Crystal Polaritons: Fundamentals & Applications

• Crystal polaritons are hybrid light–matter quasiparticles arising from strong coupling between photons and excitations in periodic media.
• They enable precise engineering of dispersion, symmetry, and loss through designed dielectric structures and anisotropic materials.
• Advanced spectroscopy and near-field nanoimaging reveal their complex band structure and dynamic light–matter interactions.

Crystal polaritons are hybrid light–matter quasiparticles whose dispersion is set by the joint dynamics of electromagnetic fields and collective excitations in a crystalline or periodic medium. In the historical sense, the term denotes phonon–polaritons and exciton–polaritons in bulk ionic or excitonic crystals, where photons strongly couple to optical phonons or excitons; in modern usage it also encompasses excitons coupled to Bloch modes of photonic crystal slabs, polaritons in anisotropic van der Waals crystals, and Bloch polaritons in engineered polaritonic lattices (0907.2813, Lerose et al., 2013, Zanotti et al., 2022). Across these realizations, the defining feature is that the “crystal” is not merely a host medium but a structure—atomic or photonic—that shapes the allowed polaritonic bands, their symmetry, and their losses.

1. Historical scope and terminology

Historically, crystal polaritons arose in bulk crystals in which infrared or optical photons hybridize with lattice vibrations or excitons. In ionic crystals, photons couple to optical phonons to form phonon–polaritons; in excitonic crystals, photons couple to Wannier excitons to form exciton–polaritons (0907.2813). A microscopic classical treatment of ionic crystals makes this precise by identifying crystal polaritons as collective normal modes of the full system “ions + electromagnetic field,” in which transverse optical phonons and photons hybridize, whereas longitudinal optical modes remain distinct in the conventional picture (Lerose et al., 2013).

Modern work extends the term beyond atomic lattices. In a two-dimensional photonic crystal slab containing quantum wells, the “crystal” can be a lithographically defined dielectric lattice, and the polaritons are mixed exciton–photon modes whose photonic component is a photonic-crystal Bloch mode (0907.2813). More generally, periodically patterned multilayer waveguides support “photonic crystal polaritons,” hybrid radiation–matter excitations whose band structure inherits the full symmetry of the underlying lattice and whose losses can be treated within a non-Hermitian Hopfield framework (Zanotti et al., 2022). This broader usage preserves the central idea of crystal polaritons while shifting emphasis from the atomic lattice alone to any periodic environment that imposes Bloch structure on the electromagnetic component.

This dual usage is now standard. In bulk materials, the crystal sets the microscopic dipolar resonance and long-range Coulomb coupling. In patterned photonic systems, the periodic dielectric environment controls Bloch dispersion, bound states in the continuum, and symmetry-protected selection rules. In anisotropic van der Waals materials, the dielectric tensor of the crystal itself sets the topology of the isofrequency contours, producing hyperbolic, canalized, and shear polaritons (Duan et al., 2021, Díaz-Núñez et al., 31 Jan 2025).

2. Microscopic descriptions and gauge structure

At microscopic level, crystal polaritons can be derived without postulating a macroscopic polarization field. A classical first-principles treatment for ionic crystals models ions as point charges subject to short-range repulsion, Coulomb forces, retarded dipole fields, and radiation reaction. The normal-mode problem is then formulated for the full retarded dynamical matrix, and the polaritonic splitting about the electromagnetic dispersion line $\omega = ck$ emerges directly from retardation (Lerose et al., 2013). In this description, the most important term is the long-distance reciprocal-space contribution with denominator $k^2 - \omega^2/c^2$, which produces the avoided crossing between the transverse optical phonon and the photon.

The usual classification into transverse and longitudinal modes becomes subtler when gauge choice is made explicit. In Coulomb gauge, transverse matter excitations couple to transverse photons and form the standard upper and lower polariton branches, while longitudinal excitations are treated as pure matter modes dressed by Coulomb interactions. In Lorenz gauge, however, both transverse and longitudinal excitations become polaritons: transverse dipoles couple to transverse photons, whereas longitudinal dipoles couple to longitudinal and scalar photons (Barros et al., 2023). In that formulation, the long-wavelength limit restores the expected three-fold degeneracy of dipolar excitations in an isotropic three-dimensional crystal, and the usual longitudinal–transverse splitting is reinterpreted as a consequence of how light–matter dressing is partitioned by the gauge choice.

