Resonant Structures: Theory & Applications

• Resonant structures are physical systems defined by their ability to support frequency-selective states that govern wave propagation, tunneling, and radiation.
• They are applied in photonics, quantum transport, MEMS, and acoustics to achieve band gap formation, enhanced tunneling, and precision filtering.
• Analytical, numerical, and inverse-design methods reveal how modal interactions and engineered poles and zeros drive performance while addressing common misconceptions.

Resonant structures are physical systems whose geometry, material response, or both support sharply frequency-selective states or modes, so that wave propagation, scattering, tunneling, vibration, or radiation is dominated by poles, quasi-bound states, guided modes, leaky modes, or collective oscillations. In the cited literature, the term encompasses periodic slabs near guided-mode conditions, photonic crystals and quasicrystals under resonant Bragg tuning, double-barrier tunneling heterostructures, open quantum wells with Siegert boundary conditions, highly resonant flexible structures, resonant MEMS, magnetoacoustic hybrids, locally resonant metamaterials, and nonlinear optical cavities for SPDC (Neale et al., 2019, Tanimu et al., 2018, Petersen, 2013).

1. Modal foundations and spectral characterization

A recurring definition of resonance in open systems is based on eigenproblems with purely outgoing boundary conditions. In planar photonic-crystal slabs, resonant states or quasi-normal modes are solutions of Maxwell’s equations with outgoing radiation boundary conditions and complex eigenfrequencies

$$\omega_n=\Omega_n-i\Gamma_n,\qquad Q_n=\frac{\Omega_n}{2\Gamma_n},$$

so that leakage to infinity is built into the eigenvalue itself (Neale et al., 2019). In biperiodic photonic structures, the outgoing condition is enforced by Rayleigh–Bloch expansions in the homogeneous half-spaces above and below the patterned region; resonant modes are then Bloch modes with real in-plane Bloch vector and complex $\omega$, while guided modes have real $\omega$, and bound states in the continuum satisfy $\Im(\omega)=0$ even though they lie in the radiation continuum (Zhang et al., 2024).

In one-dimensional quantum mechanics, resonant states are likewise defined by Siegert boundary conditions,

$\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$

for the Schrödinger equation

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$

yielding a discrete set of complex wave numbers $k_n=p_n+i\varkappa_n$ and energies $E_n=k_n^2$. The literature distinguishes bound states, anti-bound states, and normal resonant (Gamow) states, all of which belong to the full resonant-state spectrum and all of which can contribute to observables such as transmission (Tanimu et al., 2018). A closely related formulation underlies the resonant-state expansion for one-dimensional quantum systems, where bound, antibound, and normal resonant states constitute the basis for a rigorous perturbative treatment of open structures (Tanimu et al., 2019).

For periodic dielectric slabs, resonant scattering is organized by nonrobust guided modes embedded in the continuum. When a symmetric lossless slab supports such a guided mode at $(\kappa_0,\omega_0)$, the transmission and reflection anomalies are governed by analytic functions $a(\kappa,\omega)$, $\omega$0, and $\omega$1 such that

$\omega$2

with $\omega$3 for real $\omega$4. Under the sufficient conditions established in the analysis, the loci of total transmission and total reflection are real-analytic curves in $\omega$5-space intersecting quadratically at the guided-mode point (Shipman et al., 2011).

A general implication is that resonant structures are most naturally understood through their analytic continuation: poles determine linewidths and lifetimes, zeros determine suppression or cancellation channels, and complete modal descriptions require not only discrete guided or leaky modes but also, in photonic crystal slabs, cut contributions associated with Rayleigh–Wood anomalies (Neale et al., 2019).

2. Optical and photonic resonant structures

One class of resonant structures is formed by one-dimensional multiple-quantum-well stacks and deterministic aperiodic sequences built from highly doped quantum wells. In these systems, bound excitons are destroyed by strong doping, the relevant material resonance is the Mahan-type edge at

$\omega$6

and the single-well optical response is described by a reflection coefficient with a power-law singularity parameterized by $\omega$7 and $\omega$8. For a periodic structure with period $\omega$9, the resonant Bragg condition is

$\omega$0

and for $\omega$1 it yields the widest band gap. Near this condition, finite stacks exhibit a superradiant regime with $\omega$2-enhanced reflection and linewidth growth with $\omega$3, whereas large $\omega$4 yields a genuine photonic-crystal regime with a band gap whose lower edge is

$\omega$5

Because the resonance is a Mahan edge rather than a discrete exciton, both reflection and absorption are strongly asymmetric with respect to $\omega$6, and Fibonacci quasicrystals produce self-similar multi-peak spectra rather than a single broad reflection band (Voronov et al., 2010).

