Skip to main content

Homework 9

If not stated otherwise, we assume that the metric spaces R\mathbb{R} and Q\mathbb{Q} are quipped with the Euclidean metric d(x,y)=xyd(x,y)=|x-y|.

Exercise: If KRK\subset\mathbb{R} and f:KRf:K\rightarrow\mathbb{R} is continuous. Show that either f(x)=0f(x)=0 for some xKx\in K or there exists nNn\in\mathbb{N} such that f(x)>1/n|f(x)|>1/n for all xKx\in K.

Proof. Since ff is continuos and KK is compact, f(K)f(K) is a compact subset of R\mathbb{R}, hence is closed. If 0f(K)0\notin f(K), then there exists nNn\in\mathbb{N} such that [1/n,1/n]f(K)=[-1/n, 1/n]\cap f(K)=\emptyset, which gives us that f(x)>1/n|f(x)|>1/n for all xKx\in K.

Exercise: Let I=[0,1]I=[0,1] be the closed unit interval. Suppose ff is a continuous mapping from II to II. Prove that there exists at least one xIx\in I such that f(x)=xf(x)=x. Hint: Consider the function g(x)=f(x)xg(x)=f(x)-x.

Proof. Since g(0)=f(0)00g(0)=f(0)-0\geq 0 and g(1)=f(1)10g(1)=f(1)-1\leq 0, by the intermediate value theorem, there exists at least one xIx\in I such that g(x)=0g(x)=0, i.e., f(x)=xf(x)=x.

Exercise: Give an example (with proof) of two real-valued functions ff and gg which are uniformly continuous, but whose product fgf\cdot g is not uniformly continuous.

Proof. (An example without proof) Let f(x)=g(x)=xf(x)=g(x)=x for all xRx\in\mathbb{R}.

Exercise: Let f:ERf:E\rightarrow\mathbb{R} be a uniformly continuous function on the open interval E=(a,b)RE=(a,b)\subset\mathbb{R}. Show that ff can be extended to a uniformly continuos function on the closed interval [a,b][a,b]. More precisely, show that there is a uniformly continuous function g:[a,b]Rg:[a,b]\rightarrow\mathbb{R}such that f(x)=g(x)f(x)=g(x) for all x(a,b)x\in (a,b). Hint: show first that limxaf(x)\lim_{x\rightarrow a}f(x) and limxbf(x)\lim_{x\rightarrow b}f(x) exist.

Proof. (Sketch) Following the hint, we first prove that the limits exist.

Claim: For any sequence {an}\{a_n\} that converges to aa, the sequence {f(an)}\{f(a_n)\} is actually a Cauchy sequence in {R}\{\mathbb{R}\}, thus converges.

Proof of claim: to show that {f(an)}\{f(a_n)\} is a Cauchy sequence, it’s sufficient to show that for every ϵ>0\epsilon>0, there is NN such that if m,nNm,n\geq N then f(am)f(an)<ϵ|f(a_m)-f(a_n)|<\epsilon. Pick your favorite ϵ\epsilon first. As ff is uniformly continuous, there is δ>0\delta> 0 such that for any x,y(a,b)x,y\in(a,b) if xyδ|x-y|\delta, f(x)f(y)<ϵ|f(x)-f(y)|<\epsilon. As {an}\{a_n\} converges to aa, there is NN such that for all nNn\geq N, ana<δ/2|a_n-a|<\delta/2. Thus for any m,nNm,n\geq N, amanama+ana<δ|a_m-a_n|\leq|a_m-a|+|a_n-a|<\delta, whence f(am)f(an)<ϵ|f(a_m)-f(a_n)|<\epsilon.

Now, we’ve shown that for every sequence {an}\{a_n\} that converges to aa, the sequence {f(an)}\{f(a_n)\} converges. But how do you know for two different sequences {an}\{a_n\} and {bn}\{b_n\} that converge to aa, {f(an)}\{f(a_n)\} and {f(bn)}\{f(b_n)\} have the same limit?

There is a slick way to resolve this. We can mingle {an}\{a_n\} and {bn}\{b_n\} together. Specifically, consider a new sequence {a1,b1,a2,b2,}\{a_1, b_1, a_2, b_2, \ldots\}, denoted as cnc_n. Now cnc_n converges to aa, so {f(cn)}\{f(c_n)\} converges. Meanwhile, we know that {f(an)}\{f(a_n)\} and {f(bn)}\{f(b_n)\} are two subsequences of {f(cn)}\{f(c_n)\}, so they share the same limit.

Now we can extend ff to gg by feeding in the limits at aa and bb. Since gg is a continuous function on a closed interval [a,b][a,b], its uniformly continuity follows immediately.

Exercise: Let a<ba< b be real numbers. A function f:(a,b)Rf:(a,b)\rightarrow\mathbb{R} is said to be convex if

f(λx+(1λ)y)f(x)+(1λ)f(y) for all x,y(a,b) and all λ(0,1).f(\lambda x+(1-\lambda)y)\leq f(x)+(1-\lambda)f(y) \text{ for all } x,y\in(a,b) \text{ and all } \lambda\in (0,1).

Prove that every convex function is continuos. Hint: prove first that if a<s<t<u<ba<s<t<u<b, then

f(t)f(s)tsf(u)f(s)usf(u)f(t)ut\frac{f(t)-f(s)}{t-s}\leq\frac{f(u)-f(s)}{u-s}\leq\frac{f(u)-f(t)}{u-t}

Proof. First, we prove the first inequality in the hint. Notice that

f(t)f(s)tsf(u)f(s)us\frac{f(t)-f(s)}{t-s}\leq\frac{f(u)-f(s)}{u-s}

is equivalent to

f(t)tsusf(u)+utusf(s).f(t)\leq\frac{t-s}{u-s}f(u)+\frac{u-t}{u-s}f(s).

But notice that if we take x=ux=u, y=sy=s and λ=tsus\lambda=\frac{t-s}{u-s} and plug them into the convexity inequality, the above inequality holds immediately. Similarly, we can prove the second inequality in the hint.

For a<x<y<ba<x<y<b, let k(x,y)=f(x)f(y)xyk(x,y)=\frac{f(x)-f(y)}{x-y}. By the inequality chain in the hint, we know that if a<s<t<u<ba<s<t<u<b, k(s,t)k(s,u)k(t,u)k(s,t)\leq k(s,u)\leq k(t,u). Now for any x(a,b)x\in(a,b) and ϵ>0\epsilon>0, pick δ(0,ϵ)\delta\in(0,\epsilon) such that a<xδ<x+δ<ba<x-\delta<x+\delta<b and x1(a,xδ),x2(x+δ,b)x_1\in(a,x-\delta), x_2\in(x+\delta,b). For any y(xδ,x+δ)y\in(x-\delta,x+\delta), by the chain inequality in the hint, k(x,y)k(x,y) is bounded below by k(x1,x)k(x_1,x) and is bounded above k(x,x2)k(x,x_2). Thus there is BB such that k(x,y)<B|k(x,y)|<B for all y(xδ,x+δ)y\in(x-\delta,x+\delta). Whence, for all y(xδ/B,x+δ/B)y\in(x-\delta/B, x+\delta/B), f(x)f(y)<Bxy<δ<epsilon|f(x)-f(y)|<B|x-y|<\delta<epsilon. This shows that ff is continuous.