REVIEW 5 minor 12 references
Sharpness of the avalanche phase transition in the Bak--Sneppen model
T0 review · 0 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read The Bak–Sneppen avalanche transition is sharp: for both update rules the three critical thresholds coincide, and stationary fitnesses converge to independent exponentials above the common critical level.
desk verdict Resolves a long-standing open problem in the Bak–Sneppen model; the main proof chain checks out, and the paper deserves serious refereeing. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The engine is a triangular system for the tail probabilities of the one-sided avalanche range. Let $X_t$ be the shifted range (number of refreshed sites minus one) at fitness level $t$, started from one forced update, and set $q_k(t)=\mathbb{P}(X_t\ge k)$. A regeneration lemma says that, conditionally on a finite avalanche, the terminal fitnesses are independent $t+\mathrm{Exp}(1)$ variables independent of the explored genealogy; iterating it shows that $X_t$ evolves by a self-interacting jump process whose tails satisfy the triangular system $$q'_k=\sum_{u=1}^{k-1}q_u(q_{k-u}-q_k)+q_k(1-q_k),\quad k\ge2,$$ with $q_1\equiv1$ and $q_k(0)=0$. This system is the unique input from the stochastic construction used in the analytic part. Summing the system and applying an extremal inequality for decreasing probability masses—for independent identically distributed $K,K'$ with decreasing mass function, $\mathbb{E}[\max\{K,K'\}]\ge \frac43\mathbb{E}[K]-\frac13$—yields the critical lower tail $\mathbb{P}(X_{t^{\mathrm{os}}_{\mathrm{sus}}}\ge k)\ge c k^{-2/3}$. Because the exponent $2/3$ is below one, a process whose graft law (the law of the independent sub-avalanche attached at each proposed update) is frozen at the critical tails explodes with positive probability in arbitrarily small positive time, and a comparison of the true tails with the frozen ones transfers this explosion to every level $t^{\mathrm{os}}_{\mathrm{sus}}+s$. The isotropic model has no closed scalar evolution, but its left endpoint satisfies an integrated form of the same tail system as an inequality, so restarting the tail flow from a truncated endpoint law transfers the explosion there as well. Finally, the range–duration identity $D(t)=\exp(\int_0^t R(s)\,ds)$ identifies $t_d$ with $t_{\mathrm{sus}}$, and a locking-threshold argument converts sharpness into the stationary product limit.
What would settle it
Simulate many finite $t$-avalanches in the one-sided model near the suspected critical level, recording the full genealogy and all terminal fitnesses: if the terminal excesses above $t$ are not independent $\mathrm{Exp}(1)$ variables conditionally on the genealogy, the regeneration lemma fails and the triangular tail system would not describe the true range.
Extended reading notes
Core claim
For each update rule $\gamma\in\{\mathrm{os},\mathrm{iso}\}$, the paper establishes the equality $t^{\gamma}_d=t^{\gamma}_{\mathrm{sus}}=t^{\gamma}_{\infty}=:t^{\gamma}_c$, so the mean-duration threshold, the mean-range threshold, and the first infinite-avalanche threshold all coincide. The critical values satisfy $1\le t^{\mathrm{os}}_c\le 2\log 2$ and $\tfrac13\le t^{\mathrm{iso}}_c\le t^{\mathrm{os}}_c$. At the critical level the avalanche range is almost surely finite, while for every $s>0$ an avalanche at level $t^{\gamma}_c+s$ is infinite with positive probability. For every $t<t^{\gamma}_c$ the range has exponential moments. As a consequence, for any fixed distinct sites $x_1,\dots,x_\ell$, the stationary fitness vector converges in law to $\nu_{t^{\gamma}_c}^{\otimes\ell}$, where $\nu_t$ is the law of $t+\mathrm{Exp}(1)$; equivalently, in the usual uniform parametrisation, the limiting fitnesses are independent and uniform on $[q^{\gamma}_c,1]$ with $q^{\gamma}_c=1-e^{-t^{\gamma}_c}$.
Load-bearing premise
The load-bearing premise is the regeneration lemma: once a finite threshold-$t$ avalanche stops, its terminal fitnesses are independent $t+\mathrm{Exp}(1)$ variables, independent of the explored genealogy. If this self-similar memory loss failed, the triangular tail system would not govern the true avalanche range, and the sharpness argument would lack its central object.
Editorial extensions
If this is right
- The susceptibility threshold is also the threshold for actual infinite avalanches, so diverging mean range is not an artifact of averaging over rare events.
- For every fixed finite set of sites, the stationary fitnesses converge to independent $t^{\gamma}_c+\mathrm{Exp}(1)$ variables; in the original coordinates the limiting law is uniform on $[q^{\gamma}_c,1]$.
- At the critical level avalanches are almost surely finite, but adding any positive amount to the threshold gives a strictly positive probability of an infinite avalanche.
- Below criticality the avalanche range has exponential moments, and the mean duration is finite exactly when the mean range is finite.
