In this paper we review a series of results obtained for 1D and 2D simple N-body dynamical models with infinite-range attractive interactions and without short distance singularities. The free energy of both models can be exactly obtained in the canonical ensemble, while microcanonical results can be derived from numerical simulations. Both models show a phase transition from a low energy clustered phase to a high energy gaseous state, in analogy with the models introduced in the early 70's by Thirring and Hertel. The phase transition is second order for the 1D model, first order for the 2D model. Negative specific heat appears in both models near the phase transition point. For both models, in the presence of a negative specific heat, a cluster of collapsed particles coexists with a halo of higher energy particles which perform long correlated flights, which lead to anomalous diffusion. The dynamical origin of the "superdiffusion" is different in the two models, being related to particle trapping and untrapping in the cluster in 1D, while in 2D the channelling of particles in an egg-crate effective potential is responsible of the effect. Both models are Lyapunov unstable and the maximal Lyapunov exponent $\lambda$ has a peak just in the region preceeding the phase transition. Moreover, in the low energy limit $\lambda$ increases proportionally to the square root of the internal energy, while in the high energy region it vanishes as $N^{-1/3}$.

A system of N classical particles in a 2D periodic cell interacting via long-range attractive potential is studied. For low energy density $U$ a collapsed phase is identified, while in the high energy limit the particles are homogeneously distributed. A phase transition from the collapsed to the homogeneous state occurs at critical energy U_c. A theoretical analysis within the canonical ensemble identifies such a transition as first order. But microcanonical simulations reveal a negative specific heat regime near $U_c$. The dynamical behaviour of the system is affected by this transition : below U_c anomalous diffusion is observed, while for U > U_c the motion of the particles is almost ballistic. In the collapsed phase, finite $N$-effects act like a noise source of variance O(1/N), that restores normal diffusion on a time scale diverging with N. As a consequence, the asymptotic diffusion coefficient will also diverge algebraically with N and superdiffusion will be observable at any time in the limit N \to \infty. A Lyapunov analysis reveals that for U > U_c the maximal exponent \lambda decreases proportionally to N^{-1/3} and vanishes in the mean-field limit. For sufficiently small energy, in spite of a clear non ergodicity of the system, a common scaling law \lambda \propto U^{1/2} is observed for any initial conditions.

A two dimensional self-gravitating Hamiltonian model made by $N$ fully-coupled classical particles exhibits a transition from a collapsing phase (CP) at low energy to a homogeneous phase (HP) at high energy. From a dynamical point of view, the two phases are characterized by two distinct single-particle motions : namely, superdiffusive in the CP and ballistic in the HP. Anomalous diffusion is observed up to a time $\tau$ that increases linearly with $N$. Therefore, the finite particle number acts like a white noise source for the system, inhibiting anomalous transport at longer times.