1720 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 57, NO. 6, JUNE 2009

Free-Space Imaging Beyond the Diffraction LimitUsing a Veselago-Pendry Transmission-Line

Metamaterial SuperlensAshwin K. Iyer, Student Member, IEEE, and George V. Eleftheriades, Fellow, IEEE

Abstract—Focusing using conventional lenses relies on the col-lection and interference of propagating waves, but discounts theevanescent waves that decay rapidly from the source. Since theseevanescent waves contain the finest spatial details of the source,the image suffers a loss of resolution and is referred to as “diffrac-tion-limited.” Superlensing is the ability to create an image withfine features beyond the diffraction limit, and can be achieved witha “Veselago-Pendry” lens made from a metamaterial. Such a Vese-lago-Pendry superlens for imaging in free space must be stringentlydesigned to restore both propagating and evanescent waves, butmeeting these design conditions (isotropic � � � � �

�) has proven difficult and has made its realization elusive. Wedemonstrate free-space imaging with a resolution over three timesbetter than the diffraction limit at microwave frequencies usinga Veselago-Pendry metamaterial superlens based on the negative-refractive-index transmission-line (NRI-TL) approach, which af-fords precise control over its properties and is also less susceptibleto losses than other approaches. A microwave superlens can beparticularly useful for illumination and discrimination of closelyspaced buried objects over practical distances by way of back-scat-tering, e.g., in tumour or landmine detection, or for targeted irradi-ation over electrically small regions in tomography/hyperthermiaapplications.

Index Terms—Diffraction limit, focusing, left-handed (LH),metamaterial, negative refractive index (NRI), periodic struc-tures, superlenses, superresolution, transmission line.

I. INTRODUCTION

T HE long-held interest in synthesizing known materialproperties artificially has been revived by the impressive

recent developments in metamaterials, which are artificialmaterials engineered to exhibit electromagnetic phenomena notavailable or not readily available in nature, such as a negativepermeability ( ) and a negative permittivity ( ), as well astheir most unusual product: a negative refractive index (NRI).Through the theoretical work of V. Veselago [1] and J. B.Pendry [2], it is known that flat NRI lenses designed to meetcertain strict conditions are able to focus the propagating-wavecomponents of a source without geometric aberration whilesimultaneously restoring the amplitude of its evanescent-wavecomponents, which decay quickly as they depart the source,

Manuscript received January 16, 2008; revised October 27, 2008. Current ver-sion published June 03, 2009. This work was supported by the Natural Sciencesand Engineering Research Council of Canada (NSERC).

The authors are with the Edward S. Rogers Sr. Department of Electrical andComputer Engineering, University of Toronto, Toronto, ON M5S 3G4, Canada(e-mail: iyer@waves.utoronto.ca; gelefth@waves.utoronto.ca).

Digital Object Identifier 10.1109/TAP.2009.2019890

such that the focal plane contains an exact real image of thesource, down to its finest features. As a result, such a lens,appropriately termed a “Veselago-Pendry superlens,” is ableto overcome the constraints of the classical diffraction limit,which restricts focusing with conventional lenses to a resolutionon the order of half the wavelength of illumination. PracticalVeselago-Pendry superlenses have many potential biomedical,microelectronics, and defense-related applications in subd-iffraction microscopy, lithography, tomography, and sensing.

Although such a lens has been realized using negative-re-fractive-index transmission-line (NRI-TL) metamaterials forembedded sources in planar and three-dimensional form [3],[4] and inside waveguide environments [5], [6], it has, so far,eluded practical realization in a form capable of interactingwith sources in free space, the form in which it was first en-visioned. This is largely due to the fact that a true free-spaceVeselago-Pendry superlens has a number of stringent design re-quirements: first, the lens must possess andfor the polarization(s) concerned (where and are thefree-space permeability and permittivity, respectively) in orderto be impedance-matched to free space and simultaneously pos-sess an effective refractive index , which also renders itaberration-free. Since these materials are necessarily dispersive,achieving these values with adequate precision requires a meansof tightly controlling the metamaterial’s frequency response.Second, the lens must be extremely low loss and be adequatelythin, since both loss and electrical thickness serve to quicklydegrade the resonant evanescent enhancement contributing tosubdiffraction imaging. A third condition follows from theprevious: the unit cells comprising the lens must themselvesbe deeply subwavelength in size in order to minimize spatialanisotropy and to ensure that the structure possesses the desiredbulk response. Last, the transverse dimensions of the lens mustbe large enough that the lens can be illuminated by a source infree space.

