Series Editors: Chennupati Jagadish

SERIES EDITORS
CHENNUPATI JAGADISH
Distinguished Professor
Department of Electronic Materials Engineering
Research School of Physics and Engineering
Australian National University
Canberra, ACT2601, Australia
ZETIAN MI
Professor
Department of Electrical Engineering and Computer Science
University of Michigan
1310 Beal Avenue
Ann Arbor, MI 48109
United States of America
Academic Press is an imprint of Elsevier
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
525 B Street, Suite 1650, San Diego, CA 92101, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
125 London Wall, London, EC2Y 5AS, United Kingdom
First edition 2019
© 2019 Elsevier Inc. All rights reserved
No part of this publication may be reproduced or transmitted in any form or by any means, electronic
or mechanical, including photocopying, recording, or any information storage and retrieval system,
without permission in writing from the publisher. Details on how to seek permission, further
information about the Publisher’s permissions policies and our arrangements with organizations such
as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website:
www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating
and using any information, methods, compounds, or experiments described herein. In using such
information or methods they should be mindful of their own safety and the safety of others, including
parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume
any liability for any injury and/or damage to persons or property as a matter of products liability,
negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas
contained in the material herein.
ISBN: 978-0-12-817542-2
ISSN: 0080-8784
For information on all Academic Press publications

visit our website at https://www.elsevier.com/books-and-journals
Publisher: Zoe Kruze

Acquisition Editor: Jason Mitchell
Editorial Project Manager: Peter Llewellyn
Production Project Manager: Abdulla Sait
Cover Designer: Greg Harris
Typeset by SPi Global, India
Contributors
Isabel Barth
The Department of Physics, University of York, York, United Kingdom. (ch5)
Alexander Drayton
Shanhui Fan
Ginzton Laboratory and Department of Electrical Engineering, Stanford University,
Stanford, CA, United States. (ch1,4,6)
Cheng Guo
Department of Applied Physics, Stanford University, Stanford, CA, United States. (ch4)
Yu Guo
Electrical Engineering and Ginzton Laboratory, Stanford University, Stanford, CA,
United States. (ch3)
Yuri S. Kivshar
Department of Physics and Engineering, ITMO University, St. Petersburg, Russia;
Nonlinear Physics Center, Australian National University, Canberra, ACT, Australia. (ch2)
Thomas F. Krauss
Mikhail F. Limonov
Ioffe Institute; Department of Physics and Engineering, ITMO University, St. Petersburg,
Russia. (ch2)
Mikhail V. Rybin
Ioffe Institute; Department of Physics and Engineering, ITMO University, St. Petersburg,
Russia. (ch2)
Ken Xingze Wang
School of Physics, Huazhong University of Science and Technology, Wuhan, Hubei, China.
(ch3)
Zongfu Yu
University of Wisconsin, Department of Electrical and Computer Engineering, Madison,
WI, United States. (ch3)
Weidong Zhou
Department of Electrical Engineering, University of Texas at Arlington, Arlington, TX,
United States. (ch1,6,7)
ix
Preface
Fano resonance, with its characteristics asymmetric lineshape, differs signifi-

cantly from conventional resonance with an asymmetric Lorentzian lineshape.
In photonics, Fano resonance is first observed from the transmission spectra
of gratings and is responsible for one variety of Wood’s anomalies. This phe-
nomenon and the underlying mechanism, being common and ubiquitous in
many realms of physical sciences, are found in a wide variety of nanophotonic
structures and quantum systems, such as quantum dots, photonic crystals
(PhCs), plasmonics, and metamaterials. The asymmetric and steep dispersion
of the Fano resonance profile promises applicability in a wide range of pho-
tonic devices, such as optical filters, switches, sensors, broadband reflectors,
lasers, detectors, and slow-light and nonlinear devices. With the advances
in nanofabrication processes and the development of integration techniques,
Fano resonance-based materials, structures, devices, and systems have been
widely investigated and reported over the last decade.
This volume collects the state-of-the-art results on Fano resonance pho-
tonic crystal and dielectric metamaterials for photonic and optoelectronic
applications. In Chapter 1, the basic principle of Fano resonance in photonic
crystal slabs was reviewed, which establishes a basic theoretical framework
for the wide varieties of properties observed and to be discussed in this book.
In addition to photonic crystal slabs, dielectric metamaterials also create
a low-loss platform for a variety of applications in photonics involving a
complex manipulation of the amplitude, polarization, and phase of light.
However, for many applications it is important to understand how an arti-
ficial periodic system operating as a photonic crystal with the properties dom-
inated by the Bragg scattering can transform into a metamaterial described by
effective parameters. Chapter 2 provides a comprehensive overview of the
transitions between photonic crystals and metamaterials. Photonic phase
diagram concept was proposed to understand the differences and transition
between photonic crystals and dielectric metamaterials.
In Chapter 3, light trapping in photonic crystal structures was introduced
for solar cell applications. A statistical temporal coupled mode theory of light
trapping was introduced based on a rigorous electromagnetic approach. The
upper limit of the angle-integrated light trapping absorption enhancement is
proportional to the density of states in any given structure. The theory
reveals that the conventional limit can be substantially surpassed when opti-
cal modes exhibit deep-subwavelength-scale field confinement.
xi
xii Preface
Chapter 4 discusses the implementation of several isotropic image filters

in the wavevector domain using a single photonic crystal slab device. Such a
slab is designed so that the guided resonance near the Γ point exhibits an
isotropic band structure. Depending on the light frequency and the choice
of transmission or reflection mode, the device realizes isotropic high-pass,
low-pass, band-reject, and band-pass filtering in wavevector space. These
filter functions are important for various image processing tasks, including
edge detection, smoothing, white noise suppression, and suppression or
extraction of periodic noises.
In Chapter 5, label-free biosensing was discussed the versatility of the
photonic crystal guided resonances offers new functionalities, such as mul-
tiparameter sensing and imaging, while simultaneously offering competitive
sensing performance.
In Chapter 6, Fano resonance filters based on single and couple bi-layer
photonic crystal slabs are discussed. These structures possess potentials
toward extreme high quality factor filter via the manipulation of dark and
bright resonance modes.
In Chapter 7, photonic crystal membrane lasers were discussed for
on-chip integration. Based on Fano resonance principles, broadband reflec-
tors were incorporated in a new types of vertical cavity surface-emitting
lasers. Based on bandedge mode, photonic crystal surface-emitting lasers
were also introduced for on-chip integration based on transfer printing
process for hybrid integration.
As seen above, photonic crystals and dielectric metamaterials offer great
potential for a wide range of photonic applications. While this book offers a
few selected application cases, its potential goes beyond linear photonic
applications. Nonlinear and quantum applications are also expected to show
much more great progresses. We hope that this volume will be a useful
reference book for the scientists and engineers striving for photonic crystal
and dielectric metamaterials for new devices and integrated photonic
applications.
WEIDONG ZHOU
University of Texas at Arlington
SHANHUI FAN
Stanford University
CHAPTER ONE
Fano resonance principles

in photonic crystal slabs
Weidong Zhoua,*, Shanhui Fanb
a
Department of Electrical Engineering, University of Texas at Arlington, Arlington, TX, United States
b
Department of Electrical Engineering, Stanford University, Stanford, CA, United States
*Corresponding author: e-mail address: wzhou@uta.edu
Contents
1. Introduction 1
2. Principles of Fano resonance in photonic crystal slabs 2
2.1 2D photonic crystal slabs 2
2.2 The presence of Fano resonance in photonic crystal slabs 3
2.3 Characteristics of Fano resonance in photonic crystal slabs 5
References 10
1. Introduction
In contrast to a conventional symmetric-shaped Lorentzian resonance,
Fano resonance is predominantly used to describe asymmetric resonance
(Fano, 1961; Ott et al., 2013), which arises from the constructive and
destructive interference of discrete resonance states by broadband continuum
states. First observed from the transmission spectra of gratings, Fano reso-
nance is responsible for one variety of Wood’s anomalies (Wood, 1902). This
phenomenon and the underlying mechanism, being common and ubiquitous
in many realms of physical sciences (Fano, 1961; Miroshnichenko et al.,
2010), are found in a wide variety of nanophotonic structures and quantum
systems, such as quantum dots, photonic crystals (PhCs), plasmonics, and
metamaterials. The asymmetric and steep dispersion of the Fano resonance
profile promises applicability in a wide range of photonic devices, such
as optical filters, switches, sensors, broadband reflectors, lasers, detectors,
slow-light, and nonlinear devices. With the advances in nanofabrication pro-
cesses and the development of integration techniques, Fano resonance-based
materials, structures, devices, and systems have been widely investigated and
reported over the last few years. There are also excellent reviews written
Semiconductors and Semimetals, Volume 100 # 2019 Elsevier Inc. 1

ISSN 0080-8784 All rights reserved.
https://doi.org/10.1016/bs.semsem.2019.04.001
2 Weidong Zhou and Shanhui Fan
by many experts in the field. Miroshnichenko et al. (2010) presented a detailed

review on the principles and origins of Fano resonance in various nano-
structures and physical systems. Luk’yanchuk et al. (Luk’yanchuk et al.,
2010; Rahmani et al., 2013) and Giannini et al. (Giannini et al., 2011)
presented reviews on progress made on Fano resonance in plasmonic
nanostructures and metamaterials.
Optical gratings, which can also be viewed as 1D PhCs, are a research
topic with a long history, including guided mode resonances (GMRs) in
1D gratings, reported by Magnusson et al. (Kikuta et al., 2003; Magnusson
and Shokooh-Saremi, 2008; Magnusson and Wang, 1992). Recently, due
to the unique characteristics and potentials for high performance optic and
photonic devices, subwavelength high-index-contrast gratings (HCGs) have
been proposed and investigated by Chang-Hasnain et al. (Boutami et al.,
2007; Carletti et al., 2011; Chang-Hasnain, 2011; Chen et al., 2006) for
applications in mirrors, lasers, and focusing elements (Chang-Hasnain,
2011). Here we focus our review on progress made in PhC Fano resonance
photonics, with a specific focus on Fano resonances in two-dimensional
photonic crystal slabs (2D PCS) (Fan, 2002; Fan and Joannopoulos, 2002;
Zhou et al., 2014).
2D PCS structures also support guided resonances (Fan and Joannopoulos,
2002; Tikhodeev et al., 2002). In one of the original papers on this topic, Fan
and Joannopoulos (2002) presented a detailed analysis of the guided reso-
nances in 2D PCS structures and investigated the transmission/reflection
spectra properties in such systems when these modes are coupled with vertical
radiation modes. Subsequently, many theoretical and experimental papers
appeared on this topic with investigations on associated physical characteristics
of such resonances and related device demonstrations.
2. Principles of Fano resonance in photonic crystal slabs

2.1 2D photonic crystal slabs
2D PCSs are a particularly important class of PhC structures ( Joannopoulos
et al., 2008; Johnson et al., 1999). A 2D PCS consists of a two-dimensionally
periodic index contrast introduced into a high-index guiding layer. As
shown in Fig. 1A, PCSs are quasi-3D structures in which the light confine-
ment arises from the in-plane 2D photonic bandgap (PBG) confinement and
vertical waveguide index confinement (through the total internal reflection
or TIR principle). 2D PCS structures are certainly more amenable to
Fano resonance principles in photonic crystal slabs 3
A B
Photonic crystal slab (PCS) ment
Sp Em
issi
nfine on
co co
on nt
ot PBG r
ol
Ph
Photonic
bandgap
En
e
GVA PDOS
tim
han
Group Photonic
interaction
velocity density of
ced absorption
anomaly states
PhC
0.8 Photonic
crystals
anced
Frequency (c/a)
0.6 SPP NRI

Surface Negative
Enh
refractive
Sup
plasmon
0.4 Photonic bandgap index
NLO
er
TM
n
Nonlinear
pr
tio
is
0.2 a M K a optical m
2r
er properties
en An
G cg iso
TE oni trop
0 Second harm y
G M K G
Fig. 1 Two-dimensional photonic crystal slabs (2D PCS): (A) Schematic representation of
triangular lattice PCS, with simulated dispersion plot for two (TE/TM) polarizations; and
(B) Overview of photonic crystal properties and applications. Reproduced with permission
from Zhou, W., Zhao, D., Shuai, Y.-C., Yang, H., Chuwongin, S., Chadha, A., Seo, J.-H.,
Wang, K. X., Liu, V., Ma, Z., 2014. Progress in 2D photonic crystal Fano resonance photonics.
Prog. Quantum Electron. 38 (1), 1–74.
fabrication of devices using current semiconductor technology than three-

dimensional structures, yet they provide much the same amount of func-
tionality (Fig. 1B) (Zhou, 2011). 2D PCS can be easily integrated with other
photonic components for photonic integrated circuits (PICs) using planar
micro- and nanofabrication processes ( John et al., 2004; Johnson et al.,
2001; Mekis et al., 1996; Noda and Baba, 2003; Noda et al., 2002;
Notomi, 2011; Seassal et al., 2002). One of the most widely studied
structures is the triangular lattice-based air column 2D PCS structure, which
possesses a large transverse-electric (TE) and transverse-magnetic (TM) like
mode and a large complete TE/TM PBG. A typical dispersion plot is shown
in Fig. 1A, where a complete TE/TM bandgap is shown. Note that a light
core region is formed due to the vertical confinement. These structures
support in-plane-guided modes below the lightlines, which are completely
confined by the slab without any coupling to external radiations.
2.2 The presence of Fano resonance in photonic crystal slabs

In addition to in-plane waveguiding, PCS can also interact with external
radiation in complex and interesting ways. Of particular importance here
is the presence of guided resonance in the structures (Fan and Joannopoulos,

2002; Tikhodeev et al., 2002). Similar to the guided mode, for guided
resonance the electro-magnetic power is also strongly confined within
the slab. Unlike the guided mode, however, the resonance can be coupled
with external radiation (the leaky mode is above the lightline Fig. 1A).
Therefore, guided resonance can provide an efficient way to channel light
from within the slab to the external environment. A detailed theoretical
analysis of guided resonances in 2D PCS’s was first presented by Fan and
Joannopoulos (2002). Kanskar et al. (1997) presented one of the first exper-
imental reports of guided resonances at optical frequencies in a 2D PCS
structure based on a suspended AlGaAs layer.
For light incident on PCS from out of the plane direction (e.g., surface-
normal direction), photon transport can take two pathways: a direct or
background pathway that does not involve the guided resonance or a
resonance-assisted pathway in which photons are coupled into the reso-
nances and subsequently leak out to free space. Assuming that the slab
has mirror symmetry in the vertical direction, and taking into consideration
the interference between these two pathways, we express the transmission
amplitude t as a function of frequency ω
γ
t ¼ td + f (1)
iðω ω0 Þ + γ
γ
r ¼ rd f (2)
iðω ω0 Þ + γ
where td and rd are the direct transmission and reflection coefficients, ω0 and
γ are the center frequency and width of the resonance, and f is the normal-
ized complex amplitude of the resonant mode. The plus and minus sign
correspond with even and odd resonant modes for structures with the mirror
symmetry. Energy conservation requires
f ¼ ðtd rd Þ (3)
For 0 < jtd j < 1, Eq. (1) gives a Fano resonance lineshape. For example,
we consider a system consisting of a square lattice of air holes introduced into
a dielectric slab (Fig. 2A). The intensity transmission spectrum is given in
Fig. 2B and C, exhibiting Fano resonance lineshapes with strong asymmetric
profiles. Using finite-difference time-domain (FDTD) simulations, one can
determine the parameters for Eq. (1) and compare the spectrum from this
theoretical derivation to that from the FDTD simulation (Fig. 2B and C).
A B 1
Transmission
0.8
0.6
0.4
0.2
0
0.25 0.3 0.35 0.4 0.45
Frequency (c/a)
C
1
Transmission
0.8
0.6
0.4
0.2
0
0.36 0.38 0.4 0.42
Frequency (c/a)
Fig. 2 Fano resonances in 2D photonic crystal slabs: (A) Photonic crystal structure con-
sisting of a square lattice of air holes with a radius of r ¼ 0.2a in a dielectric slab with
dielectric constant nh ¼ 12 and a thickness (t) of 0.5a, where a is the lattice constant.
The arrow indicates the direction of the incident light. (B) The intensity transmission
spectrum through such a structure. The circles are the result of the finite-difference
time-domain simulations. The solid curve is determined by the analytic theory as repre-
sented by Eq. (1). (C) The same plot as (B), except the frequency range is now restricted
to [0.36(c/a), 0.42(c/a)] to exhibit further details of the resonance line shape. Reproduced
with permission from Fan, S., Suh, W., Joannopoulos, J., 2003. Temporal coupled-mode
theory for the Fano resonance in optical resonators. J. Opt. Soc. Am. 20 (3), 569–572.
In the vicinity of the Fano resonance, the intensity transmission in

Fig. 2B undergoes a transition between 0% and 100%, meaning the con-
structive and destructive interferences are located close to each other. This
phenomenon is due to the rapid variation of the phase along the localized
(resonance-assisted) pathway compared to the background pathway.
2.3 Characteristics of Fano resonance in photonic crystal slabs

While Eq. (1) intuitively captures the interference between the background
and the resonance-assisted pathways, it can be naturally derived from the
more general and phenomenological temporal coupled-mode theory
( Joannopoulos et al., 2008), in which a system is modeled as a set of localized
and propagating modes. Instead of introducing the general formalism
dealing with an arbitrary number of resonances and waveguides (Suh
et al., 2004), Fan et al. (2003) focused on the simple case of a single resonance
coupled to two ports to better illustrate the essence of the physical process.
This two-port configuration could be guaranteed in the structure when the
light’s wavelength is greater than the lattice constant such that there is no
diffraction above or below the slab, as shown in Fig. 2A.
The dynamics of this lossless single-resonance double-port system is
described by the coupled-mode equations

da 1 s1 +
¼ iω0 a + ðκ 1 κ2 Þ (4)
dt τ s2 +

s1 s1 + d1
¼C +a (5)
s2 s2 + d2
where a is the amplitude of the resonant mode such that j aj2 represents the
electromagnetic energy stored in the resonance. ω0 and γ ¼ 1/τ are the cen-
ter frequency and the width of the resonance, respectively. s1+ and s2+ are
the incoming amplitudes of the plane waves on the two ports such that j s1+j2
and js2+j2 correspond with the intensities. s1 and s2 are the outgoing
amplitudes of the plane waves from the two ports such that j s1j2 and js2j2
correspond with the intensities. κ1 and κ2 are the coupling constants between
the incoming waves and the resonance. d1 and d2 are the coupling constants
between the outgoing waves and the resonance. C is the background scatter-
ing matrix.
The coupling constants are constrained by energy conservation and
time-reversal symmetry as follows (Suh et al., 2004):
2
jd1 j2 + jd2 j2 ¼ (6)
τ
κ 1 ¼ d1 ; κ2 ¼ d2 (7)
∗
d1 d1
wC ∗ ¼ (8)
d2 d2
By choosing the appropriate positions of the reference planes, without
loss of generality, the direct scattering matrix C can be expressed as

iϕ r it
C¼e (9)
it r
where ϕ is a phase factor, and r and t, both being real, are the reflection and
transmission coefficients of the direct process satisfying r2 + t2 ¼ 1 (Haus,
1984). Combining Eqs. (6–9), one can solve for the coupling constants in
terms of the decay rate and the elements in the direct scattering matrix
(Wang et al., 2013), and the theory accounts for asymmetric photonic crystal
slabs (Rosenberg et al., 2005). For the structure with mirror symmetry in
Fig. 2A, one can further simplify the solution by requiring d1 ¼ d2, and
the intensity reflection coefficient R is therefore
r 2 ðω ω0 Þ2 + t2 ð1=τÞ2 2rtðω ω0 Þð1=τÞ

R¼ (10)
ðω ω0 Þ2 + ð1=τÞ2
which agrees with Eq. (2).

Typically there are two techniques for the simulation of the transmission
and reflection properties of PCS structures. The first approach is the rigorous
coupled wave analysis (RCWA) technique (Moharam and Gaylord, 1981), a
relatively straightforward technique for obtaining the exact solution of
Maxwell’s equations for the electromagnetic diffraction of grating structures.
It is a noniterative, deterministic technique utilizing a state-variable method
that converges with the proper solution without inherent numerical insta-
bilities. The accuracy of the solution obtained depends solely on the number
of terms in the field space-harmonic expansion, ensuring that the energy is
always conserved (Moharam et al., 1995). Recently, a Fourier Modal
Method-based package was developed by Liu and Fan, and it is now freely
available as Stanford Stratified Structure Solver (S4) software package (Liu
and Fan, 2012). Compared to the commercial RCWA package GD-Calc
( Johnson, 1981), the S4 software package offers much higher resolution
due to the consideration of higher diffraction orders and also takes much less
computation time (Shuai et al., 2013).
The other widely used simulation approach is the three-dimensional
finite-difference time domain (3D FDTD) technique (Taflove and
Hagness, 2000). In this case, a unit cell is defined by using a periodic boundary
condition (PBC) and a perfectly matched layer (PML) in the four lateral and
two vertical directions (Qiang et al., 2008), as shown in Fig. 3A. A Gaussian
source is launched from the top of the PCS structure, with two power mon-
itors monitoring both the reflected and the transmitted power. For practical
structures, where the number of the PhC lattice period is finite, full vectorial
3D FDTD can also offer very precise and accurate simulation results.
Depending on the PCS lattice parameters, either Fano resonance filters or
broadband reflectors can be realized with different Q values. A set of simulated
spectra for transmission and reflection is shown in Fig. 3B and C, respectively.
High Q filters with Q greater than 1000 can be easily obtained when a smaller
r/a value is used (r/a ¼ 0.08, t/a ¼ 0.34, a ¼ 765 nm for high Q case here).
A B
1
PML
Source 0.8 (i) High Q
Transmission
0.6
PT
0.4 (ii) Medium Q
C
PB
PBC 0.2
PB (iii) Low Q
0
1.44 1.49 1.54 1.59 1.64
PML Wavelength (µm)
C
1
(iii) Low Q
0.8
Reflection
0.6 (ii) Medium Q
0.4
(i) High Q
0.2
0
1.44 1.49 1.54 1.59 1.64
Wavelength (µm)
Fig. 3 Simulation and spectral control of Fano resonances in 2D PCS: (A) Schematic
of an unit cell used in the simulation based on 3D-FDTD technique; Simulated
(B) transmission and (C) reflection spectra for surface-normal Fano filters with different
quality factors (Qs) for either highly spectrally selective filters or broadband reflectors.
Reproduced with permission from Zhou, W., Ma, Z., Yang, H., Qiang, Z., Qin, G., Pang, H.,
Chen, L., Yang, W., Chuwongin, S., Zhao, D., 2009. Flexible photonic-crystal Fano filters
based on transferred semiconductor nanomembranes. J. Phys. D Appl. Phys. 42, 234007.
On the other hand, a larger r/a value can lead to a lower Q filters for broad-
band reflector design (r/a ¼ 0.28, t/a ¼ 0.347, a ¼ 980 nm for low Q case
here). A medium Q design was also presented with r/a ¼ 0.19, t/a ¼ 0.417,
and a ¼ 600nm.
The resonance modes were further verified with the field propagation
plots based on 3D FDTD simulations. Snapshots of field propagation for
the on- and the off-resonance modes are shown in Fig. 4A and B, respec-
tively. Note that, for the on-resonance mode (λ1), the surface-normal
incident light is reflected from the patterned PCS structure due to the coher-
ence (in-phase) reflection, which leads to a dip in the transmission spectra.
A On resonance (l1) B Off resonance

Max
–Max
Fig. 4 Snap shots of electric field distribution of Fano filters at (A) on-resonance and
(B) off-resonance conditions. Reproduced with permission from Zhou, W., Ma, Z.,
Yang, H., Qiang, Z., Qin, G., Pang, H., Chen, L., Yang, W., Chuwongin, S., Zhao, D., 2009.
Flexible photonic-crystal Fano filters based on transferred semiconductor nanomembranes.
J. Phys. D Appl. Phys. 42, 234007.
On the other hand, light at other spectra locations (off-resonance) can

pass through the patterned PCS structure with maximum transmission
efficiency.
Shown in Fig. 5 is the simulated dispersion plot for the square lattice PCS
structure, based on a three-dimensional (3D) plane wave expansion (PWE)
technique (Zhou et al., 2009). The Fano resonance modes operate above the
lightline region, denoted as the shaded area. The properties of a few Fano
resonance modes (ω1, ω2, ω3) will be discussed in later sections (Qiang
et al., 2008). Note also that the existence of PBG is not essential for the
design of Fano resonance PCS structures.
In summary, Fano resonances in 2D PCSs arise from the coupling
of in-plane discrete guide modes with the vertical continuum free-space
radiation mode. Depending on the coupling strength, the output spectral
resonance can vary from symmetric to asymmetric, with sharp transition
and large phase change between the resonance peak and dip, which can lead
to extremely high quality factor (or infinite) optical cavities, or very broad-
band high reflectivity reflectors. Local field enhancements associated with
the Fano resonance cavity can result in strong light-matter interactions
for enhanced light emission, detection, sensing, and enhanced optical forces,
as well as other nonlinear effects. The existence of the photonic bandgap is
not essential in Fano resonance PCS structures. The flexibility in dispersion
engineering and the associated slow-light effects can further enhance light-
matter interactions (Boutami et al., 2006; Chen et al., 2006; Fan and
Joannopoulos, 2002; Fan et al., 2003; Kanamori et al., 2007; Lin et al.,
2005; Rosenberg et al., 2005; Suh and Fan, 2004; Taillaert et al., 2002).
0.5
I II w3
Frequency (wa/2pc=a/l)
w2
0.4
w1
0.3
0.2 M
0.1 G X
0.0
G X M G
Fig. 5 Simulated dispersion characteristics for the square lattice Si PCS structure on low
index glass substrate. The lattice parameters are r/a ¼ 0.19, t/a ¼ 0.417, the refractive
indices of silicon and glass are 3.48 and 1.5, respectively. Reproduced with permission
from Qiang, Z., Yang, H., Chen, L., Pang, H., Ma, Z., Zhou, W., 2008. Fano filters based
on transferred silicon nanomembranes on plastic substrates. Appl. Phys. Lett. 93, 061106.
References
Boutami, S., Bakir, B.B., Hattori, H., Letartre, X., Leclercq, J.L., Rojo-Romeo, P.,
Garrigues, M., Seassal, C., Viktorovitch, P., 2006. Broadband and compact 2-D
photonic crystal reflectors with controllable polarization dependence. IEEE Photon.
Technol. Lett. 18 (7), 835–837.
Boutami, S., Bakir, B., Regreny, P., Leclercq, J., Viktorovitch, P., 2007. Compact 1.55 μm
room-temperature optically pumped VCSEL using photonic crystal mirror. Electron.
Lett. 43, 282.
Carletti, L., Malureanu, R., Mørk, J., Chung, I.S., 2011. High-index-contrast grating
reflector with beam steering ability for the transmitted beam. Opt. Express 19 (23),
23567–23572.
Chang-Hasnain, C.J., 2011. High-contrast gratings as a new platform for integrated
optoelectronics. Semicond. Sci. Technol. 26, 014043.
Chen, L., Huang, M.C.Y., Mateus, C.F.R., Chang-Hasnain, C.J., Suzuki, Y., 2006.
Fabrication and design of an integrable subwavelength ultrabroadband dielectric mirror.
Appl. Phys. Lett. 88, 031102.
Fan, S., 2002. Sharp asymmetric line shapes in side-coupled waveguide-cavity systems. Appl.
Phys. Lett. 80 (6), 908.
Fan, S., Joannopoulos, J.D., 2002. Analysis of guided resonances in photonic crystal slabs.
Phys. Rev. B 65 (23), 235112.
Fan, S., Suh, W., Joannopoulos, J., 2003. Temporal coupled-mode theory for the Fano
resonance in optical resonators. J. Opt. Soc. Am. 20 (3), 569–572.
Fano, U., 1961. Effects of configuration interaction on intensities and phase shifts. Phys. Rev.
124 (6), 1866.
Giannini, V., Francescato, Y., Amrania, H., Phillips, C.C., Maier, S.A., 2011. Fano
resonances in nanoscale plasmonic systems: a parameter-free modeling approach. Nano
Lett. 11 (7), 2835–2840.
Haus, H.A., 1984. Waves and Fields in Optoelectronics. Prentice-Hall, Englewood
Cliffs, NJ.
Joannopoulos, J.D., Johnson, S.G., Winn, J.N., Meade, R.D., 2008. Photonic Crystals:
Molding the Flow of Light, second ed. Princeton University Press.
John, S., Toader, O., Chutinan, A., 2004. Photonic band gap architectures for
micro-fabrication and diffractionless optical networking. In: IEICE Transactions on
Electronics. vol. E87-C(3). The Institute of Electronics, Information and Communication
Engineers, pp. 266–273.
Johnson, K., 1981. Coupled scalar wave diffraction theory. Appl. Phys. 24 (3), 249–260.
Johnson, S.G., Shanhui, F., Villeneuve, P.R., Joannopoulos, J.D., Kolodziejski, L.A.,
1999. Guided modes in photonic crystal slabs. Phys. Rev. B Condens. Matter 60 (8),
5751–5758.
Johnson, S.G., Povinelli, M.L., Joannopoulos, J.D., 2001. New photonic-crystal
system for integrated optics. In: Active and Passive Optical Components for WDM
Communication. SPIE-Int. Soc. Opt. Eng, Denver, CO.
Kanamori, Y., Kitani, T., Hane, K., 2007. Control of guided resonance in a photonic crystal
slab using microelectromechanical actuators. Appl. Phys. Lett. 90, 031911.
Kanskar, M., Paddon, P., Pacradouni, V., Morin, R., Busch, A., Young, J.F., Johnson, S.R.,
MacKenzie, J., Tiedje, T., 1997. Observation of leaky slab modes in an air-bridged
semiconductor waveguide with a two-dimensional photonic lattice. Appl. Phys. Lett.
70 (11), 1438–1440.
Kikuta, H., Toyota, H., Yu, W., 2003. Optical elements with subwavelength structured
surfaces. Opt. Rev. 10 (2), 63–73.
Lin, C., Lu, Z., Shi, S., Jin, G., Prather, D.W., 2005. Experimentally demonstrated filters
based on guided resonance of photonic-crystal films. Appl. Phys. Lett. 87, 091102.
Liu, V., Fan, S., 2012. S4: a free electromagnetic solver for layered periodic structures.
Comput. Phys. Commun. 183 (10), 2233–2244.
Luk’yanchuk, B., Zheludev, N.I., Maier, S.A., Halas, N.J., Nordlander, P., Giessen, H.,
Chong, C.T., 2010. The Fano resonance in plasmonic nanostructures and metamaterials.
Nat. Mater. 9 (9), 707–715.
Magnusson, R., Shokooh-Saremi, M., 2008. Physical basis for wideband resonant reflectors.
Opt. Express 16 (5), 3456–3462.
Magnusson, R., Wang, S.S., 1992. New principle for optical filters. Appl. Phys. Lett.
61, 1022.
Mekis, A., Chen, J.C., Kurland, I., Fan, S., Villeneuve, P.R., Joannopoulos, J.D., 1996. High
transmission through sharp bends in photonic crystal waveguides. Phys. Rev. Lett.
77 (18), 3787–3790.
Miroshnichenko, A.E., Flach, S., Kivshar, Y.S., 2010. Fano resonances in nanoscale structures.
Rev. Mod. Phys. 82 (3), 2257.
Moharam, M., Gaylord, T., 1981. Rigorous coupled-wave analysis of planar-grating diffraction.
J. Opt. Soc. Am. 71 (7), 811–818.
Moharam, M., Grann, E.B., Pommet, D.A., Gaylord, T., 1995. Formulation for stable and
efficient implementation of the rigorous coupled-wave analysis of binary gratings. J. Opt.
Soc. Am. A 12 (5), 1068–1076.
Noda, S., Baba, T., 2003. Roadmap on Photonic Crystals. Springer.
Noda, S., Imada, M., Okano, M., Ogawa, S., Mochizuki, M., Chutinan, A., 2002.
Semiconductor three-dimensional and two-dimensional photonic crystals and devices.
IEEE J. Quantum Electron. 38 (7), 726–735.
Notomi, M., 2011. Strong light confinement with periodicity. Proc. IEEE 99 (10),
1768–1779.
Ott, C., Kaldun, A., Raith, P., Meyer, K., Laux, M., Evers, J., Keitel, C.H., Greene, C.H.,
Pfeifer, T., 2013. Lorentz meets Fano in spectral line shapes: a universal phase and its laser
control. Science 340 (6133), 716–720.
Qiang, Z., Yang, H., Chen, L., Pang, H., Ma, Z., Zhou, W., 2008. Fano filters based
on transferred silicon nanomembranes on plastic substrates. Appl. Phys. Lett. 93,
061106.
Rahmani, M., Luk’yanchuk, B., Hong, M., 2013. Fano resonance in novel plasmonic
nanostructures. Laser Photonics Rev. 7 (3), 329–349.
Rosenberg, A., Carter, M., Casey, J., Kim, M., Holm, R., Henry, R., Eddy, C.,
Shamamian, V., Bussmann, K., Shi, S., Prather, D.W., 2005. Guided resonances in
asymmetrical GaN photonic crystal slabs observed in the visible spectrum. Opt. Express
13 (17), 6564–6571.
Seassal, C.D.Y., Letartre, X., Grillet, C., Rojo-Romeo, P., Viktorovitch, P., Benyattou, T.,
2002. Optical coupling between a two-dimensional photonic crystal-based microcavity
and single-line defect waveguide on InP membranes. IEEE J. Quantum Electron. 38 (7),
811–815.
Shuai, Y., Zhao, D., Tian, Z., Seo, J., Plant, D.V., Ma, Z., Fan, S., Zhou, W., 2013. Double-
layer Fano resonance photonic crystal filters. Opt. Express 21 (21), 24582–24589.
Suh, W., Fan, S., 2004. All-pass transmission or flattop reflection filters using a single photonic
crystal slab. Appl. Phys. Lett. 84, 4905.
Suh, W., Wang, Z., Fan, S., 2004. Temporal coupled-mode theory and the presence of
non-orthogonal modes in lossless multimode cavities. IEEE J. Quantum Electron.
40 (10), 1511–1518.
Taflove, A., Hagness, S.C., 2000. Computational Electrodynamics: The Finite-Difference
Time-Domain Method. Artech House.
Taillaert, D., Bogaerts, W., Bienstman, P., Krauss, T., Van Daele, P., Moerman, I.,
Verstuyft, S., De Mesel, K., Baets, R., 2002. An out-of-plane grating coupler for efficient
butt-coupling between compact planar waveguides and single-mode fibers. IEEE J.
Quantum Electron. 38 (7), 949–955.
Tikhodeev, S.G., Yablonskii, A.L., Muljarov, E.A., Gippius, N.A., Ishihara, T., 2002.
Quasiguided modes and optical properties of photonic crystal slabs. Phys. Rev. B
66 (4), 45102.
Wang, K.X., Yu, Z., Sandhu, S., Fan, S., 2013. Fundamental bounds on decay rates in
asymmetric single-mode optical resonators. Opt. Lett. 38 (2), 100–102.
Wood, R., 1902. XLII. On a remarkable case of uneven distribution of light in a diffraction
grating spectrum. Lond. Edinb. Phil. Mag. 4 (21), 396–402.
Zhou, W., 2011. Photonic crystal surface-emitting lasers and infrared sensors. In: Nalwa, H.S.
(Ed.), Encyclopedia of Nanoscience and Nanotechnology. vol. 20. American Scientific
Publishers.
Zhou, W., Ma, Z., Yang, H., Qiang, Z., Qin, G., Pang, H., Chen, L., Yang, W.,
Chuwongin, S., Zhao, D., 2009. Flexible photonic-crystal Fano filters based on
transferred semiconductor nanomembranes. J. Phys. D Appl. Phys. 42, 234007.
Zhou, W., Zhao, D., Shuai, Y.-C., Yang, H., Chuwongin, S., Chadha, A., Seo, J.-H.,
Wang, K.X., Liu, V., Ma, Z., 2014. Progress in 2D photonic crystal Fano resonance
photonics. Prog. Quantum Electron. 38 (1), 1–74.
CHAPTER TWO
Transition from photonic crystals

to dielectric metamaterials
Mikhail V. Rybina,b,*, Mikhail F. Limonova,b, Yuri S. Kivsharb,c
a
Ioffe Institute, St. Petersburg, Russia
b
Department of Physics and Engineering, ITMO University, St. Petersburg, Russia
c
Nonlinear Physics Center, Australian National University, Canberra, ACT, Australia
*Corresponding author: e-mail address: m.rybin@mail.ioffe.ru
Contents
1. Introduction 13
2. Photonic crystals vs metamaterials 15
3. Dielectric rods arranged in a square lattice 19
4. Experimental observation of the photonic phase transitions 24
5. Phase diagram of periodic dielectric photonic structures 27
6. Dielectric metamaterials with near-zero effective parameters 33
7. Toward practical realization of metamaterials in the visible frequency range 35
8. Conclusion and outlook 41
Acknowledgments 42
References 42
1. Introduction
In the late 1980s, Yablonovich (1987) and John (1987) introduced a
new class of artificial dielectric media—photonic crystals—that are periodic
structures having a spatial modulation of the dielectric index with a period
of half a wavelength and larger. That time, advanced nanotechnologies
became mature enough to support the fabrication of such structures with
features of several 100 nm in size. The idea of that concept was to apply
the solid-state theory of electrons to the case of electromagnetic waves
and to develop a class of “semiconductors for light” materials, which would
allow many novel applications. These photonic crystals operate due to spa-
tially extended Bragg resonances that appear as a result of the structure
periodicity.

14 Mikhail V. Rybin et al.
At the turn of the century, another class of artificial electromagnetic

media—metamaterials—was proposed by Pendry et al. (1999) and Smith
et al. (2000). Although there are many definitions of metamaterials, here
we follow to the seminal papers and define metamaterials as artificial media
composed of resonant meta-atoms and described by the effective material
parameters (dielectric permittivity ε and magnetic permeability μ) similar
to many natural materials. The main feature of metamaterials is a weak spatial
dispersion (Simovski, 2018) in contrast to photonic crystals, whose resonant
frequencies correspond to the case of strong spatial dispersion because
the Bragg resonances are formed by periodically arranged meta-atoms.
Metamaterials allow to design media with engineered on demand spatial
distribution of ε and μ, which has applications in invisibility (Pendry et al.,
2012), hyperlenses (Liu et al., 2007), etc. Unlike dielectric permittivity, natural
materials have negligible magnetic response and metamaterials pave the way
for achieving negative μ in the optical range. In particular, double negative
ε(r) and μ(r) metamaterials make it possible to realize Veselago’s left-handed
media (Shelby et al., 2001; Veselago, 1968).
The first metamaterial was composed of meta-atoms being thin wire
RF antennas. Split-ring resonators are widely used to achieve an effective
resonant magnetic response of artificial media with some values of negative
μ. The split-ring resonators can be much smaller relative to the resonant
wavelength that enables an effective media description, however at high
frequencies the electric field demonstrates essential Ohmic losses, which
is a great disadvantage of metal-based meta-atoms. In 2002, O’Brien and
Pendry suggested high-index dielectrics meta-atoms, which support mag-
netic Mie resonances for the TE polarization (magnetic field oscillates along
the rods). They considered the rods with ε ¼ 200 and evaluated an effective
magnetic permeability μ. Although the correctness of their effective param-
eter is a subject of debates (Simovski, 2009), the existence of dielectric
metamaterials with negative magnetic response was established.
The main difficulty in the experimental realization of dielectric
metamaterials is that the size of meta-atoms (e.g., the diameter of rods)
for materials with ε ¼ 12 (the value available for many high-index optical
materials) is about 5 times smaller than the resonant wavelength, and this
makes the theoretical assumption about the subwavelength operational
regime of such structures not applicable. However, numerical simulations
suggest that the effective homogenization can work even for relatively large
meta-atoms. On the other hand, low-index dielectric periodic structures are
Transition from photonic crystals to metamaterials 15
presented by the well-studied photonic crystals (see, e.g., Joannopoulos

et al., 2008) and thus one may ask a simple question: When a photonic crystal
becomes a metamaterial provided one varies the permittivity from low to high values?
In this chapter, we review the recent studies related to this problem and
introduce the concept of a photonic phase diagram (see Rybin et al., 2015).
2. Photonic crystals vs metamaterials

Phase transition terminology has been used in a number of articles to
discuss effects in artificial photonic structures (Gorlach and Belov, 2014;
Mirmoosa et al., 2015; Reyes-Gómez et al., 2014). We briefly remind what
the phase transitions in thermodynamics are. First we notice that phase tran-
sitions of the first order (i.e., solids become liquids) involving dramatic
changes in location of atoms and their interaction has nothing in common
with the effects observed in photonics. Other thermodynamic phase transi-
tions are related to second order or continuous phase transitions. In most
cases they are associated with changes in electronic properties, which make
it possible to observe features, for example, in the temperature dependences
of the heat capacity, changes in susceptibility, and so on. The physical origin
of the continuous phase transitions is that a change in external parameters
(temperature, pressure, etc.) leads to a continuous transformation of the spa-
tial arrangement of atoms that in turn results in a rapid modification of elec-
tronic subsystem, in particular, a small displacement of atoms results in a
dramatic transformation of band structure.
Unlike conventional materials, photonic structures are fabricated by an
arrangement of desired meta-atoms according to a chosen design. In prin-
ciple, such artificial structures allow one to change the distances between
meta-atoms and crystal-lattice symmetry without involving of external
influence (e.g., temperature or pressure). However, this modification of sys-
tem parameters might result in dramatic changes in photonic properties
being similar to the continuous phase transitions in the conventional mate-
rials. We discuss one important difference in these two cases. The Pauli’s
exclusive principle prohibits to occupy the same state by different particles
does not apply to photons, and hence, Fermi energy is not applicable to pho-
tonic structures. Instead, the energy (frequency) of photons is defined by
sources of the electromagnetic waves. Thus, frequency is often considered
as an additional structure parameter along with lattice spacing and constit-
uent permittivity. For example, a frequency-dependent phase transition
between hyperbolic and elliptic regimes of metamaterials was considered

(Reyes-Gómez et al., 2014). However, the term “frequency-dependent
phase transition” is a confusing point, because a true phase cannot depend
on probing parameters, which include the radiation frequency.
Before we give a definition of phase for photonic crystals or metamaterials,
let us briefly consider their physical properties and band diagrams assuming the
periodic systems. The Bloch theorem makes it possible to classify all solutions
by a (quasi) wave vector k in the form of exp(ikr)X(r), where X(r) is a peri-
odic function with the translational symmetry of the crystal lattice. We notice
that multiplying the function X(r) by the exponent exp(igr) does not change
the translational symmetry of the product, given g is any vector of the recip-
rocal lattice. It means that the unique solutions related to different wave
vectors are restricted by the Brillouin zone only, which is defined as a region
of the reciprocal space containing wave vectors being closer to the origin
than to other reciprocal lattice nodes. In the case of a small dielectric
contrast, i.e., in a photonic crystal regime, the deviation of the permittivity
from its average value can be considered as a small perturbation. The photonic
band diagram (a dependence of structure eigenfrequency ω on wave vector k)
for the unperturbed problem is the light cone defined by the dispersion
equation ω2 ¼ c2 jkj2ε1 (c is the light velocity constant), but the ambiguity
of the choice of the wave vector in the periodic structure results in a set of
light cones with apex at each nodes of the reciprocal lattice (Fig. 1A). As long
as the dielectric contrast is zero, the modes corresponding to different light
A B
Wave vector Wave vector

Fig. 1 Difference between the regimes of photonic crystals and metamaterials.
(A) Typical bandgap diagram of a photonic crystal: the interaction between two light
cones (gray dashed lines) results in a frequency gap (shaded in green) at the boundary
of the Brillouin zone (for example, at the X point). (B) Typical bandgap diagram of a
metamaterial: a coupling of the almost flat Mie resonance branch (orange dashed line)
and the light cone (gray dashed line) leads to a polariton-like feature and the Mie gap
(shaded in orange). Reproduced with permission from Li, S.V., Kivshar, Y.S., Rybin, M.V.,
2018. Toward silicon-based metamaterials. ACS Photonics 5, 4751–4757. Copyright
(2018) ACS.
cones do not mix each other, and only the main light cone at the origin k ¼ 0
is enough for a complete description of the electrodynamic problem. How-
ever, a nonzero dielectric contrast perturbs the states between different light
cones at the boundary of the Brillouin zone that results in opening of
bandgaps related to the certain direction of light. These bandgaps arise
due to Bragg scattering on the periodic structure. Fig. 1A sketches a pair
of unperturbed light cones with vertices at the Γ and Γ 0 points that cross
at the boundary of the Brillouin zone at the X point defining the lowest
intersection point for a square lattice. We consider the unique region of
wave vectors on the ΓX segment. The branch corresponding to the light
cone starts from the Γ point at a certain angle defining the effective refractive
index and ends at the X point. Near the X point the dispersion dependence
ω(k) of the lowest branch demonstrates a deviation from a linear function.
At high frequencies the second branch appears going from the X point to
the Γ point, and the Bragg bandgap is observed between the high- and
low-frequency branches.
Now we consider the band diagram of metamaterials, whose meta-atoms
support a local resonance at some frequency. Owing to the uncertainty
relation, the wave vector of the resonant state is delocalized in the reciprocal
space. At the same time, being localized in the real space, the resonance can
be considered as a quasi-discrete state with a certain frequency. Fig. 1B
exhibits such states by the horizontal orange dotted line, which means that
the frequency is well determined and the wave vector is uncertain. Beside
this, Fig. 1B demonstrates light cones corresponding to propagating waves
(gray dotted lines). A polariton-like feature appears in the band diagram due
to a mixing of the propagating waves with a localized state (Kaina et al.,
2015; Yves et al., 2017). The branch starting at the Γ point at ω ¼ 0 dem-
onstrates a linear dependence at low frequencies and a deviation from the
line as it approaches to the eigenfrequency of the localized mode and trans-
forms into a horizontal line merging with an unperturbed local state. With a
further increase in frequency, a flat branch emerges from the Γ point, which
merges in turn with the continuation of the light cone. As a result a bandgap
arises between the high- and low-frequency branches, which is associated
with a local resonance supported by the meta-atom. We notice the differ-
ence: the spectral position of the Bragg gap is related to the shape of the
Brillouin zone and therefore varies depending on the propagation direction
in the reciprocal space, whereas the position of the band gap associated with
a local resonance does not have dependence on k. It means that in a
metamaterial the resonant frequencies correspond to the weak spatial
dispersion and the structure can be described by effective material parameters

ε and μ, in contrast to the photonic crystals.
Fig. 1B shows the polariton-like feature below the Bragg gap. We
remind that the wavelength of Bragg scattering λB is governed by the Bragg
law λB ¼ 2deffcos(α), where deff is an effective lattice spacing and α is the
angle between the wave vector and the normal to the crystal planes. There-
fore, there exists the lowest frequency of the Bragg gap, while the high-
frequency limit is infinite. It means that in general a range of the weak spatial
dispersion is below the lowest Bragg gap. Thus, the criterion for the
metamaterial phase is the appearance of the polariton-like feature below
the lowest Bragg gap.
We illustrate several problems of dielectric metamaterials by considering
one-dimensional structures. Let the first structure consists of a sequence of
layers A and B. To achieve resonances on the layers A we set εA ≫ εB leading
pffiffiffiffiffi
to strong Fabry–Perot resonances at ωFP ¼ π= wA εA , where wA is the
width of layer A. It is straightforward to demonstrate that the Fabry–Perot
frequencies are always higher than the lowest Bragg gap centered at
pffiffiffiffiffi pffiffiffiffiffi
ωB ¼ π= wA εA + wB εB . Another possibility to obtain resonance
response is to use quantum wells as the layer A with the exciton energy (fre-
quency) almost independent on the layer width. An analysis of this system
reveals that the Bragg gap width is increasing dramatically when the fre-
quency of exciton resonance is similar to the Bragg frequency (Ivchenko,
2005). Thus, the difference between the Bragg frequency and the frequency
of an isolated meta-atom should be large enough to yield a metamaterial
with the prototypical polariton-like feature below the lowest Bragg gap.
Let us point out another possibility of achieving the weak spatial disper-
sion above the lowest Bragg gap. Changes in the structure parameters might
lead to an accidental vanishing of spatial dispersion resulting in a photonic
crystal with an effective left-handed properties (Decoopman et al., 2006)
or a photonic crystal with an effective near-zero refraction index (Huang
et al., 2011). Although the accidental vanishing is of some interest, in this
chapter we will focus on the regular mechanisms resulting to the weak spatial
dispersion around the structure resonances, i.e., dielectric metamaterials.
Here we use the following criterion to distinguish a phase of photonic
structure, based on the behavior of the second dispersion branch, which
determines the type of the lowest bandgap: (i) in the phase of photonic crys-
tal, the second branch has a minimum frequency at the boundary of the
Brillouin zone (see Fig. 1A) and (ii) in the phase of metamaterial, the second
branch has a minimum frequency at the center of the Brillouin zone (the Γ
point) forming the polariton-like feature (see Fig. 1B).
3. Dielectric rods arranged in a square lattice

In this section, we describe the appearance of metamaterial phase in a
square lattice of dielectric rods surrounded by air (Rybin et al., 2015). The
rods have a circular profile of radius r and permittivity ε and the lattice con-
stant is a. These three parameters completely determine the photonic struc-
ture. However, the scalability of the Maxwell’s equations allows us to reduce
it to a two parameter set by measuring all distances in the lattice constant a,
i.e., we will use a normalized radius r/a and a frequency a/λ, where λ is the
vacuum wavelength. We assume that wave vectors are orthogonal to the
rod axis, which makes it possible to separate polarization as TE (transverse
electric field relative to the rod axis) and TM (transverse magnetic field). The
dielectric rods are known to support electric Mie resonances for the TM
polarized incident waves and magnetic Mie resonances for the TE polariza-
tion (Bohren and Huffman, 1998). First, we consider TE waves because
magnetic Mie resonances allow metamaterials with artificial magnetic
permeability, which can takes negative values μ < 0.
We have discussed above that the Fabry–Perot resonances in structural
elements cannot form the fundamental gap (i.e., the bandgap with the lowest
energy in spectrum) in a layered structure. However it is possible for the
structures of higher dimensions, for example a periodic system composed
of parallel rods. Here we consider two scenarios for the bandgap related
to the Mie resonance in rod to become the fundamental gap (Fig. 2).
The first one is in reducing the lattice spacing a that increases the Bragg
frequency keeping the Mie resonance almost unchanged. At a certain value
r/a the order of the Mie and Bragg gaps swaps and the metamaterial phase
appears. The second scenario is in increasing the rod permittivity to reduce
the Mie resonance frequency, while the lattice geometry (that defines the
Bragg bandgaps) are kept. It also might result in the metamaterial phase.
Now, we follow the second scenario to uncover the physics of the
metamaterial phase appearance. Three different datasets are computed:
(i) Mie scattering spectra on an isolated dielectric rod, (ii) photonic band
diagram of a perfect square lattice consisting of homogeneous rods, and
(iii) transmission spectra of a structure consisting of 10 layers of rods along
the wave incidence direction and being periodic in the orthogonal direction.
First, let us consider the Mie scattering on an isolated rod. Following the
Mie theory, the scattering waves are expanded into the multipole cylindrical
harmonics H(1) n (kr) exp(inθ) in the far field radiation zone. Here Hn is a
(1)
Hankel function of order n, and r and θ are the polar coordinates. The har-
monic amplitudes are determined by the resonant cylindrical Lorenz–Mie
r /a-scenario -scenario
a′ (a) (b) a
(c)
a
decreasing lattice constant a ′ < a increasing rod permittivity

Metamaterial Photonic crystal Metamaterial
Fig. 2 Transition from photonic crystals to metamaterials. Schematic illustrates two pos-
sible transformations of the PhC (B) to MM (A), (C) through the competition between the
Bragg and Mie scattering wavelengths. To satisfy the classical homogenization condi-
tion in the low-frequency long-wavelength limit (λBragg < λMie), one may decrease the
lattice constant and increase the filling ratio r/a (r/a-scenario) or, alternatively, increase
the rod permittivity ε the so-called ε-scenario. Reproduced from the original paper by
Rybin, M.V., Filonov, D.S., Samusev, K.B., Belov, P.A., Kivshar, Y.S., Limonov, M.F., 2015. Phase
diagram for the transition from photonic crystals to dielectric metamaterials. Nat.
Commun. 6, 10102 under a Creative Commons License (http://creativecommons.org/
licenses/by/4.0/).
coefficients an and bn for the TE and the TM polarizations, respectively (see,

e.g., Bohren and Huffman, 1998). Fig. 3 shows the calculated Mie scattering
efficiency spectra for the dipole cylindrical harmonic with resonances TE01,
TE02 and higher multipoles with resonances TEnm (n defines the angular
behavior and m enumerates resonances with the same n). The spectra reveal
a strong decrease of the dimensionless frequencies a/λ and the narrowing of
the Mie bands TEnm with increasing ε.
The photonic band diagrams shown in Fig. 3 are calculated numerically by
exploiting a plane wave expansion method ( Johnson and Joannopoulos,
2001). A large number of plane waves in the expansion (128 128) enable
a high precision of eigenfrequencies in the photonic diagrams for both pho-
tonic crystal and metamaterial phases. We remind that the Brillouin zone of a
square lattice has a square shape as well and has three points of high symmetry:
Γ at k ¼ 0, X at k ¼ (π/a;0), M at k ¼ (π/a; π/a), which restrict an irreducible
part of the Brillouin zone. Since the lowest frequency of the Bragg resonance
in the square lattice corresponds to the ΓX we consider band diagrams for this
direction only. For convenience, Fig. 3 shows extended band diagrams for the
wave vector in the ΓXΓ interval with clear extrema of the dispersion
branches.
The transmission spectra in Fig. 3 are calculated by using the CST
Microwave Studio software for the wave vector of the incident beam parallel
to the ΓX direction. We find that transmission spectra of structure with the
a e=4 b e=9
0.8 TE11
TE01
0.6
TE01
0.4
Bragg Bragg
0.2
0.0
c e = 12.8 d e = 19.5
0.8 TE02
TE11
Frequency a /λ
TE21
0.6
Mie + Bragg
TE11
0.4 Mie + Bragg
TE01
TE01
0.2
0.0
e e = 25 f e = 40
0.8
TE02 TE12
TE31
0.6 TE21 TE02
TE21
0.4 TE11
TE11
TE01 Mie
0.2 TE01 Mie
0.0
0 1 2 3 4 0 1 0 1 2 3 4 X 0 1
Field intensity Wavevector Transmission Field intensity Wavevector Transmission
Fig. 3 (Left-hand sections in all panels) Calculated Mie scattering efficiency for an iso-
lated dielectric circular rod for TEn (n 0) modes. (Central sections) The band structure
for 2D square lattice of rods with r ¼ 0.25a in air for the TE polarization. The band struc-
ture is shown between the Γ point (wave vector k ¼ 0) and the X point (j k j ¼ π/a along
the x direction). (Right-hand panels) The transmittance calculated for 10 lattice layers of
the 2D square structure of rods in air for the TE polarization. The frequency and wave
vector are plotted in dimensionless units a/λ, where λ is vacuum wavelength and a
denotes the lattice constant. (A) ε ¼ 4, (B) ε ¼ 9, (C) ε ¼ 12.8, (D) ε ¼ 19.5, (E) ε ¼ 25,
and (F) ε ¼ 40. Reproduced from Rybin, M.V., Filonov, D.S., Samusev, K.B., Belov, P.A.,
Kivshar, Y.S., Limonov, M.F., 2015. Phase diagram for the transition from photonic crystals
to dielectric metamaterials. Nat. Commun. 6, 10102 under a Creative Commons License
(http://creativecommons.org/licenses/by/4.0/).
thickness of 10 rod layers demonstrate an intensive dip for both photonic

crystal and metamaterial phases.
The three datasets calculated by independent methods make it possible to
analyze an evolution of low-frequency bands of the photonic band diagram
being a result of interference between the local Mie resonances and the
Bragg resonances extended over the structure. When the permittivity is 4,
the Mie resonances are at the higher frequencies and the fundamental gap is
related to the Bragg resonance (compare Fig. 3A with Fig. 1A). With the
increase of ε the Bragg gap is affected by the Mie resonance TE01.
Fig. 3B demonstrates that the fundamental gap shifts down slightly, and both
the center and the low-frequency wing of Mie peak do not correspond to
any band gap. At ε ¼ 12.8 the fundamental gap degenerates (Fig. 3C) and
opens again until ε reaches 19.5 when the second band degenerates into a
horizontal line (Fig. 3D). For greater ε > 19.5, the band structure transforms
from the photonic crystal type to the metamaterial type with a pronounced
polariton-like feature (Figs. 1 and 3E and F). We notice that the bandgap in
the polariton-like feature matches the Mie resonance of isolated rod, which
means that it is a Mie gap. Thus, the square lattice of the dielectric rods trans-
forms from the photonic crystal phase to the metamaterial phase at critical
values ε ¼ 19.5 and r ¼ 0.25a.
The picture of metamaterial phase appearance in the photonic structure
becomes clear with an analysis of bandgap map. We plot in Fig. 4 the depen-
dence of the frequency gap for the ΓX direction as a function of the rod per-
mittivity. Also, we add the spectral frequencies of low-index Mie resonances
0.6 TE12
TE21 TE02
TE31
TE01
0.4
TE11
a/λ
TE0k
TE1k
Mie
0.2 TE2k
TE3k
Bragg
Photonic
crystal Metamaterial
(a) TE (b) TM (c)
0.0
10 20 30 40 50 60 G XG X
e Wave vector
Fig. 4 Bandgap diagram for the 2D square lattice of circular rods. (A) The bandgap dia-
gram obtained from the photonic band structures calculated for 1 ε 60 with steps of
Δε ¼ 1. The rods are embedded in air, r/a ¼ 0.25, and TE polarization is considered. The
Bragg and Mie gaps obtained from the band structure calculations are marked by dark
green. (B and C) The photonic band structure for the 2D square lattice of rods with
ε ¼ 60 for TE and TM polarization, respectively, for the ΓX scan of the wave vector k.
Reproduced from Rybin, M.V., Filonov, D.S., Samusev, K.B., Belov, P.A., Kivshar, Y.S.,
Limonov, M.F., 2015. Phase diagram for the transition from photonic crystals to dielectric
metamaterials. Nat. Commun. 6, 10102 under a Creative Commons License (http://
creativecommons.org/licenses/by/4.0/).
TEnm to the map. The Bragg gap appears at a/λ ¼ 0.5 when the rod permit-
tivity is low. With the increase of the permittivity the TEnm Mie resonances
shift down to a/λ ¼ 0.5, and the Bragg gap becomes broad similar to the case
of the resonant photonic crystals (Ivchenko, 2005). The interference
between Mie and broaden Bragg resonances results in a nontrivial picture
of alternating allowed bands and band gaps. However at ε ¼ 19.5 the
TE01 Mie gap splits off from the broaden Bragg gap and for the higher per-
mittivities the fundamental (lowest) band gap is related to the Mie resonance
on the structure elements only.
Recently, Maslova et al. (2018) found conditions for metamaterials with
electric response operating for the TM polarization. The TM polarization
is known to demonstrate stronger effects with respect to the case of TE
polarization. For this reason the effect of the resonant Bragg gap broadening
is stronger, which disables metamaterial regime for the structure with and
electric metamaterials with r ¼ 0.25a (see Fig. 4C). However, sparse struc-
tures make is possible to observe metamaterials with electric response
(Fig. 5). When the rod permittivity is ε ¼ 30 that corresponds to Ge2Sb2Te5
0.7
0.6
0.5
0.4
a/l
0.3
0.2
0.1
(a) (b) (c)
0.0
X M X M –4 0 4
Wave vector e eff
Fig. 5 Band diagrams of dielectric rods arranged in a square lattice for TM polarization.
(A) Photonic crystal with ε ¼ 12 and r/a ¼ 0.1. (B) Metamaterial with ε ¼ 30 and r/a ¼ 0.1.
The second branches in the diagrams are marked by thick lines. Green circles correspond
to the bottom boundary of the second band, the red dotted lines mark frequencies of the
maximum field enhancement at a/λ ¼ 0.518 and a/λ ¼ 0.472 in a structure of 5 by 250
rods. (C) Evaluated effective permittivity εeff of the metamaterial with ε ¼ 30 and
r/a ¼ 0.1. Gray shading in (B and C) corresponds to the Mie gap. Reproduced with permis-
sion from Maslova, E.E., Limonov, M.F., Rybin, M.V., 2018. Dielectric metamaterials with
electric response. Opt. Lett. 43, 5516–5519. Copyright (2018) OSA.
compound in infrared (Wuttig et al., 2017) the band diagram demonstrates

the polariton-like feature.
Metamaterials composed of sparse high-index dielectric rods allow a
homogenization procedure by using the approach proposed by O’Brien
and Pendry (2002). An effective dielectric index is evaluated as εeff ¼ hDi/
hε0Ei, where the average of the displacement field D is taken over the
2D unit-cell area and the average of the electric field E is taken along a line
at the unit-cell boundary parallel to the rod. The small radius of dielectric
rod allows us to perform the averaging over a circle instead of the square
unit-cell by setting the circle area to match the area of the unit-cell. This
approximation makes it possible to calculate the effective dielectric index
analytically by the following formula
Z r Z R
pffiffiffiffiffiffiffi ð1Þ
εrod d0 ðkρÞJ0 ð εrod kρÞdρ + J0 ðkρÞ + b0 ðkρÞH0 ðkρÞ dρ
2π r
εeff ¼ 2 0 ,
a ð1Þ
J0 ðkRÞ + b0 ðkRÞH0 ðkRÞ
where k is the wave number, b0 and d0 are the cylindrical Lorenz–Mie

coefficients of the electric field outside and inside the rod, respectively, for
pffiffiffi
the TM polarization (e.g., see Bohren and Huffman, 1998), and R ¼ a= π
is the radius of the averaging circle. We plot the effective dielectric index
as a function of frequency in Fig. 5C. The interval of the negative εeff
(0.289 < a/λ < 0.471) is in an excellent agreement with the photonic band
gap 0.282 < a/λ < 0.471 confirming the metamaterial nature of this bandgap.
4. Experimental observation of the photonic phase

transitions
The aim of the experimental investigations (Rybin et al., 2015) was to
confirm the two scenarios of transitions in dielectric structure from photonic
crystal to metamaterial shown schematically in Fig. 2. Experiments for the
dielectric metamaterials with magnetic response were carried out with the
sample designed for the microwave range. This range was chosen because
the dielectric index of the rod should be at least 20 for metamaterials, and
a sample with a variable lattice constant is easier to create on a centimeter
scale. For experiments, it is also useful that distilled water, being an ideal
dielectric, has a strong temperature dependence of the permittivity from
ε ¼ 80 at 20°C to ε ¼ 50 at 90°C (e.g., see Andryieuski et al., 2015) in
the microwave range (1–6 GHz).
The sample called a metacrystal is shown in Fig. 6 composed of 5 by

