# Brownian excursion
A **Brownian excursion** is a stochastic process \\( \eta(t) \\), \\( 0 \\le t \\le 1 \\), that behaves like a one-dimensional [Brownian motion](https://www.edgepedia.org/brownian-motion) conditioned to stay nonnegative throughout the interval and to return to 0 at both endpoints. It is a Markov process with \\( \mathsf{P}\{\eta(0)=0\} = \mathsf{P}\{\eta(1)=0\} = 1 \\) and \\( \mathsf{P}\{\eta(t) \ge 0\} = 1 \\) for \\( 0 \\le t \\le 1 \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion). Informally, it is a Brownian path started at the origin and conditioned to remain positive and return to the origin at a fixed time[2](https://web.stanford.edu/~lindrew/PIMS-percolation.pdf). The process belongs to a family of constrained Brownian motions that also includes the [Brownian bridge](https://www.edgepedia.org/brownian-bridge), which returns to 0 but may go negative, and the [Brownian meander](https://www.edgepedia.org/brownian-meander), which stays positive but has no fixed endpoint[3](https://academic.oup.com/mnras/article/438/2/1630/1014270).
| Key fact | Detail |
|---|---|
| Definition | Markov process on \\( [0,1] \\) with \\( \eta(0)=\eta(1)=0 \\) and \\( \eta(t) \ge 0 \\) throughout; for \\( 00 \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion) |
| Discrete origin | Scaling limit of Bernoulli (Dyck path) excursions: \\( \eta_{[2nt]}/\sqrt{2n} \\) converges in finite-dimensional distributions to \\( \eta(t) \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion) |
| Bessel identity | A Brownian excursion of length \\( t \\) is identical in law to a BES(3) bridge from 0 to 0 of length \\( t \\), an observation due to David Williams[4](https://celebratio.org/Yor_M/article/1126/) |
| Vervaat transform | Cyclically shifting a standard Brownian bridge at the time of its minimum turns it into a Brownian excursion (W. Vervaat, 1978)[5](https://www.cambridge.org/core/journals/advances-in-applied-probability/article/abs/relation-between-the-brownian-bridge-and-the-brownian-excursion/A3C511C947C4BB04CAF924E1D537EE07) |
| Scaling | An excursion of length \\( l \\) is \\( B^{\mathrm{ex},l}_t = \sqrt{l}\, B^{\mathrm{ex}}_{t/l} \\) for \\( 0 \le t \le l \\)[6](https://www.researchgate.net/publication/257811919_The_Vervaat_transform_of_Brownian_bridges_and_Brownian_motion) |
| Role in random trees | The rescaled contour process of a conditioned Galton-Watson tree converges to the normalized Brownian excursion (Aldous' theorem)[7](https://arxiv.org/abs/math/0511515) |
## Definition and basic structure
The process arises as a limit of discrete excursions. If \\( \{\eta_0, \dots, \eta_{2n}\} \\) is a Bernoulli excursion, meaning a walk of \\( \pm 1 \\) steps that stays nonnegative and returns to 0, then as \\( n \to \infty \\) the rescaled processes \\( \{\eta_{[2nt]}/\sqrt{2n}\} \\) converge in finite-dimensional distributions to the Brownian excursion \\( \eta(t) \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion). This is the same statement in continuous language: Dyck paths, which are finite paths of \\( \pm 1 \\) steps from \\( (0,0) \\) to \\( (2n,0) \\) staying nonnegative, have the Brownian excursion as their scaling limit[2](https://web.stanford.edu/~lindrew/PIMS-percolation.pdf).
Because \\( \eta \\) is Markov, its finite-dimensional distributions are determined by the one-dimensional density \\( f(t,x) \\) and the transition density \\( f(t,x;u,y) \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion). For \\( 00 \\), the density is explicit in terms of the normal density \\( \phi \\), and \\( f(t,x)=0 \\) for \\( x \le 0 \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion).
Excursions of other lengths follow from Brownian scaling. Writing \\( B^{\mathrm{ex}} \\) for the normalized excursion on \\( [0,1] \\), the excursion of length \\( l>0 \\) is defined by \\( B^{\mathrm{ex},l}_t := \sqrt{l}\, B^{\mathrm{ex}}_{t/l} \\) for \\( 0 \le t \le l \\)[6](https://www.researchgate.net/publication/257811919_The_Vervaat_transform_of_Brownian_bridges_and_Brownian_motion).
