Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Perspective

Cite This: ACS Photonics 2019, 6, 4−17 pubs.acs.org/journal/apchd5

Structural Slow Waves: Parallels between Photonic Crystals and


Plasmonic Waveguides
Philippe Lalanne,*,† Steṕ hane Coudert,‡ Guillaume Duchateau,‡ Stefan Dilhaire,§ and Kevin Vynck†

LP2N, Institut d’Optique Graduate School, CNRS, Université Bordeaux, 33400 Talence, France

Université de Bordeaux-CNRS-CEA, Centre Lasers Intenses et Applications, UMR 5107, 351 Cours de la Libération, 33405
Talence, France
§
Laboratoire Onde et Matière d’Aquitaine (LOMA), UMR 5798, CNRS-Université de Bordeaux, 33400 Talence, France
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: The slowdown of propagating electromagnetic waves by engineer-


ing waveguide structures is receiving considerable attention in nanophotonics.
Two types of structures are independently investigated, namely, diffraction-
Downloaded via UNIV OF BORDEAUX on August 29, 2019 at 09:09:29 (UTC).

limited photonic-crystal waveguides and highly confined plasmonic waveguides.


Here we propose a unified viewpoint on slow waveguide structures by analyzing
both approaches within a single theoretical framework. This allows us to clarify
the physical mechanisms underlying slownessthereby highlighting their
inherent differencesand to remove some inaccurate ideas on the topic. We
further discuss the strengths and weaknesses of plasmonic and photonic slow
waveguides, assess the technical and scientific challenges recently solved, and critically identify those that still stand in the way.
KEYWORDS: slow waveguide, slow surface plasmon, slow photonic-crystal waveguide, LDOS enhancement,
roughness-induced backscattering

■ INTRODUCTION
In general, the velocity at which a pulse of light propagates
plasmonic waveguides, e.g., metal/insulator/metal (MIM) or
insulator/metal/insulator (IMI) heterostructures, metal nano-
through a medium (group velocity) is given by vg = cng−1 = c(n + wires, or hyperbolic materials.6 We will indifferently use the
ω dn/dω)−1, where c is the velocity of light in a vacuum, ω is the terms “slow photon” or “slow photonic waveguide” for dielectric
light angular frequency, n is the refractive index of the medium, structures and consistently “slow plasmon” or “slow plasmonic
and ng denotes the group index. Theoretically, there is no waveguide” for metallic ones and “slow wave” or “slow
fundamental limit on the value to which light may be slowed waveguide” in general terms. Photonic-crystal and plasmonic
down. Very low propagation velocities (17 m/s)1 and even slow waves share many common features, starting from the field-
stopped light2 have been reported in ultracold atomic gases. enhancement property, and are studied, quite naturally, for the
These ultraslow regimes are reached when the velocity of light same applications in science and technology, e.g., light
pulses can be described fully in terms of an extremely frequency- harvesting,7 biosensing,8 nonclassical light sources,9,10 las-
dependent refractive index of the material, offering a sharp dip in ing,5,11−14 quantum information,15,16 nonlinearities and switch-
the absorption or gain spectrum in a narrow spectral region. ing,17−20 and integrated photonic circuits.21,22 Photonic crystals
Simple saturation effects can lead to such behavior, as well as are definitely more advantageous for applications such as delay
more advanced effects such as electromagnetically induced lines, buffering, or bit memory,23 whereas plasmonic structures
transparency.3,4 Slowness induced by material dispersion (dn/ are considered to be more promising for nanoscale applica-
dω) will not be discussed further in this Perspective article. tions.10,24−29
It is believed that if slow light is delivered at a micro- or In photonics, the energy is purely electromagnetic. Slowness
nanoscale on-chip, for instance in a fully integrated config- originates from waves that bounce back and forth as they
uration compatible with standard technological platforms, propagate through a periodic structure (see Figure 1a for an
science and technology will be strongly impacted. This route illustration). It immediately appears that a phase-matching
is mainly followed by relying on structural dispersion. It refers to condition should be satisfied and that the characteristic
situations in which the velocity of light pulses is modified by dimension of the structure, i.e., the period a, should be
structuring and/or assembling matter at the micro- or nanoscale. comparable to the wavelength λ, i.e., a ≈ λ/(2n). Quite
The group velocity is no longer related to the refractive index n differently, slowness in plasmonics does not necessarily require
of a medium, but to an effective index neff that depends on ω. In periodic back-reflections and can be implemented at deep-
marked contrast with material slowness, structural slowness is
accompanied by a strong electromagnetic field enhancement. Received: September 25, 2018
There are essentially two schemes for structural slowness: (i) Revised: November 22, 2018
photonic-crystal waveguides, e.g., single-row-missing photonic- Accepted: November 25, 2018
crystal waveguides or coupled optical resonators,5 and (ii) Published: November 26, 2018

© 2018 American Chemical Society 4 DOI: 10.1021/acsphotonics.8b01337


ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Figure 1. Structural slow electromagnetic waves. (a) Slowness at the microscale with photonic periodic waveguides. It is achieved for lattice periods
that are comparable to the wavelength, and the transversal confinement is limited by the diffraction. The slow-mode intensity pattern is periodic along
the direction of propagation. (b) Slowness at the nanoscale with plasmonic translation-invariant waveguides. Slow plasmons are achieved with
heterostructures having at least one transverse dimension much smaller than the wavelength. The slow-mode intensity pattern is uniform along the
direction of propagation. (Courtesy of S. I. Bozhevolnyi.)

subwavelength scales with translational-invariant wave- knowledge is very important since many physical effects in the
guides.30−34 Slow surface plasmons result from a coupling of linear and nonlinear regimes depend on how the electric field
electromagnetic fields with the free electrons of metals. As a increases with ng.
result, slow plasmons appear when the characteristic dimensions Section 4 highlights the advantages and drawbacks of
are much smaller than the wavelength, for instance with a MIM plasmonic and photonic slowness. It provides a comprehensive
dielectric gap (see Figure 1b for an illustration). Consequently, picture of their inherent forces and limitations, including the
contrary to photonic slow light that occurs in a quite narrow bandwidth, the attenuation due to field enhancement, and
frequency range with effective mode areas on the order of λ2, fabrication imperfections (roughness, absorption, etc.), which
slow plasmons are inherently broadband with deep-subwave- are known to considerably reduce the propagation distance of
length mode areas. slow modes in realistic structures.31,37 We also try to understand
The literature on structural slow waves appears to be why slow plasmons seem to promise much for reaching
disjointed between slow plasmons and slow photons. Review
ultraslow regimes and even possibly for stopping light,6 whereas
articles, wherein applications of slow waves are analyzed and
imperfections are known to ruin any hope of observing ultraslow
discussed,5,6,15,16,18,19,22,35,36 are available for each individual
scheme, but to our knowledge, there is no comprehensive transport with photonic waveguides.37,38
overview of structural slow waves in general. Our primary Section 5 addresses the issue of engineering the local density
motivation is to address this shortcoming. By treating absorption of states (LDOS) with slow waveguides on the illustrative and
loss for plasmons as a perturbation, the same theoretical important example of light emission. Slowness is indeed
approach, notably the Lorentz reciprocity theorem in its attached to field enhancement and may thus be used to boost
conjugate form, applies indifferently to each scheme. The light emission. After a rapid analysis of light emission in slow
emergence of a “unified” vision of both schemes is additionally photonic waveguides, which is nowadays well understood, we
favored by the systematic reference to a single and common focus on emission in slow plasmonic waveguides. The analysis
physical parameter to quantify the slowness, the group index ng. starts with the textbook case of a single metal/insulator (MI)
Section 2 is concerned with the study of the emblematic interface, for which we highlight the role of slow plasmons for
example of gap modes of MIM waveguides with classical quenching the emission. The case of MIMs is considered
electrodynamic models. The derivation is found in many afterward. We evidence that the large LDOS of the slow gap
textbooks on electrodynamics. Our prime interest is to look plasmon mode for tiny insulator gaps counterbalances
more specifically to what is happening for small gaps when the quenching and further discuss how this effect may be exploited
group velocity vanishes and to introduce, with a concrete for optimizing the efficiency of photon/plasmon sources.
example, many ingredients that will be discussed afterward more Section 6 finally points out some perspectives for the use of
abstractly. Although the results are well known, it is slow waveguides in devices and summarizes the work.
recommended that potential readers do not jump over this In the following, it might appear that the ohmic loss is
section. somewhat neglected. For instance, absorption is disregarded in
Section 3 is devoted to the basics of the electrodynamics of Sections 2 and 3 and is treated as a perturbation in Section 4.
slow modes. We review the two physical mechanisms underlying There is a good reason for not rooting complex permittivities
slowness and analyze their inherent differences. First, we clarify
directly from the beginning into Maxwell’s equations. This
how ultraslow transport is achieved by associating two counter-
simplifies the mathematical treatment of the next sections and
propagative power flows. It is generally admitted that the two
power flows exactly balance at vanishing speeds. For instance, allows us to use exactly the same Hermitian theoretical
for slow plasmons, owing to the negative contribution, the formalism to analyze photons and plasmons, making the
power flow in the metal is expected to balance that in the intercomparison more quantitative. We would like to emphasize
positive-permittivity dielectric cladding layers.6 We revisit this that (i) the role of ohmic losses will be discussed whenever it
widely admitted picture for photons and plasmons, showing that impacts the reasoning, (ii) the perturbation treatment of Section
the power flows are not perfectly balanced and that the slight 4 is exact for translation-invariant waveguides, and exact
difference is important, as for instance it allows perfect adiabatic numerical values will be provided for the main formula. The
tapering. Then, we derive the scaling laws with ng of the electric importance of ohmic losses, which are critical and sometimes
and magnetic fields of slow modes, showing that they are underestimated in the literature, is therefore not overlooked
considerably different for plasmons and photons. Their hereafter.
5 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Figure 2. Symmetric gap plasmon mode of MIM waveguides translationally invariant along the y axis. (a) z is the direction of propagation, and x
denotes the transverse direction. εm < 0 and εd > 0 are the dielectric permittivities of the metal and insulator. As the gap width g becomes much smaller
than the wavelength, the mode has no cutoff and the propagation constant β diverges as g−1. (b) Group index for εd = 2.25, εm = −28.0 + 1.5i, and
∂εm
∂λ
= − 84.23 + 2.63i (silver at λ = 800 nm39) in the limit of small gaps (i.e., large β’s). Black circles are numerical data obtained by computing the
ω dβ
complex-valued effective index β/c with the aperiodic rigorous coupled-wave analysis,40 and then the group index ng = β /c + ( ).

