(2012) arXiv 1109.0585 joint : Darren Long, Stephan Tillmann ( G , X - - PowerPoint PPT Presentation

Projective Structures on Manifolds

Daryl Cooper

U.C.S.B

September 19, 2013

(2012) arXiv 1109.0585 joint: Darren Long, Stephan Tillmann

(G, X) = geometric structure X φ ψ ψ ◦ φ−1 ∈ G Mn ⇒ dev : ˜ M − → X hol : π1M − → G Projective geometry = (PGL(n + 1, R), RPn)

(G, X) = geometric structure X φ ψ ψ ◦ φ−1 ∈ G Mn ⇒ dev : ˜ M − → X hol : π1M − → G Projective geometry = (PGL(n + 1, R), RPn) Constant curvature geometries are “subgeometries” of projective geometry ∴ Every constant curvature n-manifold has underlying projective structure. ⇒ Every surface has projective structure. Mn closed, π1Mn = 1, M real projective ⇒ M = Sn

projective representation of (G,X) geometry (ρ, dev) : (G, X) − → (PGL(n + 1, R), RPn) dev(g · x) = (ρg)(dev x) Molnar (1990), Thiel (1994) 8 Thurston geometries (virtually) have projective representations ⇒ 3-manifold with Thurston geometry (virtually) projective.

projective representation of (G,X) geometry (ρ, dev) : (G, X) − → (PGL(n + 1, R), RPn) dev(g · x) = (ρg)(dev x) Molnar (1990), Thiel (1994) 8 Thurston geometries (virtually) have projective representations ⇒ 3-manifold with Thurston geometry (virtually) projective. Benoist (2006) ∃ projective M3 with non-trivial JSJ.

projective representation of (G,X) geometry (ρ, dev) : (G, X) − → (PGL(n + 1, R), RPn) dev(g · x) = (ρg)(dev x) Molnar (1990), Thiel (1994) 8 Thurston geometries (virtually) have projective representations ⇒ 3-manifold with Thurston geometry (virtually) projective. Benoist (2006) ∃ projective M3 with non-trivial JSJ. C, Goldman (2012 arXiv 1207-2007) RP3#RP3 has no projective structure. Question are there other connected 3-manifolds with no projective structure ?

projective representation of (G,X) geometry (ρ, dev) : (G, X) − → (PGL(n + 1, R), RPn) dev(g · x) = (ρg)(dev x) Molnar (1990), Thiel (1994) 8 Thurston geometries (virtually) have projective representations ⇒ 3-manifold with Thurston geometry (virtually) projective. Benoist (2006) ∃ projective M3 with non-trivial JSJ. C, Goldman (2012 arXiv 1207-2007) RP3#RP3 has no projective structure. Question are there other connected 3-manifolds with no projective structure ?

• M. Kapovich (2007) ∃ non-hyperbolic projective projective W 4

with sec. curv. −1 ≤ K ≤ −1 + ǫ

Projective Surfaces are classified Goldman (1990) convex projective structures TconvRP(F) is cell dimension 8 · |χ(F)| extended by Marquis (2012) to finite area (+ cusps)

Projective Surfaces are classified Goldman (1990) convex projective structures TconvRP(F) is cell dimension 8 · |χ(F)| extended by Marquis (2012) to finite area (+ cusps) Choi + Goldman (1993) holonomies = Hitchin compt of Hom(π1F, PGL(3, R))

Projective Surfaces are classified Goldman (1990) convex projective structures TconvRP(F) is cell dimension 8 · |χ(F)| extended by Marquis (2012) to finite area (+ cusps) Choi + Goldman (1993) holonomies = Hitchin compt of Hom(π1F, PGL(3, R)) Choi + Goldman (1997) Every projective structure obtained by grafting convex one.

Projective Surfaces are classified Goldman (1990) convex projective structures TconvRP(F) is cell dimension 8 · |χ(F)| extended by Marquis (2012) to finite area (+ cusps) Choi + Goldman (1993) holonomies = Hitchin compt of Hom(π1F, PGL(3, R)) Choi + Goldman (1997) Every projective structure obtained by grafting convex one. Labourie (∼2007), Loftin (∼2006) Convex projective structure ↔ (Conf. str + holo cubic diff.) ∴ TconvRP(F) is vector bundle over T (F) Fock + Goncharov (∼2007) Nice coordinates for (br cover of) TconvRP(F)