## The Vervaat transform
In 1978, W. Vervaat of Katholieke Universiteit Nijmegen published a direct relation between the Brownian bridge and the Brownian excursion[5](https://www.cambridge.org/core/journals/advances-in-applied-probability/article/abs/relation-between-the-brownian-bridge-and-the-brownian-excursion/A3C511C947C4BB04CAF924E1D537EE07). For a continuous function \\( f \\) on \\( [0,1] \\), the transform shifts the path cyclically so that it starts at the first time \\( \tau(f) \\) at which the minimum is attained, subtracting the minimum value:
\\[ V(f)(t) := f(\tau(f)+t \bmod 1) + f(1)\,1_{\{t+\tau(f) \ge 1\}} - f(\tau(f)). \\]
Applied to a Brownian bridge, this operation produces a Brownian excursion[8](https://arxiv.org/abs/1310.3889). The transform has a useful path decomposition: for a Vervaat bridge with endpoint \\( \lambda<0 \\), if \\( Z_\lambda \\) is the time of first return to 0, with density \\( f_{Z_\lambda}(t) = \frac{|\lambda|}{\sqrt{2\pi t (1-t)^3}} \exp\!\left(-\frac{\lambda^2 t}{2(1-t)}\right) \\), then the path splits into two conditionally independent pieces: a Brownian excursion of length \\( Z_\lambda \\) followed by a first passage bridge through level \\( \lambda \\) of length \\( 1-Z_\lambda \\)[6](https://www.researchgate.net/publication/257811919_The_Vervaat_transform_of_Brownian_bridges_and_Brownian_motion). When the two endpoints of the bridge differ, the Vervaat transform is not Markovian[8](https://arxiv.org/abs/1310.3889).
## Connection with the Bessel process
David Williams made the key observation that a Brownian excursion of length \\( t \\) is identical in law to a \\( \operatorname{BES}(3) \\) bridge from 0 to 0 of length \\( t \\), where \\( \operatorname{BES}(\delta) \\) denotes a \\( \delta \\)-dimensional Bessel process, the radial part of \\( \delta \\)-dimensional Brownian motion[4](https://celebratio.org/Yor_M/article/1126/). The BES(3) process is a representation of Brownian motion on the positive half-line conditioned never to return to 0, in the sense of Doob's \\( h \\)-transforms[4](https://celebratio.org/Yor_M/article/1126/).
Itô's excursion law for positive Brownian excursions has two equivalent descriptions built on this identity[4](https://celebratio.org/Yor_M/article/1126/):
1. **Conditioning on the lifetime.** Pick a lifetime \\( t \\) according to the \\( \sigma \\)-finite density \\( (2\pi)^{-1/2} t^{-3/2}\,dt \\) on \\( (0,\infty) \\); then run a BES(3) bridge from 0 to 0 over time \\( t \\).
2. **Conditioning on the maximum.** Pick a maximum \\( m \\) according to the density \\( m^{-2}\,dm \\); then join back to back two independent BES(3) processes, each started at 0 and run until it first hits \\( m \\).
[Jim Pitman](https://www.edgepedia.org/jim-pitman), professor of statistics at the [University of California](https://www.edgepedia.org/university-of-california), Berkeley, and [Marc Yor](https://www.edgepedia.org/marc-yor), probabilist at Université Paris VI et VII, generalized this equivalence to Bessel processes of arbitrary real dimension \\( \delta \ge 0 \\)[4](https://celebratio.org/Yor_M/article/1126/).
## Functionals: maximum and area
The maximum of the excursion has a distribution connected to the [Riemann zeta function](https://www.edgepedia.org/riemann-zeta-function). If \\( M \\) denotes the maximum of a Brownian excursion of length 1 and \\( Y := \sqrt{2/\pi}\, M \\), then \\( \mathsf{E}(Y^s) = 2\xi(s) \\) for the Riemann \\( \xi \\) function, and Chung's identity gives \\( Y^2 \stackrel{d}{=} \frac{\pi}{2}(T + \hat{T}) \\), where \\( T \\) and \\( \hat{T} \\) are the hitting times of level 1 by two independent BES(3) processes started at 0[4](https://celebratio.org/Yor_M/article/1126/). The distribution function of the maximum has the explicit series \\( F(x) = \sum_{j=-\infty}^{\infty} (1-4j^2x^2) e^{-2j^2x^2} \\)[1](https://encyclopediaofmath.org/wiki/Brownian_excursion).
The area under the excursion follows the Airy distribution. This distribution function, first studied for the area under a Brownian excursion, also appears in the physics of fluctuating interfaces, where it describes the maximal height of the interface[9](https://arxiv.org/abs/cond-mat/0409566). The joint law of the maximum \\( M \\) and the time \\( t_m \\) at which it is achieved has been derived in explicit form for excursions, meanders, and reflected bridges, with the marginal density of \\( t_m \\) given as a doubly infinite series[10](https://arxiv.org/abs/0802.2619).
## Random trees and the continuum random tree
The Brownian excursion is the coding function of the Brownian continuum random tree (CRT), introduced by [David Aldous](https://www.edgepedia.org/david-aldous), professor of statistics at the University of California, Berkeley. Aldous' theorem states that the contour process of a Galton-Watson tree conditioned on its total population size converges in distribution, after rescaling, to the normalized Brownian excursion[7](https://arxiv.org/abs/math/0511515). The same convergence holds for the contour function of a tree drawn uniformly from ordered trees of a given size, as \\( k \to \infty \\)[11](https://www.claymath.org/library/proceedings/cmip015c.pdf).
