BIC-Based Sensing Strategies

• BIC-based sensing is a method that exploits bound states in the continuum to achieve strong field confinement and exceptionally narrow resonances via controlled symmetry breaking.
• The approach enables diverse sensing modalities—including refractive-index, SEIRAS, gas, and biosensing—by modulating radiation loss with precise structural perturbations.
• Optimization of materials, asymmetry parameters, and environmental conditions is key to balancing quality factors and maximizing sensor performance across different spectral regimes.

Searching arXiv for recent BIC-based sensing papers to ground the article. BIC-based sensing denotes sensing strategies that exploit bound states in the continuum (BICs) and their experimentally accessible leaky counterparts, quasi-BICs, to combine strong field confinement with exceptionally narrow resonances. In photonic, plasmonic, hybrid, terahertz, mid-infrared, visible, and acoustic implementations, the central idea is to suppress radiation loss by symmetry or modal interference, then reopen a weak radiation channel for readout by controlled perturbation. The resulting resonances support spectral-shift, absorbance, amplitude, phase, linewidth, angular, and multi-parameter sensing modalities, and have been used for refractive-index sensing, SEIRAS, gas detection, biosensing, and single-particle detection (Maksimov et al., 2020, Jiang et al., 23 Jun 2025, Srivastava et al., 2019, Watanabe et al., 12 Mar 2026, Nazarov et al., 21 Jul 2025, Wu et al., 15 Jul 2025).

1. Physical basis of BIC-enabled sensing

A BIC is a non-radiating eigenmode embedded in the radiation continuum. In ideal lossless and infinite systems it has infinite radiative lifetime and, equivalently, infinite radiative quality factor. In sensing platforms, such states are usually accessed indirectly through quasi-BICs, produced by a small perturbation that opens a weak leakage channel while retaining a high $Q$ factor and a strongly localized near field (Srivastava et al., 2019, Maksimov et al., 2020). For symmetry-protected BICs, the perturbation is commonly a structural asymmetry parameter such as a gap displacement, a rod-length imbalance, a tilt angle, or an off-centered channel; in many implementations the radiative quality factor follows the inverse-square law $Q_{\mathrm{rad}} \propto 1/\alpha^2$ or equivalent forms such as $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$ (Srivastava et al., 2019, Kumar et al., 8 Jul 2025, Watanabe et al., 12 Mar 2026).

Two additional mechanisms recur across the literature. Friedrich–Wintgen BICs arise from destructive interference between coupled resonances sharing the same radiation channel; in the acoustic–solid case, localized cavity modes interfere with Fabry–Pérot standing-wave modes sustained by thin aluminum plates, producing an FW-type quasi-BIC when the port radiation is cancelled (Wu et al., 15 Jul 2025). In hybrid dielectric–plasmonic waveguides, destructive interference between a dielectric waveguide mode and a surface plasmon polariton yields a topologically protected hybrid BIC whose radiative channel is suppressed while the modal sensitivity remains plasmon-enhanced (Meudt et al., 2019).

The BIC concept is not restricted to large periodic arrays. A boundary-induced embedded eigenstate has been realized in a single metallic disk resonator placed inside a rectangular metallic waveguide, where lateral perfect electric conductor walls create an image array that reproduces the radiation-cancellation condition of an extended array while localizing the field to a single site (Jacobsen et al., 2021). Topological descriptions are also common: off-$\Gamma$ BICs appear as polarization vortices in $k$-space with quantized charge, and the motion of these vortices under environmental perturbation can itself be used as a sensing observable (Nazarov et al., 21 Jul 2025, Meudt et al., 2019).

2. Sensing observables and figures of merit

The most conventional BIC-sensing observable is the resonance shift induced by a perturbation of refractive index or permittivity. For dielectric gratings, perturbation theory yields a differential sensitivity $S = d\lambda/dn$, and the maximal sensitivity of an isolated BIC is given as $S_{\max} = \lambda_{\mathrm{BIC}} n_c$ when the modal energy is maximally shifted into the cladding region (Maksimov et al., 2020). Closely related definitions recur across the literature, including $S = \Delta \lambda/\Delta n$, $S = \partial f_{\mathrm{qBIC}}/\partial t$, $S_t = \Delta f/\Delta t$, $Q_{\mathrm{rad}} \propto 1/\alpha^2$0, and $Q_{\mathrm{rad}} \propto 1/\alpha^2$1 for gas concentration (Sarkar et al., 12 Oct 2025, Srivastava et al., 2019, Wu et al., 15 Jul 2025).

