Chapter 7: Tangent Planes

Chapter 7 Tangent Planes Reading: Do Carmo sections 2.4 and 3.2 Today I am discussing 1. Differentials of maps between surfaces 2. Geometry of Gauss map 7.1 Tangent Planes; Differentials of Maps Between Surfaces 7.1.1 Tangent Planes Recall from previous lectures the definition of tangent plane. (Proposition 2-4-1). Let x : U ⊂ R2 → S be a parameterization of a regular surface S and let q ∈ U . The vector subspace of dimension 2, dxq (R2 ) ⊂ R3 (7.1) coincides with the set of tangent vectors to S at x(q). We call the plane dxq (R2 ) the Tangent Plane to S at p, denoted by Tp (S). 37 p α’(0) S α x v β’(0) −ε 0 q ε t β U u Figure 7.1: Graphical representation of the map dxq that sends β ′ (0) ∈ Tq (R2 to α′ (0) ∈ Tp (S). Note that the plane dxq (R2 ) does not depend on the parameterization x. However, the choice of the parameterization determines the basis on Tp (S), )(q), ( ∂x )(q)}, or {xu (q), xv (q)}. namely {( ∂x ∂u ∂v 7.1.2 Coordinates of w ∈ Tp (S) in the Basis Associated to Parameterization x Let w be the velocity vector α′ (0), where α = x ◦ β is a curve in the surface S, and the map β : (−ǫ, ǫ) → U , β(t) = (u(t), v(t)). Then in the basis of {xu (q), xv (q)}, we have w = (u′ (0), v ′ (0)) (7.2) 38 7.1.3 Differential of a (Differentiable) Map Between Surfaces It is natural to extend the idea of differential map from T (R2 ) → T (S) to T (S1 ) → T (S2 ). Let S1 , S2 be two regular surfaces, and a differential mapping ϕ ⊂ S1 → S2 where V is open. Let p ∈ V , then all the vectors w ∈ Tp (S1 ) are velocity vectors α′ (0) of some differentiable parameterized curve α : (−ǫ, ǫ) → V with α(0) = p. Define β = ϕ ◦ α with β(0) = ϕ(p), then β ′ (0) is a vector of Tϕ(p) (S2 ). (Proposition 2-4-2). Given w, the velocity vector β ′ (0) does not depend on the choice of α. Moreover, the map dϕp : Tp (S1 ) → Tϕ(p) (S2 ) dϕp (w) = β ′ (0) (7.3) (7.4) is linear. We call the linear map dϕp to be the differential of ϕ at p ∈ S1 . Proof. Suppose ϕ is expressed in ϕ(u, v) = (ϕ1 (u, v), ϕ2 (u, v)), and α(t) = (u(t), v(t)), t ∈ (−ǫ, ǫ) is a regular curve on the surface S1 . Then β(t) = (ϕ1 (u(t), v(t)), ϕ2 (u(t), v(t)). (7.5) Differentiating β w.r.t. t, we have ′ β (0) = µ ¶ ∂ϕ1 ′ ϕ1 ′ ∂ϕ2 ′ ϕ2 ′ u (0) + v (0), u (0) + v (0) ∂u ∂v ∂u ∂v (7.6) in the basis of (x̄u , x̄v ). As shown above, β ′ (0) depends on the map ϕ and the coordinates of (u′ (0), v ′ (0) in the basis of {xu , xv }. Therefore, it is independent on the choice of α. 39 ϕ ϕ(p) α’(0) β ’(0) S2 V p S1 α β x x v v −ε 0 ε t q q u u x Figure 7.2: Graphical representation of the map dϕp that sends α′ (0) ∈ Tq (S1 ) to β ′ (0) ∈ Tp (S2 ). Moreover, Equation 7.6 can be expressed as à ∂ϕ1 ∂u β ′ (0) = dϕp (w) = ∂ϕ2 ∂u ! ∂ϕ1 ! à ′ u (0) ∂v ∂ϕ2 v ′ (0) ∂v (7.7) which shows that the map dϕp is a mapping from Tp (S1 ) to Tϕ(p) (S2 ). Note that the 2 × 2 matrix is respect to the basis {xu , xv } of Tp (S1 ) and {x̄u , x̄v } of Tϕ(p) (S2 ) respectively. We can define the differential of a (differentiable) function f : U ⊂ S → R at p ∈ U as a linear map dfp : Tp (S) → R. Example 2-4-1: Differential of the height function Let v ∈ R3 . Con40 sider the map h : S ⊂ R3 → R (7.8) h(p) = v · p, p ∈ S (7.9) We want to compute the differential dhp (w), w ∈ Tp (S). We can choose a differential curve α : (−ǫ, ǫ)) → S such that α(0) = p and α′ (0) = w. We are able to choose such α since the differential dhp (w) is independent on the choice of α. Thus h(α(t)) = α(t) · v (7.10) Taking derivatives, we have dhp (w) = d h(α(t))|t=0 = α′ (0) · v = w · v dt (7.11) Example 2-4-2: Differential of the rotation Let S 2 ⊂ R3 be the unit sphere S 2 = {(x, y, z) ∈ R3 ; x2 + y 2 + z 2 = 1} (7.12) Consider the map Rz,θ : R3 → R3 (7.13) be the rotation of angle θ about the z axis. When Rz,θ is restricted to S 2 , it becomes a differential map that maps S 2 into itself. For simplicity, we denote the restriction map Rz,θ . We want to compute the differential (dRz,θ )p (w), p ∈ S 2 , w ∈ Tp (S 2 ). Let α : (−ǫ, ǫ) → S 2 be a curve on S 2 such that α(0) = p, α′ (0) = w. Now (dRz,θ )(w) = d (Rz,θ ◦ α(t))t=0 = Rz,θ (α′ (0)) = Rz,θ (w) dt 41 (7.14) 7.1.4 Inverse Function Theorem All we have done is extending differential calculus in R2 to regular surfaces. Thus, it is natural to have the Inverse Function Theorem extended to the regular surfaces. A mapping ϕ : U ⊂ S1 → S2 is a local diffeomorphism at p ∈ U if there exists a neighborhood V ⊂ U of p, such that ϕ restricted to V is a diffeomorphism onto the open set ϕ(V ) ⊂ S2 . (Proposition 2-4-3). Let S1 , S2 be regular surfaces and ϕ : U ⊂ S1 → S2 a differentiable mapping. If dϕp : Tp (S1 ) → Tϕ(p) (S2 ) at p ∈ U is an isomorphism, then ϕ is a local diffeomorphism at p. The proof is a direct application of the inverse function theorem in R2 . 7.2 The Geometry of Gauss Map In this section we will extend the idea of curvature in curves to regular surfaces. Thus, we want to study how rapidly a surface S pulls away from the tangent plane Tp (S) in a neighborhood of p ∈ S. This is equivalent to measuring the rate of change of a unit normal vector field N on a neighborhood of p. We will show that this rate of change is a linear map on Tp (S) which is self adjoint. 7.2.1 Orientation of Surfaces Given a parameterization x : U ⊂ R2 → S of a regular surface S at a point p ∈ S, we choose a unit normal vector at each point x(U ) by N (q) = xu ∧ xv (q), q ∈ x(U ) |xu ∧ xv | (7.15) We can think of N to be a map N : x(U ) → R3 . Thus, each point q ∈ x(U ) has a normal vector associated to it. We say that N is a differential field 42 of unit normal vectors on U . We say that a regular surface is orientable if it has a differentiable field of unit normal vectors defined on the whole surface. The choice of such a field N is called an orientation of S. An example of non-orientable surface is Möbius strip (see Figure 3). Figure 7.3: Möbius strip, an example of non-orientable surface. In this section (and probably for the rest of the course), we will only study regular orientable surface. We will denote S to be such a surface with an orientation N which has been chosen. 