We consider a set of cellular users associated with a base station (BS) in a cellular network that employs Device-to-device (D2D) communication. A subset of the users request for some files from the BS. Now, some of the users can potentially act as relays and forward the requested files, or partitions of files, from the BS to some of the requesting users (destination nodes) over D2D links. However, this requires cooperation among the cellular users. In this paper, we seek conditions under which users have an incentive to cooperate with each other. We model the above scenario using the frameworks of cooperative game theory and stable partitions in coalitional games. We consider two different models for file transfer within a coalition: (i) Model A, in which the BS can split a file into multiple partitions and send these partitions to different relays, which multicast the partitions to the destination nodes of the coalition, and (ii) Model B, in which for each file, the BS sends the entire file to a single relay, which multicasts it to the destination nodes of the coalition. First, we explore the question of whether it is beneficial for all the cellular users to cooperate, i.e., whether the grand coalition is stable. For this we use the solution concept of core from cooperative game theory. We show that, in general, the above coalitional game under Model A may have an empty core. Next, we provide conditions under which the core is always non-empty and a D_c-stable partition always exists. Also, we show that under Model B, the problem of assigning relays to destination nodes so as to maximize the sum of utilities of all the users is NP-Complete. Finally, we show via numerical computations that a significant reduction in the energy expenditure of cellular users can be achieved via cooperation.