BIC platforms also support non-spectral readout. In mid-infrared SEIRAS, one study defines the sensing metric as the reflectance-based absorbance

$Q_{\mathrm{rad}} \propto 1/\alpha^2$2

with $Q_{\mathrm{rad}} \propto 1/\alpha^2$3 the bare metasurface reflectance in solvent and $Q_{\mathrm{rad}} \propto 1/\alpha^2$4 the reflectance after coating a thin analyte layer; the peak value $Q_{\mathrm{rad}} \propto 1/\alpha^2$5 at $Q_{\mathrm{rad}} \propto 1/\alpha^2$6 is used as the sensitivity proxy rather than a spectral shift (Jiang et al., 23 Jun 2025). In terahertz thin-film sensing, differential observables are defined as $Q_{\mathrm{rad}} \propto 1/\alpha^2$7 and $Q_{\mathrm{rad}} \propto 1/\alpha^2$8, enabling detection even when the frequency shift approaches the instrumental resolution (Srivastava et al., 2019).

Several recent works move further toward amplitude- or contrast-based sensing. In a loss-driven 3D BIC metasurface, intensity sensitivity is defined as

$Q_{\mathrm{rad}} \propto 1/\alpha^2$9

and the platform reports peak intensity sensitivity up to $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$0 while remaining compatible with $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$1 bandwidth LEDs (Sun et al., 2024). In a permittivity-asymmetric qBIC, the single-wavelength transmittance sensitivity is written

$Q_{\mathrm{rad}} = Q_0 \delta^{-2}$2

and numerical results reach $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$3 under single-wavelength conditions (Yang et al., 29 Aug 2025). In hybrid BIC waveguides the figure of merit is defined as $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$4, and in diffraction-mode readout it reduces to $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$5, which the study argues is the relevant metric near the hybrid BIC rather than a purely spectral $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$6 measure (Meudt et al., 2019).

Angular and multi-parameter observables are equally important. In polarization-vortex sensing, the angular sensitivity $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$7 can exceed the spectral sensitivity $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$8 near $Q_{\mathrm{rad}} = Q_0 \delta^{-2}$9, with the vortex position obeying a square-root law $\Gamma$0 and reported values up to $\Gamma$1 in favorable directions (Nazarov et al., 21 Jul 2025). In single-particle qBIC sensing, binding events are tracked not only by redshifts but also by linewidth and amplitude steps, since local binding modifies the effective asymmetry factor and thus the radiative coupling (Watanabe et al., 12 Mar 2026).

3. Materials, asymmetry engineering, and loss channels

The dominant design problem in BIC-based sensing is not simply how to maximize $\Gamma$2, but how to distribute radiative, intrinsic, and environmental losses while preserving analyte overlap. A recurring decomposition is

$\Gamma$3

or, equivalently, $\Gamma$4, with $\Gamma$5 controlled by asymmetry and $\Gamma$6 limited by material absorption, scattering, or substrate loss (Kumar et al., 8 Jul 2025, Watanabe et al., 12 Mar 2026). In practical metasurfaces, asymmetry is implemented as a tilt angle $\Gamma$7, a length imbalance $\Gamma$8, a gap displacement $\Gamma$9, or a lateral offset $k$0, all of which tune radiative damping while leaving the basic BIC mode profile recognizable (Jiang et al., 23 Jun 2025, Srivastava et al., 2019, Sharma et al., 4 Sep 2025).

Material choice sets the trade-off between field localization and intrinsic loss. Dielectric qBICs in silicon or GaP can support very high $k$1 in low-loss environments, but their near fields are often distributed through the resonator volume and may be strongly quenched by absorptive surroundings (Jiang et al., 23 Jun 2025, Sarkar et al., 12 Oct 2025). Plasmonic qBICs in gold have higher ohmic loss, yet their fields concentrate into nanoscale hotspots at surfaces, tips, or gaps, which improves overlap with ultrathin analytes and can make them more robust in lossy solvents (Jiang et al., 23 Jun 2025). Hybrid structures deliberately combine these attributes. In MXene-coated silicon nanodisks, symmetry protection suppresses $k$2 so that the lossy MXene layer can be balanced against radiation leakage, yielding narrow high-absorption resonances near $k$3 despite the broadband plasmonic response of Ti$k$4C$k$5 (Sharma et al., 4 Sep 2025). In hybrid dielectric–plasmonic waveguides, the BIC preserves dielectric-like radiative suppression while inheriting plasmonic analyte overlap (Meudt et al., 2019).

