Join us this week for a long and interesting conversation with Tom Ouldridge of Imperial College London on Maxwell’s demon, Szilard’s engine, what people get wrong about thermodynamics and information theory, how this all relates to biology, and how his lab is using these ideas to develop exciting new approaches to molecular programming.
Tom Ouldridge is a Royal Society University Research Fellow in the Bioengineering Department, where he leads the “Principles of Biomolecular Systems” group. His group probes the fundamental principles underlying complex biochemical systems through theoretical modelling, simulation and experiment. In particular, they focus on the interplay between the detailed biochemistry and the overall output of a process such as sensing, replication or self-assembly. They are inspired by natural systems, and aim to explore the possibilities of engineering artificial analogs.
We start by discussing Maxwell’s demon and Szilard’s engine—thought experiments from the 19th and early 20th centuries about the interplay of thermodynamics and information-processing. These have long captured the imagination of theoretical physicists. There is renewed interest in these thought experiments due our increasing ability to control systems at the molecular level. Many still disagree about the interpretation of these ideas, the implications for the second law of thermodynamics, and the consequences for thermodynamics of computation.
Szilard’s engine is a simpler version of Maxwell’s thought experiment, but which is mathematically tractable, considering only a single particle separated by a divider attached to a weight. If the particle and the weight are on the same side, then the particle can bounce against the divider and lift the weight, doing work. By resetting the divider, this step can be repeated to extract more work. Tom talks about how this seeming paradox may be resolved.
Tom discusses how his group has implemented a theoretical Szilard engine in biomolecules; by explicitly rendering each step of the engine as a biochemical process (using cell surface receptors). This helps demystify the whole process by rendering all “information theoretic” steps as concrete, real, processes. Doing so is helpful not only in resolving old thought experiments, but because the crucial idea—that the generation of correlation between non-interacting degrees of freedom is thermodynamically costly—is of fundamental significance to natural and synthetic molecular information-processing systems.
Find more information at the episode page here: