Vol 461 | 8 October 2009 | doi:10.

1038/nature08293

LETTERS
Observation of unidirectional backscattering-immune
topological electromagnetic states
Zheng Wang1*, Yidong Chong1{*, J. D. Joannopoulos1 & Marin Soljačić1

One of the most striking phenomena in condensed-matter physics Ð
LAnn 

LAnn
is the quantum Hall effect, which arises in two-dimensional
1
Cn ~ 2pi d2 k y
Lkx { x
Lky
BZ
electron systems1–4 subject to a large magnetic field applied
perpendicular to the plane in which the electrons reside. In such where the k-space integral is performed over the first Brillouin zone
circumstances, current is carried by electrons along the edges of and the Berry connection6 is given by
ð
the system, in so-called chiral edge states (CESs). These are states
that, as a consequence of nontrivial topological properties of the Ann’ (k):ihEnk j+k jEn’k i~i d2 re(r)E
nk (r) ½+k En’k (r) .

bulk electronic band structure, have a unique directionality and

are robust against scattering from disorder. Recently, it was where Enk is the periodic part of the electric-field Bloch function16, an
theoretically predicted5–7 that electromagnetic analogues of such asterisk denotes complex conjugation and e(r) denotes the dielectric
electronic edge states could be observed in photonic crystals, function. Because the Chern number characterizes the winding number
which are materials having refractive-index variations with a of the phase of the Bloch functions around the boundary of the first
periodicity comparable to the wavelength of the light passing Brillouin zone11, it is a ‘global’ or ‘topological’ property of the entire
band and is very robust against structural perturbations10. Notably, it
through them. Here we report the experimental realization and
can be non-zero only if the system lacks time-reversal symmetry9.
observation of such electromagnetic CESs in a magneto-optical
photonic crystal7 fabricated in the microwave regime. We One of the most interesting properties of QHE systems is that the
Chern numbers have a direct physical significance: a finite crystal that
demonstrate that, like their electronic counterparts8–13, electro-
supports bulk bands with non-zero Chern numbers also supports
magnetic CESs can travel in only one direction and are very robust
unidirectional CESs at its boundary at energies within bulk band
against scattering from disorder; we find that even large metallic
gaps opened by the applied d.c. magnetic field. Moreover, the
scatterers placed in the path of the propagating edge modes do not
number of CESs turns out to be equal to the sum of the Chern
induce reflections. These modes may enable the production of
numbers of all the bulk bands of lower energy13. Although this result
new classes of electromagnetic device and experiments that would
has been formally proven only in a tight-binding QHE system, it is
be impossible using conventional reciprocal photonic states alone. believed to be independent of the details of the underlying model,
Furthermore, our experimental demonstration and study of photo- such as the structure of the lattice and the edge. Its validity in
nic CESs provides strong support for the generalization and applica- photonic-crystal systems was originally predicted in refs 5, 6, and
tion of topological band theories to classical and bosonic systems, corroborated through a formal mapping7 to a ‘zero-field QHE’
and may lead to the realization and observation of topological phe- system12 and ab initio numerical simulations of Maxwell’s equations7.
nomena in a generally much more controlled and customizable It is important to emphasize that although CESs have so far been
fashion than is typically possible with electronic systems. experimentally observed only in electronic (that is, fermionic)
The existence of photonic CESs was first predicted5,6 by an analogy systems, the phenomenon should actually be independent of the
between a photonic crystal14–16 with broken time-reversal symmetry underlying particle statistics because the Chern number is defined
and a system exhibiting the quantum Hall effect (QHE). In this ana- in terms of single-particle Bloch functions. An experimental verifica-
logy, the electromagnetic fields play the part of the electronic current, tion would therefore provide strong support for the generalization of
the variations of permittivity and permeability within the photonic topological band theories and their applications to classical and
crystal play the part of the periodic potential and the gradients of the bosonic systems.
gyrotropic components of the permeability tensor play the part of the The ability to work with photonic-crystal band structures without
external d.c. magnetic field that breaks the time-reversal symmetry5–7. Dirac points has allowed us to design an experimentally viable photonic-
The defining feature of a photonic CES is that its group velocity points crystal system7 for the observation of CESs. Our experimental system
in only one direction, which is determined by the sign of the field that (Fig. 1) involves a gyromagnetic, 2D-periodic photonic crystal consist-
breaks the time-reversal symmetry and the resulting unusual topo- ing of a square lattice of ferrite rods in air (details of the structure and
logical properties of the bulk band structure. To detect the possible materials used can be found in Methods), bounded on one side by a
presence of non-trivial topological band properties in a photonic- non-magnetic metallic cladding. The interface between the photonic
crystal system it is sufficient5–7 to compute its Chern numbers. crystal and the cladding acts as a confining edge or waveguide for
(Although the original proposal5,6 focused on ‘Dirac points’, it is CESs. (Without this cladding, the CESs at the air edges of the photonic
not necessary to restrict to such band structures; thus, the use of a crystal would simply radiate away.) Neglecting absorption losses and
variety of photonic-crystal systems is possible7.) The Chern number of nonlinear effects, we would expect power transmission of a CES along
band n of a two-dimensional (2D) periodic photonic crystal is an this waveguide to be independent of the waveguide geometry and also
integer defined by6 immune to backscattering from disorder, obstacles and defects.
1
Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA. {Present address: Yale University, New Haven, Connecticut 06520, USA.
*These authors contributed equally to this work.

