Two-Giant-Atom Waveguide-QED Model

Updated 5 July 2026

The two-giant-atom waveguide-QED model is a quantum system where two nonlocally coupled two-level atoms interact with a 1D waveguide via multiple connection points.
Different geometries—separate, braided, and nested—dictate the interference patterns that shape individual decay, collective decay, Lamb shifts, and coherent exchange.
This architecture enables decoherence-free interaction and robust entanglement generation, with applications spanning both Markovian and non-Markovian dynamics.

The two-giant-atom waveguide-QED model denotes a class of waveguide quantum electrodynamics systems in which two two-level atoms, usually artificial superconducting atoms, couple nonlocally to a common one-dimensional waveguide through multiple spatially separated connection points. Its defining feature is that geometry enters the reduced dynamics through propagation phases between all pairs of coupling points, so interference reshapes individual decay, collective decay, Lamb shifts, and waveguide-mediated exchange. In the canonical two-point-per-atom setting, the model admits three inequivalent orderings of the four connection points—separate, braided, and nested—and, in the braided case, can realize finite coherent exchange with vanishing waveguide-induced decay, a possibility absent for ordinary pointlike emitters (Kockum et al., 2017).

1. Definition, degrees of freedom, and canonical geometries

A giant atom differs from a small atom by coupling to the waveguide at more than one position. For two giant atoms, each atom is typically taken to be a two-level system with lowering and raising operators $\sigma_-^{(j)}$ and $\sigma_+^{(j)}$ , coupled to the same bidirectional 1D waveguide at two or more points. The geometric data are the ordered connection-point positions $x_{j_n}$ , the local coupling strengths $\gamma_{j_n}$ or $g_{j_n}$ , and the propagation phases accumulated between points. In equally spaced formulations, the nearest-neighbor phase is often written as $\theta_0=k_0 d$ or $\varphi=k(x_{j+1}-x_j)$ , where $d$ is the spacing between adjacent points (Yin et al., 2023).

For exactly two two-point giant atoms, the canonical orderings are:

Separate: $a_1,a_2,b_1,b_2$ .
Braided: $a_1,b_1,a_2,b_2$ .
Nested: $\sigma_+^{(j)}$ 0.

These orderings are not merely pictorial. They determine which path amplitudes interfere in self-emission and which interfere in atom-atom exchange. That distinction is central to the giant-atom effect. In the infinite-waveguide two-point model, separate and nested arrangements do not support decoherence-free exchange under the same conditions that suppress local decay, whereas braided arrangements do (Kockum et al., 2017).

Several later generalizations preserve this core definition while changing the bath or the coupling architecture. Structured-waveguide versions replace the linear continuum by a coupled-resonator array with a finite cosine band, semi-infinite variants add a mirror and reflected paths, and some models add direct coherent coupling between the atoms with a physical phase $\sigma_+^{(j)}$ 1 that cannot be gauged away (Wu et al., 3 Nov 2025).

2. Microscopic formulations and reduced dynamical equations

At the microscopic level, the model is usually built from an atomic Hamiltonian, a 1D bosonic waveguide, and local couplings at each connection point. In the Markovian continuum formulation, the reduced dynamics are more important than the explicit continuum Hamiltonian: after tracing out the waveguide, the geometry survives only through sums of $\sigma_+^{(j)}$ 2 and $\sigma_+^{(j)}$ 3 of propagation phases. In that reduced description, coherent terms are sine-weighted, while dissipative terms are cosine-weighted (Kockum et al., 2017).

A standard Markovian master equation for two giant atoms takes the form

$\sigma_+^{(j)}$ 4

with coherent Hamiltonian

$\sigma_+^{(j)}$ 5

For the two-point-per-atom case, the effective coefficients are sums over all four point-to-point paths,

$\sigma_+^{(j)}$ 6

$\sigma_+^{(j)}$ 7

$\sigma_+^{(j)}$ 8

with $\sigma_+^{(j)}$ 9 in the equal-coupling convention (Yin et al., 2023).

In the single-excitation sector, many treatments pass to the non-Hermitian effective Hamiltonian

$x_{j_n}$ 0

acting on

$x_{j_n}$ 1

This form is the minimal effective two-giant-atom model used in entanglement-generation studies, transfer problems, and some scattering reductions (Liu et al., 3 Jan 2025).

Structured-bath versions replace the continuum by a coupled-resonator array. For two identical giant atoms with two coupling points each, the rotating-frame Hamiltonian is

$x_{j_n}$ 2

with $x_{j_n}$ 3, so the bath spectrum is the finite band $x_{j_n}$ 4. In that setting, the relevant objects become the spectral densities $x_{j_n}$ 5 and $x_{j_n}$ 6, which encode self- and interatomic couplings, respectively (Wu et al., 3 Nov 2025).

3. Geometry-induced interference and decoherence-free interaction

The central infinite-waveguide result is the decoherence-free interaction of two braided giant atoms. For the symmetric two-point case with equal spacing between consecutive coupling points, equal bare decay $x_{j_n}$ 7, and a single segment phase $x_{j_n}$ 8, the braided coefficients are

$x_{j_n}$ 9

$\gamma_{j_n}$ 0

$\gamma_{j_n}$ 1

$\gamma_{j_n}$ 2

$\gamma_{j_n}$ 3

one has

$\gamma_{j_n}$ 4

This is the decoherence-free interaction point: the waveguide induces coherent exchange without local or collective relaxation (Kockum et al., 2017).

Its mechanism is interference at the level of the system-bath coupling operator, not merely at the level of a particular collective eigenstate. Accordingly, the protected object is not a dark subspace of the usual small-atom type. Rather, “the entire multi-atom Hilbert space is protected from decoherence, not just a subspace” under the braided giant-atom phase condition (Kockum et al., 2017).

The topology dependence is equally sharp. In the same infinite-waveguide Markovian setting, if separate or nested two-point giant atoms are tuned so that each atom’s own waveguide emission vanishes, then the waveguide-mediated atom-atom interaction also vanishes. By contrast, for a braided pair, the same condition still allows $\gamma_{j_n}$ 5. This is the basic design rule of the two-giant-atom model: suppress each atom’s net coupling to the outgoing radiation channel while arranging the interatomic path sums so that exchange remains (Kockum et al., 2017).

Boundary conditions modify that rule. In a semi-infinite waveguide terminated by a perfect mirror, reflected paths add terms with phases $\gamma_{j_n}$ 6 on top of the direct-path terms $\gamma_{j_n}$ 7. The corresponding decoherence-free interaction condition remains

$\gamma_{j_n}$ 8

In that semi-infinite model, braided giant atoms satisfy it at

$\gamma_{j_n}$ 9

and nested giant atoms satisfy it at

$g_{j_n}$ 0

The separate geometry does not satisfy the same condition for any $g_{j_n}$ 1, and the corresponding semi-infinite small-atom system also fails to realize decoherence-free interaction (Liu et al., 25 Aug 2025).

4. Structured baths, bound states, and non-Markovian formulations

Once the waveguide is structured rather than linearized, the two-giant-atom model becomes intrinsically spectral. In the coupled-resonator-array formulation, the exact single-excitation amplitudes satisfy the integro-differential equation

$g_{j_n}$ 2

with memory kernel

$g_{j_n}$ 3

For two atoms, the problem diagonalizes into even and odd sectors governed by $g_{j_n}$ 4. The long-time dynamics are then organized by the bound-state content of the full atom-waveguide spectrum: continuum only implies complete decay, one bound state implies finite population or entanglement trapping, and multiple bound states imply persistent lossless Rabi-like oscillations with frequencies set by bound-state energy differences (Wu et al., 3 Nov 2025).

The same paper gives explicit spectral densities for the two-point-per-atom case,

$g_{j_n}$ 5

$g_{j_n}$ 6

where $g_{j_n}$ 7 is the $g_{j_n}$ 8-th Chebyshev polynomial. These formulas make the dependence on intra-atom separation $g_{j_n}$ 9, inter-atom separation $\theta_0=k_0 d$ 0, and band-edge singularities explicit (Wu et al., 3 Nov 2025).

A complementary structured-bath extension adds direct coherent coupling between the two giant atoms,

$\theta_0=k_0 d$ 1

in a braided two-giant-atom coupled-resonator-waveguide model. There the coupling phase $\theta_0=k_0 d$ 2 is physically observable because the direct and indirect couplings form a loop, so $\theta_0=k_0 d$ 3 cannot be removed by a unitary transformation. In that setting the number of bound states in the continuum becomes phase-controlled: one BIC, two BICs, or no BICs can occur, and the dynamics follow the same spectral rule—fractional trapping for one BIC, complete radiative decay for no BIC, and long-lived Rabi-like oscillations for two BICs (Qi et al., 18 May 2026).

Band-gap physics yields another extension. In the two-giant-atom coupled-resonator model of dressed states, the bound-state energies are determined by an energy-dependent $\theta_0=k_0 d$ 4 self-energy matrix, and the strong-coupling limit reduces to an effective dressed-atom dimer,

$\theta_0=k_0 d$ 5

The effective coupling $\theta_0=k_0 d$ 6 depends strongly on whether the geometry is separate, braided, or nested, because each case has a different overlap structure of localized photonic clouds (Jia et al., 2023).

