Hybrid-Frequency Synthetic-Dimension Simulator

• Hybrid-Frequency Synthetic-Dimension Simulator Architecture is a method that maps discrete resonator frequency modes into a controllable synthetic lattice with engineered hopping and boundaries.
• The approach integrates hardware modules such as multimode resonators, dynamic modulators, and auxiliary elements to implement complex gauge potentials, nonreciprocal transport, and topological effects.
• This architecture supports quantum simulations and operator synthesis, paving the way for studies in photon interactions, quantum state readout, and programmable linear transformations.

Across recent work on synthetic dimensions, a coherent architectural pattern has emerged in which discrete frequency modes are treated as lattice sites and then combined with additional control or state-space resources—dynamic modulation, auxiliary resonators, real-space arrays, multiband legs, nonlinear media, or qubit interfaces—to realize programmable synthetic lattices. This pattern is aptly described as a hybrid-frequency synthetic-dimension simulator architecture. In its canonical form, a resonator supports nearly equally spaced modes

$\omega_m=\omega_0+m\Omega_R,$

and modulation at a frequency commensurate with the free spectral range turns the mode index $m$ into a synthetic coordinate with controllable hopping, while supplementary hardware defines boundaries, gauge structure, interactions, dimensional extension, and readout (Dutt et al., 2019, Dutt et al., 2022, Javid et al., 2022, Xie et al., 14 Feb 2026).

1. Frequency-space mapping and the architectural idea

The foundational mapping is the reinterpretation of a resonator’s discrete frequency ladder as a lattice. In dynamically modulated ring resonators, modulation at $\Omega_M=\Omega_R$ couples modes $m\leftrightarrow m\pm1$, yielding the standard nearest-neighbor synthetic-frequency Hamiltonian

$H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$

with the usual assumptions of negligible dispersion over the used bandwidth, resonant modulation, and the rotating-wave approximation. In the simplest translationally invariant case, the band is cosine-like,

$\varepsilon(k)=2J\cos k,$

and the reciprocal coordinate of the synthetic frequency lattice can be read out directly in time, via the relation $k\leftrightarrow t$, in time-resolved transmission spectroscopy (Dutt et al., 2019).

What makes the architecture hybrid is that the frequency lattice is rarely used in isolation. In one major branch, a synthetic frequency axis is combined with a real-space axis, as in a one-dimensional array of ring resonators whose internal mode ladders create an effective two-dimensional $(l,m)$ lattice. In that setting, spatial hopping comes from inter-ring coupling, while frequency-axis hopping comes from resonant modulation, and modulation phase plays the role of a Peierls phase or gauge potential (Yuan et al., 2015). In another branch, the frequency lattice is combined with an auxiliary spectral element that selectively removes resonant sites and thereby carves sharp open boundaries into what would otherwise be an effectively unbounded comb (Dutt et al., 2022). Other realizations hybridize the frequency lattice with a qubit I/O node and a SQUID boundary modulator in superconducting circuits, or with biphoton correlation space in thin-film lithium niobate, where the operative state space is not a single-particle chain but a two-photon correlation lattice (Xie et al., 14 Feb 2026, Javid et al., 2022).

A recurring misconception is to equate a synthetic frequency dimension with generic sideband generation. The surveyed architectures are more specific. They require a controlled mapping from mode indices to lattice sites, well-defined intersite couplings, and a simulator-level interpretation of hopping range, onsite tilt, boundary conditions, gauge phase, or interaction sector. Sidebands alone do not provide that structure.

2. Core hardware primitives

The hardware implementations vary, but they repeatedly decompose into a small set of modules: a multimode frequency backbone, a modulation engine, optional boundary or truncation elements, and a readout interface. The backbone may be a fiber ring, an integrated microresonator, a superconducting CPW resonator, or a long superconducting coaxial cable. The modulation engine may be an electro-optic phase modulator, a flux-driven SQUID termination, or an acousto-optic cavity actuator. Boundary control may be introduced by auxiliary resonators, by dispersion-engineered cutoff, or by selective mode hybridization. Readout may be optical transmission spectroscopy, heterodyne detection, photon-coincidence mapping, or qubit-assisted tomography.