10 series-connected tubes with outer radius of 20 mm, internal radius of
17 mm, length of 1000 mm and a negligible dielectric index. The tubes were
arranged in a square lattice by adjustable holders which make it possible to
modify the lattice spacing in the interval 0.65 < r/a < 1.9. All tubes were
connected in a closed series circuit together with a thermostat with heaters
and a water pump that support homogeneous temperature of the water
inside all tubes of the metacrystal (the difference between the inlet and outlet
temperature did not exceed 1°C).
The transmission spectra of the metacrystal in the interval 1–3 GHz were
measured in an anechoic chamber. An approximation to the TE-polarized
incident plane wave was generated by a rectangular horn antenna (TRIM
0.75–18 GHz; DR) connected to an output port of the vector network
analyzer Agilent E8362C. Another similar antenna was connected to an
A B
y 2r
a
TE z a
H
y
E
x
x
z
C
Reciever
E
H k
Y
Transmitter Z
X
Fig. 6 Experimental setup. (A) An anechoic chamber with a metacrystal composed of

10 tubes in length (x-axis) and 5 tubes in width (y-axis), 50 tubes in total. Each tube
has a length of 1 m (along z-axis), an outer radius of 2 cm and an inner radius of
1.7 cm. All tubes are connected in a closed series circuit together with a water heater
and a temperature stabilizer. (B) Schematic illustration of the transformation of the 2D
square lattice composed of dielectric rods as the lattice constant a increase. (C) The
scheme shows the square lattice from above. The rods with dielectric permittivity ε
and fixed radius r are embedded in air. The material is homogeneous along the z direc-
tion and periodic along x and y. Panels (A) and (B) reproduced from Rybin, M.V., Filonov,
D.S., Samusev, K.B., Belov, P.A., Kivshar, Y.S., Limonov, M.F., 2015. Phase diagram for the
transition from photonic crystals to dielectric metamaterials. Nat. Commun. 6, 10102 under
a Creative Commons License (http://creativecommons.org/licenses/by/4.0/).
input port of the analyzer. The spectra were normalized to the transmission
of the free space.
Fig. 7 demonstrates the experimental and theoretical spectra of the
metacrystal as a function of the sample volume related to the r/a parameter
as well as its temperature, which determines the rod permittivity. During the
measurements for the r/a-scenario the temperature of water was stabilized
at 90°C while the lattice spacing of the square lattice was reduced from
a ¼ 212 mm (r/a ¼ 0.08) to 89 mm (r/a ¼ 0.19). The transmission dip
corresponding to the lowest Bragg gap shifts to the higher frequencies
and almost disappears at r/a 0.09 when it reaches to the frequency of
the lowest Mie dipole resonance TE01. For the shorter lattice spacing the
Bragg dip appears again at r/a 0.095, however at the higher frequency side
of the Mie resonance. As a result, the lowest Bragg band and the TE01 Mie
Fig. 7 Transmission spectra for a square lattice of circular rods. TE polarization. (A–C)
r/a-scenario, ε ¼ 62. The calculated (B) and experimentally measured (C) transmission
spectra of the metacrystal as a function of the filling ratio r/a. The calculated spectra
are shown in the range 0.05 r/a 0.21, the experimental spectra are shown in the
range 0.08 r/a 0.19. (D–F) ε-scenario, r/a ¼ 0.08. The calculated (E) and experimen-
tally measured (F) transmission spectra of the metacrystal as a function of the dielectric
permittivity. The calculated spectra are shown in the range 30 ε 94, the experimen-
tal spectra are shown in the range 62 ε 94. The spectra are shifted vertically by the
constant value. Reproduced from Rybin, M.V., Filonov, D.S., Samusev, K.B., Belov, P.A.,
Kivshar, Y.S., Limonov, M.F., 2015. Phase diagram for the transition from photonic crystals
to dielectric metamaterials. Nat. Commun. 6, 10102 under a Creative Commons License
(http://creativecommons.org/licenses/by/4.0/).
band change their positions in the frequency scale indicating the transition
into the metamaterial phase.
The ε-scenario for the metacrystal with r/a ¼ 0.08 was realized by
varying of the temperature of water leading to modifying the rod permittiv-
ity. Increasing of the dielectric permittivity results in the spectral shift of the
Mie resonances, and the Bragg dip almost keeps its position. Fig. 7F shows a
significant proximity of the Mie and Bragg resonances, although their inter-
section was not achieved under the experimental conditions. In theory the
Mie resonance becomes the fundamental gap and the metacrystal turns into
the metamaterial regime at r/a ¼ 0.08 and the rod permittivity about
90 (Fig. 7E).
Let us discuss this transition of the metacrystal sample. We remind that
the fundamental Mie gap is associated with the negative effective magnetic
permeability μeff < 0 (e.g., see O’Brien and Pendry, 2002) and strong special
dispersion above the lowest Bragg gap disables μeff for the photonic crystal
regime. Thus, the continuous change in temperature or sample volume leads
to the appearance of the effective magnetic permeability deviating from
unity at a certain frequency interval. At the same time thermodynamic phase
transitions are known to be associated with dramatic changes in a generalized
susceptibility (e.g., Landau et al., 1980). Hence, the observed transition in
metacrystal (appearance of the nonzero magnetic susceptibility) owning
to the change in temperature or volume can be treated as a phase transition.
However, the changes of the thermodynamic quantities (temperature and
volume) in such phase transitions manifesting in the photonic response
are just means rather than physical causes. The physical reason for the pho-
tonic transition is the change of the water permittivity or the lattice spacing.
Therefore, the observed transitions are not thermodynamic phase transitions
but they are transitions of a new kind that should be referred to as photonic
phase transitions.
5. Phase diagram of periodic dielectric photonic

structures
Since the transformation between the regimes of photonic crystals and
metamaterial is considered as the photonic phase transition, we have to
develop corresponding phase diagrams. From Fig. 4 is clearly seen that the
TE01 Mie gap splits from the Bragg stop-band at ε ¼ 19.5 for r/a ¼ 0.25.
For ε > 19.5 the band structure has the typical polariton-like dispersion at
the low frequencies. Similar maps of the bandgaps vs rod permittivity plotted
for a variety of values of r/a reveal the boundary of the metamaterial phase that
is shown in Fig. 8 presenting the phase diagram as a function of rod permit-
tivity and the radius normalized by the square lattice constant for the case of
TE polarization. The phase diagram unveils the structure parameters required
for the metamaterial regime. The left side of the phase boundary corresponds
to the r/a- and ε-scenarios described earlier in Fig. 2. However, there exists a
competing process resulting in the right side of the boundary. This process is
related to the interaction between the neighboring rods and collective modes
driving the strong spatial dispersion typical to the photonic crystals.
We also construct the phase diagram for the metamaterials with electric
response in TM polarization. The metamaterial phase appears as a narrow
area at the left of Fig. 8 corresponding to sparse structures. The reason is
in the stronger electric multipole–multipole interaction relative to their
magnetic counterparts. Also the minimum rod permittivity for the electric
metamaterials with a square lattice is ε ¼ 27. We notice that the
metamaterials with magnetic response are of interest due to their effective
magnetic permeability μeff 6¼ 1, however a variety of applications demand
control of the electric response as well as the low-loss dielectric
metamaterials with engineered effective permittivity εeff can significantly
improve the light-matter interaction.
100
Electric met
Rod permittivity
80
Magnetic metamaterials
amaterials
60
40
20
Photonic crystals
0.0 0.1 0.2 0.3 0.4 0.5
r/a
Fig. 8 TM + TE photonic phase diagrams of dielectric rods arranged in a square lattice.
The region of the electric metamaterials in the TM polarization is marked by blue. The
region of the magnetic metamaterials in the TE polarization is marked by red. Circles are
phase boundary obtained from the band diagram analysis. Solid curves are guides for
eyes only. Reproduced with permission from Maslova, E.E., Limonov, M.F., Rybin, M.V.,
2018. Dielectric metamaterials with electric response. Opt. Lett. 43, 5516–5519. Copyright
(2018) OSA.
Following the intuition one might think that transitions between the
photonic crystal and metamaterial regimes are gradual. For example, let
us consider the r/a parameter continuously decreasing from a value
corresponding to the metamaterial phase (see Fig. 8) through the boundary
to a photonic crystal value. The common believe is that the “quantity” of
metamaterial is becoming weaker until it disappears completely. However
it is not correct description of the transitions. In reality, the field distribution
(photonic subsystem) undergoes a dramatic transformation when the struc-
ture parameters (ε and r/a) cross the phase boundary. Fig. 9 illustrates the
field distribution under the photonic phase transition. We examine the filed
profiles in two prism-shaped structures consisting of dielectric rods (ε ¼ 25)
arranged in a square lattice. The systems are illuminated by a Gaussian beam
directed along the Γ-X (the first prism) and the Γ-M (the second prism) at
the lowest frequency of the second dispersion band. We are decreasing the
volume of samples by varying the lattice spacing from r/a ¼ 0.18 (photonic
crystal according to the phase diagram in Fig. 8) to r/a ¼ 0.20 (metamaterial).
Photonic crystal Metamaterial
(a) (b) (c) (d)

r = 0.180a r = 0.189a r = 0.190a r = 0.200a
(e) (f) (g) (h)

r = 0.180a r = 0.189a r = 0.190a r = 0.200a
Fig. 9 A dramatic change in the magnetic field pattern depending on the operational
photonic crystal or metamaterial regimes (or the corresponding phase). Simulated field
patterns for the lowest frequency of the second band vs nanorod radius. An incident
Gaussian beam propagates along the ΓX direction (A)–(D) or along the ΓM direction
(E)–(H). The structure boundary is marked by a gray dashed line; arrows show the inci-
dent direction, nanorod permittivity ε ¼ 25. Simulations are performed within the mul-
tiple scattering theory approach. Reproduced from Li, S.V., Kivshar, Y.S., Rybin, M.V., 2018.
Toward silicon-based metamaterials. ACS Photonics 5, 4751–4757 with permission, ACS.
The prisms in the photonic crystal regime demonstrate a striped pattern

being typical for the Bragg scattering (the neighbor crystalline planes are
in anti-phase) for the Γ-X propagation and evanescent wave along the
Γ-M (see Fig. 9A and E). In contrast, the metamaterial phase corresponds
to almost homogeneous patterns without dependence on the lattice orienta-
tion because the structure can be described by a μ-near zero effective param-
eter (Fig. 9D and H). We notice that the stripped-type pattern is observed for
any parameters corresponding to the photonic crystal regime in the phase dia-
gram (Fig. 8), and the homogeneous pattern corresponds to any parameters of
metamaterial phase. The pattern is changed right on the phase boundary
revealing a sharp transition between the manifestations of these two regimes.
The transition takes place at least when the r/a parameter variation is as small as
0.001 (Fig. 9B, C and F, G).
For a deeper insight into the physics of the photonic phase transition we
also analyze isofrequency contours for the second dispersion band that plays a
crucial role. Fig. 10A demonstrates that in the photonic crystal phase with
parameters near the phase boundary (r/a ¼ 0.189, ε ¼ 25) there exist modes
with almost a circular dispersion around the Γ point (e.g., at a/λ ¼ 0.4242)
due to the Mie resonances of dielectric rods. For higher frequencies the
dispersionless contours around the Γ point transforms into a four-pointed
star (a/λ ¼ 0.4243) and the effects of spatial dispersion becomes strong.
A B
Fig. 10 A dramatic change in isofrequency contours depending on the operational

photonic crystal (A) or metamaterial (B) regimes (or the corresponding phase). The iso-
frequency contours correspond to the second dispersion band. Structure parameters
correspond (A) r/a ¼ 0.189, ε ¼ 25 and (B) r/a ¼ 0.190, ε ¼ 25 to structures in Fig. 9B, F,
C, and G.
However, event at a/λ ¼ 0.4242 the modes of another type located around
the X points coexist to the dimensionless modes around the Γ point that
indicates the strong special dispersion regime as well. We notice that the
modes around the X points on the Brillouin zone surface appear due to
the Bragg scattering on the periodic structure.
Now we consider the metamaterial with (r/a ¼ 0.190, ε ¼ 25) which is
described as only one isofrequency contour per frequency. Because the struc-
ture parameters are near the phase boundary most frequencies correspond to
the four-pointed star revealing the strong spatial dispersion. However for the
low-frequency interval below the a/λ ¼ 0.41033 the counters become circles
and the structure has the frequency interval with the weak spatial dispersion
indicating the metamaterial phase. Now we can describe the transition
through the boundary. In the isofrequency representation the Bragg reso-
nances manifest themselves as contours around points on the Brillouin zone
surface, whereas the Mie resonances exhibit circles around the Γ point similar
to the light cone (dispersion of free photons in homogeneous media). At some
frequencies the contours of Bragg modes are superimposed with the dis-
persionless Mie contours resulting in more the one mode for certain directions
of the wave vector. This regime is described by the strong spatial dispersion
and we consider it as photonic crystal regime because of the existence of the
Bragg scattering. When the frequency of Bragg resonance increases the con-
tours around the Brillouin zone surface disappears and the Mie contours
remain the exclusive resonant feature in the structure. Below some frequency
the contour becomes circular exhibiting the metamaterial properties of the
Mie resonant-driven dielectric structures.
Also we discuss an important consequence from the different nature of
the resonances defining the fundamental gap in photonic crystals and
metamaterials. In photonic crystals the Bragg scattering requires the period-
icity of the structure and the dielectric contrast is not a crucial factor. In con-
trast, for the metamaterial phase the Mie resonant modes play the decisive role,
while the periodicity is important only for the theoretical description of pho-
tonic properties by exploiting terms of the band theory. Therefore, a position
disorder has to influence on the photonic crystal Bragg bandgaps, unlike the
robust Mie bandgaps of the metamaterial regime. We analyze a transmission of
the TE-polarized plane wave through a photonic structure in different phases.
Fig. 11A shows the light incidence at a central frequency of the Bragg bandgap
(a/λ ¼ 0.3, ε ¼ 4, r/a ¼ 0.25) with the striped pattern along the exponential
decay that makes the sample opaque. The introduction of the position disor-
der breaks the periodicity, and the Bragg scattering disappears making the
Fig. 11 Transmission of the plane wave through periodic photonic structure with and
without position disorder. (A and B) Photonic crystal regime ε ¼ 4, a/λ ¼ 0.3. (C and D)
Metamaterial regime ε ¼ 25, a/λ ¼ 0.45. All structures r/a ¼ 0.25, TE polarization. The
plane wave is incident from the bottom side. Transmission spectra of photonic crystal
(E) and metamaterial (F). Spectra of prefect structures are shown by black solid curve.
Spectra of structures with position disorder are shown by red dashed curve.
structure to be transparent (Fig. 11B). The dielectric metamaterial (a/λ ¼ 0.45,

ε ¼ 25, r/a ¼ 0.25) operates due to the Mie resonances exhibited as intensive
spots inside the rods (Fig. 11C). Similar to the photonic crystal there is an
evanescent wave, which dissipates during propagation inside the structure.
However, the position disorder of rods has no effect on the propagation
length. Fig. 11D demonstrates the robustness of the Mie bandgap with the
introduction position disorder that does not make the metamaterial to be
transparent, unlike the photonic crystal.
To study this effect in details, we simulated transmission of the structure
shown in Fig. 11A–D. The spectra are shown in Fig. 11E and F. The log-
arithmic scale allows for the curve shapes do not depend on the number of
rods along the sample. For the case of photonic crystal without disorder the
spectrum exhibit symmetric Bragg gap (at a/λ 0.44) and a knife-blade like
dip corresponding to the TE01 Mie gap (at a/λ 0.47). The introduction of
position disorder leads to the degradation of features corresponding to both
gaps by several orders of magnitude. The spectra of metamaterials shown in
Fig. 11F reveal a rich structure of dips, which can be identified either as
Bragg gap or as a specific TEnk Mie gaps. Fig. 11F uncovers that after the
photonic phase transition to the metamaterial regime the fundamental
Mie gap (below all Bragg gaps) is not affected by position disorder in contrast
to other gaps Bragg and Mie both. This difference becomes another impor-
tant feature of photonic structure appearing with the phase transition.
6. Dielectric metamaterials with near-zero effective

parameters
First metamaterials were considered as artificial structures with the
left-hand behavior owning to the double negative effective ε and μ
(Shelby et al., 2001). Another application of metamaterials is in a realization
of media with engineered spatial distribution of ε and μ for an invisibility
cloaking designed by transformation optics (e.g., see Xu and Chen,
2015). Besides it, metamaterials with near-zero effective parameters ε and/or
μ have also attracted a lot of attentions (Liberal and Engheta, 2017). In par-
ticular medial with ε-near zero improves the light-matter interaction and
enhances nonlinear processes.
Now we discuss the possibility of dielectric ε-near zero metamaterials.
The phase diagram for the TM polarization shows that metamaterial regime
is allowed for the dielectric rods with ε > 27 (Fig. 8). Materials with such
permittivity are not available for the visible range, however at infrared
ε ¼ 30 can be achieved by using Ge-Sb-Te (GST) alloy (Wuttig et al.,
2017). The most studied compound Ge2Sb2Te5 has a crystalline and meta-
stable amorphous phases with ε ¼ 30 and ε ¼ 15, respectively.
Let us demonstrate the metamaterial behavior by existence of the near-
zero index regime with a homogeneous field distribution. Fig. 9 reveals that
the mode profile does not depend on the orientation of the crystalline lattice.
Now we analyze ε-near zero by changing the shape of a metamaterial and by
comparing its frequency with the frequency the second band at the Γ point
for the infinite structure (Fig. 5B). The structures are composed of about
1050 rods (it depends on specific shape and orientation) with the radius
r/a ¼ 0.1 having shape of a triangle, a square, and a circle. Fig. 12 demon-
strates homogeneous field distributions corresponding to the ε-near zero
regime. The distributions do not depend on the lattice orientation. How-
ever, the intensities are different is likely because of different boundary
conditions. The frequency of the ε-near zero mode a/λ ¼ 0.471 exactly
matches the lowest frequency of the second dispersion band in Fig. 5B.
(a) (c) (e)
(b) (d) (f)

Fig. 12 Distributions of the electric field for the ε-near zero modes in electric
metamaterials with different shapes [triangle (A and B), square (C and D), and circle
(E and F)] and different lattice orientations [ΓM (A, C, E) and ΓX (B, D, F) directions]. Green
dashed lines show the metamaterial structure boundaries. Black arrows indicate the
Gaussian beam incident direction. Insets in panels (A) and (B) show the electric field dis-
tribution on a larger scale. a/λ ¼ 0.471, ε ¼ 30, r/a ¼ 0.1, TM polarization. Reproduced with
permission from Maslova, E.E., Limonov, M.F., Rybin, M.V., 2018. Dielectric metamaterials
with electric response. Opt. Lett. 43, 5516–5519. Copyright (2018) OSA.
According to the phase diagram in Fig. 8, the dielectric metamaterials

with electric response exist for the sparse systems a > 8.3r, that is the most
of volume is occupied by air. Thus, the ε-near zero can be exploited to
enhance light-mater interaction for detecting atoms dissolved in the air by
means of an infrared absorbtion line analysis. The sample is illuminated by
a Gaussian beam and the modes in the near-zero regime tend to fill all available
sample volume because of the infinite wavelength in the metamaterial. Here
we use the following figure of merit I/I0, where I is the electric field intensity
integrated over the volume outside the rods, and I0 is the electric field intensity
of the free space integrated over the volume occupied by the sample.
As a reference we consider a photonic crystal having the same geometry
as a metamaterial sample (250 by 5 rods, r/a ¼ 0.1) but the rod permittivity is
12 (silicon in infrared). It is well-known effect of the electric field localiza-
tion in the region with lower dielectric index at the low frequencies of the
second dispersion band ( Joannopoulos et al., 2008). The structure ehibits a
well-recognized stripped pattern of the field distribution (Fig. 13A) and the
figure of merit demonstrates growth by almost twice (Fig. 13B) with the
maximum at a/λ ¼ 0.518. Now we study the effect of the electric field
enhancement by the metamaterial layer composed of the GST rods. In
A B C D
Fig. 13 Electric field localization in the plates being the photonic crystal with ε ¼ 12 and
r/a ¼ 0.10 or the metamaterial with ε ¼ 30 and r/a ¼ 0.10. (A) Distribution of the electric
field intensity in the photonic crystal plate. (B) The electric field intensity normalized to
the intensity in the free space without account of the field inside the cylinders for the
photonic crystal structure. (C) The electric field intensity normalized to the intensity in
the free space without account of the field inside the cylinders for the metamaterial
regime. (D) Distribution of the electric field intensity in the metamaterial plate. Green
dashed lines show the plate boundaries. Black arrows show the Gaussian beam incident
direction. Reproduced with permission from Maslova, E.E., Limonov, M.F., Rybin, M.V.,
2018. Dielectric metamaterials with electric response. Opt. Lett. 43, 5516–5519. Copyright
(2018) OSA.
the near-zero regime, the optical size of the sample is extremely small rel-
ative to the wavelength. As a result the Gaussian wave excites a mode
extended over the entire structure and the I/I0 ratio increases over
50 (Fig. 13C and D). The peak frequency a/λ ¼ 0.4717 differs by a small
value from the lowest frequency of the second dispersion band (in Fig. 5)
due to the finite size of the sample (it contains only 5 rods along the beam
incidence).
7. Toward practical realization of metamaterials in the

visible frequency range
In the previous sections, we have demonstrated that the metamaterial
phase in a photonic structure with a square lattice demands rod permittivity
at least 20 (for magnetic metamaterials in the TE polarization). There are a
lot of materials with required permittivity in infrared range, (Te, GeSbTe
alloys, etc.) but in the visible range the real part of permittivity growth with
its imaginary part. Also we did not optimize the crystalline lattice yet and
considered materials with negligible losses only. The aim of this section is
to discuss the approaches to yield the metamaterial phase in visible range. We

consider here silicon-based structures because of the advanced technologies
and the great experience of using silicon for photonics (Staude and Schilling,
2017). In crystalline silicon, electronic transitions between 1.3 and 3.4 eV
(365 nm < λ < 1130 nm) are indirect, i.e., they take place between states
of different wave vectors and only the simultaneous absorption or emission
of a photon allows for a weak absorption in that region. For the wavelength
shorter than 1000 nm the real part of silicon permittivity grows, while the
losses remain relatively weak. In experiments, Mie resonances in silicon
nanoparticles are observed almost in entire visible range λ > 465 nm
(Kuznetsov et al., 2016). The complex frequency-dependent dielectric
function of silicon does not allow plane wave expansion methods we have
used in the previous sections, since the method solves an eigenproblem for
square frequency ω2 and the permittivity-dependent operator ( Johnson and
Joannopoulos, 2001). Instead, here we exploit the inverse dispersion
method, which solves an eigenproblem for the wave numbers k acting as
complex eigenvalues (Rybin and Limonov, 2016).
Fig. 14 compares band diagrams of TE modes calculated by the plane
wave expansion method ω(k) and the inverse dispersion methods k(ω).
We consider two structures with the same geometry (square lattice,
r/a ¼ 0.3) but with different permittivity ε ¼ 4 (photonic crystal according
to the phase diagram in Fig. 8) and 30 (metamaterial). Fig. 14A and
B exhibits an excellent agreement between the data obtained by the different
methods since they solve the same mathematical problem. However the
inverse dispersion method finds additional modes with complex wave vec-
tors, including the evanescent waves corresponding to the bandgap frequen-
cies. It appears that the complex band diagrams uncover important
information when weak losses are taken into account. Fig. 14C and
D shows the diagrams for a dispersion of real part of wave vector as a function
of frequency for the structures with permittivity ε ¼ 4(1 + i0.01) and
30(1 +i0.01), respectively. The losses lift the degeneracy of wavevectors
(here and below, we imply the real part of wave vectors in the diagram)
for frequencies within the bandgap and show that all bands are connected
continuously and the second dispersion branch in not well-defined any-
more. Overall, the phase determination procedure based on the analysis
of the second branch is not applicable to the case of constituents with losses.
However, the complex band diagrams unveil another criterion for determi-
nation of the metamaterial phase: for the case of Bragg bandgap the modes
Photonic crystal Metamaterial

A G X G B G X G
1.2 0.5
Frequency a/l
Frequency a/l
1 0.4
Lossless
0.8 0.3
0.6
0.2
0.4
0.2 0.1
0 0
0 0.5 1 0 0.5 1
Wavevector 2pk/a Wavevector 2pk/a
C D
G X G G X G
0.5 0.25
Frequency a/l
Frequency a/l
0.24
With losses
0.48
0.23
0.46
0.22
0.44
0.21
0.42 0.2
0.475 0.5 0.525 0.4 0.5 0.6
Wavevector 2pk/a Wavevector 2pk/a
Fig. 14 Phase criteria for the lossless and lossy structures. (A and B) Band diagrams of a
square lattice of nanorods with r/a ¼ 0.3 and lossless permittivity ε ¼ 4 (the photonic crys-
tal phase) and 30 (the metamaterial phase). Data obtained by the direct method are
shown by black solid curves. Results of the inverse dispersion method are shown by cyan
circles. (C and D) Band diagrams calculated by the inverse dispersion method for the
similar structure as in (A and B) but with the nanorod permittivity having an additional
imaginary part ε00 ¼ 0.01ε0 (cyan solid curves). Black dashed curves show the results for loss-
less structures. Reproduced with permission from Li, S.V., Kivshar, Y.S., Rybin, M.V., 2018.
Toward silicon-based metamaterials. ACS Photonics 5, 4751–4757. Copyright (2018) ACS.
cross each other (Fig. 14C), while for the case of Mie bandgap the modes
exhibit avoided crossing (Fig. 14D).
Besides, we optimize the crystalline lattice taking into account two com-
peting conditions preventing the metamaterial phase discussed in Section 5:
the lowest Bragg frequency has to be higher than the Mie frequency and the
distance between the neighbor rods has to be large enough to preclude from
the collective modes formation. Having the densest packing, a hexagonal
lattice provides the smallest distance between the crystal planes among other
lattices, thus here we consider this type of lattice.
Fig. 15 compares photonic crystal–metamaterial phase diagrams of struc-
tures with the hexagonal and square lattices obtained by both considered
criteria. First we notice a significant decrease of the minimum permittivity
required for the metamaterial phase: the value ε ¼ 14 for the hexagonal lat-
tice is more promising for a practical realization than ε ¼ 19.5 for the square
lattice. Besides, we determine another phase boundary (shown by circles in
Fig. 15) by the crossing–anticrossing criterion for photonic sutures
with losses. For this case we take complex permittivity by the formula
ε ¼ ε0 (1 + i0.01) and the axis of ordinates shows ε0 (the real part of ε). The
phase diagrams reveal the excellent agreement of the two criteria. The small
difference for the denser structures (r/a > 0.35) can be explained as differ-
ent numbers of plane waves involved in calculations. The Hermitian eigen-
value problem solved in the ω(k) method allows expansion over 128 128
plane waves, whereas the inverse dispersion method k(ω) solves the non-
Hermitian eigenvalue problem that limits us by resolution of 25 25 plane
waves. Another reason for deviation between the phase boundaries evalu-
ated by two approaches is in the permittivity real-to-imaginary parts ratio
ε00 /ε0 . We find that, when the imaginary part of permittivity grows, an error
in the phase boundary is increasing. In particular, for the imaginary part of
dielectric permittivity described by ε00 /ε0 ¼ 0.35, this deviation is about 5%.
Now we have a complete tool for constructing the phase diagram of
periodic photonic structures, which constituents have a complex
frequency-dependent permittivity. The phase diagram in Fig. 15 shows that
50
Metamaterial
40
Photonic crystal
Photonic crystal
Dielectric permitivity
30
20
10
Hexagonal lattice Square lattice
0
0 0.1 0.2 0.3 0.4 0.5
r /a
Fig. 15 Phase diagram that allows differentiating metamaterials and photonic crystals.
The boundary is calculated for both hexagonal (red line) and square (black dashed line)
lattices, the circles mark the data obtained by the direct and inverse dispersion methods,
respectively. Green and red shadings correspond to the photonic crystal and
metamaterial regimes of a hexagonal lattice, respectively. Reproduced with permission
from Li, S.V., Kivshar, Y.S., Rybin, M.V., 2018. Toward silicon-based metamaterials. ACS Pho-
tonics 5, 4751–4757. Copyright (2018) ACS.
silicon-based metamaterials is possible for wavelength shorter than a certain

wavelength that corresponds to the critical value of the frequency-
dependent permittivity. By using the crossing/anticrossing criterion and
the dispersion of silicon permittivity reported by Green and Keevers
(1995), we determine the phase boundary for the hexagonal lattice of silicon
nanorods (Fig. 16). We notice that Maxwell’s equations are not scalable in
this case because of the frequency dependence of permittivity. However it
allows an unambiguous mapping of the permittivity to the wavelength,
which is used as ordinate axes instead. Since the normalized radius r/a does
not provide exhaustive information about the structure geometry, Fig. 16
shows additional diagrams for radius vs wavelength and lattice constant vs
wavelength, as well.
The silicon permittivity grows with shortening of the wavelength that
reverses the phase diagram relative to the horizontal (compare the diagrams
in Figs. 15 and 16A). Since the real part of the silicon permittivity growths up
to 42 the metamaterial phase is allowed even for the square lattice (dashed
curves in Fig. 16), however the strong absorption makes such metamaterials
impractical. In contrast, the design of silicon-bases structures with the hex-
agonal lattice allows to achieve the metamaterial phase for the wavelength
below 850 nm, where material losses are relatively weak. In practical reali-
zations, silicon metamaterials have to be an array of silicon nanopillars.
Advanced nanoimprint lithography, combined with reactive ion etching,
has been used to fabricate arrays of silicon nanopillars (see, e.g., a review
by Zhang et al., 2017). This method allows the fabrication of well-defined
A B C
900
Wavelength (nm)
800 Photonic crystal Photonic crystal

Photonic crystal
Metamaterial Metamaterial
700
Metamaterial
600
500
400
0.1 0.2 0.3 0.4 0.5 50 100 150 200 250 300 350 20 30 40 50 60 70 80 90 100
r/a Lattice constant (nm) Radius (nm)
Fig. 16 Phase diagram of a periodic silicon nanorod structure with a hexagonal lattice.
Red circles show the phase boundary, solid line is a guide for eyes only. Green and red
shadings correspond to the photonic crystal and metamaterial regimes, respectively.
The phase boundary for the square lattice is shown by a black dashed line. Phase dia-
gram is plotted in the difference axes: (A) ratio r/a vs wavelength; (B) lattice spacing a vs
wavelength; and (C) rod radius r vs wavelength. Reproduced with permission from Li, S.V.,
Kivshar, Y.S., Rybin, M.V., 2018. Toward silicon-based metamaterials. ACS Photonics 5,
4751–4757. Copyright (2018) ACS.
arrays with precise control in position, density. The reported diameter is as

small as 40 nm with the lattice spacing of 200 nm. Fig. 16 shows that those
sizes allow silicon-based metamaterials operating almost in entire visible
range from 460 to 850 nm. Besides, optimal etching parameters can lead
to uniform vertical nanowires with high aspect ratios 60:1, being sufficient
for experimental measurements.
Finally, we illustrate metamaterial behavior by considering refraction of a
TE-polarized Gaussian beam through a prism made of silicon-based
metamaterial. Here we make a number of assumptions: the Snell’s law is
applicable for description of the beam refraction at the prism surface; the
function of effective permeability μeff is unity at the low frequencies and
has a resonant feature around the Mie bandgap TE01, while the effective per-
mittivity εeff does not depend on frequency; and the effective refractive
pffiffiffiffiffiffiffipffiffiffiffiffiffiffi
index reads neff ¼ εeff μeff .
The prism has a shape of right triangle with an acute angle of 30 degree
consisting of 1220 silicon nanorods of radius 50 nm arranged in a hexagonal
lattice with period a ¼ 167 nm. The wide beam (waist is 10a) is incident on the
long side of prism and the transmitted beam is deflected at a certain angle
according to the Snell’s law. First, we evaluate the effective permittivity of
the metamaterial by simulating the beam with the wavelength of λ ¼ 1500 nm
that is far enough from the Mie bandgap TE01 (Fig. 17A and B). The trans-
mitted beam propagates at the 44 degree to the normal, which makes it
possible to evaluate the effective refractive index neff ¼ 1.39 and the effective
permittivity εeff ¼ 1.93 (we assume μeff ¼ 1).
The metamaterial prism limits the minimum and maximum values of the
permeability can be measured by the simulation of beam refraction. The
maximum is determined by the total internal reflection occurring for
neff > 2, and for the small values the transmission is suppressed by the strong
impedance mismatch (compare the μ-near zero regime with no transmitted
light shown in Fig. 9D). Because of the causality, the higher values of the
effective permeability correspond to the modes below the Mie bandgap.
In Fig. 17C and D, we show the beam refraction at the wavelength of
700 nm and evaluate the effective refractive index as neff ¼ 1.78, which yields
μeff ¼ 1.64 (we assume εeff ¼ 1.93). At the wavelength of 495 nm, the prism
deviates the Gaussian beam path 3 degree off the incident direction that gives
the effective refractive index neff ¼ 0.92, which yields μeff ¼ 0.44 both less
than one. Thus, the silicon-based photonic structure exhibits the typical
metamaterial behavior.
l = 1500 nm l = 700 nm l = 495 nm
A C
E
44° 63°
27°
B D
F
Fig. 17 Refraction of a Gaussian beam on a silicon-based metamaterial prism, for
a ¼ 167 nm and r ¼ 50 nm. The top panels are the magnetic field profiles (red are positive
values and blue are negative values of Hz); the bottom panels are the electromagnetic
field intensities. (A and B) Off-resonance case; λ ¼ 1500 nm. (C and D) the low-frequency
edge λ ¼ 700 nm; (E and F) the high-frequency edge λ ¼ 495 nm. Green dashed triangles
mark the prism. Black arrows in the top panels show the propagation direction of the
incident beam. The results are for the TE polarization. Reproduced with permission from
Li, S.V., Kivshar, Y.S., Rybin, M.V., 2018. Toward silicon-based metamaterials. ACS Photonics
5, 4751–4757. Copyright (2018) ACS.
8. Conclusion and outlook

Owing to their unique optical properties, metamaterials have been
intensively studied as promising novel optical materials enabling a control
over the electromagnetic waves. Although the dielectric metamaterials ben-
efit from their low-losses constituents, the conditions required for the struc-
ture to possess the effective parameters characterizing the metamaterial
properties have been understood recently. The approach based on the con-
cept of the photonic phase transitions shed light on the underlying physics
explaining how the metamaterial properties appear in periodic structures. In
this chapter, we have provided a comprehensive overview of the photonic
phase transitions in photonic structures between photonic crystal and
metamaterial regimes (or “phases”). We have discussed how the
corresponding phase diagrams can be constructed. For the lossless case,
the appearance of the metamaterial phase is identified by the polariton-like
feature formed around the bandgap between the first and second dispersion
branches in bandgap diagrams. However, for the realistic case of materials
with absorption, the first and second branches are merged into continuous
branches and the metamaterial phase can be distinguished by the criterion
based on crossing–anticrossing of the branches in complex bandgap dia-
grams. Although the considered photonic phase transitions are quite dissim-
ilar to conventional thermodynamic phase transitions, the phase transition
terminology is feasible because it describes abrupt changes in the photonic
patterns observed at the boundary between photonic crystal and
metamaterial regimes. Besides, more important is that the developed phase
diagrams as a function of structure parameters (permittivity and sizes)
become very useful in a design of metamaterials with desired properties.
Here we have reviewed several aspects of the metamaterial regimes includ-
ing near-zero parameters, stability against position disorder, and practical
realization of silicon-based metamaterials in the visible frequency range.
The study of photonic phase transitions has been started in the recent years,
so we anticipate the further development of this concept.
Acknowledgments
This work was supported by the Ministry of Education and Science of the Russian Federation
(Grant 3.1500.2017/4.6) and the Strategic Fund of the Australian National University.
References
Andryieuski, A., Kuznetsova, S.M., Zhukovsky, S.V., Kivshar, Y.S., Lavrinenko, A.V.,
2015. Water: promising opportunities for tunable all-dielectric electromagnetic
metamaterials. Sci. Rep. 5, 13535.
Bohren, C.F., Huffman, D.R., 1998. Absorption and Scattering of Light by Small Particles.
Wiley-VCH.
Decoopman, T., Tayeb, G., Enoch, S., Maystre, D., Gralak, B., 2006. Photonic crystal lens:
from negative refraction and negative index to negative permittivity and permeability.
Phys. Rev. Lett. 97, 073905.
Gorlach, M.A., Belov, P.A., 2014. Effect of spatial dispersion on the topological transition in
metamaterials. Phys. Rev. B 90 115136.
Green, M., Keevers, M., 1995. Optical properties of intrinsic silicon at 300 K. Prog. Photo-
volt. 3, 189–192.
Huang, X., Lai, Y., Hang, Z.H., Zheng, H., Chan, C.T., 2011. Dirac cones induced by acci-
dental degeneracy in photonic crystals and zero-refractive-index materials. Nat. Mater.
10, 582.
Ivchenko, E.L., 2005. Optical Spectroscopy of Semiconductor Nanostructures. Alpha
Science, Harrow, UK.
John, S., 1987. Strong localization of photons in certain disordered dielectric superlattices.
Phys. Rev. Lett. 58, 2486–2489.
Johnson, S.G., Joannopoulos, J.D., 2001. Block-iterative frequency domain methods for
Maxwell’s equations in a plane-wave basis. Opt. Express 8, 173.
Kaina, N., Lemoult, F., Fink, M., Lerosey, G., 2015. Negative refractive index and acoustic
superlens from multiple scattering in single negative metamaterials. Nature 525, 77.
Kuznetsov, A.I., Miroshnichenko, A.E., Brongersma, M.L., Kivshar, Y.S., Luk’yanchuk, B.,
2016. Optically resonant dielectric nanostructures. Science 354, aag2472.
Landau, L.D., Lifshitz, E.M., Pitaevskii, L.P., 1980. Statistical Physics, Part I. Pergamon
Press, Oxford.
Liberal, I., Engheta, N., 2017. Near-zero refractive index photonics. Nat. Photonics 11, 149.
Liu, Z., Lee, H., Xiong, Y., Sun, C., Zhang, X., 2007. Far-field optical hyperlens magnifying
sub-diffraction limited objects. Science 315, 1686.
Maslova, E.E., Limonov, M.F., Rybin, M.V., 2018. Dielectric metamaterials with electric
response. Opt. Lett. 43, 5516–5519.
Mirmoosa, M.S., Kosulnikov, S.Y., Simovski, C.R., 2015. Unbounded spatial spectrum of
propagating waves in a polaritonic wire medium. Phys. Rev. B 92, 075139.
O’Brien, S., Pendry, J.B., 2002. Photonic bandgap effects and magnetic activity in dielectric
composites. J. Phys. Condens. Matter 14, 4035.
Pendry, J.B., Holden, A.J., Robbins, D.J., Stewart, W.J., 1999. Magnetism from conductors
and enhanced nonlinear phenomena. IEEE Trans. Microwave Theory Technol.
47, 2075.
Pendry, J.B., Aubry, A., Smith, D.R., Maier, S.A., 2012. Transformation optics and sub-
wavelength control of light. Science 337, 549.
Reyes-Gómez, E., Cavalcanti, S.B., Oliveira, L.E., De Carvalho, C.A.A., 2014. Metric-
signature topological transitions in dispersive metamaterials. Phys. Rev. E 89, 033202.
Rybin, M.V., Limonov, M.F., 2016. Inverse dispersion method for calculation of complex
photonic band diagram and PT symmetry. Phys. Rev. B 93, 165132.
Rybin, M.V., Filonov, D.S., Samusev, K.B., Belov, P.A., Kivshar, Y.S., Limonov, M.F.,
2015. Phase diagram for the transition from photonic crystals to dielectric metamaterials.
Nat. Commun. 6, 10102.
Shelby, R.A., Smith, D.R., Schultz, S., 2001. Experimental verification of a negative index
of refraction. Science 292, 77–79.
Simovski, C.R., 2009. Material parameters of metamaterials. Opt. Spectrosc. 107, 726.
Simovski, C.R., 2018. Composite Media with Weak Spatial Dispersion. Pan Stanford.
Smith, D.R., Padilla, W.J., Vier, D.C., Nemat-Nasser, S.C., Schultz, S., 2000. Composite
medium with simultaneously negative permeability and permittivity. Phys. Rev. Lett.
84, 4184.
Staude, I., Schilling, J., 2017. Metamaterial-inspired silicon nanophotonics. Nat. Photonics
11, 274–284.
Veselago, V.G., 1968. The electrodynamics of substances with simultaneously negative
values of ε and μ. Sov. Phys. Usp. 10, 509.
Wuttig, M., Bhaskaran, H., Taubner, T., 2017. Phase-change materials for non-volatile pho-
tonic applications. Nat. Photonics 11, 465.
Xu, L., Chen, H., 2015. Conformal transformation optics. Nat. Photonics 9, 15.
Yablonovich, E., 1987. Inhibited spontaneous emission in solid-state physics and electronics.
Phys. Rev. Lett. 58, 2059–2062.
Yves, S., Fleury, R., Berthelot, T., Fink, M., Lemoult, F., Lerosey, G., 2017. Crystalline
metamaterials for topological properties at subwavelength scales. Nat. Commun. 8,
16023.
Zhang, B., Jie, J., Zhang, X., Ou, X., Zhang, X., 2017. Large-scale fabrication of silicon
nanowires for solar energy applications. ACS Appl. Mater. Interfaces 9, 34527–34543.
CHAPTER THREE
Light trapping in photonic

structures
Ken Xingze Wanga,*, Yu Guob, Zongfu Yuc,*
a
School of Physics, Huazhong University of Science and Technology, Wuhan, Hubei, China
b
Electrical Engineering and Ginzton Laboratory, Stanford University, Stanford, CA, United States
c
University of Wisconsin, Department of Electrical and Computer Engineering, Madison, WI, United States
*Corresponding authors: e-mail address: wxz@hust.edu.cn; zyu54@wisc.edu
Contents
1. Introduction 45
2. Ray optics light trapping theory 47
3. Wave optics light trapping theory 50
4. Fundamental limit of light trapping 54
4.1 Light trapping in structures with wavelength-scale periodicity 55
4.2 Light trapping in thin films 57
5. Light trapping in photonic crystals 61
5.1 Light trapping in 2D photonic crystals 62
5.2 Light trapping in 3D photonic crystals 73
6. Absorption enhancement in ultrathin crystalline silicon solar cells with
antireflection and light-trapping nanocone gratings 77
6.1 Motivation 77
6.2 Absorption enhancement at Yablonovitch limit 79
7. Conclusion 86
References 86
1. Introduction
Photon management strategies, including light trapping, have been
tremendously successful in efficiency improvement and cost reduction in
practical solar cells (Mokkapati and Catchpole, 2012). Light trapping allows
cells to absorb sunlight using an active material layer that is much thinner
than the material’s intrinsic absorption length. This effect then reduces
the amount of materials used in photovoltaic cells, which cuts cell cost in
general, and moreover facilitates mass production of photovoltaic cells that
are based on less abundant materials. In addition, light trapping can improve

46 Ken Xingze Wang et al.
cell efficiency, because thinner cells provide better collection of photogenerated

charge carriers, and potentially a higher open circuit voltage (Atwater and
Polman, 2010; Battaglia et al., 2012a; Bozzola et al., 2012; Callahan et al.,
2012; Munday, 2012; Mutitu et al., 2008; Narasimhan and Cui, 2013;
Polman and Atwater, 2012; Sandhu et al., 2013; Sheng et al., 2011; Wang
et al., 2012, 2013a,b; Wiesendanger et al., 2013; Wong and Morales, 2009;
Yu and Fan, 2011; Yu et al., 2010a,b, 2014).
It is known that the light trapping enhancement limit is related to the
photonic density of states (DOS) (Sheng et al., 1983). For a bulk silicon cell,
this leads to the well known 4n2 limit (Deckman et al., 1983; Yablonovitch,
1982; Yablonovitch and Cody, 1982), also referred to as the Yablonovitch
limit, the Lambertian limit, the ray-optic limit, or the conventional limit.
For thin films, the 4n2 limit can be modified depending on the details of
guided modes in the film (Agrawal and Peumans, 2008; Haug et al.,
2011; Munday et al., 2012; Stuart and Hall, 1997; Yu et al., 2010a). In addi-
tion, it has been shown that an elevated local DOS is an important element
for improved light trapping (Callahan et al., 2012; Tsakalakos et al., 2007;
Wang and Menon, 2014). That description is consistent with our wave
optics light trapping theory (Brongersma et al., 2014). The local DOS
describes the enhancement of absorption at a given spatial location of the
structure. To account for the performance of a solar cell, one needs to inte-
grate over all spatial points in the solar cell where absorption occurs. One
thus needs to perform spatial integration of the local DOS, which leads to
the DOS picture that we use in this chapter.
Given the importance of DOS in defining the light trapping limit, one
naturally should look beyond the relatively simple uniform absorbing layer
and consider more complex medium. In particular, the photonic crystal
structure is known to be able to drastically influence the DOS
(McPhedran et al., 2004), and its ability of dispersion and DOS engineering
has been extensively explored and applied to a wide range of applications
( Joannopoulos et al., 2008). However, while light trapping in an absorber
consisting of photonic crystals has been considered before (Chutinan and
John, 2008; Chutinan et al., 2009; Du et al., 2011; Duche et al., 2008,
2011; Eyderman et al., 2013; Gomard et al., 2012; Guldin et al., 2010;
John, 2012; Ko et al., 2009; Kuang et al., 2013; Le and John, 2014;
Mallick et al., 2010, 2012; Mihi and Mı́guez, 2005; Oskooi et al., 2014;
Park et al., 2009; Tumbleston et al., 2009a,b), the connection between
the DOS and the light trapping limit in a photonic crystal has not been
explicitly established.
Light trapping in photonic structures 47
In this chapter, we describe a statistical coupled-mode theory that

describes light trapping in general from a rigorous electromagnetic perspec-
tive (Yu et al., 2010a). We quantitatively establish the connection between
the absorption enhancement and the DOS (Wang et al., 2014a). Applying
this relationship, we show that the conventional limit of 4n2 is only correct
in bulk structures. In the nanophotonic regime, the absorption enhance-
ment factor can go far beyond this limit with proper design.
We further use the connection between the absorption enhancement
and the DOS to guide photonic crystal absorber designs. To maximize
the absorption by light trapping, one should aim at maximizing the number
of accessible modes (used interchangeably with states and resonances)
supported by an absorbing structure. Moreover, we show that to take full
advantage of the benefit of the photonic DOS, one need to consider addi-
tional issues including accessibility to these photonic states, as well as the
overlap of the photonic states with the absorber.
As an application of the light trapping theory, we show that the
Yablonovitch limit can be reached in ultrathin crystalline silicon (c-Si) solar
cells with nanocone gratings. The key to achieve the Yablonovitch limit is to
realize that antireflection and light trapping have drastically different struc-
tural requirements, and therefore should be optimized differently.
This chapter is organized as follows: In Sections 2 and 3, we briefly
review the ray-optic and the wave-optic theory of light trapping, respec-
tively, and derive a direct correspondence between the angle-integrated
absorption enhancement and the photonic DOS. In Section 4, we derive
the conventional 4n2 limit in bulk structures and demonstrate that in the
nanophotonic regime the absorption enhancement factor can go far beyond
this limit with proper design. In Section 5, we numerically study light trap-
ping in several two-dimensional (2D) photonic crystal absorbers, and
account for the numerical results using the theoretical understanding devel-
oped in Section 3, highlighting in particular the requirements to observe sig-
nificant DOS effects for light trapping in photonic crystals. We also briefly
discuss three-dimensional (3D) photonic crystals in Section 5. We apply the
light trapping theory to enhance absorption in ultrathin crystalline silicon
solar cells in Section 6 and conclude in Section 7.
2. Ray optics light trapping theory

The ray optics theory describes conventional light trapping structures
in the absence of wave effects. When the front surface is a Lambertian,
isotropic surface, the 3D absorption enhancement factor for light incident

from any angle is the same and is equal to 4n2 (Goetzberger, 1981;
Yablonovitch, 1982). This result is commonly referred to as the
Yablonovitch limit or the 4n2 limit. We refer all such ray optics limits as con-
ventional limits.
The uniform system in Fig. 1 consists of a uniform slab with a refractive
index n and an absorption coefficient α. The thickness of the uniform
slab, deff, is chosen such that the amount of dielectric materials per unit area
(in 3D) or unit length (in 2D) horizontally is the same as that of the photonic
crystal systems considered in Section 5. deff is therefore typically smaller than
the physical thickness, d, of the photonic crystal. They are related by an area
(in 2D) or volume (in 3D) fraction η
deff ¼ ηd: (1)
Except stated explicitly, the thickness of the photonic crystal is chosen or

assumed such that deff is at least equal to a few wavelengths.
We consider two different uniform systems that contain the same
amount of absorbing material. Fig. 1 is a slab where both the front and back
surfaces are flat, with no mirror at the back surface. We assume perfect ant-
ireflection for all frequencies and all angles on both surfaces. The absorption
of light incident normally to the top surface of such a slab is defined by the
single-pass absorption, αdeff, where α ¼ 2k0 Im {n} is the absorption coeffi-
cient of the material, and k0 is the wavevector of light in vaccum and Im{n}
is the imaginary part of the refractive index.
The goal of any light trapping structure is to achieve absorption greater
than the single-pass absorption, while keeping the same amount of the
absorbing material. For such light trapping structure, in 3D, we define
the absorption enhancement factor as
Aðω, θ, φÞ
f ðω, θ, φÞ ¼ : (2)
αdeff
Here, ω, θ, and φ are the frequency, the angle of incident light, and the
azimuthal angle of light, respectively. A is the absorption coefficient of the
Fig. 1 A dielectric slab with flat surfaces.

Fig. 2 A dielectric slab structure with a randomly textured front surface and a back mir-
ror. The equivalent thickness is the same as in Fig. 1.
structure. There will not be a φ dependency in 2D. The second uniform

system represents the conventional light-trapping structure (Fig. 2), where
one roughens the front surface of the uniform slab and places a perfect mirror
at the back.
Both the isotropic and the anisotropic cases above can be described in a
unified fashion by defining an angle-integrated enhancement factor
Z π=2 Z 2π
F3D ðωÞ ¼ dθ dφf ðω, θ, φÞ cosθ sin θ (3)
0 0
where f(ω, θ, φ) is defined in Eq. (3). For both the isotropic and anisotropic
cases, that is, f(ω, θ, φ) ¼ 4n2 and f ðω, θ, φÞ ¼ 4n2 = sin 2 θ0 θθ0 , respectively,
we have F3D ¼ 4πn2. Moreover, it has been shown that as an upper limit
F3D ¼ 4πn2 in fact applies to any light trapping structure of the form of
Fig. 2 with an arbitrary angular response of f(θ, φ) (Yu and Fan, 2011).
Since no actual physical light trapping structure has an ideal isotropic
angular response, to compare the performance of a physical light trapping
structure to the conventional limit, it is imperative that one performs angle
integration (Naqavi et al., 2014; Yu and Fan, 2011). Observing an enhance-
ment factor greater than 4n2 for a single angle of incidence, for example,
should not be taken as the evidence that one has overcome the conventional
limit (Ganapati et al., 2014). In this context, the 4πn2 limit, which is really
the conventional limit expressed in an angle-integrated fashion, is very useful
when one needs to compare the performance of a physical structure to the
ray optics limit.
The ray optics theory also describes the conventional light trapping
structure of Fig. 2 in 2D. For the isotropic case the enhancement factor is
f ðω, θÞ ¼ πn: (4)
The upper limit for the angle-integrated absorption enhancement factor is

Z π=2
F2D ðωÞ ¼ dθf ðω, θÞ cos θ ¼ 2πn (5)
π=2
which applies to isotropic as well as anisotropic cells in 2D. These 2D results

are not nearly as well known as the original 4n2 limit in 3D. Moreover, to the
best of our knowledge were first derived using the wave optics light trapping
theory formalism (Yu et al., 2010b). Therefore below we provide a bit more
details on the derivation of the πn limit for the 2D isotropic case using ray
optics.
For a uniform 2D structure, it is straightforward to obtain a factor of n
from the enhancement of light intensity and another factor of 2 from a back
mirror, following the same approach as in the derivation of the 4n2 limit
(Yablonovitch, 1982). For an oblique ray at an angle θ, the path length is
enhanced by 1/ cos θ. The angle-integrated enhancement of the path length
is obtained by an angular integration over the individual path length
enhancement, multiplied by a weighting factor that depends on the surface
property and the dimension of the problem (Battaglia et al., 2012b). For a
Lambertian surface, the weighting factor is cosθ normalized by an angular
integration over cosθ. Therefore, the averaged path length enhancement
in 2D is
Z π=2
1 cos θ π
dθ Z π=2 ¼ (6)
π=2 cos θ 2
dθ cos θ
π=2
instead of the 3D factor of 2. Multiplying the three enhancement factors, we

obtain the πn enhancement for 2D light trapping.
The different versions of the conventional limits are summarized in
Table 1.
3. Wave optics light trapping theory

The understanding of light trapping lies at the heart of optical design
for solar cells (Battaglia et al., 2012a; Brongersma et al., 2014; Gomard et al.,
2013; Green, 2011; Liang et al., 2012; Mann and Garnett, 2013; Naqavi
Table 1 Conventional light trapping limits in 2D and 3D structures

for a single incidence angle and after angular integration. n is the
refractive index of the weakly absorbing material
Dimension Isotropic Angle-integrated
2D πn 2πn
2
3D 4n 4πn2
et al., 2013a; Pala et al., 2013; Peretti et al., 2012; Sheng et al., 2011; Vynck
et al., 2012; Wang et al., 2012, 2013a, 2014a,b; Yu et al., 2012; Zeman et al.,
2013). Conventional light trapping theory, largely developed in the context
of c-Si solar cells in the 1980s, was initially developed using the tools of
ray optics. For solar cells with thicknesses or feature sizes that are comparable
to the wavelengths of sunlight, ray tracing is no longer applicable. It is there-
fore essential to develop a light trapping theory entirely from a wave optics
perspective.
The development of wave optics light trapping theory has a long history.
Yablonovitch, in one of his initial papers that elucidate the 4n2 limit
(Yablonovitch, 1982; Yablonovitch and Cody, 1982), used a blackbody
radiation argument that is directly related to the modal structure of the cell.
Sheng, Bloch, and Stepleman argued that the DOS plays an important role
and hinted at the possibility of modifying the ray optics limit in thin film
structures (Sheng et al., 1983). Stuart and Hall examined light trapping in
wavelength-scale thin films and concluded that such a wavelength-scale thin
film in fact has an absorption enhancement limit comparable to and some-
what below that of the bulk cell (Stuart and Hall, 1997). All these studies
have pointed out the importance of modal structures of the cells in the
understanding of the behaviors of light trapping.
As an important step toward a wave optics light trapping theory, Yu,
Raman, and Fan recently introduced a statistical temporal coupled mode
theory (TCMT) formalism to account for light trapping in the wave optics
domain (Yu et al., 2010a,b). This theory has completely reproduced all the
key results of conventional ray optics light trapping theory, including the
Yablonovitch 4n2 limit (Yu et al., 2010a), the correction to the 4n2 limit
when the single-pass absorption is nonnegligible (Yu et al., 2012), and
the correction to the 4n2 limit for an absorber with angular selective absorp-
tion (Yu and Fan, 2011; Yu et al., 2010a). Moreover, this wave optics theory
predicts that one can significantly overcome the conventional light trapping
limit in certain nanophotonic structures, a prediction that has been validated
with full-field simulations showing that the conventional limit can be over-
come for all incident angles and over a wide wavelength ranges (Wang et al.,
2014a; Yu et al., 2010a). In this section, we review this wave optics light
trapping theory.
A solar cell is modeled as a structure supporting a number of optical
modes that may couple to a number of free-space channels and thus become
resonances. Light trapping is described by the coupling of incident radiation
into the optical modes supported by the solar cell structure. Absorption is
enhanced through the aggregate contribution of these resonances. This

physics can be captured by a statistical temporal coupled-mode theory
(Yu et al., 2010a,b). We assume that free space provides N planewave chan-
nels and that the absorber structure supports M resonances. Each planewave
channel could guide incoming radiation to each resonance, and each reso-
nance could also leak to each planewave channel. The coupling between
the mth resonance to the planewaves with an incident wave from the nth
channel is described by the following temporal coupled-mode equation
0 XN 1
d γ + γ
am ¼ @jωm s¼1 m, s 0
Aam + jpffiffiffiffiffiffiffi
ffi
γ m, n Sn (7)
dt 2
The resonance amplitude am is normalized such that jam j2 is the electro-

magnetic energy stored in the resonant mode per unit area, and the incident
planewave amplitude Sn is normalized such that jSn j2 corresponds to its
intensity. (n appearing in the subscript denotes the nth channel.) ωm is the
resonant frequency of the mth resonance. γ m, s is the loss rate from the
mth resonance to the sth channel 8 s 2 {1, 2, … , N}. The loss rate is equal
to the in-coupling rate, according to energy conservation and time-reversal
symmetry. γ 0 is the loss rate of the resonance due to intrinsic material
absorption.
In the uniform absorber of Fig. 2, if the thickness is sufficient large that
the resonances completely overlap with the absorber in space, γ 0 ¼ αc/n. In a
photonic crystal (for example, in Fig. 6), however, the overlap is generally
less than unity ( Joannopoulos et al., 2008; Naqavi et al., 2013a). We define Γ
to be the overlap, or confinement factor, that characterizes the average over-
lap of the electric field with the weakly absorbing active material. Hence, the
intrinsic loss rate is in general given by
c
γ0 ¼ α Γ (8)
n
To rigorously derive Eq. (8), we can treat the absorption as a perturba-
tion to a transparent structure and integrate over the energy loss vs total
energy throughout the entire space. This intrinsic loss rate modification
generalizes the wave optics light trapping theory over the earlier work
(Yu and Fan, 2011).
We calculate the contribution of a single resonance to the absorption
spectrum excited by the incident sunlight from a single channel. For the
mth resonance and the nth channel, this contribution is
e
γ m γ m, n
A m , n ð ωÞ ¼ , (9)
ðω ωm Þ + ðe
2
γ m + γ m Þ2 =4
where ω is the frequency, ωm is the resonant frequency of the mth mode, eγ m
P
is the intrinsic material absorption rate of the mth mode, γ m ¼ N n¼1 γ m, n is
the total external coupling rate of the mth mode, γ m, n is the decay rate of the
mth mode into the nth channel (due to reciprocity, it is also the in-coupling
rate of the mth mode from the nth channel), and

πω2 L 2
N ¼2 2 (10)
c 2π
is the total number of channels in real space with a period L. For incident
sunlight from the nth channel, the overall absorption coefficient, calculated
using the concept of spectral absorption cross section (Yu et al., 2010a), is
Z ω + Δω X M
1
An ðωÞ ¼ Am, n ðωÞω
Δω ω m¼1
M Z ω + Δω
1 X
¼ Am, n ðωÞω (11)
Δω m¼1 ω
M Z ∞
1 X
Am, n ðωÞω,
Δω m¼1 ∞
where

4πn3 Δðω3 Þ L 2 d
M ¼2 (12)
3c 3 2π 2π
is the total number of resonances in the spectral range (ω, ω + Δω) supported
by the structure with a thickness d and a refractive index n. The approxima-
tion on the limits of integration in the last step of Eq. (11) is justified since Δω
is chosen to be much greater than the linewidths of the resonances, which
have high qualify factors, while still being much smaller than ω. It remains to
justify the summation in Eq. (11), which holds only if the M modes are
orthogonal.
From Eq. (7), we can calculate the broadband absorption enhancement
by the mth resonance when light is incident from the nth channel. By sum-
ming over the contributions from all M resonances and N channels in the
frequency range of [ω, ω + Δω], and by comparing the absorption to the
single-pass absorption αdeff, we can calculate the angle-integrated absorption
enhancement factor F, which in the k space translates into a summation over

all the channels
X
γ m, n γ 0
X 2π X n 2πcΓ
F¼ fn ¼ X < ρðωÞ (13)
n
αdeff Δω m γ m, n + γ 0 nηd
n
where ρ(ω) ¼ M/Δω is the spectral DOS, and fm is the contribution to the
enhancement factor from the mth resonance. η is the volume or area fraction
P
of the absorbing material. The equality is approached if N n¼1 γ m, n ≫γ 0 for
all m, that is, all the resonances are in the over-coupling regime. Since
the material is weakly absorptive, the equality holds as long as all resonances
are coupled to external radiation. In other words, all photonic states in
the structure contribute to light trapping, and the averaging over the band-
width results in the appearance of DOS in the enhancement factor F. In
next section, we use Eq. (13) to show that the limit of 4n2 is only correct
in bulk structures and the absorption enhancement factor can go far beyond
this limit with proper design in the nanophotonic regime.
4. Fundamental limit of light trapping

Eq. (13) establishes a tight upper limit for absorption enhancement in
light-trapping structures. In a bulk structure, Γ ¼ 1, η ¼ 1, the various ver-
sions of the conventional limit are readily reproduced by the theory
(Table 1) (Yu and Fan, 2011; Yu et al., 2010a,b). In particular, we consider
a structure with period L and thickness d that are both much larger than the
wavelength. In this case, the resonance can be approximated as propagating
plane waves inside the bulk structure. The total number of resonances in the
frequency range (ω, ω + Δω) and the number of channels are provided by
Eqs. (12) and (10), respectively. With Eq. (13) divided by the number of
channels, we find that for the isotropic case