True Veselago-Pendry superlenses (that is, metamaterials ca-pable of restoring both the propagating and evanescent spec-trum of a source) have not yet been realized for subdiffractionimaging in free space, but many other varieties of metamate-rial have experimentally demonstrated phenomena akin to sub-diffraction imaging by associated physical mechanisms. Theseinclude the plasmonic silver film [7], the magneto-inductivelens [8], and the swiss-roll structure [9] which, although theydo not possess a negative refractive index and so cannot focusthe propagating wave numbers, do recover fine spatial featuresthrough evanescent enhancement; unfortunately, this means that

0018-926X/$25.00 © 2009 IEEE

IYER AND ELEFTHERIADES: FREE-SPACE IMAGING BEYOND THE DIFFRACTION LIMIT 1721

sources must be placed very near to, if not directly against, thelens faces, which imposes an extremely short working distance.Although the Veselago-Pendry superlens also requires that thesource be placed in the near field of the lens, these distancesare on the order of , which may be appreciable at RF/mi-crowave frequencies. Subdiffraction imaging phenomena havebeen successfully extended to the far field using magnifying“hyperlenses” [10]–[12], but this class of superlenses relies ex-plicitly on anisotropy and sources are, once again, applied di-rectly to the hyperlens face; as a result, the working distance onthe source side remains limited. Subdiffraction imaging usingprinted split-ring-resonator- (SRR-) based structures has beenreported in the way of transversely and longitudinally confinedsubwavelength focal “spots” [13], [14] in spite of their highlosses and/or large electrical thickness; however, calculationsbased on [15] suggest that the reported loss, lens thickness, andobserved resolution ability are inconsistent, if the structures areto be regarded as true Veselago-Pendry superlenses. Further-more, the lenses described in these works employ a refractiveindex of , which defies even the basic requirement thatthe refractive index of a free-space Veselago-Pendry superlensbe ; contrary to the authors’ claim in [13], the resolutionability of a Veselago-Pendry superlens is not tolerant to suchlarge deviations in its effective parameters. Moreover, the use ofa higher index reduces the working distance between lens andimage. Last, the authors’ use of large (resonant) antennas to de-tect the fields precludes the observation of the true evanescentfield magnitudes, which necessarily manifest themselves onlyin a transverse (and not longitudinal) confinement of the fieldsat the focal plane [3] – in fact, as discussed in [16], the use ofsuch large antennas strongly perturbs the fields, making the ob-served resolution more a function of the antenna measurementthan of the fields produced by the metamaterial. Thus, the resultsdescribed in these works are inconsistent with the imaging prin-ciples of the Veselago-Pendry superlens and the requirementsoutlined above; indeed, in attempting to explain these inconsis-tencies, the authors of [13], [14] have speculated that anisotropy,rather than subdiffraction imaging by way of the restorationof evanescent waves, may be responsible for these phenomena[14]. Although these factors preclude their description as Vese-lago-Pendry superlenses, further research into such structuresmay reveal other intriguing mechanisms by which subdiffrac-tion imaging can be achieved – one possibility was suggested in[17].

The first successful attempt at superlensing [3], althoughin a planar transmission-line (TL) environment rather thana free-space environment, employed NRI-TL metamaterials,which consist of a fine TL grid loaded with inductors andcapacitors to control their effective-medium response [18].Subsequently, it was considered that a “volumetric” NRI-TLmetamaterial could be realized for free-space excitation bystacking planar NRI-TL metamaterials in a multilayer fashion[19], [20]. While not three-dimensionally isotropic and alsopolarization-specific, such a structure would appear isotropic totwo-dimensional excitations (i.e., an infinite line source) in freespace and could be fabricated easily and rapidly using widelyavailable printed-circuit-board (PCB) technologies. Previously,such a structure was realized using fully printed loading el-

ements (interdigitated capacitors and meandered inductors,and without any vias) and demonstrated diffraction-limitedfocusing of a free-space magnetic dipole source consistent withthe use of large unit cells and an electrically thick lens [21].However, it was also suggested in that work that the unit cellsmay be made simultaneously low-loss and electrically smallby exploiting the strong lumped loading afforded by discretechip inductors and capacitors with high quality factors, as inthe planar case, which would also enable the realization of anadequately thin, NRI-TL metamaterial, free-space superlens.Accordingly, this work presents a free-space NRI-TL superlensdesigned to demonstrate Veselago-Pendry superlensing at

( ). The use of discrete(chip), low-loss inductors and capacitors results in unit cellsizes of and a total lens thickness less than at thedesign frequency. Experimental verifications of both free-spacesubdiffraction focusing and free-space superresolution arepresented. These results show that the so-far elusive free-spaceVeselago-Pendry superlens is, indeed, realizable, and arguablyrepresent the first such realization.

