On keen Heegaard splitting (joint work with Ayako Ido and Yeonhee Jang) Tsuyoshi Kobayashi Nara Women’s Univ. Topology and Geometry of Low-dimensional Manifolds Tsuyoshi Kobayashi ( Nara Women’s Univ. October 27, 2014 Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 1 / 41 Contents Part 1 Part 2 Part 3 Part 4 Preliminaries Historical Background and Main Result Technical Background Idea of proof of Main Result Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 2 / 41 Part 1 Preliminaries Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 3 / 41 (Hempel) distance of Heegaard splittings C : compression-body if it is obtained from S × [0, 1] ∪ (3-ball) (S : closed orientable surface) by attaching“ 1-handles ”to S × {1} ∪ ∂3-ball. Notation: ∂− C := S × {0}, ∂+ C := ∂C \ ∂− C, The genus of ∂+ C is called the genus of the compression-body C. A compression body C is called a handlebody if ∂− C = ∅. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 4 / 41 Heegaard splittings Heegaard splittings M : compact orientable 3-manifold C1 ∪Σ C2 : (genus-g) Heegaard splitting of M (Σ: Heegaard surface) if Ci : genus-g compression-body (⊂ M) s.t.C1 ∪ C2 = M, and C1 ∩ C2 = ∂+ C1 = ∂+ C2 = Σ. It is known that any M admits a Heegaard splitting. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 5 / 41 An alternate expression of H.S. Σ = Σg : orientable surface of genus g ≥ 2. MCG(Σ): Mapping class group of Σ: the group of automorphisms of Σ. An alternate expression of the H.S. C1 ∪Σ C2 is: C1 ∪Σ C2 = C1 ∪h C2 , where h : Σ(= ∂+ C1 ) → Σ(= ∂+ C2 ) ∈ MCG(Σ) Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 6 / 41 Curve complex The curve complex was introduced by Harvey and has been used to prove many deep results about the mapping class group ( [Harv]). [Harv] W. J. Harvey. Boundary structure of the modular group. In I. Kra and B. Maskit, editors, Riemann Surfaces and Related Topics: Proceedings of the 1978 Stony Brook Conference, volume 97 of Ann. of Math. Stud. Princeton, Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 7 / 41 Curve complex The curve complex C(Σ) is a simplicial complex defined as follows: 0-simplex ↔ (isotopy class of) an essential simple closed curve on Σ, two 0-simplices are joined by a 1-simplex if they can be realized by disjoint curves. n-simplex ↔ disjoint non-parallel n + 1 simple closed curves on Σ. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 8 / 41 Distance in the curve complex For vertices a and b (∈ C(Σ)(0) ), d(a, b) = dC(Σ) (a, b) := (the smallest number of 1-simplexes in a path connecting a and b in C(Σ)). A shortest path [a0 , a1 , a2 , . . . , an ] connecting two vertices a0 and an is called a geodesic. For sets A and B(⊂ C(Σ)(0) ), d(A, B) = dC(Σ) (A, B) := min{d(a, b)|a ∈ A, b ∈ B}. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of Low-dimensi October 27, 2014 9 / 41 Let: M: compact orientable 3-manifold C1 ∪Σ C2 : Heegaard splitting of M C(Σ): curve complex of Σ D(Ci ) (⊂ C(Σ)(0) ): disk complex of Ci , (i.e. a ∈ D(Ci ) ⇔ a bounds a disk in Ci .) Definition (Distance of Heegaard splitting) d(C1 ∪Σ C2 ) (:= dC(Σ) (D(C1 ), D(C2 )) ) = min{d(a, b)|a ∈ D(C1 ), b ∈ D(C2 )} John Hempel, 3-manifolds as viewed from the curve complex, Topology 40 (2001) 631-657 Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 10 / 41 Note: h ∈ MCG(Σ) induces a bijection on C(Σ)(0) , and it naturally extends to an isometry on C(Σ) denoted by h̃. Theorem(Hempel) ∃g: Σ → Σ s.t. d(C1 ∪gn C2 ) → ∞ (n → ∞) Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 11 / 41 Part 2 Historical Background and Main Result Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 12 / 41 Historical background Masur-Minsky: C(Σ) is δ-hyperbolic. (Roughly speaking: C(Σ) is quasi-equivalent to hyperbolic space.) [MM1] H. Masur and Y. Minsky. Geometry of the complex of curves I: hyperbolicity, Invent. Math. 138 (1999), 103–149. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 13 / 41 Historical background Note: h ∈ MCG(Σ) induces a bijection on C(Σ)(0) , and it naturally extends to an isometry on C(Σ) denoted by h̃. Expected: If h ∈ MCG(Σ) is pseudo-Anosov, then, h̃ behaves like a hyperbolic isometry on a hyperbolic space. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 14 / 41 Historical background This is justified by Abrams-Schleimer to show: Theorem Suppose h ∈ MCG(Σ): pseudo-Anosov with certain technical property. Then: d(V1 ∪hn V2 ) has quasi-linear growth with respect to n. A.Abrams, and S.Schleimer, Distance of Heegaard splittings, Geometry & Topology, 9(2005), 95–119. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 15 / 41 Historical background Summary Geometry of C(Σ) up to quasi-equivalence (Coarse geometry of C(Σ)) is similar to hyperbolic geometry (and, this intuition works very well). However, Theorem (Birman-Menasco) ∃geodesic segment γ in C(Σ) which cannot be extended to a longer geodesic segment. cf. Any geodesic segment in Hn is infinitely extendable. Birman-Menasco, The curve complex has dead ends, Geometry Dedicata, Published online: 07 April, 2014. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 16 / 41 Historical background This fact represents a difference between Coarse Geometry and (Fine) Geometry of C(Σ). Motivation It may be interesting to study Fine Geometry of C(Σ) (like the paper of Birman-Menasco) particularly from the viewpoint of Heegaard Theory. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 17 / 41 keen Heegaard splitting Definition (keen Heegaard splitting) C1 ∪Σ C2 is keen. ⇐⇒ d(C1 ∪Σ C2 ) is realized by unique pair of Tsuyoshi Kobayashi ( Nara Women’s Univ. def. elements of D(C1 ) and D(C2 ). Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 18 / 41 Remarks Let C1 ∪Σ C2 : a Heegaard splitting. Let [a0 , a1 , . . . , an ]: a path in C(Σ) realizing d(C1 ∪Σ C2 ) = n. In this talk ai (∈ C(Σ)(0) ) often stands for geometric representative of ai . Remark 1 If a0 (or an ) is separating in Σ, then C1 ∪Σ C2 is not keen. [a′0 , a1 , . . . , an ] is also a path realizing d(C1 ∪Σ C2 ). Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 19 / 41 Remarks Remark 2 If ∃a∗ ∈ C(Σ)(0) s.t. |a∗ ∩| a0 | = 1, a∗ ∩ a1 = ∅, then C1 ∪Σ C2 is not keen. [a′0 , a1 , . . . , an ] is also a path realizing d(C1 ∪Σ C2 ). Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 20 / 41 Remarks Remark 1. If a0 (or an ) is separating in Σ, then C1 ∪Σ C2 is not keen. Remark 2. If ∃a∗ ∈ C(Σ)(0) s.t. |a∗ ∩| a0 | = 1, a∗ ∩ a1 = ∅, then C1 ∪Σ C2 is not keen. Hence, If C1 ∪Σ C2 is keen, then: • a0 : non-separating • a1 : non-separating • a0 ∪ a1 : separating This implies: If g(Σ) = 2, then C1 ∪Σ C2 is not keen. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 21 / 41 Main Result Theorem (Ido-Jang-K) ∀g ≥ 3, ∀n ≥ 2, ∃M: closed 3-manifold with a keen H.S. C1 ∪Σ C2 s.t. g(Σ) = g, d(C1 ∪Σ C2 ) = n. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 22 / 41 Technical Background Part 3 Technical Background Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 23 / 41 Technical Background: Subsurface projection X: essential non-simple subsurface of Σ. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 24 / 41 Technical Background: Subsurface projection X : essential non-simple subsurface of Σ, (0) 2C(X) : the power set of C(X)(0) . (0) Subsurface projection πX : C(Σ)(0) → 2C(X) is defined by Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 25 / 41 Technical Background: Subsurface projection Lemma 1 (Masur-Minsky’00) πX is 2-Lipschitz, hence, ∀[a0 , a1 , . . . , am ] a geodesic in C(Σ) such that ai ∩ X , ∅ (i = 0, 1, . . . , m), diamC(X) (πX (a0 ) ∪ · · · ∪ πX (am )) ≤ 2m. Lemma 2 (Masur-Minsky) Let C be a compression-body. If every essential disk in C intersects ∂X (+ technical conditions), then diamC(X) (πX (D(C))) < ∞. Remark Tao Li gave a sharp upper bound: diamC(X) (πX (D(C))) ≤ 12. T. Li, Images of the disk complex , Geom. Dedicata. 158 (2012), 121-136. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 26 / 41 Technical Background: Subsurface projection Lemma 3 (Masur-Minsky) Let Σ be as above. ∃c > 0, s.t. for ∀h: Σ → Σ: pseudo-Anosov, ∀γ ∈ C(Σ)(0) , ∀n ∈ Z, dC(Σ) (h̃n (γ), γ) ≥ c|n|. H. Masur and Y. Minsky. Geometry of the complex of curves I: hyperbolicity, Invent. Math. 138 (1999), 103–149. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 27 / 41 Warmup 1: Extending geodesics Proposition 1 Let [a0 , . . . , an−2 ], [an−2 , an−1 , an ] (n ≥ 4, even) be geodesics in C(Σ), where an−2 : non-separating in Σ. Suppose diamC(Xn−2 ) (πXn−2 (an−4 ) ∪ πXn−2 (an )) ≥ 4n, where Xn−2 := Cl(Σ \ N(an−2 )). Then [a0 , a1 , . . . , an ] is a geodesic in C(Σ). Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 28 / 41 Outline of proof of Proposition Suppose, for a contradiction, ∃[b0 , b1 , . . . , bm ] s.t. m < n, b0 = a0 , bm = an . Claim ∃j s.t. bj = an−2 Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 29 / 41 Claim ∃j s.t. bj = an−2 proof. If not, we can apply Lemma 1. Lemma 1 πX is 2-Lipschitz, hence, ∀[a0 , a1 , . . . , am ] a geodesic in C(Σ) such that ai ∩ X , ∅ (i = 0, 1, . . . , m), diamC(X) (πX (a0 ) ∪ · · · ∪ πX (am )) ≤ 2m. In fact: If bj ∩ Xn−2 = ∅, then bj = an−2 . ⇒ diamC(Xn−2 ) (πXn−2 ([b0 , b1 , . . . , bm ])) ≤ 2m < 2n. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 30 / 41 diamC(Xn−2 ) (πXn−2 ([b0 , b1 , . . . , bm ])) < 2n. Contradicting the assumption diamC(Xn−2 ) (πXn−2 (an−4 ) ∪ πXn−2 (an )) > 4n. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 31 / 41 By Claim, ⇒ j = n − 2, m − j = 2 ⇒ m = n: a contradiction. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 32 / 41 Existence of [an−2, an−1, an] Let [a0 , a1 , . . . , an−2 ]: geodesic in C(Σ), with an−2 : non-separating in Σ. The existence of a geodesic [an−2 , an−1 , an ] as in Proposition 1 is given as follows. Let a∗n−1 = an−3 , a∗n = an−4 . Take h(∈ MCG(Xn−2 )) : a p.A. By Lemma 3, ∃p(> 0) s.t. diamC(Xn−2 ) (πXn−2 (an−4 ) ∪ h̃p (πXn−2 (an−4 ))) ≥ 4n. Then, let an−1 = h̃p (an−3 ), an = h̃p (an−4 ). Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 33 / 41 Warmup 2: Heegaard splitting with distance n Ido-Jang-K., Heegaard splitting of distance exactly n, AGT 14 (2014), 1395-1411. Theorem ∀n ≥ 2, ∀g ≥ 2, ∃C1 ∪Σ C2 : Heegaard splitting of a closed 3-mfd. M s.t. g(Σ) = g, d(C1 ∪Σ C2 ) = n. (For simplicity, we suppose n ≥ 4, even.) Let Ci (i = 1, 2) be a genus g compression body with ∂− Ci ≃ Σg−1 . Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 34 / 41 Let Σ = ∂+ C1 . [a0 , a1 , . . . , an+2 ]: a geodesic in C(Σ) constructed in Warmup 1 with We identify a0 with ∂D1 . Take a homeomorphism f : ∂+ C1 → ∂+ C2 s.t. f (an+2 ) = ∂D2 . By analyzing D(Ci ), we can show that d(C1 ∪f C2 ) = n. Tsuyoshi Kobayashi ( Nara Women’s Univ. We attach a genus g − 1 handlebody. H to close the boundary component ∂− C1 , in a “sufficiently complicated” manner. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 35 / 41 “sufficient complicated”? Notation h : ∂H → ∂− C1 : homeomorphism, X = the component of cl(Σ \ N(a1 )) πX : C(Σ)(0) → 2C(X) : subsurface projection, P : C(X)(0) → C(∂− C1 )(0) : natural projection. (0) Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 36 / 41 “sufficient complicated”? By Lemma 2, we can show diamC(∂− C1 ) (P ◦ πX (f −1 (D(C2 )))) is bounded (in fact, ≤ 12 by T. Li.). By Hempel, we may take h so that: dC(∂− C1 ) (P ◦ πX (f −1 (D(C2 ))), h(D(H))) > 2n (: sufficiently complicated h.) Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 37 / 41 Note (C1 ∪h H) ∪f C2 is a Heegaard splitting. Proposition 2 d((C1 ∪h H) ∪f C2 ) = n Proof. Suppose, for a contradiction, d((C1 ∪h H) ∪f C2 ) = m ≤ n − 1. Take a ∈ D(C1 ∪h H), b ∈ D(C1 ), d(a, b) = m, and [a = a0 , a1 , . . . , am = b] a geodesic. By Lemma 1, diamC(X) (πX (a) ∪ πX (b)) ≤ 2m ≤ 2(n − 1). This implies: diamC(∂− C1 ) (P(πX (a))∪P(πX (b))) ≤ 2(n−1). Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 38 / 41 By using elementary cut and paste (outermost disk) arguments, we can show: d(P(πX (a)), h(D(H))) ≤ 2. These imply: d(P ◦ πX (f −1 (D(C2 ))), h(D(H))) ≤ 2(n − 1) + 2 = 2n. : a contradiction. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 39 / 41 Idea of proof of Main Result Part 4 Idea of proof of Main Result Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 40 / 41 Idea of the proof of Main Result Theorem (Ido-Jang-K) ∀g ≥ 3, ∀n ≥ 2, M: closed 3-mfd. with a keen H.S. C1 ∪Σ C2 s.t. g(Σ) = g, d(C1 ∪Σ C2 ) = n. C1 ∪Σ C2 : a genus g Heegaard splitting of a closed 3- manifold M. [a0 , a1 , . . . , an−2 ]: a geodesic in C(Σ) with a0 ∈ D(C1 ), an ∈ D(C2 ). Suppose: 1. 2. ∀a∗ ∈ D(C1 ) \ {a0 }, a1 ∩ a∗ , ∅, ∀b∗ ∈ D(C2 ) \ {an }, an−1 ∩ b∗ , ∅. 3. ∃k(∈ {2, . . . , n − 2}) s.t. ak : non-sep. with diamC(X1 ) (πX1 (a0 ) ∪ πX1 (ak )) ≥ 4n + 16, diamC(Xn−1 ) (πXn−1 (ak ) ∪ πXn−1 (an )) ≥ 4n + 16. Then, C1 ∪Σ C2 : keen. Tsuyoshi Kobayashi ( Nara Women’s Univ. Topology On keen Heegaard andsplitting Geometry of October Low-dimensi 27, 2014 41 / 41