Sequential non-determinism in tile self-assembly: a general framework and an application to efficient temperature-1 self-assembly of squares

TL;DR

This paper introduces a framework called sequential non-determinism to analyze probabilistic tile assembly systems, enabling efficient construction of large squares with high probability using fewer tile types than previous methods.

Contribution

The paper develops a novel analytical framework for probabilistic tile assembly, and demonstrates its application in creating efficient, high-probability self-assembly of large squares with fewer tile types.

Findings

Probabilistic correctness can be expressed as a product of individual attachment probabilities.

Constructed tile assembly systems can form N×N squares with high probability using logarithmic tile types.

The method improves upon previous bounds for tile complexity in square self-assembly.

Abstract

In this paper, we work in a 2D version of the probabilistic variant of Winfree's abstract Tile Assembly Model defined by Chandran, Gopalkrishnan and Reif (SICOMP 2012) in which attaching tiles are sampled uniformly with replacement. First, we develop a framework called ``sequential non-determinism'' for analyzing the probabilistic correctness of a non-deterministic, temperature-1 tile assembly system (TAS) in which most (but not all) tile attachments are deterministic and the non-deterministic attachments always occur in a specific order. Our main sequential non-determinism result equates the probabilistic correctness of such a TAS to a finite product of probabilities, each of which 1) corresponds to the probability of the correct type of tile attaching at a point where it is possible for two different types to attach, and 2) ignores all other tile attachments that do not affect the non-deterministic attachment. We then show that sequential non-determinism allows for efficient and geometrically expressive self-assembly. To that end, we constructively prove that for any positive integer $N$ and any real $\delta \in (0,1)$, there exists a TAS that self-assembles into an $N \times N$ square with probability at least $1 - \delta$ using only $O\left( \log N + \log \frac{1}{\delta} \right)$ types of tiles. Our bound improves upon the previous state-of-the-art bound for this problem by Cook, Fu and Schweller (SODA 2011).