Sticky Reflection in Stochastic Systems

• Sticky reflection is a phenomenon where stochastic processes adhere to boundaries for non-negligible time, characterized by extended Wentzell conditions and explicit local time mechanisms.
• Analytical approaches using Dirichlet forms and mixed measures reveal key spectral properties, convergence rates, and functional inequalities of these processes.
• Applications span quantum transport, wetting phenomena in interacting particle systems, and numerical methods for simulating sticky diffusions in complex geometries.

Sticky reflection refers to a broad class of dynamical, probabilistic, and analytic phenomena in which a stochastic process or physical system interacting with a boundary spends a positive, non-negligible time adhered to that boundary, in contrast to instantaneous (Neumann) reflection or absorption. Instances of sticky reflection arise throughout stochastic analysis, PDEs, quantum transport, interacting particle systems, and functional analysis, with unifying mathematical character provided by generalized (Wentzell) boundary conditions, Dirichlet form techniques, and an explicit local time mechanism capturing adhesion at the interface.

1. Mathematical Formulation of Sticky Reflection

The prototypical model is the sticky reflected diffusion on a domain $D\subseteq\mathbb{R}^d$ or a manifold-with-boundary $M$. In one dimension, for $X_t\in[0,\infty)$, the process is governed by an SDE: $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$ where $L^0_t(X)$ is the right-continuous local time at the boundary point 0, $K_t$ is the increasing process responsible for the stickiness, and $\mu>0$ is the stickiness parameter. Trajectories that reach $x=0$ can linger for positive Lebesgue time, with the expected occupation time at the boundary proportional to the local time and to $\mu$ (Sharma, 8 Aug 2025, Anagnostakis, 2022, Anagnostakis et al., 2024).

More generally, on higher-dimensional domains or manifolds, the generator of sticky diffusion acquires boundary-local contributions. For example, a sticky reflecting Brownian motion on a smooth compact Riemannian manifold $(M, g)$ with smooth boundary $M$0 is described in terms of the operator

$M$1

with $M$2 controlling boundary diffusion and $M$3 the stickiness parameter. The boundary conditions take an extended Wentzell (sometimes called Feller–Wentzell or Robin–Wentzell) form, coupling normal derivatives, values, and intrinsic boundary Laplacians (Bormann et al., 2023, Bormann, 2024).

On star graphs and more general networks, the Dirichlet domain and generator encode stickiness at distinguished nodes or vertices, typically with occupation-time-based constraints (Berry et al., 2024).

2. Probabilistic and Analytical Characterization

Sticky reflection is characterized probabilistically by positive occupation time on the boundary, in sharp contrast to purely reflecting or absorbing boundary conditions where the process is almost surely instantly reflected or killed. Analytically, this translates to speed measures for the process which have singular ("atomic") parts supported on the boundary, i.e.,

$M$4

with $M$5 (Anagnostakis, 2022, Anagnostakis et al., 2024). The presence of the term $M$6 is responsible for the boundary sojourns. Associated Dirichlet forms in $M$7-spaces are constructed using mixed measures which assign weight both to the interior and boundary, enabling automatic positivity of the boundary capacity and ensuring ergodic occupation of the boundary set (Grothaus et al., 2014, Fattler et al., 2014, Grothaus et al., 2014).

Sticky processes are time-changes of nonsticky processes using the boundary's local time: $M$8 where $M$9 is strictly increasing, and $X_t\in[0,\infty)$0 is the right-continuous inverse (Anagnostakis, 2022, Berry et al., 2024).

3. Boundary Conditions: Wentzell-Type and Spectral Implications

The boundary behavior of sticky processes is encoded by extended Wentzell boundary conditions. In one dimension, this reduces to

$X_t\in[0,\infty)$1

where $X_t\in[0,\infty)$2 is the infinitesimal generator in the interior (Sharma, 8 Aug 2025). On manifolds, the boundary operator involves both normal derivatives and (for $X_t\in[0,\infty)$3) a boundary Laplacian: $X_t\in[0,\infty)$4 see (Bormann, 2024). These boundary conditions induce nontrivial interaction between the bulk and boundary and are responsible for subtle changes in the spectrum and invariant measures, e.g., the emergence of boundary-supported invariant or reversible measures, non-standard trace-class operators, and nontrivial Steklov eigenvalues (Bormann et al., 2023).

4. Functional Inequalities and Long-Time Behavior

Sticky reflection fundamentally alters the functional inequalities satisfied by the process. Dirichlet forms with sticky reflection satisfy mixed Poincaré and logarithmic Sobolev inequalities wherein the constants depend on both the geometry and the stickiness parameter. Interpolation-type estimates between the interior and boundary energies yield explicit upper bounds for the spectral gap (Poincaré constant), log-Sobolev constant, Steklov eigenvalue, and the norm of the boundary trace (Bormann et al., 2023, Bormann, 2024, Wang, 26 Aug 2025).

These results imply ultraboundedness, semi-group hypercontractivity, and uniform integrability of the semigroups generated by sticky diffusions, with precisely characterized decay rates. Notably, super and weak Poincaré inequalities provide detailed two-sided $X_t\in[0,\infty)$5-estimates and determine polynomial or sub-exponential convergence rates to equilibrium for the associated Markov semigroups (Wang, 26 Aug 2025).

For example, in the half-space model with polynomial confinement, the decay rates and sharp constants are determined by the stickiness, whereas in the compact manifold case, ultraboundedness corresponds to Gaussian-type heat kernel estimates and exponential ergodicity (Wang, 26 Aug 2025, Bormann et al., 2023).

