Abstract
In 2012, M. M. Graev associated to a compact homogeneous space G/H a nerve \({\text {X}}_{G/H}\), whose non-contractibility implies the existence of a G-invariant Einstein metric on G/H. The nerve \({\text {X}}_{G/H}\) is a compact, semi-algebraic set, defined purely Lie theoretically by intermediate subgroups. In this paper we present a detailed description of the work of Graev and the curvature estimates given by Böhm in 2004.
Similar content being viewed by others
Data availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during this work.
References
Alekseevsky, D.V., Kimel’fel’d, B.N.: Structure of homogeneous Riemann spaces with zero Ricci curvature. Funct. Anal. Appl. 9, 97–102 (1975)
Alekseevsky, D.V., Dotti, I., Ferraris, C.: Homogeneous Ricci positive \(5\)-manifolds. Pacific J. Math. 175, 1–12 (1996)
Arvanitoyeorgos, A.: New invariant Einstein metrics on generalized flag manifolds. Trans. Amer. Math. Soc. 337, 981–995 (1993)
Aubin, T.: Équations du type Monge-Ampère sur les variétés kähleriennes compactes. Bull. Sci. Math. (2) 102(1), 63–95 (1978)
Belgun, F., Cortés, V., Haupt, A.S., Lindemann, D.: Left-invariant Einstein metrics on \(S^3\times S^3\). J. Geom. Phys. 128, 128–139 (2018)
Ballmann, W., Thorbergsson, G., Ziller, W.: Existence of closed geodesics on positively curved manifolds. J. Differential Geom. 18, 221–252 (1983)
Benedetti, R., Risler, J.J.: Real Algebraic and Semi-algebraic Sets. Actualités Mathématiques, Hermann, Paris (1990)
Berger, M.: Quelques formules de variation pour une structure Riemannienne. Ann. Sci. Ec. Norm. Super. 3, \(4^e\) serie, 285–294 (1970)
Besse, A.L.: Einstein Manifolds. Springer, Berlin (1987)
Böhm, C.: Inhomogeneous Einstein metrics on low-dimensional spheres and other low-dimensional spaces. Invent. Math. 134, 145–176 (1998)
Böhm, C.: Homogeneous Einstein metrics and simplicial complexes. J. Differential Geom. 67, 79–165 (2004)
Böhm, C.: Non-existence of homogeneous Einstein metrics. Comment. Math. Helv. 80, 123–146 (2005)
Böhm, C.: On the long time behavior of homogeneous Ricci flows. Comment. Math. Helv. 90, 543–571 (2015)
Böhm, C., Kerr, M.M.: Low-dimensional homogeneous Einstein manifolds. Trans. Amer. Math. Soc. 358, 1455–1468 (2006)
Böhm, C., Lafuente, R.: Homogeneous Einstein metrics on Euclidean spaces are Einstein solvmanifolds. Geom. Topol. 26, 899–936 (2022)
Böhm, C., Lafuente, R.: Non-compact Einstein manifolds with symmetry. JAMS 36, 591–651 (2023)
Böhm, C., Wang, M.Y., Ziller, W.: A variational approach for compact homogeneous Einstein manifolds. GAFA 14, 681–733 (2004)
Boyer, C.P., Galicki, K.: On Sasakian–Einstein geometry. Internat. J. Math. 11, 873–909 (2000)
Boyer, C.P., Galicki, K., Kollár, J.: Einstein metrics on spheres. Ann. Math. 557–580 (2005)
Bredon, G.E.: Introduction to Compact Transformation Groups, Pure and Applied Mathematics, vol. 46. Academic Press, New York (1972)
Bröcker, T., tom Dieck, T.: Graduate Texts in Mathematics. Representations of Compact Lie Groups, vol. 98. Springer, Berlin (1995)
Cheeger, J., Colding, T.H.: Lower bounds on Ricci curvature and the almost rigidity of warped products. Ann. Math. 144, 189–237 (1996)
Cheeger, J., Colding, T.H.: On the structure of spaces with Ricci curvature bounded below I. J. Differential Geom. 46, 406–480 (1997)
Chen, X., Donaldson, S., Sun, S.: Kähler–Einstein metrics on Fano manifolds. I: approximation of metrics with cone singularities. J. Amer. Math. Soc. 28, 183–197 (2015)
Chen, X., Donaldson, S., Sun, S.: Kähler–Einstein metrics on Fano manifolds. II: Limits with cone angle less than \(2\pi \). J. Amer. Math. Soc. 28, 199–234 (2015)
Chen, X., Donaldson, S., Sun, S.: Kähler-Einstein metrics on Fano manifolds. III: Limits as cone angle approaches \(2\pi \) and completion of the main proof. J. Amer. Math. Soc. 28, 235–278 (2015)
Chrysikos, I., Sakane, Y.: Non-naturally reductive Einstein metrics on exceptional Lie groups. J. Geom. Phys. 116, 152–186 (2017)
D’Atri, J., Ziller, W.: Naturally reductive metrics and Einstein metrics on compact Lie groups, Memoirs of the AMS, 215 (1979)
Delfs, H., Knebusch, M.: Semialgebraic topology over a real closed field II: Basic theory of semialgebraic spaces. Math. Z. 178, 175–213 (1981)
Delfs, H., Knebusch, M.: Separation, Retractions and Homotopy extension in semi-algebraic spaces. Pacific J. Math. 114, 47–71 (1984)
Delfs, H., Knebusch, M.: Locally Semialgebraic Spaces. Lecture Notes in Math, vol. 1173. Springer, Berlin (1985)
Dickinson, W., Kerr, M.M.: The geometry of compact homogeneous spaces with two isotropy summands. Ann. Global Anal. Geom. 34, 329–350 (2008)
Dynkin, E.B.: Maximal subgroups of the classical groups. Transl. Amer. Math. Soc. Series 2 6, 245–378 (1957)
Eschenburg, J.-H.: Lecture notes on symmetric spaces, preprint (1997)
Foscolo, L., Haskins, M.: New \(G_2\)-holonomy cones and exotic nearly Kähler structures on \(S^6\) and \(S^3\times S^3\). Ann. Math. 2, 59–130 (2017)
Fulton, W., Harris, J.: Representation Theory: A First Course Graduate Texts in Mathematics, vol. 129. Springer, Berlin (1991)
Ghoussoub, N.: Location, multiplicity and Morse indices of min-max critical points. J. Reine Angew. Math. 417, 27–76 (1991)
Graev, M.M.: On the number of invariant Einstein metrics on a compact homogeneous space, Newton polytopes and contractions of Lie algebras. Int. J. Geom. Methods Mod. Phys. 3(5–6), 1047–1075 (2006)
Graev, M.M.: The existence of invariant Einstein metrics on a compact homogeneous space. Trans. Moscow Math. Soc. 73, 1–28 (2012)
Graev, M.M.: On the compactness of the set of invariant Einstein metrics. Ann. Global Anal. Geom. 44(4), 471–500 (2013)
Hatcher, A.: Algebraic Topology. Cambridge University Press, Cambridge (2002)
He, C.: Cohomogeneity one manifolds with a small family of invariant metrics. Geom. Dedicata 157, 41–90 (2012)
Heber, J.: Noncompact homogeneous Einstein spaces. Invent. Math. 133, 279–352 (1998)
Helgason, S.: Differential Geometry, Lie Groups, and Symmetric Spaces, 2nd edn. Academic Press, New York (1978)
Hilbert, D.: Die Grundlagen der Physik, Nachr. Akad. Wiss. Gött., pp. 395–407 (1915)
Hironaka, H.: Triangulations of algebraic sets. Proc. Sympos. Pure Math. 29, 165–185 (1975)
Jensen, G.: Homogeneous Einstein spaces of dimension 4. J. Differential Geom. 3, 309–349 (1969)
Joyce, D.D.: Compact \(8\)-manifolds with holonomy \({\rm Spin}(7)\). Invent. Math. 123, 507–552 (1996)
Joyce, D.D.: Compact Riemannian \(7\)-manifolds with holonomy \(G_2\). I, II. J. Differ. Geom. 43, 291–328, 329–375 (1996)
Joyce, D.: Compact Manifolds with Special Holonomy. Oxford Mathematical Monographs. Oxford University Press, Oxford (2000)
Kerr, M.M.: Some new homogeneous Einstein metrics on symmetric spaces. Trans. Amer. Math. Soc. 348, 153–171 (1996)
Klaus, S.: Einfach-zusammenhängende Kompakte Homogene Räume bis zur Dimension Neun. Diplomarbeit, Johannes Gutenberg Universität (1988)
Kreck, M., Stolz, S.: A diffeomorphism classification of 7-dimensional homogeneous Einstein manifolds with \(SU(3) \times SU(2) \times U(1)\)-symmetry. Ann. Math. (2) 127(2), 373–388 (1988)
