1: \begin{abstract}

2:   Monte Carlo simulations of systems of particles such as hard spheres

3:   or soft spheres with singular kernels can display around a phase

4:   transition prohibitively long convergence times when using

5:   traditional Hasting-Metropolis reversible schemes. Efficient

6:   algorithms known as event-chain Monte Carlo were then developed to

7:   reach necessary accelerations. They are based on non-reversible

8:   continuous-time Markov processes. Proving invariance and ergodicity

9:   for such schemes cannot be done as for discrete-time schemes and a

10:   theoretical framework to do so was lacking, impeding the

11:   generalisation of ECMC algorithms to more sophisticated systems or

12:   processes. In this work, we characterize the Markov processes

13:   generated in ECMC as piecewise deterministic Markov processes. It first

14:   allows us to propose more general schemes, for instance regarding

15:   the direction refreshment. We then prove the invariance of the correct

16:   stationary distribution. Finally, we show the ergodicity of the processes in soft-

17:   and hard-sphere systems, with a density condition for the latter.

18: 

19:   \keywords{Markov-chain Monte Carlo, ECMC, PDMP, particle systems,

20:     hard spheres, soft spheres, ergodicity, generator}

21: \end{abstract}

22: