Integral geometric properties of non-compact harmonic spaces (Q2256824)

A complete Riemannian manifold \((X,g)\) of dimension \(n+1\) is called harmonic if the volume density function in normal coordinates around a point depends only on the distance from the point. The authors explain in the introduction that rank-one symmetric spaces are harmonic, and for compact simply connected spaces, harmonic spaces must be flat or rank-one symmetric. For non-compact manifolds the only simply connected homogeneous harmonic spaces are the examples constructed by \textit{E. Damek} and \textit{F. Ricci} [J. Geom. Anal. 2, No. 3, 213--248 (1992; Zbl 0788.43008)]. Harmonic manifolds are analytic spaces for which all harmonic functions satisfy the mean value property (i.e., the average of \(f\) over any geodesic sphere equals the value of \(f\) at its centre). Recall that in a harmonic space every function satisfying the mean value property \textit{for all points and all radii} must be harmonic (the authors recall that this is not true if a function only satisfies the mean value property for all points in a single radius \(r>0\) and give an example in the introduction of the paper). The main result (Theorem 4.3) proves that in an arbitrary non-compact harmonic space the mean value property for two generic chosen radii \(r_1,r_2\) implies harmonicity of the function. In fact the authors prove the following: Assume that \(\phi _{\lambda}\) is the unique smooth function in the spherical projection from a point in \(x_0\in X\) of the space of smooth functions on \(X\) such that \(\Delta \phi _{\lambda}+\left({\lambda}^2+\frac{H^2}{4}\right)\phi _{\lambda}=0\) and \(\phi _{\lambda}(x_0)=1\). (The existence of this function is proved in Proposition 2.3.) Let \((X,g)\) be a simply connected, non-compact harmonic manifold and \(r_1,r_2 > 0\), such that the equations \(\phi _{\lambda} (r_j)=1\) for \(j=1,2\) have no common solution \(\lambda \in \mathbb{C} \backslash \{\pm i H/2\}\). Then \(f\in C^{\infty}\) is harmonic if and only if \({\displaystyle \frac{1}{\mathrm{vol}(S_r(x))} \int _{S_r(x)}f=f(x)}\) for \(r=r_1,r_2\) and all \(x\in X\). The authors also prove (Theorem 4.2) that under the same conditions for \(X\): (i) if \(\phi _{\lambda} (r_j)=0\) for \(j=1,2\) have no common solution \(\lambda \in \mathbb{C}\) and if for \(f\in C(X)\) we have \(\int _{S_r(x)}f=0\) for \(r=r_1,r_2\) and all \(x\in X\), then \(f=0\). (ii) For \(\lambda \in \mathbb{C}\) if the equations \({\displaystyle \int_0^{r_j}\theta(\rho)\phi _{\lambda}(\rho)d\rho}=0\) for \(j=1,2\) and \(r_1,r_2 > 0\) have no common solution \(\lambda \in \mathbb{C}\) and if for \(f\in C(X)\) we have \(\int _{B_r(x)}f=0\) for \(r=r_1,r_2\) and all \(x\in X\), then \(f=0\). A direct consequence of both these results is that the vanishing of the integral of a function over all spheres, or over all balls, of radii \(r_1,r_2\) implies that the function will be equal to zero if the pair \((r_1, r_2)\) is in a specific set defined in Theorem 4.1. Let \(h(X):= {\displaystyle \inf_{K\subset X} \frac{\mathrm{area}(\partial (K))}{\mathrm{vol}(K)}}\) be the \textit{Cheeger constant}, and \({\displaystyle\mu(X):= \limsup_{r\rightarrow \infty} \frac{\log\mathrm{vol} (B_r(x))}{r}}\) be the \textit{volume growth exponent} of \(X\). The authors also prove (Theorem 5.1) that if \((X,g)\) is a simply connected, non-compact harmonic manifold and \(H\geq 0\) where \(H\) is the mean curvature of its horospheres, then \(h(X) =H= \mu(X)\). The authors remind the reader that it was already proved that the Cheeger constant and the volume growth exponent of a simply connected strictly negatively and homogeneous spaces are equal [\textit{C. Connell}, Commun. Anal. Geom. 8, No. 3, 575--633 (2000; Zbl 0978.53094)]. This result fails without the curvature assumption. Note that in the present theorem there is no condition on the curvature. Let \(p_t^X (x,y)\) be the unique heat kernel for a complete Riemannian manifold \((X,g)\) with Ricci curvature bounded from below. If \(X\) is harmonic, \(p_t^X\) is a radial kernel function and is uniquely determined by \(k_t^X(x):=p_t^X(x_0,x)\), where \(x_0\in X\) is a fixed reference point. The last result (Theorem 5.6) states that for \((X,g)\) as in Theorem 5.1 the Abel transform \(\mathcal{A}k_t^X\) of the heat kernel is \((\mathcal{A}k_t^X)(s)=e^{-H^2t/4}\frac{1}{\sqrt{4\pi t}}e^{-s^2/4t}\), i.e., it agrees up to a factor with the Euclidean heat kernel. The article includes a section with detailed proofs of radial eigenfunctions and convolutions (Section 2) and another (Section 3) with fundamental results for the Abel transform and the spherical Fourier transform that includes the proof that the Abel transform and its dual are topological isomorphisms (Theorem 3.8). Most proofs in the article are given in detail. The proofs of the major theorems in Section 4 are outlined since they follow closely the analogous results proved for Damek-Ricci spaces in the authors' article [Ark. Mat. 48, No. 1, 131--147 (2010; Zbl 1189.43005)].