Diameter and Laplace eigenvalue estimates for compact homogeneous Riemannian manifolds (Q6185776): Difference between revisions

Let \(G/K\) be a compact connected homogeneous space, where \(G\) is a compact connected Lie group and \(K\) is a closed subgroup of \(G\). For a given \(G\)-invariant Riemannian metric \(g\) on \(G/K\), let \(\operatorname{diam}(G/K, g)\) be the diameter of \(G/K\) with respect to the Riemannian distance induced by \(g\), and let \(\lambda_1(G/K, g)\) be the first eigenvalue (spectral gap) of the Laplace--Beltrami operator \(\Delta_g\). Scaling considerations suggest the rough relationship \(\lambda_1(G/K, g) \approx \operatorname{diam}(G/K, g)^{-2}\), and indeed, a celebrated result of Peter Li gives the universal lower bound \(\lambda_1(G/K, g) \operatorname{diam}(G/K, g)^2 \ge \pi^2/4\) [\textit{P. Li}, Comment. Math. Helv. 55, 347--363 (1980; Zbl 0451.53036)]. There is no corresponding universal upper bound, as can be seen by considering round spheres of increasing dimensions. However, for any fixed \(G/K\), the paper under review studies the question of whether there is a uniform upper bound over all \(G\)-invariant Riemannian metrics \(g\) on \(G/K\). That is, for a fixed \(G/K\), whether there exists a constant \(C = C(G,K)\) such that \[ \lambda_1(G/K, g) \operatorname{diam}(G/K, g)^2 \le C, \quad \text{for all \(G\)-invariant metrics \(g\).} \tag{*} \] The author begins by showing (Theorem 3) that if the uniform upper bound (*) holds for a compact Lie group \(G\) itself (i.e. (*) with \(K\) trivial), then it holds for all its homogeneous spaces \(G/K\), with the same constant \(C\). The reviewer and coauthors have conjectured, as part of a larger conjecture about uniform volume doubling, that the former actually does hold for all compact connected Lie groups \(G\); but so far this has only been proved for \(G = \mathrm{SU}(2)\) (as well as tori \(G = (S^1)^n\) for which it is trivial). See [\textit{N.~Eldredge, M.~Gordina, and L.~Saloff-Coste}, Geom. Funct. Anal. 28, No. 5, 1321--1367 (2018; Zbl 1407.53034)]. To state the main technical result of the paper (Theorem 29), we introduce some notation. Suppose that \(G\) is semisimple, and fix an \(\operatorname{Ad}(G)\)-invariant inner product \(\langle \cdot, \cdot \rangle_0\) on the Lie algebra \(\mathfrak{g}\). Let \(\mathfrak{k} \subset \mathfrak{g}\) be the Lie algebra of the subgroup \(K \le G\), and let \(\mathfrak{p}\) be its orthogonal complement under \(\langle \cdot, \cdot \rangle_0\), which is \(\operatorname{Ad}(K)\)-invariant. Then a \(G\)-invariant Riemannian metric \(g\) on \(G/K\) can be identified with an \(\operatorname{Ad}(K)\)-invariant inner product \(\langle \cdot, \cdot\rangle_g\) on \(\mathfrak{p}\). We may also write \(\langle X,Y \rangle_g = \langle \Phi_g(X), Y \rangle_0\) for a symmetric and positive definite linear endomorphism \(\Phi_g\) of \(\mathfrak{p}\) that commutes with \(\operatorname{Ad}(a)\) for all \(a \in K\). Now fix a decomposition of \(\mathfrak{p}\) as a direct sum of \(\operatorname{Ad}(K)\)-invariant subspaces \(\mathfrak{p} = \mathfrak{p}_1 \oplus \dots \oplus \mathfrak{p}_q\), and consider the class of \(G\)-invariant Riemannian metrics \(g\) on \(G/K\) for which the corresponding \(\Phi_g\) is of the form \(\Phi_g = \bigoplus_{i=1}^q c_i \operatorname{Id}_{\mathfrak{p}_i}\); or in other words, for which the corresponding inner product \(\langle \cdot, \cdot \rangle_g\) makes the \(\mathfrak{p}_i\) orthogonal and is a scalar multiple of \(\langle \cdot, \cdot\rangle_0\) on each \(\mathfrak{p}_i\). The main technical result of the paper (Theorem 29) is that there is a uniform upper bound on \(\lambda_1(G/K, g) \operatorname{diam}(G/K, g)^2\) over this class of metrics \(g\). (The reader should be advised that the statement of Theorem 29 in the paper is slightly incorrect, as it is missing the hypothesis that \(\Phi = \bigoplus_{i=1}^q c_i \operatorname{Id}_{\mathfrak{p}_i}\), and instead writes the weaker hypothesis that \(\Phi(\mathfrak{p}_i) \subset \mathfrak{p}_i\). It is in fact the stronger hypothesis that is needed for the proof. This stronger hypothesis is actually satisfied in every case where Theorem 29 is used within the paper, so the correctness of the paper's other results are unaffected. The author has confirmed this correction in personal communication with the reviewer.) The proof of Theorem 29 is based, roughly, on showing that the specified class of \(G\)-invariant metrics have their diameters bounded above by that of a certain sub-Riemannian metric on \(G/K\), and their spectral gaps bounded above by that of a certain singular Riemannian metric. As consequences of Theorem 29 the author shows: \begin{itemize} \item (Theorem 4) Suppose that \(G\) is semisimple and that the isotropy representation of \(G/K\) is multiplicity free; that is, that it decomposes as a direct sum of \emph{inequivalent} irreducible representations. (See references on page 1631 of the paper for examples, including in particular those where the isotropy representation is itself irreducible.) Then we have the uniform upper bound (*), over \emph{all} \(G\)-invariant Riemannian metrics on \(G/K\). Indeed, Schur's lemma implies that, fixing such a decomposition, the hypothesis of Theorem 29 is satisfied for every \(g\). \item (Theorem 5) The uniform upper bound (*) holds whenever \(G/K\) is a compact rank-one symmetric space. The proof is based on the classification of such spaces and their isotropy representations [see \textit{W. Ziller}, Math. Ann. 259, 351--358 (1982; Zbl 0469.53043)]. Most of them fall into the setting of Theorem 4, and for the others, there are results that their \(G\)-invariant metrics, up to isometry, form a class to which Theorem 29 can be applied. \end{itemize}