Yang-Baxter and braid dynamics
Transcription
Yang-Baxter and braid dynamics
Yang-Baxter and braid dynamics Alexander Veselov, Loughborough University GADUDIS, Glasgow, April 1, 2009 Plan I Dynamics and groups Plan I Dynamics and groups I Braid group and modular group Plan I Dynamics and groups I Braid group and modular group I Yang-Baxter versus braid relations Plan I Dynamics and groups I Braid group and modular group I Yang-Baxter versus braid relations I Diophantine analysis and Markov equation Plan I Dynamics and groups I Braid group and modular group I Yang-Baxter versus braid relations I Diophantine analysis and Markov equation I Markov dynamics Plan I Dynamics and groups I Braid group and modular group I Yang-Baxter versus braid relations I Diophantine analysis and Markov equation I Markov dynamics I Links with Painleve-VI and Frobenius manifolds Plan I Dynamics and groups I Braid group and modular group I Yang-Baxter versus braid relations I Diophantine analysis and Markov equation I Markov dynamics I Links with Painleve-VI and Frobenius manifolds Thanks to Leonid Chekhov, Boris Dubrovin, Andy Hone, Oleg Lisovyy for illuminating discussions and to Vsevolod Adler for very useful comments and help with computer simulations. Groups and dynamics Dynamical Erlangen programme: GROUPS ——– DYNAMICS Usual discrete dynamical system can be considered as nonlinear representation of Z. Groups and dynamics Dynamical Erlangen programme: GROUPS ——– DYNAMICS Usual discrete dynamical system can be considered as nonlinear representation of Z. INTEGRABILITY ∼ POLYNOMIAL GROWTH = ALMOST NILPOTENT GROUPS (Gromov’s theorem) Groups and dynamics Dynamical Erlangen programme: GROUPS ——– DYNAMICS Usual discrete dynamical system can be considered as nonlinear representation of Z. INTEGRABILITY ∼ POLYNOMIAL GROWTH = ALMOST NILPOTENT GROUPS (Gromov’s theorem) Examples: Julia, Fatou, Ritt ; AV: commuting maps Okamoto; Noumi, Yamada: affine Weyl group actions Groups and dynamics Dynamical Erlangen programme: GROUPS ——– DYNAMICS Usual discrete dynamical system can be considered as nonlinear representation of Z. INTEGRABILITY ∼ POLYNOMIAL GROWTH = ALMOST NILPOTENT GROUPS (Gromov’s theorem) Examples: Julia, Fatou, Ritt ; AV: commuting maps Okamoto; Noumi, Yamada: affine Weyl group actions I would like to discuss today the actions of the braid group B3 , which has exponential growth. Braid group Braid group Bn can be defined as the fundamental group π1 (Mn ) of the configuration space Mn of n different (unordered) points in C. Braid group Braid group Bn can be defined as the fundamental group π1 (Mn ) of the configuration space Mn of n different (unordered) points in C. It is generated by n − 1 elements σ1 , . . . , σn−1 with the following relations σi σi+1 σi = σi+1 σi σi+1 σi σj = σj σi , |i − j| > 1. Braid group Braid group Bn can be defined as the fundamental group π1 (Mn ) of the configuration space Mn of n different (unordered) points in C. It is generated by n − 1 elements σ1 , . . . , σn−1 with the following relations σi σi+1 σi = σi+1 σi σi+1 σi σj = σj σi , |i − j| > 1. If we add the relations σi2 = 1 the braid group Bn is reduced to the symmetric group Sn . Braid group B3 bx bz bx bz = bx bz bx bz bz bx bz = bx bz bx braid group defining relations (bx bz) 3 Braid group B3 and the modular group Braid group B3 is generated by σ1 and σ2 with the only relation σ1 σ2 σ1 = σ2 σ1 σ2 . It has the centre Z generated by (σ1 σ2 )3 = (σ1 σ2 σ1 )(σ2 σ1 σ2 ). Braid group B3 and the modular group Braid