Superposition Coding-Based Bounds and Capacity for the Cognitive Z-Interference Channels

This paper considers the cognitive interference channel (CIC) with two transmitters and two receivers, in which the cognitive transmitter non-causally knows the message and codeword of the primary transmitter. We first introduce a discrete memoryless more capable CIC, which is an extension to the more capable broadcast channel (BC). Using superposition coding, we propose an inner bound and an outer bound on its capacity region. The outer bound is also valid when the primary user is under strong interference. For the Gaussian CIC, this outer bound applies for $|a| \geq 1 $, where $a$ is the gain of interference link from secondary user to primary receiver. These capacity inner and outer bounds are then applied to the Gaussian cognitive Z-interference channel (GCZIC) where only the primary receiver suffers interference. Upon showing that jointly Gaussian input maximizes these bounds for the GCZIC, we evaluate the bounds for this channel. The new outer bound is strictly tighter than other outer bounds on the capacity of the GCZIC at strong interference ($a² \geq 1 $). Especially, the outer bound coincides with the inner bound for $|a| \geq \sqrt{1 + P_1}$ and thus, establishes the capacity of the GCZIC at this range. For such a large $a$, superposition encoding at the cognitive transmitter and successive decoding at the primary receiver are capacity-achieving.