We study the intermittent dynamics and the fluctuations of the dynamic correlation function of a simple aging system. Given its size L and its coherence length ξ, the system can be divided into N independent subsystems, where N=(L/ξ)^d, and d is the dimension of space. Each of them is considered as an aging subsystem which evolves according to an activated dynamics between energy levels. We compute analytically the distribution of trapping times for the global system, which can take power-law,

more »

... tretched-exponential or exponential forms according to the values of N and the regime of times considered. An effective number of subsystems at age t_w, N_eff(t_w), can be defined, which decreases as t_w increases, as well as an effective coherence length, ξ(t_w) ∼ t_w^(1-μ)/d, where μ <1 characterizes the trapping times distribution of a single subsystem. We also compute the probability distribution functions of the time intervals between large decorrelations, which exhibit different power-law behaviours as t_w increases (or N decreases), and which should be accessible experimentally. Finally, we calculate the probability distribution function of the two-time correlator. We show that in a phenomenological approach, where N is replaced by the effective number of subsystems N_eff(t_w), the same qualitative behaviour as in experiments and simulations of several glassy systems can be obtained.