Feller process

Feller processMain

Community hub

Feller process

7 pages, 0 posts

0 subscribers

Recent from talks

Be the first to start a discussion here.

Recent from talks

Be the first to start a discussion here.

Contribute something

About hubMembersContent overviewUpdatesRules

Main reference articles

Feller process

View on Wikipedia

from Wikipedia

This article needs additional citations for verification. Please help improve this article by adding citations to reliable sources. Unsourced material may be challenged and removed.
Find sources: "Feller process" – news · newspapers · books · scholar · JSTOR (September 2009) (Learn how and when to remove this message)

In probability theory relating to stochastic processes, a Feller process is a particular kind of Markov process.

Definitions

[edit]

Let X be a locally compact Hausdorff space with a countable base. Let C₀(X) denote the space of all real-valued continuous functions on X that vanish at infinity, equipped with the sup-norm ||f ||. From analysis, we know that C₀(X) with the sup norm is a Banach space.

A Feller semigroup on C₀(X) is a collection {T_t}_t ≥ 0 of positive linear maps from C₀(X) to itself such that

||T_tf || ≤ ||f || for all t ≥ 0 and f in C₀(X), i.e., it is a contraction (in the weak sense);
the semigroup property: T_t + s = T_t $\circ$ T_s for all s, t ≥ 0;
lim_t → 0||T_tf − f || = 0 for every f in C₀(X). Using the semigroup property, this is equivalent to the map T_tf from t in [0,∞) to C₀(X) being right continuous for every f.

Warning: This terminology is not uniform across the literature. In particular, the assumption that T_t maps C₀(X) into itself is replaced by some authors by the condition that it maps C_b(X), the space of bounded continuous functions, into itself. The reason for this is twofold: first, it allows including processes that enter "from infinity" in finite time. Second, it is more suitable to the treatment of spaces that are not locally compact and for which the notion of "vanishing at infinity" makes no sense.

A Feller transition function is a probability transition function associated with a Feller semigroup.

A Feller process is a Markov process with a Feller transition function.

Generator

[edit]

Feller processes (or transition semigroups) can be described by their infinitesimal generator. A function f in C₀ is said to be in the domain of the generator if the uniform limit

Af=\lim _{t\rightarrow 0}{\frac {T_{t}f-f}{t}},

exists. The operator A is the generator of T_t, and the space of functions on which it is defined is written as D_A.

A characterization of operators that can occur as the infinitesimal generator of Feller processes is given by the Hille–Yosida theorem. This uses the resolvent of the Feller semigroup, defined below.

Resolvent

[edit]

The resolvent of a Feller process (or semigroup) is a collection of maps (R_λ)_λ > 0 from C₀(X) to itself defined by

R_{\lambda }f=\int _{0}^{\infty }e^{-\lambda t}T_{t}f\,dt.

It can be shown that it satisfies the identity

R_{\lambda }R_{\mu }=R_{\mu }R_{\lambda }=(R_{\mu }-R_{\lambda })/(\lambda -\mu ).

Furthermore, for any fixed λ > 0, the image of R_λ is equal to the domain D_A of the generator A, and

{\begin{aligned}&R_{\lambda }=(\lambda -A)^{-1},\\&A=\lambda -R_{\lambda }^{-1}.\end{aligned}}

Examples

[edit]

Brownian motion and the Poisson process are examples of Feller processes. More generally, every Lévy process is a Feller process.
Bessel processes are Feller processes.
Solutions to stochastic differential equations with Lipschitz continuous coefficients are Feller processes.^{[citation needed]}
Every adapted right continuous Feller process on a filtered probability space $(\Omega ,{\mathcal {F}},({\mathcal {F}}_{t})_{t\geq 0})$ satisfies the strong Markov property with respect to the filtration $({\mathcal {F}}_{t^{+}})_{t\geq 0}$ , i.e., for each $({\mathcal {F}}_{t^{+}})_{t\geq 0}$ -stopping time $\tau$ , conditioned on the event $\{\tau <\infty \}$ , we have that for each $t\geq 0$ , $X_{\tau +t}$ is independent of ${\mathcal {F}}_{\tau ^{+}}$ given $X_{\tau }$ .^[1]

References

[edit]

