For a single qubit under continuous (diffusive) measurement with detection efficiency η<1\eta<1, Jiang, Wang, Martin and Whaley [Phys. Rev. A 102, 022612 (2020)] computed the globally optimal feedback protocol that maximizes the expected final Bloch-vector length, showing that feedback speeds purification for every 0<η<10<\eta<1, including the experimentally routine regime η<0.5\eta<0.5. Their paper leaves open whether this global-optimality structure extends beyond single-qubit purification: (a) no global-optimality analysis exists for purification of multi-qubit or higher-dimensional systems under finite measurement efficiency; (b) for measurement-based generation of a two-qubit Bell state, globally optimal protocols are proven only at unit efficiency (η=1\eta=1, pure initial states) for half- and full-parity measurements, while the known multi-qubit entanglement-generation protocols are only locally optimal (at perfect efficiency) and the noise-canceling constructions are idealized (unit efficiency, zero delay); and (c) the net advantage of feedback under realistic hardware non-idealities — finite feedback delay, extra decoherence, non-Markovian readout backaction — remains unquantified for any concrete architecture. A 2026 preprint line (Liu et al.) proves protocol-uniform ceilings on impurity-moment decay in every finite dimension and the exact frozen optimum for a qubit at 0<η<10<\eta<1, but explicitly leaves the higher-dimensional full-rank optimum and all finite-horizon optimality questions open, and does not address entanglement generation. The problem is therefore open: characterize the globally optimal measurement-based feedback law and its achievable performance as a function of η\eta for multi-qubit purification and two-qubit Bell-state generation, and establish whether the predicted advantage survives architecture imperfections.

Background

A quantum state is purified when continuous measurement collapses it toward a pure state. In the diffusive-measurement setting, a qubit is monitored continuously (e.g., in circuit QED, the architecture of superconducting qubits coupled to a microwave resonator, where homodyne detection — continuous measurement of one phase-sensitive quadrature of the outgoing microwave field — yields a noisy current record), and its conditional state evolves under a stochastic master equation (SME) — a Markovian evolution driven by the noisy measurement record. Measurement efficiency η\eta is the fraction of the information leaking from the system that is actually captured by the detector; the unobserved fraction (1−η1-\eta) causes decoherence without informing the observer, so η<1\eta<1 leaves the conditioned state mixed even in principle. The state of a qubit is described by its Bloch vector of length r≤1r\le 1; r=1r=1 is a pure state, so purification is the growth of rr, and impurity is 1−r21-r^2-type quantities. Feedback control means applying unitary rotations to the qubit in real time, with the rotation angle u(r,t)u(r,t) chosen as a function of the current conditional state (the measurement record). A feedback protocol is globally optimal if it maximizes the chosen performance functional over all admissible feedback laws, not merely over a parametrized family; local optimality only requires being instantaneously best.

For a single qubit the optimality theory is essentially complete. Jacobs (2003) showed that continuously rotating the measurement basis so the state stays unbiased relative to the measurement axis projects the qubit faster than fixed-axis measurement, and Wiseman and Ralph (2006) and Wiseman and Bouten (2008) proved that at perfect efficiency (η=1\eta=1) this unbiased-basis feedback is globally optimal for minimizing mean impurity, while analyses of that protocol suggested a speedup requires η>0.5\eta>0.5 — problematic because routine experimental efficiencies are around 0.4. Li, Shabani, Sarovar and Whaley (2013) proved, using the verification theorem of Hamilton–Jacobi–Bellman (HJB) dynamic programming (a standard stochastic-control tool: a candidate value function that satisfies the HJB inequality certifies global optimality of the associated policy), that for η≤1/2\eta\le 1/2 the globally optimal finite-horizon strategy for mean purity is no feedback at all — measure along a fixed, quantum-non-demolition (QND, i.e., not disturbing the measured observable) axis — leaving 1/2<η<11/2<\eta<1 unresolved. Jiang, Wang, Martin and Whaley (2020), the source of this problem, closed the single-qubit case: using backward-iteration dynamic programming with the Bloch-vector length (rather than purity) as the objective, they constructed the globally optimal feedback law u(r,t)u(r,t) for every 0<η<10<\eta<1 (a lookup table in which feedback is switched on only in a region near small rr and late times), proved via the stochastic Pontryagin maximum principle (an optimal-control theorem giving necessary conditions that any optimal control must satisfy) that the optimal control takes values in a discrete set, and showed by Monte Carlo simulation that the globally optimal protocol beats both no-feedback and always-on-feedback for all η∈(0,1)\eta\in(0,1) — so a speedup is available well below η=0.5\eta=0.5.

