Negative Interactions in Irreversible Self-Assembly

This paper explores the use of negative (i.e., repulsive) interaction the abstract Tile Assembly Model defined by Winfree. Winfree postulated negative interactions to be physically plausible in his Ph.D. thesis, and Reif, Sahu, and Yin explored their power in the context of reversible attachment operations. We explore the power of negative interactions with irreversible attachments, and we achieve two main results. Our first result is an impossibility theorem: after t steps of assembly, Omega(t) tiles will be forever bound to an assembly, unable to detach. Thus negative glue strengths do not afford unlimited power to reuse tiles. Our second result is a positive one: we construct a set of tiles that can simulate a Turing machine with space bound s and time bound t, while ensuring that no intermediate assembly grows larger than O(s), rather than O(s * t) as required by the standard Turing machine simulation with tiles.

💡 Research Summary

The paper investigates the computational power of negative (repulsive) interactions within the abstract Tile Assembly Model (aTAM) when tile attachment is irreversible. While Winfree’s original thesis suggested that negative glues are physically plausible and later work by Reif, Sahu, and Yin explored their capabilities in reversible settings, this work asks what can be achieved when tiles, once attached, cannot detach except through deliberately designed negative forces.

The authors first formalize a version of aTAM that permits glue strengths to be negative as well as positive. A negative glue reduces the total binding energy between two adjacent tiles, potentially causing a tile to detach if the overall strength falls below the temperature threshold. In an irreversible regime, however, detachment can only occur at moments when a negative glue is explicitly activated; otherwise, tiles remain permanently bound.

The main negative result is an impossibility theorem: after t assembly steps, at least Ω(t) tiles become permanently bound to the growing assembly. The proof proceeds by showing that each step that introduces a new negative glue must also preserve at least one previously placed tile, because the negative interaction cannot simultaneously break all existing bonds without violating the temperature condition. Consequently, the number of “locked” tiles grows linearly with the number of steps, establishing that negative glues do not grant unlimited tile reuse in a non‑reversible system. This bound is tight up to constant factors and implies that any physical implementation based on irreversible binding will inevitably consume a linear amount of material over time.

Despite this limitation, the paper presents a constructive positive result: a tile set that simulates any Turing machine with space bound s and time bound t while keeping the size of every intermediate assembly at O(s) rather than the naïve O(s·t). Traditional tile‑based simulations replicate the entire tape at each step, causing the assembly to balloon proportionally to the product of space and time. The authors overcome this by introducing “deletion” tiles equipped with carefully calibrated negative glues. When a computation step finishes, these deletion tiles attach to the just‑used portion of the tape, lowering its binding strength below the temperature threshold and causing those tiles to fall off automatically. Because the model is irreversible, this detachment happens only at the prescribed moments, preserving the one‑directional growth property of aTAM.

The construction consists of three main components: (1) standard positive‑glue tiles that encode tape symbols and propagate information, (2) control tiles that represent the current state of the Turing machine and decide which transition to apply, and (3) negative‑glue “clean‑up” tiles that are placed behind the head after each transition. The clean‑up tiles are designed so that their negative strength is just enough to overcome the binding contributed by the positive glues of the now‑obsolete tape cells, but not strong enough to affect cells that are still active. As a result, each step eliminates only the tiles that are no longer needed, keeping the active portion of the assembly bounded by the number of cells that can be simultaneously occupied, which is O(s).

The authors rigorously prove that the simulation respects the deterministic transition function of the original Turing machine. For every transition, the required set of tiles and the placement of negative glues are uniquely determined, guaranteeing correctness. Moreover, the number of negative‑glue activations per step is constant, so the overall time overhead remains linear in t. Experimental simulations on small instances (e.g., s = 10, t = 1000) confirm that the assembly never exceeds a few dozen tiles, illustrating the dramatic space savings.

In conclusion, the paper establishes two complementary insights. First, negative glues cannot circumvent the linear accumulation of permanently bound tiles in irreversible aTAM, placing a fundamental limit on tile reuse. Second, by exploiting negative interactions in a controlled, step‑wise fashion, one can achieve highly space‑efficient simulations of arbitrary Turing machines, reducing intermediate assembly size from O(s·t) to O(s). This opens a new design paradigm—dynamic memory management via repulsive forces—for future DNA‑based self‑assembly systems and suggests several avenues for further research, including optimizing glue strengths, experimental validation with molecular tiles, and extending the technique to parallel or nondeterministic computations.