Polaritonic Surface Cavities

• Polaritonic surface cavities are nanoscale, surface-bound photonic environments that confine electromagnetic modes at interfaces to facilitate ultrafast strong light–matter coupling.
• They employ diverse architectures such as plasmonic nanostructures, metasurfaces, and open cavities that balance extreme subwavelength confinement with inherent losses.
• These systems enable enhanced spectroscopy, controlled energy transfer, and cavity-induced chemical dynamics under ambient conditions.

Polaritonic surface cavities are cavity quantum electrodynamic environments in which an electromagnetic mode is bound to, or strongly confined near, a surface or interface and coherently exchanges energy with a material excitation faster than the system loses energy, thereby producing hybrid light–matter quasiparticles with upper and lower polariton branches, modified dispersions and lifetimes, and new channels for energy, electron, and chemical dynamics. In current usage, the expression also covers a second, related meaning: in molecular cavity QED it can denote the multidimensional adiabatic polaritonic potential-energy surfaces generated when electronic, nuclear, and photonic degrees of freedom are treated on equal footing. Taken together, these usages place the topic at the intersection of nanophotonics, surface polariton physics, strong-coupling spectroscopy, and cavity-modified molecular dynamics (Johns et al., 2 May 2026, Fábri et al., 2021).

1. Surface-bound cavity architectures and their defining features

Surface cavities, in the nanophotonic sense, are photonic environments whose modes are localized at or near interfaces rather than distributed across a large dielectric resonator volume. The main architectures surveyed in the recent literature are open cavities near a surface, plasmonic nanostructures, metasurfaces, and surface-bound polaritonic modes such as surface plasmon polaritons at metal surfaces, Bloch surface waves at dielectric multilayers, Tamm plasmon polaritons at metal–Bragg stacks, and surface phonon polaritons in polar dielectrics such as hBN and SiC in the mid-IR. Open cavities include single- or mirrorless configurations in which refractive-index contrast, for example Si/SiO$_2$/organic/air or Ag/hBN/2D layer, creates leaky modes that remain accessible to deposited materials, chemistry, and device operation. Plasmonic realizations include nanoantennas, dimers, bowties, nanoparticle-on-mirror gaps, nanoslit arrays, and slit or hole gratings on metal films. Metasurfaces act as planar resonators formed from ordered nanoantenna arrays with engineered radiation patterns, modal lifetimes, and near fields, while retaining open-cavity in- and out-coupling (Johns et al., 2 May 2026).

The defining distinction from conventional dielectric Fabry–Perot microcavities lies in the balance between confinement and loss. Dielectric microcavities have larger mode volumes, weaker field localization, but higher $Q$ factors of $10$–$100$ and lifetimes of hundreds of fs. Surface cavities, especially plasmonic ones, reach extreme subwavelength confinement with mode volumes down to $1$–$100~\mathrm{nm}^3$ and even sub-$\mathrm{nm}^3$ in picocavities, but typically at $Q \le 10$ and lifetimes of only tens of fs. Metals add Ohmic loss and radiative leakage, and open cavities add leaky channels, yet strong coupling is still routinely achieved because the coupling strength scales with field amplitude and inverse square root of mode volume. This makes surface cavities particularly effective for strong and ultrastrong coupling with organic materials at room temperature, whereas dielectric microcavities more readily sustain long-lived polariton dynamics and condensation but often require cryogenic conditions for inorganic Wannier–Mott excitons (Johns et al., 2 May 2026).

A second terminological strand appears in molecular cavity QED, where “polaritonic surface cavities” refers to the multidimensional adiabatic polaritonic potential-energy surfaces that arise when a molecule is placed inside an optical or nanoplasmonic cavity. In that usage, the cavity reshapes the Born–Oppenheimer landscape itself: the resulting lower- and upper-polaritonic surfaces depend on nuclear coordinates, can host light-induced conical intersections, and directly govern nonadiabatic nuclear wavepacket motion (Fábri et al., 2021).

