Homework 3 Solutions - Math 321,Spring 2015 (1) 1a. We want to find a sequence {pn } of polynomial such that pn → f uniformly on any bounded subset of R. By The First Weierstrass Approximation Theorem,for each positive integer N , we can find a sequence of polynomial {pn } such that pn → f uniformly on [−N, N ]. Hence for each positive integer N we can find a polynomial qN (picked one from the sequence) such that |f (x) − qN (x)| ≤ N1 for all x ∈ [−N, N ]. Define our (new) sequence of polynomials {pn } by pn := qn . Now if E is a bounded subset of R then there is an M > 0 a positive integer such that E ⊆ [−M, M ]. Now given > 0 and N such that N1 < , if n > max{N, M } then |pn (x) − f (x)| < 1 < n for all x ∈ [−M, M ] hence all x ∈ E. 1b. If P (x) is a non-constant polynomial then limx→∞ |P (x)| = ∞. Since sin(x) is bounded by 1, given M > 0 we can find x0 such that |P (x0 )| > M then | sin(x0 ) − P (x0 )| ≥ |P (x0 )| − | sin(x0 )| > M − 1 Since M could be arbitrarily large. We have k sin(x) − P (x)k∞ = ∞ for any nonconstant polynomial P . Hence if pn → sin uniformly on R. Then pn (x) ≡ Cn are constant polynomials. Then for all sufficiently large n, | sin(x) − Cn | < for all x ∈ R (with out loss of generality < 1). So | sin(0) − pn (0)| = |Cn | < . • Case 1: C ≥ 0 then | sin( −π ) − C| = C + 1 > , which is a contradiction. 2 • Case 2: C < 0 then | sin( π2 ) − C| = 1 − C > , which is a contradiction. Hence no such sequence of polynomials could exist. (2) Let p(x) be a polynomial such that |p(x) − f (x)| < for all x ∈ [a, b]. Now since ´b f (x)xn dx = 0 for all n, by linearity of integral, we have a ˆ b p(x)f (x)dx = 0 a for any polynomial p. Now f is continuous on [a, b] hence bounded by M ,say. So ˆ b ˆ b ˆ b 2 f (x) dx = f (x)(f (x) − p(x)) + f (x)p(x)dx = f (x)(f (x) − p(x))dx a a a ≤ M (b − a). Since M, b − a are fixed constants and > 0 could be arbitrarily small. Hence ˆ b f (x)2 dx = 0 a 1 2 Since f is continuous we have that f ≡ 0 on [a, b]. (For example, if f is continuous at x0 and f (x0 ) > 0 then f (x0 ) > 0 on a small neighborhood around x0 which could be shown by definition ´ of continuity, then would contribute positive quantity to the integral). In general, |f (x)|2 dx = 0 does not implies f ≡ 0 though f is nonzero on a very small set. (3) It would be easy easier to start from derivative and do integration (Integration makes f behaves better and easier to deal with). Since f 0 is continuous on [a, b] there is a polynomial q such that kf 0 − qk∞ < < b−a+1 (here sup norm is over [a, b].) Define a polynomial p by ˆ x p(x) := f (a) + q(t)dt a 0 Then p = q and ˆ x ˆ x 0 0 |f (x) − p(x)| = (f (a) + f (t)dt) − (f (a) + p (t)dt) a a ˆ x |f 0 (t) − p0 (t)|dt ≤ a < (b − a) < . b−a+1 Next we show separability by finding a countable dense subset which will be the set of polynomials with rational coefficients. Let Sn be the set of polynomials with rational coefficients with degree n hence |SnS| ≤ |Qn | is countable. Then the set of all polynomials with rational coefficients is n Sn is countable. Since Q is dense in R we will show that any polynomial could be approximated with the norm k · kC 1 [a,b] by a polynomial of rational coefficients. Let p be a polynomial of degree N (so N is fixed here), let M := sup max{|x|, |x|2 , · · · , |x|N } x∈[a,b] P be a constant. Give a polynomial p such that p(x) = N ak xk and for each ak k=0 P N k choose a rational bk such that |bk − ak | < (N +1) 2 M . Let q(x) = k=0 bk x be another polynomial then |p(x) − q(x)| ≤ , |p0 (x) − q 0 (x)| ≤ so C 1 [a, b] is separable. 3 (4) 4a. Give an example of a non-polynomial continuous function with reasons why she is not a polynomial. 4b. Suppose such a sequence of polynomials converging to an f whose degrees bounded by N exists. Since the set of polynomials of degree ≤ N forms a vector space, we could find {φ}N i=0 , an orthonormal basis. For any polynomial pn of degree ≤ N then ˆ b n X pn = pn · φi dx an,i φi , an,i = hpn , φi i = i=0 a Let > 0 then for sufficiently large n, kpn − f k∞ < so kpn k∞ < kf k∞ + . Also by Cauchy-Schwartz’s inequality, |hpn , φi i| ≤ kpn k2 kφi k2 = kpn k2 So for each i, {an,i := hpn , φi i}∞ n=1 is a bounded sequence hence has a convergent sequence converging to ai ∈ R. Now start with a subsequence A1 of {an,1 } converging to a1 . Consider {an,2 }n∈A1 is also a bounded sequence and hence has a convergent subsequence converging to a2 . Continue this way until ak we have a sequence of polynomial {pn } (renumbered) which is a subsequence of the origin sequence such that kpn − qk∞ → 0 where q(x) = i=0 ai x is a polynomial. So if the original converges, it must converge to a polynomial which is a contradiction. PN i