Equality of the critical inverse temperatures for the one- and two-sided Dyson models
Noam Berger, Anders Johansson, and Anders Öberg · 2025
Read Equality of the critical inverse temperatures for the one- and two-sided Dyson models by Noam Berger, Anders Johansson, and Anders Öberg (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-2512-17451.txt.
Introduction
Noam Berger, Anders Johansson, and Anders Öberg settle a threshold question for the Dyson model. In that model, a chain of spins feels every other spin, with a grip that fades with distance. The team proves the freezing point is the same either way. Stated as a critical inverse temperature, it matches whether the chain stretches one way or both ways. The proof covers fade powers between 1 and 2. They conjecture the same match holds at the borderline power 2.
Catalog record
| CREATOR | Noam Berger, Anders Johansson, and Anders Öberg |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 17 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-17451
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Equality of the critical inverse temperatures for the one- and two-sided Dyson models Noam Berger, Anders Johansson, and Anders Öberg Abstract . We prove that the critical inverse temperatures β c ℕ ( α ) \beta_{c}^{\mathbb{N}}(\alpha) and β c ℤ ( α ) \beta_{c}^{\mathbb{Z}}(\alpha) for the one- and two-sided Dyson models are the same when the power of the interaction strength α \alpha satisfies 1 < α < 2 1<\alpha<2 . We conjecture that this is true also in the remaining case of α = 2 \alpha=2 . 2020 Mathematics Subject Classification Primary 37D35, 37A60, 82B20, 82B26, 82C27 1. Introduction There has been a recent interest in the study of the one-sided and two-sided Dyson models (or phrased differently: the Dyson model on the half-line and the line) and how these models are related. Consider the state space X={\{-1,+1\}}^{V} , for some set V . The one-sided and two-sided long range Ising–Dyson potentials are given by \Phi_{V}(x)=\sum_{i,j\in V}\frac{x_{i}x_{j}}{|i-j|^{\alpha}}, when V=\mathbb{N} and \mathbb{Z} , respectively. In Johansson, Öberg and Pollicott [ 6 ] , the random cluster model was used to prove the existence of a continuous eigenfunction for the transfer operator with respect to a continuous one-point Dyson potential \phi , \phi(x)=\beta\cdot x_{0}\sum_{k=1}^{\infty}\frac{x_{k}}{k^{\alpha}}. where \beta>0 and \alpha>1 . Note that we have \Phi_{V}(x)=\sum_{k\in V}\phi(T^{k}x) . In [ 6 ] it was proved that there exists a continuous eigenfunction for the transfer operator defined by the Dyson one-point potential if the one-sided Dyson model on \{-1,+1\}^{\mathbb{N}} has a continuous density with respect to the marginal of the two-sided model on \{-1,+1\}^{\mathbb{Z}} . This is shown to hold if the interaction strength parameter \alpha is greater than 3/2 and if the inverse temperature parameter \beta(\alpha) is less than the critical inverse temperature for the two-sided Dyson model: \beta_{c}^{\mathbb{Z}}(\alpha) . Since it follows from Griffiths inequality that the critical inverse temperature for the one-sided model, \beta_{c}^{\mathbb{N}}(\alpha) , is greater than or equal to \beta_{c}^{\mathbb{Z}}(\alpha) , one would like to show the existence of a continuous eigenfunction for \beta<\beta_{c}^{\mathbb{N}}(\alpha) , or to show that \beta_{c}^{\mathbb{Z}}(\alpha)=\beta_{c}^{\mathbb{N}}(\alpha) for all 1<\alpha\leq 2 . Such an equality is of course interesting in its own right. Note that Johansson, Öberg and Pollicott [ 5 ] proved that 8\beta_{c}^{\mathbb{Z}}(\alpha)\geq\beta_{c}^{\mathbb{N}}(\alpha) , for 1<\alpha\leq 2 . Here we prove that \beta_{c}^{\mathbb{Z}}(\alpha)=\beta_{c}^{\mathbb{N}}(\alpha),\text{ for all }1<\alpha<2, omitting the case when \alpha=2 . We prove our result in the context of the random cluster model, using a modification of a renormalization idea which originally appeared in [ 1 ] . We conjecture that our result is true also in the case when \alpha=2 . Since the appearance of [ 6 ] , the recent paper by van Enter, Fernández, Makhmudov and Verbitskiy [ 3 ] , work in a more general setting to prove the existence of a continuous eigenfuction (and hence Ruelle theorems) where the random cluster interpretation does not directly apply, but at the cost of assuming the Dobrushin uniqueness condition. However, Makhmudov has in [ 8 ] improved on the paper by van Enter et al. to include the full uniqueness region, in the sense of \beta<\beta_{c}^{\mathbb{Z}}\leq\beta_{c}^{\mathbb{N}} , as well as generalising the result in [ 6 ] , using concentration inequalities also used in [ 3 ] . He also considered in [ 7 ] the interesting question of the regularity of an eigenfunction of the transfer operator in the presence of an external field in the Dyson potential. Acknowledgement : Thanks to Jeff Steif for conversations! 