Characterization of eigenfunctions of the Laplace-Beltrami operator using Fourier multipliers (Q2204348)

In 1980, \textit{J. Roe} [Math. Proc. Camb. Philos. Soc. 87, 69--73 (1980; Zbl 0463.33002)] proved that, given a bi-infinite sequence of functions \(f_k\), \(k\in \mathbb Z\), defined on \(\mathbb R\), such that \(f_{k+1}(x)={f_k'}{(x)}\) and \(|f_k(x)|\le C\) for all \(k\in \mathbb Z\), for all \(x\in \mathbb R\) and for some \(C>0\), then \(f_0(x)=a\sin (b+x)\). This provided a characterization of the sine function, which was extended by \textit{R. S. Strichartz} [Trans. Am. Math. Soc. 338, No.~2, 971--979 (1993; Zbl 0815.35012)] and by \textit{R. Howard} and \textit{M. Reese} [Can. Math. Bull. 35, No.~2, 204--213 (1992; Zbl 0764.35068)] to eigenfunctions of the standard Laplacian in \(\mathbb R^n\) associated to some particular eigenvalues. Strichartz also proved that a similar characterization was not possible in the hyperbolic \(3\)-space and it is nowadays known that counterexamples can be constructed in any Riemannian symmetric space of noncompact type. In light of these negative results, in this paper, in order to prove analogous characterizations in a rank one Riemannian symmetric space \(X\) of noncompact type, the authors replace the Laplace-Beltrami operator \(\Delta\) with a translation invariant continuous linear operator \(\Theta\), defined by \(\Theta:\, f\mapsto f *\mu\), where \(\mu\) is a radial (possibly complex) measure on \(X\). These operators are called multipliers and include, in particular, ball, sphere and annular averages and the heat operator. As an example, the authors assume that \(f_k\), \(k\in\mathbb Z\), is a bi-infinite sequence of functions defined on \(X\), such that, for all \(k\in \mathbb Z\), \({\Theta f_k}=A f_{k+1}\) for some \(A\in \mathbb C\), and \(\|f_k\|\le M\) for some \(M>0\). Here the norm \(\|\cdot\|\) is a suitable weak \(L^p\)-norm. Starting from these hypotheses, they try to prove that \( f_0\), and therefore every \(f_k\), is an eigenfunction of \(\Theta\). The results are stronger in the case \(p=2\), where one can take advantage of the one-dimensionality of the \(L^2\)-spectrum.