Twists and turns for metamaterials Mingkai Liu,1 Yue Sun,1 David A. Powell,1 Ilya V. Shadrivov,1 Mikhail Lapine,2 Ross C. McPhedran,2 and Yuri S. Kivshar1 1Nonlinear Physics Centre and Centre for Ultrahigh-bandwidth Devices for Optical Systems (CUDOS), Research School of Physics and Engineering, Australian National University, Canberra ACT 0200, Australia 2Centre for Ultrahigh-bandwidth Devices for Optical Systems (CUDOS), School of Physics, University of Sydney, Sydney NSW 2006, Australia We propose and verify experimentally a new concept for achieving strong nonlinear coupling be- tweentheelectromagneticandelasticpropertiesinmetamaterials. Thiscouplingisprovidedthrough 3 anoveldegreeoffreedominmetamaterialdesign: internalrotationwithinstructuralelements. Our 1 meta-atoms have high sensitivity to electromagnetic wave power, and the elastic and electromag- 0 netic properties can be independently designed to optimise the response. We demonstrate a rich 2 range of nonlinear phenomena including self-tuning and bistability, and provide a comprehensive experimental demonstration of the predicted effects. n a J Metamaterialsresearchhasgrownrapidlyoverthepast changes in the mutual orientation. Instead of a consid- 5 2 decade,exhibitingawidevarietyofnewwavephenomena erable displacement of an entire array [9], it is sufficient [1, 2]. Being initially conceivedin the domain of electro- to move the crucial parts of the resonant particles with ] magnetics [3–5], the metamaterial concept also proved respectto eachother—forexample,the gapsofthe two s c to be fruitful in other areas of physics [6–8]. Until re- coupled split-ring resonators — which involves a more i cently, however, direct interplay between different types gentle geometric alteration. t p of physical effects within the same metamaterial wasnot Let us now introduce our novel concept: nonlinear o considered,althoughmechanicalcontroloverelectromag- metamaterials with intrinsic rotation. As a building ele- . s neticmetamaterialpropertieswasemployedinstructural ment of the structure (see Fig. 1), we considertwo coax- c tuning [9, 10]. ialsplit ring resonators(SRRs) with elastic feedback be- i s It turns out that introducing a mechanical degree of tween them. The rings are allowed to rotate about the y h freedom into electromagnetic metamaterials leads to an common axis, while the elastic feedback is provided by p interestingrangeofnonlineareffects,givingrisetoanew connecting a thin elastic wire. Indeed, the use of elastic [ class of magnetoelastic metamaterials [11] and to wide- wires has a prominent history in physics, being utilised bandoperation[12]. Therangeofpossibleeffectsachiev- inthemilestoneachievementoftheexperimentaldemon- 1 v able in this way promisesto be richerthan inthe promi- strationoflightpressurebyP.N.Lebedev[18]. Here,we 0 nentareaofoptomechanics[13],becausethegreaterflex- employ the electromagnetic (EM) torque to construct a 6 ibility in metamaterial design overcomes the limits of “light-driven”meta-atom, which enables us to modulate 9 available material functionalities, and offers wider pos- the resonant frequencies by twisting the rotatable ele- 5 sibilities for optimisation. At the same time, the imple- mentdirectly withEMwaves. ThecomponentoftheEM . 