Platform	Key hybridization	Representative capability
Dynamically modulated ring plus auxiliary ring	Synthetic lattice + static spectral site removal	Sharp open boundaries in frequency space
1D ring array with internal mode ladders	Real-space array + synthetic frequency axis	Gauge fields and one-way edge transport
LN racetrack microresonator	SPDC source + EO hopping in one cavity	Nearly $400\times400$ biphoton correlation lattice
Superconducting cable + SQUID + qubit	Multimode memory + programmable couplings + quantum I/O	Single-photon random walks, Bloch oscillations, gauge flux
Cascaded modulated micro-rings or cavity AOM	Frequency lattice + dense linear scattering layer	Arbitrary linear transformations or full-range spectral mixing

In optical ring platforms, the ring free spectral range sets the lattice spacing in frequency. In one integrated lithium-niobate implementation, the resonator is pumped at $776\,\mathrm{nm}$, generates a telecom-band SPDC comb centered around $m$0, and has $m$1 mode spacing with an inferred $m$2-dB bandwidth of about $m$3, corresponding to roughly $m$4 cavity modes. Because signal and idler occupy opposite sides of the degenerate center, the relevant correlation space is nearly $m$5 (Javid et al., 2022). In a superconducting implementation, a $m$6 low-loss aluminum coaxial cable provides a very small $m$7, more than $m$8 adjacent modes within a few hundred MHz, and a transmon qubit provides site-selective state preparation and readout at the single-photon level (Xie et al., 14 Feb 2026).

Architecturally, these platforms separate functions. The resonator supplies a dense frequency register. Modulation synthesizes the graph connectivity. Additional resonant or nonlinear blocks impose geometry, interactions, or symmetry breaking. This separation is one reason the architecture is reconfigurable: changing the drive waveform, phase, or auxiliary alignment modifies the effective Hamiltonian without altering the physical lattice backbone.

3. Boundaries, truncation, and finite synthetic geometry

Ordinary synthetic frequency lattices do not naturally possess hard walls. In a single modulated ring, many modes participate, the modulation couples them on a regular grid, and the resulting system approximates an effectively infinite chain rather than a finite one. Loss limits propagation distance, but the papers are explicit that dissipation is not a boundary: it does not generate reflection, discrete finite-chain eigenmodes, or bulk-edge correspondence in the topological sense (Dutt et al., 2022).

A decisive architectural advance is the introduction of auxiliary resonators that selectively hybridize with chosen modes of the primary ring. In the coupled-ring boundary scheme, the primary resonator creates the synthetic frequency lattice, while a statically coupled auxiliary ring with a different free spectral range strongly hybridizes with aligned resonances. Those resonances split into doublets shifted away from the regular $m$9 grid, so modulation at $\Omega_M=\Omega_R$0 can no longer connect through them efficiently. In the effective lattice language, selected sites are removed and the chain is cut into finite segments with open boundaries. The paper further identifies a practical condition for strong confinement: the induced splitting should exceed $\Omega_M=\Omega_R$1 (Dutt et al., 2022).

This boundary engineering has direct physical consequences. Experiments in fiber rings show confinement of sideband generation, interference fringes interpreted as reflections from synthetic boundaries, and discretization of the band structure. In reciprocal-space measurements, the continuous cosine-like band of the unbounded chain becomes a discrete set of resonances matching both full Floquet scattering simulations and a finite open-boundary tight-binding model (Dutt et al., 2022). Related boundary logic also appears in other platforms. In a 2D microring topological lattice with a synthetic frequency axis, selectively modulated dynamic link rings create a screw dislocation along the frequency dimension, so that encircling the defect changes the mode index and binds a one-way topological channel in synthetic frequency space (Lin et al., 2018). In theoretical flat-band architectures, dispersion and FSR commensurability determine where practical synthetic boundaries appear, while phase-engineered flux can isolate flat bands from dispersive ones and make localization more robust to finite-bandwidth excitation and group-velocity dispersion (Yu et al., 2020).

