustic impedance that is orders of
figure-of-merit, i.e., in terms of the accessible magnitude larger than the impedance of air,
phase shift per absorption length or the acces- such that close to 100% of the acoustic wave
sible frequency-conversion efficiency per ab- is reflected at the interface. Thereby, the
sorption length. Again, as for negative-index phase velocity in the medium becomes irrel-
metamaterials, losses inherited from metallic evant. Labyrinthine metamaterials [102, 103]
ingredients are responsible [85]. Purely di- (see Fig. 3(d)) can roughly match the acoustic
electric 3D metamaterials [82, 86, 87] gen- impedance of air. Yet, by providing a winding
erally exhibit less pronounced enhancements detour by labyrinthine channels for sound in-
but might turn out to be more useful in prac- side of the unit cell bounded by rigid walls, on
tice [88]. a scale much smaller than the acoustic wave-
length, the phase velocity of the wave is effec-
Acoustic and mechanical metamaterials tively slowed down. Experiments have been
Let us start with the conceptually simplest reported for 2D [104, 105] and 3D metamate-
case of airborn (or waterborn) acoustics. For rials [103, 106].
acoustic waves in isotropic continua, the com-
pressibility κ is mathematically analogous to An interesting twist which has no counter-
the electric permittivity r and the mass den- part in electromagnetism is that the back-
sity ρ is analogous to the magnetic perme- ground air can also be actively driven in the
ability µr [89]. Therefore, negative refractive channels of an acoustic metamaterial. This
indices can be obtained in acoustics in anal- motion breaks reciprocity for the pressure
ogy to optics [90–93]. For example, a nega- wave propagating in the moving air. Metama-
tive mass density at finite frequency ω 6= 0 terials based on unit cells in which the air is
can be achieved slightly above a mass-and- locally circulating, driven by fans, have been
spring resonance, where the instantaneous ac- demonstrated [114]. By constrictions in the
celeration and force are 180 degrees out of channels, which locally modify the fluid mo-
phase [94] (see Fig. 3(b)). Taking different tion, effectively gain and loss regions can be
springs in the three spatial directions, the mimicked [115]. On this basis, metamaterials
simple scalar mass density ρ can turn into that are symmetric with respect to simultane-
a frequency-dependent rank-2 mass-density ous space-inversion and time-inversion (PT-
tensor ρ(ω) [95, 96]. 3D blueprints for such symmetric metamaterials) can be constructed
metamaterials [96–98] exploiting this freedom [115].
have been suggested theoretically (see [99]). Cauchy elasticity. The generalization of
Independent adjustment of the magnitudes Hooke’s law in one dimension to 3D solids led
of scalar B(ω) and ρ(ω) also enables acous- Cauchy to his elasticity tensor, C , which con-
tic impedance matching (analogous to optics nects stresses and strains [116]. The rank-
above). Combined with finite absorption, per- 4 Cauchy elasticity tensor is already much
fect acoustic absorbers exhibiting zero reflec- richer than the rank-2 electric permittivity
tion and zero transmission become possible tensor in optics. In a homogeneous isotropic
[93]. For a single Lorentzian resonance, ab- medium, the permittivity tensor reduces to
sorption can only be close to 100% over a a scalar, whereas the elasticity tensor does
limited frequency range. However, using a not. It can be parameterized by two scalars
clever distribution of folded Fabry-Pérot res- though, e.g., by the bulk modulus B, equal
onators with different resonance frequencies to the inverse of the compressibility κ, and
in each unit cell, close to 100% absorption has the shear modulus G. Alternatively, one can
been obtained experimentally over more than use the Young’s modulus E and the Poisson’s
two octaves of frequency from 500-3000 Hz ratio ν, or other combinations [116]. In an
[100]. The absorber thickness of about 11 cm unconstrained stable passive medium under
has been close to the fundamental limit de- stationary conditions (ω = 0), B and G can-
termined by causality. Some of the authors not be negative. Likewise, a Hooke’s spring
have commercialized these 3D metamaterials constant cannot be negative. The bounds on
for applications in noise reduction. other elastic parameters (such as E, ν, or
In acoustics, in contrast to optics, obtain- the Lamé parameters) follow from that [117].
