Abstract:
In classical computing, the energy scales of even the smallest devices, such as transistors and gates, are large compared to that of the thermal energy, k_B T. As a result, there is a clear separation of scales between signal and noise, enabling deterministic computation. This determinism comes with an energy cost: classical computers operate far above the limits of thermodynamic efficiency, and require large amounts of power and heat dissipation to ensure their reliability.

In this talk I will discuss energy-efficient computing at the k_B T scale. There, the comparable scales of signal and noise make deterministic computation challenging: efficient time-dependent protocols are required to do even the simplest logic operations. I will talk about finding such protocols for a laboratory realization of a one-bit memory. I will also talk about the approach taken in the field of thermodynamic computing, which views thermal fluctuations as a resource rather than a problem, and makes use of the tendency of physical systems to evolve toward thermal equilibrium to do computation.

Bio:
Stephen Whitelam got his Ph.D. in theoretical physics in 2004 from Oxford University, where he used statistical mechanics to study the dynamics of model glass-forming liquids. He was supervised by Juan P. Garrahan and David Sherrington. From 2004 – 2007 he did a postdoc with Phillip Geissler at UC Berkeley, using theory and simulation to study protein complex self-assembly and DNA overstretching. From 2007–2008 he was a postdoc with Nigel Burroughs at Warwick University’s Systems Biology Centre, where he worked on actin pattern formation in cells. Since 2008 he has been a staff scientist in the Theory Facility of the Molecular Foundry at Lawrence Berkeley National Laboratory. He uses statistical mechanics and machine learning to study classical phenomena, including self-assembly and unconventional forms of computing, at the nanoscale.