Another canonical photonic realization is the distributed Bragg reflector and the planar microcavity. For an ideal quarter-wave DBR with index ratio $\omega$7 and $\omega$8 periods, the peak reflectivity at the design wavelength $\omega$9 is

$\Im(\omega)=0$0

and the stop-band width is

$\Im(\omega)=0$1

For disordered unit cells with transient layers, the effective refractive-index approximation introduces an ideality factor $\Im(\omega)=0$2 and disorder factor $\Im(\omega)=0$3, with

$\Im(\omega)=0$4

and maps the imperfect DBR to an ideal DBR with reduced effective index ratio

$\Im(\omega)=0$5

This exposes a quantitative equivalence between increasing structural disorder and decreasing refractive-index contrast, and the same mapping carries over to microcavity $\Im(\omega)=0$6-factor degradation through the increased leakage of the disordered mirrors (Gačević et al., 2024).

Composite optical filters built from several identical resonant diffractive structures provide a different use of resonance: rather than isolating a single narrow Lorentzian line, they synthesize higher-order flat-top responses. If a single resonant diffractive structure has Lorentzian transmittance, then two identical structures separated by a properly chosen phase-shift layer implement a second-order Butterworth filter, while four such structures implement a third-order Butterworth filter. In the analytical construction, the spacer thickness sets the phase $\Im(\omega)=0$7, and with the condition

$\Im(\omega)=0$8

the two-resonator composite yields

$\Im(\omega)=0$9

where $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$0. A W-structure, consisting of a high-index core layer and two low-index cladding layers in a high-index dielectric environment, serves as the numerical building block, and the linewidth can be reduced to a significantly subnanometer size simply by changing the thicknesses of the cladding layers (Doskolovich et al., 2019).

Quantum optical resonant structures enter again in SPDC. In a low-gain etalon, the paper proposes a simplified model in which pair generation and resonant enhancement are treated as separate processes: $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$1 Here $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$2 is the thin-film SPDC probability, while $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$3 is a resonant filtering factor built from intracavity pump enhancement and out-coupling amplitudes. For the etalon, the forward and backward intracavity pump amplitudes are

$\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$4

and the model agrees with both the rigorous theory and experiment in the low-gain regime (Sorensen et al., 9 Jan 2025).

3. Quantum and electronic resonant structures

In nanoelectronic transport, a resonant structure often means a double-barrier resonant tunnelling device. For an InAs/GaAs/InAs resonant tunnelling structure with transport along $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$5, confinement in the well produces quasi-bound levels with finite linewidth $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$6, and a perpendicular magnetic field quantizes the in-plane motion into Landau levels. The nonequilibrium Green’s-function formulation includes coherent tunnelling, contact self-energies, LO phonon scattering, interface roughness, Zeeman splitting, and Rashba spin–orbit interaction. The Hamiltonian is

$\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$7

with

$\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$8

The model predicts quasi-bound Landau levels, phonon replicas, magnetopolaron anticrossings, and current spin polarization larger than ten percent for magnetic fields above 2 Tesla, while also showing that the Rashba effect is observable as a beating pattern in the density of states but is too small to affect the tunnelling current appreciably (Isić et al., 2011).

Double graphene-layer structures provide a different resonant electronic platform. Two graphene layers separated by a thin tunnel-transparent barrier exhibit a resonant-tunnelling current peak at the alignment voltage $\psi_n(x)\propto e^{ik_n|x|},\qquad |x|\to\infty,$9, where the Dirac points align. The local RT current density is modeled as

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$0

with differential conductivity

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$1

The same structure supports self-consistent electron–hole plasma oscillations along the layers, with characteristic plasma frequency

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$2

and dispersion relation

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$3

The resulting admittance exhibits resonant features, yet the realistic parameter range gives $$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$4, so the plasma remains stable even in the presence of negative differential conductivity. This directly supports the paper’s conclusion that negative differential conductivity does not automatically imply self-excitation of plasma oscillations (Ryzhii et al., 2013).

Open quantum wells furnish the canonical one-dimensional quantum resonant structure. For a symmetric double delta potential,

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$5

the resonant-state wave numbers satisfy

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$6

and the spectrum contains bound, anti-bound, and normal resonant states. In the triple-delta system,

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$7

continuous parameter changes induce transitions between these classes; for example, in the symmetric triple structure with a central well, the second even bound state exists only if

$$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$8

where $$\left[-\frac{d^2}{dx^2}+V(x)\right]\psi_n(x)=k_n^2\psi_n(x),$$9 and $k_n=p_n+i\varkappa_n$0. The transmission amplitude is meromorphic, its poles coincide with the resonant states, and its Mittag–Leffler expansion makes explicit how normal resonant states produce sharp transmission peaks while bound states mainly contribute a background (Tanimu et al., 2018).

