Dimers with layered disorder

2025

Read Dimers with layered disorder (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-2507-11964.txt.

Introduction

Quentin Moulard and Fabio Toninelli study dimers, pairs of neighboring tiles on a square grid, when the pairing weights are random but arranged in layers. The layered randomness copies the McCoy-Wu model that famously changed how disorder was understood in magnetism. In the clean model, the middle liquid phase has a smooth free energy and correlations that fade like inverse distance squared. Writing in twenty twenty-five, the authors show the layered disorder roughens everything. The free energy gains an essential singularity, a infinitely sharp but smooth-looking kink, and correlations decay instead in a stretched exponential way.

At the boundary where the clean model passes from gas-like to liquid behavior, the classic transition exponent is smeared. Disorder bends it into a range that varies continuously upward from its clean value. The results warn that one-dimensional-style randomness can rewrite phase behavior in two-dimensional systems.

Catalog record

YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE12 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-11964

Browse the full library · About the Tower of Babel

Full text

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

Dimers with layered disorder Abstract We study the dimer model on the square grid, with quenched random edge weights. Randomness is chosen to have a layered structure, similar to that of the celebrated McCoy-Wu disordered Ising model. Disorder has a highly non-trivial effect and it produces an essential singularity of the free energy, with e − distance e^{-\sqrt{{\rm distance}}} decay of dimer-dimer correlations, at a point of the “liquid” (or “massless”) phase where the homogeneous dimer model has instead a real analytic free energy and correlations decaying like 1 / ( distance ) 2 1/({\rm distance})^{2} . Moreover, at a point where the homogeneous model has a transition between a massive (gaseous) and massless (liquid) phase, the critical exponent 3 / 2 3/2 (Pokrovsky-Talapov law), characteristic of the transition between the two regimes, is modified by disorder into an exponent that ranges continuously between 3 / 2 3/2 and infinity. Quentin Moulard, Fabio Toninelli TU Wien, Austria Key words and phrases. Dimer model; quenched disorder; disorder relevance; McCoy-Wu model; Lyapunov exponent; random matrix product. 1 Introduction Disorder can have a drastic, non-perturbative effect on phase transitions: even an arbitrarily small amount of quenched, random impurities can discontinuously change critical exponents or remove phase transitions altogether, especially in low-dimensional systems. Famous examples include: the 2D Random Field Ising Model, where a random magnetic field with arbitrarily small variance eliminates the ferromagnetic phase transition [ ImryMa , AW ] , and induces exponential decay of correlations [ Ding , Aizenman ] ; the 2D Ising model again, where random ferromagnetic couplings conjecturally make the ferromagnetic transition smoother than in the homogeneous case [ DD , DD1 ] , turning the \mathop{\mathrm{log}}\nolimits[1/(\beta-\beta_{c})] divergence of the specific heat into a \mathop{\mathrm{log}}\nolimits\mathop{\mathrm{log}}\nolimits[1/(\beta-\beta_{c})] one, and the spin-spin critical exponent from 1/4 to 0 ; the pinning/wetting model, where disorder has a smoothing effect and makes the localization transition always at least second order [ giacomin2006smoothing ] , in contrast with the homogeneous case; the lattice Gaussian Free Field, where interaction with a quenched disordered substrate turns the wetting (localization) transition from first to second order in dimension d\geq 3 [ HubertGB-GFF ] and from (almost) first to infinite order in dimension d=2 [ HubertGFF ] ; the directed polymer in random environment, where in low dimension ( d=1,2 ) the polymer is “localized” for any arbitrarily small disorder strength [ Comets ] while in dimension d\geq 3 an infinite-order phase transition between weak and strong disorder phases appears [ lacoin2025localization ] . Despite these examples and certain general guiding heuristic principles such as the “Harris criterion” [ harris ] , the question of relevance (in the Renormalization Group sense) of quenched randomness on phase transitions is still far from understood in general. In this work, we address this question in the framework of the dimer model [ Kasteleyn , mccoy1973two , Gorin ] on the (periodized) square