1 mentation of magnetoelastic metamaterials [11] remains forces which twists the rings with respect to each other 0 challengingandinsomecases,suchastheconformational 3 1 nonlinearityinresonantspirals[14],remainsinaccessible : foroptics. Thereasonforthisisthatthemagneticforces, v employed in the initial designs, are relatively weak, so i X such materials require either high power or extremely r small elastic restoring forces, which poses a considerable a manufacturing challenge. We recall, however, that earlier research on struc- turally tunable metamaterials [9] indicated that near- field interactionmay significantly improvethe tunability range,leadingtovariouseffectsassociatedwithnear-field coupling [15]. In particular, changing the mutual orien- tationbetweentheneighbouringelementshasaprofound effectontheresonatorcouplingandthestructureoftheir FIG. 1: Conceptual layout of a new metamaterial and its modes [16, 17]. rotational “meta-atom”. Incident wave propagates along y We therefore expect that the most efficient approach direction, having a linear polarisation with the electric field toimplementdynamiccouplingbetweenelectromagnetic along xandmagnetic field along z. Theinducedelectromag- and mechanical effects should rely on near-field interac- netictorquebetweentheresonatorschangesthemutualtwist tion, which canhavea powerfulinfluence evenfor subtle angle θ between therings, connected by an elastic wire. 2 is normally not the strongest among the forces involved, Tostartwith,weuseasemi-analyticalmodeltostudy buttheprominentadvantageofusingEMtorqueinstead the dynamics of an isolated meta-atom in free space. It ofcollinearEMforcetodriveameta-atomisthattheef- will subsequently be demonstrated that this model ex- fective leverarmofthe azimuthalEMforcecanbe much plains all qualitative features of an array. As shown in larger than that of the azimuthal restoring force from a Fig.1,the twocoaxialidenticalSRRsareoffsetbya dis- thinwire;thiscaneffectivelymagnifythedeformationin tance s in the z direction, and the twist angle between azimuthal direction by orders of magnitude. them is θ. The incident wave propagates along the y Technically, there are a number of ways to implement direction, with its magnetic field in the z direction and this general scheme; in our design, one of the rings is electricfieldinthexdirection. Tostudythenonlinearbe- fixed to a substrate and the other one is suspended on haviouroftherotatablemeta-atom,weutiliseanefficient the long wire. The three symmetrically positioned wires analyticalmodelbasedonthesinglemodeapproximation which attach the suspended ring to the string provide and the near-field interaction [16]. This model can pro- stability against tilt. In this design, therefore, the only videareasonablepredictionoftheEMresponseaswellas favourable movement is the rotation of the suspended the optomechanical properties of the structures [19]. As ringwithrespecttothefixedringoverthe commonaxis, our previous studies showed [20], the current and charge and all other mechanical degrees of freedom can be ne- of the SRR can be separated into frequency-dependent glected. mode amplitudes and spatially-dependent distributions: Suppose the initial position is such that the ring slits J(r,ω) = −jωQ(ω)j(r) and ρ(r,ω) = Q(ω)q(r). The have a certain angle between them with respect to the mode amplitudes Q1,2 can be obtained after solving the common axis (Fig. 1). An EM wave then induces a cer- coupled equations: tain distribution of charges and currents in the two res- onators, and the resonance of the system is determined Q1 =(E2Fm−E1Fs)/(Fs2−Fm2) (1) by their mutual orientation [16]. These chargesand cur- Q2 =(E1Fm−E2Fs)/(Fs2−Fm2), rentsalsoresultinEMtorquebetweenthetworings[19], which drives