772
©2009 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 461 | 8 October 2009 LETTERS

a A
Scatterer of Antenna B
a
variable length l

CES waveguide

Antenna A
Metal wall

A
b

y x
b
B
c

4 cm

Figure 1 | Microwave waveguide supporting CESs. a, Schematic of the Ez

waveguide composed of an interface between a gyromagnetic photonic- Negative 0 Positive
crystal slab (blue rods) and a metal wall (yellow). The structure is
sandwiched between two parallel copper plates (yellow) for confinement in Figure 2 | Photonic CESs and effects of a large scatterer. a, CES field
the z direction and surrounded with microwave-absorbing foams (grey distribution (Ez) at 4.5 GHz in the absence of the scatterer, calculated from
regions). Two dipole antennas, A and B, serve as feeds and/or probes. A finite-element steady-state analysis (COMSOL Multiphysics). The feed
variable-length (l) metal obstacle (orange) with a height equal to that of the antenna (star), which is omnidirectional in homogeneous media
waveguide (7.0 mm) is inserted between the antennas to study scattering. A (Supplementary Information), radiates only to the right along the CES
0.20-T d.c. magnetic field is applied along the z direction using an waveguide. The black arrow represents the direction of the power flow.
electromagnet (not shown). b, Top view (photograph) of the actual b, When a large obstacle (three lattice constants long) is inserted, forward
waveguide with the top plate removed. transmission remains unchanged because backscattering and side-scattering
are entirely suppressed. The calculated field pattern (colour scale) illustrates
how the CES wraps around the scatterer. c, When antenna B is used as feed
Before we discuss the results of our measurements, we will first antenna, negligible power is transmitted to the left, as the backwards-
describe how we arrived at this particular choice of experimental propagating modes are evanescent. a, lattice constant.
system. We chose rods in air for the basic photonic-crystal geometry
because of ease of fabrication. We then performed a series of numerical obstacle results only in a change of the phase (compare Fig. 2a and
simulations for a variety of rod sizes and lattice constants on a model Fig. 2b) of the transmitted radiation, with no reduction in amplitude.
2D photonic-crystal system to optimize the band structure and For CESs to be readily measurable in the laboratory (where it is
compute corresponding band Chern numbers using material necessary to use a photonic crystal of finite and manageable size) they
parameters appropriate to a low-loss ferrite (Methods). Our numerical must be spatially well localized, and this requires the photonic band
simulations predicted that when the ferrite rods in this photonic gaps containing the states to be large. The sizes of the band gaps that
crystal are magnetized to manifest gyrotropic permeability (which contain CESs (and the frequencies at which they occur) are determined
breaks time-reversal symmetry), a gap opens between the second by the gyromagnetic constants of the ferrite rods constituting the
and third transverse magnetic (TM) bands. Moreover, the second, photonic crystal. Under a d.c. magnetic field, microwave ferrites
third and fourth bands of this photonic crystal acquire Chern numbers exhibit a ferromagnetic resonance at a frequency determined by the
of 1, 22 and 1, respectively. This result follows from the C4v symmetry strength of the applied field18. Near this frequency, the Voigt
of a non-magnetized crystal17. The results of our simulations for the parameter, V 5 jmxyj/jmxxj (where mxx and mxy are diagonal and off-
photonic crystal with metallic cladding are presented in Fig. 2. (Similar diagonal elements of the permeability tensor, respectively), which is
numerical results were obtained in ref. 7, albeit using a different a direct measure of the strength of the gyromagnetic effect, is of order
material system and geometry.) Here we show the calculated field one. Such ferromagnetic resonances are among the strongest low-loss
patterns of a photonic CES residing in the second TM band gap gyrotropic effects at room temperature and subtesla magnetic fields.
(between the second and the third bands). Because the sum of the Using ferrite rods composed of vanadium-doped calcium–iron–
Chern numbers over the first and second bands is 1, exactly one CES garnet under a biasing magnetic field of 0.20 T (Methods and
is predicted to exist at the interface between the photonic crystal and Supplementary Information), we achieved a relative bandwidth of
the metal cladding. The simulations clearly predict that this photonic 6% for the second TM band gap (around 4.5 GHz in Fig. 3b). As
CES is unidirectional. As side-scattering is prohibited by the bulk discussed earlier, this is the gap predicted to support a CES at the
photonic band gaps in the photonic crystal and in the metallic interface of the photonic crystal with the metallic wall. We emphasize
cladding, the existence of the CES forces the feed dipole antennas again that band gaps with trivial topological properties (that is, for
(which would radiate omnidirectionally in a homogeneous medium) which the Chern numbers of the bulk bands of lower frequencies sum
to radiate only towards the right (Fig. 2a, c). Moreover, the lack of to zero), such as the first TM band gap (around 3 GHz in Fig. 3b), do
any backwards-propagating mode eliminates the possibility of not support CESs. All of the insight gained from the model 2D photo-
backscattering, meaning that the fields can continuously navigate nic-crystal system was then incorporated into the final design (Fig. 1).
around obstacles, as shown in Fig. 2b. Hence, the scattering from the To emulate the states of the 2D photonic crystal, the final design
773
©2009 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 461 | 8 October 2009

