Internal Diffusion Limited Aggregation with Critical Branching Random Walks

Basic Information

• Paper ID: 2510.13733
• Title: Internal Diffusion Limited Aggregation with Critical Branching Random Walks
• Authors: Amine Asselah, Vittoria Silvestri, Lorenzo Taggi
• Classification: math.PR (Probability Theory)
• Publication Date: October 16, 2025
• Paper Link: https://arxiv.org/abs/2510.13733

Abstract

Internal Diffusion Limited Aggregation (IDLA) is an interacting particle system describing random cluster growth, where growth is controlled by the boundary harmonic measure observed from interior points. This paper studies IDLA on $\mathbb{Z}^d$ driven by critical branching random walks. We prove that, unlike classical IDLA, this process exhibits a phase transition with respect to dimension. More precisely, we establish the existence of a shape theorem for spherical shapes when $d \geq 3$, and its absence when $d \leq 2$. Our bounds on the inner and outer worst-case deviations are polynomial in nature, which we expect to be characteristic of this model.

Research Background and Motivation

Research Problem

The core problem investigated in this paper is to replace simple random walks in classical IDLA with critical branching random walks, forming the branching IDLA (BIDLA) model, and to analyze the existence of its shape theorem.

Problem Significance

1. Limitations of Classical IDLA: Classical IDLA possesses a spherical shape theorem in all dimensions $d \geq 1$, lacking dimension-dependent phase transition phenomena
2. Complexity of Branching Processes: Branching random walks introduce particle creation and annihilation, making the aggregation growth mechanism more complex
3. Theoretical Significance: Reveals the critical role of dimension in random aggregation processes, providing new perspectives for understanding broader classes of random growth models

Limitations of Existing Methods

1. Failure of Classical Approaches: Traditional IDLA analysis methods (such as the Lawler-Bramson-Griffeath method) no longer apply in the branching case
2. Non-Conservation of Particle Number: Branching processes cause random variation in particle count, breaking key properties in classical analysis
3. Long-Range Correlations: Correlations between particles generated by branching increase analytical difficulty

Core Contributions

1. Established Dimensional Phase Transition for BIDLA: Proved the existence of a spherical shape theorem when $d \geq 3$ and its absence when $d \leq 2$
2. Introduced Random Barrier Growth (RBG) Auxiliary Process: Innovatively introduced a more controllable auxiliary process to bound BIDLA growth
3. Developed New Technical Methods:
   • Extended second moment estimates of local time to branching random walks
   • Established inner-outer bound estimation methods applicable to the branching case
4. Provided Polynomial Deviation Bounds: For $d > 2$, proved inner deviation $\delta_I(t) \leq t^{1/2+\epsilon}$ and outer deviation $\delta_O(t) \leq t^{1-1/(2d)+\epsilon}$

Methodology Details

Task Definition

Input: Initial aggregate $A(0) = \emptyset$, $A(1) = \{0\}$Process: At each time step $t$, release a critical branching random walk from the origin; stop when a particle reaches an empty site Output: Aggregate sequence $(A(t))_{t \geq 0}$, analyzing its asymptotic shape properties

Model Architecture

1. Branching Random Walk (BRW)

Let $T$ be a critical Bienaymé-Galton-Watson tree with offspring distribution $\nu$ satisfying:

• $\sum_{k=0}^{\infty} k\nu(k) = 1$ (criticality)
• $\sigma^2 = \sum_{k=0}^{\infty} (k-1)^2\nu(k) > 0$ (positive variance)
• $\exists \lambda > 0: \sum_{k=0}^{\infty} e^{\lambda k}\nu(k) < \infty$ (exponential moment condition)

2. BIDLA Stabilization Process

For particle configuration $\eta: \mathbb{Z}^d \to \mathbb{N}$:

• Unstable sites: $\eta(z) > 1$
• Topological operation: One particle dies, producing a random number of offspring, each offspring independently performs a simple random walk step

3. Random Barrier Growth (RBG)

Innovative auxiliary process with key features:

• Sedimentation on randomly selected spherical layers
• Layer distribution: $P(Z_i = h) = \frac{h^{d-1}}{\sum_{r=1}^H r^{d-1}}$
• Provides upper bounds for BIDLA growth

Technical Innovations

1. Second Moment Estimates of Local Time

For local time $\ell^x_{B_R}(z)$ of BRW in ball $B_R$: $E[(\ell^x_{B_R}(z))^2] \leq G_R(x,z) + \sigma^2 \sum_{y \in B_R} G_R(x,y) \cdot G_R^2(y,z)$

2. Dimension-Dependent Bound Estimates

• $d = 3$: $\sum_{y \in B_R} G_R(x,y)G_R^2(y,z) \leq C \frac{R-\|x\|+1}{\|x-z\|^3} \log\|x-z\|$
• $d > 3$: $\sum_{y \in B_R} G_R(x,y)G_R^2(y,z) \leq C \frac{R-\|x\|+1}{\|x-z\|^d}$

3. RBG Contraction Mechanism

By appropriately choosing shell width $H$, achieve contraction of expected particle number: $E[N_2] \leq N_1\left(J\frac{N_1}{H^d} + \gamma_H\right)$ where $\gamma_H = 1 - c/\log H$ (when $d=3$) or $\gamma_H = 1 - c$ (when $d>3$)

Experimental Setup

Theoretical Verification Methods

This paper is purely theoretical research, primarily verifying results through mathematical proofs:

1. Non-Shape Theorem for $d \leq 2$: Constructive proof showing positive probability that the aggregate becomes asymmetric
2. Shape Theorem for $d \geq 3$: Proved in three steps
   • Inner bound: Filling appropriately sized balls
   • Boundary particle control: Bounding the number of frozen particles
   • Outer bound: Controlling growth through RBG

Numerical Simulations

The paper provides numerical simulations of BIDLA aggregates on $\mathbb{Z}^2$ (Figure 1), showing four realizations at $t = 20000$, validating the theoretically predicted non-spherical behavior.