ing off-resonant moderately large positive re- Intuitively, the difference in the rank of the
7
Figure 3. Gallery of designed 3D acoustical and mechanical metamaterial unit cells and
corresponding experimental realizations.| Unit cells used in mechanics/acoustics: (a) pentamode
cell leading to small shear modulus [107] (compare with auxetic cell in Figure. 1(d)) leading to small
bulk modulus [108]), (b) internal mass-spring resonance leading to negative effective mass density
[94], (c) programmable mechanical metamaterial [109], (d) 3D labyrinthine channel system leading to
an isotropic slowing down of sound propagation [103], (e) 3D chiral mechanical metamaterial [110],
(f) buckling elements leading to multistable and programmable behavior [111], and (g) truss lattices
with large coordination number leading to strong ultralight behavior [112]. (h) Unit cells for stimuli-
responsive behavior: two-component cell supporting sign reversal of thermal expansion [113].
tensors is connected to the fact that electro- above), the shear modulus is generally not
magnetism usually only supports transverse zero, G 6= 0. In auxetic elastic metamate-
waves, whereas elasticity generally supports rials, the effective shear modulus G∗ can even
transverse and longitudinal waves at the same be made larger than the effective bulk mod-
time. For the special case of acoustics, e.g., ulus, i.e., G∗ > B ∗ . Equivalently, the ef-
with air or water as the medium, only the lon- fective Poisson’s ratio is negative, ν ∗ < 0.
gitudinal pressure waves remain as the shear Such materials contract laterally when push-
modulus is zero, G = 0 (see above). ing onto them (see Fig. 1(d)). For the limit
For elastic solids, unlike air or water (see of 3D isotropic dilational metamaterials with
8
large-sample limit [130, 131]. Fig. 3(f)) out of such buckling beams, in par-
Recent theoretical work has yet gone one allel and in serial, couples them. Under strain
step further and considered gyroelastic meta- control, the resulting behavior is nonlocal be-
materials [5, 132–134] – the counterpart of cause the buckling of one beam influences
Faraday active metamaterials in optics. A beams far away as the displacements of all
built-in continuously rotating gyroscope re- unit cells have to add up to the pre-described
places the static magnetic field required in overall displacement. The resulting stress-
optics. Thereby, the mechanical metamate- strain behavior is multistable [143, 144, 147].
rial unit cell not only breaks space-inversion Upon loading and unloading such metamate-
symmetry via chirality but also breaks time- rial, it does not follow the same path. This
inversion symmetry. This leads to an asym- means that energy is irreversibly dissipated
metric cross-coupling between torques and into heat during each cycle. As the con-
displacements on the one hand, and between stituent material can be purely linearly elas-
forces and rotations on the other hand, re- tic in this process, the cycling is repeatable
spectively [69, 127] (see optics above for com- many times. Applications in terms of shock
parison). absorbers have been suggested [147]. The
buckling beams can also be designed to be
We should like to emphasize, however, that
multi-stable by themselves. In this case, the
the generalizations of Cauchy elasticity dis-
3D metamaterial has multiple stable states
cussed so far are not the only possible ones.
at zero external force. The different states
The rotations in Eringen continuum mechan-
generally have different linear elastic behav-
ics are in a way perturbations to regular elas-
ior [109]. Therefore, the elastic behavior can
ticity. By contrast, in ”strain-gradient meta-
be programmed in this sense (also see Origami
materials” [135], some components of the ef-
metamaterials [148–151]). Finally, in the non-
fective elasticity tensor are zero, such that
linear regime, e.g., for a buckling metama-
the main contribution is the strain gradient
terial, the stress-strain curve, σ(), does not
term [136–141]. Again, Cauchy elasticity is
need to by symmetric with respect to push-
not recovered in the large-sample limit. An-
ing or pulling, i.e., σ(−) 6= −σ(). This
other generalization are Willis metamaterials
asymmetry in the nonlinear response for a 1D
[96]. There, stress also depends on accelera-
“fishbone” metamaterial has been interpreted
tion, and momentum also depends on the dis-
in terms of static non-reciprocity [152]. In
placement gradient, not just the velocity.
