Pith. sign in

REVIEW 3 major objections 5 minor 6 cited by

Under a no-dark-subspace condition, repeated measurements purify quantum states exponentially fast — at an explicit, computable rate that also governs how quickly a filter learns the state.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 09:23 UTC pith:U32UQ5OU

load-bearing objection Main result unproven as written due to a gap in the uniform-contraction argument, but the Lyapunov proof of K-M is solid and the rate formula is a promising new idea. the 3 major comments →

arxiv 2601.14023 v2 pith:U32UQ5OU submitted 2026-01-20 quant-ph math-phmath.MP

The rate of purification of quantum trajectories

Ma\"el Bompais , Nina H. Amini , Juan P. Garrahan , M\u{a}d\u{a}lin Gu\c{t}\u{a} This is my paper
classification quant-ph math-phmath.MP MSC 81P1560J05
keywords quantum trajectoriespurificationexponential convergenceLyapunov functiondark subspacesquantum state estimationfidelityKraus operators
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper asks how fast a quantum system under repeated indirect measurements becomes pure. Earlier work established that, when no 'dark subspace' traps the dynamics, the trajectory almost surely purifies in the long run. The paper strengthens that qualitative picture in two ways. First, using the Lyapunov function V(ρ) = sqrt(1 − tr ρ²), it proves that the purity gap decays exponentially in expectation, E[V(ρ_n)] ≤ V(ρ_0) e^{−γ⌊n/p̄⌋}, with an explicit rate γ computed from the Kraus operators and a finite block length p̄ ≤ d² over which non-darkness is detectable. Second, it shows the same exponential rate governs state estimation: a trajectory started from an arbitrary initial guess converges to the true trajectory, so a filter forgets its initialization error at the same speed. These rates are not just qualitative milestones — they bound how fast measurement information flows out of the system, which sets limits on quantum filtering, feedback control, and real-time monitoring.

Core claim

Under the no-dark-subspace condition (Pur) — no two-dimensional subspace stays isometric under every word of Kraus operators — a measured quantum trajectory purifies exponentially fast in expectation. With V(ρ) = sqrt(1 − tr ρ²), Theorem 2 gives E[V(ρ_n)] ≤ V(ρ_0) e^{−γ⌊n/p̄⌋}, with p̄ ≤ d² and γ > 0 explicit: the worst-case contraction of two-dimensional areas under all words of length p̄. Theorem 3 extends this to estimation: if the true initial state lies inside the initial guess's support, E[1 − F(ρ_n, ρ̂_n)] ≤ C e^{−γ⌊n/p̄⌋} with C = ‖ρ̂_0^{−1/2} ρ_0 ρ̂_0^{−1/2}‖_∞ V(ρ̂_0), so a filter starting from any compatible guess forgets its error at the same exponential rate. The paper also repr

What carries the argument

The engine is the Lyapunov function V(ρ) = sqrt(1 − tr ρ²) and its one-block contraction ratio λ(ρ) = E[V(ρ_{n+p̄}) | ρ_n = ρ] / V(ρ). The core step is proving sup λ(ρ) < 1 over all mixed states — a uniform gap — by contradiction: a sequence of near-pure states with λ → 1 would, via a limiting-weight computation, force a two-dimensional subspace with πM_I π ∝ π for every word I of length p̄, a dark subspace, contradicting (Pur). The block length p̄ ≤ d² comes from Proposition 2: the subspaces E_p spanned by V†_I V_I stabilize in at most d² steps, so darkness is detectable by words of length at most d². The rate γ is the supremum over eigenbases and weights of a sum of two-dimensional determi

Load-bearing premise

The exponential bound stands on one gap claim — that over a block of p̄ steps the expected drop in V stays uniformly bounded away from zero for every mixed state — and the proof of that gap needs the finite detection length p̄ ≤ d² (cited to a private communication, not proved elsewhere), a limiting-weight computation that as printed is off by a factor of two, and continuity of λ at degenerate near-pure spectra, which is asserted but not demonstrated.

