A complete analogue of Hardy's theorem on \(SL_2(\mathbb{R})\) and characterization of the heat kernel (Q1860621)

Let \(f\) be a measurable function on \(\mathbb{R}\) such that \(|f(x) |\leq Ae^{-\mu x^2}\) and \(|\widehat f(y)|\leq Be^{-\lambda y^2}\) for positive constants \(A,B,\mu\), and \(\lambda\). Here \(\widehat f\) denotes the Fourier transform of \(f\). A theorem of Hardy states that (i) \(\mu\lambda> {1\over 4}\) implies \(f=0\), while (ii) \(\mu\lambda= {1\over 4}\) implies \(f\) is a constant multiple of the Gaussian \(e^{-\mu x^2}\). Since this Gaussian is the heat kernel on \(\mathbb{R}\) associated to the standard Laplacian at a fixed time, Hardy's theorem can be summarized as providing a characterization of the heat kernel on \(\mathbb{R}\) in terms of growth of the function and its Fourier transform. Analogous results have been sought in the setting of Lie groups and symmetric spaces. In the past most of these results have generalized only part (i) above. The present paper extends both parts (i) and (ii) of Hardy's theorem to \(SL_2 (\mathbb{R})\). The generalization of part (i) states that if a measurable function \(f\) on \(SL_2(\mathbb{R})\) and its group Fourier transform decay sufficiently rapidly, then \(f\) is the zero function. The generalization of part (ii) identifies nonzero functions on \(SL_2(\mathbb{R})\) for which the joint rate of decay of the function and its Fourier transform is as sharp as possible. Interestingly, this result can be interpreted as a characterization of the pseudo-heat kernel on \(SL_2(\mathbb{R})\) associated to the Casimir operator \(\Omega\). The author adds the prefix ``pseudo'' since the Casimir operator on \(SL_2 (\mathbb{R})\) is hyperbolic and not elliptic. Because of this, the full action of the semigroup \(e^{t\Omega}\) may not be represented as integration against a ``nice'' function. However, if one restricts this semigroup to certain subspaces of functions, then the action of the semigroup \(e^{t\Omega}\) on each subspace is given by convolution with a unique \(t\)-dependent function and it is these kernel functions that are identified in part (ii) of the theorem. The appropriate function subspaces are those that satisfy a certain finiteness condition stated in terms of the usual left and right actions of the subgroup \(SO_2(\mathbb{R})\) on functions.