sharp contrast, in the linear elastic regime,
Nonlinearities. So far, we have focused the stress-strain curve for passive media must
on linear mechanics. However, nonlinearities obey the condition σ(−) = −σ() – even for
play a tremendous role in engineering. We asymmetric structures.
distinguish between reversibly nonlinear me-
chanical behavior, such as buckling instabili- With respect to irreversible mechanical fail-
ties [142–144], and irreversibly nonlinear be- ure and fracture, the idea of 3D light-weight
havior, such as failure or fracture of the con- metamaterials has sparked considerable inter-
stituent material(s). est [112, 145, 153–156]. When decreasing the
While reversible nonlinearities are often volume filling fraction, f , of the constituent
only a minor correction to the linear be- material, the effective mass density ρ∗ obvi-
havior in optics, geometrical nonlinearities ously decreases proportionally, ρ∗ ∝ f . At
can be huge effects in mechanics – even if the same time, the stiffness and the strength,
the constituent material behaves perfectly lin- the stress σ at failure, also decrease unavoid-
ear. The nonlinearities can be tailored by ably. However, the scaling exponent η be-
the metamaterial unit cell geometry. It has tween effective strength σ and mass density,
been proven by construction that essentially σ ∗ ∝ (ρ∗ )η , can be influenced by the 3D meta-
any nonlinear mechanical behavior can be material architecture [154, 155, 157]. It, e.g.,
achieved [145, 146]. Even multi-stable behav- depends on whether the behavior is domi-
ior is possible based on the buckling of beams. nated by beam bending or stretching. Truss-
This classic mechanism is closely related to based lattices with large coordination num-
the “click” you hear when pushing onto the lid ber (or structural connectivity) turn out to
of a marmalade jar. Beyond a certain strain, be favorable [157] (see Fig. 3(g)). Along these
the stress no longer increases but rather de- lines, 3D microlattices approaching the max-
creases. Assembling a 3D metamaterial (see imum possible theoretical strength have been
10
achieved [154–156]. All of these metamate- bound can be broken in the presence of ad-
rials are not scalable in the sense that their ditional tailored spatial distributions of static
properties change if all unit cell dimensions para- and diamagnetic constituents within the
are scaled down by the same factor. To avoid composite [164].
confusion, we emphasize that loss of scalabil- The dimensionless non-negative ZT -value,
ity in the sense used here is distinct from that with ZT = σS 2 T /κ and the thermodynamic
mentioned in the context of micropolar 3D temperature T , is the commonly used figure-
metamaterials (see above). of-merit for devices aiming at efficient ther-
moelectric energy conversion [166]. It has
Transport metamaterials
been shown theoretically, that for an ar-
Being response functions (see above), sim-
bitrary composite in a small or zero mag-
ple transport coefficients such as the elec-
netic field, the effective ZT ∗ -value cannot ex-
tric conductivity σ, the diffusivity D, and
ceed the largest ZT -value of its constituents
the thermal conductivity κ of passive (meta-
[166]. This technologically unfortunate bound
)materials cannot be negative under station-
can be broken in the presence of large mag-
ary conditions (i.e., ω = 0) due to energy
netic fields though [166, 167] (complement of
conservation and the second law of thermo-
Fig. 4(a)). Furthermore, at least one of the
dynamics. The specific heat c of a stable ma-
constituents of the metamaterial has to have
terial cannot be negative either. Therefore,
a strong thermoelectric response and another
parameter sign reversals of these quantities
constituent has to have a strong Hall effect,
in analogy to the refractive index (see above)
i.e., a large Hall mobility (see above). It has
are not allowed. This rule does not apply
been argued though that the power factor (or
to active metamaterials though (see counter-
conversion capacity) may be equally impor-
example given below).
