Ilke Ercan
Bogazici University

Energy Efficiency Limits in Brownian Circuits

The saturation in the efficiency and performance scaling of conventional electronic technologies instigates the development of novel computational paradigms. Brownian computing is among the promising alternatives that can exploit fluctuations in circuits to increase the efficiency of information processing in electronics. A Brownian cellular automaton, where signals propagate randomly and are driven by local transition rules, can be made computationally universal by embedding arbitrary asynchronous circuits on it.
In this paper, we design a Brownian half adder based on the methods proposed by Peper et al and perform a physical-information-theoretic analysis on the efficiency limitations in Brownian circuits. The theoretical approach we propose here goes beyond the earlier methodologies we developed by capturing the stochasticity as well as the asynchronicity in Brownian circuits and yields fundamental lower bounds on energy efficiency of information processing in Brownian computing. Single Electron Tunnelling (SET) devices enable the simulation of noise and fluctuations in a fashion similar to Brownian search and are therefore employed as an illustrative example in realizing the circuit we design. The method we develop establishes a solid ground that enables studying computational and physical features of this emerging technology on an equal footing, and yield fundamental lower bounds that provide valuable insights into how far the efficiency of this computing strategy can be improved in principle. In determining the fundamental bounds on energy efficiency in Brownian circuits we also provide a physical-information-theoretic comparison of the proposed performance improvement of this technology proposal against its alternatives.