II. DESIGN

The NRI-TL approach is known to offer intrinsically largeNRI bandwidths and minimize losses through the tight cou-pling between unit cells [22]. The strength of the lumped el-ements loading the host TL medium renders the unit-cell sizedeeply subwavelength, allows the realization of an adequatelythin Veselago-Pendry lens, affords precise control over its ef-fective-medium properties, and mitigates losses. The TL hostmedium constituting the layers of the volumetric NRI-TL lensemploys a topology known as the “series” NRI-TL node, whichintersects four 1D NRI-TLs in series and so obviates the need forvias [19]. The most suitable TL for this topology is the fully uni-planar co-planar strip (CPS) TL. When appropriately loaded byinductors and capacitors in the NRI-TL configuration [18], thistopology may alternately be viewed as a uniplanar array of ca-pacitively loaded rings connected to each other using inductors.The substrate medium was chosen to be 0.5-oz.-copper-cladRogers RO3003 60-mil (1.524 mm) microwave substrate with

and . The surface-mount elementsselected were Coilcraft high- ultrasmall wirewound induc-tors (19.3 nH at 2.40 GHz) and American Technical Ceramicshigh- capacitors (1.2 pF). The particular geometrical featuresof the ring were decided to accommodate the mounting patternsof the surface-mount components suggested by their manufac-turers, and are shown in Fig. 1.

The optimal vertical layer spacing of 3.476 mm was decidedin concert with an equivalent-circuit model [19], [20] that ac-curately describes the effective-medium properties of the meta-material to yield effective-medium parameters and

at an operating frequency of( in free space) when illuminated by -polarizedfields propagating in the layer planes (i.e., -polarized magneticfields). As previously mentioned, the unit cells in the present de-sign were chosen to be nearly in size (7.14-mm square),and so a lumped description is justified.

1722 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 57, NO. 6, JUNE 2009

Fig. 1. Relevant dimensions of co-planar strip (CPS) transmission-line (TL)host medium.

III. SIMULATION

Simulations of the designed volumetric NRI-TL topologywere performed using Ansoft’s High-Frequency Structure Sim-ulator (HFSS) [23], which is a full-wave finite-element-method(FEM) electromagnetic field simulator. The ultimate goalof extracting the effective-medium parameters of a periodicmedium requires knowledge of its intrinsic effective propaga-tion constant, , and wave impedance, . The former may beobtained as a function of frequency by examining the spectrumof resonances within a single unit cell that are associated withparticular periodic boundary-phase conditions. In HFSS, thistype of simulation is known as an “eigenmode simulation,”which can yield in a particular direction of propagationas a function of (the “dispersion diagram”), or over alldirections of propagation at a particular frequency (the “isofre-quency” or “equifrequency” surface). The wave impedancemay be obtained in a TL environment of known geometry(e.g., parallel-plate waveguide) by way of its characteristicimpedance, . In HFSS, this, along with a measure of , isobtained through the scattering ( -) parameters produced by aslab of the periodic medium that is finite (i.e., consisting of afinite number of unit cells) in the direction of propagation, butmay be rendered infinite in other directions through the use ofappropriate periodic boundary conditions. This is known as a“driven simulation,” since a practical source (e.g., a normallyincident plane wave) is used. By associating the obtained

-parameters with a homogeneous slab of equal thickness, itseffective homogeneous properties can be obtained by way of anextraction procedure that is approximately valid for electricallythin slabs [24] consisting of electrically very small unit cells;certainly, the volumetric NRI-TL lens falls into this category,and it has further been verified through simulations of slabsof three-, four-, and five-cell thicknesses that the extractedparameters are unique.

In all cases, the lumped components were modelled bycurrent sheets endowed with the appropriate lumped-elementboundary conditions; dielectric loss tangents and component

Fig. 2. Dispersion characteristics and transmission/reflection magnitudes of afive-unit-cell-thick volumetric NRI-TL metamaterial, (full-wave simulations –circles, equivalent-circuit model – curves): (a) �� (solid red curve and circles).The inset presents � ����� (dashed blue curve and circles) and � �����(dotted green curve and circles) near � � �� � ���� ��, where the su-perlensing condition � �� ��� � � �� ��� � � is met; (b) transmis-sion magnitude �� � (solid red curve and circles); reflection magnitude �� �(dashed blue curve and circles).

losses (as specified in their data sheets in terms of either equiv-alent series resistance or quality factor) were included, and themetallic features were specified as copper with a bulk conduc-tivity reduced by over 70% from the nominal value to accountfor surface roughness, by way of HFSS’s “finite-conductivity”boundary condition.

The band structure and transmission/reflection magnitudesfor propagation of a normally incident plane wave througha five-unit-cell-thick volumetric NRI-TL metamaterial slab(thickness ) are shown in Fig. 2.Also shown in the inset of Fig. 2(a) are the relative effectivepermittivity and permeability extracted from the -parameters,in the vicinity of the design frequency, 2.40 GHz. These arerepresented by circles superposed on the theoretical curvesobtained from the volumetric NRI-TL equivalent-circuit modeldescribed in [19], [20]. These data prove the excellent agree-ment between the two, especially where the phase shift per unitcell is small (corresponding to the effective-medium limitin which effective material parameters and aredefined). It should be noted that there are certain frequencyregions in which the extracted are omitted; these are regionsin which the extraction methods fail due to the resonances thatcause large phase shifts per unit cell. The data that remainare generally confined to the region , where theextraction procedure can be regarded to be valid. The NRIregion, exhibiting a prominent backward-wave characteristic(opposite phase and group velocities), is evident between ap-proximately 2.10 GHz and 2.60 GHz, representing a fractionalNRI bandwidth of over 21%. It is also apparent from the reflec-tion magnitude ( – dashed blue curve and circles) that themetamaterial is very well matched over this frequency range,

