{"id":"d524cb3c-d558-4191-b500-cfbb7b91a856","arxiv_id":"2608.09639","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For one monitored jump operator commuting with the Hamiltonian, quantum trajectories converge to an eigenstate of L†L in the long-time limit if and only if the squared jump amplitudes |l_i|² are all distinct.","lead":"Watching a quantum system continuously can, in the long run, do the same job as a textbook measurement that collapses the system, but only if certain rates are all different. This paper proves when that happens and how long it takes, which helps experimentalists design reliable quantum readouts.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section III B proves only convergence in probability; the claimed almost-sure convergence of individual trajectories is not established.","rationale":"The reader's weakest_assumption concerns imperfect detection, a known scope limitation that the authors explicitly defer to future work. The more central issue is internal to the proof of the stated theorem: the derivation around Eqs. (30)-(32) controls the probability of crossing intervals at each fixed time t, which is convergence in probability, while the paper's language and the reader's strongest_claim describe individual trajectories converging almost surely. This distinction matters for the central claim because the conclusion 'implements a projective measurement' is about what happens to a single trajectory in the long-time limit, not just to the ensemble at each time. The gap is likely fixable: the bad probabilities decay exponentially, so a Borel-Cantelli argument with control of excursions between integer times should establish almost sure convergence. Because the fix is straightforward and the underlying physics and Poisson-mode separation are sound, the conditional verdict remains appropriate rather than a rejection. The factor-2 diffusive-rate discrepancy with Ref. [42] is probably a definitional difference and is not load-bearing for the convergence theorem.","tokens_in":28369,"tokens_out":41034,"duration_ms":388321,"concrete_test":"Check whether the bad-crossing probabilities are summable and excursions controlled: evaluate S = sum_{n=1}^∞ Pt_n(bad) with t_n=n using the Chernoff bounds in Appendix B, and bound the probability that, starting from a good count at time n, a Poisson number of jumps in [n,n+1] moves the count across a boundary. If S<∞ and the excursion probability decays exponentially in n, Borel-Cantelli upgrades the proof to almost sure convergence, validating the 'with unit probability' wording. If the excursion bound fails, the theorem should be restated as convergence in probability (error probability tends to zero), which is weaker than individual-trajectory convergence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central theorem rests on Eq. (32), which bounds the probability that the jump count at a fixed time t falls in a crossing interval. This establishes that, for each t, the conditional state converges in probability to an eigenstate as t→∞. It does not show that an individual trajectory eventually remains in one good interval for all sufficiently large times. The text in Section III B concludes that 'any realisation ... will, in the long-time limit, generate jump counts m that fall within one of the intervals ... with unit probability'; if 'with unit probability' is meant as almost sure convergence (as in the reader's strongest_claim and Ref. [42]), the proof is missing a Borel-Cantelli step or an equivalent uniform-in-time control. The crossing probabilities from the Chernoff bounds decay exponentially, so an almost-sure proof is likely available, but it is not in the paper. Because the abstract and conclusion assert convergence of individual trajectories, this gap is load-bearing: the theorem as written is stronger than the proof supports. The reader's perfect-detection concern, by contrast, applies outside the stated model and is acknowledged in Section VI.