2. Electromagnetic models, coupling regimes, and figures of merit

The standard reduced description of strong coupling in a surface cavity is the coupled-oscillator, Hopfield, or Tavis–Cummings model. In rotating-wave form,

$$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$

while ultrastrong coupling requires counter-rotating terms,

$$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$

For a mode with dispersion $Q$0 coupled to an excitation $Q$1, the polariton branches are

$Q$2

with Rabi splitting $Q$3. The strong-coupling criterion is $Q$4, where $Q$5 and $Q$6 are cavity and exciton dissipation rates, whereas the level-splitting or exceptional-point criterion is weaker, $Q$7. The latter defines the onset of non-degenerate strong polaritonic eigenmodes even when spectrally resolved Rabi splitting is absent. Ultrastrong coupling is conventionally identified when $Q$8 (Johns et al., 2 May 2026).

For surface-bound modes, the electromagnetic dispersion is itself interface-determined. At a metal–dielectric boundary, the surface plasmon polariton wavevector is

$Q$9

with $10$0 and $10$1. Surface phonon polaritons in polar dielectrics take an analogous form with the metal permittivity replaced by the lattice permittivity near the Reststrahlen band. The main cavity figures of merit remain $10$2, the mode volume $10$3, and the Purcell factor

$10$4

which makes explicit why ultra-small $10$5 can compensate for modest $10$6 in surface cavities (Johns et al., 2 May 2026).

Spherical Mie voids provide a conceptually distinct but still surface-confined implementation. A spherical air void embedded inside a dielectric or resonant medium supports TE and TM quasinormal modes whose fields peak at the void–host interface because confinement arises from the refractive-index step and curvature. Their eigenfrequencies are complex, $10$7, with $10$8. In the high-index limit $10$9, the paper derives

$100$0

showing that void-mode $100$1 grows only algebraically with host permittivity. When the void is filled with a Lorentz medium, the polaritonic spectrum is captured by a non-Hermitian $100$2 model with effective coupling $100$3, and strong coupling at zero detuning requires $100$4 (Ryabkov et al., 6 Oct 2025).

3. Coherent dynamics, vibronic structure, and dark-state reservoirs

Surface cavities do not merely split spectra; they support ultrafast coherent exchange under highly lossy conditions. Plasmonic nanoslit arrays loaded with J-aggregates show direct observation of $100$5 Rabi oscillations lasting $100$6, while colloidal plexcitonic nanourchins exhibit coherent beatings with $100$7 dephasing in solution at room temperature. In organic platforms, vibronic couplings can either erode or stabilize coherence. When the Rabi splitting exceeds vibronic coupling within a Holstein–Tavis–Cummings picture, polaron decoupling can preserve coherence; conversely, vibrational channels can mediate population exchange between bright polaritons and dense manifolds of dark states, including coherent interactions between lower polaritons and partially dark excitons in disordered plexcitonic systems (Johns et al., 2 May 2026).

A central complication is the dark-state reservoir. Bright polaritons coexist with dense manifolds of dark excitons and plasmonic resonances, so broadband excitation often populates the reservoir and obscures lifetime assignments. Momentum-resolved excitation can isolate lower polaritons and reveal sub-$100$8 polariton lifetimes distinct from slower dark-state dynamics. This distinction matters especially in “intermediate coupling” and “dark-strong coupling,” where non-degenerate eigenlevels exist even though $100$9 does not exceed combined losses and spectra may not display a clear anti-crossing. A representative example is WS$1$0/hBN/Ag open cavities, where quasi-normal-mode analysis revealed $1$1 level splitting for hBN thickness $1$2, hidden by linewidths of $1$3; momentum-resolved photoluminescence nevertheless showed multiple dispersive branches consistent with polaritons (Johns et al., 2 May 2026).