IYER AND ELEFTHERIADES: FREE-SPACE IMAGING BEYOND THE DIFFRACTION LIMIT 1723

Fig. 3. Isofrequency contours at 2.409 GHz for the volumetric NRI-TL meta-material describing propagation in the layer planes (solid blue curve – obtainedthrough full-wave FEM simulations) and for propagation in free space (redsquares). Their near-perfect coincidence suggests that the volumetric NRI-TLmetamaterial exhibits a nearly isotropic refractive index of�� at this frequency.

and particularly well matched near 2.40 GHz, where the returnloss is better than . Although not depicted in the figure,the NRI metamaterial dispersion curve intersects the (PRI) lightline near 2.40 GHz, which, along with good matching and theelectrical thinness of the slab at this frequency, uniquely sug-gests that the metamaterial possesses real effective permittivityand permeability values andat the design frequency. Indeed, the extracted permeability andpermittivity correspond to the equivalent-circuit theory andintersect at relative values of almost exactly at 2.40 GHz.The insertion losses in the passband remain near 0.2 dB per unitcell, which is consistent with the use of low-loss materials andlumped components with high quality factors, and also with therequirements of the Veselago-Pendry superlens. This shouldbe compared to an insertion loss of over 2 dB per unit cell forother metamaterials reporting similar characteristics (see, forexample, [14, Fig. 1]).

An isofrequency surface at 2.409 GHz (a difference of lessthan 0.4% from the design frequency, and, well within the tol-erance due to discretization of the FEM mesh) is depicted inFig. 3 and is shown superposed on the isofrequency surfacecorresponding to propagation in free space (the “light cone”).The coincidence of the two curves suggests that the volumetricNRI-TL metamaterial possesses isotropic and

at this frequency for all directions of propa-gation within the plane, as required by a true Veselago-Pendrysuperlens.

Although the discussion so far has concentrated on real ef-fective permeability and permittivity, the inclusion of loss in theequivalent-circuit model yields the full complex parameters. Atthe design frequency, the complex effective permeability andpermittivity are and

. It is evident that islarger than , which raises concerns given that losses quicklydegrade the resolution ability of Veselago-Pendry superlenses.However, the volumetric NRI-TL metamaterial operates on the

-polarized fields that, in the electrostatic limit, are susceptibleto alone. Indeed, the resolution of the lens can be esti-mated by inserting into the resolution enhancement equationof [15], suitably adjusted for the -polarization, which yields

. Although these numbers are approximate and relyon estimates of component and material losses, they suggest thatthe volumetric NRI-TL superlens, as designed, may be able torecover evanescent wavenumbers nearly three times larger thanthe largest propagating transverse wavenumber .

IV. EXPERIMENT

The fabricated multilayer NRI-TL metamaterial Veselago-Pendry lens is shown in Fig. 4(a). The inset depicts the loadingon a single layer, from which the reader may make out the CPSTL, series capacitors (oriented at with respect to the CPSTL axes) and shunt inductors (oriented at 0 or 90 with respectto the CPS TL axes).

The lens consists of 43 layers, each containing a 21 5array of NRI-TL unit cells. Since each unit cell measures

square, the lens measures wideby thick. The layers were held rigidly in placeusing a plastic frame designed to maintain the layer period of3.476 mm, resulting in a total lens height of .These dimensions are such that a source placed a distance of

from the front face of the lens encounters a numerical aper-ture of 0.97, which collects most of its propagating spectrum(the restoration of the evanescent spectrum is a function of thelens design and the proximity of the source and lens); thus, thephysical size of the NRI-TL does not impose severe restrictionson its imaging ability. The transverse dimensions are approxi-mately , which provides a sufficient illumination area andminimizes diffraction around the edges.

Following chemical etching of the substrate materialand contour routing of the layers, the nearly 26000 sur-face-mount components were precisely placed using a SiemensSIPLACE assembly and placement system. The boards werethen cut manually and inserted into the plastic frame. Themeasurement apparatus consisted of a source loop antenna( ) and detector loop antenna( ) connected to the terminals of anAgilent E8364B performance network analyzer (PNA). Theantennas were constructed from a semi-rigid 50- microwavecoaxial cable (1.19-mm outer diameter), and employed a“shielded” topology that provides magnetic-dipole-type fieldswhile simultaneously minimizing unwanted radiation fromunbalanced currents on the coaxial feeding structure [25]. Ithas been shown that the fields in the volumetric metamaterialremain strongly confined to the layer planes, and so, providedthat the fields at the output are measured in the same horizontalplane, the single loop antenna appears to excite the affectedlayers like an infinite line source [21]. The source (illuminating)antenna was placed at a distance of from thefront face of the lens, and the detector loop antenna was affixedto a computer-controlled -translator and scanned behind

