Article

pubs.acs.org/journal/apchd5

Near-to-Far Field Transformations for Radiative and Guided Waves

Jianji Yang,*,†,# Jean-Paul Hugonin,*,‡ and Philippe Lalanne*,†
†
Laboratoire Photonique Numérique et Nanosciences (LP2N, UMR 5298), Institut d’Optique d’Aquitaine,
CNRS−IOGS−Université Bordeaux, 33400 Talence, France
‡
Laboratoire Charles Fabry (LCF, UMR 8501), CNRS−IOGS−Université Paris-Sud, 2 Avenue Augustin Fresnel, 91127 Palaiseau,
France
*
S Supporting Information
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: Light emitters or scatterers embedded in

stratified media may couple energy to both free-space modes
and guided modes of the stratified structure. For a
Downloaded via UNIV OF BORDEAUX on October 1, 2019 at 07:26:44 (UTC).

comprehensive analysis, it is important to evaluate the angular

intensity distribution of both the free-space modes and guided
modes excited in such systems. In the present work, we
propose an original method based on Lorentz reciprocity
theorem to efficiently calculate the free-space and guided
radiation diagrams with a high accuracy from the sole
knowledge of the near-field around the emitters or scatterers.
Compared to conventional near-to-far field transformation
techniques, the proposal allows one to easily evaluate the guided-mode radiation diagrams, even if material dissipation is present
in the stack, and thus to simultaneously track the coupling of light to all channels (i.e., free-space and guided ones). We also
provide an open-source code that may be used with essentially any Maxwell’s equation solver. The numerical tool may help to
engineer various devices, such as light-emitting diodes or nanoantennas, to achieve directional and efficient radiative spontaneous
decays in free-space and guided optics.
KEYWORDS: light emission, light scattering, mode expansion, radiation diagram, stratified structure, Lorentz reciprocity,
computational electrodynamics

field distributions are usually referred to as near-to-far field

W hen excited either by propagative electromagnetic waves
(e.g., plane waves or guided waves) or by local sources
such as dipole emitters, individual scatterers or aggregates of
transformation (NFFT) approaches. The latter are mature
techniques in microwave engineering.21 Nearly all numerical
scatterers embedded in a stratified medium may couple light implementations of NFFT techniques rely on the field
into multiple channels including guided modes of the stratified equivalence principle (or Huygens’ principle),21 which allows
structure and free-space radiative modes of the claddings. The us to replace the radiated field of an antenna in the near-field
coupling mechanism plays a central role in a large variety of zone by equivalent current sources and then to generate far-
research topics in photonics, for instance, for improving the zone fields using the free-space Green’s function and the
performance of thin-film solar cells using metal nanoparticles,1,2 equivalent sources. For antennas placed in a stratified substrate,
enhancing light extraction in light-emitting devices using the free-space Green’s function must be replaced by Green’s
nanoparticles,3−5 achieving directional light scattering6,7 or functions that account for the strata.22,23
emission8−10 using nanoantennas,11 realizing various optical Many authors have contributed to the elaboration of
functionalities (e.g., beam steering12,13 and shaping14) with numerical methods on NFFT approaches, in various fields of
metasurfaces based on nanoresonator arrays, or engineering the physics, maybe starting from initial works on the emission and
scattering properties of plasmonic nano-objects such as metallic
propagation of radio waves above the earth.24 However, it
slits/grooves15−17 and metallic holes.18−20
appears that the literature contains only scattered electro-
In electromagnetic simulations with Maxwell’s equation
magnetic contributions and that a unified numerical tool that
solvers, usually one can easily compute only the near-zone
field around the scatterers. However, the far-zone field is often retrieves both free-space and guided-mode radiation diagrams is
of great interest because it contains all the information on the still lacking, at least in the optical domain.23 Hereafter, we
so-called f ree-space radiation diagram, which describes the remove this deficiency and establish a formalism to retrieve the
angular distribution of the energy flux density radiated into free far-zone field from the near-zone field for electromagnetism
space, and the guided-mode radiation diagram, which describes problems due to the scattering by local inhomogeneities in
the angular distribution intensity radiated into guided modes of
the stratified medium. Numerical techniques capable of Received: September 30, 2015
computing far-zone radiation diagrams only using near-zone Published: February 16, 2016

© 2016 American Chemical Society 395 DOI: 10.1021/acsphotonics.5b00559

ACS Photonics 2016, 3, 395−402
ACS Photonics Article

stratified media. We also provide an open-source NFFT one is able to calculate the electromagnetic field (E,H) on the
toolbox,25 which retrieves the far-field radiation diagrams box. (E,H) can be either
from the knowledge of the permittivities and permeabilities • for J 2 = 0, the scattered field induced by the
of the layers (assumed uniform and isotropic) of the stratified inhomogeneities and obtained as the difference between
structure and of the near-zone field that can be obtained with the total field and the background field (which is
virtually any Maxwell’s equation solver on a closed surface generated by the source J1 illuminating the stratified
surrounding the local inhomogeneity. medium in the absence of the inhomogeneities);
To be more precise, let us consider a planar stratified
• for J1 = 0, the total field generated by the source J2 that
medium, in which a few scatterers are embedded (see Figure
illuminates the embedded scatterers (this situation
1a). We further consider a parallelepipedic box that fully
represents a typical extraction problem encountered
with light-emitting devices);
• any linear combination of the two previous cases.
For all cases, (E,H) on the box is formed only by outgoing
(regarding the “box”) plane waves and outgoing guided modes
(no incoming fields), and the present NFFT method and open-
source toolbox may accurately retrieve the far-field radiation
diagrams for guided and free-space modes from the sole
knowledge of (E,H) and the permittivities and permeabilities of
the layers (assumed uniform and isotropic) of the stratified
structure. The approach is very general, as it applies to
inhomogeneities with arbitrary shapes, any materials, and any
thin-film stack, even for lossy dielectric or metallic materials.
The only requirement is that all material must be reciprocal, i.e.,
ε = εT and μ = μT, where the superscript “T” denotes the
transpose operator.
This article is organized as follows. First, we briefly describe a
general NFFT method for calculating the free-space radiation
diagrams. Then, as the main contribution of the present work,
the formulation of a novel and substantially analytical method
to retrieve guided-mode radiation diagrams is proposed. Third,
the efficiency and applicability of the numerical tool is
exemplified by studying some basic light scattering/emission
problems in stratified media, and the validity and accuracy of
the open-source code is tested. In the last section, the usage of
the open-source tool and the conditions under which the tool
may be used safely are summarized.

