On the boundaries, asymptotic law and Bernoulli-Doob representation of homogeneous bounded martingales

Damiano Brigo; Fr\'ed\'eric Vrins

REVIEW 2 major objections 2 minor 1 cited by

Every bounded homogeneous diffusion martingale is a Bernoulli-Doob martingale, and conversely.

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 · grok-4.5

2026-07-15 06:40 UTC pith:DRJ6IJ25

load-bearing objection Clean two-way equivalence claim packaging bounded homogeneous diffusion martingales as Bernoulli-Doob processes; abstract-only, so proofs unchecked but nothing looks broken. the 2 major comments →

arxiv 2607.12365 v1 pith:DRJ6IJ25 submitted 2026-07-14 math.PR

On the boundaries, asymptotic law and Bernoulli-Doob representation of homogeneous bounded martingales

Damiano Brigo , Fr\'ed\'eric Vrins This is my paper

classification math.PR MSC 60G4460H1060J60

keywords homogeneous diffusion martingalesBernoulli-Doob martingalesbounded state spaceFeller boundary classificationabsorptionasymptotic Bernoulli lawJacobi martingalecredit-risk modelling

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.

This paper claims that any homogeneous diffusion martingale living in a closed interval whose endpoints are zeros of the diffusion coefficient must be a Bernoulli-Doob martingale: it equals the conditional expectation of a single Bernoulli random variable that takes only the two boundary values. The converse also holds for continuous time-homogeneous Markov martingales on a Brownian filtration. The intuition is that a bounded martingale keeps a fixed mean while steadily accumulating variance, so its only possible terminal law is the two-point distribution of maximum variance on that interval. The paper further shows that this Bernoulli limit is reached only asymptotically: at every finite time the process still has positive probability of remaining inside the open interval, even when the boundaries are attainable. Finally it reconciles Feller’s classical boundary classification with the pathwise SDE description by proving that the martingale property itself forces absorption at every attainable boundary. The resulting picture unifies several well-known examples (the Φ-martingale, the Jacobi martingale) and supplies a clean modelling tool for credit-risk dynamics that must stay between 0 and 1.

Core claim

Every homogeneous diffusion martingale evolving inside a bounded interval whose endpoints annihilate the diffusion coefficient is a Bernoulli-Doob martingale Z_t = E[B | F_t], and every continuous time-homogeneous Markov Bernoulli-Doob martingale on a Brownian filtration arises from such a diffusion. The Bernoulli terminal law is approached only as time tends to infinity; for every finite horizon the probability of still being strictly inside the interval remains positive.

What carries the argument

The Bernoulli-Doob representation Z_t = E[B | F_t], where B is a Bernoulli random variable taking the two boundary values of the state space; this identity is forced by the simultaneous requirements of constant mean and maximal variance accumulation under the martingale and diffusion constraints.

Load-bearing premise

The martingale property forces absorption at every attainable boundary, reconciling Feller’s classification with the pathwise SDE description; without that absorption the Bernoulli terminal law and the claimed equivalence both collapse.

What would settle it

Construct or simulate a continuous time-homogeneous Markov martingale on a Brownian filtration that remains strictly inside (a,b) with positive probability for all time, or that hits a boundary and then leaves it while preserving the martingale property; either object would contradict the claimed equivalence.

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

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

Clean two-way equivalence claim packaging bounded homogeneous diffusion martingales as Bernoulli-Doob processes; abstract-only, so proofs unchecked but nothing looks broken.

read the letter

The punchline is a two-way equivalence: every homogeneous diffusion martingale on a bounded interval where the diffusion vanishes at the endpoints is a Bernoulli-Doob martingale Z_t = E[B | F_t], and conversely every continuous time-homogeneous Markov Bernoulli-Doob martingale on a Brownian filtration comes from such a diffusion. They also claim that for any fixed finite T the process has not yet hit the boundary with strictly positive probability, even when boundaries are accessible.

What is actually new is the packaging. The variance-maximisation intuition (constant mean, accumulating variance → Bernoulli terminal law) is standard, and absorption under the martingale constraint is classical optional-stopping material. The contribution is organising that class under a single representation, clarifying how Feller’s boundary classification sits with the pathwise SDE and the martingale constraint, and spelling out the finite-horizon non-absorption statement, with pointers to Φ-martingales, Jacobi, and credit-risk models. That is useful within the subfield if the proofs hold.

