QUANTUM HYPERBOLIC INVARIANTS Stéphane Baseilhac Université de Montpellier Nara, October 2014 1/26 The work presented here comes from collaborations with Benedetti, in particular: S. B., R. Benedetti, Analytic families of quantum hyperbolic invariants, to appear in Algebraic & Geometric Topology. 2/26 We are going to discuss some of the mathematics behind the volume conjecture (Kashaev ’97, Murakami-Murakami ’99): Volume Conjecture: If K is a hyperbolic knot in S 3 , then log |JN (K )(e N→∞ N lim 2πi N )| = 1 Vol(S 3 \ K ) 2π Here JN (K ) ∈ Z[q ± ] is the normalized N-th colored Jones polynomial of the knot K , such that JN (unknot) = 1. 3/26 Here are some motivations: 4/26 I Unfold the asymptotical geometric content of quantum invariants; the vol. conjectures are predictions for 3-dim. quantum gravity. I Clarify the “quantization processes” lying in between. I Find new interplays between algebra and geometry: new QFTs, new mapping class group representations, etc. 5/26 Rather than the Jones polynomials, we will consider the quantum hyperbolic invariants HN (M) (QHI), and the following version of the volume conjecture for the QHI: For any one-cusped hyperbolic manifold M there exist sequences of points xN ∈ XN such that HN (M)(xN ) ∼N→+∞ exp( N (Vol(M) + iπ 2 CS(M))). 2π Here, for each odd N ≥ 3, HN (M) : XN → C/µN is a rational function with complex values modulo the N-th roots of 1, and XN a determined finite cover of the “geometric” compt of the variety of PSL(2, C)-characters of M. Since the quantum invariants are defined combinatorially, we will build in a similar way spaces and maps fitting in a diagram like: XON X∞ HN (M), rational / C/µN S(M), analytic /C where again we have covering maps X∞ /Z2 / XN /(Z/NZ)2 /X over the geometric component X of the variety of PSL(2, C) characters of M, and S(M) is the Chern-Simons function. 6/26 Next talk : The quantum hyperbolic invariants; it requires a combinatorial definition of the Chern-Simons function S(M). This talk: The combinatorial definition of S(M), and the modifications required by the quantum setup. 1. Gluing varieties 2. The coverings 3. The Chern-Simons function 7/26 1. Gluing varieties 2. The coverings 3. The Chern-Simons function 8/26 Denote by X the geometric component of augmented PSL(2, C) valued characters of M, and X (∂ M̄) the character variety of ∂ M̄. The restriction map res : X → X (∂ M̄) is regular. Theorem (Dunfield) The map res : X → X (∂ M̄) is birational onto its image. Fixing a cusp basis (l, m), denote the induced map and image by h : X → C∗ × C∗ A := h(X ) 9/26 10/26 Let T be an ideal triangulation of M without null-homotopic edges. Then: I The gluing variety G (T ) is non empty (Segerman-Tilmann); I G (T ) is a complex curve (Neumann-Zagier); I There exists a point zhyp ∈ G (T ) with d.f. holonomy ρhyp . A question: Is Dunfield’s theorem true by replacing the geometric compt X of the character variety with some component of G (T ) ? Some issues: I zhyp may not be a regular point of G (T ); hence it may be contained in several components. I Dunfield’s proof uses the volume rigidity for closed hyperbolic Dehn fillings of M, and the variation formula of the volume function of characters. 