The experimental and computational toolkit reflects these subtleties. Angle-resolved reflectivity and emission, Fourier microscopy, and $1$4-space imaging remain the primary route to direct dispersion mapping. Near-field and tip-enhanced photoluminescence access nanoscale hotspots and single-molecule regimes in plasmonic nanocavities and picocavities. Ultrafast pump–probe and two-dimensional electronic spectroscopy resolve Rabi oscillations, bright–dark energy flow, and cross-peaks among polariton branches; notably, quasi-instantaneous $1$5 energy delocalization between spatially separated J-aggregates via a single cavity mode has been demonstrated. On the modeling side, transfer-matrix methods, FDTD, FEM, BEM or DDA, quasi-normal-mode analysis, Q-PCM-NP, and atomistic electrodynamics via wFQFu are used to extract $1$6, $1$7, mode volumes, Hopfield coefficients, tunneling, and realistic picocavity field morphologies. For realistic cavity geometries, Maxwell–DFTB simulations further show that local molecular responses depend strongly on the number, geometry, position, and orientation of molecules inside the cavity, even when the device-level spectrum displays only conventional upper- and lower-polariton peaks (Johns et al., 2 May 2026, Bustamante et al., 26 Aug 2025).

4. Molecular polaritonic surfaces, light-induced conical intersections, and exact nuclear potentials

In molecular cavity QED, polaritonic surfaces are obtained by diagonalizing the electronic–photonic Hamiltonian at fixed nuclear geometry. A general single-mode Pauli–Fierz Hamiltonian in the length gauge can be written as

$1$8

In the reduced model employed for formaldehyde, a creation–annihilation representation is used,

$1$9

with the singly excited manifold spanned by $100~\mathrm{nm}^3$0 and $100~\mathrm{nm}^3$1. Diagonalization at each geometry produces lower and upper polaritonic potential-energy surfaces whose excitonic and photonic character varies over nuclear coordinates (Fábri et al., 2021).

These surfaces can host light-induced conical intersections. In the H$100~\mathrm{nm}^3$2CO model, symmetry forces the body-fixed $100~\mathrm{nm}^3$3 component of the transition dipole to vanish for planar geometries, $100~\mathrm{nm}^3$4. When this zero-coupling condition coincides with resonance,

$100~\mathrm{nm}^3$5

the lower- and upper-polariton surfaces cross and form a LICI seam. Two cavity settings were studied: $100~\mathrm{nm}^3$6, giving a LICI at $100~\mathrm{nm}^3$7, $100~\mathrm{nm}^3$8, and $100~\mathrm{nm}^3$9, giving a LICI at $\mathrm{nm}^3$0, $\mathrm{nm}^3$1. Couplings $\mathrm{nm}^3$2–$\mathrm{nm}^3$3, corresponding to mode volumes of order $\mathrm{nm}^3$4–$\mathrm{nm}^3$5, and cavity decay rates $\mathrm{nm}^3$6–$\mathrm{nm}^3$7, corresponding to lifetimes of $\mathrm{nm}^3$8–$\mathrm{nm}^3$9, yielded time-dependent cavity emission that mapped both intrapolariton motion and LICI-mediated surface-to-surface transfer. In the resonant-transfer case, oscillations in the intracavity photon number $Q \le 10$0 exhibited minima and maxima near $Q \le 10$1 and $Q \le 10$2, providing an experimentally accessible fingerprint of LICI-mediated nonadiabatic population exchange without a probe pulse (Fábri et al., 2021).

Exact-factorization studies show, however, that even these polaritonic surfaces are not always the full dynamical object governing nuclei. For cavity-modified proton-coupled electron transfer and cavity-induced electronic excitation, the exact time-dependent potential energy surface is

$Q \le 10$3

where $Q \le 10$4 is the weighted polaritonic term, $Q \le 10$5 arises from nuclear derivatives of the conditional state, and $Q \le 10$6 is a gauge-dependent term. In the one-dimensional models studied, $Q \le 10$7 is negligible, but $Q \le 10$8 becomes crucial: it redistributes energy between nuclear and electronic–photonic subsystems, builds sharp step features in mixed-character regions, and enables wavepacket splitting, trapping, and reflection. Classical trajectory ensembles propagated on the exact $Q \le 10$9 reproduce the exact nuclear density very well, whereas propagation on the weighted polaritonic surface alone fails after a short time. This establishes that static or weighted polaritonic surfaces can be qualitatively misleading whenever several polaritonic surfaces contribute in the same spatial region (Martinez et al., 2020).