4. Mechanical, acoustic, and flexible resonant structures

Highly resonant flexible structures are mechanical systems dominated by lightly damped vibration modes, such as flexible beams and plates, large space structures, nanopositioning stages, and flexible manipulators. With collocated force actuators and position sensors, their transfer matrix has the modal form

$k_n=p_n+i\varkappa_n$1

and such systems are negative imaginary. The frequency-domain NI condition is

$k_n=p_n+i\varkappa_n$2

with the strict version requiring $k_n=p_n+i\varkappa_n$3. The main robust-stability theorem states that for positive feedback of an NI plant and an SNI controller,

$k_n=p_n+i\varkappa_n$4

is necessary and sufficient for internal stability under the stated assumptions. This unifies positive position feedback, resonant controllers, and integral force feedback as instances of strictly negative imaginary control laws, and explains their spillover robustness in highly resonant flexible structures (Petersen, 2013).

At microscale, resonant MEMS structures are shaped not only by their elastic resonance but also by microscale gas damping and gas spring effects. In the gyroscope model,

$k_n=p_n+i\varkappa_n$5

$k_n=p_n+i\varkappa_n$6

while squeeze-film damping yields frequency-dependent damping $k_n=p_n+i\varkappa_n$7 and elastic contribution $k_n=p_n+i\varkappa_n$8. The equivalent mechano-thermal noise force obeys the Nyquist relation

$k_n=p_n+i\varkappa_n$9

so the white-spectrum approximation fails once the damping is frequency-shaped. The paper therefore advocates semi-automatic behavioral macromodel extraction from finite-element data and sensor optimization with respect to signal-to-noise ratio, explicitly using the shaped noise spectrum rather than a constant damping approximation (0805.0927).

Magnetoacoustic resonant structures add a coupled magnon–phonon mechanism. In the epitaxial Fe$E_n=k_n^2$0Si/GaAs device, Rayleigh surface acoustic waves at

$E_n=k_n^2$1

drive a 74 nm Fe$E_n=k_n^2$2Si film with cubic anisotropy $E_n=k_n^2$3. The magnetoelastic energy density is

$E_n=k_n^2$4

and the dynamics obey the Landau–Lifshitz–Gilbert equation. The system exhibits both sharp resonant magnetoelastic coupling near $E_n=k_n^2$5 and broad off-resonant magnetoacoustic response peaking near $E_n=k_n^2$6. The shear magnetoelastic constant extracted from micromagnetic fits,

$E_n=k_n^2$7

is much larger than the one found in Nickel, and the experiments show that magnetoacoustic waves persist over a broad field range rather than existing only at resonance (Rovirola et al., 2022).

Locally resonant metamaterials recast resonance in terms of dynamic mass. For a resonator attached to a host beam, the effective dynamic mass is modeled as

$E_n=k_n^2$8

so each mode with significant modal effective mass contributes a local-resonant band gap. In the multimodal spiral resonators considered in the inverse-design study, the two targeted modes are the first two modes with the largest vertical modal effective mass, and periodic attachment of such resonators to a beam generates two targeted stop bands whose measured dispersion relation agrees closely with the desired one (Dedoncker et al., 2023).

5. Analytical, numerical, and inverse-design methodologies

A striking feature of the literature is that resonant structures are often best understood through reduced representations rather than full brute-force field solutions. In doped quantum-well photonic crystals, the entire finite stack is built from a single-well reflection coefficient via a transfer-matrix description, and the infinite periodic limit is encoded in the polariton dispersion equation

$E_n=k_n^2$9

with $(\kappa_0,\omega_0)$0 and $(\kappa_0,\omega_0)$1 for negligible dielectric mismatch (Voronov et al., 2010). In planar multilayer filters, the scattering-matrix formalism and the Redheffer star product provide exact cascaded descriptions of composite resonators, making pole placement and Butterworth synthesis analytically transparent (Doskolovich et al., 2019).

For open quantum transport, the nonequilibrium Green’s-function formalism provides the natural resonant-structure description because it unifies quasi-bound states, contact coupling, and scattering. In the resonant tunnelling study, the retarded Green’s function is computed from

$(\kappa_0,\omega_0)$2

with self-energies for the emitter, collector, interface roughness, and LO phonons, and the current follows a Meir–Wingreen-type expression in terms of $(\kappa_0,\omega_0)$3 and $(\kappa_0,\omega_0)$4 (Isić et al., 2011).

