Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Beyond the arcsine law: exact time-in-lead statistics of a continuously monitored qubit

Tomas Pødenphant Lund

Published Jul 12, 2026• 0 citations
Publisher

Abstract

For a continuously monitored qubit under an ideal diffusive quantum-nondemolition measurement, the fraction of the pre-collapse trajectory during which a given outcome leads is derived exactly—and it is not Lévy’s arcsine law, but an explicitly solvable one-parameter deformation of it. In half-log-odds coordinates the measurement diffusion becomes the exactly solvable tanh-drift (Beneš) filter, whose path law is exactly the initial-population-weighted mixture of two unit-drift Brownian motions—the sequential-testing posterior of Shiryaev. The time-in-lead fraction up to a localization threshold then equals a mixture of drifted-Brownian exit-occupation laws, with a joint Laplace transform in closed form; the family runs from an arcsine-adjacent (but numerically non-arcsine) law to a Bernoulli law in the collapse limit. A second exact identity gives the mean time the eventually-losing outcome leads: exactly half a measurement time from a symmetric start. The reduction is classical filtering theory; what is new is the observable and its exact law for the monitored qubit. All results are verified by simulation, with a step-size study and bootstrap confidence intervals. Target venue: Journal of Statistical Mechanics: Theory and Experiment. v2 update (August 2026). Adds the classical discrete-walk references for the arcsine tradition (Sparre Andersen 1953, 1954, Mathematica Scandinavica) to the introduction's lineage. No results, proofs, or claims change. v3 (August 2026) — the promised inversion is now performed. The closed-form joint transform is inverted numerically with a stated error budget rather than left as a promise; the analytic steps the inversion relies on are justified rather than asserted; and an appendix lemma bound is corrected and proved. Earlier versions remain in the version history.

DOI
10.5281/zenodo.21471598

Related papers

The Decoherence Exponent: Stable Phase Noise and Constraints on Objective State Reduction

Milen V. Velev, Ivo D. Dinov

Let $L_u$ be a symmetric Levy process representing an unobserved phase lift, coupling to a quantum system via charge $Q$. Averaging $e^{iL_uQ}$ gives a completely positive dephasing semigroup. If the characteristic exponent lacks an intrinsic charge scale, the Levy-Khintchine theorem forces $η(ξ)=D | ξ|^α$ ($0<α\le2$), so coherences decay at $Γ_{ab}=D | q_a-q_b |^α$. For integer charge differences, the process descends to the wrapped variable $Θ_u=L_u\bmod 2π$; for real charges, the lift is the relevant phase. Schoenberg's theorem ensures complete positivity for any finite real spectrum when $0<α\le2$, while $α>2$ yields an analytic obstruction. This boundary is exact, verified numerically here. A Gaussian integrated phase gives quadratic charge dependence, though colored Gaussian noise need not yield a Markov semigroup. Two idealized non-Gaussian mechanisms are analyzed: inverse-power Poisson shot noise ($α=d/p$) and Bochner subordination of Brownian phase by an $α/2$-stable clock. For a bounded reduction walk, we prove Born probabilities for every symmetric bounded proposal law. Simulations of truncated stable laws ($α=0.5,1,1.5,2$) confirm this law-independence. In continuous time, $α$-stable drivers fail to reduce: the exact-flow (Marcus) equation oscillates, and the naive Ito jump equation fails to preserve state. For Gaussian white noise, the Itô model collapses to Born probabilities, whereas the exact-flow (Stratonovich) model does not collapse without a threshold and yields non-Born exit probabilities. The quadratic case recovers Milburn's small-step limit, not his exact Poisson dynamics. Finally, a bi-temporal geometry is examined as a possible compact phase source, but closed timelike curves, nonunitary evolution, and an unstable mode tower prevent a consistent field-theoretic realization.

math-phCross-listedarXiv:2608.18335•Aug 20, 2026•0 citations

Universal first-passage time statistics for quantum diffusion

Guido Ladenburger, Finn Schmolke, Eric Lutz

First-passage phenomena play a fundamental role in classical stochastic processes. We here exactly solve a quantum first-passage time problem for quantum diffusion driven by measurement noise, a generalization of classical Brownian motion. Such continuous monitoring may trap the measured quantum system in a decoherence-free subspace, a fraction of the available state space that is isolated from the surroundings, and thus plays an important role in quantum information science. We analytically determine the first-passage time distribution, whose form neither depends on the system Hamiltonian nor on the measurement operator, and is therefore universal. These results provide a general framework to investigate the first-passage statistics of diffusive quantum trajectories.