^ Rogers, L.C.G. and Williams, David Diffusions, Markov Processes and Martingales volume One: Foundations, second edition, John Wiley and Sons Ltd, 1979. (page 247, Theorem 8.3)

v t e Stochastic processes
Discrete time	Bernoulli process Branching process Chinese restaurant process Galton–Watson process Independent and identically distributed random variables Markov chain Moran process Random walk Loop-erased Self-avoiding Biased Maximal entropy
Continuous time	Additive process Airy process Bessel process Birth–death process pure birth Brownian motion Bridge Dyson Excursion Fractional Geometric Meander Cauchy process Contact process Continuous-time random walk Cox process Diffusion process Empirical process Feller process Fleming–Viot process Gamma process Geometric process Hawkes process Hunt process Interacting particle systems Itô diffusion Itô process Jump diffusion Jump process Lévy process Local time Markov additive process McKean–Vlasov process Ornstein–Uhlenbeck process Poisson process Compound Non-homogeneous Quasimartingale Schramm–Loewner evolution Semimartingale Sigma-martingale Stable process Superprocess Telegraph process Variance gamma process Wiener process Wiener sausage
Both	Branching process Gaussian process Hidden Markov model (HMM) Markov process Martingale Differences Local Sub- Super- Random dynamical system Regenerative process Renewal process Stochastic chains with memory of variable length White noise
Fields and other	Dirichlet process Gaussian random field Gibbs measure Hopfield model Ising model Potts model Boolean network Markov random field Percolation Pitman–Yor process Point process Cox Determinantal Poisson Random field Random graph
Time series models	Autoregressive conditional heteroskedasticity (ARCH) model Autoregressive integrated moving average (ARIMA) model Autoregressive (AR) model Autoregressive–moving-average (ARMA) model Generalized autoregressive conditional heteroskedasticity (GARCH) model Moving-average (MA) model
Financial models	Binomial options pricing model Black–Derman–Toy Black–Karasinski Black–Scholes Chan–Karolyi–Longstaff–Sanders (CKLS) Chen Constant elasticity of variance (CEV) Cox–Ingersoll–Ross (CIR) Garman–Kohlhagen Heath–Jarrow–Morton (HJM) Heston Ho–Lee Hull–White Korn-Kreer-Lenssen LIBOR market Rendleman–Bartter SABR volatility Vašíček Wilkie
Actuarial models	Bühlmann Cramér–Lundberg Risk process Sparre–Anderson
Queueing models	Bulk Fluid Generalized queueing network M/G/1 M/M/1 M/M/c
Properties	Càdlàg paths Continuous Continuous paths Ergodic Exchangeable Feller-continuous Gauss–Markov Markov Mixing Piecewise-deterministic Predictable Progressively measurable Self-similar Stationary Time-reversible
Limit theorems	Central limit theorem Donsker's theorem Doob's martingale convergence theorems Ergodic theorem Fisher–Tippett–Gnedenko theorem Large deviation principle Law of large numbers (weak/strong) Law of the iterated logarithm Maximal ergodic theorem Sanov's theorem Zero–one laws (Blumenthal, Borel–Cantelli, Engelbert–Schmidt, Hewitt–Savage, Kolmogorov, Lévy)
Inequalities	Burkholder–Davis–Gundy Doob's martingale Doob's upcrossing Kunita–Watanabe Marcinkiewicz–Zygmund
Tools	Cameron–Martin formula Convergence of random variables Doléans-Dade exponential Doob decomposition theorem Doob–Meyer decomposition theorem Doob's optional stopping theorem Dynkin's formula Feynman–Kac formula Filtration Girsanov theorem Infinitesimal generator Itô integral Itô's lemma Karhunen–Loève theorem Kolmogorov continuity theorem Kolmogorov extension theorem Lévy–Prokhorov metric Malliavin calculus Martingale representation theorem Optional stopping theorem Prokhorov's theorem Quadratic variation Reflection principle Skorokhod integral Skorokhod's representation theorem Skorokhod space Snell envelope Stochastic differential equation Tanaka Stopping time Stratonovich integral Uniform integrability Usual hypotheses Wiener space Classical Abstract
Disciplines	Actuarial mathematics Control theory Econometrics Ergodic theory Extreme value theory (EVT) Large deviations theory Mathematical finance Mathematical statistics Probability theory Queueing theory Renewal theory Ruin theory Signal processing Statistics Stochastic analysis Time series analysis Machine learning
List of topics Category

