Edge Coupling: Boundary-Mediated Interactions

• Edge coupling is defined as boundary-mediated interactions in which edge-localized states hybridize with bulk modes, enabling engineered transport and dispersion properties.
• It spans platforms from photonic Chern insulators to quantum Hall systems, where differential edge engineering controls slow-light windows, charge equilibration, and mode splitting.
• Applications include impedance-matching in integrated photonics and long-range magnetic exchange in graphene, exploiting boundary design to enhance device performance.

Searching arXiv for recent and foundational papers on edge coupling across photonics, quantum Hall, and topological materials. Searching arXiv for "edge coupling topological edge states flat bands photonic crystal quantum Hall" and related terms. Edge coupling denotes a class of boundary-mediated interactions in which edge-localized degrees of freedom are hybridized with other edge modes, bulk excitations, localized resonances, reservoirs, spins, molecules, or external waveguides. The term is used in several partially overlapping senses: it can mean coupling a chiral edge mode to flat-band resonators, equilibration between co-propagating or counterpropagating quantum Hall edge channels, exchange processes mediated by edge states, coupling of localized corner modes through edge bands, or butt coupling through a chip facet in integrated photonics. A common structural feature is that the boundary is not merely a geometric termination; it acts as a spectrally selective and often topologically significant region in which localization, momentum matching, impedance matching, or Coulomb screening can be engineered.

1. Terminological scope and conceptual structure

In topological and mesoscopic physics, edge coupling usually refers to interactions involving states localized at, or strongly shaped by, a boundary. In photonic Chern insulators, it can mean coupling a propagating chiral edge mode to localized cavity resonances attached to the boundary, thereby reshaping the edge dispersion without altering the bulk lattice (Yu et al., 2020). In quantum Hall systems, it can denote controlled mixing of co-propagating spin-resolved edge states, interlayer equilibration between closely spaced bilayer edge channels, or Coulomb coupling between a compressible edge and localized bulk quasiparticles in an interferometer (Karmakar et al., 2011). In finite lattices, it can also refer to the dependence of edge-state existence on boundary termination, as shown for kagome ribbons where certain terminations suppress crystalline edge modes completely, while topological gaps generated by Kane–Mele spin-orbit coupling restore helical edge states insensitive to termination details (Sekh et al., 12 Feb 2026).

A central distinction runs between termination-induced and topology-enforced edge coupling. In the first case, the boundary geometry itself determines whether localized edge modes exist and how strongly they overlap. In the second, bulk invariants constrain the number and connectivity of edge channels, while local coupling modifies dispersion, linewidth, or interaction strength but does not eliminate the topologically required modes as long as the bulk gap remains open. A related distinction is between edge-only engineering, where the bulk is left unchanged, and bulk-sensitive coupling, where the boundary response is slaved to bulk filling, Chern number, scalar spin chirality, or non-Hermitian topology.

A common misconception is that edge coupling necessarily destroys protected transport. That is not generally correct. In Ta-intercalated epitaxial graphene on SiC(0001), quantized edge transport persists in the presence of inter-edge coupling because orbital-dependent decay allows coupling between only one pair of edge states rather than two, yielding two perfect Dirac cones with different Fermi velocities, of which only one maintains the edge-state feature (Li, 2016).

2. Boundary hybridization in photonics and wave systems

A particularly explicit form of edge coupling is the local hybridization of a propagating chiral edge mode with flat-band resonances attached to the edge of a photonic topological insulator. In a gyromagnetic photonic Chern insulator, additional dielectric resonators placed only at the lower boundary form flat bands when mutually decoupled. Opening the coupling channels hybridizes these resonances with the chiral edge mode, creating avoided crossings and increasing the momentum-space winding of the edge dispersion. For $n=1,2,3$ coupled flat-band resonances per unit cell, the hybridized bottom-edge mode winds once, twice, and three times around the Brillouin zone, respectively, and the slow-light window broadens with $n$; for $n=3$, the broadband topological slow-light region occupies about $34\%$ of the topological bandgap (Yu et al., 2020). This demonstrates an edge-only route to broadband slow light in which the bulk topology remains fixed and only boundary resonators are varied.

An older but related photonic example is the woodpile photonic crystal, where optical surface edge Bloch modes are localized at the intersection of two crystal terminations. These modes lie inside a complete 3D photonic bandgap, have field maxima in vacuum near the dielectric-vacuum edge, and exhibit a mode area $A = 0.066(\lambda_0/2)^2$ together with a radiation quality factor $Q_{\mathrm{rad}} \approx 3\times 10^7$ (Su et al., 2012). Their coupling is governed by evanescent-field overlap: two identical surface Bloch modes on opposite surfaces form symmetric and antisymmetric supermodes whose frequency splitting determines the coupling coefficient. Because the coupling decreases only slowly with separation $N_x$, large photonic crystals or careful surface design are required to suppress unwanted cross-talk (Su et al., 2012).

In waveguide quantum electrodynamics, the relevant “edge” is spectral rather than spatial. A single-mode waveguide has a lower cut-off frequency $\omega_0$ and a density of states that diverges as $1/\sqrt{\omega-\omega_0}$ near the continuum edge. Coupling atomic transitions to this waveguide continuum edge produces non-Markovian dynamics, vacuum Rabi oscillations, and bound polariton states, even though the bath is a continuum rather than a discrete cavity mode (Chen et al., 2011). The waveguide edge therefore plays a cavity-like role, but with memory effects that invalidate the flat-density-of-states approximation used in Markovian treatments.

The same logic extends to topological plasmonics and non-Hermitian photonics. In a plasmonic SSH2D nanoparticle array, molecules can couple selectively to topological edge states of the array. In the weak-coupling regime, the energy-transfer enhancement factor can reach $\sim 10^6$ near bright edge resonances, and in the strong-coupling regime, edge polaritons mediate enhancements up to $n$0, enabling highly directional long-range molecular energy transfer along the boundary (Buendía et al., 2024). In a parity-time-symmetric dimerized microwave lattice, staggered loss drives spontaneous PT-symmetry breaking of the edge sector while preserving a non-Hermitian topological invariant. There, the strong coupling between a topological photonic edge mode and a YIG magnon mode is enhanced from $n$1 in the Hermitian case to $n$2 after non-Hermitian engineering of the edge states (Qian et al., 2023). This suggests that edge coupling can be strengthened not only by confinement and overlap but also by controlled modification of photonic density of states and loss asymmetry at the boundary.

Higher-order topological systems supply another variant. In an exciton-polariton Benalcazar–Bernevig–Hughes lattice, corner modes do not couple directly in the linear regime. Instead, polariton nonlinearity drives optical parametric scattering from a pumped corner state into edge modes, and then from edge modes into another corner state, producing effective corner–edge–corner coupling (Banerjee et al., 2020). Edge states thus function as a nonlinear bus connecting zero-dimensional topological modes.

3. Quantum Hall transport, reservoir coupling, and interferometry

Quantum Hall systems provide a canonical setting for edge coupling because bulk transport is frozen out and the relevant dynamics is concentrated on one-dimensional chiral channels. A direct and coherent implementation is the controlled coupling of co-propagating spin-resolved edge states at $n$3. An array of Cobalt nano-magnets placed along the boundary of a GaAs 2DEG produces a spatially periodic in-plane magnetic field, which simultaneously flips spin and supplies momentum. The coupling is resonant when the modulation period satisfies $n$4, where $n$5 is the wave-vector mismatch between the two spin-resolved edge states. Experimentally, a maximum charge/spin transfer of about $n$6 is achieved at $n$7 (Karmakar et al., 2011). The non-monotonic dependence on the array period indicates coherent addition of scattering amplitudes rather than incoherent mode mixing.

A different form of edge coupling appears in bilayer quantum Hall counterflow. In a GaAs double quantum well with a $n$8 barrier, the edge modes of the two 2DEGs lie in close proximity, and counterflow measurements show nearly vanishing Hall resistances at integer fillings together with filling-dependent equilibration lengths. Using an exponential equilibration model, the edge-channel equilibration length is extracted as $n$9 at $n=3$0 and $n=3$1 at $n=3$2, a nearly fourfold change (Marty et al., 2023). The interpretation advanced there is enhanced one-dimensional interlayer coupling via edge channels, plausibly strengthened by exciton-like correlations confined to the edge geometry (Marty et al., 2023).

Coupling to reservoirs may occur through either particle exchange or density–density interactions. In the heat-transport problem for a single chiral quantum Hall edge coupled to two reservoirs, tunneling coupling and capacitive coupling lead to qualitatively different thermal conductance laws. For tunneling coupling, the thermal conductance grows linearly with temperature, whereas for capacitive coupling it scales as $n=3$3 at low temperature and saturates at larger temperatures (Aita et al., 2013). The boundary is therefore not a passive conduit: the microscopic form of edge–reservoir coupling determines which low-energy modes are available for heat transport and how finite-size effects enter through the mean level spacing $n=3$4.