- The critical values are bracketed, with $1\le t^{\mathrm{os}}_c\le 2\log 2$ and $\tfrac13\le t^{\mathrm{iso}}_c\le t^{\mathrm{os}}_c$, so both models have a genuine finite critical level.
Reading between the lines
- The proof only needs the critical tail exponent to be strictly below $1$, not the particular value $2/3$; the sharpness conclusion would survive a different polynomial lower bound, so the true avalanche-size exponent could differ without affecting the threshold equality.
- The same triangular-system-plus-explosion mechanism may be reusable for other one-dimensional extremal-update processes whose avalanche range admits a scalar tail equation with a subcritical power tail.
- A testable finite-size prediction is that, on a cycle of size $N$, the finite-volume avalanche range distribution stays close to the infinite-volume one up to $t^{\gamma}_c$ and then develops a sharp mass at the full-system range, with the crossover width shrinking as $N$ grows.
- For the isotropic model, the endpoint-supersolution transfer suggests that exact scalar equations for the full range may not be necessary for sharpness in two-sided dynamics; one-sided tail flows started from dominated initial data may be a general transfer tool.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proves sharpness of the avalanche phase transition in the one-dimensional Bak–Sneppen model for both the one-sided and isotropic update rules. The main theorem identifies the susceptibility, infinite-avalanche, and duration thresholds, t_sus = t_∞ = t_d, and shows that the stationary finite-dimensional marginals converge to independent exponentials conditioned above the common critical level. The proof reduces the one-sided avalanche range to a scalar self-interacting jump process whose tail probabilities satisfy the triangular system (2), derives a polynomial lower bound on the critical tail via an extremal inequality for decreasing probability mass functions, and transfers the sharpness to the isotropic model through an endpoint domination argument. The finite-volume reduction and stationary limit use the Meester–Znamenski framework, adapted in Appendix A.
Significance. If correct, the paper resolves a substantial open problem in the mathematical treatment of the Bak–Sneppen model: it establishes that divergence of the mean avalanche range occurs exactly at the first appearance of infinite avalanches, and it confirms the conjectured product-form stationary limit in both update rules. The result is sharp and the proof is analytic and largely self-contained. Among the paper's specific strengths are the parameter-free compensation estimate A(t) ≤ m(t)^{8/3} (Proposition 4.3), the Paley–Zygmund critical lower tail q_k(t*) ≥ c k^{-2/3} (Proposition 4.4), the frozen-process explosion mechanism (Proposition 5.1), and the delicate monotone comparison for isotropic endpoints (Proposition 7.4). The potential circularity concern about Lemma 2.2 leading to system (2) does not land on reading: the graphical representation is a representation of the physical avalanche obtained by iterating the regeneration property, and the tail system is then derived by a compensator calculation. I found no load-bearing mathematical error.
minor comments (5)
- [Section 2, Lemma 2.2 and following paragraph] The passage from the regeneration lemma to the infinite-volume self-similar range process is correct, but it would be helpful to state explicitly that the marks in the graphical representation are not an independent construction of the physical law but rather a pathwise representation obtained by iterated regeneration, with μ_s denoting the law of the physical s-avalanche; as written, a reader could momentarily suspect circularity.
- [Proposition 4.4] The sentence 'decreasing the constant proves the estimate for all n≥1' should specify that the estimate is for all n≥1 at the critical level t*, since at t=0 one has q_n(0)=0 for n≥2 and the lower bound cannot hold there.
- [Proposition 5.2] In the comparison step, the inequality q_k(1-q_k) ≥ Q_k(1-q_k) uses both q_k ≥ Q_k and 1-q_k ≥ 0; this is correct, but making those two inequalities explicit would improve readability.
- [Appendix A.2, Proposition A.4] There are typographical repetitions of 'spannings' instead of 'spanning' in the text of Proposition A.4 and its proof; these should be corrected.
- [Proposition 4.5] The parenthetical note 'Let us point this continuity statement is not logically needed to derive the equality of one-sided critical thresholds' is informal for a main text; it would fit better as a formal remark after the proposition.
Circularity Check
No significant circularity: the sharpness theorem is derived from the avalanche range system and external Meester–Znamenski inputs, not assumed in the hypotheses.