Resonant-state expansion plays a parallel role in quantum mechanics and photonics. In one-dimensional quantum systems, the perturbed resonant states are expanded as

$(\kappa_0,\omega_0)$5

leading to the matrix eigenvalue problem

$(\kappa_0,\omega_0)$6

This gives a rigorous perturbation theory for open quantum wells and finite periodic systems, with convergence improving as the resonant-state basis grows (Tanimu et al., 2019). In photonic-crystal slabs, the corresponding PC-RSE expands the Green’s function and the perturbed fields over the resonant states of a homogeneous slab, including discretized cut modes, and reduces the problem to a linear eigenvalue equation in $(\kappa_0,\omega_0)$7,

$(\kappa_0,\omega_0)$8

thereby providing complete modal spectra, including guided modes, leaky modes, and the ingredients needed to identify BIC formation (Neale et al., 2019).

For doubly periodic photonic structures, a complementary route avoids perfectly matched layers entirely. The PML-free method defines transverse impedance operators for the homogeneous exterior and the structured interior, truncates the tangential fields to $(\kappa_0,\omega_0)$9 Fourier unknowns, and reduces the open Maxwell eigenproblem to a small nonlinear matrix equation

$a(\kappa,\omega)$0

which is then solved by contour integrals. The method is reported to be efficient with respect to memory usage and CPU time, free of spurious solutions, and able to determine degenerate resonant modes without difficulty (Zhang et al., 2024).

Inverse design enters when the resonant structure itself is the unknown. In the multimodal metamaterial work, a conditional variational autoencoder learns nontrivial patterns between six geometric variables of a spiral resonator and four modal attributes $a(\kappa,\omega)$1. After training on a dataset generated from a few hundred to a few thousand numerical modal analyses, designs satisfying arbitrary requested modal frequencies and masses can be generated at negligible marginal cost, and the best generated designs are confirmed by 3D-printed prototypes and measured metamaterial-beam dispersion curves (Dedoncker et al., 2023).

6. Recurring principles, misconceptions, and limits

A recurring principle is the interplay between structural and material resonance. In doped multiple-quantum-well photonic crystals, the resonant Bragg condition explicitly locks a structural phase condition to a material edge singularity (Voronov et al., 2010). In etalon-based SPDC, the pair spectrum factors into a bare generation term and a resonant filtering term only because the structure acts linearly on the signal, idler, and pump fields in the low-gain regime (Sorensen et al., 9 Jan 2025). In locally resonant metamaterials, geometric shape determines modal effective masses and frequencies, which then control the host dispersion through the dynamic mass (Dedoncker et al., 2023). This suggests a common viewpoint: resonance engineering is often the engineering of how a compact set of poles and zeros is embedded into a larger continuum.

Several misconceptions are corrected explicitly in the literature. One is that resonance requires a discrete bound state such as an exciton; the doped-quantum-well work shows instead that a continuum-edge Mahan singularity can still support superradiant enhancement, broad reflection bands, and photonic gap formation under the resonant Bragg condition (Voronov et al., 2010). Another is that negative differential conductivity necessarily destabilizes resonant plasma systems; the double-graphene-layer analysis shows that realistic structures satisfy $a(\kappa,\omega)$2, so the plasma remains stable despite inter-layer resonant-tunnelling negative conductivity (Ryzhii et al., 2013). A third is that magnetoacoustic waves occur only at the exact resonance field; direct XPEEM imaging in Fe$a(\kappa,\omega)$3Si/GaAs shows substantial off-resonant coherent precession over a broad field range, with resonance providing a sharp enhancement rather than the existence condition itself (Rovirola et al., 2022). A fourth is that mechano-thermal noise in resonant MEMS can be treated as white; when gas damping is frequency dependent, the fluctuation–dissipation relation implies a shaped noise spectrum that materially changes SNR optimization (0805.0927).

The limits of resonant-structure models are equally domain-specific. The simplified SPDC model is supported only in the low-gain regime and breaks down when internal amplification becomes important (Sorensen et al., 9 Jan 2025). Negative-imaginary control theory relies on collocated force actuators and position sensors and on linear time-invariant assumptions (Petersen, 2013). The effective refractive-index approximation for DBRs presumes approximately constant disorder from cell to cell and small per-cell reflection, so that unit-cell optical response can be summarized by a single disorder factor (Gačević et al., 2024). The PML-free contour-integral method is formulated for biperiodic structures bounded in $a(\kappa,\omega)$4 and sandwiched between homogeneous half-spaces (Zhang et al., 2024). These are not defects of the concept of resonance; they are reminders that resonant structures are analytically powerful precisely because they reduce broad wave phenomena to a small number of dominant modes, and that the validity of such reduction must be checked against leakage, coupling, disorder, gain, and nonlinearity in each concrete system.