1724 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 57, NO. 6, JUNE 2009

Fig. 4. (a) Photograph of fabricated NRI-TL lens with inset showing lumpedloading of the host coplanar-strip TL structure using discrete surface-mount in-ductors and capacitors. (� � � � � � ����� �� � ����� �� � ���� ��,� � ���� ��); (b) Measurement arrangement – the marquee indicates the re-gion in which the field data presented in the following figures are measured.

the lens in the plane of excitation for field magnitude and phasedistributions suggestive of focusing and evanescent decay. Theuse of a larger loop for the source produces strong fields thatare easier to detect, and the use of a small loop for the receiverallows the detection of these fields without disturbing them. Themeasurement apparatus was covered in a microwave absorberand transmission measurements between the two antennas weretaken in intervals of 4 mm ( ). The data were averaged30 times to minimize noise in the measurement. Fig. 4(b)

Fig. 5. Raw measured magnitude and phase data for excitation with a single-loop source when (a) the lens is absent and (b) the lens is present. The blackcurves trace the half-power contours referenced to the maximum field magni-tude at the focal plane (dashed line); (c) A comparison of the normalized magni-tude profiles (linear scale) at the focal plane when the lens is absent (blue circles)and when the lens is present (red squares), along with the fields at a distance of�� from the source when the lens is absent (solid black curve). The dottedhorizontal line indicates the half-power levels and shows that the NRI-TL su-perlens is able to produce an image of the source with a half-power beam widthof ���� (minimum peak-to-null width of ���� ).

shows the measurement arrangement; the marquee identifiesthe region in which the fields are sampled and represents theregion in which the data of Figs. 5–6 are taken.

Fig. 5(a) and (b) presents the raw measured field magnitudeand phase data over the measurement region at the operatingfrequency of 2.40 GHz for two cases: Fig. 5(a) shows the re-sults of a control experiment in which the fields in free space

IYER AND ELEFTHERIADES: FREE-SPACE IMAGING BEYOND THE DIFFRACTION LIMIT 1725

Fig. 6. Raw measured magnitude and phase data for excitation with two loopsources separated by � ��when (a) the lens is absent and (b) the lens is present.The black curves trace the half-power contours referenced to the maximum fieldmagnitude at the focal plane (dashed line); (c) A comparison of the normalizedmagnitude profiles (linear scale) at the focal plane when the lens is absent (bluecircles) and when the lens is present (red squares), along with the fields at adistance of ��� from the source when the lens is absent (solid black curve). Thedotted horizontal line indicates the half-power levels and shows that the NRI-TLsuperlens comfortably differentiates the sources.

are measured before the lens is inserted; Fig. 5(b) presents mea-surements over the same spatial region with the lens in place.

The black curves indicate half-power contours referenced tothe maximum field magnitudes at the expected focal plane, in-dicated in each case by a dashed line. Since the focal planelies at a distance of from the source plane, the evanes-cent-wave components have all but disappeared in the field mag-nitude distribution of Fig. 5(a), and the half-power contour is

approximately in width. The nature of the phase frontsin Fig. 5(a) also suggests that we detect only propagating fieldsemanating away from a source located at their phase centre.However, the situation is dramatically different when the lensis inserted: the field magnitudes of Fig. 5(b) indicate the for-mation of a tightly confined focal region whose transverse half-power width at the focal plane is less than (minimumpeak-to-null width of ), over four times narrower thanwithout the lens and over 3.3 times narrower than that predictedfor diffraction-limited images ( in free space). The nor-malized magnitude profiles at the focal plane, as well as themagnitude profile a distance of from the source are com-pared in Fig. 5(c). The evanescent nature of the fields is sug-gested both by the expected decay in the field magnitude dis-tribution of Fig. 5(b) and also by the phase data of Fig. 5(b),which remain nearly constant in the focal region and assume apropagating characteristic further from the focal plane, wherethe strong evanescent fields have decayed and only the propa-gating fields remain. It is worth noting that the unnormalizedpeak field intensity of the image measured at the focal planewith the lens in place is 1.7 dB higher than that observed overthe same free-space distance without the lens in place, attestingto the low-loss nature of the NRI-TL superlens.

Although the equivalent-circuit theory and full-wave simula-tion data predict that and at2.40 GHz, the existence of appreciable sidelobes in the imageof Fig. 5(c) indicate that these parameters may deviate slightlyfrom their ideal values at 2.40 GHz. Indeed, it has been veri-fied analytically (although the results are not presented in thiswork) that a shift of even 0.5% from the operating frequency(12 MHz), although it does not severely degrade the resolutionability, results in appreciable sidelobe levels for a lens with in-finite transverse dimensions; the finite transverse dimensions ofthe lens in the present study may serve to further enhance thesidelobe levels through reflections at the edges.

