Complex lambda length as parameter for SL(2, C) representation spaces of punctured surface groups Toshihiro Nakanishi, Shimane Univeristy Introduction Let Fg,n be an oriented closed surface of genus g with n punctures at x1 ,..., xn . The fundamental group π1 (Fg,n ) has a canonical generator system (a1 , b1 , ..., ag , bg , d1 , ..., dn ) satisfying [a1 , b1 ] · · · [ag , bg ]d1 · · · dn = 1, where dj represents a loop which goes around the jth puncture xj in the positive direction. Introduction In this talk, it is always assumed that 2g − 2 + n > 0. SL(2, C) = �� a b c d � � : a, b, c, d ∈ C, ad − bc = 1 An element P of SL(2, C) is parabolic if P is conjugate to � � 1 1 ± 0 1 P = {P ∈ SL(2, C) : P is parabolic and trP = −2} � � −1 −1 = the conjugacy class of in SL(2, C). 0 −1 Introduction Let Rg,n = � ρ is faithful representation ρ : π1 (Fg,n ) → SL(2, C) : and ρ(dj ) ∈ P, j = 1, ..., n Remark. If ρ : π1 (Fg,n ) → P SL(2, C) is a faithful and discrete representation such that ρ(dj ) is parabolic for j = 1, ..., n, then there is a lift ρ̃ : π1 (Fg,n ) → SL(2, C) of ρ which belongs to Rg,n (Kra 1985) Purpose of this talk is: to introduce a coordinate-system to Rg,n . � / ∼ conj Penner’s λ-length coordinates For the punctured surfaces Fg,n , n > 0, we employ R.C.Penner’s lambda lengths Penner [1] introduced a coordinate-system by λ-lengths of ideal arcs on Fg,n to the decorated Teichmüller space T̃ (g, n), and obtained a cell decomposition of T̃ (g, n) invariant under the action of the mapping class group M Cg,n . [1] Penner, R. C., The decorated Teichmüller space of punctured surfaces, Comm. Math. Phys. 113 (1987), no. 2, 299–339. Penner’s λ-length coordinates For each ideal triangulation ∆ = (c1 , c2 , ..., cD ) of Fg,n , D = 6g − 6 + 3n, the tuple of their λ-lengths (λ(c1 ), λ(c2 ), ..., λ(cN )) gives a homeomorphism λ∆ : T̃ (g, n) → RD +. For two ideal triangulations ∆ and ∆� , T̃ (g, n) id� T̃ (g, n) λ∆ −−−−→ λ ∆� −−−−→ the coordinate change λ∆� ◦ λ∆ (T̃(g, n)) = −1 λ ◦λ � � ∆ ∆ D R+ λ∆� (T̃ (g, n)) = RD + −1 λ∆ is a rational transformation. Penner’s λ-length coordinates In particular, if ϕ : Fg,n → Fg,n is an orientation preserving homeomorphim, then ϕ∗ (λ(c1 ), ..., λ(cD )) = (λ(ϕ −1 (c1 )), ..., λ(ϕ −1 (cD ))) is a rational transformation. So the mapping class group acts on RD + as a group of rational transformations. Since the Teichmüller space T (g, n) is naturally embedded in T̃ (g, n), the coordinates can also parametrize T (g, n). Example Once Punctured Torus Group Let A and B be a canonical generator system of a once punctured torus group, normalized so that AB(∞) = 0 By some examples, we see how λ-length coordinates recover and a generator system of a�surface group � = −1 0 λ2 λ3 λ1 − λ2 3 −λ1 