the suspended ring to rotate until the EM where E1 and E2 correspond to the effective voltage ap- torque is compensated by the elasticity of the twisted plied to the lower (index 1) and top (index 2) SRRs by wire. Meanwhile, the entire pattern of charges and cur- the external fields; Fs and Fm are the self and mutual rents gets modified and the torque also changes, so the impedance terms, with their explicit forms given in the final stable equilibrium is only achieved via a complex supplemental materials [21]. nonlinearfeedback. Thetwistedwireprovidesarestoring Once the frequency-dependent mode amplitudes are torque to balance the EM torque, so that we can control known, we can calculate the EM torque experienced by thetwistangle(andthustheresonance)bychangingthe the SRRs. Here, we are particularly interested in the external field signal. torque on the top rotating ring: MEM = R ρ(r2)r2 × V2 An additional feature of the proposed design is that E+r2×[J(r2)×B]dV2,wheretheintegrationisperformed the nonlineardynamicsoftherotatableparticlenotonly over the volume V2 of the top SRR. We decompose the depend on the parameters of the wire, but also on the total torque into two parts: external torque Mext con- EMmode initially excitedinthe resonators,whichis de- tributed by the externalincident fields [19], and internal termined by the starting angle between the gaps. The torqueMint duetothenear-fieldinteractionbetweenthe lattercanbemadearbitrary,andanimplementationhas two SRRs. We assume that the lower SRR is fixed while the possibility to deliberately adjust it, introducing tun- the top SRR is only allowed to rotate about the z axis. ability to the system. See the supplemental materials [21] for explicit expres- The above design, therefore, offers a tunable resonant sions for the internal and external torque. nonlinear system with elastic feedback and, as we show Figures 2(a) and 2(b) depict the mode amplitude Q2 below,yields arichpatternofnonlinearresponseinclud- and the total EM torque MEM = Mext +Mint experi- ing self-tuning and nonlinear bistability, which are much encedbythetopSRRasfunctionsoffrequencyandtwist strongerthanthoseprovidedbyusingnonlinearsemicon- angle θ. Since radiation losses are taken into account in ductor components. Below, we present a detailed theo- themodel[20],weareabletoaccuratelydescribetheevo- retical analysis of rotational meta-atoms. We then pro- lution of the mode amplitudes, phases and lineshapes of ceed to the experimental results obtained with a fabri- the resonances. As expected, this chiral meta-atom sup- cated prototype of the rotational“meta-atom”placed in ports two hybrid resonances,which can be characterised a rectangular waveguide, and confirm all important fea- as symmetric (lowerfrequency branch)and antisymmet- tures predicted by theory. Finally, we perform full-wave ric (higher frequency branch) modes, according to the numericalsimulationsofthe arrayof“meta-atoms”(i.e., symmetry of the H component [16]. z metamaterial),anddemonstratethatallthenonlinearef- The directions of the EM torque at these two res- fects observedforthe single meta-atominthe waveguide onances are also opposite. For the symmetric mode, ◦ are qualitatively the same in the array. θ = 0 corresponds to the configuration of highest po- 3 tential energy (unstable point), and thus the two repel ◦ each other once θ > 0 , until they come to the stable ◦ state at θ = 180 , while the reverse is true for the an- tisymmetric mode [22]. The evaluated external torque