Isometries of Hermitian symmetric spaces
Makiko Sumi Tanaka
The 6th OCAMI-KNUGRG Joint Differential Geometry Workshop on Submanifold Theory in Symmetric Spaces and Lie Theory in Finite and Infinite Dimensions
Osaka City University February 1–3, 2013
1 Joint with Jost-Hinrich Eschenburg and Peter Quast
Contents
1 Introduction
2 Extrinsically symmetric spaces
3 Isometries of Hermitian symmetric spaces
4 Another result
2 1. Introduction A Riemannian symmetric space M is called a symmetric R-space if M is realized as an orbit of a linear isotropy representation of a Riemannian symmetric space of compact type.
Every Hermitian symmetric space M of compact type is a symmetric R-space since M is realized as an adjoint orbit of a compact semisimple Lie group.
3 Every symmetric R-space is realized as a real form of a Hermitian symmetric space of compact type and vice- versa (Takeuchi 1984). Here a real form of a Hermitian symmetric space M is the fixed point set of an involutive anti-holomorphic isometry of M. Every real form is con- nected and a totally geodesic Lagrangian submanifold of M.
4 M : a Riemannian manifold τ : an involutive isometry of M A connected component of F (τ, M) := {x ∈ M | τ(x) = x} is called a reflective submanifold.
Reflective submanifolds in simply connected irreducible Riemannian symmetric spaces of compact type are clas- sified by Leung (1974, ’75 and ’79).
5 Every symmetric R-space is considered as a reflective submanifold of a Hermitian symmetric space of compact type.
Q. Is every reflective submanifold of a symmetric R-space a symmetric R-space ?
If a symmetric R-space is simply connected and irre- ducible, we know it is true by Leung’s classification.
6 2. Extrinsically symmetric spaces M ⊂ Rn : a submanifold of Euclidean space M is called extrinsically symmetric if it is preserved by ⊥ the reflection ρp at the affine normal space p + (TpM) n n for every p ∈ M. Here ρp : R → R is the affine isometry with ρ (p) = p, ρ | = −Id and ρ | ⊥ = Id. p p TpM p (TpM)
Every extrinsically symmetric submanifold is a Rieman- nian symmetric space w.r.t. the induced metric. In fact, the geodesic symmetry sp at p ∈ M is given by sp = ρp|M .
7 Theorem (Ferus 1974, Eschenburg-Heintze 1995) Every symmetric R-space is extrinsically symmetric and every full compact extrinsically symmetric submanifold of Euclidean space is a symmetric R-space.
8 P = G/K : a Riemannian symmetric space of compact type
G = I0(M), K = {g ∈ G | g(o) = o}, o ∈ P g = k ⊕ p : the canonical decomposition ∼ g = Lie(G), k = Lie(K), p = ToP
Take ξ(≠ 0) ∈ p with ad(ξ)3 = −ad(ξ), then M = AdG(K)ξ ⊂ p : a symmetric R-space
9 τ : an involutive isometry of M L : a reflective submanifold determined by τ
If we assume ∃T : a linear isometry of p with T (M) ⊂ M s.t. T |M = τ, L is an extrinsically symmetric submanifold in a linear subspace F (T, p) of p.
10 I(M) : the group of isometries of M T(M) : the transvection group of M, i.e., a subgroup of
I(M) generated by {sp ◦ sq | p, q ∈ M}, which is coincides with the identity component I0(M) of I(M)
If f ∈ T(M), f can be extended to a linear isometry of p since every geodesic symmetry sp has the extension ρp.
But what about arbitrary isometry?
11 Remark
We have a list of I(M)/I0(M) for irreducible M in p.156 of Loos’s book “Symmetric spaces II”.
12 3. Isometries of Hermitian symmetric spaces P : a semisimple Hermitian symmetric space (i.e., of compact type or noncompact type) I(P ) : the isometry group of P A(P ) : the holomorphic isometry group of P T(P ) : the transvection group of P
G := I0(P ) = A0(P ) = T(P ) g = Lie(G) : semisimple J : the complex structure of P p ∈ P
13 → 7→ ◦ ◦ −1 Isp : G G, g sp g sp : an involutive automorphism of G → (Isp)∗ : g g : an involutive automorphism of g ⊕ − g = kp pp,(Isp)∗ = id on kp,(Isp)∗ = id on pp ∼ ∼ kp = Lie(Kp),Kp = {g ∈ G | g(p) = p}, pp = TpP
Jp ∈ C(kp), the center C(kp) of kp
(If P is irreducible, Jp generates C(kp).) ∼ | The action of Jp on TpP = pp is given by ad(Jp) pp. 3 Thus ad(Jp) has eigenvalues ±i and 0, so that ad(Jp) =
−ad(Jp).
14 Consider the map ι : P → g, p 7→ Jp, which is a G- equivariant embedding of P into g with the image Ad(G)Jo, the adjoint orbit for a chosen base point o ∈ P . ι is called the canonical embedding of P .