Soft spots are mostly about what we cannot check. Only the abstract is here, so the SDE coefficients, boundary-integral calculations, and the precise hypotheses of Theorems 2–4 are invisible. The load-bearing step—that the martingale property forces absorption at attainable boundaries—is not a leap; it is standard. Naming “Bernoulli-Doob” is definitional rather than circular. Nothing in the abstract is internally inconsistent or overclaimed relative to what a theorem paper can deliver. The main risk is that the equivalence is thinner than it sounds once the technical conditions are written out, or that parts of it already sit in the literature under different names.

This is for people who work with bounded continuous martingales in stochastic analysis and quantitative finance (Jacobi-type models, credit). A serious referee in that area should see it. Send it to peer review rather than desk-reject; the claim is clean enough and the subfield cares about the organisation. I would not put it in next week’s reading group until the full text is available, and I would not cite it yet.

Referee Report

2 major / 2 minor

Summary. The manuscript studies continuous homogeneous diffusion martingales on a bounded interval D=[a,b] whose diffusion coefficient vanishes at the endpoints a and b. It introduces Bernoulli-Doob martingales Z_t=E[B|F_t] for a Bernoulli random variable B and claims a two-way equivalence: every such diffusion martingale is a Bernoulli-Doob martingale (Theorem 2), and every continuous time-homogeneous Markov Bernoulli-Doob martingale on a Brownian filtration arises from such a diffusion (Theorem 3). The paper further asserts that the Bernoulli terminal law is only asymptotic: for any finite T the probability of not having hit the boundary remains strictly positive (Theorem 4). It also clarifies that the martingale property forces absorption at attainable boundaries, reconciling Feller classification with the pathwise SDE setting, and illustrates the theory with the Φ-martingale, the Jacobi martingale, and credit-risk applications.

Significance. If the claimed equivalence holds, the paper supplies a clean structural characterisation of a natural class of bounded continuous martingales and links it to the classical maximum-variance property of the Bernoulli law. The asymptotic non-absorption statement (Theorem 4) and the absorption-under-martingale clarification are useful for applications (e.g., credit-risk modelling) that rely on continuous martingale dynamics with absorbing barriers. The results appear to rest on standard tools (optional stopping, Feller boundary classification, SDE uniqueness) rather than exotic machinery, so the contribution is primarily organisational and conceptual; that is still of interest to the stochastic-processes community provided the proofs are complete.

major comments (2)

Only the abstract is available for review. Theorems 2–4 are stated cleanly and the variance-maximisation intuition is standard, but no proofs, SDE coefficients, boundary-integral calculations, or uniqueness arguments can be inspected. A full technical assessment of the load-bearing claims (especially the absorption-under-martingale premise and the converse construction in Theorem 3) is therefore impossible; the recommendation is necessarily provisional pending the complete manuscript.
The abstract asserts that the martingale property forces absorption at any attainable boundary and that this underpins Theorems 2–4. While this is classically true for continuous bounded martingales (optional stopping / local-martingale analysis), the paper must supply an explicit argument that reconciles Feller’s classification with the pathwise SDE framework under the precise regularity assumed on the diffusion coefficient. Without that argument the claimed equivalence remains incompletely justified.

minor comments (2)

The abstract introduces the term “Bernoulli-Doob martingale” without a literature pointer; a short remark on whether the notion is new or already appears under another name would help the reader.
The three illustrative examples (Φ-martingale, Jacobi martingale, credit-risk) are named but not described; even a one-sentence characterisation of each would make the abstract more self-contained.

Circularity Check

0 steps flagged

No circularity detectable from the abstract: claimed equivalences are stated as theorems, not forced by definition or fit.

full rationale

Only the abstract is available, so no internal derivation chain, equations, or self-citations can be inspected for reduction-by-construction. On the given text, Bernoulli-Doob is introduced by definition as Z_t=E[B|F_t] and then claimed equivalent (Theorems 2–3) to the class of homogeneous diffusion martingales on D=[a,b] with vanishing diffusion coefficient at the endpoints; that is ordinary naming plus a two-way mathematical claim, not a self-definitional loop or a fitted parameter renamed as prediction. The load-bearing absorption premise (martingale property forces absorption at attainable boundaries) is presented as a clarifying reconciliation of Feller classification with the pathwise SDE framework and is a classical consequence of optional stopping for continuous bounded martingales, not an author-imported uniqueness theorem or ansatz smuggled via self-citation. Theorem 4’s strict positive non-absorption probability at finite T is an independent asymptotic claim, not forced by the terminal Bernoulli law. No fitted inputs, no self-citation chain, and no renaming of a known empirical pattern appear in the abstract. Per the analyzer rules, absence of quotable circular reduction yields score 0 with empty steps; residual risk that the full proofs might later introduce circularity cannot be scored without the text.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 1 invented entities

Abstract-only pure-probability paper. No numerical free parameters appear. Background axioms are standard stochastic calculus and Feller boundary theory; the main modelling restriction (diffusion coefficient vanishes at a and b, continuous time-homogeneous Markov structure on a Brownian filtration) is a domain assumption that defines the class under study. Bernoulli-Doob is a named construction, not a new physical entity.

axioms (5)

standard math Standard Itô calculus / continuous semimartingale theory on a Brownian filtration
Required for diffusion martingales and conditional-expectation representations; assumed throughout the abstract’s setting.
standard math Feller’s boundary classification for one-dimensional diffusions
Invoked explicitly when the abstract claims to clarify the relationship between Feller classification, pathwise SDEs, and the martingale constraint.
domain assumption Diffusion coefficient vanishes at the endpoints a and b of the state space
Defines the class of processes studied; without it the process need not stay in [a,b] or become absorbed.
domain assumption The process is continuous, time-homogeneous, and Markov, and is a true martingale
Load-bearing for both directions of the equivalence (Theorems 2–3) and for the absorption claim.
ad hoc to paper Martingale property forces absorption at any attainable boundary
Stated as a clarifying result of the paper; if false, the Bernoulli terminal law and the claimed equivalence would fail for attainable boundaries.

invented entities (1)

Bernoulli-Doob martingale independent evidence
purpose: Name for processes of the form Z_t = E[B | F_t] with B Bernoulli; used as one side of the equivalence.
Not a new physical object; it is a standard Doob martingale of a Bernoulli random variable, given a convenient name for the characterisation theorems.

pith-pipeline@v1.1.0-grok45 · 6150 in / 2732 out tokens · 33402 ms · 2026-07-15T06:40:11.312440+00:00 · methodology

0 comments

read the original abstract

We study homogeneous diffusion martingales evolving in a bounded state space $D=[a,b]$, where $a$ and $b$ are zeros of the diffusion coefficient. We call a process of the form $Z_t=\mathbb{E}[B\mid\mathcal{F}_t]$, with $B$ a Bernoulli random variable, a Bernoulli-Doob martingale. Our main results establish a complete equivalence: every such diffusion martingale is a Bernoulli-Doob martingale (Theorem 2) and, conversely, every continuous time-homogeneous Markov Bernoulli-Doob martingale on a Brownian filtration arises from such a diffusion (Theorem 3). The intuitive reason is that a bounded martingale has constant expectation while accumulating variance, so it converges to the maximum-variance distribution with given mean and range, namely the Bernoulli. We further show that this Bernoulli limit is truly asymptotic: for any fixed finite horizon $T$, the probability of not yet having reached the boundary is strictly positive (Theorem 4), even when the individual boundaries are accessible. We clarify the relationship between Feller's boundary classification, the pathwise SDE framework, and the martingale constraint, showing that the martingale property forces absorption at any attainable boundary. The theory is illustrated with the $\Phi$-martingale, the Jacobi martingale, and applications to credit-risk modelling.

discussion (0)

Forward citations

Cited by 1 Pith paper

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

Embedding martingale diffusions as binary posteriors in sequential inference
math.PR 2026-07 accept novelty 7.0

Every autonomous win-martingale on [0,1] is the exact posterior of an explicit binary diffusion experiment, and current-state posteriors exist exactly under a Riccati condition on the drift gap.