APS/123-QED

Phase Conjugation and Negative Refraction Using Nonlinear

Active Metamaterials

Alexander R. Katko1 , Shi Gu1 , John P. Barrett1 , Bogdan-Ioan

Popa1 , Gennady Shvets2 , and Steven A. Cummer1∗
1
Department of Electrical and Computer Engineering and
arXiv:1008.1999v1 [physics.optics] 11 Aug 2010

Center for Metamaterials and Integrated Plasmonics,

Duke University, Durham NC 27708
2
Department of Physics, The University of Texas at Austin, Austin TX 78712
(Dated: October 22, 2018)

Abstract
We present experimental demonstration of phase conjugation using nonlinear metamaterial el-
ements. Active split-ring resonators loaded with varactor diodes are demonstrated theoretically
to act as phase-conjugating or time-reversing discrete elements when parametrically pumped and
illuminated with appropriate frequencies. The metamaterial elements were fabricated and shown
experimentally to produce a time reversed signal. Measurements confirm that a discrete array of
phase-conjugating elements act as a negatively-refracting time reversal RF lens only 0.12λ thick.

1
 Metamaterials are subwavelength structures which, when viewed macroscopically, achieve
electromagnetic properties different from their constituent elements. This may include an
effective negative refractive index [1], negative effective permeability [2], effective permit-
tivity near zero [3], or many other unusual properties. Equally importantly, functionality
can be embedded into metamaterials that can be difficult if not impossible to find in con-
ventional materials. Examples of this approach include active metamaterials with [4–6] and
without [7–9] gain for loss compensation and frequency tuning, respectively, as well as am-
plification [10]. Embedding nonlinear response into metamaterials has also been used for
power-dependent tuning [11–13]. Coherent harmonic generation has also been demonstrated
across the electromagnetic spectrum [14, 15] using nonlinear metamaterials.
In this Letter we introduce a new class of nonlinear active metamaterials (NAMMs). Com-
bining metamaterials’ nonlinear response with an active high-frequency external excitation
in the simplest metamaterial building block, the split-ring resonator (SRR), we experimen-
tally demonstrate phase conjugation of microwave signals. Using numerical simulations, we
demonstrate that such NAMM can be used as a building block for a negatively-refracting
lens. Negative refraction [16] is one of the most discussed and intriguing phenomena exhib-
ited by a special class of metamaterials with negative permittivity and permeability. It was
shown theoretically by Pendry [17], however, that negative refraction could also be produced
by a thin slab of a material exhibiting time reversal (or, equivalently, phase conjugation).
Traditional thick slabs of phase-conjugating (PC) media have been used for the useful prop-
erty of generating negative frequency waves that focus back on the source, sometimes called
retrodirectivity [18]. Discrete arrays of PC elements have also been used for retrodirectivity
by the microwave community [19, 20]. Pendry showed that a thin slab would indeed generate
these retrodirective waves but would also generate a wave on the other side of the material
exhibiting negative refraction. A few recent works have examined using discrete PC elements
for forward focusing [21, 22], but these cannot be extended to create an effective medium.
As an example, [21] used an array of patch antennas for receiving; a phase-conjugating cir-
cuit for each antenna composed of PLLs, filters, phase shifters, IQ modulators, and other
components; and a separate array of patch antennas for retransmission. Using NAMMs, we
demonstrate negative refraction using simple metamaterial elements with varactors embed-
ded, allowing easy extension to a true volumetric approach.
We demonstrate theoretically and experimentally that nonlinear metamaterial elements,