What would settle it

Compute the closed-form qubit rate γ = −ln(Σ_{I∈O^{p̄}} det(M_I)) for a two-dimensional channel satisfying (Pur) with non-rank-one Kraus operators; a channel with γ = 0 would falsify the theorem. Less specifically, a numerical sweep maximizing λ(ρ) over states approaching a pure boundary that finds a supremum of 1 would show the uniform contraction gap, and with it the exponential bound, fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Purification is not just almost sure but exponential in expectation: E[V(ρ_n)] ≤ V(ρ_0) e^{−γ⌊n/p̄⌋}, so uncertainty about a measured quantum system is erased at a definite, computable rate.
  • State estimation inherits the rate: the estimated trajectory converges to the true one with E[1 − F(ρ_n, ρ̂_n)] ≤ C e^{−γ⌊n/p̄⌋}, quantifying how quickly the measurement record pins down the unknown state.
  • The rate is computable data: γ is built from determinants of the Kraus operators over blocks of length p̄ ≤ d², and longer blocks give better bounds because γ_{p+q} ≥ γ_p + γ_q.
  • The no-dark-subspace assumption is necessary: on a dark subspace purity and fidelity are exactly conserved, so neither purification nor estimation can proceed.
  • For rank-one Kraus operators the rate is infinite: the state purifies in a single step, consistent with projective-type measurements.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the explicit γ formula turns the purification rate into a numerically checkable quantity — one could certify a measurement setup's information rate by maximizing λ(ρ) near the pure boundary, as the paper's spin-chain simulation does at small scale.
  • Editorial extension: since γ_p is super-additive, the asymptotic rate lim γ_p/p exists; comparing it with observed decay in larger systems could link this per-measurement rate to the scaling behavior studied in measurement-induced phase transitions.
  • Editorial extension: the appearance of 2×2 determinants suggests γ is the top Lyapunov exponent of the channel's induced action on the exterior square; an ergodic-theoretic formulation might carry the result over to random or time-dependent measurement protocols.
  • Editorial extension: in infinite dimensions the paper notes that non-purification no longer coincides with dark subspaces, so the exponential claim likely needs a different hypothesis there — a uniform-contraction or bounded-information condition in place of (Pur).

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper studies discrete-time quantum trajectories on a finite-dimensional Hilbert space, generated by a finite family of Kraus operators. Under the no-dark-subspace condition (Pur), the authors first give an alternative Lyapunov proof of the almost-sure purification theorem of Kümmerer and Maassen, and then claim two quantitative upgrades: (i) exponential purification in expectation with an explicit variational rate γ defined from the Kraus operators (Theorem 2), and (ii) exponential convergence of an estimated trajectory to the true trajectory under the same measurement record (Theorem 3), at the same rate γ. The argument for Theorem 2 proceeds by proving a uniform contraction gap for the Lyapunov function V(ρ)=√(1−tr ρ²) over blocks of length p̄, using a compactness/contradiction argument. Theorem 3 is derived as a corollary of Theorem 2. A numerical study on a four-qubit Ising chain with boundary measurements illustrates the claimed decay.

Significance. If Theorem 2 is correct, the paper provides a meaningful quantitative strengthening of a classical result, with an explicit, computable, parameter-free rate, and it transfers that rate to quantum filter stability. The Lyapunov supermartingale (Prop. 3) and the alternative proof of almost-sure purification (Prop. 4) are elegant and appear sound. The rate formula is also potentially useful for numerics and for comparing measurement schemes. However, the proof of the main quantitative theorem currently contains a gap in the central contradiction argument and a normalization error, and the filter-stability result inherits these problems. Thus the paper's principal new contribution is not established as written, although the almost-sure purification part and the overall strategy are valuable.