Revisions and contributors Edit on Wikipedia Read on Wikipedia

View on Grokipedia

from Grokipedia

A Feller process is a continuous-time Markov process defined on a locally compact Hausdorff topological space, where the associated transition semigroup acts as a strongly continuous semigroup of operators on the Banach space

C_0

of continuous functions vanishing at infinity, ensuring that the process exhibits continuity in both the initial state and time parameters.^[1] This means that for any continuous function

f \in C_0

and starting point

x

, the expected value

\mathbb{E}_x[f(X_t)] \to f(x)

t \to 0^+

, and the transition probabilities converge in distribution when the starting point varies continuously.^[2] Named after the mathematician William Feller, who pioneered the semigroup approach to Markov processes in the mid-20th century, these processes generalize diffusion processes and provide a framework for analyzing path regularity and generator properties in stochastic analysis.^[3] Feller processes are distinguished by several key properties that facilitate their study and application. They admit a càdlàg (right-continuous with left limits) version, ensuring well-behaved sample paths suitable for applications in physics.^[1] Additionally, they satisfy the strong Markov property with respect to their natural filtration, allowing conditional independence at stopping times, and obey the Blumenthal 0-1 law, where events observable at time zero have probability 0 or 1.^[1] The infinitesimal generator of the semigroup, often a differential operator, enables connections to partial differential equations, making Feller processes essential for modeling phenomena like Brownian motion and Lévy processes.^[1] Historically, Feller's contributions during the 1930s to 1960s integrated measure theory, functional analysis, and probabilistic methods, transforming the understanding of one-dimensional diffusions and extending to higher dimensions through semigroup theory.^[3] His seminal works, including those on boundary conditions for diffusions, laid the groundwork for modern stochastic processes, influencing fields from genetics to queueing theory.^[3] Today, Feller processes remain a cornerstone in probability theory, bridging abstract operator semigroups with concrete stochastic models.^[2]

Introduction

Definition

A Feller process is a continuous-time Markov process

(X_t)_{t \geq 0}

taking values in a locally compact Hausdorff topological space

X

, which is often assumed to possess a countable basis for its topology, and associated with a transition semigroup

\{T_t\}_{t \geq 0}

that acts on the Banach space

C_0(X)

of all continuous real-valued functions on

X

that vanish at infinity, where

C_0(X)

is equipped with the supremum norm

\|f\|_\infty = \sup_{x \in X} |f(x)|

.^[2] This setup ensures that the process can be analyzed through its action on a suitable function space that captures the topological structure of the state space without requiring compactness. The transition semigroup

\{T_t\}_{t \geq 0}

satisfies the following axioms: each

T_t

is a positive linear operator mapping

C_0(X)

into itself;

T_0

is the identity operator on

C_0(X)

; the Chapman-Kolmogorov equation holds, so

T_{s+t} = T_s T_t

for all

s, t \geq 0

; and the semigroup is strongly continuous at

t = 0

, meaning that for every

f \in C_0(X)

\lim_{t \to 0^+} \|T_t f - f\|_\infty = 0

.^[2] This strong continuity implies right-continuity in

t

at 0 and guarantees that the semigroup behaves well with respect to the topology of

X

. The semigroup is realized probabilistically via an associated transition function

P(t, x, \mathrm{dy})

, a kernel on

X

such that for every

f \in C_0(X)

and

x \in X

T_t f(x) = \int_X f(y) \, P(t, x, \mathrm{dy}),

where the integral is understood in the sense of a measure on the Borel

\sigma

-algebra of

X

.^[2] The defining Feller condition requires that

\lim_{t \to 0^+} T_t f(x) = f(x)

for all

x \in X

and all

f \in C_0(X)

, with the convergence holding uniformly on compact subsets of

X

when

f

has compact support. This condition ensures continuity of the process paths in probability as

t \to 0

and aligns the semigroup with the spatial continuity of functions in

C_0(X)

.^[2] Under this semigroup structure, the Feller process

(X_t)_{t \geq 0}

inherits the strong Markov property, meaning that for any stopping time

\tau

, the shifted process

(X_{\tau + t})_{t \geq 0}

conditional on

\mathcal{F}_\tau

is a Markov process with the same transition semigroup starting from

X_\tau

.^[2] This property facilitates the analysis of the process at random times and underpins its applications in potential theory and operator semigroups.

