RETLab: A fast design-automation framework for arbitrary RET networks

Resonance energy transfer (RET) circuits are networks of photo-active molecules that can implement arbitrary logic functions. The nanoscale size of these structures can bring high-density computation to new domains, e.g., in vivo sensing and computation. A key challenge in the design of a RET network is to find, among a huge set of configurations (i.e., design space), the optimum choice and arrangement of molecules on a nanostructure. The prohibitively large size of the design space makes it impractical to evaluate every possible configuration, motivating the need for design-space pruning to be integrated into the design flow. To this end, we have developed a computer-aided design framework, called RETLab, that enables structured pruning of the design space to extract a sufficiently small subset, which is fully evaluated and ranked based on user-defined metrics to yield the best configuration. More importantly, we have developed a new RET-simulation algorithm, which is several orders of magnitude (e.g., for a 4-node network, one million times) faster than the conventional Monte-Carlo-based simulation (MCS). This speedup in configuration evaluation enables a significantly more extensive design-space exploration with fewer and less constrained heuristics, compared to existing RET-network design methods which are ad-hoc and rely on MCS for configuration evaluation.