quant-pharXiv:2511.03455•Nov 6, 2025•3 citations

Definite Outcomes at Infinity: The Bare Binary Biconditional, Repeated QND Measurement, and the Quantum Convergence Conjecture

Christopher M Struck

The Bare Binary Biconditional selected 1/2 as the unique exchange-invariant law on a bare binary alternative and connected independent repetitions of that law to half-frequency and Lebesgue measure. This paper develops the separate dynamic question that originally motivated the biconditional and the present research series: in what precise sense can an accumulating observation record approach a definite binary outcome? A self-contained classical result is given first. Let Θ ∈ {0,1} have the symmetric prior, let observations be conditionally independent with strictly positive and distinguishable likelihoods, and let 𝑞𝑛 = 𝑃(Θ = 1 ∣ 𝑌1,…,𝑌𝑛). Then (𝑞𝑛) is a bounded martingale beginning at 1/2, and identifiability, not martingale convergence alone, gives 𝑞𝑛 → 1{Θ=1} almost surely. Its two endpoints therefore have equal probability. A binary-probe specialization places the observation record itself in {0,1}ℕ; a Gaussian model is retained as a continuous-observation comparison. Repeated independent runs, rather than the dependent observations within one run, produce the fair binary endpoint record connected to [1]. Bauer and Bernard’s repeated quantum non-demolition measurement theorem supplies a bounded quantum counterpart: pointer weights are martingales; under their nondegeneracy condition they converge to pointer-state endpoints; the endpoint probabilities are the initial pointer weights; and relative entropy governs the asymptotic decay of losing weights. Their theorem does not choose 1/2. This paper imports that starting value only under the explicit bare symmetric premise of [1]. Reproducible classical simulations illustrate finite-depth posterior behavior and information-rate scaling without simulating quantum dynamics or treating finite posteriors as completed limits. The remaining claim that observation gives physical significance to this convergence structure is stated separately as the author’s quantum convergence conjecture. Note clarifying this paper’s conclusion: “Particle-like” refers here to a definite observation outcome. In the repeated-QND setting analyzed, the quantum wave-function measurement model mathematically predicts convergence to a definite, particle-like observation outcome. The phrase “not evidence that the object has ceased to be a wave” means only that a particle-like observation is consistent with how the quantum wave function predicts the system should behave upon measurement. This clarification does not deny wave-function collapse or address wave or particle behavior outside the observation process discussed.

•Aug 29, 2026•0 citations

Martingale Collapse in Quantum Measurement A Record-Gated Dynamical Framework for the Emergence of Single Outcomes and Born-Rule Weights

Fahad Almukairish

Abstract Standard quantum mechanics asks us to live with two laws of time. The first is the continuous, linear, unitary evolution of the Schrödinger equation. The second is the discontinuous, non-unitary reduction that appears when a measurement yields one result rather than many. We have learned to use this double description with remarkable success, yet the success of the practice does not remove the tension in the principle. This paper proposes a conditional framework, Record-Gated Martingale Collapse (R-MC), for a narrower and sharper question: how can a state that has branched through ordinary unitary interaction among a system, an apparatus, and an environment settle into a single recorded outcome without adding an external projection rule? The framework begins where ordinary dynamics leaves us: unitary branching and decoherence. It then adds three elements. First, a non-selective record gate opens the collapse mechanism only after inter-branch overlap has weakened and a sufficiently independent environmental record has accumulated. Second, the collapse rate is amplified by the effective number of independent recording channels, not by the raw number of environmental degrees of freedom. Third, an uncontrollable objective seed generates the stochastic path along which branch weights move. Under quantum non-demolition (QND) conditions these weights are bounded martingales that diffusively localize toward a single branch, and Born-rule weights follow from conservation of their mean rather than from a postulated projection. We separate this gradual convergence from the operational recording time, and we distinguish a safe external form, in which the gate depends on local environmental variables and the ensemble mean closes into a linear Lindblad equation, from a more ambitious internal form that requires operational locality of the gate and locality of the apparatus pointers. A distinctive prediction, record-threshold collapse delay, states that two environments with comparable decoherence may nevertheless yield different outcome times if they differ in their capacity for independent, amplifiable recording.

