Is Computation Reversible?

Abstract

Recent investigations into the physical nature of information and fundamental limits to information transmission have revealed questions such as the possibility of superluminal data transfer or not; and whether reversible computation (information processing) is feasible. In some respects these uncertainties stem from the determination of whether information is inherent in points of non-analyticity (discontinuities) or smoother functions. The close relationship between information and entropy is also well known, e.g. Brillouin's concept of negentropy (negative entropy) as a measure for information. Since the leading edge of a step-discontinuity propagates in any dispersive medium at the speed of light in vacuum as a precursor to the main body of the dispersed pulse, we propose in this paper to treat information as being intrinsic to points of non-analyticity (discontinuities). This allows us to construct a theory addressing these dilemmas in a fashion consistent with causality, and the fundamental laws of thermodynamics. A consequence of our proposition is that the movement of information is always associated with the dissipation of heat, and therefore that the concept of reversible classical computation is not tenable.