92 Chapter 3. <strong>Surfaces</strong>: Further Topics*4. Us<strong>in</strong>g Proposition 1.2, prove that horizontal l<strong>in</strong>es have constant geodesic curvature κ g =1<strong>in</strong>H; prove that this is also true for circles tangent to the u-axis. (These curves are classicallycalled horocycles.)5. Prove that every curve <strong>in</strong> H of constant geodesic curvature κ g =1is of the form given <strong>in</strong>Exercise 4. (H<strong>in</strong>ts: Assume we start with an arclength parametrization (u(s),v(s)), <strong>and</strong> useProposition 1.2 to show that we have 1 = u′v + θ′ <strong>and</strong> u ′2 + v ′2 = v 2 . Obta<strong>in</strong> the differentialequation(v d2 v( dv) ) 2 3/2 ( dv) 2du 2 = 1+− − 1,dudu<strong>and</strong> solve this by substitut<strong>in</strong>g z = dv/du <strong>and</strong> gett<strong>in</strong>g a separable differential equation fordz/dv.)6. Let T a,b,c,d (z) = az + b , a, b, c, d ∈ R, with ad − bc =1.cz + da. Suppose a ′ ,b ′ ,c ′ ,d ′ ∈ R <strong>and</strong> a ′ d ′ − b ′ c ′ =1. Check thatT a ′ ,b ′ ,c ′ ,d ′◦ T a,b,c,d = T a ′ a+b ′ c,a ′ b+b ′ d,c ′ a+d ′ c,c ′ b+d ′ d(a ′ a + b ′ c)(c ′ b + d ′ d) − (a ′ b + b ′ d)(c ′ a + d ′ c)=1.Show, moreover, that T d,−b,−c,a = T −1a,b,c,d . (Note that T a,b,c,d = T −a,−b,−c,−d . The readerwho’s taken group theory will recognize that we’re def<strong>in</strong><strong>in</strong>g an isomorphism between thegroup of l<strong>in</strong>ear fractional transformations <strong>and</strong> the group SL(2, R)/{±I} of 2 × 2 matriceswith determ<strong>in</strong>ant 1, identify<strong>in</strong>g a matrix <strong>and</strong> its additive <strong>in</strong>verse.)b. Let T = T a,b,c,d . Prove that if z = u + iv <strong>and</strong> v>0, then T (z) =x + iy with y>0.Deduce that T maps H to itself bijectively.7. Show that reflection across the geodesic u =0isgiven by r(z) =−z. Use this to determ<strong>in</strong>ethe form of the reflection across a general geodesic.8. The geodesic circle of radius R centered at P is the set of po<strong>in</strong>ts Q so that d(P, Q) =R. Provethat geodesic circles <strong>in</strong> H are Euclidean circles. One way to proceed is as follows: The geodesiccircle centered at P =(0, 1) with radius R =lna must pass through (0,a) <strong>and</strong> (0, 1/a), <strong>and</strong>hence ought to be a Euclidean circle centered at (0, 1 2(a +1/a)). Check that all the po<strong>in</strong>ts onthis circle are <strong>in</strong> fact a hyperbolic distance R away from P . (H<strong>in</strong>t: It is probably easiest towork with the cartesian equation of the circle.)Remark. This result can be established rather pa<strong>in</strong>lessly by us<strong>in</strong>g the Po<strong>in</strong>caré disk modelgiven <strong>in</strong> Exercise 15.<strong>and</strong>*9. What is the geodesic curvature of a geodesic circle of radius R <strong>in</strong> H? (See Exercises 4 <strong>and</strong> 8.)10. Recall (see, for example, p. 298 <strong>and</strong> pp. 350–1 of Shifr<strong>in</strong>’s Abstract Algebra: A GeometricApproach) that the cross ratio of four numbers A, B, P, Q ∈ C ∪ {∞} is def<strong>in</strong>ed to be[A : B : P : Q] = Q − A / Q − BP − A P − B .
