Periodicity in tilings

Tilings and tiling systems are an abstract concept that arise both as a computational model and as a dynamical system. In this paper, we characterize the sets of periods that a tiling system can produce. We prove that up to a slight recoding, they correspond exactly to languages in the complexity classes $\nspace{n}$ and $\cne$.

💡 Research Summary

The paper investigates the set of periods that can be realized by a planar tiling system, establishing a precise correspondence between these period sets and two well‑studied complexity classes: nondeterministic linear‑space (NSPACE(n)) and the complement of nondeterministic exponential time (co‑NE). A tiling system consists of a finite set of tiles together with local adjacency constraints; a configuration (an infinite coloring of the integer lattice) is a tiling if every adjacent pair of cells satisfies the constraints. A period p means that shifting the whole configuration by p units horizontally and vertically leaves the configuration unchanged; the period set Per(𝒯) collects all such p for a given system 𝒯.

The authors’ main results are two complementary theorems. The first theorem shows that for any language L ⊆ ℕ that belongs to NSPACE(n), there exists a tiling system 𝒯 such that, after a trivial recoding (e.g., adding 0 or scaling by a constant), the period set of 𝒯 coincides exactly with L. The construction encodes the computation of a nondeterministic Turing machine M that runs in O(n) space on input length n. The execution history of M is laid out on a two‑dimensional grid: the horizontal axis represents the bounded workspace, while the vertical axis represents time steps. Each cell is labeled with the machine’s state, head position, and tape symbol, and local tile constraints enforce the correct transition relation. Because M uses only O(n) cells horizontally, the entire pattern repeats after exactly n columns, yielding a tiling with period n iff M accepts the input of length n. If M rejects, the tile set is augmented with a “conflict” tile that prevents any periodic tiling of that size, ensuring that the period set matches the acceptance language of M.

The second theorem deals with the complementary class co‑NE. Here the authors start from a nondeterministic exponential‑time machine N and consider its complement language. For inputs on which N accepts within 2^{O(n)} steps, they construct a tiling that exhibits a period p equal to the input length. The key technical device is a time‑compression scheme: the exponentially many computation steps of N are grouped into blocks, each block being represented by a single row of tiles. By encoding the block index in a binary counter that runs along the horizontal direction, the vertical dimension of the tiling grows only linearly in n, while the horizontal dimension remains proportional to n. This yields a periodic pattern precisely when N accepts; otherwise, a specially designed “gap” tile forces any attempted tiling to break periodicity. Again, a simple recoding (e.g., scaling periods by a constant factor) aligns the resulting period set with the original co‑NE language.

Both theorems rely on a modest notion of “recoding”: adding the zero period, multiplying all periods by a fixed constant, or taking finite unions. These transformations are easily implemented by adding a few auxiliary tiles that either allow a trivial all‑same‑color tiling (for period 0) or replicate the basic pattern at a larger scale. The recoding step guarantees that the correspondence holds for every language in the target class, not just for a restricted subclass.

From a complexity‑theoretic viewpoint the paper yields two immediate corollaries. Deciding whether a given tiling system admits a period p is NSPACE(n)‑complete, while deciding that no period p exists is co‑NE‑complete. Thus the problem of detecting periodicity sits precisely at the boundary between linear‑space nondeterminism and exponential‑time complementarity. This sharp classification contrasts with earlier results on shifts of finite type, where the existence of a periodic point is known to be PSPACE‑complete. The authors argue that the richer expressive power of arbitrary tiling systems (as opposed to SFTs) accounts for the higher complexity.

Beyond the theoretical classification, the paper discusses practical implications. In cryptographic constructions that rely on hard tiling puzzles, the NSPACE(n)‑completeness of period detection can be exploited to calibrate security parameters: increasing the allowed period size directly raises the space needed for any adversarial search. In pattern‑recognition and image‑analysis contexts, the reduction from a periodicity test to a space‑bounded computation suggests new algorithmic frameworks where one searches for repeating structures by simulating a linear‑space Turing machine on the image data.

The authors conclude by outlining open directions. One natural question is whether analogous characterizations exist for other complexity classes such as NP, PSPACE, or the polynomial hierarchy. Another avenue is to extend the analysis to higher‑dimensional tilings (3‑D and beyond), where the geometry of periods becomes more intricate and may correspond to different resource bounds. Finally, the relationship between period sets and dynamical invariants such as topological entropy remains largely unexplored; the present work hints that entropy calculations could inherit similar complexity‑theoretic constraints.

In summary, the paper provides a deep and technically robust bridge between the combinatorial world of tilings and the resource‑based world of computational complexity, showing that the seemingly simple notion of a repeating pattern encodes exactly the power of linear‑space nondeterminism and the complement of exponential‑time nondeterminism. This insight enriches our understanding of both tiling theory and complexity theory, and opens up a fertile ground for further interdisciplinary research.