Historical Development

The concept of the Feller process emerged in the mid-20th century as part of William Feller's efforts to develop a boundary theory for Markov processes, particularly emphasizing continuity properties at boundaries and the use of semigroups on function spaces. The term "Feller process" was coined by Eugene B. Dynkin in 1956, abstracting from Feller's earlier work. Feller's foundational contributions began in the 1940s, building toward his seminal 1952 paper "The Parabolic Differential Equations and the Associated Semi-Groups of Transformations", where he analyzed parabolic differential equations associated with Markov processes and introduced the semigroup framework that would define the Feller property. This work was further elaborated in his multi-volume treatise An Introduction to Probability Theory and Its Applications, with Volume II (first published in 1966, based on research from the 1950s) providing a comprehensive treatment of Markov processes, including diffusions and their boundary behaviors. Feller's developments were influenced by earlier advancements in Markov process theory, notably Andrey Kolmogorov's 1931 paper on analytical methods in probability, which laid the groundwork for continuous-parameter Markov processes, and Joseph Doob's 1940s innovations in martingale methods that facilitated the probabilistic analysis of Markov chains and processes. A key operator-theoretic milestone underpinning Feller's approach was the Hille-Yosida theorem, independently established in 1948, which characterized strongly continuous semigroups generated by dissipative operators and provided the analytic tools for handling transition semigroups in infinite-dimensional spaces. In the 1950s, Eugene Dynkin extended Feller's ideas to general state spaces, introducing probabilistic constructions of Markov processes via entrance laws and the Dynkin system, which complemented Feller's analytic perspective and helped formalize the class now known as Feller-Dynkin processes. By the 1960s and 1970s, the theory evolved through connections to Lévy processes—subordinated stable processes serving as prototypical examples of Feller processes—and multidimensional diffusions, influencing stochastic analysis and potential theory. Named after Feller for his pioneering role, the framework has left a lasting legacy in probability, enabling rigorous treatments of irregular boundaries in potential theory and modern stochastic modeling; research in the 2020s continues to explore boundary Feller-Dynkin processes in contexts like coherent families and intertwiners for non-local operators.^[4]

Mathematical Framework

Transition Semigroup

The transition semigroup

\{T_t\}_{t \geq 0}

of a Feller process is constructed as a family of linear operators on the Banach space

C_0(E)

of continuous functions on a locally compact Hausdorff space

E

that vanish at infinity, equipped with the supremum norm. Specifically, for each

t \geq 0

and

f \in C_0(E)

, the operator is defined by

T_t f(x) = \mathbb{E}_x [f(X_t)],

where

\mathbb{E}_x

denotes the conditional expectation given the initial condition

X_0 = x \in E

, and

X = (X_t)_{t \geq 0}

is the underlying Markov process with right-continuous paths. This definition bridges the probabilistic dynamics of the process to the abstract evolution of observables, as

T_t f(x)

represents the expected value of the function

f

at time

t

starting from

x

. The semigroup property

T_{s+t} = T_s T_t

for all

s, t \geq 0

follows directly from the Markov property of the process, which ensures that the future evolution depends only on the current state. This composition reflects the Chapman-Kolmogorov equations in integral form, governing the consistency of transition probabilities over time intervals. Additionally, each

T_t

preserves positivity: if

f \geq 0

, then

T_t f \geq 0

, since expectations of non-negative functions are non-negative. The operators are contractions in the supremum norm, satisfying

\|T_t f\|_\infty \leq \|f\|_\infty

for all

f \in C_0(E)

, which aligns with the preservation of probability measures in the underlying stochastic dynamics.^[1] The transition semigroup is intimately related to a transition kernel

P(t, x, \cdot)

defined on the Borel

\sigma

-algebra of

E

, such that

T_t f(x) = \int_E f(y) \, P(t, x, dy)

for bounded continuous

f

. This kernel is sub-probabilistic, meaning

\int_E P(t, x, dy) \leq 1

for all

t \geq 0

and

x \in E

, allowing for possible mass loss corresponding to absorption or explosion in the process. The kernel is Feller-compatible in the sense that the map

x \mapsto \int f(y) \, P(t, x, dy)

is continuous on

E

for each fixed

t

and

f \in C_0(E)

