SOLUTION OF HW11 MINGFENG ZHAO December 07, 2011 1. [The Problem 7, in Page 186] Let f be a polynomial. A real number α is said to be a zero of f of multiplicity m if f (x) = (x − α)m g(x), where g(α) 6= 0. a. If f has r zeros in an interval [a, b], prove that f 0 has at least r − 1 zeros, and in general, when k < r, the k-th derivative f (k) has at least r − k zeros in [a, b]. b. If the k-th derivative f (k) has exactly r zeros in [a, b], what can you conclude about the number of zeros of f in [a, b]? Proof. a. Assume f is a polynomial and fixed Claim I: If α is a zero of f with multiplicity m ≥ 1, then α is a zeros of f 0 with multiplicity m − 1 ≥ 0. If m = 1, then there exist some polynomial g such that f (x) = (x − α)g(x), where g(α) 6= 0. Differentiating it, we get f 0 (x) = g(x) + (x − a)g 0 (x). So f 0 (α) = g(α) 6= 0, that is, α is not a zero of f 0 , so the Claim I is true. If m ≥ 2, then there exist some polynomial g such that f (x) = (x − α)m g(x), where g(α) 6= 0. Differentiating it, we get f 0 (x) = m(x − α)m−1 g(x) + (x − α)m g 0 (x) = (x − α)m−1 [mg(x) + (x − α)g 0 (x)]. Let h(x) = mg(x) + (x − α)g 0 (x), then h(α) = mg(α) 6= 0, so α is a zero of f 0 with multiplicity m − 1. Claim II: If f has r zeros in an interval [a, b], then f 0 has at least r − 1 zeros in [a, b]. 1 2 MINGFENG ZHAO Assume f has r zeros in [a, b], that is, there exists some {αi }ni=1 ⊂ [a, b] such that a ≤ α1 < α2 < · · · < αn ≤ b, and for each αi , αi is a zero of f with multiplicity mi ≥ 1, and n X mi = r. So f 0 has i=1 zeros α1 , · · · , αn with multiplicity m1 − 1, · · · , mn − 1, respectively. For each αi , 1 ≤ i ≤ n, since αi is a zero of f with multiplicity mi ≥ 1, then by the Claim I, we know that αi is a zero of f 0 with multiplicity mi − 1 ≥ 0. On the other hand, for each 1 ≤ i ≤ n − 1, since αi and αi+1 are zeros of f , that is, f (αi ) = f (αi+1 ) = 0, then by the Rolle’s Theorem, we know that there exists some βi ∈ (αi , αi+1 ) such that f 0 (βi ) = 0. So f 0 has zeros β1 , · · · , βn with multiplicity at least 1. In a summary, we know that f 0 at least has n X i=1 (mi − 1) + (n − 1) = n X mi − n + (n − 1) = r − 1 i=1 many zeros in [a, b]. Claim III: If f has r zeros in an interval [a, b], for any 1 ≤ k < r, then f k has at least r − k zeros in [a, b]. If k = 1, by the Claim I, we know that the Claim III is true. Now assume when k = m < r, and the Claim III is true, that is, f m has at least r − m zeros in [a, b]. If m + 1 = r, we are done. If m + 1 < r, since f m has r − m zeros in [a, b], by the Claim I, we know that f m+1 has at least r − m − 1 = r − (m + 1) zeros in [a, b], that is, the Claim III is true for k = m + 1. By induction, we can conclude that for any 1 ≤ k < r, then f k has at least r − k zeros in [a, b]. b. If the k-th derivative f (k) has exactly r zeros in [a, b], we can not say f at least has how many zeros in [a, b], but we can say that f at most has k + r zeros in [a, b]. Since if f has more than k + r zeros in [a, b], say f at least has k + m + 1 zeros in [a, b]. Then by the result of part a, we know that f k at least has k + r + 1 − k = r + 1 zeros in [a, b], which contradicts with the assumption that the k-th derivative f (k) has exactly r zeros in [a, b]. For example, k = 1, and r = 2, consider f (x) = x3 + 10 in [−1, 1], then f 0 (x) = 3x2 in [−1, 1], then 0 is a zero in [−1, 1] of f 0 with multiplicity 2, but f has no solutions in [−1, 1], since f (x) = x3 + 10 ≥ (−1)3 + 10 = 7 > 0 for all x ∈ [−1, 1]. SOLUTION OF HW11 3 2. [The Problem 9, in Page 187] A