c dω

SLOW GAP PLASMONS IN electrodynamic models cease to be valid. With a fully quantum
METAL/INSULATOR/METAL WAVEGUIDES mechanical approach,41 the effects of electron spill-out on gap
In this section, we study the properties of slow gap modes guided plasmons have been analyzed, and it has been shown that the
in MIM heterostructures with tiny gaps employing a classical mode index of ultranarrow gap plasmons converges to the bulk
electrodynamic model. Our prime intention is to introduce, on a refractive index in the limit of vanishing gap, thereby rectifying
concrete example, the concepts and electromagnetic formalisms, the unphysical divergence found with classical models. These
e.g., complex group indices, antiparallel power flows, and the effects that take place at the MI interfaces, 1−2 Å from the walls,
Lorentz reciprocity theorem, which will be used further at a and introduce additional ohmic losses, are however negligible
more abstract level in the following sections. A second for the gap widths (g > 1 nm) considered in the following, as we
motivation is to have a complete analysis of an important checked with simplified nonlocal models.42
element that plays a key role in nanophotonics, e.g., for energy Solving the time-harmonic source-free Maxwell’s equations
transport though metal slits or grooves in metallic thin films, for the MIM geometry is classical, as it looks very similar to the
artificial magnetism at optical frequency, and anomalous textbook case of dielectric slab waveguides for TM polarization.
dispersion in hyperbolic media. A small additional complexity, related to the definition of the
MIM heterostructures support both transverse-electric (TE) square root function in the complex plane, should be addressed
and transverse-magnetic (TM) gap modes.30,32 TE modes may if the metal loss is not neglected, Im εm ≠ 0. We will simply
propagate over distances that may even exceed decay lengths consider the case of εm negative and real and introduce
observed for TM modes. Here, we are concerned by the absorption with perturbation theory, thereby making all the
fundamental gap-plasmon TM mode (see Figure 2), which does following quantities, e.g., the propagation constant β and the
not exhibit cutoff and offers slow propagation as the gap width group index ng, also real.
becomes much smaller than the wavelength. The magnetic field of the symmetric gap mode for TM
For ultrasmall gap widths, gap plasmons are sensitive to the polarization, propagating along the positive z direction in a MIM
electron density profile at the metal surfaces and classical waveguide with gap width g, can be written as

l
o i i i g yy i i g yyy
o
o
o
o Hy(x) = h0jjjjexpjjjγdjjjx − zzzzzz + expjjj−γdjjjx + zzzzzzzzzzexp(iβz), |x| ≤
o
o k k k {{ k k {{{
g
o
m
o
o
2 2 2
o
o i i g yy
o
o
o Hy(x) = h0(1 + exp(−γdg ))expjjj−γmjjj|x|− zzzzzzexp(iβz), |x| >
o k k 2 {{
g
n 2 (1)

1/2
ω2 ω
(
with γα = β 2 − εα c 2 ) , with α ≡ m, d. Note that γm and γd the limit of small gaps, c g ≪ 1. By assuming that γdg ≪ 1 and γd
are both positive. The magnetic field amplitude h0 is a constant ≈ γm ≈ β, we get
that will be normalized afterward. By using the tangential-field 2ε
g βg = − d
continuity equation for Ez at x = ± 2 , we obtain the tran- εm (3)
scendental dispersion equation
Equation 3 is important, as it directly shows that, as the gap-
γ γ plasmon spatial extension is shrunk by the gap size, the gap-
− m (1 + exp(−γdg )) = d (1 − exp(−γdg )) plasmon wavelength 2π/β can be as small as a few nanometers,
εm εd (2)
which is much smaller than that in any known dielectric
Effective Index. Equation 2 can be solved accurately by medium. This is the physical origin of the subwavelength
numerically computing β(ω). It is easily solved analytically in localization beyond the diffraction limit (that is, the diffraction
6 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

limit in a dielectric medium). Equation 3 also implies that both important role and can be disregarded. The net benefits are that
transverse wave vectors, γd and γm, scale as g−1. the group velocity coincides with the energy velocity, providing a
Field Profile and Mode Normalization. Always in the simpler interpretation, and that the same intuitive theoretical
limit of small gaps, the gap-plasmon field distribution takes the formalism can be used for both plasmonic and photonic slow
following asymptotic form: waveguides, enabling a direct comparison of each scheme.
General Electromagnetic Theorem for Slow Modes.
Hy(x) = 2h0 exp(iβz)
We start by deriving a relation between the power flow and the
Ex (x) = 2βh0 /(ωε0εd) exp(iβz) group velocity of slow guided modes. Throughout the
Perspective, we place no restriction on the geometry, and we
Ez(x)=iβh0 /(ωε0εd)[exp(γd(x − g /2)) − exp(− γd(x + g /2))] denote by ε0ε(x, y) the arbitrary cross-section permittivity
× exp(iβz) distribution of the plasmonic waveguide. The waveguide
(4a) permeability is equal to the vacuum permeability μ0 everywhere
g for the sake of simplicity.
for |x| ≤ 2 , and Our analysis relies on basic electromagnetic laws of z-invariant
Hy(x) = 2h0 exp(−γm(|x| − g /2)) exp(iβz) waveguides.43 We consider a guided mode at frequency ω,
[En(r), H n(r)] = [e(x , y), h(x , y)]exp i(βz − ωt ) (8)
Ex(x) = 2βh0 /(ωε0εm) exp(−γm(|x| − g /2)) exp(iβz)
x with a field distribution that is exponentially decaying away in
Ez(x) = 2iγ h0 /(ωε0εm) exp( −γm(|x| − g /2)) exp(iβz) the claddings and a propagation constant denoted by β. Note
|x | m that we drop the index “n” in e(x, y) and h(x, y) for simplicity.
(4b)
For any infinitely large transverse cross-section Σ of the
g
for |x| > We straightforwardly infer that the Poynting-vector
2
. waveguide and for Im(ε) = 0, the Lorentz reciprocity theorem
z-components in the gap and in the metals have opposite applied to waveguides43 leads to the simple relationship
directions and that the associated power flows are −4h02/ 1

ÄÅ ÉÑ
(ωε0εm) and 2h02/(ωε0εm), respectively, so that the net power ∬Σ Re(e × h*)·ẑ dx dy
ÅÅ ∂(ωε) 2 2Ñ
Ñ
ÅÅε ÑÑdx dy
2
flow is −2h02/(ωε0εm). Normalizing this net positive flow to 1,

ΣÅ
Å ÑÑ
ÅÇ ÑÖ
we find that h0 is independent of g and must be chosen so that dω
=

∬ 0
∂ω
·| e| + μ0
| h|
(9)
h0 = ( −ωε0εm /2)1/2 (5)
For photonic waveguides, the modes are Bloch modes. The
Absorption. Because we neglect absorption, Im εm = 0 and latter are pseudoperiodic functions of z, with a field distribution
eq 3 predicts a real propagation constant. The effect of still given by eq 8, except that e(x, y) and h(x, y) are now periodic
absorption can be taken into account by applying the Poynting functions of z, e.g., e(x, y, z + a) = e(x, y, z). Accounting for this
theorem and perturbation theory to a lossy waveguide (Δε = extraperiodicity, eq 9 becomes44
Im(εm) ≠ 0) with the mode profile of eq 4b. The integration of
the exponentially decaying field in the metal leads to a complex 1
∬Σ Re(e × h*)·ẑ dx dy
i y
∭ jjjjε0 ∂(∂ωε |e| + μ0 |h|2 zzzzdx dy dz
propagation constant, whose imaginary part is given by 2

k {
εd Im(εm) 1 dω ) 2
Im β = 2 g −1 =
|εm|2
(6) 4a dβ ω (10)

showing that the imaginary part of β also scales as g−1. It is where the triple integral runs over any unit cell of the periodic
interesting to note that this result can be obtained directly from waveguide delimited by two infinitely large cross-sections
eq 3 by taking a complex-valued εm. In fact, one can show that eq separated by the period a. The energetic interpretation of the
3 rigorously holds in the presence of absorption.30 classical eqs 9 and 10 evidences that the group velocity dω/dβ of
Group Index. Using the expression of the group velocity ng = the wave packet is equal to the velocity at which energy is
c∂β/∂ω, eq 3 eventually leads to the following formula for the conveyed by the mode along the waveguide.
group index: Physics of Slow Waves. There are different ways to
understand the wave deceleration in slow waveguides. For
λ 2 εd ∂εm photonic waveguides, deceleration is due to a distributed back-
ng = −
g πεm 2 ∂λ (7) reflection, and for plasmonic waveguides an equivalent could be
a negative Goos−Hänchen phase shift at the metal/insulator
Like for β, eq 7 is valid in the presence of absorption. We just interfaces. Perhaps, a most unified way, that we will use
need to plug in the complex-valued permittivity of the metal to hereafter, consists in considering that deceleration originates
obtain a complex group velocity c/ng, whose real and imaginary from antiparallel power flows. As beautifully evidenced by
parts represent the velocity and the damping of the light pulse, at coupled-wave theory,45 slow light in photonic waveguides
least when the pulse attenuation is weak. The predictive force of results from two counter-propagating waves, each carrying some
eq 7, in its complex-valued version, is tested in Figure 2b for a energy, which together constitute a quasi-stationary pattern.45,46
silver MIM waveguide at a wavelength of 800 nm.