major comments (3)
  1. [§3.1, Eqs. (54)–(56)] The limiting-weight computation has a factor-of-2 error. With w^{(n)}_{kl}=2p^{(n)}_k p^{(n)}_l/(1−Σ_j(p^{(n)}_j)^2), the printed limit in (54) should be w^{(n)}_{k1}→2p'_k, where p'_k is defined in the text as lim p^{(n)}_k/(1−Σ_j(p^{(n)}_j)^2). Under that printed definition one has Σ_k p'_k=1/2, so Σ_I \bar M_I = 1/2, and the Cauchy–Schwarz step in (56) gives λ^* ≤ 2^{-1/2}, not λ^* ≤ 1. This normalization error propagates into Eqs. (57)–(58) and must be corrected before the subsequent argument can be assessed.
  2. [§3.1, Eqs. (61)–(63)] The step from the scalar relation M_11^I = Tr(B_I σ) to the existence of a single normalized vector |ψ⟩∈H_1 with ⟨ψ|B_I|ψ⟩=M_11^I for all I is invalid. The connectedness/continuity argument only gives, for each word I separately, a vector ψ_I satisfying that equality; it does not ensure a common ψ for all I. Without a common ψ, the definition S=span(|ψ_1⟩,|ψ⟩) and the assertion πM_Iπ=M_11^Iπ for all I do not follow. Since this dark-subspace construction is the only contradiction to (Pur), the uniform contraction gap sup_{ρ∈S_mix} λ(ρ)<1 is not established. No additional semigroup structure or convexity argument is supplied to rule out empty intersections of the level sets.
  3. [§3.1, Eq. (47)] The compactness argument relies on continuity of λ(ρ) on S_mix, asserted without proof. The conditional-expectation definition of λ is basis-independent by construction, but the spectral-decomposition formula (50) is written in terms of an eigenbasis ψ, which is not unique for degenerate states. If λ is discontinuous at a mixed accumulation point, the conclusion λ(ρ^*)=1 from λ(ρ^{(n)})→1 does not follow. A proof of the required continuity (or a replacement compactness argument avoiding it) is needed.
minor comments (5)
  1. [§2.2, Eq. (38)] The symbol p is overloaded: it denotes both the probability vector in the definition of W and the block length. This makes (38) and §3.2 confusing; use e.g. q or μ for the probability vector.
  2. [§3.1, Eq. (56)] The notation M_I is reused for the full word operator V_I†V_I and for the weighted sum Σ_{k≠1} p'_k M_kk^I. Define the latter with a different symbol (e.g. \bar M_I) consistently.
  3. [§3.2] The statement 'γ_p=0 for 1≤p<p̄' is asserted without proof. A short justification from the minimality of p̄ would improve readability and rigor.
  4. [§5] The comparison between γ_2 and the empirical rate is only qualitative. Please specify how γ_emp is extracted (fit window, number of trajectories, error bars) so the reader can judge the agreement.
  5. [References] Reference [18] is listed as a private communication. Since Prop. 2 is load-bearing and a proof is included in the text, this is acceptable, but a citable published source would be preferable if one exists.

Circularity Check

0 steps flagged

No significant circularity: the purification rate is computed from the Kraus operators and the stability bound is a corollary, not an imported conclusion.

full rationale

The derivation chain is self-contained. Theorem 2's rate γ is defined directly from the Kraus operators and the finite block length p̄ in Eq. (40), and the proof derives the contraction factor λ(ρ) from the definition E[V(ρ_{n+p̄})|ρ_n=ρ]/V(ρ) in Eqs. (45)–(51); no parameter is fitted to purification data. The uniform contraction gap is argued from (Pur), compactness, and Proposition 2, which is proved in the text; the citation to private communication [18] signals missing external support but is not a circular input to the main derivation. Theorem 3 is proved as a corollary of Theorem 2 via the decomposition ρ̂0 = μρ0 + (1−μ)ρ0^c and linearity of the relevant expectation; it uses the same rate rather than assuming filter stability. Self-citations [13,15,16] and [17] provide context and prior stability statements but are not load-bearing for the new deductions. Section 5 compares a numerically evaluated theoretical rate with an empirical decay rate; it does not fit the former to the latter. The Section 6 caveat about infinite dimensions is a limitation, not circularity. No step reduces a prediction to its input by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 0 invented entities

The central claim uses no fitted numbers: the rate gamma in Theorem 2 is a variational supremum over determinants of Kraus words, and the weight set W merely reparameterizes the eigen-spectrum of the state. The example (Sec. 5) selects J, tau, B_x, B_z by hand and fits an empirical rate gamma_emp to simulated trajectories, but these enter only the illustration, not the theorems. The proof does contain as-printed slips (limiting weights p'_k off by a factor 2; qubit rate missing a square root) and an unsupported sub-claim (Sec. 3.2, gamma_p = 0 for p < pbar); none appears to invalidate the main theorem.

axioms (5)
  • domain assumption Assumption (Pur): for every projection pi with rank >= 2 there is a measurement word (i_1,...,i_p) such that pi Vdag_{i_p}...Vdag_{i_1} V_{i_1}...V_{i_p} pi is not proportional to pi
    The central structural hypothesis, inherited from Kuemmerer-Maassen; all quantitative results (Theorems 2 and 3) are conditional on it. Section 2.2, Eq. (8).
  • domain assumption Finite-dimensional discrete-time setting: dim H = d < infinity, finite outcome set O, Kraus decomposition with completeness relation
    Compactness of S, the bound pbar <= d^2 (Prop. 2), and the Kushner omega-limit theorem (A.1) all rely on finite dimension; the paper leaves the infinite-dimensional extension open (Sec. 6, Ref. [21]). Section 2.1, Eqs. (1)-(5).
  • standard math Kushner-type omega-limit theorem (Theorem A.1): for a Markov chain on a compact state space with continuous nonpositive drift Q, accumulation points are almost surely in {Q = 0}
    Adapted from [24, Thm 1, Ch. 8] and [25, Thm A.1]; used to localize the accumulation set of (rho_n) in Proposition 4. Appendix.
  • domain assumption Support inclusion rho_0 << rhoh_0 for the filter-stability theorem
    Theorem 3 requires the true initial state to be absolutely continuous w.r.t. the initial guess so that realized measurement records never have zero probability under the estimate (a full-rank guess suffices); the proof constant mu = 1 / ||rhoh_0^{-1/2} rho_0 rhoh_0^{-1/2}||_infty is positive only under this condition. Section 4, Eqs. (75)-(77).
  • standard math Standard matrix inequalities: Cauchy-Schwarz, strict convexity of the Frobenius norm, polar decomposition, fidelity inequality F >= tr(rho sigma), Fekete's lemma
    Used throughout: Prop. 3 (Cauchy-Schwarz/triangle inequality), Prop. 1 (polar decomposition), Theorem 3 (F >= tr(rho sigma)), Cor. 1 (Fekete).

