Atom-Photon Bound States

• Atom-photon bound states are hybrid light-matter excitations where an emitter shares an excitation with a spatially localized photonic cloud, typically appearing outside the propagating continuum.
• They are formed in structured photonic environments via dispersion engineering, nonlocal coupling, and interference, offering tunable localization and coherent dynamics.
• Detection and control of these states are achieved through scattering signatures and time-domain dynamics, paving the way for quantum technology applications.

Atom-photon bound states are hybrid light-matter states in which an excitation is shared between an emitter and a photonic environment, while the photonic component remains spatially localized rather than forming an extended scattering wave. In finite-band and band-gap waveguide QED, they typically appear when the dressed-state energy lies outside the propagating continuum, producing an evanescent photonic cloud around the emitter; in open systems they can instead occur as complex-energy localized modes; and in special interference settings they can remain embedded in the continuum as bound states in the continuum (BICs) (Calajo et al., 2015, Wang et al., 2018, Fong et al., 2017). More recent work extends the concept to topological reservoirs, giant-atom geometries, dynamically prepared band-edge states, and many-photon composites generated by emitter-induced nonlinearity (Zhao et al., 2024, Castillo-Moreno et al., 2024, Mahmoodian et al., 2019).

1. Definitions and spectral classifications

A standard single-excitation form is a dressed state

$$|\Psi\rangle = c_E \,\sigma_+|g,0\rangle + \sum_i c_i\, a_i^\dagger |g,0\rangle,$$

with an emitter-like component and a localized photonic cloud. In structured reservoirs, the localization arises because propagating modes are absent in a gap or strongly modified near a band edge, so the emitter cannot radiate away efficiently (Castillo-Moreno et al., 2024). In slow-light waveguide QED, the same object is described as the continuum analogue of a cavity-QED dressed state: the atom is dressed not by a single cavity mode but by a localized wavepacket formed from a finite-band continuum (Calajo et al., 2015).

The literature also uses a dissipative definition. For propagating photons interacting with $N$ two-level atoms in an open chiral channel, bound states are single-excitation eigenstates of an effective non-Hermitian spin model,

$$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$

for which the photonic envelope decays exponentially away from the atoms with localization length $\sim 1/|\operatorname{Im}E_\alpha|$ (Wang et al., 2018). Here the overall wavefunction keeps its shape while decaying through additional reservoir channels.

A further extension is the BIC, a localized eigenstate whose energy lies inside a continuum but which does not couple to propagating modes because of destructive interference. In a coupled-cavity array with two spatially separated atomic ensembles, the BIC traps excitation inside the internal atomic and cavity degrees of freedom while the external fields remain in vacuum (Fong et al., 2017). By contrast, a “virtual” bound state is not a stationary eigenstate at all: it is a finite-time suppression of spontaneous decay engineered through an initially entangled atom-photon state (Longhi, 30 Apr 2025).

Regime	Defining property	Representative sources
Finite-band / band-gap bound state	Energy outside the propagating band; evanescent photon cloud	(Calajo et al., 2015, Castillo-Moreno et al., 2024)
Dissipative bound state	Complex eigenvalue with $\operatorname{Im}E_\alpha<0$; localized but decaying profile	(Wang et al., 2018)
Bound state in the continuum	Localized state inside a continuum due to destructive interference	(Fong et al., 2017)
Virtual bound state	Finite-time population freezing from initial atom-photon entanglement	(Longhi, 30 Apr 2025)

Taken together, these works indicate that “atom-photon bound state” is a family resemblance term rather than a single universal definition.

2. Formation in finite-band and band-edge photonic media

The canonical mechanism is dispersion engineering. For a one-dimensional tight-binding photonic channel with

$\omega_k=\omega_c-2J\cos k,$

the propagating band is finite, $[\omega_c-2J,\omega_c+2J]$. A localized emitter coupled to this bath can support discrete eigenvalues outside the band, where the photon component is evanescent rather than propagating (Calajo et al., 2015, Lu et al., 2024). For a single atom in slow-light waveguide QED, the bound-state energies $E_\pm$ satisfy

$$E_{\pm}-\delta=\frac{g^2}{E_{\pm}\sqrt{1-\frac{4J^2}{E_\pm^2}}},$$

and the localization length obeys

$$\frac{1}{\lambda}=\operatorname{arccosh}\left(\frac{|E|}{2J}\right),$$

so detuning from the band edge directly controls the size of the photonic cloud (Calajo et al., 2015).