Figure 1. Radiation diagram calculation due to the scattering of local

inhomogeneities in a stratified medium. (a) The scatterers (rectangle,
■ A CLASSICAL APPROACH TO RETRIEVE THE
FREE-SPACE RADIATION DIAGRAMS
sphere, and triangle) may be excited by either a far-field source J1 or a In this section, we describe an NFFT method to retrieve the
near-field source J2. They excite both free-space radiative modes and far-zone free-space radiation diagrams. This is a classical
guided modes of the stratified structure. (b) Retrieval of the angular problem in electromagnetism, which has already been treated
distribution of plane-wave amplitudes cf(θ, φ) and guided-mode by many authors; see for instance the recent works in refs 22
amplitudes cg(φ). The blue closed “box” Σ fully encloses all the local and 23 based on a Green’s dyadics approach and the field
inhomogeneity, so that the geometry outside the box should be the
equivalence principle. A largely equivalent approach based on
stratified medium. The retrieval of cf(θ, φ) and cg(φ) requires the
knowledge of the field (E,H) on the “box”, the relative permittivity, Lorentz reciprocity theorem is briefly outlined hereafter. This is
and permeability of every layer. In practice, a small-sized box can be not the main focus of the present work; however, we think it is
used and (E,H) on the box can be obtained with any Maxwell’s still worth being documented concisely for the sake of
equation solver. The accompanying open-source code performs rapid completeness, since it is implemented in the open-source
computation of the radiation diagrams (amplitude and intensity) of NFFT tool and is inherently linked to the original contribution,
the free-space radiated waves and the guided waves. Inset in (a) shows the NFFT of guided modes, which is presented in the next
the coordinates. section.
We consider the radiation diagram into the upper half-space
(z → ∞) of the field distribution (E,H), assuming lm(ε1) = 0
encloses the scatterers, so that the geometry outside the box is and lm(μ1) = 0. The radiation in the lower half-space is treated
strictly the stratified medium without any inhomogeneity. The similarly. Since the radiation diagram essentially describes the
blue closed “box” Σ of Figure 1a provides an example. The angular distribution of the energy flux density, the retrieval
inhomogeneities might be illuminated by an external source, consists in performing a plane-wave decomposition of (E,H)
such as J1 in Figure 1a, which is placed outside the box (if the and to compute the amplitudes of the radiated plane waves for
source is very far, then the illumination is a plane wave) or by a every direction. Both TE- and TM-polarized waves have to be
localized source, such as J2 in Figure 1a, which is placed inside taken into account for every direction, as implemented in the
the box. We assume that with some Maxwell’s equation solver open-source code; however, because plane waves with
396 DOI: 10.1021/acsphotonics.5b00559
ACS Photonics 2016, 3, 395−402
ACS Photonics Article

orthogonal polarizations are independently retrieved, in the structure may be launched in all in-plane directions [i.e., 0 < φ
following we consider purely TE- or TM-polarized waves to < 2π].
reduce cumbersome notations. Before going to details, we would like to briefly address the
It is convenient to first consider the upward-going plane following question: bearing in mind the NFFT method for free-
wave (either TE- or TM-polarized) that propagates away from space waves described in the former section, can we extract the
the scatterers in the directions defined by (θ, φ); see the inset guided-mode radiation diagram by performing, similarly to eq
in Figure 1a. The plane wave is denoted by 2, an overlap integral on the closed “box” Σ between the field
+ +
(E0̂ , Ĥ 0 ) exp[i(ux + vy + χz)], with the superscript “+” of (E,H) and the field of the associated incident guided mode
+ + propagating in the direction φ, i.e., replacing the incident plane
E0̂ and Ĥ 0 labeling upward-going waves (a minus sign is used − −
wave (E ̂ , Ĥ ) by an incident guided mode with an in-plane
hereafter for downward-going waves), and wavevector parallel to the azimuthal direction φ? The answer
u = k 0 ε1μ1 sin(θ ) cos(φ), v = k 0 ε1μ1 sin(θ ) sin(φ), and that can be easily derived from the Lorentz reciprocity theorem
χ = k 0 ε1μ1 cos(θ), with k0 denoting the wavenumber in a is yes for nonlossy materials, but the treatment is problematic
vacuum. In an horizontal plane at z = z1 in the upper half-space, for lossy modes because one needs to handle exponentially
(E,H) can be expanded as large fields.
To remove this difficulty, we formulate a novel and
E(x , y , z1) = ∬ cf+(θ , φ)E0̂ + exp[i(ux + vy + χz1)] du substantially analytical approach that allows one to evaluate
the radiation diagram of guided modes with a high accuracy, for
dv (1.1) lossy (dielectric or metallic) stratified structures. The method
requires only the knowledge of the field (E,H) on the closed
box Σ and of the guided-mode profile. All the derivation steps
H(x , y , z1) = ∬ cf+(θ , φ)Ĥ 0+ exp[i(ux + vy + χz1)] du are provided hereafter, with some technical details being
documented in the SI.
dv (1.2) Cylindrical Waveguide Modes. We start by expanding
where c+f (θ, φ) denotes the plane-wave amplitude coefficient. In the radiated field (E,H) using waveguide modes of the stratified
principle, c+f (θ, φ) can be found with inverse Fourier transform. structure defined in a cylindrical coordinate system (r, φ, z);
However, this requires the knowledge of the radiated field in see Figure 2. When solving source-free Maxwell’s equations
the entire (infinite) plane z = z1, which is not suitable for
numerical calculations.
Indeed, according to Lorentz reciprocity theorem, c+f (θ, φ)
can be obtained by computing a surface integral on a closed
“box” Σ surrounding the local inhomogeneities (see Figure 1b):
k 0Z0μ1
c f+(θ , φ) =
8π 2 + −
χE0̂ ·E0̂
∮Σ (E × Ĥ − − E−̂ × H) dS
(2)
− −
where Z0 denotes the vacuum impedance and (E ̂ , Ĥ )
represents the field that is created by a downward plane wave
− −
(E0̂ , Ĥ 0 )exp[−i(ux + vy + χz)] impinging onto the stratified
medium (without the inhomogeneities). The flux density at
direction (θ, φ) can be found as Figure 2. Domain Ω with its boundaries. ΣA (vertical black dashed
1 +
2
ε0ε1/(μ0 μ1) |c f+(θ , φ)|2 |E0̂ |2 . In passing, we emphasize lines) defines a cylindrical surface of radius r that extends from z =
− − −∞ to + ∞, ΣB is a closed “box” that encloses the inhomogeneity, and
that, because the vectorial field (E ̂ , Ĥ ) induced by a plane ΣC corresponds to the upper and lower horizontal surfaces (blue
wave in a stratified structure can be calculated rather easily with dashed lines) of the cylinder at |z| = ∞. ΣA and ΣC together form the
2 × 2 matrix products for every incidence and polarization,22,26 outer boundary of the volume Ω, and ΣB is the inner boundary. The
the overlap integral of eq 2 can be computed in parallel for all outward surface normal of Ω is denoted by n.
directions and the c+f (θ, φ) coefficients can be computed very
efficiently (see Technical Remarks). (eigenmode problems) of stratified planar waveguides, the
We do not provide a demonstration of eq 2 with the Lorentz eigenmodes can be written as the product of functions with
reciprocity theorem hereafter; it is technical and essentially separated variables, z, φ, and r, with the r-dependent part
echoes the classical NFFT techniques based on the field expressed with Hankel functions (see SI). Hereafter, the mth-
equivalence principle,21 without explicitly considering the order TE- or TM-polarized modes with propagation constants
equivalent current sources. + +
km will be denoted as Φ+m , n = (Em̂ , n exp(inφ), Ĥ m , n exp(inφ))

■ SEMIANALYTICAL APPROACH TO RETRIEVE THE

GUIDED-MODE RADIATION DIAGRAMS
for outgoing (to r → ∞) modes propagating and
− −
Φ−m , n = (Em̂ , n exp( −inφ), Ĥ m , n exp( −inφ)) for ingoing (to r
→ 0) modes, with n (n = 0, ±1, ...) determining the azimuthal
In this section, an original NFFT approach for retrieving ± ±
guided-mode radiation diagrams for planar stratified structure is pattern and (Em̂ , n , Ĥ m , n) describing the r- and z-dependent
formulated. We again consider the emission problem sketched profiles. Note that due to the separation of variables r and φ, for
in Figure 1b, in which the guided modes of the planar stratified a given polarization, each order m embraces a series of
397 DOI: 10.1021/acsphotonics.5b00559
ACS Photonics 2016, 3, 395−402
ACS Photonics Article

azimuthal harmonics with the same propagation constant km but For that, we define the inner product between two modes
σ σ
with different azimuthal pattern determined by exp(±inφ). The Φσm , n = (Em̂ , n exp(σinφ), Ĥ m , n exp(σinφ)) and
explicit expressions of mode profiles (including both the ξ ξ
general expression and asymptotic expression at r → ∞) are Φξp , q = (Ep̂ , q exp(ξiqφ), Ĥ p , q exp(ξiqφ)) (here we use σ and
provided in the SI for TE and TM modes. ξ to label the propagation direction, i.e., σ = ±1 and ξ = ±1) as
Mode Expansion. As in the previous section, we assume 2π
that the modes are either TE- or TM-polarized. The total
electric field E can be decomposed into a set of outgoing
Φσm , n ⊗ Φξp , q = ∫0 exp[iφ(σn + ξq)]
+∞
guided cylindrical modes and a continuum of radiation modes: σ ξ ξ σ
dφ × ∫−∞ (Em̂ , n × Ĥ p , q − Ep̂ , q × Ĥ m , n)·
E(r , φ , z) = ∑ ∑ cm+,nEm̂ +,n(kmr , z) exp(inφ)
m n
n r r dz (6)

