Geodesic flows on quotients of Bruhat-Tits buildings and Number Theory Anish Ghosh University of East Anglia joint with J. Athreya (Urbana) and A. Prasad (Chennai) 1 / 127 Diagonal Actions 2 / 127 Diagonal Actions Recurrence 3 / 127 Diagonal Actions Recurrence Buildings 4 / 127 Diagonal Actions Recurrence Buildings Shrinking Targets 5 / 127 Shrinking 6 / 127 Shrinking Target 7 / 127 Shrinking Target Problems 8 / 127 Shrinking Target Problems 0 − 1 laws 9 / 127 Shrinking Target Problems 0 − 1 laws Return time asymptotics 10 / 127 Shrinking Target Problems 0 − 1 laws Return time asymptotics Heirarchy 11 / 127 Excursions in hyperbolic 3 manifolds 12 / 127 Sullivan (1982) Y = Hk+1 /Γ, y0 ∈ Y , {rt } a sequence of real numbers 13 / 127 Sullivan (1982) Y = Hk+1 /Γ, y0 ∈ Y , {rt } a sequence of real numbers For any y ∈ Y and almost every ξ ∈ Ty (Y ) there are infinitely many t ∈ N such that dist(y0 , γt (y , ξ)) ≥ rt 14 / 127 Sullivan (1982) Y = Hk+1 /Γ, y0 ∈ Y , {rt } a sequence of real numbers For any y ∈ Y and almost every ξ ∈ Ty (Y ) there are infinitely many t ∈ N such that dist(y0 , γt (y , ξ)) ≥ rt if and only if ∞ X e −krt = ∞ t=1 15 / 127 Logarithm Laws Corollary For almost every y ∈ Y and almost all ξ ∈ Ty (Y ) lim sup t→∞ 1 dist(y , γt (y , ξ)) = log t k 16 / 127 Some Number theory √ For almost every z, there exist infinitely many pairs (p, q) ∈ O( −d) such that √ p ψ(|q|) |z − | < and (p, q) = O( −d) 2 q |q| 17 / 127 Some Number theory √ For almost every z, there exist infinitely many pairs (p, q) ∈ O( −d) such that √ p ψ(|q|) |z − | < and (p, q) = O( −d) 2 q |q| if and only if ∞ X ψ(t) t=1 t2 =∞ 18 / 127 Sullivan’s strategy Use the mixing property of the geodesic flow (in the tradition of Margulis) 19 / 127 Sullivan’s strategy Use the mixing property of the geodesic flow (in the tradition of Margulis) Along with geometric arguments 20 / 127 Sullivan’s strategy Use the mixing property of the geodesic flow (in the tradition of Margulis) Along with geometric arguments To force quasi-independence 21 / 127 Sullivan’s strategy Use the mixing property of the geodesic flow (in the tradition of Margulis) Along with geometric arguments To force quasi-independence And then apply the Borel-Cantelli Lemma 22 / 127 Kleinbock-Margulis (1999) Generalized Sullivan’s results 23 / 127 Kleinbock-Margulis (1999) Generalized Sullivan’s results G semisimple Lie group, K maximal compact subgroup, Γ non-uniform lattice 24 / 127 Kleinbock-Margulis (1999) Generalized Sullivan’s results G semisimple Lie group, K maximal compact subgroup, Γ non-uniform lattice Y = K \G /Γ a locally symmetric space 25 / 127 Kleinbock-Margulis (1999) Generalized Sullivan’s results G semisimple Lie group, K maximal compact subgroup, Γ non-uniform lattice Y = K \G /Γ a locally symmetric space Logarithm laws for locally symmetric spaces 26 / 127 Khintchine-Groshev Theorem For almost all A ∈ Matm×n (R), there exist infinitely many q ∈ Zn , p ∈ Zm such that kAq + pkm < ψ(kqkn ) 27 / 127 Khintchine-Groshev Theorem For almost all A ∈ Matm×n (R), there exist infinitely many q ∈ Zn , p ∈ Zm such that kAq + pkm < ψ(kqkn ) if and only if ∞ X ψ(t) = ∞ t=1 28 / 127 Khintchine-Groshev Theorem For almost all A ∈ Matm×n (R), there exist infinitely many q ∈ Zn , p ∈ Zm such that kAq + pkm < ψ(kqkn ) if and only if ∞ X ψ(t) = ∞ t=1 If time permits More general multiplicative results 29 / 127 SLn (R)/ SLn (Z) can be identified with the space Xn of unimodular lattices in Rn 30 / 127 SLn (R)/ SLn (Z) can be identified