Fabry–Perot interferometers introduce yet another layer of edge coupling: Coulomb interaction between the interfering edge and localized bulk quasiparticles. In that setting, the interference phase contains a Coulomb term proportional to $n=3$5, where $n=3$6 parametrizes bulk–edge Coulomb coupling and $n=3$7 is the filling difference between the central plateau and the next outer plateau (Keyserlingk et al., 2014). Because $n=3$8 can be small on fractional plateaus, moderate microscopic coupling can produce substantial area readjustment and large frequency shifts in both $n=3$9-sweeps and $34\%$0-sweeps. For $34\%$1, the analysis further includes tunnel coupling between localized bulk Majorana states and Majorana edge states. The conclusion is deliberately cautious: the data are consistent with Moore–Read topological order, but experiments may have measured Coulomb effects rather than an “even-odd effect” due to non-abelian braiding (Keyserlingk et al., 2014). This is one of the clearest demonstrations that edge coupling can masquerade as a topological interferometric signature.

4. Electronic materials, exchange interactions, and termination-controlled coupling

In graphene nanoflakes, edge states can mediate magnetic coupling over unexpectedly long distances. Exact-diagonalization studies of the RKKY interaction show that zigzag edges host zero-energy edge states at half filling, whereas armchair edges do not and instead exhibit a finite gap around zero energy (Canbolat et al., 2023). When impurities are placed on zigzag edges, the edge-localized states supply a large local density of states at the Fermi level and turn the effective exchange into a strong, long-ranged interaction. In zigzag hexagonal flakes, the edge-to-edge RKKY coupling between impurities at midpoints of opposite zigzag edges initially increases with flake size and reaches a maximum around $34\%$2 atoms per edge before decreasing again; even a $34\%$3 edge-vacancy ratio causes an abrupt decay of the enhanced coupling (Canbolat et al., 2023). This is a boundary-mediated magnetic interaction in which geometry, sublattice structure, and defect density directly control exchange.

Inter-edge coupling in topological graphene-derived systems need not be destructive. In Ta-intercalated epitaxial graphene on SiC(0001), a non-trivial gap of $34\%$4 coexists with two perfect Dirac cones with different Fermi velocities, yet only one maintains the edge-state feature (Li, 2016). The proposed mechanism is orbital-dependent decay into the bulk: one pair of edge states overlaps strongly across the ribbon width, while the other remains effectively decoupled. The result is an intermediate regime between trivial and fully decoupled quantum spin Hall behavior in which quantized edge transport persists even though inter-edge coupling is present (Li, 2016). This directly contradicts the simplistic expectation that any finite inter-edge overlap must gap out and destroy the quantum spin Hall effect.

The kagome lattice provides a broader taxonomy. In the pristine tight-binding limit, the existence of localized edge states depends strongly on termination: armchair, zigzag, and cove edges support boundary states, while flat termination can eliminate them completely (Sekh et al., 12 Feb 2026). Adding Kane–Mele spin-orbit coupling opens a bulk gap and yields a robust $34\%$5 phase with helical edge states insensitive to termination details. By contrast, a Zeeman field plus Rashba spin-orbit coupling generates QAH phases whose Chern numbers match the number of chiral edge modes, and non-coplanar magnetic textures produce multiple Chern phases through finite scalar spin chirality (Sekh et al., 12 Feb 2026). Here edge coupling spans the full range from local, termination-controlled crystalline modes to topology-enforced chiral or helical channels whose multiplicity is dictated by bulk invariants.

5. Edge coupling as an interface technology in integrated photonics

In integrated photonics, “edge coupling” often has a more literal meaning: butt coupling through the chip facet. This usage is technologically distinct from topological or mesoscopic edge-state hybridization, but it shares the same boundary-centric logic of mode matching, impedance control, and adiabatic transformation.

On lithium niobate on insulator, inverse tapers are used to expand the on-chip mode near the facet and reduce mismatch to a lensed fiber. In one implementation on LNOI, crystal cut and etch anisotropy determine how narrow a tip can be fabricated. Using $34\%$6 tip mode-matching tapers, the reported butt-coupling loss is $34\%$7 for $34\%$8-cut LNOI and $34\%$9 for $A = 0.066(\lambda_0/2)^2$0-cut MgO:LNOI, both with low propagation loss; the performance difference is traced to sidewall angle and anisotropy limits set by the material orientation (Krasnokutska et al., 2019). In this context, edge coupling is a fabrication-limited problem once taper length is sufficient for adiabaticity.

