[Paper Review] First order transition for the branching random walk at the critical parameter

Thomas Madaule|arXiv (Cornell University)|Jun 18, 2012

Stochastic processes and statistical mechanics19 references4 citations

TL;DR

This paper establishes a first-order transition in the critical additive martingale $ W_\beta $ of a branching random walk at $ \beta = 1 $, showing that $ \lim_{\beta \uparrow 1} \frac{W_\beta}{1 - \beta} = 2D_\infty $ in $ \mathbf{P}^* $-probability, where $ D_\infty $ is the derivative martingale. The analysis relies on studying the trajectory of particles under the polymer measure, revealing a sharp discontinuity in the martingale's behavior at the critical parameter.

ABSTRACT

Considering a critical branching random walk on the real line. From a study of the law of the trajectory of a particle chosen under the polymer measure, we establish a first order transition for the partition function at the critical parameter. This result is strongly related to a recent paper of Aïdékon and Shi in which they solved the problem of the normalisation of the partition function in the critical regime.

Motivation & Objective

• To analyze the regularity of the critical additive martingale $ W_\beta $ at $ \beta = 1 $ in the boundary case of branching random walks.
• To establish a first-order transition in the behavior of $ W_\beta $ as $ \beta \uparrow 1 $, contrasting with analyticity for $ \beta < 1 $.
• To investigate the law of the trajectory of a particle sampled under the polymer measure at the critical point.
• To connect the transition to the derivative martingale $ D_\infty $, building on Aïdékon and Shi's normalization result.

Proposed method

• Define the normalized partition function $ W_{\beta,n} $ as the critical additive martingale for the branching random walk on a Galton-Watson tree.
• Use the polymer measure $ \mu_n^{(\beta)} $ to study the law of particle trajectories under $ \beta $-weighted exploration.
• Analyze the convergence of $ W_\beta $ as $ \beta \uparrow 1 $ using moment conditions, particularly $ \mathbf{E}\left(\left(\sum_{|x|=1} e^{-(1-2\delta_-)V(x)}\right)^{1+2\epsilon_0}\right) < \infty $.
• Relate the limit of $ W_\beta / (1 - \beta) $ to the derivative martingale $ D_\infty = \lim_{n \to \infty} \sum_{|u|=n} V(u) e^{-V(u)} $ via pathwise and probabilistic estimates.
• Apply functional central limit theorems and Brownian motion approximations to the rescaled trajectories under the polymer measure.
• Use coupling and moment bounds to control the probability of large deviations in the trajectory, establishing convergence in $ \mathbf{P}^* $-probability.

Experimental results

Research questions

• RQ1 Does the critical additive martingale $ W_\beta $ exhibit a first-order transition at $ \beta = 1 $?
• RQ2 How does the law of the particle trajectory under the polymer measure behave as $ \beta \uparrow 1 $?
• RQ3 Can the limit $ \lim_{\beta \uparrow 1} \frac{W_\beta}{1 - \beta} $ be expressed in terms of the derivative martingale $ D_\infty $?
• RQ4 What moment conditions ensure the existence and convergence of this limit in probability?
• RQ5 How does the critical behavior of $ W_\beta $ differ from the analytic behavior for $ \beta < 1 $?

Key findings

• The paper establishes that $ \lim_{\beta \uparrow 1} \frac{W_\beta}{1 - \beta} = 2D_\infty $ holds in $ \mathbf{P}^* $-probability under the moment condition (1.5).
• The convergence is not almost sure, but occurs in probability, indicating a discontinuous transition at $ \beta = 1 $, characteristic of a first-order transition.
• The derivative martingale $ D_\infty $, which arises in the study of the minimal position in branching random walks, plays a central role in characterizing the critical behavior.
• The analysis of the polymer measure trajectory reveals that particles at criticality concentrate on paths with specific scaling properties, linked to Brownian motion limits.
• The proof relies on delicate moment estimates and coupling arguments to control the probability of rare events in the trajectory of the particle under $ \mu_n^{(\beta)} $.
• The result confirms a sharp transition in the partition function's behavior at the critical parameter, resolving a key question in the theory of directed polymers on trees.