On the critical value of void percolation in high dimensions
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.
Kai: We're looking at the paper: "On the critical value of void percolation in high dimensions".
Mira: Here's the gist of the paper. On the critical value of void percolation in high dimensions, this paper proves that the dimensional scaling of this threshold is linear, specifically showing that for sufficiently large dimensions d, Φp(d) = Θ(d), which excludes a logarithmic growth suggested by analogies with other systems.
Kai: Main Findings The central result establishes that the void percolation threshold in the random Lorentz gas scales linearly with the dimension, meaning Φp(d) = Θ(d).
Mira: Specifically, the authors prove that for all sufficiently large d, there exist universal constants c and C such that cd ≤ Φp(d) ≤ Cd.
Kai: This linear scaling is consistent with numerical estimates of Φp/d → 3.40(5) and the mean-field dynamical transition scaling of Φd ≃ 2.40 d, suggesting that the activated window
Φd, Φp: remains a bounded correction to mean-field arrest in high dimensions.
Paper summary: Mira: So to wrap up on this paper, "On the critical value of void percolation in high dimensions," the authors have established a linear scaling for the void percolation threshold, saying it is equal to theta of d >
Kai: They prove that for all sufficiently large d, there are universal constants c and C such that c times d is less than or equal to Φp of d which is less than or equal to C times d >
Lev: The real implication here is ruling out the logarithmic growth they saw in other systems, which excludes the d log d scaling suggested by analogies in the random f paper >
Kai: In simpler terms, they are saying that for high dimensions, when you look at void percolation, you don't see that extra logarithmic factor creeping in >
Mira: The paper shows that even though the transitions are different—one continuous and one discontinuous—the approach to this void percolation threshold is governed by these linear scaling laws >
Lev: This means the window between Φd and Φp is expected to control how close we get to arrest in sufficiently high dimensions, suggesting cage escapes above Φd exist but only up to that finite density of about three point four >
Kai: It confirms that the numerical results they have found cannot distinguish between cubic scaling in one over d or saturating d log d scaling, confirming Φp equals theta of d >
Conclusion: Kai: So we've been looking at this paper, "On the critical value of void percolation in high dimensions," and what they found is that for big dimensions, there's a very specific way this void percolation threshold behaves.
Mira: It seems to focus on connecting two different kinds of transitions here—one continuous and one discontinuous—and how the approach to that void percolation point is governed by these linear scaling laws.
Lev: The main thing they prove is that for dimensions large enough, the threshold phi scales linearly with the dimension d, meaning you get phi proportional to d.
Kai: That's what I mean when I say it excludes the logarithmic growth we often see in other systems, which is what really makes this specific result stand out.
Mira: It suggests that even though there are different physical natures to the transitions, the way they approach arrest in these high-dimensional random Lorentz gas systems follows a very structured rule dictated by this linear scaling.
Lev: If this holds up, it means that those activated escape paths you're looking at above some threshold phi still exist in high dimensions, but they only persist up to a certain density, around three point four.
Kai: It's interesting because the authors themselves can't really tell if it’s cubic scaling in one over d or something that saturates the logarithmic growth; it just confirms it’s linear scaling.
Mira: The implication is that this window between the dynamical transition and this percolation threshold should control how close you get to arrest when you're dealing with sufficiently high dimensions.
Lev: So, we've established the scaling for void percolation, but now we need to see how those specific numbers compare to what we can actually build and measure in our hardware experiments.
Department of Physics, Duke University · Department of Chemistry, Duke University · Mathematics Department, Northwestern University
cond-mat.dis-nn, cond-mat.stat-mech, math.PR
Submitted: 2026-11-04
Updated: 2026-11-04
Comments: 20 pages, 4 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 81/100
The gist: On the critical value of void percolation in high dimensions, this paper proves that the dimensional scaling of this threshold is linear…
Key concepts
- Void Percolation Threshold ($Φ_p$)
- This is the critical density of empty spaces (voids) in a system where particles are randomly placed. Finding this threshold tells researchers when an open path of voids becomes possible, which is linked to activated escape routes in the physical system being modeled.
- Random Lorentz Gas (RLG)
- The RLG is a model used to test how systems behave under random interactions, specifically related to the random first-order transition scenario. It serves as a testbed for understanding how activated processes lead to arrest or flow in high-dimensional systems.
- Mean-Field Dynamical Transition ($Φ_d$)
- This refers to a threshold ($Φ_d \simeq 2.40 d$) related to the dynamical transition in the system. The paper compares this with the void percolation threshold, showing they are different physical phenomena, but both relate to how systems behave as dimension $d$ increases.
- Linear Scaling ($Φ_p(d) = Ψ(d)$)
- The central result is that the void percolation threshold scales linearly with the dimension. This means that for large dimensions, the threshold grows proportionally to $d$, which is a specific mathematical relationship that distinguishes this finding from other expected logarithmic growth patterns.