5. Local Time, High-Frequency Occupation, and Statistical Inference

Local time at the sticky boundary plays a central role both in defining the sticky SDEs and in their statistical inference. High-frequency statistics for discrete-time observations of sticky diffusions (in particular, for the occupation of the boundary) converge to multiples of the local time, allowing for consistent estimation of the stickiness parameter $X_t\in[0,\infty)$6 via

$X_t\in[0,\infty)$7

where $X_t\in[0,\infty)$8 is a functional built from test functions $X_t\in[0,\infty)$9 vanishing at the boundary and $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$0, $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$1. Uniform convergence results and Monte Carlo validation confirm the practical viability of these estimators (Anagnostakis, 2022, Anagnostakis et al., 2024).

These constructions extend to more complex situations involving "SOS" (sticky-oscillating-skew) boundaries with discontinuous volatility and biased reflection, supporting statistical inference for both stickiness and skewness parameters (Anagnostakis et al., 2024).

6. Numerical Schemes and Computational Challenges

Numerical approximation of SDEs with sticky boundary conditions is nontrivial due to the necessity of handling random sojourn times at the boundary. Recent work has introduced robust weak approximation algorithms: first, the half-order projected Euler method, and second, a first-order sticky Euler scheme utilizing symmetrization and local time tracking to achieve order $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$2 convergence in expectation. These schemes can be used to approximate solutions to linear parabolic PDEs with second-order sticky boundary conditions, which arise naturally via the Feynman-Kac formula for sticky diffusions (Sharma, 8 Aug 2025).

Tables of empirical error versus timestep confirm the expected order of convergence for these schemes. The persistence of nonzero occupation time on the boundary requires careful time-stepping and boundary correction strategies, particularly close to the interface.

7. Extensions and Applications

Sticky reflection is central in a variety of applied and theoretical contexts:

• Quantum Transport: At normal-superconductor (NS) interfaces, the quantum Goos–Hänchen effect produces a dwell-time amplification by $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$3, leading to "sticky" interfaces with long Andreev reflection times (Lee et al., 2012).
• Interacting Particle Systems: Sticky reflected models naturally describe wetting phenomena, such as the dynamical Ginzburg–Landau wetting model, with pinning at a hard wall manifest as sticky occupation of the boundary (Fattler et al., 2014).
• Graph-Based Models: Sticky diffusions generalize naturally to star graphs and networks, where stickiness at the central vertex plays a crucial role in the transport and ergodic properties (Berry et al., 2024).
• Markov Process Theory: Construction via Dirichlet forms, Skorokhod decompositions, and strong Feller properties for sticky reflected diffusions underpin a wide spectrum of Markov processes, with applications to stochastic calculus, invariant measure theory, and boundary-loitering phenomena (Grothaus et al., 2014, Grothaus et al., 2014).

A notable connection is to optimal transport: the Fokker-Planck equation for reflected sticky Brownian motion can be formulated as a Wasserstein gradient flow, where the invariant measure has both an absolutely continuous interior part and a singular (boundary-supported) component (Casteras et al., 2024).

Table: Canonical Formulations of Sticky Reflection

Setting	Generator / SDE	Sticky Parameter
$$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$4 (1D)	$$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$5	$$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$6 (stickiness at $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$7)
$$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$8 with $$dX_t = b(X_t)\,dt + \sigma(X_t)\,dW_t + dK_t, \qquad K_t = \mu\,L^0_t(X),$$9	$L^0_t(X)$0 interior; $L^0_t(X)$1 with $L^0_t(X)$2 and tangential Laplacian on $L^0_t(X)$3	$L^0_t(X)$4 (via occupation time proportion)
Star graph, vertex $L^0_t(X)$5	$L^0_t(X)$6 on edge; $L^0_t(X)$7	$L^0_t(X)$8 (stickiness at $L^0_t(X)$9)

The details of the process, spectral estimates, and ergodic occupation time universally reflect the presence and magnitude of the stickiness parameter.

• (Sharma, 8 Aug 2025) A. Sharma, "Weak approximation of stochastic differential equations with sticky boundary conditions"
• (Anagnostakis, 2022) P. Anagnostakis, "Functional convergence to the local time of a sticky diffusion"
• (Anagnostakis et al., 2024) A. Anagnostakis & S. Mazzonetto, "Sticky-threshold diffusions, local time approximation and parameter estimation"
• (Fattler et al., 2014) T. Fattler, M. Grothaus, K. Voßhall, "Construction and analysis of a sticky reflected distorted Brownian motion"
• (Grothaus et al., 2014) M. Grothaus, K. Voßhall, "Construction and analysis of sticky reflected diffusions"
• (Grothaus et al., 2014) M. Grothaus, K. Voßhall, "Strong Feller property of sticky reflected distorted Brownian motion"
• (Bormann et al., 2023) F.-Y. Wang, "Functional Inequalities for Brownian Motion on Riemannian Manifolds with Sticky-Reflecting Boundary Diffusion"
• (Bormann, 2024) M. Bormann, "Functional Inequalities for doubly weighted Brownian Motion with Sticky-Reflecting Boundary Diffusion"
• (Wang, 26 Aug 2025) F.-Y. Wang, "Super and Weak Poincaré Inequalities for Sticky-Reflected Diffusion Processes"
• (Casteras et al., 2024) S. Daneri, "Sticky-reflecting diffusion as a Wasserstein gradient flow"
• (Berry et al., 2024) J. Berry, F. Colantoni, "Sticky diffusions on star graphs: characterization and Itô formula"
• (Lee et al., 2012) Wei-Feng Lee et al., "Sticky Normal-Superconductor Interface"