λ2 λ3 −1 ABA B −1 Then A= λ21 +λ23 , λ-length 1 −1 = P −1 , B= − λ22 +λ23 λ1 λ3 −λ2 λ2 − λ2 3 λ1 − λ3 . λ-length AB = 2 λ1 + + λ1 λ2 2 λ2 2 λ3 λ3 1 − λ3 0 λ-length Here we used λ-lengths defined by λ1 = −trA − trB AB, λ2 = −trB − trA λ3 = trAB + trBA, −1 satisfying 2 λ1 + 2 λ2 + 2 λ3 = 1 2 λ1 λ2 λ3 . −1 BA The mapping class group MC 1,1 of the once punctured torus is identified with a subgroup of the group of outer automorphisms of the once punctured group G. Consider the group automorphisms ϕ1 and ϕ2 of G determined by the following transformations on the generator system: ϕ1 : (A, B) �→ (A, BA), ϕ2 : (A, B) �→ (AB, B), These automorphisms acts on λ-length coordinates by � � λ21 + λ23 ϕ1∗ (λ1 , λ2 , λ3 ) = λ1 , λ3 , , λ2 ϕ2∗ (λ1 , λ2 , λ3 ) = � λ3 , λ 2 , λ22 + λ1 λ23 � . However, it is more convenient to use λ1 = − 16 (trA + trB −1 AB), λ2 = − 16 (trB + trA−1 BA) λ3 = 1 6 (trAB + trBA), which satisfy the Markoff equation 2 λ1 2 λ2 + + P = � In this case 2 λ3 = 3λ1 λ2 λ3 . −1 0 6 −1 � . About McShane’s identity for once puncture torus groups McShane’s identity for a once punctured torus with a complete, finite volume hyperbolic structure is � γ 1 1 = l(γ) 2 1+e The sum is over all simple closed geodesics γ on the torus, and l(γ) denotes the hyperbolic length of γ. McShane’s identity Let {A, B} be a canonical generator system of a once punctured torus group G. Let G0 = {(A, B), (B −1 , A)} and define Gn (n = 1, 2, ...) by induction: (g, h) ∈ Gn if and only if there exist a pair (a, b) in Gn−1 and a positive integer k such that either (g, h) = ((ab) k−1 Let a, (ab) a), or (g, h) = ((ba) b, (ba) k G= k ∞ � n=0 Gn . k−1 b). McShane’s identity Then McShane’s identity can be expressed as � � |tr(gh)| − (trgh)2 − 4 = 1. |tr(gh)| (g,h)∈G However, if the canonical generator system (A, B) is chosen so that trA < −2 and trB < −2, then � � tr(gh) − (trgh)2 − 4 = 1. tr(gh) (g,h)∈G McShane’s identity Let G be the once punctured torus group in SL(2, Z) with canonical generators −2 1 −2 −1 , B = . A= 1 −1 −1 −1 McShane’s identity in this case would be √ � 3m − 9m2 − 4 √ √ (3 − 5) + (3 − 2 2) + 2 = 1, m m where m runs over all Markoff numbers ≥ 5: m = 5, 13, 29, 34, 89, 169, 194, · · · , if a conjecture about Markoff numbers were settled. For the partial sum for m from 5 to 610, the value is 0.9999898618 · · · Why we need complex lambda lengths? Example: Twice punctured torus Let (A, B, C, D), ABA−1 B −1 CD = 1, be a canonical generator system of a twice punctured torus group G in SL(2, C). Define λ1 λ2 λ3 λ4 λ5 = −trA − trBA B C, −1 −1 = trABA + trB C, = −trA−1 − trB −1 CAB, = −trAB − trA−1 B −1 C, = −trC − trABA−1 B −1 . −1 −1 Why we need complex lambda length? Then under the normalization D(∞) = ∞ and C(∞) = 0, λ2 λ4 λ25 + λ3 λ24 λ5 + λ22 λ3 λ5 + λ1 λ23 λ5 + λ2 λ23 λ4 − λ1 λ2 λ4 λ25 , λ3 λ5 2 λ2 λ5 + λ1 λ3 λ5 + λ2 λ3 λ4 − λ1 λ4 λ25 , 2 2 2 2 λ2 λ4 λ5 + λ1 λ2 λ5 + λ3 λ4 λ5 λ1 λ3 λ5 + λ1 λ2 λ3 λ4 λ1 λ3 λ4 λ5 λ24 λ5 + λ22 λ5 + λ1 λ3 λ5 + λ2 λ3 λ4 λ2 λ4 λ5 A= λ1 λ-length λ2 − λ5 B= λ1 λ2 + λ4 λ5 λ3 C= 0 λ5 1 − λ5 , λ-length −2 λ-length Why we need complex lambda length? Here if λ3 λ5 (λ21 + λ22 + λ24 ) + λ1 λ5 (λ22 + λ23 + λ24 ) + λ2 λ4 (λ21 + λ23 + λ25 ) = (λ5 − 2)λ1 λ2 λ3 λ4 , D −1 = ABA −1 B −1 C= � −1 0 1 −1 � . Remark. We can replace the identity above by λ3 λ5 (λ21 + λ22 + λ24 ) + λ1 λ5 (λ22 + λ23 + λ24 ) + λ2 λ4 (λ21 + λ23 + λ25 ) = (11λ5 − 2)λ1 λ2 λ3 λ4 , so that this has a solution (λ1 , λ2 , λ3 , λ4 , λ5 , λ6 , λ7 , λ8 , λ9 ) = (1, 1, 1, 1, 1, 1, 1, 1, 1) In this case D −1 = ABA −1 B −1 C= � −1 11 0 −1 � . Why we need complex lambda length? The mapping class group MC 1,2 of the twice punctured torus is identified with a subgroup of the group of outer automorphisms of G. Consider the group automorphisms ϕ1 , ϕ2 and ϕ3 of G determined by the following transformations on the generator system: ϕ1 : (A, B, C) �→ (A, BA, C), ϕ2 : (A, B, C) �→ (ABA−1 , A−1 , C), ϕ3 : (A, B, C) �→ (A, BA−1 B −1 CB, BA−1 B −1 CBAB −1 ). Why we need complex lambda length? These automorphisms acts on λ-length coordinates by ϕ1∗ (λ1 , λ2 , λ3 , λ4 , λ5 ) = ϕ2∗ (λ1 , λ2 , λ3 , λ4 , λ5 ) = � � λ1 , λ 4 , λ 3 , λ2 , λ 3 , λ24 λ22 λ5 � + λ1 λ3 , λ5 , λ2 + λ1 λ3 λ5 + λ2 λ3 λ4 , λ1 λ4 λ22 � + λ1 λ3 , λ5 , λ4 and ϕ3∗ (λ1 , λ2 , λ3 , λ4 , λ5 ) � λ2 λ4 λ5 + λ1 λ22 + λ21 λ3 λ1 λ24 λ5 + λ2 λ4 λ25 + λ21 λ2 λ4 + λ1 λ22 λ5 + λ21 λ3 λ5 = λ1 , , , λ3 λ4 λ2 λ3 λ4 � 2 2 2 λ1 λ2 + λ4 λ5 (λ4 λ5 + λ1 λ2 λ4 + λ1 λ2 λ4 λ5 + λ2 λ5 + λ1 λ3 λ5 + λ2 λ3 λ4 ) , . 2 2 λ3 λ2 λ4 λ5 Why we need complex lambda length? We consider ψ= −1 ϕ1 ϕ3 ϕ2 : (A, B, C) �→ (ABA −1 ,C −1 2 −1 B A ,C −1 BCB −1 C) Then the transformation ψ∗ induced by ψ fixes the point √ √ τ0 = (−4, 4, −4 + 8 −1, −8, −8 − 8 −1). If (A, B, C) ∈ R1,2 be the point which has the coordinates τ0 , then there is M ∈ SL(2, C) such that (ABA−1 , C −1 B 2 A−1 , C −1 BCB −1 C) = (M −1 AM, M −1 BM, M −1 CM ). This fixed point can be found, because complex lambda lengths are treated. Why we need complex lambda length? The group Γ = �A, B, C, M � acts on the 3-dimensional hyperbolic space H3 and H3 /Γ is the Whitehead link complement. This space is fibered over the circle with fiber twice punctured torus. from Wikipedia Example Once Punctured Genus 2 Surface Group Let (A, B, C, D) be a canonical generator system of a once punctured genus 2 surface group with −1 −1 −1 −1 P = ABA B CDC D parabolic. Let x1 = −tr(DCD−1 ) − tr(DCD−1 P ), x2 = tr(DCD−1 C −1 D−1 ) − tr(DCD−1 C −1 D−1 P ), x3 = −tr(C −1 D−1 ) − tr(C −1 D−1 P ), x4 = −tr(DC 2 DC −1 D−1 ) − tr(DC 2 DC −1 D−1 P ), x5 = −tr(BAB −1 A−1 ) − tr(BAB −1 A−1 P ), x6 = −tr(AB −1 A−1 ) − tr(AB −1 A−1 P ), x7 = −tr(ABAB −1 A−1 ) − tr(ABAB −1 A−1 P ), x8 = −tr(AB) − tr(ABP ), x9 = −tr(ABA−1 B −2 A−1 ) − tr(ABA−1 B −2 A−1 P ). Then A= B= C= D= − x25 x6 x7 +x5 x27 x8 +x5 x26 x9 +x6 x7 x8 x9 +x5 x28 x9 x5 x6 x8 x9 − x5 x27 +x6 x7 x9 +x5 x8 x9 x25 x6 x8 x5 x6 +x7 x8 x9 x7 x5 x6 x5 x5 x7 +x6 x9 x25 x8 x5 x26 +x6 x7 x8 +x5 x8 x9 − x7 x9 x25 x6 x7 +x5 x27 x8 +x5 x26 x9 +x6 x7 x8 x9 +x5 x8 x29 − x5 x7 x8 x9 x1 x5 2 x5 − x1 xx43+x x25 x1 x2 x3 +x21 x5 +x3 x4 x5 x2 x4 x1 x2 x3 x4 +x22 x3 x5 +x21 x4 x5 +x3 x24 x5 +x1 x2 x25 − x2 x3 x4 x5 x1 x2 x3 x4 +x22 x3 x5 +x21 x4 x5 +x23 x4 x5 +x1 x2 x25 − x1 x3 x4 x5 x1 x2 x4 +x22 x5 +x3 x4 x5 x1 x3 x25 1 x5 − x2 x3x+x 4 x2 x5 where −1 ABA B −1 CDC −1 if D −1 = � −1 0 1 −1 � , x1 x2 x3 x4 x5 x6 x7 x8 x9 = x1 x2 x3 x4 x25 x6 x7 + x1 x2 x3 x4 x5 x26 x8 + x1 x2 x3 x4 x5 x27 x8 +x1 x2 x3 x4 x6 x7 x28 + x1 x2 x3 x4 x5 x26 x9 + x1 x2 x3 x4 x5 x27 x9 +x1 x2 x23 x6 x7 x8 x9 + x1 x2 x24 x6 x7 x8 x9 + x21 x3 x5 x6 x7 x8 x9 +x22 x3 x5 x6 x7 x8 x9 + x21 x4 x5 x6 x7 x8 x9 + x22 x4 x5 x6 x7 x8 x9 2 +x1 x2 x5 x6 x7 x8 x9 +x1 x2 x3 x4 x5 x8 x29 + 2 x1 x2 x3 x4 x5 x8 x9 + 2 x1 x2 x3 x4 x6 x7 x9 Definition of lambda length Definition of λ-length If P1 and P2 in P do not commute, then there is Q ∈ SL(2, C) such that P1 P2 = −Q . 2 This Q necessarily satisfies P2 = Q −1 P1 Q. In what follows, the diagram Q P1 − → P2 means that P1 , P2 ∈ P do not commute and Q2 = −P1 P2 . Definition of lambda length Define λ(P1 , P2 ) = trQ. Remark. λ(P1 , P2 ) is defined up to sign. If P2 = A−1 P1 A for A ∈ SL(2, C), then λ = λ(P1 , P2 ) = ±(trA + trAP1 ) � � −1 1 If AB = = P −1 , then 0 −1 A= � ∗ λ ∗ ∗ � , B= −trA − trAP � ∗ λ ∗ ∗ � −trB − trBP Definition of lambda length P1 In this diagram, Q3 P2 P2 = −1 Q3 P1 Q3 , P3 = Q2 Q1 −1 Q1 P2 Q1 , P1 P3 = −1 Q2 P3 Q2 . Thus, Q2 Q1 Q3 commutes with P1 and hence trQ2 Q1 Q3 = ±2. If trQ2 Q1 Q3 = −2, then (Q1 , Q2 , Q3 ) is called a (-)-system. Definition of lambda length Ideal Ptolemy Identity Consider the following diagram. We assume that trQ5 = trQ�5 , trQ6 = trQ�6 , which means that � Q5 = −1 P1 Q 5 P1 , � Q6 = −1 P4 Q 6 P4 . Definition of lambda length If three of (Q1 , Q2 , Q5 ), (Q�5 , Q3 , Q4 ), (Q1 , Q6 , Q4 ), (Q2 , Q3 , Q�6 ) are (−)-systems, then the remaining one is also a (−)-system. In this case trQ1 trQ3 + trQ2 trQ4 = trQ5 trQ6 . Ideal Ptolemy identity Definition of lambda length An ideal arc on Fg,n is an arc connecting two (not necessarily distinct) punctures. An ideal triangulation (c1 , ...., cD ) on Fg,n is a triangulation of Fg,n by simple ideal arcs. By using the Euler characteristic, it is easy to see D = 6g − 6 + 3n An ideal