On the localization regime of high-dimensional directed polymers in time-correlated random field

Jiaming Chen · 2024

Read On the localization regime of high-dimensional directed polymers in time-correlated random field by Jiaming Chen (2024) 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-2412-14712.txt.

Introduction

Jiaming Chen studies, in a 2024 probability paper, how a directed polymer behaves in a random field that changes with time in a linked way. A directed polymer is a chain that threads through a messy medium, and it can either wander freely or lock onto favorable pockets, a shift called localization.

Working in high dimensions, Chen proves the basic laws for this system at every temperature: the chain's average behavior steadies, and a limiting free energy exists and varies smoothly. The paper then separates the wandering and locked regimes with a method that untangles how the random field binds the chain. The results extend polymer theory to messier, time-linked disorder.

Catalog record

CREATORJiaming Chen
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE101 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-14712

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On the localization regime of high-dimensional directed polymers in time-correlated random field Jiaming Chen Abstract This paper describes directed polymer on general time-correlated random field. Law of large numbers, existence and smoothness of limiting free energies are proved at all temperature. We also display the delocalized-localized transition, via separating techniques for entanglement of the random field. Contents 1 Introduction Take N particles in a fluid and assume their interactions are connected by harmonic strings. With external forces and thermal fluctuation, the shape of these particles should be understood as a random configuration. This is a very primitive model for a polymer chain consisting of N particles wafting in water [ 34 ] . Admitting the effect of the external water molecules that randomly kick the particles making up our string, we furthermore allow that these kicks occur randomly and could correlate in both space and time. In the framework of statistical mechanics [ 21 ] , the question we address here is: How does the stochastic impurities affect the macroscopic shape of the polymer chain? In this work, we try to answer this question in discrete models and in the perspective of the localization regime. The classical tréatise usually suppresses the entanglement and interactions among particles; But we find this compromising too simplified and we novelly take into account the space and time-correlations among the underlying random field. As in classical modeling [ 10 ] , we shall represent the polymer chain as a N -particle graph \{(j,x_{j})\}_{j=1}^{N} in \mathbb{N}\times\mathbb{Z}^{d} so that the polymer configuration lives in (d+1) -dimensional lattice, and stretches dependently in time direction. Each point (j,x_{j})\in\mathbb{N}\times\mathbb{Z}^{d} on the graph stands for the position of j th particle in the chain. And we also assume that the transversal motion (S_{j})_{j=1}^{N} performs a finite-range simple random walk for consecutive particles in the chain taking all possible configurations at a fixed distance one from another. The complexity of the underlying random field is nonetheless a difficult topic where classical martingale structure [ 1 ] is destroyed and has been avoided by most literature on directed polymers. Indeed, pioneering work on the subtle non-i.i.d. environment include [ 27 ] where the nonlinear diffusion with correlated space-noise is discussed in the context of random interface growth; And in [ 32 ] the authors introduced the Brownian polymer in space-correlated Gaussian field. Furthermore, the superdiffusivity is investigated in [ 5 , 23 ] in the same model. That said, there has not been much discussion on the challenging time-correlated scenario. Following previous work [ 29 ] on the scaling limits of partition function with omitted space-correlation [ 26 ] and specific Gaussian dependence in time direction, we novelly show the existence and transition between localization and delocalization regimes where the polymer lives in transient dimensions, in the presence of both more general space and time-correlation. Heuristically, striking results characterizing the delocalized-localized transition under i.i.d. random field have been given in [ 7 , 11 ] . Intuitively, delocalization implies that the polymer chain behaves like (S_{n})_{n\geq 1} in transient dimensions, wheres the localization means that the polymer is extremely affected by its favorable medium and thus concentrates in just a few coordinates. Looking beyond, we should also remark that polymers can be defined with long-range random walks. See for instance [ 28 ] where the law of increments belongs to the domain of attraction of an \alpha -stable law; And [ 38 ] where they discussed phase transitions when the random walk has very heavy tails. I also think it is possible to extend my techniques to long-range model of polymer chains. But one should pay particular notice to the specific cone-mixing structure in Section 3 . Perhaps a superposition of countably many cones suffices. Nonetheless, the picture is still obscure, and I encourage researchers (including myself) to future studies. Acknowledgment . The author wishes to thank his PhD advisor Prof. Dr. Alejandro Ramírez at NYU Shanghai for pointing out the notion of time-correlation, and for reading this manuscript as well as raising valuable comments. 