•Jun 28, 2026•0 citations

A divergence norm for the fourth-order coefficient and the invariance of the counting law

Moustafa A. Radwan

The fourth-order coefficient that controls the relative-entropy disturbance of a reduced quantum state under Gaussian coherent noise is identified here as one half of the quadratic form of the chi-squared divergence, evaluated not on a state but on the second-order mean drift of that state. The coefficient had appeared in an existing series of papers as a computed number attached to each noise family separately, with no statement of what kind of object it is. Naming it as a divergence norm on the drift is what makes the three results below possible, and each of them is a statement the series could not have reached while the coefficient remained a number. The first result is an identification, not a new computation. Five symbols are used for that coefficient across the series, carrying three different normalisations; they are shown to denote one quantity, and the conversion between them is recorded in a table. The normalisation that each symbol carries is stated explicitly, because the three differ by factors of the monitored dimension and of the total dimension, and a reader comparing two of the papers without that table would conclude that they disagree. The second result is a deformation. When the generating function of the expansion is replaced by the one-parameter family whose member at unit parameter is the logarithm, the coefficient becomes the quadratic form of the corresponding divergence of that family. The reason is local and explicit: the confluent value of the deformed kernel is the second derivative of the generator, so the kernel that the expansion produces is the Hessian of the generating function and nothing else. Closed forms follow for both noise families already studied, expressed in the power sums of the deformed theory. For unitarily invariant noise the coefficient is a single quadratic combination of three power sums. For Pauli-string noise on a stabilizer code only one power sum moves. The third result follows from the second. Because the only dependence on the deformation parameter sits in the diagonal of the kernel, whose sum over the spectrum is that single power sum, the counting law for stabilizer codes is invariant in form under the entire deformation, with the power sum read at the working parameter. The law is therefore not a feature of the relative entropy in particular; it is a feature of the whole family, and the choice of generating function changes one scalar in it and nothing of its structure. A structural correction to the series is also recorded, and it is the result with the clearest consequence for error correction. Protection at fourth order requires two independent conditions: the vanishing of every first-order amplitude, and the vanishing of the drift. The Knill-Laflamme condition supplies only the first. A code satisfying it may therefore carry a fourth-order error, and the residual is exactly the coefficient above together with the connected channel. That residual is thus the obstruction to full protection at this order, and a protection statement elsewhere in the series, which asserts the two conditions to be equivalent, is strictly weaker than it claims. A code and a noise family exhibiting the gap are given, and the residual is measured on them. The manuscript is self-contained. Three appendices derive, from the definitions, every ingredient that the results use: the two expansion coefficients and the quadratic form through the divided-difference Frechet derivative, the drift of each noise family from the completeness identities of the mode algebra, and the two vanishing arguments. No statement in the paper depends on work that is not yet published, and one appendix reproduces, from a convention-free law for the drift, the published coefficient of an earlier paper of the series, including the factor that the geometry of four tensor factors contributes to it. Every identity is certified numerically against the integral definitions of the deformed theory rather than against the closed forms used to derive it: the power sums are computed by quadrature in the substitution t = exp(u), with the interval split at the logarithm of every eigenvalue, at ninety working digits, and only then compared. Every statement of the appendices is checked separately on explicit matrices built from the definitions, with the perturbation applied by exact matrix exponentiation. Each check reports the discrepancy it attains rather than a verdict, so that the resolution of the test is visible; where a finite difference sets that resolution, the tolerance of the row is the resolution attained and not a wish. Three rational values reported as exactly verified in an earlier paper of the series are reproduced here from a single spectrum by a route not used when they were obtained. The archive deposited with this record carries the four self-contained checks, a README describing what each one measures, and the licence. Four further guards belong to the manuscript and are not in the archive, because none of them can run without files that are not part of it; they check the documents and the deposit rather than the mathematics, and all four are named in the README. The limitations of the work are stated in the manuscript as six paragraphs, among them the restriction of the drift to the diagonal case, the factor outside the weight hypothesis, and the extent of the prior-art search behind the disclaimer of novelty, which was two queries and the reading of two abstracts: the verdict recorded for the new statements is therefore "not found", which is weaker than "new". The use of an AI tool in the research and in the preparation of the manuscript is disclosed in the Acknowledgements, with the model and its version named, and described in the section on numerical certification.