triangulation of the once punctured genus 2 surface Definition of lambda length It is possible that a component of the ideal triangulation (a triangle) looks like the figure below. Let Γ be a Fuchsian group of type (g, n) and identify Fg,n with H/Γ. Γ = �A1 , B1 , ..., Ag , Bg , D1 , ..., Dn � where [A1 , B1 ] · · · [Ag , Bg ]D1 · · · Dn = 1 and D1 , ..., Dn ∈ P Definition of lambda length Let c be an ideal arc and choose a lift c̃ of c in H. Then The end points of c̃ are fixed points of parabolic elements P1 and P2 , which are conjuate with some Dj , respectively in Γ. Let [ρ] ∈ Rg,n , where we identify π1 (Fg,n ) with Γ. Then ρ(P1 ), ρ(P2 ) ∈ P, and they do not commute. Define λ(c, ρ) = λ(ρ(P1 ), ρ(P2 )) Remark. λ(c, ρ) is defined up to sign. Definition of lambda length Let ∆ = (c1 , ..., cN ) be an ideal triangulation. A tuple of λ-lengths (λ(c1 , ·), ..., λ(cN , ·)) : Rg,n → CN is subordinate to (−)-systems if for each triangle (ci , cj , ck ) in ∆, the triple (Q3 , Q1 , Q2 ) with λ(ci ) = trQ1 , λ(cj ) = trQ3 , λ(ck ) = trQ3 , (Q3 , Q2 , Q1 ) is a (−)-system. P1 Q3 P2 Q2 Q1 P3 Definition of lambda length For each ideal triangulation ∆, there is a choice of λ-lengths so that (λ(c1 , ·), ..., λ(cN , ·)) : Rg,n → CN is subordinate to (−)-systems. Remark. The image of (λ(c1 , ·), ..., λ(cN , ·)) is contained in the intersection of zero sets of n polynomial equations. Expression of the λ-length of an ideal arc Let ∆ = (c1 , c2 , ..., cd ), d = 6g − 6 + 3n, be an ideal triangulation of Fg,n , and λj = λ(cj ) be the λ-length coordinate associated to cj . We assume that (λ1 , ..., λd ) is subordinate to (−)-systems. If c is an ideal arc, then where Pc (λ1 , ..., λd ) λ(c) = a1 a2 ad , λ 1 λ2 · · · λ d so that we can apply the ideal Ptolemy Pc (λ1 , ..., λd ) is a homogeneous polynomial with positive integer coefficients of degree a1 + a2 + · · · + ad + 1, and aj is the geometric intersection number of c and cj . The expression above is not necessarily in the lowest terms. We consider only once punctured surfaces Fg,1 , g ≥ 1. In this case, if c is simple, then the expression Pc (λ1 , ..., λd ) λ(c) = a1 a2 ad , λ 1 λ2 · · · λ d in the lowest terms is with a1 ,..., ad (d = 6g − 5 + 2n) defined as follows: The blue lines are the lift of ∆ in H, and the green line of c̃. Remark. For the case of once punctured torus, aj is the geometric intersection number of c and cj . Closed surface of genus 2 We identify the Teichmüller space Tg (g ≥ 2) with the space of marked Fuchsian groups. A point of Tg is a simultaneous conjugacy class of a generator system (A1 , B1 , ...., Ag , Bg ) satisfying the relation [A1 , B1 ] · · · [Ag , Bg ] = 1. We fix a Fuchsian group Γ0 as a reference point. A point of Tg is represented also by a