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies continuous monitoring described by a single jump operator L on a finite-dimensional Hilbert space, with [H,L]=0. It shows that the conditional quantum-jump state after time t and m detected clicks depends only on the total count m, and that the distribution of m is a weighted sum of Poisson modes with means γt|l_i|^2. The authors identify a non-degeneracy condition on the eigenvalues |l_i|^2 of L†L, prove through Chernoff tail bounds that the probability of obtaining a jump count in the \"crossing\" intervals vanishes as t→∞, and conclude that individual trajectories converge to eigenstates of L†L, thereby asymptotically implementing a projective measurement. They derive convergence rates and finite-time error relations, extend the analysis to degenerate eigenvalues, dark states, and the fixed-m first-passage-time protocol, and discuss the diffusive limit through a displacement of the jump operator. The paper also connects the non-degeneracy condition to the absence of similar symmetry subspaces of the strong symmetry L†L.","tokens_in":28483,"tokens_out":13785,"duration_ms":132001,"significance":"If fully established, the central result is significant and practically useful: it gives a simple, falsifiable criterion for when continuous monitoring can be trusted to implement a projective measurement, together with explicit convergence times and confidence estimates. The derivation is largely self-contained, starting from the measurement operators rather than assuming the Poisson-mixture structure, and the Chernoff-bound analysis is detailed and reproducible. The treatment of degeneracy, dark states, and the fixed-m protocol adds genuine breadth. The connection to strong symmetries and dissipative freezing is insightful. The main caveats are that the proof as written establishes convergence in probability rather than the claimed almost-sure convergence of individual trajectories, and that an unexplained factor-2 discrepancy with the diffusive rate of Ref. [42] remains unresolved in Section IV D.","major_comments":[{"comment":"The central convergence claim is stated as almost-sure behavior of individual trajectories, but the proof establishes only convergence in probability at each fixed time. Equation (32) controls P_t(m outside the good intervals) for each t via lim_{ε→0} lim_{t→∞} Σ_i P_t(˜m^+_i ≤ m ≤ ˜m^-_{i+1}) = 0. This does not imply P(eventually all sufficiently large times have m in a good interval) = 1; a Borel-Cantelli or law-of-large-numbers argument is needed. Such an argument is available because N_t is a finite mixture of Poisson processes (N_t/t → γ|l_i|^2 almost surely for the realized mode), but it is not present in the manuscript. The abstract, Section III B, and the conclusion should either include this step or be weakened to convergence in probability.","section":"Section III B, Eq. (32)"},{"comment":"The paper states that its diffusive convergence rate is 2 times larger than the rate found in Ref. [42], but leaves the discrepancy unexplained. If the two rates are defined in the same way, one of the calculations must be wrong; if they are defined differently, the relation between the definitions must be spelled out. Because the paper claims to recover and unify the diffusive non-degeneracy condition and convergence rates, this unresolved factor undermines the diffusive extension. The authors should either derive the diffusive rate directly, compare with Ref. [42] under an explicitly matched convention, or correct the calculation in Appendix C.","section":"Section IV D, Eq. (50)"},{"comment":"The diffusive limit is obtained by first computing convergence rates for finite α and then taking |α|→∞. A true diffusive unravelling is a different stochastic process, and the interchange of the limits t→∞ and α→∞ is not automatic. The factor-2 discrepancy with Ref. [42] may be a symptom of this non-commutation. Please justify the order of limits, or derive the diffusive rates directly from the diffusive stochastic master equation.","section":"Section IV D and Appendix C"}],"minor_comments":[{"comment":"Equation (40b) appears to have a typographical error: the left-hand side is written as P_t(1|m>0), but the surrounding text and the meaning of the equation indicate it should be P_t(i|m>0) for i≠1, namely that after at least one jump the dark state is inaccessible.","section":"Section IV A, Eq. (40b)"},{"comment":"The expression for t_i(ε) in Eq. (B6) is missing a factor of γ. The derivation from Eq. (B2) yields γ t_i(ε) ≥ max{...