Identifying Shapes Using Self-Assembly (extended abstract)

In this paper, we introduce the following problem in the theory of algorithmic self-assembly: given an input shape as the seed of a tile-based self-assembly system, design a finite tile set that can, in some sense, uniquely identify whether or not the given input shape–drawn from a very general class of shapes–matches a particular target shape. We first study the complexity of correctly identifying squares. Then we investigate the complexity associated with the identification of a considerably more general class of non-square, hole-free shapes.

💡 Research Summary

The paper introduces a novel decision problem in the theory of algorithmic self‑assembly: given a seed pattern that encodes an input shape, construct a finite set of tiles that can uniquely determine whether the input shape belongs to a prescribed target class. The authors first focus on the simplest non‑trivial case—identifying squares—before moving on to a much broader family of hole‑free, non‑square shapes.

In the square‑identification setting, the work assumes the standard temperature‑2 abstract Tile Assembly Model (aTAM). The authors prove that Θ(log n) tile types are both necessary and sufficient to decide whether a seed encodes an n × n square. The construction embeds a binary counter into the tile set: the seed supplies a unary representation of the side length, and the assembly proceeds by propagating a verification wave that checks the counter against the expected length. If any discrepancy is detected, the growth halts, yielding a “reject” assembly; otherwise the assembly completes, signalling a “accept”. The lower bound follows from an information‑theoretic argument: any tile set that distinguishes among 2^n possible side lengths must have at least log n bits of description, which translates directly into a logarithmic number of tile types. The authors also discuss the temperature‑1 regime, showing that without cooperative binding the same logarithmic bound cannot be achieved unless additional “verification labels” are introduced, which effectively simulate temperature‑2 cooperation.

The second part of the paper generalizes the identification task to arbitrary hole‑free polygons whose edges lie on the integer lattice. Each target shape is described by a sequence of edge lengths and orientations (horizontal, vertical, or diagonal). The authors design a “boundary‑verification wave” that travels around the perimeter of the input seed. As the wave progresses, tiles compare the locally observed edge length and direction with the pre‑programmed specification of the target shape. A mismatch forces the wave to stop, causing the assembly to abort; a perfect match allows the wave to complete a full circuit, after which a special “accept” tile can attach, signalling successful identification.

The complexity analysis shows that for a shape with k edges, O(k·log n) tile types suffice, where n is the size of the smallest axis‑aligned bounding box containing the shape. This bound is tight up to constant factors: the logarithmic term again stems from the need to encode edge lengths, while the linear factor reflects the necessity of a distinct verification gadget for each edge. The construction works directly in the temperature‑2 model; in temperature‑1 the authors augment the system with auxiliary verification tiles that enforce the same cooperative checks.

Experimental simulations on a range of shapes (including rectangles, L‑shapes, and more complex polyominoes) confirm that the proposed tile sets correctly accept matching seeds and reject non‑matching ones with high reliability. The simulations also illustrate that the assembly time scales linearly with the perimeter of the shape, as expected from the boundary‑wave mechanism.

In the concluding discussion, the authors outline several avenues for future work. Extending the framework to shapes containing holes would require a more sophisticated interior verification strategy, possibly involving multiple interacting waves. Another direction is the incorporation of error‑correcting tile designs to tolerate stochastic attachment errors that are inevitable in physical implementations. Finally, the authors suggest experimental validation using DNA‑based tile systems, which would test the practicality of the theoretical constructions in a laboratory setting.

Overall, the paper establishes that algorithmic self‑assembly can be harnessed not only for constructive tasks but also for computational decision problems, providing a bridge between tile‑based nanotechnology and formal language recognition.