+ continuum (3) where ⊗ denotes the inner product operator. Equation 6 is also
defined for a vertical infinite cylindrical surface, as for eq 5. On
where c+m,n denotes the amplitude of the nth azimuthal harmonic the basis of the Lorentz reciprocity theorem, we establish the
of the mth outgoing mode for the polarization considered. The unconjugated form of mode orthogonality relation (see SI)
second term, “continuum”, in eq 3 corresponds to a summation
over the continuum of radiation modes of the stratified Φσm , n ⊗ Φξp , q = δσm , −ξpδn , qNm , n (σ = ±1, ξ = ±1)
waveguide; we provide some additional discussions in the SI (7)
about the orthogonality and normalization of radiation modes. From eq 3 by using the orthogonality relation of eq 7, we find
Asymptotic Behavior and Radiation Diagram. The
electric field of the mth outgoing mode can be expressed as a cm+, n = (E, H) ⊗ Φ−m , n
summation of all the relevant azimuthal harmonics, i.e., 2π
+ 1
+∞
Em(r , φ , z) = ∑n =−∞ cm+, nEm̂ , n(kmr , z) exp(inφ). The asymp- =
Nm , n
∫0 exp( −inφ) dφ
totic behavior at r → ∞ can be found with asymptotic forms of
+∞
Hankel functions, and for a TM mode, for instance, we have − −
∫−∞ (E × Ĥ m , n − Em̂ , n × H)·n rr dz
(8)
Em(r → ∞ , φ , z)
As the orthogonality and normalization relations are both r-
= fm (φ) exp(ikmr )/ r [eL(z)n r , 0nφ, e T(z)nz] (4) independent (see SI), eq 8 is r-independent as long as the
cylindrical surface, such as ΣA shown in Figure 2, fully
+∞
where fm (φ) = ∑n =−∞ cm+, n exp[−i(2n + 1)π /4] 2/(πkm) surrounds the inhomogeneity.
Nevertheless, eq 8 is not suitable for numerical implementa-
exp(inφ) tion, since it requires an integral over the entire transverse
describes the angular distribution of the amplitude of the mth direction, −∞ < z < +∞. Thus, we consider a closed source-f ree
mode in the far zone and eL(z) and eT(z) represent the volume Ω with its boundary Σ formed by three surfaces, ΣA, ΣB,
longitudinal and transverse electric field components of the z- and ΣC; see Figure 2. Applying (E,H) and Φ−m,n to the Lorentz
dependent mode profile (see SI). |f m(φ)|2 characterizes the reciprocity theorem (see SI or ref 27) for the volume Ω, a
angular intensity distribution in the far zone, i.e., the guided- simple but vital relation can be found as
− −
mode radiation diagram. ∮Σ exp(−inφ)(E × Ĥ m , n − Em̂ , n × H)·n dS = 0, owing to the
If the stack materials are lossless, the total power carried by fact that (E,H) and Φ−m,n share the same permittivity and
the mth mode can be found by summing the power of all permeability distribution in Ω. Furthermore, as we consider the
azimuthal harmonics as Pm = ∫ 2π 0 |f m(φ)| dφ = 4/|km|∑n=−∞|
2 +∞
mth mode to be a bound mode with a field that exponentially
c+m,n|2, provided that the modes are normalized in such a way as vanishes at z = ±∞,
Nm,n = (−1)n 16/km, with Nm,n denoting the normalization
coefficient of Φ±m,n defined as (see SI)
∮Σc exp(−inφ)(E × Ĥ m−,n − Em̂ −,n × H)·n dS = 0 (9)
2π +∞ + − − +
Nm , n = ∫0 dφ ∫−∞ (Em̂ , n × Ĥ m , n − Em̂ , n × Ĥ m , n)·n rr and we get
− −
dz (5) ∫Σ A
exp(−inφ)(E × Ĥ m , n − Em̂ , n × H)·n dS
The integral in eq 5 runs over a vertical cylindrical surface of − −
radius r with an infinite transverse cross-section from z = −∞ =− ∮Σ exp(−inφ)(E × Ĥ m , n − Em̂ , n × H)·n dS
(10)
to z = +∞ (e.g., surface ΣA in Figure 2), and noticeably it B

involves both the ingoing and the outgoing modes. It is The left-hand side of eq 10 is the same as the integral in eq 8,
independent of r even in the presence of absorption. A similar and therefore finally eq 8 can be replaced by
normalization approach for guided modes can be found in ref
1 − −
27.
Calculation of the Mode Amplitude. Here we will
cm+, n =
Nm , n
∮Σ B
exp(−inφ)(E × Ĥ m , n − Em̂ , n × H)·(−n)
+
demonstrate that the amplitude coefficient cm,n of the mode (11)
Φ+m,n in eq 3 can be found with a simple overlap integral [see dS
eq 11] involving (E,H) and the ingoing mode Φ−m,n on the Equation 11 constitutes the main result of the present analysis,
closed “box” ΣB defined in Figure 2. and it can be computed easily on a surface with a finite size (in
398 DOI: 10.1021/acsphotonics.5b00559
ACS Photonics 2016, 3, 395−402
ACS Photonics Article

Figure 3. Plane wave scattering by a silicon (n = 3.5) nanosphere (radius R = 285 nm) in air at a wavelength of 1.5 μm. (a) The silicon sphere is
illuminated by an x-polarized plane wave propagating along the +z direction. (b and c) Radiation diagrams of the scattered field in the x−y plane and
the y−z plane. Results obtained with the numerical tool and Mie theory are plotted with red curves and black circles.

Figure 4. Emission of two dipole sources inside a dielectric slab waveguide at a wavelength of 1 μm. (a) Sketch of the slab waveguide. The core layer
(thickness 200 nm and refractive index 1.5) is surrounded by two semi-infinite media with indices 1 (above) and 1.2 (below). The dipole coordinates
are (−100 nm, 0, 0) and (50 nm, 0, 0). (b) 3D free-space radiation diagram. The upper spherical (lower three-armed) part corresponds to radiation
into the superstrate (substrate). (c and d) Radiation diagrams of TE0 and TM0 guided modes. Results obtained with the numerical toolbox and
(semi)analytical calculation are shown with red curves and black circles, respectively.

