Martin boundary for homogeneous Riemannian manifolds of neagtive curvature at the bottom of the spectrum (Q5960828)

From MaRDI portal
scientific article; zbMATH DE number 1730500
Language Label Description Also known as
English
Martin boundary for homogeneous Riemannian manifolds of neagtive curvature at the bottom of the spectrum
scientific article; zbMATH DE number 1730500

    Statements

    Martin boundary for homogeneous Riemannian manifolds of neagtive curvature at the bottom of the spectrum (English)
    0 references
    0 references
    0 references
    0 references
    0 references
    5 January 2003
    0 references
    The authors determine the Martin boundary for elliptic operators given as the sum of squares of vector fields on a simply connected homogeneous Riemannian manifold \(M\) with negative curvature. Since such a manifold \(M\) is isometric to a solvable Lie group \(S\) which is a semi-direct product \(N\rtimes A\) of a nilpotent Lie group \(N\) and \(A=\mathbb R^+\), the analysis can be simplified to \(S\), resp. \(N\) and \(A\). Consider on \(S\) the sum of squares of left-invariant vector fields \[ \mathcal L = \mathcal L_\gamma = \sum_{j=0}^m Y_j + Y \] where \(\{Y_0,Y_1,\ldots Y_m\}\) generate the Lie algebra \(\mathfrak s\). The \(\gamma\in\mathbb R \) signals the \(\mathfrak a\)-component of the vector field \(Y\). The main aim of the paper is to determine the Martin boundary for \(\mathcal L\) both in the coercive and the non-coercive case. The first case was already solved by \textit{E. Damek} [Ann. Fac. Sci. Toulouse (6) 5, 421--441 (1996; Zbl 0876.22008)] employing \textit{A. Ancona}'s method [Ann. Math. (2) 125, 495--535 (1987; Zbl 0652.31008)]. The non-coercive case and \(\gamma=0\) require a different technique and the present paper uses the probabilistic methods developed by two of the authors in [Studia Math. 103, 239--264 (1992; Zbl 0824.43002)] and extended in [Studia Math. 126, 114--148 (1997; Zbl 0888.22007)]. Roughly speaking, the idea is to split the \(\mathcal L_\gamma\)-diffusion process into a horizontal ``Bessel'' process and a vertical component (which is determined by an explicit evolution equation). This allows to obtain two-sided estimates for the Poisson kernel \(m_\gamma\) of \(\mathcal L_\gamma\) (even in the case where \(\gamma = 0\)) of the form \(m_\gamma(x)\sim (|x|+1)^{-Q-\gamma}\). From this it is then possible to determine the Martin boundary of \(\mathcal L_\gamma\) and, in particular, of \(\mathcal L_0\).
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    sum of squares of vector fields
    0 references
    nilpotent Lie group
    0 references
    homogeneous Riemannian manifold
    0 references
    negative curvature
    0 references
    Nash inequality
    0 references
    Green function
    0 references
    Poisson kernel
    0 references
    Martin boundary
    0 references
    Bessel process
    0 references
    diffusion
    0 references
    0 references