In cognitive radio networks, secondary user performs spectrum sensing to detect primary user activities and to access the channel opportunistically. We investigate the problem of developing energy efficient opportunistic spectrum access strategy for a secondary user with energy harvesting capability. We formulate the problem to determine the optimal sensing and access policy as a partially observable Markov decision process (POMDP). Since obtaining the optimal policy is computationally prohibitive, we provide a suboptimal myopic policy that maximizes expected immediate throughput instead of expected total throughput. Simulation results demonstrate that the proposed myopic policy yields improved throughput compared to a random channel selection strategy.