Symmetric Cruciform Regions in Domino Tiling Theory

• Symmetric cruciform regions are cross-shaped areas on square lattices, defined by overlapping Aztec rectangles with matched arm parameters.
• Their domino tilings are enumerated using explicit product formulas involving hyperfactorials and power-of-two factors, even in cases with internal holes.
• Geometric symmetry through reflection and half-turn rotations is key to balancing conditions and simplifying both unweighted and weighted tiling formulations.

Symmetric cruciform regions are cross-shaped regions on the square lattice that arise in domino-tiling theory in two closely related ways: as superpositions of Aztec rectangles with matched arm parameters, and as the outer boundaries obtained after repeated complementation of Aztec rectangles with symmetrically placed odd Aztec windows. In the literature they are encoded either by the six-parameter family $C_{m,n}^{a,b,c,d}$ and its symmetric specializations, or by evolved outer boundaries such as $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$. Their central feature is enumerative: despite nonconvexity and, in some cases, internal holes, their domino tilings admit explicit product formulas, weighted generating functions, and divisibility relations linking them to Aztec diamonds, elbow regions, and Di Francesco’s Aztec triangle (Ciucu, 2021, Ciucu, 8 Aug 2025, Lai et al., 5 Apr 2026).

1. Geometric definition and parameterization

In Ciucu’s framework, the cruciform regions $C_{m,n}^{a,b,c,d}$ are obtained by superimposing two Aztec rectangles so that their overlap forms a central cross and the four outer protrusions become “piers” or arms. The parameters $a,b,c,d$ record the lengths of the northwest, northeast, southeast, and southwest piers, respectively, while $m$ and $n$ encode the two underlying Aztec-rectangle scales. The boundary is a zigzag boundary on the square lattice, and the resulting region has the visual form of a plus-shaped cross with a thicker central body and four protruding arms (Ciucu, 2021).

The same family is described in a later weighted treatment by a rotated-picture notation $\mathcal{C}^{a,b,c,d}_{m,n}$. There one starts with $\mathcal{AR}_{m+b+d+1,n}$ and $\mathcal{AR}_{m,n+a+c+1}$, aligns their central $m\times n$ core, and lets $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$0 control the vertical extensions and $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$1 control the horizontal extensions. Domino tilings of $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$2 are identified with perfect matchings of the dual cruciform graph $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$3 (Lai et al., 5 Apr 2026).

Ciucu’s definition permits negative arm parameters by interpreting them as retractions of zigzag strips. If a pier has length $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$4, it is a chain of diagonally adjacent unit squares; if its length is $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$5, one attaches $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$6 zigzag strips of width $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$7; if its length is $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$8, one removes $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$9 such strips. This extension is important because several symmetric families encountered after complementation naturally cross from a rectangle-with-windows regime into a genuine cruciform regime with formally negative arm data (Ciucu, 2021).

A distinct but related definition appears in the study of odd Aztec windows. There the cruciform is not introduced from scratch; instead it appears as the outer boundary obtained from repeated complementation of an $C_{m,n}^{a,b,c,d}$0-type Aztec rectangle with symmetric odd windows. The resulting cruciform has the form

$C_{m,n}^{a,b,c,d}$1

with a horizontal bar, a vertical bar intersecting at the center, and arm parameters inherited from the extremal windows of the original configuration (Ciucu, 8 Aug 2025).

2. Symmetry classes and balancedness

Tileability is governed first by a balancing condition. For the general cruciform region $C_{m,n}^{a,b,c,d}$2, domino tilings can occur only when

$C_{m,n}^{a,b,c,d}$3

Ciucu proves that this equality is equivalent to balancedness, i.e. equality of black and white square counts. The same paper also records necessary arm bounds: if $C_{m,n}^{a,b,c,d}$4 or $C_{m,n}^{a,b,c,d}$5 exceeds $C_{m,n}^{a,b,c,d}$6, or $C_{m,n}^{a,b,c,d}$7 or $C_{m,n}^{a,b,c,d}$8 exceeds $C_{m,n}^{a,b,c,d}$9, then the region has no domino tilings (Ciucu, 2021).

