Math 147, Homework 2 Solutions Due: April 17, 2012 1. Show that 0 is the only critical value of the map f : R3 → R1 defined by: f (x, y, z) = x2 + y 2 − z 2 . Prove that if a and b are either both positive or both negative, then f −1 (a) and f −1 (b) are diffeomorphic and sketch what happens at f −1 (0). Solution. The derivative of f is df(x,y,z) = 2x 2y −2z . Note that df(x,y,z) has rank less than 1 only if (x, y, z) = (0, 0, 0). Hence (0, 0, 0) is the only critical point of f and f (0, 0, 0) = 0 is the only critical value of f . Suppose that a and b are either both positive or both negative (we will handle both cases at once). Let G : R3 → R3 be defined by p p p G(x, y, z) = (x b/a, y b/a, z b/a). p That is, G is scalar multiplication by b/a. Then G is clearly a diffeomorphism from R3 to R3 , and the restriction of G to f −1 (a) is a diffeomorphism from f −1 (a) to f −1 (b). Hence f −1 (a) and f −1 (b) are diffeomorphic. The set f −1 (0) is a cone, as drawn in Figure 1-2 in Guillemin and Pollack. 2. Show that 1 is a regular value for the function: p f (x, y, z) = z 2 + ( x2 + y 2 − 2)2 . √ What is the tangent space to the level set f −1 (1) at (1, 3, 1)? Solution. The derivative of f is 2 2 2y 1 − √ 2 2 2z . df(x,y,z) = 2x 1 − √x2 +y2 x +y So if (x, y, z) is a critical point of f , then all entries of this matrix are zero, meaning z = 0 and either x = y = 0 or x2 + y 2 = 4. Hence f (x, y, z) = 4 or 0. Thus 1 is a regular value of f . √ The tangent space to the level set f −1 (1) at (1, 3, 1) is the set of all points orthogonal to the gradient vector √ 2 2 p p , 2y 1 − , 2z |(1,√3,1) ∇f (1, 3, 1) = 2x 1 − x2 + y 2 x2 + y 2 = (0, 0, 2). 1 Hence T f −1 (1) (1,√3,1) = {(x, y, z) ∈ R3 | 0x + 0y + 2z = 0} = {(x, y, z) ∈ R3 | z = 0}. 3. Let M ⊂ Rk be a smooth manifold, x be a point in M and v be a vector in T Mx . Show that there is a smooth map from an interval in R1 to M : f : (−t, t) → M, with f (0) = x and df0 (1) = v. Solution. Let M be a manifold of dimension m. Choose a diffeomorphism g : U → W ∩ M where U is an open subset of Rm containing ~0, where W is an open set in Rk containing x, and where g(~0) = x. Since T Mx = dg(~0) by definition, there exists a vector u ∈ Rm satisfying dg~0 (u) = v. Since U is an open subset of Rm containing ~0, we can pick t > 0 such that {su | s ∈ (−t, t)} ⊂ U . Define the function h : (−t, t) → U by h(s) = su. Note that h(0) = ~0 and dh0 (1) = u. Define f : (−t, t) → M by f = g ◦ h. Function f is smooth because it is the composition of the smooth functions g and h. Note f (0) = g(h(0)) = g(~0) = x. and df0 (1) = d(g ◦ h)0 (1) = dgh(0) (dh0 (1)) by the chain rule = dg~0 (u) = v by definition of u. Hence we have constructed a smooth map satisfying the necessary properties. 4. The tangent bundle of a smooth m-manifold M in Rk is the subset of M × Rk : T M = {(x, v) : x ∈ M, v ∈ T Mx } . Show that T M is a smooth manifold of dimension 2m. Solution. This is a Proposition on the top of page 51 in Chapter 1 §8 of Guillemin and Pollack. See pages 50 and 51 for a complete proof. 5. Find the critical values of the smooth function: det : Mn (R) → R1 . 2 sending an n × n matrix to its determinant. Show that the special linear group SLn (R) consisting of n × n matrices of determinant 1 is a smooth manifold. Compute the tangent space of SLn (R) at the identity matrix. Solution. 2 Let A ∈ Mn (R). We identify Mn (R) with Rn , and hence T (Mn (R))A = Mn (R). Also T (R1 )det(A) = R1 . So d(det)A : Mn (R) → R1 . First we show that 0 is a critical value of det. Let 0 ∈ Mn (R) be the matrix of all zeros. Let A be any matrix in Mn (R). We have det(0 + tA) − det(0) t→0 t det(tA) = lim t→0 t tn det(A) = lim t→0 t = 0. d(det)0 (A) = lim Since A ∈ Mn (R) was arbitrary, this means that d(det)0 is singular, that 0 is a critical point of det, and that 0 = det(0) is a critical value of det. Next we show that det has no other critical values. Suppose det(A) 6= 0. We’ll show that d(det)A is nonsingular by