This reformulation has direct implications for near-field optics. Longitudinal polaritons are predicted to interact with longitudinal near fields at surfaces, which provides additional excitation channels for scanning near-field microscopy and surface-enhanced spectroscopy (Barros et al., 2023). A common misconception is therefore that only transverse crystal excitations are genuinely polaritonic. The covariant treatment suggests instead that longitudinal collective modes in crystals are also hybrid light–matter states, even if their photonic content is encoded in longitudinal and scalar components rather than free-space transverse radiation.

3. Periodic photonic implementations

A central modern realization of crystal polaritons is the exciton–photon strong-coupling regime in photonic crystal slabs. In a GaAs/AlGaAs slab containing three 8 nm In$_{0.05}$Ga$_{0.95}$As quantum wells, strong coupling between the heavy-hole exciton at $E_X \simeq 1.485~\text{eV}$ and two-dimensional photonic-crystal modes was established by resonant scattering and photoluminescence. The system exhibited a clear anticrossing, a Rabi splitting of about $7~\text{meV}$, and polariton linewidths below $1~\text{meV}$; changing the lattice constant by only $2\%$ drastically reshaped the polariton branches, including a distinctive “diamond-like” dispersion unlike the usual planar-microcavity $S$-shape (0907.2813). This showed that the photonic lattice can be used to engineer polariton band curvature and phase-matching conditions.

A generalized theory for such systems is provided by the non-Hermitian Hopfield treatment of periodically patterned multilayer waveguides. In that framework, photonic Bloch bands from guided-mode expansion are coupled to excitonic Bloch bands in a generalized Hopfield matrix that includes radiative and excitonic losses. The same formalism applies to GaAs/AlGaAs multi-quantum-well slabs, perovskite metasurfaces, and structures designed to host bound states in the continuum, and it predicts that detailed polariton dispersion and loss depend on both material composition and lattice symmetry (Zanotti et al., 2022). In particular, excitonic coupling to photonic BICs can produce polariton BICs whose radiative linewidth vanishes while the residual linewidth is limited by excitonic non-radiative loss.

One-dimensional photonic crystal slabs have provided an especially compact route to exciton–polaritons in atomically thin semiconductors. In monolayer WSe$_2$ and WS$k^2 - \omega^2/c^2$0 transferred onto SiN photonic-crystal gratings, angle-resolved reflectance and photoluminescence revealed strong coupling between the monolayer A exciton and the photonic-crystal mode, with room-temperature anti-crossing in WS$k^2 - \omega^2/c^2$1, highly anisotropic dispersion, and adjustable Fano resonances in reflectance (Zhang et al., 2017). The same architecture was later used to realize electrostatically gated photonic-crystal polaritons in monolayer MoSe$k^2 - \omega^2/c^2$2, where modulation of the dielectric environment within the unit cell created two neutral excitons and two trions, and gate voltage continuously tuned the system between neutral-exciton and trion-polariton regimes (Khestanova et al., 2024). In that device, two resonances with significantly different nonlinear response enabled optical switching with ultrashort laser pulses.

These implementations establish a distinctive branch of crystal-polariton physics: the matter resonance may remain atomically thin or quantum-well-like, but the photonic component is a Bloch mode of a periodic optical crystal. The result is a polariton band structure with engineered anisotropy, controllable leakage, symmetry selection rules, and access to BIC physics, none of which are available in the same form in a planar Fabry–Pérot microcavity.

4. Anisotropic, hyperbolic, canalized, and image crystal polaritons

A second major branch of the subject concerns polaritons governed primarily by the anisotropic dielectric tensor of a crystal. In biaxial $k^2 - \omega^2/c^2$3-MoO$k^2 - \omega^2/c^2$4, phonon polaritons in the $k^2 - \omega^2/c^2$5–$k^2 - \omega^2/c^2$6 range were directly visualized with in-plane hyperbolic dispersion, confinement below $k^2 - \omega^2/c^2$7, and a lower limit on lifetime of more than $k^2 - \omega^2/c^2$8 (Oliveira et al., 2020). In the mid-infrared, the same material supports in-plane hyperbolic phonon polaritons whose allowed propagation sectors are determined by the topology of the isofrequency contour. When the slab is placed on 4H-SiC, a low-loss optical topological transition near the SiC surface optical phonon rotates the hyperbolic response and enables propagation along directions that are forbidden in bare $k^2 - \omega^2/c^2$9-MoO$_{0.05}$0 (Duan et al., 2021).