pith-pipeline@v1.3.0-alltime-deepseek · 14186 in / 45537 out tokens · 422595 ms · 2026-08-03T09:23:30.216624+00:00 · methodology

0 comments
read the original abstract

We investigate the behavior of quantum trajectories conditioned on measurement outcomes. Under a condition related to the absence of so-called dark subspaces, K\"{u}mmerer and Maassen had shown that such trajectories almost surely purify in the long run. In this article, we first present a simple alternative proof of this result using Lyapunov methods. We then strengthen the conclusion by proving that purification actually occurs at an exponential rate in expectation, again using a Lyapunov approach. Furthermore, we address the quantum state estimation problem by propagating two trajectories under the same measurement record--one from the true initial state and the other from an arbitrary initial guess--and show that the estimated trajectory converges exponentially fast to the true one, thus quantifying the rate at which information is progressively revealed through the measurement process.

Figures

Figures reproduced from arXiv: 2601.14023 by Juan P. Garrahan, Ma\"el Bompais, M\u{a}d\u{a}lin Gu\c{t}\u{a}, Nina H. Amini.

Figure 1
Figure 1. Figure 1: (a) Geometric action of a Kraus operator [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Spin chain subject to a uniform magnetic field [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) 300 trajectories of Ve(ρn) (blue) and the corresponding exponential decrease of the empirical average (thick red). (b) Logarithmic-scale plot of 300 trajectories of Ve(ρn)(blue) together with the empirical average (thick red) showing the expected linear behavior. Theoretical upper bounds with rates γp defined in Equation (66) are plotted in blue dashed line for different values of p = 1, . . . , 5. (a)… view at source ↗
Figure 4
Figure 4. Figure 4: (a) The rate γ2/(2τ ) as a function of J and τ , for Bx = 1 and Bz = 1. (b) The empirical rate γemp/τ as a function of J and τ , for Bx = 1 and Bz = 1. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: (a) The rate γ2/2 as a function of (Bx, Bz), for J = 1 and τ = 1. (b) The rate γemp as a function of (Bx, Bz), for J = 1 and τ = 1. 6 Conclusions In this paper we have shown that, in the absence of dark subspaces, quantum trajectories purify at an exponential rate in expectation. This result allowed us to establish exponential stability of the quantum filter, showing that any estimated trajectory converges… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Record Loss Sets a Rare-Trajectory Limit on Quantum Purification

    quant-ph 2026-07 accept novelty 8.0

    For a qubit under lossy continuous monitoring, every impurity moment of order at least 1/2 decays at the same optimized rate 2ηℓ², set by the distinguishability of two ambiguous measurement-record laws.

  2. Efficiency-Induced Freezing in Quantum-State Purification

    quant-ph 2026-07 accept novelty 8.0

    Non-Hermitian dynamics amplify sensor response but not quantum Fisher information, which is bounded by the spectral width of the Hamiltonian generator alone.

  3. Efficiency-Induced Freezing in Quantum-State Purification

    quant-ph 2026-07 accept novelty 8.0

    For inefficient detectors (0<η<1), the fastest possible decay rate of impurity moments with order θ≥1/2 freezes at 2ηℓ², independent of moment order and dimension; for qubits this is the exact optimum.

  4. Generating quantum ensembles via reverse-time quantum diffusions

    quant-ph 2026-06 unverdicted novelty 8.0

    The paper establishes a reverse-time quantum diffusion framework that generates complex quantum ensembles from simple distributions by deriving and learning a feedback Hamiltonian from forward trajectory data.

  5. Universal purification dynamics of monitored Clifford circuits

    quant-ph 2026-07 accept novelty 7.5

    Purification of weakly monitored Clifford circuits on prime-dimensional qudits reduces exactly to a pure-death Markov process on the density-matrix rank, producing compact universal scaling functions for all Rényi entropies.

