Isohedral Number Problem: Tiling Complexity

• The isohedral number problem is defined as the minimum number of transitivity classes required in a periodic tiling by congruent copies of a single shape.
• It plays a crucial role in understanding tiling symmetry by linking local geometric matching constraints with global periodic order, as seen in classical tilings and complex polyforms.
• Algorithmic methods, including SAT formulations and boundary-word algorithms, offer efficient tests for isohedral tilability while highlighting challenges in determining higher isohedral numbers.

Searching arXiv for recent and relevant papers on the isohedral number problem, isohedral tilings, and algorithmic detection of isohedrality. The isohedral number problem asks how complicated periodic monohedral tilings of the Euclidean plane can be when complexity is measured by tile-transitivity under the tiling’s symmetry group. In the standard formulation, one starts with a single shape $S$ that tiles the plane and asks for the minimum possible number of transitivity classes of tiles among all tilings by congruent copies of $S$; that minimum is the isohedral number of $S$. The problem then becomes twofold: which positive integers occur as isohedral numbers, and whether these numbers are bounded over all plane shapes that admit periodic tilings. In Craig Kaplan’s exposition, isohedral numbers, Heesch numbers, and aperiodic monotiles are treated as three views of the same underlying question: how much global order can be forced or obstructed by purely local geometric constraints (Kaplan, 2 Sep 2025).

1. Definitions and formal framework

A shape is a topological disk in the plane. Given a finite set of shapes $\mathcal{S}=\{S_1,\dots,S_k\}$, a tiling from $\mathcal{S}$ is a countably infinite collection of tiles $\mathcal{T}=\{T_1,T_2,\dots\}$ such that each tile is congruent to one of the prototiles in $\mathcal{S}$, and the tiles cover the entire plane with no gaps and no overlaps except possibly along boundaries. A monotile is a single shape $S$ that admits such a tiling as the singleton set $\{S\}$ (Kaplan, 2 Sep 2025).

For a tiling $\mathcal{T}$, its symmetry group is the set of rigid motions of the plane that map the tiling to itself. These isometries may include translations, rotations, reflections, and glide reflections. Two tiles $S$0 are transitively equivalent if some symmetry of the tiling maps $S$1 to $S$2. This partitions the tiles into transitivity classes, i.e. orbits of the symmetry group acting on tiles. A tiling is isohedral precisely when there is a single such orbit: every tile is symmetry-equivalent to every other tile.

The isohedral number of a tiling is the number of transitivity classes in that tiling. The isohedral number of a shape $S$3 is the minimum of that quantity over all tilings by congruent copies of $S$4. Thus isohedral number $S$5 means that $S$6 admits an isohedral tiling; isohedral number $S$7 means that every tiling by $S$8 has at least $S$9 tile orbits, while some tiling has exactly $S$0.

This notion is meaningful particularly for periodic tilings. A tiling is periodic if its symmetry group contains translations in two linearly independent directions; equivalently, its symmetry group is one of the $S$1 wallpaper groups. In a periodic tiling there are only finitely many tile orbits, so every shape that admits a periodic tiling has finite isohedral number. Kaplan also gives a heuristic interpretation: the isohedral number often reflects how many copies of the prototile must first be assembled into a larger patch before the pattern “settles” into a simple isohedral repetition (Kaplan, 2 Sep 2025).

2. Statement of the problem and its scope

The isohedral number problem is usually posed for a single shape in the Euclidean plane. Its basic questions are:

1. Occurrence: which positive integers occur as isohedral numbers of plane shapes?
2. Boundedness: is there a universal upper bound on isohedral numbers among shapes that admit periodic tilings?

Kaplan explicitly frames this as the periodic-tiling analogue of Heesch’s problem. Heesch numbers ask how many concentric layers a non-tiler can support before failure; isohedral numbers ask how far a periodic tiler can be forced away from the maximally symmetric case of a one-orbit tiling (Kaplan, 2 Sep 2025).

The problem has an immediate algorithmic interpretation. If there were a universal bound $S$2 on isohedral numbers, then one could determine whether a shape admits a periodic tiling by constructing all patches up to size $S$3 and checking whether any of them tiles isohedrally. Kaplan emphasizes this consequence directly: a bound on isohedral numbers would yield an algorithm for detecting periodic tilings. This places the problem near broader decidability questions in tiling theory. In the same discussion, the tiling problem for sets of as few as three polygons is noted as undecidable, whereas the corresponding question for a single prototile remains open (Kaplan, 2 Sep 2025).

A common misunderstanding is to identify isohedral number with the number of distinct tilings of a shape. The notion is different: it counts the minimum number of tile orbits under the symmetry group within a tiling, not the number of noncongruent tilings admitted by the shape. A second misconception is that a high isohedral number signals nonperiodicity. The opposite is true: isohedral numbers are defined through periodic structure, and a high value indicates that periodicity, when it exists, may require a comparatively intricate orbit structure before symmetry repeats.