contrast to eq 8). It implies that the amplitude of an outgoing polarized plane wave at a wavelength of 1.5 μm; see Figure 3a.
mode can be formally computed by sending its associated The radiation diagram of the scattered field is calculated with
ingoing mode into the stratified medium, as in the former COMSOL on a rectangular “box” that surrounds the
section (i.e., the retrieval approach for free-space radiation nanosphere and with the present numerical toolbox. The
diagram). We emphasize that the closed “box” ΣB that fully results are displayed with the red curves in Figure 3b and c and
encloses the local inhomogeneity has finite dimensions. Its are compared with radiation diagrams obtained with Mie
surface normal n is directed inward; see Figure 2. theory28 that are shown with black circles. The excellent
We remark that, because the r-dependent profiles are agreement between the fully numerical approach and the
depicted by Hankel functions that diverge at r = 0, both analytical solution corroborates the accuracy of the numerical
Φ+m,n and Φ−m,n diverge at r = 0; however their sum Φ+m,n + tool for retrieving the free-space radiation diagram.
Φ−m,n remains finite. According to eq 7, one can easily find that Example 2: Emission of Two Dipole Sources in a
in eqs 8 and 11 Φ−m,n can be replaced by Φ+m,n + Φ−m,n, i.e., cm,n
+
Dielectric Slab Waveguide. In example 2, we consider a
= (E,H) ⊗ (Φ+m,n + Φ−m,n). We emphasize that this more intricate problem, in which two electric dipole sources are
replacement, which is strictly rigorous from a mathematical placed in a high-index dielectric slab waveguide (see Figure 4a)
standpoint, is crucial for numerical calculations (additional deposited on a substrate with a low refractive index. The
details can be found in the SI). polarizations of the two electric dipole sources are J1 = (√2/2,

■ NUMERICAL TESTS
In this section, four examples are analyzed to show the
0, √2/2) and J2 = (0, 1, 0), respectively. The dipoles are
separated by 150 nm in the x direction. At the radiation
wavelength of 1 μm, the waveguide supports only two modes,
capability of the open-source numerical toolbox based on the TE0 (neff = 1.24) and TM0 (neff = 1.20), with all higher order
proposed retrieval approach. Additionally, the validity and modes being cut off. The near-field distribution on a
accuracy of the numerical tool are tested in examples 1 and 2, rectangular “box” that encloses the two sources is obtained
for which the radiation diagrams retrieved by the numerical tool with COMSOL and is used with the numerical toolbox to
from the near-field obtained with COMSOL multiphysics are compute the free-space and guided-mode radiation diagrams.
compared to diagrams obtained analytically. Examples 3 and 4 The results are shown in Figure 4b and 4c,d with red curves. As
present two emblematic examples in the area of nanoantennas. the refractive index of the substrate is larger than the one of the
Also, implementations of the examples 1 and 2 are provided in superstrate (air), light is dominantly radiated into the substrate.
ref 25. We additionally see that the radiation diagrams of TE and TM
Example 1: Scattering by a Silicon Nanosphere in modes are very different. We have also calculated the radiation
Free Space. We consider a silicon (n = 3.5) nanosphere of diagrams with an analytical expression for the field radiated by a
radius R = 285 nm placed in air and illuminated by an x- dipole source in a stratified medium using Green’s dyadics
399 DOI: 10.1021/acsphotonics.5b00559
ACS Photonics 2016, 3, 395−402
ACS Photonics Article

approaches.23,29,30 The results shown with black circles are in dominantly launched toward the x-direction, parallel to the
quantitative agreement with those obtained with the numerical incident polarization.
toolbox. Similar agreement has been obtained for the free space Example 4: Dipole Emission inside a Metal Patch
radiation diagram (not shown). In addition, we have checked Nanoantenna. Figure 6a−c depict a metal nanoantenna,
the energy conservation with a relative error of <1%, by
comparing the total power radiated by the source doublet
computed either directly by the total Poynting flux on a surface
surrounding the doublet or indirectly from the radiation
diagrams, summing over all free-space and guided modes.
The excellent agreement between the analytical calculations
(black circles) and the fully numerical results (red curves)
confirms the effectiveness of the present tool for treating guided
modes of the stratified substrate.
Example 3: Scattering at a Metallic Hole. The launching
of surface plasmon polaritons (SPPs) by metallic holes or slits
has been widely studied in plasmonics.15,16,18−20 In the third Figure 6. Dipole emission inside a metal nanoantenna. (a) 3D sketch
example, we consider the light scattering by a finite-depth of the nanoantenna. A silver nanocube (65 × 65 × 65 nm) with a 3 nm
rectangular metallic air hole etched in a semi-infinite metal polymer coating is placed above a gold substrate covered by a 5 nm
substrate (z < 0) and illuminated by a normally incident plane polymer (8 nm spacer in total). (b and c) Side and top views of the
wave; see the inset in Figure 5a. The hole creates a nanoantenna. A z-polarized electric dipole source shown with a red dot
lies midway between the silver cube and gold substrate and is located
near the edge of the cube. (d) Radiative (γrad/γtot) and plasmonic
(γSPP/γtot) quantum-yield spectra. The top-left inset (red solid) shows
the free-space radiation diagram in the x−z plane at λ0 = 650 nm. The
top-right inset (blue dashed) shows the SPP radiation diagram in the
horizontal x−y plane at the same wavelength. In simulation, the
refractive index of the polymer is 1.4; for the relative permittivities of
silver and gold, a Drude model εAg = 1 − ω2Ag/(ω2 − iωΓAg) and a
Drude−Lorentz model εAu = 8.1 − ω2Au/(ω2 − iωΓAu) − 0.7ω2L/(ω2 −
ω2L − 2iωΓL) are adopted, with ωAg = 1.3 × 1016 s−1, ΓAg = 1.6 × 1014
s−1, ωAu = 1.3 × 1016 s−1, ΓAu = 4.13 × 1013 s−1, ωL = 4.1 × 1015 s−1,
and ΓL = 4.7 × 1014 s−1.

composed of a silver nanocube placed on a polymer film coated

on a gold substrate and doped with a quantum emitter. Via the
Figure 5. Scattering by a rectangular metallic hole etched in a metal nanoantenna, the quantum emitter radiates into free space and
substrate. (a) The hole is illuminated by an x-polarized plane wave at launches SPPs on the dielectric−gold interface, with decay rates
λ0 = 700 nm impinging at normal incidence, as shown in the inset. denoted by γrad and γSPP, respectively. We further denote by γtot
Generation efficiencies of free-space scattered waves and SPPs as a
function of hole depth are plotted as black circles and red triangles,
the total decay rate. Nanocube antennas have been recently
respectively. (b) Free-space radiation diagram of the scattered wave in fabricated and tested, and extremely high Purcell factors and
the y−z plane (red dashed) and the x−z plane (green solid). (c) SPP large radiative efficiencies have been reported.10
radiation diagram in the x−y plane. The metal permittivity (gold) With COMSOL, we calculate the field radiated by the
taken for the simulation is εm = −15.83 + 1.28i at λ0 = 700 nm. quantum emitter (modeled as a vertical electric dipole) on a
box surrounding the nanocube and infer the radiation diagrams
into the SPP and free-space modes. We also calculate the total
power radiated by integrating the Poynting flux over a small
backscattered field in free space and launches SPPs at the box surrounding the dipole in the polymer layer. The SPP and
air−metal interface. With the present tool, we quantify the free-space emission efficiencies are shown in Figure 6d. The
amount of backscattered waves and of launched SPPs. The size computation is made over a spectral range covering the
of the hole is fixed as ax = 100 nm and ay = 400 nm. We define nanoantenna resonance at λ0 = 650 nm. The radiation diagrams
the term generation ef f iciency as the ratio of the scattered power at λ0 are shown in the insets. Note that at resonance, γrad/γtot is
(of either free-space waves or SPPs) to the power incident over about 20%. A larger value (∼50%) was predicted in ref 10, but
the hole area (ax × ay). The efficiencies are displayed in Figure we note that the authors in ref 10 are not equipped with the
5a, as a function of the hole depth. present numerical tool and that they infer γrad from the
For the present example, the scattered electromagnetic field difference between the γtot and the nonradiative decay rate γnrad
on the “box” that encloses the hole is obtained using an computed as the Joule loss inside the finite computational
aperiodic Fourier modal method.31,32 Quantitative justification domain. Such an approach overestimates γrad, especially if SPPs
of the retrieved result by comparison with already known that carry energy far away from the nanoantenna are efficiently
results is not made in this case, due to the lack of available launched, like in the present case. We believe that this
solutions for the same problem in the literature. However, the highlights the importance of the present tool. Also, we remark
response of subwavelength metallic holes can be approximately that, as there is a fairly large portion of radiation that is directed
described as that of an effective magnetic dipole my,33 and this into SPPs on the dielectric−metal interface, this type of device
approximation shows good agreement with the free-space and may be considered a good candidate for plasmon source.6,34,35
SPP radiation diagrams shown in Figure 5b and c. The SPPs are A detailed study of the decay of quantum emitters in the
400 DOI: 10.1021/acsphotonics.5b00559
ACS Photonics 2016, 3, 395−402
ACS Photonics Article