A more recent TFLN design replaces ultra-narrow lateral tapering by a vertically tapered wedge in the lithium-niobate film combined with a silicon oxynitride cladding waveguide. The measured coupling loss between the TFLN PIC and a $A = 0.066(\lambda_0/2)^2$1 mode-field-diameter lensed fiber is $A = 0.066(\lambda_0/2)^2$2, with a theoretical potential of $A = 0.066(\lambda_0/2)^2$3; the coupling loss to a UHNA7 fiber with $A = 0.066(\lambda_0/2)^2$4 MFD is $A = 0.066(\lambda_0/2)^2$5 (Jia et al., 2023). The minimum linewidth of the coupler is $A = 0.066(\lambda_0/2)^2$6, compatible with i-line stepper lithography (Jia et al., 2023). Here “edge coupling” refers to fiber-to-chip boundary engineering rather than topological edge states, yet the underlying design principle is still boundary-local mode conversion.

This terminological divergence is useful rather than problematic. It indicates that the edge can function either as a state-bearing boundary in topological systems or as an impedance-transforming interface in photonic packaging. In both cases, performance is dictated by how sharply or adiabatically boundary-localized fields are allowed to overlap.

6. Design principles, observables, and recurrent debates

Across these literatures, several design principles recur. The first is momentum or phase matching. The Cobalt finger arrays in the quantum Hall spin-coupler work only because the magnetic modulation period matches the wave-vector difference $A = 0.066(\lambda_0/2)^2$7 between the two edge channels (Karmakar et al., 2011). Flat-band-assisted slow light works because avoided crossings are inserted at spectrally selected points along the edge dispersion (Yu et al., 2020). Surface Bloch mode couplers and molecular topological edge-state transfer rely on evanescent overlap and supermode splitting (Su et al., 2012). Fabry–Perot interferometers translate bulk–edge Coulomb coupling into modified oscillation frequencies because the edge area readjusts coherently with bulk charge (Keyserlingk et al., 2014).

The second is boundary selectivity without bulk redesign. The gyromagnetic slow-light scheme changes only the bottom-most row and attached resonators, not the photonic Chern-insulator bulk (Yu et al., 2020). The non-Hermitian magnon–edge system modifies edge-state loss and density of states while keeping the bulk in a topological phase (Qian et al., 2023). The TFLN wedge coupler reshapes only the chip-edge mode transformer, leaving the functional circuit interior unchanged (Jia et al., 2023).

The third is robustness conditioned by the relevant invariant or symmetry. Helical edge states stabilized by Kane–Mele SOC in kagome ribbons are insensitive to termination details as long as time-reversal symmetry and the bulk gap persist (Sekh et al., 12 Feb 2026). Chiral edge channels in Chern phases remain guaranteed even when their local dispersion is altered (Yu et al., 2020). Conversely, purely crystalline edge states in pristine kagome or woodpile structures are termination-sensitive and can be suppressed altogether (Sekh et al., 12 Feb 2026).

A recurring debate concerns interpretation. Not every boundary mode is topological, and not every oscillation associated with an edge reflects the intended statistics or invariant. Edge-state existence can be purely geometric in the absence of bulk topology (Vedula et al., 6 Feb 2026). Inter-edge coupling need not destroy quantized transport when it is orbital selective (Li, 2016). At $A = 0.066(\lambda_0/2)^2$8, observed interferometric frequencies may reflect enhanced bulk–edge Coulomb coupling and parity-randomized Abelian interference rather than a clean even–odd signature of non-Abelian braiding (Keyserlingk et al., 2014). These cases show that edge coupling is often diagnostically ambiguous unless accompanied by a careful account of boundary geometry, interaction channel, and symmetry constraints.

Taken together, the literature treats edge coupling not as a single mechanism but as a boundary-centered design paradigm. Edges can act as slow-light dispersion shapers, thermal or electrical equilibration channels, exchange mediators, nonlinear buses between higher-order modes, non-Hermitian density-of-states enhancers, or practical optical interfaces. The technical details differ sharply, but the unifying theme is that a boundary can be engineered as an active degree of freedom whose couplings control transport, localization, and spectroscopy in ways the bulk alone does not permit.