---
title: 'Toroidal Aztec Diamond: Graphs & Dimer Models'
url: https://www.emergentmind.com/topics/toroidal-aztec-diamond
type: topic
---

# Toroidal Aztec Diamond: Graphs & Dimer Models

Searching arXiv for the specified papers on toroidal Aztec diamonds, periodic Aztec diamonds, and related dimer models.
The **toroidal Aztec diamond** has two closely related meanings in recent arXiv literature. In one, it is a finite bipartite toroidal graph obtained from an Aztec rectangle graph by identifying corresponding vertices on opposite boundaries, with the **toroidal Aztec diamond** as the special case \(T_{n,n}\) [2508.06451]. In the other, it denotes the Aztec diamond dimer model with **doubly periodic edge weights**, analyzed through a toroidal fundamental domain, a magnetically altered Kasteleyn matrix \(K_{G_1}(z,w)\), and the associated characteristic polynomial \(P(z,w)=\det K_{G_1}(z,w)\) [2508.04938]. These viewpoints are complementary: one emphasizes exact finite counting of perfect matchings under the insertion and evolution of holes, while the other emphasizes spectral, conformal, and geometric structure of periodic dimer models. A further integrable perspective treats doubly periodic Aztec-diamond weights through the octahedron equation and exact density recursions, yielding explicit arctic curves for toroidal weight patterns [1402.4493].

## 1. Graph-theoretic definition and basic combinatorial objects

In the finite toroidal graph framework, the starting point is the Aztec rectangle graph \(AR_{m,n}\), defined from a \((2m+1)\times(2n+1)\) chessboard with black corners by taking vertices at white squares and edges between vertices whose squares share a corner [2508.06451]. The **toroidal Aztec rectangle** \(T_{m,n}\) is obtained from \(AR_{m,n}\) by identifying corresponding vertices along top with bottom and left with right boundary; this yields a bipartite toroidal graph. The **toroidal Aztec diamond** is the special case \(T_{n,n}\) [2508.06451].

The same paper introduces holes in \(T_{m,n}\) as removed vertex sets inducing subgraphs isomorphic to **odd Aztec rectangles** \(O_{k,\ell}\). A hole is **white-placed** if its majority vertices are white, and **black-placed** otherwise. The associated **charge** is defined as the number of white vertices minus black vertices in the hole. For \(O_{k,\ell}\), the difference is \(k+\ell+1\), so
\[
q(O_{k,\ell})=+(k+\ell+1)
\]
if white-placed, and
\[
q(O_{k,\ell})=-(k+\ell+1)
\]
if black-placed [2508.06451]. Since \(T_{m,n}\) is balanced, removal of holes preserves the possibility of perfect matchings only when the total hole charge is zero.

The periodic weighted-dimer formulation instead begins with the tilted square lattice and imposes doubly periodic edge weights with vertical period \(k\) and horizontal period \(\ell\). One fixes a smallest non-repeating fundamental domain, passes to a torus, and introduces complex phases \(z,w\) along the two torus cycles in the magnetically altered Kasteleyn matrix \(K_{G_1}(z,w)\). The resulting characteristic polynomial
\[
P(z,w)=\det K_{G_1}(z,w)
\]
defines the spectral curve
\[
R^\circ=\{(z,w)\in (\mathbb{C}^*)^2: P(z,w)=0\},
\]
whose closure is a compact Riemann surface [2508.04938]. In that setting, the toroidal Aztec diamond is the periodic Aztec dimer model viewed through this toroidal spectral data.

These two definitions are not interchangeable, but they concern the same broader object class: Aztec-diamond dimers under toroidal identification or toroidal periodicity. This suggests that “toroidal Aztec diamond” is best understood as a family of toroidal dimer systems rather than a single canonical graph.

## 2. Holes, charge balance, and the natural evolution mechanism

A central feature of the finite toroidal theory is the insertion of odd Aztec-rectangle holes and their **natural evolution** [2508.06451]. For a white-placed hole \(O_{k,\ell}\) with \(\ell\ge 1\), the evolved form is \(e(O)=O_{k+1,\ell-1}\), centered at the same place. For a black-placed hole \(O_{k,\ell}\) with \(k\ge 1\), the evolved form is \(e(O)=O_{k-1,\ell+1}\), also with the same center. Base cases produce **separation defects**: if \(O\) is white-placed with shape \(O_{k,0}\), then \(e(O)\) is a vertical defect of \(k+1\) contiguous separations; if \(O\) is black-placed with shape \(O_{0,\ell}\), then \(e(O)\) is a horizontal defect of \(\ell+1\) contiguous separations [2508.06451].

