The power of being explicit: demystifying work, heat, and free energy in the physics of computation

Back to all publications

Publication date
Reference T.E. Ouldridge, R.A. Brittain and P.R. ten Wolde: The power of being explicit: demystifying work, heat, and free energy in the physics of computation In: The Energetics of Computing in Life and Machines, SFI Press, 2018.
Group Biochemical Networks

Interest in the thermodynamics of computation has revived in recent years,driven by developments in science, economics and technology. Given theconsequences of the growing demand for computational power, the idea ofreducing the energy cost of computations has gained new importance.Simultaneously, many biological networks are now interpreted asinformation-processing or computational systems constrained by their underlyingthermodynamics. Indeed, some suggest that low-cost, high-density biologicalsystems may help to mitigate the rising demand for computational power and the”end” of Moore’s law of exponential growth in the density of transistors. In this chapter we address widespread misconceptions about thermodynamics andthe thermodynamics of computation. In particular, we will argue against thegeneral perception that a measurement or copy operation can be performed at nocost, against the emphasis placed on the significance of erasure operations,and against the careless discussion of heat and work. While not universal,these misconceptions are sufficiently prevalent (particularly withininterdisciplinary contexts) to warrant a detailed discussion. In the process,we will argue that explicitly representing fundamental processes is a usefultool, serving to demystify key concepts. We first give a brief overview of thermodynamics, then the history of thethermodynamics of computation – particularly in terms of copy and measurementoperations inherent to classic thought experiments. Subsequently, we analysethese ideas via an explicit biochemical representation of the entire cycle ofSzilard’s engine. In doing so we show that molecular computation is both apromising engineering paradigm, and a valuable tool in providing fundamentalunderstanding.