tant technologically [168]. Power factors of
Other stationary transport coefficients such thermoelectric materials, unlike the figure-of-
as the Hall coefficient [158–160] AH , the Et- merit, can be improved by orders of magni-
tingshausen coefficient PE , the Seebeck coeffi- tude through metamaterials, e.g., by lamina-
cient S, the Peltier coefficient Π, and the rela- tion (see Fig. 2(a)) of two or more thermoelec-
tive magnetoresistance change ∆R/R can be tric constituent materials.
positive or negative in the stationary regime
[161]. While obtaining non-reciprocal behav- Transport experiments. In analogy to op-
ior is a hot and demanding topic in optics, tics and mechanics, passive laminates have led
acoustics, and mechanics (see above), it is to anisotropic electrical, thermal, and diffu-
standard in transport due to the existence of sive transport, using isotropic constituent ma-
semiconductor pn-diodes. terials [5, 169], as well as to improved thermo-
electric power factors [168] (see Fig. 4(b)).
Additional bounds apply for composites.
The isotropic Hall coefficient A∗H of a 2D Following an earlier theoretical suggestion,
metamaterial, made of positive (negative) absolute negative mobility was shown in ac-
constituents, in a perpendicular magnetic tive 2D microfluidic systems (not called meta-
field cannot be negative (positive) [159]. The materials then) [170]. There, micrometer-
same holds true for 3D hierarchical laminates sized polystyrene beads always moved in a di-
[5]. In 3D structures beyond laminates, this rection opposite to the net acting force as a
restriction does not apply and parameter sign result of an interplay between thermal noise,
reversal is allowed [160, 162, 163]. The same a periodic and symmetric microfluidic struc-
reasoning analogously applies to the Etting- ture, and a biased alternating-current electric
shausen coefficient PE . For Hall-effect based field [170].
magnetic-field sensors, which are, e.g., used Following preceding theoretical work [161],
in the compass apps of many modern mobile passive microstructures (not called metama-
phones, the modulus of the Hall mobility µH , terials then) composed of Hall bars made
which is given by the product µH = AH σ, out of a 3D layer of n-type GaAs, punc-
determines the sensitivity and the signal-to- tured by a square array of cylindrical voids
noise ratio of the sensor. In a metamaterial in the xy-plane (see Fig. 4(a)), have led to
made of non-magnetic constituents, the effec- a highly anisotropic effective classical magne-
tive isotropic Hall mobility µ∗H is fundamen- toresistance [161]. Its modulus at T = 90K
tally bounded by the Hall mobilities of the was found to be as large as ∆Rxx /Rxx ≈ 65%
constituents (up to a factor of two) [164]. This at an in-plane magnetic field corresponding to
11
Figure 4. Gallery of designed transport metamaterial unit cells and corresponding exper-
imental realizations.| (a) square array of cylindrical holes in a plate [161], leading to anisotropic,
indefinite, or hyperbolic behavior. (b) hierarchical or Maxwell laminate. (c) chainmail-like arrangement
of tori leading to sign reversal of the isotropic Hall coefficient [163] and (d) anisotropic cell supporting
the parallel Hall effect [165].
By = 12 T [161]. In sharp contrast, the con- Hall voltage parallel (rather than perpendic-
stituent GaAs crystal showed essentially zero ular) to the external magnetic field [165] (see
magnetoresistance. In the strict 2D limit of a Fig. 4(d)). The parallel Hall effect could be
thin platelet, the magnetoresistance tends to used for sensors measuring the local circu-
zero theoretically [158]. Intuitively, the origin lation of a magnetic field [165]. It appears
of the metamaterial magnetoresistance was that any Hall tensor that is compatible with
interpreted in terms of a “geometrical shadow” the Onsager relations [171] can be realized by
of the current cast by the voids in the semi- 3D metamaterials composed of isotropic con-
conductor along the direction of the magnetic stituents [164]. However, a constructive proof
field [161]. for this conjecture is absent so far.
