Publications

    Information means a useful message (or its amount) about a state of a system's component, which can be transferred to another component. In physics, it indicates how much a subsystem learns about the state of another subsystem, revealing the structure of measurement and feedback.

    Our idea starts from the fact that, even when the subsystems play the identical role in average, their roles as information sender/recipient can be spontaneously divided at the level of the fluctuating trajectories. To capture this trajectory-dependent information-theoretic effect, we focus on the stochastic information flow (SIF), and develop the analytical and neural method to calculate the quantity. It is shown that the variance of SIF reveals rich dynamical structure of various time-series data which could not be captured by previous 'averaged' concepts.

Chemokinesis refers to a phenomenon where the speed of active particle is modulated depending on the concentration of chemicals around the particle. We explore how does the collective phenomena of active matter is modulated by this effect. It is shown that, 

Heat engine consisting of active particles are known to exhibit unconventionally high performance; its efficiency can apparently exceed the Carnot bound. Beyond the phenomenological description, starting from a Langevin framework with guaranteed equilibration, we construct a thermodynamically consistent model of active engine where the source of activity (chemical fuel consumption) is explicit. By identifying the energy flows involved, we then construct a novel concept of efficiency which is properly bounded by the second law of thermodynamics.