Affine patch = RPn \ RPn−1 ∼ = Rn Ω ⊂ RPn properly convex if interior of compact convex in affine patch strictly convex if ∄ line segment ⊂ ∂Ω

Affine patch = RPn \ RPn−1 ∼ = Rn Ω ⊂ RPn properly convex if interior of compact convex in affine patch strictly convex if ∄ line segment ⊂ ∂Ω (properly/strictly) convex projective orbifold Q = Ω/Γ PGL(n + 1, R) ⊃ Aut(Ω) = subgroup preserving Ω. Γ = discrete subgroup ⊂ Aut(Ω) Hyperbolic n-space: Ω = D = unit ball Aut(D) = PO(n, 1) ∼ = Isom(Hn)

Affine patch = RPn \ RPn−1 ∼ = Rn Ω ⊂ RPn properly convex if interior of compact convex in affine patch strictly convex if ∄ line segment ⊂ ∂Ω (properly/strictly) convex projective orbifold Q = Ω/Γ PGL(n + 1, R) ⊃ Aut(Ω) = subgroup preserving Ω. Γ = discrete subgroup ⊂ Aut(Ω) Hyperbolic n-space: Ω = D = unit ball Aut(D) = PO(n, 1) ∼ = Isom(Hn) Strictly convex is generalization of hyperbolic manifold Properly convex much more general: like curvature ≤ 0. Benoist (2005) If Mn closed properly convex then M strictly convex ⇔ π1M is δ-hyperbolic.

Hilbert metric on Ω dΩ(a, b) = | log CR(x, a, b, y)|

Ω

x y a b

• Finsler metric.
• Ω = D

dΩ = 2 · dHn

• Ω = interior of triangle dΩ = HEX
• Complete
• preserved by Aut(Ω)
• ⇒ volume form

Strictly convex ⇒ Aut(Ω) = Isom(dΩ) (false for Ω =open simplex) de la Harpe (1991) HEX is norm on C. unit ball is regular hexagon [ex0 : ex1 : ex2] → x0 + x1ω + x2ω2 ω = e2πi/3 biblical metric: unique Finsler metric with π = 3

Isometries of properly convex Ω are hyperbolic/elliptic/parabolic permute pencil of hyperplanes H+ H H− p+ p− C ℓ C Ω Ω Hyperbolic Parabolic Pt Pt Question Is the set of points moved less than d by isometry connected ? convex ??

Benz´ ecri compactness (1960) ∀ p ∈ Ω ⊂ RPn properly convex. ∃ τ ∈ PGL(n + 1, R) st. τ(p) = 0 and B(1) ⊂ Ω ⊂ B(5n). Ω p τ 1 5n τ(Ω) ∴ Hilbert geometries locally uniformly bilipschitz to Euclidean.

V ∼ = Rn Q = S2(V ) := all quadratic forms on V = vector space dim n(n − 1)/2 homo σ2 : GL(V ) − → GL(Q)

V ∼ = Rn Q = S2(V ) := all quadratic forms on V = vector space dim n(n − 1)/2 homo σ2 : GL(V ) − → GL(Q) finitely many orbits ↔ signature q.f. Pos ⊂ Q = convex cone all positive definite q.f. Ωn := P(Pos) ⊂ P(Q) properly convex ∼ = SL(m, R)/SO(m) symmetric space n = −1 + m(m − 1)/2

V ∼ = Rn Q = S2(V ) := all quadratic forms on V = vector space dim n(n − 1)/2 homo σ2 : GL(V ) − → GL(Q) finitely many orbits ↔ signature q.f. Pos ⊂ Q = convex cone all positive definite q.f. Ωn := P(Pos) ⊂ P(Q) properly convex ∼ = SL(m, R)/SO(m) symmetric space n = −1 + m(m − 1)/2 n = 2 ⇒ Ω2 ∼ = H2 S1

∞ ↔ qfs rank 1

Mobius band outside S1

∞ ↔ qf signature (1,1)

V ∼ = Rn Q = S2(V ) := all quadratic forms on V = vector space dim n(n − 1)/2 homo σ2 : GL(V ) − → GL(Q) finitely many orbits ↔ signature q.f. Pos ⊂ Q = convex cone all positive definite q.f. Ωn := P(Pos) ⊂ P(Q) properly convex ∼ = SL(m, R)/SO(m) symmetric space n = −1 + m(m − 1)/2 n = 2 ⇒ Ω2 ∼ = H2 S1

