Aztec Window Domino Tilings

• Aztec Window is a domino-tiling region derived from an Aztec diamond with a central hole, challenging the classical exact product formulas.
• It serves as a test case for advanced combinatorial methods, including non-intersecting lattice paths, Pfaffians, and Kasteleyn theory.
• Variants like odd Aztec-rectangle windows and toroidal configurations yield exact tiling formulas with determinant and product structures.

Searching arXiv for recent and foundational papers on “Aztec window” and related Aztec-diamond tilings. {"query":"all: \"Aztec window\" OR title:\"Aztec\" domino tilings hole Aztec diamond", "max_results": 10} An Aztec window is a domino-tiling region obtained by taking an Aztec diamond and removing a smaller Aztec-diamond-shaped hole from its center. In the standard notation of the cited literature, the Aztec diamond of order $n$ is the union of unit squares associated with

$AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$

intersected with the integer lattice, and the classical Aztec diamond theorem gives

$M(AD(n))=2^{\frac{n(n+1)}{2}}.$

The study of Aztec windows begins by asking how this exact product formula changes when a central hole is introduced. The answer is subtle: the original window problem appears not to admit a simple product formula, while several closely related families—especially Aztec rectangles with odd Aztec-rectangle holes, symmetry-constrained diamonds with defects, and toroidal analogues—do admit exact Pfaffian, determinant, or product expressions (Ciucu, 8 Aug 2025, Lee, 2024).

1. Classical definition and historical setting

The classical Aztec window is an Aztec diamond with an Aztec-diamond-shaped hole. Interest in such regions arose shortly after the 1992 Aztec diamond theorem of Elkies, Kuperberg, Larsen, and Propp. In the formulation emphasized in recent work, the outer region is $AD_n$ and the hole is a smaller centrally placed $AD_m$ (Ciucu, 8 Aug 2025).

This problem is naturally framed in the language of domino tilings and perfect matchings. An Aztec diamond is balanced, so it admits many domino tilings. By contrast, the odd Aztec diamond $O_n$, obtained by stacking $2n+1$ strips of lengths

$1,3,5,\dots,2n-1,2n+1,2n-1,\dots,1,$

is not tileable by dominoes, because it has a color imbalance of $2n+1$. That imbalance becomes useful in generalized window problems, because Aztec rectangles themselves can carry a compensating imbalance (Ciucu, 8 Aug 2025).

The combinatorial importance of Aztec windows comes from the tension between two facts. First, the unpunctured Aztec diamond has one of the simplest exact tiling formulas in enumerative combinatorics. Second, introducing a central hole preserves much of the geometric symmetry while drastically altering the matching structure. This makes Aztec windows a canonical test case for methods based on Kasteleyn theory, non-intersecting lattice paths, Schur-function identities, and the complementation theorem.

2. Why the original window problem is difficult

The central empirical observation about classical Aztec windows is that their tiling numbers display regularity without becoming “round.” In the terminology of the recent literature, a round enumeration is a simple product formula, typically a power of $2$ times factorial or Vandermonde-type factors. Classical Aztec windows do not seem to behave that way (Ciucu, 8 Aug 2025).

A representative example is the Aztec diamond $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$0 with a central $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$1-shaped hole. Its number of domino tilings is

$AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$2

The appearance of large prime squares is one of the standard indicators that no simple product formula is presently visible in this family (Ciucu, 8 Aug 2025).

This negative phenomenon is significant because it separates Aztec windows from the original Aztec diamond theorem. It also motivates the search for nearby classes of punctured Aztec regions whose tiling numbers are still exactly computable in a structured way. Much of the modern theory can be understood as identifying which deformations of the window problem preserve product structure and which force genuinely more complicated enumerative behavior.

3. Round variants: odd Aztec-rectangle windows

A major development is the replacement of a central Aztec-diamond hole by odd Aztec rectangles placed inside Aztec rectangles rather than Aztec diamonds. In this setting, a large class of regions does admit round formulas (Ciucu, 8 Aug 2025).

The basic mechanism is to start with an Aztec rectangle region $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$3 and remove holes of shapes $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$4, where $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$5 is an odd Aztec rectangle. For suitable balancing conditions, the resulting region is tileable. Four principal families arise, depending on whether the frame is wider or taller and on the parity of the relevant parameter. In the simplest symmetric case, with $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$6 even, $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$7, and $AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$8, one has

$AD(n)=\{(x,y)\in \mathbb{R}^2:\ |x|+|y|\le n+1\}$9

where the right-hand side involves a previously enumerated line-hole graph $M(AD(n))=2^{\frac{n(n+1)}{2}}.$0 (Ciucu, 8 Aug 2025).

Analogous formulas hold for the other three parity-and-geometry cases, replacing $M(AD(n))=2^{\frac{n(n+1)}{2}}.$1 by $M(AD(n))=2^{\frac{n(n+1)}{2}}.$2, $M(AD(n))=2^{\frac{n(n+1)}{2}}.$3, and $M(AD(n))=2^{\frac{n(n+1)}{2}}.$4. The importance of these reductions is that the graphs $M(AD(n))=2^{\frac{n(n+1)}{2}}.$5 already have explicit product formulas. Consequently, the odd-window regions inherit exact tiling formulas that are powers of $M(AD(n))=2^{\frac{n(n+1)}{2}}.$6 times factorial and Vandermonde-type products (Ciucu, 8 Aug 2025).

This family is often described as solving a problem “very close” to the original Aztec window problem. The outer boundary is still Aztec-like, the holes are centrally organized, and the regions retain strong bilateral symmetry. What changes is the charge balance. That balance is precisely what allows the combinatorics to collapse to round expressions.