This band-edge picture persists in superconducting implementations. In a microwave metamaterial formed by a chain of 21 nearest-neighbor coupled lumped-element resonators, a flux-tunable transmon coupled to site 13 develops a bound state pinned below the lower band edge as the bare qubit frequency is tuned upward. The measured band spans about $N$0–$N$1 GHz, and the static signature of localization is that the dressed qubit frequency deviates from the bare transmon dispersion and becomes pinned below the band (Castillo-Moreno et al., 2024). A related architecture based on 21 compact high-impedance superconducting resonators realizes a 1 GHz-wide pass band and accesses strong multimode coupling with $N$2 MHz, enabling direct spectroscopy, lifetime measurements, and controlled interactions between bound states (Scigliuzzo et al., 2021).

The same finite-band logic extends to emitters coupled nonlocally to two adjacent resonators. In a one-dimensional coupled-resonator array with a two-level emitter coupled to sites $N$3 and $N$4, the system supports out-of-band discrete levels $N$5 whose number depends on $N$6, $N$7, and $N$8, rather than being fixed at two as in the local-coupling case (Lu et al., 2024). The bound-state photon distribution is exponentially localized but becomes spatially asymmetric when $N$9, with both bound states preferring the same direction determined primarily by the larger coupling. This asymmetry is a direct consequence of nonlocal coupling geometry rather than chirality (Lu et al., 2024).

3. Scattering, detection, and dynamical signatures

Bound states leave clear signatures in scattering observables. In dissipative chiral waveguide QED, the single-photon transmission amplitude can be written as

$$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$0

and the appearance of a bound state is in one-to-one correspondence with a zero of transmission. The same zeros generate divergent bunching in the output two-photon correlation,

$$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$1

and the number of bound states is encoded in the winding of the transmission phase through a dissipative Levinson theorem (Wang et al., 2018). This makes bound-state counting experimentally accessible through phase-sensitive transmission measurements.

Time-domain dynamics provide complementary diagnostics. In the two-adjacent-resonator model, spontaneous emission from an initially excited emitter decays exponentially to zero in the weak-coupling regime, but atom-photon bound states produce either a non-zero constant or a stationary oscillation at long times. If only one bound state contributes, the excited-state population approaches a constant; if two bound states contribute, their interference yields persistent oscillation (Lu et al., 2024). In the strong-coupling limit, the oscillation frequency is set by $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$2 (Lu et al., 2024).

At a band edge, bound-state preparation can itself be dynamical. In the transmon–metamaterial experiment, a $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$3-pulse first excites the transmon, after which a trapezoidal flux pulse sweeps the emitter from an emitter-like point $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$4 to a final point $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$5 near or inside the band edge. For rise and fall times $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$6 ns, the dynamics are close to adiabatic and the system follows the instantaneous lowest eigenstate into the APBS; for $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$7 ns, the evolution is clearly non-adiabatic and exhibits multilevel Landau–Zener-type transitions, reduced recovered emitter population, and beating at differences between dressed-mode frequencies (Castillo-Moreno et al., 2024). Rapid quenching back from the APBS-supporting regime causes the localized photonic component to “melt” into propagating excitations, with nine prominent peaks observed in the output spectrum, one for each of the nine lowest-frequency metamaterial modes that significantly hybridize with the emitter (Castillo-Moreno et al., 2024).

These diagnostics suggest a useful distinction: spectroscopy identifies the discrete spectrum, while time-domain measurements reveal whether the localized mode acts as a trapped component, a slowly decaying dressed state, or a reservoir for coherent population exchange.