Lauret, J.: Ricci soliton homogeneous nilmanifolds. Math. Ann. 319, 715–733 (2001)
Lauret, J.: Einstein solvmanifolds are standard. Ann. Math. (2) 172(3), 1859–1877 (2010)
Lauret, J.: On the stability of homogeneous Einstein manifolds, preprint arXiv:2105.06336 (2021) Asien J. Math. (to appear)
Lauret, E.A., Lauret, J.: The stability of standard homogeneous Einstein manifolds. Math. Z. 303(1), 36 (2023)
Lauret, J., Will, C.: On the stability of homogeneous Einstein manifolds, II. J. Lond. Math. Soc. 106, 3638–3669 (2022)
Lauret, J., Will, C.: Homogeneous Einstein metrics and local maxima of the Hilbert action. J. Geom. Phys. 178, 8 (2022)
Maunder, C.R.F.: Algebraic Topology. Cambridge University Press, Cambridge (1980)
Milnor, J.: Construction of Universal Bundles, II. Ann. Math. 63, 430–436 (1956)
Nienhaus, J., Wink, M.: Einstein metrics on the Ten-sphere (2023). arXiv:2303.04832
Nikonorov, Yu.G., Rodionov, E.D.: Compact \(6\)-dimensional homogeneous Einstein manifolds. Dokl. Math. RAS 336, 599–601 (1999)
Nikonorov, Yu.G.: Compact homogeneous Einstein \(7\)-manifolds. Geom. Dedicata 109, 7–30 (2004)
Nikonorov, Yu.G., Classification of generalized Wallach spaces. Geom. Dedicata 181, 193–212 (2016), Correction Geom. Dedicata 214, 849–851 (2021)
Nikonorov, Yu.G.: On left-invariant Einstein Riemannian metrics that are not geodesic orbit. Transform. Groups 24(2), 511–530 (2019)
Park, J.-S., Sakane, Y.: Invariant Einstein metrics on certain homogeneous spaces. Tokyo J. Math. 20, 51–61 (1997)
Quillen, D.: Homotopy Properties of the Poset of Nontrivial p-Subgroups of a Group. Adv. Math. 28, 101–128 (1978)
Rauße, C.: Simplicial complexes of compact homogeneous spaces. Ph.D.-thesis Münster, (2016)
Schwahn, P., Semmelmann, U., Weingart, G.: Stability of the Non-Symmetric Space \(E_7/{\rm PSO}(8)\) (2022). arXiv:2203.10138
Spiro, A.: A remark on locally homogeneous Riemannian spaces. Results Math. 24, 318–325 (1993)
Tian, G.: K-stability and Kähler-Einstein metrics. Commun. Pure Appl. Math. 68, 1085–1156 (2015). Corrigendum 68, 2082–2083 (2015)
Wang, M.Y.: Some examples of homogeneous Einstein manifolds in dimension seven. Duke Math. J. 49, 23–28 (1982)
Wang, M.Y.: Einstein metrics from symmetry and bundle construction in Surveys in differential geometry VI: essays on Einstein manifolds. C. LeBrun, M. Y. Wang (eds.), pp. 287–325. International Press, Boston (1999)
Wang, M.Y.: Einstein metrics from symmetry and bundle constructions: a sequel. Adv. Lect. Math. 22, 253–309 (2012)
Wang, M.Y., Ziller, W.: On normal homogeneous Einstein manifolds. Ann. scient. Éc. Norm. Sup. \(4^e\) série t. 18, 563–633 (1985)
Wang, M.Y., Ziller, W.: Existence and non-existence of homogeneous Einstein metrics. Invent. Math. 84, 177–194 (1986)
Wang, M.Y., Ziller, W.: Einstein metrics on principal torus bundles. J. Differential Geom. 31, 215–248 (1990)
Wang, M.Y., Ziller, W.: On isotropy irreducible Riemannian manifolds. Acta Math. 199, 223–261 (1991)
Wolf, J.A.: The geometry and structure of isotropy irreducible homogeneous spaces. Acta. Math. 120, 59–148 (1968); correction. Acta. Math. 152, 141–142 (1984)
Yau, S.T.: On the Ricci curvature of a compact Kähler manifold and the complex Monge-Ampère equation I. Commun. Pure Appl. Math. 31, 339–411 (1978)
Ziller, W.: Homogeneous Einstein metrics on spheres and projective spaces. Math. Ann. 259, 351–358 (1982)
Acknowledgements
Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy EXC 2044—390685587, Mathematics Münster: Dynamics–Geometry–Structure, and the Collaborative Research Centre CRC 1442, Geometry: Deformations and Rigidity.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no financial or proprietary interests in any material discussed in this article.
Additional information
This paper is dedicated to the memory of M. M. Graev.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In this section we will make a brief digression, first into semi-algebraic geometry. We also provide helpful background on the isotropy representation of a compact homogeneous space G/H.
1.1 Semi-algebraic sets
We refer the reader to Benedetti and Risler for more details [7]. A set \(X\subset {{{\mathbb {R}}}}^m\) is called semi-algebraic if X is defined by finitely many polynomial equations and inequalities. The inequalities are allowed to be both strict and non-strict.
Given two semi-algebraic sets \(X\subset {{{\mathbb {R}}}}^m\) and \(Y\subset {{{\mathbb {R}}}}^n\), we say a map \(f:X\rightarrow Y\) is semi-algebraic if the graph of f is a semi-algebraic subset of \({{{\mathbb {R}}}}^{m+n}\): see [7, Definition 2.3.2]. A semi-algebraic map is not necessarily continuous, see [7, Example 2.7.3], though some authors include a requirement of continuity in their definition of a semi-algebraic map: see [46, 29, Theorem 7.6], [31, I, Proposition 3.13]. The paradigm for semi-algebraic maps are polynomials.
A fundamental result in semi-algebraic geometry is the following:
Theorem 7.1
(Tarski-Seidenberg) The image of a semi-algebraic subset of \({{{\mathbb {R}}}}^n\times {{{\mathbb {R}}}}^m\) under the projection \(\pi :{{{\mathbb {R}}}}^n \times {{{\mathbb {R}}}}^m \rightarrow {{{\mathbb {R}}}}^n\) such that \((x,y)\mapsto x\) is a semi-algebraic set.
The theorem implies that intersections, unions, and complements of semi-algebraic sets are semi-algebraic. Moreover, it is possible to use first-order formulae (in the language of ordered fields with parameters in \({{{\mathbb {R}}}}\)) to obtain new semi-algebraic sets from known ones. For instance, given \(X_1\subset {{{\mathbb {R}}}}^n,X_2 \subset {{{\mathbb {R}}}}^m\) semi-algebraic sets and a semi-algebraic map \(F:{{{\mathbb {R}}}}^n\times {{{\mathbb {R}}}}^m \rightarrow {{{\mathbb {R}}}}^k\), the set
is semi-algebraic. As another example, given endomorphisms A, B of \({{{\mathbb {R}}}}^n\), the condition \(\ker (A) \subsetneq \ker (B)\) is a semi-algebraic condition because it is equivalent to
Connected components of semi-algebraic sets are semi-algebraic, and the closure, interior, or boundary of a semi-algebraic set is semi-algebraic.
The key property of compact semi-algebraic sets used in Sects. 3.1 and 3.2 is that such sets are absolute neighborhood retracts. The following result states a useful generalization of that.
Lemma 7.2
[39] Let \(X,Y \subset {{{\mathbb {R}}}}^N\) be compact, semi-algebraic sets with \(X \subset Y\) and \(X \ne \emptyset \). Suppose that for some \(\delta _0>0\), for all \(\delta \in (0,\delta _0)\) there exists a continuous map
with the following properties: \(H_\delta (0,y)=y\) for all \(y \in Y\), \(H_\delta (t,x)=x\) for all \(x \in X\) and all \(t \in [0,1]\), and
Then X is a strong deformation retract of Y.
Proof
It follows from Theorem 1 of [30], see also [31, III, Theorem 1.1], that there exists an open semi-algebraic neighborhood U of X in Y and a semi-algebraic, continuous map \(G:[0,1]\times {\overline{U}}\rightarrow {\overline{U}}\), such that the restriction \(G\vert _{[0,1]\times U}\) yields a strong deformation retraction from U to X.