Following preceding theoretical work [159,
160, 162], a sign reversal of the effective 3D Stimuli-responsive metamaterials
isotropic Hall coefficient with respect to the The properties and associated effective-
n-type ZnO constituent (i.e., sign(A∗H AH ) = medium parameters of metamaterials are not
−1) has been observed [163] at room temper- necessarily fixed, they can be influenced by
ature in cubic symmetry chainmail-like meta- external stimuli and thereby be made to
materials (see Fig. 4(c)), composed of inter- change deterministically versus time. This
linked hollow tori. The origin of this sign aspect requires that the constituent materi-
reversal can be traced back to the different als respond to some sort of stimulus, e.g.,
topologies of a bulk material and a (hollow) to the local light intensity, or local electric
torus made out of it [163]. Two further con- field, magnetic field, pressure, temperature,
ceptually different paths to a sign reversal of etc. Liquid crystal displays are a widespread
A∗H in 3D have been suggested theoretically example. Here, the optical properties are
[164]. They are based on exchanging the pick- changed by the application of electric fields
up leads and reversing the local direction of which locally modulate the local liquid crystal
the magnetic field, respectively, both in cubic orientation in each pixel. Using light as the
symmetry. In lower symmetry metamaterials stimulus for transparent metamaterials allows
(see Fig. 2(d)), off-diagonal components of the to selectively address individual unit cells in
Hall-tensor can dominate over the diagonal three dimensions, which tends to be difficult
components [164]. Experiments on 3D meta- conceptually for other stimuli. 3D stimuli-
materials have shown the resulting unusual responsive constituent materials are an ac-
12
tive field of research on their own [172–181]. field pattern. Macroscopically, the wave am-
Ideally, the metamaterial unit cell leverages plitude may blow-up exponentially or remain
these changes in some way, e.g., by resonances bounded according to whether one is in the
or geometrical nonlinearities. Provided the case of broken or unbroken PT-symmetry (see
metamaterial is multi-stable, the stimulus can above). In the unbroken case, the wake be-
even program the metamaterial in the sense hind a wavefront does not die away, rather
that the reversibly induced change persists af- the material remains excited.
ter the stimulus is being switched off. Table So far, however, the concept of space-
1 gives an overview on the experimental sta- time metamaterials is theoretical [211]. Ex-
tus of stimuli-responsive and programmable periments need to identify suitable stimuli-
metamaterials (see e.g. Fig. 3(c)), albeit not responsive mechanisms (see Table 1) with suf-
all in 3D. The columns refer to different ma- ficiently small response times and correspond-
terial properties, the rows to different types ing constituent materials, allowing for the
of stimuli. manufacturing of such architectures.
Stimulating all unit cells of a 3D metamate- Conclusions and perspectives
rial in the same way leads to tunable metama- 3D metamaterials are rationally designed
terials, which could find applications as mod- composites, aiming at effective material pa-
ulators. The metamaterial unit cells can be rameters that go beyond those of their bulk
the same initially and then change in differ- ingredients – qualitatively and/or quantita-
ent ways, such that a specific functionality is tively. While the interest exploded about
induced by a stimulus [182–187]. Moreover, twenty years ago, triggered by the advance
the time-dependent stimulation can be such of negative-index metamaterials, we should
that it creates a circulation of some property humbly admit that not all new ideas are
within one metamaterial unit cell. The axis good and that not all good ideas are new.