8πn3 ω2 L 2 d
2πc
c3 2π 2π
Fiso ¼ 2 ¼ 4n2 (14)
πω L2
2n 2 d
c 2π
which reproduces the 4n2 conventional limit, appropriate for the Lam-
bertian emission case with sinθ ¼ 1, where θ is the angle of the emission cone
in the medium surrounding the cell. The theory can be generalized to the
case of a restricted emission cone and reproduces the standard result of
4n2/sin2θ (Yu et al., 2010a).
The analysis here also points to scenarios where the conventional limit is
no longer applicable. Eq. (10) is not applicable when the periodicity is com-
parable to the wavelength, whereas Eq. (12) is not valid when the film thick-
ness is much smaller than the wavelength. Below, we consider both of
these cases.
4.1 Light trapping in structures with wavelength-scale

periodicity
We consider a square lattice where the periodicity L is comparable to the
wavelength λ. The discrete nature of the channels becomes important
(Fig. 3A). To illustrate this effect, we assume that the film has a high refrac-
tive index (for example, silicon), such that the wavelength in the material is
small compared with the periodicity. We also assume that the film has a
thickness of a few wavelengths. In this case, all modes have approximately
the same decay rate, and Eq. (12) can still be used to count the number of
resonances.
Using Eq. (13), for normally incident light, we calculate the upper limit
of the absorption enhancement factor as a function of L/λ (Fig. 3B) when
the structure has a square lattice. The discontinuous changes in Fig. 3B cor-
respond to the emergence of new channels. In particular, when λ > L, there
A ky B
12n2
8n2
k0 F
kx
4n2
2p/L 0
0 1 2 3 4 5 6 7 8
L/l
Fig. 3 Light trapping in periodic structures. (A) Blue dots represent channels in the k
space. Channels in the circle correspond to free-space propagating modes.
(B) Theoretical upper limit of the absorption enhancement factor using a light-trapping
scheme where a square-lattice periodic grating structure is introduced into a thin film.
Red area represents a spectral range where the upper limit of the absorption enhance-
ment factor Fiso is above 4n2.
is only a single channel independent of frequency. On the other hand, the

number of resonances is frequency dependent. As a result, the maximum
enhancement factor increases quadratically as a function of frequency. In
order to maximize the absorption, one should choose the periodicity to
be slightly smaller than the wavelength range of interest (red region in
Fig. 3B). We note that the upper limit for the absorption enhancement
factor approaches 4n2 for a large period, L ≫ λ.
The above analysis can be used to provide considerable insight into the
behavior of grating structures. In particular, one expects that a 2D grating
structure is superior to a 1D grating, because a 2D grating can provide access
to a significantly larger number of resonances. Also, an asymmetric grating
profile should be beneficial, because with a symmetric profile there are res-
onances that cannot be coupled to incident light due to symmetry con-
straints. These findings are consistent with existing literature (Heine and
Morf, 1995; Tobias et al., 2008).
The use of grating structures on a relatively thick film to enhance optical
absorption has been extensively explored (Bermel et al., 2007; Sheng et al.,
1983; Tobias et al., 2008). This approach is practically important because it
allows one to tailor the device response for specific material parameters and
operating conditions such as concentration. From a fundamental perspec-
tive, Sheng et al. have argued (Sheng et al., 1983) that the grating may alter
the density of state within the structure, leading to enhancement beyond 4n2
over particular frequency ranges. However, the cases we consider here
involve shallow gratings on the surface of a thick medium. In such a case,
the change of density of state in the structure is substantial only in very lim-
ited frequency ranges (Fan and Joannopoulos, 2002). Instead, our analysis
shows that enhancement beyond 4n2 is nevertheless achievable because
the grating restricts the number of channels available in free space. Also,
in Tobias et al. (2008) and Saeta et al. (2009), enhancement factors above
4n2 were predicted using approximate approaches involving a summation
of various scattering events in an incoherent fashion. The analysis presented
here is more general in the sense that it is based upon electromagnetic anal-
ysis. Moreover, our analysis indicates that the potential of significantly
exceeding the conventional limit, defined in terms of 4n2/sin2θ, is rather
limited in these structures; this conclusion arises because, to achieve high-
enhancement factors, one needs to use a periodicity comparable to the
wavelength of interest, which leads directly to strong angular and spectral
dependency, in consistency with previous results (Tobias et al., 2008).
Below, we present a strategy that overcomes these issues and exceeds the
conventional limit over a large range of angles and frequencies.
4.2 Light trapping in thin films

When the thickness d of the film is comparable to half wavelength in the
material, one can reach the single-mode regime where the film supports a
single waveguide mode band for each of the two polarizations. In such a
case, Eq. (7) is no longer applicable. Instead, the number of resonances in
the frequency range of [ω, ω + δω] can be calculated as

2πn2wg ω L 2
M ¼2 δω (15)
c2 2π
where the first factor of 2 arises from counting both polarizations. (Here, to
facilitate the comparison to the standard conventional limit, for simplicity,
we have assumed that the two polarizations have the same group index nwg.)
Notice that, in this case, the number of modes no longer explicitly depends
upon the thickness d of the film.
In order to highlight the effect of such strong light confinement, we
choose the periodicity to be a few wavelengths, in which case the number
of channels can still be calculated using Eq. (10). As a result, we obtain the
upper limit for the absorption enhancement factor
λ
Fiso ¼ 2 4n2wg V (16)
4nwg d
where the factor V ¼ αwg/α0 characterizes the overlapping between the pro-
file of the guided mode and the absorptive active layer. The absorption coef-
ficient and group index of the waveguide mode are αwg and nwg, respectively.
Eq. (16) in fact becomes 4n2 in a dielectric waveguide of d λ/(2n).
Therefore, reaching the single-mode regime is not sufficient to exceed
the conventional limit. Instead, to achieve the full benefit of nanophotonics,
one must either ensure that the modes exhibit deep-subwavelength-scale
electric-field confinement, or enhance the group index to be substantially
larger than the refractive index of the active material, over a substantial
wavelength range. Below, using both exact numerical simulations and ana-
lytic theory, we will design geometries that simultaneously satisfy both these
requirements.
Guided by the theory above, we now numerically demonstrate a nano-
photonic scheme with an absorption enhancement factor significantly exceed-
ing the conventional limit. We consider a thin absorbing film with a thickness
of 5 nm (Fig. 4A), consisting of a material with a refractive index and a
wavelength-independent absorption length of 25 μm. The film is placed on
a mirror that is approximated to be a perfect electric conductor (PEC).
Scattering layer e = 12.5 Active layer e = 2.5

Cladding layer e = 12.5 Mirror
Scattering
Cladding
Active
Mirror
Fig. 4 Structure for overcoming the conventional light-trapping limit. (A) A nano-
photonic light-trapping structure. The scattering layer consists of a square lattice of
air groove patterns with periodicity L ¼ 1200 nm. The thicknesses of the scattering,
cladding, and active layers is 80, 60, and 5 nm, respectively. The mirror layer is a perfect
electric conductor. (B) The profile of electric-field intensity for the fundamental wave-
guide mode. Fields are strongly confined in the active layer. To obtain the waveguide
mode profile, the scattering layer is modeled by a uniform slab with an averaged dielec-
tric constant.
A PEC mirror is used for simulation convenience. In practice, it can be rep-

laced by a dielectric cladding layer, which produces similar results (Yu et al.,
2010a). Our aim here is to highlight the essential physics of nanophotonic
absorption enhancement. The choice of material parameters therefore rep-
resents a simplification of actual material response. Nevertheless, we note
that both the index and the absorption strength here are characteristic of typ-
ical organic photovoltaic absorbers in the weakly absorptive regime (Hoppe
and Sariciftci, 2004). Furthermore, there is general interest in using thinner
absorbers in organic solar cells given their short exciton diffusion lengths of
about 3–10 nm (Huynh et al., 2002; Mayer et al., 2007; Yu et al., 1995).
In order to enhance the absorption in the active layer, we place a trans-
pffiffiffiffiffiffiffiffiffi
parent cladding layer nH ¼ 12:5 on top of the active layer. Such a cladding
layer serves two purposes. First, it enhances density of state. The overall
structure supports a fundamental mode with group index nwg close to nH,
which is much higher than that of the absorbing material. Second, the index
contrast between active and cladding layer provides nanoscale field confine-
ment. Fig. 4B shows the fundamental waveguide mode. The field is highly
concentrated in the low-index active layer, due to the well-known slot-
waveguide effect (Almeida et al., 2004). Thus, the geometry here allows
the creation of a broadband high-index guided mode, with its energy highly
concentrated in the active layer, satisfying the requirement in Eq. (16) for
high absorption enhancement.
In order to couple incident light into such nanoscale guided modes, we
introduce a scattering layer with a periodic pattern on top of the cladding
layer, with a periodicity L much larger than our wavelength range of inter-
est. Each unit cell consists of a number of air grooves. These grooves are
oriented along different directions to ensure that scattering strength does
not strongly depend on the angles and polarizations of the incident light.
We emphasize that there is no stringent requirement on these grooves as
long as the scattering strength dominates over resonance absorption rates.
We simulate the proposed structure by numerically solving Maxwell’s
equations (Fig. 5A). The device has a spectrally averaged absorption
enhancement factor of Fiso ¼ 119 (red line) for normally incident light.
(All the absorption spectra and enhancement factors are obtained by averag-
ing s and p polarized incident light.) This enhancement factor is well above
the conventional limit for both the active material (4n2L ¼ 10) and the clad-
ding material (4n2H ¼ 50). Moreover, the angular response is nearly isotropic
(Fig. 5C and D). Thus such enhancement cannot be attributed to the
narrowing of angular range in the emission cone, and instead is due entirely
to the nanoscale field confinement effect.
Using our theory, we calculate the theoretical upper limit of light-
trapping enhancement in this structure. For wavelength λ ¼ 500 nm, we
obtain an upper limit of Fiso ¼ 147. The enhancement factor observed in
the simulation is thus consistent with this predicted upper limit. The actual
enhancement factor obtained for this structure falls below the calculated the-
oretical upper limit because some of the resonances are not in the strong
overcoupling regime.
A B
100
10–1
10–2
10–3
10–4
400 500 600 400 500 600
wavelength (nm)
C 0 D
0
–30 30 –30 30
–60 60 –60 60
–90 90 90
–90
0 50 100 150 0 50 100 150
F F
j = 0° j = 45°
Fig. 5 Absorption with the light-trapping structures. (A) Absorption spectrum for nor-
mally incident light for the structure shown in Fig. 4. The spectrally averaged absorption
(red solid line) is much higher than both the single-pass absorption (light-gray dashed
line) and the absorption as predicted by the limit of 4n2L (dark-gray dashed line). The ver-
tical axis is the absorption coefficient. (B) Absorption spectrum without nanoscale light
confinement. The structure is the same as that of (A) except that the dielectric constant
of the active layer is now the same as the cladding layer. The dark-gray dashed line rep-
resents the absorption as predicted by the limit of 4n2H. (C and D) Angular dependence of
the spectrally averaged absorption enhancement factor for the structure in Fig. 4. Inci-
dent angles are labeled on top of the semicircles. Incident planes are oriented at 0
(C) and 45 (D) degrees (azimuthal angles) with respect to the [10] direction of the lattice.
The red circles represent the 4n2L limit.
To illustrate the importance of nanoscale field confinement enabled by

the slot-waveguide effect, we change the index of the material in the absorp-
tive layer to nH. Such a structure does not exhibit the slot-waveguide effect.
The average enhancement in this case is only 37, falling below the conven-
tional limit of 50 (Fig. 5B).
In this section, we have shown that properly designed nanophotonic
structures can achieve enhancement factors that far exceed the conven-
tional limit. Our results indicate that substantial opportunities for nano-
photonic light trapping exist using only low-loss dielectric components.
The basic theory, moreover, is generally applicable to any photonic structure,
including nanowire (Kayes et al., 2005; Law et al., 2005) and plasmonic struc-
tures (Atwater and Polman, 2010). In plasmonic structures, the presence of
nanoscale guided modes may also provide opportunities to overcome the
conventional limit.
5. Light trapping in photonic crystals

To highlight the unique physics of light trapping in photonic crystals,
and as an illustration of the model system, we consider the structure shown
in Fig. 6. It consists of a photonic crystal with a finite number of periods in a
vertical direction, with a perfect mirror on its bottom, and with light inci-
dent from the top. The crystal is assumed to be of infinite extent in the hor-
izontal directions, and to consist of a periodic array of dielectric elements in
air. The dielectric has a real part of refractive index of n, which is assumed to
be frequency independent throughout this section, and to be weakly absorb-
ing with a small absorption coefficient α ≪ 1/d where d is the thickness of the
structure. We assume that the single-pass absorption is negligible. Knowing
Fig. 6 A 2D photonic crystal structure consisting of a square lattice of dielectric rods in

air. The radius of each rod is 0.2a, where a is the lattice constant. The dielectric constant
of the rods, colored in blue, is 12, with an extremely small imaginary component. In the
absorption simulations, the absorption is obtained through 10 layers of such rods with
the other dimension being infinite and with a perfect mirror at the bottom. Light is inci-
dent on top. In the band structure and overlap factor calculations, the structure is
treated as a lossless 2D photonic crystal infinite in both dimensions. The equivalent
thickness of the structure is equal to the thickness of the slab in Fig. 1 so that the
two structures contain the same amount of absorbing materials.
the absorption enhancement factor in such a weak absorption limit, one can
derive the enhancement factor where the single-pass absorption is no longer
negligible (Green, 2002; Yu et al., 2012). Focusing on the weak absorption
limit therefore allows us to focus on illustrating the most prominent effects of
the DOS engineering in photonic crystals on light trapping absorption
enhancement.
Real solar cells are of course 3D structures. To simplify numerical cal-
culations, however, many theoretical literatures on light trapping in solar
cells have considered 2D structures (Chutinan and John, 2008; Chutinan
et al., 2009; Duche et al., 2011; Gomard et al., 2012; Park et al., 2009;
Yu et al., 2010b). By 2D, we mean structures that are uniform in the third
dimension. Moreover, we consider only light propagating within a 2D plane
perpendicular to the third dimension.
The wave optics theory in Section 3 points to several unique aspects of
light trapping in photonic crystals. We examine the most prominent devi-
ations of the optical properties of photonic crystals from those of bulk struc-
tures. First, the DOS of a photonic crystal (for example, in Fig. 6) can differ
significantly from that of the bulk structure (for example, in Fig. 2). Second,
for a bulk structure, if the thickness is greater than a few wavelengths, one
can assume complete overlap of the modes with the absorbing material. In
contrast, in a photonic crystal, the electromagnetic fields do not completely
overlap with the absorbing material, and therefore the modal absorption loss
rate is related to the photonic crystal geometries as well as material constants.
Third, to access all the modes in a bulk structure, one typically employs sur-
face roughness. In contrast, we will show that a photonic crystal by itself,
with appropriate configurations, provides complete access to all its modes
without surface disorders.
5.1 Light trapping in 2D photonic crystals

5.1.1 Relevant aspects of the band structure of a 2D photonic crystal
As a model system, we start by considering a lossless and infinite 2D photonic
crystal consisting of a square lattice of dielectric rods in air (Fig. 6). The
radius of each rod is r ¼ 0.2a, where a is the lattice constant. The dielectric
material has a nondispersive dielectric constant of 12, which is close to the
value of silicon or gallium arsenide at optical frequencies.
We used the MIT Photonic Band (MPB) package to calculate the band
structure ( Johnson and Joannopoulos, 2001). In Fig. 7, we show the band
structure of the photonic crystal along the line segments connecting the high
symmetry points. There exist photonic bandgaps for the transverse magnetic
0.8
Frequency (c/a)
0.6
0.4
TE modes
M
0.2 K1 K2
X
G
0 TM modes
G X M G
Fig. 7 Band structure of the photonic crystal in Fig. 6. The first Brillouin zone of the
square lattice is shown in the inset. Γ denotes the k point (0, 0), X denotes (0, π/a),
and M denotes (π/a, π/a).
Fig. 8 Projected TM band structure obtained from Fig. 7. The lightline is given by ω ¼ ckx,
where kx is the projected wave vector.
(TM) polarization, but not for the transverse electric (TE) polarization. In
order to highlight the features in the DOS of a 2D photonic crystal, we
choose to work with the TM polarization for all subsequent discussions
in 2D structures.
The same photonic band structure in Fig. 7 can alternatively be presented
as a projected band diagram (Fig. 8), where we project the 2D band structure
ω(kx, kz) onto the ω kx plane. The shaded regions in Fig. 8 corresponds to
states in the photonic crystal, the unshaded region is the gap region. We
notice that a significant part of the first band and some parts of the second
band lie below the light line ω ¼ ckx. Such a projected band structure will be
useful for the discussion of mode coupling issues.
We can calculate the DOS of this photonic crystal by a uniform sampling
of all the k-points in the first Brillouin zone. In Fig. 9, we observe two
bandgaps at frequencies close to 0.3c/a and 0.7c/a, as well as a number of
van Hove singularities where the DOS is divergent (Bassani and
Parravicini, 1975; Van Hove, 1953).
For each mode, we can also calculate the overlap factor Γ between the
electric field and the dielectric rods using the Dirichlet-to-Neumann (DTN)
method (Liu and Fan, 2011). At any given frequency, we select two k points
from the segments connecting Γ to X to M and back to Γ. The selected k
points correspond to the two leftmost intersections of horizontal cuts with
the band structure plot in Fig. 7. Fig. 10 shows the overlap factors for such
Fig. 9 Density of TM states per area for the 2D photonic crystal of Fig. 6.
Fig. 10 Overlap factor of the TM mode for two selected k points at each frequency. The
k points are selected from left to right, for example, at a frequency of 0.2c/a, the first two
k points are shown as K1 and K2, in Fig. 7. Inside the bandgap, the overlap factor is not
defined.
two k points for a range of frequencies. We observe that, at each frequency,

the modes at different k points tend to have very similar overlap factors. In
the limiting case at zero frequency, the field becomes uniform, and the over-
lap factors approach the area fraction η ¼ πr2/a2 0.126. The first and sec-
ond bands have quite different overlap factors, which can be explained
by the variational theorem ( Joannopoulos et al., 2008). The first band tends
to concentrate more of its electromagnetic energy in the dielectric, hence
the overlap factors are higher than the area fraction η. Within the first
band, there is an increase in overlap factors with respect to frequency,
following the trend in the DOS. Γ is not defined in the photonic bandgaps,
where light propagation and absorption are forbidden.
5.1.2 Absorption enhancement factor in the 2D photonic crystal

and comparison to the density of states
We calculate the absorption in a lossy and finite version of the 2D photonic
crystal shown in Fig. 6. The crystal contains a finite number of periods in a
vertical direction, with a perfect mirror on its bottom and with light incident
from the top. The crystal is assumed to be of infinite extent in the horizontal
directions. In the numerical simulations, we chose an extremely weak atten-
uation coefficient α ¼ 108/d, and this particular numerical value disappears
after taking the ratio between the calculated absorption and the single-pass
absorption defined in Eq. (3). The 2D photonic crystal contains 10 layers in
the normal direction of incident light. Because the absorption is weak and
there are a sufficient number of layers, the DOS and overlap factors can be
well approximately by the corresponding lossless and infinite photonic
crystal.
We calculate the absorption enhancement factor f(θ) for the entire range
of incident angles θ with the Fourier Modal Method using the Stanford
Stratified Structure Solver (S4) package (Liu and Fan, 2012). From
f(ω, θ), we obtain the angle-integrated absorption enhancement factor
F2D using Eq. (5). Numerically, we observe that F2D is not influenced by
the choice of number of layers as long as the number is sufficiently large.
We plot F2D in Fig. 11 and compare it to both the DOS bound given by
Eq. (13) and the conventional limit of 2πn.
To calculate the DOS bound on the right-hand side in Eq. (13), we first
convert the summation in the real space as an angular integration in 2D to a
summation over channels in the wave vector space. The 2D angle-
integrated absorption enhancement factor can be written as
Fig. 11 Comparison of angle-integrated light trapping absorption enhancement factor

and its theoretical upper bound for the 2D photonic crystal in Fig. 6.
Z π=2
F2D ¼ dθf ðθÞ cos θ
π=2
Z (17)
1
¼ dkx f ðkx Þ
k0 |kx |<k0
Δk X
¼ fn (18)
k0 n
where θ is the angle of incidence and kx ¼ k0 sin θ

ω
k0 ¼
c
2π
Δk ¼
L
where L is the lattice constant of the photonic crystal or an artificially
imposed period of a bulk structure. Using Eqs. (13) and (5), we obtain
ð2πc Þ2 Γ
F2d < ρ ð ωÞ (19)
nηω 2D
where ρ2D(ω) is the DOS per unit area.
We calculated ρ2D(ω) (Fig. 9), as well as Γ(ω) (Fig. 10) for selected k
points. Since the overlap factors Γ are approximately equal for different k
points at the same frequency, we simply take an average of the two overlap
factors shown in Fig. 10, and use the averaged overlap factor Γ in Eq. (19).
With these information and other parameters of the structure, we can obtain
the right-hand side expression (2πc)2Γρ2D(ω)/nηω of Eq. (19), which gives
an upper bound for F2D. We plot this bound for a range of frequencies in
Fig. 11, referring to it as the “Theoretical Bound” curve.
Fig. 11 verifies our theoretical result on the correspondence between the
absorption enhancement and the theoretical bound. We note in particular
the following features. First, over a few narrow ranges of frequencies, the
Yablonovitch limit is surpassed due to the van Hove singularities in the pho-
tonic DOS. Second, overall the absorption enhancement is below the con-
ventional limit. Third, the photonic crystal absorber, without grating or
surface roughness, is itself a very efficient scatterer. Over most frequency
ranges above 0.3c/a, there is a close match between the DOS and the absorp-
tion curves, implying that most accessible resonances are excited. Unlike the
conventional case of uniform slab, there is no need to use a surface grating to
achieve light trapping for photonic crystals. However, on the other hand, for
the first band from a frequency of 0 to approximately 0.3c/a, the numerically
obtained light trapping absorption enhancement is significantly lower than
its upper bound imposed by the DOS. This is because many modes are
below the light line and decoupled from external radiation, as we have
eluded to in the discussion of Fig. 8.
5.1.3 van Hove singularities in the density of states

For absorption enhancement, we can engineer the photonic crystals and take
advantage of the van Hove singularities in the DOS (Bassani and Parravicini,
1975). For practical applications, we might be able to apply the narrowband
divergence of DOS to achieve strong absorption improvement, for example,
at the bandedge in c-Si where light trapping is crucial (Wang et al., 2012; Yu
et al., 2010b).
Van Hove singularities are generally more prominent in lower dimen-
sions (Bassani and Parravicini, 1975). In fact, the DOS can diverge in 2D
(Fig. 9, for example) while only its derivative can diverge in 3D (Fig. 19,
for example). As a result, in 2D, the angle-integrated absorption enhance-
ment factor indeed surpasses the Yablonovitch limit at van Hove singular-
ities (Fig. 11); however, in 3D, similar absorption enhancement is much
more difficult to achieve.
The van Hove singularities in some sense redistribute the optical prop-
erties of photonic crystals (Gaponenko, 2010). For real materials, redistribu-
tion of the DOS is constrained by the DOS sum rule (Barnett and Loudon,
1996; Scheel, 2008), which implies that an increase in DOS in one fre-
quency region needs to be compensated by a decrease in other frequency
regions, limiting the bandwidth and magnitude of the van Hove
singularities. In this photonic crystal, the theoretical bound exceeds the con-
ventional limit in the frequency range approximately from 0.22c/a to 0.27c/a,
due to the van Hove singularity in DOS. Assuming that the frequency range
is centered at a free space wavelength of 1000 nm, this frequency range then
corresponds to a wavelength range of 907 to 1114 nm. Therefore, it is pos-
sible to use van Hove singularities to achieve light trapping limit above the
conventional limit over a bandwidth that is relevant for solar cell light trap-
ping at the crystalline silicon bandedge. We may also stack multiple photonic
crystals to use multiple van Hove singularities for an even broader bandwidth
of light trapping absorption enhancement.
5.1.4 Long wavelength limit

We observe that away from the van Hove singularities, the absorption
enhancement factor in a photonic crystal is in general lower than the 2πn
limit. This is partly due to the fact that the overlap factor Γ is lower than
the area or volume fraction, and, more importantly, due to the lower effec-
tive DOS of the photonic crystal (Fig. 6) compared to the DOS in the bulk
structure (Fig. 2). We can use the effective medium theory ( Joannopoulos
et al., 2008), which is accurate in the long wavelength limit, to provide a
more in depth discussion of the DOS.
In the long wavelength (low frequency) limit, the light does not probe
the fine details of the photonic crystal (Lin et al., 2005). At low frequencies in
Fig. 9, the DOS of the 2D photonic crystal is linear with respect to fre-
quency, which follows the form of the DOS in a homogeneous dielectric
material in 2D. In the long wavelength limit, we can therefore estimate
the DOS of the photonic crystal by treating it as a homogeneous effective
medium and using the effective dielectric constant of this effective medium.
In general, consider a composite material constructed from two constituents
with dielectric constants ε1 and ε2, and area or volume filling fractions of f1
and f2, respectively. The effective dielectric constant ε of the composite is
then bounded as follows (Bergman, 1978, 1980; Milton, 1981, 2002):
1 1
f1 ε1 + f2 ε1
2 <ε<f1 ε1 + f2 ε2 (20)
In 2D, imagine that we start with a bulk structure of weakly absorbing
material with ε1 and area A, and compare the structure to a photonic crystal
structure for which we use the same amount of absorbing material with ε1
but dilute it with a transparent material with ε2 to a total area of A/η, where η
is the area fraction of the absorbing material. Then f1 ¼ η and f2 ¼ 1 η.
The bound of light trapping enhancement in 2D, as pointed out in

Eq. (19), is related to the DOS ρ(ω), as well as to the area fraction η and
the overlap factor Γ. Since the overlap factor Γ η under the assumption
that the electric field is distributed uniformly in space, the area fraction η
and the overlap factor Γ cancel out in Eq. (19), and the bound of light trap-
ping enhancement is approximately proportional to ρ(ω). Furthermore, the
2D DOS is proportional to n2 or ε:
4εω
ρ2D ðωÞ ¼ (21)
πc 2
We further assume that all modes are accessible, which can be made pos-
sible by methods discussed in this section. Using Eqs. (20) and (21), the
upper bound of the angle-integrated light trapping enhancement factor
pffiffiffiffi
F2D/2πn, where n ¼ ε1 , is then given by
F2D ε2
<η + ð1 ηÞ (22)
2πn ε1
In Fig. 12, we plot this upper bound in Eq. (22) as a function of the area
fraction η of the absorbing material. In Fig. 12A we consider the hypothet-
ical case where the absorbing material has a low dielectric constant ε1 and is
embedded in a transparent background of a high dielectric constant with
ε2 ¼ 12ε1. In this case, F2D/2πn significantly exceeds unity, indicating that
such a structure has a light trapping potential that is significantly above con-
ventional light trapping with the same absorbing material in a uniform bulk
form. This effect was noted (Yu et al., 2010a) and is due to the enhanced
DOS from the high index dielectric material at the background.
In Fig. 12B, we consider the more typical case, where the absorbing
material has a high dielectric constant ε1 ¼ 12 and is embedded in a transpar-
ent background with a low dielectric constant ε2 ¼ 1. In this case, F2D/2πn is
always less than unity. In other words, in the long wavelength limit, such a
structure will always underperform, in terms of its light trapping capability,
as compared to a uniform bulk absorbing medium with the same dielectric
constant ε1. The underperformance arises because the presence of the low
index material significantly reduces the DOS of the overall structure. The
theoretical results illustrated in Fig. 12B are consistent with all numerical
results presented in this section. In particular, while these theoretical results
were derived using the effective medium theory, which strictly speaking is
correct only in the long wavelength limit, all our numerical results indicate
that away from the van Hove singularities, the photonic crystal structure that
Fig. 12 Upper bounds of 2D angle-integrated absorption enhancement factors by

Eq. (22). η is the area fraction of the absorber whose index is n. (A) ε2/ε1 ¼ 12.
(B) ε1/ε2 ¼ 12.
we examine has an absorption enhancement that significantly under-

performs the conventional limit for the corresponding uniform bulk
medium. While the DOS away from the long wavelength limit certainly
cannot be estimated from the effective medium theory, the overall trend,
which is that the use of a low index material should reduce the average
DOS, should in general be valid. This effect also limits the bandwidth over
which the use of van Hove singularities can outperform the conventional
limit, since it sets a lower floor on the enhancement factor away from
van Hove singularities.
Besides the considerations on light trapping, antireflection is also crucial
for any practical solar cells (Raut et al., 2011). In particular, optimal anti-
reflection is a prerequisite for optimal light trapping (Yu et al., 2012).
Although the DOS of a photonic crystal structure might be lower than that
of the bulk structure due to “dilution” of the active material on average, this
dilution also lowers the effective refractive index of the photonic crystal
structure and provides better optical impedance matching to free space. This
mechanism of antireflection adds another dimension to the optimization
problem of designing photonic crystal absorbers.
5.1.5 Mode coupling

Eq. (13) provides a tight upper limit for the angle-integrated absorption
enhancement factor F. The upper limit can be reached if all the modes
are in the over-coupling regime. Since we have assumed that the single-pass
absorption is weak, any mode that can couple to the external radiation is
over-coupled. Therefore, to reach the upper limit one only needs to ensure
that all modes can couple to external radiation. This is evident from Fig. 11,
where the absorption enhancement is significantly lower than its limit given
by Eq. (13) in the first band, and the limit can be approximately reached only
for the second and higher bands. As we noted above, if the modes in the first
band were able to couple to external radiation, the 2πn light trapping limit
would also be surpassed over a sizable bandwidth, approximately from
0.22c/a to 0.27c/a in Fig. 11, at the lowest van Hove singularity. Therefore,
it is important to investigate possible methods that would lead to complete
accessibility of all the modes.
For the structure shown in Fig. 6, the mode coupling aspect can be
understood by considering the projected band diagram shown in Fig. 8.
Only modes above the light line, that is, modes satisfying ω ckx, can couple
to external radiation. Since kx 2 [0, π/ax], where ax ¼ a in this case is the peri-
odicity of the truncated structure along the x direction, if follows that sig-
nificant number of modes below ω ¼ 0:5 2πc a , which includes a great portion
of the first band and a small portion of the second band, cannot couple to
external radiation and thus cannot contribute to light trapping. This effect
is seen in Fig. 11, where in the frequency range of the first band, the numer-
ically determined enhancement factor falls far below the theoretical bound
of Eq. (5), whereas the numerically determined enhancement factor agrees
quite well with the theoretical bound in the upper bands.
For a bulk structure, efficient coupling can be achieved when the light
rays are randomized by surface roughness, or when the light is scattered by a
grating structure with a period much larger than the wavelength of light
(Naqavi et al., 2013b; Yu and Fan, 2011; Yu et al., 2010b). The same under-
lying physics applies to photonic crystals in the wave optics regime. In the
following, we explicitly demonstrate two simple methods to achieve com-
plete coupling. Both methods seek to increase the spatial periodicity ax of the
truncated photonic crystal along the x direction.
In the first method, we simply choose a different crystal truncation. As an
example, in Fig. 13 we choose the orientation of the crystal such that the
pffiffiffi
period along the surface ax ¼ 5a. In this case, one can show that all the
modes in the crystal are above the light line. Many other different trunca-
tions are possible.
In the simulation results shown in Fig. 14, we indeed observe significant
improvement of light trapping enhancement in the first band, as compared
to the structure in Fig. 6. Also, the light trapping enhancement in the upper
bands remain essentially unchanged as we vary the crystal orientation, as
expected from the mode coupling argument.
Fig. 13 A different truncation of the 2D photonic crystal in Fig. 6. The structure is rotated
pffiffiffi
by 26.565∘ from the original structure in Fig. 6. The effectively period is 5a, where a is
the lattice constant of the square lattice.
Fig. 14 Comparison of angle-integrated absorption enhancement spectra for the pho-

tonic crystals in Fig. 6 (labeled as PC1) and Fig. 13 (labeled as PC2).
In the second method, we add a layer of grating on top of the original

structure. In Fig. 15, we reuse the structure from Fig. 6, but add a top layer
of transparent rods to keep deff unchanged. We vary the radii of the five
transparent rods in that top layer, therefore creating random scattering of
incoming light. Such a layer effectively increases the period to 5a.
In Fig. 16, we compare the light trapping enhancement without and
with the random layer. Again, the second and higher bands are nearly
unchanged, but, in the presence of scattering, the absorption in the first band
is vastly improved up to the theoretical bound given in Eq. (19).
With either method, we are able to achieve nearly complete coupling of
all the modes. Near the frequency range of the first van Hove singularity,
roughly between 0.2c/a and 0.3c/a, the calculated light trapping absorption
enhancement surpasses the conventional 2πn limit.
5.2 Light trapping in 3D photonic crystals

The formalism that we have developed for 2D systems can be readily
extended to 3D systems. For example, in a 3D system, using Eq. (13), the
angle-integrated light trapping enhancement factor can be expressed as
Fig. 15 The photonic crystal in Fig. 6 with one additional scattering layer on top. The
dielectric rods in the first layer on top are transparent while the other rods are weakly
absorptive. The real part of the dielectric constant is 12. The radius of each absorptive
rod is 0.2a, where a is the lattice constant. The radii of the transparent rods are alter-
nating in every five rods, being 0.2a, 0.18a, 0.22a, 0.16a, and 0.24a, thus forming a
scattering layer.
Fig. 16 Comparison of angle-integrated absorption enhancement spectra for the pho-

tonic crystals in Fig. 6 (labeled as PC1) and Fig. 15 (labeled as PC3).
Z π=2 Z 2π
F3D ¼ dθ dφf ðθ, φÞcos θ sinθ,
0
Z Z 0 (23)
1
¼ 2 dkx dky f kx , ky ,
k0 k2x + k2y <k20
Δk2 X
¼ 2 fn (24)
k0 n
where θ is the incidence angle, and φ is the azimuthal angle, and
kx ¼ k0 sin θ cos φ,
ky ¼ k0 sinθ sinφ,
ω
k0 ¼ ,
c
2π
Δk ¼ :
L
Similar to Eq. (19), we have
ð2πc Þ3 Γ
F3D < ρ ðωÞ (25)
nηω2 3D
where ρ3D(ω) is the DOS per unit volume.
Here, we highlight several notable differences between the 3D and 2D
photonic crystals. In 3D, one can show that in the long wavelength limit, for
a high index absorbing material embedded in a low index transparent back-
ground, the same index contrast and volume fraction would cause greater
DOS reduction than in 2D. The result of the effective medium theory for
this case gives
3
F3D ε2 2
< η + ð1 ηÞ (26)
4πn2 ε1
as plotted in Fig. 17 with ε1 ¼ 12 and ε2 ¼ 1. As a result, the absorption

enhancement is significantly lower than the conventional limit of 4πn2
(Table 1). In addition, the van Hove singularities are much less prominent
since the DOS at van Hove singularities is not divergent (although its deriv-
ative is divergent). Therefore, the angle-integrated light trapping enhance-
ment factor does not exceed 4πn2, although at individual angles and
wavelengths the 4n2 limit can be surpassed.
We study a 3D woodpile structure (Ho et al., 1994; S€ oz€
uer and Dowling,
1994), consisting of the same dielectric with ε ¼ 12 stacked in air as in the 2D
simulations, as shown in Fig. 18.
Fig. 17 Upper bounds of 3D angle-integrated absorption enhancement factors by

Eq. (26). η is the area fraction of the absorber whose index is n. (A) ε2/ε1 ¼ 12.
(B) ε1/ε2 ¼ 12.
Fig. 18 3D photonic crystal. The woodpile structure consists of four alternating layers.
Each layer has a thickness of 0.25a and consists of a 1D array of infinitely long square rods
with widths of 0.25a, where a is the lattice constant. The horizontal locations of the first
and the second layers are shifted laterally by 0.5a from the horizontal locations of the
third and the fourth layers. The lattice orientations of the first and the third layers are
orthogonal to the lattice orientations of the second and the fourth layers.
We obtain ρ3D(ω) by MPB for an infinite and lossless structure, and F3D
by S4 for the corresponding finite and lossy structure. We plot the right-
hand side expression (2πc)3ρ3D(ω)/nω2 in Eq. (25) as the “Theoretical
Bound” curve in Fig. 19, assuming that the overlap factor Γ is equal to
the volume fraction η. We plot F3D as the “Absorption Calculation” curve
in Fig. 19. The absorption enhancement follows the DOS in the same man-
ner as in the 2D case. We observe that both the theoretical bound and the
actual angle-integrated absorption enhancement factor fall significantly
below the conventional limit of 4πn2 ¼ 150 in this case.
To summarize this section, we have considered light trapping in pho-
tonic crystals, where the photonic crystals themselves operate as inhomoge-
neous absorbing media. We show that the DOS of photonic crystals strongly
influences the light trapping behaviors. We also note the importance of
modal overlap and the need for efficient mode coupling. Our numerical
results show that, in a 2D photonic crystal, with the use of van Hove singu-
larities in the DOS, the angle-integrated light trapping absorption enhance-
ment factor can exceed the conventional limit over substantial bandwidths.
In a 3D photonic crystal, it is more difficult to use photonic crystals to
overcome the conventional limit, due to the weakening of the van Hove
singularity and the fact that embedding a high index absorbing material in
a low index background reduces the overall DOS away from van Hove
singularities.
Fig. 19 Comparison of angle-integrated absorption enhancement factor and its theo-

retical bound calculated by Eq. (25) for the 3D woodpile structure in Fig. 18. The angle-
integrated limit 4πn2 ¼ 150.
6. Absorption enhancement in ultrathin crystalline

silicon solar cells with antireflection and
light-trapping nanocone gratings
In this section, we apply the light trapping theory to enhance the light
absorption in ultrathin-film silicon solar cells for efficiency improvement
and cost reduction. We introduce a double-sided grating design, where
the front and back surfaces of the cell are separately optimized for anti-
reflection and light trapping, respectively. The optimized structure yields
a photocurrent of 34.6 mA/cm2 at an equivalent thickness of 2 μm, close
to the Yablonovitch limit. This approach is applicable to various thicknesses
and is robust against metallic loss in the back reflector.
6.1 Motivation
There is significant recent interest in designing ultrathin c-Si solar cells with
active layer thickness of a few microns (Biswas and Xu, 2011; Dewan et al.,
2011; Fahr et al., 2011; Gjessing et al., 2011; Grandidier et al., 2011; Han
and Chen, 2010; Kelzenberg et al., 2010; Li et al., 2010; Mallick et al.,
2010; Nagel and Scarpulla, 2010; Paetzold et al., 2011; Park et al., 2009;
Sheng et al., 2011; Wang et al., 2010; Zhang et al., 2011). In Fig. 20, we
plot the complex dielectric constant of c-Si in comparison with the Air Mass
1.5 (AM 1.5) solar irradiance spectrum. The AM 1.5 spectrum is broadband,
hence it is beneficial to increase solar absorption over the entire optical spec-
trum. The real part of the dielectric constant is much greater than unity,
which implies that there is a strong optical impedance mismatch at the air
1.5 AM 1.5 Solar irradiance spectrum

W/m2/nm
0.5
50
Dielectric constants of crystalline silicon
40 er
30 ei
20
10
300 600 900 1200

nm
Fig. 20 The AM 1.5 solar irradiance spectrum and the real and imaginary parts of the
dielectric constant of c-Si. The blue curve is the real part and the green curve is the
imaginary part.
and c-Si interface, and as a result antireflection is desirable. Close to the c-Si
bandedge, the imaginary part of the dielectric constant is negligible, which
implies that the absorption depth of infrared photons is extremely large, and
as a result light trapping should be performed.
Efficient light absorption in such thin films therefore requires both
broadband antireflection coatings and effective light trapping techniques,
which often have different design considerations. In this section, we show
that by employing a double-sided grating design, we can separately optimize
the geometries for antireflection and light trapping purposes to achieve
broadband light absorption enhancement. The photocurrent generated by
the proposed thin film absorber is able to reach the Yablonovitch limit
(Green, 2002; Yablonovitch, 1982; Yablonovitch and Cody, 1982).
We use nanocones as the basic building elements for the grating geom-
etry because of their unique optical properties and compatibility with inex-
pensive fabrication techniques (Zhu et al., 2009, 2010a,b). The structure we
consider, as shown in Fig. 21, contains a c-Si thin film with nanocone grat-
ings also made of c-Si. The circular nanocones form 2D square lattices on
both the front and the back surfaces. The film is placed on a mirror. As a
starting point we assume the mirror is made of a PEC.
Fig. 21 3D c-Si thin film structures in air. Blue represents c-Si, gray represents PEC, and
yellow represents nonabsorbing silicon nitride. The nanocones are made of c-Si, as is the
uniform layers, and they are placed in a 2D square lattice on both surfaces of the film.
6.2 Absorption enhancement at Yablonovitch limit

The optimization process is as follows: For a given structure with 2D
nanocone gratings, using the rigorous coupled wave analysis (RCWA)
(Li, 1997; Liu and Fan, 2012; Tikhodeev et al., 2002), we calculate the
absorption spectrum, from which we determine the short circuit current
assuming an AM 1.5 incident solar irradiance. In a supercell of period
1000 nm, we optimize the geometry over six parameters—the numbers
of primitive cells and the base radii and heights of the nanocones on both
sides—for the greatest photocurrent generated from the structure. In the
optimization, we adjust those geometrical parameters, as well as the thick-
ness of the uniform layer sandwiched between the top and bottom gratings,
while ensuring that the structures always consist of the same amount of sil-
icon as a flat thin film structure with a predetermined thickness. We refer to
this thickness as the equivalent thickness of our nanostructured thin film.
Our optimized structure for an equivalent thickness of 2 μm is shown in
Fig. 22A. For the top nanocones, the period is 500 nm, the base radius is
250 nm, and the height is 710 nm; for the bottom nanocones, the period
is 1000 nm, the base radius is 475 nm, and the height is 330 nm. The
thickness of the uniform layer of the thin film is 1736 nm
For this structure, the absorption spectrum is shown in Fig. 23A. This
structure generates a short circuit current of 34.6 mA/cm2 assuming normal
incidence from an AM 1.5 solar spectrum.
In Fig. 23, we compare the performance of our structure to both the
single-pass absorption and the Yablonovitch limit of 2 μm-thick film.
Assuming perfect antireflection but no light trapping, that is, light passing
Fig. 22 3D c-Si thin film structures in air. In all panels, blue represents silicon, gray rep-
resents a PEC, and yellow represents nonabsorbing silicon nitride. The nanocones are
made of silicon, as is the uniform layers, and they are placed in a 2D square lattice either
on the front or on the back surface of the film. (A) The optimized double-sided nano-
structure, (B) The optimized top-only nanostructure. (C) The optimized bottom-only
nanostructure with a thin layer of nonabsorbing silicon nitride on top. (D) The flat film
with a thin layer of nonabsorbing silicon nitride on top.
through the material only once, and assuming normal incidence, the single-
pass absorption spectrum (the green curves in Fig. 23) in a thin film with
thickness d is given by
Asinglepass ¼ 1 eαd , (27)

where α is the absorption coefficient. Assuming perfect antireflection and
perfect light trapping, the absorption spectrum (the red curves in Fig. 23)
in a thin film with thickness d is given by the Yablonovitch limit (Green,
2002; Yablonovitch, 1982; Yablonovitch and Cody, 1982)
1
Ayablonovitch ¼ 1 , (28)
1 + 4n2 αd
where n is the real part of the refractive index. The absorption spectrum
of our optimized structure is much higher than the single-pass absorption
spectrum given by Eq. (27), and it is very close to the Yablonovitch-limit
spectrum given by Eq. (28). The short circuit current of 34.6 mA/cm2
in our optimized design is very close to 35.5 mA/cm2, the short-circuit
current corresponding to the Yablonovitch limit at the equivalent
thickness of 2 μm.
Fig. 23 Absorption spectra under normal incidence from AM 1.5 solar irradiance. The
red curves represent the Yablonovitch limit given by Eq. (28), the green curves represent
the single-pass absorption spectra given by Eq. (27), and the thick black curves are run-
ning averages of the absorption spectra for the corresponding structures in Fig. 22. The
kinks at the longer wavelength range in the curves are due to the discontinuity of the
material constant from the reference book (Palik, 1997). (A) The absorption spectrum of
the double-sided structure in Fig. 22A. (B) The absorption spectrum of the “top-only”
structure in Fig. 22B. (C) The absorption spectrum of the “bottom-only” structure in
Fig. 22C. (D) The absorption spectrum of the flat thin film in Fig. 22D. (E) Comparison
of the short-circuit currents generated by the four structures (gray bars), the
Yablonovitch limit (red line), the single-pass absorption (green line), and the full absorp-
tion (blue line).
The device physics for the absorption enhancement strategy in our struc-
ture is as follows: The usable solar spectrum for a c-Si cell extends from 300
to 1100 nm. To achieve significant absorption enhancement, one needs
broad-band antireflection over the entire usable solar spectrum due to the
strong impedance mismatch between silicon and air. One also needs efficient
light trapping from roughly 800 to 1100 nm where a silicon cell with an
equivalent thickness of a few microns is weakly absorbing. Our strategy is
to use the front surface grating for the goal of antireflection, and the back
surface grating for the goal of light trapping. Doing so allows us to separately
address the different structural requirements in order to achieve these two
separate goals. Below, we illustrate this strategy by comparing our optimized
structure to optimized “top-only” and “bottom-only” grating structures. All
these structures have an equivalent thickness of 2 μm. In the study of these
optimized “top-only” or “bottom-only” structures, we will compare their
performance to a flat thin film structure with the same equivalent thickness
and with a nonabsorbing silicon nitride antireflection coating on top, as
shown in Fig. 22D.
The optimized “top-only” structure and its absorption spectrum are
shown in Figs. 22B and 23B, respectively. The periodicity of the nanocone
grating is 500 nm, and the height is 900 nm, four times the base radius of
225 nm. The planar part of the structure has a thickness of 1809 nm. Com-
paring the absorption spectrum of such an optimized “top-only” structure
(see Fig. 23B) to that of the unpatterned flat thin film in Fig. 23D, we
observe substantial absorption enhancement over the entire usable solar
spectrum. The short-circuit photocurrent for the “top-only” structure is
31.7 mA/cm2, compared to 19.7 mA/cm2 for the flat thin film. From 300
to 700 nm, the absorption curve closely follows the Yablonovitch limit.
The contribution for light absorption enhancement in Fig. 23B, com-
pared to the flat thin film absorption in Fig. 23D, originates mainly from
antireflection. Nanocone arrays suppress reflection because the cone geom-
etry provides an averaged, graded index from air to silicon as the radius of its
cross section increases from zero to its maximum at the planar film surface.
The reflection suppression is broadband since the index-matching is largely
independent of wavelength. To achieve effective antireflection, the period-
icity of the array has to be in the subwavelength regime for the incoming
light to see an effective averaged index. In addition, a high aspect ratio is
preferred to provide a smooth index transition from air to silicon. These
structural aspects are precisely what we see in the optimized “top-only”
structure. However, for longer wavelengths, between 700 to 1100 nm,
the absorption of the optimized “top-only” structure falls significantly below
the Yablonovitch limit. Therefore, the structural feature of a nanocone that
is optimal for antireflection purposes is suboptimal for light trapping. Nev-
ertheless, such a nanocone structure for antireflection has excellent perfor-
mance in a wide range of light incident angles since the gradual change of
refractive index is maintained over this range (Zhu et al., 2009, 2010b).
The optimized “bottom-only” grating structure and its absorption spec-

trum are shown in Figs. 22C and 23C. The periodicity of the nanocone grat-
ing is 1000 nm. For an individual nanocone, the height is 400 nm, and the
base radius is 500 nm. The planar part of the structure has a thickness of
1895 nm to ensure the total 2 μm equivalent thickness. We also place a
90 nm thick nonabsorbing silicon nitride antireflection coating on top
(Han and Chen, 2010). This structure has a short circuit current of
30.5 mA/cm2. Comparing the absorption spectrum of the optimized
“bottom-only” structure to that of the planar structure in Fig. 22D, we
observe significant absorption enhancement beyond 600 nm, where the
absorption of the optimized “bottom-only” structure closely follows the
Yablonovitch limit. However, in the wavelength range below 600 nm,
light absorption is below the Yablonovitch limit and even below the
single-pass absorption spectrum due to optical loss from the reflection at
the silicon–air interface. The nanocone grating structure on the bottom sur-
face certainly does not contribute to antireflection.
A nanocone structure that is optimal for light trapping, as shown in
Fig. 22C, has very different structural characteristics, compared with the
optimal nanocone structure for antireflection in Fig. 22B. The spectrum
in Fig. 23C shows that the absorption enhancement due to light trapping
is primarily in the longer wavelength range where the silicon absorption
is weak. Light trapping relies on the excitation of guided resonances
(Yu and Fan, 2011; Yu et al., 2010a,b), and therefore a careful choice of
periodicity is important. The requirements on periodicity depend on two
considerations. On one hand, more guided resonances lead to higher
absorption, for which a large period is preferred. On the other hand, each
of the guided resonances should not couple and leak to many external chan-
nels, that is diffraction directions, for which a small period is preferred. Con-
sidering the trade-off between these two requirements, the optimal
periodicity for light trapping should be close to the targeted wavelength
(Yu et al., 2010a,b). For silicon, light trapping is most important for the
800 to 1100 nm wavelength range. We therefore expect that the optimal
periodicity is approximately 1000 nm, the wavelength at which the nor-
mally incident plane wave is efficiently coupled to guided resonances.
Taking into account the significant difference in the structural require-
ments for antireflection and light trapping, our optimized design with a
double-sided grating structure significantly increases light absorption over
the entire solar spectrum and provides a performance that approaches the
theoretical limit (see Fig. 23E). Furthermore, the double-sided strategy is
a generic approach that can be applied to a range of thicknesses. Fig. 24
Fig. 24 Photocurrents generated by structures as a function of their equivalent thick-

nesses. The black curve is the Yablonovitch limit calculated by integrating Eq. (28) over
the AM 1.5 solar spectrum.
shows that for ultrathin films, our approach can consistently outperform
both the “top-only” and “bottom-only” grating designs. The photocurrents
from the optimized double-sided grating structures are very close to the
theoretical Yablonovitch limit for a range of thicknesses in the few
microns range.
To characterize the loss of real mirror, we replace the PEC mirror by a
flat silver layer for the optimized structure in Fig. 22A. A thin layer of silicon
dioxide is placed between the silicon film and silver as a spacer to reduce
metallic loss. We observe only a small reduction in short-circuit photocur-
rent, from 34.63 to 33.86 mA/cm2 (see Fig. 25A). Since the grating at the
bottom of the silicon layer is sufficiently far from the silver surface, there is
no surface plasmon excitation (see Fig. 25B and C). Therefore, our flat silver
back-reflector induces no significant local or long-range surface plasmon
excitation and its resultant substantial parasitic loss (Biswas and Xu, 2011).
When we extend our optimization to include nanocone gratings made of
nonabsorbing dielectric materials instead of silicon, we observe similar geo-
metrical configurations and absorption enhancements. Therefore, the strat-
egy has great flexibility in the nanocone grating design regarding material
choice for either the grating or the back-reflector.
As a final remark, the doubled-sided structure could be fabricated
by applying the Langmuir–Blodgett (LB) assembly method of silica
nanospheres and the reactive ion etching (RIE) to each side sequentially
(Hsu et al., 2008; Zhu et al., 2009, 2010a). One could use the size of the
silica nanoparticles to control the periodicity, and the RIE to control the
shapes of the nanocones. This fabrication process could also be applied to
Fig. 25 (A) Comparison of the absorption spectra with PEC and with silver (plus a spacer
between the nanostructure and the silver back-reflector). The red curve is the absorption
spectrum of the structure in Fig. 22A, and the blue curve is the absorption spectrum of
the same structure, except that the PEC is replaced with real silver (Palik, 1997), and a
silicon dioxide layer of thickness 2 μm is placed between silicon and silver as a spacer.
(B) and (C) The double-sided grating structure and the electric field profile at the wave-
length of 950 nm.
a large scale (Hsu et al., 2008; Zhu et al., 2010a). Since our structure involves
only a slight surface modification of a silicon film with nanocones of rela-
tively low aspect ratio, the issues associated with surface recombination
and degradation of electronic properties should be less severe than those
of other high aspect ratio nanoscale structures. The double-sided nanocone
grating design provides an experimentally realistic strategy in efficiency
improvement and cost reduction for c-Si solar cells. Although we focus
on c-Si, many of the strategies can be applied to other materials (Liang
et al., 2012).
7. Conclusion
Light trapping is desirable for weak absorbers because the single-pass
absorption is usually too small for a significant percentage of light to be
absorbed by the solar absorber. In this chapter, we describe a statistical
TCMT formalism to construct a wave optics light trapping theory, which
reduces to the conventional light trapping theory in the ray optics regime.
We show that the DOS, especially the van Hove singularities, of photonic
crystals strongly influences the light trapping behaviors. We validate the
wave optics light trapping theory and show that nanophotonic light trapping
performance can surpass the conventional benchmarks, such as the
Yablonovitch 4n2 limit, even after a complete angle integration. We apply
light trapping with antireflection to practical photovoltaic cells. Using
jointly optimized nanocone gratings on both sides of ultrathin c-Si solar
cells, we showed that it is possible to reach the 4n2 limit. These results pro-
vide a theoretical guidance for the design of optical absorbers that are
inhomogeneous in general, including nanostructured photovoltaic and
photoelectrochemical cells. We show that to design and experimentally real-
ize high-efficiency photonic crystal solar cells, one should include consider-
ations on the DOS including the effects of the effective medium and van
Hove singularities, the overlap factor, the mode coupling, and antireflection
designs.
References
Agrawal, M., Peumans, P., 2008. Broadband optical absorption enhancement through
coherent light trapping in thin-film photovoltaic cells. Opt. Express 16 (8), 5385–5396.
Almeida, V.R., Xu, Q., Barrios, C.A., Lipson, M., 2004. Guiding and confining light in void
nanostructure. Opt. Lett. 29 (11), 1209–1211.
Atwater, H.A., Polman, A., 2010. Plasmonics for improved photovoltaic devices. Nat.
Mater. 9, 205–213.
Barnett, S.M., Loudon, R., 1996. Sum rule for modified spontaneous emission rates. Phys.
Rev. Lett. 77 (12), 2444–2446.
Bassani, F., Parravicini, G.P., 1975. Electronic States and Optical Transistions in Solids.
Pergamon Press.
Battaglia, C., Hsu, C.-M., S€ oderstr€om, K., Escarre, J., Haug, F.-J., Charrière, M.,
Boccard, M., Despeisse, M., Alexander, D.T.L., Cantoni, M., Yi, C., Ballif, C.,
2012a. Light trapping in solar cells: can periodic beat random? ACS Nano 6 (3),
2790–2797.
Battaglia, C., Boccard, M., Haug, F.-J., Ballif, C., 2012b. Light trapping in solar cells: when
does a Lambertian scatterer scatter Lambertianly? J. Appl. Phys. 112, 094504.
Bergman, D.J., 1978. The dielectric constant of a composite material—a problem in classical
physics. Phys. Rep. 43 (9), 377–407.
Bergman, D.J., 1980. Exactly solvable microscopic geometries and rigorous bounds for the
complex dielectric constant of a two-component composite material. Phys. Rev. Lett.
44, 1285.
Bermel, P., Luo, C., Zeng, L., Kimerling, L.C., Joannopoulos, J.D., 2007. Improving thin-
film crystalline silicon solar cell efficiencies with photonic crystals. Opt. Express 15 (25),
16986–17000.
Biswas, R., Xu, C., 2011. Nano-crystalline silicon solar cell architecture with absorption at
the classical 4n2 limit. Opt. Express 19 (S4), A664–A672.
Bozzola, A., Liscidini, M., Andreani, L.C., 2012. Photonic light-trapping versus Lambertian
limits in thin film silicon solar cells with 1D and 2D periodic patterns. Opt. Express
20 (S2), A224–A244.
Brongersma, M.L., Cui, Y., Fan, S., 2014. Light management for photovoltaics using high-
index nanostructures. Nat. Mater. 13, 451–460.
Callahan, D.M., Munday, J.N., Atwater, H.A., 2012. Solar cell light trapping beyond the ray
optic limit. Nano Lett. 12, 214–218.
Chutinan, A., John, S., 2008. Light trapping and absorption optimization in certain thin-film
photonic crystal architectures. Phys. Rev. A 78, 023825.
Chutinan, A., Kherani, N.P., Zukotynski, S., 2009. High-efficiency photonic crystal solar
cell architecture. Opt. Express 17 (11), 8871–8878.
Deckman, H.W., Roxlo, C.B., Yablonovitch, E., 1983. Maximum statistical increase of
optical absorption in textured semiconductor films. Opt. Lett. 8 (9), 491–493.
Dewan, R., Vasilev, I., Jovanov, V., Knipp, D., 2011. Optical enhancement and losses of
pyramid textured thin-film silicon solar cells. J. Appl. Phys. 110, 013101.
Du, Q.G., Kam, C.H., Demir, H.V., Yu, H.Y., Sun, X.W., 2011. Enhanced optical absorp-
tion in nanopatterned silicon thin films with a nano-cone-hole structure for photovoltaic
applications. Opt. Lett. 36 (9), 1713–1715.
Duche, D., Escoubas, L., Simon, J.-J., Torchio, P., Vervisch, W., Flory, F., 2008. Slow
Bloch modes for enhancing the absorption of light in thin films for photovoltaic cells.
Duche, D., Drouard, E., Simon, J.J., Escoubas, L., Torchio, P., Le Rouzo, J., Vedraine, S.,
2011. Light harvesting in organic solar cells. Sol. Energy Mater. Sol. Cells 95, S18–S25.
Eyderman, S., John, S., Deinega, A., 2013. Solar light trapping in slanted conical-pore pho-
tonic crystals: beyond statistical ray trapping. J. Appl. Phys. 113, 154315.
Fahr, S., Kirchartz, T., Rockstuhl, C., Lederer, F., 2011. Approaching the Lambertian limit
in randomly textured thin-film solar cells. Opt. Express 19 (S4), A865–A874.
Phys. Rev. B 65, 235112.
Ganapati, V., Miller, O.D., Yablonovitch, E., 2014. Light trapping textures designed by elec-
tromagnetic optimization for subwavelength thick solar cells. IEEE J. Photovoltaics 4 (1),
175.
Gaponenko, S.V., 2010. Introduction to Nanophotonics. Cambridge University Press.
Gjessing, J., Sudbø, A.S., Marstein, E.S., 2011. Comparison of periodic light-trapping struc-
tures in thin crystalline silicon solar cells. J. Appl. Phys. 110, 033104.
Goetzberger, A., 1981. Optical confinement in thin Si-solar cells by diffuse back reflectors.
In: 15th Photovoltaic Specialists Conference. IEEE, pp. 867–870.
Gomard, G., Meng, X., Drouard, E., El Hajjam, K., Gerelli, E., Peretti, R., Fave, A.,
Orobtchouk, R., Lemiti, M., Seassal, C., 2012. Light harvesting by planar photonic crys-
tals in solar cells: the case of amorphous silicon. J. Opt. 14 (2), 024011.
Gomard, G., Peretti, R., Drouard, E., Meng, X., Seassal, C., 2013. Photonic crystals and
optical mode engineering for thin film photovoltaics. Opt. Express 21 (8), A515.
Grandidier, J., Callahan, D.M., Munday, J.N., Atwater, H.A., 2011. Light absorption
enhancement in thin-film solar cells using whispering gallery modes in dielectric
nanospheres. Adv. Mater. 23, 1272–1276.
Green, M.A., 2002. Lambertian light trapping in textured solar cells and light-emitting
diodes: analytical solutions. Prog. Photovolt. Res. Appl. 10 (4), 235–241.
Green, M.A., 2011. Enhanced evanescent mode light trapping in organic solar cells and other
low index optoelectronic devices. Prog. Photovolt. Res. Appl. 19 (4), 473–477.
Guldin, S., H€ uttner, S., Kolle, M., Welland, M.E., M€ uller-Buschbaum, P., Friend, R.H.,
Steiner, U., Tetreault, N., 2010. Dye-sensitized solar cell based on a three-dimensional
photonic crystal. Nano Lett. 10 (7), 2303–2309.
Han, S.E., Chen, G., 2010. Toward the Lambertian limit of light trapping in thin nanostruc-
tured silicon solar cells. Nano Lett. 10 (11), 4692–4696.
Haug, F.-J., S€ oderstr€
om, K., Naqavi, A., Ballif, C., 2011. Resonances and absorption
enhancement in thin film silicon solar cells with periodic interface texture. J. Appl. Phys.
109, 084516.
Heine, C., Morf, R.H., 1995. Submicrometer gratings for solar energy applications. Appl.
Opt. 34 (14), 2476–2482.
Ho, K.M., Chan, C.T., Soukoulis, C.M., Biswas, R., Sigalas, M., 1994. Photonic band gaps
in three dimensions: new layer-by-layer periodic structures. Solid State Commun. 89 (5),
413–416.
Hoppe, H., Sariciftci, N.S., 2004. Organic solar cells: an overview. J. Mater. Res. 19 (7),
1924–1945.
Hsu, C.-M., Connor, S.T., Tang, M.X., Cui, Y., 2008. Wafer-scale silicon nanopillars and
nanocones by Langmuir-Blodgett assembly and etching. Appl. Phys. Lett. 93, 133109.
Huynh, W.U., Dittmer, J.J., Alivisatos, P., 2002. Hybrid nanorod-polymer solar cells.
Science 295 (5564), 2425–2427.
John, S., 2012. Why trap light? Nat. Mater. 11, 997–999.
Johnson, S.G., Joannopoulos, J.D., 2001. Block-iterative frequency-domain methods for
Maxwell’s equations in a planewave basis. Opt. Express 8 (3), 173–190.
Kayes, B.M., Atwater, H.A., Lewis, N.S., 2005. Comparison of the device physics principles
of planar and radial p-n junction nanorod solar cells. J. Appl. Phys. 97 (11), 114302.
Kelzenberg, M.D., Boettcher, S.W., Petykiewicz, J.A., Turner-Evans, D.B., Putnam, M.C.,
Warren, E.L., Spurgeon, J.M., Briggs, R.M., Lewis, N.S., Atwater, H.A., 2010.
Enhanced absorption and carrier collection in Si wire arrays for photovoltaic applica-
tions. Nat. Mater. 9, 239–244.
Ko, D.-H., Tumbleston, J.R., Zhang, L., Williams, S., DeSimone, J.M., Lopez, R.,
Samulski, E.T., 2009. Photonic crystal geometry for organic solar cells. Nano Lett.
9 (7), 2742–2746.
Kuang, P., Deinega, A., Hsieh, M.-L., John, S., Lin, S.-Y., 2013. Light trapping and near-
unity solar absorption in a three-dimensional photonic-crystal. Opt. Lett. 38 (20),
4200–4203.
Law, M., Greene, L.E., Johnson, J.C., Saykally, R., Yang, P., 2005. Nanowire dye-sensitized
solar cells. Nat. Mater. 4 (6), 455.
Le, K.Q., John, S., 2014. Synergistic plasmonic and photonic crystal light-trapping: architec-
tures for optical up-conversion in thin-film solar cells. Opt. Express 22 (S1), A1–A12.
Li, L., 1997. New formulation of the Fourier modal method for crossed surface-relief grat-
ings. J. Opt. Soc. Am. A 14 (10), 2758–2767.
Li, J., Yu, H.Y., Wong, S.M., Zhang, G., Lo, G.-Q., Kwong, D.-L., 2010. Si nanocone array
optimization on crystalline Si thin films for solar energy harvesting. J. Phys. D. Appl.
Phys. 43 (25), 255101.
Liang, D., Huo, Y., Kang, Y., Wang, K.X., Gu, A., Tan, M., Yu, Z., Li, S., Jia, J., Bao, X.,
Wang, S., Yao, Y., Wong, H.-S.P., Fan, S., Cui, Y., Harris, J.S., 2012. Optical absorp-
tion enhancement in freestanding GaAs thin film nanopyramid arrays. Adv. Energy
Mater. 2 (10), 1254–1260.
Lin, C., Chen, C., Sharkawy, A., Schneider, G.J., Venkataraman, S., Prather, D.W., 2005.
Efficient terahertz coupling lens based on planar photonic crystals on silicon on insulator.
Opt. Lett. 30 (11), 1330–1332.
Liu, V., Fan, S., 2011. Efficient computation of equifrequency surfaces and density of states in
photonic crystals using Dirichlet-to-Neumann maps. J. Opt. Soc. Am. B 28 (8), 1837.
Liu, V., Fan, S., 2012. S4: a free electromagnetic solver for layered periodic structures. Com-
put. Phys. Commun. 183, 2233–2244.
Mallick, S.B., Agrawal, M., Peumans, P., 2010. Optimal light trapping in ultra-thin photonic
crystal crystalline silicon solar cells. Opt. Express 18 (6), 5691–5706.
Mallick, S.B., Agrawal, M., Wangperawong, A., Barnard, E.S., Singh, K.K., Visser, R.J.,
Brongersma, M.L., Peumans, P., 2012. Ultrathin crystalline-silicon solar cells with
embedded photonic crystals. Appl. Phys. Lett. 100, 053113.
Mann, S.A., Garnett, E.C., 2013. Extreme light absorption in thin semiconductor films
wrapped around metal nanowires. Nano Lett. 13, 3173–3178.
Mayer, A.C., Scully, S.R., Hardin, B.E., Rowell, M.W., McGehee, M.D., 2007. Polymer-
based solar cells. Mater. Today 10 (11), 28–33.
McPhedran, R.C., Botten, L.C., McOrist, J., Asatryan, A.A., de Sterke, C.M.,
Nicorovici, N.A., 2004. Density of states functions for photonic crystals. Phys. Rev.
E 69, 016609.
Mihi, A., Mı́guez, H., 2005. Origin of light-harvesting enhancement in colloidal-photonic-
crystal-based dye-sensitized solar cells. J. Phys. Chem. B 109 (33), 15968–15976.
Milton, G.W., 1981. Bounds on the complex permittivity of a two-component composite
material. J. Appl. Phys. 52, 5286.
Milton, G.W., 2002. The Theory of Composites. Cambridge University Press.
Mokkapati, S., Catchpole, K.R., 2012. Nanophotonic light trapping in solar cells. J. Appl.
Phys. 112, 101101.
Munday, J.N., 2012. The effect of photonic bandgap materials on the shockley-queisser
limit. J. Appl. Phys. 112 (6), 064501.
Munday, J.N., Callahan, D.M., Atwater, H.A., 2012. Light trapping beyond the 4n2 limit in
thin waveguides. Appl. Phys. Lett. 100 (121121).
Mutitu, J.G., Shi, S., Chen, C., Creazzo, T., Barnett, A., Honsberg, C., Prather, D.W., 2008.
Thin film silicon solar cell design based on photonic crystal and diffractive grating struc-
tures. Opt. Express 16 (19), 15238–15248.
Nagel, J.R., Scarpulla, M.A., 2010. Enhanced absorption in optically thin solar cells by scat-
tering from embedded dielectric nanoparticles. Opt. Express 18 (S2), A139–A146.
Naqavi, A., Haug, F.-J., Battaglia, C., Herzig, H.P., Ballif, C., 2013a. Light trapping in solar
cells at the extreme coupling limit. J. Opt. Soc. Am. B 30 (1), 13.
Naqavi, A., Haug, F.-J., Ballif, C., Scharf, T., Herzig, H.P., 2013b. Limit of light coupling
strength in solar cells. Appl. Phys. Lett. 102, 131113.
Naqavi, A., Haug, F.-J., S€ oderstr€
om, K., Battaglia, C., Paeder, V., Scharf, T., Herzig, H.P.,
Ballif, C., 2014. Angular behavior of the absorption limit in thin film silicon solar cells.
Prog. Photovolt. Res. Appl. 22 (11), 1147–1158.
Narasimhan, V.K., Cui, Y., 2013. Nanostructures for photon management in solar cells.
Nanophotonics 2 (3), 187–210.
Oskooi, A., Tanaka, Y., Noda, S., 2014. Tandem photonic-crystal thin films surpassing Lam-
bertian light-trapping limit over broad bandwidth and angular range. Appl. Phys. Lett.
104, 091121.
Paetzold, U.W., Moulin, E., Michaelis, D., B€ ottler, W., W€achter, C., Hagemann, V.,
Meier, M., Carius, R., Rau, U., 2011. Plasmonic reflection grating back contacts for
microcrystalline silicon solar cells. Appl. Phys. Lett. 99, 181105.
Pala, R.A., Liu, J.S.Q., Barnard, E.S., Askarov, D., Garnett, E.C., Fan, S.,
Brongersma, M.L., 2013. Optimization of non-periodic plasmonic light-trapping layers
for thin-film solar cells. Nat. Commun. 4, 2095.
Palik, E.D. (Ed.), 1997. Handbook of Optical Constants of Solids. Elsevier Inc.
Park, Y., Drouard, E., El Daif, O., Letartre, X., Viktorovitch, P., Fave, A., Kaminski, A.,
Lemiti, M., Seassal, C., 2009. Absorption enhancement using photonic crystals for sil-
icon thin film solar cells. Opt. Express 17 (16), 14312–14321.
Peretti, R., Gomard, G., Seassal, C., Letartre, X., Drouard, E., 2012. Modal approach for
tailoring the absorption in a photonic crystal membrane. J. Appl. Phys. 111, 123114.
Polman, A., Atwater, H.A., 2012. Photonic design principles for ultrahigh-efficiency pho-
tovoltaics. Nat. Mater. 11, 174–177.
Raut, H.K., Ganesh, V.A., Nair, A.S., Ramakrishna, S., 2011. Anti-reflective coatings: a
critical, in-depth review. Energy Environ. Sci. 4, 3779.
Saeta, P.N., Ferry, V.E., Pacifici, D., Munday, J.N., Atwater, H.A., 2009. How much can
guided modes enhance absorption in thin solar cells? Opt. Express 17 (23), 20975–20990.
Sandhu, S., Yu, Z., Fan, S., 2013. Detailed balance analysis of nanophotonic solar cells. Opt.
Express 21 (1), 1209–1217.
Scheel, S., 2008. Sum rule for local densities of states in absorbing dielectrics. Phys. Rev. A
78 (013841).
Sheng, P., Bloch, A.N., Stepleman, R.S., 1983. Wavelength-selective absorption enhance-
ment in thin-film solar cells. Appl. Phys. Lett. 43 (6), 579.
Sheng, X., Johnson, S.G., Michel, J., Kimerling, L.C., 2011. Optimization-based design of
surface textures for thin-film Si solar cells. Opt. Express 19 (S4), A841–A850.
S€
oz€uer, H.S., Dowling, J.P., 1994. Photonic band calculations for woodpile structures.
J. Mod. Opt. 41 (2), 231–239.
Stuart, H.R., Hall, D.G., 1997. Thermodynamic limit to light trapping in thin planar struc-
tures. J. Opt. Soc. Am. A 14 (11), 3001–3008.
66, 045102.
Tobias, I., Luque, A., Marti, A., 2008. Light intensity enhancement by diffracting structures
in solar cells. J. Appl. Phys. 104, (3), 034502.
Tsakalakos, L., Balch, J., Fronheiser, J., Shih, M.-Y., LeBoeuf, S.F., Pietrzykowski, M.,
Codella, P.J., Korevaar, B.A., Sulima, O.V., Rand, J., Davuluru, A., Rapol, U.,
2007. Strong broadband optical absorption in silicon nanowire films.
J. Nanophotonics 1 (1), 013552.
Tumbleston, J.R., Ko, D.-H., Samulski, E.T., Lopez, R., 2009a. Electrophotonic enhance-
ment of bulk heterojunction organic solar cells through photonic crystal photoactive
layer. Appl. Phys. Lett. 94, 043305.
Tumbleston, J.R., Ko, D.-H., Samulski, E.T., Lopez, R., 2009b. Absorption and
quasiguided mode analysis of organic solar cells with photonic crystal photoactive layers.
Opt. Express 17 (9), 7670–7681.
Van Hove, L., 1953. The occurence of singularities in the elastic frequency distribution of a
crystal. Phys. Rev. 89, 1189.
Vynck, K., Burresi, M., Riboli, F., Wiersma, D.S., 2012. Photon management in two-
dimensional disordered media. Nat. Mater. 11, 1017.
Wang, P., Menon, R., 2014. Optimization of generalized dielectric nanostructures for
enhanced light trapping in thin-film photovoltaics via boosting the local density of opti-
cal states. Opt. Express 22 (S1), A99–A110.
Wang, W., Wu, S., Reinhardt, K., Lu, Y., Chen, S., 2010. Broadband light absorption
enhancement in thin-film silicon solar cells. Nano Lett. 10 (6), 2012–2018.
Wang, K.X., Yu, Z., Liu, V., Cui, Y., Fan, S., 2012. Absorption enhancement in ultrathin
crystalline silicon solar cells with antireflection and light-trapping nanocone gratings.
Nano Lett. 12 (3), 1616–1619.
Wang, S., Weil, B.D., Li, Y., Wang, K.X., Garnett, E., Fan, S., Yi, C., 2013a. Large-area
free-standing ultrathin single-crystal silicon as processable materials. Nano Lett. 13 (9),
4393–4398.
Wang, C., Yu, S., Chen, W., Sun, C., 2013b. Highly efficient light-trapping structure design
inspired by natural evolution. Sci. Rep. 3, 1025.
Wang, K.X., Yu, Z., Liu, V., Raman, A., Cui, Y., Fan, S., 2014a. Light trapping in photonic
crystals. Energy Environ. Sci. 7, 2725–2738.
Wang, K.X., Zongfu, Y., Liu, V., Brongersma, M.L., Jaramillo, T.F., Fan, S., 2014b. Nearly
total solar absorption in ultrathin nanostructured iron oxide for efficient photo-
electrochemical water splitting. ACS Photonics 1 (3), 235–240.
Wiesendanger, S., Zilk, M., Pertsch, T., Rockstuhl, C., Lederer, F., 2013. Combining ran-
domly textured surfaces and photonic crystals for the photon management in thin film
microcrystalline silicon solar cells. Opt. Express 21 (103), A450–A459.
Wong, B.M., Morales, A.M., 2009. Enhanced photocurrent efficiency of a carbon nanotube
p-n junction electromagnetically coupld to a photonic structure. J. Phys. D. Appl. Phys.
42, 055111.
Yablonovitch, E., 1982. Statistical ray optics. J. Opt. Soc. Am. 72 (7), 899–907.
Yablonovitch, E., Cody, G.D., 1982. Intensity enhancement in textured optical sheets for
solar cells. IEEE Trans. Electron Devices ED-29 (2), 300.
Yu, Z., Fan, S., 2011. Angular constraint on light-trapping absorption enhancement in solar
cells. Appl. Phys. Lett. 98, 011106.
Yu, G., Gao, J., Hummelen, J.C., Wudl, F., JHeeger, A., 1995. Polymer photovoltaic cells:
enhanced efficiencies via a network of internal donor-acceptor heterojunctions. Science
270 (5243), 1789–1791.
Yu, Z., Raman, A., Fan, S., 2010a. Fundamental limit of light trapping in solar cells. Proc.
Natl. Acad. Sci. U. S. A. 107 (41), 17491–17496.
Yu, Z., Raman, A., Fan, S., 2010b. Fundamental limit of light trapping in grating structures.
Opt. Express 18 (S3), A366.
Yu, Z., Raman, A., Fan, S., 2012. Thermodynamic upper bound on broadband light cou-
pling with photonic structures. Phys. Rev. Lett. 109, 173901.
Yu, Z., Sandhu, S., Fan, S., 2014. Efficiency above the Shockley-Queisser limit by using
nanophotonic effects to create multiple effective bandgaps with a single semiconductor.
Nano Lett. 14 (1), 66–70.
Zeman, M., Isabella, O., Solntsev, S., J€ager, K., 2013. Modeling of thin-film silicon solar
cells. Sol. Energy Mater. Sol. Cells 119, 94–111.
Zhang, R.Y., Shao, B., Dong, J.R., Zhang, J.C., Yang, H., 2011. Absorption enhancement
analysis of crystallline Si thin film solar cells based on broadband antireflection nanocone
grating. J. Appl. Phys. 110, 113105.
Zhu, J., Yu, Z., Burkhard, G.F., Hsu, C.-M., Connor, S.T., Xu, Y., Wang, Q.,
McGehee, M., Fan, S., Cui, Y., 2009. Optical absorption enhancement in amorphous
silicon nanowire and nanocone arrays. Nano Lett. 9 (1), 279–282.
Zhu, J., Yu, Z., Fan, S., Cui, Y., 2010a. Nanostructured photon management for high per-
formance solar cells. Mater. Sci. Eng. R 70, 330–340.
Zhu, J., Hsu, C.-M., Yu, Z., Fan, S., Cui, Y., 2010b. Nanodome solar cells with efficient
light management and self-cleaning. Nano Lett. 10 (6), 1979–1984.
CHAPTER FOUR
Optical image processing

