Math 317 HW #11 Solutions 1. Exercise 5.3.1. Recall from Exercise 4.4.9 that a function f : A → R is “Lipschitz on A” if there exists an M > 0 such that f (x) − f (y) ≤M x−y for all x, y ∈ A. Show that if f is differentiable on a closed interval [a, b] and if f 0 is continuous on [a, b], then f is Lipschitz on [a, b]. Proof. Suppose f is differentiable and f 0 is continuous on [a, b]. Then, since [a, b] is compact, f 0 is bounded. In other words, there exists M > 0 such that |f 0 (x)| ≤ M for all x ∈ [a, b]. Then, for any x, y ∈ [a, b], the Mean Value Theorem guarantees that there exists c between x and y such that f (x) − f (y) = |f 0 (c)| ≤ M. x−y Since the choice of x and y was arbitrary, we see that the inequality f (x) − f (y) ≤M x−y holds for all x, y ∈ [a, b], so f is Lipschitz on [a, b]. 2. Exercise 5.3.6. Let g : [0, 1] → R be twice-differentiable (i.e., both g and g 0 are differentiable functions) with g 00 (x) > 0 for all x ∈ [0, 1]. If g(0) > 0 and g(1) = 1, show that g(d) = d for some point d ∈ (0, 1) if and only if g 0 (1) > 1. Proof. (⇒) First, suppose g(d) = d for some d ∈ (0, 1). Then, by the Mean Value Theorem, there exists c1 ∈ (d, 1) such that g 0 (c1 ) = g(1) − g(d) 1−d = = 1. 1−d 1−d Using the Mean Value Theorem again, there exists c2 ∈ (c1 , 1) such that g 00 (c2 ) = g 0 (1) − g 0 (c1 ) g 0 (1) − 1 = . 1 − c1 1 − c1 We know that g 00 (c2 ) > 0; therefore, since the denominator on the right hand side is positive, the numerator must be positive as well, meaning that g 0 (1) > 1. (⇐) Now, I intend to prove the contrapositive of the reverse direction. Suppose g(x) 6= x for all x ∈ (0, 1). Let h(x) = g(x) − x. Then g(x) 6= x is equivalent to saying h(x) 6= 0 for all x ∈ (0, 1). We know that h(0) = g(0)−0 = g(0) > 0. Since h(x) 6= 0 for all x ∈ (0, 1) and h is continuous, we know, by the converse of the Intermediate Value Theorem, that h(x) 6≤ 0 for all x ∈ (0, 1). Thus, we know that h(x) > 0 for all x ∈ (0, 1), so h(1) − h(x) 0 − h(x) = <0 1−x 1−x 1 for all x ∈ (0, 1). Hence, if (xn ) is a sequence in (0, 1) converging to 1, then for each n there exists cn ∈ (xn , 1) such that h(1) − h(xn ) h0 (cn ) = < 0, 1 − xn so h0 (cn ) < 0 for all n. We know that h0 is differentiable and, hence, continuous, so Theorem 4.3.2 implies that limn→∞ h0 (cn ) = h0 (1). Since h0 (cn ) < 0 for all n, this means that h0 (1) ≤ 0. But now, since h0 (x) = g 0 (x) − 1, this implies that g 0 (1) ≤ 1, as desired. 3. Exercise 5.3.8. Assume g : (a, b) → R is differentiable at some point c ∈ (a, b). If g 0 (c) 6= 0, show that there exists a δ-neighborhood Vδ (c) ⊆ (a, b) for which g(x) 6= g(c) for all x ∈ Vδ (c) with x 6= c. Compare this result with Exercise 5.3.7. Proof. Suppose 0 6= g 0 (c) = lim x→c g(x) − g(c) . x−c Let = |g 0 (c)| > 0. Then, by definition of the functional limit, there exists δ > 0 such that 0 < |x − c| < δ implies g(x) − g(c) 0 − g (c) < = |g 0 (c)|. x−c This means both terms on the left have the same sign and hence, in particular, that g(x) − g(c) 6= 0. x−c Therefore, for all x ∈ Vδ (c) with x 6= c, we have that g(x) 6= g(c). Exercise 5.3.7 gives a function g such that g(0) = 0 and g 0 (0) > 0, but g is not increasing on any interval containing zero. The simplest way for this to happen would seem to be for g to take on negative values on any interval containing zero. However, by the Intermediate Value Theorem, this would imply that g(x) = 0 for some x 6= 0 in any interval containing zero, which would contradict the statement we just proved. It turns out that for all x > 0, g(x) > 0, and for all x < 0, g(x) < 0. However, in any interval containing zero there will be some y > x > 0 (in fact, infinitely many such x and y) such that g(x) > g(y). The below figure gives some sense of this: 2 0.2 0.16 0.12 0.08 0.04 -0.28 -0.24 -0.2 -0.16 -0.12 -0.08 -0.04 0 0.04 0.08 0.12 0.16 0.2 0.24 0.28 -0.04 -0.08 -0.12 -0.16 -0.2 4. Exercise 5.3.13. Prove the following version of L’Hò‚pital’s Rule: Theorem 1. Assume f and g are differentiable on an interval containing a, except possibly at a itself. If limx→a f (x) = 0 and limx→a g(x) = 0, then f 0 (x) =L x→a g 0 (x) lim implies lim x→a f (x) = L. g(x) Proof. Since f and g are differentiable except