Experimental Results

Main Theorems

Theorem 1.1 (Main Result)

Let $(A(t))_{t \in \mathbb{N}}$ be the BIDLA process on $\mathbb{Z}^d$ satisfying assumption (H):

• $d = 1,2$: $\exists \epsilon > 0$ such that $P(A(t) \notin S_\epsilon \text{ occurs infinitely often}) = 1$
• $d \geq 3$: $\forall \epsilon > 0$, $P(A(t) \in S_\epsilon \text{ eventually holds}) = 1$

Theorem 1.2 (Deviation Bounds)

For $d > 2$, any $\epsilon > 0$: $P\left(\{\delta_I(t) > t^{1/2+\epsilon}\} \cup \{\delta_O(t) > t^{1-1/(2d)+\epsilon}\} \text{ occurs infinitely often in } t\right) = 0$

Key Lemma Results

Lemma 1.3 (Coverage Probability)

For $d > 2$, there exists a sufficiently large constant $\alpha$ such that for any $n$ and particle configuration $\eta$ supported on $B_{n/2}$ with $|\eta| \geq \alpha|B_n|$: $P(B_n \not\subseteq S(\eta)) \leq \exp\left(-C \frac{\alpha \cdot n}{\log n \cdot 1_{d=3} + 1_{d>3}}\right)$

Related Work

Classical IDLA Research

1. Lawler-Bramson-Griffeath (1992): Established the spherical shape theorem for classical IDLA
2. Asselah-Gaudillière, Jerison-Levine-Sheffield: Proved logarithmic deviation bounds

Related Random Growth Models

1. Diffusion Limited Aggregation (DLA): Driven by external harmonic measure, exhibiting fractal structure
2. Activated Random Walk (ARW): Particle system with sleep parameter
3. Oil-Water Model: Branching processes with random offspring distribution

Branching Random Walk Theory

Based on work by Asselah-Schapira and others on survival probabilities and local time estimates for critical branching random walks.

Conclusions and Discussion

Main Conclusions

1. Existence of Dimensional Phase Transition: BIDLA exhibits a sharp phase transition between dimensions 2 and 3, contrasting sharply with classical IDLA
2. Polynomial Deviations: In high dimensions, deviations grow polynomially rather than logarithmically as in classical IDLA
3. New Technical Methods: The RBG method provides new tools for analyzing complex random growth processes

Limitations

1. Complete Characterization for $d=2$: Only ruled out spherical shapes; did not fully determine whether other deterministic shapes exist
2. Subcritical Case: Does not address subcritical offspring distributions
3. Optimal Deviation Orders: Deviation bounds may not be optimal

Future Directions

1. Complete Theory for $d=2$: Determine whether non-spherical deterministic limit shapes exist
2. Continuous Limits: Seek continuous Markov dynamics for the $d=2$ case
3. Subcritical Branching: Study behavior under subcritical offspring distributions
4. Precise Fluctuations: Determine the exact order of deviations

In-Depth Evaluation

Strengths

1. Strong Theoretical Innovation: First to establish dimensional phase transition for IDLA-type models, with significant theoretical importance
2. Novel Technical Methods: The introduction of the RBG auxiliary process is highly creative, providing new ideas for related problems
3. Mathematical Rigor: Proofs are complete and rigorous with careful technical details
4. Profound Result Significance: Reveals the fundamental role of dimension in random aggregation

Weaknesses

1. Limited Application Background: As pure mathematical research, practical application scenarios are not sufficiently clear
2. Computational Complexity: Numerical verification and computational implementation of theoretical results are quite difficult
3. Numerous Open Problems: Leaves multiple important open questions; the theoretical framework remains incomplete

Impact

1. Academic Value: Makes important contributions to probability theory and stochastic processes
2. Methodological Significance: The RBG method may be applicable to other random growth models
3. Inspirational Value: Provides new perspectives for understanding the role of dimension in random systems

Applicable Scenarios

1. Theoretical Probability Research: Provides methods for studying other branching-driven random processes
2. Physical Modeling: Can be used to model diffusion aggregation processes with particle creation and annihilation
3. Mathematical Education: Serves as a classical example for demonstrating dimensional effects and phase transition phenomena

References

The paper cites 44 related references, primarily including:

• Lawler, Bramson, Griffeath (1992): Foundational work on classical IDLA
• Asselah, Gaudillière series: Important progress on IDLA deviation estimates
• Asselah, Schapira (2024): Theoretical foundations of branching random walks
• Rolla, Sidoravicius: Abelian property theory for activated random walks