with the space Xn of unimodular lattices in Rn Mahler’s compactness criterion describes compact subsets of Xn 31 / 127 SLn (R)/ SLn (Z) can be identified with the space Xn of unimodular lattices in Rn Mahler’s compactness criterion describes compact subsets of Xn The systole of a lattice s(Λ) := min kv k v ∈Λ 32 / 127 Dani Correspondence x ! ux := 1 x 0 1 33 / 127 Dani Correspondence x ! ux := gt := et 0 1 x 0 1 0 e −t 34 / 127 Dani Correspondence x ! ux := gt := et 0 1 x 0 1 0 e −t Read Diophantine properties of x from the gt (semi) orbits of ux Z2 35 / 127 Dani Correspondence x ! ux := gt := et 0 1 x 0 1 0 e −t Read Diophantine properties of x from the gt (semi) orbits of ux Z2 For example, x is badly approximable if and only if the orbit of ux Z2 is bounded 36 / 127 More generally There are infinitely many solutions to |qx + p| < ψ(q) 37 / 127 More generally There are infinitely many solutions to |qx + p| < ψ(q) if and only if there are infinitely many t > 0 such that s(gt ux Z2 ) ≤ r (t) 38 / 127 r (t) defines the complement Xn (t) of a compact set 39 / 127 r (t) defines the complement Xn (t) of a compact set Shrinking systolic neighborhoods 40 / 127 r (t) defines the complement Xn (t) of a compact set Shrinking systolic neighborhoods If r (t) → 0 very fast we should expect few solutions 41 / 127 r (t) defines the complement Xn (t) of a compact set Shrinking systolic neighborhoods If r (t) → 0 very fast we should expect few solutions The speed is governed by convergence of ∞ X vol(Xn (t)) t=0 42 / 127 Similar story for logarithm laws 43 / 127 Similar story for logarithm laws Model geodesic flow on T 1 (K \G /Γ) by the action of diagonal one-parameter subgroups on G /Γ 44 / 127 Similar story for logarithm laws Model geodesic flow on T 1 (K \G /Γ) by the action of diagonal one-parameter subgroups on G /Γ Compute vol({x ∈ G /Γ : d(x0 , gt x) ≥ r (t)}) 45 / 127 Similar story for logarithm laws Model geodesic flow on T 1 (K \G /Γ) by the action of diagonal one-parameter subgroups on G /Γ Compute vol({x ∈ G /Γ : d(x0 , gt x) ≥ r (t)}) Use quantitative mixing, i.e. effective decay of matrix coefficients for automorphic representation 46 / 127 Similar story for logarithm laws Model geodesic flow on T 1 (K \G /Γ) by the action of diagonal one-parameter subgroups on G /Γ Compute vol({x ∈ G /Γ : d(x0 , gt x) ≥ r (t)}) Use quantitative mixing, i.e. effective decay of matrix coefficients for automorphic representation For φ, ψ ∈ L2 (G /Γ), and g ∈ G |hg φ, ψi| L2l (φ)L2l (ψ)H(g )−λ 47 / 127 Similar story for logarithm laws Model geodesic flow on T 1 (K \G /Γ) by the action of diagonal one-parameter subgroups on G /Γ Compute vol({x ∈ G /Γ : d(x0 , gt x) ≥ r (t)}) Use quantitative mixing, i.e. effective decay of matrix coefficients for automorphic representation For φ, ψ ∈ L2 (G /Γ), and g ∈ G |hg φ, ψi| L2l (φ)L2l (ψ)H(g )−λ To force quasi-independence on average 48 / 127 Much recent activity 49 / 127 Much recent activity Maucourant, Hersonsky-Paulin, Parkonnen-Paulin, Chaika 50 / 127 Much recent activity Maucourant, Hersonsky-Paulin, Parkonnen-Paulin, Chaika Nogueira, Laurent-Nogueira, G-Gorodnik-Nevo 51 / 127 Key technical point In order to apply decay of matrix coefficients we have to smoothen cusp neighborhoods 52 / 127 Key technical point In order to apply decay of matrix coefficients we have to smoothen cusp neighborhoods And use suitable Sobolev norms 53 / 127 Key technical point In order to apply decay of matrix coefficients we have to smoothen cusp