Loss and anisotropy introduce an additional level of structure. In $_{0.05}$1-MoO$_{0.05}$2, the complex wavevector $_{0.05}$3 is generically non-collinear: the directions of phase propagation and exponential decay do not coincide. Near-field nanoimaging showed that, for polaritons launched by a localized source, the physically selected decay vector satisfies $_{0.05}$4 rather than $_{0.05}$5, and ignoring this misalignment can produce errors in the out-of-plane wavevector of about $_{0.05}$6 (Voronin et al., 21 Aug 2025). The same analysis yielded the low-loss lifetime formula $_{0.05}$7, clarifying how decay, energy flow, and complex dispersion are related in lossy anisotropic crystals.

Low-symmetry crystals broaden the taxonomy further. In exfoliated gypsum thin films, monoclinic symmetry and non-orthogonal phonon resonances generate hyperbolic shear, canalized shear, and elliptical shear phonon polaritons; the measured group velocity reached values as low as $_{0.05}$8 (Díaz-Núñez et al., 31 Jan 2025). In biaxial MoOCl$_{0.05}$9, intrinsic plasmon polaritons undergo an elliptical-to-hyperbolic transition that produces room-temperature canalization in the $_{0.95}$0–$_{0.95}$1 range, and the canalization wavelength can be tuned by more than $_{0.95}$2 by changing the flake thickness (Aghashirinov et al., 9 Jun 2026). The same crystal also supports visible-range hyperbolic plasmon polaritons: in the $_{0.95}$3–$_{0.95}$4 interval one in-plane permittivity component is negative and the other positive, and gold rod nanoantennas launch directional, V-shaped polariton beams with up to a four-fold increase in launching efficiency at Fabry–Pérot resonance (Clemente-Marcuello et al., 9 Feb 2026).

Periodic modulation can be imposed without etching the polaritonic host. A twist-tunable in-plane anisotropic polaritonic crystal was realized by placing a pristine $_{0.95}$5-MoO$_{0.95}$6 flake atop a metallic hole array; this preserves the intrinsic low-loss phonon-polariton response while introducing Bloch modes whose orientation rotates with the twist angle (Capote-Robayna et al., 2024). A complementary “polaritonic Fourier crystal” in hBN uses harmonic modulation of the image-polariton momentum rather than binary nanostructuring, producing a clean Bloch band structure dominated by the first-order mode and a polaritonic bandgap even in naturally abundant hBN (Menabde et al., 2024). In parallel, the broader class of image polaritons—modes formed when a polaritonic vdW material is brought close to a highly conductive metal—shows exceptional field compression and lower normalized propagation loss than conventional modes on non-metallic substrates (Menabde et al., 2021). Together, these results show that crystal polaritons can be engineered not only by the intrinsic dielectric tensor but also by mirror coupling, periodic metallic screening, and mechanical twist.

5. Spectroscopy, nanoimaging, and extraction of complex band structure

Crystal polaritons are observed through a combination of angle-resolved far-field spectroscopy and near-field nanoimaging. In photonic crystal slabs, resonant scattering, crossed-polarization reflectance, and angle-resolved photoluminescence directly reveal anticrossings between a nearly flat exciton line and dispersive photonic bands (0907.2813, Zhang et al., 2017). In patterned multilayer waveguides, rigorous coupled-wave analysis and scattering-matrix methods complement the Hopfield description by connecting complex eigenfrequencies to Fano-like optical spectra and polarization-dependent selection rules (Zanotti et al., 2022).

In van der Waals crystals and other deeply subwavelength systems, the dominant tool is scattering-type scanning near-field optical microscopy. In $_{0.95}$7-MoO$_{0.95}$8, hBN, gypsum, and MoOCl$_{0.95}$9, s-SNOM images the complex near field launched by a tip, disk, or gold nanoantenna and reconstructs the polariton wavelength from fringe spacing and the isofrequency contour from spatial Fourier transforms (Oliveira et al., 2020, Menabde et al., 2024, Díaz-Núñez et al., 31 Jan 2025, Clemente-Marcuello et al., 9 Feb 2026). In anisotropic crystals the interpretation of such images requires care. The self-consistent method introduced for $E_X \simeq 1.485~\text{eV}$0-MoO$E_X \simeq 1.485~\text{eV}$1 first determines the direction of $E_X \simeq 1.485~\text{eV}$2 from wavefront normals, then imposes $E_X \simeq 1.485~\text{eV}$3 for the dominant far-field contribution of a localized source, and finally fits the complex field profile along each ray to extract the full complex wavevector (Voronin et al., 21 Aug 2025). This procedure resolves a widespread ambiguity: the attenuation extracted from fringe contrast depends only on the projection of $E_X \simeq 1.485~\text{eV}$4 along the observation direction, so a scalar propagation length can be correct even when the full direction of decay is misidentified.

