First-Passage Percolation

• First-passage percolation is a probabilistic model assigning i.i.d. nonnegative passage times to edges to study optimal paths and growth processes.
• The shape theorem establishes a deterministic limit shape that quantifies the anisotropic large-scale geometry of the infected region.
• Recent research explores fluctuation exponents, geodesic structure, and competition models, presenting challenging open problems in probability theory.

First-passage percolation (FPP) is a paradigmatic probabilistic growth model on graphs, originally introduced to model the spread of a fluid (or infection) through a disordered medium. The standard setting involves assigning i.i.d. nonnegative random variables ("passage times") to the edges of a lattice or a more general graph and studying induced random metrics, growth processes, geodesics, and fluctuation properties. Over the last several decades, FPP has become a central object in probability theory, with deep connections to statistical mechanics, random geometry, and combinatorial optimization.

1. Model Definition and Core Properties

FPP is most commonly defined on the $d$-dimensional cubic lattice $(\mathbb{Z}^d, \mathcal{E}^d)$. Each edge $e$ receives an i.i.d. nonnegative random variable $\tau_e$ (the "passage time") with common distribution $F$. For any finite path $\Gamma$, the passage time is $T(\Gamma) = \sum_{e\in\Gamma} \tau_e$, and for vertices $x,y$, the geodesic (optimal) passage time is $T(x,y) = \inf_\Gamma T(\Gamma)$ where the infimum is over nearest-neighbor paths connecting $x$ to $y$.

A central quantity is the time constant,

$\mu(x) = \lim_{n \to \infty}\frac{T(0, n x)}{n},$

which is proven to exist almost surely and in $L^1$ under suitable ergodicity and integrability assumptions (Auffinger et al., 2015, Blair-Stahn, 2010). The time constant typically defines a norm on $\mathbb{R}^d$, dictating large-scale anisotropic growth.

2. Shape Theorems and Asymptotic Geometry

A foundational result is the shape theorem: Under mild moment conditions (usually $F(0) < p_c(d)$ and $$\mathbb{E}\min\{\tau_1^d,\dots,\tau_{2d}^d\}<\infty$$), the infected (or wetted) region

$B(t) = \{x \in \mathbb{Z}^d : T(0, x) \le t\}$

satisfies

$$(1-\varepsilon) \mathcal{B} \subset t^{-1} B(t) \subset (1+\varepsilon)\mathcal{B}, \quad \text{for large } t,$$

where $$\mathcal{B} = \{x \in \mathbb{R}^d : \mu(x) \le 1\}$$ is the deterministic limit shape (Auffinger et al., 2015, Blair-Stahn, 2010, Alm et al., 2014). The precise geometry of $\mathcal{B}$—its convexity, regularity, and symmetry—is a deep open question, e.g., strict convexity (uniform curvature) remains largely unproven.

In Euclidean generalizations or on random tessellations, analogous shape theorems hold for ergodic random pseudometrics, with the limit shape inherited from the invariant norm defined by the time constant (Ziesche, 2016).

3. Fluctuations, Universality, and Scaling Exponents

Beyond the deterministic limit, FPP exhibits rich fluctuation phenomena. Variance bounds for the point-to-point passage time $T(0,nx)$ are central (Auffinger et al., 2015, 1901.10325). The sharpest known result in typical lattice models is

$\mathrm{Var}\, T(0, ne_1) = O(n / \log n)$

for a wide class of distributions (1901.10325). Lower bounds (in $d=2$) are at least logarithmic (Auffinger et al., 2015). The expected universal scaling exponents in $d=2$ are KPZ-type: longitudinal exponent $\chi=1/3$, transversal (wandering) exponent $\xi=2/3$; proofs are only available in integrable exactly-solvable directed models.

Extreme disorder leads to phase transitions between universality classes. For large variability, passage times in FPP balls initially mimic critical bond-percolation clusters (fractal, percolation exponents), with a crossover to KPZ scaling at a disorder-dependent characteristic length (Villarrubia et al., 2019).