grid \mathbb{Z}^{2} . In contrast with the Ising model, that in its homogeneous version has a single critical point separating the uniqueness and phase transition region, the phase diagram of the dimer model exhibits a whole critical region where correlations decay polynomially (like the inverse of the distance squared) and the height function scales to the Gaussian Free Field (GFF) [ kenyon2001dominos ] . In the dimer model literature [ KOS ] , this is called “liquid” or “massless” phase. The model exhibits also massive (or “gaseous”) phases where correlation decay is exponential, as well as “frozen” regions where correlations are zero. The transition between liquid and either gaseous or frozen regions is characterized by a critical exponent 3/2 (Pokrovsky-Talapov law [ PT , KO , Duse ] ). Our goal is to understand how randomness affects this picture. The type of randomness we introduce is inspired by the McCoy-Wu 2D random Ising model [ McCoyWu196 , mccoy1973two ] , where ferromagnetic spin-spin couplings are random but constant along rows of \mathbb{Z}^{2} . Similarly, we randomize the edge weights of the dimer model, in a layered way. For the McCoy-Wu model it is expected that disorder turns the second-order phase transition at \beta_{c} into an infinite-order one (for recent progress, supporting but not fully proving this conjecture, see [ comets2019continuum , Quentin ] . Some of the ideas we use in this work were developed by one of us in the latter reference). For the dimer model, the effect of disorder is in a sense even more surprising. Our first result (Theorem ) shows that an infinite-order phase transition appears at a point of the phase diagram where the free energy of the homogeneous model is real analytic. At this particular point, dimer-dimer correlations of the disordered model decay stretched-exponentially fast (approximately like the exponential of minus the square root of the distance ) in the direction perpendicular to the disorder layering (Theorem ). We emphasize that this effect is present even for an arbitrarily small disorder strength, and that it has no analog for the homogeneous model, where correlations decay either polynomially or exponentially anywhere in the phase diagram. It is interesting to mention that an analogous \mathop{\mathrm{exp}}\nolimits(-\sqrt{{\rm distance}}) decay has been predicted in the physics literature [ Shankar ] for energy-energy correlations of the McCoy-Wu model. Secondly, the phase diagram of the disordered dimer model (that is parametrized by two “magnetic fields” H_{1},H_{2} that suitably fix the edge occupation densities) exhibits regions, that have no analog for the homogeneous model, where the free energy is affine with respect to just one of the two magnetic fields (Theorem and Remark ). Yet another striking finding is that, at a specific point of the phase diagram where the homogeneous model is at the boundary between liquid and gaseous phases, the critical exponent 3/2 is modified into a continuously varying exponent, that ranges between 3/2 and +\infty (Theorem ). On the other hand, other features of the homogeneous dimer model, such as the existence of a massless phase with correlations decaying like the inverse squared distance, and the 3/2 exponent at the liquid/frozen boundary, turn out to be robust to the presence of disorder (Theorems and ). Technically, a first step of our analysis is to rewrite, via Kasteleyn’s theory, both free energy and correlations of the disordered dimer model in terms of a product of an i.i.d. sequence of 2\times 2 complex random matrices \{M^{\theta}_{y}\}_{y\in\mathbb{N}} , with \theta\in[0,\pi/2] a Fourier (momentum) variable. In particular, the free energy can be expressed through an integral over \theta involving the top Lyapunov exponent {\rm Lyap}(M^{\theta}) . As realized earlier on by Derrida-Hilhorst [ DH ] and Nieuwenhuizen-Luck [ Luck ] , the top Lyapunov exponent has a very singular behavior in situations where the matrices M_{y} “almost commute”, which is especially relevant at the point of essential singularity and at the gas-liquid transition point. Obtaining sharp bounds on {\rm Lyap}(M) in such situations is a notoriously subtle issue, and we rely on recent beautiful developments in this area, in particular [ collin2024large , de2024scaling , collin2025lyapunov ] and [ GGG , Havret ] , the latter making the Derrida-Hilhorst predictions rigorous. In addition, we prove in Section several crucial monotonicity properties of the Lyapunov exponent, seen as a function defined on the

Read the complete text