associated to this circulation acts analogous More than a Century ago, Maxwell discussed
to a static magnetic field in Faraday-active laminates, providing an anisotropic response
optical materials and can thus lead to non- from isotropic ingredients. In 1920, Lind-
reciprocal propagation (see above) of electro- man investigated arrangements of metal he-
magnetic [188] or acoustic waves [114] in 2D lices, leading to a giant effective chiral re-
metamaterial lattices made out of 3D unit sponse (optical activity) from achiral ingre-
cells. dients. The list goes on [5]. However, at least
In addition, the properties can vary not just three things have changed. The rise of nan-
in space but also in time. Rather than hav- otechnology has enabled the making of op-
ing a ”unit-cell” of periodicity in space, one tical metamaterials, composed of unit cells
has a unit cell of periodicity in 4D space-time with sub-wavelength feature sizes. Further-
[179, 206, 207]. Modulation of material prop- more, reliable 3D Additive Manufacturing on
erties in time can be achieved by the exter- various spatial scales has enabled the fabrica-
nal stimuli as discussed above. Many unusual tion of complex 3D architectures for electro-
behaviors are theoretically predicted in space- magnetism, optics, acoustics, mechanics, and
time metamaterials, although it should be em- transport that seemed very difficult if not im-
phasized that theory, so far, has focused on possible to make twenty years ago. In paral-
only one spatial dimension plus time. One lel, we have seen significant progress and new
aspect is that such metamaterials at rest in approaches in theoretical design efforts, which
the laboratory frame can mimic the Doppler could build on substantial advances in numer-
frequency shift for media moving in the labo- ical computation and inverse design.
ratory frame when waves are reflected off their The metamaterial field can be proud. A
interface [208]. In addition, one can get arti- long-standing dream of solid-state physics is
ficial magnetic fields for photons with an ar- to design materials on the computer, to avoid
tificial Lorentz effect so they travel along cir- tedious trial-and-error procedures and exces-
cular arcs [209]. Tilted band structures can sive experimentation. However, the truth is
provide unidirectional band gaps [210]. Fur- that this dream has been realized in only few
thermore, strikingly novel types of wave be- exceptions to date. Metamaterials are an en-
havior are possible, referred to as field pat- tire class of such exceptions. A vast variety
terns. The material wave and the propagat- of highly unusual 3D metamaterial properties
ing wave are concentrated on a pattern, the has been creatively predicted and then con-
13
Acad. Sci. USA 110, 3276 (2013). Nature 436, 928 (2005).
[149] Z. Y. Wei, Z. V. Guo, L. Dudte, H. Y. [171] L. Onsager, Phys. Rev. 37, 405 (1931).
Liang, and L. Mahadevan, Phys. Rev. Lett. [172] J. D. Debord and L. A. Lyon, The Journal
110, 215501 (2013). of Physical Chemistry B 104, 6327 (2000).
[150] S. Waitukaitis, R. Menaut, B. G.-g. Chen, [173] M. A. C. Stuart, W. T. S. Huck, J. Gen-
and M. van Hecke, Phys. Rev. Lett. 114, zer, M. Müller, C. Ober, M. Stamm, G. B.
055503 (2015). Sukhorukov, I. Szleifer, V. V. Tsukruk,
[151] J. L. Silverberg, J.-H. Na, A. A. Evans, M. Urban, F. Winnik, S. Zauscher, I. Luzi-
B. Liu, T. C. Hull, C. D. Santangelo, R. J. nov, and S. Minko, Nat. Mater. 9, 101–113
Lang, R. C. Hayward, and I. Cohen, Nat. (2010).
Mater. 14, 389 (2015). [174] R. C. Schroden, M. Al-Daous, C. F. Blan-
[152] C. Coulais, D. Sounas, and A. Alù, Nature ford, and A. Stein, Chem. Mater. 14, 3305
542, 461–464 (2017). (2002).
[153] H. Gao, B. Ji, I. L. Jäger, E. Arzt, and [175] P. Theato, B. S. Sumerlin, R. K. O’Reilly,
P. Fratzl, Proc. Nat. Acad. Sci. USA 100, and T. H. Epps, III, Chem. Soc. Rev. 42,
5597 (2003). 7055 (2013).