using photonic crystal slab
Cheng Guoa, Shanhui Fanb,*
a
Department of Applied Physics, Stanford University, Stanford, CA, United States
b
Ginzton Laboratory and Department of Electrical Engineering, Stanford University, Stanford,
CA, United States
*Corresponding author: e-mail address: shanhui@stanford.edu
Contents
1. Introduction 93
2. Filtering in the wavevector domain using guided resonances 95
3. Theoretical analysis of isotropic filters using a photonic crystal slab 97
4. Numerical demonstration of the isotropic filters 104
4.1 Isotropic high-pass filter (the Laplacian) 105
4.2 Isotropic low-pass filter 105
4.3 Isotropic band-reject filter 108
4.4 Isotropic band-pass filter 108
5. Conclusion 111
Acknowledgments 111
References 112
1. Introduction
In this chapter we will discuss the use of photonic crystal slab to
accomplish a number of imaging processing tasks, including edge detection,
image smoothing, white noise suppression, and suppression or extraction of
periodic features. All these tasks involve filtering in the wavevector domain.
These tasks are of broad applications ranging from microscopy and medical
imaging to industrial inspection and object detection (Abràmoff et al., 2004;
Brosnan and Sun, 2004; Dalal and Triggs, 2005; Gonzalez and Woods, 2008;
Markham et al., 1963).
All these image processing operations, as mentioned above, can certainly
be implemented using conventional electronic computing. However,
nowadays there are many big-data applications that require real-time and
high-throughput image processing, for which it is of interest to develop

94 Cheng Guo and Shanhui Fan
techniques to perform these computations with reduced energy consump-

tion and increased speed. Optical image processing is particularly suitable
for this purpose, since the images are naturally in the optical domain. More-
over, many of these tasks can be achieved using linear optical devices that
have no power consumption except the propagation loss of light through
the device, and have very low latency since the latency is determined only
by the time for light to propagate through these devices.
Traditionally, optical image processing can be performed using the tech-
niques of Fourier optics. For instance, to achieve filtering in the wavevector
domain, one first obtains a Fourier transformation of an image on a Fourier
plane by passing the image through a lens. One then performs a spatial fil-
tering on the Fourier plane, followed by a Fourier transformation again
through a second lens (Goodman, 2005; Yu et al., 2001). Such technique
requires long propagation distance and therefore results in a bulky system
(Chang et al., 1990; Eu et al., 1973; G€ orlitz and Lanzl, 1975; Sirohi and
Mohan, 1977).
Thanks to the recent progress in nanophotonics, optical image processing
can instead be achieved by compact photonic devices (AbdollahRamezani
et al., 2015; Bykov et al., 2014, 2018; Chizari et al., 2016; Dong et al.,
2018; Fang and Ruan, 2018; Fang et al., 2017; Golovastikov et al., 2014;
Guo et al., 2018b; Hwang et al., 2018; Kwon et al., 2018; Nejad and
Fleury, 2018; Pors et al., 2015; Roberts et al., 2018; Silva et al., 2014; Solli
and Jalali, 2015; Wu et al., 2017; Youssefi et al., 2016; Zhu et al., 2017).
In this chapter, we provide a brief review of optical image processing using
the guided resonance in a two-dimensional photonic crystal slab device. In
particular, we focus on a particular class of photonic crystal slab structure
where the guided resonance near the Γ point exhibits an isotropic band
structure. Depending on the light frequency and the choice of transmission
or reflection mode, this compact device realizes high-pass, low-pass, band-
reject, and band-pass filtering directly in the optical domain. These filtering
operations are applied isotropically in the two-dimensional plane, which is
particularly attractive for image processing.
The rest of this chapter is organized as follows. In Section 2 we elucidate
the underlying physics mechanism of filtering in the wavevector domain
using guided resonances. This mechanism underlies the operation of many
compact nanophotonic devices for image filtering, including ours. In
Section 3 we theoretically analyze how to realize isotropic filters by a pho-
tonic crystal slab device using guided resonances with carefully designed
isotropic band structures. In Section 4 we numerically demonstrate the
isotropic filtering functions of the device. We conclude in Section 5.
Optical image processing using photonic crystal slab 95
2. Filtering in the wavevector domain using guided

resonances
We consider a filter in the wavevector domain as characterized by
the filter transfer function H(kx, ky) (Bracewell, 1986). In general, for a
near-normally incident light beam along the z axis with a transverse field
profile Sin(x, y), upon the operation of the filter, the output beam has a pro-

file Sout ðx,yÞ ¼ F 1 ½Hðkx , ky ÞS in ðkx , ky Þ, where F 1 denotes the inverse

Fourier transform and S in ðkx , ky Þ is the Fourier transform of the input image
(Gonzalez and Woods, 2008). If H(kx, ky) ¼ H(jkj), i.e., H only depends on
the magnitude of the wavevector jkj, the filter is isotropic. Here k ¼ (kx, ky)
refers to the in-plane wavevector.
Filtering in the wavevector domain can be realized using guided reso-
nances hosted by compact photonic device. In general, guided resonances
can induce sharp Fano resonance features in the transmission and reflection
spectra (Fan and Joannopoulos, 2002; Ochiai and Sakoda, 2001; Tikhodeev
et al., 2002; Zhou et al., 2014). Since the band structures of the guided
resonances are usually dispersive, i.e., the resonant frequency is wavevector
dependent, the transmission and reflection spectra are also wavevector
dependent. Filtering in the wavevector domain can thus be realized in both
transmission and reflection modes at a specific frequency.
As an illustration, we first show how to use guided resonances to realize
second-order differentiation, a specific type of high-pass filter, at trans-
mission along the direction normal to the slab (Guo et al., 2018b). The
task is to design an optical system with a transfer function in the wave-
vector space (k-space) of the form t(k) ¼ αjkj2. For one-dimensional
differentiation, k ¼ kx; for two-dimensional differentiation, k ¼ (kx, ky).
We consider the transmission coefficient of a photonic structure which sup-
ports a single band of guided resonances as characterized by k-dependent
resonant frequencies ω(k) and radiative linewidths γ(k) (Fig. 1A). Near
the resonant frequencies, the transmitted amplitude t is expressed as
(Fan and Joannopoulos, 2002)
γðkÞ
tðω, kÞ ¼ td + f , (1)
i½ω ωðkÞ + γðkÞ
where ω is the incident light frequency, td is the direct transmission coeffi-

cient, and f is related to the complex decaying amplitude of the resonance to
the transmission side of the slab.
Fig. 1 Operating principle of filtering in the wavevector domain using guided reso-
nances. (A) Dispersion for a single band of guided resonance ω(k). (B) Transmission
spectrum t(ω) at k ¼ 0. (C) k-dependent transmission coefficient t(k) at the resonant
frequency at k ¼ 0: ω ¼ ω0. (D) Transmission spectrum t(ω) at k 6¼ 0.
In general, f is constrained by the direct process if the system satisfies

energy conservation and time-reversal symmetry (Fan and Joannopoulos,
2002; Wang et al., 2013). In particular, if td ¼ 1, that is, the direct pathway
has a 100% transmission coefficient, then
f ¼ td ¼ 1 (2)
as has been derived in Wang et al. (2013). In this special case,
γðkÞ
tðω, kÞ ¼ 1 : (3)
i½ω ωðkÞ + γðkÞ
We plot the spectra of transmission coefficient amplitude t(ω) at k ¼ 0 and

k 6¼ 0 in Fig. 1B and D, respectively. Using Eq. (3), zero transmission occurs
at the k-dependent resonant frequency:
ω ¼ ωðkÞ: (4)
We denote ω0 ¼ ω(k ¼ 0) and γ 0 ¼ γ(k ¼ 0). If we fix the frequency ω ¼ ω0,

the transmission coefficient t(ω0, k) is then k-dependent. We can perform
an expansion of t near k ¼ 0:

∂t ∂t
tðω0 , kÞ ¼ 0 + δωðkÞ + δγðkÞ, (5)
∂ωðkÞΓ ∂γðkÞΓ
where
δωðkÞ ¼ ωðkÞ ω0 , δγðkÞ ¼ γðkÞ γ 0 , (6)

∂t i ∂t
¼ , ¼ 0, (7)
∂ωðkÞΓ γ0 ∂γðkÞΓ
therefore,
i
tðω0 , kÞ ¼ δωðkÞ: (8)
γ0
In this special case, t(k) is simply proportional to the band dispersion δω(k)
near k ¼ 0. If δωðkÞ∝jkj2 , then tðkÞ∝ jkj2 as well.
The analysis above indicates that to use guided resonances to perform
second-order differentiation, it is sufficient to satisfy the following three
conditions:
1. td ¼ 1.
2. Only one guided resonance band is coupled.
3. The band satisfies the dispersion ðωðkÞ ω0 Þ∝ jkj2 .
While the discussion above is specific to differentiation, it is relevant for
understanding of some of the other filter functionalities as well. Based
on the discussion here, in the next section we will illustrate how to use a
photonic crystal slab to achieve these filter functionalities.
3. Theoretical analysis of isotropic filters using

a photonic crystal slab
In this section, based on the discussion above, we theoretically analyze
how to realize isotropic filters using a photonic crystal slab device. Depending
on the light frequency and the choice of transmission or reflection mode,
this compact device realizes isotropic high-pass, low-pass, band-reject, and
band-pass filtering directly in the optical domain (Guo et al., 2018a).
The device is shown in Fig. 2. The dielectric constant of the material
for the slabs is E ¼ 4.67, which approximates that of Si3N4 in the visible wave-
length range. Here we use a dielectric constant that is real, as the extinction
coefficient of Si3N4 at visible wavelength is negligible. The photonic crystal
slab has a thickness of d ¼ 0.50a and a square lattice of air holes with
Fig. 2 Geometry of the device, which consists of a photonic crystal slab with a square
lattice of air holes, separated from a uniform dielectric slab by an air gap. The slabs
have a dielectric constant E ¼ 4.67. The geometry parameters are: d ¼ 0.50a, r ¼ 0.11a,
ds ¼ 0.12a, dg ¼ 1.93a, where a is the lattice constant. The coordinate system is shown
above the device. The red arrow indicates the direction of the incident light. The electric
field directions of the S and P-polarized light are also indicated. Reprinted with permission
from Guo, C., Xiao, M., Minkov, M., Shi, Y., Fan, S., 2018a. Isotropic wavevector domain image
filters by a photonic crystal slab device. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 35 (10),
1685–1691, the Optical Society.
radii r ¼ 0.11a, where a is the lattice constant. The thickness d and radius r are
chosen to realize isotropic bands of guided resonances. A uniform dielectric
slab with a thickness ds ¼ 0.12a is placed in the vicinity of the photonic crystal
slab. The air gap between the two slabs has a thickness dg ¼ 1.93a. As a side
note, the two-layer structure we use in the chapter is quite compact. For
a resonant wavelength λ ¼ 500 nm and a ¼ 650 nm the total thickness is
d + dg + ds ¼ 1.66 μm. Such structure can be fabricated using focused ion
beam-assisted laser interference lithography (Vogelaar et al., 2001).
To satisfy the first criterion in Section 2, we note that the direct trans-
mission coefficient td is related to the nonresonant transmission pathway
(Ochiai and Sakoda, 2001). Hence, it is possible to realize td ¼ 1 by chang-
ing the thickness of the slab. In the structure as shown in Fig. 2, we achieve
td ¼ 1 by placing a uniform dielectric slab in the vicinity of the photon crystal
slab, and by tuning the distance between the slabs. This has the advantage
that we can tune td without significantly affecting the band structure of
the photonic crystal slab.
To satisfy the second and third criterion in Section 2, one will need to
design the band structure of the guided resonances. The design here is in fact
quite nontrivial due to the vectorial nature of electromagnetic waves. Since
the required filter transfer function H(k) is isotropic in k-space, it is natural to
consider a photonic crystal slab structure that possesses rotational symmetry.
As an illustration, here we consider a slab structure with a square lattice of air
holes that has C4v symmetry. For concreteness, we consider circular holes;
the same analysis will apply to other hole shapes that preserve C4v symmetry.
For such a slab, it is known that at the Γ point, which corresponds to jkj ¼ 0,
the only modes that can couple to external plane waves must be twofold
degenerate, belonging to the two-dimensional irreducible representation
of the C4v group (Fan and Joannopoulos, 2002; Ochiai and Sakoda,
2001). Near such modes, in the vicinity of the Γ point, from the k p analysis,
the band structure in general can be described by the following 2 2
effective Hamiltonian. (As we consider only small wavevector range near
the Γ point, it is sufficient to include only the two modes at the Γ point
in our k p analysis.)
^
HðkÞ ¼ ðω0 iγ 0 + ajkj2 ÞÎ + bðk2x k2y Þ^
σ z + ckx ky σ^ x , (9)
where a, b, c are three complex coefficients and the σ’s are the Pauli matrices.
This Hamiltonian has two eigenvalues of
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ω ðkÞ iγ ðkÞ ¼ ω0 iγ 0 + ajkj2 b2 jkj4 + ðc 2 4b2 Þk2x k2y : (10)
For a general choice of parameters b and c, the band structure as described

by Eq. (10) is not isotropic. On the other hand, Eq. (10) also indicates that
when c ¼ 2b, both bands will be isotropic:
ω ðkÞ iγ ðkÞ ¼ ω0 iγ 0 + ða bÞjkj2 : (11)

To achieve such an isotropic band structure requires detailed tuning of
the parameters. Remarkably, with the carefully tuned geometry parameters
as above, both bands of the guided resonances that are degenerate at Γ with a
frequency ω0 ¼ 0.77098 2πc/a exhibit almost complete circular symmetry
in the real part of the eigenfrequencies:
ω ðkÞ ω0 + A jkj2 , (12)

where A+ ¼ 4.35, A ¼ 1.41, and the upper (lower) sign corresponds to

the upper (lower) band. In Eq. (12), and throughout the rest of the chapter,
the wavevector will be in the unit of 2π/a.
The nearly isotropic photonic band structure (ω(k) ω0) for the struc-
ture shown in Fig. 2 is plotted in Fig. 3A, C, and D. Fig. 3A shows that for
both bands, the dispersions along the Γ–M and Γ–X directions have almost
identical effective masses. Here the effective mass tensor is defined as
m*ij ¼ ħ½∂2 ωðkÞ=∂ki ∂kj 1 . Fig. 3C and D shows that the constant frequency
contours for both bands are almost circular. We note that the radiative lin-
ewidths γ (k), unlike ω(k), are anisotropic. Nonetheless, (γ (k) γ 0) are
much smaller than (ω(k) ω0), thus do not affect the circular symmetry of
the transfer functions much, as we will show later.
As we mentioned above, for a structure with C4v symmetry, the modes
that a normally incident plane wave can couple to at the Γ point are always
two-fold degenerate. Consequently, near the Γ point, there are always two
bands of guided resonances present. Moreover, off the normal direction, if
the direction of incident waves is away from the high symmetry planes, both
S and P polarized lights may couple to both bands, leading to complicated
polarization conversion effects. Thus, in general, in addition to having an
anisotropic band structure near Γ, a photonic crystal slab structure also does
not satisfy the criterion 2 above in Section 2 regarding the excitation of a
single guided resonance band. Remarkably, however, below we show that
once the condition for isotropic bands is satisfied, each of the two bands in
fact only couples to one single polarization, for every direction of incidence.
Below, we refer to this effect, where each polarization excites only a single
band away from normal incidence, as the effect of single-band excitation. Due
to the single-band excitation effect, S and P polarizations are decoupled in
this special case.
To illustrate the effect of single-band excitation for every direction when
the bands are isotropic, we first show that single-band excitation always
occurs when the incident direction is in a mirror symmetry plane of the
structure. Using the group theory notation in Sakoda (2005), as our system
possesses C4v symmetry, the doubly degenerate states at Γ are E modes.
Along the Γ–X direction, this pair of E modes splits into two singly degen-
erate states of A and B modes which are even and odd, respectively, with
respect to the reflection operator associated with the mirror plane y ¼ 0.
On the other hand, since the S and P polarized lights are odd and even with
respect to the same mirror plane, respectively, along the Γ–X direction, the
S(P) polarized light can only couple to the B(A) mode. Thus, in general, we
Fig. 3 (A, C, D) Nearly isotropic photonic band structure of the photonic crystal slab
shown in Fig. 1 near the frequency ω0 ¼ 0.77098 2πc/a. (A) Band dispersions along
Γ–X and Γ–M directions. (C) Constant frequency contours of the lower band.
(D) Constant frequency contours of the upper band. (B) Scheme of multiple filtering func-
tions of the device. When the light frequency is on resonance at the normal incidence
(ω ¼ ω0, labeled by red arrows in (B) and red horizontal lines in (E,F)), the device realizes
isotropic high-pass filter (Laplacian) at transmission and low-pass filter at reflection. When
the light frequency is detuned from the resonance at the normal incidence (ω ¼ ωk,
labeled by green arrows in (B) and green horizontal lines in (E,F)), the device realizes isotro-
pic band-reject filter at transmission and band-pass filter at reflection. (E) Transmittance jtj
for S polarized light as a function of ω and jkj near ω0 ¼ 0.77098 2πc/a along a general
wavevector direction (φ ¼ 14°). Due to the isotropic band structure, S light only excites
the upper band, and the transmission spectra are almost identical along any wavevector
direction φ. jtj ¼ 0 when ω ¼ ω(k). (F) Reflectance jrj for S polarized light as a function of
ω and jkj near ω0 ¼ 0.77098 2πc/a along a general wavevector direction (φ ¼ 14°). The
reflection spectra are identical along any wavevector direction. jrj ¼ 1 when ω ¼ ω(k). In all
the plots, frequency (ω ω0) is in the units of 104 2πc/a, while jkj, kx, and ky are in the
units of 103 2π/a. Reprinted with permission from Guo, C., Xiao, M., Minkov, M., Shi, Y.,
Fan, S., 2018a. Isotropic wavevector domain image filters by a photonic crystal slab device.
J. Opt. Soc. Am. A Opt. Image Sci. Vis. 35 (10), 1685–1691, the Optical Society.
have the effect of single-band excitation when the direction of incidence is

in a high symmetry plane such as the y ¼ 0 plane (Tikhodeev et al., 2002).
Next, we prove the following statement: if the two-band Hamiltonian is
isotropic, that is,
^
RðφÞ ^ R
HðkÞ ^ 1 ðφÞ ¼ HðRðφÞkÞ,
^ (13)
for every φ 2 (0, 2π), then we have the effect of single-band excitation along
^
all directions. Here, R(φ) and RðφÞ are the rotation operators that describe
the rotation around the z-axis by an angle φ, in the k space and the two-
dimensional Hilbert space, respectively. To prove this, we denote the
eigenstates of the two bands as jk, Ai and jk, Bi, which connect to the A
and B modes along the Γ-X direction, respectively. Eq. (13) implies that:
^
jRðφÞk, Ai ¼ RðφÞjk, Ai,
(14)
^
jRðφÞk, Bi ¼ RðφÞjk, Bi:
We denote the S and P polarized modes as jk, Si and jk, Pi, respectively.
By definition,
^
jRðφÞk,Si ¼ RðφÞjk, Si,
(15)
^
jRðφÞk,Pi ¼ RðφÞjk, Pi:
For k ¼ kx along the Γ–X direction, the A(B) mode does not couple to the
S(P) polarization, i.e.,
hkx , Sjkx , Ai ¼ 0,
(16)
hkx , Pjkx , Bi ¼ 0:
Then for k ¼ R(φ)kx along any direction, using Eqs. (14) and (15),
we also have
^ 1 ðφÞRðφÞjk
hk,Sjk,Ai ¼ hkx , SjR ^ x , Ai ¼ 0,
^ 1 ðφÞRðφÞjk
^ (17)
hk,Pjk,Bi ¼ hkx , PjR x , Bi ¼ 0:
Due to the single-band excitation effect, as well as the presence of the

uniform dielectric slab which sets the background transmission coefficient
(Fan and Joannopoulos, 2002) to be unity, for this structure shown in
Fig. 2, the transmission and reflection coefficients are:
i½ω ω ðkÞ
t ðω, kÞ ¼ , (18)
i½ω ω ðkÞ + γ ðkÞ
γ ðkÞ
r ðω, kÞ ¼ eiϕ , (19)
i½ω ω ðkÞ + γ ðkÞ
where the upper (lower) sign corresponds to S(P) polarized light and upper
(lower) band; ω is the incident light frequency. Therefore on resonance,
t ðω ðkÞ,kÞ ¼ 0, r ðω ðkÞ, kÞ ¼ eiϕ : (20)
The numerically determined transmission and reflection spectra for S
polarized light are plotted in Fig. 3E and F. Due to the effect of single-band
excitation, S-polarized light only excites the upper band of guided reso-
nances. Moreover, as expected from Eqs. (12), (18), and (19), the resultant
transmission and reflection spectra are isotropic, i.e., the spectra are identical
along any wavevector direction as defined by the angle φ in Fig. 2. When
ω ¼ ω(k), the transmittance exhibits sharp dips with jtj ¼ 0, while the reflec-
tance exhibits peaks with jrj ¼ 1, as expected from Eq. (20).
Depending on the operating conditions, the structure as shown in Fig. 2
can be used to perform several very useful image-processing functionalities
(Guo et al., 2018a):
1. Isotropic high-pass filter
At the frequency ω ¼ ω0 ω(k ¼ 0), the k-dependent transmittance is:
jω ðkÞ ω0 j
jt ðω0 ,kÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
½ω ðkÞ ω0 2 + γ ðkÞ2
(21)
jω ðkÞ ω0 j A 2
¼ jkj :
γ0 γ0
This transmittance realizes the Laplacian, a special isotropic high-pass
filter (Guo et al., 2018b).
2. Isotropic low-pass filter
At the frequency ω ¼ ω0, if one considers instead the reflected light,
the transfer function is:
1
jr ðω0 , kÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 + ½ω ðkÞ ω0 2 =γ ðkÞ2
(22)
1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
1 + A2 jkj4 =γ 20
This transfer function realizes an isotropic low-pass filter with reflection
peak jrj ¼ 1 at the Γ point.
3. Isotropic band-reject filter

Suppose instead we operate away from the frequency ω0, but at the
frequency ω ¼ ω, q, where q is the amplitude of a nonzero in-plane
wavevector, the transfer function then becomes
jω ðkÞ ω, q j
jt ðω, q ,kÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (23)
½ω ðkÞ ω, q 2 + γ ðkÞ2
This transfer function realizes an isotropic band-reject filter, with
transmission dip jtj ¼ 0 at jkj ¼ q.
4. Isotropic band-pass filter
At the frequency ω ¼ ω, q, the reflection has a transfer function
1
jr ðωq ,kÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (24)
1 + ½ω ðkÞ ω, q 2 =γ ðkÞ2
This transfer function realizes an isotropic band-pass filter, with reflection
peak jrj ¼ 1 at jkj ¼ q.
Therefore, our device can be used to achieve multiple filtering functions.
As schematically shown in Fig. 3B, when the light frequency is on resonance
at normal incidence (ω ¼ ω0), the device operates as an isotropic high-pass
filter (Laplacian) at the transmission mode and an isotropic low-pass filter
at the reflection mode. When the light frequency is detuned a bit from
the resonance at normal incidence, the device operates as an isotropic
band-reject filter at the transmission mode and an isotropic band-pass filter
at the reflection mode, where the rejected or passed wavevector components
are determined by the light frequency detuning and polarization. For trans-
mission mode, the transmitted image is the required filtered result. For
reflection mode, the reflected image is the required filtered result, which
can be separated from the incident image by using a beam splitter.
4. Numerical demonstration of the isotropic filters

Now we numerically demonstrate the filtering functions of our
device. In all the numerical demonstration below, the incident beam is S
polarized. The transmitted image is calculated following the standard way
in image processing (Gonzalez and Woods, 2008). (1) Compute the Fourier

transform S in ðkx ,ky Þ of the incident field profile Sin(x, y). Note the incident
image is jSin(x, y)j2. (2) Compute the Fourier transform of the output field

profile, S out ðkx ,ky Þ ¼ Hðkx , ky ÞS in ðkx , ky Þ, where H(kx, ky) is the transfer
function. (3) Obtain the output field profile Sout(x, y) by inverse Fourier
transform. Calculate the output image jSout(x, y)j2.
4.1 Isotropic high-pass filter (the Laplacian)

Fig. 4 illustrates the isotropic high-pass filter (the Laplacian). Fig. 4A
plots the filter transfer function: the transmittance for S-polarized light
jtsj(kx, ky) at the frequency ω0 ¼ 0.77098 2πc/a. The transfer function
is almost isotropic. Fig. 4B plots jtsj as a function of jkj along a general
wavevector direction (φ ¼ 14°), and the fitting result of Eq. (21). The fitting
is very good for jkj up to 0.6 102 2π/a. These plots confirm that the
device indeed operates as an isotropic high-pass filter (the Laplacian) in this
case. In Fig. 4A (and also in Fig. 5A as shown later), at larger wavevectors,
the transfer function exhibits some anisotropy due to the dependency of the
radiative linewidth γ(k) on the direction of k.
The Laplacian enables image sharpening and edge detection (Guo et al.,
2018b). Fig. 4E shows an incident image of the Stanford emblem. Fig. 4C
plots the Fourier transform of the field profile for this incident image.
Fig. 4D shows the calculated Fourier spectrum for the transmitted image,
which is obtained by a pointwise product of Fig. 4A and C. The low
wavevector components have been filtered out. Fig. 4F is the calculated
transmitted image, which contains only the edge of the incident image.
4.2 Isotropic low-pass filter

Fig. 5 illustrates the isotropic low-pass filter. Fig. 5A plots the filter transfer
function: the reflectance for S-polarized light jrsj(kx, ky) at the frequency
ω0 ¼ 0.77098 2πc/a. The transfer function is almost isotropic. Fig. 5B plots
jrsj as a function of jkj along a general wavevector direction (φ ¼ 14°), and
the fitting result of Eq. (22). The fitting is almost perfect in the wavevector
range as shown. These plots confirm that the device indeed operates as an
isotropic low-pass filter in this case.
The isotropic low-pass filter accomplishes image smoothing, with appli-
cations ranging from character recognition in machine perception, pre-
processing functions in printing and publishing industry, to satellite and
aerial image processing (Buades et al., 2005; Gonzalez and Woods, 2008).
Here we show one specific application of the low-pass filter in white noise
reduction. Fig. 5E shows an incident image of the Stanford emblem with
substantial white noise components, while Fig. 5C plots the Fourier
Fig. 4 Isotropic high-pass filter (Laplacian). (A) Transmittance for S-polarized light
jtsj(kx, ky) at the frequency ω0 ¼ 0.77098 2πc/a. (B) jtsj as a function of jkj along a
general wavevector direction (φ ¼ 14°), and the fitting result of Eq. (21). The fit is very
good for jkj up to 0.6 102 2π/a. (C) Log plot of the Fourier transform of the field

profile for the incident image (E): log jS in jðkx ,ky Þ. (D) Log plot of the Fourier transform

of the field profile for the reflected image (F): log jS out jðkx , ky Þ. The low wavevector
components have been filtered out. (E) Incident image jSinj2 of the Stanford emblem.
The image size is 5220a 3456a. (F) Calculated transmitted image jSoutj2, which shows
the edges with different orientations. jkj, kx, and ky are in the units of 102 2π/a.
Reprinted with permission from Guo, C., Xiao, M., Minkov, M., Shi, Y., Fan, S., 2018a.
Isotropic wavevector domain image filters by a photonic crystal slab device. J. Opt.
Soc. Am. A Opt. Image Sci. Vis. 35 (10), 1685–1691, the Optical Society.
Fig. 5 Isotropic low-pass filter. (A) Reflectance for S-polarized light jrsj(kx, ky) at the
frequency ω0 ¼ 0.77098 2πc/a. (B) jrsj as a function of jkj along a general wavevector
direction (φ ¼ 14°), and the fitting result of Eq. (22). (C) Log plot of the Fourier transform of

the field profile for the incident image (E): log jS in jðkx ,ky Þ. (D) Log plot of the Fourier trans-

form of the field profile for the reflected image (F): logjS out jðkx , ky Þ. The high wavevector
components have been filtered out. (E) Incident image jSinj2 of the Stanford emblem with
substantial white noise component. The image size is 5220a 3456a. (F) Calculated
reflected image jSoutj2, which reduces the white noise by image smoothing. jkj, kx,
and ky are in the units of 102 2π/a. Reprinted with permission from Guo, C., Xiao, M.,
Minkov, M., Shi, Y., Fan, S., 2018a. Isotropic wavevector domain image filters by a photonic
crystal slab device. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 35 (10), 1685–1691, the Optical
Society.
transform of the field profile for this incident image. Fig. 5D shows the
calculated Fourier spectrum for the reflected image, which is obtained by
a pointwise product of Fig. 5A and C. The high wavevector components
have been filtered out. Fig. 5F shows the calculated reflected image, where
the noise component has indeed been significantly reduced, demonstrating
image smoothing.
4.3 Isotropic band-reject filter

Fig. 6 illustrates the isotropic band-reject filter. Fig. 6A plots the filter transfer
function: the transmittance for S-polarized light jtsj(kx, ky) at the frequency
ωq ¼ 0.77130 2πc/a. The transfer function is almost isotropic. Figure 6B
plots jtsj as a function of jkj along a general wavevector direction (φ ¼ 14°),
which shows jtsj ¼ 0 at jkj ¼ 0.84 102 2π/a. We fit the transmittance
curve with Eq. (23) together with Eq. (12) and γ(k) γ 0. The fitting is almost
perfect in the wavevector range as shown. These plots confirm that the device
indeed operates as an isotropic band-reject filter in this case.
The isotropic band-reject filter can effectively eliminate periodic noise, a
common type of noise arising typically from electrical or electromechanical
interference during image acquisition (Aizenberg and Butakoff, 2002;
Banham and Katsaggelos, 1997; Gonzalez and Woods, 2008). As periodic
noise appears as peaks in the Fourier transform at locations corresponding
to the wavevectors of the periodic interference, it can be isolated and filtered
by band-reject filters. Here we show an example of periodic noise reduction
with the isotropic band-reject filter. Fig. 6E shows an incident image of the
Stanford emblem incorporating a periodic noise component. Fig. 6C plots the
Fourier transform of the field profile for this incident image. In the wavevector
space, the periodic noise appears as spectral peaks which lie on an approximate
circle around the origin. Fig. 6D shows the calculated Fourier spectrum for the
transmitted image, which is obtained by a pointwise product of Fig. 6A and C.
The spectral peaks corresponding to the periodic noise have been filtered out.
Fig. 6F shows the calculated transmitted image, where the periodic noise has
indeed been eliminated effectively. Here we note that the rejected wavevector
where the maximum rejection occurs in our band-reject filter can be easily
tuned by tuning the light frequency.
4.4 Isotropic band-pass filter

Fig. 7 illustrates the isotropic band-pass filter. Fig. 7A plots the filter transfer
function: the reflectance for S-polarized light jrsj(kx, ky) at the frequency
Fig. 6 Isotropic band-reject filter. (A) Transmittance for S-polarized light jtsj(kx, ky) at the
frequency ωq ¼ 0.77130 2πc/a. (B) jtsj as a function of jkj along a general wavevector
direction (φ ¼ 14°), and the fitting result of Eq. (23) along with Eq. (12) and γ(k) γ 0.
(C) Log plot of the Fourier transform of the field profile for the incident image

(E): log jS in jðkx , ky Þ. The sinusoidal noise appears as peaks in the spectrum which lie
on an approximate circle around the origin. (D) Log plot of the Fourier transform of

the field profile for the transmitted image (F): log jS out jðkx , ky Þ. The wavevector compo-
nents corresponding to the periodic noise have been filtered out. (E) Incident image
jSinj2 of the Stanford emblem with periodic noise. The image size is 5220a 3456a.
(F) Calculated transmitted image jSoutj2, which eliminates the periodic noise. jkj,
kx, and ky are in the units of 102 2π/a. Reprinted with permission from Guo, C.,
Xiao, M., Minkov, M., Shi, Y., Fan, S., 2018a. Isotropic wavevector domain image filters
by a photonic crystal slab device. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 35 (10),
Fig. 7 Isotropic band-pass filter. (A) Reflectance for S-polarized light jrsj(kx, ky) at the
frequency ωq ¼ 0.77130 2πc/a. (B) jrsj as a function of jkj along a general wavevector
direction (φ ¼ 14°), and the fitting result of Eq. (24) together with Eq. (12) and γ(k) γ 0.
(C) Log plot of the Fourier transform of the field profile for the incident image

(E): log jS in jðkx , ky Þ. The sinusoidal noise appears as impulses in the spectrum which
lie on an approximate circle around the origin. (D) Log plot of the Fourier transform

of the field profile for the reflected image (F): log jS out jðkx , ky Þ. (E) Incident image jSinj2
of the Stanford emblem with periodic noise. The image size is 5220a 3456a.
(F) Calculated reflected image jSoutj2, which isolates the periodic noise and simplifies
its analysis. jkj, kx, and ky are in the units of 102 2π/a. Reprinted with permission from
Guo, C., Xiao, M., Minkov, M., Shi, Y., Fan, S., 2018a. Isotropic wavevector domain image
ωq ¼ 0.77130 2πc/a. The transfer function is almost isotropic. Fig. 7B

plots jrsj as a function of jkj along a general wavevector direction (φ ¼ 14°),
which shows jrsj ¼ 1 at jkj ¼ 0.84 102 2π/a. We fit the reflectance
curve with Eq. (24) together with Eq. (12) and γ(k) γ 0. The fitting is
almost perfect in the wavevector range as shown. These plots confirm that
the device indeed operates as an isotropic band-pass filter in this case.
The isotropic band-pass filter performs the opposite operation of the
band-reject filter. It is quite useful in isolating the effects on an image caused
by selected wavevector bands (Gonzalez and Woods, 2008). Here we show
an example of extracting periodic noise patterns with the isotropic band-pass
filter. Fig. 7E shows an incident image of the Stanford emblem incorporating
periodic noise (same as Fig. 6E), while Fig. 7C plots the Fourier transform of
the field profile for this incident image. Fig. 7D shows the calculated Fourier
spectrum for the reflected image, which is obtained by a pointwise product
of Fig. 7A and C. Fig. 7F shows the calculated reflected image, where the
periodic noise pattern is isolated and appears more clearly. This is useful
because it simplifies the analysis of the noise, largely independent of the
image content (Gonzalez and Woods, 2008).
5. Conclusion
Our design of isotropic wavevector domain image filters using a pho-
tonic crystal slab is based on the guided resonances with isotropic band struc-
ture. The same idea can extend to other photonic structures that host resonant
modes with isotropic band structure. In particular, the phase-shifted Bragg
grating in Bykov et al. (2014) can also perform the four filtering functionalities
of our device, but with the transmission/reflection mode flipped.
In conclusion, we have shown that isotropic high-pass, low-pass, band-
reject, and band-pass filters can be implemented at visible wavelengths using
one photonic crystal slab device. Such a simple but multifunctional photonic
device may find various applications involving image processing. For instance,
it can be used as the first layer of image recognition systems. As such our work
indicates the new opportunities for compact, high-throughput, and low-
energy-consumption optical analog computing as provided by nanophotonic
structures.
Acknowledgments
We acknowledge Dr. Meng Xiao, Dr. Momchil Minkov, and Dr. Yu Shi for collaboration,
and Dr. Bo Zhao, Dr. Alexander Cerjan, and Yu Guo for helpful discussion. This work is
supported by Samsung Electronics and US Air Force (USAF) (FA9550-17-1-0002).
References
AbdollahRamezani, S., Arik, K., Khavasi, A., Kavehvash, Z., 2015. Analog computing using
graphene-based metalines. Opt. Lett. 40 (22), 5239. ISSN 0146-9592. https://doi.org/
10.1364/OL.40.005239. https://www.osapublishing.org/abstract.cfm?URI¼ol-40-22-
5239.
Abràmoff, M.D., Magalhães, P.J., Ram, S.J., 2004. Image processing with ImageJ.
Biophotonics Int. 11 (7), 36–42. ISSN 1081-8693.
Aizenberg, I.N., Butakoff, C., 2002. In: Frequency domain medianlike filter for periodic
and quasi-periodic noise removal. Proceedings of SPIE, vol. 4667. San Jose, http://
proceedings.spiedigitallibrary.org/proceeding.aspx?articleid¼876767. http://link.aip.
org/link/?PSI/4667/181/1&Agg¼doi.
Banham, M.R., Katsaggelos, A.K., 1997. Digital image restoration. IEEE Signal Process.
Mag. 14 (2), 24–41. ISSN 10535888. https://doi.org/10.1109/79.581363. http://
ieeexplore.ieee.org/document/581363/.
Bracewell, R.N., 1986. The Fourier Transform and Its Applications. McGraw-Hill,
New York. ISBN 0073039381.
Brosnan, T., Sun, D.W., 2004. Improving quality inspection of food products by computer
vision—a review. J. Food Eng. 61 (1), 3–16. ISSN 02608774. https://doi.org/10.1016/
S0260-8774(03)00183-3.
Buades, A., Coll, B., Morel, J.M., 2005. A review of image denoising algorithms, with a new
one. Multiscale Model. Simul. 4 (2), 490–530. ISSN 1540-3459. https://doi.org/
10.1137/040616024. http://epubs.siam.org/doi/10.1137/040616024.
Bykov, D.A., Doskolovich, L.L., Bezus, E.A., Soifer, V.A., 2014. Optical computation
of the Laplace operator using phase-shifted Bragg grating. Opt. Express 22 (21),
25084. ISSN 1094-4087. https://doi.org/10.1364/OE.22.025084. https://www.
osapublishing.org/oe/abstract.cfm?uri¼oe-22-21-25084.
Bykov, D.A., Doskolovich, L.L., Morozov, A.A., Podlipnov, V.V., Bezus, E.A., Verma, P.,
Soifer, V.A., 2018. First-order optical spatial differentiator based on a guided-mode
resonant grating. Opt. Express 26 (8), 10997–11006. ISSN 1094-4087. https://doi.
org/10.1364/OE.26.010997. https://www.osapublishing.org/abstract.cfm?URI¼oe-
26-8-10997.
Chang, T.Y., Hong, J.H., Yeh, P., 1990. Spatial amplification: an image-processing
technique using the selective amplification of spatial frequencies. Opt. Lett. 15 (13), 743.
ISSN 0146-9592. https://doi.org/10.1364/OL.15.000743. https://www.osapublishing.
org/abstract.cfm?URI¼ol-15-13-743.
Chizari, A., Abdollahramezani, S., Jamali, M.V., Salehi, J.A., 2016. Analog optical computing
based on a dielectric meta-reflect array. Opt. Lett. 41 (15), 3451. ISSN 0146-9592.
https://doi.org/10.1364/OL.41.003451. https://www.osapublishing.org/abstract.cfm?URI
¼ol-41-15-3451. https://arxiv.org/abs/1605.07150%0Ahttp://doi.org/10.1364/OL.
41.003451.
Dalal, N., Triggs, B., 2005. Histograms of oriented gradients for human detection. In: 2005
IEEE Computer Society Conference on Computer Vision and Pattern Recognition
(CVPR’05) IEEE, pp. 886–893.
Dong, Z., Si, J., Yu, X., Deng, X., 2018. Optical spatial differentiator based on subwavelength
high-contrast gratings. Appl. Phys. Lett. 112 (18), 181102. ISSN 0003–6951. https://doi.
org/10.1063/1.5026309. http://aip.scitation.org/toc/apl/112/18. http://aip.scitation.
org/doi/10.1063/1.5026309. https://doi.org/10.1063/1.5026309.
Eu, J.K.T., Liu, C.Y.C., Lohmann, A.W., 1973. Spatial filters for differentiation.
Opt. Commun. 9 (2), 168–171. ISSN 00304018. https://doi.org/10.1016/0030-4018
(73)90251-4. http://linkinghub.elsevier.com/retrieve/pii/0030401873902514.
Phys. Rev. B 65 (23), 235112. ISSN 0163-1829. https://doi.org/10.1103/
PhysRevB.65.235112.
Fang, Y., Ruan, Z., 2018. Optical spatial differentiator for a synthetic three-dimensional
optical field. Opt. Lett. 43 (23), 5893–5896.
Fang, Y., Lou, Y., Ruan, Z., 2017. On-grating graphene surface plasmons enabling spatial
differentiation in the terahertz region. Opt. Lett. 42 (19), 3840. ISSN 0146-9592.
https://doi.org/10.1364/OL.42.003840. https://www.osapublishing.org/abstract.cfm?
URI¼ol-42-19-3840.
Golovastikov, N.V., Bykov, D.A., Doskolovich, L.L., 2014. Resonant diffraction gratings
for spatial differentiation of optical beams. Quantum Electron. 44 (10), 984–988. ISSN
1063-7818. https://doi.org/10.1070/QE2014v044n10ABEH015477. http://stacks.iop.
org/1063-7818/44/i¼10/a¼984?key¼crossref.8bee7b769ad2412f3500ff65fffe3a5e%5Cn.
http://iopscience.iop.org/1063-7818/44/10/984. http://stacks.iop.org/1063-7818/44/i¼
10/a¼984?key¼crossref.8bee7b769ad2412f3500ff65fffe3a5e.
Gonzalez, R., Woods, R., 2008. Digital Image Processing, third ed. Prentice Hall, Upper
Saddle River, NJ. ISBN 9780131687288.
Goodman, J.W., 2005. Introduction to Fourier Optics, third ed. Roberts and Company,
Englewood. ISBN 0974707724.
G€orlitz, D., Lanzl, F., 1975. A holographic spatial filter for direction independent differen-
tiation. Jpn. J. Appl. Phys. 14 (S1), 223. ISSN 0021-4922. https://doi.org/10.7567/
JJAPS.14S1.223. http://stacks.iop.org/1347-4065/14/223.
Guo, C., Xiao, M., Minkov, M., Shi, Y., Fan, S., 2018a. Isotropic wavevector domain image
1685–1691.
Guo, C., Xiao, M., Minkov, M., Shi, Y., Fan, S., 2018b. Photonic crystal slab Laplace operator
for image differentiation. Optica 5 (3), 251. ISSN 2334-2536. https://doi.org/10.1364/
OPTICA.5.000251. https://www.osapublishing.org/abstract.cfm?URI¼optica-5-3-251.
Hwang, Y., Davis, T.J., Lin, J., Yuan, X.C., 2018. Plasmonic circuit for second-order spatial
differentiation at the subwavelength scale. Opt. Express 26 (6), 7368. ISSN 1094-4087.
https://doi.org/10.1364/OE.26.007368. https://www.osapublishing.org/abstract.cfm?
URI¼oe-26-6-7368.
Kwon, H., Sounas, D., Cordaro, A., Polman, A., Alù, A., 2018. Nonlocal metasurfaces for
optical signal processing. Phys. Rev. Lett. 121 (17), 173004.
Markham, R., Frey, S., Hills, G.J., 1963. Methods for the enhancement of image detail
and accentuation of structure in electron microscopy. Virology 20 (1), 88–102. ISSN
00426822. https://doi.org/10.1016/0042-6822(63)90143-0. http://linkinghub.elsevier.
com/retrieve/pii/0042682263901430.
Nejad, F.Z., Fleury, R., 2018. Performing mathematical operations using high-index
acoustic metamaterials. New J. Phys. 20, 073001.
Ochiai, T., Sakoda, K., 2001. Dispersion relation and optical transmittance of a hexagonal
photonic crystal slab. Phys. Rev. B 63 (12), 125107. ISSN 0163-1829. https://doi.
org/10.1103/PhysRevB.63.125107.
Pors, A., Nielsen, M.G., Bozhevolnyi, S.I., 2015. Analog computing using reflective plasmonic
metasurfaces. Nano Lett. 15 (1), 791–797. ISSN 1530-6984. https://doi.org/10.1021/
nl5047297.
Roberts, A., Gómez, D.E., Davis, T.J., 2018. Optical image processing with metasurface
dark modes. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 35 (9), 1575–1584.
Sakoda, K., 2005. Optical properties of photonic crystals, Springer Series in Optical Sciences,
vol. 80. Springer-Verlag, Berlin, Heidelberg. ISBN 3-540-20682-5.
Silva, A., Monticone, F., Castaldi, G., Galdi, V., Alu, A., Engheta, N., 2014. Performing
mathematical operations with metamaterials. Science. 343 (6167), 160–163. ISSN
0036-8075. https://doi.org/10.1126/science.1242818.
Sirohi, R.S., Mohan, V.R., 1977. Differentiation by spatial filtering. Opt. Acta 24 (11),
1105–1113. ISSN 0030-3909. https://doi.org/10.1080/713819447. http://www.
tandfonline.com/doi/abs/10.1080/713819447.
Solli, D.R., Jalali, B., 2015. Analog optical computing. Nat. Photonics 9 (11), 704–706.
ISSN 1749-4885. https://doi.org/10.1038/nphoton.2015.208. http://www.nature.com/
doifinder/10.1038/nphoton.2015.208.
66 (4), 045102. ISSN 0163-1829. https://doi.org/10.1103/PhysRevB.66.045102.
Vogelaar, L., Nijdam, W., van Wolferen, H.A.G.M., de Ridder, R.M., Segerink, F.B., Fl€ uck, E.,
Kuipers, L., van Hulst, N.F., 2001. Large area photonic crystal slabs for visible light with
waveguiding defect structures: fabrication with focused ion beam assisted laser interference
lithography. Adv. Mater. 13 (20), 1551. ISSN 09359648. https://doi.org/10.1002/1521-
4095(200110)13:20¡1551::AID-ADMA1551¿3.0.CO;2-V. http://doi.wiley.com/10.1002/
1521-4095%28200110%2913%3A20%3C1551%3A%3AAID-ADMA1551%3E3.0.CO%
3B2-V.
Wang, K.X., Yu, Z., Sandhu, S., Fan, S., 2013. Fundamental bounds on decay rates in asym-
metric single-mode optical resonators. Opt. Lett. 38 (2), 100. ISSN 0146-9592. https://
doi.org/10.1364/OL.38.000100. https://www.osapublishing.org/abstract.cfm?URI¼ol-
38-2-100.
Wu, W., Jiang, W., Yang, J., Gong, S., Ma, Y., 2017. Multilayered analog optical differen-
tiating device: performance analysis on structural parameters. Opt. Lett. 42 (24), 5270.
ISSN 0146-9592. https://doi.org/10.1364/OL.42.005270. https://www.osapublishing.
org/abstract.cfm?URI¼ol-42-24-5270.
Youssefi, A., Zangeneh-Nejad, F., Abdollahramezani, S., Khavasi, A., 2016. Analog
computing by Brewster effect. Opt. Lett. 41 (15), 3467. ISSN 0146-9592. https://
doi.org/10.1364/OL.41.003467. https://www.osapublishing.org/abstract.cfm?URI¼ol-
41-15-3467.
Yu, F.T.S., Jutamulia, S., Yin, S., 2001. Introduction to Information Optics. Academic Press,
San Diego. ISBN 0127748113.
Zhou, W., Zhao, D., Shuai, Y.C., Yang, H., Chuwongin, S., Chadha, A., Seo, J.H.,
Wang, K.X., Liu, V., Ma, Z., Fan, S., 2014. Progress in 2D photonic crystal Fano res-
onance photonics. Prog. Quantum Electron. 38 (1), 1–74. ISSN 00796727. https://doi.
org/10.1016/j.pquantelec.2014.01.001. http://linkinghub.elsevier.com/retrieve/pii/
S0079672714000020. https://doi.org/10.1016/j.pquantelec.2014.01.001.
Zhu, T., Zhou, Y., Lou, Y., Ye, H., Qiu, M., Ruan, Z., Fan, S., 2017. Plasmonic computing
of spatial differentiation. Nat. Commun. 8, 15391. ISSN 2041-1723. https://doi.org/
10.1038/ncomms15391. http://www.nature.com/doifinder/10.1038/ncomms15391.
https://doi.org/10.1038/ncomms15391.
CHAPTER FIVE
Guided mode resonances and

photonic crystals for biosensing
and imaging
Alexander Drayton, Isabel Barth, Thomas F. Krauss*
The Department of Physics, University of York, York, United Kingdom
*Corresponding author: e-mail address: thomas.krauss@york.ac.uk
Contents
1. Introduction 116
1.1 A historical note 117
2. Figures of merit for sensing 118
2.1 Photonic factors 119
2.2 Electronic factors 119
2.3 Biochemical factors 120
3. Operation of GMR sensors 121
3.1 Optical signature 121
3.2 Polarization dependence 122
3.3 Q-factor and sensitivity 123
3.4 Bulk sensitivity 124
3.5 Surface sensitivity 125
4. Comparison between GMR sensors and other sensing modalities 128
4.1 Surface plasmon resonance 128
4.2 Nanohole arrays 129
4.3 Microrings 130
4.4 Bimodal waveguide 131
4.5 Concluding remarks 132
5. Resonant hyperspectral imaging 132
5.1 Sensing and imaging 135
5.2 Multiparameter imaging 137
5.3 Plasmonic nanohole arrays 137
6. Spatial-spectral equivalence 139
7. Phase-sensitive measurement 142
7.1 Realization of phase-sensitive measurements 143
8. Conclusion 145
References 146

