An Information Inequality Motivated by the Gaussian Z-Interference Channel

We establish an information inequality that is motivated by the capacity region computation for the Gaussian Z-interference channel. This yields an improved slope for the capacity region at Costa's corner point. We believe the inequality may also be of independent interest as it provides a non-trivial upper bound on the entropy of sums of independent random variables.