, ensuring the semigroup acts consistently on the function space.^[2]

Feller Property

The Feller property characterizes a class of Markov processes through continuity conditions imposed on their associated transition semigroup

\{T_t\}_{t \geq 0}

, acting on the space

C_0(X)

of continuous real-valued functions on a locally compact Hausdorff space

X

that vanish at infinity. Specifically, for every

f \in C_0(X)

, the operators satisfy

\lim_{y \to x} T_t f(y) = T_t f(x)

pointwise for each fixed

t > 0

, reflecting the continuity of

T_t f

as a function on

X

, and

\lim_{t \to 0} T_t f(x) = f(x)

pointwise for each

x \in X

, with the latter limit holding uniformly on compact subsets of

X

.^[1]^[2] A key aspect of the Feller property is the preservation of

C_0(X)

under the semigroup action: each

T_t

maps

C_0(X)

into itself for

t \geq 0

, ensuring that the operators maintain the vanishing-at-infinity condition while preserving continuity and boundedness on compacts. This mapping property is essential for handling processes on non-compact state spaces, such as

\mathbb{R}^d

, where standard bounded continuous functions may not suffice, and it aligns the semigroup with the topology of

X

to facilitate analysis of long-range behavior.^[1]^[5] The Feller property has significant implications for the underlying Markov processes, guaranteeing right-continuity in probability for sample paths at fixed times—that is, for any

t > 0

and

x \in X

, the distribution of the process starting near

x

converges to that starting at

x

as the starting point approaches

x

. This distinguishes Feller processes from more general Markov processes by embedding them firmly within semigroup theory, allowing the use of functional analytic tools to study dynamics without requiring pathwise regularity a priori, and ensuring the existence of càdlàg modifications under mild conditions.^[2]^[6] Variations of the Feller property include the weak Feller condition, which requires only continuity in probability for the transition measures (i.e.,

\lim_{y \to x} P(t, y, \cdot) = P(t, x, \cdot)

in the weak topology for

t > 0

), without necessarily preserving

C_0(X)

, and the strong Feller condition, where

T_t

maps the space of bounded Borel measurable functions

B_b(X)

into

C_b(X)

for

t > 0

, implying denser images under the transition operators and often absolute continuity of measures with respect to a reference measure. These variants extend the framework to broader classes of processes while retaining key analytical benefits.^[7]

Operator-Theoretic Aspects

Infinitesimal Generator

The infinitesimal generator

A

of a Feller semigroup

(T_t)_{t \geq 0}

on the Banach space

C_0(X)

of continuous functions vanishing at infinity is defined by

Af = \lim_{t \to 0^+} \frac{T_t f - f}{t},

where the limit is taken in the supremum norm, for all

f

in the domain

D(A) = \{ f \in C_0(X) : \text{the limit exists in } \| \cdot \|_\infty \}

.^[1] This operator

A: D(A) \to C_0(X)

is densely defined and closed, with

D(A)

dense in

C_0(X)

, ensuring the semigroup's strong continuity.^[8] Key properties of

A

include its dissipativity, which follows from the contraction property of the Feller semigroup (

\|T_t\| \leq 1

), implying

\|T_t f - f\| \leq \|f\|

for small

t > 0

and

f \in C_0(X)

.^[8] Additionally,

A

satisfies the positive maximum principle: if

f \in D(A)

and

f(x_0) = \sup_{x \in X} f(x) \geq 0

, then

Af(x_0) \leq 0

.^[1] The semigroup is generated by

A

in the sense that it solves the abstract Cauchy problem

\frac{d}{dt} u(t) = A u(t)

u(0) = f

, for

f \in D(A)

, with solutions

u(t) = T_t f

.^[8] By the Hille-Yosida theorem adapted to Feller semigroups,

A

generates

(T_t)

if and only if

D(A)

is dense in

C_0(X)

A

satisfies the positive maximum principle, and there exists

\alpha > 0

such that the range of

\alpha - A

is all of

C_0(X)