a c

Transmission (dB)
0 Bulk crystal
0
–20

–40

Transmission (dB)
Forwards, |SBA|2 –20
–60 Backwards, |SAB|2

b –0.5 Edge waveguide

Wavevector (2π/a)

–40
0 Forwards, |SBA|2
Backwards, |SAB|2
0 1 –2 1 –60 4.5
0.5 4.0
0 1 2 3 4 5 6 Frequency (GHz)
Frequency (GHz)

Figure 3 | CES-facilitated waveguiding in a photonic crystal. a, Forward crystal. The grey areas are bulk bands with ill-defined band-edges due to
and backward transmission spectra measured using only the bulk photonic large absorption near the ferromagnetic resonance. Each band’s Chern
crystal in the set-up shown in Fig. 1 (that is, without the metal cladding and number is labelled. c, Measured transmission spectra upon inclusion of the
obstacle), with the antennas placed in the interior of the photonic crystal, in metal cladding and antennas placed as shown in Fig. 1. In the resulting CES
a 0.20-T d.c. magnetic field. The bulk transmission is reciprocal, with waveguide, there is very high contrast between the forward and backward
photonic band gaps at 3.3 and 4.5 GHz. b, Calculated projected photonic- transmissions for frequencies in the second band gap (yellow), around
crystal band structure (blue and grey areas). Included is the CES (red curve) 4.5 GHz. This striking unidirectionality indicates the existence of a CES.
that exists at the interface between the metal cladding and the photonic

