Self-Assembly with Geometric Tiles

In this work we propose a generalization of Winfree’s abstract Tile Assembly Model (aTAM) in which tile types are assigned rigid shapes, or geometries, along each tile face. We examine the number of distinct tile types needed to assemble shapes within this model, the temperature required for efficient assembly, and the problem of designing compact geometric faces to meet given compatibility specifications. Our results show a dramatic decrease in the number of tile types needed to assemble $n \times n$ squares to $\Theta(\sqrt{\log n})$ at temperature 1 for the most simple model which meets a lower bound from Kolmogorov complexity, and $O(\log\log n)$ in a model in which tile aggregates must move together through obstacle free paths within the plane. This stands in contrast to the $\Theta(\log n / \log\log n)$ tile types at temperature 2 needed in the basic aTAM. We also provide a general method for simulating a large and computationally universal class of temperature 2 aTAM systems with geometric tiles at temperature 1. Finally, we consider the problem of computing a set of compact geometric faces for a tile system to implement a given set of compatibility specifications. We show a number of bounds on the complexity of geometry size needed for various classes of compatibility specifications, many of which we directly apply to our tile assembly results to achieve non-trivial reductions in geometry size.

💡 Research Summary

**
The paper introduces a substantial extension to Winfree’s abstract Tile Assembly Model (aTAM) by endowing each tile side with a rigid geometric pattern, forming the Geometric Tile Assembly Model (GT‑AM). In GT‑AM, a tile’s “geometry” – a set of protrusions and indentations on its face – can physically block attachment even when the glues match, thereby emulating non‑diagonal glue functions without requiring higher temperature thresholds. The authors show that this geometric restriction enables temperature‑1 assembly of complex shapes that previously needed temperature 2 in the classic aTAM.

The first major result is that an n × n square can be assembled using only Θ(√log n) distinct tile types at temperature 1. This matches an information‑theoretic lower bound derived from Kolmogorov complexity and dramatically improves on the Θ(log n / log log n) tile‑type requirement for temperature‑2 aTAM constructions. The construction encodes binary strings into the geometry of tile edges, using “jigsaw‑like” teeth to control growth direction and enforce correct binding.

Next, the authors demonstrate that a broad class of temperature‑2 aTAM systems known as zig‑zag systems – which are capable of simulating arbitrary Turing machines – can be simulated by temperature‑1 GT‑AM systems without increasing tile‑type count or assembly size. This provides indirect evidence that temperature‑1 systems, when equipped with geometric constraints, may achieve computational universality, a property previously conjectured to be impossible for plain temperature‑1 aTAM.

The paper then extends the model to the two‑handed setting, defining the Two‑Handed Geometric Tile Assembly Model (2GAM). In 2GAM, large supertiles may attach to each other only if a collision‑free planar path exists for them to slide into position, respecting both glue equality and geometric compatibility. Leveraging this restriction, the authors construct an O(log log n)‑tile‑type temperature‑1 assembly of an n × n square. The construction uses intricate geometric “teeth” patterns of size O(log n log log n); a three‑dimensional variant is also presented that keeps all components connected while preserving the same asymptotic tile‑type bound.

A substantial portion of the work is devoted to the algorithmic problem of designing minimal‑size geometries that satisfy a given compatibility matrix (a binary specification of which tile faces may bind). The authors formalize this as a matrix‑realization problem over subsets of ℤ_w × ℤ_ℓ and provide upper and lower bounds on the required geometry size for various matrix families. For arbitrary n × m binary matrices, they give an O(n m)‑time algorithm producing geometries of size Θ(log n). Special cases such as diagonal‑one, diagonal‑zero, and block‑diagonal matrices admit tighter bounds, often logarithmic in the dimension. These results directly support the constructions above, showing that the geometric patterns needed for the efficient assemblies can be generated efficiently and with provably small size.

Overall, the paper contributes three intertwined advances: (1) a geometric augmentation of aTAM that replaces higher temperature with physical shape constraints; (2) concrete temperature‑1 constructions that dramatically reduce tile‑type complexity for classic shapes and for universal computation; (3) a rigorous analysis of the geometry‑design problem, establishing both algorithmic feasibility and theoretical limits. The work bridges theoretical self‑assembly with practical nanofabrication techniques such as DNA origami, where “jigsaw‑like” faces are already experimentally realizable. It opens new avenues for low‑temperature, low‑complexity self‑assembly in both two‑ and three‑dimensional settings, and suggests future research directions in error‑tolerant geometry design, automated collision‑free path planning for two‑handed assemblies, and integration with CAD tools for DNA‑based nanostructure fabrication.