}, matching Eq. (33) in the main text; as written, the dimensions in Eq. (B6) are inconsistent.","section":"Appendix B, Eq. (B6)"},{"comment":"The quantity |K^α_{ij,i}| in Eq. (50) is not defined in the text. From the context and Appendix C it should be |R^α_{i,i±1}|; please introduce it explicitly or replace it with the rate notation used elsewhere.","section":"Equation (50)"},{"comment":"The proof assumes p_i∈(0,1) for all i, but this restriction is not stated in the non-degeneracy condition in Section III A 1. Please state this assumption where the condition is formulated, or discuss how the argument changes if some p_i vanish.","section":"Section III B"},{"comment":"In Fig. 1(d), the lower panel is referenced in terms of ˜m^±_i/γt, but the horizontal axis is not labelled. Adding an explicit label would help readers connect the figure to Eqs. (27)-(29).","section":"Figure 1"}],"recommendation":"major_revision","confidential_remarks":"The almost-sure convergence gap is the main technical issue; it is fixable by adding a short SLLN/Borel-Cantelli argument for finite Poisson mixtures. The factor-2 discrepancy with Ref. [42] also needs a clear resolution before the paper can be accepted. The core quantum-jump result appears sound and well motivated; I would be willing to support acceptance after these points are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth your time if you care about when continuous monitoring really collapses a state. The fixed-time convergence theorem is a re-derivation of Benoist–Pellegrini, and the paper says so plainly. What is actually new is the full-counting-statistics proof, the fixed-m first-passage-time convergence theorem, and the clean treatment of dark states, degenerate eigenspaces, and the diffusive limit as a large-displacement limit of the jump process. The appendices are serious: Chernoff bounds are deployed carefully, and the Poisson-mixture logic is transparent. Credit where it is due.\n\nThe main soft spot is the gap between what is proved and what is claimed. Section III B shows that for each fixed t, the probability of landing in a crossing interval goes to zero. That is convergence in probability. The text and abstract go further, saying individual trajectories converge with unit probability. To get almost-sure convergence you need a Borel–Cantelli step or uniform-in-time control. The exponential decay of the Chernoff bounds suggests the fix is straightforward, but it is not in the paper, and the current wording overstates the theorem. This is addressable, not destructive.\n\nThe factor-2 discrepancy with the diffusive convergence rate in Ref. [42] (Section IV D) is also a real loose end. The authors flag it as curious but do not resolve it. It could be a definitional difference or a small algebraic error, but a reader cannot tell which. That needs to be sorted before publication.\n\nTwo smaller things. The claim in Section III C of good agreement with numerical simulations is not backed by any plot or quantitative comparison; either add the data or soften the claim. And the perfect-detection assumption is acknowledged as an open problem in Section VI, which is fine, but it should be stated as a model assumption in Section II rather than left implicit.\n\nThe paper is honest about what it re-derives, the new material is real, and the mathematical core is sound up to the a.s. issue. This deserves a serious referee: send it to peer review, with instructions to require a Borel–Cantelli argument (or a clear statement that only convergence in probability is claimed) and an explanation of the factor-2 rate gap.","headline":"Solid counting-statistics re-derivation with a genuinely new fixed-m first-passage theorem; the almost-sure convergence claim is stronger than the proof, and the factor-2 diffusive-rate discrepancy needs resolving.","tokens_in":29063,"tokens_out":2061,"would_cite":true,"duration_ms":21691,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P15","81S22"],"pacs":["03.65.Ta","42.50.Lc"],"model":"deepseek-v4-flash","headline":"Continuous monitoring of a single jump operator asymptotically realizes an ideal projective measurement of $L^{\\dagger}L$ if and only if the squared jump eigenvalues are pairwise distinct.","keywords":["projective measurements","continuous monitoring","quantum trajectories","quantum jumps","full counting statistics","first-passage times","non-demolition measurements","strong symmetries"],"falsifier":"Take a three-level system with $[H,L]=0$ and $|l_2|^2=|l_3|^2$, start in a superposition of $|2\\rangle$ and $|3\\rangle$, and track single trajectories after long monitoring: the paper predicts the conditional state remains a mixture over $\\{|2\\rangle,|3\\rangle\\}$ with probabilities fixed by the initial state, never collapsing to one of them; seeing collapse to a definite $|2\\rangle$ or $|3\\rangle$ would falsify the non-degeneracy condition.","tokens_in":28114,"feed_emoji":"⚛️","tokens_out":7586,"duration_ms":65441,"temperature":0.7,"pith_summary":"The paper