neighborhoods And use suitable Sobolev norms This destroys sensitive information about exponents 54 / 127 Key technical point In order to apply decay of matrix coefficients we have to smoothen cusp neighborhoods And use suitable Sobolev norms This destroys sensitive information about exponents Which is crucial in G-Gorodnik-Nevo 55 / 127 Volume estimates Γ arithmetic lattice in G , S Siegel set for Γ in G 56 / 127 Volume estimates Γ arithmetic lattice in G , S Siegel set for Γ in G ¯ metric on G /Γ defined by dist metric on G , dist ¯ Γ, hΓ) := inf {dist(g γ, h) : γ ∈ Γ} dist(g γ∈Γ 57 / 127 Volume estimates Γ arithmetic lattice in G , S Siegel set for Γ in G ¯ metric on G /Γ defined by dist metric on G , dist ¯ Γ, hΓ) := inf {dist(g γ, h) : γ ∈ Γ} dist(g γ∈Γ Coarse isometry (Leuzinger + Ji) ¯ Γ, hΓ) ≤ dist(g , h) ≤ dist(g ¯ Γ, hΓ) + C dist(g 58 / 127 Volume estimates Γ arithmetic lattice in G , S Siegel set for Γ in G ¯ metric on G /Γ defined by dist metric on G , dist ¯ Γ, hΓ) := inf {dist(g γ, h) : γ ∈ Γ} dist(g γ∈Γ Coarse isometry (Leuzinger + Ji) ¯ Γ, hΓ) ≤ dist(g , h) ≤ dist(g ¯ Γ, hΓ) + C dist(g For g , h ∈ S 59 / 127 Trees and Buildings If G is a p-adic Lie group then 60 / 127 Trees and Buildings If G is a p-adic Lie group then The analogue of a symmetric space is a Bruhat-Tits building 61 / 127 Trees and Buildings If G is a p-adic Lie group then The analogue of a symmetric space is a Bruhat-Tits building Example from Shahar Mozes’ talk 62 / 127 Trees and Buildings If G is a p-adic Lie group then The analogue of a symmetric space is a Bruhat-Tits building Example from Shahar Mozes’ talk G = PGL2 (Qp ) × PGL2 (Qq ) 63 / 127 Trees and Buildings If G is a p-adic Lie group then The analogue of a symmetric space is a Bruhat-Tits building Example from Shahar Mozes’ talk G = PGL2 (Qp ) × PGL2 (Qq ) Γ a co-compact lattice coming from a quaternion algebra 64 / 127 Trees and Buildings If G is a p-adic Lie group then The analogue of a symmetric space is a Bruhat-Tits building Example from Shahar Mozes’ talk G = PGL2 (Qp ) × PGL2 (Qq ) Γ a co-compact lattice coming from a quaternion algebra X /Γ is a finite graph 65 / 127 Trees and Buildings If G is a p-adic Lie group then The analogue of a symmetric space is a Bruhat-Tits building Example from Shahar Mozes’ talk G = PGL2 (Qp ) × PGL2 (Qq ) Γ a co-compact lattice coming from a quaternion algebra X /Γ is a finite graph In fact, any lattice in a p-adic Lie group is necessarily cocompact (Tamagawa) 66 / 127 Trees and Buildings However, we could consider G = PGL2 (Fp ((t))) × PGL2 (Fp ((t −1 ))) 67 / 127 Trees and Buildings However, we could consider G = PGL2 (Fp ((t))) × PGL2 (Fp ((t −1 ))) And Γ = PGL2 (Fp [t, t −1 ]) 68 / 127 Trees and Buildings However, we could consider G = PGL2 (Fp ((t))) × PGL2 (Fp ((t −1 ))) And Γ = PGL2 (Fp [t, t −1 ]) Which turns out to be a non-uniform lattice in G 69 / 127 Trees and Buildings The Tree of SL2 (F2 ((X −1 ))) 70 / 127 Trees and Buildings Ultrametric logarithm laws So it makes sense to investigate analogues of logarithm laws for 71 / 127 Trees and Buildings Ultrametric logarithm laws So it makes sense to investigate analogues of logarithm laws for The graph K \G /Γ where G is an algebraic group 72 / 127 Trees and Buildings Ultrametric logarithm laws So it makes sense to investigate analogues of logarithm laws for The graph K \G /Γ where G is an algebraic group Defined over a totally disconnected field 73 / 127 Trees and Buildings Ultrametric logarithm laws So it makes sense