10/26 Let T be an ideal triangulation of M without null-homotopic edges. Then: I The gluing variety G (T ) is non empty (Segerman-Tilmann); I G (T ) is a complex curve (Neumann-Zagier); I There exists a point zhyp ∈ G (T ) with d.f. holonomy ρhyp . A question: Is Dunfield’s theorem true by replacing the geometric compt X of the character variety with some component of G (T ) ? Some issues: I zhyp may not be a regular point of G (T ); hence it may be contained in several components. I Dunfield’s proof uses the volume rigidity for closed hyperbolic Dehn fillings of M, and the variation formula of the volume function of characters. Definition An irreducible component of the gluing variety G (T ) is rich if it contains the point zhyp and also an infinite sequence of closed hyperbolic Dehn fillings of M with shape parameters zn converging to zhyp in G (T ). Proposition (Petronio-Porti) For any ideal triangulation T of M such that zhyp has coordinates with non negative imaginary part, the gluing variety G (T ) has rich components. In particular, the max subdivisions of the EP cellulation of M have rich components. Using the proposition and arguments similar to Dunfield’s one can prove: Corollary For any rich component Z of a gluing variety G (T ) of M, the (regular) map hZ : Z 11/26 holonomy /X h / A is birational. Definition 11/26 An irreducible component of the gluing variety G (T ) is rich if it contains the point zhyp and also an infinite sequence of closed hyperbolic Dehn fillings of M with shape parameters zn converging to zhyp in G (T ). Proposition (Petronio-Porti) For any ideal triangulation T of M such that zhyp has coordinates with non negative imaginary part, the gluing variety G (T ) has rich components. In particular, the max subdivisions of the EP cellulation of M have rich components. Using the proposition and arguments similar to Dunfield’s one can prove: Corollary For any rich component Z of a gluing variety G (T ) of M, the (regular) map hZ : Z holonomy /X h / A is birational. 1. Gluing varieties 2. The coverings 3. The Chern-Simons function 12/26 Now we are going to complete a square: ?? Z hZ ,∞ hZ / A∞ /Z2 /A where as before A := h(X ) is the image of the geometric cpt under the restriction map, hZ is the map we have just defined, and we set A∞ := {(u, v ) ∈ C2 | (e u , e v ) ∈ A}. We need to take logarithms of the coordinates of points z ∈ Z . 13/26 14/26 Define the analytic set (s is the number of tetrahedra of T ) Z∞ ⊂ C3s by requiring that a point (l01 , l11 , l21 , . . . , l0s , l1s , l2s ) ∈ Z∞ satisfies: I I in tetrahedra: ∀ j ∈ {1, . . . , s}, l0j + l1j + l2j = 0, and for all r ∈ {0, 1, 2}, exp(lrj ) = ±zrj for some point z = (zrj ) ∈ Z ; P about edges: ∀ E ∈ E (T ), j,r lrj (E ) = 0. If non empty, Z∞ is a space of logarithms of ± shape parameters in Z . Note that the edge equations are the same for all edges. Any point l = (lrj ) ∈ Z∞ determines: I a holonomy ρ(l) ∈ X (note that the tetrahedral relations allow one to determine z = (zrj ) from l); I a class γ(l) ∈ H 1 (∂ M̄; C), defined on a loop transverse to the cusp triangulation by summing the coordinates lrj viewed from the loop, with signs according to the orientation (consistency follows from by the tetrahedral and edge relations). I a class γ2 (l) ∈ H 1 (∂ M̄; Z/2), defined similarly, using normal loops in T and taking the sum mod(2) of the variables (lrj − log(zrj ))/πi attached to the edges we face along the loops. 15/26 16/26 Denote by dz ∈ H 1 (∂ M̄; C/2iπZ) the log of the linear part of the restriction of ρ(l) to π1 (∂ M̄). For all a ∈ H1 (∂ M̄; Z) we have the compatibility relations: γ(l)(a) = dz (a) mod(iπ) (γ(l)(a) − dz (a))/iπ = i ∗ (γ2 (l))(a) mod(2). where i ∗ : H 1 (M̄; Z/2Z) → H 1 (∂ M̄; Z/2Z) is induced by inclusion. Below we will assume that γ2 (l) = 0 (for simplicity). Recall that A∞ := {(u, v ) ∈ C2 | (e u , e v ) ∈ A}. Then we can put hZ ,∞ : Z∞ −→ A∞ l 7−→ (γ(l)(l), γ(l)(m)) Similarly define the algebraic set ZN ⊂ C3s by requiring that (w01 , w11 , w21 , . . . , w0s , w1s , w2s ) ∈ ZN satisfies: I I N−1 in tetrahedra: ∀ j ∈ {1, . . . , s}, w01 w11 w21 = −ζ 2 , and for all r ∈ {0, 1, 2}, (wrj )N = ±zrj for some point z = (zrj ) ∈ Z ; Q about edges: ∀ E ∈ E (T ), j,r wrj (E ) = ζ −1 . If non empty, ZN is a space of N-th roots of the shape parameters in Z . Note that the edge equations are the same for all edges. 