Talk II : the quantum hyperbolic invariants. Plan of the talk : 1. QH State sums over weakly branched triangulations 2. Some relations with other quantum invariants 3. Examples 1/34 1. QH State sums over weakly branched triangulations 2. Some relations with other quantum invariants 3. Examples 2/34 The quantum hyperbolic invariants of a one-cusped manifold M are defined by state sums HN (T , b) : ZN → C/µN over weakly branched ideal triangulations (T , b) of M carrying the spaces ZN defined previously. The state sums HN (T , b) are non-commutative analogues of the function H1 : Z∞ → C∗ that defined the Chern-Simons function S(M) in Talk 1. 3/34 Definition A 3-dim. pseudo-manifold triangulation is pre-branched if each 2-face is co-oriented and two co-orientations point inwards and two outwards each tetrahedron. The triangulation is weakly branched if its tetrahedra are branched and have compatible pre-branchings. A B B A Figure: A pre-branched tetrahedron with its square edges oriented. 4/34 A B Figure: Branched tetrahedra inducing the same pre-branching. Positive orientation on the left (“∗b = +1”), negative on the right (“∗b = −1”). 5/34 Simple facts: 1. Any triangulation has some pre-branchings, hence also weak branchings. 2. The pre-branchings correspond to solutions of the gluing eqs of the form (1, 1, −1) on each tetrahedron (‘Z/2-taut angle structures”). 2 0 3 1 Figure: Graph encoding of a branched tetrahedron with ∗b = +1. 6/34 7/34 A graph representing a weak branching on a triangulation T of the “figure eight sister” cusped manifold: 2 1 NB. There is no system of edge orientations on T (a “branching”) that define a branching by restriction to each tetrahedron of T . NB. The gluing maps between two 2-faces is given by an even permutation. This gives a color r ∈ Z/3Z for each 2-face of T . Take a weakly branched triangulation (T , b) of M. Assign copies V0 , . . . , V3 of CN to the faces of the (branched) tetra as follows: 1 0 2 0 2 3 3 1 −1 +1 Then, given an automorphism R ∈ GL(CN ⊗ CN ) and a branched tetrahedron ∆, set ( i,j (Rk,l ) : V3 ⊗ V1 → V2 ⊗ V0 if ∗b = +1 R(∆) = k,l (R̄i,j ) : V2 ⊗ V0 → V3 ⊗ V1 if ∗b = −1. Similarly, given a 3-periodic automorphism Q ∈ GL(CN ), assign Qr (e) to each (co-oriented) 2-face of T with color r (e) ∈ Z/3. 8/34 Definition of the state sums Let M be a one-cusped manifold, and (T , b) a weakly branched triangulation of M such that its gluing variety G (T ) has a rich component Z , as defined in Talk 1. Recall the space ZN ⊂ C3s of Nth-roots of the shape parameters in Z . A point w ∈ ZN has coordinates w = (w01 , w11 , w21 , . . . , w0s , w1s , w2s ) | {z } | {z } w1 ws where wj , j ∈ {1, . . . , s}, is associated to the tetrahedron ∆j of T . 9/34 For any point w = (wj )j ∈ ZN we set: HN (T , b)(w) := s X Y σ j=1 RN (∆j )(wj )σ Y r (e) (QN )σ faces e where: I for every j and every face e, RN (·)(wj ) ∈ GL(CN ⊗ CN ) and r (e) QN ∈ GL(CN ) are associated to ∆j and e as we explained. I σ runs over all maps {2−faces of T } → {0, 1, . . . , N − 1}; I (. . .)σ means that we take the entry of the tensor (...) which is determined by σ. RN (·)(·) is a tensor valued function of wj = (w01 , w11 , w21 ) called matrix dilogarithm; its behaviour under tetrahedral symmetries determine the matrix QN . 10/34 11/34 The main things to know about RN are: I I RN is derived from the 6j-symbols of the cyclic reps of Uq sl2 (first computed by Kashaev); b N −→ Aut(CN ⊗ CN ), RN eventually reduces to a rat. map C b N is the Riemann surface of the Nth root map where C C \ {0, 1} −→ C2 , z 7−→ (e log(z) N ,e j j k Figure: R(e log(z)+πip N ,e l − log(1−z)+πiq N i ). k i *b = 1 − log(1−z) N *b = −1 l ), where z := w0 and 1/(1 − z) = w1 . I Set l0,N := log(z)+πip , l1,N N RN (e l0,N , e l1,N )i,j k,i+j ∝ ϕ(e l0,N ,e l1,N := − log(1−z)+πiq . N We have √ √ √ 2 −1π 2 −1π 2 −1π i, j, k ) N N N where N ϕ(e l0,N , e l1,N )(u, v , t) = e 2iπ ((t−u)l1,N +vt+t 2 /2) × Sπ/N (−i(l0,N + u − t) + (1 − N)π/N) and... 12/34 ...Sπ/N (. . .) is a value of the meromorphic function Z 1 e zx dx . Sγ (z) = exp 4 R+i0 x sinh(πx) sinh(γx) where 0 < arg γ < π/2 and |Re(z)| < π + |Re(γ)| (Barnes’ double sine 1900, Faddeev ’95). Write Sγ = exp(Iγ (z)). One has a uniform asymptotic expansion on compact subsets of |Re(z)| < π: ∞ 1 1 X 2n−1 22n B2n (1/2) 2n iz Iγ (z) = Li2 (−e )+ γ ∂z Li2 (−e iz ) 2iγ 2i (2n)! n=1 as γ → 0. 