involved fabrication of a three-dimensional (3D) photonic-crystal slab dimensions (Supplementary Information), a similar barrier length
structure equivalent to the model 2D photonic-crystal system, made of 1.65 lattice constants reduces forward transmission by four
from gyromagnetic rods with parallel metallic plates on the top and orders of magnitude. This measurement further confirms that the
bottom, spaced to support only transverse electromagnetic modes backwards-propagating modes are purely evanescent, and not merely
(identical to the TM modes in the 2D photonic crystal; see lossy. If lossy backwards-propagating modes existed in the system, a
Methods). A copper wall was then added at the edge of the photo- large defect would scatter a significant portion of energy into them,
nic-crystal slab to provide the required cladding. essentially converting backscattering into loss. The forward transmis-
In our experiments, the band gaps and the CES waveguide were sion in the presence of the large defect would be much smaller than in
characterized using two-port vector network analysis using a pair of
dipole antennas, labelled A and B in Fig. 1a (Methods). First, to 0
characterize the band gap, we inserted antennas A and B into the l = 0.40a l = 0.65a
interior of the photonic crystal far from the edges and eight lattice
–20
constants apart. We observed the second band gap with a 50-dB
extinction for both forward and backward transmission (with
|SAB|2
respective transmission coefficients SBA and SAB; Fig. 3a). The –40
frequency ranges of both the first and the second band gaps agree |SBA|2
well with our predicted band structure calculations (no adjustable –60
parameters; Fig. 3b). Next, to characterize the CESs, we measured the 0
l = 0.90a l = 1.15a
transmission spectra with the apparatus as illustrated in Fig. 1a
Transmission (dB)

(Methods). At frequencies within the second band gap, we observed –20

a strong forward transmission, approximately 50 dB greater than the
backward transmission at mid-gap frequencies (Fig. 3c). Over much
of this frequency range, the backward transmission was below the –40
noise floor of the network analyser, which suggests an even greater
actual contrast. This difference of more than five orders of magnitude –60
in power transmission, over a distance of only eight lattice constants, 0
l = 1.40a l = 1.65a
confirms that backwards-propagating modes are highly evanescent,
as predicted. –20
We tested the robustness of the unidirectional propagation by
studying the effect of a large obstacle on transmission. We gradually
–40
inserted a conducting barrier across the waveguide, blocking the direct
path between antennas A and B. The measured transmission behaviour
at different stages of the insertion (Fig. 4) remains basically the same as –60
4.0 4.5 4.0 4.5
that in Fig. 3c: the transmission between 4.35 and 4.62 GHz remains Frequency (GHz)
strongly non-reciprocal, with a 40–50-dB difference between the
forward and backward transmissions. This finding agrees with the Figure 4 | CES transmission spectra in the presence of a large scatterer.
theoretical prediction that power transmission by means of CESs is The length of the obstacle, l, was gradually varied from 0.40a to 1.65a (lattice
fundamentally insensitive to scattering from arbitrarily large defects constant, a 5 40 mm); this induced only minor differences in the forward
transmission near the mid-gap frequency of 4.5 GHz. The lack of any
(Fig. 2b). This behaviour is a distinguishing feature of the present significant changes in the forward transmission, and non-reciprocity
waveguide. In a conventional waveguide, insertion of such a large ( | SAB | = | SBA | ) with large increases in the size of the scatterer, indicate that
obstacle would cause very large backscattering and significantly the CES can travel around the obstacle without scattering or reflections, as
reduced transmission to the output. For example, in a photonic predicted by simulations. The experimental parameters remained
crystal constructed using ordinary dielectric rods and with identical unchanged from the measurement in Fig. 3c.
774
©2009 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 461 | 8 October 2009 LETTERS