•Nov 2, 2026•0 citations

Optimal Estimation of Qubit States with Continuous Time Measurements

Mădălin Guţǎ, Bas Janssens, Jonas Kahn

arXiv:quant-ph/0608074•Communications in Mathematical Physics•Oct 26, 2007•41 citations
For a continuously monitored qubit under an ideal diffusive quantum-nondemolition measurement, the fraction of the pre-collapse trajectory during which a given outcome leads is derived exactly—and it is not Lévy’s arcsine law, but an explicitly solvable one-parameter deformation of it. In half-log-odds coordinates the measurement diffusion becomes the exactly solvable tanh-drift (Beneš) filter, whose path law is exactly the initial-population-weighted mixture of two unit-drift Brownian motions—the sequential-testing posterior of Shiryaev. The time-in-lead fraction up to a localization threshold then equals a mixture of drifted-Brownian exit-occupation laws, with a joint Laplace transform in closed form; the family runs from an arcsine-adjacent (but numerically non-arcsine) law to a Bernoulli law in the collapse limit. A second exact identity gives the mean time the eventually-losing outcome leads: exactly half a measurement time from a symmetric start. The reduction is classical filtering theory; what is new is the observable and its exact law for the monitored qubit. All results are verified by simulation, with a step-size study and bootstrap confidence intervals. Target venue: Journal of Statistical Mechanics: Theory and Experiment. v2 update (August 2026). Adds the classical discrete-walk references for the arcsine tradition (Sparre Andersen 1953, 1954, Mathematica Scandinavica) to the introduction's lineage. No results, proofs, or claims change. v3 (August 2026) — the promised inversion is now performed. The closed-form joint transform is inverted numerically with a stated error budget rather than left as a promise; the analytic steps the inversion relies on are justified rather than asserted; and an appendix lemma bound is corrected and proved. Earlier versions remain in the version history.
Let LuL_uLu​ be a symmetric Levy process representing an unobserved phase lift, coupling to a quantum system via charge QQQ. Averaging eiLuQe^{iL_uQ}eiLu​Q gives a completely positive dephasing semigroup. If the characteristic exponent lacks an intrinsic charge scale, the Levy-Khintchine theorem forces η(ξ)=D∣ξ∣αη(ξ)=D | ξ|^αη(ξ)=D∣ξ∣α (0<α≤20<α\le20<α≤2), so coherences decay at Γab=D∣qa−qb∣αΓ_{ab}=D | q_a-q_b |^αΓab​=D∣qa​−qb​∣α. For integer charge differences, the process descends to the wrapped variable Θu=Lu mod 2πΘ_u=L_u\bmod 2πΘu​=Lu​mod2π; for real charges, the lift is the relevant phase. Schoenberg's theorem ensures complete positivity for any finite real spectrum when 0<α≤20<α\le20<α≤2, while α>2α>2α>2 yields an analytic obstruction. This boundary is exact, verified numerically here. A Gaussian integrated phase gives quadratic charge dependence, though colored Gaussian noise need not yield a Markov semigroup. Two idealized non-Gaussian mechanisms are analyzed: inverse-power Poisson shot noise (α=d/pα=d/pα=d/p) and Bochner subordination of Brownian phase by an α/2α/2α/2-stable clock. For a bounded reduction walk, we prove Born probabilities for every symmetric bounded proposal law. Simulations of truncated stable laws (α=0.5,1,1.5,2α=0.5,1,1.5,2α=0.5,1,1.5,2) confirm this law-independence. In continuous time, ααα-stable drivers fail to reduce: the exact-flow (Marcus) equation oscillates, and the naive Ito jump equation fails to preserve state. For Gaussian white noise, the Itô model collapses to Born probabilities, whereas the exact-flow (Stratonovich) model does not collapse without a threshold and yields non-Born exit probabilities. The quadratic case recovers Milburn's small-step limit, not his exact Poisson dynamics. Finally, a bi-temporal geometry is examined as a possible compact phase source, but closed timelike curves, nonunitary evolution, and an unstable mode tower prevent a consistent field-theoretic realization.
First-passage phenomena play a fundamental role in classical stochastic processes. We here exactly solve a quantum first-passage time problem for quantum diffusion driven by measurement noise, a generalization of classical Brownian motion. Such continuous monitoring may trap the measured quantum system in a decoherence-free subspace, a fraction of the available state space that is isolated from the surroundings, and thus plays an important role in quantum information science. We analytically determine the first-passage time distribution, whose form neither depends on the system Hamiltonian nor on the measurement operator, and is therefore universal. These results provide a general framework to investigate the first-passage statistics of diffusive quantum trajectories.