Beyond one qubit the picture is sparse. Combes, Wiseman and Jacobs (2008) showed feedback accelerates the measurement of qudits (systems with dimension higher than a qubit) with larger relative gains in higher dimension, and Stockton, van Handel and Mabuchi (2004) demonstrated deterministic preparation of Dicke states (symmetric multi-qubit states with a fixed number of excitations) by continuous measurement and feedback, but neither addresses global optimality under inefficient detection. For entanglement, Martin, Sayrafi and Whaley (2017) posed the question 'What is the optimal way to prepare a Bell state using measurement and feedback?' (a Bell state being a maximally entangled two-qubit state) and, for continuous half-parity XH=(σz1+σz2)/2X_H=(\sigma_{z1}+\sigma_{z2})/2 and full-parity XF=σz1σz2/2X_F=\sigma_{z1}\sigma_{z2}/2 measurements with local unitary feedback, proved globally optimal switching-type protocols at unit efficiency (η=1\eta=1) with pure initial states: for the half-parity measurement, global optimality for maximal concurrence, maximal fidelity, and minimal time to a target concurrence up to 1/21/\sqrt{2}; for the full-parity measurement, global optimality for maximal concurrence and fidelity, with measurement-only globally optimal for minimal time. Their Section V explicitly leaves open optimality at finite measurement efficiency, in the presence of loss, feedback delay, and finite qubit coherence, for other cost functions, and in the half-parity minimal-time regime above concurrence 1/21/\sqrt{2}. Zhang, Martin and Whaley (2020) subsequently gave locally optimal measurement-based feedback (PaQS, proportional-and-quantum-state-based feedback) for generating W, Dicke and GHZ entangled states (the W state being the single-excitation Dicke state; a GHZ state being a maximally entangled superposition such as (∣00⋯0⟩+∣11⋯1⟩)/2(|00\cdots 0\rangle+|11\cdots 1\rangle)/\sqrt{2}) of up to N∼100N\sim 100 qubits — but assuming perfect measurement efficiency and local (gradient) optimality only. Recent work has begun to establish fundamental limits rather than optimal protocols: Liu, Wang, Li and Ma (2026 preprints) prove a sharp, dimension-independent ceiling on the decay rate of trajectory-averaged impurity moments that holds uniformly over all admissible predictable feedback protocols in any finite dimension when η<1\eta<1, and show that for a qubit this 'frozen' rate is the exact optimum, set by rare, persistently mixed measurement records — while explicitly leaving the higher-dimensional full-rank optimum open; Bompais, Amini, Garrahan and Guță (2026) prove exponential purification rates for quantum trajectories without feedback optimization.

The source paper's own outlook (Section V) states the open program: quantify, for a concrete architecture, whether the feedback advantage survives imperfections such as feedback delay and non-Markovian circuit-QED readout backaction (non-Markovian meaning the readout retains memory, invalidating the standard SME); extend the global-optimality analysis to multi-qubit purification, where 'no work has addressed the effect of measurement inefficiency in the multi-qubit context'; and settle optimality for measurement-based generation of entanglement, where 'open problems regarding optimality remain even in the case of two qubit Bell state generation' — in context, the finite-efficiency and related optimality items that Martin et al. (2017) themselves declare open. The outlook poses these directions programmatically and does not fix a unique measurement configuration (measured observables, measurement strengths, initial states, admissible feedback unitaries) for the multi-qubit tasks; the cited lineage supplies canonical choices — qudit/collective measurements as in Combes et al. (2008), the half-/full-parity two-qubit model of Martin et al. (2017), and collective-measurement multi-qubit feedback as in Zhang et al. (2020) — so a solution must state explicitly which configuration it treats, and resolving one well-specified canonical configuration answers the corresponding instance.

