Superconducting Bose-Hubbard Qutrit Arrays

• Superconducting Bose-Hubbard Qutrit Arrays are platforms where qutrits (three-level systems) simulate bosonic hopping and tailored on-site interactions.
• They integrate hardware-native high-dimensional control using transmon and fluxonium devices, enabling complex entanglement and communication protocols.
• Advanced techniques like Floquet engineering and dispersive control yield tunable interactions, effective error mitigation, and many-body phase simulations beyond standard models.

Superconducting Bose-Hubbard qutrit arrays are superconducting-circuit many-body platforms in which the local Hilbert space is restricted to three states while intersite couplings emulate bosonic hopping, on-site interactions, and, in several important variants, explicitly driven-dissipative dynamics. In the most common implementations, the three lowest states $|g\rangle, |e\rangle, |f\rangle$ of a transmon encode a qutrit and can be interpreted as $0,1,2$ bosons per site; closely related constructions use resonator arrays with Josephson-induced Kerr nonlinearities, fluxonium qutrit arrays with strongly tunable matrix elements, or modular qutrit nodes linked by microwave channels. Across these realizations, the field has developed along two coupled directions: Bose-Hubbard quantum simulation and hardware-native high-dimensional control, entanglement, and communication (Li et al., 28 Jun 2026, Amelio et al., 29 Jan 2026, Castillo-Moreno et al., 21 Aug 2025, Tsomokos et al., 2010).

1. Local Hilbert space and qutrit encodings

The defining structural choice is the restriction of each site to a three-dimensional local space. In transmon-based arrays, the $|g\rangle,|e\rangle,|f\rangle$ manifold is used directly as a qutrit, and in one important mapping these states correspond to $|S=1,S_z=-1\rangle$, $|S=1,S_z=0\rangle$, and $|S=1,S_z=+1\rangle$, allowing a spin-1 representation of many-body physics (Wang et al., 2023). In modular communication architectures, the same three transmon levels are interpreted as $0,1,2$ bosons per site, so that qutrit state transfer can be read as boson hopping in a truncated Bose-Hubbard picture (Li et al., 28 Jun 2026).

Fluxonium extends the same idea with greater Hamiltonian tunability. The three qutrit states need not be the lowest three levels; rather, one can select three sufficiently isolated local states and tune their energies and matrix elements with an external flux bias. This yields four operational regimes—plasmon-plasmon, fluxon-fluxon, plasmon-fluxon, and fluxon-plasmon—distinguished by the physical character of the $0\leftrightarrow1$ and $1\leftrightarrow2$ transitions (Amelio et al., 29 Jan 2026).

The broader motivation for qutrits is twofold. First, they enlarge the local Hilbert space without the redundant basis overhead of qubit encodings for intrinsically three-state problems; this was demonstrated explicitly in single-qutrit simulation of three-flavor neutrino oscillations on a superconducting transmon (Nguyen et al., 2022). Second, the higher transmon levels can be promoted from a source of leakage to a computational and simulational resource, a viewpoint made explicit in recent qutrit-control and qutrit-link experiments (Yu et al., 9 Jul 2025, Li et al., 28 Jun 2026).

Historically, superconducting qutrits were already studied outside the Bose-Hubbard context. The rf SQUTRID realized a quasi-three-level coherent system based on a triple-well potential and a quantum superposition of three macroscopically distinct flux states, while resonant cavity protocols used a single superconducting qutrit as a coupler to generate $n$-photon GHZ states in $0,1,2$0 cavities (Shnyrkov et al., 2011, Yang et al., 2012). These antecedents established the physical legitimacy of three-level superconducting devices before their systematic use as lattice sites and bosonic truncations.

2. Bose-Hubbard Hamiltonians and their qutrit truncations

The canonical superconducting realization of Bose-Hubbard physics uses resonators as bosonic sites and superconducting nonlinear elements to generate hopping and on-site interactions. A standard form is

$0,1,2$1

with $0,1,2$2 induced by coupling each resonator to a superconducting qubit in the dispersive regime (Tsomokos et al., 2010). In qutrit arrays, the same bosonic language is retained, but the local occupation is truncated to $0,1,2$3, yielding a three-state on-site space rather than a full harmonic mode.

