Snapping Out Brownian Motion (SOBM)

• Snapping Out Brownian Motion (SOBM) is a family of strong Markov processes that model diffusion across semi‐permeable interfaces using reflected Brownian excursions and exponential local time clocks.
• Its generator acts as the Laplacian on disjoint subdomains with complex interface conditions, while the Dirichlet form encapsulates transmission properties through explicit snapping parameters.
• SOBM underpins analytical and numerical methods for interface problems with applications in mathematical physics, biology, chemistry, and networked stochastic processes.

Snapping Out Brownian Motion (SOBM) is a family of strong Markov processes that model diffusion across interfaces exhibiting partial permeability, characterized by stochastic mechanisms for interface crossing, reflection, and possible sticking or killing. SOBM realizes semi-permeable barriers by piecing together reflected Brownian excursions on disjoint subdomains; at interface local times determined by exponential clocks, the particle “snaps out” (jumps) either across the interface, into a trapping cemetery state, or resumes on the same side, depending on parameterization. SOBM is central in the rigorous characterization of interface problems in one or more dimensions, with applications from mathematical physics to biology and chemistry. It extends to settings on the real line, star-graphs, and multidimensional configurations, generalizing classical transmission boundary value problems by a probabilistically explicit piecing-out construction and Dirichlet-form analysis.

1. Construction and Canonical Definition

A quintessential one-dimensional SOBM is defined on the state space $G = (-\infty,0^-] \cup [0^+,\infty)$, where $0^-$ and $0^+$ distinguish the origins approached from either side. On each half-line, the process evolves as a standard reflected Brownian motion. At the interface ($0^-,0^+$), the process is governed by local time: when the accumulated boundary local time exceeds an independent exponential threshold, the process restarts at the opposite side (snapping), remains (reflection), possibly sticks for an exponentially distributed time (stickiness), or is sent to a cemetery state (killing), according to rates $a^\pm$ (snapping), $c_2^\pm$ (reflection), $c_3^\pm$ (stickiness), and $c_1^\pm$ (killing) assigned at each endpoint. The process thus realizes a rich mixture of interface phenomena and, in the purely symmetric, non-sticky, non-killed case ($a^+=a^-=\kappa/2$, $c_2^\pm=1$, $0^-$0), reduces to the symmetric snapping-out Brownian motion (SNOB) of parameter $0^-$1 (Lejay, 2016, Erhard et al., 21 Dec 2025).

This piecing-out construction admits pathwise representation: the process switches across the interface at Poisson epochs with intensity given by the semi-permeability parameter times the accrued local time at the interface. Away from the interface, the process is standard reflected Brownian motion.

2. Infinitesimal Generator, Boundary Conditions, and Dirichlet Form

Let $0^-$2 be the Banach space of continuous functions with suitable limits at infinity and at $0^-$3. The infinitesimal generator $0^-$4 of SOBM acts on $0^-$5 as the Laplacian $0^-$6 away from $0^-$7, and its domain comprises functions twice differentiable on each half-line and satisfying the interface conditions: $0^-$8 Here, $0^-$9 governs the rate of snapping (across jump), $0^+$0 is the local reflection, $0^+$1 is the stickiness, and $0^+$2 is the killing term. For the symmetric SNOB, these reduce to: $0^+$3 (Lejay, 2016, Erhard et al., 21 Dec 2025, Erhard et al., 2019).

The associated regular Dirichlet form on $0^+$4 is

$0^+$5

with $0^+$6 (Lejay, 2016, Erhard et al., 2019).

For multidimensional or graph settings (e.g., star graphs), boundary transmission conditions generalize to matching fluxes and jumps with edge- and angle-dependent rates (Bobrowski et al., 2024, Li et al., 2018).

3. Analytical and Probabilistic Structure

SOBM is a Feller process and admits a strongly continuous contraction semigroup solving classical parabolic equations on each side of the interface, subject to boundary transmission conditions encoding the partial permeability. The process is symmetric with respect to Lebesgue measure (or, on rays, $0^+$7). The transition kernel can be written as a mixture of free heat kernels and interface interaction terms, often constructed by the method of images and Laplace inversion, although explicit closed forms are limited to some special cases.

The excursion picture is central: excursions away from the interface are reflected BMs, and interface events are determined by exponentially distributed local times. The process can be simulated by concatenating reflected excursions, combined with i.i.d. random switching times determined by the local time process (Lejay, 2016, Schumm et al., 2023).

For Walsh-type multidimensional settings, the state space becomes a star-graph (rays from a vertex), and the interface (the origin) is replaced by a circle or a vertex, with angular redistribution at the interface governed by a probability measure on angles (Li et al., 2018, Bobrowski et al., 2024). The Dirichlet form and generator are adapted correspondingly, with jumps between rays and parameter-dependent boundary conditions.

4. Scaling Limits, Approximation, and Relations to Other Processes

SOBM arises as the scaling limit of random walks with slow bonds (rate-penalized edges): at a critical scaling, the limit is a SNOB with $0^+$8 proportional to the slow bond rate (Erhard et al., 2019). If the semi-permeable parameters $0^+$9 with fixed ratio, the family of SOBMs converges strongly in operator norm (semigroup, resolvent, cosine family) to skew Brownian motion with skewness $0^-,0^+$0 (Bobrowski et al., 2024, Bobrowski et al., 2023). This convergence is understood both on the level of solutions to the heat equation with transmission conditions and via probabilistic approximations.

In the graph-theoretic context, high-permeability ($0^-,0^+$1) limits of SOBM on stars yield Walsh's spider process, with interface parameters determining the edge selection probabilities at the vertex (Bobrowski et al., 2024, Li et al., 2018).

Boundary conditions for SOBM and their limits can be organized in complementary pairs, such that, in the strong-permeability regime, interface continuity (continuity of $0^-,0^+$2 and weighted balance of derivatives) becomes dominant (Bobrowski et al., 2023, Bobrowski et al., 2024).

5. Multidimensional and Graph Extensions

SO-type processes extend naturally to higher dimensions and graphs:

• In two or more spatial dimensions separated by smooth hypersurfaces (e.g., curve/interface in 2D), the interface mechanism uses local times accumulated on the interface and switches side according to prescribed Bernoulli probabilities after exponentially distributed local time increments (Schumm et al., 2023).
• On star-graphs (e.g., $0^-,0^+$3 rays meeting at a vertex), the process behaves as reflected Brownian motion on each edge, switching to other edges at random interface epochs dictated by local time clocks and edge-specific permeability rates (Bobrowski et al., 2024).
• The Dirichlet forms, resolvent kernels, and semigroups are systematically extendable by accounting for the geometry and the statistical mechanism for interface jumps.

The analytical framework enables error-controlled simulation schemes in 2D, such as walk-on-spheres algorithms with explicit Skorokhod local time computation, and underlies explicit solution expressions for mean first-passage time and splitting probabilities for partially-reactive targets (Schumm et al., 2023).

6. Interface Problems, Applications, and Phase Transitions

SOBM provides a rigorous probabilistic model for interface (barrier or membrane) boundary conditions of the form

$0^-,0^+$4

signifying a jump in solution proportional to flux, as occurs in thermal, chemical, or biological settings. As the permeability parameter tends from $0^-,0^+$5 (barrier disappears) to $0^-,0^+$6 (perfect reflection), a phase transition occurs:

• $0^-,0^+$7: process is free BM.
• $0^-,0^+$8: process is reflected BM.
• $0^-,0^+$9: process is SOBM/SNOB (Lejay, 2016, Li et al., 2018).

In multidimensional media, SNOBM models are used in diffusion MRI, gas permeation in soils, composite-barrier transport, and synaptic receptor motion across gap junctions (Schumm et al., 2023). The SNOBM framework yields analytical and numerical methods for first-passage times, narrow-capture statistics, and splitting probabilities in heterogeneous or composite media.

7. Key Formulas and Statistical Properties

Table: Principal Parameters and Correspondences for SOBM

Parameter/Feature	1D SOBM/SNOB Interpretation	Multidimensional / Graph Interpretation
$a^\pm$0	Interface permeability (snapping rate)	Angle-dependent, edge-dependent jump rates
Dirichlet form	$a^\pm$1	Edgewise, with jump/coupling term at vertex/interface
Transmission condition	$a^\pm$2, $a^\pm$3	$a^\pm$4 $a^\pm$5
Scaling limit	Skew BM as $a^\pm$6 with fixed ratio	Walsh's spider as $a^\pm$7

Statistical properties:

• Number of interface crossings up to time $a^\pm$8 is asymptotically Poisson with mean proportional to $a^\pm$9.
• Transition kernels combine reflected BM kernels with exponentially decaying boundary interaction terms.
• Hitting time distributions and explicit Laplace transforms can be computed in the symmetric case (Erhard et al., 2019, Lejay, 2016).

In summary, SOBM/SNOB provides a versatile toolkit for the probabilistic and analytical treatment of diffusion with semi-permeable, reactive, or sticky interfaces, underpinning a wide variety of modern interface and composite media problems with explicit connections to classical transmission problems, skew Brownian motion, and networked stochastic processes (Lejay, 2016, Li et al., 2018, Erhard et al., 21 Dec 2025, Bobrowski et al., 2024, Bobrowski et al., 2024).