Problem Statement

Overarching objective: determine whether the global-optimality theory of measurement-based feedback — complete for single-qubit purification, where for every detection efficiency 0<η<10<\eta<1 the feedback law maximizing the expected final Bloch-vector length is known and provably beats all fixed alternatives — extends to the basic state-preparation tasks of multi-qubit quantum control under the same inefficient diffusive monitoring. Concretely, in a model of two or more qubits under continuous measurement with efficiency 0<η<10<\eta<1 and real-time unitary feedback conditioned on the measurement record, address the following linked instances of this one optimality question:

(a) Multi-qubit purification. Determine the globally optimal feedback protocol and its achievable purification performance (e.g., the optimal expected purity, Bloch-sphere analog, or impurity-moment decay rate as a function of η\eta) for purifying a two-qubit or higher-dimensional system, either at finite horizon (the direct dynamic-programming generalization of the single-qubit lookup tables) or in the long-time/asymptotic regime where the current state of the art brackets the higher-dimensional optimum between a rank-two QND ceiling (set by QND measurements of observables with two-dimensional eigenspaces) and full-rank bounds without identifying it.

(b) Bell-state generation. For two qubits prepared in a Bell state by continuous measurement and feedback, determine the globally optimal protocol and the achievable fidelity or concurrence (or entanglement generation rate) as a function of η<1\eta<1, extending the canonical half-/full-parity measurement model with local unitary feedback of Martin et al. (2017) — whose globally optimal protocols are proven only at unit efficiency (η=1\eta=1) with pure initial states — and replacing the idealized noise-canceling constructions (which assume unit efficiency and zero delay).

(c) Net advantage under realistic non-idealities. For at least one concrete architecture (e.g., superconducting circuit QED), quantify whether the measurement-feedback advantage survives the dominant imperfections — finite feedback delay, feedback-induced extra decoherence, and non-Markovian readout backaction — so that the theoretical gain is a net gain in practice.

Deliverables (a)–(c) are instances of a single question — does the globally optimal measurement-based feedback law, and its η\eta-dependent performance, admit a characterization beyond the single qubit — with (c) assessing whether the feedback gain is net-positive on real hardware. Solving at least one instance resolves the question. For (a) or (b), a resolution must explicitly specify the measurement configuration analyzed — measured observable(s) (e.g., collective or qudit observables as in Combes et al. 2008, or the half- and full-parity observables XH=(σz1+σz2)/2X_H=(\sigma_{z1}+\sigma_{z2})/2 and XF=σz1σz2/2X_F=\sigma_{z1}\sigma_{z2}/2 of Martin et al. 2017), measurement strength(s), initial state, admissible feedback unitaries, and efficiency η∈(0,1)\eta\in(0,1) — and must provide either a proof of global optimality over all admissible feedback laws of that stated model (analytic, or numerical with a certification of optimality within the stated discretization and an error analysis covering interior regions of the control table, not only switching boundaries), or a proven fundamental bound for that model that is attained or shown unattainable; a locally optimal or heuristic protocol without a global-optimality guarantee does not resolve (a) or (b). For (c), a resolution is a quantitative net-advantage analysis within an explicitly stated architecture-specific measurement model covering the relevant imperfections (delay, feedback-induced decoherence, non-Markovian readout).

