A new optimal estimate for the norm of time-frequency localization operators

In this paper we provide an optimal estimate for the operator norm of time-frequency localization operators L_F,φ : L^2(R^d) -> L^2(R^d), with Gaussian window φ and weight F, under the assumption that F is an element of L^p(R^2d) and L^q(R^2d) for some p and q in (1, ∞). We are also able to characterize optimal weight functions, whose shape turns out to depend on the ratio||F||_q/||F||_p. Roughly speaking, if this ratio is "sufficiently large" or "sufficiently small" optimal weight functions are certain Gaussians, while if it is in the intermediate regime the optimal functions are no longer Gaussians. As an application, we extend Lieb's uncertainty inequality to the space L^p + L^q.