group B3 is generated by σ1 and σ2 with the only relation σ1 σ2 σ1 = σ2 σ1 σ2 . It has the centre Z generated by (σ1 σ2 )3 = (σ1 σ2 σ1 )(σ2 σ1 σ2 ). The quotient B3 /Z is isomorphic to the modular group PSL(2, Z) = SL(2, Z)/{±I }, which is itself the free product Z2 ∗ Z3 . Braid group B3 and the modular group Braid group B3 is generated by σ1 and σ2 with the only relation σ1 σ2 σ1 = σ2 σ1 σ2 . It has the centre Z generated by (σ1 σ2 )3 = (σ1 σ2 σ1 )(σ2 σ1 σ2 ). The quotient B3 /Z is isomorphic to the modular group PSL(2, Z) = SL(2, Z)/{±I }, which is itself the free product Z2 ∗ Z3 . A natural homomorphism PSL(2, Z) → SL(2, Z2 ) ∼ = S3 maps it to the symmetric group. Yang-Baxter versus braid relations Let X be any set and R be a map: R : X × X → X × X. Let Rij : X n → X n , X n = X × X × ..... × X be the maps which acts as R on i-th and j-th factors and identically on the others. If P : X 2 → X 2 is the permutation: P(x, y ) = (y , x), then R21 = PRP. Yang-Baxter versus braid relations Let X be any set and R be a map: R : X × X → X × X. Let Rij : X n → X n , X n = X × X × ..... × X be the maps which acts as R on i-th and j-th factors and identically on the others. If P : X 2 → X 2 is the permutation: P(x, y ) = (y , x), then R21 = PRP. The map R is called Yang-Baxter map if it satisfies the Yang-Baxter relation R12 R13 R23 = R23 R13 R12 . Yang-Baxter versus braid relations Let X be any set and R be a map: R : X × X → X × X. Let Rij : X n → X n , X n = X × X × ..... × X be the maps which acts as R on i-th and j-th factors and identically on the others. If P : X 2 → X 2 is the permutation: P(x, y ) = (y , x), then R21 = PRP. The map R is called Yang-Baxter map if it satisfies the Yang-Baxter relation R12 R13 R23 = R23 R13 R12 . The twisted maps Si = Pii+1 Rii+1 satisfy the braid relations Si Si+1 Si = Si+1 Si Si+1 , S12 S23 S12 = S23 S12 S23 . Yang-Baxter versus braid relations Let X be any set and R be a map: R : X × X → X × X. Let Rij : X n → X n , X n = X × X × ..... × X be the maps which acts as R on i-th and j-th factors and identically on the others. If P : X 2 → X 2 is the permutation: P(x, y ) = (y , x), then R21 = PRP. The map R is called Yang-Baxter map if it satisfies the Yang-Baxter relation R12 R13 R23 = R23 R13 R12 . The twisted maps Si = Pii+1 Rii+1 satisfy the braid relations Si Si+1 Si = Si+1 Si Si+1 , S12 S23 S12 = S23 S12 S23 . The reversible Yang-Baxter maps satisfy the relation R21 R = Id, corresponding to Si2 = Id. Yang-Baxter relations 1 1 2 = 2 3 3 Figure: Yang-Baxter relation 1 1 = 2 2 Figure: Reversibility Transfer dynamics Define the transfer maps (n) Ti : X n → X n , i = 1, . . . , n by (n) Ti = Rii+n−1 Rii+n−2 . . . Rii+1 , where the indices are considered modulo n. In particular (n) T1 = R1n R1n−1 . . . R12 . Transfer dynamics Define the transfer maps (n) Ti : X n → X n , i = 1, . . . , n by (n) Ti = Rii+n−1 Rii+n−2 . . . Rii+1 , where the indices are considered modulo n. In particular (n) T1 = R1n R1n−1 . . . R12 . (n) For any reversible Yang-Baxter map R the transfer maps Ti each other: (n) (n) (n) (n) T i Tj = T j T i commute with and satisfy the property (n) (n) T1 T2 . . . Tn(n) = Id. (n) Conversely, if Ti satisfy these properties then R is a reversible YB map. Transfer dynamics Define the transfer maps (n) Ti : X n → X n , i = 1, . . . , n by (n) Ti = Rii+n−1 Rii+n−2 . . . Rii+1 , where the indices are considered modulo n. In particular (n) T1 = R1n R1n−1 . . . R12 . (n) For any reversible Yang-Baxter map R the transfer maps Ti each other: (n) (n) (n) (n) T i Tj = T j T i commute with and satisfy the property (n) (n) T1 T2 . . . Tn(n) = Id. (n) Conversely, if Ti satisfy these properties then R is a reversible YB map. Transfer dynamics corresponds to the translations for a natural action of the (1) extended affine Weyl group Ãn−1 . Note that the action of the corresponding braid group is reduced simply to the symmetric group Sn . Commutativity of the transfer maps i+1 i i-1 i+1 i i-1 = j-1 j-1 j j j+1 j+1 Figure: Commutativity of the transfer maps Classification Quadrirational case, X = CP 1 : Adler, Bobenko, Suris Y X V U Figure: A quadrirational map on a pair of conics History: golden ratio and Hurwitz theorem It is well-known that any irrational real number α has a rational approximation p (given by continued fraction expansion) such that q |α − p 1 | < 2. q q History: golden ratio and Hurwitz theorem It is well-known that any irrational real number α has a rational approximation p (given by continued fraction expansion) such that q |α − p 1 | < 2. q q Hurwitz: For any α one can find infinitely many |α − p q such that p 1 |< √ . q 5q 2 For the golden ratio α=φ= √ 1+ 5 = [1, 1, 1, ...] 2 (and their equivalents) this is the best possible approximation. History: golden ratio and Hurwitz theorem It is well-known that any irrational real number α has a rational approximation p (given by continued fraction expansion) such that q |α − p 1 | < 2. q q Hurwitz: For any α one can find infinitely many |α − p q such that p 1 |< √ . q 5q 2 For the golden ratio α=φ= √ 1+ 5 = [1, 1, 1, ...] 2 (and their equivalents) this is the best possible approximation. In other words, φ is the most irrational number. What is the next one ? Markov spectrum Define Markov constant as the minimal possible c in |α − pq | < c : q2 µ(α) = lim inf q||αq||, q→∞ where ||x|| = min |x − n|, n ∈ Z. The set of all possible values of µ(α) is called Markov spectrum. Markov spectrum Define Markov constant as the minimal possible c in |α − pq | < c : q2 µ(α) = lim inf q||αq||, q→∞ where ||x|| = min |x − n|, n ∈ Z. The set of all possible values of µ(α) is called Markov spectrum. Markov theorem. The Markov spectrum above 1/3 is discrete and consists of the numbers m µ= √ , 9m2 − 4 where m is one of the Markov numbers: m = 1, 2, 5, 13, 29, 34, 89, 169, 194, 233, 433, 610, 985, ... Markov spectrum Define Markov constant as the minimal possible c in |α − pq | < c : q2 µ(α) = lim inf q||αq||, q→∞ where ||x|| = min |x − n|, n ∈ Z. The set of all possible values of µ(α) is called Markov spectrum. Markov theorem. The Markov spectrum above 1/3 is discrete and consists of the numbers m µ= √ , 9m2 − 4 where m is one of the Markov numbers: m = 1, 2, 5, 13, 29, 34, 89, 169, 194, 233, 433, 610, 985, ... For example, the next two most irrational numbers after φ are √ 2 = [1, 2, 2, 2....] and α= 9+ √ 221 = [2, 2, 1, 1]. 10 Markov triples Markov numbers are parts of the Markov triples, which are the positive integer solutions of the Markov equation x 2 + y 2 + z 2 − 3xyz = 0. Markov triples Markov numbers are parts of the Markov triples, which are the positive integer solutions of the Markov equation x 2 + y 2 + z 2 − 3xyz = 0. Theorem (Markov) All Markov triples can be obtained from (1, 1, 1) by the involution σ : (x, y , z) → (x, y , 3xy − z) and permutations. Markov triples Markov numbers are parts of the Markov triples, which are the positive integer solutions of the Markov equation x 2 + y 2 + z 2 − 3xyz = 0. Theorem (Markov) All Markov triples can be obtained