the defect-free case. Existing optical isolators, such as those relying on 5. Haldane, F. D. M. & Raghu, S. Possible realization of directional optical
waveguides in photonic crystals with broken time-reversal symmetry. Phys. Rev.
Faraday rotation or non-reciprocal phase shifts, absorb or radiate Lett. 100, 013904 (2008).
backwards-propagating light in this way. Thus, the unidirectional 6. Raghu, S. & Haldane, F. D. M. Analogs of quantum-Hall-effect edge states in
guiding of a CES is fundamentally different from how optical isolators photonic crystals. Phys. Rev. A 78, 033834 (2008).
operate. 7. Wang, Z., Chong, Y. D., Joannopoulos, J. D. & Soljacic, M. Reflection-free one-way
The experimental establishment of topological photonic states edge modes in a gyromagnetic photonic crystal. Phys. Rev. Lett. 100, 013905
(2008).
opens a wide range of future opportunities. First, our realization of 8. Prange, R. E. & Girvin, S. M. (eds). The Quantum Hall effect (Springer, 1987).
nontrivial topological Chern numbers in a classical photonic system 9. Thouless, D. J., Kohmoto, M., Nightingale, M. P. & Dennijs, M. Quantized hall
raises the possibility of using photonic systems to realize other classes conductance in a two-dimensional periodic potential. Phys. Rev. Lett. 49,
of topological quantum numbers that are of interest in condensed- 405–408 (1982).
matter physics. Examples include the Z2 topological number asso- 10. Simon, B. Holonomy, the quantum adiabatic theorem, and Berry phase. Phys. Rev.
Lett. 51, 2167–2170 (1983).
ciated with the quantum spin Hall effect19–22 and the ‘Hopf number’ 11. Kohmoto, M. Topological invariant and the quantization of the Hall conductance.
in certain 3D insulators23. Photonic crystals are attractive for such Ann. Phys. 160, 343–354 (1985).
investigations because parameters such as lattice constants and unit- 12. Haldane, F. D. M. Model for a quantum Hall effect without Landau levels:
cell geometries can be chosen in a fully controlled manner16, unlike in condensed-matter realization of the ‘‘parity anomaly’’. Phys. Rev. Lett. 61,
2015–2018 (1988).
most electronic systems. Second, the fact that the CESs in the present 13. Hatsugai, Y. Chern number and edge states in the integer quantum Hall effect.
system are immune to scattering from disorder ensures that the Phys. Rev. Lett. 71, 3697–3700 (1993).
design is tolerant of fabrication imperfections, such as variations in 14. Yablonovitch, E. Inhibited spontaneous emission in solid-state physics and
the lattice constant or the exact position of the guiding edge; this electronics. Phys. Rev. Lett. 58, 2059–2062 (1987).
could enable implementation of extremely robust waveguides. 15. John, S. Strong localization of photons in certain disordered dielectric
superlattices. Phys. Rev. Lett. 58, 2486–2489 (1987).
Finally, photonic CESs might prove useful in applications involving 16. Joannopoulos, J. D., Johnson, S. G., Winn, J. N. & Meade, R. D. Photonic Crystals:
isolators24 or slow light25,26. In conventional slow-light systems, Molding the Flow of Light (Princeton Univ. Press, 2008).
disorder induces backscattering that increases quadratically with 17. Chong, Y. D., Wen, X. G. & Soljacic, M. Effective theory of quadratic degeneracies.
reduced group velocity27, making them very sensitive to disorder. Phys. Rev. B 77, 235125 (2008).
18. Pozar, D. M. Microwave Engineering 2nd edn (Wiley, 1998).
Although the experiments described here were conducted at
19. Murakami, S., Nagaosa, N. & Zhang, S.-C. Dissipationless quantum spin current at
gigahertz frequencies, this operating frequency can be increased room temperature. Science 301, 1348–1351 (2003).
simply by applying a stronger d.c. magnetic field18. Extension into 20. Kane, C. L., Mele, E. J. & Z. (2) topological order and the quantum spin Hall effect.
the terahertz range might be achieved using metamaterials that Phys. Rev. Lett. 95, 146802 (2005).
resonantly enhance the magnetic activity28–30. Further extension to 21. Bernevig, B. A., Hughes, T. L. & Zhang, S. C. Quantum spin Hall effect and
topological phase transition in HgTe quantum wells. Science 314, 1757–1761
the optical regime is challenging, given the losses and weak gyrotropic (2006).