.^[8] This ensures that for all

\lambda > 0

\lambda

belongs to the resolvent set of

A

, with the resolvent operator

R(\lambda, A) = (\lambda - A)^{-1}

bounded by

\|R(\lambda, A)\| \leq 1/\lambda

.^[8] The domain

D(A)

consists precisely of those functions in

C_0(X)

for which the limit defining

Af

exists in the supremum norm. For Feller processes corresponding to diffusions,

D(A)

often comprises functions where

A

takes the form of a second-order elliptic differential operator, such as

Af = \Delta f + b \cdot \nabla f

, with suitable coefficients ensuring the Feller property.^[1]

Resolvent Operator

The resolvent operator associated with a Feller semigroup

(T_t)_{t \geq 0}

acting on the Banach space

C_0(X)

of continuous functions vanishing at infinity on a locally compact Hausdorff space

X

is defined, for

\lambda > 0

, by

R_\lambda f(x) = \int_0^\infty e^{-\lambda t} T_t f(x) \, dt, \quad f \in C_0(X).

^[9] This integral representation establishes

R_\lambda

as a bounded positive linear operator on

C_0(X)

, inheriting the positivity from the semigroup.^[10] A fundamental property of the resolvent is the resolvent equation, which holds for distinct

\lambda, \mu > 0

R_\lambda - R_\mu = (\mu - \lambda) R_\lambda R_\mu.

^[9] This equation underscores the algebraic structure of the family

(R_\lambda)_{\lambda > 0}

and facilitates derivations in spectral theory for Feller processes.^[11] The resolvent is intimately related to the infinitesimal generator

A

of the semigroup, satisfying

R_\lambda = (\lambda I - A)^{-1}

for

\lambda > 0

, where the resolvent set includes the positive half-line.^[9] On the range of

R_\lambda

, the generator can be recovered via

A = \lambda I - R_\lambda^{-1}

.^[1] Moreover, the Post–Widder inversion formula allows recovery of the semigroup from the resolvent for suitable

f

, providing a pointwise reconstruction under the Feller continuity assumptions. The resolvent

R_\lambda

is analytic in

\lambda > 0

, with operator norm bounded by

\|R_\lambda\| \leq 1/\lambda

.^[9] In the Feller context,

R_\lambda

maps

C_0(X)

into the domain

D(A)

of the generator, ensuring that resolvents align with the core properties of the process's transition mechanism.^[10] As

\lambda \to 0^+

, the resolvent

R_\lambda

relates to the potential kernel or Green function, defined as

U f = \int_0^\infty T_t f \, dt

when the limit exists in an appropriate sense, representing the expected occupation measure for the process.^[11] For transient Feller processes, this yields a bounded potential operator on

C_0(X)

, central to potential-theoretic applications.^[12]

Key Properties

Continuity and Regularity

Feller processes are characterized by path properties that ensure a high degree of regularity, particularly in their sample paths. Specifically, every Feller process admits a version with càdlàg (right-continuous with left limits) paths, which follows from the strong continuity of the associated Feller semigroup on the space of continuous functions vanishing at infinity. This version is unique up to indistinguishability and holds almost surely with respect to the process measure. The strong Markov property is a fundamental regularity condition satisfied by Feller processes, stating that for any stopping time

\tau

, the process shifted by

\tau

is conditionally independent of the past given

X_\tau

, and follows the same transition law starting from

X_\tau

. This property enables the decomposition of the process at stopping times and is established via the semigroup structure on locally compact Hausdorff spaces. Hunt's theorem further guarantees the existence of a Hunt process—a right-continuous strong Markov process with left limits—associated with the Feller semigroup, providing a regular realization on such state spaces. In examples involving diffusions, such as Brownian motion, the paths of Feller processes are almost surely Hölder continuous with any exponent less than

1/2

. More generally, Feller processes may feature jumps, as in Lévy processes, where the jump measure is controlled by the Lévy-Khinchin representation, ensuring the overall path regularity remains càdlàg despite discontinuities. The transition probabilities

P(t, x, \cdot)

of a Feller process are inner regular, meaning for every Borel set

B

and

x \in E

\sup \{ P(t, x, K) : K \subset B, K \text{ compact} \} = P(t, x, B)

. This inner regularity, inherent to the vague continuity of the semigroup on Radon measures, supports tightness criteria for sequences of measures and facilitates weak convergence results in the space of càdlàg paths.