from (1, 1, 1) by the involution σ : (x, y , z) → (x, y , 3xy − z) and permutations. (1, 1, 1) → (1, 1, 2) → (1, 2, 1) → (1, 2, 5) → (1, 5, 2) → (1, 5, 13) → ... Markov triples Markov numbers are parts of the Markov triples, which are the positive integer solutions of the Markov equation x 2 + y 2 + z 2 − 3xyz = 0. Theorem (Markov) All Markov triples can be obtained from (1, 1, 1) by the involution σ : (x, y , z) → (x, y , 3xy − z) and permutations. (1, 1, 1) → (1, 1, 2) → (1, 2, 1) → (1, 2, 5) → (1, 5, 2) → (1, 5, 13) → ... Markov dynamics is the action of PSL2 (Z) (and hence of the braid group B3 ) on C3 determined by σ and τ (x, y , z) = (z, x, y ). It has an obvious integral F = x 2 + y 2 + z 2 − 3xyz. Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov dynamics: F = 0. Markov triples in logarithmic scale: Markov tree Markov tree Markov tree Markov tree Markov tree Markov tree Markov tree Markov tree F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. F = 49 : ”integrable chaos” On the special level F = 4 9 the dynamics can be linearised by substitution 2 2 2 2 x = cos φ, y = cos ψ, z = cos(φ + ψ), z 0 = cos(φ − ψ). 3 3 3 3 The hyperbolic elements of SL(2, Z) act as Anosov maps, so the dynamics is nevertheless chaotic. Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Generic level of F : chaos ? What happens at generic level is an interesting question (cf. Hone) Tropical Markov dynamics It is not clear how to tropicalize the formula z 0 = 3xy − z, so we can use another relation zz 0 = x 2 + y 2 + c to define the following tropical Markov dynamics: σ : (X , Y , Z ) → (X , Y , 2 max(X , Y , C ) − Z ) and τ : (X , Y , Z ) → (Y , Z , X ). It has the following tropical integral F = 2 max(X , Y , Z , C ) − (X + Y + Z ), corresponding to F = x2 + y2 + z2 + c . xyz Tropical Markov orbit Markov orbit and quantum cohomology A remarkable observation due to Dubrovin is that in the theory of Frobenius manifolds Markov orbit corresponds to quantum cohomology of projective plane CP 2 Markov orbit and quantum cohomology A remarkable observation due to Dubrovin is that in the theory of Frobenius manifolds Markov orbit corresponds to quantum cohomology of projective plane CP 2 The corresponding potential is F = 1 t1 (t1 t3 + t22 ) + G (t2 , t3 ), 2 where G (t2 , t3 ) = ∞ X d=1 Nd t33d−1 e dt2 , (3d − 1)! Nd is the number of rational curves in CP 2 passing by 3d − 1 points in general position. Markov orbit and quantum cohomology A remarkable observation due to Dubrovin is that in the theory of Frobenius manifolds Markov orbit corresponds to quantum cohomology of projective plane CP 2 The corresponding potential is F = 1 t1 (t1 t3 + t22 ) + G (t2 , t3 ), 2 where G (t2 , t3 ) = ∞ X d=1 Nd t33d−1 e dt2 , (3d − 1)! Nd is the number of rational curves in CP 2 passing by 3d − 1 points in general position. Kontsevich, Manin: WDVV equation leads to recurrent formula for Nd : ! ! X 3d − 1 3d − 1 Nd = Nk Nl k 2 l[l −k ]. 3k − 2 3k − 1 k+l=d Painlevé-VI and braid group orbits Painlevé VI equation describes monodromy preserving deformations of Fuchsian systems „ « dΦ A2 A3 A1 = + + Φ, Φ ∈ SL(2, C). dλ λ − u1 λ − u2 λ − u3 Painlevé-VI and braid group orbits Painlevé VI equation describes monodromy preserving deformations of Fuchsian systems „ « dΦ A2 A3 A1 = + + Φ, Φ ∈ SL(2, C). dλ λ − u1 λ − u2 λ − u3 Analytic continuation induces an action of the braid group B3 on the triples of the monodromy matrices M1 , M2 , M3 : σ1 (M1 , M2 , M3 ) = (M2 , M2 M1 M2−1 , M3 ), σ2 (M1 , M2 , M3 ) = (M1 , M3 , M3 M2 M3−1 ). In 2 × 2 case this leads to the generalised Markov dynamics with σ : (x, y , z) → (x, y , xy − z − ω3 ) and τ : (x, y , z; ω1 , ω2 , ω3 ) → (y , z, x; ω2 , ω3 , ω1 ), preserving F = x 2 + y 2 + z 2 − xyz − ω1 x − ω2 y − ω3 z. Markov dynamics and special solutions of PVI Markov orbit corresponds to quantum cohomology solution (Dubrovin) Markov dynamics and special solutions of PVI Markov orbit corresponds to quantum cohomology solution (Dubrovin) Integrable case F = 4 9 corresponds to the Picard solutions (Fuchs, Hitchin) Markov dynamics and special solutions of PVI Markov orbit corresponds to quantum cohomology solution (Dubrovin) Integrable case F = 4 9 corresponds to the Picard solutions (Fuchs, Hitchin) Dubrovin: Finite orbits correspond to the algebraic solutions of Painleve-VI (Dubrovin, Mazzocco) Lisovyy, Tikhiy: classification of the finite braid orbits Markov dynamics and special solutions of PVI Markov orbit corresponds to quantum cohomology solution (Dubrovin) Integrable case F = 4 9 corresponds to the Picard solutions (Fuchs, Hitchin) Dubrovin: Finite orbits correspond to the algebraic solutions of Painleve-VI (Dubrovin, Mazzocco) Lisovyy, Tikhiy: classification of the finite braid orbits Comparison with Watanabe result suggests that in general Markov dynamics can not be ”integrated in classical functions” (see, however, Picard solutions). Markov dynamics and special solutions of PVI Markov orbit corresponds to quantum cohomology solution (Dubrovin) Integrable case F = 4 9 corresponds to the Picard solutions (Fuchs, Hitchin) Dubrovin: Finite orbits correspond to the algebraic solutions of Painleve-VI (Dubrovin, Mazzocco) Lisovyy, Tikhiy: classification of the finite braid orbits Comparison with Watanabe result suggests that in general Markov dynamics can not be ”integrated in classical functions” (see, however, Picard solutions). Relation with Okamoto: affine Weyl groups are acting on the set of all solutions while braid group describes the monodromy of a given solution. Other relations and questions Teichmüller spaces (Chekhov-Penner, Fock-Goncharov) Other relations and questions Teichmüller spaces (Chekhov-Penner, Fock-Goncharov) Cluster algebras and Laurent phenomenon (Berenstein, Fomin, Zelevinsky; Hone) 2 2 xn+1 xn = xn+1 + xn+2 + c. Other relations and questions Teichmüller spaces (Chekhov-Penner, Fock-Goncharov) Cluster algebras and Laurent phenomenon (Berenstein, Fomin, Zelevinsky; Hone) 2 2 xn+1 xn = xn+1 + xn+2 + c. Poisson structures (Goldman, Chekhov-Fock) Other relations and questions Teichmüller spaces (Chekhov-Penner, Fock-Goncharov) Cluster algebras and Laurent phenomenon (Berenstein, Fomin, Zelevinsky; Hone) 2 2 xn+1 xn = xn+1 + xn+2 + c. Poisson structures (Goldman, Chekhov-Fock) Generalisations to K3 surfaces in (CP 1 )3 (Silverman) Other relations and questions Teichmüller spaces (Chekhov-Penner, Fock-Goncharov) Cluster algebras and Laurent phenomenon (Berenstein, Fomin, Zelevinsky; Hone) 2 2 xn+1 xn = xn+1 + xn+2 + c. Poisson structures (Goldman, Chekhov-Fock) Generalisations to K3 surfaces in (CP 1 )3 (Silverman) Tropical and geometrical versions Other relations and questions Teichmüller spaces (Chekhov-Penner, Fock-Goncharov) Cluster algebras and Laurent phenomenon (Berenstein, Fomin, Zelevinsky; Hone) 2 2 xn+1 xn = xn+1 + xn+2 + c. Poisson structures (Goldman, Chekhov-Fock) Generalisations to K3 surfaces in (CP 1 )3 (Silverman) Tropical and geometrical versions Relations with arithmetic (Markov, Mordell, Silverman)