Mutual entropy production in bipartite systems

It was recently shown by Barato et al (2013 Phys. Rev. E 87 042104) that the mutual information at the trajectory level of a bipartite Markovian system is not bounded by the entropy production. In the same way as Gaspard showed (2004 J. Stat. Phys. 117 599) that the entropy production is not directly related to the Shannon entropy at the trajectory level but is in fact equal to its difference from the so-called time-reversed Shannon entropy, we show in this paper that the difference between the mutual information and its time-reversed form is equal to the mutual entropy production (MEP), i.e. the difference between the full entropy production and that of the two marginal processes. Evaluation of the MEP is in general a difficult task due to non-Markovian effects. For bipartite systems, we provide closed expressions in various limiting regimes which we verify by numerical simulations. (paper)