2. The inverse critical temperatures are equal when 1<\alpha<2 For a graph G let \mathcal{C}(G)=\{C_{1},C_{2},\dots\} denote the partition of V(G) into clusters (connected components) and let \hat{C}(G) denote a cluster of maximum size. We use \omega(G)=|\mathcal{C}(G)| to denote the number of clusters. We consider (random) graphs with integer intervals V\subset\mathbb{Z} as vertex sets. The Bernoulli graph model \eta(G;p) is parameterised by a edge probability function p:V^{(2)}\to[0,1] and we define the model by having each edge ij\in V^{(2)} independently chosen with probability p(ij) . The edge probability function corresponding to the interaction strength function J(ij)\geq 0 , i,j\in V , is (1) p(ij)=1-\exp(-\beta\mathrm{J}(ij)),\quad i,j\in\mathbb{Z}. We obtain the Dyson model by setting (2) \mathrm{J}(ij)=|i-j|^{-\alpha},\quad i,j\in\mathbb{Z} for parameters \alpha>1 and \beta\geq 0 . We use p|_{V} to denote the restriction p(ij)\mathbf{1}_{\{i,j\}\subset V} of p to V^{(2)} . For an interval V\subset\mathbb{Z} and p=p_{\beta,\alpha} given by ( 1 ) and ( 2 ), we denote by \nu_{\beta}^{V}(G)=\mathsf{FK}_{q}(G;p|_{V}), the (Dyson-) random cluster model or FK-model ( [ 4 ] ) with q\geq 1 on V . For finite V , we obtain the FK-model from the corresponding (Dyson-) Bernoulli model \eta_{\beta}^{V}(G)=\eta(G;p|_{V}) by weighting each graph G with the factor q^{\omega(G)} . For infinite V , we obtain the (free-boundary) distribution as a weak limit of FK-models on finite sub-intervals. Thus \mu=\nu_{\beta}^{\mathbb{Z}} is the translation invariant two-sided model , while \nu=\nu_{\beta}^{\mathbb{N}} is what we call the one-sided model . Let \eta_{\beta}^{V}(H)=\mathsf{FK}_{1}(H;p) denote the Bernoulli graph model with edge probabilities p|_{V} . The event of percolation is the event that the random graph has an infinite component, i.e. event that |\hat{C}(G)|=\infty . The critical inverse temperature \beta_{c}^{V}\in[0,\infty] for the graph model \nu_{\beta}^{V} is \beta_{c}^{V}=\sup\{\beta:\nu^{V}_{\beta}(|\hat{C}(G)|=\infty)=0\}=\inf\{\beta:\nu^{V}_{\beta}(|\hat{C}(G)|=\infty)=1\}. Our aim in this section is to prove the following theorem of equality of critical temperatures for the one-sided and two-sided random cluster models. Theorem 1 . We have \beta_{c}^{\mathbb{N}}=\beta_{c}^{\mathbb{Z}} when 1<\alpha<2 . Remark 1 . It is well known that the value of the critical temperature is the same regardless of whether we consider free or wired boundary conditions. The choice of free boundary conditions makes our analysis easier, but does not restrict the generality of the result. 2.1. Proof of Theorem 1 We often assume that there is an underlying probability space (\Omega,\mathcal{F},\mathsf{P}) that carries the random graphs we consider. We consider graphs ordered under the subgraph relation H\subset G and a graph distribution \nu^{\prime}(H) is stochastically dominated by a graph distribution \nu(G) , written \nu^{\prime}(H)\prec\nu(G) or just H\prec G , if there exists a coupling \phi(H,G) such that H\subset G with probability one. Before proving the theorem, we state two relations regarding stochastic dominance (see [ 4 ] ). If V\subset W then (3) \nu_{\beta}^{V}(G)\prec\nu^{W}_{\beta}(G\mid\text{any condition on $G\setminus G[V]$}), where G[V] is the graph G induced on V . Furthermore, we have the “sprinkling relation” (4) \nu_{\beta}^{V}(G)\prec\nu_{\beta}^{V}\otimes\eta_{\delta}(G\cup H)\prec\nu^{V}_{\beta+2\delta}(\tilde{G}) where (G,H)\sim\nu_{\beta}\otimes\eta^{V}_{\delta} . By stochastic dominance ( 3 ), we have \beta_{c}^{\mathbb{N}}\geq\beta_{c}^{\mathbb{Z}} and thus it is enough to show that for percolation occur G\sim\nu_{\beta}^{\mathbb{N}} with positive probability if \beta>\beta_{c}^{\mathbb{Z}} . Furthermore, by the sprinkling relation ( 4 ), this statement follows if we show that percolation occur in the random graph G\cup H , where H\sim\eta^{\mathbb{N}}_{\delta} . We base the proof of Theorem 1 on the following lemma. For an integer interval I\subset V let (5) \mathcal{C}_{I}(G)=\{C\cap I:C\in\mathcal{C}(G)\} be the partition induced on I , where the elements are not necessarily connected in the graph G[I] induced on I . Let \hat{C}_{I}(G) be an element in \mathcal{C}_{I}(G) of maximum size. Lemma 2 . Assume 1<\alpha<2 and \beta>\beta_{c}^{\mathbb{Z}} and 1>\gamma>\alpha/2 . For any \epsilon>0 there exist arbitrary large integers N with the following property: For any integer interval I\subset\mathbb{N} of length N , taking G\sim\nu_{\beta}^{I} , we have (6) \mathsf{P}\left(|\hat{C}_{I}(G)|\geq N^{\gamma}\right)>1-\epsilon. Note that, by stochastic