116 Alexander Drayton et al.
1. Introduction
This chapter considers applications in the area of sensing. Since the
very nature of the resonance is less relevant in this context, we shall refer
to it more generally as a guided mode resonance, GMR, which applies
whether the Fano nature of the resonance is apparent or not.
The key property of a sensor is its ability to translate a quantity that is
difficult to measure into one that can be measured easily, so it acts as a trans-
ducer. In the case of GMR sensors, the fact that a GMR is a surface wave that
interacts closely with its environment is exploited, so GMR sensors fall into
the general class of surface affinity sensors. Surface affinity sensors are com-
monly used in biomedical applications and they detect target molecules
by their selective binding to a capture molecule such as an antibody, the anti-
body being anchored to the surface. The target molecule may be a marker
for cancer or some other disease. When the target binds to the antibody, the
optical mass on the surface increases and so the refractive index on the sur-
face increases. This local increase in refractive index changes the effective
index of the guided mode and thereby its resonance wavelength. By mea-
suring the change in wavelength, one can directly infer back to the change in
refractive index and the concentration of target molecules in the biological
medium, i.e., the analyte. Ultimately, a GMR sensor is therefore a refractive
index sensor and its performance is best described by the smallest refractive
index change it can detect; this principle is illustrated in Fig. 1, and the rel-
evant performance metrics will be discussed in Section 3.
A key point to note at the very beginning of this chapter is that GMR
sensors are not the most highly performing of all available sensors; other
types of sensors offer lower limits of detection. Nevertheless, as the chapter
develops, it will become apparent that GMRs offer the most versatile
approach toward sensing. They can be readily incorporated into low-cost
point-of-care devices because of their ease of coupling, they can perform
both the sensing and the imaging function, and they offer a multiparameter
approach via the wavelength, the amplitude, the Q-factor, and the phase of
the resonance. All of these effects will be discussed in detail in this chapter.
The driver behind much of the discussion is therefore not the quest for the
highest possible performance, but the desire for versatility, i.e., suitability for
a variety of applications, and the highest performance that can be built at the
lowest cost, which is much needed by healthcare providers both in the
developed and in the developing world.
Biosensing and imaging 117
Fig. 1 Sketch of the operation of a GMR surface affinity sensor, which translates the
binding of target molecules into a shift of the resonance wavelength.
1.1 A historical note

A guided mode resonance can be understood as the extension of the “Wood
anomaly” first described in 1902, when Wood, observing the spectrum of a
continuous light source given by a metallic diffraction grating, noticed a sur-
prising phenomenon: “I was astounded to find that under certain conditions,
the drop from maximum illumination to minimum, a drop certainly of from
10 to 1, occurred within a range of wavelengths not greater than the distance
between the sodium lines” (Wood, 1902). Wood was unable to explain the
observation, so he termed it a “singular anomaly.” Rayleigh later explained
that the anomaly occurs at the wavelength for which a scattered wave
emerges tangentially to the grating surface (Rayleigh, 1907), in other words,
when the diffracted light lies in the plane of the grating. In the simplest
case, i.e., for normal incidence, for the first diffraction order (m ¼ 1), the
grating equation.
mλ
sin ðθm Þ ¼ (1)
ng
Eq. (1) tells us that the diffraction angle θm ¼ 90 degrees occurs when the
wavelength λ equals the grating period g, which is the point at which the
sharp drop in intensity observed by Wood occurs. The refractive index n
in the denominator highlights that it is the wavelength in the material where

the diffracted wave travels that needs to be considered, which is the same
property that provides the sensing function as highlighted in Fig. 1.
A much more comprehensive explanation of the Wood anomaly can be
found in the chapter by Maystre (2012).
With the emergence of guided-wave optics in the 1960s, the Anomaly
was reconsidered in the context of a guided wave approach. (Hessel and
Oliner, 1965). Magnusson, in the 1990s, then recognized that the effect
could be extended by creating the grating in a higher refractive index layer
that supports guided modes (Wang et al., 1990). The basic physics remained
the same, so Eq. (1) still holds, except that one needs to consider the effective
index of the guided mode neff, so the wavelength λ in Eq. (1) becomes λ/neff.
The idea to track the sharp “drop” observed by Wood as a function of
refractive index and to use it for sensing applications was first mentioned in
1992 (Magnusson and Wang, 1992) and the first basic GMR sensor was
shown to exhibit sensitive shifts in wavelength to changes in refractive index
in the late 1990s. A comprehensive summary of GMR sensor history is avail-
able from Magnusson et al. (2011).
2. Figures of merit for sensing

In order to describe the performance of a GMR sensor and to compare
it to other sensor modalities, it is essential to define a universal figure of
merit (FOM).
Such FOM allows us to benchmark the sensor and to identify routes for
optimizing its performance. The FOM then yields the limit of detection
(LOD), which is the single most important number used to describe a sensor.
In terms of the optical performance, the LOD is often expressed as the min-
imum resolvable wavelength shift, e.g., as an LOD of Δn ¼ 106. Using the
refractive index is convenient in terms of sensor design and assessment and a
useful figure to optimize for the photonics engineer. The smallest measur-
able LOD (Armbruster and Pry, 2008) is defined as:
LOD ¼ LOB + 3σ (2)
where limit of blank (LOB) is the analyte concentration measured when no
analyte is present which means LOB is merely an offset for the LOD hence
Eq. (2) can actually be thought of as LOD 3σ. This consideration specifies
the LOD in terms of the refractive index shift that is three times larger than
the noise level.
The real purpose of the sensor, however, is to detect the concentration of

a biomarker. Therefore, the LOD can also be expressed as a biological con-
centration. For example, the clinically relevant concentration of prostate
specific antigen (PSA) in blood is considered to be 4 ng/mL, so a given
PSA sensor should have an LOD better than this value in order to allow
for error-free clinical decisions to be made.
Using the biological concentration as the LOD adds a significant level of
complication as there are a number of additional factors influencing the
operation; devices with very similar optical performance may yield very dif-
ferent biological LODs, as the surface chemistry used to functionalize the
sample may not always be exactly reproducible when compared to the
nanofabrication, which is usually easier to control. Overall, we therefore
classify the factors influencing the LOD as photonic, electronic and
biochemical.
2.1 Photonic factors

Photonic factors include the linewidth or Q-factor of the resonance and its
sensitivity to refractive index changes. The Q-factor, typically defined as the
ratio of the resonance’s central wavelength over its full width half maximum
describes the sharpness of the resonance. This is beneficial to sensor opera-
tion because a higher Q-factor allows the central wavelength of the reso-
nance to be more accurately defined and can therefore be tracked better.
The sensitivity then expresses the shift of the resonance in terms of the wave-
length shift Δλ as a function of external refractive index changes, Δn, thus
sensitivity S ¼ Δλ/Δn, commonly expressed in nm/RIU, with RIU the
dimensionless “refractive index unit.” It is clear that the product of Q-factor
and sensitivity is a good starting point for the desired FOM. For instance, a
GMR sensor exhibits a typical Q-factor of 100–200 and a sensitivity of a
similar magnitude, i.e., 100–200 nm/RIU. By comparison, a surface
plasmon resonance (SPR) sensor has a sensitivity of 1000 nm/RIU or more,
but only a Q of 10 or 20, hence S Q 104–105 is similar in both cases
(Altintas et al., 2011; Cunningham et al., 2002).
2.2 Electronic factors

Electronic factors mainly describe noise. The light source has instabilities
described by its intensity and phase noise, with lasers having additional
wavelength noise. The detectors add further thermal noise as well as noise
related to the discretization/pixilation in case detector arrays or cameras are
being used. Many of these noise sources can be mitigated, e.g., by using
advanced algorithms or simply by incrementing the integration time, but
it is often difficult to quantify the overall impact of these factors. Several
authors have suggested that optical sensing methods are ultimately limited
by the photon shot noise, but in practice, all the other sources of noise tend
to dominate well before the shot noise limit is reached.
Because noise sources are variable and very much depend on the partic-
ular setup being used, including the stability of the laboratory equipment,
the FOM is typically not expressed as LOD ¼ k Q S, with k some pro-
portionality factor that accounts for the electronic noise factors. In fact, ear-
lier work by White and Fan (2008) attempted to define such a universal
FOM, but we note that their figure is not used in the field. Instead,
researchers tend to quote Q-factor, sensitivity and LOD separately.
As a general rule, high performance sensors achieve LOD values of
Δn 105–106. Putting these values into context with the refractive index
variation of water, which is of order Δn ¼ 104/°C (Bashkatov and Genina,
2003), we note that achieving an LOD of Δn ¼ 106 requires a temperature
stability of ΔT ¼ 102 degrees. Even though LODs of Δn 107–108 have
been reported, it is clear that such values can only be achieved with highly
stabilized laboratory equipment and by using liquid analytes that have been
preconditioned.
2.3 Biochemical factors

Biochemical factors arguably provide the largest variability. To illustrate this
point, we refer to the detection of the carcinoembryonic antigen (CEA) as
an example, which is a well-known colorectal cancer marker of mass
180 kDa. Using a nanobeam cavity with a bulk sensitivity of 70 nm/RIU
and a Q-factor of 104, Liang et al. demonstrated an LOD of 0.1 pg/mL
(Liang et al., 2013). For the same biomarker, SPR-based devices are only
able to reach 3 ng/mL, using the Biacore system (Altintas et al., 2011), which
has Q 10 and S 3000 nm/RIU. The S Q product yields an advantage
of one order of magnitude in favor of the nanobeam, while the LOD in
terms of concentration yields an advantage of four orders of magnitude
for the same system. Given the limited information available in the respective
papers, one can only assume that lower noise and variations in the bio-
chemical functionalization strategies are responsible for these large differences
in observed performance.
These biochemical factors include the choice of capture molecule, e.g.,
as naturally derived antibodies or synthetically manufactured molecules such
as affimers or DNA/Peptide aptamers; the choice of surface preparation, e.g.,

with silane monolayers that act as intermediaries between the sensor surface
and the capture molecule; the density of coverage, the pH, the microfluidic
environment, but also whether the target molecule is present in a more com-
plex matrix such as blood plasma or urine. The latter point is particularly
important as many papers use controlled buffers to determine their limits
of detection, which yields somewhat idealized results; if the measurements
are carried out in real matrices, nonspecific binding can easily screen the
detection of the desired binding events and yield inferior results. It is for
these rather complex reasons that sensors with comparable photonic prop-
erties may report LODs ranging from the ng/mL range into the sub-pg/mL
range based on differences in biochemistry alone.
3. Operation of GMR sensors

Having established the most relevant figures of merit, we can consider
the operation of GMR sensors in more detail as well as illustrating their
operational principle.
First of all, a significant characteristic of a GMR structure is the ease of
coupling of light into the grating with out-of-plane illumination. This is in
contrast to nonleaky waveguide-based sensors such as microrings (Barrios,
2012) or bimodal waveguides (Herranz et al., 2017). Nonleaky waveguide
sensors may achieve higher sensitivity, but they are also more difficult to
implement in a practical point-of-care device, because coupling to a wave-
guide involves careful alignment with tight tolerances, even when a grating
coupler is used. A GMR does not require such tolerances and high efficiency
coupling can readily be achieved, even with simple collimated LEDs in a
rudimentary setup (Block et al., 2008; Triggs et al., 2015).
3.1 Optical signature

As described in Section 1, a characteristic peak in reflection is observed
when the diffracted light couples into the waveguide layer. In fact, the thick-
ness of the waveguide layer can be adjusted to yield the inverse, i.e., a peak in
transmission, because the leaky mode may interfere constructively or
destructively with the directly transmitted light (Rosenblatt et al., 1997).
For the sensing application, however, it is more convenient to operate
in reflection as it eases the integration of the photonics/electronics,
implementing them on one side of the sensor and incorporating the
biology/fluidics on the other.
3.2 Polarization dependence

A typical GMR that consists of a 1D grating exhibits birefringence and thus
supports two different modes depending on the polarization of the incident
light. When the incident light is polarized perpendicular to the grating vec-
tor, the resulting mode is called a TE mode, and it is called a TM mode for
polarization along the grating vector. TE and TM modes interact differently
with the surrounding medium due to the polarization dependent boundary
conditions at the interfaces, which yields different distributions of the elec-
tric field amplitude, as illustrated by the cross-sectional plots shown in
Fig. 2B and C. In particular, the signature of the TM mode exhibits an elec-
tric field component in the out-of-plane direction (vertical direction in
Fig. 2) because the propagation direction and the electrical field cannot
be collinear. Because of these different boundary conditions, the polariza-
tion also impacts on the penetration depth of a given mode into the grating,
with TM modes propagating for a longer distance before being Bragg-
reflected. This longer propagation length means that TM modes experience
more grating elements and correspondingly, they exhibit a higher quality
factor compared to TE modes.
Fig. 2 Examples for the reflectance (A) and spatial distribution of the dominant electric
field component at resonance (B, C) for TE and TM modes, respectively. The simulated
structure is a Si3N4-grating (nSi3N4 ¼ 2) with a thickness of 150 nm, a period of 555 nm and
a filling factor of 80% on a SiO2 substrate (nSiO2 ¼ 1.45). Water is assumed to fill the grat-
ing grooves and the superstrate (nliquid ¼ 1.333) at normal incidence from the
substrate layer.
A more detailed explanation for the difference between TE and TM is as

follows. The difference lies in their symmetry and the fact that the Q-factor
is limited by scattering/radiation loss. As shown in Fig. 2B and C, the field of
the TE mode is symmetric with respect to the grating grooves, i.e., the sign
of the field is identical on both sides of the groove. This means that light can
readily scatter by exciting an out-of-plane radiating wave. The field of the
TM mode, however, is asymmetric with respect to the grating grooves, so
the sign of the field on one side of the groove is the opposite to the sign on
the other. This means that the two scattered components cancel out each
other and cannot excite a radiating plane wave (Lalanne et al., 2006). There-
fore, the TM mode propagates much further into the grating before even-
tually scattering out.
3.3 Q-factor and sensitivity

The Q-factor difference between the TE and the TM mode is clearly evi-
dent in the simulation of Fig. 2, which shows a TM mode with a Q-factor of
Q ¼ 1669 and a TE mode in the same structure with a much lower Q-factor
of Q ¼ 104. As the Q-factor is an important element in determining the
FOM for sensing, it is clear that the TM mode is advantageous in this respect.
It is useful to understand that GMRs have no intrinsic loss, in contrast
to, e.g., the plasmonic structures discussed elsewhere, so the reflectance
peaks at 100%. In terms of experimental realization, however, lower reflec-
tance and quality factor are typically being observed, especially for the high-
Q TM mode. One obvious limitation is material roughness, which leads to
random scattering. The other is the angular dependence of the resonance,
i.e., the fact that the resonance sensitively depends on the angle of the
incoming beam. While the simulation assumes perfectly collimated light,
a real beam tends to cover an angular range, leading to a broadening of
the resonance by the nature of the measurement.
The second contributor to the FOM introduced in Section 2 is the sen-
sitivity S. The sensitivity describes the change of the resonance wavelength
as a function of refractive index, which is directly related to the overlap of
the field with the analyte (Block et al., 2008). The ideal situation would be a
perfect overlap of the field with the biomolecular binding region in the ana-
lyte, but since we are dealing with guided modes that are bound to a high
index region, only the tail of the mode overlaps with the analyte, commonly
known as evanescence wave sensing, which is discussed in more detail in
Section 3.5.
In terms of sensitivity, we need to distinguish between bulk sensitivity

and surface sensitivity. The bulk sensitivity is what is usually quoted, and it
measures the overlap of the mode with the analyte. In contrast, the surface
sensitivity only measures the overlap between the mode and a thin film on
the surface, typically the first 10–20 nm. These numbers are explained by the
small size of the relevant molecules; for example, a typical protein (e.g.,
40–50 kDa) is of order 5–8 nm in size. Since the purpose of biosensing is usu-
ally to measure refractive index changes in a thin film of, e.g., proteins bound
to the surface, the surface sensitivity is, among others, the most useful param-
eter. Unfortunately, surface sensitivity is more difficult to measure and is
hence rarely quoted. Instead, most papers quote the bulk sensitivity.
3.4 Bulk sensitivity

The bulk sensitivity measures the response to the refractive index of the
entire superstrate (also referred to as the cover). In the case of Fig. 2, the bulk
sensitivity of the two modes is similar with 148 nm/RIU (TE) and
154 nm/RIU (TM). The nature of the field confinement is clearly different,
however, with 43% of the overall bulk sensitivity of the TE mode relying on
the field residing in the grooves, compared to less than 5% for the TM mode.
This means that the analyte has to enter the grooves to be sensed by the
TE mode. Given the small size of the grooves (typ. 100 nm) there may be
wetting issues in a practical sensing situation, such that the analyte may not
be able to diffuse into these grooves rapidly. This consideration therefore
suggests that the TE mode is the less favorable choice and that the TM mode
is more advantageous, i.e., the same conclusion as in the Q-factor discussion.
Fig. 3 shows the reflectance of the same GMR grating for TM polariza-
tion as a function of the superstrate refractive index. It is clear that the depen-
dence is not linear, which is due to variations in mode confinement; if the
index of the superstrate layer is low, the mode is strongly confined to the
waveguide layer and the overlap between the mode and the analyte is
low, resulting in low sensitivity. For higher index, the mode is drawn into
the analyte and sensitivity increases. The interesting region for biosensing is a
superstrate index around 1.33–1.34, i.e., the index of water or slightly
above, since the goal of biosensing is to measure low concentrations of bio-
markers in water-based matrices such as blood plasma or urine. In this
regime, the resonance shift with respect to the water baseline is sufficiently
linear to justify describing it as a linear FOM in terms of nm/RIU.
Fig. 3 (A) Resonance map for TM mode showing reflectance as a function of the inci-
dence wavelength and refractive index of the cover layer. Incidence is from the sub-
strate layer. (B) Linear range of sensitivity around the baseline index of water as
typically used in biosensing. (C,D) Field plots illustrating the mode confinement for
the extremes of lowest and highest refractive index of the cover layer.
3.5 Surface sensitivity

As already stated, the bulk sensitivity is only a proxy for the performance of a
sensor and it would be pertinent to quote the surface sensitivity instead. Ulti-
mately, the sensitivity is determined by the overlap of the resonant mode
with the analyte. As is well known from guided-mode theory (Yariv,
1991), the evanescent tail of the guided mode can be described in terms
of the effective index of the mode neff and the refractive index of the sup-
erstrate nsup.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
z
p 2π
EðzÞ ¼ E0 e p¼ n2eff n2sup (3)
λ
with p being the decay-constant in the transverse direction (here: z-direc-

tion, using the same convention as in Figs. 2 and 3). Typical values for the
decay length 1/p in the guided mode systems considered here, i.e., dielectric
waveguides, surface plasmons, and GMRs are of order 100–200 nm. This
length scale highlights the key issue, namely, that the size of the mode tail
tends to be much larger than the biolayer. In order to account for this dis-
crepancy, we need to define another parameter, namely, the surface sensi-
tivity, which describes the overlap integral between the mode and a thin film
on the surface of the sensor. Unfortunately, there is no accepted, standard-
ized method for determining the surface sensitivity. Cunningham et al., for
example, define a “detection zone” of 25 nm (Block et al., 2008), whereas
we believe that it is more realistic to assume a thickness of t ¼ 10 nm, which
is closer to the size of an antibody-bound protein. To illustrate the impact of
the effective index (Eq. 3) we exemplify the mode size and corresponding
surface sensitivity on two separate structures, i.e., a GMR realized in silicon
nitride (nSi3N4 ¼ 2) and one realized in silicon (nSi ¼ 3.56), with refractive
index being the only variable and the other parameters being kept constant.
The first observation is that the modal field of the Si3N4 structure extends
much further into the cladding, in direct agreement with Eq. (3), because its
effective index is lower. The second observation is that the Q-factor of the
silicon grating is much lower; this is because the light propagating into
the grating is scattered much more strongly at the silicon interface than
at the nitride interface, so the mode in the silicon structure has a much
shorter penetration depth Lp, as also discussed in Section 5 (Eq. 4), and a
correspondingly lower Q-factor.
Using the definition of a 10 nm thick biolayer, one can then compare the
surface sensitivity of the different structures. In terms of refractive index of
the biolayer, it is now understood that biomolecules have a refractive index
of n 1.56 (Young et al., 2018). Using this value would assume that the
entire surface layer is densely packed with molecules, however. In reality,
packing is not as dense, and one needs to assume some average effective
index somewhere between n ¼ 1.56 and n ¼ 1.33 (water). We chose a value
of n ¼ 1.4, in part also because it yields a maximum wavelength shift of
around Δλ 1 nm (Table 1), which corresponds to experimental observa-
tions. In our model, the Si3N4 grating shows a surface sensitivity of
Δλ ¼ 0.9 nm and the Si grating of Δλ ¼ 1.5 nm. All the relevant parameters
are compiled in Table 1 for ease of comparison.
Table 1 clearly highlights the difference between bulk sensitivity and
surface sensitivity. While both structures exhibit similar bulk sensitivities of
around 180 nm/RIU, the surface sensitivity differs almost by a factor 2.
Table 1 Parameter comparison between Si and Si3N4 gratings as displayed in Fig. 4 and
their resulting bulk and surface sensitivities
Grating Refractive Effective Bulk sensitivity Surface sensitivity
material index index (nm/RIU) (nm; n 5 1.4, t 5 10 nm)
Silicon 3.56 2.63 187 1.50
Silicon 2.00 1.75 176 0.87
nitride
Fig. 4 (A) Electric field profiles of TM modes at resonance for two different dielectric
materials and (B) the resulting GMR spectra. All parameters are the same as in Fig. 2
except for the material refractive index.
This difference is a direct consequence of Eq. (3); since the effective index in
the silicon structure is higher, the evanescent tail is shorter and the overlap
with the biolayer is larger.
An alternative way of highlighting the difference in surface sensitivity is
shown in Fig. 5, which is a plot of wavelength shift as a function of biolayer
thickness, similar to the study performed in Otte et al. (2009). Using the
same structures as in Fig. 4, the graph illustrates the higher surface sensitivity
of the higher index silicon grating, especially for a small biolayer thickness.
For a large thickness, the two curves converge, as is expected for the sim-
ilarity in bulk sensitivity.
Fig. 5 Simulated resonance wavelength shift for varying biolayer thicknesses assuming
a constant refractive index nbiolayer ¼ 1.4 on the sensing surface for the two gratings
defined in Fig. 4
4. Comparison between GMR sensors and other

sensing modalities
Some of the advantages of GMR sensors such as the ease of light
in-coupling have already been mentioned. This advantage is traded off
against lower Q-factor, so it is very instructive to tension the GMR against
competing approaches. The most obvious of the competitors are (a) the
widely considered “gold standard” of label-free biosensing, namely, SPR
commercialized by Biacore(now GE®), the microring resonator, commer-
cialized by Genalyte®, and the bimodal waveguide interferometric biosen-
sor, pioneered by Lechuga’s group at ICN2, Barcelona.
4.1 Surface plasmon resonance

SPR biosensors were developed in the 1980s as one of the first embodiments
of optical biosensors (Liedberg et al., 1983). Like GMR, they use out-of-plane
excitation, which makes it relatively easy to couple to. The mode that is used
for transduction is a surface plasmon mode, however, which is based on a
surface charge oscillation at a metal-dielectric interface. The surface plasmon
can be understood as a traveling wave and as such, it is excited by k-vector
matching at the interface; the incident light is directed through a prism or
a high-index hemisphere, and by adjusting the angle, its propagation constant
can be adjusted to match that of the plasmon mode, illustrated in
Fig. 6A. When the k-vectors do not match, the light from the prism does
not couple into the plasmon and the light is reflected. When the angle varies
Fig. 6 (A) Operation principle of a surface plasmon resonant sensor and (B) observation
of the characteristic dip upon coupling.
and they do match, the light does couple and the plasmon is excited. Since the
plasmon is a lossy mode, a lot of the light is dissipated before it couples back, so
a characteristic dip is observed in reflection (Fig. 6B). It is the traveling wave
nature of the surface plasmon that gives it its exquisite sensitivity, because the
smallest changes in effective index of the plasmon mode will lead to a reso-
nance shift (Yu and Fan, 2011). As a result, sensitivities of several
1000 nm/RIU have been reported (Pfeifer et al., 1999).
Because of the lossy nature of the plasmon mode, SPR systems have
very low quality factors, however, i.e., of the order of Q ¼ 10–20 which
is the lowest of the sensing modalities we explore here. The low Q factor
is offset by the high sensitivity, however, so the overall FOM S Q
104–105 nm/RIU which is comparable to most biosensing platforms includ-
ing the GMR. As a result, the typical implementation of SPR, such as in
Biacor’s SPR system, exhibits a LOD in the ng/mL range (Altintas et al.,
2011; Homola, 2008), making it useful for immunoassays, while it is not
the most sensitive of all available biosensors.
4.2 Nanohole arrays

Nanohole arrays represent a hybrid between SPR and GMR, because they
merge plasmon modes with periodic arrays. Nanohole arrays first came to
the fore upon the observation of extraordinary optical transmission
(EOT). EOT describes the phenomenon whereby light is observed to trans-
mit through an array of subwavelength holes with a transmission that is
much higher than expected from geometry alone. EOT is typically
explained as a three-step process whereby light first couples to the surface
plasmon at the metal interface, the plasmon then propagates through the
holes and couples back to a free-space mode on the other side (Genet and
Ebbesen, 2007). The conversion to plasmons explains how light can transmit
through tiny holes that are well below any classical cutoff (Liu and Lalanne,
2008; Martı́n-Moreno et al., 2001), which is why the effect is termed
“extraordinary” optical transmission.
The recognition that nanohole arrays can be used for sensing came later
with key demonstrations by Altug et al. who explored the resonance char-
acteristics and relevant sensing applications (Cetin et al., 2015).
The resonances they studied usually exhibit Q factors of the order of
Q 100 when working with liquids (Yanik et al., 2011) and sensitivities
of up to 650 nm/RIU which yields a FOM of S Q 104–105, which
is comparable albeit slightly higher than SPR and GMR structures. As a
result, nanoholes have exhibited impressive detection limits down to
500 pg/mL for the real-time molecular binding of IgG and 145 pg/mL
for vascular endothelial growth factor (VEGF) (Li et al., 2017).
These values demonstrate the strength of the nanohole array method as it
combines the benefits of GMR with its ease of coupling with the sensitivity
of SPR. The only downside is that the transmission is relatively low
(T 10%–20%) which results in weak signals. Since nanoholes only work
in transmission, it also means that the collection optics interferes with the
fluidic access, i.e., detection electronics and fluidics are on the same side
of the device. As stated previously, the fact that GMR and SPR can operate
in reflection means that the optics and the electronics can be placed on one
side of the device and the fluidics on the other, which is much more elegant
for practical implementations.
4.3 Microrings
Microrings are guided-mode devices where light propagates in the form of a
whispering gallery mode which is resonant when an integer number of
wavelengths fit into the circumference of the ring. The mode is strongly
confined to the waveguide (Fig. 7) and is a proper long-lived guided mode
(unlike the leaky mode of the GMR or the very lossy mode of the SPR), so
very high Q-factors can be achieved. In fact, Q-factors into the Q 108–109
have been achieved with microdisk resonators (Armani et al., 2003) but
these are not typically used for sensing as they are impractically high; as
explored by Hu et al. (2009), Q-factors above 105 are not useful in this con-
text because the wavelength noise of typical lasers then limits the accuracy of
determining the resonance. In terms of sensitivity, microrings tend to be
Fig. 7 Operation principle of microring resonator sensor with the top view also showing
the access waveguide.
worse than the other modalities, because the same strong confinement to the
waveguide that is needed for the high Q-factor tends to limit the overlap of
the mode with the analyte, such that sensitivities of S 50–80 nm/RIU are
typically being achieved (Bogaerts et al., 2012; Juan-Colás et al., 2017b).
Nevertheless, due to the high Q-factor, the overall FOM tends to be quite
high, i.e., S Q 105–106, which is higher than the other modalities
explored so far.
4.4 Bimodal waveguide

Bimodal waveguides represent the class of interferometric waveguide sen-
sors. Bimodal waveguides obtain the sensing information by interfering
two modes that coexist in the same waveguide. The advantage of this
approach is the “common-path” nature of the interferometer, which ensures
high stability and which means that the signal and the reference experience
exactly the same external influences. This feature is in contrast to the more
commonly employed Mach–Zehnder geometry whereby the signal and the
reference experience different paths, so any noise source that is specific to
one path does not occur on the other so cannot be normalized out.
The principle of the bimodal interferometer is sketched in Fig. 8. The
device starts with a single mode input. The waveguide thickness is then
changed at the mode splitter which serves to excite the first order mode.
The two modes then propagate with different propagation constants and
have a different overlap with the analyte, so they experience a different phase
delay on interaction. At the output, the two modes are made to interfere and
Fig. 8 Operation principle of a bimodal waveguide sensor.
the interference pattern represents a direct measure for the relative phase
change between the two modes (Dante et al., 2015). Bimodal waveguides
are an elegant solution because they are less complex than conventional inter-
ferometers and they have demonstrated detection limits of 2.5 107 RIU
(Herranz et al., 2017).
4.5 Concluding remarks

The various forms of optical biosensor all have benefits and drawbacks, each
taking advantage of the modal properties of light. By considering the various
modalities, it becomes clear that a higher Q value tends to dominate the
S Q FOM, however, due to biological considerations, this does not
always lead to a low LOD in terms of bioparticle detection. A high Q tends
also to reduce the sensitivity, although this reduction is small compared to
the gain in Q-factor.
5. Resonant hyperspectral imaging

A key advantage of GMR over most other sensing modalities is its
imaging functionality, whereby one can interpret each sensing element as
the pixel of an image. By its very nature, the guided mode is spatially con-
fined with a finite lateral extent, so it can be used to detect local changes in
refractive index. The lateral extent of the GMR mode can be estimated from
Bragg mirror theory (Block et al., 2009) and is expressed as a propagation
length Lp, i.e., the distance at which the field has decayed to 1/e of its original
amplitude,
λ2
Lp ¼ (4)
2πΔλ
Fig. 9 Illustration of the spatial resolution dependence on orientation.
For the TE mode shown in Fig. 2, i.e., for a center wavelength of

λ ¼ 840 nm and a linewidth of 8 nm, the equation yields a resolution of
Lp ¼ 14 μm. In practice, it turns out that the 1/e criterion is rather severe
and that much better resolution values can be achieved, with 3.5 μm reported
in Block et al. (2009) and 6 μm reported in Triggs et al. (2015). A key point to
recognize is that the resolution depends on the grating orientation, as illus-
trated in Fig. 9.
The resolution values typically quoted refer to the resolution along the
grating vector, so along the x-direction in Fig. 9, as is also clear from the
definition via the propagation length Lp. This configuration assumes
the light to be incident in z-direction and to be perfectly collimated, so
to have no wavevector components kx and ky. The grating then imparts
its grating vector G onto the wave, such that kx ¼ G, so the light is deflected
into x-direction. Therefore, the spatial resolution is determined by the grat-
ing vector and Lp in x-direction. In y-direction, there is no grating vector,
so the resolution in y-direction is typically much better and is only limited
by the diffraction limit. This observation is somewhat counterintuitive:
even though the grating is essential for generating the GMR and for provid-
ing the sensing functionality, it actually deteriorates the spatial resolution.
Additionally, the resolution depends on the refractive index contrast, with
lower contrast yielding poorer resolution (Triggs et al., 2015).
Together, these considerations highlight the trade-off between sensing
and imaging. Sensing aims to achieve high Q-factors for a high sensing
FOM S Q, while imaging requires low Q-factors for a high spatial
resolution.
Recording an image then requires three-dimensional data, i.e., the x–y
coordinates of each pixel together with its wavelength information. Any two
of these dimensions can be acquired simultaneously while the third dimen-

sion needs to be acquired sequentially for each pixel. The two main record-
ing methods described in the literature are (a) white light illumination in
combination with an imaging spectrometer, i.e., a spectrometer that dis-
perses a line of pixels and scans the image line by line (Zhuo et al., 2017)
or (b) scanning the illumination wavelength and taking full frame pictures
at every wavelength (Triggs et al., 2015). Both methods are in principle
equivalent but have different experimental requirements. Arguably, scan-
ning the illumination wavelength (b) is simpler because the imaging optics
remains static, while scanning the image (a) either requires scanning the
beam or the moving the sample, both of which need to be performed with
high accuracy.
This recording method is commonly known as hyperspectral imaging.
The resulting dataset, based on its three-dimensional nature, is known as
a hyperspectral cube. Hyperspectral imaging was first applied to GMR by
Cunningham et al., who termed it photonic crystal enhanced microscopy,
PCEM (Chen et al., 2013) although here, we prefer to use the more generic
term of resonant hyperspectral imaging.
Valuable examples for the application of resonant hyperspectral imag-
ing can be found in Chen et al. (2013) who demonstrated that it was pos-
sible to monitor cell attachment to the surface over extended periods of
time (12–24 h) with high stability. This ability to measure over extended
periods of time is important when studying stem cell differentiation and
related processes. Similarly, drug-induced cell death was monitored over
several hours and cell movement, in response to a concentration gradient
(chemotaxis), over several days. These examples highlight some of the
unique advantages of resonant label-free imaging because alternative
methods such as fluorescence have difficulty in maintaining a constant sig-
nal over extended periods of time because fluorophores bleach during long
exposures.
Similarly, GMR imaging can be used for the early detection of bio-
films. The early detection of biofilms is particularly important in an indus-
trial context, where labeled methods such as fluorescence cannot be used.
Many commercial biofilm detectors use bacterial autofluorescence, but
this requires relatively high surface concentrations to be present; instead,
the GMR approach can already see the initial formation of the protein
layer which precedes the actual biofilm formation, so it detects biofilms
at a much earlier stage (Fischer et al., 2016).
Fig. 10 A representation of how a hyperspectral dataset is used to construct a reso-

nance map from a stack of monochromatic images.
Fig. 11 Illustration of resonant hyperspectral imaging.
5.1 Sensing and imaging

The examples quoted above have illustrated the scope for using guided
mode resonances as an alternative imaging modality, mainly for monitoring
the behavior of cells. Naturally, this modality can be further enhanced by
adding the sensing functionality discussed in Section 3. Combining imaging
with sensing therefore offers a unique combination that is difficult to achieve
otherwise; other sensing methods cannot provide imaging information,
while other imaging methods do not provide sensing information. For
example, the well-established imaging method of phase contrast microscopy
is sensitive to phase changes, but it is very difficult to detect the presence of a
single protein layer. Resonant imaging can combine both functionalities, as
illustrated in Figs. 10 and 11.
Fig. 12 provides a schematic and Fig. 13 shows experimental evidence to
illustrate the combined imaging and sensing capabilities. The BHK cell in
Fig. 13 secrets thrombopoietin, a cytokine essential for maintaining physi-
ological levels of circulating platelets in blood. The purpose of the study was
Fig. 12 Schematic of combining both imaging and sensing on the same GMR platform.
Cells attached to the surface change the resonance wavelength of the hyperspectral
image from blue to red, as in Fig. 11. When suitably stimulated, they secrete protein bio-
markers that are detected via complementary antibodies immobilized on the surface;
on binding, the resulting phase change is detected via a change in the hyperspectral
image from blue to yellow.
Fig. 13 Comparison of resonant imaging vs phase contrast imaging. All three images
show a baby hamster kidney (BHK) cell secreting thrombopoietin (TPO). Panel
(A) represents the resonant image at the beginning of the experiment, when only
the cell is present (t ¼ 0 min). Panel (B) shows a resonant image combining the cell
and the secreted proteins at t ¼ 85 min. Panel (C) shows a phase contrast micrograph
of the cell for reference, which remains unchanged in the presence of the proteins. It
is clear that the resonant image can see both the cell and the proteins it secrets while
the phase contrast image can only see the cell.
to understand how TPO secretion is regulated in the human body by the

concentration of desialylated (aged) platelets in the bloodstream. More
detail can be found in Juan-Colas et al. (2018). While the cell can be more
clearly distinguished in the phase contrast image of Fig. 13C, TPO secre-
tion cannot be identified. In Fig. 13B, however, the footprint of the cell is
significantly larger, and a “halo” of resonance change emerges surrounding
the cell due to the antigen-bound TPO, thereby indicating the secretion of
proteins.
5.2 Multiparameter imaging

The methods described in the previous sections have enabled the imaging of
refractive index changes due to the presence of cells, the binding of proteins,
or both. Refractive index changes are detected via the effective index of
the guided mode by tracking the central wavelength of the resonance.
The resonance has other properties that can be used for sensing purposes,
however. In particular, its height and Q-factor are indicative of optical losses in
the system. While the ideal GMR exhibits a reflectance on resonance of 100%
(see Figs. 2 and 4), the presence of losses reduces the magnitude of the reso-
nance peak. This reduction allows us to extract more information from the
biological system. For example, areas of rapid change of density act as scattering
centers that lead to a reduction in the height of the resonance peak. Such rapid
changes of density occur around, for instance, cellular “focal adhesions,” i.e.,
the areas at the periphery of a cell associated with adhesion which tend to be
higher in molecular density than the cell body itself. The fact that the resonance
peak drops around focal adhesions was first recognized by Zhuo et al. (2018)
and is illustrated in Fig. 14 taken from Juan-Colás and Krauss (2018).
Fig. 14 makes it clear that the guided mode resonance can detect two
separate effects thereby providing more information about the biological
system than the wavelength shift alone.
Alternatively, the multiparameter capability of the GMR can provide
depth information by a combination of resonance shift and Q-factor. The
depth-dependence directly relates to the extent of the evanescent tail of the
mode as discussed in Eq. (3), so it applies as long the thickness of the object
is within the tail of the resonant mode, while the Q-factor provides informa-
tion on the refractive index. This capability was first highlighted by
Nazirizadeh et al. (2013) where the authors observed that the refractive index
of the object impacts on the observed Q-factor of the resonance, similar to the
discussion in Figs. 3 and 4; the higher the index of the cover, the lower the
index contrast seen by the guided mode, so the further its penetration depth Lp
(Eq. 4) and the higher its Q-factor. Using the Q-factor for refractive index
information showed an excellent correlation with both phase contrast and
atomic force microscopy, thereby providing another impressive example of
the multiparameter capability of the GMR approach.
5.3 Plasmonic nanohole arrays

While imaging is an exciting capability of the GMR approach, it is not a
unique capability; other platforms offer similar functionality and they are
therefore worth mentioning for context. In particular, the plasmonic
A Common cell-matrix
≡
Increase in
adhesion components refractive index
Increase in refractive
Focal adhesions (FAs) ≡
index and scattering
B Δλ (PWS)
1
Reflectance (AU)
BG Sh Δ/ (PIS)
1 ift
Cell 0.9
0.8
Reflectance (AU)
0.8
0.6
624 626 628
0.4
Wavelength (nm)
0.2
0
610 620 630 640 650
Wavelength (nm)
Fig. 14 Illustration of the separate effects of a cell on the optical resonance; while the
cell body increases the effective index of the resonant mode and shifts the resonance
compared to the background, the additional scattering at the edges of the cell respon-
sible for adhesion (focal adhesions) can be separately identified via a reduction in the
amplitude of the resonance peak. (A) Illustration of the scattering at focal adhesions
due to their rapid change in density; (B) Corresponding experimental evidence
for the refractive index change and additional scattering loss. PWS refers to peak
wavelength shift and PIS to peak intensity shift. Panel (A) taken from Juan-Colás,
J. & Krauss, T. F., 2018. Multiparameter resonant imaging for studying cell interactions.
Light Sci. Appl., 7. Panel (B) taken from fig. 1c of Zhuo, Y. et al., 2018. Quantitative anal-
ysis of focal adhesion dynamics using photonic resonator outcoupler microscopy (PROM).
Light Sci. Appl., 7(1), 9.
nanohole approach has demonstrated imaging with high sensitivity and high
spatial resolution. Plasmonic nanohole arrays fabricated in optically thick
metal films and they support the EOT phenomenon, which is attributed
to the excitation of plasmons by grating coupling of light at normal incidence;
these plasmons result in a strong field enhancement in the vicinity of the
nanoholes. This strong light confinement enables nanohole arrays to be highly
sensitive to minute local refractive index changes, such as those induced by
the binding of biomolecules to the sensor surface. In addition, the localization
of light around the nanohole provides high spatial resolution. A good example
of this capability is provided by Li et al. (2017) where the authors demonstrate
high resolution imaging (spatial resolution not quantified) and high sensitivity
sensing of VEGF, a signaling protein, directly in complex cell media. The
VEGF was secreted from cancer cells attached to the sensor surface, providing
a similar functionality as the resonant image shown in Fig. 13.
Nevertheless, the nanohole approach also has its downsides. Intrinsically,
even though the transmission is considered “extraordinary,” it is relatively
low in absolute terms, i.e., around 10%, which limits the signal to noise ratio.
In terms of practicality, EOT being observed in transmission also means that
the fluidics needs to be integrated into the optical system, i.e., source and
detector are on opposite sides of the microfluidic circuitry. In contrast, the
GMR, preferentially working in reflection, allows a clean separation between
the optical system on one side of the setup and the fluidics on the other;
working in reflection therefore mimics the inverted microscope arrangement
favored by biologists. In terms of substrate, the nanohole array tends to be
fabricated on free-standing silicon nitride membranes (Li et al., 2017), which
are fragile and expensive and therefore not suited for low-cost point-of-care
applications. Clearly, when identifying the most suitable approach for a given
application, it is important to consider all aspects ( Juan-Colás et al., 2017a;
Pitruzzello and Krauss, 2018) (Table 2).
6. Spatial-spectral equivalence
It is intrinsic to a resonance-based optical sensor that its response is
qualified as a function of wavelength; in fact, most of the figures shown
in this chapter use wavelength as the parameter to describe a particular aspect
of sensor performance. While it is easy to perform spectral characterization
in a research laboratory, doing so in the field is more difficult and it means
that a spectrometer has to be included into the system in order to provide the
readout. Spectrometers are not cheap; despite the efforts of a number of
Table 2 Comparison of different sensing modalities
Bulk sensitivity
Structure Analyte Q factor (nm/RIU) FOM (nm/RIU) LOD References
SPR (Biacore) CEA antigen 20 10 3
2 10 4
3 ng/mL Altintas et al. (2011)
GMR Different 85 510 4.3 10 4
— Huang et al. (2009)
Solutions
GMR Streptavi-din 150 88 1.3 104 1 ng/mL Cunningham et al. (2002)
1
Microrings MCP-1 2 10 4
— — 0.5 pg mL Valera et al. (2016)
and Washburn et al. (2009)a
Bmodal waveguide HCL Solutions — 6 102 πrad/RIU — 2.5 107 RIU Herranz et al. (2017)
Bimodal Waveguide hGH — 6 102 πrad/RIUb — 10 pg mL1 González-Guerrero et al.
(2017)
Nanobeam Cavity CEA antigen 1 104 70 7105 0.1 pg mL1 Liang et al. (2013)
1
Nanohole Array IgG <100 700 7 10 4
145 pg mL Li et al. (2017)
a
Multiple references used as specific details on Genalyte Maverick™ systems are not openly available.
b
Assumed to be the same as other work(s) from the same research group due to the same structure being used.
commercial companies who have made major advances in bringing low-cost

spectrometers to the market, these devices still represent a significant cost
element amounting to US$ 100s or 1000s, depending on the desired spectral
resolution. Integrated solutions based on silicon photonics processed in
CMOS fabs are emerging that challenge this view (Martens et al., 2018),
but nevertheless, adding a spectrometer into the sensor is still an overhead.
The very nature of the GMR approach offers an alternative. As high-
lighted at the beginning of this chapter, a GMR is based on the grating equa-
tion, sin θ ¼ mλ/ng, and the excitation of a resonance requires a particular
angle θ, typically close or equal to 90 degrees. In order to obtain the desired
value for θ, one can either adjust the wavelength λ or the period g. In com-
bination with the imaging capability, it is then possible to place different grat-
ings in the same field of view, each of which exhibits a different resonance
wavelength, so different locations correspond to different wavelengths; this
method translates the spectral coordinate into a spatial coordinate. The change
in period/resonance wavelength can either be discrete or it can be continuous
over the length of the grating, the latter giving rise to the idea of the “chirped
GMR” introduced by Triggs et al. (2017) and illustrated in Fig. 15.
When illuminating the chirped grating with a fixed wavelength source
such as a laser, the resonance will occur as a horizontal line in the field of
view. Upon binding of the target molecules, the refractive index increases
as illustrated in Fig. 1, which will lead to the line moving in the field of view;
the sensing information is therefore provided by the position of the line.
The same idea was pursued by Tittl et al. (2018), who opted for the dis-
crete translation of the spectral into the spatial information and refer to their
approach as “barcoding”; instead of using a 1D GMR grating, they used a
Fig. 15 Illustration of the chirped GMR concept which translates spectral dimension
into spatial dimension and combines the sensing and the spectral readout function
in a single structure. (A) Design of grating with spatially varying period in one direction.
(B) Hyperspectral image illustrating the continuous change of the resonance wave-
length as a color code. (C) Illumination with a single wavelength source yields a single
line of resonance.
2D pixelated metasurface composed of an array of high-Q resonant

metapixels with their resonance frequencies tuned by scaling their size.
The idea is the same as in the chirped GMR, namely, that different positions
in the field of view map onto different wavelengths. The authors applied this
method in the mid-IR, around 6 μm wavelength, to identify various poly-
mers attached to the surface by their vibrational absorption lines. Both the
chirped GMR and the pixelated metasurface provide impressive illustrations
of the versatility of this approach, which, ultimately only depends on the
very simple grating equation everybody learns in High School.
7. Phase-sensitive measurement
All the modalities described so far record the amplitude of the light
interacting with the biological material, i.e., the cells and proteins. We
record the amplitude of the transmission or reflection, the amplitude of
the resonance shift or the Q-factor, which is the spectral width of the
amplitude response. Nevertheless, amplitude describes only one aspect
of the optical interaction and arguably, most advanced photonic devices
make use of the phase for advanced functionality.
Recording phase requires an interferometric approach, which can
indeed yield excellent results. For example, some of the most sensitive
photonic biosensors are based on interferometry, such as the bimodal
waveguide (Section 4.3), also (Schmitt et al., 2007) which exploits the
interference between two different modes in a waveguide, or the Farfield
approach (Cross et al., 2003). These approaches can achieve exquisite
LODs of Δn < 108 RIU. Is it possible to also benefit from the phase
approach in the context of GMR?
In a way, the guided mode resonance is already a result of the in-phase
interaction between multiply scattered waves, but it is still susceptible to
amplitude fluctuations, so is ultimately limited by amplitude noise. Instead,
a truly phase-sensitive measurement is limited only by phase noise.
Phase noise tends to be significantly lower than amplitude noise for common
light sources used in biosensing, including low-cost VCSELs. The phase
noise of lasers is inversely proportional to their coherence length lc:
Δϕ λ
¼ (5)
ϕ lc
Coherence lengths of up to 10s of cm up can be assumed for typical lasers,

which results in phase noise of the order of Δφ/φ ¼ 105–106 for lasers in
the visible or near-infrared range. In contrast, the intensity noise tends to be
orders of magnitudes higher with ΔI/I 103–102 typically being
observed (Kabashin et al., 2009). For example, a typical HeNe laser exhibits
a coherence length of about 30 cm which corresponds to a phase noise of
632.8 nm/30 cm 2*106 (Thorlabs Inc., 2019a) and an RMS intensity
noise of 0.5% ¼ 5*103 (Thorlabs Inc., 2019b). The performance of a single
mode VCSEL is similar with coherence lengths of up to several meters
(Dudzik and Rzepka, 2010; Pruijmboom et al., 2008) and RMS noise of
2% typically being reported (Thorlabs Inc., 2019c).
Another argument in favor of the phase approach is illustrated in Fig. 16.
The phase changes most rapidly on resonance, where the amplitude curve
peaks. At the peak, the amplitude curve is flat, so finding the peak means find-
ing the center of a flat curve, which is difficult. Instead, the slope of the phase
curve is largest on resonance and the center is therefore much easier to find.
The third argument in favor of interferometry is slightly more subtle and
relates to the ability of minimizing the noise in the measurement. When
recording an amplitude spectrum, one can decrease the noise by averaging
over multiple spectra, so averaging in time by decreasing the bandwidth of
the measurement. When recording phase, one can average over multiple
periods of the interferogram without loosing bandwidth, which means
that it is possible to reduce noise in postprocessing without loosing time-
resolution.
7.1 Realization of phase-sensitive measurements

Phase-sensitive GMR measurements were considered theoretically as early
as 2004 (Magnusson et al., 2004). The authors predicted an enhanced
detection sensitivity when integrating the GMR into an interferometric
experimental setup and discussed the correlation between Q-factor and
phase sensitivity. The Q-factor, as discussed in the context of Fig. 4, directly
relates to the penetration depth Lp, which is analogous to the propagation
length L in an interferometric sensor. Recall that in a traveling wave-type
arrangement such as a Mach–Zehnder or a bimodal waveguide, the relation
between the active sensing pathlength L and the phase change due to effec-
tive index changes in the signal arm Δneff is given by
2πL
Δϕ ¼ Δneff (6)
λ
So the phase change is directly proportional to the propagation length L,

which is proportional to Lp for a GMR. In the traveling-wave geometry, L
can be as long as millimeters or even centimeters, while Lp is restricted to
10s or 100s of micrometers. Even though the GMR is a resonance, which
means it is a multipass device whereby the light oscillates back and forth mul-
tiple times, the overall interaction length is shorter than in the traveling-wave
approach, which explains the higher sensitivity of the interferometric sensor.
This higher sensitivity is traded-off against ease of coupling as discussed before.
Furthermore, as discussed in Section 2.2, it is questionable whether sensitivities
of Δn < 108 RIU are achievable except in the most highly stabilized labora-
tory environments, given that temperature fluctuations tend to limit the prac-
tical sensitivity achievable in the field to Δn 106–107 RIU. It is therefore
not clear whether some of the ultrahigh sensitivities reported with interfero-
metric sensors are practically useful. Nevertheless, it is an interesting question
whether interferometry can improve the performance of a GMR sensor.
Realizing a phase-sensitive interferometric measurement requires a ref-
erence wave. An obvious way of implementing such a reference wave is to
build a classical Mach–Zehnder setup whereby the beam is split into a signal
and a reference path with the GMR acting as one of the mirrors in the signal
path. This approach was first realized by Sahoo et al. (2017) who highlighted
the ability to achieve higher sensitivity by using the phase approach. In
Fig. 16, the corresponding reflectance and phase response of a GMR grating
are shown, highlighting the strong variation in phase at the point of maximal
reflectance corresponding to minimal intensity variation.
Fig. 16 Reflectance and phase response of GMR grating proposed in Sahoo et al. (2017).
Instead of placing the GMR into an external interferometer, which

is very sensitive to alignment tolerances and vibrations, it is preferable
to implement a “common-path interferometer” approach, which is
much more elegant because the signal and reference beams experience
the same geometrical path. Common-path interferometry is similar to
the Nomarski differential interference contrast method known in micros-
copy, whereby the differences in optical pathlength experienced by two
orthogonally polarized beams are exploited to enhance imaging contrast
interferometrically.
One can learn a lot about common-path interferometry from the
related SPR approach. A common-path SPR sensor was first demonstrated
in 2005 (Su et al., 2005). This sensor exploits the fact that the SPR is
polarization sensitive, i.e., only p-polarized light excites a surface plasmon
wave, while s-polarized light is reflected specularly from the gold layer.
The s-polarization therefore does not interact with the analyte so can serve
as a reference. The idea was developed further by Kabashin et al. (2009) who
demonstrated a LOD of Δn 108 RIU, which is comparable to the
traveling-wave geometries discussed above, even though the interaction
length in an SPR is much shorter.
8. Conclusion
Having now reviewed the various properties of guided mode reso-
nances for sensing applications, one can conclude that GMR arguably pro-
vides the most versatile of all optical sensing approaches. It affords high
sensitivity, albeit not the ultrahigh sensitivity achieved by other sensor
modalities, while also offering a range of other functionalities, such as imag-
ing and multiparameter sensing which are much more difficult or not pos-
sible to achieve with other modalities. Similarly, the idea of mapping the
spectral response onto the spatial domain as exemplified by the chirped
GMR approach is a unique feature that allows combining the sensor element
with the wavelength-selective readout element in the same structure; most
other optical sensing modalities require an external spectrometer. Together
with the ease of coupling to the GMR directly with a collimated beam, the
versatility and inherent simplicity make the GMR very suitable for high per-
formance, yet low-cost diagnostic devices that may be deployed in the field
and that are much needed by healthcare providers both in the developed and
in the developing world.
In terms of scientific developments, the idea of exploiting the phase

change on resonance is yet in its infancy and further improvements can
be expected, which is further evidence for the versatility of the GMR sens-
ing modality.
References
Altintas, Z., Uludag, Y., Gurbuz, Y., Tothill, I.E., 2011. Surface plasmon resonance based
immunosensor for the detection of the cancer biomarker carcinoembryonic antigen.
Talanta 86, 377–383. 30 10.
Armani, D.K., Kippenberg, T.J., Spillane, S.M., Vahala, K.J., 2003. Ultra-high-Q toroid
microcavity on a chip. Nature 421 (6926), 925.
Armbruster, D.A., Pry, T., 2008. Limit of blank, limit of detection and limit of quantitation.
Clin. Biochem. Rev. 29 (Suppl. 1), S49.
Barrios, C.A., 2012. Integrated microring resonator sensor arrays for labs-on-chips. Anal.
Bioanal. Chem. 403 (6), 1467–1475.
Bashkatov, A.N., Genina, E.A., 2003. Water refractive index in dependence on temperature
and wavelength: a simple approximation. Optical Technologies in Biophysics and Med-
icine IV. vol. 5068. pp. 393–396.
Block, I.D., Ganesh, N., Lu, M., Cunningham, B.T., 2008. A sensitivity model for predicting
photonic crystal biosensor performance. IEEE Sensors J. 8 (3), 274–280.
Block, I.D., et al., 2009. Optimizing the spatial resolution of photonic crystal label-free imag-
ing. Appl. Opt. 48 (34), 6567–6574.
Bogaerts, W., et al., 2012. Silicon microring resonators. Laser Photonics Rev. 6 (1), 47–73.
Cetin, A.E., et al., 2015. Plasmonic nanohole arrays on a robust hybrid substrate for highly
sensitive label-free biosensing. ACS Photonics 2 (8), 1167–1174.
Chen, W., et al., 2013. Photonic crystal enhanced microscopy for imaging of live cell adhe-
sion. Analyst 138 (20), 5886–5894.
Cross, G.H., et al., 2003. The metrics of surface adsorbed small molecules on the Young’s
fringe dual-slab waveguide interferometer. J. Phys. D. Appl. Phys. 37 (1), 74.
Cunningham, B., Li, P., Lin, B., Pepper, J., 2002. Colorimetric resonant reflection as a direct
biochemical assay technique. Sensors Actuators B Chem. 81 (2–3), 316–328.
Dante, S., et al., 2015. Linear readout of integrated interferometric biosensors using a peri-
odic wavelength modulation. Laser Photonics Rev. 9 (2), 248–255.
Dudzik, G., Rzepka, J., 2010. Single frequency stable VCSEL as a compact source for inter-
ferometry and vibrometry. AIP Conference Proceedings. vol. 1253 (1), pp. 317–320.
Fischer, M., Triggs, G.J., Krauss, T.F., 2016. Optical sensing of microbial life on surfaces.
Appl. Environ. Microbiol. 82 (5), 1362–1371.
Genet, C., Ebbesen, T.W., 2007. Light in tiny holes. Nature 7123, 39.
González-Guerrero, A.B., et al., 2017. Direct and label-free detection of the human growth
hormone in urine by an ultrasensitive bimodal waveguide biosensor. J. Biophotonics
10 (1), 61–67.
Herranz, S., Gavela, A.F., Lechuga, L.M., 2017. Label-free biosensors based on bimodal
waveguide (BiMW) interferometers. In: Methods in Molecular Biology. Springer,
pp. 161–185.
Hessel, A., Oliner, A.A., 1965. A new theory of Wood’s anomalies on optical gratings. Appl.
Opt. 4 (10), 1275–1297.
Homola, J., 2008. Surface plasmon resonance sensors for detection of chemical and biological
species. Chem. Rev. 108 (2), 462–493.
Hu, J., Sun, X., Agarwal, A., Kimerling, L.C., 2009. Design guidelines for optical resonator
biochemical sensors. J. Opt. Soc. Am. B 25 (5), 1032–1041.
Huang, M., Yanik, A.A., Chang, T.-Y., Altug, H., 2009. Sub-wavelength nanofluidics in
photonic crystal sensors. Opt. Express 17 (26), 24224–24233.
Juan-Colás, J., Krauss, T.F., 2018. Multiparameter resonant imaging for studying cell inter-
actions. Light Sci. Appl. 7, 45.
Juan-Colás, J., Johnson, S., Krauss, T.F., 2017a. Dual-mode electro-optical techniques for
biosensing applications: a review. Sensors (Switzerland) 17 (9), 1424–8220.
Juan-Colás, J., Krauss, T.F., Johnson, S.D., 2017b. Real-time analysis of molecular confor-
mation using silicon electrophotonic biosensors. ACS Photonics 4 (9), 2320–2326.
Juan-Colas, J., et al., 2018. Quantifying single cell secretion in real time using resonant hyper-
spectral imaging. Proc. Natl. Acad. Sci. U. S. A. 115 (52), 13204–13209.
Kabashin, A.V., Patskovsky, S., Grigorenko, A.N., 2009. Phase and amplitude sensitivities in
surface plasmon resonance bio and chemical sensing. Opt. Express 17 (23),
21191–21204.
Lalanne, P., Hugonin, J.P., Chavel, P., 2006. Optical properties of deep lamellar gratings: a
coupled Bloch-mode insight. J. Lightwave Technol. 24 (6), 2442–2449.
Li, X., et al., 2017. Plasmonic nanohole array biosensor for label-free and real-time analysis of
live cell secretion. Lab Chip 17 (13), 2208–2217.
Liang, F., et al., 2013. Scalable photonic crystal chips for high sensitivity protein detection.
Opt. Express 21 (26), 32306–32312.
Liedberg, B., Nylander, C., Lunstr€ om, I., 1983. Surface plasmon resonance for gas detection
and biosensing. Sensors Actuators 4, 299–304.
Liu, H., Lalanne, P., 2008. Microscopic theory of the extraordinary optical transmission.
Nature 452 (7188), 728.
Magnusson, R., Wang, S.S., 1992. New principle for optical filters. Appl. Phys. Lett. 61 (9),
1022–1024.
Magnusson, R., Lee, K.J., Wawro, D., 2004. Guided- mode resonance biosensors employing
phase detection. In: Diffractive Optics and Micro-Optics (p. DTuC2). Optical Society
of America.
Magnusson, R., Wawro, D., Zimmerman, S., Ding, Y., 2011. Resonant photonic biosensors
with polarization-based multiparametric discrimination in each channel. Sensors 11 (2),
1476–1488.
Martens, D., et al., 2018. Low-cost integrated biosensing platform based on SiN
nanophotonics for biomarker detection in urine. Anal. Methods 10, 3066–3073.
Martı́n-Moreno, L., et al., 2001. Theory of extraordinary optical transmission through sub-
wavelength hole arrays. Phys. Rev. Lett. 86 (6), 1114.
Maystre, D., 2012. Theory of Wood’s anomalies. In: Plasmonics. Springer, pp. 39–83.
Nazirizadeh, Y., et al., 2013. Material-based three-dimensional imaging with nanostructured
surfaces. Appl. Phys. Lett. 102 (1), 11116.
Otte, M.A., et al., 2009. Identification of the optimal spectral region for plasmonic and
nanoplasmonic sensing. ACS Nano 4 (1), 349–357.
Pfeifer, P., et al., 1999. Real time sensing of specific molecular binding using surface plasmon
resonance spectroscopy. Sensors Actuators B Chem. 54 (1–2), 166–175.
Pitruzzello, G., Krauss, T.F., 2018. Photonic crystal resonances for sensing and imaging.
J. Opt. 20 (7), 73004.
Pruijmboom, A., Schemmann, M., Hellmig, J., Schutte, J., Moench, H., Pankert, J., 2008.
VCSEL-based miniature laser-Doppler interferometer. In: Vertical-Cavity Surface-
Emitting Lasers XII. vol. 6908. International Society for Optics and Photonics, p. 69080I.
Rayleigh, L., 1907. III. Note on the remarkable case of diffraction spectra described by Prof.
Wood. In: The London, Edinburgh, and Dublin Philosophical Magazine and Journal of
Science. vol. 14. Taylor & Francis, pp. 60–65. 79.
Rosenblatt, D., Sharon, A., Friesem, A.A., 1997. Resonant grating waveguide structures.
IEEE J. Quantum Electron. 33 (11), 2038–2059.
Sahoo, P.K., Sarkar, S., Joseph, J., 2017. High sensitivity guided-mode-resonance optical
sensor employing phase detection. Sci. Rep. 7 (1), 7607.
Schmitt, K., et al., 2007. Interferometric biosensor based on planar optical waveguide sensor
chips for label-free detection of surface bound bioreactions. Biosens. Bioelectron.
22 (11), 2591–2597.
Su, Y.-D., Chen, S.-J., Yeh, T.-L., 2005. Common-path phase-shift interferometry surface
plasmon resonance imaging system. Opt. Lett. 30 (12), 1488–1490.
Thorlabs Inc., 2019a. HeNe Tutorial. [Online] Available at: https://www.thorlabs.
com/tutorials.cfm?tabID¼9F3258E0-FB6B-46D4-944D-B31A84EF1DAA [Accessed
23 11 2018].
Thorlabs Inc., 2019b. HeNe Lasers: Red. [Online] Available at: https://www.thorlabs.com/
newgrouppage9.cfm?objectgroup_id¼1516 [Accessed 23 11 2018].
Thorlabs Inc., 2019c. VCSEL Diode Module (CPS850V). [Online] Available at: https://
www.thorlabs.com/thorproduct.cfm?partnumber¼CPS850V.
Tittl, A., et al., 2018. Imaging-based molecular barcoding with pixelated dielectric meta-
surfaces. Science 360 (6393), 1105–1109.
Triggs, G.J., et al., 2015. Spatial resolution and refractive index contrast of resonant photonic
crystal surfaces for biosensing. IEEE Photonics J. 7 (3), 6801810.
Triggs, G.J., et al., 2017. Chirped guided-mode resonance biosensor. Optica 4 (2), 229.
Valera, E., Shia, W.W., Bailey, R.C., 2016. Development and validation of an immuno-
sensor for monocyte chemotactic protein 1 using a silicon photonic microring resonator
biosensing platform. Clin. Biochem. 49 (1), 121–126.
Wang, S.S., Magnusson, R., Bagby, J.S., Moharam, M.G., 1990. Guided-mode resonances
in planar dielectric-layer diffraction gratings. J. Opt. Soc. Am. A 7 (8), 1470–1474.
Washburn, A.L., Gunn, L.C., Bailey, R.C., 2009. Label-free quantitation of a cancer bio-
marker in complex media using silicon photonic microring resonators. Anal. Chem.
81 (22), 9499–9506.
White, I.M., Fan, X., 2008. On the performance quantification of resonant refractive index
sensors. Opt. Express 16 (2), 1020.
Wood, R.W., 1902. XLII. On a remarkable case of uneven distribution of light in a diffrac-
tion grating spectrum. In: The London, Edinburgh, and Dublin Philosophical Magazine
and Journal of Science. vol 4. Taylor & Francis, pp. 396–402. (21).
Yanik, A.A., et al., 2011. Seeing protein monolayers with naked eye through plasmonic Fano
resonances. Proc. Natl. Acad. Sci. U. S. A. 108 (29), 11784–11789.
Yariv, A., 1991. Optical Electronics. Saunders College Publishings.
Young, G., et al., 2018. Quantitative mass imaging of single biological macromolecules.
Science 360 (6387), 423–427.
Yu, Z., Fan, S., 2011. Extraordinarily high spectral sensitivity in refractive index sensors using
multiple optical modes. Opt. Express 19 (11), 10029–10040.
Zhuo, Y., et al., 2017. Label-free imaging of stem cell adhesion and dynamic tracking of
boundary evolution using photonic crystal enhanced microscopy (PCEM). Microsc.
Microanal. 23 (S1), 1142–1143.
Zhuo, Y., et al., 2018. Quantitative analysis of focal adhesion dynamics using photonic res-
onator outcoupler microscopy (PROM). Light Sci. Appl. 7 (1), 9.
CHAPTER SIX
Fano resonance photonic crystal

filters and modulators
Weidong Zhoua,*, Shanhui Fanb
a
b
Department of Electrical Engineering, Stanford University, Stanford, CA, United States
Contents
1. Fano resonance photonic crystal filters 149
1.1 Single-layer filters 149
1.2 Coupled double-layer filters 152
1.3 Fano resonance filter configurations 157
2. Fano resonance photonic crystal cavity modulators 168
3. Fano resonance photonic crystal cavity optomechanics 177
3.1 Optical forces in nanoscale cavities 177
3.2 Optical forces in coupled Fano resonance PCSs 178
3.3 Optomechanical control and MOMES applications 184
Acknowledgments 186
References 186
1. Fano resonance photonic crystal filters

