Improved bounds for connection probabilities in random loop models

Volker Betz, Andreas Klippel, Julian Nauth · 2025

Read Improved bounds for connection probabilities in random loop models by Volker Betz, Andreas Klippel, Julian Nauth (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-2503-17857.txt.

Introduction

Volker Betz, Andreas Klippel, and Julian Nauth sharpen results on random loop models, in a 2025 mathematics paper. Loop models scatter random closed curves on a lattice and stand in for quantum spin systems. A key question is when long loops appear, since that marks a change of phase in the system described.

The authors revisit a method called reflection positivity. By using extra freedom in how it is applied, they squeeze more out of it than before. They prove long loops exist across a wider range of settings, and they set new lower bounds for connection chances and the critical value where behavior flips. They compare their bounds with recent computer simulations, shedding light on the phase diagram of the spin systems behind the loops.

Catalog record

CREATORVolker Betz, Andreas Klippel, Julian Nauth
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-17857

Browse the full library · About the Tower of Babel

Full text

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

Improved bounds for connection probabilities in random loop models Volker Betz, Andreas Klippel, Julian Nauth Abstract We revisit and extend results by Ueltschi [ 19 ] on the application of reflection positivity to loop models with θ ∈ ℕ ≥ 2 \theta\in\mathbb{N}_{\geq 2} . By exploiting additional flexibility in the method, we prove the existence of long loops over a broader range of parameters u u and θ \theta , and establish new lower bounds for connection probabilities and the critical parameter β c \beta_{c} . Our results are compared with recent numerical simulations, providing further insight into the phase diagram of quantum spin systems. 1 Introduction In the last decades, many physical models have been investigated using probabilistic graphical representations. A class of such models are random loop models, which represent certain quantum spin systems. For the ferromagnetic quantum Heisenberg model, such a loop model was introduced by Powers [ 16 ] and then used by Tóth [ 18 ] to give a lower bound on the pressure. A different random loop model serves as a representation of the anti-ferromagnetic Heisenberg model, as found by Aizenman and Nachtergaele [ 1 ] . Ueltschi [ 19 ] observed that these are two instances of a one-parameter family of loop models representing XXZ-models. For the definition of this family we refer to section 2. This family of loop models is built using two independent Poisson processes on the space E\times[0,1)_{\text{per}} , where E are nearest neighbor edges. The first process, with intensity u\beta , generates crosses, while the second, with intensity (1-u)\beta , generates double bars. Starting from a vertex x at time t , we construct loops by moving upwards along the time axis until encountering either a cross or a double bar. Upon reaching a cross or double bar, we jump to the nearest right neighbor. If we encounter a double bar, we reverse direction along the time axis, whereas if we encounter a cross, we also change direction. We continue this process until we return to the starting point, forming a closed trajectory in V\times[0,1)_{\text{per}} . The connection to spin- S quantum systems requires an additional parameter \theta , which reweights the measure by a factor of \theta^{|\mathcal{L}|} with \theta=2S+1 . This introduces an additional dependency into the model, making its analysis particularly challenging. The main question of interest is whether a phase transition occurs or not. For graphs with a sufficiently high vertex degree, one expects a phase transition in the sense that there exists a critical (loop) parameter \beta_{c}>0 such that: 1. for \beta<\beta_{c} , only finite loops occur, and 2. for \beta>\beta_{c} , infinite loops appear almost surely. A necessary condition for (b) is the existence of an infinite percolation cluster induced by the Poisson processes. A remarkable result by Schramm [ 17 ] shows in the case u=1 and \theta=1 that on the complete graph K_{n} , where edge rates are \beta/n , the existence of infinite loop and the emergence of an infinite percolation cluster are asymptotically equivalent as n\to\infty , implying that their critical parameters coincide. Moreover, it is shown that the normalised ordered sizes of the loops in the giant component exhibit a Poisson-Dirichlet \operatorname{PD}(1) structure. For graphs with uniformly bounded degree, it was shown in [ 15 ] that the existence of infinite percolation is not a sufficient condition. In this setting, there exists a gap between critical values, for which a lower bound is given in [ 5 ] . A similar phenomenon was observed for a wide class of trees in [ 14 ] . In general to show (b) is a challenging problem. Nonetheless, there are some cases which are well understood. On regular trees, Angel [ 2 ] established the existence of two distinct phases for \theta=1 on d -regular trees with d\geq 4 and showed that infinite loops appear for \beta\in(d^{-1}+\tfrac{13}{6}d^{-2},\log(3)) when d is sufficiently large. Hammond [ 12 ] later proved that for d\geq 2 , there exists a critical value \beta_{0} above which infinite loops emerge in the case u=1 , a result extended by Hammond and Hegde [ 13 ] to all u\in[0,1] . Björnberg and Ueltschi provided an asymptotic expression for the critical parameter for loops up to second order in d^{-1} [ 7 ] and further extended this analysis to the case \theta\neq 1 [ 8 ] . A comprehensive picture for d -regular trees with d\geq 3 , \theta=1 , and u\in[0,1] was established by Betz et al. in [ 4 ] , proving a (locally) sharp phase transition for infinite loops in the interval (0,d^{-1/2}) and deriving explicit bounds on the critical parameter up to order 5 in d^{-1} . A first result for (b) on \mathbb{Z}^{d} for d\geq 5 in the case of \theta=1 and u=1 was given in [ 10 ] . For \theta\neq 1 and \theta\in\mathbb{N} , there are results proving the existence of a phase transition in dimensions three and higher using the method of reflection positivity introduced by Fröhlich, Simon, and Spencer for the classical Heisenberg model in [ 11 ] , and introduced to the quantum case by Dyson, Lieb, and Simon [ 9 ] . The latter corresponds to the quantum Heisenberg antiferromagnet, which is equivalent to the loop model with \theta=2 and u=0 . Ueltschi managed to rigorously adapt the method to the random loop model in [ 19 ] and demonstrated that it remains applicable for various values of \theta\in\mathbb{N}_{\geq 2} and u\in[0,1/2] . The purpose of this paper is to revisit end extend the results of [ 19 ] . We use some additional flexibility that is inherent in the method, but not exploited in [ 19 ] . This allows to prove long loops for a larger range of values \theta and u , and to obtain lower bounds on nearest-neighbor, finite-range, and long-range connections probabilities. We also give lower bounds for the critical value \beta_{c} and compare them with numerical studies from [ 3 ] . Outline of the remaining sections. In Section 2, we introduce the model and some of its basic properties. Moreover, we state our main theorem and provide its proof. In Section 3, we present all the consequences of our theorem and prove them. 2 Model and results The random loop model is well-established and precisely defined in various papers, such as in [ 19 ] and [ 4 ] . However, we provide here a brief overview of the main objects. The model can be defined on any graph G , but in this paper, we only consider the d -dimensional torus \mathcal{G}_{L}=(\Lambda_{L},\mathcal{E}_{L}) with side length L , where \Lambda_{L}\coloneqq\mathbb{Z}^{d}/L\mathbb{Z}^{d} is the set of vertices, and the edge set \mathcal{E}_{L} consists of nearest neighbors, with periodic boundary conditions. Let u\in[0,1] and \beta>0 . We attach the interval [0,\beta) to each edge of the graph, and will interpret t\in[0,\beta) as a time parameter. On the space \mathcal{E}_{L}\times[0,\beta) , we define two independent Poisson point processes: one with intensity u that produces so-called crosses, and another with intensity 1-u that produces double bars. A link is either a cross or a double bar. The combined configuration space is denoted by \Omega(G) , with

Read the complete text