different decay channels (γtot, γrad, and γSPP) in such a device can (3) Mode orthogonality and normalization. (4) Remov-
be found in ref 36 as a function of the gap thickness, together ing the singularity of Hankel functions at r = 0. (5)
with a thorough analysis of why quenching does not impose Relation with the field equivalence principle (PDF)

■
severe limitations on the quantum efficiency.

■ CONCLUSION
We have developed a general method for computing far-field
AUTHOR INFORMATION
Corresponding Authors
*E-mail: jianji.yang.photonique@gmail.com.
free-space and guided-mode radiation diagrams of local *E-mail: jean-paul.hugonin@institutoptique.fr.
inhomogeneities in stratified media. The far-field retrieval *E-mail: philippe.lalanne@institutoptique.fr.
requires only the knowledge of the near-zone electromagnetic Present Address
field (obtained by a Maxwell’s equation solver) on a closed #
Department of Electrical Engineering, Stanford University,
surface that envelops the local inhomogeneity. Moreover, the Stanford, California 94305, United States.
retrieval is based on rigorous mode decomposition. The
Notes
accuracy of the tool has been corroborated by comparisons
The authors declare no competing financial interest.

■
between fully numerical results and analytical results for two
simple examples, and its range of applicability has been further
ACKNOWLEDGMENTS
illustrated with two examples taken from the recent literature
on nanoplasmonics. Additionally, we have already successfully Part of the study was carried out with financial support from
applied this numerical tool to various problems, including light “the Investments for the Future” Programme IdEx Bordeaux-
extraction with optimized nanoparticles,5 broadband absorb- LAPHIA (ANR-10-IDEX-03-02).
ers,37 and emission in metal nanogap structures36 and metallic
patch antennas.38 Thus, we are confident that the tool can be
widely used for various light emission and scattering problems
■ REFERENCES
(1) Ferry, V. E.; Sweatlock, L. A.; Pacifici, D.; Atwater, H. A.
in photonics and is compatible with virtually any Maxwell’s Plasmonic nanostructure design for efficient light coupling into solar
equation solver. cells. Nano Lett. 2008, 8, 4391.
Technical Remarks. This section briefly gathers some (2) Atwater, H. A.; Polman, A. Plasmonics for improved photovoltaic
devices. Nat. Mater. 2010, 9, 205.
information on the usage of the numerical tool. (1) Retrieval (3) Gu, X.; Qiu, T.; Zhang, W.; Chu, P. K. Light-emitting diodes
procedure: To retrieve a free-space or guided-mode radiation enhanced by localized surface plasmon resonance. Nanoscale Res. Lett.
diagram with the Matlab-based open-source software, before- 2011, 6, 199.
hand one has to calculate the radiated or scattered electro- (4) Zhmakin, A. I. Enhancement of light extraction from light
magnetic field with any Maxwell’s equation solver on a emitting diodes. Phys. Rep. 2011, 498, 189−241.
rectangular “box” that f ully encloses the local inhomogeneity. (5) Jouanin, A.; Hugonin, J.-P.; Besbes, M.; Lalanne, P. Improved
Users have to set the refractive index and thickness of each light extraction with nano-particles offering directional radiation
layer (the online version handles nonmagnetic materials only). diagrams. Appl. Phys. Lett. 2014, 104, 021119.
The numerical tool and its user guide are provided in ref 25. (6) Liu, Y.; Palomba, S.; Park, Y.; Zentgraf, T.; Yin, X.; Zhang, X.
(2) Guided-mode radiation diagram: Waveguide-mode decom- Compact magnetic antennas for directional excitation of surface
plasmons. Nano Lett. 2012, 12, 4853.
position requires the profiles of the guided modes in the (7) Staude, I.; Miroshnichenko, A. E.; Decker, M.; Fofang, N. T.; Liu,
stratified media, which are found by a mode solver included in S.; Gonzales, E.; Dominguez, J.; Luk, T. S.; Neshev, D. N.; Brener, I.;
the open-source software. In addition, users should provide the Kivshar, Y. Tailoring directional scattering through magnetic and
modal effective index neff (potentially a c-number) and set the electric resonances in subwavelength silicon nanodisks. ACS Nano
polarization of the desired mode. An accurate value of neff is not 2013, 7, 7824.
required, since it is computed precisely by the mode solver. (8) Coenen, T.; Arango, F. B.; Koenderink, A. F.; Polman, A.
Note that the neff of the cylindrical waveguide mode of a 3D Directional emission from a single plasmonic scatterer. Nat. Commun.
stratified structure is the same as that of the 2D stratified 2014, 5, 3250.
structure (see SI). (3) CPU time: The open-source code is (9) Hancu, I. M.; Curto, A. G.; Castro-López, M.; Kuttge, M.; van
rather efficient. For instance, based on an ordinary workstation Hulst, N. F. Multipolar interference for directed light emission. Nano
Lett. 2014, 14, 166.
with a processor main frequency of 1.80 GHz, for the example (10) Akselrod, G. M.; Argyropoulos, C.; Hoang, T. B.; Ciracì, C.;
1 or 2 in this paper, the retrieval of the free-space radiation Fang, C.; Huang, J.; Smith, D. R.; Mikkelsen, M. H. Probing the
diagram (including both upper and lower free space) consumes mechanisms of large Purcell enhancement in plasmonic nanoantennas.
about 100 s, and it costs just 2 s to compute the radiation Nat. Photonics 2014, 8, 835.
diagrams of both the TE and TM modes for the example 2. (11) Biagioni, P.; Huang, J. S.; Hecht, B. Nanoantennas for visible
CPU time is directly determined by the number of sample and infrared radiation. Rep. Prog. Phys. 2012, 75, 024402.
points on the “box” and the number of far-field directions that (12) Yu, N.; Genevet, P.; Kats, M. A.; Aieta, F.; Tetienne, J.-P.;
are computed. Capasso, F.; Gaburro, Z. Light propagation with phase discontinuities:

■
generalized laws of reflection and refraction. Science 2011, 334, 333−
337.
ASSOCIATED CONTENT (13) Lin, D.; Fan, P.; Hasman, E.; Brongersma, M. L. Dielectric
* Supporting Information
S gradient metasurface optical elements. Science 2014, 345, 298.
The Supporting Information is available free of charge on the (14) Zheng, G.; Mühlenbernd, H.; Kenney, M.; Li, G.; Zentgraf, T.;
ACS Publications website at DOI: 10.1021/acsphoto- Zhang, S. Metasurface holograms reaching 80% efficiency. Nat.
nics.5b00559. Nanotechnol. 2015, 10, 308−312.
(15) Lalanne, P.; Hugonin, J. P.; Liu, H. T.; Wang, B. A microscopic
(1) Unconjugated form Lorentz reciprocity theorem. (2) view of the electromagnetic properties of subwavelength metallic
Expressions of the cylindrical waveguide mode profiles. surfaces. Surf. Sci. Rep. 2009, 64, 453−469.

