---
title: Directional Waveguide–Emitter Coupling Interface
url: https://www.emergentmind.com/topics/directional-waveguide-emitter-coupling-interface
type: topic
---

# Directional Waveguide–Emitter Coupling Interface

Directional waveguide–emitter coupling interface denotes a class of light–matter interfaces in which a localized emitter couples asymmetrically to the two counterpropagating modes of a waveguide, so that spontaneous emission, scattering, absorption, or collective decay become direction-selective. In the literature, this asymmetry is realized through non-transversal guided fields and spin–orbit interaction of light, mirror-symmetry breaking and evanescent angular spectra, multi-emitter interference at fixed propagation phase, and engineered antenna or cavity geometries. Representative platforms include vacuum-clad silica nanofibers with atoms, glide-plane photonic-crystal waveguides with quantum dots, quarter-wavelength-spaced superconducting qubits in microwave transmission lines, planar TiO$_2$ waveguides assisted by Yagi–Uda antennas, and wavelength-scale deformed microdisks [1406.0896][1406.4295][2203.01430][2402.13557][1110.3344].

## 1. Electrodynamic definition and symmetry structure

A standard formulation resolves the spontaneous-emission rate of an emitter with dipole operator $\mathbf d$ at position $\mathbf r_0$ into the two counterpropagating guided modes,
$$
\Gamma_+ = \frac{2\pi}{\hbar}\,|\mathbf d\cdot \mathbf E_+(\mathbf r_0)|^2\,\rho_+(\omega_0),\qquad
\Gamma_- = \frac{2\pi}{\hbar}\,|\mathbf d\cdot \mathbf E_-(\mathbf r_0)|^2\,\rho_-(\omega_0),
$$
or, equivalently,
$$
\Gamma_{\pm}=\frac{2\omega_0^2}{\hbar\epsilon_0 c^2}\,\mathbf d^{T}\cdot \mathrm{Im}\,G_{\pm}(\mathbf r_0,\mathbf r_0;\omega_0)\cdot \mathbf d^*.
$$
In tightly confined guides, $\mathbf E_{\pm}(\mathbf r_0)$ is neither purely transverse nor symmetric under $z\rightarrow -z$, and the local polarization ellipticity depends on propagation direction. In the nanofiber HE$_{11}$ mode, the electric spin density
$$
S_e(\mathbf r)=\frac{\epsilon_0}{4\omega}\,\mathrm{Im}\!\left[\mathbf E^*(\mathbf r)\times \mathbf E(\mathbf r)\right]
$$
reverses sign when the propagation direction is reversed, which allows a dipole of given helicity to couple preferentially to one direction [1406.0896].

A compact measure of asymmetry is the directionality parameter
$$
D=\frac{P_+-P_-}{P_++P_-}\in[-1,+1].
$$
A symmetry-based treatment shows that, for emitters near a waveguide invariant under mirror operations, directionality is mostly due to mirror-symmetry breaking caused by the axial character of the angular momentum of the emitted light. For a centered emitter with both relevant mirror symmetries, $D=0$ at $m_z=0$, while for single-mode planar waveguides the dependence takes the explicit form
$$
D(m_z)=\tanh\!\bigl(m_z\,F\,\ln 10\bigr),
$$
with the exponential dependence traced to a property of the evanescent angular spectrum [1910.11228]. This establishes that the sign of the angular momentum along an axis transverse to the waveguide determines the preferential coupling direction, while helicity-dependent effects can appear when displacement breaks another mirror symmetry.

## 2. Physical mechanisms of directional coupling

One mechanism is local spin–orbit interaction of light in subwavelength confinement. In the nanofiber geometry of Mitsch et al., a circular dipole $\mathbf d\propto (\hat{\mathbf e}_x\mp i\hat{\mathbf e}_z)/\sqrt2$ can couple nearly exclusively to one propagation direction, whereas a $\pi$ transition $\mathbf d\propto \hat{\mathbf e}_y$ couples equally to both directions because the HE$_{11}$ modes are always $\pi$-polarized along $y$ [1406.0896]. Closely related behavior underlies chiral photonic-crystal waveguides, where time-reversal symmetry implies $\mathbf E_{-k}(\mathbf r)=\mathbf E_k^*(\mathbf r)$, so a $\sigma^+$ dipole overlaps strongly with one guided mode and weakly with its counterpropagating partner [1406.4295].