Similarly, for plasmonic waveguides, because the permittivities
of metallic and dielectric layers are opposite, the z-component of
ELECTROMAGNETISM OF SLOW GUIDED MODES the Poynting vector is positive in dielectric layers, whereas it is
This section is devoted to the derivation of some fundamentals negative in the metallic layers.
of the electrodynamics of slow modes. For this, absorption loss In this picture, ultimate slowdowns as the group (or the
due to metal in plasmonic waveguides does not play an energy) velocity approaches zero are often seen as resulting from
7 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

a perfect balance, for photons when the quasi-stationary pattern flows of the slow mode on the right side of the taper tend to
of photonic waveguides becomes perfectly stationary or for exactly balance as vg → 0 and Φs → 0. Applying now the
plasmons when the energy flows in the metal and in the Poynting theorem to the blue boxes, Φ1 + Φf + Φs = 0, one gets
dielectrics are exactly opposite. Albeit intuitive and widespread, Φ1 = −Φf since Φs → 0: we arrive at the contradictory
see for instance the introduction of the recent review article,6 conclusion that any slowdown is automatically accompanied by
this picture however carries some flaw. radiation on the lateral sides of the taper. Perfectly (or
The power flows indeed never exactly cancel out, even as the adiabatically) stopping light would then be impossible. This is
energy velocity asymptotically tends toward zero. A convincing indeed wrong, as we know from slowdown practice for lossless
counterexample is found by considering the MIM gap mode, for 1D systems, for which there is no loss by absorption nor by
which, as shown above, the power flows in the dielectric gap and scattering in the clads, by construction.
in the metal clads are respectively Pfor = −4h02/(ωε0εm) > 0 and It is important to find out where we get the misconception of
Pback = 2h02/(ωε0εm) < 0 in the limit of small gap widths (i.e., Pfor the exactly balanced flow and to understand in detail the
= 2 and Pback = −1 for a normalized gap mode). The energy flows principle of adiabatic tapering. For that purpose, refer to eq 9 or
do not balance: no matter the smallness of vg and no matter the 10 and consider that the power f low (left-hand-side term) is
convention for the normalization, the power flow in the gap is constant (Φs = −Φf) for any cross-section of the taper, even for
twice as larger as the power flow in the metal on that example. instance for cross-section planes close to the tip apex where the
For a general and intuitive argument clearly evidencing that group velocity vanishes. The right-hand-side term of the
power flows cannot exactly cancel, imagine an adiabatic taper equation should then remain constant as vg → 0, implying that

ÅÄ ÑÉ
that progressively transforms a fast mode with a power flow Φf

∬Σ ÅÅÅÅÅε0 ∂(∂ωε |e| + μ0 |h|2 ÑÑÑÑdx dy should diverge with ng, both for
the surface integral of the electromagnetic energy density

Ç ÑÖ
incident from the left into an outgoing slow mode with a power ) 2
flow Φs; see Figure 3. Further wrongly assume that the opposite ω
photonic and plasmonic waveguides. There is no problem with
energy conservation, as the product of the vanishing velocity and
the diverging electromagnetic energy density remains constant.
The fact that the divergence of the electromagnetic density is not
observed in practice, because of absorption loss or imperfec-
tions, will be discussed in Section 4.
Scaling Laws at Small Group Velocities. Now that the
physics of slow waves is clarified, let us see how the
electromagnetic fields scale as vg vanishes. With the power
flow being nonzero even for vanishing group velocities, slow
modes can be normalized such that their power flow is 1. This is
the classical convention adopted hereafter.
Figure 3. Impossibility of adiabatically slowing down up to vanishing vg
We start by slow photons. Since the left-hand side of eq 10
(stopping light) if modes with vg = 0 have a null power flow. Indeed, equals 1, the equality of the electric and magnetic energy
perfect adiabaticity is theoretically possible as shown with studies on 1D densities directly implies that both the electric and magnetic
lossless thin-film slow light channels, for which Φ1 is null by fields scale as
construction.
|En| , |H n| ∝ ng (11)

Figure 4. Field enhancement of normalized slow waves with unitary power flows. (a) Photonic case: both the electric and magnetic field intensities
scale as the first power of ng. Since the taper is almost perfect (0.997 efficiency for this 2D simulation47), a strong ×20 intensity enhancement is
observed between the slow mode (ng = 100) and the fast one (ng ≈ 3). Note that the expected ×30 enhancement is not really implemented because the
transverse profiles of the slow and fast modes are slightly different. (b) Plasmonic MIM case: because of the additional transversal geometry change
(gap width reduction) as the group velocity is lowered, the electric field intensity (not the magnetic one) increases at a much faster rate, proportionally
to ng2. The normalized intensity profiles are computed with the aperiodic Fourier modal method.40

8 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Table 1. Scaling of the Main Properties of Structural Slow Light and Slow Plasmons (Representative Measured Values Are Given
in Square Brackets)

To derive eq 11, we have additionally neglected the spatial are indeed differences between the photonic and plasmonic
mode profile dispersion. This assumption, which is only cases.
approximately valid for intermediately small values of the Let us denote again by Pfor and Pback the positive powers
group velocity, becomes perfectly valid for lower speeds, as vg flowing forward or backward. The ratio u = Pback/Pfor (0 ≤ u < 1)
only significantly decreases in a narrow spectral range. This well- is an important parameter that quantifies the slowness. For
known scaling law,38 illustrated in Figure 4a, represents the photonic waveguide modes operating outside a photonic gap, u
building block for slow light applications based on a strong boost is strictly smaller than 1, and for a normalized Bloch mode with a
of light−matter interactions. unitary power flow (Pfor − Pback = 1), we have Pfor ≈ Pback ∝ ng ≫
For translation-invariant plasmonic waveguides, eq 9 holds, 1 with (Pback − Pfor)/Pfor ∝ vg/c, since the electromagnetic fields
and by assuming unitary power flows, we obtain scale proportionally to ng . This implies that, in the slow light

∬ ijjjkε0 ∂∂ωε y
|e|2 + μ0 |h|2 zzzdx dy = 1
regime, the quasi-standing-wave patterns of normalized

{
c 1
× photonic waveguide modes are composed of two counter-
ng 4 ω (12) propagating waves, each carrying a huge power flow that
diverges linearly with ng.
In sharp contrast with the photonic case, it is essential to The plasmonic case is markedly different. As the group
realize that one cannot obtain directly the scaling laws for e or h velocity vanishes, the plasmonic electric field scales with ng, but
from eq 12, owing to the fact that one should change at least one since the confinement also scales with 1/ng, Pback and Pfor both
transverse characteristic dimension of the z-invariant plasmonic become constant. For instance, for the MIM waveguide studied
waveguide to tune down the group velocity, thereby radically in Section 2, Pfor = 2 and Pback = 1 for a normalized gap mode. In
changing the transverse mode profile. At large ng for deep- sharp contrast with photonic waveguides, each flow becomes
subwavelength confinements, self-sustained energy oscillations independent of ng in the limit of small gaps.
between magnetic and electric energies no longer hold, and the By way of summary, Table 1 provides a recapitulation of main
oscillations are restored by considering the kinetic energy of the properties of structural slow photons and slow plasmons.


free carriers involved in the subwavelength plasmon mode.3 A
precise analysis of the scaling of e and h with ng requires knowing LIMITATIONS OF REAL WAVEGUIDES
the exact geometry, but, as shown in Section 2 on the textbook
example of MIM waveguides with asymptotically small dielectric Despite their interesting physical properties, slow waves suffer in
gaps, the magnetic energy becomes negligible and the general practice from important limitations, in particular regarding the
trend is imposed by the balance between the electron-kinetic operation bandwidth and the propagation distance. The
and electric-field energies:48 available bandwidth, that is, the information content that can
be slowed down or stored, is limited by dispersion, whereas the
|En| ∝ ng , |H n| ∝ Cte (13) propagation distance, which defines the storage time and the
interaction length, is limited by bulk-material absorption and
for normalized plasmonic modes. Noticeably, the electric-field scattering loss due to fabrication imperfections. The latter are
increase rate for small group velocities is much larger for well documented in the context of photonic-crystal wave-
plasmons than for photons. The reason comes from the guides.38,50−52 For plasmonic waveguides, there are only a few
additional transversal geometry change (gap width reduction) studies on roughness-induced attenuation, and one generally
that is accompanying plasmon slowdowns, as illustrated in assumes that the attenuation is dominantly due to absorption. It
Figure 4b. is important to have the intrinsic differences and limitations in
Based on the previously presented difference in scaling mind to optimally exploit slowness with the appropriate scheme
between plasmonic and photonic slow waves, it is interesting to in applications. The following subsections briefly review key
reconsider how the antiparallel power flows scale with ng. There aspects.
9 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Bandwidth. A major source of limitation of slow light


