PaperlayerView exact revision on arXiv

Source-linked AI summary

Recent advances in percolation theory and its applications

Abbas Ali Saberi

arXiv:1504.02898v2cond-mat.stat-mech

TL;DR

Percolation theory addresses how simple random occupancy rules generate phase transitions, scaling, and geometric structure across physical, technological, social, and natural systems. This review synthesizes ordinary and modified models, directed percolation, SLE and conformal invariance, magnetic-model connections, and landscape applications. It reports broad applications and specific critical descriptions, including Earth’s mean sea level as a singled-out critical level and watershed curves with κ = 1.734±0.005.

  • Problem

    Percolation theory seeks to characterize critical thresholds, scaling behavior, and variants of a simple clustering model across diverse systems, while some transition types and higher-dimensional formulations remain unresolved.

  • Method

    The review synthesizes ordinary and modified percolation, directed percolation, scaling theory, SLE, conformal invariance, magnetic-model mappings, and landscape applications.

  • Results

    Earth’s present mean sea level is singled out as a critical level in a percolation model of topography, while watershed curves are characterized by κ = 1.734±0.005.

  • Takeaways & Limitations

    Percolation provides a geometric framework spanning critical phenomena and applications from networks and landscapes to magnetic models and mathematical physics.

  • Takeaways & Limitations

    The transition type of explosive percolation in Euclidean space has not yet been clarified.

Abstract

from arXiv · show

Percolation is the simplest fundamental model in statistical mechanics that exhibits phase transitions signaled by the emergence of a giant connected component. Despite its very simple rules, percolation theory has successfully been applied to describe a large variety of natural, technological and social systems. Percolation models serve as important universality classes in critical phenomena characterized by a set of critical exponents which correspond to a rich fractal and scaling structure of their geometric features. In this review we will first outline the basic features of the ordinary model and take a glimpse at a number of selective variations and modifications of the original model. Directed percolation process will be also discussed as a prototype of systems displaying a nonequilibrium phase transition. After a short review on SLE, we will provide an overview on existence of the scaling limit and conformal invariance of the critical percolation. We will also establish a connection with the magnetic models. Recent applications of the percolation theory in natural and artificial landscapes are also reviewed.

1. Introduction

Percolation models describe the emergence of spanning clusters as critical phenomena with universal scaling, fractal structure, and connections to conformal, magnetic, and landscape models. The review surveys ordinary and modified percolation, SLE, critical Ising connections, and applications to natural and artificial landscapes.

  • Ordinary percolation: A spanning cluster emerges when occupation exceeds a critical threshold, producing the transition from an insulating to a metallic phase.In the circuit example, the spanning cluster connects opposite sides and lights the bulb.
  • Critical phenomena: Percolation criticality forms a universality class characterized by scaling laws and critical exponents largely independent of microscopic details.The model exhibits a sudden spanning-cluster onset and rich large-scale critical behavior.
  • Variants and open problems: Modified models include explosive percolation, whose transition is mathematically established as continuous on random graphs but remains unresolved in Euclidean space.The review presents this as an open issue concerning the transition type in Euclidean systems.
  • Scaling limits and conformal invariance: At criticality, two-dimensional percolation exhibits scale and conformal invariance, while SLE provides a framework for conformally invariant random curves.The review also discusses critical conformal field theories, including percolation with central charge c = 0 and the Ising model with c = 1/2.
  • Connections to magnetic models: The review connects percolation with magnetic models through cluster representations of Potts and Ising systems and discusses a two-dimensional mapping for 3D Ising criticality.The proposed mapping replaces difficult three-dimensional lattice-surface descriptions with immersed curves on a two-dimensional cross section.
  • Landscape applications: Percolation theory models landscape drainage, watersheds, lakes, and Earth’s topography through clustering and critical transitions.Reported applications include watershed fractality and a transition at present mean sea level accompanied by continental aggregation.

2. Basic properties of the percolation model

The basic percolation model is formulated through cluster observables that distinguish finite and spanning structures across subcritical, critical, and supercritical regimes. These observables describe cluster size, geometry, correlation range, and critical decay.

  • Model formulation: Percolation studies clustering of identical objects randomly distributed through space with occupation probability p.The basic formulation introduces geometric observables for an infinite system.
  • Percolation strength: The percolation strength P∞(p) is zero below pc and positive above pc, becoming nonanalytic near the threshold.It serves as an order parameter distinguishing the subcritical and supercritical phases.
  • Cluster-size observables: Mean cluster size increases toward the threshold, where the spanning cluster dominates, then decreases above pc after the infinite cluster is excluded.This definition focuses on finite clusters in the supercritical regime.
  • Cluster geometry: The radius of gyration measures cluster compactness and spatial extent, complementing cluster-size observables that lack structural information.Smaller radius of gyration indicates greater compactness and lower spatial extent.
  • Cluster distributions: At criticality, finite-cluster probabilities decay as a power law, while supercritical clusters are compact and have smoother-decaying size distributions.The critical decay is ws(pc) ≈ s^-1-1/δ, with δ = δ(d) > 0.
  • Correlation length: The correlation length ξ characterizes the distance over which two points belong to the same finite cluster and bounds the self-similar scaling region.The associated two-point correlation function typically decays exponentially with ξ.