considering d(det)A (A). We have det(A + tA) − det(A) t→0 t det((1 + t)A) − det(A) = lim t→0 t (1 + t)n det(A) − det(A) = lim t→0 t (1 + t)n − 1 = det(A) lim t→0 t = det(A) n 6= 0, d(det)A (A) = lim since the limit in the third to the last line is the derivative of (1 + x)n evaluated at x = 0, which is equal to n(1 + x)n−1 evaluated at x = 0, which is n. Since d(det)A (A) 6= 0, this means that d(det)A : Mn (R) → R1 is nonsingular. Hence det(A) 6= 0 implies that A is a regular value of det, and so the only critical value of det is 0. Since 1 is a regular value of det and since SLn (R) = det−1 (1), Lemma 1 on page 11 in Chapter 1 of Milnor implies that SLn (R) is a smooth manifold of dimension n2 − 1. 3 Finally, we compute the tangent space of SLn (R) at the identity matrix I. By the Proposition on page 24 in Chapter 1 §4 of Guillemin and Pollack, we have T (SLn (R))I = ker(d(det)I ). To find ker(d(det)I ), first note that for A ∈ Mn (R) we have det(I + tA) = 1 + t Tr(A) + O(t2 ) where Tr(A) is the trace of A and O(t2 ) is a polynomial in t with all terms of degree at least two. Hence det(I + tA) − det(I) t→0 t 1 + t Tr(A) + O(t2 ) − 1 = lim t→0 t = Tr(A). d(det)I (A) = lim So T (SLn (R))I = ker(d(det)I ) is the set of all matrices with trace equal to zero. 6. (a) Show that the function f : R1 → R1 defined by: ( 2 e−1/x x > 0 f (x) = 0 x≤0 is smooth. Solution. We need to show that all derivatives f (i) of f exist and are continuous. For x < 0, all derivatives of f are identically zero. Hence we can restrict attention to x > 0. We need to show all derivatives f (k) exist on x > 0 and satisfy limx→0+ f (i) = 0 on x > 0. Note f 0 = 2x−3 f . In fact, every derivative f (i) is of the form f (i) = p(x−1 )f where p is a polynomial and hence p(x−1 ) is a polynomial in x−1 . We prove this by induction. The base case is above. The inductive step is 0 f (i+1) = f (i) = [p(x−1 )f ]0 = −x−2 p0 (x−1 )f + 2x−3 pf = (−x−2 p0 (x−1 ) + 2x−3 p)f. Hence each f (k) is a linear combination of terms of the form x−k f for k ≥ 0. It remains to show that limx→0+ x−k f = 0. The logarithm of x−k f is −k log x − 1/x2 , and we have h h 1 k log x i 1i lim+ −k log x − 2 = lim+ − 2 + 1 . x→0 x→0 x x x2 The first term on the right hand side goes to −∞ and the second term goes to 1 by L’Hopital’s rule. Hence the limit of the logarithm is −∞, and therefore the limit of its expoential is limx→0+ x−k f = 0, as desired. 4 (b) Show that, for a < b, the function g(x) = f (x − a)f (b − x) is smooth, positive on (a, b) and zero elsewhere. Solution. Function g is smooth because it is the product of two smooth functions. It is positive on (a, b) because both f (x − a) and f (b − x) are positive on (a, b). It is zero on x < a because f (x − a) is zero here, and it is zero on x > b because f (b − x) is zero here. Rx R∞ (c) Show that the function h(x) = −∞ g(t)dt/ −∞ g(t)dt is smooth and satisfies h(x) = 0 for x < a, h(x) = 1 for x > b and 0 < h(x) < 1 for x in (a, b). Solution. R∞ Function h is smooth because its first derivative is g/ −∞ g(t)dt, and we showed R x in part (b) that g is smooth. If x < a then g(t) = 0 for all t < x, so −∞ g(t)dt = 0, so h(x) = 0. If x > b then g(t) = 0 for all t > x, so Rx R∞ Rx R∞ g(t)dt = −∞ g(t)dt, so h(x) = 1. Since 0 < −∞ g(t)dt < −∞ g(t)dt for x −∞ in (a, b), we have 0 < h(x) < 1 for x in (a, b). (d) Construct a smooth function on Rk that equals 1 on the ball of radius a, zero outside a ball of radius b, and is strictly between 0 and 1 at intermediate points. (A function of this form is sometimes called a bump function.) Solution. Define g : Rk → R by g(x) = 1 − h(kxk). Note that g equals 1 on the ball of radius a, zero outside a ball of radius b, and is strictly between 0 and 1 at intermediate points. Since the function x → kxk is smooth on Rk \{~0}, function g is clearly smooth on Rk \ {~0}. But since g is constant on the ball of radius a, function g is also smooth at the origin, and hence is smooth on all of Rk . 5