possibly at a, f and g must be continuous except possibly at a. Since limx→a f (x) = 0 and limx→0 g(x) = 0, we can define continuous functions fe and ge by ( f (x) if x 6= a fe(x) := 0 if x = a and ( g(x) ge(x) := 0 if x 6= a if x = a. Since fe(a) = 0 and ge(a) = 0, Theorem 5.3.6 applies to fe and ge, so we know that fe0 (x) = L implies x→a g e0 (x) lim Now, since fe0 (x) g e0 (x) = f 0 (x) g 0 (x) and fe(x) g e(x) = f (x) g(x) fe(x) = L. x→a g e(x) lim for all x 6= a, Lemma 2 below implies that f 0 (x) fe0 (x) = lim 0 0 x→a g (x) x→a g e (x) lim and f (x) fe(x) = lim . x→a g(x) x→a g e(x) lim Combining this with (1), then, we can conclude that f 0 (x) = L implies x→a g 0 (x) lim as desired. 3 lim x→a f (x) = L, g(x) (1) Lemma 2. Suppose two functions h1 and h2 are defined on a set A, that a ∈ A is a limit point of A, and that h1 (x) = h2 (x) for all x = 6 a. Then lim h1 (x) = lim h2 (x) x→a x→a unless neither limit exists. Proof. Suppose one of the limits exists; without loss of generality, assume limx→a h1 (x) = L. Let > 0. Then, by definition of the functional limit, there exists δ > 0 such that 0 < |x − a| < δ implies that |h1 (x) − L| < . Then, if 0 < |x − a| < δ, we have that |h2 (x) − L| = |h1 (x) − L| < , where the equality follows from the fact that h1 (x) = h2 (x) for all x 6= a. Since the choice of > 0 was arbitrary, we see that for all > 0 there exists such a δ, so limx→a h2 (x) = L as well. 5. Exercise 5.4.6. (a) Without working too hard, explain why the partial sum gm = h0 + h1 + · · · + hm is differentiable at x. Now, prove that, for every value of m, we have 0 0 |gm+1 (x) − gm (x)| = 1. Proof. Note that none of h0 , h1 , . . . , hm has a corner between xm and ym , so each of these functions is linear on (xm , ym ). Since the sum of linear functions is linear, we know that gm is linear and, hence, differentiable on (xm , ym ). In fact, since the derivative of a linear 0 is constant on (x , y ). We will use this fact in function is constant, we know that gm m m part (b), but notice that this is more than enough to conclude that gm is differentiable at x for all m. Now, gm+1 (x) − gm (x) = (h0 (x) + h1 (x) + · · · + hm (x) + hm+1 (x)) − (h0 (x) + h1 (x) + · · · + hm (x)) = hm+1 (x). Since each hk is a line of slope ±1 away from its corners, we know that h0m+1 (x) = ±1. Therefore, 0 0 |gm+1 (x) − gm (x)| = |h0m+1 (x)| = 1. (b) Prove the two inequalities g(ym ) − g(x) g(xm ) − g(x) 0 < gm (x) < . ym − x xm − x 4 Proof. Note that, for all n ≥ m, 1 p+1 1 = n h 2n−m (p + 1) = 0 hn (ym ) = n h 2n · m 2 2 2 since h is zero on every even integer. Hence, g(ym ) = ∞ X hn (ym ) = n=0 m X hn (ym ) = gm (ym ). n=0 Now, this implies that g(ym ) − g(x) = gm (ym ) − g(x) = gm (ym ) − gm (x) + gm (x) − g(x). Notice that gm (x) − g(x) = − ∞ X (2) hn (x) < 0 n=m+1 since hn (x) > 0 for all n. This, along with (2), implies that g(ym ) − g(x) < gm (ym ) − gm (x), and, hence, that g(ym ) − g(x) gm (ym ) − gm (x) < . ym − x ym − x (3) On the other hand, by the Mean Value Theorem, there exists c ∈ (x, ym ) such that 0 gm (c) = gm (ym ) − gm (x) . ym − x Since, as we argued in (a), gm is linear on (xm , ym ), g 0 (c) = g 0 (x) so combining the above with (3) yields g(ym ) − g(x) 0 < gm (x). ym − x A completely analoguous argument gives the other inequality. (c) Use parts (a) and (b) to show that g 0 (x) does not exist. Proof. If g 0 (x) = limz→x lim m→∞ g(z)−g(x) z−x exists, then Theorem 4.2.3 implies that g(xm ) − g(x) g(ym ) − g(x) = g 0 (x) = lim . m→∞ xm − x ym − x In other words, lim n→∞ g(xm ) − g(x) g(ym ) − g(x) − = 0. xm − x ym − x However, (b) implies that 0 0 < gm (x) − g(ym ) − g(x) g(xm ) − g(x) g(ym ) − g(x) < − . ym − x xm − x ym − x 5 Then, by the Squeeze Theorem, g(ym ) − g(x) 0 , 0 = lim gm (x) − m→∞ ym − x so, adding limm→∞ implies g(ym )−g(x) ym −x to both sides and using the Algebraic Limit Theorem g(ym ) − g(x) g(ym ) − g(x) g(ym ) − g(x) 0 lim + lim = lim gm (x) − m→∞ m→∞ m→∞ ym − x ym − x ym − x g(ym ) − g(x) g(ym ) − g(x) 0 = lim gm (x) − + m→∞ ym − x ym − x 0 = lim gm (x). m→∞ 0 (x)) is not Cauchy, so it cannot converge. But part (a) implies that the sequence (gm From this contradiction, then, we can conclude that g 0 (x) does not exist. 6