By compactness of X, there exists \(\delta _0>0\) such that for all \(0<\delta \le \delta _0\), we have
Hence we can define
The function H is continuous since for \(t=\tfrac{1}{2}\) we have \(G(2t-1,H_\delta (1,y))=H_{\delta }(2t,y)\). Moreover \(H(t,x)=x\) for all \(t \in [0,1]\) and all \(x \in X\) and \(H(1,Y)=X\). This shows the claim. \(\square \)
1.2 Lie-theory I
In this section we provide details concerning the isotropy representation of a compact homogeneous space G/H and the normalizer \(N_G(H)\) of H in G. We assume, as we have throughout the paper, that G and H are connected.
Definition 7.3
For any compact homogeneous space G/H, we denote by \({{{\mathfrak {m}}}}_0\) the subspace of \({{{\mathfrak {m}}}}\) on which \({\text {Ad}}(H)\) acts trivially.
Notice that if \(\text {rk} \,G=\text {rk} \,H\), then \({{{\mathfrak {m}}}}_0 = \{0\}\), while if \(H=\{e\}\), \({{{\mathfrak {m}}}}_0={{{\mathfrak {g}}}}\). Recall also that
for all \(X \in {{{\mathfrak {g}}}}\), where \(\exp _G:{{{\mathfrak {g}}}}\rightarrow G\) denotes the exponential of G and \(\exp (D)=\sum _{k=0}^\infty \tfrac{D^k}{k!}\) for any endomorphism \(D:{{{\mathfrak {g}}}}\rightarrow {{{\mathfrak {g}}}}\) [44, II].
Lemma 7.4
A subspace \(\tilde{{{\mathfrak {m}}}}\) of \({{{\mathfrak {m}}}}\) is \({\text {Ad}}(H)\)-invariant if and only if \([{{{\mathfrak {h}}}},\tilde{{{{\mathfrak {m}}}}}]\subset \tilde{{{{\mathfrak {m}}}}}\). Moreover, \([{{{\mathfrak {h}}}},{{{\mathfrak {m}}}}_0]=0\) and for any \({\text {Ad}}(H)\)-invariant subspace \({\tilde{{{{\mathfrak {m}}}}}}\) of \({{{\mathfrak {m}}}}\ominus {{{\mathfrak {m}}}}_0\), we have \(\{0\} \ne [{{{\mathfrak {h}}}},{\tilde{{{{\mathfrak {m}}}}}}]\).
Proof
If \(\tilde{{{{\mathfrak {m}}}}}\) is \({{\text {Ad}}}(H)\)-invariant, then differentiation yields \([{{{\mathfrak {h}}}},\tilde{{{{\mathfrak {m}}}}}]\subset \tilde{{{{\mathfrak {m}}}}}\). Conversely, suppose we know \([{{{\mathfrak {h}}}},\tilde{{{{\mathfrak {m}}}}}]\subset \tilde{{{{\mathfrak {m}}}}}\). Since H is connected, for any \(Z \in {{{\mathfrak {h}}}}\), \({{\text {Ad}}}(\exp (Z))=\exp ({{\text {ad}}}(Z))\). Hence, \(\tilde{{{{\mathfrak {m}}}}}\) is \({{\text {Ad}}}(H)\)-invariant. This shows the first claim. The second and the third claim follow immediately. \(\square \)
Let \({{{\mathfrak {n}}}}({{{\mathfrak {h}}}})\) denote the Lie algebra of the normalizer \(N_G(H)\) of H in G.
Lemma 7.5
Let G/H be a compact homogeneous space. Then \({{{\mathfrak {n}}}}({{{\mathfrak {h}}}})= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0\). Moreover, \({{{\mathfrak {m}}}}_0\) is a compact subalgebra.
Proof
Since \([{{{\mathfrak {h}}}},{{{\mathfrak {m}}}}_0]=0 \in {{{\mathfrak {h}}}}\), see above, we conclude \({{{\mathfrak {m}}}}_0 \subset {{{\mathfrak {n}}}}({{{\mathfrak {h}}}})\). Conversely, suppose that \(\tilde{{{\mathfrak {m}}}}\subset {{{\mathfrak {n}}}}({{{\mathfrak {h}}}}) \cap {{{\mathfrak {m}}}}\). Then on the one hand, \([{{{\mathfrak {h}}}}, \tilde{{{\mathfrak {m}}}}]\subset {{{\mathfrak {h}}}}\) by definition of \({{{\mathfrak {n}}}}({{{\mathfrak {h}}}})\), while on the other hand, \([{{{\mathfrak {h}}}}, \tilde{{{\mathfrak {m}}}}] \subset {{{\mathfrak {m}}}}\) because \([{{{\mathfrak {h}}}},{{{\mathfrak {m}}}}]\subset {{{\mathfrak {m}}}}\). This shows \([{{{\mathfrak {h}}}}, \tilde{{{\mathfrak {m}}}}]=0\), and thus \({{{\mathfrak {m}}}}\cap {{{\mathfrak {n}}}}({{{\mathfrak {h}}}})\subset {{{\mathfrak {m}}}}_0\), proving our equality.
Since \([{{{\mathfrak {m}}}}_0,{{{\mathfrak {m}}}}_0] \subset {{{\mathfrak {n}}}}({{{\mathfrak {h}}}})={{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0\) and \([{{{\mathfrak {h}}}},{{{\mathfrak {m}}}}_0]=0\), then because Q is self-adjoint (4.3), we know that \({{{\mathfrak {m}}}}_0\) is a subalgebra of \({{{\mathfrak {n}}}}({{{\mathfrak {h}}}})\) and thus is a subalgebra of \({{{\mathfrak {g}}}}\). In order to show that the corresponding connected subgroup of G is compact we now use the special property of the biinvariant metric Q chosen in Sect. 2.1. We decompose the compact Lie algebra \({{{\mathfrak {n}}}}({{{\mathfrak {h}}}})={{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0\) into the Q-orthogonal sum of its semisimple part \(({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)_s= {{{\mathfrak {h}}}}_s \oplus ({{{\mathfrak {m}}}}_0)_s\) and its center \({{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\). Each of the Lie algebras \({{{\mathfrak {h}}}}_s\), \(({{{\mathfrak {m}}}}_0)_s\) and \({{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) is compact, and so \({{{\mathfrak {h}}}}\cap {{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) is also compact. It remains to show that the Q-orthogonal complement of \({{{\mathfrak {h}}}}\cap {{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) in \({{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) is compact.
By the definition of Q, the compact, abelian subalgebras \({{{\mathfrak {h}}}}\cap {{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) and \({{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) can be viewed as subalgebras of \({{{\mathfrak {s}} {\mathfrak {o}}}}(6N)\). Notice we can choose a Q-orthonormal (standard) basis \(({\hat{e}}_1,\ldots ,{\hat{e}}_{3N})\) for the standard maximal (diagonal) torus \({{{\mathfrak {t}}}}_{6N}\) of \({{{\mathfrak {s}} {\mathfrak {o}}}}(6N)\) so that every compact subtorus \({{{\mathfrak {t}}}}\) of \({{{\mathfrak {t}}}}_{6N}\) has a basis \((e_1,\ldots ,e_{\dim {{{{\mathfrak {t}}}}}})\), for which each \(e_i\) is a rational linear combination of the above basis elements \(({\hat{e}}_i)\) of \({{{\mathfrak {s}} {\mathfrak {o}}}}(6N)\). Then, the Q-orthogonal complement in \({{{\mathfrak {t}}}}_{6N}\) of the a subtorus \({{{\mathfrak {t}}}}\) must also have such a basis, and hence it corresponds to a compact subtorus in \({{{\mathfrak {s}} {\mathfrak {o}}}}(6N)\). As a consequence, the Q-orthogonal complement of \({{{\mathfrak {h}}}}\cap {{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) in \({{{\mathfrak {z}}}}({{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_0)\) is the intersection of two compact subalgebras, thus a compact subalgebra. \(\square \)
Remark 7.6
We decompose \({{{\mathfrak {m}}}}\) into its \({\text {Ad}}(H)\)-invariant isotypical summands (see [21, II, Proposition 6.9]):
Each \({{{\mathfrak {p}}}}_i\) is a direct sum of \({\text {Ad}}(H)\)-irreducible summands which are equivalent (as \({\text {Ad}}(H)\)-representations), while, for \(i \ne j\), irreducible summands in \({{{\mathfrak {p}}}}_i\) and \({{{\mathfrak {p}}}}_j\) are inequivalent. Note by Lemma 7.4, if \({{{\mathfrak {m}}}}_0\) is non-trivial, then \({{{\mathfrak {m}}}}_0\) is one of the isotypical summands of \({{{\mathfrak {m}}}}\).