13/34 1 NB. By the Jacobi triple product formula, Sγ is a multiplier 2 : Θ(y , τ ) = ξγ Sγ (z)Sγ (−z)Θ(y /τ, −1/τ ) where Θ(y , τ ) := 14/34 P∞ n=0 q n2 x n ; , ξγ ∈ C. I RN satisfies fundamental “5-term” identities called pentagon relations, and RN (∆) has some tetrahedral symmetries. Here is a graph representing a pentagon relation between matrix dilogarithms (NB: it is a non Abelian 3-cocycle relation): x3 y = x1 x2 x Figure: x1 = y /x, x2 = y (1 − x)/x(1 − y ), x3 = (1 − x)/(1 − y ), where 0 < y > x < 1. Each of the arguments x, y , x1 , x2 and x3 stands for the shape parameter z0j of the corresponding tetrahedron. The products of N-th roots wrj are required to be equal along the edges on both sides. 15/34 Any such identity corresponds to a move (T , b, w) → (T 0 , b 0 , w0 ) which preserves the branchings locally, between weakly branched triangulations having their tetrahedra equipped with the edge variables wrj . The above identity corresponds to: (1-x)/(1-y) y x y(1-x)/x(1-y) 16/34 y/x Recall that: I any point w in ZN defines a holonomy ρ(w) ∈ X and a weight hZ ,N (w) ∈ H 1 (∂ M̄; C∗ ); I the set of such weights is a Zariski open subset of the covering AN of the A-polynomial curve A = h(Z ). Theorem (B-Benedetti 2012) The state sum HN (T , b)(w) is an invariant of (M, ρ(w), hZ ,N (w)) up to multiplication by N-th roots of 1. By varying w in ZN , these invariants define rational functions HN (M) : AN → C/µN . 17/34 Strategy of the proof: I The state sums are invariant under any change of the weak branching (⇒ one needs a “calculus” for such objects). I One can define moves between triangulations of the form (T , b, w), which preserve holonomy and weight, and are compatible with the pre-branching. I The state sums are invariant under such moves. I Any two triangulations of M of the form (T , b, w) and having the same holonomy and weight are related by such moves. A similar proof applies to show the invariance of H1 mentionned at the end of Talk 1. 18/34 Example of invariance under moves preserving the pre-branching: (0321) (01) (12) (12) (12) (01) (12) (01) (0123) (01) (12) (01) (12) 19/34 (01) 1. QH State sums over weakly branched triangulations 2. Some relations with other quantum invariants 3. Examples 20/34 Relation with quantum Teichmüller theory The following papers imply that the QHI of mapping cylinders of punctured surfaces contain the (finite dim.) q-Teichmüller theory: 21/34 I F. Bonahon, X. Liu, Representations of the quantum Teichmüller space and invariants of surface diffeomorphisms, Geom. Topol. 11 (2007) 889–937 I H. Bai, Quantum Teichmüller spaces and Kashaev’s 6j-symbols, Alg. Geom. Topol. 7 (2007) 1541–1560 I H. Bai, F. Bonahon, X. Liu, Local representations of the quantum Teichmüller space, ArXiv:math.GT/?? (2007) I J. Toulisse, Irreducible decomposition for local representations of the quantum Teichmüller spaces, ArXiv:math.GT/14044938 Relations with invariants of closed manifolds 22/34 The QHI are defined also for links in arbitrary compact oriented 3-mfds with PSL(2, C)-characters. In the case of S 3 we have: Theorem (B-Benedetti 2011). For every link L in S 3 and every odd integer N ≥ 3 we have HN (S 3 , L, ρtriv , 0) ≡N JN (L)(e 2iπ/N ) where JN (L) is the normalized N-th colored Jones polynomial of L such that JN (unknot) = 1, ρtriv is the trivial holonomy, and 0 is the null weight. Here ≡N denotes equality up to multiplication by N-th roots of 1. 23/34 Similarly to the behaviour of Vol + iCS we have: Theorem (B-Benedetti 2007) For any sequence of closed hyp. Dehn fillings Mn of M with surgery cores Ln and holonomies ρn converging to ρhyp in the character variety, we have lim HN (Mn , Ln , ρn , 0) ≡N HN (M)(ρhyp , 0) n→∞ where 0 denotes the 0 weights. 