Scaled penalization of Brownian motion with drift and the Brownian ascent

TL;DR

This paper extends the analysis of a Brownian penalization model by studying the effects of scaling and identifying new phase transitions, including a novel process called the Brownian ascent, across the entire parameter plane.

Contribution

It introduces a scaled penalization framework for Brownian motion, characterizes the limit processes, and identifies new critical phases including the Brownian ascent.

Findings

Identification of the Brownian ascent as a limit process at critical phases

Extension of Roynette-Yor results to the entire parameter plane

Connection of the Brownian ascent to known Brownian path fragments and transformations

Abstract

We study a scaled version of a two-parameter Brownian penalization model introduced by Roynette-Vallois-Yor in arXiv:math/0511102. The original model penalizes Brownian motion with drift $h\in\mathbb{R}$ by the weight process ${\big(\exp(\nu S_t):t\geq 0\big)}$ where $\nu\in\mathbb{R}$ and $\big(S_t:t\geq 0\big)$ is the running maximum of the Brownian motion. It was shown there that the resulting penalized process exhibits three distinct phases corresponding to different regions of the $(\nu,h)$-plane. In this paper, we investigate the effect of penalizing the Brownian motion concurrently with scaling and identify the limit process. This extends a result of Roynette-Yor for the ${\nu<0,~h=0}$ case to the whole parameter plane and reveals two additional "critical" phases occurring at the boundaries between the parameter regions. One of these novel phases is Brownian motion conditioned to end at its maximum, a process we call the Brownian ascent. We then relate the Brownian ascent to some well-known Brownian path fragments and to a random scaling transformation of Brownian motion recently studied by Rosenbaum-Yor.