17/26 18/26 Again, any point w = (wrj ) ∈ ZN determines: I a holonomy ρ(w) ∈ X ; I a class γN (w) ∈ H 1 (∂ M̄; C∗ ); I a class γN,2 (w) ∈ H 1 (∂ M̄; {±1}). Below we will assume that γN,2 (w) = 1 (for simplicity). Then, setting AN := {(u, v ) ∈ C2 | (u N , v N ) ∈ A} we can put hZ ,N : ZN w −→ AN 7−→ (γN (w)(l), γN (w)(m)). We deduce a commutative diagram Z∞ hZ ,∞ classical / A∞ 1 e N (·) ZN (·)N Z hZ ,N /(NZ)2 quantum / AN hZ /(Z/NZ)2 /A We call hZ ,∞ (l) and hZ ,N (w) the weights of l and w. But do we have (Z 6= ∅ ⇒ Z∞ , ZN 6= ∅) ? What can be said about the image of hZ ,∞ and hZ ,N ? 19/26 Theorem BIG (follows from Neumann’s work) We have: (1) The map hZ ,∞ : Z∞ → A∞ maps onto a dense open subset (no lift is missed). (2) The fibers of the covering Z∞ → Z are affine spaces over an abelian group C that fits in an exact sequence (γ,γ2 ) 0 → Zn(edges) → C −→ H 1 (∂ M̄; Z) ⊕ H 1 (M̄; Z/2Z) r −i ∗ −→ H 1 (∂V ; Z/2Z) → 0 (3) The isomorphism γ : C ⊗ Q → H 1 (∂ M̄; Q) is symplectic with respect to Neumann-Zagier’s 2-form and the intersection product. Rq. Given any two points l, l0 ∈ Z∞ , (γ(l) − γ(l0 ), γ2 (l) − γ2 (l0 )) lies in Ker(r − i ∗ ). This helps proving (1) from (2). 20/26 1. Gluing varieties 2. The coverings 3. The Chern-Simons function 21/26 In 1992, Neumann introduced a function (a sum of dilogarithms) that can be written as H1 : Z∞ −→ C/2πiZ. He showed: Theorem (Neumann) Let Y be M or a closed hyperbolic Dehn filling M 0 of M, and ρY the hyperbolic holonomy of Y . Then: (1) H1 (l) is constant on points l ∈ Z∞ with holonomy ρ(l) = ρY and weight hZ ,∞ (l) = 0. (2) For such an l we have H1 (l) = exp π2 Vol(Y ) + 2πiCS(Y ) . 22/26 23/26 The proof is homological: Neumann produced an isomorphism b H3 (BPSL(2, C); Z) ∼ = B(C) with an extended Bloch group of scissors congruences in H3 , and used Dupont’s 3-cochain representing p̂1 in terms of dilogarithms. Another, direct proof (knowing the answer), is due to J. Marché; he computed the Chern-Simons functional of flat connections on permutohedra with “nice” representatives of the connections. For any compact oriented 3-manifold Y , Chern-Simons theory with gauge group PSL(2, C) defines: I the Chern-Simons bundle L∂Y → X (∂Y ), which is a C∗ -bundle with connection and inner product; I the parallel Chern-Simons section sY : X (Y ) → res ∗ L∂Y , where res : X (Y ) → X (∂Y ) is the restriction map. When Y is closed, sY is just a function, and if Y is hyperbolic, then Yoshida and Dupont proved independently that 2 Vol(Y ) + 2πiCS(Y ) sY (ρY ) = exp π 24/26 25/26 If Y = V ∪ϕ (D 2 × S 1 ) is a closed hyperbolic Dehn filling of M with holonomy ρY close to ρM in X , then Kirk-Klassen proved: I Fixing the gauge on i ∗ L∂ M̄ by taking as local coordinates on X the standard logarithms log(λ) and log(µ), we have 2 Vol(M) + 2πiCS(M) . sM̄ (ρhyp ) = exp π I If the Dehn filling instruction maps the meridian ∂(D 2 × ∗) to l q mp and the longitude ∗ × S 1 to l s mr (so ps − qr = 1), then sY (ρY ) = sM̄ (ρhyp ) ! Z ρY 1 × exp − (log(λ)d log(µ) − log(µ)d log(λ)) 2πi ρhyp × exp (−(s log(λ) + r log(µ))) . 26/26 Putting together all these results gives a characterization of sM̄ as a sol. of a differential problem. Using Neumann’s BIG Theorem, one can deduce: Proposition The function S(M) := exp(H1 (·)) : Z∞ → C∗ factors through hZ ,∞ to define a function on A∞ , and it descends to a parallel section s(M) of a flat trivial C∗ -bundle L(M) → A with canonical connection 1-form and inner product, such that h∗ (L(M), s(M)) ∼ = (i ∗ L∂ M̄ , sM̄ ).