A further misconception is to treat any spectral cutoff as equivalent to a boundary. The surveyed work distinguishes sharply between three cases: lossy decay, off-resonant suppression from dispersion, and engineered site removal by auxiliary hybridization. Only the last gives the hard-wall truncation needed for the boundary phenomena emphasized in these simulator architectures (Dutt et al., 2022).

4. Gauge fields, topology, and multiband synthetic matter

Once frequency modes are organized as lattice sites, modulation phase becomes a natural gauge control. In a one-dimensional array of coupled ring resonators, each ring supports modes $\Omega_M=\Omega_R$2, spatial hopping links neighboring rings at fixed $\Omega_M=\Omega_R$3, and modulation at $\Omega_M=\Omega_R$4 links $\Omega_M=\Omega_R$5 within each ring. After the rotating-wave approximation, the effective hopping along the synthetic frequency axis acquires the phase $\Omega_M=\Omega_R$6, so a spatially varying $\Omega_M=\Omega_R$7 implements a gauge potential in the hybrid $\Omega_M=\Omega_R$8 space. Choosing $\Omega_M=\Omega_R$9 gives a Landau-gauge realization of a uniform magnetic field, with a Harper–Hofstadter-type bandstructure and one-way edge states in the hybrid real-space/frequency lattice (Yuan et al., 2015).

Synthetic-frequency gauge engineering is not restricted to real-space arrays. In a single modulated ring, multiple harmonics of the modulation produce short- and long-range couplings $m\leftrightarrow m\pm1$0 and complex hopping phases, allowing asymmetric, nonreciprocal dispersions and direct band-structure spectroscopy through time-resolved transmission. A drive of the form

$m\leftrightarrow m\pm1$1

generates nearest- and next-nearest-neighbor couplings with a synthetic gauge phase $m\leftrightarrow m\pm1$2, and the measured band can become strongly asymmetric for $m\leftrightarrow m\pm1$3 and $m\leftrightarrow m\pm1$4 (Dutt et al., 2019).

Boundary engineering closes the loop between bulk topology and edge transport. In a two-leg quantum Hall ladder realized with two coupled main rings, the ring index forms the ladder legs and the frequency index forms the rung direction. An auxiliary ring truncates one leg, so $m\leftrightarrow m\pm1$5, and the effective Hamiltonian includes a modulation-phase flux $m\leftrightarrow m\pm1$6. For $m\leftrightarrow m\pm1$7, the response reflects from the synthetic boundary and produces fringes. For $m\leftrightarrow m\pm1$8, the system supports one-way chiral modes; the response propagates only in one frequency direction and no interference fringes appear despite the presence of the boundary, in agreement with bulk-edge correspondence (Dutt et al., 2022).

A different multiband use of the architecture appears in time-boundary physics. Two coupled dynamically modulated rings, with opposite-sign intraring hoppings $m\leftrightarrow m\pm1$9 and $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$0, realize a two-band synthetic ladder with Bloch Hamiltonian

$H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$1

and band energies

$H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$2

Because the bands are centered around a non-zero reference energy $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$3, time reflection and time refraction can be induced by changing $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$4 and $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$5 on a timescale set by the band splitting rather than the optical carrier frequency. The experimentally cited scales are $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$6 and $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$7, so microwave-rate control is sufficient for optical time-boundary effects in synthetic frequency space (Long et al., 2022).

Taken together, these works show that hybrid-frequency architectures support not only one-dimensional tight-binding transport but also magnetic flux, nonreciprocal band engineering, real-space/frequency topological models, and multiband temporal-scattering phenomena.

5. Quantum, interacting, and correlated extensions

The transition from linear single-particle transport to genuinely quantum or interacting simulation is a defining extension of the architecture. One route uses quantum-correlated sources. In a thin-film LN racetrack microresonator, a periodically poled region generates a broadband biphoton frequency comb by SPDC, while intracavity electro-optic modulation couples adjacent frequency modes. Because the state space is a two-photon entangled one, the natural simulator coordinates are $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$8, and the experimentally relevant observables are joint spectral intensity and biphoton temporal correlations rather than single-particle mode amplitudes. This produces a nearly $H = J\sum_m b_m^\dagger b_{m+1} + \mathrm{H.c.},$9 quantum-correlated synthetic lattice in which the authors demonstrate quantum random walks, Bloch oscillations, and multi-level Rabi oscillations in time and frequency correlation space (Javid et al., 2022).