401 DOI: 10.1021/acsphotonics.5b00559

ACS Photonics 2016, 3, 395−402
ACS Photonics Article

(16) Liu, H.; Lalanne, P.; Yang, X. Y.; Hugonin, X. Y. Surface

plasmon generation by subwavelength isolated objects. IEEE J. Sel.
Top. Quantum Electron. 2008, 14, 1522−1529.
(17) Yao, W.; Liu, S.; Liao, H.; Li, Z.; Sun, C.; Chen, J.; Gong, Q.
Efficient directional excitation of surface plasmons by a single-element
nanoantennas. Nano Lett. 2015, 15, 3115.
(18) Kihm, H. W.; Koo, S. M.; Kim, Q. H.; Bao, K.; Kihm, J. E.; Bak,
W. S.; Eah, S. H.; Lienau, C.; Kim, H.; Nordlander, P.; Halas, N. J.;
Park, N. K.; Kim, D.-S. Bethe-hole polarization analyser for the
magnetic vector of light. Nat. Commun. 2011, 2, 451.
(19) Rotenberg, N.; Spasenović, M.; Krijger, T. L.; le Feber, B.;
García de Abajo, F. J.; Kuipers, L. Plasmon scattering from single
subwavelength holes. Phys. Rev. Lett. 2012, 108, 127402.
(20) Coenen, T.; Polman, A. Optical properties of single plasmonic
holes probed with local electron beam excitation. ACS Nano 2014, 8,
7350.
(21) Balanis, C. A. Antenna Theory Analysis and Design, 3rd ed.;
Wiley-Interscience, 2005.
(22) Demarest, K.; Huang, Z.; Plumb, R. An FDTD near- to far-zone
transformation for scatterers buried in stratified grounds. IEEE Trans.
Antennas Propag. 1996, 44, 1150.
(23) Pors, A.; Bozhevolnyi, S. I. Quantum emitters near layered
plasmonic nanostructures: decay rate contributions. ACS Photonics
2015, 2, 228.
(24) Zenneck, I. Propagation of plane electromagnetic waves along a
plane conducting surface and its bearing on the theory of transmission
in wireless telegraphy. Ann. Phys. 1907, 23, 846.
(25) The toolbox will soon be available at www.lp2n.institutoptique.
fr/Membres-Services/Responsables-d-equipe/LALANNE-Philippe
and can be obtained on request by then.
(26) Yeh, P. Optical Waves in Layered Media; J. Wiley and Sons: New
York, 1988.
(27) Snyder, A. W.; Love, J. Optical Waveguide Theory; Springer:
USA, 1983.
(28) Bohren, C. F.; Huffman, D. R. Absorption and Scattering of Light
by Small Particles; Wiley, 1983.
(29) Sipe, J. E. New Green-function formalism for surface optics. J.
Opt. Soc. Am. B 1987, 4, 481−489.
(30) Paulus, M.; Martin, O. J. F. Green’s tensor technique for
scattering in two-dimensional stratified media. Phys. Rev. E: Stat. Phys.,
Plasmas, Fluids, Relat. Interdiscip. Top. 2001, 63, 066615.
(31) Silberstein, E.; Lalanne, P.; Hugonin, J.-P.; Cao, Q. Use of
grating theories in integrated optics. J. Opt. Soc. Am. A 2001, 18, 2865.
(32) Hugonin, J.-P.; Lalanne, J.-P. Perfectly matched layers as
nonlinear coordinate transforms: a generalized formalization. J. Opt.
Soc. Am. A 2005, 22, 1844.
(33) Jackson, J. D. Classical Electrodynamics; John Wiley & Sons Ltd.,
1962.
(34) Falk, A. L.; Koppens, F. H. L.; Yu, C. L.; Kang, K.; de Leon
Snapp, N.; Akimov, A. V.; Jo, M.-H.; Lukin, M. D.; Park, H. Near-field
electrical detection of optical plasmons and single-plasmon sources.
Nat. Phys. 2009, 5, 475−479.
(35) Gan, C. H.; Hugonin, J.-P.; Lalanne, P. Proposal for compact
solid-state III-V single-plasmon sources. Phys. Rev. X 2012, 2, 021008.
(36) Faggiani, R.; Yang, J.; Lalanne, P. Quenching, plasmonic, and
radiative decays in nanogap emitting devices. ACS Photonics 2015, 2,
1739.
(37) Langlais, M.; Hugonin, J.-P.; Besbes, M.; Ben-Abdallah, P.
Cooperative electromagnetic interactions between nanoparticles for
solar energy harvesting. Opt. Express 2014, 22, A577.
(38) Bigourdan, F.; Hugonin, J.-P.; Lalanne, P. Aperiodic-Fourier
modal method for analysis of body-of-revolution photonic structures. J.
Opt. Soc. Am. A 2014, 31, 1303.

402 DOI: 10.1021/acsphotonics.5b00559

ACS Photonics 2016, 3, 395−402
 Supporting Information for
“Near-to-far field transformations for radiative and guided waves”

Jianji Yang1, Jean-Paul Hugonin2 and Philippe Lalanne1

1 Laboratoire Photonique Numérique et Nanosciences (LP2N), Institut d‟Optique d‟Aquitaine,

Université Bordeaux, CNRS, 33400 Talence, France.
2 Laboratoire Charles Fabry, Institut d‟Optique Graduate School, CNRS, Univ. Paris-Sud, 2 Avenue
Augustin Fresnel, 91127 Palaiseau, France
E-mail: jianji.yang.photonique@gmail.com
jean-paul.hugonin@institutoptique.fr
philippe.lalanne@institutoptique.fr

Content

1. Unconjugated form Lorentz reciprocity theorem

2. Expressions of the cylindrical waveguide mode profiles
3. Mode orthogonality and normalization
4. Removing the singularity of Hankel functions at r = 0
5. Relation with the field equivalence principle
1. Unconjugated form of Lorentz reciprocity theorem
A key basis of the present work is the unconjugated form of Lorentz reciprocity theorem [1], which is
repeatedly used in the work and hence derived here. Assuming the stack materials are reciprocal, we
consider two solutions (labeled by the subscripts 1 and 2) of Maxwell‟s equations,
  E1  i11H1  M1 and   H1  i11E1  J1 , (S1a)
  E 2  i2  2H2  M2 and   H2  i2  2E 2  J2 . (S1b)
By applying the divergence theorem to E 2  H1  E1  H2 , the reciprocity theorem can be found as

 E2  H1  E1  H2   dS
, (S2a)
  iH1  H2  22  11   iE1  E2 11   22   H1  M2  E2  J1  H2  M1  E1  J2 dV

where  denotes a surface enclosing the domain  , which contains the source terms (J1, J2, M1,
and M2). If the domain  is source-free, Eq. (S2a) becomes

 E2  H1  E1  H2   dS  0 . (S2b)

2. Expression of the waveguide mode profiles in cylindrical coordinates

Generally, when solving the source-free Maxwell‟s equations of a stratified waveguide in a cylindrical
coordinate r , ,z  , the mode profile can be casted into independent functions of z  ,   , and r  .
The z-dependent profiles correspond to the usual mode profiles obtained in Cartesian coordinates
that are derived in all textbooks with an invariance along one of the longitudinal directions , see
Fig. S1. Due to the revolution symmetry, the  -dependence is simply exp in  (n = 0, ±1 …). And
finally, the r-dependence is characterized by Hankel functions. The propagation constant (along the
radial direction r) is the same as the one obtained in Cartesian coordinate.