The environment can become the dominant loss channel. A controlled comparison of silicon and gold qBIC metasurfaces for mid-IR SEIRAS shows that solvent imaginary refractive index $k$6 governs the crossover between dielectric and plasmonic superiority. In air or very low-loss media, silicon provides higher $k$7, but for $k$8 at $k$9, gold yields larger optimized absorbance, and at $S = d\lambda/dn$0 its mean field exceeds silicon’s (Jiang et al., 23 Jun 2025). Related environmental-loss management appears in water-compatible near-infrared qBIC sensing, where heavy water is used to reduce intrinsic optical loss and permit an experimental $S = d\lambda/dn$1 at critical coupling (Watanabe et al., 12 Mar 2026), and in terahertz sensing, where low-index, low-loss cyclic olefin copolymer supports narrower quasi-BIC resonances than Kapton (Srivastava et al., 2019).

4. Representative implementations and measured performance

A wide range of BIC-enabled sensors has been demonstrated or numerically validated, spanning thin-film, molecular, gas, biosensing, and single-particle regimes.

Platform and regime	Readout modality	Representative result
THz asymmetric split-ring metasurface on COC (Srivastava et al., 2019)	Frequency shift, $S = d\lambda/dn$2, $S = d\lambda/dn$3	Minimum detected Ge thickness $S = d\lambda/dn$4, corresponding to $S = d\lambda/dn$5; measured $S = d\lambda/dn$6 and $S = d\lambda/dn$7 for $S = d\lambda/dn$8
Mid-IR Au and Si qBIC metasurfaces for SEIRAS (Jiang et al., 23 Jun 2025)	Reflectance-based absorbance at $S = d\lambda/dn$9	Optimized absorbance crossover at $S_{\max} = \lambda_{\mathrm{BIC}} n_c$0; field crossover at $S_{\max} = \lambda_{\mathrm{BIC}} n_c$1
GaP bow-tie nanohole dielectric metasurface (Sarkar et al., 12 Oct 2025)	Bulk RI shift	Sensitivity up to $S_{\max} = \lambda_{\mathrm{BIC}} n_c$2; FOM up to $S_{\max} = \lambda_{\mathrm{BIC}} n_c$3; $S_{\max} = \lambda_{\mathrm{BIC}} n_c$4 up to $S_{\max} = \lambda_{\mathrm{BIC}} n_c$5
Low-contrast silicon qBIC metasurface in D$S_{\max} = \lambda_{\mathrm{BIC}} n_c$6O (Watanabe et al., 12 Mar 2026)	Wavelength, linewidth, amplitude steps	Experimental $S_{\max} = \lambda_{\mathrm{BIC}} n_c$7; $S_{\max} = \lambda_{\mathrm{BIC}} n_c$8 particles give mean $S_{\max} = \lambda_{\mathrm{BIC}} n_c$9
Polarization-vortex dielectric rod array (Nazarov et al., 21 Jul 2025)	Angular and spectral tracking of BIC motion	$S = \Delta \lambda/\Delta n$0 up to $S = \Delta \lambda/\Delta n$1; $S = \Delta \lambda/\Delta n$2 up to $S = \Delta \lambda/\Delta n$3
MXene-coated absorptive dielectric metasurface (Sharma et al., 4 Sep 2025)	Bulk RI and methane-induced wavelength shift	$S = \Delta \lambda/\Delta n$4, FOM $S = \Delta \lambda/\Delta n$5, methane sensitivity $S = \Delta \lambda/\Delta n$6 per unit percentage concentration
Quasi-BIC laser with double HCG resonator (Zhang et al., 2021)	Lasing wavelength shift in gas	Threshold fluence $S = \Delta \lambda/\Delta n$7; sensitivity $S = \Delta \lambda/\Delta n$8; FOM $S = \Delta \lambda/\Delta n$9
Loss-driven 3D BIC biosensor (Sun et al., 2024)	Peak-intensity readout	Peak intensity sensitivity up to $S = \partial f_{\mathrm{qBIC}}/\partial t$0; bulk LOD $S = \partial f_{\mathrm{qBIC}}/\partial t$1; extracellular vesicle LOD $S = \partial f_{\mathrm{qBIC}}/\partial t$2
Permittivity-asymmetric $S = \partial f_{\mathrm{qBIC}}/\partial t$3-qBIC metasurface (Yang et al., 29 Aug 2025)	Single-wavelength transmittance	$S = \partial f_{\mathrm{qBIC}}/\partial t$4; linear window area $S = \partial f_{\mathrm{qBIC}}/\partial t$5 times larger than geometry-asymmetric qBIC
Boundary-induced single-resonator BIC (Jacobsen et al., 2021)	Reflection shift versus liquid volume or salinity	Volume resolution less than $S = \partial f_{\mathrm{qBIC}}/\partial t$6; evaporation rates $S = \partial f_{\mathrm{qBIC}}/\partial t$7 for distilled water and $S = \partial f_{\mathrm{qBIC}}/\partial t$8 for $S = \partial f_{\mathrm{qBIC}}/\partial t$9 NaCl
Acoustic–solid FW quasi-BIC cavity (Wu et al., 15 Jul 2025)	Acoustic resonance shift with gas concentration	Measured $S_t = \Delta f/\Delta t$0; simulated $S_t = \Delta f/\Delta t$1 at the FW-BIC point; sensitivity $S_t = \Delta f/\Delta t$2 and $S_t = \Delta f/\Delta t$3 for the $S_t = \Delta f/\Delta t$4 cavity case