A second route uses genuine photon-photon interactions. In a dynamically modulated ring resonator with $\varepsilon(k)=2J\cos k,$0 nonlinearity, bare four-wave mixing would ordinarily yield long-range interactions along the frequency axis. The key design move is to build the ring from two waveguide sections with opposite group-velocity dispersion, so the overall mode spacing remains approximately uniform while unwanted nonlinear channels become phase mismatched. In the appropriate regime, only self-phase modulation and cross-phase modulation survive effectively, and in a fixed-photon-number sector the interaction reduces to a Bose–Hubbard form with local onsite interaction. Numerical demonstrations on a $\varepsilon(k)=2J\cos k,$1-site synthetic chain show photon blockade in synthetic frequency space and agreement between the engineered nonlinear Hamiltonian and the ideal Bose–Hubbard model (Yuan et al., 2019).

A further interacting extension studies chaos and localization. In a ring resonator with tailored $\varepsilon(k)=2J\cos k,$2, periodic modulation with $\varepsilon(k)=2J\cos k,$3 harmonics, and an auxiliary resonator imposing open boundaries, the effective Hamiltonian contains long-range hopping $\varepsilon(k)=2J\cos k,$4, a linear tilt $\varepsilon(k)=2J\cos k,$5, and quartic interaction terms. By tuning $\varepsilon(k)=2J\cos k,$6, $\varepsilon(k)=2J\cos k,$7, $\varepsilon(k)=2J\cos k,$8, $\varepsilon(k)=2J\cos k,$9, $k\leftrightarrow t$0, and $k\leftrightarrow t$1, the same hardware is shown to support integrable, chaotic, and Stark-localized regimes. The paper also develops an interferometric method to access the spectral form factor and uses the phase structure to generate and then localize controllable frequency-entangled multiphoton states (Wang et al., 2024).

Superconducting circuits supply a complementary quantum route. A $k\leftrightarrow t$2 superconducting coaxial cable, a SQUID-based boundary modulator, and a transmon qubit realize a programmable single-photon synthetic lattice with effective Hamiltonian

$k\leftrightarrow t$3

The qubit provides quantum-state initialization and mode-resolved detection; the SQUID sets hopping range, hopping phase, and synthetic force. Experiments observe single-photon quantum random walks, Bloch oscillations with measured period $k\leftrightarrow t$4 consistent with $k\leftrightarrow t$5 for $k\leftrightarrow t$6, nonadiabatic unidirectional frequency conversion under rapid force reversal, and flux-controlled band asymmetry in a triangular-chain connectivity graph (Xie et al., 14 Feb 2026).

Hybridization with discrete emitters also enables waveguide-QED analogues. A dynamically modulated superconducting resonator plus a tailored three-level artificial atom can be reduced to an effective giant atom coupled to two separated synthetic frequency sites, with the external drive phase controlling chirality and cascaded interactions along the synthetic dimension (Du et al., 2021).

6. Readout, programmability, and operator synthesis

The architecture is unusually rich in readout channels because the synthetic coordinate is frequency. In dynamically modulated rings, time-resolved transmission directly samples the synthetic quasimomentum through $k\leftrightarrow t$7, so scanning laser detuning and recording the transmitted waveform reconstructs the bandstructure. Time-averaged transmission instead measures the synthetic density of states and reveals van Hove singularity behavior in the nearest-neighbor cosine band (Dutt et al., 2019). In bounded synthetic lattices, heterodyne-resolved sideband spectra reconstruct mode occupations and display boundary-induced interference patterns or chiral boundary transport (Dutt et al., 2022). In quantum-correlated LN platforms, the operative observables are joint spectral intensity and temporal coincidence histograms $k\leftrightarrow t$8, with Bloch oscillations visible only in second-order correlations rather than average intensities (Javid et al., 2022). In superconducting implementations, mode-selective swap-back to a qubit reconstructs site populations and quadratures, enabling direct band spectroscopy from the complex wavefunction $k\leftrightarrow t$9 (Xie et al., 14 Feb 2026).