To ensure that the resolution ability of the NRI-TL super-lens enables the discrimination of two closely spaced sources,the single source antenna was substituted with two identicalshielded-loop antennas (each with a 21-mm diameter) fed co-herently using a passive microwave power splitter. Due to mu-tual coupling, the use of practical sources at close range limitsthe available resolution; in fact, the minimum separation of thesources required to resolve them at their half-power levels evenat the front face of the lens, where the decaying evanescent spec-trum is collected, was experimentally determined to be between

and ; the former separation distance was chosen tosimplify the analysis of the findings. Fig. 6(a) and (b) presentsthe raw measured field data in the absence and presence, respec-tively, of the NRI-TL superlens.

Once again, the black curves are half-power contours normal-ized to the maximum field amplitudes at the focal plane (dashedline). Fig. 6(c) presents the normalized magnitude profiles ofthe two sources at the front face of the lens, at the image planewithout the lens in place, and at the image plane with the lensin place. It is evident from these data that the NRI-TL superlensis able to recover the fine distinguishing features that are lostwhen the lens is absent, and nearly reproduces the sources sep-arated by . This result represents a resolution ability nearly

1726 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 57, NO. 6, JUNE 2009

twice as good ( ) as that offered by conventional lensesconstrained by the diffraction limit, and further testing of moreclosely spaced sources promises to reveal an even better reso-lution ability, possibly in line with that measured for a singlesource. The evanescent decay of the images from the exit faceof the superlens is also evident. The difference in the levels ofthe two recovered images can be attributed to the constructionof the small antennas, and slight horizontal misalignment be-tween the two sources and also between the source and detectoras the fields are scanned. The unnormalized peak field intensityof the two images is approximately 1.4 dB higher than that ob-served over the same distance without the lens in place, onceagain attesting to the low-loss nature of the NRI-TL superlens.Furthermore, although the NRI-TL superlens is designed for asingle operating frequency of 2.40 GHz, it was observed thatthese two sources could be comfortably resolved at theirlevels from 2.38 to 2.42 GHz [17]; this represents a fractionalbandwidth of nearly 1.7% over which the superlens is able tomaintain a resolution enhancement of .

From its very early stages, this work was motivated by thedesire to see metamaterial superlenses applied to the imagingof small scatterers at practical focal distances, as in close-rangenon-invasive tumour detection or land-mine detection. Al-though the experimental results presented so far pertain toluminous sources, the detection of non-luminous objects byway of backscattered fields may also be facilitated by a meta-material superlens. For example, it has been proposed that asingle luminous source at the front of the lens can be used bothto illuminate a scatterer behind the lens and detect it by wayof its backscattered secondary fields [26]. Alternatively, it ispossible to illuminate the front side of the lens using a normallyincident plane wave. Since the Veselago-Pendry lens does notpossess a unique principal axis, such a plane wave passes di-rectly through the lens and impinges upon the scatterer, whosebackscattered fields are then refocused at the front side of thelens where they may be detected. Both cases are depicted inFig. 7, where the red solid arrows indicate the directions of theilluminating fields and the blue open arrows indicate those ofthe backscattered fields. Since the antennas employed wouldnecessarily be small, matching techniques or time-gating couldbe employed to isolate the desired backscattered signals.

V. CONCLUSION

We have presented experimental evidence of free-spaceimaging beyond the diffraction limit at 2.40 GHz using a Vese-lago-Pendry superlens based on a multilayer implementationof the NRI-TL metamaterial. The NRI-TL approach affordsprecise control over the material parameters of the lens and ismuch less susceptible to losses than other methods. Moreover,this approach is supported by a good agreement between anequivalent-circuit theory and full-wave simulations. The exper-imental results reveal focusing of a single source to a minimumpeak-to-null beamwidth of less than one-sixth of a wavelengthand a resolution of two sources displaced transversely by adistance of one-third of a wavelength, both well below the clas-sical diffraction limit. The multilayer NRI-TL implementationis attractive because its constituent layers may be easily andrapidly fabricated using existing PCB fabrication techniques

Fig. 7. Detection of non-luminous objects behind the lens via backscatteredfields. The objects may be illuminated either by the detecting antenna or by anormally incident plane wave, which passes directly through the lens withoutfocusing. The red arrows indicate the directions of the illuminating fields, andthe blue arrows indicate those of the backscattered fields.

and facilities. This simplified approach to the realization ofVeselago-Pendry superlenses should encourage their applica-tion to imaging problems in biomedicine, microelectronics, anddefense.

ACKNOWLEDGMENT

The authors would like to thank Rogers Corporation, SaturnElectronics, and J. McIntyre and M. Forge of George BrownCollege for their help in fabricating the components of theNRI-TL metamaterial superlens.

REFERENCES

[1] V. G. Veselago, “The electrodynamics of substances with simultane-ously negative values of � and �,” Sov. Phys. Usp, vol. 10, no. 4, pp.509–514, Jan.-Feb. 1968.

