On the excision of Brownian bridge paths

Motivated by Pitman and Yor's construction of the 3-dimensional Bessel process, this paper demonstrates that applying a similar excision procedure to a Brownian bridge yields a 3-dimensional Bessel bridge.

The Gist

Imagine you are watching a drunkard walking randomly on a tightrope. This is Brownian motion. Sometimes, this walk starts at a point, wanders around, and must return to the starting point at a specific time. This specific journey is called a Brownian Bridge.

Now, imagine this walk has a "past maximum" line. Think of it as a ceiling that rises whenever the walker reaches a new high point. The walker constantly dips down below this ceiling, creating little "valleys" or excursions.

The Big Idea: The "Pruning" Game

This paper is about a specific game of "pruning" or "excising" these valleys.

The Rules of the Game:

1. Look at the Brownian Bridge.
2. Identify every time the walker dips down from their current highest point (the ceiling).
3. The Cut: If a dip goes all the way down to the ground (level 0), cut that entire valley out of the timeline.
4. The Stitch: Glue the remaining parts of the walk together, closing the gaps.

The authors ask: "If we do this to a Brownian Bridge, what does the resulting path look like?"

The Analogy: The Rice Paddy and the River

To understand the answer, the authors use a visual metaphor (mentioned in the acknowledgments) involving rice paddies.

• The Brownian Bridge is like a river that flows from a source, winds through a landscape, and returns to the sea.
• The Excursions are the little side-streams or tributaries that flow off the main river and eventually dry up (hit level 0).
• The Pruning: Imagine you are a gardener. You go along the river and cut off every side-stream that dries up completely. You then splice the main river back together, skipping the dry sections.

The Surprise Result:
When you do this to a Brownian Bridge, the resulting "spliced" river doesn't look like a random mess. It transforms into something very specific and elegant: a 3-dimensional Bessel Bridge.

In the world of math, a Bessel Bridge is like a "happy" path. It starts at zero, wanders up, and comes back down, but it is forbidden from touching zero in the middle. It's like a tightrope walker who is magically repelled by the ground; they can get close, but they can never actually step on it.

Why is this important?

This paper connects two very different-looking worlds:

1. The "Sad" Path: A path that is allowed to hit zero and get stuck there (the excised bridge).
2. The "Happy" Path: A path that is strictly positive and bounces off zero (the Bessel bridge).

The authors show that if you take the "sad" path, surgically remove the parts where it hits the ground, and stitch the rest together, you get the "happy" path.

The "Secret Sauce" (The Math Part)

The paper also calculates the probability of this happening. It turns out that the length of the new, stitched-up path isn't random in a simple way. It depends on how high the original bridge went.

• The Analogy: Imagine the original bridge reached a peak height of $H$. The new, stitched path will have a specific length distribution that is mathematically linked to that height $H$.
• The authors derived a formula (Theorem 1.1) that acts like a "translation dictionary." It tells you exactly how to translate the statistics of the original bridge into the statistics of the new, stitched-up bridge.

Summary in Plain English

1. Start with a random walk that must return to zero (Brownian Bridge).
2. Cut out every time the walk dips down and touches the ground.
3. Glue the remaining floating pieces together.
4. Result: You get a new path that looks like a 3D Bessel Bridge—a path that starts and ends at zero but never touches the ground in between.

This discovery is significant because it provides a new, constructive way to build these complex mathematical objects. Instead of defining them with complicated equations, you can now think of them as "cleaned-up" versions of simpler, messier paths. It's like taking a messy sketch, erasing the parts that touch the bottom of the page, and realizing the remaining lines form a perfect, elegant sculpture.

Technical Summary

Here is a detailed technical summary of the paper "On the excision of Brownian bridge paths" by Gabriel Berzunza Ojeda and Ju-Yi Yen.

1. Problem Statement and Motivation

The paper addresses a fundamental question in stochastic process theory regarding path transformations of Brownian motion.

• Background: A cornerstone result by Pitman and Yor (2003) demonstrated that a 3-dimensional Bessel process ($BES(3)$) can be constructed from a standard Brownian path by excising (removing) all excursions below the past maximum process that reach zero, and then concatenating (joining) the remaining excursions.
• The Gap: While this relationship is well-established for the standard Brownian motion (leading to $BES(3)$), it was unclear whether a similar excision procedure applied to a Brownian bridge (a Brownian motion conditioned to return to zero at time $t=1$) would yield a related process, specifically a 3-dimensional Bessel bridge (or equivalently, a normalized Brownian excursion).
• Objective: The authors aim to formalize this excision procedure for the Brownian bridge, define the resulting process, and establish the precise distributional relationship between the original bridge, the excised process, and the Bessel bridge.