The Bare Binary Biconditional selected 1/2 as the unique exchange-invariant law on a bare binary alternative and connected independent repetitions of that law to half-frequency and Lebesgue measure. This paper develops the separate dynamic question that originally motivated the biconditional and the present research series: in what precise sense can an accumulating observation record approach a definite binary outcome? A self-contained classical result is given first. Let Θ ∈ {0,1} have the symmetric prior, let observations be conditionally independent with strictly positive and distinguishable likelihoods, and let 𝑞𝑛 = 𝑃(Θ = 1 ∣ 𝑌1,…,𝑌𝑛). Then (𝑞𝑛) is a bounded martingale beginning at 1/2, and identifiability, not martingale convergence alone, gives 𝑞𝑛 → 1{Θ=1} almost surely. Its two endpoints therefore have equal probability. A binary-probe specialization places the observation record itself in {0,1}ℕ; a Gaussian model is retained as a continuous-observation comparison. Repeated independent runs, rather than the dependent observations within one run, produce the fair binary endpoint record connected to [1]. Bauer and Bernard’s repeated quantum non-demolition measurement theorem supplies a bounded quantum counterpart: pointer weights are martingales; under their nondegeneracy condition they converge to pointer-state endpoints; the endpoint probabilities are the initial pointer weights; and relative entropy governs the asymptotic decay of losing weights. Their theorem does not choose 1/2. This paper imports that starting value only under the explicit bare symmetric premise of [1]. Reproducible classical simulations illustrate finite-depth posterior behavior and information-rate scaling without simulating quantum dynamics or treating finite posteriors as completed limits. The remaining claim that observation gives physical significance to this convergence structure is stated separately as the author’s quantum convergence conjecture. Note clarifying this paper’s conclusion: “Particle-like” refers here to a definite observation outcome. In the repeated-QND setting analyzed, the quantum wave-function measurement model mathematically predicts convergence to a definite, particle-like observation outcome. The phrase “not evidence that the object has ceased to be a wave” means only that a particle-like observation is consistent with how the quantum wave function predicts the system should behave upon measurement. This clarification does not deny wave-function collapse or address wave or particle behavior outside the observation process discussed.
Abstract Standard quantum mechanics asks us to live with two laws of time. The first is the continuous, linear, unitary evolution of the Schrödinger equation. The second is the discontinuous, non-unitary reduction that appears when a measurement yields one result rather than many. We have learned to use this double description with remarkable success, yet the success of the practice does not remove the tension in the principle. This paper proposes a conditional framework, Record-Gated Martingale Collapse (R-MC), for a narrower and sharper question: how can a state that has branched through ordinary unitary interaction among a system, an apparatus, and an environment settle into a single recorded outcome without adding an external projection rule? The framework begins where ordinary dynamics leaves us: unitary branching and decoherence. It then adds three elements. First, a non-selective record gate opens the collapse mechanism only after inter-branch overlap has weakened and a sufficiently independent environmental record has accumulated. Second, the collapse rate is amplified by the effective number of independent recording channels, not by the raw number of environmental degrees of freedom. Third, an uncontrollable objective seed generates the stochastic path along which branch weights move. Under quantum non-demolition (QND) conditions these weights are bounded martingales that diffusively localize toward a single branch, and Born-rule weights follow from conservation of their mean rather than from a postulated projection. We separate this gradual convergence from the operational recording time, and we distinguish a safe external form, in which the gate depends on local environmental variables and the ensemble mean closes into a linear Lindblad equation, from a more ambitious internal form that requires operational locality of the gate and locality of the apparatus pointers. A distinctive prediction, record-threshold collapse delay, states that two environments with comparable decoherence may nevertheless yield different outcome times if they differ in their capacity for independent, amplifiable recording.