A second mechanism is interference between spatially separated emitters. For two qubits separated by $\lambda/4$, the right- and left-going output fields can be written as
$$
b_{\mathrm{out}}^\rightarrow(t)=-i\sqrt{\gamma(t)/2}\,[L_1 e^{+i\pi/4}- iL_2 e^{-i\pi/4}],
$$
$$
b_{\mathrm{out}}^\leftarrow(t)=-i\sqrt{\gamma(t)/2}\,[L_1 e^{-i\pi/4}- iL_2 e^{+i\pi/4}],
$$
and the off-diagonal coherences produce destructive interference in one direction and constructive interference in the other. Preparing
$$
|\psi_\pm\rangle=\frac{1}{\sqrt2}\left(|eg\rangle+e^{\pm i\pi/2}|ge\rangle\right)
$$
yields purely right- or left-going emission in the idealized lossless case [2004.01924].

A third mechanism uses antenna-mediated phase engineering. In a planar TiO$_2$ waveguide, a gold Yagi–Uda antenna consisting of a feed rod, one reflector and three directors is designed so that scattering from its elements acquires an approximately $\pi/2$ phase difference. Constructive interference on the forward side and destructive interference on the backward side then yield in-plane directivity into the guided TE$_0$ mode [2402.13557].

A fourth mechanism is wave-optical splitting of clockwise and counter-clockwise pseudo-orbits in wavelength-scale resonators. In a deformed microdisk, the Goos–Hänchen shift
$$
\Delta_{\rm GH}=-\frac{\lambda}{2\pi}\frac{d\Phi_r}{d\chi}
$$
and Fresnel filtering generate a spatial separation of the CW and CCW ray orbits; by placing a waveguide tangentially at a suitable boundary location, one selectively extracts one circulation sense while leaving the opposite circulation inside the cavity [1110.3344].

## 3. Optical realizations

The paradigmatic nanofiber–atom realization uses cesium atoms near a vacuum-clad silica nanofiber with fiber radius $a=250\,\mathrm{nm}$, refractive indices $n_1=1.45$ and $n_2=1.00$, single-mode guidance of the fundamental HE$_{11}$ mode at $852\,\mathrm{nm}$, and an evanescent decay length $\Lambda\approx 200\,\mathrm{nm}$. The atoms are trapped in a single-color dipole trap at $1064\,\mathrm{nm}$ plus $783\,\mathrm{nm}$ blue-detuned light, with trap minima at $r-a=230\,\mathrm{nm}$ and a bias field of $28\,\mathrm{G}$ along $y$. For the cycling transition $m_F=-4\rightarrow m_F'=-5$, only $\sigma^-$ emission is possible. On the left side, $92\%\pm 3\%$ of the guided photons go into $+z$ and $8\%$ into $-z$; on the right side, $86\%\pm 3\%$ go into $-z$ and $14\%$ into $+z$, corresponding to an imbalance ratio $\Gamma_+/\Gamma_->10:1$. For the multibranch decay $m_F=0\rightarrow m_F'=-1$ with $\sigma^+:\sigma^-:\pi=2:5:8$, the measured asymmetries were $\eta_1=66\%\pm2\%, \eta_2=34\%$ on the left side and $\eta_1=43\%\pm1\%, \eta_2=57\%$ on the right side, consistent with symmetric $\pi$ decay and chiral $\sigma$ decay [1406.0896].

A photonic-crystal implementation employs a glide-plane waveguide in a GaAs membrane. The structure is engineered with hole radius $r=0.30\,a$, membrane thickness $t=\tfrac23 a$, GaAs refractive index $n=3.4638$, and a guided band below the light line at $a/\lambda\approx 0.26$ with group index $n_g>20$. The glide operation removes the mirror plane and produces modes with local circular polarization. Experimentally, a single quantum dot emitted into the waveguide with directionality of more than $90\%$, waveguide-coupling efficiency $\beta\gtrsim 90\%$, and directional $\beta$-factor $\beta_{\rm dir}\approx 98\%$. For the best single-QD line, the measured directionality factor was $F_{\rm dir}=90\%\pm 1.3\%$ [1406.4295].

A hybrid plasmonic/photonic realization places a gold Yagi–Uda antenna atop an $80\,\mathrm{nm}$ TiO$_2$ slab of refractive index $n=2.375$ on glass. A CdSe/CdS quantum dot placed $5\,\mathrm{nm}$ from the feed tip excites the antenna at $670\,\mathrm{nm}$. After optimization, the antenna dimensions are $L_f=92\,\mathrm{nm}$, $L_r=98\,\mathrm{nm}$, $L_d=82\,\mathrm{nm}$, $y_r=71\,\mathrm{nm}$, and $y_d=78\,\mathrm{nm}$. Full-wave FEM gives $\eta_{\rm sim}\approx 64\%$ and directivity $G=6.3$ with $F/B=4.7$, corresponding to $82\%$ of the plasmonically enhanced emission propagating toward $+y$. Experimentally, the coupling efficiency was $\eta_{\rm exp}\simeq 39\%$ and the front-to-back ratio was $F/B_{\rm exp}\simeq 1.5$, so that $60\%$ of guided photons propagate in the $+y$ direction; polarization analysis confirmed the TE nature of the guided mode [2402.13557].