Geometric symmetry is reflected algebraically by relations among the arm parameters. Horizontal and vertical reflection symmetries correspond to $a,b,c,d$0 and $a,b,c,d$1, respectively, while half-turn symmetry corresponds to imposing both simultaneously. A particularly important symmetric specialization is

$a,b,c,d$2

which is symmetric across both coordinate axes in the planar picture. However, the balancing condition then becomes $a,b,c,d$3, which is impossible for integers. Thus the region $a,b,c,d$4 is geometrically symmetric but not balanced as a standalone domino region (Ciucu, 2021).

This obstruction clarifies an important point: exact four-arm symmetry is natural at the graph level, but not every graph-level symmetry class corresponds to a tileable region. In fact, Ciucu observes that true $a,b,c,d$5 rotational symmetry would require

$a,b,c,d$6

and then balancing would force $a,b,c,d$7, again impossible over the integers. The natural exact lattice symmetries for tileable cruciforms are therefore reflection symmetries and $a,b,c,d$8 rotational symmetry rather than full fourfold rotational symmetry (Ciucu, 2021).

Balanced symmetric halves do exist. The elbow region $a,b,c,d$9 is defined as the portion of $m$0 above its central horizontal row, and it is tileable exactly when

$m$1

This family may be viewed as a symmetric half-cruciform. At the opposite extreme, the region

$m$2

is a highly symmetric balanced cruciform obtained by repeated symmetrization of Di Francesco’s region $m$3; it decomposes into eight congruent fundamental regions, one of which is $m$4 (Ciucu, 2021).

In the Aztec-window setting, symmetry is inherited dynamically rather than imposed statically. Starting from an $m$5-type Aztec rectangle with windows centered on the horizontal axis, repeated complementation preserves symmetry with respect to the horizontal axis, the vertical axis, and the central point. Once the extremal windows reach and pass the outer boundary, the evolving outer shape becomes a symmetric cruciform whose inner holes remain placed symmetrically along the horizontal symmetry axis (Ciucu, 8 Aug 2025).

3. Exact enumeration and product formulas

The main enumeration theorem for general cruciforms is expressed in terms of the hyperfactorial

$m$6

For every tileable cruciform region $m$7,

$m$8

This is Ciucu’s closed product formula for domino tilings of cruciform regions, and it places the family alongside the Aztec diamond theorem as an exactly solvable class with explicit factorization (Ciucu, 2021).

The symmetric specialization $m$9, $n$0, $n$1 yields the algebraic expression

$n$2

Because the balancing condition fails for integer $n$3, this formula is naturally interpreted at the graph level rather than as the tiling count of a balanced region. For actual domino tilings, the balanced symmetric half-cruciforms $n$4 provide the relevant specialization (Ciucu, 2021).

For elbow regions, the product formula becomes especially compact. If $n$5, then

$n$6

When $n$7 and $n$8, the region reduces, after removing forced dominoes, to the Aztec diamond $n$9, and the formula recovers $\mathcal{C}^{a,b,c,d}_{m,n}$0 (Ciucu, 2021).

The Di Francesco–related symmetric cruciform

$\mathcal{C}^{a,b,c,d}_{m,n}$1

also has an explicit specialization: $\mathcal{C}^{a,b,c,d}_{m,n}$2 where $\mathcal{C}^{a,b,c,d}_{m,n}$3 is the explicit quadratic exponent obtained from the general theorem. This region is crucial because $\mathcal{C}^{a,b,c,d}_{m,n}$4 divides its tiling number (Ciucu, 2021).

