Lecturer: Dr. Yael Roichman, School of Chemistry, Tel Aviv University
Abstract:
 We realize experimentally and in simulation an information machine which converts information to work. In our realization, a single particle is diffusing in a one dimensional channel. Fluid flow in the channel drags the particle along one direction preferentially. A barrier placed at the end of the channel confines the particle to one side of the channel. If the fluid flow is weak in comparison to thermal forces, the particle is able to diffuse upstream, against the fluid flow, occasionally. During the machine’s operation the particle is monitored continuously, each time it diffuses far enough upstream, away from the barrier, a feedback mechanism moves the barrier slightly closer to the particle, without applying force on the particle or the fluid. In a quasi-static mode of operation the amount of information used by the feedback protocol is easily calculated through the Shannon entropy of uncorrelated steps. However, at faster rates of operation temporal correlations arise and their full characterization is required for assessment of the information used to operate the machine. We develop a scheme to calculate information at steady-state operation based on the measured temporal correlations. We use this calculation to characterize the efficiency of our information machine. Surprisingly, we find that while the output power and information gain are optimized for short feedback cycle times, maximum efficiency is obtained at quasi-static conditions.