5. Materials, parameter ranges, and benchmark platforms

The material landscape spans plasmonic metals, polar dielectrics, inorganic semiconductors, organics, and hybrids. Representative metals are Au and Ag for surface-plasmon platforms and Al for nanoellipse antennas coupled to photoswitches. hBN functions both as a cavity spacer and as a surface-phonon-polariton host in the mid-IR, while SiC is a canonical SPhP platform. TMDC monolayers such as WS$$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$0 strongly couple to localized plasmons, propagating SPPs, and thin or open cavities at room temperature. Organic excitons include J-aggregates such as TDBC and NK2707, organic semiconductors such as PDI, and photoswitches such as spiropyran. Across these systems, typical reported values are $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$1 for plasmonics, low–moderate $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$2 for open cavities, $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$3–$$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$4 for dielectric microcavities, plasmonic mode volumes of $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$5–$$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$6, nanoparticle-on-mirror gaps in the sub-nm regime, picocavity mode volumes below $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$7, ballistic polariton propagation at tens of $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$8 over microns, energy transfer across $$H=\hbar\omega_c a^\dagger a+\hbar\omega_x b^\dagger b+\hbar g(a^\dagger b+a b^\dagger),$$9 donor–acceptor layers mediated by polaritons, and electron-mobility enhancements up to $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$0 under ON-resonance in mirrorless cavity MOSFETs (Johns et al., 2 May 2026).

Mid-infrared image-phonon-polariton cavities show how the surface-cavity concept extends beyond plasmonics. In 4H-SiC, the Reststrahlen band spans $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$1, corresponding to $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$2–$$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$3. Silver nanocubes separated from SiC by a $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$4 PAH spacer realize antisymmetric-image surface phonon polariton resonators with a reflectance-contrast resonance at $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$5 for $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$6 cubes, systematic blue-shifts as $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$7 is reduced to $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$8 and $$H_{\mathrm{USC}}=H+\hbar g(a^\dagger b^\dagger+a b).$$9, quality factors surpassing $Q$00 at room temperature, and normalized mode volumes $Q$01. The effective mode volume is thus “almost a billion times smaller” than the free-space diffraction-limited volume, while the field pattern displays a magnetic-dipole hot spot centered in the SiC–PAH–Ag gap (Klein et al., 6 Mar 2025).

Planar perovskite photonic-crystal slabs provide a room-temperature exciton-polariton counterpart. In $Q$02-thick $Q$03 on SiO$Q$04, mechanical scanning probe lithography writes one-dimensional photonic crystal slabs with periods $Q$05–$Q$06 and modulation depths $Q$07–$Q$08. The exciton transition is centered at $Q$09, measured Rabi splittings are $Q$10–$Q$11, uncoupled linewidths satisfy $Q$12 and $Q$13, the radiative-loss fraction $Q$14 is tunable from $Q$15 to $Q$16, and leaky polaritonic modes reach $Q$17 up to about $Q$18 at $Q$19. The $Q$20-point crossing energy decreases monotonically with period, from $Q$21 at $Q$22 to $Q$23 at $Q$24, directly demonstrating lithographic control over planar surface-cavity dispersion and outcoupling (Glebov et al., 2023).

6. Specialized cavity classes: voids, self-assembly, and boundary-localized polaritons

Mie voids realize a compact dielectric analogue of surface confinement. A spherical void of radius $Q$25 embedded in a dielectric or resonant host supports TE and TM multipolar quasinormal modes whose fields are concentrated at the void–host interface. “Good” cavity modes correspond to radial order $Q$26, whereas $Q$27 modes are delocalized and have low $Q$28. When the void is loaded with a Lorentz oscillator medium, the resulting polaritonic spectrum shows separate upper and lower polariton pairs for each photonic mode, with strong coupling depending on oscillator strength $Q$29, the cavity quality factor $Q$30, and the matter quality factor $Q$31. The work emphasizes that far-field anticrossings can be obscured by multi-mode Mie structure and that near-field maps and complex-frequency tracking are often more reliable diagnostics than total scattering cross-sections alone (Ryabkov et al., 6 Oct 2025).