The fourth-order coefficient that controls the relative-entropy disturbance of a reduced quantum state under Gaussian coherent noise is identified here as one half of the quadratic form of the chi-squared divergence, evaluated not on a state but on the second-order mean drift of that state. The coefficient had appeared in an existing series of papers as a computed number attached to each noise family separately, with no statement of what kind of object it is. Naming it as a divergence norm on the drift is what makes the three results below possible, and each of them is a statement the series could not have reached while the coefficient remained a number. The first result is an identification, not a new computation. Five symbols are used for that coefficient across the series, carrying three different normalisations; they are shown to denote one quantity, and the conversion between them is recorded in a table. The normalisation that each symbol carries is stated explicitly, because the three differ by factors of the monitored dimension and of the total dimension, and a reader comparing two of the papers without that table would conclude that they disagree. The second result is a deformation. When the generating function of the expansion is replaced by the one-parameter family whose member at unit parameter is the logarithm, the coefficient becomes the quadratic form of the corresponding divergence of that family. The reason is local and explicit: the confluent value of the deformed kernel is the second derivative of the generator, so the kernel that the expansion produces is the Hessian of the generating function and nothing else. Closed forms follow for both noise families already studied, expressed in the power sums of the deformed theory. For unitarily invariant noise the coefficient is a single quadratic combination of three power sums. For Pauli-string noise on a stabilizer code only one power sum moves. The third result follows from the second. Because the only dependence on the deformation parameter sits in the diagonal of the kernel, whose sum over the spectrum is that single power sum, the counting law for stabilizer codes is invariant in form under the entire deformation, with the power sum read at the working parameter. The law is therefore not a feature of the relative entropy in particular; it is a feature of the whole family, and the choice of generating function changes one scalar in it and nothing of its structure. A structural correction to the series is also recorded, and it is the result with the clearest consequence for error correction. Protection at fourth order requires two independent conditions: the vanishing of every first-order amplitude, and the vanishing of the drift. The Knill-Laflamme condition supplies only the first. A code satisfying it may therefore carry a fourth-order error, and the residual is exactly the coefficient above together with the connected channel. That residual is thus the obstruction to full protection at this order, and a protection statement elsewhere in the series, which asserts the two conditions to be equivalent, is strictly weaker than it claims. A code and a noise family exhibiting the gap are given, and the residual is measured on them. The manuscript is self-contained. Three appendices derive, from the definitions, every ingredient that the results use: the two expansion coefficients and the quadratic form through the divided-difference Frechet derivative, the drift of each noise family from the completeness identities of the mode algebra, and the two vanishing arguments. No statement in the paper depends on work that is not yet published, and one appendix reproduces, from a convention-free law for the drift, the published coefficient of an earlier paper of the series, including the factor that the geometry of four tensor factors contributes to it. Every identity is certified numerically against the integral definitions of the deformed theory rather than against the closed forms used to derive it: the power sums are computed by quadrature in the substitution t = exp(u), with the interval split at the logarithm of every eigenvalue, at ninety working digits, and only then compared. Every statement of the appendices is checked separately on explicit matrices built from the definitions, with the perturbation applied by exact matrix exponentiation. Each check reports the discrepancy it attains rather than a verdict, so that the resolution of the test is visible; where a finite difference sets that resolution, the tolerance of the row is the resolution attained and not a wish. Three rational values reported as exactly verified in an earlier paper of the series are reproduced here from a single spectrum by a route not used when they were obtained. The archive deposited with this record carries the four self-contained checks, a README describing what each one measures, and the licence. Four further guards belong to the manuscript and are not in the archive, because none of them can run without files that are not part of it; they check the documents and the deposit rather than the mathematics, and all four are named in the README. The limitations of the work are stated in the manuscript as six paragraphs, among them the restriction of the drift to the diagonal case, the factor outside the weight hypothesis, and the extent of the prior-art search behind the disclaimer of novelty, which was two queries and the reading of two abstracts: the verdict recorded for the new statements is therefore "not found", which is weaker than "new". The use of an AI tool in the research and in the preparation of the manuscript is disclosed in the Acknowledgements, with the model and its version named, and described in the section on numerical certification.