∞ ↔ qfs rank 1

Mobius band outside S1

∞ ↔ qf signature (1,1)

n = 3 dim Ω3 = 5 ∂Ω3 ↔ positive semi-definite qfs. ⊃ ∞·(copies Ω2) ∴ Many JNF in GL(Ωn)

V ∼ = Rn Q = S2(V ) := all quadratic forms on V = vector space dim n(n − 1)/2 homo σ2 : GL(V ) − → GL(Q) finitely many orbits ↔ signature q.f. Pos ⊂ Q = convex cone all positive definite q.f. Ωn := P(Pos) ⊂ P(Q) properly convex ∼ = SL(m, R)/SO(m) symmetric space n = −1 + m(m − 1)/2 n = 2 ⇒ Ω2 ∼ = H2 S1

∞ ↔ qfs rank 1

Mobius band outside S1

∞ ↔ qf signature (1,1)

n = 3 dim Ω3 = 5 ∂Ω3 ↔ positive semi-definite qfs. ⊃ ∞·(copies Ω2) ∴ Many JNF in GL(Ωn) Auslander and Swan (1967) every polycyclic group ⊂ GL(n, Z) some n G f.g. nilpotent ⇒ polycyclic. ∴ ∃ G ⊂ PGL(Pos) discrete group parabolics exotic cusps But maximal cusps (vol < ∞) are Bieberbach groups !

Deformations of convex projective manifolds PC(M) ⊂ Hom(π1M, PGL(n + 1, R)) holonomies of properly convex structures on Mn SC(M) · · · strictly convex · · · (∃ codimension-1 flat embedded projective submanifold in M) ⇒ M deforms. Koszul (1965) M closed ⇒ PC(M) open (C, Long, Tillmann (WIP) extend to finite volume properly convex case + end condition) M closed ⇒ SC(M) is closed Choi + Goldman (1993 n = 2), I. Kim (2005 n = 3), Benoist (2005 all n)

∃ Finite dimensional moduli space of deformations of closed hyperbolic M3

• ften locally rigid but:

C, Long, Thistlethwaite (2006, 2007) Of first 4500 closed hyperbolic 3-manifolds in census 61 infinitesimally deform: H1 = 0 Why ?? Of these it is proved: 25 deform and 3 are rigid Some that deform (e.g. vol3) are non-Haken. For vol3 get free action on building for PGL4 (cf: Culler-Shalen action on trees) Under smoothness assumption on rep. variety: (∃ RPn deformations of Mn

R) ⇔ (∃ CHn deformations of MC)

Cusp C = Ω/Γ Γ = group parabolics fixes p ∈ ∂Ω and H = supp. hyperplane at p

• preserves algebraic horospheres

⇒ C ∼ = (∂C) × [0, ∞) Maximal cusp if ∂C compact ∃ Margulis constant µn for properly convex manifolds

• virt. nilpotent

Thickish-thinnish Mn strictly convex ⇒ M = A ∪ B A ∩ B = ∂A = ∂B

• If inj(x) ≤ 3−(n+1)µn then x ∈ A
• If x ∈ A, then inj(x) ≤ µn/2
• Each component of A is a Margulis tube or a cusp

Mn strictly convex and vol(M) < ∞ then every end is maximal cusp Every maximal cusp in properly convex is hyperbolic = Ω/Γ with Γ ⊂ O(n, 1) fixes p ∈ ∂Ω properly convex Q = SL(n, Z)\SL(n, R)/SO(n) and vol(Q) < ∞ but end not a cusp

”strictly convex = properly convex + rel. δ-hyperbolic + parabolic ends” properly convex M = Ω/Γ & vol(M) < ∞ & M = interior compact N & holonomy each ∂N is parabolic. TFAE:

• M is strictly convex
• ∂Ω is C 1
• π1N is δ-hyperbolic relative to ∪π1∂N

Benoist: compact case. Topological Finiteness

• strictly convex dim = 3 ⇒ fin. many. top. types with vol < V
• properly convex fin. many top. types with diam < D and ∃p inj(p) > ǫ

“Volume bounds diameter when n ≥ 4” : strictly convex dim = n ≥ 4 ∃Cn ∀Mn closed diam(M) ≤ 9 · diam(thick(M)) ≤ Cn · vol(M) true for hyperbolic manifolds.

Theory of geometric transitions + Danciger, Wienhard Continuous family of projective structures Mt on M

• (G, X)-structure for t > 0
• (G ′, X ′)-structure at t = 0

(G, X) (G ′, X ′) e.g Hn En and Sn En but H3 H2 × R