Thus by Schur’s Lemma, for every \(P_g \in \mathcal {M}^G\) we have
where, for each \(1 \le i \le \ell _{\text {iso} }\), \((P_g)_i:{{{\mathfrak {p}}}}_i \rightarrow {{{\mathfrak {p}}}}_i\) is an \({\text {Ad}}(H)\)-equivariant, self-adjoint, positive definite endomorphism. Consequently, every \(P_g \in \mathcal {M}^G\) respects the decomposition \({{{\mathfrak {m}}}}=\oplus _{i=1}^{\ell _{\text {iso} }}\,{{{\mathfrak {p}}}}_i\). Notice this is also true for every \(v \in \text {Sym} _{{{{\mathfrak {m}}}}}^{H}\).
For \({{{\mathfrak {k}}}}\) a subalgebra in \(\mathcal {K}\), let \({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}}:={{{\mathfrak {m}}}}\cap {{{\mathfrak {k}}}}\). For convenience, \({{{\mathfrak {m}}}}_{{{{{\mathfrak {h}}}}}}:= {{{\mathfrak {m}}}}\).
Lemma 7.7
Let \({{{\mathfrak {k}}}}\) be a subalgebra, \({{{\mathfrak {h}}}}\le {{{\mathfrak {k}}}}< {{{\mathfrak {g}}}}\), let K be the connected subgroup of G corresponding to \({{{\mathfrak {k}}}}\), and let \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) with \({{{\mathfrak {k}}}}\subset \ker (A)\). Then, the condition \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\) is equivalent to the \({\text {Ad}}(K)\)-equivariance of A. Furthermore, \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\) if and only if \([A\vert _{{{{\mathfrak {k}}}}^\perp },{\text {ad}}({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}})\vert _{{{{\mathfrak {k}}}}^\perp }]=0\) if and only if \([A\vert _{{{{{\mathfrak {m}}}}}},{\text {ad}}({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}})\vert _{{{{{\mathfrak {m}}}}}}]=0\).
Proof
Let K be the connected subgroup of G with Lie algebra \({{{\mathfrak {k}}}}\). We know by differentiation that \({\text {Ad}}(K)\)-equivariance of A implies \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\). To see the converse, let \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) and suppose that \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\). Using (7.1) shows that A is \({\text {Ad}}(K)\)-equivariant.
Next, let \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) with \({{{\mathfrak {k}}}}\subset \ker (A)\). Since \({\text {Ad}}(K)\), \({\text {ad}}({{{\mathfrak {k}}}})\) respect the decomposition \({{{\mathfrak {g}}}}= {{{\mathfrak {k}}}}\oplus {{{\mathfrak {k}}}}^\perp \), it follows that \(A\vert _{{{{\mathfrak {k}}}}^\perp }\) is \({\text {Ad}}(K)\vert _{{{{\mathfrak {k}}}}^\perp }\)-equivariant if and only if \([A\vert _{{{{\mathfrak {k}}}}^\perp },{\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{\mathfrak {k}}}}^\perp }]=0\).
Since \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) with \({{{\mathfrak {k}}}}\subset \ker (A)\) is \({\text {Ad}}(K)\)-equivariant if and only if \(A\vert _{{{{\mathfrak {k}}}}^\perp }\) is \({\text {Ad}}(K)\vert _{{{{\mathfrak {k}}}}^\perp }\)-equivariant we deduce that \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\) is equivalent to \([A\vert _{{{{\mathfrak {k}}}}^\perp },{\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{\mathfrak {k}}}}^\perp }]=0\).
Since for \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) with \({{{\mathfrak {k}}}}\subset \ker (A)\) A preserves \({{{\mathfrak {m}}}}\) we deduce from \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\) that \([A\vert _{{{{{\mathfrak {m}}}}}},{\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{{\mathfrak {m}}}}}}]=0\), which implies \([A\vert _{{{{{\mathfrak {m}}}}}},{\text {ad}}({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}})\vert _{{{{{\mathfrak {m}}}}}}]=0\) since \({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}}\subset {{{\mathfrak {k}}}}\).
Suppose now \([A\vert _{{{{{\mathfrak {m}}}}}},{\text {ad}}({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}})\vert _{{{{{\mathfrak {m}}}}}}]=0\). All elements of \(\text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) are \({\text {Ad}}(H)\)-equivariant, by definition, thus \([A\vert _{{{{{\mathfrak {m}}}}}},{\text {ad}}({{{\mathfrak {h}}}})\vert _{{{{{\mathfrak {m}}}}}}]=0\). Using \({{{\mathfrak {k}}}}={{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}}\) we conclude \([A\vert _{{{{{\mathfrak {m}}}}}},{\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{{\mathfrak {m}}}}}}]=0\). Since \({\text {ad}}({{{\mathfrak {k}}}})\) preserves the decomposition \({{{\mathfrak {g}}}}={{{\mathfrak {k}}}}\oplus {{{\mathfrak {k}}}}^\perp \) it follows that \({\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{{\mathfrak {m}}}}}}\) preserves the decomposition \({{{\mathfrak {m}}}}={{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}} \oplus {{{\mathfrak {k}}}}^\perp \). We deduce \([A\vert _{{{{\mathfrak {k}}}}^\perp },{\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{\mathfrak {k}}}}^\perp }]=0\). Using that for \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) with \({{{\mathfrak {k}}}}\subset \ker (A)\), this is equivalent to \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\), proving the claim. \(\square \)
Any such A induces a submersion metric on G/H with respect to \(K/H \rightarrow G/H \rightarrow G/K\). Notice however, that K need not be a compact subgroup of G.
Corollary 7.8
Let \(A \in \text {Sym} _{{{{\mathfrak {g}}}}}^{H}\) (\(v \in {{\text {S}}}\)) and let \({{{\mathfrak {m}}}}_{I_1},\dots , {{{\mathfrak {m}}}}_{I_\ell } \subset {{{\mathfrak {m}}}}\) denote the eigenspaces of \(A\vert _{{{{\mathfrak {m}}}}}\) corresponding to distinct eigenvalues in increasing order.
-
(i)
Then \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_1}]\perp {{{\mathfrak {m}}}}_{I_p}\) (i.e. \([I_1 I_1 I_p]=0\)) for all \(1<p \le \ell \) if and only if \({{{\mathfrak {k}}}}= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{I_1}\) is a subalgebra.
-
(ii)
Furthermore, when \({{{\mathfrak {k}}}}= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{I_1}\) is a subalgebra, \(A \in {\text {D}}({{{\mathfrak {k}}}})\) if and only if \({{{\mathfrak {k}}}}\le \ker (A)\) and for all \(1 \le p \ne q \le \ell \), we have \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_p}]\perp {{{\mathfrak {m}}}}_{I_q}\) (i.e. \([I_1 I_p I_q]=0\)).
Proof
(i) If for all \(1 < p \) we have \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_1}]\perp {{{\mathfrak {m}}}}_{I_p}\), then since \([{{{\mathfrak {h}}}},{{{\mathfrak {m}}}}_{I_1}] \subset {{{\mathfrak {m}}}}_{I_1}\) we have that \({{{\mathfrak {k}}}}= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{I_1}\) is a subalgebra of \({{{\mathfrak {g}}}}\). Conversely, if \({{{\mathfrak {k}}}}= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{I_1}\in \mathcal {K}\) is a subalgebra, then we know \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_1}] \subset {{{\mathfrak {k}}}}\perp {{{\mathfrak {m}}}}_{I_p}\) for all \(1<p\).