15 Theorem 1 (Eschenburg-Quast-T., J. of Lie Theory 2013) Let P ⊂ g be a canonically embedded semisimple Hermi- tian symmetric space and let f be an isometry of P , then there exists a linear isometry F of g (w.r.t. a suitable invariant inner product) whose restriction to P coincides with f.
16 Proof:
P = P1 × · · · × Pr : a decomposition into irreducible fac- tors
G = G1 × · · · × Gr : a decomposition of the transvection group G, where Gj is the transvectin group of Pj g = g1 ⊕ · · · ⊕ gr : the direct sum decomposition of g = Lie(G), where gj = Lie(Gj) is simple
ι = ι1 × · · · × ιr where ιj : Pj → gj is the canonical em- bedding of Pj.
17 I(P ) is generated by I(P1) × · · · × I(Pr) and by all per- mutations of isometric irreducible factors of P . Since permutations of isometric irreducible factors of P extend to permutations of the corresponding simple fac- tors of g (possibly up to sign on some factors), we may assume that P is irreducible.
18 Let f ∈ I(P ). Let m be the midpoint of a geodesic arc between o to f(o), then sm ◦ f(o) = o. Since P is an extrinsically symmetric submanifold of g and sm has the extension ρm, we may assume that f(o) = o.
19 −1 If : G → G, g 7→ f ◦ g ◦ f is an automorphism of G
(If )∗ : g → g is an automorphism of g ⟨ , ⟩ : an inner product on g which is proportional to the Killing form
⟨(If )∗(X), (If )∗(Y )⟩ = ⟨X,Y ⟩ (X,Y ∈ g).
20 Lemma (If )∗(Jo) ∈ {±Jo}.
Proof. Since Jo ∈ C(ko) and (If )∗ is a Lie algebra auto- 3 morphism, we have (If )∗(Jo) ∈ C(ko) and ad((If )∗(Jo)) =
−ad((If )∗(Jo)). −1 If k ∈ Ko = {g ∈ G | g(o) = o}, If (k)(o) = f ◦ k ◦ f (o) = | → o. Hence If Ko : Ko Ko is an automorphism of Ko.
Thus (If )∗(C(ko)) = C(ko) = RJo. So (If )∗(Jo) ∈ {±Jo}.
21 Let F := ±(If )∗ if (If )∗(Jo) = ±Jo, respectively. To prove f = F |P , we show ι ◦ f = F ◦ ι. Since we have (ι ◦ f)(o) = (F ◦ ι)(o) and ι ◦ f ◦ γ = F ◦ ι ◦ γ for arbitrary geodesic γ starting from o, we obtain the conclusion.
22 Remark It is a known fact that every reflective sub- manifold of a semisimple Hermitian symmetric space is a symmetric R-space. Because we know every reflective submanifold of a semisimple Hermitian symmetric space is either a complex submanifold or a real form from the next proposition.
23 Proposition (Murakami 1952, Takeuchi 1964) Let P be an irreducible semisimple Hermitian symmetric space.
Then I(P )/I0(P ) and A(P )/A0(P ) are given as follows.
(1) If P is isometric to Q2m(C)(m ≥ 2), 2m Gm(C )(m ≥ 2) or their noncompact duals, ∼ ∼ I(P )/I0(P ) = Z2 × Z2,A(P )/A0(P ) = Z2 (2) Otherwise, ∼ I(P )/I0(P ) = Z2,A(P ) = A0(P ).
∼ Remark I(P )/A(P ) = Z2.
24 4. Another result Theorem 2 (Loos 1985) Let M ⊂ Rn be a compact extrinsically symmetric space. Then the maximal torus of M is a Riemannian product of round circles.
M : a compact Riemannian symmetric space o ∈ M
A connected component of F (so,M) is called a polar.
For p ∈ F (so,M), the connected component of F (sp ◦ so,M) through p is called the meridian.
25 Proposition 1 (Chen-Nagano 1978) If M is a compact Riemannian symmetric space of rank k, then any meridian M− ⊂ M has the same rank k.
Proposition 2 A meridian M− in an extrinsically symmetric space M is extrinsically symmetric.
26 Proposition 3 Let M be a compact Riemannian symmetric space which satisfies that every polar is a single point. Then M is the Riemannian product of round spheres and possibly a torus.
Proposition 4 Let P ⊂ Rn be a full extrinsically symmetric flat torus. Then P splits extrinsically as a product of round circles, = 1 × · · · × 1 ⊂ R2n. P Sr1 Srn
27 Proof of Theorem 2 (Eschenburg-Quast-T.) We prove it by induction over the dimension. The beginning of the induction is the observation that an extrinsically symmetric flat torus is a product of Eu- clidean circles (Proposition 4). As induction step, we show an alternative for arbitrary compact symmetric spaces X: Either X is a Riemannian product of Euclidean spheres and possibly flat torus or it contains a certain totally geodesic submanifold , a so called meridian (Proposition 1).
28 Further, a meridian M− of X is extrinsically symmetric, then so is M− (Proposition 2). Thus passing to meridi- ans again and again, we will lower the dimension preserv- ing preserving the maximal torus unless we reach a space which is a Riemannian product of spheres and possibly a torus and whose maximal torus is a Riemannian product of circles.
29