[2] J. B. Pendry, “Negative refraction makes a perfect lens,” Phys. Rev.Lett., vol. 85, no. 18, pp. 3966–3969, Oct. 2000.

[3] A. Grbic and G. V. Eleftheriades, “Overcoming the diffraction limitwith a planar left-handed transmission-line lens,” Phys. Rev. Lett., vol.92, p. 117403, Mar. 2004.

[4] P. Alitalo, S. Maslovski, and S. Tretyakov, “Experimental verificationof the key properties of a three-dimensional isotropic transmission-linesuperlens,” J. Appl. Phys., vol. 99, p. 124910, Jun. 2006.

[5] S. M. Rudolph and A. Grbic, “Volumetric negative-refractive-indexmedium exhibiting broadband negative permeability,” J. App. Phys.,vol. 102, p. 013904, Jul. 2007.

[6] M. Stickel, F. Elek, J. Zhu, and G. V. Eleftheriades, “Volumetric nega-tive-refractive-index metamaterials based upon the shunt-node trans-mission-line configuration,” J. App. Phys., vol. 102, p. 094903, Jul.2007.

[7] N. Fang, H. Lee, C. Sun, and X. Zhang, “Sub-diffraction-limited op-tical imaging with a silver superlens,” Science, vol. 308, no. 5721, pp.534–537, Apr. 2005.

[8] F. Mesa, M. J. Freire, R. Marquès, and J. D. Baena, “Three-dimensionalsuperresolution in metamaterial slab lenses: Experiment and theory,”Phys. Rev. B, vol. 72, p. 235117, Dec. 2005.

IYER AND ELEFTHERIADES: FREE-SPACE IMAGING BEYOND THE DIFFRACTION LIMIT 1727

[9] M. C. K. Wiltshire, “Radio-frequency (RF) metamaterials,” Phys.Status Solidi B, vol. 244, no. 4, pp. 1227–1236, Mar. 2005.

[10] A. Salandrino and N. Engheta, “Far-field subdiffraction optical mi-croscopy using metamaterial crystals: Theory and simulations,” Phys.Rev. B, vol. 74, p. 075103, Aug. 2006.

[11] Z. Jacob, L. V. Alekseyev, and E. Narimanov, “Optical hyperlens: Far-field imaging beyond the diffraction limit,” Opt. Express, vol. 14, no.18, p. 8247, Sep. 2006.

[12] Z. Liu, H. Lee, Y. Xiong, C. Sun, and X. Zhang, “Far-field optical hy-perlens magnifying sub-diffraction-limited objects,” Science, vol. 315,no. 5819, p. 1686, Mar. 2007.

[13] K. Aydin and E. Ozbay, “Focusing of electromagnetic waves by a left-handed metamaterial flat lens,” Opt. Express, vol. 13, no. 22, p. 8759,Oct. 2005.

[14] K. Aydin, I. Bulu, and E. Ozbay, “Subwavelength resolution with anegative-index metamaterial superlens,” App. Phys. Lett., vol. 90, p.254102, Jun. 2007.

[15] D. R. Smith, D. Schurig, M. Rosenbluth, S. Schultz, S. A. Ramakr-ishna, and J. B. Pendry, “Limitations on subdiffraction imagingwith a negative refractive index slab,” App. Phys. Lett., vol. 82, pp.1506–1508, Mar. 2003.

[16] R. Marquès, M. J. Freire, and J. D. Baena, “Theory of three-dimen-sional subdiffraction imaging,” App. Phys. Lett., vol. 89, p. 211113,Nov. 2006.

[17] A. K. Iyer and G. V. Eleftheriades, “Mechanisms of subdiffraction free-space imaging using a transmission-line metamaterial superlens: Anexperimental verification,” Appl. Phys. Lett., vol. 92, no. 13, p. 131105,Mar. 2008.

[18] G. V. Eleftheriades, A. K. Iyer, and P. C. Kremer, “Planar negative re-fractive index media using periodically L-C loaded transmission lines,”IEEE Trans. Microw. Theory Tech., vol. 50, no. 12, pp. 2702–2712,Dec. 2002.

[19] A. K. Iyer and G. V. Eleftheriades, “A volumetric layered transmission-line metamaterial exhibiting a negative refractive index,” J. Opt. Soc.Am. B. (JOSA-B), Focus Issue on Metamater., vol. 23, pp. 553–570,Mar. 2006.

[20] A. K. Iyer and G. V. Eleftheriades, “Characterization of a multilay-ered negative-refractive-index transmission-line (NRI-TL) metamate-rial,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig., San Francisco,CA, Jun. 11–16, 2006, pp. 428–431.

[21] A. K. Iyer and G. V. Eleftheriades, “A multilayer negative-refrac-tive-index transmission-line (NRI-TL) metamaterial free-space lensat X-band,” IEEE Trans. Antennas Propag., vol. 55, no. 10, pp.2746–2753, Oct. 2007.