(ii) Suppose \({{{\mathfrak {k}}}}= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{I_1}\in \mathcal {K}\) is a subalgebra, and \(A \in {\text {D}}({{{\mathfrak {k}}}})\) or equivalently, \(v \in {\text {D}}({{{\mathfrak {k}}}})^{{\text {S}}}\). Since \(A \in {\text {D}}({{{\mathfrak {k}}}})\) we know \([A,{\text {ad}}({{{\mathfrak {k}}}})]=0\). Thus, by Lemma 7.7 we deduce \([A\vert _{{{{\mathfrak {k}}}}^\perp },{\text {ad}}({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}})\vert _{{{{\mathfrak {k}}}}^\perp }]=0\). This shows that \({\text {ad}}({{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}})\) respects the eigenspaces of \(A\vert _{{{{{\mathfrak {m}}}}}}\). That is, \([{{{\mathfrak {m}}}}_{I_1}, {{{\mathfrak {m}}}}_{I_p}] \subset {{{\mathfrak {m}}}}_{I_p} \perp {{{\mathfrak {m}}}}_{I_q}\) for \(p \ne q\). Conversely, suppose \({{{\mathfrak {k}}}}= {{{\mathfrak {h}}}}\oplus {{{\mathfrak {m}}}}_{I_1} \le \ker (A)\) and for all \(p \ne q\), we have \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_p}]\perp {{{\mathfrak {m}}}}_{I_q}\). Thus \({\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{\mathfrak {k}}}}^\perp }\) preserves the eigenspaces of \(A\vert _{{{{\mathfrak {k}}}}^\perp }\). By Lemma 7.7, to see that \(A \in {\text {D}}({{{\mathfrak {k}}}})\) we need that \([A\vert _{{{{\mathfrak {k}}}}^\perp },{\text {ad}}({{{\mathfrak {k}}}})\vert _{{{{\mathfrak {k}}}}^\perp }]=0\) which follows from \([A|_{{{{\mathfrak {m}}}}_{I_p}},{\text {ad}}({{{\mathfrak {m}}}}_{I_1})|_{{{{\mathfrak {m}}}}_{I_p}}] =0\) for every \(p>1\). Let \(X_1 \in {{{\mathfrak {m}}}}_{I_i}\) and \(X_p \in {{{\mathfrak {m}}}}_{I_p}\), \(p>1\). Then
since by hypothesis, \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_p}] \subset {{{\mathfrak {m}}}}_{I_p}\). Thus our equality holds. \(\square \)
Here we provide a brief overview of representation theory of compact Lie groups K. Let V be a vector space over \({{{\mathbb {C}}}}\). We call a representation \(\rho _{{{{{\mathbb {C}}}}}}:K \rightarrow {\text {U}}(V) \subset \textrm{Gl}(V,{{{\mathbb {C}}}})\) an irreducible complex representation if there exists no proper, non-trivial \({{{\mathbb {C}}}}\)-vector subspace \({\tilde{V}}\) of V which is \(\rho _{{{{{\mathbb {C}}}}}}(K)\)-invariant. The realization of an irreducible complex representation \(\rho _{{{{{\mathbb {C}}}}}}:K \rightarrow {\text {U}}(V,{{{\mathbb {C}}}})\) is given by \(\rho _{{{{{\mathbb {R}}}}}}:K\rightarrow {\text {SO}}(W) \subset \text {Gl} (W,{{{\mathbb {R}}}})\,;\,\, k \mapsto \rho _{{{{{\mathbb {C}}}}}}(k)\), where W is the \({{{\mathbb {R}}}}\)-vector space induced by V; that is, \(W=V\) as sets, but \(\dim _{{{{{\mathbb {R}}}}}} W =2 \dim _{{{{{\mathbb {C}}}}}} V\). Such representations are called real representations. Notice that the realization of an irreducible complex representation may not be irreducible.
An irreducible real representation \(\rho _{{{{{\mathbb {R}}}}}}:K \rightarrow {\text {SO}}(W)\) is called of real type, complex type or quaternionic type, respectively, if the group of intertwining operators
is isomorphic to \({{{\mathbb {R}}}}\), \({{{\mathbb {C}}}}\) or \({{{\mathbb {H}}}}\), respectively: see [21, II, 6.2, II, Theorem 6.7].
Lemma 7.9
Let \(\rho _{{{{{\mathbb {C}}}}}}:{\text {SU}}(n)\rightarrow {\text {U}}(V,{{{\mathbb {C}}}})\) be an irreducible complex representation, \(n \ge 2\). Suppose that \(\rho _{{{{{\mathbb {C}}}}}}\) can be extended to a representation of \({\text {U}}(n)\) such that the center of \({\text {U}}(n)\) acts non-trivially on V by multiples of the identity. Then the realization \(\rho _{{{{{\mathbb {R}}}}}}:{\text {U}}(n)\rightarrow {\text {SO}}(W)\), \(W=V\), is an irreducible real representation.
Proof
For a contradiction, suppose that \({\tilde{W}}\) is a non-trivial proper real subspace of W which is \(\rho _{{{{{\mathbb {R}}}}}}({\text {U}}(n))\)-invariant. By assumption there exists some \(\alpha \in {{{\mathbb {R}}}}\) such that for every \(e^{i\varphi } \cdot I_n\) in the center of \({\text {U}}(n)\), we have \(\rho _{{{{{\mathbb {C}}}}}}(e^{i\varphi } \cdot I_n)=e^{i\varphi \alpha }\cdot I_V\). Thus whenever \({\tilde{w}} \in {\tilde{W}}\), then \(i\cdot {\tilde{w}} \in {\tilde{W}}\) as well. Consequently \({\tilde{W}}\) is an \({\text {SU}}(n)\)-invariant complex subspace of \(W=V\), contradicting our hypothesis. Consequently W is an irreducible real representation. \(\square \)
1.3 Lie-theory II
In this section we describe the Casimir operator of the isotropy representation of a compact homogeneous space G/H and its relation to the Killing form and the structure constants.
Let \((Z_m)_{1\le m \le \dim H}\) denote a Q-orthonormal basis of \({{{\mathfrak {h}}}}\). Let \({{{\mathfrak {m}}}}_i\) be an \({\text {Ad}}(H)\)-irreducible summand of \({{{\mathfrak {m}}}}\). The Casimir operator on \({{{\mathfrak {m}}}}_i\) is given by
and it satisfies
A short computation using (4.3) shows \(c_i\ge 0\). Clearly, \(c_i=0\) if and only if \({{{\mathfrak {m}}}}_i \subset {{{\mathfrak {m}}}}_0\): see Lemma 7.5.
Recall, the definitions of \(d_i\), \(b_i\) and [ijk] can be found in Sect. 4.1 and that these numbers depend on the decomposition \(f=\oplus _{i=1}^\ell {{{\mathfrak {m}}}}_i\) of \({{{\mathfrak {m}}}}\) choosen.
Lemma 7.10
[77] Let G/H be a compact homogeneous space and let f be a decomposition of \({{{\mathfrak {m}}}}\). Then for all \(1 \le i \le \ell \),
Moreover, the expression is zero if and only if \({{{\mathfrak {m}}}}_i \subset {{{\mathfrak {z}}}}({{{\mathfrak {g}}}})\).
Proof
The equality is proved in [77, La.1.5].
Suppose that \(i=1\) and the left-hand side is zero. Then \(c_1=0\), thus \([{{{\mathfrak {h}}}},{{{\mathfrak {m}}}}_1]=0\) and \({{{\mathfrak {m}}}}_1 \subset {{{\mathfrak {m}}}}_0\) by Lemma 7.4. Since \([11j]=0\) for all \(1\le j\le \ell \), the module \({{{\mathfrak {m}}}}_1\) is an abelian subalgebra of \({{{\mathfrak {m}}}}_0\). Moreover, since for all \(2 \le j,k\le \ell \) we have \([1jk]=0\), the subspace \({{{\mathfrak {g}}}}':={{{\mathfrak {h}}}}\oplus (\bigoplus _{j=2}^\ell {{{\mathfrak {m}}}}_j)\) is a subalgebra of \({{{\mathfrak {g}}}}\) commuting with \({{{\mathfrak {m}}}}_1\): see Lemma 5.21. It follows that \({{{\mathfrak {m}}}}_1 \subset {{{\mathfrak {z}}}}({{{\mathfrak {g}}}})\).
The converse direction is clear. \(\square \)
Lemma 7.11
Let \({{{\mathfrak {k}}}}\in \mathcal {K}\) and let f be a decomposition of \({{{\mathfrak {m}}}}\) such that \({{{\mathfrak {k}}}}\) is f-adapted, and let \(I_1=I^{{{{\mathfrak {k}}}}}_1\). Then
Furthermore, \({{{\mathfrak {k}}}}\) is toral if and only if \(a_{{{{\mathfrak {k}}}}} =0\).