Entrance and Exit Boundaries

In the theory of Feller processes, boundaries play a crucial role in describing the asymptotic behavior of the process, particularly at the edges of the state space or at infinity. For one-dimensional diffusions, which form the foundation of Feller's boundary classification, boundaries are categorized into four types based on attainability and accessibility: natural boundaries, which are unattainable in finite time from the interior; entrance boundaries, from which the process can enter the interior but cannot exit back to the boundary; exit boundaries, to which the process can reach from the interior but cannot re-enter after leaving; and regular boundaries, which allow both entry and exit. This classification determines whether boundary conditions must be imposed for the associated differential equations and influences the long-term dynamics of the process. The classification of boundaries in one-dimensional Feller diffusions relies on scale functions, which measure the "distance" to the boundary in terms of hitting probabilities, and speed measures, which quantify local time spent near the boundary. For more general state spaces, Feller's framework extends through the Martin boundary, a compactification of the state space constructed from positive harmonic functions that are minimal with respect to the transition semigroup; this boundary captures inaccessible points at infinity and classifies process behavior analogous to entrance and exit types in higher dimensions or non-locally compact spaces. Entrance laws provide a mechanism to initiate a Feller process from an entrance boundary or infinity, defined as sigma-finite measures

\nu

(0, \infty)

such that the family of measures

\mu_t(f) = \int_0^t T_s \nu (f) \, ds

for

f

in the space of continuous functions vanishing at infinity satisfies the consistency conditions of Kolmogorov's extension theorem, yielding a process with marginal distributions

\mu_t

. These laws are particularly useful for conditioning Feller processes on non-explosion, allowing the construction of processes that "enter" from infinity in finite time while maintaining the Feller property. Exit laws, conversely, describe the behavior leading to absorption or killing at an exit boundary, capturing the distribution of the process up to the exit time. In the context of one-dimensional Feller diffusions, the exit time from a finite interval has an explicit distribution derived from the scale and speed measures, enabling computation of absorption probabilities and expected exit locations without direct simulation. Feller's boundary theory generalizes beyond diffusions to arbitrary state spaces for Feller processes via potential theory, where boundaries correspond to points in the Martin compactification, and harmonic functions on the extended space solve boundary value problems like the Dirichlet problem. This extension links entrance and exit behaviors to the resolvent operator, facilitating the study of killed processes and their connections to minimal positive solutions of the Kolmogorov backward equation.

Examples and Extensions

Classical Examples

One of the most fundamental classical examples of a Feller process is the standard Brownian motion, or Wiener process, defined on the state space

\mathbb{R}^d

. Its infinitesimal generator is the operator

\frac{1}{2} \Delta

, where

\Delta

is the Laplacian, and it satisfies the Feller property through the heat semigroup, which acts by convolution with the Gaussian kernel

(2\pi t)^{-d/2} \exp(-|x|^2/(2t))

.^[10] The paths of Brownian motion are continuous almost surely, and the process is recurrent in dimensions 1 and 2, meaning it returns to any neighborhood of the starting point with probability 1.^[13] The Poisson process, a pure-jump process on the non-negative integers

\mathbb{N}_0

, serves as another canonical Feller process, with infinitesimal generator given by

\mathcal{A} f(n) = \lambda (f(n+1) - f(n))

for a function

f \in C_0(\mathbb{N}_0)

and rate parameter

\lambda > 0

. Its transition semigroup is the Poisson distribution with parameter

\lambda t

, ensuring strong continuity on the space of continuous functions vanishing at infinity. Compound Poisson processes, which generalize this by allowing jumps of random sizes, form a subclass of Lévy processes and inherit the Feller property.^[10] Lévy processes provide a broad class of Feller processes on

\mathbb{R}^d

, encompassing processes with stationary and independent increments that include a deterministic drift, a Brownian component, and a pure-jump part. The infinitesimal generator is a pseudo-differential operator whose symbol is the characteristic exponent

\psi(\xi)

from the Lévy-Khintchin formula,

\psi(\xi) = i b \cdot \xi - \frac{1}{2} \xi^T A \xi + \int_{\mathbb{R}^d \setminus \{0\}} (e^{i \xi \cdot y} - 1 - i \xi \cdot y \mathbf{1}_{|y|<1}) \nu(dy)

