Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | x, dH is symmetric. By the definition of η(x, y) in Corol- Proof. Since x 1(1) = 0, so 1( lary 2.3.4, dH is non-negative. If x = y, then cosh− dH(x, y) = 0. Conversely, if dH(x, y) = 0, then by the definition of η(x, y), the vectors x and y must be positive scalar multiples of each other. But x and y have the same norm, which implies x = y. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The topology induced on H n by the distance function dH is the same as the subspace topology of H n, where H n is regarded as a subspace of Rn+1 with its usual metric topology. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | then x, y and z are hyperbolically collinear, that is, there is a hyperbolic line of H n containing x, y and z. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Suppose such vectors x and y exist. Then λ satisfies the differential equation λ = 0. By Proposition 2.3.11, the restriction of λ to any interval [a, b] is a λ′′ geodesic curve, and so λ is a geodesic line. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The vectors λ(0) and λ′(0) are Lorentz orthonormal by the same argument as in the proof of Proposition 2.3.11. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Hyperbolic trigonometry We now restrict our attention to the hyperbolic plane H 2, and discuss hyperbolic trigonometry. For reasons of space, the treatment here is cursory; again, see [21] for the details. In particular, the area of regions in the hyperbolic plane is not needed at all in later chapters, so we do not define area. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The trigonometric identities below may be proved using Lemma 2.3.5 and Corol- laries 2.3.6–2.3.8. In each of these identities, α, β and γ are the angles of a hyperbolic triangle, and a, b and c are the lengths of the opposite sides. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | We now generalise the definition of hyperbolic triangle to include the possibility of vertices at infinity. Suppose L1 and L2 are hyperbolic lines, with L1 the intersec- tion of H 2 with the plane span , and L2 the intersection of H 2 with the plane { is a one-dimensional subspace of Rn+1 x2, y2} x1, y1} ∩ span consisting of light-like vectors, then L1 and L2 are said to meet at infinity. Hyper- bolic lines which meet at infinity are disjoint, but become asymptotically close in one direction. A generalised hyperbolic triangle, then, is defined in the same way as a hyperbolic triangle, except that the hyperbolic lines defining adjacent sides may meet at infinity. The angle at a vertex at infinity is defined to be zero. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | This shows that there is an absolute unit of length in hyperbolic geometry, as the length c depends only on the angle α. Theorem 2.3.17. Let T be a generalised hyperbolic triangle. If the angles of T are α, β and γ, then the area of T is π − Corollary 2.3.18. The angle sum of a generalised hyperbolic triangle is less than π. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The projective disc model of hyperbolic n-space is the metric space consisting of Dn together with the distance function dD. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | A subset L of Dn is called a hyperbolic line of Dn if µ(L) is a hyperbolic line of H n. Since µ : Dn H n is an isometry, and the geodesics of H n are its hyperbolic lines, the geodesics of Dn are its hyperbolic lines. The following characterisation of the hyperbolic lines of D2 shows that in this model, hyperbolic lines are the same as Euclidean lines. Proposition 2.4.1. A subset L of D2 is a hyperbolic line of D2 if and only if L is an open chord of D2. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Let L be a hyperbolic line of H 2. Then L is the intersection of H 2 with the plane P (u, 0), where u is some unit vector in R3. Radial projection maps L onto the line of intersection of the plane P (u, 0) and the plane P (e3, 1). The translation 1(L) is an open chord of D2. This by the vector argument can be reversed for the converse. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The topology induced on Bn by dB is the same as the subspace topology of Bn as a subspace of Rn. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | A subset L of Bn is defined to be a hyperbolic line of Bn if ζ(L) is a hyperbolic line of H n. By Corollary 2.3.13, and the fact that ζ is an isometry, the geodesics of Bn are its hyperbolic lines. We now characterise the hyperbolic lines of B2. Proposition 2.5.1. A subset L is a hyperbolic line of B2 if and only if L is either an open diameter of B2, or the intersection of B2 with a circle orthogonal to S1. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Let L be an open diameter of B2. Then L is the intersection of B2 with the one-dimensional subspace of R3 spanned by, say, the vector v. The projection ζ maps L onto the hyperbolic line of H 2 obtained by intersecting H 2 with the . Thus L is a hyperbolic line of B2. This two-dimensional subspace span { argument can be reversed to show that if ζ(L) is a hyperbolic line of H 2 which passes through the point e3, then L is an open diameter of B2. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Finally, if a > 1 and a2 > 1 + r2, then choose δ > 0 so that a2 B2. We have σ( and the reflection σ does not leave B2 invariant. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Theorem 2.5.4. Every M¨obius transformation of ˆEn which leaves Bn invariant restricts to an isometry of Bn, and every isometry of Bn extends to a unique M¨obius transformation of ˆEn which leaves Bn invariant. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. The claim that M¨obius transformations restrict to isometries follows imme- diately from the explicit formula for dB together with Lemma 2.5.3. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Conversely, let φ : Bn follows. First, for any b τb leaves Bn invariant, and τb(0) = b. Next, we consider the map τ − fixes 0, and show that this map is orthogonal. It follows that τ − M¨obius transformation of ˆEn. We then prove that this extension is unique. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | formation of Bn which extends φ, and let ξ be the composition υ− continuity of ξ, we have ξ(x) = x for all x identity. Therefore τφ(0)A is the unique M¨obius extension of φ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | We can now explain why the hyperbolic angles of Bn are the same as its Eu- clidean angles. From the geometric definition of the map ζ : Bn H n, this projec- tion preserves the Euclidean angle between any two hyperbolic lines of Bn which intersect at the origin. By considering the tangent space to H n at ζ(0) = en+1, it can be proved that the hyperbolic angle between two hyperbolic lines of H n in- tersecting at ζ(0) is the same as the Euclidean angle. So, the hyperbolic angle between two hyperbolic lines of Bn which intersect at the origin is the same as the Euclidean angle. Moreover, the isometries of Bn are M¨obius transformations, which are conformal and act transitively on the set of hyperbolic lines of Bn. Therefore, the hyperbolic angle between any two intersecting hyperbolic lines in Bn is the same as the Euclidean angle between these lines. This is why Bn is called the conformal ball model of hyperbolic n-space. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | only if L is the intersection of U 2 with either a Euclidean line or a Euclidean circle which is orthogonal to the real axis. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | We will need the definitions of several groups of matrices. First, GL(n, C) is the n complex matrices. Then, GL(n, C) contains the subgroup group of invertible n SL(n, C), which is the group of n n complex matrices which have determinant 1. The groups SL(n, R) and SL(n, Z) are defined similarly. The group SO(n, R) is the group of real orthogonal matrices with determinant 1. So SO(n, R) is a subgroup of SL(n, R). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The groups P SL(n, R) and P SL(n, Z) are defined similarly. Then, if n is even, If n is an odd P SL(n, R) is SL(n, R)/ integer, we have P SL(n, R) = SL(n, R) and P SL(n, Z) = SL(n, Z). The fact that the groups P GL(2, C) are P SL(2, C) isomorphic is used to prove the following corollary. Corollary 2.6.6. The group LF ( ˆC) is isomorphic to the group P SL(2, C). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Rearranging, we obtain the equation cz2 + (d a)z + b = 0. Since this equation holds for all z, we have c = 0, d = a and b = 0. Hence, the kernel of Ξ is the . Therefore LF ( ˆC) ∼= P SL(2, C). subgroup Proposition 2.6.7. Every linear fractional transformation of ˆC is a M¨obius trans- formation of ˆC. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Since φ maps the upper half-plane onto itself, Im(φ(i)) > 0. We find that bc must be positive. bc. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | H which is both a group isomorphism and a homeomorphism. The isomorphism of LF ( ˆC) and P SL(2, C) established in Proposition 2.6.6 is an isomorphism of topological groups. Let Γ be a subgroup of a topological group G. Then Γ is said to be a discrete Γ there exists an open neighbourhood Ω of γ in G such subgroup of G if for all γ . Recall that the groups SL(2, R) and P SL(2, R) act on U 2 by that Ω Γ = isometries. We now show that SL(n, Z) is a discrete subgroup of SL(n, R), and that P SL(n, Z) is a discrete subgroup of P SL(n, R). Lemma 2.6.11. Let G be a topological group which is also a metric space, and let Γ be a subgroup of G. Suppose every convergent subsequence in Γ is eventually constant. Then Γ is discrete. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Suppose that every convergent subsequence in Γ is eventually constant, but that Γ is not discrete. Then there is a point γ 1, the ball ≥ γ, but the B(γ, 1/n) contains a point γn of Γ which is different from γ. Then γn → γn} ∞n=1 is not eventually constant, which is a contradiction. We conclude sequence that Γ must be discrete. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Corollary 2.6.14. The group P SL(n, Z) is a discrete subgroup of P SL(n, R). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | set X and let x be an element of X. The subset of X given by Gx = ∈ is called the G-orbit of x, or the orbit of x under G. The G-orbits partition X. Lemma 2.6.15. If Γ is a discrete group of isometries of a metric space X, and K is compact in X, then K contains only finitely many points of each orbit Γx. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Let K be compact in X. Suppose K contains infinitely many points of an orbit Γx. Then K contains a convergent subsequence of Γx, and there are infinitely many distinct points γnx. Since Γ is topologised with the compact open γ in Γ. As Γ is discrete, the sequence topology, γnx γnx γn} is eventually constant, that is, { } it contains only finitely many distinct elements. This is a contradiction. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | kθeri = eri. When θ = π/2, we get kθeri = e− ri. For 0 < θ < π/2, the real part of kθeri is negative. Hence, the image of [0, π/2] is those points of the circle Cr with non-positive real part. Similarly, the image of the interval [π/2, π] is those points of Cr with non-negative real part. Therefore, the orbit of the point eri under SO(2, R) is the whole circle Cr. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Let (X, d) be a metric space and let Γ be a non-trivial group of isometries of X. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | a)2 + 4bc)/2c, and Z is count- d) then by the quadratic formula z = ((a able. Some points which do have trivial stabilisers are all the points ti, where t > 1. We now show that in the upper half-plane, a Dirichlet domain with centre at p0 is a locally finite fundamental domain. To show that D(p0) is open, we will show that its complement is closed in U 2. In general, a union of closed sets need not be closed. However, if a collection of closed subsets is locally finite, then the union of the members of is closed. Proposition 2.6.18. Let Γ = P SL(2, Z) and let p0 be a point of U 2 such that Γp0 is trivial. Then the Dirichlet domain D(p0) is a locally finite fundamental domain for the action of Γ on U 2. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | D(p0). From each orbit Γz, we may select a point which is the minimum distance from the point p0 (if there is more than one closest point, choose any). Let F be the set of these points. Then F is a fundamental set for Γ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | for all γ is contained in D(p0), and so z D(p0) is a fundamental region. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Since Γ is discrete, dU (p, γp0) < 2r for only finitely many elements γ. Hence, the ball B(p0, r) meets only finitely many sets γD. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proposition 2.6.19. Let p0 = ti for some t > 1. The Dirichlet domain D(p0) is equal to T ◦. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | that is, Im(w) < Im(z). Now replace z by w and δ by γ in the above argument. We find that Im(γw) < Im(w), hence Im(z) < Im(w). This is a contradiction. Therefore, D = T ◦. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Let δT be a triangle in , and let V be the union of all the triangles γT where there exists a finite sequence γ0T, γ1T, . . . , γmT such that δT = γ0T , γT = γmT , and the triangles γi m. Since V 1T and γiT share a common side for 1 ≤ contains δT , V is non-empty. The tesselation is a locally finite collection of closed subsets of U 2. Hence, V is closed. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | of S(z, ε). This restriction is a connected tesselation, so any triangle in which has z on a side is in V . Thus V contains B(z, ε). We have shown that V is both open and closed, and is non-empty. Since U 2 is connected, this means V = U 2. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | is thus a product of the identity γ0 and elements of Σ, hence Σ generates Γ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | tesselation of X. This tesselation is locally finite, and any compact set in X meets only finitely many tiles of the tesselation. Finally, the group Γ may be generated by those finitely many elements of Γ which map D to a tile which shares a side with D. For the details of the considerable amount of theory needed to establish these results, see Chapter 4 of [8] and Chapter 6 of [21]. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Lipschitz equivalence is an equivalence relation. We write d tions d and d′ are in the same equivalence class. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The letter R is used since d is a Riemannian distance function. From the properties of the distance function d on X, it follows immediately that dR is non-negative and symmetric, that dR satisfies the triangle inequality, and that γ1 = γ2 implies dR(γ1, γ2) = 0. Now suppose dR(γ1, γ2) = 0. Then d(γ1p0, γ2p0) = 0, and so 1 γ2 is in the stabiliser γ1p0 = γ2p0. Therefore, p0 = γ− subgroup of p0. But this subgroup is trivial, hence γ1 = γ2. Thus dR really is a distance function. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The point p0 used to define the geometric distance function is not, in general, the only point of X whose stabiliser subgroup in Γ is trivial. For instance, in the upper half-plane, all points ti, where t > 1, have trivial stabiliser subgroups in P SL(2, Z). The geometric distance functions which are then defined using, say, the points 2i and 3i will not be the same. For our purposes, this difference does not matter. We are interested in Lipschitz equivalence classes of distance functions, and the following result shows that all geometric distance functions on Γ acting on the space (X, d) are Lipschitz equivalent. Lemma 3.3.1. Let Γ be a discrete group of isometries of the symmetric space (X, d). The geometric distance function on Γ depends on the choice of point in X with trivial stabiliser subgroup, but only up to Lipschitz equivalence. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | This lemma shows that changing the point p0 in the definition of a geometric distance function dR does not affect Lipschitz equivalence. Now, the definition of dR also depends on the distance function d on X. It is reasonable to wonder whether the geometric distance is well-defined, at least up to Lipschitz equivalence, if the distance function on the space X is changed. It turns out that, because X is a symmetric space, any two Riemannian distance functions d and d′ on X are scalar multiples of each other. Hence, the respective induced geometric distance functions dR and d′R are indeed Lipschitz equivalent. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | If x is in X and γ1 in Γ, the distance in X along the geodesic joining x and γ1x Γ such that γ2x = p0. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | So the geometric distance provides a way of relating a distance in the space X between points x and γx to a distance in the group Γ. In particular, the distance along the geodesic joining the points p0 and γp0 is Lipschitz equivalent to dR(1, γ). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | proving the triangle inequality. Lemma 3.3.2. The word distance function on Γ depends on the choice of genera- tors, but only up to Lipschitz equivalence. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | As with the geometric distance function, this result allows us to abuse notation. If we are interested only in Lipschitz equivalence of distance functions, we will now write dW for any word distance function, and refer to dW as the word distance function on Γ. We will continue to denote by dΣ the word distance function induced by a particular set of generators Σ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | is a word for γ with respect to a finite set of generators for Γ. Its length is the same as the number of tiles of the tesselation which are crossed by the geodesic segment [p0, γp0]. There is no shorter word for γ with respect to this set of generators (this is proved for a case when Γ = P SL(2, Z) in Section 4.3, and the proof given there generalises). Therefore, dW (1, γ) has a geometric significance as well as its algebraic meaning: it is Lipschitz equivalent to the number of tiles in the tesselation which are crossed by the geodesic segment [p0, γp0]. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The following theorem shows that this intuition is correct. In spaces, like this one, where the induced tesselation has compact tiles, the geometric and word distance functions are Lipschitz equivalent. Theorem 3.4.1. [14] Let D be a fundamental domain for the action of Γ on (X, d). Suppose D has compact closure. Then the geometric distance function dR on Γ is Lipschitz equivalent to the word distance function dW on Γ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Since the closed ball B(p0, K0 + R0) is compact, it meets only finitely many images of D. Thus Θ is a finite set. Therefore, if x and y are points in D, and the point θy is in the ball B(x, K0), the element θ belongs to the finite set Θ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | min λi < 1 then det(A) < 1, so min λi = 1. Therefore, λi = 1 for all i, and A is the identity matrix. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | which is the special orthogonal group SO(n, R). This fact is used to prove the following decomposition of SL(n, R). Lemma 3.5.2. Let M SL(n, R). Then there exist matrices K1 and K2 in SO(n, R), and a diagonal matrix A in SL(n, R) with positive diagonal entries, such that M = K1AK2. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Let G = SL(2, R) and K = SO(2, R). Recall that K is the stabiliser subgroup of i when G is acting on U 2 (see the proof of Lemma 2.6.16), and of I when G is acting on P (2). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | We now, finally, define the action of the group P SL(n, R) on P (n). Let M be in SL(n, R) and S in P (n). Then MSM T = ( M)T , so M S. ◦ − S = MSM T is a Write [M] for the coset of M in the group P SL(n, R). Then [M] well-defined group action of P SL(n, R) on the space P (n). So, the group P SL(n, R) acts transitively and by isometries on the metric space P (n). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | exists a constant K > 0 such that dW (1, γ) k operator norm of the matrix g. This step is Theorem 4.4.1. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | For the word distance function, it will often be convenient to consider words in SL(d, Z), write [g] for the the group SL(d, Z) rather than in P SL(d, Z). For g gn is a word for g in terms of elements of coset of g in P SL(d, Z). Then if g1g2 · · · [gn]. Going in the other direction, if SL(d, Z), a word for the coset [g] is [g1][g2] gn, [g] = [g1][g2] I adds 1 to the word length, or g = ( which will not affect Lipschitz equivalence. Thus, it is reasonable to move back and forth between words in SL(d, Z) and words in P SL(d, Z) when constructing words and proving results about the word distance function dW . We will usually write SL(d, Z) such that P SL(d, Z), and for an element g γ both for the coset γ [g] = γ; we may describe γ as a matrix rather than a coset. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | where the signs in the last set of matrices are arranged so that the determinant in each case is 1. The proof that the sets Σd do generate the groups P SL(d, Z) is by induction on d. For the inductive step, it is possible by multiplying by elements to transform a matrix γ in P SL(d, Z) into a matrix θ which has first of Σd ∪ column (1, 0, . . . , 0)T and first row (1, 0, . . . , 0). Then, by the inductive assumption, 1) matrix in the lower right-hand corner of θ may be written the (d × in terms of elements of Σd 1 may then be 1. The set of generators Σd − considered a subset of Σd, by embedding P SL(d 1, Z) in the lower right-hand corner of P SL(d, Z). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | and let dΣ be the word distance function induced by the set of generators 2. n, we consider the tesselation of the upper half-plane induced by the action of Γ = P SL(2, Z). Recall that the point p0 = 2i lies in the tile T , where T is the hyperbolic triangle with 2 + √3 wl be any word for un in terms of the . Let w1w2 · · · 2 i and vertices at ± 1. Let W0 be the identity in Γ. For 1 l, let Wj be the partial elements of Σ Σ− ∪ ≤ l wj. Then the sequence of partial words, word w1w2 · · · j=0, corresponds to a j=0, such that W0T = T and WlT = unT . In this sequence l sequence of tiles, of tiles, for each j such that 1 1T and WjT share a boundary. Z, let Lk be the geodesic of U 2 which is given by Re(z) = k + 1/2. Then Lk For k is a boundary between pairs of tiles. See the picture of the tesselation on page 9. Since unp0 = n + 2i, the n boundaries L0, L1, . . . , Ln 1 all lie between the points p0 and unp0. Thus, the sequence l j=0 must contain at least n successive pairs } of tiles which have a geodesic Lk as their common boundary. Therefore, l n. So n; hence dΣ(1, un) = n. This implies that the word distance dW (1, un) dΣ(1, un) ≥ grows linearly in n. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | As claimed in the introduction to this chapter, the proof in Theorem 3.4.1 suffices. This is because the proof of inequality (4.1) in Theorem 3.4.1 did not depend on the compactness of tiles in the tesselation induced by Γ. Indeed, inequality (4.1) holds in the case Γ = P SL(2, Z). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. Since x is unimodular, x just put γ1 = I. If x1 = 0, then xj 6 the first component of γ1x is non-zero. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | ad = gcd(x′1, x2, . . . , xd) = gcd(x1, x2, . . . , xd) = 1. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Then p is a fixed polynomial in the components of x, and the entries of each γi, for d, are bounded by p. The matrix γ1 is either the identity, or E1j(1) for 2 some j, so in either case the entries of γ1 are bounded by 1. The entries of γ1 are thus bounded by the polynomial p as well. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. We prove this corollary by induction on d. The idea is to multiply γ by appropriate matrices on the left to obtain a matrix γ′ with first column (1, 0, . . . , 0)T . Then, we multiply γ′ by appropriate matrices on the right to obtain a matrix γ′′ with first column (1, 0, . . . , 0)T and first row (1, 0, . . . , 0). The inductive assumption is then applied to γ′′. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | is an element of SL(d, Z) which has as its first column (1, 0, . . . , 0)T . Also by Lemma 4.4.2, each of the matrices γi has entries bounded by the fixed polynomial p in the components xj, where 1 d. Since xj = γj1, which is an entry in the first column of the matrix γ, we may now consider p to be a fixed polynomial in the entries of γ. Therefore, each matrix γi has entries bounded by a fixed polynomial p in the entries of γ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Thus the entries of the matrices of the form γ′j, for 2 fixed polynomial (p(γ)q(γ))2 + 1 in the entries of γ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Applying the inductive hypothesis to the matrix γ′′, we may write γ′′ as a 1)2, each δk belongs to some product γ′′ = δ1δ2 · · · ≤ SLs,t(2, Z), and each δk has entries bounded by a fixed polynomial in the entries of the (d 1) submatrix in the lower right-hand corner of γ′′. So we may 1) consider each δk to have entries bounded by a fixed polynomial in the entries of γ′′. Now, γ′′ is the product of the matrix γ, and matrices of the form γi and γ′j. So, each entry of γ′′ is some polynomial in the entries of γ, the γi and the γ′j. But the matrices γ, γi and γ′j all have entries bounded by fixed polynomials in the entries of γ. This means the entries of γ′′ are bounded by a (higher degree) fixed polynomial in the entries of γ. So we can find a fixed polynomial P in the entries of γ which is sufficiently large to bound the entries of the γi, the γ′j and the δk. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | inverse of γ′j = are unchanged. We conclude that γ may be expressed as a product of the required form. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. First, we show why it suffices to prove this result for just the case where γ is E13(1) SL(3, Z). Then, we identify a subgroup V of SL(3, R) with the vector space R2, and identify a discrete subgroup of V with the lattice Z2. The matrix Z2 such that every E13(1) is in this discrete subgroup. We construct a set S point of Z2 is at a bounded distance from some point of S. For a vector v S, we find a bound on the norm of v, and a bound on the word length of the element of SL(3, R) which corresponds to v. This allows us to bound the word length of those elements of SL(3, R) which are identified with the lattice Z2. To make the proof of this proposition more digestible, most of these steps are presented as lemmas. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | To show why it suffices to prove this proposition for just the case where γ is SL(3, Z), let γ and δ be elements of SL(d, Z), and suppose that γ is U1. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | If n = 0 then both sides are equal to 0. This shows that if γ is U1, so are all conjugates of γ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The relationship between the sets S1 and S2 and the eigenspaces W1 and W2 is described in the following lemma. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Lemma 4.4.5. For k = 1, 2, each point of the eigenspace Wk is within a bounded distance of some point of the set Sk. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | and if we put ai = 0 for l < i < k, we obtain a set ak 6 is bounded. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | To show that every point of W2 is within a bounded distance of some point of S2, the proof is similar. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Let S be the set S1 + S2. Note that S is a subset of the integer lattice L. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Corollary 4.4.6. Each point of L is within a bounded distance of some point in S. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Proof. The eigenvectors v1 and v2 are linearly independent, so R2 = W1 + W2. Each point of L, then, can be written as w1 + w2 for some w1 ∈ W2. By Lemma 4.4.5, w1 is within a bounded distance of some point of S1, and w2 is within a bounded distance of some point of S2. Thus w1 + w2 is within a bounded distance of some point of S1 + S2. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | We now establish a lower bound on the norm of vectors in S. Let v be in S. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Note that the value of M in Lemma 4.4.7 depends only on the angle between the vectors w1 and w2. When we use Lemma 4.4.7 in the next proof, the value of M will thus depend only on the angle between the fixed eigenvectors v1 and v2. So M, as far as it is used here, is a constant. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Corollary 4.4.6 and Lemma 4.4.8 will be used in the conclusion of the proof of Proposition 4.4.4. Before this, we return to the interpretation of V as a subgroup of SL(3, Z). If v S is considered as an element of SL(3, Z), then its word length is bounded as follows. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Lemma 4.4.9. There exists a constant c2 > 0 such that, for any v in S, if k is the maximum of m and n where v has the form (4.7), then v may be written as a word in the elements A and Y0 of SL(3, Z) of length at most c2k. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | pj} a sequence ≤ and pj = θjp0. In this product, no two consecutive matrices rj 1) and rj = u(nj), where nj respective forms rj integers. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | The following lemma shows that, roughly speaking, the length of α is not too different from the length of the geodesic joining the points p0 and δp0. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | and so cos φ < cos φ′. This completes the proof of the case p′ = 1 + 2i. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | So q = a + reiφ and q′ = a + reiφ′ . To establish (4.10), it suffices to prove that φ′ > φ. Since cos φ < 0 and cos φ′ < 0, we wish to show that cos φ > cos φ′; this is again a straightforward inequality. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | In the special case where δ is the identity, Lemma 4.4.11 is trivial. The other special case is when δ = v. Here, we have dU (p0, p1) = dU (q0, q1), so the path α is identical to the geodesic segment [p0, δp0]. Provided K2 ≥ 1, there is nothing more to prove in this case. So if, later, we arrive at a value of K2 which is less than 1, we will take K2 = 1 instead. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | and which, since p0 ∈ may, as above, parametrise z 0 < φ < π. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | Now, we find the hyperbolic distance between q and q′ in terms of a, r and n. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | where K > 0 is a constant. This implies a lower bound on dU (q, q′). |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | k rm be a word for δ as in Proposition 4.4.10, where each rj belongs Proof. Let r1r2 · · · to SLs,t(2, Z). Let Σ be a fixed finite set of generators of SL(d, Z), and let dΣ be the word distance function on SL(d, Z) induced by Σ. |
Processed_Hyperbolic_Geometry_and_Distance_Functions_on_Disc.txt | In Jeremy Gray, editor, The Symbolic Universe: Geometry and Physics 1890– 1930, pages 128–144. Oxford University Press, 1999. |
Processed_Hazard_function_models_to_estimate_mortality_rates.txt | data were derived to estimate a constant natural mortality rate, catchability and age-specific gear selectivity. |
Processed_Hazard_function_models_to_estimate_mortality_rates.txt | that the provision of effort data allowed to estimate both catchability (q) and natural mortality. |
Processed_Hazard_function_models_to_estimate_mortality_rates.txt | t=ai+1 theless, given effort in every interval ( t=ai R t=∞ t=0 f2(t; θ) dt. |
Processed_Hazard_function_models_to_estimate_mortality_rates.txt | cohort, random constant natural mortality, random catchability and random fishing effort in each year (Tab. 1). |
Processed_Hazard_function_models_to_estimate_mortality_rates.txt | Table 1: Distribution and range of value taken by different type of random variable in simulations. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.