full rationale
The central derivation chain is self-contained and does not reduce to its inputs. The triangular tail system (2) is not an ansatz or a renamed conclusion: Proposition 2.4 derives it from the marked-Poisson graphical construction and the regeneration Lemma 2.2, which is itself proved from the lack-of-memory property of independent exponentials. All subsequent analytic steps, including the first-moment equation (5), the compensation estimate A(t)<=m(t)^{8/3}, the critical lower tail q_k(t*)>=c k^{-2/3}, the no-infinite-avalanche-at-criticality statement, and the frozen explosion transfer, are obtained by explicit differential and comparison arguments from that system; none of them substitutes the target equality t_sus=t_infty into the hypotheses. The isotropic sharpness proof likewise derives endpoint tail inequalities from the same graphical construction and compares them with the one-sided tail flow, so no isotropic conclusion is assumed in the one-sided argument. The range-duration identity and locking-threshold reduction in Appendix A are proved in the paper and are attributed to Meester and Znamenski as independent external inputs; they are used to convert avalanche sharpness into the stationary product limit, not to prove the sharpness itself. There are no fitted parameters renamed as predictions, no uniqueness theorem imported from the author's own prior work, and no self-citations at all: references [8], [9], and [10] are external prior results. The only structural input, the self-similar regeneration of avalanches, is a property of the Poisson construction rather than a restatement of the theorem. Accordingly, no circular step is present and the circularity score is 0.
Assumptions & free parameters
assumptions (4)
- standard math Standard probability tools: Markov process theory, regeneration, renewal-reward theorem, Paley-Zygmund inequality, Cesaro convergence.
- domain assumption Exponential coordinate transformation u -> -log(1-u) preserves order and maps uniform fitnesses to Exp(1).
- domain assumption Meester-Znamenski graphical representation and range-duration identity (adapted in Appendix A) are valid.
- standard math The regenerative limit theorem for finite-state Markov chains (Asmussen [1]) applies to the finite Bak-Sneppen chain.
Cite this review
Pith. "Pith review of Sharpness of the avalanche phase transition in the Bak--Sneppen model." pith.science (2026). https://pith.science/paper/VAT2JQFO
@misc{pith2026260810992,
author = {Pith},
title = {Pith review of: Sharpness of the avalanche phase transition in the Bak--Sneppen model},
year = {2026},
howpublished = {\url{https://pith.science/paper/VAT2JQFO}},
note = {Machine review of arXiv:2608.10992}
}
read the original abstract
We prove sharpness of the avalanche phase transition in the one-dimensional Bak--Sneppen model, for both the one-sided and isotropic update rules. In the one-sided model, the avalanche range is described by a nonlinear self-interacting jump process whose tails satisfy a triangular system. An extremal inequality for decreasing probability masses yields a polynomial lower bound on the range tail at the susceptibility threshold; this bound forces the appearance of infinite avalanches immediately above criticality and hence identifies the susceptibility and infinite-avalanche thresholds. A comparison of avalanche endpoints transfers this equality to the isotropic model. In both models, no infinite avalanche occurs at criticality, while the range has exponential moments at every strictly subcritical level. Combined with the range--duration and stationary-limit results of Meester and Znamenski, this identifies all three avalanche critical thresholds and proves the conjectured stationary product limit in both models.
Reference graph
Works this paper leans on
-
[1]
S. Asmussen,Applied Probability and Queues, second ed., Stochastic Modelling and Applied Probability, vol. 51, Springer-Verlag, New York, 2003
work page 2003
- [2]
-
[3]
A. Gillett, R. Meester and M. Nuyens,Bounds for avalanche critical values of the Bak–Sneppen model, Markov Process. Related Fields12(2006), no. 4, 679–694
work page 2006
-
[4]
A. J. Gillett, R. W. J. Meester and P. van der Wal,Maximal avalanches in the Bak– Sneppen model, J. Appl. Probab.43(2006), no. 3, 840–851
work page 2006
-
[5]
M. Grinfeld, P. A. Knight and A. R. Wade,Bak–Sneppen-type models and rank-driven processes, Phys. Rev. E84(2011), 041123
work page 2011
-
[6]
M. Grinfeld, P. A. Knight and A. R. Wade,Rank-driven Markov processes, J. Stat. Phys.146(2012), no. 2, 378–407
work page 2012
-
[7]
Maslov,Infinite series of exact equations in the Bak–Sneppen model of biological evolution, Phys
S. Maslov,Infinite series of exact equations in the Bak–Sneppen model of biological evolution, Phys. Rev. Lett.77(1996), 1182–1185
work page 1996
- [8]
Show all 12 references
-
[9]
Meester and D
R. Meester and D. Znamenski,Limit behavior of the Bak–Sneppen evolution model, Ann. Probab.31(2003), 1986–2002
2003
-
[10]
Meester and D
R. Meester and D. Znamenski,Critical thresholds and the limit distribution in the Bak–Sneppen model, Comm. Math. Phys.246(2004), 63–86
2004
-
[11]
Paczuski, S
M. Paczuski, S. Maslov and P. Bak,Avalanche dynamics in evolution, growth, and depinning models, Phys. Rev. E53(1996), 414–443
1996
-
[12]
Volkov,Rigorous upper bound for the discrete Bak–Sneppen model, J
S. Volkov,Rigorous upper bound for the discrete Bak–Sneppen model, J. Stat. Phys. 186(2022), article 8. Institut de math ´ematique d’Orsay, Universit ´e Paris-Saclay, 91405 Orsay, France Email address:olivier.henard@universite-paris-saclay.fr
2022
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.