4. Topology, giant atoms, and interference-based localization

A major recent direction combines structured dispersion with nonlocal coupling and topology. In a coupled-resonator waveguide coupled at two distant sites to a finite Su-Schrieffer-Heeger chain—a “topological giant atom” (TGA)—the waveguide has dispersion

$$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$8

with continuum $$M|e_\alpha\rangle = E_\alpha |e_\alpha\rangle, \qquad \operatorname{Im} E_\alpha < 0,$$9, while the SSH chain supplies topological structure through $\sim 1/|\operatorname{Im}E_\alpha|$0, $\sim 1/|\operatorname{Im}E_\alpha|$1, and a boundary hopping $\sim 1/|\operatorname{Im}E_\alpha|$2 (Zhao et al., 2024). Because the TGA is coupled nonlocally at $\sim 1/|\operatorname{Im}E_\alpha|$3 and $\sim 1/|\operatorname{Im}E_\alpha|$4, photons entering the covered region acquire multiple emission and reabsorption pathways. For $\sim 1/|\operatorname{Im}E_\alpha|$5, the topologically trivial phase $\sim 1/|\operatorname{Im}E_\alpha|$6 yields mainly constructive interference near resonance and complete transmission, whereas the topologically nontrivial phase $\sim 1/|\operatorname{Im}E_\alpha|$7 yields destructive interference and complete reflection (Zhao et al., 2024). Off resonance, the reflection develops a Fano lineshape, and larger $\sim 1/|\operatorname{Im}E_\alpha|$8 can generate multiple full-reflection frequencies (Zhao et al., 2024).

The same system supports two pairs of atom-photon bound states lying above and below the continuum in the topologically nontrivial phase. Under periodic boundary condition $\sim 1/|\operatorname{Im}E_\alpha|$9, the two bound-state pairs are nondegenerate, with

$\operatorname{Im}E_\alpha<0$0

on the order of $\operatorname{Im}E_\alpha<0$1 MHz for $\operatorname{Im}E_\alpha<0$2 MHz and $\operatorname{Im}E_\alpha<0$3. Under open boundary condition $\operatorname{Im}E_\alpha<0$4, the states become degenerate, with

$\operatorname{Im}E_\alpha<0$5

so the boundary condition controls whether the doublets are split or collapsed (Zhao et al., 2024). A probe atom placed near the left coupling site decays exponentially for the split PBC spectrum but undergoes Rabi oscillations for the degenerate OBC spectrum, providing a coherent boundary-condition diagnostic (Zhao et al., 2024).

Topology can also seed weak-coupling bound states. In a semi-infinite SSH reservoir, only the finite-end edge state survives the semi-infinite extension, and an emitter coupled to that boundary can hybridize with the topological zero mode. For $\operatorname{Im}E_\alpha<0$6 and

$\operatorname{Im}E_\alpha<0$7

the system supports weak-coupling mid-gap bound states with $\operatorname{Im}E_\alpha<0$8, partial sublattice localization, and reversible oscillations between emitter-like and edge-state-like configurations (Garmon et al., 10 Mar 2025). This differs qualitatively from standard strong-coupling bound states outside the continuum: localization is supplied by the pre-existing topological edge mode rather than by repeated emission and reabsorption alone.

Giant-atom nonlocality also produces purely interference-based bound states. A single artificial atom coupled at multiple points to a one-dimensional continuum obeys a delay-differential equation in which delayed terms encode feedback from radiation emitted at one coupling point and reencountered at another. In the non-Markovian regime, a boson can be trapped between the outermost coupling points; for more than two coupling points, multiple dark modes can coexist and generate persistent oscillating bound states through coherent beating (Guo et al., 2019). With multiple giant atoms coupled to a coupled-resonator waveguide, these dressed states can hybridize into effective interacting chains and SSH-like metabands in the photonic gap (Jia et al., 2023).

5. From single excitations to correlated many-photon bound states

The emitter-induced nonlinearity that generates single-excitation localization also supports correlated few- and many-photon bound states. In chiral waveguide QED, exact Bethe-ansatz scattering eigenstates show that the two-photon bound state has wavefunction

$\operatorname{Im}E_\alpha<0$9

so the state is delocalized in center-of-mass coordinate but exponentially localized in relative coordinate (Mahmoodian et al., 2019). For general $\omega_k=\omega_c-2J\cos k,$0, the fully bound $\omega_k=\omega_c-2J\cos k,$1-string state remains clustered, and its transmission coefficient is

$\omega_k=\omega_c-2J\cos k,$2

which yields a Wigner delay

$\omega_k=\omega_c-2J\cos k,$3

The $\omega_k=\omega_c-2J\cos k,$4 scaling implies that higher-photon-number bound states propagate with smaller delay and reduced distortion (Mahmoodian et al., 2019).