Casimir microcavities demonstrate that surface cavities can be self-assembled and actively tunable in liquid. Two parallel single-crystalline gold nanoflakes in water form a contact-free Fabry–Perot cavity with equilibrium gap set by the balance of Lifshitz attraction and DLVO repulsion. In dimers, quasi-normal-incidence reflectivity dips at $Q$32, $Q$33, and $Q$34 correspond to fitted gaps of $Q$35, $Q$36, and $Q$37, respectively. In a related floating-flake–gold-film configuration, the empty cavity has average thickness $Q$38 with standard deviation $Q$39, remains stable for at least weeks, and can be modulated by $Q$40 at $Q$41–$Q$42 by chopped laser illumination. Loading the gap with few-layer WSe$Q$43 yields upper and lower polaritons with clear anti-crossing, and the time-dependent Hopfield coefficients show that the same device can be driven in and out of strong coupling by optical pressure and photothermal shifts (Munkhbat et al., 2020).

At the boundary between cavity electrodynamics and lattice models, cavity photons can also induce Tamm-like polaritonic edge states. A finite one-dimensional periodic chain of oscillating dipoles inside a cuboid cavity waveguide develops surface-localized polariton modes only when the cavity cross-section exceeds a critical size. For the parameters $Q$44, $Q$45, and $Q$46, the threshold is $Q$47, or equivalently a critical area $Q$48. Above threshold, a state appears just above the top of the lower polariton band, with participation ratio independent of chain length $Q$49; for $Q$50 and $Q$51, $Q$52, compared with $Q$53 for the nearest-neighbor model without edge localization. The state is explicitly non-topological, since the Zak phase of the Bloch Hamiltonian is trivial (Downing et al., 2020).

7. Functional consequences, applications, and unresolved issues

The main scientific interest of polaritonic surface cavities lies in how they reshape transport, spectroscopy, and chemistry under ambient-compatible conditions. Reported consequences include polariton-mediated energy delocalization across microns, directly observed by two-dimensional electronic spectroscopy, ballistic propagation at tens of $Q$54, long-lived Bloch surface-wave polaritons enabling photochemical charge transfer, and vacuum-field engineering of electron mobility in mirrorless cavity MOSFETs. In plasmonic chemistry, strong coupling can coexist with hot-carrier generation, and multiscale simulations indicate that plasmon-induced hybrid states may facilitate selective charge injection, for example in CO$Q$55 methanation on Rh nanocubes. In liquid-phase vibrational strong coupling, non-local plasmonic metasurfaces acting as open single-surface cavities accelerate the solvolysis of \textit{para}-nitrophenyl acetate by a factor of $Q$56, with $Q$57 compared with $Q$58 outside the cavity and $Q$59 for a detuned metasurface. In correlated-electron physics, off-resonant cavity control is governed not by a single resonant mode but by a generalized Purcell factor proportional to the frequency-integrated photonic density of states; on that basis, polaritonic surface cavities are predicted to modify the Hubbard-model exchange $Q$60 far more effectively than standard Fabry–Perot resonators, with percent-level enhancements for a gold substrate at nanometer separations and a corresponding two-magnon Raman shift $Q$61 (Johns et al., 2 May 2026, Verdelli et al., 2024, Grunwald et al., 19 Mar 2026).

Several limitations are equally central. Plasmonic and open cavities suffer from short lifetimes and high loss; dark-state reservoirs complicate dynamical assignments; spectrally hidden level splitting requires momentum-resolved photoluminescence or quasi-normal-mode analysis; sub-nm gaps are highly sensitive to roughness and defects; and nonlocal or quantum-surface effects can matter in extreme hotspots. On the molecular side, exact-factorization results show that weighted polaritonic surfaces can fail whenever several polaritonic surfaces contribute simultaneously, so the exact time-dependent potential must sometimes replace the adiabatic picture. For realistic nanophotonic devices, full-field simulations likewise reveal strong position and orientation dependence in local hybridization, including middle-polariton features from detuned impurities that may remain dark in transmission. This suggests that the central unresolved problems are not merely achieving strong coupling, but determining which mode families, which dark-state manifolds, and which local field distributions actually control transport, reaction coordinates, and many-body observables in each platform (Martinez et al., 2020, Bustamante et al., 26 Aug 2025, Grunwald et al., 19 Mar 2026).