Information-powered engines

Information engines are a modern realization of a Gedankenexperiment proposed by James Clerk Maxwell in 1867. They can convert thermal fluctuations into useful work by applying feedback based on the available information about a system's state. The engine was conceived by Maxwell to refine our understanding of the second law of thermodynamics; however, it was not possible to realize one physically until recently. Significant improvements in technology and theory in the last few decades have led to recent experimental realizations of these engines, after nearly 150 years. The first generation of experiments was designed to understand the second law of thermodynamics and to show that information is also a physical quantity, governed by the laws of thermodynamics. In this thesis, we build and study information engines to produce energy for practical purposes. First, we discuss the necessary theoretical and experimental background required to understand the studies presented in this thesis. Then, we build an information engine that converts thermal fluctuations into stored gravitational potential energy, which can be used later on demand. Next, we find the optimization principles that maximize the rate of stored energy and directed velocity; the outputs achieved by the engine were at least ten times higher than previously reported engines. As the bead undergoes stochastic motion, it is challenging to control the trajectory of the bead. Thus, we develop a feedback algorithm that can control the position of a microscopic bead to follow a desired trajectory, without doing any work on it. Usually the performance of information engines degrades in the presence of measurement noise. We find that, above a critical noise level, the information engine shows a phase transition to a non-operating state. Here, we present a Bayesian information engine that can optimally estimate the bead position from noisy measurements to make better feedback decisions and improve performance. Finally, we use stochastic electrical forces to create a nonequilibrium bath and find that our information engine can extract up to ten times more energy from a nonequilibrium environment compared to a thermal bath. We find that the principles of operating information engines are similar to those of energy harvesting micro- and macro-scopic engines. This thesis aims to broaden the perception of information engines from an academic concept to useful devices. At the end of the thesis, we ask whether molecular motors also use fluctuations in their environment. We perform a study that presents methods to identify information engines in simple cases.