5. Entanglement generation, disentanglement, and transfer

Entanglement studies typically use concurrence. In the single-excitation manifold,

$\theta_0=k_0 d$ 7

the concurrence reduces to

$\theta_0=k_0 d$ 8

Starting from a single-excitation separable state, Markovian two-point models with separated, braided, or nested couplings all support steady-state entanglement due to the appearance of a dark state. Starting from the double-excitation state, the same models exhibit entanglement sudden birth as the phase shift is varied. The paper reporting this comparison states that the maximal entanglement for the nested coupling is about one order of magnitude larger than those of separate and braided couplings (Yin et al., 2023).

When retardation is retained, the same two-giant-atom problem becomes non-Markovian. In the delayed real-space model, the amplitudes satisfy time-delayed equations involving both self-feedback and mutual delayed coupling. The resulting entanglement dynamics is non-exponential and shows revivals. The steady-state entanglement depends on the time delay under certain conditions in separated, braided, and nested configurations, and the nested geometry is identified as the best geometry for achieving large stationary entanglement (Yin et al., 2022).

Two-giant-atom dynamics also appears as a subsystem in entanglement-transfer problems. In the four-atom construction consisting of two independent two-emitter waveguide subsystems, the relevant two-giant-atom coefficients are again $\theta_0=k_0 d$ 9, $\varphi=k(x_{j+1}-x_j)$ 0, and $\varphi=k(x_{j+1}-x_j)$ 1. In the braided configuration at

$\varphi=k(x_{j+1}-x_j)$ 2

the dissipative terms vanish while coherent exchange remains. Under that condition,

$\varphi=k(x_{j+1}-x_j)$ 3

so complete entanglement transfer to $\varphi=k(x_{j+1}-x_j)$ 4 occurs whenever

$\varphi=k(x_{j+1}-x_j)$ 5

For transfer to $\varphi=k(x_{j+1}-x_j)$ 6, the same geometry yields

$\varphi=k(x_{j+1}-x_j)$ 7

so the maximum transferred concurrence is $\varphi=k(x_{j+1}-x_j)$ 8 (Liu et al., 3 Jan 2025).

These results establish a common pattern. Dark-state trapping, decoherence-free exchange, and delayed feedback are not separate topics in the two-giant-atom literature; they are three manifestations of the same phase-controlled interference structure.

6. Scattering, chirality, topology, and boundary-assisted extensions

The two-giant-atom model also serves as a scattering platform. In the “giant artificial molecule” formulation, two directly coupled giant atoms each connect to a bidirectional waveguide at two points, and exact real-space single-photon transmission and reflection amplitudes can be derived for separated, braided, and nested configurations. In the Markovian regime, the spectra display geometry-dependent Fano line shapes, Lorentzian peaks, tunable transmission windows, and vacuum-Rabi-splitting-like doublets. In the non-Markovian regime, finite propagation time revives multiple reflection peaks and dips that disappear under the Markov approximation (Yin et al., 2022).

Topology enriches the same model class. In the separate-coupling SSH-waveguide version, each atom still has two coupling points, but each point can couple to sublattice $\varphi=k(x_{j+1}-x_j)$ 9 or $d$ 0, yielding 16 coupling configurations. The entanglement dynamics of 14 of these 16 configurations depends on the dimerization parameter of the SSH waveguide, 10 can be grouped into five paired classes by their self-energies, and the delayed sudden birth of entanglement is strongly enhanced in those five pairs (Luo et al., 2023).

Chirality and extra connection points yield another extension. In the bidirectional-chiral model with three connection points per atom, five geometries are distinguished: separate, fully braided, partially braided, fully nested, and partially nested. In the nonchiral case, the concurrence of each configuration exhibits steady-state properties attributable to the presence of a dark state. In the chiral case, the entanglement is maximally enhanced compared to the nonchiral case, and in the fully braided configuration the concurrence reaches 1 and is robust to chirality (Liu et al., 2024).

Finally, mirror boundaries feed back into the same classification. In the semi-infinite model, braided and nested two-giant-atom systems both generate maximally entangled states because decoherence-free interaction survives in both topologies, while separate giant atoms do not support the same protected interaction even though their maximal concurrence can exceed $d$ 1 (Liu et al., 25 Aug 2025).

Taken together, these variants show that the “two-giant-atom waveguide-QED model” is not a single equation but a modeling family. Its common core is two nonlocally coupled emitters and geometry-encoded interference; its main branches differ by whether the waveguide is linear or structured, infinite or semi-infinite, nonchiral or chiral, and Markovian or explicitly delayed. Across those branches, the central technical fact remains the same: coherent and dissipative couplings are separately programmable because emission amplitudes from different connection points interfere.