Using this construction with a standard Brownian excursion gives the Brownian CRT[12](https://www.stat.berkeley.edu/~aldous/Research/research-crt.html). The resulting random compact metric space has dimension 2, just as a sample path of Brownian motion does[12](https://www.stat.berkeley.edu/~aldous/Research/research-crt.html). The excursion also controls dynamics on the limit tree: Brownian motion on the continuum random tree is the scaling limit of simple random walks on any family of discrete \\( n \\)-vertex ordered trees whose search-depth functions converge to the Brownian excursion[13](https://arxiv.org/abs/0710.0460).
Beyond trees, the CRT coded by the excursion appears as a scaling limit of other combinatorial objects, including random dissections of a polygon, which rescaled by \\( n^{-1/2} \\) converge in the Gromov-Hausdorff sense to a multiple of the Brownian tree[14](https://arxiv.org/abs/1305.3534).
## Applications in physics and other fields
In cosmology, the excursion set formalism uses conditioned Brownian paths to model the density contrast field: a Brownian excursion is defined there as a cumulative random walk of density contrast of length \\( S \\) conditioned to stay positive in the interval, describing regions that collapse, while the Brownian bridge with negative barriers describes merging regions[3](https://academic.oup.com/mnras/article/438/2/1630/1014270).
In statistical physics, the area under the Brownian excursion and related functionals govern quantities such as maximal interface heights[9](https://arxiv.org/abs/cond-mat/0409566). Within probability itself, the excursion is a building block: the Brownian path viewed through Itô's excursion theory is a concatenation of a countable collection of random path fragments timed by a local time clock, and the excursion law supplies the distribution of these fragments[4](https://celebratio.org/Yor_M/article/1126/).
## References
1. [Brownian excursion, Encyclopedia of Mathematics](https://encyclopediaofmath.org/wiki/Brownian_excursion)
2. [Scaling limits and Dyck paths, PIMS lecture notes (Lindrew)](https://web.stanford.edu/~lindrew/PIMS-percolation.pdf)
3. [Extended void merging tree algorithm for self-similar models, Monthly Notices of the Royal Astronomical Society](https://academic.oup.com/mnras/article/438/2/1630/1014270)
4. [Jim Pitman, Marc Yor and Brownian excursions, Celebratio Mathematica (2015)](https://celebratio.org/Yor_M/article/1126/)
5. [W. Vervaat, A relation between the Brownian bridge and the Brownian excursion, Advances in Applied Probability 10 (1978), p. 299](https://www.cambridge.org/core/journals/advances-in-applied-probability/article/abs/relation-between-the-brownian-bridge-and-the-brownian-excursion/A3C511C947C4BB04CAF924E1D537EE07)
6. [Titus Lupu, Jim Pitman, Wenpin Tang, The Vervaat transform of Brownian bridges and Brownian motion, Electronic Journal of Probability 20 (2015)](https://www.researchgate.net/publication/257811919_The_Vervaat_transform_of_Brownian_bridges_and_Brownian_motion)
7. [Jean-François Le Gall, Random trees and applications (2005)](https://arxiv.org/abs/math/0511515)
8. [The Vervaat transform of Brownian bridges and Brownian motion, arXiv:1310.3889](https://arxiv.org/abs/1310.3889)
9. [Airy Distribution Function: From the Area Under a Brownian Excursion to the Maximal Height of Fluctuating Interfaces, arXiv:cond-mat/0409566](https://arxiv.org/abs/cond-mat/0409566)
10. [Satya N. Majumdar, Julien Randon-Furling, Michael J. Kearney, Marc Yor, On the time to reach maximum for a variety of constrained Brownian motions, arXiv:0802.2619](https://arxiv.org/abs/0802.2619)
11. [Scaling limits of random trees and planar maps, Clay Mathematics Institute Proceedings](https://www.claymath.org/library/proceedings/cmip015c.pdf)
12. [David Aldous, The Continuum Random Tree, research summary](https://www.stat.berkeley.edu/~aldous/Research/research-crt.html)
13. [David Croydon, Convergence of simple random walks on random discrete trees to Brownian motion on the continuum random tree, arXiv:0710.0460](https://arxiv.org/abs/0710.0460)
14. [The CRT is the scaling limit of random dissections, arXiv:1305.3534](https://arxiv.org/abs/1305.3534)
---
*Topic: Encyclopedia › Physical world and mathematics › Mathematics and statistics › Statistics and probability › Stochastic processes › Continuous-time and continuous-state processes › Gaussian and Wiener processes › Brownian bridge, excursion, and meander*
*Initially written Sep 17, 2026 · Reviewed: Sep 30, 2026; Oct 15, 2026 · Edited: Sep 30, 2026; Oct 15, 2026 · Last review: Oct 15, 2026*
*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*
License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license