structure is dispersion. For any system, the operating frequency
is fixed, but in general, the transported signal possesses a
bandwidth, and one should consider how the slowness changes
over the bandwidth. To quantify the limiting impact of group-
velocity dispersion, it is convenient to define a dimensionless
coefficient ) as Figure 5. Attenuation channels for a tiny imperfection characterized by
a localized permittivity change Δεδ(r − r0) in a slow waveguide. Energy
Δω Δng can be either radiated into the cladding, backscattered, or absorbed if
=) Im(Δε) ≠ 0. Radiation into the cladding is equivalent to absorption in
ω ng (14) metal clads for MIM waveguides.
which quantifies the variation Δω of ω due to a small variation
Δng at a fixed operating value ng. Large bandwidths are achieved En|2. The absorption by metal cladding for plasmonic wave-
for large ) s, and reversely small bandwidths for small values. guides deserves special care.54 The backscattered power Pback is
) can be evaluated by considering the dispersion relation proportional to |En·J|2 (the excitation modal coefficient of the
ω(β). By derivation, it is possible to infer another relation 1
normalized back-propagating mode is 4 En·J ),44 so that Pback ∝ |
linking the frequency ω and the group velocity, or equivalently
the group index ng. Consider first a photonic waveguide with a En|4. These are only general trends, but they all tell us that the
quadratic dispersion curve ω(β) = (β − π/a)2/2m + ω0 near a influence of material loss or fabrication imperfection on mode
−1 structure is strong at small group velocities, where even a tiny
∂ 2ω
band-edge ω0, where a is the period and m = ( )
∂k 2
is the amount of loss produces large changes in the dispersion, such
that the group velocity is never zero.55−57 We will investigate this
effective photon mass, which describes the flatness of the
aspect in more detail now.
dispersion curve. For ng = (mc2/2)1/2(ω − ω0)1/2, we find
Roughness-Induced Attenuation of Photonic Wave-
mc 2 guides. The propagation of light in real photonic waveguides is
)= well documented in the literature; absorption can be neglected
ω0ng 2 (15) with photonic materials, so that roughness of the etched pattern
−2 is the main limiting technological factor. Since |En| ∝ ng , we
which evidences that ) scales with ng . The bandwidth near a
Brillouin zone boundary therefore rapidly decreases as the have that Prad ∝ ng and Pback ∝ ng2, so that the corresponding
operating group velocity decreases. attenuation lengths Lp scale with ng−1 and ng−2, respectively.38
To estimate ) for slow plasmons, we conveniently use the These simple scaling laws have been confirmed by ab initio
2ε electromagnetic computations performed with Bloch-mode-
dispersion relation of MIM waveguides, βg = − ε d ; see eq 3.
m expansion methods for realistic waveguide geometries incorpo-
dεm
We obtain ng = (2cεd /g ) ( dω
/εm 2 ), and using a Drude model rating roughness on the sidewalls of the etched holes.50 Theory
predicts typical values for the propagation length that vary
εm = ε∞(1 − ωp /ω ), ) then reads as
2 2
between Lp = 50 and 500 μm for ng = 100 and between Lp = 103
) = −1/3 (16) and 104 μm for ng = 30, depending on the accuracy of the
fabrication processes generally on the order of one nanometer or
and is independent of ng.
less. These theoretical predictions are confirmed by exper-
The bandwidth of photonic waveguides with quadratic
imental transmission and reflection measurements performed
dispersion curves can be enhanced by engineering to locally
with 100−1000λ long waveguides,18,23,37,51,58 showing a weak
flatten the dispersion curve,53 but even so, slow photonic
reflected signal for low group indices and a weak transmitted one
waveguides operate in a quite narrow frequency range that
at large group indices.
remains much smaller than that of slow plasmonic waveguides.
Absorption Loss in Plasmonic Waveguides. The
Propagation Length. Since light cannot propagate
extremely tight confinement in plasmonic components is due
indefinitely at low speed due to fabrication imperfections
to the large wavevectors of plasmons. It is inevitably
(surface roughness) or intrinsic material limitations (ohmic
accompanied by substantial absorption loss, essentially given
losses), the delay is finite. In general, the longer the delay for a
by the inverse of the damping rate of the metal. Thus, it is
given device, the narrower the bandwidth, which is why the so-
fundamentally difficult to achieve a long propagation distance at
called “delay−bandwidth product” is such a useful figure of
slow speeds.
merit. Let us then consider the propagation length Lp (defined
To analyze the attenuation due to metal absorption, one
for a 1/e field attenuation) of slow waveguides.
naturally may consider the expression of the absorbed power
Qualitatively first, the impact of imperfections can be
estimated by considering a local permittivity change Δε, a Pabs = ω/2 ∬ ε0 Im(εm)|En|2 dS , where the integral runs over
“Dirac” defect, in a waveguide. In the Born approximation, this the transverse cross-section of the plasmonic waveguide. Since
defect acts as a local current source, J = −jωΔεEnδ(r − r0), En ∝ ng for MIM waveguides (see eq 13), we may infer that Pabs
which is proportional to Δε and to the driving field (the slow ∝ ng2 at first glance. This however neglects the dispersion of the
incident mode En). As illustrated in Figure 5, this local source transverse mode profile and therefore largely overestimates the
may radiate into the cladding and induce absorption if the absorption. Instead, we directly rely on eq 9, which is valid for all
cladding or the defect absorbs or backscatter light into the reciprocal geometries, and we find that the integral of |e|2 or |En|2
waveguide by exciting the counter-propagating slow mode. The over a cross-section scales as ng. This evidences that Pabs actually
power Prad radiated into the cladding is proportional to |J|2, and scales as ng.
since J is proportional to En, Prad ∝ |En|2. Similarly, since the More insight into the role of mode-profile dispersion can be
power Pabs absorbed by the defect is proportional to J·En, Pabs ∝ | gained by considering the textbook case of MIM waveguides.
10 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

We showed in Section 2 that, as the gap width decreases, the gap- group velocities lower than c/1000 were seemingly observed.62
plasmon effective index neff = cβ/ω ∝ λ/g increases; see eq 3. For slow plasmons, experiments seemed to evidence that novel
Thus, the penetration depth of the gap plasmon mode into the metallic waveguide structures drastically reduce the propagation
c
metal, ω (neff 2 − εm)−1, decreases and becomes shorter than the speed and may even stop broadband light, producing the so-
called “trapped rainbows”.63,64 As we will now show, the very
skin depth. The effective area contributing to the cross-section
notion of group velocity loses all meaning at such small group
integral in Pabs = ω/2 ∬ ε0 Im(εm)|En|2 dS diminishes. It is thus velocities for both photons and plasmons.
the field confinement in the immediate vicinity of the MI There are two regimes for slow photons.52,56,65 Far from the
interfaces that is responsible for the ng scaling of the absorbed band-edges, where the concept of group velocity applies, tiny
power and not the ng2 scaling of our naive approach. Despite this fabrication errors introduce random spatial dephasing of the
lowering, the attenuation due to absorption is drastic in propagating field, and, as the interference between multiply
plasmonic waveguides. For instance, for a Ag/SiO2/Ag wave- scattered propagating waves may be destructive, propagating
guide with εAg = −27 + i1.5 at λ = 800, one finds Lp ≈ 300 nm for light is exponentially damped with an attenuation coefficient
ng = 10 and barely 100 nm for ng = 30. proportional to ng2 due to backscattering.38 This well-known
Roughness-Induced Attenuation in Plasmonic Wave- scaling is valid only in a double limit, when the disorder level
guides. Since Pback ∝ |En|4 and |En| ∝ ng for plasmonic tends toward zero at a fast enough rate compared to vg to
waveguides (compared to |En| ∝ ng for photonic waveguides), guarantee that the impact of random imperfections on transport
roughness-induced backscattering is plasmonic waveguides is remains perturbative.52 In practice, the disorder level, albeit very
scaling as ∝ ng4 and is expected to significantly alter the weak, is fixed and given by the employed nanofabrication
propagation of slow plasmons. Moreover, since Pabs only scales technique. A sudden breakdown of the perturbative regime,
as ng comparatively, one may wonder if backscattering may which inevitably ceases to be valid at small speeds, thus
become the dominant attenuation mechanism at small group eventually occurs. In this “ultraslow” regime, light propagating in
velocities. the randomly perturbed periodic waveguide not only experi-
This possibility has not been explored experimentally. ences random phase-shifts, but also strong back-reflection due to
Theoretically since imperfections at metal−dielectric interfaces vg-enhanced impedance mismatches or stop-band reflection for
result in a large Δε, the local current source J is not simply slow light operated near band-edges.52 The back-reflection
proportional to the plasmon mode field and local-field results in strong interference between multiply scattered waves,
corrections should be considered.59 Additionally, the plasmon- leading to the formation of wavelength-scale localized modes,
mode profile disperses as the group velocity is lowered. This may similar in nature to photonic-crystal microcavity modes. In this
explain why theoretical analyses on roughness-induced attenu- regime, the wavenumber k cannot be considered as a good
ation in plasmonic waveguides are rare. quantum number anymore, and the group velocity evidently
Imperfections in MIM waveguides with large gap widths loses physical significance.56,57
The same physics hold for slow plasmons, except that the
(operating away from the slow regime) have first been
situation is even worse since absorption loss additionally comes
documented in ref 60; a strong impact has been predicted but
into play and the localization regime is even not reached in
for relatively large roughness (∼4 nm) obtained for films
practice; the plasmon is damped in space or in time before
deposited by use of conventional e-beam evaporation. Slow
forming disordered resonant cavities. The literature is quite
propagation and small roughness have been analyzed49 for either
misleading on that matter.
chemically synthesized metal films or flakes, for which the metal
surfaces are essentially atomically flat, except for some potential
monatomic adlayer defects or high-quality polycrystalline metal
films composed of large grains with a small residual roughness
■ LIGHT EMISSION AND LDOS ENGINEERING IN
SLOW WAVEGUIDES
and separated by deep valleys that are remnants of grain Controlling light emission at the single-photon level is one of the
boundaries.61 The predictive conclusions are that back- major challenges in modern photonics, with great promises for
reflection due to imperfection is completely negligible for quantum communications, quantum computing, and optoelec-
chemically synthesized metal films. For polycrystalline films, tronics thanks to devices such as highly efficient nonclassical
back-reflection is predicted to be much stronger than for light sources, low-threshold lasers, and single-photon switches
photonic crystal waveguides, in accordance with the scaling laws, or diodes. Probably the most important problem to address to
and become more impactful than absorption for very small reach this aim is the realization of efficient coupling between an
speed regimes (ng > 30). emitter and a single guided mode. This coupling is described by
We might think that slow photons are much more sensitive to the β-factor that is the ratio between the power emitted into the
roughness than slow plasmons. This is actually not true, as mode of interest and the total emitted power. It may thus be
evidenced by the different scaling laws in ng4 for plasmons and made close to unity by boosting the emission into the mode
ng2 for photons. The strong impact of roughness is in fact hidden and/or reducing the emission into the other decay channels.
by absorption for usual slowness (ng ≈ 5−20). Despite having an Quantitatively, the amplitude coupling coefficient between a
obvious weaker scaling ∝ng2 than backscattering, absorption is dipole-current J0δ(r − r0) and translation-invariant43 or periodic
the dominant attenuation channel.49 The reason is simply that waveguides44 is proportional to the electric field of the
absorption (∝ng) occurs over the entire metal volume, whereas normalized mode at the source position r0. Slow modes with
backscattering (∝ng 4 ) is due only to surface-localized defects. large field enhancements (∝ ng for photons and ∝ng for
Impossibility to Reach Ultraslow Regimes. Some of the plasmons) are thus expected to offer near-unity couplings. In
first experimental investigations of the speed of light transport in this section, we review the basic principles underlying light
monomode photonic-crystal waveguides reported group veloc- emission in slow waveguides and identify the most promising
ities of c/50 and c/150,37 whereas in other types of experiments strategies for near-unity β-factors.
11 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Figure 6. Efficient funneling of the spontaneous decay of individual optical emitters into slow modes. (a) Slow photons: near-unity funneling into a
single slow mode can be achieved provided that the LDOS of the slow mode is much higher than the LDOS of the clad modes. (b) Slow plasmons: even
if tiny gaps lead to gap plasmons with extremely large LDOS, near-unity funneling is prohibited because of near-field nonradiative decay in the metal
(quenching).

