Speaker
Prof.
Udo Seifert
(Univ. Stuttgart)
Description
The framework of stochastic thermodynamics can be applied to
Brownian information machines for which information about
the system acquired in a measurement is used to extract work
from a single heat bath. Fluctuation theorems have been
generalized to such feedback-driven non-autonomous machines
following an almost standard recipe also allowing to discuss
their efficiency and efficiency at maximum power.
After briefly recalling this (reasonably well-understood)
class, I will describe our recent work dealing with autonomous
machines. First, I will discuss a fully stochastic,
reversible variant of the demon recently introduced by
Mandal and Jarzynski [PNAS 109, 11641, 2012]. Our
generalization which includes genuine equilibrium allows to
identify Onsager coefficients and the linear response theory
of such a demon [1]. Second, within a minimal model for
cellular sensing, I will discuss the relation between
information-theoretic and thermodynamic entropy production.
While one could naively expect the rate of information to be
bounded by the thermodynamic cost of acquiring it, based on
our new bound on the rate of mutual information for
time-continuous processes, I will show that there is no such
inequality [2].
[1] AC Barato and US, arXiv:1302.3089
[2] AC Barato, D. Hartich and US, arXiv:1212.3186