Taken together, these implementations show that BIC-based sensing is not confined to one spectral band or one readout strategy. It spans absorbance-enhanced infrared spectroscopy in liquids, thin-film detection deep below the free-space wavelength, active laser-based gas sensing, topological angular tracking, intensity-only interrogation, and single-particle event detection (Jiang et al., 23 Jun 2025, Srivastava et al., 2019, Zhang et al., 2021, Nazarov et al., 21 Jul 2025, Sun et al., 2024, Watanabe et al., 12 Mar 2026). A particularly notable transition is from bulk refractive-index sensing to local perturbation sensing: low-contrast silicon metasurfaces have resolved step-like signatures of individual $S_t = \Delta f/\Delta t$5 particles, including simultaneous wavelength, linewidth, and amplitude changes (Watanabe et al., 12 Mar 2026).

5. Optimization principles and recurring controversies

A recurrent but incomplete assumption in the field is that maximum $S_t = \Delta f/\Delta t$6 or critical coupling automatically yields the best sensor. Several studies qualify that view. For dielectric gratings, $S_t = \Delta f/\Delta t$7 depends on cladding overlap rather than $S_t = \Delta f/\Delta t$8 alone, and the analytic bound $S_t = \Delta f/\Delta t$9 makes clear that modal energy placement is as important as linewidth (Maksimov et al., 2020). In THz metallic qBIC sensors, a dedicated analysis of limit of detection shows a non-monotonic dependence on asymmetry and predicts the optimum not at $Q_{\mathrm{rad}} \propto 1/\alpha^2$00 but below critical coupling. In the constant-sensitivity model the optimum occurs at $Q_{\mathrm{rad}} \propto 1/\alpha^2$01, corresponding to $Q_{\mathrm{rad}} \propto 1/\alpha^2$02, and measured FoM curves peak in the window $Q_{\mathrm{rad}} \propto 1/\alpha^2$03 for transmission sensing (Kumar et al., 8 Jul 2025).

A second recurring issue is whether dielectrics are always preferable because they can achieve higher $Q_{\mathrm{rad}} \propto 1/\alpha^2$04. The mid-IR SEIRAS comparison of silicon and gold qBIC metasurfaces shows that this is environment dependent. Silicon is superior for $Q_{\mathrm{rad}} \propto 1/\alpha^2$05, but gold becomes superior for $Q_{\mathrm{rad}} \propto 1/\alpha^2$06, and the optimal asymmetry angles differ by platform even in the same solvent; at $Q_{\mathrm{rad}} \propto 1/\alpha^2$07, $Q_{\mathrm{rad}} \propto 1/\alpha^2$08 whereas $Q_{\mathrm{rad}} \propto 1/\alpha^2$09 (Jiang et al., 23 Jun 2025). The implication is that asymmetry must be optimized jointly with material, analyte thickness, and solvent loss rather than transferred across platforms.