Proof
Suppose that \({{{\mathfrak {k}}}}={{{\mathfrak {h}}}}\oplus {{{\mathfrak {a}}}}\) is a toral subalgebra of \({{{\mathfrak {g}}}}\), that is, \({{{\mathfrak {a}}}}\subset {{{\mathfrak {m}}}}_0\). Then clearly \(a_{{{{\mathfrak {k}}}}}=0\). Conversely, when \(a_{{{{\mathfrak {k}}}}}=0\), then \(\sum _{j\in I_1}d_jc_j=0\), so we know \({{{\mathfrak {m}}}}_{I_1}:={{{\mathfrak {m}}}}_{{{{\mathfrak {k}}}}}\subset {{{\mathfrak {m}}}}_0\). Because \({{{\mathfrak {k}}}}\) is a subalgebra we conclude \([{{{\mathfrak {m}}}}_{I_1},{{{\mathfrak {m}}}}_{I_1}]\subset {{{\mathfrak {m}}}}_{I_1}\). Since \([I_1I_1I_1]=0\), the claim follows. \(\square \)
Corollary 7.12
For every a compact homogeneous space G/H, there exists a constant \(n_{G/H}>0\) such that the following holds: given any non-toral subalgebra \({{{\mathfrak {k}}}}\), with f a decomposition of \({{{\mathfrak {m}}}}\) such that \({{{\mathfrak {k}}}}\) is f-adapted, and letting \(I_1=I^{{{{\mathfrak {k}}}}}_1\), then
Proof
For a contradiction, suppose that there exists a sequence of non-toral subalgebras \(({{{\mathfrak {k}}}}_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\) and a sequence \((f_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\) of decompositions of \({{{\mathfrak {m}}}}\), such that for each \(\alpha \in {{{\mathbb {N}}}}\), \({{{\mathfrak {k}}}}_\alpha \) is \(f_\alpha \)-adapted, with \(a_{{{{\mathfrak {k}}}}_\alpha } \rightarrow 0\) as \(\alpha \rightarrow \infty \). Because the Casimir operator \(C\vert _{{{{\mathfrak {m}}}}}:{{{\mathfrak {m}}}}\rightarrow {{{\mathfrak {m}}}}\) is non-negative with kernel \({{{\mathfrak {m}}}}_0\), for \(\alpha \) large we must have \({{{\mathfrak {m}}}}_{{{{{\mathfrak {k}}}}}_\alpha }\subset {{{\mathfrak {m}}}}_0\). But since \({{{\mathfrak {m}}}}_0\) and \({{{\mathfrak {k}}}}_\alpha \) are subalgebras of \({{{\mathfrak {g}}}}\), so is \({{{\mathfrak {k}}}}_\alpha ^0:={{{\mathfrak {m}}}}_{{{{{\mathfrak {k}}}}}_\alpha }={{{\mathfrak {m}}}}\cap {{{\mathfrak {k}}}}_{\alpha }\). Since each \({{{\mathfrak {k}}}}_\alpha \) is non-toral by assumption, \({{{\mathfrak {k}}}}_\alpha ^0\) is a non-abelian subalgebra of \({{{\mathfrak {m}}}}_0\) and consequently, \({{{\mathfrak {k}}}}_\alpha \) has a nonvanishing semisimple part \(({{{\mathfrak {k}}}}_\alpha ^0)^{ss}\). Passing to a subsequence, we may assume that these semisimple subalgebras converge to a limit subalgebra \({{{\mathfrak {k}}}}_\infty ^0\) of \({{{\mathfrak {m}}}}_0 \subset {{{\mathfrak {g}}}}\), and also that \(f_\alpha \rightarrow f_\infty \) as \(\alpha \rightarrow \infty \). Up to conjugation, \({{{\mathfrak {g}}}}\) has only finitely many semisimple subalgebras, see [17, Corollary 4.5]; thus \({{{\mathfrak {k}}}}_\infty ^0\) must be semisimple. Hence \([I_1I_1I_1]_{f_\infty }>0\), a contradiction. \(\square \)
Corollary 7.13
Let G/H be a compact homogeneous space with finite fundamental group and let f be a decomposition of \({{{\mathfrak {m}}}}\) into \({\text {Ad}}(H)\)-irreducible summands. For each \(1\le i \le \ell \),
and consequently
Proof
By Lemma 7.10 we know if \(2d_ic_i+\frac{1}{2}\sum _{j,k=1}^\ell [ijk]=0\), then \({{{\mathfrak {m}}}}_i\) is an abelian subalgebra in \({{{\mathfrak {g}}}}\) and \({{{\mathfrak {m}}}}_i\) commutes with the subalgebra \({{{\mathfrak {h}}}}\oplus (\bigoplus _{j=1,j\ne i}^\ell {{{\mathfrak {m}}}}_j)\). This would imply that G/H has infinite fundamental group, see Lemma 5.21, contradicting our hypothesis. This proves the claim. \(\square \)
1.4 Lie theory III: A Łojasiewicz inequality for structure constants
We recall the Łojasiewicz inequality in semi-algebraic geometry (cf. [7, Proposition 2.3.11]): Let \(K \subset {{{\mathbb {R}}}}^n\) be a compact semi-algebraic set and let \(f,g:K \rightarrow {{{\mathbb {R}}}}\) be continuous and semi-algebraic. Suppose that for all \(x \in K\) we have: \(f(x)=0 \Rightarrow g(x)=0\). Thus \(f^{-1}(0) \subset g^{-1}(0)\). Then there exists \(N\in {{{\mathbb {N}}}}\) and \(C=C(f,g,K)\ge 0\) such that for all \(x \in K\) we have
The Łojasiewicz inequality stated in Proposition 7.14 provides the key estimate for the proof of Theorem 4.10. Note that standard estimation methods cannot be applied to the above situation.
Let \(I^{\textsf {c}}=\{1,\ldots ,\dim {{{\mathfrak {g}}}}\}\backslash I\).
Proposition 7.14
Let G be a compact, connected Lie group endowed with a biinvariant metric Q. Let \(I\subset \{1,\ldots ,\dim \,{{{\mathfrak {g}}}}\}\) with \(1\in I\). Let \(Z_{\vert I\vert }\) denote the set of all Q-orthonormal bases \(b:=(e_1,\ldots ,e_{\dim {{{{\mathfrak {g}}}}}})\) of \(({{{\mathfrak {g}}}},Q)\) such that
is an abelian subalgebra. Then there exists a constant \(C=C(G,Q)>0\) and an open neighborhood \(U_{\vert I\vert }\) of \(Z_{\vert I\vert }\) in the space of all Q-orthonormal bases, such that for all \(\tilde{b}:=({\tilde{e}}_1,\ldots ,{\tilde{e}}_{\dim {{{{\mathfrak {g}}}}}}) \in U_{\vert I\vert }\) we have
Proof
First note that if \(I^{\textsf {c}}=\emptyset \), then \({{{\mathfrak {g}}}}\) is abelian and the above claim is trivially true. Hence we may assume that \(I^{\textsf {c}}\ne \emptyset \) and that \({{{\mathfrak {g}}}}\) is not abelian. We prove the result by induction on \(\vert I\vert \). It is clear that for \(\vert I\vert =1,2\) the above claim is true, since the left hand side of (7.4) is zero by \({\text {Ad}}\)-invariance of Q. Thus we may assume that \(\vert I\vert \ge 3\).
We first show that for an arbitrary choice of Q-orthonormal basis \(b \in Z_{\vert I\vert }\) and an arbitrary choice of sequence \((b_\alpha )_{\alpha \in {{{\mathbb {N}}}}}:=(e_1(\alpha ),\ldots ,e_{\dim {{{{\mathfrak {g}}}}}}(\alpha ))_{\alpha \in {{{\mathbb {N}}}}}\) of Q-orthonormal bases for \({{{\mathfrak {g}}}}\) with \(\lim _{\alpha \rightarrow \infty }b_\alpha =b\), such a constant \(C>0\) exists (a priori, C may depend on b and the sequence \((b_\alpha )\)).
For a contradiction, suppose that no such constant \(C>0\) exists. In what follows we will pass to a subsequence whenever convenient, without explicitly mentioning it. For \(1 \le j,k \le \dim {{{\mathfrak {g}}}}\), we set
We set \(I':= I\backslash \{1\}\) and get
where \(g:{{{\mathbb {R}}}}\rightarrow {{{\mathbb {R}}}}\) with \(\lim _{\alpha \rightarrow \infty }g(\alpha )=+\infty \).