Figure S1. Field components of the modes of a stratified waveguide in Cartesian

coordinates (all fields components are assumed to be independent of the coordinate along
the direction perpendicular to the figure). (a) and (b) TM and TE modes. For a TM mode,
the field component triplet is (eT, eL, hT); and for a TE mode, the triplet is (eT, hL, hT). „T‟ and
„L‟ label the transverse and longitudinal components. k indicates the propagation constant
and the mode profiles are only z-dependent. All details can be found in Ref. [2] for instance.
 Throughout the work, we denote the nth azimuthal harmonic of the mth outgoing (labeled by “+”)

and ingoing (labeled by “–”) waveguide modes as 

  m,n  Eˆ m ,n expin,H 
ˆ  expin
m ,n and


  m,n  Eˆ m ,n expin,H 
ˆ  expin , where Ê  ( Ĥ ) depends on z and r.
m ,n m, n m, n

The explicit expressions of TM- and TE-mode profiles and their asymptotic expressions as
r   are provided in the following Eqs. (S3a)-(S3d), where we drop exp in for the sake of

simplicity and organize the field components with the usual sequence e r , e  , e z ,hr , h , hz  .

(a) TM outgoing modes

i
  
 
 2 H n 1 k m r   H n 1 k m r  e L z  
1   e L z 
 
 H n1 k m r   H n1 k m r  e L z    0 
 
2 
Eˆ m ,n   H n k m r eT z   2  2n  1  eT z 
ˆ      
r 
exp i  k m r     , (S3a)
Hm,n    1 H  k r   H  k r  h z  k m r
 2
 n 1 m n 1 m T

 4   0 
hT z 
 i   
  

 2 H n 1 k m r   H n 1 k m r  hT z    0 
 
 0 
(b) TM ingoing modes
i
  
 
 2 H  n 1 k m r   H  n 1 k m r  e L z  
1    e L z 
 
 H n 1 k m r   H n 1 k m r  e L z    0 
 
2 
Eˆ m ,n   H n k m r eT z   2   2n  1   eT z  
ˆ      
r 
exp   i  k m r     , (S3b)
Hm,n    1 H  k r   H  k r  h z  k m r
 2
  n 1 m  n 1 m  T

  4   0 
  hT z 
 i   
  

 2 H  n 1 k m r   H  n 1 k m r  hT z    0 
 
 0 
(c) TE outgoing modes
1 
 
 2 H n 1 k m r   H n 1 k m r  eT z 
 
 i   0 
 
 H n1 k m r   H n1 k m r  eT z   e z 
 T 
 2 
Eˆ m ,n   0  2  2n  1   0 
ˆ       exp i  k m r     , (S3c)
   hL z 
r 
Hm,n   i H  k r   H  k r  h z   k m r  4
n 1 n 1
2
m m L
  0 
 

  
 2 H n 1 k m r   H n 1 k m r  hL z  
1  
hT z 
 H n k m r hT z  
 
(d) TE ingoing modes
 1 
 
 2 H  n 1 k m r   H  n 1 k m r  eT z 
 
 i   0 
 
 H n 1 k m r   H n 1 k m r  eT z     e z 
 T 
 2 
Eˆ m ,n   0  2   2n  1   0 

ˆ       exp   i  k m r     . (S3d)
     hL z 
r 
Hm,n   i H n 1 k m r   H n 1 k m r  hL z   k m r   4
2   0 
 

1  
 
 2 H  n 1 k m r   H  n 1 k m r  hL z    hT z  
 H n k m r hT z  
 

In Eqs. (S3a)-(S3d), e L z  and eT z  h z  and

L hT z  represent the longitudinal (labeled by “L”)
and transverse (labeled by “T”) electric (magnetic) components of the z-dependent mode profiles,
which correspond to the classical mode profiles calculated in Cartesian coordinates, see Fig. S1;
H n k m r  and H n k m r  denote the Hankel function of the first and second kind of order n. The
asymptotic expressions are derived using the asymptotic forms of Hankel functions [3].

3. Unconjugated form of mode orthogonality and normalization

The mode decomposition technique formulated in the main text relies on mode orthogonality and
normalization relations that are established here.
We consider two cylindrical modes, the nth azimuthal harmonic of the mth mode

 
ˆ  expin and the qth azimuthal harmonic of the pth mode 
 m,n  Eˆ m,n ,H m,n p,q  E p,q , H p,q exp iq ,
ˆ ˆ  
with   1 and   1 dictating the mode propagation direction (with „+‟ and „–‟ corresponding to
outgoing and ingoing modes, respectively).
We define the inner product of two modes as
2  

 m,n   p,q  0  expin  q  Eˆ m,n  H
ˆ   Eˆ   H
p,q p,q m,n r 
ˆ   n rdzd , (   1 ,   1 ) (S4)

where  denotes the inner product operator and nr is the normalized radial vector of the cylindrical
basis. Note that the integral in Eq. (S4) runs over a cylindrical surface of radius r from z   to
z   , such as the surfaces  1 or  2 sketched in Fig. S2. If two modes have different polarizations,
they are orthogonal to each other because the cross-products are null. In Section 3.2, we additionally
show the orthogonality relation for two modes with the same polarization

 m,n   p,q   m, p  n,q Nm,n , (S5)

where N m,n denotes the normalization coefficient of mode   m,n (see Section 3.3)

Nm,n    m,n   m,n   1

n 8 
 
 e m,T  hm,T  nr dz .
k m 
(S6)
 Figure S2. Sketch of a cylindrical-shell domain Ω. (a) Domain  and its outer

surface  1 (green), inner surface  2 (red); the surface  3 (blue) is situated at z   .

(b) Cross-sectional view. Cylindrical surfaces  1 (r = r1, vertical green lines) and  2
(r = r2, vertical red lines) are the outer and inner surfaces of the compact domain  .  3
(shown with four horizontal blue lines) presents the horizontal boundaries of  at

z   .

3.1 Radial independence of Eqs. (S5) and (S6) for guided modes
In this sub-section, we will demonstrate the r-independence property of Eqs. (S5) and (S6); this
independence is important for establishing the mode orthogonality and normalization in Section 3.2.

Yet we consider two modes,  m,n and  p,q (   1 and   1 , labeling the propagating

direction), and a cylindrical-shell volume  shown in Fig. S2. Since modes are solution of

source-free Maxwell‟s equations, we can inject  m,n [as solution 1 (E1,H1)] and  p,q [as solution 2

(E2,H2)] into Eq. (S2b) (Lorentz reciprocity theorem) for the source-free domain  . We easily obtain

1  2  3 E2  H1  E1  H2   dS  0 . Furthermore, if the modes are truly guided modes that are

exponentially damped at z   , the integral over 3 is null and hence we obtain

1 E2  H1  E1  H2   dS    2 E 2  H1  E1  H2   dS , (S7)

the minus sign resulting from the orientation of the normal vectors of the surfaces. According to
Eq. (S4), it is straightforward to find that

 m,n   p,q  1 E 2  H1  E1  H2   dS , (   1 and   1 ) (S8a)

r1

 m,n   p,q    2 E 2  H1  E1  H2   dS , (S8b)

r2

where  m,n   p,q r and  m,n   p,q r represent inner products performed at cylindrical surfaces
1 2

r = r1 and r = r2. By substituting Eqs. (S8a) and (S8b) into Eq. (S7), we find that
 θm,n   p,q   σm,n   p,q , (S9)
r1 r2

indicating that the orthogonality and normalization relations given by Eqs. (S5) and (S6) are
r-independent.