effects in currently known materials. 22. Hsieh, D. et al. A topological Dirac insulator in a quantum spin Hall phase. Nature
452, 970–975 (2008).
METHODS SUMMARY 23. Moore, J. E., Ran, Y. & Wen, X.-G. Topological surface states in three-dimensional
The gyromagnetic photonic crystal was constructed using a square array (lattice magnetic insulators. Phys. Rev. Lett. 101, 186805 (2008).
24. Yu, Z. F. & Fan, S. H. Complete optical isolation created by indirect interband
constant, a 5 40 mm) of vanadium-doped calcium-iron-garnet (VCIG; TCI
photonic transitions. Nature Photon. 3, 91–94 (2009).
ceramics NG-1850) rods. Balancing the need for a large Voigt parameter against 25. Baba, T. Slow light in photonic crystals. Nature Photon. 2, 465–473 (2008).
the drawback of absorption loss in the vicinity of the ferromagnetic resonance 26. Thevenaz, L. Slow and fast light in optical fibres. Nature Photon. 2, 474–481
(5.6 GHz), we designed the rod radius to be 3.9 mm and a to be 40 mm to (2008).
maximize the bandwidth of the band gap without suffering excessive loss. A 27. Povinelli, M. L. et al. Effect of a photonic band gap on scattering from waveguide
16 3 10 array was used to measure the band gap of a bulk crystal and a 16 3 7 disorder. Appl. Phys. Lett. 84, 3639–3641 (2004).
array was used to study the waveguide and the effect of scattering. The VCIG 28. Pendry, J. B., Holden, A. J., Robbins, D. J. & Stewart, W. J. Magnetism from
ferrite has a measured relative permittivity of er 5 14.63 and a loss tangent of conductors and enhanced nonlinear phenomena. IEEE Trans. Microw. Theory Tech.
tan d 5 0.00010. The saturation magnetization was measured to be 47, 2075–2084 (1999).
Ms 5 1.52 3 105 A m21, with a 3-dB linewidth of the ferromagnetic resonance 29. Yen, T. J. et al. Terahertz magnetic response from artificial materials. Science 303,
1494–1496 (2004).
at DH 5 1.03 3 103 A m21. Using the cyclotron electromagnet at Massachusetts
30. Linden, S. et al. Magnetic response of metamaterials at 100 terahertz. Science 306,
Institute of Technology, we applied a d.c. magnetic field of 0.20 T along the out- 1351–1353 (2004).
of-plane z direction, with a spatial non-uniformity of less than 1.5%. The d.c.
magnetic field breaks the time-reversal symmetry in the photonic crystal. The Supplementary Information is linked to the online version of the paper at
magnetic field strength was measured and calibrated using a LakeShore Model www.nature.com/nature.
410 gaussmeter. Acknowledgements We are very grateful to P. Fisher and U. J. Becker for
generously providing access to the synchrotron magnet at Massachusetts Institute
Full Methods and any associated references are available in the online version of of Technology. We should like to thank I. Chuang, P. Bermel, J. Bravo-Abad,
the paper at www.nature.com/nature. S. Johnson and P. Rakich for comments. This work was supported in part by the
Materials Research Science and Engineering Program of the US National Science
Received 1 June; accepted 15 July 2009. Foundation under award number DMR-0819762, and also in part by the US Army
1. von Klitzing, K., Dorda, G. & Pepper, M. New method for high-accuracy Research Office through the Institute for Soldier Nanotechnologies under contract
determination of the fine-structure constant based on quantized Hall resistance. no. W911NF-07-D-0004.
Phys. Rev. Lett. 45, 494–497 (1980). Author Contributions Z.W., Y.C, J.D.J and M.S. designed the photonic-crystal
2. Tsui, D. C., Stormer, H. L. & Gossard, A. C. Two-dimensional magnetotransport in system, analysed the data and wrote the manuscript. Z.W. and Y.C. fabricated the
the extreme quantum limit. Phys. Rev. Lett. 48, 1559–1562 (1982). structure and performed all the experimental measurements.
3. Novoselov, K. S. et al. Two-dimensional gas of massless Dirac fermions in
graphene. Nature 438, 197–200 (2005). Author Information Reprints and permissions information is available at
4. Zhang, Y. B., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation of the www.nature.com/reprints. Correspondence and requests for materials should be
quantum Hall effect and Berry’s phase in graphene. Nature 438, 201–204 (2005). addressed to Z.W. (zhwang@mit.edu).