2 Directed polymer A random walk on \mathbb{Z}^{d} , d\geq 1 is a sequence (S_{n})_{n\geq 0} starting from S_{0}\in\mathbb{Z}^{d} and moves over the lattice in discrete time. Letting \mathcal{P}(\mathbb{Z}^{d}) denote the canonical product Borel \sigma -algebra, we can alternatively describe its trajectories in the path space ((\mathbb{Z}^{d})^{\mathbb{N}},\mathcal{P}(\mathbb{Z}^{d})^{\otimes\mathbb{N}}) via a probability measure P^{S}_{x} whenever S_{0}=x . Throughout this paper, (S_{n})_{n\geq 0} represents a random walk bounded i.i.d. increments, i.e. \lVert S_{1}\rVert_{1}<\infty . From now on, we consider only transient (S_{n})_{n\geq 0} , i.e. d\geq 3 . Independently, we introduce a time-correlated Markovian random field \omega=(\omega_{n,z}:\,n\in\mathbb{N},z\in\mathbb{Z}^{d}) with its law \mathbb{P} and product \sigma -algebra \mathscr{F}_{\Omega} on \mathbb{N}\times\mathbb{Z}^{d} satisfying \mathbb{E}\big[\exp(\beta\omega_{n,z})\big]<\infty,\qquad\forall\penalty\ \beta\in\mathbb{R},\quad n\in\mathbb{N}\quad\text{and}\quad z\in\mathbb{Z}^{d}. (2.1) Here we adopt the following convention that E^{S}_{x}\coloneqq E_{P^{S}_{x}} and \mathbb{E}\coloneqq E_{\mathbb{P}} . And \beta refers to inverse temperature in physics literature and will more or less reflect the level of disorder in our subsequent setting. And without loss of generality [ 12 , Remark 1.1] , we will restrict to the case \beta\geq 0 . Before stating the main model of this paper, we need first describe has exactly the time-correlation is defined. For this end, we need to define the distance between sets in the space-time lattice. For instance with A,B\subseteq\mathbb{N}\times\mathbb{Z}^{d} , we let d_{1}(A,B)\coloneqq\inf\{\absolutevalue{x-y}_{1}+\absolutevalue{m-k}:\,(m,x)\in A,\,(k,y)\in B\} stand for the \ell_{1} -distance between subsets A and B of \mathbb{N}\times\mathbb{Z}^{d} . By saying \omega Markovian, we mean \mathbb{P}\big((\omega_{n,z})_{(n,z)\in V}\in\dotproduct\big|\mathscr{F}_{V^{c}}\big)=\mathbb{P}\big((\omega_{n,z})_{(n,z)\in V}\in\dotproduct\big|\mathscr{F}_{\partial V}\big),\qquad\forall\penalty\ V\subseteq\mathbb{N}\times\mathbb{Z}^{d},\quad\mathbb{P}\text{-a.s.,} where we follow the convention \mathscr{F}_{\Lambda}\coloneqq\sigma(\omega_{n,z}:\,(n,z)\in\Lambda) for any \Lambda\subseteq\mathbb{N}\times\mathbb{Z}^{d} . The exact mixing nature of the time-correlated field is described as follows. Figure 1: An illustration of simple random in transient dimensions. Definition 2.1 . Let C and g be given and fixed positive real numbers. A Markovian field \omega with law (\mathbb{P},\mathscr{F}_{\Omega}) on \mathbb{N}\times\mathbb{Z}^{d} is said to satisfy the time-correlated condition \textbf{(TC)}_{C,g} if for any finite subsets \Delta\subseteq V\subseteq\mathbb{N}\times\mathbb{Z}^{d} with d_{1}(\Delta,V^{c})>1 and A\subseteq V^{c} , \frac{d\mathbb{P}((\omega_{n,z})_{(n,z)\in\Delta}\in\dotproduct|\eta)}{d\mathbb{P}((\omega_{n,z})_{(n,z)\in\Delta}\in\dotproduct|\eta^{\prime})}\leq\exp\bigg(C\sum_{(m,x)\in\partial\Delta,(k,y)\in\partial A}e^{-g\abs{x-y}_1-g\abs{m-k}}\bigg) simultaneously for all pairs of configurations \eta,\eta^{\prime}\in\Omega\coloneqq([0,\infty))^{\mathbb{N}\times\mathbb{Z}^{d}} which agree on V^{c}\backslash A , \mathbb{P} -a.s. Notice that here we follow the convention \mathbb{P}\big((\omega_{n,z})_{(n,z)\in\Delta}\in\dotproduct\big|\eta\big)=\mathbb{P}\big((\omega_{n,z})_{(n,z)\in\Delta}\in\dotproduct\big|\mathscr{F}_{V^{c}}\big)\penalty\ \text{given that}\penalty\ (\omega_{n,z})_{(n,z)\in V^{c}}=\eta. Apart from this time-correlated condition \textbf{(TC)}_{C,g} , another type of correlation is incorporated in our discussion as well. The following correlation originates from X. Guo [ 20 ] in the context of describing the limiting velocity of random walks in random environments, and models a large class of disordered systems quite naturally. To this recognition, we have the following definition. Definition 2.2 . Let C and g be given and fixed positive constants. A Markovian field \omega with law (\mathbb{P},\mathscr{F}_{\Omega}) on \mathbb{N}\times\mathbb{Z}^{d} is said to satisfy the Guo’s time-correlated condition \textbf{(TCG)}_{C,g} if for any finite subsets \Delta\subseteq V\subseteq\mathbb{N}\times\mathbb{Z}^{d} with d_{1}(\Delta,V^{c})>1 and A\subseteq V^{c} , \frac{d\mathbb{P}((\omega_{n,z})_{(n,z)\in\Delta}\in\dotproduct|\eta)}{d\mathbb{P}((\omega_{n,z})_{(n,z)\in\Delta}\in\dotproduct|\eta^{\prime})}\leq\exp\bigg(C\sum_{(m,x)\in\Delta,(k,y)\in A}e^{-g\abs{x-y}_1-g\abs{m-k}}\bigg) simultaneously for all pairs of configurations \eta,\eta^{\prime}\in\Omega which agree on V^{c}\backslash A , \mathbb{P} -a.s. Intuitively \textbf{(TCG)}_{C,g} is an asymptotic more general assumption. Strictly speaking, the former is not implied by condition \textbf{(TC)}_{C,g} , but in asymptotic terms Guo’s time-correlated condition is harder to work with. Our standing assumption on the random field \omega is that it is Markovian and satisfies either the time-correlated condition in Definition 2.1 or the Guo’s condition in Definition 2.2 . Let us now define the \mathbb{Z}^{d} -valued

Read the complete text