1.1 Single-layer filters
As an example of how one might apply the theory in Chapter 1, we can
construct single-pole filters using an individual resonance. For r ¼ 1 and
therefore t ¼ 0, Eq. (1) gives the transmission spectrum
ð1=τÞ2
T¼ (1)
which exhibits a Lorentzian peak and such a configuration functions as a band-
pass filter (Fig. 1A). For r ¼ 0 and therefore t ¼ 1, the transmission spectrum
ðω ω0 Þ2
T¼ (2)

A B C
1
Transmission
0.5
0
–10 –5 0 5 10 –10 –5 0 5 10 –10 –5 0 5 10
(w –w 0)␶ (w –w 0)␶ (w –w 0)␶
Fig. 1 Single-pole filter lineshapes. ω0 and γ ¼ 1/τ are the center frequency and width
of the resonance, respectively. (A) A bandpass filter constructed by a resonance
imposed on a total reflection background. (B) A notch filter constructed by a reso-
nance imposed on a zero reflection background. (C) A Fano filter constructed by a
resonance imposed on a partial reflection background. Reproduced with permission
Wang, K. X., Liu, V., Ma, Z., 2014. Progress in 2D photonic crystal Fano resonance
features a band-rejection filter (Fig. 1B). With all intermediate background

parameters being 0 < r < 1 and 0 < t < 1, the transmission spectrum given by
Eq. (1) exhibits sharp asymmetric Fano lines (Fig. 1C). Many aspects in the
spectral response can be controlled if appropriately designed (Thurman and
Morris, 2003). Although we have been dealing with planewaves in the tem-
poral coupled-mode theory discussions, similar phenomena can be seen with
focused illumination as well (Crozier et al., 2006).
To construct double-pole filters, one should introduce an additional res-
onance in the photon transport through PCSs. This can be achieved by cou-
pling two PCSs, which we will discuss later, or by using two resonances in a
single PCS. A variety of double-pole filters can be constructed depending
on, for example, the symmetry of the two resonances and the difference
in frequency between the two resonances compared to the width of each
individual resonance. In Fig. 2, we show graphs corresponding with differ-
ent configurations. Fig. 2A is a collection of graphs displaying photon trans-
port through two orthogonal resonances, in addition to the direct pathway.
Fig. 2B displays what occurs when the two resonances are nonorthogonal.
Because it is relatively difficult to construct two nonorthogonal resonances
in a single-layer PCS, we will focus on the orthogonal case in this section and
continue the discussion of the nonorthogonal case in Section 1.2, which is
on coupled double-layer filters.
Relying on two resonances with opposite mirror symmetries in a single
PCS, the structure can serve as an all-pass transmission or a flattop reflection
filter (Suh and Fan, 2004). To have two resonances in the vicinity of a
signal frequency, one needs to choose appropriate geometrical parameters
for the photonic crystal slab. Fig. 3 shows such a configuration.
Fano resonance photonic crystal filters and modulators 151
A B
1.0 1.0
even–odd even–even
Transmission
0.8 Δw = 6g 0.8 Δw = 6g
0.6 0.6
0.4 0.4
0.2 0.2
0.0
–4 –2 0 2 4 0.0
–4 –2 0 2 4
1.0 1.0
even–even
even–odd
Transmission
0.8 Δw = 2g 0.8 Δw = 2g
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
–4 –2 0 2 4
–4 –2 0 2 4
1.0 1.0
even–even
even–odd
Δw = g
Transmission
0.8 Δw = g 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
–4 –2 0 2 4 –4 –2 0 2 4
1.0 1.0 even–even

even–odd
Δw = 0.02g
Transmission
0.8 Δw = 0.02g 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
–4 –2 0 2 4 –4 –2 0 2 4
(w – w 0)/g (w – w 0)/g
Fig. 2 Double-pole filter lineshapes resulting from two resonances. ω0 and γ ¼ 1/τ are
the center frequency and width of the resonances, respectively. Δω is the frequency
difference between the two resonances. (A) Transmission spectra through two orthog-
onal, even and odd, resonances. When Δω ≫ γ, the system behaves in essence as two
separate Lorentzian filters. At Δω ¼ 2γ, the structure is a flattop filter. When Δω ≪ γ, the
structure is an all-pass filter. (B) Transmission spectra through two nonorthogonal, both
even, resonances. The structure exhibits electromagnetically induced transparency (EIT)
lineshapes. Reproduced with permission from Suh, W., Wang, Z., Fan, S., 2004. Temporal
coupled-mode theory and the presence of non-orthogonal modes in lossless multimode
cavities. IEEE J. Quantum Electron. 40 (10), 1511–1518.
Because the resonances are orthogonal, similar to Eq. (1), the complex
amplitude transmission coefficient can be written as
γ even γ odd
t ¼1 (3)
iω iωeven + γ even iω iωodd + γ odd
Fig. 3 (A) Photonic crystal slab structure consisting of a square lattice of air holes
introduced into a high-index dielectric slab. The slab is 2.05a, thick and the radius of
the holes is 0.12a, where a is the lattice constant. The arrow represents the direction
of the incident light. (B) Schematic of a theoretical model for a resonator system that
supports two resonant modes with opposite symmetry with respect to the mirror plane
perpendicular to the incident light. Reproduced with permission from Suh, W., Fan, S.,
2004. All-pass transmission or flattop reflection filters using a single photonic crystal slab.
When γ even ¼ γ odd ¼ γ, ωeven ¼ ωodd ¼ ω0, Eq. (3) can be simplified to:
iðω ω0 Þ γ
t¼ (4)
iðω ω0 Þ + γ
and the photonic crystal slab becomes an all-pass filter. The intensity trans-
mission is uniform for both on and off resonances. However, the phase delay
varies from 0 to 2π in the vicinity of the resonance. This can be realized in
the structure if we choose a dielectric constant of 10.07, as shown in Fig. 3.
The all-pass transmission and the phase delay spectra are shown in Fig. 4.
The theoretical derivation (Eq. 4) is verified by FDTD simulations (Suh
and Fan, 2004).
When γ even ¼ γ odd ¼ γ, j ωeven ωodd j ¼ 2γ, Eq. (3) becomes
ðω ω0 Þ4
T ¼ jtj2 ¼ (5)
ðω ω0 Þ4 + 4γ 4
where ω0 ¼ (ωeven + ωodd)/2 and flattop reflection is achieved in the vicinity
of the resonance. This can also be realized in the structure shown in Fig. 3, if
we choose a dielectric constant of 10.9. In Fig. 5, we plot the transmission
spectrum. The theoretical derivation (Eq. 5) is again verified against FDTD
simulations (Suh and Fan, 2004).
1.2 Coupled double-layer filters

All the filters we have shown so far are constructed on a single PCS. These
filters can also be made with two coupled PCSs. This can be assumed
because with additional layers, PCS filters offer higher quality factor and
Fig. 4 Spectral response functions (intensity transmission and group delay) for the one-
slab structure in Fig. 3 with a dialectic constant of 10.07. Reproduced with permission
Wang, K. X., Liu, V. & Ma, Z., 2014. Progress in 2D photonic crystal Fano resonance photon-
ics. Prog. Quantum Electron. 38 (1), 1–74.
Fig. 5 Intensity transmission for the one-slab structure in Fig. 3 with a dialectic constant
of 10.9. Reproduced with permission from Zhou, W., Zhao, D., Shuai, Y.-C., Yang, H.,
Chuwongin, S., Chadha, A., Seo, J.-H., Wang, K. X., Liu, V. & Ma, Z., 2014. Progress in 2D
photonic crystal Fano resonance photonics. Prog. Quantum Electron. 38 (1), 1–74.
dispersion engineering capabilities (Shuai et al., 2013b). For example, by

spacing two PCSs together closer than the wavelength, one can construct a
structure for which the transmission is highly sensitive to the distance between
the slabs, providing single-pole Fano and Lorentzian filters (Suh et al., 2005),
as well as double-pole all-pass and flattop filters (Suh and Fan, 2003).
It is relatively straightforward to couple two resonances with the same sym-

metry in double-layer structures. Because the resonances are nonorthogonal,
the transmission coefficient can no longer be intuitively written as a sum
of the three pathways (two resonances plus the direct pathway) as in Eq. (3).
Instead, one needs to solve the more general temporal coupled-mode
equations (Suh et al., 2004). The amplitude transmission coefficient can
be calculated as
ðrd td Þ½γ 1 ðiω iω1 Þ + γ 2 ðiω iω2 Þ

t ¼ td (6)
ðiω iω1 + γ 1 Þðiω iω2 + γ 2 Þ γ 1 γ 2
where the top signs are used when both modes are even, and the bottom
signs are used when both modes are odd. In Fig. 2B, we assume that both
modes are even, and that ω1 ¼ ω0 Δω/2, ω2 ¼ ω0 + Δω/2, and γ 1 ¼ γ 2 ¼ γ.
The transmission behavior in Fig. 2B resembles that of electromagnetically
induced transparency (EIT). At the resonant frequencies ω1 and ω2, the
transmission is zero regardless of the spacing between the two resonant
frequencies; while at ω0, the transmission is at unity. This configuration
was demonstrated using coupled PCSs (Suh and Fan, 2003).
In coupled PCSs, there exists extra opportunity to engineer dark reso-
nances (Liu et al., 2009). A dark resonance is a guided resonance decoupled
from particular free space radiation modes due to symmetry mismatch.
We also refer to externally excited guided resonances as bright resonances.
A dark resonance in coupled PCSs can arise from the coupling of dark
resonances or bright resonances in individual slabs.
For two PCSs that each possess a dark resonance, bringing them close to
each other allows their resonances to couple. If both slabs are aligned such
that the proper symmetries are preserved, these coupled dark resonances
remain dark. If we shift the relative position of these two slabs to break
the symmetry, these coupled guided resonances are externally excited and
become bright. This will result in two Fano resonances in close proximity,
and the quality factors depend strongly on the lateral displacement. This type
of dark resonance is illustrated in Fig. 6.
It is also possible to couple two bright resonances together to make them
dark. For a pair of PCSs each with a bright resonance, we can decouple them
from radiation modes by placing a Fabry-Perot resonance between the two
slabs. At a frequency where the reflectivity of each individual slab is at unity,
the bright resonances lose access to external radiation. This type of dark res-
onance is illustrated in Fig. 7.
A B
1
Transmission
0
0.3 0.4 0.5 0.6 0.7
w (2pc/a)
C
1
D
Transmission
0
0.3 0.4 0.5 0.6 0.7
w (2pc/a)
E
1
F
Transmission
0
0.3 0.4 0.5 0.6 0.7
w (2pc/a)
Fig. 6 Dark states in a two photonic crystal slab system that arise from coupled dark
resonances from single slabs. (A) Single-slab transmission spectrum. (B) Mode profile
at the dark state frequency (top) and field profile at the frequency indicted by the arrow
in (A). (C) Double-slab transmission spectrum for d ¼ 0.5a, where d is the spacing and a is
the lattice constant. (D) Field profile at frequency indicated by the arrow in (C).
(E) Double-slab transmission spectrum for Δx ¼ 0.15a, where Δx is the lateral displace-
ment. (F) Field profile at lowest frequency resonance from (E). Reproduced with permis-
sion from Liu, V., Povinelli, M., Fan, S., 2009. Resonance-enhanced optical forces between
coupled photonic crystal slabs. Opt. Express 17 (24), 21897–21909.
A B
1
Transmission
0
0.3 0.4 0.5 0.6 0.7
w (2pc/a)
C
1
Transmission
0
0.3 0.4 0.5 0.6 0.7
w (2pc/a)
E
1
Transmission
0
0.3 0.4 0.5 0.6 0.7
w (2pc/a)
Fig. 7 Dark states in a two photonic crystal slab system that arise from coupled bright
resonances from single slabs. (A) Single-slab transmission spectrum. (B) Mode profile at
the dark state frequency (top) and field profile at the frequency indicted by the arrow in
(A). (C) Double-slab transmission spectrum for d ¼ 0.5a, where d is the spacing and a is
the lattice constant. (D) Field profile at frequency indicated by the arrow in (C).
(E) Double slab transmission spectrum for d ¼ 0.65a. (F) Field profile at peak of reso-
nance in (E). Reproduced with permission from Liu, V., Povinelli, M., Fan, S., 2009.
Resonance-enhanced optical forces between coupled photonic crystal slabs. Opt. Express
17 (24), 21897–21909.
1.3 Fano resonance filter configurations

1.3.1 Single-layer Fano resonance filters
Following initial experimental demonstrations (Kanskar et al., 1997) and
theoretical investigations (Fan and Joannopoulos, 2002; Peng and Morris,
1996; Suh and Fan, 2004), 2D PCS Fano resonance filters have been
reported from a few other groups (Fan and Joannopoulos, 2002; Lin
et al., 2005; McKerracher et al., 2013; Rosenberg et al., 2005).
Over the last few years, based on the PDMS transfer printing techniques,
many different types of Fano resonance filters have been reported, based on
single- or double-layer PCSs on Si, glass, and flexible substrates (Fig. 8). The
key lattice parameters are the air hole radius (r), the lattice period (a),
Fig. 8 Schematics of Fano resonance PCS filter configurations: (A) Single-layer PCS on
glass (top) or flexible (bottom) substrates; (B) double-PCS Fano filter 3D sketch; (C) Key
parameters defined for the square lattice double-layer PCS filter including lattice con-
stant (a), air hole radius (r), thin oxide buffer layer thickness (tb), and two single layer PCS
layer thicknesses (t1 and t2); (D) 3D close-up view of double-PCS Fano filters with lattice
displacement; and (E) Illustration and definition of key design parameters, including lat-
tice displacement (Δx, Δy) between two layers.
thicknesses for the top (t1) and bottom Si PhC layers (t2), and the oxide
buffer layer in between (tb).
Yang et al. (Qiang et al., 2008) reported the first experimental surface-
normal Fano filters based on transferred silicon nanomembranes on glass
and on flexible PET substrates. Subsequently, with improved designs
and processes, single-layer PCS Fano filters have been demonstrated with
much improved performance. Shown in Fig. 9A and B are scanning
electron microscope (SEM) images of the fabricated single-layer Fano
filters on glass substrate. The measured (blue solid line) and simulated
(red dashed line) transmission spectra are shown in Fig. 9C, with enlarged
images shown in Fig. 9D. Two transmission dips were found at 1529.88
and 1564.62 nm. For the 1564.62 nm dip, the Q value was 1737, with a
26 dB extinction ratio.
A B
C D
Transmission power (dBm)
–25 –25
–30 –30
–35 –35
–40 –40
–45 –45
–50 –50
Measured Measured
–55 –55
Simulated Simulated
1520 1530 1540 1550 1560 1570 1580 1560 1562 1564 1566 1566
Wavelength (nm) Wavelength (nm)
Fig. 9 Experimental results for Design S1: (A) Top and (B) cross-section views of fabri-
cated single-layer PhC Fano resonance filters on oxide buffer; (C) Measured (blue solid
line) and simulated (red dash line) transmission spectra for the fabricated single-layer
PhC Fano resonance filter transferred on glass substrates; and (D) Enlargement of
(C) over the second dip (λ ¼ 1564.62 nm) region. Reproduced with permission from
Shuai, Y., Zhao, D., Tian, Z., Seo, J., Plant, D. V., Ma, Z., Fan, S., Zhou, W., 2013b. Double-
layer Fano resonance photonic crystal filters. Opt. Express 21 (21), 24582–24589.
The angle and polarization properties of these filters have been investi-
gated in detail by Qiang et al.(Qiang et al., 2008) and Li et al. (Chen et al.,
2009), focusing on measurement and simulated dispersion properties.
Detailed analysis of the angle-dependent Fano resonant mode dispersion
properties is presented in Fig. 10. Following the conventional definition,
the light incident angles are specified by two polar angles, the colatitude
angle θ (an angle from the surface-normal direction) and the azimuth
angle ϕ (an angle from the positive x-axis to the orthogonal projection of
the incident beam in the x–y plane) (Lousse et al., 2004). The incident
beam with the wave-vector k was kept within the Γ–X direction in the
k-space, i.e., for the case shown in Fig. 10A with ϕ ¼ 0°, the momentum
A B
q = 0° 10° 20° 30° 40° 1400 1
Normalized frequency (a/l)
0.42
Wavelength (nm)
1450 0.8
1500 0.6
0.40
1550 0.4
0.38 0.2
Experiment 1600
Simulation
1650 0
0.36 0 5 10 15 20
0 0.08 0.16 0.24 0.32
Incident angle (degree)
Kx (2p /a, G → X)
C D
q = 40° 30° 20° 10° 0° 1400 1
Normalized frequency (a/l)
0.42
Wavelength (nm)
1450 0.8
1500 0.6
0.40
1550 0.4
0.38
Experiment 1600 0.2
Simulation
0.36 1650 0
20 15 10 5 0
1.40 1.46 1.52 1.58 1.65 1.71
Incident angle (degree)
Kx (2p /a, M → G )
Fig. 10 (A, C) Simulated dispersion plot (red lines) zoomed-in along (Γ–X, Γ–M) direc-
tions, along with measured Fano resonances (blue dots) for different incident angles θ
with (A) ϕ ¼ 0° and Ψ ¼ 0°, and (C) ϕ ¼ 45° and Ψ ¼ 45°. The straight gray lines represent
the relationships between the normalized frequency and the normalized wave-vectors
along (Γ–X, Γ–M) directions. (B, D) Measured transmission intensity contour plots for
angle-dependent transmissions at different incident angles θ with (B) ϕ ¼ 0° and
Ψ ¼ 0°, and (D) ϕ ¼ 45° and Ψ ¼ 45°. Reproduced with permission from Qiang, Z.,
Yang, H., Chen, L., Pang, H., Ma, Z., Zhou, W., 2008. Fano filters based on transferred silicon
nanomembranes on plastic substrates. Appl. Phys. Lett. 93, 061106.
vector kx relates to the normalized momentum kΓ–X along the Γ–X direction

as kx ¼ kΓx 2πa cosϕ. Thus, the normalized frequency (a/λ) can be related

to the normalized momentum kΓ–X in the dispersion plot as kΓx ¼ λa sinθ.
A set of straight lines can be superimposed on the normalized dispersion
plot along the Γ–X direction, as shown in Fig. 10A. The measured Fano
resonant modes were normalized and plotted in Fig. 10A, blue circles.
Excellent agreement was achieved between the experimental and simulation
results. The transmission intensity contour plots for the measured angle-
dependent transmission spectra, with ϕ ¼ 0°, are also shown in Fig. 10B,
where the intensity dips/peaks are shown in blue and red. The intensity con-
tour plot clearly resembles the spectra shifting behavior shown in Fig. 10A.
Following a similar procedure, the measured transmission results shown
in Fig. 10B, with ϕ ¼ 45°, were analyzed such that wave vector k lies in
the Γ–M plane, with results shown in Fig. 10C and D. Again, the experi-
mental and simulation results agree very well. It was confirmed that
angle-independent filters are feasible based on the dispersion engineering
of the Fano resonances in PCS. Details can be found in a short review by
Zhou et al. (2009).
1.3.2 Double-layer Fano resonance filters

On the other hand, much higher Q’s can be achieved in coupled double-
layer Fano resonance PhC structures. Liu and Fan et al. (Liu et al., 2009)
reported earlier that much higher Q’s can arise from the coupled dark
states in double-layer stacked PCS structures. Suspended double-layer
PCS filters have also been experimentally demonstrated in GaAs (Stomeo
et al., 2010), InP (Roh et al., 2010), and Si (Shuai et al., 2013b) materials.
Both single- and double-layer Fano resonance Si PhC filters were designed
and optimized for high Q’s around 1550 nm on the spectral band. Shown in
Table 1 are key design parameters for a few optimized design structures, where
S1–S3 are for single-layer designs and D1–D3 are for double-layer designs.
Notice that the Fano filter Q increases from 4500 for single-layer filter to
over 98,000 for a double-layer filter. Shown in Fig. 11 are the simulated trans-
mission spectra for Designs S3 and D3, with Q’s of 4500 and 98,000, respec-
tively (Shuai et al., 2013b).
Additionally, it was predicted that the double-layer PhC structure could
produce an extremely high Q mode (infinite in theory), by varying the cou-
pling condition between two PhC layers (Liu et al., 2009). Based on the
Design D2 parameters, transmission spectra were simulated by varying the
buffer layer oxide thickness tb.
Table 1 Key design parameters and Q’s for selected single

(S1–S3)- and double (D1–D3)-layer filters
a r/a t1 t2 tb Q
Unit nm nm nm nm
S1 765 0.08 260 – – 4100
S2 765 0.1 260 – – 1900
S3 780 0.08 260 – – 4500
D1 1000 0.2 230 230 160 1.2 104
D2 1000 0.08 230 230 20 2.2 104
D3 1000 0.05 230 230 20 9.8 104
A B
100 100
Transmission (%)
Transmission (%)
80 80
60 60
40 40
20 20
S3 S3 Q = 4500
0 0
1350 1450 1550 1650 1570 1580 1590 1600
C D
100 100
Transmission (%)
Transmission (%)
80 80
60 60
40 40
20 20
D3 D3 Q = 98,000
0 0
1450 1500 1550 1600 1650 1566.1 1566.6 1567.1 1567.6
Fig. 11 Simulated transmission spectra for (A, B) single- and (C, D) double-layer Fano
resonance PhC filters, where (B) and (D) are enlarged plots of (A) and (C), respectively.
The design parameters are summarized in Table 1 for Case S3 and D3, respectively.
Reproduced with permission from Shuai, Y., Zhao, D., Tian, Z., Seo, J., Plant, D. V.,
Ma, Z., Fan, S., Zhou, W., 2013b. Double-layer Fano resonance photonic crystal filters.
Opt. Express 21 (21), 24582–24589.
For the double-layer structure, the simulated transmission spectra are

plotted in Fig. 12A, with oxide buffers ranging from 0 to 160 nm thick. With
the increased oxide buffer layer thicknesses, the high Q modes (shorter
wavelength modes shown in Fig. 12A shift toward shorter wavelengths,
with the filter Q value maximizing around 10,000,000 for a buffer layer
of tb ¼ 60 nm, as shown in Fig. 12B. Shown in Fig. 12C is the magnified
spectral plot for the transmission dip with a Q of 10,000,000. Simulated field
distribution profiles for three cases close to the maximum Q are shown in
Fig. 12D, where strong field confinement is evident for the high Q trans-
mission dips at the optimal buffer layer thickness.
Two types of structures were prepared for double-layer PhC Fano filters
on silicon and on quartz substrates. For the double-layer PhC Fano filters on
A B
7
tb = 0 nm Q ~ 5K 1600 l 10
8
Q ~ 22K Q
Resonant l (nm)
20 nm
1570 106
Transmission
Q ~ 110K 40 nm
6
Q ~ 1e7
Q factor
60 nm
1540
4
Q ~ 125K 80 nm 105
Q ~ 37K 100 nm 1510
Q ~ 21K 120 nm
2 104
Q ~ 15K 140 nm 1480
Q ~ 12K 160 nm
0 1450 103
1450 1500 1550 1600 1650 0 40 80 120 160
Wavelength (nm) Buffer thickness (nm)
C D
1
tb = 20 nm
0.8 l = 1559 nm
Transmission
0.6
tb = 60 nm
0.4 l = 1515 nm
0.2
tb = 60 nm Q = 107 tb = 100 nm
0
1515.277 1515.279 l = 1489 nm
Wavelength (nm)
Fig. 12 Simulation results for Design D2 with different buffer layer thicknesses tb:
(A) Transmission spectra with different tb from 0 to 160 nm; (B) High Q resonant wave-
lengths and the corresponding Q values for different buffer thicknesses tb; (C) Zoom-in
spectrum for the buffer thickness tb ¼ 60 nm and filter Q of 10,000,000; and
(D) Simulated E-field intensity profile at resonant wavelengths for three different tb
values, where tb ¼ 60 nm representing the highest Q condition for this design.
Opt. Express 21 (21), 24582–24589.
SOI, a low index oxide buffer layer was first formed using thermal oxidation
of a single-crystalline Si layer on the SOI substrate, followed by a low pres-
sure chemical vapor deposition (LPCVD) poly-Si deposition process, to
form a poly-Si/thermal SiO2/crystalline-Si double-Si-layer structure. The
other double-layer PhC Fano filters on quartz substrates were formed by
two rounds of the LPCVD poly-Si deposition process, with a plasma-
enhanced chemical vapor deposited (PECVD) SiO2 layer sandwiched in
between the two LPCVD poly-Si layers. Shown in Figs. 13A and 5C are
cross-sectional SEM images for the double-layer poly-Si/SiO2/c-Si filter
structures, with different thermal oxide thicknesses. The same E-beam pat-
terning and etching processes were utilized for the 2D-PhC patterning. An
SEM image is shown in Fig. 13D.
Fig. 13 Cross-sectional SEM images for fabricated double-layer PhC Fano resonance
filters based on Design D2 parameters: (A–C) Double-layer PhC structure was formed
by poly-Si deposition on top of the SOI substrates; and (D) Double-layer PhC structure
was formed by two steps of poly-Si deposition on quartz substrates. Notice the oxide
buffer thicknesses are 20, 160, and 20 nm for cases (B), (C), and (D), respectively.
Opt. Express 21 (21), 24582–24589.
The double-layer Fano filters were characterized by measuring trans-

mission or reflection spectra for the two different configurations. For the
transmission measurement on the double-layer filter quartz substrate, a trans-
mission dip at 1545.2nm was obtained, with an estimated Q of 5000 and an
extinction ratio of 2.7 dB (Figs. 14A and 6B). For the reflection measurement
on the double-layer filter on SOI substrate, a reflection peak was obtained at
1567 nm with a Q factor of 9734 and an extinction ratio of 8 dB (Figs. 14C
and 6D). All of these measured spectra match well with the simulated ones at
these resonance locations (with spectral dips or peaks). The measured Q values
are lower than the simulated ones, which may be due to the imperfect process
of etching the air holes, such as the conical shape and different hole sizes in
the trilayer structure. We expect the filter performance to be improved with
much higher Q factors when we optimize the fabrication process.
A B
–9 –9
–10 –10
–11 –11
Measured Measured
–12 –12 On quartz
Simulated On quartz Simulated
1530 1540 1550 1560 1543.7 1544.7 1545.7 1546.7
C D
–12 –12
Measured
Reflection power (dBm)
On SOI
Reflection power (dBm)
–14 On SOI Simulated

–14
–16
–18 –16
–20
–22 –18
–24
–26 –20
–28 Measured
–30 Simulated –22
1500 1525 1550 1575 1600 1625 1650 1565 1566 1567 1568 1569 1570
Fig. 14 (A, B) Measured (blue solid line) and simulated (red dash line) transmission spec-
tra for the double-layer PhC Fano resonance filters on quartz; and (C, D) Measured (blue
solid line) and simulated (red dash line) reflection spectra for the double-layer PhC Fano
resonance filters on SOI. (B) and (D) are magnifications of (A), (C), respectively.
Opt. Express 21 (21), 24582–24589.
1.3.3 Double-layer Fano resonance filters with controlled lattice

displacement
It was theoretically understood that allowing lateral displacement of the lat-
tices in the two layers with respect to each other enables an additional degree
of freedom in controlling the filter characteristics of the two-layer structure
(Liu et al., 2009; Suh et al., 2003). However, experimental demonstration of
coupled double-layer PCS structures with controlled lattice displacement
has never been reported, due to the challenges associated with the fabrica-
tion. But, by employing the PDMS (polydimethylsiloxane) nanomembrane
transfer printing technique, it is possible to stack multiple layers of semicon-
ductor and other dielectric materials (Rogers et al., 2011; Yang et al., 2012).
Recently Shuai et al. (2013a) reported experimental demonstration of
coupled double-layer PCS with precisely controlled lattice displacement,
based on transfer printing and multilayer e-beam patterning alignment pro-
cesses (Rogers et al., 2011; Zhang et al., 2012). Both simulation and mea-
surement suggest that optical filter Q factors are very sensitive to the lattice
displacement. For one structure with relatively large lattice displacement,
they demonstrated a Q factor of 80,000 experimentally. Further design
and process optimization leads them to expect much higher Q-factors, par-
ticularly due to a simulation Q factor of 2.11 108 for coupled double-PCS
structures with optimal lattice displacement (Shuai et al., 2013a).
Shown in Fig. 15 are simulated transmission and reflection spectra for
double-layer stacked Fano filters without (Fig. 15A and C) and with
(Fig. 15B and D) lattice offset (Δx ¼ 0.2a, Δy ¼ 0.5a), with the insets show-
ing the magnified plot for the dominant (highest Q) resonance in each case.
Also shown in the insets are the schematics of the structures considered in the
simulation. For the transmission spectra shown in Fig. 15A and B, a glass
substrate was used. For the reflection spectra shown in Fig. 15C and D,
an SOI substrate was employed. The design parameters are a ¼ 1000 nm,
r/a ¼ 0.08, t1 ¼ t2 ¼ 230 nm, and tb ¼ 20 nm. Simulations were carried out
to find the Q factors for the different lattice displacements. The results shown
in Fig. 15B and D correspond with the highest Q value obtained for this set
of design parameters. Notice that the Fano filter Q increases from 26,000 and
28,000 for perfectly aligned lattice structures (Fig. 15A and C) to 2.11 108
and 1.76 108 for lattice misaligned structures (Fig. 15B and D) with opti-
mal lattice displacement being Δx ¼ 0.2a and Δy ¼ 0.5a. It is also worth
mentioning that the low index buffer layer thickness (e.g., SiO2 used here)
can also be varied for Q-factors approaching infinity (Liu et al., 2009; Shuai
et al., 2013b).
100 100 y = 90°

y = 0°
Transmission (%)
Transmission (%)
80 80
60 60 100
100
80
80
40 60
40 60
40
40
20 T 20 20
T
20
0
0 1559.3764 1576.3765
1559 1559.5 15560
0 0
1450 1500 1550 1600 1650 1450 1500 1550 1600 1650
(A) Wavelength (nm) (B) Wavelength (nm)
100 100 100

100
80
80
80
60
R 80 60
Reflection (%)
Reflection (%)
40
R
40
20 20
60 0
60
0
1559 1559.5 15560 1576.3764 1576.3765
40 40
20 20 y = 90°
0 y = 0°
0
1450 1500 1550 1600 1650 1450 1500 1550 1600 1650
(C) Wavelength (nm) (D) Wavelength (nm)
Fig. 15 Simulated transmission and reflection spectra for double-layer stacked
Fano filters with (B, D) and without (A, C) lattice offset (Δx ¼ 0.2a, Δy ¼ 0.5a), with
the insets showing enlarged plots for the dominant (high Qs) resonance for each cases.
Also shown in the insets are the schematics of the structures considered in the sim-
ulation. For the transmission spectra shown in (A, B), glass substrate was considered.
For the reflection spectra shown in (C, D), SOI substrate was considered. Ψ represents
the incident beam polarization (E-vector) defined as the angle from the positive x-axis
to the polarization direction. Reproduced with permission from Shuai, Y., Zhao, D.,
Chadha, A. S., Seo, J.-H., Yang, H., Fan, S., Ma, Z., Zhou, W., 2013a. Coupled double-layer
Fano resonance photonic crystal filters with lattice-displacement. Appl. Phys. Lett. 103
(24), 241106.
Following a short thermal oxidation process for the formation of a thin

SiO2 buffer layer (tb ¼ 20 nm) on top of the silicon-on-insulator (SOI) sub-
strate, single-layer Fano filters were fabricated with standard electron-beam
lithography and dry etching processes. An unpatterned single crystalline Si
nanomembrane was then stacked onto the patterned single-layer Fano filter
using the PDMS stamp-assisted membrane transfer printing process. Filters
with different displacements were fabricated. Shown in Fig. 16A–C are scan-
ning electron microscope (SEM) images of fabricated double-layer Si-NM
Fano filters with displaced lattices between the coupled layers. Shown in
Fig. 16 Fabricated Fano filter scanning electron micrograph (SEM) images: (A, B) Top
view and (C) cross-sectional view of double-layer Fano filter on SOI, with controlled lat-
tice misalignment offset (Δx ¼ 0.2a, Δy ¼ 0). Reproduced with permission from Shuai, Y.,
Zhao, D., Chadha, A. S., Seo, J.-H., Yang, H., Fan, S., Ma, Z., Zhou, W., 2013a. Coupled double-
layer Fano resonance photonic crystal filters with lattice-displacement. Appl. Phys. Lett. 103
(24), 241106.
Fig. 17 are the results obtained for the coupled double-layer PCSs, with large
lattice displacement (Δx ¼ 0.495a, Δy ¼ 0.45a). A measured reflection spec-
trum is shown in Fig. 17A, along with a top SEM image shown in the inset.
We obtained two resonant locations, around 1548.6 and 1567 nm, with a
magnified plot around 1548.6nm resonant shown in Fig. 17B. Again, simu-
lation and Fano fit were carried out and the results match the experimental
results well. In this case, we obtained a Q factor of 80,000. This value is the
highest experimental Q factor reported so far, for all the different cases we fab-
ricated. While the value is still far from the optimal Q factor of 2.11 108, the
results demonstrate the great potential of coupled double-layer PCS structures
with controlled lattice displacement. During testing, we also noticed that the
performance of these filters is very sensitive to incident beam direction, colli-
mation, and polarization orientation. We anticipate much higher Q-factors
can be obtained with precisely controlled displacement and an optimized
fabrication process.
A B
0
–4
Y = 0° Y = 0°
–2
Reflection (dB)
Reflection (dB)
–6
–4
–8
–6
–10 –8 Measured
1 mm Δx = 0.495a Simulated
–12 Δy = 0.45a Fano fitted
–10
1540 1550 1560 1570 1548 1548.5 1549
Fig. 17 Measured displaced Fano filter (Δx ¼ 0.495a, Δy ¼ 0.45a) reflection spectra for
p-polarization state, with theoretical and experimental Q-factors of 115,000 and 72,000,
respectively. Shown in the inset is the magnification of the measured reflection spec-
trum. Reproduced with permission from Shuai, Y., Zhao, D., Chadha, A. S., Seo, J.-H.,
Yang, H., Fan, S., Ma, Z., Zhou, W., 2013a. Coupled double-layer Fano resonance photonic
crystal filters with lattice-displacement. Appl. Phys. Lett. 103 (24), 241106.
2. Fano resonance photonic crystal cavity modulators

The basic idea of the modulator discussed here is illustrated in Fig. 18
(Shuai et al., 2017). Two 230 nm thick silicon photonic crystal slabs, each
patterned with an identical square lattice of circular air holes, are separated
by a 20 nm layer of SiO2. The silicon slabs are p- and n-doped, with the
oxide layer providing electrical isolation between them. The lattice period-
icity “a” is 1 μm and the hole radius “r” is 60 nm. These parameters are cho-
sen so as to offer a high quality optical resonance in the vicinity of 1550 nm.
The resonance can be excited by surface-normal incident light and the
resulting field distribution can be calculated using rigorous coupled-wave
analysis. Fig. 18B illustrates the electric field distribution of the high-Q res-
onance, calculated for our parameters using the free software package S4 (Liu
and Fan, 2012).
Electrical coupling is enabled by the plasma dispersion effect (Chmielak
et al., 2011; Jacobsen et al., 2006; Soref and Bennett, 1987), in which the
concentration of the free carriers in silicon changes the real and imaginary
parts of the refractive index. This mechanism was studied extensively
in silicon (Soref and Bennett, 1987). The doping configuration employed
in our device is that of a PIN structure, comprised of a p-doped
(P) bottom layer of crystalline silicon, an electrical isolation (I) layer of
SiO2, and an n-doped (N) top layer of polycrystalline silicon (Fig. 18C).
A
R
T
X1 (Center) X2 (Edge)
C
B d
1 Cathode
60 nm N+ = 1 × 1019 cm–3 X
a 2r y2
170 nm N = 1 × 10 17
cm –3
y1
y1
tSiO2 P = 1 × 10 cm
17 –3
y2 170 nm
Ṽ 60 nm P+ = 1 × 1019 cm–3
–1 Anode Anode
D
Y
Fig. 18 A 3D sketch of stacked double-layer Photonic Crystal Slabs (PCSs) of two-

dimensional square lattice holes’ array was shown in (A). Light incidence from top
was reflected or transmitted as the arrows in red. A high resonance mode excited by
surface normal incidence was built in the stacked double-layer PCSs. The electrical field
intensity was monitored in color as defined from “1” to “1” as shown in (B). In simplified
electrical simulation 2D model (C), electrical bias was applied to the capacitance-like
stacked bilayer PCSs structure for electro-optical modulation. Device lateral size is
“D” ¼ 20 μm and the electrodes widths are set to be “d” ¼ 5 μm. Total thickness of top
or bottom Si layer is 60 nm + 170 nm ¼ 230 nm as shown in (C). The “tSiO2” is thickness
of SiO2 between two Si layers. In the X–Y plane, we set a few monitoring positions: “x1”
center and “x1” edge of the device; “y1” near Si/SiO2 interfaces and “y2” at middle region
of the Si layers. Reproduced with permission from Shuai, Y.-C., Zhao, D., Liu, Y.,
Stambaugh, C., Lawall, J., Zhou, W., 2017. Coupled bilayer photonic crystal slab electro-
optic spatial light modulators. IEEE Photonics J. 9 (2), 1–11.
The SiO2 layer provides electrical isolation between the doped layers, so that
the device functions electrically as a parallel-plate capacitor. When a poten-
tial difference is imposed between the slabs, the accompanying charge accu-
mulation changes their refractive indices, modifying the optical frequency of
the high-Q mode. For a fixed optical frequency, the transmission and reflec-
tion are both affected, with slope efficiency proportional to the optical qual-
ity factor. We modeled the electrical response of our device in 2D (X–Y)
using Taurus Medici (Synopsys) (Synopsys, 2016), a semiconductor device
simulator. In the simulations, the doping level was taken to 1019 cm3 at the
60 nm outer regions of the silicon slabs and 1017 cm3 at the 170 nm inner
regions (Fig. 18C). The device width was taken to be D ¼ 20 μm limited by
the allowed number of grid nodes, and the electrode width to be d ¼ 5 μm, as
illustrated in Fig. 18C. The nonuniformity in the electric charge distribution
owing to the etched holes was ignored.
Optimized device geometry and Si/SiO2/Si layers’ thicknesses are deter-
mined by systematic investigation and analysis considering many factors in
our device model. In the proposed Si/SiO2/Si PIN structure, different sand-
wiched SiO2 thicknesses and Si doping levels are compared by combination
of optical and electrical analysis. Sandwiched SiO2 thickness of 20 nm was
chosen to make a balance between optical and electrical effects. Multiple fac-
tors were considered such as: its desired optical Q value, stacked Si/SiO2/Si
layers RC effect, free carrier absorption in Si, effective refractive index
changes, sufficient electrical isolation between Si layers as well as process
control preference.
In our optical study, we started from a transmission configuration as in
Fig. 19A. Different SiO2 thicknesses offer Q-factors varying from 5000 to
15,000 as shown in Fig. 19B. Two SiO2 thickness cases are compared in
Fig. 19C and D when applying forward bias, the 10 nm oxide case shows
larger spectral shifting than 20 nm oxide case; however, strong free carrier
absorption starts to occur with decreased oxide thickness, resulting in
reduced Q factor and extinction ratio. To avoid unintended free carrier
absorption and possible thermal effect, 20 nm oxide thickness is determined
for later reflection configuration design. It was noticed that, in Fig. 19E, this
interfaces accumulation (as “y1” positions, in Fig. 18C) produces a larger
blue shifting on spectra for the resonance mode when comparing to free car-
rier accumulation appearing in the middle of each Si layer (as “y2” positions,
in Fig. 18C). Owing to the semiconductor carriers equilibrium in these two
stacked Si PCS, our proposed P/N doping structure (Fig. 18C) can be mod-
ulated by either forward or reverse bias for blue or red spectral shift, respec-
tively, as shown in Fig. 19F. By applying the carrier refraction (Soref and
Bennett, 1987) and semiconductor carriers equilibrium np ¼ ni2, one could
imagine there would be a returning point of red-shifting with increased
reverse bias voltage and this returning point was observed during device
characterization.
A reflection configuration was further investigated by optical simulation.
In Fig. 20A, a 2D sketch shows the device geometry design for two kinds of
device sizes. It shows our electrical bias convention. We applied 1 μm buffer
oxide layer and 380 nm PECVD oxide for effective reflection spectra as well
as for electrical interconnection purposes. Simulation results show a dra-
matic spectral shifting in Fig. 20B when applying electrical bias. It reveals
A B C
Transmission configuration 100
tSiO2 = 20 nm Q ~ 15 K
Transmission (%)
4 80
tSiO2 = 15 nm Q ~ 15 K
Transmission
Si
60
SiO2 tSiO2 = 10 nm Q ~ 12 K
Q ~ 30 K
Si 2 40 tSiO2 = 10 nm
tSiO2 = 5 nm Q ~ 10 K
Q ~ 15 K 0V
20 +5 V
SiO2 substrate tSiO2 = 0 nm Q ~ 10 K Q~5 K
+10 V
0 0
1450 1500 1550 1600 1650 1572 1573 1574 1575
D E F
100 100 100
80 tSiO2 = 20 nm 80 tSiO2 = 20 nm 80 tSiO2 = 20 nm

Transmission (%)
Transmission (%)
Transmission (%)
60 0V 60 60 0V
+5 V +5 V
40 +10 V 40 Δn<0 40 –5 V
0V
(y1) y1
20 20 20
Δn<0 y2
Δn=0
(y2)
0 0 0
1558 1559 1560 1561 1558 1559 1560 1561 1558 1559 1560 1561
Wavelength (nm) Wavelength (nm) Wavelength (nm)
Fig. 19 Device electro-optical (EO) study in transmission configuration: a 2D sketch of the simulation lattice is shown in (A); for lattice constant
“a” ¼ 1 μm and “r/a” ¼ 0.08, the transmission spectral centered at 1550 nm are shown in (B) with different sandwiched SiO2 thicknesses varying
from 0 to 20 nm. Two different sandwiched SiO2 thicknesses, 10 nm in (C) and 20 nm in (D), are also compared for EO spectral shifting; Spectral
shifting by free carrier accumulation occurring at either middle or interfaces of Si layers is compared in (E), by 5 V forward bias. Spectral shifting
is also compared by 5 V forward/reverse bias, as shown in (F). Reproduced with permission from Shuai, Y.-C., Zhao, D., Liu, Y., Stambaugh, C.,
Lawall, J., Zhou, W., 2017. Coupled bilayer photonic crystal slab electro-optic spatial light modulators. IEEE Photonics J. 9 (2), 1–11.
A
~
350 or 80 µm V
300 or 50 µm
poly-Si n
380 nm
PECVD p
c-
oxide
Si
Buffer oxide
Silicon substrate
B
60
0V Q ~ 15 K
50 +5 V
+10 V
Reflection (%)
40
–5 V
30
20
10
0
1560 1561 1562
Wavelength (nm)
Fig. 20 (A) A sketch of device geometry cross-sectional view shows device design with
two kinds of lateral dimension (350 μm mesa with 300 μm PC; 80 μm mesa with 50 μm
PC) that are investigated in experiments. It has the same sandwiched SiO2 thickness, “r,”
“a,” Si layer thickness and doping profiles as those applied in transmission configuration
(Fig. 19D and F). The simulated reflection spectra with different biases are studied in
(B) with correspondence to transmission configuration. Reproduced with permission from
Shuai, Y.-C., Zhao, D., Liu, Y., Stambaugh, C., Lawall, J., Zhou, W., 2017. Coupled bilayer
photonic crystal slab electro-optic spatial light modulators. IEEE Photonics J. 9 (2), 1–11.
that the free carrier absorption-induced extinction ratio and Q-factor

decreasing is more critical in reflection configuration.
For experimental demonstration, we fabricated devices with a somewhat
larger lateral size than the simulated one (20 μm lateral size) for better
surface-normal light coupling from free space and to facilitate optical align-
ment during characterization. In practice, we applied two lateral sizes for
individual device. Photolithography and reactive ion etching defined a
350 μm 350 μm (or 80 μm 80 μm) mesa area inside of which e-beam
lithography defined a 300 μm 300 μm (or 50 μm 50 μm) photonic crys-
tal. Fabrication started with a SOI wafer with a 260 μm silicon thickness.
A thermal oxide layer was grown on top of the Si. The growth process
consumed a 30 nm layer of silicon from the surface, leaving 230 nm. The
thermal grown SiO2 layer was subsequently thinned down to 20 nm by
means of a HF wet etching. A polycrystalline silicon layer of 230 nm thick-
ness was deposited atop the thermal oxide layer by means of low-pressure
chemical vapor deposition. The stacked bilayers were subsequently
boron/phosphorous (p/n) doped by ion implantation into different depths,
followed by rapid thermal activation. The array of holes for the photonic
crystals was then patterned by electron-beam lithography on a 60 nm Cr
layer that is used as a hard mask. The stacked poly-Si/SiO2/Si layers were
etched through to the bottom buffer oxide layer by reactive ion etching,
after which the Cr mask was chemically removed. Finally, a 380 nm SiO2
layer was deposited on top by PECVD for interconnection.
Fig. 21A shows a microscope image of a 6 6 array of devices fabricated
with 350 μm 350 μm mesa area. The optical resonance mode reflection
measurement for an individual 350 μm 350 μm mesa area device shows
a good match to the simulation as shown in Fig. 21B. The measured optical
power for each wavelength was collected after passing through two crossed
polarizers to improve the extinction ratio (Nazirizadeh et al., 2008). Its mea-
sured reflection resonance peak wavelengths were recorded for applied elec-
trical biases varying from 10 to +10 V. Spectral blue-shifting and red-
shifting were found when applying forward and reverse bias, respectively.
A maximum red-shifting returning point was observed at (5 1) V as
shown in Fig. 21C. This spectral shifting is induced by overall free carrier
concentrations variation in two Si layers when electrical charges were
applied onto this capacitor like structure. Both spectral blue- and red-shift
by forward and reverse bias were obtained in the p/n-type stacking silicon
layers.
A 6 6 array of 80 μm 80 μm mesa devices was also fabricated for high
frequency EO modulation (Fig. 22). It was attached to a dielectric substrate
with a coplanar waveguide (CPW) line that was designed for 10 GHz oper-
ation. A single 80 μm 80 μm mesa area device was then wire-bonded. The
top and bottom electrodes were connected to 150 μm ground/signal pads for
wire-bonding. A scanning electron microscope (SEM) top view of a device
region with 80 μm 80 μm mesa area and a zoomed-in SEM image of the
50 μm 50 μm e-beam patterned array of holes are shown in Fig. 22B
and C.
While a modulator such as the one described here can in principle func-
tion in either transmission or reflection, we focus on the reflection mode of
operation in this work. A schematic diagram of the setup used for tests is
A B
70 3.5
–10 (dBm)
Simulation
60 Measurement –11 3
–12
50 2.5
Reflection (%)
Power (μW)
1520 1550
40 2
30 1.5
20 1
10 0.5
0 0
1500 1550 1600
Wavelength (nm)
C
1553.95
Wavelength (nm)
1553.9
1553.85
Red-shifting
Return point
1553.8
1553.75
–10 –8 –6 –4 –2 0 2 4 6 8 10
Voltage (V)
Fig. 21 A microscope top-view of a 6 6 array of 350 μm 350 μm mesa area devices is
shown in (A). Optical resonance mode reflection spectral for a single device is measured
in (B) with crossed-polarizers method to improve the extinction ratio (Nazirizadeh et al.,
2008). As shown in (C), continuous resonance mode spectral peak shifting was observed
by applying electrical bias from 10 to +10 V. The measurement points in blue circles
were connected by line, which shows a red-shifting return point between 4 and
6 V of reverse bias. Reproduced with permission from Shuai, Y.-C., Zhao, D., Liu, Y.,
Stambaugh, C., Lawall, J., Zhou, W., 2017. Coupled bilayer photonic crystal slab electro-
optic spatial light modulators. IEEE Photonics J. 9 (2), 1–11.
shown in Fig. 23A. Light from a narrow-linewidth tunable laser is sent

through a circulator, collimated, and passed through a polarizing beam split-
ter (PBS) to provide a well-defined linear polarization. The light is focused
and directed at the Si-NM device such that the waist of the incident beam
falls on the device at normal incidence. A quarter-wave plate can be set with
its fast axis parallel to the polarization of the incident light, so that light
reflected from the device retraces itself through the collimator and is
Fig. 22 A die of 6 6 array of 80 μm 80 μm mesa devices is shown in (A), one selected

device was wire-bonded to CPW line. In (B), a SEM top-view image for a single modulator
device with 80 μm 80 μm mesa area shows the details of interconnection and elec-
trodes, with a 50 μm 50 μm PCS area that is defined by the white square dash line.
A zoomed-in SEM top-view of its PCS center region is shown on right side in (C).
Reproduced with permission from Shuai, Y.-C., Zhao, D., Liu, Y., Stambaugh, C.,
Lawall, J., Zhou, W., 2017. Coupled bilayer photonic crystal slab electro-optic spatial light
modulators. IEEE Photonics J. 9 (2), 1–11.
diverted by the circulator for detection, or with its fast axis at 45° to the inci-
dent polarization, so that light falling on the device is circularly polarized and
the reflected light is deflected by the PBS and subsequently detected.
We supplemented the spectral shifting measurements with frequency-
domain EO modulation measurements using a network analyzer. RF ampli-
fier and a high-speed receiver were also used for device EO modulation fre-
quency domain characterization. The predicted optical resonance was found
near 1497.5 nm for an 80 μm 80 μm mesa area device. For subsequent fre-
quency domain EO modulation tests, the laser wavelength was fixed at
1499 nm where the slope of the reflection spectrum takes its highest value.
Fig. 23B shows the measured frequency response of the particular
80 μm 80 μm mesa device, where 0 dB corresponds to the static (low-
frequency) response. As shown in the figure, the response falls off by 3 dB
at a frequency of (200 20) MHz.
In summary, we have experimentally demonstrated a novel electro-optic
intensity modulator exploiting a Fano resonance in a bilayer photonic crystal
architecture. Both blue and red spectral shifts in the resonance were
observed, by applying opposite electrical biases. The 3 dB bandwidth of
one of our devices, with a lateral size of 80 μm 80 μm, was measured to
be (200 20) MHz. These spatial light modulators are polarization indepen-
dent owing to highly symmetric 2D square lattice photonic crystal structure.
Calculations suggest that bandwidths in the GHz range can be achievable if
Fig. 23 A characterization setup shown in (A) was used for device optical reflection
spectral collection and electro-optical (EO) modulation characterization. A network
analyzer, RF amplifier, and a high-speed receiver were used for device EO modulation
frequency domain characterization. As shown in (B), a 200 MHz 3 dB cutoff fre-
quency, which is about 100 times of a 350 μm 350 μm mesa area device’s perfor-
mance, was found for the device with 80 μm 80 μm mesa area. Reproduced with
permission from Shuai, Y.-C., Zhao, D., Liu, Y., Stambaugh, C., Lawall, J., Zhou, W.,
2017. Coupled bilayer photonic crystal slab electro-optic spatial light modulators. IEEE
Photonics J. 9 (2), 1–11.
the device size is reduced. High modulation depth should be possible as well,
by using a higher optical Q factor PCS stacking structure (Shuai et al., 2013a,
2013b) with process-controlled PCS hole size. These bilayer coupled PCS-
based modulators offer novel opportunities in vertical photonic integration,
including free-space communications, 3D optical interconnect, and spatial
control of light via extended horizontal arrays.
3. Fano resonance photonic crystal cavity

optomechanics
3.1 Optical forces in nanoscale cavities
Maxwell, dealing mathematically with the stresses in an electromagnetic
field, reached the conclusion that “in the medium in which waves are prop-
agated there is a pressure normal to the waves and numerically equal to the
energy in unit volume” (Kilic et al., 2007). Universal to physics from meso-
scale and microscale down to nanoscale, optomechanical interactions are a
rich subject of research and enable an especially rich set of phenomena and
opportunities because of the presence of both attractive and repulsive forces
(Nxcnons and Huu, 1901). Rapid developments have been made in this area
in the past decade. Efforts were made to higher optomechanical coupling
coefficients and several experimental demonstrations and approaches for
enhancing the strength of the effect were achieved during 2007–09
(Nxcnons and Huu, 1901).
Optical forces resulting from interacting modes and cavities can scale to
remarkably large values as the optical modes shrink to nanometer dimen-
sions (Kippenberg and Vahala, 2008). It has been noted that the remarkable
spatial confinement and large field enhancements afforded by nanometer-
scale guided modes can result in significant forces when optical modes
and cavities interact (Rakich et al., 2007; Van Thourhout and Roels,
2010). Scaled down to nanoscale dimensions, transverse optical forces in
an integrated silicon photonic circuit through an embedded nanomechanical
resonator were reported in 2008 (Povinelli et al., 2005a). In addition, inten-
sive progresses were made in 2009. Both attractive and repulsive optical
forces can be obtained between planarly coupled nanophotonic waveguides
(Povinelli et al., 2005b); force is tuned from attractive to repulsive by con-
trolling the relative phase of the optical fields injected into the waveguides
(Li et al., 2008). A static mechanical deformation of up to 20 nm in a silicon
nitride structure was experimentally demonstrated in the stacked double-
ring resonator structure with resonance shifting spanning 80 times the
intrinsic resonance linewidth with low optical power (Li et al., 2009).
Micrometer-scale displacements of the waveguide are observed for
milliwatt-level optical input powers in a structure with a high-Q optical
microresonator coupled with a micrometer-scale waveguide (Roels et al.,
2009). Seamless wavelength routing over a range of 3000 times the intrinsic
channel width was achieved in a specially designed nanooptomechanical sys-
tem (Wiederhecker et al., 2009). Eichenfield et al. proposed the idea of
“optomechanical crystals,” i.e., that in a photonic crystal, the periodicity
of the host medium is used to manipulate the properties of light, whereas
a phononic crystal uses periodicity to manipulate mechanical vibrations
(Eichenfield et al., 2007). A picogram- and nanometer-scale in a
photonic-crystal optomechanical cavity was measured in a pair of specially
patterned nanoscale beams, resulting in a scale of the per-photon force and
the mass of the structure (Eichenfield et al., 2009).
All of these results indicate that there is potential for radiation pressure
actuated devices to be used in a variety of photonic applications, such as
channel routing/switching, buffering, dispersion compensation, pulse trap-
ping/release, widely tunable lasers, and other light signal manipulations on
chip (Ma and Povinelli, 2012).
3.2 Optical forces in coupled Fano resonance PCSs

This concept was extended to 2D PCS’s where the radiation force can be
further enhanced due to the stored and strongly localized electromagnetic
field inside the PCS’s. Many systems have been investigated, including
coupled waveguides, waveguides coupled to substrates, as well as a variety
of resonator structures such as microrings, disks, or toroids that support
whispering gallery modes, and point defects in photonic crystal slab struc-
tures that support standing-wave optical modes. Recently, the behavior
of lateral and normal optical forces between coupled photonic crystal slabs
attracts a lot of interest, because near the frequencies of guided resonances,
we see significant enhancement of the optical forces (Liu et al., 2009;
Notomi et al., 2006).
Optical forces in coupled photonic crystal slabs have received a great deal
of attention recently (Notomi et al., 2006; Rodriguez et al., 2011; Roh
et al., 2010; Woolf et al., 2013). It was shown theoretically that resonances
arising from coupled photonic crystal slabs can generate optical forces that
are proportional to the quality factor of the resonances (Liu et al., 2009).
These forces can be attractive or repulsive and may also be engineered in
y x
a
Δx
Fig. 24 Schematic of the double-slab system. The red arrows indicate the directions of
the incident light. Each slab consists of an array of high index rods (ε ¼ 12) surrounded
by air (ε ¼ 1). The empty rectangle indicates the surface over which the integration of the
stress tensor is performed. In the right panel, the bottom slab is shifted relative to the
first slab by a distance of Δx. Reproduced with permission from Liu, V., Povinelli, M., Fan, S.,
2009. Resonance-enhanced optical forces between coupled photonic crystal slabs. Opt.
Express 17 (24), 21897–21909.
asymmetric structures (Rodriguez et al., 2011). We will first review some of

the basic principles for manipulating optical forces in coupled slab structures,
as shown in Fig. 24. Due to their different origins, the two types of dark
states in Section 1.2 exhibit qualitatively different force divergence behav-
iors. Three cases were considered (Liu et al., 2009).
In the first case shown in Fig. 24A, we consider the dark resonance from
coupled bright resonances as in Fig. 7. As the slab separation is changed, the
quality factor of the dark resonance becomes finite. The resonant properties
are plotted in Fig. 25. Since the resonant decay rate must vanish at the sep-
aration, during which the resonance is decoupled from radiation, the lowest
order term in the resonant decay rate is proportional to the square of the
separation. Hence the optical force, proportional to the quality factor,
diverges quadratically with respect to the separation (Fig. 26).
In the second case, shown in Fig. 24B, we consider a dark state from
coupled dark resonances as in Fig. 6. With a relative shift, the dark state splits
into two near-dark resonances, and the resonant properties of these two near-
dark resonances are plotted in Fig. 27. The quality factor diverges as Δx2,
which is the lowest order of dependence. The normal force, Fz, diverges
in the same way as Q, but the lateral force, Fx, diverges as Δx1 because
the average lateral force per photon to the lowest order is proportional to
Δx. The divergence behavior of the optical forces is plotted in Fig. 28.
In the third case, we consider a dark state from coupled bright reso-
nances, similar to Case 1, except that the two slabs are shifted relative to each
d¥
0.635
w
Resonance frequency FWHM
0.63 w¥
0.625
0.62
0.615
0.6 0.65 0.7 0.75 0.8
Slab spacing (d/a)
Fig. 25 Resonance peak and linewidth in the vicinity of the dark state arising from
coupled bright resonances as in Fig. 7 (Case 1). Reproduced with permission from
Liu, V., Povinelli, M., Fan, S., 2009. Resonance-enhanced optical forces between coupled
photonic crystal slabs. Opt. Express 17 (24), 21897–21909.
106 107
102 ´ Fz (w = 0.630)
105 Fz 106
Q
104 105
Force (l0 /c)
103 104
Q
102 103
10 102
10–5 10–4 10–3 10–2 10–1 1

|d–d¥|/a
Fig. 26 Divergence behavior of normal force on resonance compared to the Q as the
slab spacing is varied away from the dark state in Case 1. For reference, the force var-
iation at a constant frequency near the dark state frequency is also plotted with a dashed
line. Reproduced with permission from Liu, V., Povinelli, M., Fan, S., 2009. Resonance-
enhanced optical forces between coupled photonic crystal slabs. Opt. Express 17 (24),
21897–21909.
0.584
Resonance frequency
0.582
w
FWHM
0.58
0.578
0 0.1 0.2 0.3 0.4 0.5

Relative shift (Δx/a)
Fig. 27 Resonance peak and linewidth in the vicinity of the dark state arising from
coupled dark resonances as in Fig. 6(Case 2). Reproduced with permission from Liu, V.,
Povinelli, M., Fan, S., 2009. Resonance-enhanced optical forces between coupled photonic
crystal slabs. Opt. Express 17 (24), 21897–21909.
106 107
Q
105 Fz 106
Fx
104 105
Force (l0/c)
103 104
Q
102 103
10 102
10−3 10−2 10−1 1