An asymmetry-based dielectric implementation uses a GaAs photonic wire waveguide with a two-step stair-like cross section. For the optimum asymmetry $(w'/w,h'/h)=(0.3,0.5)$, FEM yields $\Gamma_+/\Gamma_-\approx 3$ and $\mathrm{DOU}\approx 0.59$, with $\Gamma_+$ only approximately $26\%$ below the symmetric rectangular-waveguide $\Gamma_{\rm tot}$. Full 3D FDTD with $15$ independent circular dipoles gives $\mathrm{DOU}\approx 0.42$, confirming that approximately $71\%$ of the ensemble emission is steered into a single direction. Wrapping the same stair profile into a $2\,\mu\mathrm m$-diameter microdisk gives a TE$_{18}$ mode at $\lambda=1.31\,\mu\mathrm m$ with $Q\approx 3.3\times 10^4$, and the bus-coupled structure yields $\mathrm{DOU}\approx 0.51$ [1902.08385].

In deformed microdisk lasers, a straight dielectric waveguide separated by a sub-wavelength gap $g\approx 100\,\mathrm{nm}$ from a deformed cavity of refractive index $n=3.13$ enables selective extraction of CW or CCW lasing emission. Full-wave calculations and experiment show peaks of $\eta\approx 80\%$ near $\theta=150^\circ$ and $\theta=90^\circ$, and experimentally $\eta_{\rm CW}(150^\circ)\approx 80\%\pm 5\%$, $\eta_{\rm CW}(110^\circ)\approx 60\%\pm 5\%$, and $\eta_{\rm CW}(90^\circ)\approx 30\%\pm 5\%$. Lasing threshold at $\theta=150^\circ$ was $P_{\rm th}\approx 250\,\mu\mathrm W\pm 10\%$ [1110.3344].

## 4. Microwave and collective waveguide-QED interfaces

In superconducting circuits, a central architecture is the artificial molecule of two qubits coupled to a bidirectional waveguide a quarter wavelength apart. For $d=\lambda/4$, waveguide-mediated exchange gives $J(t)=\gamma(t)/2$, and a direct tunable coupling $g_c(t)=-J(t)$ cancels it. The same device can operate as a directional emitter, a directional receiver, or a transparent pass-through element. For a time-symmetric $\mathrm{sech}$ wave packet,
$$
\phi(t)=\frac12\sqrt{\gamma_{\rm ph}\,\mathrm{sech}(\gamma_{\rm ph} t/2)},
$$
the required decay modulation is
$$
\gamma(t)=\gamma_{\rm ph}\,\mathrm{sech}^2(\gamma_{\rm ph} t/2)/[1-\tanh(\gamma_{\rm ph} t/2)].
$$
In resonator-mediated schemes, $\gamma_{\rm ph}\lesssim 0.5\,\gamma$ gives pulse fidelity above $99\%$; residual coupling-cancellation errors of order $10\%$ reduce directionality by less than $1\%$, and separation errors $\delta d\approx \pm 0.02\,\lambda$ or frequency mismatch $|\omega_2-\omega_1|/\gamma_{\rm ph}\approx 0.008$ still keep $D>0.99$ [2004.01924].

An experimental realization uses two frequency-tunable transmon qubits strongly coupled to a bidirectional coplanar waveguide with $\Delta x\approx \lambda/4$. The total exchange is $J_\Sigma=\gamma/2+J_c$, and the operation point $J_c=-\gamma/2$ cancels the static waveguide-mediated term. When the qubits are prepared in
$$
|\psi^\pm\rangle=\frac{1}{\sqrt2}\left(|eg\rangle+e^{\pm i\pi/2}|ge\rangle\right),
$$
interference routes the emitted photon to a chosen direction. Experimentally, for $|\psi^+\rangle$ the measured photon number was $\langle a_R^\dagger a_R\rangle=0.96(\pm0.003)$ with $\langle a_L^\dagger a_L\rangle\lesssim 0.02$, giving single-photon fidelity $F=0.960\pm 0.003$ to $|0_L,1_R\rangle$; for $|\psi^-\rangle$ the results were $\langle a_L^\dagger a_L\rangle=0.954(\pm0.001)$ and fidelity $F=0.954\pm 0.001$ to $|1_L,0_R\rangle$. The extinction ratio exceeded $20\,\mathrm{dB}$ and the emission efficiency was approximately $95\%$ [2203.01430].