Figure 7. Slow plasmons on single metal−dielectric interfaces. (a) Radiation of a vertical dipole current above a Ag/polymer interface. The color scale
represents log10(|E|2) of the total electric field for a dipole−metal separation distance d = 4 nm. The oscillating dipole decay is coupled into radiation
modes detected in the far field, into surface plasmons launched on the surface, or directly transformed by near-field coupling as Joule heating in the
metal just beneath the source. (b) Usual dispersion relation, Re k̃ − ω, of surface plasmons at a flat interface between a Drude-metal with a relative
permittivity εm = 1 − ωp2/(ω2 − iγω) and a dielectric material with a relative permittivity εd = 1. (c) Quasi-normal mode dispersion relation, Re ω̃ − k.
Slow plasmons with large k show up, but their lifetime is vanishingly small. The horizontal asymptote (dashed curve) occurs for a complex frequency
ω̃ SP, verifying εm(ω̃ SP) + εd = 0. (b and c) Both curves are obtained by solving eq 17. ωp is the plasma frequency and γ = 0.02ωp. The black line is the
light line of the air clad.

Emission in Slow Photonic Waveguides. The possibility in the presence of disorder scales as m ,67 with m the effective
of achieving near-unity β-factors is well comprehended for slow photon mass. One should therefore preferably opt for photonic
photonic waveguides; see Figure 6a. On one hand, the waveguides with a large effective photon mass, i.e., flat dispersion
enhancement (Purcell effect) of the LDOS scales with ng in curves, for optimal performance.
1D systems, thereby boosting the emission into the slow mode To date, near-unity β-factors (β ≈ 0.98) have been reported
when approaching the band-edge. On the other hand, photonic- with quantum dots embedded in semiclosed photonic-crystal
crystal waveguides offer great flexibility in loss engineering, such waveguides operating at ng > 50.9 Such a high efficiency is
that it is also possible to drastically reduce emission into the achieved on a narrow spectral range, yet it may be broadened up
other decay channels.44 β-Factors above 90% can hence be to a few tens of nanometers at the cost of a slightly lower β-factor
achieved even without strong LDOS enhancement, i.e., by by operating at lower ng values. As an alternative approach to
operating at moderate ng values. embedded quantum dots, important efforts are currently being
Unsurprisingly, the main limitation in terms of performance made to trap cold atoms near photonic structures.68 A recent
comes from fabrication imperfections, which yield the formation study reports β ≈ 0.5 for cold atoms trapped in the so-called
of localized modes with unpredictable spatial positions, resonant corrugated “alligator” waveguide designed to achieve long-range
frequencies, and quality factors.66 This limitation may be interactions between distant atoms for ng ≈ 11.69 The structure
mitigated by using waveguides shorter than the localization suffers from a much lower effective photon mass and fewer
length. In addition, when using emitters with a sharp spectral degrees of freedom for design compared to a photonic-crystal-
response, it is necessary to match the emission frequency with a type waveguide,67 suggesting that higher performances may be
desired ng. One can show that the highest ng that may be aimed at achieved with new geometries.
12 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Emission in Slow Plasmonic Waveguides. Compared to the inverse of Im ω̃ . The relation dispersion, Re ω̃ − k, shown in
light emission in photonic structures, the decay rate of emitters Figure 7c, is calculated by solving the transcendental eq 17 at
into slow plasmons70 is larger and the bandwidth is definitely complex frequencies for fixed k. Now, plasmon QNMs with
much broader. The only gray area is the efficiency issue because arbitrary large k’s (since quantum effects are not considered71)
of the prominence of near-field nonradiative decay in the metal are obtained for a certain complex frequency ω̃ SP, such that
for tiny gaps; see Figure 6b. The whole literature is not always εm(ω̃ SP) + εd = 0 i.e., ω̃SP = ωp (1 + εd) + iγ /2 for γ ≪ ωp.
clear on the role played by plasmons, especially “slow” plasmons
with large parallel momenta, into the decay process. In the The QNMs form a complete basis, so that the field radiated by
remainder of this section, we analyze this issue, step by step, first the dipole at the (real) excitation frequency ω can be
revisiting the classical light emission by a dipole-current located reconstructed from a sum over all QNMs.74
just above a single MI interface, and then considering the case of For illumination by a local oscillating dipole-current placed in
light emission by a dipole-current located in an MIM waveguide, the near field, QNMs with arbitrarily large k’s are excited.
the objective being to completely clarify the role of slow Intuitively, we anticipate that all modes are excited in phase at
plasmons. the dipole position, so that their respective contributions add up
Dipolar Emission near a Metal Surface. This problem has constructively especially just beneath the dipole, where strong
been reviewed by many authors, including ref 71. It contains a absorption takes place, as evidenced by the hot spot in Figure 7a.
rich variety of physical phenomena. In particular, we know that Away from the source on the surface, all the sinusoidal QNMs
the lifetime of an atom placed close to the surface can be are not phased and the sum of their field is no longer
typically reduced by several orders of magnitude,72 implying that constructive, so that the contribution of high-k modes to the
the LDOS close to an interface is drastic. The reduction is due to reconstructed field rapidly decreases away from the dipole in the
several decay channels. A part of the oscillating dipole decay is metal, as shown in Figure 7a. Moreover, very high-k modes have
coupled into radiation modes detected in the far field, another a very fast transversal decay (∝k−1) and are thus excited only
part launches surface plasmons on the surface, and another when the dipole is very close to the surface. This implies that, as
significant part is directly transformed by near-field coupling the distance d between the dipole and the surface decreases, the
into ohmic loss in the metal just beneath the source, resulting in hot spot becomes more intense and localized. Thus, with the
a quenching of the far-field emission especially when the dipole QNM interpretation, in contrast with the guided mode
is very close to the surface.72 interpretation, high-k slow plasmons play an essential (albeit
The plasmon dispersion relation of flat MI interfaces is well detrimental) role in the light emission process; quenching
known:73 drastically increases when d decreases (the decay rate scales as
ω (ωd/c)−3 in the static limit71) and when the dipole frequency ω
k= εdεm /(εm + εd) matches Re ω̃ SP, since the excitation coefficients of QNMs with a
c (17)
frequency ω̃ are proportional to (ω − ω̃ )−1.74
with k the parallel in-plane wavevector, ω the frequency, εm and The intuitive description of the role of “slow” plasmons has
εd the relative permittivities of the metal and dielectric materials. been carefully confirmed with numerical simulations obtained
We will use a Drude metal εm = 1 − ωp2/(ω2 − iγω) for for a silver nanorod in air. By computing several hundreds of
illustration hereafter. Since εm is complex, either k or ω is QNMs, it has been shown that the hot spot can be accurately
complex and two interpretations, in relation with slow plasmons, reconstructed in the QNM basis up to d as small as 1.5 nm; see
are usually considered. Figure 3 in ref 75. The spectrum of the plasmon
The first interpretation relies on the dispersion relation eigenfrequencies does not form a continuum for the nanorod
obtained by fixing the frequency to a real value (imposed by the like for the infinite flat interface. Rather they take discrete values.
driving dipole) and by directly solving eq 17 for complex k̃’s, However, for very small separation distance between the dipole
denoted with a tilde to emphasize their complex nature; the and the antenna, the dipole “sees” an almost flat surface. The
imaginary part of k̃ is related to the inverse of the attenuation nanorod spectrum presents an accumulation point at the
length of the plasmon mode. The dispersion relation Re k̃ − ω is plasmon resonance frequency of the flat interface for which
shown in Figure 7b. This plasmon mode is the analogue of the εm(ω̃ SP) + εd = 0,75 showing that these nanorod QNMs are
photonic-crystal mode of the previous section or the gap- indeed very similar to those of the flat surface.
plasmon mode analyzed in Section 2; it is a guided mode with a Dipolar Emission in MIM Waveguides. Let us now consider
spatial damping due to absorption. the emission of an oscillating dipole in the dielectric gap of a
Because realistic values for the loss are taken into account in MIM waveguide, in the limit of tiny gaps (see Figure 6b). The
Figure 7, the relation exhibits a back bending for a relatively low conclusions on the role of slow plasmons will be completely
value of Re k̃, Re k ̃ ≈ different. Indeed, in full analogy with the single-interface or
εd
ωp/c . In this real frequency
2(1 + εd) antenna cases, the nonradiative decay in the metal is again
representation, no slow plasmons with large k are launched,73 present. However, as discussed above, the slow gap plasmons
even when the emission frequency is close to ωp/(1 + εd)1/2 and (Re k̃ − ω) that are launched in the gap have an extremely large
the dipole is very close to the surface. This result contrasts with LDOS. Actually, it can be shown analytically that the
the emission in MIM waveguides with small gaps, presented nonradiative decay rate (quenching) and the slow-plasmon
below, for which we will show that guided gap plasmons with decay rate both scale as (k0g)−3. None of the two decay channels
large k̃’s are excited at any frequency. prevail as the gap width g becomes very small. The branching
Another representation of the same physics relies on the ratio η (the fraction of the total decay that effectively decay into
quasinormal modes (QNMs)74 that are the natural plasmon the gap mode) becomes an intrinsic quantity, as it is
resonances with a complex frequency ω̃ of the MI interface. The independent of g, depending solely on εd and on the metal
QNMs possess a sinusoidal spatial dependence parallel to the −1
surface (no spatial damping) and a finite lifetime proportional to (
loss Im εm, η ≈ 1 +
Imεm
2εd ) .54,76
13 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