Δx/a
Fig. 28 Divergence behavior of normal and lateral forces on resonance compared to the
Q for various shifts near the symmetry point Δx ¼ 0 for Case 2. Reproduced with permis-
sion from Liu, V., Povinelli, M., Fan, S., 2009. Resonance-enhanced optical forces between
coupled photonic crystal slabs. Opt. Express 17 (24), 21897–21909.
other by 0.2a, where a is the lattice constant. A dark state still exists, as seen in
Fig. 29. As the slab separation is changed, both the normal force, Fz, and the
lateral force, Fx, vary in the same way as Q, inversely proportional to the
square of the separation (Fig. 30).
In conclusion, in a coupled photonic crystal slab system, the optical
forces are strongly influenced by the symmetries of the system as well as
its resonance properties. It is worth noting that these proposed structures
d∞
0.64
w
FWHM
Resonance frequency 0.635
w∞
0.63
0.625
0.62
0.56 0.6 0.64 0.68 0.72
Slab spacing (d/a)
Fig. 29 Resonance peak and linewidth in the vicinity of the dark state for Case 3 at
Δx ¼ 0.2a. Reproduced with permission from Liu, V., Povinelli, M., Fan, S., 2009.
Resonance-enhanced optical forces between coupled photonic crystal slabs. Opt. Express
17 (24), 21897–21909.
105 106
Q
104 Fx 105
Fz
103 104
Force (l0/c)
102 103
Q
10 102
1 10
10−3 10−2 10−1

|d–d∞|/a
Fig. 30 Divergence behavior of normal and lateral forces on resonance compared to the
Q for various slab spacings near the dark state in Fig. 6. Reproduced with permission from
can lead to enhanced light-induced pressure over larger areas, in a config-

uration that is directly accessible to the externally incident, free-space optical
beams. The force scales with the Q of the resonances in the structure. By
placing the system configuration in the vicinity of a dark state, a dramatic
enhancement of the optical forces may be obtained due to the large resonant
enhancement of the fields.
Followed by initial theoretical investigations of high-Q and strong
optomechanical coupling in defect-mode double-layer PCS cavities
(Notomi et al., 2006), Notomi et al. (Roh et al., 2010) reported an exper-
imental demonstration of strong optomechanical interactions in InP-based
defect-free bilayer PCS cavities, based on bandedge or Fano resonance
modes. Significant radiation force of the stored light (F) should appear as
the electromagnetic energy stored in the cavity (U) changes with the change
in spatial separations (z) between two coupled PCS slabs, as shown in the
following equation:

dU U dω PQ dω
F ¼ ¼ ¼
dz ω dz ω2 dz
where z is the spatial coordinate of vertical direction, and the electro-

magnetic energy stored in the cavity is U ¼ Nħωc ¼ PQ/ω. P is the incident
optical power. For the mode with dω dz > 0, F < 0, attractive force appears
between the two PCS’s. Based on the modal dispersion properties
(Notomi et al., 2006), the estimated radiation force per unit of electromag-
netic energy in the cavity is 0.23 μN/pJ, with the force estimated to be
0.4 μN/pJ. For an incident power of 19.7 mW, a displacement of
3.6 nm was obtained for a cavity Q of 1600 at 1568 nm, as a result of an opti-
cal force of 10.8 nN (Fig. 31).
Theoretical analysis showed that the optical force is periodic,
corresponding with displacement, resulting in stable and unstable equilib-
rium positions. Both attractive and repulsive force were induced depending
Fig. 31 (A) Schematics of the experiment. The structure is fully clamped (all-side
clamped); (B) Top-view scanning electron microscope image; and (C) Bird’s-eye view
(upper) and cross section (lower) of a bilayer PCS. The period is 750 nm and the hole
radius at the center is 270 nm. Reproduced with permission from Roh, Y.-G., Tanabe, T.,
Shinya, A., Taniyama, H., Kuramochi, E., Matsuo, S., Sato, T., Notomi, M., 2010. Strong
optomechanical interaction in a bilayer photonic crystal. Phys. Rev. B 81 (12), 121101.
A B
a 2R
h1
s
h2
a
Fig. 32 (A, B) Schematic and (C) SEM image of single-membrane (asymmetric) structure:
a photonic-crystal (holey) membrane consisting of a square-lattice of air holes on silicon
is suspended (separation s) on top of an unpatterned (homogeneous) silicon slab (thick-
ness h2 ¼ 0.2a) sitting on top of a semiinfinite silica substrate. Light is apparent on the
membrane from the normal direction (top). Key parameters are: a ¼ 920 nm, R ¼ 207 nm,
h1 ¼ h2 ¼ 185 nm, and s ¼ 265 nm. Reproduced with permissions from Rodriguez, A. W.,
McCauley, A. P., Hui, P.-C., Woolf, D., Iwase, E., Capasso, F., Loncar, M., Johnson, S. G.,
2011. Bonding, antibonding and tunable optical forces in asymmetric membranes. Opt.
Express 19 (3), 2225–2241; Hui, P.-C., Woolf, D., Iwase, E., Sohn, Y.-I., Ramos, D.,
Khan, M., Rodriguez, A. W., Johnson, S. G., Capasso, F., Loncar, M., 2013. Optical bistability
with a repulsive optical force in coupled silicon photonic crystal membranes. Appl. Phys.
Lett. 103 (2), 021102 (021102-5).
on the spatial symmetry of the stored electromagnetic field profile in the

double-layer PC slab cavities and air-slot cavities (Roh et al., 2010).
Similar to the symmetric double-layer PCS systems, tunable attractive
and repulsive forces can also arise in highly asymmetric structures coupled
with external radiation. One recent theoretical study by Rodriguez et al.
(2011) shows this generality by demonstrating that tunable attractive and
repulsive forces can also arise in highly asymmetric structures coupled with
external radiation (Fig. 32A). Recently, Hui et al. (2013) reported an exper-
imental observation of repulsive forces with a displacement of 1 nm/mW in
such an Si-based hybrid PCS membrane/unpatterned Si membrane cavity
(Fig. 32B). It is also predicted that such configurations can extend the dis-
placement range to many tens of nanometers, providing a simple, nonintrusive
solution to extending the actuation range of microelectromechanical systems
(MEMS) devices.
3.3 Optomechanical control and MOMES applications

There is a variety of all-optical operations that can be created by this method,
ranging from all optical switching and tuning (for telecom, sensing, and imag-
ing applications) to adaptive dispersion and filter synthesis (for applications
such as adaptive filtering of laser lines and optical clock recovery). MEMS’s are
the promising mechanisms for introducing variable functions into PC devices.
Although several devices with variable structures of PC slabs combined with
MEMS were considered, there have been very few reports on the experimen-
tal studies because of the difficulty in fabricating hybrid structures (Hui et al.,
2013; Rodriguez et al., 2011). A few experimental results, as well as principle
design, have been reported (Lu and Lee, 2009; Stomeo et al., 2010). In order
to introduce variable function to the Fano resonances in PC slabs, Suh et al.
(2003) studied two PC slabs facing each other analytically for a wavelength-
tunable PC filter. Kanamori et al. (2007) experimentally demonstrated a
wavelength-selective variable reflection filter controlled by changing the
gap with a microelectromechanical actuator (Fig. 33).
Fig. 33 (A) Cross-section diagram of the MEMS tunable Fano filter design and (B) SEM
image of the fabricated Fano filter. The inset shows the magnified view of the fabricated
PC slab. Reproduced with permission from Kanamori, Y., Kitani, T., Hane, K., 2007. Control of
guided resonance in a photonic crystal slab using microelectromechanical actuators. Appl.
Phys. Lett. 90, 031911.
Acknowledgments
The authors acknowledge contributions from their current and former students, as well as
collaborators involved in the research work reviewed here. We also thank funding
agencies and program directors for their instrumental support from US Air Force Office
of Scientific Research (AFOSR, Dr. Gernot Pomrenke), US Army Research Office
(ARO, Dr. Michael Gerhold), US National Science Foundation (NSF, Drs. John Zavada
and Dominique Dagenais). We also thank the fabrication support from the University of
Texas at Arlington Nanotechnology Research Center, and the University of Texas at
Austin, part of NSF NNIN.
References
Chen, L., Qiang, Z., Yang, H., Pang, H., Ma, Z., Zhou, W.D., 2009. Polarization and angu-
lar dependent transmissions on transferred nanomembrane Fano filters. Opt. Express
17 (10), 8396–8406.
Chmielak, B., Waldow, M., Matheisen, C., Ripperda, C., Bolten, J., Wahlbrink, T.,
Nagel, M., Merget, F., Kurz, H., 2011. Pockels effect based fully integrated, strained
silicon electro-optic modulator. Opt. Express 19 (18), 17212–17219.
Crozier, K.B., Lousse, V., Kilic, O., Kim, S., Fan, S., Solgaard, O., 2006. Air-bridged pho-
tonic crystal slabs at visible and near-infrared wavelengths. Phys. Rev. B 73 (11), 115126.
Eichenfield, M., Michael, C.P., Perahia, R., Painter, O., 2007. Actuation of micro-
optomechanical systems via cavity-enhanced optical dipole forces. Nat. Photonics
1 (7), 416–422.
Eichenfield, M., Camacho, R., Chan, J., Vahala, K.J., Painter, O., 2009. A picogram- and
nanometre-scale photonic-crystal optomechanical cavity. Nature 459 (7246), 550–555.
Phys. Rev. B 65 (23), 235112.
Hui, P.-C., Woolf, D., Iwase, E., Sohn, Y.-I., Ramos, D., Khan, M., Rodriguez, A.W.,
Johnson, S.G., Capasso, F., Loncar, M., 2013. Optical bistability with a repulsive optical
force in coupled silicon photonic crystal membranes. Appl. Phys. Lett. 103 (2), 021102,
021102-5.
Jacobsen, R.S., Andersen, K.N., Borel, P.I., Fage-Pedersen, J., Frandsen, L.H., Hansen, O.,
Kristensen, M., Lavrinenko, A.V., Moulin, G., Ou, H., 2006. Strained silicon as a new
electro-optic material. Nature 441 (7090), 199–202.
Kanamori, Y., Kitani, T., Hane, K., 2007. Control of guided resonance in a photonic crystal
slab using microelectromechanical actuators. Appl. Phys. Lett. 90, 031911.
Kanskar, M., Paddon, P., Pacradouni, V., Morin, R., Busch, A., Young, J.F., Johnson, S.R.,
MacKenzie, J., Tiedje, T., 1997. Observation of leaky slab modes in an air-bridged semi-
conductor waveguide with a two-dimensional photonic lattice. Appl. Phys. Lett. 70 (11),
1438–1440.
Kilic, O., Digonnet, M., Kino, G., Solgaard, O., 2007. External fibre Fabry-Perot acoustic
sensor based on a photonic-crystal mirror. Meas. Sci. Technol. 18 (10), 3049–3054.
Kippenberg, T., Vahala, K., 2008. Cavity optomechanics: back-action at the mesoscale.
Science 321 (5893), 1172–1176.
Li, M., Pernice, W., Xiong, C., Baehr-Jones, T., Hochberg, M., Tang, H., 2008. Harnessing
optical forces in integrated photonic circuits. Nature 456 (7221), 480–484.
Li, M., Pernice, W., Tang, H., 2009. Tunable bipolar optical interactions between guided
lightwaves. Nat. Photonics 3 (8), 464–468.
Lin, C., Lu, Z., Shi, S., Jin, G., Prather, D.W., 2005. Experimentally demonstrated filters
based on guided resonance of photonic-crystal films. Appl. Phys. Lett. 87, 091102.
Liu, V., Fan, S., 2012. S 4: a free electromagnetic solver for layered periodic structures. Com-
put. Phys. Commun. 183 (10), 2233–2244.
Lousse, V., Suh, W., Kilic, O., Kim, S., Solgaard, O., Fan, S.H., 2004. Angular and polar-
ization properties of a photonic crystal slab mirror. Opt. Express 12 (8), 1575–1582.
Lu, T.-W., Lee, P.-T., 2009. Ultra-high sensitivity optical stress sensor based on double-
layered photonic crystal microcavity. Opt. Express 17 (3), 1518–1526.
Ma, J., Povinelli, M.L., 2012. Applications of optomechanical effects for on-chip manipu-
lation of light signals. Curr. Opin. Solid State Mater. Sci. 16 (2), 82–90.
McKerracher, I., Fu, L., Tan, H., Jagadish, C., 2013. Integration of bandpass guided-mode
resonance filters with mid-wavelength infrared photodetectors. J. Phys. D Appl. Phys.
46 (9), 095104.
Nazirizadeh, Y., M€ uller, J.G., Geyer, U., Schelle, D., Kley, E.-B., T€ unnermann, A.,
Lemmer, U., Gerken, M., 2008. Optical characterization of photonic crystal slabs using
orthogonally oriented polarization filters. Opt. Express 16 (10), 7153–7160.
Notomi, M., Taniyama, H., Mitsugi, S., Kuramochi, E., 2006. Optomechanical wavelength
and energy conversion in high-Q double-layer cavities of photonic crystal slabs. Phys.
Rev. Lett. 97 (2), 023903.
Nxcnons, E., Huu, G., 1901. A preliminary communication on the pressure of heat and light
radiation. Phys. Rev. 13, 307.
Peng, S., Morris, G., 1996. Resonant scattering from two-dimensional gratings. J. Opt. Soc.
Am. A 13 (5), 993–1005.
Povinelli, M., Johnson, S., Lonèar, M., Ibanescu, M., Smythe, E., Capasso, F.,
Joannopoulos, J., 2005a. High-Q enhancement of attractive and repulsive optical forces
between coupled whispering-gallery-mode resonators. Opt. Express 13 (20),
8286–8295.
Povinelli, M.L., Loncar, M., Ibanescu, M., Smythe, E.J., Johnson, S.G., Capasso, F.,
Joannopoulos, J.D., 2005b. Evanescent-wave bonding between optical waveguides.
Opt. Lett. 30 (22), 3042–3044.
Qiang, Z., Yang, H., Chen, L., Pang, H., Ma, Z., Zhou, W., 2008. Fano filters based on
transferred silicon nanomembranes on plastic substrates. Appl. Phys. Lett. 93, 061106.
Rakich, P.T., Popovic, M.A., Soljacic, M., Ippen, E.P., 2007. Trapping, corralling and spec-
tral bonding of optical resonances through optically induced potentials. Nat. Photonics
1 (11), 658–665.
Rodriguez, A.W., McCauley, A.P., Hui, P.-C., Woolf, D., Iwase, E., Capasso, F.,
Loncar, M., Johnson, S.G., 2011. Bonding, antibonding and tunable optical forces in
asymmetric membranes. Opt. Express 19 (3), 2225–2241.
Roels, J., De Vlaminck, I., Lagae, L., Maes, B., Van Thourhout, D., Baets, R., 2009. Tunable
optical forces between nanophotonic waveguides. Nat. Nanotechnol. 4 (8), 510–513.
Rogers, J.A., Lagally, M.G., Nuzzo, R.G., 2011. Synthesis, assembly and applications of
semiconductor nanomembranes. Nature 477 (7362), 45–53.
Roh, Y.-G., Tanabe, T., Shinya, A., Taniyama, H., Kuramochi, E., Matsuo, S., Sato, T.,
Notomi, M., 2010. Strong optomechanical interaction in a bilayer photonic crystal.
Phys. Rev. B 81 (12), 121101.
Rosenberg, A., Carter, M., Casey, J., Kim, M., Holm, R., Henry, R., Eddy, C.,
Shamamian, V., Bussmann, K., Shi, S., Prather, D.W., 2005. Guided resonances in
asymmetrical GaN photonic crystal slabs observed in the visible spectrum. Opt. Express
13 (17), 6564–6571.
Shuai, Y., Zhao, D., Singh Chadha, A., Seo, J.-H., Yang, H., Fan, S., Ma, Z., Zhou, W.,
2013a. Coupled double-layer Fano resonance photonic crystal filters with lattice-
displacement. Appl. Phys. Lett. 103 (24), 241106.
Shuai, Y., Zhao, D., Tian, Z., Seo, J.-H., Plant, D.V., Ma, Z., Fan, S., Zhou, W., 2013b.
Double-layer Fano resonance photonic crystal filters. Opt. Express 21 (21),
24582–24589.
Shuai, Y.-C., Zhao, D., Liu, Y., Stambaugh, C., Lawall, J., Zhou, W., 2017. Coupled bilayer
photonic crystal slab electro-optic spatial light modulators. IEEE Photonics J. 9 (2), 1–11.
Soref, R., Bennett, B., 1987. Electrooptical effects in silicon. IEEE J. Quantum Electron.
23 (1), 123–129.
Stomeo, T., Grande, M., Rainò, G., Passaseo, A., D’Orazio, A., Cingolani, R., Locatelli, A.,
Modotto, D., De Angelis, C., De Vittorio, M., 2010. Optical filter based on two coupled
PhC GaAs-membranes. Opt. Lett. 35 (3), 411–413.
Suh, W., Fan, S., 2003. Mechanically switchable photonic crystal filter with either all-pass
transmission or flat-top reflection characteristics. Opt. Lett. 28 (19), 1763–1765.
Suh, W., Fan, S., 2004. All-pass transmission or flattop reflection filters using a single pho-
tonic crystal slab. Appl. Phys. Lett. 84, 4905.
Suh, W., Yanik, M.F., Solgaard, O., Fan, S., 2003. Displacement-sensitive photonic crystal
structures based on guided resonance in photonic crystal slabs. Appl. Phys. Lett. 82 (13),
1999.
Suh, W., Wang, Z., Fan, S., 2004. Temporal coupled-mode theory and the presence of non-
orthogonal modes in lossless multimode cavities. IEEE J. Quantum Electron. 40 (10),
1511–1518.
Suh, W., Solgaard, O., Fan, S., 2005. Displacement sensing using evanescent tunneling
between guided resonances in photonic crystal slabs. J. Appl. Phys. 98, 033102.
Synopsys, 2016. Taurus Medici [Online]. Sysnopsys.
Thurman, S.T., Morris, G.M., 2003. Controlling the spectral response in guided-mode res-
onance filter design. Appl. Optics 42 (16), 3225–3233.
Van Thourhout, D., Roels, J., 2010. Optomechanical device actuation through the optical
gradient force. Nat. Photonics 4 (4), 211–217.
Wiederhecker, G.S., Chen, L., Gondarenko, A., Lipson, M., 2009. Controlling photonic
structures using optical forces. Nature 462 (7273), 633–636.
Woolf, D., Hui, P.-C., Iwase, E., Khan, M., Rodriguez, A.W., Deotare, P., Bulu, I.,
Johnson, S.G., Capasso, F., Loncar, M., 2013. Optomechanical and photothermal inter-
actions in suspended photonic crystalmembranes. Opt. Express 21 (6), 7258–7275.
Yang, H., Zhao, D., Chuwongin, S., Seo, J.H., Yang, W., Shuai, Y., Berggren, J.,
Hammar, M., Ma, Z., Zhou, W., 2012. Transfer-printed stacked nanomembrane lasers
on silicon. Nat. Photonics 6 (9), 615–620.
Zhang, K., Seo, J.H., Zhou, W., Ma, Z., 2012. Fast flexible electronics using transferrable
silicon nanomembranes (topical review). J. Phys. D Appl. Phys. 45 (14), 143001.
Zhou, W., Ma, Z., Yang, H., Qiang, Z., Qin, G., Pang, H., Chen, L., Yang, W.,
Chuwongin, S., Zhao, D., 2009. Flexible photonic-crystal Fano filters based on trans-
ferred semiconductor nanomembranes. J. Phys. D Appl. Phys. 42, 234007.
CHAPTER SEVEN
On-chip photonic crystal

surface-emitting lasers
Weidong Zhou*
Contents
1. Introduction 189
2. Overview of photonic crystal lasers 190
3. Photonic crystal MR-VCSEL cavity design 196
3.1 Photonic crystal MRs 196
3.2 Energy and phase penetration properties in MRs 198
3.3 MR-VCSEL 200
4. Photonic crystal bandedge surface-emitting laser cavity design 201
4.1 Bandedge mode design 203
4.2 Lateral cavity scaling 204
5. Printed membrane laser cavities 206
6. MR-VCSELs based on 1D PCS/grating mirrors 208
7. MR-VCSEL characterization 209
8. PCSEL characterization 212
9. PCSEL on bulk silicon substrate 213
10. PCSEL thermal performance analysis 215
11. Electrically pumped PCSEL characterization 218
12. Summary 220
Acknowledgments 221
References 221
1. Introduction
Silicon-based light sources have witnessed significant advances
over the last decade, with one of the most promising approaches being
heterogeneous integration of direct-bandgap group III–V materials on various
Si-based laser cavities on silicon-on-insulator (SOI) substrates (Bakir et al.,
2006; Balakrishnan et al., 2006; Chen et al., 2011; Liang et al., 2010; Mi
et al., 2005; Park et al., 2006; Roelkens et al., 2010; Stankovic et al., 2011;

190 Weidong Zhou
Van Campenhout et al., 2007). Photonic crystals have been known to be an

excellent candidate for ultracompact energy efficient lasers via spontaneous
emission control and dispersion engineering. Based on defect-mode cavities
and defect-free bandedge effects, lasers have been reported for low power
single-photon sources as well as high power single-mode broad-area laser
sources, mostly on III–V substrates (Altug et al., 2006; Hirose et al., 2014;
Noda, 2010; Painter et al., 1999; Park et al., 2004; Yoshie et al., 2004;
Zhou et al., 2001). Additionally, based on Fano resonance or guided
resonance principles, broadband single-layer membrane reflectors (MRs)
can also be designed based on photonic crystal cavities in constructing
VCSELs (Fan and Joannopoulos, 2002; Yang et al., 2012; Zhou et al., 2014).
Here, we report design and analysis of photonic crystal surface-emitting
membrane lasers for silicon photonics, based on printed IV–V QW hetero-
structures heterogeneously integrated on Si-based photonic crystal cavities.
Based on the evanescent mode coupling, membrane lasers can be realized by
vertical coupling of III–V QW membranes on defect-mode and bandedge
photonic crystal cavities on Si. The effect of coupling strength and mode
overlap on confinement factor and lasing threshold will be evaluated and
optimized. On the other hand, based on photonic crystal MRs,
MR-VCSEL cavities can be design with the proper selection of low index
buffer layers between the photonic crystal MRs and III–V QW structures.
Both photonic crystal guided resonance mode and Fabry–Perot (FP) lasing
cavity mode need to be properly controlled.
Based on the membrane transfer printing technique for heterogeneous
integration of III–V with Si materials (Yang et al., 2012), we demonstrated
printed large-area single-mode photonic bandedge lasers (PBELs) as well as
MR-VCSELs on SOI substrates. Optically pumped laser characteristics will
be discussed, with the results compared with the gain threshold conditions
for lasing model property analysis. Cavity scaling and electrical injection will
also be discussed for practical and energy efficient source on-chip for inte-
grated photonics.
2. Overview of photonic crystal lasers

System integration requires photonic devices to be compact yet efficient,
high performance yet low power consumption, highly uniform and simply controled,
alignment free and scalable, low cost and reliable. In search for the ideal
2D surface-emitting laser arrays, vertical-cavity surface-emitting laser
(VCSEL) seems to be the most mature technology to date, with successful
Photonic crystal surface-emitting lasers 191
demonstrations of ultralow threshold, high single-mode power, and large

size of arrays, and arrays on silicon substrates (Iga, 2000). VCSELs are rou-
tinely used for short distance data communication application, where typical
array sizes are 6 or 12. It is expected that large array sizes (10s to 100s) are also
needed for future optical interconnect and image processing systems. Size,
uniformity, and power dissipation will be the dominant factors in realizing
these large array sizes on surface-emitting lasers. In addition, current VCSEL
technologies also face a number of challenges, such as transverse and polar-
ization mode control, aperture size uniformity control, linewidth reduction,
DBR resistance, limited thermal performances, and challenges on suitable
DBRs for long wavelength VCSELs and complicated growth requirement
(Iga, 2000).
Photonic crystals offer the most promising alternative to the current
DBR-based VCSEL technology. Since the first proposal of spontaneous
emission control with photonic crystals (Huang et al., 2007), a large variety
of light-emitting devices have been demonstrated, mostly with 2D photonic
crystal slab (PCS) air hole lattice structures. These structures are incorpo-
rated in LEDs for improved extraction efficiency and directionality, in
edge-emitting lasers as cleave free end mirrors and/or PBG line defect
defined waveguide, in DBR-based VCSELs for mode control, and in
photonic band edge surface-emitting lasers (PBELs) as 2D DFB lasers. Most
of these applications utilize the photonic bandgap (PBG) purely as mirrors or
filters, where there is little impact on spontaneous emission rate. On the
other hand, spontaneous emission control is enabled in photonic bandgap
defect-mode lasers (PBDLs) by embedding the active region inside the
photonic crystals, where 2D air columns are etched through the cavity,
including active regions. A summary of different surface-emitting lasers is
given in Table 1.
Broadly speaking, there are four types of PCS-based light sources:
(I) Photonic crystal confined VCSELs: By etching into the top portion of the
DBRs in the otherwise conventional DBR-based VCSELs, transverse
modal control is possible (Furukawa et al., 2004; Lundeberg et al.,
2005; Song et al., 2002; Yokouchi et al., 2003). With the proper
design of these photonic crystal structures, VCSEL beam properties
can be engineered for improved performances.
(II) PBDL: By etching through the active region of the light-emitting
structure, the spontaneous emission process can be altered in the
defect (point and line defect, surface wave modes) cavity region,
where enhanced/suppressed spontaneous emission is feasible when
Table 1 Photonic crystal surface-emitting lasers (PCSELs)
Configurations PhC functions Characteristics References
DBR- Vertical-cavity surface-emitting Pros: Iga (2000)
Based lasers:
VCSEL • DBR (1D PhC): HR mirror • Mature technology
for high Q cavity formation • Short vertical cavity
• Relative low
threshold
• High output power
Photonic crystal confined VCSELs: • Proven reliability Song et al. (2002), Yokouchi et al. (2003),
• Possible high power Furukawa et al. (2004), and Lundeberg
• DBR (1D PhC): HR mirror single mode in et al. (2005)
for high Q cavity formation PC-VCSELs
• 2D PhC air hole: Lateral mode Cons/Challenges:
confinement
• Wafer growth
for DBRs
• Long wavelengths
• Uniform small
aperture size control
• Wavelength specific
design
Table 1 Photonic crystal surface-emitting lasers (PCSELs)—cont’d
DBR- Photonic bandgap defect-mode lasers: Pros: Hwang et al. (2000), Painter et al. (1999),
free • 2D PhC air hole: Defect mode Park et al. (2004), and Zhou et al. (2000)
PCSEL at the center of the PBG for • DBR free with simple
enhanced spontaneous wafer growth (LED
emission like structure)
• Spontaneous emission
control with high
spontaneous emission
factor
Photonic crystal DFBs or photonic • Potentially zero Dowling et al. (1994), Meier et al.
bandedge lasers: threshold (1999b), Noda et al. (2001), Susa (2001),
2D PhC air hole: • Wavelength agile Monat et al. (2003), and Kwon et al.
fabrication (2004)
• Distributed feedback • Energy efficient
• Enhanced gain at bandedge • Scalable power output
due to flat dispersion (zero • Precise cavity control
group velocity) enable high
uniformity
Cons/Challenges:
• Efficient electrical
injection
• Simplified process
development
Continued
Table 1 Photonic crystal surface-emitting lasers (PCSELs)—cont’d
DBR- Photonic crystal membrane reflector Pros: Sciancalepore et al. (2012) and Yang et al.
free VCSELs: (2012)
VCSEL • 1D high contrast grating or 2D • DBR free with simple
photonic crystal-based single- wafer growth (LED
layer reflectors like structure)
• Either one or both MRs are • Spontaneous emission
used to replace DBRs control with high
spontaneous emission
factor
• Ultracompact cavity
• Simple cavity
resonance tuning with
patterning process
Cons/Challenges:
• Efficient electrical
injection
• Simplified process
development
the active light emission region is placed inside the PBG defect cavity
(Noda et al., 2007; Yablonovitch, 1987). With extremely high
Q/Vmode (Q and Vmode being the corresponding quality factor and
modal volume, respectively), strong spontaneous emission control
is feasible for ultralow threshold single-mode lasers. Typically these
defect modes are located below lightline. This configuration can lead
to potentially thresholdless lasers (or single-mode LEDs) and single-
photon sources (Bhattacharya et al., 2005; Colombelli et al., 2003;
Ellis et al., 2011; Hwang et al., 2000; Painter, 1999; Park et al.,
2004; Zhou et al., 2001). Another architecture reported recently
(Wu et al., 2015) is the coupling of unpatterned monolayer tungsten
diselenide monolayer gain medium with the defect-mode photonic
crystal cavity. This configuration can lead to higher efficiency lasers
as no-patterned or etching is done on the gain medium. Challenges
may lie on the trade-offs in the desired high Q cavity inside the
PBG and the high coupling efficiency between the high index gain
medium and photonic crystal cavities.
(III) PBELs: The bandedge effect in defect-free photonic crystal structures,
where the group velocity is close to zero, has advantages in achieving
high performance semiconductor surface-emitting lasers, including
low lasing threshold, single longitudinal and transverse mode over a
large lasing area, narrow linewidth, high power output, small beam
divergence angle, polarization control, and output beam pattern con-
trol (Chua et al., 2014; Hirose et al., 2014; Imada et al., 1999; Meier
et al., 1999b; Miyai et al., 2006; Noda et al., 2001). Based on the
bandedge effect, high performance lasers on GaAs have been demon-
strated, including a very recent demonstration of high power, single
mode, and large-area PC bandedge lasers with unpatterned QW
active region using a regrowth process (Hirose et al., 2014). Different
from PBDL discussed above, the modes considered here are typically
located above lightline, thus these lasers are intrinsically surface emit-
ting with excellent beam qualities.
(IV) Membrane reflector VCSEL (MR-VCSELs): Based on the Fano reso-
nance principles, broadband reflection can be achieved in single-layer
MRs based on 1D or 2D PCS. These ultracompact high performance
MR structures have been proposed to replace the bulky DBR struc-
tures, to create ultracompact MR-VCSELs (Sciancalepore et al.,
2012; Yang et al., 2012). Taking advantage of Fano resonance prin-
ciples, ultracompact high performance VCSEL cavities can be formed
196 Weidong Zhou
on any substrate, such as III–V and Si, because the mirror cavity and
the gain medium can be heterogeneously integrated with different
material systems. This is one of the most promising ways to form lasers
on Si, and to form high performance long wavelength VCSELs, in
which DBR remains a great challenge.
Among four configurations discussed earlier, PBDL utilizes the localized or
defect mode in the PBG. On the other hand, for the latter two configura-
tions (PBEL and MR-VCSEL), it is not essential to design PCS’s with PBG,
as both utilize propagating modes in defect-free PCS structures. While
PBEL can utilize the mode both below and above the light lines,
MR-VCSEL’s only use the modes above the light lines. Fano resonance
modes can be designed in PBEL, where lasing is feasible due to distributed
feedback (Chua et al., 2011). Most notably, the other type of Fano
resonance-based laser is the recently demonstrated MR-VCSEL, where
Fano resonance MRs are incorporated in the VCSEL cavity design to form
ultracompact VCSEL’s and lasers on Si. It is also worth mentioning that since
PBG is not required in PBEL, the requirement on high index contrast can be
relaxed. Lasing cavities can be built directly on patterned bulk semiconduc-
tor InP or Si substrates.
In this chapter, we will focus on detailed discussions related to on-chip
photonic crystal lasers where the lasing cavity is formed directly on silicon
substrates, the active gain medium is a conventional compound semicon-
ductor quantum well (QW) structures.
3. Photonic crystal MR-VCSEL cavity design

3.1 Photonic crystal MRs
One of the key design targets for broadband reflectors is high reflectivity
spectral bandwidth. The spectral bandwidth and reflection properties have
been well investigated in multilayer DBR-based reflectors, where ultra-
broadband reflectors can be achieved with the control of index contrast
between pairs of dielectric layers. However, it remains a challenge to under-
stand the correlations between spectral bandwidth and the design parameters
in 2D PCS-based MR structures. On the other hand, details of theoretical
investigations have been reported indicating the physical basis for wideband
resonators in 1D grating structures. Magnusson et al. attributed the broad-
band reflection in 1D grating to the coupling of leaky modes and their
excitation conditions (Magnusson and Shokooh-Saremi, 2008). Karagodsky

et al. (2010) presented an analytic analysis of the ultrahigh reflectivity of sub-
wavelength high contrast grating reflectors, and it was classified as a destruc-
tive interference effect between the two grating modes. For 2D PCS-based
MRs, the designs are largely based on the trial-and-error approaches based
on complete 3D computations, such as rigorous coupled-wave analysis and
finite-difference time domain (FDTD) techniques. Based on the different
designs for high Q filters and broadband MRs, researchers found that the
Q factor (spectral linewidth) can be controlled by varying the radius, with
very small r/a values (0.1) favoring high Q filter designs and large r/a values
(0.3–0.49) favoring broadband reflector designs. Additionally, the layer
thickness is optimal at around 0.5λ, where λ is the target wavelength. Other
design parameters include the lattice structure and the air hole shape, the
buffer layer selection below and above the functional layer, effective indices,
and so on (Zhou et al., 2014).
High reflectivity for surface-normal direction originates from the modal
coupling between vertical radiation modes and discrete in-plane guide
modes. As shown in Fig. 1A, multiple Fano resonance modes can be iden-
tified from the simulated transmission spectrum (log plot). At Fano reso-
nances (e.g., 1500 nm shown in Fig. 1B), there is a coupling between
in-plane guided mode (inside Si-MR) and vertical-cavity mode which
results in a destructive interference in the direction of transmission.
A B
100 100
l = 1500 nm
80
Reflection (%)
Transmission
60
40
20
R
T (Log)
0
1400 1450 1500 1550 1600
Wavelength (nm)
Fig. 1 Fano resonance membrane reflector (MR) design and modal coupling:
(A) simulated reflection (blue) and transmission (red) spectra for the Fano MR and
(B) simulated field distribution profiles for the design shown in (A) at one Fano reso-
nance location (λ ¼ 1500 nm).
198 Weidong Zhou
Designs of MRs on SOI were reported by Qiang et al. (2010) who

demonstrated control of reflection bands based on the control of vertical
confinement with controlled refractive indices for the buffer layers below
and above the MR layer. Detailed experimental results were subsequently
reported by Yang et al. (2009) with a controlled index tuning range over
50 nm for both blue and red shifts, in which they controlled the buffer oxide
index with controlled etching and the index above the slab with controlled
oxide deposition.
Further design optimization leads to a much broader reflection band in
Si-based 2D PCS MRs. In addition to the lattice parameters, the refractive
index of the buffer layer is also very important for achieving high perfor-
mance MRs. Though the requirement of high index contrast is relaxed here
as PBG is not essential in these MR designs, it was found that sufficient index
contrast below and above the single-layer MR is still critical in achieving
high reflection with broad spectral band.
3.2 Energy and phase penetration properties in MRs

In Kim et al. (2007), the authors reported excellent work on the phase
discontinuity Φr and the energy penetration depth Le, which were esti-
mated from the mode spacing in the FP cavity. In 2009, the in-plane optical
property of 2D PCS was reported by Sauvan et al. (2009). With fully vectorial
calculations, in heterostructure mirrors, the penetration length associated
with the delay was much larger than the characteristic damping length of
the energy distribution inside the mirror. This amazing property is evidence
that these two characteristic lengths are not necessarily correlated. Conse-
quently, the usual trade-off between short damping lengths and large pene-
tration lengths that is encountered in classical DBRs can be overcome with
carefully designed PhC structures. Later Zhao et al. (2010) investigated in
more detail the phase and energy penetration depths, and field distributions
of 1D subwavelength grating and 2D PCS reflectors.
In dielectric mirrors, the reflection is not an instantaneous process. It
includes a reflection time delay (τ) and energy storage in the mirrors. The
reflection delay increases the laser cavity round-trip time or the photon life-
time. The energy storage in mirrors decreases the modal volume and the
confinement factor in the small cavity, such that the cavity round-trip time
and the cavity volume have a comparable magnitude as the reflection delay
and the mirror storage (Babic and Corzine, 1922; Sauvan et al., 2009). The
reflection delay is directly related to the slope of the reflection phase shift Φr.
The relation can be expressed as τ ¼ ∂ Φr/∂ ω. The phase penetration depth,

Lp, is defined as the half-distance that light propagates in the incident
medium during this delay time,
1 vg ∂Φr
Lp ¼ vg τ ¼ (1)
2 2 ∂ω
where vg is the group velocity of the incident wave.
The energy storage is always associated with the parameter of energy pen-
etration depth, Le. It is the length that the field intensity decays to 1/e of its max-
imum from the edge of the cavity into the mirrors. However, this method is not
suitable for calculating the Le of the PC mirrors we discussed here, because the
guided modes are excited inside the mirrors. But Le can be estimated from the
mirror transmission or reflection based on the following equation,

h
T ¼ 1 R ¼ exp (2)
Le
where T is the transmission and h is the mirror thickness (Sauvan et al., 2009;
Zhao et al., 2010). While, for DBRs, Le can be obtained with (Coldren and
Corzine, 1995):

meff λ λ
Le ¼ + (3)
2 4n1 4n2
where meff ¼ tanh(2mr)/(2r) is the effective period number seen by the inci-
dent light. r ¼ (n1 n2)/(n1 + n2) and m are the actual period number in
DBRs. n1 and n2 are the refractive index of the two materials in DBR. In
the following we will numerically investigate the phase and energy penetra-
tion depth according to the above definitions.
Based on the simulation results shown in Fig. 2, compared to the DBR
reflectors, these new types of single-layer ultracompact broadband reflectors
can have larger and more complicated phase delays and smaller energy pen-
etration properties, which can be engineered via dispersion engineering for
large spectral-dependent phase delays, and ultrasmall energy penetration
depths. The work reported was mostly based on one set of optimized design
parameters for 1550 nm band reflectors. Following similar procedures, other
design parameters can be used for reflectors with different reflection require-
ments, as well as different phase delays, energy penetrations, and field distri-
butions. All the results and conclusions can be very helpful for the design of
resonant cavities for a wide range of photonic devices.
200 Weidong Zhou
A B
5 0.4
Lp,DBR Lp,1DSWG Lp,2DPCS Le,DBR Le,1DSWG Le,2DPCS
4
0.3
Le (μm)
Lp (μm)
3
0.2
2
0.1
1
0 0
1.5 1.55 1.6 1.5 1.55 1.6
Wavelength (μm) Wavelength (μm)
Fig. 2 (A) Calculated phase penetration depths for three types of reflectors and (B) the
energy penetration depths for three types of reflectors. Reproduced with permission from
Zhao, D., Ma, Z., Zhou, W., 2010. Field penetrations in photonic crystal Fano reflectors. Opt.
Express 18 (13), 14152–14158.
3.3 MR-VCSEL
Once two good MRs with high reflectivity and broader overlapped spectral
band are obtained, a multilayer Si-MR/QW/Si-MR (MR-VCSEL) cavity
can be build up (Fig. 3A). Here, we need to decide the cavity resonant (lon-
gitude) mode by selecting a suitable cavity length, i.e., the thickness of the
buffer layers between Si-MR and QW. Fig. 3B shows the reflection spectrum
(blue line) of one designed MR-VCSEL cavity. To further confirm this cavity
mode, the phase resonant condition (total phase change of one round-trip in
cavity is equal to an integer times of 2π) is also calculated according the reflec-
tion phase change ϕ of the top and bottom MRs, which is plotted as red-
dotted line in Fig. 3B. One can find the mode located at 1478 nm has a
m2π phase change, which means 1478 nm is a real cavity mode.
By employing FDTD, a short temporal Gaussian pulse is used to excite
the cavity modes and the quality factor of the cavity mode at 1478 nm is cal-
culated to be 4300 according to Q ¼ Re(ω)/2Im(ω) (Oskooi et al., 2010).
Then a longer temporal Gaussian pulse is used to excite the single cavity
mode and the stable field is recorded. The E-field distribution is demon-
strated in Fig. 3C with the red line, where the cavity index profile is also
plotted with the blue line.
Another design concern that needs to be taken care of is the separation
the lasing cavity modes from the QW layer as mentioned in Zhao et al.
(2012). Despite the relatively large field concentration in Si-MR layers,
the lasing mode field confinement factor is 5.6% for MR-VCSELs, which
is comparable to DBR-based VCSELs. This critical feature enables low
threshold lasing in the MR-VCSEL cavities.
t5: Top Si–MR

t4: SiO2
t3: QW
t2: SiO2
t1: Bottom Si–MR
SiO2
Si substrate
Lcav
B C
100 2 4
Electrical field intensity

Resonant phase (p) 3
Refractive index
1.5 2
Reflection (%)
80 1 1
0.8 0
1
0.6 –1
60 0.4 –2
t2 = 400 nm 0.5
t4 = 400 nm 0.2 –3
0 –4
40 0 –3 –2 –1 0 1 2 3
1410 1440 1470 1500 1530
Position in cavity (μm)
Wavelength (nm)
Fig. 3 Characteristics of designed MR-VCSEL cavity: (A) schematic of MR-VCSEL;
(B) calculated cavity resonance mode based on cavity reflection and phase resonant
condition; and (C) field distribution of cavity mode. Reproduced with permission from
Zhao, D., Yang, H., Chuwongin, S., Seo, J., Ma, Z., Zhou, W., 2012. Design of photonic crystal
membrane-reflector-based VCSELs. IEEE Photonics J. 4 (6), 2169–2175.
4. Photonic crystal bandedge surface-emitting laser

cavity design
The principle of the lasing modes in the photonic crystal surface-
emitting laser (PCSEL) cavities is based on the band edge modes of two-
dimensional (2D) PC close to the second-order Γ-point (Imada et al.,
2002; Meier et al., 1999a; Sakai et al., 2005), where the group velocities
of these modes are close to zero. In such 2D PC cavities, the periodic pat-
terned structure is designed close to the unpatented active region to provide
strong in-plane distributed feedback (DFB) and the standing wave is formed
in multiple directions due to the 2D DFB effect (Imada et al., 2002). Simul-
taneously, the in-plane propagating waves are diffracted toward the vertical
direction by the PC structure itself due to the first-order Bragg diffraction.
It is because of this multidirectional Bragg diffraction in these 2D PC
202 Weidong Zhou
A B
1540 D
Wavelength (nm)
C
1500 B
MQW A
1460
z Si PC
TE(s) TM(p)
q y 1420
SiO2 0 0.005 0.01
x SOI substrate kx (2p/a) → G – X
C D
4
3 Refractive index
1 2
1
|E|2
0.8 QW
y
0.6 0
Si–PC
0.4 –1
0.2
–2
0
0.5 1.5 2.5 3.5 x
Position in cavity (a)
Fig. 4 (A) Schematic of PC bandedge membrane laser on SOI. (B) Photonic band dia-
gram along Γ–X direction of Si-PC bandedge membrane laser cavity structure, with
the transferred InGaAsP QW mesas on top. (C) Field distribution of cavity mode and
(D) simulated field distributions. Reproduced with permission from Liu, S.-C., Zhao, D.,
Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterskio€ld-Hedlund, C., Hammar, M., Zhou, W., 2017.
Photonic crystal bandedge membrane lasers on silicon. Appl. Optics 56 (31), H67–H73.
structures, the cavity mode can lase over a large 2D area and gives a high
power output (Chua et al., 2014; Hirose et al., 2014).
A schematic of the PC bandedge membrane laser on Si is shown in
Fig. 4A. The laser cavity consists of an InGaAsP MQW membrane
heterostructure, and a Si-PC slab patterned on SOI with the BOx layer
thickness of 400 nm. The InGaAsP MQW layer design is based on the pre-
viously reported low temperature (LT) PC bandedge membrane laser,
which consists of a bottom InGaAs contact layer, top and bottom InP clad-
ding layers, and eight pairs of strain-compensated In0.76Ga0.24As0.83P0.17/
In0.485Ga0.515As0.83P0.17 with the center emission wavelength of 1535 nm
at room temperature (Zhao et al., 2016). In order to achieve room temper-
ature PC bandedge membrane lasers in such a cavity, it is highly desirable to
design a cavity with the following characteristics: (1) two dimension square
lattices Si-PC structure is designed to have a lattice constant of 480 nm,
based on the second-order Bragg diffraction condition (Chua et al.,
2014; Hirose et al., 2014; Imada et al., 1999, 2002; Martınez et al., 2009;
Meier et al., 1999; Noda et al., 2001). (2) The cavity mode wavelength
should match with MQW emission peak, with the hole radius and thickness
of Si-PC slab designed to be 72 nm (r ¼ 0.15a) and 190 nm, respectively. (3)
The separation between the InGaAsP MQW layer and Si-PC slab is reduced
to enhance the confinement factors and the in-plane coupling efficiency
(Sakai et al., 2006; Sato et al., 2012).
4.1 Bandedge mode design

The approximate effective refractive index neff of the whole cavity (InGaAsP
MQW active layer and Si-PC) is around 3.2 based on the average dielectric
constant in the initial design. In order to achieve surface emission at the
wavelength close to the communication wavelength range of 1550 nm,
the lattice constant a of the cavity is calculated to be 480 nm according to
second Bragg diffraction condition a ¼ λ/neff. Second, the air hole radius
is designed to be 72 nm (r ¼ 0.15a) in order to reduce the vertical radiation
loss to obtain high Q and low threshold (Zhao et al., 2016). Finally, to opti-
mize the cavity mode to match the MQW emission peak, the thickness of
the Si-PC slab is chosen to be around 190 nm (Kurosaka et al., 2009).
The band diagram for the Si-PC bandedge lasing cavity is shown in
Fig. 4B. It is obtained from the Fano resonances peaks or dips in the simu-
lated reflection or transmission spectra according to the relation between the
in-plane wave vector k// and the incident angle θ, k// ¼ 2π/λsin θ (Lee et al.,
2012), where λ is the wavelength and θ is the incident angle off the surface-
normal direction. By changing the incident angle from the surface-normal
direction (z-axis) to in-plane along the x-axis, the photonic bands in the
Γ–X direction can be obtained. The Γ-point corresponds to the surface-
normal incidence θ ¼ 0 degree. Here we mainly focus on a small value range
of incident angle θ from 0 to 2 degrees, which corresponds to the k// range
from 0 to 0.01 (2π/a) in the vicinity of the band edge. The spectral simu-
lation is done by the Fourier modal method with stanford stratified structure
solver (S4) software package (Liu and Fan, 2012). In Fig. 4B, the four bands in
the normalized wavelength range (1520 nm, 1550nm) are displayed with
blue-solid lines for the transverse electric (TE, or s) polarization, and red-
dashed lines for the transverse magnetic (TM, or p) polarization. One can
see that all bands are very flat at the edges close to the Г point, which indicate
a very small group velocity and longer light-matter interaction time for these
bandedge modes.
204 Weidong Zhou
The field profiles of mode D at λ ¼ 1540 nm is displayed in Fig. 4C. The

intensity of E-field is plotted with a red line, and the cavity index profile is
plotted with a blue line. The confinement factor can be calculated as follows:
Z
εjE j2 ∙dv
ГQW ¼ Z
QW
(4)
εjE j2 ∙dv
cav
The quality factor and the field distribution of this cavity mode are also
investigated. Their quality factors are obtained by Fano fitting the guided
resonances in the reflection spectra (Luk’yanchuk et al., 2010). In our cal-
culations, only the vertical radiation loss is considered as the in-plane loss and
the internal loss are negligible (Hirose et al., 2014). Hence, only the radiation
loss is considered for the infinite cavity using the following relationship:
2π
α¼ (5)
Q∙a
Therefore, the gain threshold can be derived as follows:
α
gth ¼ (6)
Г QW
The E-field along the center of cavity (left) and the cross-sectional views
of the E-field intensity of modes (right) are shown in Fig. 4D. One can see
the field is strongly confined inside the cavity. Based on the above funda-
mental equations, the confinement factor, quality factor, and radiation loss
of mode D (λ ¼ 1540 nm) is calculated to be 16.83%, 1.9 105 and
0.68 cm1, respectively. Single-mode lasing can be predicted to occur at
mode D which has the lowest gain threshold of 4.04 cm1.
4.2 Lateral cavity scaling

However, the scaling of on-chip lasers demands compact cavity sizes in
order to achieve high energy efficiency and high speed operation. The
reduction in lateral cavity size for PCSELs may result in a reduced
in-plane and out-of-plane confinement and thus increased lasing threshold
(Liang et al., 2012).
To investigate the lateral confinement effect, we considered two control
parameters. (Liu et al., 2018) First, we adjust the lateral cavity size by con-
trolling the pumping beam spot size, as shown in Fig. 5C. We also
A B
Layer Thickness
p-InP 130 nm
InGaAsP MQW 127 nm
n-InP 30 nm
z Si–PC
SiO 2 n-InGaAs 40 nm
q y
te 190 nm, a = 480 nm, and
bstra
Si su Si–PC
r/a = 0.14
x
SiO2 400 nm
Si substrate 525 μm
C
D
Etch
depth
z z
q y q y
x x
Fig. 5 (A) Schematic of PCSEL with 600 × 600 μm2 size. (B) Cavity layer definitions.
(C) Schematic of PCSEL with variation in lateral dimension (L). (D) Schematic of PCSEL
with variation in vertical etch depth. Reproduced with permission from Liu, S.-C., Zhao, D.,
Ge, X., Reuterskio€ld-Hedlund, C., Hammar, M., Fan, S., Ma, Z., Zhou, W., 2018. Size scaling of
photonic crystal surface emitting lasers on silicon substrates. IEEE Photonics J. 10 (3), 1–6.
experimentally fabricated cavities with various lateral dimensions. Second,

for a fixed lateral cavity size, we adjust the lateral confinement by controlling
the etch depth defining the cavity region as shown in Fig. 5D. Theoretical
investigations were first carried out to estimate the cavity Q factor and the
associated gain thresholds for different lateral and vertical dimensions and
then the corresponding threshold gains were estimated (Hirose et al.,
2014; Liu et al., 2017). In our calculations, only the in-plane loss and vertical
radiation loss are considered, and the internal loss is ignored (Hirose
et al., 2014).
Shown in Fig. 6A are the calculated threshold gains and Q factors for
different lateral cavity sizes. The Q factor drops rapidly while the threshold
gain increases with decreasing device size from 80 to 20 μm. This indicates
that the influence of in-plane loss has to be taken into account for finite
device sizes. As a result, power escapes from the edges of the laser cavity
due to weak lateral confinement.
To reduce the cavity loss due to the finite lateral dimensions, some lateral
confinement can be introduced. Here we investigate a simple lateral
206 Weidong Zhou
A B
350 7
1600 30
Gain threshold (cm–1)
Gain threshold (cm–1)

Quality factor (103)
Quality factor (103)

300 6
25
1200
20 250 5
800 15
200 4
10
400 150 3
5
0 0 100 2
10 20 30 40 50 60 70 80 90 0 100 200 300 400 500
Lateral dimension (μm) Etch depth (nm)
Fig. 6 (A) The simulated gain thresholds and Q factors for different device lateral dimen-
sions and (B) the simulated gain thresholds and Q factors for a device with lateral dimen-
sion of 48 μm (L ¼ 100 a) at different etching depths. Reproduced with permission from
Liu, S.-C., Zhao, D., Ge, X., Reuterskio€ld-Hedlund, C., Hammar, M., Fan, S., Ma, Z., Zhou, W.,
2018. Size scaling of photonic crystal surface emitting lasers on silicon substrates. IEEE
Photonics J. 10 (3), 1–6.
confinement scheme by controlling the etch depth of the lasing cavity (lat-
eral index contrast). The result is shown in Fig. 6B for a PCSEL on Si with a
lateral cavity dimension of 48 μm (L ¼ 100 a). By increasing the etch depth
from 0 to 500 nm, which corresponds to the complete removal of the
InGaAsP QW heterostructure and Si-PC cavity (Fig. 5D), the cavity Q fac-
tor increases from 2300 to 6800. We attribute this to the increased lateral
confinement due to the increased lateral index contrast.
5. Printed membrane laser cavities

The integration of InGaAsP QW heterostructure with the Si photonic
crystal structure was carried out with a printing process. For Si-MR/
InGaAsP QW/Si-MR MR-VCSEL cavity, two step printing processes
are involved (Yang et al., 2012). For the desired vertical index contrast,
low index SiO2 buffer layers are prepared between Si-MR and InGaAsP
QW layers. It is noted such a requirement on the low index buffer can limit
device performance due to poor thermal conductivity of the presence of low
index SiO2 buffer layers.
Different from the MR-VCSEL (Yang et al., 2012), PCSEL structure
only requires a one-step printing process for the heterogeneous integration
of III–V gain structure on top of the patterned Si-PC cavity. It does not
require a low index buffer layer to be placed in between the active region
and the Si-PC layer. In fact, it is critical to minimize the separation between
QW layer and Si-PC cavity layer in order to obtain a high confinement fac-
tor which is critical to low threshold lasing.
An SOI wafer with 230 nm top Si layer and 400 nm BOx layer was used
here. The top Si layer was first thinned down to 190 nm by thermal oxida-
tion and selective wet etching processes. Next, the 2D square lattice PC pat-
tern with a size of 1 1 mm2 was defined with ZEP-520A resist by using the
standard electron beam lithography (EBL). The sample was then etched with
the ZEP-520A resist pattern as the mask to create a high index contrast
between Si and air by using the dry etching process. The scanning electron
microscopic (SEM) top view of Si-PC on SOI substrate is shown in Fig. 7A.
The lattice constant a and air hole radius r of the fabricated PC are 480 and
72 nm, respectively.
Fig. 7 (A) Zoomed-in SEM top view of the defect-free Si-PC cavity. (B) An SEM image of
InGaAsP MQW mesa transferred onto a Si-PC. (C) Cross-sectional view of the cavity and
(D) micrograph image of the transferred mesa array on a Si-PC. Reproduced with permis-
sion from Liu, S.-C., Zhao, D., Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterskio€ld-Hedlund, C.,
Hammar, M., Zhou, W., 2017. Photonic crystal bandedge membrane lasers on silicon. Appl.
Optics 56 (31), H67–H73.
208 Weidong Zhou
InGaAsP MQW structure was grown on (001) n-InP substrate by metal

organic chemical vapor deposition, with a 500 nm InGaAs sacrificial layer
between the top QW structure and the bottom InP substrate. Substrate
removal and wax protection processes were used here for the release of
the large-area InGaAsP QW structure (Zhao et al., 2016). The transfer
printing process was utilized to attach the InGaAsP MQW to the SOI sub-
strate. The Si-PC on SOI substrate was first cleaned by solvent, followed by
standard RCA cleaning to remove organic residues from the surface. The
lifted InGaAsP MQW membrane is then carefully printed on a patterned
Si-PC cavity. Subsequently, wax was removed with trichloroethylene
(TCE) and isopropyl alcohol (IPA) solutions. Finally, a 300 300 μm2 active
MQW square size is defined by photolithography and selective wet etching
process. Fig. 7B and C shows the SEM top view and cross-sectional view of a
transferred InGaAsP MQW membrane on a Si-PC cavity on a SOI sub-
strate. Shown in Fig. 7D is a micrograph image of PC bandedge membrane
lasers on SOI substrate. It is worth noting that this process could be applied at
room temperature and is capable of a 5-μm alignment accuracy.
6. MR-VCSELs based on 1D PCS/grating mirrors

In addition to the 2D PCS-based MRs, 1D photonic crystal mirrors or
high contrast grating structures have been proposed and demonstrated as
well (Boutami et al., 2007; Huang et al., 2007). The most straight forward
way of incorporating MR’s is to grow the top MR layer monolithically on
the laser heterostructure, to form hybrid VCSEL cavities with top MR and
bottom DBR configurations. Huang et al. (2007) reported an electrically
pumped 850 nm VCSEL on a GaAs substrate (Fig. 8). Excellent performance
was reported with single-mode emission and a side-mode suppression ratio
(SMSR) of 45 dB. Boutami et al. (2007) reported an optical 1550 nm
VCSEL on an InP substrate, with pulsed operation at room temperature
and a threshold power around 15 mW.
The ultimate goal for this configuration is to replace both top and bottom
MR’s with single-layer MR’s. However, in this case, monolithic growth is
not possible. One straight forward technique is to use the wafer bonding
technique for the integration of a gain medium on top of the bottom
MR’s. Sciancalepore et al. (2012) reported an optically pumped 1550 nm
VCSELs on Si. They used a double 1D PCM employing III–V/SiO2 molec-
ular bonding and amorphous-Si deposition processes. A 2-in. III–V wafer
was bonded to a SiO2 coated 200 mm SOI wafer, followed by an InP
Fig. 8 Hybrid top MR/bottom DBR VCSEL configuration: suspended GaAs high contract
grating (HCG) reflector monolithically grown on GaAs substrate. Reproduced with per-
mission from Huang, M.C.Y., Zhou, Y., Chang-Hasnain, C.J., 2007. A surface-emitting laser
incorporating a high-index-contrast subwavelength grating. Nat. Photonics 1 (2), 119–122.
substrate removal process. The polarization-dependent top 1-D PCM was

formed on a deposited amorphous-Si film using deep-UV lithography
and reactive ion etching.
7. MR-VCSEL characterization
Optically pumped devices are mounted inside a cryostat and charac-
terized with a monochromator-based microphotoluminescence (μ-PL) set
up using quasi-continuous wave (c.w.) 532 nm laser pumping (with 50%
duty cycle).
Shown in Fig. 9A is the L–L plot (light output for different pump pow-
ers) and the corresponding spectral linewidths measured at T ¼ 50 K. The
threshold pump power is 8 mW, or 0.32 kW/cm2. The measured spectral
linewidths reduce from 30 nm below threshold to 0.6–0.8 nm above thresh-
old. The measured spectral outputs are shown in Fig. 9B, for pump powers
below, at, and above thresholds (points (i, ii, iii, iv) at L–L curve). The lasing
spectral linewidth is 8 Å, which is limited by the measurable resolution of
monochromator. The relative peak location shift shown in Fig. 9B from bias
levels (ii) and (iii, iv) is mostly related to mode hopping and temperature rise
inside the active region at higher pump power levels.
210 Weidong Zhou
A B
80 100 80
T = 50 K, D = 100 µm iv
Output power (µW)

Output power (µW)
Linewidth (nm)
60 60 iv
10
40 40
iii
20 1 20 iii
i ii ii
i
0 0
0 4 8 12 16 20 1430 1450 1470 1490 1510
Pump power (mW) Wavelength (nm)
Fig. 9 (A) The L–L plot (light output for different pump powers) and the corresponding
spectral linewidths at T ¼ 50 K and (B) the measured spectral outputs for pump powers
below, at, and above thresholds. Reproduced with permission from Yang, H., Zhao, D.,
Chuwongin, S., Seo, J.-H., Yang, W., Shuai, Y., Berggren, J., Hammar, M., Ma, Z.,
Zhou, W., 2012. Transfer-printed stacked nanomembrane lasers on silicon. Nat. Photonics
6 (9), 615–620.
A B
20
Power (a.u.)
Lth (mW)
1
15
0.5
0 120 10
1520 90 T0 = 125 K
)
70 (K
Wa 1500 ure 5
vele 1480 50 t
ngth p era 20 40 60 80 100 120
(nm 1460 15 m Temperature (K)
) Te
Fig. 10 (A) Measured lasing spectral at different temperatures under 532 nm laser
pumping and (B) lasing threshold as a function of temperature. Reproduced with permis-
sion from Yang, H., Zhao, D., Chuwongin, S., Seo, J.-H., Yang, W., Shuai, Y., Berggren, J.,
Hammar, M., Ma, Z., Zhou, W., 2012. Transfer-printed stacked nanomembrane lasers on
silicon. Nat. Photonics 6 (9), 615–620.
This MR-VCSEL LT device was also characterized at different temper-

atures up to 120 K. Fig. 10A shows the normalized lasing spectra (arbitrary
units, a.u.) above the threshold at T ¼ 15, 50, 70, 90, and 120 K. The lasing
peak red shifts as T increases and multimode lasing for most T cases at a rate
(dλc/dT) close to the simulated 0.088 nm/K. There is a mode hopping
occurred below and above the operation temperature of 80 K. As T rises
higher than 125 K, there was no lasing any more, which is mainly limited
by the MR reflector bandwidth. The LT design cavity has three lasing wave-
lengths of 1448 nm (at 10 K), 1478 nm (at 50 K), and 1520 nm (at 120 K)
which match well with the cavity resonances. Fig. 10B shows the lasing
threshold (Lth) as a function of T. Lth lineally increases with T, which
matches the QW PL efficiency linearly decreases as T.
In order to estimate the lasing gain threshold in MR-VCSEL cavity,
detailed investigation was also carried out to estimate the associated phase
discontinuities in MR mirrors, as well the field penetration depth and effec-
tive cavity length in the optimized MR-VCSEL cavity. The results are sum-
marized in Table 2. Based on lasing gain threshold conditions,
Γgth ¼ αi + αSi + 1=Lcav lnð 1=Rt Rb Þ (7)
Finally, lasing gain threshold can then be estimated, based on the cavity
gain and loss characteristics reported earlier from similar QW structures
(Rapp et al., 1999). Based on our design, MR-VCSEL can lase with a gain
threshold of 840–1300 cm1, depending on the MR reflectivity.
The cavity design here can definitely lead to high performance MR-
VCSELs with very low threshold and high efficiency, as shown in Table 2.
For comparison, also shown in Table 2 are the typical parameters for long
wavelength VCSELs based on conventional InGaAsP/InP DBRs. As can
be seen, the MR-VCSEL has a very similar lasing gain threshold with con-
ventional InGaAsP/InP DBR-based VCSEL, even though MR-VCSELs
have smaller cavities with much shorter effective cavity lengths (2.4 μm vs
5–20 μm) (Karim et al., 2000; Rapp et al., 1997).
Table 2 MR-VCSEL lasing threshold estimates

Cavity types MR-VCSEL on Si DBR VCSEL on InP
Top reflector Si NM MR Five pairs GaAs/AlxOy DBR
Bottom reflector Si NM MR InGaAsP/InP DBR
Wavelength λres 1.513 μm 1.54 μm
Reflection Rt 98.2% (99.6%) 99.6%
Reflection Rb 98.9% (99.6%) 99.6%
Confinement Γ 5.6% 4.5%
Cavity length Lcav 1.21 μm 1.0 μm
Effective cavity Leff 3.783 μm 13.1 μm
Cavity loss αi + αSi 40 cm1 43.1 cm1
Gain threshold gth 1283 cm1 (843 cm1) 1224 cm1
212 Weidong Zhou
A B
100
80 On SOI 1 P = 0.32 kW/cm2
T = 300 K T = 300 K
Peak intensity (a.u.)
c.w. 0.8
Intensity (a.u.)
Linewidth (nm)
60
10
0.6
40
0.4
20 1
0.2
0 0
0.1 0.2 0.3 0.4 0.5 0.6 1510 1530 1550 1570
Power density (kW/cm2) Wavelength (nm)
Fig. 11 (A) Lasing power and linewidth vs input power density at room temperature
and (B) emission spectrum of the laser above threshold. Reproduced with permission
from Liu, S.-C., Zhao, D., Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterskio€ld-Hedlund, C.,
Hammar, M., Zhou, W., 2017. Photonic crystal bandedge membrane lasers on silicon. Appl.
Optics 56 (31), H67–H73.
8. PCSEL characterization
PCSEL devices were optically pumped from the surface-normal direc-
tion via a long working distance objective lens (10) by a c.w. green (532 nm)
laser. The diameter of the excitation spot is around 110 μm. The optical char-
acteristics of this PC bandedge membrane laser device were then measured by
a monochromator-based μ-PL set-up. Fig. 11A shows the light output and
linewidth as a function of the pump power measured on the surface of the
focusing plane. A threshold power density of 0.25 kW/cm2 was obtained.
The measured spectral linewidth was reduced from 14nm (below threshold)
to 0.54nm (above threshold). Notice a thermal roll-over was observed at opti-
cal pumping intensities higher than 0.4 kW/cm2. A spectrum of the laser
above the threshold is shown in Fig. 11B where the lasing wavelength is
1542 nm, very close to the predicted value of mode D in the simulation.
The lower threshold pump power density is a result of the high Q of mode
D. All the other modes are possible to get lasing with sufficient pump energy.
Fig. 12 shows the lasing spectrum measured at a pump power density of
0.32 kW/cm2. A clear single-mode operation was observed. A zoomed-in
lasing spectrum plotted in semilog scale is shown in Fig. 12A. The peak
wavelength, full width at half maximum (FWHM) of the spectrum, and
SMSR are 1542 nm, 0.54 nm, and 31.8 dB, respectively. Fig. 12B is a mea-
sured far-field lasing profile with single-mode output at the focusing plane of
the collimator.
A B
0 P = 0.32 kW/cm
2 3584
3328
3072
T = 300 K 2816
Relative intensity (dB)
–5 FWHM = 0.54 2560

2304
SMSR = 31.8 dB
2048
–10 nm 1792
1536
1280
–15 1024
768
512
–20 256
0
–25
–30
–35
–40
1525 1535 1545 1555
Wavelength (nm)
Fig. 12 (A) Lasing spectral output plotted in semilog scale above the pumping thresh-
old and (B) 3D far-field laser profile. Reproduced with permission from Liu, S.-C., Zhao, D.,
Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterskio€ld-Hedlund, C., Hammar, M., Zhou, W., 2017.
Photonic crystal bandedge membrane lasers on silicon. Appl. Optics 56 (31), H67–H73.
9. PCSEL on bulk silicon substrate

As mentioned earlier, the low index buried oxide layer underneath the
Si-PC layer limits the heat dissipation performance of the laser cavity due to
relatively low thermal conductivity of oxide. Thus it is of great importance
to introduce the PC bandedge membrane laser directly built on bulk Si sub-
strate, where the Si-PC structure is used as the bottom cladding layer and the
MQW membrane can be printed directly on bulk Si substrate without low
index oxide buffer layer. The design parameters were carefully chosen for
the bottom Si-PC as more restrict index constrain was involved as compared
with laser on SOI substrate with low index buffer. A schematic of a success-
ful design of this structure is shown in Fig. 13A, which consists of a trans-
ferred InGaAsP MQW membrane on a Si-PC cavity on a bulk Si substrate.
A bulk Si wafer with 550 μm thickness was used here. The 2D square lattice
PC pattern on bulk Si substrate was made following the same EBL patterning
and dry etching processes. The SEM top view of Si-PC cavity on bulk Si
substrate is shown in Fig. 13B. The lattice constant (a), air hole radius (r),
and etched depth of the fabricated PC are 540, 243, and 400 nm, respec-
tively. The approximate refractive index neff of the InGaAsP MQW active
layer and Si-PC was around 3.2 and 2.2 based on the average dielectric con-
stant in the design. Different from the PC bandedge membrane laser on SOI
substrate, the PC structure here was considered as the low refractive index
cladding layer of the cavity. The field profile of the lasing mode is displayed
214 Weidong Zhou
A B
MQW
MQW
Si PC Si–PC
z Bulk Si
substrate Bulk Si
y
200 nm
x
C D
4
3
Refractive index
1 2
QW
0.8 1 y
|H|2
0.6 0 Si–PC
0.4 –1
0.2
–2
0 x
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Position in cavity (μm)
Fig. 13 (A) Schematic of PC bandedge membrane laser on bulk Si substrate. (B) Cross-
sectional view of the Si-PC bandedge membrane laser cavity structure on bulk Si sub-
strate, with the transferred InGaAsP QW mesas on top. (C) Field distribution of the cavity
mode and (D) simulated in-plane and cross-sectional field distributions. Reproduced with
permission from Liu, S.-C., Zhao, D., Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterskio€ld-Hedlund,
C., Hammar, M., Zhou, W., 2017. Photonic crystal bandedge membrane lasers on silicon.
Appl. Optics 56 (31), H67–H73.
in Fig. 13C, where the integrated H-field intensity of the whole cavity along
the z-axis direction is plotted together with the cavity index profile. The
cross-sectional view of H-field intensity of lasing mode is shown in
Fig. 13D. Note here that the large air hole radius was designed to ensure
the high refractive index contrast. One can see the field was confined inside
the thin cavity due to high index contrast above and below the laser cavity.
Moreover, the thickness of the Si-PC slab was chosen to be 400 nm to obtain
high Q value by reducing the vertical radiation loss. Finally, The Q value of
the lasing mode can be theoretically calculated around 3000, which would
be able to achieve the single-mode lasing in the 160 K operation temperature
(Zhao et al., 2016).
To evaluate the laser performance experimentally, the PC bandedge
membrane laser on bulk Si substrate was mounted on a temperature-
controlled stage inside a cryostat and characterized by the same μ-PL set-
up system in our previous works (Yang et al., 2012; Zhao et al., 2016).
Fig. 14A shows the light output and linewidth as a function of the pump
A B
3 100
On bulk Si 1 P = 0.5 kW/cm2
2.5 T = 160 K T = 160 K
Peak intensity (a.u.)
c.w. 0.8
Linewidth (nm)
Intensity (a.u.)
2 10
0.6
1.5
1 0.4
1
0.5 0.2
0 0
0.2 0.3 0.4 0.5 0.6 0.7 1420 1440 1460 1480
Power density (kW/cm2) Wavelength (nm)
Fig. 14 (A) Lasing intensity and linewidth vs input power density at T ¼ 160 K for PCSEL
on bulk Si substrate and (B) emission spectrum of the laser above threshold. Reproduced
with permission from Liu, S.-C., Zhao, D., Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterskio€ld-
Hedlund, C., Hammar, M., Zhou, W., 2017. Photonic crystal bandedge membrane lasers
on silicon. Appl. Optics 56 (31), H67–H73.
power measured on the surface of the PC bandedge membrane laser on bulk

Si substrate. The laser threshold pump power density of 0.4 kW/cm2 was
obtained, where it was around a factor of two greater than PC bandedge
membrane laser on SOI substrate at similar operation temperatures. We
attribute this to the much lower optical confinement factor and less Q value
within the cavity due to larger air hole radius. The calculated Q factor is
3.27 103, which is two orders lower than the Q for PCSEL on SOI. How-
ever, there are few strategies to improve the laser characteristics of present
cavity design on bulk Si substrate, such as the introduction of vertical asym-
metry in the air hole structure and effective utilization of tailoring the ver-
tical etched profile. The Q value can be theoretically increased much higher
above 3000, which would achieve lower threshold pump power density in
the 160 K operation temperature. Fig. 14B presents the measured lasing
spectrum above the threshold. The single-mode lasing peak wavelength
and linewidth are 1452 and 0.87 nm in the 160 K operation temperature.
10. PCSEL thermal performance analysis

In order to compare the thermal performance, here we prepared two
PC bandedge membrane lasers built onto bulk Si substrate and SOI substrate,
respectively. The laser cavity for both devices is designed to operate at similar
temperatures and emission wavelengths. The thicknesses of each layer of the
InGaAsP MQW membrane and the parameter of Si-PC are listed in Table 3.
216 Weidong Zhou
Table 3 Cavity design structure of laser on bulk Si and SOI

Layer materials Laser on bulk Si Laser on SOI
p-InP 130 nm
InGaAsP MQW 127 nm
n-InP 130 nm
n-InGaAs 40 nm
Si-PC 400 nm, a ¼ 540 nm 190 nm, a ¼ 480 nm
r/a ¼ 0.45 r/a ¼ 0.15
SiO2 0 400 nm
Si substrate 525 μm
The bulk Si substrate is expected to dissipate more heat along the vertical
direction from the surface of the cavity to the bottom substrate and heat sink.
A figure of merit to quantify the heat dissipation out of the cavity is the ther-
mal resistance (Ferrara et al., 2015; Flick et al., 2011). By extracting the shift
of emission peak wavelength with varying the temperature and pump
power, the thermal resistance of the laser can be mathematically defined
by the following ratio:
Δλ Δλ
|T ¼ const |T ¼ const
ΔPheat ΔPp
Rth ¼ (8)
Δλ Δλ
|P ¼ const |P ¼ const
ΔT ΔT
Here, the ΔPheat is the heat generated from the cavity and ΔPp is the total
pump power. We assume heat generation is primarily in the QW section
(Lindberg et al., 2005), and it is linearly proportional to the optical pump
power for the operation window between threshold and roll-over. We first
characterize the PC bandedge membrane laser on bulk Si substrate. The
wavelength spectrum dependence of the pumping power from the PC
bandedge membrane laser on bulk Si substrate was measured as shown in
Fig. 15A. The peak of the emission wavelength slightly shifted toward lon-
ger wavelengths when the pump power increased from 25 to 50 mW
at 160 K. In our measurement, the spectral shift vs power ratio presents
the spectral shift vs power generated from the QW region. Second, the
spectral dependence of the operation temperature varying from 120 to
200 K at a constant incident power was also measured and shown in
Fig. 15B. Please note the cavity design of the PC bandedge membrane laser
A B
T = 160 K P = 40 mW
Normalized relative intensity (a.u.)