This truncation can be engineered dynamically. In superconducting Bose-Hubbard qutrit arrays with Floquet control, periodic flux modulation produces the effective Hamiltonian

$0,1,2$4

with

$0,1,2$5

By choosing $0,1,2$6 and $0,1,2$7 so that $0,1,2$8 with $0,1,2$9, the engineered model matches the canonical Bose-Hubbard form with adjustable $|g\rangle,|e\rangle,|f\rangle$0 (Wang et al., 2 Sep 2025).

Driven-dissipative generalizations are equally central. In a two-dimensional square lattice, pair driving and nonlinear two-boson loss lead to the non-equilibrium Hamiltonian

$|g\rangle,|e\rangle,|f\rangle$1

with $|g\rangle,|e\rangle,|f\rangle$2, where $|g\rangle,|e\rangle,|f\rangle$3 is the interaction and $|g\rangle,|e\rangle,|f\rangle$4 represents nonlinear two-boson loss (Wang et al., 2020). This model is not an equilibrium Bose-Hubbard Hamiltonian with weak perturbations; its steady states are fixed points of open-system dynamics.

Fluxonium qutrit arrays go further and realize an extended bosonic theory beyond the textbook Bose-Hubbard paradigm: $|g\rangle,|e\rangle,|f\rangle$5 Here $|g\rangle,|e\rangle,|f\rangle$6 is correlated single-particle hopping, $|g\rangle,|e\rangle,|f\rangle$7 controls occupation-dependent hopping, $|g\rangle,|e\rangle,|f\rangle$8 is the on-site interaction, $|g\rangle,|e\rangle,|f\rangle$9 is pair hopping, and $|S=1,S_z=-1\rangle$0 collects non-local diagonal interactions; the local truncation imposes a three-body hard-core constraint $|S=1,S_z=-1\rangle$1 (Amelio et al., 29 Jan 2026). A common misconception is that superconducting qutrit arrays merely reproduce a soft-core Bose-Hubbard model with three local states. The fluxonium results show that, depending on circuit regime and coupling type, pair hopping and non-local interactions can be leading-order terms rather than perturbative corrections.

3. Hardware architectures, coherent control, and verification

Several architectures coexist under the same thematic umbrella. One route uses one-dimensional or two-dimensional resonator lattices with Josephson arrays as nonlinear superinductors; a 21-site chain of lumped-element resonators realized a driven Bose-Hubbard simulator with nearest-neighbor capacitive coupling and direct end-port pumping and readout (Castillo-Moreno et al., 21 Aug 2025). Another route uses transmon qutrits laid out in one dimension, with each neighboring pair sharing a common linear microwave resonator in the strong dispersive regime (Wang et al., 2023). A third route is modular: each node contains a transmon qutrit, a transmission resonator, a readout resonator, and a tunable Purcell filter, and remote nodes are connected by a one-meter microwave channel (Li et al., 28 Jun 2026).

Programmability depends on hardware-native qutrit control. In a $|S=1,S_z=-1\rangle$2-type superconducting transmon, direct coherent SU(3) control has been used to implement qutrit Hadamard and $|S=1,S_z=-1\rangle$3 gates in $|S=1,S_z=-1\rangle$4, with average fidelity $|S=1,S_z=-1\rangle$5 verified by randomized benchmarking. The same work implemented the full single-qutrit Clifford group $|S=1,S_z=-1\rangle$6 of 216 elements and used virtual $|S=1,S_z=-1\rangle$7 control for phase Clifford generators (Yu et al., 9 Jul 2025). This addresses a central hardware constraint of superconducting qutrit arrays: the absence of a direct $|S=1,S_z=-1\rangle$8 transition in the practical $|S=1,S_z=-1\rangle$9-type configuration.

State verification in qutrit devices has developed both operational and information-theoretic forms. For a single superconducting qutrit implemented with Josephson junctions, the positivity of entropy-energy and von Neumann entropic relations was proposed as a consistency check for tomography. The entropy-energy inequality

$|S=1,S_z=0\rangle$0

and the qubit-portrait analog of subadditivity for a noncomposite qutrit provide tests of the physicality of reconstructed density matrices (Man'ko et al., 2016).