  6. Record Loss Sets a Rare-Trajectory Limit on Quantum Purification

    quant-ph 2026-07 accept novelty 7.0

    Protocol-independent purification speed limits in all finite dimensions are set by sqrt(impurity) with matched rates 2:4:8 ηℓ², freezing at the half-moment and attained by QND measurement.

Reference graph

Works this paper leans on

25 extracted references · 2 linked inside Pith · cited by 4 Pith papers

  1. [1]

    H. M. Wiseman and G. J. Milburn,Quantum measurement and control(2009)

    2009

  2. [2]

    Gardiner and P

    C. Gardiner and P. Zoller,Quantum noise: a handbook of Markovian and non-Markovian quantum stochastic methods with applications to quantum optics(2004)

    2004

  3. [3]

    H. J. Carmichael, Phys. Rev. Lett.70, 2273 (1993)

    1993

  4. [4]

    V. P. Belavkin, J. Multivar. Anal.42, 171 (1992)

    1992

  5. [5]

    V. P. Belavkin, Autom. Remote Control , 50 (1983)

    1983

  6. [6]

    V. Belavkin, inInformation Complexity and Control in Quantum Physics: Proceedings of the 4th International Seminar on Mathematical Theory of Dynamical Systems and Micro- physics Udine, September 4–13, 1985(Springer, 1987) pp. 311–329

    1985

  7. [7]

    Bouten, R

    L. Bouten, R. Van Handel, and M. R. James, SIAM Rev.51, 239 (2009). 18

    2009

  8. [8]

    Van Handel, J

    R. Van Handel, J. K. Stockton, and H. Mabuchi, J. Opt. B: Quantum Semiclassical Opt. 7, S179 (2005)

    2005

  9. [9]

    Jacobs,Quantum measurement theory and its applications(2014)

    K. Jacobs,Quantum measurement theory and its applications(2014)

    2014

  10. [10]

    M. A. Nielsen, arXiv:quant-ph/0011036 (2000)

    Pith/arXiv arXiv 2000

  11. [11]

    M. A. Nielsen and I. L. Chuang,Quantum computation and quantum information, Vol. 2 (2001)

    2001

  12. [12]

    Maassen and B

    H. Maassen and B. K¨ ummerer, IMS Lecture Notes Monogr. Ser. , 252 (2006)

    2006

  13. [13]

    N. H. Amini, M. Bompais, and C. Pellegrini, J. Phys. A54, 385304 (2021)

    2021

  14. [14]

    Benoist, M

    T. Benoist, M. Fraas, Y. Pautrat, and C. Pellegrini, Probab. Theory Relat. Fields174, 307 (2019)

    2019

  15. [15]

    Bompais,Asymptotic stability and feedback stabilisation of quantum trajectories, Ph.D

    M. Bompais,Asymptotic stability and feedback stabilisation of quantum trajectories, Ph.D. thesis, Paris-Saclay University (2024)

    2024

  16. [16]

    N. H. Amini, T. Benoist, M. Bompais, and C. Pellegrini, arXiv:2512.05770 (2025)

    arXiv 2025

  17. [17]

    Benoist, C

    T. Benoist, C. Pellegrini, and A. Szczepanek, arXiv:2409.18655 (2024)

    Pith/arXiv arXiv 2024

  18. [18]

    Private communication with Cl´ ement Pellegrini and Tristan Benoist

  19. [19]

    M. A. Nielsen, Phys. Rev. A63, 022114 (2001)

    2001

  20. [20]

    Jozsa, J

    R. Jozsa, J. Mod. Opt.41, 2315 (1994)

    1994

  21. [21]

    Girotti and A

    F. Girotti and A. Vitale, arXiv:2509.13377 (2025)

    arXiv 2025

  22. [22]

    M. P. Fisher, V. Khemani, A. Nahum, and S. Vijay, Annu. Rev. Condens. Matter Phys. 14, 335 (2023)

    2023

  23. [23]

    M. J. Gullans and D. A. Huse, Phys. Rev. X10, 041020 (2020)

    2020

  24. [24]

    H. J. Kushner,Introduction to stochastic control(1971)

    1971

  25. [25]

    Amini, P

    H. Amini, P. Rouchon, and M. Mirrahimi, in2011 50th IEEE conference on decision and control and European control conference(IEEE, 2011) pp. 8193–8198. Appendix The following theorem is a slight adaptation of [24, Theorem 1, Chapter 8] (see also [25, Theorem A.1]). Theorem A.1.LetX k be a Markov chain in the compact state spaceS. Suppose that there exists ...

    2011