is about one order of magnitude smaller than the in- ternal torque, and the total torque is of the order of 10−10Nm when the structure is pumped with a power density PI=1mW/mm2. This is confirmed by the full- wave simulation (CST Microwave Studio) followed by calculation based on the Maxwell stress tensor, which yields the EM torque through a surface integral of the FIG.3: Schematicoftheexperimentalset-up. Thepumpand field components around the object [23]. probe signals are combined by a 3dB combiner. The sample is positioned in the centre of the waveguide. 1: vector net- The overall mechanism of achieving a nonlinear ef- work analyser; 2: 3dB combiner; 3: rectangular waveguide; fect is presented in Figs. 2(c,d). As an example, we 4: 20dB attenuator; 5: power amplifier; 6: signal generator; choose a pump frequency (3.5GHz) at the symmetric 7: sample. mode [regime denoted by the black dashed line in Fig. 2 (b)]. It can be seen that MEM is a Lorentz-like func- tion of the twist angle, while the restoring torques MR gle,itnaturallyleadstononlinearsolutions. Asshownin under different initial twist angles θ0 are approximated Fig.2(d),the power-dependenttwistanglesunderdiffer- by linear functions (Hooke’s law). The intersections of entinitialanglesdemonstratetheevolutionfromsmooth these two functions, MEM(θe,PI)+MR(θe) = 0, corre- nonlineartobistable responseasθ0 departsfromthe an- spond to the equilibrium angles θe. However, only the gle of maximum EM torque. In principle, as θ0 moves angles with ∂∂θ [MEM(θe)+MR(θe)] < 0 are stable. As further away from the resonance, more noticeable rota- the pumppowerPI increasesfromzerotomaximumand tionandhysteresiseffectsareexpected,buthigherpump ◦ then reduces, the stable angles also change accordingly. powerisrequired(seethecaseforθ0 =45 ). Suchevolu- Withthis method, wecannumericallyfinda sequenceof tion of the power-dependent nonlinear response can also stable angles under different pump power PI. be observed by fixing the initial twist angle but chang- ing the pump frequency, as will be demonstrated in the SincetheEMtorqueisanonlinearfunctionoftwistan- experiment below. To confirm the feasibility of the proposed nonlinear (cid:1) e(cid:5)H(cid:7)Hq (cid:31) e(cid:13)!(cid:1)"q rotatable meta-atoms we carry out a pump-probe mi- 563 4 563 7(cid:31) crowave experiment. To experimentally realise a strong 5g3 5g3 nonlinear or even bistable effect, the restoring torque (cid:11)(cid:25)q543 (cid:11)(cid:25)q543 from the wire has to be sufficiently small so that the (cid:11) (cid:11) (cid:24)(cid:10) 93 (cid:24)(cid:10) 93 structure can be twisted by a large enough angle within (cid:11) (cid:11) (cid:22) e(cid:23) s3 (cid:22) e(cid:23) s3 the maximum available power. We found that rubber is θ3 θ3 a good candidate, since the shear modulus of rubber is 3 3 of the order of 0.2MPa – 2.4MPa [24], which is at least θ θHg d dHg g θ θHg d dHg g three orders of magnitude smaller than it is for other (cid:9)(cid:10)(cid:11)(cid:12)(cid:7)(cid:11)(cid:13)(cid:14)(cid:15) e/(cid:17)(cid:18)q (cid:9)(cid:10)(cid:11)(cid:12)(cid:7)(cid:11)(cid:13)(cid:14)(cid:15) e/(cid:17)(cid:18)q polymers. (cid:1)(cid:3)(cid:4)(cid:5)(cid:1) 543 (cid:1)(cid:31) e(cid:13)!"q& 33HH54 (cid:1)(cid:1)(cid:1)(cid:2)(cid:2)(cid:2)(cid:2)(cid:3)(cid:3)(cid:3)(cid:5)(cid:6)(cid:7)(cid:5)(cid:2)(cid:5)(cid:1)(cid:1)(cid:1) (cid:31)#7$(cid:31) (cid:24)(cid:10)(cid:11)(cid:11)(cid:25)q53633 (cid:1)(cid:1)(cid:2)(cid:2)==(cid:3)(cid:3)==(cid:7)(cid:6)(cid:5)(cid:2)==(cid:1)(cid:1) FraidgA.iu3ss.crhWe=meah3ta.i2vcmeoumfs,etdhtretawcekoxpwseeirpdiamtrhaent1etmdalmcos,eptcpuoepprpSiesRrRsthhsoicw(kinnnneeisnrs f G(cid:31) 3 (cid:22) e(cid:23)(cid:11) s3 (cid:1)(cid:1)(cid:2)==(cid:3)(cid:3)==(cid:5)(cid:4)(cid:5)(cid:5)==(cid:1)(cid:1) 0R.4003053µsmubasntrdatselist(wǫrid=th3.g5,=los0s.2tamnmge)ntpr0i.n0t0e2d7,osnubRsotrgaetres (cid:31)7 G(cid:31)& e(cid:14)q d3 (cid:2) e(cid:23)q thickness0.5mm). ThelowerSRRisfixedandpositioned g3 0g 533 54g 3 3H4g 3Hg 3H0g 5 at the centre of a WR229 rectangular waveguide with ◦ (cid:22) e(cid:23)(cid:11)(cid:24)(cid:10)(cid:11)(cid:11)(cid:25)q # e")z""4q Φ = 0 , and the top SRR is suspended with a thin rub- $ ber wire (radius a = 50µm, length d = 20mm), so that it canrotate aboutthe commonaxis. The twoSRRs are FIG. 2: The principle of nonlinear response in rotatable alignedcoaxially,withafacetofacedistanceof0.75mm, meta-atoms. (a) The mode amplitude Q2 and (b) the EM torque MEM of the top rotatable ring. (c) The EM torque separated by air. The horizontal positions of the SRRs at 3.5GHz for different pump powers from 0 to 1 mW/mm2 are carefully adjusted and the initial twist angle θ0 is in 0.2mW/mm2 steps, and the restoring torque for different set at around 70◦. The mass of the suspended sample initial twist angle θ0; (d) the corresponding paths of power- is 101mg, which leads to a 6.1% elongation of the wire. dependenttwist angles underdifferent θ0. TheYoung’smodulusisthusestimatedas2.06MPa,and 4 (cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)(cid:4)v(cid:10)(cid:4)(cid:2)(cid:11)(cid:6)(cid:12)(cid:4)(cid:13)(cid:4)(cid:14)(cid:15) 5(cid:12)(cid:12)(cid:2)(cid:17)v(cid:18)(cid:7)(cid:13)(cid:6)(cid:19)(cid:2)(cid:15)(cid:7)(cid:20)(cid:14) which indicates that the twist angle is increased. When (cid:28)rv )y(c Gl the pump frequency is at the red tail of the resonance,a w (cid:25)(cid:17)vsP )y(( GHql &vs(cid:8)(cid:4)(cid:5) llainregweisdptehc)trcaaln“jbuemopb”se(ravbeoduwtthherneethteimpeusmopftphoewreesropnaasnscees (cid:6)(cid:4)(cid:14) )yoH Hq (cid:12)(cid:4)(cid:4) a certain threshold value [Fig. 4(a)]. The thresholds are b(cid:12)(cid:4)(cid:24) )yoF s(cid:2)r s)r zl (cid:11)r different for increasing and decreasing pump powers. As v oq oH (o (F (G oq oH (o (F (G the pump frequency approaches to the initial resonance, the spectral “jump” becomes smaller [Fig. 4(c)] and fi- w(cid:28)rv )y(q G( nally disappears [Fig. 4(e)]. We also observed similar (cid:25)(cid:17)vsP )y(( GHcl &vs(cid:8)(cid:4)(cid:5) ereffdecttasil(noofttshheowannt)iswymhemnetthriecpmuomdpe,frienquwehniccyh icsaaset tthhee (cid:4)(cid:24)(cid:6)(cid:4)(cid:14) )yoz s(cid:25)r s(cid:8)r HF (cid:12)(cid:4)(cid:4)(cid:11)r tdwiroecrteisoonnoafnctheseaEpMprotoarcqhueea.ch other due to the opposite b(cid:12) )yoc HH v o( oG (( (Go( oG (( (G Tofurthervalidatetheobservedeffect,wenumerically rv simulated the exact experimental geometry and found (cid:28) w )y(q G( excellent agreement[21]. The estimated optimum initial (cid:17)vsP Gq &vs(cid:8)(cid:4) twist angle is around 72.5◦, and the maximum twist an- (cid:4)(cid:24)(cid:6)(cid:4)(cid:14)(cid:25) )y(( s(cid:4)r s-r GHHo (cid:5)(cid:12)(cid:4)(cid:4)(cid:11)r g3l.e2s1GobHtazinaneddf3o.2r3tGheHtzh)raeerepaurmoupndfre9q0u◦e,n8c5i◦esan(3d.1882G◦,Hrez-, b(cid:12) )yoz spectively [21]. Finally we remap these angles back to v ol oF oH (( (c ol oF oH (( (c the corresponding resonant frequencies according to the *(cid:20)+(cid:4)(cid:12)vs(cid:8)7(cid:13)r *(cid:20)+(cid:4)(cid:12)vs(cid:8)7(cid:13)r simulation spectra (see Fig. S3 (b), (d) and (f) of sup- plementary file). FIG. 4: Comparisons of resonant frequencies demonstrated The single meta-atom used in the waveguide exper- in a waveguide experiment and numerically calculated for an iment induces image currents in the waveguide walls. array. (a) and (b) pump at 3.18GHz; (c) and (d) pump at Thus, our experimental system is analogous to an array 3.21GHz; (e) and (f) pump at 3.23GHz. The stable twist angles forthearray system areshown on theright axes. The where neighbouring elements interact. Since the nonlin- power in thearray is thepower incident on each unit cell. earity arises from individual rotatable meta-atoms, the behaviourinanarraywillnotshowqualitativedifference from a single meta-atom as predicted in the analytical the shear modulus follows as G≈0.69MPa. The exper- model, as long as the neighbour interaction is relatively imental realization varies slightly from Fig. 1, however weak. Weusefullnumericalsimulationtomodelanarray the underlying physical mechanism of the nonlinear re- with thickness of a single cell and periodicity of 45mm sponseduetothedynamiccouplingbetweenelectromag- in the transverse directions, arranged as in Fig. 1. The neticandelasticpropertiesisthesame. Thisequivalence parametersofthe meta-atomsusedinthe simulationare is verified in the supplemental material [21]. thesameasintheexperiment. Thearrayandthewaveg- The transmission spectrum measurements are per- uide experiment show quite good qualitative agreement formedbyavectornetworkanalyser(RohdeandSchwarz (see Fig. 4(b), (d) and (f) for the power-dependent reso- ZVB-20). The CW pump signal is generated by a signal nanceandFig. S1(b)fortheEMtorque),thusjustifying generator(HP8673B)andisfurtheramplifiedbyapower that the nonlinear behaviour observed in the waveguide amplifier (HP 83020A)before being sentinto the waveg- experiment is similar to that of the analogous array sys- uide. Three pump frequencies (3.18, 3.21 and 3.23GHz) tem. are chosen in order to capture the evolution of the non- The novel nonlinear “meta-atom” described in this linear responseatdifferentdistances fromthe initial res- workprovedtopossessasensitiveelasticfeedbackbring- onance. The pump power is increased in 1dB steps; for ing nonlinearity to the interaction of EM modes of the each step, the sample reaches steady state after 30 sec- resonators. The resulting nonlinearity and bistability of onds. The mechanical rotation is quite significant and the response was successfully observed in experiments can be visually observed in the experiment, thus ruling anditturnsoutthatthese resultscanbe accuratelypre- outotherpossiblenonlinearmechanismssuchasheating. dictedwiththeoreticalmodelling. Wealsonotethatthis The experimentally observed transmission spectra are structure is chiral (except in the high symmetry cases ◦ ◦ shown in Fig. S2 of the supplementary file, and the cor- of 0 and 180 angle between the rings), it should also respondingresonantfrequenciesaredepictedinFig.4(a), exhibit nonlinear optical activity, an effect which is rela- (c)and(e),wherethepredictedevolutionfrombistability tivelyweakinnaturalmedia[25],butcanbequitestrong tosmoothnonlinearityisclearlyshown. Theinitialreso- in metamaterials [26]. nance(symmetricmode)withoutpumpislocatedaround Althoughtheexperimentaldemonstrationinthiswork 3.256GHz,anditred-shiftsasthepumppowerincreases, was performed in the microwave frequency range, the 5 general principle of operation is valid at any frequency Y. S. Magnetoelastic metamaterials. Nature Materials where aresonantresponsecanbe excitedin suchor sim- 11, 30–33 (2012). ilar metamaterials elements, and the way to analyse the [12] Lapine,M.,Shadrivov,I.