7.2.2 Gauss Map (Definition 3-2-1). Let S ⊂ R3 be a surface with an orientation N and S 2 ⊂ R3 be the unit sphere S 2 = {(x, y, z) ∈ R3 ; x2 + y 2 + z 2 = 1}. (7.16) The map N : S → S 2 is called the Gauss map. The map N is differentiable since the differential, dNp : Tp (S) → TN (p) (S 2 ) 43 (7.17) at p ∈ S is a linear map. For a point p ∈ S, we look at each curve α(t) with α(0) = p and compute N ◦ α(t) = N (t) where we define that map N : (−ǫ, ǫ) → S 2 with the same notation as the normal field. By this method, we restrict the normal vector N to the curve α(t). The tangent vector N ′ (0) ∈ Tp (S 2 ) thus measures the rate of change of the normal vector N restrict to the curve α(t) at t = 0. In other words, dNp measure how N pulls away from N (p) in a neighborhood of p. In the case of the surfaces, this measure is given by a linear map. Example 3-2-1 (Trivial) Consider S to be the plane ax + by + cz + d = 0, the tangent vector at any point p ∈ S is given by N=√ (a, b, c) + b2 + c 2 a2 (7.18) Since N is a constant throughout S, dN = 0. Example 3-2-2 (Gauss map on the Unit Sphere) Consider S = S 2 ⊂ R3 , the unit sphere in the space R3 . Let α(t) = (x(t), y(t), z(t)) be a curve on S, then we have 2xx′ + 2yy ′ + 2zz ′ = 0 (7.19) which means that the vector (x, y, z) is normal to the surface at the point (x,y,z). We will choose N = (−x, −y, −z) to be the normal field of S. Restricting to the curve α(t), we have N (t) = (−x(t), −y(t), −z(t)) (7.20) dN (x′ (t), y ′ (t), z ′ (t)) = (−x′ (t), −y ′ (t), −z ′ (t)) (7.21) and therefore or dNp (v) = −v for all p ∈ S and v ∈ Tp (S 2 ). 44 Example 3-2-4 (Exercise: Gauss map on a hyperbolic paraboloid) Find the differential dNp=(0,0,0) of the normal field of the paraboloid S ⊂ R3 defined by x(u, v) = (u, v, v 2 − u2 ) (7.22) under the parameterization x : U ⊂ R2 → S. 7.2.3 Self-Adjoint Linear Maps and Quadratic Forms Let V now be a vector space of dimension 2 endowed with an inner product h , i. Let A : V → V be a linear map. If hAv, wi = hv, Awi for all v, w ∈ V , then we call A to be a self-adjoint linear map. Let {e1 , e2 } be a orthonormal basis for V and (αij ), i, j = 1, 2 be the matrix elements of A in this basis. Then, according to the axiom of selfadjoint, we have hAei , ej i = αij = hei , Aej i = hAej , ei i = αji (7.23) There A is symmetric. To each self-adjoint linear map, there is a bilinear map B : V × V → R given by B(v, w) = hAv, wi (7.24) It is easy to prove that B is a bilinear symmetric form in V . For each bilinear form B in V , there is a quadratic form Q : V → R given by Q(v) = B(v, v), v ∈ V. (7.25) Exercise (Trivial): Show that B(u, v) = 1 [Q(u + v) − Q(v) − Q(u)] 2 45 (7.26) Therefore, there is a 1-1 correspondence between quadratic form and selfadjoint linear maps of V . Goal for the rest of this section: Show that given a self-adjoint linear map A : V → V , there exist a orthonormal basis for V such that, relative to this basis, the matrix A is diagonal