2
when driven with an active source, act as PC elements. The nonlinear characteristics of the
metamaterials are obtained by embedding varactor diodes in each metamaterial element and
exciting each element with a monochromatic RF signal, allowing easy extension to construc-
tion of a bulk medium of phase-conjugating SRRs with no external circuitry. We then demon-
strate experimentally that nonlinear metamaterials do indeed act as phase-conjugating and
thus negatively-refracting elements, and in an array they act as a time-reversing material.
Parametric resonance can be used to realize phase-conjugation in a single cell of a
NAMM. Consider a single dipole-like nonlinear resonant scatterer with the parameter-
ically modulated natural frequency Ω(t) given by Ω2 (t) = Ω20 (1 + h0 sin (2ω0 t)), where
p
the Ω(t) = 1/ L(t)C(t) is related to the resonator’s inductance L and capacitance C,
one (or both) of which are modulated with the frequency 2ω0 by the externally pro-
vided signal. The dimensionless parameter h0 characterizes the strength of the paramet-
ric drive. Our experimental implementation relies on the modulation of the SRR’s ca-
pacitance by inserting a nonlinear varactor whose capacitance depends on the externally-
applied high-frequency voltage. Further assume that SRRs are arrayed in the z = 0
plane along the y− axis and subjected to an external monochromatic electromagnetic wave
P
E(y, t) = ky Eky exp [−i(ω1 t − ky y)] + c. c. incident from the negative z direction and

produced by a far-field source located at zsource < 0. The response of a single j’s SRR,
characterized by its induced dipole moment pj , is calculated according to

d2 pj dpj
2
+ γ + Ω20 (1 + h0 sin (2ω0 t)) pj =
dt
X dt
Ω20 E(ky )a3 e−i(ω1 t−ky yj ) + c.c., (1)
ky

where yj is the spatial location of the SRR-based NAMM, a is its effective size, and γ is the
SRR’s loss rate that includes dissipative and radiative losses. The effects of inter-element
interaction are assumed to be negligible owing to the tight field confinement. Eq. (1) models
a single SRR as a linear electric dipole with the low-frequency polarizability a3 which is
resonantly-driven by the RF field of the antenna source. The solution of Eq. (1) can be
(j) (j) (j) (j)
expressed as pj = a1 exp (−iω1 t) + b2 exp (iω2 t), where ω2 = 2ω0 − ω1 , and a1 , b2 are
computed by substituting pj into Eq. (1) to obtain:

(j) Ω2 − ω22 + 2iγω2 iky yj (j) ih0 Ω2 iky yj

a1 = e , b2 = e , (2)
D(ω1 ) 2D(ω1 )

3
where the dispersion function D(ω1 ) is given by

D = (Ω2 − ω12 − 2iγω1 )(Ω2 − ω22 + 2iγω2 ) − h20 Ω4 /4. (3)

The electric field at z > 0 behind the plane of NAMMs is given by the sum of the incident
fields and the fields produced by the induced dipole moments pj of all NAMMs. Because
the components of pj proportional to a1 have the same spatio-temporal dependence as the
incident electric field E(y, t), they produce scattered radiation which is spatially diverging
away from the source. However, the contributions to pj proportional to b2 produce a phase-
conjugated signal which, assuming that ω2 ≈ ω1 , produces a spatially-converging field with
the focus at zimage = −zsource > 0, i. e. the plane of NAMMs constitutes a flat negative-
refraction lens.
The strongest phase-conjugated signal is produced when Ω0 = ω0 and ω1 ≈ ω0 . How-
ever, ω2 can deviate from ω1 , significantly simplifying our experiments because the phase-
conjugated signal has a different frequency from that of the incident field and is not over-
whelmed by it. Note that, for certain values of ω0 , Ω0 , γ, and h0 , there exist complex
(r) (i)
solutions ω∗ ≡ ω∗ + iω∗ of D(ω∗ ) = 0 corresponding to the parametric instability [23] with
(i)
the growth rate ω∗ . For example, for ω = Ω0 instability occurs if h0 ω0 > 2γ0 . For the rest
of this Letter we assume that the strength of the parametric drive h0 is small enough for
the NAMM to remain stable.
Following the above analysis, we designed and fabricated SRR-varactor nonlinear meta-
material elements. They were designed to resonate near 950 MHz and have a pump signal
delivered via cables directly to the varactors. Fig. 1 shows the layout of an SRR with
dimensions and corresponding photo of a fabricated SRR.
Pump (SMA)
(a) (b)

40mm Source k
B

1mm 1mm

FIG. 1: (Color online) (a): Layout of SRR element. Red elements are varactor diodes and blue
elements are isolation resistors. (b): Photo of fabricated SRR element.

4
 The SRRs were pumped with a signal at a specified pump frequency fpump corresponding
to 2ω0 . A monopole antenna located above a ground plane was used as the illumination
source, radiating a signal at a specified source frequency fsource corresponding to ω1 . As
described above, a nonlinearly generated, phase conjugated signal should be reradiated at
fpc = fpump − fsource corresponding to ω2 .
Measurements were made using a spectrum analyzer with a receiving antenna to deter-
mine interferene-free frequency bands and the frequencies were selected so that the mixed
signal at fpc was clearly distinct from the source signal at fsource . These experiments used fre-
quencies of fpump = 1850MHz, fsource = 950MHz, and subsequently fpc = 900MHz. Output
from the spectrum analyzer is shown in Fig. 2 showing both the background signal levels and
signal levels for the experiments with the pump and source signal generators switched on. As
expected with our chosen frequencies, the signals at fpump = 1850MHz, fsource = 950MHz,
and fpc = 900MHz are only due to our experimental setup and not background radiation,
demonstrating that the nonlinear SRRs are indeed generating a nonlinear mixed signal. The
maximum measured value of the signal at fsource was -25.2 dBm. Through the experiments
the maximum recorded value of the PC signal at fpc was -49.6 dBm, yielding a minimum
ratio of -24.4 dB between the source and PC signal. The source of this difference comes from
the mixing efficiency in our experiments, which was not optimized in this work. Improv-
ing the coupling efficiency of the pump signal into the nonlinear metamaterial to overcome
losses in the SRR will result in more efficient and practically-relevant NAMMs. The signal
fpump
at 925MHz corresponds to the sub-harmonic of 2
and is present due to a power amplifier
used in connection with the pump. The broad signal from approximately 868 to 888MHz
corresponds to the downlink band of GSM-850 cellular phone signals. This and the signal
at 929MHz are independent of our experiments but did not interfere at any frequencies.
No higher order mixed frequency signals were detected and, along with the noted weak PC
signal at least 24.4 dB down from the source, the assumptions in the analysis section are
shown to be valid.
We conducted multiple experiments to demonstrate phase conjugation of these elements.
To verify that individual elements were PC, we measured the interference pattern generated
by two isolated elements. We also conducted an experiment to examine phase conjugation
with an array of elements, described later. To map interference patterns, two monopole
antennas above a ground plane were connected to a signal generator and spaced 1.5λ apart.

5
 −15
Sources turned off Source signal
−25 Sources turned on

Measured Signal [dBm]

−35

−45
PC
Signal
−55
GSM-850
Downlink
−65 Pump Sub-
Harmonic

−75

−85
800 825 850 875 900 925 950 975 1000
Frequency [MHz]

FIG. 2: (Color online) Data from spectrum analyzer for background signal (dashed black) and
pump and source signals switched on (solid blue). Notable frequencies include 900MHz (fpc ),
950MHz (fsource ), 1850MHz (not shown).

A microwave absorber was placed between the antennas to isolate them. They were excited
both in phase and –90°out of phase, and the interference patterns were mapped at a constant
distance λ from the line connecting the antennas, at the signal frequency, as illustrated in
Fig. 3(b). For two normal, non-conjugating elements excited in phase and then –90°out of
phase, the maximum of the pattern moved to the right. Then the same interference patterns
were measured at the mixed frequency with the SRRs in place. The absorber was used to
ensure that only one antenna was illuminating each SRR. Again the antennas were excited
in phase and then –90°out of phase and the interference pattern was mapped along the same
line as before, at the mixed frequency. The measurements were interpolated and a filter was
applied in the spatial Fourier domain to remove noise. The results are shown in Fig. 3(a)
and a schematic is shown in Fig. 3(b).
Exciting two radiating elements with 0°phase shift and the same spatial positioning
yielded field patterns roughly independent of the elements used, as shown by the solid
lines in Fig. 3. The small-scale features in each curve are artifacts of a low dynamic range in
these measurements. However, the important feature is the clear difference in the locations
of the global extrema. Introducing a phase shift to the non-conjugating elements creates a
maximum to the right of the center. However, with the NAMM elements, the peak moves
to the opposite side of the center, showing that exciting normal radiating elements with a
–90°phase shift is equivalent to exciting the NAMM elements with a +90°phase shift. This

6
 1
(a) 0° Antennas
0.9 Signal freq
0.8

Normalized Intensity [au]