asks when a weak, continuous measurement that records discrete detector clicks can stand in for an ideal projective measurement. It answers: for a finite-dimensional system with one diagonalisable jump operator $L$ commuting with the Hamiltonian, individual quantum-jump trajectories converge to a single eigenstate of $L^{\\dagger}L$ in the long-time limit exactly when the squared amplitudes $|l_i|^2$ of the jump operator are all distinct. When some squared amplitudes coincide, trajectories converge only to the corresponding degenerate eigenspace, so a full projective measurement is impossible. If correct, this gives experimentalists a transparent criterion for when a monitored record can be trusted to have collapsed the state, with explicit rates and confidence times.","feed_headline":"Distinct jump rates make continuous monitoring projective","feed_subtitle":"With distinct squared jump eigenvalues, every trajectory converges to one eigenstate, recovering Born and Lüders rules.","key_machinery":"The load-bearing identity is the full counting statistics of a single jump operator: $P_t(m)=\\sum_i p_i\\,\\mathrm{Pois}(\\gamma t|l_i|^2;m)$. Because $[H,L]=0$, the conditional state depends only on the total number of clicks $m$, not their times, so the unnormalised trajectory factorises as $S(t)J^m\\rho(0)$. The proof then studies the conditional distribution $P_t(i|m)$, writes it as $1/(1+\\sum_{j\\neq i}\\exp[x_{ij}])$ with $x_{ij}=\\gamma t(|l_i|^2-|l_j|^2)-m\\ln(|l_i|^2/|l_j|^2)-\\ln(p_i/p_j)$, and shows via exponential tail-probability bounds that the windows of $m$ where these probabilities cross have vanishing cumulative probability as $t\\to\\infty$. The first-passage-time distribution, a mixture of Erlang distributions, carries the analogous argument for the fixed-$m$ protocol.","core_discovery":"The central claim is that the counting statistics of the detected jumps are sufficient to implement an ideal projective measurement. Under the assumptions $[H,L]=0$ and $L$ diagonalisable with eigenstates $|i\\rangle$, the probability of $m$ clicks after time $t$ is a mixture of Poisson distributions, $P_t(m)=\\sum_i p_i\\,\\mathrm{Pois}(\\gamma t|l_i|^2;m)$, one mode per eigenvalue of $L^{\\dagger}L$. The conditional probability that the state is $|i\\rangle$ given $m$ clicks has a logistic form, and the paper proves that in the long-time limit the jump count $m$ falls, with probability one, into a window where this conditional probability exceeds $1-\\varepsilon$ for a single eigenvalue. Hence each trajectory collapses to a single eigenstate selected with the Born probability $p_i$, and the post-measurement state follows L\\\"uders' rule; if eigenvalues coincide, the collapse is only to the degenerate eigenspace. The same conclusion holds when conditioning on a fixed jump count $m$, using first-passage-time statistics, and the diffusive (homodyne) limit emerges from shifting $L$ by a large amplitude $\\alpha$, which lifts degeneracies.","pith_inferences":["Beyond the paper: if the same reasoning extends to several mutually commuting diagonalisable jump operators each satisfying the non-degeneracy condition, the converged eigenstate would be selected by the joint counting statistics; this is a natural next step the authors flag but do not carry out.","Beyond the paper: the Poisson-mode picture predicts a direct experimental test in a qutrit with two near-degenerate $|l_i|^2$ values: the convergence time diverges as the gap shrinks, so monitoring for a fixed duration should show a sharp drop in assignment fidelity as degeneracy is approached.","Beyond the paper: real detectors have finite efficiency, which thins the click record; the paper's rate formulas would acquire an efficiency-dependent rescaling of $\\gamma t$, and the proof's factorisation $S(t)J^m$ would need modification, so robustness under imperfect detection is a concrete open extension."],"forward_implications":["For a qubit measured via quantum jumps with $L=\\sigma_z$, no projective measurement is possible in the jump picture because $L^{\\dagger}L=I$ is fully degenerate; adding a local-oscillator shift $\\alpha$ makes the eigenvalues $|\\alpha\\pm1|^2$ distinct and restores convergence, so homodyne detection of $\\sigma_z$ implements the projection while photon counting alone does not.","If some squared eigenvalues coincide, the non-degeneracy condition fails and individual trajectories converge only to the eigenspace of $L^{\\dagger}L$ belonging to that eigenvalue; the conditional probability