References

  1. Yuxiao Jiang, Xiyue Wang, Leigh Martin, K. Birgitta Whaley, "Optimality of feedback control for qubit purification under inefficient measurement", Phys. Rev. A 102, 022612 (2020). DOI: 10.1103/PhysRevA.102.022612; arXiv:1910.02487. https://doi.org/10.1103/PhysRevA.102.022612
  2. K. Jacobs, "How to project qubits faster using quantum feedback", Phys. Rev. A 67, 030301(R) (2003). DOI: 10.1103/PhysRevA.67.030301. https://doi.org/10.1103/PhysRevA.67.030301
  3. H. M. Wiseman, J. F. Ralph, "Reconsidering rapid qubit purification by feedback", New J. Phys. 8, 90 (2006). DOI: 10.1088/1367-2630/8/6/090. https://doi.org/10.1088/1367-2630/8/6/090
  4. Howard M. Wiseman, Luc Bouten, "Optimality of Feedback Control Strategies for Qubit Purification", Quantum Inf. Process. 7, 71 (2008). DOI: 10.1007/s11128-008-0075-8. https://doi.org/10.1007/s11128-008-0075-8
  5. H. Li, A. Shabani, M. Sarovar, K. B. Whaley, "Optimality of qubit purification protocols in the presence of imperfections", Phys. Rev. A 87, 032334 (2013). DOI: 10.1103/PhysRevA.87.032334. https://doi.org/10.1103/PhysRevA.87.032334
  6. J. Combes, H. M. Wiseman, K. Jacobs, "Rapid Measurement of Quantum Systems Using Feedback Control", Phys. Rev. Lett. 100, 160503 (2008). DOI: 10.1103/PhysRevLett.100.160503. https://doi.org/10.1103/PhysRevLett.100.160503
  7. J. K. Stockton, R. van Handel, H. Mabuchi, "Deterministic Dicke-state preparation with continuous measurement and control", Phys. Rev. A 70, 022106 (2004). DOI: 10.1103/PhysRevA.70.022106. https://doi.org/10.1103/PhysRevA.70.022106
  8. Leigh Martin, Mahrud Sayrafi, K. Birgitta Whaley, "What is the optimal way to prepare a Bell state using measurement and feedback?", Quantum Sci. Technol. 2, 044006 (2017). DOI: 10.1088/2058-9565/aa804c; arXiv:1704.00332. https://doi.org/10.1088/2058-9565/aa804c
  9. Song Zhang, Leigh Martin, K. Birgitta Whaley, "Locally optimal measurement-based quantum feedback with application to multiqubit entanglement generation", Phys. Rev. A 102, 062418 (2020). DOI: 10.1103/PhysRevA.102.062418; arXiv:1807.02029. https://doi.org/10.1103/PhysRevA.102.062418
  10. Jiaxin Liu, Zuoxian Wang, Feng Li, Danyue Ma, "Efficiency-Induced Freezing in Quantum-State Purification", arXiv:2607.07434 (2026, preprint). https://arxiv.org/abs/2607.07434
  11. Jiaxin Liu, Zuoxian Wang, Feng Li, Danyue Ma, "Record Loss Sets a Rare-Trajectory Limit on Quantum Purification", arXiv:2607.09118 (2026, preprint). https://arxiv.org/abs/2607.09118
  12. Maël Bompais, Nina H. Amini, Juan P. Garrahan, Mădălin Guță, "The rate of purification of quantum trajectories", arXiv:2601.14023 (2026, preprint). https://arxiv.org/abs/2601.14023
  13. Tathagata Karmakar, Philippe Lewalle, Yipei Zhang, K. Birgitta Whaley, "Noise-canceling quantum feedback: Non-Hermitian dynamics with applications to state preparation and magic state distillation", Phys. Rev. A 113, 042403 (2026). DOI: 10.1103/cnxj-m1wn; arXiv:2507.05611. https://doi.org/10.1103/cnxj-m1wn
  14. Dominic Shea, Alessandro Romito, "Stochastic action for the entanglement of a noisy monitored two-qubit system", Phys. Rev. A 111, 012442 (2025). DOI: 10.1103/PhysRevA.111.012442; arXiv:2403.08422. https://doi.org/10.1103/PhysRevA.111.012442
  15. Yanan Liu, Daoyi Dong, Sen Kuang, Ian R. Petersen, Hidehiro Yonezawa, "Two-step feedback preparation of entanglement for qubit systems with time delay", Automatica 122, 109174 (2021). DOI: 10.1016/j.automatica.2020.109174; arXiv:2307.14599. https://doi.org/10.1016/j.automatica.2020.109174