Brownian Snake: Scaling, Genealogy & Applications

• Brownian snake is a path-valued Markov process that encodes both the genealogical structure and spatial evolution of branching systems.
• It emerges as the universal scaling limit of discrete branching trees with spatial displacements and underpins the construction of random metric spaces like the Brownian map.
• The process provides a framework for analyzing superprocesses, excursion theory, and high-dimensional potential theory through explicit measure formulations and martingale problems.

The Brownian snake is a fundamental object in probability theory and random geometry, describing a rich continuum limit of spatial branching structures. It is formulated as a path-valued Markov process encoding both the genealogical tree and the spatial evolution of a branching system, and arises as the universal scaling limit of discrete branching trees equipped with spatial displacements. The Brownian snake plays a central role in the study of superprocesses, random planar maps, and associated random metric spaces, exhibiting deep connections with Lévy trees, Brownian excursions, and super-Brownian motion.

1. Construction and Canonical Properties

The Brownian snake takes values in the space $\mathcal{W}$ of finite continuous paths $w:[0,\sigma(w)]\to\mathbb{R}^d$, where $\sigma(w) \ge 0$ denotes the lifetime. The process $(W_t)_{t\ge 0}$ is a strong Markov process on $\mathcal{W}$, whose lifetime process $(\zeta_t)_{t\ge 0},\ \zeta_t = \sigma(W_t)$, evolves as a reflecting Brownian motion on $[0,\infty)$ under the canonical excursion measure $\mathcal{N}_x$ for $x\in\mathbb{R}^d$ (Gall, 2014, Bai et al., 2024).

The evolution is characterized as follows:

• Growth: When the lifetime increases by $d\zeta>0$, an independent $w:[0,\sigma(w)]\to\mathbb{R}^d$0-dimensional Brownian path segment of length $w:[0,\sigma(w)]\to\mathbb{R}^d$1 is appended to the current path.
• Pruning: When lifetime decreases, a terminal segment of the path is erased, shortening the path accordingly.
• Tip process: Conditionally on $w:[0,\sigma(w)]\to\mathbb{R}^d$2, the tip $w:[0,\sigma(w)]\to\mathbb{R}^d$3 evolves as a time-changed Brownian motion.

A path-valued description encodes the genealogy: for every $w:[0,\sigma(w)]\to\mathbb{R}^d$4, $w:[0,\sigma(w)]\to\mathbb{R}^d$5 records the historical path from the root to the current tip of the genealogical tree as coded by the lifetime process.

2. Scaling Limits and Discrete Approximations

The Brownian snake arises as the functional scaling limit of discrete spatial trees. Consider size-conditioned critical Bienaymé–Galton–Watson trees $w:[0,\sigma(w)]\to\mathbb{R}^d$6 with i.i.d. edge displacements. Vertices are spatially labeled via sums of increments along the ancestral line, creating the discrete snake (Marzouk, 2018, Addario-Berry et al., 27 May 2025).

The main regime is as follows:

• Stable Branching: If the offspring distribution belongs to the domain of attraction of a stable law with index $w:[0,\sigma(w)]\to\mathbb{R}^d$7 and spatial increments $w:[0,\sigma(w)]\to\mathbb{R}^d$8 satisfy $w:[0,\sigma(w)]\to\mathbb{R}^d$9, after suitable scaling of the height and label processes, convergence holds in $\sigma(w) \ge 0$0:

$\sigma(w) \ge 0$1

where $\sigma(w) \ge 0$2 codes an $\sigma(w) \ge 0$3-stable Lévy tree and, given $\sigma(w) \ge 0$4, $\sigma(w) \ge 0$5 is a mean-zero Gaussian process with

$\sigma(w) \ge 0$6

This process $\sigma(w) \ge 0$7 may be termed the “Brownian-snake head” (Editor's term), i.e., Brownian motion indexed by the random tree coded by $\sigma(w) \ge 0$8 (Marzouk, 2018).

• Heavy-Tailed Displacements: If spatial increments have heavier tails, the rescaled label process converges instead to a “hairy snake”: a continuous Gaussian motion indexed by the tree, decorated with a Poisson cloud of macroscopic vertical jumps corresponding to rare large increments (Marzouk, 2018, Addario-Berry et al., 27 May 2025).
• Applications: Scaling limits for differences between height, Łukasiewicz, and looptree codes have been established for discrete snakes, with the limiting processes expressible in terms of the Brownian snake head (Addario-Berry et al., 27 May 2025).