4. Pfaffian and path formulations for window-type constraints

A parallel line of work studies symmetry-constrained Aztec diamonds with defects. Although the phrase “Aztec Window” is not used there as a formal definition, the paper on off-diagonally symmetric domino tilings explicitly presents these objects as relevant to Aztec Window–type questions because they combine global symmetry with local constraints on a boundary or symmetry axis (Lee, 2024).

For off-diagonally symmetric tilings with boundary defect set $M(AD(n))=2^{\frac{n(n+1)}{2}}.$7, the count is given by a Pfaffian: $M(AD(n))=2^{\frac{n(n+1)}{2}}.$8 where $M(AD(n))=2^{\frac{n(n+1)}{2}}.$9 is an infinite skew-symmetric matrix whose entries satisfy a Delannoy-type recurrence (Lee, 2024). In odd order, a main theorem states that

$AD_n$0

so the number of off-diagonally symmetric tilings is invariant under reflection of the defect position across the boundary midpoint (Lee, 2024).

A second family, “nearly” off-diagonally symmetric tilings, allows exactly one nonzero diagonal cell. Its total count is also Pfaffian: $AD_n$1 with $AD_n$2 built from the same matrix $AD_n$3 together with a Pell-type sequence $AD_n$4 satisfying

$AD_n$5

These counts satisfy linear relations governed by matrices whose entries are Delannoy numbers: $AD_n$6 The same paper also conjectures log-concavity and entropy behavior for these symmetric classes (Lee, 2024).

Methodologically, this literature places Aztec-window phenomena squarely inside the standard toolbox of domino enumerations: non-intersecting lattice paths, Stembridge-type Pfaffian formulas, determinant-to-Pfaffian transitions caused by paired endpoints, and explicit matrix recurrences. A plausible implication is that many punctured or defected Aztec regions are best viewed not as isolated counting problems but as manifestations of a shared Pfaffian path geometry.

5. Toroidal windows, correlations, and duality

The 2025 work on round Aztec windows also develops a toroidal theory. One starts with a toroidal Aztec rectangle $AD_n$7 and removes odd Aztec-rectangle holes $AD_n$8 whose total charge is zero, so that perfect matchings remain possible (Ciucu, 8 Aug 2025).

The central structural result is an evolution law for holes. If $AD_n$9 is the naturally evolved hole and $AD_m$0 is its flank charge, then

$AD_m$1

Thus the effect of evolving all holes is governed by a pure power of $AD_m$2 determined only by flank charges (Ciucu, 8 Aug 2025).

Iterating this evolution reduces odd Aztec-rectangle holes to diagonal multiplets, or slits. In a symmetric configuration consisting of $AD_m$3 horizontal white-placed slits and $AD_m$4 vertical black-placed slits, the finite-size matching number is invariant under flipping every slit from horizontal to vertical or vice versa: $AD_m$5 This produces an unexpected symmetry in the correlation of diagonal slits (Ciucu, 8 Aug 2025).

The same framework yields what is described as a natural dual of the Aztec diamond theorem. For the correlation $AD_m$6 of an odd Aztec rectangle viewed as a hole in the infinite-volume limit,

$AD_m$7

and in particular

$AD_m$8

Here the comparison is between the correlation cost of an odd Aztec diamond and that of a diagonal slit of equal total length (Ciucu, 8 Aug 2025).

These results connect Aztec windows to the electrostatic picture of dimer correlations. The cited paper states the electrostatic conjecture in the form

$AD_m$9

and shows that equal-height odd Aztec rectangles aligned along a diagonal satisfy the conjecture in the weak sense, via exact reduction to monomer-run correlations (Ciucu, 8 Aug 2025).

6. Probabilistic and asymptotic uses of the “window” viewpoint

In a different usage, “Aztec window” can denote a local observation window inside a large random tiling. This is not a formal definition of the classical Aztec window region, but it has become a useful interpretive language in the asymptotic theory of domino tilings (Duits et al., 2017, Chhita et al., 2012).

For weighted Aztec diamonds, local windows in different macroscopic regions exhibit different limiting processes. Near the northern frozen boundary, the southern-domino process converges to a thinned Airy point process; near the southern boundary, the process of holes converges to a thickened Airy point process; and in the unfrozen region, local statistics converge to the translation-invariant Gibbs measure $O_n$0 (Chhita et al., 2012).

For the two-periodic Aztec diamond, a local window can lie in the solid, liquid, or gas phase. In the gas region, the limiting kernel is an explicit translation-invariant gas kernel with exponential decay of correlations; in the liquid region, the local process is of rough-phase dimer type; and at cusp points of the liquid-gas boundary, after subtracting the gas background and using anisotropic scaling, the rescaled kernel converges to a Pearcey-type kernel (Duits et al., 2017). This literature effectively treats a “window” as a microscope on phase structure.

A related restriction phenomenon appears for Aztec half-diamonds and Novak half-hexagons. The half-diamond limit shape is the restriction of the Aztec-diamond limit shape, and the corresponding half-hexagon inherits an “arctic parabola” as a corollary of the Arctic Circle theorem (Nordenstam et al., 2011). This suggests that subregions, punctures, and slices of Aztec-type tilings can often be analyzed by restricting or projecting a better-understood global limit shape.

The broader significance of Aztec windows lies in exactly this dual role. As planar regions with holes, they test the limits of exact dimer enumeration. As local windows in random tilings, they provide a language for phase-sensitive asymptotics, from Airy and Gibbs limits to gas kernels and Pearcey cusps. Across both viewpoints, the subject remains a focal point for understanding how sensitive domino statistics are to boundary geometry, internal defects, and symmetry constraints (Ciucu, 8 Aug 2025, Duits et al., 2017).