0° SRRs
0.7 Mixed freq
0.6
-90° Antennas
0.5 Signal freq
0.4
-90° SRRs
0.3
Mixed freq
0.2

0.1

0
−0.25 0 0.25
Lateral displacement [λ]

(b) Measurement Line

SRR SRR

Antenna Antenna
Absorber

Phase
Shift

Signal Feed

Setup Diagram

FIG. 3: (Color online) (a) Interference pattern for: in-phase antennas at fsignal (solid blue), –
90°out-of-phase antennas at fsignal (dashed blue), in-phase SRRs at fmixed (solid red), –90°out-of-
phase SRRs at fmixed (dashed red). The opposite lateral shifts in extrema between the SRRs and
antennas, with the same phase shift applied to each, clearly demonstrates that the SRRs act as
PC elements. (b) Schematic of setup (to scale). Antennas were fed at fsignal with both 0°and –
90°phase shifts. Absorber was used to isolate antennas and SRRs. The pump signal for the SRRs,
as described above, is not shown.

is a clear demonstration that the nonlinear SRRs act as phase-conjugating elements.

Three NAMM elements were then used to approximate a discrete slab of time-reversal
medium. The thickness of the SRR array was approximately 0.12λ. The SRRs were pumped
in phase and illuminated with a single antenna. As before fpump = 1850MHz, fsource =
950MHz, and fpc = 900MHz. The SRRs were placed λ/4 apart, creating an array a total of
λ/2 in width. The source antenna was placed a distance one λ normal to the array and λ/2
from the center of the array. The field power distribution of the mixed and phase-conjugated
signal was then measured spatially to create a two-dimensional field map.
A simple calculation was used to generate the expected relative field distribution for
phase-conjugating elements without considering details of the mixing efficiency of individual

7
elements. The calculation assumed a r−1 dependence on the original signal, the generation
of the nonlinear difference frequency in linear proportion to the incident signal, and a r−1
dependence and the measured radiation pattern of the SRRs for the reradiated signal. Its
primary purpose was to predict the relative interference pattern for PC elements in a given
geometry for comparison with measurements.
The field map was generated over a measurement area of dimensions −3λ/2 to 3λ/2
tangential to the array and λ/4 to λ normal to the array again by using the spectrum
analyzer. The raw data was interpolated and a filter was applied in the spatial Fourier
domain to remove noise from the measurements. Fig. 4 shows the results overlaid on the
experimental setup.

Scale [dBm]
−50

Tangential Displacement
Source
m
SRRs Bea −55
/PC
NR
−60

−65

−70

−75
0.25 0.5 0.75 1
Simulation Normal Displacement [λ]

FIG. 4: (Color online) Measured abs(E)dBm and experimental setup. Setup is approximately to
scale. The source radiates from the upper left in the figure and SRRs produce a mixed signal at
fpc . The inset shows calculated results, exhibiting good agreement. The measurements show that
the beam negatively refracts.

The main beam at the phase-conjugated frequency fpc clearly refracts to the opposite side
of the normal compared to a conventional material. Moreover, there is excellent agreement
between measurement and calculation, as shown by the inset in the figure. The same cal-
culation but assuming non-phase-conjugating elements was also conducted and yielded very
different results. The field power along a line of constant distance from the array was exam-
ined to illustrate this more quantitatively, comparing conjugate calculation, non-conjugate
calculation, and filtered measured data. This is presented in Fig. 5.
An array of PC metamaterials is limited in far-field imaging ability only by normal
imaging considerations like aperture size and spacing between elements. Creating a larger

8
 1
PC
Simulation

0.8 Non-PC
Simulation

Normalized |E|2
Filtered
0.6 Measurement

0.4

0.2

0
−0.75 −0.5 −0.25 0 0.25 0.5 0.75
Vertical Position [λ]

FIG. 5: (Color online) Normalized intensity at constant distance from array for various data sets:
calculated data for phase-conjugating elements (black); calculated data for non-phase-conjugating
elements (blue); and filtered measured data (red). The measured data match calculated data very
well and are negatively refracted when compared to non-PC simulated data.