Suppose that for each \(\alpha \in {{{\mathbb {N}}}}\), \(e_1(\alpha )\in {{{\mathfrak {t}}}}(\alpha )\), where \({{{\mathfrak {t}}}}(\alpha )\) denotes a maximal abelian subalgebra of \({{{\mathfrak {g}}}}\). Since maximal abelian subalgebras of \({{{\mathfrak {g}}}}\) are conjugate, there exists a maximal torus \({{{\mathfrak {t}}}}(\infty )\) and a sequence \((g_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\) of group elements in G with \(\textrm{Ad}(g_\alpha )({{{\mathfrak {t}}}}(\alpha ))={{{\mathfrak {t}}}}(\infty )\). We obtain a sequence \((b_\alpha )\) of Q-orthonormal bases of \({{{\mathfrak {g}}}}\) converging to b, with \(e_1(\alpha )\in {{{\mathfrak {t}}}}(\infty )\) for all \(\alpha \in {{{\mathbb {N}}}}\), which we again denote by \(b_\alpha =(e_1(\alpha ),\ldots , e_{\dim {{{{\mathfrak {g}}}}}}(\alpha ))\). Since all terms in (7.5) are \(\textrm{Ad}(G)\)-invariant, Eq. (7.5) holds for each \(b_\alpha \) in this sequence as well.
Step 1: Let \(V_1(\alpha ):=\langle e_2(\alpha ),\ldots , e_{\vert I\vert }(\alpha )\rangle _{{{{{\mathbb {R}}}}}}\), and let \(V_1(\alpha )^\perp \) denote the Q-orthogonal complement of \(V_1(\alpha )\) in \({{{\mathfrak {g}}}}\). Let \(j\in I'\). Since \(e_1(\alpha )\in {{{\mathfrak {t}}}}(\infty )\) we have
As mentioned above, we prove our proposition by induction on \(\vert I\vert \). Suppose that, after renumbering, \(I=\{1,\ldots ,\vert I\vert \}\), and that (after passing to a subsequence)
for all \(j,k\in I'\) and all \(\alpha \in {{{\mathbb {N}}}}\). By (7.5) we know that \(\vert s_{23}(\alpha )\vert >0\) for all sufficiently large \(\alpha \). Thus, for all \(j,k\in I'\)
Note that \(\lim _{\alpha \rightarrow \infty } s_{23}(\alpha )=0\), since \({{{\mathfrak {t}}}}=\oplus _{i=1}^{\vert I \vert }\langle e_i \rangle \) is abelian.
By (7.5) we claim that for each \(j\in I'\) there exists some \(k\in I'\) such that the limit in (7.6) is positive. For \(\vert I\vert =3\) this is trivially true. To see the claim, suppose that there exists \(j_0 \in I'\backslash \{2,3\}\) such that this limit is zero for all \(k\in I'\). It follows that there exists a function \({\tilde{g}}:{{{\mathbb {R}}}}\rightarrow {{{\mathbb {R}}}}\) with \({\displaystyle \lim _{\alpha \rightarrow \infty } {\tilde{g}}(\alpha )=+\infty }\) and
for all \(k \in I'\) and all \(\alpha \in {{{\mathbb {N}}}}\). Let \(I_{j_0}:= I \backslash \{j_0\}\) and \(I'_{j_0}:= I'\backslash \{j_0\}\). Then
and
We deduce from (7.5)
By (7.7) this yields
Since \(\sum _{j,k \in I'_{j_0}} [1jk]_\alpha \ge [123]_\alpha \), using \(j_0 \ne 2,3\), this yields
This shows that for the torus \({\displaystyle {{{\mathfrak {t}}}}_{j_0}:=\oplus _{i\in I_{j_0}} \langle e_i \rangle }\) of dimension \(\vert I\vert -1\) we obtain an inequality like (7.5), because as \(\alpha \rightarrow \infty \),
By the induction hypothesis, we obtain a contradiction. This proves for each \(j\in I'\) there exists some \(k\in I'\) such that the limit in (7.6) is positive.
Step 2: For every \(j\in I'\) there exists some \(k(j)\in I'\backslash \{j\}\) with
for all \(k\in I'\) and \(\alpha \in {{{\mathbb {N}}}}\). By Step 1 we have
which yields \(\vert s_{jk(j)}(\alpha )\vert >0\) for all sufficiently large \(\alpha \in {{{\mathbb {N}}}}\). Thus we obtain, for all \(j \in I'\),
with
The left hand side of (7.5) is bounded from above by \(C' \cdot (s_{23}(\alpha ))^2\) by the very definition of \(s_{23}(\alpha )\), where \(C'\) is independent of \(\alpha \). Meanwhile, \(\Vert X_j^1(\alpha )\Vert ^2 \cdot s_{jk(j)}(\alpha )^2 \) contributes to the right hand side. Since by (7.8), the limit behavior of \(s_{jk(j)}(\alpha )\) is like that of \(s_{23}(\alpha )\), in order for the inequality in (7.5) to hold while \(g(\alpha ) \rightarrow \infty \), we must have
Next, let \({{{\mathfrak {c}}}}(e_1)\le {{{\mathfrak {g}}}}\) denote the centralizer of \(e_1\). Since \(e_1 \in {{{\mathfrak {t}}}}(\infty )\) and \([e_1,{{{\mathfrak {t}}}}']=0\), where \({\displaystyle {{{\mathfrak {t}}}}':=\oplus _{i\in I'} \langle e_j\rangle \subset {{{\mathfrak {t}}}}}\) is abelian, we obtain
Using that \(1 \le \Vert E_j^1(\alpha )\Vert \le C''\) by the definition of \(s_{jk(j)}(\alpha )\) this implies
This is seen as follows: By (7.9) we have \(E_j^1(\alpha )+X_j^1(\alpha ) \in {{{\mathfrak {t}}}}(\infty )^\perp \). Because we know \(\lim _{\alpha \rightarrow \infty }X_j^1(\alpha )=0\) this implies \(E_j(\infty ) \in {{{\mathfrak {t}}}}(\infty )^\perp \). Moreover, we have \(E_j^1(\alpha )\in V_1(\alpha )\) by definition. Since \(V_1(\alpha )\rightarrow {{{\mathfrak {t}}}}'\) as \(\alpha \rightarrow \infty \), we deduce \(E_j(\infty )\in {{{\mathfrak {t}}}}'\), so it follows that \(E_j(\infty ) \in {{{\mathfrak {c}}}}(e_1)\) by (7.11).
An important consequence is that
In fact we have that
Step 3: We show by induction on p, for \(1 \le p \le \dim {{{\mathfrak {t}}}}(\infty )\), that \({{{\mathfrak {t}}}}(\infty )\subsetneq {{{\mathfrak {c}}}}({{{\mathfrak {t}}}}(\infty ))\), which is clearly impossible. This will prove our contradiction for the specific choice of Q-orthonormal basis \(b \in Z_{\vert I\vert }\) and sequence \((b_\alpha )_{\alpha \in {{{\mathbb {N}}}}}:=(e_1(\alpha ),\ldots ,e_{\dim {{{{\mathfrak {g}}}}}}(\alpha ))_{\alpha \in {{{\mathbb {N}}}}}\) of Q-orthonormal bases for \({{{\mathfrak {g}}}}\) with \(\lim _{\alpha \rightarrow \infty }b_\alpha =b\). Our induction hypotheses for each p, denoted by \((A1)_p\),...,\((A4)_p\), are defined here:
Assumption \((A1)_p\): There exist a Q-orthonormal frame \((e_1^1:=e_1,\ldots ,e_1^p)\) in \({{{\mathfrak {t}}}}(\infty )\), positive numbers \(d_1(\alpha ),\ldots ,d_p(\alpha )>0\), and \(r_1^{p+1}(\alpha )\in {{{\mathfrak {t}}}}(\infty )\), \(r_1^{p+1}(\alpha ) \perp e_1^1,\ldots ,e_1^p\), such that
Before we define \((A2)_p\), we set \({{{\mathfrak {c}}}}(0):={{{\mathfrak {g}}}}\) and
For \(j \in I'\) let
Let
and let \(V_p(\alpha )^\perp \) denote the Q-orthogonal complement of \(V_p(\alpha )\) in \({{{\mathfrak {c}}}}(p-1)\). Let
recalling, as we showed in Step 2, that \(s_{jk(j)}(\alpha )>0\) for all sufficiently large \(\alpha \in {{{\mathbb {N}}}}\).