2.1. Percolation in d-dimensions

Percolation on lattices and networks is governed by critical thresholds separating finite-cluster and giant-component regimes. The review describes threshold behavior and structural properties for hypercubic lattices, trees, and random graphs.

  • For d = 1, the critical threshold is pc = 1, whereas for d > 1 it satisfies 0 < pc < 1.
  • On periodic graphs, the number of infinite clusters can be 0, 1, or ∞, while trees can support infinitely many coexisting infinite clusters.
  • The Bethe lattice has coordination number z, no closed loops, and critical threshold pc = 1/(z −1).At pc, the probability of an infinite cluster is zero; above pc it is strictly increasing.
  • On the Bethe lattice, the correlation function is gc(k) = pk and the correlation length remains finite for every 0 < p < 1.The correlation length scales as ξ(p) ∼ −1/ln p.
  • In ERn(p), λc = 1 separates logarithmic components from a critical O(n2/3) largest component and a supercritical unique giant component of O(n).For λ > 1, the giant component contains a positive fraction of vertices.

2.2. Percolation at and near criticality

Near criticality, percolation is described through cluster-size scaling, critical exponents, and renormalization-group transformations. The review emphasizes universality while noting limits of real-space renormalization.

  • Below pc, clusters are finite with exponentially decaying size distributions; above pc, an infinite cluster coexists with finite clusters having slower-than-exponential tails.
  • Scaling hypotheses express the finite-cluster density through a scaling function and critical exponents such as τ, σ, β, γ, α, ∆, and ν.Cluster-size moments are defined as Mm(p) = ∑s smns(p).
  • At pc, the two-site connectivity probability decays as gc(r) ≃ r^(2−d−η), introducing the anomalous dimension η.
  • Critical exponents satisfy scaling and hyperscaling relations, with hyperscaling believed valid only below the upper critical dimension dc = 6.The mean-field exponent values α = −1, β = 1, γ = 1, and τ = 5 imply dc = 6.
  • Real-space renormalization coarse-grains and rescales blocks, identifying pc with a nontrivial fixed point of the transformation.It gives limited results for critical exponents except on the Bethe lattice.

2.3. Fractal structure of the critical percolation clusters

Critical percolation clusters exhibit scale-dependent fractal geometry governed by correlation length and mass-scaling laws. Their characterization requires multiple dimensions, including shortest paths, chemical structure, and perimeters.

  • At criticality, the correlation length diverges and cluster mass scales as a power of observation-window size, producing self-similar geometry.For scales larger than ξ above pc, the infinite cluster becomes effectively homogeneous.
  • The cluster fractal dimension obeys dc_f = d − β/ν and is universal because β and ν are universal.Exact values are known in 2D and for d ≥ 6, while other dimensions rely on numerical estimates.
  • The shortest-path length scales as lmin ∼ R^dmin_f, while chemical dimension is defined through M ∼ l^dch.The shortest-path fractal dimension is known exactly for d ≥ 6 but not in 2D.
  • In 3D, dmin_f distinguishes percolation clusters from DLA, with approximate values 1.38 and 1, respectively.
  • In 2D, perimeters are conformally invariant SLE curves with κ = 6 and ˜κ = 8/3, related by the duality relation κ˜κ = 16.
  • For scale-free networks with 3 < τ < 4, the spanning-cluster fractal dimension is dc_f = 2(τ −2)/(τ −3).For τ > 4, the result agrees with infinite-dimensional regular percolation, where dc_f = 2 and dch = 2.

3. Variants of percolation

Percolation has been extended beyond independent Bernoulli occupation through models incorporating correlations, anisotropy, nonlocality, and altered cluster-merging rules. These variants also connect percolation with lattice statistical models.

  • The Fortuin–Kasteleyn construction connects bond percolation to lattice statistical models and formulates it as a limiting case of the Potts model.
  • Other percolation variations address spatial correlations, anisotropy, nonlocality, and explosivity.Explosive percolation modifies cluster-merging rules to seek a transition from continuous to discontinuous behavior.

3.1. Explosive percolation

Explosive percolation modifies edge-selection or bond-occupation rules to alter how giant or spanning clusters emerge. The review contrasts rigorously continuous Achlioptas transitions with discontinuous behavior in half-restricted and, conditionally, spanning cluster-avoiding processes.