conjugacy class of Fuchsian representations ρ : Γ0 → SL(2, C). Each element g of Γ0 defines a function (the trace function) by χ g : Tg → R χg (ρ) = trρ(g). Theorem (Schmutz 1993, Okumura 1996, Feng Luo 1998, Hamenstädt 2003) There are d = 6g − 5 elements g1 ,..., gd of Γ0 such that the tuple of their absolute trace functions embeds Tg into d R+ . (|χg1 |, ..., |χgd |) Remark It is impossible to embed Tg into R6g−6 by any tuple of 6g − 6 trace functions (Wolpert). For g = 2, let (A, B, C, D) be a point of T2 such that trA > 0, trB > 0, trC > 0, trD > 0. Then the following seven traces (which are all positive) embed T2 into R7+ . a = trA, v = −trACD2 , b = trB, w = −trACD, z = trAB, t = trCD u = −trACDC −1 . They satisfy awt + a2 + w2 + t2 + K 2 + S 2 + 4 � −w (K 2 + 4)(S 2 + 4) = 0, where � � K = abz − a2 − b2 − z 2 , S = uvt − u2 − v 2 − t2 Lemma. Let G be a group generated by n elements A1 , A2 , ...., An in SL(2, C), then for any element X ∈ G, trX is a polynomial with integer coefficients in the variables � trAj1 Aj2 · · · Ajk 1 ≤ j1 < j2 < · · · < jk ≤ n, : k = 1, ..., n � For the Fuchsian group �A, B, C, D� = �A1 , A1 , A2 , A4 �, each element of � � 1 ≤ j1 < j2 < · · · < jk ≤ 4, trAj1 Aj2 · · · Ajk : k = 1, ..., 4 is a rational function of a = trA, v = −trACD2 , b = trB, w = −trACD, z = trAB, t = trCD u = −trACDC −1 . (K 2 + S 2 + t2 + a2 + 4)u + w(2atu − 2av − uw + t2 uw − tvw) trC = , 2 2 w(S + t ) (K 2 + S 2 + t2 + a2 + 4)v + w(2au + twu − vw) trD = . 2 2 w(S + t ) uw(2a + tw) + v(4 + a2 + K 2 − w2 ) trAC = − , 2 2 S +t (4 + a2 + K 2 − w2 )v + wu(2a + tw) , trAD = (ad + u − cw) + t 2 2 S +t c(2b + a2 b − 2az + bK 2 ) − tuz + dw(ab + z + zK 2 ) − v(ab + zK 2 ) trBC = , 2 2 K +a 2(adz − bd) − u(ab + K 2 z) + tv(ab + z + K 2 z) + (c − dt)w(ab + z + K 2 z) trBD = , 2 2 K +a −2cz − btu + avz + wd(b − az) trABC = , 2 2 K +a d(K 2 + a2 + 2) + auz + vt(b − az) + w(bc − bdt − acz + adtz) trABD = , 2 2 K +a t(2b + a2 b − 2az + bK 2 ) + dvz + w(ab + K 2 z) + u(cz − dtz) trBCD = , 2 2 K +a −2tz + b(c − dt)u + bdv − awz trABCD = K 2 + a2 So for each X ∈ Γ = �A, B, C, D�, trX is a rational function of a, b, z, u, v, w, t. A mapping class ϕ induces an outer automorphism ϕ∗ of Γ. and (a, b, ..., w, t) = (trA, trB, ..., −trACD, trCD) ↓ ϕ∗ (trϕ∗ (A), trϕ∗ (B), ..., −trϕ∗ (ACD), ϕ∗ (CD)) is a rational transformation. One may need to change the sign of elements. As a corollary, we can conclude: The mapping class group M C2 is represented by a group Mercifin beaucoup of rational transformations in the variables. a, b, z, u, v, w and t. The speaker thanks the organizers of the Nara conference, Teruaki Kitano, Takayuki Morifuji, Ken’ichi Ohshika and Yasushi Yamashita. He also thanks Marjatta Näätänen and Gou Nakamura. fin