2. Methodology

The authors employ a combination of deterministic path transformations, excursion theory, and Poisson point process (PPP) techniques.

A. Deterministic Path Transformations

The paper first establishes a rigorous framework for manipulating continuous functions:

1. Bridge to Meander Transformation: They define a transformation ($T^{me}$) that maps a Brownian bridge to a time-reversed Brownian meander by splitting the bridge at its unique maximum and reversing the second half.
2. Excision Operator ($H$): They define a deterministic operator that:
   • Identifies excursions of a path below its running maximum.
   • Excises (removes) those excursions that touch the level 0.
   • Concatenates the remaining excursions to form a new path of reduced length.
3. Scaling: The resulting path is scaled (Brownian scaling) to fit the time interval $[0,1]$.

B. Stochastic Analysis via Poisson Point Processes

To analyze the distribution of the excised process, the authors leverage the Lévy-Itô excursion theory:

• They decompose the Brownian path into a Poisson point process (PPP) of excursions away from the past maximum.
• They apply thinning to this PPP based on specific rules:
  • For excursions below the level $x/2$, an excursion is kept only if its height is strictly less than its level.
  • For excursions between $x/2$ and $x$, an excursion is kept only if its height is strictly less than its level minus $x/2$.
• This allows them to characterize the total length of the non-excised path ($\tau_x$) and the structure of the resulting path ($X_x$) using the properties of the PPP and the Itô excursion measure.

C. Linking to Bessel Processes

The authors utilize known results connecting the excursions of Brownian motion to the excursions of $BES(3)$ processes. Specifically, they show that the concatenation of non-excised excursions of a Brownian path up to a hitting time $T_x$ is distributionally equivalent to a $BES(3)$ process stopped at its hitting time of $x/2$, followed by an independent shifted $BES(3)$ process.

3. Key Contributions and Results

A. The Main Theorem (Theorem 1.1)

The central result characterizes the joint law of the scaled excised bridge process ($\tilde{Y}^{br}$) and the scaling factor ($\tau_{br}$).
For any positive or bounded measurable function $G$:
$$\mathbb{E}\left[G(\tilde{Y}^{br}) (B_{br}(\mu))^{-1} (\tau_{br})^{1/2}\right] = \int_0^\infty \mathbb{E}\left[G(B_{exc}) \phi_x(\bar{B}_{exc})\right] dx$$
Where:

• $B_{br}$ is the Brownian bridge, $\mu$ is the time of its maximum.
• $\tilde{Y}^{br}$ is the Brownian scaling of the excised bridge.
• $B_{exc}$ is the normalized Brownian excursion (equivalent to a 3D Bessel bridge).
• $\bar{B}_{exc}$ is the maximum of the excursion.
• $\phi_x(t)$ is a specific kernel function defined via integrals involving the density of the maximum of a Brownian excursion.

This formula establishes that the law of the excised bridge is a mixture of laws of Brownian excursions, weighted by the kernel $\phi_x$.

B. Corollary on the Maximum

The paper derives a corollary (Corollary 1.2) specifically for the maximum of the scaled process, showing how the maximum of the excised bridge relates to the maximum of the standard excursion through the same integral transform.

C. Technical Continuity Results

A significant portion of the paper (Sections 3 and Appendices) is dedicated to proving the continuity and measurability of these path transformations.

• They define a subset of "regular" paths ($B^{reg}_1$) where the maximum is unique and 0 is not a strict local minimum.
• They prove that the excision and concatenation maps are continuous on this set with respect to the uniform topology.
• They show that standard Brownian bridges and meanders belong to this regular set with probability 1, ensuring the transformations are well-defined almost surely.

4. Significance and Implications

• Extension of Pitman-Yor Theory: The paper successfully extends the famous Pitman-Yor "excision" theorem from the realm of infinite-time Brownian motion (leading to $BES(3)$) to the finite-time Brownian bridge (leading to the Bessel bridge/Excursion).
• Unified Framework: It provides a unified path-transformation perspective linking Brownian bridges, meanders, first-passage bridges, and Bessel processes.
• New Construction of Bessel Bridges: It offers a novel constructive method for generating a 3-dimensional Bessel bridge by simply "pruning" a Brownian bridge.
• Rigorous Foundation: By establishing the continuity of these transformations on specific function spaces, the authors provide a robust mathematical foundation for future applications in stochastic analysis, particularly in the study of random trees, fragmentation processes, and statistical mechanics models where such path transformations are relevant.

In summary, the paper resolves a specific open question by demonstrating that the "pruning" of a Brownian bridge yields a process intimately connected to the Brownian excursion, quantified by a precise integral identity involving the maximum of the excursion.