The good news is that the emission is no longer quenched for of the plasmon amplitude. Nanofocusing devices have been used
tiny gaps so that, provided that the gap plasmon mode is in many applications, yet many new ones appear every year for
efficiently tapered into a photonic mode,54 large efficiencies are high-harmonic generation and hot-electron plasmon-assisted
achievable. The downside of the existence of a branching ratio is generation; see ref 63 for a review.
that the efficiency cannot reach unity. This imposes a limitation The second important slow-plasmon device is nanocavity or
on the β-factors that usually not exceed 80% for MIM54,77−79 or nanoantenna. At deep-subwavelength dimensions, these com-
IMI80,81 waveguides and remain significantly smaller than their ponents are essentially “Fabry−Perot cavities” involving slow
photonics counterparts. plasmons that propagate in a slow waveguide and are reflected at

■ PERSPECTIVES AND CONCLUSIONS


The applications of slow photonic waveguides in both linear and
the waveguide terminations. For instance, the dipolar resonance
of a simple nanorod immersed in a dielectric background can be
considered as resulting from the bouncing back and forth of two
nonlinear optics are so diverse that it is impossible to cover them IMI plasmon modes between the wire facets. Similarly, two
all even briefly. Slow photons are useful to realize compact all- MIM plasmon modes are involved in the dipole and quadrupole
optical buffer memories, which is one of the most important modes of cut-wire-pair resonators and nanopatch antennas.
missing components for the construction of various all-optical As the transverse dimensions (the rod diameter or gap width)
processing devices such as photonic routers. They are also useful are scaled down, the physics is the same. The fundamental IMI
to funnel light emission into a single channel and to enhance or MIM plasmon modes experience a monotonic increase in
various light−matter interactions in nonlinear optics. We have transverse localization and effective index (see eq 3), implying
no choice but to note that slow photon applications rarely that by reducing the nanoresonator length by the same scaling
feature a speed reduction greater than 10−20. For instance, factor, the resonance wavelength is maintained.90,91 This simple
moderate ng values (∼30) are used in photonic-crystal and model that promotes slow plasmons as an integral part of the
coupled-resonator waveguides to implement delay lines over near-field response of deep-subwavelength nanoresonators
millimeter distances.23,82,83 For short devices such as low-power predicts well the resonance frequency. Pushed further, it also
optical switches,20 typical ng values of ∼50 are met in practice. explains how important quantities such as the Q factor, the
Interesting perspectives may arise with regimes of smaller absorption, or the facet reflectivity scale.90 In particular, it
group velocities. Remembering that the attenuation length Lp of predicts that the Purcell factor scales as g−3,54 making slow-
photonic waveguides at λ = 1500 nm and ng = 100 varies from 50 plasmon nanoresonators or nanoantennas a serious candidate
to 500 μm according to the fabrication technology and that Lp ∝ for an ultimate control of light emission,10,91,92 especially if an
ng−2, we anticipate that slow light at ng = 300−500 may accurate control of the conversion of the slow gap plasmons into
propagate over 10-period long distances before backscattering free-space photons is implemented by engineering the tapered
takes place. Since remarkably short and efficient tapers have facets to boost the photon generation yield.
been designed47,84 (see Figure 4a) and successfully imple- In summary, we have provided a comparative overview of the
mented in many slow-light experiments,9,85−87 devices that physical mechanisms by which wave deceleration is achieved in
feature a sudden reduction of the group velocity followed by a plasmonic and photonic waveguides. The scaling laws of various
few-periods-long ultraslow section and a reciprocal acceleration important physical quantities (field enhancement, bandwidth,
to return to the initial speed appear within reach. Applications attenuation) as a function of ng have been presented; see Table
may include all-optical processing on chip, biosensors, or new 1. Slow plasmonic and photonic waveguides share many
resonator-like devices with anomalous step-like spectral trans- common features, but they have also distinct properties. We
mission. However, as ultraslow photonic waveguides and think that the key fundamental property responsible for the
microcavities have a lot in common, especially for large ng difference, notably for the mode intensities (∝ng for slow
when it becomes difficult to engineer waveguide dispersion,84 photons and ∝ng2 for slow plasmons), is that the slow mode of
attention should be paid not to repeat earlier works with high-Q photonic waveguides retains the same profile as light is slowed
microcavities. down, whereas the plasmon-mode profile drastically changes for
Comparatively, slow-plasmon studies are more prospective. plasmonic waveguides.
Slow plasmons can only propagate over very small distances, due Slow-plasmon devices are generally operating at much lower
to absorption. Their applications must therefore inevitably rely ng than slow-photon devices (ng = 15 corresponds to insulator
on components with very short lengths, not longer than a few gap width of g = 2 nm at visible wavelengths). One should
hundreds of nanometers. Two main slow-plasmon components nevertheless remember that slow plasmons and photons have
have emerged in the past decade, namely, nanofocusing tapers
different field-enhancement scalings (|En| ∝ ng for plasmons and
and slow-plasmon nanocavities or nanoantennas.
|En| ∝ ng for photons), so that it is rather illogical to compare
Nanofocusing devices that taper and concentrate optical
energy into nanoscale objects and structures probably constitute plasmonic and photonic devices for the same ng. In this respect, a
the most important device using slow plasmons. Various plasmonic taper operating at ng = 15 would be equivalent to a
geometries based on MIM and IMI configurations have been photonic taper operating at ng ≈ 225 in our analogy.
successfully implemented, e.g., conical metal rods with a Toward strong field enhancement and ultimate miniatur-
rounded tip for near-field imaging, metal wedges on substrates ization, the recent literature has promoted slow plasmons as the
such as in Figure 3, or exotic structures such as metal− pinnacle of slowness in nanophotonics. Indeed, field enhance-
insulator−insulator (MII) wedges88 or tapered chains of ment and bandwidth are decisive figures of merit that render
coupled nanoparticles;89 see ref 36 for a review. The whole plasmon wave deceleration incomparably more attractive than
literature suggests that adiabatic tapering into deep-subwave- photon wave deceleration. However, from other perspectives,
length slow plasmons is relatively easy to implement, since for the benefit of the deep-subwavelength confinement (or
sufficiently short tapers the local field enhancement along the miniaturization) brought by plasmons is not as clear. All-optical
taper could efficiently compensate for the dissipative reduction switches are perhaps the most important building blocks for on-
14 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

chip all-optical processing. Switching is generally achieved by (13) Ma, R.-M.; Oulton, R. F.; Sorger, V. J.; Zhang, X. Plasmon lasers:
modulating the cavity transmission intensity by illuminating the Coherent light source at molecular scales. Laser Photon. Rev. 2013, 7,
cavity with a control light beam, which shifts the resonance 1−21.
wavelength via the intensity-dependent refractive index of the (14) Xiong, C.; Monat, C.; Clark, A. S.; Grillet, C.; Marshall, G. D.;
cavity material. The required switching power scales as V/Q2,18 Steel, M. J.; Li, J.; O’Faolain, L.; Krauss, T. F.; Rarity, J. G.; Eggleton, B.
J. Slow-light enhanced correlated photon pair generation in a silicon
because the light intensity in the cavity is proportional to Q/V,
photonic crystal waveguide. Opt. Lett. 2011, 36, 3413−3415.
and the required wavelength shift is proportional to Q−1. In (15) Tame, M. S.; McEnery, K. R.; Ö zdemir, S. K.; Lee, J.; Maier, S. A.;
contrast to spontaneous rate (Purcell) enhancements,10 it thus Kim, M. S. Quantum plasmonics. Nat. Phys. 2013, 9, 329−340.
becomes unclear if plasmons, with their drastic reduction of V (16) Krauss, T. F. Photonic Crystals shine on. Phys. World 2006, 19,
often counterbalanced by an inevitable Q reduction due to 32−36.
absorption, can really be competitive in devices exploiting, for (17) Kauranen, M.; Zayats, A. V. Nonlinear plasmonics. Nat. Photonics
example, optical nonlinearities. 2012, 6, 737.
We expect that this short comparison leads to a balanced point (18) Notomi, M. Manipulating light with strongly modulated
of view between slow plasmonic and photonic waveguides. photonic crystals. Rep. Prog. Phys. 2010, 73, No. 096501.


(19) Soljačić, M.; Joannopoulos, J. D. Enhancement of nonlinear
AUTHOR INFORMATION effects using photonic crystals. Nat. Mater. 2004, 3, 211−219.
(20) Beggs, D. M.; White, T. P.; O’Faolain, L.; Krauss, T. F.
Corresponding Author Ultracompact and low-power optical switch based on silicon photonic
*E-mail: philippe.lalanne@institutoptique.fr. crystals. Opt. Lett. 2008, 33, 147−149.
ORCID (21) Vlasov, Y. A.; O’Boyle, M.; Hamann, H. F.; McNab, S. J. Active
control of slow light on a chip with photonic crystal waveguides. Nature
Philippe Lalanne: 0000-0003-1979-2290 2005, 438, 65−69.
Notes (22) Stockman, M. I. Nanoplasmonics: The physics behind the
The authors declare no competing financial interest. applications. Phys. Today 2011, 64, 39−44.


(23) Sancho, J.; Bourderionnet, J.; Lloret, J.; Combrié, S.; Gasulla, I.;
ACKNOWLEDGMENTS Xavier, S.; Sales, S.; Colman, P.; Lehoucq, G.; Dolfi, D.; Capmany, J.;
De Rossi, A. Integrable microwave filter based on a photonic crystal
The work was partly supported by the French State managed by delay line. Nat. Commun. 2012, 3, 1075−84.
the French National Agency for Research (ANR) in the frame of (24) Giugni, A.; Torre, B.; Toma, A.; Francardi, M.; Malerba, M.;
the “Investments for the Future” Programme IdEx Bordeaux− Alabastri, A.; Proietti Zaccaria, R.; Stockman, M. I.; Di Fabrizio, E. Hot-
LAPHIA (Grant No. ANR-10-IDEX-03-02) and by Bordeaux electron nanoscopy using adiabatic compression of surface plasmons.
University. Nat. Nanotechnol. 2013, 8, 845−852.