3.2. Mode orthogonality relation

As  m,n   p,q (   1 and   1 , labeling the propagating direction) is r-independent, for

simplicity we may derive the mode orthogonality relation Eq. (S5) by considering r   . Referring to

Eqs. (S3a)-(S3d), we calculate  m,n   p,q for r   using the asymptotic expressions for the

modes

 m,n   p,q 
2
 
exp i k m  k p r exp i 2n  q       4 
 kmk p . (S10)
0 expin  q d e m,T  h p,T  e p,T  hm,T   nr dz
2 

To show that Eq. (S10) leads to Eq. (S5), we consider several cases:
(a) Different mode polarization: From the asymptotic expressions of TM and TE modes, one easily

finds Eˆ TE ˆ TM ˆ TM ˆ TE
m,n  H p,q  E p,q  Hm,n  0 . Therefore,  TE
m,n   p,q  0
TM
for two modes with different

polarizations.

(b) Different azimuthal indices: For n  q , 0 expin  q d  0 and hence  m,n   p,q  0 .
2

(c) Different propagation constant (km ≠ kp): Note that in Eqs. (S3a)-(S3d), the mth outgoing (   1 )
and ingoing (   1 ) modes have the same value of propagation constant km, but their propagation
directions (i.e., being outgoing or ingoing) is discriminated by the notation „±‟ in front of km. When

km ≠ kp, a simple relation   e m,T  hp,T  e p,T  hm,T   nr dz  0 can be found [4], which is the


unconjugated form of mode orthogonality relation for planar waveguides in a Cartesian coordinate,

and it ensures  m,n   p,q  0 .

(d) The same value of propagation constant km = kp and the same propagation direction (i.e.,    ):
In this case, we have

  e m,T  hp,T  e p,T  hm,T   nr dz    e p,T  hp,T  e p,T  hp,T   nr dz  0 ,
 

and it again ensures  m,n   p,q  0 .

3.3. Mode normalization

  
ˆ  exp in , we use the counter-propagative mode
To normalize an outgoing mode   m,n  Eˆ m ,n , H m, n

 
ˆ  exp in . By substituting the two modes into Eq. (S5), we find that
  m,n  Eˆ m ,n , H m, n

  m,n    m,n   1
n 1 8 
 
 em,T  hm,T  nr dz .
km  
(S11)

It is convenient to normalize the modes such that

  e m,T  hm,T   nr dz  2 ,

(S12)

If the planar waveguide is non-lossy [i.e., Im(km) = 0], a planar mode normalized according to Eq. (S12)

carries a unitary Poynting flux, i.e.

1
2

  
Re   e m,T  h*m,T  nr dz  1 .

Injecting Eq. (S12) into Eq. (S11),  m,n is normalized as

Nm,n    m,n    m,n  16  1 k m .

n
(S13)

3.4 Remark on the orthogonality and normalization of radiation modes

In the former sub-sections, mode orthogonality and mode normalization have been established for
guided modes, which are of main interest in practice. Here, we make a short remark on the extension
of orthogonality and normalization issue to radiation modes. According to Eq. (3) in the main text, the
total field is expanded into the waveguide mode basis that includes both a discrete set of truly guided
modes (bound states) and a continuum of radiation modes. Therefore, mode orthogonality has to be
established for all the modes, especially the orthogonality between bound and continuum states
should be established. Such orthogonalities can be derived directly from Maxwell‟s equations with a
formalism based on the Lorentz reciprocity theorem. The formalism and its difficulties to handle
leakage of continuum states are documented in textbooks of optical waveguides [4]. Moreover it was
generalized to periodic waveguides in [5], by introducing complex spatial coordinate transforms [6] ─ a
sort of perfectly matched layer (PML) ─ that map the open problem with its associated continuum of
radiation states to an approximated closed problem with a countable number of discrete states, called
quasi-normal Bloch modes. For brevity, the introduction of complex spatial coordinate transforms
allows one to handle the orthogonality and normalization issues for both guided and radiation modes
of stratified waveguides that is treated in the current work. The PML-assisted approach has been
effectively applied in several theoretical studies on waveguides, such as the computation of scattering
loss at waveguide terminations [7], the coupling of quantum emitters with photonic crystal waveguides
[8], and the localization lengths of periodic waveguides with tiny imperfections [9].

4. Removing the singularity of Hankel functions at r = 0

Hankel function H n r   J n r   iYn r  is a linear combination of Bessel functions of the first J n r  and
second Yn r  kinds [3]. J n r  is finite at r = 0, however Yn r  diverges at r = 0 and therefore the
cylindrical modes (including both the outgoing mode   m,n and ingoing mode   m,n ) expressed by

Hankel functions have singularity at r = 0. This results in difficulty to numerically implement the
calculations of Eqs. (8)-(11) in the main text. Note that, H n r   H n r   2J n r  is finite and therefore

  m,n    m,n does not diverge at r = 0. According to the mode orthogonality relation given by Eq. (7)

in the main text or Eq. (S5), one can easily find   m,n    m,n  0 and then

 
N m, n    m , n    m, n    m, n    m, n    m, n . (S14)

So in Eq. (8) of the main text,   m,n can be replaced by   m,n   m,n to compute the excitation

coefficient, i.e., c m ,n  E, H   m,n  E, H    m,n   m,n  . Furthermore, this important replacement is

applied into Eqs. (8)-(11) of the main text to avoid any singularity in numerical calculations.

5. Relation with the field equivalence principle

The field equivalence principle classically used for NFFT [1] has not been used explicitly in the present
derivations. Nevertheless, it is helpful for understanding the proposed methods, e.g., Eqs. (2) and (11)
in the main text, to establish the link between the present method and the field equivalent principle.
Let‟s take Eq. (11) as an example, with some simple vector identities, one can rewrite Eq. (11) as

cm ,n 
1
Nm,n B


 exp in Eˆ m,n  J  Hˆ m,n  M dS ,

 (S15)

with electric surface current J  n  H and magnetic surface current M  n  E . Converting Eq. (11)
to Eq. (S15) amounts to casting E, H on  B into fictitious sources according to the field equivalent
principle. Note that the amplitude of an outgoing mode excited by local sources is determined by the
corresponding ingoing mode and the same sources, as the modes conforms to the unconjugated form
of orthogonality, see for instance the dipole emission in a periodic waveguide in Ref. [5].

Reference
[1] C. A. Balanis, Antenna Theory Analysis and Design, 3rd edition, Wiley-Interscience, 2005.
[2] P. Yeh, Optical Waves in Layered Media, J. Wiley and Sons eds., New York, 1988.
[3] G. Arfken, Mathematical Methods for Physicists, 3rd editon, Academic Press, 1985.
[4] A.W. Snyder, J. Love, Optical waveguide theory, Springer US, 1983.
[5] G. Lecamp, J. P. Hugonin, and P. Lalanne, ʹʹTheoretical and computational concepts for periodic
optical waveguidesʹʹ, Opt. Express 15, 11042 (2007).
[6] J. P. Hugonin and P. Lalanne, ʹʹPerfectly matched layers as nonlinear coordinate transforms: a
generalized formalizationʹʹ, J. Opt. Soc. Am. A 22, 1844 (2005)
[7] C. Sauvan, P. Lalanne, and J. P. Hugonin, ʹʹSlow-wave effect and mode-profile matching in
photonic crystal microcavitiesʹʹ, Phys. Rev. B 71, 165118 (2005).
[8] G. Lecamp, P. Lalanne, and J. Hugonin, ʹʹVery large spontaneous-emission β factors in
photonic-crystal waveguidesʹʹ, Phys. Rev. Lett. 99, 023902 (2007).
[9] A. Baron, S. Mazoyer, W. Smigaj, and P. Lalanne, ʹʹAttenuation coefficient of single mode periodic
waveguidesʹʹ, Phys. Rev. Lett. 107, 153901 (2011).