§2. An Introduction to Hyperbolic Geometry 93a. Show that A, B, P , <strong>and</strong> Q lie on a l<strong>in</strong>e or circle if <strong>and</strong> only if their cross ratio is a realnumber.b. Prove that if S is a l<strong>in</strong>ear fractional transformation with S(A) = 0, S(B) = ∞, <strong>and</strong>S(P )=1, then S(Q) =[A : B : P : Q]. Use this to deduce that for any l<strong>in</strong>ear fractionaltransformation T ,wehave[T (A) :T (B) :T (P ):T (Q)] = [A : B : P : Q].c. Prove that l<strong>in</strong>ear fractional transformations map l<strong>in</strong>es <strong>and</strong> circles to either l<strong>in</strong>es or circles.(For which such transformations do l<strong>in</strong>es necessarily map to l<strong>in</strong>es?)d. Show that if A, B, P , <strong>and</strong> Q lie on a l<strong>in</strong>e or circle, then[A : B : P : Q] = AQ / BQAP BP .Conclude that d(P, Q) =| ln[A : B : P : Q]|, where A, B, P , <strong>and</strong> Q are as illustrated <strong>in</strong>Figure 2.2.e. Check that if T is a l<strong>in</strong>ear fractional transformation carry<strong>in</strong>g A to 0, B to ∞, P to P ′ ,<strong>and</strong> Q to Q ′ , then d(P, Q) =d(P ′ ,Q ′ ).11. a. Let O be any po<strong>in</strong>t not ly<strong>in</strong>g on a circle C <strong>and</strong> let P <strong>and</strong> Q be po<strong>in</strong>ts on the circle C sothat O, P , <strong>and</strong> Q are coll<strong>in</strong>ear. Let T be the po<strong>in</strong>t on C so that OT is tangent to C. Provethat (OP)(OQ) =(OT) 2 .b. Def<strong>in</strong>e <strong>in</strong>version <strong>in</strong> the circle of radius R centered at O by send<strong>in</strong>g a po<strong>in</strong>t P to the po<strong>in</strong>tP ′ on the ray OP with (OP)(OP ′ )=R 2 . Show that an <strong>in</strong>version <strong>in</strong> a circle centered atthe orig<strong>in</strong> maps a circle C centered on the u-axis <strong>and</strong> not pass<strong>in</strong>g through O to anothercircle C ′ centered on the u-axis. (H<strong>in</strong>t: For any P ∈ C, let Q be the other po<strong>in</strong>t on Ccoll<strong>in</strong>ear with O <strong>and</strong> P , <strong>and</strong> let Q ′ be the image of Q under <strong>in</strong>version. Use the result ofpart a to show that OP/OQ ′ is constant. If C is the center of C, let C ′ be the po<strong>in</strong>t onthe u-axis so that C ′ Q ′ ‖CP. Show that Q ′ traces out a circle C ′ centered at C ′ .)c. Show that <strong>in</strong>version <strong>in</strong> the circle of radius R centered at O maps vertical l<strong>in</strong>es to circlescentered on the u-axis <strong>and</strong> pass<strong>in</strong>g through O <strong>and</strong> vice-versa.12. a. Prove that every (orientation-preserv<strong>in</strong>g) isometry of H can be written as the compositionof l<strong>in</strong>ear fractional transformations of the formT 1 (z) =z + b for some b ∈ R <strong>and</strong> T 2 (z) =− 1 z .(H<strong>in</strong>t:[It’s]probably[ ]easiest to work[with]matrices. Show that you have matrices of the1 b 0 −11 0form , , <strong>and</strong> therefore , <strong>and</strong> that any matrix of determ<strong>in</strong>ant 1 can be0 1 1 0b 1obta<strong>in</strong>ed by row <strong>and</strong> column operations from these.)b. Prove that T 2 maps circles centered on the u-axis <strong>and</strong> vertical l<strong>in</strong>es to circles centered onthe u-axis <strong>and</strong> vertical l<strong>in</strong>es (not necessarily respectively). Either do this algebraically oruse Exercise 11.c. Use the results of parts a <strong>and</strong> b to prove that isometries of H map geodesics to geodesics.13. We say a l<strong>in</strong>ear fractional transformation T = T a,b,c,d is elliptic if it has one fixed po<strong>in</strong>t, parabolicif it has one fixed po<strong>in</strong>t at <strong>in</strong>f<strong>in</strong>ity, <strong>and</strong> hyperbolic if it has two fixed po<strong>in</strong>ts at <strong>in</strong>f<strong>in</strong>ity.a. Show that T is elliptic if |a + d| < 2, parabolic if |a + d| =2,<strong>and</strong> hyperbolic if |a + d| > 2.