Remote qutrit operations have also moved beyond local control. A superconducting qutrit link between independently packaged nodes transferred arbitrary qutrit states with a mean transferred-state fidelity of $|S=1,S_z=0\rangle$1 and a qutrit process fidelity of $|S=1,S_z=0\rangle$2, exceeding both the classical qutrit-transfer benchmark and the best possible average fidelity of an effective qubit channel used to transmit an arbitrary qutrit. Using partial-transfer operations, the same platform reconstructed a remote two-qutrit state with negativity $|S=1,S_z=0\rangle$3, dense-coding capacity $|S=1,S_z=0\rangle$4 bits, and CGLMP parameter $|S=1,S_z=0\rangle$5 (Li et al., 28 Jun 2026).

Because qutrit circuits are more noise-sensitive than qubit circuits, mitigation has become part of the hardware stack. On a superconducting transmon qutrit processor, randomized compiling, noiseless output extrapolation, and readout calibration improved a three-qutrit GHZ-state fidelity from $|S=1,S_z=0\rangle$6 to $|S=1,S_z=0\rangle$7, and produced up to $|S=1,S_z=0\rangle$8 improvement in multipartite entanglement and random-circuit-sampling benchmarks (Goss et al., 2023).

4. Many-body phases and non-equilibrium phenomena

A major branch of the literature studies open-system Bose-Hubbard physics realized in superconducting circuits. In the pair-driven, pair-damped two-dimensional model, the most unstable modes satisfy $|S=1,S_z=0\rangle$9, forming a closed contour in momentum space termed the Bose surface. For generic $|S=1,S_z=+1\rangle$0, the steady state selects a single pair of opposite momenta from this contour and forms a striped density wave. For the nested square-lattice case $|S=1,S_z=+1\rangle$1, all modes on the Bose surface are equally populated, producing a homogeneous state with uniform density but nontrivial phase pattern. Fluctuations around the condensate are characterized by a purely diffusive mode $|S=1,S_z=+1\rangle$2 rather than propagating superfluid sound (Wang et al., 2020).

Experimentally, dissipative Bose-Hubbard phase transitions have been observed in a 21-resonator superconducting chain. As pump frequency or power is swept, all resonator modes exhibit an abrupt and simultaneous frequency jump, indicating a multimode dim-to-bright first-order transition. The phase diagram can be constructed from the emitted photon number, and time-resolved switching measurements near the transition show synchronized jumps between metastable phases with dwell times ranging from a few ms up to $|S=1,S_z=+1\rangle$3. At moderate pump powers, a single-mode mean-field theory with cross-Kerr interactions quantitatively reproduces the transition line (Castillo-Moreno et al., 21 Aug 2025).

Transport and localization phenomena provide a complementary probe. In a five-site chain of capacitively coupled Xmon transmons simulating a driven-dissipative Bose-Hubbard model, direct transmission spectroscopy and cross-Kerr spectroscopy resolved normal modes, doublon bands, and the transition from the linear regime to photon blockade. For vanishing disorder the system retained many-particle coherence despite strong coupling to open spaces, whereas controllable disorder suppressed non-local photon transmission (Fedorov et al., 2020).

Attractive interactions change the equilibrium phase structure qualitatively. In the one-dimensional disordered attractive Bose-Hubbard model realized by a chain of superconducting transmon qubits or cold atoms, the ground state may be superfluid, a W state, or a localized collapsed state. At strong disorder all bosons localize near a single site, unlike the Bose glass of the repulsive model; at weak disorder and small hopping the W phase is a multi-site, multi-particle entangled superposition of states in which all bosons occupy a single site, but its robustness against disorder decreases as the total number of bosons increases (Mansikkamäki et al., 2021).

