On Brownian computation

as a way of demonstrating that logically reversible mathematical operations can be computed by physical processes that are thermodynamically reversible, or nearly so. In a recent paper 4 , I have argued that the computational processes they describe are not thermodynamically reversible and that their misidentification rested on a misapplication of thermal and statistical physics. A question left open by my critique was whether there is some way to repair the computational protocol used in a Brownian computer so that its operation becomes thermodynamically reversible; or at least as close to thermodynamic reversibility for a single step process as the “no-go” result of Ref. 5 (Part II) allows. My purpose in this note is to describe such a repair. It must be emphasized here that the repair is my proposal and that it conflicts with an assumption used in the Bennett and Landauer analysis. The proposed repair leads to a Brownian computer that will compute extremely slowly. Bennett (Ref. 2, pp. 905-906, 922-23; Ref. 1, pp. 531-32) has urged that if the Brownian computer’s operation becomes too slow, a thermodynamically dissipative driving force must be introduced to drive it to completion. In the case of logically irreversible computation, the resulting dissipation is identified with the dissipation required by Landauer’s Principle. My view is that In t. J. M od . P hy s. C on f. S er . 2 01 4. 33 . D ow nl oa de d fr om w w w .w or ld sc ie nt if ic .c om