matrix. Moreover, the elements of the diagonal are the maximum and minimum of the corresponding quadratic form restricted to the unit circle of V . (Lemma (Exercise)). If Q(x, y) = ax2 = 2bxy+cy 2 restricted to {(x, y); x2 + y 2 = 1} has a maximum at (1, 0), then b = 0 Hint: Reparametrize (x, y) using x = cos t, y = cos t, t ∈ (−ǫ, 2π + ǫ) and set dQ | = 0. dt t=0 (Proposition 3A-1). Given a quadratic form Q in V , there exists an orthonormal basis {e1 , e2 } of V such that if v ∈ V is given by v = xe1 + ye2 , then Q(v) = λ1 x2 + λ2 y 2 (7.27) where λi , i = 1, 2 are the maximum and minimum of the map Q on |v| = 1 respectively. Proof. Let λ1 be the maximum of Q on the circle |v| = 1, and e1 to be the unit vector with Q(e1 ) = λ1 . Such e1 exists by continuity of Q on the compact set |v| = 1. Now let e2 to be the unit vector orthonormal to e1 , and let λ2 = Q(e2 ). We will show that this set of basis satisfy the proposition. Let B be a bilinear form associated to Q. If v = xe1 + ye2 , then Q(v) = B(v, v) = λ1 x2 + 2bxy + λ2 y 2 (7.28) where b = B(e1 , e2 ). From previous lemma, we know that b = 0. So now it suffices to show that λ2 is the minimum of Q on |v| = 1. This is trivial since 46 we know that x2 + y 2 = 1 and Q(v) = λ1 x2 + λ2 y 2 ≥ λ2 (x2 + y 2 ) = λ2 (7.29) as λ2 ≤ λ1. If v 6= 0, then v is called the eigenvector of A : V → V if Av = λv for some real λ. We call the λ the corresponding eigenvalue. (Theorem 3A-1). Let A : V → V be a self-adjoint linear map, then there exist an orthonormal basis {e1 , e2 } of V such that A(e1 ) = λ1 e1 , A(e2 ) = λ2 e2 . (7.30) Thus, A is diagonal in this basis and λi , i = 1, 2 are the maximum and minimum of Q(v) = hAv, vi on the unit circle of V . Proof. Consider Q(v) = hAv, vi where v = (x, y) in the basis of ei , i = 1, 2. Recall from the previous lemma that Q(x, y) = ax2 + cy 2 for some a, c ∈ R. We have Q(e1 ) = Q(1, 0) = a, Q(e2 ) = Q(0, 1) = c, therefore Q(e1 + e2 ) = Q(1, 1) = a + c and 1 B(e1 , e2 ) = [Q(e1 + e2 ) − Q(e1 ) − Q(e2 )] = 0 2 (7.31) Thus, Ae1 is either parallel to e1 or equal to 0. In any case, we have Ae1 = λ1 e1 . Using B(e1 , e2 ) = hAe2 , e1 i = 0 and hAe2 , e2 i = λ2 , we have Ae2 = λ2 e2 . Now let us go back to the discussion of Gauss map. (Proposition 3-2-1). The differential map dNp : Tp (S) → Tp (S) of the Gauss map is a self-adjoint linear map. Proof. It suffices to show that hdNp (w1 ), w2 i = hw1 , dNp (w2 )i 47 (7.32) for the basis {w1 , w2 } of Tp (S). Let x(u, v) be a parameterization of S at p, then xu , xv is a basis of Tp (S). Let α(t) = x(u(t), v(t)) be a parameterized curve in S with α(0) = p, we have dNp (α′ (0)) = dNp (xu u′ (0) + xv v ′ (0)) d = N (u(t), v(t))|t=0 dt = Nu u′ (0) + Nv v ′ (0) (7.33) (7.34) (7.35) with dNp (xu ) = Nu and dNp (xv ) = Nv . So now it suffices to show that hNu , xv i = hxu , Nv i (7.36) If we take the derivative of hN, xu i = 0 and hN, xv i = 0, we have hNv , xu i + hN, xu vi = 0 (7.37) hNu , xv i + hN, xv ui = 0 (7.38) hNu , xv i = − hN, xu vi = hNv , xu i (7.39) Therefore 48