1 P = 25 mW 1 T = 120 K
P = 30 mW T = 140 K
Relative intensity (a.u.)
P = 40 mW T = 160 K
0.8 P = 50 mW 0.8 T = 180 K
T = 200 K
0.6 0.6
0.4 0.4
0.2 0.2
0 0
1440 1445 1450 1455 1460 1440 1445 1450 1455 1460
Fig. 15 (A) Lasing spectra at different pump power above threshold at constant
160 K. (B) Lasing spectra at different operation temperature with constant 40 mW
pump power. Reproduced with permission from Liu, S.-C., Zhao, D., Liu, Y., Yang, H.,
Sun, Y., Ma, Z., Reuterskio€ld-Hedlund, C., Hammar, M., Zhou, W., 2017. Photonic crystal
bandedge membrane lasers on silicon. Appl. Optics 56 (31), H67–H73.
A B
1 Laser on SOI Laser on SOI
5
Laser on bulk Si Laser on bulk Si
0.8 4
Peak shift (nm)
Peak shift (nm)
0.6 3
0.4 2
0.2 1
0 0
25 30 35 40 45 50 120 140 160 180 200
Pump power (mW) Temperature (K)
Fig. 16 (A) Emission peak wavelength shift vs pump powers for PC bandedge mem-
brane laser on SOI and on bulk Si substrate. (B) Emission peak wavelength shift vs oper-
ation temperatures for PC bandedge membrane laser on SOI and on bulk Si substrate.
Reproduced with permission from Liu, S.-C., Zhao, D., Liu, Y., Yang, H., Sun, Y., Ma, Z.,
Reuterskio€ld-Hedlund, C., Hammar, M., Zhou, W., 2017. Photonic crystal bandedge mem-
brane lasers on silicon. Appl. Optics 56 (31), H67–H73.
on bulk Si substrate was optimized for device operation around 160 K. Single-
mode lasing operation was obtained at the 120–200 K range for the pump
power at 40 mW. The wavelength shift vs pumping power is calculated to
be 0.0125 nm/mW, and the shift vs temperature was 0.0412 nm/K.
Fig. 16A and B shows the measured emission wavelength peak shift at
different pump powers and varying operation temperatures for both PC
bandedge membrane lasers on bulk Si substrate and on SOI substrate
218 Weidong Zhou
operation at 160–180 K. For the PC bandedge membrane laser on bulk Si

substrate, the thermal resistance was extracted from Fig. 16A and B and cal-
culated to be 303.4 K/W. On the other hand, the thermal resistance of PC
bandedge membrane laser on SOI was calculated to be 558.5 K/W. The
thermal resistance of PC bandedge membrane laser on bulk Si substrate is
therefore around a factor of two times lower than that of a structure with
a BOx layer. We attributed the higher thermal resistance to the low thermal
conductivity of the BOx layer. It is also worth noting that the thermal resis-
tance values calculated here is a relative value, as the absolute value should be
based on the exact heat source (ΔPheat), which is not known here.
11. Electrically pumped PCSEL characterization

Finally, we present some recent results on electrically pumped
PCSELs. A buried tunnel junction (BTJ) is incorporated here for efficient
charge injection and lateral carrier confinement. As shown schematically in
Fig. 17A, the PCSEL cavity consists of a transferred InGaAsP multiquantum
well (MQW) heterostructure active region, printed on a single-layer Si-PC
cavity. The thicknesses of each layer of the InGaAsP MQW membrane
and the parameters of Si-PC are listed in Fig. 17B. BTJ n++/p++ current
injection structures with different sizes were formed using an etch and epi-
taxial regrowth process, including pass marks for continued processing on
the Si substrate. The laser cavity structure parameters are optimized for
low threshold lasing, following the similar procedures reported earlier
(Liu et al., 2017; Zhao et al., 2016). Shown in Fig. 18A are the simulated
A B
Description Materials BTJ-PCSEL
Contact layer: n InGaAs 40 nm
Cladding layer: n InP 80 nm
BTJ: n++ InGaAs 20 nm
BTJ: p++ InAlGaAs 10 nm
Cladding layer: p InP 265 nm
MQW InGaAsP 127 nm
Cladding layer: n InP 31.5 nm

Contact layer: n InGaAs 40 nm
Photonic crystal Air hole Si–PC 125 nm, r/a = 0.15
BOX SiO2 400 nm
Substrate Si substrate 525 μm
Fig. 17 Electrically pumped PCSELs on SOI substrate: (A) schematic and (B) detailed
heterostructure and cavity design parameters.
A B
1 10–1
10–2
0.8 10–3
Transmission
Current (mA)
10–4
0.6
10–5
0.4 10–6
10–7
0.2
TE 10–8
TM
0 10–9
1480 1500 1520 1540 –1 0 1 2 3 4
Wavelength (nm) Voltage (V)
Fig. 18 Electrically pumped PCSELs on SOI substrate: (A) simulated transmission spectra
for both TE and TM modes and (B) measured IV curve for the electrically pumped PCSEL.
transmission spectra of the cavity, with blue lines for the transverse electric
(TE, or s) polarization, and red lines for the transverse magnetic (TM, or p)
polarization. Substrate removal process is used in the device fabrication for
the release of top large-area InGaAsP MQW membrane from the native InP
substrate. The membrane is then transferred onto Si-PC cavity. Top n-type
ring contact is formed by e-beam evaporation, with alloy contact consisting
of Ni/Ge/Au/Ti/Au (10/20/30/10/200 nm). Mesa etching is then carried
out down to the bottom n-contact layer, followed with another e-beam evap-
oration process for the bottom n-contact formation. A plasma-enhanced
chemical vapor deposition SiO2 layer is deposited for laser planarization,
followed by interconnect layer formation. The complete device is then
mounted on a chip carrier with wire bonding process for device characteri-
zation. The current–voltage (IV) characteristics were measured and evaluated
for different BTJ sizes, with one curve plotted in semilog scale for electrically
pumped PCSEL shown in Fig. 18B. Excellent IV characteristics were shown
here, essential for efficient charge injection and lasing.
The device characterization was carried out with c.w. current injection
at room temperature. The output was measured in the direction normal to
the surface via a long working distance objective lens and sent into a
monochromator. Fig. 19A shows the measured electroluminescence (EL)
spectrum at various biased currents for a PCSEL with an 80 80 μm2 BTJ
mesa region. A dominant single-mode lasing emission peak at 1504 nm
appears above the threshold current. Fig. 19B shows the light–current–voltage
(LIV) characteristics. The turn on voltage is about 0.8 V. The threshold cur-
rent is about 2 mA.
220 Weidong Zhou
A B
8 4 4
I = 15 mA
I = 5 mA
EL intensity (a.u.)
6 I = 2.5 mA 3 3
Intensity (a.u.)
Voltage (V)
4 2 2
2 1 1
0 0 0
1400 1450 1500 1550 1600 0 2 4 6 8 10
Wavelength (nm) Current (mA)
Fig. 19 (A) The emission spectra at different current injection levels measured at room
temperature and (B) the light–current–voltage (LIV) characteristics for an 80 80 μm2
electrically pumped PCSEL on Si.
12. Summary
In summary, we review here recent progresses on photonic crystal
surface-emitting membrane lasers for on-chip integration with CMOS elec-
tronics. Defect-free photonic crystal structures are incorporated into laser
cavities as ultracompact MRs for DBR-free VCSELs and as bandedge cav-
ities evanescently coupled with gain medium for PCSELs. Lateral cavity size
effect was also investigated for cavity scaling. Electrically injected PCSELs
with BTJ-based InGaAsP QW heterostructures were also demonstrated.
Further research is needed to address device efficiencies. While high effi-
ciency VCSELs and PCSELs on native III–V substrates have been reported
earlier with wall-plug efficiencies greater than 50%, the efficiencies in the
heterogeneously integrated DBR-free MR-VCSELs and Si-based PCSELs
are largely impacted by the cavity coupling and thin membrane surface and
interface qualities. Device scaling in both vertical and lateral directions can
also lead to improved device modulation speed and operation data transfer
efficiency, with ultimate objective of subfemto joule per bit energy
efficiency.
Thermal and reliability are another set of issues to be addressed. Since
PBG is not essential in these photonic crystal structures reported here,
relaxed index contrast is possible to have photonic crystal lasing cavities built
on low index contract material systems. However certain index contract is
still needed, which may result in the need of low index buffers in the lasing
cavity. This issue needs to be further addressed to avoid the use of low index
dielectric (e.g., SiO2, Si3N4) due to the poor thermal conductivities and the
strain-induced reliabilities issues.
On the other hand, new functional properties should also be expected
from photonic crystal surface-emitting membrane lasers, as the modulation
of photonic crystal cavity and the gain/loss modulation of the active media
can result in lasers with unique spectral, spatial, and beam properties.
Engineered beam shapes can be designed. Large-area single-mode lasers
are also feasible. Laser beams can also be steered with certain angles. With
the proper dispersion engineering, it is possible to have beam emission
in-plane.
Acknowledgments
The authors acknowledge contributions from their current and former students, as well as
collaborators involved in the research work reviewed here. Special thanks to Mattias
Hammar from Royal Institute of Technology who provided QW materials for the work.
We also thank the fabrication support from the University of Texas at Arlington
Nanotechnology Research Center, and the University of Texas at Austin, part of
NSF NNIN.
References
Altug, H., Englund, D., Vuckovic, J., 2006. Ultrafast photonic crystal nanocavity laser. Nat.
Phys. 2 (7), 484–488.
Babic, D., Corzine, S., 1922. Analytic expressions for the reflection delay, penetration depth,
and absorptance of quarter-wave dielectric mirrors. IEEE J. Quantum Electron.
28, 514–524.
Bakir, B.B., Seassal, C., Letartre, X., Viktorovitch, P., Zussy, M., Di Cioccio, L.,
Fedeli, J.M., 2006. Surface-emitting microlaser combining two-dimensional photonic
crystal membrane and vertical Bragg mirror. Appl. Phys. Lett. 88, 081113.
Balakrishnan, G., Jallipalli, A., Rotella, P., Huang, S., Khoshakhlagh, A., Amtout, A.,
Krishna, S., Dawson, L.R., Huffaker, D.L., 2006. Room-temperature optically pumped
(Al) GaSb vertical-cavity surface-emitting laser monolithically grown on an Si (100) sub-
strate. IEEE J. Sel. Top. Quantum Electron. 12 (6), 1636–1641.
Bhattacharya, P., Sabarinathan, J., Topol’ancik, J., Chakravarty, S., Yu, P., Zhou, W., 2005.
Quantum dot photonic crystal light sources. Proc. IEEE 93 (10), 1825–1838.
Boutami, S., Bakir, B., Regreny, P., Leclercq, J., Viktorovitch, P., 2007. Compact 1.55 μm
room-temperature optically pumped VCSEL using photonic crystal mirror. Electron.
Lett. 43, 282.
Chen, R., Tran, T.T.D., Ng, K.W., Ko, W.S., Chuang, L.C., Sedgwick, F.G.,
Chang-Hasnain, C., 2011. Nanolasers grown on silicon. Nat. Photonics 5 (3), 170–175.
Chua, S.L., Chong, Y., Stone, A.D., Soljacic, M., Bravo-Abad, J., 2011. Low-threshold las-
ing action in photonic crystal slabs enabled by Fano resonances. Opt. Express
19, 1539–1562.
Chua, S.L., Lu, L., Bravo-Abad, J., Joannopoulos, J.D., Soljacic, M., 2014. Larger-area
single-mode photonic crystal surface-emitting lasers enabled by an accidental Dirac
point. Opt. Lett. 39 (7), 2072–2075.
222 Weidong Zhou
Coldren, L., Corzine, S., 1995. Diode Lasers and Photonic Integrated Circuits. Wiley,
New York.
Colombelli, R., Srinivasan, K., Troccoli, M., Painter, O., Gmachl, C.F., Tennant, D.M.,
Sergent, A.M., Sivco, D.L., Cho, A.Y., Capasso, F., 2003. Quantum cascade surface-
emitting photonic crystal laser. Science 302 (5649), 1374–1377.
Dowling, J.P., Scalora, M., Bloemer, M.J., Bowden, C.M., 1994. The photonic band edge
laser: a new approach to gain enhancement. J. Appl. Phys. 75 (4), 1896–1899.
Ellis, B., Mayer, M.A., Shambat, G., Sarmiento, T., Harris, J., Haller, E.E., Vuckovic, J.,
2011. Ultralow-threshold electrically pumped quantum-dot photonic-crystal nanocavity
laser. Nat. Photonics 5 (5), 297–300.
Phys. Rev. B 65 (23), 235112.
Ferrara, J., Yang, W., Zhu, L., Qiao, P., Chang-Hasnain, C.J., 2015. Heterogeneously inte-
grated long-wavelength VCSEL using silicon high contrast grating on an SOI substrate.
Opt. Express 23, 2512–2523.
Flick, T., Becks, K.H., Dopke, J., M€attig, P., Tepel, P., 2011. Measurement of the thermal
resistance of VCSEL devices. J. Instrum. 6, C01021.
Furukawa, A., Sasaki, S., Hoshi, M., Matsuzono, A., Moritoh, K., Baba, T., 2004. High-
power single-mode vertical-cavity surface-emitting lasers with triangular holey structure.
Appl. Phys. Lett. 85, 5161–5163.
Hirose, K., Liang, Y., Kurosaka, Y., Watanabe, A., Sugiyama, T., Noda, S., 2014. Watt-class
high-power, high-beam-quality photonic-crystal lasers. Nat. Photonics 8 (5), 406–411.
Huang, M.C.Y., Zhou, Y., Chang-Hasnain, C.J., 2007. A surface-emitting laser incorporat-
ing a high-index-contrast subwavelength grating. Nat. Photonics 1 (2), 119–122.
Hwang, J., Ryu, H., Song, D., Han, I., Park, H., Jang, D., Lee, Y., 2000. Continuous room-
temperature operation of optically pumped two-dimensional photonic crystal lasers at
1.6 um. IEEE Photon. Technol. Lett. 12 (10), 1295–1297.
Iga, K., 2000. Surface-emitting laser-its birth and generation of new optoelectronics field.
IEEE J. Sel. Top. Quantum Electron. 6 (6), 1201–1215.
Imada, M., Noda, S., Chutinan, A., Tokuda, T., Murata, M., Sasaki, G., 1999. Coherent
two-dimensional lasing action in surface-emitting laser with triangular-lattice photonic
crystal structure. Appl. Phys. Lett. 75, 316.
Imada, M., Chutinan, A., Noda, S., Mochizuki, M., 2002. Multidirectionally distributed
feedback photonic crystal lasers. Phys. Rev. B 65 (19), 195306.
Karagodsky, V., Sedgwick, F.G., Chang-Hasnain, C.J., 2010. Theoretical analysis of
subwavelength high contrast grating reflectors. Opt. Express 18 (16), 16973–16988.
Karim, A., Bjorlin, S., Piprek, J., Bowers, J., 2000. Long-wavelength vertical-cavity lasers
and amplifiers. IEEE J. Sel. Top. Quantum Electron. 6 (6), 1244–1253.
Kim, J., Chrostowski, L., Bisaillon, E., Plant, D., 2007. DBR, sub-wavelength grating, and
photonic crystal slab Fabry–Perot cavity design using phase analysis by FDTD. Opt.
Express 15 (16), 10330–10339.
Kurosaka, Y., Iwahashi, S., Sakai, K., Miyai, E., Kunishi, W., Ohnishi, D., Noda, S., 2009.
Band structure observation of 2D photonic crystal with various V-shaped air-hole
arrangements. IEICE Electron. Exp. 6, 966–971.
Kwon, S.H., Kim, S.H., Kim, S.K., Lee, Y.H., 2004. Small, low-loss heterogeneous
photonic bandedge laser. Opt. Express 12 (12), 5356–5361.
Lee, J., Zhen, B., Chua, S.-L., Qiu, W., Joannopoulos, J.D., Soljacic, M., Shapira, O., 2012.
Observation and differentiation of unique high-Q optical resonances near zero wave
vector in macroscopic photonic crystal slabs. Phys. Rev. Lett. 109, 067401.
Liang, D., Roelkens, G., Baets, R., Bowers, J., 2010. Hybrid integrated platforms for silicon
photonics. Materials 3 (3), 1782–1802.
Liang, Y., Peng, C., Sakai, K., Iwahashi, S., Noda, S., 2012. Three-dimensional coupled-
wave analysis for square-lattice photonic crystal surface emitting lasers with
transverse-electric polarization: finite-size effects. Opt. Express 20 (14), 15945–15961.
Lindberg, H., Strassner, M., Gerster, E., Bengtsson, J., Larsson, A., 2005. Thermal manage-
ment of optically pumped long-wavelength InP-based semiconductor disk lasers. IEEE J.
Sel. Top. Quantum Electron. 11, 1126–1134.
Liu, V., Fan, S., 2012. S4: a free electromagnetic solver for layered periodic structures.
Comput. Phys. Commun. 183 (10), 2233–2244.
Liu, S.-C., Zhao, D., Liu, Y., Yang, H., Sun, Y., Ma, Z., Reuterski€ old-Hedlund, C.,
Hammar, M., Zhou, W., 2017. Photonic crystal bandedge membrane lasers on silicon.
Appl. Optics 56 (31), H67–H73.
Liu, S.-C., Zhao, D., Ge, X., Reuterski€ old-Hedlund, C., Hammar, M., Fan, S., Ma, Z.,
Zhou, W., 2018. Size scaling of photonic crystal surface emitting lasers on silicon sub-
strates. IEEE Photonics J. 10 (3), 1–6.
Luk’yanchuk, B., Zheludev, N.I., Maier, S.A., Halas, N.J., Nordlander, P., Giessen, H.,
Chong, C.T., 2010. The Fano resonance in plasmonic nanostructures and metamaterials.
Nat. Mater. 9, 707–715.
Lundeberg, L.D., Boiko, D.L., Kapon, E., 2005. Coupled islands of photonic crystal hetero-
structures implemented with vertical-cavity surface-emitting lasers. Appl. Phys. Lett.
87, 241120.
Magnusson, R., Shokooh-Saremi, M., 2008. Physical basis for wideband resonant reflectors.
Opt. Express 16 (5), 3456–3462.
Martınez, L.J., Alen, B., Prieto, I., Galisteo-Lopez, J.F., Galli, M., Andreani, L.C.,
Seassal, C., Viktorovitch, P., Postigo, P.A., 2009. Two-dimensional surface emitting
photonic crystal laser with hybrid triangular-graphite structure. Opt. Express
17, 15043–15051.
Meier, M., Mekis, A., Dodabalapur, A., Timko, A., Slusher, R.E., Joannopoulos, J.D.,
Nalamasu, O., 1999. Laser action from two-dimensional distributed feedback in
photonic crystals. Appl. Phys. Lett. 74, 7–9.
Mi, Z., Bhattacharya, P., Yang, J., Pipe, K., 2005. Room-temperature self-organised
In0.5Ga0.5As quantum dot laser on silicon. Electron. Lett. 41 (13), 742–744.
Miyai, E., Sakai, K., Okano, T., Kunishi, W., Ohnishi, D., Noda, S., 2006. Photonics: lasers
producing tailored beams. Nature 441 (7096), 946.
Monat, C., Seassal, C., Letarte, X., Regreny, P., Gendry, M., Romeo, P.R.,
Viktorovitch, P., D’Yerville, M.L.V., Cassagne, D., Albert, J.P., Jalaguier, E.,
Pocas, S., Aspar, B., 2003. Two-dimensional hexagonal-shaped microcavities formed
in a two-dimensional photonic crystal on an InP membrane. J. Appl. Phys. 93 (1), 23.
Noda, S., 2010. Photonic crystal lasers—ultimate nanolasers and broad-area coherent lasers
[invited]. JOSA B 27 (11), B1–B8.
Noda, S., Yokoyama, M., Imada, M., Chutinan, A., Mochizuki, M., 2001. Polarization
mode control of two-dimensional photonic crystal laser by unit cell structure design.
Science 293 (5532), 1123–1125.
Noda, S., Fujita, M., Asano, T., 2007. Spontaneous-emission control by photonic crystals
and nanocavities. Nat. Photonics 1 (8), 449–458.
Oskooi, A.F., Roundy, D., Ibanescu, M., Bermel, P., Joannopoulos, J., Johnson, S.G., 2010.
MEEP: a flexible free-software package for electromagnetic simulations by the FDTD
method. Comput. Phys. Commun. 181 (3), 687–702.
Painter, O., 1999. Two-dimensional photonic band-gap defect mode. Science 284 (1819),
1999.
Painter, O., Lee, R.K., Scherer, A., Yariv, A., O’Brien, J.D., Dapkus, P.D., Kim, I., 1999.
Two-dimensional photonic band-gap defect mode laser. Science 284 (5421), 1819–1821.
224 Weidong Zhou
Park, H.-G., Kim, S.-H., Kwon, S.-H., Ju, Y.-G., Yang, J.-K., Baek, J.-H., Kim, S.-B.,
Lee, Y.-H., 2004. Electrically driven single-cell photonic crystal laser. Science
305 (5689), 1444–1447.
Park, H., Fang, A.W., Cohen, O., Jones, R., Paniccia, M.J., Bowers, J.E., 2006. Design and
fabrication of optically pumped hybrid silicon-AlGaInAs evanescent lasers. IEEE J. Sel.
Top. Quantum Electron. 12 (6), 1657–1663.
Qiang, Z., Yang, H., Chuwongin, S., Zhao, D., Ma, Z., Zhou, W., 2010. Design of Fano
broadband reflectors on SOI. IEEE Photon. Technol. Lett. 22 (15), 1108–1110.
Rapp, S., Piprek, J., Streubel, K., Andre, J., Wallin, J., 1997. Temperature sensitivity of
1.54-m vertical-cavity lasers with an InP-based Bragg reflector. IEEE J. Quantum
Electron. 33 (10), 1839.
Rapp, S., Salomonsson, F., Streubel, K., Mogg, S., Wennekes, F., Bentell, J., Hammar, M.,
1999. All-epitaxial single-fused 1.55μm vertical cavity laser based on an InP Bragg
reflector. Jpn. J. Appl. Phys 38, 1261–1264.
Roelkens, G., Liu, L., Liang, D., Jones, R., Fang, A., Koch, B., Bowers, J., 2010. III–V/
silicon photonics for on-chip and intra-chip optical interconnects. Laser Photonics
Rev. 4 (6), 751–779.
Sakai, K., Miyai, E., Sakaguchi, T., Ohnishi, D., Okano, T., Noda, S., 2005. Lasing
band-edge identification for a surface-emitting photonic crystal laser. IEEE J. Sel. Areas
Commun. 23 (7), 1335–1340.
Sakai, K., Miyai, E., Noda, S., 2006. Coupled-wave model for square-lattice two-
dimensional photonic crystal with transverse-electric-like mode. Appl. Phys. Lett.
89, 021101.
Sato, Y., Tanaka, Y., Upham, J., Takahashi, Y., Asano, T., Noda, S., 2012. Strong coupling
between distant photonic nanocavities and its dynamic control. Nat. Photonics 6, 56–61.
Sauvan, C., Hugonin, J., Lalanne, P., 2009. Difference between penetration and damping
lengths in photonic crystal mirrors. Appl. Phys. Lett. 95, 211101.
Sciancalepore, C., Bakir, B.B., Letartre, X., Harduin, J., Olivier, N., Seassal, C., Fedeli, J.,
Viktorovitch, P., 2012. CMOS-compatible ultra-compact 1.55-um emitting VCSELs
using double photonic crystal mirrors. IEEE Photon. Technol. Lett. 24 (6), 455–457.
Song, D.S., Kim, S.H., Park, H.G., Kim, C.K., Lee, Y.H., 2002. Single-fundamental-mode
photonic-crystal vertical-cavity surface-emitting lasers. Appl. Phys. Lett. 80 (21),
3901–3903.
Stankovic, S., Jones, R., Sysak, M.N., Heck, J.M., Roelkens, G., Van Thourhout, D., 2011.
1310-nm hybrid III–V/Si Fabry–Perot laser based on adhesive bonding. IEEE Photon.
Technol. Lett. 23 (23), 1781–1783.
Susa, N., 2001. Threshold gain and gain-enhancement due to distributed-feedback in
two-dimensional photonic-crystal lasers. J. Appl. Phys. 89 (2), 815–823.
Van Campenhout, J., Rojo Romeo, P., Regreny, P., Seassal, C., Van Thourhout, D.,
Verstuyft, S., Di Cioccio, L., Fedeli, J.M., Lagahe, C., Baets, R., 2007. Electrically
pumped InP-based microdisk lasers integrated with a nanophotonic silicon-on-insulator
waveguide circuit. Opt. Express 15 (11), 6744–6749.
Wu, S., Buckley, S., Schaibley, J.R., Feng, L., Yan, J., Mandrus, D.G., Hatami, F., Yao, W.,
Vuckovic, J., Majumdar, A., Xu, X., 2015. Monolayer semiconductor nanocavity lasers
with ultralow thresholds. Nature 520, 69–72. https://doi.org/10.1038/nature14290(.
Yablonovitch, E., 1987. Inhibited spontaneous emission in solid-state physics and electron-
ics. Phys. Rev. Lett. 58 (20), 2059–2062.
Yang, H., Chuwongin, S., Qiang, Z., Chen, L., Pang, H., Ma, Z., Zhou, W., 2009. Res-
onance control of membrane reflectors with effective index engineering. Appl. Phys.
Lett. 95, 023110.
Yang, H., Zhao, D., Chuwongin, S., Seo, J.-H., Yang, W., Shuai, Y., Berggren, J.,
Hammar, M., Ma, Z., Zhou, W., 2012. Transfer-printed stacked nanomembrane lasers
on silicon. Nat. Photonics 6 (9), 615–620.
Yokouchi, N., Danner, A., Choquette, K.D., 2003. Two-dimensional photonic crystal con-
fined vertical-cavity surface-emitting lasers. IEEE J. Sel. Top. Quantum Electron. 9 (5),
1439–1445.
Yoshie, T., Scherer, A., Hendrickson, J., Khitrova, G., Gibbs, H., Rupper, G., Ell, C.,
Shchekin, O., Deppe, D., 2004. Vacuum Rabi splitting with a single quantum dot in
a photonic crystal nanocavity. Nature 432 (7014), 200–203.
Zhao, D., Ma, Z., Zhou, W., 2010. Field penetrations in photonic crystal Fano reflectors.
Opt. Express 18 (13), 14152–14158.
Zhao, D., Yang, H., Chuwongin, S., Seo, J., Ma, Z., Zhou, W., 2012. Design of photonic
crystal membrane-reflector-based VCSELs. IEEE Photonics J. 4 (6), 2169–2175.
Zhao, D., Liu, S., Yang, H., Ma, Z., Reuterski€ old-Hedlund, C., Hammar, M., Zhou, W.,
2016. Printed large-area single-mode photonic crystal bandedge surface-emitting lasers
on silicon. Sci. Rep. 6.
Zhou, W.D., Sabarinathan, J., Kochman, B., Berg, E., Qasaimeh, O., Pang, S.,
Bhattacharya, P., 2000. Electrically injected single-defect photonic bandgap surface-
emitting laser at room temperature. Electron. Lett. 36 (18), 1541–1542.
Zhou, W., Sabarinathan, J., Bhattacharya, P., Kochman, B., Berg, E.W., Yu, P.-C.,
Pang, S.W., 2001. Characteristics of a photonic bandgap single defect microcavity
electroluminescent device. IEEE J. Quantum Electron. 37 (9), 1153–1160.
Zhou, W., Zhao, D., Shuai, Y.-C., Yang, H., Chuwongin, S., Chadha, A., Seo, J.-H.,
Wang, K.X., Liu, V., Ma, Z., 2014. Progress in 2D photonic crystal Fano resonance
Index
Note: Page numbers followed by “f ” indicate figures, and “t” indicate tables.
A C
Amplitude, 6, 52, 95–97, 132, 142–143, Carcinoembryonic antigen (CEA), 120
151–152, 154 Chirped GMR approach, 141–142, 141f,
Angle-integrated light trapping absorption 145
enhancement factor, 49–50, 65–67, Common-path interferometry, 145
66f, 69–70, 70f, 72–76, 72f, 74–75f Controlled lattice displacement,
Artificial magnetic permeability, 19 165–167
Artificial structures, 15–16, 33 Coplanar waveguide (CPW), 173
Asymmetric photonic crystal slab, 6–7 CST Microwave Studio software, 20–21
Cylindrical Lorenz–Mie coefficients, 24
B
Bandgaps vs. rod permittivity, 27–28
Band structure, 2D photonic crystal, 63f D
density of TM states, 64, 64f Density of states (DOS), light trapping,
MPB package, 62–63 46–47, 51, 53–54
overlap factor of TM mode, 64–65, 64f elevated local DOS, 46
projected TM band structure, 63–64, 63f 3D photonic crystals, 74–76
Barcoding, 141–142 2D photonic crystals, 62–63
Bimodal waveguide sensor, 121, 128, absorption enhancement factor,
131–132, 140t 65–67
Biosensing and imaging, GMR sensor. density of TM states, 64, 64f
See Guided mode resonance (GMR) long wavelength limit, 68–70
sensor van Hove singularities, 67–68
Bloch theorem, 16–17 Dielectric metamaterials, 31–32
Bragg bandgaps, 16–17, 31–32, 36–37 benefit, 41–42
Bragg gaps, 17–18, 21–23, 26–27 one-dimensional structures, 18
Mie gaps, 32–33 with near-zero effective parameters,
resonant Bragg gap broadening, 23–24 33–35, 34–35f
swap, 19 with negative magnetic response, 14
symmetric, 32–33 Dielectric permittivity, 14, 27,
width, 18 37–38
Bragg mirror theory, 132 Dirichlet-to-Neumann (DTN) method,
Bragg resonances, 13–14, 20–23, 27, 31 64–65
Bragg scattering, 31 Double-layer PhC Fano resonance filters
on periodic structure, 16–17, 30–31 amplitude transmission coefficient, 154
periodicity, 31–32 controlled lattice displacement,
wavelength, 18 165–167
Brillouin zone, 16–17 coupled bright resonances, dark resonance
Bragg gap, 17–18 from, 154, 156f
minimum frequency, 18 coupled dark resonances, dark resonance
of square lattice, 20 from, 154, 155f
surface, 30–31 design parameters and Q’s, 160, 161t
Buried tunnel junction (BTJ), 218–219, 218f lattice parameters, 157–158, 157f
227
228 Index
Double-layer PhC Fano resonance filters SOI substrate, 162–164, 163–164f

(Continued ) suspended double-layer PCS filters,
quartz substrates 160
measured and simulated transmission transmission behavior, EIT, 151f, 154
spectra, 164, 164f optomechanical cavity
poly-Si deposition on, 162–163, 163f coupled photonic crystal slabs, optical
simulated transmission spectra, 160, forces in, 178–184
161–162f, 162 MOMES applications, 184–185
SOI substrate nanoscale cavities, optical forces in,
measured and simulated reflection 177–178
spectra, 164, 164f PBEL, 196
poly-Si deposition on, 162–163, 163f single-layer filters (see Single-layer PCS
suspended double-layer PCS filters, 160 Fano resonance filters)
3D close-up view of, 157f 2D photonic crystal slabs (see Two-
transmission behavior, EIT, 151f, 154 dimensional photonic crystal slabs
Dry etching process, 166–167, 207, (2D PCSs), Fano resonance)
213–214 ultracompact MR-VCSELs, 195–197,
197f
E Farfield approach, 142
EBL. See Electron beam lithography (EBL) Figure of merit (FOM), GMR sensor, 140t
Electrical coupling, 168–170 biochemical factors, 120–121
Electrically pumped PCSELs, 218–219 electronic factors, 119–120
Electric field intensity, 34, 35f limit of detection, 118–119
Electromagnetically induced transparency photonic factors, 119
(EIT), 154 Finite-difference time domain (FDTD), 4,
Electromagnetic wave, 13, 15–16, 41–42, 99 5f, 152, 196–197, 200
Electron beam lithography (EBL), 166–167, First-order Bragg diffraction, 201–202
172–173, 207, 213–214 Fourier modal method, 65, 203
Electronic transition, 35–36 Fourier optics, 94
Electro-optical (EO) modulation, 168–177 Frequency-dependent phase transition,
Extraordinary optical transmission (EOT), 15–16
129–130, 137–139 Full width at half maximum (FWHM), 212
F G
Fano resonance, in photonic crystal slabs Gaussian beam, 29–30, 29f, 34, 34f, 40, 41f
coupled double-layer filters, 152–153 Ge-Sb-Te (GST) alloy, 33
amplitude transmission coefficient, 154 Guided mode resonance (GMR) sensor, 2
controlled lattice displacement, advantages, 128, 132
165–167 vs. bimodal waveguides, 121, 128,
coupled bright resonances, dark 131–132, 140t
resonance from, 154, 156f Bragg mirror theory, 132
coupled dark resonances, dark chirped GMR, 141–142, 141f
resonance from, 154, 155f figures of merit
design parameters and Q’s, 160, 161t biochemical factors, 120–121
lattice parameters, 157–158, 157f electronic factors, 119–120
quartz substrates, 162–164, 163–164f limit of detection, 118–119
simulated transmission spectra, 160, photonic factors, 119
161–162f, 162 history, 117–118
Index 229
vs. microrings, 121, 128, 130–131, 140t incident and output field profile,
vs. nanohole arrays, 129–130, 140t Fourier transform of, 104–105
operation of isotropic band-pass filter, 108–111,
bulk sensitivity, 124, 127t 110f
optical signature, 121 isotropic band-reject filter, 108, 109f
polarization dependence, 122–123 isotropic high-pass filter, 105, 106f
Q-factor, 123–124 isotropic low-pass filter, 105–108, 107f
surface sensitivity, 124–127, 127t photonic crystal slab, guided resonance
phase-sensitive measurement dielectric constant, 97–98, 98f
phase noise, 142–143 direct transmission coefficient, 98–99
realization of, 143–145 filter transfer function, 99
reflectance and phase response, geometry parameters, 97–100, 98f
143–144, 144f Hamiltonian, 99
resonant hyperspectral imaging, 134 isotropic band-pass filter, 104
biofilms, early detection of, 134 isotropic band-reject filter, 104
hyperspectral cube, 134 isotropic high-pass filter, 103–104
multiparameter imaging, 137 isotropic low-pass filter, 103–104
plasmonic nanohole arrays, 137–139 nearly isotropic photonic band
sensing and imaging, 135–136 structure, 100–103, 101f
scanning illumination wavelength, single-band excitation effect, 100–103
133–134 square lattice of air holes, 97–99, 98f
spatial resolution dependence transmission and reflection coefficients,
on orientation, 133, 133f 102–103
refractive index contrast, 133 uniform dielectric slab, 97–99, 98f,
vs. SPR sensor, 128–129, 140t 102–103
surface affinity sensor, operation of, 116,
117f L
2D pixelated metasurface, 141–142 Lambertian limit, 46
Lateral cavity size scaling, 204–206
H Lattice constant vs. wavelength, 38–39
Hankel function, 19–20 Light-emitting devices, 191
Harmonic amplitude, 19–20 Light trapping, 45–47
Hermitian eigenvalue problem, 37–38 in bulk structure, 54–55
High-index-contrast gratings (HCGs), 2 cell efficiency, 45–46
High-index dielectrics meta-atoms, 14 conventional limits, in 2D and 3D
Hyperspectral imaging, GMR sensor. structures, 50, 50t
See Resonant hyperspectral imaging, in periodic structures, 55–56, 55f
GMR sensor in plasmonic structures, 60–61
ray optics theory, 47–50
I in thin films, 57–61
Image processing, 93–94. See also Optical in 3D photonic crystals
image processing “absorption calculation” curve, 76
Infrared absorbtion line analysis, 34 angle-integrated absorption
Interferometry, 142–145 enhancement factors, 73–76, 75f, 77f
Ion beam-assisted laser interference “theoretical bound” curve, 76
lithography, 97–98 vs. 2D photonic crystals, 74–75
Isotropic filters woodpile structure, 75, 76f
numerical demonstration in 2D photonic crystals
230 Index
Light trapping (Continued ) 2D PCS-based MRs, 196–197

absorption enhancement factor and characterization, 209–211
comparison to DOS, 65–67 1D PCS/grating mirrors, 208–209
band structure, 62–65 MEMS. See Microelectromechanical
long wavelength limit, 68–70 systems (MEMS)
mode coupling, 71–73 Metacrystal, 25–27
square lattice of dielectric rods in air, Metamaterials
61–62, 61f definitions, 14
van Hove singularity in DOS, 67–68 dielectric metamaterials (see Dielectric
ultrathin c-Si solar cells metamaterials)
AM 1.5 solar irradiance spectrum and electric response, 23–24
dielectric constant, 77–78, 78f high-index dielectrics meta-atoms, 14
double-sided grating design, 77–78 Ohmic losses, 14
3D c-Si thin film structures in air, 78, phase transition, 19, 20f, 22–23
79f vs. photonic crystals
Yablonovitch limit, absorption band diagram, 16–18
enhancement at, 47, 79–85 Bloch theorem, 16–17
wave-optic theory, 50–54 Bragg bandgap, 16–17
Limit of blank (LOB), 118 Brillouin zone, 16–17
Limit of detection (LOD), 118 continuous phase transitions, 15–16
biochemical factors, 120–121 near-zero refraction index, 18
biological concentration, 119 polariton-like feature, 17–18
electronic factors, 120 regimes, 16–17, 16f
LOB, 118 resonance response, 18
optical performance, 118 thermodynamic phase transitions, 15
refractive index shift, 118 sparse high-index dielectric rods, 24
Long wavelength limit, 68–70, 74–75 split-ring resonators, 14
Low-index dielectric periodic structure, in visible frequency range
14–15 crystalline lattice, 37
Low-loss dielectric metamaterials, 28 Gaussian beam refraction, 40, 41f
Low pressure chemical vapor deposition lossless and lossy structures, phase
(LPCVD) poly-si deposition process, criteria, 36–37, 37f
162–163, 172–173 metamaterial prism, 40
periodic silicon nanorod structure,
M phase diagram of, 39–40, 39f
Magnetic Mie resonance, 14, 19 photonic crystal–metamaterial phase
Magnetic permeability, 14, 27–28 diagrams, 37–39, 38f
Membrane reflector vertical-cavity surface- silicon-based structures, 35–36
emitting lasers (MR-VCSELs), 190, silicon permittivity, 39–40
192–194t, 195 TE-polarized Gaussian beam, 40
cavity design, 190 Metastable amorphous phase, 33
energy and phase penetration Microelectromechanical systems (MEMS),
properties, 198–199 184–185, 185f
Fano resonance MR design and modal Microring resonator sensor, 121, 128,
coupling, 196–197, 197f 130–131, 140t
Si-based 2D PCS MRs, 198 Mie resonance, 31–32
Si-MR/QW/Si-MR (MR-VCSEL) and Bragg resonance, 21–22
cavity, 200 of dielectric rods, 30–31
Index 231
frequency, 19, 26–27 isotropic high-pass filter, 105, 106f

low-index, 22–23 isotropic low-pass filter, 105–108, 107f
polariton-like feature, 21–22 isotropic filters, photonic crystal slab
in silicon nanoparticles, 35–36 structure
spectral shift, 27 dielectric constant, 97–98, 98f
for TE polarization, 14, 19 direct transmission coefficient, 98–99
for TM polarization, 19 filter transfer function, 99
Mie scattering geometry parameters, 97–100, 98f
efficiency, 19–20, 21f Hamiltonian, 99
spectra on isolated dielectric rod, 19–20 isotropic band-pass filter, 104
wavelengths, 20f isotropic band-reject filter, 104
MIT Photonic Band (MPB), 62–63, 76 isotropic high-pass filter, 103–104
Mode coupling, 71–73, 86 isotropic low-pass filter, 103–104
Multidirectional Bragg diffraction, 201–202 nearly isotropic photonic band
structure, 100–103, 101f
N single-band excitation effect, 100–103
Nanohole array method, 129–130, 140t square lattice of air holes, 97–99, 98f
Nanomembrane transfer printing transmission and reflection coefficients,
technique, 165 102–103
Nano-photonic light trapping, 57–61, 58f uniform dielectric slab, 97–99, 98f,
Near-zero effective parameter, 33–35 102–103
Nomarski differential interference contrast Optical signature, 121
method, 145 Optomechanical cavity, Fano resonance
Non-Hermitian eigenvalue problem, 37–38 PCSs
Nonleaky waveguide sensor, 121 MOMES applications, 184–185
Nonzero dielectric contrast, 16–17 optical forces in
coupled photonic crystal slabs, 178–184
O nanoscale cavities, 177–178
Ohmic losses, 14
On-chip photonic crystal surface-emitting P
lasers. See Photonic crystal surface- PBDLs. See Photonic bandgap defect-mode
emitting lasers (PCSELs) lasers (PBDLs)
Optical biosensor, 132. See also Guided PBELs. See Photonic band edge surface-
mode resonance (GMR) sensor emitting lasers (PBELs)
Optical gratings, 2 PBG. See Photonic bandgap (PBG)
Optical image processing, 93–94 PBS. See Polarizing beam splitter (PBS)
filtering, in wavevector domain PCSEL. See Photonic crystal surface-
filter transfer function, 95 emitting lasers (PCSELs)
guided resonances, second-order PECVD. See Plasma-enhanced chemical
differentiation, 95–97, 96f vapor deposited (PECVD)
Fourier optics, 94 Perfect electric conductor (PEC), 57–59
isotropic filters, numerical Perfectly matched layer (PML), 7–8
demonstration of Periodic boundary condition (PBC), 7–8
incident and output field profile, Periodic dielectric photonic structure,
Fourier transform of, 104–105 27–33
isotropic band-pass filter, 108–111, Permeability, 14, 19, 27–28, 40
110f Permittivity real-to-imaginary parts ratio,
isotropic band-reject filter, 108, 109f 37–38
232 Index
Phase contrast microscopy, 135 on-chip photonic crystal surface-

Phase determination procedure, 36–37 emitting lasers (see Photonic
Phase diagram, 14–15, 27–34, 36–42 crystal surface-emitting lasers
Phase-sensitive GMR measurement, (PCSELs))
142–145 periodic dielectric photonic structures,
Phase-shifted Bragg grating, 111 phase diagram of
Phase transition terminology, 15, 41–42 bandgaps vs. rod permittivity, 27–28
Photolithography, 172–173 Brillouin zone surface, 31
Photonic band diagram, 16–17, 19–22, 21f isofrequency contours, change in,
Photonic band edge surface-emitting 30–31, 30f
lasers (PBELs), 191, 192–194t, magnetic field pattern, change in,
195–196 29–30, 29f
Photonic bandgap defect-mode lasers plane wave transmission, 31–33, 32f
(PBDLs), 191, 192–194t, 196 TM+TE photonic phase diagrams,
Photonic bandgap (PBG), 2–3, 24, 191 27–28, 28f
Photonic crystals (PhCs) phase, 21–22, 30–31
biosensing and imaging, GMR sensor photonic phase transitions, experimental
(see Guided mode resonance (GMR) observation
sensor) dielectric index, 24–25
dielectric rods, square lattice of experimental setup, 25, 25f
bandgap diagram, 22–23, 22f lattice spacing, 27
Bragg gap, 22–23 metacrystal, 25–27
Bragg resonances, 21–22 thermodynamic phase transitions, 27
Brillouin zone, 20 transmission spectra for square lattice of
effective dielectric index, 24 circular rods, 26–27, 26f
Mie resonances, 19 water permittivity, 27
Mie scattering efficiency, 20, 21f resonant photonic crystal, 22–23
photonic band diagrams, 20–22, 21f spatially extended Bragg resonances, 13
photonic crystals to metamaterials type, 21–22
transition, 19, 20f value, 29–30
sparse high-index dielectric rods, 24 Photonic crystal slabs (PCSs), 191
TE polarization, 19 Fano resonance (see Fano resonance, in
TM polarization, 19, 23–24, 23f photonic crystal slabs)
transmission spectra, 20–21 guided resonance, external radiation, 3–4
light trapping (see Light trapping) optical image processing (see Optical
low-index dielectric periodic image processing)
structures, 14–15 Photonic crystal surface-emitting lasers
vs. metamaterials (PCSELs), 190
band diagram, 16–18 bulk Si substrate, PC bandedge membrane
Bloch theorem, 16–17 laser on
Bragg bandgap, 16–17 emission spectrum of laser above
Brillouin zone, 16–17 threshold, 214–215, 215f
continuous phase transitions, 15–16 field distribution of cavity mode,
near-zero refraction index, 18 213–214, 214f
polariton-like feature, 17–18 lasing intensity and linewidth vs. input
regimes, 16–17, 16f power density, 214–215, 215f
resonance response, 18 refractive index, 213–214
thermodynamic phase transitions, 15 schematic of, 213–214, 214f
Index 233
simulated field distributions, 213–214, R

214f Ray optics light trapping theory, 47–50
thermal performance analysis, 215–218 RCWA. See Rigorous coupled wave
characterization, 212 analysis (RCWA)
electrically pumped PCSEL Reactive ion etching (RIE), 84–85
characterization, 218–219 Refractive index, 198–199, 203, 213–214
membrane reflector VCSEL (see Membrane Resonant frequency, 14, 17–18, 52–53, 95,
reflector vertical-cavity surface- 96f
emitting lasers (MR-VCSELs)) Resonant hyperspectral imaging, GMR
PBELs, 191, 192–194t, 195–196 sensor, 134
photonic crystal confined VCSELs, 191, biofilms, early detection of, 134
192–194t hyperspectral cube, 134
photonic crystal DFBs/PBDLs, 191, multiparameter imaging, 137
192–194t, 196 plasmonic nanohole arrays, 137–139
SOI substrate, PC bandedge membrane sensing and imaging, 135–136
laser on Rigorous coupled wave analysis (RCWA),
band diagram for, 202f, 203 7, 79, 168, 196–197
effective refractive index, 203
field distribution of cavity mode, 202f,
204 S
lateral cavity size scaling, 204–206 Scattering wave, 19–20
printed membrane laser cavities, Second-order Bragg diffraction, 202–203
206–208 Side-mode suppression ratio (SMSR), 208,
quality factor, 204 212
schematic of, 202–203, 202f Silicon-based metamaterial, 40
simulated field distributions, 202f, 204 Silicon-on-insulator (SOI) substrates,
thermal performance analysis, 215, 166–167, 189–190
216t, 217–218, 217f double-layer PhC Fano resonance filters
2D PC cavities, 201–202 measured and simulated reflection
Photonic integrated circuits (PICs), 2–3 spectra, 164, 164f
Photonic phase transition, 24–33, 41–42 poly-Si deposition on, 162–163, 163f
Plane wave expansion (PWE), 9 electrically pumped PCSELs, 218–219
Plasma dispersion effect, 168–170 MR designs, 198
Plasma-enhanced chemical vapor deposited PC bandedge membrane laser
(PECVD), 162–163 band diagram for, 202f, 203
Plasmonic nanohole array, 137–139 effective refractive index, 203
Polarizing beam splitter (PBS), 173–175 field distribution of cavity mode, 202f,
Polydimethylsiloxane (PDMS) transfer 204
printing technique, 157–158, lateral cavity size scaling, 204–206
165–167 printed membrane laser cavities,
Printed membrane laser cavities, 206–208 206–208
Prostate specific antigen (PSA) sensor, 119 quality factor, 204
schematic of, 202–203, 202f
Q simulated field distributions,
Quality factor, 116, 119–120, 123–124, 126, 202f, 204
132, 137, 142–144, 152–154, thermal performance analysis, 215,
168–172, 178–179, 191–196, 200, 216t, 217–218, 217f
204 Silicon permittivity, 38–40
234 Index
Single-layer PCS Fano resonance filters angle-integrated absorption enhancement

all-pass transmission/flattop reflection factors, 73–76, 75f, 77f
filter, 150–152, 152f “theoretical bound” curve, 76
angle-dependent Fano resonant mode vs. 2D photonic crystals, 74–75
dispersion properties, 159–160, 159f woodpile structure, 75, 76f
angle-independent filters, 160 Transverse electric (TE), 2–3, 3f, 19, 203,
design parameters and Q’s, 160, 161t 218–219
double-pole filters, 150, 151f Transverse-magnetic (TM), 2–3, 3f, 19, 203,
dry etching process, 166–167 218–219
electron-beam lithography, 166–167 Transverse magnetic (TM) band structure,
FDTD simulations, 152 of 2D photonic crystal, 62–63
flexible substrates, 157–158, 157f density of states, 64, 64f
glass substrate, 157–158, 157–158f overlap factors, 64–65, 64f
intensity transmission, 152, 153f projected TM band structure, 63–64, 63f
phase delay spectra, 152, 153f Two-dimensional (2D) photonic crystals,
simulated transmission spectra, 160, 161f light trapping in
single-pole filters, 149–150, 150f absorption enhancement factor and
Single-pass absorption spectrum, 79–80, comparison to DOS, 65–67
81f, 83 band structure, 63f
Slot-waveguide effect, 59–60 density of TM states, 64, 64f
SOI. See Silicon-on-insulator (SOI) MPB package, 62–63
substrates overlap factor of TM mode, 64–65, 64f
Spectrometer, 133–134, 139–141, 145 projected TM band structure, 63–64,
Split-ring resonators, 14 63f
SPR. See Surface plasmon resonance (SPR) long wavelength limit, 68–70
sensor mode coupling, 71–73
Stanford stratified structure solver (S4) square lattice of dielectric rods in air,
software package, 7, 65, 168, 203 61–62, 61f
Surface affinity sensor van Hove singularities in DOS, 67–68
biomedical applications, 116 Two-dimensional photonic crystal slabs (2D
GMR (see Guided mode resonance PCSs), 94
(GMR) sensor) Fano resonance
Surface-emitting lasers. See Photonic crystal characteristics, 5–9
surface-emitting lasers (PCSELs) energy conservation, 4
Surface plasmon resonance (SPR) sensor, guided resonances, 3–4
119, 128–129, 140t, 145 intensity transmission spectrum,
4–5, 5f
T photonic crystal properties and
TE. See Transverse electric (TE) applications, 2–3, 3f
Temporal coupled mode theory (TCMT), photon transport, direct/background
5–6, 51–52, 149–150 pathway, 4
Thermal performance analysis, PCSELs, square lattice of air holes, 4, 5f
215–218 triangular lattice PCS, TE/TM
Three-dimensional finite-difference time bandgap, 2–3, 3f
domain (3D FDTD), 7–9, 8f MR-VCSELs, 195–197
Three-dimensional (3D) photonic crystals, energy and phase penetration
light trapping in properties, 198–199, 200f
“absorption calculation” curve, 76 Si-based 2D PCS MRs, 198
Index 235
U membrane reflector VCSEL

Ultrathin crystalline silicon (c-Si) solar cells (see Membrane reflector vertical-
AM 1.5 solar irradiance spectrum and cavity surface-emitting lasers
dielectric constant, 77–78, 78f (MR-VCSELs))
double-sided grating design, 77–78 photonic crystal confined VCSELs, 191,
3D c-Si thin film structures in air, 78, 79f 192–194t
Yablonovitch limit, absorption short distance data communication
enhancement at, 47, 79–85 application, 190–191
W
V Wafer bonding technique, 208–209
van Hove singularity, 67–71, 73–76 Wave optics light trapping theory, 50–54
Vascular endothelial growth factor (VEGF), Wax protection process, 208
130 Wire bonding process, 218–219
Vertical-cavity surface-emitting laser Wood anomaly, 1–2, 117–118
(VCSEL), 191, 192–194t
broadband single-layer membrane Y
reflectors, 189–190 Yablonovitch limit, 46–50, 53–54, 67,
challenges, 190–191 79–85, 81f, 84f

Series Editors: Chennupati Jagadish

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Series Editors: Chennupati Jagadish

Uploaded by

Copyright:

Available Formats

SERIES EDITORS

First edition 2019

© 2019 Elsevier Inc. All rights reserved

For information on all Academic Press publications

Publisher: Zoe Kruze

Fano resonance, with its characteristics asymmetric lineshape, differs signifi-

Chapter 4 discusses the implementation of several isotropic image filters

Fano resonance principles

Semiconductors and Semimetals, Volume 100 # 2019 Elsevier Inc. 1

by many experts in the field. Miroshnichenko et al. (2010) presented a detailed

2. Principles of Fano resonance in photonic crystal slabs

0.6 SPP NRI

fabrication of devices using current semiconductor technology than three-

2.2 The presence of Fano resonance in photonic crystal slabs

is the presence of guided resonance in the structures (Fan and Joannopoulos,

In the vicinity of the Fano resonance, the intensity transmission in

2.3 Characteristics of Fano resonance in photonic crystal slabs

r 2 ðω ω0 Þ2 + t2 ð1=τÞ2 2rtðω ω0 Þð1=τÞ

which agrees with Eq. (2).

0.6 (ii) Medium Q

A On resonance (l1) B Off resonance

On the other hand, light at other spectra locations (off-resonance) can

Transition from photonic crystals

Semiconductors and Semimetals, Volume 100 # 2019 Elsevier Inc. 13

At the turn of the century, another class of artificial electromagnetic

presented by the well-studied photonic crystals (see, e.g., Joannopoulos

2. Photonic crystals vs metamaterials

between hyperbolic and elliptic regimes of metamaterials was considered

Wave vector Wave vector

dispersion and the structure can be described by effective material parameters

3. Dielectric rods arranged in a square lattice

decreasing lattice constant a ′ < a increasing rod permittivity

coefficients an and bn for the TE and the TM polarizations, respectively (see,

thickness of 10 rod layers demonstrate an intensive dip for both photonic

compound in infrared (Wuttig et al., 2017) the band diagram demonstrates

where k is the wave number, b0 and d0 are the cylindrical Lorenz–Mie

4. Experimental observation of the photonic phase

The sample called a metacrystal is shown in Fig. 6 composed of 5 by

Fig. 6 Experimental setup. (A) An anechoic chamber with a metacrystal composed of

5. Phase diagram of periodic dielectric photonic

Photonic crystal Metamaterial

(a) (b) (c) (d)

(e) (f) (g) (h)

The prisms in the photonic crystal regime demonstrate a striped pattern

Fig. 10 A dramatic change in isofrequency contours depending on the operational

structure to be transparent (Fig. 11B). The dielectric metamaterial (a/λ ¼ 0.45,

6. Dielectric metamaterials with near-zero effective

(a) (c) (e)

(b) (d) (f)

According to the phase diagram in Fig. 8, the dielectric metamaterials

7. Toward practical realization of metamaterials in the

to discuss the approaches to yield the metamaterial phase in visible range. We

Photonic crystal Metamaterial

silicon-based metamaterials is possible for wavelength shorter than a certain

800 Photonic crystal Photonic crystal

arrays with precise control in position, density. The reported diameter is as

l = 1500 nm l = 700 nm l = 495 nm

8. Conclusion and outlook

Light trapping in photonic

Semiconductors and Semimetals, Volume 100 # 2019 Elsevier Inc. 45

cell efficiency, because thinner cells provide better collection of photogenerated

In this chapter, we describe a statistical coupled-mode theory that

2. Ray optics light trapping theory

ω ðkÞ iγ ðkÞ ¼ ω0 iγ 0 + ða bÞjkj2 : (11)

ω ðkÞ ω0 + A jkj2 , (12)