A weighted generalization replaces the unweighted count by the generating function $\mathcal{C}^{a,b,c,d}_{m,n}$5, where odd vertical dominoes have weight $\mathcal{C}^{a,b,c,d}_{m,n}$6, even horizontal dominoes have weight $\mathcal{C}^{a,b,c,d}_{m,n}$7, even vertical dominoes containing a white square at $\mathcal{C}^{a,b,c,d}_{m,n}$8 have weight $\mathcal{C}^{a,b,c,d}_{m,n}$9, and odd horizontal dominoes containing a white square at $\mathcal{AR}_{m+b+d+1,n}$0 have weight $\mathcal{AR}_{m+b+d+1,n}$1. The main theorem of the weighted paper requires

$\mathcal{AR}_{m+b+d+1,n}$2

and produces a closed product formula depending on the parity of $\mathcal{AR}_{m+b+d+1,n}$3. In the top-bottom symmetric case $\mathcal{AR}_{m+b+d+1,n}$4, one has $\mathcal{AR}_{m+b+d+1,n}$5, so all factors involving $\mathcal{AR}_{m+b+d+1,n}$6, the product over $\mathcal{AR}_{m+b+d+1,n}$7, and the extra $\mathcal{AR}_{m+b+d+1,n}$8-factors disappear. Equivalently, the Schur correction term vanishes, and the generating function collapses to a substantially simpler product expression (Lai et al., 5 Apr 2026).

4. Symmetric cruciform regions from Aztec-window evolution

A second line of work produces symmetric cruciform regions as evolved outer boundaries of Aztec rectangles with odd windows. One begins with

$\mathcal{AR}_{m+b+d+1,n}$9

where the windows are odd Aztec rectangles placed symmetrically along the horizontal axis. Repeated applications of the complementation theorem transform the configuration by shrinking the outer rectangle and evolving each window according to

$\mathcal{AR}_{m,n+a+c+1}$0

After $\mathcal{AR}_{m,n+a+c+1}$1 steps one obtains

$\mathcal{AR}_{m,n+a+c+1}$2

When the extremal windows communicate with the exterior and $\mathcal{AR}_{m,n+a+c+1}$3, $\mathcal{AR}_{m,n+a+c+1}$4, the outer boundary becomes the cruciform

$\mathcal{AR}_{m,n+a+c+1}$5

and the remaining inner holes $\mathcal{AR}_{m,n+a+c+1}$6 lie symmetrically along the horizontal symmetry axis (Ciucu, 8 Aug 2025).

The enumeration in this cruciform phase is inherited directly from the windowed-rectangle phase. If $\mathcal{AR}_{m,n+a+c+1}$7 is odd and the hypotheses of Theorem 3.4 hold, then

$\mathcal{AR}_{m,n+a+c+1}$8

and Remark 4 states that the same right-hand side gives the number of domino tilings of the resulting symmetric cruciform region with holes. The graph quantity $\mathcal{AR}_{m,n+a+c+1}$9 is itself given by an explicit product formula due to Krattenthaler, so the cruciform with symmetric holes inherits a fully explicit enumeration (Ciucu, 8 Aug 2025).

This construction shows that symmetric cruciforms are not merely static families of regions; they are also terminal shapes in a controlled evolution. The symmetry is rigid throughout the process: the evolving regions remain symmetric with respect to the horizontal axis, the vertical axis, and the central point, and no new combinatorial factor appears when the geometry changes from “rectangle with deeper windows” to “genuine cruciform with centered holes” (Ciucu, 8 Aug 2025).

A concrete example uses $m\times n$0, $m\times n$1, $m\times n$2, $m\times n$3, $m\times n$4, $m\times n$5, giving $m\times n$6. The outer boundary becomes

$m\times n$7

with a single internal hole $m\times n$8 on the horizontal symmetry axis. The tiling count is then

$m\times n$9

followed by substitution of Krattenthaler’s product formula for $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$00 (Ciucu, 8 Aug 2025).

5. Proof architecture

The unweighted product formula for $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$01 is obtained by applying the complementation theorem to the dual graph of the cruciform. In this setting the complement of the dual graph of $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$02 is the dual graph of

$C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$03

and the path-of-cells count yields the recurrence

$C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$04

Iterating $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$05 times reduces enumeration to a terminal configuration $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$06, whose dual graph is a doubly intruded Aztec rectangle. Krattenthaler’s product formula for such graphs then gives the hyperfactorial expression above (Ciucu, 2021).