A third design shift concerns how asymmetry is encoded. Conventional geometry-asymmetric qBICs mainly respond by a horizontal translation of the resonance spectrum. In contrast, permittivity-asymmetric qBICs use the environment itself as the asymmetry factor, $Q_{\mathrm{rad}} \propto 1/\alpha^2$10, so refractive-index changes simultaneously modulate resonance position and modulation depth. This produces a substantially wider linear operating region for single-wavelength readout and can even optically restore a geometry-broken qBIC to an ultra-high-$Q_{\mathrm{rad}} \propto 1/\alpha^2$11 state with $Q_{\mathrm{rad}} \propto 1/\alpha^2$12 when the total asymmetry is cancelled (Yang et al., 29 Aug 2025).

High-$Q_{\mathrm{rad}} \propto 1/\alpha^2$13 operation is also not synonymous with wavelength-interrogated sensing. A loss-driven 3D BIC biosensor explicitly argues that increasing $Q_{\mathrm{rad}} \propto 1/\alpha^2$14 in a conventional wavelength-shift sensor can deteriorate wavelength sensitivity and complicate the system, whereas Q-switched intensity sensing uses the refractive-index perturbation to modulate radiative damping and thereby boost peak intensity sensitivity (Sun et al., 2024). That viewpoint complements the THz review’s observation that high-$Q_{\mathrm{rad}} \propto 1/\alpha^2$15, low-intensity modes are advantageous for drip-dry sensing, while lower-$Q_{\mathrm{rad}} \propto 1/\alpha^2$16, higher-intensity modes can be preferable for liquid-phase absorption sensing (Abdulaal et al., 30 Sep 2025).

6. Limitations, robustness, and emerging directions

The principal limitations of BIC-based sensing remain intrinsic loss, environmental absorption, fabrication tolerances, and readout stability. Metallic structures suffer ohmic loss; dielectric structures can be severely quenched by lossy solvents; and in both cases finite size, disorder, roughness, and imperfect asymmetry control cap the accessible $Q_{\mathrm{rad}} \propto 1/\alpha^2$17 factor (Jiang et al., 23 Jun 2025, Kumar et al., 8 Jul 2025, Abdulaal et al., 30 Sep 2025). In water-rich environments, terahertz absorption is so strong that liquid handling must be redesigned through low-path microcapillaries, ultrathin films, or alternative assay geometries (Abdulaal et al., 30 Sep 2025). Optical implementations also inherit temperature cross-sensitivity, angle sensitivity, and source or detector noise, which become increasingly consequential as the quasi-BIC linewidth collapses (Yang et al., 29 Aug 2025, Sun et al., 2024).

At the same time, the field has diversified into architectures that explicitly address those constraints. Flexible low-loss polymer substrates and large-mechanical-strength qBIC metamaterials have been proposed for wearable THz sensing (Srivastava et al., 2019). Boundary-induced embedded eigenstates reduce footprint and disorder sensitivity by eliminating the need for large periodic arrays (Jacobsen et al., 2021). Low-contrast shallow-etched silicon metasurfaces provide free-space access, large-area illumination, and position-insensitive collective response for single-particle sensing in microfluidic environments (Watanabe et al., 12 Mar 2026). Active control has also become a design goal: polarization rotation in dielectric metasurfaces provides on/off switching and multiplexed resonances, while thermo-optic tuning, electro-optic overlays, graphene integration, and microfluidic modulation are identified as natural routes to dynamic control (Sarkar et al., 12 Oct 2025, Abdulaal et al., 30 Sep 2025).

Current extensions are therefore less about demonstrating that BICs can enhance sensing, and more about choosing the appropriate BIC regime for a specific measurement. Reported directions include solvent-aware mid-IR platform selection, multi-parameter wavelength/linewidth/amplitude readout, angle-based vortex tracking, restored-BIC operation via environmental asymmetry, hybrid plasmonic–dielectric or lossy-overlay designs, functionalized biosurfaces for specificity, and reconfigurable THz platforms for integrated sensing and communications (Jiang et al., 23 Jun 2025, Nazarov et al., 21 Jul 2025, Yang et al., 29 Aug 2025, Sharma et al., 4 Sep 2025, Watanabe et al., 12 Mar 2026, Abdulaal et al., 30 Sep 2025). This suggests that BIC-based sensing is best understood not as a single sensor architecture, but as a family of radiation-engineered measurement strategies in which symmetry, interference, and controlled leakage are used to tailor which observable carries the information and which loss channel sets the ultimate performance.