A related two-qubit analysis identifies the jump operators
$$
J_R=-i\sqrt{\gamma/2}\,[\sigma_1+e^{-i\phi}\sigma_2],\qquad
J_L=-i\sqrt{\gamma/2}\,[\sigma_1+e^{+i\phi}\sigma_2],
$$
with null-eigenstates
$$
|\psi_R\rangle=\frac{|eg\rangle-e^{-i\phi}|ge\rangle}{\sqrt2},\qquad
|\psi_L\rangle=\frac{|eg\rangle-e^{+i\phi}|ge\rangle}{\sqrt2}.
$$
Under the controlled antiresonance conditions $\phi=(n+\tfrac12)\pi$ and $g_c=\sin\phi=\pm 1$, one obtains $\Gamma_L(|\psi_R\rangle)=0$ and $\Gamma_R(|\psi_R\rangle)=\gamma$. Starting from $|ee\rangle$, the first quantum jump projects onto $|\psi_R\rangle$ or $|\psi_L\rangle$ with equal probability $1/2$, and the second photon must emerge in the same direction, yielding the two-photon N00N state $\tfrac12[|2_R0_L\rangle-|0_R2_L\rangle]$ up to overall phase [2402.01286].

Long-range collective directionality has also been realized with two quantum dots embedded in a bidirectional photonic-crystal waveguide and separated by $13\,\mu\mathrm m$, corresponding to $26$ effective wavelengths. The guided mode mediates dissipative coupling $\Gamma_{12}=\sqrt{\gamma_1^{\rm wg}\gamma_2^{\rm wg}}\cos\phi$ and dispersive coupling $J=\tfrac12\sqrt{\gamma_1^{\rm wg}\gamma_2^{\rm wg}}\sin\phi$, producing dressed states $|\pm\phi\rangle=(|eg\rangle\pm e^{i\phi}|ge\rangle)/\sqrt2$ with shifted energies $\pm J$. Choosing the relative driving phase $\theta_d=\pi-\phi$ yields $I_L=0$ and emission entirely to the right, while $\theta_d=\pi+\phi$ yields $I_R=0$ and emission entirely to the left. In the same platform, continuous driving gives direction-dependent photon statistics, and under pulsed full inversion the measured same-port correlations $g^{(2)}_{LL}(0)$ and $g^{(2)}_{RR}(0)$ show partial bunching of approximately $0.7$–$0.8$, while $g^{(2)}_{LR}(0)=g^{(2)}_{RL}(0)\approx 0.4$ [2604.06410].

## 5. Figures of merit and design rules

The literature uses several closely related performance measures. For guided-mode loading, the usual $\beta$-factor is
$$
\beta=\frac{\Gamma_{\rm wg}}{\Gamma_{\rm wg}+\Gamma_{\rm rad}},
$$
with directional variants
$$
\beta_R=\frac{\Gamma_R}{\Gamma_R+\Gamma_L+\Gamma_{\rm rad}},\qquad
\beta_L=\frac{\Gamma_L}{\Gamma_R+\Gamma_L+\Gamma_{\rm rad}},
$$
and
$$
\beta_{\rm dir}=\beta\,F_{\rm dir},\qquad
F_{\rm dir}=\frac{\max\{\Gamma_R,\Gamma_L\}}{\Gamma_R+\Gamma_L}.
$$
Other common measures are the power asymmetry
$$
D=\frac{P_+-P_-}{P_++P_-},
$$
the coupling efficiency $\eta=P_{\rm wg}/P_{\rm tot}$, and the antenna front-to-back ratio $F/B$ [1406.4295][1910.11228][2402.13557].

Practical design rules differ by platform but show recurring themes. For nanofibers, reported guidelines are $a\lesssim 0.3\,\lambda_0$, emitter placement at $r_0-a\approx (0.2\ldots 0.3)\lambda_0$, use of a pure circular dipole transition $\sigma^\pm$ aligned to the local transverse spin, a quantization axis perpendicular to the fiber axis and in the plane of dipole emission, and suppression of $\pi$-dipole decay to reduce symmetric background [1406.0896]. From the symmetry perspective, maximizing the steepness of $D$ versus transverse angular momentum favors operation near the single-mode cutoff so that $\beta/k_0$ is as large as possible, with emitters kept close to the evanescent field and the transverse axis chosen for quantization to achieve strict sign locking [1910.11228].

