Busemann functions in a harmonic manifold (Q1419412): Difference between revisions

A Riemannian manifold \((M^d,g)\) is said to be harmonic if for any \(m\in M\) the volume density function in polar coordinates \(\Theta_m=r^{d-1}(\det g_{ij})^{1/2}\) of a geodesic sphere \(S_r(m)\) with centre \(m\) and radius \(r\) is a radial function. In this case \(\Theta_m(r)=\Theta_n(r)\) for all \(m,n\in M\). Consequently, the function \(\Theta\) is an important invariant of harmonic manifolds. Moreover, harmonic manifolds are Einstein manifolds. In 1944 \textit{A. Lichnerowicz} [Bull. Soc. Math. Fr. 72, 146--168 (1944; Zbl 0060.38506)] proved that harmonic manifolds of dimension \(d\leq 4\) are either flat or rank one symmetric spaces. This led Lichnerowicz to conjecture that any complete simply connected harmonic manifold is either flat or a rank one symmetric space. This turns out to be false in general. Let \(\kappa\) denote the scalar curvature of \((M,g)\). Then \((M,g)\) is flat if it is harmonic with \(\kappa=0\), by the Cheeger-Gromoll splitting theorem, while if \((M,g)\) is harmonic with \(\kappa>0\), then \(M\) is a compact rank one symmetric space by a result of \textit{Z. I. Szabó} [J. Differ. Geom. 31, 1--28 (1990; Zbl 0686.53042)]. On the other hand, \textit{E. Damek} and \textit{F. Ricci} [Bull. Am. Math. Soc., New Ser. 27, No. 1, 27, 139--142 (1992; Zbl 0755.53032)] have constructed examples of noncompact harmonic manifolds with \(\kappa<0\) that are not symmetric spaces. For compact negatively curved harmonic manifolds the conjecture was proved by \textit{G. Besson}, \textit{G. Courtois} and \textit{S. Gallot} [Geom. Funct. Anal. 5, No. 5, 731--799 (1995; Zbl 0851.53032]. Here the authors continue their previous study [\textit{R. Ranjan} and \textit{H. Shah}, J. Geom. Anal. 12, 683--694 (2002; Zbl 1066.53086] of complete simply connected harmonic manifolds. Any exponential map of such a manifold is a diffeomorphism between the appropriate tangent space and the manifold. Let \(\gamma_v\) denote the geodesic tangent to the unit vector \(v\). The Busemann function corresponding to \(\gamma_v\) is defined by \[ b_v^-(x) = \lim_{t\rightarrow -\infty} d(x,\gamma_v(t))+t. \] The authors first show that \(K_\infty:=\lim_{r\rightarrow\infty} \Theta'(r)/\Theta(r)\) exists and is nonnegative. The main result of the paper is the analyticity of the Busemann functions. This is proved by showing that \(\Delta b_v=K_0\). Since \(b_v\) is analytic, the level sets of \(b_v\), called horospheres, are analytic hypersurfaces of constant mean curvature \(K_0\), and one may try to classify harmonic manifolds according to the mean curvature of their horospheres. The authors discuss the bi-asymptotic property of harmonic spaces with \(K_\infty=0\). They conjecture that this condition should force \(M\) to be flat. Their results provide some evidence for this.