If (an)n=1∞ and (bn)n=1∞ are equivalent, then ∀ϵ>0,∃N>1, s.t. (an)n=N∞ and (bn)n=N∞ are ϵ-close. Thus ∣an−bn∣<ϵ,∀n>N This means limn→∞(an−bn)=0, and all steps above can be reversed.
Exercise 9.9.2
( a ) implies ( b ): ∀ϵ>0,∃δ>0, s.t. ∣f(x)−f(y)∣<ϵ whenever ∣x−y∣<δ. As (xn)n=1∞ and (yn)n=1∞ are equivalent, we can find N such that ∣xn−yn∣<δ,∀n>N, so we shall get ∣f(xn)−f(yn)∣<ϵ,∀n>N This means the sequence (f(xn))n=1∞ and (f(yn))n=1∞ are equivalent. ( b ) implies ( a ): Assume f is not uniformly continuous on X, then we shall be able to find a ϵ0>0 s.t. for every n∈N,∃xn,yn∈X,∣xn−yn∣<1/n, but ∣f(xn)−f(yn)∣≥ϵ0. By Archimedean principle we know the sequence (xn)n=1∞ and (yn)n=1∞ are equivalent, but n→∞lim(f(xn)−f(yn))≥ϵ0>0 Which means the sequence (f(xn))n=1∞ and (f(yn))n=1∞ are not equivalent, a contradiction.
Exercise 9.9.3
f is uniformly continuous on X means ∀ϵ>0,∃δ>0, s.t. ∣f(x)−f(y)∣<ϵ whenever ∣x−y∣<δ. As (xn)n=0∞∈X is a Cauchy sequence, we can find N>0 s.t. ∣xn−xm∣<δ,∀n,m>N This means ∣f(xm)−f(xn)∣<ϵ,∀n,m>N Which shows (f(xn))n=0∞ is a Cauchy sequence.
Exercise 9.9.4
Since x0 is an adherent point of X, there’s a sequence (xn)n=1∞∈X which converges to x0, in particular, this sequence is a Cauchy sequence by Theorem 6.4.18, thus by Proposition 9.9.12, the sequence (f(xn))n=1∞ is a Cauchy sequence, by Theorem 6.4.18, (f(xn))n=1∞ is convergent, let the number being convergent be L∈R. To finish the proof, let (xn′)n=1∞∈X be any sequence which converges to x0, then (xn′)n=1∞ and (xn)n=1∞ are equivalent sequence, thus (f(xn′))n=1∞ and (f(xn))n=1∞ are also equivalent sequence, thus (f(xn′))n=1∞ converges to L. By Proposition 9.3.9, f converges to L at x0. An alternate demonstration of Example 9.9.10: if f is uniformly continuous, then since 0 is an adherent point of the sequence (0,2), the limit limx→0;x∈(0,2)1/x should exist, but in fact limx→0;x∈(0,2)1/x=+∞.
Exercise 9.9.5
Assume f(E) is unbounded, then ∀n∈N+, we can find xn∈E, s.t. ∣f(xn)∣>n. Since E is bounded, we know the sequence (xn)n=1∞ is bounded, thus has a convergent subsequence (xnj)j=1∞, we denote limj→∞xnj=x0, by Corollary 9.9.14, limj→∞f(xnj) exists, thus is bounded, but ∣f(xnj)∣>nj>j, which is a contradiction.
Exercise 9.9.6
Let ∀ϵ>0, then since g is uniformly continuous, ∃δ1>0, s.t. ∣g(y)−g(y′)∣<ϵ,∀y,y′∈Y,∣y−y′∣<δ1 For this δ1, since f is uniformly continuous, ∃δ2>0, s.t. ∣f(x)−f(x′)∣<δ1,∀x,x′∈X,∣x−x′∣<δ2 So we have ∣(g∘f)(x)−(g∘f)(x′)∣=∣g(f(x))−g(f(x′))∣<ϵ,∀x,x′∈X,∣x−x′∣<δ2 Thus g∘f is uniformly continuous on X.