Résumé:

We use a standard Monte Carlo algorithm to study the slow dynamics of a binary Lennard-Jones glass-forming mixture at low temperature. We find that the Monte Carlo approach is by far the most efficient way to simulate a stochastic dynamics since the relaxation is about 10 times faster than in Brownian dynamics and about 30 times faster than in stochastic dynamics. Moreover, the average dynamical behaviour of the system is in quantitative agreement with that obtained using Newtonian dynamics, apart from at very short times where thermal vibrations are suppressed. We show, however, that dynamic fluctuations quantified by four-point dynamic susceptibilities do retain a dependence on the microscopic dynamics, as recently predicted theoretically.