to investigate analogues of logarithm laws for The graph K \G /Γ where G is an algebraic group Defined over a totally disconnected field K is a compact open subgroup and Γ is a lattice in G 74 / 127 Trees and Buildings Local fields of positive characteristic Fp - finite field 75 / 127 Trees and Buildings Local fields of positive characteristic Fp - finite field Fp [X ] ! Z 76 / 127 Trees and Buildings Local fields of positive characteristic Fp - finite field Fp [X ] ! Z Fp (X ) ! Q 77 / 127 Trees and Buildings Local fields of positive characteristic Fp - finite field Fp [X ] ! Z Fp (X ) ! Q Fp ((X −1 )) ! R 78 / 127 Trees and Buildings Euclidean buildings G has a BN pair 79 / 127 Trees and Buildings Euclidean buildings G has a BN pair Corresponding to which there is a Euclidean building 80 / 127 Trees and Buildings Euclidean buildings G has a BN pair Corresponding to which there is a Euclidean building A Euclidean building is polyhedral complex 81 / 127 Trees and Buildings Euclidean buildings G has a BN pair Corresponding to which there is a Euclidean building A Euclidean building is polyhedral complex With a family of subcomplexes called apartments 82 / 127 Trees and Buildings Euclidean buildings G has a BN pair Corresponding to which there is a Euclidean building A Euclidean building is polyhedral complex With a family of subcomplexes called apartments There is a nice (CAT(0)) metric on the building 83 / 127 Trees and Buildings Euclidean buildings G has a BN pair Corresponding to which there is a Euclidean building A Euclidean building is polyhedral complex With a family of subcomplexes called apartments There is a nice (CAT(0)) metric on the building And G acts by isometries 84 / 127 Trees and Buildings Borel-Harish-Chandra, Behr-Harder G (Fp [X ]) is a lattice in G (Fp ((X −1 ))) 85 / 127 Trees and Buildings Borel-Harish-Chandra, Behr-Harder G (Fp [X ]) is a lattice in G (Fp ((X −1 ))) Lubotzky G (rank 1) has non-uniform and uniform, arithmetic and non-arithmetic lattices. 86 / 127 Trees and Buildings Logarithm laws for trees 87 / 127 Trees and Buildings Logarithm laws for trees T - a regular tree, then Aut(T ) is a locally compact group 88 / 127 Trees and Buildings Logarithm laws for trees T - a regular tree, then Aut(T ) is a locally compact group Hersonsky and Paulin obtained logarithm laws for T /Γ 89 / 127 Trees and Buildings Logarithm laws for trees T - a regular tree, then Aut(T ) is a locally compact group Hersonsky and Paulin obtained logarithm laws for T /Γ Where Γ is a geometrically finite lattice in Aut(T ) 90 / 127 Trees and Buildings Logarithm laws for trees T - a regular tree, then Aut(T ) is a locally compact group Hersonsky and Paulin obtained logarithm laws for T /Γ Where Γ is a geometrically finite lattice in Aut(T ) By Serre + Raghunathan + Lubotzky 91 / 127 Trees and Buildings Logarithm laws for trees T - a regular tree, then Aut(T ) is a locally compact group Hersonsky and Paulin obtained logarithm laws for T /Γ Where Γ is a geometrically finite lattice in Aut(T ) By Serre + Raghunathan + Lubotzky This includes all algebraic examples 92 / 127 Trees and Buildings Logarithm laws for trees T - a regular tree, then Aut(T ) is a locally compact group Hersonsky and Paulin obtained logarithm laws for T /Γ Where Γ is a geometrically finite lattice in Aut(T ) By Serre + Raghunathan + Lubotzky This includes all algebraic examples Namely lattices in rank 1 algebraic groups 93 / 127 Trees and Buildings Trees continued They also obtained