Abstract
The second lecture tried to show how, starting from a microscopic Hamiltonian dynamics, one can end up with a Markovian dynamics, by partitioning the phase space into cells and approximating the Hamiltonian dynamics by jump probabilities between these cells. Once this Markovian description is adopted, the dynamics becomes irreversible and it can be shown that the microscopically defined entropy is no longer constant, as predicted by Liouville's theorem, but becomes an increasing function of time.
The lecture continued with an introduction to stochastic thermodynamics, showing how the notions of heat, work and dissipation can be defined in the context of any Markov process. Finally, it described several examples of thermostat representation within the framework of Markov dynamics: Andersen thermostat, Monte Carlo dynamics, Langevin equation.