(25) Harutyunyan, H.; Martinson, A. B. F.; Rosenmann, D.; Khosravi
REFERENCES Khorashad, L.; Besteiro, L. V.; Govorov, A. O.; Wiederrecht, G. P.
Anomalous ultrafast dynamics of hot plasmonic electrons in
(1) Hau, L. V.; Harris, S. E.; Dutton, Z.; Behroozi, C. H. Light speed nanostructures with hot spots. Nat. Nanotechnol. 2015, 10, 770−775.
reduction to 17 m/s in a cold atomic gas. Nature 1999, 397, 594. (26) Sorger, V. J.; Ye, Z.; Oulton, R. F.; Wang, Y.; Bartal, G.; Yin, X.;
(2) Philips, D. F.; Fleischhauer, A.; Mair, A.; Walsworth, R. L.; Lukin, Zhang, X. Experimental demonstration of low-loss optical waveguiding
M. D. Storage of light in atomic vapor. Phys. Rev. Lett. 2001, 86, 783− at deep sub-wavelength scales. Nat. Commun. 2011, 2, 331.
786. (27) Eggleston, M. S.; Messer, K.; Zhang, L.; Yablonovitch, E.; Wu, M.
(3) Khurgin, J. B. Slow light in various media: a tutorial. Adv. Opt. C. Optical antenna enhanced spontaneous emission. Proc. Natl. Acad.
Photonics 2010, 2, 287. Sci. U. S. A. 2015, 112, 1704−1709.
(4) Boyd, R. W. Material slow light and structural slow light: (28) Demichel, O.; Petit, M.; Viarbitskaya, S.; Méjard, R.; de Fornel,
similarities and differences for nonlinear optics. J. Opt. Soc. Am. B 2011, F.; Hertz, E.; Billard, F.; Bouhelier, A.; Cluzel, B. Dynamics, Efficiency,
28, A38. and Energy Distribution of Nonlinear Plasmon-Assisted Generation of
(5) Baba, T. Slow light in Photonic crystals. Nat. Photonics 2008, 2,
Hot Carriers. ACS Photonics 2016, 3, 791−795.
465−473.
(29) Volkov, V. S.; Bozhevolnyi, S. I.; Devaux, E.; Ebbesen, T. W.
(6) Tsakmakidis, K. L.; Hess, O.; Boyd, R. W.; Zhang, X. Ultraslow
Bend loss for channel plasmon polaritons. Extremely Large Group-
waves on the nanoscale. Science 2017, 358, eaan5196.
Velocity Dispersion of Line-Defect Waveguides in Photonic Crystal
(7) Atwater, H. A.; Polman, A. Plasmonics for improved photovoltaic
devices. Nat. Mater. 2010, 9, 205−213. Slabs. Appl. Phys. Lett. 2006, 89, 143108.
(8) Scullion, M. G.; Di Falco, A.; Krauss, T. F. Slotted photonic crystal (30) Takahara, J.; Yamagishi, S.; Taki, H.; Morimoto, A.; Kobayashi,
cavities with integrated microfluidics for biosensing applications. T. Guiding of a one-dimensional optical beam with nanometer
Biosens. Bioelectron. 2011, 27, 101−105. diameter. Opt. Lett. 1997, 22, 475−477.
(9) Arcari, M.; Sö llner, I.; Javadi, A.; Lindskov Hansen, S.; (31) Dionne, J. A.; Sweatlock, L. A.; Atwater, H. A.; Polman, A. Planar
Mahmoodian, S.; Liu, J.; Thyrrestrup, H.; Lee, E. H.; Song, J. D.; metal plasmon waveguides: frequency-dependent dispersion, prop-
Stobbe, S.; Lodahl, P. Near-Unity Coupling Efficiency of a Quantum agation, localization, and loss beyond the free electron model. Phys. Rev.
Emitter to a Photonic Crystal Waveguide. Phys. Rev. Lett. 2014, 113, B: Condens. Matter Mater. Phys. 2005, 72, No. 075405.
No. 093603. (32) Dionne, J. A.; Sweatlock, L. A.; Atwater, H. A.; Polman, A.
(10) Akselrod, G. M.; Argyropoulos, C.; Hoang, T. B.; Ciracì, C.; Plasmon slot waveguides: Towards chip-scale propagation with
Fang, C.; Huang, J.; Smith, D. R.; Mikkelsen, M. H. Probing the subwavelength-scale localization. Phys. Rev. B: Condens. Matter Mater.
mechanisms of large Purcell enhancement in plasmonic nanoantennas. Phys. 2006, 73, No. 035407.
Nat. Photonics 2014, 8, 835. (33) Kurokawa, Y.; Miyazaki, H. T. Metal-insulator-metal plasmon
(11) Berini, P.; De Leon, I. Surface plasmon-polariton amplifiers and nanocavities: Analysis of optical properties. Phys. Rev. B: Condens.
lasers. Nat. Photonics 2012, 6, 16−24. Matter Mater. Phys. 2007, 75, No. 035411.
(12) Yang, J.-K.; Noh, H.; Rooks, M. J.; Solomon, G. S.; Vollmer, F.; (34) Oulton, R. F.; Bartal, G.; Pile, D. F. P.; Zhang, X. Confinement
Cao, H. Lasing in localized modes of a slow light photonic crystal and propagation characteristics of subwavelength plasmonic modes.
waveguide. Appl. Phys. Lett. 2011, 98, 241107. New J. Phys. 2008, 10, 105018.

