Underlay femtocells have recently emerged as a key technology that can significantly improve the coverage and performance of next-generation wireless networks. In this paper, we propose a novel approach for interference management that enables a number of femtocells to cooperate and improve their downlink rate, by sharing spectral resources and suppressing intra-tier interference using interference alignment. We formulate a coalitional game in partition form among the femtocells and propose a distributed algorithm for coalition formation. Using our approach, the femtocell access points can make individual decisions on whether to cooperate or not, while maximizing a utility function that captures the cooperative gains and the costs in terms of transmit power for information exchange. We show that, using the proposed coalition formation algorithm, the femtocells can self-organize into a network partition composed of disjoint femtocell coalitions, which constitutes the recursive core of the game. Simulation results show significant gains in terms of average payoff per femtocell, reaching up to 30% relative to the non-cooperative scheme.