Symmetry enters again through Ciucu’s factorization theorem for graphs invariant under reflection. It is used to derive the elbow-region formula, to decompose symmetric graphs along an axis with $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$07-weighted cut edges, and to prove divisibility statements relating $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$08, auxiliary symmetrized regions, and the symmetric cruciform $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$09 (Ciucu, 2021).

The weighted generating-function proof follows a different but compatible route. It works on dual graphs and uses the vertex-splitting lemma, the star lemma, and the spider lemma, followed by Sandwich and Mega-sandwich lemmas that transform weighted Aztec rectangles into weighted semi-honeycomb graphs. The latter are then interpreted as lozenge-tiling problems on semi-hexagons with dents, and their generating functions are expressed through Schur functions. The final product formulas arise from Schur summations whose evaluation depends on whether $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$10 is even or odd. In the symmetric case $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$11, the relevant Schur sum becomes trivial, which is exactly why the additional correction factor disappears (Lai et al., 5 Apr 2026).

In the odd-window framework, the complementation theorem is iterated through the phase where windows touch the boundary and continue past it. The resulting cruciform phase carries the same power-of-two factor

$C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$12

as in the rectangular phase, so the passage from rectangle to cruciform changes geometry but not the underlying factorization mechanism. This persistence is the core reason that symmetric cruciforms with holes inherit explicit product formulas rather than requiring a new enumeration theory (Ciucu, 8 Aug 2025).

6. Connections, special roles, and significance

Symmetric cruciform regions occupy a junction between several classical themes in exact tiling enumeration. On one side stands the Aztec diamond theorem,

$C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$13

and on the other stands Di Francesco’s Aztec triangle $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$14, whose conjectural product formula motivated the original introduction of cruciform regions. Ciucu shows that

$C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$15

and similarly

$C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$16

so symmetric cruciforms and symmetric half-cruciforms function as explicit “containers” whose tiling counts are known in closed form (Ciucu, 2021).

The later odd-window paper places symmetric cruciforms in a different historical narrative. Ordinary Aztec diamonds with central Aztec windows exhibited intriguing but not “round” factorization patterns, whereas odd Aztec rectangles with odd windows do admit clean product formulas. Remark 4 shows that the same mechanism extends beyond rectangular outer shapes: once the evolving windows force the boundary to become cruciform, the symmetric cruciform region with symmetric holes is still enumerated by a simple power of $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$17 times a Krattenthaler-type product. This shows that the round-product phenomenon survives the transition from convex or nearly convex rectangles to genuinely cross-shaped regions (Ciucu, 8 Aug 2025).

The weighted theory broadens the same picture from counting to refinement. Its $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$18-weighted generating function separates arm parameters into bulk factors and a parity-sensitive interaction term controlled by $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$19. The top-bottom symmetric case $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$20 is therefore structurally special: the parity distinction disappears, the Schur-sum correction vanishes, and the formula becomes markedly simpler. This suggests that symmetry is not merely geometric decoration but an algebraic simplifier in the exact solution (Lai et al., 5 Apr 2026).

Several clarifications are essential. First, not every visually symmetric cruciform is tileable: $C_{m-a-k,k}^{(m-1)/2-k,a_s+1-k,(m-1)/2-k,a_1+1-k}$21 is symmetric but unbalanced. Second, exact fourfold rotational symmetry is obstructed by the integer balancing condition. Third, the cruciform regions arising in the odd-window setting may carry internal odd-Aztec-rectangle holes, so “symmetric cruciform region” can denote either a hole-free cross-shaped region or a symmetric cruciform outer boundary with symmetrically placed holes, depending on context (Ciucu, 2021, Ciucu, 8 Aug 2025).

The broader significance stated in the literature is that cruciform regions enlarge the class of planar regions with exact factorized tiling counts. They generalize Aztec diamonds, interact naturally with factorization theorems for symmetric graphs, and provide partial progress toward Di Francesco’s conjecture. The weighted extension further indicates a refined enumerative structure, while the odd-window construction suggests that the same complementation mechanism remains robust even after the outer boundary leaves the rectangular regime and becomes cruciform (Ciucu, 2021, Ciucu, 8 Aug 2025, Lai et al., 5 Apr 2026).