[Paper Review] Last Branching in Directed Last Passage Percolation

Patrik L. Ferrari, Herbert Spohn|ArXiv.org|Nov 18, 2002

Random Matrices and Applications11 references3 citations

TL;DR

This paper studies the geometry of maximal-length directed polymers in a Poissonian random environment in 1+1 dimensions, focusing on the last branching point where two polymers starting from the same origin and ending at distinct points on a line diverge. It establishes that the last branching point scales with exponent $2/3$ in transverse fluctuations, confirming universality with the KPZ class, and proves that with high probability, multiple branches persist at distances $t^\mu$ from the endpoint line for $\mu < 2\nu - 1/3$, with $\nu = 2/3$ yielding the critical scaling.

ABSTRACT

The 1+1 dimensional directed polymers in a Poissonean random environment is studied. For two polymers of maximal length with the same origin and distinct end points we establish that the point of last branching is governed by the exponent for the transversal fluctuations of a single polymer. We also investigate the density of branches.

Motivation & Objective

• To understand the spatial geometry of maximal-length directed polymers in a 1+1 dimensional Poissonian random environment.
• To determine the scaling exponent of the last branching point where two polymers with the same origin but different endpoints diverge.
• To analyze the density and persistence of branches in the polymer network formed by maximal paths.
• To confirm that the transverse fluctuation exponent for the last branching point matches the known $2/3$ exponent of single polymer fluctuations.
• To establish the existence of a non-degenerate number of unmerged branches at distances $t^\mu$ from the endpoint line under appropriate scaling.

Proposed method

• The authors model the system as directed polymers in a Poisson point process on $\mathbb{R}_+^2$, with path length defined by the number of visited points.
• They analyze the set of maximizers $\Pi_{\max}(S,E,\omega)$ from origin $S=(0,0)$ to end points $E$ on the line $U_t = \{(t-x,t+x)\}$.
• Using large deviation estimates and Fredholm determinant techniques, they derive bounds on the probability that the last branching occurs within distance $t^\mu$ from $U_t$.
• They apply a geometric argument involving a region $D(w,l)$ around the line $U_t$, with $w = t^\nu/2$ and $l = t^\mu$, to control the location of branching points.
• The proof relies on comparing path lengths from origin to end points via intermediate points in $D(w,l)$, using estimates on $\sqrt{a(0,z)} + \sqrt{a(z,E)}$ to bound deviations from optimality.
• They combine large deviation bounds with probabilistic estimates to show that the probability of last branching within $t^\mu$ decays as $t^{-2}$ under the condition $\mu < 2\nu - 1/3$.

Experimental results

Research questions

• RQ1 What is the scaling exponent of the transverse distance to the last branching point of two maximal-length polymers with the same origin but different endpoints?
• RQ2 How many unmerged branches persist at a distance $t^\mu$ from the endpoint line $U_t$ in the limit of large $t$?
• RQ3 Does the last branching point exhibit the same $2/3$ transverse fluctuation exponent as a single polymer?
• RQ4 What is the critical scaling $\mu$ such that the number of unmerged branches at distance $t^\mu$ from $U_t$ remains non-degenerate as $t \to \infty$?
• RQ5 Under what conditions does the joint distribution of polymer lengths from a common origin to nearby endpoints remain non-degenerate?

Key findings

• The last branching point of two maximal-length polymers from the same origin to distinct endpoints on $U_t$ is governed by the $2/3$ transverse fluctuation exponent of a single polymer.
• With high probability, at least $t^{1-\nu}$ branches remain unmerged at distance $t^\mu$ from $U_t$ for $\mu < 2\nu - 1/3$ and $\nu = 2/3$.
• The probability that the last branching occurs within distance $t^\mu$ from $U_t$ decays as $t^{-2}$ for $\mu < 2\nu - 1/3$, implying typical branching occurs further out.
• For $\mu < 5/6 - \mu/2$, the number of unmerged branches at distance $t^\mu$ from $U_t$ grows as $t^{\sigma}$ with $\sigma = 1 - \nu$, and this number tends to infinity in probability as $t \to \infty$.
• The result confirms that the geometry of the polymer network is governed by the same KPZ universality class exponents as the single polymer fluctuation, with $2/3$ as the critical transverse exponent.
• The analysis confirms that the critical scaling $\nu = 2/3$ is essential for non-degenerate joint statistics, consistent with earlier results in the literature.