array with more PC elements provides a better approximation of an infinite, continuous

medium and thus provides a closer approximation to an ideal far-field lens. The backward-
focusing (retrodirectivity) of an array of PC elements has been explored to some extent [19–
22]. However, these approaches have relied on much more complex structures or circuitry
and cannot be easily extended to volumetric devices. Our approach relies on much simpler
elements which could be used as building blocks for PC effective media. Moreover, the
primary application of retrodirective arrays has been for backward focusing. We have shown
that NAMMs can be used to create a simple and effective phase-conjugating medium and
provided a path towards realizing time-reversal metamaterial media.
In summary, we first demonstrated theoretically that metamaterials with embedded non-
linear elements produce a phase-conjugated or time-reversed signal through parametric
pumping. Two element experiments unambiguously demonstrated the phase-conjugation
of the mixed signal, and measurements of the field distribution produced by a 3 element
array showed the expected effective negative refraction behavior of the time-reversed signal.
Such elements form a simple but powerful building block for more complex phase conjuga-
tion media and devices, and we have shown that this class of active nonlinear metamaterials
is promising for realising designs utilizing time-reversal media and thin-slab RF imaging.
Acknowledgements: this work was supported in part by DARPA under Contract No.
HR0011-05-C-0068. G.S. would like to acknowledge the support of the U.S. Air Force Office

9
of Scientific Research (AFOSR) MURI grants FA 9550-06-1-0279 and FA 9550-08-1-0394.

∗ Electronic address: [email protected]

[1] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz, Phys. Rev. Lett.
84, 4184 (2000).
[2] J. Pendry, A. Holden, D. Robbins, and W. Stewart, IEEE Trans. Microwave Th. and Tech.
47, 2075 (1999).
[3] R. W. Ziolkowski, Phys. Rev. E 70, 046608 (2004).
[4] Y. Yuan, B.-I. Popa, and S. Cummer, Opt. Express 17, 16135 (2009).
[5] B.-I. Popa and S. Cummer, Microwave and Optical Technology Letters 49, 2574 (2007).
[6] A. Fang, T. Koschny, M. Wegener, and C. M. Soukoulis, Phys. Rev. B 79, 241104(R) (2009).
[7] H.-T. Chen, W. J. Padilla, J. M. O. Zide, A. C. Gossard, A. J. Taylor, and R. D. Averitt1,
Nature 444, 597 (2006).
[8] O. Reynet and O. Acher, Applied Physics Letters 84, 1198 (2004).
[9] D. Wang, L. Ran, H. Chen, M. Mu, J. A. Kong, and B.-I. Wu, Applied Physics Letters 91,
164101 (2007).
[10] R. Syms, L. Solymar, and I. Young, Metamaterials 2, 122 (2008).
[11] B. Wang, J. Zhou, T. Koschny, and C. M. Soukoulis, Opt. Express 16, 16058 (2008).
[12] D. Huang, E. Poutrina, and D. Smith, Applied Physics Letters 96, 104104 (2010).
[13] I. Shadrivov, S. Morrison, and Y. Kivshar, Opt. Express 14, 9344 (2006).
[14] Z. Wang, Y. Luo, L. Peng, J. Huangfu, T. Jiang, D. Wang, H. Chen, and L. Ran, Applied
Physics Letters 94, 134102 (2009).
[15] M. W. Klein, C. Enkrich, M. Wegener, and S. Linden, Science 313, 502 (2006).
[16] V. Veselago, Sov. Phys. USPEKHI 10, 509 (1968).
[17] J. B. Pendry, Science 322, 71 (2008).
[18] A. Yariv, IEEE J. Quant. Elec. QE-14, 650 (1978).
[19] L. Chiu, T. Yum, W. Chang, Q. Xue, and C. Chan, Microwave and Optical Technology Letters
48, 409 (2006).
[20] C. Pon, IEEE Trans. Ant. Prop. 12, 176 (1964).
[21] V. F. Fusco, N. B. Buchanan, and O. Malyuskin, IEEE Trans. Ant. Prop. 58, 798 (2010).

10
[22] O. Malyuskin and V. Fusco, IEEE Trans. Ant. Prop. 58, 459 (2010).
[23] L. D. Landau and E. M. Lifshitz, Mechanics (Pergamon Press, New York, 1976), 3rd ed.

11