775
©2009 Macmillan Publishers Limited. All rights reserved
doi:10.1038/nature08293

METHODS Microwave transmission measurement for bulk crystals and for CESs. Two
identically constructed antennas were inserted through the top copper plate,
Parallel-plate waveguide for out-of-plane confinement. The unidirectional
extending to contact the bottom copper plate. These antennas, labelled A and B
CES waveguide was designed to reproduce the dispersion relation and the modal
in Fig. 1a, were connected by coaxial cables to the two ports of a Hewlett Packard
profile of a topological edge mode of a 2D gyromagnetic photonic crystal, using a
8719C vector network analyser, which measures the transmission coefficients SAB
3D structure with a finite height. The out-of-plane confinement in the z direction
and SBA. Two-port short–open–load–through calibrations were performed at the
was achieved using two parallel horizontal copper plates, separated by 7.0 mm.
coaxial adaptor. Therefore, measured S parameters contain a frequency-dependent
This structure is known as a parallel-plate waveguide in microwave engineer-
insertion loss from the impedance mismatch between the antenna, the feed coaxial
ing18. It supports TEM modes with electric fields pointing in the out-of-plane
cable and the photonic-crystal waveguide, and from the transition between the
z direction and magnetic fields parallel to the x–y plane. This polarization is balanced parallel plates and the unbalanced coax cable. This loss is reciprocal and
identical to the TM modes in 2D photonic crystals where topological modes does not affect the ratio of the transmission coefficients, jSAB/SBAj. Therefore,
have been proposed to exist7. Between the two plates, the electromagnetic fields any substantial difference between jSABj and jSBAj is an experimental signature
of TEM modes are also uniform along the z direction, as in a 2D system. This 3D of the unidirectionality of CESs. We extracted the forward and backward trans-
structure therefore closely mimics a 2D system and is considered to be quasi-2D. mission spectra in a frequency sweep from 1 to 6 GHz. Each measurement was
When operated below 21 GHz, the waveguide supports only TEM modes. performed with an intermediate frequency of 20 Hz and four averages, with the
Single-mode microwave CES waveguide and absorbing boundaries. Similar to power level normalized to the level at the band edges. To measure bulk band gaps
the case of conventional waveguides, if the edge waveguide has too large a cross- (Fig. 3a), antennas A and B were located along the long axis of a 16 3 10 photonic
sectional area it could lead to multimode operation, causing both a uni- crystal, eight lattice constants apart (Supplementary Information). For the CES
directional CES as well as conventional bidirectional modes to be present in waveguide (Figs 3c and 4), we performed the measurement with the feed and probe
the waveguide. To ensure that only a CES is present in the measurement set- antennas located between the copper wall and the 16 3 7 photonic crystal, also
up, we chose the distance between the photonic crystal and the conducting eight lattice constants apart (Fig. 1a), and with the metal wall 9 mm away from each
copper wall to be 25 mm, which is narrow enough to eliminate all bidirectional antenna.
modes at the frequencies of the second band gap. With a 6% relative bandwidth Effects of material absorption loss. Most of the propagation loss in the present
for this band gap, a CES is confined within three lattice constants of the edge, system may be attributed to two sources: the radiation losses originating from the
even around a large scatterer. The copper scatterer had a height of 7.0 mm and a finite width of the photonic-crystal cladding and the intrinsic material absorption
width of 7.2 mm, with its maximum length mainly limited by the finite size of the associated with the ferromagnetic resonance. The radiation loss could be further
crystal used in this experiment. Microwave-absorbing foam pieces were placed reduced simply by increasing the number of unit cells in the lateral direction,
along the other three edges of the photonic crystal, to prevent the CES from whereas the absorption loss could in principle be further reduced by using mono-
circulating all the way around the boundary of the crystal. In addition, these crystalline yttrium–iron-garnet as the ferrite material18. The resultant attenuation
foam pieces shielded the system from external interference. length would be on the order of hundreds of lattice constants.

©2009 Macmillan Publishers Limited. All rights reserved