Convergence from the Log-Gamma Polymer to the Directed Landscape
Xinyi Zhang · 2025
Read Convergence from the Log-Gamma Polymer to the Directed Landscape by Xinyi Zhang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-05685.txt.
Introduction
Xinyi Zhang presents a 2025 math paper on a model of random growth called the log-gamma polymer. Zhang defines new sheet and landscape forms based on the free energy of that model, for two and four inputs. The goal is to show that these forms approach two known limit objects as the scale grows.
Those limits are the Airy sheet and the directed landscape, which sit at the heart of a broad class of random growth models. The proof for the sheet uses a known match between free energy and a combinatorial rule, plus the fact that free energy only moves one way. For the stronger result, Zhang builds new local bounds on swings away from the main line, since the discrete model lacks a handy scaling trait.
Catalog record
| CREATOR | Xinyi Zhang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 108 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-05685
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Convergence from the Log-Gamma Polymer to the Directed Landscape Xinyi Zhang Abstract . We define the log-gamma sheet and the log-gamma landscape in terms of the 2 2 -parameter and 4 4 -parameter free energy of the log-gamma polymer model and prove that they converge to the Airy sheet and the directed landscape, which are central objects in Kardar-Parisi-Zhang (KPZ) universality class. Our proof of the convergence to the Airy sheet relies on the invariance of free energy through the geometric RSK correspondence and the monotonicity of the free energy. To upgrade the convergence to the directed landscape, tail bounds in both spatial and temporal directions are required. However, due to the lack of scaling invariance in the discrete log-gamma polymer—unlike the Brownian setting of the O’Connell–Yor model—existing on-diagonal fluctuation bounds are insufficient. We therefore develop new off-diagonal local fluctuation estimates for the log-gamma polymer. 1. Introduction The study of stochastic growth models has long been one of the central topics in probability theory [ BS95 , CSY04 , KL13 , HHZ95 , Mea98 , Maj07 , Spo06 ] . The well-known central limit theorem characterizes the magnitude of the fluctuation of a sum of independent and identically distributed (i.i.d.) random variables with 1/2 exponent. The limiting distribution of the fluctuation after proper scaling will be Gaussian. However, with certain dependence of variables introduced to the system, we observe totally different limiting behavior. As suggested by various experimental evidence and numerical simulations [ TS10 , MMK+97 , MMMT05 , WZG03 ] , a large class of random growth models including random growth interfaces, interacting particle systems, and directed polymer models have fluctuations of 1/3 scaling exponent and have completely different limiting distributions. These random growth models fall into the Kardar–Parisi–Zhang (KPZ) universality class . There have been many important breakthroughs in the last twenty-five years in understanding the KPZ universality, including the work [ DOV22 ] where Dauvergne, Ortmann and Virág constructed central objects in the KPZ universality class: the Airy sheets and the directed landscape. Not only do the Airy sheets and the directed landscape capture the intricate geometry of random growth phenomena but they are also believed to be the final limiting objects of growth models in the KPZ universality class. The Airy sheets and the directed landscape have been identified as scaling limits for a few zero-temperature models. Notable examples include Poissonian, exponential, and geometric Last Passage Percolation, as demonstrated in the works of Dauvergne-Nica-Virág [ DV21b ] and [ DNV23 ] . However, proving such convergences are more challenging in positive temperature models and hence there are few results. Namely, the solutions to the KPZ equation by Wu [ Wu23 ] and the colored ASEP by Aggarwal, Corwin, and Hegde [ ACH24 ] and the recent results by Dauvergne and Zhang that upgrade the KPZ fixed point convergence to the Directed landscape convergence [ DZ24 ] . The challenges of the positive temperature models come from the following two reasons. First, there are generally no determinantal structures from which we can extract exact