This paper focuses on energy efficiency that is a key performance metric in heterogeneous cellular networks to two key areas. First, based on the Poisson point process distributions of small-cell base stations (SBSs) and macrocell base stations (MBSs), the energy-efficiency model is formulated, and the effect of base stations’ distribution on energy efficiency is analyzed. For maximizing energy efficiency, the joint optimal densities of SBSs and MBSs are deduced under the constraint of quality of service. Second, according to this, we propose a joint sleep strategy of MBSs and that of SBSs. We deduce the optimal threshold of traffic load according to the joint optimal densities. If the traffic load of SBSs (or MBSs) is less than the optimal threshold of traffic load, these SBSs (or MBSs) go to sleep; otherwise, it is activated. This makes the SBSs and MBSs adaptively and distributively sleep according to their own traffic loads. The simulation results verify that the deduced joint optimal densities of the SBS and the MBS are accurate, and energy efficiency is improved when SBSs and MBSs adaptively sleep.