The experimental observables depend strongly on geometry. Antennas and hole arrays provide momentum matching for visible and mid-infrared plasmon polaritons in MoOCl$E_X \simeq 1.485~\text{eV}$5 (Clemente-Marcuello et al., 9 Feb 2026); drilled rings and hole-array crystals launch Bloch phonon polaritons in twist-tunable $E_X \simeq 1.485~\text{eV}$6-MoO$E_X \simeq 1.485~\text{eV}$7 structures (Capote-Robayna et al., 2024); metallic disks beneath gypsum reveal the continuous transition from hyperbolic to canalized to elliptical shear regimes (Díaz-Núñez et al., 31 Jan 2025). In all of these cases, the measured near-field pattern is not merely a visualization tool but a direct map of the underlying complex band structure.

6. Functional regimes, misconceptions, and outlook

The practical importance of crystal polaritons lies in the fact that their dispersion is highly designable while their matter component supplies strong nonlinearity or magnetic, excitonic, or phononic selectivity. In photonic crystal slabs, tuning the lattice period reshapes the polariton bands and was explicitly proposed as a route toward entangled photon-pair generation through polariton stimulated scattering (0907.2813). In generalized photonic-crystal-polariton platforms, excitonic coupling to BICs yields long-lived propagating modes and topological polarization vortices, with potential applications ranging from low-threshold lasers to symmetry-protected hybrid states (Zanotti et al., 2022). In gated monolayer-semiconductor devices, electrically controlled redistribution of oscillator strength between neutral and charged excitons produces strong polariton nonlinearity and optical switching with ultrashort laser pulses (Khestanova et al., 2024).

Anisotropic crystal polaritons underpin a different application space: canalization-based imaging, nanoscale waveguiding, directional couplers, nanofocusing, hyperbolic reflection, and sensing in frequency windows overlapping vibrational resonances (Voronin et al., 21 Aug 2025, Aghashirinov et al., 9 Jun 2026). Visible-range hyperbolic plasmon polaritons in MoOCl$E_X \simeq 1.485~\text{eV}$8 bring this functionality into the $E_X \simeq 1.485~\text{eV}$9–$7~\text{meV}$0 range, while low-loss twist-tunable polaritonic crystals in $7~\text{meV}$1-MoO$7~\text{meV}$2 suggest mechanically reconfigurable nano-optical components for lasing, sensing, and energy harvesting (Clemente-Marcuello et al., 9 Feb 2026, Capote-Robayna et al., 2024). The Fourier-crystal concept in hBN suggests that smooth momentum-space modulation may become a general strategy for suppressing the severe scattering losses associated with sharply etched polaritonic crystals (Menabde et al., 2024).

Two recurring misconceptions have been corrected by recent work. First, in anisotropic lossy crystals the directions of propagation and decay are generally not the same, so interpreting near-field fringes with $7~\text{meV}$3 can distort the inferred out-of-plane field structure (Voronin et al., 21 Aug 2025). Second, longitudinal collective modes in crystals need not be regarded as purely matter excitations; in Lorenz gauge they are longitudinal polaritons that couple to longitudinal and scalar photons and therefore belong within the broader taxonomy of crystal polaritons (Barros et al., 2023).

A further step toward active control has already been demonstrated in CrSBr photonic crystal slabs, where self-hybridized exciton–polaritons track the antiferromagnetic-to-ferromagnetic spin-flip transition and the sign of the group velocity can be reversed by a change of only $7~\text{meV}$4 in the external magnetic field (Gorelkina et al., 13 Apr 2026). This suggests that future crystal-polariton devices will not be limited to passive band engineering. They may combine periodic photonic structuring, intrinsic crystal anisotropy, electrostatic tuning, magnetic switching, and near-field launching in a single platform, while retaining the defining feature of the subject: polariton bands whose properties are inseparable from the symmetry and structure of the crystal environment.