3. Examples and empirical landscape

The simplest case is isohedral number $S$4. Any isohedral tiling realizes this value; familiar examples include square-grid tilings by unit squares and honeycomb tilings by regular hexagons. In such tilings, translations alone already act transitively on tiles.

More informative are shapes whose periodic tilings are necessarily non-isohedral. Kaplan highlights Heesch’s 1935 tile, a polygon of isohedral number $S$5: it tiles periodically, but no periodic tiling by that polygon is tile-transitive. The simplest periodic realizations therefore require at least two tile orbits. At the higher end of current examples, Kaplan describes Joseph Myers’s 16-hex, a polyhex of $S$6 hexagonal cells whose isohedral number is $S$7; Myers found a periodic tiling with $S$8 transitivity classes and showed that no periodic tiling by that shape can have fewer (Kaplan, 2 Sep 2025).

Myers’s computational work is the principal empirical source reported in Kaplan’s survey. It produced polyominoes, polyhexes, polyiamonds, and polykites realizing every isohedral number from $S$9 through $\mathcal{S}=\{S_1,\dots,S_k\}$0 except $\mathcal{S}=\{S_1,\dots,S_k\}$1. The reported realized values are therefore

$\mathcal{S}=\{S_1,\dots,S_k\}$2

No example with isohedral number $\mathcal{S}=\{S_1,\dots,S_k\}$3 is known in that dataset.

Isohedral number	Representative evidence	Reported status
$\mathcal{S}=\{S_1,\dots,S_k\}$4	Classical isohedral tilings	Realized
$\mathcal{S}=\{S_1,\dots,S_k\}$5	Heesch’s 1935 polygon	Realized
$\mathcal{S}=\{S_1,\dots,S_k\}$6	Myers polyforms	Realized
$\mathcal{S}=\{S_1,\dots,S_k\}$7	No example in Myers’s data	Open gap
$\mathcal{S}=\{S_1,\dots,S_k\}$8	Myers’s 16-hex	Realized

This evidence does not settle boundedness, but it pushes the known range well beyond the trivial cases. Kaplan explicitly notes that there is no known reason for a global bound to exist, and that if one did exist, he would be “extremely surprised” if it were $\mathcal{S}=\{S_1,\dots,S_k\}$9 (Kaplan, 2 Sep 2025). That statement is an informed research judgment rather than a theorem, but it captures the present asymmetry between substantial computational evidence and the absence of general structure theory.

4. Algorithms and computational methodologies

Algorithmic work on the isohedral number problem is presently strongest at the threshold case of isohedral number $\mathcal{S}$0, i.e. isohedral tilability. Kaplan’s survey cites Langerman and Winslow’s result that a polyomino with perimeter $\mathcal{S}$1 can be tested for isohedral tilability in $\mathcal{S}$2 time (Kaplan, 2 Sep 2025). The underlying paper gives a boundary-word algorithm for deciding whether a polyomino admits any isohedral tiling, improving an earlier $\mathcal{S}$3-time algorithm of Keating and Vince (Langerman et al., 2015).

The quasilinear algorithm uses Heesch–Kienzle boundary criteria. The boundary word of a polyomino is treated as a circular word over $\mathcal{S}$4, and isohedral tilability is reduced to the existence of one of seven specific factorizations corresponding to the square-lattice isohedral types. The method relies on admissible palindromes, reflect squares, $\mathcal{S}$5-dromes, and gapped mirrors; the slowest subroutine is the half-turn case, which determines the global $\mathcal{S}$6 complexity. The same paper defines a tiling to be $\mathcal{S}$7-isohedral when its tiles split into $\mathcal{S}$8 orbits under the symmetry group, and explicitly leaves efficient detection of $\mathcal{S}$9-isohedral polyomino tilings as an open problem (Langerman et al., 2015).

A different computational line uses SAT. Kaplan shows how to encode the question “does this polyform tile the plane isohedrally?” as a Boolean satisfiability problem (Kaplan, 2024). The formulation is based on surrounds and the Local Theorem: a shape admits an isohedral tiling iff it has a surround in which every surrounding tile is extendable. The SAT instance combines halo-coverage constraints, non-overlap constraints, hole elimination, and extendability clauses induced by compositions of rigid motions. A central feature is that the method does not explicitly invoke the classical classification into $\mathcal{T}=\{T_1,T_2,\dots\}$0 isohedral tiling types; instead it works directly with congruent surrounds and local propagation. Kaplan reports agreement with Myers’s tables up to 12-ominoes, 12-hexes, 13-iamonds, and 12-kites, but also states that the implementation does not compute exact isohedral numbers for $\mathcal{T}=\{T_1,T_2,\dots\}$1-anisohedral polyforms with $\mathcal{T}=\{T_1,T_2,\dots\}$2 (Kaplan, 2024).

A more restrictive enumerative tradition studies isohedral tilings in which the tile is a fundamental domain of the symmetry group. For polyominoes and polyiamonds with $\mathcal{T}=\{T_1,T_2,\dots\}$3-, $\mathcal{T}=\{T_1,T_2,\dots\}$4-, or $\mathcal{T}=\{T_1,T_2,\dots\}$5-fold rotational symmetry, computer algorithms were used to enumerate all such tilings for small $\mathcal{T}=\{T_1,T_2,\dots\}$6, with symmetry groups of types $\mathcal{T}=\{T_1,T_2,\dots\}$7, $\mathcal{T}=\{T_1,T_2,\dots\}$8, $\mathcal{T}=\{T_1,T_2,\dots\}$9, $\mathcal{S}$0, and $\mathcal{S}$1. In the same setting it was proved that there are no isohedral tilings with groups $\mathcal{S}$2, $\mathcal{S}$3, or $\mathcal{S}$4 having polyominoes or polyiamonds as fundamental domains (Fukuda et al., 2010). These results do not address full isohedral numbers, but they provide explicit finite classifications of highly symmetric subfamilies.

5. Open problems and conceptual significance

Three open questions dominate the area. The first is boundedness: whether $\mathcal{S}$5 is finite over all plane shapes $\mathcal{S}$6 that admit periodic tilings. The second is realizability: which integers occur at all. The missing value $\mathcal{S}$7 in Myers’s data makes the latter question concrete, but the real difficulty lies in distinguishing a local computational gap from a genuine structural obstruction (Kaplan, 2 Sep 2025).

The third concerns decidability of periodic tilings. If there were a universal bound on isohedral numbers, periodic tilability could in principle be decided by a bounded search over patches. No such bound is known. This connects the isohedral number problem to the broader boundary between local constraints and global order: a shape might support arbitrarily large finite patches, yet periodicity might remain detectable only at arbitrarily large orbit complexity.

Kaplan makes this relation explicit through a joint Heesch–isohedral thought experiment. One can imagine simultaneously increasing the search depth for finite surrounds and for isohedral patches. If the surround search terminates first, the shape does not tile. If the isohedral search terminates first, the shape tiles periodically with bounded isohedral number. If neither search terminates, one is led toward the possibility of an aperiodic monotile: a shape admitting tilings but no periodic tiling at all (Kaplan, 2 Sep 2025). This is not an algorithm, because no a priori bounds are available, but it clarifies why Heesch numbers, isohedral numbers, and aperiodicity form a coherent research cluster.

A further point of clarification is that current algorithmic success for isohedral number $\mathcal{S}$8 should not be mistaken for progress on exact higher isohedral numbers. The SAT method decides isohedral tilability; the quasilinear boundary-word method decides isohedral tilability for polyominoes; neither presently computes the minimum orbit count once the answer is at least $\mathcal{S}$9 (Kaplan, 2024).

6. Related settings and contrastive results

Although the isohedral number problem is formulated in the Euclidean plane, related work in spherical tiling theory illuminates how tile-transitivity can fail, bifurcate, or become rigid under additional geometric constraints. In spherical tilings by congruent quadrangles over pseudo-double wheels, Akama proves that isohedrality is equivalent to a specific metric-combinatorial condition $S$0 and to the presence of a $S$1 symmetry group consisting of one $S$2-fold axis and $S$3 perpendicular $S$4-fold axes (Akama, 2018). In that setting, a quadratic equation in $S$5,

$S$6

classifies the admissible edge lengths of the tile.

This spherical classification exhibits a phenomenon absent from the simplest Euclidean examples: for certain angle pairs $S$7, the same pseudo-double-wheel skeleton and the same cyclic list of tile angles support two non-congruent isohedral tilings. Akama also gives a special $S$8-face case in which the quadratic degenerates to $S$9, and the same spherical quadrangle organizes both an isohedral tiling and a non-isohedral tiling over the same skeleton (Akama, 2018). These are not statements about Euclidean isohedral numbers, but they show that orbit structure can depend subtly on the metric realization even when the combinatorial scaffold is fixed.

A complementary spherical rigidity result appears in work on monohedral quadrangular tilings topologically equivalent to trapezohedra. There it is proved that if the number of faces is $\{S\}$0 or $\{S\}$1, or if the tile is a kite, dart, or rhombus, or if the tile is convex, then the tiling is isohedral (Akama et al., 2013). In the language of isohedral numbers, these are classes in which the value is forced to be $\{S\}$2. The contrast with the Euclidean-plane problem is instructive: strong convexity and topology hypotheses can collapse orbit complexity, whereas the unrestricted Euclidean problem remains open even at the level of boundedness.

Taken together, these related settings suggest that the isohedral number problem is not only about enumerating orbit counts. It is a problem at the intersection of Euclidean symmetry, local matching structure, combinatorial patch growth, and algorithmic decidability. Current knowledge establishes a nontrivial range of examples, efficient tests for isohedrality in several important classes, and multiple analogues showing both rigidity and multiplicity. What remains unknown is the central global picture: whether periodic monohedral tilings can force arbitrarily many tile orbits, and whether the integers that already appear computationally are the beginning of an unbounded sequence or the visible edge of a much more constrained theory.