  • The Erdős–Rényi random graph has a continuous transition at critical time t_c = 1/2 in the fraction of vertices belonging to its largest cluster.
  • Achlioptas process: Achlioptas processes impose nonrandom edge-selection rules intended to delay or accelerate formation of a large percolating cluster.
  • Achlioptas process: The largest cluster can grow from at most √n to at least n/2 within at most 2n^2/3 steps, but the transition is rigorously continuous rather than discontinuous.
  • Half-restricted process: The half-restricted process forms a restricted set from the ⌊fn⌋ vertices in the smallest components and exhibits a discontinuous transition for every f < 1.
  • Spanning cluster-avoiding process: In the spanning cluster-avoiding model, the transition can be continuous or discontinuous for d < d_c = 6 depending on the number m of potential bonds.
  • Earlier acceptance methods also suppressed largest-cluster formation, while cluster perimeters at threshold had fractal dimension 1.23 ± 0.03.

3.2. Non-self-averaging percolation

Fractional percolation models crackling-noise-like growth through stochastic jumps in the largest component. Its order parameter remains non-self-averaging because infinitely many discontinuities persist near the first transition.

  • Fractional percolation uses a fractional growth rule to model crackling noise in slowly driven systems such as crumpled paper, earthquakes, and magnetization.
  • After the first transition, the largest component either stays constant or increases by at least the fraction r = f/(1 + f) as n →∞.
  • The order parameter increases through infinitely many discontinuous transitions arbitrarily close to the first dynamical transition point p_c.
  • The kth staircase step scales as δ_k ∼ (1 + r)^−k, producing a jump-size distribution d_s ∼ s^−1.
  • Jump sizes and transition points remain stochastic in the thermodynamic limit, so the process is non-self-averaging.

3.3. Correlated percolation

Correlations modify percolation criticality according to their spatial range. Short-range correlations leave critical behavior unchanged, whereas long-range correlations alter critical exponents through the correlation decay.

  • Short-range correlations do not change critical behavior because percolation satisfies dν − 2 = −α > 0.
  • For long-range correlations g(r) ∼ r^−2H with 2H < d, the extended Harris criterion gives ν_H = 1/H when Hν^−1 < 0.
  • Power-law correlations in self-affine surfaces make the transition critical only for H = 0 in the thermodynamic limit.
  • For two-dimensional Euclidean lattices, ν_H follows 1/H for 0 < H < 1/ν and ν for H ≥ 1/ν, with ν = 4/3.
  • Numerical results find H-dependent fractal dimensions for clusters, external perimeters, shortest paths, backbones, and red sites.
  • The hyperscaling and duality relations are numerically supported, but theoretical verification remains lacking.

3.4. Bootstrap percolation

Bootstrap percolation models activation dynamics in highly coupled systems and has been studied across lattices, random graphs, and social-network settings. Its thresholds and finite-size behavior connect to metastability and k-core structure.

  • Model and applications: Bootstrap percolation describes systems whose elements activate according to the states of their close neighbors.Applications include crack propagation, neuronal activity, and magnetic systems.
  • Thresholds and metastability: Sharp metastability thresholds have been proved in two-dimensional lattices and generalized to arbitrary dimensions.The finite-size behavior of bootstrap percolation is also known as metastability.
  • Thresholds and metastability: On the infinite lattice Z^d, pc(Z^d, k) = 0 if k ≤ d and pc(Z^d, k) = 1 otherwise.This result gives an all-or-nothing threshold classification by dimension and activation parameter.
  • Finite-size behavior: Simulations estimated pc(L, d = 2, k = 2) ln L = 0.245 ± 0.015, differing from the rigorous value π^2/18 = 0.548311 · · · because convergence is very slow.The discrepancy was rigorously attributed to the asymptotic limit L →∞.
  • Networks and k-core relations: On power-law random graphs, bootstrap percolation models the spread of ideas or trends, while its relation to k-core percolation is close but not identical.Bootstrap percolation is an infection process beginning from source nodes, whereas the k-core is a maximal subgraph whose nodes have at least k neighbors within it.

3.5. Directed percolation

Directed percolation introduces a preferred spatial or temporal direction to model anisotropic spreading and nonequilibrium phase transitions. The review describes its dynamics, scaling laws, critical exponents, and proposed universality criterion.