generalizations of Khintchine’s theorem 94 / 127 Trees and Buildings Trees continued They also obtained generalizations of Khintchine’s theorem In positive characteristic 95 / 127 Trees and Buildings Trees continued They also obtained generalizations of Khintchine’s theorem In positive characteristic Their method is very geometric 96 / 127 Trees and Buildings Trees continued They also obtained generalizations of Khintchine’s theorem In positive characteristic Their method is very geometric And reminiscent of Sullivan 97 / 127 Trees and Buildings Let G be a connected, semisimple linear algebraic group defined and split over Fp 98 / 127 Trees and Buildings Let G be a connected, semisimple linear algebraic group defined and split over Fp Let Γ be an S arithmetic lattice in GS 99 / 127 Trees and Buildings Let G be a connected, semisimple linear algebraic group defined and split over Fp Let Γ be an S arithmetic lattice in GS By a result of Margulis + Venkataramana 100 / 127 Trees and Buildings Let G be a connected, semisimple linear algebraic group defined and split over Fp Let Γ be an S arithmetic lattice in GS By a result of Margulis + Venkataramana This includes any irreducible lattice in higher rank 101 / 127 Trees and Buildings And their quotients We prove analogues of the results of Kleinbock-Margulis 102 / 127 Trees and Buildings And their quotients We prove analogues of the results of Kleinbock-Margulis Namely logarithm laws for arithmetic quotients of 103 / 127 Trees and Buildings And their quotients We prove analogues of the results of Kleinbock-Margulis Namely logarithm laws for arithmetic quotients of Bruhat-Tits buildings of semisimple groups 104 / 127 Trees and Buildings And their quotients We prove analogues of the results of Kleinbock-Margulis Namely logarithm laws for arithmetic quotients of Bruhat-Tits buildings of semisimple groups And several results in Diophantine Approximation 105 / 127 Trees and Buildings Main Result Let B be a family of measurable subsets and F = {fn } denote a sequence of µ-preserving transformations of G /Γ 106 / 127 Trees and Buildings Main Result Let B be a family of measurable subsets and F = {fn } denote a sequence of µ-preserving transformations of G /Γ B is Borel-Cantelli for F if for every sequence {An : n ∈ N} of sets from B µ({x ∈ G /Γ | fn (x) ∈ An for infinitely many n ∈ N}) P∞ 0 if n=1 µ(An ) < ∞ = P∞ 1 if n=1 µ(An ) = ∞ 107 / 127 Trees and Buildings ∆ a function on G /Γ define the tail distribution function Φ∆ (n) := µ({x ∈ G /Γ | ∆(x) ≥ s n }) 108 / 127 Trees and Buildings ∆ a function on G /Γ define the tail distribution function Φ∆ (n) := µ({x ∈ G /Γ | ∆(x) ≥ s n }) Call ∆ “distance-like” if Φ∆ (n) s −κn ∀ n ∈ Z 109 / 127 Trees and Buildings Let F = {fn | n ∈ N} be a sequence of elements of G satisfying sup ∞ X m∈N n=1 kfn fm−1 k−β < ∞ ∀ β > 0, 110 / 127 Trees and Buildings Let F = {fn | n ∈ N} be a sequence of elements of G satisfying sup ∞ X m∈N n=1 kfn fm−1 k−β < ∞ ∀ β > 0, and let ∆ be a “distance-like” function on G /Γ. Then B(∆) := {{x ∈ G /Γ | ∆(x) ≥ s n } | n ∈ Z} 111 / 127 Trees and Buildings Let F = {fn | n ∈ N} be a sequence of elements of G satisfying sup ∞ X m∈N n=1 kfn fm−1 k−β < ∞ ∀ β > 0, and let ∆ be a “distance-like” function on G /Γ. Then B(∆) := {{x ∈ G /Γ | ∆(x) ≥ s n } | n ∈ Z} is Borel-Cantelli for F 112 / 127 Trees and Buildings Ergodic ideas We establish effective decay of matrix coefficients for semisimple groups in positive characteristic 113 / 127 Trees and Buildings Ergodic ideas We establish effective decay of matrix coefficients for semisimple groups in positive characteristic Well known to experts but unavailable in the literature 114 / 127 Trees and Buildings Ergodic ideas We establish effective decay of matrix coefficients for semisimple groups in positive characteristic Well known to experts but unavailable in the literature Main idea is to establish strong spectral gap for the regular representation of G on G /Γ 115 / 127 Trees and Buildings Ergodic ideas We establish effective decay of matrix coefficients for semisimple groups in positive characteristic Well known to experts but unavailable in the literature Main idea is to establish strong spectral gap for the regular representation of G on G /Γ 116 / 127 Trees and Buildings Ergodic ideas We establish effective decay of matrix coefficients for semisimple groups in positive characteristic Well known to experts but unavailable in the literature Main idea is to establish strong spectral gap for the regular representation of G on G /Γ 117 / 127 Trees and Buildings Effective mixing Let G be a connected, semisimple, linear algebraic group, defined and quasi-split over k and Γ be an non-uniform irreducible lattice in G . For any g ∈ G and any smooth functions φ, ψ ∈ L20 (G /Γ), |hρ0 (g )φ, ψi| kφk2 kψk2 H(g )−1/b . 118 / 127 Trees and Buildings Effective mixing Let G be a connected, semisimple, linear algebraic group, defined and quasi-split over k and Γ be an non-uniform irreducible lattice in G . For any g ∈ G and any smooth functions φ, ψ ∈ L20 (G /Γ), |hρ0 (g )φ, ψi| kφk2 kψk2 H(g )−1/b . We use the Burger-Sarnak method as extended to p-adic and positive characteristic fields by 119 / 127 Trees and Buildings Effective mixing Let G be a connected, semisimple, linear algebraic group, defined and quasi-split over k and Γ be an non-uniform irreducible lattice in G . For any g ∈ G and any smooth functions φ, ψ ∈ L20 (G /Γ), |hρ0 (g )φ, ψi| kφk2 kψk2 H(g )−1/b . We use the Burger-Sarnak method as extended to p-adic and positive characteristic fields by Clozel and Ullmo 120 / 127 Trees and Buildings Effective mixing Let G be a connected, semisimple, linear algebraic group, defined and quasi-split over k and Γ be an non-uniform irreducible lattice in G . For any g ∈ G and any smooth functions φ, ψ ∈ L20 (G /Γ), |hρ0 (g )φ, ψi| kφk2 kψk2 H(g )−1/b . We use the Burger-Sarnak method as extended to p-adic and positive characteristic fields by Clozel and Ullmo Along with the solution of the Ramanujan conjecture for GL2 due to Drinfield 121 / 127 Trees and Buildings Volumes Complements of compacta µ({x ∈ G /Γ : d(x0 , x) ≥ s n }) s −κn 122 / 127 Trees and Buildings Volumes Complements of compacta µ({x ∈ G /Γ : d(x0 , x) ≥ s n }) s −κn We compute volumes of cusps directly 123 / 127 Trees and Buildings Volumes Complements of compacta µ({x ∈ G /Γ : d(x0 , x) ≥ s n }) s −κn We compute volumes of cusps directly Using an explicit description in terms of algebraic data attached to G 124 / 127 Trees and Buildings Volumes Complements of compacta µ({x ∈ G /Γ : d(x0 , x) ≥ s n }) s −κn We compute volumes of cusps directly Using an explicit description in terms of algebraic data attached to G Building on work of Harder, Soulé and others 125 / 127 Trees and Buildings Picture credits D. Sullivan, Acta Math 1982 126 / 127 Trees and Buildings Picture credits D. Sullivan, Acta Math 1982 Ulrich Görtz, Modular forms and special cycles on Shimura curves, S. Kudla, M. Rapoport and T. Yang 127 / 127