A closely related line of work treats the frequency lattice as a programmable linear operator. In cascaded dynamically modulated micro-rings side-coupled to a common waveguide, the scattering matrix

$(l,m)$0

acts between input and output frequency channels, with $(l,m)$1 set by modulation harmonics. By inverse design of the short- and long-range couplings, arbitrary unitary and contractive non-unitary transformations can be synthesized with near-unity fidelity on truncated synthetic spaces, and auxiliary rings provide the finite boundaries required for a controlled $(l,m)$2-site subspace (Buddhiraju et al., 2020). A complementary integrated cavity acousto-optic modulator on an AlN-on-SOI platform acts as a dense, complex-valued linear map on frequency bins. At $(l,m)$3, the device achieves $(l,m)$4, a maximum $(l,m)$5, roughly $(l,m)$6 synthetic sites over $(l,m)$7, and a $(l,m)$8 coherent matrix-vector-multiplication demonstration at $(l,m)$9 (Zhao et al., 2021).

These operator-synthesis papers clarify another common misconception. “Fully connected” in the frequency-domain layer means that one input site can couple coherently to many or all accessible output sites within the sideband span. It does not mean that a single modulation stage independently sets every matrix element. The realized operator remains structured by the physics of the cavity response and modulation waveform (Zhao et al., 2021, Buddhiraju et al., 2020).

7. Constraints, interpretive issues, and development trajectory

Several constraints recur across the literature. The synthetic-lattice picture depends on nearly equally spaced modes; group-velocity dispersion, imperfect FSR matching, and finite usable bandwidth limit lattice size and translational invariance (Dutt et al., 2019, Dutt et al., 2022). Boundary sharpness depends on sufficiently strong auxiliary-induced splitting relative to the modulation scale, while topological or gauge implementations require stable phase control across multiple drives or resonators (Dutt et al., 2022, Yuan et al., 2015). In integrated quantum platforms, comb bandwidth, detector bandwidth, and pump scattering under strong modulation constrain how much of the nominal lattice can be populated or read out (Javid et al., 2022). In superconducting implementations, small FSR enables many sites but increases mode crowding and makes readout serial rather than parallel (Xie et al., 14 Feb 2026). In interacting optical designs, realizing simultaneously strong nonlinearity, large and controlled GVD, low loss, and broad equal mode spacing remains technically demanding (Yuan et al., 2019, Wang et al., 2024).

Interpretively, “hybrid” has more than one meaning in this area. In some papers it means real-space plus frequency space, as in resonator arrays or 3D weak topological-insulator constructions (Yuan et al., 2015, Lin et al., 2018). In others it means dynamic modulation plus static spectral boundary engineering, as in the boundary-forming auxiliary-ring architecture (Dutt et al., 2022). In still others it means frequency lattices enlarged by biphoton correlation space or interfaced with qubit I/O and boundary modulators (Javid et al., 2022, Xie et al., 14 Feb 2026). The term therefore denotes a family resemblance rather than a single fixed hardware recipe.

The development trajectory is nonetheless clear. Early work established the frequency lattice itself, together with direct band spectroscopy, long-range synthetic hopping, and gauge potentials (Dutt et al., 2019). Subsequent architectures added sharp boundaries, hybrid real-space/frequency topology, programmable linear scattering, dense all-to-all spectral mixing, and chip-scale quantum-correlated state spaces (Dutt et al., 2022, Yuan et al., 2015, Buddhiraju et al., 2020, Zhao et al., 2021, Javid et al., 2022). More recent work extends the architecture into the single-photon superconducting regime, multiband temporal-boundary physics, and interacting few-body dynamics with chaos, localization, and controllable frequency entanglement (Xie et al., 14 Feb 2026, Long et al., 2022, Wang et al., 2024).

A plausible implication is that the mature form of the architecture will be modular rather than monolithic: a frequency backbone for large synthetic extent, modulation layers for programmable graph synthesis, auxiliary spectral elements for boundaries and defects, nonlinear or quantum nodes for interactions, and frequency-resolved tomography for diagnostics. That modular picture is strongly suggested by the surveyed platforms, even though different papers emphasize different modules.