[22] G. V. Eleftheriades, “Analysis of bandwidth and loss in negative-re-fractive-index transmission-line (NRI–TL) media using coupledresonators,” IEEE Microw. Wireless Comp. Lett., vol. 17, no. 6, pp.412–414, Jun. 2007.

[23] Ansoft Corporation [Online]. Available: http://www.ansoft.com/prod-ucts/hf/hfss/

[24] D. R. Smith, S. Schultz, P. Markos, and C. M. Soukoulis, “Determina-tion of effective permittivity and permeability of metamaterials fromreflection and transmission coefficients,” Phys. Rev. B, vol. 65, no. 19,p. 195104, May 2002.

[25] R. W. King, “The loop antenna for transmission and reflection,” in An-tenna Theory: Part I, R. E. Collin and F. J. Zucker, Eds. Toronto:McGraw-Hill, 1969, ch. 11, pp. 478–480.

[26] G. Wang, J. R. Fang, and X. T. Dong, “Refocusing of backscatteredmicrowaves in target detection by using LHM flat lens,” Opt. Express,vol. 15, no. 6, pp. 3312–3317, Mar. 2007.

Ashwin K. Iyer (S’01) received the BASc., MASc.,and Ph.D. degrees in electrical engineering from theUniversity of Toronto, ON, Canada, in 2001, 2003,and 2009, respectively.

In 1999 and 2000, he held summer positions asan ASIC Designer for the Microelectronics andHigh-Performance Optical Components divisions ofNortel Networks, respectively, in Ottawa, Canada.While with the University of Toronto, he was in-volved with the development and characterization ofengineered electromagnetic materials, also known

as metamaterials, that exhibit a negative refractive index. He has coauthored achapter in a book on the subject of metamaterials, Negative-Refraction Meta-materials: Fundamental Principles and Applications (Wiley and IEEE Press,2005). He has also coauthored a number of refereed journal and conferencepapers and has given many seminars, invited talks, and workshops on thesubject of his research. His current research interests include RF/microwavecircuits and transmission-line techniques, fundamental electromagnetic theory,antennas, periodic structures, and wave propagation in engineered artificial(“meta-”) materials/surfaces at the micro- and nanoscales, and their applica-tions to novel leaky-wave antennas, imaging, microwave and optical devices,and biomedicine.

Dr. Iyer is the recipient of the 2008 R. W. P. King Award, presented by theIEEE Antennas and Propagation Society. He received second place in the stu-dent-paper competitions of both the 2002 and 2006 IEEE MTT-S InternationalMicrowave Symposia, held in Seattle, WA in June 2002, and San Francisco, CA,in June 2006, respectively. He received an honorable mention prize in the stu-dent-paper competition of the 2008 IEEE AP-S Antennas and Propagation Sym-posium held in San Diego, California. He was awarded the best student paperprize at the ANTEM/URSI Symposium, Banff, AB, Canada, in February 2009.He has received several awards for his graduate work, including the CanadaGraduate Scholarship, presented by the Natural Sciences and Engineering Re-search Council (NSERC) of Canada.

George V. Eleftheriades (S’86–M’88–SM’02–F’06) received the Ph.D. and M.S.E.E. degreesin electrical engineering from the University ofMichigan, Ann Arbor, in 1993 and 1989 respec-tively, and the diploma (with distinction) in electricalengineering from the National Technical Universityof Athens, Greece in 1988.

In the period 1994 to 1997, he was with the SwissFederal Institute of Technology in Lausanne wherehe was engaged in the design of millimeter andsubmillimeter wave receivers, and the creation of

fast CAD tools for planar packaged microwave circuits. Currently, he is a Pro-fessor in the Department of Electrical and Computer Engineering, Universityof Toronto, Toronto, ON, Canada, where he holds the Canada Research/VelmaM. Rogers Graham Chair in Engineering. Currently, he is leading a sizablegroup of graduate students in the areas of electromagnetic negative-refractionmicrowave and optical metamaterials, IC antennas and components for broad-band wireless communications, novel antenna beam-steering techniques, andelectromagnetic design for high-speed digital circuits.

Prof. Eleftheriades received the Ontario Premier’s Research ExcellenceAward and the Gordon Slemon Award for the “teaching of design” from theUniversity of Toronto both in 2001. In 2004, he received an E. W. R. SteacieFellowship from the Natural Sciences and Engineering Research Councilof Canada. In 2006, he was elected a Fellow of the IEEE “for contributionsto conception, analysis and fabrication of electromagnetic materials andtheir applications.” He serves as an IEEE AP-S Distinguished Lecturer andamongst his other scholarly achievements he is the recipient of the 2008 IEEEKiyo Tomiyasu Technical Field Award “for pioneering contributions to thescience and technological applications of negative-refraction electromagneticmaterials.” Presently, he serves as an elected member of the AP-S AdComand as an Associate Editor of the IEEE TRANSACTIONS ON ANTENNAS AND

PROPAGATION. He is a member of Technical Co-ordination Committee MTT-15(Microwave Field Theory).