, where

b \in \mathbb{R}^d

A

is a positive semidefinite matrix, and

\nu

is the Lévy measure. The Feller property holds provided the semigroup maps

C_0(\mathbb{R}^d)

continuously into itself, a condition satisfied by all Lévy processes without fixed discontinuities; notable subclasses include stable processes with index

\alpha \in (0,2]

.^[10]^[14] Bessel processes, which arise as the radial parts of multidimensional Brownian motion, are Feller processes on the state space

(0, \infty)

. For dimension

\delta > 0

, the squared Bessel process (BESQ

^\delta

) has infinitesimal generator

\mathcal{L} f(r) = 2r f''(r) + (\delta - 1) f'(r)

for

r > 0

, related to the Laguerre differential operator. The origin 0 acts as an entrance boundary for

\delta > 2

, allowing the process to enter from 0 but not exit to it instantaneously, while for

\delta \leq 2

, it is a regular or exit boundary depending on the dimension.^[15] Solutions to stochastic differential equations (SDEs) of the form

dX_t = b(X_t) dt + \sigma(X_t) dW_t

\mathbb{R}^d

, where

b

and

\sigma

are Lipschitz continuous, yield Feller processes under standard existence and uniqueness conditions. The infinitesimal generator is the second-order elliptic operator

\mathcal{L} f(x) = b(x) \cdot \nabla f(x) + \frac{1}{2} \sum_{i,j=1}^d a_{ij}(x) \partial_i \partial_j f(x)

, with

a(x) = \sigma(x) \sigma(x)^T

, and the Feller property follows from the strong continuity of the associated transition semigroup on

C_0(\mathbb{R}^d)

.^[16]^[17]

Applications in Probability and Beyond

Feller processes find significant applications in stochastic modeling, particularly in queueing theory where birth-death processes, such as the M/M/1 queue, serve as continuous approximations that exhibit the Feller property, enabling the analysis of steady-state distributions and transient behaviors under Poisson arrivals and exponential service times.^[18] In risk theory, these processes model surplus dynamics, with exit laws providing explicit computations for ruin probabilities in diffusion approximations of claim processes, facilitating the evaluation of insolvency risks in insurance portfolios.^[19] In physics, Feller processes underpin diffusion approximations for particle systems, capturing spatial spreads in systems like Brownian motion variants with boundaries, which model transport phenomena in heterogeneous media. In finance, the Cox-Ingersoll-Ross (CIR) model, a prototypical Feller diffusion, prices interest rate derivatives by ensuring non-negative rates through entrance boundary conditions at zero, with its semigroup yielding closed-form bond valuations under equilibrium assumptions.^[20] Potential theory leverages the resolvent operators of Feller processes to solve Dirichlet and Poisson problems on state spaces, where the resolvent kernel represents expected occupation times, analogous to Green's functions in electrostatics and providing probabilistic solutions to boundary value problems for random walks and diffusions.^[21] Modern extensions include score-based generative models in machine learning, where Feller semigroups govern the reverse diffusion processes in stochastic differential equations, enabling high-fidelity data synthesis by estimating score functions for complex distributions like images. In biological population dynamics, Feller diffusions approximate branching processes with immigration at entrance boundaries, modeling gene frequency drifts and population growth under selection pressures, as originally explored in genetic contexts.^[22]^[23] Addressing contemporary gaps, 21st-century applications incorporate non-local Feller operators, such as Lévy-Feller processes, to describe anomalous diffusion in materials science, where fractional derivatives capture superdiffusive or subdiffusive behaviors in disordered structures like porous media or composites, aiding simulations of heat conduction and solute transport.^[24]

History

Feller process

Recent from talks

Recent from talks

Contribute something

Contribute something

Media Pages

Timelines

Articles

Notes collections

Notes

Notes

Days in Chronicle

Feller process

Definitions

Generator

Resolvent

Examples

See also

References

Feller process

Introduction

Definition

Historical Development

Mathematical Framework

Transition Semigroup

Feller Property

Operator-Theoretic Aspects

Infinitesimal Generator

Resolvent Operator

Key Properties

Continuity and Regularity

Entrance and Exit Boundaries

Examples and Extensions

Classical Examples

Applications in Probability and Beyond

References

Add your contribution

Related Hubs

Contribute something