In antenna-assisted planar waveguides, the design process starts from effective wavelengths $\lambda_{\rm spp}=\lambda_0/n_{\rm spp}$ and $\lambda_{\rm wg}=\lambda_0/n_{\rm wg}$, then scales feed, reflector, director, and spacing dimensions according to RF Yagi–Uda rules before full FEM optimization. The cited design uses $n_{\rm spp}\approx 3.64$, $n_{\rm wg}\approx 1.72$, a dipole parallel to $x$ at $670\,\mathrm{nm}$, and a $5\,\mathrm{nm}$ feed gap [2402.13557]. In microwave two-emitter interfaces, optimal directionality requires inter-emitter spacing $d/\lambda=1/4+n/2$ and control coupling chosen so that the coherent exchange is canceled, namely $J/\gamma=-(1/2)\sin\phi$ or, equivalently in the antiresonant condition, $g_c=\sin\phi=\pm1$ [2402.01286]. For long-range photonic-crystal molecules, the reported guidelines are to choose the coupling phase near $\phi\approx (N+\tfrac12)\pi$, work near the band edge to boost $n_g$, and use independently tunable segments to align emitters into resonance; in the cited device, $\beta>0.8$ was achieved [2604.06410].

## 6. Conceptual clarifications, applications, and extensions

A recurrent misconception is that all directional coupling can be reduced to a single notion of spin–momentum locking. The symmetry analysis shows a more specific statement: independently of emitter displacement, directionality is mostly due to mirror-symmetry breaking caused by the axial character of the angular momentum of the emitted light, whereas the chiral character of the handedness of the emission yields a binary and less pronounced effect when displacement breaks another mirror symmetry. The same work also shows that choosing a different angular-momentum axis changes whether strict locking exists and how directionality depends on angular momentum [1910.11228].

Functionally, directional interfaces support single-photon routing, quantum switching, non-reciprocal emission, and cascaded interactions. The nanofiber platform explicitly identifies integrated quantum photonic circuits, on-chip optical isolators, chiral quantum networks, and single-photon diodes and circulators as target applications [1406.0896]. In chiral photonic circuits, near-unity directional coupling enables optical diodes and circulators at the single-photon level and motivates a deterministic photon–photon CNOT gate with entanglement fidelity $\mathcal F=\beta_{\rm dir}^2\approx 0.96$ and worst-case fidelity $\mathcal F_{\min}\approx 0.92$ [1406.4295]. A $\Lambda$-type emitter coupled to a chiral waveguide extends the same interface concept to quantum-memory and gate operations: for a Gaussian photon pulse of bandwidth $\sigma_\omega=5\gamma$ and coupling ratio $\Gamma/\gamma=50$, the reported averages are $\bar F_{\rm swap}\approx 0.9982$, $\bar\eta_{\rm swap}\approx 0.9810$, $\bar F_{\rm ent}\approx 0.9996$, $\bar\eta_{\rm ent}\approx 0.9901$, and $\bar F_{\rm mem}\gtrsim 0.9928$, $\bar\eta_{\rm mem}\gtrsim 0.9345$ [1803.05626].

The interface also has non-Markovian and network-level extensions. Near a waveguide continuum edge, the density of states singularity produces a non-local memory kernel and allows exact control-pulse prescriptions for shaped emission and absorption in a $\Lambda$-type system; with free-space leakage $\gamma'\le 0.06\gamma$, the cited fidelities exceed $96\%$, and for slow pulses with partial photonic-crystal cladding the interface fidelity exceeds $99\%$ [1110.5386]. At the network scale, phase-tunable waveguide-QED architectures use engineered propagation and coupling phases so that coherent exchange and collective dissipation balance to suppress the backward channel while retaining a finite forward channel. In that framework, perfect nonreciprocity corresponds to $g_+=0$, equivalently $kL+\Delta\varphi=(n+\tfrac12)\pi$, and the giant-small-emitter mirror-terminated configuration simultaneously achieves perfect nonreciprocity and battery-dominated storage [2605.21909].

Taken together, these results suggest that a directional waveguide–emitter coupling interface is not a single device architecture but a unifying operational principle. Across atomic, semiconductor, plasmonic, cavity, and superconducting implementations, the central task is the same: engineer the local vectorial mode structure or the collective interference pattern so that one propagation channel becomes bright while the opposite channel becomes dark.

Source: https://www.emergentmind.com/topics/directional-waveguide-emitter-coupling-interface