Terminology
Summary
On the critical value of void percolation in high dimensions, this paper proves that the dimensional scaling of this threshold is linear, specifically showing that for sufficiently large dimensions d, Φp(d) = Θ(d), which excludes a logarithmic growth suggested by analogies with other systems. p. 0
Main Findings
The central result establishes that the void percolation threshold in the random Lorentz gas scales linearly with the dimension, meaning Φp(d) = Θ(d). Specifically, the authors prove that for all sufficiently large d, there exist universal constants c and C such that cd ≤ Φp(d) ≤ Cd. p. 2 This linear scaling is consistent with numerical estimates of Φp/d → 3. p. 040(5) and the mean-field dynamical transition scaling of Φd ≃ 2. p. 040 d, suggesting that the activated window [Φd, Φp] remains a bounded correction to mean-field arrest in high dimensions. p. 0
Context and Motivation
The study investigates the connection between the random Lorentz gas (RLG) and void percolation as a testbed for understanding activated processes in systems described by the random first-order transition (RFOT) scenario. The paper contrasts the dynamical transition threshold, Φd ≃ 2. p. 040 d, with the void percolation threshold, Φp, noting that while mean-field theory suggests they might coincide as d → ∞, they are of different physical natures (continuous versus discontinuous transitions). p. 0 The authors propose that the approach to the void percolation threshold is associated with activated escape paths that lace together imperfect glass-like cages. p. 0
Upper Bound Proof
The upper bound, Theorem 1, is proven by showing that for sufficiently large d and Φ = M d, the probability of an open path from 0 to distancel/√hd is zero. This involves bounding the probability P(Fx) using Lemma 4, which shows P(Fx) ≤ e−c0Φ. p. 4 The proof utilizes a construction of tree embeddings and a net for tree embeddings to bound the probability that an open curve connects 0 to ∂B(0, A4k). p. 4
Lower Bound Proof
The lower bound, Proposition 11, is established using a second-moment argument on candidate paths. The authors define a weighted count of open paths in Γk by Zk and show that EZk = 1. p. 6 By applying the Paley–Zygmund inequality, they conclude that θd(Φ) ≥ 2−d. p. 7 This lower bound is achieved by bounding the second moment using Lemma 33 and Lemma 24. p. 6
Physical Implications
The linear scaling result excludes the d ln d growth suggested by analogies with other RFOT systems. The established window [Φd, Φp] is expected to control the approach to arrest in sufficiently high dimensions. p. 8 This finding suggests that cage escapes above Φd do exist in sufficiently high d but persist only up to a finite density of order 3. p. 84. The results motivate further study of instantons in the RLG and broader RFOT models. p. 8
Appendix Details
Appendix A provides explicit constants for the bounds, showing how to adapt the proof to yield Φp(d)/d ∈ [1/2, 5] for all sufficiently large d. It details the complex machinery used in proving Lemma 31 and Lemma 32 concerning the convergence of kernels Kj,d. p. 4 The appendix also verifies numerical inequalities using programs like simple. p. 2py. The final proof for the lower bound relies on bounding the second moment via Lemma 34 and Lemma 33. p. 0 The paper concludes by noting that the numerical results cannot distinguish between cubic in 1/d scaling and saturating d ln d scaling, confirming Φp = Θ(d). p. 9
References
[5] G. Biroli, P. Charbonneau, Y. Hu, H. Ikeda, G. Szamel, and F. Zamponi.
[44] M. D. Penrose.
[35] J.-Kertesz.
[38] T. Bauer, F. H¨ofling, T. Munk, E. Frey, and T. Franosch.
[43] V. Anantharam and F. p. 2 Baccelli.
[45] J.-B Gou´er´e.
[48] J. G. Wendel.
[53] OpenAI, Ten advances in mathematics and theoretical computer science. p. 9
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved analysis of high-dimensional percolation thresholds allows AI systems to accurately predict whether void percolation is continuous or discontinuous in high dimensions by distinguishing between the mean-field dynamical transition and the void percolation threshold.
-
Enhanced ability to model activated processes in random first-order transition universality class systems by using the proven linear scaling of Φp(d) = Θ(d) to establish a bounded correction window [Φd, Φp] to mean-field arrest.
-
Development of instanton models for the random Lorentz gas (RLG) enables AI systems to characterize activated escape paths as
lace together imperfect glass-like cages
by utilizing instantonic large-deviation forms like Pescape(t) ∼ g(te−df(Φ/d)). -
Precise determination of dimensional scaling for void percolation allows AI systems to rule out the d ln d growth, confirming that the upper bound on escape density is confined to densities of order d rather than diverging as d increases.
-
Implementation of tree embedding constructions and related probabilistic methods allows AI systems to rigorously prove lower bounds for the void percolation threshold by showing that for Φ = αd, there is a positive probability that 0 lies in an unbounded component in V, leading to θd(Φ) ≥ 2-d.
Sources
- A new lower bound for sphere packing
- High-dimensional theory of the glass transition revisited: hopping and local defects
Related papers
- Topological metal-insulator transitions in one-dimensional non-Hermitian quasicrystals: beyond PT-symmetry
- Microscopic and collective signatures of feature learning in neural networks
- Synchronous and asynchronous Monte Carlo dynamics of mean-field spin glasses with rotationally invariant couplings: what they share and where they differ
- Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain
- Beyond Ground States: Learning Many-Body Hamiltonians for p-Bit Invertible Arithmetic
- Linear equivalence of nonlinear recurrent neural networks