The counting factor attached to this evolution is expressed through the **flank charge**. Using the width-based definition,
\[
f(O)=\ell
\]
for a white-placed \(O_{k,\ell}\), and
\[
f(O)=-(\ell+1)
\]
for a black-placed \(O_{k,\ell}\) [2508.06451]. An equivalent height-based formulation is also given, and the two agree when the total hole charge is zero.

This leads to the main evolution identity. If \(O_1,\dots,O_r\) are disjoint odd Aztec rectangle windows in \(T_{m,n}\) of total charge zero, and \(T_{m,n}\setminus (O_1\cup\cdots\cup O_r)\) has at least one perfect matching, then
\[
M(T_{m,n}\setminus O_1\cup\cdots\cup O_r)
=
2^{\sum_{i=1}^r f(O_i)}
\,M(T_{m,n}\setminus e(O_1)\cup\cdots\cup e(O_r)).
\]
This theorem is the paper’s “simple formula” governing evolution on the torus [2508.06451].

The proof uses the complementation theorem for perfect matchings of cellular graphs. On the torus, shaded cells are partitioned into vertical paths, some of which are rings. Rings contribute \(0\); the remaining paths contribute a type equal to the number of extremal vertices in the subgraph minus \(1\). Only vertical paths crossing holes contribute nontrivially. For white-placed \(O_{k,\ell}\), exactly \(\ell\) vertical paths contribute \(+1\); for black-placed \(O_{k,\ell}\), exactly \(\ell+1\) contribute \(-1\). This produces the exponent \(\sum f(O_i)\), while the complementation operation itself geometrically transforms the holes into their evolved forms [2508.06451].

A further lemma states that if the original holes are disjoint and the punctured toroidal graph admits a perfect matching, then their evolved forms remain disjoint [2508.06451]. This ensures that the evolution process can be iterated.

## 3. Reduction to diagonal multiplets and flip symmetry

The evolution formalism yields several structured consequences. For congruent windows, if there are \(s\) white-placed windows of shape \(O_{k,\ell}\) and \(s\) black-placed windows of shape \(O_{\ell,k}\), then
\[
M(T_{m,n}\setminus O_1\cup\cdots\cup O_{2s})
=
2^{s(\ell-k-1)}
\,M(T_{m,n}\setminus O_1^{+,-}\cup\cdots\cup O_s^{+,-}\cup O_{s+1}^{-,+}\cup\cdots\cup O_{2s}^{-,+}),
\]
with the evolved shapes defined in the paper [2508.06451].

A more striking consequence is the reduction to **diagonal multiplets**. A horizontal \(k\)-multiplet is a hole of shape \(O_{0,k-1}\), and a vertical \(k\)-multiplet is a hole of shape \(O_{k-1,0}\). For a general \(O_{k,\ell}\), the paper defines \(h(O)\) as the horizontal diagonal multiplet of length \(k+\ell+1\), and \(v(O)\) as the vertical diagonal multiplet of the same length [2508.06451]. Then, for \(s\) white-placed \(O_{k,\ell}\) holes and \(s\) black-placed \(O_{\ell,k}\) holes,
\[
M(T_{m,n}\setminus O_1\cup\cdots\cup O_{2s})
=
2^{-s\ell k}
\,M(T_{m,n}\setminus h(O_1)\cup\cdots\cup h(O_s)\cup v(O_{s+1})\cup\cdots\cup v(O_{2s})).
\]
Thus systems of congruent odd Aztec-rectangle windows can be “straightened” into diagonal slits of equal length, at the cost of an explicit power of \(2\) [2508.06451].

Corollary 4.4 of the same paper establishes a nontrivial **flip symmetry**. Let \(O_1,\dots,O_{2s}\) be diagonal multiplets of length \(k\), with \(s\) horizontal white-placed and \(s\) vertical black-placed, and suppose at least one perfect matching exists. If \(\overline{O}\) denotes the \(90^\circ\)-rotated multiplet at the same center, then
\[
M(T_{m,n}\setminus O_1\cup\cdots\cup O_{2s})
=
M(T_{m,n}\setminus \overline{O}_1\cup\cdots\cup \overline{O}_{2s}).
\]
In the language of finite-size correlation, flipping every diagonal slit by \(90^\circ\) leaves the correlation invariant [2508.06451].

