Technical Summary: Universal Monitored Dynamics in Multimode Bosonic Systems

Problem Statement

While measurement-induced phase transitions (MIPTs) have been extensively studied in qubit-based circuit models and probed experimentally in superconducting and trapped-ion systems, the behavior of monitored dynamics in multimode bosonic systems remains less understood. Previous theoretical work on monitored free bosons with Gaussian unitaries and Gaussian measurements suggested no MIPT at finite measurement rates, with systems flowing to an area-law entangled phase. However, circuit quantum electrodynamics (cQED) platforms offer access to non-Gaussian ingredients, such as parity measurements and interacting gates, which are absent in purely Gaussian models. The central problem addressed is whether these non-Gaussian elements in bosonic systems lead to novel entanglement phases and MIPTs distinct from the qubit case, and whether such dynamics can be realized on near-term hardware.

Methodology

The authors propose an experimental and theoretical framework to study monitored bosonic dynamics using circuit QED.

Theoretical Models

The study utilizes brickwork circuits composed of $L$ bosonic modes with a fixed total photon number $Q = L/2$. The dynamics consist of discrete layers containing:

1. Unitary Gates:
   • Beam-Splitter Gates: Implemented as nearest-neighbor hopping operations. The authors distinguish between Fixed-Phase (BSFP) gates, which generate SU(2) rotations about a single axis, and Randomized-Phase (BSRP) gates, which induce stronger mode mixing and scrambling.
   • On-Site Hubbard Interactions: Non-Gaussian gates implemented via Selective Number-dependent Arbitrary Phase (SNAP) gates, introducing interactions of the form $e^{-iU(\hat{a}^\dagger \hat{a})^2}$.
2. Measurements: Local Quantum Non-Demolition (QND) mid-circuit measurements performed with probability $p$. The study focuses primarily on parity measurements (projecting onto even/odd photon number subspaces) but also considers photon-number (density) measurements.

Diagnostics

To probe the dynamics without the exponential overhead of experimental post-selection, the authors employ two primary diagnostics:

1. Ancilla Purification: An ancillary reference qubit is prepared in a globally entangled state with the bosonic modes. The transition is diagnosed via the entanglement entropy of the ancilla ($S_R$) conditioned on measurement trajectories. A transition from volume-law to area-law entanglement corresponds to a transition from slow (exponential in $L$) to fast ($O(1)$) purification times.
2. Learnability Probe: A classical decoder attempts to distinguish between two initially orthogonal random states based on the measurement record. The accuracy of this decoder serves as a post-selection-free order parameter.

Experimental Proposal

The authors propose realizing these circuits on two near-term cQED architectures:

1. Linear Cavity Array: A 1D array of coaxial $\lambda/4$ cavities coupled by tunable couplers, with individual transmon ancillas for each mode.
2. Cascaded Random Access Quantum Memory (RAQM): A multimode storage cavity coupled to a buffer cavity via a tunable coupler. The buffer cavity interfaces with a transmon for universal control (SNAP gates) and QND measurements. This architecture suppresses spurious cross-Kerr interactions and reduces crosstalk.

The authors perform end-to-end Lindblad master-equation simulations to assess feasibility, modeling dominant noise sources including cavity decay ($T_1$), transmon relaxation/dephasing, and gate infidelities.

Key Results

1. Standard MIPT in Interacting/Scrambling Circuits

For circuits containing on-site Hubbard interactions ($U=2$) or randomized-phase beam splitters (BSRP), the system exhibits behavior consistent with a conventional MIPT.

• Transition: A clear phase transition occurs at a critical measurement rate $p_c \approx 0.3$.
• Phases: Below $p_c$, the system is in a volume-law phase with slow purification. Above $p_c$, it enters an area-law phase with $O(1)$ purification times.
• Critical Exponent: The data collapses with a dynamical critical exponent $z \approx 1$, similar to qubit systems.
• Learnability: The decoder accuracy mirrors the entanglement entropy transition, validating the learnability probe.

2. Novel Critical Phase in Pure BSFP Circuits

A striking deviation from standard MIPT behavior is observed in circuits composed purely of fixed-phase beam-splitter gates ($U=0$, BSFP) with parity measurements.

• Absence of Standard Transition: No clear crossing of entropy curves indicative of a standard $z=1$ transition is found.
• Critical-like High-Measurement Regime: For high measurement rates ($p \gtrsim 0.3$), the system enters a phase where the ancilla purification time scales linearly with system size ($O(L)$).
• Significance: This linear scaling is typically associated only with the critical point in standard MIPTs, yet here it persists across a range of $p$. The authors refer to this as a critical phase.
• Role of Scrambling: Counterintuitively, increasing the scrambling power (by adding Hubbard gates or randomizing phases) decreases the purification time in the high-measurement regime from $O(L)$ to $O(1)$. This suggests that in bosonic systems with parity measurements, scrambling aids measurements in revealing information rather than hiding it.

3. Measurement Type Dependence

The nature of the phase depends critically on the measurement type. If density (photon number) measurements are used instead of parity measurements, the ancilla purifies on $O(1)$ timescales for all $p$, even in the purely BSFP case. This highlights that the incomplete collapse of the local Hilbert space by parity measurements is essential for the novel critical phase.

4. Experimental Feasibility

Simulations incorporating realistic noise parameters (cavity lifetimes of 1.5 ms to 10 ms, transmon $T_1 \approx 200$ $\mu$s) indicate that the signatures of these phases are observable.

• Residual Entropy: Noise induces a finite residual entropy at high measurement rates.
• Mitigation: The authors show that subtracting this noise-induced residual entropy allows the recovery of the underlying ideal dynamics, a technique previously used in qubit MIPT experiments.
• Resource Estimates: For a system size of $L=4$, the total acquisition time is estimated to be on the order of a few hundred hours, which is within reach of current hardware.

Significance and Claims

The paper claims to advance the understanding of monitored bosonic dynamics on two fronts:

1. Theoretical Discovery: It uncovers a novel critical-like phase in monitored bosonic systems driven by fixed-phase beam splitters and parity measurements, characterized by system-size-dependent purification times. This phase is distinct from the standard MIPT found in qubit systems and generic bosonic circuits.
2. Experimental Roadmap: It provides a detailed, feasible protocol for observing these phenomena on near-term multimode cQED platforms (specifically the RAQM architecture). The work demonstrates that the unique physics of bosonic statistics (bunching) and the tunability of measurement strength (parity vs. number) allow for the exploration of entanglement phases inaccessible to qubit-based platforms.

The authors modestly note that while numerical evidence supports the existence of the critical phase, analytical work is required to distinguish whether this is a true phase or an artifact of finite system sizes. They also highlight the potential connection between these transitions and the classical simulability of monitored beam-splitter dynamics (related to boson sampling) and fault-tolerance thresholds.