1. QH State sums over weakly branched triangulations 2. Some relations with other quantum invariants 3. Examples 24/34 25/34 Summarizing, in order to compute HN (M) one needs: I an ideal triangulation T of M having a solution zhyp ∈ G (T ) of the gluing eqs with d.f. holonomy and coordinates with non-negative imaginary part (for example a subdivision of the EP cellulation); this provides rich components Z of G (T ). I Explicit formulas for the solutions w ∈ Z of the gluing eqs, and for their N-th roots w ∈ ZN . This is essentially provided by Snappea, snap, and Zickert’s algo. Moreover one needs: I a weak-branching b of T (immediate); I the state sum formula of HN (M) on (T , b) (immediate); I a very good computer for large values of N ! Numerical computations show that: I The invariants depend on the characters and the weights. I This dependence persists as N → +∞. I For some weights the invariants grow exponentially with N, and yield instances of the volume conjecture, and for some other weights they do not. I There are symmetries of the invariants at the hyperbolic holonomy, which disappear when moving away from it. I For knot complements, the absolute values of the invariants do not depend on the weights of the meridian as N → ∞. In practice, any example offers a whole world to explore ! 26/34 Example 1: 27/34 M = complement of the figure eight knot. We take the canonical triangulation. It has a branching. We fix one, and find the state sum expression (up to a scalar which is negligeable as N → +∞): HN (M)(u, v) ∝ N−1 X ζβ 2 −α2 α,β=0 ω(u0 , u−1 1 |N − β) ω(v0 /ζ, v1−1 |N − α) Here u := (u0 , u1 , u2 ) and v := (v0 , v1 , v2 ) are the triangulation’s 2πi N-th roots of shape parameters, and we set ζ := e N and ω(x, y |n) := n Y j=1 y (1 − xζ j )−1 . Take (u, v) with d.f. holonomy and longitude weight as in the left πi column, and deform it by u0 (A, B) := e 3 + A + iB as in the top row. You get the values of |H15 (M)(u, v)| listed below: 28/34 (0, 0) (1, 0) (1, 0.5) (1, 1) (A, B)(2,3) −8πi 2.5587... 2.6504... 2.0018... 1.6118.... 5.0307.... −6πi 58.5466... 58.3761... 47.0533... 39.9892... 95.0326... −4πi 2.1356... 2.0279... 1.9058... 1.8138... 2.7491... −2πi 77.4851... 77.5401... 77.5885... 77.6997... 76.4850... 0 0.1118... 0.1620... 0.1672... 0.1746... 0.0650... 2πi 77.4851... 77.5401... 95.8738... 112.7032... 47.4247... 4πi 2.1356... 2.0279... 2.7647... 3.5183... 1.0549... 29/34 Take (u, v) with d.f. holonomy and longitude weight κ(l) ∈ 2πiZ. The ratio GN (κ(l)) := π log |HN+2 (M)(u, v)| |HN (M)(u, v)| gives the real part of the exponential growth rate of (HN (M))N for large values of N. At N = 151 we get: G151 (−6πi) = 2.03069... , G151 (−4πi) = −0.49036... G151 (−2πi) = 2.02968... , G151 (0) = 0.48922... Further numerical computations show that GN (2πi) converges to Vol(M) ≈ 2.02988321... as N >> 1. The values of GN at points of 4πiZ behave differently. Example 2: Msister = the figure eight knot “sister” manifold. again we consider the weakly branched triangulation dually represented by the graph: 2 30/34 1 On this weakly branched triangulation we have the state sums: HN (Msister )(u, v) ∝ N−1 X ζ (i+k)(j+I )−ik+ j 2 +I 2 2 −1 ω(u0 , u−1 1 |i −k)ω(v0 , v1 |I −j). i,j,k,I =0 Here are some values of |H15 (Msister )(u, v)| (conventions as before): 31/34 (0, 0) (1, 0) (1, 1) (2, 7) 2 4.7755... 4.5346... 2.6295... 1.9024... 1 173.2621... 173.3850... 126.3516... 101.7810... 0 0.2500... 0.3624... 0.1498... 0.3701... −1 173.2621... 173.3850... 236.1129... 302.5352... −2 4.7755... 4.5346... 9.3949... 14.6569... Some immediate perspectives: Recall that the volume conjecture for the QHI deals with H∞ ((xN )) := limsupN→ ∞ log |HN (xN )| N ∈ R ∪ {+∞}. In order to get a robust formulation of the conjecture one should: I Study numerically the invariants of many one-cusped mfds; I When does (HN (xN ))N have an exponential growth rate ? When this growth rate is given by complex Chern-Simons invariants ? Or what measures the discrepancy ? I What can we say about the variation of the growth rate as a function of the sequence (xN )N , about its regularity ? Etc. NB. H∞ ((xN )) should depend only on the limit point of (xN )N in the Hausdoff limit of the sequence of curves (XN )N (which is = (amoeba of the A-polynomial) ×T 2 ). 32/34 More ambitious perspectives: I Lift HN : XN → C/µN to a C-valued function; I HN is a rational function: expression in terms of augmented characters of meridian/longitude ? Poles, periods ? I Find a skein theoretic construction of HN I Find a pure geometric quantization construction of HN from the Chern-Simons function and line bundle. Work in progress with Benedetti and Frohman. 33/34 JOIN US ! 34/34