4. Geodesic Structure: Directionality, Coalescence, and Busemann Functions

Finite and infinite geodesics—paths achieving the optimal passage time—encode the geometry and competition in the model. For continuous distributions, finite geodesics are almost surely unique. The central questions for infinite geodesics include:

• Existence and uniqueness in directions: In $d=2$ and under strong regularity of the limit shape, for almost every direction, there exists a unique infinite geodesic (Blair-Stahn, 2010, Auffinger et al., 2015, Ahlberg, 2020).
• Coalescence: Geodesics with the same asymptotic direction eventually merge (Auffinger et al., 2015, Ahlberg, 2020).
• Busemann functions: These additive functionals $$B_g(x,y) = \lim_{k \to \infty} [T(x, v_k) - T(y, v_k)]$$ (for $g = (v_1, v_2, \ldots)$ a geodesic) are essential for analyzing geodesic structure and are closely tied to exposed (tangent) points of the limit shape (Auffinger et al., 2015, Ahlberg, 2020).

In hyperbolic groups, analogous results hold with additional ergodic and geometric tools, leading to almost sure coalescence and linear variance growth in all boundary directions (Basu et al., 2019).

5. Variations and Generalizations

A diversity of FPP variants has been rigorously explored:

• Correlated weights: Under long-range dependencies (e.g., from Gaussian free field, random interlacements), shape theorems and positive time constants still hold under decoupling and ergodicity (sprinkling) conditions (Andres et al., 2021).
• Random environments: On random graphs (e.g., random triangulations, configuration models), FPP distances typically concentrate, often at novel scaling exponents (e.g., $n^{1/4}$ for planar maps) and with universality classes distinct from the lattice (Stufler, 2022, Bhamidi et al., 2024).
• Inhomogeneous models: Assigning independent but non-identically distributed weights (e.g., per half-plane) leads to surprising phenomena such as enhanced vertical growth and "pyramid" defects in the limit shape (Ahlberg et al., 2013).
• Critical and infinite parameters: At criticality ($F(0)=p_c$ in $\mathbb{Z}^2$), passage times grow logarithmically in distance, with limit (central limit theorem) behaviors tied to near-critical percolation, invasion percolation, and sharp "min-summability" criteria (Damron et al., 2015, Bhamidi et al., 2024).
• Edge or face-based models: FPP has been formulated on boundaries of tessellations, in the Euclidean continuum, and in the Brochette model where passage times have strong linewise dependence (Ziesche, 2016, Marivain, 20 Jan 2025).
• Competition and recovery: Extensions include multi-type Richardson models (competition), first-passage percolation with recovery (where vertices can revert to inactive states), first-contact percolation (random contact time sets), and variants with escape strategies (Ahlberg, 2020, Candellero et al., 2024, Jahnel et al., 2024, Andjel et al., 2012).

6. Competition, Growth Models, and Open Problems

FPP is the backbone for stochastic growth and competition models:

• Richardson and Eden models model competitive spread with the geometry and number of coexisting types linked to the existence of disjoint infinite geodesics and the number of exposed points of the limit shape (Auffinger et al., 2015, Blair-Stahn, 2010, Ahlberg, 2020).
• Multi-type coexistence: In $d=2$, the duality between coexistence of $k$ types and the existence of $k$ disjoint geodesics is sharp for generic continuous distributions (Ahlberg, 2020).
• Open problems span strict convexity/uniform curvature of the limit shape, fluctuations exponents, universality class characterizations (especially in higher dimensions), structure and coalescence of infinite geodesics, exact characterization of geodesic fluctuation exponents, and properties of competition interfaces (Auffinger et al., 2015).

7. Current Directions and Impact

Recent years have seen advances in sublinear variance bounds, explicit shape descriptions in integrable and non-integrable models, geometry/topology of FPP balls (number and size of "holes" scales as $t^{d-1}$ and $\log t$ respectively for non-deterministic passage times; (Damron et al., 2022)), the exploration of disorder universality, and applications to random conductance models and random Schrödinger operators (Andres et al., 2021).

FPP remains a central model interfacing with percolation, random network optimization, kinetic growth, integrable systems, and interacting particle systems. Its open conjectures—particularly on fluctuation exponents, the shape theorem, and geodesic structure—continue to motivate active research at the interface of probability, mathematical physics, and combinatorics.