of any single state inside the degenerate manifold stays strictly below one.","The convergence rate for state $|i\\rangle$ is set by the two neighbouring eigenvalues of $L^{\\dagger}L$, not by the full spectrum, and explicit formulas give the measurement time needed for a chosen error $\\varepsilon$ and confidence in standard deviations.","Dark states (eigenvalues $|l_i|^2=0$) act as switches: a single click excludes the dark state forever, while a long click-free record projects the state into it with the initial dark-state probability.","Conditioning on a fixed jump count $m$ instead of a fixed time $t$ yields the same convergence, with the rate given by an analogous formula in terms of Erlang distributions, so experiments can choose the protocol that reaches a given accuracy faster."],"supporting_citations":[{"why":"Supplies the full-counting-statistics and first-passage-time identities from which the Poisson and Erlang mode decompositions are derived.","marker":"[37]"},{"why":"Introduces the quantum-jump trajectory treatment of non-destructive quadrature measurement that this paper generalises to arbitrary discrete observables.","marker":"[38]"},{"why":"Establishes earlier convergence conditions and rates for diagonal jump operators in finite dimension, which the paper's counting-statistics proof recovers and refines.","marker":"[42]"},{"why":"Identifies strong symmetries and similar symmetry subspaces as governing dissipative freezing, which the paper reinterprets in terms of the observable $L^{\\dagger}L$.","marker":"[45]"},{"why":"Provides the exponential tail-probability bounds used to show that the jump-count intervals where conditional probabilities cross have vanishing probability.","marker":"[49]"},{"why":"Provides the analogous tail-probability bounds for the Erlang tails in the fixed-$m$ first-passage-time proof.","marker":"[52]"}],"fun_headline_variants":["Jump count statistics alone yield projective measurements","Nondegenerate jump rates force trajectory collapse","Continuous monitoring becomes projective with distinct jump rates","Projective measurement emerges from jump count statistics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire argument rests on the idealisation that the click record is complete: every jump of the single monitored channel is detected and there are no unmonitored loss channels, so the conditional state can be written as $S(t)J^m\\rho(0)$ and the click count is a pure Poisson mixture.","fun_headline_variants_meta":{"raw":{"variants":["Jump count statistics alone yield projective measurements","Nondegenerate jump rates force trajectory collapse","Continuous monitoring becomes projective with distinct jump rates","Projective measurement emerges from jump count statistics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000519,"raw_usage":{"total_tokens":2508,"prompt_tokens":932,"completion_tokens":1576,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":548,"completion_tokens_details":{"reasoning_tokens":1522}},"tokens_in":548,"tokens_out":1576,"duration_ms":11709,"temperature":1.0,"reasoning_tokens":1522,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T13:37:02.822851+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a three-level system with $[H,L]=0$ and $|l_2|^2=|l_3|^2$, start in a superposition of $|2\\rangle$ and $|3\\rangle$, and track single trajectories after long monitoring: the paper predicts the conditional state remains a mixture over $\\{|2\\rangle,|3\\rangle\\}$ with probabilities fixed by the initial state, never collapsing to one of them; seeing collapse to a definite $|2\\rangle$ or $|3\\rangle$ would falsify the non-degeneracy condition.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes earlier convergence conditions and rates for diagonal jump operators in finite dimension, which the paper's counting-statistics proof recovers and refines."},{"cited_title":"Bauer, D","cited_arxiv_id":null,"evidence_quote":"Identifies strong symmetries and similar symmetry subspaces as governing dissipative freezing, which the paper reinterprets in terms of the observable $L^{\\dagger}L$."},{"cited_title":"K¨ ummerer and H","cited_arxiv_id":null,"evidence_quote":"Provides the exponential tail-probability bounds used to show that the jump-count intervals where conditional probabilities cross have vanishing probability."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the analogous tail-probability bounds for the Erlang tails in the fixed-$m$ first-passage-time proof."}],"review_version":1}