3. Genealogy, Markov Properties, and Excursion Theory

The genealogical structure of the Brownian snake is governed by the height (lifetime) excursion, which codes a real tree via the pseudo-metric

$\sigma(w) \ge 0$9

The random tree $(W_t)_{t\ge 0}$0 results from collapsing points with $(W_t)_{t\ge 0}$1. Labels projected to tree points, $(W_t)_{t\ge 0}$2, propagate the spatial process along the genealogy.

Key properties:

• Markov Property: The process is Markovian in $(W_t)_{t\ge 0}$3 with infinite-dimensional state $(W_t)_{t\ge 0}$4.
• Excursion Decomposition: Cutting the lifetime process at zeros yields i.i.d. snake excursions.
• Conditional Gaussianity: Given the tree, the label process is a centered Gaussian field with variance given by genealogical distance.

The Brownian snake's excursion measure underpins profound decompositions, such as Poissonian decompositions of subtrees ("spine" decompositions) and Williams-type decompositions at the minimum, allowing detailed analysis of extremal paths via Bessel processes (dimension $(W_t)_{t\ge 0}$5) (Gall, 2014).

4. Brownian Snake in Random Geometry

The Brownian snake with Brownian excursion lifetime and spatial labels is the canonical object underlying the scaling limits of large planar maps:

• Brownian Sphere and Disk: The Brownian sphere, a random metric space homeomorphic to $(W_t)_{t\ge 0}$6 arising as the scaling limit of random quadrangulations, can be constructed via a labeled continuum random tree (CRT) with Brownian labels—the Brownian snake—through the continuum CVS bijection or “mating-of-trees” formalism (Angel et al., 18 Feb 2025, Gall, 2017).
• Metric Construction: The label process determines a random metric on the CRT via

$(W_t)_{t\ge 0}$7

and the induced quotient metric space is the Brownian map or disk, depending on boundary conditions.

• Boundary and Exit Measures: Excursion theory for the snake leads to explicit constructions of uniform boundary measures and decompositions of random surfaces into independent Brownian disks along metric nets (Gall, 2017).

5. Interactions with Superprocesses and Measure-Valued Dynamics

The Brownian snake provides a pathwise representation of super-Brownian motion and more general measure-valued branching processes, including those in random environments (Gall, 2014, Mytnik et al., 2011, Buckland et al., 2 Apr 2026).

• Historical Process: The snake structure encodes the full ancestral lineages, leading to detailed "historical" superprocesses.
• Snake Martingale Problems: The generator and martingale problem for the Brownian snake yields identification of the associated measure-valued process and its spatial-temporal properties, as in on/off super-Brownian motion with dormant and active phases (Buckland et al., 2 Apr 2026).
• Random Environment: In random (and possibly correlated) environments, the snake formalism extends to incorporate environmental covariances, influencing branching rates and yielding modified martingale characterizations tied to stochastic PDEs (Mytnik et al., 2011).

6. Brownian Snake Capacity and High-Dimensional Potential Theory

For $(W_t)_{t\ge 0}$8, the Brownian snake supports a non-linear capacity theory. The Brownian-snake capacity $(W_t)_{t\ge 0}$9 of a Borel set $\mathcal{W}$0 is defined by the scaling limit

$\mathcal{W}$1

where $\mathcal{W}$2 is the snake's spatial range. This capacity is homogeneous of degree $\mathcal{W}$3 and satisfies strong Choquet sub-additivity properties, with explicit comparisons to Riesz capacities (Bai et al., 2024). Discrete branching random walk capacities, under scaling, converge vaguely to this continuous Brownian-snake capacity, establishing a universality principle in high-dimensional branching systems.

7. Generalizations and Recent Developments

Extensions and variants of the Brownian snake include:

• On/Off Brownian Snake: Addition of a two-state Markov switch (active/dormant) modifies genealogical and spatial dynamics, leading to on/off super-Brownian motion and applications to dormant population models (Buckland et al., 2 Apr 2026).
• Heavy-Tailed and Hairy Snakes: In regimes with insufficient moment/tail decay, the standard Gaussian Brownian snake is replaced by Poisson-decorated discontinuous limits, with vertical “hairs” reflecting rare large spatial displacements (Marzouk, 2018, Addario-Berry et al., 27 May 2025).
• Inverse Bijections and Recoverability: The Brownian snake is a measurable invariant of the Brownian sphere, and can be reconstructed uniquely (up to orientation) from the metric structure and marked points on the sphere (Angel et al., 18 Feb 2025).

The Brownian snake thus functions as a universal scaling object interconnecting spatial branching, continuum random trees, random geometry, superprocesses, and non-linear potential theory.