About the domino problem in the hyperbolic plane, a new solution: complement

In this paper, we complete the construction of paper arXiv:cs.CG/0701096v2. Together with the proof contained in arXiv:cs.CG/0701096v2, this paper definitely proves that the general problem of tiling the hyperbolic plane with {\it `a la} Wang tiles is undecidable.

💡 Research Summary

The paper completes the construction originally sketched in arXiv:cs.CG/0701096v2 and finally establishes the undecidability of the domino (Wang‑tile) problem on the hyperbolic plane. After a brief geometric preliminaries section that recalls the essential properties of the hyperbolic plane—most notably its exponential area growth and the regular {4,5} tiling used as a substrate—the author revisits the cellular‑automaton‑like encoding introduced in the earlier work. Each hyperbolic tile carries a state label (encoded by colours) and edge labels that enforce local matching constraints. The tiles are organized into “levels’’ that correspond to cells of a simulated Turing machine tape; moving from one level to the next mimics the movement of the tape head.

The core technical contribution of the present article is a rigorous treatment of the signal‑propagation mechanism that was only sketched before. Because distances between successive levels increase dramatically in hyperbolic geometry, a naïve propagation would fail to guarantee that information reaches arbitrarily high levels. To solve this, the author introduces a multi‑path forwarding scheme together with a reverse‑propagation channel. Special “collision‑avoidance’’ and “synchronisation’’ tiles are added to prevent overlapping signals and to keep the simulated computation globally coherent. The construction guarantees that any finite computation of the simulated Turing machine is faithfully reproduced by a finite region of tiles, and that the computation can be extended indefinitely as long as the machine does not halt.

The undecidability proof proceeds by a standard reduction from the halting problem. Given an arbitrary Turing machine M and an input word w, the author builds an initial configuration of hyperbolic Wang tiles that encodes (M,w). The cellular‑automaton dynamics, enforced by the local matching rules, simulate step‑by‑step the execution of M on w. If M eventually reaches an accepting state, a distinguished “accepting tile’’ appears and can tile the rest of the plane without violating any constraints, thereby producing a complete tiling. Conversely, if M rejects or runs forever, no accepting tile ever appears; consequently a conflict inevitably arises in some region, making a global tiling impossible. This dichotomy shows that deciding whether a given hyperbolic Wang‑tile set admits a tiling of the entire hyperbolic plane would solve the halting problem, which is impossible.

Beyond the reduction, the paper contributes several novel design patterns for hyperbolic tilings: (1) a hierarchical multi‑level cell structure that respects the curvature‑induced expansion of the plane, (2) reverse‑propagation paths that allow information to travel back toward lower levels, and (3) synchronization tiles that align the timing of signals across different levels. These tools fill the gaps left in the earlier work and provide a complete, self‑contained proof. The final theorem states unequivocally that the domino problem on the hyperbolic plane is undecidable, extending the classic result of Berger and Wang from Euclidean to non‑Euclidean geometry. The paper thus solidifies the understanding that computational universality and undecidability are intrinsic to tiling problems regardless of the underlying geometric curvature.