Combinatorial substitutions and sofic tilings

A combinatorial substitution is a map over tilings which allows to define sets of tilings with a strong hierarchical structure. In this paper, we show that such sets of tilings are sofic, that is, can be enforced by finitely many local constraints. This extends some similar previous results (Mozes'90, Goodman-Strauss'98) in a much shorter presentation.

💡 Research Summary

The paper introduces the notion of a combinatorial substitution, a map that operates on tilings by specifying how each tile type is replaced by a finite configuration of smaller tiles together with adjacency labels. Unlike classical geometric substitutions that rely on scaling and rotation, a combinatorial substitution is purely symbolic: it works on a finite alphabet of tile labels and a finite set of edge colors (or matching constraints). By iterating the substitution, one obtains an infinite hierarchical tiling in which every level refines the previous one according to the same local rule.

The central claim is that the family of tilings generated by any such substitution is a sofic subshift, i.e., it can be enforced by a finite set of local constraints. To prove this, the authors first encode the substitution rules into a finite directed graph. Vertices correspond to tile labels, and directed edges encode the allowed adjacency relations after one substitution step. From this graph they extract a finite “boundary language”: the set of all possible patterns that can appear on the interface between a substituted macro‑tile and its neighbours at any level. Because the substitution alphabet and the graph are finite, this boundary language is also finite, regardless of how many substitution steps are performed.

Next, the authors construct a finite set of Wang tiles (or, equivalently, a finite set of colored squares) that carries three pieces of information: (1) the intrinsic label of the tile, (2) a “state” that records which edge of the substitution hierarchy the tile belongs to, and (3) a boundary label drawn from the previously computed boundary language. The matching rules for these Wang tiles require that adjacent tiles have identical boundary labels and that the state transitions follow the edges of the substitution graph. Consequently, any global tiling that satisfies these local matching rules automatically respects the hierarchical structure imposed by the original substitution. Conversely, any tiling produced by iterating the combinatorial substitution can be projected onto a tiling of the constructed Wang tiles by assigning to each tile its corresponding state and boundary label. This establishes a bijective factor map between the substitution tilings and the tiling space defined by the finite Wang set, proving that the former is a factor of a shift of finite type and therefore sofic.

The result generalizes earlier work. Mozes (1990) showed that certain self‑similar tilings are sofic, while Goodman‑Strauss (1998) proved that a class of “enforced” substitution tilings is sofic, but both required strong geometric assumptions. The present paper abstracts away geometry entirely; the only requirements are finiteness of the substitution alphabet and the combinatorial nature of the replacement rules. This yields a much broader theorem: every combinatorial substitution, no matter how intricate its hierarchical pattern, gives rise to a sofic tiling space.

From a theoretical standpoint, the theorem bridges symbolic dynamics, formal language theory, and tiling theory. Sofic subshifts are precisely the class of languages recognized by finite‑state automata; thus the paper shows that hierarchical tilings generated by combinatorial substitutions are recognisable by a finite automaton operating on two‑dimensional configurations. This has several implications: (i) decidability results that hold for shifts of finite type extend to these hierarchical tilings; (ii) the existence of a finite set of local constraints implies that algorithmic generation, verification, and even classification of such tilings become feasible; (iii) the construction provides a systematic method to translate high‑level substitution specifications into low‑level tile sets, which can be useful for designing aperiodic tile sets or for engineering programmable materials.

Finally, the authors discuss potential applications. In computer graphics and procedural generation, combinatorial substitutions offer a compact description of complex patterns; the sofic encoding guarantees that these patterns can be rendered using only local rules, which is advantageous for parallel processing. In mathematical physics, hierarchical structures such as quasicrystals can be modeled by combinatorial substitutions, and the sofic representation may aid in studying their spectral properties via symbolic dynamics. In theoretical computer science, the result contributes to the understanding of two‑dimensional language hierarchies, showing that the class of languages generated by hierarchical substitution is contained within the sofic class.

Overall, the paper provides a concise yet powerful proof that any tiling space defined by a finite combinatorial substitution is sofic, thereby unifying and extending previous isolated results and opening new avenues for both theoretical investigation and practical implementation.