Assumption \((A2)_p\): We have
Notice that because \(e_1(\alpha )\in {{{\mathfrak {t}}}}(\infty )\subset {{{\mathfrak {c}}}}(p)\subset {{{\mathfrak {c}}}}(p-1)\), we have
Assumption \((A3)_p\): For every \(j \in I'\) we know \({\displaystyle \lim _{\alpha \rightarrow \infty }X_j^p(\alpha )=0}\) and
Assumption \((A4)_p\): For every \(j \in I'\), we have
The base case \(p=1\) of our induction was essentially established in Step 2: By definition we have \({\displaystyle \lim _{\alpha \rightarrow \infty }e_1(\alpha )=e_1}\), that is, we can write \(e_1(\alpha )=d_1(\alpha )\cdot e_1 + r_1^2(\alpha )\) with \(d_1(\alpha )>0\), \(r_1^2(\alpha )\in {{{\mathfrak {t}}}}(\infty )\) and \(e_1 \perp r_1^2(\alpha )\). Thus \((A1)_1\) holds. Next, \((A2)_1\) is established in (7.13). Equations (7.10) and (7.12) give \((A3)_1\). And then, because \(e_j^1(\alpha )=e_j(\alpha ) \rightarrow e_j \in {{{\mathfrak {t}}}}' \subset {{{\mathfrak {c}}}}(e_1)\), \((A4)_1\) is also clear.
We now prove the induction step \(p \rightarrow p+1\le \dim {{{\mathfrak {t}}}}(\infty )\): By the induction hypothesis \((A2)_p\) we have \({{{\mathfrak {t}}}}(\infty ) \subset {{{\mathfrak {c}}}}(p)\), thus \([{{{\mathfrak {t}}}}_\infty ,{{{\mathfrak {c}}}}(p)^\perp ]\subset {{{\mathfrak {c}}}}(p)^\perp \). For \(j \in I'\) we define \(e_j^{p+1}(\alpha ):=\textrm{pr}_{{{{{\mathfrak {c}}}}}(p)}(e_j(\alpha ))\). Then by \((A1)_p\) we have
Using \((A2)_p\), the equation after (7.15), and \({{{\mathfrak {c}}}}(p-1) \supset {{{\mathfrak {c}}}}(p)\), we see
By \((A3)_p\) we know
Since \(s_{jk(j)}(\alpha )>0\) we must have \(r_1^{p+1}(\alpha )\ne 0\) for all sufficiently large \(\alpha \in {{{\mathbb {N}}}}\). Set
Then, by passing to a further subsequence, we may assume \({\displaystyle \lim _{\alpha \rightarrow \infty } e_1^{p+1}(\alpha )=e_1^{p+1} \in {{{\mathfrak {t}}}}(\infty )}\), with \(e_1^{p+1}\perp e_1,\ldots ,e_1^p\) and \(\Vert e_1^{p+1} \Vert =1\). This shows \((A1)_{p+1}\).
The induction hypothesis \((A4)_p\) guarantees for all \(j \in I'\),
Thus for all sufficiently large \(\alpha \in {{{\mathbb {N}}}}\),
is a subspace of \({{{\mathfrak {c}}}}(p)\) of dimension \(\vert I'\vert \), converging to \({{{\mathfrak {t}}}}'\) as \(\alpha \rightarrow \infty \). Using (7.14), the definition of \({{{\mathfrak {c}}}}(p)\), and \(e_1^{p+1}(\alpha )\) we deduce from (7.18) for all \(j \in I'\)
where \({\displaystyle E^{p+1}_j(\alpha ):= \text {pr} _{V_{p+1}(\alpha ) }(\tilde{E}_j^{p+1}(\alpha ))\in {{{\mathfrak {c}}}}(p)}\) and \({\displaystyle X^{p+1}_j(\alpha ):= \textrm{pr}_{V_{p+1}(\alpha )^\perp }(\tilde{E}_j^{p+1}(\alpha ))\in {{{\mathfrak {c}}}}(p)}\).
Notice, by Eq. (7.21),
Since by (7.19) \({\displaystyle \lim _{\alpha \rightarrow \infty }\tilde{E}^{p+1}_j(\alpha )=E_j(\infty )\in {{{\mathfrak {t}}}}'}\) and \(V_{p+1}(\alpha )\rightarrow {{{\mathfrak {t}}}}'\) as \(\alpha \rightarrow \infty \), we get
This shows the first part of \((A3)_{p+1}\). Next, we claim \({\displaystyle \lim _{\alpha \rightarrow \infty }\tfrac{s_{jk(j)}(\alpha )}{\Vert r_1^{p+1}(\alpha ) \Vert } =0}\). To see this, notice that \(\lim _{\alpha \rightarrow \infty }e_1^{p+1}(\alpha )=e_1^{p+1}\in {{{\mathfrak {t}}}}(\infty )\), that \(\lim _{\alpha \rightarrow \infty } e_j^{p+1}(\alpha ) =e_j \in {{{\mathfrak {t}}}}'\) by (7.20) and that \(\lim _{\alpha \rightarrow \infty } (E_j^{p+1}(\alpha )+X_j^{p+1}(\alpha ))=E_j(\infty ) \in {{{\mathfrak {t}}}}'\) for all \(j \in I'\). Now if \({\displaystyle \tfrac{s_{jk(j)}(\alpha )}{\Vert r_1^{p+1}(\alpha ) \Vert } }\) had a uniform positive lower bound, then by passing to a subsequence if necessary, then we deduce \(Q([e_1^{p+1},e_j],E_j(\infty ))>0\) by (7.21), because \(E_j(\infty )\ne 0\). But since \(e_j,E_j(\infty )\in {{{\mathfrak {t}}}}'\) and since \({{{\mathfrak {t}}}}'\) is abelian, the left-hand side equals zero. This shows the above claim.
By (7.21) we now deduce \([e_1^{p+1},e_j]=0\) for all \(j \in I'\), that is,
Using this and \(E_j(\infty )\in {{{\mathfrak {c}}}}(p)\cap {{{\mathfrak {t}}}}'\) we deduce
where \({{{\mathfrak {c}}}}(p+1)={{{\mathfrak {c}}}}(p)\cap {{{\mathfrak {c}}}}(e_1^{p+1})\). By (7.20), this yields \((A2)_{p+1}\):
Notice that this implies
implying both the second part of \((A3)_{p+1}\) and also \((A4)_{p+1}\).
So far we have proved that for a given \(b\in Z_{\vert I\vert }\) and a given sequence \((b_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\) converging to b, such a constant C exists for all sufficiently large \(\alpha \). We next show that C can be chosen independently of both \(b\in Z_{\vert I\vert }\) and the sequence \((b_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\).
To this end, suppose that there exists a sequence \((b_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\) in \(Z_{\vert I\vert }\) and corresponding sequences \((b_{\alpha ,\beta })_{\beta \in {{{\mathbb {N}}}}}\) with \(\lim _{\beta \rightarrow \infty }b_{\alpha ,\beta }=b_\alpha \) for all \(\alpha \in {{{\mathbb {N}}}}\), such that for the corresponding optimal constants, as \(\alpha \rightarrow \infty \), \(C_\alpha \rightarrow +\infty \). Here by optimal we mean that the constants \(C_\alpha -1\) would not work. Since the constants \(C_\alpha \) are chosen to be optimal, for each \(\alpha \in {{{\mathbb {N}}}}\) there exists \(N(\alpha ) \in {{{\mathbb {N}}}}\), such that we may assume (after passing to a subsequence) that
for all \(\beta \ge N(\alpha )\). By again passing to a subsequence, we may assume that this holds for all \(\alpha ,\beta \ge 1\). Since \(Z_{\vert I\vert }\) is compact, we may assume that \(\lim _{\alpha \rightarrow \infty }b_\alpha =b\). Moreover, it is clear that we can construct a sequence \(({\tilde{b}}_\alpha )_{\alpha \in {\textbf{N}} }\) converging to b and satisfying (7.22). Since \(C_\alpha \rightarrow + \infty \) as \(\alpha \rightarrow \infty \), we obtain a contradiction. This shows that C can be chosen independently.
To prove the theorem, suppose now that no such constant \(C=C(G,Q)\) and no such open neighborhood \(U_{\vert I\vert }\) of \(Z_{\vert I\vert }\) exist. Then there exists a sequence of bases \((b_\alpha )_{\alpha \in {{{\mathbb {N}}}}}\) converging to \(b \in Z_{\vert I\vert }\) for which (7.22) holds. Contradiction. \(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Böhm, C., Kerr, M.M. Homogeneous Einstein metrics and butterflies. Ann Glob Anal Geom 63, 29 (2023). https://doi.org/10.1007/s10455-023-09905-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10455-023-09905-0