Information friction limits on computation

The recently proposed “Information friction” model accounts for energy losses incurred in moving bits on a computational substrate and was first studied in the context of encoding and decoding computations for communication. Information friction loss is modeled as being proportional to bit-meters, the sum of the lengths over which the bits are transported during the computation. Its analysis provides us with an understanding of the fundamental energy requirements for computation. In this paper, we obtain lower bounds on information friction for several canonical computations that have been analyzed to obtain “AT²” bounds in the context of what is called “VLSI complexity” and, more recently, in deriving computation throughput in the context of wireless sensor networks.