15 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

(35) Stockman, M. I. Nanoplasmonics: past, present, and glimpse into (57) Topolancik, J.; Ilic, B.; Vollmer, F. Experimental Observation of
future. Opt. Express 2011, 19, 22029. Strong Photon Localization in Disordered Photonic Crystal Wave-
(36) Gramotnev, D. K.; Bozhevolnyi, S. I. Nanofocusing of guides. Phys. Rev. Lett. 2007, 99, 253901.
electromagnetic radiation. Nat. Photonics 2014, 8, 13−22. (58) Johnson, S. G.; Povinelli, M. L.; Soljačić, M.; Karalis, A.; Jacobs,
(37) Notomi, M.; Yamada, K.; Shinya, A.; Takahashi, J.; Takahashi, S.; Joannopoulos, J. D. Roughness losses and volume-current methods
C.; Yokohama, I. Extremely large group-velocity dispersion of line- in photonic-crystal waveguides. Appl. Phys. B: Lasers Opt. 2005, 81,
defect waveguides in photonic crystal slabs. Phys. Rev. Lett. 2001, 87, 283−293.
253902. (59) Min, C.; Veronis, G. Theoretical investigation of fabrication-
(38) Hughes, S.; Ramunno, L.; Young, J. F.; Sipe, J. E. Extrinsic optical related disorders on the properties of subwavelength metal-dielectric-
scattering loss in photonic crystal waveguides: role of fabrication metal plasmonic waveguides. Opt. Express 2010, 18, 20939−20948.
disorder and photon group velocity. Phys. Rev. Lett. 2005, 94, (60) Munoz, R. C.; Concha, A.; Mora, F.; Espejo, R.; Vidal, G.;
No. 033903. Mulsow, M.; Arenas, C.; Kremer, G.; Moraga, L.; Esparza, R.; Haberle,
(39) Palik, E.D. Handbook of Optical Constants of Solids; Academic P. Surface roughness and size effects of thin gold films on mica:
Press: NY, Part II, 1985. Application of quantitative scanning tunneling microscopy. Phys. Rev.
(40) Lalanne, P.; Besbes, M.; Hugonin, J. P.; van Haver, S.; Janssen, O. B: Condens. Matter Mater. Phys. 2000, 61, 4514−17.
T. A.; Nugrowati, A. M.; Xu, M.; Pereira, S. F.; Urbach, H. P.; van de (61) Gersen, H.; Karle, T. J.; Engelen, R. J. P.; Bogaerts, W.; Korterik,
Nes, A. S.; Bienstman, P.; Granet, G.; Moreau, A.; Helfert, S.; Sukharev, J. P.; van Hulst, N. F.; Krauss, T. F.; Kuipers, L. Real-Space Observation
M.; Seideman, T.; Baida, F. I.; Guizal, B.; Van Labeke, D. Numerical of Ultraslow Light in Photonic Crystal Waveguides. Phys. Rev. Lett.
analysis of a slit-groove diffraction problem. J. Eur. Opt. Soc. Rapid Publ. 2005, 94, No. 073903.
2007, 2, No. 07022. (62) Tsakmakidis, K. L.; Boardman, A. D.; Hess, O. ’Trapped rainbow’
(41) Tserkezis, C.; Mortensen, N. A.; Wubs, M. How nonlocal storage of light in metamaterials. Nature 2007, 450, 397−401.
damping reduces plasmon-enhanced fluorescence in ultranarrow gaps. (63) Gana, Q.; Gao, Y.; Wagner, K.; Vezenov, D.; Ding, Y. J.; Bartoli,
Phys. Rev. B: Condens. Matter Mater. Phys. 2017, 96, No. 085413. F. J. Experimental verification of the rainbow trapping effect in adiabatic
(42) Luo, Y.; Fernandez-Dominguez, A. I.; Wiener, A.; Maier, S. A.; plasmonic gratings. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 5169−5173.
Pendry, J. B. Surface Plasmons and Nonlocality: A Simple Model. Phys. (64) Engelen, R. J. P.; Mori, D.; Baba, T.; Kuipers, L. Two regimes of
Rev. Lett. 2013, 111, No. 093901. slow-light losses revealed by adiabatic reduction of group velocity. Phys.
(43) Snyder, A. W.; Love, J. D. Optical Waveguide Theory; Chapman Rev. Lett. 2008, 101, 103901.
and Hall: NY, 1983. (65) Smolka, S.; Thyrrestrup, H.; Sapienza, L.; Lehmann, T. B.; Rix, K.
(44) Lecamp, G.; Hugonin, J. P.; Lalanne, P. Theoretical and
R.; Froufe-Pérez, L. S.; García, P. D.; Lodahl, P. Probing the statistical
computational concepts for periodic optical waveguides. Opt. Express
properties of Anderson localization with quantum emitters. New J. Phys.
2007, 15, 11042−60.
2011, 13, No. 063044.
(45) Yeh, P. Optical Waves in Layered Media; J. Wiley and Sons: New
(66) Zang, X.; Yang, J.; Faggiani, R.; Gill, C.; Petrov, P. G.; Hugonin,
York, 1988.
J.-P.; Vynck, K.; Bernon, S.; Bouyer, P.; Boyer, V.; Lalanne, P.
(46) Bendickson, J. M.; Dowling, J. P.; Scalora, M. Analytic
Interaction between atoms and slow light: a study in waveguide design.
expressions for the electromagnetic mode density in finite, one-
dimensional, photonic band-gap structure. Phys. Rev. E: Stat. Phys., Phys. Rev. Appl. 2016, 5, No. 024003.
(67) Goban, A.; Hung, C.-L.; Yu, S. P.; Hood, J. D.; Muniz, J. A.; Lee, J.
Plasmas, Fluids, Relat. Interdiscip. Top. 1996, 53, 4107−4121.
(47) Hugonin, J. P.; Lalanne, P.; White, T.; Krauss, T. F. Coupling into H.; Martin, M. J.; McClung, A. C.; Choi, K. S.; Chang, D. E.; Painter,
slow-mode photonic crystal waveguides. Opt. Lett. 2007, 32, 2638− O.; Kimble, H. J. Atom−light interactions in photonic crystals. Nat.
2640. Commun. 2014, 5, 3808.
(48) Khurgin, J. B. How to deal with the loss in plasmonics and (68) Goban, A.; Hung, C.-L.; Hood, J. D.; Yu, S.-P.; Muniz, J. A.;
metamaterials. Nat. Nanotechnol. 2015, 10, 1−6. Painter, O.; Kimble, H. J. Superradiance for atoms trapped along a
(49) Coudert, S.; Duchateau, G.; Dilhaire, S.; Lalanne, P. Attenuation photonic crystal waveguide. Phys. Rev. Lett. 2015, 115, No. 063601.
of slow Metal-Insulator-Metal plasmonic waveguides, from Joule (69) Bozhevolnyi, S. I.; Khurgin, J. B. Fundamental limitations in
absorption to roughness-induced backscattering. arXiv:1810.e07389. spontaneous emission rate of single-photon sources. Optica 2016, 3,
(50) Mazoyer, S.; Hugonin, J. P.; Lalanne, P. Disorder-induced 1418−1421.
multiple-scattering in photonic-crystal waveguides. Phys. Rev. Lett. (70) Ford, G. W.; Weber, W. H. Electromagnetic interactions of
2009, 103, No. 063903. molecules with metal surfaces. Phys. Rep. 1984, 113, 195−287.
(51) Le Thomas, N.; Diao, Z.; Zhang, H.; Houdré, R. Statistical (71) Drexhage, K. H. Interaction of light with monomolecular dye
analysis of subnanometer residual disorder in photonic crystal layers. Prog. Opt. 1974, 12, 165.
waveguides: Correlation between slow light properties and structural (72) Raether, H. Surface Plasmons on Smooth and Rough Surfaces and
properties. J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.: Mater., on Gratings; Springer-Verlag: Berlin, 1988.
Process., Meas., Phenom. 2011, 29, No. 051601. (73) Lalanne, P.; Yan, W.; Vynck, K.; Sauvan, C.; Hugonin, J.-P. Light
(52) Faggiani, R.; Baron, A.; Zang, X.; Lalouat, L.; Schulz, S. A.; interaction with photonic and plasmonic resonances. Laser Photonics
O’Regan, B.; Vynck, K.; Cluzel, B.; de Fornel, F.; Krauss, T. F.; Lalanne, Rev. 2018, 12, 1700113.
P. Lower bounds for the spatial extent of optical localized modes in 1D (74) Yan, W.; Faggiani, R.; Lalanne, P. Rigorous modal analysis of
disordered periodic media. Sci. Rep. 2016, 6, 27037. plasmonic resonances. Phys. Rev. B: Condens. Matter Mater. Phys. 2018,
(53) Frandsen, L. H.; Lavrinenko, A. V.; Fage-Pedersen, J.; Borel, P. I. 97, 205422.
Photonic crystal waveguides with semi-slow light and tailored (75) Yang, J.; Faggiani, R.; Lalanne, P. Light emission in nanogaps:
dispersion properties. Opt. Express 2006, 14, 9444−9450. overcoming quenching. Nanoscale Horizons 2016, 1, 11−13.
(54) Reza, A.; Dignam, M. M.; Hughes, S. Can light be stopped in (76) Martín-Cano, D.; Martín-Moreno, L.; García-Vidal, F. J.;
realistic metamaterials? Nature 2008, 455, E10−E11. Moreno, E. Resonance Energy Transfer and Superradiance Mediated
(55) Le Thomas, N.; Zhang, H.; Houdré, R. Light transport regimes in by Plasmonic Nanowaveguides. Nano Lett. 2010, 10, 3129−3134.
slow light photonic crystal waveguides. Phys. Rev. B: Condens. Matter (77) Sorger, V. J.; Pholchai, N.; Cubukcu, E.; Oulton, R. F.; Kolchin,
Mater. Phys. 2009, 80, 125332. P.; Borschel, C.; Gnauck, M.; Ronning, C.; Zhang, X. Strongly
(56) Huisman, S. R.; Ctistis, G.; Stobbe, S.; Mosk, A. P.; Herek, J. L.; Enhanced Molecular Fluorescence inside a Nanoscale Waveguide Gap.
Lagendijk, A.; Lodahl, P.; Vos, W. L.; Pinkse, P. W. H. Measurement of Nano Lett. 2011, 11, 49077−4911.
a band-edge tail in the density of states of a photonic-crystal waveguide. (78) Kongsuwan, N.; Demetriadou, A.; Chikkaraddy, R.; Benz, F.;
Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 86, 155154. Turek, V. A.; Keyser, U. F.; Baumberg, J. J.; Hess, O. Suppressed

16 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17
ACS Photonics Perspective

Quenching and Strong-Coupling of Purcell-Enhanced Single-Molecule


Emission in Plasmonic Nanocavities. ACS Photonics 2018, 5, 186−191.
(79) Chen, Y.; Nielsen, T. R.; Gregersen, N.; Lodahl, P.; Mørk, J.
Finite-element modeling of spontaneous emission of a quantum emitter
at nanoscale proximity to plasmonic waveguides. Phys. Rev. B: Condens.
Matter Mater. Phys. 2010, 81, 125431.
(80) Yang, Y.; Zhen, B.; Hsu, C. W.; Miller, O. D.; Joannopoulos, J. D.;
Soljačić, M. Optically Thin Metallic Films for High-Radiative-Efficiency
Plasmonics. Nano Lett. 2016, 16, 4110−4117.
(81) Notomi, M.; Kuramochi, E.; Tanabe, T. Large-scale arrays of
ultrahigh-Q coupled nanocavities. Nat. Photonics 2008, 2, 741−747.
(82) Melloni, A.; Canciamilla, A.; Ferrari, C.; Morichetti, F.;
O’Faolain, L.; Krauss, T. F.; De La Rue, R.; Samarelli, A.; Sorel, M.
Tunable delay lines in silicon photonics: coupled resonators and
photonic crystals, a comparison. IEEE Photonics J. 2010, 2, 181−194.
(83) Faggiani, R.; Jang, J.; Hostein, R.; Lalanne, P. Implementing
structural slow light on short length scales: the photonic speed bump.
Optica 2017, 4, 393−399.
(84) Corcoran, B.; Monat, C.; Grillet, C.; Moss, D. J.; Eggleton, B. J.;
White, T. P.; O’Faolain, L.; Krauss, T. F. Green light emission in silicon
through slow-light enhanced third-harmonic generation in photonic-
crystal waveguides. Nat. Photonics 2009, 3, 206−210.
(85) Spasenovic, M.; et al. Experimental observation of evanescent
modes at the interface to slow-light photonic crystal waveguides. Opt.
Lett. 2011, 36, 1170.
(86) Lodahl, P.; Mahmoodian, S.; Stobbe, S. Interfacing single
photons and single quantum dots with photonic nanostructures. Rev.
Mod. Phys. 2015, 87, 347.
(87) Verhagen, E.; Kuipers, L. K.; Polman, A. Plasmonic nanofocusing
in a dielectric wedge. Nano Lett. 2010, 10, 3665−3669.
(88) Fernández-Domínguez, A. I.; Maier, S. A.; Pendry, J. B.
Collection and concentration of light by touching spheres: a
transformation optics approach. Phys. Rev. Lett. 2010, 105, 266807.
(89) Yang, J.; Sauvan, C.; Jouanin, A.; Collin, S.; Pelouard, J. L.;
Lalanne, P. Ultrasmall metal-insulator-metal nanoresonators: impact of
slow-wave effects on the quality factor. Opt. Express 2012, 20, 16880−
16891.
(90) Dorfmüller, J.; Vogelgesang, R.; Weitz, R. T.; Rockstuhl, C.;
Etrich, C.; Pertsch, T.; Lederer, F.; Kern, K. Fabry-Perot resonances in
one-dimensional plasmonic nanostructures. Nano Lett. 2009, 9, 2372−
2377.
(91) Aizpurua, J.; Bryant, G. W.; Richter, L. J.; García de Abajo, F. J.
Optical properties of coupled metallic nanorods for field-enhanced
spectroscopy. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71,
235420.
(92) Russel, K. J.; Liu, T. L.; Cui, S.; Hu, E. L. Large spontaneous
emission enhancement in plasmonic cavities. Nat. Photonics 2012, 6,
459−462.

17 DOI: 10.1021/acsphotonics.8b01337
ACS Photonics 2019, 6, 4−17

You might also like