The paper emphasizes that this symmetry is non-obvious and calls for a direct proof not based on complementation. This identifies an explicit combinatorial invariant of toroidal slit configurations and clarifies that geometrically distinct defect systems may have identical matching enumerations.

## 4. Correlation functions and the dual Aztec diamond theorem

The finite toroidal setting also supports correlation functions for holes. For holes \(O_1,\dots,O_k\) of total charge zero placed on \(T_{m,n}\), the **finite size correlation** is defined by
\[
\omega_{m,n}(O_1,\dots,O_k):=
\frac{M(T_{m,n}\setminus O_1\cup\cdots\cup O_k)}{M(T_{m,n})},
\]
and the **infinite correlation** is
\[
\omega(O_1,\dots,O_k):=
\lim_{n\to\infty}
\frac{M(T_{n,n}\setminus O_1\cup\cdots\cup O_k)}{M(T_{n,n})}
\]
[2508.06451].

Assuming the **electrostatic conjecture** for square-grid dimers, the paper derives a relation between single-cluster correlations of odd Aztec rectangles and diagonal multiplets:
\[
\omega(O_{k,\ell}) = 2^{-k\ell/2}\,\omega(O_{0,k+\ell}).
\]
In particular, for the odd Aztec diamond \(O_{n,n}\),
\[
\omega(O_{n,n}) = 2^{-n^2/2}\,\omega(O_{0,2n}).
\]
The paper describes this as a natural **dual of the classical Aztec diamond theorem** [2508.06451]. The classical theorem counts interior tilings of the ordinary Aztec diamond \(AD_n\) by
\[
M(AD_n)=2^{n(n+1)/2},
\]
whereas the toroidal hole-correlation statement gives a power-of-two law for an exterior accommodation problem.

This duality should be interpreted carefully. It is not an equality between matching numbers of ordinary Aztec diamonds and toroidal hole systems. Rather, it is a correspondence in structural form: both statements produce clean powers of \(2\), but in different enumerative regimes.

The same paper recalls explicit values for \(\omega(O_{0,m})\) from prior exact results on diagonal monomer runs, allowing closed forms for \(\omega(O_{k,\ell})\) to be obtained by substitution [2508.06451]. A plausible implication is that odd Aztec rectangles occupy a distinguished place among toroidal defects because their correlations reduce to previously tractable diagonal objects.

## 5. Doubly periodic weighted models, spectral curves, and conformal structure

A different, more analytic theory studies the toroidal Aztec diamond through periodic edge weights and spectral geometry. In this setting, one fixes a smallest non-repeating fundamental domain of the tilted square lattice, imposes periodicity with periods \(k\) and \(\ell\), and considers the magnetically altered Kasteleyn matrix \(K_{G_1}(z,w)\) on the torus [2508.04938]. Its determinant
\[
P(z,w)=\det K_{G_1}(z,w)
\]
defines the spectral curve, whose amoeba and Ronkin function encode the phase diagram and the limiting height slopes.

The Ronkin function is
\[
N(x,y)=\frac{1}{(2\pi)^2}\int_0^{2\pi}\int_0^{2\pi}
\log |P(e^{x+i\theta},e^{y+i\phi})|\,d\theta\,d\phi,
\]
and its gradient \(\nabla N(x,y)\) gives slopes of the limiting height function [2508.04938]. The liquid region is controlled by the **Kenyon–Okounkov conformal structure**, realized through a critical point map \(\Omega\) that sends liquid points to points of the spectral curve where the action function is stationary [2508.04938].

For \(1\times \ell\) periodic weights, the paper gives a scalar-symbol factorization
\[
\Phi(z)=\prod_{i=1}^{\ell}\frac{\gamma_i+\alpha_i z^{-1}}{1-\beta_i z^{-1}},
\qquad
P(z,w)=\prod_i(1-\beta_i z^{-1})(\Phi(z)-w),
\]
so the spectral curve has genus \(0\) [2508.04938]. For \(2\times \ell\) periodic weights, the symbol becomes \(2\times 2\), the spectral curve is generically of genus \(g=\ell-1\), and compact real ovals correspond to gas components [2508.04938].

This framework makes precise that toroidal periodicity is not merely a boundary identification; it is a spectral construction. The torus supports Fourier variables \(z,w\), and these variables control both macroscopic phases and conformal data. In this sense, the toroidal Aztec diamond is a model in which discrete dimer combinatorics and complex-algebraic geometry are tightly coupled.

## 6. Perfect \(t\)-embeddings, maximal surfaces, and phase geometry

The same periodic-weight paper develops **perfect \(t\)-embeddings** and **origami maps** for Aztec diamonds with periodic edge weights [2508.04938]. Given Coulomb gauge functions \(F^\bullet\) on black vertices and \(F^\circ\) on white vertices satisfying Kasteleyn-harmonic equations, one defines closed edge-forms on the augmented dual graph:
\[
dT(bw^*) := F^\bullet(b)K_R(b,w)F^\circ(w),
\qquad
dO(bw^*) := F^\bullet(b)K_R(b,w)\overline{F^\circ(w)}.
\]
These integrate to an embedding \(T\) and an origami map \(O\), forming a **\(t\)-surface** [2508.04938].

Viewing
\[
X=(\Re T,\Im T,\Re O,\Im O)\in \mathbb{R}^4,
\]
the paper equips \(\mathbb{R}^4\) with the Minkowski metric of signature \((2,2)\),
\[
\langle X,Y\rangle_{2,2}=X_1Y_1+X_2Y_2-X_3Y_3-X_4Y_4.
\]
The induced quadratic form is \(|dT|^2-|dO|^2\). Space-like means this is positive; light-like means it is zero [2508.04938].

The main convergence theorem states that for Aztec diamonds of size \(k\ell N\) with \(k\times \ell\) periodic weights, the perfect \(t\)-embedding and origami map converge uniformly on compact subsets of the liquid region to a limiting surface \((Z\circ \Omega,\vartheta\circ \Omega)\). For \(k=1\), the limiting surface lies entirely in \(\mathbb{R}^{2,1}\). For \(k=2\), gas is allowed, and the limiting surface is **space-like and maximal in \(\mathbb{R}^{2,2}\)** [2508.04938].

The phase geometry has a particularly rigid form. All frozen regions collapse to four boundary points, regardless of the number of frozen regions, while each gas region collapses to a distinct light-like cusp in the interior [2508.04938]. The four boundary vertices are determined by a single parameter \(a\), extracted from inverse Kasteleyn data; in periodic Aztec settings this is \(a=\sqrt{\gamma_1\beta_\ell/\alpha_1}\) for \(k=1\) and \(a=\sqrt{\alpha_1/\beta_\ell}\) for \(k=2\) [2508.04938].

The paper further states that cusp locations depend on the spectral divisor parameter \(t\), whereas the conformal structure \(\Omega\) does not. This separation between conformal structure and cusp position leads to the conjecture that cusp locations encode the discrete Gaussian shift appearing in global fluctuations of toroidal dimers [2508.04938]. This suggests that the toroidal Aztec diamond carries geometric information beyond the standard limit shape.

## 7. Integrable toroidal weights, density recursions, and arctic curves

A third perspective comes from the octahedron equation. For Aztec-diamond dimer coverings with doubly periodic face weights, the partition function satisfies the \(A_\infty\) \(T\)-system
\[
T_{i,j,k+1}T_{i,j,k-1}
=
T_{i+1,j,k}T_{i-1,j,k}
+
T_{i,j+1,k}T_{i,j-1,k},
\]
with flat stepped-surface initial data \(T_{i,j,i+j+1\!\!\!\!\pmod 2}=t_{i,j}\) [1402.4493]. The solution \(T_{i,j,k}\) equals the dimer partition function \(Z_{i,j,k}\), so periodic initial data define toroidal dimer weights on Aztec graphs [1402.4493].

Differentiating the \(T\)-system yields a density recursion with periodic coefficients:
\[
\rho_{i,j,k+1}^{(\epsilon,\eta)}+\rho_{i,j,k-1}^{(\epsilon,\eta)}
=
L_{i,j,k}\bigl(\rho_{i+1,j,k}^{(\epsilon,\eta)}+\rho_{i-1,j,k}^{(\epsilon,\eta)}\bigr)
+
R_{i,j,k}\bigl(\rho_{i,j+1,k}^{(\epsilon,\eta)}+\rho_{i,j-1,k}^{(\epsilon,\eta)}\bigr),
\]
where
\[
L_{i,j,k}=\frac{T_{i+1,j,k}T_{i-1,j,k}}{T_{i,j,k+1}T_{i,j,k-1}},
\qquad
R_{i,j,k}=1-L_{i,j,k}
\]
[1402.4493]. For toroidal initial data, \(L\) and \(R\) are periodic, so the infinite recursion reduces to a finite linear system for generating functions.