5. Correlation dynamics, entanglement, and dissipative state engineering

Qutrit arrays make it possible to study few-particle correlation dynamics in a locally bosonic but strongly constrained setting. In tunable superconducting Bose-Hubbard qutrit arrays, quantum walks with Floquet-engineered on-site interaction $|S=1,S_z=+1\rangle$4 show the crossover from bosonic bunching at $|S=1,S_z=+1\rangle$5 to fermionic antibunching as $|S=1,S_z=+1\rangle$6 increases. The corresponding two-particle entanglement and quantum correlation dynamics were quantified through negativity and quantum discord. Depending on the initial state, increasing $|S=1,S_z=+1\rangle$7 can strongly suppress entanglement propagation while leaving discord with considerably larger amplitude, or make both measures relatively insensitive to $|S=1,S_z=+1\rangle$8 (Wang et al., 2 Sep 2025).

Dissipative preparation of many-body entangled states is an equally important theme. In a one-dimensional transmon qutrit array, nearest-neighbor pairs share dissipative microwave resonators, and the elimination of population in the $|S=1,S_z=+1\rangle$9 subspace stabilizes the AKLT state up to the edge-mode configuration. The protocol is formulated through pairwise dissipative processes and pulse-level simulations of real devices; for increasing size from two to four qutrits, the reported fidelity decreases from $0,1,2$0 to $0,1,2$1, while stabilization time increases (Wang et al., 2023).

Entanglement-generation protocols using superconducting qutrits preceded these array studies and remain relevant as building blocks. A single superconducting qutrit can act as a central coupler for $0,1,2$2 microwave cavities and generate the photonic GHZ state

$0,1,2$3

using only resonant qutrit-cavity and qutrit-pulse operations, with no measurement on the qutrit or cavity photons (Yang et al., 2012). This suggests a modular route from single-qutrit control to distributed many-body state engineering.

High-dimensional communication metrics reinforce the same theme. In the remote-link setting, negativity, dense-coding capacity, and CGLMP violation all exceeded corresponding qubit or local bounds, showing that the three-level structure of superconducting nodes can support genuinely high-dimensional entanglement rather than a convex mixture of qubit resources (Li et al., 28 Jun 2026).

6. Scalability, topology, and directions beyond the standard Bose-Hubbard model

The superconducting platform is attractive partly because array topology is not fixed by geometry alone. Tunable couplers and multilayer circuit fabrication were proposed early on as a route to arbitrary graph connectivity, allowing higher-dimensional, fractal, and higher-genus lattice realizations, including topology quenches between structures such as tori and Klein bottles (Tsomokos et al., 2010). Later roadmaps for floating-transmon arrays emphasized that superconducting devices can prepare highly excited many-body states, not only ground states or thermal states, and can access observables such as correlation lengths and entanglement-entropy scaling from area law to volume law behavior (Yanay et al., 2019).

At the same time, several limitations recur across the literature. Qutrits suffer more complex errors and greater noise sensitivity than qubits, making scaling difficult without explicit mitigation (Goss et al., 2023). In dissipative spin-1 stabilization protocols, imperfect matching of dispersive shifts, finite $0,1,2$4, qutrit decoherence, and the requirement that each bond have its own resonator and control infrastructure all reduce fidelity as system size increases (Wang et al., 2023). For Bose-Hubbard simulation with transmons, the attractive interaction inherited from negative anharmonicity and the finite on-site Hilbert space limit the accessible boson number per site (Mansikkamäki et al., 2021). Even in single-qutrit settings, extending from local control to multi-qutrit entanglement requires native high-fidelity entangling gates rather than only precise single-qutrit rotations (Nguyen et al., 2022).

A further conceptual point is that the field has already moved beyond the narrow interpretation of “qutrit Bose-Hubbard array” as a three-level approximation to a standard lattice boson model. Fluxonium qutrit arrays naturally generate correlated single-particle hopping, pair hopping, non-local interactions, and a three-body hard-core constraint, and were explicitly proposed as simulators of strongly correlated bosonic matter beyond the Bose-Hubbard paradigm, with possible applications to lattice gauge theories and non-Abelian topological states (Amelio et al., 29 Jan 2026). This suggests that the Bose-Hubbard language remains foundational but is no longer exhaustive of what superconducting qutrit arrays are designed to realize.