  • Motivation and scope: Directed percolation models anisotropic propagation, such as water penetration through porous rock, where gravity selects a preferred direction.It is a nonequilibrium universality class with applications to contact processes, epidemics without immunization, and forest fires.
  • Dynamics: Adding temporal freedom makes directed percolation a dynamical process in d + 1 dimensions, and dynamic-network percolation can be mapped to directed percolation in infinite dimensions.This mapping follows from treating time as an additional degree of freedom.
  • Dynamics: In directed bond percolation, a source active vertex at t = 0 propagates activity row by row through bonds occupied with probability p until reaching an absorbing state.The vertical direction represents time, and active sites are those connected to the source by directed paths.
  • Scaling and criticality: The stationary active-site density near criticality follows ρs ∼ (p−pc)^β, with β a universal exponent determined by dimensionality.The density serves as the order parameter of the spreading process.
  • Scaling and criticality: The scaling regime relates correlation lengths through ξ∥ ∼ ξ⊥^z, where z = ν∥/ν⊥; for d ≥ 4, mean-field values are β = 1, ν∥ = 1, and ν⊥ = 1/2.For d < 4, exact critical exponents and thresholds are unavailable, but precise estimates exist for several dimensions and lattices.
  • Universality: The Janssen–Grassberger conjecture assigns models to the directed-percolation universality class when they have a continuous transition, a unique absorbing state, and a positive one-component order parameter.The supplied passage begins a broader list of stated conditions for this conjecture.

4. Percolation in two dimensions

Two-dimensional critical percolation develops scale- and conformally invariant structures that can be analyzed through SLE and related exact methods. Its geometric clusters connect to magnetic critical behavior through the Fortuin–Kasteleyn representation.

  • Scaling and conformal invariance: At the critical point, the correlation length diverges and scale invariance emerges, while scaling alone does not determine all critical exponents.Conformal invariance provides a stronger symmetry used to obtain exact two-dimensional results.
  • Scaling and conformal invariance: Cardy’s crossing probability depends on conformal geometry through a cross-ratio and is invariant under conformal transformations at criticality.Away from the critical point, its scaling-limit value becomes 0 below pc and 1 above pc.
  • Stochastic Loewner evolution: SLEκ describes scaling limits of interfaces in two-dimensional critical lattice models; Markov and stationary-increment properties lead to Brownian driving with diffusivity κ.Percolation corresponds to κ = 6, whose cluster boundaries are described by SLE6.
  • Stochastic Loewner evolution: For SLE, curve roughness increases with κ, with fractal dimension df = 1 + κ/8 for κ ≤8 and space filling for κ ≥8.The curves are dilute for 0 ≤κ ≤4 and dense for κ > 4.
  • Scaling and conformal invariance: Smirnov proved that triangular-lattice site percolation has a conformally invariant limiting crossing probability satisfying Cardy’s formula.This established existence and conformal invariance of the limit for that model.
  • Percolation and magnetic models: The Fortuin–Kasteleyn mapping represents Potts-model configurations as occupied-bond clusters, and the limit q →1 reproduces percolation.For the Ising model, FK cluster observables correspond to magnetic quantities such as correlation length and susceptibility.

5. Percolation description of landscapes

Percolation theory is used to model fractal and scale-invariant features of natural and artificial landscapes, including coastlines, watersheds, and global topography. These applications connect landscape geometry with critical transitions, correlations, and conformal invariance.

  • Global topography: The Earth’s topography has a power spectrum S(k) ∼ k^-2 across a wide range of scales, a scaling relation also found in bathymetry, rock surfaces, and Venusian topography.
  • Percolation models describe statistical properties of natural landscapes, including global topography, coastlines, river basins, drainage networks, and watersheds.
  • Fractal geometry of coastlines: Coastline models combining rapid mechanical erosion and slow chemical weakening reach a stationary fractal geometry with dimension close to 4/3, matching observed coasts and critical percolation boundaries.
  • Fractal geometry of coastlines: Long-range lithological correlations produce a critical sea force fc and make coastline fractal dimension depend on the Hurst exponent; below fc erosion stops, whereas above fc it continues.
  • Statistical properties of watersheds: Watershed fractal dimensions are df = 1.2168 ± 0.0005 in 2D and 2.487 ± 0.003 in 3D for uncorrelated artificial landscapes, while natural 2D values are 1.10 ≤ df ≤ 1.15.
  • Statistical properties of watersheds: Watershed lines show conformal-invariant behavior compatible with SLEκ, with κ = 1.734 ± 0.005 in the scaling limit.
  • Global topography: Global topography exhibits a percolation transition at a critical level coinciding with Earth’s present mean sea level, where most landmass joins together.

6. Conclusions

The review synthesizes percolation’s broad scientific reach and its simple rules, rich critical geometry, and combination of physical conjectures with rigorous mathematics. It emphasizes emergent long-range correlated structure and robust universal scaling.

  • Percolation theory spans social and network sciences, string theory, particle physics, and mathematical probability.
  • Simple percolation rules can generate long-ranged correlated geometry despite an uncorrelated microscopic structure.
  • Critical percolation has fractal structure and scaling laws characterized by universal critical exponents.
  • Percolation research combines exact conjectures motivated by physical insight with rigorous mathematical proofs.
  • Percolation theory is robust against small perturbations.
Loading 1504.02898v2…