[154] T. A. Schaedler, A. J. Jacobsen, A. Tor- [176] T. Skylar, Architectural Design 84, 116
rents, A. E. Sorensen, J. Lian, J. R. Greer, (2014).
L. Valdevit, and W. B. Carter, Science 334, [177] Z. Hao, W. Piotr, P. Camilla, M. Daniele,
962 (2011). B. Matteo, and W. D. Sybolt, Adv. Mater.
[155] X. Zheng, H. Lee, T. H. Weisgraber, 27, 3883 (2015).
M. Shusteff, J. DeOtte, E. B. Duoss, J. D. [178] D. Martella, D. Antonioli, S. Nocentini,
Kuntz, M. M. Biener, Q. Ge, J. A. Jackson, D. S. Wiersma, G. Galli, M. Laus, and
S. O. Kucheyev, N. X. Fang, and C. M. C. Parmeggiani, RSC Adv. 7, 19940 (2017).
Spadaccini, Science 344, 1373 (2014). [179] F. Momeni, S. M. M. H. N, X. Liu, and
[156] J. Bauer, S. Hengsbach, I. Tesari, J. Ni, Materials and Design 122, 42 (2017).
R. Schwaiger, and O. Kraft, Proc. [180] N. Akihiro, M. Ahmed, Z. Hang, and
Nat. Acad. Sci. USA 111, 2453 (2014). M. Martin, Adv. Sci. 5, 1700038 (2017).
[157] J. Bauer, R. Meza Lucas, T. A. Schaedler, [181] P. Haedong, K. Hyungho, A. Yongsan,
R. Schwaiger, X. Zheng, and L. Valdevit, Y. Woong-Ryeol, M. Myoung-Woon, and
Adv. Mater. 29, 1701850 (2002). H. Kahyun, Phys. Status Solidi (RRL) 12,
[158] R. S. Popovic, Hall Effect Devices (Institute 1800040.
of Physics Publishing, Bristol, Philadelphia, [182] O. Hess, J. B. Pendry, S. A. Maier, R. F.
2004). Oulton, J. M. Hamm, and K. L. Tsak-
[159] M. Briane, G. W. Milton, and V. Nesi, makidis, Nat. Mater. 11, 573 (2012).
Arch. Rational Mech. Anal. 173, 133 [183] I. Shadrivov, M. Lapine, and Y. Kivshar,
(2004). Nonlinear, Tunable and Active Metamate-
[160] M. Briane and G. W. Milton, SIAM J. Appl. rials, Springer Series in Materials Science
Math. 70, 1810 (2010). (Springer International Publishing, 2014).
[161] M. Tornow, D. Weiss, K. von Klitzing, [184] K. Fan and W. J. Padilla, Mater. Today 18,
K. Eberl, D. J. Bergman, and Y. M. Strel- 39 (2015).
niker, Phys. Rev. Lett. 77, 147 (1996). [185] X. Tong, Functional Metamaterials and
[162] M. Kadic, R. Schittny, T. Bückmann, Metadevices, Springer Series in Materials
C. Kern, and M. Wegener, Phys. Rev. X Science (Springer International Publishing,
5, 021030 (2015). 2017).
[163] C. Kern, M. Kadic, and M. Wegener, Phys. [186] S. Rout and S. Sonkusale, Active Metama-
Rev. Lett. 118, 016601 (2017). terials: Terahertz Modulators and Detectors
[164] C. Kern, W. M. Graeme, M. Kadic, and (Springer International Publishing, 2017).
M. Wegener, New J. Phys. 20, 083034 [187] K. Yu, N.-X. Fang, G. Huang, and
(2018). Q. Wang, Adv. Mater. 30, 1706348 (2018).