In the uniform case \(t_{i,j}=1\), one recovers
\[
T_{i,j,k}=2^{k(k-1)/2},
\]
with \(L=R=1/2\), and singularity analysis of the corresponding generating function yields the arctic circle [1402.4493]. For the period-\(2\times 2\) case, the paper gives an exact factorized solution, a \(4\times 4\) linear system for density generating functions, and a denominator \(D(x,y,z)\) whose blow-up at \((1,1,1)\) produces the **fortress** arctic curve \(P_\alpha(u,v)=0\) [1402.4493]. The parameter
\[
\alpha=16\sigma(1-\sigma)\tau(1-\tau)
\]
controls the deformation: \(\alpha=1\) reproduces the arctic circle, whereas \(\alpha=0\) yields the inscribed square [1402.4493].

For general \(m\)-toroidal initial data, the paper derives an exact factorized solution in terms of periodic sequences \(a_i,b_i,c_i,d_i\), explicit cross-ratios \(L,R\), and a \(4m\times 4m\) linear system whose determinant governs the arctic curve [1402.4493]. Generically, the curves exhibit multiple inner facet regions, with the number and arrangement depending on the toroidal periodicity and parameters. This provides an integrable derivation of multi-facet limit shapes for toroidal Aztec-diamond dimers.

This approach differs from the Kasteleyn–spectral-curve framework of periodic toroidal dimers, but the paper explicitly notes that \(\det M\) plays the role of a characteristic polynomial or spectral curve in the density analysis [1402.4493]. A plausible implication is that the toroidal Aztec diamond admits several mathematically equivalent “spectral” encodings, depending on whether one studies matchings, densities, or large-scale geometry.

## 8. Examples, scope, and open directions

Several explicit examples illustrate the theory. In the toroidal hole framework, for \(T_{8,10}\) with four length-\(2\) diagonal slits—two horizontal white-placed and two vertical black-placed—flipping all slits preserves the count; direct evaluation yields
\[
\omega_{8,10}(\text{four slits})=
\frac{(\pi-4)(54\pi^2-339\pi+532)}{54\pi^4},
\]
and the flipped configuration has the same value [2508.06451]. In the planar context surrounding the toroidal theory, the same paper contrasts a non-round Aztec window count with “round” product forms for odd windows, such as an \(AR_{23,18}\) region with an \(O_2\) window [2508.06451].

In the periodic-geometry framework, the two-periodic \(2\times 2\) model provides a concrete genus-\(1\) example. Four weight assignments are listed, with corresponding values of the parameter \(a\) and divisor parameter \(t\). The liquid region has one gas hole, hence one cusp apex. The paper states that weights 1 and 2 produce surfaces in \(\mathbb{R}^{2,1}\), while weights 3 and 4 produce surfaces not embeddable in \(\mathbb{R}^{2,1}\), even locally near the cusp [2508.04938].

Across these works, several open problems and interpretive themes recur. One is the **electrostatic conjecture**, which is assumed in the derivation of the dual Aztec diamond relation for single-cluster correlations [2508.06451]. Another is the call for a direct proof of the flip symmetry of diagonal slits [2508.06451]. In the geometric direction, the conjecture that cusp locations encode the discrete Gaussian shift parameter \(t\) remains an explicit research question [2508.04938].

Taken together, these results place the toroidal Aztec diamond at the intersection of exact enumeration, periodic dimer spectral theory, integrable recurrences, and Lorentzian surface geometry. In finite toroidal graphs, its structure is governed by charge balance and complementation-driven hole evolution. In doubly periodic weighted models, it is organized by Kasteleyn spectral curves, Ronkin functions, and perfect \(t\)-embeddings. In integrable formulations, it yields exact density systems and algebraic arctic curves. The common theme is that toroidal periodicity does not merely modify boundary conditions; it introduces new invariants—charge, spectral divisor, cusp data, and periodic cross-ratios—that fundamentally alter both combinatorics and asymptotic geometry [2508.06451] [2508.04938] [1402.4493].

Source: https://www.emergentmind.com/topics/toroidal-aztec-diamond