formulas in the positive temperature environment. Second, there is no direct metric composition law as in the zero-temperature models and in the limit of the directed landscape. However, the lack of direct metric composition law is remedied by a variational formula of the free energy in the cases of the KPZ equation and the log-gamma polymer model. In this paper, we establish the convergence of another positive temperature model, the log‐gamma polymer, to the directed landscape. Our analysis builds on the geometric Robinson–Schensted–Knuth (RSK) correspondence, which plays a crucial role in relating the partition functions of the log‐gamma polymer to the corresponding free energy on the dual line ensemble. Inspired by the continuous limit of the geometric RSK correspondence [ Cor20 ] that preserves certain polymer free energy used in [ Wu23 ] and the melon map used in [ DOV22 ] that preserves certain last passage values, we exploit the invariance properties of the free energy via the discrete geometric RSK to obtain an equality ( 2.20 ). This equality together with the variational formula ( 2.23 ) provides a key tool to prove the convergence to the Airy sheet as it enables us to disintegrate the free energy from the bottom-left point to the top-right point of the line ensemble into two parts that only pertain the information of few top curves of the line ensemble. Similar to [ Wu23 ] , we also leverage the monotonicity properties of Busemann functions to show that the contribution coming from the bottom part of the line ensemble becomes negligible in the scaling limit. Due to lack of a nice coupling between the KPZ line ensemble and the KPZ sheet at the time that [ Wu23 ] was written, Wu worked with the O’Connell Yor polymer which is the pre-limit of the KPZ line ensemble and has nice Brownian environment. In our case, the situation simplifies as there already exists a nice coupling between the log-gamma polymer and log-gamma line ensemble through the geometric RSK correspondence and hence we only need to take limit once instead of twice. However, the continuous Brownian environment in the work of [ Wu23 ] is not present in the log-gamma polymer model. Although the Brownian symmetry is approximated by the i.i.d. environment of the polymer, lack of Brownian scaling has created challenges in proving the tightness of the log-gamma landscape. It is no longer sufficient to prove a tail bound for local fluctuations of the free energy along the diagonal line. Untypical off-diagonal fluctuations are also needed. We solved this by generalizing the methods used in [ BSS24 ] for the diagonal case to the off-diagonal case. 1.1. Log-gamma polymer Definition 1.1 . An inverse-gamma random variable X with parameter \theta>0 is a continuous random variable with density given by (1.1) f_{\theta}(x)=\frac{\indic\{x>0\}}{\Gamma(\theta)}x^{-\theta-1}\exp(-x^{-1}). Fix some \theta>0 . Let \{d_{i,j}\}_{i,j\in\Z} be i.i.d inverse-gamma random variables with parameter \theta and (\Omega,\mathcal{F},\PP) be the probability space on which \{d_{i,j}\}_{i,j\in\Z} are defined. Let us use \preceq to denote the order on \Z^{2} where (x_{1},y_{1})\preceq(x_{2},y_{2}) if x_{1}<x_{2} or x_{1}=x_{2},y_{1}\leq y_{2} . Let u,v\in\Z^{2} . An up-right lattice path \pi connecting u to v is a set of vertices that can be ordered as u=(x_{1},y_{1}),\cdots,(x_{n},y_{n})=v\in\Z^{2} such that (x_{i},y_{i})\preceq(x_{i+1},y_{i+1}) and x_{i+1}-x_{i}+y_{i+1}-y_{i}=1 for all i=1,\cdots,n-1 . Let \Pi(u\rightarrow v) denote the set of up-right lattice path \pi that connects u to v . Let \pi=(\pi_{1},\cdots,\pi_{k}) where \pi_{i} is an up-right path. We call \pi a multipath if for all 1\leq i\neq j\leq k , \pi_{i}\cap\pi_{j}=\emptyset . Let U and V be two finite subsets of \Z^{2} with the same cardinality. We call (U,V) an endpoint pair if there exists a multipath \pi=(\pi_{1},\cdots,\pi_{k}) such that for every u\in U , there exists a path \pi_{i} and v\in V such that \pi_{i} connects u to v . Similarly, we use \Pi[U\rightarrow V] to denote the set of multipaths \pi that connects U to V . Definition 1.2 (Polymer partition function and polymer free energy) . Let u,v\in\Z^{2} . We define the partition function of the inverse-gamma polymer from u to v by (1.2) Z[u\rightarrow v]=\sum_{\pi\in\Pi[u\rightarrow v]}\prod_{(i,j)\in\pi}d_{i,j}. Moreover, we call (1.3) \log Z[u\rightarrow v] the free energy of the inverse-gamma