[165] C. Kern, V. Schuster, M. Kadic, and [188] J. S. Nicholas A. Estep, Dimitrios L. Sounas
M. Wegener, Phys. Rev. Applied 7, 044001 and A. Alu, Nat. Phys. 10, 923–927 (2014).
(2017). [189] M. Ozaki, Y. Shimoda, M. Kasano, and
[166] D. J. Bergman and Y. M. Strelniker, Phys. K. Yoshino, Adv. Mater. 14, 514–518
Rev. B 49, 16256 (1994). (2002).
[167] Y. M. Strelniker and D. J. Bergman, Phys. [190] M. Kamenjicki, I. Lednev, and S. A. Asher,
Rev. B 96, 235308 (2017). Adv. Funct. Mater. 15 (2005).
[168] L. Liu, New J. Phys. 16, 123019 (2014). [191] T. J. Cui, M. Q. Qi, X. Wan, J. Zhao, and
[169] R. Schittny, M. Kadic, S. Guenneau, and Q. Cheng, Light Sci. Appl. 3, e218 (2014).
M. Wegener, Phys. Rev. Lett. 110, 195901 [192] Z. Jie et al., arXiv:1806.04414 (2018).
(2013). [193] N. Courjal, S. Benchabane, J. Dahdah,
[170] A. Ros, R. Eichhorn, J. Regtmeier, T. T. G. Ulliac, Y. Gruson, and V. Laude, Appl.
Duong, P. Reimann, and D. Anselmetti, Phys. Lett. 96, 131103 (2010).
18
[194] D. Shin, J. Kim, C. Kim, K. Bae, S. Baek, gener, Sci. Rep. 7, 40643 (2017).
G. Kang, Y. Urzhumov, D. R. Smith, and [203] H. Zhang, X. Guo, J. Wu, D. Fang, and
K. Kim, Nat. Commun. 8, 16090 (2017). Y. Zhang, Sci. Adv. 4, 8535 (2018).
[195] J. M. Weissman, H. B. Sunkara, A. S. [204] X. Zhang, J. Liu, M. Chu, and B. Chu,
Tse, and S. A. Asher, Science 274, 959–963 Appl. Phys. Lett. 109, 072903 (2016).
(1996). [205] M. Kamenjicki, I. k. Lednev, A. Mikhonin,
[196] S. Kubo, Z.-Z. Gu, K. Takahashi, A. Fu- R. Kesavamoorthy, and S.-A. Asher, Adv.
jishima, H. Segawa, and O. Sato, J. Am. Funct. Mater. 13 (2003).
Chem. Soc. 126, 8314 (2004). [206] V. Bacot, M. Labousse, A. Eddi, M. Fink,
[197] D. Roy, J. N. Cambre, and B. S. Sumerlin, and E. Fort, Nature Physics 12, 972 (2016).
Prog. Polym. Sci. 35, 278 (2010), special [207] Deck-Léger, Zoé-Lise, Akbarzadeh, Alireza,
Issue on Stimuli-Responsive Materials. and C. Caloz, Phys. Rev. B 97, 104305
[198] D. H. Gracias, Current Opinion in Chemical (2018).
Engineering 2, 112 (2013). [208] J. C. Halimeh, R. T. Thompson, and
[199] V. Laude, A. Belkhir, A. F. Alabiad, M. Ad- M. Wegener, Phys. Rev. A 93, 013850
douche, S. Benchabane, A. Khelif, and F. I. (2016).
Baida, Optica 4, 1245 (2017). [209] K. Fang, Z. Yu, and S. Fan, Nature Photon.
[200] S. Babaee, N. Viard, P. Wang, N. X. Fang, 6, 782 (2012).
and K. Bertoldi, Adv. Mater. 28, 1631–1635 [210] H. Nassar, H. Chen, A. N. Norris, and G. L.
(2016). Huang, Phys. Rev. B 97, 014305 (2018).
[201] Z. Nicolaou and A. Motter, Nat. Mater. 11, [211] S. Xu and C. Wu, Phys. Rev. Lett. 120,
608 (2012). 096401 (2018).
[202] J. Qu, M. Kadic, A. Naber, and M. We-