This hierarchy has been observed directly in a one-sided optical cavity containing a single semiconductor quantum dot. Scattering a weak coherent pulse from the cavity-QED system, the measured delays are

$\omega_k=\omega_c-2J\cos k,$5

for the one-, two-, and three-photon components, respectively (Tomm et al., 2022). The reduction $\omega_k=\omega_c-2J\cos k,$6 is interpreted as a direct dynamical fingerprint of stimulated emission: the arrival of two photons within the emitter lifetime shortens the interaction time relative to a single photon (Tomm et al., 2022).

In a broader many-body perspective, matrix-product-state simulations show that the large-photon-number limit of these bound-state dynamics coincides with self-induced transparency or immediate generalizations thereof across chiral waveguides, bidirectional waveguides, and Rydberg media (Calajo et al., 2021). This suggests that classical solitonic pulses can emerge as coherent superpositions of many-photon bound states rather than as a separate phenomenon.

Rydberg-EIT media provide a distinct two-body mechanism. After adiabatic elimination, the relative coordinate of two polaritons obeys an effective Schrödinger equation with

$\omega_k=\omega_c-2J\cos k,$7

whose singularity near the blockade radius behaves like a Coulomb potential. The resulting spectrum contains multiple hydrogen-like branches of metastable Coulomb bound states, and under some conditions the wavefunction becomes double-peaked near $\omega_k=\omega_c-2J\cos k,$8, producing a diatomic-molecule-like photonic state (Maghrebi et al., 2015).

6. Mathematical generalizations and technological uses

Recent work broadens the geometry and mathematical formulation of atom-photon bound states. In self-similar fractal photonic lattices, the bound-state Green’s function can be written as a heat-kernel Laplace transform,

$\omega_k=\omega_c-2J\cos k,$9

and the far-field localization length scales as

$[\omega_c-2J,\omega_c+2J]$0

where $[\omega_c-2J,\omega_c+2J]$1 is the walk dimension of the underlying fractal (Bönsel et al., 22 May 2026). This scaling does not rely on translational invariance or a band-edge effective-mass approximation. In the near field, the bound-state amplitude exhibits an additional algebraic variation; for nested finitely ramified fractals the exponent agrees with resistance and first-passage scaling, whereas Sierpiński carpets deviate from that simple law (Bönsel et al., 22 May 2026).

A complementary mathematical formulation reduces single-photon quantum optics with two-level atoms to a nonlinear eigenproblem for nonlocal PDEs. In that setting, bound states correspond to negative eigenvalues of a self-adjoint operator, resonances to outgoing solutions with positive real part and nonpositive imaginary part, and a Birman–Schwinger principle gives necessary and sufficient conditions for bound-state existence together with an upper bound on their number (Hiltunen et al., 2023). This approach makes explicit that the existence question can be posed without committing to a lattice model.

On the control side, atom-photon bound states already function as engineered resources. In high-impedance resonator arrays, coherent interactions between two atom-photon bound states have been demonstrated in resonant and dispersive regimes suitable for SWAP and CZ two-qubit gates, with interaction strengths tunable by the overlap of evanescent photonic clouds (Scigliuzzo et al., 2021). In topological giant-atom settings, topology and interference toggle between cloaking-like complete transmission and mirror-like complete reflection and supply a route to coherent switches, transistors, and boundary-condition sensing (Zhao et al., 2024). Even when the Hamiltonian has no true stationary bound state, an initially entangled atom-photon state can produce a transient “virtual” bound state for which

$[\omega_c-2J,\omega_c+2J]$2

or can slow spontaneous emission to an engineered rate $[\omega_c-2J,\omega_c+2J]$3 (Longhi, 30 Apr 2025).

A plausible implication of these developments is that atom-photon bound states now serve less as a single specialized effect than as a unifying language for localization, coherent transport control, topological hybridization, many-photon correlation, and programmable interactions in waveguide and cavity QED.