&Kivshar,Y. Wide-bandneg- ativepermeability ofnonlinear metamaterials. Scientific samephenomenainTHzoropticalrangeisconceptually Reports 2, 412 (2012). the same. [13] Marquardt, F. & Girvin, S. M. Optomechanics. Physics We believe that this work provides a substantial con- 2, 40 (2009). tribution to the emerging area of optomechanical and [14] Lapine, M., Shadrivov, I. V., Powell, D. A. & Kivshar, magnetoelasticmetamaterials,andoffersanefficientand Y. S. Metamaterials with conformational nonlinearity. convenient design for practical applications. Scientific Reports 1, 138 (2011). This work was supported by Australian Research [15] Powell, D. A., Lapine, M., Gorkunov, M., Shadrivov, I. V. & Kivshar, Y. S. Metamaterial tuning by manipu- Council. The authors are grateful to A.A. Sukhorukov lation of near-field interaction. Phys. Rev. B 82, 155128 for helpful discussions. (2010). [16] Powell, D., Hannam, K., Shadrivov, I. & Kivshar, Y. Near-field interaction of twisted split-ring resonators. Phys. Rev. B 83, 235420 (2011). [17] Liu, H. et al. Coupled magnetic plasmons in metamate- [1] Solymar, L. & Shamonina, E. Waves in Metamaterials rials. Phys. Stat. Sol. B 246, 1397–1406 (2009). (Oxford University Press, 2009). [18] Lebedev, P. N. Experimental examination of light pres- [2] Marqu´es, R., Mart´ın, F. & Sorolla, M. Metamaterials sure. Ann. der Physik 6, 433 (1901). with negative parameters (Wiley,2008). [19] Liu, M., Powell, D. A. & Shadrivov, I. V. Chiral meta- [3] Smith,D.R.,Padilla, W.J., Vier,D.C., Nemat-Nasser, atoms rotated by light. Appl. Phys. Lett. 101, 031105 S. C. & Schultz, S. Composite medium with simultane- (2012). ouslynegativepermeabilityandpermittivity. Phys. Rev. [20] Liu,M.,Powell,D.A.,Shadrivov,I.V.&Kivshar,Y.S. Lett. 84, 4184–4187 (2000). Optical activity and coupling in twisted dimer meta- [4] Pendry,J. B. Electromagnetic materials enter the nega- atoms. Appl. Phys. Lett. 100, 111114 (2012). tiveage. Physics World 14, 47 (2001). [21] See SupplementalMaterial [5] Shalaev, V. M. Optical negative-index metamaterials. [22] Povinelli, M.L.et al. Evanescent-wavebondingbetween Nature Photonics 1, 41–48 (2007). opticalwaveguides. OpticsLetters 30,3042–3044(2005). [6] Li, J. & Chan, C. T. Double-negative acoustic metama- [23] Jackson, J. D. Classical electrodynamics (Wiley (New terial. Phys. Rev. E 70, 055602 (2004). York),1999). [7] Norris, A. N. Acoustic cloaking theory. Proc. Royal [24] Gent, A. N. Engineering with rubber: How to design Soc. A 464, 2411–2434 (2008). rubber components (HanserGardner Pubns,2001). [8] Nicolaou, Z. & Motter, A. Mechanical metamaterials [25] Kielich, S. Natural and magneto-optical rotation in the with negative compressibility transitions. Nature Mate- presenceofan intenselight beam. Physics Letters A25, rials 11, 608–613 (2012). 517–518 (1967). [9] Lapine,M.et al. Structuraltunabilityinmetamaterials. [26] Shadrivov,I.V.,Fedotov,V.A.,Powell, D.A.,Kivshar, Appl. Phys. Lett. 95, 084105 (2009). Y.S.&Zheludev,N.I.Electromagneticwaveanalogueof [10] Liu, A.-Q., Zhu, W. M., Tsai, D. P. & Zheludev, N. I. an electronic diode. New Journal of Physics 13, 033025 Micromachinedtunablemetamaterials: areview. J.Opt. (2011). 14, 114009 (2012). [11] Lapine, M., Shadrivov, I. V., Powell, D. A. & Kivshar,