function f , continuous on [a, b], has a second derivative f 00 everywhere on the open interval (a, b). The line segment joining (a, f (a)) and (b, f (b)) intersects the graph of f at a third point (c, f (c)), where a < c < b. Prove that f 00 (t) = 0 for at least one point t in (a, b). Proof. Since the line segment joining (a, f (a)) and (b, f (b)) intersects the graph of f at a third point (c, f (c)), where a < c < b, then we have f (b) − f (c) f (c) − f (a) = . b−c c−a By the Mean Value Theorem, there exists some d ∈ (a, c), and e ∈ (c, b) such that f (c) − f (a) = f 0 (d)(c − a), and f (b) − f (c) = f 0 (e)(b − c). That is, f (c) − f (a) = f 0 (d), c−a and f (b) − f (c) = f 0 (e). b−c So we get f 0 (d) = f 0 (e). Notice that a < d < c < e < b, by the Rolle’s Theorem, we know that there exists some t ∈ (d, e) ⊂ (a, b) such that f 00 (t) = 0. 3. [The Problem 19, in Page 209] Give a function g, continuous everywhere, such that g(1) = 5 and Z 1 Z 1 x g(t) dt = 2. Let f (x) = (x − t)2 g(t) dt. Prove that 2 0 0 0 Z x Z g(t) dt − f (x) = x 0 x tg(t) dt. 0 Then compute f 00 (1) and f 000 (1). Proof. Since f (x) = 1 2 Z x (x − t)2 g(t) dt for all x ∈ R, and g is continuous on R, then we know that 0 f is differentiable on R, and f 0 (x) = 1 1 (x − x)2 g(x) + 2 2 Z x 2(x − t)g(t) dt, 0 for all x ∈ R. 4 MINGFENG ZHAO That is, f 0 (x) = x Z x Z g(t) dt − 0 tg(t) dt, 0 Z for all x ∈ R. 0 x x Z tg(t) dt are differentiable on R, which implies g(t) dt and Since g is continuous on R, then x x Z (x − t)g(t) dt = x 0 0 that f 0 is differentiable on R, and x Z 00 Z 0 In particular, f 00 (1) = x g(t) dt + xg(x) − xg(x) = f (x) = for all x ∈ R. g(t) dt, 0 R1 0 g(t) dt = 2. Since g is continuous on R, then Rx 0 g(t) dt is differentiable on R, which implies that f 00 (x) is differentiable on R, and f 000 (x) = g(x), for all x ∈ R. In particular, f 000 (1) = g(1) = 5. 4. [The Problem 35, in Page 238] A function f , continuous on the positive real axis, has the property that Z xy Z f (t) dt = y 1 x y Z f (t) dt + x f (t) dt, 1 for all x > 0 and all y > 0. 1 If f (1) = 3, compute f (x) for each x > 0. Proof. For any y > 0 and fix, we define Z gy (x) = xy Z f (t) dt, and hy (x) = y 1 x Z f (t) dt + x 1 y f (t) dt, for all x > 0. 1 Then we know that gy (x) ≡ hy (x) for all x > 0. Since f is continuous on (0, ∞), then both gy and hy are differentiable on (0, ∞), and 0 (gy ) (x) = yf (xy), and y Z 0 f (t) dt, (hy ) (x) = yf (x) + for all x > 0. 1 Since gy (x) ≡ hy (x) for all x > 0, then (gy )0 (x) ≡ (hy )0 (x) for all x > 0, that is, Z yf (xy) = yf (x) + y f (t) dt, 1 for all x > 0. SOLUTION OF HW11 5 In particular, taking x = 1, then y Z yf (y) = yf (1) + Z f (t) dt = 3y + 1 y f (t) dt. 1 By the arbitrary of y > 0, so y Z yf (y) = 3y + f (t) dt, for all y > 0. 1 That is, Ry f (y) = 3 + Since Ry 1 f (t) dt , y 1 for all y > 0. f (t) dt and y are differentiable functions on (0, ∞), then we know that f is also differen- tiable on (0, ∞), which implies that yf (y) is differentiable on (0, ∞). Since y Z yf (y) = 3y + f (t) dt, for all y > 0. f (y) + yf 0 (y) = 3 + f (y), for all y > 0. 1 Differentiating both sides, so That is, f 0 (y) = 3 , y for all y > 0. For any x > 0, integrating both sides along [1, x], then x Z f 0 (y) dy = 1 Z 1 x 3 dy, y for all x > 0. That is, f (x) − f (1) = 3 ln x, So we have f (x) = 3 + 3 ln x for all x > 0. for all x > 0. 6 MINGFENG ZHAO Department of Mathematics, University of Connecticut, 196 Auditorium Road, Unit 3009, Storrs, CT 06269-3009 E-mail address: mingfeng.zhao@uconn.edu